D. S. Chandrasekharaiah and Lokenath Debnath (Auth.) - Continuum Mechanics-Elsevier Inc, Academic Press (1994)

Download as pdf or txt
Download as pdf or txt
You are on page 1of 599

D. S.

Chandrasekharaiah
Department of Mathematics
Bangalore University
Central College Campus
Bangalore, 560001
India
Lokenath Debnath
Department of Mathematics and
Department of Mechanical and Aerospace Engineering
University of Central Florida
Orlando, Florida 32816-1364
U.S.A.

CONTINUUM MECHANICS
D. S. Chandrasekharaiah
Bangalore University
Bangalore, India
Lokenath Debnath
University of Central Florida
Orlando, Florida

ACADEMIC PRESS
Boston San Diego New York
London Sydney Tokyo Toronto

This book is printed on acid-free paper. @


Copyright 1994 by Academic Press, Inc.
All rights reserved.
No part of this publication may be reproduced or
transmitted in any form or by any means, electronic
or mechanical, including photocopy, recording, or
any information storage and retrieval system, without
permission in writing from the publisher.
ACADEMIC PRESS, INC.
525 B Street, Suite 1900, San Diego, CA 92101-4495

United Kingdom Edition published by


ACADEMIC PRESS LIMITED
24-28 Oval Road, London NW1 7DX

Library of Congress Cataloging-in-Publication Data


Chandrasekharaiah, D. S., 1943Continuum mechanics / D. S. Chandrasekharaiah, Lokenath Debnath.
p. cm.
Includes bibliographical references and index.
ISBN 0-12-167880-6 (acid-free)
1. Continuum mechanics. I. Debnath, Lokenath. II. Title.
QA808.2.C482 1994
531dc20
93-41278
CIP

PRINTED IN THE UNITED STATES OF AMERICA


94 95 96 97

BB 9 8 7 6 5 4 3 2 1

To our children
PRASHANTH, PRATHIBHA, POORNIMA
and
JAYANTA

PREFACE

Solid and fluid mechanics are two major subjects studied by all students
of applied mathematics, physics and engineering. Traditionally, these two
subjects are taught separately by two different specialists whose approach,
orientation and notation are in general different. In such separate treatments, it has not always been clear to students that the fundamental ideas
and general principles are indeed common to these subjects. The modern
trend is therefore to make a unified presentation of the ideas and general
principles common to all branches of solid and fluid mechanics under the
general heading of Continuum Mechanics. This unified course develops the
fundamentals and foundations more carefully than the traditional separate
courses where normal tendency is to put emphasis on applications. Once
familiar with the basic concepts and general principles of continuum
mechanics, the student will find little difficulty in specializing in various
branches of solid and fluid mechanics at a later stage.
There appear to be many books available for use by students studying
continuum mechanics. Some are excellent but too sophisticated and terse
for the beginner. Some are too elementary or have only limited scope in
their contents. While teaching continuum mechanics, the authors have
found difficulty over the choice of textbooks to accompany the lectures.
They have felt the need of a detailed and self-contained textbook primarily
intended for the beginners. This book is an attempt to meet this need.
It is based upon courses of lectures given by the authors over a number of
years to the first year graduate students in Bangalore University, Calcutta
University, East Carolina University and the University of Central Florida.
XI

Preface

The book assumes only a limited knowledge of mechanics, and the material
in it has been selected to introduce the reader to the fundamental ideas,
general principles and applications of continuum mechanics. Despite its
bulk the book is genuinely an introduction to continuum mechanics; hence
no attempt is made to present a detailed account of solid and fluid
mechanics except for the formulation of their governing equations and
immediate simple applications. It is hoped the book will prepare the reader
for further study of various branches of solid and fluid mechanics including
nonlinear elasticity, plasticity, thermoelasticity, viscoelasticity and nonNewtonian fluid dynamics.
A good knowledge of vectors and tensors is essential for a full appreciation of continuum mechanics. A simple and self-contained presentation
of these topics primarily tailored to the needs of continuum mechanics
is therefore included in the first three chapters of the book. Since the
Cartesian tensor formulation is sufficient for the development of continuum
mechanics at an elementary level, we have limited our discussion of tensors
to Cartesian tensors only. Bearing in mind the mathematical background
and skill of the students for whom the book is primarily intended, we have
made only minimal use of abstract mathematics. The reader is assumed to
be familiar with traditional mathematics including matrices, geometry,
differential and integral calculus, and ordinary and partial differential
equations, in three-dimensional space.
Chapters 4-8 discuss the fundamental concepts, general principles and
major results of nonlinear continuum mechanics in a detailed and systematic
way. Chapter 4 introduces the continuum hypothesis, basic definitions and
the meanings of the Lagrangian and Eulerian formulations of continuum
mechanics.
The study of deformation of a continuum is the major topic of Chapter
5. Stretch and strain tensors are introduced and their respective geometrical
significances explained. The strain-displacement relations in the general
(nonlinear) and linearized forms are obtained. The compatibility condition
for the linearized case is derived. Principal strains and principal directions
of strain are discussed in some detail.
Chapter 6 deals with the instantaneous motion of a continuum. The
concept of material derivative is defined and the velocity and acceleration
vectors are introduced. The stretching tensor and the vorticity tensor/vector
are discussed along with their physical significance. The transport formulas
are then proved. The concepts of path lines, stream lines, vortex lines and
circulation are introduced for subsequent references in fluid mechanics.
The seventh chapter is concerned with the concept of stress in a continuum.
Based upon the Cauchy's stress principle, the stress vector and the stress
tensor are defined and their relationship discussed. In addition, the normal

Preface

Xi

stress, the shear stress, the principal stresses and the principal directions
of stress are defined and their basic properties examined. The PiolaKirchhoff stress tensors are also introduced.
The field equations of continuum mechanics are presented in Chapter 8.
The equation of continuity and the equations of motion and equilibrium
are obtained by using the laws of balance of mass and momentum. Some
general solutions of the equilibrium equation in terms of various stress
functions are presented. The first law of thermodynamics is used to establish
the energy equation. The Clausius-Duhem inequality is obtained by the use
of the second law of thermodynamics. It is pointed out that all the field
equations are applicable to all continua representing solids, liquids and gases
regardless of their internal physical structure. The crucial need for the
so-called constitutive equations which distinguish one class of materials from
the other while studying the individual branches of continuum mechanics is
indicated. This is the key chapter in the sense that every specialized branch
of solid or fluid mechanics is just an offshoot of this chapter. Discussion of
the constitutive theory falls beyond the scope of the book.
The last two chapters are devoted to the development of the governing
equations of two basic areas of continuum mechanics: linear elasticity and
mechanics of nonviscous and Newtonian viscous fluids. Chapter nine deals
with the fundamental equations of the linear theory of elastic solids. The constitutive equation for a linear elastic solid (generalized Hooke's law) is
postulated and then specialized to homogeneous and isotropic solids. The
governing equations of elastostatics and elastodynamics are derived and the
uniqueness of solutions established. Some standard elastostatic problems
including extension, bending and torsion of beams and the pressure-vessel
problems are discussed. Finally, wave propagation problems including
plane waves, Rayleigh waves and Love waves are studied in some detail.
The final chapter deals with the fundamental equations of fluid mechanics.
Based upon the appropriate constitutive relations, the Euler's equation for a
non-viscous fluid, and the Navier-Stokes equation for a viscous fluid are
derived and their consequences studied. Some standard viscous flow problems are considered. Further, a brief introduction to water waves is given.
Throughout the book, major emphasis is given to the logical development
of the fundamental principles and unified treatment of solid and fluid
mechanics. All the mathematical preliminaries are presented in Chapters 1
through 3 in order to develop a systematic theory of continuum mechanics.
However, it is not necessary for the reader to know everything contained in
these Chapters before taking up the study of continuum mechanics which
begins with Chapter 4. One can start with Chapter 4 after having just a
broad review of Chapters 1 to 3 and return to appropriate sections of these
chapters for a detailed study as and when the need arises. The theory of

XIV

Preface

continuum mechanics dealt with in Chapters 4 to 8 is essentially a systematic


mathematical theory. Special effort is made to present this theory in concise
and clear terms from a mathematical point of view; physical considerations
and motivating arguments are not emphasized beyond a certain point.
However, the last two chapters, which provide links with the traditional
developments of the theory of elasticity and fluid mechanics, are appliedoriented and contain sufficient physical explanations.
The book contains over 250 worked examples and over 500 exercises.
Some of these are elementary and some are challenging. These should help
the student in the process of understanding and mastering so analytical a
subject as continuum mechanics. Answers and hints to some selected
exercises are provided at the end of the book.
This is a text book designed for use by the beginners in continuum
mechanics. We have therefore made no attempt to present any new material
as such. Also, we have refrained from burdening the reader with historical
notes and references to original sources. Those interested in an advanced
treatment of the topics covered in the book along with full references may
consult the encyclopedia articles of Gurtin, Serrin, Sneddon and Berry,
Truesdell and Noll, and Truesdell and Toupin listed in the Bibliography
given at the end of the book. The Bibliography also includes some other
works recommended for further study and reference. Many of the results
and problems presented in the book are either motivated by or borrowed
from the works cited in the Bibliography. We wish to acknowledge our
indebtness to these works.
In preparing the book, the authors have been encouraged by and have
benefited from the helpful comments/criticisms of a number of students
and faculty members of several universities in the United States and India.
Comments and suggestions made by anonymous reviewers have helped to
improve the quality of the book. The authors are thankful to all these
individuals for their interest in the book.
A major portion of the present version of the book was written when the
first author was a visiting Fulbright scholar at the University of Central
Florida, Orlando, under the Indo-American Fellowship Program. He
records his grateful thanks to the agencies sponsoring the program.
Our special thanks to Jackie Callahan and June Wingler who typed the
manuscript and cheerfully put up with constant revisions and changes. In
spite of the best efforts of everyone involved, doubtless some typographical
errors remain. We do hope that these are both few and obvious, and will
cause minimal confusion. Finally, we thank Brian Miller, Editor, and the
staff of Academic Press for their assistance and cooperation.
D. S. Chandrasekharaiah
Lokenath Debnath

CHAPTER 1

SUFFIX NOTATION

1.1
INTRODUCTION
The language of tensors is best suited for the development of the subject of
continuum mechanics. The compactness as well as the efficiency of the
tensor notation is very useful for the study of this subject and gives the
subject a great beauty. The use of Cartesian tensors is sufficient for the
development of the theory of continuum mechanics; for the solution of
specific problems, orthogonal curvilinear coordinates suitable to the
geometry of the problem may lead to simplification of the analysis. We
therefore use Cartesian tensors in the main body of the text, and in the first
three chapters we present a detailed and self-contained account of Cartesian
tensors primarily tailored to the needs of continuum mechanics. A shorthand notation, known as the suffix notation (or subscript notation or index
notation), employed in the treatment of Cartesian tensors is introduced in
this chapter. It is assumed that the reader has a basic knowledge of vector
algebra, matrix theory and three-dimensional analytic geometry.
1

SUFFIX NOTATION

1-2
RANGE AND S U M M A T I O N CONVENTIONS
Consider the following system of algebraic equations
# * + anx2

+ #13*3 = bx

a2lxl + a22x2 + #23*3 = b2

(1.2.1)

#31*1 + #32*2 + #33*3 = *3

We can write these equations as


#/i*i + #/2*2 + #/3*3 = bi9

i = 1, 2, 3

(1.2.2)

By using summation sign, this can be rewritten as


3

aikxk = bi9

k= \

i = 1,2,3

(1.2.3)

We say that the values 1, 2, 3 form the range of the suffixes i and k.
Let us adopt the following convention.
CONVENTION 1 All the suffixes we employ have the range 1, 2, 3. (This
is known as the range convention.)
Then (1.2.3) may be shortened to
#/*** = bi
k

(1.2.4)

We note that the suffixes / and k play different roles in (1.2.4); although
summation is indicated over the suffix k, the suffix / is left "free." We
observe that whereas the suffix k (over which summation is taken) is
repeated in the term included under the summation sign, the suffix /
appears exactly once in every term. This observation prompts us to adopt
another convention stated as follows.
CONVENTION 2 Whenever a suffix is repeated (once) in a term,
summation is required to be taken over that suffix. (This is known as the
summation convention.)
Then, we can write kaikk
rewritten as

as

J ust #/*** Consequently, (1.2.4) can be

aikxk = bi
This is a concise form of the system of equations (1.2.1).

(1.2.5)

1.3

FREE AND D U M M Y SUFFIXES

An equation such as (1.2.5) containing suffixes for which the range and
summation conventions are applicable is referred to as an equation written
in the suffix notation, subscript notation or index notation. Depending on
the context, such an equation may be viewed either as a representation (in
a concise form) of a system of equations or as a representative (typical
equation) of the system.
It should be emphasized that, according to the summation convention, a
summation over a suffix is implied only if the suffix is repeated (only once)
in the same term. Thus, the symbol ak + bk does not represent the sum
= l(ak + bk) and the symbol auXi makes no sense. Repetition of a suffix
more than once (in a term) is prohibited under the summation convention.
The summation convention is useful in writing a double sum, a triple
sum, etc., also in a short form. For example, we can write
simply as a^b^
which has nine terms. Similarly, the symbol aijbjkcki represents the triple sum

which contains 27 terms.

1.3
FREE AND D U M M Y SUFFIXES
It has been noted that (1.2.5) is a concise form of the system of equations
(1.2.1). A suffix such as k in (1.2.5) that is summed over is called a summation suffix or a dummy suffix. A suffix such as / in (1.2.5) that is free of
summation is called a free suffix or a live suffix.
Since a dummy suffix just indicates summation, the letter used to denote
it is of no consequence, because expressions such as akbk and ambm represent the same sum as ax bl + a2b2 + a3 b3. As such, a dummy suffix may be
replaced by any other suffix within the provisions of the summation
convention. For example, in the expression aikxk the suffix k may be
changed to any suffix, say m, other than /. (Note that changing k to / leads
to auxi9 which makes no sense under the summation convention!) The
concise form (1.2.5) of equations (1.2.1) may therefore be written also as
"im^m

(1.3.1)

1 SUFFIX NOTATION

On the other hand, (1.2.5) has the same meaning as


0;*** = bj

(1.3.2)

because, by the range and summation conventions, (1.3.2) actually stands


for
3

ajkxk = bj,

j = 1,2,3

A: = 1

which are precisely equations (1.2.1). In (1.3.2), y is a free suffix and this
equation may be obtained from (1.2.5) by changing the free suffix i toy in
every term of (1.2.5). This illustrates the fact that a free suffix may also be
changed to any other suffix (of course within the provisions of the summation convention) provided the same change is made in every term.
Often we deal with systems of equations whose concise form requires the
use of more than one free suffix. For example, consider the following
system of equations:
anx\ + anx2xx + 013*3*! = bn
anxlx2 + aX2x\ + al3x3x2 = bl2
anXxXs
021*1

+ tf12*2*3 + 013*3 = *13


022*2*1 ' 023*3*1

021*1*2 + tf22*f

^21

^23*3*2 = *22

(1.3.3)

<*21*1*3 + 022*2*3 + #23*3 = *23


031*? + 032*2*1 + 033*3*1 = *31
031*1*2 + 032*2 + 033*3*2

#32

031*1*3 + 032*2*3 + 033*3 = &33

By using the range and summation conventions, the first three equations in
(1.3.3) may be represented as
0i****, = bv

(1.3.4)

Similarly, the next three and the last three equations in (1.3.3) can be
represented, respectively, as
02****, = b2j

(1.3.5)

03****, = b3j

(1.3.6)

Equations (1.3.4) to (1.3.6) may be put together as


0/****7 = bij

(1.3.7)

1.3

FREE AND D U M M Y SUFFIXES

This is a concise form of equations (1.3.3) written in the suffix notation.


Note that this concise form contains two free suffixes, / and y, and one
dummy suffix, k.
Since every suffix has the range 1, 2, 3 according to the range convention,
an equation that contains one free suffix represents three equations, as in
the case of (1.2.5); an equation that contains two free suffixes represents
nine equations as in the case of (1.3.7), and so forth. An equation containing N free suffixes represents 3^ equations.
The concepts of free and dummy suffixes as well as the range and summation conventions adopted to systems of equations, as just described, may be
employed to other mathematical systems also. Thus, the set of three
numbers (ax, a2,a3) can be represented as just (,). The 3 x 3 matrix
^
#21

#12 An]
a

#22

23

L#31 #32 #33j

may be represented as just (a^) or [a^\.


In general, the symbol aijk involving N free suffixes i, y, k... represents
3^ numbers.
EXAMPLE 1-3.1
of ct.
Solution

If UT, = cx^, and bt = ijCj, write down at in terms

It is given that
a, = "ijbj

(1.3.8)

bi = ucj

(1.3.9)

and
If we change the free suffix from / to j and the dummy suffix from j to
k in (1.3.9), we get
(1.3.10)
bj = jkck
This has the same meaning as that of (1.3.9). Substituting for bj from
(1.3.10) in (1.3.8) gives
*i = oLUjkck
This is the expression for at in terms of c,.
EXAMPLE 1.3.2
(i)

(1.3.11)

Show that
aubu = ab

(ii) (aUk + ajki + akUyXiXjXk = 3aijkxixjxk

(1.3.12)
(1.3.13)

SUFFIX NOTATION

Solution (i) In a^by both / and y are dummy suffixes. Hence / may be
replaced by any suffix and j may be replaced by any suffix. In particular,
we may replace / by j , and j by /. Thus
a b

u u = anbJi

(ii) We first note that


(ijk + ajki + <*kij)XiXj*k = <*ijkXiXjXk + <*jkixixjxk

+ <*kijxixj*k

(1.3.14)

In ajkiXiXjXk, all of /,y, k are dummy suffixes. As such / may be replaced


by kfj may be replaced by / and k may be replaced byy. Thus,
a

jkixixjxk

aijkxkxixj

kijxixjxk

ijkxixjxk

Similarly,
a

ijkxjxkXi

ijkxixjxk

Thus, all the three terms in the righthand side of (1.3.14) are equal; their
sum is therefore 3aijkxixJxk. This proves (1.3.13).

EXAMPLE 1.3.3

Suppose that the equations


a

ipaJpcj=

b c

(1.3.15)

OipOjp = bij

(1.3.16)

UJ

hold for arbitrary c,. Show that


Solution Since equations (1.3.15) hold for arbitrary c z , these hold when
cx = 1, c2 = 0, c3 = 0, and we obtain
< V I P = bn

(1.3.17)

Similarly, taking cx = c 3 = 0, c2 = 1; and cx = c2 = 0, c 3 = 1 in (1.3.15),


we get, respectively,
<*<*2 = bi2;

aiP<iip = bi2t

Clearly, (1.3.17) and (1.3.18) constitute the required result (1.3.16).

(1.3.18)

1.4
SUMMARY OF RESULTS IN VECTOR ALGEBRA
A vector is an entity that has two characteristics: magnitude and direction.
Force and velocity are two typical examples of a vector. Geometrically, a
vector is represented by a directed line segment, with the length of the
segment representing the magnitude of the vector and the direction of the

1.4

SUMMARY OF RESULTS IN VECTOR ALGEBRA

segment indicating the direction of the vector. Evidently, the magnitude of


a vector is a nonnegative real number.
Vectors are denoted by boldface letters in print (like a) or letters with a
superposed arrow in manuscripts (like a). In this text, only lowercase letters
are used to denote vectors, reserving capital letters for tensors (to be introduced later). The magnitude of a vector a is denoted by |a| or a. A vector
a is called a unit vector if |a| = 1. A unit vector directed along a vector a is
denoted by . A vector whose magnitude is 0 is called the zero vector; it is
denoted by 0 (or 0).
Two vectors are said to be equal if they have the same magnitude and the
same direction. If a is a vector and a is a real number, then aa, called the
scalar multiple of a and a, is a vector whose magnitude is |a| |a|and whose
direction is that of a or opposite to a accordingly as a > 0 or a < 0. If
a = 0, then aa = 0. The vector (-l)a is denoted by - a . Every vector a can
be written as a = |a|.
Two vectors are said to be collinear if their directions are either the same
or opposite. If a and b are collinear vectors, then a = ab for some real
number a.
The sum a + b and the difference a - b of vectors a and b are defined by
the parallelogram law (see Figure 1.1). It is easy to verify that b + a =
a + b, b - a = -(a - b), a + 0 = a, and a - a = 0.
Suppose that there is a righthanded rectangular system of Cartesian axes
with a fixed origin O. We refer to these axes as the X\,x2, *3-axes, or briefly
the jC/-axes. Also, we denote the unit vectors directed along the positive
X\ *2 > *3-axes by ej, e 2 , e 3 , respectively (see Figure 1.2). Then e, are called
the base vectors of the xx;-system.
By virtue of the definition of equality of vectors and the parallelogram
law of addition of vectors, every vector a can be expressed as a linear

Figure 1.1. Parallelogram law.

SUFFIX NOTATION

Figure 1.2. Base vectors.

combination of e!, e 2 , e3 as follows:


a = alel + a2e2 + 3e3

(1.4.1)

Here al9a2,a3 are real numbers representing the projections of a on the


axes; these are called the components of a along e<, e2 and e 3 , respectively.
Since
e! = lej + 0e2 + 0e3
the components of e! are (1,0,0). Similarly, the components of e2 are
(0, 1, 0) and those of e3 are (0, 0, 1).
By employing the summation convention, expression (1.4.1) can be
rewritten as
(1.4.2)
a = <*k*k
The three components al9a2, a3 of a are referred to briefly as at. However,
the symbol at is also frequently used to denote a typical component of a; so
ai is referred to as the /th component of a.
We denote the /th component of a also by [a],; thus
M; = as

(1.4.3)

If a is a real number, then the vector aa is given by


aa = (aa^i

+ (afl2)e2 + (atf3)e3 = aakek

(1.4.4)

Thus, the components of cxa are ^; that is,


[aa]z = ctdi = a [a]/

(1.4.5)

As a direct consequence of this, it follows that -at are components of - a .

1.4

SUMMARY OF RESULTS IN VECTOR ALGEBRA

Let b be a vector with components bt\ that is,


b = bxtx + b2e2 + 3 e 3 = bktk

(1.4.6)

Then, by the definition of equality of vectors, it follows that a = b if and


only if ax = bi9 a2 = b2, a3 = b3, or briefly
*/ = *>/

(1.4.7)

Thus, the vector equation a = b is equivalent to the equations at = bt


written in the suffix notation.
By the parallelogram law of addition of vectors, it follows that the sum
a + b and the difference a - b of a and b are given by
a b = ta bx)ex + (a2 b2)e2 + (a3 b3)e3 = (ak bk)ek

(1.4.8)

The components of a b are thus a{ Z?f; that is,


[a b], = a{ b, = [a], [b],

(1.4.9)

A direct consequence of (1.4.9) is that the components of the zero vector


0 are all 0.
The scalar product or the dot product a b of a and b is a real number
defined by
a b = |a||b|cos0

(1.4.10)

where is the angle between the directions of a and b. Evidently,


b a = a b.
Two vectors are said to be orthogonal if the angle between their directions
is /2. It follows that a and b are orthogonal if and only if a b = 0.
Obviously, a 0 = 0 for every vector a. Also, if a b = 0 for every vector
a, then b must be the zero vector. This means that the zero vector is the only
vector that is orthogonal to every vector.
In terms of components, a b is given as follows:
a b = albl + a2b2 + a3b3 = akbk

(1.4.11)

In particular, a ej = ax, a e2 = a2, a e3 = a3, or briefly,


a -6, = ^ = [a],

(1.4.12)

Note that for specified e, equation (1.4.1) determines a when at are known
and equations (1.4.12) determine at when a is known. Thus, for specified e,
a vector a is completely determined by the ordered triplet (tfz) = (ax, a2, a3).
This triplet serves as a representation of a in the x,-system.
From expressions (1.4.10) and (1.4.11), it follows that
a a = |a| 2 = a\ + a\ + a\ = akak

(1.4.13)

10

SUFFIX NOTATION

From expressions (1.4.10) and (1.4.12), we find that


(1.4.14)

a, = |a|cos<9f

where 0, is the angle that (the direction of) a makes with the xr axis. The
numbers cos 0, are therefore the direction cosines of a. From (1.4.14) we
note that the numbers at are proportional to these direction cosines; at are
therefore the direction ratios of a. Thus, for any vector a the components
ai are its direction ratios.
In particular, if a is a unit vector, we have |a| = 1, and expression
(1.4.14) shows that at are the direction cosines of a. Thus, for any unit
vector a, the components at are its direction cosines. It follows that the
direction cosines of a vector are the components of the unit vector directed
along the vector.
The vector product or the cross product a x b of a and b (Figure 1.3) is
the vector defined by
a x b = |a||b||sin0|n

(1.4.15)

where 0 is the angle between the directions of a and b and n is the unit vector
that is perpendicular to both a and b and that is in the direction of the
advancement of a righthanded screw (corkscrew) rotated from a to b. It
follows that (i) a x a = 0, (ii) b x a = - ( a x b ) , and (iii) a and b are
collinear if and only if a X b = 0.
Obviously, a x 0 = 0 for every vector a. Also, if a x b = 0 for every
vector a, then b must be the zero vector. Hence, the zero vector is the only
vector that is collinear with every vector.
In terms of components, the cross product a x b is given by
a x b = (a2b3 - tf32)ei + (a3bi - axb3)z2 + (fl\b2 - ^2^i)e3 (1.4.16)
From (1.4.15), we find that |a x b| represents the area of the parallelogram whose sides are represented by a and b.
.axb

n{ y
l1

laxbl
a

*s

Figure 1.3. Vector product of a and b.

1.4

SUMMARY OF RESULTS IN VECTOR ALGEBRA

11

Figure 1.4. Parallelopiped determined by vectors a, b, c.

If c is a vector with components ci9 then the scalar triple product


a (b x c) is the scalar product of a and b x c; it is given by
a (b x c) = ax{b2c3 - b3c2) + a2(b3cx - bxc3) + abxc2 - b2c{)
ax

a2

a3

bx

b2

b3

Ci

c2

c3

(1.4.17)

It can be easily verified that


a (b x c) = (a x b) c

(1.4.18)

The absolute value of a (b x c) represents the volume of the parallelopiped


whose adjacent edges with a common corner are represented by a, b, c
(see Figure 1.4).
It follows that a, b, c are coplanar (that is, a, b, c are all parallel to the
same plane) if and only if a (b x c) = 0. In such a situation, each of a, b, c
can be expressed as a linear combination of the other two; for example,
a = ab + c for some real numbers a and . The products a (b x c) and
(a x b) c are also denoted by [a, b, c].
The vector triple products a x (b x c) and (a x b) x c are the vector
products of the vectors a and b x c and a x b and c, respectively, in the
order indicated. These two products are different in general; in fact, the

12

SUFFIX NOTATION

products have the following explicit forms:


a x (b x c) = (a c)b - (a b)c

(1.4.19)

(a x b) x c = (a c)b - (b c)a

(1.4.20)

It can be proven that a x (b x c) = (a x b) x c if and only if


b x (c x a) = 0.

1.5
SUMMARY OF RESULTS IN MATRIX ALGEBRA
By a matrix we mean a rectangular array of elements. The elements may
be real or complex numbers or any other mathematical objects. A matrix
having m rows and n columns (where m and n are positive integers) is
referred to as an m x n matrix; if m = n, the matrix is called a square
matrix of order n. In this text, we will be mainly dealing with square
matrices of order 3 whose elements are real numbers. We denote such a
matrix by [a^] whose explicit form is as follows:

[*(/] =

an

al2

a3l

a32

2\

12

al3
a

2Z

(1.5.1)

a33

Note that for a given / and given j, ^ denotes the element in the /th row and
theyth column in the matrix [au]. This element is referred to as the (/y)th or
typical element of the matrix. In particular, the elements an, a22 and a33 are
called the diagonal elements of [a^]. If in a square matrix all elements
except the diagonal elements are 0, the matrix is called a diagonal matrix.
Two matrices [^] and [by] are said to be equal if their corresponding
elements are equal; that is, an = bn, al2 = bl2,..., a33 = b33, or, briefly,
au = by. Thus, the matrix equation [au] = [by] is equivalent to the
equations ay = bi} in the suffix notation.
The matrix obtained by interchanging the rows and the columns of a
matrix [a^] is called the transpose of [a^] denoted by [%] r . Thus,
[*</Ir =

#11

#21

#31

\2

#22

#32

#13

#23

033

(1.5.2)

1.5

SUMMARY OF RESULTS IN MATRIX ALGEBRA

13

It is evident that the typical element of [au] is an\ therefore,

KV = []

(1.5.3)

( M 1 r = [*</]

(1.5.4)

It is trivial to verify that

The determinant formed by the elements au of a matrix [ay] is called the


determinant of the matrix [au]; it is denoted by det [#{,]. Thus,

detfo,] =

^11

#12

#13

22

a23

#31

#32

#33

(1.5.5)

Since the value of a determinant remains unchanged if its rows and


columns are interchanged, it follows that
(1.5.6)

det[au] = dettyi] = det[auY

The matrix obtained by multiplying every element of a matrix [ay] by a


number a is called the scalar multiple of a and [a,,]; denoted by a[a^\.
Thus,
<*[*#] = [afltf] =

aflrn

afl12

a 13

ar

ar

22

23

(1.5.7)

ar31 oca32 aa33


Evidently,
[aau\T = [aa] = [aau\T

(1.5.8)

det[aur/y] = a 3 det [#{,-]

(1.5.9)

Also,
In particular, the scalar multiple of - 1 and [7] is called the negative of
[ay], denoted by -[#<,]. Thus,
-[*</] =

[-Oij]

(1.5.10)

A matrix [] is said to be symmetric or skew-symmetric, accordingly, as


laij\T = [ff</] o r ' equivalently, ^- = 0^. Thus [##] is symmetric if and
only if a2X = ai2, a3X = al3 and a32 = a23, and skew-symmetric if and only
if an = a22 = a33 = 0, a2l = -al2, a3l = -ai3 and a32 = -a23.
Given the matrices [au] and [&,,], the matrix obtained by adding the
corresponding elements of [aiy] and [by] is called the sum of [a#] and [,,],

14

SUFFIX NOTATION

denoted by [ay] + [by]. Thus, by definition,


(1.5.11)
Similarly, the matrix obtained by subtracting the elements of [by] from the
corresponding elements of [ay] is called the difference of [ay] and [bu] in
that order; it is denoted by [au] - [by]. Thus, by definition,
[Oij] - [bij] = [au - bij]

(1.5.12)

[by] - [aij] = -{[aij] - [bu]}

(1.5.13)

It is trivial that

For any matrix [ay], the matrix [ay] - [ay] is obviously a matrix all of
whose elements are equal to 0. Such a matrix is called the zero matrix; we
denote it by [0]. Thus,
(1.5.14)

[ay] - [ay] = [0]


where
|~0 0
[0] =

0 0 0

(1.5.15)

[ 0 0
For any two matrices [au] and [by], we have
([ay] [by])T = [ay]T

[byf

(1.5.16)

For any matrix [ay], it can be verified that [ay] + [ay]T is a symmetric
matrix, and [ay] - [ay]T is a skew-symmetric matrix. Also,

M = *([*(/] + M 7 ) + < - [*/])

(1.5.17)

The first matrix in the righthand side of (1.5.17), namely j([ay] + [ay]T),
is symmetric and the second matrix, namely j([ay] - [ay]T), is skewsymmetric. As such, every matrix [ay] can be represented uniquely as a
sum of a symmetric matrix and a skew-symmetric matrix. The matrix
i([ffi/] + [ay]T) is called the symmetric part of [ay] and the matrix
iflfftf] - [Qij]T) the skew-symmetric part.
Given the matrices [a,,] and [by], the product of [<*#] and [&,,] in that
order is defined as a matrix [cy] where
c

aikbkj = riAA,

*= i

(1.5.18)

1.5 SUMMARY OF RESULTS IN MATRIX ALGEBRA

15

This product is denoted by [ay][by]. Thus,


[au][bu] = [aikbkJ]

(1.5.19)

[by][ay] = [bikakJ]

(1.5.20)

In the same way


From (1.5.19) and (1.5.20), it is evident that [au][by] * [bu][au] in general.
The product [][] is called the square of the matrix [ay], denoted by
[a0]2. Thus
[au)2 = [a][au] = [aikakJ]

(1.5.21)

Higher powers of [a,,] are defined in an analogous way.


By virtue of expression (1.5.19), it follows that, for any two matrices [ay]
and [by], we have
[ay][by]T = [aikbjk]

(1.5.22)

(1.5.23)

[ay] [by] = [akibkj]


T

([ay][by]) = [aJkbki] = (iby] [ay] )

(1.5.24)

det([ay][by]) = (det[ay])(det[by])

(1.5.25)

Also, for any three matrices [], [by] and [cy], we have
[ay]([by][cy]) = ([ayUbyUcy]

(1.5.26)
(1.5.27)

[ay]([by] [Cy]) = [ay][by] iay][Cy)


<[*</] [6y])tc,7] = K][Cy] [by][Cy]

(1.5.28)
-1

Given a matrix [ay], if there exists a matrix, denoted by [fly] , such that
[au\[auYl = [ ^ ' = [/]

(1.5.29)

where [/] is the matrix given by


"10
[/] =

0 1

0
(1.5.30)

0 0
then we say that [ay] is an invertible matrix or a nonsingular matrix; the
matrix [ay]~l is then called the (multiplicative) inverse of [ay]. Otherwise,
[ay] is said to be a singular matrix. It can be proved that [ay] is nonsingular
if and only if det[ay] ^ 0.
The matrix [/] given by (1.5.30) is called the unit matrix or the identity
matrix (of order 3). It is trivial that det[/] = 1; as such [/] is nonsingular.

16

SUFFIX NOTATION

The inverse of [/] is itself. Also, for any matrix [a^], we have
[/][*</] = K i m = [*<,]

(1.5.31)

From results (1.5.25) and (1.5.29), it follows that for any non-singular
matrix [au],

de,[

1 "'- s i b

<|5 32)

It can be proven that if [^] and [by] are nonsingular matrices, then
[fli/Hoi/] is also nonsingular and
( f a i r 1 = [

(1.5.33)

A matrix [au] is said to be orthogonal if it is nonsingular and its inverse


is equal to its transpose; that is, if
lau]Tlau] = lau][au]T = [I]

(1.5.34)

It follows that for an orthogonal matrix [a^]9


det[au] = 1

(1.5.35)

An orthogonal matrix [a^\ for which det [au] = 1 is called a proper


orthogonal matrix.

1.6
THE SYMBOL lf
Recall that the base vectors tlyt2 and e3 are the unit vectors directed along
the xl9 x2 and x3 axes, respectively. Since the axes are assumed to be
rectangular, expression (1.4.10) yields the following relations:
ei = e2 e2 = e* e3 = 1
e

l * e2 =

2 * el =

l " e3 =

3 * el =

2 * e3 =

3 * e2 = 0

(1.6.1)

These nine relations may be represented in a condensed form by defining


a symbol ou as follows:
iJ

_ 1
^0

if / and j take the same value


if / and j take different values

(1.6.2)

That is,
^11 = ^22 = ^33 =

<*12 = <*21 = ^13 = ^31 = ^23 = <*32 =

1.6

THE SYMBOL <?0

17

Clearly, a condensed form of relations (1.6.1) is


*i-*j

(1.6.3)

= u

The symbol ou defined by (1.6.2) is known as the Kronecker symbol,


named after the German mathematician Leopold Kronecker (1823-1891).
The following property is inherent in the definition of :

(1.6.4)

= ji

It is important to note that, by summation convention,


<* = <*11 +<*22 + <*33 = 3

(1.6.5)

From (1.6.2), it is evident that u are the elements of the unit matrix [/]
defined by (1.5.30). Thus

"i o ol
[I] = [] =

1 0

(1.6.6)

0 0 lj
Consequently,
det(<5<,.) = 1

(1.6.7)

Write down the following equations in a condensed

EXAMPLE 1.6.1

form:

an = a(bn + b22 + b33) + bn


a22 = oi(bn + b22 + 633) + b22
a33 = <x(bn + b22 + 33) + b33

Solution

al2 = bl2,

a23 = b23,

a31 = b3i

a2l = b2U

a32 = b32,

al3 = b13

We have
an = cc(bu + b22 + b33) + bn = abkk + bn
= 0Lubkk + bn,

because = 1

Expressions for a22 and 33 may be written similarly. Also,


al2 = bl2=

2 bkk + bl2,

because 2 = 0. Expressions for a23y 39 2, a32, a13 may be written


similarly.

18

SUFFIX NOTATION

It is now obvious that the given equations may be expressed in the following condensed form:
a

ij

EXAMPLE 1.6.2

^Ubkk

+ bij

Given that
a

U=

Ubkk + bu

(1.6.8)

where * 0, 3 + ^ 0, find bu in terms of au.


Solution From (1.6.8), we get
app = 0LOppbkk + bpp = (3 + )bpp,
because = 3 and bkk = bPP

so that

bpp =

(1.6.9)

3 +

It also follows from (1.6.8) that


b

la -

(1.6.10)

a b

ij pp\

Substituting for bpp from (1.6.9) in (1.6.10), we obtain


a

EXAMPLE 1.6.3

(1.6.11)

*-JZT*a_

Prove the following:

(i) ijOj = ai9

(ii)

(iv) 3uau = au,

(iii) ^^ = aki

SuaJk = aik9

(v) uJkakm = aim.

(Results of this type are called substitution properties of ^.)


Solution

(i) Put bt = uaj. Then


b

^11^1 + ^12^2 + ^13^3

Similarly, b2 = a2, b3 = a3. Thus, bt = {. That is, uaj = ?,.


(ii) Put bik = Suajk. Then
bn = ] =
Similarly, b22 = 0 22 , b^
bl2

= a m

+ 122 +

= #11

Also,

= 2

+ 222

+ 13*32 = ^12

1.7

THE SYMBOL ,

19

Similarly, b2l = 21 , bn = 13 , bn = 31 , ft23 = 23 , b32 = a32. Thus,


bik = 0,-ifc. That is, uajk = T/:.
(iii) Put cik = aki. Then result (ii) yields ucjk = cik. That is,
ijakj, =

aki.

(iv) ijuij = an, by result (iii).


( v ) ijJkakm = ikakm = aim, by result (ii).
EXAMPLE 1.6.4

Prove the matrix identity (1.5.31).

Solution Since [/] = [Jl7] = [], expression (1.5.19) and results (ii) and
(iii) of Example 1.6.3 yield
[/][*</] = [ikakj] = [dij] = [Jkaik] = [aikkj] = [a^I]

1-7
THE SYMBOL eijk
It has been assumed that the coordinate system for which e, are base
vectors is a righthanded system; see Section 1.4. Consequently, by virtue of
expression (1.4.15), the vectors e, obey the following relations, in addition
to (1.6.1):
ei x e2 = - ( e 2 x ex) = e3
e 2 x e 3 = - ( e 3 x e2) = el

(1.7.1)

e3 x ex = -(el x e3) = e2
From relations (1.6.1) and (1.7.1), we obtain the following 27 relations:
e r (e2 x e3) = e2 (e3 x ex) = e3* (ci x e2) = 1
*1"

(e3 x e2) = e2 (ei X e3) = e3* (e2 x ej) = - 1

ei (ei x e2) = e2 (e2 x e3) = e 3 * (e3 x ex) = 0

e r (e2 x e t ) = e2 (e3 x e2) = e 3 - (ei x e3) = 0


e

i '

(e! x e3) = e2 " (e2 x eO =e 3 (e3 x e2) = 0

e r (e3 x ex) = e2 (ei x e2) = e 3 (e2 x e3) = 0


e r (e2 x e2) = e2 ' (e3 x e3) = e 3 (ei x e t ) = 0
ei (e3 x e3) = e2 (et x eO = e3* (e2 x e2) = 0
e

i '

(ei x ex) = e2 (e2 x e2) = e 3 (e3 X e3) = 0

(1.7.2)

20

SUFFIX NOTATION

These 27 relations may be represented in a condensed form by defining a


symbol eiJk as follows:
if /,y, k take values in the cyclic order,
if /,y, k take values in the acylic order,

(1.7.3)

if two or all of /,y, k take the same value.


That is,

123 1>

132

1>

111 112 ~ 0>

etc.

It is obvious that a condensed form of relations (1.7.2) is


(1.7.4)
The symbol eiJk defined by (1.7.3) is known as the Levi-Civita -symbol,
named after the Italian mathematician Tullio Levi-Civita (1873-1941). The
-symbol is also referred to as the permutation symbol, the alternating
symbol or the alternator.
The following facts can be easily verified:
(1.7.5)

eUk = Ui-J)U-k)(k-i)

ijk

eixeJ

jki

kij

~ikj

~jik

~~ekji

(1.7.6)

(1.7.7)

= eiJkek

The -symbol is useful in expressing the vector product a x b and the


scalar triple product [a, b, c] in the suffix notation as described in the
following.
For any vectors a and b with components #, and bt, respectively, we have
a x b = foe,) x (,-e,) = ^ - ( e ^ e * )
using (1.7.7)
= majbk*i

= ijk<tjbk*i,

(1.7.8)

on changing dummy suffixes and using (1.7.6). Thus, the symbol eiJkajbk
denotes the ith component of the vector a x b; that is,
[a x b], = eUkajbk = eUk[*]j\b]k

(1.7.9)

Since a x a = 0, it follows that


eukajak = 0
This expression can also be verified directly.

(1.7.10)

1.7

THE SYMBOL

21

For any vector c with components c, we have


[a, b, c] = (a x b) c = [a x bjjc], = eUkajbki

(1.7.11)

on using (1.7.9). Interchanging the dummy suffixes and using (1.7.6),


(1.7.11) can be rewritten as
[a, b, c] = e^ibjCf,

(1.7.12)

Recalling that [a, b, c] = a (b x c) also, we find from (1.4.17) and (1.7.12)


that

ijkaibjck

ax

a2

a3

bl

b2

b3

Ci

c2

c3

(1.7.13)

This expression can also be verified directly.


EXAMPLE 1.7.1

ijk

Show that
,

e, e 2

e, e 3

*/'ei

*j e 2

e, e 3

e* ej

e*e3

* ' e2

an an
an *n

k\

^ki

k2

(1.7.14)

Solution We begin with expression (1.4.17). Noting that for any vector a
with components at, we have at = a e,, this expression can be rewritten as
a ej

a e2

a e3

a - ( b x c) = | b - e t

b e2

b e3 |

c e!

c e2

c e3

(1.7.15)

Setting a = e,, b = e,, c = e* in (1.7.15) and using (1.7.4), it turns out that
e, e1

ijk

e, e2

k ' el

k ' e2

which, on using (1.6.3), gives

ijk

e, e3

<*

i3

j\

j3

k\

kl

k3

fc * e 3

22

SUFFIX NOTATION

EXAMPLE 1.7.2
Solution

Show that au = an if and only if eijkajk = 0.

For / = 1, we have
3
e

ijkajk

ijkajk

3
E

7=1k=1

\jkajk

= (vijl + eWan

+ *WaJ*)

7=1
3

= tel/2tf/2 + 173^3)
7=1

1 3 2 # 3 2 + ^123 #23

2 3 ~~ #32

by using the definition of the symbol.


Similarly, we get
&Ukajk = #3i - # n

for / = 2

ijk<*jk = <*\i - #21

for / = 3

Thus
r23 Zijk<*jk =

tf32

< 31 -

for / = 1

for / = 2

13

(1.7.16)

for / = 3

axl - alx

If eijkajk = 0, then from (1.7.16) it follows that a23 = # 32 , #31 = 13 and


12

21 > that is ij =

Conversely, if a^ = aJi9 then an = alx, al3 = a3i and a13 = 32 Consequently, the lefthand side of (1.7.16) is 0 for all /'; that is, eiJkaJk = 0.
EXAMPLE 1.7.3

(i) Show that

^pqr &ip &jq &kr

Oil

Oil

J\

J2

k\

k2

0/3
a

J3

(1.7.17)

k3

(ii) Deduce that


det(ur0) = epqalpa2qa3r

epqraplaq2ar3

(1.7.18)

1.7

Solution
pqraipajqakr

23

(i) We have
3

THE SYMBOL <

L
Li
p=lq=\

p=lq=\

Li
r=\

pqraipajqakr

<*ip<*jq(Zpq\<k\ + epq2<*k2 +

^pqS^ki)

3
=

<*ip[<tk\(p2\<*j2 + 31/3) + <*k2ip\2(*j\ + Cp32*/3)

+ ^ 3 ( ^ 1 3 ^ / 1 + ^23^/2

(other terms being 0 by the definition of the -symbol)


= **(321*|3*/2 + 2310/20,3) + 0*2^312 0/3 0yl + 132^/1 % )
+ 0*3^213 0/2 0,1 +

ei23<*il<lj2)

(other terms being 0)


= an(aj2ak3 - anak2) + ai2(aklan - ak^ajx)
+ 0/3(0*2 0 y i -

0*10,2)

(on using that 123 = 1, i32 = - 1 , etc.)


0/1

0/2

0/3

0,1

0y2

0,3

0*1

0*2

0*3

(ii) In the above result if we set / = 1, j = 2, k = 3, we obtain


^ r 0 i p 0 2 ^ 0 3 r = det(f l7 )

(1.7.19)

Suppose we set bu = #,,. Then


e

pqraplaq2ar3

pqr^lp^2q^3r

= det(bu),

by (1.7.19)

= deHaji) = det(uri7)

(1.7.20)

because det(a^) = det(a^). Expressions (1.7.19) and (1.7.20) are the


required results in (1.7.18).

24

1 SUFFIX NOTATION

EXAMPLE 1 . 7 . 4 (i) Let D = det(a l7 ). Verify that

ijk pqr*-'

(*ip

<*iq

<*ir

jp

1 akp

JQ

kq

(1.7.21)

Jr

kr

(ii) Hence deduce the following results:


(a)

ijkpqr "~

(b)

ijk pqk

(c)

ijkpjk

(d)

ijkijk

sip

siq sir

Sjp

Sjq

SJr

Skp

Skq

Skr

SipSjg

(1.7.22)
(1.7.23)

SiqSJp

(1.7.24)

2Sip

(1.7.25)

Solution (i) If (at least) two of i9j, k or two of /?, q, r are equal, then both
sides of (1.7.21) are 0. If ij, k and p, q, r are both cyclic or acyclic, then
each side of (1.7.21) is equal to D. If /,y, k are cyclic but p , q, r are acyclic
or vice versa, then each side of (1.7.21) is equal to -D. Thus, in all possible
cases, result (1.7.21) is verified.
(ii) (a) If we set ay = Su in (1.7.21) and recall that det(<?0) = 1, we
readily get (1.7.22).
(b) It follows from (1.7.22) that

ijkpqk

Siq

Sik

*JP

Sjq

Sjk

Skp

Skq

Skk

= Sip(Sjqkk

- SjkSkq) + Siq(jkSkp

+ Sik(SjpSkq

Sjpkk)

SjqSkp)

= Sip(3Sjq - Sjq) + Siq(SJp - 3Sjp) + {SjpSiq = SipSjq - SiqSjp

(c) Hence
inSu - S
uSin = 3Sin - Sin
ijk PJk = S
'W^JJ
'U"JP

(d) Consequently
E

ijkeijk 20ji 6

2:,

SjqSip)

1.7

THE SYMBOL ,,,

25

Note: The identity (1.7.21) and its particular cases (1.7.22) to (1.7.25) are
of great utility in subsequent discussions. In particular, the identity (1.7.23)
is known as the - identity or the permutation identity. Since a cyclic
permutation of suffixes does not change the sign of the permutation
symbol, this identity may also be expressed in the following alternative
forms:

ijkpqk

ijkkpq

EXAMPLE 1.7.5

= e

kijkpq

= e

jkiqkp

^ip^jq ~ ^iq^jp

(1-7.23)'

Show that

det(uri7) = }eUkePQraipajqakr

= iUkePQrapiaqjark

= 6(<*u<ijj(tkk + 2auajkaki

(1.7.26)
(1.7.27)

- }^)

If D = det(ur/7), deduce that


ZijkD = Zpqr<lip<jq<tkr = pqr<*pi<tqj<lrk

(1.7.28)

Solution Expanding the determinant on the righthand side of (1.7.17),


we obtain
epqr<tipajq<*kr = 0 i l ( * / 2 * * 3 ~

+ an(anak2

j3akl)

+ *i2(*/3**l ~

^jl^kl)

(1.7.29)

- aklaJ2)

Now
ijk<tii(aj2ak3 - aj3ak2) = ^ijkaixaj2ak3

= *Ukailaj2ak*

eUkanaJ3ak2

- Cikj<*il<lj2<lk3

(on interchanging the dummy suffixes j and k in the second term)


= 2eUkanaJ2ak3
since eiJk = -ikj.

(1.7.30)

Similarly,
i/**i2(**i*/3 - <*k*<*j\) = l*ukai2aklan

(1.7.31)

Interchanging the dummy suffixes / and j , j and k, and k and / in


ijkai2akiaj3 and noting that eijk = ekij, we find that

ijk^i2^ki^J3

= kij^il^J2^k3

Zijk<til<tj2<tkl

Hence, (1.7.31) becomes


ijk<ii2(aki<ij3 - <*ki<ij\) =

teukanaj2aki

(1.7.32)

In the same way we obtain


ZUkail(ak2<*j\

k\<*j2) =

2ijk<til<tj2<tk3

(1.7.33)

26

SUFFIX NOTATION

From (1.7.29), (1.7.30), (1.7.32) and (1.7.33), we get


ijkEpgrOiPajgakr =

6eijkananaki

(1.7.34)

Noting that eiJkepqraipajqakr = eijkepqrapiaqJakr (interchanging the dummies


/ and/?,y and q, and k and r) and noting from (1.7.18) that ijkanaj2ak3 =
det(uri7), we obtain (1.7.26) from (1.7.34).
Next, it follows from (1.7.22) that

ijkpqraipajqakr

(ipajqakr)

**

iT

jP

*jr

kp

kq

kr

= {iP(jqkr - kqjr) + iq(jrokp - jpkr)


+ ir(jpkq - kpjq))aipajqakr

(1.7.35)

Using the substitution property of otj and changing the dummy suffixes
appropriately, we find that
ip(ojqkr - jrkq)aipajqakr = ^ - }^

(1.7.36)

iq(jrkp - jpkr)aipajqakr = auajkaki - auaakk

(1.7.37)

ir(jpkq - kqjp)aipajqakr = auajkaki - auaakk

(1.7.38)

Substituting (1.7.36)-(1.7.38) in the righthand side of (1.7.35) and using


the resulting expression in (1.7.26) we arrive at result (1.7.27).
Finally, multiplying (1.7.26) throughout by eijk and using (1.7.25), we
obtain (1.7.28).

EXAMPLE 1.7.6
Show that

Given the matrix [a^], consider the matrix [afj] where


&ij ~

l^ipqSjrsQprQqs

[au][afj]T = [ajjflay] = D[I]

(1.7.39)
(1.7.40)

where D = det(a0). Deduce that, if D * 0,


1
[/1 = l f i l
Solution

(1.7.41)

By virtue of ( 1.5.22), we have

= li*M

1<$][] = [a*kiakj]

(1.7.42)

1-7

THE SYMBOL ,,

27

Using (1.7.39), (1.7.28) and (1.7.24) we find that


<*ik<*jk = jejPQekrsaikapraQS = \ejpq{peipq) = ^
a*ki<*kj = \ZkpqZirs<lpr<lqs<lkj

= J^rsEirsD

= }

(1.7.43)
(1.7.44)

Using (1.7.43) and (1.7.44) in (1.7.42) and noting that [<5/7] = [/], we get
(1.7.40).
If D 5* 0, then the matrix [au] is invertible, and it follows from (1.7.40)
that
[ , - 7 , = D[auyl[I]
which on using (1.5.29) and (1.5.31) yields
[afjf = D[aurl
This is the expression (1.7.41).

Note: The matrix [afj] is called the adjugate or co factor and its transpose
[aj]T = [] is called the adjoint of the matrix [au]. Expression (1.7.41) is
useful to compute the inverse of a nonsingular matrix.
EXAMPLE 1-7-7

Given the vectors a,b,c:

(i) Write down the /th component of a x (b x c).


(ii) Use the vector identity (1.4.19) to prove the - identity (1.7.23).
(iii) Use the - identity to prove the identity (1.4.19).
Solution
obtain

(i) If we put d = b x c, then by virtue of expression (1.7.9), we


^ = [ b x c]* = kpqbpcq

(1.7.45)

Also,
[a x (b x c)L = [a x d], = eUkajdk
which by use of (1.7.45), becomes
[a x (b x c)L = eijk8kpqajbpcq

(1.7.46)

Thus, ijkekpqajbpcq is the /th component of a x (b x c).


(ii) It follows from the vector identity (1.4.19) that
[a x (b x c)]f = (fljCjWi - (ajbj)Ci

(1.7.47)

Using (1.7.46), (1.7.47) becomes


ZijkZkPq<ijbpcq

= ajCjbi

- ajbjCi

(1.7.48)

28

SUFFIX NOTATION

By using the substitution properties of ij9 we find that


ajCjbi = aj(jqcq)(ipbp)

ipjqajbpcq

ajbjCi = aj(jpbp)(iqcq)

iqjpajbpcq

By use of these expressions, (1.7.48) becomes


iijktkpq - ipjq + iqjp)ajbpcq

= 0

(1.7.49)

This should hold for arbitrary a,, &f and ct. Hence
+ iqjp

ZijkZkpq - ipjq

= 0

This is the - identity (1.7.23). (This serves as an alternative proof for the
- identity.)
(iii) If the - identity (1.7.23) holds, then (1.7.49) is valid for arbitrary
ai9bi9Ci. But, (1.7.49) is equivalent to (1.7.47). Thus, (1.7.23) yields
(1.7.47), which is nothing but (1.4.19).

EXAMPLE 1.7.8

Prove the vector identity

(a x b) (c x d) = (a c)(b d) - (a d)(b c)

(1.7.50)

Deduce that
|[(a x b) (b x c) + (a b)(b c)] = (a b)(b c) - \b\*

c).

(1.7.51)

Solution By use of (1.4.11), (1.7.9), (1.7.23) and the substitution


property of Su, we find that
(a x b) (c x d) = [ a x b]Jc x d]^
=

(kpqapbq)(kmnCmdn)

= (pmqn ~
= (ombn -

qm)(apbqcmd')

anbm)(cmdn)

= (omcm)(bndn)

(andn)(bmcm)

= (a-c)(b-d)-(a-d)(b-c),
which is (1.7.50). As a particular case, we get
(a x b) (b x c) = (a b)(b c) - (a c)(b b)
Adding (a b)(b c) to both sides of (1.7.52) gives
(a x b) (b x c) + (a b)(b c) = 2(a b)(b c) - (a c)b2
which is (1.7.51).

(1.7.52)

1.8

EXERCISES

29

1.8
EXERCISES
1. State which of the following expressions are meaningful in the suffix notation.
Write out the unabridged versions of the meaningful expressions:
(i)

(ii) aubj

(iii) aubi

(iv) aubi

(v) aubjj

(vi) aub

(vii) arsbsr

(viii) arsbss

(ix) aijkbik

2 . Which of the following expressions have the same meaning?


dijbj,

arsbs,

apqbp,

a^bj,

apqbpbq,

asrbsbr

3 . State which of the following equations are meaningful in the suffix notation.
(i) Xg = (Xijyj

(ii) y =

(iii) au = ctuOLjj

(iv) au = otipotjp

(v) au = oLimoijnbmn
(vii) au = otimajnbrs

aikxk

(vi) ars = <xircLjsbu


(viii) apq =

apraqsb

4. Which of the following expressions have the same meaning as { =


(i) ap = oipibi

apibpl

(ii) ap = aipbi

(iii) am = 0Ljmbj

(iv) ar = asrbs

(v) ak = otkmbm

(vi) ak = arkbr

5. Write down the following system of equations in the suffix notation:


Oli=bll9

22 = ^22

33

13 = i(^13 + ^3l) = 31

^33>

12 = 2 + *>2l) = 21

23 = 1(^23 + ^32> = 32

6. If ij = -0,7, show that a^Xj = 0 for all xt.


7. If au = a and 6,, = -bJi9 show that ^ ^ = 0.
8. If au = and bu = y(c/7 + c,,), show that aubu = aucu.
9. If au = j(bij + &,,) and cu = %(bu - b), show that ayCy = 0.
1 0 . In the system of axes with base vectors ef, show that a = (a ef) e, for every
vector a.
1 1 . Prove the expressions (1.4.16) through (1.4.18).
1 2 . The absolute value of the scalar triple product [a, b, c] represents the volume of
the parallelopiped whose adjacent edges with a common corner are represented by
a, b, c. Justify this statement.

30

SUFFIX NOTATION

1 3 . Prove the identities (1.4.19) and (1.4.20). Deduce that a x (b x c) =


(a x b) x c if and only if b x (c x a) = 0. Interpret the condition geometrically.
1 4 . Given the matrix
1 0 -1
[*</] =

-1

0 -1

3
find*?,,, dijji and

dijdij.

1 5 . Represent the following matrix as a sum of a symmetric matrix and a skewsymmetric matrix:

\flu\ =

-6

10 - 8

1 6 . For the matrix given in the previous exercise, compute [^][^]9

[^^]

[ay]2.

and

1 7 . Prove the expressions (1.5.4), (1.5.6), (1.5.16), (1.5.24) to (1.5.28), (1.5.33)


and (1.5.35).
1 8 . Prove that a matrix [^\ has a multiplicative inverse if and only if det[a^] ^ 0.
1 9 . Verify that the following matrices are nonsingular. Find their inverses.
1 0

0) 0 0 2
0

(ii)

2 3

-1

0-1

2 0 . Verify that the following matrices are orthogonal.

o o"
(i)

0 1

(ii)

COS0

-sine?

sin0

COS0

[o i o
1
V2
(iii)

-T=

<2

"Vf V2
0

0
-1

(iv)

V2
1
V2

"7=

V2

7 =

V2

1.8

EXERCISES

31

2 1 . Show that
(i) ijij = 3
(iii) ikjmij

(ii) ijjkik

= 3

= km

2 2 . Simplify the following


(ii)

(i) u{au - a)
(iii) (u + au)(u -

ipjqapbjCq

a^

2 3 . Write down the following equations in matrix forms:


(i) otipOtjp = ou
(iii) au = aubkk

(ii) oLpiOLpj = u
+ bu

2 4 . Write down the following sets of equations in condensed forms:


(i) au = a22 = a33 = -p
<*\2 = 021 = #23 = #32 = 13 = tf31 = 0

(ii) ani + a122 + ff133 + ^ = 0,


#211 + ^222 + #233 +

2 = >

#311 + #322 + tf333 + b3 = 0.

2 5 . If 0f = eijkbjk,

evaluate ^, <72, a3.

2 6 . If aik = eijkbj, show that aik =

-aki.

2 7 . Prove the following:


(i) ekki = 0

(ii) ^,,* = 0

(iii) epqrerqs = -2ps

(iv) epqrerqp = - 6

2 8 . Verify the relation (1.7.7), and using this relation deduce the relation (1.7.4).
2 9 . Show that

ijk<*ibj

a2

a3

by

b*

3 0 . Show that vectors with components ai9bi9Ci


Zijk<*ibjCk

are coplanar if and only if

= 0.

3 1 . If au = - and bt = jeijkajk,

show that apq = epqrbr.

3 2 . If n is a unit vector with components nt and v is a vector with components vi9


show that
Vi = vknkni +

eijkekrsnjvrns

32

SUFFIX NOTATION

33. Write down the following equations in the suffix notation:


(i) x = a(a x b) + jffa
(ii) (a - b) (c x d) = 0
(Hi) x = b + exe x {(a - b) X c)
(iv) (a x b)2 = a2b2 - (a b)2
34. Prove the following vector identities by using the suffix notation:
(i) (a x b) x c = (a c)b - (b c)a
(ii) (a x b) x (c x d) = [a, b, d]c - [a, b, c]d = [c, d, a]b - [c, d, b]a
(iii) [b x c, e x a, a x b] = [a, b, c]2
35. Deduce the - identity by using the vector identity (1.4.20).

CHAPTER 2

ALGEBRA OF TENSORS

2,1
INTRODUCTION
This chapter is devoted to the study of some algebraic aspects of Cartesian
tensors. The concept of a Cartesian tensor is introduced through certain
rules of coordinate transformations, and these rules are employed to establish some basic algebraic properties of tensors. Second-order tensors are
interpreted as linear operators on vectors; these are studied in some detail
with the aid of matrices.

2.2
COORDINATE TRANSFORMATIONS
Let us consider a righthanded system of rectangular Cartesian axes with a
fixed origin O and denote the coordinates of a general point P with respect
to (w.r.t.) this system by (xx, x2, x3), or briefly, (*,). Also, let the position
vector of P w.r.t. O be denoted by x. Then, xt are the components of x along
the axes, and by virtue of (1.4.1) we have
x = xlei + x2e2 + x 3 e 3 = xpep
33

(2.2.1)

34

ALGEBRA OF TENSORS

Figure 2.1. Two sets of axes at 0.

Now, let us suppose that there is another righthanded system of


rectangular Cartesian axes with the same (fixed) origin O and denote
the coordinates of P w.r.t. this "new" coordinate system by (*/). See
Figure 2.1. The new system of axes may be regarded as having been
obtained by rotating the "old" system of axes (namely, the jcraxes) about
some line in space through O. The position vector x of P has the following
representation in the new system
x = x[e[ + x^2 + x& = x'pt'p

(2.2.2)

where ei is the unit vector directed along the positive ArJ-axis and so forth.
Since the x\ system is also rectangular and righthanded, e satisfy equations
that are analogous to those satisfied by e,, namely (1.6.3) and (1.7.4). Thus,
(2.2.3)
ei ' / = <*</
(2.2.4)
[e'i, ej, e^] ijk{
From (2.2.1) and (2.2.2), we note that the vector x is represented as the
ordered triplet (#,) in the old system of axes and the ordered triplet (x[) in
the new system. These two triplets are different from one another as long as
the two coordinate systems are noncoincident and x ^ 0. We now proceed
to obtain relations connecting xt and x\.
We first introduce a matrix [^] whose elements a^ are defined as

U = e/ " ej = cos(x'i9Xj)
(2.2.5)
where COS(JC/, Xj) denotes the cosine of the angle between the positive x\ axis
and the positive Xj axis. Thus, with respect to the old system of axes,

> i2> 13 a r e the direction cosines of the x[ axis; 2 > 22 23 a r e the


a

2.2

COORDINATE TRANSFORMATIONS

35

Table 2.1. Table of Direction Cosines


*1

e2

e3

11

12

13

21

22

23

e3

31

32

33

direction cosines of the x'2 axis; and 31 , 32, a33 are the direction cosines of
the * 3 axis. The relations (2.2.5) may be displayed in a tabular form as
shown in Table 2.1, which is referred to as the table of direction cosines.
Taking the scalar product with e on both sides of (2.2.1) and using (2.2.5)
as well as the fact that x e = x\ (which follows from (2.2.2)) we get
x'i = + ai2x2 + ai3x3 = aipxp

(2.2.6)

Similarly, we obtain, from (2.2.1) to (2.2.5),


Xi = x- e, = (xpe'p) ef- = xp(e'p e,) = apixp

(2.2.7)

When the orientations of the new axes w.r.t. the old axes are known, the
coefficients au are known. Expressions (2.2.6) then determine x\ in terms of
xi9 and (2.2.7) determine *, in terms of x[. In other words, (2.2.6) represent
the law that transforms the triplet (x,) to the triplet (x) and (2.2.7) represent
the inverse law.
Substituting for xp from (2.2.7) in (2.2.6) gives aipajpx'j = x[. Since
xl = ijXj, we get aipaJpXj = ^x]. This relation holds for arbitrary xj. It
therefore follows that (see Example 1.3.3)
a

ipajp

(2.2.8)

Similarly, substitution of x'p in (2.2.7) from (2.2.6) leads to


(2.2.9)

apiapj = ij

When written in the unabridged form, the nine relations in (2.2.8) read as
follows
?1 + 2 + ?3 = 1
21

222 + 23

Il +

32

= 1

+ a323 = 1

11 21

12 22

1323

= 0

21 31

22 32

23 33

= 0

31 11

32 12

3313

= 0

(2.2.10)

36

ALGEBRA OF TENSORS

The first three relations in (2.2.10) show that the sums of the squares of
the direction cosines of the JCJ, JC^ and x'3 axes (w.r.t. je, axes) are each equal
to 1, as expected. The last three equations in (2.2.10) verify the fact that the
JCJ, x'2 and x'3 axes are mutually orthogonal. The corresponding relations for
the direction cosines of the xx, x2 and JC3 axes w.r.t. the x[ axes follow from
(2.2.9). The relations (2.2.8) and (2.2.9) are referred to as the orthonormal
relations for aiy.
In matrix notation, relations (2.2.8) and (2.2.9) may be represented,
respectively, as follows:
W k / = [ / ]

(2.2.11)

[u]Tl<*ul = W

(2.2.12)

These expressions show that the matrix [a,,] is nonsingular and that
[a//]" 1 = [<*i/]r; in other words, the matrix [a,,] is orthogonal. Therefore,
the transformation laws (2.2.6) and (2.2.7), determined by [a 0 ], are called
orthogonal transformations. The matrix [a,, ] is referred to as the matrix of
the transformation from the xt system to the x- system.
EXAMPLE 2.2.1 The JC/ system is obtained by rotating the JC, system
about the JC3 axis through an angle in the sense of the righthanded screw.
Find the transformation matrix. If a point P has coordinates (1, 1, 1) in the
Xi system, find its coordinates in the JC/ system. If a point Q has coordinates
(1,1,1) in the JC system, find its coordinates in the *, system.
Solution Figure 2.2 illustrates how the JC/ system is related to the jcf
system. From this figure we readily get Table 2.2 of direction cosines for the
given transformation (on using Table 2.1). Hence the matrix of the given

Figure 2.2. Rotation about the x3 axis.

2.2

COORDINATE TRANSFORMATIONS

37

Table 2.2. Direction Cosines for Example 2.2.1

e'i

e2

e3

COS0

sind
cos0
0

0
0
1

-sin#
0

transformation is
COS

[(/] =

0 sin0

-sin 0 COS0
0

0"

(2.2.13)

0
1

The coordinates of the point P in the x# system are given as xl = 1,


x2 = \9 JC3 = 1. Its coordinates x\ in the JC system may be obtained by
employing the transformation law (2.2.6). Using the values of ay given by
(2.2.13), the law (2.2.6) yields
*1 = l p * p = 11*1 + 12*2 + 13*3 =
X

*2 = <*2p p = 21*1 + 22*2 + 23*3 =

COS

0 + SHI 0

COS

0 - SHI 0

(2.2.14)

*3 = <*3 = 31*1 + 32*2 + 33*3 = 1

Thus, the coordinates of Pin the JC system are (cos 0 + sin 0, cos 0 - sin 0,1).
The coordinates of the point Q in the x\ system are given as x[ = 1,
x'2 = l, ^ = l. its coordinates JCf in the Arf system follow from the law
(2.2.7), which, on using (2.2.13), yields
Xl = OipiXp = OL\\X\ + 21*2 + 31*3 =

COS

0 - SHI 0

2 - <Xp2Xp = 12*1 + 22*2 + 32*3 = COS 0 + Sin 0

(2.2.15)

*3 = />3*p = 13*1 + 23*2 + 33*3 = 1

Hence, the coordinates of Q in the xt system are (cos 0 - sin 0,


cos 0 + sin 0, 1).

EXAMPLE 2.2.2

Complete Table 2.3 of direction cosines.

Solution By data, an = a2i = yvf, a12 = OL22 = 13 = -23 = 1/2V2.


The orthonormal relations (2.2.10) now yield the following conditions on
31 > 32 3 3
31 + 32 + 33 = 1

V531 + a32 + 0:33 = 0


V3a31 + a32 - 33 = 0

38

ALGEBRA OF TENSORS

Table 2.3. Direction Cosines for Example 2.2.2

-?-

1
2V2

2yj2

2V2

2V2
1
~2

2yj2

Solving these equations, we obtain


a3i =

32 = +

V3
2 '

33=0

and the Table 2.3 is completed.


Show that

EXAMPLE 2 . 2 . 3

[ej, ej, e'k] =

il

i2

i3

/I

<*/2

<*/3

A:l

*2

*3

(2.2.16)

Hence, deduce that


det[a l7 ] = 1
Solution

(2.2.17)

For any vector a, we have


a = (a e ^

+ (a e2)e2 + (a e3)e3

Accordingly,
ej = (e; e^ej + (c{ e2)e2 + (e{ e3)e3
= aneY + ai2e2 + a i 3 e 3
on using the definition of au. Similarly,
ej = antx

+ aj2t2 + a- 3 e 3

ei = oiklel + ak2e2 + *33


Hence, by virtue of expression (1.4.17), we obtain
il

,-2

i3

[ei,ej, e^] = | o ^

a, 2

a y3

*1

A:2

A:3

2.3 CARTESIAN TENSORS

39

In particular,
I 11

[ei,e 2 ,e^] =

12

13

ot2l a22 a23 I = det[a/7]


I 31

32

(2.2.18)

33

On the other hand,


[el, e 2 , ej] = (ei x e) ej = ej ej = 1.

(2.2.19)

From (2.2.18) and (2.2.19), we get


det[al7] = [ei,e 2 ,e^] = 1

Note: Expressions (2.2.11), (2.2.12) and (2.2.17) show that [a 0 ] is a


proper orthogonal matrix. The coordinate transformations determined by
(Xij are therefore called proper orthogonal transformations. Only such
coordinate transformations are dealt with in our further discussions; we will
refer to them simply as coordinate transformations (with no adjective).

2.3
CARTESIAN TENSORS
In the previous section, we saw that if (jt,) are the coordinates of a point in
one coordinate system and (x[) are the coordinates of the same point in any
other coordinate system with the same fixed origin, then xt and x- are related
by equations (2.2.6) and (2.2.7). This is equivalent to saying that if the
position vector x of a point is represented as the ordered triplet (jtf) in one
coordinate system and as the ordered triplet (*) in another coordinate
system, then xt and x\ transform to each other according to the rules given
by (2.2.6) and (2.2.7). We now show that precisely the same transformation
rules hold for the components of an arbitrary vector when it is referred to
two systems of coordinates (with the same fixed origin).
Let a be a vector having components at along the Jt, axes and components
a- along the x\ axes, so that a is represented as the ordered triplet (a,) in the
coordinate system (jcf) and the ordered triplet (a) in the coordinate system
(x). Then we have
a = alel + a2e2 + a3e3 = apep

(2.3.1)

a = a[e[ + a2e'2 + a& = ape'p

(2.3.2)

Taking the scalar product with e on both sides of (2.3.1), and noting that
a e = a-, we get
0/ = 0Lipap

(2.3.3)

40

ALGEBRA OF TENSORS

where ay are as defined by (2.2.5). Similarly, taking the scalar product with
e, on both sides of (2.3.2) and noting that a e, = ai9 we obtain
*i = <xPiap

(2.3.4)

Relations (2.3.3) and (2.3.4) are the rules that determine a- in terms of at
and vice-versa. Evidently, these relations are analogous to (2.2.6) and
(2.2.7). (Indeed, if we set a = x, relations (2.3.3) and (2.3.4) reduce to
(2.2.6) and (2.2.7), respectively.)
The foregoing analysis may be summarized thusly. With respect to a
(every) system of axes, a vector a may be represented as an ordered triplet
of real numbers; if (a,) and (a) are the representations of a vector in the xt
and x[ systems, respectively, then at and a[ obey the transformation rules
(2.3.3) and (2.3.4). This result describes a characteristic property of vectors,
and the property may be used to define a vector in an alternative way. When
defined in this "new" way, a vector is referred to as a Cartesian tensor of
order 1.
Definition A Cartesian tensor of order 1 (or a vector) is an entity that
may be represented as an ordered triplet of (real) numbers in every
Cartesian coordinate system with the property that if the ordered triplet (at)
is the representation of the entity in the JC, system and the ordered triplet (a)
is the representation of the entity in the x[ system, then at and a\ obey the
following transformation rules:
<*l = <*ip<*p\

<*i = <Xpi<tp

(2.3.5a, b)

Then a, and a\ are called the components of a first-order Cartesian tensor


in the x{ and x\ systems, respectively.
2.3.1

TENSOR PRODUCT OF VECTORS

Consider now a vector represented as the triplet (,) in the xt system and the
triplet (b[) in the x[ system. Then
b'i = aipbp;

bif = apibp

(2.3.6)

As usual, let a denote the vector for which the transformation rules
(2.3.5) hold and b denote the vector for which (2.3.6) hold.
From (2.3.5) and (2.3.6) we obtain
} = (oiipap)(ajQbQ) = 0Lipajqapbq
<*ibj = (apiupXoiqjbq) = apiaqjapbq

.0,1)

Note that [a^bj] is a 3 x 3 matrix in the system and [a-bj] is a 3 x 3


matrix in the x\ system; the elements of these two matrices transform to each

2.3

CARTESIAN TENSORS

41

other according to the rules given by (2.3.7). If we treat [#,,] and [a-bj] as
representations of a certain entity in the xt and x[ systems, respectively, then
that entity is called the outer product or tensor product of the vectors a and
b in that order, denoted by a (g) b. This product serves as an example of
what is called a Cartesian tensor of order 2.
2.3.2

CARTESIAN TENSOR OF SECOND ORDER

Definition A Cartesian tensor of order 2 is an entity that may be represented as a 3 x 3 (real) matrix in every Cartesian coordinate system with the
property that, if the matrix [a^] is the representation of the entity in the xk
system and [a'u] is the representation of the entity in the x\ system, then au
and a-j obey the following transformation rules:
a-j = aipajgapg;

} = ^^'^

(2.3.8a, b)

Then } and a'u are called the components of a second-order Cartesian


tensor in the xt and x[ systems, respectively. Note that the transformation
rules (2.3.8) are motivated by and generalizations of the rules (2.3.7).
2.3.3

CARTESIAN TENSORS OF HIGHER ORDER

Cartesian tensors of higher orders are defined by generalizing the transformation rules (2.3.8) in a natural way. The definition of a Cartesian
tensor of order N, where N is a positive integer, is given as follows.
Definition A Cartesian tensor of order N, where N is a positive integer,
is an entity that may be represented as a set of 3N real numbers in every
Cartesian coordinate system with the property that if (aijk^) is the representation of the entity in the xrsystem and (a'ijk ) is the representation
of the entity in the x[ system, then aijk . and a'ijk^ obey the following
transformation rules:
<*Uk- = a i p a j q a k r N suffixes
*</*...
TV suffixes

N factors
= <Xpi<Xqj<Xrk
N factors

<*pqr...

(2.3.9a)

N suffixes
Q'pqr...

(2.3.9b)

N suffixes

Then #(,*... and a'ijk are called the components of a Cartesian tensor of
order N in the *, and x[ systems, respectively.
We have seen that if a and b are two vectors, then the tensor product
a (x) b, represented as the matrix [,-/] in the xt system and the matrix [a-bj]
in the x- system, is an example of a Cartesian tensor of order 2. One may
show that if c is a vector with components c, and c\ in the A:, and x\ systems,

42

ALGEBRA OF TENSORS

respectively, then the entity a (x) b (x) c represented as (tf/fyc*) in the x,


system and (a-bjC^) in the x\ system is a Cartesian tensor of order 3.
Examples of Cartesian tensors of higher orders may be constructed in an
analogous way.

2.3.4

SCALAR INVARIANTS

Quite often we deal with quantities that may be represented by single


numbers not dependent on any coordinate system. Temperature, density,
mass, etc., are examples of such quantities; these are referred to as scalar
invariants. Hence, if a scalar invariant is represented as a number a in the
Xi system and a number a' in the x[ system, then a' = a. An entity obeying
a transformation law of this type is referred to as a Cartesian tensor of
order 0. Thus, a Cartesian tensor of order 0 is precisely a scalar invariant.
In the preceding paragraphs we defined Cartesian tensors of order
0,1,2,... through appropriate transformation rules. The adjective Cartesian
was used to emphasize the fact that the transformation rules pertained to
Cartesian coodinate systems. It should be borne in mind that for an entity
to be a Cartesian tensor, first of all it should have a representation as a
set of real numbers in every Cartesian coordinate system and next, the
individual numbers constituting the set should obey the appropriate transformation rules. Then, these individual numbers are called the Cartesian
tensor components in the coordinate system in which they are defined. For
a Cartesian tensor of order N (>0), there exist 3^ components in every
Cartesian coordinate system. While the tensor components are always
defined with respect to one or another coordinate system, the tensor as such
is an entity by itself; it depends on no coordinate system whatsoever.
2.3.5

TERMINOLOGY AND NOTATION

Henceforth, a Cartesian tensor will be referred to simply as a tensor.


Further, the term tensor will be used only when the order is higher than or
equal to 2; a tensor of order 1 will be referred to as a vector and a tensor of
order 0 as a scalar invariant or just a scalar.
Like vectors, tensors are also denoted by boldface letters (in print) or
letters with an arrow above them (in manuscripts). If not used with proper
care, this notation often leads to confusion, particularly when vectors and
tensors are treated together. In order to avoid such possible confusion, we
shall consistently denote a tensor by an uppercase letter with an overscript
indicating its order, in constrast to a vector always denoted by a lowercase
letter. Thus, a symbol such as a (or a) will denote a vector as usual, while
a symbol such as A [or A) will denote a tensor of order N (>2). In

2.3

CARTESIAN TENSORS

43

continuum mechanics, we deal mainly with second-order tensors, in


addition to scalars and vectors. Therefore, we relax the notation for secondorder tensors a little. We will denote a second-order tensor by an uppercase
letter without the superscript 2 (which will be automatically understood); in
other words, we will use just A (or A) for A (or A). Also, when we refer to
a tensor without explicitly indicating its order, it will be understood that the
order is 2.
If the vector defined by the transformation rules (2.3.5) is denoted as
usual by a, then we write
Mi = *,
(2.3.10)
to mean that the ith (typical) component of a in the xt system is a^ see
expression (1.4.3).
If the tensor defined by the transformation rules (2.3.8) is denoted by A,
then the matrix [au] that represents the tensor in the xt system is called the
matrix of the tensor A in the jcf system, denoted by [A]. Thus, by definition,
(2.3.11)

[A] = [au]

so that the elements of the matrix [A] are precisely the components of the
tensor A in the xt system. We write
(2312)

M = <*u

to mean that the ijth (typical) component of A in the xf system is au.


Thus, for the tensor product a (x) b of vectors a and b,
[*b]ij = aibj=[*]i[b]j

(2.3.13)

Similarly, we write
[AL*...

= *</*...

(2.3.14)
N

to mean that the (j]k...)th (typical) component of A in the xt system is aijk .


It has been pointed out that the components of a vector or a tensor are
always associated with a system of axes. However, for brevity, we will make
statements like "tf<,*... are the components of a tensor" to actually mean
that "ciijk... are the components of a tensor in the JC, system." (Similar
terminology will be used for vectors as well.)
Although tensors are denoted by capital letters, no confusion should arise
if the components of tensors are denoted by lowercase letters, like those of
vectors. For, the components of vectors appear with a single suffix, such as
ai and those of tensors appear with multiple suffixes, such as au in the case
of a second-order tensor and aijk in the case of a third-order tensor. The
number of free suffixes appended to a symbol indicates the actual order of

44

ALGEBRA OF TENSORS

the tensor being dealt with. This is indeed one of the greatest advantages of
the suffix notation in the treatment of tensors. In what follows, unless
stated to the contrary, the components of a tensor, denoted by an uppercase
letter, will be denoted by the corresponding small letter appended with an
appropriate number of suffixes, and vice-ver sa.
In the x{ system, a vector a has components

EXAMPLE 2.3.1

ax = - 1 ,

a2 = 0,

(2.3.15)

a3=l

and a tensor A has its matrix


0

[*(/] = - 1

(2.3.16)

0 -2
The x[ system is obtained by rotating the JC7 system about the x3 axis
through an angle of 45 in the sense of the righthanded screw. Find the
components of a and the matrix of A in the x\ system.
Solution For the given transformation, Table 2.4 is the table of direction
cosines (see Fig. 2.2 and Table 2.2 with = 45). If a- are the components
of the given vector a in the x[ system, the transformation rule (2.3.5a) for a
vector yields
<*i =

0L

ipap

= 11*1 + i2*2 +

i 3 * 3 = - i l + /3

using (2.3.15). Therefore, on using Table 2.4,


a[ = - a n

C*M

2 = - 2 1 + 23 =

1
V2

V2

*3 = - 3 1 + 33 = 1

Thus, the components of a in the x\ system are


ff =

V2'

(2.3.17)

= l

Table 2.4. Direction Cosines of Example 2.3.1

*'i

e2
ei

e2

e3

l/\/2
-1/V2
0

1/V2
1/V2
0

0
0
1

2.3

CARTESIAN TENSORS

45

Next, if a\j are the components of the given tensor A in the x\ system, the
transformation law (2.3.8a) for second-order tensor yields
a'

ipaJQaPQ

= oLip(oLjXapl + oLj2ap2 + 0,-30,3)


= jl(ilril + i2*21 + 13^31) + 0/2(il^l2 + i2*22 + i3*32)
+ 0/3(il^l3 + <*i2<*23 + i3*33>
= - / i a / 2 + j2/i - ^ ^ + 2a y 3 a / 2
on using (2.3.16). Therefore, on using Table 2.4, we get
# = 11 of12 + c*i2aii ~" 2ai2n + 2 1 3 1 2 = 0
^ 2 = 1,

13 = V2,

^3! = -V2,

r2i = - 1 ,

^ 2 = -V2,

a22 = 0,

#23 = V2

#33 = 0

Thus, in the x\ system, the matrix of A is

[a'ij] =

V2

-1

V2 I

-V2 -V2

(2.3.18)

EXAMPLE 2 . 3 . 2 Show that each of the two rules of transformation


given in (2.3.8) follows from the other.
Solution
yields

First, consider the rule (2.3.8a); namely, a-j = aipajgaPQ.

This

^mp Ofiq **pq

using (2.2.9);
= a
Thus, amn = oiimaJna-j. This is precisely the rule (2.3.8b).
Next, let us start with the rule (2.3.8b); namely, au = ^.
yields

Omp Onq **pq

using (2.2.8);
Thus a'mn = amianjaij.

= a'
This is precisely the rule (2.3.8a).

This

46

ALGEBRA OF TENSORS

Note: In the same way, one may show that each of the two transformation rules governing a vector and a tensor of (any) order N (>2) may
be obtained from the other. Consequently, while treating a tensor, it is
sufficient to consider only one of the two transformation rules governing
the tensor.
EXAMPLE 2.3-3 Let a and b be vectors with components at and , and
A be a tensor with components au. Show that ^ and au are scalar
invariants.
Solution If a\ and b\ are components of a and b in the x\ system, we have,
by the transformation rule of vector components,
aib'i = (aipap)(aigbg) = (aipaig)apbg = Spqapbq = apbp = a.b, (2.3.19)
Thus, Qibi has the same value in all coordinate systems; it is therefore a
scalar invariant. Note that this scalar invariant is actually the scalar product
of a and b, namely a b.
If a\j are the components of A in the x[ system, we have, by the transformation rule of tensor components of order 2, a[j = aipajgapg. Hence
a'u = oiipaigapg = Spgapg = app = au

(2.3.20)

Thus, au has the same value in all coordinate systems; it is therefore a scalar
invariant. This invariant is called the trace of A, denoted by tr A.

2.4
PROPERTIES OF TENSORS
We now proceed to obtain some basic algebraic properties of tensors.
2.4.1

ZERO TENSOR

It is obvious that if all the components of a vector are (equal to) 0 in one
coordinate system, then they are 0 in all coordinate systems. (Because of
this property, the 0 vector is often defined as the vector all of whose
components are 0 in a coordinate system.) This property is true for
tensors of all orders, and this is an important property. Proof of the
property follows here; the definition of zero tensor will follow immediately
thereafter.
PROPERTY 1 If all components of a tensor are 0 in one coordinate
system, then they are 0 in all coordinate systems.

2.4

PROPERTIES OF TENSORS

47

Proof Let au and a-j be components of a second-order tensor A in two


coordinate systems *, and x'i9 respectively. Suppose that ai} = 0. Then the
transformation rule yields
a\j = otipotjqapq = 0

(2.4.1)

This proves the property for second-order tensors. The proof for higher

order tensors is similar.


Definition A tensor of order N all of whose components are 0 in a
coordinate system is called the zero tensor of order N, denoted by 0.
2.4.2

EQUALITY OF TENSORS

If the corresponding components of two vectors a and b are equal in a


coordinate system, then the components of their difference a - b are all 0
in that coordinate system. Consequently, the components of a - b are 0 in
all coordinate systems. (Because of this property, we say that a and b are
equal, that is a = b, if the corresponding components of a and b are equal
in a coordinate system.) This property is true for tensors of all orders, and
this is another important property. Proof of the property follows here; the
definition of equality of tensors will follow immediately thereafter.
PROPERTY 2 If the corresponding components of two tensors of the
same order are equal in one coordinate system, then they are equal in all
coordinate systems.
Proof Let A and B be two second-order tensors and atj and b^ be their
components in the x{ system. Suppose a^ = Z^. If a-j and bjj are the
components of A and B in the x[ system, the transformation rule gives
**ij

(since apq = bpq)

-ip&jq**pq

&iP&jq"Pq

= b'u

(2.4.2)

This proves the property for second-order tensors. The proof for higher

order tensors is analogous.


Definition Two tensors of the same order whose corresponding components are equal in a coordinate system (and hence in all coordinate
systems, by the property just proven) are called equal tensors,
N

Note: If A and B are equal tensors, we write A = B. It is obvious that


N
N
N
N
A = B if and only if A - B is the zero tensor (of order N).

48

2.4.3

ALGEBRA OF TENSORS

TENSOR EQUATIONS

From the definition of equality of tensors it follows that if aijk


N

and bijk

are components of two tensors A and B, then the 3 equations


(243)

aijk... = hk...
are equivalent to the equation
N

A=B

(2.4.4)

Thus, in order to prove that two tensors are equal, it is sufficient to show
that their corresponding components are equal in any one of the coordinate
systems. Equations of the type (2.4.3) and (2.4.4) are called tensor
equations. It is to be emphasized that equations of the type (2.4.3) actually
imply the equality of two tensors. Equations like (2.4.4) are referred to as
tensor equations expressed in the direct notation, while those like (2.4.3) are
referred to as tensor equations in the suffix notation.
2.4.4

SCALAR MULTIPLE OF A TENSOR

If a is a scalar and a, are components of a vector a, we have seen that aat


are components of the vector aa, called the scalar multiple of a and a. This
property and notation may be extended to tensors as well, as follows.
N

PROPERTY 3 If a is a scalar and aiJk^ are components of a tensor A,


then aaijk are components of a tensor of order N. (This tensor is denoted
N

'"

by A and referred to as the scalar multiple of a and A.)


Proof Consider a second-order tensor A with components a^ and put
Cij = cLuy in all coordinate systems. This means that c-j = a'a-j in the x[
system.
Since a is a scalar, we have a' = a, so that
c[j = aa[j
= <x<XipOcjqaPQ =
= <*ip<XjqCpq

0Lipccjq(aapq)
(2.4.5)

Thus, Cij = aau obey the transformation rule of a second-order tensor;


therefore aa^ are components of a second-order tensor.
This proves the property for second-order tensors. The proof for higherorder tensors is analogous.

2.4

PROPERTIES OF TENSORS

49

Note: From the property 3, it follows that

(2.4.6)

[al]ulc_._ = [A]^...
N

The scalar multiple of - 1 and A, namely, (-l)A, is denoted by - A ,


N
referred to as the negative of A.
2.4.5

S U M AND DIFFERENCE OF TENSORS

If ai are components of a vector a and bt are components of a vector b, we


have seen that a, + bt are components of the vector sum a + b and - 6,
are components of the vector difference a - b. This property and notation
may be extended to tensors as well, as follows.
N

PROPERTY 4

Let aiJkm be components of a tensor A and biJk_ be

components of a tensor B. Then (aiJk

+ bUk ) are components of a tensor


N

of the same order N (called the tensor sum A + B). Also, (aijk^ - bijk^)
are components of a tensor of the same order (called the tensor difference
N

A - B).
Proof Consider two second-order tensors A and B with components a^
and bij9 respectively. Put c,, = a^ + by in all coordinate systems. Then,
4 = a\j + b'u
= aipajqapq + aipajqbpq = aipajq(apq + bpq)
= otipaJqcpq

(2.4.7)

Thus, Cij = au + bij obey the transformation rule of a second-order tensor;


therefore au + by are components of a second-order tensor. Similarly,
ay - bu are components of a second-order tensor.
This proves the property for second-order tensors. The proof for higher
order tensors is analogous.

Note: From the property 4, it follows that


(2.4.8)

It is easy to see that


(2.4.9)
and
(2.4.10)

50

ALGEBRA OF TENSORS

Remark: The tensor sum and tensor difference are defined for tensors of
the same order.
2.4.6

TENSOR MULTIPLICATION

If a and b are vectors with components ak and ft, (respectively), we have seen
that Qibj are components of a second-order tensor, called the tensor
product a (x) b. Such products may also be obtained by combining a vector
and a tensor or two tensors; the outcome is a tensor whose order is the sum
of the original orders. This process of obtaining new tensors from given
vectors or tensors is called tensor multiplication; the process is illustrated in
the following.
PROPERTY 5 (i) If au are components of a second-order tensor A and
bi are components of a vector b, then aubk are components of a third-order
tensor (known as the tensor product of A and b in that order, denoted
A(g)b).
(ii) If ciij and bu are components of two second-order tensors A and B,
respectively, then aikbmj are components of a fourth-order tensor (called the
tensor product of A and B in that order, denoted A B).
Proof (i) Note that ( t f ^ ) is a system of 33 = 27 numbers. Put
Uk = aijbk in all coordinate systems. Then

c'vk = aub'k = ioLipoLjqapQ){akrbr)


= aipajqakr(apqbr)

= OLipoLjqotkrcpqr

(2.4.11)

Thus, the numbers cijk obey the transformation rule of a third-order


tensor. Hence cijk = aijbk are components of a third-order tensor.
(ii) Note that (aikbmj) is a system of 34 = 81 numbers. Put cikmj =
aikbmj in all coordinate systems. Then
c'ikmj = a'ikb'mj = (otipakqapq)(amraJsbrs)
= ^ip^kq^mr^js^pqbrs

= ^ip^kq^mrOLjsCpqrs

(2.4.12)

Thus, the numbers cikmj obey the transformation rule of a fourth-order


tensor. Hence cikmj = aikbmj are components of a fourth-order tensor.

Note: From the property 5, it follows that


[A b ] ^ = aubk = [Ahjlb],
[A <g> B ] ^ , = aikbmj = [ALJB]^

(2.4.13)
(2.4.14)

2.4

2.4.7

PROPERTIES OF TENSORS

51

CONTRACTION

Let us again look at the tensor product a (x) b of two vectors a and b.
Suppose that we change the suffix j to the suffix i in the symbol a^bj
representing the components of a (x) b. Then ,-, becomes ,-,-, which
represents a scalar, namely a b. Thus, the replacement of j by i in the
symbol a^bj reduces the second-order tensor a (x) b to a scalar. Such a
process is called a contraction operation. Also, the tensor resulting from a
contraction operation is called a contraction of the original tensor. Contraction operations are applicable to tensors of all orders (higher than 1);
each such operation reduces the order of a tensor by 2. Some illustrations
follow.
PROPERTY 6 If au and bu are components of second-order tensors A
and B and c, are components of a vector c, then (i) auCj are components of
a vector, called the product of A and c in that order and denoted Ac;
(ii) aikbkj are components of a second-order tensor, called the product of A.
and B in that order and denoted AB; (iii) a^b^ is a scalar, called the scalar
product of A and B and denoted A B.
Proof (i) We note that, by property 5(i), auck are components of a thirdorder tensor and a^Cj can be obtained from a^ck by a contraction operation
(namely, changing k toy). We have to show that a^Cj are components of a
vector.
Since a^ck are components of a tensor, we have
Hence
a

UCJ

<*ip<Xjq<*jrapqCr

= 0Lipoqrapqcr

= aipapqcq

(2.4.15)

This transformation rule shows that a^Cj are indeed components of a


vector. The result illustrates that a third-order tensor reduces to a vector as
a result of a contraction operation.
(ii) We note that, by property 5(ii), aikbmj are components of a fourthorder tensor, and that aikbkj can be obtained from aikbmj by a contraction
operation (namely, changing m to k). We have to show that aikbkj are
components of a second-order tensor.
Since aikbmj are components of a fourth-order tensor, we have
a'ikb'mj =

oLipakqamroLjsapqbrs

52

ALGEBRA OF TENSORS

Hence
a'ikKj = <XiP<Xkq<Xkr<*jSaPqbrs
=

^iP^qr0LjSapqbTS
(2.4.16)

= 0Lip0Ljsapqbqs

This transformation rule shows that aikbkj are indeed components of a


second-order tensor. The result illustrates that a fourth-order tensor reduces
to a second-order tensor as a result of a contraction operation.
(iii) We note that a^by can be obtained from aikbmj by two contraction
operations (namely, replacing k by j and m by /). Whereas aikbmj are
components of a fourth-order tensor, we have to show that aubu is a scalar.
Since aikbmj are components of a fourth-order tensor, we have
a'ikb'mj =

oiipakqamrajsapqbrs

Hence
a

UbU

ipaJQair<Xjs<*Pqbrs

OprOqsQpqOrs

(2.4.17)

= apqbpq = dijbij

Thus djjbij is a scalar. This result illustrates that a fourth-order tensor


reduces to a scalar (zeroth order tensor) as a result of two contraction
operations.

Note: From the property 6, it follows that


[Ac],i

(2.4.18)

aucj = [A]u[e]j

[AB]l7 = aikbkj = [A]ik[B]kj

(2.4.19)

A B = ^

(2.4.20)

= [A]/7[B]l7

It is not hard to verify that


[AB] = [A][B],

AB * BA,

A B = B A

(2.4.21)

The product A A is called the square of the tensor A and is denoted by A 2 .


The cube A3 of A is defined by A3 = (A2)A. Higher powers of A are defined
in an analogous way. Thus,
[A2] = [aikakJl
and so forth.

[A3] = [aikakpapj]

(2.4.22)

2.4

PROPERTIES OF TENSORS

53

The following identities can be easily verified.


(i)

(aA)a = a(Aa)

(2.4.23)

(ii)

(A B)a = Aa Ba

(2.4.24)

(iii)

(AB)a = A(Ba)

(2.4.25)

(iv)

(AB)Ca = A[B(Ca)]

(2.4.26)

(v)

[A(BC)]a = A[(BC)a] = A[B(Ca)]

(2.4.27)

2.4.8

QUOTIENT LAWS

If a, and bi are components of vectors a and b we have noted that /,-, is a


scalar, namely, a b, and that if au are components of a tensor A, then a^bj
are components of a vector, namely, Ab. Partial converses of these results
are obtained in the following.
PROPERTY 7 (i) Let (a{) be an ordered triplet related to the xt system.
For an arbitrary vector with components bi9 if a^i is a scalar, then at are
components of a vector.
(ii) Let [ij] be a 3 X 3 matrix related to the Jt, system. For an arbitrary
vector with components bi9 if a^bj are components of a vector, then are
components of a tensor.
Proof (i) Since & is a scalar, we have ,-,- = a\b\. Since bt are components of a vector, we get ,, = a-(aipbp); that is,
(ap - 0Lipa[)bp = 0

(2.4.28)

Since bi are arbitrary, it follows from (2.4.28) that ap = aipa. Thus at obey
the transformation rule of a vector. Hence at are components of a vector.
(ii) Put Ci = dijbj in all coordinate systems. Then c[ = a^bj, or
aipcp = alj(ajQbg)

(2.4.29)

because bx and c, are components of vectors, by data. Multiplying (2.4.29)


throughout by air and noting that oiipair = pr and that prcp = cr = arQbq9
we obtain
(2.4.30)
(arq - cLirajqa[j)bq = 0
Since f are arbitrary, it follows from (2.4.30) that arq = airajqa\j. Thus, au
obey the transformation rule of a second-order tensor. Hence a^ are
components of a second-order tensor.

54

ALGEBRA OF TENSORS

Note: The two results proved above may be employed as tests to decide
whether a given ordered triplet/(3 x 3) matrix represents a vector/secondorder tensor. Analogous results may be established for systems of the form
(aUk), (aiJkm) and so on. Such results are all referred to as quotient laws. Some
other useful quotient laws are contained in Examples 2.4.4, 2.4.5 and 2.4.7.
EXAMPLE 2.4.1 (i) Show that are components of a second-order
tensor.
(ii) Deduce that 'u = u.
Solution (i) Since ^ = e, e,, [^] may be treated as a 3 x 3 matrix
related to the JC, system. Also, for any vector b with components 6,, we have
Sjjbj = bi9 which are components of a vector, namely, b itself. Hence, by a
quotient law (Property 7(ii)) it follows that u are components of a tensor.
(ii) Consequently,
' = CLipUjqpq = OLipOLjp = y

(2.4.31)

Note: The tensor whose components are u is referred to as the unit


tensor or the identity tensor, denoted by I. Expression (2.4.31) shows that
the components of I remain unaltered under all coordinate transformations.
In other words, the unit matrix of order 3, namely, [/] = [,,], is the matrix
of I in all Cartesian coordinate systems.
EXAMPLE 2.4.2 Prove the following identities:
(i)

la = a

(2.4.32)

(ii)

IA = AI = A

(2.4.33)

(iii)

I - A = A - I = trA

(2.4.34)

(iv)

I-a(g)b = a<g)b-I = tr(a <g) b) = a b

(2.4.35)

ek e* = I
(2.4.36)
(v)
Solution (i) Since I is a tensor and a is a vector, la is a vector, by property
6. Since the components of I are 9 expression (2.4.18) yields
[la], = ijdj = ai = [a],

This proves (2.4.32).


(ii) By property 6, we note that IA and AI are tensors. Also, expression
(2.4.19) yields
[IA]l7 = ikakj = au = [A\ij
[M\u = aikkj = au = [A\u
This proves (2.4.33).

2.4

PROPERTIES OF TENSORS

55

(iii) By property 6, we note that I A and A I are scalars. Also,


expression (2.4.20) yields
I A = = an = }} = A I
Since an = tr A (see Example 2.3.3), identity (2.4.34) is proved.
(iv) Taking A = a b in (2.4.34), we get
I (a b) = (a b) I = tr(a b)
On the other hand, by using (2.4.20), we get
I (a b) = ijidibj)

= ibi = a b

Thus, (2.4.35) is proved.


(v) We first note that
*k ek = e l e l + e 2 e 2 + e 3 e 3

Thus, ek (x) e* is a sum of tensor products of vectors. Therefore, it is a


tensor. Since [a]/ = a e, for any vector a, we have
Hence, by using expression (2.3.13), we obtain
[e* *kh = l*khl*k]j = Sklkj
This proves (2.4.36).

= u = [\]u

Prove the following identities:

EXAMPLE 2 . 4 . 3
(i)

(a <g> b)c = (b c)a

(ii)

(a b - b (g) a)c = c x (a x b)

(iii)

A(a b) = (Aa) b

(iv)

(a b)(c (g) d) = (b c)(a d)

Solution
we get

(2.4.37)
(2.4.38)
(2.4.39)
(2.4.40)

(i) Taking A = a b in expression (2.4.18) and using (2.3.13),

[(a b , . = [a b] l7 [c], = dibjcj = (b c ) ^ = [(b c)a],


This proves (2.4.37).
(ii) Consequently,
(a b - b a)c = (a b)c - (b a)c
= (b c)a - (a c)b = c x (a x b)
which is (2.4.38).

56

2 ALGEBRA OF TENSORS

(iii) Taking B = a b in (2.4.19) and using (2.3.13) and (2.4.18), we get


[A(a b)]l7 = [A],Ja b]*, = aikakbj
= [AaL[b]7=[(Aa)b]/7
This proves (2.4.39).
(iv) Taking A = a b a n d B = c d i n (2.4.19) and using (2.3.13), we get
[(a b)(c d)] l7 = [a b]lit[c d]*, = aibkckdj
= (b eyiidj = (b c)[a d] l7
This proves (2.4.40).

EXAMPLE 2 . 4 . 4 Let [a^] be a 3 x 3 matrix related to the xt system. For


two arbitrary vectors with components bt and ci9 if aubicj is a scalar, show
that jj are components of a second-order tensor.
Solution

Since a^bjCj is a scalar, we have


(2.4.41)

aubiCj = a'ub'iC}

Since , and ct are vector components, (2.4.41) yields, on noting that


aubiCj = apqbpcq,
(pq - OLipOLjqalj)bpCq

= 0

Consequently, since b{ and cf are arbitrary, we should have


Hence au are components of a second-order tensor.

EXAMPLE 2 . 4 . 5 (i) Let (aijk) be a system of 3 3 = 27 numbers related to


the Xi axes. For arbitrary vectors with components bi9 ct and dh \aijkbiCjdk
is a scalar, show that aijk are components of a third-order tensor.
(ii) Show that eijk are components of a third-order tensor.
(iii) Deduce that e'ijk = eijk.
Solution

(i) Since aijkbiCjdk is a scalar, we have


(2.4.42)

aijkbiCjdk = a'Ukblc}di

Since bi9 ct and dt are components of vectors, (2.4.42) yields, on noting that
aijkbiCjdk = apqrbpcqdr
the following expression:
(pqr - ^ipOijqOikraljk)bpCqdr

= 0

Consequently, since bi9 ct and dt are arbitrary, we should have


a

pqr

ipajqakraijk

2.4

PROPERTIES OF TENSORS

57

This transformation rule shows that aiJk are components of a third-order


tensor.
(ii) Note that (eijk) is a system of 27 numbers. Since eijk = [ef-, e,, ek]9 the
system (eijk) may be treated as one related to the *, axes. Also, for any three
vectors, b, c, d, [b, c, d] = eijkbiCjdk is a scalar representing the volume of
the parallelopiped determined by the vectors b, c, d. Hence, from the result
proven in (i), it follows that eijk are components of a third-order tensor.
(iii) Consequently,

ijk

(x

see (1.7.17);

ip0Ljq0Lkrpqr

il

<*/3

/l

0/2

/S

itl

<**2

Jt3

see (2.2.16). By use of expression (2.2.4), we now get


e'Uk = eUk

(2.4.43)

Note: The tensor whose components are eijk is referred to as the permutation tensor or the alternating tensor. The result proven in (iii) means that,
like the components of the identity tensor, the components of the permutation tensor also remain unaltered under all coordinate transformations.
EXAMPLE 2.4.6 By using a transformation law for a vector, show that
the vector product a x b of two vectors a and b is a vector.
Solution

We recall that
[a x b] = eUkajbk

and set c, = ijkajbk in all coordinate systems. Then


c

ijka'jbk

By using the fact that eijk are components of a tensor and ai9 bt are
components of vectors, this becomes
c'i =

(oiipaJQakrepQr)(ajmam)(aknbn)

Oiip Oqm Orn Gpqr am On opmn u m On

This shows that cf = eijkajbk are components of a vector; in other words


a x b is a vector.

58

2 ALGEBRA OF TENSORS

EXAMPLE 2.4.7 (i) If cljkm and Qjj are components of tensors of orders
4 and 2, respectively, show that cijkmakm are components of a second-order
tensor.
(ii) Let (cijkm) be a system of 81 numbers related to the xt axes. For an
arbitrary second-order tensor with components a^ if cijkmakm are components of a second-order tensor, show that cijkm are components of a
fourth-order tensor.
Solution (i) Put du = cijkmakm in all coordinate systems. Then d-j =
c-jkmakm. Since cijkm and } are tensor components, this relation yields
dlj = (oiipOijqakramscpgrs)(akhamnahn)
=

ipajqrhSnCpqrsahn

Oijp(XjqCpqrs(lrs

OLipOLjqUpq

This transformation rule shows that d^ = cijkmakm are components of a


second-order tensor.
(ii) Put dtj = cijkmakm in all coordinate systems so that d-j = c'ijkmakm as
previously. But, now, ^ and dtj are tensor components. As such,
ocipOLjqdpq = c[jkmotkramsars

(2.4.44)

Multiplying (2.4.44) throughout by aihajn and noting that aihaip = hp,


jnajQ = nq and hpnqdPq = dhn, we get

(2.4.45)

dhn = 0Lihajnakr0Lmsc[jkmars
Since dhn = chnrsars, we get from (2.4.45),
(Chnrs - oiihajnakramscljkm)ars

=0

(2.4.46)

Since ars are arbitrary, it follows from (2.4.46) that


Chnrs

^ih^jn^kr^ms^ijkm

This transformation rule shows that cijkm are indeed components of a


fourth-order tensor.

Note: The result proved in (ii) is a quotient law for a fourth-order tensor.

2,5
ISOTROPIC TENSORS
A tensor whose components remain unchanged under all coordinate
transformations is called an Isotropie tensor. In other words, a tensor

2.5

A having components aijk


isotropic if and only if

ISOTROPIC TENSORS

in the ,- system and a'ijk

59

in x[ system is

(2.5.1)
= a,ijk..
for all choices of the xt and x\ systems.
We have seen that the identity tensor and the permutation tensor possess
this property; these are therefore examples of isotropic tensors. The
definition of a scalar itself reveals that a scalar is an isotropic tensor. The
geometrical meaning of components of a nonzero vector suggests that the
components do change under coordinate transformations; as such, a
nonzero vector cannot be an isotropic tensor. On the other hand, zero
tensors of all orders are isotropic.
It is easy to verify that a scalar multiple of an isotropic tensor is an
isotropic tensor and that the sum and the difference of two isotropic tensors
is an isotropic tensor. Since the identity tensor I is isotropic, it follows that
every scalar multiple of I is an isotropic tensor of order 2. The converse of
this result, namely, that every second-order isotropic tensor is a scalar
multiple of I, is also true. The following theorem establishes this fact.
Mjk..

THEOREM 2.5.1 If } are components of an isotropic tensor (of second


order), then au = ^ for some scalar a.
Proof Since the given tensor is isotropic, we have a[j = a^ for all choices
of the x[ system.
We first consider the coordinate transformation defined by
ei = e 2 ,

e2 = e 3 ,

e3 = e!

(2.5.2)

all other a^ = 0

(2.5.3)

For this transformation we have


12 = 23 = or31 = 1;
see Figure 2.3 and Table 2.5.
I x'7

y.
Figure 2.3. Theorem 2.5.1.

60

ALGEBRA OF TENSORS
Table 2.5. Theorem 2.5.1

5
i
3

e2

e3

0
0
1

1
0
0

0
1
0

Using (2.5.3), the transformation rule of a second-order tensor together


with the fact that a-j = } gives
au = a'n = alpalQaPQ

= a22

(2.5.4)

Similarly, we find that


#22 = #33 = #n

(2.5.5)

#i2 = #23 = a3l;

a2l = a32 = #13

(2.5.6)

Next, consider the coordinate transformation defined by


e

'l = e2>

2 =

_e

3 = e3

(2.5.7)

This transformation corresponds to the rotation of the JC, axes about the x3
axis through an angle of 90. Hence
12 = 33 = 1

2i = -U

all other al7 = 0

(2.5.8)

see Figure 2.2 and Table 2.2 with = 90.


Using (2.5.8), the transformation rule of a second-order tensor together
with the fact that a\j = atj gives
#13 = #13 = <X\p<Xlq<lpq = #23

#23 = #23 = oc2poi3qapq = -an

(2.5.9)

(2.5.10)

These equations show that


#13 = #23 = 0

(2.5.11)

#12 = #21 = #31 = #32 = 0

(2.5.12)

Equations (2.5.6) now yield

Expressions (2.5.11) and (2.5.12) show that all the nondiagonal terms of
the matrix [au] are 0. Also, the diagonal terms are equal, by (2.5.5). If we
set these diagonal terms equal to a, say, then we have ^ = ^. This
completes the proof.

2.6 ISOTROPIC TENSOR OF ORDER 4

61

Note: Since the permutation tensor is isotropic, every scalar multiple of


this tensor is an isotropic tensor of order 3. The converse of this result,
namely, that if aijk are components of an isotropic tensor of third order,
then aijk = aeijk for some scalar a is also true. The proof is analogous to
that of Theorem 2.5.1.

2.6

ISOTROPIC TENSOR OF ORDER 4


Since u are components of an isotropic tensor, it is easy to verify that the
fourth-order tensors whose components are ijkm, ikjm and imjk are
also isotropic. Consequently, it follows that if
<*ijkm = <xuOkm + ikOjm

yimjk

where , , y are scalars, then aijkm are components of an isotropic tensor.


The converse of this result is also true; it is proved in the following theorem.
THEOREM 2.6-1
tensor, then

If aijkm are components of a fourth-order isotropic

<*km = <xijkm + oikjm

for some scalars

yimjk

(2.6.1)

,,.

Proof If the given tensor is the zero tensor, the proof is trivial. We
therefore suppose that at least 1 of the 81 components aijkm is not 0.
Since the proof is lengthy, we will give it in four stages. In the first stage,
we will classify the 81 components of the tensor into five different classes
(classes I to V) depending upon the nature of the suffixes present in the
components. In the second stage of the proof, we will consider some
particular coordinate transformations and show that the components
belonging to each of the classes I-IV are equal among themselves. In the
third stage, we will show, again by considering some particular transformations, that all the components belonging to class V are identically 0. In the
last stage of the proof, we deduce the representation (2.6.1). The arguments
throughout essentially depend on the fact that a'ijkm = aijkm for all coordinate transformations, whatsoever.
First Stage of the Proof, The 81 components aijkm may be classified into
the following five classes:
Class I.

Components in which all suffixes are the same: flnn ^2222

3333

62

2 ALGEBRA OF TENSORS

Class II. Components in which the first two suffixes are the same and
the last two suffixes are the same, second and third suffixes
being different: aU22, #2233 03311 tf22ii> 03322 01133
Class III. Components in which the first and the third suffixes are the
same, and the second and the fourth suffixes are the same,
first and second suffixes being different: tf12i2, #2323 03131
#2121 #3232 01313

Class IV. Components in which the first and the fourth suffixes are the
same, and the second and third suffixes are same, first and
the second suffixes being different: tf1221, tf2332, a 3113 , #2112
#3223 0 1 3 3 1

Class V. The remaining 60 components (in which one of the suffixes is


different from all the other three).
Second Stage of the Proof After this classification, we now proceed to
consider some particular transformations. The first of these transformations is
e
e
e
l = e2>
2 = e3>
3 =el
(2.6.2)
For this transformation we have, from (2.5.3),
12 = 23 = 31 = 1;

other au = 0

(2.6.3)

The transformation rule for a fourth-order tensor together with the fact
that a[jkrn = aijkm yields, on using (2.6.3), the following
1111 = tfllll = <*\p<X\q<*\r<*U<*pqrS = #2222

()

(2.6.4)

Similarly,
*2222 = 03333
()

01122 = 01122 = oLlpalqalra2sapqrs

(2.6.5)
= a2233

(2.6.6)

Similarly,
02233 = 03311
(")

02211 = 02211 = ^2pOLlqalrauapqrs

(2.6.7)
= a3322

(2.6.8)

Similarly,
03322 = 01133
1"V)

01212 = 01212 = OLlpOL2qauOL2sapqrs

(2.6.9)
= ?2323

(2.6.10)

Similarly,
02323 = 03131
M

a2l2l

= 72i2i = <X2p<XiqOi2raisapgrs

(2.6.11)
= a3232

(2.6.12)

Similarly,
'3232

= 01313

(2.6.13)

2.6 ISOTROPIC TENSOR OF ORDER 4


(Vi)

0i
O a a
'1221 = 01221 = lp 2q 2r ls pqrs

2332

63

(2.6.14)

Similarly,
=

#2332
< V ")

(2.6.15)

#3113

^2112 = #2112 = <X2p<X\q<X\r<X2s<*pqrs = 3223

(2.6.16)

Similarly,
=

#3223

tf

1331

(2.6.17)

The next coordinate transformation which we consider is


e

l e 3>

2>

3 ~~

(2.6.18)

For this transformation, we have (see Figure 2.4 and Table 2.6)
-1

13 = 22 = 31 =

other oLij = 0

>

(2.6.19)
Using (2.6.19), the transformation rule for a fourth-order tensor together
with the fact that aijkm = a\jkm yields
=

#1122 = 0122 = l p l ? 2 r 2 s W s

3322

(2.6.20)

Similarly,
#1212 = #3232

(2.6.21)

#1221 = #3223

(2.6.22)

Relations (2.6.4) and (2.6.5) show that all the components belonging to
class I are equal; that is,
tfiiii

= #2222 = 3333 = , s a y
A *3

>^

I
I
I

Figure 2.4. Theorem 2.6.1, Stage 2.


Table 2.6. Theorem 2.6.1, Stage 2

e'i

e>

ei

e3

0
0
-1

0
-1
0

-1
0
0

(2.6.23)

64

2 ALGEBRA OF TENSORS

Relations (2.6.6) to (2.6.9) and (2.6.20) show that all the components
belonging to class II are equal; that is,

S a v

(2.6.24)
= a2233 #3311 = 2211 = #3322 = 01133 = <*>
Relations (2.6.11) to (2.6.13) and (2.6.21) show that all the components
belonging to class III are equal; that is,
'1122

#1212 = 02323 = #3131 = #3232 = #1313 = #2121 =

S a v

(2.6.25)

Relations (2.6.14) to (2.6.17) and (2.6.22) show that all the components
belonging to class IV are equal; that is,
01221 = 02332 = 03113 = ^2112 = 03223 = 1331 = 7

Sa

(2.6.26)

From their definitions it is evident that //, a, and y have the same values
in all coordinate systems.
Third Stage of the Proof. We now attend to the components belonging to
class V, in which one of the suffixes is different from the other three. We
first take up the components of the form aUJk or anjk or aiJlk or aijkl, where
/, j , k are different from 1. Let us consider the transformation defined by
e'i = if

e2 = - e 2 ,

e^ = - e 3

(2.6.27)

For this transformation, we have (see Figure 2.5 and Table 2.7)
<*n = <*33 = 1,

22 = - 1 ;

other OLU = 0

, * 3

0
I
I

'

Figure 2.5. Theorem 2.6.1, Stage 3.


Table 2.7. Theorem 2.6.1, Stage 3

(2.6.28)

2.6 ISOTROPIC TENSOR OF ORDER 4

65

On using (2.6.28), the transformation rule for a fourth-order tensor and


the fact that a'ijkm = aijkm yield
a

\ijk

= a'lijk =

^Xp^iq^jr^ks^pqrs

= oiiqajraksalqrs

for i,j\ k * 1

= -aWk,

Thus
<*w = 0,

for ij, k * 1

(2.6.29)

Similarly,
Oiijk = <*ij\k = <iijk\ = 0,

for i,j, k * 1

(2.6.30)

By considering the transformations


e'i = - e j ,

e2 = e 2 ,

e[ = -ei9

e2 = - e 2 ,

e3 = - e 3

(2.6.31)

e 3 = e3

(2.6.32)

we may show that


(tujk = Qiijk = <*ij2k = a-ijki = 0,

for ij, k*2

<*Wk = oi3jk = aijU = aim = 0,

for ij9 k * 3

(2.6.33)
(2.6.34)

Thus, all components belonging to class V are identically 0. Consequently, at least one of the components belonging to classes I to IV is
nonzero; that is, at least one of , a, /?, y defined by (2.6.23) through
(2.6.26) should be nonzero.
Fourth Stage of the Proof. We now introduce the numbers cijkm defined as
Cijkm = <*ijkm - (<xijkm + ikjm

+ yimjk)

(2.6.35)

Obviously, cijkm form the components of a fourth-order isotropic tensor.


Using relations (2.6.23) through (2.6.26) and the fact that all aijkm belonging
to class V are identically 0, it is easy to verify from (2.6.35) that
other cijkm = 0

C\u\ = C2222 = C3333 = n - (<* + + 7)\

(2.6.36)

Finally, we consider the transformation given by


el = 7 j ( e i

+ e

2>>

2 = ^ | ( e 2 - e0,

e3 = e3

(2.6.37)

This transformation corresponds to the rotation of the jcf axes about the x3
axis through an angle of 45, and we have (see Figure 2.2 and Table 2.2
with = 45)

<*2 =

*2

22

7^>

other a y = 0

(2.6.38)

66

2 ALGEBRA OF TENSORS

The transformation rule for a fourth-order tensor and the fact that
c

ijkm cijkm yield


C

llll

llll

lpalq0ilralsCpqrs

= ( + lKllll

by using (2.6.36);

Tcnii

by using (2.6.38). Therefore c l l n = 0. Consequently, (2.6.36) yields


C2222 = c3333 = 0. Thus all the 81 numbers ciJkm defined by (2.6.35) vanish
identically, and we get the representation (2.6.1). The proof of the theorem

is complete.
It may be mentioned that the representation (2.6.1) is useful in constitutive modeling; see, for example, sections 9.2 and 10.10.
EXAMPLE 2.6.1 The components a^ and by of two tensors A and B are
related through the expressions a^ = cijkmbkm. If cijkm are components of
an isotropic tensor of order 4 and by = b, show that
au = kbkkSu + 2bu

(2.6.39)

for some scalar s and .


Solution If cijkm are components of an isotropic tensor of order 4,
ij = cijkmbkm becomes, on using (2.6.1)

Ou = (aijkm + ikjm + ySimjk)bkm


= aukk + bu + ybji

(2.6.40)

If by = b, (2.6.40) reduces to (2.6.39) on setting a = and \( + ) = .

m
2-7
TENSORS AS LINEAR OPERATORS
If ay are components of a second-order tensor A and ct are components of
a vector c, it has been shown that a^Cj are components of a vector, denoted
by Ac (see Section 2.4, property 6). Suppose we redesignate this vector as d;
that is we set
Ac = d
(2.7.1)
For a given A, the equation (2.7.1) defines a transformation that transforms the vector c to the vector d, with A effecting the transformation

2.7

TENSORS AS LINEAR OPERATORS

67

operation. The tensor A may therefore be thought of as an operator that


acts on a vector c and generates a vector d = Ac.
For example, the identity (2.4.32), namely la = a, shows that the tensor
I operating on a generates a itself. This explains why I is called the identity
tensor.
We now prove a theorem that demonstrates a characteristic property of
second order tensors.
THEOREM 2.7.1 If A is a second-order tensor, then it is a linear
operator on vectors and its components are given by
au = e, Ae,

(2.7.2)

Conversely, if A is a linear operator on vectors and au are defined by


(2.7.2), then au are components of a second-order tensor.
Proof First suppose that A is a second-order tensor with components a^.
Consider any two scalars a, and any two vectors b and c. Then Ab, Ac
and A(ab + c) are vectors. Also,
[A(o* + c)]j = a^ab + c]j = ijiabj + cj)
= a(aubj) + (aucj) = [Ab], + /?[Ac]f
so that

= [(Ab) + (Ac)L
A(ab + c) = cx(Ab) + (Ac)

(2.7.3)

This expression shows that A is a linear operator on vectors.


