MAE 256 A&B Course Reader: Professor Ajit Mal
MAE 256 A&B Course Reader: Professor Ajit Mal
2005
c by Ajit Mal
Contents
I Linear Elasticity 7
1 Cartesian Tensors 9
1.1 Review of Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1.1 Definitions and Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1.2 Rules of Matrix Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.1.3 Eigenvalues and Eigenvectors of a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.1.4 The Positive-Definite matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1.5 Polar Decomposition of a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 ALGEBRAIC DEFINITION OF A VECTOR; THE TRANSFORMATION MATRIX . . . . . . . . . 15
1.2.1 The Vector Transformation Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3 THE INDEX NOTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.1 The Kronecker Delta and the Alternating Symbol . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4 GENERAL DEFINITION AND ALGEBRA OF CARTESIAN TENSORS . . . . . . . . . . . . . . 22
1.4.1 Inner and Outer Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4.2 The Quotient Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5 SOME SPECIAL PROPERTIES OF SECOND-ORDER TENSORS . . . . . . . . . . . . . . . . . . 25
1.6 PRINCIPAL AXES OF A REAL SYMMETRIC SECOND-ORDER TENSOR . . . . . . . . . . . . 28
1.7 ISOTROPIC TENSORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.8 VECTOR AND TENSOR CALCULUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.9 PHYSICAL INTERPRETATION OF SOME VECTOR DIFFERENTIAL OPERATIONS . . . . . . . 37
1.10 DYADICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.11 SOME VECTOR IDENTITIES IN CYLINDRICAL AND SPHERICAL COORDINATES . . . . . . 40
1.11.1 Cylindrical Coordinates (r, φ, z) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.11.2 Spherical coordinates (R, θ, φ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3
4 CONTENTS
2.1 STRAIN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 STRESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3 EQUATIONS OF EQUILIBRIUM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.4 HOOKE’S LAW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.5 THERMAL EFFECTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3 Linear Elastostatics 65
3.1 FIELD EQUATIONS OF LINEAR ELASTICITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2 NAVIER’S EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3 UNIQUENESS OF SOLUTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.1 Clapeyron’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4 BETTI-RAYLEIGH RECIPROCITY (B-R-R) RELATIONS . . . . . . . . . . . . . . . . . . . . . . 71
3.4.1 Applications of the B-R-R Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5 EXACT SOLUTION OF SOME LINEAR ELASTIC
PROBLEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.6 EQUATIONS OF COMPATIBILITY FOR INFINITESIMAL STRAINS . . . . . . . . . . . . . . . . 82
3.7 Anisotropic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.7.1 Medium with one plane of symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.7.2 Orthotropic media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.7.3 Transversely isotropic media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.7.4 Cubic Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.7.5 Isotropic media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.8 CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES . . . . . . . . . . . . . 89
2005
c by Ajit Mal
CONTENTS 5
2005
c by Ajit Mal
6 CONTENTS
2005
c by Ajit Mal
Part I
Linear Elasticity
7
Chapter 1
Cartesian Tensors
In this section we present certain results of matrix algebra that are useful for the development of elasticity theory. The
treatment is quite terse, more detailed discussion can be found in the reference cited at the end of the appendix.
An aggregate of numbers, in general complex, arranged in the form of an array containing m rows and n columns in
the form
a11 a12 . . a1n
a21 a22 . . a2n
. . . . . (1.1)
. . . . .
am1 am2 . . amn
is called an (m × n) matrix. The matrix will be denoted by A or a (if there is no cause for confusion) and also by
the notations [aij ] where i and j are understood to take on the values 1, 2, . . . , m and 1, 2, . . . , n, respectively. The
numbers a11 , a12 , . . . are called the elements of the matrix A. We shall be concerned with real matrices for which
the elements aij are all real. If m 6= n, the matrix is called rectangular, whereas if m = n, it is a square matrix of
order n. A matrix with all zero elements is called a null matrix and is denoted by 0.
The transpose of A is obtained by interchanging the rows and columns of A and denoted by AT . Thus AT is a
(n × m) matrix given by
a11 a21 . . am1
a12 a22 . . am2
. . . . . (1.2)
. . . . .
a1n a2n . . amn
A matrix containing only one row is called a row vector. Thus, [v1 , v2 , ...vn ] is a row vector, and it is denoted by
[vi ]. A matrix containing only one column is called a column vector. A column vector has the following alternative
representations:
[V ] = {vi } = [v1 , v2 . . . vn ]T (1.3)
Square matrices play an important role in the mathematical treatment of elasticity and the following definitions
apply to a square matrix of order n.
9
10 CHAPTER 1. CARTESIAN TENSORS
1. The determinant
a11 a12 . . a1n
a21 a22 . . a2n
. . . . . (1.4)
. . . . .
an1 an2 . . ann
is called the determinant of A and is denoted by |A| or det{aij }.
2. The sum of the diagonal elements is called the trace of A and is denoted by trA or tr[aij ], trA = a11 + a22 +
. . . + ann
3. A is symmetric if aij = aji and is skew symmetric or antisymmetric if aij = −aji . Clearly, for a symmetric
matrix A, AT = A, and for an antisymmetric matrix, AT = −A. Any matrix A can be decomposed as the sum
of a symmetric and an antisymmetric matrix:
1 1
A + AT + A − AT
A=
2 2
4. If the nonzero elements of a matrix lie on the main diagonal only, then it is called a diagonal matrix. A diagonal
matrix D has the general form
d1 0 . . 0
0 d2 . . 0
D= . . . . .
. . . . .
0 0 . . dn
and is often denoted by diag [d1 , d2 , . . . , dn ]. The diagonal matrix [1, 1, . . . , 1] is called a unit matrix and is
denoted by I. Then
I = [δij ] (1.5)
where δij is the Kronecker symbol defined through
δij = 1, i =j
= 0, i 6= j (1.6)
Clearly, |I| = 1.
Two matrices A and B are equal if they have the same dimensions (m × n) and their corresponding elements are
equal. Thus,
A = B ⇔ aij ; i = 1, 2, . . . , m; j = 1, 2, . . . , n
The elements of the sum A + B are aij + bij ; similar notation holds for their difference. The product of a matrix
A with a scalar α is the matrix
αA = [αaij ]
The product AB of two matrices A (m × n) and B (n × p) is defined only under the restriction that the number of
columns (n) of A is equal to the number of rows of B. Then
AB = C (1.7)
2005
c by Ajit Mal
1.1. REVIEW OF MATRICES 11
In general, the commutative law does not hold for matrix products; i.e., AB 6= BA. Furthermore,
T
(AB) = B T AT (1.9)
It can be easily seen that that for a square (n × n) matrix A and the (n × n) unit matrix I,
AI = IA = A (1.10)
and mij is the value of the determinant obtained by striking out the row and column containing aij in |A|. It should be
pointed out that in prectice the numerical evaluation of C is, in general, extremely laborious, and other more efficient
computational techniques are available for the numerical computation of A−1 . These can be found in modern texts on
numerical methods.
A real square matrix A is called orthogonal if the sum of the squares of the elemnts of each column is equal to 1
and the sum of the products of the corresponding elements of two different columns is zero. In other words, if A is
orthogonal, then
a1i a1j + a2i a2j + · · · + ani anj = δij , (i, j = 1, 2, · · · , n) (1.15)
Equation (1.15) implies that AT A = I, so that
A−1 = AT (1.16)
T T th (i)
It can be shown that A A = I ⇔ AA = I. If the i column o A is denoted by the column vector A , then
equation (1.15) may be expressed in the form
If |A| = 1, then A is called proper orthogonal; whereas if |A| = −1, it is called improper orthogonal.
Let A = [aij ] be a square matrix, V = {vi } be a column vector of dimension n, and λ be a scalar. Consider the matrix
equation
AV − λV = 0 or [A − λI] V = 0 (1.19)
2005
c by Ajit Mal
12 CHAPTER 1. CARTESIAN TENSORS
Equation (1.19) represents the following system of n linear homogeneous algebraic equations in the components vi
of V:
(a11 − λ) v1 + a12 v2 + · · · + a1n vn = 0
a21 v1 + (a22 − λ) v2 + · · · + a2n vn = 0
···································· (1.20)
an1 v1 + an2 v2 + · · · + (ann − λ) vn = 0
2005
c by Ajit Mal
1.1. REVIEW OF MATRICES 13
or
AQ = QΛ
A = QΛQ−1 (1.27)
Relations of this type are called similarity transf ormations. It should be noted that the eigenvalues of a real matrix
can, in general, be complex (in conjugate pairs) , so that the modal matrixcan also be complex-valued. Furthermore,
some of the eigenvalues can be equal; consequently, the number of linearly independent eigenvectors may be less than
n. In the mathematical treatment of elasticity, the matrices that arise are often real and symmetric. The eigenvalues and
eigenvectors of such matrices have certain properties that are of special interest in elasticity theory. These properties
are as follows:
A = QΛQT (1.29)
Λ = QT AQ (1.30)
Properties (a)-(c) are proved in Chapter 2 for the case n = 3 and property (d) follows from the definition of orthogo-
nality (1.15, 1.17) and (1.28).
The similarity transformations can be used to define and calculate certain f unctions f (A) of the matrix A. The
general formula is
f (A) = Qf (Λ) Q−1 (1.31)
which reduces to
f (A) = Qf (Λ) QT (1.32)
if A is symmetric. Thus, if we can define a function of the diagonal matrix Λ, we can define and calculate the
corresponding function of the matrix A. As an example of the calculation of the function of a matrix, consider the
function A1/q where q is a positive integer, and A is a real symmetric matrix whose eigenvalues (and, therefore, Λ)
and the eigenvectors (and, therefore, Q) are known. Then (1.32) gives
In analogy with scalar algebra, Λ1/q is defined to be another matrix whose q th power is Λ. Clearly,
h i
1/q 1/q
Λ1/q = diag λ1 , λ2 , . . . , λ1/q
n (1.34)
and A1/q can be calculated from equation (1.33). It should be noted that there are q values for each element of Λ1/q ,
so that there are q n values of Λ1/q and hence of A1/q . Other functions of A are defined in a similar manner.
Example Problem 1,2 (Lecture 1 Page 26,31)
Caley-Hamilton Theorem. Every sqaure matrix satisf ies it′ s own eigenvalue equation; i.e., if D(λ) =
|A − λI| = 0, then D(A) = 0. This theorem is quite useful in the development of the constitutive equations of
elasticity. The proof can be found in standard texts on matrix theory listed at the end of the appendix and will be
omitted.
2005
c by Ajit Mal
14 CHAPTER 1. CARTESIAN TENSORS
Let C be a matrix and X = {xi } be an arbitrary n-dimensional nonzero column vector. Then C is called positive-
def inite if X T CX > 0, for all X.
Consider the matrix C defined through
C = AT A (1.35)
where A is a real nonsingular matrix. Then C is nonsingular and symmetric. The proof is given below.
Let X = {xi } be an arbitrary n-dimensional nonzero column vector, and assume that
f (x1 , x2 , . . . , xn ) = X T CX (1.36)
CV = λV (1.38)
V T CV
λ= (1.39)
V TV
Since C is positive-definite and V is nonzero by definition, the right-hand side of (1.39) is always positive. Thus
all the eigenvalues of C are positive. Furthermore, since C is a real symmetric matrix, it ha n linearly independent
eigenvectors, and its modal matrix Q is nonsingular and orthogonal. Thus,
where h i
1/2 1/2
Λ1/2 = diag λ1 , λ2 , . . . , λ1/2
n (1.41)
is a real symmetric matrix and can be uniquely defined by requiring that the square roots in (1.41) are positive. It is
easy to verify that U is nonsingular and positive-definite. Now, consider the matrix R defined through
A = RU
equation Then
1/2 −1/2
U = AT A , R = AU −1 = A AT A (1.43)
and T −1 T
RRT = AU −1 = AU −2 AT = A AT A
AU −1 A =I
Hence R is orthogonal. Further, −1/2
|R| = A AT A = +1
So R is proper. Equation (1.42) implies that the nonsingular matrix A can be decomposed into the product of the
proper orthogonal matrix R and the positive-definite symmetric U . The decomposition (1.42) is unique, i.e., R and U
are uniquely related to A, and is called the lef t decomposition of A.
2005
c by Ajit Mal
1.2. ALGEBRAIC DEFINITION OF A VECTOR; THE TRANSFORMATION MATRIX 15
Many physical quantities e.g., mass, volume, and temperature of a solid body can be specified by using their numerical
values in appropriate units. These are called scalars. Others e.g., displacement, velocity, and linear acceleration of a
moving particle or the force acting on an object require a magnitude as well as a direction for their description. These
quantities are called vectors. A vector by its very nature can be represented by a directed line segment or an arrow in
space whose length is proportional to the magnitude of the vector and whose direction and sense coincide with those
of the vector. This is the geometrical representation of a vector, and together with the geometrical rules of operation
between vectors (e.g., the parallelogram law of addition, the cosine law for the dot product, and the sine law for the
cross product) form the basis of the vectorial treatment of elementary mechanics. In more advanced treatments it is
very convenient to give an algebraic definition of a vector and to develop algebraic rules of operation between vectors.
If we associate the symbol x (in direct notation) with a given directed line segment, then the definition of a vector
implies that any other directed line segment of the same length and sense is also x. For example, two forces with the
same magnitude and direction but acting at two different points of a rigid body are represented by the same vector,
although the forces may have different influence on the equilibrium or motion of the body.
We introduce a right-handed Cartesian coordinate system with axes numbered as 1, 2, and 3 at an arbitrarily
⇀
chosen origin O (Fig. 1-1) and assume that the directed line segment OP is the vector x. Let x1 , x2 , and x3 denote
⇀
the projections of OP on the coordinate axes numbered 1, 2 and 3, respectively, or alternatively, the coordinates of
⇀
P relative to the chosen axes. Then, for a given vector OP , the numbers x1 , x2 , x3 can be uniquely determined.
⇀
Conversely, if the numbers x1 , x2 , x3 are given, then the vector OP can be uniquely found. In other words, with a
⇀
given choice of the coordinate system, the ordered triplet x1 , x2 , x3 is completely equivalent to the vector OP . The
order in which the numbers are given is important, since, in general, the triplet x1 , x3 , x2 is a vector different from
⇀
OP . The triplet x1 , x2 , x3 will be denoted by [xi ] when written as a row matrix and by {xi } as a column matrix.
Let e1 , e2 , e3 be three vectors of unit magnitude directed along the coordinate axes (Fig. 1-1). In analogy with the
notations introduced in the appendix, the array of these vectors will be denoted by [ei ] as a row and {ei } as a column.
They define the reference coordinate system for describing all vectors that lie in a three-dimensional Euclidean space
and are called a vector basis; the unit vectors e1 , e2 , e3 are called the base vectors. Then the vector x can be uniquely
expressed in terms of the base vectors and its components x1 , x2 , x3 along the base vectors in the form
3
X
x = x1 e1 + x2 e2 + x3 e3 = xi ei (1.47)
i=1
An obvious consequence of the representation (1.47) is that the sum, difference, and (dot or cross) products between
two or more vectors can be expressed in terms of their components. Thus, if u and v are two vectors with components
u1 , u2 , u3 and v1 , v2 , v3 , respectively, in the same vector basis {ei }, then
3
X
u±v = (ui ± vi )ei (1.48)
i=1
2005
c by Ajit Mal
16 CHAPTER 1. CARTESIAN TENSORS
x 3
P (x1 ; x2 ; x3)
e3
e2 x2
e1
x 1
Fig. 1-1. The coordinate system.
3
X
u·v = ui v i (1.49)
i=1
u×v = (u2 v3 − u3 v2 )e1 + (u3 v1 − u1 v3 )e2 + (u1 v2 − u2 v1 )e3
e1 e2 e3
= u1 u2 u3 (1.50)
v 1 u2 v 3
i, j, = 1, 2, 3
and
e2 × e3 = −e3 × e2 = e1
e3 × e1 = −e1 × e3 = e2 ( 1.51b)
e1 × e2 = −e2 × e1 = e3
In (1.51a), δij is the Kronecker delta defined via
1, if i = j
δij = (1.52)
0, if i 6= j
The relations (1.51) and (1.52) can be easily verified from the geometrical definition of the dot and cross products, the
orthogonality, and the right-handedness of the vector basis {ei }.
We first note that if u = v = x in (1.49) we obtain the well-known formula for the ”length”, or magnitude, |x| of the
vector x,
|x|2 = x21 + x22 + x23 (1.53)
2005
c by Ajit Mal
1.2. ALGEBRAIC DEFINITION OF A VECTOR; THE TRANSFORMATION MATRIX 17
e3
e03 P
x
e02
o e2
e1
e01
Fig. 1-2. Rotation of axes.
⇀
It should be noted that the vector OP itself is not dependent upon the choice of the vector basis and that the triplet
⇀
{xi } is simply the algebraic representation of OP in the chosen vector basis {ei }. If we choose another vector basis
′ ⇀ ′
{ei } at the common origin O (Fig. 1-2), then the same vector OP is represented by a different triplet {xi } in the new
vector basis. Thus,
3 3
X ′ ′ X
x= xi ei = xi ei (1.54)
i=1 i=1
′ ′ ′
Taking the dot product of (1.54) successively with e1 , e2 , e3 and using the orthogonality property of the vector
′ ′ ′
basis {ei }(i.e., ei · ej = δij ), the following results are obtained;
′
x1 = ℓ11 x1 + ℓ12 x2 + ℓ13 x3
′
x2 = ℓ21 x1 + ℓ22 x2 + ℓ23 x3 (1.55)
′
x3 = ℓ31 x1 + ℓ32 x2 + ℓ33 x3
where ′
ℓij = ei · ej , i, j, = 1, 2, 3 (1.56)
Similarly, by taking the dot product of (1.54) successively with e1 , e2 , e3 , we obtain
′ ′ ′
x1 = ℓ11 x1 + ℓ21 x2 + ℓ31 x3
′ ′ ′
x2 = ℓ12 x1 + ℓ22 x2 + ℓ32 x3 (1.57)
′ ′ ′
x3 = ℓ13 x1 + ℓ23 x2 + ℓ33 x3
and ′ ′
{xi } = LT {xi } or X = LT X ( 1.58b)
where
ℓ11 ℓ12 ℓ13
L = [ℓij ] = ℓ21 ℓ22 ℓ23 (1.59)
ℓ31 ℓ32 ℓ33
2005
c by Ajit Mal
18 CHAPTER 1. CARTESIAN TENSORS
Equations (1.55) and (1.57)—or, alternatively, the matrix equations (1.58) give the transformation properties of the
′
components of x in the two vector bases {ei } and {ei } through the linear transformation matrix L. It can be easily
seen from (1.55) and (1.57) that the column vectors of L are the direction cosines of the base vectors e1 , e2 , and e3
′ ′ ′ ′
relative to the vector basis {ei } and the row vectors of L are the direction cosines of e1 , e2 , and e3 relative to the
vector basis {ei }. For this reason L is also called the direction cosine matrix of the vector basis transformation from
′
{ei } to {ei }. Since the unit vectors e1 , e2 , and e3 are mutually orthogonal, the sum of the squares of the elements of
each column (or row) of L is unity and the sum of the products of the corresponding elements of any two columns (or
rows) is zero. Thus L is an orthogonal matrix [see the definition in (1.15)] and, therefore,
L−1 = LT (1.60)
in agreement with (1.58) and
|L|−1 = 1 (1.61)
Example Problem (Lecture 1 Page 48) ′
Clearly, ℓ11 , ℓ12 , ℓ13 are the components of the unit vector e1 in the vector basis {ei } and similarly for ℓ21 , ℓ22 , ℓ23
and ℓ31 , ℓ32 , ℓ33 . These results can be expressed in matrix form as
′
{ei } = L{ei } (1.62)
Then, from (1.50),
e1 e2 e3
′ ′
e2 × e3 = ℓ21 ℓ22 ℓ23
ℓ31 ℓ32 ℓ33
and ′ ′ ′
e1 · (e2 × e3 ) = |L| (1.63)
′
But, since the vector basis {ei } is right-handed,
′ ′ ′
e2 × e3 = e1
|L| = +1
′ ′ ′
then the same vector x is represented by the array {x1 , x2 , x3 }, where
′ ′
{xi } = L{xi } or X = LX (1.66)
′
It should be noted that the preceding definition also holds for a vector in n dimensions. In that case {xi } and {xi } are
′
(n × 1) arrays, each of the vector bases {ei } and {ei } has n mutually orthogonal base vectors, and L is a matrix with
the stated properties. We shall be concerned with two- and three-dimensional vectors only.
Another major advantage of the algebraic definition of a vector is that it can be used to define and manipulate
other physical entities calledtensors, which cannot be represented by directed line segments. Before we introduce
the concept of tensors, it is necessary to develop certain notations and conventions for ease and compactness in the
manipulation of tensors.
2005
c by Ajit Mal
1.3. THE INDEX NOTATION 19
In the mathematical formulation of the theory of elasticity, it is often necessary to work with differential or algebraic
expressions and equations containing a large number of similar terms. It is possible to reduce them to more convenient
compact forms through the introduction of two simple conventions. The first of these is the range convention, which
stipulates that every subscript (or index) takes on the values 1, 2, 3 unless otherwise stated. Thus, the symbol xi stands
for the collection of the three numbers x1 , x2 , x3 and the symbol aij stands for the collection of the nine numbers
a11 , a12 , ..., a33 . Similarly, the equation xi = yi represents three separate equations corresponding to i = 1, 2, 3.
The indices i and j in these examples are called free indices. The same free indices must appear on both sides of an
equation.
A second simplification is achieved by means of the summation convention. According to this convention, if an
index i appears twice in a given expression, a summation over i must be performed by assigning the values 1, 2, 3
to i successively. Any exception to this rule must be stated explicitly. A repeated index is called a dummy index. A
dummy index cannot occur more than twice in an expression.
It can be seen that many of the equations that appear in the previous section can be written in compact forms
through the use of these conventions. For example, (1.54), (1.55), and (1.57) may be rewritten as
′ ′
x = xi ei = xi ei
′
xi = ℓij xj
′
xj = ℓij xi
As another example, a homogeneous quadratic function f (x1 , x2 , x3 ) may be written as aij xi xj , where [aij ] is a
constant matrix. Thus, f (x1 , x2 , x3 ) has the alternative expressions
f (x1 , x2 .x3 ) = aij xi xj
= a11 x21 + a22 x22 + a33 x23 + (a23 + a32 )x2 x3 (1.67)
+ (a31 + a13 )x1 x3 + (a12 + a21 )x1 x2
The Kronecker delta, δij , introduced in (1.51) plays an important role in the manipulation of equations written in index
notation. A particularly useful property of δij is illustrated by the results
xi δij = xj , aik δij = ajk (1.68)
Thus, the presence of the dummy index i and δij in (1.68) results in the replacement of i by j in xi and aik . This is
known as the substitution property of the Kronecker delta.
The orthogonality of the vector basis {ei } is expressed by the equation
ei · ej = δij (1.69)
and the dot product between any two vectors u and v is given by
u · v = ui v i (1.70)
On the other hand, the right-handedness of the vector basis {ei } is reflected in the cross product rule between the base
vectors given in (1.52) and between any two vectors given in (1.50). In order to express these results in index notation,
it is necessary to introduce the alternating (or permutation or Levi-Civita) symbol ǫijk , defined via
Thus,
1 if ijk are even permutations of 123, i.e., are incyclic order
ǫijk = −1 if ijk are odd permutations of 123, i.e., are inanticyclic order
0 otherwise, i.e., if two or three indices are equal.
2005
c by Ajit Mal
20 CHAPTER 1. CARTESIAN TENSORS
ei × ej = ǫijk ek ( 1.72a)
(ei × ej ) · ek = ǫijk ( 1.72b)
u×v = ǫijk ui vj ek ( 1.72c)
It should be noted that equations written in the index notation as introduced here are valid in Cartesian coordinates
only, in contrast to those written in the direct vector notation, which are independent of the coordinate system.
Example 1.3-1
show that
Solution. To prove (a), we note that the quantity ǫijk′ ǫpqk′ (where a primed index implies that no sum is to be carried
′ ′
out over that index) is nonzero if and only if i, j, k are all different and p, q, k are all different. Therefore, either
i = p and j = q or i = q and j = p. In the first case, ǫijk′ = ǫpqk′ and, therefore, ǫijk′ ǫpqk′ = 1 . In the second
case, ǫijk′ = −ǫpqk′ , because the interchange of two indices changes the sign of the value of ǫijk′ . Consequently,
ǫijk′ ǫpqk′ = −1. Now, if i, j, p, and q are fixed and k runs through the values 1, 2, 3, then, as explained before, only
one term in the sum is nonzero. Hence, the whole sum is either +1 or -1. This can be written as δip δjq − δiq δjp because
this expression equals +1 for i = p, j = q, i 6= j and equals -1 for i = q, j = p, i 6= j; if neither is the case, it equals
0. Putting q = j in (a), we get
Example 1.3-2
Show that
δip δiq δir
ǫijk ǫpqr = δjp δjq δjr
(i)
δkp δkq δkr
2005
c by Ajit Mal
1.3. THE INDEX NOTATION 21
Therefore,
δi1 δi2 δi3
δp1 δp2 δp3
ǫijk ǫpqr =
δj1 δj2 δj3
δq1 δq2 δq3
δk1 δk2 δk3 δr1 δr2 δr3
δi1 δi2 δi3
δp1 δq1 δr1
=
δj1 δj2 δj3
δp2 δq2 δr2
δk1 δk2 δk3 δp3 δq3 δr3
The last step follows immediately on noting that, for example, the (1, 1) element of the product determinant will be
Putting r = k in (i) and expanding the determinant, we get Eq.(i) of Example 1.3-1.
Example 1.3-3
Show that
(a) det [aij ] = ǫijk ai1 aj2 ak3 = ǫijk a1i a2j a3k
= 61 ǫijk ǫlmn ail ajm akn
(b) ǫlmn det [aij ] = ǫijk ail ajm akn
Solution.
(a) Note that for three vectors u, v, w,
u1 v1 w1
u2 v2 w2 = (u × v) · w = ǫijk ui vj wk (i)
u3 v3 w3
In (i) let ui = ai1 , vj = aj2, wk = ak3 , proving the first result in (a). Similarly, letting ui = a1i , vj = a2j , wk = a3k
gives the second result in (a). To prove the third result, we note that
ǫ123 ǫlmn a1l a2m a3n + ǫ213 ǫlmn a2l a1m a3n = ǫlmn a1l a2m a3n − ǫmln a1l a2m a3n = 2 det [aaij ]
from the second result in (a). Writing two more relations of this type and adding all three, we get the desired result.
(b) There are three possibilities: l, m, n a cyclic permutation of 1 2 3, l, m, n an anticyclic permutation of 1 2 3 and at
2005
c by Ajit Mal
22 CHAPTER 1. CARTESIAN TENSORS
since if any two columns of a determinant are identical, its value is zero and ǫlmn = 0 if two of l, m, n are equal. This
proves the result.
Cartesian tensors are defined through a generalization of the algebraic definition of a vector discussed in Sec. 2.1. Let
a physical entity be defined in a three-dimensional Euclidean space such that for a given choice of a vector basis {ei }
it is possible to specify the entity completely in terms of an ordered set of 3n numbers denoted by tijk ... containing
a total of n indices. The collection of these numbers will be denoted by (the boldface capital) T or by t if there is no
confusion with the direct notation for a vector, and the numbers tijk ... will be called the components of T. Let a new
′
vector basis {ei } be introduced through a proper rotation of {ei } described by means of the direction cosine matrix
′
L = [ℓij ]. If the components of T in the new vector basis become tijk... , where
′
tijk... = ℓip ℓjq ℓkr · · · tpqr... (1.73)
It can be easily seen that the multiplication of a tensor T by a scalar α produces another tensor αT of the same order
with components αtij... , since
′
αtij... = ℓik ℓjm · · · αtkm...
The sum (or difference) of S and T is another tensor denoted by S + T (or S + T) with components sij... + tij... (or
sij... − tij... ), since from (1.74)
′ ′
sij... ± tij... = ℓik ℓjm · · · (skm... ± tkm... )
2005
c by Ajit Mal
1.4. GENERAL DEFINITION AND ALGEBRA OF CARTESIAN TENSORS 23
The two tensors S and T are called equal if their difference is the null tensor, i.e., if
sij... − tij... = 0
The multiplication between two (or more) tensors can be classified into two categories called outer and inner products.
In order to illustrate the construction and the properties of these products, we assume that S and T are tensors of second
order and U and V are vectors (or tensors of first order). Then, the components of U and V in the two vector bases
′
{ei } and {ei } are related by the equations
′ ′
ui = ℓik uk , vi = ℓik vk (1.75)
aij = ui vj (1.77)
We would like to examine whether the collection of these numbers is a tensor. Since ui and vi are the components of
U and V in the vector basis {ei }, the numerical values as well as the ordering of the elements of the array {aij } are
clearly dependent on the choice of the vector basis. Furthermore, by changing the free index in the second equation of
(1.75) from i to j and the dummy index from k to m, it can be written in the form
′
vj = ℓjm vm
so that ′ ′
ui vj = ℓik ℓjm uk vm (1.78)
′ ′ ′
Hence, if we use the symbol aij for the transformed quantities ui vj , then (1.78) becomes
′
aij = ℓik ℓjm akm
which is the transformation law of a second order tensor A. Thus, the product ui vj defines a second-order tensor. It
can be shown in a similar manner that the set of 27 numbers bijk defined by the product
represents a third-order tensor, and the set of 81 numbers defined by the product
represents a tensor of the fourth order. The products defined in (1.77), (1.79), and (1.80) are called outer products
between the tensors involved, and they can be extended to higher-order tensors. An outer product between two tensors
always generates a tensor whose order is the sum of the orders of the constituent tensors.
Inner products between two tensors are of the following types:
These products are characterized by the presence of one or more dummy suffixes. It can be easily shown that α is a
tensor of order 0 or a scalar, pi is a component of a tensor of order 1 or a vector, and qik is a component of a tensor of
order 2. Since the total number of indices in the left-hand sides of (1.81) is reduced from that of the right-hand sides,
2005
c by Ajit Mal
24 CHAPTER 1. CARTESIAN TENSORS
the order of the resulting tensors is always less in inner products than in the outer products.
It can be seen that the inner products defined in (1.81) are related to the outer products (1.77), (1.79), and (1.80)
through the equations
α = aii , pi = bijj , qik = cijkj (1.82)
The quantities in the right-hand sides of (1.82) are obtained by putting two of the indices in the components of the
corresponding tensors equal and then summing over this index according to the summation convention. This operation
is called a contraction, and it results in the reduction of the order of the tensor. Thus the scalar α is obtained by
contracting the second-order tensor A, the vector P is obtained by contracting the third-order tensor B, and the second-
order tensor Q is obtained by contracting the fourth-order tensor C. In general, the contraction of a tensor of order
n(≥ 2) results in a tensor of order n − 2.
As another application of the inner product, consider an array [aij ]. Let U be an arbitrary vector with components ui
in the vector basis {ei } and let
vi = aij uj (1.83)
If v1 , v2 , v3 are the components of another vector V in the vector basis {ei }, then aij are the components of a second
′ ′
order tensor A. To prove this result we note that in the vector basis {ei } the elements of the array [aij ] become [aij ],
where ′ ′ ′
vi = aij uj (1.84)
Furthermore, since U and V are vectors, their components transform according to
′
vi = ℓik vk (1.85)
′
uj = ℓmj um (1.86)
Also, (1.83) may be rewritten as
vk = akj uj (1.87)
Then, by (1.85) - (1.87),
′ ′ ′
vi = ℓik akj ℓmj um = ℓik akm ℓjm uj
and by (1.84)
′ ′
(aij − ℓik ℓjm akm )uj = 0 (1.88)
But, since U is an arbitrary vector, (1.88) must hold for
′
uj = δjr
This gives
′
air = ℓik ℓrm akm
which is the transformation law for a second order-tensor A. This result is known as the quotient rule, and it can be
easily extended to tensors of higher order. The quotient rule can be used to show that δij are the components of a
second-order tensor and ǫijk are the components of a third-order tensor.
Example 1.4-1
If A is a second order tensor and U, V are arbitrary vectors, show that the expression aij ui vj is an invariant.
Solution. Let the components of A, U and V in the vector basis {ei } be aij , ui and vi , respectively, and let the
′ ′ ′ ′
corresponding components in the vector basis {ei } be aij , ui and vi . Then
2005
c by Ajit Mal
1.5. SOME SPECIAL PROPERTIES OF SECOND-ORDER TENSORS 25
′ ′ ′
aij = ℓim ℓjn amn , ui = ℓip up , vj = ℓjq vq
Therefore,
′ ′ ′
aij ui vj = (ℓim ℓjn amn )(ℓip up )(ℓjq vq )
= (ℓim ℓip )(ℓjn ℓjq )amn up vq
= δmp δnq amn up vq = apq up vq
This shows that aij ui vj is an invariant. In general, if T is a tensor of the nth order and U, V,..., Z are n arbitrary
vectors, then
tij...s ui vj · · · zs
is an invariant. This property can be used as a definition of the nth-order tensor because the converse is also true.
The definitions and results discussed in the previous section hold for tensors of any order n. Most of the tensors that
appear in the theory of elasticity are of second order. A number of important special properties of such tensors are
described in this section.
An important property of a second-order tensor A is that its components aij are identical with the elements of the
square matrix A = [aij ]. Furthermore, the transformation law
′
aij = ℓik ℓjm akm (1.89)
becomes
′
aji = ℓjm ℓik amk
by a simple interchange of the indices i and j as well as k and m. Thus, the matrix AT = [aji ] also represents a tensor
of order two and is denoted by AT . The determinant of A can be defined as in Sec. 1.1 of of the appendix, and is
denoted by |A| or by det [aij ]. Similarly, the tensor A is symmetric or antisymmetric if the corresponding matrix A
is symmetric or antisymmetric. A diagonal tensor is one whose nonzero elements lie along the diagonal. Clearly, the
unit tensor I formed by the array [δij ] is diagonal.
All the rules of matrix algebra for square matrices discussed in Sec. 1.2 apply to Cartesian tensors of order 2.
Thus, the components of the product of the tensors A and B are identical with the elements of the matrix product AB,
and the components of (AB)T are identical with the elements of the matrix B T AT .
For a given tensor A the inverse of its matrix, A−1 , can be defined and calculated as in (1.12) and (1.13) provided
|A| =6 0, although the elements of the inverse do not, in general, constitute a tensor. A tensor A will be called
orthogonal if the matrix A is orthogonal, i.e., if its column vectors A(i) satisfy the property
T
A(i) A(j) = δij (1.90)
The transformation law of a second order tensor may be expressed by the equation
′
A = LALT (1.91)
′
It should be understood that the elements of A and those of A are the components of the same tensor A in the two
′
vector bases {ei } and {ei }, respectively. It is sometimes more convenient to use the matrix form of the transformation
rule (1.91) instead of the corresponding rule (1.89) in index notation.
As a special case of (1.91), we consider the transformation of a tensor A whose components are given by the matrix
2005
c by Ajit Mal
26 CHAPTER 1. CARTESIAN TENSORS
e2
0
e2
0
e1
o e1
0
e3 , e3
a11 a12 0
A = a21 a22 0
0 0 1
′
in the vector basis {ei }. The new vector basis {ei } is obtained by giving a rotation to {ei } about e3 through an angle
′
θ in a counterclockwise sense (Fig. 1-3). We wish to calculate the components of A in {ei }.
First, it is necessary to calculate the elements of the direction cosine matrix L. Clearly, the direction cosines of
′ ′ ′
e1 , e2 , and e3 relative to ei are (cos θ, sin θ, 0), (− sin θ, cos θ, 0) and (0, 0, 1), respectively. Thus in accordance
with (1.59),
c s 0
L = −s c 0
0 0 1
′ ′
where we have written c for cos θ and s for sin θ. Then, the matrix of the components aij of A in {ei } is given by
c s 0 a11 a12 0 c −s 0
LALT = −s c 0 a21 a22 0 s c 0
0 0 1 0 0 1 0 0 1
′ ′
a11 a12 0
′ ′
= a21 a22 0
0 0 1
Calculation of the matrix product gives
′
a11 = c2 a11 + s2 a22 + cs(a12 + a21 )
′
a12 = −cs(a11 − a22 ) + c2 a12 − s2 a21
′
a21 = −cs(a11 − a22 ) − s2 a12 + c2 a21 ( 1.92a)
′
a22 = s2 a11 + c2 a22 − cs(a12 + a21 )
′
If A is symmetric, i.e., if a12 = a21 , then A is also symmetric, and
′
a11 = a11 cos2 θ + a22 sin2 θ + a12 sin 2θ
′
a22 = a11 sin2 θ + a22 cos2 θ − a12 sin 2θ ( 1.92b)
′ 1
a12 = a21 = − (a11 − a22 ) sin 2θ + a12 cos 2θ
2
2005
c by Ajit Mal
1.6. PRINCIPAL AXES OF A REAL SYMMETRIC SECOND-ORDER TENSOR 27
′ ′
The angle for which a12 is maximum (or minimum) is given by (∂a12 /∂θ) = 0 i.e.,
1
2 (a11 = a22 ) cos 2θ + a12 sin 2θ = 0
Therefore,
′ 1
|(a12 )max | = [ (a11 − a22 )2 + a212 ]1/2 ( 1.92c)
4
′ 1
|(a12 )min | = [ (a11 − a22 )2 − a212 ]1/2 ( 1.92d)
4
A special antisymmetric tensor Ω is useful in the description of the deformation of a solid and will be introduced
here. Let u, v,θ be three vectors related through the equation
v =θ×u ( 1.93a)
It is well known from elementary mechanics that if u denotes the position vector of a particle within a rigid body and
θ is the (small) rotation vector of the body, then v represents the displacement of the particle. Equation (1.93a) implies
that
vi = ǫijk θj uk = ωik uk ( 1.93b)
where
ωik = ǫijk θj (1.94)
Since ui and vi are the components of first-order tensors in {ei }, according to the quotient rule, ωij are the
components of a second order tensor Ω in {ei }. Furthermore, it can be easily seen from (1.94) and the properties of
ǫijk that
0 −θ3 θ2
Ω = [ωij ] = θ3 0 −θ1 (1.95)
−θ2 θ1 0
Thus, Ω is an antisymmetric tensor. Equation (1.93) expresses the fact that the components of the displacement vi of
a particle with position vector ui produced by an infinitesimal rotation θi can be expressed by the matrix equation
V = ΩU (1.96)
It follows from (1.95) that if the displacement can be expressed in the product form (1.96), where Ω is an antisymmetric
matrix, then this displacement can be produced by an infinitesimal rotation θ of the body, where
θ1 = −ω23 = ω32
θ2 = −ω31 = ω13 (1.97)
θ3 = −ω12 = ω21
i.e.,
θi = − 21 ǫijk ωjk
Let A be a real symmetric second order tensor with components aij in the vector basis {ei }. Let V be an unknown
first order tensor whose components vi in {ei } are to be determined from the system of equations
2005
c by Ajit Mal
28 CHAPTER 1. CARTESIAN TENSORS
The preceding system of homogeneous, linear equations in the unknowns v1 , v2 , v3 has nontrivial solution if and
only if λ is a solution of
det [aij − λδij ]vj = 0 ( 1.99a)
or
a11 − λ a12 a13
a21 a 22 − λ a23 =0 ( 1.99b)
a31 a32 a33 − λ
where
D(λ) is called the characteristic polynomial, and (1.99) is the characteristic equation; its roots are called the charac-
teristic values (or eigenvalues). The nontrivial solutions of (1.98) are called the characteristic vectors or eigenvectors
of the tensor bf A. Example Problem (Lecture 2 Page 36)
Since D(λ) is a cubic in λ, there are three eigenvalues λ1 , λ2 , λ3 of A, and corresponding to each eigenvalue λk ,
there is an eigenvector V (k) . Clearly, if V is a solution of the homogeneous system (1.98), so is αV, where α is a
scalar. Thus the ”length” of the eigenvector is arbitrary. We shall assume that all eigenvectors have unit length.
The eigenvalues and eigenvectors of a real and symmetric tensor A have certain special properties that are relevant
in the mathematical treatment of elasticity. The most important of these are contained in the following theorems.
Proof. We first assume that λ is complex and show that its imaginary part must vanish if A is real and symmet-
ric.
If λ is complex, then V must, in general, also be complex-valued. Thus, taking the complex conjugate of (1.98a),
aij v j = λv i
where the overbar indicates complex conjugate and we have used the fact that aij is real. Thus,
aij v j vi = λv i vi
Interchanging the dummy indices i, j in the left-hand side of the preceding equation and using the symmetry property
of A (i.e., aij = aji ), it becomes
aij vj v i = λvi v i (1.101)
By (1.98a) and (1.101) we have
(λ − λ)vi v i = 0 (1.102)
But
2005
c by Ajit Mal
1.6. PRINCIPAL AXES OF A REAL SYMMETRIC SECOND-ORDER TENSOR 29
where |vk | is the modulus of the complex quantity vk . Since, by definition, V is a nonzero vector, vi v i is always
positive. Thus (1.102) implies that λ = λ, i.e., λ is real.
Theorem 1.6-2. The eigenvectors corresponding to two distinct eigenvalues of A are orthogonal.
Proof. Let V(1) , V(2) denote the two eigenvectors corresponding to two distinct eigenvalues λ1 , λ2 of A and
let vk1 , vk2 , vk3 denote the components of the vector V(k) in the vector basis {ei }. Then, by (1.98a),
and
aij v2j = λ2 v2i ( 1.103b)
Upon multiplying both sides of (1.103a) by v2i , interchanging the dummy indices i and j in the left-hand side of the
result, and using the symmetry property of A, we obtain
v1i v2i = 0
i.e., the eigenvectors V(1) and V(2) are orthogonal. Since the eigenvectors are normalized, the orthogonality property
may be expressed in the general form
vki vmi = δkm (1.105)
provided λk 6= λm .
Proof. If all three eigenvalues of A are distinct, then it is obvious from Theorem 1.6-2 that A has three mutu-
ally orthogonal eigenvectors. However, the proof of the theorem fails if some or all of the eigenvalues are equal.
We first assume that A has two distinct eigenvalues λ1 , λ2 . Let V(1) and V(2) be the eigenvectors corresponding
to the eigenvalues λ1 and λ2 , respectively. Let P be a vector orthogonal to both V(1) and V(2) . Then
pi v1i = pi v2i = 0
and, by (1.103),
aij v1i pi = λ1 v1i pi = 0, aij v2i pi = λ2 v2i pi = 0 (1.106)
Interchanging the dummy indices i and j in the left-hand sides of (1.106) and using the symmetry of A, we have
Let
aij pj = qi (1.108)
Since A is a second order tensor and P is a vector, by the inner product rule discussed in Sec. 2.3, q1 , q2 , q3 are the
components of a vector Q. Then, (1.107) implies that
qi v1i = qi v2i = 0
2005
c by Ajit Mal
30 CHAPTER 1. CARTESIAN TENSORS
i.e., Q is orthogonal to both V(1) and V(2) . Thus, Q must be parallel to P, and we may write
where α is a scalar. Equation (1.109) shows that must be an eigenvalue and P must be the corresponding eigenvector
of A. Since A has only three eigenvalues λ1 , λ2 , α, and two of these are equal, α must be either λ1 or λ2 . Furthermore,
the eigenvector P is orthogonal to both V(1) and V(2) , and we may denote its normalized form by V(3) . From (1.103)
and (1.109), it can be seen that any vector that lies in the plane containing the two eigenvectors corresponding to the
repeated eigenvalues is also an eigenvector of A. Thus, if the eigenvalues of A are λ1 , λ2 , λ2 , then any two orthogonal
vectors that lie on the plane normal to V(1) can be chosen as the other two eigenvectors of A. If all the eigenvalues of
A are equal, any three mutually orthogonal vectors may be chosen as eigenvectors of A.
