Vous êtes sur la page 1sur 230







Professor of Mechanics Michigan State University



New York, St. Louis, San Francisco, London, Sydney, Toronto, Mexico, and Panama

Copyright © 1H7() hy MeGraw-HilI, Iru-. All Right;; Reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher.

ISBN 07-040663-4

7 8 9 10 II 12 13 14 15 SH SH 8 7 6 5 4


Because of its emphasis on basic concepts and fundamental principles, Continuum Mechanics has an important role in modern engineering and technology. Several undergraduate courses which utilize the continuum concept and its dependent theories in the training of engineers and scientists are well established in today's curricula and their number continues to grow. Graduate programs in Mechanics and associated areas have long recognized the value of a substantial exposure to the subject. This book has been written in an attempt to assist both undergraduate and first year graduate students in understanding the fundamental principles of continuum theory. By including a number of solved problems in each chapter of the book, it is further hoped that the student will be able to develop his skill in solving problems in both continuum theory and its related fields of application.

In the arrangement and development of the subject matter a sufficient degree of continuity is provided so that the book may be suitable as a text for an introductory course in Continuum Mechanics. Otherwise, the book should prove especially useful as a supplementary reference for a number of courses for which continuum methods provide the basic structure. Thus courses in the areas of Strength of Materials, Fluid Mechanics, Elasticity, Plasticity and Viscoelasticity relate closely to the substance of the book and may very well draw upon its contents.

Throughout most of the book the important equations and fundamental relationships are presented in both the indicial or "tensor" notation and the classical symbolic or "vector" notation. This affords the student the opportunity to compare equivalent expressions and to gain some familiarity with each notation. Only Cartesian tensors are employed in the text because it is intended as an introductory volume and since the essence of much of the theory can be achieved in this context.

The work is essentially divided into two parts. The first five chapters deal with the basic continuum theory while the final four chapters cover certain portions of specific areas of application. Following an initial chapter on the mathematics relevant to the study, the theory portion contains additional chapters on the Analysis of Stress, Deformation and Strain, Motion and Flow, and Fundamental Continuum Laws. Applications are treated in the final four chapters on Elasticity, Fluids, Plasticity and Viscoelasticity. At the end of each chapter a collection of solved problems together with several exercises for the student serve to illustrate and reinforce the ideas presented in the text.

The author acknowledges his indebtedness to many persons and wishes to express his gratitude to all for their help. Special thanks are due the following: to my colleagues, Professors W. A. Bradley, L. E. Malvern, D. H. Y. Yen, J. F. Foss and G. LaPalm each of whom read various chapters of the text and made valuable suggestions for improvement; to Professor D. J. Montgomery for his support and assistance in a great many ways; to Dr. Richard Hartung of the Lockheed Research Laboratory, Palo Alto, California, who read the preliminary version of the manuscript and gave numerous helpful suggestions; to Professor M. C. Stippes, University of Illinois, for his invaluable comments and suggestions; to Mrs. Thelma Liszewski for the care and patience she displayed in typing the manuscript; to Mr. Daniel Schaum and Mr. Nicola Monti for their continuing interest and guidance throughout the work. The author also wishes to express thanks to his wife and children for their encouragement during the writing of the book.

Michigan State University June 1970


Chapter 1
Chapter 2


Tensors and Continuum Mechanics General Tensors. Cartesian Tensors.

Tensor Rank .


1 1

8 10


20 21 22 23 23

Vectors and Scalars 2

Vector Addition. Multiplication of a Vector by a Scalar.................. 2

Dot and Cross Products of Vectors 3

Dyads and Dyadics 4

Coordinate Systems. Base Vectors. Unit Vector Triads.................. 6

Linear Vector Functions. Dyadics as Linear Vector Operators 8

Indicial Notation. Range and Summation Conventions Summation Convention Used with Symbolic Convention

Coordinate Transformation. General Tensors 11

The Metric Tensor. Cartesian Tensors 12

Transformation Laws for Cartesian Tensors. The Kronecker Delta.

Orthogonality Conditions 13

Addition of Cartesian Tensors. Multiplication by a Scalar 15

Tensor Multiplication 15

Vector Cross Product. Permutation Symbol. Dual Vectors 16

Matrices. Matrix Representation of Cartesian Tensors 17

Symmetry of Dyadics, Matrices and Tensors

Principal Values and Principal Directions of Symmetric

Second-Order Tensors .

Powers of Second-Order Tensors. Hamilton-Cayley Equation .

Tensor Fields. Derivatives of Tensors .

Line Integrals.

Stokes' Theorem

The Divergence Theorem of Gauss


The Continuum Concept .

Homogeneity. Isotropy. Mass-Density .

Body Forces. Surface Forces .

Cauchy's Stress Principle. The Stress Vector .

State of Stress at a Point. Stress Tensor .

The Stress Tensor-Stress Vector Relationship .

Force and Moment. Equilibrium. Stress Tensor Symmetry .

Stress Transformation Laws

Stress Quadric of Cauchy .

Principal Stresses. Stress Invariants. Stress Ellipsoid .

Maximum and Minimum Shear Stress Values .

Mohr's Circles for Stress .

Plane Stress

Deviator and Spherical Stress Tensors 57

44 44 45 45 46 47 48 49 50 51 52 54 56

Chapter 3
Chapter 4
Chapter 5
Chapter 6



Particles and Points


Continuum Configuration. Deformation and Flow Concepts 77

Position Vector. Displacement Vector


Lagrangian and Eulerian Descriptions 79

Deformation Gradients. Displacement Gradients 80

Deformation Tensors. Finite Strain Tensors 81

Small Deformation Theory. Infinitesimal Strain Tensors


Relative Displacements. Linear Rotation Tensor. Rotation Vector 83

Interpretation of the Linear Strain Tensors Stretch Ratio. Finite Strain Interpretation

84 86

Stretch Tensors. Rotation Tensor 87

Transformation Properties of Strain Tensors 88

Principal Strains. Strain Invariants. Cubical Dilatation 89

Spherical and Deviator Strain Tensors 91

Plane Strain. Mohr's Circles for Strain 91

Compatibility Equations for Linear Strains 92


Motion. Flow. Material Derivative

110 111 112 112 113 114 116

Velocity. Acceleration. Instantaneous Velocity Field .

Path Lines. Stream Lines. Steady Motion .

Rate of Deformation. Vorticity. Natural Strain .

Physical Interpretation of Rate of Deformation and Vorticity Tensors .

Material Derivatives of Volume, Area and Line Elements .

Material Derivatives of Volume, Surface and Line Integrals .


Conservation of Mass. Continuity Equation 126

Linear Momentum Principle. Equations of Motion.

Equilibrium Equations 127

Moment of Momentum (Angular Momentum) Principle 128

Conservation of Energy. First Law of Thermodynamics.

Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

Equations of State. Entropy. Second Law of Thermodynamics 130

The Clausius-Duhem Inequality. Dissipation Function 131

Constitutive Equations. Thermomechanical and Mechanical Continua 132


Generalized Hooke's Law. Strain Energy Function 140

Isotropy. Anisotropy. Elastic Symmetry 141

Isotropic Media. Elastic Constants


Elastostatic Problems. Elastodynamic Problems . . . . . . . . . . . . . . . . . . . . . . . . 143

Theorem of Superposition. Uniqueness of Solutions.

St. Venant's Principle 145

Two-Dimensional Elasticity. Plane Stress and Plane Strain 145

Chapter 7
Chapter 8
Chapter 9

Page Airy's Stress Function .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

Two-Dimensional Elastostatic Problems in Polar Coordinates 148

Hyperelasticity. Hypoelasticity 149

Linear Thermoelasticity 149


Fluid Pressure. Viscous Stress Tensor. Barotropic Flow 160

Constitutive Equations. Stokesian Fluids. Newtonian Fluids 161

Basic Equations for Newtonian Fluids. Navier-Stokes-Duhem Equations. . . . 162

Steady Flow. Hydrostatics. Irrotational Flow 163

Perfect Fluids. Bernoulli Equation. Circulation 164

Potential Flow. Plane Potential Flow 165


Basic Concepts and Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

Idealized Plastic Behavior 176

Yield Conditions. Tresca and von Mises Criteria 177

Stress Space. The II-Plane. Yield Surfaces 17:)

Post- Yield Behavior. Isotropic and Kinematic Hardening Plastic Stress-Strain Equations. Plastic Potential Theory

180 181

Equivalent Stress. Equivalent Plastic Strain Increment 182

Plastic Work. Strain Hardening Hypotheses 182

Total Deformation Theory 183

Elastoplastic Problems


Elementary Slip Line Theory for Plane Plastic Strain 184


Linear Viscoelastic Behavior Simple Viscoelastic Models

196 196

Generalized Models. Linear Differential Operator Equation 198

Creep and Relaxation 199

Creep Function. Relaxation Function. Hereditary Integrals 200

Complex Moduli and Compliances 202

Three Dimensional Theory 203

Viscoelastic Stress Analysis. Correspondence Principle 204

INDEX.... 217

Chapter 1

Mathematical Foundations


Continuum mechanics deals with physical quantities which are independent of any particular coordinate system that may be used to describe them. At the same time, these physical quantities are very often specified most conveniently by referring to an appropriate system of coordinates. Mathematically, such quantities are represented by tensors.

As a mathematical entity, a tensor has an existence independent of any coordinate system. Yet it may be specified in a particular coordinate system by a certain set of quantities, known as its components. Specifying the components of a tensor in one coordinate system determines the components in any other system. Indeed, the law of transformation of the components of a tensor is used here as a means for defining the tensor. Precise statements of the definitions of various kinds of tensors are given at the point of their introduction in the material that follows.

The physical laws of continuum mechanics are expressed by tensor equations. Because tensor transformations are linear and homogeneous, such tensor equations, if they are valid in one coordinate system, are valid in any other coordinate system. This invariance of tensor equations under a coordinate transformation is one of the principal reasons for the usefulness of tensor methods in continuum mechanics.


In dealing with general coordinate transformations between arbitrary curvilinear coordinate systems, the tensors defined are known as general tensors. When attention is restricted to transformations from one homogeneous coordinate system to another, the tensors involved are referred to as Cartesian tensors. Since much of the theory of continuum mechanics may be developed in terms of Cartesian tensors, the word "tensor" in this book means "Cartesian tensor" unless specifically stated otherwise.

Tensors may be classified by rank, or order, according to the particular form of the transformation law they obey. This same classification is also reflected in the number of components a given tensor possesses in an n-dimensional space. Thus in a three-dimensional Euclidean space such as ordinary physical space, the number of components of a tensor is 3N, where N is the order of the tensor. Accordingly a tensor of order zero is specified in any coordinate system in three-dimensional space by one component. Tensors of order zero are called scalars. Physical quantities having magnitude only are represented by scalars. Tensors of order one have three coordinate components in physical space and are known as vectors. Quantities possessing both magnitude and direction are represented by vectors. Second-order tensors correspond to dyadics. Several important quantities in continuum mechanics are represented by tensors of rank two. Higher order tensors such as triadics, or tensors of order three, and tetradics, or tensors of order four are also defined and appear often in the mathematics of continuum mechanics.




[CHAP. 1


Certain physical quantities, such as force and velocity, which possess both magnitude and direction, may be represented in a three-dimensional space by directed line segments that obey the parallelogram law of addition. Such directed line segments are the geometrical representations of first-order tensors and are called vectors. Pictorially, a vector is simply an arrow pointing in the appropriate direction and having a length proportional to the magnitude of the vector. Equal vectors have the same direction and equal magnitudes. A unit vector is a vector of unit length. The null or zero vector is one having zero length and an unspecified direction. The negative of a vector is that vector having the same magnitude but opposite direction.

Those physical quantities, such as mass and energy, which possess magnitude only are represented by tensors of order zero which are called scalars.

In the symbolic, or Gibbs notation, vectors are designated by bold-faced letters such as a, b, etc. Scalars are denoted by italic letters such as a, b, A, etc. Unit vectors are further distinguished by a caret placed over the bold-faced letter. In Fig. 1-1, arbitrary vectors a and b are shown along with the unit vector e and the pair of equal vectors c and d.

A e


Fig. 1-1

The magnitude of an arbitrary vector a is written simply as a, or for emphasis it may be denoted by the vector symbol between vertical bars as far.


Vector addition obeys the parallelogram law, which defines the vector sum of two vectors as the diagonal of a parallelogram having the component vectors as adjacent sides. This law for vector addition is equivalent to the triangle rule which defines the sum of two vectors as the vector extending from the tail of the first to the head of the second when the summed vectors are adjoined head to tail. The graphical construction for the addition of a and b by the parallelogram law is shown in Fig. 1-2(a). Algebraically, the addition process is expressed by the vector equation

a + b = b + a = c (1.1)

Vector subtraction is accomplished by addition of the negative vector as shown, for example, in Fig. 1-2(b) where the triangle rule is used. Thus

a - b = -b + a = d (1.2)

The operations of vector addition and subtraction are commutative and associative as illustrated in Fig. 1-2(c), for which the appropriate equations are

(a + b) + g = a + (b + g) = h (1.3)



~ g g





Fig. 1-2

CHAP. 1]



Multiplication of a vector by a scalar produces in general a new vector having the same direction as the original but a different length. Exceptions are multiplication by zero to produce the null vector, and multiplication by unity which does not change a vector. Multiplication of the vector b by the scalar m results in one of the three possible cases shown in Fig. 1~3, depending upon the numerical value of m.




m > 1

0< m < 1 Fig. 1-3

m < 0

Multiplication of a vector by a scalar is associative and distributive. Thus

m(nb) = (mn)b = n(mb) (1.4)

(m+n)b = (n+m)b = mb+nb (1.5)

m(a + b) = m(b + a) = ma + mb (1.6)

In the important case of a vector multiplied by the reciprocal of its magnitude, the result is a unit vector in the direction of the original vector. This relationship is expressed

by the equation "

b = bib (1.7)


The dot or scalar product of two vectors a and b is the scalar

A = a·b = b'a = abcos8


in which 8 is the smaller angle between the two vectors as shown in Fig. 1-4(a). The dot product of a with a unit vector e gives the projection of a in the direction of e.



A e



Fig. 1-4

The cross or vector product of a into b is the vector v given by v = a X b = -b X a = (ab sin 8) e


in which 8 is the angle less than 1800 between the vectors a and b, and e is a unit vector perpendicular to their plane such that a right-handed rotation about e through the angle 8 carries a into b. The magnitude of v is equal to the area of the parallelogram having a and b as adjacent sides, shown shaded in Fig. 1-4(b). The cross product is not commutative.




The scalar triple product is a dot product of two vectors, one of which is a cross product.

a'(bXc) = (axb)'c = a·(bxc)=.\


As indicated by (1.10) the dot and cross operation may be interchanged in this product. Also, since the cross operation must be carried out first, the parentheses are unnecessary and may be deleted as shown. This product is sometimes written [abc] and called the box product. The magnitude .\ of the scalar triple product is equal to the volume of the parallelepiped having a, b, c as coterminous edges.

The vector triple product is a cross product of two vectors, one of which is itself a cross product. The following identity is frequently useful in expressing the product of a crossed into b X c.

a X (b X c) = (a' c)b - (a' b)c = w

From (1.11), the product vector w is observed to lie in the plane of band c.



The indeterminate vector product of a and b, defined by writing the vectors in juxtaposition as ab is called a dyad. The indeterminate product is not in general commutative, i.e. ab =F ba. The first vector in a dyad is known as the antecedent, the second is called the consequent. A dyadic D corresponds to a tensor of order two and may always be represented as a finite sum of dyads


which is, however, never unique. In symbolic notation, dyadics are denoted by bold-faced sans-serif letters as above.

If in each dyad of (1.12) the antecedents and consequents are interchanged, the resulting dyadic is called the conjugate dyadic of D and is written


If each dyad of D in (1.12) is replaced by the dot product of the two vectors, the result is a scalar known as the scalar of the dyadic D and is written

Ds = al' b1 + a2' b2 + ... + aN' bN


If each dyad of D in (1.12) is replaced by the cross product of the two vectors, the result is called the vector of the dyadic D and is written

Dv = ar X b, + a2 X bs + ... + a» X b»


It can be shown that Dc, Ds and Dv are independent of the representation (1.12).

The indeterminate vector product obeys the distributive laws a(b + c) = ab + ac

(a + b)c ac + be

(a + b)(c + d) = ac + ad + be + bd

(1.16) (1.17) (1.18)

and if .\ and }-t are any scalars,

(x + }-t)ab = xab + }-tab (.\a)b = a(.\b) = Xab

(1.19) (1.20)




If v is any vector, the dot products V' 0 and O· v are the vectors defined respectively by

V' 0 (v- al)bl + (v- a2)b2 + ... + (v- aN)bN U (1.21)




In (1.21) 0 is called the postfactor, and in (1.22) it is called the prefactor. Two dyadics 0 and E are equal if and only if for every vector v, either

V' 0 = V' E or o· v = E· v (1.23)

The unit dyadic, or idemfactor I, is the dyadic which can be represented as


where CI, C2, C3 constitute any orthonormal basis for three-dimensional Euclidean space (see Section 1.7). The dyadic I is characterized by the property

for all vectors v.

The cross products v X 0 and 0 x v are the dyadics defined respectively by

v x 0 (v X al)bl + (v X a2)b2 + ... + (v X aN)bN F (1.26)

Iv v = v r l = v





The dot product of the dyads ab and cd is the dyad defined by ah+cd = (b'c)ad

From (1.28), the dot product of any two dyadics 0 and E is the dyadic

o· E = (a.b. + a2b2 + ... + aNbN) • (cid. + c2d2 + ... + cNdN) = (bl' cI)aldl + (b1 • c2)ald2 + ... + (bN' cN)aNdN = G The dyadics 0 and E are said to be reciprocal of each other if

E'O = O'E = I

For reciprocal dyadics, the notation E = 0-1 and 0 = E-l is often used.




Double dot and cross products are also defined for the dyads ab and cd as follows,
ab : cd (a' c)(b· d) A, a scalar (1.31)
ab ~ cd (a x c)(b' d) h, a vector (1.32)
ab ~ cd (a'c)(bxd) g, a vector (1.33)
ab ~ cd (a x c)(b X d) uw, a dyad (1.34) From these definitions, double dot and cross products of dyadics may be readily developed. Also, some authors use the double dot product defined by

ab· • cd = (b- c)(a' d) = A, a scalar A dyadic 0 is said to be self-conjugate, or symmetric, if

o = n,

and anti-self-conjugate, or anti-symmetric, if

o = -Dc




Every dyadic may be expressed uniquely as the sum of a symmetric and anti-symmetric dyadic. For the arbitrary dyadic 0 the decomposition is


o = t(O + Dc) + t(0 - Dc) = G + H




[CHAP. 1


Gc = !(Dc + (Dc)c) = !(Dc + D) = G (symmetric)

Hc = !(Dc - (Dc)c) = !(Dc - D) = -H (anti-symmetric)

(1.39) , (1.40)

for which

Uniqueness is established by assuming a second decomposition, 0 = G* + H*. Then G* + H* = G + H

and the conjugate of this equation is

G* - H* = G - H



Adding and subtracting (1.41) and (1.42) in turn yields respectively the desired equalities, G* = G and H* = H.


A vector may be defined with respect to a particular coordinate system by specifying the components of the vector in that system. The choice of coordinate system is arbitrary, but in certain situations a particular choice may be advantageous. The reference system of coordinate axes provides units for measuring vector magnitudes and assigns directions in space by which the orientation of vectors may be determined.

The well-known rectangular Cartesian coordi-

nate system is often represented by the mutually perpendicular axes, Oxyz shown in Fig. 1-5. Any vector v in this system may be expressed as a linear combination of three arbitrary, nonzero, noncoplanar vectors of the system, which are called base vectors. For base vectors a, b, c and suitably chosen scalar coefficients A, p., v the vector v is given by

v = Aa + p.b + vC


Base vectors are by hypothesis linearly independent, i.e. the equation

Aa + p.b + vC = 0 (1.44)

is satisfied only if A = p. = v = O. A set of base vectors in a given coordinate system is said to constitute a basis for that system.

Fig. 1-5

The most frequent choice of base vectors for the rectangular Cartesian system is the set of unit vectors t, j, k along the coordinate axes as shown in Fig. 1-5. These base vectors constitute a right-handed unit vector triad, for which

i x j = k, j x k = t, k x i



i 0 i

joj = kok = 1




A A .At.. '" A A


Such a set of base vectors is often called an orthonormal basis.

In terms of the unit triad i, j, k, the vector v shown in Fig. 1-6 below may be expressed by

v = vxi + v3 + vzk

in which the Cartesian components



CHAP. 1]



Cv vl»

= (cos a) i + (cos In j + (cos y)k (1.48) Since v is arbitrary, it follows that any unit vector will have the direction cosines of that vector as its Cartesian components.

In Cartesian component form the dot product of a and b is given by

a·b = (as +ayj +azk)' (bxi +byj +bzk)

= axbx + ayby + azbz (1.49)

For the same two vectors, the cross product a x b is

a x b = (aybz-azby)i + (azbx-axbz)j + (axby-aybx)k This result is often presented in the determinant form



v- i

V cos a



v- j

v cos f3


VZ v- k v cos y

are the projections of v onto the coordinate axes. The unit vector in the direction of v is given according to (1.7) by



i j k

Fig. 1-6



bx by bz

in which the elements are treated as ordinary numbers. The triple scalar product may also be represented in component form by the determinant


Cx Cy Cz


In Cartesian component form, the dyad ab is given by

A A A A. A A

ab (axi+ayj+azk)(bxi+byj+bzk)


axbxi i + axbyi j + axbzi k


+ aybxj i + aybyj j + aybzj k


+ azbxk i + azbykj + azbzkk (1.53)

Because of the nine terms involved, (1.53) is known as the nonion form of the dyad abo It is possible to put any dyadic into non ion form. The nonion form of the idemfactor in terms of the unit triad i,1, k is given by


I = i i + j j + kk


In addition to the rectangular Cartesian coordinate system already discussed, curvilinear coordinate systems such as the cylindrical (R, 8, z) and spherical (r, 8, </» systems show~ in Fig. 1-7 below are also widely used. Unit triads (CR' Ce, cz) and (CT' ee, ect» of base vectors illustrated in the figure are associated with these systems. However, the base vectors here do not all have fixed directions and are therefore, in general, functions of position.