Further,
Oijbj = [Ab], = e, Ab

by property (2.7.3);

= e, A(bjej) = e, ^(Ae,)
= (ef Ae,)*/

(274)

Since this is true for any vector b, expression (2.7.2) is proven.


Conversely, suppose A is a linear operator on vectors and atj are defined
by (2.7.2). From (2.7.2) we note that au are defined with respect to the x,
system and are uniquely determined by A with respect to this system. Let a-j
be the corresponding numbers determined by A and defined with respect to
the X- system; that is,
a'u = *'r Ae;

(2.7.5)

68

ALGEBRA OF TENSORS

Since e{ = (ej tp)tp = aipep, (2.7.5) yields


a[j = (aipep) H<*jQeQ)
= (a/Pep) (cLjqAcq)
because A is a linear operator;
= ip^(ep *

Ae

g)

<*ipajqapq

by (2.7.2). This transformation rule shows that au are components of a


second-order tensor. The proof of the theorem is complete.

By virtue of this theorem, the terms linear operator on a vector and


second-order tensor are treated as synonyms.
EXAMPLE 2.7.1 A transformation A transforms every vector into a
fixed vector. Show that A is not a tensor.
Solution Let a be the vector into which all vectors are transformed under
A. Then, for any scalar a and any vector c, we have Ac = a and A(ac) = a,
so that A(ac) = Ac. Since a is arbitrary, Ac 5* ocAc, in general. Hence
A(ac) 5* OLAC, in general. As such, A is not a linear operator and hence not
a tensor.

EXAMPLE 2.7.2 Show that two tensors A and B are equal if and only
if (i) Ae^ = Be*, or (ii) Aa = Ba for any vector a.
Solution
note that

(i) Let ay and by be components of A and B, respectively. We


[Ae*], = aij[ek]j = ^ e,) = = aik

Similarly, [BeJ, = bik. Hence the condition aik = bik holds if and only if
[Aek]i = [Be*];. That is, A = B if and only if Ae*. = Be^.
(ii) Let a1 be components of a. Then, since A and B are linear operators,
we have
Aa = A(ur*e*) = ak(Aek)
Ba = B(akek) = ak(Bek)
Suppose that Aa = Ba. Then, ak(Aek) = ak(Bek). Since a is arbitrary, it
follows that Ae*. = Be*. Hence A = B by what has been proven already.
The converse is trivially true.

2.8 TRANSPOSE OF A TENSOR

69

EXAMPLE 2-7.3 Show that every tensor A with components can be


represented in the form
A = autt e,
Solution

(2.7.6)

Let c be any vector with components cf, Then


Ac = A(cjej) = Cj(Aej)

because A is a linear operator;


= c,(e, Aeyte
because any vector a = (e, a)e,;
using (2.7.2);

= <*ucJ*i

= <tijCkkjei = aijCk(ek ey)ef


by (2.4.37);

= auck(*i <g> tj)ek


= ^ ( e , <g> e,)c

Since c is an arbitrary vector, (2.7.7) yields (2.7.6).

(2.7.7)

2.8
TRANSPOSE OF A TENSOR
In the remaining part of this chapter we consider second-order tensors
only. Given a 3 x 3 matrix [a^], the transpose of [a^] has been defined by
[aij]T = [jil If [tf(/] is the matrix of a tensor A, a natural question that
arises is, is [ay]Talso a matrix of a tensor? The answer is in the affirmative;
the justification is as follows.
Suppose that au are tensor components. Put [^\ = [by] in all coordinate systems so that bu = and b-j = . Then
This transformation rule shows that b^ are components of a tensor.
Equivalently, [by] = [] = [tf//]ris the matrix of a tensor. This new tensor
is called the transpose of the tensor A, denoted A r . Thus, if A is a tensor
with matrix [au], then A r , the transpose of A, is the tensor with matrix
[aij]T; that is,
(2.8.1)
[AT] = [A]T

70

ALGEBRA OF TENSORS

The following relations, which hold for any number a and any tensors A
and B, may be verified:
(i)

(AT)T = A

(ii)

(c*A) = OLK

(iii)

(A B) r = AT BT

(iv)

(2.8.2)
T

() =

(2.8.3)

(2.8.4)
(2.8.5)

If Vi are components of a vector v, it is easy to verify that


[ATv]i = avj = [A][v]j
2.8.1

(2.8.6)

INVERSE OF A TENSOR AND THE


ORTHOGONAL TENSOR

Recall that a matrix [] is said to be invertible (or nonsingular) if there is


a matrix [tf/,] -1 such that
[au\[auYl = M" 1 !*/] = m

(2.8.7)

Then [ij]' is called the inverse of [au]. Further, [au] is said to be


orthogonal if [au] is invertible and [tf/,] -1 = [^\. These definitions can be
extended to tensors as follows.
A tensor A is said to be an invertible tensor if there exists a tensor, which
we denote by A"1, such that
AA"1 = A _ 1 A = I

(2.8.8)

Then A - 1 is called the inverse of A. Further, A is said be an orthogonal


tensor if A is invertible and A"l = AT. Thus A is an orthogonal tensor if
and only if
AA r = A r A = I
(2.8.9)
Conditions (2.8.8) and (2.8.9) show that a tensor A is invertible if and
only if its matrix [A] is invertible with [A] - 1 = [A -1 ] and that A is orthogonal if an only if [A] is orthogonal.
It is easy to verify that the following relations are true for any invertible
tensors A and B:
(i)
(ii)
(iii)

(A"1)"1=A

(2.8.10)
r

(A-y=(A r
_1

A (Aa) = a;

(2.8.11)
_1

A(A a) = a

(2.8.12)

for any vector a;


(iv)

(AB)"1 = B^A" 1

(2.8.13)

2.8

TRANSPOSE OF A TENSOR

71

Let A be a second-order tensor. For any two vectors a

EXAMPLE 2.8-1
and b, show that

a Ab = b A r a

(2.8.14)

Solution Let ai9 bt and a y be components of a, b and A respectively.


Then c = Ab is a vector whose components are c, = a^bj, and d = A r a is
a vector whose components are rff = . Therefore
a Ab = a c = a^c,
= (tiQijbj =

ajjibi

(interchanging the dummy suffixes / and y);


= bidi = b d = b A r a

Prove that for any two vectors a and b,

EXAMPLE 2 . 8 . 2

(ab)r = b a
Solution

(2.8.15)

We have
[(a b) r ] l 7 = [a b] = ajbi = b^j = [b <g> a],,

which yields (2.8.15).

EXAMPLE 2 . 8 . 3 Show that a tensor Q is orthogonal if and only if


Qa Qb = a b for all vectors a and b. Deduce that if Q is an orthogonal
tensor, then |Qa| = |a| for all vectors a.
Solution

With the aid of (2.8.14), we note that

(Qa Qb) - (a b) = (a QrQb) - (a b) = a (Q r Q - I)b


It now follows that Qa Qb = a b for arbitrary a and b if and only if
Q r Q = I or, equivalently, Q is orthogonal.
If Q is an orthogonal tensor, then, by the result just proven, we have
Qa Qa = a a; or equivalently, |Qa| = |a| for every vector a.

EXAMPLE 2 . 8 . 4
that

Let A be a tensor. For any vectors, u, a, b and n, show


[(u a)(A r b)] n = a [(u An)b]

(2.8.16)

Solution Let ut, at, bt and , be components of u, a, b and n, respectively,


and au be components of A. Then we get
[(u a)(A r b)] n = &&)(<*jkbj)nk =

a^a^n^bj

= [a],[u An] /y [b], = a [(u An)b]

72

ALGEBRA OF TENSORS

EXAMPLE 2 . 8 . 5 If au are components of a tensor A and bt are components of a vector b, show that eirsbraJS are components of a tensor. If this
tensor is denoted by b A, show that, for any vector a,
(b A)a = b x (A r a)

(i)

(b A) a = A(a x b)

(ii)
Solution

(2.8.17)
(2.8.18)

(i) Let c^ = sirsbrajs in all coordinate systems. Then

This transformation rule shows that cu = eirsbrajs are components of a


tensor. If this tensor is denoted by b A, we have
[b A]ij =

eirsbrajs

(2.8.19)

Consequently,
[(b A)a], = eirsbrajsaj = f irs [b] r [A r a] s = [b x A^a],
from which (2.8.17) follows.
(ii) Also,

from which (2.8.18) follows


EXAMPLE 2 . 8 . 6 Let [au] be a matrix related to xt system. For an
arbitrary vector with components bi9 if Obj are components of a vector,
show that ay are components of a tensor.
Solution Let cu = an. Then [cu] = [^ is a matrix related to the x{
system. By data, c^bj = a^bj are components of a vector. Hence, by a
quotient law (see Section 2.4, property 7(ii)), cu are components of a tensor.
Consequently, c = ^ are also components of a tensor.

2.9
SYMMETRIC AND SKEW TENSORS
If A is a (second-order) tensor, we have seen that the transpose of A,
namely, A r , is a tensor whose matrix is [A] r . If A is such that AT = A,
we say that A is a symmetric tensor; if A r = - A , we say that A is a

2.9

SYMMETRIC AND SKEW TENSORS

73

skew-symmetric tensor or just skew tensor. It follows that A is symmetric


or skew accordingly as the matrix [A] is symmetric or skew-symmetric.
Equivalently, if au are the components of A, then A is symmetric or skew
accordingly as a = au or = -}. It is easily seen that the identity tensor
is a symmetric tensor and the tensor considered in Example 2.3.1 is a skew
tensor. It is obvious that only six of the nine components of a symmetric
tensor are independent. In the case of a skew tensor, three components
represented as diagonal elements of the matrix of the tensor are 0 and only
three of the remaining six components are independent.
For any tensor A, we have from (2.8.2) and (2.8.4)
(A + AT)T = AT + A = A + A r

(2.9.1)

(A - AT)T = AT - A = - ( A - A 7 )

(2.9.2)

Accordingly, the tensor A + A is always symmetric and the tensor A - A r


is always skew.
Also, we note that
T

A = | ( A + A r ) + ^(A - A r )

(2.9.3)

for any tensor A. Evidently, the first term in the righthand side of (2.9.3) is
a symmetric tensor and the second term is a skew tensor. Thus, every tensor
A may be represented as a sum of a symmetric tensor and a skew tensor. It
can be proven that such a representation is unique. The tensor y (A + A r )
is referred to as the symmetric part of A and denoted sym A. The tensor
y (A - A r ) is referred to as the skew part of A and denoted skw A. Thus,
for every tensor A,
sym A = | ( A + AT)

(2.9.4)

(2.9.5)

skw A = ^(A - A )
A = sym A + skw A

(2.9.6)

If au are components of A, it is obvious that \(au + ) are components


of sym A and \{^ - ) are components of skw A.
EXAMPLE 2.9-1 If A is a symmetric tensor with components and B
is a skew tensor with components bij9 then show that
A B = aubu = 0

(2.9.7)

Solution In the expression ify-o^, both the suffixes / and y are dummies.
As such, aubu = ab. Since A is symmetric and B is skew, we have
Qji = au and b = -bu. Hence
a

UbU = ab = aiA~bij)

= ~aUbU

74

ALGEBRA OF TENSORS

so that ijbij = 0. Since A B = aubu,


proven.

If A is a symmetric tensor and B is any tensor, show

EXAMPLE 2 . 9 . 2
that

Solution

see (2.4.20), the result (2.9.7) is

A B = A-(symB)

(2.9.8)

A B = A (sym B + skw B) = A sym B + A skw B

(2.9.9)

We have

Since the tensor skw B is skew, and the tensor A is symmetric by data, we have
A skw B = 0 by Example 2.9.1, and we get (2.9.8) from (2.9.9).

2.10

DUAL VECTOR OF A SKEW TENSOR


One of the most important properties of a skew tensor is the existence of
what is called the dual vector of the tensor. The property may be stated in
the form of the following theorem.
THEOREM 2 . 1 0 . 1
such that

Given a skew tensor A, there exists a unique vector


Au = x u

(2.10.1)

for every vector u. Conversely, given a vector , there exists a unique skew
tensor A such that (2.10.1) holds for every vector u. (The vector is called
the dual vector or the axial vector of the tensor A.)
Proof
write

Let [A],, = a^. Since A is skew, we have a^ = aJi9 and we may


a

U = i(au " a) = 2(ipJq -

jpiq)apq

= \ijkkPqaPq

(2.10.2)

on using the substituting property of ij and the - identity.


Let
<k = -\ZkPqapq

(2.10.3)

Then k are components of a vector, say, , and this vector is determined


uniquely by A. Substituting for k from (2.10.3) in (2.10.2), we get
au = -eijkcok

(2.10.4)

2.10 DUAL VECTOR OF A SKEW TENSOR

75

For any vector u with components ui9 (2.10.4) yields


(2.10.5)
which is precisely (2.10.1) in the direct notation.
Thus, given a skew tensor A, there exists a unique vector , whose
components are defined by (2.10.3), such that (2.10.1) holds for an
arbitrary vector u. The existence as well as the uniqueness of the dual vector
of A has thus been established.
Conversely, given a vector <o, suppose we define a^ by (2.10.4). Then, au
are components of a skew tensor, say, A, and this tensor is determined
uniquely by <o. Since (2.10.4) yields (2.10.1), it follows that A is the tensor
of which is the dual vector. The proof of the theorem is complete.

Note: Expanding the righthand side of (2.10.3) with the aid of the
definition of the -symbol, we get
= - ( t f 2 3 e l + 031 e 2 + #12 e 3)

(2.10.6)

Similarly, we find from (2.10.4),


0 3
[A] = [au\ =

3
- 2

0 -,

(2.10.7)

Expression (2.10.6) is used to find when A is known, and (2.10.7) is


used to find the matrix of A when is known.
EXAMPLE 2.10.1
whose matrix is

Find the dual vector of the skew part of the tensor A


"1-2
[A] =

0'
2

0 -1
Solution The transpose of the given matrix is
1 0
r

f A ] = ' -2 1
0 2

76

ALGEBRA OF TENSORS

Hence
"0 - 1
r

[skwA]=i([A]-[A] ) =

-1

1 -1

By using (2.10.6), we get the dual vector of skw A as


= -e,
EXAMPLE 2 . 1 0 . 2

e> + e,

If is the dual vector of a skew tensor A, show that


||=^||

(2.10.8)

where || is, as usual, the magnitude of , and |A| = VA A.


Solution If ,- are components of and au are components of A, we get
from expression (2.10.3),

k k

4kpqekrsapgars

which, on using the - identity, becomes


u>k<k = }(Pr<>s ~ 4\QpqQpq

psqr)apqars
&pq@qp)

==

l^pq^pq

(2.10.9)

When written in the direct notation, (2.10.9) reads

||2 = +

(2.10.10)

from which (2.10.8) is immediate.


EXAMPLE 2 . 1 0 . 3 Let Q be an orthogonal tensor and a be a vector such
that Qa = a. Show that (i) Q r a = a and (ii) the dual vector of skw Q is
collinear with a.
Solution

(i) By data, a = Qa. Hence


Q 7 * = Qr(Qa) = (QrQ)a

But, since Q is an orthogonal tensor, Q r Q = I. Also, la = a. Therefore,


Qra = a

(2.10.11)

(ii) Since Qa = Q r a = a, we have


(skwQ)a = i ( Q - Q r )a = (Qa " Q 7 ) =

(2.10.12)

2.11

INVARIANTS OF A TENSOR

77

On the other hand, if is the dual vector of skw Q, we have


(skwQ)a = x a

(2.10.13)

From (2.10.12) and (2.10.13), we get x a = 0; that is, is collinear


with a.

2.11
INVARIANTS OF A TENSOR
Given a tensor A with components aU9 we have noted in Example 2.3.3 that
aa (which is the contraction of a^) is a scalar invariant called the trace of A
and denoted tr A. Besides this invariant, two other invariants associated
with A are encountered. Keeping this in view, tr A is often referred to as the
first invariant of A and denoted also by 7 A . Thus,
(2.11.1)

IA = trA = au

The second and the third invariants of A are introduced in the following.
We have noted that the square of a tensor A, defined by A 2 = AA, is also
a tensor and that the components of A2 are aikakj\ see (2.4.22). Hence tr A 2 ,
given by
txA2 = aikaki

(2.11.2)

is a scalar. Consequently, y[(tr A) 2 - tr A2] is also a scalar. This scalar is


called the second invariant of A and denoted IIA. Thus,
7/A = i [ ( t r A ) 2 - t r A 2 ]

(2.11.3)

When written in the suffix notation, this expression reads


(2.11.4)

HK = \[au<tkk - <*ik<*ki]

Writing the righthand side of (2.11.4) in the unabridged form we get


tfn

a12

a2l

a22

*11

013

031

033

022

023

032

033

(2.11.5)

Some authors take j[A A - (trA) 2 ] as the second invariant of A,


denoted 7/ A . Thus,
/7A = i [ A - A - ( t r A ) 2 ]
(2.11.6)
or, in the suffix notation,
//A = WijOij - ouajj]
For a symmetric tensor A, HA =

-IlA.

(2.11.7)

78

ALGEBRA OF TENSORS

It has been noted that the cube of a tensor A, defined by A3 = A 2 A, is


also a tensor and its components are aikakmamJ; see (2.4.22). Hence tr A 3 ,
given by
trA 3 = aikakmami

(2.11.8)

is a scalar. Consequently, it follows that


[(trA) 3 + 2trA 3 - 3(trA 2 )(trA)]
is also a scalar. This scalar is called the third invariant of A and denoted
/ / / A . Thus,
IIIA = [(tr A) 3 + 2 tr A3 - 3(tr A2)(tr A)]

(2.11.9)

When written in the suffix notation, this expression reads


IIIA = i[auajjakk + 2aikakmami - 3aikakiajj]
(2.11.10)
The three invariants IA, IIA and IIIA are usually referred to as the fundamental or principal invariants of A. It is easy to verify that A and A r have
the same principal invariants. The significance of these invariants will
become clear in Section 2.13; see equation (2.13.2) in particular.
2.11.1

DETERMINANT OF A TENSOR

Using expression (1.7.27), we note that the righthand side of (2.11.10)


represents det(r^) Consequently, by use of (1.7.18), we obtain the following simpler representation for IIIA :
IIIA = detfo,] = eUkanaj2ak3

= eUkaua2ja3k

(2.11.11)

This expression shows that, although the individual tensor components


ay change from one set of axes to the other, the determinant of [au] as a
whole retains the same value IIIA in all sets of axes. The invariant IIIA is
(therefore) referred to as the determinant of the tensor A and denoted
det A. Thus, the determinant of A is nothing but the determinant of the
matrix [A] in any of the coordinate systems.
It has been noted that A is invertible if and only if [A] = [0] is invertible
and that [au] is invertible if and only if detfo,] ^ 0. Hence, A is an
invertible tensor if and only if det A ^ 0.
From the properties of matrices and determinants, it readily follows that
(i)

det I = 1

(ii)

det A = det AT

(iii)

det(aA) = a 3 det A

(iv)

det(AB) = (det A)(det B)

for any tensors A and B and any number a.

(2.11.12)

2.11

INVARIANTS OF A TENSOR

79

If A is invertible we find from (i) and (iv) of (2.11.12), on setting


B = A" 1 , that
det A ' 1 =^
(2.11.13)
det A
From (ii) of (2.11.12), and (2.11.13), we find that det A = 1 for an
orthogonal tensor A. An orthogonal tensor A for which det A = 1 is called
a proper orthogonal tensor.
EXAMPLE 2 . 1 1 . 1 Let a and b be arbitrary vectors. Show that the
second and third principal invariants of a (x) b are 0. If a and b are
orthogonal, show that the first principal invariant of a b is also 0.
Solution If at and b-t are components of a and b, then a^j are components of a (x) b so that
7 a(g)b = t r a ( x ) b = & = a b

(2.11.14)

If a and b are orthogonal, then a b = 0 and we get 7 a(g)b = 0.


Using (2.11.4), we get
//.b = j[(<*ibi)(akbk) - ((^)]

=0

(2.11.15)

Also, using (2.11.10) we get


IIh*h

= il(aibi)(ajbj)(akbk)
= 0

+ 2{aibk)(akbm)(ambi)

- 3(eA)(*^)(*A)]

(2.11.16)

Note: This example illustrates an interesting fact that the principal


invariants of a nonzero tensor can all be 0.
EXAMPLE 2 . 1 1 . 2 For a skew tensor A, show that IA = IIIA = 0 and
IIA = | | 2 , where is the dual vector of A.
Solution Since A is skew, all the diagonal elements of its matrix [^] are
equal to 0. Hence tr A = an = 0; that is,
7A = 0

(2.11.17)

Expressions (2.11.9) and (2.11.10) then give


IIIX = |It tr
r AA33 = I | aikakmami
Since au = -aJi9 this yields
IHK

~~lakiamkaim

interchanging dummy suffixes k and m\


= -HU

~laikakmami

(2.11.18)

80

ALGEBRA OF TENSORS

so that
/// A = 0

(2.11.19)

Also, using (2.11.17), expression (2.11.4) gives


IIA = ~\aikaki = \aikaik = (A A)

(2.11.20)

Expression (2.10.10) now yields


IIA = || 2
EXAMPLE 2.11.3

(2.11.21)

Let a^ be components of a tensor A. If


4 =

teipqtjnapraq*

(2.11.22)

show that dfj are components of a tensor. If this tensor is denoted by A*,
prove the following:
A(A*)r = (A*)rA = (det A)I

(i)

(2.11.23)

(ii) If A is invertible, then


_,

= ^ ( * )

(2 1L24)

(iii) For all vectors a, b


A*(a x b) = Aa x Ab

(2.11.25)

(iv) For all vectors a, b, c,


(A*)ra (b x c) = a Ab x Ac

(2.11.26)

Here A* is called the adjugate or cofactor of A, and (A*)r is called the


adjoint of A.
Solution Since a^ and eijk are components of tensors, the righthand side
of (2.11.22) represents a contraction of a product of tensors. As such afj are
components of a tensor, A*. Consequently, expressions (2.11.23) and
(2.11.24) follow from expressions (1.7.40) and (1.7.41).
Further, for any vectors a and b with components at and bi9 we have
[A*(a x b)]f = <rj}[a x b], = afj(ejmnambn)
2 ^ipq fyrs ^pr^qs ^jmn &m &n

using (2.11.22);
=

2ipq(rmSn

rnsn^Qpraqsambn

2.12

DEVIATORIC TENSORS

81

using the - identity;


= |[Aa x Ab], - [[Ab x Aa]f = [Aa x Ab],
which proves (2.11.25).
By use of (2.8.14), we have
(A*)ra (b x c) = a A*(b x c)
which by use of (2.11.25), yields (2.11.26).

Note: From (2.11.22) and (2.11.24), it follows that the components a~l of
A"1 are given by

^=dL**=2^)*''

(2 27)

2.12
DEVIATORIC TENSORS
Given a tensor A, suppose we define a new tensor A( as follows:
A(d) = A - j(trA)I

(2.12.1)

If Oij are components of A, then from (2.12.1) it follows that the


components afp of A(f) are given by
of = aij-\akkSij

(2.12.2)

Hence ajf} = an - \{?>akk) = 0; that is,


trA(rf) = 0

(2.12.3)

If we set
a = itrA

= iakk

(2.12.4)

then expression (2.12.1) can be rewritten as


A = I + A(fi

(2.12.5)

Thus, every tensor A can be represented as a sum of two tensors; one of


these, ai, is an isotropic tensor, and the other, A(d), is a tensor whose trace
is 0. Expressions (2.12.1) and (2.12.4) show that A(d) and a are uniquely
determined by A. As such, the representation (2.12.5) is unique. The part I
of this representation is called the spherical part of A and the part A(i/) is
called the deviatoric part or the deviator part of A. It is obvious that the

82

ALGEBRA OF TENSORS

spherical part of a tensor is always symmetric and that the deviator part is
symmetric (skew) if an only if the tensor is symmetric (skew). The deviator
part of a given tensor is sometimes referred to as the deviatoric tensor
associated with the given tensor.
From (2.12.3), we note that the first invariant of A(d) is always 0; that is,
/Aw) = 0

(2.12.6)

By use of expressions (2.11.3) and (2.11.9), we find that the second and
the third invariants of A(d) are given by
//A(rf) = -i-tr(A (d) ) 2
(c

IIIAm = jtr(A )

(2.12.7)

(2.12.8)

To suppress the negative sign appearing in the righthand side of (2.12.7),


some authors take -7/A^) as the second invariant of A(d).
EXAMPLE 2.12.1
A whose matrix is

Find the spherical and deviatoric parts of the tensor

1]

[_1

1J

[A] = [au] =

Solution We readily see that a = j tr A = 1, and


[af] = [au] - a[tj] = [au - aotj]

1 0

[l

l"
1

(2.12.9)

oj

Thus, the spherical part of the given tensor is I, and the deviatoric part is the
tensor whose matrix is [afp] given by (2.12.9).

EXAMPLE 2.12.2 Show that the deviator part of a tensor A is 0 if and


only if A is isotropic.
Solution If A is istropic, then by Theorem 2.5.1, A = l for some scalar
, so that tr A = 3. Expression (2.12.1) then gives A(d) = 0. Conversely, if
A(c = 0, then (2.12.5) shows that A is a scalar multiple of I. Hence A is
isotropic.

2.13 EIGENVALUES AND EIGENVECTORS


EXAMPLE 2.12.3

Show that

K<* = {(tr A)2 - tr A2} = // A - \l\


Solution

83

(2.12.10)

From (2.12.7) and (2.12.2), we get

= ~l(ik - -3<tmnik)(<lki ~ i ^ / m A i )

= jiWkk - aikaki] = i(|(tr A)2 - tr A2}

(2.12.11)

With the aid of (2.11.1) and (2.11.4), we find that


(Wkk - aikaki) = ( j / 2 + 2/7A - 72} = 2(//A - \l\)
Expressions (2.12.11) and (2.12.12) together yield (2.12.10).

(2.12.12)

2.13
EIGENVALUES AND EIGENVECTORS
It has been noted that a (second-order) tensor A operating on a vector v
gives rise to a vector Av. A case of great interest is the one in which Av is
collinear with v so that Av can be written as
Av =

(2.13.1)

for some real number . A unit vector v possessing this property, for a
given tensor A, is called an eigenvector or a principal vector of A, and the
real number is called the corresponding eigenvalue or the principal value.
Also, the direction of v is called a principal direction of A.
It is evident that, if v is an eigenvector of a tensor A and is the corresponding eigenvalue, then - v is also an eigenvector of A corresponding to
the same eigenvalue A. Thus, more than one eigenvector can correspond to
the same eigenvalue. But, for a given eigenvector there corresponds only
one eigenvalue.
As an example, we may note that (since Iv = v for every vector v) every
unit vector is an eigenvector of the identity tensor I, and the number 1 is the
eigenvalue corresponding to all these eigenvectors. In other words, for the
tensor I, every direction is a principal direction and all these directions
correspond to the principal value 1.
It may be noted that according to the definition of an eigenvalue and
eigenvector just given, an eigenvalue is a real number and an eigenvector is
a unit vector. The terms eigenvalue and eigenvector are used in this sense
throughout our discussion.

84

2.13.1

ALGEBRA OF TENSORS

EXISTENCE OF EIGENVALUES

The following theorem ensures that every tensor has an eigenvalue and
hence an eigenvector.
THEOREM 2 . 1 3 . 1 A number A is an eigenvalue of a tensor A if and
only if it is a real root of the cubic equation
A3 - 7AA2 + 7/ A A - 7// A = 0

(2.13.2)

where 7 A , IlA and 77/A are the principal invariants of A.


Proof Let [v], = t;, and [A]l7 = au. Then the vector equation (2.13.1)
reads as follows in the suffix notation:
or

ctijVj = Aviy

(au - Au)Vj = 0

(2.13.3)

When written in the expanded form, this equation yields the following
system of three equations:
(an - A)vl + al2v2 + anv3

=0

021^1 + (22 - ) 2 + #23^3 =


anVx

(2.13.4)

+ 032^2 + (33 ~ )>3 = 0

The determinant of the coefficients of these equations is


Un - A

al2

au

a2\

a22 - A

a23

a32

a33 - A

det(A - AI) =
|
s -A

T31
3

+ 7AA - 7/ A A + IIIA

(2.13.5)

With this information, we now suppose that A is an eigenvalue of A.


Then there exists a unit vector v such that (2.13.1) holds. Consequently, the
system of equations (2.13.4) has a nontrivial solution. Therefore, the determinant of the coefficients of the system is equal to 0. Expression (2.13.5)
now shows that A is a root of equation (2.13.2). Since A is a real number,
being an eigenvalue of A, it is a real root of equation (2.13.2).
Conversely, suppose that A is a real root of the cubic equation (2.13.2).
Then the determinant of the coefficients of equations (2.13.4) is 0 for this
A, and consequently the system (2.13.4) has a nontrivial solution, say w,.
This implies that
Au = Au
(2.13.6)
where u is the vector whose components are w^. If we set v = u / | u | ,
equation (2.13.6) becomes (2.13.1) and A is therefore an eigenvalue of A.

2.13 EIGENVALUES AND EIGENVECTORS

85

Thus, is an eigenvalue of A if and only if it is a real root of equation


(2.13.2). The proof of the theorem is complete.

It is important to note that, since (2.13.2) is a cubic equation with real


coefficients, it has at least one real root for ; consequently, A has at
least one eigenvalue (and therefore at least one eigenvector). Further, since
the coefficients / A , IIA and IIIA in equation (2.13.2) are invariant under
coordinate transformations, the roots of this equation are also invariant
under coordinate transformations. In other words, the cubic equation
(2.13.2) yields the same three roots in every choice of the coordinate axes.
This means that the eigenvalues of a tensor are independent of the coordinate system chosen for the purpose of their computation.
The polynomial on the righthand side of (2.13.5) is called the characteristic polynomial of the tensor A. Also, the cubic equation (2.13.2) (or its
negative version) is called the characteristic equation of A. Since A and AT
have the same principal invariants, it follows that the characteristic
equation of A r is identical with that of A. Hence A and AT have the same
eigenvalues. From equation (2.13.2) it also follows that 0 is an eigenvalue of
A if and only if IIIA = det A = 0.
It is obvious that A cannot have more than one eigenvalue if the cubic
(2.13.2) has a complex root. If A is symmetric, we will prove below that all
the three roots of (2.13.2) are real and each of these roots is an eigenvalue
of A.
2.13.2

EIGENVALUES OF A SYMMETRIC TENSOR

THEOREM 2.13.2 If A is a symmetric tensor, then all the three roots of


the characteristic equation of A are real, and therefore A has exactly three
(not necessarily distinct) eigenvalues.
Proof We have noted that at least one of the roots of the characteristic
equation (2.13.2) is real and this real root is an eigenvalue of A. Denote this
eigenvalue by Ai and let Vj be a corresponding eigenvector. Then, if vu are
components of Vj and a^ = a are components of A, we have
(2.13.7)

aijvlj = A^u

Introduce a coordinate system x\ such that the x[ axis is along the vector yx ;
that is, e\ = Vj. For this system, we have
i/ = ei ef- = V! e, = vu

(2.13.8)

If a\j are components of A in the x\ system, we have


tfn = <xip<xiQtiP = apqvlpvlq

= AxvXpvlp = Ai

(2.13.9)

86

ALGEBRA OF TENSORS

Figure 2.6. Coordinate system x[.

In obtaining (2.13.9), we have used (2.13.7), (2.13.8) and the fact the \l is
a unit vector. Also,
#12 = OLlpa2qapq =

0L

otlqa2paqp

2papqV\q

using (2.13.8) and that apq = agp;


=

^lOi2pVip

using (2.13.7);
= \21

= 21

using (2.13.8) and the orthogonality of a,,;


= 0

(2.13.10)

Similarly,
a'u = 0

(2.13.11)

In view of (2.13.9), (2.13.10) and (2.13.11), the characteristic equation of


A in the x[ system is given by
Ax - A

a'22 - A

*23

a23

"33 - A

= 0

which simplifies to
(Ax - A)[A2 - (a'22 + a'^)A + a'22a^ - a2j] = 0

(2.13.12)

Evidently, one of the roots of this equation is Ax and the other two roots,
say, A 2 and A 3 , are determined by the quadratic equation
A2 - (a'22 + a^)A + ('22a^ - a) = 0

(2.13.13)

2.13 EIGENVALUES AND EIGENVECTORS

87

The discriminant D of this equation is


D = (a22 + a'^f - 4(22^3 -

a)

= (*22 - ^33>2 + 4a > 0


Therefore 2 and 3 are also real.
Thus, all the three roots of the characteristic equation of A are real.
Since each real root of the characteristic equation is an eigenvalue of A by
Theorem 2.13.1, it follows that A has exactly three eigenvalues. Further,
since the three roots Al, 2 and 3 are not necessarily distinct, the three
eigenvalues of A need not be different from one another. From equation
(2.13.2) it follows that
/ A = Al + 2 + 3
IIA = A t A 2 + 2 3 + 3 !,
IIIA = A!A 2 A 3
2.13.3

(2.13.14a)
(2.13.14b)
(2.13.14c)

PROPERTIES OF EIGENVECTORS

The next three theorems provide some important properties of eigenvectors


of a symmetric tensor.
THEOREM 2 . 1 3 . 3 Eigenvectors corresponding to two distinct eigenvalues of a symmetric tensor A are orthogonal. (Or, the principal directions
corresponding to distinct principal values of A are orthogonal.)
Proof Let Aj and A 2 be two distinct eigenvalues of A, and \ l and v2 be
the corresponding eigenvectors. Then we have
Avj =

AJVJ;

Av 2 = A 2 v 2

(2.13.15)

Using (2.13.15), (2.8.14) and the fact that A is symmetric, we find that

" v2 = (AvO v2 = (Av2) \x = A 2 v 2 \x

Since A{ ^ A 2 , it follows that v t v2 = 0. The vectors \{ and v2 are thus


orthogonal.

THEOREM 2 . 1 3 . 4 For a symmetric tensor A, there exist (at least) three


eigenvectors that are mutually orthogonal. (Or, a symmetric tensor has at
least three mutually perpendicular principal directions.)
Proof Since A is a symmetric tensor, it has got exactly three eigenvalues,
A j , 2 , 3 , by Theorem 2.13.2. Let vl9 v2 and v3 be the corresponding
eigenvectors. Since Al, 2 and 3 are not necessarily distinct, the following
three cases arise.

88

ALGEBRA OF TENSORS

Figure 2.7. Theorem 2.13.4, case (ii).

Case (i). Suppose Al9 A2 and 3 are all distinct. Then, by Theorem
2.13.3, we have yl v2 = v2 v3 = v3 yx = 0. Thus, v l 5 v2 and v3 are
mutually orthogonal.
Case (ii). Suppose two of the three eigenvalues are equal; say, Ar =
2 ^ 3 . Then yx v3 = v2 v3 = 0, by Theorem 2.13.3; thus v3 is orthogonal to both V! and v2 (see Figure 2.7).
Take any unit vector u coplanar with Vj and v 2 . Then, u = avj + y2 for
some numbers a and , and we have
Au = A[avj + y2]
= aAvj + /?Av2
by the linear property of A;
= a(A l V l ) + (A2y2)

= Al(awl + y2)

since Ax = A 2 ;
= Aju

Thus u is an eigenvector corresponding to A j . Since u is arbitrary, it follows


that every unit vector coplanar with Vj and v2 is an eigenvector corresponding to A!, and therefore there are infinitely many coplanar eigenvectors. Of
these vectors choose a vector u2 that is orthogonal to v x . Then
Vj U 2 = U 2 V3 = V3 V! = 0

showing that there exist (at least) three mutually orthogonal eigenvectors.
Case (Hi). Suppose all the three eigenvalues are equal; that is, Aj = A 2 =
A 3 . Then the characteristic equation should be of the following form:
(A1 - A) 3 = 0

2.13 EIGENVALUES AND EIGENVECTORS

89

or
A! -

Aj - A

A! - A

(2.13.16)

Comparing the determinant in the lefthand side of (2.13.16) with that in


(2.13.5), we find that an = a22 = #33 = A t , other atj = 0. Thus A = A t I
so that for any vector u, we have Au = ^ . Hence, A has infinitely many
eigenvectors, all of which correspond to the same eigenvalue A j . Of these
vectors we can certainly choose (at least) three vectors yx, v2 and v3 that are
mutually orthogonal. Thus, in this case also, there exist three mutually
orthogonal eigenvectors for A. This completes the proof.

THEOREM 2 . 1 3 . 5 Given a symmetric tensor A, there exists a (at least


one) coordinate system with respect to which the matrix of A is diagonal.
(Or, given a symmetric tensor A, if the coordinate axes are chosen along the
principal directions of A, then with respect to these axes the matrix of A is
diagonal.) (Such axes are called principal axes of A.)
Proof By Theorem 2.13.4, there exist at least three eigenvectors yx, v 2 , v3
of A that are mutually orthogonal. Since - v is also an eigenvector when v
is an eigenvector, we may choose these three mutually orthogonal eigenvectors in such a way that they form base vectors of a righthanded
rectangular coordinate system.
We recall from (2.7.2) that, with respect to the base vectors e,, the
components a^ of A are given by a^ = e, Ae,. Hence, with respect to the
coordinate system with vz as base vectors, we have
au

= Vj Av! = V! ^ ! = Aj

(2.13.17)

al2 = V l Av 2 = yx A 2 v 2 = 0
Similarly,
#22 =

2>

#33 = A 3 ,

al3

(2.13.18)

= a23 = 0

In view of (2.13.17) and (2.3.18), it follows that, with respect to the


coordinate system having vf as base vectors, the matrix of A is

[I =

*1

which is a diagonal matrix and the proof is complete.

(2.13.19)

90

ALGEBRA OF TENSORS

Note that the diagonal elements of the matrix (2.3.19) are the eigenvalues
of A.
Summary: The results proved in Theorems 2.13.2 to 2.13.5 may be summarized as follows. A symmetric tensor possesses exactly three eigenvalues
that are not necessarily distinct. The principal directions corresponding to
distinct eigenvalues are orthogonal. There exist at least three mutually
orthogonal principal directions. When the axes are chosen along these
directions, the matrix of the tensor is purely diagonal with the eigenvalues
as the diagonal elements.
EXAMPLE 2 . 1 3 . 1
A whose matrix is

Find the eigenvalues and eigenvectors of the tensor


[2

0"

[A] =

[o 4 -3
Solution

The characteristic equation of the given tensor A is


2 -

3-

| = 0

-3 -

Expanding the determinant we get the cubic equation


(2 - )(2 - 25) = 0
whose roots are
A, = 2 ,

A2 = 5,

A3 = - 5

These are the eigenvalues of the given tensor A.


To find an eigenvector v of A, associated with the eigenvalue Al = 2, we
substitute for atj from the given matrix and take A = 2 in equations
(2.13.4). Thus, we obtain the following three equations for the three
components y, of v:
2vx = 2vl
3v2 + 4t>3 = 2v2
4v2 - 3i>3 = 2v3

The first of these equation is identically satisfied for any vl, and the last
two equations give v2 = v3 = 0. Since v has to be a unit vector, we must
have ViVi = 1 from which it follows that vx = 1 . Thus, an eigenvector
associated with the eigenvalue A z is Vj e j .

2.13 EIGENVALUES AND EIGENVECTORS

91

For = A 2 = 5, equations (2.13.4) yield 3vx = 0, v2 - 2v3 = 0. These


equations together with the condition ,-;, = 1 give vx = 0, v2 = 2v3 =
2/V5. Thus, an eigenvector associated with the eigenvalue A 2 = 5 is
v2 = (1/V5)(2e 2 + e 3 ).
Similarly, an eigenvector associated with the eigenvalue 3 = - 5 can be
found to be v3 = (1/V5)(e 2 - 2e 3 ).

EXAMPLE 2 . 1 3 . 2 Show that for an orthogonal tensor Q, the only


eigenvalues are 1 . Show further that if Q is proper, then +1 is an eigenvalue, and that if Q is not proper, then - 1 is an eigenvalue.
Solution Let be an eigenvalue of Q with v as a corresponding eigenvector. Then
A2 = A2(v v) = Av Av = Qv Qv
Since Qv Qv = v v (see Example 2.8.3), it follows that A2 = v v = 1.
Thus, an eigenvalue of Q should be 1.
Now, suppose that Q is proper; that is, detQ = 1. Then
det(Q - I) = det(Q - l)T = det(Q r - I)
= det(Q r - Q r Q) = det Q r (I - Q)
= (det QT) det(I - Q) = det(I - Q)
= -det(Q - I)
so that det(Q - I) = 0, showing that 1 is an eigenvalue of Q.
Next, suppose that Q is not proper; that is, det Q = 1. Then we obtain,
following the steps of the previous case,
det(Q + I) = det(Q r ) det(Q + I) = -det(Q + I)
so that det(Q + I) = 0, showing that - 1 is an eigenvalue of Q.

EXAMPLE 2 . 1 3 . 3 If A is a skew tensor, show that (i) A has only one


eigenvalue, and (ii) there exists a unit vector v such that Av = 0.
Solution In Example 2.11.2, it has been shown that for a skew tensor A,
we have IA = IIIA = 0 and IIA = || 2 , where is the dual vector of A.
Hence in this case the characteristic equation (2.13.2) reduces to
A3 + || 2 = 0
This equation has only one real root, equal to 0. Thus A has only one
eigenvalue, namely, 0. A corresponding eigenvector v is such that
Av = Ov = 0.

92

ALGEBRA OF TENSORS

EXAMPLE 2 . 1 3 - 4 Let A be tensor and A be an eigenvalue of A with v


as a corresponding eigenvector. For any positive integer n9 show that An is
an eigenvalue of A" with v continuing as a corresponding eigenvector.
Deduce that if A is invertible, then A~n, A ^ 0, is an eigenvalue of
n
A~ = (A" 1 )" with v continuing as a corresponding eigenvector.
Solution

Given that Av = Av, we have to show that


=

(2.13.20)

for all positive integers n. Let us employ the method of induction.


For n = 1, (2.13.20) is true by data. We assume that (2.13.20) is true
for n = r; that is,
Arv = Arv
Then A r+1 v = A(Arv) = A(Arv) = Ar(Av) = A'(Av) = +1 .
This shows that (2.13.20) is valid for n = r + 1 as well. Hence, by
induction, (2.13.20) is valid for all positive integers n.
If A is invertible, then A has a nonzero eigenvalue A and equation
Av = Av gives
A _1 v = - v
A

(2.13.21)

showing that 1/A is an eigenvalue of A - 1 with v as a corresponding


eigenvector. Applying the result (2.13.20) to A - 1 and denoting (A - 1 )" by
~, we obtain
A" n v = A""v
(2.13.22)
This shows that A - " is an eigenvalue of A" n with v continuing as a
corresponding eigenvector.

EXAMPLE 2 . 1 3 . 5 Let A be a symmetric tensor with A, as eigenvalues


and v, as corresponding eigenvectors. Show that A can be represented as
A=

*(**)

k= 1

(2.13.23)

This is known as the spectral representation of A.


Solution In Example 2.7.3, it has been shown that every tensor A has the
following representation in a coordinate system having e, as base vectors:
A = ijti (x) e,. Hence, with respect to a coordinate system with the eigenvectors v, as base vectors, a symmetric tensor A has the representation
A = dijVi V;

(2.13.24)

2.13 EIGENVALUES AND EIGENVECTORS

93

Also, with respect to this coordinate system, the matrix of A is given by


(2.3.19). Using (2.3.19), the representation (2.13.24) becomes
A = A^V! vO + A2(v2 (x) v2) + A3(v3 <g> v 3 ),
which is (2.13.23).
EXAMPLE 2 . 1 3 . 6

Show that for every tensor A,


A3 - / A A 2 + / / A A - / / / A I = 0

This result is known as the Cayley-Hamilton

(2.13.25)

theorem,

Solution Consider the tensor A - a i , where a. is an arbitrary scalar, and


denote its adjoint by H. Then (2.11.23) and (2.13.5) yield
H(A - ocl) = {det(A - al)jl
= (-a3

+ IAa2

- IIAa + 7// A )I

(2.13.26)

The form of the righthand side of (2.13.26) suggests that H must be of the
form
(2.13.27)
H = 2 + aF + G
for some tensor E, F, G. Substituting for H from (2.13.27) in the lefthand
side of (2.13.26) and equating the corresponding terms (tensors), we get
E = I,

EA - F = 7 A I,

G - FA = 7/ A I,

Using these relations, we find that

GA = 7// A I
(2.13.28)

A 3 - 7AA2 + 7/ A A - 7// A I = A 3 - (A - F)A2 + (G - FA)A - GA


(2.13.29)
Evidently, the righthand side of (2.13.29) is the zero tensor, and (2.13.25) is
proven.

Note: When expressed in words, the identity (2.13.25) reads: Every tensor
satisfies its own characteristic equation.
EXAMPLE 2 . 1 3 . 7 (i) If is an eigenvalue of a tensor A with v as a
corresponding eigenvector, show that ( - j tr A) is an eigenvalue of A(cf),
with v as a corresponding eigenvector.
(ii) Deduce that, if A is symmetric and , are the eigenvalues of A, then
the eigenvalues of A(d) are
A?> = i(2A t - 2 - 3)

= |(2 2 - 3 - A t )

?> = (23 - Ai - 2) J

(2.13.30)

94

ALGEBRA OF TENSORS

(iii) Further, show that


(2.13.31)
Solution (i) If is an eigenvalue of A with v as a corresponding eigenvector, (2.12.1) yields
A(</)v = { -}(trA)}v
showing that { - j(tr A)} is an eigenvalue of A(rf) with v as a corresponding eigenvector.
(ii) If A is symmetric, then A(i/) is also symmetric, and from the result just
proven it follows that the principal directions of A(d) are coincident with
those of A and that if , are the eigenvalues of A, then ^ = {, - j(tr A))
are the eigenvalues of A(c. Since tr A = A{ + 2 + 3 by (2.13.14a), we get
? = (, - |(, + 2 + 3)}

(2.13.32)

from which expressions in (2.13.30) follow.


(iii) By virtue of the expressions (2.13.14b, c), we have
//A(.) = ?>?> + ?>?> + A f < >

///Ao =

(2.13.33)

Substituting for ^ from expressions (2.13.30) into these, we readily


obtain the expressions (2.13.31).

2.14
POLAR DECOMPOSITION
In Section 2.9, it has been shown that every tensor A can be represented as
a sum of two tensors, one symmetric and the other skew. Another useful
representation of a tensor involving products of tensors is obtained in this
section. We need some definitions and preliminary results before taking up
the derivation of the representation.
2.14.1

POSITIVE DEFINITE TENSOR

A tensor A is said to be positive definite if


a-Aa>0
for every nonzero vector a.

(2.14.1)

2.14

POLAR DECOMPOSITION

95

A trivial example of a positive definite tensor is the identity tensor I. In


fact, every invertible tensor A yields two positive definite symmetric
tensors: A r A and AA r . This is proved in the following theorem.
THEOREM 2 . 1 4 . 1 For any tensor A, the tensors A A r and A r A are
symmetric and if A is invertible, A A r and A r A are positive definite.
Proof

For any tensor A, we have


( ) = ( ) =
( ) = ( ) =

(2.14.2)

Hence, A r A and A A r are symmetric.


If A is invertible, then for any nonzero vector a, we have Aa ^ 0.
Because, if Aa = 0, the A _1 (Aa) = 0 or a = 0. Similarly, A r a 5* 0. Consequently, we obtain, on using (2.8.14),
a (A r A)a = a A r (Aa) = Aa Aa > 0
a (AA r )a = a A(A r a) = A r a A r a > 0

(2.14.3)

Hence, A r A and A A r are positive definite. This completes the proof of the
theorem.

The next theorem contains a characteristic property of positive definite


tensors.
THEOREM 2 . 1 4 . 2
values.

A positive definite tensor has only positive eigen-

Proof Consider an eigenvalue and an associated eigenvector v of a


positive definite tensor A. Then,
v Av = v = ( v) =
Since a Aa > 0 for any a, it follows that > 0, and the proof is
complete.

COROLLARY

A positive definite symmetric tensor is invertible.

Proof Since a symmetric tensor has three eigenvalues, from Theorem


2.14.2 it follows that all the three eigenvalues of a positive definite symmetric tensor are positive. Consequently, the determinant of such a tensor
is positive, and therefore, the tensor is invertible.

96

2.14.2

ALGEBRA OF TENSORS

SQUARE ROOT OF A TENSOR

Given a tensor A, if there exists a tensor B such that B2 = A, then B is


called a Square root of A, denoted VA.
For example a square root of the unit tensor I is the tensor whose matrix is

[vT] =

0 0'

0 -1 0

L l.

THEOREM 2.14.3 Every positive definite symmetric tensor has a


unique square root that itself is symmetric and positive definite.
Proof Let A be a positive definite symmetric tensor. Then the eigenvalues
, of A are all positive and the corresponding eigenvectors v, may be chosen
to be mutually orthogonal. Also, A has the following spectral representation; see (2.13.23).
3

A = 4(* v t )
k=\

(2.14.4)

Let us now introduce a tensor B by


B = V^(v,(g)v,)
k=\

(2.14.5)

The fact that B is a tensor is obvious from the nature of the righthand side
of (2.14.5). By use of (2.8.15) we find from (2.14.5) that B r = B; hence B
is symmetric. Also, for any nonzero vector a, we get
3

Ba = VX^(v* v*)a = VX^(v* a)v*


k=\

k=\

by using (2.4.37). Hence


a-Ba =

VX^-a)2 > 0

Thus, B is positive definite


Further,
3

B2

VA^(V*V*)|

V^(vmvj

k= 1
3
3

^ V ^ ( v * vk)(ym vm)

k=1m=1
3
3

= >/^>/^(*)(<8) y m)
k=\

m=l

(2.14.6)

2.14

POLAR DECOMPOSITION

97

see (2.4.40). Noting that v* \m = 0 for k 5* m and v* v* = 1, and using


(2.14.4), we find from (2.14.6) that
3

B 2 = *(* v*) = A
k=\

(2.14.7)

Thus, B is a positive definite symmetric tensor whose square is A. That is,


B is a square root of A.
Suppose that there is another positive definite symmetric tensor B' whose
square is also A. If , are the eigenvalues of B' and wf are corresponding
eigenvectors, then, from Example 2.13.4, it follows that 2 are eigenvalues
of A and wf are corresponding eigenvectors. Thus, ] = , and w, = vf. As
such, the spectral representation for B' is
B'= VX^(v*(g)v*)
k=\

Comparing this with (2.14.5), we find that B = B'. Thus B is unique, and
the proof of the theorem is complete.

COROLLARY If A is a positive definite symmetric tensor, then VX is


invertible.
Proof The result follows from Theorem 2.14.3 and corollary to Theorem
2.14.2.

We are now ready to establish the representation of a tensor as indicated


in the beginning of the section.
2.14.3

POLAR DECOMPOSITION THEOREM

THEOREM 2.14.4
form

Every invertible tensor A can be represented in the


A = QU = VQ

(2.14.8)

where Q is an orthogonal tensor and U and V are positive definite symmetric tensors such that U2 = ArA and V2 = AA r . Furthermore, the
representations are unique.
Proof Since A is invertible, the tensors ArA and AA r are symmetric and
positive definite (by Theorem 2.14.1). Hence, these have unique positive
definite symmetric square roots which are invertible (by Theorem 2.14.3
and its corollary). Let us denote the square root of ArA by U and define the
tensor Q by
(2.14.9)
Q = AU"1

98

ALGEBRA OF TENSORS

Then
QrQ = ( 1 ) =

UT)-lATAV-1

= VlV2Vl = I

(2.14.10)

Hence Q is an orthogonal tensor. From expression (2.14.9), we also get


A = QU

(2.14.11)

This is the first of the two representations specified in (2.14.8). To prove


the uniqueness of this representation, we assume that A has another
representation
A = Q

(2.14.12)

where Q is orthogonal and 0 is symmetric and positive definite. Then


ATA = (Q) r (Q) = (Q r Q) = 2

(2.14.13)

showing that is a square root of A T A. But ATA has a unique square root
U. As such = U. From (2.14.12) and (2.14.9) we then get
Q = AU-1 = AU-1 = Q
Thus, the representation (2.14.11) is unique.
Next, let us define V by
V = QUQ r
Then
V2 = (QUQ r )(QUQ r ) = QU 2 Q r = (QU)(QU) r = A A r

(2.14.14)

(2.14.15)
(2.14.16)

Thus, V is the square root of AA r ; as such it is symmetric and positive


definite. Also,
A = QU = (QUQ r )Q = VQ
(2.14.17)
This is the second of the representations specified in (2.14.8). Expression
(2.14.15) shows that V is uniquely determined by Q and U. Hence this
representation is also unique. The proof of the theorem is complete.

Note: The representation A = QU is called the right polar decomposition


of A, and the representation A = VQ the left polar decomposition. The
name polar decomposition stems from the analogy between these representations and the polar representation z = relB of a complex number z
having r as its modulus and as the argument.

2.14

EXAMPLE 2 . 1 4 . 1

POLAR DECOMPOSITION

99

Find the square root of the tensor B whose matrix is


1

[B] = | 0

6 |

(2.14.18)

13

Solution We note that B is a symmetric tensor. It is easy to verify that it


is positive definite also. If U is the square root of B, we have U 2 = B. Hence
the components wl7 of U are given by uikukj = bij9 where by are components
of B. Using (2.14.18), we find that
= 1,

ulku2k = 0,

ulku3k = 0

"2* "2* = 4,

u2ku3k = 6,

u3kUsk = 13

ulkuik

Solving these equations, we obtain


"ll =

23 = "32 =

2,

*33

and other u^ = 0

3;

Thus,
[U] =

(2.14.19)

This is the matrix of VS.


EXAMPLE 2 . 1 4 . 2
matrix is

Find the polar decompositions of the tensor A whose


"0
[A] =

0 -2

(2.14.20)

Solution For the given A, we readily note that det A ^ 0. Hence A is


invertible.
From (2.14.20), we find that
r

[A A] = [A ] [A] =

13

(2.14.21)

100

ALGEBRA OF TENSORS

If U is the square root of A r A, the use of the result proved in Example


2.14.1 gives
" 1 0 0
[U] = I 0

(2.14.22)

The inverse of the matrix [U] can be found as


1

_1
4

1
2

1
2

[U]- = [IT ] =

ol
(2.14.23)

-1

If we set Q = A U , then we get, from (2.14.20) and (2.14.23),


1

[Q] = [ ] =

0 -1

0 1

(2.14.24)

With [U] and [Q] determined by (2.14.22) and (2.14.24), the polar
decomposition A = QU is obtained.
Since A = VQ is the other polar decomposition, we have
[V] = [AHQ]-1 = [A][Qf
Using (2.14.20) and (2.4.24), we obtain
0
[V] =

0-2

-2

(2.14.25)

With [Q] and [V] known from (2.14.24) and (2.14.25), the decomposition
A = VQ is also obtained.

EXAMPLE 2 . 1 4 . 3 Let A be an invertible tensor and A = QU = VQ be


the polar decompositions of A. Show that U and V have identical eigenvalues. Also, find the eigenvectors of V in terms of those of U. Deduce the
representations of Q and A in terms of the eigenvalues and eigenvectors of
U and V.
Solution We first note that U and V are symmetric tensors, and each of
these has three eigenvalues and mutually orthogonal eigenvectors. Let be

2.15

EXERCISES

101

an eigenvalue of U and v be a corresponding eigenvector. Then


(QU)v = Q(Uv) = Q(Av) = A(Qv)

(2.14.26)

Since QU = VQ, (2.14.26) yields


V(Qv) = (VQ)v = (QU)v = A(Qv)

(2.14.27)

This shows that is also an eigenvalue of V with Qv is a corresponding


eigenvector.
Hence, if A, are the eigenvalues of U and v, are corresponding eigenvectors, then Ai are also eigenvalues of V with Qv, as corresponding
eigenvectors.
If v, are taken as base vectors, we have from (2.4.36) and (2.13.23),
3

I = v* v*.

U = *(* v*)
k=\

(2.14.28)

Hence
Q = Ql = Q(v* yk) = (Qv*) v*

(2.14.29)

A = QU = Q *(* yk)
k=\

= A*Q(v*v*) = Ak(Qvkvk)
k=\

k=\

(2.14.30)

These are the representations for Q and A in terms of the eigenvalues and
eigenvectors of U and V.

2.15
EXERCISES
1. Find the matrix [,,] if the x\ system is obtained by rotating the xt system in the
sense of a righthanded screw through an angle
(a) 90 about the jq axis.
(b) 60 about the x2 axis.
2. Compute a,, for a pair of coordinate systems related by ei = - e 3 , e'2 = e 2 ,
e'3 = el. By what kind of rotation of the xt system the x\ system can be obtained? If
a point P has coordinates(1, 1,0) in the x, system, find its coordinates in the jcf
system. Also, if a point Q has coordinates (1,0, 1) in the x\ system, find its
coordinates in the xt system.

102

ALGEBRA OF TENSORS

3 . Complete the following table of direction cosines:

ei
e2
e3

*i

e2

e3

1/V3
0
-2/V6

1/V3
1/V2
1/V6

?
?
?

4. The x[ system is obtained by rotating the xt system about the xl axis through an
angle of 60 in the sense of a righthanded screw. If a plane has the equation
xx + 2x2 - x 3 = 1 in the x, system, find its equation in the x\ system.
5. Show that

Eiik
ijk &i
pqr
a

kp

kq

kr

Deduce that
x

*ij

= 4-,
e

2 ipq tjrs &pr *-*qs

6. Show that
OipOjgakparsijkr

qs

7. The x[ system is obtained by rotating the *, system through an angle of 90 about


the x2 axis in the sense of a righthanded screw. A certain entity, which can be represented as an ordered triplet in every coordinate system, is represented by the triplet
(1, 2, 3) in the xt system and by the triplet ( 2 , - 1 , 3) in the x\ system. Verify whether
this entity is a vector.
8. Show that e = ,-,,; e, = ,-,-}. Hence deduce the relations (2.2.8) and (2.2.9).
9. Using (2.2.8) and (2.2.9), deduce the relations (2.3.3) from (2.3.4) and
vice-ver sa.
1 0 . In the x, system, a vector a has components
ax = 0,

a, =

a3 =

vT

"V2

and a tensor A has its matrix


1 0 -1
[*</]

0
-1

0 -1

If the x\ system is as defined in Exercise 4, find the components of a and A in the


x[ system.
1 1 . If a and are scalar s and a^ and by are components of tensors, show that
ctciij bij are components of a tensor.

2.15

EXERCISES

103

1 2 . If at are components of a vector and by are components of a tensor, show that


Oibij are components of a vector.
1 3 . Let at be components of a vector and btj be elements of a matrix, referred to
the Xi system. If afi^ are components of a vector, show that bu are components of
a tensor.
1 4 . If atj and bijk are tensor components, show that so are djjbkrs.
1 5 . Assuming that } = \j9 show that u are components of a tensor.
1 6 . Assuming that eijk = ,*, show that eijk are components of a tensor.
1 7 . If at are components of a vector, show that ^ *
tensor.

are

components of a skew

1 8 . If a,, are components of a tensor, show that siJkajk are components of a vector.
1 9 . If aijk are components of a third-order tensor, show that eijkaijk is a scalar
invariant.
2 0 . If aijkm are components of a fourth-order tensor, show that aijkk
components of a second-order tensor.

are

2 1 . If a is a unit vector, show that the tensors a a and I - a (x) a are equal to
their own squares.
2 2 . Prove the following identities:
(i)

(aA)a = a(Aa)

(ii)

(A B)a = Aa Ba

(iii)

(AB)a = A(Ba)

(iv)

(AB)(Ca) = A[B(Ca)]

(v)

[A(BC)]a = A[(BC)a] = A[B(Ca)]

2 3 . If A is a tensor such that A B = 0 for every tensor B, show that A is the zero
tensor.
2 4 . Show that there is no pair of vectors a and b such that a b = I.
2 5 . Show that
(i) (a <g> b) (c d) = (a c)(b - d)

(ii) (e, e,) (e* em) =

ikjm

2 6 . Prove the following identities:


(i)
(ii)
(iii)

a Ab = A (a <g> b)
(a b)A = a <g> (A r b)
(Ae p ) e p = A

2 7 . If b is a vector such that a b = 0 for every vector a, show that b = 0.


2 8 . If a is a unit vector, show that I - 2a a is an orthogonal tensor. Is this a
proper orthogonal tensor?

104

ALGEBRA OF TENSORS

29. A linear transformation A transforms the base vectors e, as follows:


Aej = 2e2, Ae2 = et + e3, Ae3 = e^ Find the components of A. Also find the
vector to which v = ex - e2 + e3 is transformed under A.
30. A transformation A transforms every vector a to the vector aa where a is a
scalar. Show that A is an isotropic tensor.
31. Suppose A is an entity defined by A = ,-,, (x) e,, where } are real numbers.
Show that A is a tensor with a^ as components.
32. Show that a vector is isotropic if and only if it is the zero vector.
33. Let aijk be components of a tensor of order 3. Show that this tensor is isotropic
if and only if aijk = oteijk for some scalar a.
34. Verify the identities (2.8.2) to (2.8.5) and (2.8.10) to (2.8.13).
35. Let A, B, C be tensors. Show that
A (BC) = (BrA) C = (AC7) B
36. If the matrix of a tensor is symmetric in one coordinate system, show that it
is symmetric in all coordinate systems.
37. If the matrix of a tensor is skew-symmetric in one coordinate system, show that
it is skew-symmetric in all coordinate systems.
38. The product b A of a vector b and a tensor A is defined in Example 2.8.5.
Show that (b I)c = b x c for any vector c.
39. If A is a symmetric tensor and B is an arbitrary tensor, show that
A B = A B r = | A (B + B r )
40. If A is a skew tensor and B is an arbitrary tensor, show that
A B = - A B r = | A (B - B r )
41. Prove that the representation of a second-order tensor as a sum of a symmetric
tensor and a skew tensor is unique.
42. Prove the following:
(i) If A B = 0 for every symmetric tensor B, then the tensor A is skew.
(ii) If A B = 0 for every skew tensor B, then the tensor A is symmetric.
43. For every vector a and every tensor A, show that
(i)
a (skw A)a = 0
(ii)
a (sym A)a = a Aa

2.15

EXERCISES

105

4 4 . Find the dual vector of the skew part of the tensor A whose matrix is
1-2
[A] = I 2
-5

5"
0-3

If u = 2ej - 3e2 + e 3 , verify that (skw A)u = u.


4 5 . Find the skew tensor whose dual vector is = ! - 2e2 + 3e 3 .
4 6 . If ax are components of a vector, show that
0

a3

-a2

-a3

ax

a2 ax

is the matrix of a tensor.


4 7 . Let a and b be arbitrary vectors. Show that (a (x) b - b (g) a) is a skew tensor
and that b x a is its dual vector. Deduce that j(b x a) is the dual vector of
skw(a (g) b). (The tensor (a (g) b - b (g) a) is called the exterior product of a and b.)
4 8 . Let A and B be tensors; prove the following
(i)

tr A = tr(symA)

(ii)

tr(AB) = tr(BA)

(iii) If A is symmetric and B is skew, then tr(AB) = 0


tr(A r B) = A B

(iv)

4 9 . Let A and B be skew tensors and u and v be their respective dual vectors. Show
that
(i) AB = v

(u v)I

(ii) tr(AB) = 2(u v)

5 0 . Show that the equation AB = BA holds for two skew tensors A and B if and
only if they have the same dual vectors.
5 1 . Prove the identities (2.11.12).
5 2 . Show that a skew tensor is not invertible.
5 3 . Show that e* (g) e'k is a proper orthogonal tensor.
5 4 . Let A* be the adjugate of a tensor A (see Example 2.11.3). Prove the
following:

(0

(A r )* = (A*) r

(ii)

(A*)* = (det A)A

(Hi)

trA* = // A

(iv)

A(a x A r b) = (A*a) x b, for all vectors a and b

106

2 ALGEBRA OF TENSORS

55. By employing the Cayley-Hamilton theorem, show that


(*)= A2-IAA

(i)

+ IIAI

If A is invertible, A" = (IIIAy iA

(ii)

- 7AA + 7/AI)

56. Obtain the following expressions for the principal invariants of a tensor A.
*

HA

HI A

~2ipqejpqaij
2ijkrskairajs

Zeijkpqraipajqakr

57. For any tensor A and any invertible tensor B, prove that
^ 1

HBAE1

= IA,

= H A >

IH*AB-i

= HI A

58. For any tensor A and any vectors u and v, prove that
Ar(Au X Av) = IIIA(u x v)
59. If A is a tensor, show that for any noncoplanar vectors u, v, w
(i)

[u, v, w]/ A = [Au, v, w] + [u, Av, w] + [u, v, Aw]

(ii)

[u, v, w]// A = [Au, Av, w] + [Au, v, Aw] + [u, Av, Aw]

(iii)

[u, v, yv]IIIA = [Au, Av, Aw]

60. For any vectors a, b, c, u, v, prove that


[a, b, c] tr(u v) = [(a v)u, b, c] + [a, (b v)u, c] + [a, b, (c v)u]
= (a u)[u, b, c] + (b v)[a, u, c] + (c v)[a, b, u]
61. Show that
IIIA = |{tr A3 - IA tr A2 + IIA tr A)
= |(tr A3 - f(tr A2)(tr A) + \{ix A)3)
62. If Ix = tr A, I2 = tr A2, I3 = tr A3, show that

63. If A is an invertible tensor, show that


'A1

mA

\-
A

ni A

III A -' =

64. Show that


IIIAm = /// A - | / A / / A + M

IIIA

2.15

EXERCISES

107

6 5 . Find the principal invariants of tensors whose matrices are given below:
1 0

0)

2 3

1
(ii)

0 -1

2 - 1 0
0

1 -2

nee find their ei genvalues.


6 6 . For the tensors with the following matrices, find the eigenvalues and the
principal directions:

0)

0"

0 3

' 3 - 1
(ii)

1 1

1 1

1 1 1 "
(Hi)

-1

0]
0

0 1J

"10
(iv)

2]

-2j

6 7 . If is an eigenvalue of a tensor A, show that for any vectors a, b, c,


[Aa - Aa, Ab - Ab, Ac - Ac] = 0
6 8 . For the tensor a (x) b, show that a b is an eigenvalue and a is a corresponding
eigenvector.
6 9 . Let A be a symmetric tensor and a be a unit vector. Show that a is an eigenvector of A if and only if A(a (x) a) = (a a)A.
7 0 . If A is a positive definite symmetric tensor, show that A - 1 is also positive
definite and symmetric and that V A - 1 = (VX)" 1 .
7 1 . Show that a symmetric tensor is positive definite if and only if its fundamental
invariants are positive.
7 2 . Find the square roots of the tensors whose matrices follow:
1 0 -1

(0

- 1 0

1 0
0

-3

(ii)

7 3 . Find the polar decompositions of the tensors whose matrices follow:


1 0
(i)

1 0

1 0

0 1

(ii)

2 3

108

(Hi)

ALGEBRA OF TENSORS

2 -1

0"

0 -1

(iv)

1 0

-2

74. In the polar decomposition A = QU = VQ of an invertible tensor A, show


that U and V have the same fundamental invariants.
75. Let A = QU = VQ be the polar decomposition of an invertible tensor A. If ,
are the eigenvalues of U and v, are the corresponding eigenvectors, show that A -1
can be represented in the form

A_1 = (*)-1****)
A: = 1

CHAPTER 3
CALCULUS OF TENSORS

3.1
INTRODUCTION
This chapter deals with a brief study of differential and integral calculus of
vector and tensor functions. It is assumed that the reader is already familiar
with calculus of several variables, particularly vector calculus.

3.2

SCALAR, VECTOR AND TENSOR FUNCTIONS


By a vector or a tensor function, we mean a vector or a tensor (respectively)
whose components are real-valued functions of one or more real variables.
Thus, if the components at of a vector a are real-valued functions of a
(single) real variable /, we say that a is a vector function of /, denoted a(0Similarly, if the components atj of a tensor A are real-valued functions of
t, we say that A is a tensor function of t, denoted A(t). Also, if a scalar
is a real-valued function of /, it is denoted as usual by </>(/). Generally, t is
assumed to vary over a specified interval on the real line.
109

110

3 CALCULUS OF TENSORS

In studying scalar, vector and tensor functions, we assume that the components being dealt with are defined with respect to fixed sets of coordinate
axes and that the components possess derivatives of any desired order.
Thus, if 0,(0 are components of a vector function a(/), we assume that
the nth derivatives [dn/dtn\[ai{t)] exist for any desired n. It is easy to see
that [dn/dtn][ai(t)} are also components of a vector. Because, if we set
Ci = [dn/dtn]{ai(t)) in all coordinate systems, we find that
ci = -^nia(t)}

niaikak(t)l

using (2.3.5a);
dn

= OLik

{ak{t)}

= OLikCk

The vector whose components are [dn/dtn]{ai(t)) is called the nth derivative
of a, denoted dna/dtn. In particular, the first derivative da/dt is the vector
whose components are da^/di; that is,
da
dt

-*

(3.2.1)

Obviously, if a is a constant vector, then its derivatives of all orders are


equal to the zero vector.
Similarly, if a^(t) are components of a tensor function A(0, we assume
that the nth derivatives [dn/dtn]{ajj(t)) exist for any desired n. It may be
verified that these derivatives are components of a tensor; this tensor is
called the nth derivative of A, denoted dnA/dtn. In particular, the first
derivative dA/dt is a tensor whose components are [/\[^{)}\ that is,
dA
dt

^KWI^HAM

(3.2.2)

Obviously, if A is a constant tensor, then its derivatives of all orders are


equal to the zero tensor.
The following identities follow from the usual rules of differentiation:
d ,
^
da
db
(a b) = dt
dt
dt
d
dt

dt

(3.2.3)
da
dt

d /
db
da ^
(a b) = a + b
dt
dt
dt

(3.2.4)
(3.2.5)

3.2

SCALAR, VECTOR AND TENSOR FUNCTIONS

111

(3.2.6)
(3.2.7)
(3.2.8)
(3.2.9)
(3.2.10)
(3.2.11)
(3.2.12)
(3.2.13)
where 0 is a real-valued (scalar) function of t, a and b are vector functions
of t, and A and B are tensor functions of t.
Next, let us consider scalar, vector and tensor functions of more than one
real variable. The independent real variables of our main interest in this text
are the three coordinates xt of a point x. Generally, point x is assumed to
vary over some region of three-dimensional space. A vector a whose components ai are real-valued functions of xt is called a vector point function
or a vector field, denoted aC*,) or a(x). Similarly, a tensor A whose components dy are real-valued functions of xt is called a tensor point function
or a tensor field, denoted A(Xj) or A(x). Also, a scalar that is a real-valued
function of x{ is called a scalar field, denoted () or ().
Scalar s, vectors and tensors arising in physical situations are generally
functions of both , and /, where / is the time variable. Such functions are
analyzed by the combined use of properties of scalar, vector and tensor
functions of / and those of scalar, vector and tensor fields.

EXAMPLE 3.2.1

Prove the identities (3.2.11) and (3.2.13).

112

CALCULUS OF TENSORS

Solution Let au(t) and bu(t) be components of A and B, respectively.


Then

I(AB)

= jt\[PM\u}
da,k b
dt kj

[rfAj
~^B

+ Oik

[dt J y

jt(aikbkj)
db,kj
d t

. dB]
+ A
dt

This proves the identity (3.2.11). Also,


d

= {[A I= {[AI

dt

^ 1

d\
dt

(dA>T
\dt

This proves the identity (3.2.13).


EXAMPLE 3.2.2
is a skew tensor.

If Q(r) is an orthogonal tensor, show that (dQ/dt)QT

Solution Since Q is an orthogonal tensor, we have QQ r = I. Differentiating both sides, we get


dQ
dt

so that

Hence

"+<%)-

@y-m-it>

showing that the tensor (dQ/dt)QT is skew.


EXAMPLE 3-2.3 If w(/) is the dual vector of a skew tensor W(0, show
that dvi/dt is the dual vector of dW/dt.

3.3

COMMA NOTATION

113

Solution The fact that dW/dt is skew when W is skew is obvious from
(3.2.13). Let a be any vector (constant or function of t). Then

idA

because w is the dual vector of W;

Hence dyt/dt is the dual vector of dVf/dt.

3.3
C O M M A NOTATION
Consider a real-valued function/ = /(JC,) that is a scalar, a component of a
vector or a component of a tensor. As stated in Section 3.2, we assume that
/ has partial derivatives of all desired orders (w.r.t. xt). We employ the
following comma notation:
dXi

dx2

dx3

or, briefly,
/.i = | ;

/ = 1,2,3

(3.3.1)

Thus, for a specified value of i( = 1, 2, 3) the symbol ( )> f denotes the partial
differential operator (/, ).
We will employ all the rules and conventions of the suffix notation to this
comma notationthe comma not interfering with these rules and conventions. In particular, the summation convention will be employed across the
comma.
Accordingly, the expression / = 1,2,3 can be suppressed in (3.3.1).
Further, an expression like (ftifti) will stand for ?=(///) Thus, for
example, the differential df of / , which is given by
df
df
df=/-dxl+/-dx2
3
dx2

df
+ 7+-dx3
dx3

114

CALCULUS OF TENSORS

may be written in a compact form as follows:


(3.3.2)

df = lkdxk

The comma notation can be extended to second-order partial derivatives


also. We employ the following notations:

d fdf\ _ d2f
dx2\dxi/

dx2dxx'

and so forth. In general,


dxi \dXjJ

dXi dxj

Since the functions dealt with are assumed to be differentiate up to the


desired order, the order of partial differentiation is taken to be interchangeable. That is,

a2/

^f

/
,U

Jl

dXidXj

- f,ll

(3.3.3)

dXjdXi

Also,
fji

a2/

a2/

a2/

dx\

bx\

a2/

+ f,22 + f,33 - T~2 ++ T~2+ + "^2

bx\

The sum (d f/dx + d f/dx\ + d f/dxj) is called the Laplacian of/, usually
denoted V2/. (The symbol V is pronounced as del or nabla). Thus, we have
f.u = V2/

(3.3.4)

the symbol ( ) u denoting the Laplacian operator V2( ).


The comma notation for third and higher order partial derivatives is
analogous. Thus,

_ay
f
7,111 - ~ 3 >
a^3'

a3/

7,123
-TT^7
- 7 ' 1 2 3 fljC3 3JC2 fljC! '

7,1123*/,U23

a4/
dX^X2dx\

and so forth.
EXAMPLE 3.3.1

Show that
imnJ,mn

^imnJ,nm

(J.J.D)

3.3

COMMA NOTATION

115

Solution The result is obvious if we note that


(interchange of dummies)

timnf.mn = ^inmf,nm
e

= ~ imnf,nm

(definition of the symbol)

= -imnftmn

(changing the order of differentiation)

EXAMPLE 3-3.2

Prove the following:

(0

xu = u\

*,,, = 3

(xmXn)j = imxn + inxm

()
OU)

(XmXn\ = imjn +

(IV)

2 ( ) = Mmn

(v)

rf. = (l/r)x if

Sinjm
(3.3.6)

r*0

(vi)

ir \u = Mu
V2(A2) = 6

(vii)

Here r2 = jcfjcf-. (7/5 notation is used throughout.)


Solution

(i) We note that


1,1

dxx

Similarly, x2t2 = *3,3 = 1. Also,

Similarly, Jt1>3 = x 3tl = Jt2>3 = x3>2 = *2,i = 0. Thus,


if / and j take the same value
if i and j take different values
Consequently, x/fl = = 3.
(ii) By the use of the product rule of differentiation, we get
\XmXn)yi

by (i).

XmtiXfi ' XmXft,i = ^mi^n + Xm^ni

116

CALCULUS OF TENSORS

(iii) Hence
\XmXn)tij

Y\xmXn)ti\j

= imiXn)j

(Xm0ni)j

OmiXnJ "*" Xm^jPni

(because the symbol is a constant)


= minj +

mjni

by (i);
= imjn + injm
(iv) Consequently, by using (3.3.4) we obtain
v\xmXn) = (xmxn),u = imin + imin = 2mn
(v) We have (r2)ti = Irr;,. On the other hand, since r2 = xmxm> we get,
from the result (ii),
(r2)ti = (xmxm),i = l*xmmi = 2xt
Hence r f = ,/, if r ^ 0.
(vi) From the result (iii), we get
v ),ij

\xmxm),ij

^im^jm

^dy

(vii) Consequently,
V2(r2) = (r2)tii = 2 = 6
EXAMPLE 3.3.3

Ufk = /*(*,), prove the following:

(0

(**/*),/ =fi + Xkfkj

(")
(**/*), = ; + fj.i + **/*.</
(337)
(iii)
v W * ) = yti*+^(V2A)
Solution (i) By the use of the product rule of differentiation, we get
(xkfk),i

xkjfk

+ xkfkJ

^kifk + xkfk,i

fi +

kfk,i

(ii) Hence
(xkfk),ij

=
=

(fi + xkfk,i)j
fij

+ ^kjfkj

fij
x

+ kfkjj

+ xkjfk,i
=

fij

kfk,ij

+ //",* +

kfk,ij

(iii) Consequently
2 (*) = (**/*). = 2/; , + **>|7 = 2,^ + **(2/*)

3.4 GRADIENT OF A SCALAR

117

3.4
GRADIENT OF A SCALAR,
DIVERGENCE AND CURL OF A VECTOR
In this section we consider three important entities that arise frequently in
practical applications. The first one is a vector field associated with a scalar
field and the other two are scalar and vector fields associated with a vector
field.
3.4.1

GRADIENT OF A SCALAR FIELD

Consider a scalar field and put , = -, in all coordinate systems. Then


,

By using the chain rule of differentiation, this can be rewritten as


'

abc* dx't

* dx'i

From the transformation rule (2.2.7) we find that


dX

Hence

,i
(3A1)

=a

a* *
<*} = 0Lik<t>ik =

cLikak

This transformation rule shows that = ^ are components of a vector.


This vector is evidently a function of xt in general; it is called the gradient
of , denoted grad , or .
Thus, if is a scalar field then V</> is a vector field with components given
by
[V4>]i = 4>.i
(3.4.2)
It is obvious that the gradient of a constant scalar field is the zero vector.
Also,
( ) =
(3.4.3)
7() = () + ()

(3.4.4)

for all scalar fields and and scalar constants a and .


By virtue of (3.4.2), expression (3.3.2) can be rewritten as
df= [Vf]kdxk = Vf-dx

(3.4.5)

118

CALCULUS OF TENSORS

3.4.2

DIRECTIONAL DERIVATIVE AND


NORMAL DERIVATIVE

From geometry, we recall that an equation of the form () s


(, 2, *3) = constant represents a surface in three-dimensional space,
and that a normal to the surface has direction ratios (3/3, 3/2,
/3). From (3.4.2) we note that these direction ratios are indeed the
components of V</>. It therefore follows that at any point x of the surface
() = constant, the vector V</> is directed along a normal to the surface.
This is illustrated in Figure 3.1.
If n is the unit vector along V</>, then
V0
(3.4.6)
V0
Let a be a unit vector inclined at an angle to the direction of . Then
n=

V0 a = ||( a) = |V0| cos

(3.4.7)

The scalar V</> a, which represents the component of along a, is called


the directional derivative of along a, usually denoted 3</>/da. Thus,
3

- a

(3.4.8)

In particular, the directional derivative of along n is called the normal


derivative of 0, denoted /dn. Thus,
dn

Figure 3.1. The vector V</>,

(3.4.9)

3.4

GRADIENT OF A SCALAR

119

From (3.4.7) and (3.4.8) it is evident that |30/3a| is maximum when 0 = 0,


that is, a = n, so that
max

3
-

(3.4.10)

Thus, the normal derivative of is the maximum of all the directional


derivatives of .
From (3.4.6) and (3.4.9) we get the useful expression
V0 = ||= M j n
3.4.3

(3.4.11)

DIVERGENCE OF A VECTOR FIELD

Consider a vector field u and put a = ukk in all coordinate systems. Then
dx'k
By using the chain rule of differentiation, this takes the form
g, =

bu'k dxn
dxn dxk

Noting that u'k = akmum and using (3.4.1) this becomes


<*' = (<XkmUm)tn<*kn

^km^kn^m,n

This transformation rule shows that a = ukk is a scalar (invariant). This


scalar is evidently a function of JC in general; it is called the divergence of
u, denoted div u, or V u.
Thus, if u is a vector field, then div u is a scalar field given by
divu = uk%k = f[u]J>it

(3.4.12)

It is obvious that the divergence of a constant vector field is equal to 0.


If u = V0, then ut = ^ and ukk = ^ = V2</>, or div u = V2</>. Thus,
we have the identity
(3.4.13)
div(V0) = 2
If v = 0u, then i>f = ui and divv = vkk = ^uk)k
u V0 + div u. Thus we have another identity,

= tkuk + uk k =

div(0u) = 0(div u) + V0 u

(3.4.14)

120

CALCULUS OF TENSORS

3.4.4

CURL OF A VECTOR FIELD

Consider again a vector field u and put at = eimnunm


systems. Then

in all coordinate

**

* = *

OJim

Using the fact that eijk are components of a third-order tensor and the
chain rule of differentiation, we get
a

ipamq(Xnrpqr

~ ,

Noting that u'n = otnkuk and using (3.4.1) this becomes


a\ =

oLipotmqocnramsankepqruktS

= <XipqSrkpqrUk,s

= ^ip^pqrUryq

<*ipap

This transformation rule shows that a, = eimnun m are components of a


vector. This vector is a function of xt in general, called the curl of u,
denoted curl u or V x u.
Thus, if u is a vector field, then curlu is also a vector field with components given by
[curl u], = imnuntm = eimn{[u]n}tm

(3.4.15)

It is obvious that the curl of a constant vector field is the zero vector.
If u = V0, then (3.4.15) gives [curl u], = = 0, see (3.3.5). Thus,
we have the identity
curlV0 = O

(3.4.16)

From (3.4.12) and (3.4.15), we get another identity:


div(curlu) = {[curlu]*},* = ekmnunymk

=0

(3.4.17)

If v = , then
[curlv], = eijkvkJ
= Zijk^UkJ

= eUk(<l>uk)j
+

Zijk$jUk

= </>[curl u], + [V</> x u],


Thus, we have one more identity:
curl(</>u) = </>curlu + x u

(3.4.18)

3.4

GRADIENT OF A SCALAR

121

From the definitions of div u and curl u , it is easy to verify that div and
curl are linear differential operators on vectors; that is,
div(au + y) = a div u + div v
curl(au + \) = a curl u + curl vj

(3.4.19)

for all vectors u and v and all scalar constants a and /?.
3.4.5

THE OPERATOR

u -V

In computations involving scalar s and vectors, we often deal with differential operators of the form Uj(d/dxj)( ), where Uj are components of a vector
u. We denote this operator by u V. That is,
u-V = W ,
'dXj

(3.4.20)

so that
(u V)</> = uj^

= uj4>j = u ()

(3.4.21)

for a scalar field , and


(u V)v = (u V^e, = (UjVijye;

(3.4.22)

for a vector field v. From (3.4.22), we note that


[(u V)v]f = ujvu = WjlMAj
3.4.6

(3.4.23)

SOME IDENTITIES

The following are some other well-known identities:


div(u X v) = v curl u - u curlv
curl(u X v) = (divv)u - (divu)v + (v V)u - (u V)v

(3.4.24)
(3.4.25)

V(u v) = v x curl u + u x curl v + (v V)u + (u V)v (3.4.26)


curl curl u = V div u - V2u

(3.4.27)

Here, the vector V u, called the Laplacian o/u, is defined by


[V2!!], = V2ui = V2[u]f

(3.4.28)

Taking v = u in (3.4.26), we get the following useful identity:


(u V)u = iV(w2) - u x curl u = \V(u2) + (curl u) x u (3.4.29)
where u2 = u u.

122

CALCULUS OF TENSORS

If = () and / = / ( ) , show that

EXAMPLE 3.4.1
(i)

/=

(ii)

/()</| = /()()

Solution (i) S i n c e / = /() and = *,), we have by the chain rule of


differentiation/^ = {/'()}^ so that V / = {/'()}(). Hence
X V/ = {/'()}() X () = 0
(ii) If we set /() = ^ ) , then we get F , = {F'()0 f i = {/()},/ so
that VF = {/()}().

EXAMPLE 3 . 4 . 2 Write down div(u x v) in the suffix notation and


hence prove the identity (3.4.24).
Solution By the expression (1.7.9), we have [u x v]^ = ekUUiVj. Hence
(3.4.12) gives
div(u x ) = {[ii x w]k)9k = ekU(UiVj\k

(3.4.30)

This is the expression for div(u x v) in the suffix notation.


Carrying out the indicated differentiation in (3.4.30), we get
div(u x v) = ekij(uitkvj

+ UiVjtk) = (ejkiui>k)Vj - (eikjVjtk)Ui

= [curlu]jVj - [curly],*/,
(by (3.4.15))
= v curl u - u curl v
EXAMPLE 3 . 4 . 3

Show that
[u x curlv] z = (Vjj - Vij)Uj

(3.4.31)

Deduce the identity (3.4.26).


Solution

By virtue of the expressions (1.7.9) and (3.4.15), we have


[u x curlv], = ^Julyfcurlv]* = eiJkekmnUjVntm

Using the - identity, this becomes


[u X curlv], = (imjn - injm)UjVntm = (VJJ -

vu)Uj

This is the required result (3.4.31).


From (3.4.31) and (3.4.23), we get
[u x curlv + (u V)v], = UjVjj

(3.4.32)

3.4

GRADIENT OF A SCALAR

123

Similarly
[v x curlu + (v V)u]f = VjUjj.

(3.4.33)

Expressions (3.4.32) and (3.4.33) yield


[u X curl v + (u V)v + v X curl u + (v V)u]f
=

UJVJJ

+ VjUjj = (UjVj)ti = (u v) t/ = [V(u v)L

(by (3.4.2)). This proves the validity of the identity (3.4.26).


EXAMPLE 3 . 4 . 4

Prove the identity:


V2(x v) = 2 div v + x V2v

Solution

(3.4.34)

By use of the relation (iii) of Example 3.3.3 we have


V2(**v*) = 2vktk + xkV2vk

When written in the direct notation, this becomes the identity (3.4.34).
EXAMPLE 3 . 4 . 5
notation:

Write down the following equation in the direct


V2Ui + (A + )ukiki + p0bi = p0i

(3.4.35)

where A, and p0 are constants.


Solution

We first recall that


V2W/ = [ 2 ],

and

ukki = (uktk)ti

= [V(divu)]f

Hence the given equation is


//[V2u]f + (A + //)[V(div u)L + p0[b]i = p0[*]i
Therefore, in the direct notation the equation reads
2 + ( + //)V(div u) + p0b = p 0 a

(3.4.35)'

EXAMPLE 3 . 4 . 6 Show that v is orthogonal to curl v if and only if


v = for some scalar functions and .
Solution First, suppose that v = . Then by using identities (3.4.16)
and (3.4.18), we get
curlv = V0 x
Hence
v curl v = ( ) = 0
Thus, v is orthogonal to curl v.

124

CALCULUS OF TENSORS

Next, suppose that v is orthogonal to curl v; that is, v curl v = 0, or


equivalently,
V\(v3t2 ~ 2, + ^2(^1,3 - ^3,1) + v3(v2tl - vh2) = 0
This implies that the total differential equation
vi dxx + v2 dx2 + v3 dx3 = 0
is integrable. Hence, there exist functions 5* 0 and such that
( dxx + v2 dx2 + v3 dx3) =
Setting = / and using (3.3.2), this can be rewritten as vkdxk =
(<t>W,k)dxk> which yields vt = ^ so that v = , as required.

EXAMPLE 3-4.7 A vector field v is said to be a Beltrami field if


curlv = for some scalar field . (Then is called the abnormality
factor of v.) Show that for such a field,
= -jr v curl v
v
_ curl v curl curl v
=

|curlv| 2

(3.4.36)
(3.4.37)

If v = + {), show that


=
Solution

v> x2772
v^

/r?^2
2

(V0) -

\)

(3.4.38)

By data, curl v = \. Hence


v curl v = v (>) = ||

This is the required result (3.4.36). Also,


curl curl v = curl(a;v) = curl v + Va> x v
by the identity (3.4.18). Hence
curl v curl curl v = co|curl v| 2 + curl v (Vco x v)
Since
curl v (Va> x v) = (Va> x v) = 0
the required result (3.4.37) follows.
If y = V</> + ^(Vx), we find, by using the identities (3.4.16) and (3.4.18),
that
curlv = x

3.5

GRADIENT OF A VECTOR

125

Hence
(V^ X ) = (V</> - ) (V^ x V*) = ( - ^) curl v
= (V0 - ) \ = ( - ^) (V</> + ^)
= ()2 - 2()2}
This is the required result (3.4.38).

3.5
GRADIENT OF A VECTOR,
DIVERGENCE AND CURL OF A TENSOR
In this section we introduce the gradient, divergence, curl and Laplacian
operators frequently encountered in tensor calculus. These are extensions of
and analogous to the operators defined in the preceding section.
3.5.1

GRADIENT OF A VECTOR FIELD

Consider a vector field u with components w, and put au = uitj in all


coordinate systems. Then
y

ax}

By using the chain rule of differentiation, this becomes


,

a li

'

du'j dxn
dxn dx'j

Noting that u[ = aimum and using (3.4.1) this becomes


l; = Ot;
MCL;Um
**ij
^im^jn^m,

n =

(X;^im^jn^mn
mCX;Clm

This transformation rule shows that ay = uitJ are components of a (secondorder) tensor. This tensor is a function of jcf in general and is called the
gradient o/u, denoted grad u or Vu.
Thus, ifu is a vector field then Vu is a tensor field with components given
by
[Vu]l7 = uiJ=i[u]i]j
(3.5.1)
The transpose of this tensor is denoted by Vur. Thus, we have
[VuT]u=uj9i

= [[u]j)9i

(3.5.2)

126

CALCULUS OF TENSORS

Also, if v is a vector, then by virtue of (2.8.19), it follows that v Vu is


a tensor whose components are given by
[vAVu]i:/ = eirsvruj>s

(3.5.3)

It is obvious that the gradient of a constant vector field is the zero tensor of
the second order. It is straightforward to verify that
(i)

Vx = I

(ii)

I Vu = Vu I = tr(Vu) = divu

(iii)

tr(v Vu) = v curl u

(3.5.4)
(3.5.5)
(3.5.6)

If u = V</>, then Vu = V(V</>) is called the second gradient of </>, denoted


. Thus
[VV<t>hj = <i>,

(3.5.7)

(VV0) r = VV0

(3.5.8)

Evidently,

3.5.2

GRADIENT OF A TENSOR FIELD

Consider a tensor field A with components au and put cijk = aijk


coordinate systems. Then

in all

By using the chain rule of differentiation, this becomes

Noting that a'u = aimajnamn

and using (3.4.1) this becomes

This transformation rule shows that cijk = a^ k are components of a thirdorder tensor. This tensor is in general a function of xi9 called the gradient
of A, denoted grad A or VA.
Thus, if A is a (second-order) tensor field, then VA is a third-order tensor
field with components given by
(3.5.9)
Obviously, the gradient of a constant second-order tensor field is the zero
tensor of order three.

3.5

GRADIENT OF A VECTOR

127

If A = Vu, then VA = V(Vu) is called the second gradient of u, denoted


VVu. Thus
[ W u ] ^ = uiJk = [[u],}^

(3.5.10)

By taking contraction in (3.5.10) we find that V tij Uj kk are components


of a vector. This vector is the Laplacian of u, denoted V2u; see (3.4.28).
3.5.3

DIVERGENCE OF A TENSOR FIELD

If ay are components of a tensor field A, it already has been seen that aijtk
are components of the tensor VA. Consequently, it follows (by contraction)
that aiJtj are components of a vector field. This vector field is called the
divergence of A, denoted div A.
Thus, if A is a second-order tensor field, then div A is a vector field with
components given by
[divAL^^-^ilA^J,,

(3.5.11)

Obviously, the divergence of a constant (second-order) tensor field is the


zero vector.
From (3.5.11) it readily follows that d i v A r is a vector field with
components
[div AT]i = ajU = l[A]}j

(3.5.12)

Setting A = Vu, we find from (3.5.11) and (3.5.12), that


[div Vu], = uUJ = V2Ui
so that
div Vu = V2u

(3.5.13)

and
[div Vu r ], = Ujjj = ujji = [Vdivu],
so that

3.5.4

d i v V u r = V(divu)

(3.5.14)

CURL OF A TENSOR FIELD

Consider again a tensor field A with components atj. Let us now put
U = imnajn,m ^n a ^ coordinate systems. Then

c'.. = e'-

UJL

128

CALCULUS OF TENSORS

Using the fact that eimn are components of a third-order tensor and the
chain rule of differentiation, this expression becomes
dxs
dxs " dx'm

-ip ^mq &nr ^pqr n y \fljn


jn/ )

Noting that a'jn = oijhankahk and using (3.4.1), we get


^ipamq^nr^jhOinkOimspqrahks

C' =
=

<Xip<Xjhqsrkpqr<thk,s

= aipajh8pqrahrg

aipaJhcph

This transformation rule shows that q, = eimnaJntm are components of a


second-order tensor. This tensor is a function of A:, in general and is called
curl of A, denoted curl A.
Thus, if A is a second-order tensor field, then curl A is also a secondorder tensor field with components given by
[curl A],, = eimnaJntm = eimn{[A]Jn]}m

(3.5.15)

It is obvious that the curl of a constant tensor field is the zero tensor.
From (3.5.15) it readily follows that curl A r and (curl A) r are tensors with
components given by
[curl % = imnanjym = ^ { [ A ] ^ ) ^
T

[(curl A) ]U = eJmnainfm = ^{[A],,,)^

(3.5.16)
(3.5.17)

Evidently, (curl A ) r ^ curl A r , in general.


If A = Vu then from (3.5.15) and (3.5.16), we find that
[curlVu],, = eimnuJtnm = 0
so that
curl Vu = 0

(3.5.18)

and
[curl Vu r ] 0 = eimnunJm

(eimHuntm)j

= {[curlu])fy = [Vcurlu],,
so that
curlVu r = Vcurlu
3.5.5

(3.5.19)

LAPLACIAN OF A TENSOR FIELD

If au are components of a tensor A, it has been noted that cijk = aijk are
components of a third-order tensor VA. It is not hard to verify that
c
ijk,m = au,km a r e components of a fourth-order tensor field. This tensor is

3.5

GRADIENT OF A VECTOR

129

called the second-gradient of A,, denoted VVA. Consequently, it follows (by


contraction) that cijkk = aiJkk are components of a second-order tensor
field. This tensor is called the Laplacian of A, denoted V2A.
Thus, if A is a second-order tensor field, then V2A is also a second-order
tensor field with components given by
[V2A]l7 = v\au)

= V2[A]U

(3.5.20)

Obviously, the Laplacian of a constant tensor field is the zero tensor. From
(3.5.20) it readily follows that
(V 2 A) r = V 2 A r

(3.5.21)

It is easy to verify that V, div, curl and V are linear differential operators
in tensor calculus also; that is,
V(au -I- \) = Vu + V\
V(aA + B) = aVA + VB
div(aA + B) = cxdiv A + div B

(3.5.22)

curl(aA + B) = a curl A + curl B


V2(c*A + B) = aV 2 A + V2B
for any vectors u, v, any tensors A and B and any constant scalars a and .
EXAMPLE 3-5.1 Let ui9 at and bt be components of vectors u, a and b,
and eu, w^ and /7 be components of tensors E, W and T, respectively. Write
down the following equations in the direct notation:
e

U = i(uu

+ u

J,i)>

U = i(uij

- uJ,i)>

TjU + p0bi = pQai

(3.5.23a, b)
(3.5.24)

Solution By data and expressions (3.5.1) and (3.5.2), equations (3.5.23a)


may be rewritten as
[E]/7 = i[Vu + V u %
In the direct notation, these read
E = |[Vu + Vu r ] = sym(Vu)

(3.5.23a)'

Similarly, equations (3.5.23b) read as follows in the direct notation:


W = |[Vu - Vu r ] = skw(Vu)

(3.5.23b)'

By the data and expression (3.5.12), equations (3.5.24) may be rewritten as


[divTr];( + A>[b], = A>M

130

CALCULUS OF TENSORS

In the direct notation, this reads


div T r + p0b = p02i

(3.5.24)'

xjx^),
EXAMPLE 3 . 5 . 2 For the vector field u = a(x^x2e1 + *f* 3 e 2 +
where a is a constant, find (i) the gradient of u, (ii) the divergences of Vu
and Vu r and (iii) the curl of Vu r .
For the given vector field, we have

Solution

(3.5.25)
These give
(3.5.26)
This is the matrix of Vu.
From (3.5.26), we get
(3.5.27)
This is the matrix of Vu r .
On using (3.5.25), expression (3.5.13) yields
[div Vu]! = V2ul = 2ax2,

[div Vu]2 = V2w2 = 2oa3,

[div Vu] 3 = V2w3 = 2axx


Thus

div Vu = V2u = 2of(jc2e1 + * 3 e 2 + x^3)

(3.5.28)

On using (3.5.25), expression (3.5.14) yields


[divVu r ]! = 2OL(X2 + 3),

[div Vu r ] 3 =

[divVu r ] 2 = 2CL(XX + x 3 ),
2OL(X2

+ xx).

Thus
div(Vu r ) =

2OL[{X2

+ x3)ei + (xl + * 3 )e 2 + (x2 + *i)e3] (3.5.29)

It may be verified that this is equal to V div u.


From (3.5.19) and (3.5.25), we find that
[curl Vu r ] 1 2 = - 2 x 2 ,

[curl Vu r ] 2 3 =

-2 3 ,

[curl Vu r ] 3 1 =2xi

3.5

GRADIENT OF A VECTOR

1 31

and
[curl ]

=0

for other values of i,j. Hence


[curl Vu r ]

(3.5.30)

This is the matrix of curlVu r . It may be verified that curlVu r =


V curl u.

EXAMPLE 3 . 5 . 3 Let u be a given vector, and A be a given tensor. For


any fixed vector c prove the following:
(Vu r )c = V(u-c)

(i)

(3.5.31)

(div A) c = div(A r c)

(ii)

(3.5.32)

(iii)

(curl A)c = curl(A c)

(3.5.33)

Solution Let au be components of A and ut and c, be the components of


u and c, respectively. Then
(i)

[(Vu r )c] f = [Vu%[c] 7 = ujticj = (ujCj)ti = (u c) pi = [V(u c)],

This proves (3.5.31).


(ii) (div A) c = [div A ] , ^ = aUtjCi = ajUCj = (aCj)ti = ([A^],),,
= div A r c
(iii) [(curlA)c]i = [curlA]0-[c]y- = (eimnaJnfm)Cj =
T

= eimn([A c]n),m

eimn(ajnCj)im

= [curKA^L-

This proves (3.5.33).


EXAMPLE 3 . 5 . 4

Prove the following identities:

(i)

(u V)v = (Vv)u

(3.5.34)

(ii)

div(0A) = 0divA +

(3.5.35)

(iii)

div(Au) = u div AT + AT - Vu

(3.5.36)

(iv)
(v)

div(curl A) = curl(div AT)


div(curl A ) r = 0

(3.5.37)
(3.5.38)

132

3 CALCULUS OF TENSORS

(vi)

tr(curlA) = 0

(3.5.39)

if A is symmetric;
(vii)
curl(curl A r ) = (curl curl A)7"
(3.5.40)
Solution (i) [(Vv)u]( = [Vv]y[u]7- = v.jiij = [(u V)y\it see (3.4.23). This
proves (3.5.34).
(ii) [div(0A)]( = {[<t>A\ij}j = (^ = ^ + au4>j. This proves (3.5.35).
(iii) [div(Au)] = {[An],),, = {auuj\t
= aUilUj

+ dijUjj

= djijUi

auu

= [u],[div A ] ( + [A ]0[Vu]y = u div AT + A7" (Vu)


(iv) [()]( = [-1](7},,
~~ ^imn^jn,mj

~~

^imn\jnj)ym

= eimn{[diwAT]nlm = [curl(div A71)],.


This proves (3.5.37).
= {[curl A]}j = jmnain^mj = 0.
(v) [div(curlA)r], = {[(curl Af]^
This proves (3.5.38).
(vi) tr(curl A) = [curl A]l7 = ^imnain%m. If A is symmetric, we have
Eimn"in,m

~~ ^inm"ni,m

^nmi^*in,m

~^imn^*in,m

so that eimnainm = 0. Hence tr(curl A) = 0, if A is symmetric.


(vii) [curl(curl A7)],, = eimni[cm\AT]jn}tm
~ ^imn^jpq^qn,pm

= eimn(ejpq[[AT]nq]tP)tm

~~

^jpq\Pimrfiqn,m),p

= eJpqi[cmlA\iq}tP = [curl(curl A)]


This proves (3.5.40).

EXAMPLE 3.5-5 Let a and b be constant vectors. For any vectors u and
v and any tensor A, prove the following:
(i)

div{(u a)Arb} = a {(Vu)Ar + u <g> div A}b


r

(3.5.41)
r

(ii) curl{(u a)A b} v = a [(v Vu) A + u <g) (curl A) v]b

(3.5.42)

Solution Let ai9 bif w, and vt be components of a, b, u and v, respectively, and au be components of A. Then,
(i)

div{(u a)Arb) = (ukakabj)ti

= akbj(ukjari + ukaJ)
T

= akbj{[(Vu)A ]kj + [u <g> div A]*,}


= ak[{(Vu)AT + u <g> div A}b]*
= a {(Vu)Ar + u (g) div A}b

3.5

(ii) curl{(u a)A r b) v =

GRADIENT OF A VECTOR

133

eirs(ukakajsbj)trVi

= ZirS<*kl>jVi(uktrajS +
= Okl^sirViUktr)ajs +
= ak{[vAVu]sk[A]js

ukajsr)

uk(irsajStr)Vi]bj
+ [uUcurlAJ^vLJ,

= ak{[(y V u ) A % + [u (g) (curl A) r v]^}6 7


= a {(v Vu) r A r + u (g) (curl A)rv}b
EXAMPLE 3 . 5 . 6
show that

If W is a skew tensor and w is the dual vector thereof,


curlW = (divw)I - Vw

(3.5.43)

Solution Let wu be components of W and wf be components of w. Then


Wij and Wi are related as follows (see (2.10.4)): wu = -eijkwk.
Hence
^imn^jntm

~ ~^imn^jnp*^p,m ~

= (ijmp

^imn^jpn^p,m

- ipmj)wpm

= ijWpp -

witj

When written in direct notation, this becomes (3.5.43).


EXAMPLE 3-5-7

Let u be a vector, E = sym Vu, W = skw Vu and w be

the dual vector of W. Prove the following:


(i)
(ii)

w = jcurlu
2

|Vu| = |E| + |W|

(3.5.44)
2

= |E| + fleuri u|

= div(Vu r )u - u V 2 u.
(iii)

(3.5.45)

Vu Vu r = |E| 2 - |W| 2 = |E| 2 - |curl u| 2


= div(u V)u - (u V)(div u)
= div{(Vu)u - (divu)u) + (divu) 2

(iv)

EW + WE = skw(Vu)2

(3.5.46)
(3.5.47)

Solution (i) Let ui9 wi9 eu and wu be components of u, w, E and W,


respectively. Then by use of (2.10.2) we get

When written in the direct notation, this becomes (3.5.44).

134

CALCULUS OF TENSORS

(ii) Consequently, by use of (2.10.8), we have


|W| 2 = ||curlu| 2

(3.5.48)

Noting that Vu = E + W, we find


|Vu|2 = Vu Vu = (E + W) (E + W) = |E| 2 + |W| 2

(3.5.49)

because E W = 0. On the other hand,


| V u | 2 = UijUij

= (UiUu)j

= il(ynT)u]j]j

UiUijj

[uU^uh

= div(Vu r )u - u V2u

(3.5.50)

Expressions (3.5.48), (3.5.49) and (3.5.50) constitute the relations (3.5.45).


(iii) Since Vu r = E - W, we get
Vu Vu r = (E + W) (E - W) = |E| 2 - |W| 2

(3.5.51)

On the other hand,


Vu V u r = UijUjj

= (UijUjXi

UjUiyU

= (u-V)u] | .} >I .-(u-V)( | . >f )


= div{(u V)u) - (u V)(div u)

(3.5.52)

Also,
Vu Vu r = uuUjj

= (uuuj - uktkUi\i + (ukykf

= {[(Vu)uL - [(Uk,k)*h)9i + ("k,k)2


= div{(Vu)u - (divu)u) + (divu) 2

(3.5.53)

Expressions (3.5.48), (3.5.51), (3.5.52) and (3.5.53) give (3.5.46).


(iv) Next,
(Vu)2 = (Vu)(Vu) = (E + W)(E + W)
= E 2 + W2 + EW + WE.

(3.5.54)

[(Vu) 2 ] r = E 2 + W2 - (EW + WE)

(3.5.55)

Hence
Expressions (3.5.54) and (3.5.55) yield
(Vu)2 - [(Vu)T = 2(EW -h WE)
from which (3.5.47) is immediate.

(3.5.56)

3.5

EXAMPLE 3-5.8

GRADIENT OF A VECTOR

135

Show that for any tensor A,

curl curl A = [V2(tr A) - div(div A)]I + V div AT


+ (Vdiv A)T - VV(tr A) - V 2 A r

(3.5.57)

Deduce that, if A is symmetric, then the equation


curl curl A = 0

(3.5.58)

V2A + VV(tr A) - V(div A) - (V div A ) r = 0

(3.5.59)

is equivalent to the equation

Solution

If Ojj are components of A, we note that


[curl curl A ] 0 =

(3.5.60)

eirsejmnasntmr

By using the identity (1.7.22) this becomes


[curl curl A],, = (aSfttmr)

ij

Sim in

rj

rm ^rn

(3.5.61)

On expanding the determinant on the righthand side, this simplifies to


[curl curl A] l7 = u(atmm

- amryVnr) + ajrJr + amitmj - ajUrr - aSSfU

= {V (tr A) - div (div A ) ^ + [(V div A ) %


+ [V(div A % - [ V 2 A % - [VV(tr A)] l7

(3.5.62)

This is precisely the identity (3.5.57).


If A is symmetric, the identity (3.5.57) reads as follows in the suffix
notation:
^irs^jmn^sn,mr

~ \y*ss,mm

**mrymr)^ij

+ <*jm,im -

ij,mm

" ^im,mj

~ <*mmyij

(3.5.63)

Now, suppose that (3.5.58) holds; then we have


^irs tjmn &sn, mr

Consequently, eirseimnasnmr

(3.5.64)

= 0, which, on using the - identity, reduces to


asS,mm

- amr,mr

= 0

(3.5.65)

In view of (3.5.64) and (3.5.65), identity (3.5.63) yields


<lij,mm + <*mm,ij ~ <*im,mj ~ <*jm,im = 0

(3.5.66)

136

3 CALCULUS OF TENSORS

When written in the direct notation, these equations become the tensor
equation (3.5.59). Thus equation (3.5.58) implies equation (3.5.59).
Conversely, suppose that the tensor equation (3.5.59) holds. Then
(3.5.66) hold. Taking the contraction of (3.5.66), we find that (3.5.65)
holds. Since (3.5.65) and (3.5.66) hold, identity (3.5.63) yields (3.5.64). This
means that the tensor equation (3.5.58) holds.
Thus, equations (3.5.58) and (3.5.59) imply each other; the equations are
therefore equivalent.

3.6
INTEGRAL THEOREMS FOR VECTORS
Here we state the divergence theorem and Stokes's theorem in vector
integral calculus, with which the reader should be already familiar. Some
consequences of these theorems are also summarized.
3.6.1

DIVERGENCE THEOREM

Let V be the volume of a three-dimensional region bounded by a closed


regular surface S; see Figure 3.2. Then for a vector field u defined in V
and on 5,
(divu)rfK=

(u-n)rfS

(3.6.1)

where n is the unit outward normal to S. In the suffix notation, the expression (3.6.1) reads:
\uktkdV=

\uknkdS

^J-Surface element dS
Volume V

Figure 3.2.

/.
^ Surface S

Regular region.

(3.6.1)'

3.6

INTEGRAL THEOREMS FOR VECTORS

1 37

The following are some immediate consequences of expression (3.6.1):


(i)

]v

(ii)

=\

]s

(curlu)rfF=
]V

0nt/5,

or

(nxu)tfS

=\

}v '

or

}V

}S

(V2u)</K =

(iv)
]V

<t>nkdS

(3.6.2)

I 6ijkuk jdV = I eijknjUkdS


Js
(3.6.3)

]V

{V2<j>)dV= (n-V)^rfS or

(iii)

Js

<t>,kkdV = \ nk<t>,kdS
]V

(n-V)urfS or
h

u,tkkdV=
]V

Js

(3.6.4)

nkui<kdS
(3.6.5)

]S

In (3.6.2) and (3.6.4), is a scalar field defined in Fand on S.


3.6.2

SOLENOIDAL VECTORS

The surface integral J s u n dS is usually called the outward normal flux or


just//wjc of u across S. A vector is said to be solenoidal in a region if its flux
across every closed regular surface in the region is 0. From the divergence
theorem, it follows that u is solenoidal in a simply connected region if and
only if div u = 0 in that region.
A vector field whose divergence is exactly 0 is called a divergence-free
vector. From the statement just made, it follows that a vector field is
solenoidal in a simply connected region if and only if it is divergence free.
Since div(curlu) = 0, see (3.4.17), therefore curlu is a divergence-free
vector for every vector u. It can be proven that every divergence-free vector
u defined in a simply connected region can be represented as
u = curl w

(3.6.6)

where w is itself a divergence-free vector. Here w is called a vector potential


of u.
3.6.3

STOKES S THEOREM

Let C be a simple closed curve in three-dimensional space and S be an open


regular surface bounded by C; see Figure 3.3. Then for a vector field u
defined on S as well as C,
c

u t ds =

Js

(curl u) n dS

(3.6.7)

138

CALCULUS OF TENSORS

Figure 3.3.

Open regular surface.

where t is a unit tangent vector to C that is assumed to be positively oriented


relative to the unit normal n to S.
In the suffix notation, expression (3.6.7) reads
uitids=\
eijkukjnidS
(3.6.7)'
c
Js
If S is a closed surface, the lefthand side of (3.6.7) reduces to 0, then
(3.6.8)
(curl u) n dS = 0 or
eUkukJnidS = 0
s
Js
This expression also follows from the divergence theorem applied to curl u.
The particular case of (3.6.7), where C lies entirely in thex l jc2-plane and S
is the part of the plane bounded by C, is important in its own right. In this
case, expression (3.6.7) reduces to the form
(3.6.9)
(i/j dxx + u2dx2) = I (w2,l _ ult2)dxidx2
c
Js
where ux and u2 are the xx and .^-components of u. This particular case of
Stokes's theorem is usually referred to as Green's theorem in the plane,
Following are some immediate consequences of the expression (3.6.7):
(i) <l>tds = \ nxV<l)dS
Jt

JS

or

(D 0^ife=

JC

JS

eUknj<l>tkdS (3.6.10)

3.6

()

(u x t) ds =\ [(div u)n - (Vu)rn] dS


ic
Js
c

(iii)

INTEGRAL THEOREMS FOR VECTORS

zukujtkds

curl w ids =
CijkUkJids --

Js

[Uk^rii -

Js dn

139

or
(3.6.11)

ukJnk]dS

(div u) - n (V2u) dS

or

d_
("*,*) - "iHkk dS
Js dn

(3.6.12)

In (3.6.10), is a scalar field defined on S as well as C.


Note: The arc element t ds is often denoted dx. Consequently, we write
lc()dx for IcOtds.
3.6.4

CONSERVATIVE AND IRROTATIONAL VECTORS

The line integral in the lefthand side of expression (3.6.7), namely,


c u t ds, which represents the integral of u t around C, is called the
circulation of u round C. A vector u defined in a region is said to be
conservative if its circulation round every simple closed curve in the region
is 0, or equivalently, if the value of the line integral {* u t ds, defined in the
region, depends only on the endpoints A and B and not on the curve joining
A and B on which integration is carried out. Also, a vector is said to be
irrotational (or curl free) in a region if its curl is exactly 0 in the region.
From Stokes's theorem, it follows that, in a simply connected region, a
vector is conservative if and only if it is irrotational in the region.
Since curl V</> = 0, see (3.4.16), therefore V</> is an irrotational vector for
every scalar field . It can be proven that every irrotational vector u defined
in a simply connected region can be represented
u =

(3.6.13)

Then is called a scalar potential of u.


If a vector u is both divergence free and irrotational, then the identity
(3.4.27) shows that V2u = 0; in this case we say that u is a harmonic vector.
3.6.5

HELMHOLTZ'S REPRESENTATION

It has been noted that a divergence-free vector has a representation as


given by (3.6.6) and an irrotational vector has a representation as given by
(3.6.13). A representation valid for a general vector, known as the
Helmholtz's representation follows.

140

CALCULUS OF TENSORS

Consider a vector field u. Define the vector field v by


V(X) = - - L [ r^rdV
4 ]v\\ - \\

(3.6.14)

where V is some volume in the region of definition of u and the integral is


taken by varying x over F, keeping x as a fixed point. It can be proven that
V2v = u. Then by using the identity (3.4.27) we get
u

= v</> + curlw

(3.6.15)

where
= divv,

w = -curlv

(3.6.16)

Thus, given a vector field u, there exist a scalar field and a vector field
w, defined by (3.6.16) and (3.6.14), such that u has a representation as given
by (3.6.15). This is the Helmholtz's representation. Note that the vector w
present in this representation is divergence free.
Prove expressions (3.6.3) and (3.6.11).

EXAMPLE 3.6.1

Solution Let a be an arbitrary constant vector field. Employing the


divergence theorem to the vector a x u, we get

div(a xu)dV=

Js

(axu)-nrfS

(3.6.17)

Using identities (3.4.24) and (1.4.18), expression (3.6.17) becomes


a \\

curl udV -

nxurfS} = 0

Since a is arbitrary, this yields (3.6.3).


On the other hand, if we employ Stokes's theorem to the vector a x u, we
get
c

(a x u) ids =

Js

curl(a x u) n dS

(3.6.18)

Using identity (1.4.18) and


[curl(a x u)] n = [(div u)a - (a V)u] n = [(div u)a - (Vu)a] n
= a [(div u)n - (Vu) r n]
which is obtained by using (3.4.25), (3.5.34) and (2.8.14), expression

3.6

INTEGRAL THEOREMS FOR VECTORS

141

(3.6.18) becomes
a I u x t ds -

(div u - Vur)n dS = 0

Since a is arbitrary, this yields (3.6.11).


EXAMPLE 3.6.2

Show that

n x curl u = (Vu)rn - (n V)u

(3.6.19)

Hence, deduce the following alternative version of (3.6.11):


u x t ds =\ [(div u)n - (n V)u - n x curl u] dS
Jc
Js
Solution

(3.6.20)

By use of (3.4.31), (3.5.2) and (3.4.23), we get


[n x curlu], = (ukti - uitk)nk = [(Vu)rn - (n V)u]f

from which (3.6.19) is immediate.


Substituting for (Vu)rn from the identity (3.6.19) in the expression
(3.6.11), we readily get the expression (3.6.20).

EXAMPLE 3.6.3 If A is a tensor field such that div A r = 0 in a simply


connected region, show that A = curl W for some tensor W.
Solution As usual, let e, be any one of the base vectors of a fixed
coordinate system. Then
see (3.5.32);

div(Ae/) = e# div AT

=0
by data. Hence, by virtue of (3.6.6), Ae, can be represented as
Aef = curlw,

(3.6.21)

for some divergence-free vector w,. Let W = e* (x) w^. Then


W^e, = (w* ek)et = (ek e^w* =

W | .

(3.6.22)

Hence (3.6.21) and (3.6.22) give


Ae, = curlw, = cur^W 7 ^) = (curlWJe,
see (3.5.33). By virtue of the result proven in Example 2.7.2, it follows that
A = curl W.

142

3 CALCULUS OF TENSORS

EXAMPLE 3 . 6 . 4 If A is a tensor field such that curl A = 0 in a simply


connected region, show that A = Vu for some vector u. If, further,
tr A = 0, show that A = curl W for some skew tensor W.
Solution As usual, let ef be any one of the base vectors of a fixed coordinate system. Then,
cm\(ATei) = (curlAJe,
see (3.5.33);
=0
by data. Hence by virtue of (3.6.13), there exists a scalar field </>, such that
, = ,

(3.6.23)

Let u = <t>kek. Then u e, = </>, so that, by using (3.5.31),


V0f = V(u e,) = (Vu^e,

(3.6.24)

Hence (3.6.23) and (3.6.24) give A 7 ^ = (Vu r )e,. By virtue of the result
proven in Example 2.7.2, it follows that AT = Vu r or A = Vu. Consequently, tr A = tr Vu = div u. Hence, if tr A = 0, then div u = 0.
Let W be the skew tensor of which - u is the dual vector. Then, from
(3.5.43), it follows that
curlW = div(-u)I - V(-u)
Since div u = 0, we get
curl W = Vu = A

EXAMPLE 3 . 6 . 5 Let u be a vector field defined on and inside a closed


regular surface S enclosing a volume V, and E = sym Vu and W = skw Vu.
If u = 0 on 5, prove the following:
|W| 2 rfK<

(i)

\E\2dV

(3.6.25)

(ii) Korn's inequality:


\Vu\2dV<2

(3.6.26)

Sv

Solution

\E\2dV

(i) From (3.5.46), we get

{|E|2 - |W| 2 }rfK=

div{(Vu)u - (divu)u)dV +

(divu) 2 rfK

3.7

INTEGRAL THEOREMS FOR TENSORS

143

Employing the divergence theorem to the first of the integrals on the


righthand side, and recalling that (Vu) u = (u V)u, we get
[|E|2 - |W|2} dV =\ {(u V)(u n) - (div u)(u n)} dS +
]V

]S

]V

(div u)2 dV
(3.6.27)

Since u = 0 on S by data, this yields


|E| 2 rfK-

\W\2dV=

(divu) 2 rfF>0

from which (3.6.25) is immediate.


(ii) From (3.5.46) we also get
|Vu| 2 rfF=

|E| 2 </F+

\W\2dV

(3.6.28)

Using the inequality (3.6.25), expression (3.6.28) yields


v

|Vu| 2 rfF<

]v

\E\2dV+

]v

\E\2dV

which is (3.6.26).

3.7
INTEGRAL THEOREMS FOR TENSORS
We now obtain some extensions to tensor fields of the divergence theorem
and Stokes's theorem stated in Section 3.6. As in the case of a vector field,
an integral of a given tensor field is defined as a tensor field whose
components are the integrals of the components of the given field.
THEOREM 3.7.1 (Divergence Theorem for a Tensor)

Let V be the

volume of a three-dimensional region bounded by a closed regular surface


S. Then for a tensor field A defined in V and on 5,
divAdV=
An dS
v
is
where n is, as usual, the unit outward normal to 5.

(3 JA)

144

3 CALCULUS OF TENSORS

Proof

Let c be an arbitrary constant vector. Then


c

c-(divA)rfK

di\AdV=

see (3.5.32);

so that
J

-I.

div(A'c)dV

(ATc)-ndS

Js

divArfK-

]s

c-(An)rfS

AnrfS = 0

Since c is an arbitrary vector, this expression yields the result (3.7.1).


