D. S. Chandrasekharaiah and Lokenath Debnath (Auth.) - Continuum Mechanics-Elsevier Inc, Academic Press (1994)
D. S. Chandrasekharaiah and Lokenath Debnath (Auth.) - Continuum Mechanics-Elsevier Inc, Academic Press (1994)
D. S. Chandrasekharaiah and Lokenath Debnath (Auth.) - Continuum Mechanics-Elsevier Inc, Academic Press (1994)
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
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
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
(1.2.1)
i = 1, 2, 3
(1.2.2)
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
aikxk = bi
This is a concise form of the system of equations (1.2.1).
(1.2.5)
1.3
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
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
(1.3.2)
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
021*1*2 + tf22*f
^21
^23*3*2 = *22
(1.3.3)
#32
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)
(1.3.7)
1.3
#12 An]
a
#22
23
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
(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
+ <*kijxixj*k
(1.3.14)
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
ipaJpcj=
b c
(1.3.15)
OipOjp = bij
(1.3.16)
UJ
(1.3.17)
aiP<iip = bi2t
(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
SUFFIX NOTATION
(1.4.1)
(1.4.3)
(1.4.4)
(1.4.5)
1.4
(1.4.6)
(1.4.7)
(1.4.8)
(1.4.9)
(1.4.10)
(1.4.11)
(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
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
1.4
11
a2
a3
bx
b2
b3
Ci
c2
c3
(1.4.17)
(1.4.18)
12
SUFFIX NOTATION
(1.4.19)
(a x b) x c = (a c)b - (b c)a
(1.4.20)
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
13
KV = []
(1.5.3)
( M 1 r = [*</]
(1.5.4)
detfo,] =
^11
#12
#13
22
a23
#31
#32
#33
(1.5.5)
aflrn
afl12
a 13
ar
ar
22
23
(1.5.7)
(1.5.8)
(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)
14
SUFFIX NOTATION
(1.5.12)
(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)
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,
(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)
15
(1.5.19)
[by][ay] = [bikakJ]
(1.5.20)
(1.5.21)
(1.5.22)
(1.5.23)
(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)
(1.5.28)
-1
Given a matrix [ay], if there exists a matrix, denoted by [fly] , such that
[au\[auYl = [ ^ ' = [/]
(1.5.29)
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)
(1.5.34)
(1.5.35)
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)
_ 1
^0
(1.6.2)
That is,
^11 = ^22 = ^33 =
1.6
17
(1.6.3)
= u
(1.6.4)
= ji
(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)
EXAMPLE 1.6.1
form:
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
2 bkk + bl2,
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)
so that
bpp =
(1.6.9)
3 +
la -
(1.6.10)
a b
ij pp\
EXAMPLE 1.6.3
(1.6.11)
*-JZT*a_
(ii)
(iii) ^^ = aki
SuaJk = aik9
= a m
+ 122 +
= #11
Also,
= 2
+ 222
+ 13*32 = ^12
1.7
THE SYMBOL ,
19
aki.
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"
i '
i '
(1.7.2)
20
SUFFIX NOTATION
(1.7.3)
123 1>
132
1>
etc.
eUk = Ui-J)U-k)(k-i)
ijk
eixeJ
jki
kij
~ikj
~jik
~~ekji
(1.7.6)
(1.7.7)
= eiJkek
= 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)
(1.7.10)
1.7
THE SYMBOL
21
(1.7.11)
(1.7.12)
ijkaibjck
ax
a2
a3
bl
b2
b3
Ci
c2
c3
(1.7.13)
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
ijk
e, e3
<*
i3
j\
j3
k\
kl
k3
fc * e 3
22
SUFFIX NOTATION
EXAMPLE 1.7.2
Solution
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
for / = 2
for / = 3
Thus
r23 Zijk<*jk =
tf32
< 31 -
for / = 1
for / = 2
13
(1.7.16)
for / = 3
axl - alx
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
Oil
Oil
J\
J2
k\
k2
0/3
a
J3
(1.7.17)
k3
epqraplaq2ar3
(1.7.18)
1.7
Solution
pqraipajqakr
23
(i) We have
3
L
Li
p=lq=\
p=lq=\
Li
r=\
pqraipajqakr
<*ip<*jq(Zpq\<k\ + epq2<*k2 +
^pqS^ki)
3
=
+ ^ 3 ( ^ 1 3 ^ / 1 + ^23^/2
ei23<*il<lj2)
0*10,2)
0/2
0/3
0,1
0y2
0,3
0*1
0*2
0*3
(1.7.19)
pqraplaq2ar3
pqr^lp^2q^3r
= det(bu),
by (1.7.19)
= deHaji) = det(uri7)
(1.7.20)
24
1 SUFFIX NOTATION
ijk pqr*-'
(*ip
<*iq
<*ir
jp
1 akp
JQ
kq
(1.7.21)
Jr
kr
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)
(c) Hence
inSu - S
uSin = 3Sin - Sin
ijk PJk = S
'W^JJ
'U"JP
(d) Consequently
E
ijkeijk 20ji 6
2:,
SjqSip)
1.7
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)
- }^)
(1.7.28)
+ an(anak2
j3akl)
+ *i2(*/3**l ~
^jl^kl)
(1.7.29)
- aklaJ2)
Now
ijk<tii(aj2ak3 - aj3ak2) = ^ijkaixaj2ak3
= *Ukailaj2ak*
eUkanaJ3ak2
- Cikj<*il<lj2<lk3
(1.7.30)
Similarly,
i/**i2(**i*/3 - <*k*<*j\) = l*ukai2aklan
(1.7.31)
ijk^i2^ki^J3
= kij^il^J2^k3
Zijk<til<tj2<tkl
teukanaj2aki
(1.7.32)
k\<*j2) =
2ijk<til<tj2<tk3
(1.7.33)
26
SUFFIX NOTATION
6eijkananaki
(1.7.34)
ijkpqraipajqakr
(ipajqakr)
**
iT
jP
*jr
kp
kq
kr
(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)
(1.7.37)
(1.7.38)
EXAMPLE 1.7.6
Show that
l^ipqSjrsQprQqs
(1.7.39)
(1.7.40)
(1.7.41)
= li*M
1<$][] = [a*kiakj]
(1.7.42)
1-7
THE SYMBOL ,,
27
= 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
(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)
(1.7.47)
= ajCjbi
- ajbjCi
(1.7.48)
28
SUFFIX NOTATION
ipjqajbpcq
ajbjCi = aj(jpbp)(iqcq)
iqjpajbpcq
= 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
(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)
(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
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
apraqsb
apibpl
(ii) ap = aipbi
(iii) am = 0Ljmbj
(iv) ar = asrbs
(v) ak = otkmbm
(vi) ak = arkbr
22 = ^22
33
13 = i(^13 + ^3l) = 31
^33>
12 = 2 + *>2l) = 21
23 = 1(^23 + ^32> = 32
30
SUFFIX NOTATION
-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
[^^]
[ay]2.
and
0) 0 0 2
0
(ii)
2 3
-1
0-1
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
(i) u{au - a)
(iii) (u + au)(u -
ipjqapbjCq
a^
(ii) oLpiOLpj = u
+ bu
2 = >
2 5 . If 0f = eijkbjk,
-aki.
(ii) ^,,* = 0
(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*
= 0.
3 1 . If au = - and bt = jeijkajk,
eijkekrsnjvrns
32
SUFFIX NOTATION
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
(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,
2.2
COORDINATE TRANSFORMATIONS
35
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)
(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)
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
2.2
COORDINATE TRANSFORMATIONS
37
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
COS
0 + SHI 0
COS
0 - SHI 0
(2.2.14)
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
(2.2.15)
EXAMPLE 2.2.2
38
ALGEBRA OF TENSORS
-?-
1
2V2
2yj2
2V2
2V2
1
~2
2yj2
32 = +
V3
2 '
33=0
EXAMPLE 2 . 2 . 3
il
i2
i3
/I
<*/2
<*/3
A:l
*2
*3
(2.2.16)
(2.2.17)
+ (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
,-2
i3
[ei,ej, e^] = | o ^
a, 2
a y3
*1
A:2
A:3
39
In particular,
I 11
[ei,e 2 ,e^] =
12
13
32
(2.2.18)
33
(2.2.19)
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)
(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)
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)
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
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)
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
2.3.4
SCALAR INVARIANTS
2.3
CARTESIAN TENSORS
43
[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
(2.3.13)
Similarly, we write
[AL*...
= *</*...
(2.3.14)
N
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
[*(/] = - 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
C*M
2 = - 2 1 + 23 =
1
V2
V2
*3 = - 3 1 + 33 = 1
V2'
(2.3.17)
= l
*'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
13 = V2,
^3! = -V2,
r2i = - 1 ,
^ 2 = -V2,
a22 = 0,
#23 = V2
#33 = 0
[a'ij] =
V2
-1
V2 I
-V2 -V2
(2.3.18)
This
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
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
(2.4.1)
This proves the property for second-order tensors. The proof for higher
EQUALITY OF TENSORS
-ip&jq**pq
&iP&jq"Pq
= b'u
(2.4.2)
This proves the property for second-order tensors. The proof for higher
48
2.4.3
ALGEBRA OF TENSORS
TENSOR EQUATIONS
and bijk
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
'"
0Lipccjq(aapq)
(2.4.5)
2.4
PROPERTIES OF TENSORS
49
(2.4.6)
[al]ulc_._ = [A]^...
N
PROPERTY 4
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)
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
= OLipoLjqotkrcpqr
(2.4.11)
= ^ip^kq^mrOLjsCpqrs
(2.4.12)
(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)
oLipakqamroLjsapqbrs
52
ALGEBRA OF TENSORS
Hence
a'ikKj = <XiP<Xkq<Xkr<*jSaPqbrs
=
^iP^qr0LjSapqbTS
(2.4.16)
= 0Lip0Ljsapqbqs
oiipakqamrajsapqbrs
Hence
a
UbU
ipaJQair<Xjs<*Pqbrs
OprOqsQpqOrs
(2.4.17)
= apqbpq = dijbij
(2.4.18)
aucj = [A]u[e]j
(2.4.19)
A B = ^
(2.4.20)
= [A]/7[B]l7
AB * BA,
A B = B A
(2.4.21)
[A3] = [aikakpapj]
(2.4.22)
2.4
PROPERTIES OF TENSORS
53
(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)
(2.4.27)
2.4.8
QUOTIENT LAWS
(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)
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)
la = a
(2.4.32)
(ii)
IA = AI = A
(2.4.33)
(iii)
I - A = A - I = trA
(2.4.34)
(iv)
(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],
2.4
PROPERTIES OF TENSORS
55
= ibi = a b
= u = [\]u
EXAMPLE 2 . 4 . 3
(i)
(ii)
(a b - b (g) a)c = c x (a x b)
(iii)
A(a b) = (Aa) b
(iv)
Solution
we get
(2.4.37)
(2.4.38)
(2.4.39)
(2.4.40)
56
2 ALGEBRA OF TENSORS
aubiCj = a'ub'iC}
= 0
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
pqr
ipajqakraijk
2.4
PROPERTIES OF TENSORS
57
ijk
(x
see (1.7.17);
ip0Ljq0Lkrpqr
il
<*/3
/l
0/2
/S
itl
<**2
Jt3
(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
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)
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
(2.4.44)
(2.4.45)
dhn = 0Lihajnakr0Lmsc[jkmars
Since dhn = chnrsars, we get from (2.4.45),
(Chnrs - oiihajnakramscljkm)ars
=0
(2.4.46)
^ih^jn^kr^ms^ijkm
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
ISOTROPIC TENSORS
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..
e2 = e 3 ,
e3 = e!