[CHAP. 1





(a) Cylindrical

(b) Spherical

Fig. 1-7

1.8 LINEAR VECTOR FUNCTIONS. DYADICS AS LINEAR VECTOR OPERATORS A vector a is said to be a function of a second vector b if a is determined whenever b is given. This functional relationship is expressed by the equation

a = £(b)

The function f is said to be linear when the conditions


£(b+c) = £(b) + £(c) £('\b) = '\£(b)

are satisfied for all vectors band c, and for any scalar ,\.

Writing b in Cartesian component form, equation (1.55) becomes

(1.56) (1.57)


a = £(bxi + byj + bzk)

which, if f is linear, may be written




In (1.59) let £ei) = u, £(j) = v, £(k) = w, so that now

a = u(i.b)+v(j·b)+w(k·b) = (ui+vj+wk)·b which is recognized as a dyadic-vector dot product and may be written a = D·b



where D = u i + v j + wk. This demonstrates that any linear vector function f may be expressed as a dyadic-vector product. In (1.61) the dyadic D serves as a linear vector operator which operates on the argument vector b to produce the image vector a.


The components of a tensor of any order, and indeed the tensor itself, may be represented clearly and concisely by the use of the indicial notation. In this notation, letter indices, either subscripts or superscripts, are appended to the generic or kernel letter representing the tensor quantity of interest. Typical examples illustrating use of indices are the tensor symbols

CHAP. 1]



In the "mixed" form, where both subscripts and superscripts appear, the dot shows that j is the second index.

Under the rules of indicial notation, a letter index may occur either once or twice in a given term. When an index occurs unrepeated in a term, that index is understood to take on the values 1,2, ... , N where N is a specified integer that determines the range of the index. Un repeated indices are known as free indices. The tensorial rank of a given term is equal to the number of free indices appearing in that term. Also, correctly written tensor equations have the same letters as free indices in every term.

When an index appears twice in a term, that index is understood to take on all the values of its range, and the resulting terms summed. In this so-called summation convention, repeated indices are often referred to as dummy indices, since their replacement by any other letter not appearing as a free index does not change the meaning of the term in which they occur. In general, no index occurs more than twice in a properly written term. If it is absolutely necessary to use some index more than twice to satisfactorily express a certain quantity, the summation convention must be suspended.

The number and location of the free indices reveal directly the exact tensorial character of the quantity expressed in the indicial notation. Tensors of first order are denoted by kernel letters bearing one free index. Thus the arbitrary vector a is represented by a symbol having a single subscript or superscript, i.e. in one or the other of the two forms,

The following terms, having only one free index, are also recognized as first-order tensor quantities:

Second-order tensors are denoted by symbols having two free indices. Thus the arbitrary dyadic D will appear in one of the three possible forms

In the "mixed" form, the dot shows that j is the second index. Second-order tensor quantities may also appear in various forms as, for example,

By a logical continuation of the above scheme, third-order tensors are expressed by symbols with three free indices. Also, a symbol such as ,\ which has no indices attached, represents a scalar, or tensor of zero order.

In ordinary physical space a basis is composed of three, non coplanar vectors, and so any vector in this space is completely specified by its three components. Therefore the range on the index of a;, which represents a vector in physical three-space, is 1,2,3. Accordingly the symbol ai is understood to represent the three components al,~, aa. Also, a; is sometimes interpreted to represent the ith component of the vector or indeed to represent the vector itself. For a range of three on both indices, the symbol Aij represents nine components (of the second-order tensor (dyadic) A). The tensor Ai; is often presented explicitly by giving the nine components in a square array enclosed by large parentheses as


A21 Aal

A12 A22 A32








In the same way, the components of a first-order tensor (vector) in three-space may be displayed explicitly by a row or column arrangement of the form



In general, for a range of N, an nth order tensor will have N» components.

The usefulness of the indicial notation in presenting systems of equations in compact form is illustrated by the following two typical examples. For a range of three on both i and j the indicial equation

represents in expanded form the three equations



For a range of two on i and j, the indicial equation ' Aij = BipCjqDpq represents, in expanded form, the four equations

All e.c;»; + BllC1ZD12 + B1ZCllD21 + B1ZC12D22


BllC21Dll + BllC22D12 + B12C21D21 + B12C22D2Z B21CllDll + BZ1C12D1Z + B22CllDz1 + B22C1ZD2Z


Azz B21C21Dll + BZ1C22D12 + BZ2CZ1Dz1 + BZZC22Dz2

For a range of three on both i and i. (1.66) would represent nine equations, each having nine terms on the right-hand side.


The summation convention is very often employed in connection with the representation of vectors and tensors by indexed base vectors written in the symbolic notation. Thus if the rectangular Cartesian axes and unit base vectors of Fig. 1-5 are relabeled as shown by Fig. 1-8, the arbitrary vector v may be written


in which V1, V2, V3 are the rectangular Cartesian components of v. Applying the summation convention to (1.68), the equation may be written in the abbreviated form

A V = Viei


where i is a summed index. The notation here is essentially symbolic, but with the added feature of the summation convention. In such a "combination" style of notation, tensor character is not given by the free indices rule as it is in true indicial notation.

Fig. 1-8

CHAP. 1]



Second-order tensors may also be represented by summation on indexed base vectors.

Accordingly the dyad ab given in nonion form by (1.53) may be written


It is essential that the sequence of the base vectors be preserved in this expression. In similar fashion, the nonion form of the arbitrary dyadic D may be expressed in compact notation by



Let Xi represent the arbitrary system of coordinates Xl, X2, x3 in a three-dimensional Euclidean space, and let (}i represent any other coordinate system (}l, (}2, (}3 in the same space. Here the numerical superscripts are labels and not exponents. Powers of x may be expressed by use of parentheses as in (X)2 or (X)3. The letter superscripts are indices as already noted. The coordinate transformation equations


assign to any point (xl, X2, x3) in the Xi system a new set of coordinates ((}l, (}2, (}3) in the Oi system. The functions (}i relating the two sets of variables (coordinates) are assumed to be single-valued, continuous, differentiable functions. The determinant

a(}l aOl aOl[
axl ax2 ax3
J a02 a02 a02 (1.73)
axl ax2 ax3
a(}3 a(}3 a(}3
axl ax2 ax3 or, in compact form,


I a(}i·i



is called the Jacobian of the transformation. If the Jacobian does not vanish, (1.72) possesses a unique inverse set of the form


The coordinate systems represented by Xi and Oi in (1.72) and (1.75) are completely general and may be any curvilinear or Cartesian systems.

From (1.72), the differential vector d(}i is given by


This equation is a prototype of the equation which defines the class of tensors known as contravariant vectors. In general, a set of quantities bi associated with a point P are said to be the components of a contravariant tensor of order one if they transform, under a coordinate transformation, according to the equation


where the partial derivatives are evaluated at P. In (1.77), b' are the components of the tensor in the Xi coordinate system, while b" are the components in the (}i system. In general



[CHAP. 1

tensor theory, contravariant tensors are recognized by the use of superscripts as indices. It is for this reason that the coordinates are labeled Xi here rather than Xi, but it must be noted that it is only the differentials dx', and not the coordinates themselves, which have tensor character.

By a logical extension of the tensor concept expressed in (1.77), the definition of contravariant tensors of order two requires the tensor components to obey the transformation law


Contravariant tensors of third, fourth and higher orders are defined in a similar manner.

The word contravariant is used above to distinguish that class of tensors from the class known as covariant tensors. In general tensor theory, covariant tensors are recognized by the use of subscripts as indices. The prototype of the covariant vector is the partial . derivative of a scalar function of the coordinates. Thus if 1> = 1>(xl, x2, x3) is such a function,

a1> axJ axJ a()i


In general, a set of quantities b, are said to be the components of a covariant, tensor of order one if they transform according to the equation


In (1.80), If. are the covariant components in the ()i system, b, the components in the Xi system. Second-order covariant tensors obey the transformation law



Covariant tensors of higher order and mixed tensors, such as



are defined in the obvious way.


Let Xi represent a system of rectangular Cartesian coordinates in a Euclidean threespace, and let ()i represent any system of rectangular or curvilinear coordinates (e.g. cylindricalor spherical coordinates) in the same space. The vector x having Cartesian components Xi is called the position vector of the arbitrary point Ptx', X2, x3) referred to the rectangular Cartesian axes. The square of the differential element of distance between neighboring points P(x) and Q(x + dx) is given by



From the coordinate transformation

Xi = Xi( ()t, ()2, ()3) relating the systems, the distance differential is

axi d()p a()p



CHAP. 1]



and therefore (1.83) becomes



where the second-order tensor gpq = (axi/a(Jp)(axi/a(Jq) is called the metric tensor, or fundamental tensor of the space. If (Ji represents a rectangular Cartesian system, say the X'i system, then


where Spq is the Kronecker delta (see Section 1.13) defined by Spq = 0 if p ~ q and Spq = 1 if p = q.

Any system of coordinates for which the squared differential element of distance takes the form of (1.83) is called a system of homogeneous coordinates. Coordinate transformations between systems of homogeneous coordinates are orthogonal transformations, and when attention is restricted to such transformations, the tensors so defined are called Cartesiam tensors. In particular, this is the case for transformation laws between systems of rectangular Cartesian coordinates with a common origin. For Cartesian tensors there is no distinction between contravariant and covariant components and therefore it is customary to use SUbscripts exclusively in expressions representing Cartesian tensors. As will be shown next, in the transformation laws defining Cartesian tensors, the partial derivatives appearing in general tensor definitions, such as (1.80) and (1.81), are replaced by constants.



Let the axes OXIX2X3 and Ox~x~x~ represent two rectangular Cartesian coordinate systems with a common origin at an arbitrary point 0 as shown in Fig. 1-9. The primed system may be imagined to be obtained from the unprimed by a rotation of the axes about the origin, or by a reflection of axes in one of the coordinate planes, or by a combination of these. If the symbol aij denotes the cosine of the angle between the ith primed and jth unprimed coordinate axes, i.e. tu, = cos (x[, Xj), the relative orientation of the individual axes of each system with respect to the other is conveniently given by the table

Xl X2 X3
Xl all al2 al3
X2 a21 a22 a23
x£ a3l a32 a33 or alternatively by the transformation tensor


Fig. 1-9



[CHAP. 1

From this definition of aij, the unit vector e; along the x~ axis is given according to (1.48) and the summation convention by


An obvious generalization of this equation gives the arbitrary unit base vector e; as


In component form, the arbitrary vector v shown in Fig. 1-9 may be expressed in the unprimed system by the equation


and in the primed system by


Replacing e; in (1.91) by its equivalent form (1.89) yields the result


Comparing (1.92) with (1.90) reveals that the vector components in the primed and unprimed systems are related by the equations


The expression (1.93) is the transformation law for first-order Cartesian tensors, and as such is seen to be a special case of the general form of first-order tensor transformations, expressed by (1.80) and (1.77). By interchanging the roles of the primed and un primed base vectors in the above development, the inverse of (1.93) is found to be


Vi = aijVj


It is important to note that in (1.93) the free index on aij appears as the second index. In (1.94), however, the free index appears as the first index.

By an appropriate choice of dummy indices, (1.93) and (1.94) may be combined to produce the equation


Since the vector v is arbitrary, (1.95) must reduce to the identity v, = Vj. Therefore the coefficient tu.a«, whose value depends upon the subscripts j and k, must equal 1 or 0 according to whether the numerical values of j and k are the same or different. The Kronecker delta, defined by


for i = j for i # j


may be used to represent quantities such as aijaik. Thus with the help of the Kronecker delta the conditions on the coefficient in (1.95) may be written


In expanded form, (1.97) consists of nine equations which are known as the orthogonality or orthonormality conditions on the direction cosines aij. Finally, (1.93) and (1.94) may also be combined to produce Vi = aijakjV~ from which the orthogonality conditions appear in the alternative form


A linear transformation such as (1.93) or (1.94), whose coefficients satisfy (1.97) or (1.98), is said to be an orthogonal transformation. Coordinate axes rotations and reflections of the axes in a coordinate plane both lead to orthogonal transformations.

CHAP. 1]



The Kronecker delta is sometimes called the substitution operator, since, for example,

(1.99) (1.100)

and, likewise,

It is clear from this property that the Kronecker delta is the indicial counterpart to the symbolic idemfactor I, which is given by (1.54).

According to the transformation law (1.94), the dyad U;Vj has components in the primed coordinate system given by


In an obvious generalization of (1.101), any second-order Cartesian tensor Tij obeys the transformation law


With the help of the orthogonality conditions it is a simple calculation to invert (1.102), thereby giving the transformation rule from primed components to unprimed components:


The transformation laws for first and second-order Cartesian tensors generalize for an Nth order Cartesian tensor to


1.14 ADDITION OF CARTESIAN TENSORS. MULTIPLICATION BY A SCALAR Cartesian tensors of the same order may be added (or subtracted) component by component in accordance with the rule

AUk ... ± Bijk ... = Tijk ...


The sum is a tensor of the same order as those added. Note that like indices appear in the same sequence in each term.

Multiplication of every component of a tensor by a given scalar produces a new tensor of the same order. For the scalar multiplier A, typical examples written in both indicial and symbolic notation are

b, = sa, or b = Aa

Bii = AAu or B = AA

(1.106) (1.107)


The outer product of two tensors of arbitrary order is the tensor whose components . are formed by multiplying each component of one of the tensors by every component of the other. This process produces a tensor having an order which is the sum of the orders of the factor tensors. Typical examples of outer products are

(a) tub, = Ti, (b) viFjk = ltijk

(c) DijTkm = <Pijkm (d) €ijkVm = 0ijkm

As indicated by the above examples, outer products are formed by simply setting down the factor tensors in juxtaposition. (Note that a dyad is formed from two vectors by this very procedure.)



[CHAP. 1

Contraction of a tensor with respect to two free indices is the operation of assigning to both indices the same letter subscript, thereby changing these indices to dummy indices. Contraction produces a tensor having an order two less than the original. Typical examples of contraction are the following.

(a) Contractions of Tij and UiVj

r; Tll + T22 + T33

(b) Contractions of Eijak

Eijaj bi
Eijai Cj
Eiiak dk (c) Contractions of EijFkm
EijFim Gjm e.r.. Pij
EijFki Hjk EijFjm o:
e.r«; «.; EijFkj ti; An inner product of two tensors is the result of a contraction, involving one index from each tensor, performed on the outer product of the two tensors. Several inner products important to continuum mechanics are listed here for reference, in both the indicial and symbolic notations.

Outer Prod uct

Inner Product

Indicial Notation

Symbolic Notation a·b

1. aibj

2. aiEjk

aiEik h
aiEji hj
EijFjm c;
EijEjm e.; a·E


E·a h

3. EijFkm

4. EijEkm


G (E)2

Multiple contractions of fourth-order and higher tensors are sometimes useful. Two such examples are

1. EijFkm contracted to EijFij, or E: F

2. EijEkmEpq contracted to EijEjmEmq, or (E)3


In order to express the cross product a X b in the indicial notation, the third-order tensor £ijk' known as the permutation symbol or alternating tensor, must be introduced. This useful tensor is defined by

1 if the values of i, i. k are an even permutation of 1,2,3 (i.e. if they appear in sequence as in the arrangement 1 2 3 1 2).

-1 if the values of i, i. k are an odd permutation of 1,2,3 (i.e. if

£;jk they appear in sequence as in the arrangement 32132).

o if the values of i, j, k are not a permutation of 1,2,3 (i.e. if two or more of the indices have the same value).

CHAP. 1]



From this definition, the cross product a x b = c is written in indicial notation by


Using this relationship, the box product a x b· c =,\ may be written


Since the same box product is given in the form of a determinant by (1.52), it is not surprising that the permutation symbol is frequently used to express the value of a 3 x 3 determinant.

It is worthwhile to note that €..k obeys the tensor transformation law for third order


Cartesian tensors as long as the transformation is a proper one (det c., = 1) such as arises

from a rotation of axes. If the transformation is improper (det c., = -1), e.g. a reflection in one of the coordinate planes whereby a right-handed coordinate system is transformed into a left-handed one, a minus sign must be inserted into the transformation law for Eijk• Such tensors are called pseudo-tensors.

The dual vector of an arbitrary second-order Cartesian tensor Tij is defined by


which is observed to be the indicial equivalent of Tv, the "vector of the dyadic T", as defined by (1.15).


A rectangular array of elements, enclosed by square brackets and subject to certain laws of combination, is called a matrix. An M x N matrix is one having M (horizontal) rows and N (vertical) columns of elements. In the symbol Aij, used to represent the typical element of a matrix, the first subscript denotes the row, the second subscript the column occupied by the element. The matrix itself is designated by enclosing the typical element symbol in square brackets, or alternatively, by the kernel letter of the matrix in script. For example, the M x N matrix cA, or [Aij] is the array given by


A matrix for which M = N, is called a square matrix. A 1 x N matrix, written [alk], is called a row matrix. An M x 1 matrix, written [akl], is called a column matrix. A matrix having only zeros as elements is called the zero matrix. A square matrix with zeros everywhere except on the main diagonal (from All to ANN) is called a diagonal matrix. If the nonzero elements of a diagonal matrix are all unity, the matrix is called the unit or identity matrix. The N x M matrix cAT, formed by interchanging rows and columns of the M x N matrix cA, is called the transpose matrix of cA.

Matrices having the same number of rows and columns may be added (or subtracted) element by element. Multiplication of the matrix [Aij] by a scalar ,\ results in the matrix ['\Au]. The product of two matrices, cACf3, is defined only if the matrices are conformable, i.e. if the prefactor matrix cA has the same number of columns as the postfactor matrix Cf3 has rows. The product of an M x P matrix multiplied into a P x N matrix is an M x N matrix. Matrix multiplication is usually denoted by simply setting down the matrix symbols in juxtaposition as in




[CHAP. 1

Matrix multiplication is not, in general, commutative: eA'R # 'ReA.

A square matrix eA whose determinant IAiil is zero is called a singular matrix. The cofactor of the element Aii of the square matrix eA, denoted here by Aii, is defined by


in which Mii is the minor of Aii; i.e. the determinant of the square array remaining after the row and column of Aii are deleted. The adjoint matrix of eA is obtained by replacing each element by its cofactor and then interchanging rows and columns. If a square matrix eA = [Aii] is non-singular, it possesses a unique inverse matrix eA-1 which is defined as the adjoint matrix of eA divided by the determinant of eA. Thus






From the inverse matrix definition (1.114) it may be shown that eA-1eA = eAeA-1 = !J


where !J is the identity matrix, having ones on the principal diagonal and zeros elsewhere, and so named because of the property

!JeA = eA!J = eA


It is clear, of course, that !J is the matrix representation of 8ii, the Kronecker delta, and of I, the unit dyadic. Any matrix eA for which the condition eAT = eA-1 is satisfied is called an orthogonal matrix. Accordingly, if eA is orthogonal,


As suggested by the fact that any dyadic may be expressed in the nonion form (1.53), and, equivalently, since the components of a second-order tensor may be displayed in the square array (1.62), it proves extremely useful to represent second-order tensors (dyadics) by square, 3 x 3 matrices. A first-order tensor (vector) may be represented by either a 1 x 3 row matrix, or by a 3 x 1 column matrix. Although every Cartesian tensor of order two or less (dyadics, vectors, scalars) may be represented by a matrix, not every matrix represents a tensor.

If both matrices in the product eA'R = Care 3 x 3 matrices representing second-order tensors, the multiplication is equivalent to the inner product expressed in indicial notation by


where the range is three. Expansion of (1.118) duplicates the "row by column" multiplication of matrices wherein the elements of the ith row of the prefactor matrix are multiplied in turn by the elements of the kth column of the postfactor matrix, and these products summed to give the element in the ith row and kth column of the product matrix. Several such products occur repeatedly in continuum mechanics and are recorded here in the various notations for reference and comparison.

(a) Vector dot product aob

[ali] [bid = [A]

[a,a,a{~:J ~ [a,b, + a,b,+a,b,]


CHAP. 1]



(b) Vector-dyadic dot product

a·E = b

(c) Dyadic-vector dot product

E'a = c

ac = CJ3

[alEll + a2E2l + a3E31, alE 12 + a2E 22 + a3E 32, alE13 + a2E23 + a3E33]


ca = c



According to (1.36) (or (1.37)), a dyadic D is said to be symmetric (anti-symmetric) if it is equal to (the negative of) its conjugate Dc. Similarly the second-order tensor Dij is symmetric if

and is anti-symmetric, or skew-symmetric, if



Therefore the decomposition of Dij analogous to (1.38) is

Dij = t(Dij + Dji) + t(Dij - Dji) or, in an equivalent abbreviated form often employed,



where parentheses around the indices denote the symmetric part of Dij, and square brackets on the indices denote the anti-symmetric part.

Since the interchange of indices of a second-order tensor is equivalent to the interchange of rows and columns in its matrix representation, a square matrix cA is symmetric if it is equal to its transpose cAT. Consequently a symmetric 3 x 3 matrix has only six independent components as illustrated by



A12 A13


An anti-symmetric matrix is one that equals the negative of its transpose. Consequently a 3 x 3 anti-symmetric matrix CJ3 has zeros on the main diagonal, and therefore only three independent components as illustrated by






[CHAP. 1

Symmetry properties may be extended to tensors of higher order than two. In general, an arbitrary tensor is said to be symmetric with respect to a pair of indices if the value of the typical component is unchanged by interchanging these two indices. A tensor is antisymmetric in a pair of indices if an interchange of these indices leads to a change of sign without a change of absolute value in the component. Examples of symmetry properties in higher-order tensors are

(a) Riikm = Rikim (symmetric in k and j)

(b) €iik = -€kji (anti-symmetric in k and i)

(c) Giikm = Giimk (symmetric in i and j; k and m)

(d) {3iik = {3iki = {3kji = {3jik (symmetric in all indices)


In the following analysis, only symmetric tensors with real components are considered.

