44 Springer Series in Solid-State Sciences

Edited by Ekkehart Kroner

Springer Series in Solid-State Sciences
Editors: M. Cardona P. Fulde H.-J. Queisser

Volume 40 Semiconductor Physics - An Introduction By K Seeger

Volume 41 The LMTO Method By H.L. Skriver
Volume 42 Crystal Optics with Spatial Dispersion and the Theory of Excitations
By V.M. Agranovich and V.L. Ginzburg

Volume 43 Resonant Nonlinear Interactions of Light with Matter

By V.S. Butylkin, A.E. Kaplan, Yu.G. Khronopulo, and E.L Yakubovich

Volume 44 Elastic Media with Microstructure II Three-Dimensional Models

By LA. Kunin
Volume 45 Electronic Properties of Doped Semiconductors
By B. L Shklovsky and A. L. Efros
Volume 46 Topological Disorder in Condensed Matter
Editors: E Yonezawa and T. Ninomiya

Volumes 1 - 39 are listed on the back inside cover

IsaakA. Kunin

Elastic Media with

Microstructure II
Three-Dimensional Models

With 20 Figures

Springer-Verlag
Berlin Heidelberg New York Tokyo 1983
Professor Isaak A. Kunin
Department of Mechanical Engineering, University of Houston
Houston, TX 77004, USA

Guest Editor:
Professor Dr. Ekkehart Kroner
Institut flir Physik, Universitat Stuttgart
D-7000 Stuttgart 80, Fed. Rep. of Germany

Series Editors:
Professor Dr. Manuel Cardona
Professor Dr. Peter Fulde
Professor Dr. Hans-Joachim Queisser
Max-Planck-Institut flir Festk6rperforschung, Heisenbergstrasse 1
D-7000 Stuttgart 80, Fed. Rep. of Germany

Title of the original Russian edition: Teoriya uprugikh sred s mikrostrukturoi

© by Izdatel'stvo Nauka, Moscow 1975

ISBN-13:978-3-642-81962-9 e-ISBN-13:978-3-642-81960-5
DOl: 10.1007/978-3-642-81960-5

Library of Congress Cataloging in Publication Data. (Revised for volume 2). Kunin, I. A. (Isaak
Abramovich), 1924- Elastic media with microstructure. (Springer series in solid-state sciences; ) Rev. and
updated translation of: Teoriia uprugikh sred s mikrostrukturoI. Bibliography: v. 1, p. Includes index.
Contents: 1. One-dimensional models. - v. 2. Three-dimensional models. 1. Elasticity. I. Title. II. Series.
QA931.K8913 1982 531'.3823 81-18268
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically those of translation, reprinting, reuse of illustrations, broadcasting, reproduction
by photocopying machine or similar means, and storage in data banks. Under § 54 ofthe German Copyright
Law, where copies are made for other than private use, a fee is payable to "Verwertungsgesellschaft Wort",
Munich.
© by Springer-Verlag Berlin Heidelberg 1983
Softcover reprint of the hardcover 1st edition 1983
The use of registered names, trademarks, etc. in this publication does not imply, even in the absence
of a specific statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.

2153/3130-543210
Preface

Crystals and polycrystals, composites and polymers, grids and multibar systems
can be considered as examples of media with microstructure. A characteristic
feature of all such models is the existence of scale parameters which are con-
nected with microgeometry or long-range interacting forces. As a result the cor-
responding theory must essentially be a nonlocal one.
This treatment provides a systematic investigation of the effects of micro-
structure, inner degrees of freedom and non locality in elastic media. The prop-
agation of linear and nonlinear waves in dispersive media, static, deterministic
and stochastic problems, and the theory of local defects and dislocations are
considered in detail. Especial attention is paid to approximate models and lim-
iting transitions to classical elasticity.
The book forms the second part of a revised and updated edition of the
author's monograph published under the same title in Russian in 1975. The
first part (Vol. 26 of Springer Series in Solid-State Sciences) presents a self-
contained theory of one-dimensional models. The theory of three-dimensional
models is considered in this volume.
I would like to thank E. Kroner and A. Seeger for supporting the idea of
an English edition of my original Russian book. I am also grateful to E. Borie,
H. Lotsch and H. Zorski who read the manuscript and offered many sugges-
tions.

Houston, Texas Isaak A. Kunin

January, 1983
Contents

1. Introduction ............................................ .
2. Medium of Simple Structure 6
2.1 Quasicontinuum .................................... . 6
2.2 Equations of Motion ................................ . 12
2.3 Elastic Energy Operator ............................... . 13
2.4 Symmetric Stress Tensor and Energy Density ............. . 17
2.5 Homogeneous Media ................................ . 22
2.6 Approximate Models ................................ . 27
2.7 Cubic Lattice ...................................... . 30
2.8 Isotropic Homogeneous Medium, ........................ . 33
2.9 Debye Quasicontinuum .............................. . 37
2.10 Boundary-Value Problems and Surface Waves ............. . 41
2.11 Notes ............................................ . 44

3. Medium of Complex Structure ............................... 45

3.1 Equations of Motion ................................. 45
3.2 Energy Operator ..................................... 49
3.3 Approximate Models and Comparison with Couple-Stress Theories 54
3.4 Exclusion of Internal Degrees of Freedom in the Acoustic Region 57
3.5 Cosserat Model ...................................... 61
3.6 Notes ............................................. 66
4. Local Defects ............................................ 68
4.1 General Scheme ..................................... 68
4.2 Impurity Atom in a Lattice ............................ 70
4.3 Point Defects in a Quasicontinuum ........ . . . . . . . . . . . . . .. 70
4.4 System of Point Defects ............................... 75
4.5 Local Inhomogeneity in an Elastic Medium ................ 77
4.6 Homogeneous Elastic Medium .......................... 80
4.7 The Interface of Two Media ............................ 86
4.8 Integral Equations for an Inhomogeneous Medium .......... 89
4.9 Ellipsoidal Inhomogeneity ............................. 96
4.1 0 Ellipsoidal Crack and Needle .......................... 104
4.11 Crack in a Homogeneous Medium .... . . . . . . . . . . . . . . . . .. 111
4.12 Elliptic Crack ...................................... 115
4.13 Interaction Between Ellipsoidal Inhomogeneities ........... 118
4.14 Notes ............................................ 121
VIII Contents

5. Internal Stress and Point Defects 122

5.1 Internal Stress in the Nonlocal Theory ................... 122
5.2 Geometry of a Medium with Sources of Internal Stress 127
5.3 Green's Tensors for Internal Stress ...................... 131
5.4 Isolated Point Defect ................................ 136
5.5 System of Point Defects .............................. 141
5.6 Notes ............................................ 143

6. Dislocations 144
6.1 Elements of the Continuum Theory of Dislocations ........ . 144
6.2 Some Three-Dimensional Problems .................... . 150
6.3 Two-Dimensional Problems 152
6.4 Screw Dislocations ................................. . 154
6.5 Influence of Change of the Force Constants in Cores of Screw
Dislocations ...................................... . 158
6.6 Edge Dislocations .................................. . 160
6.7 Notes ........................................... . 163

7. Elastic Medium with Random Fields of Inhomogeneities .......... 165

7.1 Background ...................................... . 165
7.2 Formulation of the Problem .......................... . 167
7.3 The Effective Field ................................. . 170
7.4 Several Mean Values of Homogeneous Random Fields ...... . 174
7.5 General Scheme for Constructing First Statistical Moments of
the Solution ...................................... . 177
7.6 Random Field of Ellipsoidal Inhomogeneities ............ . 183
7.7 Regular Structures ................................. . 188
7.8 Fields of Elliptic Cracks ............................. . 193
7.9 Two-Dimensional Systems of Rectilinear Cuts ............ . 201
7.10 Random Field of Point Defects ....................... . 209
7.11 Correlation Functions in the Approximation by Point Defects . 216
7.12 Conclusions ...................................... . 224
7.13 Notes ........................................... . 227

Appendices ................................................ 229

A 1. Fourth-Order Tensors of Special Structure ................ 229
A 2. Green's Operators of Elasticity ......................... 232
A 3. Green's Operators K and S in x-Representation ............ 235
A 4. Calculation of Certain Conditional Means ................ 239

References 249

Bibliography 255

Subject Index 271

1. Introduction

In recent years, new physical and mathematical models of material media, which
can be considered far-reaching generalizations of classical theories of elasticity,
plasticity, and ideal and viscous liquids, have been developed intensively. Such
models have appealed for a number of reasons. Primary among them are the
use of new construction materials in extreme conditions and the intensification
of technological processes. The increasing tendency toward rapprochement of
mechanics with physics is closely connected with these factors. The internal logic
of the development of continuum mechanics as a science is also important.
This treatment is devoted to the study of models of elastic media with mi-
crostructure and to the development of the nonlocal theory of elasticity. Starting
from such models as a crystal lattice and simple discrete mechanical systems,
we develop the theory and its applications in a systematic way.
The Cosserat continuum was historically one of the first models of elastic
media which could not be described within the scope of classical elasticity [1.1]
However, the memoirs of E. and F. Cosserat (1909) remained unnoticeJ for a
long time, and only around 1960 did the generalized models of the Cosserat
continuum start to be developed intensively. They are known as oriented media,
asymmetric, multipolar, micromorphic, couple-stress, etc., theories. For short
we shall call them couple-stress theories. Essential contributions to the develop-
ment of couple-stress theories were made, for example, by Aero, Eringen, Green,
Grioli, Gunther, Herrmann, Koiter, Kuvshinsky, Mindlin, Naghdi, Nowacki,
Palmov, Rivlin, Sternberg, Toupin, and Wozniak; their fundamental works
are listed in the Bibliography. The Bibliography is far from being complete. A
survey of works before 1960 can be found in the fundamental treatment of
Truesdell and Toupin [1.2]; later ones are quoted in papers by Wozniak [1.3],
Savin and Nemish [1.4], Iliushin and Lomakin [1.5], as well as in monographs by
Misicu [1.6] and Nowacki [1.7].
From the very beginning of the development of the generalized Cosserat
models, attention was turned to their connections with the continuum theory of
dislocations. In 1967 a symposium was organized by the International Union of
Theoretical and Applied Mechanics, which had great significance in summing up
the ten-year period of development [1.8]. In the symposium a new trend closely
connected with the theory of the crystal lattice was also presented which con-
tained the above-indicated models as a long-wave approximation, namely, a
nonlocal theory of elasticity. The nonlocal theory of elasticity was also devel-
oped in works of Edelen, Eringen, Green, Kroner, Kunin, Laws and others,
2 1. Introduction

given in the Bibliography. A rather complete listing on media with microstruc-

ture is contained in [1.9]. It is worth mentioning here that the very term "non-
local elasticy" seems to have been introduced by Kroner in 1963 [1.10], and the
first monograph on the subject was published by the present author in 1975 [1. I 1].
We start out with a brief classification of the theories of elastic media with
microstructure. Explicit or implicit nonlocality is the characteristic feature of
all such theories. The latter, in its turn, displays itself in that the theories contain
parameters which have the dimension of length. These scale parameters can have
different physcial meanings: a distance between particles in discrete structures,
the dimension of a grain or a cell, a characteristic radius of correlation or action-
at-a-distance forces, etc. However, we shall always assume that the scale pa-
rameters are small in comparison with dimensions of the body.
One has to distinguish the cases of strong and weak nonlocality. If the "re-
solving power" of the model has the order of the scale parameter, i.e., if, in the
corresponding theory, it is physically acceptable to consider wavelengths com-
parable with the scale parameter, then we shall call the theory nonlocal or
strongly nonlocal (when intending to emphasize this). In such models, one can
consider elements of the medium of the order of the scale parameter, but, as a
rule, distances much smaller than the parameter have no physcial meaning.
The equations of motion of a consistently nonlocal theory necessarily contain
integral, integrodifferential, or finite-difference operators in the spatial variables.
In nonlocal models, the velocity of wave propagation depends on wavelength;
therefore, the term "medium with spatial dispersion" is also used frequently.
Let us emphasize that nonlocality or spatial dispersion can have different
origins. They can be caused by a microstructure of the medium (in particular,
by the discreteness of the micromodel) or by approximate consideration of such
parameters as thickness of a lod or plate. One can speak therefore about the
physical or geometrical nature of nonlocality. In the latter case, the nonlocal
model is, as a rule, one- or two-dimensional and serves as an effective approxi-
mate description of a local three-dimensional medium.
If the scale parameter is small in comparison with the wavelengths con-
sidered, but the effects of nonlocality cannot be neglected completely, then a
transition is possible to approximate models, for which integral and finite-
difference operators are teplaced by differential operators with small parameters
attached to their highest derivatives. In such a case, one can speak about the
model of the medium with weak spatial dispersion. The corresponding theory
will be called weakly nonlocal. All above-mentioned couple-stress theories be-
long to this type, although they are usually constructed on a purely phenomeno-
logical basis.
Finally, the consideration of sufficiently long waves (zeroth long-wave ap-
proximation) leads to a transition to a local theory in the limit, already con-
taining no scale parameters. This property of locality, i.e., the possibility of
considering "infinitesimally small" elements of the medium, is inherent in all the
classical models of the mechanics of continuous media.
 1. Introduction 3

Let us return to nonlocal models. They can be divided into two classes:
discrete and continuous. Discrete structure of a medium could be taken into
account in the usual way, for example, as is done in the theory of the crystal
lattice. However, the apparatus of discrete mathematics is most cumbersome;
therefore, we shall also use the mathematical model of quasicontinuum for an
adequate description of the discrete medium. Its essence is an interpolation of
functions of discrete argument by a special class of analytical functions in such
a way that the condition of one-to-one correspondence between quasicontinuum
and the discrete medium is fulfilled. The advantages of such an approach con-
sist in an ability to describe discrete and continuous media within the scope of a
unified formalism and, in particular, to generalize correctly such concepts of
continuum mechanics as strain and stress. It is to be emphasized that the model
of quasicontinuum is applicable not only to crystal lattices but also to macrosys-
terns.
We shall also distinguish media of simple and complex structures. In the first
case, the displacement vector is the only kinematic variable and it determines a
state of the medium completely. Body forces are the corresponding force vari-
able. To describe a medium of complex structure, a set of microrotations and
microdeformations of different orders characterizing the internal degrees of
freedom and the corresponding force micromoments is additionally introduced.
The difference between media of simple and complex structures, generally
speaking, is conserved in the approximation of weak nonlocality, but this is
displayed only for high enough frequencies of the order of the natural fre-
quencies of the internal degrees of freedom. At low frequencies, the internal
degrees of freedom can be excluded from the equations of motion so that they
will contribute to the effective characteristics of the medium only. The difference
between the quasicontinuum and the continuous medium completely disappears
in the approximation of weak nonlocality.
In the zeroth long-wave approximation, at not very high frequencies, a
complete identification of different models of a medium with microstructure
takes place: all of them are equivalent to the classical model of elastic con-
tinua which was obtained on the basis of general phenomenological postulates.
Only effective elastic moduli "know" about the structure of the initial mic-
romodel, but this information cannot, of course, be derived from them. It fol-
lows that an explicit consideration of miclOstructure effects and, in particular,
of the internal degrees of freedom is possible only with the simultaneous con-
sideration of nonlocality, i.e., a consistent theory of elastic medium with
microstructure must necessarily be nonlocal.
Schematically, the connections between different theories are shown in
Fig. 1.1.
Our main purpose is the investigation of the effects of microstructure and
nonlocality. In addition, we wish to elucidate the domain of applicability of
different theories of media with microstructure. Such theories are considerably
more complex than the usual theory of elasticity, although they reduce to this in
4 1. Introduction

Nonlocal theory of
Fig. 1.1. Illustrating
elasliclly the interconnections
of the various theories

Weakly nonlocal

certain limits; their application is, as a rule, reasonable only when they describe
qualitatively new effects which are not derivable from the local theory.
We find it advisable to consider a number of simple models of media with
microstructure in order to acquire some nonlocal intuition before proceeding to
generalizations.
For these reasons, the treatment is divided into two parts. One-dimensional
models, for which cumbersome tensor algebra is not needed, are studied in the
first volume [1.12]. At the same time, one can trace a number of distinctions of
the nonlocal theory from the local one already in one-dimensional models.
These distinctions have both physical and methodological character. Particu-
larly, one has to analyze critically the possibility of using in nonlocal theories
such habitual notions as stress, strain, energy, density and flux.
In this volume, three-dimensional models of media with microstructure
are considered, the main attention being payed to specific three-dimensional
effects. The second and third chapters are devoted to the general theory of
media with simple and complex structures. Local defects in media with mi-
crostructure as well as in elasticity are considered in Chapter 4. Internal stress
and dislocations in the nonlocal elasticity are considered in Chapters 5 and 6.
Chapter 7 (written by S.K. Kanaun) is devoted to random fields of inhomo-
geneities.
The author has tried to avoid complex mathematical methods in the first
part. It is assumed that the reader is acquainted with the Fourier transform and
has some skill in working with a-functions (though it is easy to acquire it in
the process of reading the book).
The necessity of simultaneous use of space-time and frequency-wave re-
presentations is one of the peculiarities of the formalism used in the nonlocal
theory. This way of thinking is quite habitual for physicists, but could be, to
some extent, unusual for engineers. Because of this, the rate of presenting the
material is slow in the beginning and then speeds up gradually. For the same
 1. Introduction 5

purpose, a number of results are presented in the form of problems, which are
considered a part of the text and they are referred to in the later text.
In conclusion, let us note that a number of important questions are omitted.
In particular, this is related to the thermodynamics of the nonlocal models,
which was contained in the original plan of the book. Unfortunately, the author
has not succeeded in representing this problem in a sufficiently simple and phy-
sically motivated form. An axiomatic approach to the nonIocal thermody-
namics was developed by Edelen [1.13].
2. Medium of Simple Structure

In this chapter, we consider in detail the three-dimensional elastic medium of

simple structure. To describe, in the same language, discrete and continuum
non-local models, we introduce the three-dimensional quasicontinuum. Then,
starting from crystal dynamics, we obtain the general nonlocal equations of
motion for the medium of simple structure. We investigate the consequences
of such physical requirements as energy conservation, finite action-at-a-distance,
stability and invariance of the energy with respect to rigid body motions. In
particular, the last requirement leads to the existence of a symmetric stress
tensor. In contrast to the local elasticity, the stress and energy density are not
defined uniquely in nonlocal elasticity.
The second part of the chapter is devoted to homogeneous media. Wave
propagation (including surface waves) in dispersive media and the correspond-
ing physical phemomena are examined. Approximate theories that can describe
partially these effects are discussed. As examples of non local media, a cubic
lattice and isotropic models are considered.

2.1 Quasicontinuum

Let us suppose that in a three-dimensional Euclidean space E3 with points x,

a triple of noncoplanar vectors ea(a = 1, 2, 3) with a common origin at the
point Xo is given. The set of points in E 3, which is obtained through all dis-
placements of the origin by vectors! naein and a being aribtrary integers)
forms a (simple) lattice with an elementary cell of the form of parallelepiped,
constructed on ea' Points of this lattice are also called knots.
It is convenient to introduce oblique lattice coordinates x a with the origin
at the point xo, basis vectors ea and a metric tensor gaf3' which is equal to the
scalar product of the basis vectors: gail = (e a, ef3). Then, to knots of a lattice
there correspond combinations of vectors n = naea with integer components na.
Let u(n) be a scalar or tensor, generally a complexvalued function, which is
given at knots and which increases no faster than a power of Inl as Inl -+ 00.
Then analogously to the one-dimensional case [1.12], the function u(n) can be
interpolated by a generalized analytic function u(x), which is uniquely deter-

IHere and in the following material by identical superscripts and subscripts we contemplate
summation.
 2.1 QuasicontinuuID 7

mined by some natural conditions. In order to avoid repetition, we shall con-

centrate on the special features of the three-dimensional case in the following.
Together with the interpolating function u(x) let us also consider its Fourier
transform:

u(k) = Ju(x)e- ikx dx . (2.1.1)

Here, kx is, strictly speaking, a linear function, which is conveniently written

in the form of a scalar product. As is well-known, the space E3 of all linear
functions of x is a three-dimensional vector space (k-space), where the basis
ef3 is introduced; it is reciprocal to the basis ea:

(2.1.2)

Then k = kf3ef3 and kx = kaxa. Note that the last expression is also under-
stood as an ordinary scalar product (k, x), if between the spaces E3 (with fixed
origin) and E~ an identification is established with the help of the metric tensor
gafJ' i.e., the operation of raising and lowering of indices. For our purpose such
an identification is not convenient and we shall consider separately the "physi-
cal" x-space E3 and the dual k-space E~.
Let us construct a parallelepiped B { - 11: :s; kfJ :s; 11:}, in the k-space, whose
edges are parallel to vectors of the reciprocal basis ef3 and set

o(x) ~ _1_ Je
B (211:)3 B
ixk dk = _1~ IT
11:3Vo fJ=l
sin(xef3 )
xe f3 '
(2.1.3)

where Vo is the volume of an elementary cell constructed on the vectors ea' It is

obvious that OB(O) = VOl and oB(n) = 0 for all other knots.
Let us assume that u(n) decreases sufficiently rapidly as Inl --+ 00. Then in
B one can define a function u(k), such that the coefficients of its Fourier series
are equal to vou(n), i.e.

u(k) = Vo L; u(n)eink, kEB. (2.1.4)

n

Defining u(k) to be zero outside B, and taking into account (2.1.3) we find
for the inverse Fourier transform u(x)

u(x) = Vo L; u(n)oB(x - n) . (2.1.5)

n

It is possible to prove that the function u(x) can be continued analytically

into the complex plane as an entire function of exponential type :s; 11:.
In view of the properties of OB(X), it is easy to see that u(x), defined by (2.1.5),
is in fact the required interpolating function. The one-to-one correspondence
8 2. Medium of Simple Structure

u(n) ~ u(x) ~ u(k) is well guaranteed by the condition of truncating the Fou-
rier-transform u(k) of the function u(x) and the uniqueness of the expansion
(2.1.4).
Let us proceed to the general case of the space N'(B), whose elements are
functions u(n), which increase no faster than some power of Inl as Inl --+ 00.
The space N'(B) can also be regarded as a space of generalized functions on the
space N(8), the latter being the space of rapidly decreasing functions.
Recall that in the one-dimensional case [1.12], when interpolating functions
u(n) E N'(B), it turned out to be necessary to identify the ends of the segment B,
transforming it into a circle. Analogously, let us identify the opposite faces of
the present parallelepiped B or, in other words, let us transform it into a three-
dimensional torus.
The series (2.1.4) now converges only in the sense of the generalized func-
tiom [2.1], and u(k) belongs to the space of the generalized functions K'(B),
defined on the space K(B) of infinitely differentiable test functions with sup-
ports in B.
Let us denote the inverse Fourier transforms of the spaces K(B) and K'(B)
by X(B) and X'(8), respectively. The inverse Fourier transform u(x) E X'(B) of a
function u(k) E K'(8) is defined, as usual, with the help of the Parseval equality

(2.1.6)

where q>(x) E X(B) and q>(k) E K(B) are connected with each other by the usual
Fourier transform.
It can be proved [2.1] that u(x) E X'(B) are regular generalized functions (i.e.
conesponding to ordinary functions), which increase no faster than a power
of Ixl as Ixl --+ 00 and which can be analytically continued to a complex region
as entire functions of the exponential type :5; Te. The proof of the fact that the
function u(x) has values u(n) at the knots, can be performed analogously to the
one-dimensional case [l.12].
The uniqueness of interpolation is ensured by the identification of the
opposite faces of the parallelepiped B. This excludes from X'(B) functions of
the type P(x) sin[Te(xe)~], [P(x) being a polynomial], which vanish at all the
knots.
Thus we have:
Theorem. The formulae (2.1.4, 6) fix linear isomorphisms between the spaces

N'(B) ~ X'(B) ~ K'(B) , (2.1.7)

such that u(x) E X'(B) interpolates u(n) E N'(B) .

We may now consider u(n), u(x) and u(k) as representations of one and the
same function in different functional bases. Of course, to this end it is also
 2.1 Quasicontinuum 9

necessary to know the corresponding representations of operations over these

functions.
A function q(n), which decreases rapidly, as Inl -+ 00 (in particular, a func-
tion with finite support), determines the linear functional

(qlu) ~ ~ q(n)u(n) = Jq(x)u(x)dx = (2~)3 Jq(k)u(k)dk , (2.1.8)

which is invariant with respect to the n -, x -, and k-representations, if the

kernels q(n), q(x) and q(k) are connected by the relations

q(k) = 1: q(n)e-ink,
n
kEB,

q(x) = 1: q(n)oB(x -
n
n) . (2.1.9)

Problem 2.1.1. Prove the above proposition.

Note that, as distinct from (2.1.4, 5), in the correspondence q(n) +-+ q(x) +-+
q(x) the factor Vo is absent. This is connected with the fact that in the further
consideration the quantities q(n) will have the meaning of forces acting at the
knots of a lattice. Thus q(x) can be interpreted as the corresponding density of
body forces, q(k) as its Fourier transform and the functional (qlu) as the work,
done by the forces q on the displacement u.
As an important example, consider the functional

(olu) = u(n) In=o = u(x) Ix=o = Ju(k)dk (2.1.10)

B

with the kernels

o(n) +-+ OB(X) +-+ l(k) . (2.1.11)

Here o(n) = 1 at n = 0, and it is equal to zero at all other knots. Evidently,

OB(X) plays in X'(B) the role of an ordinary o-function but, as distinct from the
latter OB(X) is not singular. In what follows, we shall use the notation o(x)
instead of oix) when this does not lead to any misunderstandings.

Problem 2.1.2. Show that the following identities are generated by the func-
tional with the kernel I (x) :

Vo 1: u(n) = Ju(x)dx = u(k) Ik=O . (2.1.12)

10 2. Medium of Simple Structure

Analogously to the one-dimensional case [1.12] we define the invariant form

<ul([Jlw) <le( 1: u(n)([J(n, n')w(n') = SSu(x)([J(x, x')w(x')dx dx'

nn'

= (2~)3- Sf u(k)([J(k, k')w(k')dk dk' = <wl([J+lu) . (2.Ll3)

Here

([J(k, k') = _1-3 1: ([J(n, n')e-iCkn-k'n'l

(21t) nn'

= _1 _ SS "'(x
G1tP ~,
x')e-iCkx-k'x'ldx dx',
, (2.1.14)

([J+(k, k') = ([J(k',k), ([J+(x, x') = ([J(x', x) . (2.LlS)

One can establish the correspondences between the discrete and continuous
operations of multiplication, convolution, differentiation and so on, distinc-
tions compared with the one-dimensional case being small.
Thus, the functions u(n) E N'(B) constitute a ring with respect to the opera-
tion of multiplication. At the same time, the ordinary product of two functions
from X'(B), generally speaking, does not belong to X'(B). Nevertheless, it is
possible to introduce in X'(B) such an operation of multiplication, which cor-
responds to the multiplication of the inverse transforms in N'(B) and whose
result is still contained in X'(B).
In fact, on the torus B we can define an operation of displacement and,
hence, an integral convolution

u(k) * w(k) = Su(k - k')w(k')dk', (2.1.16)

B

with respect to which K'(B) is a ring.

Problem 2.1.3. Verify the correspondence

fen) = u(n)w(n) +-> f(k) = u(k) * w(k) . (2.1.17)

This leads to the natural definition of the product, in X '(B):

f(x) = u(x)· w(x) +-> f(k) = u(k) * w(k) . (2.Ll8)

Problem 2.1.4. Show thatf(x) coincides withf(n) at all knotes.

In particular, the following formulae are valid:

u(x)o(x) = u(O)o(x), o(x)o(x) = voo(x) . (2.1.19)

 2.1 Quasicontinuum 11

Here, we observe from a new point of view, the expedience of transforming

the parrallelepiped B into a torus, since it is impossible to find a reasonable de-
finition of a convolution on a parallelepiped such that the support of u(k)*w(k)
is always contained in the parallelepiped.
Note that if the supports of u(k), w(k) E K'(B) are contained in a paral-
lelepied similar to B, but twice smaller, then, the convolution in (2.1.16) coin-
cides with the ordinary one and hence the above-defined multiplication also
coincides with the ordinary one. Of course, this will also happen if one of the
factors is a polynomial (in the x-representation), since its support (in the k-
representation) is concentrated at the point k = o.
The algorithm considered above for passing from functions of discrete
argument to functions of continuous argument may be clearly interpreted as
spanning some analytic structure over the lattice in such a way that one-to-one
correspondence between the structures is conserved. The Euclidean space in
which only functions from X'(B) are allowed, is conveniently interpreted as a
quasi continuum, its geometry being isomorphic to that of a lattice. When the
size of elementary cell of the lattice tends to zero (this is equivalent tp the
consideration of functions which change only slightly over distances of the
order of the cell size), the region B extends over all k-space and the quasi-
continuum is transformed into a continuum with the ordinary geometry.
The idea of quasicontinuum can be naturally generalized by taking, instead
of the parallelepiped B, another manifold. The isotropic continuum will be
of considerable importance, the corresponding region B being a sphere. The
replacement of the parallelepiped by a sphere in k-space was first carried out
by Debye in order to calculate the heat capacity of a crystal and was further
widely used for qualitative estimates of physical properties of crystals (without
any connection with the idea of the quasicontinuum). The radius K of the sphere
is usually determined by the condition of equality of the volumes of the sphere
and the parallelepiped. The exact correspondence with the lattice structure is
lost in this case, but the corresponding quasicontinuum model, which will be
called the Debye model, describes qualitatively correctly the presence of a dis-
crete structure with the characteristic scale parameter a ~ 1':,/ K.
The Debye model can be somewhat improved by identifying the opposite
points of the boundary sphere analogously to identifying opposite faces of the
parallelepiped. In this case, the manifold B, will be homeomorphic to a three-
dimensional projective space.
As another example we can mention a layered medium, in which the char-
acteristic scale parameter is to be introduced in one direction only. Accordingly,
in the k-space, transform is to be carried out with respect to one coordinate
only.
In the general case, the quasicontinuum of class B may be defined as the
set of two objects: an Euclidean space and a space of admissible functions X'(B)
given on that space, B being a fixed manifold.
12 2. Medium of Simple Structure

2.2 Equations of Motion

The model for a medium with simple structure will be based on Born's model
of a simple lattice with the basis ea in the harmonic approximation. For the
elastic energy rp we have, analogously to the case of a one-dimensional chain
[1.12],

rp = 21 L: uaCn) rpafJ(n, n')uf3(n') , (2.2.1)

nn'

where rpaf3 (n, n') is a given tensor of the force constants, and uaCn) is the displace-
ment of the particles from the equilibrium position, i.e. it is assumed that initial
external forces are absent.
This model may be interpreted, if necessary, as a system of pointwise parti-
cles situated at the knots of the lattice and interacting by means of linear elastic
bonds of the most general nature.
The Lagrangian of this system has the form

L = i ~ gafJ m(n)uaCn)uf3(n)

-i~ uaCn)rpafJ(n, n')uf3(n') - ~ qa(n)ua(n) , (2.2.2)

where men) is the mass of a particle at the point n, and qa(n) is the external
force acting on the particle. The dependence of the field parameters on time is
not indicated explicitly.
In the usual way we find the equations of motion

m(n)iia(n) + L: rpaf3 (n, n')uin') = qa(n) . (2.2.3)

n'

Using the above-described algorithm, let us proceed to a quasicontinuum

representation. For the elastic energy rp we have, cf. (2.1.13)

rp = i (ualrpafJlufJ) = iss ua(x)rpa f3 (x, x') ulx')dx dx'

= (2;)3 SSua(k)rpafJ(k, k')ulk')dk dk' , (2.2.4)

where rpafJ(x, x') and rpaf3 (k, k') are expressed in terms of the force constants

rpafJ (k, k') = _1-3 L: rpafJ(n, n')e-iCkn-k'n')

(2rc) nn'

= _1_ SS mafJ(x, x')e-iCkx-k'x') dx dx'

(2rc) 3 'V •
(2.2.5)
 2.3 Elastic Energy Operator 13

Introducing the mass density p(x) and force density q(x)

p(x) = 1:n m(n)o(x - n), q(x) = 1:

n
q(n)o(x - n) , (2.2.6)

we write the Lagrangian (2.2.2) in the form

(2.2.7)

where the operator p has the kernel p(x)o(x - x').