Theorem 1.6-4. A can be transformed into a diagonal tensor by a proper rotation of the vector basis {ei }.
Proof. We have shown that the tensor A has three real eigenvalues, λ1 , λ2 , λ3 and three mutually orthogonal
eigenvectors, V(1) , V(2) , V(3) . Since the ordering of the eigenvalues is arbitrary, we may always order them so that
′ ′
V(1) , V(2) , V(3) form a right-handed triad {ei }. The base vectors ei are related to ei through the equation
′
ei = vij ej (1.110)
′
since (vi1 , vi2 , vi3 ) are the direction cosines of ei relative to {ei }. Thus, the matrix [vij ] is the direction cosine matrix
′
for a proper rotation from {ei } to {ei } as introduced in Sec. 2.1. The orthogonality relation between the unit base
′ ′ ′
vectors e1 , e2 , e3 is expressed by the equation
vik vjk = δij (1.111)
′ ′
Let aij denote the components of A in the vector basis {ei }. By the tensor transformation rule (1.89),
′
aij = vik vjm akm (1.112)
where a bar under an index implies that no sum is to be carried out over that index. Substituting from (1.113) into
(1.112) and using the orthogonality relation (1.111), we obtain
′
aij = vik λj vjk = λj δij (1.114)
′
Thus, the only nonzero values of aij are given by
′ ′ ′
a11 = λ1 , a22 = λ2 , a33 = λ3
′
Hence, when referred to {ei }, the tensor A becomes diagonal with its eigenvalues along the diagonal. This proves the
theorem. It can be shown that the result holds even if some or all of the eigenvalues are equal.
′
The base vectors ek or, equivalently, the specially ordered eigenvectors V(k) are called the principal axes of A and
the corresponding eigenvalues λk are called the principal values of A.
It should be noted that a given tensor has a unique set of principal values, i.e., λ1 , λ2 , λ3 are independent of the
vector basis in which the components of A are measured. Since λ1 , λ2 , λ3 are the roots of the characteristic equation
(1.99c), from the theory of equations, we have
where IA , IIA , IIIA are given in terms of the components of A by (1.100). Equations (1.115) imply that the values of
the expressions in the right-hand sides of (1.100) remain unchanged and equal to those in the left-hand sides of (1.115)
if A is referred to any other coordinate system. For this reason IA , IIA , IIIA are called the principal invariants of A.
2005
c by Ajit Mal
1.7. ISOTROPIC TENSORS 31
Clearly, all scalar functions of the invariants IA , IIA , IIIA are also invariants with respect to these coordinate trans-
formations.
Example 1.6-1
2
Show that if A is a symmetric second-order tensor, then I1 = tr(A) = aii , I2 = tr(A ) = aij aji and I3 =
3
tr(A ) = aij ajk aki are invariants and that
2 3
I1 = IA , I2 = IA − 2IIA , I3 = IA − 3IA IIA + 3IIIA (i - iii)
′ ′
Solution. Let the components of A in the vector bases {ei } and {ei } be aij and aij , respectively. Then
′
aij = ℓim ℓjn amn
Therefore,
′
aii = ℓim ℓin amn = δmn amn = amm
showing that I2 is invariant. We can similarly show that I3 is invariant. Equations (i) and (ii) follow from the definitions
given in (2.54). To derive (iii), we note from the Cayley-Hamilton theorem (see appendix) that
Inserting the values of I1 and I2 from (i) and (ii) in (viii), we obtain (iii). Equations (iv)-(vi) follow directly from the
definitions of I1 , I2 and I3 .
A tensor is called isotropic if its components retain the same values however the axes are rotated. It follows imme-
diately from the definition of zero order tensors (scalars) that they are isotropic. The following results can be easily
shown to be true.
Theorem 1.7-1. There are no nontrivial isotropic tensors of the first order.
Theorem 1.7-2. Any isotropic tensor of order 2 can be written as αI, where α is a scalar.
2005
c by Ajit Mal
32 CHAPTER 1. CARTESIAN TENSORS
Theorem 1.7-3. The alternating symbol ǫijk is a component of a third order isotropic tensor, ǫ, and any third-order
isotropic tensor can be written as αǫ for some scalar α.
Of these, the last result is of considerable interest in elasticity theory, and we present a simple and straightfor-
ward proof.
Each of i, j, k, and m can assume only the values 1, 2, or 3. Hence, we can divide the components of C into the
following four groups:
(i) Components c1111 , c2222 , c3333 in which all the suffixes are equal;
(ii) Components of the type c1112 (24 in all), in which three suffixes are equal and the fourth suffix is different;
(iii) Components of the type c1122 , c1212 , c1221 (18 in all), which have two different pairs of equal suffixes.
(iv) Components of the type c1123 , c1321 (36 in all), which have one and only one pair of equal suffixes.
Consider a rotation of axes about Ox3 through 180o , represented by the direction cosine matrix L given by
−1 0 0
L = 0 −1 0 (1.118)
0 0 1
i.e.,
c1113 = c1123 = 0 (1.119)
But c1113 and c1123 are but typical components from groups (ii) and (iv), respectively. Hence if C is isotropic, all the
components in groups (ii) and (iv) must vanish.
Next consider a rotation about Ox3 through 90o . The corresponding direction cosine matrix is given by
0 1 0
L = −1 0 0 (1.120)
0 0 1
Equation (1.117) then gives c1111 = c2222 and a cyclic change of suffixes yields
Considering the effect of the rotation (1.120) on the component c1122 , we find
2005
c by Ajit Mal
1.8. VECTOR AND TENSOR CALCULUS 33
we obtain
c1122 = c2233 (1.125)
Advancing the suffixes cyclically, we get
Thus,
c1122 = c1133 = c2211 = c2233 = c3311 = c3322 = α (say) ( 1.127a)
Similarly, it can be shown that
According to (1.117),
c1111 = ℓ1i ℓ1j ℓ1k ℓ1p cijkp (1.129)
It is apparent from (1.128) that i, j, k, p can take on only the values 1 and 2. Therefore, there are only 16 terms in
the right-hand side of (1.129). Out of the corresponding 16 values of cijkp , 8 belong to group (ii) and, therefore, they
vanish. In this manner, we obtain
1
c1111 = (c1111 + c2222 + c2211 + c1212 + c2121 + c1221 + c2112 ) (1.130)
4
On using (1.121) and (1.127), we find
d = 12 (d + α + β + γ)
i.e.,
d=α+β+γ (1.131)
We thus conclude that if C is isotropic, it has only 21 nonzero components given in (1.121) and (1.127). The
component cpqrs is nonzero if and only if either all its suffixes are equal or if it has two distinct pairs of equal suffixes.
From (1.121), (1.127), and (1.131), we get (1.116).
In the mathematical treatment of elasticity, it is very often necessary to work with tensors (and vectors) that describe
the physical properties of a material at a specific location within its bulk, and these properties may change with a
change of location or with time. Thus, the components of these tensors are, in general, functions of space and time,
and they are said to form a tensor (or vector) field. It is also often necessary to introduce the rate of change of their
components with respect to space or time within the region of space or in the time interval of interest. Differentiation
of the components may change the properties of a given tensor. Some of the relevant properties are discussed next.
2005
c by Ajit Mal
34 CHAPTER 1. CARTESIAN TENSORS
Let D be a region of space spanned by a fixed vector basis {ei }, (x1 , x2 , x3 ) be the coordinates of a point within
D and t be the time. A scalar function of the independent variables x1 , x2 , x3 , t will be denoted by f (x, t). Let
vi (x, t) and aij (x, t) be the components of the vector V and the second order tensor A in the vector basis {ei }. The
symbols V(x, t) and A(x, t) will be used to indicate that the vector and the tensor are functions of space and time.
The arguments x, t may be omitted if this can be done without confusion. We shall assume that all functions are
continuously differentiable in D to an arbitrary order.
The first partial spatial derivatives of a function f (x, t) will be denoted by
∂f
= f,k (1.132)
∂xk
where k = 1, 2, 3. Similarly, the spatial derivatives of vi (x, t) and aij (x, t) will be denoted by vi,k and aij,k respec-
tively. Thus the augmentation of the indices of a tensor by , k indicates differentiation with respect to xk . The higher
partials can be constructed by simply adding other indices to the right of the comma; for example,
∂ 2 aij
= aij,km (1.133)
∂xk ∂xm
The partial derivatives with respect to t will be denoted by f˙ or by f,t . Clearly, for ordinary functions, the order of
differentiation with respect to any two or more independent variables can be interchanged.
The preceding notation for the partial derivatives is called the comma notation and is often quite useful in express-
ing the equations of elasticity theory in compact form. The notation applies to higher-order tensors as well.
To see how differentiations alter the properties of tensors, consider the aggregate of the functions ui (x, t) given by
ui = f,i (1.134)
′
Then {ui } is an ordered triplet, and if the vector basis is changed to {ei }, resulting in a change in the coordinates from
′ ′ ′ ′
(x1 , x2 , x3 ) to (x1 , x2 , x3 ), {ui } changes to {ui }, where
′ ∂f
ui = ′ (1.135)
∂xi
Further, ′ ′
xi = ℓij xj , xj = ℓij xi (1.136)
where [ℓij ] is the direction cosine matrix introduced in Sec. 2.1. Thus,
′ ∂f ∂xj
ui = ′ = ℓij f,j = ℓij uj (1.137)
∂xj ∂xi
where we have used the fact that
∂xj
′ = ℓij (1.138)
∂xi
Equation (1.137) shows that u1 , u2 , u3 are the components of a vector U. This vector is called the gradient of f
and is written as grad f , or ∇f, i.e.,
gradf ≡ ∇f = f,i ei (1.139)
It can easily be seen that the functions f,ij are the components of a second- order symmetric tensor and f,ii is a scalar,
denoted by ∇2 f , where ∇2 is the well-known Laplacian operator
f,ii = ∇2 f (1.140)
2005
c by Ajit Mal
1.8. VECTOR AND TENSOR CALCULUS 35
′ ′ ′
where vi are the components of V in {ei }, i.e., vi = ℓik vk Thus,
′ ∂ ∂ ∂xm
bij = ′ (ℓik vk ) = (ℓik vk ) ′
∂xj ∂xm ∂xj
= ℓik ℓjm vk,m = ℓik ℓjm bkm
where we have used (1.138) and the fact that ℓik is independent of (x1 , x2 , x3 ). Thus bij is a component of the second-
order tensor B, not necessarily symmetric. Also, bii = vi,i̇ , i.e., a contraction of the tensor B is a scalar, called the
divergence of v which is denoted by
div v ≡ ∇ · v = vi,i (1.143)
Similarly, the functions vi,jk are the components of a third order tensor with the symmetry property vi,jk = vi,kj , and
so on.
The preceding results can be easily extended to higher order tensors. For example, cijk = aij,k are the components
of a third order tensor C and dijkm = aij,km are the components of a fourth-order tensor D with the symmetry property
dijkm = dijmk . In general, if T is a tensor of order n, then tijk...,m are the components of a tensor of order n + 1,
which is sometimes called the gradient of T. Clearly, differentiations with respect to t do not change the order of a
tensor. Furthermore, symbols such as (∂T/∂x) are meaningless in the context of our notations and conventions.
A number of identities involving (vector and) tensor functions are useful in the analysis of the deformation of
solids. In writing these we shall use the notations ∇f for the gradient of a scalar function f (x, t), ∇ · v for the
divergence of a vector v, and ∇ × v for the curl of a vector v. In index notation,
∇ × v = ǫijk vj,i ek
∇ × (∇f ) = 0 ( 1.144a)
2
∇·(∇f ) = ∇ f ( 1.144b)
∇·(∇ × v) = 0 ( 1.144c)
∇ × (∇ × v) = ∇(∇·v) − ∇2 v ( 1.144d)
It should be noted that all of these equations are written in direct vector notation introduced in Sec. 2.1 and thus they
are valid in Cartesian as well as curvilinear coordinate systems, notably cylindrical and spherical.
Two integral identities involving the components of vectors and tensors are also useful. These are provided by the
Green-Gauss theorem or the divergence theorem and Stokes’ theorem, as stated next. The divergence theorem. Let
V denote the volume of a region of space bounded by a piecewise smooth and closed surface S and its interior and let
ui (x) be the components of a vector field in a fixed vector basis {ei } that spans the space. Assume that the first partial
derivatives ui,j exist and are continuous in V . Then,
Z Z
u·ndS = ∇·udV (1.145)
S V
where n is the outward unit normal vector to S at the surface element dS (Fig. 1-4) and dV is a volume element.
The proof of this well known theorem can be found in standard texts on analysis (e.g., Kellogg 1939) and is omitted.
Equation (1.145) can be written in index notation as
Z Z
ui ni dS = ui,i dV (1.146)
S V
Let aij (x) be the components of a second-order tensor field in {ei } and p be an arbitrary constant vector with com-
ponents pi . Then the functions ui = aij pj are the components of a vector u and (1.146) becomes
Z Z
aij ni dS − aij,i dV pj = 0
S V
2005
c by Ajit Mal
36 CHAPTER 1. CARTESIAN TENSORS
S n
V dS
dS
c
ds
Fig. 1-5. The domain for Stokes’ theorem.
Stokes’ theorem. Let S be an open piecewise smooth surface bounded by a simple closed curve C (i.e., C
does not intersect itself) and let ui (x) be the components of a vector field in the vector basis {ei }. Assume that the
first partial derivatives of ui (x) exist and are continuous in the domain containing S and C. Then
I Z
u · ds = (∇ × u)·ndS (1.148)
C S
where ds is an oriented element on C and n is the unit normal to S at dS along its positive curvature (Fig. 1-5). This
result is also known as the circulation theorem. The proof of this is also omitted, since it can be found in standard texts
on analysis (e.g., Kellogg, 1939).
2005
c by Ajit Mal
1.9. PHYSICAL INTERPRETATION OF SOME VECTOR DIFFERENTIAL OPERATIONS 37
The vector field generated by taking the gradient of a scalar field can be given a useful physical interpretation. Let
φ(x) denote a scalar field representing a physical property (such as temperature or gravitational potential) of a solid.
At a neighboring point x + δx, its value can be approximated by the expression
Let δx = eδx, where e is a unit vector along δx. Then, the preceding equation gives
Proceeding to the limit δx → 0, we see that e · ∇φ represents the rate of change of φ in the direction e. Consider the
surfaces represented by the equation
φ(x) = λ
where λ is a parameter. These are called level surfaces for the physical property described by φ (i.e., equitemperature
or equipotential surfaces). Clearly, if e is tangential to the level surface passing through x, then e · ∇φ = 0 (since φ is
constant along e). Thus, the vector ∇φ is along the normal to the level surface at x, and the unit normal to φ(x) = λ
is given by ∇φ/|∇φ|.
The scalar fields generated by taking the divergence of certain vector fields can also be given physical interpreta-
tions. As an example, consider an aggregate of material particles contained within a fixed volume element V bounded
by the surface S. At a given instant of time, let q(x, t) denote the velocity of a material particle located at a point x
within V . The rate of outflow of material, Ḟ across S is given by the surface integral
Z
Ḟ = q · ndS
s
In other words, (∇ · q) represents the rate of outflow of material per unit volume at a spatial position x.
Finally, the vector field obtained by taking the curl of another vector field may also be given a physical interpreta-
tion. To this end, we consider the rotation of a rigid body through a small angle δθ, measured in radian, about a given
axis (say e3 ). Then the displacement of a material particle located initially at x is given by
U = δθe3 × (x1 e1 + x2 e2 + x3 e3 )
= δθ(−x2 e1 + x1 e2 )
Taking the curl of both sides, we obtain
∇ × U = e3 δθ
Hence ∇ × U represents the infinitesimal rotation vector of a rigid body through |∇ × U| radian about the axis
(∇ × U)/|∇ × U|.
2005
c by Ajit Mal
38 CHAPTER 1. CARTESIAN TENSORS
1.10 DYADICS
In the direct notation introduced in Sec. 2.1, a vector is represented by the expression
v = v i ei
in which the components vi of v in and the base vectors ei occur explicitly. All algebraic operations with this vector
can be performed by using the algebraic relations between the unit vectors ei as in (1.48), (1.49), and (1.50). If the
′ ′
components of v are vi when measured in another vector basis {ei }, then
′ ′
v = vi ei = vi ei (1.149)
The direct representation can also be extended to tensors. As an example, let aij be the components of a second-
order tensor A in the vector basis . Consider the expression
a = aij ei ej (1.150)
′
in which the unit vectors ei and ej are juxtaposed in a specific order and the summation convention applies. Let {ei }
be another vector basis whose base vectors are related to through the usual transformation law,
′
ei = ℓki ek
′
akm = ℓki ℓmj aij
′ ′ ′ ′ ′ ′
a = akm ek em = aij ei ej
and by (1.150),
′ ′ ′
a = aij ei ej = aij ei ej (1.152)
′
These expressions for a are similar to those for v in (1.149) except for the fact that the unit vectors ei and ei have
′ ′
been replaced by the juxtaposed pairs ei ej and ei ej . Thus, (1.150) can be interpreted as a direct representation of the
second-order tensor A, and algebraic operations can be carried out directly with a provided certain rules of operation
on the pair ei ej can be established. To this end, we recall that the components of v are obtained from the relation
v · ei = ei ·v = vi
and
(ek · a) · em = (ek ·aij ei ej )·em = aij δki ej ·em = akj δjm = akm
2005
c by Ajit Mal
1.11. SOME VECTOR IDENTITIES IN CYLINDRICAL AND SPHERICAL COORDINATES 39
2005
c by Ajit Mal
40 CHAPTER 1. CARTESIAN TENSORS
x3
x3 = z
o x2
r
x1
Fig. 1-6. The cylindrical coordinates.
2005
c by Ajit Mal
1.11. SOME VECTOR IDENTITIES IN CYLINDRICAL AND SPHERICAL COORDINATES 41
er · er = 1, eφ · ez = 0
we obtain
∂ ∂ ∂
∇·u = er + eφ + ez · (ur er + uφ eφ + uz ez )
∂r r∂φ ∂z
∂ur 1 1 ∂uφ ∂uz
= + ur + + (1.164)
∂r r r ∂φ ∂z
Similarly, noting that
er × er = 0, eφ × ez = −ez × eφ = er , ...
we find
∂ ∂ ∂
∇×u = er + eφ + ez × (ur er + uφ eφ + uz ez ) (1.165)
∂r r∂φ ∂z
1 ∂uz ∂uφ ∂ur ∂uz ∂uφ uφ 1 ∂ur
= er − + eφ − + ez + −
r ∂φ ∂z ∂z ∂r ∂r r r ∂φ
This result is more easily remembered in the operator determinant form
er reφ ez
1 ∂
∇ × u = ∂ r ∂∂φ ∂∂z (1.166)
r
ur ruφ uz
∇2 f = ∇ · ∇f
∂2f 1 ∂f 1 ∂2f ∂2f
= 2
+ + 2 2+ 2 ( 1.167a)
∂r r ∂r r ∂φ ∂z
ur 2 ∂uφ uφ 2 ∂ur
∇2 u = er [∇2 ur − − 2 ] + eφ [∇2 uφ − 2 + 2 ] + ez ∇2 uz ( 1.167b)
r2 r ∂φ r r ∂φ
2005
c by Ajit Mal
42 CHAPTER 1. CARTESIAN TENSORS
x3
P
x3
o x2
x1
x2
x1
Therefore, we have
∂eR ∂eR
= 0, = (cos θ cos φ)e1 + (cos θ sin φ)e2 − (sin θ)e3 = eθ
∂R ∂θ
∂eR
= −(sin θ sin φ)e1 + (sin θ cos φ)e2 = (sin θ)eφ (1.170)
∂φ
∂eθ ∂eθ
= 0, = −(sin θ cos φ)e1 − (sin θ sin φ)e2 e2 − (cos θ)e3 = −eR
∂R ∂θ
∂eθ
= −(cos θ sin φ)e1 + (cos θ cos φ)e2 = (cos θ)eφ
∂φ
∂eφ ∂eφ
= =0
∂R ∂θ
∂eφ
= −(cos φ)e1 − (sin φ)e2 = −(sin θ)eR − (cos θ)eθ
∂φ
2005
c by Ajit Mal
1.11. SOME VECTOR IDENTITIES IN CYLINDRICAL AND SPHERICAL COORDINATES 43
1 ∂uφ 1 ∂uθ
+ eθ eφ + − (cos θ)uφ eφ eθ
R ∂θ R sin θ ∂φ
1 ∂uR ∂uφ
+ − (sin θ)uφ eφ eR + eR eφ
R sin θ ∂φ ∂R
∂uθ 1 ∂uR
+ eR eθ + − uθ eθ eR
∂R R ∂θ
∂2f
2
1 ∂2f
2 2 ∂f 1 ∂ f ∂f
∇ f = + + + cot θ + ( 1.173a)
∂R2 R ∂R R2 ∂θ2 ∂θ sin2 θ ∂φ2
2uR 2 ∂ 2 ∂uφ
∇2 u = eR [∇2 uR − 2 − 2 (sin θuθ ) − 2 ]+
R R sin θ ∂θ R sin θ ∂φ
uθ 2 ∂uR 2 cos θ ∂uφ
+ eθ [∇2 uθ − 2 2 + 2 − 2 2 ] ( 1.173b)
R sin θ R ∂θ R sin θ ∂φ
uφ 2 ∂uR 2 cos θ ∂uθ
+ eφ [∇2 uφ − 2 2 + 2 + 2 2 ]
R sin θ R sin θ ∂φ R sin θ ∂φ
2005
c by Ajit Mal
44 CHAPTER 1. CARTESIAN TENSORS
PROBLEMS
bi = 21 ǫijk akj
ǫipq bi = aqp
b × c = ac
1.3. Show that the condition for three vectors a, b, and c to be coplanar can be written as
ǫijk ai bj ck = 0
1.4. Show that if aij is an antisymmetric tensor in one coordinates system, it remains antisymmetric in all coordinates
systems.
1.5. The components of a second-order tensor A in the vector basis are given by the matrix
1 3 0
A= 3 1 0
0 0 1
′
Calculate its components in a vector basis {ei } that is obtained by a proper rotation of through an angle of 30o
about e1 .
1.6. a. If ǫij are the components of a second-order tensor and λ, µ are scalars, deduce that
σij = 2µǫij + λǫkk δij
are also the components of a second order tensor, and show that
σkk = (3λ + 2µ)ǫkk
σij ǫij = 2µǫij ǫij + λ(ǫkk )2
σij σij = 4µ2 ǫij ǫij + λ(3λ + 4µ)(ǫkk )2
(d) (d)
b. If ǫij and σij are defined by
2005
c by Ajit Mal
1.11. SOME VECTOR IDENTITIES IN CYLINDRICAL AND SPHERICAL COORDINATES 45
1.7. Let A be a second-order tensor and B = A−1 . Using Cayley Hamilton theorem show that the principal invariants
of B can be expressed in terms of the principal invariants of A in the form
IIA IA 1
IB = , IIB = , IIIB = , |A| 6= 0
IIIA IIIA IIIA
1.8. Suppose is an orthogonal basis and that e3 is the eigenvector of a symmetric second-tensor A corresponding to
the eigenvalue λ3 . If λ1 .λ2 are the other two eigenvalues, show that there exists an angle θ such that A has the
representation
A = (λ1 cos2 θ + λ2 sin2 θ)e1 e1 + (λ1 sin2 θ + λ2 cos2 θ)e2 e2
+ (λ1 − λ2 ) sin θ cos θ(e1 e2 + e2 e1 ) + λ3 e3 e3
1.13. Given that f, g are scalars, and u, v are vectors, prove the following identites using index notation.
(a) Equation 1.144 a, b, c, d
(b) ∇2 (f g) = (∇2 f )g + f (∇2 g) + 2∇f · ∇g
(c) ∇2 (f u) = (∇2 f )u + f (∇2 u) + 2∇f · ∇u
(d) ∇2 (uv) = (∇2 u)v + u(∇2 v) + 2(∇u)T · ∇v
Note that ∇u and ∇v are dyadics.
1.14 Write down the components of ∇f, ∇ · u, ∇2 f, ∇2 u in cylindrical and spherical coordinates under symmetry
conditions.
BIBLIOGRAPHY
1. Brand, L., Vector and Tensor Analysis; Wiley, New York, 1947.
2. Drew, T. B., Handbook of Vector and Polyadic Analysis; Reinhold, New York, 1961.
3. Gibbs, J. W. and Wilson, E. B., Vector Analysis; Dover, New York, 1960.
4. Jeffreys, H., Cartesian Tensors, Cambridge University Press, Cambridge, 1931. Kellogg, O. D., Foundations of
Potential Theory, Springer, Berlin, 1939.
5. Kellogg, O. D., Foundations of Potential Theory, Springer, Berlin, 1939.
2005
c by Ajit Mal
46
2005
c by Ajit Mal
Chapter 2
In this chapter we present a brief exposition of the linear theory of elasticity as it can be found in most undergraduate
texts. This will set the stage for more advanced treatment, which is the primary objective of this book.
2.1 STRAIN
To begin with, let us consider a deformation that is independent of one Cartesian coordinate, say z, and parallel to the
xy-plane. This type of deformation is known as plane strain. We shall return to the problem of three-dimensional
deformation after we have completed our study of deformation in a plane.
Let a point P with coordinates (X, Y ) in the undeformed state be displaced to the point P ′ with coordinates (x, y)
due to the deformation of the body (Fig. 2-1). The vector displacement of the point P has Cartesian components
(U, V ), where
U = x − X, V =y−Y (2.1)
It will be assumed that the displacement components are continuous and twice differentiable functions of X, Y
or x, y. Consider a small rectangular element P QRS in the undeformed state, with sides (dX, dY ) parallel to the
coordinate axes (Fig. 2-2). Let the points P, Q, R and S move, respectively, to P ′ , Q′ , R′ and S ′ after deformation.
The coordinates of Q relative to P are (dX, 0) and the coordinates of Q′ relative to P ′ are (dx, dy), where
∂U
dx = dX + dU = dX + dX (2.2a)
∂X
∂V
dy = dV = dX (2.2b)
∂X
We thus have
" 2 2 #1/2
′ ′ ∂U ∂V
P Q = dX + dX + dX
∂X ∂X
∂U
≈ dX + dX
∂X
where we have neglected the squares and higher powers of (∂U/∂X) and (∂V /∂X). Therefore, the increase in length
per unit length of the line P Q, denoted by ǫxx is given by
P ′ Q′ − P Q ∂U
ǫxx = = (2.3a)
PQ ∂X
47
48 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
P′
U
y
x
0 X
x
Fig. 2-1. The displacement vector.
R’
S’
2
S
R
Q’
@V
@X dX
dY
P’
1
dx + @U
@X dX
Q
P
dX
2005
c by Ajit Mal
2.2. STRESS 49
∂U ∂V
2ǫxy = γ1 + γ2 = + = γxy (2.5)
∂Y ∂X
The quantity ǫxy is the shearing strain. The symbol γxy = 2ǫxy is often used to denote the shearing strain in
engineering applications.
Relations (2.3) and (2.5) are obtained under the assumption that the displacement derivatives are small compared
to unity. The theory of elasticity in which the products and squares of derivatives of the displacement components with
respect to the space coordinates are neglected in comparison with the derivatives themselves is known as infinitesimal
or linear theory. In this chapter we shall confine ourselves to the linear theory only.
We have seen that in the two-dimensional problem of plane strain, there are two normal strains, ǫxx and ǫyy ,
and one shearing strain, ǫxy . In the general case of three-dimensional deformation, there are three normal strains,
ǫxx , ǫyy , ǫzz , and three shearing strains, ǫyz , ǫzx and ǫxy . They are related to the displacement components U, V , and
W through the equations
∂U ∂V ∂W
ǫxx = , ǫyy = , ǫzz = (2.6a)
∂X
∂Y ∂Z
1 ∂V ∂W 1 1 ∂W ∂U 1
ǫyz = + = γyz , ǫxz = + = γxz (2.6b)
2 ∂Z ∂Y 2 2 ∂X ∂Z 2
1 ∂U ∂V 1
ǫxy = + = γxy (2.6c)
2 ∂Y ∂X 2
2.2 STRESS
A solid may be acted upon by two types of external forces: body forces and surface forces. Body forces act upon
every volume element of the solid. Surface forces, in contrast, are forces that act upon every element of the surface of
the solid. In addition to the external forces, there are internal forces which arise from the mutual interaction between
various parts of the body.
Let a deformable solid be in its unstrained state with no forces acting on it and let a system of forces be then applied
to it. On account of the application of these forces, the solid becomes deformed, and a system of internal forces is set
up within it to oppose this deformation. These internal forces give rise to what is known as stress within the solid.
Let us consider a part of the solid occupying a region V enclosed by the surface S in the deformed state. The
boundary S is acted upon by surface forces caused by the action of the material exterior to V on that within V . It will
be assumed that these surface forces are continuously distributed over S. A suitable measure of such forces is their
intensity, i.e., the amount of force per unit area of the surface on which they act.
To specify the stress acting on a small area δS at a point P on S, we assume that the forces acting across this
elementary area, due to the action of the material outside V , can be reduced to a single force δp (Fig. 2-3) and that the
limit
δp
lim (2.7)
δS→0 δS
exists. This limit is known as the traction, or the stress vector on δS at the point P .
In general, the traction is inclined to the area δS on which it acts, and we can resolve it into two components: a
normal stress perpendicular to the area and a shearing stress acting in the plane of the area δS. Alternatively, we can
2005
c by Ajit Mal
50 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
δp
n
P
δs
S
Z
R V
O
Y
resolve the traction into its Cartesian components. Let the outward drawn unit normal to δS be n and the x-, y-, and
(n) (n) (n)
z- components of the traction acting on δS be denoted by tx , ty , and tz , respectively.
Consider a small cubic element with sides parallel to the coordinate axes (Fig. 2-4). The components of the traction
(1) (1) (1)
acting on the face with normal in the positive x−direction are tx , ty , tz . We use the notations
t(1)
x = σxx , t(1)
y = σxy , t(1)
z = σxz (2.8)
Similarly, the components of the traction acting on the face of the cube with normal in the positive y-direction are
σyx , σyy , σyz , and the components of the traction acting on the face of the cube with normal in the positive z-direction
are σzx , σzy , σzz . Thus the first suffix indicates the direction of the normal to the face and the second suffix indicates
the direction of the traction component. In all, we have nine components, σxx , σxy , ..., σzz , which are known as the
components of stress. We can display these components in the form of the matrix
σxx σxy σxz
σyx σyy σyz (2.9)
σzx σzy σzz
Obviously the components σxx , σyy , σzz represent normal stresses and the components σxy , σyx , σxz , σzx , σyz , σzy
represent shearing stresses.
We use the convention that the normal stress on a surface is positive when it produces tension and negative when it
produces compression of the material within the element. The positive direction of a component of the shearing stress
on any face of the cubic element are taken in the positive (negative) direction of the coordinate axis if a tensile stress
on the same face is in the positive (negative) direction of the corresponding axis. This rules is illustrated in Fig. 2-4
by indicating the positive directions of thecomponents σxx , σxy , and σxz for the two faces of the cubic element with
normals in the positive x-direction and the negative x-direction, respectively.
2005
c by Ajit Mal
2.3. EQUATIONS OF EQUILIBRIUM 51
y
yy
yx
xy xy xy
yz
zy xx xx
xx
x
xz xz
zx xz
zz
z
Consider the equilibrium of a small rectangular parallelepiped with its center at P (x, y, z) and edges δx, δy, δz parallel
to the coordinate axes (Fig. 2-5). The centers of the six faces of the parallelepiped are at the points
1 1 1
(x ± δx, y, z), (x, y ± δy, z), (x, y, z ± δz)
2 2 2
If the components of stress at P are σxx , σxy , ..., σzz , then the components of the traction acting on the face with
x
(x 2 ; y; z ) (x + x
2 ; y; z )
y z
y
x
x
z
since the outward normal to this face is in the negative x-direction. Similar expressions can be written down for the
remaining four faces. If (fx , fy , fz ) denote the components of the external body force per unit volume, then summing
2005
c by Ajit Mal
52 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
The symmetry relations (2.11) imply that there are only six independent components of stress.
We now prove an important result namely, the traction at a point across an arbitrarily oriented surface element can
be expressed as a linear combination of the stress components at that point. To this end, we assume, for simplicity,
that
σzx = σzy = σzz = 0 (2.12)
Consider the equilibrium of a small element of volume in the form of a right triangular prism (Fig. 2-6). Let the
external body force per unit volume be (fx , fy , 0). Because of the conditions (2.12), the two triangular end sections of
the prism are traction-free. Therefore, the material inside the prism is in equilibrium under the tractions acting on the
three rectangular faces of the prism and the body force (fx , fy , 0). Noting that the outward drawn normals to the faces
in the yz- and xz-planes are, respectively, in the negative x- and y-directions, the components of tractions on these
faces are (−σxx , −σxy , 0) and (−σyx , −σyy , 0). Let the unit normal to the slanted face be n and the components of
(n) (n) (n)
traction on it be denoted by (tx , ty , tz ). Then, summing forces along the x-axis, we find
1
t(n)
x AB − σ xx OB − σ yx OA + f x OP · AB =0
2
where P is the foot of the perpendicular drawn from the origin to the line AB. Noting that OA = AB sin θ and
OB = AB cos θ and taking the limit as OP → 0, keeping n fixed, we find that
t(n)
x = σxx cos θ + σyx sin θ (2.13a)
2005
c by Ajit Mal
2.3. EQUATIONS OF EQUILIBRIUM 53
ty
B n
xx P tx
xy
o yx A x
yy
Fig. 2-6. Equilibrium of a prism element.
where (nx , ny , nz ) = (cos θ, sin θ, 0) are the components of n. This result can be extended to the general three-
dimensional stress state, as will be seen later.
Equilibrium of the element shown in Fig. 2-6 leads to another important property of the stress, namely, its trans-
formation due to a rotation of the coordinate axes. To derive this property, let the system (Ox′ y ′ z ′ ) be obtained from
the system (Oxyz) by a rotation about the z-axis through an angle θ. The components of stress relative to the system
′ ′
(Ox′ y ′ z ′ ) are denoted by σxx , σyz , etc. (Fig. 2-7). Summing forces along the x′ −, y ′ −, and z ′ -directions and taking
the limit as OP → 0, we obtain the relations
′
σxx = σxx cos2 θ + σyy sin2 θ + 2σxy sin θ cos θ
′
σxy = (σyy − σxx ) sin θ cos θ + σxy (cos2 θ − sin2 θ)
′
σxz = 0 (2.15a)
We can similarly show that
′
σyy = σxx sin2 θ + σyy cos2 θ − 2σxy sin θ cos θ
′ ′
σyz = σzz = 0
2005
c by Ajit Mal
54 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
The transformation property (2.15) can be used to calculate the stress components on an element oriented in an arbi-
trary direction. The transformation equation for the three-dimensional stress state will be obtained later.
B xy
0
xx x
0 0
y 0
xx
P
xy
o A x
yx
yy
Fig. 2-7. Transformation of stress due to rotation of xy-axes.
Up to this point, we have considered the mathematical description of strain, stress, and the equations of equilibrium
of a solid without paying any attention as to the specific nature of the material. In order to complete the mathematical
formulation, it is necessary to introduce additional equations that relate the stress and strain components in the solid
under consideration. We shall now derive these equations for the special case of a linear elastic material. By elastic
we mean that the material will return to its original shape and size upon the removal of the deforming loads. By linear
we mean that the stress-strain relationship is linear.
The concept of linear elasticity was first introduced by Robert Hooke in 1678 when he postulated the law ut tensio
sic vis, which can be translated as ”the extension is proportional to the force.” A typical observed stress-strain law
for a material subjected to uniaxial stress is sketched in Fig. 2-8. The law is linear and elastic up to a stress called
the proportional limit. The behavior is nonlinear but still elastic up to a stress called the yield stress. The material
undergoes permanent deformation if loaded beyond the yield point; i.e., a certain amount of residual strain remains
after the load is removed completely. The terms ultimate stress and rupture stress are self-explanatory. Thus Hooke’s
law holds below the proportional limit.
For three-dimensional bodies, we do not have just extension but, instead, require six components of strain to
describe the deformation at a point. Similarly, the complete specification of the internal force operating at a point
requires the use of six components of stress. Cauchy and others developed the relationship between the components
of stress and strain that is known as generalized Hooke’s law. According to this law, each component of stress at any
point of a solid is a linear and homogeneous function of the components of strain at that point.
If the elastic properties of a solid at given point are the same in every direction, it is called elastically isotropic. A
2005
c by Ajit Mal
2.4. HOOKE’S LAW 55
medium that is not isotropic is called aeolotropic, or anisotropic. If the material property is the same at all points within
the solid, it is called homogeneous. In this chapter, we shall confine our discussions to isotropic and homogeneous
solids only.
Consider a unit cube of a material with its edges parallel to the coordinate axes and subjected to the action of the
Yield Point
Rupture Stress
Ultimate Stress
Proportional Limit
Fig. 2-8. Typical observed stress-strain curve for materials under uniaxial tension.
normal stress σxx uniformly distributed over opposite faces perpendicular to the x-axis. Then, according to Hooke’s
law the extension of the element is given by
1
ǫ(1)
xx = σxx
E
in which E is a material constant known as Young’s modulus. Clearly, from physical considerations, E > 0 and the
extension of the element in the x-direction is accompanied by lateral contraction resulting in normal strains in the y-
(1)
and z-directions. It is reasonable to assume that these strains are proportional to ǫxx :
ν
ǫ(1) (1) (1)
yy = ǫzz = −νǫxx = − σxx
E
The material constant ν is called the Poisson’s ratio of the solid. Similarly, corresponding to the normal stress σyy ,
we have
ν 1 ν
ǫ(2)
xx = − σyy , ǫ(2)
yy = σyy , ǫ(2)
zz = − σyy
E E E
and corresponding to the normal stress σzz ,
ν ν 1
ǫ(3)
xx = − σzz , ǫ(3)
yy = − σzz , ǫ(3)
zz = σzz
E E E
Since the stress-strain relationship is linear, the principle of superposition can be applied. Thus, if the element of
volume is subjected simultaneously to the action of normal stresses σxx , σyy , and σzz uniformly distributed over the
faces of the cube, the resultant components of strain can be obtained from the preceding as
1
ǫxx = [σxx − ν(σyy + σzz )] (2.16a)
E
1
ǫyy = [σyy − ν(σzz + σxx )] (2.16b)
E
1
ǫzz = [σzz − ν(σxx + σyy )] (2.16c)
E
2005
c by Ajit Mal
56 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
If shearing stresses σyz , σzx , and σxy act on all the sides of the cube, the change in the angle between any two of its
intersecting faces depends only on the corresponding shearing-stress component. We have seen in Sec. 2.1 that 2ǫxy is
equal to the decrease in the right angle between two line elements that were parallel to the x-and y-axes, respectively,
before deformation. We can therefore assume that
1 1 1
2ǫyz = σyz , 2ǫzx = σzx , 2ǫxy = σxy (2.16d)
µ µ µ
The material constant µ is called the modulus of rigidity, or shear modulus, and from physical considerations, µ > 0.
The extensions given by (2.16a, b, c) and the distortions given by (2.16d) are independent of each other. The
general strains, produced by simultaneous application of the three normal and three shearing components of stress,
can be obtained by superposition.
Sometimes it is desirable to express the stress components in terms of the strain components. This is easily achieved
through inversion of (2.16), and the results can be expressed in the form
E
σxx = λd + ǫxx (2.17a)
1+ν
E
σyy = λd + ǫyy (2.17b)
1+ν
E
σzz = λd + ǫzz (2.17c)
1+ν
σxy = 2µǫxy , σyz = 2µǫyz , σxz = 2µǫxz (2.17d)
where
d = ǫxx + ǫyy + ǫzz (2.17e)
and
νE
λ= (2.18a)
(1 + ν)(1 − 2ν)
The three material constants E, µ, and ν are related by the equation
E
µ= (2.18b)
2(1 + ν)
Similarly, from (2.19), the components of strain in the primed system are
′ 1+ν ′ ′ ′ ′ ′
ǫxy = − σo , ǫxx = ǫyy = σzz = ǫxz = ǫyz = 0
E
Since the medium under consideration is isotropic, we can use relations (2.16) even in the primed system. Thus,
the preceding equations give
′ 1 ′
2ǫxy = σ
µ xy
2005
c by Ajit Mal
2.4. HOOKE’S LAW 57
3κν Eν
λ= 1+ν (1+ν)(1−2ν)
λ(1−2ν) 3κ(1−2ν) E
µ= 2ν 2(1+ν) 2(1+ν)
2µ λ(1+ν) E
κ= λ+ 3 3ν 3(1−2ν)
µ(3λ+2µ) λ(1+ν)(1−2ν)
E= λ+µ ν 3κ(1 − 2ν)
λ
ν= 2(λ+µ)
i.e.,
1+ν 1
−2 σo = (−σo )
E µ
which leads to the relation (2.18b). Hence, of the three constants introduced before, only two are independent; i.e., an
isotropic material is characterized by two independent material constants.
With the help of (2.18b), we may write (2.17a-d) in the form
The constants λ and µ are known as Lame constants. From (2.18a,b), we can express E, ν in terms of λ, µ as
µ(3λ + 2µ) λ
E= , ν= (2.21a)
λ+µ 2(λ + µ)
2005
c by Ajit Mal
58 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
equilibrium derived in Sec. 2.3, are for an element of the solid in its deformed state, so that the derivatives in (2.10)
are with respect to the current coordinates x, y, z. It will be shown later that in a consistent linear theory in which all
nonlinear effects are ignored, no distinction need to be made between the coordinates (X, Y, Z) and (x, y, z).
In many elementary engineering applications, the primary quantities of interest are the stress components produced
in a solid under given boundary loads only. In such cases, the body force terms in the equations of equilibrium (2.10)
are absent. It is often possible to construct solutions of the resulting homogeneous system of equations to obtain stress
states, which can then be subjected to the appropriate boundary conditions. A few simple examples of this approach
are presented next.
In the first example we consider a beam of symmetric cross section subjected to uniform tensile load σ0 applied to
its end faces shown in Fig. 2-9. The objective is to determine the stresses produced within the beam.
In order to construct a solution of the problem, we locate a coordinate system with x-axis parallel to the beam (Fig.
2-9). Let the normal vector to the boundary of the cross section be (0, ny , nz ). Then by (2.14) the components of the
traction or stress vector at a point on the curved boundary of the beam, tx , ty , tz are related to the stress components
at the point through the equations
Since there is no applied load on the curved boundary, the components of the stress vector must vanish on it. Further,
the boundary conditions at both end sections are easily seen to be
σxx = σ0 (2.23a)
The boundary conditions as well as the equilibrium equations in the absence of body forces are clearly satisfied by the
stress field
σxx = σ0 , σyy = σzz = σxy = σyz = σxz = 0 (2.23b)
The strain components within the beam can be easily calculated from (1.16) and (1.23) as
σ0 νσ0 νσ0
ǫxx = , ǫyy = − , ǫzz = −
E E E
ǫxy = ǫyz = ǫxz = 0 (2.24)
Thus, (2.22) and (2.23) give a possible solution for the stress and strain components within the block. However,
not all stress states that satisfy the equilibrium equations and boundary conditions are physically possible. Some of
these solutions may lead to multiple-valued displacements within the material, leading to creation of cracks and other
features that are absent in the original problem. The stress state given by (2.23) happens to be the correct solution of
the present problem. A detailed discussion of this and related issues will be given later.