This simplifies the mathematics somewhat, and since the important tensors of continuum mechanics are usually symmetric there is little sacrifice in this restriction.

For every symmetric tensor Tii, defined at some point in space, there is associated with each direction (specified by the unit normal ni) at that point, a vector given by the inner product


Here Tii may be envisioned as a linear vector operator which produces the vector Vi conjugate to the direction ni. If the direction is one for which Vi is parallel to n;, the inner product may be expressed as a scalar multiple of ni. For this case,


and the direction n; is called a principal direction, or principal axis of Tii• With the help of the identity ni = 8iinj, (1.129) can be put in the form


which represents a system of three equations for the four unknowns, n; and A, associated with each principal direction. In expanded form, the system to be solved is

(Tu - A)nl + Tl2n2 + Tlana 0



Tal nl + Taznz + (Taa - A)na 0

Note first that for every A, the trivial solution n; = 0 satisfies the equations. The purpose here, however, is to obtain non-trivial solutions. Also, from the homogeneity of the system (1.131) it follows that no loss of generality is incurred by restricting attention to solutions for which n;n; = 1, and this condition is imposed from now on.

For (1.130) or, equivalently, (1.131) to have a non-trivial solution, the determinant of coefficients must be zero, that is,

ITii - A8ijl = 0

Expansion of this determinant leads to a cubic polynomial in A, namely, Aa - ITAz + IITA - Ifl- = 0



which is known as the characteristic equation of Tii, and for which the scalar coefficients,

CHAP. 1]



Il l- = ITijl = det Tij

(1.134) (1.135) (1.136)

IT = T« = tr Tij (tr31ce of Tij) IIr = -HTiiTjj - TijTij)

are called the first, second and third invariants, respectively, of Tij. The three roots of the cubic (1.133), labeled A(1), A(2), A(3), are called the principal values of Tij. For a symmetric tensor with real components, the principal values are real; and if these values are distinct, the three principal directions are mutually orthogonal. When referred to principal axes, both the tensor array and its matrix appear in diagonal form. Thus






If A(1) = A(2), the tensor has a diagonal form which is independent of the choice of A(1) and A(2) axes, once the principal axis associated with A(3) has been established. If all principal values are equal, any direction is a principal direction. If the principal values are ordered, it is customary to write them as Am, A (II) ,A(III) and to display the ordering as in A (I) > A(II) > A (III) •

For principal axes labeled Oxix~x~, the transformation from OXlX2X3 axes is given by the elements of the table

Xl X2 X3
X* all = nil) al2 = n~l) al3 = n~l)
x* a2l = ni2) a22 = n~2) a23 = n~2)
X* a3l = ni3) a32 = n~3) a33 = n~3)
3 in which n~j) are the direction cosines of the jth principal direction.

1.20 POWERS OF SECOND-ORDER TENSORS. HAMILTON-CAYLEY EQUATION By direct matrix multiplication, the square of the tensor Tij is given as the inner product Tik Tkj; the cube as Tik Tkm T mj; etc. Therefore with Tij written in the diagonal form (1.137), the nth power of the tensor is given by






A comparison of (1.138) and (1.137) indicates that Tij and all its integer powers have the same principal axes.

Since each of the principal values satisfies (1.133), and because of the diagonal matrix form of 'P given by (1.138), the tensor itself will satisfy (1.133). Thus


in which /J is the identity matrix. This equation is called the Hamilton-Cayley equation. Matrix multiplication of each term in (1.139) by T produces the equation,




[CHAP. 1

Combining (1.140) and (1.139) by direct substitution,

']'4 = (I~ - IIT}'T2 + (H l- - ITIIT}'T + ITIIIrc1


Continuation of this procedure yields the positive powers of 'T as linear combinations of 'T2, 'T and !J.


A tensor field assigns a tensor T(x, t) to every pair (x, t) where the position vector x varies over a particular region of space and t varies over a particular interval of time. The tensor field is said to be continuous (or differentiable) if the components of T(x, t} are continuous (or differentiable) functions of x and t. If the components are functions of x only, the tensor field is said to be steady.

With respect to a rectangular Cartesian coordinate system, for which the position vector of an arbitrary point is x = xiei (1.142)

tensor fields of various orders are represented in indicial and symbolic notation as follows,

(a) scalar field: cP cp(Xi, t} or cp cp(x, t} (1.143)

(b) vector field: Vi Vi(X,t} or v = v(x,t} (1.144)

(c) second-order tensor field:

Ti, = Tij(x, t} or T = T(x, t}


Coordinate differentiation of tensor components with respect to Xi is expressed by the differential operator a/aXi, or briefly in indicial form by ai, indicating an operator of tensor rank one. In symbolic notation, the corresponding symbol is the well-known differential vector operator V, pronounced del and written explicitly

'" a er r+



Frequently, partial differentiation with respect to the variable Xi is represented by the comma-subscript convention as illustrated by the following examples.

(a) acp CP,i (d) a2Vi
= Vi,Jk
aXi aXjaXk
(b) aVi (e) aTij Tij,k
Vi,i =
aXi aXk aVi

(c) a = v ..

Xj t,)

From these examples it is seen that the operator ai produces a tensor of order one higher if i remains a free index ((a) and (c) above}, and a tensor of order one lower if i becomes a dummy index ((b) above} in the derivative.

Several important differential operators appear often in continuum mechanics and are given here for reference.

grad e = VCP acp '" or a,cp = Cp" (1.147)
divv = V·v or a,v, = v", (1.148)
curl v = V Xv or (ljk aj vk = (ljk vk,J (1.149)
\72cp = V·Vcp or aliCP = CP,I' (1.150) CHAP. 1]




In a given region of space the vector function of position, F = F(x), is defined at every point of the piecewise smooth curve C shown in Fig. 1-10. If the differential tangent vector to the curve at the arbitrary point P is dx, the integral

iF, dx - i:B F· dx

taken along the curve from A to B is known as the line integral of F along C. In the indiciaI notation, (1.151) becomes





Stokes' theorem says that the line integral of F taken around a closedreducible curve C, as pictured in Fig. 1-11, may be expressed in terms of an integral over any two-sided surface S which has C as its boundary. Explicitly,

f F·dx Is n'("\7 xF)dS (1.153)

in which n is the unit normal on the positive side of S, and dS is the differential element of surface as shown by the figure. In the indicial notation, (1.153) is written

f r. dx, J: n_;€ijkFk,j dS (1.154)


The divergence theorem of Gauss relates a volume integral to a surface integral. In its traditional form the theorem says that for the vector field v = v(x),

Iv div v dV

Is n'v dS


where n is the outward unit normal to the bounding surface S, of the volume V in which the vector field is defined .. In the indicial notation, (1.155) is written

Iv Vi,i dV Is Vini dS (1.156)

The divergence theorem of Gauss as expressed by (1.156) may be generalized to incorporate a tensor field of any order. Thus for the arbitrary tensor field Tijk ... the theorem is written

i r.; .. ,p dV




[CHAP. 1

Solved Problems ALGEBRA OF VECTORS AND DYADICS (Sec. 1.1-1.8) 1.1. Determine in rectangular Cartesian form the unit vector which is (a) parallel to the vector


V = 2 i + 3 j - 6k, (b) along the line joining

points P(l, 0, 3) and Q(O, 2,1).

(a) Ivl = v = V(2)2 + (3)2 + (-6)2 = 7


1.2. Prove that the vector v = a i + b j + ck is normal

to the plane whose equation is ax + by + cz = ,\.

Let P(x" y" z,) and Q(X2' Y2, z2) be any two points in the plane. Then ax, + by, + cz, = h and aX2 + bY2 + CZ2 = h


and the vector joining these points is u = (X2 - x,) i +


(Y2 - y,) j + (Z2 - z,)k. The projection of v in the direction

of u is

u r v 1 A A

= - [(x2 - x,) i + (Y2 - y,) j

U U A A .At. A

+ (z2-z,)k] • [ai + bj + ck]

1 h - h x

= - (ax2 + bY2 + cZ2 - ax, - by, - cz,) 0

u u

Since u is any vector in the plane, v is .L to the plane.


V = v/v = (2/7) i + (3/7) j - (6/7)k

(b) The vector extending from P to Q is


U - (0 -1) i + (2 - 0) j + (1- 3)k


= -i + 2j - 2k

u -

V(-1)2 + (2)2 + (_2)2 = 3



U = -(1/3) i + (2/3) j - (2/3)k




(1/3) i - (2/3) j + (2/3)k



Fig. 1-12

directed from P to Q directed from Q to P



Fig. 1-13


1.3. If r = xi + y j + zk is the vector extending from the origin to the arbitrary point


P(x, y, z) and d = a i + b j + ck is a constant vector, show that (r - d) . r = ° is the

vector equation of a sphere.

Expanding the indicated dot product,

A A .At. A A A

(r-d)·r = [(x-a)i+(y-b)j+(z-c)k]·[xi+yj+zk]

= x2 + y2 + z2 - ax - by - cz = 0

Adding d2/4 = (a2 + b2 + c2)/4 to each side of this equation gives the desired equation (x - a/2)2 + (y - b/2)2 + (z - C/2)2 = (d/2)2

which is the equation of the sphere centered at d/2 with radius d/2.

1.4. Prove that [a· b X c]r = (a· r)b x c + (b· r)c X a + (c· r)a X b.

Consider the product a X [(b X c) X r]. By direct expansion of the cross product in brackets, a X [(b X c) X r] = a X [(b· r)c - (c- r)bJ = -(b· r)c X a - (c· r)a X b

Also, setting b X c = v,

a X [(b X c) X r] = a X (v X r) = (a- r)b X c - (a· b X c)r

CHAP. 1]



Thus -(b· r)c X a - (c· r)a X b = (a· r)b X c - (a- b X c)r and so

(a· b X c)r = (a· r)b X c + (b· r)c X a + (c· r)a X b

This identity is useful in specifying the displacement of a rigid body in terms of three arbitrary points of the body.

1.5. Show that if the vectors a, band c are linearly dependent, a· b X c = O. Check the linear dependence or independence of the basis


u 3i + j - 2k


v 4i-j-k



The vectors a, band c are linearly dependent if there exist constants A, p. and v, not all zero,

such that xa + p.b + vC = O. The component scalar equations of this vector equation are

Aax + pbx + vCx 0

Aay + p.by + vCy 0

Aa. + p.b. + vc. 0

This set has a nonzero solution for A, p. and v provided the determinant of coefficients vanishes, ax i, Cx

ay by cy 0

az bz c.

which is equivalent to a- b X c = O. For the proposed basis u, v, w,

3 1-2

4 -1 -1 0

1 -2 1

Hence the vectors u, v, ware linearly dependent, and indeed v = u + w.

1.6. Show that any dyadic of N terms may be reduced to a dyadic of three terms in a form having the base vectors e" e2, ea as (a) antecedents, (b) consequents.

Let 0 = alb, + a2b2 + ... + aNbN = a.b, (i = 1,2, ... , N).

(a) In terms of base vectors, ai = ali~' + a2i~2 + aai~a = aji-ej and so 0 = aji~bi = ~j(ajibi) = -ejCj with j = 1,2,3.

(b) Likewise setting b, = bji~j it follows that 0 = aibji~j = (bjiai)~j = gj~j where j = 1,2,3.

1.7. For the arbitrary dyadic D and vector v, show that D· v = v· Dc.

Let 0 = alb, + a2b2 + ... + aNbN. Then

o·v a,(b,·v) + a2(b2·v) + + aN(bN·v)

(v· b,)a, + (v· b2)a2 + + (v· bN)aN v· Oc

1.8. Prove that (Dc· D)c = Dc· D.


0c·O == Dji~i~j·Dpqep~q == DjiDpq(~j·-ep)-ei-eq DjiDpq(~·~P)~Q~i = Dpq~q(~·~j)~iDji == Dpqeqep·Dji~j-ei



[CHAP. 1

1.9. Show that (D x v), = -v x n,

o x v = al(bl X v) + a2(b2 X v) + ... + aN(bN X v)

(bl x v)al + (b2 x v)a2 + + (bN x v)aN

= -(v X bl)a! - (v X bz)az - - (v X bN)aN -v x Dc


1.10. If D = a i i + b j j + ckk and r is the position vector r xi + y j + zk, show

that r : D • r = 1 represents the ellipsoid ax2 + by2 + cz2 = 1.


(xi + yj + zk) • (ai i + bj j + ck k j > (xi + yj + zk)


= (x i + y j + z k) • (ax i + by j + czk) = ax2 + by2 + cz2 = 1


1.11. For the dyadics D = 3 i i + 2 j j - j k + 5kk and F = 4 i k + 6 j j - 3kj + kk, com-

pute and compare the double dot products D: F and D" F.

From the definition ab: cd = (a· c)(b • d) it is seen that 0: F = 12 + 5 = 17. Also, from ab- • cd = (b- c)(a • d) it follows that D" F = 12 + 3 + 5 = 20.

1.12. Determine the dyadics G = D· F and H = F' D if D and F are the dyadics given in Problem 1.11.

From the definition ab - cd = (b- clad,


G (3ii+2jj-jk+5kk)·(4ik+6jj-3kj+kk)


12 i k + 12 j j + 3 j j - j k - 15 k j + 5k k




(4 i k + 6 j j - 3 k j + k k) • (3 i i + 2 j j - j k + 5 k k)


20 i k + 12 j j - 6 j k - 6 k j + 8 k k

1.13. Show directly from the non ion form of the dyadic D that D


(D' k)k and also i· D' i = Dxx, i' D' j = Dxy, etc.

Writing 0 in non ion form and regrouping terms,


o (Dxxi + Dyxj + Dzxk) i + (Dxy i + Dyyj + Dzyk) j + (Dxzi + Dyzj + Dzzk)k


(D, i) i + (D· j) j +

Also now


j • D' i = Dyx,


j • d2 = j. 0 • j = Dyy, etc.

1.14. For an antisymmetric dyadic A and the arbitrary vector b, show that 2b· A = Av x b. From Problem 1.6(a), A = ~ICI + ~2C2 + ~aca; and because it is antisymmetric, 2A = (A - Ac)

2A (~ICI + ~2C2 + ~aca - CI~1 - C(e2 - ca~)


and so


[(b' ~!)CI - (b· Cl) ~d + [(b, ~2)C2 - (b > C2) ~2l + [(b' ~a)ca - (b > ca) ~al [(~IXCI)Xb+(~2XC2)Xb+(~aXca)Xbl = (AvXb)

1.15. If D = 6if + 3ij + 4kk and u = 2i + k, v = 5j, show by direct calculation that D' (u x v) = (D x u) . v.


Since u X v = (2 i + k) x 5 j


10k - 5 i ,


D' (u x v) = (6i i + 3i j + 4kk)· (-5i + 10k)


-30 i + 40k


Next, DXu = (6ii+3ij+4kk)X(2i+k) = -6ik+8kj-6ij+3ii


and (0 x u) • v = (3 i i - 6 i j - 6 i k + 8 k j ) • 5 j = -30 i + 40k

CHAP. 1]



1.16. Considering the dyadic


D = 3 i i - 4 i j + 2 j i + j j + kk

as a linear vector operator, determine the vector r' produced when D operates on

r = 4 i + 2 j + 5 k.

r' o· r


12 i + 8 j - 8 i + 2 j + 5k

'" '" '"

4 i + 10 j + 5k

Fig. 1-14

1.17. Determine the dyadic D which serves as a linear vector operator for the vector

'" '" '"

function a = f(b) = b + b X r where r = xi + y j + zk and b is a constant vector.

In accordance with (1.59) and (1.60), construct the vectors

'" '" '" '" '" '"
u - f( i) i + i X r i - z j + yk
'" '" '" '" '" '"
v f( j ) j + j X r z i + j - xk
'" '" '" '" '" '"
w f(k) k + k X r -y i + x j + k Then



U i + v j + w k = (i - z j + y k) i + (z i + j - x k ) j + (-y i + x j + k) k


As a check the same result may be obtained by direct expansion of the vector function,

1.18. Express the unit triad ~<jJ' ~o' ~r in terms of


i, j, k and confirm that the curvilinear triad

is right-handed by showing that ~<jJ X ~o = ~r'

By direct projection from Fig. 1-15,

A '" A

(cos ¢ cos 0) i + (cos ¢ sin 0) j - (sin ¢) k

A '"

(-sino)i + (coso)j


(sin ¢ cos 0) i + (sin ¢ sin 0) j + (cos ¢)k


Fig. 1-15
i j k
'" '" cos ¢ sin 0 -sin¢
e<jJ X eo cos ¢ cos 0
- sin 0 cos 0 0
(sin ¢ cos 0) i + (sin ¢ sin 0) j + [(cos20 + sin? 0) cos ¢] k e, and so


1.19. Resolve the dyadic D = 3 i i + 4 i k + 6 j i + 7 j j + 10 k i + 2 k j into its symmetric

and antisymmetric parts.

Let 0 = E + F where E = r, and F = -F c' Then


E = (1/2)(0 + Dc) = (1/2)(6 i i + 4 i k + 4 k i + 6 j i + 6 i j + 14 j j


+ 10 k i + 10 i k + 2 k j + 2 j k)


(1/2)(0 - Dc)


3 i i + 3 i j + 7 i k + 3 j i + 7 j j + j k + 7 k i + k j = e,


(1/2)(4 i k - 4 k i + 6 j i - 6 i j + 10 k i - 10 i k + 2 k j - 2 j k)


-3 i j - 3 i k + 3 j i - j k + 3 k i + k j =-Fc



[CHAP. 1

1.20. With respect to the set of base vectors a!, a2, a3 (not necessarily unit vectors), the set at, a2, a3 is said to be a reciprocal basis if ar- a' = 8;;. Determine the necessary relationships for constructing the reciprocal base vectors and carry out the calculations for the basis

A A A,IIt.. A AA A

hi = 3 i + 4 j, h2 = - i + 2 j + 2k, ba = i + j + k

By definition, al· a1 = 1, a2• a1 = 0, a3· a1 = O. Hence a1 is perpendicular to both a2 and a3' Therefore it is parallel to a2 X a3, i.e. a1 = A(a2 X a3)' Since al· a1 = 1, al· Aa2 X a3 = 1 and 1.= 1/(a1 • a2 X a3) = 1/[ala2a3]' Thus, in general,

For the basis b., b2, b3, 1/1. = b1 • b2 X b3 = 12 and so

(b2 X b3)/12 (b3 X b1)/12 (b1 X b2)/12


(j -k)/4


- i/3 + j /4 + k/12


2 i/3 - j/2 + 5k/6


1.21. For a range of three on the indices, give the meaning of the following Cartesian tensor symbols: Aii, Bijj, Rij, ai Tij, a.bS«:

Aii represents the single sum Aii = All + A2Z + A33•

Bijj represents three sums: (1) For i = 1, Blll + B122 + B133• (2) For i = 2, B2ll + B222 + B233• (3) For i = 3, B3ll + B322 + B333.

Rij represents the nine components Rll, R12, R13, R2l> R22, R23, R3b R32, R33• aiTij represents three sums: (1) For j = 1, a1Tll + a2T2i + a3T31'

(2) For j = 2, a1 T12 + a2 T22 + a3 T3Z'

(3) For j = 3, alT13 + aZT23 + a3T33'

aibjSij represents a single sum of nine terms. Summing first on i, aibjSij = albjSlj + aZbjS2j + a3bjSaj. Now summing each of these three terms on j,

a;bjSij = a1b1Sll + a1b2S12 + a1b3S13 + a2b1S21 + a2b2S22 + a2b3S23 + a3blS31 + a3b2S32 + a3b3S33

1.22. Evaluate the following expressions involving the Kronecker delta 8ij for a range of three on the indices.

(a) Ilii = Illl + 1122 + 1133 = 3

(b) Ilijllij = 1l1jll1j + 1l2jll2j + 1l3jll3j = 3

(c) Ilijlliklljk = Illjlllklljk + 1l2jllZklljk + 1l3jll3klljk 3

(d) Ilijlljk = Ilililik + lli21l2k + lli31l3k = Ilik

(e) lli;Aik = 1l1;Alk + 1l2jA2k + 1l3jA3k = Ajk

1.23. For the permutation symbol £ijk show by direct expansion that (a) £ijk£kij 6,

(b) £ijkajak = O.

(a) First sum on i,

CHAP. 1]



Next sum on i- The nonzero terms are

fijkfkij = f12kfk12 + f1akfk1a + f21kfk21 + f2akfk2a + fa1kfkat + fa2kfka2 Finally summing on k, the nonzero terms are

= (1)(1) + (-1)(-1) + (-1)(-1) + (1)(1) + (1)(1) + (-1)(-1) 6

(b) Summing on j and k in turn,

filkatak + fi2ka2ak + fiakaaak

fi12ata2 + fi1aa1aa + fi21a2a1 + fi2aa2aa + fia1aaal + fia2aaa2

From this expression,
when i = 1, f1jkajak a2aa - aaa2 0
when i = 2, f2jkajak a1aa - aaal 0
when i = 3, fajkajak = a1a2 - a2a1 0 Note that fijkajak is the indicial form of the vector a crossed into itself, and so a X a = O.

1.24. Determine the component h for the vector expressions given below. (a) t, fijkTjk

12 f2jkTjk = f21aT1a + f2a1Ta1 -T1a + Tal

(b) t, 12

Ci,jbj - Cj,ibj

C2,lb1 + C2.2b2 + C2.aba - C1.2b1 - C2.2b2 - Ca.2ba (C2.1 - cl,2)b1 + (C2.a - ca.2)ba


B2d: + B2zf; + B2a/:

(c) t, 12

1.25. Expand and simplify where possible the expression DijXiXj for (a) Di, D.«,

(b) Du = -Dji•

D1jX1Xj + D2jX2Xj + Dajxaxj

D11X1x1 + D12X1X2 + D1aX1Xa + D21x2x1 + D22X2X2 + D2aX2Xa + Da1XaX1 + Da2Xaxz + Daaxaxa

(a) DijXiXj (b) DijXiXj

DU(X1)2 + D22(X2)2 + Daa(xa)2 + 2D12X1X2 + 2D2aX2X3 + 2D13X1X3 o since D11 = -Du, D12 = -D2l> etc.