(2.5.2)
all other a^ = 0
(2.5.3)
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
= a22
(2.5.4)
(2.5.5)
(2.5.6)
'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
(2.5.8)
(2.5.9)
(2.5.10)
(2.5.11)
(2.5.12)
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.
61
2.6
yimjk
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.
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
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
()
(2.6.5)
= a2233
(2.6.6)
Similarly,
02233 = 03311
(")
(2.6.7)
= a3322
(2.6.8)
Similarly,
03322 = 01133
1"V)
(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)
0i
O a a
'1221 = 01221 = lp 2q 2r ls pqrs
2332
63
(2.6.14)
Similarly,
=
#2332
< V ")
(2.6.15)
#3113
(2.6.16)
Similarly,
=
#3223
tf
1331
(2.6.17)
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
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
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
'
(2.6.28)
65
\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)
e2 = e 2 ,
e[ = -ei9
e2 = - e 2 ,
e3 = - e 3
(2.6.31)
e 3 = e3
(2.6.32)
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)
(2.6.36)
+ 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
llll
llll
lpalq0ilralsCpqrs
= ( + lKllll
by using (2.6.36);
Tcnii
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)
(2.6.40)
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
67
(2.7.2)
= [(Ab) + (Ac)L
A(ab + c) = cx(Ab) + (Ac)
(2.7.3)
by property (2.7.3);
= e, A(bjej) = e, ^(Ae,)
= (ef Ae,)*/
(274)
(2.7.5)
68
ALGEBRA OF TENSORS
Ae
g)
<*ipajqapq
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
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.
69
(2.7.6)
= <*ucJ*i
(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)
(2.8.6)
(2.8.7)
(2.8.8)
(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)
(AB)"1 = B^A" 1
(2.8.13)
2.8
TRANSPOSE OF A TENSOR
71
EXAMPLE 2.8-1
and b, show that
a Ab = b A r a
(2.8.14)
ajjibi
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],,
EXAMPLE 2 . 8 . 4
that
(2.8.16)
a^a^n^bj
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)
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,
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
73
(2.9.1)
(A - AT)T = AT - A = - ( A - A 7 )
(2.9.2)
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)
(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
EXAMPLE 2 . 9 . 2
that
Solution
A B = A-(symB)
(2.9.8)
(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
(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
jpiq)apq
= \ijkkPqaPq
(2.10.2)
(2.10.3)
(2.10.4)
75
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)
3
- 2
0 -,
(2.10.7)
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
e> + e,
(2.10.8)
k k
4kpqekrsapgars
psqr)apqars
&pq@qp)
==
l^pq^pq
(2.10.9)
||2 = +
(2.10.10)
(2.10.11)
(2.10.12)
2.11
INVARIANTS OF A TENSOR
77
(2.10.13)
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)
(2.11.3)
HK = \[au<tkk - <*ik<*ki]
a12
a2l
a22
*11
013
031
033
022
023
032
033
(2.11.5)
-IlA.
(2.11.7)
78
ALGEBRA OF TENSORS
(2.11.8)
(2.11.9)
DETERMINANT OF A TENSOR
= eUkaua2ja3k
(2.11.11)
det I = 1
(ii)
det A = det AT
(iii)
det(aA) = a 3 det A
(iv)
(2.11.12)
2.11
INVARIANTS OF A TENSOR
79
(2.11.14)
=0
(2.11.15)
= il(aibi)(ajbj)(akbk)
= 0
+ 2{aibk)(akbm)(ambi)
- 3(eA)(*^)(*A)]
(2.11.16)
(2.11.17)
~~lakiamkaim
~laikakmami
(2.11.18)
80
ALGEBRA OF TENSORS
so that
/// A = 0
(2.11.19)
(2.11.20)
(2.11.21)
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)
= ^ ( * )
(2 1L24)
(2.11.25)
(2.11.26)
using (2.11.22);
=
2ipq(rmSn
rnsn^Qpraqsambn
2.12
DEVIATORIC TENSORS
81
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)
(2.12.2)
(2.12.3)
If we set
a = itrA
= iakk
(2.12.4)
(2.12.5)
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)
1]
[_1
1J
[A] = [au] =
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).
Show that
83
(2.12.10)
= ~l(ik - -3<tmnik)(<lki ~ i ^ / m A i )
(2.12.11)
(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)
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
(2.13.4)
al2
au
a2\
a22 - A
a23
a32
a33 - A
det(A - AI) =
|
s -A
T31
3
+ 7AA - 7/ A A + IIIA
(2.13.5)
85
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)
= AxvXpvlp = Ai
(2.13.9)
86
ALGEBRA OF TENSORS
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
^lOi2pVip
using (2.13.7);
= \21
= 21
(2.13.10)
Similarly,
a'u = 0
(2.13.11)
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)
87
a)
(2.13.14a)
(2.13.14b)
(2.13.14c)
PROPERTIES OF EIGENVECTORS
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
88
ALGEBRA OF TENSORS
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
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
89
or
A! -
Aj - A
A! - A
(2.13.16)
= 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
[I =
*1
(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
0"
[A] =
[o 4 -3
Solution
3-
| = 0
-3 -
A2 = 5,
A3 = - 5
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 .
91
92
ALGEBRA OF TENSORS
(2.13.20)
(2.13.21)
*(**)
k= 1
(2.13.23)
(2.13.24)
93
(2.13.25)
theorem,
+ 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,
GA = 7// A I
(2.13.28)
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
(2.13.32)
///Ao =
(2.13.33)
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
(2.14.1)
2.14
POLAR DECOMPOSITION
95
(2.14.2)
(2.14.3)
Hence, A r A and A A r are positive definite. This completes the proof of the
theorem.
COROLLARY
96
2.14.2
ALGEBRA OF TENSORS
[vT] =
0 0'
0 -1 0
L l.
A = 4(* v t )
k=\
(2.14.4)
(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
k=\
VX^-a)2 > 0
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
B 2 = *(* v*) = A
k=\
(2.14.7)
Comparing this with (2.14.5), we find that B = B'. Thus B is unique, and
the proof of the theorem is complete.
THEOREM 2.14.4
form
(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)
(2.14.11)
(2.14.12)
(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)
2.14
EXAMPLE 2 . 1 4 . 1
POLAR DECOMPOSITION
99
[B] = | 0
6 |
(2.14.18)
13
ulku2k = 0,
ulku3k = 0
"2* "2* = 4,
u2ku3k = 6,
u3kUsk = 13
ulkuik
23 = "32 =
2,
*33
and other u^ = 0
3;
Thus,
[U] =
(2.14.19)
0 -2
(2.14.20)
[A A] = [A ] [A] =
13
(2.14.21)
100
ALGEBRA OF TENSORS
(2.14.22)
_1
4
1
2
1
2
[U]- = [IT ] =
ol
(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.
2.15
EXERCISES
101
(2.14.26)
(2.14.27)
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
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
6. Show that
OipOjgakparsijkr
qs
a, =
a3 =
vT
"V2
0
-1
0 -1
2.15
EXERCISES
103
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)
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)
ikjm
a Ab = A (a <g> b)
(a b)A = a <g> (A r b)
(Ae p ) e p = A
104
ALGEBRA OF TENSORS
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
a3
-a2
-a3
ax
a2 ax
tr A = tr(symA)
(ii)
tr(AB) = tr(BA)
(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
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)
(Hi)
trA* = // A
(iv)
106
2 ALGEBRA OF TENSORS
(i)
+ IIAI
(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)
(ii)
(iii)
mA
\-
A
ni A
III A -' =
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
0)
0"
0 3
' 3 - 1
(ii)
1 1
1 1
1 1 1 "
(Hi)
-1
0]
0
0 1J
"10
(iv)
2]
-2j
(0
- 1 0
1 0
0
-3
(ii)
1 0
1 0
0 1
(ii)
2 3
108
(Hi)
ALGEBRA OF TENSORS
2 -1
0"
0 -1
(iv)
1 0
-2
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
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)
^KWI^HAM
(3.2.2)
dt
(3.2.3)
da
dt
d /
db
da ^
(a b) = a + b
dt
dt
dt
(3.2.4)
(3.2.5)
3.2
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
112
CALCULUS OF TENSORS
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
= {[A I= {[AI
dt
^ 1
d\
dt
(dA>T
\dt
so that
Hence
"+<%)-
@y-m-it>
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
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
df = lkdxk
d fdf\ _ d2f
dx2\dxi/
dx2dxx'
dXi dxj
a2/
^f
/
,U
Jl
dXidXj
- f,ll
(3.3.3)
dXjdXi
Also,
fji
a2/
a2/
a2/
dx\
bx\
a2/
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)
_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
timnf.mn = ^inmf,nm
e
= ~ imnf,nm
= -imnftmn
EXAMPLE 3-3.2
(0
xu = u\
*,,, = 3
()
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)
dxx
by (i).
116
CALCULUS OF TENSORS
(iii) Hence
\XmXn)tij
Y\xmXn)ti\j
= imiXn)j
(Xm0ni)j
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
(0
(")
(**/*), = ; + 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/*)
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
abc* dx't
* dx'i
Hence
,i
(3A1)
=a
a* *
<*} = 0Lik<t>ik =
cLikak
(3.4.4)
(3.4.5)
118
CALCULUS OF TENSORS
3.4.2
(3.4.7)
- a
(3.4.8)
(3.4.9)
3.4
GRADIENT OF A SCALAR
119
3
-
(3.4.10)
(3.4.11)
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
^km^kn^m,n
(3.4.12)
= tkuk + uk k =
div(0u) = 0(div u) + V0 u
(3.4.14)
120
CALCULUS OF TENSORS
3.4.4
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
~ ,
oLipotmqocnramsankepqruktS
= <XipqSrkpqrUk,s
= ^ip^pqrUryq
<*ipap
(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)
=0
(3.4.17)
If v = , then
[curlv], = eijkvkJ
= Zijk^UkJ
= eUk(<l>uk)j
+
Zijk$jUk
(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)
(3.4.22)
(3.4.23)
SOME IDENTITIES
(3.4.24)
(3.4.25)
(3.4.27)
(3.4.28)
122
CALCULUS OF TENSORS
EXAMPLE 3.4.1
(i)
/=
(ii)
/()</| = /()()
(3.4.30)
= [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)
vu)Uj
(3.4.32)
3.4
GRADIENT OF A SCALAR
123
Similarly
[v x curlu + (v V)u]f = VjUjj.