If we do not assume the existence of a discrete micromodel, then the La-
grangian in the form or (2.2.7) corresponds to the most general nonlocal theory
of an elastic medium with simple structure. In this case, the admissible functions
are not restricted by truncation of their Fourier transforms, i.e., all the field
quantities are considered to be given on an ordinary manifold. The quasi-
continuum (discrete) model is, in fact, a particular case of the general theory.
The equations of motion, which correspond to the Lagrangian (2.2.7) have the
form

p(x)jja(x) + J<pa~(x, x')uix')dx' = qa(x) , (2.2.8)

(21t)-3 p(k) * jja(k) + J<pa~(k, k')u~(k')dk' = qa(k) (2.2.9)

in the x- and k-representations, respectively. In the case of the quasicontinuum,

the n-representation (2.2.3) is to be added.
The equations of motion are conveniently considered in the ev-representation.
In operator form, they have the form cf. (2.1.37),

(2.2.10)

Obviously, the physical content of the equations of motion is determined,

first of all, by properties of the operator of the elastic energy <pall; the next
section will be devoted to its consideration.

2.3 Elastic Energy Operator

The properties of the operator <pa~(x, x') are in many aspects analogous to the
properties of the operator <P(x, x'), which were considered in details in [Ref.
1.12, Sect. 2.4]. We shall therefore concentrate on features specific to the
three-dimensional case.
Hermiticity. From (2.2.4), it follows that

<pa~(x, x') = <p~a(x', x), <pa~(k, k') = <p~a(k', k) . (2.3.1)

14 2. Medium of Simple Structure

Also, since cpa~(x, x') is real

cpa~(k, k') = cpa~( - k, - k') . (2.3.2)

Finiteness of action-at-a-distance: In what follows it is assumed that cpa~(x, x') =

°if Ix - x'l > I for any value of x. In other words, it is assumed that there
exists a characteristic radius of interaction, I. An argument analogous to that
discussed in [Ref.1.l2, Sect. 2.4] implies that cpap(k, k') is an entire function of
exponential type :::; I with respect to the argument k + k'.
If, moreover, the medium is bounded by a characteristic dimension L,
then cpap(k, k') is also an entire function of exponential type:::; L of the argu-
ment k - k'.
Invariance of cp with respect to translation and rotation, as will be shown in
the next section, is equivalent to the conditions2

Scpa~(x, x')dx = 0, Sx[).cpaJ~(x, x')dx = ° (2.3.3)

or, in the quasicontinuum (lattice) terminology,

L.: cpa~(n, n') =

n
0, L.: n[).cpaJ~(n, n') =
n
°. (2.3.4)

From this follows the possibility of representing cpa~(k, k') in the form

(2.3.5)

where c).aJ1.~(k, k') is an analytic function of the same type as cpa~(k, k'); without
loss of generality it can be considered to be symmetric in the indices inside the
first and second pairs and to be Hermitian with respect to permutation of the
pairs, i.e.

(2.3.6)

Local approximation: If the field parameters vary slowly over distances of the
order of I, then

cAaJ1.~(x, x') = c~aJ1.~(x)o(x - x') , (2.3.7)

where c$aJ1.~(x) is a tensor of the elastic moduli of the corresponding local

model of the zero-order approximation (with respect to I).
Stability: This requirement is equivalent to positive definiteness of the elastic
energy cp for all admissible displacements, excluding translation and rigid
body rotation, for which cp = 0.

2Here and subsequently indices which are contained in brackets (parentheses) are assumed
to be subject to the operation of alternation (symmetrization).
 2.3 Elastic Energy Operator 15

Media with binary interaction: One of the possible specific properties is the
following one. The forces appearing due to a change of the distance between
points x and X', are proportional to this change and are directed along the
line, which connects these two points. It is easy to show that such a medium can
be described by an operator of elastic bonds 1Jfup with the kernel

1JfuP(x x') = (xu - x'U)(xP - x'P) 1Jf(x x') (2.3.8)

, Ix - x'12 "

where 1Jf(x, x') = 1Jf(x', x) is a scalar characteristic of the elastic bond which
connects the points x and x', such that 1Jf(x, x') = 0 when Ix - x'i > L.
For the energy operator rp we may write

rpUP(x - x') = cpuP(x)o(x - x') - 1JfuP(x, x') , (2.3.9)

cpUP(x) = f 1JfuP(x, x')dx' . (2.3.10)

The conditions of invariance with respect to rotation are here satisfied au-
tomatically.
Homogeneous media: In this case, which is important for applications

rpuP(x, x') = rpaP(x - x') • (2.3.11)

From (2.2.5) we find

rpuP(k, k') = rpuP(k)o(k - k') , (2.3.12)

rpuP(k) = f rpuP(x)e- ikx dx = rpPu(k). (2.3.13)

In case of the quasicontinuum we also have

rpuP(n, n') = rpuP(n - n') , (2.3.14)

(2.3.15)

The finiteness of action-at-a-distance is now equivalent to the functions

rpaP(x) or rpaP(n) having finite support, from which it follows that rpaP(k) is an
entire function of exponential type ~ L.
The conditions (2.3.3, 4) now take the form

f rpaP(x)dx = 0, f x[).rpulP(x)dx = 0, (2.3.16)

16 2. Medium of Simple Structure

.L; (/jafJ(n) = 0, .L; n[)'(/jalfJ(n) = 0, (2.3.17)

n n

and to (2.3.5, 6) there correspond

(2.3.18)

(2.3.19)

The static Green's tensor Ga{l is defined as an operator which is inverse with
respect to the energy operator (/jafJ and hence

J(/ja).(x, x")G).ix", x')dx" = o~o(x - x'),

J(/ja).(k, k")G).{l(k", k')dk" = o~o(k - k'), (2.3.20)

Galx, x') satisfying the corresponding boundary conditions as x, x' --+ 00.
For a homogeneous medium,

(2.3.21 )

i.e. in the k representation, the kernel of the Green's tensor can be obtained by
the purely algebraic operation of inverting the matrix (/jafJ(k).
It is easy to show that from the definition of the Green's tensor, the elastic
energy can be represented in the static case, as

(2.3.22)

where the qa's are body forces. Conversely, this relation can be taken as a de-
finition of the static Green's tensor.

Problem 2.3.1. Prove the equivalence of the two definitions of GafJ'

It is obvious that the kernels Ga{l(x, x'), Ga{l(k, k') possess the symmetry
properties (2.3.1, 2). For the quasicontinuum, due to the condition k E B, the
kernel Galx, x') is an entire function of the corresponding arguments.
The dynamic Green's tensor GafJ is defined as an operator inverse to the
operator

UfJ clef pgafJ OF + (/ja{l . (2.3.23)

We leave it to the reader to write the analogs of the equations (2.3.20) for the
kernels GafJ(x, x', t), GafJ(x, x', ev) and Galk, k', ev).
 2.4 Symmetric Stress Tensor and Energy Density 17

2.4 Symmetric Stress Tensor and Energy Density

In this section, we introduce a symmetric stress tensor and the corresponding

elastic energy density. In order to achieve this, the relations (7.3.5 and 6) have
to be proved.
We need three auxiliary propositions.
Lemma 1. Let A(k) be an analytic tensor function (indices are omitted) and
A(O) = O. Then

A(k) = kJ.AJ.(k) , (2.4.1)

where the analytic function AJ.(k) is uniquely determed by the condition that
coefficients of its Taylor series
00

AJ.(k) = ~ AJ.']···'nk'l ... k'n (2.4.2)

n=O

are symmetric with respect to the indices A, AI, ... , An'

The proof is obvious: it is sufficient to compare the Taylor series for A(k)
and AJ.(k).
Lemma 2. Let AJ.af/ be a tensor, symmetric with respect to the indices A, p. Then
it can be associated with the tensor

(2.4.3)

which is symmetric with respect to the indices A, a; the one-to-one correspon-

dence:

(2.4.4)

holds.
Lemma 3. If an analytic function Aa(k) satisfies the conditions

(2.4.5)

then it admits the representation

(2.4.6)

where the analytic function BJ.a(k) is symmetric with respect to the indices A, a.
In order to prove this let us write the function Aa(k) in the form

(2.4.7)
18 2. Medium of Simple Structure

The constant tensor A&a appearing here is symmetric because of the second
of the conditions (2.4.5); the tensor function A}.a"(k) is symmetric with respect
to the indices Afl- and, according to Lemma 1, is defined uniquely. This function
can be replaced by a tensor function A{Aa")(k) from Lemma 2, which is sym-
metric with respect to the indices A, a, since the factor k}.k" provides the opera-
tion of symmetrization needed in the identity (2.4.4). As a result, we find

(2.4.8)

This completes the proof.

Let us now prove the relations (2.3.5). The main physical assumption that
the elastic medium is situated in a homogeneous isotropic Euclidean space and
is equivalent to requiring the in variance of all properties of the medium with
respect to translation and rigid rotation of the whole medium. As it is known
from Lie group theory, it is sufficient to restrict ourselves to invariance with
respect to infinitesimal transformations.
Let the state of the medium be described by a displacement field U a • In the
linear theory the principle of superposition is always valid and to an additional
infinitesimal displacement, there corresponds the transformation U a -> U a +
ug, where, in the case of translation,
U£(x) = aa = const, u£(k) = (2'Jt)3aao(k) , (2.4.9)

and for an infinitesimal rigid body rotation

(2.4.10)

The requirement concerning invariance of the energy (/J with respect to the
transformation Ua -> Ua + u£, (2.2.4) and the Hermiticity of the operator
(/Ja{3 lead to the following equation

2(u£l(/Ja{3lu{3> + (u£l(/Ja{3lu~> = 0, (2.4.11)

which is to be satisfied for any u£ and u{3' From this one can obtain conditions
which must be satisfied by the kernel of the operator (/Ja{3'
First, let us make one remark. In the case of an infinite medium, the ex-
pression for the energy (2.2.4) and hence the second term in (2.4.11) are not a
priori defined on displacements ug of the form (2.4.9) or (2.4.10). Consequently
let us consider next a bounded medium and then perform the transition to the
limit.
Taking into account the arbitrariness of u{3' we conclude that (2.4.11) is
equivalent to the condition
Ju£(x)(/Ja{3(x, x')dx = 0 (2.4.12)
 2.4 Symmetric Stress Tensor and Energy Density 19

or in the k representation

Ju~(k)f/Ja~(k, k')dk = 0. (2.4.13)

Using (2.4.9), aa being arbitrary, gives two equivalent conditions for f/Ja~

Jrpa~(x, x')dx = 0, f/Jap(k, k') Ik=O = 0 • (2.4.14)

Due to the assumption about the boundedness of the medium the function
x') has finite support and hence the function f/Jap(k, k') is analytic with
f/Ja~(x,
respect to the pair of points k, k'. Taking into account also the conditions of
Hermiticity (2.3.1) we find that the relations (2.4.14) are equivalent to the
possibility of representing f/Ja~(k, k') in the form

f/Ja~(k, k') = k).k~f/J).ap.(3(k, k') , (2.4.15)

where, according to Lemma 1, the analytic function f/JAap.~(k, k') is determined

uniquely.
The substitution of (2.4.10) into (2.4.13), aa). being an arbitrary antisymme-
tric tensor, gives in the x-representation

JX[Arpa]~(X, x')dx = 0 . (2.4.16)

In the k representation, taking into account (2.4.15) and the identity

k;.'iJp.o(k) = - oqo(k) , (2.4.17)

we have

(2.4.18)

and, due to the Hermitian symmetry of rpa~(k, k'),

k ).f/JAa[p.(31(k, 0) = O. (2.4.19)

We now apply Lemma 3 to the function f/Ja~(k, k') for each argument k and
k'. This leads to the required result:

(2.4.20)

Here, the tensor cAap.~(k, k') satisfies the symmetry conditions (2.3.6) and
is connected with f/JAap.~(k, k') by

(2.4.21)
20 2. Medium of Simple Structure

Problem 2.4.1. Check that c Aap{3(k, k') satisfies the conditions (2.3.6).

It was assumed above that the medium is finite. However, the derived
result about the representability of l/Ja{3(k, k') in the form of (2.3.5) with the
tensor c Aap{3(k, k') possessing the symmetry (2.3.6) depends neither upon the
shape nor upon the size of the elastic body. Hence this remains valid also
for the limiting case of an infinite medium. According to the same reasoning,
if the second term in (2.4.11), vanishes for an abritrary finite medium, it can
be considered to be equal to zero also in the limiting case of an infinite medium
(natural regularization).
Let us now introduce the strain tensor

(2.4.22)

and the symmetric tensor

(2.4.23)

or in the k-representation

O'Aa(k) = JcAap{3 (k, k') ep{3(k')dk' . (2.4.24)

The equations of motion (2.8) take the form

(2.4.25)

or, in direct notations,

pu - div 0' = q. (2.4.26)

The expression for elastic energy can be represented in the form

(2.4.27)

or

l/J = Jrp(x)dx, (2.4.28)

From the above expressions it follows that O'Aa(x) and rp(x) may be inter-
preted as the stress tensor and the elastic energy density, respectively, both
quantities being invariant with respect to translation and rotation, as they
should be according to their physical meaning.
 2.4 Symmetric Stress Tensor and Energy Density 21

The expressions (2.4.23) and (2.4.24) which can be rewritten in operator

form as

(1 = Co:, (2.4.29)

can be interpreted naturally as Hooke's law in operator form with C playing

the role of the operator of the elastic moduli. In essence, we have proved the
following
Theorem. In the nonlocal theory of a linear elastic medium of simple structure
with finite action-at-a-distance, it is always possible to introduce a symmetric
stress tensor and an energy density, which can be expressed in terms of stress
and strain in the usual way.
At first glance, it may seem that we have proved even more, since the
expression obtained above determines the stress tensor and the energy density
uniquely. However, this conclusion is of a formal nature, since the derivation
itself contains an element of arbitrariness. From reasoning analogous to that
described in [Ref. 1.12, Chap. 2], it follows that the stress and energy density
in a nonlocal theory cannot be determined uniquely from a physical point
of view, if only the operator of the elastic energy (]Jafi(x, x') is given; information
on the structure of the elastic bonds is also needed. At the same time, in the
three-dimensional case additional difficulties arise as compared with the one-
dimensional one. The first difficulty is connected with the nonuniqueness of
tensor representations of the type (2.4.20). The second and more important
one is due to the fact that, generally speaking, even the effective bonds do
not allow representation in the form of a binary interaction. As a consequence,
the stress tensor (2.4.23) only approximately determines the force which acts
on a small area, and the larger the latter is (in comparison with the charac-
teristic length I), the more accurate is the relation considered.
Nevertheless, the conclusion which we have drawn is of principle signi-
ficance. This is because it is a widespread opinion that as distinct from the
ordinary theory of elasticity, a nonsymmetric stress tensor is necessarily as-
sociated with a weakly nonlocal theory of media of simple structure. As an
example, we mention the so-called couple-stress theory of elasticity with
constrained rotation. The corresponding proofs cannot be considered as
entirely correct.
This is connected with the fact that in the above-mentioned theories the
stress tensor is introduced in a formal manner, by analogy with the local
theory of elasticity and without due consideration of specific features of the
nonlocal model. In particular, no attention is paid to the nonuniqueness of
tensor representations of. the type (2.4.20) and, as a rule, the natural principle
of correspondence is not taken into account: in the long-wave length approxi-
mation, the stress tensor must coincide with the classical symmetric stress
tensor.
22 2. Medium of Simple Structure

In conclusion, let us note that, as in the one-dimensional model, one can

also introduce two-point measures of deformation, stress and energy density,
which have, however, a clear physical meaning only in the presence of binary
interaction.

2.5 Homogeneous Media

Let us consider an infinite homogeneous medium with the difference operator

(/JafJ(x - x') and the constant density p. In the k-representation

(/JafJ(k, k') = (/JafJ(k)o(k - k') , (2.5.1)

where (/JafJ(k) is defined by (2.3.13) or (2.3.15).

In view of the essentially non-analytic nature of (/JafJ(k, k') a direct gener-
alization of the results of the last section to the case of a homogeneous medium
is impossible; exact limiting transitions are necessary, but these are connected
with some difficulties. Therefore we prefer a more direct approach.
Let us begin with the properties which (/JafJ(k) possesses due to the invariance
of the energy with respect to translation and rotation.

Problem 2.5.1. Show that a necessary condition for this invariance is the re-
presentability of (/JafJ(k) in the form

(2.5.2)

[Hint: Expand (/JafJ(k) in a series, write (2.4.12) and (2.4.16) in the k-represen-
tation and take into account the condition of Hermiticity.]

However, the condition (2.5.2) is not sufficient. In order to obtain an

additional condition, let us take into account that due to (2.4.10), this can
be imposed only on (/JJ.af'fJ, i.e. on the constants of the zeroth long-wavelength
approximation. But in this limiting case there must exist a uniquely determined
elastic energy density 'Po(x), which is invariant with respect to translation and
rotation, and also makes sense for displacements which increase linearly as Ixl
-+ 00 (as distinct from the energy (/J).
To the operator of the zeroth-order approximation

(2.5.3)

corresponds the Lagrangian density

'Po(x) = ~ chua(x)(M/lfJfJ0f'ufJ(x). (2.5.4)

The energy density 'Po(x) differs from q>Q(x) by divergence terms, which must
 2.5 Homogeneous Media 23

be chosen in accOldance with the conditions of invariance of <po(x) with respect

to translation and rotation. Since ~(x), which is determined by (2.5.4), is
already invariant with respect to translation, the divergence part must also
possess this property.

Problem 2.5.2. Verify that the most general expression for the divergence
part, which is quadratic with respect to the displacement and invariant with
respect to translation, has the form

(2.5.5)

where b).a}J~ = -b}Ja).~ is a constant tensor.

Problem 2.5.3. Show that without loss of generality it is possible to replace
b).a}J~ by a tensor, which is antisymmetric with respect to Af-t and a{3 and is
symmetric with respect to permutations pf these pairs.

Thus, <po(x) can be represented in the form

(2.5.6)

where ~}J~ and b).al'~ possess the above-described symmetries. The conditions
of invariance with respect to rotation yield

(2.5.7)

Regarding (2.5.7) as an equation for b).a}J~, let us find conditions for its
solvability. For this purpose it is necessary to employ a simple but somewhat
cumbersome procedure: to symmetrize (2.5.7) with respect to Af-t and a{3
separately, then to take the difference of these expressions and, finally, to
symmetrize the difference with respect to Aa. The final result will contain
only two components of the tensor (/loa}J~ with permuted pairs of indices,
~}J~ = (/Yij).~}J. (2.5.8)

This is, in fact, the desired additional condition which ensures the invariance
of <po(x) with respect to rotation.

Problem 2.5.4. Show that if the condition (2.5.8) is satisfied, (2.5.7) has only
one solution:
b).a}J~ = (/J{f}J~ - rpArf"~, (2.5.9)
which possesses the necessary symmetry.
24 2. Medium of Simple Structure

The substitution of (2.5.9) in to (2.5.6) leads to the final expression for

the elastic energy density

(2.5.10)

where

(2.5.11)

is the tensor of the elastic constants in zeroth approximation. As desired, this

tensor possesses the symmetry with respect to the first and second pairs of
indices as well as with respect to permutation of the pairs, i.e.

(2.5.12)

It can easily be seen that (5.3) now admits another form:

(2.5.13)

Such equalities reflect features specific to the three-dimensional case. For a

one-dimensional medium (/J(k) uniquely determined the coefficient of k 2,
whereas the tensor coefficient of k),k p is given only when its symmetry is speci-
fied. The physical meaning of the tensor of elastic constants c&a p(3 is clear.
The tensor (/J$a p(3 is convenient since it is directly connected with the force
constants. In fact, from (2.5.2) and (2.3.13) we have

(2.5.14)

or in the n-representation, taking into account (2.3.15),

(2.5.15)

These expressions enable us to express explicitly C$a p(3 in terms of the force
constants and to interpret (2.5.8) as restrictions on the force constants, due
to the requirements of invariance of the energy with respect to rotation.
Thus, we have completed the first part of the problem. Now it is necessary
to show that (2.5.2) admits a representation in the form

(2.5.16)

where the operator of the elastic moduli c),ap(3(k) possesses the symmetry (2.3.19),
and to find a connection between (/JJ.a p(3(k) and cJ.ap(3(k).
Let us now write down (2.2.4) for the elastic energy (/J in the form

(2.5.17)
 2.5 Homogeneous Media 25

It is easily seen that in this expression it is possible to replace (/)).apfJ(k) by the

tensor

(2.5.18)

which is symmetric with respect to the indices ila because of the symmetry
properties of (/)).apfJ(k) (Lemma 2). This enables us to write (2.5.17) in the form

(2.5.19)

where cdk) is the strain tensor (2.4.22).

Let

(2.5.20)

Taking into account (2.5.11) and (2.5.18), we have

(2.5.21)

from which it follows that

a~a[pfJJ (0) = 0, (2.5.22)

or

(2.5.23)

where a~apfJT(k) is a tensor, antisymmetric with respect to pf3.

The expression for (/) may be now rewritten in the form

(2.5.24)

where

(2.5.25)

is an antisymmetric rotation tensor.

The first term on the right-hand side of (2.5.24) does not change if we replace
afa(pfJ) by the tensor

r.'apfJ _
U2 - ~
1 (a).a(pfJ)
1
+ a(p{3),a)
1, (2.5.26)

which has the required symmetry.

Taking into account the identity

(2.5.27)
26 2. Medium of Simple Structure

let us represent the second term in (2.5.24) in the form

<c).ala~a,ut3/c,ut3> ,

where

(2.5.28)

also has the required symmetry.

Finally, combining both terms, we obtain

(2.5.29)

where the operator of the elastic moduli

(2.5.30)

satisfies the symmetry conditions (2.3.19) and is connected with <pAa,utS(k) by the
above-described expressions, which implicitly determine the operator S.
Hooke's law in the operator form (2.4.23, 24) takes now the form

(JAa(x) = JcAa,uf3(x - x') c,uf3(x')dx' , (2.5.31)

(2.5.32)

Problem 2.5.5. Show that the real (imaginary) part of c Aa ,uf3(k) is an even (odd)
function of k and hence, for an isotropic medium and for a medium with central
symmetry, c Aa,uf3(k) is a real function.

The equation of motion (2.2.8) for a homogeneous medium has the form
(2.5.33)

or in the (k, w)-representation:

(2.5.34)

Taking into acount (2.5.16), the equations of motion may be also written
in the following form

J
pua(x) - 0). cAa,utS(x - x')0,uuix') dx' = qa(x) , (2.5.35)

pw 2ua(k) - k).k,ucAa,utS(k)utS(k) = -qa(k). (2.5.36)

For an infinite medium, free vibrations have a special significance. By free

vibrations we mean solutions of the equations of motion with q = 0, which
 2.6 Approximate Models 27

correspond to propagating waves. It is clear that nontrivial solutions ale pos-

sible only when wand k are connected by the dispersion equation

det[pw 2ga ,8 - q)a,8(k)] = 0, (2.5.37)

the three roots of which

Wi = w,{k), i = 1, 2, 3 (2.5.38)

give the dependence of the frequencies on the wave vector for the corresponding
mode of free vibrations. At the same time, according to (2.5.2) all the modes
pass through the points w = 0, k = O.
To real values of wand k, correspond nondecaying waves of the type

vix, t) = uik)ei(OJI-kx) , (2.5.39)

which propagate with the group velocity

V,~(k) -def 'Owi(k)

'Ok), . (2.5.40)

Generally speaking, the group velocity depends upon the modulus of the wave
vector k, i.e. spatial dispersion takes place. The absence of nonzero real roots
of the equation wi(k) = 0 is a necessary condition of stability of the medium.
In fact, such roots would correspond to static displacements different from
translation and rotation, which do not contribute to the elastic energy; this is
not possible for a stable elastic medium.
However, as in the one-dimensional case, it is possible to have complex
roots of the function w,(k), which are caused by the action-at-a-distance. To
these correspond exponentially increasing (or decreasing) displacements ua(x),
which cannot be considered to be admissible functions for an infinite medium.
At the same time such solutions of the static equation should be taken into ac-
count in boundary-value problems, where they can make a considerable con-
tribution to the elastic energy by means of the forces which act in the boundary
region.
As a rule, for high frequencies only complex values of the wave vector can
satisfy (2.5.37). Solutions having such frequencies must decay, while propagat-
ing, but this is not connected with dissipation of energy, since for such waves
the energy flux is equal to zero.

2.6 Approximate Models

As in the case of a one-dimensional medium [1.12], when we consider fields

which change slowly enough over distances of the order of the characteristic
28 2. Medium of Simple Structure

scale parameter I, we can perform a transition to approximate models by

replacing the integral operators by differential ones. As we shall see, in this
situation, there are possibly some qualitative distinctions between the three-
dimensional medium and the one-dimensional one.
We start with the simpler case of a homogeneous medium with an operator
of elastic moduli cAal'fJ(k) which, as usual, will be assumed to be an analytic
function, i.e.

= .6 c~al'fJTj ... Tn(ikTl)

00

cAal'fJ(k) ... (ik Tn ) , (2.6.1)

n=O

where the cn's are real constant tensors, which can be explicitly expressed in
terms of force constants of the micro-model.
For example, in the zeroth approximation, taking into account that

~8fJ(XfJ = [oAol'~afJ(k)]k=O = - JXAXI'~afJ(x) dx - = vI .6 nAnl'~afJ(n) , (2.6.2)

o n

(the last expression is for a quasicontinuum) from (2.5.4), we have

(2.6.3)

which coincides with the expression for the elastic moduli, known in the theory
of a crystal lattice [2.2].
Thus, in the zeroth approximation we obtain the equations of the classical
theory of elasticity

(2.6.4)

with the elastic-constants tensor c8al'fJ. Note that this approximation is universal
in the sense that it is physically correct for an infinite medium, as well as for
boundary-value problems.
As we saw in the examples of one-dimensional models, when constructing
successive approximations, it is usually necessary to distinguish the cases of
infinite and finite media. In this section, we consider infinite media only. As
distinct form the classical theory of elasticity, for nonlocal and weakly nonlocal
theories this case has a fundamental significance.
Recall that in models of media with weak dispersion, which correspond to
the long-wavelength approximation, only the fields u(x) whose Fourier trans-
forms u(k) are localized in the region of small (as compared with [-I) wave num-
bers are strictly speaking, to be considered. This enables one to restrict oneself
to a finite number of terms in (2.6.1), or, equivalently in the expansion of the
functions of bounded support CAal'fJ(x) in a multipole series.
To a first approximation
(2.6.5)
 2.6 Approximate Models 29

The operator Hooke's law (2.5.20) takes the form

(2.6.6)

and the equations of motion (2.5.35) are differential equations of third order
with respect to space derivatives.
It is known [2.2] that such models in crystal optics describe the specific
phenomenon of a rotation of the polarization plane of a propogating electro-
magnetic wave. Media for which this effect takes place are called gyrotropic.
An anologous phenomenon is possible also for an elastic medium, when cfPi3< #-
O. Unfortunately, as far as the author knows, no experimental investigations of
rotating of the polarization plane of elastic waves in a gyrotropic medium have
been carried out up to now. At the same time, such an investigation must be of
interest, because gyrotropy is one of not very many qualitatively new phenomena
which arise already in taking into account weak spatial dispersion. This is con-
nected with the fact that, despite the small value of the second term in (2.6.5),
the effect accumulates over a sufficiently large path and hence the plane of
polarization rotates through a finite angle.
For media with central symmetry, as well as for the isotropic medium, we
have

rpaf3(x) = rpaf3( - x) = rpf3a(x) , (2.6.7)

and, as can be easily shown (Problem 2.5.5), Cfpf3< = 0, i.e. such media are not
gyrotropic. For such media the next representation after the zeroth one is the
second approximation

(2.6.8)

to which corresponds Hooke's law

(T.l.a(x) = (C$a pf3 + c~apf3TJ<2aTJa<2)Cpf3(x) . (2.6.9)

It is convenient to set

C2 = /2C2' (2.6.10)

where C2 has the dimension of ordinary elastic moduli and / is the characteristic
scale parameter. The latter can also be considered as a small parameter with
respect to which the expansion is carried out. The equations of motion (2.5.35)
are now written in the form

(2.6.11)

For an inhomogeneous medium, the elastic moduli Co, Ch C2 will be functions

of the point x, and it is natural to regard them as changing slowly over distances
30 2. Medium of Simple Structure

of the order of I. The connection between the elastic moduli and rpap(x, x')
or cJ.app(x, x') can be found, as in [Ref. 1.12, Sect. 2.4].
Thus the equations for the medium with weak dispersion belong to the
well studied class of equations with a small coefficient of higher derivatives.
It is necessary to take this coefficient into account when constructing solutions
as well as when determining the field of applicability. In particular, such equa-
tions are also connected with phenomena of the boundary-layer type.
In an obvious way, the equations for higher-order approximations can be
written down, but they have no significance, since they do not describe qualita-
tively new phenomena and may provide at most a small correction.
Equation (2.6.11) of a medium of simple structure with weak spatial
dispersion does not differ in essence from the equations of the so-called couple-
stress theory of elasticity with constrained rotation, which was developed
by a number of authors [2.3-5]. A superficial distinction is only in the form
in which the equations of motion are written down, since in the couple-stress
theory of elasticity a nonsymmetric stress tensor usually appears. Such a
distinction is connected with the different definitions of the stress tensor and
has mainly a methodological character: the equations of motion expressed
in terms of displacements coincide.
However, the above-accepted physical interpretation of the approximate
model (2.6.11) differs fundamentally from the usual interpretation of the
couple-stress theory of elasticity. The latter is introduced purely phenomenolo-
gically in analogy with the classical theory of elasticity, and the parameter I
is not regarded as small. 3 As a result, typical problems, which are often solved
in connection with couple-stress theories, concern calculations of stress con-
centration on inhomogeneties having dimensions of the order of or even less
than I. It is natural that values of the stress concentration can differ strongly
from values obtained on the basis of the classical theory of elasticity, and
this creates the illusion of a theory being significantly improved. The preceeding
analysis shows that these results should rather be considered as a brave ex-
trapolation of the approximate model (2.6.11) in a field in which, generally
speaking, only the strongly nonlocal model is applicable.