As another example of the elementary solution, assume that the beam of the previous problem is subjected to an
end moment of magnitude Mz about the z-axis, as shown in Fig. 2-10. The boundary tractions on the curved surface
are still given by (2.22), and these must vanish.
The conditions at the end sections are specified not in terms of the distributed stress vector, but their resultant
2005
c by Ajit Mal
2.4. HOOKE’S LAW 59
2005
c by Ajit Mal
60 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
At the end sections, the stress vector is subjected to the conditions (2.25a, b, c) and
Z
(yσxz − zσxy ) dA = Mx (2.29a)
ZA
(zσxx − xσxz ) dA = 0 (2.29b)
ZA
(xσxy − yσxx ) dA = 0 (2.29c)
A
Assume that the stress state within the cylinder is of the form
where c is a constant. The equilibrium equations (2.10) without body forces as well as the boundary conditions (2.28)
are identically satisfied by these stresses. The conditions (2.25a, b, c) and (2.29a, b, c) are satisfied if the x-axis passes
through the center of the cross section and if
Z
Mx = c (y 2 + z 2 ) dA
A
Thus,
Mx
Z
c= , Io = (y 2 + z 2 ) dA (2.31)
Io A
where Io is the polar area moment of inertia of the cross section and can be easily shown to be equal to πa4 /2. It
should again be pointed out that the preceding solution is exact if the stress vector at the end sections are prescribed in
accordance with (2.30). Otherwise it is valid at some distance away from the ends.
Temperature changes may be a source of significant stresses in elastic solids. It is reasonable to assume that a linear
relationship exists between the temperature difference and the normal strains caused by it and that thermal strains do
not involve shearing of the material. Thus, if Θ0 denotes the uniform temperature of the body in its reference state and
Θ the temperature in the current deformed state, we may write
where
T (x, y, z) = Θ − Θ0
2005
c by Ajit Mal
2.5. THERMAL EFFECTS 61
Equations (2.33) or (2.34), together with the strain-displacement relations (2.6) and the equilibrium equations
(2.10), form the basis for the solution of linear thermoelastic problems for given temperature field and boundary con-
ditions. It should again be noted that the independent variables in (2.6) and (2.10) are not the same, and a consistent
theory needs to be developed before the solution of general problems of thermoelasticity can be attempted.
2005
c by Ajit Mal
62 CHAPTER 2. A REVIEW OF THE ELEMENTARY THEORY OF ELASTICITY
PROBLEMS
U = 3X 2 Y, V = Y 2 + 6XZ, W = 6Z 2 + 2Y Z
calculate the strain components at the point (1, 0, 2). What is the extension of a line element dX (parallel to the
x-axis) at this point?
2.2. Show that equal extension and contraction of two orthogonal linear elements is equivalent to a shearing strain of
equal magnitude, that is associated with directions bisecting the angles between the elements.
where σ0 is a constant. Determine the components of the stress vector, tx , ty , tz , at the point P on a plane with
normal in the direction (2, 2, 1). Also determine the normal and shearing stresses at the point P on this plane.
2.5. Determine the body forces for which the following stress field describes a state of equilibrium:
σxx = −2x2 − 3y 2 − 5z
σyy = −2y 2 + 7
σzz = 4x + y + 3z − 5
σyz = 0
σzx = −3x + 2y + 1
σxy = z + 4xy − 6
2.6. Determine whether the following stress field is admissible in an elastic body when the body forces are negligible:
2.7. Determine f (x, y) so that the stress distribution given below may be in equilibrium in the absence of body forces.
σxx = y 2 + c(x2 − y 2 )
σyy = x2 + c(y 2 − x2 )
σzz = x2 + y 2
σyz = σzx = 0, σxy = f (x, y)
2.8. Show that equal tension and compression across two orthogonal planes is equivalent to a shear stress of equal
magnitude across a plane that bisects the angle between the two given planes.
2005
c by Ajit Mal
2.5. THERMAL EFFECTS 63
2.9. Determine the stresses in a loaded semicircular arch of rectangular cross section shown in Fig. 7-4, under the
assumption that its plane sections remain plane (strength of materials approach). Assume P = Q = 0.
BIBLIOGRAPHY
1. Timoshenko, S. P. and Goodier, J. N., Theory of Elasticity, McGraw-Hill, New York, 1951.
2005
c by Ajit Mal
64
2005
c by Ajit Mal
Chapter 3
Linear Elastostatics
The theory of elasticity in which the displacement gradients ui,j are assumed to be small compared to unity is known
as the linear theory of elasticity, and the corresponding strain tensor is called the infinitesimal strain tensor. For ready
reference, we list the essential equations of the linear theory of elasticity derived in Ch. 2 in index notation We assume
that the body whose deformation is to be studied occupies the region V bounded by the surface S in the deformed state.
Equilibrium equations
σij,j + fi = 0 x∈V (3.1)
where fi is the body force per unit volume. It should be emphasized that these equations are not valid at a boundary or
at an interface between two media; the boundary or interface values of the field variables must be obtained as limiting
cases from their interior values.
where the components of the elastic tensor c satisfy the symmetry relations
Strain-displacement relations
1
ǫij = (ui,j + uj,i ) (3.4)
2
Strain Compatibility conditions
or
σij nj = ti (B) (prescribed tractions) (3.6b)
where ui (B) and ti (B) are given functions and n is a unit outward drawn normal to S. In the so-called mixed
boundary-value problems, the traction is prescribed on one part of the boundary, and the displacement is prescribed on
65
66 CHAPTER 3. LINEAR ELASTOSTATICS
the remainder. At an interface between two different materials, certain continuity conditions must be enforced. If the
materials are perfectly bonded, or welded, then both displacement and stress vectors must be continuous across their
common interface. Conditions at other types of interfaces depend upon the specific nature of the bond.
The stored energy per unit volume in the solid W, also called the strain energy density, can be expressed in the
form
1 1
W = σij ǫij = cijkl ǫij ǫkl (3.7)
2 2
It should be emphasized that (3.7) and (3.8) are not valid in the nonlinear case.
For an isotropic medium, cijkl is given by
λ 1 ν 1+ν
ǫij = − sδij + σij = − sδij + σij (3.12)
2µ(3λ + 2µ) 2µ E E
where Θ is the temperature increase and α is the thermal coefficient of linear expansion. Similarly, (3.12) is to be
replaced by the Duhamel-Neumann relations:
ν 1+ν
ǫij = − sδij + σij + α∆Θδij (3.14)
E E
In cylindrical coordinates (r, φ, z), the components of the equilibrium equation are
2005
c by Ajit Mal
3.1. FIELD EQUATIONS OF LINEAR ELASTICITY 67
x3
zz
z
dz
zr
r
rz
r rr
z
z
o
d x2
r rd
x1 dr
where
d = ǫrr + ǫφφ + ǫzz = ∇ · u (3.16e)
The strain-displacement relations are
∂ur 1 ∂uφ ∂uz
ǫrr = , ǫφφ = + ur , ǫzz = (3.17a)
∂r r ∂φ ∂z
1 ∂uφ 1 ∂uz 1 ∂ur ∂uz
ǫφz = ǫzφ = + , ǫzr = ǫrz = + (3.17b)
2 ∂z r ∂φ 2 ∂z ∂r
1 1 ∂ur ∂uφ uφ
ǫrφ = ǫφr = + − (3.17c)
2 r ∂φ ∂r r
2005
c by Ajit Mal
68 CHAPTER 3. LINEAR ELASTOSTATICS
x3
RR
R
R
R
R
d
x2
o
d
R Rd
x1 dR
where
∂uR 1 ∂uθ
ǫRR = , ǫθθ = + uR (3.20a)
∂R R ∂θ
1 ∂uφ
ǫφφ = + (sin θ)uR + (cos θ)uθ (3.20b)
R sin θ ∂φ
1 1 ∂uθ ∂uφ
ǫθφ = ǫφθ = + − (cot θ)uφ (3.20c)
2R sin θ ∂φ ∂θ
1 1 1 ∂uR ∂uφ uφ
ǫφR = ǫRφ = + − (3.20d)
2 R sin θ ∂φ ∂R R
1 1 ∂uR ∂uθ uθ
ǫθR = ǫRθ = + − (3.20e)
2 R ∂θ ∂R R
2005
c by Ajit Mal
3.2. NAVIER’S EQUATIONS 69
cijkl ǫkl,j + fi = 0
cijkl uk,lj + fi = 0
The Navier’s equation for an isotropic, homogeneous, thermoelastic medium follows from (3.1) and (3.13):
Cylindrical coordinates(r, φ, z)
∂ ∂ ∂
∇f (r, φ, z) = er + eφ + ez f
∂r r∂φ ∂z
∂ur 1 1 ∂uφ ∂uz
∇·u = + ur + +
∂r r r ∂φ ∂z
2 2
∂ f 1 ∂f 1 ∂ f ∂2f
∇2 f = + + +
∂r2 r ∂r r2 ∂φ2 ∂z 2
ur 2 ∂uφ uφ 2 ∂ur
∇2 u = er [∇2 ur − 2 − 2 ] + eφ [∇2 uφ − 2 + 2 ] + ez ∇2 uz
r r ∂φ r r ∂φ
Spherical coordinates(R, θ, φ)
∂ ∂ 1 ∂
∇f (R, θ, φ) = eR + eθ + eφ f
∂R R∂θ R sin θ ∂φ
∂uR 2 1 ∂uθ 1 ∂uφ
∇·u = + uR + + (cot θ)uθ +
∂R R R ∂θ R sin θ ∂φ
2 2
1 ∂2f
∂ f 2 ∂f 1 ∂ f ∂f
∇2 f = + + + cot θ +
∂R2 R ∂R R2 ∂θ2 ∂θ sin2 θ ∂φ2
2uR 2 ∂ 2 ∂uφ
∇2 u = eR [∇2 uR − 2 − 2 (sin θuθ ) − 2 ]+
R R sin θ ∂θ R sin θ ∂φ
uθ 2 ∂uR 2 cos θ ∂uφ
+ eθ [∇2 uθ − 2 2 + 2 − 2 2 ]
R sin θ R ∂θ R sin θ ∂φ
uφ 2 ∂uR 2 cos θ ∂uθ
+ eφ [∇2 uφ − 2 2 + 2 + 2 2 ]
R sin θ R sin θ ∂φ R sin θ ∂φ
2005
c by Ajit Mal
70 CHAPTER 3. LINEAR ELASTOSTATICS
If the traction is prescribed over a part Sσ of the boundary S and the displacement is prescribed over the remaining
part Su , the solution of the elastostatic problem as specified by (3.1) and (3.6) is unique.
(1) (2)
To prove this uniqueness theorem, let us assume that two solutions ui and ui are possible. Then the equilibrium
equations are
(1) (2)
σij,j + fi = 0, σij,j + fi = 0
where
(1) (1) (2) (2)
σij = cijkl uk,l σij = cijkl uk,l
and
Then
σij,j = 0 in V (3.24a)
and
ui = 0 on Su ; σij nj = 0 on Sσ (3.24b)
Let W be the strain energy function corresponding to the displacement field ui . Then,
Z Z
2 W dV = σij ǫij dV
V V
1
Z
= σij (ui,j + uj,i ) dV
2
ZV
= σij ui,j dV (since σij = σji ) (3.25)
ZV Z
= (σij ui ),j dV − σij,j ui dV
V V
Z Z
= σij nj ui dS − σij,j ui dV
S V
where we have used the divergence theorem in converting the volume integral into the surface integral. Equations
(3.24) and (3.25) yield Z
W dV = 0 (3.26)
V
Since W is nonnegative, (3.26) implies that W = 0 in V . Further, since W is a positive-definite quadratic form in
the strain components, it cannot vanish unless all the strain components vanish. Thus, W = 0 implies that ǫij = 0,
(1) (2)
which, in turn, implies that ui = ui − ui represents a rigid-body displacement. Thus the solution is unique, except
for a rigid-body displacement. However, since ui = 0 on Su , we must have ui = 0 everywhere in the body. This
(1) (2)
means that ui = ui , which proves the uniqueness of the solution. If surface traction is prescribed over the whole
(1) (2)
boundary S, then ui and ui may differ by a rigid-body displacement.
It should be noted that this uniqueness proof applies to the linear case only.
2005
c by Ajit Mal
3.4. BETTI-RAYLEIGH RECIPROCITY (B-R-R) RELATIONS 71
If a solid is in equilibrium under the action of a given system of body and surface forces, then the strain energy of
deformation is equal to one-half the static work that would be done by the external forces.
If an elastic body is subject to two systems of body and surface forces, 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 is known as the Betti-Rayleigh
reciprocity (B-R-R) theorem, and it can be expressed in a number of alternative integral relations between the various
field variables.
(1)
To derive these relations, consider two equilibrium states of an elastic solid: one with displacement ui due to
(1) (1) (2) (2)
body force fi and surface forces ti and the other with displacement ui due to body force fi and surface force
(2) (1) (1) (2)
ti . Then, the work Ω that would be done by the forces fi , ti if they acted through the displacement ui is given
by
Z Z
(1) (2) (1) (2)
Ω = fi ui dV + ti ui dS
V S
Z Z
(1) (2) (1) (2)
= fi ui dV + σij nj ui dS
V S
Z Z
(1) (2) (1) (2)
= fi ui dV + (σij ui ),j dV (by divergence theorem)
V V
Z Z
(1) (1) (2) (1) (2)
= {fi + σij,j }ui dV + σij ui,j dV (3.28)
V V
Z
(1) (2)
= σij ui,j dV (using equilibrium equations)
V
1 (1) (2)
Z
(2)
= σij {ui,j + uj,i } dV (since σij = σji )
V 2
Z Z
(1) (2) (1) (2)
= σij ǫij dV = cijkm ǫkm ǫij dV
V V
(1) (2) (1) (2)
But, since cijkl = cklij , the expression cijkl ǫkl ǫij is symmetric in ǫij and ǫij . Therefore,
Z Z Z Z
(1) (2) (1) (2) (2) (1) (2) (1)
fi ui dV + ti ui dS = fi ui dV + ti ui dS (3.29a)
V S V S
which proves the desired result. Other useful forms of (3.29a) are
Z Z
(1) (2) (2) (1)
σij ǫij dV = σij ǫij dV (3.29b)
V V
Z Z Z
(1) (2) (2) (1) (2) (1)
σij ǫij dV = fi ui dV + ti ui dS (3.29c)
V V S
(1) (2)
If we assume that ui = ui = ui in (3.29c), it reduces to Clapeyron’s theorem (3.27).
2005
c by Ajit Mal
72 CHAPTER 3. LINEAR ELASTOSTATICS
But,
Z
ǫii dV = ∆V (3.32)
V
where ∆V denotes the increase in volume of the body corresponding to the displacement ui . Thus, the increase
in volume of an elastic body caused by the body force fi and surface traction ti is given by
Z
1
Z
∆V = fi xi dV + ti xi dS (3.33)
3k V S
2. Let
(1) (1) (1) (1)
ui = Ax1 , u2 = u3 = 0, fi =0 (3.34a)
Then
(1) (1) (1) (1)
ǫ11 = A, σ11 = (λ + 2µ)A, σ22 = σ33 = λA (3.34b)
the strain and stress components not listed being zero. The Betti-Rayleigh relation in the form (3.29b) yields
Z Z Z
(λ + 2µ)ǫ11 dV + λ(ǫ22 + ǫ33 ) dV = σ11 dV
V V V
or,
1 (λ + 2µ) λ
Z Z Z
σ11 dV = u1,1 dV + (u2,2 + u3,3 ) dV
V V V V V
ZV
(λ + 2µ) λ
Z
= u1 n1 dS + (u2 n2 + u3 n3 ) dS (3.35)
V S V S
by Problem 2.16. Similar expressions can be obtained for the average values of the stress components σ22 and
σ33 over V .
For some problems in elastostatics, Navier’s equations can be integrated directly. The integration constants can then
be found from the boundary conditions. In this section we illustrate this method of solving elastostatic problems with
the help of a few examples.
2005
c by Ajit Mal
3.5. EXACT SOLUTION OF SOME LINEAR ELASTIC PROBLEMS 73
Consider a cylindrical tube of internal radius a and external radius b, subjected to a uniform internal pressure p and
external pressure q only (Fig. 3-3). Assume that there is no deformation along the axial direction. Then working in
terms of cylindrical coordinates (r, φ, z), we can assume that the displacement field in the cylinder is given by
ur = u(r), u φ = uz = 0 (i)
The nontrivial boundary conditions on the curved surface of the cylinder are
σrr = −p at r = a; σrr = −q at r = b (ii)
The Navier equation (3.21b) becomes
d du u
+ =0 (iii)
dr dr r
whose general solution is
B
u = Ar + (iv)
r
where A and B are constants. Therefore, using (3.16) and (3.17),
2µB
σrr = 2(λ + µ)A −
r2
2µB
σφφ = 2(λ + µ)A + 2 (v)
r
σzz = ν(σrr + σφφ )
σφz = σzr = σrφ = 0
The constants A and B are obtained from the boundary conditions as
pa2 − qb2 a2 b2 (p − q)
A= , B= (vi)
2(λ + µ)(b2 − a2 ) 2µ(b2 − a2 )
Thus
(pa2 − qb2 )r a2 b2 (p − q)
ur = + (viia)
2(λ + µ)(b − a ) 2µ(b2 − a2 )r
2 2
2005
c by Ajit Mal
74 CHAPTER 3. LINEAR ELASTOSTATICS
p a
Similarly, the minimum value of σφφ occurs at the outer boundary, and
2a2 p
(σφφ )min = (x)
b2 − a2
If the thickness of the cylinder, d = b − a, is small, then a, b, and r are nearly equal, and (viii) yields
pa2 b2
σφφ = 1 + 2 ≈ pa/d (xi)
d(a + b) r
Thus, for a thin walled cylinder, σφφ is inversely proportional to the thickness of the tube.
If the outer radius b is much larger than the inner radius a, (vii) may be approximated as
qr (p − q)a2
ur = − + (xii)
2(λ + µ) 2µr
(p − q)a2
σrr = −q − (xiiia)
r2
(p − q)a2
σφφ = −q + (xiiib)
r2
At the inner boundary, r = a,
λ + 2µ a
ur = p− q (xiva)
λ+µ 2µ
σrr = −p, σφφ = p − 2q (xivb)
Thus, if p = 0, then σφφ = −2q, which is twice the stress that would exist without the hole. Finally, if r is large
compared to a, (xii) and (xiii) yield
qr
ur = − , σrr = σφφ = −q (xv)
2(λ + µ)
2005
c by Ajit Mal
3.5. EXACT SOLUTION OF SOME LINEAR ELASTIC PROBLEMS 75
ρω2r
ω
r
Fig. 3-4. A rotating shaft; ρω 2 r is the force of inertia per unit volume.
Consider a solid circular cylinder of radius a rotating with uniform angular velocity ω about its axis. We assume
that the cylinder is not free to deform longitudinally. The differential equation of motion of a rotating shaft can be ob-
tained from the equation of equilibrium by invoking D’Alembert’s principle and taking the force of inertia fr = ρω 2 r
as the body force in the radial direction (Fig. 3-4). Therefore, assuming
ur = u(r), u φ = uz = 0
Π
σrr = 2(λ + µ)A − (2λ + 3µ)r2 (iva)
4
Π
σφφ = 2(λ + µ)A − (2λ + µ)r2 (ivb)
4
Π 2
σzz = 2λA − λr (ivc)
2
If the surface of the shaft is traction-free, the boundary condition is
σrr = 0 at r = a
2005
c by Ajit Mal
76 CHAPTER 3. LINEAR ELASTOSTATICS
This yields
(2λ + 3µ)Πa2
A= (vi)
8(λ + µ)
We therefore have
Πr 2λ + 3µ 2
ur = a − r2 (vii)
8 λ+µ
Π
σrr = (2λ + 3µ)(a2 − r2 ) (viiia)
4
Π 2λ + 3µ 2
σφφ = (2λ + µ) a − r2 (viiib)
4 2λ + µ
Πλ 2λ + 3µ 2
σzz = a − r2 (viiic)
2 2(λ + µ)
On the surface r = a,
Πa3 (λ + 2µ)
ur = (ix)
8(λ + µ)
1 2
σφφ = Πa µ (xa)
2
Πa2 λµ
σzz = (xb)
4(λ + µ)
The maximum stress occurs at the center (r = 0) and is
Πa2
σrr = σφφ = (2λ + 3µ) (xi)
4
In the case of a cylindrical shell, B =
6 0. Assuming traction-free boundary conditions, it can be shown that
Π(2λ + 3µ) 2 Π(2λ + 3µ) 2 2
A= (a + b2 ), B= a b (xii)
8(λ + µ) 8µ
where b is the inner radius.
Assume that a cylinder (an inclusion) of infinite length, and radius a and composed of an elastic material is em-
bedded in another unbounded elastic material (the matrix) (Fig. 3-5). The interface between the two solids is assumed
to be perfectly bonded, or welded, and the composite is subjected to a hydrostatic pressure σ0 at infinity. The elastic
constants of the matrix are λ1 , µ1 and those of the inclusion are λ2 , µ2 . The objective is to determine the stress and
displacement components in the solid.
We assume that plane strain condition is valid and that the stress and displacement fields are cylindrically symmet-
ric for this problem. Then the nonzero displacement and stress components in the two materials are of the forms (see
Example 3.5-1)
Bi
u(i) (r) = Ai r + (i)
r
(i) 2µi Bi
σrr (r) = 2(λi + µi )Ai − (iia)
r2
(i) 2µi Bi
σφφ (r) = 2(λi + µi )Ai + (iib)
r2
(i) (i) (i)
σzz (r) = νi (σrr + σφφ ) = 4νi (λi + µi )Ai (iic)
2005
c by Ajit Mal
3.5. EXACT SOLUTION OF SOME LINEAR ELASTIC PROBLEMS 77
σo λ 2,µ 2 o σ0
r Φ
x
λ 1,µ 1
2a
where i = 1 for r > a and i = 2 for r < a. There is no sum on i and Ai , Bi are unknown constants.
Since u(2) (r) must be finite at all points within the inclusion, we must have
B2 = 0 (iii)
(1)
Also, since σrr → σ0 for r → ∞,
At the interface between the two solids, the displacement ur and the stress σrr must be continuous. Using (i)-(iv), this
gives
σ0 a B1
+ = A2 a (v)
2(λ1 + µ1 ) a
2µ1 B1
σ0 − = 2(λ2 + µ2 )A2 (vi)
a2
Solving for A1 , A2 , B1 from (iv), (v), and (vi),
σ0 (λ1 + 2µ1 ) σ0
A2 = , A1 = (vii)
2(λ1 + µ1 )(λ2 + µ2 + µ1 ) 2(λ1 + µ1 )
σ0 a2 (λ1 + µ1 − λ2 − µ2 )
B1 = (viii)
2(λ1 + µ1 )(λ2 + µ2 + µ1 )
The displacement and stress components can be obtained from (i) and (ii), on substitution of the constants. In particu-
lar, the interfacial stresses are given by
(1) (2)
σrr (a) = 2(λ2 + µ2 )A2 = σrr (a) (ixa)
(1) 2µ1 B1 (2)
σφφ (a) = 2(λ1 + µ1 )A1 + , σφφ (a) = 2(λ2 + µ2 )A2 (ixb)
a2
(i)
σzz (a) = 4νi (λi + µi )Ai , i = 1, 2 (ixc)
Consider a cylindrical tube of internal radius a and external radius b, with fixed ends. Assuming axially symmet-
ric temperature distribution Θ(r), we have
ur = u(r), uφ = uz = 0, f =0 (i)
2005
c by Ajit Mal
78 CHAPTER 3. LINEAR ELASTOSTATICS
The stresses can be obtained by using (3.13) and (3.17). Noting that on account of the assumption (i),
du u
ǫrr = , ǫφφ = , ǫzz = ǫφz = ǫzr = ǫrφ = 0 (iv)
dr r
we have
Z r
E A B αE
σrr = − 2 − Θ(r)r dr (va)
1 + ν 1 − 2ν r (1 − ν)r2 a
Z r
E A B αE αEΘ
σφφ = + 2 + 2
Θ(r)r dr − (vb)
1 + ν 1 − 2ν r (1 − ν)r a 1−ν
Assuming the walls of the cylinder to be traction-free, the boundary conditions are
Thus,
b
A B A B 1+ν α
Z
− 2 =0 − 2 = Θ(r)r dr
1 − 2ν a 1 − 2ν b 1 − ν b2 a
Solving for A and B and substituting the results in (v), we obtain
" #
r 2 − a2 b
Z r
αE
Z
σrr = Θr dr − Θr dr
(1 − ν)r2 b2 − aa a a
" #
r 2 + a2 b
Z r
αE
Z
2
σφφ = Θr dr + Θr dr − r Θ (viii)
(1 − ν)r2 b2 − a2 a a
" Z b #
αE 2ν
σzz = Θr dr − Θ
1 − ν b2 − a a a
Clearly, σzz is the normal stress distribution that must be applied to the ends of the cylinder to keep ǫzz = 0 throughout.
The displacement field is
Z b
a2 α(1 + ν) r
α(1 + ν)
Z
u= (1 − 2ν)r + Θr dr + Θr dr (ix)
(1 − ν)(b2 − a2 ) r a (1 − ν)r a
The corresponding results for a solid circular cylinder of radius b can be obtained from (viii) and (ix) with a = 0.
Consider a spherical shell of inner radius a and outer radius b, subjected to pressures p and q at the inner and
outer surfaces, respectively. The boundary conditions are
2005
c by Ajit Mal
3.5. EXACT SOLUTION OF SOME LINEAR ELASTIC PROBLEMS 79
uR = U (R), u θ = uφ = 0 (ii)
Integrating, we get
B
U = AR + , a<R<b (iv)
R2
where A and B are arbitrary constants. From (3.19), (3.20), and (iv),
dU 2λ 4µB
σRR = (λ + 2µ) + U = (3λ + 2µ)A − 3 (va)
dR R R
2µB
σθθ = σφφ = (3λ + 2µ)A + , σθφ = σφR = σRθ = 0 (vb)
R3
Equations (i) and (v) yield
We thus have
(pa3 − qb3 )R a3 b3 (p − q)
uR = + (vii)
(3λ + 2µ)(b3 − a3 ) 4µ(b3 − a3 )R2
pa3 b3
4µ
uR = R + (ix)
4µ(b3 − a3 ) 3λ + 2µ R2
pa3 b3 pa3 b3
σRR = − − 1 ≤ 0, σθθ = σφφ = 1+ >0 (x)
b − a3
3 R3 b − a3
3 2R3
2005
c by Ajit Mal
80 CHAPTER 3. LINEAR ELASTOSTATICS
pa3 b3
(σθθ )max = (σφφ )max = 1+ (xi)
b − a3
3 2a3
at the inner surface R = a.
If the shell is of small thickness d = b − a ≪ a, (x) yields
pa3 b3
pa
σθθ = σφφ = 2 1+ ≈ (xii)
(a + ab + b2 )d 2R3 2d
qR (p − q)a3
uR = − + (xiiia)
3λ + 2µ 4µR2
a3
σRR = −q − (p − q) 3 (xiiib)
R
a3
σθθ = σφφ = −q + (p − q) 3 (xiiic)
2R
At R = a,
p−q q
uR = − a (xiv)
4µ 3λ + 2µ
1
σRR = −p, σθθ = σφφ = (p − 3q) (xv)
2
3
Therefore, if p = 0, σθθ = σφφ = −3q/2, which is 2 times the pressure that would exist if there were no spherical
cavity. If R is large compared to a, (xiii) gives
qR
uR = − , σRR = σθθ = σφφ = −q (xvi)
3λ + 2µ
We consider the problem of the straining of an elastic sphere due to mutual attraction of its parts. The force of
gravity at any point inside the sphere is directly proportional to the distance of the point from the center and is directed
towards the center. Therefore,
fR = −CR, fθ = f φ = 0 (i)
d dU 2U
(λ + 2µ) + − CR = 0 (ii)
dR dR R
Integrating and using the fact that U must be finite at R = 0,
CR3
U = AR + (iii)
10(λ + 2µ)
This yields
(5λ + 6µ)CR2
σRR = (3λ + 2µ)A +
10(λ + 2µ)
(5λ + 2µ)CR2
σθθ = σφφ = (3λ + 2µ)A + (iv)
10(λ + 2µ)
Assuming the surface of the solid sphere of radius a to be traction-free, the boundary condition σRR = 0 at R = a
gives
2005
c by Ajit Mal
3.5. EXACT SOLUTION OF SOME LINEAR ELASTIC PROBLEMS 81
Consider a hollow sphere a < R < b in which the temperature distribution is symmetrical with respect to the center
and is therefore a function of R alone. We assume that
uR = U (R), u θ = uφ = 0 (i)
The Navier equation (3.22b) now takes the form
d dU 2U (1 + ν)α dΘ
+ =
dR dR R 1 − ν dR
Integrating this equation, we find
R
B (1 + ν)α
Z
U = AR + + ΘR2 dR (ii)
R2 (1 − ν)R2 a
where a is a constant taken equal to zero for a solid sphere and equal to the inner radius for a hollow sphere.
The stresses can be obtained by using (3.19) and (3.20). Noting that on account of the assumption (i),
dU U
ǫRR = , ǫθθ = ǫφφ =
dR R
ǫθφ = ǫφR = ǫRθ = 0
we obtain
R
EA 2EB 2αE
Z
σRR = − − ΘR2 dR
1 − 2ν (1 + ν)R3 (1 − ν)R3 a
R
EA EB αE αEΘ
Z
σθθ = σφφ = + + ΘR2 dR − (iii)
1 − 2ν (1 + ν)R3 (1 − ν)R3 a 1−ν
σθφ = σφR = σRθ = 0
2005
c by Ajit Mal
82 CHAPTER 3. LINEAR ELASTOSTATICS
If the walls of the sphere are traction-free, the boundary conditions are
As a special case of the preceding, let Θ0 be the temperature at the inner surface and let the temperature at the
outer surface be zero. Then at a distance R from the center,
Θ0 a b
Θ= −1 (vii)
b−a R
Substituting this value of Θ in (vi), we find
a 2 b2
αEΘ0 ab 1 2 2
σRR = a+b− (a + ab + b ) + 3
(1 − ν)(b3 − a3 ) R R
a 2 b2
αEΘ0 ab 1 2 2
σθθ = a+b− (a + ab + b ) + (viii)
(1 − ν)(b3 − a3 ) 2R 2R3
We see that σRR is zero for R = a and R = b. For Θ0 > 0 , the stress σθθ increases as R increases. At R = a,
αEΘ0 b(a + 2b)
σθθ = −
2(1 − ν)(a2 + ab + b2 )
while, at R = b,
αEΘ0 a(2a + b)
σθθ = (ix)
2(1 − ν)(a2 + ab + b2 )
2005
c by Ajit Mal
3.6. EQUATIONS OF COMPATIBILITY FOR INFINITESIMAL STRAINS 83
Equation (3.41) gives a set of necessary conditions for the existence of continuous single-valued displacements.
We next show that these conditions are not only necessary but sufficient as well.
Let us assume that the strain components ǫij are known and the displacements u0j and the rotations ωij 0
are known
at a given point P 0 (x0 , t). We express the displacement uj at an arbitrary point P ′ (x′ , t) as a line integral along a
simple continuous curve joining the points P 0 and P ′ in the form
Z P′ Z P′
′
uj (x , t) = u0j + duj = u0j + uj,k dxk
P0 P0
Z P′
= u0j + (ǫjk + ωjk )dxk (3.42)
P0
1
where ωij = 2 (uj,k − uk,j ). Integrating the second term by parts, we get
Z P′ Z P′
ωjk dxk = ωjk d(xk − x′k )
P0 P0
Z P′
= (x′k − x0k )ωjk
0
− (xk − x′k )ωjk,l dxl
P0
Z P′
= (x′k − x0k )ωjk
0
− (xk − x′k )(ǫlj,k − ǫlk,j )dxl
P0
the first step following from the fact that, since the point P ′ is fixed with respect to the integration, dx′k = 0.
Equation (3.42) now becomes
Z P′
′
uj (x , t) = u0j + (x′k − x0k )ωjk
0
+ [ǫjl + (x′k − xk )(ǫlj,k − ǫlk,j )]dxl (3.43)
P0
Since the displacement uj must be single-valued, the integral in (3.43) must be independent of the path of integra-
tion. Therefore,
Z Z
Ujl dxl = Ujl dxl
P 0 AP ′ P 0 BP ′
where
Ujl = ǫjl + (x′k − xk )(ǫlj,k − ǫlk,j ) (3.44)
0 ′ 0 ′ 0 ′
and P AP and P BP are two arbitrary nonintersecting paths connecting P and P (Fig. 3-6). We thus have
Z
Ujl dxl = 0 (3.45a)
C
where C is the closed contour P 0 BP ′ AP 0 . If the region under consideration is simply connected, then Stokes’
2005
c by Ajit Mal
84 CHAPTER 3. LINEAR ELASTOSTATICS
x2 B
o x1
Fig. 3-6. The integration contour for (3.45).
where S is an arbitrary open surface bounded by C, and nm are the direction cosines of the normal to S at dS.
Using (3.45a) and noting that the surface of integration S is arbitrary, (3.45b) holds if
i.e.,
Since this equation must hold for an arbitrary (x′k − xk ), it follows that
This is identical to (3.41), which therefore expresses the necessary and sufficient conditions for the existence of a
unique solution of (3.37) for given ǫij . Some of the equations are identically satisfied (e.g., for i = l or j = k), and
some are repeated because of the symmetry in indices (e.g., for i = j = 2, k = l = 3 and i = j = 3, k = l = 2).
In fact, only 6 of the 81 equations in (3.41) are independent. These six equations correspond to the cases for which
2005
c by Ajit Mal
3.7. ANISOTROPIC MATERIALS 85
Equation (3.46) is the necessary and sufficient condition for the single-valuedness of the integral in (3.43) if the
region is simply connected. For multiply connected regions, both equations (3.41) and (3.45a) must be satisfied by the
strain field. An excellent discussion of these can be found in many classical texts (e.g., Fung 1965) and so is omitted
here.
2005
c by Ajit Mal
86 CHAPTER 3. LINEAR ELASTOSTATICS
Most solids exhibit symmetry properties with respect to certain rotations of the body or reflection about one or
more planes. The effect of these symmetries is to further reduce the number of elastic constants. It is obvious that
the elastic constants cij , in general, depend upon the chosen reference frame, since the stress components σi and the
strain components ǫi vary with the choice of the coordinate system. For certain solids the elastic constants cij may
remain invariant under a given transformation of coordinates. It is this invariance property that determines the elastic
symmetry of the solid under consideration.
Let the plane of symmetry be taken as the x1 x2 -plane. Then cij must be invariant under the transformation
x′1 = x1 , x′2 = x2 , x′3 = −x3
Under this coordinate transformation,
ǫ′23 = −ǫ23 , ǫ′31 = −ǫ31
all other components remain unchanged. Thus
ǫ′4 = −ǫ4 , ǫ′5 = −ǫ5 , ǫ′i = ǫi , i = 1, 2, 3, 6
The strain energy function W ′ is obtained from the right-hand side of (3.52) on changing the signs of ǫ4 and ǫ5 . But,
since the material is symmetric about the x1 x2 -plane, W ′ = W . Hence
c14 = c15 = c24 = c25 = c34 = c35 = c46 = c56 = 0 (3.53)
Thus, for a material with one plane of elastic symmetry, there are 21 − 8 = 13 independent elastic constants. Such
materials are called monoclinic.
For an elastic material having x1 x2 -plane as a plane of symmetry, the generalized Hooke’s law may be expressed
in the form
σ1 c11 c12 c13 0 0 c16 ǫ1
σ2 c12 c22 c23 0 0 c26 ǫ2
σ3 c13 c23 c33 0 0 c36 ǫ3
σ4 = 0 (3.54)
0 0 c44 c45 0 ǫ4
σ5 0 0 0 c45 c55 0 ǫ5
σ6 c16 c26 c36 0 0 c66 ǫ6
If a material has three mutually orthogonal planes of elastic symmetry, it is called orthotropic. Wood is a common
example of an orthotropic material. Let the planes of symmetry be taken as the coordinate planes. Then, on account
of the symmetry about the x1 x2 -plane, the relations (3.53) hold. Similarly, on account of the symmetry about the
x2 x3 -plane,
c16 = c26 = c36 = c45 = 0 (3.55)
and the number of independent elastic constants reduces to 9.
It can be easily seen that if x1 x2 - and x2 x3 -planes are planes of symmetry, then the x1 x3 -plane is also a plane
of symmetry. Thus, no further reduction in the number of independent elastic constants is achieved by considering
symmetry in the x1 x3 -plane. The generalized Hooke’s law for an orthotropic hyperelastic material can therefore be
expressed in the form
σ1 c11 c12 c13 0 0 0 ǫ1
σ2 c12 c22 c23 0 0 0
ǫ2
σ3 c13 c23 c33 0 0 0 ǫ3
σ4 = 0 (3.56)
0 0 c44 0 0
ǫ4
σ5 0 0 0 0 c55 0 ǫ5
σ6 0 0 0 0 0 c66 ǫ6
2005
c by Ajit Mal
3.7. ANISOTROPIC MATERIALS 87
If an orthotropic solid exhibits symmetry with respect to arbitrary rotations about one of the axes (such as the x3 -axis),
then it is called transversely isotropic. Let the system Ox′1 x′2 x′3 be obtained from the system Ox1 x2 x3 by a rotation
about the x3 -axis through an angle θ, so that
{x′i } = [L]{xi }
where
cos θ sin θ 0
L= − sin θ cos θ 0 (3.57)
0 0 1
Then cij must be invariant under the transformation (3.57). In the new coordinate system, the strain components are
given by
where,
The strain energy function W ′ is obtained from the right-hand side of (3.52) on replacing ǫi by ǫ′i given in (3.58).
But since the solid is symmetric with respect to arbitrary rotations about the x3 -axis, W ′ = W for arbitrary θ. Equating
the coefficients of ǫ21 , we obtain
2005
c by Ajit Mal
88 CHAPTER 3. LINEAR ELASTOSTATICS
i.e.,
Equations (3.59) imply that a transversely isotropic elastic solid has 9 − 4 = 5 independent elastic constants. The
generalized Hooke’s law (3.56) now takes the form
σ1 c11 c12 c13 0 0 0 ǫ1
σ2 c12 c11 c13 0 0 0 ǫ2
σ3 c13 c13 c33 0 0 0 ǫ3
= (3.60)
σ4 0 0 0 c44 0 0 ǫ4
σ5 0 0 0 0 c44 0 ǫ5
σ6 0 0 0 0 0 c66 ǫ6
where
c11 − c12
c66 = (3.61)
2
It should be noted that the preceding stress-strain equations are valid only if the symmetry axis is parallel to the
x3 -axis. If the symmetry-axis is inclined at an arbitrary angle to the coordinate axes, these equations become much
more complex. Several such cases are discussed in the following sections.
For a cubic crystal with its symmetry axes along the coordinate axes,
1 1
c11 ǫ211 + ǫ222 + ǫ233 + c12 (ǫ11 ǫ22 + ǫ22 ǫ33 + ǫ33 ǫ11 ) + c44 ǫ212 + ǫ223 + ǫ213
W =
2 2
σ11 = c11 ǫ11 + c12 (ǫ22 + ǫ33 )
σ22 = c12 (ǫ11 + ǫ33 ) + c11 ǫ22
σ33 = c12 (ǫ11 + ǫ22 ) + c11 ǫ33
σij = c44 ǫij , i 6= j
In the case of an isotropic material, the elastic constants cij are independent of the orientation of the coordinate axes.
We begin with the stress-strain relations (3.56) for an orthotropic elastic solid. The symmetry about the x3 -axis yields
the relations (3.59-3.61). Similarly, the symmetry about the x1 -axis implies
1
c22 = c33 , c44 = (c22 − c23 ), c12 = c13 , c55 = c66 (3.62)
2
2005
c by Ajit Mal
3.8. CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES 89
Thus
c11 = c22 = c33
c12 = c13 = c23 = λ (3.63a)
1
c44 = c55 = c66 = (c11 − c12 ) = µ
2
Where λ and µ are constants. Then
c11 = c22 = c33 = λ + 2µ (3.63b)
We note that the number of independent elastic constants for an isotropic material is just two. Inserting the values
of the elastic constants cij from (3.63) into (3.48) and reverting to the double-suffix notation, we arrive at the Hooke’s
law for an isotropic elastic material in the form
σij = λǫkk δij + 2µǫij (3.64a)
which coincides with (3.10). It can be further shown that the elastic tensor c can be expressed in the form
cijkm = λδij δkm + µ(δik δjm + δim δjk ) (3.64b)
The strain energy function can be expressed as
1
W = σij ǫij
2
1 2 1
= λd + µǫij ǫij = λIǫ 2 + µ(Iǫ 2 − 2πǫ ) (3.65)
2 2
1 2 λ
= λd + µ(ǫ211 + ǫ222 + ǫ233 + 2ǫ223 + 2ǫ231 + 2ǫ212 ) = ( + µ)Iǫ 2 − 2µπǫ
2 2
where
d = ǫkk
1
IIǫ = (ǫii ǫjj − ǫij ǫij ) (3.66)
2
ǫij ǫij = Iǫ 2 − 2IIǫ
Fiber-reinforced composites are used in a variety of structures due to their low weight and high strength. A brief
discussion of the linear stress-strain equations for a special class of these materials is presented in this section. The
material is assumed to contain a single system of infinitely long, parallel fibers embedded in a matrix. The diameter
of the fibers is small compared to all other length dimensions of the body under consideration, so that the overall, or
macroscopic, behavior of the material can be adequately described as transversely isotropic, with symmetry axis par-
allel to the fibers. Then, as shown in Sec. 3.7.3, the overall linear elastic behavior of the material can be characterized
by means of five elastic constants. The various possible forms of the stress-strain law for these materials containing
alternate sets of constants are discussed here.
For a transversely isotropic material with its symmetry axis directed along one of the coordinate axes, the linear
stress-strain law can be expressed in the general form given in (3.60). In the more general case, when the symmetry
axis is in an arbitrary direction, say a, as shown in Fig. 3-7, the equations can be obtained in a number of alternative
ways. The details of these can be found in several books and monographs (e.g., Christensen 1979, Spencer 1984). We
present the result only. The elastic tensor c can be expressed as in the form
cijkl = λδij δkl + µT (δik δjl + δjk δil ) + α(ak al δij + ai aj δkl )
+ (µL − µT )(ai ak δjl + ai al δjk + aj ak δil + aj al δik ) (3.67a)
+ βai aj ak al
2005
c by Ajit Mal
90 CHAPTER 3. LINEAR ELASTOSTATICS
3 o 1
Fig. 3-7. Unidirectional graphite-epoxy compsite with fibers along a.
where λ, µL , µT , α, and β are elastic constants. Note that the constants α, β, and µL − µT describe the degree
of anisotropy of the material. The isotropic case is obtained by setting α = β = µL − µT = 0. The constants
λ, µL , µT , α, and β may be thought of as the generalized Lamé constants of the solid. The stress-strain law can be
written in the form
If we choose the fiber direction as (1, 0, 0), then the stress-strain law can also be written in terms of the stiffness
matrix [cij ] in the form
σ11 c11 c12 c12 0 0 0 ǫ11
σ22 c12 c22 c23 0 0 0 ǫ22
σ33 c12 c23 c22 0 0 0 ǫ33
σ23 = 0 (3.68)
0 0 c44 0 0
2ǫ23
σ31 0 0 0 0 c55 0 2ǫ31
σ12 0 0 0 0 0 c55 2ǫ12
with
c22 − c23
c44 =
2
From (3.67) and (3.68) the stiffness constants can be related to the Lamé constants by
The engineering constants of the material will be denoted by the extension, or Young’s, moduli EL , ET , the shear
moduli µL , µT , and the Poisson’s ratios νLT , νT L , νT T = νT , where the subscripts L and T indicate properties along
and at right angles to the fibers. The extension moduli EL and ET are related to the Lamé constants through
2005
c by Ajit Mal
3.8. CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES 91
The stiffness constants cij can be shown to be related to the engineering constants through
From (3.71)
The generalized Lamé constants α, β, λ are obtained from (3.68) and (3.69b) as
It should be noted that for general orientation of the fibers, the stress-strain equation is far more complicated than
that for the case when the fiber orientation coincides with one of the coordinate axes. Unfortunately, the coordinate
system cannot always be chosen arbitrarily, since structural composites often consist of a number of layers of uniaxial
laminae with different orientations. Analysis of such problems are beyond the scope of this book. Interested readers
are referred to Christensen (1979) for details.