(3.4.33)
UJVJJ
Solution
(3.4.34)
When written in the direct notation, this becomes the identity (3.4.34).
EXAMPLE 3 . 4 . 5
notation:
(3.4.35)
and
ukki = (uktk)ti
= [V(divu)]f
(3.4.35)'
124
CALCULUS OF TENSORS
|curlv| 2
(3.4.36)
(3.4.37)
v> x2772
v^
/r?^2
2
(V0) -
\)
(3.4.38)
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
ax}
a li
'
du'j dxn
dxn dx'j
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
(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)
(3.5.4)
(3.5.5)
(3.5.6)
(3.5.7)
(VV0) r = VV0
(3.5.8)
Evidently,
3.5.2
in all
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
(3.5.10)
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)
(3.5.12)
(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)
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
C' =
=
<Xip<Xjhqsrkpqr<thk,s
= aipajh8pqrahrg
aipaJhcph
(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
(3.5.16)
(3.5.17)
(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)
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
= 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)
U = i(uu
+ u
J,i)>
U = i(uij
- uJ,i)>
(3.5.23a, b)
(3.5.24)
(3.5.23a)'
(3.5.23b)'
130
CALCULUS OF TENSORS
(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,
(3.5.28)
[div Vu r ] 3 =
[divVu r ] 2 = 2CL(XX + x 3 ),
2OL(X2
+ xx).
Thus
div(Vu r ) =
2OL[{X2
[curl Vu r ] 2 3 =
-2 3 ,
[curl Vu r ] 3 1 =2xi
3.5
GRADIENT OF A VECTOR
1 31
and
[curl ]
=0
(3.5.30)
(i)
(3.5.31)
(div A) c = div(A r c)
(ii)
(3.5.32)
(iii)
(3.5.33)
= eimn([A c]n),m
eimn(ajnCj)im
= [curKA^L-
(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)
(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
~~
^imn\jnj)ym
~~ ^inm"ni,m
^nmi^*in,m
~^imn^*in,m
= eimn(ejpq[[AT]nq]tP)tm
~~
^jpq\Pimrfiqn,m),p
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)
(3.5.41)
r
(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)
= akbj(ukjari + ukaJ)
T
3.5
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,
(3.5.43)
~ ~^imn^jnp*^p,m ~
= (ijmp
^imn^jpn^p,m
- ipmj)wpm
= ijWpp -
witj
w = jcurlu
2
(3.5.44)
2
= |E| + fleuri u|
= div(Vu r )u - u V 2 u.
(iii)
(3.5.45)
(iv)
EW + WE = skw(Vu)2
(3.5.46)
(3.5.47)
134
CALCULUS OF TENSORS
(3.5.48)
(3.5.49)
= (UiUu)j
= il(ynT)u]j]j
UiUijj
[uU^uh
= div(Vu r )u - u V2u
(3.5.50)
(3.5.51)
= (UijUjXi
UjUiyU
(3.5.52)
Also,
Vu Vu r = uuUjj
(3.5.53)
(3.5.54)
(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
(3.5.57)
(3.5.58)
(3.5.59)
Solution
(3.5.60)
eirsejmnasntmr
ij
Sim in
rj
rm ^rn
(3.5.61)
(3.5.62)
~ \y*ss,mm
**mrymr)^ij
+ <*jm,im -
ij,mm
" ^im,mj
~ <*mmyij
(3.5.63)
Consequently, eirseimnasnmr
(3.5.64)
- amr,mr
= 0
(3.5.65)
(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
(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
1 37
]v
(ii)
=\
]s
(curlu)rfF=
]V
0nt/5,
or
(nxu)tfS
=\
}v '
or
}V
}S
(V2u)</K =
(iv)
]V
<t>nkdS
(3.6.2)
]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
SOLENOIDAL VECTORS
(3.6.6)
STOKES S THEOREM
u t ds =
Js
(curl u) n dS
(3.6.7)
138
CALCULUS OF TENSORS
Figure 3.3.
JS
or
(D 0^ife=
JC
JS
eUknj<l>tkdS (3.6.10)
3.6
()
(iii)
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)
(3.6.13)
HELMHOLTZ'S REPRESENTATION
140
CALCULUS OF TENSORS
(3.6.14)
= 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
div(a xu)dV=
Js
(axu)-nrfS
(3.6.17)
curl udV -
nxurfS} = 0
(a x u) ids =
Js
curl(a x u) n dS
(3.6.18)
3.6
141
(3.6.18) becomes
a I u x t ds -
(div u - Vur)n dS = 0
Show that
(3.6.19)
(3.6.20)
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)
W | .
(3.6.22)
142
3 CALCULUS OF TENSORS
(3.6.23)
(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
(i)
\E\2dV
(3.6.25)
(3.6.26)
Sv
Solution
\E\2dV
div{(Vu)u - (divu)u)dV +
(divu) 2 rfK
3.7
143
]S
]V
(div u)2 dV
(3.6.27)
\W\2dV=
(divu) 2 rfF>0
|E| 2 </F+
\W\2dV
(3.6.28)
|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
(3 JA)
144
3 CALCULUS OF TENSORS
Proof
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
\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)'
1,"-
=
(3.7.3)
3.7
V and on S,
lndr-\.
VndV=
145
(3.7.4)
nndS
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
. . .
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
Let C be a
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
n curl(A r a) dS =
=
Jc
Jc
(A r a) t ds
a (At) ds = a
Jc
At ds
(3.7.7)
(3.7.5)'
(3.7.8)
(curl A) W S = 0
or
Js
ejrsair>snjdS
=0
(3.7.9)
t(g>uds=
js
(nAVu)dS
(3.7.11)
3.7
Proof
147
Is
by (3.6.10);
=
n x V(u a) dS
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)
(3.7.13)
148
CALCULUS OF TENSORS
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
(Wu-VuJ)cdV=\
Js
( u n - nu)c</S
(3.7.14)
(3.7.15)
(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
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
149
(3.7.19)
(3.7.20)
(3.7.21)
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
(u (x) At) ds =
}s
(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
(3.7.23)
150
CALCULUS OF TENSORS
[u (curl ) + ( Vu)rAr] dS b
js
(ut)ds =
}s
(nAVufdS
(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)
3.8
EXERCISES
151
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
(ii) UJJ
+ Ujti)
(iv) (Uij -
ujti)
152
CALCULUS OF TENSORS
(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 + 2 + 2V0 .
3.8 EXERCISES
153
(ii)
]v
U^-w^\dS=\
W>V> +
(0VV - 2)
curlufK= 0
1 54
CALCULUS OF TENSORS
x <g) n dS = V\
aijtidV=
Js
dnias
(i)
(u -curl curl u) d V =\
(ii)
(Hi)
(u a)V2(u a) dV = -
(iv)
|a-curlu|2fK
|V(u a)|2 dV
,333 dV
h curl curl h dV =
]v
|curlh|2rfF
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
4.3
CONFIGURATION OF A CONTINUUM
157
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
4.4
1 59
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
'>0
(4.5.1)
for / = 0
(4.5.2)
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.
162
CONTINUUM HYPOTHESIS
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
163
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)
164
CONTINUUM HYPOTHESIS
t>0
(4.6.3)
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
165
'-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)
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
(4.6.8)
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
7=1 j
i= 1,2,3,
(5.2.1)
5.2
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)
(5.2.3)
(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
5.2.2
ROTATION
5.2
5.2.3
1 71
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)
(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
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)
= F*(dx x dx)
(5.2.8)
(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)
(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
1 73
ax
= xljdxj
(5.2.14)
d\ = (Vx)dx.
Comparing this with (5.2.13), we find that
F"' = Vx;
Consequently (since det A
-1
\-\
= 4J
(5.2.16)
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
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
1 75
rfx|,
(5.2.20)
(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)
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
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
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
Xi ^ 0 ,
x$ = x}
(5.2.28)
5.2
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)
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
Solution
\jj
L* Jy
Xjj
wirsjpqXp;r*q
(5.2.32)
%l\j
iJtirs&jpqXp.rXq.s
(5.2.33)
= i(F*)r
(5.2.34)
178
DEFORMATION
(5.2.35)
Since xfj are the components of F _ 1 , we find from (5.2.34) and (5.2.35)
that
_ J_
(5.2.36)
(5.2.37)
TF.
2 uc,jpq
x
F
Y s
irs-Y
p, r -*<7,
(0
^-] = ^-
0\Xi;j)
div(JF r j=0
(ii)
Solution
(1.7.26),
AP'1)*]*
(5.2.38)
(5.2.39)
~e^imn^pqrxi\pXm\qXn\r
*
Pj
=dx[
dx[dxL _ o
Xj iXi
dXidxp
- '->
(j.Z,*HJ)
5.2
1 79
we obtain
1
dJ
(5.2.41)
dbi-j)
[>
we have
XK
..
1
JXjil
J
dxXjdxf)
\JXJVJ
1 /dxj\
d / 1 \ dXj
~ J dxj \dxf)
1 d /dxj\
/ dXj\dxf)
1 dJ dXj~\
JdXjdxfl
(5.2.42)
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)
(5.3.3)
(ds)
1 = ^
(5.3.4)
kprp
L1C1 v
(ds0)
5.3
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
(5.3.5a)
F = VQ
(5.3.5b)
(5.3.6a)
V2 = FF r
(5.3.6b)
(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
183
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
185
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)
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,
= 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
(5.3.19)
186
DEFORMATION
(5.3.20)
(5.3.21)
EXAMPLE 5.3.4 For the deformation defined by the following equations, find Q, U and V:
xx = 2x1
x2 = -Xi
x3 = -2x1 - 3x?
(5.3.23)
5.3
Solution
187
[F] = I - 1
0 -2
-3
(5.3.24)
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)
*\i = 1,
3 = - 4
(5.3.26)
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)
(5.4.3)
(5.4.4)
2 = a0 Ca
(5.4.5)
(5.4.6)
5.4
STRAIN TENSORS
189
(5.4.7)
NORMAL STRAIN
(5.4.8)
(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
- ij)
(5.4.12)
190
DEFORMATION
+ Ig^a}1'2
- 1
(5.4.13)
(5.4.14)
SHEAR STRAIN
(5.4.1)'
2 = CtjW
(5.4.7)'
= 1+
(5.4.8)'
={\
+ 2gua?i$}l/2 - 1
(5.4.13)'
(5.4.15a)
(5.4.15b)
5.4
STRAIN TENSORS
191
(5.4.16)
Expressions (5.4.7), (5.4.7)' and (5.4.16) show that is completely determined by cu. Thus, the deformation components cu not only determine the
stretch of a material arc but also the final angle between two material arcs
whose initial orientations are known.