2.7 Cubic Lattice

As an example, let us consider a simple cubic lattice with interaction between

nearest neighbors. It is evident in that case that one cannot restrict oneself
"Note that when deriving the equations of the couple-stress theory of elasticity from a
consideration of equillibrium of the volume element, the existence of a scale parameter, i.e.
a nonlocal nature of the model, is not taken into account. At the same time, propositions
concerning the forces of short-range action and infinitesimal volume elements, which are
typical for such considerations, are justifiable only for the consistent local classical theory
of elasticity, where there is no scale parameter.
 2.7 Cubic Lattice 31

to the binary interaction (with the bond of the type of a spring) since such
a lattice would be unstable; it would be possible to fold up such a lattice without
changing the distances between neighboring knots. Therefore, for a lattice
with nearest-neighbor interaction, it is necessary to introduce bonds of the
rod type. In the simplest case, each of these rods would be characterized by
two parameters: the longitudinal and the transverse stiffnesses A; and B;,
respectively (i = 1, 2, 3).
Taking into account the above reasoning let us give, in the lattice coordinate
system, the matrix of force constants in the form (see also [2.6]):

°° 0,0, 1, ° 0,0,
0, 0, 1
-1, °
n 1, 0,
I -1, 0, -1
Al ° ° °° °° B2 B3
(2.7.1)
- vol!j)afJ(n)
°° ° ° °° ° ° °° ° °
Bl
Bl
A2
B2
B3
A3

The first condition from (2.3.17) gives (here and subsequently only nonzero
components are presented and there is no summation over identical indices):

(2.7.2)

From (2.3.15) it follows that the matrix !j)afJ(k) is diagonal. For its com-
ponents in the crystallographic system of coordinates we find (a =1= (3 =1= r)

!j)aa(k) = Aak~s2(ka) + B{3k~S2(k{3) + Brk~s(kr)'

(2.7.3)
s(k a) clef fa sin ~a ( -11: ::;; ka ::;; 11:) •

From these expressions, according to (2.5.2) we easily find the tensor !j)}.ap.{3(k),
which is symmetric with respect to the indices )" fJ-

(2.7.4)

The substitution of components of the tensor !j)}.ap.{l(O) into (2.5.8) shows

that the conditions of invariance with respect to rotation are not fulfilled.
This should not cause a surprise, since for real rod systems the elastic energy
depends not only on displacements of the ends of the rods but also on the
angles of their rotation. Consideration of such a system in the scope of the
model of simple structure is equivalent to postulating that the angles of rotation
of the rods' ends are equal to zero. This means that masses located at the knots
are displaced in slots which prevent their rotation. Howevel, such a con-
struction is not, in general invariant with respect to rotation.
For an adequate description of the system of rods a model of a medium
with complex structure is needed, as we have seen in [Ref. 1.12, Chap. 3], when
32 2. Medium of Simple Structure

considering one-dimensional models. The corresponding generalization to

the three-dimensional case will be carried out in the following chapter, and
here we make use of the incidental fact that for rods with equal transverse
stiffness, i.e. for Ba = B p, the conditions (2.5.8), of invariance with respect
to rotation, are satisfied automatically. Hence this model admits a formal
consideration in terms of the medium with simple structure, but one should
not expect all results to be valid for a real-rod system.
Thus, assuming that the conditions Ba = Bp = B are satisfied let us con-
struct the expression for the operator of elastic moduli c),ap.p(k), following
the procedure which was considered in Sect. 2.5.
To this end, first of all, let us find the tensor afP.P, which is defined by
(2.5.18). Its nonzero components have the form

afaaa = Aas2(k a), afafjfj = - BS2(ka) ,

(2.7.5)
aqaafj = afPa{3 = BS2(k a) .

For the components (2.5.20) which are symmetric and scewsymmetric with
respect to the second pair of indices we obtain

a'fP(afj) = i B[S2(k a) + s2(ka)] ,

= i
(2.7.6)
a'fPCaPl B[s2(k a) - s2(k p)] •

According to (2.5.26) for the components a~ap.fj we have

(2.7.7)

The pentavalent tensor afp.Pr of fifth rank, which is defined by (2.5.23),

has only the following nonzero components:

(2.7.8)

Substitution in (2.5.28) gives

a~fJafj = i B[S2(ka) + s2(k p) - 2] ,

(2.7.9)
(fl,aafJ = 1-. Bk 1 - s2(ka)
3 2 P ka .

Finally, according to (2.5.30), for nontrivial components of the elastic

moduli operator c)'ap.p(k) we may write down the expressions
 2.8 Isotropic Homogeneous Medium 33

(2.7.10)
capap = B[S2(ka) + s2(k p) - 1],

caapp = - i B[S2(k a) + s2(k p)] .

Note that all the computations here were carried out in the lattice coordi-
nate system and therefore the components of the tensor c"app have the dimen-
sion [Nm-6]. In order to obtain the usual dimension, i.e. [Nm-2], it is neces-
sary to consider the mixed components d~p. Since gap '" 12, where 1 is the
length of a rod, equal to the parameter of the cubic lattice, we have

(2.7.11)

where the last equality is valid only in the lattice coordinate system.
For the elastic-constants tensor in the zeroth approximation, we find

cg,:.pp == AOOlaOpp + f1.0(01P'P + O~O~) + z: riOiAOiaOipOiP'

I
(2.7.12)

where

f1.O =- AO = [4B, ri = 14(Ai - B). (2.7.13)

Note that the different signs of the Lame constants Ao and f1.0 additionally
indicate the artificial nature of the model and, in particular, the fact that the
model does not correspond to a system of rods. In fact, in the latter, a side
thrust should not result from tension along a lattice axis, and this is possible
only if Ao = O. In the next chapter we shall see that in a correct consideration
of rod systems in terms of media of complex structure, such contradictions
do not appear.

Problem 2.7.1. Write down the elastic constants in the second approximation.
Problem 2.7.2. Taking into account (2.7.3), derive an expression for the dis-
persive modes w,{k) (j = 1, 2, 3).

2.8 Isotropic Homogeneous Medium

As a start, let us define our notation in the case of isotropic media. The position
vectors of the physical and Fourier spaces will be denoted by rand k, respec-
tively. Their moduli are denoted by rand k, i.e. r = Irl, k = Ikl. Note also
34 2. Medium of Simple Structure

that in the three-dimensional case, a function <p(r) and its Fourier transform
<p(k) are connected by the following relations

(2.8.1)

<p(r) = - 4~2r r
-00
k<p(k)e irk dk , (2.8.2)

in the presence of poles, the appropriate regularization of the integrals is neces-

sary.
Obviously, for an infinite isotropic medium, in cartesian coordinates, the
operator <J)afJ(k) may be constructed only by means of combinations of the
tensor (PI' and the components of the vector k. The most general expression
which satisfies the conditions (2.3.13,18), has the form

<J)afJ(k) = fJ,(k)k2oafJ + [)'(k) + fJ,(k)]kakfJ , (2.8.3)

where )'(k), fJ,(k) are real scalar functions. When the action-at-a-distance is
finite, ), and fJ, are analytic functions of the vector k, and their expansions with
respect to k can contain only terms with k 2 (as a consequence of isotropy), i.e.
)'(k) and fJ,(k) are even analytic functions of k.
It can be easily seen that (2.8.3) for the operator <J)afJ(k) can be associated
with the corresponding operator of the elastic moduli

(2.8.4)

which has the same structure as the tensor of elastic moduli in the local theory
of elasticity. However, as distinct from the local theory, this representation for
clal'fJ(k) is not unique. In fact, without destroying the symmetry, one can add
to (2.8.4) terms of the form

which would only change the connection between the scalar coefficients of the
tensors <J)afl(k) and clal'fl(k). Thus, we have here one more example of nonuni-
queness of the operator of elastic moduli (and hence the stress tensor) in a non-
local theory.
The equations of motion (2.5.33) in direct notation have the form

pu - M tlu - (A + M)grad div u = q , (2.8.5)

which is identical with the equation for an isotropic medium in the local theory,
but with substitution of the Lame constrants ),0, fJ,0 by scalar operators A, M
 2.8 Isotropic Homogeneous Medium 3S

with the kernels il(r), p(r). The latter are inverse Fourier transforms of il(k),
p(k). Note that the nonuniqueness of the operator of elastic moduli cAap{J does
not influence the equations of motion whatsoever: these are fully determined by
the elastic-energy operator f/Ja{J.
In order to obtain solutions of the equations of motion it is sufficient to
construct the Green's tensor G{r, t). Formally this can be acheived by replacing
the Lame constants ih the Green's tensor of the local theory of elasticity by
the corresponding operators. For example, for the static Green's tensor we
have

_ 1 ( a{J _ il{k) + p(k} kak{J)

GaP(k) - k 2p{k) 0 il(k) + 2p(k) k2 . (2.8.6)

However, this representation is not very suitable for the investigation of specific
nonlocal phenomena. Consequently, we shall consider another approach.
An arbitrary vector function uP(k) can be decomposed into its longitudinal
(I) and transverse (t) components, as a result of the projection operators

(2.8.7)

In the r-representation this corresponds to the decomposition of the vector

field up(r) into potential and rotational parts.

Problem 2.S.1. Check that in the r-representation

1C = - _1_ * grad div {) = - 1

4 * rot rot. (2.8.8)
41Cr ' 1Cr
Note that f/Ja{J(k) can be represented in the form

(2.8.9)
where

(2.8.10)

Applying the operators 1C and {) to the equations of motion we find that these
equations are decomposed into independent equations with scalar operators

[ - pw2 + f/J j(k)] uJ(k, w) = qj(k, w), j = 1, t (2.8.11)

with respect to the longitudinal and transverse components of the displacements
and forces.
The dispersion equation (2.5.37) is also decomposed into an equation for
the longitudinal mode
36 2. Medium of Simple Structure

(2.8.12)

and a doubly degenerate equation for the transverse mode

(2.8.13)

To the decomposition (2.8.9) of the operator (/Ja~(k) into two orthogonal

components corresponds the representation of the Green's tensor Ga~(k, (V)
in the form

(2.8.14)

where

1 . 1 (2.8.15)
-P-'[(V"7(7ik~)---(V"'2] , J = ,t.

If the action-at-a-distance is finite, the analytic properties of the entire

functions (/Jj(k, (V) = - p(V2 + (/Jj(k) are identical to the properties of the func-
tion (/J(k, (V) for a one-dimensional model. This allows us to use the method
developed in [Ref. 1.12, Sect. 2.8, 9] for constructing the merom orphic functions
Gi(k, (V), but in order find their inverse Fourier transforms Gj(r, (V), it is neces-
sary to apply the three-dimensional inversion formula (2.8.2).
Analogous to (I.2.8.16) we have

(/J/k, (V) = - p(V2 mfi_o (1 _ kr,2m«(V)

k2 )
,
(2.8.16)

where, as (V -+ 0,

p(V2 2 p(V2
kr,o «(V) -+ ). +2 ,kt,o«(V) -+ - - . (2.8.17)
o flo flo

The inverse functions Gi(k, (V) = (/Jjl(k, (V) can be represented in the form
([Ref. 1.12., Eq. (2.9.17)])

Gi(k ) - 2 ~ k i,2m«(V) (2.8.18)

,(V - ),2m())
£....j d\~(k.
m=O 'V) (V k2 - k2i,2m((V) '

and hence, in view of (2.8.2), we obtain

GJ·(r, (V) = -4- ~

1 £....j k i2m«(V)
d\~(k. ()) exp ['k ()]
I i,2m (V r . (2.8.19)
1(;r m=O 'V) ),2m (V

Thus for the isotropic medium, the structure of the solutions is determined
 2.9 Debye Quasicontinuum 37

by the distribution of roots of the functions l/Jj(k, m) as was also true in the one-
dimensional case.
When proceeding to models with weak nonlocality, we used the long-wave-
length approximation. However, for isotropic media, in a number of cases, it
may be more suitable to use an approximation with respect to the first roots
(Sect. 2.10). In this approximation, in the static case (m = 0)

(2.8.20)

where II:j = k j ,2 and Im{II:j} > O.

From the above expressions it follows that together with the classical
asymptotic behavior - r- 1 there are also terms _,-3 and terms which decay
exponentially. Note that for one-dimensional media, the asymptotic, behavior
contained only terms -lxl- 1 and exponential terms.
When passing to the local theory, it can be assumed that Im{II:j} --+ 00,
and the Green's tensor coincides with the usual one.

2.9 Debye Quasicontinuum

In the preceding section, we considered the isotropic elastic continuum with

nonlocal interaction. In order to obtain a correct qualitative description of the
discrete structure, and to conserve isotropy simultaneously one can use the
Debye quasicontinuum model. All the formulae (2.8.1-15) remain valid, but
the vector k has to satisfy the additional condition k ~ II: where II: ~ 7C/a is
the Debye radius and a is the parameter of the discrete structure.
The role of the a-function is now played by air), whose Fourier transform
is equal to 1 when k ~ II: and equal to zero when k > 11:. Applying (2.8.2),
we obtain

air) = - II:2 (sin II:r

2 2 -- - cos II:r ) • (2.9.1)
7T: r II:r

We shall also need the Green's functions of the Laplacian g(r) and bi-
Laplacian h(r), i.e. the inverse Fourier transforms of - k 2 and k4, which satisfy
the condition
/:::..2h(r) = /:::..g(r) = air) . (2.9.2)
Simple computations yield
38 2. Medium of Simple Structure

g(r) = - - 12
2 Si(Kr), (2.9.3)
'ltr

h( r ) = - - 1
42
( .
KrSl(Kr) sin,.,
+- -+ cosKr ) , (2.9.4)
'It K Kr

where Si (x) is the integral sine.

When K -4 00 (a -4 0) or when r -4 00, all these expressions are transformed
into the ordinary ones.
The Green's tensor in the r-representation can be constructed, as in the pre-
vious section, with the help of the expansion with respect to the roots. However,
in the case of a quasicontinuum there is another possibility which is connected
with a suitable approximation of the functions (llj(k) over the segment 0 ~
k ~ K.
Various approximations for (J)j(k) which are used in the theory of specific
heat [2.7], are known. Thus in the simplest Debye model it is assumed that

(2.9.5)

where Vb VI are the logitudinal and transverse velocities of sound in the cor-
responding local model of the elastic continuum. It is evident that this model
does not take into account the spatial dispersion.
In a more realistic Born-Karman model, a dispersion law is accepted, which
is analogous to the case of a chain with interaction between nearest neighbors

2KV j 'ltk
(J)' (k)
.
J
= -~
'It
SIn -
2K
. (2.9.6)

The natural conditions

(2.9.7)

are then fulfilled.

For our purposes the polynomial approximation of the form

(2.9.8)

is more convenient, where P j(k 2) is an appropriate polynomial, which satisfies

the conditions (2.9.7.) The simplest polynomial is

(2.9.9)

it depends on the parameter n. The corresponding family of dispersion curves

 2.9 Debye Quasicontinuum 39

is shown in Fig. 2.1. The straight line corresponds to the Debye model where
w~ = vi" is the Debye frequency. The dotted curve corresponds to the Bom-
Karman model and, as is clear from the picture, practically it coincides with the
curve for rj = 0.5.

1.0 ~

wJwf· ,//
//
I} 0
/1/
/'"
,--

If y --r
/

f/j/
j
1/..-- y(1O

I
If Fig. 2.1. Dispersion curves
o k/x !.O

The parameter n may be connected with the ratio of the boundary frequency
Wj(") to the Debye frequency w~, i.e.,

(2.9.10)

from which the condition - 2 < 7i < 00 follows.

The approximation (2.9.8) allows us to construct an explicit expression for
the static Green's tensor. From (2.8.7, 14 and 15) with W = 0, we have

(2.9.11)

Taking into account (2.9.2, 8) we finally obtain

(2.9.12)

(2.9.13)
40 2. Medium of Simple Structure

where

hj(r) def P j( - fl)h(r), gj(r) = P j( - fl)g(r) , (2.9.14)

and h(r) and g(r) = M(r) are determined by (2.9.3, 4).

In particular, for the model (2.9.9)

(2.9.15)

and for the Debye model (2.9.5) we have P j( - fl) = 1.

An important property of the constructed Green's tensors is the absence of
a singularity at r = 0, since Ga~(r) is an even analytic function. This corresponds
to a physical meaning of the non local theory for a medium with an elementary
unit of length, which models the discrete medium and formally is an effect of
truncating the Fourier spectrum. As a result, all the divergencies connected
with point sources of force, which are inherent to the model of an elastic con-
tinuum are eliminated.
Let us consider for example the concentrated force

(2.9.16)

where Qa = const, and a(r) can also be replaced by air) since due to the trunca-
tion of the Fourier spectrum of the Green's tensor, the final result is the same.
This reasoning enables us, in what follows, to make no distinction between the
"true" a-function and the a-function with the truncated Fourier spectrum,
when we consider the quasicontinuum model.
For the energy of the elastic field of a concentrated force, taking into account
(2.3.16), we have

(2.9.17)

where the (Xi are dimensionless constants, which depend on the form of the
functions oJj{k),

(2.9.18)

For the Debye model and for the model (2.9.9) we find

Let us consider the asymptotic form of the Green's tensor, as r ~ 00. It is

easily seen that this is determined by the asymptotic behavior of the function
hk). The latter has the form
 2.10 Boundary-Value Problems and Surface Waves 41

hj{r) ~ - 8r
10
(1 - 22Ti2 ).
ICr
(2.9.19)

Substitution in (2.9.12) and (2.9.13) yields the dominant terms of the asymptotic
form of Gap(r) to be of the order of ,-1 and ,-3, which, as it should, coincides
with the asymptotic form of the Green's tensor for the elastic continuum with
action-at-a-distance examined in the preceeding section. As IC ~ 00, the Green's
tensor is transformed into the Green's tensor of the local theory of elasticity
and (2.9.17) diverges linearly with IC.

2.10 Boundary-Value Problems and Surface Waves

The results obtained in [Ref. 1.12, Sects. 2.6, 11] can be transferred to the three-
dimensional case but here some specific three-dimensional effects arise; in par-
ticular, special types of surface waves appear.
Let us restrict ourselves to a consideration of the simplest type of model of
a medium with binary interaction. In this case the operator of the elastic bonds
1jfap has the form of(2.3.8). Repeating almost literally the arguments carried out
earlier for the one-dimensional case, one can introduce the three-dimensional
operators (/Jr!, raP(w), etc., write down the Green's formula and formulate the
basic boundary value problems. In particular, in the obvious notation (see the
notation [Ref. 1.12, Sect. 2.6]), the Green's formula has the form

(2.10.1)

The equations of the first basic problem take the form

(2.10.2)

and the boundary conditions are written in the form

(2.10.3)

where, for example, the kernel

raP(r, r') = S(r)o(r - r')J1jfa{l(r, r")dr" - S(r)1jfap(r, r')V(r') (2.10.4)

v

corresponds to the operator raP.

Note that in the above formulation of the problem, the conditions of in-
variance with respect to rigid translations and rotations of the medium, are
automatically satisfied.
Since the radius of the action-at-a-distance is usually small in comparison
with other characteristic dimensions, th~ main interest in the theory of the
42 2. Medium of Simple Structure

elastic medium with action-at-a-distance is in the boundary-value problems for

a half space (half plane).
As an example, let us consider the problem of propagation of surface waves
in a homogeneous isotropic medium with binary interaction, for which

lffa{3(r, r') = lffafJ(r - r'), lffafJ(r) = x a: fJ Iff(r). (2.10.5)

r

The equations of the problem can be also represented in the form

(2.10.6)

where Uj denotes either the longitudinal or the transverse component UI or Ut ,

and Wj is the corresponding operator (/)1 or Wt •

Problem 2.10.1. Show that for the model under consideration.

(2.10.7)

where the functions Iffb Iff t and Iff are connected by the relations

I
/fft(r) = Sr-1/ff(r) dr . (2.10.8)
r

Problem 2.10.2. Show that the interaction of binary type implies the equality
of the Lame constants

271: I
AO = flo = -[5 Sr 4/ff(r) dr . (2.10.9)
o

Let us now assume that the medium occupies the upper half space and
choose the coordinate system so that the z-axis is directed normalIy to the
surface. The boundary conditions, corresponding to the absence of forces in
the surface layer S(z), have the form

- (J}pua(r) + Sra{J(r, r')uir') dr' = 0, r E S(z) . (2.10.10)

Let us consider a plane surface wave propagating along the x-axis, i.e. let
us seek a solution of the form

U(x, z) = eiPx v(z) , (2.10.11)

 2.10 Boundary-Value Problems and Surface Waves 43

where v(z) --+ 0 as z --+ 00, and p is a real parameter. It is easy to see that in
this case two types of surface waves can propagate independently in the medi-
um. For a wave of the first type the displacement vector lies in the plane, which
is drawn in the direction of propagation perpendicularly to the surface, and
which satisfies the boundary conditions

- (J)2pU" + Fllu" + F13U• = 0,

(2.10.12)
- (J)2pu• + F13u " + F33U• = 0.

For a wave of the second type, the displacement vector is parallel to the z-axis
and satisfies the condition

(2.10.13)

The formal decomposition of the boundary conditions is a consequence of

the fact that, for the given model, the following integrals vanish:

Jl/fa2(r) dy = x a J ~ l/f(r) dy
r
= 0 (0: -:F 2) . (2.10.14)

We carry out the further analysis, restricting ourselves to an approximate

consideration of the problem. In the zeroth approximation the conditions
(2.10.12) are transformed into usual boundary conditions for Rayleigh waves.
The condition (2.10.13), as I --+ 0, yields

au,
az = 0
. (2.10.15)

On account of (2.10.11) it follows that in the zeroth approximation

u, = O. (2.10.16)

Thus the wave of the second type does not have any classical analogy.
Let us proceed to the approximation with respect to the first root. The
equations of motion in this approximation take the form

(2.10.17)

where c2 is equal to 3p.o/p or to p.o/p, A = k l,2 or A = k t ,2 depending upon the

type of the waves under consideration (longitudinal or transverse).
Let us seek v(z) in the form of a linear combination of exponentials exp(i,tz),
where" is obtained from the condition that u(x, z) = exp[i(pA + I>z)] is a
solution of (2.10.17). For u to be a surface wave, the fulfillment of the additional
condition 1m I> > 0 is necessary.
44 2. Medium of Simple Structure

Substitution in (2.10.17) leads to an equation of sixth order with respect to "

(2.1O.IS)

At small values of lV, for the roots of the above equation we obtain

The root "0 corresponds to the usual Rayleigh wave and the roots "1,2
correspond to the new surface waves, which are absent in the classical theory.
With p ,.., lV/C, we have

1m "0 ,.., .!!!.-, 1m "1 = 1m "2 ,.., 1m A ,.., 1-1 . (2.10.20)

C

For the complete solution of the problem it is necessary to write down

the boundary conditions (2.10.12, 13) in the appropriate approximation. This
will enable us to find the coefficients of exp(i"z) and to obtain the dispersion
law of the surface waves in the given approximation. We omit these cumbersome
computations. However, even from the relations shown above, it follows that
in the nonlocal theory, new types of surface waves exist which, in contrast to
the Rayleigh waves, decay in the case of large wavelength, at a distance of the
order of the radius of interaction.

2.11 Notes

The considered model of quasicontinuum introduced independently by Krum-

hansl [B3.19], Kunin [B3.22] and Rogula [B6.36] can be used in linear field
theories only. A space quantization which can be applied to nonlinear field
theories was considered in [BS.2S, 29].
Various problems of the nonlocal elasticity were considered in the works of
Kroner [B3.13-16], Kroner and Datta [B3.17, IS], Kunin [B3.20-21], Mindlin
[B.3.29], Edelen, Green and Laws [B3.1], Eringen and Edelen [B3.2], Eringen
[B3.3-6]. Note the work of Soos [B3.30], where the case of slowly decreasing
(according to the Coulomb law) forces qf action-at-a-distance is considered.
The existence of surface waves, which have no analogy in classical elasticity
was pointed out in the work of Kunin and Vaisman [B3.2S, 31] (see also Kaliski
and Rymarz [B3.S]).
3. Medium of Complex Structure

The medium of complex structure is considered in detail with the main at-
tention being focused on specific three-dimensional effects.
Starting from the harmonic model of the crystal lattice with a basis, we
introduce the collective variables: the mass center displacement, micro rotations
and microdeformations. In terms of these variables, the general nonlocal equa-
tions of the medium of complex structure are obtained. The physical require-
ments of stability and invariance with respect to rigid body motions lead to the
existence of acoustical and optical modes of vibrations. It is shown that in the
corresponding acoustical region the microrotations and microdeformations
can be excluded from the equations of motion. As a result, the medium of com-
plex structure can be described in the acoustical region in terms of the medium
of simple structure with space and time dispersion (dynamical elastic moduli).
The detailed comparison of the theory with approximate couple-stress
theories is provided. It is shown that the latter have a very restricted range of
applicability. As an example of the medium of complex structure the multibar
system is considered in detail.

3.1 Equations of Motion

The general formalism of the three-dimensional medium of complex structure is

very similar to the one-dimensional case. The most important distinctions are
connected with replacing the scalar variables and operators by tensor quantities.
To avoid repetition, we shall take into account this circumstance when generaliz-
ing the results, obtained in [Ref.1.12, Chap. 3] for the one-dimensional model.
As a starting point for a micro-model of the medium of complex structure,
let us consider a model of an infinite complex crystal lattice in the harmonic
approximation. The introduction of collective cell variables will enable us later
on to include this model, as a particular case, in the general phenomenological
scheme.
A geometrically complex lattice is a three-dimensional discrete periodic
structure with an elementary cell, constructed on the basis of the vector ea(a =
1, 2, 3). The positions of N particles with masses mj(j = 1, ... , N) inside the
cell are given by a set of vectors ej = ei ea. The interaction between particles
both inside the cell and with particles belonging to other cells is given by the
force constant (/)a{j(n, n', j, j'), where n = naea is a vector which characterizes
the number of the cell. In the oblique system of coordinates x a with the basis
46 3. Medium of Complex Structure

ea and the metric tensor ga{3 = ea ·e{3' the vector n has the integer component

In the absence of initial forces, the Lagrangian of the lattice has the form

2L = ga{3 ~m/.vaCn,j)win,j)
n;

- 1: waCn, j) (/Ja{3(n, n',j,},) win',),) + 2 1:fa(n, j)wa(n, j), (3.1.1)

n~ ~

where wa(n, j) is the displacement of the ph particle in the nth cell, and fa(n, j)
is the external force which acts on it.
Let us generalize the model in such a way that it would be possible to take
into consideration both discrete and continuous periodic structures. To this
end, let us introduce the density p(~) of the mass distribution within the cell
and a force matrix (/Ja{3(n, n', ~, e) which determines the interaction. Here,
~ = ~aea is the position vector of a point of a cell with the number n in the
local system of coordinates with the origin situated at the center of mass of
the cell. Then the Lagrangian is given by

2L = ga{3 1: p(~)wa(n, ~)w{3(n, ~)d~

n

- 1:
nn'
SJ wa(n, ~)(/Ja{3(n, n', ~, ~')w{3(n', e) d~ d~'

+ 2 1: Jfa(n, ';)wa(n, .;) d~ . (3.1.2)

n

In order to pass to the discrete structure it is necessary to set

(/Ja{3(n, n', .;, .;') = f1 (/Ja{3(n, n', j, }')o(.; -

;;
';j)o(e - I;j') . (3.1.3)

Using the formalism of the quasicontinuum let us write down the Lagrangian
(3.1.2) in the x representation

2L = ga{3 JJp( .;)wacx, .;)w p(x, .;) d'; dx

- SJ SJ wa(x, ~)(/Ja{3(x, x', .;, .;')wp(x', .;') d'; de dx dx'
+ 2 II fa(x, ';)wacx, .;) d'; dx . (3.1.4)

As in the case of a one-dimensional structure, it is convenient to introduce

collective cell parameters in order to pass to the theory of elasticity. To this end
let us define a set of moment-of-inertia tensors of order s (s = 0, 1, ... ) of the
mass distribution in the cell
 3.1 Equations of Motion 47

where Vo is the cell volume. In an abreviated form

(3.1.6)

Here, s denotes the multi index ih ... As and the parentheses denote the cor-
responding scalar product.
Let us denote by P2; the tensor inverse to P2s and introduce the two quasi-
diagonal matrices

(3.1.7)

where oSs' and oss' are Kronecker deltas.

By a method analogous to the one considered in [Ref. 1.12, Chap. 3], let us
construct an orthonormalized system of basis polynominals es(';) = eAl--- As (.;)
with weight p(';) and a reciprocal system of functions es'(';) == el'l ___ I',,(';)' defining
them by the relations

(3.1.8)

Problem 3.1.1. Verify that

(3.1.9)
Problem 3.1.2. Taking into account the fact that the origin of the coordinate
system coincides with the mass center of a cell, show that for the first two
elements of the bi-orthogonal basis, the following relations are valid:

eO(';) = 1,
(3.1.10)

Let us expand the functions of the variables .;, .;' entering (3.1.4) in terms of
the bi-orthogonal basis

Wj3(x, .;) = wsix)es(';), fa(x, .;) = fsa(x)esC';),

(3.1.11)
rpaj3(x, x', .;, .;') = rpsas'j3(x, x')es(';)es'(.;') .

For the coefficients of the expansion, taking into account (3.1.8) we find

wsj3(x) = (e" wj3(x», fsa(x) = (e s, fa(x»,

(3.1.12)
rpsas'j3(x, x') = V~ Sf rpaJ3(x, x', .;, .;')es(';)eS'(';') d'; d.;'.
48 3. Medium of Complex Structure

In the new set of variables, the Lagrangian (3.1.4) can be written in a form
which is invariant with respect to the n, x and k representations:

(3.1.13)

where pas'fJ def gafJps'.

If we forget about the quasicontinuum, then this Lagrangian describes the
most general linear model of a three-dimensional medium of complex structure.
The corresponding equations of motion have the form [Ref. 1.12, Eq. (3.3.4.]

Jpas'fJ(x, x') ws,p(x') dx' + JJf/)saS'f3(x, x') WS'fJ(x') dx' = fsa(x) , (3.1.14)

where, in the general case, pas'f3(x, x') is the kernel of a nonlocal moment-of-
inertia operator.
Let us now consider the physical meaning of the collective cell variables
was(x) andfsa(x). In view of (3.1.10,12) we find that WOa is the displacement
of the mass center of the cell and fOa is the average body force density. It is
clear that these quantities will have a considerable significance when the medium
is described macroscopically. Hence, let us introduce the special notation

(3.1.15)

The remaining kinematic and force variables, which correspond to the

internal degrees of freedom, will be denoted by

"pa
'Yl
-def Wpa'rIIPa - , p = 1" 2
deffPa (3.1.16)

Let

1}1a == 1}).a = e).a + 0;" ,

,
e).a = def
1}().a) , Q').a =def 1}[).a] • (3.1.17)

It is easy to verify that e~a is the average microdeformation of a cell and

Q~a = - O~). is the average microrotation. Analogously,

(3.1.18)

where p.().a) is the average density of force dipoles and m},a = - ma). is the aver-
age density of micromoments. It is clear that the p.().a) and m).a are generalized
forces, which correspond to the generalized displacements e5.a and Q;". The
quantities 1}pa and p'pa, for p > 1, describe the microdeformations and micro-
moments of higher orders.
Note that the number of independent elements of the basis is determined
by the number of degrees of freedom of a cell. If the number of degrees of
 3.2 Energy Operator 49

freedom does not exceed 12 (in the case of a discrete structure, this means that
the number of particles in a cell does not exceed 4), then the basis is automati-
cally restricted to the first two elements. In this case, the role of kinematic
variables will be played by displacements of the center of mass u, micro-
deformation e' and micro rotation Q'; and the role of force variables will be
assumed by the body force density q, the dipole force density J1- and the micro-
moments density m.
If the cell consists of one particle with a finite moment of inertia, then the
internal degrees of freedom are restricted to a micro rotation Q'. This simplest
model of the medium of complex structure, known as the Cosserat medium,
will be considered in Sect 3.5.
In terms of the new variables, the equations of motion (3.1.14) take the form

pga{Jii{J + (/Ja{Ju{J + (/J+'aP'{J rJp'f3 = qa,

(3.1.19)
[pap'fJ rJ P'fJ + (/JpafJu{J + (/JPaP'fJ rJ P'fJ = J1- pa ,

where the operator coefficients are connected with the operators in (3.1.14)
in an obvious way.