2005
c by Ajit Mal
92 CHAPTER 3. LINEAR ELASTOSTATICS
PROBLEMS
3.1 Show that in the linear theory of deformation, the decrease in the angle between two orthogonal unit vectors a and
b is 2ai bj ǫij .
3.2. Consider a linear strain field associated with a simply connected region such that
Find the relationship between A and B such that it is possible to obtain a single-valued continuous displacement
field which corresponds to the given strain field including RBD.
3.3 The Cauchy stress tensor in a deformed solid is given by
where A is a constant.
(a) Examine whether there are any body forces within the slab.
(b) Calculate the resultant forces acting on the faces x1 = ±a, x2 = ±a.
3.5. The Cauchy stress tensor in a domain −∞ < x1 , x2 < ∞, x3 > 0 is given by
αxi xj x3
σij =
R5
where R = (x21 + x22 + x23 )1/2 and α is a constant. Examine whether there are any body forces in the domain.
Calculate the resultant force on the surface of the hemisphere R = a, x3 ≥ 0.
3.6. The Cauchy stress tensor at a point in a solid is given by the matrix
T 2 1
2 0 2
1 2 0
where T is a constant. Calculate T so that there will be a traction-free plane through the point and determine
the orientation of this plane.
2005
c by Ajit Mal
3.8. CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES 93
3.7. Show that the stationary values of the normal stress at a point correspond to the principal values of stress at that
point.
3.8. The Cauchy stress tensor at a point in a solid is given by the matrix
3 1 1
1 0 2
1 2 0
in arbitrary units.
(a) Calculate the x, y, z components and the direction and the magnitude of the stress vector on a surface
element with normal (0, 1, 1).
(b) Show directly that the principal directions of stress are mutually orthogonal. Also show that relative to the
coordinate system 0x′1 x′2 x′3 , where 0x′i is parallel to a principal direction, the stress tensor becomes
−2 0 0
0 1 0
0 0 4
3.14. Show that the linear isotropic stress-strain relations may be solved for the strains in terms of the stresses in the
form
−λ 1
ǫij = sδij + σij , s = σkk , µ 6= 0, 3λ + 2µ 6= 0
2µ(3λ + 2µ) 2µ
2005
c by Ajit Mal
94 CHAPTER 3. LINEAR ELASTOSTATICS
3.15. Show that the Hooke’s law for an isotropic solid may be expressed in an equivalent form
(d) (d)
σij = 2µǫij , s = (3λ + 2µ)d, s = σkk , d = ǫkk
(d)
where σij are the components of the deviatoric stress tensor:
(d) 1
σij = σij − sδij
3
(d)
and ǫij are the components of the deviatoric strain tensor:
(d) 1
ǫij = ǫij − dδij
3
3.16. Show that the strain energy function for an isotropic material may be expressed in the forms:
1
W = σij ǫij
2
1
= µǫij ǫij + (3k − 2µ)d2
6
1 1 1 1
= σij σij + − s2
4µ 6 3k 2µ
1 (d) (d) 1
= σ ǫ + sd
2 ij ij 6
(d) (d) 1 2
= µǫij ǫij + kd
2
1 (d) (d) 1 2
= σ σ + s
4µ ij ij 18k
3.17. The linear stress-strain law for a fiber reinforced composite with parallel fibers aligned with the unit vector a is
given in Sec. 3.8. Assuming that a = (cos θ, sin θ, 0), obtain the stiffness matrix [c′ij ] in terms of cij and θ.
Also determine the stress components necessary to produce the uniaxial strain ǫij = δi1 δj1 in the material.
3.18. In a rectangular block of a fiber-reinforced composite aligned with the coordinate planes, all fibers are parallel to
the x1 x2 -plane and inclined to the x1 -axis at angle θ. The block is subjected to the inplane stresses σ11 , σ12 , σ22 .
Calculate the strain components.
(λ + 2µ)∇∇ · u − µ∇ × ∇ × u + f = 0
and putting
1
∆ = ∇ · u, ω= ∇×u
2
show that the equilibrium equations in cylindrical coordinates can be expressed in the form
∂∆ 1 ∂ωz ∂ωφ
(λ + 2µ) − 2µ − + fr = 0
∂r r ∂φ ∂z
1 ∂∆ ∂ωr ∂ωz
(λ + 2µ) − 2µ − + fφ = 0
r ∂φ ∂z ∂r
∂∆ ∂ωφ ωφ 1 ∂ωr
(λ + 2µ) − 2µ + − + fz = 0
∂z ∂r r r ∂φ
2005
c by Ajit Mal
3.8. CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES 95
3.20. Show that the equilibrium equations in spherical coordinates can be expressed in the form
∂∆ 2µ ∂ωφ 1 ∂ωθ
(λ + 2µ) − + (cot θ)ωφ − + fR = 0
∂R R ∂θ sin θ ∂φ
1 ∂∆ 1 ∂ωR ∂ωφ ωφ
(λ + 2µ) − 2µ − − + fθ = 0
R ∂θ R sin θ ∂φ ∂R R
1 ∂∆ ∂ωθ ωθ 1 ∂ωR
(λ + 2µ) − 2µ + − + fφ = 0
R sin φ ∂φ ∂R R R ∂θ
3.22. (a) Show that the Navier’s equation (3.22b)for an isotropic elastic solid in the absence of body forces and steady
state temp for θ is satisfied provided the dilatation d is a solution of ∇2 d = 0.
(b) Let
3λ + 2µ
u= α∇Ω
λ + 2µ
∇(∇2 Ω − Θ) = 0
where f is the body force per unit volume and t is the surface force per unit area.
3.24. Let a beam be acted upon by point loads P1 , P2 ,... applied at the points x1 , x2 ,... and denote by αij the transverse
displacement at xi due to a unit transverse force applied at xj . Assuming that the body forces vanish, apply the
B-R-R relations to show that the influence coefficients αij are symmetric, i.e.,
αij = αji
3.25. If a constant pressure p acts on the boundary of a cylindrical hole of radius a in an infinite medium, show that
pa2 pa2
ur = , σrr = −σφφ =
2µr r2
where r is the distance from the axis of the cylinder and φ is the azimuthal angle.
2005
c by Ajit Mal
96 CHAPTER 3. LINEAR ELASTOSTATICS
3.26. An elastic solid occupies the region a ≤ r ≤ b, where r2 = x2 + y 2 . The surface r = a is fixed and the surface
r = b is subjected to a uniform pressure q. Show that the displacement in the material is given by
−qb2 a2
ur = r−
2(λ + µ)b2 + 2µa2 r
3.27. If a uniform radial displacement ur = −u0 is specified on the boundary of a solid circular cylinder of radius a,
show that
r 2(λ + µ)u0
ur = −u0 , σrr = σφφ = −
a a
3.28. Prove that for a hollow circular cylinder of inner radius a and outer radius b, rotating with uniform angular
velocity ω about its axis, the cylindrical components of stress are
a 2 b2
σrr = A a2 + b2 − 2 − r2
r
a 2 b2
1 + 2ν 2
σφφ = A a2 + b2 + 2 − r
r 3 − 2ν
2
2r
σzz = 2νA a2 + b2 −
3 − 2ν
σφz = σzr = σrφ = 0
where
3 − 2ν
A= ρω 2
8(1 − ν)
Assuming that 0 < ν < 1/2 prove that the maximum stress occurs at the inner boundary and is
1 − 2ν a2
2
σφφ = 2Ab 1 +
3 − 2ν b2
Hence show that the maximum stress is doubled when a solid cylinder has a small hole drilled through its center.
3.29. Calculate the radial displacement u(r) and the stress components σrr , σφφ in a rotating thin disc of radius b
concentrically attached to a rigid shaft of radius a.
3.30. An elastic solid extending to infinity in all directions contains a rigid spherical inclusion of radius a. If the
material is in equilibrium under the action of a uniform pressure q at infinity and body force is absent, show that
the displacement at any point of the material is purely radial and is given by
a3
q
uR = − R− 2
3λ + 2µ R
where R denotes the radial distance from the center of the sphere.
3.31. An elastic solid occupies the region a ≤ r ≤ b, r = (x2 + y 2 )1/2 . The surface r = a is held fixed, whereas
the surface r = b is displaced a distance ǫ in the z-direction. Assuming that body force is absent, show that the
displacement and stress components in the material are given by
ln(r/a) µǫ
uz = ǫ , σrz =
ln(b/a) r ln(b/a)
3.32. Write down the cylindrical and spherical components of the equations of linear elasticity under symmetry con-
ditions.
2005
c by Ajit Mal
3.8. CONSTITUTIVE EQUATIONS FOR FIBER-REINFORCED COMPOSITES 97
BIBLIOGRAPHY
1. Sokolnikoff, I. S., Mathematical Theory of Elasticity (2d ed.), McGraw-Hill, New York, 1956.
2005
c by Ajit Mal
98
2005
c by Ajit Mal
Chapter 4
From (3.4) and (3.8), the stress-displacement relations for an isotropic, linear elastic medium may be expressed in the
form
σij = λul,l δij + µ(ui,j + uj,i ) (4.1)
It is sometimes possible to assume that the displacement components are independent of a single Cartesian coordinate,
say x3 , so that ui = ui (x1 , x2 ). In such a case, (4.1) yields
On substituting the values of σ13 and σ23 from (4.2d) into (4.3c), we get an equation in u3 alone as
∂2 ∂2
µ∇2 u3 + f3 = 0, ∇2 ≡ 2 + (4.4)
∂x1 ∂x22
Similarly, (4.3a, b) yield two simultaneous equations in u1 and u2 that do not involve u3 . The deformation represented
by u3 is known as antiplane strain, whereas the deformation represented by (u1 , u2 ) is known as plane strain. It is
obvious that in both cases the body forces and the stresses must be independent of x3 .
For a solid in antiplane strain with respect to the x1 x2 -plane,
u1 = u2 = 0, u3 = u3 (x1 , x2 ) (4.5a)
σ13 = µu3,1 , σ23 = µu3,2 (4.5b)
99
100 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
Consequently, ǫ13 = ǫ23 = ǫ33 = 0, and ǫ11 , ǫ22 , and ǫ12 are independent of x3 . The stresses are given by
Differentiating (4.3a) with respect to x1 and (4.3b) with respect to x2 and adding, we find
Eliminating σ12 from (4.10) and (4.11), we obtain the compatibility equation in terms of stresses in the form
f1,1 + f2,2
∇2 (σ11 + σ22 ) + =0 (4.12)
1−ν
If the body force is constant, the stress components satisfy
which indicates that (σ11 + σ22 ) is harmonic. Then from (4.7c), (ǫ11 + ǫ22 ) is also harmonic.
It is often convenient (or necessary) to work with the cylindrical coordinates (r, φ, z). Under plane strain condi-
tions, the displacement components are
2005
c by Ajit Mal
4.3. PLANE STRESS 101
Therefore, σrz = σφz = 0 and σrr , σφφ , σzz , and σrφ are independent of z. Hence, from (3.13) the equilibrium
equations reduce to
σrφ,φ σrr − σφφ
σrr,r + + + fr = 0 (4.16a)
r r
σφφ,φ 2σrφ
σrφ,r + + + fφ = 0 (4.16b)
r r
The strain-stress relations are
(1 + ν)[(1 − ν)σrr − νσφφ ]
ǫrr = (4.17a)
E
(1 + ν)[(1 − ν)σφφ − νσrr ]
ǫφφ = (4.17b)
E
(1 + ν)σrφ
ǫrφ = (4.17c)
E
Noting that
and that
fr fφ,φ
f1,1 + f2,2 = div f = fr,r + +
r r
the compatibility equation (4.12) becomes
∂2 1 ∂2
1 ∂ 1 ∂fr fr 1 ∂fφ
+ + (σrr + σφφ ) + + + =0 (4.18)
∂r2 r ∂r r2 ∂φ2 1−ν ∂r r r ∂φ
Physically, plane strain parallel to the x1 x2 -plane is applicable whenever the dimension of the body in the x3 -
direction is large compared to its dimensions in the other directions and when such a body is under the action of
external forces that are perpendicular to the x3 -direction and independent of it. Under these conditions, we can
assume that there is no deformation in the x3 -direction and that the displacement at each crosssection perpendicular
to x3 -axis will not depend upon x3 . Thus the problem becomes truly two-dimensional.
and σ11 , σ22 , and σ12 are independent of x3 . From (3.1) and (4.19), we note that formally the equilibrium equations
for plane stress are the same as Eqs. (4.3a, b) for plane strain. The body forces must be such that f3 = 0 and f1 and
f2 are independent of x3 . If we put σ33 = 0 in the relation
we obtain
λ(ǫ11 + ǫ22 )
ǫ33 = − (4.20)
λ + 2µ
Therefore,
2µ(ǫ11 + ǫ22 )
ǫkk =
λ + 2µ
2005
c by Ajit Mal
102 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
where
2λµ
λ= (4.22)
λ + 2µ
The strains can be expressed in terms of stresses as
σ11 − νσ22
u1,1 = ǫ11 = (4.23a)
E
σ22 − νσ11
u2,2 = ǫ22 = (4.23b)
E
−ν(σ11 + σ22 )
u3,3 = ǫ33 = (4.23c)
E
2(1 + ν)σ12
u1,2 + u2,1 = 2ǫ12 = , ǫ13 = ǫ23 = 0 (4.23d)
E
In cylindrical coordinates,
σrr − νσφφ
ǫrr =
E
σφφ − νσrr
ǫφφ =
E
(1 + ν)σrφ
ǫrφ =
E
E
σrr = (ǫrr + νǫφφ )
1 − ν2
E
σφφ = (ǫφφ + νǫrr )
1 − ν2
E
σrφ = ǫrφ
1+ν
Of the six compatibility equations, two are identically satisfied, and the four remaining ones become
ǫ33 = c1 + c2 x1 + c3 x2
where c1 , c2 , and c3 are constants. Only (4.24a) is taken into account in most problems, and the requirement of (4.24b)
is ignored. This results in solutions that, although approximate, are satisfactory when the dimension of the body in
the x3 -direction is small. As in the case of plane strain, the strain-stress relations (4.23) and the equilibrium equations
(4.3a, b) transform the compatibility equation (4.24a) to
Comparing (4.7) and (4.21), it is evident that solutions of plane stress problems can be obtained from the solutions of
corresponding plane strain problems on replacing λ by λ = 2λµ/(λ + 2µ). However, it may be noted that in the case
of plane stress, the components u1 and u2 of the displacement may depend on x3 . Hence plane stress problems are
not truly two-dimensional.
2005
c by Ajit Mal
4.3. PLANE STRESS 103
The plane stress formulation is appropriate when the body is thin in the x3 -direction and can carry only stresses parallel
to this plane. Under these assumptions, the stress components σi3 are negligibly small. This can be made more explicit
through the so-called generalized plane stress formulation.
Consider a ”thin” flat plate whose thickness, 2h, is small compared to the linear dimensions of its cross section.
Assume that the following additional conditions are satisfied:
(a) The flat surfaces of the plate are free from applied forces.
(b) The forces acting on the curved surface lie in planes parallel to the middle plane (x3 = 0) and are symmetrically
distributed with respect to this plane.
(c) The component f3 of the body force vanishes and the components f1 and f2 are symmetrically distributed with
respect to the middle plane.
It is obvious from the symmetry of these problems that the displacement component u3 will be an odd function of
x3 . Therefore, points of the middle plane will undergo no displacement in x3 -direction and the mean value of u3 with
respect to the thickness of the cylinder will vanish, i.e.,
h
1
Z
u3 (x1 , x2 ) = u3 (x1 , x2 , x3 ) dx3 = 0 (4.26)
2h −h
where the overbar signifies mean value over the thickness of the plate. From assumption (a),
Therefore,
∂σ13 ∂σ23
= =0 at x3 = ±h
∂x1 ∂x2
∂σ33
=0 at x3 = ±h
∂x3
Therefore, both σ33 and its normal derivative vanish at x3 = ±h. If we expand σ33 in a Taylor series in x3 about ±h,
it is apparent that σ33 = O(h2 ).
Since the plate is thin, we assume that σ33 = 0 throughout its interior. We introduce the averaged field variables,
ui , eij and σ ij as defined in (4.25) for u3 . Then,
u1 = u1 (x1 , x2 ), u2 = u2 (x1 , x2 ), u3 = 0
σ 13 = σ 23 = σ 33 = 0
σ 11 = (λ + 2µ)e11 + λe22
σ 22 = λe11 + (λ + 2µ)e22
σ 12 = 2µe12
λ
e33 = − (e11 + e22 )
λ + 2µ
2005
c by Ajit Mal
104 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
In plane problems, it is convenient to use the standard notation, x, y, z instead of x1 , x2 , x3 for Cartesian coordinates,
and this will be done in all subsequent discussions.
Let us write the plane strain equilibrium equations (4.3a, b) on the xy-plane in the form
Further, let the external body force be conservative, so that f = −∇V , where V (x, y) is the force potential. Then
from (4.27)
Equations (4.28) can be satisfied through the introduction of a potential function Φ such that
where
∂2 ∂2
∇2 = +
∂x2 ∂y 2
Equations (4.29) and (4.30) reveal that Φ is a solution of
2 2 1 − 2ν
∇ ∇ Φ+ ∇2 V = 0 (4.31)
1−ν
The function Φ is known as Airy’s stress function. Many traction boundary-value problems of two-dimensional linear
elasticity can be formulated as boundary-value problems in Φ.
For plane stress, (4.25) and (4.29) yield
∇2 ∇2 Φ + (1 − ν)∇2 V = 0 (4.32)
If the body forces vanish, then in both plane strain and plane stress, Φ satisfies the same biharmonic equation
∇2 ∇2 Φ = 0 (4.33a)
2005
c by Ajit Mal
4.4. AIRY’S STRESS FUNCTION IN CARTESIAN COORDINATES 105
i.e.,
∂4Φ ∂4Φ ∂4Φ
+ 2 + =0 (4.33b)
∂x4 ∂x2 ∂y 2 ∂y 4
If there are no body forces, the displacement components can be expressed in the form
where
∇2 ψ = 0, ∇2 Φ = ψ,xy (4.35a)
and
For solids with rectangular geometry, solutions of (4.33) in the form of polynomials are of interest. By choosing
polynomials of various degrees and suitably adjusting their coefficients, a number of important problems can be solved.
As an example, consider a polynomial of the second degree,
a2 x 2 c2 y 2
Φ= − b2 xy + (4.36)
2 2
which satisfies (4.33) and yields the stresses
In the case of a rectangular plate or a beam with sides parallel to the coordinate axes, a positive c2 corresponds
to a uniform tension in the x-direction, whereas b2 gives rise to uniform shear. Solution of these problems by the
elementary strength of materials approach was discussed in Chapter 2.
Consider next a polynomial solution of the third degree,
a3 x 3 b3 x2 y c3 xy 2 d3 y 3
Φ= + + + (4.37)
6 2 2 6
which corresponds to the stresses
Since polynomials of the second and third degrees identically satisfy (4.33), the constants appearing in them can
be chosen arbitrarily. However, for polynomials of higher degrees, (4.33) is satisfied only if the coefficients satisfy
certain relations. Consider, for example, a polynomial of the fourth degree,
a4 x 4 b4 x 3 y c 4 x 2 y 2 d4 xy 3 e4 y 4
Φ= + + + + (4.38)
12 6 2 6 12
Clearly, Φ is biharmonic provided
a4 + 2c4 + ǫ4 = 0 (4.39)
The stresses are found to be
σxx = c4 x2 + d4 xy + ǫ4 y 2
σyy = a4 x2 + b4 xy + c4 y 2
−b4 x2 d4 y 2
σxy = − 2c4 xy −
2 2
Since (4.33) is a linear differential equation, the principle of superposition can be used; i.e., a linear combination of
its solutions is also a solution. Thus, exact solutions for certain problems involving rectangular geometries can be
obtained by selecting appropriate polynomial expressions for the stress function. In a given problem the solution is
2005
c by Ajit Mal
106 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
l
b
h
x
o
h
y P
correct only if the applied surface loads or displacements are consistent with the stresses and displacements derived
from them. These conditions are seldom satisfied in practice. However, as we shall show in the following examples,
approximate but useful simple closed form solutions of many practical engineering problems can be obtained through
this approach. Other examples can be found in standard texts on advanced strength of materials (e.g., Timoshenko
and Goodier 1951).
Consider a cantilever beam of length l, width b and a narrow, rectangular crosssection bent by a force P applied
at the end (Fig. 4-1). The upper and lower edges of the beam are free from traction. Shearing forces, having a resultant
P , are distributed along the end in the plane x = 0. To solve this problem, we consider a combination of (4.36) with
only b2 nonzero and (4.38) with only d4 nonzero,
d4 xy 3
Φ = −b2 xy + (i)
6
The stresses derived from (i) are
d4 y 2
σxx = d4 xy, σyy = 0, σxy = b2 − (ii)
2
Since the sides y = ±h are assumed to be traction-free, we must have
d4 h2
b2 =
2
The force balance condition at the loaded end gives
h h
1
Z Z
P = (−σxy )b dy = −b (b2 − d4 y 2 ) dy
−h −h 2
which yields
−3P
d4 =
2bh3
Inserting the values of b2 and d4 in (ii), we find
−P xy −P (h2 − y 2 )
σxx = , σyy = 0, σxy = (iv)
I (2I)
It should be noted that these results represent the exact solution of the problem provided the shearing forces at the
ends are distributed according to the same parabolic law as the shearing stress σxy in (iv) and if the intensity of the
2005
c by Ajit Mal
4.4. AIRY’S STRESS FUNCTION IN CARTESIAN COORDINATES 107
normal forces at the built-in end is proportional to y. If the forces at the ends are distributed in any other manner, this
solution is not exact. However, this solution can be considered to be satisfactory for crosssections of the beam at some
distance away from the ends.
The displacement components can be calculated from (4.34). From (i) and (4.35),
ψ,xy = ∇2 Φ = d4 xy (v)
d4 x 2 y 2
ψ= + f (x) + g(y) (vi)
4
where f(x) and g(y) are arbitrary functions. Since ψ is harmonic, these functions satisfy the equation
d4 (x2 + y 2 )
f ”(x) + g”(y) + =0
2
for all x, y in the domain occupied by the beam. Hence we must have
d4 x 2
f ”(x) = a0 −
2
d4 y 2
g”(y) = −a0 −
2
where a0 is a constant. Integrating these equations and substituting the results in (vi),
1 1 1
ψ= d4 x2 y 2 + a0 (x2 − y 2 ) − d4 (x4 + y 4 ) + a1 x + b1 y + c1
4 2 24
where a1 , b1 , c1 are arbitrary constants. Then (4.34) gives the displacement components
αd4 x2 y (1 + α)d4 y 3
2µu = − − (a0 α − b2 )y + b1 α (viia)
2 6
(1 − α)d4 xy 2 αd4 x3
2µv = − − + (a0 α + b2 )x + a1 α (viib)
2 6
It can easily be seen that these displacement components cannot be made to satisfy the required boundary condi-
tions at the cantilever end, namely,
The unknown constants a0 , a1 , b1 can be determined through the application of approximate ”engineering” boundary
conditions at x = ℓ. As an example, if we assume that the center of the end section is fixed and that the center line of
the beam has zero slope at this end, then the displacement components must satisfy the conditions
u = v = 0, v,x = 0 at x = ℓ, y = 0 (ix)
Then
αd4 ℓ2 −d4 ℓ3
b1 = 0, a0 α + b2 = , a1 = (x)
2 3
For plane stress, the deflection of the centerline of the beam can be shown to be given by
2005
c by Ajit Mal
108 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
a
p
x
O
Fig. 4-2. The plane strain problem of a dam.
more and more accurate as the distance from the ends increases. Solutions of this type are quite useful in many
engineering applications.
Example 4.4-2 (The dam problem)
As another example of the use of the Airy stress function in Cartesian coordinates, consider the plane strain prob-
lem of a very long vertical plate of thickness a with the lower end fixed and subjected to a linearly decreasing lateral
load (Fig. 4-2). The stress boundary conditions for the problem are
p(y − h)
σxx = , at x = 0, 0 < y < h
h
= 0, at x = a, 0 < y < h (ia)
σyy = 0, at y = h, 0 < x < a (ib)
σxy = 0, at x = 0, 0 < y < h (ic)
at x = a, 0 < y < h
at y = h, 0 < x < a
We show how an appropriate polynomial based stress function for this problem can be constructed from the stress
boundary conditions. The boundary condition (ia) can be satisfied by the assumption
q”(x) = −hg”(x)
2005
c by Ajit Mal
4.4. AIRY’S STRESS FUNCTION IN CARTESIAN COORDINATES 109
which is satisfied if
q(x) = −hg(x)
Since the third boundary condition in (ic) cannot be satisfied exactly, we enforce
Z b
σxy dx = 0
a
which gives
g(0) − g(a) = 0.
Further, since Φ is biharmonic,
for all x, y in the domain 0 < x < a, 0 < y < h. Thus, we must have
−2pf ”(x)
f (iv) (x) = 0, g (iv) (x) = (x)
h
with the conditions: f (0) = 1, f (a) = 0, f ′ (0) = f ′ (a) = 0; g ′ (0) = g ′ (a) = 0, g(0) = g(a).
The simplest polynomial solutions of (x) subject to the boundary conditions (iii) and (viii) can be easily shown to be
(x − a)2 (2x + a)
f (x) = (xia)
a3
−p(a − 2x)(a − x)2 x2
g(x) = +c (xib)
10ah3
From (vi) and (xi),
2005
c by Ajit Mal
110 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
In plane polar coordinates (r, φ), the strain-displacement relations, the equations of equilibrium, and the strain-stress
relations are the same as (4.15), (4.16), and (4.17). In the absence of body forces, the equilibrium equations (4.16)
may be expressed in the form
Equations (4.40) are satisfied identically if the stresses are derived from a potential function Φ(r, φ) in the form
where
∇2 ψ = 0 (4.44a)
2
∇ Φ = (rψ,φ ),r (4.44b)
and
1−ν for plane strain
α= (4.45)
(1 + ν)−1 for plane stress
The general solution of (4.42) in the form
∞
X
Φ(r, φ) = gn (r) [An cos(nφ) + Bn sin(nφ)] (4.46)
n=0
The expressions for gn (r) for different values of n and the corresponding expressions for the stress and displacement
components are derived next.
Substitution of (4.46) into (4.42) leads to the following ordinary differential equation for gn (r):
2
n2
2
n2
d 1 d d 1 d
+ − + − gn (r) = 0 (4.47)
dr2 r dr r2 dr2 r dr r2
The general solutions of this equation for different values of n can be obtained by standard techniques. It can be shown
that
g0 (r) = a0 r2 + b0 r2 ln r + c0 + d0 ln r (4.48a)
g1 (r) = a1 r3 + b1 r + c1 r ln r + d1 r−1 (4.48b)
gn (r) = an rn+2 + bn rn + cn r−n+2 + dn r−n , n>1 (4.48c)
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 111
For n = 1,
For n > 1,
In these equations u0r and u0φ represent rigid-body displacement components, which produce no stress or strain in the
material.
In axisymmetric (n = 0) plane strain problems, the displacement components are of the general form
1+ν
ur = [2(1 − 2ν)a0 r − b0 r + 2(1 − 2ν)b0 r ln r − d0 r−1 ] + u0r
E
(4.52a)
4(1 − ν 2 )
uφ = b0 rφ + u0φ (4.52b)
E
For axisymmetric plane stress problems,
1
ur = [2(1 − ν)a0 r − (1 + ν)b0 r + 2(1 − ν)b0 r ln r − (1 + ν)d0 r−1 ] + u0r
E
(4.53a)
4b0 rφ
uφ = + u0φ (4.53b)
E
In specific problems the form of the solution will depend on the boundary and other conditions as illustrated in
Examples 4.5-1 through 4.5-4 below.
2005
c by Ajit Mal
112 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
We suppose that a portion of a ring between two adjacent cross-sections has been cut out and the ends of the ring
have been joined again by welding or by some other process. This will generate initial stresses in the ring; i.e., there
are stresses in the ring when the external forces are absent. Let ǫ be the small angle measuring the portion of the ring
that was cut out (Fig. 4-3). Then, the displacement necessary to bring the ends of the ring together is given by
ur = 0, uφ = ǫr (i)
Assuming plane stress we note from (4.53b) that the term (4/E)b0 rφ is zero when φ = 0 but becomes (8π/E)b0 r
when we trace a circuit around the axis of symmetry and return to the same point after turning around an angle 2π.
Thus, the displacement given by (4.53b) is multivalued. The discontinuity in the displacement given by (i) can be
obtained from (4.53b) by assuming
ǫE
b0 = . (ii)
8π
Since σrr = 0 on the internal (r = a) and external (r = b) faces of the ring, the other two constants, a0 and d0 , can
be obtained from 4.49 as
ǫE(b2 − a2 + 2b2 ln b − 2a2 ln a) ǫEa2 b2 ln ab
a0 = − , d0 = . (iii)
8(b2 − a2 )π 4(b2 − a2 )π
Thus, the initial stresses are
ǫE a2 (b2 − r2 ) ln a + b2 (r2 − a2 ) ln b − r2 (b2 − a2 ) ln r
σrr =−
4πr2 (b2 − a2 )
ǫE a (b + r2 ) ln a − b2 (r2 + a2 ) ln b + r2 (b2 − a2 )(1 + ln r)
2 2
σφφ =
4πr2 (b2 − a2 )
The stress function solution given in Sec. 4.5 can be used to determine the stresses in the loaded circular arch shown
in Fig. 4-4. Let a and b be the internal and external radii of the arch, respectively. The boundary conditions on the
curved surfaces of the arch are
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 113
2P
r
Φ
Q Q
d
M M
P P
Fig. 4-4. Arch problem.
The conditions at the end sections, φ = 0, π are given in terms of the resultants as
Z b
σφφ dr = −P (iia)
a
Z b
σrφ dr = −Q (iib)
a
Z b
σφφ r dr = −M − P d (iic)
a
This problem can be solved by means of the stress function Φ in the form
f (r) = a0 r2 + b0 r2 ln r + c0 + d0 ln r (iva)
g(r) = a1 r3 + b1 r + c1 r ln r + d1 r−1 (ivb)
The stress and displacement components can be calculated from (4.49) and (4.50) as
2005
c by Ajit Mal
114 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
A = P, B = −Q (viiia)
M0 2
a0 = − [b − a2 + 2(b2 ln b − a2 ln a)] (viiib)
N2
2M0 2
b0 = (b − a2 ) (viiic)
N2
4M0 2 2 b
d0 = a b ln (viiid)
N2 a
1 −(a2 + b2 ) −a2 b2
a1 = , c1 = , d1 = (viiie)
2N1 N1 2N1
where
2 2 2 2 b
N1 = (a − b ) + (a + b ) ln (ixa)
a
2
2 2 2 2 2 b
N2 = (b − a ) − 4a b ln (ixb)
a
M0 = −M − P d (ixc)
a 2 + b2 a 2 b2
P cos φ
σrr = r− + 3 (xia)
N1 r r
2 2
a 2 b2
P cos φ a +b
σφφ = 3r − − 3 (xib)
N1 r r
2 2 2 2
P sin φ a +b a b
σrφ = r− + 3 (xic)
N1 r r
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 115
σο σ r ο
ϕ
a x
Similarly, the case of pure bending is obtained by letting P = Q = 0 in the preceding. It should again be noted that
the solutions obtained above are exact only if the stresses at the end sections are distributed in conformity with those
given in (x); otherwise they are approximate.
Consider a plate subjected to a uniform tension σ0 in the x-direction (Fig. 4-5). If a small circular hole of radius
a is made in the middle of the plate, the stress distribution in the neighborhood of the hole will be changed, but the
change is negligible at distances large compared to a.
If we use Cartesian coordinates x, y, the stress components for large values of r = (x2 + y 2 )1/2 are given by
where A, B, C, D, α, β, γ, δ are constants. The stresses follow from (4.49) and (4.51). Noting that the stresses must
be finite as r tends to infinity, we must have B = β = 0. Then
A 6γ 4δ
σrr = + 2C − 2α + + cos 2φ (iiia)
r2 r4 r2
A 6γ
σφφ = − 2 + 2C + 2α + 4 cos 2φ (iiib)
r r
6γ 2δ
σrφ = 2α − 4 − 2 sin 2φ (iiic)
r r
2005
c by Ajit Mal
116 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
σ 0 a2 σ0
A = − , C= (v)
2 4
σ0 −a4 σ0 a2 σ 0
α = − , γ= , δ= (vi)
4 4 2
Inserting the values of the various constants in (iii), we find
a2 3a4 4a2
σ0
σrr = 1 − 2 + 1 + 4 − 2 cos 2φ (viia)
2 r r r
2 4
σ0 a 3a
σφφ = 1 + 2 − 1 + 4 cos 2φ (viib)
2 r r
4 2
σ0 3a 2a
σrφ = − 1 − 4 + 2 sin 2φ (viic)
2 r r
At the boundary of the hole,
The hoop stress σφφ on the boundary of the hole is greatest when φ = π/2 or 3π/2 and
Thus the maximum tensile (hoop) stress at the boundary of the hole is three times the applied (remote) stress
If a tensile stress of magnitude σ0 is applied in the y-direction instead of the x-direction, the stress components
can be easily shown to be given by
a2 3a4 4a2
σ0
σrr = 1 − 2 − 1 + 4 − 2 cos 2φ (xa)
2 r r r
a2 3a4
σ0
σφφ = 1 + 2 + 1 + 4 cos 2φ (xb)
2 r r
3a4 2a2
σ0
σrφ = 1 − 4 + 2 sin 2φ (xc)
2 r r
Therefore, if the plate is under the simultaneous influence of equal tensile stresses in two mutually orthogonal direc-
tions in the plane of the plate, the stresses are obtained by superposition as
a2 a2
σrr = σ0 1 − 2 , σφφ = σ0 1 + 2 , σrφ = 0
r r
This corresponds to the case of hydrostatic tension discussed in Chapter 3.
If the remote tensile stress is applied at an angle α with the x-axis, then the stresses can be obtained from the above
solution simply by changing φ to φ − α.
A tensile stress S in the x′ -direction and a compressive stress (−S) in the y ′ -direction amount to pure shear S in
the xy-plane, provided the x′ y ′ -axes are oriented at 45o as shown in Fig. 4-6. In (vii) and (x), the angle φ should be
changed to φ − π/4 in order to use the equations in the x′ y ′ -system. Thus the stress components in the presence of
the hole subjected to a remote pure shear S in the xy-plane are given by
3a4 4a2
π
σrr = S 1 + 4 − 2 cos 2φ −
r r 2
4
3a π
σφφ = −S 1 + 4 cos 2φ −
r 2
4 2
3a 2a π
σrφ = −S 1 − 4 + 4 sin 2φ −
r r 2
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 117
S
S
S
y
x’
y’
o
45
x
S
S
Fig.4-6. Circular hole under pure shear.
λ2, µ 2 r
σ0 ϕ σ0 x
2a λ1, µ 1
Fig. 4-7. The cylindrical inclusion.
At r = a, we now have
σrr = σrφ = 0, σφφ = −4S sin 2φ
and, therefore,
(σφφ )max = 4S
Hence, for a large plate under pure shear, the maximum tensile stress at the boundary of the hole is four times the
applied pure shear stress.
Example 4.5-4 (The inclusion problem)
Next consider the so-called inclusion problem, in which a cylinder of one material is embedded in a different un-
bounded material (matrix) (Fig. 4-7). The elastic constants of the matrix are λ1 , µ1 and those of the cylinder are
λ2 , µ2 . The interface r = a is perfectly bonded (i.e., welded) and the medium is subjected to uniaxial tension σ0 ,
as in the previous problem. The boundary conditions for this problem are that the displacement and stress vectors
are continuous across the interface. Using superscripts (1) and (2) to denote the field variables in the matrix and the
inclusion, respectively, the interface conditions can be expressed as
(1) (2) (1) (2) (1) (2)
σrr = σrr , σrφ = σrφ , u(1) (2)
r = ur , uφ = uφ on r = a
At large distances from the inclusion, the stress state must satisfy condition (ii) of Example 4.5-6.
The appropriate stress function is of the form
Φ(n) (r, φ) = f (n) (r) + g (n) (r) cos 2φ, n = 1, 2 (i)
It has been shown in (4.48) that f (n) , g (n) are given by
f (n) = an r2 + bn r2 ln r + cn + dn ln r
g (n) = ǫn r4 + fn r2 + gn + hn r−2
Thus the stresses can be found from (4.49) and (4.51) as
(n)
σrr = 2an + bn (2 ln r + 1) + dn r−2 − (2fn + 6hn r−4 + 4gn r−2 ) cos 2φ (iia)
(n) −2 2 −4
σφφ = 2an + bn (2 ln r + 3) − dn r + (12ǫn r + 2fn + 6hn r ) cos 2φ (iib)
(n) 2 −2 −4
σrφ = 2(3ǫn r + fn − gn r − 3hn r ) sin 2φ (iic)
2005
c by Ajit Mal
118 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
d 2 = h 2 = b2 = g 2 = 0
(2)
σrr = 2a2 − 2f2 cos 2φ (va)
(2) 2
σφφ = 2a2 + 2(6ǫ2 r + f2 ) cos 2φ (vb)
(2)
σrφ = 2(3ǫ2 r2 + f2 ) sin 2φ (vc)
where
σ0 2
f (1) (r) = r + c1 + d1 ln r
4
−σ0 2
g (1) (r) = r + g1 + h1 r−2
4
f (2) (r) = a2 r 2 + c 2
g (2) (r) = ǫ2 r 4 + f2 r 2
In order to apply the interface continuity conditions, it is necessary to determine the displacement components from
(4.50) and (4.51). It can be shown that the displacements in the matrix are of the general form
4µ1 u(1)
r = −σ0 r − 2d1 r−1 + (σ0 r + 4h1 r−3 ) cos 2φ
+ 2α1 [σ0 r + 4g1 r−1 cos 2φ] + 4µ1 u0r (via)
(1)
4µ1 uφ = (−σ0 r + 4g1 r −1
+ 4h1 r −3
) sin 2φ − 8α1 g1 r −1
sin 2φ + 4µ1 u0φ (vib)
where α1 is the value of α defined in (4.45) for the matrix. The last term in each of (vi) represents rigid-body motion
and can be ignored. Thus, the displacement components in the matrix are given by
4µ1 u(1)
r = σ0 r(2α1 − 1) − 2d1 r−1 + (σ0 r + 4h1 r−3 + 8g1 α1 r−1 ) cos 2φ (viia)
(1) −1 −3
4µ1 uφ = [−σ0 r + 4g1 (1 − 2α1 )r + 4h1 r ] sin 2φ (viib)
µ2 u(2)
r = a2 r(2α2 − 1) + [2ǫ2 r3 (α2 − 1) − f2 r] cos 2φ (viiia)
(2) 3
µ2 u φ = [(2α2 + 1)ǫ2 r + f2 r] sin 2φ (viiib)
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 119
Application of the interface conditions gives the following equations for the calculation of the constants
where
µ2
β= (x)
µ1
Equations (ix) can be solved for the constants; the results are
α1 βσ0
a2 =
2(2α2 + β − 1)
(1 + 2α1 β − 2α2 − β)σ0 a2
d1 =
2(2α2 + β − 1)
α1 βσ0 (1 − β)σ0 a2
f2 = − , g1 =
1 − β + 4α1 β 2(1 − β + 4α1 β)
4
(β − 1)σ0 a
h1 = , ǫ2 = 0
4(1 − β + 4α1 β)
The other constants do not contribute to the deformation and stress in the solid.
The displacement and stress components are given by
σ0 r 1 + 2α1 β − 2α2 − β a 2
u(1)
r = (2α 1 − 1) −
4µ1 2α2 + β − 1 r
a 4
4α1 (1 − β) a 2 1−β
+ 1+ − cos 2φ (xia)
1 − β + 4α1 β r 1 − β + 4α1 β r
a 4
(1) σ 0 r 2(1 − β)(1 − 2α 1 ) a 2 1−β
uφ = −1 + + sin 2φ (xib)
4µ1 1 − β + 4α1 β r 1 − β + 4α1 β r
(1) σ0 1 + 2α1 β − 2α2 − β a 2 4(1 − β) a 2
σrr = 1+ + 1−
2 2α2 + β − 1 r 1 − β + 4α1 β r
3(1 − β) a 4
+ cos 2φ (xiia)
1 − β + 4α1 β r
(1) σ0 1 + 2α1 β − 2α2 − β a 2
σφφ = 1−
2 2α2 + β − 1 r
3(1 − β) a 4
− 1+ cos 2φ (xiib)
1 − β + 4α1 β r
(1) σ0 2(1 − β) a 2 3(1 − β) a 4
σrφ = − 1+ − sin 2φ (xiic)
2 1 − β + 4α1 β r 1 − β + 4α1 β r
σ0 α1 βr 2α2 − 1 2 cos 2φ
u(2)
r = + (xiiia)
2µ2 2α2 + β − 1 1 − β + 4α1 β
(2) σ0 α1 βr sin 2φ
uφ = − (xiiib)
µ2 (1 − β + 4α1 β)
2005
c by Ajit Mal
120 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
O
y
Φ r
O y
C −Φrr
(2) 1 2 cos 2φ
σrr = σ 0 α1 β + (xiva)
2α2 + β − 1 1 − β + 4α1 β
(2) 1 2 cos 2φ
σφφ = σ 0 α1 β − (xivb)
2α2 + β − 1 1 − β + 4α1 β
(2) 2σ0 α1 β sin 2φ
σrφ = − (xivc)
1 − β + 4α1 β
where
1 − νi for plane strain
αi =
(1 + νi )−1 for plane stress
i = 1, 2.
In some problems other forms of stress functions are needed due to different types of boundary conditions as
illustrated in the next few examples.
Example. 4.5-5 (Normal line load on a semi-infinite medium)
Let a normal line load P per unit length act at the origin on the surface of the semi-infinite elastic medium x ≥ 0 (Fig.
4-8). We shall show that this plane strain problem can be solved by the Airy stress function
P
Φ=− rφ sin φ (i)
π
The stresses follow from (4.41) as
2P
σrr = − cos φ, σφφ = 0, σrφ = 0 (ii)
πr
The maximum shearing stress at (r, φ) can be shown to be
P
τmax = cos φ (iii)
πr
From (ii) it is clear that the surface x = 0 (excluding the origin) is traction-free. Further, the locus of points where the
radial stress σrr is constant is a circle, which touches the x-axis at the origin (Fig. 4-8).
Consider a semicircle C of radius ǫ, centered at the origin, and lying in the material of the half-space, x ≥ 0. The
resultant force acting on the inside of the right cylinder of unit thickness and having the semicircle C as its base must,
from symmetry, act in the x-direction. From (ii), the magnitude of this resultant force is
π/2 π/2
2P
Z Z
(−σrr )r cos φ dφ = cos2 φ dφ = P (iv)
−π/2 π −π/2
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 121
Since the resultant force is independent of ǫ, in the limit as ǫ → 0, it remains P . Hence the stress field (ii) may be
regarded as due to a line load of magnitude P per unit length acting at the origin in the positive x-direction on the
surface of the semi-infinite region x ≥ 0.