In the suffix notation, expression (3.7.1) reads
\aikikdV=

\aiknkdS

(3.7.1)'

Consequences
(i) If A = (or ay = <t>bij) where is a scalar, then (3.7.1) and (3.7.1)'
become, respectively,

\()
\()= =

[<**!*). dV=
k

(3.7.2)

\^biknkdS

(3.7.2)'

(ii) If A = , expression (3.7.1) becomes, on noting that div(</>I) = V<,


see (3.5.35),

1,"-
=

(3.7.3)

This is identical with expression (3.6.2).


THEOREM 3.7.2 Let V be the volume of a three-dimensional region
bounded by a closed regular surface S. Then for a vector field u defined in

3.7

V and on S,

INTEGRAL THEOREMS FOR TENSORS

lndr-\.
VndV=

145

(3.7.4)

nndS

where n is, as usual, the unit outward normal to S.


Let a and c be arbitrary constant vectors. Then

Proof
c

e-(Vu)atfK =

(Vu)rfFa =

*-(Vu)TcdV

V(c u) dV = a

= a

(c u)n dS

by (3.7.3);
=

c (n <g) u)ra dS

a (n u)c dS =
is

is

. . .

u (x) n)a dS = c ] I u nrfSJa

Since c and a are arbitrary, this expression yields the result (3.7.4).
In the suffix notation, (3.7.4) reads
V

uudV=
'

JS

(3.7.4)'

UiHjdS

Note: If we take the trace on both sides of (3.7.4), we recover the


divergence theorem (3.6.1).
THEOREM 3.7.3 (Stokes s Theorem for a Tensor)

Let C be a

simple closed curve in three-dimensional space and S be an open regular


surface bounded by C. Then for a vector field defined on 5 as well as C,

At ds=
(curlA)WS
(3.7.5)
c
is
where t is the unit tangent to C, which is assumed to be positively oriented
relative to the unit normal n to S.
Proof Let a be an arbitrary constant vector. Then
a

is

(curl A)rn dS =\

is

a (curl A) rn dS =\
n-curl(A r a)f5.

is

n (curl A)a dS
(3.7.6)

146

CALCULUS OF TENSORS

By virtue of expression (3.6.7), we have


s

n curl(A r a) dS =
=

Jc

Jc

(A r a) t ds
a (At) ds = a

Jc

At ds

(3.7.7)

Since a is arbitrary, expressions (3.7.6) and (3.7.7) yield the result


(3.7.5).

In the suffix notation, expression (3.7.5) reads


aiktkds=\
eJrsaiStrnjdS
c
Js
Note: In the special case when S is closed, we have by (3.6.8),
n curl(A r a) dS = 0

(3.7.5)'

(3.7.8)

Then (3.7.5) gives

(curl A) W S = 0

or

Js

ejrsair>snjdS

=0

(3.7.9)

A Particular Case In the particular case when C lies in the jc^-plane


and S is the part of this plane bounded by C, we have nx = n2 = t3 = 0 and
n3 = 1. Then expression (3.7.5)' takes the following form:
(aildxl + ai2dx2) = (ai2il - aiU2)dS
(3.7.10)
c
Js
This is a generalization of the Green's theorem in the plane, given by
(3.6.9).
THEOREM 3 . 7 . 4 Let C be a simple closed curve in three-dimensional
space and S be an open regular surface bounded by C. Then for a vector
field u defined on S as well as C,

t(g>uds=

js

(nAVu)dS

(3.7.11)

where t is the unit tangent to C, which is assumed to be positively oriented


relative to the unit normal n to S.

3.7

Proof

INTEGRAL THEOREMS FOR TENSORS

147

Let a be a constant vector. Then


(D t uds a = ( t ( x ) u ) a * = (u-a)tfr

Is

by (3.6.10);
=

n x V(u a) dS

n x (Vu)'a dS = \ (n Vu)a i/S

by (2.8.17);
(nAVu)rfS[a
Since a is arbitrary, expression (3.7.11) follows.
In the suffix notation, expression (3.7.11) reads
tiUjds = \ eirsnrujsdS
(3.7.11)'
c
}s
If we take the trace of expression (3.7.11), we recover Stokes's theorem
(3.6.7).
EXAMPLE 3.7.1 Let 5 be a regular surface enclosing a region of
volume V. For a vector field u and a tensor field A defined on V and on 5,
show that
(u divA + (Vu)AT}dV=

v
Js
Hence deduce expressions (3.7.1) and (3.7.4).

u (An) dS

(3.7.12)

Solution Let a and b be arbitrary constant vectors. Employing the


divergence theorem (3.6.1) to the vector (u a)(A r b), we get
div((u a)(Arb)} dV =\

[(u a)(A r b)j n dS

(3.7.13)

Using identities (2.8.16) and (3.5.41), expression (3.7.13) becomes


a S

[(Vu)A r + u <g> div A] dV b = a M [u (x) An] dS b

Since a and b are arbitrary, this expression yields (3.7.12).

148

CALCULUS OF TENSORS

If u is a constant vector, then (3.7.12) becomes


u (x)

div A dV = u

An dS

Since this result is true for any u, we obtain (3.7.1). On the other hand, if
A is the identity tensor I, then (3.7.12) becomes (3.7.4).

EXAMPLE 3 . 7 . 2
(3.6.3).
Solution

Starting with expression (3.7.4) deduce expression

For any constant vector c, expression (3.7.4) yields

(Wu-VuJ)cdV=\

Js

( u n - nu)c</S

(3.7.14)

By virtue of (3.5.44), the dual vector of (Vu - Vu r ) is curl u. Hence


(Vu - Vu r )c = (curlu) x c

(3.7.15)

[u n - n (x) u]c = (n c)u - (u c)n = c x (u x n)

(3.7.16)

Also,
Using (3.7.15) and (3.7.16), expression (3.7.14) becomes
c x

curl u f i f F = c x

(n x u) dS

Since c is arbitrary, expression (3.6.3) follows.

EXAMPLE 3 . 7 . 3 Let S be a regular surface enclosing a region of


volume V. For a tensor field A defined in V and on 5, show that

u x (An) dS =

]v

(2w + u x divA)rfK

(3.7.17)

Here n is the unit outward normal to S and w is the dual vector of the skew
part of A(Vu) r . Deduce that
x x (An) dS = \ (2 + x x div A) dV
where is the dual vector of skw A.

(3.7.18)

3.7

Solution

INTEGRAL THEOREMS FOR TENSORS

149

Let a be an arbitrary constant vector. Then


a x \ u x (An) dS = \ [a x [u x (An)}] dS

(u (An) - (An) <g> u}arfS

(3.7.19)

{u (g) div A + (Vu)A'} dV

(3.7.20)

((div A) u + A(Vu)'} </F

(3.7.21)

From expression (3.7.12), we have


[u <g) (An)} dS =
so that (on taking the transpose)
{(An) <x) u) dS =

Using (3.7.20) and (3.7.21) in (3.7.19), we get


ax

u x (An)dS
[{u (div A) - (div A) <g> u}a - 2 skw{A(Vu)r}a] dV

= a x

{u x div A + 2w) dS

Since a is arbitrary, (3.7.17) follows.


If we set u = x in (3.7.17) and recall that Vx = I, we readily get
(3.7.18).

EXAMPLE 3 - 7 . 4 Let S be a regular open surface bounded by a simple


closed curve C. For a vector field u and a tensor field A defined on S as well
as C, show that
c

(u (x) At) ds =

}s

[u <g> (curl A) 7 n + (n Vu)'A 1 ] dS

(3.7.22)

where t and n are as defined in expression (3.7.5). Hence deduce (3.7.5) and
(3.7.11).
Solution Let a and b be arbitrary constant vectors. Employing the
Stokes's theorem (3.6.7) to the vector (u. a)(A r b), we get
c

{(u a)(Arb)} t ds =\

Js

curl{(u a)(A r b)} n dS

(3.7.23)

150

CALCULUS OF TENSORS

Using the identities (2.8.16) and (3.5.42), expression (3.7.23) becomes


(u <g> At) ds b = a

[u (curl ) + ( Vu)rAr] dS b

Since a and b are arbitrary, this expression yields (3.7.22).


If u is a constant vector, expression (3.7.22) becomes
u (D Aids = u <g) I (curlAyndS
jc

js

Since this result is true for any u, expression (3.7.5) follows.


On the other hand, if A is the unit tensor, expression (3.7.22) becomes

(ut)ds =

from which (3.7.11) is immediate.

}s

(nAVufdS

EXAMPLE 3.7.5 Let S be a regular open surface bounded by a simple


closed curve C. For a tensor field A defined on S and C, show that
(u X At) ds =

[2w + u X (curl A)rnj dS

(3.7.24)

where t and n are as defined in (3.7.5) and w is the dual of the skew part of
A(n Vu). Hence deduce that
(x x At)ds = \ (2i + x x (curlA)'n)dS
c
Js
where is the dual vector of skw(n A).
Solution Let a be an arbitrary constant vector. Then

(3.7.25)

a x ( D ( u x A t ) ^ = ( p ( u (At) - (At) u}a ds


Using (3.7.22) and its transpose, this expression takes the form
a x (u x At) ds
= \ [{u (x) (curl A)7!! - (curl A)rn <g) u}a - 2{skw A(n Vu)}a] dS
=ax

[u x (curl A)rn + 2w} dS

Since a is an arbitrary vector, result (3.7.24) follows.

3.8

EXERCISES

151

Setting u = x in (3.7.24) and recalling that Vx = I and () =


-(n A)T, we get (3.7.25).

3-8
EXERCISES
1- Verify the identities (3.2.3) to (3.2.7).
2 . For any vector u = u(t) with magnitude u, show that
du
du
u -- = u
dt
dt
3 . If u(/) is a unit vector, show that
du

du

ux
=
dt
dt
4. If u(0 and v(/) are such that du/dt = w x u and dy/dt = w x v, where w is a
constant vector, show that
(u x v) = (u (x) v - v (x) u)w
dt
5. Verify the identities (3.2.8) to (3.2.10) and (3.2.12).
6. If Q(0 is an orthogonal tensor, show that Q(dQ/dt)T
tensors.

and QT(dQ/dt)

are skew

7. If A(/) is an invertible tensor, show that


^(detA) = ( d e t A ) t r ^ A - M
8. If ui = 2xlx2, u2 = -*2*3> u3 = 3x 1 x 3 , find the matrices of the following:
(i) uitj
(iii) (uu

(ii) UJJ
+ Ujti)

(iv) (Uij -

ujti)

9. Verify the identities (3.4.3) and (3.4.4).


1 0 . Show that the directional derivatives of along e, are ,.
1 1 . Find the directional derivative of = *2*3 along the tangent to the curve
given by the parametric equations = t,x2 = t2, x3 = t3, at the points (1, 1, 1) and
(-1,1,-1).
1 2 . Find the direction along which the directional derivative of the function
= X\X2Xi is maximum at the point (1, 1, 1). Also, find the maximum directional
derivative.

152

CALCULUS OF TENSORS

1 3 . If = xnx + x" + x", show that x V</> = .


1 4 . I f / = /(r), show that Vf = l/'(r)/r)x, r ^ 0.
1 5 . If A is a constant tensor, show that V(Ax x) = (A + A r )x.
1 6 . Verify the identities (3.4.19).
1 7 . Show that
(i) div u = ek u k

(ii) curl u = ek x u

1 8 . Prove the identities (3.4.25) and (3.4.27) by using the suffix notation.
1 9 . If u = curl v and v = curl u, show that
(i) div(curl u x curl v) = u2 - v2
(ii) V2u = - u ,

V2v = - v

2 0 . Show that V2|x|" = 0 only if n = 0 or - 1 .


2 1 . Find = (), given that div(0x) = 0.
2 2 . Find = (), given that V2(r</>) = 0.
2 3 . If / = /(r), show that V2{/(r)) = f\r)
V 2 (l/r) = 0.
2 4 . Show that \)

+ (2/r)/'(r), r * 0. Deduce that

= 2 + 2 + 2V0 .

2 5 . Show that (u V)x = u.


2 6 . Show that curl u = y[curl((curl u) x x) - (x V) curl u].
2 7 . Show that rn\ is curl free for any n, but divergence free only if n = - 3 .
2 8 . If c is a constant vector, show that /(r)c x x is divergence free.
2 9 . If u curl v = v curl u, show that u x v is divergence free and that
curl curl(u x v) = V2(v x u)
3 0 . Show that V(v V(l/r)) + curl(v x V(l/r)) = 0, r * 0.
3 1 . If the curl of a Beltrami vector is again a Beltrami vector, show that the
abnormality factor is independent of x.
3 2 . A vector u is said to be a complex lamellar vector if u curl u = 0. If u is a
complex lamellar vector, show that for any scalar field 9 the vector is also a
complex lamellar vector.
3 3 . Show that Vu = urstr e s .
3 4 . If Vu = -(Vu) r , show that VVu = 0.
3 5 . Show that (v Vu)w = v x (Vu)rw.
3 6 . Show that V(</>u) = u V</> + 0Vu.

3.8 EXERCISES

153

3 7 . Verify the identities (3.5.4) to (3.5.6) and (3.5.22).


3 8 . If u = x]tx + x\t2 + jc|e 3 , find Vu and (Vu)u.
3 9 . If c is a constant vector, show that (V2A)c = V2(Ac).
4 0 . Prove the following identities:
(i) div(div A) = tr V(div A)
(ii) div(Vur) = div[(div u)I]
(Hi) div(0I) = V0
(iv) div(Au) = u div AT + tr[Ar(Vu)]
(v) div(u v) = (div v)u + (Vu)v
(vi) div[0(Au)] = 0((u div A 7 ) + (A r Vu)j + V0 Au
(vii) div[(Vu)u] = Vu Vu r + u V(div u)
(viii) div[(Vu)u - (div u)u] = Vu Vu r - (div u)2
(ix) div[(u <g) v)w] = (div u)(v w) + u V(v w)
(x) curl(0I) = -[curl>I)] r
(xi) tr[V2A + VV(tr A)] = 2V2(tr A)
4 1 . L e t / be a scalar field or a component of a vector or a tensor field defined in
a region R. If \vf dV = 0 for every volume V in /?, show t h a t / = 0 in R.
4 2 . Calculate the circulation of u = x\tx - Jt1Jt2e2 round the ellipse xl = a cos ,
x2 = b sin , x 3 = 0, 0 < <2.
4 3 . By using the divergence theorem or Stokes's theorem, prove the expressions
(3.6.2), (3.6.4), (3.6.5), (3.6.9), (3.6.10) and (3.6.12).
4 4 . By using the divergence theorem, prove the following:
(i) | <t>-!-dS=
s on

(ii)

]v

U^-w^\dS=\

W>V> +

(0VV - 2)

4 5 . If u is everywhere normal to a closed regular surface S bounding V, show that


v

curlufK= 0

4 6 . If v is as defined by (3.6.14), prove that V2v = u.


4 7 . Verify that the vector w present in Helmholtz's representation (3.6.15) obeys
the equation
1 curl u(x)
.
,dV
An

1 54

CALCULUS OF TENSORS

48. Deduce representations (3.6.6) and (3.6.13) from Helmholtz's representation


(3.6.15).
49. By starting with the expression (3.7.17), deduce expression (3.6.3).
50. By starting with the expression (3.7.25), derive expression (3.6.12).
51. If Fis the volume of a region bounded by a closed regular surface S, show that
s

x <g) n dS = V\

52. Prove that


v

aijtidV=

Js

dnias

53. Let u be a vector field defined in a region of volume V enclosed by a regular


closed surface S. Also, let a be a constant vector and = (3) is a scalar function
defined in V. If u is divergence free in V and u = 0 on 5, prove the following:
|curlu|2</F

(i)

(u -curl curl u) d V =\

(ii)

(u x a) V(a curl u)rfK= -

(Hi)

(u a)V2(u a) dV = -

(iv)

W> > 3 u)-V W 3 rfK=-

|a-curlu|2fK

|V(u a)|2 dV
,333 dV

54. Let h be a vector field defined in a region of volume V enclosed by a regular


closed surface S. If h is divergence free in V and n x curl h = 0 on 5, show that
v

h curl curl h dV =

]v

|curlh|2rfF

55. Show that


[(curl u) x v + (curl v) x u + u(div v) + v(div u)] dV
= l [(u (g) v + v (x) u)n - (u v)n] dS

CHAPTER 4

CONTINUUM HYPOTHESIS

4.1
INTRODUCTION
The subject of mechanics is concerned with the study of external effects,
such as forces and heat inputs, on a physical object. If the object is a solid
whose volume or shape or both generally change, the subject is called solid
mechanics. If the object is a fluid (liquid or gas), the subject is referred to
as fluid mechanics. Historically, both solid and fluid mechanics were
developed almost simultaneously. The foundations of these subjects were
laid during the latter half of the eighteenth and the first half of the nineteenth centuries by celebrated mathematicians including Leonhard Euler
(1707-1783), Augustin Louis Cauchy (1789-1857), Simeon Denis Poisson
(1781-1840), George Green (1793-1841) and George Stokes (1819-1903).
An examination of these foundations reveals that the basic postulates and
general principles upon which both solid and fluid mechanics are based are
indeed the same. The mathematical equations that describe the physical
laws, called the field equations, are common to both the subjects. But the
two subjects are not identical. Solids and fluids have individual characteristic properties, and these properties are reflected in the study of the
155

156

CONTINUUM HYPOTHESIS

respective subjects. Equations that represent the characteristic properties of


a material (or a class of materials) and distinguish one material from the
other are called constitutive equations. Obviously, solids and fluids cannot
have the same constitutive equations. Hence, although both solid and fluid
mechanics share the same field equations, the two subjects differ in
constitutive equations.
The study of solid and fluid mechanics through a unified approach
constitutes the subject of continuum mechanics. This subject can be divided
into three natural parts. The first part deals with the field equations
common to both solid and fluid mechanics. The second part is concerned
with the constitutive equations that are different for different types of solids
and fluids. The third part deals with different branches of the subject whose
governing equations are obtained by combining the field equations with
appropriate constitutive equations.
The whole theory of continuum mechanics rests on a fundamental
hypothesis known as the continuum hypothesis. The hypothesis is explained
in the following sections of this chapter. The field equations of the theory
are formulated on the basis of the fundamental laws of physics: the laws of
conservation of mass, momentum and energy. These equations are derived
in Chapter 8. The basic kinematic concepts of deformation, strain and
motion as well as the concept of stress that pave the way for the derivation
of field equations are presented in Chapters 5 to 7. Historically, the concepts of strain and stress were introduced and discussed by Cauchy during
the years 1823-1827. A large part of the theories of strain and stress, as is
known today, has stemmed from his pioneering work. The development of
kinematics of motion and field equations are essentially due to Euler.
Constitutive equations are formulated in two different ways. One way is
to postulate their basic forms on experimental grounds. Hooke's law, which
describes the behavior of elastic solids, and the Newton's law of viscosity,
which describes the behaviour of viscous fluids, are two standard examples
of constitutive equations that were first formulated in this way. The other
way is through a rigorous axiomatic approach. In this text, which is meant
primarily for the beginners in solid and fluid mechanics, we refrain from
following the axiomatic approach. We confine ourselves to dealing with
only three important examples of constitutive equations applicable to perfectly elastic solids, perfectly non viscous fluids and the so-called Newtonian
viscous fluids. When combined with the field equations, these particular
constitutive equations give rise to the governing equations of the linear
elasticity theory and the theories of nonviscous and Newtonian viscous fluid
flows. The equations of the linear elasticity theory and their immediate
consequences and simple applications are presented in Chapter 9. Chapter
10 is concerned with the equations of nonviscous and Newtonian viscous

4.3

CONFIGURATION OF A CONTINUUM

157

fluid flows and their immediate consequences and simple applications.


These two chapters exhibit the distinct advantages of having a unified basis
for solid and fluid mechanics with common notation and terminology. They
also prepare the reader for advanced studies in either of the two subjects.

4.2
NOTION OF A CONTINUUM
It is a common knowledge that every physical object is made up of molecules,
atoms and even smaller particles. These particles are not continuously
distributed over the object; microscopic observations reveal the presence of
gaps (empty spaces) between particles. While studying the external effects
on physical objects, these gaps may or may not be taken into consideration
depending on the hypothesis made. The study that takes account of the
existence of gaps is called microscopic study. On the other hand, the study
that ignores the gaps and treats a physical object as a continuous distribution of matter is called macroscopic study. The subjects of solid and fluid
mechanics are concerned mainly with macroscopic study. Although microscopic study is supposed to yield more accurate results than macroscopic
study, at least in principle, macroscopic description of mechanical behavior
of materials is adequate and useful for common engineering applications.
The macroscopic viewpoint adopted in the study of solid and fluid
mechanics leads to the notion of a continuous medium, or briefly a continuum. By a continuum, we mean a hypothetical physical object in which
the matter is continuously distributed over the entire object. Solids and
fluids whose mechanical behavior is studied from macroscopic point of
view stand as two concrete examples of a continuum. Continuum mechanics
deals with the study of deformation and motion of a continuum, and
thereby provides a unified approach to solid and fluid mechanics.

4.3
CONFIGURATION OF A CONTINUUM
Consider a physical object (B that occupies (fills) a region B in three-dimensional space, at a given instant of time /. Then every part of the object (B has
a position (or place or location) in the region B. If the object (B is a continuum, that is if (B is made up of matter that is continuously distributed
(without gaps) over B, then every part of the region B is filled by some part
of the continuum (B. In particular, every point P in the region B is the
position of a part (P of the continuum (B; such a part (P is called a particle

1 58

CONTINUUM HYPOTHESIS

Figure 4.1. The configuration of (B.

of (B. Thus, a continuum (B may be regarded as a connected (continuous)


set of particles such that (i) every particle of (B has a position in the region
B occupied by (B, and (ii) every point P in the region B is the position of
some particle (P of the continuum (B. In other words, there exists a oneto-one correspondence between the particles of a continuum and the
geometrical points of a region that the continuum occupies at any given
instant of time; see Figure 4.1. The geometrical region that a continuum
occupies at a given instant of time is called the configuration of the continuum at that instant of time.
It may be noted that the configuration of a continuum and the position
of a particle of a continuum are defined with reference to a given instant of
time. As a continuum moves, the position of every particle of the continuum and the configuration of the continuum change from one instant to
another. Thus, a continuum has infinitely many configurations and every
particle of the continuum has infinitely many positions in space. However,
at any given instant of time, the position of every particle of a continuum
and the configuration of the continuum are uniquely determined. At no
instant can a particle have two distinct positions nor can two distinct
particles have the same position.
The one-to-one correspondence between the particles of a continuum and
the points of its configuration (at a given instant of time) enables us to
speak of points, curves, surfaces and volumes in a continuum. A part of a
continuum whose position is a geometrical point is referred to as a material
point (or a body point). A material point is nothing but a particle, as

4.4

MASS AND DENSITY

1 59

defined earlier. A part of a continuum whose position is a geometrical curve


is called a material curve or a material arc. The arc-length of a material
curve at a given instant of time is defined to be identical to the arc-length of
the geometrical curve that it occupies at that instant. A material arc of
infinitesimally small length is referred to as a material arc element. A part
of a continuum whose position is a geometrical surface is called a material
surface. The surface area of a material surface at a given instant is defined
to be identical to the surface area of the geometrical surface it occupies at
that instant. A material surface of infinitesimally small area is referred to as
a material surface element. A part (or the whole) of a continuum whose
position is a three-dimensional geometrical region is called a material body.
The position of a material body at a given instant of time is nothing but its
configuration at that instant. The volume of a material body at a given
instant of time is defined to be identical to the volume of its configuration
at that instant. A material body of infinitesimally small volume is referred
to as a material body element, material volume element or just a material
element. For simplicity, material points, material arcs, material surfaces
and material bodies are often referred to as points, arcs, surfaces and
bodies. Unless stated to the contrary, all arcs, surfaces and bodies we consider are supposed to have finite lengths, finite areas and finite volumes,
respectively.

4.4
MASS AND DENSITY
The one-to-one correspondence between the particles of a continuum and
the points of its configuration allows us to study physical and kinematical
quantities associated with a continuum through appropriate functions
defined over its configuration. To illustrate this fact, let us consider the concepts of mass and density that every physical object is supposed to possess.
At an instant of time t, let B be the configuration of a material body (B
and P be the position of a particle (P of (B. Consider an infinitesimal subregion SB of B containing point P, and let (B be the material element that
occupies (fills) the subregion SB (see Figure 4.1). Because of the hypothesis
of continuity in distribution of matter in (B, the subregion SB will not be
empty, however small it may be. If dV is the volume of , then SV is the
volume of <5(B. Let dm be the mass of J(B. Like time and space, mass and
volume are primitive quantities; these are postulated to be positive scalar s.
Thus, Sm > 0; SV > 0. Consequently, the ratio (m/V) is also a positive
number. Since is not empty, however small it may be, we may think of
the limit of (m/V) as V tends to 0 in the terminology of differential

160

CONTINUUM HYPOTHESIS

calculus. We suppose that this limit exists (finitely and uniquely) and denote
it by p. That is,

/> = Hm
(4.4.1)
Thus, p is the limit of the ratio of the mass of the material element
contained in an infinitesimal region around the point P to the volume of the
region, as the volume tends to 0. The quantity p is called the mass-density
(or just the density) of the material body at the point P. Note that the relation (4.4.1) defining p is consistent with the common definition of density
and that the value of p depends on the point P chosen in the configuration
B of the body (B. Hence, p is a function defined over B. Since the configuration B itself changes with time /, p is a function of t as well, in general.
While defining density through the relation (4.4.1), we have made the
tacit assumption that mass is a differentiable function of volume. This
means that the mass is assumed to be distributed over a region without
sudden increase or decrease in any part of the region. In other words, no
part of a material body is assumed to possess concentrated mass. Such
assumptions are made throughout our discussions. In fact, every function
considered in our analysis is supposed to be differentiable up to any desired
order in the region of interest. Thus, for example, we suppose that p defined
over the region B through the relation (4.4.1) is continuously differentiable
over B. Consequently, it follows that if V denotes the volume of B then the
relation (4.4.1) is equivalent to the relation
m =

pdV

(4.4.2)

where m is the total mass of the material body (B filling the region B.
Total kinetic energy, total force, total linear momentum, total angular
momentum, etc. of a material body are defined by relations analogous to
(4.4.2). Total force on a material surface is defined in a similar way by considering a surface integral. The use of integral calculus over spatial variables
distinguishes continuum mechanics from analytical mechanics dealing with
discrete systems.

4.5
DESCRIPTIONS OF MOTION
Suppose we wish to study a motion of a material body (B starting with some
particular instant of time. It is convenient to take this instant of time as the
origin in the time scale so that the instant corresponds to the reference or

4.5

DESCRIPTIONS OF MOTION

161

initial time t = 0. Let B0 be the configuration of (B at this instant of time.


Then B0 is called the reference or initial configuration. In view of the oneto-one correspondence between the particles of (B and the points of B0,
every particle of (B can be identified by its position in B0. This means that
if a typical point P0 in B0 is the position of a typical particle (P of (B at time
t = 0, then (P may be identified by the position vector x of the point P0
with respect to some fixed origin O. Thus, the vector x serves as a label for
the particle <P (for all time), and when convenient, (P itself may be called the
particle x.
As the body (B moves out of the configuration B0 at time / = 0, every
particle of (B moves with it. At any subsequent time t > 0, let P be the
position of the particle (P and let x be the position vector of the point P
(with respect to the origin O); see Figure 4.2. Since the point P determines
and is determined by x, P is referred to as the point x. Evidently, x depends
on the particle (P, which is identified with x, and /; and the motion
being studied determines the type of dependence. Thus, a motion of (B is
described by an equation of the form
x = x(x,0>

'>0

(4.5.1)

for / = 0

(4.5.2)

subject to the consistency condition


x = x

The lefthand side of equation (4.5.1) indicates the position x that the
particle x occupies at time / > 0, and the righthand side represents the
function that determines this x. The same symbol x is used to denote both
the dependent variable and the function for simplicity in the notation.

Figure 4.2. Initial and current positions of a particle.

162

CONTINUUM HYPOTHESIS

Equation (4.5.1) determines the position x of every particle (P of (B at an


instant of time / > 0. If we focus our attention on a specific particle, this
equation determines the successive positions of that particle at different
instants of time. On the other hand, at a specific instant of time the
equation determines the positions of all particles of (B at that instant of
time, and the totality of all these positions constitutes the configuration B
of (B at that instant. Thus, for a given t, equation (4.5.1) defines a mapping
from (B B0 onto B that carries a particle x of (B to the point x in B. The
point x is referred to as the current or instantaneous position of the particle
x. Also, the configuration B is referred to as the current or instantaneous
configuration of (B.
Since no two distinct particles of (B can have the same position in any
configuration and no two distinct points in a configuration can be positions
of the same particle, equation (4.5.1) should determine not only a unique x
for a given x and a given t, but also a unique x for a given x and a given
t. In other words, the mapping defined by (4.5.1) should be a one-one
(invertible) mapping so that (4.5.1) has a unique solution of the form
x = x(x, 0,

t>0

(4.5.3)

where the function x(x, t) is the inverse of the function x(x, t).
For any point x and for any instant of time / > 0, equation (4.5.3)
specifies the particle x of which x is the position at that instant. If we focus
our attention on a specific point x, equation (4.5.3) determines all those particles of (B that pass through that point at different instants of time t > 0.
On the other hand, at a specific instant of time equation (4.5.3) specifies all
particles that are positioned at different points of the current configuration
B, and the totality of all these particles constitutes the body (B. Thus, for a
given /, equation (4.5.3) defines a mapping from B onto (B = B0 that carries
a point x of B to the particle x of (B. This mapping is the inverse of the
mapping defined by (4.5.1).
Since the function x(x, /) is the inverse of the function x(x, t) it follows
that the function x(x, t) has to be the inverse of the function x(x, t). This
means that equation (4.5.1) can be recovered as a unique solution of
equation (4.5.3), in principle. In other words, at any instant of time t > 0,
equation (4.5.3) not only assigns a unique particle x to a given point x but
also yields equation (4.5.1), which specifies a unique position x for a given
particle x. Thus, the motion described by equation (4.5.1) is described by
equation (4.5.3) also. But the ways the two equations describe the motion
are not identical; they are only equivalent. While equation (4.5.1) contains
the particle x and time t as independent variables and specifies the position
x of x for a given /, equation (4.5.3) contains the point x and time / as
independent variables and specifies the particle x that occupies x for a

4.6

MATERIAL AND SPATIAL COORDINATES

163

given t. Thus, in the description of motion given by (4.5.1), attention is


focused on a particle and we observe what is happening to the particle as it
moves. This description is called the material description, and the independent variables (x, t) present in (4.5.1) are referred to as material variables.
On the other hand, in the description of motion given by (4.5.3) attention
is given to a point in space, and we study what is happening at the point
as time passes. This description is called the spatial description, and the
independent variables (x, t) present in (4.5.3) are referred to as spatial
variables. Both material and spatial descriptions are useful in their own
right. In solid mechanics, material description is more convenient than the
spatial description. It is the other way round in fluid mechanics. Traditionally, material description is referred to as the Lagrangian description in
honor of Lagrange, and the spatial description is referred to as the Eulerian
description in honor of Euler. Both descriptions were given by Euler.
In Section 4.4, we mentioned that physical and kinematical quantities
associated with a continuum are studied through appropriate functions
defined over its configuration. If a quantity is represented by a function
defined over the initial configuration B0 of a body (B, equation (4.5.3)
enables us to express that quantity as a function defined over a current
configuration B of (B. On the other hand, if a quantity is represented by a
function defined over a current configuration, equation (4.5.1) enables us
to express that quantity as a function defined over the initial configuration.
In this way, every quantity associated with a body (B (or its motion) can be
expressed in terms of the material variables (x, t) or in terms of the spatial
variables (x, /) When a quantity is expressed in terms of (x, /)> w e say that
it is in the material or Lagrangian form or Lagrangian description. When a
quantity is expressed in terms of (x, t) we say that it is in spatial or Eulerian
form or Eulerian description.

4.6
MATERIAL AND SPATIAL COORDINATES
We often deal with the component forms of equations (4.5.1) and (4.5.3).
For this purpose, we set up at the origin O a fixed system of righthanded,
rectangular Cartesian coordinate axes and consider the components of both
x and x along these axes (see Figure 4.3). We denote the base vectors of the
coordinate system by ef, as usual, and set
x e, = x?,

x e, = Xi

(4.6.1)

so that x? are the components of x and xt are the components of x, along


e f . To avoid possible confusion that may arise while dealing with xf and *,

164

CONTINUUM HYPOTHESIS

Figure 4.3. Material and spatial coordinates.

together, we refer to the coordinate axes as 1, 2, 3 axes, or briefly the /-axes.


Thus the /-axis is the axis along which the base vector e, is directed.
Then, equations (4.5.1) and (4.5.3) read respectively as follows in the
component form:
Xi = Xi(4,0,
*>0
(4.6.2)
x? = x?(xk9t),

t>0

(4.6.3)

Also, the consistency condition (4.5.2) reads


Xi = xf

at t = 0

(4.6.4)

Since x is the position vector of the point P 0 (which is the initial position
of a particle (P), xf are the coordinates of P0. These coordinates are termed
material coordinates. Also, since x is the position vector of the point P
(which is the current position of (P), xt are the coordinates of P, called
spatial coordinates. Often, material coordinates are referred to as referential or initial coordinates, and spatial coordinates are referred to as instant
or current coordinates.
As a consequence of the one-to-one correspondence between the particles
of a material body and the points of its configuration at any instant of
time, it has been noted that the function x(x, /) and x(x, /) are inverses
of each other and either equations (4.5.1) or (4.5.3) is a unique solution
of the other. Therefore, equations (4.6.3) represent a unique solution of
equations (4.6.2) and vice versa. An immediate implication of this is that
the functions Xi(xk9 0 and xf(xk91) possess continuous partial derivatives

4.6

MATERIAL AND SPATIAL COORDINATES

165

with respect to their arguments and that the Jacobian J, defined by

'-Kl?)

(4 6 5)

is nonzero for any xf and any / of our interest. This amounts to saying that
a motion described by (4.6.2) and (4.6.3) takes place smoothly and that it
carries curves, surfaces and regions into curves, surfaces and regions,
respectively. In particular, boundary surfaces get transformed to boundary
surfaces only so that the particles that lie on a boundary surface of a
material body in one configuration continue to remain on the boundary
surface of the body in all configurations.
Another important consequence of the continuous differentiability of the
functions X/OcJ, t) and x?(xk, t) is that the vanishing of an integral of a
continuous function of spatial or material variables over an arbitrary
volume implies the vanishing of the function at every point of the volume.
That is, \f\yfdV
= 0 for arbitrary volume V, t h e n / = 0 at every point of
V. This result, referred to as the localization theorem, will be employed in
Chapter 8 for deducing the local forms of balance equations from their
integral forms.
With the view of making use of the conventions and advantages of the
suffix notation in dealing with partial derivatives of the type present in
(4.6.5), we will henceforth employ the semicolon notation:

reserving the usual comma notation for partial differentiation with respect
to A:,. Then (4.6.5) can be rewritten as
J = det(xi;d,)

(4.6.6)

Since xt are assumed to be continuously differentiable functions of xf


and /, the partial derivatives jci;7 are continuous in jcf and /. Consequently,
/ is a continuous function of xf and t. The consistency conditions (4.6.4)
require that / = det(x?.y) at / = 0. Since xf;J = Jl7 and det(^) = 1, we get
y=l

at / = 0

(4.6.7)

Thus, (i) J = 1 for / = 0, and (ii) J is continuous and nonzero for t > 0.
Therefore, / cannot be negative for any t\ that is,
J> 0

for all * > 0

(4.6.8)

The geometrical-physical meaning of this inequality will be discussed at a


later stage.

166

CONTINUUM HYPOTHESIS

It is clear that every term defined, every assumption made and every
inference drawn in this chapter stems from the hypothesis of continuity
in distribution of matter over a region of space. The whole theory of
continuum mechanics is based on this continuum hypothesis.

CHAPTER 5
DEFORMATION

5.1
INTRODUCTION
In Chapter 4, we explained what we mean by a material body and how its
motion is described. The present chapter is concerned with the analysis of
the geometrical changes that take place in a material body during its motion
from one configuration to the other. The tensors which serve to measure
these changes will be introduced and the related aspects will be considered
in some detail.

5-2
DEFORMATION GRADIENT TENSOR
For analyzing the geometrical changes that take place in a material body (B,
as it moves, we focus our attention on only two of its configurations: B0 and
B, where B0 is the reference configuration and B is the current configuration
at a specified (fixed) instant of time / > 0. We will refer to the transition of
(B from B0 to B as a deformation. Also, BQ will be referred to as the initial
167

168

DEFORMATION

or undeformed configuration, and B will be referred to as the final or


deformed configuration (see Figure 5.1).
Consider a representative particle (P of the body (B and a material arc
element d<3 issuing from (P. In the initial configuration B0 this arc lies on
some geometrical arc, say dC09 issuing from the initial position P0 of (P. Let
x be the position vector of the point P0 and d\ be the vector representing
the arc element dC0 in both length and direction. Then the initial direction
of rfC is the direction of the vector dx and the initial length of dG is
ds0= \dx\.
When the body (B moves over to the final configuration B, carrying with
it the particle (P and its neighboring particles, the material arc dG lies on
some geometrical arc, say dC, in B that initiates from the position P of (P
in B. (Recall that, by the continuum hypothesis, curves transform to
curves.) Let x be the position vector of the point P and dx be the vector
representing the arc element dC both in direction and length. Then
ds = \dx\\s the final length, and the direction of dx is the final direction of
de. We will call dC the final location of dC. On the other hand, dC0 will
be called the initial location of dQ.
In the material description, a motion of (B is described by equations of
the form (4.6.2) in which xf and / are independent variables and xt are
dependent variables, xf and xf being the components of x and x, respectively. Therefore, the differentials dxt can be expressed in terms of the
differentials dx? through the following formula of differential calculus:
3
dx
dxi= ^
X

7=1 j

i= 1,2,3,

the time variable / being held fixed.

Figure 5.1. Initial and final configurations.

(5.2.1)

5.2

DEFORMATION GRADIENT TENSOR

169

Using the conventions of the suffix notation, and the semicolon notation
defined in Section 4.6, expression (5.2.1) can be rewritten as
dXi = xrJdxj

(5.2.2)

We note that (A:I;7) is a 3 x 3 nonzero matrix related to xf system of


coordinates. For an arbitrary vector d\ with components djcf, expression
(5.2.2) shows that Xijdxj are components of a vector, dx. Therefore, by
quotient law (Section 2.4, property 7(ii)), it follows that xi;J are components
of a nonzero tensor. This tensor is immediately recognized as the gradient
of the vector x, the gradient being taken with respect to xf9 called the
deformation gradient tensor in the material form, or the material deformation gradient. We denote it by Vx or F, the symbol V denoting the gradient
operator with respect to xf. Thus,
[Vx]u = [F]/7 = xrJ

(5.2.3)

/ = det(x/;7) = det(Vx) = detF

(5.2.4)

and
Consequently, (5.2.2) can be rewritten as follows in the direct notation
dx = (Vx) dx = F dx

(5.2.5)

This expression clearly exhibits that the vector dx is generated by the action
of the tensor F on the vector dx. In other words, the effects of the deformation on the material arcrfe is represented by the tensor F. Hence, (5.2.5)
serves as a transformation rule satisfied by d(2> as it deforms from its initial
location represented by the vector dx to the final location represented by
the vector dx . Since dx is a nonzero vector, we find from (5.2.5) that dx
is also a nonzero vector. Thus, if a material arc has a nonzero length in one
configuration it continues to have a nonzero length in every configuration.
In other words, the tensor F transforms arcs of nonzero lengths to arcs of
nonzero lengths. Relation (5.2.5) is the fundamental relation upon which
the whole analysis of this chapter is based.
Let us consider two special cases of the relation (5.2.5).
5.2.1

TRANSLATION

Suppose F = I. In this (trivial) case, (5.2.5) reads dx = dx. This means that
both the length and the orientation of d<2> are now preserved under the
deformation. Such a deformation is called a rigid-translation or just a
translation; see Figure 5.2.

170

DEFORMATION

Figure 5.2. Translation.

5.2.2

ROTATION

Suppose F is an orthogonal tensor 5* I. Then by virtue of the result proved


in Example 2.8.3, we find that \dx\ = \dx\. But, since F 5* I, dx ^ dx.
Hence, in this case, the length of rfC remains unchanged but its orientation
undergoes a change. In other words, the effect of the deformation, now, is
just to change the orientation of dQ without changing its length. Such a
deformation is called a rigid rotation or just a rotation; see Figure 5.3. It is
to be noted that, under a rigid rotation, the arc dC may experience a
translation as well. This is because, F always contains I as a factor; that is,
F = FI = IF.
In the general case, where F is neither I nor an orthogonal tensor, dG
generally experiences a change in length as well as a change in orientation.
An analysis of this general case requires the use of the polar decomposition
of F, and this will be taken up in detail in Section 5.3. Pending this analysis,
we proceed to examine how a material surface element and a material
volume element change under a deformation.

Figure 5.3. Rotation.

5.2

5.2.3

DEFORMATION GRADIENT TENSOR

1 71

DEFORMATION OF A SURFACE ELEMENT

At the particle (P consider another material arc element d& so that dC and
d& together determine a material surface element rfS. Let rfC be represented
by the vectors dx and dx in the configurations B0 and B, respectively.
Then, by virtue of the geometrical meaning of the cross product of vectors,
the areas of rfS in B0 and B are respectively given by
dS0 = \dx x dx\

(5.2.6a)

dS = \dx x dx\

(5.2.6b)

Also, if n and n are unit normal vectors to rfS in B0 and B, respectively,


such that the triads (rfx, dx, n) and (dx, dx, n) are righthanded (see Fig.
5.4), we have
dx x dx = (dS0)n

(5.2.7a)

rfxxdx = (dS)n

(5.2.7b)

Since dG and tfC are of nonzero initial lengths, the initial area dS0 of rfS
is not zero, provided dQ and dC are initially noncoincident. Under the
deformation defined by (5.2.5), dQ and d<2> continue to be of nonzero
lengths in every configuration. Hence, dx and dx, which represent these arcs
in B, are nonzero vectors. Therefore, from (5.2.6b) we find that dS = 0 if
and only if dx and dx are collinear; that is, if and only if dC and dC become
coincident in B. If dC and rfC are initially noncoincident, the possibility of
f(B and dG becoming coincident in B is ruled out, because two different

Figure 5.4. Righthanded triads in two configurations.

172

DEFORMATION

particles cannot occupy the same position in the same configuration. Hence
dS 5* 0 whenever dS0 ^ 0. Thus, if a material surface element has a
nonzero area in one configuration, it continues to have a nonzero area in
every configuration. In other words, F transforms surface elements of
nonzero areas to surface elements of nonzero areas.
From (5.2.5), we find that
dxxdx

= (F d\) x (F dx)

which on using (2.11.25) becomes


dxxdx

= F*(dx x dx)

(5.2.8)

where F* is the cofactor of F. By use of (5.2.7), expression (5.2.8) yields


(dS)n = (dS0)F*n

(5.2.9)

(dS)FTn = (dS0)FTF*n

(5.2.10)

which gives
Using (2.11.23) and (5.2.4), we note that
F r F* = J\

(5.2.11)

so that (5.2.10) can be rewritten as


(dS)FTn = J(dS0)n

(5.2.12)

This expression explicitly exhibits how the oriented surface element (dS)n
representing dS in the final configuration is related to the oriented surface
element (dS0)n representing the same dS in the initial configuration. In
fact, (5.2.12) reveals an important characteristic property of the tensor F.
Recall that dS and dS0 are nonzero areas. Also, n and therefore n are
arbitrary vectors; hence F r n is a nonzero vector. Consequently, it follows
from (5.2.12) that / ^ 0, which is precisely what we noted immediately
after defining J through the relation (4.6.5). An immediate consequence of
this important relation is that the tensor F is an invertible tensor, that is, the
inverse F _ 1 of F exists.
From (5.2.5), we find that
dx = F~1dx

(5.2.13)

which shows that the tensor F" 1 carries the vector dx to the vector dx. In
other words, (5.2.13) defines the transformation that carries d<5 back to the
initial location dC0 from the final location dC.

5.2

DEFORMATION GRADIENT TENSOR

1 73

In the spatial description, xf are functions of xt and /. Hence


3
dx?
dxi= ^dxj
j= l

ax

= xljdxj

(5.2.14)

if / is held fixed. Evidently, x?j are the components of the gradient of x,


namely Vx, the gradient being taken with respect to xt. As such, (5.2.14)
can be rewritten, in the direct notation, as
(5.2.15)

d\ = (Vx)dx.
Comparing this with (5.2.13), we find that
F"' = Vx;
Consequently (since det A

-1

\-\

= 4J

(5.2.16)

= (det A)" for any invertible tensor A),

detF"1 = det(x?j) = -

(5.2.17)

The tensor F"1 is called the deformation gradient in the spatial form, or
briefly the spatial deformation gradient.
Let us return to (5.2.12) and analyze another of its important implications. To this end, consider one more material arc element de issuing from
the particle (P and initially directed along n. Then tfC is initially perpendicular to rfS. Let dx and dx be the vectors representing rfC in the initial
and final configurations respectively. Then, by use of (5.2.5), (5.2.12),
(2.8.11) and (2.8.14), we find that
rfx-n^iFrfx0)^/^^)-10

= J(

dS

0\

M)di

j = 0 . 0

"

(5 2 18)

If d0 and ds are the initial and final lengths of dG and if is the angle
between dx and n then dx n = ds0 and dx n = (ds) cos 0, and (5.2.18)
yields

<i)(3)
Evidently, J* 0 in general, the case 0 = 0 being not ruled out. This means
that in the final configuration, d& is not necessarily perpendicular to <iS. In
other words, a material arc that is perpendicular to dS in one configuration

174

DEFORMATION

Figure 5.5. Initial and final orientations of dC

need not be perpendicular to rfS in a subsequent configuration; see


Figure 5.5.
A question now arises: can be equal to /2? The answer is no, because
can become equal to n/2 only if d<2> falls onto dS and becomes a part of
rfS in the final configuration. This is not possible, as there is no gap in dS
to take in tfG. The next question is this: can be greater than /2? The
answer to this question is also in the negative, because 0, which starts from
the value 0 in the initial configuration, can acquire a value greater than n/2
only after passing through the value /2, which has already been ruled out.
(Remember that we deal with continuous transformations!) Thus, must lie
in the interval 0 < < n/2. As such, the lefthand side of (5.2.19) is
(always) positive. Since dS0,dS,d50,d$ are all positive, it follows from
(5.2.19) that / > 0, which is precisely the inequality (4.6.8) obtained by the
assumption of continuity of / . This inequality asserts that the deformation
preserves the sense of orientation of surface elements. Hence if dS is
positively oriented in one configuration, it continues to be positively
oriented in every configuration.
It is to be pointed out that the positive nature of / is a synthesis of the
various postulates and assumptions made in the course of the preceding
discussions. The geometrical meaning of J is obtained next.
5.2.4

DEFORMATION OF A VOLUME ELEMENT

Let us now consider the case where d is not necessarily orthogonaj_ to and
not a part of dS in the initial configuration. Then, dQ9 d<8 and dC determine a material volume element, dV. By virtue of the geometrical meaning
of the scalar triple product, the volumes of dV in the initial and final

5.2

DEFORMATION GRADIENT TENSOR

1 75

Figure 5.6. Volume elements dV0 and dV.

configurations are respectively given by (see Figure 5.6)


dV0 = \(dx x dx)

rfx|,

dV = \(d\ x dx) dx\.

(5.2.20)

Now, by using (5.2.8) and (5.2.5), we find that


{dx x dx) dx = F* (tfx x tfx) F dx
which, by use of (2.8.14) and (5.2.11), becomes
(dx x dx) - dx = FTF*(dx x dx) dx
= J(dx x dx) dx

(5.2.21)

Taking the absolute values on both sides of this expression and using
(5.2.20) we get
dV= JdV0

(5.2.22)

This expression explicitly exhibits how the final volume dV of dV is related


to its initial volume dV0. In fact, (5.2.22) reveals that dV is generally not
equal to dV0i the case dV = dV0 not being ruled out. Further, / represents
the ratio of the final volume of dV to its initial volume. More fundamentally, dV ^ 0 whenever dV0 ^ 0. Thus, a material element having
nonzero volume in one configuration continues to have nonzero volume in
every configuration. This characteristic property of motion is a direct consequence of the invertibility of F. Expression (5.2.22) is given by Euler.
5.2.5

ISOCHORIC DEFORMATION

From the Euler's formula (5.2.22) we note that a material volume element
at the particle (9 does not suffer a change in volume under a deformation if
and only if J = 1. In such a case we say that (9 undergoes an isochoric
(volume-preserving) deformation. If a deformation is such that every particle undergoes an isochoric deformation, we say that the deformation is an
isochoric deformation. A material body is said to be incompressible if every
deformation it undergoes is an isochoric deformation. Thus, a deformation
is isochoric if and only if J = 1 at every particle in every configuration, and

176

5 DEFORMATION

a material body is incompressible if and only if every deformation it


undergoes is an isochoric deformation.
EXAMPLE 5-2.1
Xl +

For the deformation defined by equations


x2 = -2%,

X 2,

3=

+ %-$

(5.2.23)

find F , F _ 1 and/.
Solution The given equations (5.2.23) express xt in terms of xf. Hence the
description of deformation is in the material form. From these equations,
we obtain
1

1 -2

[F] = [xi;J] =

(5.2.24)

1 -1

This is the matrix of F for the given deformation. From this matrix we find
that
1

1 -2

1
J = det F =

(5.2.25)

= 3

1 -1

Solving equations (5.2.23) for xf, we obtain


x = |(2x, + x2),

xl = i(Xi - *2)>

xt = xl-

x3 (5.2.26)

These equations describe the given deformation in the spatial form. From
these equations we get
2/3
_1

[F ] = [xfj

1/3

1/3-1/3

[ 1

(5.2.27)

0-1

This is the matrix of F _ 1 for the given deformation.


EXAMPLE 5.2.2
*? = K*i + xl),

For the deformation defined by the equations


*2 = tan" 1 ^/*!),

Xi ^ 0 ,

x$ = x}

(5.2.28)

find F and F _ 1 . Show that the deformation is an isochoric deformation.

5.2

DEFORMATION GRADIENT TENSOR

1 77

Solution From the equations (5.2.28), we observe that the given deformation is described in the spatial form. These equations give

[F"1] = [x?J =

X\

*2

R2

R2

(5.2.29)

where R2 = xj + x\, which is taken to be nonzero.


Solving equations (5.2.28), we obtain
xx = (2x?)1/2 cos*?,

x2 = (2x?)1/2 s i n 4 ,

x3 = (5.2.30)

These equations describe the given deformation in the material form. From
these equations we obtain

To
[F] = [xi;j]

cos x% - 0 sin x% 0

1
sin x%
&

(5.2.31)

0 cos x% 0

where 0 = (2x?)1/2.
Thus, (5.2.31) gives the matrix of F and (5.2.29) gives the matrix of F" 1
for the given deformation.
Also, from (5.2.31), we find that
/ = det(xi;j) = 1

for any xf

Hence the given deformation is an isochoric deformation.


EXAMPLE 5.2.3
(i)
OU

Solution

Prove the following:


[

\jj
L* Jy

Xjj

wirsjpqXp;r*q

(5.2.32)

%l\j

iJtirs&jpqXp.rXq.s

(5.2.33)

(i) From the identity (5.2.11), we get


F-

= i(F*)r

(5.2.34)

178

DEFORMATION

Since xi;j are the components of F, the components of the cofactor F* of


F are given by [see (2.11.22)]
[F \ij = jipqjrsXp,rXq;s

(5.2.35)

Since xfj are the components of F _ 1 , we find from (5.2.34) and (5.2.35)
that

_ J_

~~ *\ j ^JPQ ^irsxp; rxq; s

which is the relation (5.2.32).


(ii)From (5.2.11), we find
F = /[(F-1)*]7

(5.2.36)

Since xfj are the components of F - 1 , the components of (F - 1 )* are given by


[(F" 1 )*],, = ieipg8jrsx0Ptrxq>s

(5.2.37)

Since xi;J are components of F, we find from (5.2.36) and (5.2.37),


Xrj = [F]/, =
=

TF.
2 uc,jpq

which is the relation (5.2.33).


EXAMPLE 5 . 2 . 4

x
F
Y s
irs-Y
p, r -*<7,

Prove the following identities:

(0

^-] = ^-

0\Xi;j)

div(JF r j=0

(ii)
Solution
(1.7.26),

AP'1)*]*

(5.2.38)

(5.2.39)

(i) Since J = det(x, v ), we have, by virtue of the relation


J

~e^imn^pqrxi\pXm\qXn\r

Differentiating this partially with respect to (xi;j) and noting that


d(*i;P)
diXij)

*
Pj

=dx[

dx[dxL _ o
Xj iXi
dXidxp
- '->

(j.Z,*HJ)

5.2

DEFORMATION GRADIENT TENSOR

1 79

we obtain
1

dJ

which by use of (5.2.40) becomes


dJ

(5.2.41)

dbi-j)

This is precisely the identity (5.2.38).


(ii) Since

[>

we have

XK

..

1
JXjil
J

dxXjdxf)

\JXJVJ
1 /dxj\

d / 1 \ dXj

~ J dxj \dxf)

dxj \j) dxf

1 d /dxj\
/ dXj\dxf)

1 dJ dXj~\

JdXjdxfl

(5.2.42)

By using the chain rule of differentiation and the identity (5.2.41), we


obtain
dJ
dxj

dJ
(,*)
d(xns) dxj
dx< d idx,
dxr dXj \

Therefore
1 dJ dxj _ d fdxr\ dxj
Jdxj'dx^ = 'dxrdXj \dJ?J die?
_dx[_d_ /3
~ dxr dxf \dx)

_dx[_d_ (Xj_
~ dxr dxs \dx?

~ dxr \dxf,
In view of this relation, we find that the righthand side of (5.2.42) vanishes,
and the identity (5.2.39) is proven.

180

DEFORMATION

5.3
STRETCH AND ROTATION
We now proceed to analyze the effect of deformation defined by (5.2.5),
namely,
dx = dx
(5.3.1)
on the material arc elementtf(Bconsidered in Section 5.2, in the general case
where F is neither the identity tensor nor an orthogonal tensor.
Let a0 and a be the unit vectors directed along dx and dx, respectively
(see Figure 5.7), so that
dx = (tfs0)a

(5.3.2a)

dx = (ds)a

(5.3.2b)

Using (5.3.2), expression (5.3.1) can be rewritten as


= Fa0

(5.3.3)

(ds)
1 = ^

(5.3.4)

kprp
L1C1 v

(ds0)

Evidently, is a positive number representing the ratio of the final and


initial lengths of dQ. It follows that ofC does not change in length as it
moves from initial location dC0 to the final location dC if and only if = 1,
or equivalently F preserves the length of a0. Not every F can have this
property. As such, dQ generally changes in length as it moves from dC0 to
dC. This change is called an extension if ds > ds0 (i.e., > 1) and a contraction if ds < ds0 (i.e., < 1). The number is called the stretch of dQ
or stretch along a0.

Figure 5.7. Vectors a0 and a.

5.3

STRETCH AND ROTATION

181

From (5.3.3), we find that a is aligned with a0 if and only if Fa0 = //a0,
or equivalently a0 is an eigenvector of F with as the corresponding eigenvalue. Not every F can have a0 as its eigenvector. Hence a is generally not
aligned with a0. This means that the orientation of rf6 generally changes as
it moves from dC0 to dC.
Thus, under a deformation, dC generally changes in length as well as
orientation. Later we show that this change can be interpreted as the net
result of a translation, a rotation and proportional extensions or contractions in three mutually perpendicular directions.
5.3.1

DECOMPOSITION OF A DEFORMATION

In Section 5.2, it was noted that F is an invertible tensor. Hence, by the


polar decomposition theorem (Theorem 2.14.4), F can be decomposed
uniquely in the forms
F = QU

(5.3.5a)

F = VQ

(5.3.5b)

where Q is an orthogonal tensor, and U and V are positive definite symmetric


tensors such that
U2 = F r F

(5.3.6a)

V2 = FF r

(5.3.6b)

Using the right polar decomposition of F, given by (5.3.5a), the relation


(5.3.1) can be rewritten as
dx = (QU) dx = Q(\Jdx)

(5.3.7)

dy = Vdx

(5.3.8a)

dx = Qdy

(5.3.8b)

or, equivalently,

Thus, the transformation of tf(B from its initial location dC0 to the final
location dC can be decomposed into two parts. Under the first part of the
transformation, given by (5.3.8a), dC is carried from dC0 to an intermediate location, say dC, represented by the vector dy, and in the second
part given by (5.3.8b), it is carried from dC to the final location dC. Note
that the first part of the transformation is effected by the tensor U acting on
dx and the second part is effected by the tensor Q acting on dy.

182

5 DEFORMATION

5.3.2

TRIAXIAL STRETCH

Since U is positive definite and symmetric, it has at least one set of mutually
orthogonal eigenvectors, say p l f p 2 , p 3 , which may be taken as the base
vectors of a coordinate system, and three corresponding eigenvalues, say
9 2, //3, which are all positive; see sections 2.13 and 2.14. Hence U has
the following spectral representation (see Example 2.13.5):
3

U = 1k(Pk P*)
k=l

(5.3.9)

Substituting this expression into (5.3.8a) and using the identity (2.4.37)
we get
dy=

nk(pk'dx)pk

Ar=l

(5.3.10)

If dy'i and dxf' are the components of dy and d\ along the base vectors
p, (that is, if dy\ = dy p, and dxf' = d\ p,), then (5.3.10) yields
dy{ = nxdx\'

(5.3.11a)

{ = 22'

(5.3.11b)

dyi = n,dxf

(5.3.11c)

We note that dxf ' are the projections of the length of d<3 along the
directions of p, in the initial location dC0 and dy- are the corresponding
projections in the intermediate location dC. Expressions (5.3.11) show that
these two sets of projections are not the same. We find from (5.3.11a) that
the projection of the initial length of tf(3 in the direction of pj gets
multiplied by the factor as it moves from dC0 to dC. This means that JG
experiences a stretch equal to in the direction of pj as it moves from dC0
to dC Similarly, (5.3.11b and c) show that dG generally experiences
stretches equal to 2 and 3 along the directions of p2 and p 3 . (See Figure
5.8.) This experience of d(5 (as it moves from dC to dC) is referred to as
triaxial stretch suffered by dQ. The numbers r\i are called the principal
stretches and the directions of p, (along which are the stretches) are called
the principal directions of stretch.
5.3.3

RIGID-BODY TRANSFORMATION

Let us now consider the second part of the transformation that carries dQ
from the intermediate location dC to the final location dC. This part of the
transformation, effected by Q acting on dy, is described by the equation
(5.3.8b). Since Q is an orthogonal tensor, we have |rfx| = |rfy|. Hence, Q

5.3

STRETCH AND ROTATION

183

Figure 5.8. Triaxial stretch.

does not effect a change in the length of dy; the only (possible) effect of Q
on dy is to change the orientation of dy. This means that, under the second
part of the transformation, tf6 generally changes its orientation but retains
the length it had acquired in the first part of the transformation. That is, dC
acquires its final length in full in the first part of the transformation itself;
in the second part of the transformation, it just gets tilted (rigidly) to the
final orientation. Thus, the second part of the transformation is just apure
frigid) rotation.
In the process of the full transformation from dC0 to dC, rfC may
experience apure translation as well. The translation may occur before the
triaxial stretch, between the triaxial stretch and the rigid rotation, after the
rigid rotation, or simultaneously with the triaxial stretch and rigid rotation.
For definiteness in the interpretation, we suppose, without any loss of
generality in the analysis, that the translation occurs after the rigid rotation.
(This supposition amounts to treating Q(dy) as l(Qdy) so that the identity
tensor I represents the effect of translation.) Then, the second part of the
transformation, effected by Q, may be interpreted as a rigid-body transformation consisting of a pure translation and a pure rotation.
The preceding analysis shows that, during its transition from the initial
location to the final location, dQ may be thought of as being subjected to
triaxial stretch followed by a rigid-body transformation (see Figure 5.9).
And the net result is a change in length as well as a change in the orientation
of dG. This interpretation is based on the right polar decomposition of F
given by (5.3.5a). A similar interpretation can be arrived at on the basis of

184

DEFORMATION
Effect of U
\

Effect of Q
\

Translation

s
P
Figure 5.9. Stretch followed by rotation and translation.

the left polar decomposition given by (5.3.5b). In this case, triaxial stretch is
effected by the tensor V and the rigid-body transformation is effected (as in
the previous case) by Q. But here the rigid-body transformation precedes
the triaxial stretch. Bearing in mind that U and V have the same eigenvalues
but not necessarily the same eigenvectors (see Example 2.14.3), we may
infer that, while the principal stretches remain the same in both the cases,
the directions in which these occur are generally different in the two cases.
The tensor Q that essentially effects the rotation (in both the cases) is
referred to as the rotation tensor. The tensors U and V that effect triaxial
stretches are known as the right and left stretch tensors, respectively.
In the special case when U = V = I, we have F = Q (by 5.3.5), and
the deformation then consists of no triaxial stretch; d(2> experiences just a
rigid-body transformation. Conversely, if d6 experiences just a rigid-body
transformation, then U = V = I so that F = Q. Thus the deformation
given by (5.3.1) is a rigid-body transformation if and only if F = Q.
Since material arcs generally experience changes in their lengths and
orientations (under a deformation) as seen in the preceding analysis, the
relative distances between the particles of a material body are generally
altered as the body moves from one configuration to the other. Consequently, the size and shape of the body generally change under deformation. Material bodies whose size or shape change under deformation are
called deformable materials. Material bodies that are not deformable are
called rigid materials. All real materials are deformable to a certain degree,
and the concept of rigid material is just an abstract one. Rigid materials are
of little interest in continuum mechanics.
EXAMPLE 5-3-1 For the deformation defined by equations (5.2.23)
find (i) the direction after deformation of a material arc dQx initially having
direction ratios 1:1:1 and (ii) the direction before deformation of a
material arc d&2 finally having direction ratios 1:1:1.

5.3

STRETCH AND ROTATION

185

Solution For the given deformation, [F] and [F -1 ] are as given by


(5.2.24) and (5.2.27).
For the arc <2, we have
(5.3.12)

a = al = al = ^
Using (5.2.24) and (5.3.12) in (5.3.3) we find
=

73

2 =

1
~^3

3 =

1
73

(5.3.13)

Hence, after deformation, dQx has direction ratios 2 : - 1 : 1 .


For the arc cfC2 we have
(5.3.14)

tfi =a2 = a3 = ^
Using (5.2.27) and (5.3.14) in (5.3.3) we get
a

"? = * 7 J

2=

"3 = 0

(5.3.15)

Hence, before deformation, dQ2 had direction ratios 1:0:0. That is,

initially rf62 laid along the 1 axis.


EXAMPLE 5.3.2

Deformations defined by equations of the form


Xi

= QijXj + Ci

(5.3.16)

where a^ and c, are constants or at most functions of time are called homogeneous deformations. Show that, under a homogeneous deformation,
plane elements transform to plane elements and straight lines transform to
straight lines.
Solution From equations (5.3.16) we find that [F] = [ay]. Since F has to
be invertible, [a^\ has to be a nonsingular matrix and equations (5.3.16)
have to be solvable for xf to get equations of the form
x? = a\xj - cj)
1

(5.3.17)

-1

where aj] are elements of the matrix [F ]; that is,


lau1] = [au]'1 = [IT 1
(5.3.18)
In the initial configuration, let a material surface element be a plane
element with equation
a,*?+ 0 = 0

(5.3.19)

186

DEFORMATION

where a and are constants. Using (5.3.17), this equation becomes


< W ( * / -cj) + = 0

(5.3.20)

This is the equation of the material surface element in the deformed


configuration. Being a linear equation in JC,, equation (5.3.20) represents a
plane. As such, the surface element continues to be a plane element after
deformation also.
Thus, under a homogeneous deformation, a plane element transforms to
a plane element. Consequently, since a straight line may be thought of as
the intersection of two planes, it follows that straight line elements transform to straight line elements under homogeneous deformations.

EXAMPLE 5.3.3 (i) Show that a homogeneous deformation defined by


equations (5.3.16) is a rigid-body transformation if and only if [#,,] is an
orthogonal matrix.
(ii) Deduce that the deformation defined by the equations
xx = x\ cos + x% sin
x2 = -x* sin + je cos

(5.3.21)

where is a constant, is a rigid-body rotation.


Solution (i) We recall that a deformation is a rigid-body transformation
if and only if F = Q, or equivalently [F] is an orthogonal matrix. Since
[F] = [au] for a homogeneous deformation defined by (5.3.16), the
required result is immediate.
(ii) We observe that equations (5.3.21) are of the form (5.3.16). Therefore, the given deformation is a homogeneous deformation. From (5.3.21)
we find that
(5.3.22)
It is easy to verify that this matrix is orthogonal. Hence, F is an orthogonal
tensor and the given deformation is a rigid-body rotation.

EXAMPLE 5.3.4 For the deformation defined by the following equations, find Q, U and V:
xx = 2x1

x2 = -Xi

x3 = -2x1 - 3x?

Give a geometrical description of the deformation.

(5.3.23)

5.3

Solution

STRETCH AND ROTATION

187

From equations (5.3.23), we get


0

[F] = I - 1

0 -2

-3

(5.3.24)

In Example 2.14.2 the method of obtaining the polar decomposition of a


tensor has been illustrated. Using that method, we find that for the deformation gradient given by (5.3.24) the matrices of the right and left stretch
tensors U and V and the matrix of the rotation tensor Q are as follows:
"-1
[U] = |

0 -2 I

(5.3.25a)

0 - 2 - 3
0

[V] = | 0 - 1
2

2
0 |

(5.3.25b)

0-3

0 -1

[Q] = | 1

0 I

(5.3.25c)

From (5.3.25a), we find that the eigenvalues of U are


1i = - 1 .

*\i = 1,

3 = - 4

(5.3.26)

These are the principal stretches for the given deformation.


It can be verified that the eigenvectors of U are as follows:
Pi = e l f

2
1
p2 = ~ ^ J e 2 + ^ e 3 ,

P3 = ^ e

+ ^e

(5.3.27)

These vectors represent the principal directions of stretch for the given
deformation.
Comparing the matrix of Q given by (5.3.25c) with the transformation
matrix (2.2.13), we find that Q represents a (rigid) rotation about the 3 axis
through an angle 3/2 in the sense of a righthanded screw.
Thus, the given deformation may be thought of as the one in which a
material arc experiences triaxial stretches - 1 , 1 and 4 , respectively, along
the directions of the vectors Pi, p 2 and p 3 given by (5.3.27) followed by a

188

5 DEFORMATION

rotation through an angle 3/2 about the 3 axis in the sense of a righthanded screw. Since the given equations do not contain constant terms, no
translation takes place.

5.4
STRAIN TENSORS
In the preceding section it was shown that, under a deformation, a material
arc generally experiences a change in length and a change in orientation. We
now proceed to derive expressions that will enable us to compute these
changes.
Let us start with expressions (5.3.3) and (5.3.4):
/7a = Fa0
=

(5.4.1)
(5.4.2)

From (5.4.1), we find that


2 = //a = Fa0 Fa0 = a0 (FrF)a

(5.4.3)

by use of (2.8.14). If we set


C = F r F,

(5.4.4)

2 = a0 Ca

(5.4.5)

expression (5.4.3) becomes


From (5.4.4), (5.4.5) and (5.3.6a) we find that C is a positive definite
symmetric tensor and that C = U2. Expression (5.4.5) shows that the tensor
C serves as a means of calculating when a0 is known. This tensor was
introduced by George Green in 1841 and is known as the Green deformation
tensor.
For a rigid-body transformation, we have F = Q. Expression (5.4.4) then
becomes C = Q r Q = I. Thus, for a rigid-body transformation C is the
identity tensor. The converse is also true; because when C = I, F becomes
an orthogonal tensor that effects only a pure rotation possibly combined
with a translation.
Recalling that the components of F are xi;j9 we find from (5.4.4) that the
components cu of C are
cu = xkiixklj

(5.4.6)

5.4

STRAIN TENSORS

189

Consequently, (5.4.5) reads as follows in the suffix notation:


2 = Ctjafa? = xkiixklJa?a?

(5.4.7)

For a deformation given in the material description, the numbers c^ can


be computed from (5.4.6). The stretch of any material arc whose initial
orientation is known can then be determined by the use of (5.4.7).
5.4.1

NORMAL STRAIN

In many discussions, we will be interested in knowing the changes in the


lengths and relative orientations of material arcs rather than their stretches.
Here we introduce a tensor that enables us to compute these changes.
To this end, we first define a number e as follows:
ds - dsQ
ds0

(5.4.8)

Evidently, e represents the change in length per unit initial length of a


material arc; this number is referred to as the normal strain of the arc. We
note that e is positive or negative depending on whether the element
experiences an extension or a contraction. Also, e = 0 if and only if the arc
retains its length (during a deformation).
From (5.4.5) and (5.4.8), we obtain
1 + e = (2)/2 = (a0 Ca)1/2
= {1 + a-(C - I)a}1/2

(5.4.9)

G = i ( C - I) = i ( F r F - I)

(5.4.10)

e= {1 + 2a-Ga} 1 / 2 - 1

(5.4.11)

If we set
then (5.4.9) becomes

From (5.4.10) we note that G is a symmetric tensor and that G = 0 if and


only if the deformation is a rigid-body transformation. Expression (5.4.11)
shows that G serves as a means of calculating e for any known a 0 . The
tensor G was introduced by Green in 1841 and by B. Saint-Venant in 1844,
and it is known as the Green strain tensor.
By using (5.4.6) and (5.4.10), we obtain the following expression for the
components gu of G:
Su

2(Cij - <*</) = \{xk-jxku

- ij)

(5.4.12)

190

DEFORMATION

Consequently, (5.4.11) reads as follows in the suffix notation:


e={l

+ Ig^a}1'2

- 1

(5.4.13)

For a deformation given in the material description, the numbers gu can


be computed from (5.4.12). The normal strain of any material arc whose
initial orientation is known can then be determined from (5.4.13).
If e is so small that e2 can be neglected, then (5.4.13) becomes
e = gua?af

(5.4.14)

Thus, to the first-order approximation, g^afa*- represents the normal


strain of a material arc that was initially directed along the unit vector a0.
In the particular case when a0 is aligned with e,, we have ax = 1,
a
2 = al = 0. Expressions (5.4.7) and (5.4.14) then reduce to 2 = cn and
e gn. Thus, the component c n of C represents the square of the stretch
of a material arc that was initially directed along the 1 axis, and to the
first-approximation, the component gn of G represents the normal strain of
this arc. The components c22 and c33 of C and g22 and g33 of G have similar
geometrical meanings.
5.4.2

SHEAR STRAIN

In order to analyze the changes in the relative orientations of material arcs,


let us consider another material arc dC initiating from the particle <P and
directed along the unit vectors and in the initial and final configurations
respectively. If is the stretch and is the normal strain of this arc, then by
virtue of the relations (5.4.1), (5.4.7), (5.4.8) and (5.4.13), we have the
following expressions:
= F

(5.4.1)'

2 = CtjW

(5.4.7)'

= 1+

(5.4.8)'

={\

+ 2gua?i$}l/2 - 1

(5.4.13)'

If 0 is the angle between d<5 and f in the initial configuration and is


the angle between them in the final configuration, we have (see Figure 5.10)
cos 0O = a0 = a? ?
cos = a = ii

(5.4.15a)
(5.4.15b)

5.4

STRAIN TENSORS

191

Figure 5.10. Initial and final orientations of material arcs.

From (5.4.15b), (5.4.1), (5.4.1)' and (5.4.4) we obtain


cos = //a *
= Fa0 F
= a0 F r F
= a-C = cua?a}

(5.4.16)

Expressions (5.4.7), (5.4.7)' and (5.4.16) show that is completely determined by cu. Thus, the deformation components cu not only determine the
stretch of a material arc but also the final angle between two material arcs
whose initial orientations are known.
Using (5.4.12) and (5.4.15a), expression (5.4.16) can be rewritten in terms
of gjj as follows:
cos - cos 0O = 2gu<$cf}

(5.4.17)

Evidently, and 0 are different from one another in general. This means
that the relative orientations of material arcs generally change under a
deformation. If we set
2 = 0-

(5.4.18)

then y represents one-half the reduction in the angle between the arcs rfC
and rf due to deformation. The number y is called the shear strain between
these arcs.
If 0 = /2, expression (5.4.17) becomes, on using (5.4.8), (5.4.8)' and
(5.4.18),
(1 + ?)(l + )sin2y = 2guafaf
(5.4.19)
Expressions (5.4.13), (5.4.13)' and (5.4.19) show that y is completely determined by gij. Thus, the strain components gu not only determine the normal
strain of a material arc but also the shear strain between two initially
orthogonal arcs.
If y, e and are so small that nonlinear terms involving these can be
neglected, then (5.4.19) reduces to
y = gijaf*

(5.4.20)

192

DEFORMATION

Thus, to the first-order approximation, gijafj represents the shear strain


between the material arcs that were initially directed along two orthogonal
unit vectors a0 and .
In the particular case when a0 is aligned with e! and is aligned with e 2 ,
we have aQx = 1, a\ = a? = 0 and 2 = 1, ? = ? = 0. Then, together
with (5.4.7) and (5.4.7)', expression (5.4.16) yields c12 = Vc n c 22 cos 0, and
(5.4.20) yields g12 = . Thus, cn represents the product of the stretches suffered by the material arcs that were initially directed along the 1 and 2 axes
and the cosine of their relative orientation in the final configuration. Also,
to the first-order approximation, gn represents the shear strain between
these arcs. Other off-diagonal components of C and G have similar geometrical meanings.
Thus, the components of the two tensors C and G, whose knowledge
enables us to compute the changes in length and relative orientation of
material arcs, have definite geometrical meanings. While the diagonal components of both the tensors are associated with the changes in length of
material arcs, which were initially laying along the coordinate directions,
the off-diagonal components are associated with the changes in their
relative orientations.
5.4.3

EXPRESSIONS IN SPATIAL FORM

We observe that (5.4.5) and (5.4.13), which give the stretch and the normal
strain of a material arc, and (5.4.16) and (5.4.17), which give the final angle
and the change in the angle between two material arcs, are all expressed in
terms of the initial orientations of the arcs. Here we deduce their counterparts expressed in terms of the final orientations of the arcs.
From (5.4.1), we obtain
(5.4.21)
a0 = //F_1a
from which it follows that
-L = jLa 0 -a = F- 1 a-F- 1 a
= a (F-yOF^a)
= a-iFF7)-^
If we set

(5.4.22)

B = FF r

(5.4.23)

-2 = a-B _ 1 a.

(5.4.24)

then (5.4.22) becomes

5.4

STRAIN TENSORS

193

From (5.4.23), (5.4.24) and (5.3.6b) we find that B is a positive definite


symmetric tensor and that B = V2. Expression (5.4.24) shows that the
tensor B" 1 serves as a means of calculating when a is known. It can be
easily shown that for a rigid-body transformation, B = B" 1 = I, and conversely. The tensor B" 1 was introduced by Cauchy in 1827 and is known as
the Cauchy deformation tensor. The tensor B was introduced by Finger in
1894 and is known as the Finger deformation tensor. However, C and B are
commonly referred to as the right and left Cauchy-Green deformation
tensors in view of their direct connection with the right and left polar
decompositions of F.
Recalling that xi;j and x?-} are components of F and F" 1 , respectively, we
find from (5.4.23) that the components bu of B and the components J1 of
B" 1 are given by
bu = xi;kXj;k

(5.4.25a)

bl=49i4,j

(5.4.25b)

Consequently, (5.4.24) reads as follows in the suffix notation:


-2 = b^diOj = xl^l^aj

(5.4.26)

For a deformation given in spatial description, the numbers bj]1 can be


computed from (5.4.25b). The stretch of any material arc whose final
orientation is known can be determined by the use of (5.4.26).
From (5.4.8) and (5.4.24) we obtain
1 + e = (2)/2 = (a-B^a)" 1 7 2
= [1 - a (I - B - 1 )ar 1 / 2

(5.4.27)

Noting from (5.4.23) that


B" 1 = (FF 7 )" 1 = ( F - y F " 1
and setting

A = i(I - B"1) = ill - (F-YF" 1 )

(5.4.28)

we find from (5.4.27) that


e = (1 - 2a-Aa)" 1 / 2 - 1

(5.4.29)

From (5.4.28) we note that A is a symmetric tensor and that A = 0 if and


only if the deformation is a rigid-body transformation. Expression (5.4.29)
shows that A serves as a means of computing e for any known a. The tensor
A was introduced by E. Almansi in 1911 and by G. Hamel in 1912 and is
known as Almansi strain tensor.

194

DEFORMATION

By using (5.4.25b) and (5.4.28) we obtain the following expression for the
components a^ of A:

*(/ = Wu - bl) = fiu - 4.i4.j)

(5.4.30)

Consequently, (5.4.29) reads as follows in the suffix notation:


e = (1 - la^jY^1 - 1

(5.4.31)

For a deformation given in spatial description the numbers a^ can be


computed from (5.4.30). The normal strain of any material arc whose final
orientation is known can be determined from (5.4.31).
If e is so small that e1 can be neglected, (5.4.31) becomes
e a^j

(5.4.32)

Thus, to the first-order approximation, ^^ represents the normal


strain of a material arc that is finally directed along the unit vector a.
Computations analogous to those employed in obtaining (5.4.16) and
(5.4.17) lead to the following expressions:
(5.4.33)
(5.4.34)
When = /2, (5.4.34) becomes, on using (5.4.8), (5.4.8)' and (5.4.18),
(5.4.35)
Expressions (5.4.31) and (5.4.35) show that y is completely determined by
dij. Thus, the strain components atj not only determine the normal strain of
a material arc but also determine the shear strain between two finally
orthogonal material arcs.
If y, e and are so small that nonlinear terms involving these can be
ignored, then (5.4.35) reduces to the following form:
(5.4.36)
Thus, to the first-order approximation, ^ represents the shear strain
between the material arcs that are finally directed along two orthogonal unit
vectors a and .
By using (5.4.32) and (5.4.36), it can be verified that, to the first-order
approximation, an represents the normal strain of a material arc that is
finally oriented along the 1 direction, and al2 represents the shear strain
between two material arcs that are finally oriented along the 1 and 2
directions. Other components of A have similar geometrical meanings.

5.4

STRAIN TENSORS

195

Thus, the diagonal components of A are associated with the normal


strains of material arcs that are finally directed along the coordinate directions and the off-diagonal components of A are associated with the shear
strains between these arcs.
It is to be noted that the tensors C and G are defined in terms of the
material deformation gradient F. As such, C and G are tensors associated
with the material (Lagrangian) description of deformation. For this reason,
C and G are often referred to as the Lagrangian deformation tensor and
Lagrangian strain tensor, respectively. On the other hand, the tensors B" 1
and A are defined in terms of the spatial deformation gradient F" 1 . As
such, B" 1 and A are tensors associated with the spatial (Eulerian) description of deformation. Therefore, B _ 1 and A are often referred to as the
Eulerian deformation tensor and the Eulerian strain tensor, respectively.
EXAMPLE 5.4.1 For the deformation considered in Example 5.2.1,
find the tensors C and G. Hence find (i) the stretch and the normal strain
of a material arc rfC whose initial direction ratios are 1:1:1, and (ii) the
change in the angle between this arc and the material arc dC whose initial
direction ratios are - 1 : 2 : - 1 .
Solution
we find

For the given deformation, [F] is given by (5.2.24), from which

[Cijl = [C] = [F r F] =

o --l 1

6 --1

1J

- 1 --1
Hence

[gij] = [G] = | [ C - I] =
L

(5.4.37)

5
2

2"

1
2

1
2

~2

(5.4.38)

For the given deformation, (5.4.37) and (5.4.38) give the matrices of the
tensors C and G, respectively.
For the material arc dG, we have

? = a2

1
V3

(5.4.39)

Using these a? and cu given by (5.4.37) in (5.4.7), we get = V2 1.4142.


This is the stretch of the a r c d e . Its normal strain is e = - 1 0.4142.
Since e > 0, dQ experiences an extension.

196

5 DEFORMATION

For the material arc rfC, we have

-?=-

vs

- *

- - *

(5.4.40)

Using these <zf and c# given by (5.4.37), we get the stretch of dC as


= V5 2.2360 and the normal strain of tf as = 77 - 1 1.2360.
From (5.4.39) and (5.4.40), we find that a0 = 0. Therefore, de and
de are initially orthogonal. Using the values = V2 and = y obtained
previously and gu given by (5.4.38) in (5.4.17), we find that cos 0.6708,
or 47.9. This is the final angle between de and de. Thus, the deformation causes a decrease of about 42.1 in the angle between d<2> and
de.
M
EXAMPLE 5-4.2 For the deformation considered in Example 5.2.2,
find the tensor B" 1 . Show that one of the principal stretches is equal to 1
and that the squares of the other two principal stretches are the two roots
of the quadratic equation
(5.4.41)

xz - [R1 + i l * + 1 = 0
Deduce that detB" 1 = 1.

Solution For the given deformation, F" 1 is given by (5.2.29). Using this
in (5.4.23) we obtain

IB"1] = [F' 1 ] r [F- 1 ] =

-
1 -

~ F J*1*2
(5.4.42)

^)*1*2

This determines the tensor B" 1 .


The principal stretches are, by definition, the eigenvalues of V = VB.
Hence, if = VX is a principal stretch, then is an eigenvalue of B so
that 1/ is an eigenvalue of B" 1 . Thus, in the present case, l/ must be a
root of the following characteristic equation:

5.4 STRAIN TENSORS

197

= 0

1 Simplifying this equation, we find that must satisfy the following


equation:
( - 1)

l1 - (*+ p + 1

= 0

(5.4.43)

It is evident that one of the principal stretches is VX", = 1 and that the
squares of the other two principal stretches, 2 and 3 , are the roots of the
equation (5.4.41).
From equation (5.4.41) we note that 23 = 1, so that A t A 2 A 3 = 1.
Hence
det(B_1) = (A.AzAa) -1 = 1
EXAMPLE 5.4.3

Show that
\2

/ j \2

(dsy-(ds0f

Solution

0 JV0
= 2gijdx?dx>

(5.4.44)

(5.4.45)

IcijjdXidXj

Recalling from (5.3.2) and (5.4.8) that


dxf = (ds0)af,

ds
1+ e = =
ds0

dXi = {ds)a

we find by use of (5.4.13) and (5.4.31) that


(i)

(ds)2 - (ds0)2 = (ds0f


2

= (<fc0) [(l + ef - 1]
= 2(ds0)2gija?a?
= 2gijdxfdxJ

(5.4.46)

198

DEFORMATION

and
(ds)2 - (ds0)2 = (ds)2\l -

(ii)

(^

= (ds)2[\ - (i + *r z ]
= 2(dsfauaiaj
=

IdijdxidXj

Thus, (5.4.44) and (5.4.45) are proven.

Note: When gtj are known, (5.4.44) gives the change in the square of the
length of a material arc that was initially located at x. When a^ are known,
(5.4.45) gives the change in the square of the length of a material arc that
is finally located at x. Some authors use (5.4.44) and (5.4.45) for defining
the Lagrangian and Eulerian strain tensors, respectively.
EXAMPLE 5 . 4 . 4
and that

Show that B and C have the same principal invariants

/// B = IIIC = J2,

IIB = IIC = 7 2 (trC _ 1 )

(5.4.47)

Solution Recall that U and V have the same eigenvalues, which are the
principal stretches, and U 2 = C and V2 = B. Hence, if ,2, r\z are the
principal stretches, then r\\, \ and \ are the eigenvalues of both B and C,
and A/J"2, A/22 a n d >/J2 a r e the eigenvalues of both B _ 1 and C _ 1 . Consequently, by use of (2.13.14), we obtain

h = ic = n\ + *\\ + n\

//, = nc = n\n\ + n\n\ + n\n\

(5.4.48)

IIIB = IIIC = \\\\


which show that B and C have the same principal invariants. Also,
trCT 1 = / c - i = /72 + ni2 + ni1
= m2ri22ri32(riW2 + \\\ + \2)
= ^-//c

(5-4.49)

Further,
IIIC = det C = det(F r F) = (det F) 2 = J2
Expressions (5.4.48), (5.4.49) and (5.4.50) yield (5.4.47).

(5.4.50)

5.4 STRAIN TENSORS

199

EXAMPLE 5.4.5 Show that the stretch of a material arc element d<5
assumes an extreme value when d(5 initially lies along a principal direction
of C.
Solution In terms of its initial orientation, the stretch of dQ is given by
(5.4.7). We have to find af for which is an extremum. Since af are components of a unit vector we have afaf = 1. Thus, 2 assumes extreme values
when af obey the equations
-[2 - a(akak- l)j = 0

(5.4.51)

where a is a Lagrangian multiplier.


Substituting for 2 from (5.4.7) in (5.4.51) and noting that (da?/da?) =
Sij, we obtain the following conditions on af under which 2 is an
extremum:
(cu - au)aj = 0
(5.4.52)
These conditions are satisfied iff a0 is an eigenvector of C. Thus, is an
extremum when d<5 initially lies along a principal direction of C.
EXAMPLE 5.4.6 Consider a material surface element dS that is initially
orthogonal to a unit vector n. After deformation, suppose that it becomes
orthogonal to a unit vector n. If dS0 and dS are the initial and final areas
of dS, prove that

(dsy2
= y2n T V ) = J2{n 1
(dS0f
Solution
follows:

(5.4.53)

From (5.2.12), let us recall that dS0 and dS are related as


(</S)n = J(dS0)(T)-ln

(5.4.54)

Since n is a unit vector, (5.4.54) yields


(dS)2 = (rfS)n (rfS)n
= / 2 ( i /So) 2 ((F"yn-(F- 1 ) r n 0 )
= /VS 0 ) 2 {n ( F ^ r ' n 0 )

(5.4.55)

which on using (5.4.4) becomes


^

= yV-(C-'n)

(5.4.56)

200

DEFORMATION

By use of (5.4.54), we also get


J2(dS0)2 = J(dS0)n J(dS0)n = (rfS)2(Frn) (F r n)
= (dS)2n (FFM

(5.4.57)

which on using (5.4.23) becomes


^ 5

= / 2 (n-Bn)-'

Expressions (5.4.56) and (5.4.58) prove (5.4.53).

(5.4.58)

The results (5.4.54) and (5.4.53) are by E. J. Nanson (1878), known as


Nanson 's formulae,

5.5
STRAIN-DISPLACEMENT RELATIONS
As in Section 5.2, let P0 be the initial position and P be the final position of
a particle (P. Then the directed line segment P0P is called the displacement
vector of (P from the initial configuration to the final configuration. We
denote it by u. If x is the position vector of P0 and x that of P, then we have
(see Figure 5.11)
u = P^P = x - x
(5.5.1)
In the material description, x is a function of x and t. Hence, in this
description, u is a function of x and t so that (5.5.1) stands for
u(x, t) = x(x, 0 - x

Figure 5.11. Displacement vector.

(5.5.2)

5.5

STRAIN-DISPLACEMENT RELATIONS

201

This expression specifies the displacement of the particle (P at time t in


terms of its initial position x.
In the spatial description, x is a function of x and t. Hence, in this
description, u is a function of x and / so that (5.5.1) now stands for
u(x, 0 = x - x(x, 0

(5.5.3)

This expression specifies the displacement of the particle <P at time / in


terms of its spatial position x. In this case, u is referred to as the displacement at the point x.
We now proceed to deduce relations connecting the deformation and
strain tensors with the displacement vector. We find from (5.5.2) that
Vu = Vx - Vx = F - I.

(5.5.4)

The tensor Vu is referred to as the displacement-gradient tensor in the


material description, or briefly the material displacement-gradient.
From (5.5.4) and (5.4.4) we get
C = F r F = (Vu r + I)(Vu + I)
= I + Vu + Vu r + (Vur)(Vu)

(5.5.5)

Consequently, (5.4.10) yields


G = i ( C - I) = i[Vu + Vu r + (Vur)(Vu)]

(5.5.6)

This is called the strain-displacement relation in the material description.


In the suffix notation, (5.5.4) to (5.5.6) read respectively as follows:
"iiJ = *i;J -
Cij = u + ui;J + Uj.j + uk,iUk,j

( 5 5 4 )'
(5.5.5)'

8ij = l&iJ + UJM + Uk;ik;j)

(5.5.6)'

When the displacement components ut are known as functions of xf9


(5.5.4)' to (5.5.6)' may be employed to compute the components of Vu, C
and G, respectively.
From (5.5.3) we find that
Vu = Vx - Vx = I - F"1

(5.5.7)

The tensor Vu is referred to as the displacement-gradient tensor in the


spatial description, or briefly the spatial displacement gradient.
From (5.4.23) and (5.5.7) we obtain
B"1 = ( F - y F - 1 = (I - Vur)(I - Vu)
= I - Vu - Vu r + (Vur)(Vu)

(5.5.8)

202

5 DEFORMATION

Consequently, (5.4.28) yields


A = i ( I - B 1 ) = |[Vu + Vu r - (Vur)(Vu)]

(5.5.9)

This is called the strain-displacement relation in the spatial description.


In the suffix notation, (5.5.7) to (5.5.9) read respectively as follows:
u

u = -

Ul
a

= -

U = & UU

x
U

(5.5.7)'

U -

JJ +

J,i ~

(5.5.8)'

k,iukJ

(5.5.9)'

k,iukj)

When Ui are known as functions of*,, (5.5.7)' to (5.5.9)' may be employed


to compute the components of Vu, B" 1 and A, respectively.
EXAMPLE 5.5.1 In a certain deformation, the displacement components are found to be
Ul

= 3(JC?)2 + jcj,

u2 = 2 ( 4 ) 2 + 4 ,

w3 = 4(JC?)2 + jcf

(5.5.10)

Find (i) the Green strain tensor at the origin O, and (ii) the stretch and the
final length of a material arc of initial length ds0 that was initially based at
O and directed along the line making equal angles with the axes.
Solution (i) From (5.5.10), we find that the matrix of the displacementgradient Vu is
[Vu] = [ui;J] =

6x?

Ax\

8x?

(5.5.11)

By virtue of (5.5.6) and (5.5.6)' the matrix of the Green strain tensor G
is given by
2[gij] = 2[G] = [Vu] + [Vu] r + [Vu]r[Vu]

(5.5.12)

Substituting for [Vu] from (5.5.11) in (5.5.12), we obtain


2[gij\ = 2[G]

1 + 6x? + 36(x?) 2
1 + 6?
1 + 6x
1 + Sx2 + 16(x 2 0 ) 2
1 + 8x3

1 + Ax2

1 + 8x3
1 + 4x2

1 + 16*?

+ 64(?) 2 J

(5.5.13)

5.5

STRAIN-DISPLACEMENT RELATIONS

203

At the origin, this becomes


2fe(/] = 2[G] =

1 1

1 1

1 1

(5.5.14)

(ii) The stretch of a material arc whose initial orientation is given may be
computed by using the following expression obtained from (5.4.2), (5.4.7),
and (5.4.12):

n =

(ds)2
(*o)

= (2gij + }?

= CuaX

(5.5.15)

For the given arc, we have a*} = a% = a% = 1/V3. Using these a? and
substituting for gy from (5.5.14) in (5.5.15), we get the stretch of the arc as
= 2. Hence the final length of the arc is ds = 2 ds0.

EXAMPLE 5 . 5 . 2

(i) Show that


/ = det(I + Vu) = {det(I - Vu)}"1

(5.5.16)

(ii) Deduce that the deformation for which the displacements (displacement components) are
ul=x1+x2,

u2 = xi-x2i

(5.5.17)

u2 = 2x3

is isochoric.
Solution

(i) From (5.5.4) and (5.5.7), we have


F = Vu + I

(5.5.18a)

F _ 1 = I - Vu

(5.5.18b)
1

Recalling that / = detF and (1/7) = detF" , we readily get (5.5.16) from
(5.5.18).
(ii) From (5.5.18b), we have
- = d e t F " 1 = det(J l7 -

(5.5.19)

u u)

Computing uitj from (5.5.17) and substituting them in (5.5.19), we get


1
J

0-1
-1

01
0 = 1

0 -1 I

Hence, / = 1 so that the given deformation is isochoric.

204

5 DEFORMATION

EXAMPLE 5 . 5 . 3

Prove the identities:


[curl u], = - eJn xi;rxJ;
=

Solution

(5.5.20)

~JEkpqXp;kXq;i

(5.5.21)

Since ut = xt - xf, we get


[curlu], = eUkukJ

= eiJk(xkJ

- xkJ)

= 6ijkxkj = ijkXjtk

(5.5.22)

By using the identity (5.2.32), the relation (5.5.22) becomes


[curl u]f = eijkeJrsekpqxp.rxq.s

(5.5.23)

Employing the - identity (1.7.23) to the factor eijkekpq9


(5.5.23) becomes
[CUrlll], = -^jZjrstfipjq -

~ ~ j jrs\Xi;rXj;

_ 1

~7 ^jrsxi

expression

iqjp)Xp.rXq.s

i;sXj;r)

s ~~

;rxj;s

which proves the identity (5.5.20).


On the other hand, if we employ the - identity to the factor
expression (5.5.23) becomes
[curlu], = zkpq(kris
1
^ Y ^kpq\Xp;
27
1

~7 ^kpqXp;

which proves the identity (5.5.21).

k Xq; i

kXq; i

ksir)xp.rxq.s

q; k Xp\

i)

eijkejrs9

5.6

INFINITESIMAL STRAIN TENSOR

205

5.6
INFINITESIMAL STRAIN TENSOR
In great many applications, particularly those concerned with common
elastic materials, we deal with deformations wherein the derivatives of
displacement components are so small that their squares and products can
be ignored. Such deformations are referred to as infinitesimal or small
deformations. A deformation that is not small is referred to as a finite
deformation. In Sections 5.2 to 5.5 we dealt with finite deformations. From
now on we will deal with infinitesimal deformations.
Using the fact that ut = Ui(xk,t) in the material description and that
** = * * - "*> we get
Uij = ui;kxkJ = ui;k(xk - uk)j = (ui;j - ui;kukJ)
(5.6.1)
For small deformation, we neglect the nonlinear terms ui;kukj9
(5.6.1) becomes
uu = Uij or Vu = Vu

so that
(5.6.2)

Thus, for small deformation the material displacement gradient and the
spatial displacement gradient are (almost) equal. Accordingly, in dealing
with small deformation it is immaterial whether the displacement gradient
is formed w.r.t. x or w.r.t. x.
For small deformation, (5.5.5)', (5.5.6)', (5.5.8)' and (5.5.9)' yield
Cij

- ij = ij - bj1 = Igij = 2au = ui;J + uj.s = uu + uhi

(5.6.3)

or, in the direct notation,


C - I = I - B 1 = 2G = 2A = Vu + Vur = Vu + VuT (5.6.3)'
From these expressions we find that the tensors G and A are identical. As
such, for small deformation, the distinction between the Lagrangian and
the Eulerian strain tensors disappears. Hence in dealing with small deformation it is immaterial whether the strain components are referred to the
initial configuration or the final configuration.
In the case of small deformation, the tensor G (or A) is redesignated as
E and referred to as the infinitesimal or linear strain tensor. Thus, from
(5.6.3)' we get the following strain-displacement relation for small
deformation:
E = (Vu + Vu7) = i(Vu + V0ur)

(5.6.4)

Evidently, E is just the symmetric part of Vu = Vu, and the symmetry of


E is inherent in its definition itself. The tensor E was introduced by Cauchy

206

5 DEFORMATION

in 1827 and is also referred to as the Cauchy's strain tensor. The components of E, called infinitesimal strain components or Cauchy's strain
components, are denoted e^, Thus,
eu = gij = au = \{Uij + ujj) = \iuij

+ Uj.j)

(5.6 .4)'

Let us recall expression (5.4.13) for the normal strain e and rewrite it in
terms of ui;j by using (5.6.3) to obtain
e = (1 + liiijafa?)1'2

- 1

(5.6.5)

Since ui;J are taken to be infinitesimally small, the first term in the
righthand side of (5.6.5) can be expanded by using the binomial theorem.
On carrying out this expansion, neglecting nonlinear terms involving wi;7
and subsequently using (5.6.3) again we obtain
e = utjafa? =

(5.6.6)

gua?af

We find that this expression is identical with the expression (5.4.14), which
was obtained by neglecting the squares and higher powers of e. It therefore
follows that the small displacement-gradient approximation and the small
strain approximation are of the same order. As such, e^ have the same
geometrical meanings as gy have under the small strain approximation.
Thus, for example, en represents the normal strain of a material arc that
initially lay along the 1 axis and e12 represents the shear strain between two
material arcs that initially lay along the 1 and 2 axes. The numbers eu, e22
and e33 are called infinitesimal normal strains, and el2 = e2X, en = en and
^23 = ei2 a r e called infinitesimal shear strains,

5.6.1

DILATATION

From (5.5.18a), we have


1 + i;i

/ = det F = det(Jl7 + ui;j) =

"l;2

"l;3

2;l

1 + H2; 2

2; 3

"3;l

3; 2

1 + u3]3

(5.6.7)

Expanding the determinant at the far righthand side of (5.6.7) and neglecting the nonlinear terms in displacement gradients, we obtain the following
expression valid for small deformation:
/= 1 + uk,k=

1 +ekk

(5.6.8)

By use of the Euler's formula (5.2.22), dV = JdV0, expression (5.6.8) can

5.6

INFINITESIMAL STRAIN TENSOR

207

be rewritten as
dV - dV0
dV0

(5.6.9)

= ekk

Thus, ekk represents the change in volume per unit initial volume of a
material element undergoing small deformation. The number ekk is called
the cubical dilatation or just the dilatation.
We note from (5.6.4) and (5.6.4)' that
ekk = t r E = divu = divu

(5.6.10)

where div denotes the divergence w.r.t. the initial coordinates x. Thus, for
small deformation, tr E = divu = div u represents the change in volume
per unit initial volume.
From (5.6.8) and (5.6.10), we infer that for isochoric small deformation
ekk = divu = divu = 0 and for an incompressible continuum ekk =
divu = div u = 0 for every small deformation.
EXAMPLE 5.6.1

For small deformation, prove the following:

(i)
C
(ii)

C B I + 2E

(5.6.11a)

(5.6.11b)

B~l I - 2E

U V I + E
1

U" V * I - E
(iii)

Q F - E
1

Q1 F
Solution

(5.6.12a)
(5.6.12b)
((5.6.13a)

+ E

(5.6.13b)

(i) From (5.6.3)' and (5.6.4), we get


C = I + 2E
1

B" = I - 2E

(5.6.14a)
(5.6.14b)

From (5.5.18) and (5.6.4), we get


B = F F r = (Vu + I)(V 0 u r + I) I + Vu + Vu r = I + 2E

(5.6.15a)

and
C

= ( F r F ) _ 1 = -l(F'l)T

= (I - Vu)(I - Vu7)

I - Vu - Vu r = I - 2E
Expressions (5.6.14) and (5.6.15) constitute the relations (5.6.11).

(5.6.15b)

208

5 DEFORMATION

(ii) Expressions (5.6.11) may be rewritten as

C ~ B (I + 2E) (I + E)(I + E) = (I + E) 2

(5.6.16a)

(5.6.16b)

(I - 2E) (I - E)(I - E) = (I - E) 2

Recalling that U 2 = C, V2 = B, ( 1 ) 2 = C" 1 and (V - 1 ) 2 = B \


immediately get (5.6.12) from (5.6.16).
(iii) On using (5.5.18a) and (5.6.12b), expression (5.3.5a) yields

we

Q = F I T 1 = (Vu + I)(I - E )
Vu + I - E = F - E
which is (5.6.13a).
On using (5.5.18b) and (5.6.12a), expression (5.3.5b) yields
Q-i

= F

- i v = (i _ vu)(I + E)

= I - Vu + E = F" 1 + E
which is (5.6.13b).

EXAMPLE 5.6-2 Show that a small deformation for which E = 0 represents a rigid-body transformation. Deduce that in this case the displacement
can be represented in the form
u = x x + c

(5.6.17)

where c is a vector not dependent on x and is a vector such that


(Vu)a = x a

(5.6.18)

for every vector a.


Show also that if two displacement vectors u and u' correspond to the
same infinitesimal strain tensor, then u - u' is a rigid displacement. (This
result is known as Kirchhoff's theorem.)
Solution If E = 0, then we find from (5.6.12a) that U = V = I. Therefore, the deformation is a rigid-body transformation.
Further, when E = 0, we find from (5.6.4) that Vu is a skew tensor, so
that ui9j = -Ujj from which it follows that
u

ijk

i,kj

~uk,ij

~ukji

j,ki

jjk

~UiJk

or
"ijk = 0

(5.6.19)

5.6

INFINITESIMAL STRAIN TENSOR

209

Figure 5.12. Rigid rotation.

The displacement components w, must therefore be linear functions of A:,;


that is
"i = QikXk + Ct
(5.6.20)
where aik and ct are independent of xk. We easily note from (5.6.20) that
ik - ui,k'> hence, (5.6.20) can be rewritten as

"/ = Ui%kXk + q

or

u = (Vu)x + c

(5.6.21)

Since Vu is a skew tensor, there exists a unique dual vector such that
(5.6.18) holds for every vector a. Using (5.6.18) in (5.6.21) we obtain (5.6.17).
We observe that u given by (5.6.17) consists of two parts: c and x x.
The part c represents a rigid translation while the part x x represents a
rigid rotation through an angle || about the axis directed along (see
Figure 5.12).
If two displacement vectors u and u' correspond to the same infinitesimal
strain tensor, then u - u' corresponds to the zero infinitesimal strain. From
the result just proven, it follows that u - u' is a rigid displacement.

EXAMPLE 5 . 6 . 3 Find E for the rigid rotation defined by equations


(5.3.21) of Example 5.3.3.
Solution From (5.3.21) we obtain the following expressions for the
displacement components associated with the rotation considered here.
ux = x1 - xx

= ATJ^COS 0 - 1 )

u2 = x2 - *2 =

*i

u3 = x3 - x3 = 0

s n

+ x2 sin

+ AT20(COS -

1)

(5.6.22)

210

5 DEFORMATION

These give
cos - 1
-sin

[Vu] =

sin

0
(5.6.23)

cos - 1 0

Hence the strain-displacement relation (5.6.4) yields


0

" cos - 1

cos -- 1 0

[E] = |[Vu] + i[Vu] =

(5.6.24)

This is the matrix of the infinitesimal strain tensor E for the given rigid
rotation. Evidently, E is a nonzero tensor (for ^ 0).

Note: This example illustrates the important fact that, unlike the finite
strain tensors G and A, the infinitesimal strain tensor E need not be 0 for
a rigid-body transformation. The reason for this is that E is not an exact
measure of deformation.
EXAMPLE 5 . 6 . 4 Let V0 be the initial volume and V be the final volume
of a material volume V undergoing small deformation. If / = /(x f ), show
that
(1 +

fdV=

ekk)fdV0

(5.6.25)

Hence deduce that the change in volume of V is given by


V=
Solution
Therefore

ekkdV

ekkdV0

(5.6.26)

From the Euler's formula (5.2.22) let us recall that dV = JdV0.


fdV

JfdV0

(5.6.27)

Substituting for / from (5.6.8) in (5.6.27) we immediately obtain (5.6.25).


F o r / = 1, (5.6.25) becomes
dV=\

(1

+ekk)dV0

5.7

INFINITESIMAL STRETCH AND ROTATION

211

so that
V =V-V0=\

Jv

dV-\

J v0

dV0=\

Jv0

ekk dV0

(5.6.28)

T a k i n g / = ekk in (5.6.25), we get


ekkdV=

(1 + ekk)ekkdV0

which together with (5.6.28) yields (5.6.26).


EXAMPLE 5 . 6 . 5

ekkdV0

Show that for small deformation


n dS = n dS0

Solution

(5.6.29)

(5.6.30)

From (5.2.12) let us recall that

TndS = Jn dS0
(5.6.31)
Substituting for F from (5.5.4) and for J from (5.6.8) together with
(5.6.10) in (5.6.31), we get
n dS + (Vu)rn dS = n dS0 + (divu)n dS0

(5.6.32)

Equating the terms of the same order in displacements on the two sides of
(5.6.32), we obtain (5.6.30).

5-7
INFINITESIMAL STRETCH AND ROTATION
In Section 5.3, it was seen that a transformation of a material arc element
dG from its initial location dC0 to the final location dC under a deformation may be decomposed into two parts: the first part described by (5.3.8a),
representing a triaxial stretch of the element; and the second part described
by (5.3.8b), representing a rotation of the element. Let us now reexamine
these two parts in the case of small deformation.
For small deformation, we recall from (5.6.12a) that
U = E + I

(5.7.1)

Consequently, (5.3.8a) becomes


dy = d\ + E dx

(5.7.2)

This equation shows that the first part of the transformation (which
carries d<5 from its initial location dC0 to the intermediate location dC) is

212

5 DEFORMATION

Figure 5.13. Translation preceding by stretch.

now made up of a (pure) translation and a stretch effected by E. The


translation may follow or precede the stretch. Translation preceding by
stretch is depicted in Figure 5.13.
By use of (5.7.1) and (5.3.9), we find that
3

Edx=

nk(PkPk)dx-dx

k=\

ek(PkPk)dx

k=l

where

= ek(dx P*)P*

(5.7.3)

ek = nk-\

(5.7.4)

k= 1

Expression (5.7.3) shows that the stretch effected by E is made up of three


simple elongations directed along the principal directions of stretch with ek
as the corresponding normal strains.
Let us now look at the second part of the transformation (which carries
dQ from the intermediate location dC to the final location dC) given by
(5.3.8b). By use of (5.6.13a), (5.5.4) and (5.6.4), expression (5.3.8b)
becomes
dx = ( - E) dy = (I + Vu - E) dy
= dy + Qdy

(5.7.5)

= |(Vu - Vur) = skw Vu

(5.7.6)

where
Since is a skew tensor, (5.7.5) may be rewritten as
dx = dy + x dy

(5.7.7)

where is the dual vector of . Expression (5.7.7) shows that the second
part of the transformation of dQ is now made up of a (pure) translation and

5.7

INFINITESIMAL STRETCH AND ROTATION

213

Figure 5.14. Translation following rotation.

a pure rotation through an angle || about an axis directed along . The


translation may follow or precede rotation. Translation following rotation
is depicted in Figure 5.14.
From this analysis, it follows that, under small deformation, a material
arc element dQ may be thought as subjected to a pure translation and a
triaxial elongation as described by (5.7.2) followed by another pure translation and a pure rotation as described by (5.7.7). This interpretation is
illustrated in Figure 5.15.
It is to be noted that, in this analysis, the stretch part of the transformation is represented by the strain tensor E, which is the symmetric part of
Vu, and the rotation part is represented by the tensor , which is the skew
part of Vu. The tensor is called the infinitesimal or linear rotation
tensor. The vector , which is the dual of , is called the infinitesimal or
linear rotation vector.
Since Vu Vu for a small deformation, the expression (5.7.6) defining
may be replaced by the following expression:
= skwVu = |(Vu - VuT)

(5.7.8)

Translations
Figure 5.15. Translation and elongation followed by another translation and rotation.

214

DEFORMATION

Recalling that the dual vector of skw Vu is \ curl u (see Example 3.5.7),
we obtain the following expression for the linear rotation vector:
0)

y curl u y curlu

(5.7.9)

where curl0 is the curl taken w.r.t. JC.


If coy and , are the components of and , respectively, then (5.7.6),
(5.7.8) and (5.7.9) yield the following expressions for ^ and , in terms
of Ui :
i.
, = Wi-J

EXAMPLE 5-7-1

J;i)

2ijkUkJ

2ijkUk;j

HUiJ -

Jj)

(5.7.10)
(5.7.11)

For the small deformation defined by


(5.7.12)

u = <xxix2(ti + e2) 4- 2a(x! + Ar2)Ar3e3

where a is a constant, find the strain tensor, the rotation tensor and the
rotation vector.

Solution From (5.7.12), we find that


[Vu] = a

x2

xx

x2

x1

(5.7.13)

2x3 2x3 2(x + * 2 )_

Hence

*2

[E] = [Vu] + |[Vu] = a (*1 +


L

i(Xi
Xi)

Xi)

x3
2(, + x2)
(5.7.14)

Xi

x3

and

[] = j[Vu] - |[Vu] = a -i(*i -

z(Xi
*i)

*3

Xi) -Xi'

0
Xi

-Xi

(5.7.15)

For the given deformation, relations (5.7.14) and (5.7.15) give the matrices
of the strain and rotation tensors respectively. Also, the rotation vector,
obtained from (5.7.12), is
= j curl u = ocx3(ei - e2) + \a(x2 - *i)e 3

(5.7.16)

5.7

EXAMPLE 5-7-2

INFINITESIMAL STRETCH AND ROTATION

21 5

Consider the displacement components


ux = axxx + bxx2 + cxx3
(5.7.17)

u2 = a2Xi + b2x2 + c2x3


u3 = a3xx + b3x2 + c3x3

where ai9 bi and cf are constants. Show that these displacement components
correspond to (i) a pure deformation if and only if
b\ = a2,

cx = a3,

(5.7.18)

c2 = *3

and (ii) a pure rotation if and only if


ax = b2 = c 3 = 0,

j = -a2,

cx = - a 3 ,

Cy = ~b%

(5.7.19)

Find the rotation vector in the latter case.


Solution

From (5.7.17), we find that


[Vu] =

ax

bx

<*2

^2

c2

a3

b3

c3

(5.7.20)

so that
2a i

[E] = i 2 + 6,
_3 + d

bx + a2

cx + a3

22

c2 + b3

3 + c2

2c3

(5.7.21)

and

0
[]=|

a2 _3

6i - a2

Ci -

a3

6j

c2 - b3

Ci

b3 - c2

(5.7.22)

Evidently, there occurs a pure deformation, that is = 0, if and only if


(5.7.18) hold. Also, there occurs a pure rotation, that is E = 0, if and only
if (5.7.19) hold.
From (5.7.17), we find that
curlu = (b3 - c2)ex + (cx - a3)e2 + (a2 - bx)e3

(5.7.23)

In the case of pure rotation, where (5.7.19) hold, this yields


= j curlu = (b3ex + cxe2 + a 2 e 3 )
as the rotation vector.

(5.7.24)

216

5 DEFORMATION

EXAMPLE 5 . 7 . 3 If u is the displacement vector at a point x and u + du


is the displacement vector at a neighboring point x + dx, then du is called
the relative displacement in the neighborhood of x. Show that du is
generally a superposition of a displacement due to stretch and a rigid-body
displacement.
Solution

From the definition of du, we note that


dUi = T^dxj = Uijdxj
OXj
= eu dXj + coy dXj

or
du = Erfx + Qdx

(5.7.25)

Evidently, du is made up of two parts: E dx, which is effected by E, and


dx which is effected by . Since E causes a stretch and causes a pure
rotation, it follows that du is a superposition of a displacement caused by
stretch and a rigid-body displacement.

For small deformation, show that

EXAMPLE 5 . 7 . 4

Q = I +

(5.7.26a)

(5.7.26b)

= I -

Solution Comparing (5.7.5) with (5.3.8b), we readily get (5.7.26a). By


use of (5.6.13b), (5.5.7), (5.6.4) and (5.7.8) we obtain
Q- 1 = F
which is (5.7.26b).

+ E = (I - Vu) + |(Vu + VuT) = I -

EXAMPLE 5 . 7 . 5 Show that for small deformation the polar decompositions of F reduce to the additive decompositions of Vu into symmetric
and skew parts.
Solution Using the relations (5.5.18a), (5.6.12a) and (5.7.26a), the right
polar decomposition of F given by (5.3.5a) becomes
I + vu = (I + )( + E) = I + E +
or
Vu = E +

(5.7.27)

which is the additive decomposition of Vu into symmetric and skew parts


E and .

5.8

COMPATIBILITY CONDITIONS

217

On the other hand, using the relations (5.5.18b), (5.6.12b) and (5.7.26b),
the left polar decomposition of F given by (5.3.5b) which is equivalent to
F" 1 = Q ^ V 1 becomes
I - Vu = (I - )( - E) = I - E -
or
Vu = E +

(5.7.28)

which is again the decomposition of Vu = Vu into symmetric and skew


parts.

EXAMPLE 5.7.6

Show that, for small deformation,


//a = a0 + Ea + Qa

Solution

(5.7.29)

For small deformation, we have by (5.5.18a) and (5.7.27),


F=I+E+

(5.7.30)

Substituting this into (5.4.1), we readily get (5.7.29).

5.8
COMPATIBILITY CONDITIONS
The strain-displacement relations (5.6.4)' enable us to find E for any given
u. But, for any given E, it may not always be possible to find u from
(5.6.4)', because, for a given eij9 (5.6.4)' is a system of six partial differential equations for only three unknown functions ut, and the system may not
possess a solution. For example, suppose that e^ are given as follows:
11 = 22 = 33 = 12 = 13 = 0 ,

?23 = *2 *3

(5.8.1)

Then (5.6.4)' yield the following partial differential equations:


du

du

du

dx3

oxx

dx2

dw,

du2

dux

du3

dx2

dx{

dx3

dxx

du2
dx3

du*
dx2

= 0

(5.8.2)

Solutions of the second and the third of these equations are u2 =f(X\, ^3),
u3 = g(xl9x2), where/ and g are arbitrary functions of their arguments.

218

5 DEFORMATION

Inserting these solutions into the last of the equations, we get

(XI

3> + -( )

Jr ' *

23=

~ ^*

Since the first or second term in this equation cannot have terms of the form
, the equation can never be satisfied. Thus, for the ei} given by (5.8.1),
equations (5.6.4)' do not yield a solution for w,.
We now proceed to establish a necessary and sufficient condition on E
that ensures the existence of u as a solution of (5.6.4)'.

JC2JC3

THEOREM 5-8-1

For a given E, if the strain-displacement equation


E = (Vu + Vu r )

(5.8.4)

has a solution for u, then E should satisfy the condition:


curl curl E = 0

(5.8.5)

Conversely, if E satisfies the condition (5.8.5) in a simply connected region,


then equation (5.8.4) possesses a solution for u. Furthermore, this solution
is unique within a rigid-body displacement.
Proof First, suppose that for a given E there exists u such that equation
(5.8.4) holds. Taking curl on both sides of this equation, we get
curl E = \ curl Vu + \ curl Vu r

(5.8.6)

But, for any vector a,


curl Va = 0
r

c u r l V a = V curl a

(5.8.7a)
(5.8.7b)

see identities (3.5.18) and (3.5.19). Hence (5.8.6) becomes


curlE = ^V(curlu)

(5.8.8)

Taking curl of this equation and using (5.8.7a), we get the condition (5.8.5).
Conversely, suppose that the condition (5.8.5) holds for a given E in a
simply connected region. This condition may be written as
curlE = 0

(5.8.9)

= curlE

(5.8.10)

where
Then, since E is symmetric, the identity (3.5.39) yields
t r = tr(curlE) = 0

(5.8.11)

(5 8 3

**

5.8

COMPATIBILITY CONDITIONS

219

Consequently, by virtue of the result proved in Example 3.6.4, there


exists a skew tensor W such that
= curlW

(5.8.12)

curl(E - W) = 0

(5.8.13)

Using (5.8.10) this becomes


from which it follows that there exists a vector u such that
E - W = Vu

(5.8.14)

see again Example 3.6.4. Equating the symmetric parts on both sides of
(5.8.14) and bearing in mind that E is symmetric and W is skew, we find
that u satisfies equation (5.8.4). Equation (5.8.4) thus possesses a solution.
Suppose u' is also a solution of (5.8.4) for the same E. Then the strain
tensor corresponding to u - u' is 0 so that u - u' is a rigid-body displacement (see Example 5.6.2). Thus, two solutions of (5.8.4), when they exist,
differ at most by a rigid-body displacement.

This completes the proof of the theorem.


The criterion (5.8.5) that ensures the possibility of obtaining u by solving
the equation (5.8.4) (in a simply connected region) is known as the
integrability or compatibility condition for E. A strain tensor E that
satisfies this condition is referred to as a possible strain tensor.
By virtue of the result proved in Example 3.5.8, the compatibility
condition (5.8.5) can be expressed in the following equivalent form:
(5.8.15)
V2E + VV(tr E) - V div E - (V div E) r = 0
In the suffix notation, the conditions (5.8.5) and (5.8.15) read respectively as follows:
(5.8.5)'
e

ij,kk

kkjj

~~ eik,kj

jk,ki

(5.8.15)'

When written in the unabridged form, these conditions yield the


following six explicit and distinct conditions:
^11,22 + ^22,11 = 2e 1 2 ,i2
^22,33 + ^33,22

2^23,23

33,11 + 11,33

2e31>31

11,23 ~~ 23,11

23,21 + 21,23 ~~ 22,31 ~~ 31,22

31,32 + 32,31 ~~ 33,12 ~~ 12,33

12,13 + 13,12

(5.8.16)

220

5 DEFORMATION

These six conditions were first obtained by Saint-Venant in 1860 and are
known as Saint-Venant compatibility conditions for the infinitesimal strain
components.
The conditions (5.8.16) can also be extracted from the following set of
conditions:
eijyki + **#,</ - *ikji - ejk = 0

(5.8.17)

As such, conditions (5.8.17) are also often taken as an alternative version


of the compatibility conditions (5.8.5)'. Note that the conditions (5.8.15)'
are a contracted version of the conditions (5.8.17).
Show that the following are not possible strain

EXAMPLE 5.8.1
components:

en = k(x\ + *f),
el2 = kxlx2x3,

e12 = k(x\ + xj)


el3 = e23 = e33 = 0

(5.8.18)

Here k and k' are constants.


Solution The given numbers etj are possible strain components only if
each of the six compatibility conditions (5.8.16) is satisfied. Substituting the
given eij into the first of the conditions (5.8.16), we get
2k = 2k'x3

(5.8.19)

Since k and k' are constants, the condition (5.8.19) cannot be satisfied for
x3 iA 0. For x3 = 0, (5.8.19) gives k = 0; but then all the given e^ vanish.
Hence, the given e^ are not possible strain components.

EXAMPLE 5 . 8 . 2 (i) Find the compatibility condition for the strain


tensor E if en, e22 and el2 are independent of x3 and el3, e23 and e33 are all 0.
(ii) Find the condition under which the following are possible strain
components:
^n

k{X\ x2)9

e22 = kx{x2,

e2 = k X\X2
el3 = e23 = e33 = 0

(5.8.20)

where k and k' are constants.