3.2 Energy Operator

From (3.1.13, 14) it follows that the properties of the medium are com-
pletely determined by the kinetic energy operators pas'fJ and the elastic energy
operators (/Jsas'{J. In this section we consider the structure of these operators.
According to a reasoning analogous to that of [Ref. 1.12, Sect, 3.3] let us
assume that the operator [sas'fJ is determined by a set of constant moment-of-
inertia tensors, the following ones having the principal significance:

[OaO{J = pga{J, [lal{J == [a).p{J • (3.2.1)

From this assumption, in view of (3.1.13) we obtain the symmetry property

(3.2.2)

It is evident that it is also necessary to require the positive-definiteness of

the kinetic energy.
Let us now consider the properties of the elastic-energy operator and their
consequences.
Hermiticity: From (3.1.13), we have

(/Jsas'fJ (x, x') = (/Js'{Jsa(x', x) (3.2.3)

or in the representation (3.1.19)

50 3. Medium of Complex Structure

</Jall(x, x') = </Jlla(x', x), </Jpap'f3(x, x') = </Jplf3pa(x', x),

(3.2.4)
</J+aPIl(x, x') = </JPf3 a(x' , x) .

Boundedness of action-at-a-distance is assumed as before. The analytical

properties of the kernels in various representations, which follow from this,
are already well-known.
Invariance with respect to translation and rotation. Let the corresponding
displacements be w~a' Then for any w sa the condition of invariance of energy
must be satisfied:

(3.2.5)

Together with (3.1.13) this yields

(3.2.6)

or, due to the arbitrariness of wsfl:

(3.2.7)

For a translation by the vector aa

W~(x) = aa' w~a(x) = 0, P = I, 2, .... (3.2.8)

For a rotation determined by the tensor a).a = -aa).'

(3.2.9)

with all other wsaCx) remaining unchanged.

Problem 3.2.1. Obtain (3.2.8, 9) from the laws of transformation of waCx, ~),
taking into account (3.1.10, 12).

From this it follows that the conditions

S</J0asf3(x, x') dx' = 0, (3.2.10)

S[x[).</JOalsf3(x, x') + </J[},al sfl(x, x')] dx = 0 (3.2.11)

are to be safisfied. In the k-representation (a). = a/ok).), these become

</J0asf3(O, k') = 0,
(3.2.12)
ia[},</J0al sf3(O, k') + </J[},alsfl (0, k') = 0.
 3.2 Energy Operator 51

This enables us to represent the matrix rpsas'fJ(k, k') in (3.1.19) in the form
[Ref. 1.12, Eq. (3.3.15)]

rpsas'fJ(k k') _ (k;.k~r).apfJ (k, k') - ik;.X+).ap'fJ(k, k'») (3.2.13)

, - ik~XpapfJ(k, k') rPaP'fJ(k, k') ,

where the functions X and X+ are connected by the relation

XHaPfJ(k, k') = XPfJ).a(k', k) (3.2.14)

and satisfy the conditions

r).a[pfJ J(k, 0) + X+;'a["fJJ (k, 0) = 0,

(3.2.15)
Xpa ["fJJ (k, 0) + rPaC"fJJ(k, 0) = 0 .

Problem 3.2.2. Prove (3.2.13). [Hint: Use the Hermitian adjointness and the
condition (3.2.12).]

The equations of motion (3.1.12), after taking into account (3.2.13), become

pgafJUfJ - '(hr).api3opufJ - o;.XHaPfJ7}PfJ = qa ,

(3.2.16)
IPap'fJ~p'fJ + Xpapf30pufJ + rpap'fJ7}P'fJ = p'Pa,

where r, X, r are the corresponding operators.

Stability is equivalent to the requirement of the positive definiteness of
the elastic energy for all generalized displacements except translations and
rotations. As a consequence, the positive definiteness of the matrices r;'a pfJ
and rPaP'fJ follows. In fact, the elastic energy rp can be represented in the form
of a sum of three terms

(3.2.17)

rpo def i <o;,ua Ir AapfJ IopufJ) ,

rpl def i <7}p;' IrPap'fJ l7}p'fJ) ,

(3.2.18)

rpint !kf <7}pa IXpapfJ IopufJ) .

Here, rpo and rpl are the elastic energies, corresponding to the displacements
of the centers of masses and to the internal degrees of freedom and rpint is
the interaction energy. It is clear that the terms rpo and rpl are to be positive
definite forms.
Homogeneous Medium. As we know, in this case all the operators have dif-
ference kernels in the x representation and are reduced to a product of functions
of k in the k representation [Ref. 1.12, Eq. (3.9.7)]. The matrix (3.2.12) becomes
52 3. Medium of Complex Structure

(3.2.19)

When writing down the above matrix, the conditions of invariance of the
energy with respect to translation and rotation were used. However, additional
conditions, analogous to (3.2.15), are also to be fulfilled. One of them is obtained
immediately from the second relation in (3.2.15)

(3.2.20)

The simple form of the limiting transition is here due to the fact that the
tensors XPal'i3 and r Pa l'[3 do not possess any symmetry, which would be destroyed
by alternating with respect to the indices /1(3.
The analog of the first relation of (3.2.15) produces more difficulties. When
passing to the limiting case of a homogeneous medium, a new symmetry with
respect to the indices A/1 arises in the tensor r).al'i3, which does not commute
with the alternation with respect to /1(3. Therefore, as in the case of a medium
with simple structure, we have to obtain this condition directly for a homo-
geneous medium. The corresponding discussion was carried out in details in
Sect. 2.5. This enables us to restrict ourselves to a few principal points. Hence,
it may be worthwhile, with a view to brevity, to stress only the most salient
features.
We start with a requirement of invariance of the energy density of the
zeroth approximation ({Jo(x) with respect to rotation. Taking into account
(3.2.19), the density of the Lagrangian in the zeroth approximation can be
represented in the form

((J~(X) = 1 chuix) r6al'[3oI'U[3(X)

+ 7jpa(X) xlbal'i3 0l'u[3(X) + 1

7jpa(X) rlffP'[3 7jpl[3 (x) , (3.2.21)

which is already invariant with respect to translation. The energy density

({Jo(x) differs from ({J~(x) by divergence terms, which are to be chosen from
the conditions of the invariance of ({Jo(x) with respect to rotation.
It is not difficult to verify that the most general expression which can be
represented as a divergence and, which is invariant with respect to translation,
has the form

where b).al'i3 = - bl'a).[3 is a constant tensor.

For ({Jo(x) being invariant with respect to rotation, it is necessary and
sufficient to require that
 3.2 Energy Operator 53

(bJ.ap.{J + bp.{JJ.a + rtrp.{J + X'tpa)[p.{Jl =0. (3.2.23)

At last, analogously to the treatment given in [Ref. 1.12, Sect. 1.5], it can
be shown that the condition of solvability of (3.2.23) with respect to bJ.ap.{J
has the form

[Aa, ,u,81 = [aA, {3,u] , (3.2.24)

where
[Aa, ,u,81 = r~ap.[3 + (X~ap.[3 + X~{J).a + r~ap.(3)().p.) . (3.2.25)
Under the condition (3.2.24) the solution of (3.2.23) is unique

(3.2.26)

Thus, for the homogeneous medium, the additional conditions concerning

the invariance of the energy with respect to rotation are reduced to (3.2.20,
24). These conditions can be clearly interpreted, if together with the micro-
rotation one introduces the macroscopic rotation Q1'[3 = O[p.U[3l' Then it can
be easily shown that the conditions obtained above are equivalent to the
requirement that the energy density ~o(x) depends on the difference Q1'[3 - Q~[3
only.
The equations of motion (3.2.16) in the (k, w) representation can be written
in the form of an algebraic system

[-w 2 pga[3 + k).kp.r).ap.{J(k)] u[3(k, w)

- ik).X+).aP[3 (k) r;# (k, w) = qa(k, w) ,
(3.2.27)
[_w 2]pa p'p. + r Pap '[3 (k)]r;p'[3(k, w)
+ ix pa p.[3 (k)kp.u[3(k, w) = ,uPa(k, w) .

Quasicontinuum. Restricting to the class of admissible functions of the

types described in Sect. 2.1, the operators and the equations of motion can
be written in the discrete form, too. Particular models are obtained either
with a special choice of kinematic variables (for example, the Cosserat model,
which is considered in Sect. 3.5), or in the presence of or by imposing additional
groups of internal symmetry. Discrete symmetry groups for crystals or the
rotation group for an isotropic medium can serve as examples.

Problem 3.2.3. Write the three-dimensional analog of [Ref. 1.12, Eq. (3.2.35)]
for the case, in which the elementary cell has a centre of inversion.
Problem 3.2.4. Formulate the three-dimensional analog of the conditions
which were considered in 1.3.7 concerning the existence of an equivalent
model of medium of simple structure.
54 3. Medium of Complex Structure

In the general case the matrix (/)sas'~(k) does not possess any symmetry
with respect to a, (3. However, if this symmetry is present, then it can be given
a simple interpretation. Let us call the interaction in a discrete model binary
(trinary), if the elastic energy can be represented in the form of a sum, each
term of which depends on the difference of displacements of two (three) par-
ticles l . The generalization to the case of a continuous medium is obvious.
In [3.1] it was shown that, in the harmonic approximation, the trinary inter-
action is the most general form of interaction in a crystal lattice. It can be
proved [3.2] that the symmetry of (/)a~(n - n', ~, ~') and hence also of (/)sas'~
with respect to a, (3, is a necessary and sufficient condition of the interaction
being binary (in above-described sense). From this, in particular, it follows
that in anyone-dimensional model the interaction is always binary, and we
saw this also directly. In the three-dimensional models of simple structure,
invariance with respect to inversion
(/)a~(x) = (/)a~( - x) = (/)~a(x) , (3.2.28)
insures that the conditions of the interaction being binary, are fulfilled.
Let us point out an interesting case, when the conditions of invariance
of the equations of motion with respect to a homogeneous deformation

(3.2.29)
are satisfied. To this end, as follows from (3.2.27), it is necessary and sufficient
to satisfy the following conditions:

(3.2.30)
Together with (3.2.20) they give [Ref. 1.12, Eq.(3.8.7)]

(3.2.31)
In Sect. 3.4, we shall see that this condition considerably simplifies the com-
putation of the macroscopic tensor of elastic constants.

3.3 Approximate Models and Comparison with Couple-Stress

Theories

As in the case of a medium of simple structure, various methods of constructing

appoximate models are possible. Let us restrict ourselves to the simplest model
of the long-wavelength approximation.

IThis definition is more general than the one given in Sect. 2.3, since the invariance of
the terms of the sum with respect to rotation is not required. In connection with this, the
definition has a formal nature.
 3.3 Approximate Models and Comparison with Couple-Stress Theories 55

Let us expand the functions r(k), X(k) and F(k) entering (3.2.22) in powers
of k and consider only a finite number of terms. With k = 0, we have the
zeroth approximation, which is considered in the next section. When proceeding
to successive approximations, it is expedient to distinguish two cases (Sect.
2.6). For a gyrotropic medium, the next approximation is the first one and the
operators (3.2.17) have the form
r Aal'{3 = r6al'{3 + r~al'{3Ton Xpal'{3 = xPt,,{3 + Xfa,,{3TOr ,
(3.3.1)
FPap'{3 = F1fP'1' + F1ap'{3Tor .
As was pointed out above, this corresponds to the rotation of the polariza-
tion plane of a wave, which is typical of a gyrotropic medium. If the medium is
non gyro tropic then the approximation following the zeroth, is the second
order approximation where
r Aal'{3 = r6al'{3 + r~al'{3IlTOIlOT' XPa l'{3 = X~aJ1{3 + XP{I'{3IlTOIlOT ,
(3.3.2)
FPap'{3 = F~aP'{3 + FPz"p'{3llrOIlOT .
By the indices 1, 2 in (3.3.1, 2) we denote the corresponding coefficients in the
expansions of the functions r(k), X(k) and F(k) in powers of ik, the coefficients
obviously being real tensors.
Replacing the integral operators (3.2.17) by the differential operators (3.3.1,
2)[i.e. transforming (3.2.17) into (3.3.1)] or by analogous operators of higher
order, corresponds to the transition to a model of a medium in the approxima-
tion of weak dispersion. In other words, this means taking into account non-
local effects. From considerations of a dimensional analysis it follows that the
coefficients of the higher-order derivatives contain indirectly a scale parameter
I, which, due to the approximate character of the model, should be considered
small in comparison with a characteristic wave length or the size of the body.
Hence we conclude that the "resolving power" of the model is limited by the
parameter I, i.e. solutions of the approximate equations, in the best case, can
pretend to give a qualitatively correct description of non local effects in regions
of the order of I, but, generally speaking, should not be extrapolated to smaller
regions.
The equations of a medium with weak dispersion contain, as a particular
case, the equations of the Cosserat continuum, of an oriented medium, and of
couple-stress, multi poles, etc. theories of elasticity. The above considerations
point out the possible regions of applicability of these theories.
Usually, such theories are derived on phenomenological grounds, roughly
according to the following scheme. It is postulated, that the Lagrangian de-
pends on the strain and its derivatives and on a set of additional kinematic
variables such as microrotations and microstrain and their derivatives. By
analogy with the classical theory of elasticity the new kinematic variables and
their derivatives are put into correspondence with moment stresses of various
orders, for which equations of equilibrium (conservation laws) and equations of
56 3. Medium of Complex Structure

state of the Hooke's law type, are postulated. This enables us to obtain a closed
system of equations and a set of boundary conditions. Further formulation of
boundary-value problems usually does not differ from the formulation of an-
alogous problems in the classical theory of elasticity. For example, many papers
are devoted to investigations, in the scope of the couple-stress theory of elas-
ticity, concerning stress concentration around holes. In this situation, one would
expect that significant differences from the classical theory of elasticity arise for
holes with diameters of the order of or less than I.
Let us consider some of the difficulties which arise in such an approach.
1) All the couple-stress theories approximately account for the nonlocal
effects, and the parameter 1 which is contained in them has to be considered
to be small. Therefore, when extrapolating the results to the regions of the order
of or much smaller than I, one should be very cautious.
2) The presence of a small parameter in terms with higher-order derivatives,
as is well-known, leads to effects of the boundary-layer type. In connection with
this, it is hardly reasonable to consider boundary-value problems for com-
plicated regions in the scope of couple-stress theories. Moreover, it is also
unknown how to interpret the boundary conditions physically and from where
to take the values of the moments on the boundary.
3) The theories contain new constants (sometimes in large numbers) but
usually it is not clear, from what sort of experiments these could be determined.
It is necessary to know at least the qualitative relationship between these con-
stants and the parameters of the micromodel.
4) In an axiomatic construction of the theories without taking into account
their approximate nature, the equations of motion are often written down with
different degrees of accuracy with respect to different variables. We shall con-
sider this aspect with the example of the Cosserat model in Sect 3.5.
5) The connection between the couple-stress theories and the classical
theory of elasticity is usually discussed in a simplified manner: in order to pro-
ceed to the theory of elasticity, it is necessary to set the constants contained in
the equation equal to zero. As we have observed in the example of the one-
dimensional model of a medium of complex structure, the actual connection
is more involved: for sufficiently long wavelengths and for sufficiently low fre-
quencies the internal degrees of freedom are excluded from the equations of
motion, but they do make some contribution to the effective macroparameters.
In the next section this question will be considered for the three-dimensional
medium.
6) We observed that, in the nonlocal theory of a medium of simple structure,
the stress tensor is not a physically uniquely determined quantity. This holds
even more strongly for the couple stress in the medium of complex structure.
For its formally unique determination, it is necessary to postulate conserva-
tion laws,2 which do not have any group basis. As is well-known in field theory

2This is also pointed out in [B3.IJ.

 3.4 Exclusion of Internal Degrees of Freedom in Acoustic Region 57

[3.3], the conservation laws are a result of the invariance of the Lagrangian
with respect to the corresponding groups and in our case they are the groups
of translations and rotations. However, only displacements and microrotations
are essentially with this fundamental group; this leads to a special divergence
form of the equations of motion, which ensures the equilibrium of forces and
moments. The remaining internal degrees of freedom (microdeformations) are
invariant with respect to the fundamental group, and additional laws for couple
stress have to be postulated rather arbitrarily.
Note also that couple stresses do not satisfy the natural principle of cor-
respondence, i.e. they do not become the usual stresses in the limiting case of
long waves, when a transition to the conventional theory of elasticity should
take place.
Taking into account these arguments, one should not be surprised that the
rare attempts of experimental determination of couple stress in an nonhomo-
geneous medium have not led to any results.
Let us sum up. The couple-stress theories of elasticity and, generally, models
of a medium with weak dispersion approximately take into account compara-
tively delicate effects of nonlocality and of internal degrees of freedom. Their
application is expedient, first of all, when they yield not merely small quantita-
tive corrections, but describe qualitatively new phenomena. Therefore the
physically justified field of applicability of these theories is narrower than that
of the classical theory of elasticity, though the latter is the limiting case.

3.4 Exclusion of Internal Degrees of Freedom in the Acoustic

Region

As in the one-dimensional case, we have the dispersion equation for free vibra-
tions
det { - (Ji /,as'f3 + (/)sas'f3(k)} = 0, (3.4.1)
which connects the frequency wand the wave vector k. However this equation
now yields three acoustic and 3N-3 optical modes w(k), where 3N is the number
of degrees of freedom of a cell.
As pointed out in [Ref. 1.12, Chap. 3], in the acoustic region offrequencies
the main kinematic variable is the displacement of the center of the masses of
a cell Ua = WOa. The distinctive role of this variable is connected with the fact
that in the acoustic region the internal degrees of freedom 1jpa can be excluded
from the equations of motion by transforming the latter into an equation,
which contains only one variable U a , and into equations which express explicitly
the remaining variables in terms of displacements. Recall that the possibility of
such an operation is connected with the special law of transformation of the
kinematic variables under translation and with the stability of the medium.
58 3. Medium of Complex Structure

Let us consider, for the beginning, the equations of the zeroth approximation
[Ref. 1.12, Eq. (3.6.1)]

(- w 2pga(> + k;.kp.r6«P.f') u(>(k, w) - ik;.xdap(>r;p(>(k, w) = qa(k, w),

(3.4.2)
(- w2 JPaPI(> + r1rplf') 7Jp.1ik, w) + ixtau$kp.ur><k, w) = f1. pa(k, w) .

Problem 3.4.1. Show that the optical frequencies at k = 0 are determined by the
equation
det( - w2JPaP'(> + FPryP'f') = 0, (3.4.3)
and, moreover, from the conditions of stability, the optical frequencies cannot
be zero.

We see that the matrix (- w2J + r 0) has an inverse in the region of acoustic
frequencies. This means that in the acoustic region the second equation of the
system (3.4.2) can be solved with respect to 7Jpl(>, and 7Jn can be eliminated from
the first equation. From here it also follows that there exists a finite neighbor-
hood of the origin of the k, w-space, in which the internal degrees of freedom
can be eliminated in such a manner.
In the direct matrix notations (3.2.22) coincide with [Ref. l.12, Eq. 3.3.24)]
for the one-dimensional medium. This enables us to describe the final results
without going into details.
Let us introduce the matrix

A(k, w) ~eJ [ - w 2 J + r(k)]-l = A+(k, w) . (3.4.4)

The neighborhood of the origin of the k, w-space, in which A(k, w) exists, will
be called an admissible acoustic region.
With the help of the matrix A(k, w) the system (3.2.22) can be transformed
into the form

(3.4.5)

(3.4.6)

Here, the following notations

l/J).ap.(> = r).ap.(> - (X HaPq Apqp',XPI,p.fJ) ().p.) ,
aj,tfJ = APap', XP'qp.fJ, (3.4.7)
q~ = qa - ik).f1.).a, f1.).a = - a+J::p'(3f1.P'(3,
have been introduced.
The form of (3.4.5) coincides with that of the equation of motion (2.5.25)
of the medium of simple structure, but now l/J).ap.(> depends not only on k but also
 3.4 Exclusion of Internal Degrees of Freedom in Acoustic Region 59

on lV, or, in other words, there exists a time dispersion besides with the spatial
dispersion, the former being unconnected with the dissipation of energy.

Problem 3.4.2. Show that, when k and lV are small, the time dispersion appears
only beginning with the second approximation.

The right-hand side of (3.4.5) contains the equivalent density of external

forces q¥., which is equal to the difference between the density of external forces
qa and the divergence of the density of micromoments pAa, as is the case in the
classical theory of elasticity in the presence of distributed moments.
Repeating reasonings analogous to those described in Sect. 2.5, we can
rewrite (3.4.5) in the form

(3.4.8)

where cAal'P(k, lV) possesses the symmetry of the tensor of elastic moduli (2.3.19)
and is connected with rJ>AaI'P(k, lV) by the operator S. This operator is defined
by the set of operations described in Sect. 2.5

cAal'P(k, lV) = SrJ>AaI'P(k, lV) . (3.4.9)

We can now introduce a symmetric stress tensor

(3.4.10)

and write the equations of motion (3.4.8) in the (x, t)-representation in the usual
form

pu - div (J = q* . (3.4.11)

As distinct from the case of a medium with simple structure, the operator
Hooke's law has the form
(JAa(x, t) = H cAal'P(x - x', t - tl)CI'P(X', t') dx' dt', (3.4.12)
i.e. the strain tensor cl'P and the stress tensor (JAa have nonlocal connections not
only with respect to the spatial variables but also with respect to time.
It is important to emphasize that in the zeroth long-wavelength approxi-
mation, we obtain the equations of the classical theory of elasticity, and the
stress tensor (JAa transforms into the usual stress tensor, which is connected with
strain by the tensor of the elastic constants c«,I'P. The latter, according to (2.5.4),
is given by the expression3 [Ref. 1.12, Eq. (3.6.12)]

3For the crystal lattice, this expression coincides with that obtained by Born and Huang
[3.5, 6].
60 3. Medium of Complex Structure

ctfl43 = S(/>6al'(3 = ~I'«(3 + (/>8/l).(3 - (/>tfl'(3,

(3.4.13)
(/)6al'(3 def (/)).al'(3 (0, 0) = rtfl'(3 - [xt).aJn" A pqp'r(O, 0) X-trl'f3] <'11') •

Problem 3.4.3. Show that for the tensor cgal'(3 to possess the required symmetry,
it is necessary and sufficient to have symmetry of (/>}.al'(3 with respect to the
indices A.p. and aj3, as well as with respect to the permutation of these pairs of
indices, and this symmetry is ensured by (3.2.20).

Problem 3.4.4. Verify that if the conditions of homogeneous deformation

(3.2.26) are satisfied, then

m).al'(3 -
'Vo -
r).al'(3 _
0
1 (rl!.(3}.a
"2 0
+ r).(3tJa)
0, (3.4.14)

i.e. for the determination of caal'(3, it is not necessary to invert the matrix roo

Problem 3.4.5. On the basis of (3.2.17), show that in the long-wave length ap-
proximation the elastic energy (/) can be written in the usual form (with p. = 0)

(3.4.15)

where C).a is the strain tensor.

Problem 3.4.6. Let us write Co ~ 0 in (3.4.15) if the corresponding form is posi-

tive. Show that the following inequality is valid:

(3.4.16)

Eq. (3.4.13) and the analogous formulae for the elastic constants of higher
approximations enable us to express these constants in terms of the parameters
of the micromodel. In the following section we shall illustrate this point with
the example of the Cosserat model.
When proceeding to the higher approximations in (3.4.8) it is expedient to
distinguish the cases of gyrotropic and nongyrotropic media. However, it is
necessary to note that the property of gyrotropy in the above sense is not
invariant with respect to the transformation of the equations carried out. A
medium can be gyrotropic with respect to the initial system of equations (3.2.22)
but then the first approximation in (3.4.8) may coincide with the zeroth one.

Problem 3.4.7. Show that, if the conditions of homogeneous deformation

(3.2.26) are fulfilled and the interaction is binary, then generally speaking, the
medium is gyrotropic, but the first approximation in (3.4.8) coincides with
the zeroth one.
 3.5 Cosserat Model 61

The time dispersion appears in the second approximation only (Problem

8.4.2) and, for a nongyrotropic medium, the operator of elastic moduli in this
approximation has the form

(3.4.17)

where the constant tensors C2 and c~ are determined as the corresponding

coefficients of the expansion of the function c(k, (1) with respect to k and (1).
For an isotropic medium, the structure of c(k, (1) in the general case of
strong dispersion coincides with (2.8.4), but A(k, (1) and p(k, (1) are even analytic
functions of Ikl and (1). The evenness with respect to (1) is obvious, and evenness
with respect to Ikl follows from the fact that isotropic tensors with odd parity
do not exist.
The equation of motion for an isotropic medium is written in the (x, t)-
representation in the second approximation, as [Ref. 1.12, Eq. (3.6.18)]

Jii - (Po + /2P26.)6.u - [Ao + po + /2(A2 + P2)6.] grad div u = q* , (3.4.18)

J = def p- T2[p~6. + (A~ + p~)grad div] . (3.4.19)

Here / and T are the characteristic length and time, Ao and po are the correspond-
ing coefficients of the expansion of the functions A(k, (1) and p(k, (1), and J can
be considered to be an inertial operator characteristic of the medium.
In the static case the equations for an isotropic medium coincide with the
equations for a medium of simple structure and, hence, for solving them, one
can use the static Green's tensor constructed in Sect. 2.8.
Thus, in the admissible acoustic legion, the equations of a medium with
complex structure are reduced to the equations for an equivalent medium with
simple structure with spatial and time dispersions. In the long-wavelength
approximation this corresponds to the transition to a macroscopic description
of the medium with microstructure, when the internal degrees of freedom are
excluded from explicit consideration, but contribute to the effective elastic
constants.
Note that, conversely, in the region of high frequencies it is possible to elimi-
nate the displacements of the mass centers from the equations, analogously to
what was done in the one-dimensional case [Ref. 1.12, Sect. 3.6].

3.5 Cosserat Model

Let us consider a medium for which the internal degrees of freedom are reduced
to microrotation. The latter can be characterized by an antisymmetric tensor
Q~~ or by the pseudo vector

(3.5.1)
62 3. Medium of Complex Structure

where in the cartesian coordinate system crf.'~ is the unit anti symmetric pseu-
dotensor. 4 Analogously, the force micromoments are given by the antisym-
metric tensor rn).a or by the pseudovector rn r • Eq. (3.2. 17), after obvious simpli-
fications, takes the form

pga~ii~ - (h7).af.'Waf.'u~ - (hx+).af3D~ = qa ,

(3.5.2)
Ia~D~ + Xaf.'Waf.'u~ + Fa~D~ = rna.

A more particular model is obtained if it is assumed that the operator

Xaf.'f3 is anti symmetric with respect to p.{3. In this case the second equation
contains only the micro rotation 0' and the macrorotation 0 = 1/2 rot u.
Finally, if we restrict the obtained equations to the second approximation
and an isotropic medium, then we arrive at the well-known model of a Cosserat
continuum

pii - (70 + f27Z!:::..) I:::..u + 70 + IZ ({32 + rz)l:::..] grad div

- [{30 u
- (Xo + f2X21:::..) rot 0' = q,
ID' + (Xo + 12X21:::..)D + (Fo + F21:::..) 0' = rn . (3.5.3)

More precisely, in order to obtain the usual equations of the Cosserat conti-
nuum [3.8], it is necessary to equate the constants {32' 72' Xz to zero in (3.5.3),
but to preserve the nonzer0 5 constant F z. The equations thus obtained
appear to be slightly inconsistent with respect to the order of approximation.
The first equation is of the zeroth order with respect to the scale parameter
I, while the second equation is of second order. This inconsistency is a con-
sequence of the "symmetrical" assumption, that the Lagrangian depends on
the displacements u, microrotations 0' and their first order derivatives.
In reality, the variables u and 0' enter the equations nonsymmetrically,
in view of the differences in the laws of their transformations with respect to
the translation group. Therefore, in order to have consistent accuracy, it is
necessary to assume that, in the Lagrangian, the displacement derivatives have
an order higher by one than the order of the derivatives of microrotation.
We shall not continue the general investigation of the Cosserat medium,
but shall instead proceed to a concrete model which is interesting for its own
sake. Let us consider a system of rods which forms a cubic lattice and assume
that, at the knots of the lattice, masses connected by the weightless rods are
located, and that the masses possess finite moments of inertia. Nearest-
neighbor interaction is assumed for simplicity.
Let us introduce the following characteristics of the rods being parallel
to the axis x a of the lattice coordinate system: length I, longitudinal stiffness

4In an arbitrary affine or lattice coordinate system this is the corresponding tensor density
[B8.251.
50ther possible variants are connected with different definitions of an isotropic medium.
 3.5 Cosserat Model 63

EFa' transverse stiffness EJa (which is assumed to be independent of the axes),

torsional rigidity GJ~. The masses have a cubical symmetry with respect to the
axes, which coincide with those of the lattice, and hence, the moment of inertia
is diagonal 13f3 = I~af3. Let us also point out that, although the lattice pos-
sesses a cubical symmetry in the geometrical sense, in view of different char-
acteristics of the rods, the system has lower symmetry.
The role of the kinematic variables is played by the displacement of the
centers of mass U a and the pseudovector of rotation of the masses ()~. The
role ofthe force variables is played by the force la and the moment p'a, which
act on the masses. The equations of motion have the form

+ 1: (/)ff~(n - n')uf3(n', t)
n'

+ ~ (/)fJ1(n - n'){)fj..n', t) = la(n, t),

n
(3.5.4)
Iogaf3{)fj..n, t) + 1: (/)~~(n - n')u~(n, t)
n'
+ 1: (/)~q(n - n') ()f;(n', t) = pa(n, t).
n'

The above equations enable us to give an interpretation of the force con-

stants (/)(n). Thus, if we fix all the masses except the mass at a point n'
and give the latter a unit displacement in the direction of the axis {3, then
(/)ff~(n - n') and ~~(n - n') are equal, respectively, to the force and the moment,
which act in the direction IX on the mass situated at the point n and compen-
sate the reactions of the elastic bonds. One can draw analogous conclusions
about (/)ff~(n) and (/)af3(n).
We now make use of the well-known formulae of the strength of materials
for bending. It is convenient to solve these expressions with respect to the
displacement u and the angle of rotation ()' of the end of the rod in one plane,
representing them in the form

(I)
P
= 12EJ ( 1 -
/3
1 )
-1I - _/2
1
i
()' ,
(u) (3.5.5)
2 6

where 1 and p. are the corresponding force and moment at the clamped end.
Taking into account analogous relations for tension and twisting, we
obtain explicit expressions for the force constants in terms of the parame-
ters of the rods. The nonzero components of the force matrices have the
form
64 3. Medium of Complex Structure

o 1

I~
1 0 0 0 001
-1 0 0 o -1 0 o 0-1

(3.5.6)
-B
+ I +B I
±B1 1 1
':fB z - ZI ±B3

Here, the following notations have been adopted:

(3.5.7)

Problem 3.5.1. Verify that the force constants of self-action, found from the
conditions of invariance of (3.5.4) with respect to translation and rotation,
see (3.2.10) and (3.2.11), have the form (do not use the summation convention!)