Using (5.4.12) and (5.4.15a), expression (5.4.16) can be rewritten in terms
of gjj as follows:
cos - cos 0O = 2gu<$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
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)
5.4
STRAIN TENSORS
193
(5.4.25a)
bl=49i4,j
(5.4.25b)
(5.4.26)
(5.4.27)
(5.4.28)
(5.4.29)
194
DEFORMATION
By using (5.4.25b) and (5.4.28) we obtain the following expression for the
components a^ of A:
(5.4.30)
(5.4.31)
(5.4.32)
5.4
STRAIN TENSORS
195
[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)
196
5 DEFORMATION
-?=-
vs
- *
- - *
(5.4.40)
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
-
1 -
~ F J*1*2
(5.4.42)
^)*1*2
197
= 0
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
ds
1+ e = =
ds0
dXi = {ds)a
= (<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
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
(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\
(5.4.48)
(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)
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)
(dsy2
= y2n T V ) = J2{n 1
(dS0f
Solution
follows:
(5.4.53)
(5.4.54)
(5.4.55)
= yV-(C-'n)
(5.4.56)
200
DEFORMATION
(5.4.57)
= / 2 (n-Bn)-'
(5.4.58)
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
(5.5.2)
5.5
STRAIN-DISPLACEMENT RELATIONS
201
(5.5.3)
(5.5.4)
(5.5.5)
(5.5.6)
( 5 5 4 )'
(5.5.5)'
(5.5.6)'
(5.5.7)
(5.5.8)
202
5 DEFORMATION
(5.5.9)
u = -
Ul
a
= -
U = & UU
x
U
(5.5.7)'
U -
JJ +
J,i ~
(5.5.8)'
k,iukJ
(5.5.9)'
k,iukj)
= 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)
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
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
(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
(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)
0-1
-1
01
0 = 1
0 -1 I
204
5 DEFORMATION
EXAMPLE 5 . 5 . 3
Solution
(5.5.20)
~JEkpqXp;kXq;i
(5.5.21)
= eiJk(xkJ
- xkJ)
= 6ijkxkj = ijkXjtk
(5.5.22)
(5.5.23)
~ ~ j jrs\Xi;rXj;
_ 1
~7 ^jrsxi
expression
iqjp)Xp.rXq.s
i;sXj;r)
s ~~
;rxj;s
~7 ^kpqXp;
k Xq; i
kXq; i
ksir)xp.rxq.s
q; k Xp\
i)
eijkejrs9
5.6
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
(5.6.3)
(5.6.4)
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
"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)
5.6
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
(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)
B" = I - 2E
(5.6.14a)
(5.6.14b)
(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
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
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)
(5.6.18)
ijk
i,kj
~uk,ij
~ukji
j,ki
jjk
~UiJk
or
"ijk = 0
(5.6.19)
5.6
209
"/ = 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.
= 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
" cos - 1
cos -- 1 0
(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)
ekkdV
ekkdV0
(5.6.26)
JfdV0
(5.6.27)
(1
+ekk)dV0
5.7
211
so that
V =V-V0=\
Jv
dV-\
J v0
dV0=\
Jv0
ekk dV0
(5.6.28)
(1 + ekk)ekkdV0
ekkdV0
Solution
(5.6.29)
(5.6.30)
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)
(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
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
(5.7.5)
(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
213
(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)
EXAMPLE 5-7-1
J;i)
2ijkUkJ
2ijkUk;j
HUiJ -
Jj)
(5.7.10)
(5.7.11)
where a is a constant, find the strain tensor, the rotation tensor and the
rotation vector.
x2
xx
x2
x1
(5.7.13)
Hence
*2
i(Xi
Xi)
Xi)
x3
2(, + x2)
(5.7.14)
Xi
x3
and
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
21 5
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
j = -a2,
cx = - a 3 ,
Cy = ~b%
(5.7.19)
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)
(5.7.23)
(5.7.24)
216
5 DEFORMATION
or
du = Erfx + Qdx
(5.7.25)
EXAMPLE 5 . 7 . 4
Q = I +
(5.7.26a)
(5.7.26b)
= 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)
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)
EXAMPLE 5.7.6
Solution
(5.7.29)
(5.7.30)
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)
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
(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
(5.8.4)
(5.8.5)
(5.8.6)
c u r l V a = V curl a
(5.8.7a)
(5.8.7b)
(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
(5.8.12)
curl(E - W) = 0
(5.8.13)
(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.
ij,kk
kkjj
~~ eik,kj
jk,ki
(5.8.15)'
2^23,23
33,11 + 11,33
2e31>31
11,23 ~~ 23,11
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)
EXAMPLE 5.8.1
components:
en = k(x\ + *f),
el2 = kxlx2x3,
(5.8.18)
(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.
k{X\ x2)9
e22 = kx{x2,
e2 = k X\X2
el3 = e23 = e33 = 0
(5.8.20)
(5.8.21)
5.8
COMPATIBILITY CONDITIONS
221
= -2k,
e22fil = 0,
e12>12 = k'
(5.8.22)
k = -k'
"2,2 = kX\Xi
(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)
(5.8.27)
\kx\
- f'{x2)
=c
(5.8.28)
222
DEFORMATION
- cx2 + cx,
g(xx) = cxl + c2
(5.8.29)
(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.
v
~ 3 J = --N
eiL
22 = e
33
E9 '
V*/
() eu
e 2i
elil2 = e"
2, = en = 0 (5.8.32)
= 0,
(5.8.33)
3T*1
b
(iii) el2 -e22-
(iv) eu - e22 -
(/
Xi),
*33 = - (I - *3>>
(5.8.34)
= 0
vM
xu
e33 =
M
~Xl'
(5.8.35)
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)
ij
a2 = -b39
(5.8.39)
a3 = -bx
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,
"*>> + >* =
uX2 + u2y ! = 0
2M
^*b *
(5.8.42)
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)
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)
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 = - ~ .
(5.8.51)
5.8
COMPATIBILITY CONDITIONS
225
/,2 = -.
(5.8.52)
/ 3 f i3 = 0
(5.8.53)
vpgl
JfXl
+ a \X 2 + C l
vpgl
= -~T*2 - ** + c2
,)
(5.8.54)
where UTJ , a3, 63 and c, are constants. Using these in (5.8.49) and (5.8.50) we
obtain
vpg
(5.8.55)
M2
= -^(/-x3)x2
(5.8.56)
Xl)\
These are the displacement components associated with the strains given by
(5.8.34).
226
5 DEFORMATION
U3 3 =
'
M
~mXl
l/2,3 + "3,2 = 0
(5.8.57)
+/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)
(5.8.61)
/3f22 = 0
vM ,
^2 */ i + Site).
(5.8.62)
h
vM
7
27
&(*)
(5.8.63)
(5.8.64)
vM 2
~2Xl
+ xXl
+ Cu
gl =
+ Cl
(5
*865)
(5.8.66)
227
"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
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)
228
5 DEFORMATION
(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,
IIIE = exe2e3
(5.9.5)
229
[E] = | a
0 I
4a
(5.9.6)
(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)
IIE = 8a 2 ,
IIIE = 0
(5.9.9)
ax + a2 = 0,
4a3 = 0
so that
ax = -a2 = ^ j ,
a3 = 0
(5.9.10)
(5.9.11)
e2)
(5.9.12)
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
(5.9.13)
- a(a0ka0k - 1)) = 0
(5.9.14)
=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.
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$p] is called the strain deviator matrix.
In Example 2.13.7, it has been shown that a tensor and its deviator part
have the same eigenvectors. It therefore follows that the principal directions
of E(c coincide with those of E. With respect to a set of principal axes, the
231
strain deviator matrix is purely diagonal, the diagonal elements being the
eigenvalues of E(c. These eigenvalues, e[d), e$*\ 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),
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)
eu2)
(5.10.3)
_ JL
if(2ei - e2 - e3)(2e2 - e3 - el)(2e3 - ex - e2)
(5.10.4)
[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)
(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
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
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,
(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
x2 = x2i
x3 = xl
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,
x3 = xQ3
xl = xx sin(ax3) + x2 cos(ax3),
234
5 DEFORMATION
X2=X*
+ ,
*3 = *3
///B
= 1
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
5.11
EXERCISES
235
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
*(*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
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 +
4(JC20)2,
(here a is a constant)
u3 = 2(xx)3 + 4x3
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)
= J
ZpQrXi;pXr,
5.11
EXERCISES
237
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 )
* 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)
"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 =
en = e23 = e33 = 0
(Here a and /? are constants.)
e:23,1
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
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
(6.2.2)
X
6.2
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
(6.2.3)
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
v = ^-(x + u)
v(6.2.5)
Dt
'
,dt
dt
244
6 MOTION
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)
\dt
OX;
(6.2.7)
6.2
245
/3\ I _ 3
\dt)\xt~'t9
idxj_
\dtt
Vi
do
D7 = 7 +
do
^ = 7 +
( v
(6 2 8)
^ = 7 + y ' ( ) " = 7
+ v
-v
(6 2 9)
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)
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)
W = 0;
6.2.7
(6.2.13)
'
LINEAR M O T I O N
^>
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
247
(6.2.16)
(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
dt
dt
T\
fdQ\{dtf
\dt)\dt
-(**QT)(QdQr
\dt
J\
dt
dt
248
MOTION
/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.
w, = 0 ,
3 = *3 - i(*2 + *)
x e
'
(6.2.21)
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
(6.2.22)
(6.2.23)
r,
Dt
(6.2.24)
6.2
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+>'-\(-)-'
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.
dv2
dt
(v - v)Vl = 0 ,
dv3
dt
(v V)v2 = x2,
(v V)y3 = 3
(6.2.28)
(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,
xx
15t ~ 1 + t
(6.2.30)
(6.2.31)
x, = x?(l + 0
(6.2.32a)
(6.2.32b)
x3 = 3(1 +
(6.2.32c)
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
251
v(6.2.36)
'
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
(i)
Dl
()
(6.2.44)
= -F _ 1 (Vv)
Dt
(6.2.45)
^ ( < / x ) = (Vv)(rfx)
(iii)
(6.2.46)
(6.2.47)
(iv)
Solution
Dt(Xij)
dt
= ( ^ )
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)
Dl
D
,
D~l
= IDt
T " = Dt
^ - ( F F ) = F - ^Dt +
D
Dt
F i
Hence
-F
Dl
Dt
D
,
F"1 = Vv
Dt
(tfx)
=Dt(F</x)=^V
Dt
\Dt
= (Vv)Frfx0
by (6.2.44);
= (Vv)(i/x)
6.2
253
= ^-(dxDt
= 2dx-
by (6.2.46).
dx) = 2dx
'
-^-(dx)
Dt
(Vv)(rfx)
EXAMPLE 6-2.7
(0
(ii)
formulas:
(6.2.51)
^ r ( l o g / ) = divv
(6.2.52)
(6.2.53)
Therefore,
DJ _
(D
D7 ~ ^^^^^^^
D
D
J^(X2>q)XUpX3,r + J^(X3;r)XUpX2;q
(6.2.54)
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)
Dt
JA-
*'*
(l J)
Dt
DJ
= JDl
(6.2.57)
(6.2.58)
(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
= {(divv) -
(Vv)T)J(FT)-l(dS0)n
(6.2.60)
6.2
255
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)
(dF/dt)
j7
\VF\
(6.2.62)
DF_ fdF
Dt " \di
= 0
-l + (y . V)F = 0
(6.2.63)
256
6 MOTION
(6.2.64)
(6.2.65)
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
(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
257
(6.3.4)
JL(*) = . . D .