Solution (i) Since en, e22 and el2 are independent of x3 and el3 = e23 =
e33 = 0, we readily verify that all the conditions except the first one in
(5.8.16) are automatically satisfied. Hence the only compatibility condition
to be satisfied in this case is
^n.22 + ^22.11 = 2<?12>12

(5.8.21)

5.8

COMPATIBILITY CONDITIONS

221

(ii) For e^ given by (5.8.20), the only compatibility condition to be


satisfied is (5.8.21). From (5.8.20) we find that
ellt22

= -2k,

e22fil = 0,

e12>12 = k'

(5.8.22)

Hence (5.8.21) is satisfied if and only if


(5.8.23)

k = -k'

Thus, eu given by (5.8.20) are possible strain components if and only if


k = -k'.
M
Note: A deformation for which en, e22 and e33 are independent of x3 and
^13 = ^23 = 33 = 0 is called a plane deformation parallel to the xix2 plane.
Study of such deformations forms an important part of the elasticity
theory.
EXAMPLE 5 . 8 . 3 When e^ given by (5.8.20) are possible strain components, find the corresponding displacements, given that w3 = 0.
Solution In Example 5.8.2, it was shown that e/7 given by (5.8.20) can
yield displacements only when k = -k'. Substituting these eu with k = -k'
into the strain-displacement relations (5.6.4)', we obtain the following
partial differential equations:
"1,1 = k(xl - *!)>

"2,2 = kX\Xi

"1,2 + "2,1 = -2kxxx2


"2,3 + "3,2 = 0,

(5.8.24)
w3,i + "1,3 = 0,

w3>3 = 0

Since w3 = 0, the fourth and the fifth of equations (5.8.24) show that ux
and 2 a r e independent of x3. Integration of the first two of the equations
then yields
ux = \k(xl - 3x^1) +f(x2)
(5.8.25)
u2 = \kxxx\+

(5.8.26)

g(xx)

w h e r e / and g are arbitrary functions of x2 and xi9 respectively. Inserting


(5.8.25) and (5.8.26) into the third equation in (5.8.24), we get
g'(xl) = \kx12-f'{x2)

(5.8.27)

The lefthand side of this condition is a function of xx while the righthand


side is a function of x2. As such, this condition is possible only if
g'(Xl) = c,

\kx\

- f'{x2)

=c

(5.8.28)

222

DEFORMATION

where c is a constant. Consequently, we get


fix2) = \kx\

- cx2 + cx,

g(xx) = cxl + c2

(5.8.29)

where cl and c2 are arbitrary constants. Expressions (5.8.25) and (5.8.26)


now become
ux = ik(2x{ - 6xxxl + x\) - cx2 + cx

(5.8.30)

u2 = \kxxx\

(5.8.31)

+ cxx + c2

These are the displacement components associated with the given eu when
the compatibility conditions are obeyed.

EXAMPLE 5 . 8 . 4 . For each of the following strain systems obtain the


corresponding displacement components representing pure deformation,
given that there is no displacement at the origin:
1
(i) ew u11 = -N,
E '

v
~ 3 J = --N
eiL
22 = e
33
E9 '

V*/

() eu
e 2i

^22 e33 e\2

elil2 = e"
2, = en = 0 (5.8.32)

= 0,

(5.8.33)

3T*1

b
(iii) el2 -e22-

<?n = <?i3 = e2i

(iv) eu - e22 -

(/

Xi),

*33 = - (I - *3>>

(5.8.34)

= 0

vM
xu

e33 =

M
~Xl'

*i2 = e23 = e*\ = 0

(5.8.35)

Here, E, N, v, M, , , , Z?, /, /?, g, / are constants.


Solution It may be verified that the strain systems (5.8.32) to (5.8.35)
satisfy the compatibility conditions.
(i) Substituting for e^ from (5.8.32) in the strain-displacement relations
(5.6.4)', we obtain the following equations:
1
"1,1 = ^,
1,2 + 2,1 = 0,

v
w2)2 = w3,3 = j^N,
H2,3 + 3,2 = 0,

(5.8.36)

!l 3 f l + Wlf3 = 0 ( 5 . 8 . 3 7 )

5.8

COMPATIBILITY CONDITIONS

223

The righthand sides of equations (5.8.36) are constants; hence the solutions
of these equations are linear functions of xx, x2, x3 and are given by
ux = Nxx + axx2 + bxx3 + cx
E
v
u2 = - Nx2 + a2x3 + b2Xi + c2
E
v
u3 = - Nx3 + a3xl + b3x2 + c3

(5.8.38)

where a,, bi9 ct are constants.


Using (5.8.38) in equations (5.8.37), we obtain the following relations
connecting at and bt :
ax = -bl9
u

ij

a2 = -b39

(5.8.39)

a3 = -bx

Since there are no rigid-body rotations, we have = skw Vu = 0 so that


= u
j,i- The solutions (5.8.38) therefore yield (see Example 5.7.2):
ax = bl9

a2 = b39

(5.8.40)

a3 = bx

The conditions (5.8.39) and (5.8.40) imply that at and bt are 0. Also, since
there is no displacement at the origin, (5.8.38) yield ct = 0. Thus, (5.8.38)
become
1
ux = gNxl9

v
u2 = jjNxi*

v
u3 = - ^ ^ * 3

(5.8.41)

These are the displacement components associated with the strain components given by (5.8.32).
(ii) Substituting for eu from (5.8.33) in the strain-displacement relations
(5.6.4)', we obtain the following equations
w

+ U

i,i

0,

u22 = 0,
2M
X

~1^9 ^

u33 = 0,

"*>> + >* =

The form of these equations suggests that uX9u29u3


U

uX2 + u2y ! = 0
2M

^*b *

(5.8.42)

are of the form

\ = #1*2*3 + ^1*2 + CiX3 + d\

u2 = a2x3xx

+ b2x3 + c2xx + d2

u3 = a3xxx2

+ b3xx + c3x2 + d3

(5.8.43)

where al9 bi9 c, and dt are constants. Substituting (5.8.43) into (5.8.42) we

224

5 DEFORMATION

find that
M(aL + b*)
a2 = -a, = "y~ 3 : ' ,

'

a3 =
*3 "

bl = -c2i

b3 = -cl

b2 = -c39

M , 2L
^
^
(a - bz)
*

(5.8.44)
(5.8.45)

Since there is no rigid-body rotation, we have wl>7 - ujyi = 0. In view of


this, we find from (5.8.43) and (5.8.45) that , and c, are all 0. Further, since
there is no displacement at O, the constants rff are all 0. Thus, the solutions
(5.8.43) reduce, with the use of (5.8.44), to
U

M(a2 + b2)
l

2 =

3T3

b
M(a2 + b2)
.._ 3A3
~

*2*3

*3*1

"a = - _ 3 A5 3 ( g

Tb

(5.8.46)
b2)xix2

These are the displacement components associated with the strains given by
(5.8.33).
(iii) Substituting for eu from (5.8.34) into the strain-displacement
relations (5.6.4)', we obtain the following equations:
. = "2,2 = - ^ ( / - * 3 ) >
",2+ "2, = 0,

w2,3 +3,2 = 0,

3,3 = ( ' - * 3 )

(5.8.47)

uhi + ult3 = 0

(5.8.48)

Integrating the last of the equations in (5.8.47) gives

U =E

' E\ I*3)*3

+ MXl Xl)

'

(5 8,49)

where f3(x1, x2) is a function of , x2 only. Using (5.8.49) in the last two of
equations (5.8.48) and integrating the resulting equations we obtain
"i = -^3/3,1

+/i(*i,*2)

(5.8.50)
"2 = --^3/3,2 +(*,*2)
where/i a n d / 2 are functions of xl9 x2 only. Substituting (5.8.50) into the
first two equations in (5.8.47) gives
/ l , l =/2,2 = - ~ .

/3.II =,22 = '

(5.8.51)

5.8

COMPATIBILITY CONDITIONS

225

Similarly, from (5.8.50) and first of (5.8.48) we obtain


73.12 = 0,

/,2 = -.

(5.8.52)

Further, (5.8.49) and the last two equations in (5.8.48) yield


/ 3 ,23 = 0,

/ 3 f i3 = 0

(5.8.53)

The conditions (5.8.51) to (5.8.53) imply that./) are of the form


/l(*l>*2) =

vpgl
JfXl

+ a \X 2 + C l

vpgl
= -~T*2 - ** + c2

,)

= ~ | f <*? + *2> + r3^i + 63*2 + c3

(5.8.54)

where UTJ , a3, 63 and c, are constants. Using these in (5.8.49) and (5.8.50) we
obtain
vpg

"l = g - C - *3>*1 - <*3*3 + *1*2 +

u2 = ~ (l - x3)x2 - a3x3 - axxx + c2

(5.8.55)

"3 = 1F[(*3 ~ 2/)*3 + v(*' + x*)] + iXl + bXl + 3


Since there is no rigid-rotation, we have uu - w, , = 0, and expressions
(5.8.55) yield a3 = ax = 3 = 0. Also, since there is no displacement at the
origin, c, = 0. Consequently, (5.8.55) become

M2

= -^(/-x3)x2

3 = " | | K*3 - 2/)X3 + V(X? +

(5.8.56)
Xl)\

These are the displacement components associated with the strains given by
(5.8.34).

226

5 DEFORMATION

(iv) Substituting for eu from (5.8.35) into the strain-displacement


relations (5.6.4)' we obtain the following equations:
vM
"i,i = "2,2 = ~EjXl'
1,2 + "2,1 = 0,

U3 3 =

'

M
~mXl

l/2,3 + "3,2 = 0

(5.8.57)

t/3,1 + "1,3 = 0 (5.8.58)

From the third equation in (5.8.57), we get


M
"3 = ~YJX^

+/3(*i>*2)

(5.8.59)

where/ 3 is a function of xx, x2. The last two equations in (5.8.58) then give
"i = ^TjXl

- X3/3.1 + /i(*i>*2>

(5.8.60)

"2 = - * 3 / i , 2 + (* > *2>

(5.8.61)

where/j a n d / 2 are functions of ^ , ^ Substituting for Wj and u2 from


(5.8.60) and (5.8.61) in the first two equations in (5.8.57), we obtain the
following relations:
/3.11 = 0,

/3f22 = 0

vM ,
^2 */ i + Site).

(5.8.62)
h

vM
7
27

&(*)

(5.8.63)

where gj and g2 a r e functions of x2 and ATJ respectively.


Using (5.8.60), (5.8.61) and (5.8.63) in the first equation of (5.8.58), we
obtain the following relations:
.2 = 0
gl

(5.8.64)

vM 2
~2Xl

+ xXl

+ Cu

gl =

+ Cl

(5

*865)

where ax, cx, c2 are constants.


From (5.8.62) and (5.8.64) we find that
/ 3 = a3xl + Z?3x2 + c3
where a3, Z?3, c3 are constants.

(5.8.66)

5.9 PRINCIPAL STRAINS

227

Substituting for ft from (5.8.63), (5.8.65) and (5.8.66) in (5.8.59) to


(5.8.61), we obtain
M

"l = 2/(*3

j
VX

VX

2) + OiX2 - tf3X3 + CX

Mv
u2 = -j^X\*2 - <*\X\ - M s + c2

(5.8.67)

W3 = " ^ * 3 * 1 + *3*1 +

3*2 + C3

Since there is no rigid-body rotation, we have utJ - ujti = 0, and (5.8.67)


yield ax = a3 = b3 = 0. Also, since there is no displacement at O, we get
q = 0. Thus, (5.8.67) reduce to
^i = W + v^i - vx2),

u2 = xxx2,

u3 = ~ x3xl

(5.8.68)

These are the displacement components associated with the strains given by
(5.8.35).

5.9
PRINCIPAL STRAINS
It has been seen that, under a deformation, a material arc generally changes
in length and orientation. From a particle (9 there emerge infinitely many
material arcs, and each of these arcs generally changes in length and
orientation. We now consider the arcs that change only in length, under
small deformation.
As in Section 5.3, let a0 and a be unit vectors representing the initial and
final orientations of a material arc rf(B initiating from (P. In the case of
small deformation, the stretch of d& is fully effected by the strain tensor E
and the rotation is fully effected by the rotation tensor . Hence, dC has
no change in orientation if and only if = 0 and a = a0. A necessary and
sufficient condition for this may be obtained by setting = 0 and a = a0 in
(5.7.29). The condition so obtained is
Ea = ( - l)a = ea

(5.9.1)

where e = - 1 is the normal strain experienced by dQ. Evidently, this


condition holds if and only if a0 is an eigenvector of E and e is the corresponding eigenvalue. Since E is a symmetric tensor, it has exactly three (not

228

5 DEFORMATION

necessarily distinct) eigenvalues and at least three mutually orthogonal


eigenvectors (by virtue of Theorems 2.13.2 and 2.13.4). Hence there exist at
least three mutually orthogonal unit vectors and three normal strains for
which condition (5.9.1) holds. This means that at least three mutually
orthogonal material arcs at (P change only in length. The normal strain of
such an arc is called a principal strain, and its direction is called a. principal
direction of strain at (P. The material surface elements at (P that are
orthogonal to these principal directions are called the principal planes of
strain at (P.
Thus, at a particle (P, there exist (i) at least three mutually orthogonal
principal directions of strain, (ii) at least three mutually orthogonal principal planes of strain, and (iii) exactly three principal strains (which are not
necessarily distinct).
Since the principal strains are the eigenvalues of E, these are the roots of
the characteristic equation
- 3 + IEA2 - IIEA + IIIE = 0

(5.9.2)

where 7E, IIE and IIIE are the fundamental invariants of the tensor E,
which by virtue of (2.11.1), (2.11.4) and (2.11.11) are given by
h = ekk,
IIE = j(eii^jj - eueji)> /// E = det(^) (5.9.3)
These invariants are called the small strain invariants.
Once the three principal strains, say el, e2, e3, are found by solving the
cubic (5.9.2), the corresponding principal directions of strain can be found
by solving the vector equation (5.9.1) for a0 = a in the cases: = el9
A = e2 and A = e3. With respect to a set of axes chosen along the principal
directions so determinedsuch axes are called principal axes of strainthe
matrix of E (or briefly the strain matrix) is given as follows (by virtue of
Theorem 2.13.5):
- ?, o

M =

e2

(5.9.4)

0 0 e,
In view of the geometrical meanings of e^ noted in Section 5.6, it is
verified from (5.9.4) that ex ,e2,e3 represent the normal strains of material
arcs initially lying along the principal directions of strain. Since the shear
strains are all 0 in the representation (5.9.4), it is also verified that these arcs
remain in the same directions after deformation also.
Since ex, e2, e3 are the roots of the cubic (5.9.2), the strain invariants have
the following expressions in terms of the principal strains:
h = ex + e2 + ?3,

IIE = exe2 + e2e3 + e3el9

IIIE = exe2e3

(5.9.5)

5.9 PRINCIPAL STRAINS

229

EXAMPLE 5.9-1 For the small deformation considered in Example


5.7.1, find the strain invariants, principal strains and the principal directions of strain at the point (1,1,0).
Solution For the deformation considered here, the strain matrix is given
by (5.7.13). At the point (1, 1,0), this matrix becomes

[E] = | a

0 I

4a

(5.9.6)

whose characteristic equation is found as


( - 2)( - 4) = 0

(5.9.7)

Hence the principal strains, which are the roots of equation (5.9.7), are
ex = 0 ,

e2 = 2a,

e3 = 4a

(5.9.8)

The strain invariants now follow from (5.9.5) as


/ E = 6a,

IIE = 8a 2 ,

IIIE = 0

(5.9.9)

Since ex = 0, the eigenvector a corresponding to ex is given by Ea = ex a = 0,


which on using (5.9.6) yields
ax + a2 = 0,

ax + a2 = 0,

4a3 = 0

so that
ax = -a2 = ^ j ,

a3 = 0

(5.9.10)

Here we have used the fact that a is a unit vector.


Thus, an eigenvector associated with the principal strain ex = 0 is
a, = ^ ( . ,

(5.9.11)

e2)

Similarly, the eigenvectors corresponding to e2 = 2a and e3 = 4a are


obtained as
a2 = ^ | ( e i + e 2 ),

(5.9.12)

The vectors a ! , a 2 , a 3 just determined specify the required principal


directions.

EXAMPLE 5.9-2 For small deformation, show that the normal strain of
a material arc assumes an extremum value when the element lays along a
principal direction of strain.

230

5 DEFORMATION

Solution In terms of its initial orientation, the normal strain of a


material arc is given by
e Uijafa} = e^a*

(5.9.13)

see (5.6.6). We have to find af for which e is an extremum. Since af are


components of a unit vector, we have akak = 1. Thus, e assumes extreme
values when af obey the equations
*{e

- a(a0ka0k - 1)) = 0

(5.9.14)

where a is a Lagrangian multiplier.


Substituting for e from (5.9.13), the conditions (5.9.14) become
(eu - ^

=0

(5.9.15)

These conditions are satisfied if and only if a0 is along a principal direction of strain. Thus, e assumes an extreme value when the material arc lies
along a principal direction of strain.

Note: This result obtained is analogous to (and a consequence of) the


result obtained in Example 5.4.5.

5.10
STRAIN DEVIATOR
In Section 2.12, we have seen that every second-order tensor can be decomposed uniquely into a spherical part and a deviator part. Employing this
decomposition to the tensor E, we get
E = j(trE)I + Eid)

(5.10.1)

We note that the trace of the spherical part of E, which is equal to the
trace of E, represents dilatation. As such, the spherical part of E accounts
for the volume change of a material element, under deformation. (In fact,
the adjective spherical part is motivated by this interpretation.) Consequently, the deviator part E(c accounts for the change in shape of the
element. The tensor E (d) is called the (infinitesimal) strain deviator tensor.
The components eff* of E( are called the (infinitesimal) strain deviator
components and the matrix [e$p] is called the strain deviator matrix.
In Example 2.13.7, it has been shown that a tensor and its deviator part
have the same eigenvectors. It therefore follows that the principal directions
of E(c coincide with those of E. With respect to a set of principal axes, the

5.10 STRAIN DEVIATOR

231

strain deviator matrix is purely diagonal, the diagonal elements being the
eigenvalues of E(c. These eigenvalues, e[d), e$*\ e$*\ are called the principal deviator strains. By virtue of expressions (2.13.30), we note that ejd)
are related to the principal strains et through the following relations:
e = \(2ei -e2-

e3),

eid) = j(2e 2 -e3-

ex)9

oW

(5.10.2)

= i(2e3

e2)
Since the deviator part of a tensor has zero trace, we have 7E<d) = 0. By
virtue of expressions (2.13.31) and (2.13.33), the other two fundamental
invariants of (/) are given as follows:
//E(d)

e?>e> + e>e> + ef>e?>

= -\{(ex - e2)2 + (e2 - e3)2 + (e3


IH^id) el

eu2)

(5.10.3)

_ JL
if(2ei - e2 - e3)(2e2 - e3 - el)(2e3 - ex - e2)

These are called the strain-deviator invariants.


EXAMPLE 5.10.1 For the small deformation considered in Example
5.7.1, find (i) the strain deviator matrix, (ii) principal deviator strains, and
(iii) strain deviator invariants, at the point (1, 1,0).
Solution For the deformation considered here, the strain matrix at the
point (1, 1, 0) is given by (5.9.6). By use of (5.10.1), we find that the strain
deviator matrix is given by
[E(d)] = [E] - j(trE)[I]

(5.10.4)

which, on using (5.9.6), yields


0
(d

[E >] =

0
0

(5.10.5)

la

Recalling that the principal strains are as given by (5.9.8) and using (5.10.2),
we obtain the corresponding principal deviator strains as
}
* =
= -2a,
er'
ef ) = 0,
=fa
(5.10.6)
ef
Using (5.10.6) in (5.10.3), we get the strain deviator invariants as

//E(d) = -f a 2 ,

mEid) = 0

(5.10.7)

232

5 DEFORMATION

EXAMPLE 5.10.2

Show that
A{//w)
deu

Solution

_efiO

(5.10.8)

By virtue of expression (2.12.11), we have


//E< = Jijelk - eue]

(5.10.9)

Hence
T [//E<*>] = [\Suekk
deu

- eu] =

-efp

which is (5.10.8).

5.11
EXERCISES
1. For the deformations defined by the following sets of equations, find the tensors
F, F~\ Q, U and V.
(i) xl = oxu
()

Xx = Xx,

x2 = x2,
X 2 = X2 -

x3 = xl
X 3 = X + OX2

(iii) = ax* + x2i

x2 = -ax* + x2l

(iv) xl=xl

x2 = x2 + xx,

- x2x3l

x3 =
*3 = *?

Here a, and y are positive constants. Which of these deformations are isochoric?
Also, find the principal stretches in each case.
2. Give the geometrical descriptions of the deformations defined by the sets of
equations (i), (ii) and (iv) of Exercise 1.
3. Deduce the relation (5.2.33) from the relation (5.2.32) and vice versa.
4. Prove the following identities:
W

jmnXj,k

~~ ~j kpqXp,mXq,n

""

tkpqXj.k

~ ~7

vHM

jmnXj,k

JkpqXp,mXq,n

kpqXj\k

JEjmnXp,mXq,n

v' /

tjmnXptmXqtn

5.11

EXERCISES

233

5. Show that the deformation given by the equations


xx = *?,

x2 = xl + ax?,

x3 = xl + ctxl

where (5* 1) is a constant, is just a triaxial stretch. Deduce that the particles that
lie on the circle
*J = 0,

(x2)2 + (JC30)2 = (1 - a 2 )" 1

before deformation move over to the ellipse


xx = 0,

(1 +

OL2)XI

- 4ax2x3 + 0 +

2)2

= 1 - a2

6. Under the deformation defined by the set of equations iv of Exercise 1 find the
surface into which the cylinder (x?)2 + (xl)2 = a2 deforms.
7. Show that, under a homogeneous deformation,
(i) Parallel plane elements transform to parallel plane elements;
(ii) Parallel straight line elements transform to parallel straight line elements;
(iii) spherical surface elements transform to ellipsoidal surface elements.
8. Consider a homogeneous deformation defined by x, = auXj + ci9 where c, and
Ojj are constants. Show that, in this deformation, the particle that lies on a spherical
surface of radius a in the current configuration initially lay on the surface of an
ellipsoid. Show further that this ellipsoid is a sphere of radius b if and only if
9. For the deformations defined by the following sets of equations, find the tensors
C, B and G:
(i) xl =

JCJ,

x2 = xl + ox?,

x3 = Jtj

(ii) xx = *,

x2 = x2 - ax 3 ,

x3 = xl + ax2

(iii) xx = V(2ax? + ,

x2 = yx,

x3 = 3

(iv) JCJ = xl + x2 tana,

x2 = x2i

x3 = xl

(v) xx = xl cosiax?) + JC2 sin(ax3),

x2 = -A:? sin(ax?) + xl cos(ax3),

x3 = (1 + aflxj
Here a, , y and <5 are nonzero constants.
10. For the deformations defined by the following sets of equations, find the
tensors B _ 1 and A:
(i) x\ = 2ox? + /?,

x2 = yxl,

(ii) xl = x1 cos(ax3) - x2 sin(ax3),


xl = (1 + )3

x3 = xQ3
xl = xx sin(ax3) + x2 cos(ax3),

234

5 DEFORMATION

11. For the deformation defined by the equations


Xl = Xu

X2=X*

+ ,

*3 = *3

where a is a constant, show that


/ B = 7/B = 7//B-, = 3 + a 2 ,

///B

= 1

1 2 . Prove the following identities:


/c = 3 + 2/ G
/ / c = 3 + 4/ G + 4// G
IIIC = 1 + 2/ G + 4// G + 8/// G

13. Show that a deformation is a rigid-body transformation if and only if


Ic = IIC = 3 a n d / / / c = 1.
14. For the deformation defined by the equations
X^

X j ~r OU2

^2

"^2 * ^ ^ 1 >

-^3

=:

^3

where a is a constant, find the normal strain of the material arc initially lying along
the vector e2 + e3.
15. Show that under the deformation considered in the Exercise 11, material arcs
initially lying along the 3 axis retain their lengths.
16. For the deformation defined by the equations
Xl

ATj ,

X2

X2

-^3

==

-^3

'

0^2

where a is a constant, find the material arcs initially lying parallel to the 23 plane,
for which the normal strain is 0.
17. Derive the expressions (5.4.33) and (5.4.34).
18. Show that the final angle between two material arcs whose initial directions
are specified by the unit vectors a0 and inclined at an angle 0 is given by
cos* =

2guafJ + se0
(1 + 2gua?a)t/2V + 2gufay/2

Obtain the counterpart of this expression in the spatial description.


19. Show that the angle between two material arcs initially represented by the unit
vectors a0 and remains unchanged if and only if
a0 C = (a0 )((a Ca)( C))1/2
Express this condition in terms of the unit vectors a and along which the arcs lay
in the deformed configuration.

5.11

EXERCISES

235

20. Show that the deformation defined by the equations


xx = 2x, + JCS,

x2 = ? + 2x%, x3=x

does not change the angle between the material arcs lying along ex and e2 in the
initial configuration.
21. For the deformation defined by the equations
X\

Xi

X2

' X$ y

X2 ~ %2

*^3

' -^1

-^3

-^3

-^1 ' ^ 2

find (i) the stretch in the direction a = (1/V2)(e! + e2), and (ii) the angle in the
deformed configuration between the material arcs that were initially in the directions
of a0 and = e 2 .
22. For the deformation defined by the equations
xx = 3x? + 4,

x2 = x2 + xl

x3 = 2x1 - x2

find (i) the normal strain of a material arc initially having direction ratios 1:1:1, (ii)
the initial direction ratios of the material arc that finally ends up in the 3 direction,
and (iii) the change in the angle between the material arcs initially having direction
ratios 1:0:0 and 1:1:1.
23. For the deformation considered in Exercise 22, find (i) the direction of the
normal of the material surface element in the final configuration given that normal
had the direction ratios 1:1:1 in the initial configuration, (ii) the ratio of the area
of this surface element in the final configuration to its area in the initial configuration.
24. For the deformation defined by
x\ = 4JC? COS 2 (4/2),

X\ = 4? sin 2 (4/2),

x3 = *?,

JC? > 0

show that the lengths of sides, angles and volume of the element dxx dx2 dx3 located
initially at the point (1,1,0) remain unchanged.
25. Show that a material surface element initially orthogonal to a unit vector n
retains its area, under deformation, if and only if
n

C n

=J2

2 6 . Prove that
l

iiJ

2 dJ
Jdc,

1 3J
JdgiJ

2 7 . Show that
dV

..,

236

5 DEFORMATION

28. For the deformation defined by


X

*(*1 )>

2\x\

X2

show that one of the principal stretches is 1. Show further that the deformation is
isochoric if and only if the product of the other two principal stretches is 1.
29. If are the principal stretches, show that \(2 - 1) are the principal values of
G and y(l - i/f"2) are the principal values of A.
30. Show that the stretch of a material arc assumes an extreme value when it finally
lies along a principal direction of B.
31. For the deformations defined by the following sets of equations find the displacement components in material and spatial forms:
(i) xx = xxt2 + 2x2t + xx
x2 = 2xt2 + xlt + 4

(ii) xx = xW + *(*' - 1),


*2 = * + *(e' - O

X3 ~2X3 t ~r X3

X3

X3

32. For each of the following sets of displacement components, find the Lagrangian
and Eulerian strain tensors:
(i) ux = oa3f
2

(ii) ux = (x?) +

u2 = -a*?,
(JC20)2,

u3 = a(x2 - x?)

u2 = 3xx +

(iii) ux = (xx)2 + 2(x2)2x3 + x2x3,

4(JC20)2,

(here a is a constant)
u3 = 2(xx)3 + 4x3

u2 = xxx2 + xxx3 + 3(xx)2x3

u3 = 0
If a < 1, show that case (i) corresponds to a rigid-body transformation.
33. For the deformation defined by the displacement vector
+ * 2 * 3 2 + 2(

U = XxX^l

+ *2>e 3

find (i) the normal strain of the material arc initially having direction ratios 1 : - 1 : 1 ,
(ii) the normal strain of the material arc finally having direction ratios 1 : - 1 : 1 , (iii)
the final angle between the material arcs which initially lay along e2 and e3.
34. Prove the following:
(i) F + F"1 = 21 + (Vu)(Vu)
(ii) B = I + Vu + Vur + (Vu)(Vur)
35. Prove the following:
0)

(ii)

[CUrl 11], = - ZprXp.iXr.q

= J

ZpQrXi;pXr,

epqrxnq = Jx?.p[cunu], = /x -[curlu],

5.11

EXERCISES

237

36. For small deformation, show that


(i)

i(l-trE)

(ii)

Ic 3 + 2 tr E,

IIC 3 + 4 tr E,

(iii)

IIIC 1 + 2 tr E

E = i(F + ) - I = I - ifF" + ( F V )

37. For the small deformation defined by the equations


* i = * ? + <**3>

* 2 = *2

x 3 = JC? - ajc?

where a is a small nonzero constant, find the dilatation. Deduce the condition under
which the deformation is isochoric.
38. Find the constants a and such that the small deformation defined by
2 = x\

Wj = aJCj + 3 x 2 ,

- xi

>

3 = 3*3

is isochoric.
39. For small deformations defined by the following sets of displacements, find
the strain tensor, rotation tensor and rotation vector.
(i)
()
(iii)

M,

= -ax2x3,

W, = OJC?JC2,
u

(iv) u,
(v) u,

VJ

u2 = axUl

w2 = a x2 + x3)2,
2

- x3) ,

= 3(A:A:2 + 2),
=

~3XlX3>

u3 = 2a(jc? + 4)JC?

u2 = oa2 *?,
2

u3 = 0
w3 = -

u2 = a

2U=

~X2X3

w3 =

=:

3 a ( 2 j c j + JC2JC3)

(*3 + ^2)

Here a is a small constant, k is a constant and r2 = jt,*, ^ 0.


40. For the deformation corresponding to the case iii of Exercise 39, find (i) the
normal strain at the point (0, 2,-1) along the direction 8 : -1:4; and (ii) the change
in the right angle between the material arcs at (0, 2,-1) lying along the directions
8:-l:4and4:4:-7.
41. For the deformation corresponding to the case v of Exercise 39, find the
deformed shape of a spherical surface r = constant.
42. The linear strain tensor at a point is given by
[eij] =

"1

-3

V2

-3

-V2

V2 -V2

Find the normal strain in the direction 1 : - 1 : V2 and the shear strain between the
directions 1 : 1 :V2 and -1:1:V2.

238

5 DEFORMATION

4 3 . Find which of the following values of eti are possible linear strains. Compute
the corresponding displacements in the appropriate cases.
(i)
(ii)

\i = *i*2

other eu = 0

eu = 2*,,

el2 = Xi + 2x:

= 2x3,

= 2xl9

ei2 = e23 = 0
(Hi)

en = a(x\ + xl),
e

\3

(iv)

(v)

ii

en = ax3(x\ + x\),

e21 = oa\x3i

*13 = 23 = 33 = 0

(a = constant)

[<?]

"*1 + *2

*2

*,

X 2 + *3

*3

*2

*3

ATj

(vi)

(a = constant)

33

en = 2ax 1 x 2 ,

e22 = oal,

[eu\

x2

*1

el2 =

2oalx2x3,

*3

ATj : 3

*22
0

4 4 . Obtain the conditions under which the following values of ei} are possible
linear strains:
e22 = ax 2 ,

(i)

en = oxf,

(ii)

en = OLX3{X\ + xf),

el2 = X\X2,

e22 = ax 2 x 3 ,

el2 =

e13 = e23 = e33 = 0


xxx2x3,

en = e23 = e33 = 0
(Here a and /? are constants.)

45. Find the nature of the function/ such that


eu = o/(x 2 ,x 3 ),

e22 = e33 = /(x 2 ,x 3 ),

el2 = e23 = e33 = 0

where a and are constants, is a possible system of strain.


4 6 . If e13 and e23 are the only nonzero strain components and el3 and e23 are
independent of x3, show that the compatibility conditions may be reduced to the
following single condition:
13.2 -

e:23,1

constant

4 7 . If en = e22 = e33 = el2 = 0, el3 = >2 and e23 = , i, where is a function of


! and x2, show that must satisfy the equation
V = constant

5.11

EXERCISES

239

48. Derive the compatibility conditions (5.8.17) by differentiating the straindisplacement relations (5.6.4)'. Verify that the six conditions given in (5.8.16) are the
only distinct conditions represented by (5.8.17).
49, For the small strain tensor considered in Exercise 42, find (i) the deviator strain
components, (ii) the strain invariants, (iii) the principal strains and (iii) the direction
of the maximum normal strain.
50. For a certain small deformation, the displacement gradient at a point is given
by
4

[Vu] =

1 4

-1 -4
0

0
6

Find (i) the strain components, (ii) the deviator strain components, (iii) the strain
invariants, (iv) the principal strains and (v) the principal directions of strain.

CHAPTER 6

MOTION

6-1
INTRODUCTION
In the preceding chapter we studied some geometrical aspects of a transformation of a material body from the initial configuration to a current
configuration with the time held fixed. Such a transformation was termed
a deformation. We introduced and analyzed tensor fields that serve to
measure the changes in length and relative orientation suffered by material
arcs during a deformation. In the present chapter, we introduce and analyze
tensors that serve to measure the time rates (or briefly, rates) of these
changes during a motion, which may be viewed as a one-parameter
sequence of deformations with time as the (continuous) parameter. The
concepts of vorticity and circulation are also introduced and their
kinematic aspects analyzed. Some important transport formulas are also
proven.
241

242

6 MOTION

6.2
MATERIAL AND LOCAL TIME DERIVATIVES
In the study of motion of a continuum, we deal with the time rates of
changes of entities that vary from one continuum particle to the other. The
displacement vector, introduced in Section 5.5, is one such entity; the
velocity and the acceleration, to be introduced shortly, are also among such
entities. These entities may be expressed as functions described in the
material form or the spatial form, and the meaning of the time rate of their
change depends on the nature of the description. This basic aspect is
explained in the following paragraphs.
6.2.1

MATERIAL TIME DERIVATIVE

First, consider a real-valued function/ = f(x, t) that represents a scalar or


a component of a vector or tensor. As pointed out in Section 4.5, the point
x determines a continuum particle (P uniquely, namely the one located at x
at t = 0, and this particle is referred to as the particle x. With this notation,
we interpret/(x, /) as the value of/ experienced at time / by the particle x.
Also, the partial derivative of/ with respect to /, with x held fixed, is interpreted as the time rate of change of/ at the particle x. This derivative is
called the particle or material time derivative off, denoted Df/Dt. Thus,
Dt

dt

)\xo

(6.2.1)

where the subscript x accompanying the vertical bar indicates that x is held
constant in the differentiation of/ Note that, like/, Df/Dt is a function of
x and / by definition. In other words, Df/Dt given by (6.2.1) is a function
in the material form.
6.2.2

LOCAL TIME DERIVATIVE

Next, consider a real-valued function = (, /) that represents a scalar or


a component of a vector or tensor. Since x is a point in the current configuration of a continuum, we interpret (, t) (which represents the value
of at the point x at time /) as the value of experienced by the particle
currently located at x. Also, the partial derivative of with respect to t, with
x held fixed, is interpreted as the time rate of change of at the particle
currently located at x. This derivative is called the local time derivative of ,
denoted by the usual partial derivative symbol /dt. Thus,
3= /30(x, t)\
dt
\ dt )

(6.2.2)
X

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

243

Note that, like , /dt is a function of x and / by definition. That is, /dt
given by (6.2.2) is a function in the spatial form.
The distinction between the material time derivative and the local time
derivative should be emphasized. While both are partial derivatives with
respect to t, the former is defined for a function of x and / whereas the
latter is defined for a function of x and /. Physically, the local time
derivative of a function represents the rate at which the function changes
with time as seen by an observer currently (momentarily) stationed at a
point, whereas the material time derivative represents the rate at which the
function changes with time as seen by an observer stationed at a particle and
moving with it. The material time derivative is therefore also called the
mobile time derivative or the derivative following a particle. For brevity, the
material time derivative will be referred to as the material derivative or
material rate, and the local time derivative as the local derivative or local
rate.
6.2.3

VELOCITY AND ACCELERATION

Since x is a function of x and / in the material description of motion, the


material derivative of x, namely Dx/Dt, can be defined; we denote this
derivative by v. Thus,
_ Dx _ idx

(6.2.3)

" ~Dt " V7

By virtue of the meaning of the material derivative, it is evident that v represents the time rate of change of position, at time t, of the particle x. This
is called the velocity of the particle x at time t. If vt are the components of
v, then (6.2.3) reads as follows in the component form:
DXi
Vi=

fdxi

(6.2.3)'

Di \-3

Consequently, vt are called the velocity components of the particle x at


time t.
In Section 5.5, the displacement vector u of the particle x has been
defined by
u = x - x
(6.2.4)
It has also been noted that u may be regarded as a function of x and t, or
of x and t. Treating u as a function of x and t, we get from (6.2.3) and
(6.2.4),
Du
'du

v = ^-(x + u)
v(6.2.5)
Dt
'
,dt
dt

244

6 MOTION

Thus, the velocity of a particle at time / is precisely the rate of change of


displacement of that particle at time t. Expression (6.2.5) reads as follows
in the component form:
Du,
(6.2.5)'
Vi =
~Df
It may be pointed out that, in solid mechanics, the deformation and
motion are generally described in terms of the displacement vector. In fluid
mechanics, the motion is generally described in terms of the velocity vector.
When a motion is described in terms of velocity, it is commonly referred to
as diflow.
Since v is a function of x and t by definition, namely, (6.2.3), the
material derivative of v, namely, Dy/Dt, can be defined. This derivative is
called the acceleration of the particle x at time t. We often write v for
Dv/Dt. Thus, the acceleration of a particle at time t is the rate of change of
velocity of that particle at time t. The components of the acceleration are
denoted by Dvt/Dt or t>f.
It is to be emphasized that the velocity and acceleration are defined with
reference to a particle and are basically functions of x and t. In the spatial
description of motion, x is a function of x and /. Hence, like the displacement, velocity and acceleration can also be expressed as functions of x and
/. When v is expressed as a function of x and /, v(x, t) is referred to as the
instantaneous velocity at the point x. This actually means that v(x, t) is the
velocity at time t of the particle currently located at the point x. Similar
terminology is used in respect of acceleration also.
Next, we deduce a formula that will enable us to compute the instantaneous acceleration from the instantaneous velocity.
6.2.4

MATERIAL DERIVATIVE IN SPATIAL FORM

Consider again the function = 0(x, t) for which the local derivative was
defined by (6.2.2). This function can be expressed as a function of xf and
/ as explicitly indicated in the following:

= </>(*,, t) = 0, 0

(6.2.6)

Consequently, the material derivative of can also be defined. By the chain


rule of partial differentiation, we obtain from (6.2.6),
ro

\dt

OX;

(6.2.7)

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

245

By virtue of expressions (6.2.1), (6.2.2) and (6.2.3)', we find that


/3\ I _
\at)\xo"'Dt9

/3\ I _ 3
\dt)\xt~'t9

idxj_
\dtt

Vi

Hence, denoting (d0/3jcf)|, as just /dXi = ,, expression (6.2.7) can be


rewritten as follows
Do

do

D7 = 7 +

do

^ = 7 +

( v

(6 2 8)

When v is known as a function of x and t9 expression (6.2.8) enables us


to compute /Dt as a function of x and t. As such, (6.2.8) serves as a
formula for the material derivative in the spatial form. Note that the first
term on the righthand side of this formula, namely, /dt, represents the
local rate of change of and the second term, namely, vrfti = (v V)0, is
the contribution due to the motion. The second term is referred to as the
convective rate of change of .
It can be easily verified that the material derivative operator
D

^ = 7 + y ' ( ) " = 7

+ v

-v

(6 2 9)

which operates on functions represented in spatial form, satisfies all the


rules of partial differentiation.
The concept of the material derivative and the formula (6.2.8) are
attributed to Euler (1770) and Lagrange (1783).
6.2.5

ACCELERATION IN SPATIAL FORM

Taking = vt in (6.2.8), we get the following expression for the


acceleration:
DV;

dV;

- = ^r1 + vkvi, k
Dt
dt
'

(6.2.10a)

or
D\
by
_ = _ + (y. v)v
(6.2.10b)
Dt dt
When v is known as a function of x and t9 expression (6.2.10b) determines
Dv/Dt directly in terms of x and /; this expression therefore serves as a
formula for acceleration in the spatial form.

246

MOTION

By using the vector identity (3.4.29), formula (6.2.10b) can be put in the
following useful form:

D\

d\

Dt

dt

= + -Vv2 + (curlv) x v

(6.2.11)

From (6.2.10b) and (6.2.11), we note that the acceleration vector is


made up of two parts: (i) the local rate of change of velocity, namely dv/dt,
and (ii) the connective rate of change of velocity, namely, (v V)v =
jVv2 + (curl v) x v. Evidently, the second part is quadratically nonlinear in
nature. Thus, the acceleration depends quadratically on the velocity field,
and a given motion cannot be viewed as a superposition of two independent
motions in general.
6.2.6

STEADY A N D UNIFORM M O T I O N S

We often deal with what are called steady and uniform motions. A motion
is said to be steady if the velocity at each point is independent of time (in the
spatial description) so that v = v(x) and d\/dt = 0. On the other hand, a
motion is said to be uniform if the velocity at each instant is independent of
position (in the spatial description) so that v = v(/) and Vv = 0. Thus,
(i) for steady motion
D\
1 ,
= (v V)v = - Vv2 + (curl v) x v
Dt
2

d\
= 0;
dt

(6.2.12)

and (ii) for uniform motion,


D\
d\
=
Dt
dt

W = 0;

6.2.7

(6.2.13)
'

LINEAR M O T I O N

While analyzing infinitesimal deformations and motions for which the


velocity is very small compared with other field functions, we approximate
the material derivative operator D/Dt to the local derivative operator d/dt.
In such a case, the motion is referred to as a linear motion. For such a
motion,
du
D\
d\
d 2u
v

^>

dt

* - * T T

Dt

dt

dt2

(6.2.14)

'

EXAMPLE 6.2.1 For a continuum rotating like a rigid body about the
origin, find the velocity and acceleration in the material and spatial forms.

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

247

Solution In the material description, a motion is represented by an


equation of the form
x = x(x,0
(6.2.15)
If this motion is a rigid rotation about the origin, we should have |x| = |x|.
Then equation (6.2.15) takes the form
x = Qx

(6.2.16)

where Q is an orthogonal tensor depending solely on t.


By use of (6.2.3), we obtain from (6.2.16),
v=^ =^x
Dt
dt

(6.2.17a)

=^ x
(6.2.17b)
Dt
dt2
Thus, for rigid rotation, (6.2.17) give the velocity and acceleration in the
material form.
From (6.2.16) we get x = Q_1x = Q r x. Using this, (6.2.17) can be
written as
v =yQrx
D\

d 2Q

= -TTQTX

(6.2.18a)

(6.2.18b)
Dt
dt2
Thus, for rigid rotation, (6.2.18) give the velocity and acceleration in the
spatial form.
Expressions (6.2.18) can be expressed in more useful forms as follows.
From Example 3.2.2, let us recall that W = (dQ/dt)QT is a skew tensor.
Hence, if w is the dual vector of W expression (6.2.18a) can be rewritten as
v = wxx
(6.2.19)
This is a well-known formula for rigid rotation about an axis through O
with w as the angular velocity vector.
Further, we note that

d^Q T_d_ fdQ


dt2^ ~ dt\dtV

dt
dt

T\

fdQ\{dtf
\dt)\dt

-(**QT)(QdQr
\dt

J\

dt

dt

248

MOTION

Hence expression (6.2.18b) can be rewritten as


D\
Dt

/d\V
\dt
dvi

= x x + w x (w x x)
dt

(6.2.20)

For rigid rotation, this gives the acceleration in spatial form, in terms of
the angular velocity vector w.

EXAMPLE 6.2.2 For a material body in motion the displacement field


is given as follows:
u2 = x2 - i(*2 + *3)e~' - (2 -

w, = 0 ,

3 = *3 - i(*2 + *)

x e

'

(6.2.21)

' + i(*2 - x*)e'

Find the velocity and acceleration fields in the material and spatial forms.
Solution The given displacement field is in the spatial form. In order to
find the velocity field, it is convenient to have the equations describing the
motion rewritten in the material form. Since i/, = xt - xf, equations
(6.2.21) may be rewritten as
*? = * i

4 = i(*2 + *3)e-' + (*2 - * 3 )e'

(6.2.22)

4 = 2(*2 + xue~* - \{x2 - xje*


Solving these equations for xx, x2, x 3 , we get
x 2 = 1(2 + x3)e' + \-(2 - x)e"

(6.2.23)

x3 = i(x2 + x3)e' - i(x2 - xl)e-'


These are the equations describing the given motion in the material form.
From these we find that
DX

r,

Dt

v2 = ?l = L (xo2 + x33 >' _ 1 ( 4 _ x0)e-<


2
Dt
2

"3 = fr= ^0+ )'+ i(4 - x>~

(6.2.24)

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

249

For the given motion, these are the velocity components in the material
form. The corresponding acceleration components are
^1 = 0
Dt

= \ (4 + * 3 V + \ (x2 - x!)e-'

(6.2.25)

\{2+>'-\(-)-'

These are also expressed in the material form.


To deduce the velocity and acceleration components in the spatial form,
we substitute for xf from (6.2.22) in (6.2.24) and (6.2.25). Thus, we obtain
Vi = 0,

v2 = x3,

v3 = x2

(6.2.26)

Dv2
Dv*
Dvx
- ^ = 0,
-=r = Xi,
- = *
(6.2.27)
2
3
Dt
Dt
Dt
For the given motion, (6.2.26) give the components of velocity and (6.2.27)
give the components of acceleration in the spatial form.

Note: Expressions (6.2.27) can also be deduced from (6.2.26) by making


use of (6.2.10a), as follows. From (6.2.26) we get
dvl
dt

dv2
dt

(v - v)Vl = 0 ,

dv3
dt

(v V)v2 = x2,

(v V)y3 = 3

(6.2.28)

Using these in (6.2.10a), we obtain expressions (6.2.27).


It may be noted that d\/dt = 0 and Ds/Dt ^ 0. The motion is therefore
steady and nonuniform.
EXAMPLE 6 . 2 . 3
given by

For a certain motion of a continuum, the velocity is


vx = - ^ 1
1+ t

(6.2.29a)

v2 = - ^ 2
1+ t

(6.2.29b)

v, = - ^ 3
1+ t

(6.2.29c)

250

MOTION

Find the velocity in the material form, and the acceleration in both material
and spatial forms.
Solution

Since vt = DxJDt,

we get from equation (6.2.29a)


Dxx _

xx

15t ~ 1 + t

(6.2.30)

This is a first-order linear differential equation. Bearing in mind that


Dxy/Dt = (dxx/dt)x<>, a general solution of this equation is obtained as
AT, = (1 + /)/(**)

(6.2.31)

where/(x") is an arbitrary function of x. Since x{ - x\ at t = 0, we get

f(xt) = *?. Thus,

x, = x?(l + 0

(6.2.32a)

Similarly, from equations (6.2.29b and c), we find that


x2 = *(1 + tf

(6.2.32b)

x3 = 3(1 +

(6.2.32c)

Substituting for x, from (6.2.32) in (6.2.29) we obtain


, = jcf,
v2 = 2*?(1 + 0,
"3 = 3x?(l + tf
(6.2.33)
For the given motion, these are the velocity components in the material form.
Expressions (6.2.33) yield the following expressions for the acceleration
components in the material form:
(6.2.34)
If we substitute for x and xl from (6.2.32b, c) into (6.2.34), we get
(6.2.35a)

Dt

(1 + 0

Dv3
Dt

6x3
(1 + 0

(6.2.35b)
(6.2.35c)

For the given motion, (6.2.35) are the acceleration components in the
spatial form. [These can also be obtained by employing the expression
(6.2.10a) to the given relations (6.2.29)].

6.2

EXAMPLE 6.2.4

MATERIAL AND LOCAL TIME DERIVATIVES

If v = V0, show that


D\
/
1
= Vl^r + -v2 .
Dt
\dt
2 )

251

v(6.2.36)

'

Solution Since v = , we have curl v = 0. Consequently, (6.2.36)


follows from (6.2.11).

Note: The result just proven shows that if the velocity is the gradient of
a potential then so is the acceleration.
EXAMPLE 6.2.5 At a point x in the current configuration of a continuum a certain physical quantity associated with the motion is given by
= ->""',
r*0
(6.2.37)
r
where r = |x| and a is a positive constant. Also, the velocity field describing
the motion is given by
vx = xYx2t,

v2 = x\t,

Vs = x2x3t

(6.2.38)

Find the material rate of change of at the point (0, a, 0) of the configuration.
Solution Since and v are given in the spatial forms, the formula (6.2.8)
is convenient for the computation of /Dt.
From (6.2.37), we find that
</>,/ = ~Ke~atrj
Using (6.2.38), we now get
.

= -3*ie~at

. = -(L\tx2e-at

(6.2.39)
(6.2.40)

Also,
= --*-
(6.2.41)
dt
r
Using (6.2.40) and (6.2.41), formula (6.2.8) gives

1
(6.2.42)
- = J + . . = (a + tx2)e~at
Dt
dt
' '
r
This is the material derivative of at the point x at time / > 0. At the point
(0, a, 0), we get
M
(6.2.43)
^ = - 0 + t)e~at

252

MOTION

EXAMPLE 6 . 2 . 6

Prove the following identities:


D_
= (Vv)F
~Dt

(i)

Dl

()

(6.2.44)

= -F _ 1 (Vv)

Dt

(6.2.45)

^ ( < / x ) = (Vv)(rfx)

(iii)

(6.2.46)
(6.2.47)

(iv)
Solution

(i) Since *, and xi;j = [F] y are functions of xf and /, we have

Dt(Xij)

dt

= ( ^ )
dxJ\Dt

\dxfj r? " dxj \ dt

= *J

(6-2-48)

Treating v-, as functions of x, and /, and using the chain rule of differentiation, we get vi;j = vikxk.j. Hence (6.2.48) can be rewritten as
D
Dt

i,kxk;j

(6.2.49)

g>- -*>

(6.2.50)

(Xij)

This proves (6.2.44).


(ii) From (6.2.44), we readily get

Also, we note that


0

Dl
D
,
D~l
= IDt
T " = Dt
^ - ( F F ) = F - ^Dt +

D
Dt

F i

Hence
-F

Dl
Dt

D
,
F"1 = Vv
Dt

by (6.2.50), from which (6.2.45) is immediate.


(iii) Recalling from relation (5.2.5) that dx = dx, we get

(tfx)
=Dt(F</x)=^V
Dt
\Dt
= (Vv)Frfx0
by (6.2.44);
= (Vv)(i/x)

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

253

(iv) Since (ds)2 = dx dx, we find


^-(ds)2
Dt

= ^-(dxDt
= 2dx-

by (6.2.46).

dx) = 2dx
'

-^-(dx)
Dt

(Vv)(rfx)

EXAMPLE 6-2.7
(0

Prove the following Euler's


DJ
r
Dt=JdWy

(ii)

formulas:
(6.2.51)

^ r ( l o g / ) = divv

(6.2.52)

Solution By definition, J = det Vx = det(jc/;7). Hence by use of expression (1.7.18), we have


J = epqrXl;pX2;qX3;r

(6.2.53)

Therefore,
DJ _
(D
D7 ~ ^^^^^^^

D
D
J^(X2>q)XUpX3,r + J^(X3;r)XUpX2;q
(6.2.54)

From (6.2.49), we note that


D
^-^;/?)

l,kxk;p

Hence
D
pqr~jZ7 \xl,p)x2,qX3,r

~ pqr
^l,kXk;pX2;qX3;r
V
l,l\pqrXl;pX2;qX3;r)

+ vU3(ePQrX3.px2.gx3.r)

l,2\pqrX2;pX2;qX3;r)

(6.2.55)

With the aid of (1.7.18), we note that the parenthetical expressions in the
last two terms of (6.2.55) represent determinants with two identical rows;
these are therefore 0. Hence (6.2.53) and (6.2.55) yield
D

pqr^(XUp)X2;qX3;r

= ^1,1

(6.2.56a)

254

6 MOTION

Similarly,

pqrjrAX2;q)X\;pXi;r
Dt

= ^2,2

(6.2.56b)

"PV r>,(X3;r)Xl;pX2;q
Dt

= M,3

(6.2.56c)

Putting (6.2.56) into (6.2.54), we obtain


DJ

Dt

JA-

*'*

which is (6.2.51). Since


D

(l J)

Dt

DJ

= JDl

expression (6.2.52) follows from (6.2.51).

EXAMPLE 6 . 2 . 8 Prove the following expressions for the material


derivatives of the surface and volume elements in the current configuration:
(i)
(ii)
Solution

i(dS)n] = {(div v) - (Vv)>(tfS)


^- (dV) = (div v) dV

(6.2.57)
(6.2.58)

(i) Recall from relation (5.2.12) that


(dS)n = J(Ty\dS0)n

(6.2.59)

This gives
D
(DJ T .
DT)~l>
1
i(dS)n] = [ ^ ( F ^ - + J yD{
HdS0)n
Noting that
D(Tyl
Dt

D(~Y
Dt

/Z)F_lxr
\ Dt

and using (6.2.45) and (6.2.51), expression (6.2.60) becomes


^{(dS)n)

= {(divv) -

(Vv)T)J(FT)-l(dS0)n

which, with the aid of (6.2.59) yields (6.2.57).

(6.2.60)

6.2

MATERIAL AND LOCAL TIME DERIVATIVES

255

(ii) Recalling the Euler's formula (5.2.22), namely, dV = J(dV0), we get

t(dV)

%{dVo)

= 7 divy dK

) o

by (6.2.51);
= (divv)JK
Note: Since dS0 and dV0 are surface and volume elements in the initial
configuration, their material derivatives are 0. This fact has been used in the
preceding computations.
EXAMPLE 6 . 2 . 9 Show that a surface represented by an equation of the
form F(x, t) = 0 is a material surface if and only if
DF
= 0.
Dt

(6.2.61)

Deduce that the normal component of velocity at a point of a material


surface F(x, t) = 0 is given by
v n

(dF/dt)
j7
\VF\

(6.2.62)

n being the positively oriented unit normal to the surface.


Solution First suppose that the surface given by the equation F(x, t) = 0
is a material surface. Then the surface consists of the same set of particles
for all time. Hence, its equation in the initial configuration should be
F(x) = 0. Consequently,

DF_ fdF
Dt " \di

= 0

which is the required condition (6.2.61).


Conversely, suppose that the condition (6.2.61) holds. Then the value of
the function F(x, /) computed for a particle currently positioned at x
remains unchanged during the motion. In particular, those particles for
which F = 0 at some time retain this value for all times. This means that the
surface represented by the equation F(x, /) = 0 contains the same set of
particles for all time. Hence the surface is a material surface.
In the spatial description, the condition (6.2.61) reads, on using (6.2.8),
d

-l + (y . V)F = 0

(6.2.63)

256

6 MOTION

We note that (v V)F = v (VF) and


n = i^i

(6.2.64)

Consequently, the relation (6.2.63) may be rewritten as


dF
+ (v n)|VF| = 0
at

(6.2.65)

from which the expression (6.2.62) is immediate.

Note: Expression (6.2.62) serves as a formula for the speed with which
the surface under consideration advances in the direction normal to itself in
space.

6-3
STRETCHING AND VORTICITY
In Section 5.4, we introduced the strain tensors that serve as measures of the
changes in length and relative orientation of material arcs that generally
occur when a material body moves from one configuration to another. We
now introduce and analyze the tensors that serve as measures of the time
rates of these changes. The tensor that serves as a means of computing the
rate of rigid rotation will also be introduced.
Let us start with expression (6.2.46); namely,
^(dx)

(6.3.1)

= (yy)dx

Substituting dx = (<is)a (which holds for a material arc dQ currently


directed along a unit vector a) in (6.3.1) and carrying out the differentiation, we obtain
(ds) j + a ^ (ds) = (<fc)(W)a

(6.3.2)

Taking the scalar product with a on both sides of (6.3.2) and noting that
a a = 1 and (therefore) a Da/Dt = 0, we get
(ds)Dt(dS)

*'(VV)a

2a "(VV

+ )1

(6 3,3)

6.3

Setting

STRETCHING AND VORTICITY

D = (Vv + Vvr) = sym Vv

257

(6.3.4)

expression (6.3.3) becomes

JL(*) = . . D .

(6.3.5)

Evidently, D defined by (6.3.4) is a symmetric tensor completely


determined by the spatial velocity gradient Vv. Also, (\/(ds))(D/Dt)(ds)
represents the rate of change in length, per unit of current length, of the
element rfC; this rate is called the stretching rate (or just stretching) ofdG
or stretching along a. Expression (6.3.5) shows that, when the current
orientation of an arc element is known, the tensor D completely determines
the stretching of the element. In a rigid motion ds = ds0, and from expression (6.3.5) it follows that a necessary and sufficient condition for a motion
to be rigid is that D = 0. Thus, the tensor D serves as a measure of stretching; this tensor is called the stretching tensor or the rate of deformation
tensor and was introduced by Euler in 1770.
Next we obtain the geometrical interpretations of the components of D.
In the component form, expressions (6.3.4) and (6.3.5) read
dij = \{pu + vjti)

W-dw

(6.3.4)'
(6.3.5)'

By taking the particular case in which a is aligned with e! we find from


(6.3.5)' that dn represents the stretching along the xx-direction. The components d22 and d33 of D have similar geometrical meanings.
In order to obtain the meanings of the off-diagonal components of D, let
us consider another material arcrfC, of current length (ds), directed along
a unit vector . If 0(^0) is the current angle between dG and dC, then
cos = a , which yields, on material differentiation,
D0
Da Da
-(sin
) = a + -a
v
Dt
Dt
Dt
Noting from (6.3.2) and (6.3.5) that
Z>a
_ = (Vv)a - (a Da)a
and, similarly,

Da
= (Vv) - ( D)

(6.3.6)

258

MOTION

we obtain
a

Da
Da
+
= a (Vv)a + a (Vv)a - (a a)(a Da + a Da) (6.3.7)
Dt
Dt

Also, by use of (2.8.14) and (6.3.4) we get


a (Vv) + (Vv)a = a (Vv + Vvr) = 2a Da

(6.3.8)

Using (6.3.7) and (6.3.8), expression (6.3.6) becomes


)
(sin ) = (a )(a Da + Da) - 2a Da

(6.3.9)

This shows that D not only determines the stretchings of material arcs,
but also the time rate of change in the angles between them.
If = /2, we have a = 0 and (6.3.9) reduces to the following simple
form:
(6.3.10)
a Da = -l-^2 Dt
This expression shows that a Da represents one-half the rate of decrease
in the angle suffered by two (currently) orthogonal material arcs. This rate
is called the shearing rate or just the shearing between the arcs. Evidently,
the arcs suffer no shearing if and only if the motion is rigid.
For the particular case in which a is aligned with e! and is aligned with
e 2 , we find from (6.3.10) that dl2 represents the shearing between material
arcs currently lying along the xx and x2 directions. The other off-diagonal
components of D have similar geometrical interpretations.
Thus, all the components of D have definite geometrical interpretations.
While the diagonal components of D represent the stretching of material
arcs currently lying along the coordinate directions, the off-diagonal
components represent the shearing between such arcs. These geometrical
interpretations are analogous to those of the components of the strain
tensor A (which is also defined in the current configuration). But there is an
important difference: whereas the meanings of the components of D are
exact, the meanings of the components of A are just approximations.
Since D is a symmetric tensor, it follows from results proven in Section
2.13 that D possesses three eigenvalues and at least one set of three mutually
orthogonal principal directions and that, with respect to the axes chosen
along these principal directions, the matrix of D is purely diagonal, the
diagonal elements being the eigenvalues of D. By virtue of the geometrical
interpretations of the components of D obtained previously, it follows that
the eigenvalues of D represent the stretchings along the principal directions

6.3

STRETCHING AND VORTICITY

259

of D. The eigenvalues of D are called the principal stretchings and the


principal directions of D are called the principal directions of stretching. As
in Section 5.3, it can be shown that the stretching of a general material arc
is the net result of triaxial principal stretchings.
6.3.1

ISOCHORIC MOTION

From (6.3.4) we find that trD = divv. Hence expression (6.2.58) yields

trD = divv =

(rfTf (rfF)

(6 311)

Evidently, tr D = (div v) represents the rate of change in volume per unit


volume of a material element (as it moves with velocity v). Further, dV is
constant during the motion if and only if div v = 0. From Euler's formula
(6.2.51) we note that div v = 0 if and only if DJ/Dt = 0; that is, 7 remains
unchanged during the motion. Initially, when t = 0, we have 7 = 1 . Hence
div v = 0 if and only if 7 = 1.
Thus, the following statements are equivalent:
1.

dV = constant

(6.3.12a)

2.

divv = 0

(6.3.12b)

3.

7=1

(6.3.12c)

A motion is said to be an isochoric (volume-preserving) motion whenever


dV = constant during the motion. It follows that a motion is isochoric if
and only if div v = 0 or equivalently 7 = 1 during the motion. Note that
this statement is consistent with that made in Section 5.2 in respect of an
isochoric deformation.
In Section 5.2, it has been indicated that a continuum is said to be an
incompressible continuum if and only if every deformation it undergoes is
an isochoric deformation. Since a motion is a (one-parameter) family of
deformations, it follows that a continuum is incompressible if and only if all
its motions are isochoric motions. Consequently, it is inferred that for every
motion of an incompressible continuum each of the conditions in (6.3.12)
holds. The condition (6.3.12b), namely, div v = 0, plays a fundamental role
in the study of motions of incompressible fluids. Some further aspects of
the conditions (6.3.12) will be considered in Section 8.2.
6.3.2

ANALYSIS OF RELATIVE MOTION

We return to expression (6.3.1) and analyze it on slightly different lines. We


note that dx may be regarded as the position vector of the point (x + dx)

260

MOTION

Figure 6.1. Relative velocity.

with respect to the point x. Consequently, (D/Dt) (dx) may be interpreted


as the velocity at the point (x + dx) relative to the point x. Let us denote this
relative velocity byrfv;see Figure 6.1. Thus, (6.3.1) may be rewritten as
rfv = (Vv)(rfx)

(6.3.13)

This expression shows that the velocity gradient Vv, which is a tensor,
serves as a means of computing d\. Let us decompose Vv into symmetric
and skew parts. It has already been noted that D is the symmetric part of Vv
(by definition). Let W be the skew part of Vv; that is,
W = (Vv - Vvr) = skw Vv

(6.3.14)

Then expression (6.3.13) becomes


d\ = D(rfx) + W(rfx)

(6.3.15)

Thus, d\ is made up of two parts: (i) dyx = D(tfx), and (ii) d\2 = W(rfx).
By virtue of the geometrical significance of D, d\x may be interpreted as the
velocity caused by the stretching. To interpret d\2, let us look for the dual
vector of W. Since W = skw Vv, we find from Example 3.5.7 that the dual
vector of W is \ curl v. Hence
d\2 = W(rfx) = (\ curl v) x (dx)

(6.3.16)

This expression is analogous to the expression (6.2.19) giving the velocity


due to rigid rotation. Comparing these two expressions, we find that d\2
represents the velocity due to a rigid rotation about an axis through the
point x, with angular velocity \ curl v. Thus, d\ may be considered a superposition of the velocity caused by the stretching determined by D and the
velocity due to the rigid rotation determined by W.
In view of its significance as described earlier, the tensor W is called
the spin or vorticity tensor. Also, curl v, which is equal to twice the dual

6.3

STRETCHING AND VORTICITY

261

vector of W, is called the spin or vorticity vector. We denote this vector by


w; that is,
w = curlv
(6.3.17)
The concept of vorticity was first introduced and analyzed by Cauchy in
1841.
6.3.3

IRROTATIONAL MOTION

In the particular case when W = 0 at x, rfv is completely determined by D


and there occurs no rigid rotation about x; in such a situation we say that
the motion is irrotational. Thus, a motion is irrotational at a point x if and
only if the vorticity tensor W and (therefore) the vorticity vector w are Oatx
6.3.4

POTENTIAL FLOW

We often deal with a special type of motion wherein the velocity is the
gradient of a scalar function; that is, v = for some scalar function .
Such a motion is called a potential flow and the function is called the
velocity potential of the flow. Since curl V0 = 0 b y identity (3.4.16), it
readily follows that every potential flow is automatically an irrotational
motion. The converse is valid only if the region of interest is simply connected; the proof requires the use of the Stokes's theorem (3.6.7). Thus, in
a simply connected region, every irrotational motion is a potential flow.
From the result of Example 6.2.4, we note that, in a potential flow, the
acceleration is also the gradient of a scalar function.
Isochoric motions, irrotational motions and potential flows are of great
importance, particularly in fluid mechanics.
6.3.5

COMPARISON WITH SMALL DEFORMATION

There is a striking resemblance between the set of expressions (6.3.4) and


(6.3.14), through which D and W have been defined, and the set of
expressions (5.6.4), (5.7.6) and (5.7.38), relating the strain tensor E and
the rotation tensor with the displacement vector occurring in small
(infinitesimal) deformation. Although all of these expressions are linear, the
linearity in the expressions (6.3.4) and (6.3.14) is exact, while that in (5.6.4),
(5.7.6) and (5.7.38) is just approximate. In other words, while D and W
measure the stretching and the angular velocity (due to rigid rotation)
exactly, E and measure the stretch and angular displacement (due to rigid
rotation) only to the first-order approximation. In particular, it must be
noted that, whereas D is exactly 0 in a rigid-motion, E need not be 0 in such
a motion\ see Example 5.6.3. However, because of the formal similarity

262

6 MOTION

between the expressions (6.3.4) and (5.6.4), all the results obtained in
respect of E on the basis of (5.6.4) can be rewritten in respect of D with the
words displacement and strain tensor replaced by the words velocity and
stretching tensor.
From expression (5.6.4), we obtain
DE
~Dt

2 I Dt

2Dt

\DtJ )

= |(Vv + Vv )
using (6.2.5);
|(Vv + Vv r )
for small deformation;
= D
by (6.3.4). Similarly, from (5.7.6), (5.7.38), (6.2.5) and (6.3.14), we obtain
DQ/Dt = W.
Thus, for small deformation, we have
DE
D =
Dt

(6.3.18a)

DQ
Dt

(6.3.18b)

W =

These expressions show that, when the displacement gradient is infinitesimally small, D represents the rate of change of the strain tensor and W
represents the rate of change of the rotation tensor. For this reason, D is
often referred to as the rate of strain or strain-rate tensor, and W as the rate
of rotation or rotation-rate tensor. When the deformation is not small, the
strain tensor is either G or A. Then D does not represent the actual rate of
change of G or A; see Example 6.3.3 later.
EXAMPLE 6.3.1

Prove the following identities:

(i)

( v V)v = w x v + }v|v| 2

(ii)

div(v V)v - (v V)(div v) = |D| 2 - | | w | 2 = Vv Vv r (6.3.20)

(iii)

(6.3.19)

div(v x w) = i(V 2 |v| 2 + |w| 2 )


- f|D| 2 + ( v V)(divv))

(6.3.21)

Solution (i) By use of the vector identity (3.4.29) and expression (6.3.17)
we readily obtain (6.3.19).

6.3

STRETCHING AND VORTICITY

263

(ii) Recalling that D = sym Vv and that yw = \ curl v is the dual vector
of W = skw Vv, we obtain (6.3.20) by changing the notation appropriately
in (3.5.46).
(iii) Taking divergence of (6.3.19) and using (6.3.20), we arrive at
(6.3.21).

EXAMPLE 6 . 3 . 2
(i)
(ii)

Prove the following identities:


div( ) = -^-(div v) + |D| 2 - \ |w| 2
\Dt
Dt
2

(6.3.22)

curl( ) = + curl(w x v)

(6.3.23)

\pi) ~~dt

= + (div v)w - (w V)v


(iii)
Solution

(6.3.24)

(TF-'w) = / F " 1 c u r / ^

(6.3.25)

(i) By using (6.2.8), we get

D
/
x d ,
x
fdvk
j^(Vk,k) = j(vk,k) + t>i*>*,*i = y-^f)

dvk
dt

ViVkfi

vitkvkJ

+ (PiVkAk - Vi.kVkj
= l-^rj
\Dt/tk

vitkvkti

or, in the direct notation,


D
/D\\
(div v) = divi ) - Vv - Vv r

(6.3.26)

Recalling from (6.3.20) that


V v V v r = |D| 2 - ^-|w|2
expression (6.3.26) readily yields (6.3.22).
(ii) Taking the curl of expression (6.2.11), we obtain (6.3.23). Since
divw = div curl v = 0, the identity (3.4.25) gives
curl(w X v) = (div v)w + (v V)w - (w V)v
Using this in (6.3.23), we obtain
/Z>v\
aw
curl ) = - + (v V)w + (div v)w - (w V)v
\DtJ
dt
which in turn becomes (6.3.24) on using (6.2.8).

264

6 MOTION

(iii) Operating both sides of the expression (6.3.24) by J


use of (3.5.34) that (w V)v = (Vv)w, we obtain

D\\

= JF~l

DtJ

Dw

Dt

and noting by

+ /(div v)F~'w - 7F -1 (Vv)w

(DJ
,
- \ - y(divv)JF-'W

D
.
= ^(JV-'w)

_/

+ F 1(Vv) w

[^r "

(6 3 27)

--

Substituting for DJ/Dt and D~ /Dt from (6.2.51) and (6.2.45), respectively, in (6.3.27), we arrive at (6.3.25).

EXAMPLE 6 . 3 . 3

Show that

D = (F-yf^V1

=
Solution

w+ (A(Vv) + ()|

(6.3.28)

(6 3 29)

From (5.4.10) and (5.4.28), let us recall that


G = i ( F r F - I)

(6.3.30a)

A = |(I - ( F - y F - 1 )

(6.3.30b)

Using expressions (6.2.44) and (6.3.4) we find from (6.3.30a) that


DC
1
^ r = 2[{(Vv)F)rF

Fr(Vv)F]

= r [F r (Vv r + Vv)F] = F r DF
from which (6.3.28) is immediate.
Using expressions (6.2.45) and (6.3.4), we obtain from (6.3.30b) that
DA

-^ = -1WvTWY*-X

+ (F-yF-'(Vv)]

= ^ [ ( V v r ) ( I - 2 A ) + (I-2A)(Vv)]
= D - (Vv r )A - A(Vv)
from which (6.3.29) follows.

6.3

STRETCHING AND VORTICITY

265

Note: From expressions (6.3.28) and (6.3.29) it is evident that D does not
generally represent the material rate of change of Lagrangian or Eulerian
strain tensors.
EXAMPLE 6.3.4 For the flow defined by velocity components of the
form vx = vl(xl9x2, t), v2 = v2(xl9x2, 0> ^3 = 0 (such a flow is called a
plane flow), show that
DW + WD = (divv)W
Solution

(6.3.31)

By using the identity (3.5.47), we note that


DW + WD = skw(Vv)2

so that
[DW + WD]l7 = \(vifkvkJ

(6.3.32)

- vLkvkJ)

For the given vi9 this yields the following expression for the matrix of
DW + WD:
0
[DW + WD] = (vUi + v2f2)

2(^1,2 -

i(vU2 + 2) 0
^2,l)

= (div v)[W]
This proves (6.3.31).

EXAMPLE 6.3.5 Find the stretching and vorticity tensors for the flow
described by the velocity components
AR)

f(R)

]jT*i.

v2 =

3 = 0

(6.3.33)

where R = (jcf + xf) 1/2 ^ 0. Deduce that the motion is isochoric.


Further, show that the flow defined by
Vi =

~^'

"3 = 0

* = la>
2

(6.3.34)

is irrotational as well.
Solution For the velocity components given by (6.3.33), we find that
2
[Vv]

-{

+ 2)

-2

0
0

(6.3.35)

266

MOTION

where

= >
R2

and

f\R) - ^/(*)

jKR)

Noting that xx = R cos 0, x2 = R sin 0, where /? and are plane polar


coordinates, we obtain from (6.3.35) that
sin 2 - cos 2 0
[D] = i[Vv] + i[Vv] r =

-</> cos 2 - sin 20


0

(6.3.36)

and

0 ^ 0
[W]

= i[Vv] -- i[Vv] =

0 0

0 0

(6.3.37)

where

= \R2^

= \[f\R)

+ ]

Expressions (6.3.36) and (6.3.37) give the matrices of D and W for the
given motion. From (6.3.36), we note that trD = 0. Hence the motion is
isochoric.
Expressions (6.3.34) correspond to the particular case/(i?) = -\/R of
(6.3.33). In this case we find that = 0. Consequently, W = 0 by (6.3.37),
and the motion is irrotational as well in addition to being isochoric. [This
can be proven directly by showing that for vt given by (6.3.34), divv = 0
and curl v = 0.]

6.4
PATH LINES, STREAM LINES AND VORTEX LINES
6.4.1

P A T H LINES

Let us recall the equation (4.5.1), namely


x = x(x 0 ,/)

(6.4.1)

representing the motion of a material body in the Lagrangian description.


As already pointed out, for a specified particle (P this equation gives the
successive positions x of (P with increasing /. In other words, equation
(6.4.1) represents the path in which the particle (9 moves during the motion
of the continuum. This path, which is a space curve having its initial point
as x, is called the path line of the particle (P; see Figure 6.2.

6.4

PATH LINES, STREAM LINES AND VORTEX LINES

267

Figure 6.2. Path line.

For example, consider the motion described by the equations


xi = sin(ox? + t)
x2 = 1 - cos(oa2 + t)
x3=

(6.4.2)

where a and are constants. For the particle (P0 initially located at the
origin, equations (6.4.2) become
xx = sin t,

x2 = 1 - cos t,

x3 = 0

(6.4.3)

These equations determine the location x of the particle (P0 at any time t.
Rewriting the equations we get
x\ + (x2 - l) 2 = 1,

x3 = 0

for all /

(6.4.4)

Therefore, the path line of the particle (P0 is the circle x\ + (x2 - l) 2 = 1 in
the 3 plane.
Reverting to the general discussion, suppose that, for a certain motion,
the velocity is known at a point x of the current configuration. By the
definition of v, we have
(6.4.5)
Since v is known, equation (6.4.5) may be viewed as a differential equation
for x and integrated under the (initial) condition x = x at t = 0; the
resulting solution will be in the form of (6.4.1). Thus, (6.4.5) serves as the
differential equation for path lines.

268

MOTION

For example, consider the motion for which the velocity field is as
follows:
3*3
2x2
X\
(6.4.6)
v3 = 1 + t
v
=
2
1 + *'
7
For this field, equation (6.4.5) has been integrated in Example 6.2.3 and the
following solution has been obtained; see (6.2.32):
xx = *?(1 + 0,

*i = 4(1 + t)\

x3 = x?(l + tf

(6.4.7)

These are the equations for the path line of the particle whose current
velocity is given by (6.4.6).
6.4.2

FIELD LINES

Consider a geometrical arc y in the current configuration of a continuum.


Let its equation be x = (), where is a parameter. From threedimensional geometry, we recall that the vector dx/dr is tangential to y at
the point x on it. Suppose f is a non-zero vector known at every point of the
configuration. Then it follows that at any point x on y the tangent to y is
directed along f if and only if
f = a

dx
dx

(6.4.8)

for some positive real number a. Such an arc is called the field line of f.
Thus, the field line of a vector f at a point x is an arc whose tangent at x is
directed along f; see Figure 6.3.

Figure 6.3. Field line.

6.4

PATH LINES, STREAM LINES AND VORTEX LINES

269

For a given f, the field line of f satisfies the differential equation (6.4.8)
by definition. When integrated, this differential equation yields a solution
of the form x = (), which represents the field line of f. Equation (6.4.8)
is therefore the differential equation for the field line of f. This equation
can be rewritten in the following two forms:
fi

rfx,

dXj

dx3

dx2
" fi

6.4.3

(6.4.9)

dz
=

(6.4.10)

STREAM LINES AND VORTEX LINES

The two vectors that generally determine the nature of the motion of a continuum are the velocity and the vorticity. The field lines of our major
interest here are therefore those of velocity and vorticity. The field lines of
velocity are called the stream lines and those of vorticity are the vortex lines,
Thus, a stream line is an arc in the current configuration such that the
tangent to the arc at any of its points is directed along the velocity vector at
that point. The meaning of a vortex line is analogous. The differential
equations for a stream line and a vortex line are obtained by replacing f (or
fi) by v (or ) and w (or w,), respectively, in the equations (6.4.8) to
(6.4.10). It is important to note that, since the velocity and the vorticity
vectors are uniquely determined at a point, there cannot be two distinct
stream lines and two distinct vortex lines at a point. In other words, two
stream lines cannot cross each other; neither can vortex lines. It is obvious
that the stream lines do not make their appearance when there is no motion,
and that the vortex lines are nonexistent in an irrotational motion.
From their definitions, it is evident that the path lines and stream lines are
not the same. Whereas a path line is the trajectory of a specific particle as
it moves from its initial location, a stream line is a curve tangential to the
velocity of the particle at its current location (Figure 6.4). However, if the
motion is steady, that is, if d\/dt = 0, the path lines and the stream lines
become coincident, because, when d\/dt = 0, the differential equation for
the path lines, namely, (6.4.5), yields the solution
x = /v(x) + x

(6.4.11)

which evidently represents the curve tangential to v.


However, the converse is not true. To see this, let us consider the motion
for which the velocity is given by (6.4.6). For this motion the path lines
are given by (6.4.7), as already seen. Let us now find the stream lines. The

270

MOTION

Figure 6.4. Stream line.

differential equation for these lines, namely


dxx _ dx2 _ dx3
vl
v2
v3

(6.4.12)

are
djCj

dx2

dx3

dt

xl

2x2

3x3

\ + t

(6.4.13)

The solution of these equations obeying the condition x = x for / = 0 is


given by
xx = x 0 (l + /),

x2 = x2(l + t)\

x3 =

3(1

+ O3 (6.4.14)

These equations are identical with equations (6.4.7). Hence, for the motion
given by (6.4.6), the stream lines and the path lines are coincidental. But, we
note from (6.4.6) that d\/dt ^ 0; that is, the motion is not steady. Thus, the
path lines and stream lines can be coincident in a nonsteady motion also.
This analysis shows that whereas for a steady motion the path lines and the
stream lines are necessarily coincident, the coincidence of these lines does
not ensure that the motion is steady.
Returning to the motion for which the velocity is given by (6.4.6), we
note that w = 0 for this motion. As such, the motion is irrotational and
(therefore) the vortex lines do not exist.

6.4

PATH LINES, STREAM LINES AND VORTEX LINES

271

6.4.4 STREAM TUBES AND VORTEX TUBES


Consider a simple closed curve C in the current configuration and visualize
the tube formed by the field lines of a vector f, initiating from the points of
C. This tube is called afield tube of f at C If S is a section of the tube, then
the integral \s f n dS is called the strength of the tube at S. In particular,
the field tube of v is called the stream tube, and the field tube of w is called
the vortex tube; see Figure 6.5.
Consider two sections S{ and S2 of a field tube of f (Figure 6.6). Let V be
the volume of the region bounded by Sl9 S2 and the curved surface S' of the
tube. Then, by the divergence theorem (3.6.1), we obtain
di\fdV=\

f-nrfS+|

f-ndS+\

t-ndS

(6.4.15)

where n is as usual the unit outward normal.


Since S' is formed by the field lines of f, f n = 0 on S'. If the distance
between Sj and S2 is so small that the tube can be considered straight, then
n has opposite directions on S{ and S2. Consequently, expression (6.4.15)
becomes
divfdV

. t-ndS-\

s2

fndS

Vortex tube
<^stream tube
Figure 6.5. Vortex tube and stream tube.

Figure 6.6. Field tube between two sections.

(6.4.16)

272

MOTION

For f = w, the field tube of f is the vortex tube and in this case (6.4.16)
yields, on using the fact that div w = 0,

wnrfS
wnrfS =

w-nrfS

(6.4.17)

This expression shows that the strength of the vortex tube at Sx is the same
as that at S2. In other words, the strength of a vortex tube is constant. This
result is essentially by Helmholtz (1858).
For f = v, the field tube of f is the stream tube and in this case (6.4.16)
yields
divv</K=

vnrfSvnrfS .
(6.4.18)
(Jsi
Js 2
)
If the motion is isochoric, that is if div v = 0, we find from (6.4.18) that
]B

\-ndS

y-ndS

(6.4.19)

Thus, in an isochoric motion, the strength of a stream tube is constant.


For an incompressible continuum, every motion is isochoric. Hence, for
every motion of an incompressible continuum, the strength of a stream tube
is constant.
EXAMPLE 6-4.1 Show that the stream lines and the vortex lines are
orthogonal if and only if v = () for some scalar functions and .
Solution We note that the stream lines and the vortex lines are
orthogonal whenever v and curl v are orthogonal. From Example 3.4.6, we
recall that this is possible if and only if v = , for some and .

EXAMPLE 6-4.2

For the flow defined by the velocity field


v = (1 + at)ex + xxt2

(6.4.20)

where a is a constant, find the path lines, stream lines and vortex lines.
Solution Recall that the path lines are governed by the differential
equation (6.4.5). For the given motion, this equation yields
^

= (1 + at)

(6.4.21a)

Dx2
-^f = xi

(5.4.21b)

(6.4.21c)

= 0

6.5

TRANSPORT FORMULAS

273

Integration of the first of these equations gives, on using the fact that
ATj = Xi for t = 0, the solution
xx = JC? + (/ + \at2)

(6.4.22a)

Using this expression in equation (6.4.21b) and integrating the resulting


equation under the condition x2 = x% for t = 0, we get
x 2 = Je? + * ? ' + jt2 + at3
Equation (6.4.21c) gives
x 3 = JC?

(6.4.22b)
(6.4.22c)

For the given motion, (6.4.22) constitute the parametric equations for the
path lines.
Next, recall that the stream lines are governed by the equations (6.4.12).
For the given motion, these equations become
JC3 = constant

(6.4.23)

These equations yield, on using the fact that xt = xf for / = 0, the equations
(6.4.24)
For the given motion, these are the equations of the stream lines.
From the given velocity field, we get w = e3 ; therefore, the vortex lines
are given by dx/dr = ae3, which yields, on using the conditions xt = xf for
/ = 0, the equations
(6.4.25)
Evidently, vortex lines are straight lines in the x3 direction.
Note: For a = 0, equations (6.4.22) yield
(6.4.26)
When a = 0, equations (6.4.24) also reduce to (6.4.26). Thus, for the given
flow, the stream lines and the path lines coincide in the case of a = 0. This
result is expected because for a = 0 the given flow is steady.

6-5
TRANSPORT FORMULAS
In Examples 6.2.6 and 6.2.8, the material derivatives of a material arc
element, a material surface element and a material volume element located
in the current configuration have been computed. We now generalize these

274

MOTION

expressions by deriving formulas for the material derivatives of integrals


over material curves, material surfaces and material volumes located in
the current configuration. These formulas, which are usually called the
transport formulas; are of great utility in deriving the balance laws of mass,
momentum and energy. The formulas are contained in the following.
THEOREM 6-5.1 Let denote a scalar, a component of a vector or a
component of a tensor known in the spatial form. Then for a material curve
6, material surface S and a material body (B occupying a curve C, a surface
S and a volume V, respectively, in the current configuration, the following
formulae hold:
(0
(ii)

(iii)

- l = |

Dt

J M dS =

BH

^ + 0(Vv)[tfx

(6.5.1)

f ( ^ + 0 div vj - 4>(Vv)H n

\-\

rfS

(6.5.2)

(6 5 3)

^)

The general method of proving each of the formulas (6.5.1) to (6.5.3) is


as follows. We first transfer the expression on the lefthand side of the
formula to the initial configuration by changing the variables from xt and t
to x? and t. The range of integration then becomes a curve-surface-volume
in the initial configuration. Since the material derivative is the partial
derivative with respect to t, keeping xf fixed, the material differentiation
can be carried out under the integral sign in the initial configuration. After
this differentiation process is over, we transfer the expression back to the
current configuration to arrive at the final result. The details follow.
Proof (i) Let C0 be the initial location of the material curve G that is
currently located on C. Then dx is related to the arc element dx of C0
through the transformation dx = F dx; see (5.3.1). Hence, the change of
variables from xt to xf yields
(6.5.4)
(, t)dx=\
{(, 0, t]F dx
Jc
Jc0
Taking the material derivative on both sides of (6.5.4) and bearing in mind
that C0 is a curve in the initial configuration, we get
Dt]c

]c0DtKV

}c\Dt

6.5

TRANSPORT FORMULAS

By the use of (6.2.44), this becomes

sL*A-L(i+H'*'

275

(6 5 5)

Transferring the expression on the righthand side of (6.5.5) back to the


current configuration with the use of (5.3.1), we obtain the formula (6.5.1).
(ii) Let S0 be the initial location of the material surface S that is currently
located on 5. Then the oriented surface element (n dS) taken on S and the
oriented surface element (n dS0) taken on S0 are related to each other
through the formula (5.2.12), namely, (dS)n = J(FTyln(dS0). Hence
is

</>nrfS=

Js0

(l>J(FTyln0 dS0

(6.5.6)

Taking the material derivative on both sides of (6.5.6) and bearing in mind
that S0 is a surface in the initial configuration, we get

J M < =

jl<t>J<F-l)T}n0dSo

(6.5.7)

Carrying out the material differentiation in the righthand side of (6.5.7) and
using the relations (6.2.45) and (6.2.51), we obtain
^

\ <t>ndS =

&

+ </>divv - )](-)0dS0

(6.5.8)

Transferring the expression on the righthand side of (6.5.8) back to the


current configuration with the use of (5.2.12), we obtain the formula
(6.5.2).
(iii) Let V0 be the initial volume of the material body (B that is currently
occupying the volume V. Then the volume element dV of V is related to the
volume element dV0 of V0 through the formula dV = JdV0; see (5.2.22).
Hence
v

4>dV=\ <t>JdV0
J Vo

(6.5.9)

Taking the material derivative on both sides of (6.5.9) and bearing in mind
that V0 is a volume in the initial configuration, we get
[ D
(D
DJ\
Dt (d>J)dV0=\
\-Zj+<i>Dt
<t>dV=\

1z:)dVQ
}v
}vo
}
By use of (6.2.51), this becomes
D [

Dt

wt\v*dv=\M+d^]jdv

(6 5 )

276

MOTION

Transferring the expression on the righthand side back to the current


configuration with the use of (5.2.22) we obtain the formula (6.5.3). This
completes the proof of the theorem.

6.5.1

REYNOLD'S TRANSPORT FORMULA

Here we obtain an alternative version of the formula (6.5.3). By use of the


identity (3.4.14), the expression for the material derivative, given in (6.2.8),
yields
Do
do
- + </>divv = - ^ + div(</>v)
(6.5.11)
Dt
dt
Using this, the formula (6.5.3) becomes

<=\+)}

(6 5 2)

Uv*dv=\Mdv+\s*-nds

(6 5 13)

which on using the divergence theorem (3.6.1) reduces to

where S is the surface enclosing K.


Expression (6.5.13), which is an alternative version of (6.5.3), is known as
the Reynold's transport formula (theorem), after O. Reynold (1903).
This expression shows that the rate of change of the integral of over a
material volume is made up of (i) the integral of the local rate of change of
in the volume, and (ii) the flux of across the surface enclosing the
volume.
Note: For simplicity in the terminology, the curve C, the surface S and
the volume V, which are the current locations of a material curve 6, a
material surface S and a material volume V (respectively) are themselves
often referred to as a material curve, a material surface and a material
volume (respectively).
EXAMPLE 6.5.1

Prove the following:

(ii)
(iii)

(VvMx

(6 5 14)

ndS =\ {divv - (y\)T]ndS

(6.5.15)

i\cdx=\

(0

^-\

Dt

is

Js

dV=\

(aiyy)dV

(6.5.16)

6.5

TRANSPORT FORMULAS

277

Solution Taking = 1 in the transport formulas (6.5.1) to (6.5.3), we


readily obtain the results (6.5.14) to (6.5.16).

Note: Expression (6.5.16) can be rewritten as


= 1 div\dV=\
Dt

Jv

]v

(trD)rfF

(6.5.17)

Thus, the rate of change of a material volume is equal to the integral of


divv on that volume. Expression (6.5.17) is analogous to the expression
(5.6.26) for the change in volume.
EXAMPLE 6.5.2 Let f be a vector function defined in the current
configuration and C be a material curve in the configuration. Show that
|M

frfx =

| ^ f i f x + f(Vv)rfxj

(6.5.18)

Deduce that

Solution

(6 5 9)

i\cf-dx=\M'dx+(vy)Tf'dx}

From the transport formula (6.5.1), we have

i\c*dXj = \c(^dXj + *Vj-kdXk)


Setting = fj in this expression, where/) are the components of the given
vector f, we get

t\cf'dxj=\Mdxj+fiV-kdXk
which is precisely (6.5.18).
Taking the trace of expression (6.5.18) and using the fact that
tr{a (Ab)} = a Ab = (Ara) b, which follows by use of (2.4.35) and
(2.8.14), we obtain (6.5.19).

EXAMPLE 6.5.3 Let f be a vector function defined over the current


configuration. For a material surface S in this configuration, show that
D f

fn</S=

(Di _

_ _

. _ ,_

n + (div v)f n - f (Vv)'n} dS (6.5.20)

278

6 MOTION

Deduce that
^ f f ndS = f ^ + f(div v) - (f V)v! ndS
= I j + (divf)v + curl(f x v)j ndS
Solution

(6.5.21)
(6.5.22)

From the transport formula (6.5.2), we have


D [
Dt

IO

</>/?, tfS =

f (
n
\j^ j

+ <t>vk,k"j - <t>Vkj*k\ dS

Setting = ff in this expression, we get


D

Dt

[ r

. firijdS=

f {Vfi

nj+Mv^^nj-fiV^jn^dS

This is precisely (6.5.20).


Taking the trace of expression (6.5.20) and using (2.4.35) and the fact
that f (Vv)rn = (Vv)f n = (f V)v n, which follows by use of (2.8.14)
and (3.5.34), we obtain (6.5.21). Rewriting the righthand side of (6.5.21) by
using the vector identity (3.4.25) and the result
7- = T- + (v V)f

Dt
we obtain (6.5.22).

dt

Note: It follows from (6.5.22) that the flux of a vector f across every
material surface is constant if and only if
(6.5.23)
This is known as Zorawski's criterion.

6.6
CIRCULATION AND VORTICITY
In Section 3.6, the circulation of a vector field f round a simple closed curve
C was defined as the line integral | c f ids. We now study this concept
when f is the velocity vector and C is a simple closed curve in the current
configuration.

6.6

CIRCULATION AND VORTICITY

279

Let C be a simple closed curve (circuit) in the current configuration. Then


the circulation of the velocity vector v around C will henceforth be referred
to as the circulation round C, denoted Ic. Thus,
= \ Y-dx
(6.6.1)
ic
Jc
If C is a reducible curve, that is, if there is a regular open surface S
bounded by C (which lies completely within the configuration), then by
using Stokes's theorem (3.6.7), expression (6.6.1) can be rewritten as
Ic=

vtds

Ic=

(6.6.2)

yvndS

This relation shows that the concept of circulation is closely related to the
concept of vorticity; in fact Ic represents the flux of w across S. It is evident
that Ic = 0 for every reducible curve C if and only if w = 0. In other words,
a motion is irrotational in a given region if and only if the circulation round
every reducible circuit in the region is 0.
6.6.1

MATERIAL DERIVATIVE OF CIRCULATION

Suppose that C is a material curve in the current configuration. Then


DIc/Dt can be computed with the aid of a transport formula. In fact,
taking f = v in (6.5.19) we readily get
DI

c
Dt

1 \Ds ^
)c(Dt

+ ( V v ) ,Tv

) y dxl

(6 6 3)

Since (Vv)rv = |V(v v), we find that


)c

(Vv)rv dx = -\

2 Jc

V(v2) dx = -\

2 Jc

dv2 = 0

because C is a closed curve. Hence, expression (6.6.3) becomes


D\
, ^ dx
(6.6.4)
Dt
]cDt
This formula for the material derivative of circulation is given by Lord
Kelvin (1869).
DIC

6.6.2

KELVIN'S CIRCULATION THEOREM

As a direct consequence of (6.6.4), we now prove the following famous


result, known as the Kelvin's circulation theorem.

280

6 MOTION

THEOREM 6.6-1 If the acceleration is the gradient of a potential, then


the circulation round a material curve remains constant in time.
Proof If D\/Dt is the gradient of a potential, that is, if Dv/Dt = V^ for
some scalar function , we have
DV

dx = \ Vw - dx =

dw = 0

Then expression (6.6.4) becomes


DIr
- = 0 or Ic = constant (w.r.t. /)
Dt
This completes the proof the theorem.
6.6.3

(6.6.5)

CIRCULATION-PRESERVING MOTION

A motion is said to be circulation preserving, if the circulation round every


material curve remains constant in time. A necessary and sufficient condition for a motion to be circulation preserving follows immediately from
the Kelvin's formula (6.6.4) and is given by the following theorem.
THEOREM 6.6.2 A motion is circulation preserving if and only if the
acceleration is an irrotational vector.
Proof

By employing Stokes's theorem (3.6.7) to (6.6.4), we obtain


[
jDs
DIC
^ = curl

Dt

)s

nds

\Dt)

(666)

From this expression, it is evident that DIc/Dt = 0 for every C if and only
if cur\(D\/Dt) = 0. This completes the proof.

Note: In a potential flow, the acceleration is the gradient of a scalar


function; see Example 6.2.4. Hence it follows from this theorem that a
potential flow is circulation preserving.
COROLLARY For a circulation-preserving motion,
(/-*) = 0
Proof

The result follows by setting cur\(D\/Dt) = 0 in (6.3.25).

(6.6.7)

6.6

6.6.4

CIRCULATION AND VORTICITY

281

CAUCHY'S VORTICITY EQUATION

Integrating equation (6.6.7) and noting that J = 1 and F = I at / = 0, we


get
7F _ 1 w = w
(6.6.8)
where w = w(x) is the vorticity vector at / = 0.
Equation (6.6.8) exhibits a simple relationship between the current
vorticity and the initial vorticity, in a circulation-preserving motion. This
remarkable equation is known as Cauchy's vorticity equation. We now give
two important consequences.
6.6.5

PERMANENCE OF IRROTATIONAL MOTION

THEOREM 6 . 6 . 3 If a circulation-preserving motion is once irrotational,


then it is irrotational throughout the history of the motion. (This is known
as the Cauchy-Lagrange theorem.)
Proof

From equation (6.6.8), we readily get

w = JFW

(6.6.9)

It is evident that if w = 0, then w = 0. Hence if the vorticity is initially 0,


then it is 0 for all time. This completes the proof of the theorem.

6.6.6

MOTION OF VORTEX LINES

THEOREM 6 . 6 . 4 In a circulation-preserving motion, the vortex lines


are transported with the motion. This theorem is known as the Helmholtz
theorem.
Proof Consider a vortex line through a point x in the initial configuration. Then the tangent to this line at x lays along w and therefore an arc
element of this line is given by the equation
dx = cxw

(6.6.10)

where a is a nonzero scalar. The corresponding arc dx in the current configuration is gotten by the use of (5.3.1) as
fx = aFw

(6.6.11)

By virtue of the vorticity equation (6.6.9), equation (6.6.11) becomes


dx = (ajyw

(6.6.12)

This equation shows that dx is the arc element of the vortex line at x (in the

282

MOTION

current configuration). This means that a material arc tangential to a vortex


line in the initial configuration remains tangential to the vortex line in the
current configuration as well. Hence, the vortex lines are transported with
the motion. This completes the proof of the theorem.

EXAMPLE 6-6.1

For the flow defined by the velocity components


lot

la

vl = - - 7 2 * 2 ,

v2 = * i >

^3 = 0

(6.6.13)

where a, a and b are constants, find the circulation round the ellipse
(x\/a2) + (xl/b2) = 1.
Solution For the given ellipse, the parametric equations are xx = a cos 0,
x2 = b sin 0, 0 < < In. Hence, if C denotes this ellipse, we obtain

lc

-v

y-

dx

dxx + v2 dx2)

2 2 2 2
2
2
(a sin 0 + b cos )
- ab) 0
=

EXAMPLE 6.6-2
only if

!
b2 )
~b < +

Show that a motion is circulation preserving if and


dw
= curl(v x w)

Solution

(6.6.14)

From equation (6.6.2), we get


D f
DIr
= - wndS
Dt
Dt]s

(6.6.15)

Using (6.5.22) and noting that div w = 0, (6.6.15) becomes


DIC
Dt

f (by/
\ [dt+

c u r l ( w x y

n'

n d s

< 6 6 16 ^

From this expression it is evident that Ic is constant for every C if and only
if the integrand on the righthand side is 0 for every S. In other words, the
motion is circulation preserving if and only if
- ^ + curl(wx v) = 0
which is (6.6.14).

(6.6.17)

6.6

CIRCULATION AND VORTICITY

283

Note: From (6.6.15) and (6.6.16), we find that (6.6.17) is a necessary and
sufficient condition for the flux of w across every material surface is constant. This result is known as Helmholtz's third vorticity theorem. Observe
that the condition (6.6.17) is just a particular case of the Zorawski's
criterion (6.5.23).
EXAMPLE 6.6.3 If ds0 and ds are the arc lengths of a vortex line in the
initial and current configurations of a circulation-preserving motion, show
that
Iwl
(6.6.18)
Deduce that
(i)

(6.6.19)

(ii)

(6.6.20)

Solution

From expression (6.6.10) and (6.6.12), we obtain

Eliminating |a| from these relations, we get (6.6.18).


From (6.6.18), we get

Dt \J\w\J

Dt \Jw0|.

because both ds0 and |w| correspond to the initial configuration. The
relation (6.6.19) is thus proven.
Carrying out the differentiation in (6.6.19) we obtain
1 D

l DJ

I D . .

which, by use of (6.2.51), reduces to (6.6.20).

Note: Expressions (6.6.18) and (6.6.20) serve as formulas for the stretch
and stretching of vortex lines in circulation-preserving motions. The stretch
is completely determined by the current and initial vorticity vectors,
whereas the stretching is influenced by velocity in addition to vorticity. In
an isochoric motion, the stretch of a vortex line is the ratio of the magnitude
of the current vorticity to that of the initial vorticity, and its stretching
is equal to ratio of the material rate of change of the magnitude of vorticity
to the magnitude of vorticity, both evaluated in the current configuration.

284

MOTION

6-7
EXERCISES
1. Find the velocity and acceleration in both material and spatial forms for the
motions given by the following sets of equations:
(i) xx = xx (1 + a2t2),

x2 = jcf,

+ x2t + x3t2,

(ii) xx=xx

x3 = x3

x2 = x2+ x*t + xxt2,

x3 = 4 + xxt + x\t2

(iii) xx = xxe-* - x3(\ - e~%

x2 = x2 - x3(e' - e~%

(iv) xx = xx cos at - x2 sin at,

x2 = xx sin at + x2 cos at,

x3 = 3~
x3 = x3,

(a is a constant)
2. For a certain motion, the velocity is given by v = x/(l + /). Show that
x = x(l + t). Find the velocity in the material form and the acceleration in both
material and spatial forms.
3 . In a certain motion, the velocity is given by
vx = -k(x\ + xxxl)e~a\

v2 = k(x\x2 + xl)e~a\

v3 = 0

(k and a are positive constants). Find the acceleration of the particle currently
located at the point (0, 1, 0).
4. A motion for which the velocity is of the form v = a(a b)x, where a is a
constant and a and b are fixed unit vectors, is called a shear motion. For this
motion, show that
~Dt=Tt
5. For the motion given by the equations
xx = f(t)xl

x2 = g(t)xQ2,

x3 = h(t)x3

where/(/), g(t) and h(t) are positive valued functions of t, find v and Dv/Dt in both
material and spatial forms. I f / ( / ) = eat, g(t) = h(t) = 1, where a is a nonzero
constant, show that
= 0
dt

but

?0
Dt

6. Show that the motion defined by the velocity components vx = -ax2, v2 = axx,
v3 = /?, where a and are nonzero constants, is isochoric. Find the acceleration.
7. Show that the motion defined by the velocity components
"i = ^ f *

v2 =

(Xl

~Rl)X',

is irrotational. Find the acceleration.

3 = >

wh

ere R2 = x\ + x\ * 0

6.7

EXERCISES

285

8. Show that the plane motion defined by the velocity components


^ = V,2>

2 = - ,,

v3 = 0

where = (, 2) is a harmonic function, is irrotational. Find the acceleration


components.
9. If in an isochoric motion, w = constant, show that V2v = 0.
1 0 . For an isochoric and irrotational motion, show that V 2 |v| 2 > 0.
1 1 . Find J for the motion defined by the velocity components
v2 = x3(e' + *?"'),

1=-^9
1 2 . If dv/dt = 0, show that

tfj/Dt

v3 = 0

= 7div((div v)v).

1 3 . A certain motion is defined by the equations


x

*? + Ltx\y

X 2 = X 2,

3 = x3

where r is a constant. Find the rate of change of the function = x\ - x\ at the


current position of the particle initially located at the origin.
1 4 . At a point x in the current configuration of a continuum the temperature is
given by T = -e~t(xl + 2x2 - 3*3). Also, the motion is defined by the velocity
components = xxe~\ v2 = -x2e~\ v3 = Jt3e~'. Find the material rate of change
of T.
1 5 . The temperature at a point x in a region filled by a material body is given by
T= T0e-at(sm xjicos

x2)

where T0, a and are constants. Also, the motion of the body is defined by the
equations
xx = x\

+ x2e-\

x2 = x 2 V + x\e\

x3 = x3

Find the rate of change of T at the point (1,1,1).


1 6 . Suppose that an observer O moves with velocity v = ex - 2e2 + e 3 . Find the
rate of change of the function = (x\ - x\)e~t as seen by O, when moving past the
point (1,0, 1).
1 7 . For the velocity fields given in Exercises 2 and 3 find
(i) The stretching tensor.
(ii) The spin tensor and the vorticity vector.
(iii) The invariants of the stretching tensor.
(iv) The deviator of the stretching tensor.
1 8 . For the velocity fields given in Exercises 2 and 3 find
(i) The stretching of a material arc currently located in the direction (1:1:1).
(ii) The shearing between material arcs currently located in the directions (1:1:1)
and(l:0:-l).
(iii) The maximum stretchings.

286
19.

6 MOTION
For the motion defined by the velocity components
Vx = / ( / ? , 0*2 ,

2 = -f(R, 0*1 ,

V, = 0

where R = x\ + x\, find the principal stretchings, the principal directions of


stretching and the angular velocity vector.
20. By finding the most general integral of the equations vtJ + vJti = 0, show that
the stretching tensor is 0 if and only if the motion is rigid.
2 1 . Show that the motion defined by the velocity v = c x x, where c is a constant
vector, is a rigid motion.
22.

Verify that the motion defined by the velocity components


= 3 x 3 ,

v2 = - 4 x 3 ,

v3 = 4x2 - 3xx

is a rigid rotation. Find the spin tensor for the motion.


23.

For a rigid motion show that

(i)div

@)=4 |w|2

/D\\
(ii) curl

\DtJ

Dvf
=

(iii) W = - ^ Q

Dt

24. Check whether there exists a velocity field for which the tensor whose matrix
follows is the stretching tensor:
*2 +
\ X\ -

X2

\
\

JX\

2
3

3*1
X\

X2 I

If so, find the velocity field.


25. For the shear motion defined in Exercise 4 with the angle between a and b
equal to /2, find the principal stretchings and the principal directions of stretchings.
26. Show that the stretching tensor D corresponds to a shear motion if and only if
D is independent of x and ID = IIID = 0.
27.

Show that a motion is uniaxial if and only if IID = IIID = 0.

28. Find a necessary and sufficient condition on D for a motion to be a plane


motion.
29. Show that at each particle of a continuum there exists at least one material arc
that is instantaneously stationary. If the velocity is of potential kind, show that there
are at least three such material arcs, which are mutually orthogonal.

6.7

EXERCISES

287

3 0 . Show that
curl(Vv)v = (Vv)w - (Vw)v - (div v)w
3 1 . Show that
D D = (tr D) 2 - 2IID = Vv (Vv) r + | w w
Deduce that IID < 0 for an isochoric motion.
3 2 . If all the principal stretchings are equal, show that
(detD) 2 = (}// D ) 3 = ( | t r D ) 6
3 3 . Show that
. D\
D
1.
div = (tr D) + (tr D) 2 - 2//uD - - w 2
Dt
Dt
2
3 4 . If Dv/Dt = , show that satisfies the equations:
ZXdivv)
1.
(0 V 2 0 = KDt ; - - l w | 2 + D - D
(ii) V2( + -v2 1 = (div v) - div(v x w)
3 5 . Show that
D

fv

dx

/</x

3 6 . Express the eigenvectors of W in terms of w.


3 7 . For a plane isochoric motion in which acceleration is the gradient of a
potential, show that
Dt

= 0

3 8 . The tensor D is defined by


D =

DO

+ D(Vv) + (Vv) r D

Show that
D2
Dt2

dx) = 2dx D dx

<(0

= K ( V v )} + V(Vv)2
Dt
'

3 9 . Show that

Deduce that

V^F
=DtS
\Dt

288

MOTION

4 0 . For any vector p, show that


Dp

= Dp + Wp - (p Dp)p

Deduce that if p is an eigenvector of D, then


Dp
1
- = - w x p
Dt
2
4 1 . Prove the following:
(i)
(ii)
(iii)
(iv)
4 2 . Prove the following:
(i)
(ii)
(iii) skw V(D\/Dt)
is constant.

= 0 if and only if the flux of w through each material surface

4 3 . Find the path lines and stream lines for the motion defined by the following
velocity fields:
(i) = ax2,
v2 = -axlt
v3 = ,
(a, are constants).
(ii) vx = axl9

v2 = -oa2,

v3 = R
2

(a, are constants and R = JC? + x\)


(iii) vx = ,

v 2 = x 2,

xl
(iv) = -^^
,

v2 = -~rr~^

(v) , =

Xl

1+ t

v3 = 0
v3 = 0

6.7

EXERCISES

289

44. For the motion considered in Exercise 3, show that the stream lines are circles.
4 5 . Find the stream lines for the motion defined by the equations
x2 = x\ea\

x3 = x\eat

where a is a constant.
46. For the motion defined by the equation x = x + ya/ 2 a, where a is a real constant and a is a fixed unit vector, show that both the path lines and the stream lines
are straight lines and that the motion is not steady.
47. For the motion defined by the equations
xl = a + -e~bk sin k(a + ct)
K

x2 =b - -e~bk cos k(a + ct)


K

x3 = constant
where , b, c and k are constants, show that the path lines are circles. Show that
(D/Dt)\\\ = 0.
48. Show that the motion defined by the velocity components

where a is a constant and R2 = x\ + x\ ^ 0 is both isochoric and ir rotational. Show


further that the stream lines are circles and that = atan~l(x1/x2) is the velocity
potential.
49. Show that in a motion where v/|v| is independent of /, the stream lines and the
path lines are coincident.
50. Show that for the motion considered in Exercise 48, the vortex lines are
straight lines. Find their equations.
5 1 . Find the stream lines and the vortex lines for the motion given by the velocity
components

where a. is a constant and R2 = x2 + x\.


52. A flow for which curl = , where is a scalar field and is called a Beltrami
flow. Show that for such a flow, the stream tubes (lines) and vortex tubes (lines)
are coincident.

290

MOTION

53. Let C be a closed curve on a vortex tube not encircling the tube. Show that
v dx = 0
54. If the cross-sectional area of a vortex tube is small, show that the product of
the vorticity and the cross-sectional area is constant.
55. Let/ be a scalar function defined in the current configuration. For a material
curve C, show that

E-=L(:

%+/***)*

t being, as usual, the unit tangent to C.

56. Let f be a vector function defined in the current configuration. For a regular
material surface S bounded by a curve C, show that

Dt

f-ntfS=

+ (divv)f -nfS+

xtds

57. Show that


\ Y/dV=\

nx(vxw)tfS+

( n x ) dS

Hence deduce that


w dV =
n x + (w n)v I dS
Dt]v
)s{
Dt
')
58. Let H be a tensor function defined in the current configuration. For a material
surface S, show that
|M

HntfS=

i ^ + (divv)H-H(W)r(n^

59. Let f be a vector function defined in the current configuration. For a material
volume bounded by a regular closed surface S, show that

-1l"""\M*"n)""'-\^'"/+b''-)"is

Dt

60. Show that

(2W + (div v))v dV= \ \ (v n)v - )- v2n [ dS


Deduce that for an isochoric motion with v = 0 on 5,
Wv dV = 0
v

6.7

EXERCISES

291

6 1 . For the motion defined by the velocity components


V

* =

o*i + xi
^2

>

2 =

*\ + oa2
2

>

^3 = 0

where R2 = x\ + x\ 5* 0 and a and are constants, find the circulation round the
circle x\ + x\ = c2.
62. For the motion described by the velocity components
ax2
"i = - ^ r >

2 =

ocX\
?>

*=

where a is a constant and R2 = x\ + x\ ^ 0, show that the circulation round a


circuit is 0 or 2rck according as the circuit encloses the origin or not.
6 3 . Show that a motion defined by the equation v = Ax, where A is a constant
tensor, is circulation preserving if and only if (Vv)w = (div v)w.
64. Let Cx and C2 be two circuits drawn on a vortex tube, each encircling the tube
in the same direction. Show that the circulation round Q is equal to circulation
round C2. What does this circulation represent?
6 5 . Show that, in a circulation-preserving motion, the vortex tubes move with the
continuum.
66. Derive the expression
/D\\
Dvi
curll = - + (div v)w - (w V)v
\DtJ
Dt
by computing DIc/Dt from (6.6.2) with the aid of the transport formula (6.5.21),
and comparing the resulting expression with (6.6.6).

CHAPTER 7

STRESS

7.1
INTRODUCTION
In Chapters 5 and 6, some kinematical aspects concerned with the deformation and motion of a continuum were discussed. Deformation and motion
are generally caused by external forces that give rise to interactions between
neighboring portions in the interior parts of a continuum. Such interactions
are studied through the concept of stress. This chapter deals with the theory
of stress.

7.2

BODY FORCES AND SURFACE FORCES


In continuum mechanics, two distinct types of forces are considered: body
forces and surface forces. Body forces are forces that act on every element
of a material and hence on the entire volume of the material. Gravitational
force is an example of a body force. We postulate that the total body force
acting on a material body (B occupying a configuration B of volume V, at
293

294

7 STRESS

time /, is expressible in the form


) = \ pbdV

(7.2.1)

where p = p(\, t) is the density at a point x of B and at time / and b is a


vector with the physical dimension force per unit of mass. The vector b is
referred to as the body force per unit of mass or just body force. Like p, b
generally varies with x and t.
Surface forces (or contact forces) act on the surface of a material. This
surface may be either a part or the whole of the boundary surface, if any,
of the material or an imaginary surface visualized in the interior of the
material. In the former case, surface forces are external forces that act on
the boundary surface of the material. Wind forces and forces exerted by a
liquid on a solid immersed in it are examples of such surface forces. In the
latter case, surface forces are internal forces that arise from the action of
one part of the material upon an adjacent part across the surface. For
example, if we consider a heavy rod suspended vertically and visualize a
horizontal cross section separating the rod into upper and lower parts, then
the weight of the lower part of the rod acts as a surface force on the upper
part across the cross section.
We postulate that the total surface force acting on an oriented surface S
in the configuration B of a material body (B is expressible in the form
f(s) = \ sdS

(7.2.2)

where s is a vector with physical dimension force per unit of area. Since s is
the integrand on the righthand side of (7.2.2), s is a function of x, in general,
where x is a point of S. It may vary with time t as well. The vector s is called
the surf ace force per unit of area ofS, or the stress vector or traction on S.
For a chosen point x (in the current configuration of a material), let us
consider an oriented surface element dS containing x; see Figure 7.1. By
virtue of (7.2.2), the total surface force acting on this surface element is
(dS)s . If we consider another oriented surface element, say dSi, containing
the same point x, then the total surface force on this surface element is
(dSi)s . Obviously, if dS and dSl have different areas, then the total surface
forces acting on them are different. But if dS and dS{ have the same areas
but not necessarily the same orientation, then we cannot decisively say
whether the total surface forces acting on them are the same or different,
unless some definite postulate is made. The postulate made for this purpose
is as follows: the stress vector s depends also on the orientation of the
surface element upon which it acts; that is, s is in general a function not

7.2

BODY FORCES AND SURFACE FORCES

295

only of x and t, but also of n, where n is the unit normal to the surface
element characterizing the orientation of the element: s = s(x, /, n). This
postulate is by Cauchy and known as Cauchy's stress postulate. From this
postulate, it follows that if dS and dSi have the same areas and the same
orientation, then the total surface forces acting on them are the same and
that, if dS and rfSj have different orientations, the total surface forces
acting on them are different.
When dS is taken on the boundary surface, if any, of a material, it is conventional to choose the unit normal n to dS as the exterior normal to the
boundary surface. Then the stress vector s = s(x, /, n) is interpreted as the
external surface force per unit of area on the boundary at time / and is
usually referred to as surface traction.
When dS is considered in the interior of a material, n has two possible
directions that are opposite to each other; we choose one of these directions.
For a chosen n, the stress vector s = s(x, t, n) is interpreted as the internal
surface force per unit of area acting on dS at time / due to the action of that
part of the material into which n is directed upon the adjacent part across
dS. Consequently, s(x, t, -n) is the internal surface force per unit area
acting on dS, at the same instant /, due to the action of that part of the
material exterior to which n is directed upon the adjacent part across dS.
Motivated by the Newton's third law of motion, we postulate that
s(x, t, -n) is equal and opposite to s(x, t, n); that is,
s(x, t, -n) = -s(x, t, n)

(7.2.3)

For simplicity in the notation, we write s(n) for s(x, /, n). Then the
relation (7.2.3) can be rewritten as
s(-n) = -s(n)

(7.2.4)

This relation is known as the Cauchy's reciprocal relation; see Figure 7.2.
While interpreting the stress vector acting on an internal surface element
of a material, we have made the tacit assumption that the portions of a
material lying on the two sides of a surface element interact with each other
to produce a force across the surface. This assumption is well accepted in

Figure 7.1. Stress vector.

296

7 STRESS

Figure 7.2. Cauchy's reciprocal rotation.

the theory of conventional solids and fluids. However, apart from a force,
a couple may also exist across a surface. The inclusion of the effects of such
couples in the theory gives rise to the concept of what is called couple stress.
The effects of couple stresses are of interest only in special kinds of
materials, called polar materials; these are not treated in this text.
The resultant force f(r) acting on a material body (B in a configuration B
of volume V with boundary surface S is defined as the vector sum of the
total body force acting on V and the total surface force acting on S; that is,
f(r) =

f<*)

f<*)

pbdV+\

sdS

(7.2.5)

Here, it is postulated that the internal forces across surfaces lying in the
interior of V balance each other so that their resultant is zero.

7.3
STRESS COMPONENTS
Suppose we consider the particular case in which the surface element dS has
ej as the unit normal and denote the stress vector acting on this element by
s(1); that is, s(1) = s(ej). Resolving this vector along the coordinate axes, we
get an expression of the form
s(1) = rnex + r 12 e 2 + T13e3
= rlkek

(7.3.1)
(1)

where rik are the components of s along the axes.


The stress vectors s(2) and s(3) are defined in an analogous way, and the
following expressions are obtained:
s(2) = r2*e*,

s(3) = T3iteJt

(7.3.2)

7.3

STRESS COMPONENTS

297

Expressions (7.3.1) and (7.3.2) can be put in the following condensed


form:
s Tikek
(7.3.3)
Taking dot product with e, on both sides of (7.3.3) and noting that
ek e, = kj and rikkj = 9 we get
s(/) e, = ,,

(7.3.4)

Thus, for given / andy, ^ represents theyth component of s(/), where s(/) is
the stress vector acting on a surface element having e, as the unit normal.
For example, 23 represents the third component of the stress vector s(2)
(acting on a surface element having e2 as the unit normal). The nine real
numbers # defined in this way are called the stress components at the point
x and time t, in the JC, system. These components have dimensions: force/
(length)2. The matrix with as elements, namely,
'12
T21

22

'23

'31

^32

'33

(7.3.5)

is called the stress matrix at the point x and time t> in the system. The
three stress vectors s(l) and the nine stress components ^ can be displayed
pictorially as in Figure 7.3.

**3

Figure 7.3. Stress vectors and components.

298

7 STRESS

7.4
STRESS TENSOR
At a point P(x) in the current configuration of a continuum, consider a
"small" tetrahedron with three of its faces perpendicular to the coordinate
axes and the fourth one perpendicular to an arbitrary unit vector n (see
Figure 7.4). Let AA j , AA2 and AA3 be the areas of the faces PQR, PSR and
PQS perpendicular to the xl, x2 and x3 axes, respectively, and A A be the
area of the slant face QRS perpendicular to n. Then
AAX = (AA) cos (n, xx) = (AA)nl

(7.4.1a)

AA2 = (AA)n2

(7.4.1b)

AA3 = (AA)n3

(7.4.1c)

Figure 7.4. Stresses on the face of a tetrahedron.

7.4

STRESS TENSOR

299

where cos(n, x{) denotes the cosine of the angle between n and the positive
xx axis and are components of n. Also, the volume of the tetrahedron is
(7.4. Id)

AV = \h(AA)

where h is the perpendicular distance of the point P from the slant face
QRS.
We note that the plane element PQR is a part of the boundary surface of
the material contained in the tetrahedron. As such, the unit normal to PQR,
which is to be the exterior normal by convention, is -zx. The total surface
force on this plane element is (AAx)[s(-zx)}. In view of the reciprocal
relation (7.2.4), this surface force is equal to -(A^Msie^} or
-(^(\
Similarly, the total surface forces on the plane elements PSR and PSQ are
-(AA2)s(2) a n d - (A>l3)s (3) , respectively. Since n is the exterior normal to
the tetrahedron on the plane element QRS, the total surface force on this
plane element is (Av4)s(n). It is to be noted that while s (1) , s(2) and s(3) are
stress vectors on plane elements passing through the point P, s(n) is the
stress vector on a plane element that does not pass through P.
By virtue of the expression (7.2.5), the resultant force acting on the
tetrahedron is given by
f(r) = (AV)pb + (AA)s(n) - (AAx)s(1) - (A,42)s(2) - (A^ 3 )s (3)

(7.4.2)

where p is the density and b is the body force both evaluated at the point P.
Let Am = p(AV) be the mass of the material contained in the
tetrahedron and a be the acceleration evaluated at P. Motivated by the
Newton's second law of motion, we postulate that
/?(AF)a = f(r)

(7.4.3)

Then, on taking note of (7.4.1) and (7.4.2), we get the equation


s(n) = n{sil) + n2s(2) + n3s0) + j/>A(a - b)
Now, with n and P fixed, let us take the limit as h - 0. In this limit, the
slant face QRS tends to a plane element normal to n at P. Then s(n) represents the stress vector at P on this element. Also, s (l) are evaluated at P.
Thus, in this limit,
s(n) = /i 1 s (1) + >z2s(2) + AZ3S(3)

(7.4.4)

This linear relation enables us to find s(n) for an arbitrary n, when the
stress vectors s ( 0 , / = 1, 2, 3, are known.
Substitution for s (/) from (7.3.1) and (7.3.2) in (7.4.4) gives
s(n) = (,)/ + (Tvej)n2 +
= Tkjnktj

(^)3
(7.4.5)

300

7 STRESS

Taking the scalar product with ef on both sides of this expression, we get
(7.4.6)

^(n) = xklnk

These relations connecting ^(n) and TU show that, for any given n, 5f(n) can
be determined when the stress components 7 are known.
Recall that the nine components } are defined w.r.t. a set of axes.
Relations (7.4.6) show that for an arbitrary vector n with components ni9
xkink are components of a vector, namely, s(n). From the quotient law
proved in Example 2.8.6, it follows that ,-, are components of a secondorder tensor, called Cauchy's stress tensor, denoted T. Then, by use of
(2.7.6), we have
T = ,-,-e, tj

(7.4.7)

The relations (7.4.6) can now be expressed in the direct tensor notation as
follows:
(7.4.8)
s(n) = T r n
This relation connecting the stress vector s(n) and the stress tensor T is
known as Cauchy's law (hypothesis). The law can be expressed in terms of
T (instead of T 7 ) by interchanging the meanings of xu and . This is just
a matter of notation and convention.
We note that, through a given point, there exist infinitely many surface
elements. On every one of these elements we can define a stress vector. The
totality of all these stress vectors is called the state of stress at the point. The
relation (7.4.8) enables us to find the stress vector on any surface element at
a point by knowing the stress tensor at the point. As such, the state of stress
at a point is completely determined by the stress tensor at the point.
For computational purposes, it is convenient to have the equation (7.4.8)
rewritten in the matrix notation. The resulting matrix equation has the
explicit form
3
s2

22

'32

23

(7.4.9)

33

where we have dropped n from the symbols ^(n) for simplicity in the
notation.
It has to be pointed out that Cauchy's law (7.4.8), which asserts the
existence of the stress tensor, is a direct consequence of the postulates
(7.2.4) and (7.4.3). These two postulates are just two particular cases of a
more general and fundamental postulate called the law of balance of linear
momentum. This fundamental law will be introduced in Section 8.3.