Show that i = p = 2. indices.)

(a) Introduce i = 1, j = 2, p = 3, q = 2 and note that since k is a summed index it takes on

all values. Then

£ijk£kpq = 8ip8jq - 8iq8jp for (a) i = 1, j = q = 2, p = 3 and for (b) i = q = 1, (It is shown in Problem 1.59 that this identity holds for every choice of


(b) Introduce i = 1, j = 2, p = 2, q = 1. Then fijkfkpq = f123f321 = -1 and 8ip8jq - 8iq8jp = 812821 - 811822 = -1.

1.27. Show that the tensor Bi; = £ijkaj is skew-symmetric.

Since by definition of fijk an interchange of two indices causes a sign change, Bik = fijkaj = -(fkjiaj) = -(Bki) = -Bki



[CHAP. 1

1.28. If Bij is a skew-symmetric Cartesian tensor for which the vector show that Bpq = £pqibi'



Multiply the given equation by €pqi and use the identity given in Problem 1.26.

€pq;bi = t€pqi€!jkBjk = t(8p/lqk - 8pk8qj)Bjk = t(Bpq - Bqp) = t(Bpq + Bpq) Bpq

1.29. Determine directly the components of the metric tensor for spherical polar coordinates as shown III Fig. 1-7(b).

ax! ax!

Write (1.87) as Upq = ao 7i9 and label the co-

p q

ordinates as shown in Fig. 1-16 (1' = 01> ¢ = 0z, ° = (3), Then

XI °1 sin liz cos 03

Xz III sin liz sin 113

Hence aXI
aliI sin liz cos 113
axz sin liz sin 113
aliI cos liz
from which ax! ax!
Un aliI aOI
ax! ax!
Uzz allz allz
ax! ax!
U33 =
all3 all3 X2

Fig. 1-16
aXI aXI -°1 sin 112 sin 113
allz III cos 112 cos 113 all3
aX2 III cos 112 sin 113 aX2 III sin liz cos 113
all2 all3
aX3 -111 sin 112 aX3 °
allz all3 sin? liz cos- 113 + sin? liz sinz 03 + cos? liz 1

0~COS2I1Z COSzII3 + lI~cos2l1z sinzll3 + lI~sin2 °2 = O~

Also, Upq = ° for p # q. For example,

ax; ax; UI2 = aliI allz

Thus for spherical coordinates, (ds)2


+ (sin liz sin 113)(111 cos °2 sin 113) - (cos 112)(111 sin 112)

1.30. Show that the length of the line element ds resulting from the curvilinear coordinate increment dB! is given by ds = ~ dBi (no sum). Apply this result to the spherical coordinate system of Problem 1.29.

Write (1.86) as (ds)2 = Upq dllp dllq- Thus for the line element (dill, 0, 0), it follows that (ds)Z = Ul1(dlll)Z

and ds = vu;;, dill' Similarly for (0, dllz, 0), ds = Vi;; d1l2; and for (0,0, dIl3), ds = ~ d1l3• Therefore (Fig. 1-17),

Fig. 1-17

CHAP. 1]



(1) For (dOl' 0, 0), (2) For (0, d02, 0), (3) For (0,0, doa),

ds ds

dOl = dr

01 d02 = r d</>

01 sin O2 dOa = r sin </> do


1.31. If the angle between the line elements represented by (dOl, 0, 0) and (0, d02, 0) is denoted by f312, show that cos f312 = ~~

a« g22

Let dSl = vu;;_ dOl be the length of line element represented by (dOl' 0, 0) and dS2 = ...;g;; doz ax!

be that of (0,d02,0). Write (1.85) as dXi =~dOk' and since (ds)2 = COS/312dslds2' vOk


dXi dXi = 8ii dXi dXj

aXl aXl aX2 aX2 aXa aXa

;- - del d02 + ;- ;- dOl d02 + ;- ;- dOl d02

vOl a02 vOl u02 vOl v02

dOl d02 U12

Hence using the result of Problem 1.30, cos /312 = U12 d d = ----==--==

Sl S2 vu;;_ y-g;

1.32. A primed set of Cartesian axes Ox! x~ x; is obtained by a rotation through an angle o about the Xa axis. Determine the transformation coefficients au relating the axes, and give the primed components of the vector v = Vlel + V2e2 + vaea.

From the definition (see Section 1.13) aii =

cos (xi' Xi) and Fig. 1-18, the table of direction cosines is

Xl X2 Xa
xi cos 0 sin 0 0
x~ -sino cos II 0
xa 0 0 1 Thus the transformation tensor is


By the transformation law for vectors (1.94),

vi alivi
v~ a2jVj
v~ aajVj 1.33. The table of direction cosines relating two sets of rectangular Cartesian axes is partially given as shown on the right. Determine the entries for the bottom row of the table so that Ox!x~x; is a right-handed system.

Fig. 1-18

-VI sin 0 + V2 cos 0

Xl X2 Xa
X' 3/5 -4/5 0
X~ 0 0 1
Xa 32


[CHAP. 1

The unit vector ~; along the x~ axis is given by the first row of the table as ~~ = (3/5) ~l - (4/5) ~2.

Also from the table ~~ = ~3. For a right-handed primed system ~; = ~; X ~;, or ~~ =

X~ I -4/5 I -3/5

[(3/5) ~l - (4/5) ~21 X ~3 = (-3/5) ~2 - (4/5) ~l and the third row is

1.34. Let the angles between the primed and un primed coordinate directions be given by the table shown on the right. Determine the transformation coefficients aij and show that the orthogonality conditions are satisfied.

The coefficients aij are direction cosines and may be calculated directly from the table. Thus




Xl X2 X3
x~ 135° 60° 120°
X2 90° 45° 45°
x' 45° 60° 120°
3 1/2 1/-/2 1/2

-1/2 )



The orthogonality conditions aijaik = 8jk require:

1. For j = k = 1 that allall + a2la21 + a3la31 = 1 which is seen to be the sum of squares of the elements in the first column.

2. For j = 2, k = 3 that al2al3 + a22a23 + a32a33 = 0 which is seen to be the sum of products of corresponding elements of the second and third columns.

3. Any two columns "multiplied together element by element and summed" to be zero. The sum of squares of elements of any column to be unity.

For orthogonality conditions in the form ajiaki = 8j", the rows are multiplied together instead of the columns. All of these conditions are satisfied by the above solution.

1.35. Show that the sum "-Aij + p,Bij represents the components of a second-order tensor if Aij and Bi, are known second-order tensors.

By (1.108) and the statement of the problem, Aij = apiaqjA~q and Bij = apiaqjB;q. Hence

which demonstrates that the sum transforms as a second-order Cartesian tensor.

1.36. Show that (Pijk + Pjki + Pjik) XiXjXk = 3PijkXiXjXko

Since all indices are dummy indices and the order of the variables Xi is unimportant, each term of the sum is equivalent to the others. This may be readily shown by introducing new dummy variables. Thus replacing i, j, k in the second and third terms by p, q, r, the sum becomes

Now change dummy indices in these same terms again so that the form is

1.37. If Bi, is skew-symmetric and Aij is symmetric, show that AijBij = O.

Since Aij = Aji and Bij = -Bji, AijBij = -AjiBji or AijBij + AjiBji = AijBij + ApqBpq o.

Since all indices are dummy indices, ApqBpq = AijBij and so 2AijBij = 0, or AijBij = O.

CHAP. 1]



1.38. Show that the quadratic form Dijxixj is unchanged if Do is replaced by its symmetric part DUj).

Resolving Dij into its symmetric and anti-symmetric parts,

Dij = DUj) + Dun = t(Dij + Dji) + t(Dij - Dji)


1.39. Use indicial notation to prove the vector identities (1) a X (b X c) = (a' c)b - (a' b)c,

(2) a X b· a


(1) Let v = b X c. Then vi = EijkbjCk; and if aX v = W, then

(Ilpjllqk - Ilpkllqj)aqbjck (see Problem 1.26) aqbpcq - aqbqcp

(aqcq)bp - (aqbq}cp

Transcribing this expression into symbolic notation,

W = a X (b X c) = (a' c)b - (a' b)c

(2) Let a X b = v. Thus Vi = Eijkajbk; and if A = v • a, then A = Eijk(aiajbk). But Eijk is skewsymmetric in i and i, while (aiajbk) is symmetric in i and j. Hence the product €ijkaiajbk vanishes as may also be shown by direct expansion.

A Eijlaiajbl + Eij2aiajb2 + Eij3aiajb3 (€32la3a2 + E23la2a3)bl + '"

(-a2a3 + a2a3)bl + (O)b2 + (O)b3 0

1.40. Show that the determinant

det A«

All A12 A13

A2l A22 A23

A3l A32 A33

may be expressed in the form €ijkAliA2jA3k-

From (1.52) and (1.109) the box product [abc] may be written

A = a'bxc


al a2 a3 EijkaibjCk = bl b2 b3

Cl C2 C3

If now the substitutions ai = Ali' bi = AZi and Ci = A3i are introduced,

This result may also be obtained by direct expansion of the determinant. An equivalent expression for the determinant is EijkAilAj2Ak3'

1.41. If the vector Vi is given in terms of base vectors a, b, c by Vi

aai + f3bi + yCi,

show that

a -

£ijkVibjCk £pqrapbqCr •

V2 aa2 + f3b2 + YCz

V3 aa3 + f3b3 + yC3




VI bi CI
V2 b2 C2
Va ba Ca EijkVibjCk
By Cramer's rule, " - and by (1.52) and (1.109), ,,=
al bi CI Epqrapbqcr
a2 b2 C2
aa bs Cs MATRICES AND MATRIX METHODS (Sec. 1.17-1.20)

A A A.At.

1.42. For the vectors a = 3 i + 4k, b = 2 j - 6k


4 j j - 5k j, compute by matrix multiplication

Let a- D = v; then [VI' V2, vs] [3, 0, 4][~ -:

o -5


and the dyadic D = 3 i i + 2 i k -

the products a·D, D·h and a·D·h.

n = [9,-20,61.

Let D· b = w; then [::] [~-: ~][~] .:

Ws 0 -5 0 -6 -10


1.43. Determine the principal directions and principal values of the second-order Cartesian tensor T whose matrix representation is

3 -1 0

[Tij] -1 3 0


From (1.132), for principal values A,

3 - A -1 0

-1 3 - A 0

o 0 I-A

(1 - A)[(3 - A)2 - 1]


which results in the cubic equation AS - 7A2 + 14A - 8 = (A - 1)(A - 2)(A - 4) = 0 whose principal values are A(1) = 1, A(2) = 2, A(S) = 4.

Next let n~l) be the components of the unit normal in the principal direction associated with A(l) = 1. Then'the first two equations of (1.131) give 2nil) - n~l) = 0 and -nil) + 2n~l) = 0, from which n~l) = n~l) = 0; and from nini = 1, n~l) = ±1.

For A(2) = 2, (1.131) yields n~2) - ni2) = 0, -n~2) + n~2) = 0, and _n~2) = o. Thus n~2) = n~2) = ±1/V2, since nini = 1 and n~2) = O.

For A(S) = 4, (1.131) yields -nia) - n~a) = 0, _n~S) - n~a) = 0, and 3n~a) = O. Thus n~S) = 0, niS) = _n~S) = +1IV2.

The principal axes xj may be referred to the original axes xi through the table of direction cosines

CHAP. 1]



Xl X2 Xa
X* 0 0 ±1
X* ±1/V2 ±1/V2 0
X* +1/V2 ±1IV2 0
a from which the transformation matrix (tensor) may be written:

roo ±lj

±1/V2 ±1/V2 0

+1/V2 ±1/V2 0


o ±1/V2 ±1/Y2

1.44. Show that the principal axes determined in Problem 1.43 form a right-handed set of orthogonal axes.

Orthogonality requires that the conditions aijaik = 0jk be satisfied. Since the condition nini = 1 was used in determining the aU' orthogonality is automatically satisfied for j = k. Multiplying the corresponding elements of any row (column) by those of any other row (column) and adding the products demonstrates that the conditions for j #- k are satisfied by the solution in Problem 1.43.

Finally for the system to be right-handed, it (2) X it (3) = it (1). Thus

o o

11V2 -1/V2

As indicated by the plus and minus values of au in Problem 1.43, there are two sets of principal axes, xi and xi*. As shown by the sketch both sets are along the principal directions with xi being a right-handed system, xj* a left-handed system.

x* 3


Fig. 1-19

1.45. Show that the matrix of the tensor Tij of Problem 1.43 may be put into diagonal (principal) form by the transformation law Ti~ = aipajq T pq, (or in matrix symbols 'T* = cA'TcAT).

r 1/~ 1/~ :j' -~ -~ :j':

-lIY2 11Y2 0 l 0 0 1 II

r ~ ~ :j':

-2Y2 212 0 II

1/V2 -1/V2J

11V2 11V2

o 0







[CHAP. 1

1.46. Prove that if the principal values A(1), A(2), A(3) of a symmetric second-order tensor are all distinct, the principal directions are mutually orthogonal.

The proof is made for A(2) and A(3)' For each of these (1.129) is satisfied, so that Tijnj2) =

(2) d T (3) - (3) MI' . (3)

A(2)ni an ijn; - A(3)ni' u tiplying the first of these equations by ni and the second


Since Tij is symmetric, the dummy indices i and j may be interchanged on the left-hand side of the second of these equations and that equation subtracted from the first to yield

(A(2) - A(3»ni2)ni3) = 0

Since A(2) ¥= A(3), their difference is not zero. Hence ni2)ni3) = 0, the condition for the two directions to be perpendicular.

1.47. Compute the principal values of (T)2 of Problem 1.43 and verify that its principal axes coincide with those of T.

[T"I' [-: -: :][-: -::J [~:~::]

The characteristic equation for this matrix is


o o

1 - A

(1 - A)[(10 - A)2 - 36]

(1 - A)(A - 4)(A - 16)


10 - A -6

-6 10 - A


from which A(l) = 1, A(2) = 4, A(3) = 16. Substituting these into (1.131) and using the condition nini = 1,

For A(!) = 1,

9ni1) - 6n~l) :}
-6nilJ + 9nil)
6ni2) - 6ni2) :}
-6ni2) + 6ni2)
-6ni3) - 6ni3) :}
-6ni3) - 6n~3)
-15n~3) For A(2) = 4,

For A(3) = 16,

which are the same as the principal directions of T.

1.48. Use the fact that (T)2 has the same principal directions as the symmetrical tensor T to obtain VT when

T (-~ -~ -~)

-1 0 4

First, the principal values and principal directions of T are determined. Following the procedure of Problem 1.43, the diagonal form of T is given by

CHAP. 1]


with the transformation matrix being

[ l/Va


l/Va 11 Val

1/...;2 -1/...;2

1/"f6 1/{6


Therefore ...fi*

(Va: 0: ~:)

V 0 and using [aii] to relate this to the original axes by the

transformation VT = AcVT* A, the matrix equation is

0l[ 1/V3

: ~ -2/~


v'2 - 2 ...;2+{6+1 v'2-V6+1





11V3 1/V31

1/...;2 -1/...;2

1/{6 1/{6

[ 5 414 -0.586 .402 -0:586 4.863 -0.586 -0.035




1.49. For the function ,\ = AiiXiXj where Aij is constant, show that a,\/aXi = (Au + Aii)xj and a2,\/aXiaXj = Ai) + Aji• Simplify these derivatives for the case Aij = Aji•

1.50. Use indicial notation to prove the vector identities (a) \7 x Vcf> = 0, (b) \7' \7 x a = 0.

(a) By (1.147), V¢ is written ¢,i and so v = V X V¢ has components Vi = <ijkaj¢,k = <ijk¢,kj' But <ijk is anti-symmetric in j and k, whereas ¢,kj is symmetric in j and k; hence the product <ijk¢,kj vanishes. The same result may be found by computing individually the components of v. For example, by expansion V1 = <123¢,23 + <13Z¢,32 = (¢,23 - ¢,32) = o.

1.51. Determine the derivative of the function ,\ = (X1)Z + 2X1X2 - (X3)2 in the direction of the unit normal n = (2!7)e1 - (3!7)ez - (6!7)e3 or n = (2e1 - 3ez - 6e3)!7.

Th . d d . ti . aA n A Th

e require errva rve IS an = v A' n = A,ini' us



[CHAP. 1

1.52. If Aij is a second-order Cartesian tensor, show that its derivative with respect to Xk, namely Aij,k, is a third-order Cartesian tensor.

For the Cartesian coordinate systems Xi and x[, Xi = ajiX; and ax/ax} = aji' Hence

which is the transformation law for a third-order Cartesian tensor.

1.53. If r2 = XiXi and f(r) is an arbitrary function of r, show that (a) v(f(r)) = f'(r)x/r, and (b) \l2(f(r)) = f"(r) + 2j'/r, where primes denote derivatives with respect to r.

(a) The components of Vf are simply f,i' Thus f,i = :~ :;.;


iJr _ Xj _ af iJr _ ,

follows that ;-- - -. Thus t.. - ;- " - f x/r.

aXj r or UXi

and since iJ(r2) = 2r~!:. = 28 "x· it

aXj aXj 1J t

(b) '\]2f = f .. = (!'x./r) . = f" XiXi + !' (.1! - XiXi) = f" + 'ij_' .

,tt 1,1 r2 r r3 r

1.54. Use the divergence theorem of Gauss to show that Is xm, dS = V8ij where n, dS represents the surface element of S, the bounding surface of the volume V shown in Fig. 1-20. Xi is the position vector of n, dS, and ni its outward normal.

By (1.157),

f x· ·dV



Fig. 1-20

1.55. If the vector b = V' x v, show that i sbm, dS = Iv i\.,ibi dV where i\. = i\.(Xi) is a scalar function of the coordinates.

Since b = V X V, bi = 'ijkvk,j and so

~ AbinidS = ~ EijkAVk,jnidS

by (1.157)

f ('''kA ,Vk . + '''kAVk .. ) dV

t) .!.J 1) ,}1


CHAP. 1]




1.56. For the arbitrary vectors a and b, show that

A = (a x b) • (a X b) + (a· b)2 (ab)2

Interchange the dot and cross in the first term. Then

A a • b X (a X b) + (a· b)(a· b)

a· [(b· b)a - (b· alb] + (a· b)(a· b)

(a· a)(b· b) - (b· a)(a • b) + (a· b)(a· b) (ab)2

since the second and third terms cancel.


1.57. If U = (II X u and v = (II X v, show that dt (u X v)

(a) In symbolic notation,

(II X (u X v).


at (u X v)

u X v + u X V = (01 X u) X v + u X (01 X v) (V·OI)U - (v - u) .. + (u- v) .. - (u· OI)V

(v· .. )u - (u- OI)V = '" X (u X v)


(b) In indicial notation, let dt (u X v) = w. Then


Wi = at ('ijkUjVk)

and since uj = 'jpqWpUq and Vk = 'kmnWmVn,

Wi = 'ijk'jpqWpUqVk + 'ijk'kmnUjWmVn

and using the result of Problem 1.59(a) below,

Wi (cSimcSjn - cSincSjm - cSimcSjn + cSijcSmn)UjWmVn

which is the indicial form of .. X (u X v).

1.58. Establish the identity fpqsfmnr

8mp 8mq 8ms

8np Bnq s.,

Let the determinant of Aij be given by det A or columns causes a sign change. Thus

An Al2 Al3 A21 A22 A23 A31 A32 A33

An interchange of rows

A21 A22 A23 Al2 All Al3
All Al2 Al3 = A22 A21 A23 -detA
A31 A32 A33 A32 A3l A33
and for an arbitrary number of row changes,
Ami Am2 Am3
Ani An2 An3 'mnr detA
Arl Ar2 Ar3 40


[CHAP. 1

or column changes,

AlP Alq Ais
A2P A2q A2s Epqs detA
A3P Aaq A3s: Hence for an arbitrary row and column interchange sequence,

Amp Amq Ams

Anp Anq Ans

Arp Arq Ars

EmnrEpqs det A

When Aij == Oii' det A = 1 and the identity is established.

1.59. Use the results of Problem 1.58 to prove (a) EpqsEsnr = apn8qr - 8praqn, (b) EpqsEsqr = -28pr.

Expanding the determinant of the identity in Problem 1.58,

EpqsEmnr = 0mp(onqOrs - OnsOrq) + 0mq(onsorp - 0npors) + 0ms(onpOrq - 0nqOrp)

(a) Identifying 8 with m yields,

0sp(OnqOrs - OnsOrq) + 0sq(onsorp - 0npors) + 0ss(onpOrq - °nqOrp) OrpOnq - Op"Orq + OqnOrv - OnpOqr + 3onporq - 30nqorp

OnpOrq - onqorp

(b) Identifying q with n in (a),

1.60. If the dyadic B is skew-symmetric B = -Be, show that B" X a = 2a· B.

Writing 8 = bl~1 + b2~2 + b3~3 (see Problem 1.6),

a, b1 X ~I + b2 X ~2 + b3 X ~3


(b, X~I) X a + (b2 X~2) X a + (b3 X~3) X a

(a· bl) ~l - (a· ~l)bl + (a· b2) ~2 - (a· ~2)b2 + (a- h3) ~3 - (a· ~3)b3 a > (hI ~l + b2~ + b3~3) - a· (~lbl + ~b2 + ~3b3)

a • 8 - a· 8e = 2a· 8

1.61. Use the Hamilton-Cayley equation to obtain (B)4 for the tensor B Check the result directly by squaring (B)2.

( ~ ~ -~).

-1 0-2

The characteristic equation for 8 is given by



o -2 - A


1- A 0

o 3 - A


By the Hamilton-Cayley theorem the tensor satisfies its own characteristic equation. Hence (8)3 - 2(8)2 - 68 + 91 = 0, and multiplying this equation by 8 yields (8)4 = 2(8)3 + 6(8)2 - 98 or (8)4 = 10(8)2 + 38 - 181. Hence

CHAP. 1]




o 9 o

1) + (3 ~ -3) (18 1~ 0) (5 8~ 7)

~ -~ 0 -~ ~ 0 1~ ~ 0 2~

Checking by direct matrix multiplication of (8)2,

(~)' .: m: ::] [: 8: 2:]

1.62. Prove that (a) Aii, (b) AijAij, (c) fijkfkjpAip are invariant under the coordinate transformation represented by (1.103), i.e. show that Aii = A~i, etc.