(6.3.5)
W-dw
(6.3.4)'
(6.3.5)'
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
(6.3.8)
(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
259
ISOCHORIC MOTION
From (6.3.4) we find that trD = divv. Hence expression (6.2.58) yields
trD = divv =
(rfTf (rfF)
(6 311)
dV = constant
(6.3.12a)
2.
divv = 0
(6.3.12b)
3.
7=1
(6.3.12c)
260
MOTION
(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)
(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)
6.3
261
IRROTATIONAL MOTION
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
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
(i)
( v V)v = w x v + }v|v| 2
(ii)
(iii)
(6.3.19)
(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
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)
(6.3.22)
curl( ) = + curl(w x v)
(6.3.23)
\pi) ~~dt
(6.3.24)
(TF-'w) = / F " 1 c u r / ^
(6.3.25)
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
(6.3.26)
264
6 MOTION
D\\
= JF~l
DtJ
Dw
Dt
and noting by
(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)
(6.3.30a)
A = |(I - ( F - y F - 1 )
(6.3.30b)
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
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)
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)
~^'
"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)
(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
(6.4.1)
6.4
267
(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
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.
6.4
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)
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)
270
MOTION
(6.4.12)
are
djCj
dx2
dx3
dt
xl
2x2
3x3
\ + t
(6.4.13)
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
271
f-nrfS+|
f-ndS+\
t-ndS
(6.4.15)
. t-ndS-\
s2
fndS
Vortex tube
<^stream tube
Figure 6.5. Vortex tube and stream tube.
(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)
EXAMPLE 6-4.2
(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)
(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
(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)
^)
]c0DtKV
}c\Dt
6.5
TRANSPORT FORMULAS
sL*A-L(i+H'*'
275
(6 5 5)
</>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)
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
6.5.1
<=\+)}
(6 5 2)
Uv*dv=\Mdv+\s*-nds
(6 5 13)
(ii)
(iii)
(VvMx
(6 5 14)
(6.5.15)
i\cdx=\
(0
^-\
Dt
is
Js
dV=\
(aiyy)dV
(6.5.16)
6.5
TRANSPORT FORMULAS
277
Jv
]v
(trD)rfF
(6.5.17)
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}
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).
fn</S=
(Di _
_ _
. _ ,_
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)
IO
</>/?, tfS =
f (
n
\j^ j
+ <t>vk,k"j - <t>Vkj*k\ dS
Dt
[ r
. firijdS=
f {Vfi
nj+Mv^^nj-fiV^jn^dS
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
279
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
c
Dt
1 \Ds ^
)c(Dt
+ ( V v ) ,Tv
) y dxl
(6 6 3)
(Vv)rv dx = -\
2 Jc
V(v2) dx = -\
2 Jc
dv2 = 0
6.6.2
280
6 MOTION
dx = \ Vw - dx =
dw = 0
(6.6.5)
CIRCULATION-PRESERVING MOTION
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.
(6.6.7)
6.6
6.6.4
281
w = JFW
(6.6.9)
6.6.6
(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)
(6.6.12)
This equation shows that dx is the arc element of the vortex line at x (in the
282
MOTION
EXAMPLE 6-6.1
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 < +
Solution
(6.6.14)
(6.6.15)
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
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
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 . .
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
x3 = 4 + xxt + x\t2
x2 = x2 - x3(e' - e~%
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',
3 = >
wh
ere R2 = x\ + x\ * 0
6.7
EXERCISES
285
2 = - ,,
v3 = 0
1=-^9
1 2 . If dv/dt = 0, show that
tfj/Dt
v3 = 0
= 7div((div v)v).
*? + Ltx\y
X 2 = X 2,
3 = x3
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
286
19.
6 MOTION
For the motion defined by the velocity components
Vx = / ( / ? , 0*2 ,
2 = -f(R, 0*1 ,
V, = 0
v2 = - 4 x 3 ,
v3 = 4x2 - 3xx
(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
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
= 0
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
= Dp + Wp - (p Dp)p
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
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
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
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(:
%+/***)*
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
nx(vxw)tfS+
( n x ) dS
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
6.7
EXERCISES
291
* =
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\
?>
*=
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
294
7 STRESS
(7.2.1)
(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
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
296
7 STRESS
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)
s(3) = T3iteJt
(7.3.2)
7.3
STRESS COMPONENTS
297
(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
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)
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)
(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
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)
(7.4.12)
|s| = V3372
(7.4.13)
(7.4.14)
(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
[</] =
V/ = e, - 2x2e2 - 2x 3 e 3 ;
(7.4.16)
|V/|
V5
so that
n
\ = -7j>
2 = 0,
/2,
V5
(7.4.17)
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>
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)
(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.
(l)
(7.4.22)
(7.4.23)
(7.4.24)
n s(n') = n ( T V ) .
(7.4.25)
(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
7.5
305
= riknink
s(n) = n +
(7.5.4)
|s(n)| 2 = 2 + 2
(7.5.5)
(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
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 .
s(n) = T c o s a
(7.5.11)
|s(n)| = | 7 l | c o s 0 |
(7.5.12)
(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)
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
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
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
/,-,-
= 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).
(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)
=0
(7.6.9)
310
7 STRESS
' , 0 0
[T(/] =
(7.6.10)
s2(n) = 22,
s3(n) = 33
(7.6.11)
(7.6.12)
(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
(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)
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)
(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
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)
(7.7.5)
/// = <?\<
= ^
- 2 - 3)(22 - 3 - ^,
- 2 - 3)
(7.7.6)
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
[rf I =
0"
i
0 -i
-i
f.
4
3
u0
314
7 STRESS
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
(7.7.7)
(7.7.8)
= jlljid)
Solution
(7.7.9)
- r2r + (r2 - 3 + (3 - ,]
(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
31 5
(7.8.1)
(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)
T21 = 0,
T 31 = 0
(7.8.5)
316
7 STRESS
xi = 0
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)
J(dS0)(-lT)Tn
(7.9.3)
7.9
31 7
Setting
we get
T = J
(7.9.4)
(dS)s = (rfS0)(T0)V
(7.9.5)
(7.9.6)
where
s
= (T)Tn
(7.9.7)
(7.9.7)'
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)
(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
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.
[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
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.
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
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
for
n = -=(! + e2 + e3)
for
n = e2
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
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)
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
7.10
EXERCISES
323
tan 20 =
and the corresponding principal stresses are
-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,
324
7 STRESS
ij
~jxi\kTkj
() Tij = JXjikTkj
~Jxi\kxj\mskm
s
ikxj;k
=
ikxj,k
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
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
+ pdiv\)dV=
(8.2.3)
8.2
CONSERVATION OF MASS
327
'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)
(8.2.7)
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
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)
AN IMPORTANT CONSEQUENCE
D_
(pd>)dV =
() + div v dV
f
}[Dt
Dp
\
t.
X]
dV
)\
Dt \
J / Dt
(8.2.10)
8.2
CONSERVATION OF MASS
329
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 =
/ g + pdiv?j =0
This leads to equation (8.2.3) since 7 ^ 0 .
EXAMPLE 8 . 2 . 2
find the density in the current configuration in terms of the density in the
reference configuration.
330
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
3
1 + /
(8.2.13)
/> = 7 7 ^ 3
(i + ty
(8-2.14)
Dt
X;
1+ t
= jtf(i + /)
(8.2.15)
(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
(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)
(8.2.20)
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)
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
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)
<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
~~
ph
" divTlfrfK=
Since V is an arbitrary volume, the integrand must vanish identically (by the
localization theorem). Thus
dWTT + pb = p^
(8.3.7)
(8.3.7)'
(8.3.8)
8.3
BALANCE OF LINEAR M O M E N T U M
333
(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
]vo
Po*dV0=\
JK
0
JVQ
p0bdV0+\
J&
(T)TndS0
(8.3.9)
(8.3.10)
(8.3.11)
(8.3.12)
334
&j + Poi=Poj
Dv;
IXr.kSjklj + Pobi = Po^f
(8.3.11)'
mk + uixk)sjk].j + ^ = Po-^r
(8.3.12)'
Tj j = ,
12 =
T22 = %2
21 = 2X1*2
23 =
33 =
32 =
31 =
13 =
dToi
* 3
+
dXi
fan ,
dxx
+
dx2
fan ,
dx2
(8.3.13)
+ 2 = 0)
dx$
33
dx3
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
XT)
f(X\ 3
\ - 2*2
[
F'mdf(xx,
f(X\ >
x2
= -h
dxx
= 2
i =
T n
= Tu-
= -(X2-5)
on xx = 1
on
xx = 1
(8.3.15)
= 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
(8.3.17)
+ 27^
(8.3.18)
dT
2
+ - rr,
T = 0
dr
r
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+\
(x x s) dS
(8.4.2)
8.4
337
(8.4.3)
{xx (pb)J dV + f (x x s) dS
(8.4.4)
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)
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)
(8.4.8)
338
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)
V2h = -pb
(8.5.3)
(8.5.4)
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
339
(8.5.5)
(8.5.6)
div T + pb = 0
(8.5.7ab)
(8.5.8)
(8.5.9)
(8.5.10)
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)
340
(8.5.12)
+ */ f i ~ **,*<*</
(8.5.13)
where
aij = aji and hi^kk = -pbi
8.5.1
(8.5.14)
(8.5.15)
T0 = T - * I
(8.5.16)
where
Also, equation (8.5.2) can be rewritten as
T0 = T0r
(8.5.17)
(8.5.18)
(8519)
(8.5.20)
where
au = and hUkk = 0
(8.5.21)
8.5
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
A 2>2 + A33
(8.5.22)
23
31
PARTICULAR CASES
L fJ
_'22
,U
(8.5.23)
21 12 ~, 12
32 =
23 =
~,23
= 3 =
-,
* 0
* *
*
0
and A, = 0
342
= -
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
Suppose
vf
and
ht = 0
12
21 = ~, 12
23
32 =
13 =
(8.5.25)
31 = 0
Suppose
0 0 0
[*(/] =
0 0 0
0 0
and , = 0
8.5
343
12
13
21
31 =
(8.5.26)
~~ , 12
23 =
32 =
13 =
33 = 0
Suppose
[*</]
and
^ = 0
ll
2 2 T 33
12
T23 = - / i a f l
(8.5.27)
31 = 2
where
= *9
'
*0
344
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)
pb-\dV+
s-vrfS
(8.6.2)
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.