7.4

STRESS TENSOR

301

The stress matrix at a point P in a material is given as

EXAMPLE 7.4.1
follows:

2 -5

[T(/l =

-5

Find (i) the stress vector on a plane element p through P and parallel to the
plane 2xx + x2 - x3 = 1, (ii) the magnitude of the stress vector and (iii) the
angle that the stress vector makes with the normal to the plane.
Solution The plane element p on which the stress vector is required
is parallel to the plane 2xx + x2 - x3 = 1. Hence, the direction ratios of
the normal to the plane are (2,1,-1) and the direction cosines are
(2/V6, 1/V6, -1/V6). Thus, the components nt of the unit normal to the
plane element p are

V6'

nr = V6'

n* = - V6

(7.4.10)

Substituting for xi} from the given stress matrix and using (7.4.10) in
Cauchy's law (7.4.9), we obtain
"*l"
*2

-*3.

"3
= 1

1 4 ] 2/V6"
2-5
1/V6

4-5

oj [-1/V6

This gives
st = V372,

s2 = 3V3/7,

s3 = V372

(7.4.11)

Thus, the required stress vector is


s = V3/2(e[ + 3e2 + e3)

(7.4.12)

|s| = V3372

(7.4.13)

and its magnitude is


From (7.4.10) and (7.4.12), we get
sn =2

(7.4.14)

Hence the angle between the directions of s and n is given by


cos = = V8/33
ls||n|

(7.4.15)

302

7 STRESS

Note: Expression (7.4.15) illustrates the important fact that the stress
vector on a surface element need not be along the normal to the surface.
The stress matrix at a point P(Xi) in a material is given

EXAMPLE 7.4-2
as follows:

3-*l
x*x

*3

-X2

-x2

[</] =

Find the stress vector at the point ( l , 0, - 1 ) on the surface x\ + x\ = xx.


Solution The stress vector is required on the surface f(xx, 2, 3) =
xx - x\ - x\ = 0. We find that
|V/| = (1 + Ax\ + 4xj)1/2

V/ = e, - 2x2e2 - 2x 3 e 3 ;

(7.4.16)

At the point Q(l, 0, - 1 ) , we get Vf = el + 2e 3 , |V/| = vT. Hence the unit


outward normal to the surface / = 0 at the point Q is
n = 7=T7 = - 7 yy( e ! + 2 e 3 )

|V/|

V5

so that
n

\ = -7j>

2 = 0,

/2,

V5

(7.4.17)

At the point , the given stress matrix is


-1

1 0

[T(/] = 1 0
0

0
0

(7.4.18)

Substituting for nt and 7 from (7.4.17) and (7.4.18) in Cauchy's law (7.4.9)
and equating the corresponding elements, we get sl = -1/V5, s2 = 1/V5,
53 = 0. Hence, the required stress vector is
s = ~ ^ j ( e i - e2>

EXAMPLE 7-4.3 The state of stress at a point x is said to be uniaxial if


s(n) is along a constant unit vector a for all n. Show that such a state of
stress arises if the stress tensor at x is of the form T = ( (g) a), where T
is a scalar. Interpret T.

7.4

Solution

STRESS TENSOR

303

Suppose that
T = T(aa)

(7.4.19)

where a is a constant unit vector and T is a scalar. Then for any vector n,
= T(a a)n = T(a n)a
Using Cauchy's law (7.4.8), this becomes
s(n) = T( a n)a

(7.4.20)

Evidently, s(n) is along a and the state of stress at a point x is uniaxial.


From (7.4.20), we get
s(a) a = T(a a)(a a) = T

(7.4.21)

Since s(a) is the stress vector on a surface element with a as the unit normal,
we find from (7.4.21) that T represents the component along a of the stress
vector that acts on a plane perpendicular to a.

EXAMPLE 7 . 4 . 4 Show that s(n) = - p n for all n if and only if


Xij = -pij. (Such a state of stress occurs in fluids at rest and is called
hydrostatic stress.)
Solution

First suppose that


s(n) = -pn

for all n. Then s

(l)

(7.4.22)

= s(ef) = -pei9 and expression (7.3.4) yields


TU = s{i) e, = -pet e, = -}.

(7.4.23)

Conversely, suppose that TU are given by (7.4.23). Then T = -/?I, and by


using Cauchy's law we get (7.4.22).
EXAMPLE 7 . 4 . 5 Let s(n) and s(n') be stress vectors on two surface
elements at a point x. Prove that
n s(n') = n' s(n)

(7.4.24)

if and only if the stress tensor at x is symmetric.


Solution

By Cauchy's law (7.4.8), we obtain


n' s(n) = n' (T r n);

n s(n') = n ( T V ) .

(7.4.25)

Further, by (2.8.14) we note that


( ' ) = n' (Tn).

(7.4.26)

From (7.4.25) and (7.4.26), we find that (7.4.24) holds if and only if
T r = T.

304

7 STRESS

7.5
NORMAL AND SHEAR STRESSES
Example 7.4.1 illustrated that the stress vector s(n) need not be collinear
with n. Hence, s(n) may be resolved along n and perpendicular to n, in
general. Let
= s(n)n
(7.5.1)
and
T = s(n) t
(7.5.2)
where t is a unit vector perpendicular to n. Then is called the normal stress
and is called the shear stress along t, on the surface element considered.
The normal stress is said to be tensile if cr > 0 and compresse if < 0. If
a = 0, s(n) acts tangential to the surface and is called a pure tangential
stress or pure shear stress. When s(n) acts along or opposite to n, the shear
stress on the surface is 0 and s(n) is (then) called pure normal stress. The
pure normal stress acting opposite to n is called pressure.
In the particular case when n = e h expressions (7.5.1) and (7.3.4) yield
= s(1) ej =
Thus, is the normal stress on the surface element perpendicular to ex.
Similarly, 22 and 33 are normal stresses on surface elements perpendicular
to e2 and e3 respectively. The stress components , 22 and 33 are
therefore called normal stresses. Note that these stresses are the diagonal
elements of the stress matrix.
Suppose we take n = ex and t = e2. Then expressions (7.5.2) and (7.3.4)
yield
(1)
e2 = 12
= s

Figure 7.5. Normal and shear stresses.

7.5

NORMAL AND SHEAR STRESSES

305

Thus, 12 is the shear stress along e2 on the surface element perpendicular to


e j . Similar interpretation is given for 2 1 , 3 1 , 1 3 , 23 and 3 2 . The stress
components ru, i 5* j are therefore called the shear stresses. Note that these
components are the nondiagonal elements of the stress matrix.
Using Cauchy's law (7.4.6), expression (7.5.1) yields
= ^(, = Tkinkni
(7.5.3)

= riknink

on interchanging the dummy suffixes. Expression (7.5.3) enables us to


determine the normal stress directly from the stress components ^.
If t is chosen in the plane containing n and s(n), see Figure 7.5; then we
have, in view of (7.5.1) and (7.5.2),
This gives
so that

s(n) = n +

(7.5.4)

|s(n)| 2 = 2 + 2

(7.5.5)

= |[|s(n)| 2 - 2)! 1/2

(7.5.6)

The expression in the righthand side of (7.5.6) is called the magnitude of the
shear stress, or just the shear stress.
EXAMPLE 7.5.1
given by

The stress matrix at a certain point in a material is


'3
[*</] =

1 1
0

Find the normal stress and the shear stress on the octahedral plane element
through the point. (Note: Octahedral plane is the plane whose normal
makes equal angles with positive directions of the coordinate axes.)
Solution For the octahedral plane element, the normal n has direction
ratios (1,1,1). Hence the components of n are
nl = n2 = n3 =

(7.5.7)

73

Substituting for /7 from the given matrix and for AI, from (7.5.7) in
Cauchy's law (7.4.9), and equating the corresponding terms, we get
= -7J ,

*2 = V3 >

53 = V3

(7.5.8)

306

7 STRESS

as the components of the stress vector. The magnitude of this stress vector is
|s(n)| = V4373

(7.5.9)

From (7.5.7) and (7.5.8), we get the normal stress acting on the given
plane element as
a = Sini = 11/3
(7.5.10)
Since > 0, the normal stress on the plane is tensile.
Substituting for |s(n)| 2 and 2 from (7.5.9) and (7.5.10) in the expression
(7.5.6), we obtain the shear stress acting on the given plane element as
= 2V2/3 .

EXAMPLE 7.5.2 For the state of stress considered in Example 7.4.3,


compute the normal stress and shear stress on an arbitrary plane element.
Find the planes on which these stresses are maximum. Also, find the
maximum values.
Solution For the state of stress considered, the stress vector on an
arbitrary plane element with normal n is given by expression (7.4.20). If 0
is the angle that n makes with a, this expression can be rewritten as
so that

s(n) = T c o s a

(7.5.11)

|s(n)| = | 7 l | c o s 0 |

(7.5.12)

Hence the normal stress on the plane element considered is


= n-s(n) = 7cos 2 0

(7.5.13)

Substituting for and |s(n)| from (7.5.13) and (7.5.12) in (7.5.6), we get
the shear stress on the plane element as
= |7 2 cos 2 0 - r 2 c o s 4 0 | 1 / 2 = | r | | s i n 2 0 |

(7.5.14)

From (7.5.13) we see that is maximum for 0 = 0; this means that


maximum normal stress occurs on a plane element perpendicular to a. Also,
"max = T. From (7.5.14) we see that is maximum for 0 = /4; this means
that the maximum shear stress occurs on a plane inclined at an angle of 45
with a. Also, r max = \\T\.

7.6
PRINCIPAL STRESSES
It has been pointed out that the stress vector on a surface element need not
be collinear with the normal to the surface element. We now proceed to
investigate whether there exists any surface element at a given point on
which the stress vector is collinear with the normal to the element.

7.6

PRINCIPAL STRESSES

307

We note that the stress vector s(n) is collinear with n if and only if
s(n) = An

(7.6.1)

for some scalar . Using Cauchy's law (7.4.8), the condition (7.6.1)
becomes
(7.6.2)
T r n = An
The condition (7.6.2) holds if and only if n is an eigenvector of T r and A is
the associated eigenvalue. Since every tensor has at least one eigenvector
(Theorem 2.13.1), there exists at least one n such that the condition (7.6.2)
holds. Thus, at any given point there does exist at least one surface element
on which the stress vector is collinear with the normal. Such a surface
element is called a principal plane of stress at the point. It follows that for
a principal plane of stress, condition (7.6.1) holds and any surface element
for which the condition (7.6.1) holds is a principal plane of stress. (Consequently, there is no shear stress on a principal plane of stress.) The direction
of the normal to a principal plane of stress is called a principal direction of
stress. The normal stress on a principal plane of stress is called a principal
stress. From expressions (7.6.1) and (7.6.2), we find that a principal stress
is nothing but an eigenvalue of the tensor and a principal direction of
stress is nothing but the direction of an eigenvector of .
In Section 2.13, it was shown that for a symmetric tensor there exist
exactly three (not necessarily distinct) eigenvalues and at least three
mutually orthogonal eigenvectors with the property that, if a system of axes
is chosen along the orthogonal eigenvectors, then with respect to this system
the matrix of the tensor is diagonal, the diagonal elements being the
eigenvalues of the tensor. This property can be employed to the tensor T r ,
provided it is symmetric; that is T r = T. In Section 8.4, we will indeed
prove that T = T r . Bearing this in mind, we restrict ourselves to the case
T = . Then, from what has been noted already, it follows that at every
point of a material (i) there exist exactly three (not necessarily distinct)
principal stresses, (ii) there exist at least three mutually orthogonal principal
directions of stress; consequently there exist at least three mutually
orthogonal principal planes of stress, and (iii) if a system of axes is chosen
along the mutually orthogonal principal directions of stress, then the matrix
of the stress tensor is purely diagonal, the diagonal elements being the
principal stresses. A system of coordinate axes chosen along the principal
directions of stress is referred to as principal axes of stress.
Since the principal stresses are the eigenvalues of T, these are the roots of
the characteristic equation
det(T - AI) det(Ti7 - = - A 3 + 7TA2 - 7/TA + IIIT = 0

(7.6.3)

308

7 STRESS

where , and are the fundamental invariants of the tensor T; that is


^

kk

(7.6.4)
= det(rl7)
The scalars , and are called the fundamental stress invariants.
Once the three principal stresses, say , 2 and 3 , are found by solving
the cubic (7.6.3), the corresponding principal directions of stress can be
found by solving the vector equation (7.6.2) for n in the cases: A = 9
A = 2 and = 3 . With respect to the axes chosen along the principal
directions so determined, the matrix of the stress tensor is as follows:
[/] =

Tl

T-,

(7.6.5)

Since 92,3 are roots of the cubic (7.6.3), the fundamental stress
invariants have the following expressions:
= + 2 + r3
Hj = 2 + 2 3 + T3TJ

(7.6.6)

IIIj = 23
EXAMPLE 7.6.1

The stress matrix at a point (jcf) is given by


| " 0
0

fa =\

[ -ax2

-ax2

bxx

bxx

where a is a constant. Show that the principal stresses at the point (a, b, 0)
are 0, yf2ab. Find the principal direction for which the principal stress is
yflab. Also, find the fundamental stress invariants.
Solution The principal stresses are the roots of the characteristic
equation (7.6.3). Substituting for ,-, from the given stress matrix with
= a and x2 = b in this equation we get the equation
-

-ab

ab

-ab

ab

-A

= 0

7.6

PRINCIPAL STRESSES

309

Expanding the determinant and solving the resulting equation, we get the
three roots = 0, 2 = V2 ab, 3 = -yfab. These are the principal stresses.
To find the principal direction associated with the principal stress
2 = yfab, we have to solve equation (7.6.2) for n by using the given ^
with xx = a, x2 = b and taking A = 2 = Vlab. Thus we obtain the
following equations for # :
yf2nl + n3 = 0
y/2n2 - n3 = 0
n

\ -

2 + ^n3

=0

A solution of these equations subject to the constraint


1

/,-,-

= 1 is

Thus, the principal direction associated with the principal stress yflab has
direction cosines (-j,j,
1/V2).
Substituting = 0, 2 = yfab and 3 = -yflab in expressions (7.6.6),
we get
= 0,

= -2a2b2,

IIIj = 0

These are the fundamental stress invariants for the given state of stress at
the point (a, b, 0).

EXAMPLE 7.6-2 Show that, as the orientation of a surface element at


a point P varies, the normal stress on the surface element assumes an
extreme value when the element is a principal plane of stress at P and that
this extreme value is a principal stress.
Solution

The expression for the normal stress is


= Tij^rij

(7.6.7)

see (7.5.3). We have to find /i, for which is an extremum. Since n is a unit
vector, we have the restriction nknk = 1. Hence, assumes extreme values
when rii obey the conditions
^-[a-a(nknk-

l)j = 0

(7.6.8)

where a is a Lagrangian multiplier. Substituting for from (7.6.7), the


conditions (7.6.8) become
(Jij - au)nj

=0

(7.6.9)

310

7 STRESS

These conditions are satisfied if and only if n is a principal direction of


stress. Thus, assumes an extreme value on a principal plane of stress.
For nt given by (7.6.9), expression (7.6.7) yields = a. From (7.6.9), we
note that a is a principal stress. Thus, a principal stress is an extreme value
of .
M
EXAMPLE 7.6-3 Prove that, as the orientation of a surface element at
a point P varies, the shear stress on the element assumes the maximum value
when the normal to the element bisects the angle between the directions of
the largest and the least principal stresses at P and that the maximum shear
stress is equal to one-half the difference between these two principal
stresses.
Solution It is convenient to choose the coordinate axes to coincide with
the principal directions of stress at P. Then the stress matrix has the
following diagonal form:

' , 0 0
[T(/] =

(7.6.10)

where 9 2 , 3 are the principal stresses at P. Cauchy's law (7.4.6) then


yields
sl(n) = xinl9

s2(n) = 22,

s3(n) = 33

(7.6.11)

Hence the normal stress is


= Si(n)ni = xxn\ + 2\ + x3n\

(7.6.12)

From (7.6.11), we also get


\s(n)\2 = \\ + \\ + \\

(7.6.13)

Substituting for and |s(n)| from (7.6.12) and (7.6.13) in (7.5.6) we get
= \\\ + xini + \\ - (,/if + 22 + 33)]

(7.6.14)

We have to find the maximum value of as nf- vary. Since nknk = 1, the
extreme values of 2 are found by solving the following three equations
for rii :

a
drij

{xL - *(nknk

- 1)} = 0

(7.6.15)

7.6

PRINCIPAL STRESSES

311

where a is a Lagrangian multiplier. Using (7.6.14) and carrying out the


differentiation, equations (7.6.15) give
{( - 2) - ] =
1(2 - 2)2 - ]2 = 0

(7.6.16)

{(3 - 2)3 - ]3 = 0)
We verify that the following sets of values of , obey equations (7.6.16) as
well as the constraint nknk = 1:
(i)

ni = 1 ,

n2 = 0,

n3 = 0

(ii)

nx = 0,

n2 = 1 ,

n3 = 0

(iii)

nx = 0,

n2 = 0,

(iv)

= 2 = ,

(V)

= /23 = ,

A22 = 0

(vi)

Aij = 0,

n2 = n3 =

= = 1

= 0,

Thus 2 assumes extreme values for ni given by relations (i) and (vi). The
values of 2 corresponding to nt given by (i) to (vi) may be obtained from
(7.6.14). These values are
(a) 2 = 0,

(b) 2 = 0,

(d) 2 = i ( T l - 2),

(c) 2 = 0

(e) 2 = i ( T l - 3),

(f) 2 = | ( 2 - 3)

We note that the value of 2 given by (a) to (c), namely, 2 = 0, is the


minimum value. The maximum value of 2 is one of the three values given
by (d) to (f), depending on the values of , 2 , 3 . If denotes the largest
and 3 the least of , 2 , 3 , then the maximum value of is given by (e),
the maximum value being \( - 3). Thus the maximum value of the shear
stress is equal to one-half the difference between the largest and the smallest
principal stresses. The values of nt that correspond to this maximum value
of are given by (v). We verify that these values of A, correspond to the
directions bisecting the angle between the xx and x3 axes, which are along
the principal directions of the largest and the smallest principal stresses.
Thus, the normal to the surface element on which the maximum shear stress
occurs bisects the angle between the directions of the largest and the least
principal stresses.

312

7 STRESS

EXAMPLE 7-6.4 If the coordinate axes are chosen along the principal
directions of stress at a point P, then the normal and shear stresses at P on
the octahedral plane are called the octahedral normal and shear stresses at
P. Show that these are given by
<7oct = j(Ti + *2 + 3)

(7.6.17)

Toct = l i d ! - 2 ) 2 + <2 - 3 ) 2 + ( 3 - )1\/2

(7.6.18)

where , 2 , 3 are the principal stresses at P.


Solution When the axes are along the principal directions of stress at P,
the normal and shear stresses at P are given by (7.6.12) and (7.6.14).
For the octahedral plane, nl = n2 = n3 = 1/V3. With these values of ni9
(7.6.12) becomes (7.6.17), and (7.6.14) becomes
Toct2 = i(3(T 2 + 22 + T|) - (T, + 2 + 3)2}
which can be simplified to obtain (7.6.18).

(7.6.19)

7.7
STRESS DEVIATOR
In Section 2.12, we have seen that every second-order tensor can be
decomposed into a spherical part and a deviator part. Employing this
decomposition to the stress tensor T, we get
T = }(trT)I + T(<0

(7.7.1)

= -) = - |

(7.7.2)

Setting
(7.7.1) becomes
T = -pi + T (d)

(7.7.3)
(

The number/? is called the mean pressure. The deviator part T , which is
the excess of T over the spherical part -pi, is called the stress deviator
tensor. It is evident that T( is symmetric if T is symmetric. The components $ of T (d) are called the stress deviator components and the matrix
[ ^ ] is called the stress deviator matrix.
Example 2.13.7 showed that a tensor and its deviator part have the same
eigenvectors. It therefore follows that the principal directions of T(
coincide with those of T (which is assumed to be symmetric). With respect
to a set of axes chosen along these principal directions, the stress deviator
matrix is purely diagonal, the diagonal elements being the eigenvalues of

7.7 STRESS DEVIATOR

313

T (d) . These eigenvalues, say \\ %*\ t3d), are called the principal deviator
stresses. By virtue of expressions (2.13.30), these are related to the principal
stresses r,, 2 , r3 through the following relations:
T-r = i ( 2 T , - T 2 - T 3 ) = |(2 2 - 3 - ,)
A") = i(2r3
t2).

(7.7.4)

Since the deviator part of a tensor has its trace equal to 0, we have
IT(d) = 0. By virtue of expressions (2.13.14) and (2.13.31), the second and
third fundamental invariants of T (d) are given as follows:
// T ( d ) = T\)Ti)

Ti)Ti)

ri^rif)

= -i((Ti - 2)2 + (2 - 3)2 + (3 - T^ 2 }

(7.7.5)

/// = <?\<
= ^

- 2 - 3)(22 - 3 - ^,

- 2 - 3)

(7.7.6)

The scalars // and 7//T(<o are called the stress-deviator invariants.


EXAMPLE 7.7.1

For the state of stress given by the stress matrix


1

3-1
_1 3

[Tij] =\0

find (i) the stress deviator matrix, (ii) principal deviator stresses, and (iii)
stress deviator invariants.
Solution (i) From the given stress components, we get xkk = 7. Hence,
from expression (7.7.1), we obtain the stress deviator components as
follows:
4.
3

rff = 0

Tid)

11

_
_ 1T
11
3 kk

Tid)
22

_ JLT
22
3 kk

-2..
~ 3

id) r ( d ) _ r
23 32

_
~

id)
T

Tid)

33

33

_ 1T
3 kk

2..
3

12

13

_ -(d)
31

Thus, the stress deviator matrix is


"

[rf I =

0"

i
0 -i

-i
f.

4
3

u0

314

7 STRESS

(ii) The characteristic equation of the given matrix is


1 -

3 -

-1

3 -

|=0

Expanding the determinant and solving the resulting cubic equation we get
the three roots as = 1, 2 = 2, 3 = 4. These are the principal stresses for
the given state of stress. The corresponding principal deviator stresses may
be computed by using expressions (7.7.4). Thus, we obtain

r (d)

3>

rf>

5.
3

These may also be computed directly from the characteristic equation of the
matrix [rg].
(iii) Expressions (7.7.5) and (7.7.6) now yield the stress deviator
invariants
Iljid)

2L

Illjid)

9 >

20.

27

EXAMPLE 7-7.2 Show that the octahedral shear stress roct is given in
terms of the principal deviator stresses by
T
T2

_ _l<T(d)
oct 3 lTl

(d) , id) {d) , (rf) (dh


2 + 2 3 + 3 1 i

(7.7.7)
(7.7.8)

= jlljid)

Solution

From (7.6.18), we have


<ct = mi

(7.7.9)

- r2r + (r2 - 3 + (3 - ,]

From expressions (7.7.5), we get


(Ti - 2)2 + (2 - 3)2 + (3 - )2
= -6//(.) = -6[\>?> + 2)?> + T^T}^}
Using (7.7.10) in (7.7.9), we obtain (7.7.7) and (7.7.8).

(7.7.10)

7.8
BOUNDARY CONDITION FOR THE STRESS TENSOR
It was noted that Cauchy's law (7.4.8) connecting s(n) and T holds at every
point of the material and at every instant of time. As such, it holds also at
points of the boundary surface, if any, on the material. Thus, if we take a

7.8 BOUNDARY CONDITION FOR THE STRESS TENSOR

31 5

point P on the boundary surface of the material, we should have


= s(n)

(7.8.1)

at P. For a boundary point, n is a known vector and by convention is taken


as the exterior normal, and s(n) represents the surface force per unit of area
caused by the external surface forces. Hence (7.8.1) serves as a boundary
condition for T (for known n and known s(n)).
The boundary surface of a material is said to be a stress-free (or tractionfree) surface or just a free surface if the stress vector acting on it is 0. For
such a boundary surface, the boundary condition (7.8.1) becomes
Trn = 0

(7.8.2)

EXAMPLE 7.8-1 Obtain the boundary conditions for the stress components in the case when the boundary of the surface is subjected to a
pressure /?. Specialize these conditions for a material filling the region
xx < 0 with the plane xx = 0 as the boundary.
Solution By definition, pressure is the pure normal stress acting opposite
to the normal of a surface (see Sec. 7.5). Hence, if p(>0) is the magnitude
of pressure, we have
s(n) = -pn
By using (7.8.1), this can be rewritten as
T r n = -pn

or

(7.8.3)

,,, = -pnt

(7.8.4)

Thus, if a boundary surface is acted upon by a pressure of magnitude p,


(7.8.4) are the conditions to be satisfied by the stresses xtj on the boundary,
with rti as the components of the outward unit normal.
If the material fills the space xx < 0 with xl = 0 as the boundary, the
outward normal to the boundary is n = ex ; see Figure 7.6. The conditions
(7.8.4) then yield the following conditions to be satisfied on xx = 0:
= -P,

T21 = 0,

T 31 = 0

(7.8.5)

Note: If the boundary xx = 0 is stress free, conditions (7.8.5) become


= 2 = 3 = 0 o n x\ = 0.

EXAMPLE 7.8-2 Let P be a point on a free surface 5. Show that the


stress vector on every other surface element at P acts tangential to 5.
Solution Let n be the unit normal to the surface S at the point P. Since
S is a free surface, we have s(n) = 0. Let S' be any other surface element at
P with unit normal n'. Then the stress vector on S' is s(n'). Since the stress

316

7 STRESS

xi = 0

Figure 7.6. Example 7.8.1.

tensor is assumed to be symmetric, we have from (7.4.24), n s(n') =


n' * s(n). Since s(n) = 0, it follows that s(n') is orthogonal to n. That is,
sin') is tangential to S.

7.9
PIOLA KIRCHHOFF STRESS TENSORS
In the preceding sections, we analyzed the concept of stress by considering
points and surface elements in the current configuration of the material.
The stress vector s on a surface element dS at a point in the current
configuration was defined as the surface force per unit of area of that
element. In some discussions, particularly in nonlinear elasticity, it is often
more convenient to use the idea of a surface force on dS measured per unit
of area of the corresponding element dS0 in the initial configuration. Here
we introduce tensors that will enable us to compute such a surface force.
Like the Green strain tensor, these tensors are in Lagrangian description.
If dS and dS0 are the corresponding elements in the current and initial
configurations, respectively, we find from (5.2.12) that
(dS) n = J(dS0)(FTy1n

(7.9.1)

where n and n are the unit vectors to dS and dS0, respectively; see
Figure 7.7.
By virtue of the relations (7.4.8), the total surface force acting on dS is
(dS)s = (dS)(TTn) = TT(dS)n

(7.9.2)

Substituting for (dS)n from (7.9.1), this becomes


(dS)s =

J(dS0)(-lT)Tn

(7.9.3)

7.9

PIOLA KIRCHHOFF STRESS TENSORS

31 7

Figure 7.7. Surface elements dS0 and dS.

Setting
we get

T = J

(7.9.4)

(dS)s = (rfS0)(T0)V

(7.9.5)

This expression is analogous to expression (7.9.2). It is evident that


(T)rn represents the surface force on dS measured per unit area of dS0;
we denote this vector by s. Thus,
(dS)s = (dS0)s

(7.9.6)

where
s

= (T)Tn

(7.9.7)

When the tensor T is known, expression (7.9.7) enables us to find s for


any given n. The vector s is called the nominal stress vector and the tensor
T is called the nominal stress tensor. From (7.9.4) we note that T may not
be symmetric even when T is symmetric. The transpose of T, namely,
(), is called the first Piola-Kirchhoff stress tensor, after Piola (1833) and
Kirchhoff (1852). Some authors refer to T as the Poila-Kirchhoff stress
tensor and (T)r as the nominal stress tensor.
In components, expression (7.9.7) reads
s? = > ;

(7.9.7)'

To interpret the meanings of the components of T, let us take n to be


a base vector, say, ej. Then s represents the surface force per unit of initial
area of the surface element that initially had ej as its'unit normal. The
expressions (7.9.7)' now become s,0 = ?,. Thus represents the /th component of the surface force per unit initial area of the surface element which
initially had ex as its unit normal. The components TJ, and ^ have
analogous meanings. Thus, in general, rfj represents theyth component of

318

7 STRESS

the surface force per unit of initial area on the surface element that initially
had its normal in the /th-direction.
Setting
S = T(F"y
(7.9.8)
expression (7.9.7) becomes
s = FS r n

(7.9.9)

This expression is often used as an alternative to the expression (7.9.7).


The tensor S is called the second Piola-Kirchhoff stress tensor. Unlike
the tensors T and T, the tensor S does not admit a direct physical
interpretation.
The following expressions follow from (7.9.4) and (7.9.8):
T = -FT0 = -FSFr

(7.9.10)

S = /F^TiF" )

From these expressions, we find that S is symmetric whenever T is symmetric and vice versa.

7.10
EXERCISES
1. The stress matrix at a point P is given by
1

0-2

- 1 0

Find the stress vector at P on a plane element whose normal is e2 . Also, find its
magnitude.
2. The stress matrix at a point P is given by
2 - 1 3
[7] = 1 - 1
3

0-1

Find the stress vector s at P on a plane element with normal n = 2(el + e2) + e 3 .
Also, find (i) the angle that s makes with n, and (ii) magnitude of s.

7.10

EXERCISES

319

3 . The stress matrix at a point P is given by


1

1 0

[*</] = I 1 - 1
0

0 1

Find the stress vector at P acting on the plane element parallel to the plane
\ + xi + 2*3 = 0. Also find the normal stress and the shear stress on the element.

4. The stress matrix at a point P is given by


"1 a b
a 1 c

[0] =

b c 1
where a, b, c are constants. Find a, b, c so that the stress vector at P on the
octahedral plane is 0.
5. At a point P9 the stress matrix is given by
1

0-2

- 1 0

Find a plane through P parallel to the x3 axis on which the normal stress is 0.
6. At a point P, the stress matrix is as follows:
1

0-2"

[7] = 0 - 1

[-2

Find n such that 5j(n) = s2(n) = 0 at P. Also, determine s3(n) at P.


7. The stress matrix at a point P is given by
0

1 2

where a is a constant. Find a so that the stress vector on some plane at P is 0. Also,
specify this plane.

8. The stress matrix at a point (*,) is given by


[T(/] =

3xlx2

5x1

5x1

2*3

2x3

320

7 STRESS

Find the stress vector at a point P(l, - 1 / 2 , V3/2) on the plane that is tangential to
the cylindrical surface x\ + x\ = 1. Find also the normal and shear stresses.
9. The stress components at a point P with plane polar coordinates (/?, 0) are
r, i =

2a

2 = 2

,
cos 0,
a

22 =

sin 0 sin 20

cos 0 sin 20,

R 5* 0

To = T31 = 23 = 3 2 = 3 3 = 0

where a is a constant. Find the stress vector at P on the surface R = constant. Also,
find the normal and shear stresses on this surface. Evaluate all these at the point
(1,/4).
1 0 . At a point, the stress vectors on three planes are obtained as follows:
s(n) = el + 2e2 + 3e3

for

n = -ej

s(n) = 2yf3el + 2V3e 2

for

n = -=(! + e2 + e3)

s(n) = 2(e, + e2 + e3)

for

n = e2

Find the stress matrix at the point.


1 1 . By employing Cauchy's law, show directly (without appealing to a quotient
law) that ^ are components of a tensor.
1 2 . Let ASX and AS2 be two arbitrary plane elements at a point P such that the
stress vector on AS1 lies in 5 2 . Find a necessary and sufficient condition under
which the stress vector on 5 2 lies in ASj.
1 3 . Let s(n) and s(n') be the stress vectors at a point on two plane elements with
normals n and n'. Find the direction of the stress vector on the plane containing s(n)
and s(n'), given that the stress tensor is symmetric.
1 4 . Show that the sum of the squares of the magnitudes of stress vectors at a
point on planes parallel to the coordinate planes is invariant under coordinate
transformations.
1 5 . The state of stress at a point P is such that the stress vector on a particular
surface element AS is purely normal and the stress vector on every surface element
perpendicular to AS is 0. Show that the state of stress is uniaxial.
1 6 . If the state of stress is uniaxial, show that the stress tensor is of the form
T = T(a a), where a is a constant unit vector and T is a scalar (given that T is
symmetric).

7.10

EXERCISES

321

17. The state of stress at a point P is said to be apwre shear of magnitude T relative
to unit vectors a and b if a and b are orthogonal and if a surface element with
normal a experiences a stress of magnitude T in the direction of b and a surface
element with normal b experiences a stress of magnitude T in the direction of a. For
such a state of stress, find the stress vector on an arbitrary surface element, and the
stress tensor.
18. The stress matrix at a point is given by

M=

LI
2

2
;
2 -10

8 -10

Find the principal stresses and principal directions of stress at the point.
19, The stress matrix at a point is given as follows:
'10

M=

2"

0 1
2

0-2

Find the principal stresses. Show that the principal directions that correspond to the
largest and the smallest principal stresses both lie in the x3xx plane.
20. Find the principal stresses and the principal directions of stress at a point where
the stress matrix is as follows:

[T(/] =

21. Find the fundamental stress invariants for the state of stress defined by the
stress matrix
0 1 2^
1

22. Find the principal stresses at a point P, given that the stress matrix at P is
3 1 1

1 0 2
1 2 0

Hence, find the maximum of all the shear stresses acting on plane elements through
P. Also, determine the element on which the maximum shear stress occurs.

322

7 STRESS

2 3 . The stress matrix at a point is given by


a

lru] = 1 0

0 c

where a > b > c are constants. Find the maximum shear stress and the plane
element on which it acts.
2 4 . At a point P, the principal stresses are = 4, 2 = 1, 3 = - 2 . Find the stress
vector, the normal stress and the shear stress on the octahedral plane at P .
. The stress matrix at a point (") is

M=

0 -axl

ax2

-QXX

ax2
0

where a is a constant. Find the principal stresses and the principal directions of stress
at the point P (1, 2, 4). Also, compute the stress invariants, octahedral normal stress
and octahedral shear stress at P.
2 6 . At a point the principal stresses are = 2, 2 = tf, 3 = 1, where a is a
constant. Find a so that the octahedral shear stress is the maximum shear stress.
2 7 . At a point P , the principal stresses ,2, 3 are such that 2 = \{ + 3 ). Find
the plane element at P on which the normal stress is 2 and shear stress is \\ - 3 |.
2 8 . Prove the following:
(i)
(ii)

T2 = (T, - T2)2njnl + (T 2 T2ct

T 3 ) 2 /j| 3 2 + (T 3 -

T,)2?2

= i ( 2 / | - 6// T )

2 9 . Suppose that the xt system is coincident with the principal directions of stress
at a point P and let , 2 and 3 be the principal stresses at P . If -, are the stress
components w.r.t. the A: system also set up at P , show that

k=l

3 0 . Find the principal stresses for the state of stress that is a pure shear of
magnitude T relative to ex and e 2 , as defined in Exercise 17.
3 1 . The stress components at a point are such that 7 = with 13 = 23 = 0 and

33 = ( + 2)' Show that the maximum shear stress at the point is given by
W

= [( - 22 ) 2 + 4? 2 ] 1/2

3 2 . If a state of stress is such that 31 = 32 = 33 s 0 and 7 = , show that (i)


the x3 axis is a principal direction of stress and the corresponding principal stress is
0, and (ii) the principal directions in the xxx2 plane are inclined at an angle with

7.10

EXERCISES

323

the xx and x2 axes, respectively, where


2 :

tan 20 =
and the corresponding principal stresses are

i(T + 22) [i(r n - 22)2 + r 2 2 ] 1/2


3 3 . The stress components at a point P are given by /7 = ((, + ,)> where rf
and , are components of constant unit vectors and a is a constant. Find the
principal stresses and principal directions of stress at P.
34. Given the stress matrix
[] =

-3

V2

-3

-V2

|_V2 - V 2

find (i) the stress deviator matrix, (ii) the principal deviator stresses and (iii) the
stress deviator invariants.
35. The stresses in a circular cylinder of radius a with x3 axis as the axis are
given by
*13 =

31 = *1

23 = 32 = -/*0*2

T u = 22 = T 33 = T 12 = T 21 = 0

where and a are constants. Show that the boundary surface of the cylinder is stress
free.
36. The stresses acting at a point (x,) in a rectangular block bounded by the
surfaces xl = a, x2 = a and x3 = b are given as follows:
T

= -k(x\ - xl),

22 = k(x\ - x\)

T13 = 31 = 23 = 32 = 33 = 0

2aXiX2,

Find the total force on each face of the block.


37. The stress components in the half-space x3 > 0 are given by xu = (^/5)3,
r ^ 0, a > 0. Find the total force on the surface of the hemisphere r = a, x3 > 0.
38. A long circular cylinder with its axis along the x3 axis is subjected to a pressure
p on its surface. Write down the boundary conditions in terms of the stress
components /7.
39. A sphere with center at the origin is subjected to a pressure p on its surface.
Write down the boundary conditions in terms of the stress components ^.

324

7 STRESS

40- Prove the following relations connecting the components of T, T and S:


(0

ij

~jxi\kTkj

() Tij = JXjikTkj

(Hi) Sij = Jxitkxj,mTkm

~Jxi\kxj\mskm
s

ikxj;k
=

ikxj,k

41. Show that for small deformations, s^ ,,.

CHAPTER 8

FUNDAMENTAL LAWS OF
CONTINUUM MECHANICS

8-1
INTRODUCTION
Continuum mechanics is essentially based upon the following four fundamental mechanical principles, commonly known as conservation laws or
balance laws: (i) the law of conservation of mass, (ii) the law of balance of
linear momentum, (iii) the law of balance of angular momentum and (iv)
the law of balance of energy. These laws are postulated in the form of
equations involving certain integrals; such equations give rise to the socalled field equations that should hold at every point of a continuum and
for all time. The important feature of the field equations is that these
equations are applicable to all continua (solids, liquids, gases) regardless of
their internal physical structure. This chapter is devoted primarily to the
derivation of the field equations with some of their immediate consequences.
It will be seen that the field equations are inadequate to determine all the
unknown functions involved in the theory. These equations are therefore
to be supplemented by certain additional basic equations. The additional
325

326

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

equations appear in the form of relations representing the intrinsic physical


properties of the continuum under study. Such additional equations are
called constitutive equations. Unlike the field equations, the constitutive
equations distinguish one class of materials from the other. A brief discussion on the need for constitutive equations is given at the end of the chapter.
Consideration of the general theory of constitutive equations falls beyond
the scope of this book.

8.2
CONSERVATION OF MASS
In Section 4.4, the mass m of a material body (B in a configuration B of
volume V9 at time t was expressed as
(8.2.1)

m = \ pdV

where p = p(x, /) is the density at a point x of B and at time t. We now


postulate that mass m remains unchanged during the motion of (B. This
postulate is called the law (principle) of the conservation of mass. As a
mathematical expression, this law reads
^ =0
(8.2.2)
Dt
Substituting for m from (8.2.1) in (8.2.2) and using the transport formula
(6.5.3), we obtain
^

+ pdiv\)dV=

This equation is supposed to hold for any volume V; therefore, by


invoking the localization theorem (see Section 4.6) it follows that
Dp
-^- + />divv = 0

(8.2.3)

at every point of V and for all /.


Equation (8.2.3) represents the law of conservation of mass in the spatial
(Eulerian) form. This equation, which is a first-order partial differential
equation in the functions p and v, is one of the fundamental equations of
continuum mechanics. Traditionally this equation is called the equation of
continuity, in which the word continuity is used in the sense of constancy
(of mass). The equation was first obtained by Euler in 1757; a particular
case of the equation was given earlier, in 1752, by d'Alembert.

8.2

CONSERVATION OF MASS

327

The following equivalent forms of equation (8.2.3) are often useful:


D
(log p) + div v = 0

'G)

= divv

(8.2.4a)

(824b)

--

Using (6.2.9) and (3.4.14), equation (8.2.3) can be put in the following
alternative forms also:
dp

8.2.1

Y + v Vp + p div v = 0

(8.2.5a)

^ + div(pv) = 0

(8.2.5b)

EQUATION OF CONTINUITY IN MATERIAL FORM

Multiplying equation (8.2.3) by J and using the Euler's formula (6.2.51),


namely,
DJ
= /divv
(8.2.6)
Dt
we obtain
;(PJ) = 0

(8.2.7)

In the initial configuration where t = 0 and x = x, we have 7 = 1 . Let


p0 denote the initial density /?(x, 0). Then, integration of equation (8.2.7)
yields
(8.2.8)
pj = p0
This is the equation of continuity in the material (Lagrangian) form. This
equation is also given by Euler.
8.2.2

CASE OF INCOMPRESSIBLE CONTINUUM

From equation (8.2.3) we note that div v = 0 if and only if Dp/Dt = 0. But
Dp/Dt = 0 holds if and only if p remains unchanged during the motion;
that is, p = pQ for all time. Equation (8.2.8) shows that p = p0 holds if and
only if / = 1.

328

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

Thus, the following four conditions are all equivalent to one another.
divv = 0

(8.2.9a)
(8.2.9b)

Dt
P = A)

(8.2.9c)

/= 1

(8.2.9d)

In Section 6.3, it was noted that either of the conditions divv = 0 or


/ = 1 is a necessary and sufficient condition for a motion to be isochoric.
Since the four conditions (8.2.9) are all equivalent, it follows that each of
these four conditions is a necessary and sufficient condition for a motion to
be isochoric.
In Section 6.3, it was also pointed out that an incompressible continuum
is a continuum for which every motion is isochoric. Accordingly, each of
the conditions (8.2.9) holds for every motion of an incompressible continuum. Any of these four conditions can therefore be employed as the
equation of continuity for studying motions of such continua.
8.2.3

AN IMPORTANT CONSEQUENCE

Let = (, t) be an arbitrary scalar, vector or tensor function defined


over the configuration B. Then by using the transport formula (6.5.3) we
obtain
D
Dt

D_

(pd>)dV =

() + div v dV

f
}[Dt

Dp
\

t.

X]
dV
)\

Using (8.2.3), this reduces to


D

Dt \

J / Dt

(8.2.10)

This consequence of the equation of continuity is of great utility in our


further discussions.
EXAMPLE 8.2.1 Obtain equation (8.2.8) directly by using the law of
conservation of mass. Hence deduce equation (8.2.3).

8.2

CONSERVATION OF MASS

329

Solution If V0 is the volume of the initial configuration B0 of a body (B,


the mass of <B evaluated in this configuration is
p(x\0)dVo=\
p0dV0
(8.2.11)
J v0
Jv0
The law of conservation of mass states that m = m0. Hence, (8.2.1) and
(8.2.11) give
m0=\

PodV0=

p(x,t)dV

(8.2.12)

This is often referred to as the integral (or global) form of the law of
conservation of mass.
Changing the variable of integration from x to x in the integral on the
righthand side of (8.2.12) and bearing in mind that dV = JdV09 we get

(po-pJ)dVo

=0

Since this holds for any volume V0, it follows that p0 = pJ, which is (8.2.8).
Since p0 = p(x, 0) depends only on x, equaton (8.2.8) gives

that is,

f(^> =

Dt + JirDt =

On using the Euler's formula (8.2.6), this becomes

/ g + pdiv?j =0
This leads to equation (8.2.3) since 7 ^ 0 .
EXAMPLE 8 . 2 . 2

For the motion of a continuum given by the equations


xx = :? + atx

find the density in the current configuration in terms of the density in the
reference configuration.

330

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

Solution For the given motion, we find that 7 = 1 + 2a2t2. The equation of continuity in the Lagrangian form, namely, (8.2.8), gives the current
density as
P

JPo~

A)
l+2a2t2

EXAMPLE 8 . 2 . 3 For a certain flow of a continuum the velocity field is


given by vt = //(1 + / ) . Show that the density at time / is p = p0/(\ + t)3.
Deduce that pX\X2x^ = PoX 1X2X3
Solution

For the given velocity field, we find that


d i w = vk>k =

3
1 + /

Hence the equation of continuity (8.2.4a) yields


D
^(log/ = -

Integration of this equation gives


log/? = - 3 log(l + 0 + A

(8.2.13)

where A is, in general, a function of ;cf.


For t = 0, we have p = pQ. Hence (8.2.13) gives A = \ogp0. Putting this
into (8.2.13), we obtain
A)

/> = 7 7 ^ 3
(i + ty

(8-2.14)

The given velocity field can be rewritten as


DX;

Dt

X;

1+ t

Solving this differential equation, we obtain


x.

= jtf(i + /)

Expressions (8.2.14) and (8.2.15) yield pxxx2x^ = p0xx%x%.

(8.2.15)

EXAMPLE 8 . 2 . 4 (i) Show that in a moving continuum, the density, the


velocity and the vorticity are connected through the equation:
D /w\
/w
\
1
,Z)v
_ ( _ ) = ( _ . VJv + - c u r l
Dt\p)
\p
)
p
Dt
This is known as Beltrami's vorticity equation.

(8.2.16)

8.3

BALANCE OF LINEAR M O M E N T U M

331

(ii) Show that Cauchy's vorticity equation (6.6.8) can be expressed in the
form:
P
Solution

(8.2.17)

Po

(i) In Example 6.3.2, we have obtained the identity:


\
Dw
curl - = + (div v)w - (w V)v
Dt
Dt

(8.2.18)

Now,
lDw
D /w\
-^- ( ) =

Dt\p)

p Dt

1 Dp

9 -^-W
1

p Dt

1
-

>

X 1

l v(div v)w
Dt
' \

(8.2.19)

by the equation of continuity (8.2.3).


Substitution for (Dw/Dt) + (div v)w from (8.2.19) in (8.2.18) yields equation (8.2.16). For an incompressible material, equation (8.2.16) becomes
Dw /
D\
x
= (w V)v
+ curlt
Dt
Dt

(8.2.20)

(ii) Substituting for J from (8.2.8) in Cauchy's vorticity equation (6.6.8)


we obtain (8.2.17).

8.3
BALANCE OF LINEAR M O M E N T U M
For a material body (B occupying a configuration B of volume K, at time t,
we define the linear momentum p by
p = I p \ dV

(8.3.1)

As stated in Section 7.1, we consider two types of forces acting on (B:


body forces and surface forces. The result of these forces in the configuration B is given by (7.2.5), namely
f(r) = I phdV+

I sdS

(8.3.2)

where b is the force per unit mass and s is the surface force per unit area of
S, where S is the boundary surface enclosing V.
We postulate that the material time rate of change of p is equal to f(r).
This postulate, consistent with the Newton's second law of motion, is

332

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

known as the law (principle) of balance of linear momentum for a continuum. As a mathematical expression, this law reads
= f(r)
Dt
Using (8.3.1) and (8.3.2), equation (8.3.3) becomes
-\

p\dV=

pbdV+

sdS

(8.3.3)

(8.3.4)

By virtue of (8.2.10), we have

1" "1' '

<8 3 5)

TTndS=\

(8.3.6)

Cauchy's law connecting the stress vector and the stress tensor, namely,
(7.4.8), and the divergence theorem given by (3.7.1) yield
sdS=\

divTTdV

Substitution of (8.3.5) and (8.3.6) into (8.3.4) gives

~~

ph

" divTlfrfK=

Since V is an arbitrary volume, the integrand must vanish identically (by the
localization theorem). Thus
dWTT + pb = p^

(8.3.7)

at every point of V and for all /.


Equation (8.3.7) expresses the law of balance of linear momentum in
terms of Cauchy's stress tensor. This equation, first obtained by Cauchy in
1827, is known as Cauchy*s equation of motion.
In components, (8.3.7) reads
Tjij + pb^pjji

(8.3.7)'

For a continuum at rest (static equilibrium) or in motion with constant


velocity, we have v = 0; then equation (8.3.7) becomes
divT r + />b = 0

(8.3.8)

8.3

BALANCE OF LINEAR M O M E N T U M

333

This equation is referred to as Cauchy's equation of equilibrium. In


components, the equation reads
+ pbi = 0

(8.3.8)'

Recall that Cauchy's stress tensor T has been defined in the current
configuration. As such, the equation of motion (8.3.7) and the equation of
equilibrium (8.3.8) are in the spatial form. Next we deduce the corresponding equations in the material form.
8.3.1

EQUATION OF MOTION IN MATERIAL FORM

In view of relations (5.2.22), (7.9.5) and (8.2.8), expression (8.3.4) may be


rewritten, with reference to the initial configuration, as
Dx
Dt

]vo

Po*dV0=\

JK
0
JVQ

p0bdV0+\

J&

(T)TndS0

(8.3.9)

Employing the divergence theorem given by (3.7.1) to the surface integral


in (8.3.9) and noting that V (and therefore V0) is an arbitrary volume, we
obtain the following equation of motion expressed in terms of the first
Piola-Kirchhoff stress tensor, in the material form:
diAT)T + Pob = p0^

(8.3.10)

We can rewrite equation (8.3.10) in terms of the second Piola-Kirchhoff


stress tensor S. From expression (7.9.8), we get T = SFr. Substituting this
into (8.3.10), we obtain
div(FST) + p0b = p0^

(8.3.11)

From expression (5.4.4), we have F = I + Vu. Using this and expression


(6.2.5)', equation (8.3.11) can be expressed as
div((I + Vu)Sr} + A>b = A>^

(8.3.12)

Thus, (8.3.11) and (8.3.12) serve as alternative versions of the equation of


motion (8.3.10). In nonlinear elasticity, equations (8.3.10) to (8.3.12) are
more convenient than Cauchy's equation (8.3.7). Equations (8.3.10) and
(8.3.11) are by Piola (1833).

334

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

In components, equations (8.3.10) to (8.3.12) read, respectively, as


(8.3.10)'

&j + Poi=Poj
Dv;
IXr.kSjklj + Pobi = Po^f

(8.3.11)'

mk + uixk)sjk].j + ^ = Po-^r

(8.3.12)'

The equilibrium counterparts of equations (8.3.10) to (8.3.12) and their


corresponding component forms can be obtained by setting Ds/Dt (or
DVi/Dt) equal to 0 in these equations.
EXAMPLE 8.3.1
are given by

The stress components in a continuum in equilibrium

Tj j = ,

12 =

T22 = %2

21 = 2X1*2

23 =

33 =

32 =

31 =

13 =

Find the body force that must be acting on the continuum.


Solution Since the continuum is in equilibrium, the given stress components must obey Cauchy's equations of equilibrium (8.3.8)'. When
written in the expanded form, these equations read
dlii

dToi

* 3

+
dXi
fan ,
dxx

+
dx2
fan ,
dx2

(8.3.13)

+ 2 = 0)
dx$
33
dx3

Substituting the given expressions for the stress components in these


equations, we have
4xl + pbx = 0,

4x2 + pb2 = 0,

b3 = 0

Thus, the body force that must act on the material is given by
pb = -4(xlel + x2e2)

8.3

BALANCE OF LINEAR M O M E N T U M

335

EXAMPLE 8 . 3 . 2 At a point of a continuum in equilibrium under zero


body force, the stress matrix is given as follows:
X\ + *2
M

XT)

f(X\ 3

\ - 2*2

[
F'mdf(xx,

f(X\ >

x2

x2), given that on the plane xx = 1, the stress vector is


s = (1 + x2)ex - (x2 - 5)e2

Solution Since the continuum is in equilibrium under zero body force,


Cauchy's equations of equilibrium (8.3.13) are to be satisfied with , = 0.
Substituting for the stress components from the given stress matrix in these
equations, we find that the last equation is identically satisfied and that the
first two equations yield
dx2

= -h

dxx

= 2

Since/ depends only on xx and x2, these equations show t h a t / must be of


the form
f(xl,x2) = 2xl -x2 + A
(8.3.14)
where A is a constant.
For the plane xx = constant, n is given by nx = 1, n2 = n3 = 0, and
Cauchy's law (7.4.6) gives
s

i =

T n

= Tu-

For the given stress matrix, this yields


*2 = T 12 = f(Xl , X2)

But, it is given that


^2 = ~~(x2 - 5)
Hence
Axl9X2)

= -(X2-5)

on xx = 1
on

xx = 1

(8.3.15)

But, from (8.3.14) we get


f(xl9x2)

= A - x2 + 2,

for

xx = 1

(8.3.16)

Relations (8.3.15) and (8.3.16) give ,4 = 3. Putting this value of A back into
(8.3.14), we get
f(X\ > *2> = 2x\ - x2 + 3

336

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

EXAMPLE 8 . 3 . 3 The stress tensor in a continuum in equilibrium under


zero body force is given by T = T(y (x) y), where T = T(\) and y is the unit
vector along x. Show that
x VT+ 2T= 0

(8.3.17)

Find T in the case when T = T(r), where r2 = x x.


Solution The components of the given stress tensor are ,-, = Ty^j. Since
y is the unit vector along x, we have yt = xjr. Hence ,-, = (Tx^/r2,
so
that
*(/j = 1 ^

+ 27^

(8.3.18)

The continuum is in equilibrium under zero body force, so we have


Tjij = 0, by Cauchy's equations (8.3.8)'. We note that the given T is
symmetric. Hence it follows from (8.3.18) that
TjXj + IT = 0
which is the required result (8.3.17).
If T = T(r), then VT = (dT/dr)(x/r).

Hence (8.3.17) becomes

dT
2
+ - rr,
T = 0
dr
r

Solving this differential equation, we find that T = A/r2,


arbitrary constant.

where A is an

8-4
BALANCE OF ANGULAR M O M E N T U M
For a material body (B occupying a configuration B of volume V, at time /,
the angular momentum about the origin is defined by
h 0 = \{xx(pv)}dV

(8.4.1)

Also, the resultant moment, about the origin, of the external forces
acting on the body (B in the configuration B is defined by
go =

( x x (PW dV+\

where S is as usual the surface enclosing V.

(x x s) dS

(8.4.2)

8.4

BALANCE OF ANGULAR MOMENTUM

337

We postulate that the material time rate of change of h0 is equal to g0.


This postulate is known as the law (principle) of balance of angular
momentum for a continuum. As a mathematical expression, this law reads
* = go

(8.4.3)

On using expressions (8.4.1) and (8.4.2), equation (8.4.3) becomes


^

\[x x (pv)} dV=\

{xx (pb)J dV + f (x x s) dS

(8.4.4)

By using (8.2.10) and recalling that Dx/Dt = v, we get

Dt]

(x X py)dV=j^

\p(x X y)dV=

p^(x

X v)dV

f
Ds
= I Px x 7"

(8.4.5)

By using Cauchy's law (7.4.8) and (3.7.18), we get


js

x x s dS =\

}s

x x Trn dS
{2 + (xxdivTT)]dV

v
where is the dual vector of skw T r .
Substituting (8.4.5) and (8.4.6) in (8.4.4), we get

j fx X (pjj; - Ph - - 2] dV = 0

(8.4.6)

(8.4.7)

Using Cauchy's equation (8.3.7) and the localization theorem, equation


(8.4.7) yields = 0. Consequently, we have skwT r = 0, or
T =

(8.4.8)

at every point of the continuum and for all time.


Thus, the law of balance of angular momentum, as represented by
equation (8.4.3), leads to the interesting and important conclusion that the
stress tensor is symmetric. The symmetry relation (8.4.8) is also by Cauchy
(1827).
As a direct consequence of the expression (8.4.8) we find from (7.9.10)
that
FT0 = (T)rFr;
S = ST
(8.4.9)

338

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

In recent years, the principle of angular momentum represented by an


equation more general than (8.4.3) has also been considered. In this
generalized representation, the vector g0 includes the possible contributions
of the moments of internal forces developed within the continuum. Then
the stress tensor turns out to be nonsymmetric and one deals with what
is known as a polar continuum not considered in this text; see remarks
following the Cauchy's reciprocal relation (7.2.4).

8.5
GENERAL SOLUTIONS OF THE EQUATION
OF EQUILIBRIUM
In many static problems, the body force per unit of volume, namely, pb
(rather than the body force per unit mass, namely, b) is taken as a known
function. Then, the equilibrium equations (8.3.8)' contain ^ as the only
unknown functions. These equations are three in number and, therefore,
are generally not adequate to find all of }. However, the equations can be
solved to determine /7 in terms of some auxiliary functions, known as stress
functions. One such general solution is obtained next. Some other solutions
that follow from this solution are also presented.
For ready reference, let us write down the equations (8.3.8) and (8.4.8):
d i v T r + /?b = 0
T =

(8.5.1)
(8.5.2)

A general solution of these equations is provided by the following theorem.


THEOREM 8.5.1 Let A be an arbitrary symmetric tensor, and h an
arbitrary vector obeying the equation
Then

V2h = -pb

(8.5.3)

T = curl curl A + Vh + Vh r - (divh)I

(8.5.4)

is a solution of equations (8.5.1) and (8.5.2). Furthermore, this solution is


complete in the sense that every solution of equations (8.5.1) and (8.5.2)
admits a representation of the form (8.5.4).
Proof

If A is symmetric, the identity (3.5.40) yields


(curl curl A)T = curl curl A

showing that curl curl A is a symmetric tensor. Also, I and (Vh + Vh r ) are
symmetric tensors. Consequently, T given by (8.5.4) is a symmetric tensor
(being a sum of symmetric tensors). Thus, equation (8.5.2) is satisfied.

8.5

GENERAL SOLUTIONS OF THE EQUATION OF EQUILIBRIUM

339

Using identities (3.5.13), (3.5.14), (3.5.35), (3.5.37) and (3.5.38), we have


div curl curl A = curl div(curl A ) r = 0
div Vh = V2h,

div Vh r = V(div h),

div(div h)I = V(div h) )

(8.5.5)

Using these relations, we find from expression (8.5.4) that


divT = V2h

(8.5.6)

If h obeys equation (8.5.3), relation (8.5.6) becomes


div T + /?b = 0
showing that equation (8.5.1) is satisfied.
Thus, if A is a symmetric tensor and h is a vector obeying equation
(8.5.3), then T given by (8.5.4) is a solution of equations (8.5.1) and (8.5.2).
Next, we prove that the solution (8.5.4) is complete. Suppose T is a
solution of equations (8.5.1) and (8.5.2). Then we have
T = Tr,

div T + pb = 0

(8.5.7ab)

Let T be a tensor such that


V2T = - T

(8.5.8)

An equation of the form (8.5.8) is known to have a solution; therefore,


the existence of T is assured. Equation (8.5.7a) implies that T is symmetric.
Let
A = T - (trT)I

(8.5.9)

Then A is also symmetric. The identity (3.5.57) now implies that


curl curl A = V div T + (V div ) - div(div T)I - V2T

(8.5.10)

where we have used the results


tr A = - 2 tr T,

div A = div T - V(tr T)

Setting h = -div T and using (8.5.8), the relation (8.5.10) can be rewritten as
T = curl curl A + (Vh + Vh 7 ) - (divh)I

(8.5.11)

This expression is in the form (8.5.4). We have already noted that A is


symmetric. Since T obeys equation (8.5.7b) we verify from (8.5.11), with
the aid of the identities (8.5.5), that h obeys equation (8.5.3).
Thus, if T is any solution of equations (8.5.1) and (8.5.2), then T can be
expressed in the form (8.5.4), where A is a symmetric tensor and h is a vector
obeying equation (8.5.3). In other words, the solution of equations (8.5.1)
and (8.5.2) given by (8.5.4) is complete. The theorem is proven.

340

FUNDAMENTAL LAWS OF C O N T I N U U M MECHANICS

In the absence of the body force (b = 0), h turns out to be a harmonic


vector. This case of the solution (8.5.4), first obtained by H. Schaefer in
1953, is known as the Schaefer's solution. If h is set equal to 0 (with b = 0),
the solution (8.5.4) becomes
T = curl curl A

(8.5.12)

This is known as the Beltrami's solution, after Beltrami (1892).


In the suffix notation, the solution (8.5.4) reads:
*(/ = eimnjpq<lmptnq + Kj

+ */ f i ~ **,*<*</

(8.5.13)

where
aij = aji and hi^kk = -pbi
8.5.1

(8.5.14)

CASE OF CONSERVATIVE BODY FORCE

In many practical problems, the body force is conservative; that is,


pb = - for some scalar function . Then, using the identity = div (),
equation (8.5.1) can be written in the form
divT 0 r = 0

(8.5.15)

T0 = T - * I

(8.5.16)

where
Also, equation (8.5.2) can be rewritten as
T0 = T0r

(8.5.17)

By virtue of Theorem 8.5.1, it follows that a complete solution for


equations (8.5.15) and (8.5.17) is given by
T0 = curl curl A + Vh + Vh r - (div h)I
so that
T = curl curl A + Vh + Vh r + ( - divh)I

(8.5.18)

where A is a symmetric tensor and h is a harmonic vector.


Equations (8.5.15) to (8.5.17) are together equivalent to the following
three equations for ru:
*UJ-X.i

(8519)

Also, the solution (8.5.18) yields:


*U =

imnjp<lmp,ng + (A/J + A/ f i) + (X ~ **,*)/

(8.5.20)

where
au = and hUkk = 0

(8.5.21)

8.5

GENERAL SOLUTIONS OF THE EQUATION OF EQUILIBRIUM

341

The explicit expressions that follow from (8.5.20) and (8.5.21) are
Til = X + *22,33 + 33,22 ~ 2tf23,23 - A 2>2 ~ ^3,3 +
*22

kxx

X + 33,11 + 11,33 " 2tf 3 l,31 " ^ 3 > 3 - A M + A 2 2

^33 = X + 11,22 + 22,11 " 2tf 1 2 ,i 2 ~ * ! , ! -

A 2>2 + A33

(8.5.22)

*12 = 21 = 23,31 + 31,23 ~ 33,12 ~ 12,33 + (*1,2 + ^ 2> j)

23

32 = 31,12 + 12,31 ~~ 11,23 ~~ 23,11 + (^2,3 + ^3,2)

31

13 = 12,23 + 23,12 "~ 22,31 31,22 + (^3,1 + ^1,3)

These expressions constitute a general and complete solution of equations


(8.5.19).
8.5.2

PARTICULAR CASES

Several special solutions of equations (8.5.19) can be deduced from the


expressions (8.5.22) by making special choices of a^ and A,.
CASE i Suppose
\ 0 0]
[au\ = 0 0
and A, = 0

L fJ

Then expressions (8.5.22) become

Til = X + V,33 + .22


^22 = X + .11 + </>,33
^33 = X + 22 + 11
33

_'22

,U

(8.5.23)

21 12 ~, 12

32 =

23 =

~,23

= 3 =

-,

The solution of equations (8.5.19) as represented by (8.5.23) is known as


the Maxwell's solution, after Maxwell (1868). The function , and are
referred to as Maxwell's stress functions,
CASE ii Suppose
0
[/,]

* 0
* *

*
0

and A, = 0

342

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

Then expressions (8.5.22) become

= -

2*23

22 = X -

2%

33 = X -

2*12

(8.5.24)

\2 = *21 = ,*31 + %

- 0,*33

2 3 = 3 2 = *\ + 0*31 - C i i

31 = *13 = 0*23 + C l 2 ~ ^*22

The solution of equations (8.5.19) as represented by (8.5.24) is known as


the Morera's solution, after Morera (1892). The functions </>*, *, * are
referred to as Morera's stress functions.
CASE iii

Suppose

vf

and

ht = 0

where v is a constant and = (, :2).


Then expressions (8.5.22) become
Til = X + ,22
2 2 = * + f,ii
3 3 = + v22 + f,n)

12

21 = ~, 12

23

32 =

13 =

(8.5.25)

31 = 0

The solution of equations (8.5.19) as represented by (8.5.25) is known as


Airy's solution, after Airy (1863). The function is referred to as Airy's
stress function. This solution plays a fundamental role in the study of what
are called plane strain problems in the theory of elasticity.
CASE iv

Suppose

0 0 0
[*(/] =

0 0 0
0 0

and , = 0

8.5

GENERAL SOLUTIONS OF THE EQUATION OF EQUILIBRIUM

343

Then expressions (8.5.22) become


= X + , 22
22 = X + ,

12

13

21

31 =

(8.5.26)

~~ , 12

23 =

32 =

13 =

33 = 0

The solution of equations (8.5.19) as represented by (8.5.26) is also


known as Airy's solution. This solution plays a fundamental role in the
study of what are called plane stress problems in the theory of elasticity.
Note that solutions (8.5.25) and (8.5.26) are just particular cases of
Maxwell's solution (8.5.23).
CASE v

Suppose

[*</]

and

^ = 0

where * = *(, 2).


Then expressions (8.5.22) become
T

ll

2 2 T 33

12

T23 = - / i a f l

(8.5.27)

31 = 2
where
= *9
'

*0

The solution of equations (8.5.19) as represented by (8.5.27) is known as


PrandtVs solution, after L. Prandtl (1903). The function is known as
Prandtl's stress function. This solution is particularly useful in solving what
is known as the torsion problem in the elasticity theory.
Note that the solution (8.5.27) is just a particular case of Morera's
solution (8.5.24).

344

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

8.6
BALANCE OF ENERGY
The kinetic energy A' of a material body (B occupying a configuration B
of volume V, at time /, is defined by
K

= \ \ P(y-y)dv

(8.6.1)

Also, the rate of work or power of external forces acting on (B in B is


defined by
P=

pb-\dV+

s-vrfS

(8.6.2)

where S is as usual the surface enclosing V.


If the continuum is a heat conducting material and if there is a temperature difference between the interior of S and its exterior, then it is a physical
phenomenon that heat flows across S. In fact, if the temperature is higher
in the interior of S, heat flows out of V across S and if the temperature is
lower in the interior of 5, heat flows into V across S. We postulate that
the "amount" Q of heat flowing out of V across S per unit of time is
representable as
Q=\

qndS

(8.6.3)

where n is as usual the unit outward normal to 5, and q is known as the heat
flux vector (see Figure 8.1). It follows that the amount of heat that flows
into V across S per unit of time is given by Q.

Figure 8.1. Heat flux.

8.6

BALANCE OF ENERGY

345

If any heat is generated within V because of the presence of heat sources,


this heat contributes to the temperature difference across S. We postulate
that the amount of heat H generated within V per unit of time is representable as
(8.6.4)

H = \ phdV

where h is referred to as the heat supply or strength of the (internal) heat


source. It is obvious that the net amount of heat contained in V per unit
time is H - Q. We suppose that, apart from the heat flow, there is no other
energy flow (like that due to chemical energy or electromagnetic energy) in
the material.
We make a hypothesis that in addition to the kinetic energy the continuum possesses another type of energy called the internal energy and that
the total energy of a body of the continuum is the sum of the kinetic energy
and the internal energy. Like the concepts of time, mass, force, heat, etc.,
the concept of internal energy is also primitive. The internal energy E
possessed by a body (B in the configuration B is expressed as
E =

(8.6.5)

pedV

where is known as the internal energy per unit of mass or the specific
internal energy.
After having introduced the quantities AT, P, Q, H and E, we now
postulate a relation connecting these quantities as follows:
-(K

(8.6.6)

+ E) = P+(H-Q)

In words, this postulate states that the material time rate of change of the
total energy of a body (B is equal to the sum of the rate of work done by
external forces acting on the volume and the boundary surface of (B and the
rate of the net amount of heat contained in the volume. This is known as the
law (principle) of balance of energy or the first law of thermodynamics.
Substituting for K, E, P, Q and H from expressions (8.6.1), (8.6.5),
(8.6.2), (8.6.3) and (8.6.4) in the equation (8.6.6), we obtain the following
integral form of the law of balance of energy:
\ / ( ) + fijrfK= I pb-\dV+\

qntfS

s-\dS+\

phdV
(8.6.7)

346

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

By using (8.2.10), we note that


Dt

-\A

D\
De)
v + [dV
Dt
Dt)

(8.6.8)

By using (7.4.8), (2.8.14), (3.6.1) and (3.5.36), we get


s

s v dS = \ v dS = \ n Tv dS
js

Js

div (T\)dV

(v div T 1 + Ty Vv) dV

(8.6.9)

Also, by (3.6.1), we have


q-ntfS=

diwqdV

(8.6.10)

s
]v
With the aid of (8.6.8) to (8.6.10), expression (8.6.7) becomes
De

(D\

+ v

")

- d i v T r - pb\ - T Vv + divq - ph dV=0


(8.6.11)

Using Cauchy's equation (8.3.7) and noting that V is an arbitrary


volume, equation (8.6.11) yields
De

p - Vv + divq - ph = 0

(8.6.12)

Using the symmetry of T and noting then that T Vv = T D, equation


(8.6.12) can be rewritten as
De
p = T D - divq + ph

(8.6.13)

This equation represents the law of balance of energy. This is a first-order


partial differential equation that should hold at every point of the continuum and for all time. The equation is by Kirchhoff (1894) and usually
referred to as the energy equation. The term T Vv = T D is interpreted as
the rate of working of the stress, called the stress power. Equation (8.6.13)
is in the spatial form.

8.6

8.6.1

BALANCE OF ENERGY

347

EQUATION OF ENERGY IN MATERIAL FORM

We recall the law of balance of energy given by equation (8.6.7) and rewrite
it in the initial configuration. The resulting equation is

PohdV-\

<x-J(TyxTi0dS0

(8.6.14)
Here we have used (5.2.12), (5.2.22), (7.9.1) and (7.9.3) and the equation of
continuity (8.2.8).
From expression (7.9.4), we note that TTJ(T)~1 = (T)r. Hence
v Tr{y(Fr)-1nv) dS0 =
J SQ

() v dS0 = I n (Tv) dS0


J SQ

J SQ

div(Tv) dV0
(v div(T)r + (T0)7" (Vv)) dV0 (8.6.15)
Here we have used the divergence theorem (3.7.1) and (3.5.36).
Setting
q = 7F"'q

(8.6.16)

the last of the integrals in (8.6.14) can be rewritten as


J So

q J{Tyln dS0 =

J SQ

q n dS0

(8.6.17)

Evidently, q represents heat flux measured per unit area in the initial
configuration.
By the use of the divergence theorem, expression (8.6.17) becomes
q J(TrlndS0

=
divqdV0
(8.6.18)
)s0
}v0
With the aid of expressions (8.6.15) and (8.6.18), expression (8.6.14) can
be rewritten as

^7 + v-Lg-div(TV-A>b
- () (Vv) + divq - p0h dV0 = 0

(8.6.19)

348

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

Using the equation of motion (8.3.10) and noting that V (and therefore
V0) is an arbitrary volume, equation (8.6.19) yields
De

/ ^

M T Y (V\) - divq0 + ,

(8.6.20)

This is the equation of energy in the material form.


Prove the following formula for kinetic energy:

EXAMPLE 8-6.1
K

I [pix w x v + (x v) div vj
+ - |(x x v) (v x Vp) + (x v)(v Vp)}] dV
- 2 I /> x v) (v x n) + (x v)(v n)] dS

(8.6.21)

Deduce that for a material of constant density,


K =p

1
(x w x v) dV + I j - v x - (x v)v} n dS

This relation is known as the Lamb-Thomson


Solution
find that

(8.6.22)

formula.

By using identities (3.4.4), (3.4.14), (3.4.22) and (3.4.26), we


div(y9i;2x) = 3pv2 + x (p Vv2 + v2 Vp)

(8.6.23)

div{(x \)(p\)} = (x \)(p div v + Vp v) + pv2 + px w x v


+ \px Vv2

(8.6.24)

These relations yield


\\[\(2)

- (x \)(p\)] = jpv2 + jv2x -Vp-(x-

\)(p div v + v Vp)

- px (w x v)
(8.6.25)
Integrating both sides over the volume V and using the expression (8.6.1)
and the divergence theorem, we get
K

v (x n) - (x v)(v n) dS

1
(x v)(v Vp) - - v (x Vp) + p[(x v) div v + x w x v) dV
(8.6.26)
By use of the vector identity (1.7.51), expression (8.6.26) yields (8.6.21).

8.6

BALANCE OF ENERGY

349

If p is constant, we have divv = 0, by the equation of continuity.


Expression (8.6.26) then becomes (8.6.22).

EXAMPLE 8,6.2

Prove the following:

(i)

= \ (pb v - T D) dV +

(ii)

s v dS

(8.6.27)

= I (ph + T-O)dV-\
q-ndS
(8.6.28)
Jv
Js
Equation (8.6.27) is known as the equation of balance of mechanical
energy, and (8.6.28) is known as the equation of balance of thermal energy.
Solution

Dt

(i) Using (8.2.10), we find from equation (8.6.1) that

}\2

Dt

Dt

(8.6.29)

With the aid of (8.3.7), (8.6.9) and the symmetry of T, (8.6.29) becomes
DK

(divT +

Dt

1y

s v dS +

pb)-\dV
(pb v - T D) dV

This is the equation (8.6.27).


(ii) In view of (8.6.1) and (8.6.5), the equation of energy (8.6.7) can be
rewritten as
^-(K

+ E) =

p(b v + h)dV +

(s v - q n)dS

(8.6.30)

Subtracting equation (8.6.27) from equation (8.6.30) we get equation


(8.6.28).

Note: Equation (8.6.28) can also be deduced from the energy equation
(8.6.13) and vice versa. Indeed, equation (8.6.13) is just the local form of
equation (8.6.28).
Because of the presence of the common term \ -OdV, equations
(8.6.27) and (8.6.28) are interlinked. This means that, in general, the
motion and the thermal state of a material influence one another.
EXAMPLE 8 . 6 . 3 In the absence of body force and a heat supply, show
that the material rate of total energy in a material body contained in a

350

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

volume V is equal to the flux of the vector (Tv - q) across the surface S
enclosing V.
When b = 0 and h = 0, the energy equation (8.6.7) becomes

Solution

(K + E)=

\ (s v - q n)dS = \ (Tv - q) ndS

)s

Js

(8.6.31)

because s - v = T r n - v = T v - n . This proves the result


Note: The vector (Tv - q) is referred to as the energy flux

vector.

Show that

EXAMPLE 8 . 6 . 4

p n T D
p
DG
T D = K() = ^ - S
p0
' Dt
p0
Dt
Solution

(8.6.32)

We have
T D = tr{T(Vv)} = tr{(Vv)T)
= tr ( V v ) - F T 0

using (7.9.4);

-WTTI

J
\Dt
p n T D
= -()
Po
Dt

by (6.2.44);

(8.6.33)

Also,
< % * ) - < * % ) - < - % )
using (7.9.8);

'"

since S = ST;
,

DF\

= S . sym( FT r J + - S (FrF)

(8.6.34)

Using (5.4.3) and (5.4.8), this becomes


DC
DG
S = S
Dt
Dt
Expressions (8.6.33) and (8.6.35) together yield (8.6.32)

(8.6.35)
'

8.7

ENTROPY INEQUALITY

351

EXAMPLE 8.6-5 Show that equation (8.6.20) can be put in the


following alternative forms:

= tr T

[ S"] ~ divq + Poh

De

DG
o n
= S - divq + Poh

^
Solution

(8 6 36)

-'

(8.6.37)

We first note that

(TV (Vv) = (T)r <Vk)

= (TV ^ = tr [ T 0 ) ]

Using (8.6.35), this becomes


n T

(J)T

t\

D\s

D\J

(Vv) = - S - = S
2
Dt
Dt

(8.6.38)

Using (8.6.38) in equation (8.6.20), we obtain equations (8.6.36) and


(8.6.37).

8.7
ENTROPY INEQUALITY
In the preceding section, we postulated the law of balance of energy on the
assumption that a material body possesses an internal energy. In this section
we postulate what is called an entropy inequality on the assumption that a
material body possesses entropy, a primitive quantity that increases or
decreases accordingly as heat is supplied or withdrawn from the body. We
suppose that the entropy H of a. body < is expressible as
H=

pr\dV

(8.7.1)

where is known as the entropy per unit of mass or the specific entropy.
Since the entropy is associated with the heat content of a body, it is
directly related to the temperature, which is a measure of the degree of
hotness (or coldness) of a body. In what is called the absolute scale (or the
Kelvin scale), the temperature of a body is always positive. We denote the
absolute temperature at a point of a body (B by T. Then
Q= f ydV-

f ^-litfS

(8.7.2)

352

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

is called the rate of entropy input into the body (B in its current configuration. The integrand in the volume integral in (8.7.2), namely, (ph/T), is
called the entropy source and the integrand in the surface integral, namely,
(q/), is called the entropy flow.
With H and Q defined by (8.7.1) and (8.7.2), we now postulate that

(873)

*Q

In words, this postulate states that the material time rate of change of
entropy of a material body is no less than the rate of entropy input into the
body. This is known as the law (principle) of entropy or the second law of
thermodynamics.
Substituting for H and Q from (8.7.1) and (8.7.2) in (8.7.3) we obtain the
following integral form of the law of entropy:

Mrdv-lfdv+lr-ads^0

(8 7 4)

Rewriting the first term of this inequality by use of (8.2.1) and the last
term by use of the divergence theorem (3.6.1) and noting that V is an
arbitrary volume, we obtain the following inequality representing the law of
entropy in the local form:
Dt

\Tj

(8.7.5)

This is known as the Clausius-Duhem inequality, after R. Clausius (1854)


and P. Duhem (1901).
Using the fact that
div <* \) = - =A'
divdWq
q - 2 (V7) q
1

-T->

(8.7.6)

which follows by the use of (3.4.14) and the definition of V</>, the inequality
(8.7.5) can be put in the following useful form:

~ ph

+ divq

" Y{VT) 'q -

(8 7/7)

An alternative version of this inequality, obtained by the use of the


energy equation (8.6.13), reads
Dr\
De
1
pT-=r - PIT + T D - - ( ) q > 0
Dt
Dt
T

(8.7.8)

8.7

ENTROPY INEQUALITY

353

As mentioned earlier, it is a physical phenomenon that heat flows from a


region of higher temperature to a region of lower temperature. In other
words, heat does not flow against a temperature gradient. This means that
the vector fields q and VT act in opposite directions. That is,
(V7) q < 0

(8.7.9)

This is known as the classical heat conduction inequality. The equality in


(8.7.9) holds only when VT = 0.
In view of (8.7.9), the inequality (8.7.7) implies
Dn
pT - + divq - ph > 0

(8.7.10)

This is known as the Clausius-Planck inequality, after Clausius (1854) and


M. Planck (1887).
An alternative version of (8.7.10), obtained by the use of (8.7.8) and
(8.7.9), is

pT!- - p + T D > 0
(8.7.11)
Dt
Dt
Returning to the inequality (8.7.5), we note that the inequality is in the
spatial form. Its counterpart in the material form is obtained next.
8.7.1

ENTROPY INEQUALITY IN MATERIAL FORM

Let us recall the integral form of the entropy law as given by (8.7.4) and
rewrite it in the initial configuration of <B by use of the relations (5.2.12),
(5.2.22), (8.2.8) and (8.6.16). The resulting inequality is

Po

^tdV~\

TdV+\

^Q-n^o^0

(8.7.12)

Employing the divergence theorem to the surface integral in (8.7.12) and


noting that V0 is an arbitrary volume, we obtain

"

+ div

(^) -

(8 7 13)

-'

This is the Clausius-Duhem inequality in the material form. Observe that


this inequality is formally identical with its spatial counterpart (8.7.5).
The material versions of inequalities (8.7.7) to (8.7.11) can be written by
changing V and div to V and div, p and q to p0 and q and T D to
() Vv in (8.7.7) to (8.7.11).
The inequalities (8.7.5) and (8.7.13) are useful in studying restrictions on
the thermomechanical behavior of material models. In recent years, entropy

354

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

inequalities that are more general than these inequalities have been
proposed; their discussion falls beyond the scope of this text.
EXAMPLE 8-7.1 The specific free energy (or the Helmholtz free
energy) is defined by
= -
(8.7.14)
Show that
(0

Pj

= T D - divq + ph - ^()

(ii)

Dw
DT
1
Pjfi + pr\ - T D + -(VT)

q < 0

(8.7.15)
(8.7.16)

Solution Substituting for from (8.7.14) in the energy equation (8.6.13),


we obtain equation (8.7.15). Similarly, elimination of from (8.7.14) and
(8.7.8) yields the inequality (8.7.16).

EXAMPLE 8.7.2 Assuming that the heat-flux vector q is related to the


temperature gradient VT through the relation
q = -K(vr)

(8.7.17)

show that K is a positive definite tensor.


Solution Substituting for q from (8.7.17) into the classical heat conduction inequality (8.7.9), we find
(VD-K(vr)>0

(8.7.18)

Since this inequality holds for arbitrary VT and the equality in (8.7.9) is

valid only if VT = 0, it follows that K is a positive definite tensor.


Note: The relation (8.7.17) is known as the Fourier-Duhamel law of heat
conduction, after J. Fourier (1822) and J. M. C. Duhamel (1832). This law
is directly motivated by experimental observations at normal laboratory
conditions and is generally employed to study thermal states of solids,
liquids and gases. The tensor K is called the thermal conductivity tensor;
its components ki} are found to be either constants or functions of T.
Experimental observations also show that ku = k; that is, K is a symmetric
tensor.
A material is said to be thermally Isotropie if K is an isotropic tensor; in
such a material there is no preferred direction of heat flow. By virtue of

8.8

CONSTITUTIVE EQUATIONS

355

Theorem 2.5.1, it follows that, for a thermally isotropic material,


K = kl

(8.7.19)

where A: is a scalar. The positive definiteness of K implies that k > 0.


By use of (8.7.19) in (8.7.17), we obtain
(8.7.20)

q = -kWT

This is known as Fourier's law of heat conduction. The positive scalar k is


called the coefficient of thermal conductivity.
In recent years, heat conduction laws that are more general than (8.7.17)
have been proposed and used. These generalizations, called non-Fourier
heat conduction laws, are relevant only in specialized applications (not
considered in this text).

8.8
CONSTITUTIVE EQUATIONS
In the earlier sections of this chapter we have developed the following
equations.
Equation of continuity:
Dp

- p + p div v = 0
Po = pJ

(spatial form)
(material form)

(8.8.1a)
(8.8.1b)

Equations of motion:
divT r + pb = p .
Dt
div{(I + Vu)Sr) + p0b = p0

(spatial form)

(8.8.2a)

(material form) (8.8.2b)

Equation of energy:
De

p
De

p0

= T Vv - div q + ph

(spatial form)

= (1 + Vu)Sr- (Vv) - divq + p0h

(8.8.3a)
(material form) (8.8.3b)

356

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

These are the field equations representing the laws of conservation of


mass, balance of momenta and balance of energy, in the local form. These
are all partial differential equations and hold at every point of a continuum
and for all time. The initial density p0, the body force b and the heat supply
h are taken to be known functions.
It is readily seen that a total of 8 scalar equations are embodied in the
field equations while a total of 17 unknown field functions are involved in
these equations, the unknown functions being the density p, the three
velocity components t;, (or the three displacement components wf), the nine
stress components # (or } or s^), the three heat-flux components qt (or qf)
and the specific internal energy . Hence, these equations are inadequate to
determine all the field functions, and we need some more basic equations to
close the system.
Recall that the field equations (8.8.1) to (8.8.3) have been developed for
a general continuum without focusing on any particular material. As such,
these equations are valid for all continua irrespective of their internal
constitution. The equations developed in Chapters 5 to 7 also fall into this
category. This means that the equations developed so far hold equally well
to all solids, liquids and gases. But solids, liquids and gases have several
individual characteristic properties. We should therefore have appropriate
additional basic equations that reflect these properties. Equations that
represent the characteristic property of a material (or a class of materials)
and distinguish one material from the other are called constitutive equations.
In other words, constitutive equations are the equations describing the
relations between observable effects and the internal constitution of matter.
Thus, whereas the equations developed so far are based purely on
geometrical-dynamical principles and hold for all continua, the constitutive
equations are aimed at portraying the intrinsic physical properties that are
different for different continua. In the preceding paragraph, the need for
having more basic equations of the theory has been observed. The
constitutive equations adequately fill this need.
The constitutive equations are usually postulated directly from experimental data or by judiciously developed mathematical generalizations of
experimental data. However, in recent years, these equations are also being
developed by employing rigorous axiomatic approach based upon what are
known as the principles of consistency, coordinate invariance and material
frame indifference. The constitutive equations generally consist of (i) a law
of heat conduction, (ii) a caloric equation of state and (iii) a material law.
The law of heat conduction relates the heat-flux vector with temperature
gradient and is usually taken as given by (8.7.20). Caloric equation of state
is an equation determining the specific internal energy in terms of temperature and other field functions. This equation is again a postulate based

8.9

EXERCISES

357

upon experimental grounds, and the form of the equation depends on the
particular material under study. A material law usually consists of a set of
equations, each equation specifying a stress component in terms of other
field functions. The form of these equations again depends on the particular
material under study.
Sometimes, in the development of a theory, the number of field functions
is increased. In such cases, apart from the constitutive equations described
previously, some additional constitutive equations are postulated. The
kinetic equations of state, as represented by the Boyle's law for a "perfect
gas," is an example of such an additional constitutive equation. In many
situations, constitutive equations will be required to obey certain restrictions. The entropy inequality (8.7.5), which holds for all continua (solids,
liquids and gases), is an example of such a restriction.
The field equations (8.8.1) to (8.8.3) together with relevant equations
obtained in Chapters 5 to 7 as well as appropriate constitutive equations
generally serve as an adequate set of basic equations for determining all the
field functions. But since the constitutive equations necessarily differ from
one material to the other, we cannot have (even if we desire so) a common
set of such basic equations valid for all continua. We are thus lead to the
study of different branches of continuum mechanics separately, each branch
dealing with one set of constitutive equations. The branch of continuum
mechanics in which constitutive equations valid for solids are utilized is
known as solid mechanics. The branch in which constitutive equations valid
for fluids are utilized is known as fluid mechanics. The subjects of elasticity,
thermoelasticity, viscoelasticity, plasticity, hydrodynamics, gas dynamics
and so on, emerge as further subdivisions of continuum mechanics.
It is again emphasized that all the equations and results developed in
Chapters 5 to 7 and in the earlier sections of this chapter hold for all (nonpolar) continua and therefore remain valid in all branches of continuum
mechanics. The constitutive equations distinguish branches of continuum
mechanics.

8.9
EXERCISES
1. For the velocity and the density fields given here, check whether the equation of
continuity is satisfied.
(i) v = xleu

p = pQe~l

358

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

2. Given the velocity v = t(x2e2 + * 3 e 3 ), find the density p = p(t) such that the
equation of continuity is satisfied.
3 . Given the velocity v = {x3/(\ + /)e 3 ), find the density p = p(x3) such that the
equation of continuity is satisfied.
4. If the current density of an incompressible continuum is p = kxx, where k is a
constant, find the nature of velocity v such that v2 = 0.
5. Show that
V =

*2ei

~~ * l e 2

x\ + x\

can be velocity of an incompressible continuum.


6. Find k such that
= kx3(x2 - 2) 2 ,

v2 = -xYx2,

v3 = kxxx3

may be velocity components of an incompressible continuum.


7. Show that vt = ax^r3, where a is a constant and r2 = xkxk ^ 0 is a possible
velocity field for an incompressible continuum.
8. For an irrotational motion of a continuum with velocity v = , show that the
equation of continuity is
- i l + = 0
Dt
Deduce that if the continuum is incompressible, is a harmonic function.
9. A deformation of a continuum is defined by
xx = (1 + ? + bx2,

x2 = bxx + (1 + a)x2i

x3 = x3

where a and b are constants such that a > b - 1. Show that


p = p0{(\ + a)2 -

b2rl

1 0 . For any scalar function / , show that


Df
d
~Dt= Tt^

div(

^v>

1 1 . Show that
Dy

~Dt* 7 ( / ? v )

+ dlv(/?v

v)

Write down this expression in the suffix form.


1 2 . For any scalar function/, obtain the following generalized vorticity equation:
\
w
Df
1
D\
D /w
- V/ = - V + -curl
V/

Dt \p

Dt

Dt

8.9
1 3 . If V(div v) = 0, show that

EXERCISES

359

HdH

p = Po exp

1 4 . Verify that each of the following stress systems obeys Cauchy's equations of
equilibrium:
(i) = 22 = 12 = 23 = 13 = 0, 33 = pg* 3 , where p and g are constants,
with b = - g e 3 .
(ii) = 22 = 33 = 12 = 0, 23 = ,
constants, with b = 0.

13 = -2,

where and a are

(iii) = 22 = 33 = 12 = 0, 13 = (3/3 - 2), 23 = (3/32 + Xj)


where /i and a are constants, and = (, :2) is a harmonic function, with
b = 0.
(iv) = x2 + t;(*? - *f),
33 = v(x\ + x|),

22 = * + v(xl - x])

12 = -21^X2,

23 = 31 = 0

where v is a constant, with b = 0.


1 5 . The stress field in a continuum in equilibrium is given by

[*</] =

xlx3

x2(l - x\)

x 2 (l - x\)

x\ - 3x3

2xr?

Find the body force that must be acting.


1 6 . Repeat Exercise 15 for the stress matrix

M=

x\

2xxx2

2Xj x2

x2

2(xt + xi)

1 7 . The stress tensor in a continuum in equilibrium with zero body force is such
that 3 = 0. Show that

, + 22,22 + 212>12 = 0

1 8 . The stress tensor in a continuum in equilibrium with zero body force is given
by T = Ta a, where T = T(x) and a is a constant unit vector. Show that VT must
be everywhere perpendicular to a.
1 9 . Find the function = () such that the stresses
TU = &) + '()\ + 2U(r}ou

-'()\

obey the equations of equilibrium under zero body force. Here and are constants
and r1 = ,, ^ 0.

360

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

2 0 . In a continuum in equilibrium, the stresses are given by /7 =


ax^Xj/r5,
2
where a is a constant and r = ,, * 0. Find the body force that acts on the
continuum.
2 1 . Using the result of Exercise 11, show that Cauchy's equation of motion can be
written as

(pv) = div[T - p\ (x) v] + pb


ot
Write down this equation in the suffix form.
2 2 . For a function/ = / ( x , t), prove that
s

/ T n dS =

}v

[T(V/) + pfiy - b)] dV

2 3 . For a function f = f(x, t), prove that


((Vf)T + p (g) (v - b)) dV =\

fsdS

Deduct that

, , -1.

(i) I TdV=
(ii)

x sdS + \ px(g)(b - \)dV

T (Vu) dV = \ s u dS +

/?(b - V) u </K

2 4 . Deduce equations (7.2.4) and (7.4.3) as particular cases of equation (8.3.4).


2 5 . Deduce equations (8.3.10) and (8.3.11) from equation (8.3.7).
2 6 . Show that
s

[x x s]tdS =

Jv

eijk{Tjk + XjTmktm)dV

Using this expression and the suffix form of (8.4.5), show that equation (8.4.4)
yields eiJkTJk = 0. Hence deduce the relation (8.4.8).
2 7 . The center of mass xc of a body (B of mass m is defined by
xc = I px dV
By employing the law of balance of linear momentum, show that
D2xc
mT2 = pb
Dt
2 8 . Show that in the law of balance of angular momentum the origin may be taken
at the mass center with the velocity interpreted as that relative to the mass center.

8.9

EXERCISES

361

29. Assuming that the amount Q of heat flowing out of a volume V across its
boundary surface 5 per unit time is representable as Q = Js<7(x, n)dS, show that
q = qn.
30. For a continuum rotating like a rigid body with angular velocity w, show that
the kinetic energy is given by
K=-

}v

p{r2l - x <g> x) (w w) dV

31. For a material in which the stress tensor is of the form T = -pi, show that the
stress power is given by
p Dp
p Dt

If, further, the material is incompressible, deduce that the equation of mechanical
and thermal energy balances become uncoupled.
32. Prove that

33. Prove the following:

34. Deduce equation (8.6.20) from equation (8.6.13).


35. Show that the energy equation (8.6.13) can be expressed in the following form:

where
36. If T is uniform at each time, show that the Clausius-Duhem inequality (8.7.5)
reduces to the Clausius-Planck inequality (8.7.10).
37. If = T D - /Dt + (/))9 where is the specific free energy
defined by (8.7.14), show that the Clausius-Planck inequality is given by S > 0 and
that the Clausius-Duhamel inequality is given by TO > VT- q. (Here, is called
internal dissipation.)
38. If defined in Exercise 37 is exactly 0, show that the Clausius-Duhem
inequality reduces to the classical heat conduction inequality (8.7.9).

362

39.

FUNDAMENTAL LAWS OF CONTINUUM MECHANICS

Show that the material version of the energy equation (8.7.15) is

40.

~{)

+ )

' v v " div q + ph

Show that the material version of the inequality (8.7.16) is

CHAPTER 9

EQUATIONS OF
LINEAR ELASTICITY

9.1
INTRODUCTION
In this chapter, we consider one of the most important branches of
continuum mechanics: the classical theory of elasticity. This theory deals
with a class of continua called linear elastic solids. An elastic solid is a
deformable continuum that possesses the property of recovering its original
configuration when forces causing deformation are removed. An elastic
solid that undergoes only an infinitesimal deformation and for which the
governing material law is linear is called a linear elastic solid. From
experimental observations it is known that, under normal loadings, many
structural materials such as metals, concrete, wood and rocks behave as
linear elastic solids. The classical theory of elasticity serves as an excellent
model for studying the mechanical behavior of a wide variety of such solid
materials.
The classical elasticity theory is an essential part of solid mechanics and
its scope is vast. We restrict ourselves to the derivation of the governing
equations of the theory and some of its immediate consequences. Some
simple and standard applications are also presented.
363

364

EQUATIONS OF LINEAR ELASTICITY

9.2
GENERALIZED HOOKE'S LAW
As just mentioned, an elastic solid is a deformable continuum that recovers
its original configuration when forces causing deformation are removed.
In order to find a material law that portrays this characteristic behavior
of elastic solids undergoing infinitesimal deformation, we start with the
following set of linear relations:

11
12

33

+ "" +

3312^12 + ' " +

llll^ll +

\\\2e\2

1211^11 +

\2Yl \2

3311^11 +

+ '"

1133^33

1233^33

/Q 9 1)

3333^33

Here, are, as usual, the components of the stress tensor T arising due to
the presence of external forces, e^ are the components of Cauchy's strain
tensor E describing the infinitesimal deformation caused by these forces,
and c l i n , c 1 1 1 2 , ...,c 33 33 are 81 scalar coefficients that depend on the
physical properties of the solid and are independent of the strain components ei}. We suppose that relations (9.2.1) hold at every point of the
continuum and at every instant of time and are solvable for etj in terms of
Tij. Then it follows that ,-, are all 0 whenever eu are all 0, and that eu are all
0 whenever ^ are all 0. That is, TU and eu are homogeneous, linear functions
of each other. Physically, this means that the material for which the linear
relations (9.2.1) hold deforms in the presence of stresses and it recovers its
undeformed configuration when the stresses are removed. Further, since eu
are components of Cauchy's strain tensor, the deformation is infinitesimal.
Thus, the linear relations (9.2.1) describe the characteristic property of an
elastic solid undergoing infinitesimal deformation and constitute a material
law for such an elastic solid. An elastic solid for which the linear relations
(9.2.1) constitute the material law is referred to as a linear (or linearly)
elastic solid.
A particular case of relations (9.2.1) is
r n = Een

(9.2.2)

If the coefficient E is a constant, then the relation (9.2.2) states that for
a linear elastic solid the normal stress in the xx direction is directly proportional to the normal strain in the same direction. In effect, this relation was
first enunciated, on experimental grounds, by Robert Hooke in 1678 and is
known as Hooke's law. The material law described by relations (9.2.1) is a
natural generalization of Hooke's law, referred to as the generalized
Hooke's law. This (generalized) law is attributed to Cauchy (1822).

9.2

GENERALIZED HOOKE'S LAW

365

The nine relations in (9.2.1) can be expressed in a concise form as


T(/ = cijklekl

(9.2.3)

Since # and e^ are components of second-order tensors, using a quotient


law proved in Example 2.4.7, it follows from (9.2.3) that cijkl are components of a fourth-order tensor. This tensor characterizes the mechanical
properties of the material and is called the elasticity tensor. The 81 components cijkl of this tensor are called elastic moduli. Because of the symmetry
of TU and eiJf not more than 36 of these 81 elastic moduli can be independent. Since eu are dimensionless quantities, it follows from (9.2.3) that the
elastic moduli have the same dimensions as the stresses (force/length2).
By hypothesis, relations (9.2.3) should hold at all points xt of the material
and at all time t. The stresses /7 and the strains e^ vary in general with xt
and t. Likewise, cijkl may also vary with xt and t. If, however, cijkl are independent of Xi and t9 we say that the material is (elastically) homogeneous.
Thus, for a homogeneous elastic solid, the elastic moduli are constants so
that the mechanical properties remain the same throughout the solid for
all time.
The tensor equation (9.2.3) represents the generalized Hooke's law in the
Xi system. When written in the x\ system, the law would read
(9.2.4)

' = c[jklekl

where -, are components of T, e-j are components of E and c[jkl are


components of the elasticity tensor, in the x\ system. In general, /, and e-j
are different from ^ and e^. Likewise, c\jkl are generally different from
cijkl. If, however, c[jkl are the same as cijkl, we say that the material is
(elastically) isotropic. Thus, for an isotropic elastic solid, c-Jkl = ciJkl; that
is, the elasticity tensor is isotropic, which means to say that the mechanical
properties of the solid are independent of the orientation of the coordinate
axes.
In Section 2.6 it has been shown that every isotropic tensor of order 4 can
be represented in the form given by (2.6.1). Hence, for an isotropic elastic
solid, ciJkl are of the form
Cijki = cLijkl + ikoji

+ yujk

(9.2.5)

where , and y are scalars. Substituting for cijkl from (9.2.5) in (9.2.3) and
noting that eu = e we obtain (see Example 2.6.1),
Tij = auekk + ( + y)ei}
which on redesignating a as and + y as 2 yields
ru = k{jekk + 2eu

(9.2.6)

366

EQUATIONS OF LINEAR ELASTICITY

These relations represent the generalized Hooke's law for a linear, Isotropie
elastic solid. Evidently, the law now involves just two independent elastic
moduli A and//. These moduli are constants if the solid is also homogeneous.
The generalized Hooke's law as given by (9.2.6) was proposed by Cauchy
in 1822. The symbols A and were introduced later by Gabriel Lam in
1852; these are called the Lam moduli.
From relations (9.2.6), we get
rkk = (3A + 2)elkk

(9.2.7)

Substituting for ekk from (9.2.7) back into (9.2.6) and solving the resulting
equations for eu we obtain
2

TiJ

3A + 2

(9.2.8)

^ijTkk

provided ^ 0, 3A + 2 ^ 0; see Example 1.6.2.


Relations (9.2.8) and (9.2.6) are equivalent to one another. While (9.2.6)
express TU in terms of eU9 (9.2.8) express e^ in terms of TU.
In the direct notation, relations (9.2.6) and (9.2.8) read, respectively, as
follows:
T = A(tr E)I + 2/iE
(9.2.6)'
E =

1
2

3A + 2

(trT)I

(9.2.8)'

Relations (9.2.6) can also be rewritten in the form of a matrix equation,


as follows:
A

A + 2

+ 2

*12

'23

Til

22

+ 2

0 1 pul
0

en

<?33

ei2

e23

2\

L ^31 J

(9.2.6)"

The matrix equation equivalent to relations (9.2.8) can be written similarly.


It should be emphasized that the relations (9.2.3) and their special case
(9.2.6) are postulated to be valid only for linear elastic solids. For most solid
materials such relations hold until the stresses reach a limit called the
proportional limit. When the stresses exceed this limit, the deformation
becomes nonlinear but the elastic behavior of the material continues to exist

9.2

GENERALIZED HOOKE'S LAW

367

until the stresses reach what is called the elastic limit of the material. For
studying nonlinear deformations below the elastic limit, relations (9.2.3) are
to be replaced by more general (nonlinear) functional relations. The theory
is then called nonlinear elasticity. Consideration of this theory falls beyond
the scope of this book.
Henceforth, we will be concerned with linear elastic solids that are both
homogeneous and Isotropie. As such, relations (9.2.6) or their equivalents
(9.2.8) will represent the generalized Hooke's law for our further discussion
of the elasticity theory in this text. These relations will be referred to as
stress-strain relations or simply Hooke's law. Also, a homogeneous and
isotropic linear elastic solid will be referred to simply as an elastic body.
EXAMPLE 9.2.1 Show that for a linear isotropic elastic solid, the
principal directions of stress and the principal directions of strain coincide.
Solution Let a be a unit vector along a principal direction of strain at a
point of the solid. Then a is an eigenvector of the strain tensor E at the point
so that Ea = Aa for some scalar A. The Hooke's law (9.2.6)' then yields
Ta = [A(trE) + 2//A]a

(9.2.9)

This relation shows that the vector Ta is collinear with a; therefore, a is an


eigenvector of T as well. Hence, a is along a principal direction of stress
also.

EXAMPLE 9 . 2 . 2 Show that the stress-strain relation (9.2.6)' is equivalent to the following relations taken together:
tr T = (3A + 2) tr E

Solution

(/)

= 2//E

(i/)

(9.2.10a)
(9.2.10b)

From (7.7.1) and (5.10.1), we recall that


T*> = T - j ( t r T ) I

(9.2.11a)

E(d) = E - | ( t r E ) I

(9.2.11b)

From (9.2.6)' we find that (9.2.10a) holds and that


T ( d ) = {A(trE)I + 2) - j ( t r T ) I
using (9.2.11a);
= 2//(E-i(trE)Ij
using (9.2.10a);

= 2//E<d>

by (9.2.11b). Thus, relation (9.2.6)' implies the relations (9.2.10).

368

EQUATIONS OF LINEAR ELASTICITY

Conversely, with the aid of (9.2.11), relations (9.2.10) give


T = <*> + |(trT)I
= 2/4E - j(trE)I) + |(3A + 2//)(trE)I
= A(tr E)I + 2//E
which is the relation (9.2.6)'.

EXAMPLE 9.2.3 In a vertical elastic beam deforming under its own


weight (acting in the x3 direction) the strain components are found to be

e33 = a(b - x3),

e12 = e23 = e3l = 0

where a and b are constants. Find the stress components.


Solution To find the stress components from the strain components, it is
convenient to employ the matrix equation (9.2.6)". Substituting the given
strain components into this equation, and equating the corresponding
terms, we obtain the required stress components as follows:
Tu = 22 = 12 = 23 = 13 = 0
M3A + 2)
33 =
a(b - 3)
A ~r JU

9.3
PHYSICAL MEANINGS OF ELASTIC MODULI
With the view of obtaining the physical meanings of the elastic moduli
appearing in the Hooke's law (9.2.6), we consider the following particular
cases.
CASE i Suppose the stress tensor has only one nonzero component .
Such a stress system occurs in a beam placed along the jcraxis and subjected
to a longitudinal stress (Figure 9.1). Then, the relations (9.2.8) yield the
following expressions for the strain components:
en

~*?X+%i)T"'

e\i = e23 = e31 = 0

= 33 = - 2 3

+ 2//))

^ ^

9.3

PHYSICAL MEANINGS OF ELASTIC MODULI

369

Figure 9.1. Case i.

If we set
E =

(3 + 2)
+

.
v = 2( + )

(9.3.2a)
(9.3.2b)

then the relations (9.3.1) yield


(9.3.3)

en
e22

-33

= -v

(9.3.4)

Expression (9.3.3) is indeed the original version of the Hooke's law, given
by (9.2.2). Experiments conducted on most naturally occurring elastic
materials show that a tensile longitudinal stress produces a longitudinal
extension together with a contraction in transverse directions. Accordingly,
for > 0 we take en > 0, e21 < 0, e33 < 0. It then follows from (9.3.3)
and (9.3.4) that E > 0, v > 0.
From (9.3.3), we note that the constant E represents the ratio of the
longitudinal stress to the corresponding longitudinal strain. This constant is
referred to as Young's modulus, after Thomas Young, who gave a discussion of the elasticity theory in 1807.
From equation (9.3.4), we get
v =

22
*11

33

(9.3.5)

The constant v thus represents the numerical value of the ratio of the
contraction in a transverse direction to the corresponding extension in the
longitudinal direction. This ratio was introduced by Simon D. Poisson in
1829; it is known as Poisson's ratio.

370

EQUATIONS OF LINEAR ELASTICITY

Relations (9.3.2) can be solved for A and in terms of E and v; the


resulting relations are

- (i + Z -

">
(9 3 61

" - 2<rb>

"

Since E > 0 and v > 0, we have > 0.


The following consequences of the relations (9.3.6) are frequently needed:
.
E(l - v)
A + 2// = (1 + v)(l - 2v) '
A + 2 2(1 - v)

~ 1 - 2 '

A + /i

1
1 - 2v

A
v
A+ 2//~l-v

(9.3.7)

CASE ii Suppose the stress tensor is such that 7 = -^. Such a stress
system occurs when every point of a material is subjected to an all-around
pressure/?, as illustrated in Figure 9.2. Then we have xkk = -3/7 and (9.2.7)
yields

- -irh

If we set

(9 3 8)

(9.3.9)

=+$
it follows from (9.3.8) that

(9.3.10)

K =-
*kk

Experiments conducted on most naturally occurring elastic materials show


that an all-around pressure tends to reduce the volume of the material; that

Figure 9.2. Material with all-around pressure.

9.3

PHYSICAL MEANINGS OF ELASTIC MODULI

371

is, if p > 0, then ekk < 0. Consequently, it follows from (9.3.10) that
K > 0. Relation (9.3.10) also shows that the constant K represents the
numerical value of the ratio of the compressive stress to the dilatation. This
constant is called the modulus of compression or the bulk modulus. This
was introduced by George G. Stokes in 1845.
Substitution for A and from (9.3.6) into (9.3.9) yields

h>

(9 311)

Since K > 0, it follows from (9.3.11) that v < \. The relation (9.3.6a) then
yields A > 0. Thus, the two Lam constants A and are both positive.
From (9.2.6)" we note that
2M = T-^ = hi = Ii1
(9312)
en eu
e23
The constant 2 is thus the ratio of a shear stress component to the corresponding shear strain component; it is therefore related to the rigidity of the
material. For this reason, is called the modulus of rigidity or the shear
modulus. The other Lam constant A has no direct physical meaning.
The symbols A, , E, v and K are the five basic elastic constants that
appear in the theory of elasticity. As seen earlier, all of these are taken to
be positive on the basis of experimental observations, and barring A, all
have definite physical meanings. (The positiveness of the elastic constants
can be proven from thermodynamical considerations also.) In view of the
interrelationships that exist between these constants, only two of these are
independent. Some of these interrelationships are summarized in Table 9.1.
The constants , E, v and K that have definite physical meanings are often
called engineering elastic constants. For a typical copper material, the
approximate values of these engineering elastic constants, expressed in units
of 1010N/m2 are// = 4.8, E = 12.98, v = 0.343, K = 13.78.
Using (9.3.9) and (9.3.11), the relation (9.2.7) can be written as
ekk = JTM=

kk

(9.3.13)

From (9.3.13) we find that ekk = 0 if and only if v = y, provided of


course E and xkk remain finite. For v - y, the relations (9.3.6) and (9.3.11)
yield ->,->, = \. This limiting case corresponds to what is
called an incompressible elastic body. (Recall from Section 5.6 that, for an
incompressible continuum, ekk = 0.) Unless stated to the contrary, we will
be concerned with elastic bodies that are not incompressible; accordingly,
we take v < \ so that A and K remain finite, in general.

,
=

(1 + v)

2.
A + f

3 -

3 - A

9/f(* - A)

zv
\{-)

'

(,)

A(l + v)(l - 2v)

2v

(1 - 2)

(A, v)

2( + )

M3A + 2)

(,)

Elastic
Constant

3(3// - )

( - 2)
3-

(,)

2(\ + )

3(1 - 2)

2(1
+ v)

'

2
1 - 2

(,)

In terms of

3-
6

9/:-

3KQK - )
9 -

(,)

3-2

3(1 - 2)

2(1 + v)

(1 + )(1 - 2)

(,)

2{3 + )

3K +

9A/i

-
3

(,)

3K{\ - 2v)

3X1 - 2)
2(1 + )

1+

(,)

Table 9.1

372
EQUATIONS OF LINEAR ELASTICITY

9.3

PHYSICAL MEANINGS OF ELASTIC MODULI

373

It is often convenient to have relations (9.2.6) and (9.2.8) expressed in


terms of the constants E and v. Substituting for A and from (9.3.6) in
(9.2.6), we obtain
v
E

s 1
(9.3.14)
Similarly, with the aid of (9.3.6), the relations (9.2.8) can be rewritten as
1+ v

(9.3.15)

In the direct notation, relations (9.3.14) and (9.13.15) read as follows:


T =

1+ v

E +

E =
1 +v T
^

1 - 2v

(trE)I

(9.3.14)'

-(trT)I

(9.3.15V

Relations (9.13.14) and (9.13.15) can also be rewritten in the form of matrix
equations. For example, relations (9.13.15) read as follows:
en

\/E

-v/E

-v/E

-v/E

||

e2i

-v/E

\/E

<?33

-v/E

-v/E

\/E

*33

<?12

(1 + v)/E

Tl2

<?23

(1 + v)/E

<?31_

EXAMPLE 9.3.1

(1 +

22

723

v)/E\ LT31
(9.3.15)"

If
W = \\Xe\k + 2jteueu]

(9.3.16)

ove the following:


(i)

dW
dey -

(ii)

(9.3.17)

TiJ

(iii)
(iv)

<v)

(9.3.18)

2Tijeij

W is a scalar invariant
W>0
dW

an i

s ; " <

W= 0

(9.3.19)

if and only if e0 = 0
(9.3.20)

374

EQUATIONS OF LINEAR ELASTICITY

(The function W is called the strain-energy function; its physical meaning


is given in Section 9.4.)
Solution (i) From (9.3.16), we note that W is a function of eu. Partial
differentiation of this function with respect to e^ gives
3W
den

2Aekk

beirtr

+ 6

(9.3.21)

Since ekk = ^^, it follows that bekk/be{j = ^. Using this fact and
Hooke's law (9.2.6) in (9.3.21) we immediately get (9.3.17).
(ii) Noting that e\k = ekkeH = ekke^^y (9.3.16) may be rewritten as
W = \[kekku

+ Ie^eij

which yields (9.3.18), on using (9.2.6).


(iii) Since # and eu are components of tensors T and E, respectively,
W = jTijeij = jT - E is a scalar invariant.
(iv) Since > 0, > 0, and (ekk)2 and e^e^ are nonnegative, it follows
from (9.3.16) that W > 0 and that W = 0 if and only if ekk = 0 and eu = 0.
Since eu = 0 automatically implies that ekk = 0, therefore W = 0 holds if
and only if e^ = 0.
(v) Substituting for e^ in terms of TU from (9.3.15) in (9.3.18) we obtain
the following expression for W expressed in terms of # :
W

1 + v

kk

(9.3.22)

Differentiating this expression w.r.t. 7 , we get

dw
Noting that dxkk/dxu
(9.3.20).

1 + v

v
u~ETkk^

L
Tii

bxkk

(9.3.23)

= u and using (9.3.15) in (9.3.23) we obtain

EXAMPLE 9 . 3 . 2 Show that, for an incompressible elastic body, the


stress-strain relation can be written as
T = -pi

+ 2//E

where p = - y ( t r T ) is the mean compressive normal stress.

(9.3.24)

9.3

PHYSICAL MEANINGS OF ELASTIC MODULI

375

Solution For an incompressible elastic body, we have v = \ and = \E.


In this case, relation (9.3.15)' becomes
E = ^ - T - --(trT)I
2
6

(9.3.25)

so that

T = j(trT)I + 2
Setting p = -j(tr T) in (9.3.26), we get (9.3.24).

(9.3.26)

EXAMPLE 9.3.3 In an elastic beam placed along the x3 axis and bent by
a couple about the x2 axis, the stresses are found to be
E

11 22 T12 23 ~ 31 ~ ^

T33 7^*1

where R is a constant. Find the corresponding strains.


Solution It is convenient to employ the matrix equation (9.3.15)" to find
the strain components from the stress components. Substituting the given
stress components in (9.3.15)" and equating the corresponding terms, we
obtain the required strain components as
v

*11 = *22 = - * ! >

*33 = ~ ^ * 1 >

^12 = ^23 = *31 =

EXAMPLE 9.3.4 A beam placed along the xx axis and subjected to a


longitudinal stress at every point is so constrained that e12 and 33 are 0
at every point. Show that
22 = vr n

(9.3.27a)
2

1 -v
*n=g

Solution

(9.3.27b)

e33 = - V ( 1 * V ) T "
E
From (9.3.15)", we get
1

22

^ 22 ~^(T11 +

(9.3.27c)

33>

Since e22 = 0 and 33 = 0, this immediately yields (9.3.27a).


Using (9.3.27a) and the fact that r33 = 0 in the expressions for en and e33
contained in (9.3.15)", we get (9.3.27b,c).

376

EQUATIONS OF LINEAR ELASTICITY

9,4
GOVERNING EQUATIONS
In the preceding chapter we have obtained the following balance equations
that are valid for all continua.
Equation of balance of mass:
(9.4.1)
Equations of balance of momentum:
(9.4.2a)
(9.4.2b)
Equation of balance of energy:
(9.4.3)
Since the theory of elasticity is a branch of continuum mechanics (in which
the continuum is an elastic solid), the field equations (9.4.1)-(9.4.3) are
automatically valid in this theory.
The classical elasticity theory is based on the hypothesis that the deformation is infinitesimal. As such, the linear strain-displacement relation (5.6.4)
holds in this theory; that is, we have
Strain-displacement relation: E = y(Vu + Vu r )

(9.4.4)

The material law valid for homogeneous and isotropic linear elastic solids
is given by (9.2.6)'; that is, we have
Material law: T = A(tr E)I + 2//E

(9.4.5)

In Section 6.2, it was noted that for an infinitesimal deformation,


D/Dt d/dt so that
du
V
(9.4.6a)
~ dt
3

div v (div u)
dty

D\

d2u

~Dt "dt2

'

(9.4.6b)
(9.4.6c)

9.4

GOVERNING EQUATIONS

377

Using (9.4.6b), equation (9.4.1) becomes


1 dp
p at

d
at

(9.4.7)

The general solution of this partial differential equation is


p = p0 exp(-div u)

(9.4.8)

Thus, under the linear approximation, the continuity equation (9.4.1) has
been completely solved for p in terms of u.
Using (9.4.8), equation (9.4.2a) can be rewritten as
1
D\
T\
div T r exp(div u) + b =
(9.4.9)
Po
/
Dt
Simplifying the first term of this equation with the aid of MacLaurin's
expansion applied to exp(divu) and neglecting products of stresses and
displacements in the expansion and subsequently using (9.4.6c), we arrive at
the following linearized form of Cauchy's equation of motion (9.4.2a):
d i v T r + f = p 0^

(9.4.10)

where f = p0b is the body force per unit of undeformed volume.


Since E is a symmetric tensor, we find from (9.4.5) that T is also a
symmetric tensor. Thus, the symmetry of T is an intrinsic physical property
of an isotropic linear elastic solid. Consequently, the balance equation
(9.4.2b) is trivially satisfied, and equation (9.4.10) becomes
divT + f = p01-I 2
dt

(9.4.11)

The classical elasticity theory is an isothermal theory (that is, the


deformation is assumed to take place at a constant temperature). Therefore,
the last two terms on the righthand side of the energy equation (9.4.3) may
be set equal to 0 in this theory. Then, under the approximations given by
(9.4.6), this equation becomes
3e

1
1
/fti\
1
dE
_ T . Vv - T V U - T -
(9.4.12)
dt
p0
p0
\dtj
po
dt
Here we have used the relation (9.4.4), the symmetry of T, the relation
(2.9.8) and the approximation V V valid for small deformation.