(a) By (1.103), Aij = apiaqjA~Q" Hence Aii = apiaqiA~q = IlpqA~q = A~p = A;i' (b) AijAij = apiaqjA~qamianjA~n = Ilpm8qnA~qA~n = A~qA~q = A;jA;j

fijkfkjpamianpA~n = (8ij8jp - 8ip8jj)amianpA~n

(Ilmn - Ilmn8jj)A~n = (8mj8nj - 8mnlljj)A;"n = fmjkfkjnA;"n

1.63. Show that the dual vector of the arbitrary tensor Tij depends only upon T[iil but that the product TijSij of Ti, with the symmetric tensor S« is independent of T[iil.

By (1.110) the dual vector of Tij is vi = fijk Tjk' or Vi = fijk (T(jk) + TUk1) = fijk T[jkl since fijkT(jk) = 0 (fijk is anti-symmetric in j and k, T(jk) symmetric in j and k).

For the product TijSij = T(ij)Sij + T[iilSij. Here T[ijlSij = 0 and TijSij = TUjlSij,

1.64. Show that D: E is equal to D·· E if E is a symmetric dyadic.

Write 0 = Dij~i~j and E = Epq~p-eQ" By (1.31), 0: E = DijEpq(~i • ~p)(~j • ~q). By (1.35), D· • E = Dij E pq (~j • ~p)( ~i • ~q) = Dij E qp (~j • ~)( ~i • ~q) since E pq = E qp' N ow interchanging the role of dummy indices p and q in this last expression, D·· E = DijEpq (~j • ~q)( ~i • ~p).

1.65. Use the indicial notation to prove the vector identity V x (a X b) = b- Va - b(V . a) + a(V ·b) - a· Vb.

Let V X (a X b) = v; then vp = fpqifUkaqajbk or

vp = fpqifijk(ajbk).q = fpqi<ijk(aj,qbk + ajbk,q)

= (8pj8qk - 8pkllqj)(aj,q bk + ajbk,q) = ap,q bq - aq,q bp + apbq,q - aqbp,q Hence v = b· Va - b(V • a) + a(V • b) - a· Vb.

1.66. By means of the divergence theorem of Gauss show that 1: Ii. x (a X x) dS 2a V

where V is the volume enclosed by the surface S having the outward normal n. The position vector to any point in V is x, and a is an arbitrary constant vector.

In the indicial notation the surface integral is f fqpinpfijkajxk dS. By (1.157) this becomes the volume integral f (fqpifijkajXk),p dV and since a fs constant, the last expression becomes


f (8qj8pk - 8qk 8pj)ajXk,p dV = f (aqxp,p - apxq,p) dV

• v v

f (aq Ilpp - ap 8qp) dV = f (3aq - aq) dV = 2aq V

v v



[CHAP. 1

1.67. For the reflection of axes shown in Fig. 1-21 show that the transformation is orthogonal.

From the figure the transformation matrix is

The orthogonality conditions aijaik = Iljk or ajiaki = 13 jk are clearly satisfied. In matrix form, by (1.117),

Fig. 1-21

1.68. Show that (I x v) . D = v x D.


1 X v (i i + j j + kk) x (vxi + vyj + vzk)

A A .Art. A A A A A A

i (Vyk - vzj) + j (-vxk + vzi) + k(vxj - vyi)

(vxi)'i + (vxj)j + (vXk)k = v XI

Hence (I X v) • D

v XI· D = v X D.

Supplementary Problems


~ ~ A ~ ~

Show that u = I + J - k and v = I - J

that u, v, ~ forms a right-handed triad.

are perpendicular to one another.

Ans. w = (-l/V6)(i + j + 2k)

A Determine w so


Determine the transformation matrix between the u, v, ~ axes of Problem 1.69 and the coordinate directions.

ll/v'3 l/yS -l/v'3l

1/V2 -1/V2 °

-l/V6 -l/V6 -2/V6



Use indicial notation to prove (a) ". x = 3, (b) " X x = 0, (c) a· "x = a where x is the position vector and a is a constant vector.

1.72. Determine the principal values of the symmetric part of the tensor Tij Ans. A{J) = -15, 1..(2) = 5, 1..(3) = 10

( : =~ -:).

-3 -18 1

CHAP. 1]



1.73. For the symmetric tensor Tij of the principal axes.

(~7 :3 ~O)

determine the principal values and the directions

Ans. A(1) = 2, A(2) = 7, A(3) = 12,

Xl X2 X3
X* -3/5V2 1/V2 -4/5V2
x* 4/5 0 -3/5
X* 3/5V2 1/V2 4/5V2
3 1.74. Given the arbitrary vector v and any unit vector ~, show that v may be resolved into a component parallel and a component perpendicular to ~,i.e. v = (v 0 ~) ~ + ~ X (v X ~).

1.75. If V 0 v = 0, V X v = wand V X w = -v, show that \72v = V.

1.76. Check the result of Problem 1.48 by direct multiplication to show that yT yT = T.

1.77. Determine the square root of the tensor B


(!(y'5 + 1) !(y'5 -1) o

!(V5 -1) 0) t(y'5 + 1) 0

o 3

1.78. Using the result of Problem 1.40, det A = EijkAliA2jA3k' show that det (AB)

det A det B.

1.80. Let the axes Ox~ x;x~ be related to OXIX2x3 by the table

Xl X2 X3
, 3/5V2 1/V2 4/5V2
, 4/5 0 -3/5
, -3/5y2 1/y2 -4/5V2
X3 (a) Show that the orthogonality conditions aijaik = Iljk and apqaSq = Ilps are satisfied. (b) What are the primed coordinates of the point having position vector x = 2~1 - ~3? (c) What is the equation of the plane Xl - X2 + 3X3 = 1 in the primed system?

Ans. (b) (2/5y'2, 11/5, -2/5V2) (c) y'2 x~ - x; - 2V2 x; = 1

x is the position vector and n the unit normal to the surface. Hint: and use (1.157).

v = t f V(xox) on dS where

S f

Write V = (1/6) (Xixi).jnj dS

. s


Show that the volume V enclosed by the surface S may be given as

Chapter 2

Analysis of Stress


The molecular nature of the structure of matter is well established. In numerous investigations of material behavior, however, the individual molecule is of n~onc~~-~;_d -orily--Uleoebivloroi'the materialas-U-whoIe-is -d-eemed important. - For these c~s~s-th-e . observed macroscopic b~havior Is "usiiaIfy" eXpIalned fjy"cHs-regarding- molecular considerations and, instead, by assuming the material to be continuously distributed throughout itsvolume a~~_1;Q com.~I_t:!leIYfiIl the space it occupies. This continuum conc~pt of matter- is'· the -fundamental postulate of Continuum Mechanics. Within the limitations for which the