The strains can be calculated from the strain-stress relations (4.17) as
The displacement components can be obtained, either by direct integration of the strains or through the use of (4.43),
as
2(1 − ν 2 )P P
ur = − cos φ ln r − (1 − 2ν)(1 + ν)φ sin φ + u0r (via)
πE πE
2ν(1 + ν)P 2(1 − ν 2 )P
uφ = sin φ + sin φ ln r
πE πE
P
+ (1 − 2ν)(1 + ν)(sin φ − φ cos φ) + u0φ (vib)
πE
where u0r , u0φ correspond to rigid-body displacements. To determine these, let us apply the condition that all points on
the x-axis do not have any lateral displacement and one point at a distance d on this axis is fixed. Then (via) and (vib)
give
(1 + ν)P d
ur = 2(1 − ν) cos φ ln − (1 − 2ν)φ sin φ (viia)
πE r
(1 + ν)P d
uφ = sin φ − 2(1 − ν) sin φ ln − (1 − 2ν)φ cos φ (viib)
πE r
−P (1 + ν)(1 − 2ν)
ur = (viiia)
2E
±P (1 + ν) d
uφ = 1 − 2(1 − ν) ln (viiib)
πE r
Equation (viiia) indicates that the material on the surface is displaced toward the origin. Obviously, the solution
is not valid in the immediate neighborhood of the point of application of the line load. Moreover, the displacement
components contain a logarithmic singularity at ∞. This is peculiar to many two-dimensional elastostatic problems.
The stress and displacement components for the plane stress problem for a normal line load on a semi-infinite
elastic plate can be obtained by simply changing ν to ν/(1 + ν), keeping µ = E/[2(1 + ν)] unchanged. The results
for the displacements are
P d
ur = 2 cos φ ln − (1 − ν)φ sin φ
πE r
P d
uφ = (1 + ν) sin φ − 2 sin φ ln − (1 − ν)φ cos φ
πE r
2005
c by Ajit Mal
122 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
σrr
P r
Φ
x
O
2α
Consider a wedge of infinite length subjected to a load P per unit length at the vertex along the axis of the wedge
(Fig. 4-9).
The problem can be solved by the Airy stress function
These stresses vanish at infinity and make the straight edges of the wedge φ = ±α traction-free. The constant c can
be found by considering the equilibrium of a portion of the wedge lying inside a cylindrical surface of radius r and
center at the origin O. Summing forces along the x- and y-directions, we find
Z α
σrr r cos φ dφ + P = 0 (iiia)
−α
Z α
σrr r sin φ dφ = 0 (iiib)
−α
Substituting the value of σrr from (iia), we find that condition (iiib) is identically satisfied, and condition (iiia) gives
P
c=− (iv)
2α + sin 2α
We, therefore have
2P cos φ
σrr = − , σφφ = σrφ = 0 (v)
r(2α + sin 2α)
If we put α = π/2, we get the results for the case of a semi-infinite medium obtained in Problem 4.5-3.
The problem of a line load P acting at the apex in the y-direction (normal to the axis of the wedge) can be solved
by considering the Airy stress function
where the constant c can be found as before. The stresses are found to be
2P sin φ
σrr = − , σφφ = σrφ = 0 (vii)
r(2α − sin 2α)
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 123
q x
y
Fig. 4-10. A cantilever wedge subjected to distributed load.
The stresses for an inclined load at the apex of a wedge can be obtained by the superposition of those for the two cases
just given.
Consider the plane stress problem of a cantilever wedge subjected to the uniform load shown in Fig. 4-10. An
approximate solution of the problem can be obtained through the use of the stress function
cr2
Φ= [(α − φ) cos α − cos φ sin(α − φ)] (i)
cos α
where c is a constant.
It can be easily verified that Φ satisfied the biharmonic equation (4.42) and by (4.41), the stress components are
2c
σrr = [(α − φ) cos α − sin φ cos(α − φ)] (ii)
cos α
2c sin φ sin(α − φ)
σrφ = − (iii)
cos α
2c
σφφ = [(α − φ) cos α − cos φ sin(α − φ)] (iv)
cos α
The Cartesian components of the stress vector on a vertical section, x = a constant are given by
2005
c by Ajit Mal
124 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
r
Φ x
O
2 Φ0
These results are consistent with the overall equilibrium of the wedge. It should be noted that as in the case of Example
4.4-1, the displacement boundary conditions at the cantilever section cannot be satisfied by the solution.
Consider the wedge-shaped solid shown in Fig. 4-11. The internal angle of the wedge is 2φ0 , where 0 < φ0 < π,
and the stress state is two-dimensional (either plane strain or plane stress). The Airy’s stress function formulation
in polar coordinates can be used to determine the nature of the stresses in the solid when the faces of the wedge are
subject to various types of boundary conditions. Of particular interest in engineering applications is the possibility of
the presence of large stresses near the corner.
Let (r, φ) denote the polar coordinates of a point within the wedge (Fig. 4-11) such that the faces of the wedge
are given by
φ = ±φ0 (i)
We recall from (4.41) and (4.43) that the stress and displacement components can be expressed in terms of the functions
φ and ψ, where Φ is biharmonic and ψ is harmonic. We seek solutions of (4.42) and (4.44a) in the form
where β, m are unknown constants and F (φ), G(φ) are unknown functions. Substitution into (4.42) and (4.44a) yields
the following ordinary differential equations for F (φ) and G(φ):
2005
c by Ajit Mal
4.5. AIRY’S STRESS FUNCTION IN POLAR COORDINATES 125
where a1 , a2 , b1 , b2 , b3 , b4 are constants. Using (ii) in the auxiliary condition (4.44b), we obtain
Since (v) must be valid for all r in 0 < r < ∞, we must have
β =m+1 (vi)
4b3 4b4
a1 = − , a2 = (vii)
β−1 β−1
The displacement and stress components are obtained from (4.41) and (4.43) as
If both faces of the wedge are free, then the components of the stress vector must vanish on them, i.e.,
2005
c by Ajit Mal
126 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
and upon using (xb), (xc), the following system of linear homogeneous equations are obtained for the remaining
unknowns, b1 , b2 , b3 , b4 ,
The preceding equations show that the constants b1 , b3 are solutions of the system
It can be seen that b1 , b3 produce antisymmetric displacements and b2 , b4 produce symmetric displacements in the
wedge relative to its centerline φ = 0. For nontrivial solutions of (xiiia) and (xiiib) for b1 , b3 , we must have
Thus, for a wedge with both faces free, the constant β must be a solution of (xva) in order to produce antisymmetric
displacements and a solution of (xvb) to produce symmetric displacements. The solutions may, in general, be complex.
In order that the displacement components given by (ixa) and (ixb) be finite everywhere, Re(β) must be positive, where
Re stands for the real part of a complex number. Numerical solution of (xva) and (xvb) are shown in figure 4-12.
If the faces of the wedge are clamped, the boundary conditions become
ur = uφ = 0 at φ = ±φ0
where the plus sign is for antisymmetric case and the minus sign is for symmetric case. The constant α is defined in
(4.45).
It can be seen from (xv) that for 0 < φ0 < π/2, the minimum value of Re(β) is greater than 1. Moreover the
stress components are proportional to rRe(β)−1 , so that the stress is finite at the corner. However, these equations have
real roots in the range β > 0 if π/2 < φ0 < π, and so the stress becomes singular at the corner. The details of this can
be found in Williams (1952). It should be noted that the order of the singularity is such that the integral of the strain
energy density W is finite for all choice of the region of integration and that the net force at the corner is zero.
2005
c by Ajit Mal
4.6. COMPLEX VARIABLE METHOD 127
The general solution to the two-dimensional equations of elastostatics can be expressed in terms of two analytic but
otherwise arbitrary functions of a complex variable. The boundary conditions may then be expressed as certain func-
tional equation relating these so-called complex potentials on the boundaries of the solid. Many important engieering
problems can be solved through the use of these potentials. We present a brief discussion of the general idea behind
the technique and use it in the solution of a few elastostatic problems. A more detailed discussion can be found in
Sokolnikoff (1956) and England (1971). In the absence of external body forces, the stress components are related to
the Airy stress function Φ(x, y) through the relations
where
∇2 ∇2 Φ = 0 (4.55)
Let
z = x + iy, z = x − iy (4.56a)
so that
1 1
x= (z + z), y= (z − z) (4.56b)
2 2i
√
where i = −1. Then, any function of x and y may be regarded as a function of z and its conjugate z. Writing
Φ(x, y) = Φ̂(z, z), we have
∂Φ ∂ Φ̂ ∂z ∂ Φ̂ ∂z ∂ Φ̂ ∂ Φ̂
= + = + (4.57a)
∂x ∂z ∂x ∂z ∂x ∂z ∂z !
∂Φ ∂ Φ̂ ∂z ∂ Φ̂ ∂z ∂ Φ̂ ∂ Φ̂
= + =i − (4.57b)
∂y ∂z ∂y ∂z ∂y ∂z ∂z
Therefore,
∂ 2 Φ̂ ∂ 2 Φ̂ ∂ 2 Φ̂
∇2 Φ = + = 4 (4.58)
∂x2 ∂y 2 ∂z∂z
2005
c by Ajit Mal
128 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
∂4Φ
16 =0 (4.59)
∂z 2 ∂z 2
where the ”hat” on Φ has been dropped. Solving (4.59), we obtain
∂2Φ
σxx + σyy = ∇2 Φ = 4 (4.64a)
∂z∂z
∂2Φ
σyy − σxx + 2iσxy = 4 2 (4.64b)
∂z
Equation (4.62) gives,
∂2Φ
∂u 1 2
= ǫxx = (1 − ν)∇ Φ − (4.66a)
∂x 2µ ∂x2
∂2Φ
∂v 1
= ǫyy = (1 − ν)∇2 Φ − (4.66b)
∂y 2µ ∂y 2
∂u ∂v 1 ∂2Φ
+ = 2ǫxy = − (4.66c)
∂y ∂x µ ∂x∂y
Let
F (z) = p(x, y) + iq(x, y) (4.67)
so that, from the Cauchy-Riemann relations,
∂p ∂q ∂p ∂q
= , =− (4.68)
∂x ∂y ∂y ∂x
and
∂p ∂q
F ′ (z) = +i (4.69)
∂x ∂x
Equations (4.65a), (4.68), and (4.69) yield
∂p ∂q
∇2 Φ = 2[F ′ (z) + F ′ (z)] = 4 =4 (4.70)
∂x ∂y
2005
c by Ajit Mal
4.6. COMPLEX VARIABLE METHOD 129
Integrating (4.66a) and (4.66b), after substituting suitable expressions for ∇2 Φ from (4.70), we have
∂Φ
2µu = − + 4(1 − ν)p(x, y) + f (y) (4.71a)
∂x
∂Φ
2µv = − + 4(1 − ν)q(x, y) + g(x) (4.71b)
∂y
where f (y) and g(x) are arbitrary functions. Inserting the values of u and v from (4.71) into (4.66c) and using the
relations (4.68) once again, we find
df (y) dg(x)
+ =0
dy dx
Therefore,
f (y) = αy + β, g(x) = −αx + γ (4.72)
where α, β, and γ are arbitrary constants. The displacement field corresponding to (4.72) is a rigid-body displacement
and will be ignored. Equations (4.57), (4.62), and (4.71) yield
∂Φ ∂Φ
2µ(u + iv) = − +i + 4(1 − ν)(p + iq)
∂x ∂y
∂Φ
= −2 + 4(1 − ν)(p + iq) (4.73)
∂z
= κF (z) − zF ′ (z) − G′ (z)
for plane strain, where
λ + 3µ
κ = 3 − 4ν = (4.74)
λ+µ
For plane stress, (4.73) is replaced by
2µ(u + iv) = κ̂F (z) − zF ′ (z) − G′ (z) (4.75)
with
3−ν 5λ + 6µ
κ̂ =
= (4.76)
1+ν 3λ + 2µ
The preceding results can be summarized as follows:
σxx + σyy = 2[F ′ (z) + F ′ (z)] = 4Re[F ′ (z)] (4.77a)
σyy − σxx + 2iσxy = 2[zF ”(z) + G”(Z)] (4.77b)
2µ(u + iv) = κF (z) − zF ′ (z) − G′ (z) (4.77c)
where F (z) and G(z) are analytic and
3 − 4ν, for plane strain
κ= 3−ν (4.78a)
1+ν , for plane stress
2005
c by Ajit Mal
130 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
Therefore,
in which the constants A, B, C, D, Bn , and Dn are related to the constants αn and βn . From (4.85a)
∞
X
F ′ (z) = A(1 + ln z) + Bz −1 + nBn z n−1 (4.86)
n=−∞
If A is complex, Re(A ln z) will be a multiple-valued function. Then, from (4.77), (σxx + σyy ) will be a multiple-
valued function. Therefore, we assume that A is real. It can be further shown that if we wish that the displacements
should be single-valued everywhere, then (see, for example, Sokolnikoff 1956)
A = 0, κB + C = 0 (4.87)
A few examples of the use of the complex variable method in solving two dimensional elastostatic boundary-value
problems are given below.
Consider the plane stress problem of a circular disc of radius a subjected to uniform pressure of magnitude p0 on
its circumference (Fig. 4-13). For this problem, the appropriate complex potentials are of the form
σrr + σφφ = 4A
σφφ − σrr + 2iσrφ = 0
2005
c by Ajit Mal
4.6. COMPLEX VARIABLE METHOD 131
y
r
Φ
a x
p o
Thus,
p0
A=−
2
and from (4.82a, b) and (4.83),
2005
c by Ajit Mal
132 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
x
F
X
o ϖ y
Y
The complex variable method can be used to determine the stress field produced by a concentrated force within
the material. To show this, we assume that a force with components X and Y is present at the origin of coordinates
(Fig. 4-14), O, and consider the material outside a circle with center at O and radius ǫ. Consider the complex stress
functions given by
−(X + iY )
F (z) = ln z
2π(1 + κ)
κ(X + iY )
G′ (z) = ln z
2π(1 + κ)
Clearly, these are special cases of (4.85) and, therefore, the stresses calculated from them are possible in an isotropic
solid.
The resultant moment on r = ǫ can be shown to be zero. It should be noted that the stress is singular at the point of
application of the force and the order of the singularity is r−1 .
Example 4.6-3 (Circular disc under concentrated forces, the Brazil Nut problem)
Consider a circular disc of radius a subjected to the action of two equal and opposite concentrated fores shown in
2005
c by Ajit Mal
4.6. COMPLEX VARIABLE METHOD 133
r2
r1
P Φ1 Φ2 P
x
Fig. 4-15. For this problem, we assume plane stress condition and introduce the complex stress functions
P a+z Pz
F (z) = − ln + (ia)
2π a − z 2πa
P a+z Pa 1 1 iP
G′ (z) = ln + − − (ib)
2π a − z 2π a − z a+z 2
Since
cos3 φ1 cos3 φ2
2P P
σxx = − + −
π r1 r2 πa
2P sin2 φ1 cos φ1 sin2 φ2 cos φ2
P
σyy = − + −
π r1 r2 πa
2P sin φ1 cos2 φ1 sin φ2 cos2 φ2
σxy = − −
π r1 r2
where r1 , r2 are the distances and φ1 , φ2 are the angles shown in Fig. 4-15. It can be seen that on the centerlines of
the disc (x-axis), φ1 = φ2 = 0; thus the shear stress component σxy = 0 on it except at the points of application of
the concentrated forces, where the stress is singular, as in Example 4.6-2.
For stress fields with symmetry about one or both coordinate axes, the complex variable approach simplifies con-
siderably. As an example of this, let us consider two stress fields, one symmetric about the x-axis and the other
antisymmetric about the x-axis.
Symmetric stress field. A symmetric stress field is assumed to have the properties
2005
c by Ajit Mal
134 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
Let
F (z) = FI (z) + FII (z), G(z) = GI (z) + GII (z) (4.94)
where FI , GI correspond to the symmetric part of the stress and FII , GII correspond to the antisymmetric part of the
stress. We assume that
where a prime indicates differentiation with respect to the argument. Equations (4.77) now yield
Therefore,
where Im stands for the imaginary part. We note that the stresses corresponding to FI (z) have the property (4.89) and
the stresses corresponding to FII (z) have the property (4.91). The functions
2005
c by Ajit Mal
4.6. COMPLEX VARIABLE METHOD 135
1
2µu = (κ + 1)Im ẐII (z) + y Re ZII (z) (4.100a)
2
1
2µv = − (κ − 1) Re ẐII (z) − y Im ZII (z) (4.100b)
2
where
dẐII (z)
= ZII (z)
dz
Example 4.6-4 (Stress field of a loaded crack)
where (z 2 − a2 )1/2 denotes the principal branch with the branch cut
−a ≤ x ≤ a, y=0
Putting
a2 r
r φ1 + φ2 3(φ1 + φ2 )
σxx = σ0 cos φ − − sin φ sin (iia)
(r1 r2 )1/2 2 (r1 r2 )3/2 2
2
r φ1 + φ2 a r 3(φ1 + φ2 )
σyy = σ0 cos φ − + sin φ sin (iib)
(r1 r2 )1/2 2 (r1 r2 )3/2 2
2
a r 3(φ1 + φ2 )
σxy = σ0 sin φ cos (iic)
(r1 r2 )3/2 2
σ0 r 2
κ−1 1/2 φ1 + φ2 φ1 + φ2
2µu = σ0 (r1 r2 ) cos − sin φ sin φ − (iiia)
2 2 (r1 r2 )1/2 2
2
κ+1 φ 1 + φ 2 σ 0 r φ 1 + φ2
2µv = σ0 (r1 r2 )1/2 sin − sin φ cos φ − (iiib)
2 2 (r1 r2 )1/2 2
Thus, for points on the x-axis, σxy = 0 and
−1/2
a2
σxx (x, 0) = σyy (x, 0) = σ0 1 − 2 (iva)
x
κ−1
2µu(x, 0) = σ0 (x2 − a2 )1/2 , v(x, 0) = 0 (ivb)
2
2005
c by Ajit Mal
136 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
y
σ0
r2 r1 σ0
φ2 r
φ φ1
A O B x
2a
σ0 ar1 1/2 2 φ1 φ1
u = κ − 1 + 2 sin cos (viia)
2µ 2 2 2
σ0 ar1 1/2 φ 1 φ 1
v = κ + 1 − 2 cos2 sin (viib)
2µ 2 2 2
Equations (vi) show that the stress is singular at the crack tip; the singularity is of the square-root type. The stress
components near the tip B of the crack can be expressed in the form
KI
σij = √ fij (φ1 ) (viii)
r1
2005
c by Ajit Mal
4.7. TRANSFORM METHOD 137
where
√
KI = σ0 πa (ix)
KI is known as the stress intensity factor; it is also of great importance in fracture mechanics. Equation (vii) indicates
1/2
that the crack opening displacement increases as r1 near B. It should be noted that the crack opening displacement
given by (vb) and the singular stresses given by (vi) are the same if the remote stress σ0 is applied normal to the crack
only.
Integral transform methods are useful for solving problems in which one or more dimensions of the solid is infinite.
As an example of the application of this method, we consider the plane problem of a homogeneous, isotropic, and
elastic half-space occupying the region y ≥ 0 and subjected to a normal pressure p(x) distributed over its boundary.
Then, the boundary conditions are
σxy = 0, σyy = −p(x) at y = 0 (4.101)
In an unbounded region it is also necessary to enforce certain conditions at infinity; we shall assume, on physical
grounds, that the stresses vanish as x, y → ∞.
We recall from Sec. 4.4 that the problem can be formulated in terms of the Airy’s stress function Φ through the
relations
σxx = Φ,yy , σyy = Φ,xx , σxy = −Φ,xy (4.102)
∇ ∇2 Φ
2
= 0, −∞ < x < ∞, y>0 (4.103)
The boundary conditions (4.100) become
Φ,xy = 0, Φ,xx = −p(x), −∞ < x < ∞, y=0 (4.104)
with suitable restrictions on the behavior of Φ as x, y → ∞, so that the stresses remain finite everywhere and vanish
at infinity.
We solve this problem by means of the Fourier transform technique. The transform pair is defined through
Z ∞
Φ̃(ξ, y) = Φ(x, y)ǫiξx dx (4.105a)
−∞
Z ∞ Z ∞
1 1
Φ(x, y) = Φ̃(ξ, y)ǫ−iξx dξ = Re Φ̃(ξ, y)ǫ−iξx dξ (4.105b)
2π −∞ π 0
where Φ̃(ξ, y) is the Fourier transform of Φ(x, y) with respect to x and Φ(x, y) is the inverse transform of Φ̃(xi, y)
and Re indicates real part. The integrals in (4.104) will be assumed to exist. For a discussion of the sufficient condi-
tions under which the Fourier inversion theorem holds, see, for examples Titchmarsh (1937) and Sneddon (1951).
Using (4.8) and (4.101), and under the assumption of plane strain conditions, the displacement gradients are ob-
tained as
(1 + ν)[(1 − ν)Φ,yy − νΦ,xx ]
u,x =
E
(1 + ν)[(1 − ν)Φ,xx − νΦ,yy ] −2(1 + ν)Φ,xy
v,y = , u,y + v,x =
E E
Taking Fourier transform, these equations yield
" #
1+ν d2 Φ̃ 2
−iξ ũ = (1 − ν) 2 + νξ Φ̃
E dy
" #
dṽ 1+ν 2 d2 Φ̃
= −(1 − ν)ξ Φ̃ − ν 2
dy E dy
dũ 1 + ν dΦ̃
− iξṽ = 2iξ
dy E dy
2005
c by Ajit Mal
138 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
where we have used the fact that the Fourier transform of (∂f /∂x) is −iξ f˜. Thus,
" #
i(1 + ν) d2 Φ̃
ũ = (1 − ν) 2 + νξ 2 Φ̃ (4.106a)
ξE dy
" #
1+ν d3 Φ̃ 2 dΦ̃
ṽ = (1 − ν) 3 − (2 − ν)ξ (4.106b)
ξ2E dy dy
From (4.101) and (4.105), the stress and displacement components within the medium are given by
Z ∞ 2
1 d Φ̃ −iξx
σxx = ǫ dξ (4.107a)
2π −∞ dy 2
Z ∞
1
σyy = − ξ 2 Φ̃ǫ−iξx dξ (4.107b)
2π −∞
Z ∞
i dΦ̃ −iξx
σxy = ξ ǫ dξ (4.107c)
2π −∞ dy
" #
1+ν ∞
d2 Φ̃
Z
u = i (1 − ν) 2 + νξ 2 Φ̃ ξ −1 ǫ−iξx dξ (4.108a)
2πE−∞ dy
" #
1+ν ∞ d3 Φ̃ 2 dΦ̃
Z
v = (1 − ν) 3 − (2 − ν)ξ ξ −2 ǫ−iξx dξ (4.108b)
2πE −∞ dy dy
where A and B are arbitrary constants and |ξ| has been used in the exponential in order to ensure that the stresses are
finite for all positive y. Equations (4.108b) and (4.109) yield
p̃(ξ) |ξ|p̃(ξ)
A= , B= (4.111)
ξ2 ξ2
We next consider some simple but useful applications of the transform method in the formulation and solution of
two-dimensional boundary-value problems of elastostatics in a half-space.
Let
p0 , −a < x < a
p(x) = (ia)
0, otherwise
(ib)
Then
a
sin ξa
Z
p̃(ξ) = p0 ǫiξx dx = 2p0 (ii)
−a ξ
2005
c by Ajit Mal
4.7. TRANSFORM METHOD 139
a a
po
x
Φ2 Φ r Φ1
r1
r2
(x,y)
2p0 ∞ dξ
Z
σxx = − (1 − ξy)ǫ−ξy sin ξa cos ξx (iva)
π 0 ξ
2p0 ∞ dξ
Z
σyy = − (1 + ξy)ǫ−ξy sin ξa cos ξx (ivb)
π 0 ξ
2p0 y ∞ −ξy
Z
σxy = − ǫ sin ξa sin ξx dξ (ivc)
π 0
2005
c by Ajit Mal
140 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
The locus of points at which the maximum shearing stress has a given value is known as an isochromatic line. Equation
(vii) shows that for the present problem, the isochromatic lines are a family of circles, all passing through the points
(a, 0).
If, in the solution of Example 4.7-1, we put p0 = P/2a and then take the limit as a → 0, we get the solution of
the problem considered in Example 4.5-3 for a normal line load of magnitude P per unit length acting on a uniform
half-space. Thus,
2P x2 y 2P y 3 2P xy 2
σxx = − , σyy = − , σxy = − (i)
πr4 πr4 πr4
where r2 = x2 + y 2 . It can be verified that (i) is the solution of the boundary-value problem of a uniform half-space
y ≥ 0 with the boundary conditions
x = r sin φ, y = r cos φ,
2P
σrr = − cos φ, σφφ = σrφ = 0 (iii)
πr
Equation (iii) coincides with (ii) of Example 4.5-3 obtained by using Airy’s stress function in polar coordinates.
The crack problem discussed in Example 4.6-4 can also be solved by the transform method. We assume that all
lengths are normalized with respect to the half crack length and its faces are subjected to a normal stress σyy = σ0 .
The objective is to calculate the stress field in the vicinity of the crack.
We first note that the problem is symmetric with respect to the x- and y-axes and that the domain of the associated
boundary-value problem can be reduced to the first quadrant, x > 0, y > 0. The symmetry about the y-axis implies
that the displacement component u and the stress component σxy must vanish on the y-axis and the symmetry about
the x-axis implies that v and σxy must vanish on the x-axis. Thus the symmetry and boundary conditions for the
problem can be stated as follows:
2005
c by Ajit Mal
4.7. TRANSFORM METHOD 141
Equations (v) are called dual integral equations, and they can be solved for C(ξ) by means of several techniques.
We present an indirect but simple approach through the use of the discontinuous properties of certain integrals involv-
ing Bessel functions. The reader is referred to texts on integral transform (e.g., Titchmarsh 1937; Sneddon 1951) for
more comprehensive treatments of dual integral equations. We recall the formulas (Abramowitz and Stegun, 1965 p
487)
Z ∞
J1 (ξ)
cos(ξx) dξ = 0, x>1 (via)
0 ξ
p
= 1 − x2 , 0<x<1 (vib)
Z ∞
J1 (ξ) cos(ξx) dξ = 1, 0<x<1 (viia)
0
1
= −√ √ , x>1 (viib)
x2 − 1(x + x2 − 1)
Comparing (v), (via) and (viia) we see that a possible solution of (v) is
Assuming that C(ξ) is given by (viii), the displacement and stress components v and σyy on y = 0 are found from
(iiia,b), (vib), and (viib) as
(1 − ν)σ0 p
v(x, ±0) = ± 1 − x2 , |x| < 1 (ixa)
µ
σ0
σyy = √ √ , |x| > 1 (ixb)
x − 1(x + x2 − 1)
2
2005
c by Ajit Mal
142 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
y
Mz
z x
The nature of the solution is the same as that obtained earlier in Example 4.6-4 by means of the complex variable
method. It should be noted that the dual integral equations (v) have many solutions, but only (viii) gives rise to the
square root type singularity in the stress field at the crack tip. Other solutions lead to singularities of higher order,
which are possible only in the presence of real forces or moments at the crack tip. The reader is referred to text books
on fracture mechanics (e.g., Broek 1982) and related topics (e.g., Sneddon and Lowengrub 1966) for a more detailed
account of crack problems. Application of the transform method to contact and related problems can be found in
Gladwell (1980).
Three-dimensional problems of elasticity are complicated by the fact that their governing equations are quite cum-
bersome, often containing very long expressions. Apart from their algebraic complexity, the general equations of
equilibrium are severely coupled, and even when they can be decoupled, the resulting equations are not easily solved.
We give a brief discussion of a few topics that are of interest in engineering applications.
We first consider the problem of the torsion of a cylinder of arbitrary cross section (Fig. 4-17). The curved surface S
of the cylinder is free of traction, and its end sections are subjected to in-plane stresses resulting in an axial torque Mz .
Then the cross sections will, in general, warp as they rotate, in contrast to the case of the circular cylinder considered
in Chap. 1.
Assume a displacement field in the form
where γ is a constant and the unknown function f (x, y) is the so-called Saint-Venant’s warping function. The nonva-
nishing stresses are
∂f ∂f
σzx = µγ −y , σzy = µγ +x (4.113)
∂x ∂y
Substituting (4.112) into the stress equations of equilibrium, we have
∇2 f = f,xx + f,yy = 0
i.e., f (x, y) is harmonic. The boundary condition on the curved surface S of the cylinder may be written in the form
∂f ∂f
l −y +m +x =0
∂x ∂y
where (l, m) is the unit normal on S. Since f is harmonic, there exists a conjugate harmonic function ψ related to f
through the Cauchy-Riemann equations, namely,
Then
∂2ψ ∂2ψ
∇2 ψ = + =0 (4.115)
∂x2 ∂y 2
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 143
∇2 Φ = −2 (4.120)
Φ = constant (4.121)
∂Φ
Z Z
Fx = σxz dA = µγ dA = 0 (4.123)
A A ∂y
The last part of (4.123) can be easily proved by using Stokes’ theorem (see Sec. 1.7). Similarly,
Fy = 0
x2 y2
+ =1 (i)
a2 b2
2005
c by Ajit Mal
144 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
a 2 b2
α=− (iii)
a2 + b2
From (4.124), we obtain the relationship between the constant γ and the applied torque Mz as
2µγa2 b2 x2 y2 a 3 b3
Z
MZ = 2 1 − − dA = πµγ (iv)
a + b2 A a2 b2 a 2 + b2
−2Mz y 2Mz x
σxz = , σyz =
πab3 πa3 b
The total shear stress τ on a cross section is
2 2 1/2
τ = (σxz + σyz )
2Mz 2
= [a − (a2 − b2 ) cos2 φ]1/2 (v)
πa2 b2
where x = a cos φ, y = b sin φ. From (v) we find that the maximum shear stress on the cross section occurs at φ = π/2
and φ = 3π/2, and its values can be obtained by putting cos φ = 0 in (v).
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 145
(-2a,0) (a,0) x
Consider a cylinder of cross section in the form of an equilateral triangle as shown as Fig. 4-19. The edges of
the triangle have the equations
p p
x − a = 0, x − 3y + 2a = 0, x + 3y + 2a = 0
Consider the cylinder of circular cross section with a semicircular groove shown as Fig. 4-20. The appropriate
Prandtl stress function for this cross section is
2 x 1 1
Φ=a x−b 2 + b2 − (x2 + y 2 )
x + y2 2 2
2005
c by Ajit Mal
146 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
r
Φ a
A x
b
or
2 2 2a cos φ
(r − b ) 1 − =0
r
which is made up of the two circles
r = b and r = 2a cos φ
The stress components can be calculated in a straightforward manner. The maximum shear stress is at the point A
(Fig. 4-20) and has the value
τmax = µγ(2a − b)
In the absence of body forces, the equilibrium equation (3.19b) may be written as
We assume that the displacement vector can be expressed in the form (Helmholtz decomposition)
u = ∇φ + ∇ × ψ, ∇·ψ =0 (4.126)
Then
∇ · u = ∇2 φ (4.127)
and Eq. (4.124) gives
Thus
(λ + µ)∇φ + µu = µΦ (4.128)
where Φ is a vector harmonic function. We therefore have
λ
u=Φ− 1+ ∇φ (4.129)
µ
Substituting this expression for u in (4.127), we find
λ
∇2 φ = ∇ · Φ − 1 + ∇2 φ
µ
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 147
x3
P
x2
o
x1
or
λ 1
2+ ∇2 φ = ∇ · Φ = ∇2 (R · Φ) (4.130)
µ 2
where R is the position vector. The general solution of (4.130) is
µ
φ= (Φ0 + R · Φ) (4.131)
2(λ + 2µ)
where
∇2 Φ0 = 0, ∇2 Φ = 0 (4.133)
Equation (4.131) gives the general solution of the Navier equation (4.124) in terms of a scalar harmonic function Φ0
and a vector harmonic function Φ and is known as Papkovitch-Neuber solution. Since there are four scalar functions
Φ0 and the three components of Φ involved in the representation of the vector u, not all of them are independent. In
Cartesian coordinates, each of the four functions Φ0 , Φi are harmonic. In curvilinear coordinate systems, in general,
only Φ0 will be harmonic, because the components of a harmonic vector need not be harmonic. We next consider
some applications of the Papkovitch-Neuber solution.
Consider the problem of a concentrated force P applied at the origin in an infinite elastic medium (Fig. 4-21). This is
known as Kelvin’s problem. We show next that this problem is solved by the Papkovitch-Neuber potentials
P
Φ0 = 0. Φ= (4.134)
4πµR
Clearly, Φ is harmonic for R > 0. Equations (4.131) and (4.133) yield
1 1
u= (3 − 4ν)P + 2 R(R · P) (4.135)
16πµ(1 − ν)R R
2005
c by Ajit Mal
148 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
We find
Pm 3xi xj xm
σij = − + (1 − 2ν)(−δij xm + δim xj + δjm xi ) (4.137)
8π(1 − ν)R3 R2
Note that ui and σij are singular at the origin; the singularity in the stress is of order R−2 , typical of three-
dimensional problems with a concentrated force. Moreover, both stress and displacement vanish at infinity. These
results should be contrasted with those for two-dimensional problems discussed in Example 4.6-2.
To show that the stress field (4.137) gives rise to the concentrated force P at the orgin, assume that the x3 -axis is
along the force; i.e., P = (0, 0, P3 ). Then, from (4.136),
x23
x1 x3 x2 x3 1
u1 = A 3 , u2 = A 3 , u3 = A (3 − 4ν) + 3 (4.138)
R R R R
where
P3
A=
16πµ(1 − ν)
Consider a sphere of radius R with its center at the origin. The traction at any point on the surface S of the sphere is
given by
σij xj
ti = σij nj =
R
Then, by (4.137),
x1 x3 x2 x3
t1 = −6µA 4
, t2 = −6µA 4 (4.139a)
R R
3x23
1 − 2ν
t3 = −2µA + 4 (4.139b)
R2 R
F1 = F2 = 0, F3 = −P3
Since the sphere is in equilibrium, the resultant of surface tractions on the outer surface of the sphere must be balanced
by the singularity at the origin. This shows that the singularity at the origin must be a concentrated force of magnitude
P3 acting in the positive x3 -direction.
If the force P acts at a point y instead of the origin, the displacement and stress components are obtained from
(4.136) and (4.137) on replacing xj by xj − yj and the distance R in these formulas is defined by
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 149
x1
O
x2
x3
where
1 (xi − yi )(xj − yj )
gij (x; y) = (3 − 4ν)δij + (4.142)
16πµ(1 − ν)R R2
It can be seen that gij (x; y) is the ith component of the displacement at x produced by a concentrated force of unit
magnitude acting at y in the jth direction. Clearly gij is a tensor of the second order. Thus gij is an elastostatic
Green’s tensor for the isotropic infinite elastic solid. It can be easily verified that the elastostatic Green’s tensor has
the following symmetry properties,
gij (x; y) = gij (y; x) = gji (x; y) (4.143)
The problem of a normal concentrated force acting on the surface of a homogeneous isotropic half-space is known
as Boussinesq’s problem. Let the half-space occupy the region x3 ≥ 0 and let a concentrated force of magnitude P
be acting on the surface of the half-space at the origin in the positive x3 -direction (Fig. 4-22). Then, the boundary
conditions are
σ31 = σ32 = 0, σ33 = −P δ(x1 )δ(x2 ) at x3 = 0 (4.144)
We have seen in the last section that for a concentrated force in an unbounded medium acting at the origin in the
x3 -direction, the Papkovitch-Neuber potentials are of the form
A1
Φ0 = Φ1 = Φ2 = 0, Φ3 = (4.145)
R
These yield [Eq. (4.137)]
3x23
µA1 xi
σ3i = − 1 − 2ν + (4.146)
2(1 − ν)R3 R2
Thus, at x3 = 0 (except at the origin), we have
(1 − 2ν)µA1 x1 (1 − 2ν)µA1 x2
σ31 = − , σ32 = − , σ33 = 0 (4.147)
2(1 − ν)r3 2(1 − ν)r3
where r2 = x21 + x22 . To remove the stresses σ31 and σ32 at x3 = 0, we consider another solution, for which the
potentials are
Φ0 = A2 ln(R + x3 ), Φ=0 (4.148)
2005
c by Ajit Mal
150 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
µA2 x1 µA2 x2
σ31 = , σ32 = , σ33 = 0 (4.150)
2(1 − ν)r3 2(1 − ν)r3
From equations (4.147) and (4.150), we note that, to ensure σ3i = 0 at x3 = 0, we must choose
A2 = (1 − 2ν)A1 (4.151)
3µA1 π ∞
rx33
Z
dr = P
1−ν 0 (r2 + x23 )5/2
P
A1 = (1 − ν)
πµ
Collecting results, the Papkovitch-Neuber potentials for the Boussinesq problem are
P
Φ0 = (1 − ν)(1 − 2ν) ln(R + x3 )
πµ
(1 − ν)P
Φ1 = Φ2 = 0, Φ3 =
πµR
R2 = r 2 + z 2
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 151
x1
O P
x2
x3
Therefore, ur > 0 if
z R
> (1 − 2ν)
R R+z
Thus, on putting z = R cos θ, we find that ur > 0 for θ < θ0 , where
Points of the half-space inside (outside) the cone θ = θ0 move away from (approach) the z-axis. In particular, the
points of the boundary plane move towards the origin which is the point of application of the force. For ν = 1/4, we
have θ0 ≈ 68o 30′ .
Corresponding to the displacements (4.153), the stresses are:
2
P 3r z R
σrr = − − (1 − 2ν)
2πR2 R3 R+z
(1 − 2ν)P z R
σφφ = −
2πR2 R R+z
3P z 3 3P rz 2
σzz = − 5
, σrz = − , σrφ = σzφ = 0
2πR 2πR5
The problem of the tangential concentrated force, P ǫ1 , acting on the surface of a homogeneous, isotropic, and elastic
half-space is known as the Cerruti’s problem (Fig. 4-23). The displacement in the half-space is of the form
∇2 ψ = 0, ∇4 γ = 0
The scalar potential ψ is known as Lame’s strain potential, and the vector potential γ is known as Galerkin’s
vector. As in Sec. 4.8.4, it can be shown that, if we choose
P (1 − 2ν)x1
ψ =
2π(R + x3 )
PR P (1 − 2ν)
γ = , γ2 = 0, γ3 = x1 ln(R + x3 )
4π(1 − ν) 4π(1 − ν)
2005
c by Ajit Mal
152 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
The displacement and stress components at any point (x1 , x2 , x3 ) of the half-space x3 ≥ 0 are given by
x21 x21
P R
u1 = 1 + 2 + (1 − 2ν) −
4πµR R R + x3 (R + x3 )2
P x 1 x2 1 1 − 2ν
u2 = 2
−
4πµR R (R + x3 )2
P x1 x3 1 − 2ν
u3 = +
4πµR R2 R + x3
3x21 2Rx22
P x1 1 − 2ν 2 2
σ11 = − 2 + R − x2 −
2πR3 R (R + x3 )2 R + x3
3x22 2Rx21
P x1 1 − 2ν 2 2
σ22 = − 2 + 3R − x1 −
2πR3 R (R + x3 )2 R + x3
3P x1 x23
σ33 = −
2πR5
3P x1 x2 x3 3P x21 x3
σ23 = − 5
, σ31 = − 5
2πR 2πR
2
2Rx21
P x2 3x1 1 − 2ν 2 2
σ12 = − 2 + −R + x1 +
2πR3 R (R + x3 )2 R + x3
Consider a spherical cavity under simple tension shown as Fig. 4-24. The geometry is such that a spherical coordinate
system R, θ, φ with its origin at the center of the sphere is the most convenient. Since the problem is axisymmetric, all
field quantities are independent of the azimuthal coordinate φ.
The conditions to be satisfied by the displacement and stress are as follows:
(ii) The stress vector must vanish on the surface of the sphere, i.e.,
(iii) The stress components must reduce to the applied remote stress, namely, σzz = σ0 , as R → ∞. Then
as R → ∞.
In order to solve this boundary value problem, we assume representation of the displacement in the form
where
∇2 Φ = 0, ∇2 Ψ = 0 (4.158b)
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 153
Φο
o y
Φο
Fig. 4-24. Spherical hole in an infinite solid under uniaxial stress.
Then u satisfies Navier’s equation and its components are related to the potentials Φ and Ψ by
∂Φ ∂Ψ
2µuR = + cos θ R − (3 − 4ν)Ψ (4.159a)
∂R ∂R
1 ∂Φ ∂Ψ
2µuθ = + cos θ + (3 − 4ν)Ψ sin θ (4.159b)
R ∂θ ∂θ
The component uφ can be assumed to be zero without any loss of generality. (symmetry)
The general axisymmetric solution of Laplace’s equation in spherical coordinates is of the form
∞
X cn
bn R n + Pn (cos θ) (4.160)
n=0
Rn+1
3x2 − 1
P0 (x) = 1, P1 (x) = x, P2 (x) = (4.161)
2
Since the stress components must satisfy (4.157) for large R, the potentials must be of the form
c0 P2 (cos θ) P1 (cos θ)
Φ= + c2 , Ψ = c1 (4.162)
R R3 R2
where c0 , c1 , and c2 are unknown constants to be determined from the boundary conditions (4.156). The displacement
components can be found from (4.159) and the stress components from (3.17). We omit the algebraic details of these
2005
c by Ajit Mal
154 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
3(1 + 5ν)
σxx (0, 0, ±a) = − σ0 = −K(ν)σ0 (4.164a)
2(7 − 5ν)
3(1 + 5ν)
K= (4.164b)
2(7 − 5ν)
It can be seen that K > 1 if ν > 11 25 ; i.e., there is stress concentration at the pole of the sphere. This should be
contrasted with the corresponding two-dimensional problem discussed in Example 4.5-7, where the value of K is 1
for all values of ν.
The stress components on the equatorial plane θ = π/2 are
6σ0 a 3 a 5
σRR = − (4.165a)
7 − 5ν R R
4 − 5ν a 3 9 a 5
σθθ = σ0 1 + + (4.165b)
2(7 − 5ν) R 2(7 − 5ν) R
σ0 a 3 a 4
σφφ = −3(2 − 5ν) +3 (4.165c)
2(7 − 5ν) R R
Thus the stress concentration factor on the spherical cavity depends upon the value of ν. In particular, if ν = 0.3, then
and its maximum value is 2.046σ0 on the equatorial circumference, θ = π/2. Hence the stress concentration factor is
2.046, as compared to 3 for the corresponding two-dimensional problem discussed in Example 4.5-7. In general, the
stress concentration at a hole is more severe for two-dimensional cases than that for three-dimensional cases.
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 155
Problems
4.1. Prove the validity of the relations (4.34) and (4.35) for Cartesian coordinates and (4.43)-(4.45) as well as (4.49)-
(4.53) for polar coordinates.
4.2. Show that for plane stress, the stress-strain relations may be expressed in the form
E
σ11 = (ǫ11 + νǫ22 )
1 − ν2
E
σ22 = (ǫ22 + νǫ11 )
1 − ν2
E
σ12 = ǫ12
1+ν
4.3. Show that the Navier equations may be expressed in the form
2
∂ 2 ui
∂ ∂u1 ∂u2 ∂ ui
(λ + µ) + +µ + + fi = 0 (i = 1, 2)
∂xi ∂x1 ∂x2 ∂x21 ∂x22
for plane strain and
∂ 2 ui ∂ 2 ui
∂ ∂u1 ∂u2
(λ + µ) + +µ + + fi = 0 (i = 1, 2)
∂xi ∂x1 ∂x2 ∂x21 ∂x22
for plane stress, where
2λν
λ=
λ + 2µ
4.4. Show that any problem of a plane state of strain may be solved as a problem in a plane state of stress after
replacing the true value of the Poisson ratio ν by the ”apparent value” ν̂ = ν/(1 + ν), keeping the shear
modulus µ unchanged. Conversely, any plane stress problem may be solved as a problem of plane strain by
replacing the true value of ν by an apparent value ν/(1 − ν), keeping µ unchanged.