(/)f,ff (0) = 21-1(Aa + B~ + B r), (/)r« (0) = 0,

(/)r~ (0) = 21( Ca + 1 +1
BfJ Br),

(a '# (3 '# r)· (3.5.8)

In the usual manner, passing to the k representation, we obtain the nonzero

components of the tensors r(k), X(k) and r(k) in the form (no summation)

r':'~a = AasZ(ka), rC;Pa~ = BasZ(k a), Xap.fJ = cafJ P.Bp.s(2kp.),

r':'a = Cak~sZ(ka) + BfJ[1 - ~ k~Z(kfJ)J + Br[1 - ~ k~sZ(kr)J '

(a '# (3 '# r), s(k a) def ( . sin ;a . (3.5.9)

Problem 3.5.2. Write down the conditions of invariance with respect to rotation
(3.2.20, 24) taking into account (3.5.1), and verify that these conditions are
fulfilled for the system under consideration.

Problem 3.5.3. Write down the equations of motion (3.5.2) in the zeroth and
second approximations expressing the tensor coefficients in terms of the pa-
rameters of the system.

Problem 3.5.4. Verify that the conditions of homogeneous deformation (3.2.26)

are satisfied only for rods having the same stiffness.
 3.5 Cosserat Model 65

In the admissible acoustic region, it is possible to eliminate the micro-

rotation ()~ from the equations of motion (3.5.2) by transforming them to

°
the form (3.4.5, 6). The tensor if) Aal'{3(k, w) entering this equation possesses
for k = and w = 0, the required symmetry with respect to the indices and
in particular, satisfies the conditions of invariance with respect to the rotation
(2.5.8). Recall that this condition was not fulfilled in the consideration of the
rod system (Sect. 2.7), in the framework of a medium of simple structure
(without taking into account microrotations).
Let us now present the final expressions for the operator of the elastic
moduli c Aal'{3(k, w) confining ourselves (for simplicity) to the case of identical
rods and setting Aa = A, Ba = B, C a = C. For the essential components of
cAal'{3(k, w) we have (a #- f3 #- r)

C':.~{3 = ~ B { - s2(ka) - s2(k{3) + [s(2ka) + s(2k{3)] Pr(k, w)} ,

c':.~{3 = B {- I + s2(k a) + s2(k{3) (3.5.10)

+ ~ [I - s2(2k a) - s2(2k{3)]pr(k, w)} ,

af3. _ 1 [ 1 - s2(k{3) 1 1 - s2(2k{3) ]
c .. {3{3 - TB ka k{3 + Tka k{3 Pr(k, w) ,

where Pr(k, w) is determined by the expression

1 _ 1 C k2 2(k) .lk2 2(k ) _ .lk2 2(k ) _ Iw 2 (3.5.11)

Pr(k, w) - + 2B 'is r - 12 as a 12)3S {3 2B '

where I = VOl 10 is the density of moments of inertia.

From here, by expanding in k and w, the elastic constants of the zeroth
and successive approximations are found. In particular, for the tensor of
the elastic constants of the zeroth approximation, we have [see (2.7.12)]

(3.5.12)

It is obvious that, under tension along a coordinate axis, there is no side

thrust, as one should expect (Sect. 2.7). Note that this expression does not
contain the constant C, i.e. the stiffness of the rod with respect to twisting
does not contribute to the elastic constants in zeroth approximation. This
can be easily explained, if we take into account that twisting is connected
with an inhomogeneous deformation.
Let us give the expressions for the essential components of the elastic-
constants tensor in the case of nonidentical rods:
66 3. Medium of Complex Structure

BB
aa - A a' CfIlP -
B a fJ co·aa·PfJ -- 0 (a..J.
.,.. t'f.l) (3.5.13)
+ B'
CO··aa - U··afJ - .
a fJ

For the rod system, let us consider the dispersion equation (3.4.1), which
connects the frequency wand the wavevector k. For some characteristic direc-
tions of the cubic lattice, in particular along the diagonals (1, 1, 1) and (1, 1,0)
and along the axis (1, 0, 0), it is possible to obtain the solution w(k) of the
dispersion equation explicitly. In all these cases degeneracy takes place: there
exists one acoustical and one optical mode, these being the analogs of longi-
tudinal waves, as well as two coinciding acoustical and two coinciding optical
modes, these being the analogs of transverse waves.
Let us write down the dispersive modes for the direction (1, 0, 0,), i.e.
for the wave vector (k h 0, 0). The acoustic modes are given by (p = VOl m)

w2 = wi(k 1) = :~ sin 2 ~ ,
(3.5.14)
w2 = w!,lk1) = ~ [I(k 1) - ,.; I2(k 1) - g(k 1)] ,

and the optical modes by

(3.5.15)

Here

(3.5.16)

Problem 3.5.5. Write down the solutions of the dispersion equation for the
directions along the diagonals.

3.6 Notes

Extensive literature is devoted to different types of couple-stress theories,

see e.g. [B2.1-38]. All these theories can be obtained as long-wavelength
approximations for nonlocal elasticity developed in [B3.1-34].
 3.6 Notes 67

The theory of media of complex structure considered in this chapter is

based on [B3.23, 34) and is closely connected with the theory of crystals with
a basis [B5.2, 10). The method of calculating the elastic constants for crystals
with a basis is due to Huang [B5.4), see also [B5.2, 5, 9-] 2]. The nonlocal
Cosserat model was considered in [B3.1O).
4. Local Defects

In this chapter, we consider local inhomogeneities as defects in a homogeneous

medium. We use systematically the Green's function technique which seems to
be most adequate to the problem.
We start from a general scheme for arbitrary inhomogeneities. Then we con-
sider impurity atoms in a lattice and show that the problem is reduced to solving
a finite system of linear algebraic equations. A smilar method is used for point
defects in a quasicontinuum.
The second part of the chapter is devoted to inhomogeneities in elasticity.
Our approach is essentially different from that adopted in classical elasticity.
We formulate boundary-value problems for inhomogeneities in terms of a
special class of global pseudo-differential operators (projection operators and
Green's operators for stress and strain). It is shown that the developed techni-
que can be efficiently applied to ellipsoidal inhomogeneities and cracks. Finally
a similar technique is used in the theory of internal stress, dislocations and
stochastic medIa.

4.1 General Scheme

Let 1/)0 be a linear operator for which the Green's function (inverse operator)
Go is known and satisfies certain boundary conditions, i.e., Gol/)o = I, where 1
is the identity operator. If 1/)1 is a perturbation of the operator 1/)0 such that a
Green's function G of the operator I/) = 1/)0 + 1/)1 exists, then one can show that
the Green's function G has the representation

G = Go - GoPG o , (4.1.1)

where the operator P is defined by

(4.1.2)

In fact, substituting Pinto (4.1.1) and multiplying by I/) from the left, we have
I/)G = (1/)0 + 1/)1) [Go - G 01/)1(1/)1 + 1/)1G OI/)1)-1 1/)1G O]
= 1 + 1/)1GO - 1/)1(1/)1 + 1/)1GOI/)1)-11/)1GO - 1/)1G01/)1(1/)1 + 1/)1GOI/)1)-1 1/)1GO
= 1 + [I - (1/)1 + 1/)1Go1/)1) (1/)1 + 1/)1GOI/)1)-1] 1/)1GO = I.
 4.1 General Scheme 69

Thus, the construction of the Green's function G is reduced to finding P from

(4.1.2) or, equivalently from

(4.1.3)

Later, as a rule, the operator (/)0 will correspond to an infinite homogeneous

medium and the operator (/)1 will describe defects of the type of inhomogeneities
which are localized in small regions. It is easy to see that the kernel of the
operator P, as distinct from that of the Green's function G, is also localized in
these regions and due to this its construction both analytically and by numerical
methods appears to be more convenient.
From the Hermiticity of the operators Go and G, the Hermiticity of P follows
immediately, this being evident from (4.1.1 and 4.1.2).
In those cases for which the operator (/)11 exists, (4.1.2) can also be written
in the form

(4.1.4)

where Go is the restriction of the operator Go to the region of defects.

Let a distribution of external forces q be given and let Uo = Goq be the cor-
responding external field in the homogeneous medium without defects. Then,
for the field u = Gq in the medium with defects, taking into account (4.1.1)
we find

u = Uo - GoPuo, (4.1.5)

where the second term is the perturbation caused by the defects. It is obvious
that this equation is valid also in the case when the external field Uo is given
directly.
For the elastic energy (/), according to (2.3.22), we have

2(/) = <qIGlq> = <uol(/)oluo> - <uoIPluo>· (4.1.6)

The first term corresponds to the energy (/)0 of the external field in the me-
dium without defects and the second term is the energy (/)int of interaction of
the defects with the external field and among themselves. This enables us to in-
terpret P as an operator for interaction energy.

Problem 4.1.1. Show that from the invariance of the energy with respect to tran-
slation, it follows that the operator P can be represented in the form

P = VPV. (4.1.7)

Substituting Pinto (4.1.6) we obtain the expression for the energy of inter-
action in terms of the deformation eo of the external field and the operator 15:
70 4. Local Defects

(4.1.8)

4.2 Impurity Atom in a Lattice

Let (/Jfft9(n - n') be a matrix of the force constants for a homogeneous lattice,
and let (fla!n - n') be the corresponding static Green's tensor. Let the impurity
atom be located at the point n = 0, this atom causing a local perturbation of the
force constants. Then the perturbed matrix has the form
(/)at9(n, n') = (/Jfft9(n - n') + (/Yft9(n, n') , (4.2.1)
where (/Yft9(n, n') differs from zero only for atoms of the first coordinate spheres.
According to the above, the Green's tensor of the lattice with a defect can be
represented in the form

Ga{l(n, n') = G~{l(n - n') - 1: G~in

m,m'
- m)Plp(m, m')G~!m' - n'), (4.2.2)

where the summation is carried out over the region of perturbation and the
matrix plp(mm') is defined by
plp(m, m') = {[rpfP(m, m')]-l + (flJ.a(m - m')}-l . (4.2.3)

Since the points m, m' here belong to the finite region of perturbed bonds, the
problem is reduced to the solution of a finite system of linear algebraic equations.
When inverting the matrices, it is necessary to take into account that, due
to the conditions of invariance with respect to translation and rotation, the in-
verse matrices exist only in some special vector subspaces, which can be distin-
guished by constructing the corresponding projection operators. Also, in the
general case, expressions ofthe type (4.1.2) should be used instead of (4.2.3).
Thus, an effective analytical solution of the problem can be constructed only
for the simplest models, which we shall not dwell upon. In more complicated
situations, numerical methods are needed.
In analogous manner (by numerical methods) the problem of the dynamic
Green's tensor Gain, n', ev) can be solved for the case of a single defect as well
as for the case of finite set of defects.
In order to obtain tractable analytical expressions, it is expedient to accept
another model for the defect, which will be considered in the next section.

4.3 Point Defects in a Quasicontinuum

We assume the existence of an elementary unit of length a in the medium;

this is equivalent, as was pointed out above, to a corresponding truncation of
the Fourier spectra of the admissible functions, i.e. to a transition to the elastic
 4.3 Point Defects in a Quasicontinuum 71

quasicontinuum. This enables us to use the a-function model for the description
of local inhomogeneities, introducing a-functions and their derivatives directly
into the coefficients of the equations. Such a procedure would be incorrect for
an elastic continuum.
Suppose that there is a single defect at a point x = Xo. Let us first consider
the static case. The simplest model of a defect of the type of local inhomogeneity
of the operator for the elastic moduli in the neighborhood of the point Xo can
be described by an expression of the form

(4.3.1)

Here, the constant tensor cia i'{3 has the dimension and symmetry of the elastic-
constants tensor and characterizes the variation of the elastic properties in the
neighborhood of the point Xo. A more complicated model of a defect can be
obtained if we add additional terms with derivatives of a-functions.
Straightforward computations according to the scheme of Sect. 4.1 give for
the Green's tensor Ga{3(x, x') of the medium with the defect, the expression

(4.3.2)

The constant tensor PVi').~ appearing here has the symmetry of the tensor of
elastic constants and is determined by

P Vi').~ -- _ (....0
l> Vi').~
+ v-lb
0 Vi').~
)-1 , (4.3.3)

where b Vi').~ is the tensor which is inverse to crll).r, and

(4.3.4)

The existence of g2i').~ follows from the analyticity of G~ix) for the quasi-
continuum. From here it also follows that Ga{3(x, x') is an analytic function of
x and x'. However Gaix, x') depends on the parameter a nonanalytically.

Problem 4.3.1. Show that gO '" VOl, P '" Vo and, as a ~ 0, the second term in
(4.3.2) is of the order of Vo for the points (x "# xo, x' "# xo), of the order of a
for points (x = xo, x' "# xo) or (x "# xo, x' = xo), and tends to infinity as {l1
for the point (x = x' = xo) .

The construction of the dynamic Green's tensor for a single defect can be
carried out in an analogous way. Here it is reasonable to consider two cases.
If the density p(x) = Po = const and if the defect is caused only by a change
of the elastic properties of the medium, then the expression for the dynamic
Green's tensor Gaix, x', w) coincides with (4.3.2), after replacing G~ix) by
72 4. Local Defects

the dynamic Green's tensor for the homogeneous medium G~J3(x, w). Similarly,
p = P(w) is determined by (4.3.3) with gO = gO(w).
In the other case, when c(x, x') = co(x - x'), and the density has the form:

p(x) = Po + VOPIO(X - xo) , (4.3.5)

we find for the Green's tensor

GaJ3(x, x', w) = G~ix - x', w)

- G~.lx - Xo, w)P)."'(w)G~J3(xo - x', w), (4.3.6)

P).I'(w) = -VOPIW2[O).I'- VOPlw2G2/0, W)]-l. (4.3.7)

In the general case of a defect of mass and elastic moduli, the expression
for the Green's tensor is constructed in an analogous way, but has a more
complicated structure.
Let us now consider a perturbation of the external field u~(x), caused by the
defect, and let us confine ourselves to the static case. Taking into account (4.3.2),
we find

uix) = u~(x) + u~(x) , (4.3.8)

where u~(x) is the perturbation which coincides, as can be easily proved, with
the displacement in the homogeneous medium, caused by a force dipole at
the point Xo with density

(4.3.9)

and with the moment

(4.3.10)

where c~«xo) is the deformation at the point Xo without the defect.

Thus, the perturbation caused by a defect can be obtained, if we replace
the defect by an equivalent force dipole having the moment (4.3.10). Roughly
speaking, a local inhomogeneity generates a force dipole which is proportional
to the field. By analogy with the electric field in a dielectric, one can also speak
about the polarization effect caused by a local inhomogeneity.
Let us represent the elastic energy f/J as a sum of the energy of the field f/J o in
the absence of the defect and the energy of interaction f/Jint of the defect with
the external field. For <Pint' analogously to (4.1.8), we find

(4.3.11)

This energy depends on the position of the defect. In turn, it enables us to

 4.3 Point Defects in a Quasicontinuum 73

introduce naturally the concept of a force f, with which the external field acts
on the defect. By definition

(4.3.12)

As an example, let us consider a spherically symmetric defect located at the

origin in an isotropic quasicontinuum. The tensor Cl in this case has the form
(see (2.8.4»

Cl = 2P.lE1 + AIE2,
Ela!'fJ def ~ (aafJaJ.!, + aaJ.afJ!'), E~afJ!' def aJ.aa!'fJ . (4.3.13)

Using the general expression for the Green's tensor of the Debye quasicon-
tinuum with spatial dispersion (2.9.11) and taking into account (4.3.3, 4 and
13) we obtain

gO - 1 [( {31 (3t ) E + ( 2{3/ + 3~ot) E1] , (4.3.14)

- 15vo Ao + 2p.o - Po 2 Ao + 2p.o r

(4.3.15)

where

r(3/ I 1-1
PI = V't 3(Ao + 2p.o) + 3Al + 2P.l J '
(4.3.16)
P _ v I.l.. ( 2{3/ + 3{3t) + I 1-1
2 - 't15 Ao + 2p.o Po
2p.1 J·

The dimensionless constants {3j depend on the dispersion law

k4 .
(3j
3v?
= .1t'~ Jwr(k)
K
dk (J = I, t) . (4.3.17)

For the Debye model and for the model (2.9.8) we have, respectively,

(4.3.18)

Let us assume in addition that with the defect there is associated a force
dipole with the defect there is associated a force dipole with the density 1

1 Recall that for an isotropic medium we use the notations rand r instead of x and 1x I.
74 4. Local Defects

q).(r) = - Qo).o(r) . (4.3.19)

The internal stresses due to a vacancy, an interstitial atom and so on are

often modeled in such a manner. In the given case, the model corresponds to
a center of dilatation.
It was pointed out above that the displacement field u(r) can be represented
in the form of a sum of the unperturbed and perturbed fields uO(r) and u1(r).
In the model under consideration, uO(r) is a field of the force dipole (4.3.19) in
the homogeneous medium. The deformation of the unperturbed field at the
defect point, as can easily be verified, is equal to

o (0) _ Q5:).p. 0 _ Qf31 5: (4.3.20)

ca[3 - v ga(3).p. - 3VO(AO + 2/10) Vall .

For the moment of the effective dipole, according to (4.3.10) and taking
into account (4.3.15, 16 and 20), we have

(4.3.21)

The expression for the total deformation c = cO + c1 now has the form

(4.3.22)

or, if (2.9.12) is taken into account,

(4.3.23)

At the point of the defect

(4.3.24)

and at large distances from the defect

(4.3.25)

It is also easy to calculate the corresponding stresses (]'a[3(O) and (]'all(r).

It is interesting to compare the model under consideration with the classical
problem of a spherical inhomogeneity in an isotropic elastic medium with the
Lame constants AO' /10' Let the elastic constants of a small sphere of the volume
 4.4 System of Point Defects 75

Va be equal to Ao + Al and f.J.a + f.J.l and let, on its boundary, radial forces with
the surface density Q/vo be applied. Then it can be shown that for the Debye
model (i.e., when spatial dispersion is absent) the values of e(O) and 0'(0) coincide
with the strain and stress within the sphere and the asymptotic values of e(r)
and O'(r) coincide with the strain and stress outside the sphere. The total elastic
energies for these cases also coincide exactly.2
For the interaction energy <Pint of the defect with the external filed egir),
we find, according to (4.3.11, 15)

<Pint = J
~l [ Tr{eO(O) } + ~{Dev{eO(O) } J' (4.3.26)

where Tr{eO} and Dev{eO} are the spherical and deviatoric parts of eO.
From this it follows that the force which acts upon the defect is equal to

fA = - ~l Tr {eO(O)} oJ.Tr{eO(O)} - P2 Dev{eO(O)} OJ. Dev{eO(O)} . (4.3.27)

4.4 System of Point Defects

Let us now consider the case of a system of point defects; for definiteness, we
take it to be a system of defects of the elastic moduli. Let us also restrict ourselves
to the static case (for the simplicity of notation). We have

cJ.apf3(x, x') = cS apf3(x - x') + Vo L: cfa l'i3o(x - Xj)o(x' - x,.) . (4.4.1)

j

Omitting simple but lengthy computations, we present the final result for
the static Green's tensor

Gaix, x') = G~ix - x')

- r; ovGg/x -
IJ
x,.) PYJJ.~oJ.G~iXj - x'). (4.4.2)

The matrix P ij with tensor components entering the above expression, is the
inverse of the matrix

(4.4.3)

where

(4.4.4)

and the btpJ.a are tensors inverse to ctl'J.a"

2Such a coincidence is, of course, caused by the high symmetry of the model. In the general
case, the correspondence between point defects and local inhomogeneities is more complicated.
We return to this question later.
76 4. Local Defects

The dynamic Green's tensor for the system of point mass defects has an
analogous structure.
In many cases, the dominant term of the asymptotics of the perturbed
elastic and or interaction energy is of principal interest. It may be shown that,
in such a situation, the problem is reduced to the binary interaction of defects.
In this connection, let us consider in more detail the case of two defects, which
is by itself of interest.
When i andj = 1,2, the computation of the components P ij is conveniently
carried out according to (no summation)

Pi .. = (R jj - RijRj/Rj,)-l, (i :f,j),
(4.4.5)
Pij = (Rji - RjjRj} Rjj)-l

whose validity is verified by direct computation.

These formulae also enable us to estimate easily the asymptotics of the
matrix Pij for a large distance p = IXI - x21 between the defects. In fact, GO(p)
'" p-l, and hence gii(p) '" p-3. Therefore, taking into account (4.4.3, 5),
we obtain

p.ip) = P7 + O(p-6) ,
(4.4.6)
p .. /p) '" p-3 + O(p-9) ,

where the tensor PP corresponds to an isolated defect at the point Xj and is de-
fined by (4.3.3).
For the isotropic Debye quasicontinuum, the matrix P../p) can be computed
explicitly. Let us also present the generalization of (4.3.8) to the case of a
system of defects in the external field u~(x):

uaCX) = u~(x) + .L: u~(x) . (4.4.7)

i

Here, u~(x) is the displacement due to a defect at the point Xj' However, the
moments of the equivalent force dipoles

Q"I'
• = .L.,.l'J ",0 (x.)
" P¥f!.).< ,,).< J (4.4.8)
J

now depend on the values of the field at all points x j as well as on the elastic
characteristics of all defects. Therefore, in the general case, the perturbation is
not the superposition of perturbations caused by individual defects, i.e. the
matrix Pij depends on the perturbations of the elastic constants of individual
defects in a complicated manner. In the limiting case of large distances between
the defects it follows from formulae of the type of (4.4.6) that in the zeroth
approximation, the perturbations of individual defects can be linearly super-
posed; in the next approximation we have the superposition of binary inter-
actions.
 4.5 Local Inhomogeneity in an Elastic Medium 77

The calculation of the energy of interaction of defects with the external

field and between themselves can be carried out according to (4.1.6), which
in the present case takes the form

(4.4.9)

The force, which acts on the point defect at a point Xk is given by the
expression

rk def 0 (/J
J,l ~ - oxt int

= Q'jfO,lC8/Xk) - i i1 c8/ x j) [o~t Pir~}~~(Xj) . (4.4.10)

For defects situated far enough from each other, the second term becomes
vanishingly small and the first term coincides with the force which acts on the
isolated defect at the point Xk' Conversely in a homogeneous external field
cO = const, only the second term contributes to the force.

4.5 Local Inhomogeneity in an Elastic Medium

Strictly speaking, the problem of inhomogeneity in the elastic medium is related

not to the theory of a medium with microstructure but to the conventional
theory of elasticity. However, we shall be interested in the asymptotic behavier
of perturbed fields and of the interaction energy. We shall see that in such a
formulation of the problem, a far-reaching analogy with the model of point de-
fects considered above exists.
Let the tensor of the elastic moduli of the medium be of the form

(4.5.1)

where cff{l,lf.l is the tensor of the elastic constants of the homogeneous medium,
c7{l,lu is the perturbation caused by the defect localized in a (small) region V;.
The case Cj -+ 00, corresponds to a rigid inclusion and the case C; -+ - Co
corresponds to a cavity. Let us consider, for simplicity, an infinite elastic
medium.
Using the general scheme described in Sect. 4.1, one can show without diffi-
culty that the Green's tensor can be represented in the form

Ga{l(X, x') = G~ix - x')

- II Ga(ix, y)o).)p,lv"~(y, yl)o("G~)iy', x)dy dy' (4.5.2)

or, in operator form

78 4. Local Defects

G = GO - GOVPVGO, (4.5.3)

where the operator P for the interaction energy is a sum of operators

P = 1:; pjj. (4.5.4)

jj

The matrix pjj and its inverse matrix Rij have the structure analogous to (4.4.3):

pjj = Cj(cjVGoVCj - CAj)-lCj = (VGoV - ci1o.)-1 = Ri/ '

Rjiy, y') = - VV'GO(y - y') - ci1(y)o(y - y')Oij, (4.5.5)
y E Vj, y' E V; .

It can easily be seen that the operator P is self-adjoint; its kernel satisfies the
symmetry conditions

(4.5.6)

and is localized in the region UjjVj ® V j •

For the displacement u(x) in the external field and for the energy of the in-
teraction (jjint we obtain

u..(x) = u2(x) - JG2,,(x - y)chPJ.>"r(y, y')e~(y') dy dy' , (4.5.7)

(4.5.8)

Taking into account that P(y, y') implicitly depends on the coordinates of
the centers of mass of the defects, we have for the force fj, which acts on the
kth defect, due to the external field and other defects,

(4.5.9)

In an analogous manner one can find the moment which acts on a defect.
In the general case, the kernel P(y, y') may be found by numerical methods,
and use of the latter is facilitated by the fact that, unlike G(x, x'), P(y, y') is
localized in a bounded region.
The problem is considerably simplified and, in a number of cases, can be
solved analytically if the defects are located at distances, which are large in
comparison with their size and if only the asymptotic form of the perturbed
field is of interest. Eventually, this is equivalent to replacing local inhomo-
geneities by equivalent point defects.
First, let us consider approximations with respect to the distance r jj =
IXj - xjl between defects. In the zeroth approximation the defects do not in-
 4.5 Local Inhomogeneity in an Elastic Medium 79

teract and Pij = Piou j , where pi is the operator of the i-th defect. One can show
that in the first approximation the problem is reduced to binary interaction.
Therefore let us restrict ourselves to the case of two defects. For two defects,
according to (4.5.5), we have

(4.5.10)

The calculation of the components Pii in terms of Rij is conveniently carried

out according to (4.4.5). Keeping only dominant terms with respect to r12 =
IXI - x21, we find the explicit expressions for the components Pii in terms of the
operators pi of individual defects

(4.5.11)

Let us now proceed to approximations of another type, which enable us to

find the asymptotics of the perturbed field. With this in mind, let us approximate
kernel P(y, y') by first terms of its multi pole expansion in the neighborhood
of each defect

P(y, y') = L: (-l)m+"P{,{ir.)o(ml(y

ijmn
- x •.)o(nl(y' - Xj) (4.5.12)

(Xi E Vi' Xj E Vj) .

Substituting this expansion in the equation for P of the type of (4.1.3), it is

possible to obtain finally a linear system of equations for the moments P:i,,..
It is important that the dominant terms with respect to r ij of the first moment
of the multipoles can be determined explicitly, if the first moment P:" for each
defect, in the absence of all the other defects, is known.
Note that, in the case of a homogeneous external field E~,8 = const, we have
for the interaction energy ~int (for any r ij) the exact formula, see (4.4.9),

'" - -2
YJ'int -
1 EOa,8 pa{3A"(r
()() ij)EOA,.., P 00 -_ 4~ pij
., 00· (4.5.13)

From here it follows that only the asymptotics of the matrix ptIJ(r.) contri-
butes to the dominant term of the asymptotics of ~int in a homogeneous ex-
ternal field. In the case of an arbitrary external field, it is necessary to take into
account also the contribution to the asymptotics of ~int from the asymptotics
of the diagonal components of the matrices pM(rij) and pf6(r.). This contribu-
tion is equal to zero if the defects possess central symmetry or if the external
field changes slowly over a distance of the order of the size of a defect.
Let us sum up. For the description of the interaction of defects in an elastic
medium, it is necessary to know, in some approximation, the Green's function
for the medium with defects, or, equivalently, one has to be able to solve the
80 4. Local Defects

corresponding problems of the classical theory of elasticity, exactly or approxi-

mately. The following sections of the present chapter will be devoted to these
problems, which are of interest in themselves. The results obtained will be
used in an essential way, particularly in Chap. 7. This chapter is devoted to
elastic media with random distribution of defects.
Note that the approach to problems ofthe classical theory of elasticity, which
is developed further, differs from that presented in standard textbooks.

4.6 Homogeneous Elastic Medium

The equations of classical elasticity were written above in the form of (2.6.4).
In this section we confine ourselves to statics. Then the equations become

(4.6.1)

We assume additionally that the medium is unbounded and homogeneous;

hence, the tensor clapfJ is constant. The density of external forces q(x) is assumed
to be a generalized function, decreasing sufficiently rapidly at infinity; for
example, it can have a bounded support.
Under these conditions, the solution of (4.6. 1) which tends to zero at infinity,
is unique and is representable in the form

u = Gq, (4.6.2)

where G = G+ is the Green's operator for displacements. In a more detailed

expression

uix) = JGafJ(x - x')qfJ(x') dx' , (4.6.3)

where GafJ(x) is the tensor Green's function for displacements. Obviously, in

the k representation,

G(k) = [(/)(k)]-l , (4.6.4)

where

(4.6.5)

For the isotropic medium, the tensor of elastic moduli becomes, see (2.8.4),

c= 2p.E1 + )'E2 , (4.6.6)

where E1 and E2 are constant isotropic tensors defined in (A. 1), and )., p. are
 4.6 Homogeneous Elastic Medium 81

the Lame constants. The expression for G(k) coincides with (2.8.6), where A
and p. are taken to be constants.
For an arbitrary anisotropic medium

Garf.-k) = ga~n) , (4.6.7)

where na = kallkl and gap(n) is a function defined on the unit sphere in k-space,
which is a determined explicitly by (4.6.4, 5).
The explicit expression for Gap(x) is known only for an isotropic medium
and media with transversly isotropic and hexagonal symmetries [4.1]. However,
we shall see that in a number of important cases it is sufficient to know general
properties of Gap(x) and its expression in the k-representation.
In particular, note that it follows from (4.6.4, 5) that

(4.6.8)

Let us consider briefly the behavior of the solution (4.6.3) for a few important
types of the density of external forces qa(x). Let qa(x) = ija(x) V(x), where V(x)
is the characteristic function of a bounded region V with a sufficiently smooth
boundary (), and ija(x) is a continuous function. Then uix) is a continuous
function, which is twice differentiable inside V, analytic outside V and tends to
zero as ,-1, when r = Ixl --+ 00. Its first derivatives and consequently, the strain
and stress tensors, have a finite discontinuity on ().
If qa(x) = qffo(x), then the solution is proportional to Garf.-x) and has, as
follows from (4.6.6), a singularity of the order of ,-1 as x --+ O. From here, it
is easy to find also the behavior of uix) for external forces concentrated on
surfaces.
Proofs of all these statements can be obtained in a manner analogous to
that used in the potential theory, we shall not therefore dwell on this.
We shall see later, that in a number of applications, in particular for a
medium with defects, it is convenient to have an equivalent form of the equations
of elasticity and to have their solutions not in terms of the displacement ua ,
but in terms of the strain eap or stress (laP. For this purpose, let us rewrite (4.6.1)
in the form of the usual system

(4.6.9)

or, in the operator form,

div (l = -q, (l = Ce, e = def u, (4.6.10)

where def is the operator for the symmetrized gradient.