continuum assumption is valid, this concept provides a framework for studying the behavior of solids, liquids and gases alike.

~~~~Q!1.of_tll_!L~9_l}tin~~l!!__Y!~~PO!!lt_ .. a~_~~~_?~~~~_~~~_Ql~ __ IE_3:!ll_~matical description of material behavior means that field quantities such as stress and displacement are expressed as piecewise continuous functions'ofthespace coordin"ates and "-tIme.- - _.

'-'~"-"-"---"-'"--~... - __ '._" .. _" .. ,-- __ .

-_ - .-'------- .. ' •.. - .... ----.------~---." .. ~~-


A hOJ?1-ogene()~_!l:!llj:eri~!J~_~ne hayil1_g:_!~e!!!icaI_..E!"operties-at _::lll J>Qi_nts. With respect to someproperty, a material is isotropic if that property is the same in all directions at a point. A material is caI1ed--ani-;oir-opic~ith:r:.espi~t-to-thoseproperties which are dLrectional :at a P?Il1J. --- ---------.------ ------ -- .-." . -- -.- - "- -----._- ...

The concept of density is developed from the mass-volume ratio in the neighborhood of a point in the continuum. In Fig. 2-1 the mass in the small element of volume .:l V is denoted by .:lM. The average density of the material within .:l V is therefore


The density at some interior point P of the volume element .:l V is given mathematically in accordance with the continuum concept by the limit,

lim .:lJl;~ dM (2.2)

P av .... O .:l V d~;:;/

Mass-density p is a scalar quantity.


Fig. 2-1


CHAP. 2]




Forces are vector quantities which are best described by intuitive concepts such as push or pull. Those forces which act on all elements of volume of a continuum are known as body forces. Examples are gravity and inertia forces. These forces are represented by the symbol b, (force per unit mass), or as Pi (force per unit volume). They are related through the density by the equation

pbi = Pi or pb = p


Those forces which act on a surface element, whether it is a portion of the bounding surface of the continuum or perhaps an arbitrary internal surface, are known as surface forces. These are designated by Ii (force per unit area). Contact forces between bodies are a type of surface forces.




A material continuum occupying the region R of space, and subjected to surface forces /i and body forces b., is shown in Fig. 2-2. _ As a result of forces being transmitted from one portion of the continuum to another, the material within an arbitrary volume V enclosed by the surface S interacts with the material outside of this volume. Taking ni as the outward unit normal at point P of a small element of surface AS of S, let A/i be the resultant force exerted across AS upon the material within V by the material outside of V. Clearly the force element A/i will depend upon the choice of AS and upon ni. It should also be noted that the distribution of force on AS is not necessarily uniform. Indeed the force distribution is, in general, equipollent to a force and a moment at P, as shown in Fig. 2-2 by the vectors A/i and AMi.

Fig. 2-2

Fig. 2-3

The average force per unit area on AS is given by A/d AS. The Cauchy stress principle asserts that this ratio A/dAS tends to a definite limit d/;/dS as AS approaches zero at the point P, while at the same time the moment of st. about the point P vanishes in the limiting process: The"~esuifing veCtor d/Jds (:force per unit area) is called-the ~tr~~--vect~~""tF) "a.ndIs·shown in Fig. 2-3. If the moment at P were not to vanish in the limiting process, a couple-stress vector, shown by the double-headed arrow in Fig. 2-3, would also be defined at the point. One branch of the theory of elasticity considers such couple stresses but they are not considered in this text.



Mathematically the stress vector is defined by

1· b./i Im-

<lS ... o t,.S


[CHAP. 2

1. M Imss ... o t,.S

df dS


The notation t~~) (or t<~») is used to emphasize the fact that the stress vector at a given point P in the continuum depends explicitly upon the particular surface element t,.S chosen there, as represented by the unit normal ni (or n). For some differently oriented surface element, having a different unit normal, the associated stress vector at P will also be different. The stress vector arising from the action across t,.S at P of the material within

V upon the material outside is the vector -ti~). Thus by Newton's law of action and



The stress vector is very often referred to as the traction vector.



At an arbitrary point P in a continuum, Cauchy's stress principle associates a stress vector ti~) with each unit normal vector ru, representing the orientation of an infinitesimal surface element having P as an interior point. This is illustrated in Fig. 2-3. The totality of all possible pairs of such vectors t~~) and n; at P defines the state of stress at that point. Fortunately it is not necessary to specify every pair of stress and normal vectors to completely describe the state of stress at a given point. This may be accomplished by giving the stress vector on each of three mutually perpendicular planes at P. Coordinate transformation equations then serve to relate the stress vector on any other plane at the point to the given three.

Adopting planes perpendicular to the coordinate axes for the purpose of specifying the state of stress at a point, the appropriate stress and normal vectors are shown in Fig. 2-4.

Fig. 2-4

For convenience, the three separate diagrams in Fig. 2-4 are often combined into a single schematic representation as shown in Fig. 2-5 below.

Each of the three coordinate-plane stress vectors may be written according to (1.69) in terms of its Cartesian components as

A ti~') el + t~~,) e2 + t~~,) e3 tJ~') ej
t<~,) ti~2) el + t~~2) e2 + t~~,) e3 A
t;e,) ~ (2.6)
t(~) ti~3) el + t~~,,) e2 + t~~3) e3 t(~3) A
j e, CHAP. 2]



Fig. 2-5

The nine stress vector components,

Fig. 2-6


are the components of a second-order Cartesian tensor known as the stress tensor. The equivalent stress dyadic is designated by I, so that explicit component and matrix representations of the stress tensor, respectively, take the forms



Pictorially, the stress tensor components may be displayed with reference to the coordinate planes as shown in Fig. 2-6. The components perpendicular to the planes (u11, U22' u33) are called normal stresses. Those acting in (tangent to) the planes (u12' U13' U21' U23' U31' u32) are called shear stresses. A stress component is positive when it acts in the positive direction of the coordinate axes, and on a plane whose outer normal points in one of the positive coordinate directions. The component o .. acts in the direction of the jth


coordinate axis and on the plane whose outward normal is parallel to the ith coordinate

axis. The stress components shown in Fig. 2-6 are all positive.


The relationship between the stress tensor Uu at a point P and the stress vector t~~) on a plane of arbitrary orientation at that point may be established through the force equilibrium or momentum balance of a small tetrahedron of the continuum, having its vertex at P. The base of the tetrahedron is taken perpendicular to tu, and the three faces are taken perpendicular to the coor-

. dinate planes as shown by Fig. 2-7. Designating the area of the base ABC as dS, the areas of the faces are the projected areas, dS1 = dS n1 for face CPB, dS2 = dS n2 for face APC, dS3 = dS n3 for face BP A or

Fig. 2-7



(CHAP. 2


The average traction vectors -tf(;;) on the faces and tf(~) on the base, together with the average body forces (including inertia forces, if present), acting on the tetrahedron are shown in the figure. Equilibrium of forces on the tetrahedron requires that


If now the linear dimensions of the tetrahedron are reduced in a constant ratio to one another, the body forces, being an order higher in the small dimensions, tend to zero more rapidly than the surface forces. At the same time, the average stress vectors approach the specific values appropriate to the designated directions at P. Therefore by this limiting process and the substitution (2.9), equation (2.10) reduces to

t~~) dS = tl~') nl dS + tl~2) n2 dS + t~~3) n« dS = t~~;) nj dS Cancelling the common factor dS and using the identity t~~;) == "w (2.11) becomes

t<~) n t(~) = n' I

i = uji j or



Equation (2.12) is also often expressed in the matrix form [t~h = [n1k] [ukj]

which is written explicitly



The matrix form (2.14) is equivalent to the component equations



Equilibrium of an arbitrary volume V of a continuum, subjected to a system of surface forces ti~) and body forces b, (including inertia forces, if present) as shown in Fig. 2-8, requires that the resultant force and moment acting on the volume be zero.

Summation of surface and body forces

results in the integral relation,
Is t~~) dS + 1: pbidV 0
or Is t<~) dS (2.16)
+ Iv pbdV X2
0 Replacing t i~) here by Ujinj and converting the resulting surface integral to a volume integral by the divergence theorem of Gauss (1.157), equation (2.16) becomes

Fig. 2-8

CHAP. 2]



r (a ... + pb.) dV

Jv J~.J 1






Since the volume V is arbitrary, the integrand in (2.17) must vanish, so that

aji,j + pbi = 0 or V' • I + ph = 0 which are called the equilibrium equations.

In the absence of distributed moments or couple-stresses, the equilibrium of moments about the origin requires that




Is x x t<~) dS + Iv x x ph dV in which Xi is the position vector of the elements of surface and volume. Again, making the substitution t;~) = ajinj' applying the theorem of Gauss and using the result expressed in (2.18), the integrals of (2.19) are combined and reduced to


For the arbitrary volume V, (2.20) requires


Equation (2.21) represents the equations alZ = a2l' aZ3 = a3Z' al3 = a3l' or in all




Iv = 0


which shows that the stress tensor is symmetric. In view of (2.22), the equilibrium equations (2.18) are often written


which appear in expanded form as
aall + aa!Z + aa13 + pbl
aXI aX2 aX3
aaZ! + aaZ2 + aa23 + pb2
ax! aX2 aX3
aa3l + aa32 + aa33 + pb3

At the point P let the rectangular Cartesian coordinate systems PXIXZX3 and Px{x~x~ of Fig. 2-9 be related to one another by the table of direction cosines

Xl X2 X3
Xl all al2 al3
, a22
X2 a21 a23
X~ a3l a32 a33 o




Fig. 2-9



[CHAP. 2

or by the equivalent alternatives, the transformation matrix [au], or the transformation dyadic


According to the transformation law for Cartesian tensors of order one (1.93), the components of the stress vector t~~) referred to the unprimed axes are related to the primed axes components t;(~) by the equation

t'(~) - a··t(~) i - tJ j



Likewise, by the transformation law (1.102) for second-order Cartesian tensors, the stress tensor components in the two systems are related by


In matrix form, the stress vector transformation is written [t;i~)] = [aij][t~f)]

and the stress tensor transformation as

[<Tij] = [aip][<Tpq][aqj]



Explicitly, the matrix multiplications in (2.28) and (2.29) are given respectively by
[;<:'l lau a12 a"li:'J
t~(n) a21 a22 a23 t~n) (2.30)
t'(~) a31 a32 a33 t(~)
3 3
l":' <T:2 ":'l lau a12 a'I"u <T 12 ""lu a21 a"l
and <T21 <T;2 <T23 a21 a22 a23 <T21 <T22 <T23 a12 a22 a32 (2.31)
<T£1 <T;2 <T£3 a31 a32 a33 <T31 <T32 <T33 a13 a23 a33 2.9 STRESS QUADRIC OF CAUCHY

At the point P in a continuum, let the stress tensor have the values <T .. when referred to directions


parallel to the local Cartesian axes P'1 '2'3 shown

in Fig. 2-10. The equation


represents geometrically similar quadric surfaces having a common center at P. The plus or minus choice assures the surfaces are real.

The position vector r of an arbitrary point lying on the quadric surface has components 'i = rni, where n; is the unit normal in the direction of r. At the point P the normal component <T Nni of the stress vector t:~) has a magnitude


Fig. 2-10

Accordingly if the constant k2 of (~.32) is set equal to CTNr2, the resulting quadric <T;j'i'i = ±<TNr2





is called the stress quadric of Cauchy. From this definition it follows that the magnitude UN of the normal stress component on the surface element dS perpendicular to the position vector r of a point on Cauchy's stress quadric, is inversely proportional to r2, i.e. UN = ±k2/r2.

Furthermore it may be shown that the stress vector ti~) acting on dS at P is parallel to the normal of the tangent plane of the Cauchy quadric at the point identified by r.


At the point P for which the stress tensor components are Uij' the equation (2.12), ti~) = uji nj, a;;sociates with each direction ni a stress vector tin). Those directions for which ti~) and ni are collinear as shown in Fig. 2-11 are called principal stress directions. For a principal stress direction,

t(~) - -"'1.

i - VI"'f,



in which (J', the magnitude of the stress vector, is called a principal stress value. Substituting (2.35) into (2.12) and making use of the identities n. = S .. n.

1. tJ J

and (J'ij = uji' results in the equations


Fig. 2-11

In the three equations (2.36), there are four unknowns, namely, the three direction cosines n; and the principal stress value (J'.

For solutions of (2.36) other than the trivial one n, = 0, the determinant of coefficients, !uij - Iliju!, must vanish. Explicitly,


which upon expansion yields the cubic polynomial in (J',



(J'3 - II(J'2 + IIIu - IIII 0 (2.38)
where II aii = tr E (2.39)
III t( (J'iiUjj - (Tijrrij) (2.40)
IIII !(J'ij! = det Z (2.41) are known respectively as the first, second and third stress invariants.

The three roots of (2.38), (J'(J)' u(2)' U(3) are the three principal stress values. Associated with each principal stress UCk)' there is a principal stress direction for which the direction cosines n;kl are solutions of the equations

(I - £I I) . Ii Ck) = 0



(k = 1,2,3)



[CHAP. 2

In (2.42) letter subscripts or superscripts enclosed by parentheses are merely labels and as such do not participate in any summation process. The expanded form of (2.42) for the second principal direction, for example, is therefore

( - )nC2) + n(2) + n(2)

O"tl O"(2) 1 0"12 2 0"13 3



Because the stress tensor is real and symmetric, the principal stress values are also real.

When referred to principal stress directions, the stress matrix [O"ij] is diagonal,

in the second form of which Roman numeral subscripts are used to show that the principal stresses are ordered, i.e, O"r > O"n > O"nr. Since the principal stress directions are coincident with the principal axes of Cauchy's stress quadric, the principal stress values include both the maximum and minimum normal stress components at a point.

In a principal stress space, i.e. a space whose axes are in the principal stress directions and whose coordinate unit of measure is stress (ti~), t~~), t~~)) as shown in Fig. 2-12, the arbitrary stress vector ti~) has components


tCn) - n

2 - O"(2) 2'






Fig. 2-12


according to (2.12). But inasmuch as (nJ)2 + (n2)2 + (n3)2 = 1 for the unit vector ni, (2.45) requires the stress vector ti~) to satisfy the equation



in stress space. This equation is an ellipsoid known as the Lame stress ellipsoid.


If the stress vector ti~) is resolved into orthogonal components normal and tangential to the surface element dS upon which it acts, the magnitude of the normal component may be determined from (2.33) and the magnitude of the tangential or shearing component is given by


CHAP. 2]



This resolution is shown in Fig. 2-13 where the axes are chosen in the principal stress directions and it is assumed the principal stresses are ordered according to O'r > O'n > O'rn'

Hence from (2.12), the components of ti~) are

t(n) O'rnl
t(n) O'nn2 (2.48)
t(~) O'IIrna
3 and from (2.33), the normal component mag-

nitude is


Substituting (2 .. ~8) and (2.49) into (2.47), the squared magnitude of the shear stress as a function of the direction cosines n; is given by

Fig. 2-13


The maximum and minimum values of "s maybe obtained from (2.50) by the method of Lagrangian multipliers. The procedure is to construct the function

F = O'~ - Anini


in which the scalar A is called a Lagrangian multiplier. Equation (2.51) is clearly a function of the direction cosines ni, so that the conditions for stationary (maximum or minimum) values of F are given by aF/an; = O. Setting these partials equal to zero yields the


nl{O'~ - 20'r(O'rni + O'IIn~ + O'mn~) + A} = 0 n2{0'~r- 20'n(O'rni +O'nn~ + O'mn~) + A} = 0 na{O'~n - 20'm(O'rni + O'nn~ + O'mn~) + A} = 0

(2.52a) (2.52b) (2.52c)

which, together with the condition n;n, = 1, may be solved for A and the direction cosines nl, n2, na, conjugate to the extremum values of shear stress.

One set of solutions to (2.52), and the associated shear stresses from (2.50), are

nl ±1, n2 0, na O' for which O's = 0 (2.53a)
nl 0, n2 ±1, na O' for which O's = 0 (2.53b)
nl 0, n2 0, na ±1; for which O's = 0 (2.53c) The shear stress values in (2.53) are obviously minimum values. Furthermore, since (2.35) indicates that shear components vanish on principal planes, the directions given by (2.53) are recognized as principal stress directions.

A second set of solutions to (2.52) may be verified to be given by

nl 0, n2 ±1/y'2, na ±1/y'2; for which O's = hr - O'm)/2 (2.54a)

nl ±1/y'2, n2 0, na ±1/y'2; forwhich O's=(O'm-O'r)/2 (2.54b)

nl ±Vy'2, n2 ±Vy'2, na 0; for which "s = (O'r - O'n)/2 (2.54c)

Equation (2.54b) gives the maximum shear stress value, which is equal to half the difference of the largest and smallest principal stresses. Also from (2.54b), the maximum shear stress component acts in the plane which bisects the right angle between the directions of the maximum and minimum principal stresses.




A convenient two-dimensional graphical representation of the three-dimensional state of stress at a point is provided by the wellknown Mohr's stress circles. In developing these, the coordinate axes are again chosen in the principal stress directions at P as shown by Fig. 2-14. The principal stresses are assumed to be distinct and ordered according to


For this arrangement the stress vector ti~) has normal and shear components whose magnitudes satisfy the equations

[CHAP. 2

Fig. 2-14

(2.56) (2.57)

Combining these two expressions with the identity 11-ini = 1 and solving for the direction cosines ni, results in the equations

(nl)2 (aN - an)(aN - am) + (as)2 (2.58a)
(aj - an)( aj - am)
(n2)2 (aN - am)(aN - aj) + (as)2 (2.58b)
(an - am)( an - aj)
(n3)2 (aN - aj)(aN - an) + (as)2 (2.58c)
(am - aj)(am - an) These equations serve as the basis for Mohr's stress circles, shown in the "stress plane" of Fig. 2-15, for which the "» axis is the abscissa, and the as axis is the ordinate.

In (2.58a), since aj - all > 0 and ar - am > 0 from (2.55), and since (nl)2 is non-negative, the numerator of the right-hand side satisfies the relationship


which represents stress points in the (aN' as) plane that are on or exterior to the circle

[aN - hj+am)/2j2 + (as)2 = [hj·-am)/2j2 In Fig. 2-15, this circle is labeled Ct.


Fig. 2-15

CHAP. 2]



Similarly, for (2.58b), since an- am> 0 and au - al < 0 from (2.55), and since (n2)2 is non-negative, the right hand numerator satisfies

(aN-am)(aN-al) + (as)2 ..,c:: 0 which represents points on or interior to the circle

[aN - (am + al)/2]2 + (as)2 = [(am - al)/2)2



labeled C2 in Fig. 2-15. Finally, for (2.58c), since am - al < 0 and am - au < 0 from (2.55), and since (n3)2 is non-negative,

(aN - al)(aN - an) + (as)2 ==== 0 which represents points on or exterior to the circle

[aN - (al + aII)/2]2 + (as)2 = [(al - aII)/2)2



labeled C3 in Fig. 2-15.

Since each "stress point" (pair of values of "» and as) in the (aN' as) plane represents a particular stress vector ti~>' the state of stress at P expressed by (2.58) is represented in Fig. 2-15 as the shaded area bounded by the Mohr's stress circles. The diagram confirms a maximum shear stress of (al - am)/2 as was determined analytically in Section 2.11. Frequently, because the sign of the shear stress is not of critical importance, only the top half of this symmetrical diagram is drawn.

The relationship between Mohr's stress diagram and the physical state of stress may be established through consideration of Fig. 2-16, which shows the first octant of a sphere of the continuum centered at point P. The normal 10 at the arbitrary point Q of the spherical surface ABC simulates the normal to the surface element dS at point P. Because of the symmetry properties of the stress tensor and the fact that principal stress axes are used in Fig. 2-16, the state of stress at P is completely represented through the totality of locations Q can occupy on the surface ABC. In the figure, circle arcs KD, GE and FH designate locations for Q along which one direction cosine of 10 has a constant value. Specifically,

I /



/ /

Fig. 2-16



[CHAP. 2

n1 = cos cp on KD,

n2 = cos{3 on GE,


cos s on FH

and, on the bounding circle arcs BC, CA and AB,

n1 = cos 7r/2 = 0 on BC,

n2 = cos 7r/2 = 0 on CA,

na = cos7r/2 = 0 on AB

According to the first of these and the equation (2.58a), stress vectors for Q located on BC will have components given by stress points on the circle C1 in Fig. 2-15. Likewise, CA in Fig. 2-16 corresponds to the circle C2, and AB to the circle C« in Fig. 2-15.

The stress vector components UN and Us for an arbitrary location of Q may be determined by the construction shown in Fig. 2-17. Thus point e may be located on C« by drawing the radial line from the center of C« at the angle 2{3. Note that angles in the physical space of Fig. 2-16 are doubled in the stress space of Fig. 2-17 (arc AB subtends 900 in Fig. 2-16 whereas the conjugate stress points U1 and Un are 1800 apart on Ca). In the same way, points g, hand f are located in Fig. 2-17 and the appropriate pairs joined by circle arcs having their centers on the UN axis. The intersection of circle arcs ge and hi represents the components UN and Us of the stress vector t~~) on the plane having the normal direction ni at Q in Fig. 2-16.

Fig. 2-17


In the case where one and only one of the principal stresses is zero a state of plane stress is said to exist. Such a situation occurs at an unloaded point on the free surface bounding a body. If the principal stresses are ordered, the Mohr's stress circles will have one of the characterizations appearing in Fig. 2-18.

UUI = 0

Un = 0

Fig. 2-18

CHAP. 2]



If the principal stresses are not ordered and the direction of the zero principal stress is taken as the X3 direction, the state of stress is termed plane stress parallel to the XIXZ plane. For arbitrary choice of orientation of the orthogonal axes Xl and X2 in this case, the stress matrix has the form


The stress quadric for this plane stress is a cylinder with its base lying in the XIX2 plane and having the equation


Frequently in elementary books on Strength of Materials a state of plane stress is represented by a single Mohr's circle. As seen from Fig. 2-18 this representation is necessarily incomplete since all three circles are required to show the complete stress picture. In particular, the maximum shear stress value at a point will not be given if the single circle presented happens to be one of the inner circles of Fig. 2-18. A single circle Mohr's diagram is able, however, to display the stress points for all those planes at the point P which include the zero principal stress axis. For such planes, if the coordinate axes are chosen in accordance with the stress representation given in (2.65), the single plane stress Mohr's circle has the equation


The essential features in the construction of this circle are illustrated in Fig. 2-19. The circle is drawn by locating the center C at "» = (all + a2z)/2 and using the radius R = V[(al1 - a22)/2F + (a12)2 given in (2.67). Point A on the circle represents the stress state on the surface element whose normal is nl (the right-hand face of the rectangular parallelepiped shown in Fig. 2-19). Point B on the circle represents the stress state on the top surface of the parallelepiped with normal n2. Principal stress points ar and an are so labeled, and points E and D on the circle are points of maximum shear stress value.


Fig. 2-19


It is very often useful to split the stress tensor a .. into two component tensors, one of


which (the spherical or hydrostatic stress tensor) has the form



[CHAP. 2




where UM = -p = uk/3 is the mean normal stress, and the second (the deviator stress tensor)
has the form
C" -UM 0'12 0'13 C S12 ''')
ID 0'21 0'22 - UM U" ) - S21 S22 S23 (2.69)
0'31 0'32 0'33 - uM S31 S32 S33 This decomposition is expressed by the equations

uij = 8ijUkkl3 + Sij or I = uMI + ID


The principal directions of the deviator stress tensor Sij are the same as those of the stress tensor uij• Thus principal deviator stress values are


The characteristic equation for the deviator stress tensor, comparable to (2.38) for the stress tensor, is the cubic

S3 + IhDs - IIhD = 0 or S3 + (SISn + SnSm + SmSI)S - SISnSm = 0 (2.72)

It is easily shown that the first invariant of the deviator stress tensor hD is identically zero, which accounts for its absence in (2.72).

Solved Problems


2.1. At the point P the stress vectors t~~) and t~~*) act on the respective surface elements n; IlS and n r IlS*. Show that

the component of ti~) in the direction of nf is equal to the component of

t~~*) in the direction of n;.

It is required to show that t~~*)n. = t~~)n*

1. 1. 1. 1.


From (2.12) t~n*)ni = I1j;njni' and by (2.22)

I1ji = l1ij' so that

Fig. 2-20

CHAP. 2]



2.2. The stress tensor values at a point P are given by the array

I (~~ -~)

-2 0 4

Determine the traction (stress) vector on the plane at P whose unit normal is n = (2/3) Cl - (2/3) C2 + (1/3) C3.

From (2.12), t(~) = ~. I. The multiplication is best carried out in the matrix form of (2.13):

[2/3, -2/3, 1/3] [ : -2

[14 _ ~ -10 -4 + !]

3 3' 3 '3 3

2.3. For the traction vector of Problem 2.2, determine (a) the component perpendicular to the plane, (b) the magnitude of t~~), (c) the angle between ti~) and n.


(a) tin). ~ = (4~1 - !§2ez)· (i~l - i~2 + t~3) = 44/9

(b) Iti~)1 = V16 + 100/9 = 5.2

(c) Since ti~)· ~ = Iti~)1 coso, cos s = (44/9)/5.2 = 0.94 and 0 = 20°.

2.4. The stress vectors acting on the three coordinate planes are given by ti~l), ti~2) and t~~3).

Show that the sum of the squares of the magnitudes of these vectors is independent of the orientation of the coordinate planes.

Let S be the sum in question. Then

S = t~~l)t~~l) + t~~2)t~~2) + t~~3)t~~3)

t t t t 1. t

2.5. The state of stress at a point is given by the stress tensor

( o au bU)

au a co

b« co o