4.5. Solve the problem of a tangential line load at the surface of a semi-infinite elastic medium y > 0, by considering
an Airy stress function of the form
P rφ
Φ= cos φ
π
Show that the stresses and displacements resulting from a tangential line load, P , acting at the origin in the
positive x-direction are given by
2P x3 2P y 2 2νP x
σxx = − , σyy = − , σzz = −
πr4 πr4 πr2
2P x2 y
σxy = − , σyz = σzx = 0
πr4
2πµu d y2
= 2(1 − ν) ln − 2
P r r
xy −1 y
2πµv r 2 − (1 − 2ν)
tan x, x>0
= xy −1 y
P r 2 − (1 − 2ν) tan x + π , x<0
Also show that for a normal line load, Q,
2πµu xy x
= − (1 − 2ν) tan−1 , x > 0
Q r2 y
2
2πµv x r
= − 2 − 2(1 − ν) ln
Q r d
2Qx2 y 2Qy 3
σxx = − 4
, σyy = − ,
πr πr4
2Qxy 2
σxy = −
πr4
2005
c by Ajit Mal
156 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
4.6. The cross section of a cylinder is the region |x| ≤ a, |y| ≤ b. Examine the stress system in the cylinder corre-
sponding to the complex functions
1 1
F (z) = T z, G(z) = iSz 2
2 2
where z = x + iy and S, T are real. Also find the displacement components.
Ans.:
4.7. Show that the stress components σrr , σrφ and the displacement components ur , uφ in polar coordinates may be
expressed in the form
z
σrr + iσrφ = F ′ (z) + F i(z) − zF ”(z) − G”(z)
z
2µ(ur + iuφ ) = κǫ−iφ F (z) − rF ′ (z) − ǫ−iφ G′ (z)
4.8. Show that a thermoelastic plane strain problem parallel to the xy-plane can be solved in terms of the Airy stress
function Φ, defined as follows
αEΘ ∂2Φ
σxx − =
1 − 2ν ∂y 2
αEΘ ∂2Φ
σyy − =
1 − 2ν ∂x2
∂2Φ
σxy = −
∂x∂y
αE 2 ∂2 ∂2
∇2 ∇2 Φ = − ∇ Θ, ∇2 ≡ +
1−ν ∂x2 ∂y 2
4.9. Show that in the case of plane stress, the equations of Problem 4.8 are to be modified as follows:
αEΘ ∂2Φ
σxx − =
1−ν ∂y 2
αEΘ ∂2Φ
σyy − =
1−ν ∂x2
∂2Φ
σxy = −
∂x∂y
and
∂2 ∂2
∇2 ∇2 Φ = −αE∇2 Θ, ∇2 ≡ +
∂x2 ∂y 2
4.10. In the plane strain problem, let the body force be radial and be derived from the potential V (r) such that
∂V
fr = − , fφ = 0
∂r
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 157
Show that the equilibrium equations in polar coordinates are identically satisfied if we take
1 ∂Φ 1 ∂2Φ
σrr = + 2
r ∂r r ∂φ2
2
∂ Φ ∂V
σφφ = +r
∂r2 ∂r
∂ 1 ∂Φ
σrφ = −
∂r r ∂φ
and if the Airy stress function in polar coordinates, Φ(r, φ), satisfies the equation
1 − 2ν 2
∇2 ∇2 Φ + ∇ V =0
1−ν 1
where
∂2 1 ∂
∇21 ≡ 2
+
∂r r ∂r
4.11. Obtain approximate expressions for the stresses within the loaded arch of Fig. 4-4 using beam theory, i. e.,
under the assumption that plane sections remain plane. Compare results with those given in equations (x) and
(xi) of Example 4.5-2; through analytical and numerical calculations.
4.12. For the corner problem, numerically evaluate and plot the values of β as a function of the wedge angle φ0 from
equations (xva) and (xvb) of Example 4.5-6. Discuss the nature of stress singularity at the corner.
4.13. Beginning with the expressions for the stresses given in equation (vi) of Example 4.7-1, show that the three
principal stresses at any point (x, y) are given by
−P0 p0 2νp0 α
σ1 = (α + sin α), σ2 = − (α − sin α), σ3 = −
π π π
where α = θ2 − θ1 . Further show that the direction of the principal stress σ1 bisects the angle between the radii
r1 and r2 shown in Fig. 4-17.
4.14. Assume
2 ∂Γ
2µu = 2(1 − ν)ǫz ∇ Γ − ∇
∂z
Show that the Navier’s equation with body force (0, 0, Z) is satisfied if the potential Γ (known as Love’s strain
function) satisfies the equation
Z
∇4 Γ = −
1−ν
2005
c by Ajit Mal
158 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
Further show that the cylindrical components of the displacement and stress are given by
∂2Γ 1 ∂2Γ
2µur = − , 2µuφ = −
∂r∂z r ∂φ∂z
∂2Γ
2µuz = 2(1 − ν)∇2 Γ − 2
∂z
2
∂ ∂ Γ
σrr = ν∇2 Γ − 2
∂z ∂r
1 ∂2Γ
∂ 1 ∂Γ
σφφ = ν∇2 Γ − − 2 2
∂z r ∂r r ∂φ
2
∂ ∂ Γ
σzz = (2 − ν)∇2 Γ − 2
∂z ∂z
∂2Γ
1 ∂ 2
σφz = (1 − ν)∇ Γ − 2
r ∂φ ∂z
2
∂ ∂ Γ
σzr = (1 − ν)∇2 Γ − 2
∂r ∂z
3
∂ Γ
σrφ = −
∂r∂φ∂z r
4.15. Show that the Kelvin problem can be solved by taking Love’s strain potential of the form
Γ = B(r2 + z 2 )1/2
To determine B, consider a cylinder with bases at z = ±a. Since the cylinder is in equilibrium, the resultant of
the surface tractions on its outer surface must balance the load P acting at the origin. Show that this yields
P
B=
8π(1 − ν)
4.16. Show that for a concentrated force of magnitude P in an unbounded medium acting at the origin in the positive
z-direction, the displacement and stress components in spherical coordinates are
4.17. Show that for a concentrated force of magnitude P in an unbounded medium acting at the origin in the positive
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 159
P
uR = sin θ cos φ
4πµR
P 3 − 4ν
uθ = cos θ cos φ
16πµR 1 − ν
P 3 − 4ν
uφ = − sin φ
16πµR 1 − ν
P 2−ν
σRR = − sin θ cos φ
4πR2 1 − ν
P 1 − 2ν
σθθ = σφφ = sin θ cos φ
8πR2 1 − ν
P 1 − 2ν
σRθ = − cos θ cos φ
8πR2 1 − ν
P 1 − 2ν
σRφ = sin φ, σθφ = 0
8πR2 1 − ν
4.18. The results for the Cerruti problem when the force acts in the x2 -direction can be obtained from (4.154) and
(4.155) on interchanging x1 and x2 throughout. Combine the Boussinesq solution and the Cerruti solutions to
show that if a concentrated force (P1 , P2 , P3 ) acts at the origin on the surface of a uniform half-space (x3 ≥ 0),
the displacement components are given by
x21 x21
1 1 − 2ν
4πµu1 = P1 + + 1−
R R3 R + x3 R(R + x3 )
x1 x2 x1 x2
+ P2 − (1 − 2ν)
R3 R(R + x3 )2
x1 x2 x1
+ P3 − (1 − 2ν)
R3 R(R + x3 )
x1 x2 x1 x2
4πµu2 = P1 − (1 − 2ν)
R3 R(R + x3 )2
2
x22
1 x2 1 − 2ν
+ P2 + + 1−
R R3 R + x3 R(R + x3 )
x2 x3 x2
+ P3 − (1 − 2ν)
R3 R(R + x3 )
x1 x3 x1
4πµu3 = P1 + (1 − 2ν)
R3 R(R + x3 )
x2 x3 x2
+ P2 + (1 − 2ν)
R3 R(R + x3 )
2
1 x3
+ P3 2(1 − ν) + 3
R R
4.19. Work out the details of the spherical cavity problem discussed in Sec. 4.8.6.
2005
c by Ajit Mal
160 CHAPTER 4. SOLUTION OF LINEAR ELECTROSTATIC PROBLEMS
4.20. For a distributed normal load P (x′ , y ′ ) over area S of a half-space z > 0, show that
(x − x′ )P (x′ , y ′ ) ′ ′
Z Z
2πu(x, y, 0) = −A dx dy
R2
Z ZS
(y − y ′ )P (x′ , y ′ ) ′ ′
2πv(x, y, 0) = −A dx dy
R2
Z Z S ′ ′
P (x , y ) ′ ′
2πw(x, y, 0) = B dx dy
S R
where
R2 = (x − x′ )2 + (y − y ′ )2
1 − 2ν (1 + ν)(1 − 2ν) 1−ν 2(1 − ν 2 )
A = = , B= =
2µ E µ E
4.21. Calculate the displacement and stress components for a spherical inclusion in an infinite solid subjected to
uniaxial remote stress σ0 along the z-axis.
A cos θ B sin θ
uR = uθ = uφ = 0
R R
where R, θ, φ are the spherical coordinates of the field point. Determine A and B such that
(a) (1) is a possible field in R > a in the absence of body forces and
(b) the resultant force exerted on the sphere R = a is P along the z-axis. Discuss the far-field behavior of the
stress components.
BIBLIOGRAPHY
1. Abramowitz, M. and Stegun, I.(ed.), Handbook of Mathematical Functions, Dover, New York, 1965.
2. Broek, D., Elementary Engineering Fracture Mechanics, Martinus Nijhoff, Hague, 1982.
3. England, A. H., Complex Variable Methods in Elasticity, John Wiley, New York, 1971.
4. Fung, Y. C., Foundations of Solid Mechanics, Prentice Hall, Englewood Cliffs, N.J., 1965.
5. Gladwell, G. L. M. Contact Problems in the Classical Theory of Elasticity, Sijthoff and Nordhoff, Germantown,
M.D., 1980.
6. Green, A. E. and Zerna, W., Theoretical Elasticity, Oxford University Press, London, 1960.
7. Landau, L. D. and Lifshitz, E.M., Theory of Elasticity, Pergamon, New York, 1959.
9. Love, A. E. H., A Treatise on the Mathematical Theory of Elasticity (4th ed.), Cambridge University Press,
Cambridge, 1927.
10. Lur, A. I., Three-Dimensional Problems of the Theory of Elasticity, Inter- science, New York, 1964.
11. Muskhelishvili, N. I., Some Basic Problems of the Mathematical Theory of Elasticity, P. Noordhoff Ltd., 1969.
13. Snedden, I. N., The Use of Integral Transform, McGraw-Hill, New York, 1972.
2005
c by Ajit Mal
4.8. THREE-DIMENSIONAL PROBLEMS 161
14. Sneddon, I. N. and Lowengrub, M., Crack Problems in The Classical Theory of Elasticity, SIAM Monograph,
John Wiley, New York, 1966.
15. Sokolnikoff, I. S., Mathematical Theory of Elasticity,(2d ed.), McGraw-Hill, New York, 1956.
16. Timoshenko, S. P. and Goodier, J. N., Theory of Elasticity, McGraw-Hill, New York, 1951.
17. Titchmarsh, E. C., Introduction to the Theory of Fourier Integrals (2d ed.), Oxford University Press, London,
1937.
18. Williams, M. L., ”Stress singularities resulting from various boundary conditions in angular corners of plates in
extension,” J. Appl. Mech. 19, 1952, pp. 526-528.
2005
c by Ajit Mal
162
2005
c by Ajit Mal
Part II
Nonlinear Elasticity
163
Chapter 5
Kinematics of Deformation
The mathematical treatment of the mechanics of solids is based on the assumption that the material is distributed
throughout some region of space. At every instant of time, every point in the region is occupied by a small element of
the solid, which we call a material particle. The dimensions of the particles are small compared to all characteristic
lengths in any given situation, so they can be idealized as mathematical points but are nevertheless large compared
to atomic dimensions. In contrast to rigid bodies, application of external loads to real bodies causes changes in the
relative positions of its material particles, resulting in deformation of the body. In this section, we discuss how the
motion of the material particles in a continuously deforming solid can be quantitatively described and introduce a
number of alternative measures that can be used to analyze the deformation of infinitesimal elements of the solid.
In order to quantify the motion of the material particles, it is necessary to have some means of labeling them. To
this end, each particle is assumed to occupy a unique position in a three-dimensional Euclidean space at each instant of
time. The region of space Rt occupied by the particles is called the configuration of the solid at time t. A convenient
known configuration R0 of the body is chosen as the reference configuration, and each particle in this configuration
is identified by its coordinates XK (K = 1, 2, 3) referred to a suitably chosen Cartesian vector basis {ei }. Then,
the triplet XK or X can be regarded as a label by which a particle P , which occupies this position in the reference
configuration, can be identified at all subsequent times. It should be noted that the reference configuration need not
be one actually assumed by the solid during its motion, although we shall assume that R0 is its natural or undeformed
state, i.e., the configuration assumed by the body in the absence of any external loads.
After loads are applied, motion accompanied by deformation of the solid occurs, and the material particles initially
in R0 are carried continuously through various spatial positions. At time t, the spatial position of P will be denoted
by its coordinates xk (k = 1, 2, 3), where x ∈ Rt (Fig. 5-1). We shall use the same vector basis {ei } to specify
the positions of the particle in the reference and current configurations, although separate systems can also be used.
X1 , X2 , and X3 are called the material, or Lagrangian, coordinates and x1 , x2 , and x3 , the spatial, current, or Eulerian
coordinates.
We assume that the configuration of the solid at time t can be expressed by a functional relation of the form
where x ∈ Rt , X ∈ R0 , and t > 0. The functions xk are continuous and differentiable with respect to X1 , X2 , X3 ,
and t for all X ∈ R0 and t > 0. It is further assumed that the mapping (5.1a) is one-to-one, so that a unique inverse
exists in the form
XK = XK (x, t) or X = X(x, t) (5.1b)
in the neighborhood of each point x ∈ Rt . Then, the Jacobian of the transformation (5.1a) given by
J = det[xk,K ] (5.2)
165
166 CHAPTER 5. KINEMATICS OF DEFORMATION
x3 X x
P R pt
R0
o
x2
x1
Fig. 5-1. The material and spatial coordinates.
Example 5.1-1
xk = akK XK + bk
where
t 1 0
[akK ] = 0 t 1 = A
0 −1 1
2005
c by Ajit Mal
5.2. DISPLACEMENT, VELOCITY, AND ACCELERATION FIELDS 167
(c) The temperature field within the solid is given by its material description F (X, t) = aX1 + b, where a, b are
constants. Obtain the spatial description of the temperature field.
XK = CKk xk + dK
where
1 + t −1 1
1
C = A−1 = 0 t −t
t(t + 1)
0 t t2
1 1
d1 = − 1, d2 = d3 = −
t(t + 1) 1+t
The displacement of the material particle P due to the motion described by (5.1) is the vector x − X with components
x1 − X1 , x2 − X2 , x3 − X3 in the vector basis {ei }. These can be expressed in terms of the material coordinates as
UI (X, t) or in terms of the spatial coordinates as ui (x, t), where the indices I and i take on the values 1, 2, 3. The
velocity of P at time t is obtained by the instantaneous time rate of change of its vector position. The components of
this velocity are given by (∂/∂t)x1 (X, t), (∂/∂t)x2 (X, t), (∂/∂t)x3 (X, t), where the differentiations are carried out
by holding the material coordinate X fixed. These derivatives are called the material derivatives, and the associated
differentiation operation is denoted by the symbol (D/Dt). The velocity components can be expressed in terms of the
material coordinates as VI (X, t) or in terms of the spatial coordinates as vi (x, t).
The material description of the acceleration of P is simply given by
D
A(X, t) = V(X, t) (5.4)
Dt
It is, however, convenient and often necessary to express the acceleration components in terms of the spatial description
of the velocity components vi (x, t). This can be easily obtained by using the chain rule as
D ∂ ∂ ∂
= + vk = + (v·∇) (5.6)
Dt ∂t ∂xk ∂t
2005
c by Ajit Mal
168 CHAPTER 5. KINEMATICS OF DEFORMATION
The first term in the right-hand side of (5.5) is sometimes called the local, or nonstationary, time rate and the second
term is called the convective time rate.
It should be noted that in the Lagrangian description, the material particle with a specified displacement, velocity,
or acceleration is immediately identifiable. In the Eulerian description, the displacement, velocity, and acceleration
at time t at a given spatial position are known but the particle that occupies that position is not known explicitly; as
each particle moves through this spatial point, it acquires the displacement, velocity, and acceleration associated with
the point. The displacement, velocity, and acceleration fields associated with the motion are given by the functions
U(X, t), V(X, t), A(X, t), where X ∈ R0 , t > 0 or by the functions u(x, t), v(x, t), a(x, t), where x ∈ Rt , t > 0.
Example 5.2-1
xk = Xk + ctδk1
where c is a constant and k = 1, 2, 3. A physical property (such as the temperature) of the body is described by the
function F (XK , t), where
F (XK , t) = aX1 + b
X1 = x1 − ct, X2 = x 2 , X3 = x 3
Hence,
∂f
f (xk , t) = a(x1 − ct) + b, = −ac
∂t
and
Df ∂f ∂f ∂x1
= + = −ac + ac = 0
Dt ∂t ∂x1 ∂t
Example 5.2-2
DJ
Show that Dt = J(∇·v) = Jvi,i
Solution.
Thus,
DJ D D
= ǫLM N x1,L x2,M x3,N + ǫLM N x1,L x2,M x3,N
Dt Dt Dt
D
+ ǫLM N x1,L x2,M x3,N
Dt
2005
c by Ajit Mal
5.3. THE DEFORMATION GRADIENT TENSOR 169
Now,
D ∂ ∂ ∂
x1,L = v1 = v1 xl = v1,l xl,L
Dt ∂XL ∂xl ∂XL
and, therefore,
D
ǫLM N x1,L x2,M x3,N = ǫLM N v1,l xl,L x2,M x3,N
Dt
It can be easily verified that
so that
D
ǫLM N x1,L x2,M x3,N = v1,1, ǫLM N x1,L x2,M x3,N = v1,1 J
Dt
and the result follows.
In this section, we discuss how the deformational properties of the solid in the neighborhood of each of its material
points can be quantitatively analyzed for a given motion. We assume that the material particle P occupying the position
XK in the reference configuration R0 is displaced to the spatial position xk in the current configuration Rt , where xk
and XK are related through (5.1). Then, a neighboring particle located at XK + dXK in R0 occupies the position
xk + dxk in Rt , where
where J is the Jacobian introduced earlier. Since J 6= 0 by (5.2), the tensor F is nonsingular for all physically possible
motions described by (5.1). Then, the unique inverse of (5.7a) is
2005
c by Ajit Mal
170 CHAPTER 5. KINEMATICS OF DEFORMATION
is necessary to identify and, if possible, to separate out the part of F that describes the rigid translation and rotation of
the element. This can be accomplished through the use of the polar decomposition theorem discussed in Sec. A.5.
Since F is nonsingular, it can be uniquely decomposed in the form
F = RU (5.10)
or as
Equations (5.12b) describe two successive transformations of the line element dX, and each can be given a simple
geometric interpretation.
In (5.12b) the tensor U is a real positive-definite and symmetric tensor, and so it has three real positive eigenvalues
u1 , u2 , u3 and three mutually orthogonal eigenvectors, which are the principal axes of U. We assume, for this discus-
sion, that the base vectors e1 , e2 , e3 coincide with the principal axes of U. Then U becomes the diagonal tensor with
the elements u1 , u2 , u3 along the diagonal (see Sec. 2.5) and (5.12b) becomes
This shows that the transformation of the line element dX to dy is due to compression (if ui < 1) or stretching (if
ui > 1) along the principal axes of U. In general, the direction cosines of dX and dy are different, so that the line
element dX changes its orientation as well as its length due to the triaxial nature of the compression or stretching. The
quantities u1 , u2 , u3 are called the principal stretches of the deformation. On the other hand, in (5.12), the tensor R is
proper orthogonal, so that it describes a rigid rotation of dy to dx.
Thus the transformation described by (5.12) and hence by (5.7) may be thought of as a stretching operation of dX
along three mutually perpendicular axes whose magnitude, sign, and directions are entirely specified by the tensor U,
followed by a rigid rotation specified by the tensor R (Fig. 5-2). Clearly, if U = I, the identity tensor, then F = R and
the line element dX is rotated as a rigid rod without undergoing any deformation whatsoever.
The deformation gradient tensor F can be alternatively decomposed in the form
F = VR (5.13)
The transformation of the line element by (5.7) can then be interpreted as a rigid rotation characterized by R, followed
by a stretching operation along three mutually perpendicular axes specified by V. It should be noted that the orthogonal
tensor R defined by (5.11b) and (5.14b) are identical. Further, the eigenvalues of U and V can be shown to be identical.
2005
c by Ajit Mal
5.4. DEFORMATION OF VOLUME AND SURFACE ELEMENTS 171
dX Rt
P
e3 R0 dX dx
X p
x
o e2
e1
dS N
dX 1 dX 2
e3 dX 3 Ro ds n
dx2
dx 1
Rt
o e2
dx3
e1
In this section we describe how the transformation of surface and volume elements of the reference configuration R0
into those in the spatial configuration can be expressed in terms of the components of the deformation gradient tensor
F and its inverse G.
Consider an arbitrary parallelepiped element with edges dX(1) , dX(2) , dX(3) in the reference configuration R0 (Fig.
5-3). Let dV be the volume of the element and dS be the area of the surface element formed by dX(1) and dX(2) , with
unit normal N. Then,
(1) (2) (3)
dV = dX(3) · (dX(1) × dX(2) ) = ǫIJK dXI dXJ dXK (5.15)
Also,
(3)
dV = dSN · dX(3) = dSI dXI = dST dX(3) (5.16a)
where
dSI = NI dS, dS = NdS (5.16b)
Let the body be subjected to the deformation (5.1a), resulting in the transformation of dX(K) to dx(K) , dV to
dv, dS to ds, and N to n (Fig. 5-3). Then,
(1) (2) (3)
dv = ǫijk dxi dxj dxk (5.17)
2005
c by Ajit Mal
172 CHAPTER 5. KINEMATICS OF DEFORMATION
and
(3)
dv = dsi dxi = ds(3) · dx(3) (5.18a)
where
dsi = ni ds, ds = nds (5.18b)
By (5.7) and (5.17),
(1) (2) (3)
dv = ǫijk xi,I xj,J xk,K dXI dXJ dXK
(1) (2) (3) (k) (k)
= ǫIJK J dXI dXJ dXK = JdV dxi = xi,I dXI
(ǫijk ail ajm akn = |a|ǫlmn by Example 1.3 − 3)
or
1 T
ni ds = JXI,i NI dS, {n}ds = J[GT ]{N }dS, {N }dS = [F ]{n}ds (5.21b)
J
q
T 2 2 T T
2 T
{n) {n} ds = J {N } [G] G {N } dS ds = JdS {N } [G] [GT ] {N } (5.21c)
Equation (5.21b)is known as the Nanson’s theorem. Equation (5.19) gives the deformed volume of an arbitrarily
oriented element, and (5.21) describes the change in an arbitrarily oriented surface element within the solid due to the
deformation specified by (5.1). If J = 1, there is no change in volume of the element, and the deformation is called
isochoric.
The preceding discussions imply that the tensor field F(X, t) provides a complete description of the deformation of
the solid in the neighborhood of its material points. The stretching tensor fields U and V also provide measures of the
deformation and are better amenable to analysis because of their special properties. Five other related second order
symmetric tensors denoted by B, C, c, E, and e can be used as alternative measures of the deformation of the solid.
These are defined by
B = FFT = V2 , C = FT F = U2 (5.22a)
c = B−1 = GT G = (V−1 )2 (5.22b)
1 1
E = (C − I), e = (I − c) (5.22c)
2 2
Note that
B2 = FCFT , C2 = FT BF (5.22d)
In component form, we have
2005
c by Ajit Mal
5.5. THE STRAIN TENSORS 173
1
EIJ = (CIJ − δIJ ) (5.24a)
2
1
eij = (δij − cij ) (5.24b)
2
Clearly, B, C and c are positive-definite and they, as well as E and e, are more easily calculated from a knowledge
of F than that of U and V. Equations (5.23) and (5.24) imply that B, C, and E should be expressed in the material
coordinates X, and c and e should be expressed in the spatial coordinates x.
The tensor B is called the left Cauchy-Green tensor, C is the right Cauchy-Green tensor, c is the Cauchy’s strain
tensor, E is the Lagrangian, or Green’s, strain tensor, and e is the Eulerian, or Almansi’s, strain tensor.
The components of C, c, E, and e can be related to certain geometric properties of the deformation in the neighborhood
of the material particles of the solid. To see this, we first consider the squared lengths (dL)2 and (dℓ)2 of the line
elements dX and dx in the reference and spatial configurations, respectively (Fig. 5-2). Clearly
dL2 = dXK dXK = XK,i XK,j dxi dxj = cij dxi dxj (5.25)
2
dℓ = dxk dxk = xk,I xk,J dXI dXJ = CIJ dXI dXJ (5.26)
and thus,
Thus, a knowledge of the tensors C or E enables us to calculate the change in the squared length of a line element
of known length and orientation in the reference configuration. Similarly, a knowledge of the tensors c or e enables
the calculation of the change in the squared length of a line element of known length and orientation in the spatial
configuration.
From (5.26),
s
dℓ CIJ dXI dXJ p
= 2
= CIJ NI NJ
dL (N ) (dL)
s
(dℓ)2 1
= =√ (5.28)
(cij dxi dxj ) cij ni nj
where N and n are unit vectors along dX and dx, respectively. The length ratio (dℓ/dL) is called the stretch in the
direction N or n and is denoted by L(N ) or ℓ(n) . Thus,
p
L(N ) = CIJ NI NJ (5.29a)
1
ℓ(n) = √ (5.29b)
cij ni nj
By choosing the line elements in the directions of the base vectors, we have
p p p
L(1) = C11 , L(2) = C22 , L(3) = C33
1 1 1
ℓ(1) = √ , ℓ(2) = √ , ℓ(3) = √ (5.30)
c11 c22 c33
Hence the diagonal elements of the strain tensor C are the squares of the stretches of line elements initially aligned
with the coordinate axes. Similarly, the diagonal elements of the strain tensor c are the inverse squares of the stretches
of the line elements aligned with the coordinate axes in the current configuration.
2005
c by Ajit Mal
174 CHAPTER 5. KINEMATICS OF DEFORMATION
The change in length per unit length of a line element will be denoted by E(N ) for an element directed along N in
the reference configuration and by e(n) for an element directed along n in the spatial configuration. Then,
E(N ) = L(N ) − 1 (5.31a)
e(n) = ℓ(n) − 1 (5.31b)
and so, for elements oriented along the axes,
p p
E(1) = C11 − 1 = 1 + 2E11 − 1 (5.32a)
p p
E(2) = C22 − 1 = 1 + 2E22 − 1 (5.32b)
p p
E(3) = C33 − 1 = 1 + 2E33 − 1 (5.32c)
1 1
e(1) = √ −1= √ −1 (5.33a)
c11 1 − 2e11
1 1
e(2) = √ −1= √ −1 (5.33b)
c22 1 − 2e22
1 1
e(3) = √ −1= √ −1 (5.33c)
c33 1 − 2e33
Equations (5.32) and (5.33) give the relationship between the stretches and the diagonal elements of the strain tensors
E and e.
We now consider two perpendicular line elements dX1 e1 and dX2 e2 in the reference configuration. These are
transformed into the line elements dx(1) and dx(2) in the current configuration with components given by
(1) (2)
dxk = xk,1 dX1 = Fk1 dX1 , dxk = xk,2 dX2 = Fk2 dX2
Since the strain tensors C, c, E, and e are real-valued and symmetric second-order tensors, they can be diagonalized
by proper choice of axes in accordance with the discussions presented in Sec. 2.5. Considering the tensor C first, its
three real eigenvalues are the roots of the eigenvalue equation
λ3 − IC λ2 + IIC λ − IIIC = 0 (5.37)
2005
c by Ajit Mal
5.6. HOMOGENEOUS DEFORMATIONS 175
IC = CKK (5.38a)
1
IIC = (CKK CLL − CKL CKL ) (5.38b)
2
IIIC = |C| (5.38c)
IC = C1 + C2 + C3 (5.39a)
IIC = C2 C3 + C3 C1 + C1 C2 (5.39b)
IIIc = |C| (5.39c)
It has been shown in Theorem 2.5-4 that if the vector basis {ei } is rotated so that the base vectors are aligned
with the eigenvectors of C, then the off-diagonal terms of C vanish and it is transformed into the diagonal tensor
diag[C1 , C2 , C3 ]. This implies that a rectangular parallelepiped element whose edges are aligned with the principal
axes of C in the reference configuration√ remains
√ rectangular
√ after deformation, but the lengths of its edges are changed
(increased or decreased) by the factors C1 , C2 , C3 .
Similar remarks apply to the tensors c, E, and e except for the fact that whereas the eigenvalues of C and c are
always positive, this is not necessarily true for the eigenvalue of E and e.
The following relationships between the principal invariants of the strain tensors can be easily established:
The invariants are useful in the development of the constitutive equations for certain special solids, as will be seen
later.
A deformation in which the tensors F, C, E, c, and e are the same at all points within the solid is known as a homoge-
neous deformation. Then, the current position of the particles can be expressed in the form
xk = akK XK + bk (5.42)
where akK and bk are independent of XK . If the deformation is static, then akK and bk are independent of time t as
well. The kinematic properties of the static deformation of two special cases of (5.45) are discussed next.
1. Let
[akK ] = diag[a1 , a2 , a3 ], bk = 0
ı.e.,
x 1 = a1 X1 , x 2 = a2 X2 , x 3 = a3 X3
Clearly,
F = diag[a1 , a2 , a3 ]
2005
c by Ajit Mal
176 CHAPTER 5. KINEMATICS OF DEFORMATION
and
Then
J = a1 a2 a3
and the condition J > 0 implies that either one or all of a1 , a2 , a3 must be positive. Since C is diagonal, a
rectangular parallelepiped with edges aligned with the coordinate axes remains rectangular; the lengths of the
edges are increased or decreased by the factors a1 , a2 , a3 in the three axial directions. The invariants of C are
given by
IC = a21 + a22 + a23 , IIC = a22 a23 + a23 a21 + a21 a22 , IIIC = a21 a22 a23 (5.43)
If the constants a1 , a2 , a3 are all positive then F is positive-definite and symmetric so that
R = I, U=V=F
and there is no rigid-body motion. On the other hand, if a1 > 0 and a2 , a3 < 0, then
and the motion given by (5.45) includes a rigid rotation of the body about the 1-axis through an angle π.
If the deformation is isochoric, then a1 a2 a3 = 1. In addition, if a1 > 1, then there is an extension along the
1-direction, and if a2 = a3 , then the extension in the 1-direction must be accompanied by a contraction in the
2- and 3-directions.
If a1 = a2 = a3 = a, then
F = aI, C = a2 I and J = a3
and J > 0 implies that a must be positive. If a > 1, the deformation is uniform expansion in all directions, and
if a < 1, it is a uniform contraction. Deformations of this type are called dilatations.
2. Let a11 = a22 = a33 = 1, a12 = Λ, all other elements of A be zeros, and bk = 0. Thus,
x1 = X1 + ΛX2 , x 2 = X2 , x 3 = X3 (5.44)
where Λ is a constant.
The deformation and strain tensors are given by
1 Λ 0 1 Λ 0 1 −Λ 0
F = 0 1 0 , C = Λ 1 + Λ2 0 , c = −Λ 1 + Λ2 0
0 0 1 0 0 1 0 0 1
(5.45)
Clearly, J = 1 so that the deformation is isochoric. Line elements initially aligned with√the 1- and 3-axes are
unchanged in their lengths, while those aligned with the 2-axis are increased by a factor 1 + Λ2 . The perpen-
dicular line elements initially aligned with 1-and 2-axes are inclined at an angle θ12 in the current configuration
(Fig. 5-4), where
C12 Λ
cos θ12 = √ =√ (5.46)
C11 C22 1 + Λ2
i.e.,
1
tan θ12 =
Λ
The invariants of C are given by
IC = IIC = 3 + Λ2 , IIIC = 1 (5.47)
A deformation of this type is called a simple shear in the x1 x2 -plane.
2005
c by Ajit Mal
5.6. HOMOGENEOUS DEFORMATIONS 177
x2
12
x1
o
Fig. 5-4. Simple shear in the 1- 2- plane.
Example 5.6-1
Consider the case [akK ] = diag[a1 , a2 , a3 ]. Show that a sphere of radius 1 in the reference configuration is trans-
formed into an ellipsoid (known as strain ellipsoid) in the current configuration. Calculate the lengths of the semiaxes
of the ellipsoid and the orientations of its axes.
Solution. Consider the sphere of radius 1 in the reference configuration whose equation is
x21 x2 x2
2 + 22 + 23 = 1
a1 a2 a3
which represents an ellipsoid with semiaxes (a1 , a2 , a3 ). The axes of the ellipsoid are along the coordinate axes in the
reference configuration.
Example 5.6-2
Consider the simple shear deformation discussed in case (2). Calculate the matrices E and e and their principal
invariants.
2005
c by Ajit Mal
178 CHAPTER 5. KINEMATICS OF DEFORMATION
Let UI (XK , t) denote the components of the particle displacement vector x(X, t) − X in the Lagrangian description.
Then UI,J are the components of a second-order tensor H, not necessarily symmetric, called the displacement gradient
tensor. Clearly, H is related to F, C, and E through the equations
F = I+H (5.48a)
C = I + 2E = (I + HT )(I + H) = I + H + HT + HT H (5.48b)
1
ǫ= (H + HT ) (5.50a)
2
or
1
ǫIJ = (UI,J + UJ,I ) (5.50b)
2
If the small positive number ǫ denotes a measure of the ”size” of the displacement gradients UI,J , then ǫI,J is also
of the same size ǫ. This is often expressed by saying that UI,J and ǫI,J are small quantities of the first order and that
the product terms UK,I UK,J are small quantities of the second order. In writing the approximate strain-displacement
equation (5.53), only small quantities of the first order have been retained. Thus, in a consistent development of the
theory of linear elasticity, all small quantities of second and higher order should be neglected.
Assuming ǫ11 , ǫ22 , ǫ33 to be small and retaining only the first-order terms in the binomial expansions of the square
roots in the right-hand sides of (5.32), we obtain approximate expressions for the extensions ǫ11 , ǫ22 , ǫ33 as
Thus, for infinitesimal strains the diagonal elements of ǫ are approximately equal to the extensions (or change in length
per unit length) of line elements initially oriented along the coordinate axes.
If Γ12 denotes the change in the angle between perpendicular line elements initially oriented along the 1- and
2-axes, then (5.35) gives, after neglecting second- and higher-order terms,
This shows that Γ12 is itself a small quantity of the first order and, if it is expressed in radians,
Thus, the strain component ǫ12 is approximately equal to half the change in the angle between two perpendicular
line elements initially oriented along the 1- and 2-axes. The other off-diagonal elements of ǫ can be given a similar
geometrical interpretation. Equations (5.53), (5.54), and (5.55) agree with the results obtained in Chapter 1 from
simple geometric considerations.
To the first order of approximation, (5.38) and (5.39) give
ds3
≈ 1 + ǫ11 + ǫ22 (5.53)
dS3
2005
c by Ajit Mal
5.7. THE STRAIN-DISPLACEMENT EQUATIONS AND THE INFINITESTIMAL STRAIN TENSOR 179
and
dv
J= ≈ 1 + ǫ11 + ǫ22 + ǫ33 = 1 + tr (ǫ) (5.54)
dV
Equation (5.57) expresses the fact that the trace of the linear strain tensor ǫ is approximately equal to the change in
volume per unit volume of an element in the reference configuration. It can be seen from the preceding discussions
that the components of the infinitesimal strain tensor are simply related to the deformation of a unit cube element in
the reference configuration. The diagonal elements of ǫ describe the change in the length of its edges, the off-diagonal
elements describe half the change in the angle between its adjacent edges, and the trace of ǫ gives the change in vol-
ume. This is in contrast to the exact strains C and E, where the relationships are much more complex.
The transformation of an infinitesimal line element dX can also be described in a simple manner for small defor-
mations. If the displacement at X + dX is denoted by U + dU, then the relative displacement between particles located
at X and X + dX may be expressed in the form
where
1
ωIJ = (UI,J − UJ,I ) (5.56)
2
Clearly, ωIJ are the components of a skew symmetric second-order tensor. The first term in the right-hand side of
(5.58) represents the transformation of dX (stretch and rotation) due to the (small) deformation of the material in the
neighborhood of X. The second term represents an infinitesimal rigid rotation θ of the line element dX, where (see
Sec. 2.4)
1
θI = − ǫIJK ωJK (5.57)
2
In the Eulerian description, the displacement vector, expressed in the spatial coordinates, is denoted by u(x, t);
then the exact strain tensors e and c are related to the displacement gradients through the equations
1 1
eij = (δij − cij ) = (ui,j + uj,i − uk,i uk,j ) (5.58)
2 2
For small deformations, the displacement gradients ui,j are small quantities of the first order and the infinitesimal
strain tensor ǫij is given by
1
ǫij = (ui,j + uj,i ) (5.59)
2
The infinitesimal rotation tensor ω is defined by
1
ωij = (ui,j − uj,i ) (5.60)
2
Simple geometrical interpretation of the components of ǫ and ω can be given in terms of the deformation of line,
surface, and volume elements located in the spatial configuration Rt . However, when small quantities of the second
and higher orders are neglected, the distinction between the Lagrangian and Eulerian descriptions of the deformation
properties simply disappears. To see this, we note that in accordance with (5.3),
Thus,
But
u K = x K − XK
2005
c by Ajit Mal
180 CHAPTER 5. KINEMATICS OF DEFORMATION
Hence
∂XK ∂uK
= δKj − (5.61)
∂xj ∂xj
so that
∂ui ∂Ui ∂Ui ∂uK
= − (5.62a)
∂xj ∂Xj ∂XK ∂xj
or
h = H − Hh, h = (I + H)−1 H, H = h(I − h)−1 (5.62b)
For small deformations the product term in the right-hand side of equation (5.65) is of second order and can be
neglected; thus,
∂ui ∂Ui
≈ (5.63)
∂xj ∂Xj
Equation (5.66) implies that, to first order of approximation, the displacement gradients calculated from the Lagrangian
and Eulerian descriptions of the displacement vector are numerically equal. Thus, the infinitesimal strain and rotation
components are approximately the same in the two descriptions of the displacement vector as long as small quantities
of the second and higher orders are ignored.
It can be deduced from (5.21) and (5.64) that, for small deformations,
dsi ≈ (1 + ǫkk )dSi − uj,i dSj = (1 + ǫkk )dSi + (ωij − ǫij )dSj (5.64)
Equation (5.67) is a generalization of (5.56).
The properties of the infinitesimal strain tensor can be summarized as follows:
(i) For small deformations in which the components of the deformation gradient tensor H have values small compared
to unity, the infinitesimal strain tensor ǫ may be defined by the equation
1
ǫij = (ui,j + uj,i ) (5.65)
2
where the displacement vector u can be expressed either in terms of the material coordinates X or the spatial
coordinates x and time t.
(ii) The deformational properties of a unit cube with its edges aligned with the coordinate axes can be simply described
by the components of ǫ. The cube may be located at X in the reference state or at x in the deformed state, but
the relationships remain the same.
(iii) The diagonal elements of ǫ give the change in length of the edges of the unit cube, and these are called the normal
strain components.
(iv) The off-diagonal elements of ǫ give one-half of the change in angle between the adjacent perpendicular edges of
the unit cube and are called the shearing strain components.
(v) The trace of ǫ gives the change in volume of the unit cube, and the changes in the areas of its faces are given by
(5.56) and (5.67).
(vi) The transformation of a line element dX at X or dx at x is described by the equations
dui = ǫij dXj + ωij dXj ≈ ǫij dxj + ωij dxj (5.66)
where ω is the (infinitesimal) rotation tensor with components
The first term in (5.69) describes the transformation caused by the deformation of the material in the neighbor-
hood of X or x, and the second term describes the rigid rotation of the line element.
(vii) The principal values ǫ1 , ǫ2 , ǫ3 and the principal directions of ǫ can be obtained by calculating the eigenvalues of
ǫ. A cube oriented along the principal directions is deformed into a rectangular parallelepiped or another cube
of different size.
2005
c by Ajit Mal
5.7. THE STRAIN-DISPLACEMENT EQUATIONS AND THE INFINITESTIMAL STRAIN TENSOR 181
PROBLEMS
5.1. The material coordinates XK and the spatial coordinates xk of a point are related by the equations
x 1 = X1 , x2 = X2 + AX3 , x3 = X3 + AX2
where A is a constant. Determine the displacement components in both the material and the spatial forms. Show
that the material particles that lie on the circle
1
X1 = 0, X22 + X32 =
1 − A2
before deformation, lie on the ellipse
after deformation.
5.2. Find the value of the Jacobian J and the material and spatial descriptions of the velocity field for the motion given
by
x1 = X1 + atX22 , x2 = (1 + bt)X2 , x 3 = X3 , t ≥ 0
where a and b(≥ 0) are constants. Calculate ∂v/∂t, (v·∇)v and Dv/Dt. Verify the relation
Dv ∂v
= + (v·∇)v
Dt ∂t
x1 = X1 et + X3 (et − 1)
x2 = X2 + X3 (et − e−t )
x3 = X3
Show that J does not vanish for this motion and obtain the material and spatial descriptions of the velocity field.
x 1 = X1 , x2 = X2 − cX3 , x3 = X3 + cX2
where c(> 0) is a constant. Interpret the deformation as a sequence of stretches and a rotation.
Determine U, V, R.
Give geometric interpretations of these results.
x1 = 2X1 + X2 , x2 = X1 + 2X2 , x 3 = X3
2005
c by Ajit Mal
182 CHAPTER 5. KINEMATICS OF DEFORMATION
(a) Calculate changes in the length of sides, angles, and volume of the element dX1 dX2 dX3 located at (1, 1, 0)
in the undeformed state. Describe the nature of this deformation.
′
(b) Calculate B in cylindrical coordinates r, φ, and z.
5.8. A solid undergoes the deformation
C n = F T B n−1 F, B n = F C n−1 F T
(b) Show that the eigenvalues of B and C are equal, but their eigenvectors are not necessarily the same. (Use
Cayley Hamilton theorem).
5.10. Determine the principal axes of the strain tensors B and C for the plane shear
x1 = X1 + γX2 , x 2 = X2 , x 3 = X3
where γ is a constant.
5.11. Consider the deformation of a cylindrical region
(a) Show that the deformation is an ”extension” along the X3 -axis of α per unit length, a torsion about the
X3 -axis with twist angle ψ per unit length, and an ”inflation” of β per unit length in the radial direction.
(b) Calculate F, B, and C.
′
(c) Calculte B in cylindrical coordinates r, φ, and z.
5.12. A solid is subjected to the deformation field
x 1 = X1 − X2 , x2 = 2X1 + X2 , x 3 = X3 .