378

9 EQUATIONS OF LINEAR ELASTICITY

Substituting for T from Hooke's law (9.4.5) in (9.4.12), we get


^

de
d
- = A(trE)-(trE)

dE
2^E-

i a

dw

= - - [A(tr E) 2 + 2/iE E] =
where

W = {A(tr E) 2 + 2//E E)

(9.4.13)
(9.4.14)

Obviously, if = 0 in the reference configuration, a general solution of


the equation (9.4.13), is
0 = W
(9.4.15)
Thus, the energy equation (9.4.3) is solved.
We note that, when thermal effects are absent, the only mechanism
available for storing internal energy in a body is through deformation. As
such, the solution (9.4.15), with W defined as a positive-definite function of
E through the relation (9.4.14), is meaningful on physical grounds also. By
virtue of this solution, the function W may be interpreted as the internal
energy of an elastic body per unit of undeformed volume. The internal
energy of an elastic body is called the strain-energy. The function W therefore represents the strain-energy per unit of undeformed volume, usually
referred to as the strain-energy function or the elastic potential.
Note that (9.4.14) is the same as (9.3.16). As such, the elastic potential W
has properties given by (9.3.17) through (9.3.20). The function W was first
introduced through equations of the form (9.3.17) by George Green in
1839. Expressions (9.3.20) are by A. Castigliano (1875). Expression (9.3.18)
is attributed to B. P. E. Clapeyron (1799-1864).
In the preceding paragraphs, the equation of conservation of mass and the
equation of energy have been completely solved and the momentum equation
(9.4.2a) has been linearized. Also, it is found that the momentum equation
(9.4.2b) is identically satisfied. Thus, the linearized momentum equation as
given by (9.4.11) is the only field equation to be taken account of in the
study of the classical elasticity theory.
Fifteen individual equations are embodied in the set of equations (9.4.4),
(9.4.5) and (9.4.11) and 15 unknowns also are present in the set, the
unknowns being 3 displacement components u{, 6 strain components eu and
6 stress components xu. The set of equations (9.4.4), (9.4.5) and (9.4.11)
therefore serves as a complete set of governing equations for the linear theory
of elasticity of homogeneous and Isotropie solids. We note that equations
(9.4.4) and (9.4.11) hold for all continua undergoing linear deformation
and equation (9.4.5) is the material law portraying the characteristic

9.4 GOVERNING EQUATIONS

379

behavior of homogeneous and isotropic linear elastic solids. Since these


equations form a complete set of governing equations, no more constitutive
equations are needed. In other words, the material law (9.4.5) is the only
constitutive equation needed for the theory under consideration.
In a problem in linear elasticity the displacement vector u is generally
extracted by integrating (9.4.4). Hence, the following compatibility condition obeyed by E, namely, (5.8.15), is also required:
V2E + VV(tr E) - V div E - (V div E) r = 0

(9.4.16)

Recall that this compatibility condition is just a consequence of the


strain-displacement relation (9.4.4); as such, (9.4.16) serves just as a
supplementary condition rather than a basic equation in its own right.
We observe that the governing equations (9.4.4), (9.4.5) and (9.4.11)
are expressed in the spatial form. Let us now look at the corresponding
equations in the material form.
9.4.1

EQUATIONS IN MATERIAL FORM

In Section 5.6, it has been noted that E could be regarded as a function of


x as well and that the strain-displacement relation (5.6.4) has the following
material form:
E = (Vu + Vur)
(9.4.17)
By using (5.5.4), (5.5.7), (5.6.8) and (7.9.10), we find that in the linear
case T T. Consequently, (8.3.10) which is the equation of motion in
material form becomes
divT + f = />o^?

(9.4.18)

Here, we have used the symmetry of T and the approximations given by


(9.4.6). Since T is now regarded as a function of x and /, f and u are also
required to be regarded so.
Equations (9.4.17) and (9.4.18) are the counterparts of equations (9.4.4)
and (9.4.11) in the Lagrangian description. We observe that these two sets
of equations are identical in structure. In other words, (9.4.4) and (9.4.11)
retain their form in the Lagrangian description also. Further, since the
stress-strain relation (9.4.5) is an algebraic equation, it also retains its form
in the Lagrangian description. As such, equations (9.4.4), (9.4.5) and
(9.4.11) could be employed both in the Eulerian description and in the
Lagrangian description, with A:, regarded as the current coordinates in
the Eulerian description and as the initial coordinates in the Lagrangian
description. The usual practice in the elastic theory is to employ the
Lagrangian description. Accordingly, throughout the remaining part of

380

EQUATIONS OF LINEAR ELASTICITY

this chapter, equations (9.4.4), (9.4.5) and (9.4.11) are employed with A:,
regarded as the initial coordinates. Consequently, the region V over which
Xi vary is regarded as the initial configuration of the body. On the same
count, we will write p for p0. This arrangement is of course a matter
of notation.
EXAMPLE 9.4.1 Show that the change in volume of an elastic body of
volume V, bounded by a closed surface 5, is given by
X dV+
dt2j '

I S>XidS

(9.4.19)

Solution We first recall that the change in volume of a continuum


undergoing small deformation is given by (5.6.26):
V=

\ekkdV

Substituting for ekk from (9.3.13) in this expression, we get


SV =

1 - 2v
y

1 - 2v

jv

1 - 2v*

rkkdV
TijijdV

1 -2v

(TijXiXjdV-

TijXijdV

TijjXidV

(9.4.20)

Employing the divergence theorem (3.6.1)' to the first integral on the


righthand side of (9.4.20) and using Cauchy's equation (9.4.11) in the
second integral, we get
SV =

1 - 2v

TijrijXidS + I I/, -

p-^)XidV

(9.4.21)

The result (9.4.19) is now immediate if we note that ^} = Sj.


Note: In the absence of inertial effects, (9.4.19) reduces to
SV=

~ 2 V | | s,x,dS+ I fiXidV
s
Jv

(9.4.22)

EXAMPLE 9.4.2 If an elastic body of volume V, bounded by a closed


surface S, is in equilibrium in a deformed state under a given system of body
forces fi and surface forces s,, prove that the total strain energy is

9.4

GOVERNING EQUATIONS

381

equal to one-half the work that would be done by the external forces acting
through the displacements u{ from the undeformed state to the state of
equilibrium. (This result is known as Clapeyron's theorem.)
Solution We recall that the elastic potential W represents the strain
energy per unit volume. Hence the total strain energy U contained in a
volume V is
U =

(9.4.23)

WdV

Using (9.3.18) and the fact that


= Muu

+ uj,i) =

T u

uu

expression (9.4.23) becomes


1
2}vTijUijdV

U=Z

(jijU^jdV-

TijjUidV

(9.4.24)

Employing the divergence theorem to the first of the integrals in the


righthand side of (9.4.24) and using Cauchy's law s, = x^rij and Cauchy's
equations of equilibrium T^J + / = 0, which follow by equating the righthand side of (9.4.11) to 0, we obtain
U

=2

SiUidS +

]v

fiUidV

(9.4.25)

We note that s, w, represents the work done by the surface force s per unit
of area with displacement u. As such \sSiUidS represents the total work
done by surface forces acting on 5. Similarly, f vfi,wf dV represents the total
work done by the body forces acting in V. Thus, the righthand side of
(9.4.25) represents one-half the total work done by body forces as well as
surface forces. This proves the required result.

EXAMPLE 9 . 4 . 3 Let ^ be the stresses corresponding to the strains


e\P and ^ be the stresses corresponding to the strains eff\ in an elastic
body. Prove that
(9.4.26)
If and obey the equations of motion under the body forces ffl)
andy/ 2 ) , respectively, and if ujl)9 uj2) and s<1}, sj2) are the corresponding

382

EQUATIONS OF LINEAR ELASTICITY

displacements and stress vector components, deduce that

\$sPuPdS

+ ^(/r

pd-^p)uVdV

- p ^ f ) , ( , ) dV

= j s /> dS + ^(fV

(9.4.27)

Interpret this result in the absence of inertia effects.


Solution

By virtue of the Hooke's law (9.4.5), we have

r#> = X6ue + 2;

rff> = A V # + **$>

These give
^eJP = + 2^><>

= (Aeg* + 2 / ^V> = rf^>

which is (9.4.26).
Using the symmetry of the stress components, the strain-displacement
relations, the divergence theorem, the Cauchy's law and the Cauchy's
equations of motion, we find that

^efdV=^u<?]dV
=

\J,TW\jdV-^T\WdV
S

= \ssM2)dS

JV

+ | ^ / / - p^pju?UV

(9.4.28)

Similarly,
vxfef

dV = k W > dS + J^ (2 > - / > ^ P ) ^ </K (9.4.29)

By the result (9.4.26), the lefthand sides of (9.4.28) and (9.4.29) are
equal. Hence their righthand sides are also equal. This proves (9.4.27).
In the absence of inertia, (9.4.27) becomes
f sil>u&dS + f ffl>ufdV = f s&uWdS+ f f&uPdV
(9.4.30)
Js
}v
]s
]v
The lefthand side of (9.4.30) represents the total work that would be done
by the force system {sf1},./)(1)} in acting through the displacements uf2) while

9.5

BOUNDARY VALUE PROBLEMS

383

the righthand side represents the total work that would be done by the force
system {$\./)(2)) in acting through the displacements u$l\ These two works
are equal, according to (9.4.30).
Thus, (9.4.30) has the following interpretation: if an elastic body is
subjected to two systems of body and surface forces producing two
equilibrium states, then the work that would be done by the first system of
forces in acting through the displacements of the second system is equal to
the work that would be done by the second system of forces in acting
through the displacements of the first system. (This result is known as the
reciprocal theorem of Betti and Rayleigh.)

9.5
BOUNDARY VALUE PROBLEMS
The most general problem of the elasticity theory is to determine the
distribution of stresses and strains as well as displacements at all points of
a body and at all time when certain boundary conditions and certain initial
conditions are specified. It has been noted that, in the linear elasticity
theory, the displacements, strains and stresses are governed by equations
(9.4.4), (9.4.5) and (9.4.11). Accordingly, solving a problem in linear
elasticity generally amounts to solving these equations for u, E and T in
terms of x and t under certain specified boundary conditions and initial
conditions, the body force f being assumed to be known beforehand. For a
body occupying a region V with boundary S, the boundary conditions
specified are usually of one of the following three kinds:
(i) The displacement vector is specified at every point of S and for all time
/ > 0; that is,
u = u* on 5 for / > 0
(9.5.1)
where u* is a known function.
(ii) The stress vector is specified at every point of S and for all time t > 0;
that is,
Tn = s* on S for / > 0
(9.5.2)
where s* is a known function. (Here we have used Cauchy's law s = Tn.)
(iii) The displacement vector is specified at every point of a part Su of S
for all time t > 0 and the stress vector is specified at every point on the
remaining part ST= S - Su for all time t > 0; that is,
u = u*
Tn = s*

on Su^i
on

ST )

for

/>0

(9.5.3)

384

EQUATIONS OF LINEAR ELASTICITY

It has been assumed that at time / = 0, the body is in an undeformed


state; that is,
u = 0 in V at t = 0
(9.5.4)
This is an initial condition. In addition, the following initial condition is
also employed:
du
= v* in V at t = 0
(9.5.5)
dt
where v* is a prescribed function. This condition implies that the velocity is
specified at every point of the body at time t = 0.
The problem of solving equations (9.4.4), (9.4.5) and (9.4.11) under the
initial conditions (9.5.4) and (9.5.5) and one of the boundary conditions
(9.5.1) to (9.5.3) to determine u, E and T at every point x of V and for / > 0 is
known as a boundary value problem in elasticity. A set {u, E, T) so determined, if it exists, is called a solution of the problem. When the boundary
condition is of the form (9.5.1), the problem is referred to as the displacement
boundary value problem; when the boundary condition is of the form
(9.5.2), the problem is referred to as the traction (or stress) boundary value
problem; and when the boundary condition is of the form (9.5.3), the problem is referred to as mixed boundary value problem. The three problems are
together called the fundamental boundary value problems. The boundary
conditions valid for all the three problems can be written down in the form of
(9.5.3). For the displacement problem, S = Su and STis empty; for the traction problem S = STand Su is empty, and for the mixed problem Su ^ S ^ ST.
In many applications, we consider problems in which inertia can be
neglected. For such problems, equation (9.4.11) becomes the equilibrium
equation:
divT + f = 0
(9.5.6)
Then, the boundary conditions become time independent and the initial
conditions (9.5.4) and (9.5.5) are redundant. The problems governed by
equations (9.4.4), (9.4.5) and (9.5.6) and the boundary conditions (9.5.3)
are referred to as boundary value problems in elastostatics.
On the other hand, time-dependent problems governed by equations
(9.4.4), (9.4.5) and (9.4.11), the initial conditions (9.5.4) and (9.5.5) and the
boundary conditions (9.5.3) are referred to as boundary value problems in
elastodynamics.
When a boundary-value problem is considered, natural questions arise:
does a solution exist? and if so, is it possible to find more than one solution?
A discussion of the first question involves great mathematical difficulties
and is beyond the scope of this book. However the second question can be
answered (in the negative) in a straightforward way.

9.6

UNIQUENESS OF SOLUTION (STATIC CASE)

385

9-6
UNIQUENESS OF SOLUTION (STATIC CASE)
First, we consider the case of elastostatic problems. For these problems, the
governing equations are (9.4.4), (9.4.5) and (9.5.6); namely,
E = |(Vu + Vu r )

(9.6.1)

T = A(tr E)I -l- 2

(9.6.2)

divT + f = 0

(9.6.3)

Also, the boundary conditions are (9.5.3), namely


u = u*

on

Tn = s*

SU9

on

ST

(9.6.4)

A solution of an elastostatic problem is a set {u, E, T} where u, E and T


satisfy equations (9.6.1) to (9.6.3) in V and the boundary conditions (9.6.4)
on S, the body force f being taken to be a known function. Next we prove
that such a solution is unique.
9.6.1

UNIQUENESS THEOREM IN ELASTOSTATICS

THEOREM 9.6.1 The solution of an elastostatic problem governed by


equations (9.6.1) to (9.6.3) and the boundary conditions (9.6.4) is unique
within a rigid body displacement.
Proof Suppose {u(1), E (1) , T(1)) and [u(2), E (2) , T(2)} are two solutions of an
elastostatic problem. Then equations (9.6.1)-(9.6.3) and the boundary
conditions (9.6.4) are satisfied by both the solutions. Thus,
divT ( 1 ) + f = <0
divT ( 2 ) + f = 0

in

T (1) = A(trE (1) )I + 2//E (1)

in

T (2) = A(trE (2) )I + 2 (2)


E (1) = i(Vu (1) + V u ( 1 ) r n
E (2) = i(Vu (2) + V u ( 2 ) r ) j
and

(9.6.5)

V
V

in

u (1) = u*,

u (2) = u*

on

Su

(1)

(2)

on

ST

T n = s*,

T n = s*

(9.6.6)
(9.6.7)

(9.6.8)

386

EQUATIONS OF LINEAR ELASTICITY

Suppose we set
= u (1) - u (2) ,

= E (1) - E (2) ,

T = T (1) - T(2>

(9.6.9)

Then (9.6.5) to (9.6.8) yield


divT = 0

T = A(tr)I + 2

(9.6.10a)
in

= |(V + V ) J

(9.6.10b)

(9.6.10c)

and
= 0

on

Tn = 0

SU9

on

ST

(9.6.11)

Let
/=

(9.6.12)

-EdV

Simplifying the integrand in (9.6.12) by using (9.6.10b) we obtain


/ =

{A(tr E)2 + 2 E} dV

(9.6.13)

If Hi, ij and xu are components of , and T, respectively, we note that


T = fjjeij = Tjjij. Expression (9.6.12) then yields

i = \jtuudv
=

(//"),7 -

*UJUAdV

= \ fuinjdS-\

TijjidV

(9.6.14)

In (9.6.14), the surface integral vanishes because of the boundary


conditions (9.6.11) and the volume integral vanishes because of equation
(9.6.10a). Accordingly, we have 1=0. Expression (9.6.13) therefore yields
(A(tr)2 + 2// -}dV=0

(9.6.15)
v
Since > 0, > 0, and (tr )2 and = eyey are nonnegative, the
integrand in equation (9.6.15) must be identically 0. Consequently,
trsO,

in

(9.6.16)

Equation (9.6.10b) then yields

in

(9.6.17)

9.7

UNIQUENESS OF SOLUTION (DYNAMIC CASE)

387

Since E = 0, the corresponding displacement represents a rigid-body


displacement; see Example 5.6.2. The relations (9.6.9), (9.6.16) and (9.6.17)
together imply that the two solutions {u(1), E(1), T(1)} and {u(2), E(2), T(2))
are identical within a rigid body displacement. The theorem is thus
proved.

This theorem was first established by Kirchhoff in 1859.


The crucial point in the proof of the theorem is the nonnegativeness of
the integrand in (9.6.13). In view of expression (9.4.14), this integrand is
nothing but twice the strain-energy function associated with the strain . As
such, the uniqueness of solution just proven is a direct consequence of the
positive definiteness of the strain-energy function.

9.7
UNIQUENESS OF SOLUTION (DYNAMIC CASE)
We now consider the case of elastodynamic problems. For these problems,
the governing equations are (9.4.4), (9.4.5) and (9.4.11), namely,
divT + f = p-^

(9.7.1)

T = A(trE)I + 2//E

(9.7.2)

E = |(Vu + Vu7)

(9.7.3)

the initial conditions are (9.5.4) and (9.5.5), namely,


u = 0,

du
= v* for t = 0

(9.7.4)

and the boundary conditions are (9.5.3), namely,


n = u* on Su
Tn = s*

on S

for / > 0

(9.7.5)

A solution of an elastodynamic problem is a set {u, E, T} where u, E, T


satisfy (i) equations (9.7.1) to (9.7.3) in V for t > 0, with f as known function, (ii) the initial conditions (9.7.4) in V for / = 0, and (iii) the boundary
conditions (9.7.5) on S for / > 0. Next we prove that such a solution is
unique.

388

EQUATIONS OF LINEAR ELASTICITY

9.7.1

UNIQUENESS THEOREM IN ELASTODYNAMICS

THEOREM 9.7.1 An elastodynamic problem governed by equations


(9.7.1) to (9.7.3), the initial conditions (9.7.4) and the boundary conditions
(9.7.5) cannot have more than one solution.
Proof Suppose {u(1), E (1) , T(1)) and [u(2), E (2) , T(2)) are two solutions of an
elastodynamic problem. Then equations (9.7.1) to (9.7.3), the initial conditions (9.7.4) and the boundary conditions (9.7.5) are satisfied by both
solutions. Accordingly, if we set
= E(1>-E(2),

= U (D_ U (2) >

= )_(2)

(96)

then 0, , T satisfy the following equations and boundary conditions:


divT = p ^

in

(9.7.7a)

T = A(tr )I + 2//

in

(9.7.7b)

= f(V + V 7 )

in

(9.7.7c)

= 0 on

S,

Tn = 0

on

ST

(9.7.8)

Relations (9.7.7) and (9.7.8) must hold for / > 0.


Also, we obtain the following initial conditions from (9.7.4):
= 0

(9.7.9a)

d
- = 0

(9.7.9b)

in V for / = 0. Let

(9 7 10)

- '* %'%]

Substituting for T from (9.7.7b) in (9.7.10), we obtain


N=\

v2

du

du

jyl(trE) 2 + 2//E-E + p-\

dV

= |^ [ A ( ^ ) 2 + 1vuu + p ^ J j dV

(9.7.11)

where uit e,j and 0 are components of u, E and T, respectively. Then


DN
Dt

= 2 (-f + ^ f ) + > ? ^

*''

9.7

UNIQUENESS OF SOLUTION (DYNAMIC CASE)

389

In obtaining this expression, we have used the approximation D/Dt d/dt


and the fact that V is regarded as the initial configuration.
By use of (9.7.7b), the symmetry of T and (9.7.7c), we find that
^ + 2 / /lJ ^ = [ ^
)
u

at

\dtjj

(9.7.13)

Using this expression, (9.7.12) becomes


DN

f f- ida-\

da a2

i "0 ...
(9.7.14)

Using the divergence theorem and (9.7.7a), expression (9.7.14) yields


DN

DF

2 i

d,

_,

n dS

\, -F '

(9.7.15)

The boundary conditions (9.7.8) imply that the righthand side of (9.7.15)
is 0. Consequently, we obtain
N = N0

(9.7.16)

where N0 is independent of time t.


By the initial condition (9.7.9a), we have w, = 0 at = 0. From (9.7.7c)
it then follows that u, = 0 at t = 0. We also have bujbt = 0 at t = 0 by
condition (9.7.9b). Consequently, (9.7.10) implies that N = 0 for t = 0. It
then follows from (9.7.16) that N = 0 for all time t > 0. Hence, from
(9.7.11) we get

f Utr) 2 + 2/i- + ^ ^ j r f F = 0

(9.7.17)

Since > 0, > 0, > 0, and (tr )2, and d/dt d/dt are nonnegative, it follows that the integrand in (9.7.17) is exactly 0; consequently,
(9.7.18a)
(9.7.18b)
(9.7.18c)
in V for all t > 0. Since = 0 in V at / = 0, (9.7.18c) yields
(9.7.19)

390

EQUATIONS OF LINEAR ELASTICITY

in V for all t > 0. The relation (9.7.18b) together with (7.7.7b) yields
T =0

(9.7.20)

in V for all t > 0.


Relations (9.7.6), (9.7.18b), (9.7.19) and (9.7.20) together show that the
two solutions {u(1), E(1), T(1)} and {u(2), E(2), T(2)) are identical. The theorem
is thus proven.

This theorem was first established by Franz E. Neumann in 1855.


The crucial point in the proof of the theorem is the nonnegative character
of the integrand in (9.7.10). This integrand is nothing but twice the sum of
the strain-energy function associated with and the kinetic energy function
associated with . As such, the uniqueness of the solution just proven is a
direct consequence of the positive definiteness of the strain-energy function
and the nonnegativeness of the kinetic energy function.
Having answered the question of uniqueness of solutions in the
affirmative (in both static and dynamic cases), our next aim is to look for
ways of solving the boundary value problems, assuming of course that the
solutions do exist.
Recall that solving a boundary-value problem means the determination of
fifteen unknowns ut, eu and ^ by solving the fifteen governing equations
embodied in (9.4.4), (9.4.5) and (9.4.11) under appropriate initial and
boundary conditions. But finding a solution of fifteen equations for fifteen
unknowns is a formidable task. It is therefore convenient to have fewer
governing equations with fewer unknowns. Since the boundary conditions
are usually specified in terms of displacement or stress, it is convenient to
express the governing equations entirely in terms of displacement or stress.
This is done in the following sections.

9.8
NAVIER'S EQUATION
We now proceed to express the governing equations (9.4.4), (9.4.5) and
(9.4.11) entirely in terms of the displacement vector.
Equations (9.4.4), (9.4.5) and (9.4.11) are
E = i(Vu + Vur)

(9.8.1)

a u

divT + f = pI

T = A(trE)I + 2//E

(9.8.2)
(9.8.3)

9.8

NAVIER'S EQUATION

391

Substituting for E from (9.8.1) in (9.8.3), we get the stress-displacement


relation:
T = A(divu)I + //(Vu + Vu 7 )
(9.8.4)
Using the identities (3.5.35), (3.5.13) and (3.5.14), we find from (9.8.4)
that
div T = //V2u + (A + //)V(div u)
(9.8.5)
Substitution of (9.8.5) into (9.8.2) yields the following governing
equation expressed entirely in terms of the displacement vector:
d2u
//V2u + (A + //)V div u + f = p ^

(9.8.6)

The suffix form of this equation is


MV2Ui + (A + )uktki +fi

a 2 ,

= p-^

(9.8.6)'

Equation (9.8.6) forms a synthesis of the analysis of strain, analysis of


stress and the stress-strain relation. This fundamental partial differential
equation of the elasticity theory is known as Navier's equation of motion,
after Navier (1821).
Navier's equation (9.8.6) can be written in several different forms. By
using the vector identity (3.4.27), this equation can be put in the following
two alternative forms:
d2u
(A + 2//)V u + (A + ) curl curl u + f = p
dtT2
,

(A + 2//)V div u - curl curl u + f = p -^2


dt

(9.8.7)
(9.8.8)

On using the relation [see (9.3.7)]


A+

1
1 - 2v

(9.8.9)

we can write equation (9.8.6) in one more form as


D2u +
where

Vdivu + - f = 0
1 - 2v

i d2

(9.8.10)

D2 = V 2 - ^ f 3

(9.8.11a)

cf = -

(9.8.11b)

c\ dt2

392

EQUATIONS OF LINEAR ELASTICITY

9.8.1

STATIC CASE

In the static case, equation (9.8.6) becomes


//V2u + (A + ) div u + f = 0

(9.8.12)

which reads as follows in the suffix notation:


2, + (A + )uk>ki + / , = <>

(9.8.12)'

Equation (9.8.12) is known as Navier's equation of equilibrium.


Like (9.8.6), equation (9.8.12) can also be rewritten in several different
forms. Of these, the following is a standard one:
(9.8.13)
Note that this is the static counterpart of equation (9.8.10).
EXAMPLE 9.8.1

If

(9.8.14a)
(9.8.14b)
prove the following:
(9.8.15a)

(i)

(9.8.15b)
()

(9.8.16)

(iii)

(9.8.17)

(iv)
(v>

(9.8.18)
/9(1 + V) d

i - 2 = E(l
^ - v)
a?

(9.8.19)

Solution (i) Substituting for (A + 2//) from (9.3.7) in (9.8.14b), we get


(9.8.15a). Substitution for from (9.3.6b) in (9.8.11b) yields (9.8.15b).
(ii) Using (9.8.11b), (9.8.14b) and (9.8.15) we readily obtain (9.8.16).
(iii) Expressions (9.8.14a) and (9.8.11a) yield (9.8.17).

9.8

NAVIER'S EQUATION

393

(iv) Substituting for c\ and c\ from (9.8.15) in (9.8.17), and simplifying


the resulting expression, we obtain (9.8.18).
(v) Expression (9.18.19) follows from (9.8.11a), (9.8.14a) and
(9.8.15).

EXAMPLE 9 . 8 . 2

Prove the following

(0

D,(divu) =

jdivf

(9.8.20)

(ii)

D2(curl u) =

2 cufl '

(9.8.21)

pc2

pel . W1

(iii)

X7 r\i\r

- v uiv

(9.8.22)

)
Solution (i) Taking the divergence of Navier's equation (9.8.6) with the
aid of the identity (3.4.13), we obtain the equation
b2
(A + 2//)V2(div u) + div f = p ^ (div u)
at

(9.8.23)

Rearranging the terms in this equation with the aid of (9.8.14), we get
(9.8.20).
(ii) Taking the curl of equation (9.8.6) with the aid of the identity
(3.4.16), we obtain the equation
d2
2 (1 u) + curl f = p ^ (curl u)
at

(9.8.24)

Rearranging the terms in this equation with the aid of (9.8.11), we arrive at
(9.8.21).
(iii) Operating the equation (9.8.10) throughout by \2X and substituting
for Djidiv u) from (9.8.20) in the resulting equation, we get
D 1 D 2 (divu)

* . V(divf) + - D ^ = 0
pc2(l - 2v)

(9.8.25)

Substituting for from (9.3.6b) in (9.8.25) and simplifying the resulting


expression with the use of (9.8.16), we obtain (9.8.22).

Note: If f is constant, the preceding results become D^div u) = 0,


D2(curl u) = 0 and D!D 2 u = 0. If, further, the deformation is static, we get
V2(div u) = 0, V2(curl u) = 0, V2V2u = 0.

394

EQUATIONS OF LINEAR ELASTICITY

EXAMPLE 9 . 8 . 3
body is given by

Show that the stress vector on a surface in an elastic

s = A(div u)n + 2( V)u + x curl u

(9.8.26)

Solution Substituting for T from the stress-displacement relation (9.8.4)


into Cauchy law: s = Tn and using the identities (3.5.34) and (3.6.19), we
obtain (9.8.26).

EXAMPLE 9 . 8 . 4

Show that for an incompressible elastic body,


T (d) = //(Vu + Vu r )

(9.8.27)

Deduce that
2 u + i ( V ( t r T ) + 3f) = 0
(9.8.28)
E
Solution For an incompressible elastic body, the stress-strain relation is
given by (9.3.26). Substituting for E from the strain-displacement relation
(9.8.1) in (9.3.26) and using the definition of T (d) we readily obtain (9.8.27).
Taking the divergence throughout in (9.3.26), using (9.8.1) and the
identities (3.5.35), (3.5.13), (3.5.14) and bearing in mind that div u = 0 and
= E/3, we obtain
div T = }[V(tr T) + E V2u]

(9.8.29)

Substitution of this expression into Cauchy's equation of motion (9.8.2)


yields equation (9.8.28).

EXAMPLE 9 . 8 . 5 When an elastic body undergoes spherically symmetric


deformation, the displacement vector is of the form
u = u(r)er,

r* 0

(9.8.30)

where er is the unit vector along the radial direction. For such a displacement, compute (i) the corresponding stress components, (ii) the normal
stress on a spherical surface r = constant and (iii) the normal stress on a
radial plane. Also, determine u{r) so that Navier's equation of equilibrium
with zero body force is satisfied.
Solution

where

(i) The expression (9.8.30) can be rewritten as


u = u(f)-\
r

= <t>(r)x

() = -^()

(9.8.31)

(9.8.32)

9.8

NAVIER'S EQUATION

395

From (9.8.31), we find that ux = </>(/*)*/ SO that


ij = <Kr)3u + \)

= uj9i

(9.8.33)

Therefore
*,* = 3() + '()

(9.8.34)

Substituting for uitJ and ujti from (9.8.33) and for ukk from (9.8.34) in
the stress-displacement relation (9.8.4) rewritten in the suffix notation and
using (9.8.32), we obtain the following expressions for the stresses associated
with the given displacement field:
TU = 2 (A 4- ) - -XiXjl -u(r) + \ku + 2-^}\u'(f)

(9.8.35)

(ii) For a spherical surface r = constant, we have n = er so that nt = xjr.


Therefore, by (7.5.3), the normal stress ar on this surface is given by
ar = Tij/tiftj = (juXiXj)/r2. Using (9.8.35), we obtain
ar = 2-u(r) + (A + 2)\'()

(9.8.36)

This normal stress is called the radial stress\ see Figure 9.3.
(iii) If n is the unit normal to a radial plane, we have ii er = 0, and the
normal stress ah on the plane is given by ah = ^^. By use of (9.8.35), we
obtain the following expression for ah :
ah = 2( + )-u(r) + ku'(r)

(9.8.37)

This normal stress is called the peripheral stress or the hoop stress; see
Figure 9.3.
(iv) Finally, to determine u(r), we return to (9.8.33) and (9.8.34) and find
from these expressions that
uUj = uktki = U"(r) + Uf(r))Xi

Figure 9.3. Radial stress and hoop stress.

(9.8.38)

396

EQUATIONS OF LINEAR ELASTICITY

Substituting these into Navier's equation of equilibrium (9.8.12)' with


fi = 0, we find that this equation is satisfied if () obeys the following
differential equation:
4
2
2 + - -r. =
(9.8.39)
dr2
r dr
The general solution of this equation is

0(r) = 4 + B

(9.8.40)

where A and B are arbitrary constants. Expression (9.8.32) then yields

u(r) = 4 + Br

(9.8.41)

Thus, when u(r) is as given by (9.8.41), u given by (9.8.30) obeys Navier's


equation of equilibrium with zero body force.
EXAMPLE 9 . 8 . 6 When an elastic body undergoes a deformation symmetric about the x3 axis, the displacement vector is of the form
u = u(R)eR,

(9.8.42)

R*0

where R2 = x\ + x\ and eR is the unit vector along the radial direction in


the cylindrical polar coordinate system with x3 axis as axis. For such a
displacement, compute (i) the corresponding stress components, (ii) the
normal stress on a cylindrical surface R = constant and (iii) the normal
stress on a plane containing the x3 axis (such a plane is called an axial
plane). Also, determine u(R) such that Navier's equation of equilibrium
with zero body force is satisfied.
Solution

(i) The expression (9.8.42) can be rewritten as


Ui = u(R)-Xi

= w(R)xii

ij

=1,2;

= ujfi

for

u3 = 0

(9.8.43)

Then
Uij = w(R)u + - w,(R)xixj
^
u3J ^ 0

ij

= 1, 2

(9.8.44)

so that
uktk = 2() + Ry/'(R)

(9.8.45)

Substituting (9.8.44) and (9.8.45) into the stress-displacement relation


(9.8.4) rewritten in the suffix notation and using the fact that y/{R) =
(l/R)u(R), we obtain the following expressions for stresses associated with

9.8

NAVIER'S EQUATION

397

the given displacement field:


r = }(A + 2)0 - 1,\

^u(R)

+ j + 2 jpXiXj\u'{R),

i,j=

1,2

(9.8.46)

1
T33 = A - (A) + '(/?)
31 = r32 = 0
(ii) For a cylindrical surface /? = constant, we have n = eR so that
/i, = xt/R for / = 1,2 and w3 = 0. Therefore, by (7.5.3), the normal stress
aR on this surface is aR = ^/,-, = {xyXiX^/R2; i,j = 1,2. Using (9.8.46),
we find that
= ^4&) + ( + 2)4'&)

(9.8.47)

(iii) If is the unit outward normal to an axial plane, we have n eR = 0,


and the normal stress ah on the plane is given by oh = x^n-fij. Using
(9.8.46), we obtain the following expression for ah:
= ( + 2)-&)

+ '&)

(9.8.48)

(iv) Finally, to determine u(R), we return to (9.8.44) and (9.8.45) and find
from these expressions that
R

d?\R

-)5

(9.8.49)

Substituting these into Navier's equation of equilibrium (9.8.12)' with


ft = 0, we find that the equation is satisfied if u obeys the following
differential equation:
dR

iS0"

= 0

(9.8.50)

398

EQUATIONS OF LINEAR ELASTICITY

Integration of this equation yields


u(R) = AR +
(9.8.51)
R
where A and B are arbitrary constants.
Thus, when u(R) is as given by (9.8.51), u given by (9.8.42) obeys the
Navier's equation of equilibrium with zero body force.

9-9
DISPLACEMENT FORMULATION
Having the governing equations expressed entirely in terms of the displacement vector u, as shown in the preceding section, the boundary-value
problems of elastodynamics can now be described entirely in terms of u.
For a mixed problem the boundary conditions are given (9.5.3). By using
the stress-displacement relation (9.8.4), these boundary conditions can be
rewritten as
u = u* on Su
r

[A(div u)I + //(Vu + Vu )]n = s*

on ST

for

/ > 0 (9.9.1)

The Navier's equation (9.8.6), the initial conditions (9.5.4) and (9.5.5) as
well as the boundary conditions (9.9.1), all expressed entirely in terms of u,
constitute what is known as the displacement formulation of a mixed
boundary-value problem in elastodynamics. In this formulation, the displacement u is determined by solving the equation (9.8.6) subject to
conditions (9.5.4), (9.5.5) and (9.9.1). Once this problem is solved, the
strain E and the stress T can be determined by using the strain-displacement
relation (9.8.1) and Hooke's law (9.8.3). It could be said that almost all
problems in elastodynamics are solved in this way. In the static case, the
same procedure is adopted except that the initial conditions (9.5.4) and
(9.5.5) are not required.
By virtue of the uniqueness theorems proven in Sections 9.6 and 9.7, the
displacement boundary-value problem has a unique solution in the static as
well as dynamic cases. However, the uniqueness of solution could be proven
directly by using the displacement formulation. Here we prove one such
uniqueness theorem in the dynamic case. The static case is left for the reader.
THEOREM 9.9.1 An elastodynamic problem governed by equations
(9.8.6), the initial conditions (9.5.4) and (9.5.5) and the boundary conditions (9.9.1) with S = Su cannot have more than one solution.

9.9

DISPLACEMENT FORMULATION

399

Proof Suppose u (1) and u (2) are two solutions of one and the same
problem obeying (9.8.6), (9.5.4), (9.5.5) and (9.9.1) for S = Su, and let
= u (1) - u (2) . Then wf satisfy the following equations and the initial and
boundary conditions:
d2;
+ (A + )k>ki = ^

(9.9.2)

in V for t > 0;
x, = 0,

^ = 0

(9.9.3)

in V for t = 0;
i = 0

(9.9.4)

on 5 for t > 0. It is sufficient to show that / 0 in V for all / > 0.


Multiplying (9.9.2) by t = di/dt and integrating the resulting equation
over V and setting f = d 2 W//3/ 2 , we get
}v

[UijjUi + (A + )uktkiUi

- pUiUi\

dV=0

This can be rewritten in the following form


[iji + (A +

)ktkj]jdV
+ (A + )ktkiti

[ijij

/,-,]

rfK = 0

(9.9.5)

Employing the divergence theorem to the first of the integrals in (9.9.5)


and noting that an over dot denotes partial derivative w.r.t. t, we obtain the
equation
+ (A +

Iiji
-\y

)ktkj]njdS

1 MuVfii.j)

+ (A + /*)(**,*)2 + ,-,.] rfK = 0

(9.9.6)

In view of the boundary conditions (9.9.4), the integrand under the


surface integral in (9.9.6) vanishes identically, and we obtain
(ij)(fiij)

+ (* + /0(*,*) 2 + M-j rfK = 0

(9.9.7)

for / > 0, where K0 is independent of t. At t = 0, the integrand in the


lefthand side of (9.9.7) vanishes by virtue of the initial conditions (9.9.3);

400

9 EQUATIONS OF LINEAR ELASTICITY

hence, K0 = 0. Further, since A, , p > 0, the integrand in equation (9.9.7)


is a sum of nonnegative terms, and so each term must vanish separately.
Thus, we arrive at the following set of equations:

(ijtjdv^o,

(ktk)2dv=o,

fiidv^o

for / > 0. Since the integrand in each of these integrals is nonnegative, we


find
ij = 0,
kik = 0,
f. 0
(9.9.8)
in V for all t > 0. The last result in (9.9.8) implies that t = uf in V for
/ > 0, where uf are independent of t. In view of the initial conditions on w,,
given in (9.9.3), it follows that uf = 0. Thus, wf = 0 in V for all t > 0, and
the proof is complete.

9.10
STRESS FORMULATION
We now proceed to recast the governing equations of elasticity entirely in
terms of the stress tensor T. As in the derivation of Navier's equation, let
us start with the strain-displacement relation (9.4.4); that is,
E = |(Vu + Vu r )

(9.10.1)

and the linearized Cauchy equation (9.4.10); that is,


divT + f = p

(9.10.2)

but take Hooke's law in the form (9.3.15)'; that is,


^ = 1^zrT
+ v
E
E

-(trT)I

(9.10.3)

From equation (9.10.2), we obtain


V(divT) + V(divT) r + (Vf + Vf r ) = p(V + V r )

(9.10.4)

From (9.10.1) and (9.10.3), we find that


V + V r = 2

-^

By using this in the righthand side of (9.10.4), we obtain the equation


V(divT)

V(divT)r-^_L^

v
KtrT)
1+ v

+ Vf + Vf = 0
(9.10.5)

9.10

STRESS FORMULATION

401

In the suffix notation, this equation reads


r

ik,kj

jk,ki

2p(l + v)

tf ""^"*** + fu+fj.i

(9.10.5)'

The partial differential equation (9.10.5) expressed entirely in terms of the


stress tensor (f being regarded as known) is by Valcovici (1951) and referred
to as the stress equation of motion. This serves as a counterpart of Navier's
equation (9.8.6), which is often referred to as the displacement equation of
motion.
Since (9.10.1) to (9.10.3) hold at every point of the volume V of the body
and for all t > 0, equation (9.10.5) also holds in V for all t > 0. To this
equation, we append the initial conditions
T = T0,

f = Tj

(9.10.6)

in V at / = 0, and the boundary condition


T = T

(9.10.7)

on S for / > 0, where T 0 , and T are specified functions.


Equation (9.10.5) together with the conditions (9.10.6) and (9.10.7)
constitute a boundary value problem in the stress formulation of elastodynamics. Once this problem is solved, the strain and the displacement can
be determined by using Hooke's law (9.10.3) and the strain-displacement
relation (9.10.1). The uniqueness of the solution of this problem is proven
next.
9.10.1

UNIQUENESS OF SOLUTION

THEOREM 9 . 1 0 . 1 The solution of the problem governed by equation


(9.10.5) with the conditions (9.10.6) and (9.10.7) is unique.
Proof For given f, T 0 , T! and T, let T (1) and T^2) be two solutions of the
problem governed by (9.10.5)-(9.10.7), and let T = T (1) - T (2) . Then the
components TU of T obey the following equations and conditions:
Tik,kj + iJkiU -

2p(l

V)

(fu - ^

a,!**) = 0

(9.10.5)"

in V for / > 0;
^ = 0,

^ = 0

(9.10.6)'

in V for / = 0;
fu = 0

(9.10.7)'

on S for / > 0. It is sufficient to show that fy = 0 in V for t > 0.

402

9 EQUATIONS OF LINEAR ELASTICITY

Multiplying equation (9.10.5)" by # and making use of the relations


^ik,kj^ij " \^iktk^ij),j
^jk,ki^ij

~ ^-ik^ijj

(9.10.8)

^ik,kj^ij

which are readily verified, we obtain the equation


[(Tik,kTij),j

ik,kTij,j]

p{\ + v)

l + v kk ii

T T

= 0

(9.10.9)

Integrating this equation over V and using the divergence theorem and the
boundary conditions (9.10.7)', we get
d_

<-

v
- , , / > ( ! + v) U
,- ,2

T
J J ~ ~t
"

E
C'
1 + v~A kk)

dV=0

(ik,kTijj)

(9.10.10)

Integrating this equation w.r.t. /, and using the identity


1+ v

(JkkY - ( -

J^ijTkk

*")

1 1 -2v

kk)
3 1 + v (fkkf

(9.10.11)

which may readily be verified, we obtain the equation


?ik,k?ijjdV

/>(1 + v)

2STkk)[*

- ^UTmm)dV

+**\<-'

(9.10.12)

where F is independent of /. In view of the initial conditions (9.10.6)', we


find that F = 0. Since p > 0, E > 0 and 0 < v < , it follows that
ik.k = 0,

fy-Wn^O,

tts0

(9.10.13)

in V. The last two relations in (9.10.13) give TU = 0; that is, fy are


independent of /. By virtue of the initial conditions (9.10.6)', it follows that
fij = 0 in V for all / > 0. This completes the proof.

The uniqueness Theorem 9.10.1 and its proof were first given by J.
Ignaczak in 1963.

9.11
BELTRAMI MICHELL EQUATION
Another important governing differential equation of elasticity that is also
expressed completely in terms of the stress tensor can be deduced by making
use of the compatibility condition (9.4.16).

9.11

BELTRAMI MICHELL EQUATION

403

Substituting for E from Hooke's law (9.10.3) into the compatibility


condition (9.4.16), we obtain the equation
V2T - (V div T) - (V div T ) r

V2(tr T)I +
VV(tr T) = 0
1+ v
1+ v
(9.11.1)
Substitution for divT from Cauchy's equation (9.10.2) in (9.11.1) and
simplification of the resulting equation by use of (9.10.1) and (9.10.3) yields
the following equation expressed entirely in terms of T:
D2 T - D2 (tr T)I + ! VV(tr T) + Vf + VfT = 0
1+ v
1+ v

(9.11.2)

This tensor equation can be further simplified as described next.


Taking the trace of equation (9.11.2), we get
![(1 - 2v)D2 + V2](trT) + 2divf = 0

(9.11.3)

By use of (9.8.18), equation (9.11.3) becomes


Di(tr T) = - j - - ^ div f

(9.11.4)

D2(tr T) = - f - j - ^ div f + (D, - D2)(tr T)j

(9.11.5)

so that

Substituting this back into the second term in (9.11.2), we obtain


D2T + ! {VV + v(Dj - D2)I}(trT) + ^(divf)I + Vf + VfT = 0
(9.11.6)
This is the desired governing differential equation expressed entirely in
terms of the stress tensor. In the suffix notation, this equation reads

i
2 lJ

r a2

1 + v IdXi dxj
+ ^

)
l

lJ

kk

*Ufk.k + fu + fj.i = 0

(9.11.6)'

In the static case, equations (9.11.6) and (9.11.6)' become


V2T + VV(trT) + (divf)I + Vf + VfT = 0
1+ v
1- v

(9.11.7)

v% + ^**. + yr^*ufk.t + fu + fu = 0 (9.11.7)'

404

EQUATIONS OF LINEAR ELASTICITY

The six equations embodied in (9.11.7)' were obtained by Donati in 1894


and by Michell in 1900. Earlier, in 1892, Beltrami had obtained the equations for the case of zero body forces. The dynamic counterpart of these
equations, namely, (9.11.6)', were given by Iacovache in 1950 for the case
v = \ and by Valcovici in 1951 for the general case. Equations (9.11.6) to
(9.11.7)' are referred to as the Beltrami-Michell equations,
Notice that equation (9.11.6) has been derived by starting with the compatibility condition (9.4.16) for the strain tensor E. For this reason, equation
(9.11.6) is often referred to as the equation of compatibility for stress.
In Section 8.5, it was pointed out that Cauchy's equations of equilibrium,
(8.3.8)', which are three in number, are inadequate to determine all of the
six stress components ,,. Together with the compatibility equations
(9.11.7)', of which there are six, Cauchy's equations (8.3.8)' serve as a
closed system of governing equations for ^ in elastostatic problems.
EXAMPLE 9 . 1 1 . 1 Find whether the following stress system can be a
solution of an elastostatic problem in the absence of body forces:
T n = 2$>

22 = ^3^1

12

3>

13

23

33

Solution In order that the given stress system can be a solution of an


elastostatic problem in the absence of body forces, the following equations
are to be satisfied:
(i) Cauchy's equations of equilibrium (8.3.8)' with/) = 0; that is,

*ii,i +

2,2 + T 1 3 i 3 = 0

*21fl + 22,2 + T23,3 = 0

31,1 + 32,2 + 33,3

(9.11.8)

(ii) Beltrami-Michell equations (9.11.7)' with/; = 0; that is,


o
1
V2TU + j - ^ ; ( T l l +

22 + *33),11 = 0

?
1
V 2 T 2 2 + j - q - ^ f r i i + *22 + T 3 3 ) f 2 2 = 0

V 2 T 3 3 + ~^

2 12 + ^ (

( T + 22 + 33 ) >33 = 0

(9.11.9)
+ 22 + 33),2 = 0

1
2,
7 1 3 + ^ ( * + 22 + 33 ) >13 = 0

1
23 + ^(

+ 22 + 33 ) >23 = 0

9.11

BELTRAMI-MICHELL EQUATION

405

It is easy to check that all the equations in (9.11.8) and all except the
fourth one in (9.11.9) are satisfied by the given stress system. Since the given
stress system does not satisfy the Beltrami-Michell equations fully, it
cannot form a solution of an elastostatic problem.

Note: This example illustrates the important fact that a stress system may
not be a solution of an elasticity problem even though it satisfies Cauchy's
equilibrium equations.
EXAMPLE 9 . 1 1 . 2
Navier's equation.

Deduce the Beltrami-Michell equation starting with

Solution Let us start with the Navier's equation as given by (9.8.10).


From this we get the equation
QiVu +

VV(div u) + - Vf = 0
1 - 2v

Adding this equation to its transpose and subsequently using the straindisplacement relation (9.8.1), we obtain the following equation of motion
expressed entirely in terms of E:
(9.11.10)
D 2 E + X - VV(trE) + -!-(Vf + Vf r ) = 0
1 - 2v
2
Substituting for E from Hooke's law (9.10.3) in (9.11.10) and recalling
that 2(1 + v) = E, see Table 9.1, we arrive at equation (9.11.2), from
which the Beltrami-Michell equation follows.

EXAMPLE 9 . 1 1 . 3 If u is a solution of Navier's equation, show that


the corresponding T is automatically a solution of the Beltrami-Michell
equation.
Solution By use of relations (9.3.14)', (9.3.13), (9.8.1) as well as identity
(9.8.19), the Beltrami-Michell equation (9.11.6) can be decomposed as
follows:

#>
+

(1 + v)(1

+ Vu7") + : j ^ ( d i v u ) I
_ 2 v ) [ W ( d i v u ) + v(D, - D 2 )(divu)I]

(div f )I + VfT = 0

(9.11.11)

406

EQUATIONS OF LINEAR ELASTICITY

Multiplying this equation by 2(1 + v)/E = / and rearranging the


terms, we get
\
V div u + - f
D2u +
V(divu) + - f + VI D 7 u +
1 - 2v
1 - 2v

2v
' A.
, (1 - 2v) A.
div f I = 0
1 - 2v D, div u +
2M1 - v)
By use of the identity (9.8.18), it is easy to check that

(9.11.12)

D^divu) + ^ ~ 2V\ divf


2(\ - v)
1 - 2v
V(divu) + - f
div D 2 u +
2(1 - v)
1 - 2v

(9.11.13)

In view of the identity (9.11.13), equation (9.11.12) can be rewritten as


D2u +

Vdivu + - f
V div u + - f + V D 2 u +
1 - 2v

1 - 2v

-div D 2 u +
2

Vdivu + - f I = 0
1 - 2v

(9.11.14)

If u is a solution of Navier's equation, each term on the lefthand side of


(9.11.14) is 0 and the equation is identically satisfied. Thus, if u is a solution
of Navier's equation, the Beltrami-Michell equation is automatically
satisfied.

Note: If T obeys the Beltrami-Michell equation, then it can be similarly


proved that the corresponding displacement u determined by using (9.10.3),
and (9.10.1) is a solution of Navier's equation.
EXAMPLE 9.11,4

Show that

1
D ^ T = T ^ [ ( W - vD 2 I)divf - (1 - vJD^Vf + Vf r )]
Solution
(9.11.15).

(9.11.15)

Operating (9.11.2) by Dj and using (9.11.4), we readily get

Note: If f = constant, it follows from (9.11.3) and (9.11.15) that


D^trT) = O a n d D ^ T = 0.

9.12

SOME STATIC PROBLEMS

407

9.12
SOME STATIC PROBLEMS
In the preceding sections we formulated the governing equations of the linear
elasticity theory in terms of displacements and stresses. We now employ these
equations to solve some simple problems of practical interest. Static problems are considered in this section; the next section deals with dynamical
problems. The four problems that follow are solved using the stress formulation, and problems 5 and 6 are solved using the displacement formulation.
9.12.1

AXIAL EXTENSION OF A BEAM

Consider an elastic beam of uniform cross section bounded by a cylindrical


surface and by a pair of planes normal to this surface. The cylindrical
surface is usually referred to as the lateral surface of the beam and the
planes are the bases (or end faces). Suppose that the beam is in equilibrium
under a uniform normal stress N acting on the bases. The lateral surface is
stress free and the body force is neglected. The problem is to compute the
stresses, strains and displacements at an arbitrary point of the beam.
It is convenient to choose the xl axis along the line of centroids of cross
sections of the beam with the origin lying on one of the bases. If / is the
initial length of the beam, the other base lies on the plane = I before
deformation; see Figure 9.4. Then, by hypothesis, the following boundary
condition hold:
= N;

2 = Tu = 0 for jcx = 0,/

on the lateral surface

Tijnj = 0

(9.12.1)

From conditions (9.12.1) we note that the stresses are prescribed at every
point of the boundary surface of the beam; the problem considered is
therefore a traction boundary-value problem.
X
1 3

N Z

0
*1

Figure 9.4. Axial extension of a beam.

408

EQUATIONS OF LINEAR ELASTICITY

Bearing in mind that nx = 0 at every point of the lateral surface, we verify


that the following stress system obeys the conditions (9.12.1):
?ii = N,

12 = 13 = 22 = 23 = 33 = 0

(9.12.2)

Thus, if we take (9.12.2) as the expressions for the stress system at any point
of the beam, then the boundary conditions of the problem are identically
satisfied. In Section 9.11, it was pointed out that Cauchy's equilibrium
equations (8.3.8)' and the Beltrami-Michell compatibility equations
(9.11.7)' together serve as a set of governing equations for stresses in an
elastostatic problem. Noting that N is a constant and f = 0 by hypothesis,
we check that the stress system (9.12.2) readily satisfies all these governing
equations.
The strains eu associated with the stresses given by (9.12.2) can be
obtained by substituting for } from (9.12.2) in the matrix equation
(9.3.15)". Thus, we obtain
1

KT

en =gN'>
*12 =

23

e22 = e 33 = - -

= ^33 =

(9.12.3)

The displacements wf associated with these strains follow from the suffix
form of the strain-displacement relation (9.4.4). Thus, we obtain (see
Example 5.8.4):
ux = -Nxl9

u2 = ~^Nx2,

u3 = -^Nx3

(9.12.4)

Expressions (9.12.2), (9.12.3) and (9.12.4) give the stresses, strains and
displacements that occur at an arbitrary point x of the beam. By virtue of
the uniqueness theorem proved in Section 9.6, these expressions constitute
the only possible solution of the problem. It is important to note that this
solution is completely independent of the length and the geometrical form
of the cross section of the beam. As such, the solution is valid for a beam
of any length and any cross section. Expressions (9.12.4) show that (i) the
displacements ux and u2 have the same values in all sections of the beam,
and (ii) plane sections xx = constant deform into plane sections xx + ux =
((E + N)/E)xx = constant.
EXAMPLE 9 . 1 2 . 1 In the problem just considered, find the normal and
shear stresses on an oblique section of the beam. Also, find their extreme
values and the sections on which these occur.

9.12 SOME STATIC PROBLEMS

409

Figure 9.5. Example 9.12.1.

Solution If n denotes the unit normal to an oblique plane section p of the


beam (Figure 9.5), the components of the stress vector on this section are
obtained, by use of Cauchy's law (7.4.9) and the relations (9.12.2), as
Sl(n)

= Nnl,

52(n) = s3(n) = 0

(9.12.5)

Using these expressions in (7.5.1) and (7.5.6), the normal stress and the
shear stress acting on the plane p are found to be
a = Nn\9

= (1 - \

(9.12.6)

If is the angle that n makes with the axis of the beam (xl axis), then
nl = cos and the expressions (9.12.6) can be put in the following
alternative form:
a = Ncos2e9

= |JV|sin20|

(9.12.7)

From these expressions it is evident that a is maximum for = 0 and


minimum for = 90. Also, is maximum for = 45 and minimum for
0 = 0. Thus, the normal stress is maximum on the right cross sections of the
beam and minimum on sections parallel to the axis. The shear stress is
maximum on sections inclined at an angle of 45 with the axis of the beam
and minimum on the right sections. The extreme values of normal and shear
stresses are given by
tfmax = Af,

r max = ^V;

crmin = 0,

Tmin = 0

(9.12.8)

REMARKS The matrix of the stresses given by (9.12.2) is purely


diagonal. Hence, the coordinate axes chosen for the discussion of the
problem are actually the principal axes of stress, and the principal stresses
are = 9 22 = 0 and 33 = 0. Since TV and 0 are the maximum and
minimum values of that occur when n is along the xx and x2 axes, respectively, the principal stresses and 22 are the maximum and the minimum
normal stresses. We also observe that the maximum value of occurs on the
plane that bisects the angle between the planes on which is maximum and
minimum, and that its magnitude is equal to one half the difference between
the maximum and minimum values of . These facts illustrate the results
proven in Examples 7.6.2 and 7.6.3.

410

9.12.2

EQUATIONS OF LINEAR ELASTICITY

EXTENSION OF A BEAM BY ITS OWN WEIGHT

Consider an elastic beam of uniform cross section suspended vertically by


fixing the centroid of the upper base. Suppose that a stress system that
balances the weight of the beam acts on this base and that the lower base
and the lateral surface of the beam are stress free. Suppose that there is no
external force other than the force due to gravity and that the beam is in
equilibrium. The problem is to find the stresses, strains and displacements
at an arbitrary point of the beam.
For convenience, we choose the centroid of the upper base as the origin
of the coordinate system with the x3 axis directed vertically downward along
the line of centroids of cross sections (Figure 9.6). Then the gravitational
force per unit mass is given by b = ge3, where g is the usual gravity factor.
Consequently, we have
t = pb = pge3
(9.12.9)
If / is the undeformed length of the beam, then the weight of a column
of unit cross sectional area of the beam is pig. This weight acts along the x3
axis and the stress at the upper base, x3 = 0, balances this weight. Hence,
the following conditions are to hold:
l

33

= pig,

*31

32 =

(9.12.10)

at JC3 = 0. Since the lower base and the lateral surface are stress-free, the
following conditions are also to be satisfied:
T u >
133
32

Tijnj = 0

= 0

at

= /

on the lateral surface

h^_D^

*2

1
1

n
!

*3

Figure 9.6. Extension of a beam by gravity.

(9.12.11)

9.12

SOME STATIC PROBLEMS

411

From conditions (9.12.10) and (9.12.11), we note that the stresses are
prescribed at every point of the boundary surface of the beam; as such, this
problem is a traction boundary-value problem.
Bearing in mind that n3 = 0 at every point of the lateral surface, we verify
that all the conditions contained in (9.12.10) and (9.12.11) are satisfied by
the following stress system:
Tu = 2 2 = 1 2 = 1 3 = 2 3 = 0

(9.12.12)

That is, if we take (9.12.12) as the expressions for the stress system at any
point of the beam, then all the boundary conditions of the problem are
satisfied. It is easy to verify that this stress system obeys also Cauchy's
equilibrium equations (8.3.8)' and the Beltrami-Michell equations (9.11.7)',
with f given by (9.12.9).
The strains e^ associated with the stresses ^ given by (9.12.12) can be
obtained by using the matrix equation (9.3.15)". Thus, we get

(9.12.13)

The displacements w, associated with these strains follow from the suffix
form of the strain-displacement relation (9.4.4). Thus, we obtain (see
Example 5.8.4, equation (5.8.56)):

(9.12.14)

It is trivial to verify that expressions (9.12.14) meet the condition that the
centroid of the upper base positioned at the origin is held fixed.
Relations (9.12.12), (9.12.13) and (9.12.14) give the stresses, strains and
the displacements that occur at an arbitrary point x of the beam. By virtue
of the uniqueness theorem proven in Section 9.6, these relations constitute
the only possible solution of the problem. We note that the solution is valid
for all geometrical forms of the sections of the beam.

41 2

EQUATIONS OF LINEAR ELASTICITY

From the results (9.12.14) we find that the points on the axis of the beam
are displaced only in the vertical direction and that the displacement u3
obeys the law:
"3

*3(

/ _

^3)

(9.12.15)

From equations (9.12.14) we also note that different points of a given


section x3 = c of the beam have different vertical displacements. As such,
the plane sections x3 = constant do not deform into plane sections. In fact,
we find from the last equation in (9.12.14) that the deformed shape of such
a section is a paraboloid of revolution having the x3 axis as the axis.
From (9.12.13), we find that the lateral contractions en and e22 in a cross
section vary directly with the distance of the section from the lower base of
the beam. As such, the maximum contraction occurs at the upper base of
the beam, the maximum contraction being (vpgl)/E. Also, no contraction
occurs at the lower base. Consequently, vertical line elements of the beam
do not remain vertical after deformation. The deformed shape of the beam
is indicated by the dotted lines in Figure 9.6.
EXAMPLE 9 . 1 2 . 2 Show that the vertical displacement at the centroid
of the lower base of a beam stretched by its own weight w is the same as that
produced by the load \w applied at the base.
Solution With reference to the axes shown in Figure 9.6, the coordinates
of the centroid of the lower base of the beam are (0, 0, /). Result (9.12.15)
gives the vertical displacement at this point as
(" 3 ), 3 = / = ^ /
IE

(9.12.16)

If A is the area of cross section of the beam, the weight w of the beam is
given by w = plAg. Using this, (9.12.16) may be rewritten as
(^

wl
= 1 = ^77;
2E

(9.12.17)

On the other hand, if \w is the load (total surface force) acting on the
lower base (x3 = /) of the beam, then the normal stress acting on this base is
(33)3

(9.12.18)

Hooke's law then gives


(*33)* 3 -/ = | ( * 3 3 ) * 3 - # = 2 ^

(9.12.19)

9.12

SOME STATIC PROBLEMS

413

Consequently, the strain-displacement relation yields


(9.12.20)

2AE

Relations (9.12.17) and (9.12.20) are identical, and the required result is
proved.

9.12.3

BENDING OF A BEAM BY TERMINAL COUPLES

Consider an elastic beam of uniform cross section and of length /. Suppose


that a couple of moment m about a line perpendicular to the beam is applied
at one of the bases and an opposite couple of moment - m is applied at the
other base (so that the beam is in equilibrium). The lateral surface of the
beam is stress free and the body forces are neglected. The problem is to find
the stresses, strains and displacements that occur at an arbitrary point of the
beam due to the bending it experiences because of the applied end couples.
This kind of bending of a beam is called simple bending or pure bending.
Let us choose the centroid of the base of the beam on which the couple
- m acts as the origin of the coordinate system, with the x3 axis initially
directed along the line of centroids of cross sections of the beam. Also, the
x2 axis is chosen along the axis of the couple m and the xx axis is chosen such
that the coordinate system is righthanded (Figure 9.7).
Due to the bending caused by the applied couples, the longitudinal
elements of the beam experience elongation or contraction. Consequently,
the stress vector s acting at a point on a cross section of the beam produces
a moment x x s per unit of area. The total moment on the cross section is
therefore ^( x s)dA, where A is the area of cross section. If the cross
section is the base x3 = /, then this moment should be equal to the moment
of the applied couple; that is,
(x x s) dA = m

(9.12.21)