where a, b, c are constants and o is some stress value. Determine the constants a, b and c so that the stress vector on the octahedral plane (n = (1/V3) Cl + (1/-/3) C2 + (1/V3 )C3) vanishes.

In matrix form, ti~) = O-ijnj must be zero for the given stress tensor and normal vector.

[a: ~ ::] [~;~] = [~]

bo- Co- 0- 1/y'3 0





hence a + c

b + c

Solving these equations, a = b = c = -1/2. Therefore the solution tensor is

-0-/2 -0-/2)

0- -0-/2

-0-/2 0-



2.6. The stress tensor at point P is given by the


I (-~ -~ ~)

Determine the stress vector on the plane passing through P and parallel to the plane ABC shown in Fig. 2-21.

The equation of the plane ABC is 3xl + 6xz + 2X3 = 12, and the unit normal to the plane is therefore (see Problem 1.2)

[CHAP. 2


Fig. 2-21

From (2.14), the stress vector may be determined by matrix multiplication,

[317, 6/7, 217{ -: -: :] ~ t [-9, 6,101

2.7. The state of stress throughout a continuum is given with respect to the Cartesian axes OX1XZX3 by the array

I ( 3:~~' :: 2:.)

Determine the stress vector acting at the point P(2, 1, V3) of the plane that is tangent to the cylindrical surface x~ + x; = 4 at P.

At P the stress components are given by

The unit normal to the surface at P is determined from grad ¢ = V¢ = V(x~ + x~ - 4). Thus V¢ = 2X2~2 + 2X3~3 and so

V'¢ = 2~2 + 2v'3~3 at P

'" ~ -Js :

Therefore the unit normal at P is n = 2 + 2e3.

This may also be seen in Fig. 2-22. Finally the stress vector at P on the plane ..L to ~ is given by

Fig. 2-22

CHAP. 2]




2.8. For the distribution of the state of stress given in Problem 2.7, what form must the body force components have if the equilibrium equations (2.24) are to be satisfied everywhere.

Computing (2.24) directly from I given in Problem 2.7,

3X2 + 10x2 + 0 + pb1 0

o + 0 + 2 + pb2 = 0

o + 0 + 0 + pb3 = 0 These equations are satisfied when b1 = -13x2/p, b2 = -2/p, b3 = O.

2.9. Derive (2.20) from (2.19), page 49.

Starting with equation (2.19),

f EijkXjt~~) dS + f EijkXjPbk dV = 0

s v

in the surface integral and convert the result

substitute (1.157):


tin) = Ujinj

to a volume integral by

Carrying out the indicated differentiation in this volume integral and combining with the first volume integral gives


But from equilibrium equations, upk,p + pbk = 0; and since Xj,p = 8jp, this volume integral reduces to (2.20), f EijkUjk dV = O.



2.10. The state of stress at a point is given with respect to the Cartesian axes OXIX2X3 by the array

I (-: ~ ~)

o O-V2

Determine the stress tensor I' for the rotated axes OX;X~X3 axes by the transformation tensor

( 0

1/\1'2 -11V2

related to the unprimed


11V2 1/V2)

1/2 -112

1/2 -1/2

The stress transformation law is given by (2.27) as Ulj = aipajqUpq or I' = A' I . Ac' The detailed calculation is best carried out by the matrix multiplication [uij] = [aip][upq][aqj] given in (2.29). Thus

[ 0 1112 1/12] [ 2 -2 '][0 1112 -1112]
[ u[j] 1112 1/2 -1/2 -2 12 0 1/12 1/2 1/2
-1112 1/2 -1/2 0 o -12 ]/12 -1/2 -1/2
[: 0 1~~]
-1 62


[CHAP. 2

2.11. Show that the stress transformation law may be derived from (2.33), the equation aN = aijninj expressing the normal stress value on an arbitrary plane having the unit normal vector ni.

Since UN is a zero-order tensor, it is given with respect to an arbitrary set of primed or un primed axes in the same form as

and since by (1.94) ni = uijnj,

where new dummy indices have been introduced in the last term. Therefore

and since the directions of the unprimed axes are arbitrary,

2.12. For the unprimed axes in Fig. 2-23, the stress tensor is given by

Determine the stress tensor for the primed axes specified as shown in the figure.

It is first necessary to determine completely the transformation matrix A. Since xi makes equal angles with the Xi axes, the first row of the transformation table together with U33 is known. Thus

Fig. 2-23

Xl X2 X3
xi 1/13 1113 1113
X3 1//2 Using the orthogonality equations uijuik = 8jk, the transformation matrix is determined by computing the missing entries in the table. It is left as an exercise for the student to show that

Therefore [urj]

['/V3 '/V3 '/V3]
-2/% 1/% 1/-16
o -1//2 1//2
[1IV3 '/V3 '/V3l 0 IlIV3 -2/V3 0 J
-2/-16 1/-16 11-16 0 T o 1113 1/V6 -1/Y2
o -1/Y2 1//2 0 0 T 1/13 11V6 1IV2 CHAP. 2]




r/Vs] [1/Vs -2/16 0]

rh/6 1/Vs 1/% -1/V2

r/V2 1/Vs 1/% 1/V2


The result obtained here is not surprising when one considers the Mohr's circles for the state of stress having three equal principal stress values.


2.13. Determine the Cauchy stress quadric at P for the following states of stress: (a) uniform tension Ull = U22 = U33 = u; UI2 = UI3 = U23 = 0

(b) uniaxial tension Ull = o ; U22 = U33 = UI2 = UI3 = U23 = 0

(c) simple shear UI2 = U21 = r; Ull = U22 = U33 = UI3 = U23 = 0

(d) plane stress with Ull = U = u22; UI2 = U21 = T; U33 = U31 = U23 = o.

From (2.32), the quadric surface is given in symbolic notation by the equation r' I • r = ±k2• Thus using the matrix form,

(a) it" t" t,{ ~ ~ :][i:] ~ .t: + .t: +otl ~ ±k'

Hence the quadric surface for uniform tension is the sphere ri + r~ + r~


Hence the quadric surface for uniaxial tension is a circular cylinder along the tension axis.

(c) [rl, b r3J [~ ~ ~] [~:] = 2Trlr2 = ±k2

o 0 0 r3

and so the quadric surface for simple shear is a hyperbolic cylinder parallel to the r3 axis.

(~ it" t" t.f ~ mi:] ~ .,: + 2·t ,t, + .tl ~ ±k'

and so the quadric surface for plane stress is a general conic cylinder parallel to the zero principal axis.

2.14. Show that the Cauchy stress quadric for a state of stress represented by

I (~~~)

o 0 c

is an ellipsoid (the stress ellipsoid) when a, band c are all of the same sign.

The equation of the quadric is given by

[rl' r2' r3J [~ ~ ~] [~:] = ari + br~ + cr~

o 0 c r3


2 r2 r2

Therefore the quadric surface is the ellipsoid !~ + a~ + a~

±k2 abc "



[CHAP. 2


2.15. The stress tensor at a point P is given with respect to the axes OX1X2X3 by the values

"u (: ~ D

Determine the principal stress values and the principal stress directions represented by the axes Oxtxix:.

From (2.37) the principal stress values U are given by

3-u 1 1 1 -u 2 1 2-u

o or, upon expansion, (u + 2)(u - 4)(u - 1) = 0

The roots are the principal stress values U(1) = -2, U(2) = 1, U(3) = 4. Let the xr axis be the direction of U(1), and let nil) be the direction cosines of this axis. Then from (2.42),

(3 + 2)nil) + n~l) + n~l) 0

nil) + 2n~l) + 2n~l) 0

n(l) + 2n(1) + 2n(l) 0


Hence nill = 0; n~l) = -n~1) and since nini = 1, (n~1l)2 = 1/2. Therefore nil) = 0, n~ll = 11...;2,

n~ll = -1/"';2.

Likewise, let x; be associated with U(2). Then from (2.42),

2ni2) + n~2) + n~2) 0

ni2) - n~2) + 2n~2) 0

ni2) + 2ni2) - n~2) 0

so that ni2) = 1/..,;3, n~2) = -1/..,;3, n~2) = -1/...;3 .

Finally, let x~ be associated with U(3). Then from (2.42),

-ni3) + n~3) + n~3) 0

ni3) - 4n~3) + 2n~3) 0

ni3) + 2n~3) - 4n~3) 0

2.16. Show that the transformation tensor of direction cosines determined in Problem 2.15 transforms the original stress tensor into the diagonal principal axes stress tensor.

According to (2.29), [u~j] = [aip][upq][aqj], which for the problem at hand becomes

[u~jl [ 1/~ -:;~ =:;~] [~ ~ ~] [ 1/~2 -~;~ =:;~]

-2//6 -1/"';6 -11/6 1 2 0 -11...;2 -1/V3 -1//6

[ 0 -...;2 ...;2~ [ 0 1/V3 -2//6]

1/V3 -1/V3 -11y3 11...;2 -1/V3 -1/...;6

-8/16 -4//6 -4/..;6 -1/...;2 -1/...;3 -1/V6

CHAP. 2]



2.17. Determine the principal stress values and principal directions for the stress tensor

From (!2.37),




Hence u(1):= 0, U(Z) = 0, u(3):= 37. For u(3):= 37, (2.42) yield





:= 0 or (7 - U)[-27U + u2] + 272U := [37 - u]u2 := O.



n(3) + n(3) - 2n (3)

1 2 3

and therefore n (3) - n (3) - n (3) - 1/- '3 For U - - 0

1 - 2 - 3 - V u. (1) - U(2) - ,

(2.42) yield

which together with njnj = 1 are insufficient to determine uniquely the first and second principal directions. Thus any pair of axes perpendicular to the nj3) direction and perpendicular to each other may serve as principal axes. For example, consider the axes determined in Problem 2.12, for which the transformation matrix is

[ 1/V3 1/V3 -2/V6 1/V6 o -1/12

According to the transformation law (2.29), the principal stress matrix [u~j] is given by

[ 1/V3 1/V3 -2/% 1/% o -1/12

1/V3] [7 r 7] [l/Ys -2/V6 0]

1/~ 7 7 7 1/V3 1/% -1/12

1/y2 7 7 7 1IYs 11V6 1112

V37 o


V3 7] [1/V3 -2/% 0]

o llVs 1/V6 -1/..;2

o 11V3 1/% 1/12

2.18. Show that the axes Oxix~x~, (where xi, X3 and x~ are in the same vertical plane, and xi, Xl and X2 are in the same horizontal plane) are also principal axes for the stress tensor of Problem 2.17.

The transformation matrix [ajj] relating the two sets of axes clearly has the known elements

[au] = [- - Y2~V3]

1/V3 1IV3 1/V3

as is evident from Fig. 2-24. From the orthogonality conditions aijaik = 8jk, the remaining four elements are determined so that

Fig. 2-24



[CHAP. 2

[-11112 11-12 -1IV6 -1IV6 lI..j3 lIVa

As before,




2.19. Show that the principal stresses u~k) and the stress components uij for an arbitrary set of axes referred to the principal directions through the transformation coefficients 3

aij are related by uij = ~ apiapjUp'


From the transformation law for stress (2.27), uij = apjaqju;q; but since U;q are principal stresses, there are only three terms on the right side of this equation, and in each p = q. There- 3

fore the right hand side may be written in form Uij = ~ apjapju;,


2.20. Prove that U .. U"kUk" is an invariant of the stress tensor.

\} 1 J

By the transformation law (2.27),

aipajqUpqairaksursakmajnUmn (aipair) (ajqajn)( aksakm)UpqU rsU mn llpr8qn8smupqUrsUmn

(llprupq) (llqnUmn){ II smUTs)

2.21. Evaluate directly the invariants h, IIr, IIh for the stress tensor

u,; (-~ -~ D

Determine the principal stress values for this state of stress and show that the diagonal form of the stress tensor yields the same values for the stress invariants.

From (2.39), II = Uii = 6 + 6 + 8 = 20. From (2.40), III = (1I2)(uiiujj - UijUij)

= 0'110'22 + 0'220'33 + 0'330'11 - 0'120'12 - 0'230'23 - 0'310'31

= 36 + 48 + 48 - 9 = 123.

From (2.41), IIII = 10'01 = 6(48) + 3(-24) = 216.

The principal stress values of Uij are 0'1 = 3, un = 8, Urn = 9. In terms of principal values,

II ur + un + um = 3 + 8 + 9 = 20



CHAP. 2]



2.22. The octahedral plane is the plane which makes equal angles with the principal stress directions. Show that the shear stress on this plane, the so-called octahedral shear

stress, is given by

With respect to the principal axes, the normal to the octahedral plane is given by

A 1 A A A

n = v's ( er + e2 + es)

Hence from (2.12) the stress vector on the octahedral plane is

A 1 A A A

t(n) = _ In ( er + e2 + es)


• (O'rerer + O'lle2e2 + O'IIIe3e3)

1 (A A A

v's O'rer + O'lle2 + O'IIIeS)

and its normal component is

- A. t(~) - 1 ( )

UN - n - 11 O'r + O'Il + vm

Therefore the shear component is

Fig. 2-25

_ I A A 2

"'J t(n). t(n) - «» = {i(O'; + O';r + O';ll) - ~(O'r + O'll + O'III)2}112

i{3(0'; + O';r + O';ll) - (0'; + O'i1 + O'ill + 20'rO'll + 20'llO'III + 20'IIIO'r)}1I2 i{(O'i - 20'1O'II + O'irl + (O'i1 - 20'llO'III + O'ill) + (O'iu - 20'IIIO'r + 0';)}1I2


2.23. The stress tensor at a point is given by O'ij = (~ -~ -1~). Determine the maxio -12 1

mum shear stress at the point and show that it acts in the plane which bisects the maximum and minimum stress planes.

From (2.38) the reader should verify that the principal stresses are 0'1 = 10, <n = 5, O'III = -15. From (2.54b) the maximum shear value is Us = (O'III - O'r)/2 = -12.5. The principal

axes Ox! x; x; are related to the axes of maximum shear Ox~ xzxs by the transformation table below and are situated as shown in Fig. ::-26.

x* x* x*
1 2 a
xi 1/12 0 1112
Xz 0 1 0
x' -1112 0 1IYz
a x* O'Il 2

Fig. 2-26

The stress tensor referred to the primed axes is thus given by
[ 1IV' 0 lI~r: 0 ']["V' o -1IV'] [-2.' 0 -12.']
[u[j] = -1/~ 1 5 -1~ 1/~ 1 0 .0 5 0
0 1/V2 0 0 o 1IV2 -12.5 0 -2.5 The results here may be further clarified by showing the stresses on infinitesimal cubes at the point whose sides are perpendicular to the coordinate axes (see Fig. 2-27).


x* 2

x* I

Fig. 2-27


MOHR'S CIRCLES (Sec. 2.12-2.13)

2.24. Draw the Mohr's circles for the state of stress discussed in Problem 2.23. Label important points. Relate the axes OXIX2X3 (conjugate to uij) to the principal axes Oxt x: x~ and locate on the Mohr's diagram the points giving the stress states on the coordinate planes of OXIX2X3.

The upper half of the symmetric Mohr's circles diagram is shown in Fig. 2-28 with the maximum shear point P and principal stresses labeled. The transformation table of direction cosines is

xi x; *
XI 0 1 0
X2 -3/5 0 4/5
X3 4/5 0 3/5 from which a diagram of the relative orientation of the axes is made as shown in Fig. 2-29. The XI and x; axes are coincident. X2 and X3 are in the plane of x~ x; as shown. From the angles IX = 36.8° and f3 = 53.2° shown, the points A(-6,12) on the plane .L to x2 and B(l, 12) on the plane .L to X3 are located. Point C(5, 0) represents the stress state on the plane .L to Xl'

Fig. 2-28

Fig. 2-29

CHAP. 2]



2.25. The state of stress at a point referred to axes OXlX2X3 is given by

(1ij (-~ -~ -1~)

o -12 1

Determine analytically the stress vector components on the plane whose unit normal is n = (2/3)el + (l/3)e2 + (2/3)e3. Check the results by the Mohr's diagram for this problem.

From (2.13) and the symmetry property of the stress tensor, the stress vector on the plane of n is given by the matrix product

[-: _~ -1~] [:~:] [=~~/3]

o -12 1 2/3 -10/3

Thus t(~) -10 ~l/3 - 10 ~2 - 10 ~3/3; and from (2.33), UN = t(~) • it = -70/9. From (2.47),

Us = 70.7/9.

For (lij the principal stress values are U1 = 10, Un = -5, UIll = -15; and the principal axes are related to OXlX2X3 by the table

Xl X2 X3
X* 0 -3/5 4/5
x* 1 0 0
X* 0 4/5 3/5
3 Thus in the principal axes frame, nt = aijnj or [~-3~5 4~5] [:~:] = [~~: l. Accordingly

o 4/5 3/5 2/3 2/3 J

the angles of Fig. 2-16 are given by (J = f3 = cos-12/3 = 48.2° and ¢ = cos-l1/3 = 70.5°, and the Mohr's diagram comparable to Fig. 2-17 is as shown in Fig. 2-30.

Fig. 2-30



[CHAP. 2

2.26. Sketch the Mohr's circles for the three cases of plane stress depicted by the stresses on the small cube oriented along the coordinate axes shown in Fig. 2-31. Determine the maximum shear stress in each case.


(b) Fig. 2-31

The Mohr's circles are shown in Fig. 2-32.

(US)max = U



Fig. 2-32




2.27. Split the stress tensor au = (1~ : -~) into its spherical and deviator parts and o -2 3

show that the first invariant of the deviator is zero.

17M = "kk/3 = (12 + 9 + 3)/3 = 8. Thus

and 8ii = 4 + 1 -·5 = O.

(~ ~ ~) + (: : -~)

o 0 8 0 -2 -5

2.28. Show that the deviator stress tensor is equivalent to the superposition of five simple shear states.

The decomposition is

C 812 ''') (,;. 812 0) (' 0 .,,) (' 0 ,;,)
821 822 823 0 o + 0 0 o + 0 0
831 832 833 0 o 831 0 o 0 832
('" 0 ~) (: 0 ,D
+ 0 -811 + -833
0 0 0 where the last two tensors are seen to be equivalent to simple shear states by comparison of cases (a) and (b) in Problem 2.26. Also note that since 8;; = 0, -811 - S33 = S22'

2.29. Determine the principal deviator stress values for the stress tensor

The deviator of uii is 8ii from the determinant

and its principal values may be determined

-6 o

3-8 o


o -6-8

(-6 - 8)(S + 3)(s - 9)


3-s -6

Thus 8r = 9, 8n = -3, Sm = -6. The same result is obtained by first calculating the principal stress values of uij and then using (2.71). For Uij, as the reader should show, ur = 16, Un = 4, Urn = 1 and hence Sr = 16 - 7 = 9, Sn = 4 -7 = -3, 8m = 1 - 7 = -6.

2.30. Show that the second invariant of the stress deviator is given in terms of its principal stress values by IIl:D = (SrSn + SuSnr + SmSr), or by the alternative form IhD = -!(si + sir + sin).

In terms of the principal deviator stresses the characteristic equation of the deviator stress tensor is given by the determinant


o o

(8r - s)(sn - 8)(Sm - s)


o o

Su - 8


Sm - 8

Hence from (2.72), IIl:D = (SrBn + 8n8m + 8mSr). Since 8r + 8n + 8m = 0,


2.31. Prove that for any symmetric tensor such as the stress tensor O'ij' the transformed tensor O'~. in any other coordinate system is also symmetric.




[CHAP. 2

2.32. At the point P the principal stresses are such that 2(Tu = (T1 + (Tm' Determine the unit normal n; for the plane upon which (TN = "n and as = ((Tr - (Trn)l4.

From (2.33), (1N = ni(1r + n~«(1r + (1IlI)/2 + n~(1III = «(1r + (1m)/2; and since ni + n~ + n~ = 1, these equations may be combined to yield n1 = n3' Next from (2.47),

(1~ = ni(1i + n~«(1r + (1rn)2/4 + n~(1in - «(1r + (1rn)2/4 = «(1r - (1rn)2/16

Substituting n1 = n3 and n~ - 1 = -ni - n~ = -2ni into this equation and solving for nv the direction cosines are found to be n1 = 1/2V2, n2 = V3/2, n3 = 1/2V2. The reader should apply these

results to the stres ~n,",." (:: D


Show that the stress tensor (T .. may be decomposed into a spherical and deviatoric


part in one and only one way.

Assume two decompositions, (1ij = 8ij~ + Sij = 8ij~* + srj with Sii = 0 and S~i = O. Then (1ii = 3~ = 3~ *, so A = ~ *; and from AIlij + Sij = A8ij + srj it follows that Sij = s0'

2.34. Prove that the principal stress values are real if I is real and symmetric.

For real values of the stress components the stress invariants are real and hence the coefficients in (2.38) are all real. Thus by the theory of equations one root (principal stress) is real. Let (1(3) be this root and consider a set of primed axes xi of which x~ is in the direction of U(3)' With respect to

such axes the characteristic equation is given by the determinant

o o

= 0

o 0 (1(3)-(1

or «(1(3) - e) [«(1~1 - (1)«(1~2 - (1) - (1~~1 = O. Since the discriminant of the quadratic in square brackets D = «(1;1 + (122)2 - 4[(1;1(1~2 - «(1;2)2] = «(1;1 - (122)2 + 4(1;~ > 0, the remaining roots must be real.

2.35. Use the method of Lagrangian multipliers to show that the extremal values (maximum and minimum) of the normal stress (TN correspond to principal values.

From (2.33), (1N = (1ijninj with n;ni = 1. Thus in analogy with (2.51) construct the function R == (1N - ~nini for which oR/oni = O. Then

(1ijni,pnj + (1ijninj,p - 2~ni,pni (1ij8ipnj + (1ijni8jp - 2A8ipni

(1pjnj + (1ipni - 2A8ipni = 2«(1pi - ~8ip)ni == 0 which is equivalent to equation (2.36) defining principal stress directions.


Assume that the stress components (T .. are derivable from a symmetric tensor field


A. .. by the relationship a .. == £. e. A. • Show that in the absence of body forces the

'l"'tJ 13 Ipq rmn't'qn,pm

equilibrium equations (2.23) are satisfied.

Using the results of Problem 1.58, the stress components are given by

or explicitly

au = 8ij(9)qq,W - 9>qp,qp) + 9>pi,pj + <Pjp,pi - <Ppp,ji - <Pji,pp
(111 <P33,22 + 9>22,33 (112 (121 -9>33,21
(122 9>11,33 + 9>33,11 (123 (132 -9>11,23
(133 9>22,11 + CP11,22 (131 (113 -CP22,13 CHAP. 2]



, Substituting these values into l1ij,j = 0, 1111,1 + 1112,2 + 1113,3 1121, 1 + 1122,2 + 1123,3 1131,1 + 1132,2 + 1133,3

<1>33,221 + <1>22,331 - <1>33,212 - <1>22,133 -<1>33,211 + <1>11,332 + <1>33,112 - <1>11,233

-<1>22,131 - <1>11,232 + ¢22,113 + <1>11,223

o o o

2.37. Show that, as is asserted in Section 2.9, the normal to the Cauchy stress quadric at the point whose position vector is r is parallel to the stress vector ti~) .

Let the quadric surface be given in the form <I> = l1ijKiKj ± k2 = O. The normal at any point is then V ¢ or o¢/dKi = ¢,i. Hence ¢,p = l1ijllipij + l1ijiilljp = 2l1piii. Now since Ki = rni' this becomes

211pirni or 2r(l1pini) = 2t~~).

2.38. At a point P the stress tensor referred to axes OX1X2X3 is given by

( 15 -10 0)

CTij -10 5 0

o 0 20

If new axes Ox; x~x; are chosen by a rotation about the origin for which the transformation matrix is [au] = 13~5 ~ -4~5l, determine the traction vectors on each

l4/5 0 315 J

of the primed coordinate planes by projecting the traction vectors of the original axes onto the primed directions. In this way determine CTU. Check the result by the transformation formula (2.27).

From (2.6) and the identity ti~i) = l1ij (2.7), the traction vectors on the unprimed coordinate


axes are tee!) = 15~1 - 10~2' t(e.) = -10~1 + 5~2' t(e3) = 20~3 which correspond to the rows

of the stress tensor. Projecting these vectors onto the primed axes by the vector form of (2.12), t(~) = n1t<~!) + n2t(~.) + n3t(~3), gives

t(~;) = !(15~1 - 10~) - t(20~3) = 9~1 - 6~2 - 16~3 which by the transformation of the unit vectors becomes


- 9(3AI + 4AI) 6AI

- ije1 5e3 - e2



-6~~ + 5~; 8~;

-12~~/5 - 8~; + 84~;/5

I O'ij

I O'ij

-6 -12/5) 5 -8

-8 84/5

so that

By (2.27),

o 1 o

[ ~~5 -:

-12/5 -8






[CHAP. 2

2.39. Show that the second invariant of the deviator stress tensor IhD is related to the octahedral shear stress by the equation UOCT = V-iIIID•

From Problem 2.22, aOCT = !V(ar - au)2 + (au - am)2 + (am - ar)2, and because ar = aM + S1> an = aM + Srr, etc.,

aOCT 1V(sr - srr)2 + (su - sm)2 + (sm - sr)2

! 2(s; + s;r + sirr) - 2(srsrr + SuSm + smsr)

Also Sr + Srr + sm

o and so (sr + Su + sm)2 = 0 or

s; + sir + siu = -2(srsn + SrrSm + smsr)


2.40. The state of stress throughout a body is given bythe stress tensor

·u (+ _:~: -+)

where C is an arbitrary constant. (a) Show that the equilibrium equations are satisfied if body forces are zero. (b) At the point P(4, -4, 7) calculate the stress vector on the plane 2X1 + 2X2 - X3 = -7, and on the sphere (X1)2 + (X2)2 + (X3)2 = (9)2. (c) Determine the principal stresses, maximum shear stresses and principal deviator stresses at P. (d) Sketch the Mohr's circles for the state of stress at P.

(a) Substituting directly into (2.24) from aii' the equilibrium equations are satisfied identically.

(b) From Problem 1.2, the unit normal to the plane 2X1 + 2X2 - x3 = -7 is ~ = i ~1 + i ~2 - 1 ~3' Thus from (2.12) the stress vector on the plane at P is

t(~) (i ~1 + i ~2 - i- ~3) • (7C~1~2 + 7C~2~1 - 4C~2~3 - 4C~3~2)

C(14A +14A 8A +4A) lC(14A +18A 8A) "3 e2 "3 e1 - :3 e3 :3 e2 = IT e1 e2 - e3

The normal to the sphere Xi Xi = (9)2 at P is ni = ¢.i with ¢ = Xixi - 81, t~2 + ~~3' In the matrix form of (2.14) the stress vector at P is

A or n

[0 7C 0]

[4/9, -4/9, 719] 7C 0 -4C

o -4C 0

[-28C/9, 0, 16C/9]

(c) From (2.37), for principal stresses a, -a 7 0

"I = v'65

7 -a -4 o -4 -a

ar = {65, arr = 0, aru = -{65. The maximum shear stress value is given by (2.54b) as as = (am - arl/2 = ±y65. Since the mean normal stress at P is aM = (ar + an + alII )/3 = 0, the principal deviator stresses are the same as the principal stresses.

a(a2 - 65) = 0; hence

(d) The Mohr's circles are shown in Fig. 2-33.

Fig. 2-33

CHAP. 2]



Supplementary Problems

2.41. At point P the stress tensor is Uij ( 1~ 2: -:). Determine the stress vector on the

-7 ° 35

plane at P parallel to plane (a) BGE, (b) BGFC of the small parallelepiped shown in Fig. 2-34.

Fig. 2-34

2.42. Determine the normal and shear stress components on the plane BGFC of Problem 2.41.

Ans. UN = 63/5, Us = 37.7/5

2.43. The principal stresses at point Pare ur = 12, Un = 3, UlII = -6. Determine the stress vector and its normal component on the octahedral plane at P.

Ans. t(~) = (12~1 + 3~2 - 6~3)/Y3, UN = 3

2.44. Determine the principal stress values for

(a) Uij


(b) uij

and show that both have the same principal directions. Ans. (a) ur = 2, Un = Urn = -1, (b) Ur = 4, Un = Unr = 1


Decompose the stress tensor Uij =



30 -27


into its spherical and deviator parts and

determine the principal deviator stresses.

Ans. sr = 31, Sn = 8, snr = -39

2.46. Show that the normal component of the stress vector on the octahedral plane is equal to one third the first invariant of the stress tensor.

(2°1 Ul122 021)

2.47. The stress tensor at a point is given as uij = with U22 unspecified. Determine U22

so that the stress vector on some plane at the point will be zero. Give the unit normal for this traction-free plane.

A A A A )/_ r;;

Ans. U22 = 1, n = (el - 2 e2 + e3 Y 6



[CHAP. 2

2.48. Sketch the Mohr's circles and determine the maximum shear stress for each of the following

stress states:

(a) CTij

(b) CTij

Ans. (a) CTS = 7, (b) CTS = 37/2

2.49. Use the result given in Problem 1.58, page 39, together with the stress transformation law (2.27), page 50, to show that EijkEpqmCTipCTjqCTkm is an invariant.

2.50. In a continuum, the stress field is given by the tensor

( XiX2 (1- X;)Xl 0)

CTij (1 - :; )xl (x~ - 03X2)/3 2:~

Determine (a) the body force distribution if the equilibrium equations are to be satisfied throughout the field, (b) the principal stress values at the point P(a,0,2Va), (c) the maximum shear stress at P, (d) the principal deviator stresses at P.

Ans. (a) b3 = -4X3' (b) a, -a, 8a, (c) ±4.5a, (d) -lla/3, -5a/3, 16a/3

Chapter 3

Deformation and Strain


In the kinematics of continua, the meaning of the word "point" must be clearly understood since it may be construed to refer either to a "point" in space, or to a "point" of a continuum. To avoid misunderstanding, the term "point" __ will. beuse._d.exclll~j_vely to designate alocation in fixed space. The word "particle" will denote a small volumetric "element, or "material point", of a continuum. In brief, a point -is a·-p-lace in space,a particl.e

is a small part of a material continuum. .


At any instant of time t, a continuum having a volume V and bounding surface S will occupy a certain region R of physical space. The identification of the particles of the continuum with the points of the space it occupies at time t by reference to a suitable set of coordinate axes is said to specify the configuration of the continuum at that instant.

The term deformation refers to a change in the shape of the continuum between sOllle initial (undeformed) configuration and a subsequent (deformed) configuration. The emphasis in deformation studies is on the initial and final configurations. No attention is given to intermedfate configuraUolls or ·to-theParficula:rsequenceofConftguration-sby-whlch-The deformaIion·occurs~Y cOntrast, the. word ftowlSused. to-desrgnate-ihe-continui~gstate -of motion-of a continuum. Indeed, a configuration liistory- isilihereiifill-fiow-investigati~~s for which the specification of a time-dependent velocity field is given.


In Fig. 3-1 the undeformed configuration of a material continuum at time t = 0 is shown together with the deformed configuration of the same continuum at a later time t = t. For the present development it is useful to refer the initial and final configurations to separate coordinate axes as in the figure.

Fig. 3-1




[CHAP. 3