(a) Consider a unit cube in the reference configuration with its faces parallel to the coordinate axes. Calculate
the areas of the faces and the direction cosines of their normals in the spatial configuration.
√
(b) Consider a surface element ds(1, 1, 1)/ 3 in the spatial configuration. Determine the ratio dS/ds and the
direction cosines of the normal to dS in the reference configuration.
5.13 A circular cylinder is subjected to the deformation
x1 = X1 c − X2 s, x2 = sX1 + X2 c, x 3 = X3
where c = cos(γX3 ), s = sin(γX3 ). Consider the surface element dSN shown in the reference configuration.
After deformation the element becomes dsn. Use formula {n}ds = J[GT ]{N }dS to show that ds = dS and
n = cos(γX3 + θ)e1 + sin(γX3 + θ)e2 . Provide a geometrical interpretation to these results.
2005
c by Ajit Mal
5.7. THE STRAIN-DISPLACEMENT EQUATIONS AND THE INFINITESTIMAL STRAIN TENSOR 183
X3
dS
X2
X2
5.14 Write down the linear strain-displacement relations in (a) cylindrical coordinates with axial symmetry and (b)
spherical coordinates with central symmetry.
BIBLIOGRAPHY
1. Atkin, R. J. and Fox, N., An Introduction to the Theory of Elasticity, Longman, London, 1980.
2. Eringen, A. C., Nonlinear Theory of Continuous Media, McGraw-Hill, New York, 1962.
4. Fung, Y. C., Foundations of Solid Mechanics, Prentice Hall, Englewood Cliffs, N.J., 1965.
5. Green, A. E. and Adkins, J. E., Large Elastic Deformations and Nonlinear Continuum Mechanics, Oxford
University Press, London, 1960.
6. Malvern, L. E., Introduction to the Mechanics of a Continuous Medium, Prentice Hall, New York, 1969.
7. Murnaghan, F. D., Finite Deformation of an Elastic Solid, Dover, New York, 1951.
2005
c by Ajit Mal
184 CHAPTER 5. KINEMATICS OF DEFORMATION
2005
c by Ajit Mal
Chapter 6
In this chapter, we consider the physical principles that govern the motion of a continuum and develop their mathe-
matical statements in various forms. The most important physical principles are the conservation of mass, balance of
linear momentum, and balance of angular momentum of the aggregate of material particles that compose the body. In
addition, the motion of a continuum must be subject to the first law of thermodynamics, which plays an important role
in the formulation of problems, especially in cases where there is significant interaction between mechanical and non-
mechanical (say, thermal) processes. The balance laws of linear and angular momenta can be deduced from Newton’s
laws of motion applied to a system of particles supplemented by other assumptions regarding the nature of interacting
forces between them. The physical principles involved in the derivation of the laws discussed here should be accepted
as fundamental to continuum mechanics as results of our experience with nature. The justification of the principles
lies entirely on the usefulness of the laws in describing the true behavior of solids.
We consider a given set of particles that occupy a volume V0 (called the control volume) enclosed by the surface
S0 in the reference configuration R0 . The same set of particles occupy the volume Vt enclosed by the surface St in the
spatial configuration Rt at time t. The vector basis {ei } is assumed to be inertial, i.e., Newton’s second law of motion
holds if the velocity and acceleration of a moving particle are measured in this vector basis. A particle P labeled by
its position vector X in V0 occupies the position x in Vt and has instantaneous velocity v and acceleration a relative to
the vector basis {ei }. An element of volume dV located at X is transformed into the element dv at x. The density of
the material at X is denoted by ρ0 and at x by ρ. An element of surface dS at Y on S0 is transformed into the element
ds at y on St (Fig. 6-1) and n is the unit normal to ds at y.
Since there is no mechanism for the creation or destruction of mass within the continuum and since the same material
particles are contained within S0 and St , the principle of conservation of mass can be stated in the form
Z Z
ρ0 dV (X) = ρ dv(x)
V0 Vt
The right hand side of this equation may be expressed as an integral over the initial volume by means of the transfor-
mation (5.19), namely,
dv(x) = JdV (X)
Thus,
Z Z
ρ dv(x) = ρJ dV (X)
Vt V0
185
186 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
Fig. 6-1. The control volume in the reference and current states.
where the integrand ρJ is treated as a function of the material coordinate X. The law of conservation of mass now
becomes
Z
(ρ0 − ρJ) dV (X) = 0
V0
Assuming that the integrand is continuous in V0 and noting that V0 is an arbitrary volume in R0 , we conclude that
ρ0 − ρJ = 0, X ∈ R0 (6.1a)
or, using (5.19) once again,
ρ0 dV (X) = ρdv(x) (6.1b)
Differentiation of (6.1a) with respect to t with X fixed, gives
D
(ρJ) = 0
Dt
2005
c by Ajit Mal
6.1. BALANCE LAWS 187
or
Dρ
J + ρJvk,k = 0
Dt
where we have used the result of Example 5.2-2, (p 168):
DJ
= J∇·v = Jvk,k
Dt
Therefore,
Dρ
+ ρvk,k = 0 (6.1c)
Dt
Recalling (5.6),
Dρ ∂ρ ∂ρ
= + vk
Dt ∂t ∂xk
we obtain
∂ρ ∂
+ (ρvk ) = 0 (6.2)
∂t ∂xk
Equations (6.1) and (6.2) are alternative forms of the local balance of mass of a continuum and are called continuity
equations.
Before we discuss the balance laws of mass and momenta, we prove a result involving the time derivative of a volume
integral that is quite useful in their mathematical manipulation. Let f (x, t) denote the spatial description of a property
of the deforming continuum and consider the derivative
d
Z
ρf (x, t) dv(x)
dt Vt
Finally, transforming the integral in the right hand side of the preceding equation into the spatial domain, we obtain
d
Z Z
ρf (x, t) dv(x) = ρf˙(x, t) dv(x) (6.3)
dt Vt Vt
and Z
ρ(x × v) dv(x) (6.5b)
Vt
2005
c by Ajit Mal
188 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
respectively, where the angular momentum has been calculated with respect to the ”fixed” origin O. The rate of
increase of the total linear momentum is
d
Z
ρv dv(x)
dt Vt
By (6.3), this is equal to
Z Z
ρv̇ dv(x) = ρa(x, t) dv(x)
Vt Vt
where a(x, t) is the acceleration at x. Using a similar procedure, it can be shown that the rate of increase of total
angular momentum is
Z
ρx × a dv(x)
Vt
Thus, the balance laws of linear and angular momenta may be expressed by the equations
Z
ρa dv(x) = F (6.6a)
Vt
Z
ρx × a dv(x) = M (6.6b)
Vt
where F is the sum of all the forces acting on the particles within Vt and M is the sum of their moments about O.
Two types of forces are usually encountered in continuum mechanics, namely, body forces and surface forces.
Body forces are identical with the forces that are considered in the study of elementary mechanics (e.g., gravitational
forces), and they act throughout the interior of the continuum. For the present discussions we assume that these forces
are continuously distributed so that they can be specified by means of a volume density or mass density. The cases of
discontinuous or concentrated forces require special treatment. If b(x) denotes the body force per unit mass at x, the
resultant force on the particles within Vt is given by the integral
Z
ρb(x) dv(x) (6.7)
Vt
The boundary St of Vt is acted upon by a surface force that is caused by the interaction of the material exterior to
Vt with that within Vt . As in Sec. 1.2, we assume that the reaction forces on the surface element δs (Fig. 6-1) can be
represented by a single force δp and that the limit
δp
lim = t(y, n) (6.9)
δs→0 δs
exists. We further assume that, if δM denotes the moment of δp about any point within δs, then
δM
lim =0 (6.10)
δs→0 δs
Equations (6.9) and (6.10) are based on the stress principle of Cauchy and Euler. The vector t(y, n) is called the stress
vector at y. The resultant of surface forces acting on the particles within Vt is given by
Z
t(y, n) ds(y) (6.11)
St
2005
c by Ajit Mal
6.2. PROPERTIES OF THE STRESS VECTOR 189
Using (6.7), (6.8), (6.11), and (6.12), the balance laws of linear and angular momenta (6.6) may be expressed as
Z Z Z
ρa dv(x) = ρb(x) dv(x) + t(y, n) ds(y) (6.13)
Vt Vt St
Z Z
ρ(x × a) dv(x) = intVt ρx × b(x) dv(x) + y × t(y, n) ds(y) (6.14)
Vt St
It should be noted that b(x) defines a vector field over the volume of the body and t(y, n) defines a vector field
over surfaces drawn within the body, at a given instant of time. The time dependence of b(x) and t(y, n) has not
been made explicit in these notations, and their dependence on x, y, and n will also be omitted when this can be done
without confusion. If St happens to coincide with the boundary of Rt , then t(y, n) represents the applied force per
unit area at a boundary point y of the body (e.g., hydrostatic pressure on the boundary of a submerged body or the
contact forces between two bodies pressed together). Then t(y, n) is called the traction. If the boundary forces are
concentrated at a point, then t(y, n) is undefined, and the analysis of the motion and deformation of the solid requires
special consideration.
Equations (6.13) (6.16) give the global forms of the balance laws for the momenta, and these can be reduced to
local forms. This requires the introduction of the stress tensor, which is considered next.
In this section, we discuss two properties of the stress vector that lead to the introduction of an associated second order
tensor called the Cauchy’s stress tensor.
Let S denote an arbitrary smooth surface drawn in the spatial configuration of a solid and let m denote the unit
normal at a point y on the convex side of S. Then the stress vector on the convex side of S is t(y, m). The unit normal
to S at y on its concave side is −m; therefore, the stress vector on this side of S is t(y, −m) (Fig. 6-2). We assume
that
t(y, m) = −t(y, −m) (6.17)
i.e., the stress vectors at adjacent points on the opposite sides of a surface are equal and opposite (see Atkin and Fox
1980).
The second property of the stress vector is expressed by the equation
i.e., the stress vector on an arbitrarily oriented surface element at a point y can be expressed as a linear combination
of the stress vectors on three mutually perpendicular surface elements at that point. This is a generalization of the
two-dimensional result (1.14).
In order to prove (6.18), we consider a volume element in the form of a tetrahedron P ABC with one of its vertices
at P (y) (Fig. 6-3). Let the face ABC be of area δs and let the other three faces be mutually perpendicular and of areas
δsi (i = 1, 2, 3). Further, let the unit outward drawn normals to δs and δsi be n and −ei , respectively.
The volume of the tetrahedron is 13 h δs, where h is the distance of the face δs from the vertex P . Due to the
assumed continuity of the stress vector, the resultant force acting on the face δs of the tetrahedron is
[t(y, n) + ǫ]δs
2005
c by Ajit Mal
190 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
t(y, m)
m
- m
t(y, m)
Fig. 6-2. Sign convention for the stress vector.
C
t
e3 n
h
e2
e1
P B
A
Fig. 6-3. The tetrahedron element.
where |ǫ| → 0 as h → 0. Therefore, the principle of the balance of linear momentum, when applied to the material
within the tetrahedron, yields
2005
c by Ajit Mal
6.3. CAUCHY’S STRESS TENSOR 191
t(y3 , e3)
e3
σ33
t(y2 ,e 2)
σ 32
σ 31 y e2
t(y1,e1) 3 σ e1
23
σ13 σ 22
σ12 y2
y1 σ 21
σ11
Fig. 6-4. Stress components.
where |ǫ′ | → 0 as h → 0.
Since δsi is the projection of δs on the xi = 0 plane, δsi = ni δs. Taking the limit as the tetrahedron shrinks to the
point P (y), keeping n fixed, so that h → 0, we find
t(y, n) + t(y, −e1 )n1 + t(y, −e2 )n2 + t(y, −e3 )n3 = 0
We consider distributed surface forces acting on an infinitesimal rectangular parallelepiped element (Fig. 6-4) with its
edges parallel to the coordinate axes in the spatial configuration Rt . The stress vector at a point y1 on the face δs1
of the parallelepiped having unit outward drawn normal e1 is t(y1 , e1 ). Let σ11 , σ12 , σ13 denote the components of
t(y1 , e1 ) in the vector basis {ei }. Then
The component σ11 is called the normal stress on δs1 and the components σ12 , σ13 are called the shear stresses on
δs1 . Similarly, the stress vectors t(y2 , e2 ), t(y3 , e3 ) at points y2 , y3 on the faces δs2 , δs3 of the parallelepiped having
unit outward drawn normals e2 , e3 may be expressed as
where σ22 and σ33 are the normal stresses and σ21 , σ23 and σ31 , σ32 are the shear stresses on the surface elements
δs2 and δs3 , respectively (Fig. 6-4). In conformity with (6.17), the normal and shear stresses on opposite faces of the
parallelepiped must be drawn in opposite directions, as shown in Fig. 1-4.
In the limit, as the parallelepiped element shrinks to a point y, the preceding equations yield
2005
c by Ajit Mal
192 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
By the quotient rule (Sec. 2.3), the nine elements of the array [σij ] are the components of a second-order tensor σ,
known as the Cauchy stress tensor. The component σij represents the force per unit area in the j− direction acting on
an element of surface (in the spatial configuration) whose normal is in the i-direction. If σ is known at a given point,
then the components of the stress vector at that point on a surface element of arbitrary orientation can be obtained
from (6.21). Furthermore, the forces on other rectangular parallelepiped elements with arbitrary orientation can be
calculated from the tensor transformation law (2.43) or (2.45). Finally, if σ is known for all points in Rt , then the
surface forces on all surface elements in Rt can be calculated. Note that the diagonal elements of σ are the normal
stresses and its off diagonal elements are the shear stresses.
We now return to the balance laws of linear and angular momenta given by (6.15) and (6.16). Using (6.21), the surface
integrals in the right hand sides of these laws may be rewritten as
Z
σji nj ds(y)
St
and
Z
ǫijk yj σmk nm ds(y)
St
respectively. Using the divergence theorem (2.101), these may be transformed into the volume integrals
Z
σji,j dv(x)
Vt
and
Z Z
ǫijk (xj σmk ),m dv(x) = ǫijk (σmk δjm + σmk,m xj ) dv(x)
Vt Vt
Z
= ǫijk (σjk + xj σmk,m ) dv(x)
Vt
and these must hold for all regions Vt selected arbitrarily within the spatial configuration Rt . Therefore, the integrands
must vanish at each point of Rt . This leads to the local laws of the balance of linear and angular momenta,
and
ǫijk σij = 0
Recalling the properties of the alternating symbol ǫijk (Sec. 2.2), this gives
2005
c by Ajit Mal
6.4. CAUCHY’S EQUATION OF MOTION 193
Thus, balance of linear momentum leads to the equation of motion (6.24) and the balance of angular momentum
to (6.25), implying that the stress tensor is symmetric. Using the symmetry property, (6.21) and (6.24) may be written
in the more familiar forms
ti (y, n) = σij nj (6.26)
and
σij,j + ρbi = ρai , x ∈ Rt , t > 0 (6.27a)
Equation (6.27a) is known as Cauchy’s equation of motion. In the static case, a = 0, and (6.27a) reduces to the stress
equations of equilibrium (1.10)
σij,j + ρbi = 0 (6.27b)
It should be emphasized that (6.27) is valid for arbitrarily large deformations of the solid as no assumptions regarding
the size of the deformations have been made in their derivation. However, there are two serious difficulties in attempt-
ing to find solutions of this system of partial differential equations. First, there are only three scalar equations for the
determination of nine unknowns, namely, three components of acceleration and six components of stress. This diffi-
culty will be removed by introducing the constitutive equations, which give certain functional relationships between
the components of the stress and strain tensors. A second difficulty arises from the fact that both (6.26) and (6.27)
are derived by considering the spatial configuration Rt , and the dependent variables ρ, a, σ are all functions of the
spatial coordinates x (and time t). In order to solve the differential equations, it is necessary to provide information
on the initial state of the body as well as the conditions on the boundaries of Rt . In a severely deformed body these
boundaries are, in general, unknown at the outset and must be determined through the solution of the problem.
On the other hand, the boundaries of the reference configuration are known, and so the external loads on it can be
prescribed if the stress tensor is measured relative to an undeformed surface element in the reference state and if (6.27)
is written in terms of the material coordinates. In order to accomplish this, we return to the balance law (6.15) and,
using (6.21), rewrite it in the form
Z Z Z
ρai dv(x) = ρbi (x) dv(x) + σji nj ds(y)
Vt Vt St
Using (6.1b)
ρdv(x) = ρ0 dV (X)
the volume integrals over Vt can be transformed into those over V0 in the reference configuration, giving
Z Z Z
ρ0 ai dV (X) = ρ0 bi dV (X) + ti ds(y)
V0 V0 St
ti ds = ti dS = σ Ii NI dS
Similarly, using the surface transformation property (5.21b), nj ds = JGIj NI dS the surface integral over St can be
converted into an integral over S0 , namely,
Z
σji JXI,j NI dS(Y)
S0
Application of the divergence theorem further transforms it into the volume integral
Z
(σji JXI,j ),I dV (X)
V0
2005
c by Ajit Mal
194 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
where the integrand is now considered a function of the material coordinates X and time t. Since (6.28) must hold for
arbitrarily selected subregions of R0 , the local form of the balance law of linear momentum in material coordinates is
obtained as
σ Ii,I + ρ0 bi = ρ0 ai , X ∈ R0 , t>0 (6.29)
where
σ Ii = JXI,j σji = JGIj σji (6.30a)
or,
1
σ = JGσ, σ= Fσ (6.30b)
J
The second order tensor σ is called the first Piola stress tensor.
If δp denotes the force on the element δs, then its components can be related to the components of σ and σ through
δpj = σij ni δs = σij JGIi NI δS(Y) = σ Ij NI δS(Y), Y ∈ S0 (6.31a)
or
{δp} = [σ]T {n}δs = {t}ds = J[σ]T [G]T {N }δS = [σ]T {N }δS = {t̄}dS (6.31b)
where {t} and {t̄} are the stress vectors associated with Cauchy and Piola stresses:
{t} = [σ]{n}, {t̄} = [σ̄]T {N } (6.31c)
where we have used (6.30a). If S0 coincides with the boundary of R0 and St with the boundary of Rt , then the surface
tractions applied to the boundary of Rt can be prescribed by specifying σ Ij NI on S0 according to (6.31).
The stress tensor σ is nonsymmetric, an undesirable property in many solution techniques. Another stress tensor
σ̃, defined by
σ̃IJ = XJ,i σ Ii = GJi σ Ii , σ̃ = σGT (6.32a)
σ Ii = σ̃IJ xi,J = σ̃IJ FiJ , σ = σ̃FT (6.32b)
can be used to rewrite (6.29) in the form
(σ̃IJ xi,J ),I + ρ0 bi = ρ0 ai (6.32c)
The second-order tensor σ̃ is called the second Piola stress tensor and is related to the Cauchy stress tensor σ through
the equations
σ̃ = JGσGT , σ = J −1 Fσ̃FT (6.33)
Since σ is symmetric, (6.33) implies that σ̃ is also symmetric.
It should be emphasized that the Piola stresses are merely convenient mathematical representations of the physical
stresses in the reference configuration; they do not directly describe the traction on the surface of the solid. This is in
contrast to the Cauchy stress, which gives a simple and direct description of the traction, or the force per unit area on
the surface of the deformed solid.
We describe here a number of important properties of σ for future use. Since σ is a real, symmetric, second-order
tensor, it can be diagonalized by referring it to a new vector basis {e′i }, which is obtained by a proper rotation of {ei }.
The nonzero elements σ1 , σ2 , σ3 of the diagonalized tensor are called the principal stresses, and the directions e′i are
called the principal axes of stress. The stress vectors on the positive faces of a rectangular parallelepiped element with
edges parallel to the base vectors e′i are σ1 e′1 , σ2 e′2 and σ3 e′3 , i.e., the faces are acted upon by normal stresses only.
These mutually orthogonal planes with no shear stresses are known as the principal planes of stress.
The invariants of σ are given by
Iσ = σii = σ1 + σ2 + σ3
1
IIσ = (σii σjj − σij σij ) = σ2 σ3 + σ3 σ1 + σ1 σ2 (6.34)
2
IIIσ = |σ| = σ1 σ2 σ3
2005
c by Ajit Mal
6.6. LINEARIZATION OF THE BALANCE EQUATIONS 195
It can be shown that the second and third invariants of σ (d) are related to those of σ by the equations
is called uniaxial tension (if σ1 > 0) and uniaxial compression (if σ1 < 0) in the 1-direction.
Application of the balance laws of mass and momenta to the continuum leads to the system of differential equations
(6.2) and (6.27) for the density, velocity, and the Cauchy stress in the spatial domain Rt . However, as indicated in
Sec. 6.4, it is more convenient to formulate boundary-value problems in the reference domain R0 whose boundary B0
(Fig. 6-1) is known. The alternative forms of the differential equations are given by (6.1) and (6.29), in which all the
field variables are functions of the material coordinates X ∈ R0 and t > 0. In (6.29), the first Piola stress tensor sg is
related to the Cauchy stress tensor σ through
∂2
ai (X, t) = Ui (X, t)
∂t2
xi,K = δiK + Ui,K
J = det[δiK + Ui,K ]
From (5.21b), the surface elements ds(x) and dS(X) are related by
2005
c by Ajit Mal
196 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
or
NI dS(X) = J −1 xi,I ni ds(x) (6.41)
We now consider the simplifications that can be achieved in formulating the boundary-value problems when the
deformations are small. As in Sec. 5.7, we introduce a small number ǫ as a measure of the size of the displacement
gradients and deduce the following results;
If the boundary forces are prescribed on Bt , this is equivalent to prescribing the Cauchy stress vector in the form
dpj
tj (y, t) = σij ni (y) = , y ∈ Bt
ds(y)
{t} = σ T N ≈ [σ] {n}
N ≈ n, ds ≈ dS
where E IJ (X, t) are the components of the infinitesimal strain tensor introduced in Sec. 5.7.
Thus, it can be concluded that, in formulating boundary-value problems for small deformations, the distinction
between the reference and spatial configurations can be ignored, to the first order of approximation.
2005
c by Ajit Mal
6.7. BALANCE OF MECHANICAL ENERGY 197
We first assume that only mechanical processes are involved in the deformation of the continuum, i.e., there is no
significant interaction between mechanical and nonmechanical (e.g. thermal, electromagnetic, chemical, etc.) forms
of energy, and we consider the energy balance of the material contained within the spatial volume Vt bounded by the
surface St (Fig. 6-1). We recall that if f(x, t) denotes the spatial description of a property in Vt , then
d
Z Z
ρf (x, t) dv(x) = ρf˙(x, t) dv(x) (6.47)
dt Vt Vt
where
D
f˙(x, t) = f (x, t) = f,t + vj f,j (6.48)
Dt
In (6.48), v(x, t) is the velocity of the particle at x. Therefore, the rate of increase of the kinetic energy K of the
material particles within Vt is given by
d ρ
Z Z
K̇ = vi vi dv(x) = ρvi ai dv(x) (6.49)
dt Vt 2 Vt
We recall that for an isolated particle, Newton’s second law of motion is equivalent to PF = K̇, i.e., the mechanical
power of the external forces is exactly equal to the rate of increase of kinetic energy of the particle. Integration of this
equation along the path of the particle results in the principle of conservation of energy for conservative forces and to
the work-energy equation for general force fields. Equation (6.51) implies that this is no longer true for deformable
bodies and that the power of the external forces results in the rate of increase in the kinetic energy as well as in an
additional factor, which depends on the internal stress and deformation of the body. This latter factor is called the
stored, or internal energy, and the second term in the right-hand side of (6.51) is called the total stress power. Clearly,
a relationship between the stress power per unit volume, σji vi,j , and the internal energy exists, and this relationship is
of fundamental importance in developing constitutive equations, as will be seen in the next chapter. In order to derive
this relationship, we assume that W (x, t) is the internal energy at the spatial position x but measured per unit volume
of the undeformed material at X. Then, the rate of increase of internal energy of the material in Vt is W ˙ , where
˙ = d
Z
W W (x(X, t), t) dV (X)
dt V0
ρ
Z Z
= Ẇ (x(X, t), t) dV (X) = Ẇ (x, t) dv(x)
V0 ρ
Vt 0
2005
c by Ajit Mal
198 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
Let the energy balance equation (6.51) for the material within Vt be written as
˙
PF = K̇ + W (6.52)
ρ
Z Z
Ẇ (x, t) dv(x) = σji vi,j dv(x)
Vt ρ0 Vt
If there is significant interaction between mechanical and nonmechanical processes, the increase in the internal energy
of the body is caused by its deformation as well as by the energy supplied to it in other forms. We consider ther-
momechanical systems in which the only nonmechanical energy present is thermal. In order to include these thermal
effects in the equation of energy balance, it is necessary to apply the laws of thermodynamics to the continuum after
certain modifications. To this end, we first give a brief review of certain important concepts and principles of classical
thermodynamics.
In the language of thermodynamics, a particular collection of matter that we identify for study is called a system. If
a system does not exchange matter with its surroundings, it is called closed, and if it does not interact with its surround-
ings, it is called isolated. We shall be concerned with only closed and isolated systems. The totality of information that
completely characterizes a system for the purpose at hand is said to describe the state of the system, and the variables
whose values specify these characteristics are called state variables, each of which may describe a different property
of the system. If a state variable can be expressed as a function of a set of other variables, then this relationship is
known as an equation of state. If, for a given system, the state variables are independent of time, the system is said to
be in thermodynamic equilibrium, and if some or all of them vary with time, the system is said to undergo a process.
A system is homogeneous, or uniform, if the state variables are independent of the spatial coordinates; otherwise it is
nonhomogeneous, or nonuniform. A system is said to be thermally insulated if no changes can be produced in it by
any external agency, and any process taking place within it is called adiabatic.
According to the principle of global conservation of energy, the time rate of change of the kinetic plus internal
energies is equal to the sum of the power of the external forces plus all other energies that enter the system per unit
time. Thus, for a thermomechanical system, we have
d
Z
K̇ + ρU dv(x) = PF + H (6.54)
dt Vt
where K is the kinetic energy, U (x, t) is the internal energy per unit mass, PF is the power of the external forces, and
H is the heat energy absorbed by the system per unit time. For a system in thermodynamic equilibrium, K̇ = 0, and
(6.54) can be expressed in the form
δW = δPF + δQ (6.55)
where δW is the increase in the internal energy, δPF = PF δt is the work done on the system, and δQ = Hδt is the
heat energy absorbed by the system in time δt. Equation (6.55) is known as the first law of thermodynamics and is
valid for closed, isolated, and homogeneous systems in thermodynamic equilibrium.
The temperature Θ of a homogeneous system in thermal equilibrium is a single-valued function of state and is
a measure of the degree of hotness or coldness of the system. There is a temperature below which no system can
be cooled. Therefore, temperature scales can be constructed such that Θ is always positive. We assume that Θ is
measured in one such specific scale, called the absolute scale, and then Θ is the absolute temperature. The entropy
2005
c by Ajit Mal
6.8. THERMAL EFFECTS 199
S is another function of state that defines a fundamental property (similar to the mass or the electric charge) of a
homogeneous system in thermal equilibrium. We consider the system to be taken from one equilibrium state to a
neighboring equilibrium state by a reversible process. If Θ denotes its absolute temperature and δQ is the net heat
energy absorbed by the system, then its entropy increases by an amount δS, where
δQ
δS = (6.56a)
Θ
and δS is an exact differential. If the process is not necessarily reversible, then
δQ
δS ≥ (6.56b)
Θ
where the equality sign holds for reversible processes only. Equation (6.56a) can be used as the definition of entropy
and (6.56b) is a statement of the second law of thermodynamics. It should be noted that (6.56a, b) hold only for
homogeneous systems in thermal equilibrium.
In order to apply the laws of thermodynamics to continuum mechanics, we must extend their validity to systems
that are not necessarily homogeneous or are in thermodynamic equilibrium. The simplest approach is to assume that
the internal energy, temperature, and entropy are primitive quantities that can be defined for dynamic systems by
assigning them the values they would have as state functions if the system were held in equilibrium at that instant of
time. If the system is nonhomogeneous, we define the absolute temperature to be a function of the spatial position x
and assume that the heat energy, internal energy, and entropy can be defined per unit mass at each spatial position x.
Thus, in general, all the state functions are dependent upon x and t.
We now consider the energy balance of the material contained in the spatial volume Vt (Fig. 6-1) in the presence
of thermal effects. Let h(x, t) denote the rate of supply of heat energy per unit mass by sources within Vt . Since the
system is nonhomogeneous, it is possible to have transfer of heat energy between various material elements. This
will be represented by a heat flux rate vector q(x, t) such that q · nds(x) represents the rate of transfer of heat energy
across a surface element ds(x) at x with unit normal n from the negative side of n to its positive side. Then, the rate
of increase of heat energy within Vt is H, where
Z Z
H= ρh dv(x) − qi ni ds(y)
Vt St
Using (6.47), the global energy balance equation (6.54) may be written in the form
Z
PF + H = K̇ + ρU̇ dv(x) (6.58)
Vt
where PF , the total mechanical power of the body and surface forces, is given by (6.50). It can be seen from the
derivation of (6.51) that it is valid for all systems in which Cauchy’s equation of motion (6.27) holds and is, therefore,
valid even when thermal effects are present. Then, from (6.51) and (6.58), we have
Z Z
PF − K̇ = σji vi,j dv(x) = ρU̇ dv(x) − H
Vt Vt
Under the usual assumption regarding the continuity of the integrand, we obtain
Equation (6.60) describes how the stress power contributes to the increase in internal energy in the presence of
thermal effects and is known as the dissipation equation.
2005
c by Ajit Mal
200 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
In order to apply the second law of thermodynamics to the continuum, we assume that η(x, t) is the entropy per
unit mass at x and that the body is taken to a neighboring state at time t + δt. In the second state, the entropy per unit
mass is η + δη. Then the change in entropy of the material contained in Vt is
Z
δS = ρδη dv(x)
Vt
Thus, for an irreversible process the second law of thermodynamics (6.56b), as applied to the continuum, becomes
Z
ρh qi ni
Z Z
ρδη dv(x) ≥ dv(x) − ds(y) δt (6.61)
Vt Vt Θ St Θ
Equation (6.61) is the modified form of the it Clausius inequality of classical thermodynamics. Applying the
divergence theorem to the surface integral in the right-hand side of the preceding equation and assuming continuity of
the integrands within Rt , we obtain
ρh qi
ρδη − − δt ≥ 0, x ∈ Rt , t>0 (6.62)
Θ Θ ,i
Eliminating h by means of the energy equation (6.60) and writing δU for U δt, (6.62) becomes
qi Θ,i
ρ(Θδη − δU ) + σji vi,j − δt ≥ 0 (6.63)
Θ
The Helmholtz free energy W (x, t) of the material at x measured per unit volume of the undeformed material at X
is defined through the equation
W = ρ0 (U − ηΘ) (6.64)
Then the inequality (6.63) may be expressed in the form
qi Θ,i δW
σji vi,j − δt − ρ ηδΘ + ≥0 (6.65)
Θ ρ0
Equation (6.67) and the inequality (6.66) are extremely useful in developing the constitutive relations in the presence
of significant interaction between mechanical and thermal processes.
where b is the body force per unit mass and a is the cceleration. We may write the preceding equation as
∂ (σji ) ∂ (σji )
ek · ej =
∂xk ∂xj
2005
c by Ajit Mal
6.9. EQUATION OF MOTION IN CYLINDRICAL AND SPHERICAL COORDINATES 201
In cylindrical coordinates, we denote the components of the body force by br , bφ , bz ; the components of the accelera-
tion by ar , aφ , az , and the components of the stress dyadic by (Fig. 6-5)
σrr σrφ σrz
σφr σφφ σφz (6.71)
σzr σzφ σzz
We may write
σ = e r tr + e φ tφ + e z tz (6.72)
where er , eφ , ez are unit vectors and
tr = σrr er + σrφ eφ + σrz ez
tφ = σφr er + σφφ eφ + σφz ez (6.73)
tz = σzr er + σzφ eφ + σzz ez
Then, as in (1.164), it can be shown that
∂tr 1 1 ∂tφ ∂tz
∇·σ = + tr + + (6.74)
∂r r r ∂φ ∂z
Using relations (1.160) for the derivatives of the unit vectors er , eφ , ez , (6.72) (6.74) yield
∂σrr ∂σrφ ∂σrz 1
∇·σ = er + eφ + ez + (σrr er + σrφ eφ + σrz ez )
∂ ∂r ∂r r
1 ∂σφr ∂σφφ ∂σφz
+ er + σφr eφ + eφ − σφφ er + ez (6.75)
r ∂φ ∂φ ∂φ
∂σzr ∂σzφ ∂σzz
+ er + eφ + ez
∂z ∂z ∂z
Equations (6.69) and (6.75) yield a vector equation that is equivalent to the following three scalar equations of motion
in cylindrical coordinates:
∂σrr 1 ∂σφr ∂σzr 1
+ + + (σrr − σφφ ) + ρbr = ρar
∂r r ∂φ ∂z r
∂σrφ 1 ∂σφφ ∂σzφ 2
+ + + σrφ + ρbφ = ρaφ (6.76)
∂r r ∂φ ∂z r
∂σrz 1 ∂σφz ∂σzz 1
+ + + σrz + ρbz = ρaz
∂r r ∂φ ∂z r
For plane stress problems, the equations of motion reduce to
∂σrr 1 ∂σφr 1
+ + (σrr − σφφ ) + ρbr = ρar
∂r r ∂φ r
∂σrφ 1 ∂σφφ 2
+ + σrφ + ρbφ = ρaφ (6.77)
∂r r ∂φ r
2005
c by Ajit Mal
202 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
x3
zz
z
dz
zr
r
rz
r rr
z
z
o
d x2
r rd
x1 dr
Fig. 6-5. Stress components in cylindrical coordinates.
We list the equations of motion in spherical coordinates (Fig. 6-6) without going into the details of their derivation.
2005
c by Ajit Mal
6.9. EQUATION OF MOTION IN CYLINDRICAL AND SPHERICAL COORDINATES 203
x3
RR
R
R
R
R
d
x2
o
d
R Rd
x1 dR
PROBLEMS
6.1. Prove the following relationships between the components of the Cauchy stress tensor and the Piola stress tensors:
1 1
σ= eF T
Fσ = Fσ
J J
6.2. Show that for small deformations the distinction between the Cauchy stress tensor σ and the second Piola stress
tensor σ̃ can be ignored.
6.3. (a) Write down the global form of the balance of energy for a control volume of a solid in presence of body
forces, surface forces and a temperature field.
(b) Express the global equation in the reference configuration.
(c) obtain the local form of the equation in material coordinates involving the first Piola stress. Simplify this
equation assuming that the balance of linear momentum holds.
x 1 = X1 + X2 , x2 = −2X1 + X2 , x 3 = X3
√
A surface element, dS(e1 + e2 + e3 )/ 3, is located at (1,2,1) in the reference configuration is subjected to the
(first Piola) stress,
1 0 0
σ0 5 3 0
0 0 1
where σ0 is a constant. Determine the Cauchy stress and the associated stress vector at the corresponding point
in the spatial configuration.
2005
c by Ajit Mal
204 CHAPTER 6. BALANCE LAWS AND ANALYSIS OF STRESS
6.5. Write down the components of Cauchy’s equation, (6.27a), in cylindrical and spherical coordinates under cylin-
drical and spherical symmetries, respectively.
BIBLIOGRAPHY
1. Atkin, R. J. and Fox, N., An Introduction to the Theory of Elasticity, Longman, London, 1980.
2. Eringen, A. C., Nonlinear Theory of Continuous Media, McGraw-Hill, New York, 1962.
4. Fung, Y. C., Foundations of Solid Mechanics, Prentice Hall, Englewood Cliffs, N.J., 1965.
5. Green, A. E. and Adkins, J. E., Large Elastic Deformations and Nonlinear Continuum Mechanics, Oxford
University Press, London, 1960.
6. Malvern, L. E., Introduction to the Mechanics of a Continuous Medium, Prentice Hall, Englewood Cliffs, N.J.,
1969.
7. Murnaghan, F. D., Finite Deformation of an Elastic Solid, Dover, New York, 1951.
8. Spencer, A. J. M., Continuum Mechanics, Longman, London, 1980.
2005
c by Ajit Mal
Chapter 7
Constitutive Equations
The equations of continuity and motion derived in the last chapter provide only four scalar equations in the ten un-
knowns ρ, ui , σij . The energy balance laws provide additional equations but only at the expense of introducing other
unknowns η, q, Θ. It may be recalled that the balance equations are derived on the basis of laws that hold for all
materials regardless of their internal constitutions. It is, however, common knowledge that two bodies composed of
different materials (e.g., steel and aluminum) undergo different deformations even though all other factors influencing
them are identical. Hence, it is not surprising that the balance equations are not adequate for the determination of all
the unknowns that appear in them. As indicated in Sec. 6.4, the balance laws must be supplemented by another set
of equations that characterize the physical properties of the specific material under consideration. Such equations are
called constitutive equations.
The general theory of constitutive equations deals with the selection of the variables that enter these equations and
with the establishment of rules that must be followed in deriving the equations for very general classes of materials
subjected to a variety of interacting fields. The number of equations must be such that they, together with the balance
equations, comprise a system that can, in principle, be solved for all the unknowns. Another major objective of the
constitutive theory is to provide a framework for designing laboratory experiments that may yield a complete character-
ization of a given material. The systematic development of the constitutive theory can lead to considerable complexity
[see, for example, Truesdell and Noll (1966), Eringen (1967)], and is beyond the scope of this book. We consider only
a special class of materials and assume that all deformations are produced by either mechanical or thermal effects. The
constitutive theory for these cases is relatively simple to develop and the resulting constitutive equations are simpler
in their representations. The problems that can be handled under these restrictions are nevertheless of considerable
practical importance.
For this restricted class of problems, the dependent constitutive variables are the stress tensor σ, the heat flux
vector q, and the entropy density η. We assume that the constitutive equations are subject to the following general
rules:
Rule 1. The values of the dependent constitutive variables at a material point X at time t are completely determined
by either the instantaneous values or the time histories of the deformation and temperature in the neighborhood
of X.
Rule 2. The constitutive equations must be form invariant under arbitrary superposed rigid motion of the spatial con-
figuration Rt .
Rule 3. The constitutive equations must be form invariant under certain rigid motions of the reference configuration
R0 , depending on the symmetry properties of the material under consideration.
Rule 4. In the presence of thermal effects, the constitutive equations must be consistent with the thermodynamic
inequality (6.66).
205
206 CHAPTER 7. CONSTITUTIVE EQUATIONS
We now consider certain special classes of materials and derive the general forms of their constitutive equations on the
basis of these rules.
7.2 ELASTICITY
We first assume that there is no significant interaction between mechanical and thermal processes. Then the only
dependent constitutive variable is σ, and in accordance with Rule 1, it depends only on the properties of the local
deformation at a given point. It has been shown in Chapter 5 that the state of deformation in the neighborhood of a
point in the body is completely determined by the value of the deformation gradient tensor F at that point. An elastic
body is defined to be one in that the stress at X at time t is completely determined by the instantaneous value of F and
is not affected by the previous values (i.e., the history) of F. Thus, the constitutive equation for an elastic material may
be expressed in the form
σ = g(F) (7.1a)
or
σij = gij (F) (7.1b)
where the functions gij are single valued and are assumed to be differentiable with respect to their arguments. Clearly,
the symmetry of σ implies that
gij = gji (7.2)
The functions gij may depend explicitly on the material coordinates X. Then the constitutive equation varies
from point to point within the body, and the material is called inhomogeneous; otherwise it is homogeneous. The
definition of a homogeneous material should not be confused with that of a homogeneous deformation (Sec. 5.6). In
the discussions that follow, the explicit dependence of g on X can be omitted without loss of generality, and this will
be done throughout this chapter.
In the absence of a thermal field, Rule 4 of Sec. 7.1 plays no role in the discussions. However, (7.1) in its present
form does not necessarily satisfy the invariant properties described in Rules 2 and 3 unless the function g(F) possesses
certain special properties; these will be derived next.
Consider a second spatial configuration Rt of the body, which is obtained by a rigid rotation of the configuration
Rt at the same instant of time t (Fig. 7-1). A particle that was at X in the undeformed reference configuration R0
occupies the position x in Rt and x in Rt . Then x and x are related by the equation
where c(t) is an arbitrary time dependent vector and Q is a proper orthogonal tensor, i.e.,
Clearly, the relative positions of the particles within the continuum are identical in the two deformed configurations
Rt and t, although their absolute positions differ in accordance with (7.3). The deformation gradient tensor F for Rt
is given by
∂xi
F iK = xi,K = xj,K
∂xj
and, by (7.3),
∂xi
= Qij
∂xj
Therefore,
F iK = Qij xj,K = Qij FjK (7.5a)
or
F = QF (7.5b)
2005
c by Ajit Mal
7.2. ELASTICITY 207
x x
X
Rt
x3 Ro Rt
x2
o
x1
Since the stress depends on xk,K and not on xk as such, the translatory motion prescribed by the vector c(t) has no
effect on the stress. Furthermore, since the orientations of Rt and Rt differ by a rigid rotation only, σ in Rt is related
to σ in Rt by the rotation formula (2.45), i.e.,
σ = QσQT (7.6a)
or since Q is orthogonal,
σ = QT σQ (7.6b)
The form invariance of (7.1) under arbitrary superposed rigid-body motion of Rt (Rule 2) implies
σ = g(F) (7.7)
for arbitrary choice of Q. It follows from (7.5b), (7.6b), and (7.7) that
σ = QT g(QF)Q (7.8)
for every choice of proper orthogonal Q; i.e., (7.1) must be expressible in the form (7.8).
It was shown in Sec. 5.3 that F can be decomposed as F = RU, where R is proper orthogonal and U is positive-
definite and symmetric. Also, R and U are given in terms of F through the relations
U = UT = RT F = (FT F)1/2 (7.9a)
R = FU−1 (7.9b)
Since (7.8) holds for all proper orthogonal Q, it must also hold for Q = RT for every spatial position x ∈ Rt . Then
we must have
σ = Rg(RT F)RT = Rg(U)RT (7.10)
Recalling that the right Cauchy-Green tensor is given by C = FT F = U2 and using (7.9b), (7.10) may be expressed
in the form
σ = Ff(C)FT (7.11)
where
f(C) = C−1/2 g(C1/2 )C−1/2
and
√
C−1/2 = ( C)−1 = U−1
The definition of the square root of a matrix can be found in the appendix.
The form (7.11) of the constitutive equation is also sufficient for the form invariance property to hold. To show
this, we note that for an arbitrary rigid-body motion described by (7.3), F = QF [Eq. (7.5b)], so that
T
C = F F = FT QT QF = FT F = C
2005
c by Ajit Mal
208 CHAPTER 7. CONSTITUTIVE EQUATIONS
x
x3 Ro
x x
Ro Rt
x2
o
x1
Fig. 7-2. The rotated reference configuration R0 .
thus proving our assertion. Therefore, in order that the constitutive equation for an elastic body be form invariant, it is
necessary and sufficient that it takes the special form given by (7.11).
Further simplification in the form of the constitutive equation can be achieved through the use of certain symmetry
properties that are present in most materials in their undeformed states (Rule 3). An important special case is a material
for which there are no preferred directions at a material point X ∈ R0 with regard to its response in any stress-strain
experiment. In other words, the material is isotropic.
In order to examine the nature of restrictions imposed on the form of the constitutive equation by such symmetry
properties, we consider a second reference configuration R0 that is obtained by giving a rigid displacement to R0 .