8.6
BALANCE OF ENERGY
345
H = \ phdV
(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
-\A
D\
De)
v + [dV
Dt
Dt)
(8.6.8)
s v dS = \ v dS = \ n Tv dS
js
Js
div (T\)dV
(v div T 1 + Ty Vv) dV
(8.6.9)
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
")
p - Vv + divq - ph = 0
(8.6.12)
(8.6.13)
8.6
8.6.1
BALANCE OF ENERGY
347
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
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)
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
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)
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)
1
(x w x v) dV + I j - v x - (x v)v} n dS
(8.6.22)
formula.
(8.6.23)
(8.6.24)
- 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
EXAMPLE 8,6.2
(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
}\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
+ E) =
p(b v + h)dV +
(s v - q n)dS
(8.6.30)
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
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
Js
(8.6.31)
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)
(8.6.35)
'
8.7
ENTROPY INEQUALITY
351
= tr T
De
DG
o n
= S - divq + Poh
^
Solution
(8 6 36)
-'
(8.6.37)
= (TV ^ = tr [ T 0 ) ]
(J)T
t\
D\s
D\J
(Vv) = - S - = S
2
Dt
Dt
(8.6.38)
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
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)
-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
(8 7/7)
(8.7.8)
8.7
ENTROPY INEQUALITY
353
(8.7.9)
(8.7.10)
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
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)
"
+ div
(^) -
(8 7 13)
-'
354
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)
(8.7.17)
(8.7.18)
Since this inequality holds for arbitrary VT and the equality in (8.7.9) is
8.8
CONSTITUTIVE EQUATIONS
355
(8.7.19)
q = -kWT
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)
Equation of energy:
De
p
De
p0
= T Vv - div q + ph
(spatial form)
(8.8.3a)
(material form) (8.8.3b)
356
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
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\
v2 = -xYx2,
v3 = kxxx3
x2 = bxx + (1 + a)x2i
x3 = x3
b2rl
div(
^v>
1 1 . Show that
Dy
~Dt* 7 ( / ? v )
+ dlv(/?v
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,
22 = * + v(xl - x])
12 = -21^X2,
23 = 31 = 0
[*</] =
xlx3
x2(l - x\)
x 2 (l - x\)
x\ - 3x3
2xr?
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
/ T n dS =
}v
fsdS
Deduct that
, , -1.
(i) I TdV=
(ii)
T (Vu) dV = \ s u dS +
/?(b - V) u </K
[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
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.
40.
~{)
+ )
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
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
+ "" +
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
365
(9.2.3)
' = c[jklekl
+ 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
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
1
2
3A + 2
(trT)I
(9.2.8)'
A + 2
+ 2
*12
'23
Til
22
+ 2
0 1 pul
0
en
<?33
ei2
e23
2\
L ^31 J
(9.2.6)"
9.2
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)
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)
(9.2.11a)
E(d) = E - | ( t r E ) I
(9.2.11b)
= 2//E<d>
368
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"'
= 33 = - 2 3
+ 2//))
^ ^
9.3
369
If we set
E =
(3 + 2)
+
.
v = 2( + )
(9.3.2a)
(9.3.2b)
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
- (i + Z -
">
(9 3 61
" - 2<rb>
"
~ 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
9.3
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)
,
=
(1 + v)
2.
A + f
3 -
3 - A
9/f(* - A)
zv
\{-)
'
(,)
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
373
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)
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)
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
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
1 + v
kk
(9.3.22)
dw
Noting that dxkk/dxu
(9.3.20).
1 + v
v
u~ETkk^
L
Tii
bxkk
(9.3.23)
+ 2//E
(9.3.24)
9.3
375
(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
*33 = ~ ^ * 1 >
(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>
376
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)
div v (div u)
dty
D\
d2u
~Dt "dt2
'
(9.4.6b)
(9.4.6c)
9.4
GOVERNING EQUATIONS
377
d
at
(9.4.7)
(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)
(9.4.11)
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
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)
379
(9.4.16)
(9.4.18)
380
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)
\ekkdV
1 - 2v
y
1 - 2v
jv
1 - 2v*
rkkdV
TijijdV
1 -2v
(TijXiXjdV-
TijXijdV
TijjXidV
(9.4.20)
1 - 2v
TijrijXidS + I I/, -
p-^)XidV
(9.4.21)
~ 2 V | | s,x,dS+ I fiXidV
s
Jv
(9.4.22)
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
= Muu
+ uj,i) =
T u
uu
U=Z
(jijU^jdV-
TijjUidV
(9.4.24)
=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.
382
\$sPuPdS
+ ^(/r
pd-^p)uVdV
- p ^ f ) , ( , ) dV
= j s /> dS + ^(fV
(9.4.27)
r#> = X6ue + 2;
rff> = A V # + **$>
These give
^eJP = + 2^><>
which is (9.4.26).
Using the symmetry of the stress components, the strain-displacement
relations, the divergence theorem, the Cauchy's law and the Cauchy's
equations of motion, we find that
^efdV=^u<?]dV
=
\J,TW\jdV-^T\WdV
S
= \ssM2)dS
JV
+ | ^ / / - p^pju?UV
(9.4.28)
Similarly,
vxfef
By the result (9.4.26), the lefthand sides of (9.4.28) and (9.4.29) are
equal. Hence their righthand sides are also equal. This proves (9.4.27).
In the absence of inertia, (9.4.27) becomes
f sil>u&dS + f ffl>ufdV = f s&uWdS+ f f&uPdV
(9.4.30)
Js
}v
]s
]v
The lefthand side of (9.4.30) represents the total work that would be done
by the force system {sf1},./)(1)} in acting through the displacements uf2) while
9.5
383
the righthand side represents the total work that would be done by the force
system {$\./)(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
9.6
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)
(9.6.2)
divT + f = 0
(9.6.3)
on
Tn = s*
SU9
on
ST
(9.6.4)
in
in
(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
Suppose we set
= u (1) - u (2) ,
= E (1) - E (2) ,
T = T (1) - T(2>
(9.6.9)
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
{A(tr E)2 + 2 E} dV
(9.6.13)
i = \jtuudv
=
(//"),7 -
*UJUAdV
= \ fuinjdS-\
TijjidV
(9.6.14)
(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)
in
(9.6.17)
9.7
387
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)
du
= v* for t = 0
(9.7.4)
on S
for / > 0
(9.7.5)
388
9.7.1
= )_(2)
(96)
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)
(9.7.9a)
d
- = 0
(9.7.9b)
in V for / = 0. Let
(9 7 10)
- '* %'%]
v2
du
du
dV
= |^ [ A ( ^ ) 2 + 1vuu + p ^ J j dV
(9.7.11)
= 2 (-f + ^ f ) + > ? ^
*''
9.7
389
at
\dtjj
(9.7.13)
f f- ida-\
da a2
i "0 ...
(9.7.14)
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)
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
in V for all t > 0. The relation (9.7.18b) together with (7.7.7b) yields
T =0
(9.7.20)
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
(9.8.6)
a 2 ,
= p-^
(9.8.6)'
(9.8.7)
(9.8.8)
1
1 - 2v
(9.8.9)
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
9.8.1
STATIC CASE
(9.8.12)
(9.8.12)'
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)
9.8
NAVIER'S EQUATION
393
EXAMPLE 9 . 8 . 2
(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)
394
EXAMPLE 9 . 8 . 3
body is given by
(9.8.26)
EXAMPLE 9 . 8 . 4
(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)
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
= <t>(r)x
() = -^()
(9.8.31)
(9.8.32)
9.8
NAVIER'S EQUATION
395
= 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)
(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
(9.8.38)
396
0(r) = 4 + B
(9.8.40)
u(r) = 4 + Br
(9.8.41)
(9.8.42)
R*0
= 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)
9.8
NAVIER'S EQUATION
397
^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)
+ '&)
(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)
iS0"
= 0
(9.8.50)
398
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
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)
[UijjUi + (A + )uktkiUi
- pUiUi\
dV=0
)ktkj]jdV
+ (A + )ktkiti
[ijij
/,-,]
rfK = 0
(9.9.5)
Iiji
-\y
)ktkj]njdS
1 MuVfii.j)
(9.9.6)
(9.9.7)
400
(ijtjdv^o,
(ktk)2dv=o,
fiidv^o
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)
(9.10.2)
-(trT)I
(9.10.3)
(9.10.4)
-^
V(divT)r-^_L^
v
KtrT)
1+ v
+ Vf + Vf = 0
(9.10.5)
9.10
STRESS FORMULATION
401
ik,kj
jk,ki
2p(l + v)
tf ""^"*** + fu+fj.i
(9.10.5)'
f = Tj
(9.10.6)
(9.10.7)
UNIQUENESS OF SOLUTION
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)'
402
~ ^-ik^ijj
(9.10.8)
^ik,kj^ij
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)
(JkkY - ( -
J^ijTkk
*")
1 1 -2v
kk)
3 1 + v (fkkf
(9.10.11)
/>(1 + v)
2STkk)[*
- ^UTmm)dV
+**\<-'
(9.10.12)
fy-Wn^O,
tts0
(9.10.13)
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
403
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)
(9.11.3)
(9.11.4)
(9.11.5)
so that
i
2 lJ
r a2
1 + v IdXi dxj
+ ^
)
l
lJ
kk
*Ufk.k + fu + fj.i = 0
(9.11.6)'
(9.11.7)
404
22 = ^3^1
12
3>
13
23
33
*ii,i +
2,2 + T 1 3 i 3 = 0
(9.11.8)
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.
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.
#>
+
(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
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)
(9.11.13)
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)
Show that
1
D ^ T = T ^ [ ( W - vD 2 I)divf - (1 - vJD^Vf + Vf r )]
Solution
(9.11.15).
(9.11.15)
9.12
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
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
408
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.
409
= 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)
r max = ^V;
crmin = 0,
Tmin = 0
(9.12.8)
410
9.12.2
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
= /
h^_D^
*2
1
1
n
!
*3
(9.12.11)
9.12
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
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)
(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)
(9.12.19)
9.12
413
2AE
Relations (9.12.17) and (9.12.20) are identical, and the required result is
proved.
9.12.3
(9.12.21)
~~~~?
-7
W J
x;
'
*3
414
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
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
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
(9.12.30)
(9.12.31)
= cl
</l - ^'
(9.12.32)
9.12
I.