Our first task is to write down these equations and their solutions in a form
which contains the strain tensor only.
82 4. Local Defects

Let us represent the density of external forces in the form

(4.6.11)
where pal' = pl'a is the density of force moments, the latter being defined non-
uniquely. After some additional conditions are imposed, the correspondence
qa _ pal' becomes one-to-one, and it is not difficult to find pal' explicitly. It is to
be emphasized that, in problems concerning inhomogeneities in an external
field, pal', and not qa is the naturally given quantity (see below).
Substituting into (4.6.3) the expression (4.6.11) for qa and applying the opera-
tor defto both sides of the equality, we obtain
(4.6.12)
where
(4.6.13)
In operator notations,

c = Kp, K = - def G def . (4.6.14)

From the preceding, it follows that K is a correctly defined self-adjoint
operator, K = K+. Its kernel K(x) is a generalized function, whose Fourier
transform K(k) is, as follows from (4.6.7, 13), a homogeneous function of the
zeroth degree, i.e. K(k) is, in fact, a function of the unit vector na = ka/lkl
(4.6.15)

Problem 4.6.1. Show that, for an isotropic medium,

K(n) = -.L [E5(n) - ,A+2c E6(n)] , (4.6.16)

fJ- A+ fJ-

where E5(n) and E6(n) are defined in (A. 1).

From the theory of generalized functions, it follows [4.2] that the inverse
Fourier transform K(x) of the generalized homogeneous function of the zeroth
degree K(k), is defined with the help of the corresponding regularization. The
latter is reduced to a decomposition of K(x) into a regular part and a singular
one, which is proportional to o(x). Properties of the generalized function K(x)
are considered in detail in Appendix A.3.
From (4.6.12), it follows that the desired strain field is obtained as a con-
volution of the function K(x) with the density of force moments, p(x), i.e. K
is the Green's operator for strain. Let us consider a few features ofthis operator.
From (4.6.14), it is obvious that the operator K is annihilated by all opera-
tors which annul the operator def. In particular, this is true for the operator Rot,
 4.6 Homogeneous Elastic Medium 83

which is considered in detail in the next chapter. Thus, we have

RotK= O. (4.6.17)

Further, taking into account (4.6.4, 14) and the symmetry of the tensor of
elastic constants, we have

KCK = def GVCVG def

= - def G(I)G def = - def G def (4.6.18)

or

KCK= K. (4.6.19)

In particular, it follows from here that

U= KC, li = CK (4.6.20)

are projection operators (projections), i.e.

(4.6.21)

while

li+ = U. (4.6.22)

Let us introduce a dimensionless density of moments

m =Bp, (4.6.23)

where, as usual, B = C-l. Then, from (4.6.14), it follows that

e = Um, (J = lip , (4.6.24)

i.e. the projections U and li give one of the representations of the Green's
operators for strain and stress, respectively.

Problem 4.6.2. Show that

;h.li~p{J = o<tPp> ' (4.6.25)

from which follows the equation of equilibrium in (4.6.9).

In direct notation, in view of (4.6.17,25), we have

Rot U = 0, div li = div . (4.6.26)

84 4. Local Defects

Another representation of the Green's operators in the form of self-adjoint

operators follows from the relations

c = Kp, (J = CKCm. (4.6.27)

This representation is closely connected with the elastic energy (/). In fact,
taking into account (4.6.19), we have

2(/) = <(Jic)= <pIKlp) = <mICKClm) (4.6.28)

This gives an energy interpretation of the Green's operators in terms of the

corresponding field sources.
When considering inhomogeneities in an external field, we shall also need
an operator S, which is connected with the operator K by the relations
S = C - CKC, K = B - BSB. (4.6.29)

The connection of this operator with internal stress will be found in the
next chapter. Here, we consider only the relations which are analogous to those
for the operator K presented above.

Problem 4.6.3. Taking into account (4.6.19), show that

SBS = S. (4.6.30)

From here, it follows that the projections

e = SB, e = BS (4.6.31)

are connected with the operator S, while

e+ = e. (4.6.32)

Problem 4.6.4. Taking into account (4.6.17, 26), show that

div S = 0, Rot BS = Rot, (4.6.33)

dive = 0, Rot e= Rot. (4.6.34)

Problem 4.6.5. Verify that

(4.6.35)

Thus, the strain and the stress can be obtained by projecting the field
sources onto the corresponding subspaces of the Hilbert space. To the pro-
 4.6 Homogeneous Elastic Medium 85

jections li and ff, there correspond the usual stress a (caused by external forces
q, i.e. diva = -q) and the usual strain c (which satisfies the compatibility
conditions Rot c = 0 and is represented in terms of the displacement, as c =
def u ). Conversely, the internal stress a and the internal strain c, for which
diva == 0, but Rot c#-O correspond to the projections and e e. In this case,
the deformation c is not represented in terms of a displacement. A detailed
consideration of the internal stress and the corresponding field sources will
be given in the next two chapters.
Unfortunately, the projection operators introduced above are not self-
adjoint, i.e. the corresponding projections are not orthogonal. This leads to
some mathematical complications, since arbitrary projections do not possess
a number of nice properties inherent to orthogonal ones. This difficulty can
be easily overcome by a renormalization of the field variables, after which
the projections will become orthogonal. The loss of the usual physical meaning
of the field variables, is an inevitable price.
Positive definite tensors such as C and B = C-l, have uniquely determined
self-adjoint square roots C' and B'. Let

7: = B'a = C'c, p. = B'p = C'm. (4.6.36)

Then, (4.6.24) is equivalent to one single equation

7: = Hp., (4.6.37)

where

H = C'KC' = C'liB' = B'ffC' . (4.6.38)

It is easy to see that

HZ = H, H+ = H, (4.6.39)

i.e. we have obtained the desired result.

In terms of the new variables, the expression for the elastic energy takes
the form

(4.6.40)

and for the internal stress we have

e = B'SB' = C'eB' = B'eC' , (4.6.41)

ez = e, e+ = e, (4.6.42)

while
86 4. Local Defects

1=0+8. (4.6.43)

Thus, the solution of the problems of elasticity for both external and internal
stress is reduced to the construction of the orthogonal projections 0 and 8
in the Hilbert space with the scalar product, defined by the elastic energy
(4.6.40). The projections were explicitly expressed above by means of the
Green's operator G for displacement. It can be shown that the converse is
also valid, namely the operator G can be expressed in terms of 0 and 8.
It is essential that if one admits the definition of the operator K in the
form of (4.6.14), then all the relations, starting with (4.6.17), remain valid
for an arbitrary inhomogeneous elastic medium. In connection with this,
a closed system of equations for the projections is of interest. This system will
be deduced in Sect. 4.8.

4.7 The Interface of Two Media

Before we proceed to boundary value problems for elastic media with in-
homogeneities, let us consider a fundamental local problem about a discon-
tinuity of stress and strain at a point on the interface between two media.
Let two elastic media with tensors of elastic moduli cSIlA,u and callA,u be joined.
Assume that, on the (sufficiently smooth) interface {} between them, the
usual conditions of continuity of displacement and of the normal stress vector

(4.7.1)

are satisfied. Let us find an expression for the discontinuity of the stress and
strain at a point on the interface {}.
From the continuity of u on {}, follows the continuity of the tangential
component of the tensor Vu on {}. Introducing the projection operators onto
the normal and onto the tangent plane

(4.7.2)

let us write down this condition in the form

(4.7.3)

To transform the second condition (4.7.1) we have

(4.7.4)

where replacing CAl' by ihu,u is possible due to the symmetry of the tensor of
elastic moduli. Decomposing VtJ! into the sum of normal and tangent com-
ponents, we obtain
 4.7 The Interface of Two Media 87

Using (4.7.3), we obtain the equation for n·Vuo

(4.7.6)

where

(4.7.7)

Applying to both sides of (4.7.6) a matrix GO(n), which is inverse to.Lo(n)

[the matrix GO(n) exists due to the positive definiteness of the elastic energy],
we solve (4.7.6) with respect to n· Vu o. Then, multiplying (tensorially) the result
obtained by n, we find

(4.7.8)

Adding this to (4.7.3) and symmetrizing with respect to the indices A, fl., we
have

(4.7.9)

where

K~pV<1(n) = [n"G~(n)nv] ("p)(VI1) ,

(4.7.10)
K"pvin) = [n"Gpl1(n)nv] ("p) (VI1) .

Using direct notations, one can rewrite (4.7.9), in the form

[c] = - Ko[C]c = - K[C]co· (4.7.11)

or, equivalently,

c = (I - K[ C])co, co = (I + K o[C])c . (4.7.12)

From here it follows in particular that

K - K o + K[ C]Ko = 0. (4.7.13)

Let us present explicit formulae for the discontinuity [c] of the strain on the
interface between two isotropic media for the case, when the x 3-axis coincides
with the normal:

[cn] = [cd = [czz) = 0,

[cIs] =- ([fl.]/ fl.O)cI3, [c23] =- ([fl.]/ fl.O)c23 , (4.7.14)
[c33] = - ([A](cn + C22 + C33) + 2[fl.]c33}/(Ao + 2fl.o) •
88 4. Local Defects

In this simplest case the formulae for [e] can be obtained directly, by obvious
reasoning. The continuity of the tangent component of the tensor Vu follows
from the first condition in (4.7.1). In the coordinate system chosen, this means
the continuity of six components of oau{3 with indices la and 2f3 and, hence,
of the three components of strain ell, e12 and e22' The continuity of stress
components 0'13,0'23 and 0'33 follows from the second condition in (4.7.1). Using
the derived conditions on 0' and e and also Hooke's law for an isotropic medium,
we arrive at (4.7.14).
Calculations of such a kind, in a fixed basis, have an essential defect, namely
they do not enable one to find an invariant meaning of the coefficients in the
formulae for the discontinuities. We shall see how useful it is to have a physical
interpretation of these coefficients.
Notice. The formulae obtained for the discontinuity of the strain are a
consequence of the interface conditions (4.7.1), but are not completely equiv-
alent to the latter, since they are valid under weaker assumptions. In particular,
for the case of the isotropic media considered, the component u3 can have a
discontinuity on the interface that will lead to a a-function term in e33' However,
this a-function will not affect the discontinuity in the limiting values of e33'
It is not difficult to obtain formulae for the discontinuity of the stress
from the formulae for the discontinuity in the strain. We present these formulae
in a form, which is completely symmetric to (4.7.11, 13),

[0'] = - So[B]O' = - S[B]O'o . (4.7.15)

0' = (I - S[B])O'o, 0'0 = (I + So[B])O' (4.7.16)

S - So + S[B]So = O. (4.7.17)

Here

S(n) = C - CK(n)C, (4.7.18)

[B] = B - BO, B = C-1 , (4.7.19)

and similarly for So.

Later, we shall need an explicit relation between 0'0 and e. From (4.7.16)
after simple transformations, we obtain

(4.7.20)

where

(4.7.21)
Let us agree that, when considering a defect in an inhomogeneous medium,
we shall denote by C and Co the moduli of the medium with and without the
 4.8 Integral Equations for an Inhomogeneous Medium 89

inhomogeneity, respectively. At the same time, we introduce the notations

(4.7.22)

while an evident relation

(4.7.23)

holds.
Note, that C 1 = - Co or Bl -+ 00 corresponds to a cavity whereas C 1 -+ 00
or Bl = - Bo corresponds to a rigid inclusion.
In the notations adopted above,

(4.7.24)

while for the cavity, as follows from (4.7.18), S*(n) = So(n).

A critical reader has probably noticed a remarkable correspondence be-
tween local coefficients in the formulae for the discontinuities of stress and
strain and the Green's functions introduced in the previous section. This
correspondence was reflected above in the adopted notations. In fact, the
matrix Go(n) is, by definition, the inverse or the matrix Lo(n), which is given
by (4.7.7). Comparison with (4.6.4 and 5) shows that Go(n) coincides with
the value of the Green's function Go(k) of the homogeneous medium on the
unit sphere in k-space. As regards Ko(n) and So(n), they coincide exactly with
the Green's functions of the homogeneous medium for strain and stress in
the k representation. This correspondence will be used below, in the problem
about an ellipsoidal inclusion.

Problem 4.7.1. Verify that the analogs of (4.6.19, 30) are particular cases of
(4.7.13, 17). (Hint: Set formally C = -Co.)
Problem 4.7.2. Show that (4.7.13, 17) are equivalent to the relations

Bo(n)fi(n) = 0, ffo(n) e(n) = 0 (4.7.25)

or
no(n)B~C'e(n) = 0, eo(n)C~'n(n) = 0 (4.7.26)

where the projection operators are defined by (4.6.20, 31, 38 and 41).

4.8 Integral Equations for an Inhomogeneous Medium

In this section we shall obtain integral equations for inhomogeneous elastic

media in terms of stress and strain. As we shall see this form of the equations
90 4. Local Defects

is the most convenient one for investigating problems about inclusions in a

homogeneous medium.
Let us consider an unbounded elastic medium described by the elastic
moduli tensor C(x), which is representable in the form

(4.8.1)

where Co is a constant tensor and CI(x) is a perturbation. It is assumed that

C(x) and Co are positive definite and CI(x) -+ 0 as Ixl -+ 00.
It is also convenient to introduce the tensor of elastic compliance B(x) ,
connected with C(x) by the relations

(4.8.2)

B(x) = C-I(X), Bo = COl. (4.8.3)

It is obvious that B(x) and Bo are positive definite, too.

The cases of a cavity and a rigid inclusion are to be described separately,
since, for them, C(x) and B(x) are not positive definite. In these cases, generally
speaking, additional assumptions are needed, which ensure the uniqueness
of a solution of the corresponding equations. To avoid this we consider these
cases only as limiting ones: for a cavity V

(x E V) (4.8.4)

and for a rigid inclusion,

(x E V). (4.8.5)

The equations of elasticity in displacements are written In the form

Lu = (Lo + LI)u = q , (4.8.6)

where

(4.8.7)

The quantity q denotes an external force, decreasing sufficiently rapidly as

Ixl -+ 00. For the.se equations to have a unique solution, the conditions at
infinity must be added.
We shall understand (4.8.6) in the sense of generalized functions, so that
we can consider also piecewise continuous functions CI(x). The displacement
u(x) and the normal stress vector n· (j(x) will be automatically continuous on
the surface 0, where CI(x) has a discontinuity, if q(x) does not contain sin-
gularities of the type of the single or double layer on 0, the latter being here-
 4.8 Integral Equations for an Inhomogeneous Medium 91

after assumed. As is shown in potential theory, under the conditions pointed

out above, the (generalized) solution of (4.8.6) exists and is unique.
In order to proceed to the integral equation, let us first introduce additional
restrictions, which we shall drop later. Suppose that C1(x) be a continuous
function with bounded Impport, q(x) be a piecewise continuous function and
the required solution be in the class of functions which tend to zero at infinity.
Let us apply the Green's operator Go of the homogeneous medium to both
sides of (4.8.6). We have

(4.8.8)

where Uo = Goq is an external field which vanishes at infinity, i.e. the field
which would be present in the homogeneous medium under the action of
the force q. By virtue of our assumptions, u(x) and uo(x) are continuously
differentiable functions.
Let us apply the operator def to both sides of (4.8.8). Taking into account
the symmetry of Cb we obtain the equation for the strain t = def u

(4.8.9)

where to = def Uo is an external strain field and

Ko = -def Go def (4.8.10)

is the Green's operator for strain in the homogeneous medium, which was
introduced in Sect. 4.6.

Problem 4.8.1. Show that the equation for stress

(4.8.11)

follows from (4.8.9), where 0" = Ce, 0"0 = Coeo and

So = Co - CoKoCo (4.8.12)

is the Green's operator for internal stress introduced in Sect. 4.6, see (4.6.29).

Let us consider now, how far one can extend the field of applicability of
(4.8.9, 11). We set temporarily

e=to+tl (4.8.13)

where el is a perturbed field, caused by the inhomogeneity C1(x), satisfying

(4.8.14)
92 4. Local Defects

The right hand side contains only the restriction 4 of the external field
co to the support V of the function C1(x) with bounded support. Thus C1 will
not change, if one extends cb outside V in any (smooth) way. Therefore,
the external field co appearing in (4.8.9) can satisfy any conditions at infinity,
c - t co as Ixi - t 00 being automatically satistied. The same conclusion is also
valid for equation (4.8.11).
Let us now assume that C1(x) has a discontinuity on a certain surface D,
and that co is continuous on D. Then, by virtue of the properties of the operator
Ko (Appendix A.3) the solution also has a discontinuity, which satisfies the
conditions (4.7.11), i.e. the necessary conditions at the interface are fulfilled.
This enables us to extend (4.8.9) to piecewise continuous C1(x)'s, in particular,
to functions of the type of a product of a constant tensor and the characteristic
function of a region. Consequently, (4.8.11) can also be extended to piecewise
continuous functions B1(X),
Let us proceed to the extension of (4.8.9, 11) to the case of functions C1(x)
and B1(x) of the oscillatory type. We confine ourselves to the most interesting
case, when C1(x) and B1(x) are almost periodic functions, which are repre-
sentable as series of exponentials with arbitrary wave vectors. Let the external
fields be functions of the same type. Then, the solutions of the equations and,
consequently, the products C1c and B1c belong to the same class of almost-
periodic functions. Thus, we arrive at a problem of extending the operators
Ko and So to functions of the above-mentioned type.
Since the series are understood in the sense of generalized functions, the
action of the operators on exponentials can be taken to be termwise. The
action of an operator on an exponential with a nonzero wave vector is written
in the k representation in the form of a product of the corresponding homo-
geneous function and a a-function with support at a nonzero point. This is
a well-defined operation. Hence it remains only to complete the definition of
the action of the operators on constants, which corresponds to multiplying
the a-function concentrated at the zero point by homogeneous functions,
the latter being a quite nonunique operation.
In order to understand how to overcome this difficulty, let us consider a
model problem. Let an "inhomogeneity" be given by the tensors C1 or B1
which are constant throughout the whole space. Assume first that what is
really given is 0"0 = const. Then it is obvious that

(4.8.15)

It is natural to require that (4.8.9, 11) give the same result. This is possible only
if Ko and So act on constants as follows:

JKo(x - J
x') dx' = B o, So(x - x') dx' = 0, (4.8.16)

while, according to (4.8.12), each of the relations is a consequence of another.

 4.8 Integral Equations for an Inhomogeneous Medium 93

Analogously, if one considers an external field of strain to be really given,

then Ko and So are to act on constants as follows:

JKo(x - x'} dx' = 0, JSo(x - x'} dx' = Co· (4.8.17)

We emphasize once more that such an extension of the definition is equiv-

alent to certain additional physical conditions. 3 There is no "canonical re-
gularization" of the operators Ko and So.
Note that (4.8.9, II) obviously remain valid, if one assumes the external
medium to be also inhomogeneous, i.e. Co = Co(x}. Here, of course, the oper-
ators Ko and So will not be translationaly invariant.
Let us, finally, remove the last restriction. The external field is not to be
caused by force sources only. Sources of internal stress may also contribute to it.
In other words, no restrictions such as Rot co = 0 or div (To = 0, are imposed
on the external field. At the same time, from (4.6.17) and (4.6.33), it follows that
the conditions

Rot c = Rot co' div (T = div (To (4.8.18)

are automatically fulfilled.

Thus, we have shown that the integral' equations in the form of (4.8.9, II)
have meaning for a rather wide class of inhomogeneities and external fields,
i.e. under the indicated conditions solutions of the equations exist, are unique
and coincide with the solutions of the differential equations of elasticity with
the corresponding conditions at infinity.
The solutions of (4.8.9, II) can be represented in a form, which is sym-
metric to that of the original equations

(4.8.19)
(4.8.20)
where K and S are naturally interpreted as the corresponding Green's operators
for a medium with characteristics Cor B while - C 1 and - Bl can be considered
as "perturbations."
Let us emphasize the correspondence between the pairs of integral equations
(4.8.9, 19) and (4.8.11, 20) on the one hand and the pairs of local relations at
the discontinuities (4.7.12, 16) on the other hand. This correspondence is
not accidental. The conditions for the jumps can be obtained from the integral
equations, making use of properties of the Green's operators. Conversely, the
conditions for the discontinuities imply a natural form of the integral equations. 4

3There is an analogy between these asymptotic conditions and the first and second bound-
ary value problems.
40 ne can say that there is a remarkable finite-dimensional representation of the global
Green's and projection operators by local ones.
94 4. Local Defects

In particular, the global analogs of (4.7.13, 17) are

K - Ko + KC1Ko = 0, (4.8.21)

(4.8.22)

Problem 4.8.2. Show that (4.6.19,30) are consequences of these relations.

The fundamental equations for the projection operators

(4.8.23)

are the global analog of (4.7.25). One can prove that these equations together
with the relations

cn = nc, ce = BC (4.8.24)

n+ = n, e+ = B (4.8.25)

and the analogous ones for no, eo, determine the projections nand e uniquely,
if the projections no and eo are given. For the latter, one can take the standard
projections Do, B o, which correspond to the case Co = l. Their explicit expres-
sion is presented in (B.7). Then (4.8.23-25) uniquely determine the Green's
operators and, hence the solutions of the corresponding problems for an in-
homogeneous medium.

Problem 4.8.3. Verify that the system (4.8.23-25) is equivalent to the system

DoBoC'B = 0, BoCoB'D = 0, (4.8.26)

see (4.7.26).

In conclusion, let us consider briefly the case, when C1(x), and consequently,
Bl(X) have a bounded support (a local inhomogeneity). Let V(x) be the charac-
teristic function of a bounded region V occupied by the inhomogeneity, and V
be the complement of V. Let us denote by + (-) the restriction of a field to
the region V (V) and introduce the operators

KJ+) = VKoV, KJ-) = VKoV, (4.8.27)

S~+) = VSoV, S~-) = VSoV,

where V and V denote the operators of multiplication by the characteristic

functions V(x) and I - V(x).
 4.8 Integral Equations for an Inhomogeneous Medium 95

It is obvious that (4.8.9) is equivalent to the pair of equations

(4.8.28)
(4.8.29)
where the first equation uniquely determines the solution inside V and the
second uniquely determines its continuation to V. Analogously, for (4.8.11),
we have
(4.8.30)
(4.8.31)

The correctness of such a representation of the solutions is justified by the

analysis carried out above. In fact, under appropriate conditions on external
fields, the solutions do not have singularities on the boundary D of the region
V and admit the restriction operation. Recall that the functions Cl(x) and
Bl(X) can have discontinuities on D.
Thus, for local inhomogeneities, the problem is in principle reduced to
solving the integral equations (4.8.28) or (4.8.30) inside the region V. We shall
see below that these equations can be effectively solved for ellipsoidal inhomo-
geneities and their limiting cases.

Problem 4.8.4. Show that, for a cavity, (4.8.9, 11) take the form, using (4.8.4),

(4.8.32)

(4.8.33)
and for a rigid inclusion, (4.8.9, 11) take the form, using (4.8.5),

(4.8.34)

(4.8.35)

Note that, in the case of a cavity, c+ is finite and (1+ = O. Conversely, for a
rigid inclusion, (1+ is finite and c+ = O.

Let us now consider a field outside the inhomogeneity, e.g. for definiteness
the stress field. According to (4.8.31), for the perturbed field (1l(X), when x E V,
we have

(1l(X) = - SSo(x - x')m(x') dx' (4.8.36)

where the function
(4.8.37)
96 4. Local Defects

is determined by the solution of (4.8.30). It is convenient to consider the con-

tinuation of m(x) outside the region Vas being equal to zero.
From (4.6.33) it follows that the field O"l(X) satisfies the condition div 0"1 = 0
and can be interpreted as a field of internal stress. In the next chapter, we shall
see that m(x) then plays the role of the density of dislocation moments.
The asymptotics of the perturbed field, as Ixl -> 00, is obtained, if one re-
places m(x) by a a-function source with the dislocation moment

(4.8.38)

In this sense, one can speak about a point inhomogeneity in an elastic

medium. However, as distinct from the case of the elastic quasicontinuum,
it would here be incorrect to introduce a a-function directly into the elastic
moduli or compliances, since the corresponding equations, as one can easily
verify, have no solutions. In connection with this, a model of a point defect in
an elastic medium requires a delicate treatment. In particular, the corresponding
dislocation moment (4.8.38) cannot be given a priori, but is to be obtained from
the solution of the problem for a finite inhomogeneity. In other words, a point
defect "knows" about its origin, as distinct from a point defect in a quasicon-
tinuum (Sect. 4.3).

4.9 Ellipsoidal Inhomogeneity

In this section we consider a case which is important for applications, namely

(4.9.1)

where C l = const and Vex) is a characteristic function of an ellipsoidal domain

V. We show that the solutions of the most interesting problems for such el-
lipsoidal inhomogeneities can be obtained explicitly. This fact is based on the
remarkable property of polynomial conservation (the p-property) for the el-
lipsoidal domain.
Theorem. If the external field co(x) in the neighborhood of V is a polynomial
of degree m, then the field c+(x) induced within V is a polynomial of the same
degree.
Proof Since the statement of the theorem is invariant with respect to non-
degenerate linear transformations, it is possible without loss of generality to
consider V to be the unit sphere. As usual, we suppose that Co + Cl Vex) is
the positive-definite tensor, i.e. the limiting cases of a cavity, and a rigid in-
clusion is excluded for the moment. Then the operator 1+ KtC l in (4.8.28)
has an inverse, and to prove the theorem it is sufficient to show that a polynomial
within V is transformed by the operator KJ+l into a polynomial of the same
degree.
 4.9 Ellipsoidal Inhomogeneity 97

It follows from the definition (4.8.27) of the operator KJ+) and from the
property (C.I) of the generalized function Ko(x) that KJ+) is decomposed into
two parts: the first one is proportional to the identity operator and the second
one is determined by the regularized part of Ko(x), an arbitrary region contained
in the regularization being naturally identified with V. As a result, it is sufficient
to consider the regularized part only. The latter acts on a function q>(x) which is
smooth within V according to the rule

q>(x) -+ cf;(x) = JKo(x - x')[q>(x') - q>(x)] dx', x E V, (4.9.2)

V

where Ko(x) is the formal second derivative of Go(x) defined by (4.6.13). Taking
into account that Go(x) """ Ixl-1 and Go( -x) = Go(x) we can rewrite (4.9.2)
in the form

cf;(x) = Jgo~n) [q>(x') - q>(x)]dx', x E V (4.9.3)

v r

where

r = 1x - x' I, n = x' -r x (4.9.4)

and go (n) is a tensor function of fourth order, even in the argument n.

Let us consider the action of this operator on a homogeneous polynomial or,
equivalently, on the monomial x(m) = x'} ... x'm where (m) is the tensorial
multi-index. Set

F(m)(x) = J g(~) (x'(m) - x(m») dx', x E V. (4.9.5)

v r

Substituting from (4.9.4) x' = x + rn, we have

m
F(m)(x) = 1: x(m-k)J<k>(x) , (4.9.6)
k=l

where

J<kl(x) = Jgo(n)n(k)rk- 3 dx', x E V. (4.9.7)

V

*
Let us set dx' = r 2drdwand integrate first over an elementary cone dw(n)
with vertex at the point x E V. Then

J(k)(X) =
{J)
Jgo(n)n(k)pk(x, n)dw, x E V, (4.9.8)
98 4. Local Defects

where

p(x, n) = - x·n + .vI - x·x + (x·n)2 (4.9.9)

is the distance from the point x in the direction n to the boundary of the unit
sphere V. For pk we have

k
pk(X, n) = ~ (-I)ICL(x.n)l [.vI-x.x + (x.n)2] k-I. (4.9.10)

Since go(n) = go( -n) the contribution to J<kl(x) is due only to terms whose
product with n(kl is even. It is readily seen that they contain the radical in an
even degree and have the form

.L: A (k-21) X(k-21) .

I

As a consequence, J<kl(x) is a polynomial in x of degree k

J<kl(x) = .L: BiZL2J) X(k-21) , X E V, (4.9.11)

I

where

Bt%L 2J) = 1Jgo(n)n(klA(k_21)(n)dw

w
(4.9.12)

is a constant tensor.
Substituting (4.9.11) into (4.9.6) we find that F(ml(x) is a polynomial of
degree m. Finally, it is easy to show that the term x(ml has a nonzero coefficient.
Q.E.D.
Note 1. To extend the theorem to the cases of a cavity and a rigid inclusion
we have to make a further assumption. It follows from the proof that for the
cavity (C 1 = -Co) a weaker statement is valid: the operator 1- KJ+)Co
transforms a polynomial c:+(x) into a polynomial C:b(x) of the same degree but
equation (4.8.28) is not solvable for all c:b(x). This means that the theorem is
valid if and only if conditions of solvability are fulfilled. It is essential that for
the most important case, C:b = const, (4.8.28) always admits a solution as will
be shown below. The extension of the theorem to the case of a rigid inclusion
is performed in the same way but the corresponding field variables will be
stresses satisfying (4.8.30).
Note 2. It follows from the proof that even (odd) c:o(x) corresponds to even
(odd) c:+(x). In particular, if c:o(x) is linear then c:+(x) is also linear.
Note 3. Using the Green's functions for homogeneous elliptic operators of
the order 2p [4.2] it is easy to generalize the theorem to the case 2p < n where
n is the spatial dimension, n being arbitrary for p = 1.
 4.9 Ellipsoidal Inhomogeneity 99

Note 4. It is most likely that the property of polynomial conservation is

valid only for the ellipsoid, i.e. that this is a characteristic property of the
ellipsoid. However, a proof of this statement is not known to the author.
Before considering the consequences of the theorem let us introduce relations
which are useful for what follows. Let us write the equation for the ellipsoid D
in the form

(4.9.13)

If the Cartesian coordinate system is connected to the ellipsoid semi-axes a10

a2, a3, then the tensor aa{3 is diagonal, the semi-axes being the diagonal com-
ponents.

Problem 4.9.1. Verify the relations

(4.9.14)

These formulae establish a one-to-one correspondence between a function

f(x) on the ellipsoid D and a functionf(n) on the unit sphere n2 = 1. In partic-
ular,

p(x) == (xa- 4x)-1I2 +-> p(n) = (na 2n)-1I2. (4.9.15)

We define the mean value off(n) over the unit sphere n2 = I by the expression

<f(n) = ~ J
det a f(n) p3(n) dn . (4.9.16)

Problem 4.9.2. Show that <f(n) coincides with the mean value of f(x) over
the ellipsoid D defined by the expression

_ I f(x)
<f(x) - 4ndeta! p(x) dx, (4.9.17)

which permits one to interpret <f(n) as the mean value over the ellipsoid D.