Then, a material particle located at X ∈ R0 occupies the position X ∈ R0 , where
QQT = QT Q = I, |Q| = 1
We now consider a motion of the body in which the same particle occupies the position x in the spatial configuration
Rt (Fig. 7-2). If we use R0 as the reference configuration, then x is a function of X and t and the deformation gradient
tensor F is given by
∂xk
FkK = (X, t) = xk,K
∂XK
On the other hand, if R0 is used as the reference configuration, then x is a function of X and t, and the corresponding
deformation gradient tensor F is given by
∂xk
F kK = (X, t)
∂X K
2005
c by Ajit Mal
7.2. ELASTICITY 209
If the material is symmetric with respect to the motion described by Q, then (7.11) must be form invariant under
the transformation just described; i.e., we must have
T
Ff(C)FT = Ff(C)F = FQT f(QCQT )QFT
Since F is nonsingular and Q is orthogonal, it follows that the tensor function f must satisfy the relation
For an isotropic solid, (7.15) must hold for all proper orthogonal Q. Then f(C) is an isotropic function.
We assume that f(C) can be expanded in the power series
f(C) = a0 I + a1 C + a2 C2 + · · · + aN CN (7.16)
where I is the identity tensor and a0 , a1 , ..., aN are scalar constants. Since, for any positive integer n,
each term on the right-hand side of (7.16) is an isotropic tensor of the second order. Noting that
where B is the left Cauchy-Green tensor defined in (5.22), we have, from (7.11) and (7.16),
σ = Ff(C)FT = a0 B + a1 B2 + a2 B3 + · · · + aN BN +1 (7.19)
Furthermore, the invariants of the tensors B and C are equal (see Problem 5.9), i.e.,
According to the Cayley-Hamilton theorem (Sec. A.3), B satisfies its own eigenvalue equation, i.e.,
Thus,
Bn = IB Bn−1 − IIB Bn−2 + IIIB Bn−3 (7.21)
Therefore, (7.19) can be expressed in the form
σ = α0 I + α1 B + α2 B2 (7.22a)
where α0 , α1 , α2 are functions of the invariants IB , IIB , IIIB and, in general, X. Equation (7.22a) is the general
constitutive equation for an isotropic elastic solid. Using (7.21) it can be expressed in the alternative form
where β0 , β1 , β−1 are, in general, functions of X and the invariants IB , IIB , IIIB . The α’s and β’s are called response
functions.
2005
c by Ajit Mal
210 CHAPTER 7. CONSTITUTIVE EQUATIONS
An alternative procedure can be used to derive (7.22). In this procedure, the material symmetry argument is applied
to the postulated equation (7.1) directly, leading to the restriction
F = VR, V = FRT
Since V is positive-definite, it is uniquely determined by the left Cauchy-Green tensor B, so that we may write
σ = f(B)
The preceding equation is now subjected to the principle of invariance described in Rule 2 of Sec. 7.1, namely,
σ = f(B)
i.e.,
Qf(B)QT = f(B)
Qf(B)QT = f(QBQT )
for all proper orthogonal Q. Assuming a power series representation of f(B) in the form
σ = f(B) = b0 I + b1 B + b2 B2 + · · · + bN BN
and using the Cayley-Hamilton theorem, this can be reduced, as before, to an equation of the form (7.22). An alterna-
tive proof of (7.22) without assuming the power series expansion can be found in Gurtin (1982).
We assume that only mechanical processes are involved in the deformation of the continuum so that there is no
interaction between mechanical and nonmechanical forms of energy and the increase in the internal energy of the
body is caused by the deformation alone. Then there exists a function W (F), known as the strain energy function,
that is a single-valued function of F only (and not of the rate of deformation or its history). We recall from Sec. 6.7
that W (F) is the internal mechanical energy stored in the body, due to deformation, measured per unit volume of the
undeformed state. Such materials are often called hyperelastic.
We assume, on physical grounds, that W (F) is invariant with respect to arbitrary superposed rigid motions of the
spatial configuration Rt (see Rule 2 of Sec. 7.1). Then W (F) must satisfy the condition
2005
c by Ajit Mal
7.3. THE STRAIN ENERGY APPROACH: HYPERELASTICITY 211
for arbitrary proper orthogonal Q. Using the polar decomposition F = RU, with R and U of (7.9), (7.23a) yields
Since U = C1/2 , we conclude that W depends on F through the right Cauchy-Green tensor. In order to avoid the
introduction of another symbol, we denote this function of C by W (C) itself. Then
∂W
Ẇ = ĊIJ
∂CIJ
Now,
C = F T F, Ċ = Ḟ T F + F T Ḟ
Ḟ = [ẋi,J ] = [vi,J ] = [vi,j xj,J ] = DF
Ḟ T = F T DT
where,
D = [vi,j ] = the spin tensor
Hence,
Ċ = F T DT F + F T DF
ĊIJ = xi,I vi,j xj,J + vi,j xj,I xi,J (7.24a)
ρ ∂W
σij = (xi,I xj,J + xi,J xj,I ) (7.25)
ρ0 ∂CIJ
Since C is symmetric, (7.25) gives
ρ ∂W ∂W
σij = xi,I xj,J +
ρ0 ∂CIJ ∂CIJ
ρ ∂W
= 2 xi,I xj,J (7.26)
ρ0 ∂CIJ
2 ∂W
= √ FiI FjJ
IIIC ∂CIJ
2
σ = FPFT
J
∂W
PIJ =
∂CIJ
where we have used the result
ρ0 p
= J = IIIC
ρ
2005
c by Ajit Mal
212 CHAPTER 7. CONSTITUTIVE EQUATIONS
2
Comparing with (7.11), f (C) = J P.
it can be seen from (6.30), (6.33) and (7.26) that the first and second Piola stress tensors are related to W through the
equations
∂W
σ Ii = 2xi,J (7.27a)
∂CIJ
∂W ∂W
σ̃IJ = 2 = (7.27b)
∂CIJ ∂EIJ
The presence of material symmetry imposes additional restrictions on the forms of the constitutive equations. We
consider here only isotropic materials for which W must be invariant under arbitrary rotational motion of the reference
configuration R0 . Then we must have
W (C) = W (QCQT ) (7.28)
for an arbitrary orthogonal Q. Let us choose Q so that QCQT is diagonal. It follows from (7.28) that W (C) is a
function of the three eigenvalues C1 , C2 , C3 of C. Noting that C1 , C2 , and C3 are the three roots of the equation
λ3 − IC λ2 + IIC λ − IIIC = 0
C3 − IC C2 + IIC C − IIIC I = 0
I1 = tr(C) = IC
I2 = tr(C2 ) = IC2 − 2IIC
I3 = tr(C3 ) = IC3 − 3IC IIC + 3IIIC
2005
c by Ajit Mal
7.3. THE STRAIN ENERGY APPROACH: HYPERELASTICITY 213
∂W
= (W1 + IC W2 + IIC W3 )δIJ − (W2 + IC W3 )CIJ + W3 CIK CKJ (7.31)
∂CIJ
where
∂W ∂W ∂W
W1 = , W2 = , W3 = (7.32)
∂IC ∂IIC ∂IIIC
Thus, the constitutive equation (7.26) becomes
2
σij = √ [(W1 + IC W2 + IIC W3 )FiI FjI
IIIC
− (W2 + IC W3 )FiI CIJ FjJ + W3 FiI CIK CKJ FjJ ] (7.33a)
In tensor notation,
2
σ=√ [(W1 + IC W2 + IIC W3 )FFT − (W2 + IC W3 )FCFT + W3 FC2 FT ]
IIIC
(7.33b)
B = FFT , C = FT F
B2 = FCFT , C2 = FT BF
2
σ=√ [(W1 + IC W2 + IIC W3 )B − (W2 + IC W3 )B2 + W3 B3 ] (7.34)
IIIC
B3 = IB B2 − IIB B + IIIB I
Inserting the value of B3 in (7.34), and noting that the invariants of B and C are the same, we obtain
σ = α0 I + α1 B + α2 B2 (7.35)
where
p
α0 = 2 IIIB W3 (7.36a)
2
α1 = √ (W1 + IB W2 ) (7.36b)
IIIB
2
α2 = −√ W2 (7.36c)
IIIB
∂W ∂W ∂W
W1 = , W2 = , W3 = (7.37)
∂IB ∂IIB ∂IIIB
Using the Cayley-Hamilton theorem, equation (7.35) can be written in the alternative form
2005
c by Ajit Mal
214 CHAPTER 7. CONSTITUTIVE EQUATIONS
where
2 2 p
β0 = √ (IIB W2 + IIIB W3 ), β1 = √ W1 , β−1 = −2 IIIB W2
IIIB IIIB
(7.39)
It may be noted that the constitutive equations (7.35) and (7.38) for isotropic hyperelastic materials are more
restrictive than the corresponding equations (7.22a, b) for isotropic elastic materials in that the coefficients α0 , α1 , α2
and β0 , β1 , β−1 in the former case are not independent but, instead, depend on W through (7.36) and (7.39).
It can be shown that for a transversely isotropic solid with the symmetry axis along x1
2
σ= (IIB W2 + IIIB W3 ) I + IIIB W2 B −1 + W4 M + W5 N
J
2Mij = C1α (Fiα Fj1 + Fjα Fi1 ) , α = 2, 3, i, j = 1, 2, 3
Nij = Fi1 Fj1
For many solids (e.g., rubber or similar materials) that remain elastic at large deformations, the change in volume of
an element is small compared to its change in shape. Such solids are capable of only isochoric deformations and are
called incompressible. For an incompressible solid that suffers no deformation when subjected to only hydrostatic
stress,
p ρ0
IIIC = J = =1
ρ
Thus, the constitutive equation for an incompressible elastic solid may be expressed in the form
or
σ = −pI + f(F) (7.40b)
where p is an arbitrary scalar undetermined by the local deformation. Assuming the existence of a strain energy
function, it can be shown that the constitutive equation for an incompressible solid is of the form [see Eq. (7.26)]
∂W
σij = −pδij + 2xi,I xj,J (7.41)
∂CIJ
where the strain energy function W is now a function of only IC and IIC (since IIIC = 1). For an isotropic,
incompressible material, the constitutive relation (7.38) becomes
It can be seen from (7.26), (7.27), (7.35), (7.38), (7.41), and (7.42) that once the form of the strain energy function
is known, the constitutive equations can be obtained as an explicit system of equations relating the stress components to
the strain components. In the case of incompressible solids, the stress is determined to within an arbitrary hydrostatic
pressure, which must be determined from the equations of motion and boundary conditions. The constitutive equations
provide six equations that are needed in addition to the equations of continuity and motion to determine the ten
unknowns ρ, ui , σij
Various forms of the strain energy function for nonlinear elastic deformations have been proposed in the literature.
For incompressible materials, the simplest form of W proposed by Treloar (1974) is
W = c1 (IB − 3) (7.43a)
2005
c by Ajit Mal
7.4. THERMOELASTICITY 215
where c1 and c2 are constants, was proposed by Mooney (1940). It has been used extensively by Rivlin in his work on
rubber elasticity [see, for example, Rivlin 1949]. Materials satisfying the resulting stress-strain law
where c1 is again a constant, and no explicit analytical form of the function f was given except for the implication that
it is monotonic decreasing. Various explicit forms of f have been suggested by a number of other authors for example,
Truesdell and Noll 1966, Rivlin and Saunders 1957; these will not be discussed here.
For a homogeneous, isotropic, compressible, elastic solid, W may be expressed in a power series of the form
#
3 3 3
"
X X X
W = ai I i + aij Ii Ij + (aijk Ii Ij Ik + · · · ) (7.45)
i=1 j=1 k=1
where ai , aij ,... are constants. Since the reference state is the undeformed state, stress and strain must vanish simulta-
neously, which implies that a1 = 0 in the preceding representation.
7.4 THERMOELASTICITY
The constitutive equations derived in Secs. 7.2 and 7.3 are based on the assumption that thermal effects are negligible.
This would be the case when, for example, the temperature is constant throughout the body and no change in temper-
ature occurs during the deformation process; in other words, the thermal state is isothermal. On the other hand, if the
temperature field within the body is inhomogeneous, then there may be significant deformation or stress generated in
the material due to its nonuniform expansion. In this section, we show how these effects can be incorporated in the
constitutive equations for solids which behave elastically.
We first make the obvious observation that in the presence of significant thermal effects, the stress tensor σ must
depend on the deformation gradient tensor F as well as on the temperature Θ(X, t). If the functional dependence
were known, this would have given us six equations to supplement the three equations of motion and the continuity
condition. These ten equations are still not sufficient for the determination of all the unknowns, which now include
Θ(X, t) in addition to the original ten, namely, ρ, ui , and σij . If we include the equation of energy (6.67) as one of the
field equations, then, for a given heat source h, five additional unknowns, W, η, and qi are introduced. Thus, in order
to arrive at a sufficient number of equations, it is now necessary to consider W (the Helmholtz free energy), η, q, and
σ as the dependent constitutive variables. Since the deformation in the neighborhood of X is completely determined
by F, and the temperature field in the neighborhood of X is determined to a sufficient degree of accuracy by Θ(X, t)
and its gradient vector, g = ∇Θ(X, t), Rule 1 of Sec. 7.1 implies that for an elastic body, these constitutive variables
are single-valued functions of F, Θ, and g.
Before applying Rules 2 and 3 of Sec. 7.1, we examine what, if any, restrictions are imposed on the forms of the
constitutive functions by the thermodynamic inequality (6.66). Noting that
∂W ∂W ∂W
Ẇ (Θ, F, g) = Θ̇ + ḞiJ + ġi (7.46a)
∂Θ ∂FiJ ∂gi
ḞiJ = ẋi,J = vi,j FjJ (7.46b)
2005
c by Ajit Mal
216 CHAPTER 7. CONSTITUTIVE EQUATIONS
qi Θ,i ≤ 0 (7.49)
Equations (7.48) indicate that σ, η, and W are all independent of the temperature gradient vector g.
It is of interest to note that the use of (7.46) and (7.48) in the energy equation (6.67) reduces it to
We now impose the condition that W must be independent of arbitrary rigid rotations of the spatial configuration
Rt (Rule 2 of Sec. 7.1), i.e.,
W (F, Θ) = W (QF, Θ)
for an arbitrary choice of proper orthogonal Q. The preceding equation is similar to (7.23a) and, therefore, following
the arguments used before, we conclude that W depends upon F through the right Cauchy-Green tensor C. Using the
procedure outlined in Sec. 7.3, we obtain
2 ∂W
σij = √ FiI FjJ (7.51)
IIIC ∂CIJ
where W is a symmetric function of C and CT . Equations (7.26) and (7.51) are identical in form, but the function W
of (7.26) representing the internal energy is, in general, different from that of (7.51), which represents the Helmholtz
free energy. The constitutive equation (7.51) reduces to (7.26) when thermal effects are neglected.
Since W is a function of C and Θ, (7.48b) implies that so is the entropy density η. Then,
∂η ∂η
η̇ = Θ̇ + ĊIJ
∂Θ ∂CIJ
2
∂2W
1 ∂ W
= − Θ̇ + ĊIj (7.52a)
ρ0 ∂Θ2 ∂Θ2
and (7.50) becomes
∂2W ∂2W
ρ
qi,i = Θ Θ̇ + Θ ĊIj + ρh (7.52b)
ρ0 ∂Θ2 ∂Θ2
Equations (7.48b), (7.50), (7.51), and (7.52) together with the equations of continuity and motion provide the
required number of equations for the determination of the 12 unknowns ρ, ui , CIJ , η, and Θ, provided the functional
forms of W and q are known. In particular, for an isotropic body, W must be a function of the strain invariants
IC , IIC , IIIC , and Θ. The general form of the heat flux vector q can be found in Eringen (1967, Sec. 7.10). A
detailed discussion of the nonlinear theory of thermoelasticity is beyond the scope of this book. We shall derive the
explicit forms of the thermoelastic constitutive equations for the linear case in the next section
2005
c by Ajit Mal
7.5. LINEAR CONSTITUTIVE EQUATIONS 217
In this section, we derive the explicit expressions for the constitutive equations under the assumption that the displace-
ment gradients UI,J are small quantities of the first order. We recall that the displacement gradient tensor H with
components UI,J is related to the deformation gradient tensor F through the equation
F=I+H (7.53a)
B ≈ C ≈ I + 2ǫ (7.54)
We further recall from Sec. 6.6 that, to this order of approximation, no distinction need be made between the reference
coordinate X and the spatial coordinate x, or between the Cauchy stress tensor σ and the Piola stress tensors σ and σ̃.
7.5.1 Elasticity
We first consider an elastic material for which the constitutive equation (7.11) is of the general form
σ = Ff(C)FT (7.55)
where f(C) is a symmetric tensor function. Then, for small ǫij , we may expand f(C) in a power series in ǫ and,
retaining only the linear terms, may write
where A is a constant second-order tensor and c is a constant fourth-order tensor. The symmetry of the tensor function
f implies that A is symmetric and
cijkl = cjikl
yielding
cijkl = cijlk
σ ≈ (I + H)(A + cǫ)(I + HT )
≈ A + cǫ + HA + AHT (7.57)
If we assume that both stress and strain ǫ vanish in the reference state, then
A=0
2005
c by Ajit Mal
218 CHAPTER 7. CONSTITUTIVE EQUATIONS
σ = cǫ = cH
or
σij = cijkl ǫkl = cijkl Uk,l (7.58)
where the fourth order tensor c has the symmetry properties
Equation (7.58) is the general form of the linear constitutive relation for an elastic solid and is known as the
generalized Hooke’s law; c is called the elastic tensor. If the solid is inhomogeneous, then the elastic tensor c is a
function of X.
Although the fourth-order tensor c has 81 components cijkl , the symmetry conditions (7.59) imply that only 36 of
these components are independent. This is also obvious from the fact that (7.58) is a set of six linear homogeneous
equations relating the six independent components of σ to the six independent components of ǫ.
We next assume that there exists a strain energy function W (C) from which the stress tensors can be derived
through (7.26) and (7.27). For small strains, W may be expressed in the power series
1
W = W0 + bij ǫij + cijkl ǫij ǫkl + · · · (7.60)
2
where W0 , bij ,.... are constants. Since the strain energy of the material has been assumed to be zero in the undeformed
reference state, W0 must be zero. Furthermore, since ǫ is symmetric, the preceding expression for W may be written
in the alternative forms
1 1
W = bij ǫij + cijkl ǫij ǫkl + · · · = bji ǫij + cjl ǫij ǫkl + · · ·
2 2
1 1
= bij ǫij + cijlk ǫij ǫkl + · · · = bij ǫij + cklij ǫij ǫkl + · · · (7.61)
2 2
These equalities must hold for all possible choices of the strain tensor ǫ. We therefore have
bij = bji
and
cijkl = cjikl = cijlk = cklij (7.62)
The second Piola stress tensor is related to W through (7.27b). Since in the linear theory there is no distinction
between the Cauchy stress tensor σ and the second Piola stress tensor σ̃, we may write, using (7.54),
∂W ∂W
σij ≈ 2 ≈ (7.63)
∂Cij ∂ǫij
Noting that
∂
(cijkl ǫij ǫkl ) = 2cmnkl ǫkl
∂ǫmn
and using the power series expansion (7.60) for W , (7.63) yields
This shows that the linear terms in the stress-strain relations arise from the quadratic terms (in strains) in W . As usual,
if we assume that the material is stress-free in the undeformed state, then bij = 0. It follows that the linear constitutive
equation or the generalized Hooke’s law for an elastic solid is
2005
c by Ajit Mal
7.5. LINEAR CONSTITUTIVE EQUATIONS 219
and that
1 1
W = cijkl ǫij ǫkl = σij ǫij (7.66)
2 2
where the fourth-order tensor c satisfies the symmetry properties given by (7.62). The constitutive equation (7.65)
coincides with (7.58) derived earlier without assuming the existence of a strain energy function. However, the existence
of the strain energy function introduces the additional symmetry condition
cijkl = cklij
which reduces the number of independent components of the tensor c from 36 to 21.
The presence of material symmetries further reduces the number of independent constants that appear in the con-
stitutive equations. For isotropic bodies discussed in Sec. 7.2.1 and in Chapter 1, the constitutive equation must reduce
to the form (7.22a)
σ = α0 I + α1 B + α2 B2 (7.67)
where α0 , α1 , α2 are scalar functions of the three invariants IB , IIB , IIIB of B. Using approximation (7.54) and
retaining only the linear terms in ǫ, we find
IB ≈ 3 + 2ǫkk , IIB ≈ 3 + 4ǫkk , IIIB ≈ 1 + 2ǫkk
Therefore,
α0 ≈ α00 + α01 ǫkk
α1 ≈ α10 + α11 ǫkk
α2 ≈ α20 + α21 ǫkk
where α00 , α01 , · · · are independent of ǫ. Substituting these approximate expressions for α0 , α1 , α2 in (7.67) and
using (7.54), we obtain
σ ≈ σ 0 + λǫKkk I + 2µǫ
where
σ0 = (α00 + α10 + α20 )I
λ = α01 + α11 + α21
µ = α10 + 2α20
Assuming that the stress and strain vanish simultaneously in the reference state, so that σ 0 = 0, the linear constitutive
equation for an isotropic elastic solid reduces to
σ = λǫkk I + 2µǫ (7.68a)
or
σij = λǫkk δij + 2µǫij (7.68b)
The constants λ and µ are the Lamé constants introduced in Chapter 1, and (7.68b) can be seen to be the same as
(1.20).
Isotropic solids have the strongest possible material symmetry, namely, complete symmetry about a point. There
are other, weaker symmetries for naturally occurring (e.g., wood, certain single crystals) or engineered (e.g., fiber-
reinforced composites) materials for which the stress behavior may be symmetric about certain preferred planes or
lines. For these materials, the stress-strain law (7.65) must be form invariant under certain coordinate transformations.
Solids that do not contain point symmetry are called anisotropic, or aelotropic. In the next subsection we show how
the stress-strain equations for such solids containing the smallest possible number of material constants can be derived
from the general equation (7.65) through the use of its form invariance properties under appropriate coordinate trans-
formations.
Since in the linear case the distinction between the Lagrangian and Eulerian descriptions of the deformation disap-
pears, we shall use the symbol e to denote the infinitesimal strain tensor and the symbol u to denote the displacement
vector, although they have been used for finite deformations earlier. Also, the stress tensors will be denoted by σ,
which, in the linear theory, may be either the Cauchy’s stress or one of the Piola stresses. The use of the same symbols
for exact and approximate quantities should not cause any confusion, since all discussions in the remainder of this
chapter deal only with small deformations.
2005
c by Ajit Mal
220 CHAPTER 7. CONSTITUTIVE EQUATIONS
If an elastic solid is subjected to significant thermal effects in addition to mechanical, the relationship between the
stress tensor and the Helmholtz free energy is identical with that between the stress tensor and the internal energy in
the absence of thermal effects. However, the free energy W is now a function of the strain tensor E as well as the
absolute temperature Θ.
The general form of the thermoelastic constitutive equation has been given in Sec. 7.4. In the linear case, the
Helmholtz free energy W must be of the form
1
W = W0 (T ) + cijkl ǫij ǫkl + dij ǫij T, T = Θ − Θ0 (7.69)
2
where W0 is a suitable function of temperature only, cijkm is the elastic tensor, [dij ] is a symmetric, constant matrix,
Θ0 is the temperature of the body in the reference state and T is small compared to Θ0 .
In order to obtain the linear constitutive equations, we use the approximation (7.54) and put
Cij = δij + 2ǫij
We assume that the heat flux vector q is given by the generalized form of the Fourier law of heat conduction,
qi = −βij T,j (7.70)
where [βij ] is a constant and symmetric matrix. Then, by (7.48a) and (7.63),
∂W
σij = = cijkl ǫkl + dij T (7.71)
∂ǫij
From (7.52), (7.69), and (7.70) we obtain the generalized form of the heat conduction equation,
− βij T,ji = −ρCE Ṫ + dij ǫ̇ij Θ + ρh
≈ −ρCE Ṫ + dij ǫ̇ij Θ0 + ρh (7.72)
where
1 ∂ 2 W0
CE = − Θ (7.73)
ρ0 ∂Θ2
and we have neglected the second order term, ǫ̇ij T . In the above, CE is the specific heat (at constant deformation) of
the material. Experimental evidence indicates that CE is constant over a broad range of temperatures for most solids.
In that case, (7.73) can be integrated to yield
W0 (Θ) = −ρCE (Θ ln Θ − Θ) + constant (7.74)
For the special case of isotropy
cijkl = λδij δkl + µ(δik δjl + δil δjk )
and
dij = −α(3λ + 2µ)δij , βij = κδij (7.75)
where α is the coefficient of thermal expansion and κ is the thermal conductivity of the solid. Assuming that the
reference state is at temperature Θ0 , the constitutive equation for an isotropic solid becomes
σij = λǫkk δij + 2µǫij − α(3λ + 2µ)δij T (7.76)
and the heat conduction equation is obtained as
κT,ii = −ρh + CE ρṪ + (3λ + 2µ)αΘ0 ǫ̇kk (7.77)
Equation (7.76) can be seen to be identical with the stress-strain-temperature relation (1.34) of Chapter 1. It should be
noted that the heat conduction equation is modified by the last term in the right-hand side of (7.77), which represents
the coupling between deformation and temperature fields, and is often ignored in engineering applications.
2005
c by Ajit Mal
7.6. LINEAR THERMOELASTICITY 221
PROBLEMS
7.1. The stress-strain law for an isotropic elastic material is of the form
where α0 , α1 , α2 , α3 are constants. The linear elastic (Lamé) constants of the material are λ, µ and the reference
state is undeformed and stress-free.
(a) Obtain the linear stress-strain law from (1) and hence determine α0 , α1 , α2 , in terms of λ, µ.
(b) Obtain the σ − H relationship to second order in H.
7.2. Let W be treated as a function of the Green’s strain tensor E. Assume that the conservation of mechanical energy
of an elastic solid may be expressed in the form
∂W
σij = J −1 FiK FjL (2)
∂EKL
(b) Let
∂W
H = [UI,K ], A=
∂EKL
Jσ = A + AH T + HA + HAH T (3)
(c) Linearize relation (3) for small UI,K and calculate the next higher-order terms.
(d) Show that
∂W
σ̃KL =
∂EKL
where σ̃ is the second Piola stress tensor.
7.3 The stress strain law for a solid are assumed to be of the following alternate forms:
x1 = X1 + γX2 , x 2 = X2 , x 3 = X3
where γ and µ are positive constants. Calculate the components of the Cauchy stress and the First Piola stress
associated with this deformation.
2005
c by Ajit Mal
222 CHAPTER 7. CONSTITUTIVE EQUATIONS
7.5 Assume that the strain energy function for a hyperelastic solid is of the form:
where s = HTH/2. Explain the reason for any discrepancy between your result in (a) and the above result in (b).
(c) A bar of the material is subjected to the uniaxial stress, σ0 along the x1 direction. Determine the nonzero
strain components in terms of σ0 and the material constants, λ, µ, c3 , c4 and c5 using Eq (2). Describe the
differences between this result and that for the linear case.
2005
c by Ajit Mal
7.6. LINEAR THERMOELASTICITY 223
BIBLIOGRAPHY
1. Atkin, R. J. and Fox, N., An Introduction to the Theory of Elasticity, Longman, London, 1980.
4. Fung, Y. C., Foundations of Solid Mechanics, Prentice Hall, Englewood Cliffs, N.J., 1965.
6. Mooney, M. A., Theory of large elastic deformation, J. Appl. Phys., 11, pp. 582-592, 1940.
7. Rivlin, R. S., Large elastic deformations of isotropic materials VI: Further results in the theory of torsion, shear
and flexure, Phil. Trans. Roy. Soc. Lond., A242, pp.173-195, 1949.
8. Rivlin, R. S. and Saunders, D. W., Large elastic deformations of isotropic materials VII: Experiments on the
deformation of rubber, Phil. Trans. Roy. Soc. Lond., A243, pp. 251-288, 1957.
9. Spencer, A. J. M., Constitutive Theory for Strongly Anisotropic Solids, in Continuum Theory of The Mechanics
of Fiber-reinforced Composites, A. J. M. Spencer, Ed., Springer-Verlag, New York, 1984, pp. 1-32.
10. Treloar, L. R. G., Physics of Rubber Elasticity, 2nd ed., Oxford University Press, London, 1974.
11. Truesdell, C., and Noll, W., Continuum Mechanics I: The Mechanical Foundations of Elasticity and Fluid
Dynamics, Gordon and Breach, New York, 1966.
12. Wang, C. -C. and Truesdell, C., Introduction to Rational Elasticity, Noordhoff Int. Publ., Leyden, 1973.
2005
c by Ajit Mal
224 CHAPTER 7. CONSTITUTIVE EQUATIONS
2005
c by Ajit Mal
Chapter 8
Nonlinear Elastostatics
In this section we present the solution of a number of simple nonlinear static problems of isotropic elasticity. In
general, these problems are extremely difficult to solve by means of conventional, direct approaches, whereby the
governing equations of equilibrium are solved subject to the relevant boundary conditions. We employ an inverse
method, in which a suitable deformation of the solid is prescribed and the associated stresses are determined through
the use of the constitutive relations. We restrict our attention to those problems for which an exact solution can be
obtained without the detailed knowledge of the constitutive equation. Such solutions are extremely helpful in revealing
the general features of the nonlinear effects as well as in designing experiments to determine the response functions of
a given material.
For the simple problems to be discussed here, it is convenient to describe the state of stress in the solid by means of
the Cauchy stress tensor σ which is a solution of the equations (see Chap. 4),
σij,j + ρbi = 0 (8.1)
The stress-strain law for an isotropic elastic solid may be expressed in the alternate forms (see Chap. 5)
σ = α0 I + α1 B + α2 B2 (8.2a)
or
σ = β0 I + β1 B + β−1 B−1 (8.2b)
where the α’s and β’s are functions of the invariants of the left Cauchy-Green strain tensor B, which is related to the
deformation gradient tensor F through
B = FFT (8.3)
For a hyperelastic solid with strain energy function W , the response functions can be expressed in the forms
p 2(W1 + IB W2 ) 2W2
α0 = 2 IIIB W3 , α1 = √ , α2 = − √ (8.4)
IIIB IIIB
where IB , IIB , IIIB are the principal invariants of B and W1 , W2 , W3 are the derivatives of W with respect to
IB , IIB , IIIB respectively.
If the solid is hyperelastic and incompressible, then the stress-strain law is simplified to
σ = −pI + 2W1 B − 2W2 B−1 (8.5)
where p is an undetermined hydrostatic pressure.
225
226 CHAPTER 8. NONLINEAR ELASTOSTATICS
We first consider the homogeneous deformation discussed in Sec. 5.6 and described by the equations
xk = akK XK (8.6a)
The associated deformation gradient tensor F and the left Cauchy-Green strain tensor B are
F = A, (8.6b)
B = AAT (8.6c)
The stress within the material is given by (8.2). Since B is independent of the position x, so are its invariants IB , IIB ,
and IIIB . Then all the response functions and, consequently, the stress components are constant throughout the
region occupied by the solid in its deformed state. This implies that the equilibrium equations (8.1) are identically
satisfied in the absence of body forces and that it is always possible to maintain a homogeneous deformation within a
homogeneous solid through the application of surface tractions only.
In particular, if A is diagonal,
A = diag[a1 , a2 , a3 ] (8.7a)
then
J = a1 a2 a3 > 0 (8.7b)
Hence the response functions are constant. The stress-strain equation (8.2b) gives
a1 = a2 = a3 = a > 0 (i)
As indicated in Sec. 5.6, this corresponds to a uniform dilatation of the material. Clearly,
B = C = a2 I (ii)
2005
c by Ajit Mal
8.1. EXACT SOLUTION OF SOME NONLINEAR ELASTIC PROBLEMS 227
The response functions β0 , β1 , and β−1 depend only on a2 . The stress vector on the surface of the deformed solid with
unit normal n is given by
t = −p(a2 )n (iva)
where
Thus the deformation can be maintained through the application of the hydrostatic pressure p given by (ivb) in a
direction normal to the bounding surface of the solid. The ”bulk modulus” of the material can be defined as
−p β0 + β1 a2 + β−1 a−2
k̃(a2 ) = = (v)
J −1 a3 − 1
The bulk modulus is, in general, a function of the material properties of the solid as well as of the amount of the strain.
This is a common characteristic of nonlinear deformations and should be contrasted with the linear case, where the
moduli are independent of the strains.
The linear result can be recovered from (v) by assuming that
a=1+ǫ (vi)
where ǫ is a small quantity of the first order. Then retaining the first-order terms only,
where βij are constants depending on the material properties of the solid. Substituting these in (iii) and assuming that
the stress and strain vanish simultaneously, we obtain
It should be noted that the expression in the left-hand side of (viiia) is the value of β0 + β1 + β−1 for a = 1. Equation
(viiib) can be seen to be identical with the linear stress-strain law (2.21b) discussed in Sec. 2.4. The linear bulk
modulus k can be expressed in terms of the material constants by means of (v) as
Other conditions are imposed on the response functions based on the fact that a compressive stress should result in
a decrease in volume and that an increase in applied pressure should further decrease the volume. Thus, p > 0 when
a < 1, p < 0 when a > 1, and dp/da < 0, where p is given by (ivb).
a1 = a, a2 = a3 = b (i)
The corresponding deformation is a uniform extension (for a > 0) or contraction (for a < 0) in the 1-direction
accompanied by equal amounts of contraction or extension in the 2- and 3-directions. Then, from (8.10) and (8.11),
the shear stresses are zero and the normal stresses are given by
where β0 , β1 , β−1 depend only on a2 and b2 . These equations imply that a rectangular block of the material aligned
with the coordinate axes can be maintained in a state of equilibrium under the action of uniaxial tension
σ11 = T
2005
c by Ajit Mal
228 CHAPTER 8. NONLINEAR ELASTOSTATICS
σ 22
σ12
x2 σ11
tan (1/ Λ)
-1
x1
Fig. 8-1. Simple shear of a block.
on its end faces normal to the 1-axis, the other two faces being free of traction. Then we must have
We can define the Young’s modulus E and the Poisson’s ratio ν of the solid through the equations
T −(b − 1)
E= , ν= (ivab)
a−1 a−1
Then
β0 + β1 a2 + β−1 a−2
E= (v)
a−1
and ν is a solution of the simultaneous equations (iiia) and (ivb). As in the case of the bulk modulus just defined, both
the Young’s modulus and the Poisson’s ratio are strain-dependent. For small deformations we may write
a = 1 + ǫ, b = 1 − νǫ (vi)
where ǫ is a small quantity of the first order and obtain results similar to (viii) of Example 8.1-1. It is also possible to
deduce certain restrictions on the properties of the response functions based on physically reasonable behavior of the
solid under uniaxial tension.
As another example of homogeneous deformation, consider the simple shear discussed in Sec. 5.6(ii),
x1 = X1 + ΛX2 , x 2 = X2 , x 3 = X3 (i)
which deforms a rectangle in the X1 X2 -plane aligned with the coordinate axes in the reference configuration into a
parallelogram with angles tan−1 (1/Λ) and π − tan−1 (1/Λ) (Fig. 8-1).
For this deformation
1 + Λ2 Λ
1 Λ 0 0
F = 0 1 0 , B= Λ 1 0 (ii)
0 0 1 0 0 1
1 −Λ 0
B −1 = −Λ 1 + Λ2 0
0 0 1
2005
c by Ajit Mal
8.1. EXACT SOLUTION OF SOME NONLINEAR ELASTIC PROBLEMS 229
Thus the deformation is isochoric and the response functions are functions of Λ2 as well as the material properties.
The stress-strain law (8.2b) may be expressed in the form
0 1 0 1 0 0 0 0 0
[σij ] = µΛ 1 0 0 + β1 Λ2 0 0 0 + β−1 Λ2 0 1 0 (iva)
0 0 0 0 0 0 0 0 0
where we have assumed that the reference configuration is free from stress and strain and
The strain-dependent quantity µ(Λ2 ) may be defined to be the shear modulus of the material. Note that µ is an
even function of Λ, so that σ12 is an odd function of Λ, indicating that a change in the direction of the shear stress
results in a change in the direction of the shear strain also. We also expect that µ(Λ2 ) > 0. From (v) it follows that
if all the normal stresses are zero, then for nonzero Λ, the shear stress is also zero, and therefore the block could be
deformed without any stress whatsoever! We make the assumption that for an elastic material there can be only one
state for which the stress is zero, and we have already taken that state as the undeformed natural state. Thus from
(v) we must have σ11 6= σ22 for Λ 6= 0, so that at least one normal stress must be present in order to maintain the
simple shear. The existence of unequal normal stresses in simple shear is called the Poynting effect. In the linear case,
(Λ → 0) and µ(Λ2 ) → µ0 , the shear modulus of linear elasticity. Thus, from equations (v) we see that for small
strains, σ12 = O(Λ) whereas σ11 − σ22 = O(Λ2 ); hence for small deformations, the difference in the normal stresses
is of second order. It is interesting to note that the second order effect appears as normal stresses rather than as a
departure from linearity of the shear stress. The equation
is a universal relation between the deformation and the various components of the stress tensor, and is independent of
the material response functions.
For a hyperelastic incompressible material, the stress-strain law is given by (8.5). If the material is subjected to
the simple shear deformation described in Example 8.1-3, the stress components may be expressed in the form
In view of Eq. (iii) of Example 8.1-3, W1 and W2 are independent of position; since the stress components satisfy
the equations of equilibrium without body forces, p is a constant. As in the case of the compressible material discussed
in the previous section, a shear modulus µ can be defined through
µ = 2(W1 + W2 ) (ii)
and Eq. (v) of Example 8.1-3 still holds, indicating that the Poynting effect is present in incompressible materials as
well. However, in contrast to the compressible case, we can choose p appropriately to make any one of the normal
stresses vanish. For example, in this problem we can make σ33 = 0 by choosing
p = 2(W1 − W2 )
2005
c by Ajit Mal
230 CHAPTER 8. NONLINEAR ELASTOSTATICS
γx3
x3
0
2
1
Fig. 8-2. Simple torsion of a cylinder.
Then
σ11 = 2Λ2 W1 , σ22 = −2Λ2 W2 (iii)
σ12 = Λµ(Λ2 ), σi3 = 0
If W1 > 0, W2 > 0, then (ii) gives µ(Λ2 ) > 0. Further, from (iii) we can see that σ11 > 0 and σ22 < 0; i.e., a tensile
stress along x1 and a compressive stress along x2 are needed to maintain the deformation.
We next consider the simple torsion of an incompressible right circular cylinder subjected to end loads (Fig. 8-2).
The deformation of the cylinder can be expressed by the equations
x1 = cX1 − sX2 , x2 = sX1 + cX2 , x 3 = X3 (i)
where c = cos(γX3 ), s = sin(γX3 ), and γ is a constant. It can be easily seen that under this deformation, the radius
of the cylinder remains unchanged and all radial lines on the cross section rotate through the angle γx3 .
The deformation gradient tensor F and the Green’s strain tensor B are given by
cos(γX3 ) −sin(γX3 ) −γx2
F = sin(γX3 ) cos(γX3 ) γx1 (iia)
0 0 1
1 + γ 2 x22 −γ 2 x1 x2 −γx2
2005
c by Ajit Mal
8.1. EXACT SOLUTION OF SOME NONLINEAR ELASTIC PROBLEMS 231
Stress-strain law:
′ ′ ′
σ = −pI + 2W1 B − 2W2 B −1
σrr = −p + 2W1 − 2W2
σφφ = −p + 2W1 (1 + γ 2 r2 ) − 2W2 (iv)
σzz = −p + 2W1 − 2W2 (1 + γ 2 r2 )
σφz = 2γr(W1 + W2 ), σrφ = σrz = 0
Stresses are symmetric, i.e., independent of φ.
Equilibrium:
Equations of equilibrium in cylindrical coordinates (4.76):
∂σrr 1 ∂σφr ∂σzr 1
+ + + (σrr − σφφ ) + fr = 0
∂r r ∂φ ∂z r
∂σrφ 1 ∂σφφ ∂σzφ 2 ∂p
+ + + σrφ + fφ = 0 → =0 (v)
∂r r ∂φ ∂z r ∂φ
∂σrz 1 ∂σφz ∂σzz 1 ∂p
+ + + σrz + fz = 0 → =0
∂r r ∂φ ∂z r ∂z
φ and z components are identically zero.
r-components:
dσrr 1
+ (σrr − σφφ ) = 0 (vi)
dr r
dp 2d
= (W1 − W2 ) − 2γ 2 rW1 (vii)
dr dr Z r
p(r) = 2(W1 − W2 ) − 2γ 2 ξW1 (ξ)dξ + c (viii)
B.C.: σrr = 0 at r = a
Z a
p(r) = 2(W1 − W2 ) + 2γ 2 ξW1 (ξ)dξ (ix)
r
Hence:
Z a
2
σrr = −p + 2(W1 − W2 ) = −2γ ξW1 (ξ)dξ
r
Z a
σφφ = 2γ 2 r2 W1 − 2γ 2 ξW1 (ξ)dξ (x)
r
Z a
σzz = −2γ 2 r2 W2 − 2γ 2 ξW1 (ξ)dξ
r
σφz = 2γr(W1 + W2 )
2005
c by Ajit Mal
232 CHAPTER 8. NONLINEAR ELASTOSTATICS
Thus the resultant force on the end section is along the normal to the section and the normal component F3 of the
resultant force is given by
Z Z 2π Z a Z a
F3 = σ33 dS = −2γ 2 ξW1 dξ − 2γ 2 r2 W2 r dr dφ
S 0 0 r
Z a Z a
= −2πγ 2 2r ξW1 dξ + 2r3 W2 dr (xi)
0 r
But
a a a a
1 a 3
1 2
Z Z Z Z
r ξW1 dξ dr = r ξW1 dξ r W1 dr +
0 r 2 r 0 2 0
a
r2 1 a 3 1 a 3
Z Z
= − 2 t(er ) · er + r W1 dr = r W1 dr
4γ 0 2 0 2 0
2005
c by Ajit Mal
8.1. EXACT SOLUTION OF SOME NONLINEAR ELASTIC PROBLEMS 233
PROBLEMS
8.1. A rectangular block occupying the region A < X1 < 4A, √−5A < X2 < 5A, −A < √ X3 < A in its undeformed
state is subjected to the plane strain deformation, x1 = AX1 cos(X2 /L), x2 = AX1 sin(X2 /L), x3 = X3 ,
where A and L are constants.
(a) Using polar coordinates (r, φ) in the spatial configuration, determine the deformed shape of the boundaries
of the block. Determine the maximum or minimum value of the ratio L/A so that the material does not
penetrate into itself.
(b) Calculate B and its principal invariants in terms of r, φ.
(c) Assuming that the material is incompressible with strain energy function W , obtain the stresses produced
in the material. Determine the undetermined pressure p and the normal component of the stress vector on
the faces X3 = ±A required to maintain the deformation.
8.2. Let S0 denote the smooth surface of a control volume V0 in the undeformed state of an elastic solid and St , Vt
the corresponding surface and volume in the spatial configuration. There are no body forces. Show that
Z
(W NK − σ Ij Uj,K NI ) dS(Y) = 0
S0
where W is the strain energy density, σ is the first Piola stress tensor, U is the displacement vector, and N is the
unit normal to S0 at Y. Derive the linearized form of (1).
8.3. Consider a hollow cylinder of internal radius A0 and external radius A1 in the undeformed state. The material is
incompressible, of the Mooney-Rivlin type. The cylinder is subjected to an internal pressure p0 and end loads
with resultant force F3 and moment M3 . The deformation of the cylinder is of the form
BIBLIOGRAPHY
1. Atkin, R. J. and Fox, N., An Introduction to the Theory of Elasticity, Longman, London, 1980.
2. Fung, Y. C., Foundations of Solid Mechanics, Prentice Hall, Englewood Cliffs, N.J., 1965. New York,
1956.
2005
c by Ajit Mal