Accordingly, in the initial configuration a representative particle of the continuum occupies a point Po in space and has the position vector


with respect to the rectangular Cartesian axes OX1X2X3• Upper-case letters are used as indices in (3.1) and will appear as such in several equations that follow, but their use as summation indices is restricted to this section. In the remainder of the book upper-case subscripts or superscripts serve as labels only. Their use here is to emphasize the connection of certain expressions with the coordinates (X1X2X3), which are called the material coordinates. In the deformed configuration the particle originally at Po is located at the point P and has the position vector


when referred to the rectangular Cartesian axes OXIX2X3. Here lower-case letters are used as subscripts to identify with the coordinates (XIX2X3) which give the current position of the particle and are frequently called the spatial coordinates.

The relative orientation of the material axes OX1X2X3 and the spatial axes OXIX2X3 is specified through direction cosines CXkK and CXKk' which are defined by the dot products of unit vectors as


No summation is implied by the indices in these expressions since k and K are distinct


indices. Inasmuch as Kronecker deltas are designated by the equations h olp = 8KP and

Ck 0 cp = 8kp, the orthogo.nality conditions between spatial and material axes take the form


In Fig. 3-1 the vector u joining the points Po and P (the initial and final positions, respectively, of the particle), is known as the displacement vector. This vector may be expressed as


or alternatively as


in which the components UK and Uk are interrelated through the direction cosines CXkK• From (1.89) the unit vector Ck is expressed in terms of the material base vectors IK as

Therefore substituting (3.7) into (3.5),



from which


Since the direction cosines CXkK are constants, the components of the displacement vector are observed from (3.9) to obey the law of transformation of first-order Cartesian tensors, as they should.

The vector b in Fig. 3-1 serves to locate the origin 0 with respect to O. From the geometry of the figure,

u = b + x - X (3.10)

CHAP. 3]



Very often in continuum mechanics it is possible to consider the coordinate systems OXIX2X3 and OXIX2X3 superimposed, with b == 0, so that (3.10) becomes

u = x-x

In Cartesian component form this equation is given by the general expression



However, for superimposed axes the unit triads of base vectors for the two systems are identical, which results in the direction cosine symbols ll'kK becoming Kronecker deltas. Accordingly, (3.12) reduces to


in which only lower-case subscripts appear. In the remainder of this book, unless specifically stated otherwise, the material and spatial axes are assumed superimposed and hence only lower-case indices will be used.


When a continuum undergoes deformation (or flow), the particles of the continuum move along various paths in space. This motion may be expressed by equations of the form

Xi = Xi(XI, X2, X3, t) = Xi(X, t) or x = x(X, t)


which give the present location Xi of the particle that occupied the point (XIXzX3) at time t = o. Also, (3.14) may be interpreted as a mapping of the initial configuration into the current configuration. It is assumed that such a mapping is one-to-one and continuous, with continuous partial derivatives to whatever order is required. The description of motion or deformation expressed by (3.14) is known as the Lagrangian formulation.

If, on the other hand, the motion or deformation is given through equations of the form

Xi = X(XI, X2, X3, t) = Xi(x, t) or X = X(x, t)


in which the independent variables are the coordinates Xi and t, the description is known as the Eulerian formulation. This description may be viewed as one which provides a tracing to its original position of the particle that now occupies the location (Xl, X2, X3). If (3.15) is a continuous one-to-one mapping with continuous partial derivatives, as was also assumed for (3.14), the two mappings are the unique inverses of one another. A necessary and sufficient condition for the inverse functions to exist is that the Jacobian



should not vanish.

As a simple example, the Lagrangian description given by the equations

Xl Xl + X2(et -1)


has the inverse Eulerian formulation, Xl

-XI + x2(et -1) 1 - et - e t

xl(e-t -1) - X2 1 - et - e t




[CHAP. 3


Partial differentiation of (3.14) with respect to X, produces the tensor axi/aXj which is called the material deformation gradient. In symbolic notation, axi/aXj is represented by the dyadic




in which the differential operator V'x = a~i ei is applied from the right (as shown explicitly in the equation). The matrix form of F serves to further clarify this property of the operator V' x when it appears as the consequent of a dyad. Thus

[X ] [aXdaXl axdaX2 aXdaX3]

l' = XX3~ [a~l' a~2' a~3J aX2/aXl aX2/aX2 aX2/aX3

aX3/aXl aX3/aX2 aX3/aX3

Partial differentiation of (3.15) with respect to Xj produces the tensor aXi/aXj which is called the spatial deformation gradient. This tensor is represented by the dyadic



having a matrix form

aXdaX2 aXdaX3] aXz/axz aXZ/aX3 aX3!aX2 aX3/aX3

The material and spatial deformation tensors are interrelated through the well-known chain rule for partial differentiation,

aXi «s, aXj aXk

aXi aXj aXj aXk


Partial differentiation of the displacement vector u, with respect to the coordinates produces either the material displacement gradient auJaXj, or the spatial displacement gradient aui/aXj. From (3.13), which expresses u; as a difference of coordinates, these tensors are given in terms of the deformation gradients as the material gradient

aUi aXj

and the spatial gradient


aXj - 8ij or J == u V' x = F - I



8·· - - or K == u V' x = I - H

lJ aXj


In the usual manner, the matrix forms of J and K are respectively
[::] [ai;, a~' a~,J [aUdaX, audaXz iJUdaX']
!J aU2/aXl aU2/aX2 aU2/aX3
U3 aU3/aXl aU3/aX2 aU3/aX3
[:: ] [a~, 'a:,' ,:,J ['Udax, aUdaX2 au,/ax,]
'K._ aU2/aXl aUZ/aX2 aU2/aX3
U3 aU3/aXl aU3/aX2 aU3/aX3 (3.27)

CHAP. 3]




In Fig. 3-2 the initial (undeformed) and final (deformed) configurations of a continuum are referred to the superposed rectangular Cartesian coordinate axes OX1X2X3 and OXIX2X3. The neighboring particles which occupy points Po and Qo before deformation, move to points P and Q respectively in the deformed configuration. _

Fig. 3-2

The square of the differential element of length between Po and Qo is (dX)2 = ax- dX = as, ex, = au as. ax,

From (3.15), the distance differential tiX, is seen to be



dX = H·dx


so that the squared length (dX)2 in (3.28) may be written (dX)2 = aXkaXkdxidxj = Cijdxidxj aXi aXj



dx'C' dx


in which the second-order tensor




is known as Cauchy's deformation tensor.

In the deformed configuration, the square of the differential element of length between P and Q is


From (3.14) the distance differential here is


dx, = aXj d.X,

so that the squared length (dX)2 in (3.32) may be written


dx = F·dX


(d )2 -_ aXk aXk ax, dX· G dX dX

x aXi aXj 1 J = ij i j

in which the second-order tensor





aXk aXk

Gij = X -X--- or a i a j

is known as Green's deformation tensor.





[CHAP. 3

The difference (dX)2 - (dX)2 for two neighboring particles of a continuum is used as the measure of deformation that occurs in the neighborhood of the particles between the initial and final configurations. If this difference is identically zero for all neighboring particles of a continuum, a rigid displacement is said to occur. Using (3.34) and (3.28), this difference may be expressed in the form

d )2 2 (aXk aXk )

(X - (dX) = aXi aXj - Sij d.X, dXj = 2Lij dXi d.X,



in which the second-order tensor



is called the Lagrangian (or Green's) finite strain tensor.

Using (3.32) and (3.30), the same difference may be expressed in the form

( «x, aXk)

(dX)2 - (dX)2 = S .. - -- dx· dx· = 2E .. dx· dx·

'J aXi aXj 'J 'J' J


(dX)2 - (dX)2 = dx·(I-Hc·H)·dx = dx·2EA·dx


in which the second-order tensor

g. _ _!_ (s .. _ aXk aXk)

'J - 2 'J aXi aXj



is called the Eulerian (or Almansi's) finite strain tensor.

An especially useful form of the Lagrangian and Eulerian finite strain tensors is that in which these tensors appear as functions of the displacement gradients. Thus ifax;/aXj from (3.24) is substituted into (3.37), the result after some simple algebraic manipulations is the Lagrangian finite strain tensor in the form

t.; = ! (aui + aUj + aUk aUk) 2 aXj ex, ex, es,

In the same manner, ifaXdaxj from (3.25) Eulerian finite strain tensor in the form

Eij = !. (aUi + aUj _ aUk aUk)

2 aXj e». aXi aXj



is substituted into (3.39), the result is the


EA = t(K + Kc - Kc· K)


The matrix representations of (3.40) and (3.41) may be written directly from (3.26) and (3.27) respectively.


The so-called small deformation theory of continuum mechanics has as its basic condition the requirement that the displacement gradients be small compared to unity. The fundamental measure of deformation is the difference (dX)2 - (dX)2, which may be expressed in terms of the displacement gradients by inserting (3.40) and (3.41) into (3.36) and (3.38) respectively. If the displacement gradients are small, the finite strain tensors in (3.36) and (3.38) reduce to infinitesimal strain tensors, and the resulting equations represent small deformations.

CHAP. 3]



In (3.40), if the displacement gradient components au/aXj are each small compared to unity, the product terms are negligible and may be dropped. The resulting tensor is the Lagrangian infinitesimal strain tensor, which is denoted by


L = i(uVx + Vx u) = !(J +Jc)


Likewise for au/aXj ~ 1 in (3.41), the product terms may be dropped to yield the Eulerian infinitesimal strain tensor, which is denoted by


If both the displacement gradients and the displacements themselves are small, there is very little difference in the material and spatial coordinates of a continuum particle. Accordingly the material gradient components au/axj and spatial gradient components au/aXj are very nearly equal, so that the Eulerian and Lagrangian infinitesimal strain tensors may be taken as equal. Thus


L = E


if both the displacements and displacement gradients are sufficiently small.



In Fig. 3-3 the displacements of two neighboring

particles are represented by the vectors ujPo) and u;Qo) (see also Fig. 3-2). The vector

du = u(Qo) - u(Po)


is called the relative displacement vector of the particle originally at Qo with respect to the particle originally at Po. Assuming suitable continuity conditions on the displacement field, a Taylor series expansion for uipo) may be developed in the neighborhood of Po. Neglecting higher-order terms in this expansion, the relative displacement vector can be written as


Fig. 3-3



Here the parentheses on the partial derivatives are to emphasize the requirement that the derivatives are to be evaluated at point Po. These derivatives are actually the components of the material displacement gradient. Equation (3.46) is the Lagrangian form of the relative displacement vector.

It is also useful to define the unit relative displacement vector duJdX in which dX is the magnitude of the differential distance vector dXi. Accordingly if Vi is a unit vector in the direction of dXi so that d.X, = Vi dX, then

du; aUi d.X, au;

dX = aXj dX = ax~ Vj


du dX


Since the material displacement gradient auJaXj may be decomposed uniquely into a symmetric and an anti symmetric part, the relative displacement vector du, may be written as



[CHAP. 3


[! (aUi + aUj) + ! (aUi _ aUj)] ax,

2 aXj aXi 2 aXj aXi J

[!(uVx + Vx u) + !(uVx - Vx u)] ° dX




The first term in the square brackets in (3.48) is recognized as the linear Lagrangian strain tensor l;j. The second term is known as the linear Lagrangian rotation. tensor and is denoted by



In a displacement for which the strain tensor l;j is identically zero in the vicinity of point Po, the relative displacement at that point will be an infinitesimal rigid body rotation. This infinitesimal rotation may be represented by the rotation vector


w = !Vx Xu


in terms of which the relative displacement is given by the expression


du = w x dX


The development of the Lagrangian description of the relative displacement vector, the linear rotation tensor and the linear rotation vector is paralleled completely by an analogous development for the Eulerian counterparts of these quantities. Accordingly the Eulerian description of the relative displacement vector is given by


du = Kodx


and the unit relative displacement vector by du; = au~ dXj = aUi /L.

aXj dx aXj.1


du A A - = uV 0IA- = KOlA-

dx x


Decomposition of the Eulerian displacement gradient au;/aXj results in the expression


du, dx



The first term in the square brackets of (3.54) is the Eulerian linear strain tensor (ij' The second term is the linear Eulerian rotation tensor and is denoted by



From (3.55), the linea?" Eulerian rotation vector is defined by



in terms of which the relative displacement is given by the expression


du = (I) x dx



For small deformation theory, the finite Lagrangian strain tensor Lij in (3.36) may be replaced by the linear Lagrangian strain tensor lij, and that expression may now be written

CHAP. 3]




(dX)2 - (dXF (dX)2 - (dX)2

(dx - dX)(dx + dX) (dx - dX)(dx + dX)



Since dx "'" dX for small deformations, this equation may be put into the form

The left-hand side of (3.59) is recognized as the change in length per unit original length of the differential element and is called the normal strain for the line element originally having direction cosines dX;/dX.

When (3.59) is applied to the differential line element PoQo, located with respect to the set of local axes at Po as shown in Fig. 3-4, the result will be the normal strain for that element. Because PoQo here lies along the X2 axis,

dXt/dX = dX3/dX = 0, dX2/dX 1 and therefore (3.59) becomes

dx - dX - 1

dX - 22




Fig. 3-4

Thus the normal strain for an element originally along the X2 axis is seen to be the component l22. Likewise for elements originally situated along the Xl and X3 axes, (3.59) yields normal strain values ll1 and l33 respectively. In general, therefore, the diagonal terms of the linear strain tensor represent normal strains in the coordinate directions.

Fig. 3-5

The physical interpretation of the off-diagonal terms of lij may be obtained by a consideration of the line elements originally located along two of the coordinate axes. In Fig. 3-5 the line elements PoQo and PoMo originally along the Xz and X3 axes, respectively, become after deformation the line elements PQ and PM with respect to the parallel set of local axes with origin at P. The original right angle between the line elements becomes the angle (j. From (3.46) and the assumption of small deformation theory, a first order approximation gives the unit vector at P in the direction of Q as



[CHAP. 3

A aU! A A aU3 A

n2 = aX2 ei + e2 + aX2 es

and, for the unit vector at P in the direction of M, as





or, neglecting the product term which is of higher order,

aU2 aU3

cos B = aX3 + aX2 = 2lza


Furthermore, taking the change in the right angle between the elements as Y23 = -rr/2 - B, and remembering that for the linear theory Y23 is very small, it follows that

Y23 = sin Y23 = sin (-rr/2 - B) = cos B = 2123


Therefore the off-diagonal terms of the linear strain tensor represent one-half the angle change between two line elements originally at right angles to one another. These strain components are called shearing strains, and because of the factor 2 in (3.65) these tensor components are equal to one-half the familiar "engineering" shearing strains.

A development, essentially paralleling the one just presented for the interpretation of the components of l;j, may also be made for the linear Eulerian strain tensor £ij. The essential difference in the derivations rests in the choice of line elements, which in the Eulerian description must be those that lie along the coordinate axes after deformation. The diagonal terms of £ij are the normal strains, and the off-diagonal terms the shearing strains. For those deformations in which the assumption l. = e., is valid, no distinction

11 tJ

is made between the Eulerian and Lagrangian interpretations.


An important measure of the extensional strain of a differential line element is the ratio dx/dX, known as the stretch or stretch ratio. This quantity may be defined at either the point Po in the un deformed configuration or at the point P in the deformed configuration. Thus from (3.34) the squared stretch at point Po for the line element along the unit vector m = dX/dX, is given by


Similarly, from (3.30) the reciprocal of the squared stretch for the line element at P along the unit vector Ii = dx/dx is given by

(dX)2 .. _1_ _ .. dXi dXj 1 A A

- = 2 - G'l or -2- = n· C . n

dx p A(~) dx dx A(~)

For an element originally along the local X2 axis shown in Fig. 3-4, m = e2 and therefore dXIIdX = dX3/dX = 0, dX2/dX = 1 so that (3.66) yields for such an element


A2(~2) = G22 = 1 + 2L22 Similar results may be determined for A~~,) and A~~3).


CHAP. 3]



For an element parallel to the X2 axis after deformation, (3.67) yields the result


with similar expressions for the quantities l!A~~l) and l!A~~3)' In general, A(~2) is not equal to A(~2) since the element originally along the X2 axis will not likely lie along the X2 axis after deformation.

The stretch ratio provides a basis for interpretation of the finite strain tensors. Thus the change of length per unit of original length is

dx - dX dX

= dx _ 1 dX


and for the element PoQo along the Xz axis (of Fig. 3-4), the unit extension is therefore


This result may also be derived directly from (3.36). For small deformation theory, (3.71) reduces to (3.60). Also, the unit extensions Lei, and L(3) are given by analogous equations in terms of L11 and L33 respectively.

For the two differential line elements shown in Fig. 3-5, the change in angle YZ3 = -rr/2 - () is given in terms of A (~2) and A (~3) by

2L23 A(~2)A(~3)


When deformations are small, (3.72) reduces to (3.65).


The so-called polar decomposition of an arbitrary, nonsingular, second-order tensor is given by the product of a positive symmetric second-order tensor with an orthogonal secondorder tensor. When such a multiplicative decomposition is applied to the deformation gradient F, the result may be written


F = R'S = T·R


in which R is the orthogonal rotation tensor, and Sand T are positive symmetric tensors known as the right stretch tensor and left stretch tensor respectively.

The interpretation of (3.73) is provided through the relationship da; = (iJxdiJXj) dXj given by (3.33). Inserting the inner products of (3.73) into (3.33) results in the equations


dx = R' S • dX = T' R' dX


From these expressions the deformation of dXi into dx, as illustrated in Fig. 3-2 may be given either of two physical interpretations. In the first form of the right hand side of (3.74), the deformation consists of a sequential stretching (by S) and rotation to be followed by a rigid body displacement to the point P. In the second form, a rigid body translation to P is followed by a rotation and finally the stretching (by T). The translation, of course, does not alter the vector components relative to the axes Xi and Xi.



[CHAP. 3


The various strain tensors Lij, Eij, lij and fij defined respectively by (3.37), (3.39), (3.42) and (3.43) are all second-order Cartesian tensors as indicated by the two free indices in each. Accordingly for a set of rotated axes X: having the transformation matrix [bij] with respect to the set of local un primed axes Xi at point Po as shown in Fig. 3-6(a), the components of L:j and lij are given by

or L~

B' LG' Be

(3.75) (3.76)


or L'



Fig. 3-6

Likewise, for the rotated axes x[ having the transformation matrix [aij] in Fig. 3-6(b), the components of s; and e[j are given by




E' = A' E'Ae

(3.77) (3.78)

By analogy with the stress quadric described in Section 2.9, page 50, the Lagrangian and Eulerian linear strain quadrics may be given with reference to local Cartesian coordinates 'f}i and 'i at the points Po and P respectively as shown in Fig. 3-7. Thus the equation of the Lagrangian strain quadric is given by



Fig. 3-7

CHAP. 3]



and the equation of the Eulerian strain quadric is given by



Two important properties of the Lagrangian {Eulerian} linear strain quadric are:

1. The normal strain with respect to the original {final} length of a line element is inversely proportional to the distance squared from the origin of the quadric Po {P} to a point on its surface.

2. The relative displacement of the neighboring particle located at Qo {Q} per unit original {final} length is parallel to the normal of the quadric surface at the point of intersection with the line through PoQo {PQ}.

Additional insight into the nature of local deformations in the neighborhood of Po is provided by defining the strain ellipsoid at that point. Thus for the undeformed continuum, the equation of the bounding surface of an infinitesimal sphere of radius R is given in terms of local material coordinates by (3.28) as


(dX)2 = dX·I·dX = R2


After deformation, the equation of the surface of the same material particles is given by (3.30) as


(dX)2 = dx : C • dx = R2


which describes an ellipsoid, known as the material strain ellipsoid. Therefore a spherical volume of the continuum in the undeformed state is changed into an ellipsoid at Po by the deformation. By comparison, an infinitesimal spherical volume at P in the deformed continuum began as an ellipsoidal volume element in the un deformed state. For a sphere of radius r at P, the equations for these surfaces in terms of local coordinates are given by (3.32) for the sphere as

(dX)2 = 8ijdxidxj =r or (dX)2 = dx '1' dx = r2 (3.83)
and by (3.34) for the ellipsoid as
(dX)2 = GijdXidXj = r or (dX)2 = dX' G • dX = r2 (3.84) The ellipsoid of (3.84) is called the spatial strain ellipsoid. Such strain ellipsoids as described here are frequently known as Cauchy strain ellipsoids.


The Lagrangian and Eulerian linear strain tensors are symmetric second-order Cartesian tensors, and accordingly the determination of their principal directions and principal strain values follows the standard development presented in Section 1.19, page 20. Physically, a principal direction of the strain tensor is one for which the orientation of an element at a given point is not altered by a pure strain deformation. The principal strain value is simply the unit relative displacement (normal strain) that occurs in the principal direction.

For the Lagrangian strain tensor lij, the unit relative displacement vector is given by (3.47), which may be written

du, dX


du = (L+W)'~ dX


Calling li~) the normal strain in the direction of the unit vector iu, (3.85) yields for pure strain (Wij =0 0) the relation

l(~) - l.m, i - "i) )





[CHAP. 3

If the direction 11,; is a principal direction with a principal strain value l, then



Equating the right-hand sides of (3.86) and (3.87) leads to the relationship (lij - 8ii)nj = 0 or (L - Il) • n = 0


which together with the condition 11,;11,; = 1 on the unit vectors ni provide the necessary equations for determining the principal strain value 1 and its direction cosines ni. Nontrivial solutions of (3.88) exist if and only if the determinant of coefficients vanishes. Therefore

Ilu - 8ijll = 0


IL - III = 0


which upon expansion yields the characteristic equation of 1;j, the cubic


(3.90) (3.91)


IL = l« = tr L,

are the first, second and third Lagrangian strain invariants respectively. The roots of (3.90) are the principal strain values denoted by l(1), l(2) and l(3).

The first invariant of the Lagrangian strain tensor may be expressed in terms of the principal strains as


and has an important physical interpretation. To see this, consider a differential rectangular parallelepiped whose edges are parallel to the principal strain directions as shown in Fig. 3-8. The change in volume per unit original volume of this element is called the cubical dilatation and is given by

Do = ~Vo = dX1(1 + leo)dX2(1 + l(2))dXa(l + lea») - dX1dX2dXa

Vo dX1 dX2 ax,


For small strain theory, the first-order approximation of this ratio is the sum Do = l(1) + l(2) + l(3) = IL


Fig. 3-8

CHAP. 3)



With regard to the Eulerian strain tensor e .. and its associated unit relative displacement


vector fin), the principal directions and principal strain values f(1)' f(2)' f(3) are determined

in exactly the same way as their Lagrangian counterparts. The Eulerian strain invariants may be expressed in terms of the principal strains as


The cubical dilatation for the Eulerian description is given by .:l V/V = D = f(!) + f(2) + f(3)



The Lagrangian and Eulerian linear strain tensors may each be split into a spherical and deviator tensor in the same manner in which the stress tensor decomposition was carried out in Chapter 2. As before, if Lagrangian and Eulerian deviator tensor components are denoted by dij and eij respectively, the resolution expressions are

l.;j i: L = LD + I(tr L) (3.97)
dij + SU3 or 3
and fkk E = ED + I(tr E) (3.98)
€ij eij + Su 3 or 3 The deviator tensors are associated with shear deformation for which the cubical dilatation vanishes. Therefore it is not surprising that the first invariants rhi and eii of the deviator strain tensors are identically zero.


When one and only one of the principal strains at a point in a continuum is zero, a state of plane strain is said to exist at that point. In the Eulerian description (the Lagrangian description follows exactly the same pattern), if X3 is taken as the direction of the zero principal strain, a state of plane strain parallel to the X1X2 plane exists and the linear strain tensor is given by

(" f12 D [ '" f12 ~J
fij f12 f22 or [ fij] f12 f22 (3.99)
0 0 0 0
When Xl and X2 are also principal directions, the strain tensor has the form
('~' 0 D [,~, 0 n
fij f(2) or [ fij] f(2) (3.100)
0 0 In many books on "Strength of Materials" and "Elasticity", plane strain is referred to as plane deformation since the deformation field is identical in all planes perpendicular to the direction of the zero principal strain. For plane strain perpendicular to the X3 axis, the displacement vector may be taken as a function of Xl and X2 only. The appropriate displacement components for this case of plane strain are designated by