2h
f-
41 7
2b
1/
1
\
If
X2
1
I
t
Tx,
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
c2 + vh2 - v(jc2)2(l + -x
R
2R
h + -{c2 + v/*2
2/?
/ 2
v(*D
(9.12.35)
418
9.12.4
\x(-s{3))dA
= -m
(9.12.36)
(9.12.37)
9.12
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)
31 dA = 0,
JA
x32dA = 0,
JA
(32 - x2T3l)dA = M
(9.12.40)
(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)
13,3
= 0,
23>3 = 0,
T 3M + 32)2 = 0
(9.12.43)
420
32 = 2 cx x
a
(9.12.44)
(b2xj + a2x\)dA =
xl dA = 0,
(9.12.45)
22
33
2M
12
(9.12.47)
_2M
(9.12.48)
2 = OOC\X$>
U3 = -Oil ~2
TJ )XlX2
(9.12.49)
9.12
421
where
a2 + b2 M
>0
(9.12.50)
3^3
a =
a5b
(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)
P'Ui ,X2)
- - ^ P ( X l ,X2*
X1 .
422
2M
2
7
s = T31e! + T32e2 = 3T3 (-0 x2Ci + b ^2)
na o
(9.12.54)
(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
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)
I2 = ^ab(a2 + b2)
(9.12.63)
4n3ab
A4
424
where
D = DC = ^ An Ic
(9.12.64)
= ^a4
(9.12.65)
Ic = (Ie)b
=a
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
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
425
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
u = AR
(9.12.68)
426
(9.12.69)
(9.12.70)
l
(+)--^ - = --
(+)-^
(9.12.71)
= --
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)
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
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)
(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)
= ~^^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
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.
(9.12.84)
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)
9.12
429
(9.12.86)
(9.12.87)
+ (3A + 2)
(9.12.88)
- q)
3
-a
pa3 - ab3
(3 + 2) = *,3 _ 3
(9.12.90)
430
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)
- 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)
(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)
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
where
2d
2r
i i _ d2Tn
dx2 * dt2
a
E
=
(9.13.2)
(9.13.3)
9.13
ELASTIC WAVES
433
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
or
(9.13.10)
(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)
^ 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
(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)-
436
(9.13.16)
=0
(9.13.17)
= (
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)
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
<9 2)
"
for *! > 0, / > 0, where the functions / and g are to be determined. Using
9.13
ELASTIC WAVES
437
H)=A
f
KcJ
Xi
for
t> 0,
= A for
x,>0
(9.13.22)
-i
*i
t<
1
X
F\t- -\
for
(9.13.23)
/>ii
-1/
<"l
-rM' cj
for
t<
(9.13.24)
'>r
a>-^<--rTV<'>
K'-$-fwr",'w*
<9 13 251
Putting this expression for F(t - xx/cx) back into (9.13.23), we get the
438
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 =
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)
9.13
ELASTIC WAVES
439
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)
440
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)
(9.13.33)
K-?)
= A cos [ t
) + / sin \ t
(9.13.34)
K'-2)}
ux = 1 {/ f -
(9.13.35)
9.13
ELASTIC WAVES
441
(9.13.36)
RAYLEIGH WAVES
-axi
exp
M-?)}
(9.13.37)
442
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
\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)
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)
fB)}
(9.13.44)
*)
(9.13.45)
+ -Be~aiXi
+ (l -
\^Be~^A
32 = 0
(9.13.46)
*~<.!((*- >-***where we have suppressed the factor exp{/cc;(l - /c)} for brevity.
444
A + 2a B = 0
2 2
2OL x
(9.13.48)
= 0
20Li
~2\2
2 \ 1/2
= 41 1 -
,2\ 1/2
c\
(9.13.49)
(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
(9.13.52)
=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
which simplifies to
axc = 0
(9.13.55)
(9.13.56)
(9.13.57)
9.13.5
LOVE WAVES
9.13
ELASTIC WAVES
447
(9.13.58)
(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)
448
where
/>2 = ( l - )
(9.13.63a)
c\ = ^
P
(9.13.63b)
(9.13.64)
(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
(9.13.67)
and
in the half-space
in the layer
(9.13.68)
in the half-space
(9.13.69)
in the layer
(9.13.70)
- CcoshH
= 0
(9.13.71)
9.13
ELASTIC WAVES
449
-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-
m-'YH
(1 - C 7c 2 2 ) 1 / 2
fiic'/ci - 1)
1/2
(9.13.75)
450
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
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
u2 = 0,
M3
= ke
(ii) Wi = 23 ,
u2 = kx3xl9
u3 = kxxx2
(iii) ! = kx2x3,
u2 = x 3 * lf
w3 = k(x2 - x2)
w2 = Arv*^,
w3 = kx3xl
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
(1 - v)E
(1 + v)(l - 2v)
1+ v
dn
,./.
J J +
uk krii
1 - 2v '
where
n-V
and
u = -(uid
- ujti)
r2
.. c - ci
(ii)^
1 - 2v
c]
1
2(1 - v)
= 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)
9.14
EXERCISES
453
Axl
Ax2
U2 =
Mi = ,
u2 = Bxxxlx3,
M3 = Cxxx2x\
V2</> = 0
V2g = 0,
V A = 0,
a = constant
a = constant
V2h = 0
u= h
-2(rbj v *
454
2 8 . If h and h* are such that V2h = curl h* and V2h* = 0, show that
u = h*
1 - 2v
curl h
r2
V 2 g
- 2 ^
-(\/).
V ( d i V g )
9.14
34.
EXERCISES
455
where a and b are constants. Find the body force that must be acting. Compute the
corresponding displacements and stresses.
35.
(i) g = r2e3
(ii) g = &3
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
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.
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 )
456
1 - 2v
curl h
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
curlu = - 2
D2(curl u) = 0
9.14
EXERCISES
457
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
T22 = bx3xl9
13 = ex2,
where a,b,c,
12 = dx\,
^ = cxxx2,
23 = jX\
T33 = cxxx2x3
T22 = DX\X2x3>
Ti2 = ( + bxl)x3,
23 = (bxl + cx\)xx,
13 = (cx\ + ax\)x2
ll
,22>
22
, 11
12
, 12
12
13
33
458
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 =
(x s) dS +
s
2(3 + 2)
(x f) dV
}v
x-sdS
}v
x-fdV
9.14
EXERCISES
459
f u dV
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
Figure 10.1.
Hydrostatic pressure.
(10.2.2)
(10.2.3)
10.3
463
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)
(10.3.2)
464
10
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
465
INCOMPRESSIBLE FLUIDS
(10.4.5)
(10.4.7)
(10.4.8)
466
10
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
(10.4.10)
(10.4.11)
(10.4.12)
(10.4.13)
10.4
467
468
10
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)
(10.4.15)
(10.4.16a)
(10.4.16b)
10.4
469
P = constant
P1
T
p
= constant
(10.4.17a)
(10.4.17b)
where
y = (l + - ) > 1
(10.4.18)
(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)
(10.4.21a)
p = p(p)
(10.4.21b)
470
10
divv = 0
(10.4.5)
D\
divT + /?0b = /> 0
(10.4.7)
T = -pi
(10.4.4)
2. Equation of motion:
3. Material law:
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)
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
471
M=I^_*^
(10 .4.24)
Dt \pj
p Dt p2 Dt
equation (10.4.22) reduces to equation (10.4.23).
+- =
dp
+ constant
JP
(10.4.25)
+ - = , 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)
constant
(10.4.30)
472
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).
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)
10.5
473
p(VX-v)dV=-\
P(VX-v)dV=-\
( v * - ^ W = - j
P^dV
(10.4.37)
= - \
pxdV
(10.4.38)
= 0
(10.4.39)
(10.4.40)
= 0
(10.4.41)
(10.4.42)
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
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)
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
475
(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)
=
*'
-^
(10.5.9)
476
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
v2 = -6xy9
v3 = 0
(10.5.11)
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
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)
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
1 -,
= + y(v V)v = + w X v + - VIT
Dt
dt
dt
2
(10.6.4)
(10.6.5a)
(10.6.5b)
10.6
479
(10.6.6b)
3y
dp
dz
(10.6.6c)
(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)
)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
_--H
\ ~~ "
J>
= a>e3 x (a>e3 x x) =
~Dt
(10.6.10)
= 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
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)
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)
dp
fdy
= o,
dp
-f
dz
= -gp
482
10
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)
p = Plexp(-^)
exp
(,0 6 24)
'=a (~)
(10.6.25)
~R0)
^_j
00.6.26)
10.6
()
(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)
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
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
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)
-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)
(10.6.38)
y- 1p
(10.6.39)
10.7
485
*-*[*-
Solution
That is,
(10.6.40)
pk^ki*
(10.6.41)
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
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
VORTICITY EQUATION
fDy\
curl I = curlb
\Dt
10.7
487
EQUATIONS OF EQUILIBRIUM
(10.7.11)
-(V/>) = b
P
(10.7.12)
For b = - , equation
an (10.7.11) yields
P+X=
+x= constant
(10.7.13)
(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
constant
(10.7.18)
(10.7.20)
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
489
-ge 3
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
(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)
Recalling that
^ = - + (v V)
Dt
dt
(v.V)v
= ^(V/>) + b
p
(10.7.25)
490
10
(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)
= (?-&vkik + bkvk
+ 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
(10.7.30)
(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
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)
[F^v]' = - V
x-\v2)dt
(P +
(10.7.34)
^vV =
ay
(10.7.35)
dt2
-Di
= P
dv
, V)v
+ (v
dt
d\
(10.7.37)
(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
$(S) + *<""-
(10 71|
p ^ + A,divv = 0.
- >
(10.7.42)
(10.7.44)
This shows that (divv) also satisfies the wave equation (10.7.35).
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
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)
'-
(10.8.4)
(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
494
10
- + -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)
where C is a constant.
- + \v2 + gz = C
P 2
___^_
(10.8.11)
496
10
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
+ 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)
497
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
(10.8.19)
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
"*=-&"<
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
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
(10.8.23)
(10.8.24)
(10.8.25)
(10.8.26)
10.8
BERNOULLI'S EQUATIONS
499
(10.8.27)
Also,
= y\py~2dp
= + constant
(y - i)
(y - i)
- + constant
P
(10.8.28)
_i\l/y
(10.8.29)
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
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.
501
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)
* + 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
e/ + v
^ . 1 , 2
* . v l - + -(v)'
+ ez
2
/\
l
,
(10.9.4)
502
10
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
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)
503
(10.9.11)
(10.9.12)
(10.9.13)
(10.9.14)
(10.9.15)
(10.9.16)
(10.9.18)
504
10
<l0 9
",
'H'+i^ki
(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)
(10.9.21)
(10.9.22)
(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.
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 - ] = + \[ ()2]2 = = 0
(10.9.24)
(10.9.25)
505
where, for simplicity, we have written for /dt, for /dz, for
32/32, etc.