We note that on the base x3 = /, we have s = s(3) and m = Me 2 ,


where M = |m|. Hence (9.12.21) yields, on equating the corresponding

~~~~?

-7

W J

x;

'

Figure 9.7. Simple bending.

*3

414

EQUATIONS OF LINEAR ELASTICITY

components,
A

(233 - h32)dA = 0
(/31 - xxx33)dA = M

(32 - x2z3l)dA

on x3 = I

(9.12.22)

=0 I

On the base x3 = 0, the moment of the applied couple is -Me 2 and the
stress vector is s = -s ( 3 ) . Hence, we should have
(XXS ( 3 ) )GL4
)A

=Me 2

(9.12.23)

on x3 = 0.
Equating the corresponding components in (9.12.23) we get
x2 33 dA = 0
xxx3ldA

(\12 ~ 23)^

= -M ) on x3 = 0

(9.12.24)

= 0

The conditions (9.12.22) and (9.12.24) are to be satisfied by the stresses


at the bases x3 = I and x3 = 0, respectively. It is easy to check that the stress
system
Tu = 22 = 12 = 13 = 23 = 0,
33 =
(9.12.25)
where a is a nonzero constant, meets all of the conditions in (9.12.22) and
(9.12.24) provided that
A

2 = 0

(9.12.26a)

(9.12.26b)
a
We note that the integral in (9.12.26b) represents the moment of intertia
of a section about the x2 axis, denoted /. Then (9.12.26b) is satisfied if we
take a = -(M/I). Further, the integral in (9.12.26a) represents the product
of intertia of a section about the x2 and x3 axes. If the axes are assumed to
x\dA =

9.12

SOME STATIC PROBLEMS

415

be along the principal axes ofintertia, then this product is identically 0, and
(9.12.26a) is identically satisfied. Thus, with a = -(M/7), the stress system
(9.12.25) satisfies all of the conditions in (9.12.22) and (9.12.24).
The lateral surface of the beam has been assumed to be stress free. Hence
the conditions ^} = 0 are to hold at every point of this surface. Since
n3 = 0 at all these points, we readily find that the stress system (9.12.25)
meets these conditions as well.
Bearing in mind that f = 0, we check that the stress system (9.12.25) also
obeys Cauchy's equilibrium equations (8.3.8)' and the Beltrami-Michell
compatibility equations (9.11.7)'.
Thus, the stress system
Tu = 22 = 12 = 13 = 23 = 0
M
33 = - y * i

(9.12.27)

satisfies all the governing equations and all the boundary conditions
relevant to the problem.
The strains eu associated with the stresses ,, given by (9.12.27) follow by
using Hooke's law (9.3.15)". Thus, we get
vM
eu = e22 = - X l ,

M
e,,---*,

12 = 23 = 13 =

The corresponding displacements follow by the use of strain-displacement relation (9.4.4). Thus, (see example 5.8.4)

"i

2t

+ v

Mv
u2 = ^7*1*2
"3 =

t -*2
(9.12.29)

Relations (9.12.27), (9.12.28) and (9.12.29) give the stresses, strains and
displacements that occur at an arbitrary point x of the beam. By virtue of
the uniqueness of solutions, these relations constitute the only possible
solution of the problem. We note that this solution is valid for all geometrical forms of the sections of the beam provided that the moment of the
applied couple is directed along a principal axis of inertia of the section.

416

EQUATIONS OF LINEAR ELASTICITY

From (9.12.28), we find that longitudinal elements of the beam extend or


contract depending on whether xx > 0 or xx < 0. For xx = 0, all the strains
are 0; as such the elements initially lying on the x2x3 plane do not change in
length. The x2x3 plane is therefore referred to as the neutral plane of the
beam. From (9.12.29), we find that u2 = 0 for elements initially lying in the
x3xx plane; as such, these elements continue to remain in the same plane
after deformation. The xxx3 plane is therefore referred to as the plane of
bending. We note that the line element initially lying along the x3 axis lies
on the neutral plane as well as on the plane of bending; this element is
called the central line of the beam. From (9.12.29) we find that the points
of the central line are displaced completely in the xx direction and that the
displacement obeys the equation
* i = ^ *

(9.12.30)

We note that the curve defined by the equation (9.12.30) is a parabola in


the xxx3 plane having the xx axis as the axis. Thus, the central line of the
beam deforms into a parabola given by the equation (9.12.30). The curvature R of this parabola is given by
d2ux
M
1
d2ux/dx\
2 3/2
R " [1 + (dux/dx3) ]
~ dx\ ~ El

(9.12.31)

where the nonlinear term {dux/dx3)2 is neglected because deformation is


small.
The result (9.12.31) is of great importance in the theory of bending of
beams, usually referred to as the Bernoulli-Euler law.
Next, consider a particle initially located on the section x3 = c. Let x\ be
the coordinates of the location of this particle after deformation, so that
x[ = Xi + (wf)*3 = c Then, we find from (9.12.29) and (9.12.31) that
x'3 = c- ^cxx = cjYl - jjX'A + ^(ux)X3

= cl

</l - ^'

(9.12.32)

Equation (9.12.32) represents a plane in the deformed configuration of


the beam, and it is easy to verify that this plane is normal to the deformed
central line. Thus, planes normal to the central line before deformation are
transferred to planes normal to the deformed central line. This result is also
of importance in the theory of bending of beams.
EXAMPLE 9.12.3 Suppose that the beam considered in the above
problem is rectangular with xx = h and x2 = b as the boundary lines of
a cross section. Find the deformed shape of the cross section.

9.12

I.

2h

f-

SOME STATIC PROBLEMS

41 7

2b

1/

1
\

If

X2

1
I

t
Tx,

Figure 9.8. Example 9.12.3.

Solution Figure 9.8 shows a typical section of the beam for which x3 is a
constant, say, c. Due to deformation, suppose the particle initially located
at the point (x,, x2) of this section gets displaced to the point (x|, x'^), where
x'i = x, + H, and x'2 = x2 + u2. Then (9.12.29) and (9.12.31) give
x[

=X

2R^2

Hx2i

-4))

(9.12.33)

*2 = ^ 2 ( 1 + ^ * 1

For the particle initially located on the linesx2 = b, equations (9.12.33) give
x2 = 6(1 + - x ,
= b

1 +

R\Xi

~2R(C2

~ ****
(9.12.34)

b[ 1 + -x

Thus, the lines x2 = b transform into lines given by equations (9.12.34),


which make angles tan~l(bv/R) with the vertical.
For the particle initially located on the lines x, = h, equations (9.12.33)
give
x[ = + + -L{c 2 + v(A2 -xf))
2R
V

c2 + vh2 - v(jc2)2(l + -x
R
2R
h + -{c2 + v/*2
2/?

/ 2

v(*D

(9.12.35)

418

EQUATIONS OF LINEAR ELASTICITY

Thus, the lines xx = h transform into curves given by equations


(9.12.35), which represent parabolas whose axes are directed opposite to the
jq axis. It is easy to verify that the curvature of both these parabolas is
approximately v/R.
Thus, the boundary lines xx = h and x2 = b of a cross section transform, respectively, to the oblique lines given by (9.12.34) and the parabolas
given by (9.12.35). The deformed shape of the section is shown by dotted
lines in Figure 9.8.

9.12.4

TORSION OF ELLIPTIC AND CIRCULAR BEAMS

Consider an elastic beam of length / and uniform elliptic or circular cross


section. Suppose that a couple (of moment) m about the axis of the beam
is applied at one of the bases and an opposite couple - m is applied at the
other base. The lateral surface is stress free and body forces are neglected.
The problem is to compute the displacements, strains and stresses developed
in the beam because of the twist (or torsion) it experiences due to the
applied couples; see Figure 9.9.
Let us choose the coordinate axes such that the origin is the centroid of
the base subjected to the couple - m and the x3 axis is along the axis of the
beam.
If s is the stress vector at a point of a cross section, then the total moment
of the surface force acting on the cross section is \A x x s dA. If this section
is the base x3 = 0, we have s = - s ( 3 ) and the total moment is equal to - m ;
thus,

\x(-s{3))dA

= -m

(9.12.36)

for x3 = 0. On the other hand, if the section is the base x3 = I we have


s = s(3) and the total moment is equal to m; thus,
x x s(3) dA = m
for x3 = /.

Figure 9.9. Torsion.

(9.12.37)

9.12

SOME STATIC PROBLEMS

419

By hypothesis, the applied couples act about the axis of the beam; hence,
we may take m = Me 3 , where M = |m|. Then (9.12.36) and (9.12.37) yield
the following set of conditions to be satisfied by the stress components:
JA

x2x33dA = 0,

JA

xxx33dA = 0,

(x2x33 - /T32) A4 = 0,

JA

JA

{xxx32 - x2x3l)dA = M

(/31 - xxx33)dA = 0

(9.12.38)

Since the beam is twisted by the applied couples, the cross sections
x3 = constant experience shear strains and hence shear stresses. Therefore,
31 and 32 , are not 0. If we take all the remaining stress components to be
identically 0, that is, if
Tn = 22 = 33 = 12 = 0

(9.12.39)

then the conditions (9.12.38) become


JA

31 dA = 0,

JA

x32dA = 0,

JA

(32 - x2T3l)dA = M

(9.12.40)

By hypothesis, the lateral surface of the beam is stress free. Hence


TijHj = 0 on this surface, for which n3 = 0. Suppose that the beam is elliptic
with the boundary of a cross section given by the equation:
^ +p=l

(9.12.41)

(The circular beam corresponds to the case b = a.) Then the slope of the
normal n to the boundary curve (9.12.41) (which lies parallel to the x1x2
plane) is
dxx _ a2x2
dx2 b2xi
Hence, the components nl and n2 of n are proportional to b2Xi and a2x2,
respectively. Consequently, the conditions x^rij = 0 to be satisfied on the
lateral surface reduce to the single condition
?23* + 0232*2 = 0

(9.12.42)

where we have made use of (9.12.39).


Since the body force is neglected, Cauchy's equilibrium equations
TJJJ = 0 must be satisfied at every point of the beam. In view of (9.12.39),
these equations take the following explicit forms:
L

13,3

= 0,

23>3 = 0,

T 3M + 32)2 = 0

(9.12.43)

420

EQUATIONS OF LINEAR ELASTICITY

The simplest solution of these equations that meets the condition


(9.12.42) is
31 = cx2,

32 = 2 cx x
a

(9.12.44)

where c is a nonzero constant.


Using (9.12.44), the conditions (9.12.40) become
\ x2dA = 0,

(b2xj + a2x\)dA =

xl dA = 0,

(9.12.45)

Bearing in mind that A is the cross-sectional area bounded by the ellipse


(9.12.41), we note that the integrals \Ax\dA and \Ax\dA denote the
moments of inertia I2 and Ix of the section about the x2 and xx axes,
respectively, with Ix = nab374 and I2 = na3b/4. As such, the last condition
in (9.12.45) becomes
(9.12.46)
c = -^-3
nab
Further, the first two conditions in (9.12.45) are identically satisfied.
Hence, if c is chosen as given by (9.12.46), then Cauchy's equilibrium equations as well as all the boundary conditions of the problem are satisfied by
the stress system given by (9.12.39) and (9.12.44). It is easy to verify that
this system also satisfies the Beltrami-Michell equations.
Thus, a system of stresses that meets all the requirements of the problem is

22

33

2M

12

(9.12.47)

_2M

The corresponding system of strains follow from Hooke's law (9.3.15)".


Thus, we get
M

(9.12.48)

The displacement components associated with these strains follow from


the strain-displacement relation (9.4.4). Thus we obtain (see equation
(5.8.46)),
fa2 - b2\
Ux = - 0 * 2 * 3 ,

2 = OOC\X$>

U3 = -Oil ~2

TJ )XlX2

(9.12.49)

9.12

SOME STATIC PROBLEMS

421

where
a2 + b2 M
>0
(9.12.50)
3^3
a =
a5b

Expressions (9.12.49) show that in the case of an elliptic beam (b 5* a)


there does occur a longitudinal displacement u3, in addition to the transverse displacements ul9u2. Since u3 depends on xx and x2 and is independent
of x3, cross sections are warped according to the law given by the last
relation in (9.12.49) and the warping is the same in all sections. We find that
the lines of warping are rectangular hyperbolas and that (because of
warping) the parts of a cross section for which xx and x2 have the same signs
get pulled toward the fixed base whereas the parts for which xx and x2 have
opposite signs get pushed away from this base.
If x[ and x2 are the coordinates of the new location P' of a particle
initially located at the point P = (, x2) in a cross section, we find by the
use of (9.12.49) that
(x[f + (xi)2 = (1 + cx2x2)(x2 + xl) x\ + x\

(9.12.51)

Thus, the particle experiences a pure rotation about the axis of the beam; if
0 is the angle of rotation, we find that (see Figure 9.10)
ux = x[ - xx = rcos(y + 0) - A*cos y -2
u2 = x2 - x2 = r sin(y + 0) - r sin y 0

(9.12.52)

Comparing these results with those given in (9.12.49), we find that


0 = ax3. Thus, the angle of rotation 0 in a section is proportional to the
distance of the section from the fixed base. The constant of proportionality,
namely, a, represents the angle of twist (rotation) per unit length. The
quantity (Af/), which represents the moment of the couple required to
produce a unit angle of twist per unit of length, is called the torsional
rigidity or the torsional stiffness of the beam, usually denoted D. From

P'Ui ,X2)

- - ^ P ( X l ,X2*

X1 .

Figure 9.10. Rotation in a cross section.

422

EQUATIONS OF LINEAR ELASTICITY

(9.12.50), we find that


.>0
(9.12.53)
a + b2
From (9.12.47), we find that the stress vector at a point P of a cross
section of the beam is given by
D =

2M
2
7
s = T31e! + T32e2 = 3T3 (-0 x2Ci + b ^2)
na o

(9.12.54)

If we set xi = ka cos y, x2 = kb sin y, then (9.12.54) becomes


2Mk
s = _ 22 2 (-et g sin y + e 2 6cosy)
na b

(9.12.55)

This is the stress vector at a point P (ak cos y, bk sin y) of the section. It
follows immediately that the point P lies on the ellipse

(9.12.56)
whose normal at P is along the vector
(9.12.57)
From (9.4.55) and (9.4.57), we readily find that s n = 0, so that s is along
the tangent to the ellipse (9.12.56). Thus, at a point P of a cross section, the
stress vector is tangential to the ellipse that passes through the point and is
concentric with and similar to the boundary curve of the section. In other
words, the family of ellipses determined by (9.12.56) for different values of
k are the lines (curves) of shear stress for the elliptic beam considered.
From (9.12.55), we find that the magnitude of s at P is given by
2Mk 2
W = ^
lb + (a2 - b2) sin 2 y} 1/2
na2b2

(9.12.58)

If a and b are the lengths of the semi-major axis and the semi-minor axis,
respectively, of the boundary curve (9.12.41), we find from (9.12.58) that
|s| is maximum when k is maximum (that is k = 1) and y = /2 or 3/2.
Thus, the magnitude of shear stress assumes a maximum value at the
endpoints of the minor axis of the boundary curve, and the maximum value
is given by
max|s| = -2
nab

(9.12.59)

9.12

SOME STATIC PROBLEMS

423

For a circular beam (b = a), we find from (9.12.49) that u3 = 0. Accordingly, the cross sections of the beam remain undisplaced from their original
position, and the beam experiences no warping along its length. There occur
only transverse displacements ux and u2 at a point, and these displacements
vary from one cross section to the other. From (9.12.53) and (9.12.59) we
find the following expressions for torsional rigidity and maximum shear
stress for a circular beam of radius a:
D = ^a4

(9.12.60)

2M
max|s| = 5
na

(9.12.61)

We note from (9.12.58) that for a circular beam there is no specific point on
the boundary at which |s| is maximum; max|s| given by (9.12.61) occurs at
every point of the boundary. Further, the lines of shear stress are circles
concentric with the boundary.
It may be pointed out that the torsion problem for a circular beam was
first solved by Coloumb in 1787. The theory of torsion for noncircular
beams was later developed by Saint-Venant in 1855. The Saint-Venant's
theory of torsion is one of the most celebrated works in the mathematical
theory of elasticity.
EXAMPLE 9.12.4 Show that the torsional rigidity of an elliptic beam
with cross sectional area A and polar moment of intertia Ie is given by
uA4
De = - P -

(9.12.62a)

If Ic is the polar moment of intertia of a circular beam having the same


cross-sectional area as the previous elliptic beam and Dc is its torsional
rigidity, show that
De = (Ic/Ie)Dc
(9.12.62b)
Solution For a beam whose cross section is bounded by the ellipse
(9.12.41), the area of cross section is A = nab and the polar moment of
inertia is
Ie = Il+

I2 = ^ab(a2 + b2)

(9.12.63)

Hence, (9.12.53) may be rewritten as


na3b3

4n3ab

A4

424

EQUATIONS OF LINEAR ELASTICITY

which is (9.12.62a). For a circular beam (b = a), this becomes

where

D = DC = ^ An Ic

(9.12.64)

= ^a4

(9.12.65)

Ic = (Ie)b

=a

is the polar moment of inertia of a circular beam and A is its cross-sectional


area.
For a given A, we readily find from (9.12.62a), (9.12.63), (9.12.64) and
(9.12.65) that
_ ^ 4 ^ _ / 7 \
D

which is (9.12.62b).

* - AnX

\lJDc

Note: From (9.12.63) and (9.12.65), we find that Ic< Ie. Consequently,
from (9.12.62b) it follows that De< Dc. Thus, a circular beam has greater
torsional stiffness than an elliptic beam with same cross-sectional area.
9.12.5

CYLINDRICAL TUBE UNDER PRESSURE

Consider a long, straight elastic circular tube of inner radius a and outer
radius b. Suppose that a uniform pressure p acts on the inner surface and
a uniform pressure q(^p) acts on the outer surface. Body forces are
ignored. The problem is to determine the displacements and stresses caused
at an arbitrary point of the tube.
We note that the pressures p and q act completely along the radial
direction of the circular boundary surfaces. Also, since the tube is long, the
effects of surface forces acting on the ends of the tube (on the deformation
in a cross section) can be neglected. Further, the body forces are ignored. In
view of these facts, we naturally assume that (i) the particles get displaced
completely along the radial direction, (ii) the amount of displacement
depends solely on the distance of the particle from the central line (axis)
of the tube and (iii) the same kind of deformation occurs in every section of
the tube.
Let us choose the coordinate axes such that the xx x2 plane coincides with
a section of the tube and the JC3 axis lies along the axis of the tube. Then, in
view of these assumptions, the displacement field at any point of the tube
may be taken as
u = u(R)eR
(9.12.66)

9.12

SOME STATIC PROBLEMS

425

Figure 9.11. Pressures on a cylindrical tube.

where eR is the unit vector along the radial vector j^ej + Jt2e2, R2 =
x\ + JC| and u = u(R) is a function to be determined.
We note that the displacement field given by (9.12.66) is parallel to the
x1x2 plane and independent of x3. A problem in which this kind of displacement arises is called a plane strain problem. Such problems, together with
the so-called plane stress problems, belong to the area of plane elastostatics.
Returning to the problem under consideration, we note that the boundary
surfaces of the tube are R = a and R = b, where a and b are the inner and
outer radii of the tube, and that these surfaces are subjected to uniform
pressures p and q\ see Figure 9.11. Hence, if aR is the normal stress on a
cylindrical surface concentric with the tube, the boundary conditions for the
problem are
-p for R = a
(9.12.67)
-q for R = b

In order that u taken as in (9.12.66) may be a solution of the problem, it


must obey Navier's equation of equilibrium with zero body force and the
boundary conditions (9.12.67). In Example 9.8.6, it was shown that u as
given by (9.12.66) obeys Navier's equation of equilibrium with zero body
force if u(R) is as given by (9.8.51); that is,

u = AR

(9.12.68)

where A and B are arbitrary constants. These constants are now to be


determined by using the boundary conditions (9.12.67). In Example 9.8.6,

426

9 EQUATIONS OF LINEAR ELASTICITY

aR has been obtained in terms of u(R) as given by (9.8.47); that is,


** = | + ( + 2 / / ) ^

(9.12.69)

Substituting for u from (9.12.68) in (9.12.69) we get


2
= 2(+)--
R2

(9.12.70)

The boundary conditions (9.12.67) now yield the following simultaneous


equations for A and B:

l
(+)--^ - = --

(+)-^

(9.12.71)

= --

Solving these equations, we get


a2P - b2g
(* + AM = ^772dr.
2(b2-a2)'

a2b\p - q)

B~ = jr22-22)

(91272)

With these values of A and B, (9.12.68) and (9.12.66) give the displacement vector
2lp - b2q
a2b\p -q) ,

f a

W+

)* - )

e
+ ^-7fT:
2 - ) R

(9.12.73)

which satisfies the governing differential equation as well as the boundary


conditions of the problem. By virtue of the uniqueness of solution, this u is
unique. Thus, (9.12.73) gives the displacement that occurs at a point distant
R from the axis of the tube.
Using (9.12.72) in (9.12.70) we obtain
a2p - b2q

a2b\p - q) 1

OR = bii2-a2 2 - - 7b22-a^2 2 ^ R~
si2

(9 1 2 7 4 )

This is the expression for the normal stress acting on a cylindrical surface
R = constant in the tube.
The normal stress ah acting on an axial plane can be computed by use of
expression (9.8.48) together with (9.12.68) and (9.12.72). Thus, we obtain
_ a2p - b2q
- b2-a2
h

a2b\p - q) 1
b2-a2
R~2

(9 12 75)

9.12

SOME STATIC PROBLEMS

427

Let us now analyze the nature of aR and ah in the particular case where
the external pressure q is absent. In this case, (9.12.74) and (9.12.75)
become
a2p
aR = b2-a2
a2p
b -a2

R2

(9.12.76)

(9.12.77)

Since a < b and R < 6, we readily find that aR < 0 and > 0. This
means that every point of the tube experiences a compressive radial stress
and a tensile peripheral stress. Also, the peripheral stress ah is maximum on
the inner surface of the tube, with
maxa A

p(a2 + b2)
= b2-a2

(9.12.78)

If the thickness (b - a) of the tube is very small, (9.12.78) becomes


Pb
maxa A b- a

(9.12.79)

Thus, for a given pressure /?, max varies inversely as the thickness of the
tube.
Two other particular cases of the problem are also of interest. If b -> oo,
the tube becomes an unbounded body with a cylindrical cavity. Then
(9.12.73) becomes
U

_^ [a2p
[iR

q(

+*

\)
MRJ)R

(9.12.80)

It is evident that for u to have finite magnitude for all R, q should be 0.


Then we get
a2p
(9.12.81)
The corresponding expressions for aR and ah follow from (9.12.76) and
(9.12.77) with b - oo; the results are
R

= ~^^2'

(9.12.82)

We find that these aR and ah have the same magnitude at every point of the
body. Also, as R - oo, u, aR and ah all tend to 0.

428

EQUATIONS OF LINEAR ELASTICITY

In the case when a - 0 with b remaining finite, the tube becomes a full
cylinder (with no hole). Then (9.12.73), (9.12.74) and (9.12.75) become
(taking p to be bounded),
u =
R =

ReR
2(A + )
h = -Q

(9.12.83)

We find that oR and ah are now everywhere equal to the applied pressure.
Also, the magnitude of displacement varies directly with R and is maximum
on the surface R = b.

9.12.6 SPHERICAL SHELL UNDER PRESSURE


Consider an elastic spherical shell of inner radius a and outer radius b.
Suppose that a uniform pressure p acts on the inner surface and a uniform
pressure q(j*p) acts on the outer surface. Body forces are ignored. The
problem is to determine the displacements and stresses caused at an
arbitrary point of the shell.
We note that the pressures p and q act completely along the radial direction from the center of the shell. Further, body forces are ignored. Hence,
we assume that (i) the particles get displaced completely along the radial
direction, and (ii) the amount of displacement depends solely on the
distance of the particle from the center of the shell. Then, if we choose the
coordinate axes with the origin at the center of the shell, the displacement
vector at any point of the shell may be taken as
u = u(r)er

(9.12.84)

where er is the unit vector along the radial direction and u{r) is a function
to be determined.
The surfaces r = a and r = b are subjected to uniform pressures p and q,
respectively, where a and b are the inner and outer radii of the shell; see
Figure 9.12. Hence, if ar is the normal stress on a spherical surface
r = constant in the shell, the boundary conditions for the problem are
ar=

C-p

for

r = a

l^-q

for

r= b

(9.12.85)

In order that u taken as in (9.12.84) may be a solution of the problem, it


must satisfy Navier's equation of equilibrium with zero body force and the
boundary conditions (9.12.85). In Example 9.8.5, it was shown that u given
by (9.12.84) obeys Navier's equation of equilibrium with zero body force if

9.12

SOME STATIC PROBLEMS

429

Figure 9.12. Spherical shell under pressure.

u(r) is as given by (9.8.41); that is,


u(r) = -^ + Br
r

(9.12.86)

where A and B are arbitrary constants. These constants are now to be


determined by using the boundary conditions (9.12.85). In Example 9.8.5,
ar was obtained in terms of u(r) as given by (9.8.36); that is,
du
= 2- u(r) + (A + 2): ,
r
dr

(9.12.87)

Substituting for u from (9.12.86) in (9.12.87), we get


ar = -4-^

+ (3A + 2)

(9.12.88)

The boundary conditions (9.12.85) now yield the following simultaneous


equations for A and B:
4 - (3 + 2) = p
(9.12.89)
^-^ - (3 + 2) = g
Solving these equations, we get
alb\p
4=, 5
b

- q)
3

-a

pa3 - ab3
(3 + 2) = *,3 _ 3

(9.12.90)

430

EQUATIONS OF LINEAR ELASTICITY

With these values of A and B, (9.12.84) and (9.12.86) give the displacement vector
u = b3

a3b\p

1
-a3

- q)

4//rz

{pa3 - qb3)r
er
(3A + 2)

(9.12.91)

which satisfies the governing differential equation as well as the boundary


conditions of the problem. By virtue of the uniqueness of solution, this u is
unique. Thus, (9.12.91) gives the displacement that occurs at a point distant
r from the center of the shell.
By use of (9.12.90) in (9.12.88) we obtain
or =

{pa3 - qb3) -"-^{p

- q)

(9.12.92)

This is the expression for the normal stress acting on a spherical surface
r = constant in the shell.
The normal stress acting on a radial plane can be computed with the
use of expressions (9.3.37), (9.12.86) and (9.12.90). Thus, we obtain
* =

a'b\p-q)

+ {pa3 - qb3)

2r3

(9.12.93)

Let us now analyze the nature of ar and ah in the particular case where the
external pressure q is absent. In this case, (9.12.92) and (9.12.93) become
=

pa*
pa

-
b

l+

(9.12.94)
'~

2?

(9.12.95)

Since a < b and r < b, we readily find that ar < 0 and ah > 0. This
means that every point of the shell experiences a compressive radial stress
and a tensile peripheral stress. Also, the peripheral stress ah is maximum on
the inner surface of the shell, with
p(2a3 + b3)
max<7A = 2(b3 - a3)

(9.12.96)

If the thickness (b - a) of the shell is very small, (9.12.96) becomes


Pb
maxa A 2(0 - a)

(9.12.97)

Thus, for a given pressure, max ah varies inversely as the thickness of the
shell.

9.13

ELASTIC WAVES

431

Two other particular cases of the problem are also of interest. If b - oo,
the shell becomes an unbounded body with a spherical cavity. Then
(9.12.91) becomes
u=

a (P - 0)
4

qr
er
3 + 2

(9.12.98)

It is evident that, for u to have finite magnitude for all r, q should be 0.


Then we get
a3p
* = *
(9.12.99)
The corresponding values for ar and oh follow from (9.12.94) and (9.12.95)
with b -* oo ; they are
pa3
or=-^T>

pa3
oh = ^j-

(9.12.100)

We find that these ar and ah have the same magnitude at every point of
the body. Also, as r - oo, u, ar and ah all tend to 0.
When a - 0 with b remaining finite, the shell becomes & full sphere (with
no hole). Then (9.12.91), (9.12.92) and (9.12.93) become (taking/? to be
bounded)
=

qr

3 + 2

Or = Oh =

(9.12.101)

-Q

We find that ar and oh are now everywhere equal to the applied pressure.
Also, the magnitude of displacement varies directly with r and is maximum
on the surface r = b.
The problems on the cylindrical tube and the spherical shell subjected to
unequal pressures were first considered by Lam in 1852. These problems
are referred to as Lame's pressure-vessel problems.

9.13
ELASTIC WAVES
As indicated earlier, in the displacement formulation, a dynamical problem
in elasticity is governed by Navier's equation (9.8.6). A simple inspection of
this equation reveals that it is a hyperbolic-type partial differential equation
of the second order. A solution of this equation represents a wave motion.
In the stress formulation of an elastodynamic problem, the stress equation

432

EQUATIONS OF LINEAR ELASTICITY

of motion (9.10.5) or the Beltrami-Michell equation (9.11.6) serves as the


governing equation; these are also hyperbolic partial differential equations.
Therefore, the solutions of these equations also represent wave motions.
Accordingly, every motion of an elastic body occurs in the form of a wave
(called an elastic wave), and every dynamical problem in elasticity deals
with the study of one or another type of elastic waves. Some simple elastic
wave propagation problems are discussed in this section.
9.13.1

STRESS WAVES IN A SEMI-INFINITE BEAM

Consider a thin semi-infinite elastic beam that is initially at rest in an


undeformed state. Suppose that at time t = 0+ a time-dependent pressure
p(t) is applied to the end of the beam along the length, and this pressure is
maintained for all subsequent times; see Figure 9.13. The body forces are
ignored. The problem is to determine the stress and displacement that occur
at an arbitrary point of the beam at any subsequent time /.
Let us choose the xx axis along the axis of the beam with the origin at the
initial position of the end where the pressure is applied. Then the boundary
condition to be satisfied is
ti, = -Pit)
(9.13.1)
for xx = 0, / > 0.
It is assumed that the beam is thin. Also, the load that causes deformation
acts along the length of the beam. Hence, we suppose that at any point xx
of the beam there occurs only the longitudinal stress and that this is a
function of x1 and /. Then the stress equation of motion (9.10.5) yields the
following equation for :
a

where

2d

2r

i i _ d2Tn

dx2 * dt2
a

E
=

(9.13.2)
(9.13.3)

We have ignored the body forces.


Equation (9.13.2) is the governing partial differential equation for the
problem. We note that this is a one-dimensional wave equation with
a representing the speed of propagation. We therefore infer that the

Figure 9.13. Stress on a beam.

9.13

ELASTIC WAVES

433

distribution of the stress occurs in the form of a wave propagating with


speed a = \fE7p along the length of the beam. Such a wave is called a
stress wave.
Since the beam is initially at rest in an undeformed state, the following
initial conditions hold:
du
u = = 0
v(9.13.4)
dt
'
at / = 0 and xx > 0. Consequently, we get, on using the stress-displacement
relation (9.8.4), the following conditions for :

Tll =

"f= 0

(9 13 5)

'

at / = 0 and xx > 0.
Thus, (9.13.2) is the governing wave equation, (9.13.5) are the initial
conditions and (9.13.1) is the boundary condition for the problem. Note
that all of these are expressed in terms of the stress component . This is
the stress formulation of the problem.
To solve the problem, we change the independent variables from xl and
t to = t - (/) and = t + (X\/a). Equation (9.13.2) then becomes
^ = 0
<9136>
33
Integration of this equation yields the following general solution for :
= / ) + *(*)

(9.13.7)

where /() and g(//) are arbitrary functions. This solution is usually called
D'Alembert's solution.
Using (9.13.7), the initial conditions (9.13.5) take the form
f

n-xi/a)

+ g'(xl/a) = 0)

(9.13.8)

for xx > 0. These conditions are satisfied if we choose /() and g(ji) such
that
g(/7) = A for all
*)
(9.13.9)
) = -
for < 0 j
where A is an arbitrary constant. Then the solution (9.13.7) becomes
0
\A + / )

for

<0

for

>0

434

EQUATIONS OF LINEAR ELASTICITY

or
(9.13.10)

where F(tt - xx/a) = A + f(t - xx/a) is an arbitrary function of (t - xx/a).


The boundary condition (9.13.1) is satisfied if we choose F(t - xx/a) =
-p(t - xx/a). Thus, a solution for that satisfies the governing equation
(9.13.2) and the conditions (9.13.5), and (9.13.1) is

(9.13.11)

From this solution, we note that at any chosen point xx of the beam, no
stress occurs until the time t = xx/a, and a time-dependent compressive
stress -p(t - xx/a) occurs thereafter. This stress is due to the stress wave
that starts from the end xx = 0 at t = 0 and arrives at the point xx at time
f= XX/OL. The speed of the wave is a = \lE/p as already indicated.
The longitudinal displacement ux associated with given by (9.13.11)
can be computed by using Hooke's law (9.3.3) and the strain-displacement
relation exx = uxx. Thus we get

(9.13.12)

This result shows that the


a point xx at time t > xx/a
the pressure curve over the
proportionality being a/E =
From (9.13.12), we get

displacement ux caused by the stress wave at


is directly proportional to the area under
time interval (0, t - xx/a), the constant of
(Ep)~l/2.

^ W ) = />('-^)

dt

(9.13.13)

for t > xx/a. This gives the speed at a point xx of the beam at time
t(>xx/a). From (9.13.11), (9.13.13) and (9.13.3) we get, for t > x/a9
' / ( * ) - !

JEp

(9.13.14)

9.13

ELASTIC WAVES

435

We note that the quantity Tn/(dux/dt) represents the stress required to


generate a speed of unit magnitude in the beam, called the mechanical
impedance or wave resistance factor of the beam. From (9.13.14), we
observe that this quantity is a constant (equal to -yfp) depending on the
physical property of the beam.
9.13.2

PLANE WAVES IN A HALF-SPACE

Consider a semi-infinite elastic body (half-space) with plane boundary,


which is initially at rest in an undeformed state. Suppose that at time t = 0 +
a time-dependent pressure p(t) is applied to the boundary, and this pressure
is maintained for all subsequent times. Body forces are ignored. The
problem is to determine the displacement and stress that occur at an
arbitrary point of the body at any subsequent time t.
Let us choose the axes such that the origin lies on the boundary of
the half-space in its initial position with the xx axis perpendicular to the
boundary and directed into the half-space. The geometrical space filled by
the body is then xx > 0 with the plane xx = 0 as its boundary; see Figure
9.14. It is assumed that the deformation of the half-space is caused by the
pressure p(t) acting on the plane xl = 0 along the xx axis for t > 0. Hence
the boundary condition for the problem is
Tu = -Pit)

(9.13.15)

at xl = 0 for t > 0.
In view of this boundary condition, we assume that the displacement due
to deformation is directed along the xx axis and that it depends only on xx,

p(t)-

Figure 9.14. Half-space xx > 0.

436

EQUATIONS OF LINEAR ELASTICITY

apart from /. Thus, we seek a solution in the form


u = ^,!

(9.13.16)

for xi > 0, t > 0; where Ui(xi9t)isto be determined.


Since the body is initially at rest in an undeformed state, the following
initial conditions hold:
"i=^

=0

(9.13.17)

for xl > 0 at t = 0. By using (9.13.16) in the stress-displacement relation


(9.8.4), we get

= (

12

23

22 = 33 =

2),

31

(9.13.18)

Using the first of these expressions, the condition (9.13.15) can be rewritten
as
^i(^i,0 = - r - ^ r P ( 0
+ 2
dxl

(9.13.19)

for JC! = 0, / > 0.

Thus, (9.13.17) and (9.13.19), expressed in terms of ux, are the initial and
boundary conditions for the problem.
Since the problem is being formulated in terms of the displacement, the
governing equation is Navier's equation (9.8.6). In view of (9.13.16),
Navier's equation becomes
c -
(9.13.20)
1
v
bx\
dt2
'
where cx = [(A -l- 2)/]1/2 as usual. We have ignored the body forces.
We note that (9.13.20) is the one-dimensional wave equation with Cj as
the speed of propagation. Thus, under the assumptions made, the distribution of the displacement in the half-space occurs in the form of a plane wave
propagating in the xx direction with speed cY.
As in the case of equation (9.13.2), we find the D'Alembert's solution for
equation (9.13.20) as

" , _ / ('-i) + *(' + s)

<9 2)

"

for *! > 0, / > 0, where the functions / and g are to be determined. Using

9.13

ELASTIC WAVES

437

(9.13.21) in the initial conditions (9.13.17) we get the equations

for xx > 0, which are satisfied if we choose/ and g as follows;

H)=A
f

KcJ

Xi

for

t> 0,

= A for

x,>0

(9.13.22)

-i

where ^4 is an arbitrary constant. Then (9.13.21) reduces to the following


form:
for

*i
t<

1
X

F\t- -\

for

(9.13.23)

/>ii

-1/

<"l

where F(t - xx/cx) = A + f(t - X\/cx) is an arbitrary function. Hence


for

-rM' cj

for

t<
(9.13.24)

'>r

Using this, the boundary condition (9.13.19) becomes

a>-^<--rTV<'>

for > 0. This is satisfied if we take

K'-$-fwr",'w*

<9 13 251

Putting this expression for F(t - xx/cx) back into (9.13.23), we get the

438

EQUATIONS OF LINEAR ELASTICITY

following solution for M, that meets all the requirements of the problem:
0

for

i(*i,0= {

Cl

n x / c

73VT

isii

P('O)AO

for

(9.13.26)

/>-!

(A + 2//) Jo
Cj
The solution (9.13.26) shows that at any chosen point xx of the half-space
no displacement occurs until time f= xx/cx and a longitudinal displacement
occurs thereafter. The magnitude of this displacement is directly proportional to the area under the pressure curve over the time interval
(0, / - xx/cx), the constant of proportionality being [( + 2)]~/2.
For t > xx/cx, (9.13.18) and (9.13.26) yield the following expressions for
the normal stresses developed in the half-space:

= -/>('-I

(9.13.27)

T = T =

" ^' " %

We note that all these stresses are compressive in nature.


From (9.13.26), we also find that, for t > xl/cl,
dui,
dt

cx
(
( + 2) \

x\
cj

(9.13.28)

This gives the speed at a point xx and at time t > xl/cl. From (9.13.27) and
(9.13.28), we get the following expression for the wave resistance of the
half-space:
dt J

c,

(9.13.29)

REMARK The two problems just discussed are analogous to one


another. While the first was formulated in terms of stress, the second was
formulated in terms of displacement.
9.13.3

PLANE WAVES IN AN UNBOUNDED ELASTIC BODY

Suppose that an unbounded elastic body experiences a displacement field


that depends only on xx and t. The body forces are neglected. The problem
is (i) to find the cause for the occurrence of such a displacement, and (ii) to
analyze the associated physical phenomenon.

9.13

ELASTIC WAVES

439

When u = u(xl, / ) , Navier's equation of motion (9.8.6) yields the


following three equations if the body forces are neglected.
2d

ux _ d2ux

dxj

dt2

d2u2

d2u2

dx\
Cl

dt2

d2u3
d2u3
dx2 - dt2

(9.13.30a)
(9.13.30b)
(9.13.30c)

where
c\ = ( + 2)/,

c\ = /

(9.13.31)

Therefore, in order that the displacement is of the form u = u(xx, /) the


components w, of u must obey the equations (9.13.30). We observe that each
of the three equations in (9.13.30) is a one-dimensional wave equation
governing a plane wave propagating in the xx direction. We note that ux is
governed by the wave equation associated with the speed cx, and u2 and u3
are both governed by the wave equation associated with the speed c2. Thus, a
displacement field of the form u = u(Xj, /) occurs due to the propagation of
three plane waves, of which one moves with speed cx and the other two move
with speed c2, Since the body is unbounded, there are no boundary effects.
We note that ux is the component of u taken along the xx direction, which
is the direction of propagation. Hence, the wave associated with ux (which
propagates with speed cx) is called the longitudinal wave. When this wave is
not absent (that is, when ux 5* 0) we find that div u 5* 0; as such, the wave
causes a change in volume of the material elements in the body and is
therefore also referred to as the dilatational wave. If u2 = u3 = 0, the
deformation is caused fully by this wave, and we find that curl u = 0; hence
the motion is irrotational so that no change occurs in the shape of material
elements.
Next, we note that u2 and w3 are the components of u taken along
directions perpendicular to the direction of propagation. Hence the waves
associated with u2 and u3 (which propagate with speed c2) are called the
transverse waves. When either of these waves is not absent we find that
curl u 5* 0; as such, these waves produce a change in shape of the material
elements and are therefore also referred to as shear waves. If ux = 0, the
deformation is caused fully by these waves, and we find that div u = 0;
hence the motion is isochoric so that no change occurs in the volume of
material elements.
Expressions (9.13.31) show that cx > c2. As such, the dilatational wave
propagates faster than the shear waves. The dilatational wave is therefore

440

EQUATIONS OF LINEAR ELASTICITY

also called the primary wave or the P-wave, and the shear waves the
secondary waves or the S-waves. It is customary to take the x3 axis in the
vertical direction and the x2 axis in a horizontal direction. Then, the S-wave
with displacement u2 is referred to as the secondary horizontal wave or the
SH-wave, and the S-wave with displacement u3 is referred to as the
secondary vertical wave or the SV-wave. For a typical metal like copper, the
speeds of primary and secondary waves are estimated as cx = 4.36 x 103 m/s
and c2 = 2.13 x 10 3 m/s, respectively.
We note for a plane wave propagating in the xl direction, a function
(, t) associated with the wave is a solution of an equation in the form
***-**

<913 32)

A particular form of that is periodic in both xl and t is


= A exp{i(t - kxx)\

(9.13.33)

where A, and k are real constants. Evidently, the period of in t is 2/;


this is referred to as the period of the wave. The reciprocal of the period,
namely, /2, is called the frequency of the wave; the number is often
referred to as the angular frequency. Also, the period of in xx is 2n/k,
referred to as the wavelength. The reciprocal of the wavelength, namely,
k/2n, is called the wave number. The number /k is called the phase speed
of the wave. The maximum value A of is called the wave amplitude.
Substitution of (9.13.33) into (9.13.32) yields c2 = 2/k2. Accordingly, a
solution of (9.13.32) that corresponds to a wave propagating with frequency
/2 and speed c is
= A exp

K-?)

= A cos [ t

) + / sin \ t

(9.13.34)

This solution is just a particular form of the D'Alembert's solution of the


wave equation. A wave associated with this type of solution is referred to as
a plane harmonic wave.
In view of these remarks, we may note that when the P-wave, governed
by equation (9.13.30a), is a harmonic wave, the associated displacement is

K'-2)}

ux = 1 {/ f -

(9.13.35)

9.13

ELASTIC WAVES

441

where is a constant. The corresponding expressions for S-waves are


(u29 II3) = ( i 4 2 M 3 ) e x p W / - ^ J]

(9.13.36)

where A2,A3 are constants. Consequently, the displacement field u(xl9 /)


determined by (9.13.35) and (9.13.36) corresponds to a plane harmonic
wave propagating with frequency /2.
9.13.4

RAYLEIGH WAVES

Consider an elastic half-space with a stress-free plane boundary in which a


plane harmonic wave whose amplitude diminishes with the distance from
the boundary propagates in a direction parallel to the boundary. The body
forces are ignored. The problem is to find the speed of propagation of the
wave.
Let us choose the axes such that (i) the half-space initially fills the region
JC3 > 0 with x3 = 0 as the boundary, and (ii) the xx axis is along the direction
of wave propagation; see Figure 9.15. Then, if /2 is the frequency and
c is the speed of propagation, we seek the displacement associated with the
wave in the form
u = ae

-axi

exp

M-?)}

(9.13.37)

where a is a constant vector and a is a positive number, so that the


amplitude ae~a*3 diminishes with increasing x3. The vector a and the speed
of propagation c are to be determined by using the fact that u must be a
solution of Navier's equation of motion and that the boundary x3 = 0 is
stress free.

Figure 9.15. Half-space x3 > 0.

442

EQUATIONS OF LINEAR ELASTICITY

From (9.13.37) we obtain the following expressions for the components


of (V div u) and D 2 u:
[V div\ = -\ a\ - ia a3\ exp j\ t

I\

[V div u] 2 = 0
[V div u] 3 = [ioL ax + a2a3 j exp j/ /
r^

3z\r

, i l

j(
2

akexp\i(olt

- -^j j

Substituting these expressions into the component form of Navier's


equation (9.8.10) with f = 0 and using (9.8.16), we get the following three
equations for the three components at of a:

\cW + 2{\ - }\ +

ia(c\ - c f ) - e 3 = 0
c

cL2aL + [ 1 /

(M - cfyct i +

(9.13.38)

a2 = 0

(9.13.39)

cfa 2 + ( 1 - - j

3 = 0

(9.13.40)

We find that (9.13.39) is automatically satisfied if we take a2 = 0. Then, for


u to be nonzero, al and a3 cannot both be 0; that is, equations (9.13.38) and
(9.13.40) should have a nontrivial solution. A necessary and sufficient
condition for this is
+ 2(

cW

(c2 - c2)ia

- ^)

(c - cl)ia

l f *

C\OL2 + 2[ 1 -

= 0

(9.13.41)

which simplifies to

Fl-3

*-H

= 0

(9.13.41)'

9.13

ELASTIC WAVES

443

Thus, the two possible values of a which make a, and a3 both not 0 are
(9.13.42)
For a = j, expression (9.13.40) yields (a3/ax) = -(c/)ial9 so that we
may take a3 = alA and UTJ = (/c)iA, where /I is an arbitrary real constant. Thus, for a = ai9 ax = (co/c)iA, a2 = 0, a3 = OLXA, expression
(9.13.37) satisfies Navier's equation (9.8.10) with f = 0. In other words
u = (iel + a^Ae-^'apUU

- J

(9.13.43)

is a solution of the equation (9.8.10) with f = 0.


For a = a2, (9.13.38) yields (a3c/a>) = -(ia1/a2), so that we may take
a3 = (co/c)B and ax = ia2B, where B is an arbitrary real constant. Thus, for
a = a2, ax = ia2B, a2 = 0, a3 = (co/c)B, expression (9.13.37) satisfies
Navier's equation (9.8.10) with f = 0. Accordingly,

fB)}

u = ( ia2*i + ~ e 3 )Be~a*x* exp j ( / - ^ ) (

(9.13.44)

is another solution of the equation (9.8.10) with f = 0.


A general solution for u associated with the wave under consideration is
obtained by superposing the two solutions (9.13.43) and (9.13.44). We thus
obtain the following solutions for the components of u:
ux = i[ -Ae~aiX> + a2Be-a2X*)exp
u2 = 0
M3 = [axAe~a^

*)

(9.13.45)

) exp K' - ?)]

+ -Be~aiXi

The stresses associated with these displacement components follow by


the use of the stress-displacement relation (9.8.4). We find that
31 = -fr {laiAe-a^

+ (l -

\^Be~^A

32 = 0

(9.13.46)

*~<.!((*- >-***where we have suppressed the factor exp{/cc;(l - /c)} for brevity.

444

EQUATIONS OF LINEAR ELASTICITY

Since the boundary x3 = 0 is stress free, we have 31 = 32 = 33 = 0 for


x3 = 0. By use of (9.13.46), these boundary conditions yield the following
equations:
c2\
-B = 0
2axA + 2 c
(9.13.47)

A + 2a B = 0
2 2

From (9.13.45) we note that for u to be nonzero, A and B cannot vanish


together. Hence the two equations for A and B given by (9.13.47) should
have a nontrivial solution. A necessary and sufficient condition for this is
2 -

2OL x

(9.13.48)

= 0

20Li

With the aid of expressions (9.13.42), this condition reduces to


2 -

~2\2

2 \ 1/2

= 41 1 -

,2\ 1/2

c\

(9.13.49)

Thus, under the assumptions made, this is a necessary and sufficient


condition for the solution of the type (9.13.45) to exist. This condition
contains c as the only unknown; as such the condition serves as the equation
for the determination of c. Since ax and a2 are to be positive, we find from
expressions (9.13.42) that c has to be less than c2, which is less than cx. Of
course, c has to be positive as well. Thus, a possible speed of the wave being
considered is a root of the equation (9.13.49) that lies in the interval (0, c2).
Equation (9.13.49) can be simplified to the form
p3 - Sp2 + (24 - \6q2)p - 16(1 - q2) = 0

(9.13.50)

where
p = C-r and q = - < 1
(9.13.51)
ci
cx
If we denote the left side of (9.13.50) by /(/?), it is easily seen that
/(0) = -16(1 - q2) < 0 a n d / ( l ) = 1 > 0. Hence, equation (9.13.50) has
one or three real roots for/? between 0 and 1. We verify that/"(/?) = 6p - 16
does not vanish for/? lying between 0 and 1. Hence (9.13.50) has exactly one
root for/? between 0 and 1. Equivalently, equation (9.13.49) has exactly
one root, say cR, for c between 0 and c2. Hence, the wave being considered
does exist and propagates with only one possible speed cR.

9.13

ELASTIC WAVES

445

We note that equation (9.13.49) determining cR, contains no term


involving . As such, the speed cR of the wave is independent of the
frequency. Further, since q2 = {c\/c\) = (1 - 2v)/2(l - v) [see (9.8.16)],
cR depends entirely on the shear wave speed c2 and Poisson's ratio. For a
material with v = 0.25 (such as rock), equation (9.13.50) becomes
3/73 - 24/72 + 56/7 - 32 = 0

(9.13.52)

whose roots are/7 = 4, 2 2/V. Of these roots, only/7 = 2 - 2/VJis less


than 1, and this yields cR = 0.9194c2 as the speed of the wave.
REMARK A wave in a semi-infinite body whose amplitude diminishes
with the distance from the boundary surface of the body is called a surface
wave or Rayleigh wave. Studies on such waves are of practical importance
in areas like seismology and geophysics. Lord Rayleigh first initiated the
study of surface waves by discussing the problem just considered in 1885.
It may be noted that the displacement vector associated with a Rayleigh
wave is made up of two components: one along the direction of propagation
and the other perpendicular to the boundary surface; see (9.13.45). Hence
(if the boundary surface is taken as a horizontal surface), a Rayleigh wave
is a coupled combination of the P-wave and the SV-wave.
EXAMPLE 9 . 1 3 . 1 Show that a surface wave of the type just considered
cannot exist if w3 = 0 and 31 = 0 on the boundary x3 = 0. (Such a
boundary is called a rigid, lubricated boundary.)
Solution In view of the last relation in (9.13.45), the boundary condition
w3 = 0 for x3 = 0 becomes
ctxA +-B
=0
(9.13.53)
c
Also, the boundary condition 31 = 0 for x 3 = 0 is given by the first
relation in (9.13.47), namely,
2a{A + (l - %\ -B

=0

(9.13.54)

For the waves to exist, A and B involved in (9.13.53) and (9.13.54) cannot
both be 0. A criterion for this is

*. H)?

446

EQUATIONS OF LINEAR ELASTICITY

which simplifies to
axc = 0

(9.13.55)

This condition is satisfied only if ax = 0, because and c are not 0 when


there is a wave propagation. But, for a wave of the surface type, <xx > 0.
Hence, the condition (9.13.55) cannot be satisfied by a surface wave; such
a wave therefore cannot exist under the assumed conditions.

EXAMPLE 9.13.2 Show that an S//-wave of the surface type cannot


exist in a half-space with a stress-free horizontal boundary.
Solution Suppose that the half-space initially fills the region A:3 > 0 with
x3 = 0 as the horizontal boundary. Since the boundary is stress free, we
have 31 = 32 = 33 = 0 for x3 = 0.
If there is a pure S//-wave of surface type propagating in the xx direction,
the corresponding displacement vector is given by u = w2e2, where
u2 = Ae-aXi explico(t - xx/c)}

(9.13.56)

a > 0 and A is a nonzero constant. Also, /2 is the frequency and c is the


speed of propagation.
For u = u2e2, with u2 given by (9.13.56), the stress-displacement relation
(9.8.4) gives 31 = 33 = 0 and
32 = -aAe~aXi Qxp{i(t - xx/c)}

(9.13.57)

Hence the only boundary condition to be satisfied on x3 = 0 is 32 = 0,


which on using (9.13.57) becomes = 0. This condition cannot be satisfied, because all of//, a and A are nonzero. Hence the wave considered is
not a possible one.

9.13.5

LOVE WAVES

Consider an elastic half-space with horizontal plane boundary, covered by


an elastic layer of uniform thickness. Suppose that (i) the layer and the
half-space have different densities and different shear moduli, (ii) there is a
welded contact between the layer and the half-space and (iii) the horizontal
boundary of the composite structure is stress free. The problem is to
investigate the possibility of the propagation of an 5//-wave of the surface
type in this composite structure. Body forces are ignored.
Let us choose the axes such that the half-space initially fills the region
x3 > 0 and the layer the space -H < JC3 < 0, so that the horizontal plane
boundary of the composite structure is x3 = -H\ see Figure 9.16. Also,
let the xx axis be chosen along the direction of propagation. Then for an

9.13

ELASTIC WAVES

447

Figure 9.16. Half-space covered by a layer.

S//-wave of surface type we take u = w2e2, where


u2 = Ae-ax> exp[i(t - xx/c)\

(9.13.58)

in 0 < x3 < oo; and


u2 = f(x3) exp[ico(t - Xi/c)]

(9.13.59)

in -H < x3 < 0. Then /2 is the frequency and c is the speed of propagation of the wave (if it exists). The unknowns A, a(> 0) and/(x 3 ) are to
be determined by using the governing equation for u and the boundary
conditions.
When u = u2(xl, x3, /)e 2 , the governing equation for u, namely, Navier's
equation (9.8.6), takes the following form in the absence of body forces
(u2,u + u2t33) = p

(9.13.60)

Let us first consider the case where 0 < x3 < oo. Substitution for u2 from
(9.13.58) into (9.13.60) yields the relation

2 - ( ' - g )

<93 6)

Since a has to be greater than 0, c has to be less than c 2 .


Next, let us consider the case where -H < x3 < 0. If p is the density and
fi is the shear modulus of the layer, then the governing equation for u2 in
-H < x3 < 0 is the equation (9.13.60) with p and replaced by p and //.
Substitution for u2 from (9.13.59) in this governing equation yields the
equation
(9.13.62)
/"(*3) - 2f{x,) = 0

448

EQUATIONS OF LINEAR ELASTICITY

where
/>2 = ( l - )

(9.13.63a)

c\ = ^
P

(9.13.63b)

A general solution of the differential equation (9.13.62) is


/(JC 3 )

s cosh /?x3 + C sinh #* 3

(9.13.64)

where B and C are arbitrary constants.


Since there is a welded contact between the layer and the half-space, the
displacement and the stresses must be continuous across the interface
JC3 = 0. That is
( l / 2 > ^31 > T 32> T 33)llayer

( w 2 3 1 T 32 > T 33) I half-space

(9.13.65)

for :3 = 0.
Also, since the boundary of the composite structure, x3 = - / / , is stress
free, we have
(*3i > T32, * 33 )| layer = (0, 0, 0)

(9.13.66)
for JC3 = -H.
Since u = w2(*i, ^3, t)t2 we find from the stress-displacement relation
(9.8.4) that
T

3i = T33 = 0

in the layer and the half-space

(9.13.67)

and
in the half-space
in the layer

(9.13.68)

By use of (9.13.58), (9.13.59) and (9.13.64) in (9.13.68) we obtain

32 = -oLe *3 exp[i(t - /c)]

in the half-space

32 = M[Bsinhx3 + Ccoshx^] exp[ico(t - X\/c)]

(9.13.69)
in the layer
(9.13.70)

Using (9.13.58), (9.13.59), (9.13.64), (9.13.67) and (9.13.70), we obtain


from the interfacial conditions (9.13.65) and the boundary conditions
(9.13.66) the following three relations to be satisfied by A9 B and C:
A - B = 0
+ fiC = 0
BsinhH

- CcoshH

= 0

(9.13.71)

9.13

ELASTIC WAVES

449

For the waves under consideration to exist, A, B, C cannot all be 0; that


is, the system (9.13.71) should have a nontrivial solution. A necessary and
sufficient condition for this is
1

-1

= 0

uxfaH -cosh/?//

which simplifies to
=

-fltanhH

(9.13.72)

We note that and fi are positive constants, and a > 0 for a surface-type
wave. From (9.13.63a), we find that is either real or purely imaginary
according as c < c2 or c > c2. When is real (positive, negative or 0),
(tanhH) is nonnegative. Hence, if is real, the condition (9.13.72)
cannot be satisfied. As such, in this case, A = B = C = 0 and, consequently, u = 0 in the half-space and in the layer. Therefore, the wave
propagation considered is not a possible one.
Let us now consider the other case, where is purely imaginary. If we set
= ///, where is a nonzero real number, the condition (9.13.72) becomes
/7(tan ) =

(9.13.73)

Since 5* 0, both sides of this equation are positive. Also, tan //// takes
every value between 0 and oo; hence, //(tan////) certainly takes the value
/fi at least once. Hence, in this case the condition (9.13.72) is possible.
Setting = in (9.13.63a), we get
(9.13.74)

'-7(S-

Since is real and nonzero, c has to be necessarily greater than c 2 .


Substituting for a and from (9.13.61) and (9.13.74) into (9.13.73), we get
the following equation governing the speed c:
tan

m-'YH

(1 - C 7c 2 2 ) 1 / 2

fiic'/ci - 1)

1/2

(9.13.75)

This is a transcendental equation that yields infinitely many roots for c.


However, c has to be less than c2 and greater than c2. Thus, the possible
speeds of the wave considered are precisely the roots of the equation
(9.13.75) that lie in the interval (c2,c2). If the layer and the half-space are
such that c2-< c2, the wave considered is not a possible one.

450

EQUATIONS OF LINEAR ELASTICITY

It is important to observe that the equation (9.13.75) contains . Therefore, the values of c determined from this equation do depend on . This
means that the speed of propagation depends on the frequency. That is, the
wave is dispersive.
In the limiting case when the layer is absent, we have = fi and p = p.
Then the equation (9.13.75) leads to the impossible condition 0 = - 1 .
Hence, in this case, the wave considered cannot exist. This result is in
agreement with that proven in Example 9.13.2.
Note: The problem discussed above is by A. E. H. Love (1911), and the
wave considered in the problem is known as a Love wave. Like Rayleigh
waves, Love waves are also of practical importance, particularly in
seismology and geophysics.

9.14
EXERCISES
1. Write down the relations (9.2.8) in the matrix form.
2. Prove the identities (9.3.6), (9.3.7) and those contained in Table 9.1.
3 . Write down the relations (9.3.14) in the matrix form.
4. Prove the following relations connecting the stress invariants and the strain
invariants:
7T = (3A + 2)
= (3 + 4/i)/| + 42
IIIj = 2( + 2) + 42 + 8// 3 ///
5. Obtain the following expressions for the strain-energy function W:
W = ^

2)2, = X-K(ekk)2 +

+ \<?-

+ - TLwew =

tkke>

mm ^

ij

^ij

3 + 2

(^+\-2

- &

- 2d

= wx + w2

efef

\%KTkk +

bkkY

" fyTTu

6ij

6[JK ~

= \-]

v)//TJ ~ /

+ ^(/-3/)

kk

9.14

EXERCISES

451

where
_ 1
~2Kekk

Wl

1 2
JsKTkk

W2 = -/![(* - e22)2 + (e22 - e33)2 + (e33 - en)2 + 6(e2l2 + e23 + efO]


= ^ - - 22 ) 2 + ( 22 - 33 ) 2 + ( 33 - ) 2 + 6(22 + | 3 + 32,)]
6. Evaluate W for the stress field
Til = T 2 2 = T 3 3 = T 1 2 = 0,

T 1 3 = -/iOX2,

T23 = ^

where a is a constant.
7. If 13 = 23 = 33 = 0, show that
^ + ^22 = ^ - ( + 22) and e22 - eu + 2iel2 =^(22

- + 2hl2)

Deduce that
1- v
^ = 1^(T11

+ T

, 1+ v,
.,
") + - ^ - 1*22 - Til + 2lT12|2

8. Find the stresses associated with the following displacement fields:


(i) ux = 0,

u2 = 0,

M3

= ke

(ii) Wi = 23 ,

u2 = kx3xl9

u3 = kxxx2

(iii) ! = kx2x3,

u2 = x 3 * lf

w3 = k(x2 - x2)

(iv) Hi = - - { * ! + v(x? - xf)),

w2 = Arv*^,

w3 = kx3xl

where is a constant and = t2Ln~l(x2/xx); xx ^ 0.


9. For an elastic body subjected to the stress system ,-, = -poijt
placements are ut =
(-p/lK)xt.

show that the dis-

1 0 . When an elastic solid is immersed in a fluid whose density is the same as that
of the solid, the stresses in the solid are found to be = 22 = 33 = -p + pgx3,
where p is the pressure at the level x3 = 0 and g is the constant gravity factor.
Compute the corresponding strains and displacements.

452

EQUATIONS OF LINEAR ELASTICITY

1 1 , A rod placed along the x axis and subjected to a longitudinal stress is so


constrained that there is no lateral contraction. Show that
, =

(1 - v)E
(1 + v)(l - 2v)

1 2 . Derive the solution (9.4.8) for the equation (9.4.7).


1 3 . Show that the relation (9.8.26) can be put in the following component form:
OU:

1+ v

dn

,./.
J J +

uk krii
1 - 2v '

where
n-V

and

u = -(uid

- ujti)

1 4 . For an incompressible elastic body, show that


du
s = -pn + 2 + ( X curl u)
dn
1 5 . If f is such that div f = 0 and curl f = 0, prove the following in the static case:
(i) u, E and T are biharmonic functions.
(ii) div u, curl u, tr E and tr T are harmonic functions.
1 6 . Verify the following identities:
r2 -

r2

.. c - ci

(ii)^

1 - 2v

c]

1
2(1 - v)

1 7 . If ut are solutions of Navier's equations of equilibrium under zero body forces,


show that so are uik.
1 8 . Show that, in the absence of body forces, Navier's equation of equilibrium
may be written as follows
u + - i l + -J(divu)x

= 0

1 9 . Let
u0 = + \( + /i)(div u)x
Show that in the static case, u0 satisfies the equation
VZUn =

- f

+
- (div f )x
2(A + 2)

2 0 . For an incompressible elastic body in equilibrium, show that


(i) curl curl u + Vp = f
(ii) V2T - 2VVp + (Vf + VfT) + (div f)I = 0

9.14

EXERCISES

453

21. Show that the displacements


Wi =

Axl

Ax2

U2 =

Mi = ,

where A is a constant, satisfy the Navier's equations of equilibrium in the absence


of body forces. Find the corresponding stresses. Also compute the stress vector on
the surface r = constant.
22. In an elastic body in equilibrium under the body force f = axxx2^, where a is
a constant, the displacements are of the form
ux = Ax2x2x3,

u2 = Bxxxlx3,

M3 = Cxxx2x\

where A, B, C are constants. Find A,B9C. Evaluate the corresponding stresses.


23. Verify that the following are solutions of Navier's equation of equilibrium in
the absence of body forces:
(i) u = V0,

V2</> = 0
V2g = 0,

(ii) u = (a x)Vg - (3 - 4v)gV(a x),


(iii) u = VA x V(a x),

V A = 0,

(iv) u = (h) - ^ V(div h),


dxx
3 - 4v

a = constant

a = constant
V2h = 0

In each case find the corresponding stresses.


24. Find the functions and h such that u = h + r2(V<) is a solution of Navier's
equation of equilibrium under zero body force.
25. If and h are harmonic functions such that
(5 - 4) + x + div h = constant
show that u = h + is a solution of Navier's equation of equilibrium under zero
body force. What happens to this solution in the case v = y?
26. If and h are harmonic functions such that
<3</>

(3 - 4v) + div h = constant


dxl
show that u = h + *j(V</>) is a solution of Navier's equation of equilibrium in the
absence of body force. What happens to this solution for v = j?
27. If and h are such that 2 = div h and V2h = 0, show that

u= h

-2(rbj v *

is a solution of Navier's equation of equilibrium in the absence of body force. Prove


that this solution is complete.

454

EQUATIONS OF LINEAR ELASTICITY

2 8 . If h and h* are such that V2h = curl h* and V2h* = 0, show that
u = h*

1 - 2v

curl h

is a solution of Navier's equation of equilibrium in the absence of body force. Prove


that this solution is complete.
2 9 . If pQ and p are harmonic functions and c is a constant unit vector, show that
u = pc -
V(p0 + (x c)p]
4(1 - v)
is a solution of Navier's equation of equilibrium in the absence of body force. (This
solution is known as Boussinesq's solution).
3 0 . Let p be an arbitrary vector function obeying the equation V2p = -(1/) and
p0 be a scalar function obeying the equation V2p0 = (1/) f. Show that u given by
1
u = p-
V(/?0 + x p)
4(1 - v)
is a complete solution of Navier's equation of equilibrium. (This solution is known
as the Papkovitch-Neuber solution. The functions p and/? 0 are called PapkovitchNeuber potentials).
3 1 . If f = constant, show that in the equilibrium case the Papkovitch-Neuber
potentials may be expressed in the form
1

r2

where p j and p* are harmonic functions and


' x* = jrfo + *le 2 + *f e3
3 2 . Let g be an arbitrary vector function obeying the equation V2V2g =
Show that
U =

V 2 g

- 2 ^

-(\/).

V ( d i V g )

is a complete solution of Navier's equation of equilibrium. (This solution is known


as the Galerkin solution. The vector g is called the Galerkin vector.)
3 3 . Given that
A
g = (x2ej - xxe2)
where A is a constant, is the Galerkin vector in the equilibrium case in the absence
of body forces, compute the corresponding displacements and stresses.

9.14

34.

EXERCISES

455

In the equilibrium case, the Galerkin vector is given by


g = ar2ex + bx*e2

where a and b are constants. Find the body force that must be acting. Compute the
corresponding displacements and stresses.
35.

Repeat the Exercise 34 in the following cases:

(i) g = r2e3

(ii) g = &3

(iii) g = r2(x2e, - xxt2)

36. Suppose the body force f has magnitude / and acts in the JC3 direction. If a
scalar function g obeys the equation V4g = -(1/)/, show that

is a solution of Navier's equation of equilibrium. This solution is known as the


Love's solution. The function g is called the Love's strain function.
37. Deduce the Papkovitch-Neuber solution of Exercise 30 from the Galerkin
solution of Exercise 32 and vice versa.
38.

Using Navier's equation of motion, show that ekk satisfies the equation
kk

39.

2(1 - v)

Using Navier's equation, show that ^ = j(uifj - ujfi) satisfies the equation

= -^(>-/)

where/ =/ - w,.
40.

Find c such that


In
ul = A sin (X! ct),

u2 = u3 = 0

where A and / are constants, may satisfy Navier's equation of motion in the absence
of body forces.
41. Find/(r) such that uk = f(f)xk exp(icot) may be a solution of Navier's equation
of motion in the absence of body forces.
42. Let g be an arbitrary vector function obeying the equation D2D!g = -(\/).
Show that

u = Di8

~2(rh v ( d i V 8 )

is a complete solution of Navier's equation of motion. (This solution is known as the


Cauchy-Kovalevski-Somigliana solution.) Hence, deduce the Galerkin solution of
Exercise 32.

456

EQUATIONS OF LINEAR ELASTICITY

4 3 . Let p be an arbitrary vector function obeying the equation D2p = (-l//*)f


and pQ is a scalar function satisfying the equation 0 = 2divp - Dj(x p).
Show that
U= P

"40r^ V(Po + X*P)

is a complete solution of Navier's equation of motion. (This solution is known as the


Sternberg-Eubanks solution.) Hence deduce the Papkovitch-Neuber solution of
Exercise 30.
44. Let k be an arbitrary vector function obeying the equation Dxk =
-(1/(A + 2/i))f and h be a vector function obeying the equation D2h = curlk.
Show that
u=k

1 - 2v

curl h

is a complete solution of Navier's equation of motion. Hence deduce the solution


indicated in Exercise 28.
45. Let be an arbitrary scalar function and be an arbitrary vector function
obeying the equations
1
1
= - X ,
2 =
+ 2

where and are defined by f = (V# + curl ) with div = 0. Show that u =
V0 + curl is a complete solution of Navier's equation of motion. (This solution
is known as the Green-Lam solution, and and are called the Lam potentials.)
Show that the static counterpart of this solution is not complete.
46. If w3 = 0 and ux and u2 are functions of xi, x2 and t, show that
3
oxx

dx2

3
2

3
dxx

constitute a solution of Navier's equation of motion in the absence of body forces,


provided = 0, D2 = 0. Is this solution complete?
47. If and are the Lam potentials, show that
divu = 2 ,

curlu = - 2

Deduce that in the absence of body forces


D^div u) = 0,

D2(curl u) = 0

9.14

EXERCISES

457

4 8 . Show that every solution of Navier's equation of motion can be represented in


the form u = u(1) + u (2) , where
-,
DjU0* = - + 2

curlu (1) = 0

D 2 u (2) = curl ,

div u (2) = 0

with
f = V# + curl
4 9 . Obtain expressions for the stress tensor associated with (i) the CauchyKovalevski-Somigliana solution, (ii) the Sternberg-Eubanks solution, and the
Green-Lam solution, for the displacement vector.
5 0 . Deduce the Cauchy-Kovalevski-Somigliana solution, the Sternberg-Eubanks
solution and the Green-Lam solution from each other.
5 1 . Show that the stress system = 22 = 13 = 23 = 12 = 0, 33 = pgx3,
where p and g are constants, satisfies the equations of equilibrium and the equations
of compatibility. Find the corresponding displacements.
5 2 . Show that the following stress system cannot be a solution of an elastostatic
problem although it satisfies Cauchy's equations of equilibrium with zero body
forces:
Tll

= x\ + v(x2 - xl),

T12 = -2vx1x2i

22 = x\ + v(xf - x 2 ),

33 = v(x2 + xf)

23 = 31 = 0

5 3 . Determine whether or not the following stress components are a possible


solution in elastostatics in the absence of body forces:
= 23,

T22 = bx3xl9

13 = ex2,
where a,b,c,

12 = dx\,

^ = cxxx2,

23 = jX\

d, e,f are all constants.

5 4 . In an elastic body in equilibrium under the body force f = axxx2e3, where a is


a constant, the stresses are of the form
Tu = 23,

T33 = cxxx2x3

T22 = DX\X2x3>

Ti2 = ( + bxl)x3,

23 = (bxl + cx\)xx,

13 = (cx\ + ax\)x2

where a, b, c are constants. Find these constants.


5 5 . In the absence of body forces, a general solution of Cauchy's equations of
equilibrium is given by
T

ll

,22>

22

, 11

12

, 12

12

13

33

Show that this is a possible solution in elasticity, provided = 0.

458

EQUATIONS OF LINEAR ELASTICITY

5 6 . If is an arbitrary vector function satisfying the equation 2 = f and is a


scalar function obeying the equation V2 = (1/(1 - ) ^ , show that
T = - ( 2 ) - ( + )
is a complete solution of the Beltrami-Michell equation (9.11.7). (This is known as
the Schaefer's solution.)
5 7 . For the Beltrami-Michell equation in the static case, deduce the Schaefer
solution by starting with (i) the Galerkin solution and (ii) the Papkovitch-Neuber
solution of Navier's equation of equilibrium.
5 8 . Specialize the Beltrami-Michell equations (9.11.7)' to the case where/) = ,.
Hence deduce that if 2 = 0, then V2rkk = 0 and V2ekk = 0.
5 9 . If hi and are arbitrary functions satisfying the equations

show that
TU = (hktk - 2c22V20)Jl7 + 2c|* i ( / - (hitj + hjti)
constitute a solution of the Beltrami-Michell equations (9.11.7)'.
6 0 . If is an arbitrary vector function obeying the equation 2 = f and is a
scalar function satisfying the equation = (1/(1 - ) ^ , show that
T = - ( 2 ) - ( + )
is a complete solution of the Beltrami-Michell equation (9.11.6). (This solution is
known as the Tedorescu's solution.)
6 1 . For the Beltrami-Michell equation in the dynamic case, deduce the Teodorescu
solution by starting with (i) the Cauchy-Kovalevski-Somigliana solution, (ii) the
Sternberg-Eubanks solution, and the (iii) Green-Lam solution, of Navier's
equation of motion.
6 2 . Prove the following identities:
(i)

T dV =\

(ii)

s dS =

(/i(u n + n (g) u) + A(u n)I) dS


((A + 2//)(div u)n + ( u) x n) dS

6 3 . For an elastic body in equilibrium, show that


EdV=^-

(x s) dS +
s

2(3 + 2)

(x f) dV
}v

x-sdS

}v

x-fdV

9.14

EXERCISES

459

64. For an elastic body in equilibrium, show that


u ((A + 2/i)(div u)n + //(curl u) x n) dS +

f u dV

= \ {(A + 2/i)(divu)2 + \cm\u\2}dV


Hence deduce the uniqueness of solution of the displacement boundary value
problem in elastostatics.
6 5 . Consider a long circular cylinder rotating with a uniform angular speed
about its axis. The cylinder is not free to deform longitudinally. The body force is
ignored and the boundary is stress free. Find the displacements and stresses
produced in the cylinder.
66. Consider a solid elastic sphere deforming statically due to mutual attraction
between its parts. The force of gravity at any point inside the sphere is directly
proportional to the distance of the point from the center and is directed toward the
center. The surface of the sphere is stress free. Find the displacements and stresses
produced in the sphere.
67. Find the displacements and stresses in a spherical shell whose inner boundary
is fixed rigidly and outer boundary is subjected to a uniform pressure. Body forces
are ignored.
68. Find the displacements and stresses in a long circular cylinder whose boundary
is subjected to a uniform displacement in the radial direction. Body forces are
ignored.
69. Find the displacement and stress fields developed in an elastic half-space with
a plane boundary when the boundary is subjected to a uniform shear stress. Body
forces are ignored.
70. Consider an unbounded elastic solid having a spherical cavity. The surface of
the cavity is subjected to a time-dependent pressure p(t) for time / > 0 + . Body
forces are ignored. Find the displacement field generated.
7 1 . Compute the displacements due to Rayleigh waves and analyze their nature on
the boundary surface of the half-space.
72. Consider an elastic layer of uniform thickness and infinite lateral dimensions.
The boundary surfaces of the layer are stress free. Investigate the existence of plane
harmonic waves propagating parallel to the surfaces. Body forces are ignored.

CHAPTER 10

EQUATIONS OF
FLUID MECHANICS

10.1
INTRODUCTION
In this chapter, we consider another important branch of continuum
mechanics: the mechanics of nonviscous and Newtonian viscous fluids.
Like the classical elasticity theory, this branch also uses linear constitutive
equations. Many common fluids including water and air satisfy such
constitutive equations. Nonviscous and Newtonian viscous fluids therefore
serve as excellent models for studying the mechanical behavior of a wide
variety of common liquids and gases.
The scope of fluid mechanics is vast. We restrict ourselves to the derivation of the governing equations for nonviscous and Newtonian viscous fluid
flows and their immediate consequences. Some simple and standard
applications are also discussed.

10.2
VISCOUS AND NONVISCOUS FLUIDS
The fundamental characteristic property of a fluid, which distinguishes it
from a solid, is its inability to sustain shear stresses when it is at rest or in
461

462

10

EQUATIONS OF FLUID MECHANICS

Figure 10.1.

Hydrostatic pressure.

uniform motion. More specifically, whereas shear stresses can occur in a


solid even when the solid is in static equilibrium or in uniform motion, shear
stresses cannot occur in a fluid unless it undergoes a nonuniform motion.
Experimental observations show that, in a fluid at rest or in uniform
motion, the stress on a surface element consists only of a pressure (negative
normal stress) and that this pressure has the same magnitude on all surface
elements through a point. This pressure is usually called hydrostatic pressure
or simply static pressure. Like density, static pressure is a physical property
of the fluid and generally varies from one point to another.
Thus, on the basis of experimental observations, it is postulated that on
a surface element in a fluid at rest or in uniform motion the stress vector s
is given by
(10.2.1)
s = -p0n
where p0 is hydrostatic pressure; see Figure 10.1.
The Cauchy's law, (7.4.8), then yields (see Example 7.4.4)
T = -p0l

(10.2.2)

Taking the trace of this equation, we find that


Po=P

(10.2.3)

wherep = (l/3)(tr T) is the mean pressure. Thus, the hydrostatic pressure


is equal to the mean pressure.
Experimental observation also shows that fluids in nonuniform motion
exert shear stresses in addition to normal stresses. This means that, when a
fluid undergoes a nonuniform motion, the stress vector on a surface
element in the fluid acts obliquely to the surface; see Figure 10.2. As such,
for fluids in nonuniform motion relation (10.2.1) does not hold, in general.
However, in many practical problems, it is found that the effects of
shear stresses are small and can even be neglected. But, important realistic
situations occur where these effects are appreciable and cannot be ignored.

10.3

STRESS TENSOR FOR A NON VISCOUS FLUID

463

Figure 10.2. Stress vector acting obliquely to a surface.

In view of this observation, fluids are generally classified into two classes:
(i) fluids that exert no or negligible shear stresses, and (ii) fluids that
exert nonnegligible shear stresses. A fluid in which shear stresses are not
negligible is called a viscous fluid. On the other hand, a fluid in which no
or negligible shear stresses occur is called a nonviscous (or inviscid or
frictionless) fluid. It should be pointed out that all fluids are viscous to a
certain degree and that the concept of a nonviscous fluid is just an ideal one.
Nevertheless, the study of nonviscous fluid flows is of great practical utility
in engineering applications dealing with common liquids and gases (like
water and air, for example) and an essential part of fluid mechanics.

10-3
STRESS TENSOR FOR A NONVISCOUS FLUID
From the definition of a nonviscous fluid, it follows that the stress vector on
a surface element in such a fluid is of the form
s = -pu

(10.3.1)

even when there is a nonuniform motion, where p is a scalar function that


reduces to the hydrostatic pressure p0 when there is no motion or when the
motion is uniform. Relation (10.3.1) essentially implies that at a point of a
nonviscous fluid every surface element is a principal plane of stress and
every direction is a principal direction of stress, with p as the principal
stress.
Consequently, Cauchy's law (7.4.8) gives the following equation for the
stress tensor (as shown in Example 7.4.4):
T = -pi

(10.3.2)

We note that T, given by (10.3.2), is symmetric and isotropic. Thus, for


a nonviscous fluid the symmetry of T is a physical property as in the case of
an elastic solid governed by Hooke's law (9.2.3). Further, unlike in the

464

10

EQUATIONS OF FLUID MECHANICS

case of an elastic solid, the isotropy of T is also a physical property of a


non viscous fluid. This means that there is no preferred direction in a nonviscous fluid. Relation (10.3.2), which describes the intrinsic material
property of a nonviscous fluid, is taken as the material law for such a fluid.
It follows from (10.3.2) that
p=p
(10.3.3)
where p is, as usual, the mean pressure. Thus, p present in the material law
(10.3.2) represents the mean pressure in a nonviscous fluid in nonuniform
motion. Like the static pressure p0i the mean pressure p generally varies
from one point to another and at any point it has the same value in all
directions. When the fluid is brought to uniform motion or rest, p reduces to
p0, as already stated. Therefore p is called dynamic pressure or just pressure.

10.4
GOVERNING EQUATIONS FOR A
NONVISCOUS FLUID FLOW
In Section 8.7, it was pointed out that the field equations (8.7.1) to (8.7.3)
hold for all continua. Hence, these equations automatically hold for nonviscous fluids. Let us recall the spatial forms of these equations and record
them here for ready reference.
Equation of continuity:
Dp
j + /?divv = 0
(10.4.1)
Equations of motion:
div T r + pb = p T = Tr

(10.4.2a)
(10.4.2b)

Equation of energy:
De
p = T Vv - div q + ph

(10.4.3)

To these field equations, let us append the material law (10.3.2), which is
valid only for nonviscous fluids.
Material law:
T = -pi
(10.4.4)
As noted earlier, T is symmetric by virtue of this material law; therefore,
equation (10.4.2b) is identically satisfied.

10.4

GOVERNING EQUATIONS FOR A NONVISCOUS FLUID FLOW

465

It is easily seen that 11 scalar equations are involved in the governing


equations (10.4.1)-(10.4.4), whereas 15 unknown field functions are
present in these equations, the unknowns being 3 velocity components viy 3
heat flux components qt, 6 stress components ,, = (,,) and 3 scalars p, p
and . (The body force b and heat supply h are taken as known functions
as usual.) Accordingly, equations (10.4.1) to (10.4.4) are inadequate to
determine all the field functions. Therefore, we have to either reduce the
number of unknowns or increase the number of governing equations in
order to close the system of governing equations. These two possibilities
lead us to consider the following two important particular cases.
10.4.1.

INCOMPRESSIBLE FLUIDS

In Section 6.3, we noted that, for every motion of an incompressible


continuum,
divv = 0

(10.4.5)

In view of the equation of continuity (10.4.1), this condition is equivalent


to the condition
Dp
-^ = 0
(10.4.6)
Dt
which implies that p retains its initial value p0 throughout the motion. A
fluid for which this property of p holds is called an incompressible fluid,
and (10.4.5) is employed as the equation of continuity for the motion of
such a fluid.
In this case, equations (10.4.2) can be together rewritten as
divT + A j b ^ A , ^

(10.4.7)

Thus, for an incompressible fluid, equations (10.4.1) and (10.4.2) are


replaced by equations (10.4.5) and (10.4.7).
From equations (10.4.4) and (10.4.5), we get
T . vv = -p(\ Vv) = -p div v = 0

(10.4.8)

Hence, for an incompressible nonviscous fluid, the energy equation (10.4.3)


reduces to
De
Po = - d i v q + ph
(10.4.9)
We note that 10 scalar equations are involved in (10.4.4), (10.4.5) and
(10.4.7) and the number of unknown field functions is also 10, the functions

466

10

EQUATIONS OF FLUID MECHANICS

beingp, vt and ,,. Equation (10.4.9) does not contain these field functions
and is therefore not needed in their determination. Equations (10.4.4),
(10.4.5) and (10.4.7) thus serve as a closed system of governing differential
equations for the study of nonviscous, incompressible fluid flows. For such
flows, energy considerations are generally not important and equation
(10.4.9), which is purely an equation of balance of thermal energy, may be
discarded.
10.4.2.

COMPRESSIBLE FLUIDS

A fluid is said to be compressible if it is not incompressible; that is, if p is


not constant during the motion. In this case, some constitutive relations
from thermodynamics are appended to the system of equations (10.4.1) to
(10.4.4) to close the system.
Since the density of a compressible fluid varies during the motion, the
equation of continuity (10.4.1) shows that div v ?* 0. Hence, fluid elements
undergo a change in their volumes during the motion. A change in volume
of a material is generally caused by a change in normal stress and a change
in temperature. Thus, for a compressible fluid, it is postulated that/?, p and
T are related in a definite way, where T is absolute temperature; that is,
there exists a functional relationship of the form
f(p,p,T) = 0

(10.4.10)

where / is a known function. An equation of this type is called a kinetic


equation of state. The actual form of the equation depends on the physical
properties of the fluid being dealt with.
The equation of energy (10.4.3) shows that is related to /?, T, v and q in
a definite way. From the material law (10.4.4) and the equation of
continuity (10.4.1), we find that
T . w = -pi (Vv) = - p d i v v = ~jf

(10.4.11)

Thus, the stress-power T Vv is directly related with p and p in a definite


way. Consequently, it follows that is related to p, p and q in a definite
way. If we employ Fourier's law of heat conduction (8.7.20), namely,
q = -kVT

(10.4.12)

then q is determined by . Also, the kinetic equation of state, (10.4.10),


determines p as a function of p and . Thus, ultimaltely can be related to
p and T in a definite way. Hence, it is postulated that may be taken as a
known function of p and T\ that is,
= (, )

(10.4.13)

10.4

GOVERNING EQUATIONS FOR A NON VISCOUS FLUID FLOW

467

An equation of this type is called a caloric equation of state. The actual


form of the function (, ) depends on the physical properties of the fluid
being dealt with.
If equations (10.4.10), (10.4.12) and (10.4.13) are appended to equations
(10.4.1) to (10.4.4), then we have in all 16 scalar equations with 16 unknown
field functions, the field functions being 3 velocity components vi9 3 heatflux components qi9 6 stress components 0 and 4 scalars /?, /?, T and .
Equations (10.4.1) to (10.4.4), (10.4.10), (10.4.12) and (10.4.13) therefore
serve as a closed system of governing differential equations for the study of
nonviscous compressible fluid flows. Note that whereas (10.4.1) to (10.4.3)
are field equations, (10.4.4), (10.4.10), (10.4.12) and (10.4.13) are all
constitutive equations. Unlike in the case of an incompressible fluid, where
the material law (10.4.4) is the only constitutive equation, the constitutive
equations for a compressible fluid consist of the equations of state (10.4.10)
and (10.4.13) and a law of heat conduction (10.4.12), in addition to the
material law (10.4.4). Also, in this case, the energy equation (10.4.3) is
coupled with other governing equations.
Thus, for a nonviscous fluid flow, two sets of equations govern the
flowone set is applicable to incompressible fluids, and the other to compressible fluids. For an incompressible fluid, the density does not change
with motion, and the pressure /?, the velocity field v and the stress T serve
as the field functions. On the other hand, for the case of a compressible
fluid in which the density changes with motion, the field functions are the
density p, the pressure /?, the temperature T, the internal energy , the heat
flux q, the velocity v and the stress T. Obviously, the study of compressible
fluid flows is more difficult than that of incompressible fluid flows.
Experimental observations show that in the case of liquids and low-speed
gases, the changes in density during the motion are very small and hence
can be neglected. Liquids and low-speed gases are therefore generally
considered as incompressible fluids. In the case of gases moving with high
speed, the variations in density during the motion are considerably large
and cannot be neglected. High-speed gases are therefore regarded as
compressible fluids. In reality, no fluid is incompressible; every fluid is
compressible to a certain degree. Like the concept of nonviscous fluid, the
concept of an incompressible fluid is just an ideal one.
It is to be emphasized that fluids are classified as compressible or incompressible accordingly as the density changes or does not change during the
motion. In a fluid, the density may change from one part of the fluid to the
other, and such a change of density may occur in both compressible and
incompressible fluids. A fluid in which the density is the same everywhere
in the fluid (so that Vp = 0) is called a homogeneous fluid. On the other
hand, a fluid in which there is a change in density from one part of the

468

10

EQUATIONS OF FLUID MECHANICS

fluid to the other (so that Vp 5* 0 in general) is called a nonhomogeneous


fluid. Salt water with nonuniform concentrations with depth is a standard
example of a nonhomogeneous fluid. The governing equations considered
in the previous paragraphs hold for both homogeneous and nonhomogeneous nonviscous fluids.
A point that is to be clearly understood is the following. Recall that the
initial density p0 has been defined by p0 = p(x9 0). Evidently, p0 varies with
x in general. It is only for homogeneous fluids that we have p0 = constant.
10.4.3.

STANDARD FORMS OF EQUATIONS OF STATE

It was pointed out that the equations of state, namely, (10.4.10) and
(10.4.13), serve as constitutive equations for compressible fluids and the
exact forms of these equations depend on the physical properties of the
fluid being dealt with. Here we give some standard forms of these equations
that are needed in our further discussion.
The caloric equation of state (10.4.13) is intended to specify in terms of
p and T. One of the simplest forms of this equation is
e = cvT

(10.4.14)

where cv is either a constant or a function of p. Then cv is called the


specific heat at constant volume.
The kinetic equation of state (10.4.10) is intended to specify a functional
relationship between p, p and T. One of the simplest forms of this equation
is the Boyle's law given by
p = pRT

(10.4.15)

where R is a constant for a particular fluid under consideration.


A compressible fluid for which (10.4.14) with cv = constant and
(10.4.15) hold is called a perfect gas, and R is then known as gas constant.
Common gases like air, oxygen and nitrogen are regarded as perfect gases.
For air, the value of R is approximately equal to 2.87 x 106.
An important particular case of equation (10.4.15) is
p
= constant
P
T= constant

(10.4.16a)
(10.4.16b)

A compressible fluid for which (10.4.14) with cv = constant and (10.4.16)


hold is called an isothermal perfect gas.

10.4

GOVERNING EQUATIONS FOR A NONVISCOUS FLUID FLOW

469

An important generalization of equations (10.4.16) is

P = constant
P1
T
p

= constant

(10.4.17a)
(10.4.17b)

where
y = (l + - ) > 1

(10.4.18)

A compressible fluid for which (10.4.14) with cv = constant and


(10.4.17) hold is called an isentropicperfect gas; the constant y is called the
adiabatic constant. If
cp = c + R

(10.4.19)

then cp is called the specific heat at constant pressure, and the following
relations can be verified:
cv =

(10.4.20a)

cp = - ^ y- I

(10.4.20b)

^ = y

(10.4.20c)

We often deal with what are called barotropic fluids. By a barotropic


fluid, we mean a compressible fluid for which the kinetic equation of
state, (10.4.10), is independent of T and can be expressed in either of the
following forms:
p = p(p)

(10.4.21a)

p = p(p)

(10.4.21b)

A nonviscous barotropic fluid is often referred to as an Eulerian fluid


or elastic fluid.
Equations of state that are more general than those just summarized are
also found in the literature. For easy reference, the governing equations for
incompressible and compressible nonviscous fluid flows are summarized in
Tables 10.1 and 10.2 with the same equation numbers as in the text.

470

10

EQUATIONS OF FLUID MECHANICS

Table 10.1. Governing Equations for Nonviscous Incompressible Fluid Flows


1. Equation of continuity:

divv = 0

(10.4.5)

D\
divT + /?0b = /> 0

(10.4.7)

T = -pi

(10.4.4)

2. Equation of motion:

3. Material law:

Total number of scalar equations: 10


Total number of unknown field functions (/?, Vj and ,,): 10

Table 10.2. Governing equations for Nonviscous Compressible Fluid Flows


1. Equation of continuity:
Dp

Dt + /?divv = 0

(10.4.1)

D\
divT + pb = p
Dt

(10.4.2)

De
py = T Vv - div q + Hph
Dt

(10.4.3)

T=/7l

(10.4.4)

2. Equation of motion:

3. Equation of energy:

4. Material law:
5. Fourier's law of heat conduction:
q =

-kVT

6. Equations of state

,,)

=0

= (/>, )

(10.4.12)
(10.4.10)
(10.4.13)

Total number of scalar equations: 16


Total number of unknown field functions (/?,/?, , , qit vt and TU): 16

EXAMPLE 10.4.1 Assuming that the coefficient of thermal conductivity k is constant, show that the energy equation for a nonviscous
compressible fluid can be expressed in the following alternative forms:
De p Dp k 72
= ^2 - f + - V r + A
Dt p Dt p
D

Dt

p] = -k 2,^ + -1 -Dp
^ +h
P
pDt
P\

6 + -

v(10.4.22)

'

(10.4.23)

10.4

GOVERNING EQUATIONS FOR A NONVISCOUS FLUID FLOW

471

Solution Substituting for T Vv from (10.4.11) and for q from (10.4.12)


in the energy equation (10.4.3) we immediately arrive at equation (10.4.22).
With the use of the identity

M=I^_*^

(10 .4.24)

Dt \pj
p Dt p2 Dt
equation (10.4.22) reduces to equation (10.4.23).

Note: The quantity + (p/p) present on the lefthand side of (10.4.23) is


called specific enthalpy.
EXAMPLE 10.4.2 A flow in which there is no heat flux is called an
adiabatic flow. Assuming that the fluid is nonviscous and there is no heat
supply, prove the following for an adiabatic flow:
(i) For a baratropic fluid

+- =

dp

+ constant

JP

(10.4.25)

(ii) For an isentropic perfect gas

+ - = , Y , , - + constant
(10.4.26)
P (y - 1) P
Solution (i) For an adiabatic flow, we have q = 0, so that VT = 0 by
the Fourier's law of heat conduction (10.4.12). Consequently, div(VT) =
V2T = 0. Then in the absence of , we obtain, from (10.4.23),
]_Dp
(10.4.27)
Dt
p Dt
P
For a barotropic fluid, we have p = p(p) so that

Hence

"?

(10.4.28)

Using this in (10.4.27) and integrating the resulting equation, we obtain


(10.4.25).
(ii) Recall that for an isentropic perfect gas, p/py = constant = /?, say.
Then
constant =

constant
(10.4.30)

Using this in (10.4.25), we obtain (10.4.26).

472

1 0 EQUATIONS OF FLUID MECHANICS

EXAMPLE 10.4.3 Show that for a nonviscous perfect gas flow, the
energy equation can be expressed in the following alternative forms:
cvv = il2-^:
+ -V1T+ h
" Dt
p2 Dt p

Solution

(10.4.31)

I Dp k ~
,
DT
cD- = - - f + - 2 + h
(10.4.32)
p
Dt
p Dt p
For a perfect gas, we get from (10.4.14), (10.4.15) and (10.4.20),

= cpT-E
(10A33>
* = cT=ci-=(i>-i)E
p
R p \R
) p
p
Substituting = cvT in (10.4.22), we immediately get (10.4.31).

Substituting = cpT - (p/p) in (10.4.23), we obtain (10.4.32).


EXAMPLE 10.4.4 (a) Show that for a nonviscous incompressible fluid
flow, the equation of balance of mechanical energy is given by
^

p(b v) dV -

py-ndS

(10.4.34)

(b) If the body force is conservative so that b = -V# for some scalar
potential = (), show that (10.4.34) can be put in the form
- (K + ) = - I p\ n dS

(10.4.35)

where
=-

pXdV

(10.4.36)

is called the potential energy of the fluid contained in V.


Deduce that (i) if the fluid moves tangentially over the surface 5, then
K + remains constant during the motion, and (ii) if K + remains
constant during the motion, then the stream lines lie on surfaces of constant
pressure.
Solution (a) For a continuum, the equation of balance of mechanical
energy is given by (8.6.27). For a nonviscous incompressible fluid,
T Vv = 0, see (10.4.8). Recall also that s = -pn on 5 and T Vv = T D.
Using these results in (8.6.27), we immediately obtain equation (10.4.34).
Note: For an incompressible nonviscous fluid, the equation of balance of
thermal energy is given by (10.4.9), which has no common term with the
equation of balance of mechanical energy, (10.4.34). Hence the motion and
the thermal state of such a fluid do not influence one another.

10.5

INITIAL AND BOUNDARY CONDITIONS

473

(b) For b = - V # , we have


pb-vdV=-\' = -

p(VX-v)dV=-\
P(VX-v)dV=-\

( v * - ^ W = - j

P^dV
(10.4.37)

which, by use of (8.2.10), becomes


pb-\dV

= - \

pxdV

(10.4.38)

Using (10.4.38) and (10.4.36) in (10.4.34) we immediately arrive at (10.4.35).


If the fluid moves tangentially over the surface 5, we have v n = 0 on S.
Consequently, (10.4.35) yields
^(+)

= 0

(10.4.39)

This implies that K + is constant during the motion.


Conversely, if K + is constant during the motion, then (10.4.39) holds
and (10.4.35) yields
p(vn)rfS = 0

(10.4.40)

which, by use of the divergence theorem (3.6.1), becomes


aiv(pv)dV

= 0

(10.4.41)

Using the identity (3.4.14) and the equation of continuity (10.4.5),


expression (10.4.41) reduces to
(Vp v) dV = 0

(10.4.42)

Since V is an arbitrary volume, it follows that Vp v = 0 at all points of the


fluid; that is, v is orthogonal to Vp everywhere in the fluid. Since Vp is
orthogonal to surfaces of constant p (see Section 3.4), it follows that v is
tangential to these surfaces. Thus, stream lines lie on surfaces of constant
pressure at every point of the fluid.

10.5
INITIAL AND BOUNDARY CONDITIONS
A boundary value problem in fluid mechanics consists in determining
all the unknown field functions by solving the governing equations under

474

10

EQUATIONS OF FLUID MECHANICS

appropriate initial and boundary conditions. For incompressible nonviscous fluid flows, the unknown functions are/?, v and T and the governing
equations are (10.4.4), (10.4.5) and (10.4.7). In this case, the following
initial and boundary conditions are prescribed:
v = v(0)

in

F at

v = v*

on

Sv

for

t > 0

(10.5.2a)

Tn = s*

on

ST for

/ > 0

(10.5.2b)

t = 0

(10.5.1)

where F is a material volume of the fluid bounded by a closed surface 5,


and Sv and ST are parts of S that are complementary to each other. As
special cases Sv or ST can be equal to 5. Also, v (0) , v* and s* are prescribed
quantities in their respective domains.
The initial conditions (10.5.1) mean that the velocity of the fluid is known
at every point of the initial volume. For a fluid motion starting from rest at
time / = 0, v(0) = 0. The boundary conditions (10.5.2) mean that the
velocity is prescribed on Sv and stress is specified on ST. The conditions
(10.5.1) and (10.5.2ab) are similar to conditions (9.5.4) and (9.5.3), but
there is one difference. In the case of linear elastic solids, S can be treated
as the boundary surface in an initial configuration. In the case of fluids, S
is the boundary surface in the current configuration.
For nonviscous compressible fluid flows, the unknown functions are p, p,
T, , v, q and T; and the governing equations are (10.4.1)-(10.4.4),
(10.4.10), (10.4.12) and (10.4.13). Since p, and q can be regarded as
known functions of p and T through the equations of state (10.4.10) and
(10.4.13) and the law of heat conduction (10.4.12), the initial conditions are
usually specified in terms of v, p and T and the boundary conditions in
terms of v, T and T. The conditions for v and T are usually taken as those
given in (10.5.1) and (10.5.2ab), whereas the conditions for p and T are as
follows:
in

V at

t = 0

(10.5.3)

T= T0 in

V at

t = 0

(10.5.4)

p = p0

T = T*

on

STO

/
for / > 0
(10.5.5)
v
VT-n = q* on Sq )
'
where ST and SQ are parts of S that are complementary to each other (as
special cases 5 r and Sq can be equal to S). Also, p0 is the initial density and
T0 is the initial temperature, and T* and g* are prescribed quantities in their
respective domains in 5.

10.5

INITIAL AND BOUNDARY CONDITIONS

475

The boundary conditions (10.5.5) mean that the temperature is prescribed


on the part ST of S and the normal derivative of temperature on the
remaining part Sq of S. If T* = T0, then the points of ST experience no
change from the initial temperature, and ST is then called an isothermal
surface. On the other hand, if q* = 0, then no heat flux occurs across Sq,
and Sq is then called an adiabatic surface.
The boundary conditions (10.5.2) require special attention when the
boundary surface is a rigid impermeable solid surface contacting a fluid
and when it is a fluid surface exposed to atmosphere. A rigid solid surface
contacting a fluid could be the wall of a container or the surface of a body
immersed in the fluid. Since the fluid is non viscous, shear stress on every
surface element is 0; hence on a rigid surface 5, no resistance to the motion
of the fluid relative to S occurs and the fluid is free to slip over S. Therefore, no constraint can be imposed on the tangential component of v at a
point of S. However, for the fluid and S to be in geometrical contact, the
velocity of the fluid at the boundary must be such that its component
normal to the boundary is equal to the normal component of velocity of the
boundary. That is,
v n = vs n

(10.5.6)

on S, where v s is the velocity of the rigid boundary and n is the unit normal
to the boundary. In the particular case when the boundary is at rest, the
condition (10.5.6) becomes
vn = 0
(10.5.7)
onS.
The condition (10.5.6) can be rewritten in an alternative form by using
the geometrical equation of the boundary surface. Let F(x, /) = 0 be the
geometrical equation of the boundary surface. Since particles that initially
lie on a boundary continue to lie on a boundary, F(x, t) = 0 is a material
surface, and the speed with which this surface advances perpendicular to
itself is given by (6.2.62). That is,
(dF/dt)
(10.5.8)
=

*'

-^

Consequently, the condition (10.5.6) can be rewritten as


3F
+ (v-n)|VF| = 0

(10.5.9)

on the boundary, F(x, /) = 0.


When 5 is the boundary surface of a fluid that is exposed to atmosphere,
the fluid pressure on S must balance the atmospheric pressure pa.

476

10

EQUATIONS OF FLUID MECHANICS

Thus, on such a surface we impose the following boundary condition:


(10.5.10)

P=pa

on S. This condition is valid only if the surface tension effects are neglected.
EXAMPLE 10.5.1 For a certain flow of a nonviscous fluid in a region
bounded by the fixed solid boundary y = 0, the velocity field is given by
v = , where = x3 - 3xy2. Verify that the boundary condition
v n = 0 is satisfied and that at all points except the origin the fluid slips on
the boundary.
Note that here and in the following discussion, (x,y) or (Jt,j>, z) denote
the Cartesian coordinates.
Solution

From the given expression for v, we find that


vx = 3(x2 - y \

v2 = -6xy9

v3 = 0

(10.5.11)

From these components we note that on the boundary y = 0, we have


v n = v2 = 0. Thus, the boundary condition v n = 0 is verified.
Since v3 = 0, vl is the only tangential component of velocity that occurs
on the boundary y = 0. For y = 0, (10.5.11) yields vx = 3x29 and thus vl
vanishes only at the origin. Hence, except at the origin, everywhere else on
the boundary y = 0 the fluid slips with speed 3A:2 in the x direction.

EXAMPLE 10.5.2 A long cylinder is fixed rigidly in the flow of a nonviscous fluid. Obtain the boundary condition for velocity on the lateral
surface of the cylinder.
Solution Let us choose the coordinate axes such that the z axis is along
the axis of the cylinder. Then at any point P on the lateral surface of the
cylinder, the normal n is parallel to the xy plane so that nx = cos 0,
n2 = sin , n3 = 0, where is the angle between n and the x direction;

Figure 10.3.

Example 10.5.2.

10.6

EULERS EQUATION OF MOTION

477

see Figure 10.3. Since the cylinder is stationary, the boundary condition
(10.5.7) yields
v n Vl cos + v2 sin = 0
(10.5.12)
This is the condition to be satisfied by the velocity components of the
fluid at a point P on the lateral surface of the cylinder.

10,6
EULER'S EQUATION OF MOTION
In Section 10.4, it was shown that for a nonviscous fluid flow there are ten
field functions (namely, /?, vt and TU) for the incompressible case and sixteen
field functions (namely, p, p, e, T, vi9 qt and ,,) for the compressible case,
and as many governing equations. By combining the equations of motion
(10.4.2) and the material law (10.4.4), it is possible to eliminate the stress
components from the governing equations and thereby reduce the number
of field functions and their governing equations. In this section, we derive
the equation of motion through this method.
We start with the material law (10.4.4). Taking the divergence on both
sides of this tensor equation and using the identity (3.5.35), we obtain
divT = -V/7

(10.6.1)

Substituting this into the equation of motion (10.4.2) we obtain the


following equation of motion expressed in terms of p, p and v:
D\
_

1
__Vp + b

(10.6.2)

This equation was first obtained by Euler in 1755 and is known as Euler's
equation of motion. This equation holds for both incompressible and
compressible fluids and is a basic governing differential equation in the
theory of nonviscous fluid flows. In the case of incompressible fluid flows,
p = Po and the equation contains v and p as the field functions. In the case
of compressible fluid flows, the equation includes v, p and p as the field
functions. Thus the stresses are completely eliminated from the equation
of motion.
For a fluid at rest (static equilibrium) or in uniform motion, Euler's
equation (10.6.2) reduces to the equation of equilibrium :
-Ivp + b = 0
(10.6.3)
P
where p denotes the static pressure and p denotes time-independent density.

478

10

EQUATIONS OF FLUID MECHANICS

By using the identities (6.2.10) and (6.2.11), that is,


D\
d\

1 -,
= + y(v V)v = + w X v + - VIT
Dt
dt
dt
2

(10.6.4)

where w = curl v is the vorticity vector, equation (10.6.2) can be rewritten


in the following alternative forms:
1
d\
Vp + b = + (v V)v
p
at
d\
1 ~
= + - Vy2 + w x v
dt
2

(10.6.5a)
(10.6.5b)

Thus the Euler's equation of motion is a nonlinear partial differential


equation of the first order, the nonlinearity being caused by the acceleration
vector. This implies that in principle it is not possible to simply superpose
solutions of Euler's equation.
In the case of an incompressible fluid flow, equation (10.6.2) may be
solved under appropriate initial and boundary conditions described in
Section 10.5 to find a solution for the velocity field v for a prescribed
pressure /?; generally p is specified through boundary data. Equation of
continuity (10.4.5) serves as a constraint on v.
For compressible fluid flows, p is usually specified by a kinetic equation
of state, and equation (10.6.2) and (10.4.1) are to be solved simultaneously
under appropriate initial and boundary conditions described in Section 10.5
to find v and /?. Once v and p are so determined, the energy equation (10.4.3)
may be employed under appropriate initial and boundary conditions to
determine T. Note that is usually specified in terms of p and T by a caloric
equation of state.
EXAMPLE 1 0 . 6 . 1 A rectangular tank containing a nonviscous liquid of
constant density moves horizontally to the right with a constant acceleration. Gravitational force is the only external force. Find the pressure
distribution in the liquid and the geometrical shape of the upper surface of
the liquid.
Solution Choose the directions of the coordinate axes as indicated in
Figure 10.4. Then, D\/Dt = aex and b = - g e 3 , where a = \D\/Dt\ is a
constant and g is the (constant) acceleration due to gravity. Euler's equation
(10.6.2) now yields the following three equations for the three Cartesian
(x> y, z) components of Vp :
dp
Y = -ap
(10.6.6a)

10.6

EULER'S EQUATION OF MOTION

479

Figure 10.4. Example 10.6.1.

(10.6.6b)

3y
dp
dz

(10.6.6c)

Equation (10.6.6b) shows that/? is independent of y9 and hence (10.6.6a)


gives
p = -pax+f(z)
(10.6.7)
where/(z) is an arbitrary function of z.
Equations (10.6.6c) and (10.6.7) give/(z) = -pgz + C, where C is a
constant. Putting this/(z) into (10.6.7), we get
p = -p(ax + gz) + C

(10.6.8)

At the point where the z axis meets the upper surface of the liquid, we
have p = pa, where pa is the atmospheric pressure. If this point is at a
height A above the origin, (10.6.8) gives C = pa + pgh. Thus the pressure
distribution in the liquid is
P = Pa - P(x + gz-

gh)

(10.6.9)

For p = pa, equation (10.6.9) becomes


z =

)x + h

This is the equation of the upper surface of the liquid. Evidently, this surface
is a plane making an acute angle = ta.n~1(a/g) with the horizontal.

Note that in the limiting case when a -* 0, the liquid moves with constant
velocity and the upper surface of the liquid becomes a horizontal plane.

480

10

EQUATIONS OF FLUID MECHANICS

_--H

\ ~~ "

J>

Figure 10.5. Example 10.6.2.

EXAMPLE 1 0 . 6 . 2 A column of a nonviscous liquid of constant density


contained in a vertical circular vessel rotates like a rigid body about the axis
of the vessel with a constant angular velocity . Gravitational force is the
only external force. Find the pressure distribution in the liquid and the
geometrical form of the upper surface of the liquid.
Solution For a fluid rotating like a rigid body, the acceleration is given
by (6.2.20). Taking the axis of rotation as the z axis directed positive
upward (see Figure 10.5) and noting = we 3 is a constant angular velocity,
this relation becomes

= a>e3 x (a>e3 x x) =
~Dt

(10.6.10)

where r = xel + ye2.


Using (10.6.10) and the fact that b = - g e 3 , Euler's equation of motion
(10.6.2) reduces to
Vp = (2 - ge3)
(10.6.11)
Taking dot product with f = x/r (where r = |r|) on both sides of this
equation and noting that Vp f = dp/dr, we obtain
dp

= pr

(10.6.12)

On the other hand, if we take dot product with e3 on both sides of (10.6.11),
we find
dp
dz

(10.6.13)

10.6

EULERS EQUATION OF MOTION

481

Due to the axial symmetry of the problem, we assume that p depends only
on r and z so that
dp
dp
dp = -fdr + -fdz
(10.6.14)
or
az
Using (10.6.12) and (10.6.13) in (10.6.14) and integrating the resulting
equation, we obtain
p = \p2r2 - pgz+ C
(10.6.15)
where C is a constant of integration. At the point where the z axis meets the
upper surface of the liquid, we have p = pa, where pa is the atmospheric
pressure. If this point is at a height h above the origin, (10.6.15) gives
C = pa + pgh. Putting this value of C into (10.6.15) we obtain the pressure
distribution in the liquid as
p=pa

+ \2r2 - gp(z - h)

(10.6.16)

For/? = pa9 equation (10.6.16) becomes


z = h-^-r2
(10.6.17)
2g
This is the equation of the upper surface of the rotating liquid; the
surface is a paraboloid of revolution with the z axis as its axis and vertex
downward.

EXAMPLE 1 0 . 6 . 3 Treating the atmosphere around the earth as a nonviscous perfect gas at rest under constant gravitational field, show that the
pressure at height z above the ground level is given by
Pi exp

i-m

(10.6.18)

where px is the pressure at the earth's surface. If T = constant = 7^,


deduce that
- = - =(-^)
< 10 6 19 >
Pi
Pi
\
RTJ
where px is the density at the earth's surface.
Solution With the positive z axis vertical upward, we have b = - g e 3 ,
since the gravitational force is the only body force. The equation of
equilibrium (10.6.3) now yields
dp
-f
= o,
dx

dp
fdy

= o,

dp
-f
dz

= -gp

482

10

EQUATIONS OF FLUID MECHANICS

The first two equations show that p is independent of x and y. Hence the
last equation becomes
d

-y = -pg
(10.6.20)
az
Since the fluid under consideration is a perfect gas, we have/? = pRT by the
equation state (10.4.15), and equation (10.6.20) can be rewritten as
? +|- = 0
(10.6.21)
dz R T
Integration of this equation yields the pressure at a height z above the
ground level (z = 0) as
(10.6.22)

p = Acxp(-^ [j)

where A is a constant of integration. Using the conditions p = px at z = 0,


we find A = pl and (10.6.18) follows immediately.
If T = 0, then (10.6.18) becomes
(10.6.23)

p = Plexp(-^)

In view of the equation of state p = pRT = pRT0, (10.6.23) gives

exp

(,0 6 24)

'=a (~)

The equationp = pRT also givespx = pxRT0. Hence (10.6.24) becomes


P = PieXp(

(10.6.25)

~R0)

Expressions (10.6.23) and (10.6.25) yield (10.6.19).

EXAMPLE 10.6.4 In a part of the atmosphere lying nearer the earth's


surface, the temperature distribution is known to be of the form T =
Tx - ^ where Tx is the temperature at the ground level, z is the vertical
distance from the ground level and a is a constant. Assuming that the
atmosphere is a nonviscous perfect gas at rest under the gravitational field,
prove the following:
(i)

^_j

00.6.26)

10.6

()

EULER'S EQUATION OF MOTION

(g/Ra)-i

- = U H

Pi

(10.6.27)

l-(Ra/g)

P__
- (^
>
Pi

(,,)

483

00.6.28,

\Pu1

where px and px are the pressure and the density at the ground level.
Solution Recall that (10.6.18) is the equation for the pressure p in a
nonviscous perfect gas at rest under the gravitational field. Setting
T = Tx - ocz in this equation and simplifying it with the condition, T = Tx
at z = 0, we get

'hu

which is (10.6.26).
For z = 0, the equation of statep = pRT yields RTX = px/px.
equation p = pRT can be rewritten as
= pRTTx = ppxT
P

PXTX
so that

Pi

Pi \Tj

Hence the

(10.6.29)

Elimination of p/pl from (10.6.26) and (10.6.29) gives the result


(10.6.27). Also, (10.6.28) follows by the elimination of 77, from (10.6.26)
and (10.6.29).

EXAMPLE 10.6.5 For an adiabatic flow of a nonviscous perfect gas


with no heat supply, show that
DT
Dp _yp Dp
(10.6.30)
PCp
~Dt' ~ D~t ~ ~p D~t
Deduce that
(i) in the absence of body force,

2.(cpT+\A-l-%

Dt\p
2 ) p dt
(ii) in the absence of body force and for steady flow,
7

y-\p

- + - v2 = constant
2

(10.6.31)

(10.6.32)

484

10

EQUATIONS OF FLUID MECHANICS

Solution Recall that for a flow of a nonviscous perfect gas, the energy
equation has been expressed in two alternative forms as given by equations
(10.4.31) and (10.4.32). In the case of an adiabatic flow with no heat
supply, these equations become
Cv

DT
~Dt

p^Dp
p2 Dt

(10.6.33)

DT
~Dt

\_Dp_
' p Dt

(10.6.34)

Cp

Since p = pRT for a perfect gas, we find


DT
~Dt

R Dt \p)

pR

Dp
Dt

pDp
p Dt

Using this in (10.6.33) and taking note of (10.4.18) we get the equation
py Dp _ Dp
~p~Dt " ~Dt

(10.6.35)

Equations (10.6.34) and (10.6.35) together constitute equation (10.6.30).


In the absence of body force, Euler's equation of motion (10.6.2) gives
D\

-Vp
P

y - = - y

so that

Dt

(v2) = 2v = - 2 v - V/?
Dt
Dt
\p

(10.6.36)

Also,
Dp
Dt

dp
dt

(10.6.37)

Using (10.6.36) and (10.6.37) in (10.6.34), we obtain (10.6.31).


For steady flow, dp/dt = 0, and equation (10.6.31) yields
cpT + jv2 = constant

(10.6.38)

Since cp = yR/(y - 1) by (10.4.20b) and p = pRT, we get


c PT =

y- 1p

Results (10.6.38) and (10.6.39) together yield (10.6.32).

(10.6.39)

10.7

EQUATION OF MOTION OF AN ELASTIC FLUID

485

EXAMPLE 1 0 . 6 . 6 Show that the equation of motion of a nonviscous


fluid can be expressed in terms of x and t as follows:

*-*[*-
Solution
That is,

(10.6.40)

Treating p as a function of xf and /, we have p.t =


Vp = T(Vp)

pk^ki*

(10.6.41)

Substituting for Vp from Euler's equation (10.6.2) in (10.6.41) and


recalling that D\/Dt = D2x/Dt2, we get the equation (10.6.40).

Note: Equation (10.6.40) is the equation of motion for a nonviscous fluid


in the material (Lagrangian) form.

10.7
EQUATION OF MOTION OF AN ELASTIC FLUID
We now specialize Euler's equation of motion (10.6.2) to the case of an
elastic fluid and analyze some consequences of it. Recall from Section 10.4
that an elastic fluid is a nonviscous compressible fluid for which the kinetic
equation of state is given by (10.4.21a) or (10.4.21b).
Let us set
P=

(10.7.1)

-dp

Using (10.4.21b), that is, p = p(p), we find by virtue of the result (ii) of
Example 3.4.1 that
VP = - V p
P
Consequently, Euler's equation of motion (10.6.2) becomes

(10.7.2)

Dv

= -VP + b
(10.7.3)
Dt
This is the equation of motion for an elastic fluid. Evidently, this equation
holds for a fluid of constant density also; in this case, P = p/p9 where p is
the constant density.
Equation (10.7.3) can be expressed in an alternative form. Since/? = p(p)
for an elastic fluid, by (10.4.21a), we have

V/7 =

-f

V/9

(10.7.4)

486

10

EQUATIONS OF FLUID MECHANICS

Setting
(10.7.5)
we find from (10.7.2) and (10.7.4) that
(10.7.6)
The physical meaning of cs, which is evidently not a constant, will be given
in Example 10.7.5.
Using (10.7.6) in (10.7.3), we get the following alternative version of
(10.7.3):
^ = -^(V/>) + b
(10.7.7)
Dt
p
In many practical problems, the body force b is conservative; that is,
b = - for some scalar function (called the potential ofb). Equations
(10.7.3) and (10.7.7) then become
j = -V(P + )
= - - ( V p ) - V*
P
10.7.1

(10.7.8)
(10.7.9)

CIRCULATION THEOREM

From equation (10.7.8) we note that the acceleration is the gradient of a


scalar function. Hence, by Kelvin's circulation theorem proven in Section
6.6, it readily follows that the motion for which (10.7.8) is the governing
differential equation is circulation preserving. Thus, every motion of
an elastic fluid under conservative body force is circulation preserving.
Consequently, all the properties of circulation-preserving motions, like the
permanence of irrotational motion and transportation of vortex lines, as
discussed in Section 6.6, hold for such a motion.
10.7.2

VORTICITY EQUATION

For a continuum, the vorticity vector w is governed by Beltrami's vorticity


equation (8.2.16). For an elastic fluid, we find from (10.7.3) that

fDy\

curl I = curlb
\Dt

10.7

EQUATION OF MOTION OF AN ELASTIC FLUID

487

Substituting this in (8.2.16), we obtain the equation


^ . .
.
V)v + - c u r l b
(10.7.10)
Dt\pJ
\p
)
p
This is the vorticity equation for an elastic fluid. This equation is by
Helmholtz (1868) and Nanson (1874). If the external force is conservative,
the second term on the righthand side of the equation vanishes identically.
10.7.3

EQUATIONS OF EQUILIBRIUM

For an elastic fluid at rest (static equilibrium) or in uniform motion,


equations (10.7.3) and (10.7.7) become the equations of equilibrium:
VP = b

(10.7.11)

-(V/>) = b
P

(10.7.12)

For b = - , equation
an (10.7.11) yields
P+X=

+x= constant

(10.7.13)

EXAMPLE 1 0 . 7 . 1 For an elastic fluid in static equilibrium under the


earth's gravitational field (taken as constant), show that
dp
-- + gz = constant
(10.7.14)
P
where z is the height above the surface of the earth. Hence deduce the
following.
(i) For an isothermal perfect gas,
p
gz + log/? = constant
P
(ii) For an isentropic perfect gas,

(10.7.15)

gz + , ^ - = constant
(10.7.16)
(y - i) P
Solution Let us choose the positive z axis vertically upward from the
earth's surface. Then, if gravity is the only body force that is taken as
constant, we have b = - g e 3 = -V(gz), so that = gz is the potential of b.
Substituting = gz in the equation of equilibrium (10.7.13), we readily get
(10.7.14).

488

10

EQUATIONS OF FLUID MECHANICS

For an isothermal perfect gas, we have by (10.4.16), (p/p) = constant = A,


say, so that
= A I = A log/? + constant = log/7 + constant (10.7.17)
P
} P
P
Using this in (10.7.14), we obtain (10.7.15).
For an isentropic perfect gas, we have, by (10.4.17a), p = > where
and y(> 1) are constants, so that
+- constant
The use of this result in (10.7.14) gives (10.7.16).

constant
(10.7.18)

EXAMPLE 10.7.2 A vertical rectangular plate of width a and height b


is exposed to an atmosphere that is an isothermal perfect gas in equilibrium
under the earth's gravitational field (which is taken as constant). Show that
the magnitude of the total force acting on a face of the plate is given by
f
i 1
(10.7.19)
"exp(~^^
f =
p(z0)
g
where p(z0) is the pressure at the lower edge of the plate, and T is the
constant temperature.
R T
a

Solution Since the atmosphere to which the plate is exposed is an


isothermal perfect gas in equilibrium under the earth's gravitational field,
equation (10.7.15) holds. Making use of p = pRT in (10.7.15), we find
gz + RT\ogp = C

(10.7.20)

where C is a constant. If the lower edge of the plate is at a height z0 above


the earth's surface, see Figure 10.6, (in the particular case, z0 can be 0),
equation (10.7.20) gives
C = gzo + RT\ogp(zo)
Substituting this value of C into (10.7.20) and rewriting the resulting
equation yields
P = P(ZQ) exp

fiz-zo)

(10.7.21)

The magnitude of the total force exerted on a face of the plate due to the
pressure p is given by
f=\pdS

(10.7.22)

10.7

EQUATION OF MOTION OF AN ELASTIC FLUID

489

-ge 3

Figure 10.6. Example 10.7.2.

where S is the surface area of a face of the plate. Substituting for p from
(10.7.21) into (10.7.22) and noting that p depends only on the vertical
distance (as is evident from (10.7.21)), we obtain
Zo+b

/ =

ap(z) dz =

aRT

p(z0) 1 - expl -

gb_
RT

which is the desired result (10.7.19).


EXAMPLE 1 0 . 7 . 3 In a steady flow of an elastic fluid, show that v
satisfies the equation
<^(div v) + b v = v (v V)v = v V(u2)

(10.7.23)

Deduce that, for potential flow, the velocity potential satisfies the equation
( c f o * - <l>,k<t>,m)<i>,km + bkvk

=0

(10.7.24)

Equations (10.7.23) and (10.7.24) are usually referred to as gas dynamical


equations.
Solution

Recalling that
^ = - + (v V)
Dt
dt

the equation of motion (10.7.7) for an elastic fluid can be rewritten as


^ +
at

(v.V)v

= ^(V/>) + b
p

(10.7.25)

490

10

EQUATIONS OF FLUID MECHANICS

Similarly, the equation of continuity (10.4.1) can be rewritten as (see


equation (8.2.5a))
Y + v Vp + p div v = 0

(10.7.26)

For steady flow, dv/dt = 0 and dp/dt = 0. In this case (10.7.25) and
(10.7.26) give
v (v V)v = (?s div v + b v

(10.7.27)

By use of identity (3.4.29), we find that


v (v V)v = \V(v2) v
(10.7.28)
Identity (10.7.28) together with equation (10.7.27) yields the result
(10.7.23).
In the suffix notation, equation (10.7.27) reads
vkvmvkym

= (?-&vkik + bkvk

which can be rewritten as


(<?skm - vkvm)vkym

+ bkvk = 0

(10.7.29)

For potential flow, we have vt = </>>f, where <f> is the velocity potential.
Equation (10.7.29) then reduces to (10.7.24).

EXAMPLE 1 0 . 7 . 4 Show that for an elastic fluid moving under conservative body force the equation of motion can be expressed in the form

F^v = v - Vf[ ^ + x ~\Ad?t

(10.7.30)

where v is the initial velocity. This equation is known as Weber's equation,


after Weber who derived it in 1868.
Solution The equation of motion of an elastic fluid in the presence of
conservative body force is given by (10.7.8). Operating by the tensor F r o n
both sides of this equation, we obtain
D
DT
T
TD\
F r V(P + ) = -T- = (F r v) + - v
Dt
Dt
DT

(10.7.31)

But
DT
v = {(Vv)F}rv
Dt
by (6.2.44).

= F r (Vv) r v = \TV(y v)

(10.7.32)

10.7

EQUATION OF MOTION OF AN ELASTIC FLUID

491

Substituting this into (10.7.31) and noting that Fr(V</>) = V</> for any
function , we get the equation
Dt

(F r v) = -V(P + -

\v2)

(10.7.33)

Integrating this equation with respect to t from 0 to /, we find

[F^v]' = - V

x-\v2)dt

(P +

(10.7.34)

If we use the fact that F = I and v = v at / = 0, and note that P = Jdp/p,


equation (10.7.34) reduces to equation (10.7.30).

EXAMPLE 1 0 . 7 - 5 In the linearized case of an elastic fluid motion in the


absence of body forces, show that p, p and div v all satisfy the same wave
equation

^vV =

ay

(10.7.35)

dt2

Solution In the absence of body force, equations (10.7.2), (10.7.3) and


(10.7.7) yield
D\
7
(10.7.36)
In the linearized case, we make the approximation
D\

-Di

= P

dv
, V)v
+ (v
dt

d\

(10.7.37)

Hence (10.7.36) becomes


dv
7
Poyt = - V p = -<r8Vp

(10.7.38)

so that
P o ^ ( d i v v ) = -v2p

-c2V2p

(10.7.39)

In obtaining the term on the far right, we have neglected (Vc^) Wp.
In the linearized case, the equation of continuity (10.7.26) becomes
dp
+ p0 div v = 0
dt

(10.7.40)

492

10

EQUATIONS OF FLUID MECHANICS

Since p = p(p) for an elastic fluid, (10.7.40) can be rewritten as

or, on using (10.7.5),

$(S) + *<""-

(10 71|

p ^ + A,divv = 0.

- >

(10.7.42)

Substituting for divv from (10.7.40) and (10.7.42) in (10.7.39) we get


d 2p
1 d2p
,
, .
(10.7.43)
- = "2 TT = v /> = & P
s
dt2
c2 dt2
These show that both p and p satisfy the wave equation (10.7.35).
Taking the Laplacian of (10.7.40) and substituting for V2p from (10.7.39)
in the resulting equation, we find
c^V2(div v) = 2
(div v)
dt2

(10.7.44)

This shows that (divv) also satisfies the wave equation (10.7.35).

Note: The wave equation (10.7.35) governing p, p and divv plays a


fundamental role in the field of acoustics; it is called the acoustic wave
equation for small disturbances. The speed cs = (dp/dp)l/2 associated with
this equation represents the local speed of sound. It should be noted that cs
is not a constant, in general. For an isothermal nonviscous perfect gas (for
which (p/p) = constant), we find that cs = (p/p)l/2, which is a constant.
For an isentropic, nonviscous perfect gas (for which p = ) we obtain
cs = (yp/p)1/2 Experiments on common gases show that cs = (/)/2 is a
good approximation for computing the local speed of sound. With y = 1.4,
the value of cv for air at 0C is known to be 331.3 m/s.

10.8
BERNOULLI'S EQUATIONS
It has been shown that for an elastic fluid moving under conservative body
force the equation of motion is given by (10.7.8). Using relation (10.6.4),
equation (10.7.8) can be rewritten as
d\
+ w x v = -VH
dt

(10.8.1)

'

where
Z_T _
_ 2
H
= PD +i ,w +i \

(10.8.2)

10.8

BERNOULLI'S EQUATIONS

493

It is possible to integrate the equation of motion (10.8.1) in three important


special cases, considered next.
10.8.1

CASE I

Suppose the flow is of potential kind; that is, v = V</> for some scalar </>.
Then w = 0, and equation (10.8.1) becomes

+d

~)=

(10 8 3)

'-

This equation implies that


H + -=f(t)

(10.8.4)

where/(0 is an arbitrary function of /. Substitution for H from (10.8.2) in


(10.8.4) gives
P + \<y<i>)2 + X + Yt=f(t)
10.8.2

(10.8.5)

CASE II

Suppose the flow is steady, and the stream lines and vortex lines are
noncoincident; that is, d/dt = 0 and v x w ^ O . Then equation (10.8.1)
becomes
w x v = -VH
(10.8.6)
We note that VH is a vector normal to the surface of constant //, see
Section 3.4. Also, v x w is a vector perpendicular to both v and w. Hence,
from equation (10.8.6) it follows that v and w are tangential to a surface of
constant H. Since v and w are tangential to the stream lines and vortex lines,
respectively, it follows that these lines lie on a surface of constant H. In
other words, H is constant along stream lines and vortex lines; that is, by
(10.8.2),
P + W + X = constant
(10.8.7)
along stream lines and vortex lines.
A surface of constant H covered by a network of stream lines and vortex
lines is known as Lamb surface, after H. Lamb (1878); this is illustrated in
Figure 10.7.
10.8.3

CASE III

Suppose the flow is steady and is either an irrotational or a Beltrami


flow. (A flow is called a Beltrami flow if the velocity vector is a Beltrami
vector, as defined in Example 3.4.7.) Then d/dt = 0, and either w = 0 or

494

10

EQUATIONS OF FLUID MECHANICS

Figure 10.7. Lamb surface.

v x w = 0. Then equation (10.8.1) gives VH = 0. Since dH/dt = 0, it


follows that
H P + \v2 + = constant
(10.8.8)
everywhere in the fluid.
Thus, under the assumed conditions, relations (10.8.5), (10.8.7) and
(10.8.8) are integrals of the equation of motion of an elastic fluid. These
relations are known as the Bernoulli's equations, after Daniel Bernoulli
(1738). The function / / , defined by (10.8.2), is known as the Bernoulli
function,
Note: Since the Bernoulli's equations hold for an elastic fluid for which
P = P(P)> they automatically hold in the special case of p = constant as well.
EXAMPLE 10.8-1 For a certain flow of a nonviscous fluid of constant
density under the earth's gravitational field, the velocity distribution is
given by v = V0, where = x3 - 3xy2. Find the pressure distribution.
Solution From the given v, we
Further, since the body force is the
with = gz, where z is measured
Bernoulli's equation (10.8.8) with P

find that curl v = 0 and d\/dt = 0.


gravitational force, it is conservative
vertically upward. Accordingly, the
= p/p holds; that is,

- + -V2 + gz = C
p 2

(10.8.9)

where C is a constant.
F r o m the given v, we also find that
do
~> -y
vt = = 3(x 2 - y2),

o
v2 = = -6xy,

v3 = 0

10.8

BERNOULLI'S EQUATIONS

495

so that
v2 = v\ + v\ = 9(x2 + y2)2
Hence equation (10.8.9) becomes

- + l(x2 + y2)2 + gz = c

P 2
From this result, it is evident that C = p/p, where/?0 is the pressure at the
origin. Thus,
P = P - P[j(*2 + y2)2 + gz]
is the required pressure distribution.

(10.8.10)

EXAMPLE 10.8.2 A liquid flows out of a very large reservoir through a


small opening (Figure 10.8). Assuming that the liquid is nonviscous and of
constant density and that the flow is steady and irrotational, find the exit
speed of the liquid jet. Assume that there is no external force apart from the
gravitational force.
Solution For the given flow, the Bernoulli's equation (10.8.8) with
P = p/p and = gz, where z is measured vertically upward, holds; that is,

where C is a constant.

- + \v2 + gz = C
P 2

___^_

Figure 10.8. Example 10.8.2.

(10.8.11)

496

10

EQUATIONS OF FLUID MECHANICS

Since the reservoir is very large, at any given instant of time, the top
surface of the liquid may be assumed to be at rest. Then at a point A on this
surface (say, z = Zi) we have v = 0 and/7 = pa, wherep a is the atmospheric
pressure. Hence (10.8.11) gives
^ +
P

=c

gz

(10.8.12)

Also, at a point B on the surface of the jet (say, z = z2) we have v = ve and
P = Pa> where ve is the exit speed of the jet. Hence (10.8.11) now gives
- + \v2e+gz2
=C
(10.8.13)
P
2
Relations (10.8.12) and (10.8.13) yield the following expression for the exit
speed:
ve = -J2g(Zl - z2)
(10.8.14)
This expression is known as the Torricelli's formula; after Torricelli
(1608-1647). We note that ve is actually equal to the speed acquired by a

body falling freely (under gravity) through a height (z\ - z2)


EXAMPLE 1 0 . 8 . 3 A liquid of constant density flows down under
gravity in a long inclined pipe with a slowly tapering circular cross section.
The length of the pipe is L, the angle of inclination is a and the radius of
the upper end is N times the radius of the lower end, where N > 1. Also, the
lower end is maintained at the atmospheric pressure pa whereas the upper
end is at pressure Mpa, where M is a constant greater than 1. Assuming that
the liquid is nonviscous and that the flow is steady and irrotational, find the
exit speed of the liquid at the lower end.
Solution For the given flow, Bernoulli's equation (10.8.8) with
P = (p/p) and = gz holds, the positive z axis being taken upward as
shown in Figure 10.9. Thus,
- + \v2
P 2

+ gz = C

(10.8.15)

where C is a constant.
Let vx be the speed of entry at the upper end (z = L sin a) and ve be the
speed of exit at the lower end (z = 0). Then, for z = 0 we have v = ve and
p = pa, and for z = L sin a we have v = vx and p = Mpa. Equation
(10.8.15) therefore gives
-pa
P

+ \vl
2

=C

(10.8.16)

10.8 BERNOULLI'S EQUATIONS

497

Figure 10.9. Example 10.8.3.

and
M

1 2
2

Pa + ~ Vl + ^ S i n = C
P

(10.8.17)

so that
( M - \) + \(v\

- vl) + gLnoL = 0

(10.8.18)

This does not give ve unless vx is known. To find vx, we use the equation of
continuity, div v = 0, which holds at every point of the liquid. If V is the
volume of the liquid contained in the pipe (at any chosen instant of time),
the equation of continuity yields
(divy)dV

= 0

By divergence theorem, this becomes


v n dS = 0

(10.8.19)

where S is the surface enclosing V. We note that S is made up of the upper


and lower end surfaces of the pipe, say, A and /?, and the lateral surface of
the pipe, say, 5. Since S is a solid surface that is at rest, we have v n = 0
on S by (10.5.7). Therefore equation (10.8.19) gives, on noting that
v n = - vx on A and v n = ve on B,

vl dS + I vdS = 0

(10.8.20)

]B

We note that A and B are both circular areas such that if na2 is the area of
B, then 22 is the area of A, If we take it that the speeds on A and B

498

10

EQUATIONS OF FLUID MECHANICS

are uniform, then (10.8.20) gives {nN ) = ( )ve, so that


(10.8.21)

"*=-&"<
Substituting this back into (10.8.18), we get

vl =

2N4
(MN - 1
4

1) +
P

gLsina

(10.8.22)

which gives the exit speed at the lower end of the pipe
EXAMPLE 1 0 . 8 . 4 In a steady irrotational flow of a fluid of constant
density under zero body force, show that the pressure is maximum at a
stagnation point. (A stagnation point is a point at which the velocity is 0.)
Solution For a given flow, the Bernoulli's equation (10.8.8) with
P = p/p and = 0, holds; that is,
- + - v2 = C
P 2
where C is a constant. Consequently, since p is constant,
dp
dv =

d^p
dv2

-pv,

-P

Evidently, dp/dv - 0 for v = 0, and d2p/dv2


maximum when v = 0.

< 0 for all v. Hence p is

EXAMPLE 1 0 . 8 . 5 Consider the steady and irrotational flow of an elastic


fluid for which the pressure-density relationship is p = py, where and
{> 1) are constants. Assuming that the body force is absent, show that
(i)
(ii)

v2 +

(y - i) P

Po

(y-i)P

- 1

(7-1)/

y Po = P

(7-0/7

2c?

(iii)

= constant

vL 7/(7-0

-(,-lW

where p0 is the pressure at 0 speed.

(10.8.23)
(10.8.24)
(10.8.25)
(10.8.26)

10.8

BERNOULLI'S EQUATIONS

499

Solution (i) Under the given conditions, Bernoulli's equation (10.8.8)


holds with = 0; that is,
P + jv2 = constant

(10.8.27)

Also,

= y\py~2dp
= + constant
(y - i)
(y - i)

- + constant
P

(10.8.28)

Using this in (10.8.27), we readily get (10.8.23).


(ii) For = 0, we have p = p0. Also, the relation p = py gives
p = ((1//?))) (1/) for/7 = p0. Consequently, expression (10.8.23) yields
i 2.

_i\l/y

(10.8.29)

Substituting = p/py on the righthand side of (10.8.29) and simplifying


the resulting expression, we obtain (10.8.24).
Using the relation p = py, we find from (10.7.5) that
(10.8.30)
By use of (10.8.30), expression (10.8.24) immediately reduces to (10.8.25).
(iii) Expression (10.8.25) may be rewritten as

so that
Po
P

, (y - 1) v2
2 cl

/(7-1)

which is (10.8.26).
Note: In Example 10.6.5, expression (10.8.23) was obtained for a perfect
gas with the use of the energy equation.

500

1 0 EQUATIONS OF FLUID MECHANICS

10.9
WATER WAVES
In Section 9.13 we considered Rayleigh waves that propagate near the
surface of an elastic half-space. We now consider waves that propagate
near the surface of a nonviscous, incompressible fluid, in particular, water.
Water wave motions are of great importance; they range from waves
generated by wind or solar heating in the oceans to flood waves in rivers,
from waves caused by a moving ship in a channel to tsunami waves (tidal
waves) generated by earthquakes, and from solitary waves on the surface
of a canal caused by a disturbance to waves generated by underwater
explosions, to mention only a few. We restrict ourselves to two simple cases,
which are the starting points for the study of linear and nonlinear water
waves.
We consider a body of water (nonviscous, incompressible fluid) occupying the region -h < z < 0 with the plane z = -h as the bottom boundary
and the plane z = 0 as the upper boundary, in the undisturbed (initial)
state; see Figure 10.10. We suppose that the bottom boundary is a rigid
solid surface and the upper boundary is the surface exposed to a constant
atmospheric pressure pa. We consider plane waves propagating in the x
direction whose amplitude varies in the z direction, with the gravitational
force as the only body force. Since the motion is supposed to start from
rest, it is necessarily irrotational as a consequence of Kelvin's circulation
theorem (see Section 10.7), and we take the velocity field to be the gradient
of a potential = (, t); that is, v = V<. The equation of continuity
1,

Pa

z = 0 'LI ^
-r

*
- N

0 _y
'

z=-h

V///////////////////////////
Figure 10.10. Water waves.

10.9 WATER WAVES

501

(10.4.5) then implies that is a harmonic function at every point of the


fluid; that is,

do

* - ? ? ? = 0

(10 91)

Since the gravitational force is the only body force, which acts in the
negative z direction, we have b = - g e 3 = -V(gz), and Bernoulli's equation
(10.8.5) yields the following equation for pressure p at every point of the
fluid:
- + \ (V0)2 + gz + ^7 = /(*),
p 2
at

t > 0

(10.9.2)

Since the upper surface is exposed to the constant atmospheric pressure


pa9 we have p = pa on this surface. In the undisturbed state, the equation
of this surface is z = 0. After the motion is set up, let the equation of this
surface, denoted 5, be z = (, t) where is an unknown function of x and
/ that tends to 0 as / -* 0. The function (, t) is referred to as the surface
elevation. Thus, we have/? = pa on 5. Using (10.9.2), this condition reads

* + W + * + =/(/)
p

at

on S for / > 0. Absorbing (pa/p) and/(f) into /dt, this condition may be
rewritten as
do
Je

1
_ ( V0)2

+ W

= o

(10.9.3)

on S for > 0.
Since S is a boundary surface, it contains the same fluid particles for all
time; that is, S is a material surface. Hence it follows from (10.9.3) that
do

~Dt

= 0

on S for / > 0, or equivalently

e/ + v

^ . 1 , 2
* . v l - + -(v)'
+ ez

2
/\
l
,

= ^ + 2 V</> V( ^ ) + ^ V</> V(V</>)2 + g ^ = 0


on S for / > 0.

(10.9.4)

502

10

EQUATIONS OF FLUID MECHANICS

Since the lower boundary z = -h is a rigid solid surface at rest, the


condition to be satisfied at this boundary is, by (10.5.7),
d

7 =
dz

( 10 9 5 )

forz = -h,t> 0.
Thus, for the problem considered, Laplace's equation (10.9.1) serves as
the partial differential equation satisfied by </>, and (10.9.4) and (10.9.5)
serve as the upper and lower boundary conditions for . Once is
determined, equation (10.9.3) gives the surface elevation, z = .
Because of the presence of nonlinear terms in the boundary condition
(10.9.4), the determination of in the general case is a difficult task. We
restrict ourselves to two particular cases.
10.9.1

SMALL AMPLITUDE WAVES

We first consider the case where the motion is linear so that nonlinear terms
in velocity components may be neglected. In this case no distinction is made
between the initial and the current states of the upper boundary, and the
boundary conditions (10.9.3) and (10.9.4) are taken in the linearized forms
3
(10.9.6)
ft + OT = 0
for z = 0, t > 0;

a2o

for z = 0, / > 0.
These conditions yield

-3? 4=

(10 9 7)

(10.9.8)
for z = 0, t > 0.
For a plane wave propagating in the x direction with frequency /2 and
wavelength 2n/k, we seek in the form
= Re () exp[/(a>/ - kx)]

(10.9.9)

where () is a function to be determined.


We find that given by (10.9.9) satisfies Laplace's equation (10.9.1)
provided
2
- :2 = 0
(10.9.10)

10.9 WATER WAVES

503

The general solution of this ordinary linear differential equation is


() = Aekz + Bekz

(10.9.11)

where A and B are arbitrary constants. Using the boundary condition


(10.9.5) we get Ae~kh = Bekh. Consequently, solution (10.9.11) takes
the form
= Ccosh(z + h)

(10.9.12)

where C( = 2Ae~kh = 2Bekh) is an arbitrary constant so that the solution


(10.9.9) becomes
= Re Ccosh k(z + h) exp[i(cot - kx)]

(10.9.13)

Using (10.9.6), we find


= Rerexp[/(o>/ - kx)]

(10.9.14)

where a = (C/ig) cosh kh = max M is the amplitude. Thus, the solution


for has the final form
_ coshk(z + h)
=*<!?>
cosh kh
=

(10.9.15)

Using (10.9.15) in (10.9.7), we obtain the following dispersion relation


between the frequency and wave number:
2 = gktanhkh

(10.9.16)

This dispersion relation can be rewritten in terms of the phase velocity


c = /k as
(10.9.17)
c2 = %r = f tanh(:/0
k
k
This equation shows that the phase velocity c depends on the gravity g and
depth h as well as the wavelength 2n/k. Hence, the waves are dispersive in
nature (like the Love waves considered in Section 9.13). This means that, as
the time passes, the waves would disperse (spread out) into different groups
such that each group would consist of waves having approximately the same
wavelength. The quantity dco/dk represents the velocity of such a group in
the direction of propagation and is called group velocity, denoted C(k).
From (10.9.16) we find
C(k) = ^ = ( ^-Vtanh kh 4- kh sech2 kh)
dk
\2/

(10.9.18)

504

10

EQUATIONS OF FLUID MECHANICS

which on using (10.9.17) becomes

<l0 9

",

'H'+i^ki

Evidently, the group velocity is different from the phase velocity.


In the case where the wavelength 2n/k is large compared with the depth,
such waves are called shallow water waves, kh < 1 so that tanh kh kh
and sin 2kh 2kh. In such a situation, results (10.9.17) and (10.9.19) yield
C(k) c -Jgh

(10.9.20)

Thus, shallow water waves are nondispersive, and their speed varies as the
square root of the depth.
In the other limiting case where the wavelength is very small compared
with the depth, such waves are called deep water waves, kh > 1. In the
limit as kh - <x>, [coshA:(z + /i)]/(cosh kh) - ekz, and the corresponding
solutions for and become
= Re( ) exp[z + i(a>t - kx)] = ()ekzsin(kx

- t)

= Retfexp[/(a;/ - kx)] = acos(kx - t)

(10.9.21)
(10.9.22)

Results (10.9.16), (10.9.17) and (10.9.19) yield


2 = gk
c = (g/k)l/2

(10.9.23a)
= (gX/2n)l/2

C(k) = \c

(10.9.23b)
(10.9.23c)

Thus, deep water waves continue to be dispersive and their phase velocity
now is proportional to the square root of their wavelengths. Also, the group
velocity is equal to one-half of the phase velocity.
10.9.2.

FINITE AMPLITUDE WAVES (STOKES S WAVES)

We now consider the case where the motion is nonlinear and the amplitude
is not small. Let us recall (10.9.3) and (10.9.4) and write them for ready
reference in the form
1
2
= tfi + jiV*)
=
g
+ g<t>z]z = v + 2[V0 - ] = + \[ ()2]2 = = 0

(10.9.24)
(10.9.25)

10.9 WATER WAVES

505

where, for simplicity, we have written for /dt, for /dz, for
32/32, etc.
A systematic procedure can be employed to rewrite these boundary
conditions by using Taylor's series expansions of the potential and its
derivatives in the typical form
(10.9.26)

= , t) = ] = 0 + [] = 0 + \ [] = +

(^,

(,,

= 9t) = [] = 0 + >7[< = + M * J * = o +

(10.9.27)

Substituting these and similar Taylor's expansions into (10.9.24) gives

+ \

= --

+
=

+ ^
, +

=0

-MH,
+ \<?)2\\

+ -2
8

\--*

Jz = o

= 0

+ \

+ 0(3)

= 0

(10.9.28)

Similarly, condition (10.9.25) gives