Now, let us consider consequences of the theorem. Let the external field
co(or 0"0) be homogeneous, i.e. constant. It follows from the theorem that in
this case c+ = const and 0"+ = const. To calculate them explicitly we use (4.8.28)
and (4.8.30), rewriting them in the form

(4.9.18)

(4.9.19)
100 4. Local Defects

It is seen from here that we need to know the action of the operators KJ+) and
SJ+) on constants. Let us recall that these operators are represented in the form

KJ+) = AI + Kd+), So<+) = DI + So<+) , (4.9.20)

where A and D are constant tensors and the tilde denotes the regularized parts;
these and the constants A and D depend on the choice of the regularization
region. In the case under consideration it is natural to identify the regulariza-
tion region with the ellipsoid. Then we have for the constants A and D (Ap-
pendix C)

(4.9.21)

where the integration is carried out over the unit sphere ()} = 1. Introducing
the new variable

(4.9.22)

and taking into account that Ko(k) and SoCk) are homogeneous functions of
zero degree, we find

A = (Ko(n» , D = (So(n» . (4.9.23)

It is readily seen from (4.9.2) that in the case e+(x) = const the contribution
of the operator KJ+) is equal to zero. Hence, the action of KJ+) on a constant
results in multiplication of this constant by A. In the same way the action of
Sd+) results in multiplication by D. Finally, we have

(4.9.24)

(4.9.25)

Thus, to find e+ and ()'+ we must invert the corresponding constant matrices.
It can be shown that the inverse matrices always exist including the limiting
cases of the cavity and the rigid inclusion.
Note that (4.9.23) permits one to rewrite the equations for e+ and ()'+ in the
form

(4.9.26)

(4.9.27)

A comparison with the discontinuity relations (4.7.12, 16) shows that the
coefficients in the equations admit an interesting interpretation: they are equal
 4.9 Ellipsoidal Inhomogeneity 101

to the mean values over the ellipsoid of the coefficients in the corresponding
relations for jumps.
Now let us consider the case for which the external field is a linear function
of x

(4.9.28)

It follows from the theorem that

(4.9.29)

the tensors band d being linearly dependent. It is convenient to write this de-
pendence in the form

(4.9.30)

To find the tensor A let us multiply (4.9.18) by x and integrate over V. In a

manner analogous to the case of the homogeneous field we obtain

A = <p(n)n(l + KO(n)C1)np(n» . (4.9.31)

The solution of the problem for linear stresses can be deduced in the same way.
If the external field is a polynomial of the second or higher degree, then the
p-property permits one to find from (4.9.18, 19) recurrent algebraic equations
which connect the coefficients of the polynomials eo(x) and c+(x) or (7o(x) and
(7+(x). However, in contrast to the case ofthe homogeneous field these equations
are solvable only under additional conditions. The explanation is as follows:
formally there exist external fields (they do not satisfy the equilibrium condition
div (70 = 0) which are not perturbed by the ellipsoidal inhomogeneity. For ex-
ample, in the isotropic case this property has an external field of the type
ca{3 = (c,x)oa{3 where C is an arbitrary constant vector. As a consequence the
matrix A can be inverted only on the corresponding subspace.
Let us consider one more problem which is of interest for applications.
Let (7(x) be the limiting value, from outside, of the stress at the point x E D.
As above, instead of (7(x) we consider the functionf(n) of the normal n. The de-
pendence (7(n) upon the external field (7o(x) can be represented in the form

(4.9.32)

where F(n) is an operator concentration coefficient. For polynomial external

fields the operator F(n) can be constructed in an explicit form. For the sake of
simplicity we confine ourselves to the case of a homogeneous external field.
It is convenient to construct the operator F(n) in two steps. According to
(4.7.20) we have on the boundary D of the ellipsoidal inhomogeneity

(4.9.33)
102 4. Local Defects

where in this case $+ = const and

(4.9.34)

Using (4.9.27), we have

(4.9.35)

and consequently

(4.9.36)

In particular, for the cavity C 1 =- Co and S*(n) = So(n). Taking into account
(4.9.23) we have

F(n) = So(n)D-l . (4.9.37)

It is interesting to note that this problem as well as the previous ones, were
solved for an arbitrary anisotropy without use of the explicit form for the
Green's tensor Go(n). It is sufficient to know its Fourier transform Go(k). In
other words, we find stress concentrations even though we have not been able
to solve the corresponding boundary value problem. This is a direct conse-
quence of the property of polynomial conservation.
The representation for F(n) obtained above is especially convenient for the
investigation of the dependence of the concentration coefficient on the shape
of the ellipsoid. Indeed, the local dependence of F(n) on the normal to the el-
lipsoid (or equivalently, on the point) is given by the first factor S*(n), which
is independent of the ellipsoid parameters and always remains finite. The second
factor <S*(n»-l does not depend on the point and becomes infinite in the case
of the passage to the limit of the elliptical crack.
It follows from (4.9.16) that the entire dependence of <S*(n» on the ellip-
soid parameters is concentrated in the scalar weight factor pen). This is essential
since it permits a complete investigation of the stress concentration in the limit-
ing cases of an ellipsoidal crack and a needle for an arbitrary anisotropic me-
dium by a series expansion of pen) for small parameter.
Note the interesting relationships

<F(n» = I, <IT(n)) = lTo (4.9.38)

which immediately follow from (4.9.36) and (4.9.32), respectively.

In conclusion, we present some results for the ellipsoidal cavity in an iso-
tropic medium.
The tensors A and D in this case have the symmetry of the ellipsoid and
are defined by nine essential components. We have in the coordinate system con-
nected to the ellipsoid axes
 4.9 Ellipsoidal Inhomogeneity 103

Allll = Ko [3111 + (1 - 4]..1o)It], A1122 = Ko(I21 - It) ,

(4.9.39)
A1212 = Kl [121 + It2 + (1 - 2]..10) (It + 12)] ,

1
Dllll = To [1 - 81t (3111 + It)] ,

D1l22 = To {]..I0 - 1~7t' [121 + 112 - (1 - 4]..1o)(It + 12)]} ,

(4.9.40)
D1212 = To {I -; ]..10 - 1~7t' [121 + It2 + (1 - 2]..1o)(It + 12)]} ,

where

1 2po (4.9.41)
KO =
161tpo(1 - ]..10) , To = 1 - ]..10 '

the quantities

_ 3 2
_ 3
Ip - "2 V!
00 d~
(a~ + ~)Ll(~)' Ipq - "2 va p ! (a~ + ~)(a~ + ~)Ll(~)
00 d~

(Ll(~) = .J(ar + ~)(a~ + ~)(a§ + ~); p, q = 1, 2, 3) (4.9.42)

are expressed in terms of elliptic integrals (v is the ellipsoidal volume). The

remaining six tensor components are obtained by a cyclic replacement of the
indices 1, 2, 3.
Note that in the particular case of an ellipsoid of revolution only six essential
components remain and the formulas do not contain elliptic integrals.
In the case of a cavity according to (4.9.37) we need expressions for So(n)
and D-1. In the basis Ei (Appendix A)

(4.9.43)

As is customary, in order to evaluate D-1 we represent the tensor D of fourth

order as a sixth-order matrix. The latter is decomposed into two blocks of third-
order matrices. The components DJ.Ap.p. form a symmetric martix, and the
doubled components Dlp.lp. (A :f: p) a diagonal matrix; consequently

(4.9.44)

Let us consider two cases. First, let the external field O'ffP be the tension
along the coordinate axes, i.e. only the diagonal components 0'$" are nonzero.
Then the stress concentration is defined by the components
104 4. Local Defects

(4.9.45)

Secondly, let the external field off" be pure shear. Thd stress concentration
is determined by the components

(4.9.46)

The formulas obtained permit one to investigate the stress concentration

over the whole surface of the inhomogeneity. This is essential because the
knowledge of the concentration coefficient just at the vertices of the ellipsoid
is not sufficient and can result in incorrect qualitative deductions. For example,
for pure shear with the components oi:?, the stress tensor oa" is zero at the vertex
(l, 0, 0), but there is a stress concentration in direct proximity to this vertex.
Hence a complete investigation of the stress on the whole ellipsoid surface
is necessary, and not only at its characteristic points. Such an investigation will
be given in the next section for limiting cases of the ellipsoidal inhomogeneity.

4.10 Ellipsoidal Crack and Needle

It is convenient to introduce the dimensionless parameters of an ellipsoid

(4.10.1)

where al ;::: a2 ;::: a3 are the ellipsoid's semiaxes. The case a <:: 1, ~ '" 1 cor-
responds to the needle, while ~ « 1, a '" 1 to the crack, and e « a « 1 to a
narrow crack. We note that in all cases we have e = ~a « 1.
It follows from (4.9.36), that the solution of the stress concentration prob-
lems for a needle and a crack reduces to the computation of the principal
terms in the expansion of the tensor D-l with respect to the corresponding
small parameters. From (4.9.23) and (4.9.16) we have

D = d~~a JSo(n)p3(n) dn . (4.10.2)

First we carry out the computation for a needle. Eq. (4.10.2) is transformed
to the spherical coordinates q;, () with the polar axis along the axis of the needle,
i.e. along Xl. We perform the change of variable cos () = t and we set

(4.10.3)

Without loss of generality we can assume that D(q;, t) is an even function of

 4.10 Ellipsoidal Crack and Needle 105

t (only this component contributes to the integral (4.10.2». We have

~ dtp
D - -2 J JD(q;, t)f1(t, al) dt ,
_ 21t 1
2 (4.10.4)
11: 0 cos q; + .;2·sm2 q; -1

where

(4.10.5)

For a needle a --+ 0, .; ,...., 1 andf1(t, al) --+ oCt). Consequently, the principal term
in the expansion of D with respect to a has the form

D = ~2f1t D(q;, 0) d (4.10.6)

o 211: 0 cos 2 q; + ';2 sin 2 q; q; .

Thus, for an arbitrary anisotropic medium the problem is reduced to the

computation of a simple integral. If the tensor Do has an inverse (detDo =1= 0),
then the coefficient of the concentration tends to a constant value when a --+ 0,
i.e. it has no singularity. The computations show that det Do =1= if the sym-
metry of the medium is not below the rhombic symmetry (orthotropic, hexa-
°
gonal, cubic and others 5). Obviously, this holds also in the case of an arbitrary
anisotropy. The tensor of the elastic constants of the indicated medium has
nine nonzero components, denoted according to the usual rule by

Cff a{38 = ca{3 (a, [3 = 1, 2, 3)

(4.10.7)
C6 323 = C44, c&313 = C55, C6 212 = C66.
For an orthotropic medium all the nine components are essential, while for a
transversely isotropic medium only five components play a role:

(4.10.8)

For the case of cubic symmetry we have three constants

(4.10.9)

and, finally, for an isotropic medium

(4.10.10)

In all these cases the tensor DOl can be found in explicit form. For its

SHere and in the following we assume that the symmetry axes of the medium coincide
with the axes of the ellipsoid.
106 4. Local Defects

computation it is necessary to take n1 = 0 in the tensor So(n), insert the expres-

sion obtained into (4.10.6), integrate and invert the tensor Do. However for
the component of DOl with the indices (aa{J{3) corresponding to extension
along the axes, the expressions become cumbersome because it is necessary
to invert a third-order matrix. Therefore here we give only the nonzero shear
components

(4.10.11)

1
~(~A1VU1U2 + B1 ~ + C1vlu1D' (4.10.12)

Here Ul and U2 are the roots of the quadratic equation

(4.10.13)

The computations are considerably simplified in the isotropic case. Taking

into account (9.43), we obtain the zero components of DOl('1)o = [2,uo(1 + )./0)]-1)

(DOl)l1l1 = '1)0, (DOl)1122 = (DOl)1l33 = - )./0'1)0

(DOl)2233 = -'1)0(1 - 2)./ID, (DOlh222 = '1)0[1 + 2(1 - )./ID~] (4.10.14)

(DOl)3333 = '1)0 (1 + 2(1 ; )./2)), (DOl)1212 = 1to~

-1
(Do h313
1 (
= 4,uo 1
1)
+~ , _
(Do 1)2323 = 2"0
1 (1 +
~
~)2

Inserting DOl into (4.9.37) we can obtain, according to (4.9.32) the stress
q(n) at the surface of the needle for an arbitrary external field qo.
The expressions (4.10.12,14) give an obvious mechanism for the appearance
of the singularity in the conversion from a needle to a narrow crack, i.e. for
~ -+ 0: a singularity of order ~-1 appears only for external stresses qo con-
taining components with the index 3.
Now let us consider an arbitrary crack ~ « 1, a '" 1. In the same way
as in the case of a needle, we transform (4.10.2) to spherical coordinates 'P, 0,
but with the polar axis oriented along x 3• Setting

~1 = ~a., t = cos 0 (4.10.15)

VCOS2'P +2 2 a sm 'P
 4.10 Ellipsoidal Crack and Needle 107

and making use of the definition (4.10.3) for D('P, t), we have

_ a 210 d'P
D - -2 J f
1
D('P, t)/2(t, ~1) dt (4.10.16)
'Jt 0 cos 'P + a sm 'P-l
2 2 . 2

with

We write the expansion of D as a function of ~ in the form

(4.10.17)

In contrast to the case of the needle, for the computation of the principal
term in the expansion of D-l with respect to ~ we have to retain the first two
terms in D since the tensor Do in (4.10.17) does not have an inverse, i.e.
det Do = O. It is sufficient to show that all the components of the tensor Do
containing the index 3 are equal to zero. In fact, as ",ill be shown below, in the
integral for Do the tensor DC'P, 1), which corresponds to So(n) for nl = n2 = 0,
n3 = 1, occurs. But from the general expression (4.9.43) for So(n) it follows
that in the case of a rhombic structure, the corresponding components of
So(n) which contain the index 3 are equal to zero for nl = n2 = o. We should
expect that this is true also for an arbitrary anisotropy.
For the computation of Do and Db we consider12(t, ~l) as a generalized func-
tion on the segment It I :::;; 1 with identified points ± 1, i.e. on the circumference.
This is possible by virtue of the fact that D('P, t) is even and continuous. We
can verify that for ~l -. 0

12(t, ~l) = oCt ± 1) + 2(1 !lt 2)3/2 + O(~D·

Substitution into (4.10.16) gives

D('P, ± 1)
D = ~21O
o !
2'Jt cos 2 'P + a 2 sin 2 'P d'P,
(4.10.18)

where the latter integral is written in regularized form. We note that the ex-
pressions obtained become considerably simpler for the circular (a = 1)
and for the narrow (a « 1) crack. In the isotropic case the integrals (4.10.18)
can be easily computed and the inversion of the tensor (4.10.17) allows us to
obtain explicit expressions for the components of the principal term of the
expansion of D-l
108 4. Local Defects

Here K(a) and E(a) are complete elliptic integrals of the first and second
kind, respectively. The remaining components «an be considered equal to
zero with the accuracy of 0(1). From this it follows that a contribution to
the singular stress O"(n) at the surface of the crack is given only by the com-
ponents of the external field 0"0 which have index equal to 3.
Passing to the narrow crack a --+ 0 yields

(D-l)l3l3 = 4~0' (D- lb23 = 2~0 ~ ,

(D- lh333 = K~ ~ , (4.10.20)

which coincides with (4.10.14) for ~ --+ O. In the case when the medium has
a rhombic symmetry, the tensor D-l has a similar structure and its components
can be expressed in the general case in terms of elliptic integrals. They become
considerably simpler for the narrow crack and have the form

(4.10.21)

which, obviously, in the particular case of the isotropic medium coincide with
(4.10.20).
We emphasize that, representing an independent interest, the case of the
narrow crack not only makes obvious the mechanism of the variation of the
stress concentration when passing from a needle to a crack, but also allows
us to pass to the plane problem if we consider the median section of the crack.
We now proceed to the direct investigation of the stress at the surface of
the crack. We start with a pure extension. Since the components 0"6 1 and 0"32,
according to (4.10.19), do not contribute to the singularity, we consider only
the extension 0"83 along the axis x 3 normal to the crack.
From (4.9.32, 37) and (4.10.19-21) we find

(4.10.22)

It is convenient to carry out the investigation of O"a~(n) in a local system of

 4.10 Ellipsoidal Crack and Needle 109

coordinates X"', connected at each point of the surface of the ellipsoidal crack
with the normal n. As a local basis we take

(4.10.23)

where ea is the unit vector of the coordinate system connected with the axes
of the ellipsoid, while the vector nO is normal to the boundary of the crack
at the point (nh n2, 0). It is obvious that for the points of the surface near
the edge, which are of primary interest, the axes XlI and x 2' are, respectively,
parallel and normal to the edge. With respect to the local axes, the tensor
O'alfJ'(n) is planar since all components with index 3 are equal to zero. This
follows from the equilibrium conditions.
First we consider the isotropic case. Transforming in (4.10.22) to local
coordinates and taking into account (4.9.43) and (4.10.19, 20), we obtain
O'1 /1'(n) = l.ioO'max(1 - n§), O'2 /2/(n) = O'max(1 - n§), O'1'2 /(n) = O. (4.10.24)
Here for the finite and the narrow crack we have, respectively,

,,_ 2 .l "33,, _ .2.. _13

(4.10.25)
Vmax - E(.vl _ a 2) ~ vO, Vmax - ~ Uo •
Thus, for the given load the selected system of coordinates coincides, to the
accepted degree of accuracy, with the principal axes of the tensor O'(n). Since
the principal stresses 0'1 = 0'1'1 and 0'2 = 0'2/2' have the same sign, the maximum
1

tangential stress is 't"max = 10'21/2. It is clear from (4.10.24) that in the case of an
extension the largest value of the stress in any cross section normal to the edge
is attained for n3 = 0, i.e. on the edge itself.
In the anisotropic case the computations are similar but more cumbersome.
We give the final expressions for the stresses O'(n) at the surface of a narrow
crack in an orthotropic medium in the cross sections nl = 0 and n2 = O. In the
selected local basis, which in the present case coincides with the principle
axes of the tensor O'(n), we have for the cross section n2 = 0

O'l(n) = ~-lC55[cnC23 - C12C13) n~

+ (C12C33 - C13c23)n§] n~(D-l)3333 O'~3 , (4.10.26)

O'2(n) = ~-lC55(CnC33 - c1~)n~(D-l)3333 ~3

~ = Cn C55n f+ (Cn C33 - 2C13C55 - c~3)n~n§ + c33c55n~ . (4.10.27)

Here (D-l)3333 is given by (4.10.21). The stresses in the cross section nl = 0 are
obtained by replacing in the right-hand sides the indices 1 and 5 by 2 and 4,
respectively.
110 4. Local Defects

We consider now a pure shear. The component (J6 2 does not contribute to the
singularity and the situation for (J6 3 and (J5 3 is similar. Therefore we assume that
the external field coincides with (J6 3.
In the isotropic case, we find from (4.9.32,37) and (4.10.19)

(4.10.28)

If in the above-introduced local coordinate system we perform a rotation

around the normal n, i.e. the axis x3 /, through an angle

(4.10.29)

then the new axes are principal for (JaIW(n). For the principal stresses we find

(JI(n) .v
= (JI/I/(n) = (Jmax [(1 - ).io) (1 - nD (1 - n~) - (1 + vo) nln3] ,
.v
(J2(n) = (J2 /2/(n) = - (Jmax [(1 - vo) (1 - ni)(1 - n~) + (1 + vo) nln3] .
(4.10.30)

Here for the finite and the narrow crack we have, respectively,

(4.10.31)

The largest absolute value of the principal stresses 0"1 and (J2, equal to (Jmax,
occurs at the cross section n2 = 0 in the points

n* = (± )2,0, + )2)' n* = (± )2,0, ± )2) (4.10.32)

which are the maxima (minima) of the functions (JI(n) and (J2(n). In this cross
section (J1 and (J2 are equal to zero at the edge. It is essential that the maximum
point is situated at a very small distance from the edge of the crack, which
according to (4.9.14,30) is

(4.10.33)

We can also show that the characteristic width of the peak is of order c. By
varying the cross section from n2 = 0 to n1 = 0 the magnitude of the peak de-
creases while the point of the maximum approaches the edge coinciding with
it for n1 = O. For the cross sections in which (J1 and (J2 are of the same sign, the
character of the variation of '"max is evident. In particular, in the cross section
 4.11 Crack in a Homogeneous Medium 111

n2 = 0 the quantity 't"max has a sharp peak of height O'rnax/2 at the point n* and
vanishes at the edge. In the case when 0'1 and 0'2 have different signs

't"max = (1 - ).IO)O'max.v(1 - nt)(1- n§). (4.10.34)

This quantity reaches its maximum value equal to (1 - ).10) O'max at the edge in
the cross section n1 = O.
Thus, in the case of shear, a remarkable increase of stress takes place, which,
if not taken into account, wi11lead, in numerical computations, to qualitatively
irregular results on the stress concentration in the crack. This shows the neces-
sity of investigating the stresses not only at the characteristic points of the
contour of the crack, but also over its entire surface.
A similar phenomenon takes place in an anisotropic medium. The principal
stresses at the surface of a narrow crack in an orthotropic medium in the
cross section n2 = 0 are

O'l(n) = - 4a-1c55 [(CnC23 - C12C13) nr

+ (C12C33 - C13c23)n~] n1n3(D-1)1313 0'63 ,

0'2(n) = - 4a-1c55(CnC33 - Cr3) n1n3(D-1)1313 0'6 3 •

Here a and (D-1)1313 are given by (4.10.27 and 21), respectively. In the cross
section n1 = 0

O'l(n) = - 0'2(n) = 4 C55 c 66n 2

+ C55n23 (D-1) 13
C66n 22 1313 0'0 .

Obviously, as in the isotropic case, the increase effect takes place in the cross
section n2 = O.

4.11 Crack in a Homogeneous Medium 6

Let us first consider a flattened cavity, which occupies a finite simply-connected

region V with a smooth boundary. Then, according to (4.8.42, 43), the corre-
sponding integral equations have the form

e(x) - JKo(x - x') Coe+(x')dx' = eo, (4.11.1)

O'(x) + JSo(x - x') e+(x')dx' = 0'0 • (4.11.2)

From (4.11.1) one can find e+(x); then (4.11.2) determines the field O'(x) outside
the cavity.

6Sections 4.11 and 4.12 were written in collaboration with S. K. Kanaun.

112 4. Local Defects

Let us consider the limiting transition from the cavity to a crack, i.e. to a
cut along a smooth oriented surface D which is bounded by a closed contour
r which lies on the boundary of V. Let us choose a local coordinate system at
the point x' E D, such that its z-axis is directed along the normal n(x'). Let
h(x') be the transverse dimension of the cavity and Zl(X', h), Z2(X', h) be the co-
ordinates of the points of intersection of the z-axis with the boundary of V,
while Zh Z2 ~ 0, as h ~ 0. For a fixed point x ~ V, the kernel So(x - x') in
(4.11.2) is a smooth, bounded function. Hence,

f So(x - x') e+(x')dx' = f So(x - x') e+(x', h)dD' + O(h) , (4.11.3)

o
where
z2(x, h)
e+(x, h) = f e+(x + zn(x»dz, xED. (4.11.4)
Zl(X, h)

Note that the quantity e+(x, h) can be interpreted as the coefficient of the
principal term of the expansion of the function e+(x) V(x) in multipoles, which
are concentrated on D,

e+(x) V(x) = e+(x, h)a(D) + (4.11.5)

Here a(D) is the a-function concentrated on D (Sect. 11.1).

It is important that all other terms in this expansion vanish as h ~ 0,
i.e. the total strain field e(x) contains a singular component proportional to
a(D).
On the other hand, for the limiting case of a crack, the stress u(x) and dis-
placement u(x) satisfy the conditions

n·u/o = 0, [u]/o = b, (4.11.6)

where b(x) (x E D) is the vector of a discontinuity in displacement (or the

opening of the crack), i.e. a quantity which is to be found from the solution of
the crack problem. From here, it follows that the singular component of e(x)
for the crack has a coefficient which is equal to the symmetrized product of
the vectors band n. A comparison with (4.11.5) yields

(4.11. 7)

From this, in particular, it is seen that, as h ~ 0, the normal components of

s+(x) tend to infinity, whereas the function e+(x) is bounded.
Thus, with (4.11.2, 7) and the symmetry of the tensor So(x) taken into
account, the stress field outside the crack is connected with b(x) by the rela-
tion
 4.11 Crack in a Homogeneous Medium 113

o{x) = O"o(x) - f So(X - x')n(x')b(x') dO' . (4.11.8)

o
From this and from the first condition (4.11.6) we find the equation for the
vector field b(x) on 0

f T(x, x')b(x')dO' = n(x)O"o(x), x EO, (4.11.9)

o

where
T(x, x') = n(x)So(x - x')n(x') . (4.11.10)

The operator T, defined by these relations, admits the following regular

representation on continuously differentiable on the surface 0 functions b(x)
(Appendix A.3):

(Tb)(x) = f T(x, x') [b(x') - b(x)]dO' - n(x)r(x)b(x) , (4.11.11)

o
where r(x) is a tensor-valued function, which depends on the contour r.
The operator T is continued to a wide enough class of generalized functions
on 0 and belongs to the class of pseudo-differential operators. For functions
which one has to deal with in the crack problem, the existence and uniqueness
of a solution of an equation of the type Tb = f is shown, for example, in
[4.3]. For what follows, it is essential that the asymptoties of the solution of this
equation near r for smoothf(x) has the form

b(x) = (3(xo)r 1l2 + 0(r 312) , (4.11.12)

where r is the distance from the point x along the normal to a point Xo E r,
(3(xo) is a smooth function on r.
Let ut consider the asymptotics of the stress field O"(x) outside the crack in
the neighborhood of the edge r of the crack. Let Yh Y2, Y3 be the local cartesian
coordinates at the point Xo E r,
the Y2-axis being directed along the limiting nor-
mal to 0 at the point xo, and the Y3-axis being tangent to r.
Then the Yraxis
lies in the plane which is tangent to 0 at the point Xo. Taking into account
(4.11.12), we have

(4.11.13)

Using (4.11.8) and (4.11.3) we can write down the expression for 0" at the
point z = (- p cos (), - p sin (), 0) (p > 0)

o"(Z) = O"o(z) - .: f So(cos () + ~{, sin () + ~~, ~~) n(p~'){3(p~')"; ~l dO'

'\! P0
+ O(";p). (4.11.14)
114 4. Local Defects

Here ~ = P-1Yj and it is taken into account that So(x) is an even homogeneous
function of degree - 3. It is easy to show that, as p ~ 0, the integral tends to
a finite limit, and hence, the stress has a singularity of the order of p-ll2.
Let us introduce a tensor coefficient of stress intensity, which is of interest
for applications and defined by the expression

J(O, xo) = lim

p-+O
-.lP o{z) . (4.11.15)

From (4.11.14) it follows that

(4.11.16)

where

s(O)
00

o
00

-00
°
= - J -.I ~l d~l J So(cos + ~l> sin 0, ~3) d~3 (4.11.17)

and n(xo) is the limiting normal to () at Xo E r. In a more detailed description,

(4.11.18)

The intensity coefficient J admits an obvious interpretation, if one observes

00

that J~ooSo(~l> ~2' ~s> d~3 coincides, in essence, with the Green's tensor for stress
in the problem of plane deformation and complex shear (in dimensionless
variables ~) of a homogeneous medium with moduli Co, the normal to the
plane coinciding with the ~3-axis. Then J coincides with the stress at the point
~l = - cosO, ~2 = - sinO, when a discontinuity of the displacement vector
(J(xo) -.1ft is given along the positive semi-axis of ~l. In the isotropic case,
the function J(O) is easily found in an explicit form.
The tensor J can be represented as a sum of three tensors, which correspond
to the three components of the vector 1',

(4.11.19)

where

(4.11.20)

(do not sum over i!). The tensors J 1 and J 2 are found from the solution of
the corresponding planar problem, and J 3 is derived from the solution of an
anti-planar problem (complex shear). All these problems can be solved by
known methods for the case of an arbitrary anisotropy.
Note that in the case of linear fracture mechanics, the asymptotics of the
stress field in the neighborhood of the edge of a crack is always characterized
by the stress intensity factors K 1, K II , KIll. It can be shown that these
 4.12 Elliptic Crack 115

factors coincide with the three components of ~(xo) up to constant factors,

the latter depending on the elastic constants.
Since the vector ~(xo) depends linearly on the external field, it can be re-
presented in the form

(4.11.21)

where B(xo) is the corresponding linear operator. Then J becomes

(4.11.22)

where
(4.11.23)
This representation is convenient since s(O)n(xo) depends on a neighborbood
of the point Xo only, and the operator B(xo) depends on the whole surface Q.
Thus, the calculation of the tensor intensity coefficient J is finally reduced
to solving a certain planar problem and constructing the operator B. An
explicit expression for the latter, for the case of an elliptic crack, will be obtained
in the next section.

4.12 Elliptic Crack

An elliptic crack is the simplest case which is of interest for applications and
admits an exact solution.
Let Q be a plane cut with an elliptic boundary in an elastic medium, and
b(x) be the discontinuity of the displacement vector at the cut. We have the
following analog of the theorem about polynomial conservation: for a poly-
nomial external field O'o(x) of degree m, the vector b(x) has the form

(4.12.1)

where B(xl> xz) is a polynomial in Xl and Xz of a degree no higher than m

(here the coordinate system Xl> Xz is connected with the principal axes 2al
and 2az of the ellipse). 7
In order to prove this, let us consider the elliptic crack as the limit of an
ellipsoidal cavity, as the semiaxis a3 = h of the ellipsoid tends to zero. Accord-
ing to (4.11.3), the vector b(x) is determined by the limiting value of the quantity
hz(,,)
c+(x, h) = J
-hz(,,)
c+(x + ~n) d~, (4.12.2)

7This result was first obtained (in a very complicated way) in [4.4].
116 4. Local Defects

where X(Xlo X2) E D, n = const is the normal to D, eis a coordinate along the
normal and

(4.12.3)

From the polynomial conservation theorem, it follows that e+(x + en)

is a polynomial of a degree not higher than m with respect to the coordinates
Xl; X2, e
and, consequently, can be represented in the form

(4.12.4)

where e?i)(x) is a polynomial of a degree not higher than m - k with respect

to the coordinates Xlo X2' After this decomposition is substituted into (4.12.2),
e
the integrals of odd powers of vanish, and the integrals of terms with e2k
give expressions of the form

Q2k(Xlo X2)Z(Xlo X2) ,

where Q2k is a polynomial of a degree not higher than 2k. Hence it follows
that

(4.12.5)

where Pm is a polynomial of degree :5: m. Finally, setting h = 0, we obtain

the required result.
Let us now consider the method of computing the polynomial B(Xlo X2)'
The vector b(x) is a solution of (4.11.9), where Ob(X) is a polynomial of degree
m, given on D. In the considered case, it is convenient to write down the equa-
tion for b(x) in the equivalent form (Appendix A.3)

JT(x - x') b(x')dx' = n 'Ob(X), xED, (4.12.6)

where the integration is carried out over the whole plane, which contains D,
the vector b(x') is defined to be zero outside D, and T(x) is a generalized func-
tion, which is generated by the generalized function So(x) in three-space.
Substituting (4.12.1) for b(x) into (4.12.6) and taking into account (4.12.3),
we have

JT(x - x')z(x')B(x')dx' = n 'Ob(X), xED, (4.12.7)

where z(x) is continued by zero outside D.

According to the property of polynomial conservation, the operator T
transforms the polynomial B(x) (multiplied by z(x» into the polynomial
n 'Ob(x) of the same degree; in particular, a constant vector B is transformed
 4.12 Elliptic Crack 117

into a constant tensor ToR, which depends linearly on B. It is obvious that

the constant coefficient To is determined by the expression (the integration
is carried out over the plane)

To = JT(x - x') z(x') dx'

= J T(x)z(x)dx = 1
(2'Jt)Z J T(k)z(k) dk , (4.12.8)

where

(4.12.9)

sinlakl )
z(k h k z) = 2'Jta l aZ lakl- z ( lakl - cos lakl (4.12.10)

and lakl is the modulus of the vector (a1k h azkz) .

Conversely, a constant external field oil'tS = const being given, we have

(4.12.11)

For an arbitrary polynomial field, the solution can be constructed in the

following way. The operator T transforms the homogeneous polynomial
B(x) = xm of degree m (multiplied by z(x)) into a polynomial of the same
degree. We find the coefficients of the latter, taking values at x = 0 of the
successive derivatives in x from T(xm). Then, solving a system of linear equa-
tions, we obtain the coefficients of the polynomial Bm(x) for the external field,
which is proportional to x m , and consequently, for an arbitrary polynomial
n·O't(x).