A systematic procedure can be employed to rewrite these boundary
conditions by using Taylor's series expansions of the potential and its
derivatives in the typical form
(10.9.26)
= , t) = ] = 0 + [] = 0 + \ [] = +
(^,
(,,
= 9t) = [] = 0 + >7[< = + M * J * = o +
(10.9.27)
+ \
= --
+
=
+ ^
, +
=0
-MH,
+ \<?)2\\
+ -2
8
\--*
Jz = o
= 0
+ \
+ 0(3)
= 0
(10.9.28)
+ 8) zzh = o + "*
' V*,JJZ = 0 +
(10.9.29)
, +
+ 2[(V</))-V0 t ] z =
- 1
^)--
g
^)?--,1
o
+ i[V^-V(V^) 2 ] z =
= 0
- -
--
, +
+ **)*]* = o
[( + 8)] =
z=0
[(y<t>
^--{
8
z=o
, +
Z= 0
,2
)-- Z = 0
,)]=0
[{()\]=0
(10.9.30)
506
10
...
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)
507
(10.9.36)
(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
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
509
(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)
510
10
+ Idij
(10.10.5)
+ A(trD)}I + 2
(10.10.6)
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)'
10.10
511
(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)
(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).
Let
= (A + f//)(trD) 2 + 2\)\2
(10.10.12)
51 2
10
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
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)
10.10
513
vx = 4x^2^3,
v2 = *3,
v3 = -2x2xl
(10.10.22)
d22 = 0,
dl2 = 2 ^ * 3 ,
du = 2xxx2,
d33 =
-4x2x3
(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
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
51 5
(10.11.4)
where
AC = + f/i
(10.11.5)
STOKES'S CONDITION
(10.11.6)
divv = 0
(10.11.7)
= 0
(10.11.8)
516
10
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)
10.11
51 7
+ 2//D 0
(10.11.12)
Hence
jid)
= T
_ i.(trT)I
= {_pi
(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.
(10.11.14)
0-2
L i
f
1
518
10
[T ] = [T] + p[l] = 0 - 1
(v)
Li
i.
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
519
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
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)
(10.12.10)
10.12
521
(10.12.12)
De
p = -p(divv) - divq + ph +
(10.12.13)
(10.12.15)
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)
522
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
for = 0
(10.12.4)
(10.12.12)
5. Equations of state:
(10.12.9)
f(p,p,T) = 0
= (,)
(10.12.10)
,
+ kV2T + ph +
(10.12.16)
pcv = -j + kV2T+
Dp
,
DT
pcp = -j^t + kV2T+
Solution
ph +
ph +
(10.12.17)
(10.12.18)
(10.12.19a)
p = pRT
(10.12.19b)
10.13
523
so that
DT
DT
cv = cpp
Dt
Dt
I Dp
p Dp
- ^ + 42 - ^
p Dt
p Dt
(10.12.20)
'
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
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)
(10.13.2b)
(10.13.4)
(10.13.5)
(10.13.6a)
(10.13.6b)
(incompressible case);
10.13
525
(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)
(10.13.9)
(10.13.10)
(10.13.11)
(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
(10.13.13)
(10.13.15)
(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.
vdx =
js
wndS
(10.13.17)
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)
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)
(10.14.3)
(10.14.4)
+ (v V)v
2 + (A + //)V(divv) - Vp + pb = \
+ w X v + - Vi;2 j (10.14.5b)
(10.14.5a)
(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
REYNOLDS NUMBER
Q(Y\
Vv /
10.14
NAVIER-STOKES EQUATION
529
v3 = 0,
v2 = 2kxlx2,
(k = constant)
(10.14.10)
(
\
+
\(
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
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)
\k2p(x\ + x2)2
(10.14.16)
v x w = v ( - + -v2 + ) + vcurlw
(10.14.17)
(ii)
(10.14.18)
(iii) If
v = ^
- ^2
(10.14.19)
(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)
(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
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)
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
Pa = -pW2(,i)
(10.14.31)
p t 2 1 = -2(1)
(10.14.32)
or
VV = 0
(10.14.33)
+ ?- +
p
(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).
(10.14.39)
10.14
NAVIER-STOKES EQUATION
533
where
H* = - + - v2 + +
P 2
(10.14.40)
V2(- + + X- v2 J = div(v x w)
V 2 ( - + ) = \yf2 - D D
(10.14.41)
(10.14.42)
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
= (w . V)v + vV2w
(10.14.44)
(10.14.45)
(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)
(10.14.48)
= w 2 = o,
w3 = 2 - vU2
(10.14.49)
Consequently,
( w V)v = 0
(10.14.50)
10.14
535
( X 1 #X 2 )
"' = -R X>'
RXl'
"3 =
(10.14.51)
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)
dv
R dR
(10.14.53)
/ = 1,2
(10.14.54)
536
10
Consequently,
Dw
dw
~Dt=~dt
(10.14.55)
d2w
1 dw
= v 'dR2 + R^R
(10.14.47)'
R2\
Avt
(10.14.57)
= \
n-(vxw)rfS
(10.14.58)
W = \
(10.14.59)
dV
p(y\)dV
1 f
DY
(10.14.60)
10.14
537
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)
w z dV - \
div(v Xw)rfF
(10.14.64)
(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)
538
10
Dt
^ * dx
) c Dt
(10.14.68)
= v(t v V t f x
(10.14.69)
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
10.15
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)
(10.15.3)
dxl
dp_
dxx
(10.15.4)
540
10
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)
(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)
(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
541
(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,
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)
pv0h
2
(10.15.16)
542
10
=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
(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)
(10.15.23)
10.15
543
,,
p (x,)<0
Vo=0
(10.15.24)
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.
544
10
(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)
10.15
545
(10.15.35)
'( 3 )<0
*3
546
10
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
"2=JR>
"3
(10.15.39)
s1 = -p^
s2 = -p^
s3 = -^R
(10.15.40b)
(10.15.40c)
10.15
547
(^
$~1)
(10.15.42)
Thus,
v=
(10 15 43)
= -^j
a2b2G
2( + bz)
a2
(10.15.44)
b2
M=
)r
dA=
pa2b2G
[ / dA x2
y1
2^v^)X-7-^
( 5 45)
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
ab
cos0,
a
AI2 = sin0,
n3 = 0
(10.15.49)
where
N = (b2 cos 2 6> + a2 sin2 0) 1/2
(10.15.50)
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)
10.15
549
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)
(10.15.56a)
for i, k = 1,2;
v3 = 0
(10.15.56b)
550
10
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
551
dRe"
(10.15.59)
*<*
= 0
(10.15.60)
(10.15.61)
(10.15.63)
552
10
2b2 - 2
tf-a
( - 2) 2u2
b
B =
'
(10.15.64)
b2
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
'
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)
for ij
=1,2;
(10.15.68)
32 = 0,
^33
10.15
553
(10.15.69)
(10.15.70)
(10.15.71)
= 4
\2 - \
(10.15.72)
554
10
*3
3> 0
Xi=0
Dv
Dt
P,2 = 0
(10.15.73)
/?,3 = 0
(10.15.74)
(10.15.75)
dv
,
^
dv
- + v (v V)u
=
}
dt
dt
(10.15.76)
dv _ 1 dp
dt
p dxx
(10.15.77)
(10.15.78)
f(t)
(10.15.79)
10.15
555
(10.15.80)
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)
(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
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)
- 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
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
dv dr\
3 dx3
V ,
2\M
dv
dv
Vvx3
dt~ 3 dt ~
4(v0 3/2
d2v
dx3
V M/ ,
4vt
e~v + B
(10.15.96)
for
/7 = 0,
F(A7)-Oas/7- .
(10.15.97a,b)
exp(-/7V/7 = ^
(10.15.98)
'
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
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)
(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
559
for r = a,
(10.15.104)
p = -4*i
(10.15.105)
=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)
B = -3A,
C = D' = Aa2
560
10
(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.
10.15
561
( * + 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)
(10.15.114)
^<
r
3
r
(10.15.115)
562
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).
22
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
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
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.
. wdV =
(w n)vdS
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
= 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
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,
sriS=
]v
(divT(v) - Vp)dV =
]v
]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
Dt
vV 7
#
4
+
at
3
.
-2=/()
-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
^-(x-a)(3y2-x2)
v=
Ga4
60V3v
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
Section 1.8
1.
+ <723^2 + #3363
+ #32^23 + #33^33
+ bji)
569
570
CONTINUUM MECHANICS
-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 -
613 >
o3=.bn - b2l
-skjibj
2 pqr ejkr
l( pq
* 'ijm Omk
. = ijk
jk
= \(pjqk a
' ek
qp)
^^^
pq
n^rij
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
a'2 =
1 -V3
2V2 '
1 + V3
2V2
, = - -
K/] =
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
b).
572
49.
CONTINUUM MECHANICS
(i) Use (2.10.4) to get
<*ikbkj = eikmumekjnvn
(ii) aikbki =
a.
ij^mnt^m
= (inmj
"n = UjVi -
Suumvm
-2umvm
em)
= km(ak em) = e* ek = I
55.
56.
5 7 . For any A,
det(BAB _1 - AI) = det{B(A - AI)B _1 )
= (detB)(det(A - AIJKdetB"1) = det(A
AI)
// A = - l ,
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
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
Xi
/,=/'(),=/'()-
18. (i)
l,kel
= ~Wl,2e3 +
l,3 e 2>
21.
22.
24.
26.
27.
30.
32.
35.
40.
20.
etC
2|| = (rn)M =
+ UiVjj
(^"1-)
UjjVi
UjVu
= n(n + l)/""2
574
CONTINUUM MECHANICS
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
[Q]
= V2
-1
575
isochoric if = 1/2
0 V2
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)
576
CONTINUUM MECHANICS
2a2(*i -x3)
2a 2 (x 2 + x3)
^ f 2a(x2 + x3) - *, )
39. (iii) []
2 f2a(*2
* s ) " Xl '
-\2(3-)
+ 2}
-\2a(x2+x3)
+ x x)
[0] =
(ii) ^ ( H 9 a + 40)
u2 = 2xlx2 + x2,
u3 = x2
4 4 . (i) a = 2
Section 6 . 7
1.
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 '
cos
Dv2
Dt
Dv3
= 0
Dt
<*f + xi sin 0 ^3 = 0
Zty
= 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
577
(*!
+ 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 '
d$ = ^(5x 2 + 7xi)e'at;
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.
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,
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;
= 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 - | ,
33.
A
u2 = -xli
\0
13 ~ 2 3 -
33 -
-2-\
+ x 2 e 2 + (r + Ar3)e3].
u3 = 0
-\>
33 =
\0
\2,
12 = (*-),
580
CONTINUUM MECHANICS
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] =
( 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]
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
582
CONTINUUM MECHANICS
BIBLIOGRAPHY
583
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
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
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
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
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