\ + 8\ = + *[(* + 8)] + Wltott

+ 8) zzh = o + "*

+ 2[V</> ,] = 0 + 2[{ V0,}Jz = o + ^

' V*,JJZ = 0 +

+ * v(V0) 2 )], =o + i^[[V0 - v(V0) 2 )j z = o + i//2[{V0 - v ( V 0 ) 2 y , = o


+ ... = 0

(10.9.29)

We substitute (10.9.28) for into (10.9.29) to obtain


[ + 1 = - - +
22

, +

+ 2[(V</))-V0 t ] z =

- 1
^)--
g

^)?--,1
o

+ i[V^-V(V^) 2 ] z =
= 0

- -

--

, +

+ **)*]* = o

[( + 8)] =
z=0

[(y<t>
^--{
8
z=o

, +

Z= 0

,2

)-- Z = 0

,)]=0

[{()\]=0
(10.9.30)

506

10

EQUATIONS OF FLUID MECHANICS

The first-, second- and third-order boundary conditions on z = 0 are,


respectively, given by
( + 8) = 0 + 0(2)
(10.9.31)
W> + ) + ,] - -( + 8) = 0 + 0(<*>3) (10.9.32)
(,, + 8) + 2[V0 ,) + - [V0 V(V<*>)2] - -,[ + + 2( V0,)]
iZ
2

...

1
[* + g<t>zh +

+ g* z )J

= 0 + (04)
(10.9.33)
where 0( ) indicates the order of magnitude of the neglected terms. These
results can be used to determine the third-order expansion of plane
progressive waves.
As indicated before, the first-order plane wave potential in deep water
is given by (10.9.21). Direct substitution of the first-order velocity potential
(10.9.21) in the second-order boundary condition (10.9.32) reveals that the
second-order terms in (10.9.32) vanish. Thus the first-order potential is a
solution of the second-order boundary-value problem, and we can state that
= (^-jekzsm(kx

- cot) + 0(a*)

(10.9.34)

Substitution of this result into (10.9.28) leads to the second-order result for
in the form
= a cos(kx - ) - \ka2 + ka2 cos2(A:x - ) +
= acos(fo: - t) + \ka2cos2(kz

- cot) +

(10.9.35)

The second term in (10.9.35), which represents the second-order correction


to the surface profile, is positive at the crests kx - t = 0, 2, 4 , . . . ,
But the crests are
and negative at the troughs kx - = , 3, 5,
steeper, and the troughs flatter as a result of the nonlinear effect. The
notable feature of solution (10.9.35) is that the wave profile is no longer
sinusoidal. The actual shape of the wave profile is a curve known as a
trochoid (see Figure 10.11), whose crests are steeper and the troughs are
flatter than the sinusoidal wave.

Figure 10.11. The surface wave profile.

10.9 WATER WAVES

507

Substituting the wave potential (10.9.34) in the third-order boundary


condition (10.9.33) reveals that all nonlinear terms vanish identically except
but one term, (1/2) V(V</>)2. Thus the boundary condition for the
third-order plane-wave solution is given by
+ g<t>z + jV<t>' V(V0)2 = 0 + 0(4)

(10.9.36)

If the first-order solution (10.9.34) is substituted into the third-order


boundary condition on z = 0, the dispersion relation with second-order
effect is obtained in the form
Co2 = gk(l + a2k2) + O(fcV)

(10.9.37)

Note that this relation involves the amplitude in addition to frequency and
wave number. This nonlinear dispersion relation can be expressed in terms
of the phase velocity as
C= T =

( |

(l + fcV)1/2 ( I )

U -

(10.9.38)

Thus the phase velocity depends on the wave amplitude, and waves of
large amplitude travel faster than smaller ones. The dependence of c on
amplitude is known as the amplitude dispersion in contrast to the frequency
dispersion as given by (10.9.23a, b). The nonlinear solutions for plane
waves based on systematic power series in the wave amplitude are known as
Stokes's expansions.
We conclude this section by discussing the phenomenon of breaking of
water waves which is one of the most common observable phenomena in an
ocean beach. A wave coming from deep ocean changes shape as it moves
across a shallow beach. Its amplitude and wavelength also are modified.
The wave train is very smooth some distance offshore, but as it moves
inshore, the front of the wave steepens noticeably until, finally, it breaks.
After breaking, waves continue to move inshore as a series of bores or
hydraulic jumps, whose energy is gradually dissipated by means of the
water turbulence. Of the phenomena common to waves on beaches,
breaking is the physically most significant and mathematically least known.
In fact, it is one of the most intriguing longstanding problems of water
wave theory.
For waves of small amplitude in deep water, maximum particle velocity
is v = = ack. But the basic assumption of small amplitude theory
implies that v/c = ak < 1. Therefore, wave breaking can never be predicted
by the small-amplitude wave theory, and the possibility arises only in
the theory of finite-amplitude waves. It is to be noted that the Stokes's
expansions are limited to relatively small amplitude and cannot predict the

508

10

EQUATIONS OF FLUID MECHANICS

Figure 10.12. The steepest wave profile.

wavetrain of maximum height at which the crests are found to be very


sharp. For a wave profile of a constant shape moving at a uniform velocity,
it can be shown that the maximum total crest angle as the wave begins to
break is 120; see Figure 10.12.

10.10
STRESS TENSOR FOR A VISCOUS FLUID
In Sections 10.3 through 10.9 we dealt with equations governing the motion
of nonviscous fluids and some of their consequences. Let us now turn our
attention to viscous fluids. As mentioned in Section 10.2, a viscous fluid is
characterized by the property that when the fluid undergoes nonuniform
motion, a fluid surface experiences not only normal stress but shear stress
as well. In view of this property, it is postulated that the stress tensor at a
point in a viscous fluid is made up of two parts: one part due to normal
stress and the other due to shear stress. Since the distinction between viscous
and nonviscous fluids completely disappears when the fluid is at rest or in
uniform motion, we make an additional postulate that the part due to
normal stress is analogous to the stress tensor occurring in a nonviscous
fluid as given by (10.3.2) and that the part due to shear stress is represented
by a new tensor that vanishes when the motion is absent or uniform. We
refer to this new tensor as the viscous stress tensor, denoted T(t;). Recall
that in Section 6.3 we introduced the stretching tensor (strain-rate tensor) D
and made the observation that D = 0 if and only if the motion is absent or
uniform. As such, the requirement that T(r should vanish when the motion
is absent or uniform can be met by imposing the restriction: T (y) = 0
for D = 0. Thus, the following generalization of the expression (10.3.2),
proposed by George G. Stokes in 1845, is adopted in dealing with viscous
fluids:
T = -pi
with

()

+ ()

(10.10.1)

= 0, for D = 0, or in components,

= - M / + 4}>

(o.io.i)'

10.10

STRESS TENSOR FOR A VISCOUS FLUID

509

with x\f = 0 for du = 0, where are components of () and/? is a scalar


analogous to (but not necessarily identical with) the/? of (10.3.2), which is
also referred to as dynamic pressure.
The most direct way of accommodating the condition () = 0 for
D = 0 is to postulate that () is a homogeneous function o/D. When this
postulate is made, the fluid being dealt with is called a Stokesian fluid, after
Stokes. In the particular case when () is taken as a homogeneous linear
function of D, the fluid is called a linear viscous fluid. Thus, for linear
viscous fluids, it is postulated that
4 ) = aijkmdkm

(10.10.2)

where the coefficients aijkm are independent of dtj. Note that the relations
(10.10.2) are analogous to the relations (9.2.3) representing the generalized
Hooke's law.
Since and d^ are components of tensors T(y) and D, respectively, it
follows from (10.10.2) and the quotient law proven in Example 2.4.7 that
a
ijkm a r e components of a fourth-order tensor. This tensor is analogous to
the elasticity tensor and is called the viscosity tensor. Like the elasticity
tensor, the viscosity tensor depends on the physical properties of the
material, and its 81 components aijkm are called coefficients of viscosity.
From (10.10.1)' and (10.10.2), we find that these coefficients have dimensions: (force x time)/area.
An isotropic linear elastic solid has been defined as the material for which
the elasticity tensor is isotropic. Similarly, an isotropic linear viscous fluid
is defined as the material for which the viscosity tensor is isotropic. This
means that, like in an isotropic elastic solid, there is no preferred direction
in an isotropic viscous fluid and the relations (10.10.2) represent a physical
law that does not depend on the orientation of the axes. By virtue of the
general representation (2.6.1) for a fourth-order isotropic tensor, it follows
that, for a linear isotropic viscous fluid, aijkm are of the form
(iijkm = OLij^ + SikSjm + Vimjk

(10.10.3)

where , and y are scalar s.


Substituting for aijkm from (10.10.3) in (10.10.2) and recalling that
dij = d, we obtain
if = kijdkk + 2dij
(10.10.4)
where we have set = a and (l/2)(/? + ) = .
The relations (10.10.4) are analogous to the stress-strain relations (9.2.6).
It is to be noted that the symbols and appearing in (10.10.4) are not the

510

10

EQUATIONS OF FLUID MECHANICS

same as those appearing in (9.2.6). While in (9.2.6), A and denote the


elastic moduli, in (10.10.4) these are the coefficients of viscosity. Thus, for
a linear isotropic viscous fluid, there are only two viscosity coefficients.
Substituting for from (10.10.4) into (10.10.1)', we obtain
Tij = (-P + kdkk)5u

+ Idij

(10.10.5)

+ A(trD)}I + 2

(10.10.6)

In the direct notation, these relations read


T = l-p

This is the stress-strain rate relation valid for a linear, isotropic viscous
fluid. This relation was obtained by Stokes in 1845 and is also known as
Stokes*s law. This law is adopted as the material law in the theory of linear,
isotropic, viscous fluid flows. Note that, since D is symmetric, so is T by
virtue of this law. Thus, as in the cases of a linear, isotropic elastic solid and
a non viscous fluid, the symmetry of T is a physical property of a linear,
isotropic viscous fluid. A large class of liquids and gases possess this
property. As such, the theory of viscous fluid flows obeying Stokes's law is
very successful from practical point of view also. Henceforth we will refer
to a linear, isotropic, viscous fluid simply as a viscous fluid.
It is sometimes convenient to have the Stokes's law (10.10.6) expressed in
terms of T and v. Substitution for D from (6.3.3) in (10.10.6) yields
T = (-/? + div v)I + /i(Vv + Vv r )

(10.10.7)

or, in components,
T(/ = (-P + kvkik)u

+ (Ou + vjj)

(10.10.7)'

The stress-velocity relation (10.10.7) is analogous to the stressdisplacement relation (9.8.4).


We note that in the limiting case when dy -> 0, that is, when the motion
is absent or uniform, (10.10.6) reduces to the material law (10.3.2) valid for
a nonviscous fluid; p then represents the mean pressure.
Comparison of Stokes's law (10.10.6) with Hooke's law (9.2.6) reveals an
important difference between elastic solids and viscous fluids. Whereas in
the case of an elastic solid the Hooke's law (9.2.6) shows that the stresses
are all 0 when there is no deformation, in the case of a viscous fluid Stokes's
law (10.10.6) shows that nonzero stresses specified by -pi do occur even
when there is no deformation. The presence of the residual stress -pi
makes a fundamental difference between elasticity and fluid mechanics.
EXAMPLE 1 0 . 1 0 . 1 By using Stokes's law, show that at every point of
a viscous fluid, the principal directions of T, () and D are all coincident.

10.10

STRESS TENSOR FOR A VISCOUS FLUID

511

Solution At a point of a viscous fluid let a be an eigenvector of D. Then


a is along a principal direction of D and Da = Aa for some scalar A.
Stokes's law (10.10.6) then gives
Ta = Aa

(10.10.8)

where A = -p + A(tr D) + 2, which is a scalar. Hence a is an eigenvector of T as well. Consequently, a is along a principal direction of T.
Further, by use of (10.10.1) and (10.10.8), we get
T (y) a = (T + /?I)a = (A + /?)a

(10.10.9)

from which it is evident that a is along a principal direction of ()


also.

Note: This result is analogous to that proven in Example 9.2.1.


EXAMPLE 1 0 . 1 0 . 2 Show that Stokes's law (10.10.6) is equivalent to
the following relations taken together:
T ( d ) = 2//D (d)

(10.10.10a)

t r T = - 3 p + (3 + 2)()
(c

(10.10.10b)

(d)

Solution Let us recall that T and O are the deviator parts of T and D,
respectively, so that
To = T _ j ( t r T ) I
(10.10.11a)
D (rf) = D - j ( t r D ) I

(10.10.11b)

Let us now consider Stokes's law (10.10.6). Taking the trace throughout
in (10.10.6), we get (10.10.10b). Substituting for T and (tr T) from (10.10.6)
and (10.10.10b) in (10.10.11a) and using (10.10.11b), we obtain (10.10.10a).
Thus, (10.10.6) yields (10.10.10).
Conversely, with the use of (10.10.11), relations (10.10.10) yield
T = T*> + }(tr T)I
= 2{ - j(trD)I) + j { - 3 p + (3A + 2/i)(trD)}I
= -pi

+ A(tr D)I 4- 2

which is (10.10.6).

Note: This result is analogous to that proven in Example 9.2.2.


EXAMPLE 10.10.3

Let

= (A + f//)(trD) 2 + 2\)\2

(10.10.12)

51 2

10

EQUATIONS OF FLUID MECHANICS

Prove the following:

0)

(10.10.13)

()

= T (v) D

(10.10.14)

(iii)

= T D + (trO)p

(10.10.15)

(iv)

= A(trD) 2 + 2\\2
2

(10.10.16)
7

(10.10.17)

(v)

= A(div v) + //(Vv + Vv ) Vv

(vi)

(
D\
D
,
1 S)
= A(divv)2 + 2//jdiv - ^ ( d i v v ) 2 + - w 2 j (10.10.18)

(vii)

A(divv)2 - 2 / / ^ ( d i v v ) 2 + //w2

f Dv
+ 2\
UDt

dV
(10.10.19)

ndS

Expression (10.10.19) is known as the Bobyleff-Forsythe

formula.

Solution (i) We first note that the given function has the following
expression in the suffix notation
= (A + i)(dkk)2

+ 2$(&

(10.10.20)

Differentiating both sides of this expression w.r.t. dtj and noting that

and
^ (

= -^

f dkm - -kmdrr\

= (ikjm -

\kmij)

we obtain

= 2( + ^)dkkij
13

+ 4(oikoJm - \kmu)dkm

= 2{kudkk + 2pdu)
(10.10.21)

Expression (10.10.4) now yields (10.10.13).


(ii) Using (10.10.4) and (10.10.11b) we obtain
T<>D = [A(trD)I + 2//D} D
= A(tr D) 2 + 2//D(rf) D ( d ) + | / / ( t r D ) 2
= (A + f|/)(trD) 2 + 2//|D| 2
Expression (10.10.12) now yields (10.10.14).

10.10

STRESS TENSOR FOR A VISCOUS FLUID

513

(iii) Using (10.10.1) and (10.10.14) we get


T D = (-pi
which is (10.10.15).

+ T<">) D = -(tr D)p +

(iv) Expressions (10.10.15) and (10.10.6) yield


= T D + (tr ) = [A(tr D)I + 2//DJ D
= A(trD)2 + 2\\2
which is (10.10.16).
(v) Since D = sym Vv = (l/2)(Vv + Vv 7 ), we have tr D = div v and
(Vv + Vv7) Vv = 2D (Vv) = 2D (sym Vv + skw Vv)
= 2D D = 2|D| 2
Consequently, (10.10.16) yields (10.10.17).
(vi) By using result (i) of Example 6.3.1 in (10.10.16) and noting that
trD = div v, we immediately get (10.10.18).
(vii) Integrating both sides of (10.10.18) over a volume V and employing
the divergence theorem (3.6.1), we obtain (10.10.19).

Note: The function is analogous to the strain-energy function of the


elasticity theory and is called the viscous dissipative function. This
function is nonnegative for + (2/3) > 0 and > 0, as is evident from
(10.10.12).
EXAMPLE 1 0 . 1 0 . 4
field is given by

In a certain flow obeying Stokes's law, the velocity

vx = 4x^2^3,

v2 = *3,

v3 = -2x2xl

(10.10.22)

Find the strain-rate tensor and the stress tensor.


Solution

Substituting for vt from (10.10.22) in (6.3.4), we obtain


dn=4x2x3,

d22 = 0,

dl2 = 2 ^ * 3 ,

du = 2xxx2,

d33 =

-4x2x3

d23 = 2x3(l - x3)

(10.10.23)

These are the components of the strain-rate tensor associated with the given
velocity field.
Substituting for d^ from (10.10.23) into the Stokes's law (10.10.5), we
obtain
= p + 8//x 2 x 3 ,
Ti2 = 43
22 = -P,

23 = 43(\ - 3)

33 = - - 8//2*3>

13 = 4 / ^ * 2

(10.10.24)

514

10

EQUATIONS OF FLUID MECHANICS

These are the components of the stress tensor associated with the given
velocity field.
Note that unless p is specified, normal stresses are not completely
determined.

10.11
SHEAR VISCOSITY AND BULK VISCOSITY
In order to obtain the physical meanings of the coefficients of viscosity, let
us consider the flow of a viscous fluid for which the velocity field is as
follows:
v2 = o,
v3 = 0
(10.11.1)
Vl = Vl(x3)9
For this velocity field, Stokes's law (10.10.7) gives
Tu = 22 = T33 = -/?,

12 = 23 = 0

(10.11.2)
(10.11.3)

A velocity field of the form (10.11.1) occurs, for example, when a viscous
fluid bounded between two parallel horizontal plates is made to move due
to a uniform movement of the upper plate in the xl direction, keeping the
lower plate stationary (Figure 10.13). Relation (10.11.3) shows that there
occurs a shear stress in the direction of the flow on planes parallel to the flow
and that, this shear stress is directly proportional to the velocity gradient in
the perpendicular direction, with serving as the proportionality factor.
The coefficient thus represents the shear stress on a plane element parallel
to the direction of flow due to a unit of velocity gradient in the perpendicular

Moving plate

Figure 10.13.

Stationary plate
Shear viscosity.

10.11

SHEAR VISCOSITY AND BULK VISCOSITY

51 5

direction. This coefficient is known as the coefficient of shear viscosity, it


is analogous to the shear modulus in the elasticity theory. Experimental
observations on common fluids show that is a nonnegative number.
Relation (10.11.3) that defines and is a particular case of Stokes's law
(10.10.7) was proposed by Newton in 1687 on experimental grounds. This
relation is known as Newton's law of viscosity. Stokes's law (10.10.6) was
developed as a generalization of Newton's law. For this reason linear viscous
fluids (obeying Stokes's law) are referred to as Newtonian viscous fluids.
Returning to the general case, we find, from relation (10.10.4), that
r$ = 3Kdkk

(10.11.4)

where
AC = + f/i

(10.11.5)

Thus, K is the proportionality factor relating the mean viscous stress


(l/3)fj$ to the rate of dilatation dkk. (Recall the geometrical meaning of
dkk from Section 6.3.) The coefficient is thus analogous to the bulk
modulus of the elasticity theory and is called the coefficient of bulk
viscosity. Experimental observations on common fluids indicate that this
coefficient may also be taken to be nonnegative.
From (10.11.5), it follows that A represents the bulk viscosity diminished
by two-thirds of the shear viscosity. It is called the coefficient of dilatational
viscosity or second coefficient of viscosity. Thus, the coefficients and
appearing in Stokes's law have definite physical meanings. Experiments
show that these coefficients generally change with temperature and
pressure. However, these changes are usually small and in most problems
and are treated as constants.
10.11.1

STOKES'S CONDITION

Taking the trace of Stokes's law (10.10.7) we find that


p = p + K(divv)

(10.11.6)

wherep = -(l/3)(tr T) is as usual the mean pressure. Comparing (10.11.6)


with (10.3.3), we find that the dynamic pressure p of the viscous case is not
generally identical with the corresponding p of the nonviscous case.
Whereas the p of the nonviscous case is equal to the mean pressure, the p
of the viscous case represents the mean pressure plus the dilatation rate
multiplied by the bulk viscosity. We find that p = p if and only if one of the
following two conditions is satisfied:
(i)
(ii)

divv = 0

(10.11.7)

= 0

(10.11.8)

516

10

EQUATIONS OF FLUID MECHANICS

Condition (10.11.7) holds whenever the volume of each element of the


fluid remains constant during the motion. As mentioned in Section 6.3,
such a motion is called an isochoric motion and materials for which every
motion is isochoric are called incompressible materials. Thus, for an
incompressible fluid, the geometrical condition (10.11.7) always holds. In
fact (10.11.7) is the equation of continuity for an incompressible fluid as
already noted in the non viscous case.
In a compressible fluid, volumes of material elements generally change
with time and condition (10.11.7) is an impossibility. However, in this case
condition (10.11.8) can hold. Stokes studied compressible fluid flows by
assuming condition (10.11.8), and this condition is known as Stokes's
condition. Nearly all studies on compressible fluid flows make use of this
condition, and a large number of theoretical results so obtained are verified
even experimentally.
Thus, for incompressible viscous fluids and for compressible viscous
fluids with zero bulk viscosity, the dynamic pressure p and the mean
pressure/; are identical as in the case of nonviscous fluids. For compressible
viscous fluids not obeying Stokes's condition (10.11.8), the two pressures
are not the same. Dense gases belong to such a class of compressible fluids.
It is to be pointed out that Stokes's law (10.10.6), upon which the theory
of incompressible and compressible Newtonian viscous fluid flows is based,
is a linear relationship between T and D and that the linearity of the relationship is merely a hypothesis. In recent years, relationships between
T and D that are linear but more general than (10.10.6) and those that are
even nonlinear in nature have also been postulated, and based upon such
relationships some nonclassical theories of fluid flows have been developed.
Fluids for which material laws other than (10.10.6) are postulated are
referred to as non-Newtonian fluids, and the study of non-Newtonian fluid
flows is called rheology. This topic is not treated in this text.
EXAMPLE 1 0 . 1 1 . 1 For a compressible fluid with zero bulk viscosity,
show that Stokes's law (10.10.6) can be written in the following equivalent
forms:
(i)

T = -(/? + f / / t r D ) I + 2//D

(10.11.9)

(ii)

T = -pi

+ 2//D(rf)

(10.11.10)

(iii)

T ( d ) = 2^

and

=-3p

(10.11.11)

Solution When = 0, we have = -(2/3)// by (10.11.5) and Stokes's


law (10.10.6) becomes (10.11.9).

10.11

SHEAR VISCOSITY AND BULK VISCOSITY

51 7

Relation (10.11.9) may be rewritten as


T = -pi

+ 2//[D - |(trD)I} = -pi

+ 2//D 0

which is (10.11.10). Thus (10.11.9) yields (10.11.10).


From relation (10.11.10), we get, on noting that t r D ( d ) = 0,
t r T = -3/7

(10.11.12)

Hence
jid)

= T

_ i.(trT)I

= {_pi

+ 2Oid)] - j(-3/?)I = 2//D(rf)

Relations (10.11.12) and


(10.11.10) yields (10.11.11).
Relation (10.11.13) yields

(10.11.13)

constitute

(10.11.13)

(10.11.11).

Thus,

T - ( j t r T ) I = 2/4D - |(trD)I}
which on utilizing (10.11.12) becomes
T = -(p

+ \ tr D)I + 2//D

This is the relation (10.11.9). Thus, the relations (10.11.11) yield the
relation (10.11.9).
This proves that the relations (10.11.9) to (10.11.11) are equivalent and
hold for a compressible fluid obeying Stokes's condition = 0.

EXAMPLE 1 0 . 1 1 . 2 For a compressible viscous fluid with zero bulk


viscosity, show that the viscous stress tensor is identical with the stress
deviator tensor.
Solution When = 0, we have/7 = p = - ( 1 / 3 ) t r T . Consequently, we
find from equations (10.10.1) and (10.10.11a) that
T(w) = pi + T = - j ( t r T ) I + T = T ( d )

(10.11.14)

EXAMPLE 1 0 . 1 1 . 3 At a certain point of an incompressible viscous


fluid, the stress matrix is
-1
[T] =

0-2

L i

f
1

Find the pressure and the viscous stress tensor.


Solution For an incompressible fluid, we have/7 = -(1/3)**. The given
stress matrix yields p = 1.

518

10

EQUATIONS OF FLUID MECHANICS

Using the given TU and the fact that/7 = 1 in (10.10.1), we get


[T ] = [T] + p[l] = 0 - 1

(v)

Li

i.

This is the matrix of the required viscous stress tensor.

10.12
GOVERNING EQUATIONS FOR A
VISCOUS FLUID FLOW
With the material law applicable to viscous fluids formulated and analyzed
in Sections 10.10 and 10.11, let us summarize all the equations that serve
as governing equations for flows of such fluids. Since the equation of
continuity (10.4.1), the equations of motion (10.4.2a, b) and the equation
of energy (10.4.3) hold for all continua, they hold not only for nonviscous
fluids but also for viscous fluids. Let us record these equations for easy
reference.
Equation of Continuity:
Dp
- + p div v = 0
Dt

(10.12.1)

Equations of Motion:
d i v T r + pb = p

Dy_

~Dt

T = Tr

(10.12.2a)
(10.12.2b)

Equation of Energy:
De
p = T Vv - div q + ph

(10.12.3)

Next, in place of the material law (10.4.4), which is valid only for
nonviscous fluids, let us consider Stokes's law as given by (10.10.7) and
record it.
Material Law:
T = (-/? + div v)I + ( + Vv7)
(10.12.4)
As noted earlier, T is symmetric by virtue of this material law. Hence
equation (10.12.2b) is identically satisfied (as in the case of nonviscous
fluids).

10.12

GOVERNING EQUATIONS FOR A VISCOUS FLUID FLOW

519

It is easily seen that 11 scalar equations are involved in the governing


equations (10.12.1)-(10.2.4) whereas 15 unknown field functions are
present in these equations, the unknowns being 3 velocity components vi9 3
heat flux components qi9 6 stress components ( = ) and 3 scalars /?, p
and (as in the case of nonviscous fluids). The body force b and heat supply
h are taken as known functions as usual. Accordingly, equations (10.12.1)
to (10.12.4) are inadequate to determine all the field functions. Therefore,
we have to either reduce the number of unknowns or increase the number
of governing equations in order to close the system of governing equations.
These two possibilities lead us to consider the cases of incompressible and
compressible fluids as in the theory of nonviscous fluid flows.
The classification of fluids into the categories of compressible fluids and
incompressible fluids holds for both nonviscous and viscous fluids. Note
that the property of compressibility is associated directly with the change in
density of fluid elements during the motion, and the property of viscosity is
linked with the ability of the fluid to sustain shear stresses, again during the
motion. Since the change in density is always accompanied by change in
volume of elements, and shear stresses are always accompanied by change
in shape of elements, the properties of compressibility and viscosity are
not interlinked (at least theoretically). This means that a fluid can be
(i) nonviscous and incompressible, (ii) nonviscous and compressible,
(iii) viscous and incompressible and (iv) viscous and compressible. Thus,
there exist four different classes of fluids. The first two classes of fluids,
namely, incompressible and compressible nonviscous fluids, were analyzed
in Section 10.4. Let us now consider the cases of incompressible and
compressible viscous fluids to complete the task of obtaining a closed
system of governing equations for flow of such fluids.
10.12.1

INCOMPRESSIBLE FLUIDS

In the case of incompressible fluids, for which p retains its initial value p0
during the motion, the equations of continuity, motion and energy, namely,
(10.12.1), (10.12.2) and (10.12.3), reduce to the following equations
(respectively):
divv = 0
(10.12.5)
D\
divT + p0b = p0
(10.12.6)
De
/>o-^ = -divq + />/*
Also, in this case, the material law (10.12.4) reduces to
= -pl+ //(Vv + Vv7)

(10.12.7)
(10.12.8)

520

10

EQUATIONS OF FLUID MECHANICS

We note that ten scalar equations are involved in (10.12.5), (10.12.6) and
(10.12.8) and the number of unknown field functions is also ten, the field
functions being /?, vt and 7. Equation (10.12.7), which is purely an
equation of thermal energy does not contain these field functions and is
therefore not needed in their determination. Equations (10.12.5), (10.12.6)
and (10.12.8) thus serve as a closed system of governing equations for
incompressible viscous fluid flows.
10.12.2

COMPRESSIBLE FLUIDS

As in the case of non viscous compressible fluid flows, the equations of state
and law of heat conduction are appended to the material law for obtaining
a full set of constitutive equations for viscous compressible fluid flows.
These additional constitutive equations are the same as those considered in
Section 10.4. These equations are the kinetic equation of state (10.4.10),
namely,
f(p,p,T)

=0

(10.12.9)

the caloric equation of state (10.4.13), namely,


= (,)

(10.12.10)

and Fourier's law of heat conduction (10.4.12), namely,


q = -kVT
(10.12.11)
If equations (10.12.9) to (10.12.11) are appended to equations (10.12.1)
to (10.12.4), then we will have in all sixteen scalar equations for sixteen field
functions, the field functions being three velocity components vi9 three
heat-flux components qi9 six stress components TU and four scalars p9 p, T
and . Equations (10.12.1) to (10.12.4) and (10.12.9) to (10.12.11) therefore
serve as a closed system of governing differential equations for compressible viscous fluid flows. Note that, unlike the case of incompressible fluid
flows, energy equation (10.12.3) is now fully coupled with other governing
equations.
Thus, as in the case of nonviscous fluid flows, there are two sets of
governing equations for viscous fluid flows. One of these sets is applicable
to incompressible fluid flows and the other set to compressible fluid flows.
Comparison of the governing equations of viscous fluid flows with those of
nonviscous fluid flows reveals that the material law is the only governing
equation that brings out the difference between the theory of viscous fluid
flows and the theory of nonviscous fluid flows; (10.12.4) is the material law
employed for viscous fluids and (10.4.4) is the material law for nonviscous
fluids. The standard forms of equations of state, given by (10.4.14) to
(10.4.21) are employed in the theory of viscous fluids also.

10.12

GOVERNING EQUATIONS FOR A VISCOUS FLUID FLOW

521

In Section 10.11, it was pointed out that compressible viscous fluids


themselves fall into two classes depending on whether condition (10.11.8)
holds or not. When condition (10.11.8) holds, we have A = -(2/3) by
(10.11.5) and the material law (10.12.4) reduces to
T = - p i + M ( - f ) ( d i w ) I + Vv 4- (Vv) r )

(10.12.12)

This relation together with (10.12.1) to (10.12.3) and (10.12.9) to (10.12.11)


constitute the governing equations for compressible viscous fluid flows for
which Stokes's condition (10.11.8) holds.
Thus, the theory of viscous fluid flows is generally studied in three
different cases: (i) the incompressible case governed by equations (10.12.5),
(10.12.6) and (10.12.8); (ii) compressible case with Stokes's condition,
governed by equations (10.12.1) to (10.12.3), (10.12.12) and (10.12.9)
to (10.12.11); and (iii) general compressible case governed by equations
(10.2.1) to (10.12.4) and (10.12.9) to (10.12.11). The governing equations
for viscous fluid flows valid in different cases are summarized in Tables
10.3 and 10.4 with the same equation numbers as in the text.
EXAMPLE 1 0 . 1 2 . 1
rewritten in the form

Show that the energy equation (10.12.3) can be

De
p = -p(divv) - divq + ph +

(10.12.13)

where is the viscous dissipative function considered in Example 10.10.3.


Solution

By use of the material law (10.12.4), we find that


T . Vv = -/?(div v) + A(div v) 2 + //(Vv + Vv7) Vv (10.12.14)

Using (10.10.17), this becomes


T w = -p(div v) +
Substituting this in (10.12.3), we obtain (10.12.13).

(10.12.15)

Table 10.3. Governing Equations for Incompressible Viscous Fluid Flows


1. Equation of continuity:

divv = 0

(10.12.5)

D\
divT + p 0 b = p0

(10.12.6)

2. Equation of motion:

3. Material law:

T = -pi

+ //(Vv + Vv 7 )

(10.12.8)

Total number of scalar equations: 10


Total number of unknown field functions (namely, /?, and ^): 10

522

1 0 EQUATIONS OF FLUID MECHANICS


Table 10.4. Governing Equations for Compressible Viscous Fluid Flows
1. Equation of continuity:

2. Equation of motion:

Dp
+ />divv = 0
Dt

(10.12.1)

D\
divT + pb = p

(10.12.2)

De
p = T Vv - divq + ph

(10.12.3)

3. Equation of energy:

4. Material law:
= {-p + A(divv)}I + //(Vv + Vv 7 ),

for * 0

T = -pi + //{(-f)(divv)I + (Vv + Vv7)},

for = 0

(10.12.4)
(10.12.12)

5. Equations of state:
(10.12.9)

f(p,p,T) = 0

= (,)

(10.12.10)

6. Fourier's law of heat conduction:


q = -kVT
(10.12.11)
Total number of scalar equations: 16
Total number of unknown field functions (namely, />,/?, , , qit and ^): 16

Note: By using Fourier's law (10.12.11), equation (10.12.13) can be


expressed in terms of T as follows:
De
p = -p(di\\)

,
+ kV2T + ph +

(10.12.16)

EXAMPLE 1 0 . 1 2 . 2 Show that, for a perfect gas, the energy equation


(10.12.3) can be rewritten in the following forms:
DT
p Dp

pcv = -j + kV2T+
Dp
,
DT
pcp = -j^t + kV2T+
Solution

ph +
ph +

(10.12.17)
(10.12.18)

For a perfect gas, we have


= cvT

(10.12.19a)

p = pRT

(10.12.19b)

Substituting for from (10.12.19a) and for div v from (10.12.1) in


(10.12.16), we obtain (10.12.17).

10.13

INITIAL AND BOUNDARY CONDITIONS

523

Next, using (10.4.19), expression (10.12.19b) yields


cvT=cpT-^-

so that
DT
DT
cv = cpp
Dt
Dt

I Dp
p Dp
- ^ + 42 - ^
p Dt
p Dt

(10.12.20)
'

With the use of this expression in (10.12.17), we arrive at (10.12.18).

Note: In the absence of viscous effects, we have = 0, and equations


(10.12.17) and (10.12.18) reduce to equations (10.4.31) and (10.4.32),
respectively.

10.13
INITIAL AND BOUNDARY CONDITIONS
Since the field functions for viscous fluid flows are the same as those for
nonviscous fluid flows, initial and boundary conditions (10.5.1) to (10.5.5)
hold for the viscous case also.
However, because of the presence of shear stresses in a viscous fluid, the
condition to be satisfied on a rigid boundary surface S contacting the fluid
is different from the condition (10.5.6) employed in the nonviscous case.
For a viscous fluid flow, the condition (10.5.6) is usually replaced by the
stronger condition:
v = vs

(10.13.1)

on S, where v s is the velocity with which the surface S moves. For a surface
at rest, the condition becomes
v = 0

(10.13.1)'

on 5.
Note that condition (10.5.6) is just a particular case of condition
(10.13.1). Whereas condition (10.5.6) implies that the normal component of
fluid velocity is the same as that of the solid boundary (at the point of
contact), condition (10.3.1) implies such a restriction on the tangential
component also. In other words, condition (10.13.1) implies that the fluid
in contact with the solid surface must move with the surface. This amounts
to saying that the fluid must adhere to the solid and therefore cannot slip
over the surface of the solid. The condition (10.13.1) was first proposed
by Stokes and is known as the no-slip condition. In order to satisfy this
boundary condition, Prandtl (1905) made the hypothesis that within a thin

524

10

EQUATIONS OF FLUID MECHANICS

layer of fluid adjacent to the boundary the relative fluid velocity increases
rapidly from 0 at the solid boundary to the full value at its outer edge. This
thin layer is called the boundary layer within which the viscosity effects are
predominant. The condition (10.13.1) is employed as the standard boundary condition in common engineering problems. However, for high-altitude
aerodynamical problems, this condition is known to be invalid, and one
uses what are called slip conditions. We do not consider such conditions in
this text.
EXAMPLE 1 0 . 1 3 . 1 Show that on a viscous fluid surface, the stress
vector is given by (i) for incompressible fluid
s = - / m + 2( V)v + X w

(10.13.2a)

and (ii) for compressible fluid


s = (-/? + A div v)n + 2( V)v + X w

(10.13.2b)

Solution For an incompressible viscous fluid, the material law is given by


(10.12.8). Using this in Cauchy's law (7.4.8), we obtain
s = Tn = -pn + ( + Vv r )n = -pn + 2//(Vv)n + ( - Vv)n
(10.13.3)
By use of identities (3.5.34) and (3.6.19), we find that
(Vv)n = (n V)v;

(Vv r - Vv)n = n x curl v

(10.13.4)

Using these in (10.13.3), we obtain (10.13.2a).


For a compressible viscous fluid, the material law is given by (10.12.4).
Using this in Cauchy's law (7.4.8) we obtain the following counterpart of
(10.13.3):
s = Tn = -pn + A(div v)n + 2() + ( - Vv)n
Using (10.13.4) in (10.13.5), we obtain (10.12.2b).

(10.13.5)

Note: Expressions (10.13.2a) and (10.13.2b) can be utilized to rewrite the


boundary condition (10.5.2b) entirely in terms of v and/7 as follows
-pn + 2( V)v + ( X curl v) = s*

(10.13.6a)

(-/? + div v)n + 2( V)v + ( x curl v) = s*

(10.13.6b)

(incompressible case);

(compressible case), on ST for I > 0.

10.13

INITIAL AND BOUNDARY CONDITIONS

525

EXAMPLE 1 0 . 1 3 . 2 Show that at a fixed rigid solid surface contacting


a viscous fluid, the stress vector is given by
s = - / H I + //w x n

(10.13.7a)

(incompressible case);
s = l-p

+ (A + 2//)divv]n + w x n

(10.13.7b)

(compressible case). Hence, compute the normal and shear stresses exerted
by the fluid on the solid surface.
Solution At a fixed rigid solid surface contacting a viscous fluid, we have
v = 0, by the no-slip boundary condition (10.13.1)'. Hence if C is an
arbitrary simple closed curve chosen on this surface,
vxrfx = 0

(10.13.8)

By using a consequence of Stokes's theorem, given by (3.6.20),


expression (10.13.8) becomes
[(div v)n - (n V)v - n x w] dS = 0

(10.13.9)

where S is the area on the solid surface, bounded by C Since C is arbitrary,


it follows that the integrand of the surface integral in (10.13.9) should
vanish identically; hence,
(n V)v = (div v)n + w x n

(10.13.10)

on the solid surface considered.


Substituting for (n V)v from (10.13.10) in (10.13.2a) and noting that
divv = 0 for an incompressible fluid, we get the result (10.13.7a).
Similarly, (10.13.10) and (10.13.2b) yield the result (10.13.7b).
The normal stress exerted by the fluid on the solid surface is given by
=- s n

(10.13.11)

where n is directed into the fluid. For an incompressible fluid, (10.13.7a)


and (10.13.11) yield = p. Since p is equal to the mean pressure p for an
incompressible fluid, it follows that
=p

(10.13.12)

Thus, in the case of an incompressible viscous fluid, the fluid elements exert
just the mean stress on the surface, along the normal.

526

1 0 EQUATIONS OF FLUID MECHANICS

For a compressible fluid, we find by using (10.13.7b) in (10.13.11) that


a = p - (A + 2//)(div v)

(10.13.13)

which on using (10.11.6) and (10.11.5) may be rewritten as


a=p-i(divy)
(10.13.14)
Thus, in the case of a compressible viscous fluid the normal stress exerted
by the fluid on the surface is caused by the mean pressure as well as the rate
of change in volume of fluid elements.
If t is a unit vector tangential to the surface, the magnitude of the shear
stress exerted by the fluid on the solid surface is
= |s-t|

(10.13.15)

By using (10.13.7ab) in (10.13.15), we find that the following expression


holds for both incompressible and compressible fluids:
T = /i|[w,n,t]|

(10.13.16)

Thus, depends only on the vorticity vector w for both incompressible and
compressible viscous fluids.
Form (10.13.16) it also follows that, in irrotational motion, no shear
stress is exerted on a fixed solid boundary.

EXAMPLE 10.13.3 Show that, at a fixed rigid solid surface contacting


a viscous fluid, the vorticity vector lies tangential to the surface.
Solution Let C be an arbitrary simple closed curve on the surface and let
S be the area of the surface enclosed by C. Then, by Stokes's theorem,
we have
jc

vdx =

js

wndS

(10.13.17)

But by the no-slip boundary condition v = 0 on S. Hence the lefthand


side of (10.13.17) vanishes identically. Since C is arbitrary, it follows that
w n = 0 or that w is tangential to the surface, at every point.

10.14
NAVIER STOKES EQUATION
We now deduce the equation of motion of a viscousfluidby eliminating the
stresses from the governing equations.
Let us start with Stokes's law given by (10.12.4), namely,
T = (-/? + Adivv)I + //(Vv + Vv7)

(10.14.1)

10.14 NAVIER STOKES EQUATION

527

Taking the divergence on both sides of this tensor equation and using the
identities (3.5.35), (3.5.13) and (3.5.14), we obtain
div T = V(-/7 + div v) + ( 2 + V div v)

(10.14.2)

Substituting this expression into Cauchy's equation of motion (10.12.2) we


arrive at the following equation of motion expressed in terms of p, p and v:
,
D\
2 + ( + //)V(divv) - V/7 + pb = p

(10.14.3)

This equation, essentially by Navier (1822) and Stokes (1845), is referred to


as the Navier-Stokes equation. This equation holds for both compressible
and incompressible viscous fluid flows; in the incompressible case, p = p0
and div v = 0.
In the absence of viscosity, that is, if and are negligibly small,
equation (10.14.3) reduces to Euler's equation of motion (10.6.2).
Using the identity (10.6.4), namely,
Ds
d\
dv
1
= + (v-V)v = + w x v + - VIT
Dt
dt
dt
2

(10.14.4)

equation (10.14.3) can be rewritten in the following alternative forms:


2 + ( + //)V(div v) - Vp + pb = p\y

+ (v V)v

2 + (A + //)V(divv) - Vp + pb = \

+ w X v + - Vi;2 j (10.14.5b)

(10.14.5a)

For an incompressible viscous fluid, the Navier-Stokes equation


(10.14.3) is often rewritten in the following form:
.
1
D\
vV2v - -V/7 + b =

(10.14.6)

v = /.

(10.14.7)

where
The coefficient v has dimension area/time and is called kinematic viscosity.
By using (10.14.4), the equation (10.14.6) can be rewritten in two explicit
forms as follows:
1
dv
vV2v - - Vp + b = + (v V)v
(10.14.8a)
p
at
vV2v - - V p + b = -^ + w X v + -Vi; 2
p
ot
2

(10.14.8b)

528

10

EQUATIONS OF FLUID MECHANICS

When the external force is conservative, that is, if b = -V#, and p is a


constant we get the following useful form of equation (10.14.8b):
(10.14.9)
vV2v - V ( - + * + - t ; 2 ) = - ^ + w x v
\P
2 /
dt
In the case of a compressible viscous fluid for which Stokes's condition
(10.11.8) holds, we have = -(2/3)//; see (10.11.5). Then, equation
(10.14.3) involves only the shear viscosity .
Like Euler's equation of motion for a nonviscous fluid, the NavierStokes equation is the fundamental equation of motion for a viscous fluid.
Whereas the Euler's equation is a first-order partial differential equation in
spatial derivatives of velocity, the Navier-Stokes equation is of second
order in these derivatives.
It may be noted that the Navier-Stokes equation (10.14.3) is strikingly
analogous to Navier's equation (9.8.6) of classical elasticity. While (9.8.6)
is a linear partial differential equation containing the displacement u as the
only unknown, (10.14.3) is a nonlinear partial differential equation that
contains the velocity v, the density p and the pressure p as unknowns, the
nonlinearity of the equation stemming from the presence of (v V)v in the
acceleration. Thus, whereas (9.8.6) provides a complete set of field
equations for the determination of u, (10.14.3) does not provide a complete
set of field equations for the determination of v; it has to be supplemented
by the equation of continuity in the incompressible case and, further,
equations of state and an equation of heat conduction in the compressible
case.
10.14.1

REYNOLDS NUMBER

Due to nonlinearity of the Navier-Stokes equation, no general method of


solution is available. Exact solutions are known only in special cases.
Approximate solutions are often obtained either by neglecting the convective term (v V)v in comparison with the viscous term vV2v or by neglecting
the viscous term in comparison with the convective term. Thus, the ratio of
the order of the convective term to that of the viscous term plays a crucial
role in solving the Navier-Stokes equation approximately.
Suppose that, in a given flow, V is a reference speed and L is a reference
length. Then, the magnitude of the convective term (v V)v is of order
V(V/L) = V2/L and the magnitude of the viscous term vV2v is of order
(vV/L2). Thus,
l(y - v)v|
|v(V2v)|

Q(Y\

Vv /

10.14

NAVIER-STOKES EQUATION

529

Since v has dimension area/time, we note that (VL/v) is a dimensionless


quantity. This quantity is called the Reynolds number, after O. Reynolds,
and we denote it by Re. Thus, the convective term (v V)v is negligible in
comparison with the viscous term v(V2v) when Re is very small, and the
viscous term is negligible in comparison with the convective term when Re
is very large. For small values of Re, the Navier-Stokes equation takes a
linearized form; for example, equation (10.14.6) will reduce to equation
(10.14.28) given later. For large values of Re, the effect of viscosity is
significant only in boundary layers; at points outside a boundary layer the
Navier-Stokes equation is approximated to Euler's equation valid for
nonviscous fluids.
EXAMPLE 1 0 . 1 4 . 1 Find the pressure distribution such that the
velocity field given by
vi = k(x\ - xl),

v3 = 0,

v2 = 2kxlx2,

(k = constant)

(10.14.10)

satisfies the Navier-Stokes equation for an incompressible fluid in the


absence of body force.
Solution When written in the component form, the Navier-Stokes
equation for an incompressible fluid given by (10.14.6) reads as follows, on
using the identity (10.14.4) also taken in the component form:
dv*
dt
^dt 7

(
\
+

\(

d
dxl
v

^ +
dx
x

d
dx2
v

d\
dx3)

v
*lldx + v*ir)
dx23J
2

vv2y

1 dp
p dx1

2 - -IT
p dx2 + *2 (10.14.11)

dv3
(
3
d
d\
.
1 dp
-T7 + [Vi + v2 + v3 )v3 = W2v3 - - + b3
dt
\ dxl
dx2
dx3)
p dx3
Substituting the expressions for vt from (10.14.10) in equations (10.14.11)
and noting that , = 0, we obtain
dp_
dxx

-2k2pxx(xl

+ xi)

(10.14.12)

dp_
dx2

~2k2px2(xl

+ xi)

(10.14.13)

dp_
dx3

(10.14.14)

530

10

EQUATIONS OF FLUID MECHANICS

Thus, the pressure-gradient in the Jt3-direction should be 0. Accordingly,

P = P(X\ Xii s o that

dp
dp
dp = -dXi + - dx2

dxl l dx2 2
This yields, on using (10.14.12) and (10.14.13),
dp = -2k2p(x2 + xi)(xl dxl + x2dx2) = -k2pd{\(x\
Hence
p = -\k2p(x2 +xi)2 + C
y

+ x22)2}
(10.14.15)

where C is an arbitrary constant. If/?0 is the pressure at the origin, we get


p=p-

\k2p(x\ + x2)2

This is the required pressure distribution.

(10.14.16)

EXAMPLE 10.14.2 For steady flow of an incompressible viscous fluid


under a conservative body force, prove the following:
(i)

v x w = v ( - + -v2 + ) + vcurlw

(10.14.17)

(ii)

(v V)w - (w V)v = vV2w

(10.14.18)

(iii) If
v = ^

- ^2

(10.14.19)

where = (, 2), then


(vV),i

(VV),2

= vVV

(10.14.20)

Solution (i) For an incompressible viscous fluid moving under a conservative force, the Navier-Stokes equation is given by (10.14.9).
For steady flow d\/dt = 0. Also, for an incompressible fluid,
curlw = curl curl v = graddivv - V2v = -V2v

(10.14.21)

because divv = 0. Using these facts in equation (10.14.9), we obtain


(10.14.17).
(ii) From (10.14.21), we get
curl curlw = -V2w

(10.14.22)

Also, since divv = 0 and divw = 0, we find from identity (3.4.25) that
curl(v x w) = (w V)v - (v V)w

(10.14.23)

10.14

NAVIER STOKES EQUATION

531

Now, taking the curl on both sides of (10.14.17) and using (10.14.22),
(10.14.23) and (3.4.16), we obtain (10.14.18).
(iii) When v is given by (10.14.19), we get
w = curlv = - v V e 3

(10.14.24)

so that
(v V)w = ( y/i2-

- A j ( _ v V ) e 3

(10.14.25)

and
( W . V)v = 0

(10.14.26)

Using (10.14.24), (10.14.25) and (10.14.26) in (10.14.18), we obtain


(10.14.20).

Note: The function defined by (10.14.19) is known as two-dimensional


stream function.
EXAMPLE 1 0 . 1 4 . 3 A fluid motion for which the Reynolds number
is small (so that nonlinear terms in velocity are negligible) is known as a
creeping flow or Stokes's flow. For a steady creeping flow of an incompressible viscous fluid under zero body force, show that p is a harmonic
function. Deduce that defined by (10.14.19) is a biharmonic function in
this case.
Solution

For creeping flow,

D\

d\

d\

= + (v V)v
Dt
dt
dt
Then the Navier-Stokes equation (10.14.6) becomes

(10.14.27)

(10.14.28)
vV2v - -V/7 + b =
p
dt
For steady flow with zero body force, this equation reduces to
(10.14.29)
vv 2 v = - V p
P
Taking the divergence of this equation and using the equation of continuity
div v = 0, we get
V2p = 0
(10.14.30)
Thus, p is a harmonic function.
For defined through the relation (10.14.19), we have
Vi = ,2>

*>2 = - V , l

V3 = 0

532

10

EQUATIONS OF FLUID MECHANICS

Using these in equation (10.14.29), we get


Ai = ^,

Pa = -pW2(,i)

(10.14.31)

p t 2 1 = -2(1)

(10.14.32)

From these, we obtain, respectively,


P.i2 = PvV 2 (^ 2 2 );
so that
vpV2(W,n + ^ 2 2 ) = -(/?, 2 1 - /7>12) = 0
Thus, is a biharmonic function.

or

VV = 0

EXAMPLE 1 0 . 1 4 . 4 Consider a steady motion of an incompressible


viscous fluid under a conservative body force. If
0 = \2
2

(10.14.33)

+ ?- +
p

prove the following.


(i) H0 is constant along the field lines of the vector
f = (v x w) x curl w
(ii)

(10.14.34)

v V// = v(V // - w )
0

(10.14.35)

Solution For the motion considered, the relation (10.14.17) holds. Using
(10.4.33), this relation can be rewritten as
VH0 = (v x w) - vcurl w

(10.14.36)

From (10.14.34) and (10.14.36) we readily see that f VH0 = 0. Thus, VH0
is orthogonal to f and hence to the field line of f. But Vi/ 0 is always
orthogonal to the surfaces of constant H0. Hence f must be tangential to a
surface of constant H0. That is, HQ is constant along the field lines of f.
From (10.14.36), we get
V2H0 = div(v X w) = w2 - v curl w

(10.14.37)

v . v / / 0 = - vv curl w

(10.14.38)

and
These relations together yield the relation (10.14.35).

EXAMPLE 1 0 . 1 4 . 5 For a nonsteady flow of an incompressible viscous


fluid under conservative body force with curlw = V^ for some scalar
function 9 show that the Navier-Stokes equation becomes
d\
+ w X v = -VH*

(10.14.39)

10.14

NAVIER-STOKES EQUATION

533

where
H* = - + - v2 + +
P 2

(10.14.40)

Deduce the following.


(i) If the flow is of potential kind, then H* + (/dt) = / ( / ) , where f(t)
is an arbitrary function of t.
(ii) If the flow is steady and v x w ^ O , then H* is constant along stream
lines and vortex lines.
(iii) If the flow is steady and v x w = 0, then H* is constant everywhere
in the fluid.
Solution For an incompressible viscous fluid moving under a conservative body force, the Navier-Stokes equation is given by (10.14.9).
Since curl w = - V2v, the given condition, curl w = V, yields V2v = - V.
Using this, equation (10.14.9) becomes (10.14.39), with H* defined by
(10.14.40).
We observe that equation (10.14.39) is strikingly similar to equation
(10.8.1). Indeed, in the absence of viscous effects, function H* defined by
(10.14.40) reduces to Bernoulli's function H defined by (10.8.2). Following
the steps that led to Bernoulli's equations (10.8.4), (10.8.7) and (10.8.8)
from (10.8.1), we arrive at the desired results starting from the equation
(10.14.39).

EXAMPLE 1 0 . 1 4 . 6 For an incompressible viscous fluid moving under


a conservative body force, prove the following:
(i)
(ii)

V2(- + + X- v2 J = div(v x w)
V 2 ( - + ) = \yf2 - D D

(10.14.41)
(10.14.42)

Further, if the motion is irrotational, deduce that


(iii)

V V = 2D D > 0

(10.14.43)

Solution (i) For an incompressible viscous fluid moving under a conservative body force, the Navier-Stokes equation is given by (10.14.9).
If we take divergence of this equation and note that div v = 0, we get the
relation (10.14.41).
(ii) Substituting for div(v x w) from the identity (6.3.21) (with div v = 0)
in the righthand side of the relation (10.14.41), we get the relation (10.14.42).
(iii) For an irrotational motion, w = 0. In this case, subtracting (10.14.42)
from (10.14.41) we get v V = 2D D. Since D D = dudu > 0, we obtain
the inequality V2v2 > 0. Thus, (10.14.43) is proven.

534

10

EQUATIONS OF FLUID MECHANICS

EXAMPLE 1 0 . 1 4 . 7 For a flow of an incompressible viscous fluid


under conservative body force, show that the vorticity equation is given by
Dyv

= (w . V)v + vV2w

(10.14.44)

Deduce the following.


(i) If curl w = V, equation (10.14.44) becomes
= (w V)v
Dt

(10.14.45)

(ii) If the motion is two-dimensional (where v3 = 0 and vx and v2 are


independent of x3), equation (10.14.44) reduces to
= W2w
Dt

(10.14.46)

where w = w3.
(iii) If the motion is two-dimensional and in circles with centers on x3
axis, equation (10.14.46) reduces to
= W2w
dt

(10.14.47)

Solution For the given flow, the Navier-Stokes equation is given by


(10.14.6) with b = - V # . Taking curl on both sides of this equation, we get
c u r l ( = ^ J = vV2w

(10.14.48)

Using this expression in Beltrami's vorticity equation (8.2.16), we obtain


the equation (10.14.44).
(i) If curlw = V, then the identity (3.4.27) yields V2w = 0. Consequently, equation (10.14.44) reduces to equation (10.14.45).
(ii) If v3 = 0 and vx and v2 are independent of x3, we readily find
Wl

= w 2 = o,

w3 = 2 - vU2

(10.14.49)

Consequently,
( w V)v = 0

(10.14.50)

and equation (10.14.44) reduces to equation (10.14.46).


(iii) If the motion is two-dimensional and in circles with centers on the x3
axis, we have v = vee, with v = v(R, /), where R is the radial distance
parallel to the x1x2 plane and ee is the unit vector in the transverse direction
(see Figure 10.14).

10.14

NAVIER STOKES EQUATION

535

( X 1 #X 2 )

Figure 10.14. Example 10.14.7(iii).

From Figure 10.14, we find that


ee = (-sin)ej + (cos0)e2 = - - ^ + ^ e 2
Thus,
v = ve0 = v[-x2el + ^2\
so that
Vz =

"' = -R X>'

RXl'

"3 =

(10.14.51)

These give, on noting that v = v(R, t) and Rti = x,/R, i = 1,2,


_ v
1 dv
~~R~R
d~RR-lXl

Vl 2

2,

v
R

1 dv
RdR

v
_ v
dv 2
2 2 2
T * * = ~R-"R~2 d~RXl + ~>
R3-X2
(10.14.52a)

v
Rl

v
R

1 dv 2
R2 dR

R*
(10.14.52b)

Using these in (10.14.49), we obtain


v
W =

dv
R dR

(10.14.53)

(where w = w3), which yields


d2v
1 dv
v Xi
w,; = JR2 ~ d~R ~ ~R*_
R'

/ = 1,2

From (10.14.51), (10.14.53) and (10.14.54) we find that


(v V)w = vxw + v2w2 = 0

(10.14.54)

536

10

EQUATIONS OF FLUID MECHANICS

Consequently,
Dw
dw
~Dt=~dt

(10.14.55)

Substituting this in (10.14.46), we obtain (10.14.47).


Note: Using (10.14.53) it can be easily shown that
(10.14.56)
Then equation (10.14.47) takes the polar form
dw

d2w
1 dw
= v 'dR2 + R^R

(10.14.47)'

It can be verified that a solution of this partial differential equation is


A
w = - exp
Snvt

R2\
Avt

(10.14.57)

where A is a nonnegative constant. This solution shows that w decays


rapidly with time, a phenomenon called diffusion of vorticity.
EXAMPLE 1 0 . 1 4 . 8 Show that the rate of decrease in kinetic energy due
to viscosity in a finite volume V of an incompressible fluid is given by
w2dV-\

= \

n-(vxw)rfS

(10.14.58)

where 5 is the boundary of V.


If S is a rigid solid surface at rest, deduce that
yf2dV=

W = \

(10.14.59)

dV

Solution From (8.6.1), we recall that the kinetic energy of a continuum


contained in a volume V is given by
K = i\

p(y\)dV

Using (8.2.10), we obtain


DK

1 f

DY

(10.14.60)

10.14

NAVIER STOKES EQUATION

537

For an incompressible viscous fluid, expression (10.14.60) becomes, on


using the Navier-Stokes equation (10.14.6),
= \

v-V 2 vrfK-

p\-VpdV+\

py-bdV

(10.14.61)

The last two terms on the righthand side of this expression represents the
rate of work done by the pressure and the body force, and the first term
represents the contribution of viscosity to the rate of change of kinetic
energy. Therefore, if W denotes the rate of decrease in kinetic energy due
to viscosity, we have
W = -\

\-V2\dV

(10.14.62)

Since the fluid is incompressible, we note by using identities (3.4.27) and


(3.4.24) that
V2v = -curl curl v = -curl
(10.14.63)
f_ (
div(v X w) = w22 - v curl
Iw j
In view of these expressions, (10.14.62) becomes
W = \

w z dV - \

div(v Xw)rfF

(10.14.64)

This expression yields (10.14.58), on using the divergence theorem (3.6.1).


If S is a rigid solid surface at rest, then by the no-slip boundary condition
we have v = 0 on S. Consequently, the surface integral on the righthand
side of (10.14.58) becomes identically 0, and we obtain
W= \

(10.14.65)

vtdV

which is the first part of (10.14.59). To obtain the other part, we recall
expression (10.10.19). For an incompressible fluid, this expression becomes
= \

/2+2\

^ndS

(10.14.66)

If 5 is a rigid surface at rest, the surface integral in the expression (10.14.66)


vanishes and we get
= \ w W
(10.14.67)
v
}v
which is the second part of (10.14.59).

Expression (10.14.59) explicitly exhibits the remarkable relationship


between the kinetic energy, the vorticity and the viscous dissipation of a
fluid moving within a rigid enclosure.

538

10

EQUATIONS OF FLUID MECHANICS

EXAMPLE 10.14.9 For an incompressible viscous fluid moving under


a conservative body force, show that the circulation Ic round a circuit c
moving with the fluid is not constant in general. Deduce that Ic is constant
if and only if curl w = V, for some .
Solution For the given flow, the Navier-Stokes equation is given by
(10.14.6) with b = -V*.
From (6.6.4), we recall that the rate of change of circulation round a
material circuit is given by
DIr

Dt

^ * dx
) c Dt

(10.14.68)

Substituting for Ds/Dt from (10.14.6) with b = -V* in (10.14.68), we get


^

= v(t v V t f x

(10.14.69)

Since the fluid is viscous, v ^ 0 and consequently DIc/Dt ^ 0 when


V2v 5* 0. Thus, Ic is not constant in general.
Since V2v = -curlw for an incompressible fluid, expression (10.14.69)
can be rewritten as
DIC
= - v curlw dx
(10.14.70)
Evidently, Ic = constant if and only if curl w = V for some scalar .
(Thus, Kelvin's circulation theorem is not generally valid for viscous
fluids.)

10.15
SOME VISCOUS FLOW PROBLEMS
As remarked earlier, the presence of the nonlinear term in the NavierStokes equation makes the exact solution of the equation very difficult
except for a very few cases. In this section we consider some simple
examples of flows of an incompressible viscous fluid for which the NavierStokes equation admits exact solutions. The linearized Navier-Stokes
equation is employed in the last example.
10.15.1 STEADY LAMINAR FLOW BETWEEN
PARALLEL PLATES

First, we consider an incompressible viscous fluid bounded between two


rigid infinite flat plates distance h apart. We suppose that the fluid flows
steadily in a fixed direction parallel to the plates. Such a flow is called a

10.15

SOME VISCOUS FLOW PROBLEMS

539

*3
x3 =h

^ve<
x3 = 0
Figure 10.15. Steady laminar flow.

steady laminar flow. Assuming that there is no body force, the problem is
to find the velocity field in the fluid.
We take the coordinate axes such that the xl axis is along the direction of
flow and the x3 axis is normal to the plates, with JC3 = 0 and x3 = h as the
planes containing the plates; see Figure 10.15. It is convenient to refer to the
plate in the plane x3 = 0 as the lower plate and that in the plane x3 = h as
the upper plate. Then the velocity is of the form \ = vel9 where
v = v(xl9x2,x3),
and the Navier-Stokes equation (10.14.6) gives the
following equations for v and p , :
1

vV2v

p >*

Dv

Dl

(10.15.1)

pt2 = 0,

(10.15.2a)

Pt3 = 0

(10.15.2b)

The equation of continuity (10.12.5) becomes


= 0

(10.15.3)

If the fluid extends to infinity in the x2 direction, then the variation of v


in the x2 direction may be neglected. Consequently, it follows that v varies
only with x3, and
Dv
dv
+ (v V)v = 0
Dt
dt
Thus, under the assumptions made, inertia plays no role in the flow. From
(10.15.2), it follows that p = p(xi), and equation (10.15.1) reduces to the
ordinary linear differential equation
l

dxl

dp_
dxx

(10.15.4)

540

10

EQUATIONS OF FLUID MECHANICS

Thus the flow is induced entirely by the pressure gradient, dp/dxx = p'(xx)
in the xx direction.
Integrating (10.15.4) successively twice, we obtain the following explicit
expression for v:
= i* 3 V(*i) + XiMxi)

+(*)

(10.15.5)

where/! a n d / 2 are arbitrary functions of xx. If we suppose that the lower


plate (x3 = 0) is stationary, then, by the no-slip boundary condition, we
have v = 0 forx 3 = 0 and (10.15.5) g i v e s y ^ ) = 0. If we compute from
(10.15.5) and use (10.15.3) we get
*3/>"(*i) = -2fi(xx)

(10.15.6)

Since this equation holds for all xx and x3 in 0 < x3 < , we must have
p"(xx) = 0 and//(jc!) = 0 so that
fx(xx) = Cx

(10.15.7a)

p(Xl) = c2 - Gxx

(10.15.7b)

where CXi C2 and G are arbitrary constants. All these results allow us to
rewrite (10.15.5) in the form
= Cxx3 ~\Gxl

(10.15.8)

This shows that the flow occurs only when at least one of the constants Cx
and G is nonzero.
Since vx = v(x3), v2 = v3 = 0, we find from the material law (10.12.8) that
= ^

(10.15.9)

which, by using (10.15.8), reduces to


31 = Cx - Gx3

(10.15.10)

When the constants Cx and G are known, (10.15.8) gives the velocity
distribution and (10.15.10) gives the shear stress on plane elements parallel
to the plates. Evidently, (10.15.8) represents the parabolic profile of the
flow between the plates. Expression (10.15.10) shows that the shear stress
31 varies linearly with height (increasing x3 direction) with
Cx = [r31L3 = 0

(10.15.11)

as the shear stress at the lower plate. Thus the constant Cx involved in
the solution (10.15.8) has a definite physical interpretation. It can be seen
from (10.15.7b) that the other constant G in (10.15.8) also has physical
significance; in fact,
G = -p'(xx)
(10.15.12)

10.15

SOME VISCOUS FLOW PROBLEMS

541

Thus, G represents the negative of the pressure gradient in the direction of


the flow. Since G is a constant, the pressure gradient must be a constant for
the flow. Further, since the flow takes place in the xx direction, the pressure
does not increase in that direction; that is, G > 0.
Two particular cases of this problem are of interest.
Case i: Plane Couette Flow Suppose the flow is generated by the
movement of the upper plate with a constant speed v0 in the xx direction.
Such a flow is called plane Couette flow. Then by the no-slip boundary
condition, we have v = v0 for x3 = h, and (10.15.8) gives
(10.15.13)
Putting this expression back into (10.15.8), we obtain
= - Gx^h - x3) + v0

(10.15.14)

This determines the velocity distribution completely, provided the constant pressure-gradient G = -p'(xx) is known beforehand. For example, if
this pressure-gradient is 0, then
(10.15.15)

v = -x,

This represents a linear distribution of velocity depicted in Figure 10.16a.


In the figure, a few velocity vectors are indicated to make the flow easier to
visualize. Such a diagram is called a velocity profile. The velocity profile for
p'(Xi) < 0 is given in Figure 10.16b.
Expression (10.15.14) can be used to find the mass flow rate per unit of
width ; that is the mass of fluid that passes the plane xx = constant per unit
of distance in the x2 direction and per unit of time. This is given by
M

3 =
- v o-1

pv0
x3(h - x3) dx3 +

Vp'(x,) = 0

E
(a)

x* dx* =

Gh3
T2v

r~ v o-i
^ H p'(x,)<0

(b)

Figure 10.16. Velocity profiles for plane Coutte flow.

pv0h
2
(10.15.16)

542

10

EQUATIONS OF FLUID MECHANICS

In the case of /?'(*,) = 0, we get


(10.15.17)

=0
Thus, in this case, M is independent of viscosity.
From (10.15.10) and (10.15.13), we get
T31=^V0

(10.15.18)

+ G(^-X3)

This expression gives the shear stress on a fluid element parallel to the plates
(for Couette flow). The shear stress exerted on the plates is often called the
skin friction. The skin friction on the lower plate is given by
(10.15.19)
and that on the upper plate is given by
(10.15.20)
n

W h e n / ? ' ^ ) = 0, (10.15.18) gives


3 , =^0

(10.15.21)

Evidently, in this case, all fluid particles, including those in contact with the
plates, experience the same shear stress. This stress is directly proportional
to vQ and inversely proportional to A and acts in the direction of the flow.
Case ii: Plane Poiseullie Flow Suppose the upper plate is also held
stationary and the flow is generated only by the constant pressure gradient
p'(Xi) < 0. Such a flow is called a plane Poiseullie flow. Then v0 = 0
and G > 0, and (10.15.14) yields the following expression for velocity
distribution:
= \Gx3(h - x3)

(10.15.22)

Evidently, the velocity distribution is now symmetrical about the midplane


x3 = A/2 and the velocity is maximum on this plane, with
* W = *3*/2 = | ; 2
The velocity profile is depicted in Figure 10.17.

(10.15.23)

10.15

SOME VISCOUS FLOW PROBLEMS

543

,,

p (x,)<0

Vo=0

Figure 10.17. Plane Poiseullie flow.

Substituting for v from (10.15.22) in (10.15.9) we get the following


expression for the shear stress on a fluid element parallel to the plates:
T3I = G Q - J C 3 )

(10.15.24)

We observe that 31 = 0 on the midplane x3 = A/2 and that 31 > 0 for


0 < x3 < h/2 and 31 < 0 for /2 < x3 < . That is, plane elements laying
below the midplane experience a shear stress along the direction of the flow
while those above experience a shear stress in the opposite direction, with
the elements laying on the midplane remaining stress free.
From (10.15.24), we find that the skin friction on the lower plate is
h"3iL3 = o
and that on the upper plate is
taiL3 = /!

Gh
2

(10.15.25)

Gh
' 2

(10.15.26)

The magnitudes of the skin frictions on both the plates are equal to Gh/2.
But, while the lower plate experiences a friction along the direction of the
flow, the upper plate experiences a (backward) drag.
Expression (10.15.22) can be employed to find the mass flow rate per unit
of width. This is given by
[h
G [h
Gh3
M = \ pvdx3 = x3(h - x3)dx3 =
(10.15.27)
Evidently, for given G and A, M varies inversely as v. Expression (10.15.27)
can be used to measure v.
10.15.2.

STEADY FLOW IN A STRAIGHT CONDUIT

We now consider the steady flow of an incompressible viscous fluid through


a straight conduit of uniform cross section under a constant pressure
gradient along the direction of flow. Such a flow is called the HagenPoiseullie flow. If we choose the axes such that the x3 axis is along the

544

10

EQUATIONS OF FLUID MECHANICS

direction of the conduit so that a right cross section of the conduit is


bounded by a simple closed curve C parallel to the xlx2 plane, then
the velocity at any point of the fluid is of the form v = ve3, where v =
(92,3),
and the equation of continuity (10.12.5) yields v3 = 0.
Consequently, v = {, x2) and (v V)v = 0. Since the flow is steady, we
get (Dy/Dt) = 0. Under these conditions the Navier-Stokes equation
(10.14.6) yields the following three component equations in the absence of
body force:
= 0

(10.15.28a)

pf2 = 0

(10.15.28b)

p 3 = M(Vn

+ v22)

(10.15.28c)

The first two of these equations show that/? = p(x3). It is assumed that the
pressure-gradient is a constant; hence,
^- = -G
(10.15.29)
dx3
where G is a constant. Note that, for the fluid to flow in the x3 direction,
G has to be positive.
Equations (10.15.28c) and (10.15.29) yield the following governing
equation for v:
(10.15.30)
If the conduit is stationary, the no-slip boundary condition yields
v =0

(10.15.31)

on C; see Figure 10.18.


We note that (10.15.30) is a two-dimensional Poisson equation. When
solved under the boundary condition (10.15.31), this equation determines v
uniquely. Once v is determined, the stresses developed in the fluid can be
computed by use of the material law (10.12.8).
We now consider two particular cases for further analysis.
Case i : Circular Cross Section Suppose the conduit is of circular cross
section. Then C is a circle whose equation can be taken as
(10.15.32)
where a is the radius of the conduit.
Since v = 0 on C by (10.15.31), the equation (10.15.32) of C suggests
that v is of the form
(10.15.33)

10.15

SOME VISCOUS FLOW PROBLEMS

545

Figure 10.18. Cross section of a conduit.

where a is a constant. Substituting for v from (10.15.33) into equation


(10.15.30), we find that a = -G/. Thus,
(10.15.34)

is the expression for v that satisfies the governing equation (10.15.30) as


well as the boundary condition (10.15.31). We note that the velocity distribution is in the form of a paraboloid of revolution; the velocity profile in
a section parallel to the length of the pipe is depicted in Figure 10.19. We
find that the velocity is maximum on the axis of the conduit, namely the * 3
axis, and
G
4

(10.15.35)

'( 3 )<0
*3

Figure 10.19. Velocity profile for case (i).

546

10

EQUATIONS OF FLUID MECHANICS

Expression (10.15.34) can be used to find the rate of mass of fluid that
passes a cross section (x3 = constant); this rate is given by

pvdA = 2 I vRdR
M= I pvdA=2np\

(10.15.36)

JO

where A is the area of cross section, and R = (xl + x|) 1 / 2 . Using (10.15.34)
in (10.15.36), we get
M = ^ (10.15.37)
8
Consequently, the rate of volume of fluid that passes a cross section
(JC3 = constant) is
M nGa4
(10.15.38)
8/1

This is known as Poiseullie's formula. This formula is generally used to


measure the viscosity of a fluid.
If we consider a cylindrical fluid surface element dS coaxial with the
conduit and having radius R, then the components of the exterior unit
normal to this element are
=^>

"2=JR>

"3

(10.15.39)

The components of the stress vector on dS can be computed by use of


Cauchy's law (7.4.9), the material law (10.12.8) and expressions (10.15.34)
and (10.15.39). The stress components st thus obtained are
(10.15.40a)

s1 = -p^
s2 = -p^
s3 = -^R

(10.15.40b)
(10.15.40c)

Together with (10.15.39), expressions (10.15.40ab) yield the normal stress


on dS as a = s n = -p, as expected and expression (10.16.40c) gives the
shear stress on dS along the conduit. Evidently, this shear stress acts
opposite to the direction of flow. It is to be noted that all the three stress
components st are independent of viscosity. If dS is taken on the boundary
surface, we get -(G/2)a as the skin friction along the direction of flow.

10.15

SOME VISCOUS FLOW PROBLEMS

547

Case ii : Elliptic Cross Section Suppose the conduit is of elliptic cross


section. Then C is an ellipse whose equation can be taken as
x2
x2
4 + ^= 1
(10.15.41)
a
b"
In view of the boundary condition (10.15.31), v can be assumed in the form
V =

(^

$~1)

(10.15.42)

where is a constant. Substituting for v from (10.15.42) into equation


(10.15.30), we find that

Thus,
v=

(10 15 43)

= -^j

a2b2G
2( + bz)

a2

(10.15.44)

b2

is the expression for v that satisfies governing equation (10.15.30) as well as


boundary condition (10.15.31). We note that the velocity distribution is
paraboloidal in nature. Also, the velocity is maximum on the axis of the
conduit, and t>max = -/?.
Using (10.15.44), we find that the rate of mass of fluid that passes a cross
section is given by

M=

)r

dA=

pa2b2G

[ / dA x2

y1

2^v^)X-7-^

( 5 45)

where A is the area of cross section. Setting


xx = cos 0,

x2 = bt, sin

(10.15.46)

and noting that varies from 0 to 1 and varies from 0 to 2 as the point
(Xi,x2) varies over A, we find from (10.15.45) that
M

pa2b2G

= ^TT2

[2 [x /Jr2

lxr

pna3b3G

( - 1) = -f2

-2- (10.15.47)

2(2 + b2) Jo Jo
4(*2 + b2)
This is a generalization of Poiseullie's formula (10.15.38). If we set b = a in
(10.15.47), we recover (10.15.38).
Let us consider a fluid element dS on an elliptic surface coaxial with and
similar to the boundary surface of the conduit. On this surface element,
x2
x2
(92) = - + -A = constant

(10.15.48)

548

10

EQUATIONS OF FLUID MECHANICS

The exterior unit normal to dS is therefore n = (V</>/|V</>|). From (10.15.48)


and (10.15.46), we find that
2xx
2x2
_ 2
= ej + e2 = cos e, + a sin e2)
\\=^-

ab

(b2 cos2 + a2 sin2 )/2

Hence the components of n are


b
n1=

cos0,

a
AI2 = sin0,

n3 = 0

(10.15.49)

where
N = (b2 cos 2 6> + a2 sin2 0) 1/2

(10.15.50)

The components of the stress vector on dS can now be computed by using


Cauchy's law (7.4.9), the material law (10.12.8) and expressions (10.15.44),
(10.15.46) and (10.15.49). We find that
i = \\\

b
+ 22 = ~Pn\ = -p cosd

(10.15.51a)

s2 =

a
+ 222 = -pn2 = -/? sin0

(10.15.51b)

Tlxnx

53 = Tax! + 322 = (

GabN
+ 2) = - g 2 + bi

(10.15.51c)

Together with (10.15.49), expressions (10.15.51a,b) yield the normal stress


on dS as = s n = - / ? , as expected. Expression (10.15.51c) gives the
shear stress on dS along the conduit. Evidently, this stress acts opposite to
the direction of flow. It may be observed that all the stress components 57
are independent of the viscosity.
If dS is chosen on the boundary surface, for which = 1, we get
S3 = , where
Gab
T = 2
nN
(10.15.52)
2
a + b2
This gives the skin friction at a point on the conduit and this varies from one
point to another. By using (10.15.50), it can be shown that the skin friction
is maximum at the endpoints of the minor axis and minimum at the
endpoints of the major axis of the boundary curve (10.15.41). If a > b,

10.15

SOME VISCOUS FLOW PROBLEMS

549

the maximum and minimum values of || are


i i

G( lb

i i

Gal 2

>

,**

*r

max = - s
-^;
min = -*
-*
(10.15.53)
a 2 + b2
a2 + b2
When b = a, we have N = a and = R/a and expressions (10.15.51)
reduce to (10.15.40). Also, then, becomes equal to -(G/2)a at all points
of the boundary.
10.15.3 STEADY FLOW BETWEEN TWO COAXIAL
ROTATING CYLINDERS
Here we consider the steady flow of an incompressible viscous fluid between
two infinitely long coaxial circular cylinders due to the rotation of the
cylinders with constant but different angular velocities about the common
axis. Such a flow is called a Couette flow. Assuming that there is no body
force, the problem is to find the velocity field.
We choose the axes such that the x3 axis coincides with the axis of the
cylinders and the xlx2 plane lies on a section of the tube formed by the
cylindrical surfaces. Since the flow is caused by the rotation of the cylinders
about the axis and the flow is steady, we assume that the fluid particles
move in circular orbits centered on the axis and that the magnitude of
velocity depends only on the distance of the particle from the axis. Accordingly, the velocity at any point of the fluid is taken in the form
v = v(R)ee

(10.15.54)

where R = (xf + x\Y/2 is the distance of the point from the axis and e 0 is
the unit vector in the transverse direction (see Figure 10.20). If we denote
the unit vector directed along the radial direction perpendicular to the axis
by eR, then ee = e3x eR, so that
[eel/ = sUk[e3]j[eR]k
for i,k=

= ei3k xk

(10.15.55a)

1,2;
[ej3 = 0

(10.15.55b)

Consequently, (10.15.54) gives


Vi = ei3kv(R) xk = ei3ky/(R)xk

(10.15.56a)

for i, k = 1,2;
v3 = 0

(10.15.56b)

550

10

EQUATIONS OF FLUID MECHANICS

Figure 10.20. Cross section of a circular tube.

From (10.15.56a), we obtain the following relations valid for i9j9 k = 1,2:
1 dw
Vu = *idjW{R) + CiucXkXj-

=e

(10.15.57a)

d2w)

(3 dw

(1(U5 57b)

4*5! ssT*

i,JVJ ~

(10.15.57c)

i3jej3kXkV/

[The reader may note that explicit forms of (10.15.56a) and (10.15.57a) are
contained in (10.14.51) and (10.14.52ab).]
From (10.15.57a) and (10.15.56b) we find that divv = 0. Thus, the
equation of continuity is satisfied. Also, (10.15.55ab), (10.15.56b) and
(10.15.57bc)give
1
(V v ) - e , = vUJ[ei3k-xk)

[3dR+RdR~\

LA.

d
_d_
(J?V)
~ dR
dR RdR
{(v V)v} e e = vuvj[ei3k

xk ) = 00

>

(10.15.58a)
(10.15.58b)

10.15

SOME VISCOUS FLOW PROBLEMS

551

Because of the axisymmetric and steady nature of the flow, we suppose


that the pressure is a function of R only so that
dp
Vp =

dRe"

from which it follows that


(V/?)-e = 0

(10.15.59)

Taking the scalar product with ee on both sides of the Navier-Stokes


equation (10.14.6) with body force equal to 0 and using expressions
(10.15.58) and (10.15.59) in the resulting equation, we obtain the following
ordinary linear differential equation for the function v(R):
_d_
dR

*<*

= 0

(10.15.60)

Integration of this equation yields


v(R) = AR + ^

(10.15.61)

where A and B are arbitrary constants.


We verify that the component of the Navier-Stokes equation (10.14.6)
along eR yields, on using (10.15.57bc),
(10.15.62)
and that the component along e3 is identically satisfied.
Thus, under the assumptions made, (10.15.61) gives the velocity field and
(10.15.62) gives the pressure-gradient field when the constants A and B in
(10.15.61) are determined by the boundary conditions.
Let a and b be the radii and and 2 be the constant angular speeds of
the inner and outer cylinders, respectively. Since the cylinders rotate rigidly
about the x3 axis, the velocities at points on these cylinders can be determined by (6.2.19). Thus, we find that laee and 2bee are the velocities at
a point on the inner and outer cylinders, respectively. The no-slip boundary
condition then yields v(R) = for R = a and v(R) = lb for R = b.
Using (10.15.61), these conditions yield the following two equations for the
determination of the two constants A and B :

(10.15.63)

552

10

EQUATIONS OF FLUID MECHANICS

Solving these equations we obtain


A =

2b2 - 2

tf-a

( - 2) 2u2
b

B =

'

(10.15.64)

b2

Substituting these into (10.15.61), we obtain


v(R) =

1
a

(co2b - xa)R

a2b2
+ ( - 2)R

(10.15.65)

This determines the velocity field in the fluid. We immediately note that,
if both the cylinders rotate with the same angular velocity a>e3, then
v(R) = R. Hence, in this case, the fluid rotates like a rigid body along
with boundary surfaces.
Substituting (10.15.65) in (10.15.62) we obtain the pressure-gradient at a
point of the fluid.
Since the motion is essentially rotational, it is of interest to compute the
vorticity at a point. By use of (10.15.56ab), we find that
0,
Wi = [curlv]f = eUkvkJ

for / = 1,2

= \ 1 d

RM'

f r = 3

'

Consequently, with the use of (10.15.65), we obtain


w = curl v =

2(2b2 - , 2 )
-j
y
e3
bL - a1

(10.15.66)

Evidently, the vorticity is constant and is directed along the axis. Further,
the flow is irrotational if and only if the angular speeds of the cylinders obey
the relation
,2

co2 _ a

(10.15.67)

The stresses produced at a point of the fluid can be computed by use of


(10.15.56) in Stokes's law (10.12.8). Thus, we find

for ij

=1,2;

(10.15.68)
32 = 0,

^33

Cauchy's law (7.4.6) can be employed to find the stress vector s on a


cylindrical surface R = constant in the fluid. The tangential component of

10.15

SOME VISCOUS FLOW PROBLEMS

553

this stress vector is given by


s e, = SileJii = (*%)U*k^Xk)

(10.15.69)

Substituting for ,-, from (10.15.68) in (10.15.69), we arrive at the following


simple formula for the shear stress = s e e :

(10.15.70)

By use of (10.15.65) in (10.15.70), we get


2ua2b2 ct>2 - ,
** = -gf
R2

(10.15.71)

Evidently, the shear stress is tensile or compressive accordingly as the


outer cylinder rotates faster or slower than the inner cylinder. Also, the
magnitude of this stress is maximum on the inner cylinder and minimum on
the outer cylinder. The stress vanishes when both the cylinders rotate with
the same speed.
Expression (10.15.71) can also be employed to compute the couple
exerted by the fluid on the cylinders due to viscous drag. For example, the
magnitude M of the moment of the couple per unit length in the axial
direction on the inner cylinder is given by
M= [ \TReR2d0]R=a

= 4

\2 - \

(10.15.72)

This result is used for measuring by a rotation viscometer.


Two interesting limiting cases are worth mentioning. For a = 0, the
problem reduces to that of a flow inside a cylinder due to the rotation of the
cylinder. In this case, we find from (10.15.65) that the motion is simply a
rigid-body motion; the fluid rotates like a rigid body along with the
cylinder. For b -> <x> with a remaining finite and greater than 0, the problem
reduces to that of a flow of an infinite fluid outside a cylinder due to the
rotation of the cylinder. In this case, condition (10.15.67) is trivially
satisfied and the flow is irrotational.
10.15.4 UNSTEADY FLOW NEAR A MOVING
PLANE BOUNDARY

Here, we consider the unsteady motion of an incompressible viscous fluid


occupying the half-space x3 > 0 with a plane rigid plate at x3 = 0 as its
boundary. Suppose the fluid and the plate are initially at rest and that the
flow is generated by the movement of the plate in a direction parallel to its
own plane, say the xl direction; see Figure 10.21.

554

10

EQUATIONS OF FLUID MECHANICS

*3

3> 0

Xi=0

Figure 10.21. Unsteady flow near a plate.

When the plate moves in the xx direction it is natural to expect that


the fluid also moves in the same direction so that the velocity of the fluid at
any point is of the form v = vel. In the absence of body force, the NavierStokes equation (10.14.6) yields the following three component equations
1
p

Dv
Dt
P,2 = 0

(10.15.73)

/?,3 = 0

(10.15.74)

The equation of continuity (10.12.5) becomes


dv
= 0
dxx

(10.15.75)

Since the fluid extends to infinity in the x2 direction, the variation of v in


thejc2 direction may be neglected. Then, in view of (10.15.75), v varies only
with x3 and t. Consequently,
Dv
Dt

dv
,
^
dv
- + v (v V)u
=
}
dt
dt

(10.15.76)

and equation (10.15.73) reduces to


d2v
dxl

dv _ 1 dp
dt
p dxx

(10.15.77)

Equations (10.15.74) imply that p is independent of x2 and :3. Since v


does not depend on xx by (10.15.75), the lefthand side of (10.15.77) is
a function of x3 and /, while the righthand side is a function of xl and t.
Therefore, each side of (10.15.77) must be a function of t alone; that is,
dv
d2v
"--.=/)
dxl
dt
\_dp_
=
pdxx

(10.15.78)

f(t)

(10.15.79)

10.15

SOME VISCOUS FLOW PROBLEMS

555

w h e r e / ( 0 is an arbitrary function of t. It follows from (10.15.79) that


p = pxif(t)+fl(t)

(10.15.80)

where f(t) is again an arbitrary function of t. Since the fluid extends to


infinity in the xx direction, p - oo as xx - oo unless / ( / ) = 0. But the
infinite pressure is physically unrealistic. Therefore, we assume f(t) = 0 so
that (10.15.78) and (10.15.80) become
dv

d2v

aTvR
P=fi(t)

(1(U5 81)

(10.15.82)

Evidently, the velocity and pressure fields are now uncoupled; v can be
determined independent of p. If desired, p can be obtained from a
boundary condition prescribed at x3 = 0.
Equation (10.15.81) is usually called the diffusion equation. To determine v we have to solve this equation under appropriate initial and boundary
conditions. Since the fluid is assumed to be initially at rest, the initial
condition is
v =0
(10.15.83)
in x3 > 0 for t = 0.
As the plate is assumed to move in the xx direction, the no-slip boundary
condition requires
v= V(t)
(10.15.84)
for x3 = 0, t > 0, where V(t) is the speed of the plate. This is a timedependent boundary condition.
Since the flow is generated by the motion of the plate (at x3 = 0), we may
assume that the fluid particles far way from the plate remain unaffected by
the movement of the plate. This leads to the condition
i;-0

(10.15.85)

as x3 - oo for / > 0. This is often called a regularity condition (or boundary


condition at infinity).
We consider two particular cases of interest.
Case i : Oscillating Plate Suppose the plate oscillates with real constant
amplitude V in its own plane with a given frequency so that
V(t) = Vst

(10.15.86)

/ > 0, and we are interested in the velocity field strictly after the flow is fully
developed. Then the initial condition (10.15.83) is not required in our
analysis. We now seek a solution of (10.15.81) in the form
v(x3,t)

= Re[f(x3)ei"t]

(10.15.87)

556

10

EQUATIONS OF FLUID MECHANICS

where Re stands for the real part. Substituting this solution into (10.15.81)
yields the following ordinary differential equation forf(x3):
d2f
- 4 dxt

ii\
( ) / = 0
\vj

(10.15.88)

The general solution of this equation subject to the regularity condition


(10.15.85) is
f(x3) = A exp[-(l + )]3
(10.15.89)
where A is an arbitrary constant and = \/2 > 0. In view of (10.15.84),
(10.15.86) and (10.15.87), we find that A = V and hence the final form of
the solution for v(x3, 0 is
<*3,0 = Re[Kexp[-(l + i)x2 + it]} = Ve~6x's(t

- 3)
(10.15.90)

This shows that the fluid particles oscillate with the plate with the same
frequency as that of the plate and that their amplitude, Kexp(-&t 3 ),
decreases exponentially with distance from the plate. In other words, in the
fluid, shear waves that spread out from the oscillating plate propagate with
exponentially decaying amplitude so that fluid oscillations are essentially
confined to a layer adjacent to the plate. This layer is called the Stokes's
boundary layer and the problem is called the Stokes's problem. The thickness of the layer is on the order ~ = (2/) 1 / 2 . Evidently, the thickness of
the layer depends on the frequency of the plate: the layer becomes thicker
as the frequency decreases and thinner as the frequency increases. In other
words, a large amount of fluid oscillates with a slowly oscillating plate and
only a small amount of fluid oscillates with a rapidly oscillating plate. The
speed of oscillation is (/) = V2va>.
The use of the material law (10.12.8) and v = w , gives the shear stress on
a fluid element parallel to the plate as

3 = v,3

e~bx*{n(t

= V2oVe-6x>sm(a>t

- 3) - cos(t - 3)}
(10.15.91)

- 3 - | J

This shear stress is maximum on the plate and decays exponentially with
increasing distance from the plate, and
maxT31 = [ 31 ], 3 = 0 = yflaVsmlcot
This is the skin friction acting on the plate.

- j)

(10.15.92)

10.15

SOME VISCOUS FLOW PROBLEMS

557

Case ii: Impulsively Moved Plate This case arises when the plate is
started impulsively from rest with a constant velocity Vei and this velocity
is maintained for all time t > 0.
To determine v in this case, we transform governing equation (10.15.81)
to an ordinary differential equation by using the substitution
T/ = ^ = ,

= (),

(10.15.93)

t>0

where is called a similarity variable. Thus we find


dv
dx3

dv dr\
3 dx3

V ,
2\M

dv

dv

Vvx3

dt~ 3 dt ~

4(v0 3/2

d2v
dx3

V M/ ,
4vt

so that (10.15.81) reduces to


F"(ri) + 2//F'(>7) = 0
(10.15.94)
The general solution of this first order ordinary linear differential equation
for F\n) is
F'(/7) = Ae-*2
(10.15.95)
This gives F(//) as
F(/7) = ,4

e~v + B

(10.15.96)

where A and B are arbitrary constants.


From (10.15.84), (10.15.85) and (10.15.93), we note that F(^) satisfies the
following conditions:
F(//)=l

for

/7 = 0,

F(A7)-Oas/7- .

(10.15.97a,b)

Using these conditions in (10.15.96) and noting


Jo

exp(-/7V/7 = ^

(10.15.98)

we find B = 1 and A = -2/Vn. Thus the final form of v is given by


2

'

2
v = V 1 --7= I exp(-n )dn
(10.15.99)
" Jo
It can be verified that v decreases continuously from F to 0 as x3 increases
from 0 to infinity. As in the Stokes's problem, solution (10.15.99) exhibits
a boundary layer nature; that is, the effect of the motion of the plate on the

558

10

EQUATIONS OF FLUID MECHANICS

fluid flow decreases rapidly away from the plate. Here, the thickness of the
boundary layer is on the order = Vv7, which grows with increasing time;
the rate of growth is on the order d/dt VvT/, which decreases with
increasing time. This problem is known as the Rayleigh problem and the
associated boundary layer is called the Rayleigh layer.
Using the fact that v = vel, where v is given by (10.15.99), we find that
the vorticity vector is directed along the x2 axis with magnitude
\v3\ = -7^=exp( -^-)

luvt

\ 4vt

(10.15.100)

Evidently, the vorticity is 0 at / = 0 except at x3 = 0, and it decays


exponentially to 0 with increasing x 3 . This implies that the vorticity is
generated at the plate and diffuses outward within the Rayleigh layer.
By using Stokes's law (10.12.8), we find the shear stress on a fluid
element parallel to the plate as given by
31 = //.a = - ^ L e x p f - ^ )

(10.15.101)

Evidently, the shear stress acts opposite to the direction of motion and its
magnitude is maximum on the plate with
maxT 31 = [ 31 ], 3 = 0 = jL

(10.15.102)

This is the skin friction on the plate. Its magnitude varies inversely as vT.
10.15.5 SLOW AND STEADY FLOW PAST
A RIGID SPHERE
As already indicated, for sufficiently small Reynolds numbers, the viscous
forces exceed the inertia forces and the nonlinear terms in the
Navier-Stokes equation may be neglected. In this case the governing
equations for an incompressible fluid flow are given by (10.12.5) and
(10.14.28). As an example, we consider here the steady flow past a rigid,
fixed sphere in the absence of body force.
We assume that the fluid particles far away from the sphere are
undisturbed and move with a uniform velocity Ko in the xx direction. Then,
the velocity field satisfies the following conditions:
Vi = Ko,

v 2 = v3 = 0

as r - oo

(10.15.103)

where r is the distance measured from the center of the sphere, which is
taken as the origin (see Figure 10.22).

10.15

SOME VISCOUS FLOW PROBLEMS

559

Since the sphere is rigid and stationary, the condition to be satisfied on


the surface of the sphere is the no-slip boundary condition
v = 0 or Vi = 0

for r = a,

(10.15.104)

where a is the radius of the sphere.


Since the flow is steady and there is no body force, equations (10.14.29)
and (10.14.30), which follow from equations (10.12.5) and (10.14.28), serve
as the equations for determining p and t;,.
Noting that xx/r* (r 5* 0) is a harmonic function, we take the pressure /?,
which is a harmonic function by (10.14.30), in the form

p = -4*i

(10.15.105)

where A is a constant. Substituting this value of p into (10.14.29) we obtain


the following governing equations for vt :

=y50xl-r2)

V\

,
3
2 = xxx2

(10.15.106)

1A
2
3 = XiX3
It can be verified that the solutions for vt are
vl = V00- ^-s(Ar2
2//r

- C) + -r^iBr2
6//

- C)

(10.15.107a)

v2 = -^(2

- D')

(10.15.107b)

v3 = -^{2

- D')

(10.15.107c)

where A, B, C, D' are constants. In view of conditions (10.15.103) and


(10.15.104) we obtain
A=iaV00,

B = -3A,

C = D' = Aa2

560

10

EQUATIONS OF FLUID MECHANICS

Substituting these into (10.15.105) and (10.15.107), we obtain the following


solution for the pressure and velocity fields:
300
2r5
vl = Vo

(10.15.108)

3aVa 2
aV^ z
2
5 -(r - a )xf - ^{3r
Ar
Ar

3aV^[
(r Ar5
3aK00 (Ji
v3 = - Ar" (r

- az)

(10.15.109a)

a)xYx2

(10.15.109b)

- a )3

(10.15.109c)

This solution was first obtained by Stokes in 1851. The velocity distribution
shows that the stream lines of the flow are symmetric about the equatorial
plane normal to the direction of the flow, as depicted in Figure 10.22.
Since the fluid essentially moves in the xl direction, the resultant force on
the surface of the sphere acts in the xx direction. This force, called the drag
on the sphere, is
D=

sxdS

(10.15.110)

where sx is, as usual, the xx component of the stress vector s, and 5 is the
surface of the sphere.

Figure 10.22. Slow and steady flow past a rigid sphere.

10.15

SOME VISCOUS FLOW PROBLEMS

561

By using Cauchy's law (7.4.9) and noting that flf = ,/ on S, expression


(10.15.110) can be rewritten as
D =-\

( * + Ti2*2 + rl3x3)dS

(10.15.111)

Computing xVl by using the material law (10.12.8) and results (10.15.108)
and (10.15.109), substituting the resulting expressions onto the righthand
side of (10.15.111) and subsequently evaluating the surface integral, we
arrive at the following famous formula, known as Stokes's formula:
D = ^

(10.15.112)

Computations leading to (10.15.112) from (10.15.111) reveal that one-third


of the righthand side of (10.15.112) is due to pressure and two-thirds is due
to skin friction.
A nondimensional drag coefficient for the sphere, CD, is defined by the
relation
D = \pVlCD(na2)
(10.15.113)
Using (10.15.112), we get
CD = Y6

(10.15.114)

where Re = (2aV00/v)(< 1) is the Reynolds number based on the diameter


of the sphere.
It has to be pointed out that Stokes's solution (10.15.109) is not valid at
long distances from the sphere. In order to check this point, we estimate the
orders of magnitude of the inertia term p(\ V)v and the viscous term //V2v.
According to Stokes's solution, the inertia term is on the order p(aV2/r2),
whereas the viscous term is on the order ^/r3. Hence Stokes's solution
holds only when
aVl

^<
r

3
r

that is, r < v/V^.


The difficulty associated with Stokes's solution was first resolved by
Oseen in 1910. Using the hypothesis that vx is approximately equal to V^
at long distances from the sphere, he assumed (v V)v Vjy/) in the
Navier-Stokes equation and replaced equation (10.14.29) by the following
equation to study the problem:
pVoo^- = - V p + //V2v

(10.15.115)

562

10 EQUATIONS OF FLUID MECHANICS

We shall not attempt to find the solution of the problem on the basis of this
equation, which is known as the Oseen equation. We just mention that the
solution leads to the expression for the drag coefficient

1 + ^ )
Re\

(10.15.116)

which agrees with experimental findings up to Re = 2aVx/v = 5. Comparison of (10.15.114) and (10.15.116) shows that Stokes's formula (10.15.114)
is a first approximation of Oseen's formula (10.15.116).

10.16
EXERCISES
1. If a fluid of uniform density is at rest under a constant gravitational field, show
that the pressure varies linearly with depth.
2. For water, an approximate relationship between the pressure and density is
p = 1 + dp, where a is a constant. If pa is the density at the surface of a lake, show
that the density of water at depth h in the lake is given by p = paeagh. Assume that
p varies only with depth and gravity is the only body force.
3 . At a height h above the earth's surface the density of air is p and pressure is p.
At the surface, the corresponding values are p0 and p0. Assuming that the gravitational field is constant and p = p0e~az, where a is a constant and z is the vertical
distance above the earth's surface, show that

P=Po-(\-e-z)
a

4. Assuming that the atmosphere around the earth is a perfect gas at rest in which
the pressure-density relation at an altitude z is given by (p/p0) = (p/p0)y, where p0
and p0 are the pressure and density at a reference elevation z0, determine expressions
for , p and p in terms of T0, pQ, p0, z and z0. Here T and 7^ are temperatures at
z and ZQ , respectively.
5. A rigid sphere of radius a is fixed in an incompressible fluid at rest. The fluid
extends to infinity in all directions and the pressure-density relation is/? = ap, where
a is a constant. If every fluid particle is attracted towards the center O of the sphere
by a force of magnitude /r2, where r is the distance from O, show that the pressure
exerted on the surface of the sphere is proportional to exp(/(xa).
6. Show that in an incompressible nonviscous fluid flow, the rate of work done by
external forces is equal to the rate of change of kinetic energy and that the rate of
change of internal energy is equal to the rate of heat input.

10.16 EXERCISES

563

7. If e = + (p/p) is the specific enthalpy, show that de/dT = cp for a perfect gas.
8. Show that the energy equation for a compressible non viscous fluid flow can be
put in the following forms:
De
p = -p div v - div q + /?A

(i)

oi))
(iii)
(iv)

D /l\

De

+ ^ <, - =

De a
div(pv + q) - p(\ b) - ph = p = jt(p)

+ \\(p)

De dp
/> = - divq + pA + />(b v)
where e = + (/?//>) and = e + (t>2/2).

9. Write down Euler's equation of motion in the suffix notation.


10. Show that the velocity field given by
x\ - x\

22

where a is a nonzero constant and R2 = x\ + x\ ^ 0, obeys Euler's equation of


motion for an incompressible fluid. Determine the pressure distribution associated
with this velocity field.
11 - The velocity field in an incompressible nonviscous fluid flow is given by

v1 = vQ COS cos ,
2a
2a
. nx . nz
v2 = y0sin sin ,
2a 2a

v3 = 0

where v0 and a are nonzero constants. Show that the pressure is given by
1
nz
P T Vo ) cos

nx
cos + constant

a^

12. For a steady irrotational flow of a nonviscous fluid of constant density under
gravity, show that
P=Po~ PSZ- jpv2
where p0 is a constant.
13. For an irrotational steady flow of a nonviscous fluid under zero body force,
show that dp = -pvdv.

564

10

EQUATIONS OF FLUID MECHANICS

14. A liquid column of uniform density moves in a vertical direction (against


gravity) with a constant acceleration a. Find the pressure at a point whose depth
from the upper surface of the liquid is h.
15. For the liquid column considered in Example 10.6.2, show that the rise of the
liquid above the vertex of the upper surface along the wall of the vessel is a>2a2/2g,
where a is the radius of the vessel.
16. For a two-dimensional flow of an incompressible nonviscous fluid under zero
body force, show that the vorticity is constant.
17. In the irrotational, two-dimensional flow of an incompressible nonviscous
fluid, show that the velocity is constant in magnitude if and only if it is constant in
direction as well.
18. The velocity field for a two-dimensional flow of an incompressible nonviscous
fluid under zero body force is given by
vx =

29

v2 = -

l9

v3 = 0

where = (,2> t). By using Euler's equation of motion show that satisfies
the equation
a
,
,2
(VV)
v
V
.
i
vV,2
dr
19. Show that in a compressible nonviscous fluid flow, the rate of change of
circulation round a circuit c is given by

it=-iH)xvp-cunb}'nds

where 5 is a surface for which c is the rim.

20. Show that for an elastic fluid moving under conservative body force, Cauchy's
vorticity equation (6.6.8) can be rewritten as
- = ^-Vx
P
A)

21.

For an elastic fluid moving under conservative force, show that


D

. wdV =

(w n)vdS

where V is a material volume and S is its boundary surface.


22.

Show that the equation (10.7.8) can be rewritten as


+ w x v = -V(P + + \v2)

Deduce that V2(P + + \2) = w2 + v V2v - v V(div v) - (d/dt)(div v).

10.16 EXERCISES

565

23. Show that in the absence of body force the only possible steady flow of an
elastic fluid for which v = vxfa, p = p(xx), P = () is the one for which ,
and P are all constants, independent of xx.
24. For a nonviscous perfect gas moving in the xx direction, show that
d2p
d2
^ =- {

7
+

RT)]

25. For a steady, irrotational flow of an adiabatic isentropic gas in the absence of
body force, show that the speed of sound is given by
2

y -

cs = c0

where c0 is the speed at a reference condition.


26. Show that for a perfect gas cs = yfyRT.
27. Write equation (10.7.24) in the unabridged form.
28. Show that the Bernoulli's equation (10.8.7) reduces to
(i) p\og(p/p) + (\/2)v2 + = constant, for a perfect gas.
(ii) /( - \)p + (\/2)2 + = constant, for an isentropic perfect gas.
29. A gas, in which p and p are related by an adiabatic relation p = , flows in
a steady state in a horizontal conduit. Using Bernoulli's equation show that the
maximum speed is given by

where c0 is the speed of sound at a reference condition.


30. A nonviscous incompressible fluid flows steadily around the outside of a fixed
vertical cylinder of radius a so that the fluid particles move with speed a/R in
horizontal circles concentric with the boundary of the cylinder, where R is the radial
distance from the axis of the cylinder. Show that the flow is irrotational. If the
surface of the fluid is open to the atmosphere, show, by using the Bernoulli's
equation, that
2gz = 1 - (a2/R2)
for appropriate choice of the coordinate axes.
3 1 . Find the stress matrix associated with the following velocity field occurring in
a viscous fluid flow:
(i)

= 0 ,

v2 = 0,

v3 = x2

(ii)

vx = 0,

v2 = x\ - *3,

(iii)

vY = !;,(*,, x2),

v3 = -2x2x3

v2 = v2(xx, x2),

v3 s 0

566

10

EQUATIONS OF FLUID MECHANICS

32. For a certain motion of an incompressible viscous fluid, the velocity is of the
form v = v(xx, t)ex. Verify that the motion is necessarily irrotational and
nonisochoric. Find the stress components developed in the fluid. Deduce that the
stress vector on planes parallel to and perpendicular to the direction of flow is a pure
pressure.
3 3 . For a viscous fluid, show that
34. Show that for a compressible viscous fluid,

35. Show that


36. The stress matrix at a given point of a viscous fluid with zero bulk viscosity is

Find the viscous stress matrix [{)].


37. For a compressible viscous fluid with zero bulk viscosity contained in a volume
V, show that the total surface force on the boundary S of V is given by
is

sriS=

]v

(divT(v) - Vp)dV =

]v

[2 div D (d) - Vp] dV =

]v

[pW2\ - Vp] dV

38. Show that the energy equation (10.12.3) can be expressed in the form
De p Dp
->
p = - - p + kV2T + ph +
Dt p Dt
39. For an irrotational flow of a viscous fluid adjacent to a plane rigid wall,
determine how the tangential velocity varies with the normal direction from the wall.
40. Show that the magnitude of normal stress on a stationary boundary with which
a compressible viscous fluid is in contact is p + (4/3)v(Dp/Dt).
4 1 . For an irrotational flow of an incompressible viscous fluid in a region bounded
by a rigid boundary, show that the velocity is uniquely determined by prescribing the
normal component of velocity on the boundary.
42. Write down the Navier-Stokes equation (10.14.3) in the suffix notation.
4 3 . For an irrotational flow of an incompressible fluid, show that the NavierStokes equation reduces to Euler's equation.

10.16 EXERCISES

567

4 4 . For an incompressible viscous fluid moving under conservative body force,


show that
DIC

Dt

vV 7

4 5 . For an incompressible viscous fluid moving under conservative body force,


show that the vorticity obeys the equation
= curl(v x w) + vV2w
dt
4 6 . Generalize expression (10.14.20) to the nonsteady case, other conditions being
the same.
4 7 . Assuming that / is a constant, show that, for an irrotational motion of a
compressible fluid with zero bulk viscosity under a conservative body force, the
Navier-Stokes equation has an integral in the form
dp
1 ,
+ 2 + p
L

#
4
+
at
3

.
-2=/()

4 8 . In an incompressible fluid moving under zero body force, the velocity


components are of the form
vx = -\axx

-f(R)x2

2 = -\<*x2 + f(R)x2
v3 = ax3
2

l/2

where R = (x + x2)
and a is a positive constant. Show that the pressure
distribution is of the form
P=Po~

\pa\R2

+ 4*32) + p

R{f(R)j2 dR

where p0 is a constant.
4 9 . For the plane Poiseullie flow, show that the vorticity is a harmonic function.
5 0 . An incompressible viscous fluid bounded between two inclined parallel, rigid
plates flows down under gravity. If the upper plate is moving in the direction of
flow with speed v0(t) and the lower plate is held stationary, obtain the velocity
distribution in the fluid.
5 1 . An incompressible viscous fluid flows steadily down an inclined plane under
gravity. The fluid layer is of uniform thickness and the upper layer is exposed to
atmospheric pressure. Obtain the velocity distribution in the fluid.
5 2 . For the Hagen-Poiseullie flow in a circular conduit show that the vorticity is
perpendicular to the conduit and that its magnitude varies directly with the radial
distance from the axis.

568

10

EQUATIONS OF FLUID MECHANICS

5 3 . For the steady flow of an incompressible viscous fluid through a straight


triangular tube bounded by the planes x = a, y = (1/V3)x under a constant
pressure gradient and zero body force, show that the velocity distribution is given by

^-(x-a)(3y2-x2)

v=

Deduce that the mass flow rate in a cross section is


M

Ga4
60V3v

5 4 . An incompressible viscous fluid moves steadily in a horizontal elliptic pipe of


length L placed along the x axis due to a pressure difference P maintained across the
ends. If y1 + 4z 2 = 4 is the equation of the boundary of a cross section, show that
the total volume of fluid delivered through the pipe per unit of time is 2/5.
5 5 . An incompressible viscous fluid flows steadily through a region bounded by a
solid cylinder of radius a and a coaxial tube of radius b, b > a. The cylinder and the
tube are stationary, there is no body force and the motion is due to a constant
pressure gradient parallel to the axis. Show that the velocity distribution is given by
v = ^- [(a2 - R2) + (b2 4

a2)i\og(R/a)/\og(b/a)}]

where R is, as usual, the radial distance from the axis. Deduce that the shear stress
and rate of mass flow are, respectively,

<-!

R\og(b/a)

2R\,
J

M = 8v

(b2 - a2)2!

\og(b/a) J

5 6 . An incompressible viscous fluid flows steadily through a region bounded by a


solid cylinder of radius a and a coaxial tube of radius b, b > a. The motion is due
to the movement of the cylinder with speed v0 along its axis. The tube is stationary;
there is no body force; and there is no pressure gradient. Show that the velocity
distribution is given by v = v0 \og(b/R )/\og(b/a) and that the shear stress along the
axis is = -v0/[R \og(b/a)), where R is, as usual, the radial distance from the axis.
5 7 . Carry out the computations involved in obtaining (10.15.112) from (10.15.110).

ANSWERS AND HINTS TO


SELECTED EXERCISES

Section 1.8
1.

(i) meaningful; an + a22 + #33


(iii) meaningful; anbl + ff2i^2 + ^31^3 #12^1 + #22^2 + #32 3 >
anb1

+ <723^2 + #3363

(iv) not meaningful


(vi) not meaningful
(vii) meaningful; anbu
anbl3

+ al2b2i + al3b3l + a2lbl2 + #22^22 + ^23^32 +

+ #32^23 + #33^33

(viii) not meaningful


2. aybj and arsbs have the same meaning.
ciijbjbj, apqbpbq and asrbsbr have the same meaning.
3. (ii) not meaningful
(iii) not meaningful
(iv) meaningful
(vi) meaningful
(vii) not meaningful
(viii) not meaningful
4. (ii), (iii), (iv) and (vi).
5. au = \(bu

+ bji)
569

570

CONTINUUM MECHANICS

6. Note that a^Xj = ^ = -,-,,,.


8. Use the fact that a^C = a^Cy = ayCy.
10. Note that a e, = at.
13. (i) Show that (a x b) x c - a x (b x c) = b x (c x a).
(ii) a and c are collinear, or b is orthogonal to both a and c.
14. au = 2, aua = 0, auau = 17
15.

-3

[*</] =

2 -3

5 -8_

3 -2

0 -5

2 0 - 1 2 -16"
T

16.

\au\[au\

-12

100

32

-16

32

228

32 - 5 6 "

104
r

32

K] [^l =

164 - 8 0

-56

-80

36
2

K,] =

-60

40 - 2 4 "
64 - 2 4

-108
" l O O "
1 9 . (i)

. 0
2 2 . (i) 0

(ii)

0_

4
1
4

-i

80

96

ol

i 0
0 -lj

(iii) 3 - auau

(ii) dibjCj

23.

(0 1]1] = [I]
(ii) 1[] = [I]
(iii) [] = abkk[I] + [bij\
2 4 . (i) au = -pu
(ii) aw + b,f = 0
25.

i = 623 -

^32

2 = 31 -

2 6 . Note that eiJkbj =

613 >

o3=.bn - b2l

-skjibj

2 7 . (iii) Note that epgrerqs = -<75qr and use (1.7.24)


2 8 . / * (7 x e *) = (e, x e,) e* = (m0 em)
3 0 . Use Zuk<tibiCk = [a,b ,c]
3 1 pqr br

2 pqr ejkr

l( pq

* 'ijm Omk

. = ijk

jk

= \(pjqk a

' ek

qp)

^^^

pq

32. Use eijkekrsrijVrns = (irjs - l5jr)njVrns = (v^j - n^rij = , -

n^rij

ANSWERS AND HINTS TO SELECTED EXERCISES

33.

(i)
(ii)
(iii)
(iv)

Xi = aeijkajbk + at
eijk(ai - b^Cjdk = 0
Xi = bi + aeijkekrs(ar - br)CjCt
eim
einaamanbbn = ^i^i^j^j
a,aib,b,
imn^ipq^m^p^n^q

571

^i^i^j^j

Section 2.15

]_

1. (a) [a</] =

' 2

(b) [eg =

0
\_
2

V3
2

0 -1
4. x[ + (1 - V3/2)*2 - (1/2 + y/3)x'3 = 1

5. Use (1.7.21), (2.2.17) and orthogonal property of [au].


7. The entity is a vector.
1 0 . a[ = 0,

a'2 =

1 -V3
2V2 '

1 + V3
2V2

, = - -

K/] =

1 5 . <% = <^ = oLimajm

V3
-
2

V3

j_

"2

"2

__1
~2

V3
2

1
2

V3
~2
1
~2

^ i m **/n "1

2 1 . Use (2.4.40).
2 4 . If a b = I, then for any c, we get c = Ic = (a b)c = (b c)a, which is
not necessarily true.
2 8 . Show (I - 2a (x) a)(I - 2a <g) a ) r = I; not a proper orthogonal tensor.

m=

29.

0 0

1 0

2e2 - e3

3 1 . Use a quotient law.


3 8 . [(b I)c], = eirs brjsCj = eirsbrcs
4 0 . Use aubu = \au{bu
43.

b).

(i) Use a (skw A)a = (skw A) r a a = -(skw A)a a.


(ii) Use a (Aa) = a (sym A + skw A)a = a (sym A)a.

4 4 . Use (2.10.6) to get = 3ej + 5e2 + 2e 3 .


4 5 . Use (2.10.7).

572
49.

CONTINUUM MECHANICS
(i) Use (2.10.4) to get
<*ikbkj = eikmumekjnvn
(ii) aikbki =

a.
ij^mnt^m

= (inmj

"n = UjVi -

Suumvm

-2umvm

5 3 . (e* ei)(e m e^) r = (e* ei)(ei em) = (e* e'm)(ek

em)

= km(ak em) = e* ek = I
55.

(i) Use (2.13.25) and (2.11.23) to get


A 3 - / A A 2 + / / A A = (det A)I = (A*) r A
(ii) Use (2.11.24).

56.

(i) Use (1.7.24) and (2.11.1).


(ii) Use (1.7.23) and (2.11.4).
(iii) Use (1.7.26) and (2.11.11).

5 7 . For any A,
det(BAB _1 - AI) = det{B(A - AI)B _1 )
= (detB)(det(A - AIJKdetB"1) = det(A

AI)

5 8 . Use (2.11.23) and (2.11.25).


6 1 . Use (2.11.1), (2.11.3) and (2.11.9).
6 3 . Use (2.11.24) along with (2.11.22) and results of Exercise 56.
6 4 . Use (2.12.2) and results of Exercise 56.
6 5 . (ii) 7A = - 2 ,

// A = - l ,

66. (ii) 1,2,4;


(iv) 2 , 1 , - 3 ;
6 8 . Use (a

A = 0,

/ / / A = 0;

- 1 V2

e3,
(1/V2)( ei + e2),
(1/V2)( ei - e2)
(1/5)(2 + e3),
e2,
(1/V5)(e1 - 2e3)

b)a = (a b)a
0 -1

7 2 . (ii)

0 -2

7 3 . (i) [Q] =

0 ,

[U] =

0 -1_
1

(i) [Q] =

j_

J_-

V2

_ 0

"V2
1

"2

0
0
-1 .

o"
0
0-1

, m=

" 2V"2
,

[U] =

[~o
1

0]

V2

lj

>

0 -1

ANSWERS AND HINTS TO SELECTED EXERCISES

IV] =

573

o
o
1

Section 3.8
2. Start with u u = u2 and differentiate both sides with respect to /.
3. Using u u = 1, show u (du/dt) = 0, and use the definition of vector product
to compute |u x (du/dt)\
4x2
8 . (Uij + uJti) = '

2xx

2xl -2x3
3*3

-*2

3x3
-Xi

6xx

1 1 . 6/14 and -6/14.


14.

Xi

/,=/'(),=/'()-

17. (i) e* uk = ek {ultkex + u2ke2 + w3,*e3) = divu.


+ u2tke2 + w3,*e3);
(ii) e* x uk = ekx(uhkei
e

18. (i)

l,kel

= ~Wl,2e3 +

l,3 e 2>

21.
22.
24.
26.
27.
30.
32.
35.
40.

[curl(u x v)], = eijkekmn(umvn)j


= (imjn - SinJm)(umJvn + umvnJ)
= UijVj

20.

etC

2|| = (rn)M =

+ UiVjj

(^"1-)

UjjVi

UjVu

= n(n + l)/""2

= A/r3, where v4 is a constant.


0 = A/r2 + B/r, where A and Z? are constants.
Note that V2(</>^) = div () and use (3.4.4) and (3.4.14).
Simplify curl{(curl u) x x) using (3.4.25).
curl(r"x) = rn curl x + nrn~2\ x x = 0
div(r"x) = 3rn + nrn~2\ x = (3 + ri)rn
Use (3.4.25) and (3.4.26).
curl(0u) = </>u ( curl u + V0 x u) = 0.
[(v Vu)w], = eirs vrujt s vv,.
(ii) [div Vur]( = ukJk = (uktku)d.
(iv) div(Au) = (djiUj = Uidjij + anui%j.
(ix) div[(u <g) v)w] = (UiVjWj)ti = uitiVjWj + f/ji^w,),,.

574

CONTINUUM MECHANICS

4 1 . If/ 5* 0 at a point P of R, then for some infinitesimal region Av containing P,


we have

fdv=f(Av)*0
This is a contradiction.
4 2 . Zero.
4 4 . (i) Employ (3.6.1) to u = <t>(Vy/).
(ii) Employ (3.6.1) to u = (?) + ().
4 5 . Use (3.6.3).
4 9 . Take A = I and note that 2 = - c u r l u by Example 3.5.7.
5 3 . (i) Employ divergence theorem to u x curl u and use (3.4.24).
(iii) Employ divergence theorem to (u a)V(u a) and use (3.4.14).
5 5 . Use (3.7.1) and (3.7.3) to get
[{u v + v ujn - (u v)n] dS
[(div v)u + (div u)v + (v V)u + (u V)v - V(u v)] dV
and then use (3.4.26).

Section 5 . 1 1
a
[F] = [U] = [V] =

1- (0

[U] =

0
y

l/

L 0

1/y

Principal stretches:

0)0
0

[F" ] =

[F]

l/ct
1

(Hi)

Q = i

isochoric if = 1

,,.

-a

[F" 1 ]

V2a

yfl

1
2a
1

1
2a
1

1/y.

a + /ff -a +
[V] =

vf

-a +
0

a + )ff

V2y

ANSWERS AND HINTS TO SELECTED EXERCISES

[Q]

= V2

-1

575

isochoric if = 1/2

0 V2

Principal stretches: V2a, V2/?, y.


2.

(i) The deformation is a uniform extension or contraction in the 1 direction


according as a > 1 or a < 1. measures the extension or contraction in the
2, 3 plane according as > 1 or < 1.
(iv) The deformation displaces particles in the plane x% = constant through a
distance proportional to xf and to the distance {()2 + (xf) 2 ) 1/2 from the
3 axis; the direction of displacement is at right angles to the vector
*iei +* 2 e 2 .

6. The hyperboloid of revolution: x\ + x% = a2(\ + x32).

9.

(i)

[C] =

[G]=i

(iv)

[C] =

[G]=i
1
1 0 . (ii)

[IT ] =

1 + a2

a2

[B] =

tan a.

tan a

sec a

0
0

sec a
[B] =

tana

tan a

tan a

0
0

a
1 + a2

0 1
0

1
a

[A]
(1 + ot)2

sec a

tan a

ajff(2 + a)

1 4 , e = (1 + a 2 /2) 1 / 2 - 1
1 6 . Arcs initially lying along the 3 axis.
2 1 . (i) = V2,

(ii) = cos'^l/V)

3 1 . (ii) Material form: 1 = (x? + x3)(e' - 1), u2 = x 2 V - *"'). 3 = 0.


Spatial form: ux = (JCX + x 3 )(l - e~'), u2 = x^e* - e~'), w3 = 0.

576

CONTINUUM MECHANICS

2a2(*i -x3)

- (2a(*3 - xx) - x2]

2a 2 (x 2 + x3)

^ f 2a(x2 + x3) - *, )

39. (iii) []

2 f2a(*2

- f 2a(x3 - Xi) - x2)

* s ) " Xl '

-\2(3-)

+ 2}

-\2a(x2+x3)

+ x x)

- [2a(x1 - x3) - x2\

- - {2a(x2 + x3) + xx)

[0] =

= [{J^i + 2a(*2 + x3)]ex - [x2 - 2 ^ ! - x3))e2]


4 0 . (i) ^-(58a - 64),
4 3 . (ii) ux = Xi + x2,

(ii) ^ ( H 9 a + 40)
u2 = 2xlx2 + x2,

u3 = x2

4 4 . (i) a = 2

Section 6 . 7
1.

(i) Material description:


vx = 2a2txx,

v2 = v3 = 0;

Dv1
=
~Dt

Dv2
Dv3
~Dt~~Dt~

2a2xl

Spatial description:
=

2a2xxt

5-,
2 2

v2

= u33
2
3

1+ at
(iv) Material description:

- 0;

Dvx
Dt

2a\
1 + a2t2 '

v! = - ( x ? sin a/ - x2 cos a/), ^2 = ~(

cos

Dv2
Dt

Dv3
= 0
Dt

<*f + xi sin 0 ^3 = 0

= - ( " cos a/ + x2 sin a/)

Zty

- = a2(x? sin at + 2 cos at),


,
1
2
"

= 0
Dt

Spatial description
vx = ax2,

v2 = -ax1,

3 . ( - f o e - 0 " + 3k2e~2at)e2

Dvx
v3 = 0; - ^ = -a2xx,
Dt

Dv2
-j = -a2x2,

Dv,
-j = 0

ANSWERS AND HINTS TO SELECTED EXERCISES

577

6. (D\/Dt) = -a\xxex + x2e2)


10. Use the relation (6.3.21).
1 1 . J = .
14.

(*!

+ 2x2

3JC3)T' -

(*!

2JC2 -

3jt3)?"2'

16. '
17. For exercise 3:
(i) dn = -k(3x + 22)', d22 = k(xf + 3x22)e-ai;
(ii) H>12 = -w 2 1 = - 2 ^ X 2 ^ " " ' ; other wu = 0;
w = 2kx1x2e~ate3.
(iii) / D = -2A:(JC2 - x)e-at
IID = -k\x\

///D=0.

+ 3JC22)(3JC2 + * 2 V 2 a '

(iv) </<?> = - - ( 7 x 2 + 5xi)e-at,

d$ = ^(5x 2 + 7xi)e'at;

19. Principal stretchings: (\/2)Rdf/dR, (-\/2)R(df/dR),


Angular velocity: (l/2)[2/+ Rdf/dR]e3;
Principal directions of stretching:
cosi + - J e , + sin( + - ) e 2 ,
20.
24.
25.

43.

other tf0 = 0.

other ^

0;

-sinl + T W I + cost + - ) e 2 ,

e3,

where 0 = tan"1(x2/Xi).
Follow the method of Example 5.6.2.
Integrate the equation D = (l/2)(Vv + Vvr) for the given [D].
Principal stretchings: (l/2)a, (-1/2), 0;
1
1
axb
V + b),
V - b),
Principal directions of stretching: -7=(a
-r=(a
V2 " ""
laxbl*
V2 ""
(i) Both stream lines and path lines: xl = a cos at + b sin at, x2 = -a sin at +
b cos at, x3 = t + c, where a, b, c are constants
(iii) Stream lines: xl = aes/(l+t\ x2 = be5, x3 = c. Path lines: xx = x\\ + /),
x2 = xl, x3 = x3.
(iv) Both stream lines and path lines: Jt, = (1 + t)xf.

Section 7.10
1.
2.
3.
4.
5.
7.
9.

= 0.

s = - 2e2.
s = (1/3X56! + 6e2 + 5e3);
|s| = V86/3;
0 = cos"\9/86).
s = (2/V6)(ej + e3);
= 1,
= 1/V3.
= fc = c = -1/2.
Plane perpendicular to the vector V2 e! + e 2 .
a = 3/4; plane perpendicular to the vector ex - 4e2 + e 3 .
s = -(a/R)(2cos20ex + sin20e2);
= -(2//?) cos 0;
= 0.

578

CONTINUUM MECHANICS

10. | 2

3 2 -5
12. Symmetry of T.
13. Along the line of intersection of the elements.
17. s = T[(b n)a + (a n)b);
T = b + b a).
19. Principal stresses: 2, 1, - 3 ;
Corresponding principal vectors: (1/V5)(2e! + e3);
e2, (1/V5)(e! - 2e3).
21. /T = 0,
// T = - 9 ,
= 8.
22. Maximum shear stress of magnitude 3 occurs on a plane element perpendicular
to the vector 2ex + (1 - V3 )e2 + (1 + V3 )e3.
24. s = ( l / V ) ^ + e2 - 2e3);
= 1,
= V6.
27. Octahedral plane.
30. 0,
T.
-1
- 3 V2~
Principal deviator stresses: 0, 4;
3
- 1 -V2
34. [ =
V2 -V2 2 _
Stress deviator invariants: //TGO = -16, ///T(d) = 0.
35. Note that on the boundary surface, nl = (\/a)xlf n2 = (l/a)x2, n3 = 0.
36. Total force on each of the faces xx = a, x2 = a is (8/3)3. Faces x3 = b
are stress-free.
38. cos + 12 sin = -p cos 0;
21 cos + 22 sin = -p sin 0, where 0 is
the polar angle of the point.

Section 8.9
1. (i) and (ii): satisfied.
2. p = /?0exp(-/2).
3. p = A/x3, where is a constant.
4. v = i;e3, where v is independent of x3.
6. * = 1.
9. Find J and use equation (8.2.8).
15. pb = (xi - 1)!.
18. Note that divT = (a VT)a.
19. <t>(r) = (A/r2) + B, where and 5 are constants.
20. Zero body force.
31. Note that T D = -p div v and use the equation of continuity.
32. Use equation (8.6.32).
33. Convert equations (8.6.27) and (8.6.28) to material form.

ANSWERS AND HINTS TO SELECTED EXERCISES

579

Section 9.14
4 . Use relations (2.11.1), (2.11.4), (2.11.10) and (9.2.6).
W = (l/4)//a 2 (jti 2 + * | ) .

6.

8.

(i) = 12 = 22 = 33 = 0,
(ii) = 22 = 33 = 0,
(iii) = 22 = 33 = 0,

13 = k0A%

12 = 2kx3,
12 = 2kx3,

^11 ^22 ^33

33 =

3(1 - )

3(1 -

a [

-P*3 +

23 = k(xx - 22)

-Ekxx

(- + PgxJ

3(1 - )
l-pxi + pgx3Xih
E
u, =

31 = 2kx2

13 = ( 2 + 2 ^ ) ,

(iv) = 22 = 12 = 23 = 31 = 0,
10

23 = k0a
23 = 2kxl9

^ *

i=

1,2;

1 3 . Use identities (3.5.34) and (3.6.19) to rewrite (9.8.26) as


s = A(div u)n + 2( V)u - //(Vu - Vu r )n
Write this in the suffix notation and express A and in terms of E and v.
1 5 . Use relations (9.8.20) to (9.8.22).
21.

= 2fiA
22

xj + x32

x2

L r (r + x3)

r (r + JC3) J

x\ + x32
= 2 3
r (r + x3) ~ ? ( 7 + ^3)2J

!f_l

12 = - 2

* i * 2 ( * 3 + 2r)
r\r + JC3)2

13 = -2~,

23 = -2/^4 - | ,

On r = constant, s = {(-2/^4)/r 2 (r + 3))[^


A
ux = X2i

33.

A
u2 = -xli

\0

13 ~ 2 3 -

33 -

-2-\

+ x 2 e 2 + (r + Ar3)e3].

u3 = 0

-\>

33 =

\0

\2,

12 = (*-),

4 1 . f(r) = >1 cos(/cx)r + 2? s i n ^ / c , ) / , where .4 and B are constants.


4 8 . Use the Green-Lam solution.
5 5 . Use the fact that xkk is a harmonie function.

580

CONTINUUM MECHANICS

62. (i) Use relations (9.8.4), (3.6.1) and (3.7.4).


(ii) Use relation (9.8.26) and simplify.
64- To obtain the required identity, start with the equilibrium counterpart of
equation (9.8.8), take dot product with u, integrate over K, simplify using
identities (3.4.14) and (3.4.24) and employ the divergence theorem (3.6.1).

Section 1 0 . 1 6
2. Solve equation (10.6.3) for p, with given p and b = ge3. Express the solution in
terms of p\ note that p = pa for z = 0.
6. Consider equations (10.4.34) and (10.4.9).
12. Use the Bernoulli's equation (10.8.8).
14. p = Po + p(g + |a|); p0 is the pressure at the upper surface.
15. Use the fact that, at the point where the liquid contacts the wall, r is maximum
and equal to a.
17. Use the
DL
19. =
Dt

fact that v3 =
f D\ .
-dx =
YcDt

-p
3 1 . (ii) [T] =

0, vl2 + vi,\ = 0> y2,i _ yi,2 = 0


f
,D\
.

( 1
curl
I curl
-n</S=
n dS = 1I curl]--Vp
curl ] - - Vp++b{
b{ ( ndS.
)s
Dt
)s
I P
0

0 -p + 42
0

0
-p -

42]

34- Use equations (10.11.6) and (10.12.1).


36.
39.
40.
51.

2 2 "
[T ] = 2 - 2
1
L0 1 0
Use relation (10.13.10).
Use relation (10.13.14) and the equation of continuity.
If the x1 axis is along the direction of flow, the x3 axis is perpendicular to
the plane, is the inclination of the plane and h is the thickness of the layer, the
velocity field is
pg sin
(2 - ^^
2
(y)

Bibliography

Achenbach, J. D. (1975). Wave Propagation in Elastic Solids. North-Holland,


Amsterdam.
Aris, R. (1962). Vectors, Tensors and Basic Equations of Fluid Mechanics.
Prentice-Hall, Englewood Cliffs, NJ.
Batchelor, G. K. (1967). An Introduction to Fluid Mechanics. Cambridge University
Press, London.
Bowmen, R. M. (1989). Introduction to Continuum Mechanics for Engineers.
Plenum Press, New York.
Bowmen, R. M., and Wang, C. C. (1976). Introduction to Vectors and Tensors,
Vols. I and II. Plenum Press, New York.
Chadwick, P. (1976). Continuum Mechanics. George Allen and Unwin, London.
Debnath, L. (1994). Nonlinear Water Waves. Academic Press, Boston.
Erickson, J. L. (1960). "Tensor Fields," in Handbuch der Physik. Vol. /3
(ed. S. Flgge), pp. 794-858. Springer-Ver lag, Berlin.
Eringen, A. C. (1967). Mechanics of Continua. John Wiley and Sons, New York.
Eringen, A. C. (ed.) (1975). Continuum Physics, Vols. I and II. Academic Press,
New York.
Eringen, A. C , and Suhubi, E. S. (1974-1975). Elastodynamics, Vols. I and II.
Academic Press, New York.
Ewing, W. M., Jardetsky, W. S., and Press, F. (1957). Elastic Waves in Layered
Media. McGraw-Hill Book Co., New York.
Frederick, D., and Chang, T. S. (1965). Continuum Mechanics. Allyn and Bacon,
Boston.
581

582

CONTINUUM MECHANICS

Fung, Y. C. (1965). Foundations of Solid Mechanics. Prentice-Hall, Englewood


Cliffs, NJ.
Goldstein, S. (1938). Modern Developments in Fluid Dynamics, Vols. I and II.
Oxford University Press, London.
Gurtin, M. E. (1972). "The Linear Theory of Elasticity,,, in Handbuch der Physik
(ed. S. Flgge), Vol. VI a/2, pp. 1-273. Springer-Verlag, Berlin.
Gurtin, M. E. (1981). An Introduction to Continuum Mechanics. Academic Press,
New York.
Hunter, S. C. (1983). Mechanics of Continuous Media, 2nd ed. Ellis Horwood Ltd.,
Chichester, England.
Jaunzemis, W. (1967). Continuum Mechanics. Macmillan Publishing Co., New
York.
Jefferys, H. (1952). Cartesian Tensors. Cambridge University Press, London.
Kolsky, H. (1953). Stress Waves in Solids. Clarendon Press, Oxford.
Lamb, H. (1932). Hydrodynamics, 6th ed. Cambridge University Press, London.
Landau, L. D., and Lifshitz, E. M. (1959). Fluid Mechanics. Pergamon Press,
Oxford.
Landau, L. D., and Lifshitz, E. M. (1986). Theory of Elasticity, 3rd ed. Pergamon
Press, Oxford.
Leigh, D. C. (1968). Non-Linear Continuum Mechanics. McGraw-Hill Book Co.,
New York.
Love, A. E. H. (1944). The Mathematical Theory of Elasticity. Dover Publications,
New York.
Malvern, L. E. (1969). Introduction to Mechanics of a Continuous Medium.
Prentice-Hall, Englewood Cliffs, NJ.
Milne-Thomson, L. M. (1960). Theoretical Hydrodynamics, 4th ed. Macmillan
Publishing Company, New York.
Nowacki, W. (1963). Dynamics of Elastic Systems. Chapman and Hall, New York.
Prager, W. (1961). Introduction to Mechanics of Continua. Ginn and Co.,
Lexington, Mass.
Segal, L. A. (1977). Mathematics Applied to Continuum Mechanics. Macmillan,
New York.
Serrin, J. (1959). "Mathematical Principles of Classical Fluid Mechanics,*' in
Handbuch der Physik (ed. S. Flgge), Vol. III/l. Spring-Verlag, Berlin.
Sneddon, I. N., and Berry, D. S. (1958). "The Classical Theory of Elasticity," in
Handbuch der Physik (ed. S. Flgge), Vol. VI, pp. 1-126. Springer-Ver lag,
Berlin.
Sokolnikoff, I. S. (1956). Mathematical Theory of Elasticity, 2nd ed. McGraw-Hill
Book Co., New York.
Spencer, A. J. M. (1980). Continuum Mechanics. Longman, London.
Stoker, J. J. (1957). Water Waves. Interscience Publishers, New York.
Timoshenko, S. P., and Goodier, J. N. (1970). Theory of Elasticity, 2nd ed.
McGraw-Hill Book Co., New York.
Truesdell, C. (1977). A First Course in Rational Continuum Mechanics, Vol. 1.
Academic Press, New York.

BIBLIOGRAPHY

583

Truesdell, C. (1966). The Elements of Continuum Mechanics. Springer-Verlag,


New York.
Truesdell, C , and Noll, W. (1965). "The Non-Linear Field Theories of
Mechanics,'' in Handbuch der Physik (ed. S. Flgge), Vol. III/3. SpringerVerlag, Berlin.
Truesdell, C , and Toupin, R. A. (1960). "The Classical Field Theories," in
Handbuch der Physik (ed. S. Flgge), Vol III/l, pp. 226-790. Springer-Ver lag,
Berlin,
Wang, C. T. (1953). Applied Elasticity. McGraw-Hill Book Co., New York.

INDEX

A
Abnormality factor, 124
Absolute temperature, 351
Acceleration, 244
Acoustic wave equation, 492
Addition of
matrices, 13
tensors,49
vectors,7
Adiabatic
constant, 469
flow, 471
surface, 475
Adjoint of a
matrix, 27
tensor, 80
Adjugate of a
matrix, 27
tensor, 80
Airy's
solution, 342-343
stress function, 342
Almansi's strain tensor, 193
Alternating
symbol, 20
tensor, 57

Amplitude, 440, 503


Amplitude dispersion, 507
Angle between vectors, 9
Angular
momentum, 336
velocity, 260
Axial extension of a beam, 407
Axial vector, 74

B
Balance of
angular momentum, 336
energy, 344
linear momentum, 300, 331
mass, 376
mechanical energy, 349, 472
thermal energy, 349
Barotropic fluid, 469
Base vectors, 7
Beltrami's
field, 124
flow, 289, 493
solution, 340
vorticity equation, 330
Beltrami-Michell equation, 402-404
585

586

INDEX

Bending of a beam, 413-416


Bernoulli's
equations, 494
function, 494
Bernoulli-Euler law, 416
Biharmonic function, 393
Bobyleff-Forsythe formula, 512
Body force, 294, 377
Body point, 158
Boundary conditions
elasticity, 383
non-viscous fluid flows, 474
viscous fluid flows, 523
Boundary layer, 524
Boundary value problems
in elastostatics, 384
in elastodynamics, 387
Boussinesq's solution, 454
Boyle's law, 468
Bulk
modulus, 371
viscosity, 515

c
Caloric equation of state, 467, 520
Cayley-Hamilton theorem, 93
Cartesian
axes, 7, 34
coordinates, 33
tensors, 39-41
Cauchy's
deformation tensor, 193
equation of equilibrium, 333
general solutions of, 338
equation of motion, 332
law, 300
reciprocal relation, 295
strain tensor, 206, 364
stress postulate, 295
stress tensor, 300
Cauchy-Green deformation tensors, 193
Cauchy-Kovalevski-Somigliana solution,
455
Cauchy-Lagrange theorem, 281
Characteristic
equation, 85
polynomial, 85
Circulation
material derivative of, 279

preserving motion, 280


theorem, 279-280, 486
Clapeyron's theorem, 381
Classical heat conduction inequality, 353
Clausius-Duhem inequality, 352
Clausius-Planck inequality, 353
Coefficient of
shear viscosity, 515
thermal conductivity, 355
viscosity, 509-510
Cofactor of a
matrix, 27
tensor, 80
Collinear vectors, 7
Comma notation, 113
Compatibility conditions
for strain, 219-220
for stress, 403-404
Components of a
tensor, 40, 41
vector, 8
Compressible fluid, 466, 520
Compressive stress, 304
Configuration of a continuum, 157
current, 162
deformed (final), 168
initial (reference), 161
Continuum hypothesis, 166
Conservation laws, 325
Conservation of mass, 325
Conservative
force, 340
vector, 139
Constitutive equations, 156, 326, 355
Continuity equation, 326
material form, 327, 355
spatial form, 326, 355
Continuum (continuous medium), 157
Continuum hypothesis, 156
Contraction of a
material arc, 180
tensor, 51
Convective
part of acceleration, 246
rate of change, 245, 246
term, 528
Coordinate transformations, 33
Coordinates
Cartesian, 7
polar, 535

INDEX
Coplanar vectors, 11
Couette flow, 549
Couple stress, 296
Creeping flow, 531
Cross product of vectors, 10
Current position, 162
Curl of a
tensor, 127
vector, 120
Cylindrical tube under pressure, 424-<

D
D'Alembert's solution, 433, 436
Decomposition of deformation, 181
Deep water waves, 504
Deformable materials, 184
Deformation, 167, 241
finite, 205
homogeneous, 185
infinitesimal, 205
of a surface element, 171
volume element, 174
Deformation-gradient
material, 169
spatial, 173
tensor, 169
Deformation tensors
Cauchy, 193
Finger, 193
Green, 188
Delta symbol, 16
Density, 159-160
Derivative following a particle, 243
Determinant of a
matrix, 13
tensor, 78
Deviator stresses, 313
Deviatoric tensor, 81-82
Diagonal matrix, 12
Differential operator, 113
Difference of
tensors, 49
vectors, 7
Diffusion equation, 555
Diffusion of vorticity, 536
Dilatation, 207
Dilatational wave, 439
Direction
cosines, 10

ratios, 10
Directions of stretching, 259
Directional derivative, 118
Dispersion relation, 503, 507
Displacement equation of motion, 401
Displacement formulation, 398
Displacement vector, 200
Displacement-gradient, 201
material, 201
spatial, 201
Divergence of a
tensor field, 127
vector field, 119
Divergence theorem for a
tensor. 143
vector, 136
Dot product of vectors, 9
Drag on a
plate, 543
sphere, 560-561
Drag coefficient, 561
Dual vector, 74
Dynamic pressure, 464, 509

E
Epsilon-delta identity, 25
Eigenvalue, 83
Eigenvector, 83
Elastic
constants, 371-372
fluid, 469, 485
limit, 367
moduli, 365
potential, 378
waves, 431-450
Elasticity
tensor, 365
nonlinear, 367
Energy
equation, 346-349
flux vector, 350
internal, 345
kinetic, 344
mechanical, 349
potential, 472
thermal, 349
Entropy, 351
flow, 352
inequality, 351

587

588

INDEX

source, 352
specific, 351
Equation of
compatibility for
strain, 219
stress, 404
continuity, 326, 518
energy, 345-348, 518
equilibrium, 333
for an elastic body, 392
for a fluid, 477, 487
mechanical energy, 349, 472
motion (Cauchy), 332
for an elastic body, 391
for an elastic fluid, 485
for a perfect fluid, 477
for a viscous fluid, 527
thermal energy, 349
Euler's equation of motion, 477, 485
Euler's formulas, 253
Eulerian
description, 163
fluid, 469
form, 326
strain tensor, 195
Exterior product of vectors, 105
Extension of
a beam, 410
a material arc, 180

F
Field
equations, 325
line, 268
scalar, 111
tensor, 111
tube, 271
vector, 111
Finger's deformation tensor, 193
Finite deformation, 205
First law of thermodynamics, 345
Flow
Couette, 541, 549
creeping, 531
potential, 261
Poiseullie, 542
Stokes', 531
Fluid
barotropic, 469

compressible, 466
elastic, 469, 485
homogeneous, 467
incompressible, 465
in viscid, 463
non viscous, 463
viscous, 463
Force
body, 294
conservative, 340
gravitational, 478, 494, 500
surface, 294
Fourier's law of heat conduction, 355, 466
Fourier-Duhamel law, 354
Free
energy, 354
suffix, 3
surface, 315
Fundamental invariants, 78
strain, 228
stress, 308
tensor, 78

G
Galerkin solution, 454
Galerkin vector, 454
Gas
constant, 468
dynamical equations, 489
perfect, 468
Generalized Hooke's law, 364
Generalized vorticity equation, 358
Governing equations for
elasticity, 378
compressible non-viscous fluid flows, 470
compressible viscous fluid flows, 522
incompressible non-viscous fluid flows, 470
incompressible viscous fluid flows, 521
Gradient
of a scalar field, 117
of a tensor field, 126
of a vector field, 125
material deformation, 169
spatial deformation, 173
Gravitational force, 478
Green's
deformation tensor, 188
strain tensor, 189
theorem, 138

INDEX
Green-Lam solution, 456
Group velocity, 503

H
Hagen-Poiseullie flow, 543
Harmonie vector, 139, 340
Heat
flux vector, 344
supply, 345
Helmholtz representation, 139-140
Helmholtz free energy, 354
Helmholtz theorem, 281, 283
Homogeneous
deformation, 185
elastic solid, 365
fluid, 467
function, 509
material, 365
Hooke's law, 364, 367
Hoop stress, 395
Hydrostatic
pressure, 462
stress, 303

I
Identity
matrix, 15
tensor, 54
Incompressible
continuum, 175, 259
elastic body, 371
fluid, 465, 519
Index notation, 1
Infinitesimal
deformation, 205
normal strain, 206
rotation tensor, 213
rotation vector, 213
shear strain, 206
strain tensor, 205
stretch, 211
Initial
configuration, 161, 380
coordinates, 164, 380
Instant coordinates, 164
Instantaneous
configuration, 162
position, 162

Internal dissipation, 361


Internal energy, 345
Internal surface force, 295
Invariants
strain, 228
stress, 308
tensor, 77
Inverse
function, 162
matrix, 15
tensor, 70
Invertible mapping, 162
Invertible tensor, 70
Inviscid fluid, 463
Irrotational
motion, 261
permanence of, 281
vector, 139
Isentropic perfect gas, 487
Isochoric
deformation, 175
motion, 259, 516
Isothermal
perfect gas, 468
surface, 475
Isotropie
elastic body, 365
fluid, 509
tensor, 58, 59
thermal, 354

J
Jacobian, 165
positiveness of, 165

K
Kelvin's circulation theorem, 279
Kelvin scale of temperature, 351
KirchhofPs theorem, 208
Kinematic viscosity, 527
Kinetic energy, 344
Kinetic equation of state, 466, 520
Kronecker delta, 17

L
Lagrangian
deformation tensor, 195

589

590

INDEX

description, 163
strain tensor, 195
Lamb-Thomson formula, 348
Lamb's surfaces, 493
Lame's
moduli, 366
potentials, 456
pressure-vessel problems, 431
Laminar flow, 539
Laplace's equation, 501
Laplacian, 114, 129
Law of
balance of angular momentum, 337
balance of linear momentum, 331-332
conservation of mass, 326, 329
entropy, 351
Left
Cauchy-Green deformation tensor, 193
polar decomposition, 98
stretch tensor, 184
Levi-Civita -symbol, 20
Linear
combination of vectors, 7, 8
elastic solid, 364
momentum, 331
motion, 246
operators, 66
rotation tensor, 213
rotation vector, 213
strain tensor, 205
viscous fluid, 510
Local
rate of change of velocity, 246
speed of sound, 492
time-derivative, 242
Localization theorem, 165
Longitudinal wave, 439
Love
strain function, 455
waves, 446-450

M
Macroscopic study, 157
Magnitude of a vector, 7
Mass, 159-160
Mass conservation law, 325-326, 329
Material
arc, 159
body, 159
curve, 159

derivative, 242-243
of a line integral, 274
of a surface integral, 274
of a volume integral, 274
operator, 245
description, 163
Lagrangian form, 327, 485
law, 376, 518
point, 158
surface, 159
time derivative, 242
variables, 163
Matrix
difference, 14
nonsingular, 15
orthogonal, 16
of tensor, 43
of transformation, 36
product, 14
singular, 15
sum, 13
Maximum
normal stress, 306
shear stress, 306
Maxwell's solution, 341
Maxwell stress functions, 341
Mean pressure, 312
Microscopic study, 157
Mixed boundary value problem, 384
Mobile time derivative, 243
Modulus of
compression, 371
rigidity, 371
Moment of inertia, 414
Momentum
angular, 336
linear, 331
Morera's solution, 342
Morera's stress functions, 342
Motion
descriptions of, 160
Cauchy's equation of, 332
Euler's equation of, 477
Navier's equation of, 390-391
Navier-Stokes equation of, 527

N
Nanson's formulae, 200
Navier's equation of

INDEX
equilibrium, 392
motion, 391
Navier-Stokes equation, 527
Newton's law of viscosity, 515
Newtonian viscous fluid, 515
Nominal stress tensor, 317
Nominal stress vector, 317
Nonhomogeneous fluid, 468
Nonlinear dispersion relation, 507
Nonlinear elasticity, 367
NonNewtonian fluid, 516
Nonsingular matrix, 15
Non-viscous fluid, 463
Normal derivative, 118
Normal strain, 189
Normal stress, 304
No-slip condition, 523

o
Octahedral
plane, 305
normal stress, 312
shear stress, 312
One-dimensional wave equation, 432
Orthogonal
matrix, 16
transformation, 36
tensor, 70
vectors, 9
Orthonormal relations, 36
Oseen equation, 562

P
Papkovitch-Neuber solution, 454
Parallelogram law, 7
Partial differential operator, 113
Particle, 157
Path line, 266
Perfect gas, 468
Peripheral stress, 395
Permutation
identity, 25
symbol, 20
tensor, 57
Phase speed (velocity), 440, 503
Piola-Kirchhoff stress tensor
first, 317
second, 318

Plane
Couette flow, 541
flow, 265
harmonic wave, 440
Poiseullie flow, 542
strain, 425
stress problems, 343
waves, 435-441
Poiseullie flow, 542-548
Poiseullie's formula, 546
Poisson's ratio, 369
Polar
coordinates, 535
decomposition, 94
left, 98
right, 98
Polar decomposition theorem, 97
Polar materials, 296
Position vector, 33
Positive definite tensor, 94
Potential, 486
energy of a fluid, 472
flow, 261
Power, 344
Prandtl
boundary layer, 524
solution, 343
stress function, 343
Pressure, 464
dynamic, 464
mean, 312
static, 462
Primary wave, 440
Principal
axes of strain, 228
axes of stress, 307
axes of a tensor, 89
deviator strains, 231
directions of strain, 228
directions of stretch, 182
directions of stretching, 259
directions of stress, 307
direction of a tensor, 83
invariants of a tensor, 77
planes of strain, 228
plane of stress, 307
strains, 227
stress, 307
stretches, 182
stretchings, 259

591

592

INDEX

value of a tensor, 83
Product of two tensors, 51
Proper orthogonal tensor, 79
Proportional limit, 366
Pure shear stress, 304
P-wave, 440

Q
Quotient law, 54

R
Radial stress, 395
Range convention, 2
Rate of
deformation tensor, 257
rotation tensor, 262
strain tensor, 262
work, 344
Rayleigh
layer, 558
problem, 558
waves, 441-445
Reciprocal theorem of
Betti and Rayleigh, 383
Cauchy, 295
Reference configuration, 161
Referential coordinates, 164
Relative
displacement, 216
velocity, 260
Residual stress, 510
Reynolds
number, 529, 561
transport formula, 276
Right
Cauchy-Green deformation tensor,
193
polar decomposition, 98
stretch tensor, 184
Rigid-body transformation, 182
Rigid
materials, 184
rotation, 170
Rigid-translation, 169
Rotation-rate tensor, 262
Rotation tensor, 184

S
Saint-Venant's compatibility conditions,
220
Scalar
field, 111
invariant, 42
multiple of a
matrix, 13
tensor, 48
vector, 7
potential, 139, 486
product of
tensors, 51
vectors, 9
Schaefer's solution, 458
Second law of thermodynamics, 352
Secondary wave, 440
Semicolon notation, 165
Shallow water waves, 504
SH-wave, 440
Shear
modulus, 371
pure, 321
strain, 191
stress, 304-305
viscosity, 515
waves, 439
Shearing, 258
Similarity variable, 557
Singular matrix, 15
Skew-symmetric
matrix, 13, 14
part of a matrix, 14
Skew
part of a tensor, 73
tensor, 73
Skin friction, 542-543
Slip-condition, 524
Solenoidal vector, 137
Small deformations, 205
Spatial
coordinates, 164
deformation gradient, 173
description, 163
Eulerian form, 326
variables, 163
Spectral representation of a tensor, 92
Specific enthalpy, 471
Specific entropy, 351

INDEX
Specific heat at constant
pressure, 469
volume, 468
Speeds of elastic waves, 439
Spherical shell under pressure, 428-431
Spherical part of a tensor, 81
Square
matrix, 12
of a tensor, 52
Square-root of a tensor, 96
Stagnation point, 498
Static pressure, 462, 477
Steady laminar flow, 539
Steady motion, 246
Sternberg-Eubanks solution, 456
Stokes'
boundary layer, 556
condition, 516
expansion, 507
flow, 531
formula, 561
law, 510
problem, 556
theorem for a
tensor, 145
vector, 137
waves, 504
Stokesian fluid, 509
Strain
deviator, 230
deviator invariants, 231
energy function, 374, 378
invariants, 228
tensors, 188
Strain-displacement relations, 200, 376
Strain-energy functions, 374
Strain-rate tensor, 262
Stream function, 531
Stream
function, 531
lines, 269
tube, 271
strength of, 272
Stress
components, 297
deviator, 312
invariants, 313
equation of motion, 401
invariants, 308
matrix, 297

593

power, 346
tensor, 300, 337
for an elastic solid, 364-365
for a non-viscous fluid, 463
for a viscous fluid, 508
vector, 294
Stress-displacement relations, 391
Stress formulation, 401
uniqueness, 401
Stress-free surface, 315
Stress-strain relation, 364-365
Stress tensor for a non-viscous fluid, 463
Stress-velocity relation, 510
Stress waves, 432-433
Stretch
of a material arc, 180
tensors, 184
Stretching tensor, 257
Suffix
dummy, 3
free, 3
notation, 3
Summation convention, 2
Surface
elevation, 501
force, 294-295
traction, 295
waves, 441-445
S-waves, 440
SV waves, 440
Symmetric
matrix, 13, 14
part of a matrix, 14
part of a tensor, 73
tensor, 72
Symmetry of stress tensor, 337

T
Table of direction cosines, 35
Tangential stress, 304
Temperature, absolute, 351
Tensile stress, 304
Tensor
alternating, 57
as linear operator, 66
Cartesian, 39
difference, 49
equality, 47

594

INDEX

equation, 48
field, 111
invariants, 77
invertible, 70
isotropic, 58
multiplication, 50
orthogonal, 70
permutation, 57
product of tensors, 50
product of vectors, 40
skew,72
symmetric, 72
sum, 49
unit, 54
vorticity, 260
Thermal conductivity, 354
coefficient of, 355
tensor, 354
Thermodynamics,
first law, 345
second law, 352
Teodorecu's solution, 458
TorricellPs formula, 496
Torsion of beams, 418
Torsional rigidity, 421
Total mass, 160
Trace of a tensor, 46
Traction, 294
Traction-free surface, 315
Transformation
rule for a tensor, 41
rule for a vector, 39
Translation, 169
Transport formulas, 273-278
Transpose of a
matrix, 12
tensor, 69
Transverse waves, 439
Triaxial stretch, 182
Triple product
scalar, 11
vector, 11
Trochoid, 506

u
Undeformed configuration, 168
Uniaxial stress, 302
Uniform motion, 246
Uniqueness theorem in

elastostatics, 385-386
elastodynamics, 387-388
Unit
matrix, 15
tensor, 54
vector, 7

V
Vector
axial, 74
base, 7
components, 8
difference, 7
field, 111
gradient of, 125
magnitude of, 7
orthogonal, 9
position, 33
potential, 137
product, 10
stress, 294
sum, 7
triple product, 11
unit, 7
vorticity, 261
zero, 7
Velocity, 243
potential, 261
profile, 541
Viscosity
bulk, 515
coefficients of, 509
kinematic, 527
shear, 515
tensor, 509
Viscous
dissipative function, 513
fluid, 463, 510
stress tensor, 508
term, 528
Volume change, 207
Vortex
lines, 269
motion of, 281
tube, 271
strength of, 272
Vorticity
equation (Beltrami), 330
equation (Cauchy), 281

INDEX
equation (Helmholtz), 487
equation for elastic fluid, 487
tensor, 260
vector, 261

W
Water waves, 500-508
Wave
amplitude, 440
dilatational, 439
elastic, 431-450
harmonic, 440
length, 440
longitudinal, 439
Love, 446-450
number, 440
period, 440
plane, 435
primary (P), 440

Rayleigh, 441-445
resistance, 435
secondary (S), 440
secondary horizontal (SH), 440
secondary vertical (SV), 440
shear, 439
surface waves, 441-445
Weber's equation, 490

Y
Young's modulus, 369

z
Zero
matrix, 14
tensor, 47
vector, 7
Zorawski's criterion, 278

595

You might also like