Problem 4.12.1. Show that, if n· O'i;(x) is a linear homogeneous function (ffx;

(i = 1, 2), then Bix) = Ba;x;, where

(4.12.12)

p{3;j= - (2~)2 J ;k; [Ta{3(k)k;] z(k) dk. (4.12.13)

Problem 4.12.2. Verify that (4.12.1) for the vector b(x) has, in a neighborhood
of the point Xo E r,
the following asymptotics in the local coordinate system
Yh Yz (Sect. 4.11)

(4.12.14)
118 4. Local Defects

where

(4.12.15)
0)
a= (a1o az'
and B(xo) is the value of B(Xb xz) at the point Xo E r .

Thus, for a polynomial external field, the polynomial B(x) being considered
as a known, we have an explicit expression for {3(xo), and consequently, for the
tensor intensity coefficient

(4.12.16)

Let us consider an isotropic medium and a constant external field as an

example. An expression for s(O) is easily found from (4.11.17); therefore it
is sufficient to write down an expression for B(xo) = const. Taking into account
(4.12.8, 11), after simple calculations, we find

D~{J = 2ai{lp.az- ).I) d 0

a a{J
(do not sum!) (4.12.17)

where

d 1 d 1 d-l
1 = - C3 + ).I(
Cz 2
- C3)' z= C3 + ).I(
C1 2
- C3)' 3 - C3'

C
1=
- Cz + 3C3' Cz = C3 - E(a) - K(a)
aZ - 1 E(a)
C _
,3 _ a Z'

and E(a) and K(a) are the complete elliptic integrals of the first and second
kinds.

4.13 Interaction Between Ellipsoidal Inhomogeneities

The results of Sects. 4.9-12 enable one, in particular, to deduce an explicit
solution of the problem of interaction of inhomogeneities and cracks, which
have ellipsoidal or elliptic form. For simplicity, let us consider only the principal
asymptotics of perturbed fields. As is shown in Sect. 4.5, this means that, in
the expansion with respect to multipoles (4.5.12) for the operator P of the
interaction energy, we can restrict ourselves to the term Poo.
 4.13 Interaction Between Ellipsoidal Inhomogeneities 119

Omitting simple calculations, we present the final result for an ellipsoidal

inhomogeneity

(4.13.1)

Here v is the volume of the ellipsoid, C l is the tensor of perturbed elastic con-
stants, and A is a tensor which depends on parameters of the ellipsoid and on
the tensor of elastic constants Co of the external medium. In the general case,
the latter is given by the (4.9.21, 23), and for an isotropic medium this is given
by (4.9.39). The second equality in (4.13.1) is valid if CII exists.
For an ellipsoidal cavity, Cl = - Co and

Poo = VCoD-lC O = v(Bo - A)-I, (4.13.2)

where Bo = COl is the tensor of elastic compliance of the medium, and D is

given by (4.9.21, 23 and 40).
Using (4.5.7), we can now find an explicit expression for the asymptotics
of the perturbed field. In the case when C l is an isotropic tensor, this expression
coincides with the one obtained by ESHELBY [4.5-7]. One can find a correspond-
ence between the ellipsoidal inhomogeneity and the equivalent point defect
(4.3.1) in the Oebye's quasicontinuum with the characteristic

(4.13.3)

where gO is defined by (4.3.4). A perturbed field caused by this defect coincides

with the asymptotics of the perturbed field of the ellipsoidal inhomogeneity. In
order to show this, it is sufficient to compare (4.3.2) and (4.5.2), where P(y, y')
in the last expression is to be substituted by the dominant term of its expansion
(4.5.12). Note that, in the case of a spherical inhomogeneity, A = vgO and,
consequently, C\ = C l .
Let us consider two ellipsoidal inhomogeneities and assume

(4.13.4)

to be given. According to (4.5.11), the interaction of ellipsoidal inhomogeneities

is described by the matrix Pii(Xl - X2), which, in the approximation under
consideration, has the form

.. (
PtiJ =
pao paoVVGO(Xl - X2)P&)
(4.13.5)
P50VVGO(X2 - Xl) pao P& '

where Xl> X2 are the position vectors of the centers of the ellipsoids. As is shown
in Sect. 4.5, the central symmetry of the inclusions enables us to ignore the
matrices PM u P{6.
Substituting PM in (4.5.13), we find an expansion of the interaction energy
<Pint in powers Of'12 = IXI - x21
120 4. Local Defects

(4.13.6)

where

(4.13.7)

(4.13.8)

It is obvious, that here </Jb is the self-energy of interaction of i-th defect with the
external field co, and </J 1 is a quadratic function of the external field, which
depends on Co and C .. (i = 1, 2), these depending on the defects and the unit
vector of the straight line which connects the defects.
We present below the explicit expressions for </J6 and </J 1 in the case, when
the external field is pure dilatation c~{l = cOoa{l and the inhomogeneities are two
spheroids with a common rotation axis x 2 and with isotropic elastic constants

2
"=
</Jo - ~
2Ko6...
(3P"1 + Pi)
2,
(4.13.9)

where

Pi = (5 - 4vo) I~ - 3I{2 + 8Te(1 - Vo) fJ-o , (4.13.10)

f.1-;

P~ = 2(1 - 2vo) (If - ID ,

6. i = [4(1 - vo) I~· - 2If2 - 8Te(1 - vo) fJ-o . 1 ]

fJ-i (1 + Vi)
x [(3 - 4vo) If - I~l + 877:(1 - vo) ~.
fJ-..
g-+ Vi~] Vi

- 2[ - 12 + h2 - 8Te(1 - vo) ~. Vi ]2 .
fJ-i 1 + Vi
The quantities I p , Ipq contained in these expressions are given by (9.42), Ko
is expressed by (4.9.41), the index i is the number of a defect, and the lower
index of the quantities P:
is connected with the principal axes of the ellipsoids,
these taking the values 1, 2 only (due to the rotation symmetry). A limiting
case of a rigid inclusion is obtained as fJ-i ~ 00; the case of a cavity is obtained
as, fJ-i ~ - fJ-o, Vi ~ Vo·
If one of the ellipsoids, for example the second one, is a sphere, then P~ =
o and the expression for </J1 is greatly simplified
 4.14 Notes 121

(4.13.11)

Finally, if both ellipsoids are spheres, then P} = P~ = 0 and cj)1 = O. This

is in agreement with the result of [4.8], where it is shown that the interaction
energy of two isotropic spherical inclusions is of the order of rrl.
In conclusions, a few words about the model of point defects in a classical
elastic continuum. We have already mentioned earlier that the a-functional
model of an inhomogeneity is, strictly speaking, correct only for the quasi-
continuum. This is evident, for example, from (4.13.3), where gO -+ 00 during
limiting transition to the continuum. A point defect in a continuum has meaning
only as a limiting image which corresponds to the asymptotics of the field of a
finite inhomogeneity, and is completely characterized by the tensor P oo . In the
particular case of an ellipsoidal inhomogeneity, P oo is given by (4.13.1).
The matrix P66 describes the interaction of two point defects. It is essential
that if one extrapolates this interaction to small distances, then the interaction
energy appears to have a singularity at distances of the order of V1l3, i.e. of the
order of the dimension of the starting (finite) defect. This result can be obtained
by substituting (4.13.5) into (4.5.13) for the interaction energy. Qualitatively,
the existance of a characteristic scale parameter follows from dimensional
reasonings.
Thus, the model of a point defect in an elastic continuum possesses an inner
restriction at small distances. Later, these considerations will be taken into
account, when considering a random field of point defects in Chap. 7.

4.14 Notes

The content of Sects, 4.1-4.5 is based on [B6.l4, 20, 23]. As regards point
defects, see [B6.27, 39]. The Green's tensor for displacement in an anisotropic
medium was investigated in [B1.2] The Green's tensors for strain and stress
were introduced in [B6.l7, 24]. Elasticity in terms of projection operators was
formulated in [B6.25, 26]. Discontinuities of stress and strain on the boundary
of two media were obtained in [B6.24].
The polynomial conservation theorem for an ellipsoidal inclusion was
proved for an isotropic medium and homogeneous field in [B6.5-7] and for
an anisotropic medium and arbitrary polynomial field in [B6.22]. The com-
plete investigation of stress on an ellipsoidal inhomogeneity and its limiting
forms was carried out in [B6.2l, 24], see also [B6.l2, 31, 37, 42].
A survey on cracks can be found, for example, in [B6.8, 29]. As regards
integral equations for cracks, see [B6.43, 7.13, 8.6].
5. Internal Stress and Point Defects

Vacancies, interstitial atoms and dislocations are well known examples of

crystal lattice defects which are sources of internal stress. In this chapter we
give a general outline of the theory of internal stress in nonlocal elasticity.
Point defects, which are sources of internal stress as well as inhomogeneities,
are considered in detail.

5.1 Internal Stress in the Nonlocal Theory

In all the preceeding discussions, it was the displacement vector or the field of
the symmetric strain tensor of rank two, which is uniquely determined by
the former, that described a state of a medium of simple structure. Conversely,
if the strain field is given, satisfying the usual compatability conditions, then
the displacement field is determined to within a rigid translation and a rotation
of the medium. The deformation is connected with the stress tensor by Hooke's
law (local or nonlocal), the stress tensor being connected by means of the
equations of equilibrium, with the external body or boundary forces. In the
absence of external forces, both the strain and the stress are equal to zero.
According to such a scheme the classical as well as the nonlocal theory of
elasticity are constructed.
However, another situation is possible, occurring in a number of important
cases. In order to give an illustration, let us carry out a thought experiment.
Let us consider a medium which consists of particles connected by an
arbitrary system of elastic bonds of the type of springs. Let us also assume
that, in the initial state, the elastic energy of all the springs is equal to zero
and consequently forces of interaction between particles of the medium also
vanish. Let us take some of the springs, change their length (e.g. they are short-
ened) and return the springs to their original places. This, generally speaking,
will result in a deformation of the medium, and the latter will proceed to a
new state which corresponds to the minimum of the elastic energy. In this
state of equilibrium, as distinct from the initial state, the elastic energy of
bonds is no longer equal to zero. A convenient quantitative measure of inter-
action between particles of the medium is the stress tensor, which is not con-
nected at all with the external forces. To emphasize this fact, let us call this
tensor an internal stress tensor, while the usual stress, caused by external
forces, will be called the external stress.
 5.1 Internal Stress in the Nonlocal Theory 123

It is easily seen that the displacement of points of the medium cannot

adequately describe the internal stressed state of the medium. In fact, the
bonds could be changed in such a way that the internal forces would be self-
compensated and the particles would not be subject to any displacements.
This shows that it is necessary to introduce a new kinematic variable which
determines the state of the medium.
Consider first a general case of a stressed state, caused by the action of
the external forces as well as by sources of internal stress. The role of the
latter in the crystal can be played by point defects of the type of foreign atoms
and of vacancies, as well as by dislocations, grain boundaries and so forth.
In macroscopic bodies internal stress can be caused, for example, by a hetero-
geneous temperature field or by purely mechanical causes of the type of
prestress or prestrain. Hereafter, we confine ourselves to the static case and
therefore we shall not introduce explicitly the internal degrees of freedom.
According to the results of the Chap. 3, one can assume that they are eliminated
from the equations and make an implicit contribution to the effective operators
of elastic moduli.
Let us assume that the field variable, which uniquely determines the state
of the medium, is a nonsymmetric tensor of rank two, of the elastic distortion
Xla' and that the elastic energy (/) in the harmonic approximation is a quadratic
functional of the distortion:

(5.1.1)

Hence, it follows that the operator (/)la/lp is self-adjoint, i.e. its kernel satisfies
the following conditions

(/)la/lP(x, x') = (/)/lpla(x', x) (5.1.2)

or in the k-representation

(/)Aa/lP(k, k') = (/)/lPAa(k', k) . (5.1.3)

It is natural also to consider the form (5.1.1) as positive definite.

The distortion Xla can be represented in the form

(5.1.4)

where ,,;,. and 'la are the external and internal distortions, respectively.
The external distortion ,,;,. is defined as a gradient of the displacement Ua ,
caused by external forces, i.e.

(5.1.5)

Thus, we have
124 5. Internal Stress and Point Defects

(5.1.6)

This condition is a characteristic one for the external distortion.

It is obvious that the external distortion is invariant with respect to trans-
lation, and under rotation, given by a constant anti-symmetric tensor aJ.a' it
is transformed according to the law

(5.1.7)

Let us postulate that these transformational properties are possessed also by

the total distortion XJ.a and consequently by the internal distortion ~J.a-
The requirement of invariance of the elastic energy with respect to rotation
yields

(5.1.8)

Hence, taking into account the arbitrariness of aJ.a and Xl'fl' we obtain a condi-
tion for C/JJ.al'fl [cf. 2.4.18)]

(5.1.9)

The energy operator C/JJ. al'f3 determines a scalar product in the functional
space of distortions, transforming the latter into a Hilbert space. This Hilbert
space can be decomposed into two orthogonal (with respect to energy) sub-
spaces of external and internal distortions, and the total energy can be re-
presented in the form of a sum of the external and internal energies.
We know from the results, obtained in the Chap. 2. that the most general
expression for the external elastic energy can be represented in the form

(5.1.10)

Here,

(5.1.11)

is the symmetric strain tensor and the Hermitian operator of the elastic moduli
cJ.al'fl is symmetric with respect to the indices Aa and pl3-
The external stress tensor (j' J.a is connected with the external strain c;a by
the nonlocal Hooke's law:

(5.1.12)

and it satisfies the conditions of equilibrium

(5.1.13)
 5.1 Internal Stress in the Nonlocal Theory 125

The system of equations (5.1.11-13) (under the given conditions at infinity)

uniquely determines the displacement and consequently the external distortion,
strain and stress. In other words, there exists a Green's operator GrxfJ , in terms
of which the displacement is given by the relation

uix) = JGrxl'(x, x')qfJ(x') dx' . (5.1.14)

We require, by definition, the fulfillment of the condition of orthogonality

of external and internal distortions with respect to the energy, i.e.

(5.1.15)

this is equivalent to the representation of the total energy in the form of a

sum of external and internal energies. The corresponding elastic fields, in this
situation, do not interact and, hence, in the theory of internal stresses no ex-
ternal forces are needed. Let us emphasize that all this is valid only in the case
of statics and, in particular, is not applicable to the case of moving dislocations.!
For the internal energy we now have

(5.1.16)

The quantity

(i.lrx(x) def Jf/J.lrxl'fJ(X, x') ~l'fJ(X') dx' (5.1.17)

will be called the tensor of internal stress. From (5.1.15), taking into account
(5.1.5) we find that (i.lrx satisfies the equation of equilibrium

(5.1.18)

Problem 5.1.1. Show that (5.1.15 and 18) are equivalent.

Problem 5.1.2. Show that if the internal distortion ~.lrx satisfies the condition
(5.1.6), then
~.la = O. [Hint: Make use of the positive definiteness of the form
(5.1.16)].

The right-hand side of the equation

(5.1.19)

characterizes the density of sources of internal stress. If the physical source

of internal stress are dislocations, then the tensor a.:'ix) is called the density

lIn the case of a moving dislocation, the external forces perform work.
126 5. Internal Stress and Point Defects

of dislocations and, in the general case, the density of quasi-dislocations ac-

cording to the terminology introduced by Kroner [5.1]. In what follows,
unless otherwise stated, it is assumed that the sources of internal stress are
localized in a bounded region.
Eq. (5.1.17 to 19) constitute a complete system which determines the internal
distortion ~.l.a and the internal stress O'.l.a for a given dislocation density a~v.
The existence of a solution is proved in the usual manner and the uniqueness
follows from Problem 5.1.2.
Let us consider the structure of the operator (/J.l.app, assuming that the
action-at-a-distance is of the order of the scale parameter I. Let us represent
(/J.l.app in the form

(/J.l.apfJ = c.l.apfJ + a.l.apfJ , (5.1.20)

where c.l.app is the above-defined operator of elastic moduli which is symmetric
with respect to the indices Aa and p.fJ.
It is easily seen that the Hermitian operator a.l.app is to be localized in a region,
where the sources of internal stress are present (more precisely, in an I-neigh-
borhood of this region). In fact, outside the region, occupied by the sources
of internal stress, the equations for the internal and external distortions must
be identical, since there ~.l.a is locally representable in the form of the gradient
of a displacement. This requirement leads to the condition
O.l.O~a.l.apfJ(x, x') = 0 (5.1.21)
imposed on the kernel of the operator. Moreover, from the equations (5.1.9,
20), with the symmetry of c.l.apfJ taken into account, it follows that
a[,lalpfJ (0, k') = 0. (5.1.22)

Problem 5.1.3. Show that the expansion of an entire analytic function

in a series with respect to k, k', begins with terms of the type
a.l.app(k, k')

(5.1.23)
where atfP{Jv is a constant tensor, which is anti-symmetric with respect to the
indices A)) and symmetric with respect to the indices p.fJ. In the presence of
a central symmetry and in an isotropic medium, the expansion of a.l.apfJ(k, k')
begins with terms of the second order in k, k'.

If I -+ 0, i.e. if a transition to a model of the ordinary elastic continuum

is performed, then a.l.app -+ 0, and the operator (/J.l.app coincides with the usual
tensor of the elastic constants c8app . In this case, the tensor of internal stress
O'}.a is symmetric. In the general case of a nonlocal model, a.l.apfJ :F 0 and the
tensor of internal stress, generally speaking, is not symmetric, but only in
the region, where the sources of internal stress are present.
 5.2 Geometry of a Medium with Sources of Internal Stress 127

Here the following question can arise: why the tensor of internal stress
cannot be symmetrized with the help of transformations, analogous to those
carried out in Chap. 2. Recall, that there, besides conditions of invariance
with respect to translation and rotation, an essential use was made of the
representability of the external distortion in the form (5.l.5), which is ensured
by the condition (5.l.6). However this condition is obviously invalid in the
region, where the sources of internal stress are present.
It is assumed hereafter for the sake of simplicity that f/)AapfJ = cAapfJ. The
case aAapfJ =f. 0 can be considered in an analogous way but leads to more com-
plicated formulae.
In view of our assumption, (5.l.16) for the internal energy can now be
written in the form

(5.l.24)

and (5.1.17) takes the form of the nonlocal Hooke's law:

aAa(x) = f cAapfJ(x, x') efJiX') dx' , (5.l.25)

which connects the symmetric tensor of internal stress with the symmetric
tensor of internal strain.
From the positive definiteness of the form (5. l. 24) , it follows that the
operator cAapfJ is non degenerate, i.e. there exists an inverse operator bpfJAa ,
which will be called the operator of elastic compliance. This enables us to
invert Hooke's law and write it down in the form

(5.l.26)

In a short operator form

a = C e, e = Ba. (5.l.27)

Correspondingly, instead of (5.l.24) we have

f/)= J (ale)= J (alBia) = J (eICle). (5.l.28)

5.2 Geometry of a Medium with Sources of Internal Stress

In this section we briefly consider the geometrical meaning of the introduced

measure of internal strain. A more complete geometrical theory of the internal
stress can be found, for example, in [5.2-7].
128 5. Internal Stress and Point Defects

Due to the assumption, concerning the equality of the operators (/J),ap.~ and
c),ap.f3, the role of the main kinematic variable, which determines the internal
energy (5.1.24), is now played not by the internal distortion (),a but by the
internal strain C),a' This enables us to simplify the geometric picture and to
confine ourselves to purely metric considerations.
Two metrics can be connected with the elastic medium. The external one
gives a distance ds' between points of the medium x a and x a + dx a in a de-
formed state, caused by the external forces. The corresponding expression
has the form

dS'2 = g~ix)dxa dx~ , (5.2.1 )

where g~{3(x) = gpa(x) is the external metric tensor, which determines the
external metric in the given state. Since we are working in the harmonic ap-
proximation, i.e. we are considering small deformations and small displace-
ments, the difference between Lagrangian and Eulerian systems of coordinates
is not essential. Therefore, the tensor g~f3(x) can be considered as a field quantity,
which is related to a fixed coordinate system x a in space, in which the deformed
medium is embedded. Without loss of generality, it can be assumed that the
coordinate system x a is cartesian, with the constant metirc tensor Oa~'
Let in the initial undeformed state the external metric tensor coincide
with oa{3' Then, as a measure of deformation we can take the difference

(5.2.2)

called the tensor of external strain. But the deformed state can be also described
in terms of the field of displacement Ua(x) , with which the tensor c~{3(x) is con-
nected by the relation (5.1.11). From the linear theory of elasticity it is well
known that, for the representability of c~ix) in the form of (5.1.11), it is neces-
sary and sufficient that the deformation c~{3(x) satisfies the Saint-Venant condi-
tions of compatibility. The latter can be written in the form

(5.2.3)

or briefly

Rot C = 0, (5.2.4)

where the operator Rot is defined by the expression 2

2The operator Rot is sometimes denoted also by the symbol Ink. Attention should be
paid to the correct order of indices: when (5.2.5) is substituted in (5.2.3), the operators rot
are contracted with different indices.
 5.2 Geometry of a Medium with Sources of Internal Stress 129

The conditi(ms of compatability and the operator Rot admit an interesting

geometrical interpretation. The external metric of the medium is induced
by the metric of the Euclidian space in which the medium is embedded before
and after the deformation. Therefore, both in the initial and in the final de-
formed state, the metric tensor ga/l(x) must be Euclidean. But this means that
the corresponding Riemann-Christoffel curvature tensor vanishes. For small
changes of the metric, the curvature tensor is linearized, and the condition
that it vanishes takes the form

Rot g' = 0, (5.2.6)

and this determines the geometrical meaning of the operator Rot.

Problem 5.2.1. Verify that the operator Rot satisfies the identities

div Rot = 0, (5.2.7)

Rot grad = 0, Rot def = °, (5.2.8)

where the operators grad and def act on vector fields (recall that the operator
def is the symmetrized gradient).
Problem 5.2.2. Show that the operator Rot commutes with the operators of
symmeterization and alternation, and hence, conserves the symmetry of a
tensor of the second order.
Problem 5.2.3. Verify that analogously to the well-known identity for the
Laplacian

t::.. = grad div - rot rot (5.2.9)

there exists the following identity for the bi-Laplacian:

t::..2 = def(2t::.. - grad div) div + Rot Rot. (5.2.10)

Other properties of the operator Rot and its representations in curvilinear

coordinate systems can be found, for example, in [5.8, 10].
The internal metric of the medium

(5.2.11)

as distinct from the external one, is invariant with respect to embeddings of

the medium in the space and is connected with the defects of structure, which
130 5. Internal Stress and Point Defects

were discussed above. For the phenomenological definition of the internal

metric, let us cut out an element of the medium which contains a point x and
is small enough for the internal stress in it to be eliminated. Then distances
between points of the element in this stress-free state are identified with the
internal metric at the point x. For the nonlocal theory, this definition, strictly
speaking, is valid only in the cases when internal stresses do not change ap-
preciably over distances of the order of the scale parameter I.
Let, in the state without defects of structure, the internal metric tensor
coincide with oa{3' As a measure of internal deformation, we take the difference

(5.2.12)

i.e. the internal strain tensor.

It is easily seen that the internal metric can be Euclidean only in the absence
of sources of internal stress, i.e. when ga{3(x) = oa{3' In fact, otherwise the
internal strain ca{3(x) could be compensated by an external strain c:lx) = ca{3(x),
i.e. the internal stress could be removed (globally!) with the help of external
forces. But, from the results obtained in the preceeding section, we know that
this is impossible. Thus, the quantity

(5.2.13)

differs from zero only in the region, where there are sources of internal stress,
and it characterizes their density. This quantity is called an incompatibility.
From the properties of the operator Rot, it follows that 7]"p(x) is a symmetric
tensor, the divergence of which is equal to zero

(5.2.14)

Taking into account (5.1.18) and (5.1.27), let us write in abreviated notations,
the complete system of equations for the internal stress

div 0" = 0, Rot BO" = 7] (div 7] = 0) . (5.2.15)

Incompatibility 7] and natural conditions at the infinity being given, the

existence and uniqueness of the solution follow from the above.

Problem 5.2.4. Show that the incompatibility and the dislocation (or quasi-
dislocation) density are connected by the relation

(5.2.16)

We see that a given dislocation density determines uniquely the incom-

patibility and hence the internal stress. However, the converse it not valid. If
 5.3 Green's Tensors for Internal Stress 131

the stress (J is given then r; is defined uniquely but a is defined with some
arbitrariness.

Problem 5.2.5. Verify that a given internal strain e determines a to within

rot (def u + (0), where u is an arbitrary rotational vector, and (0 is an arbi-
trary rotational anti symmetric tensor (the latter is representable in the form
(OP> = ePI'>0l'lfJ, where lfJ is an arbitrary scalar).

While considering specific problems of the theory of dislocations, it will be

shown that this indeterminacy is not essential and is fully compensated by the
geometric clarity of the quantity a as a characteristic of dislocations.
For a number of problems it is convenient to introduce one more (the third)
characteristic of the density of sources of stress, namely the density of moments
of dislocations (or quasi-dislocations) mP>, which determines uniquely the
density of dislocations

aW(x) = roV'p mp>(x) . (5.2.17)

The three characteristics are connected by

(5.2.18)

Thus, from the viewpoint of internal stress, the density of moments of dis-
locations is defined with even mOl e indeterminancy than the density of dis-
locations itself. In particular, the anti-symmetric component mCpl'l does not
affect r;AI' and, hence, (Ja{3 (but not al'». We consider this situation in more detail
in Section 6.1, with the example of a dislocation loop.

5.3 Green's Tensors for Internal Stress

Let us represent the solution of the system (5.2.15) in two forms

(Ja{3(x) = J za{3AI'(x, x') r;AI'(X') dx' = JGa{3AI'(x, x') mAI'(x') dx' (5.3.1)

or briefly,

(J = Zr; = Gm. (5.3.2)

The tensor operators Z and G obviously, play here the role of the Green's oper-
ators for internal stress. From (5.2.18) it follows that they are connected by the
relation

G = ZRot. (5.3.3)
132 5. Internal Stress and Point Defects

Substituting (5.3.2) in (5.2.15), we find that the Green's operator G must

satisfy the following equations

div G = 0, Rot BG = Rot. (5.3.4)

For G to be uniquely determined, let us require additionally that it be Hermi-

tian: G = G+. This is equivalent to the condition that its kernel GafJAf'(x, x')
possesses the symmetry of the kernel CafJAf'(x, x') of the operator of elastic
moduli. It is also necessary to require the fufillment of natural conditions at
infinity. 3
Taking into account the first equation of the system (5.3.4) and the con-
dition of Hermiticity, let us represent the operator G in the form

G = Rot HRot, (5.3.5)

where the Hermitian operator H is determined not uniquely. Later we shall use
this indeterminacy in order to obtain the simplest expression for H.
The internal elastic energy, which is defined by (5.1.24 or 28), can be now
represented in the forms

f/J = J <ml Glm) J <alrotHrotla) J <1JIHI1J),

= = (5.3.6)

the convenience of a choice of this or that being determined by a specific nature

of sources of internal stress.
Note that (5.3.6) could be used as the basis for a definition of the operators
G and H, if we consider the conditions which must be satisfied by the sources of
internal stress as being given. Now, our concrete problem is an explicit con-
struction of the Green's tensor G(r) for the homogeneous isotropic medium
with the kernel of the operator of elastic moduli (3.8.4), which, in the brief
notation is written in the form 4

c(k) = 2p(k)El + )'(k)E2 . (5.3.7)

Here El and E2 are the constant fourth order tensors:

(5.3.8)

The kernel b(k) of the operator of the elastic compliance is obtained by

inverting the tensor c(k). In the Appendix A4, a table for multiplication of the
fourth order tensors is given, from which it can be seen that E plays the role

aIn the local limit, G coincides with the operator S of Chap. 4. Earlier, the symbol G
denoted the Green's operator for displacements, but the latter is not used in Chap. 5 and
6, which excludes a possibility of confusion.
4Here, as in Sect. 2.8, k and r denote the moduli of the vectors k and r.
 5.3 Green's Tensors for Internal Stress 133

of a unit in this tensor algebra. It can be easily proved that

1 A(k) (5.3.9)
b(k) = 2fJ-(k) E1 - 2fJ-(k) [3A(k) + 2fJ-(k)] E2 .

Let us write the second equation of the system (5.3.4) in the k-representation

(5.3.10)

where

(5.3.11)

is the Fourier transform of the symmetrized operator Rot.

The solution of (5.3.10) must satisfy the first equation of the system (5.3.4),
i.e.

(5.3.12)

and have the required singularity, as k --* 0, which ensures the fulfillment of the
conditions for G(r) at infinity. Omitting calculations, we present the final result

G(k) = R(k)H(k)R(k) , (5.3.13)

where

(5.3.14)

In order to pass to the r-representation, it is necessary to give explicitly the func-

tions fJ-(k) and A(k) or, equivalently wI(k) and wt(k) , i.e. the frequencies of the
longitudinal and transverse eigenvibrations, which are connected with fJ-(k)
and A(k) by (2.8.10,11).
In what follows we confine ourselves to the Debye model of quasi-continu-
um introduced in Sect. 2.9, i.e. we assume k ~ K = 1t/a, where K is Debye's
truncation radius and a is the scale parameter. Let us take for wJk) (i = I, t)
the simplest polynomial approximation, which, as distinct from the case con-
sidered in Sect. 2.9, is conveniently represented in the form

(5.3.15)

where Vi is the velocity of sound when k = 0, Ti is an arbitrary parameter con-

nected with the limiting value of the frequency WiCK),

(5.3.16)
134 5. Internal Stress and Point Defects

Hence - 00 < n < 2. If we assume additionally that the limiting frequencies

are less than the Debye ones, then -1 < n < 2. The value ri = 0.8 practically
corresponds to the Born-Karman model. It is also easy to verify that (5.3.15)
satisfies the conditions (2.9.7) for group velocity.
To simplify the formulae, let us assume that n = rt = r. This is equivalent
to the assumption of independence of the Poisson coefficient)) = ))0 of k. Then,
a comparison of (5.3.15) with (2.8.10, 11) yields

p(k) = popek) , (5.3.17)

where Po = p(O), and p(k) is a dimensionless function:

(5.3.18)

Eq. (5.3.14) assumes the form

(5.3.19)

Thus, the problem is reduced to the determination of the inverse Fourier trans-
form of the functionf(k) = k- 4p(k)under the condition k::;; K. Calculations,
analogous to those carried out in Sect. 2.9 yield

f(r) = p(b.)h(r) , (5.3.20)

where her) is given by (2.9.4) and

p(b.) = Po P(b.) ,
(5.3.21)

For the Debye model we set p(b.) = l.

The expression for H(r) has now the form

(5.3.22)

and taking into account (5.3.13), we obtain

(5.3.23)

This expression can also be represented in the form

G(r) = 1 2 Rot Rotp(b.)h(r) + 1 2 ))0 Rotp(b.)g(r) , (5.3.24)

- ))0 - ))0

where g(r) is given by (2.9.3).

 5.3 Green's Tensors for Internal Stress 135

It is to be emphasized that, in the Debye quasicontinuum, the Green's tensors

G and H are entire functions, and consequently, do not have singularities, as
r -+ O. This enables us to write them in the form of well converging series. In
particular, forf(r), the expansion

fer) = ~ f; (- l)n[2(2n + 1) (2n + 5) - r (2n - 1) (2n + 7)] (U)2n

3')t2,t n=O (4n 2 - 1)(2n + 3)(2n + I)!
(5.3.25)

is valid. Using the expansion for fer), it is easy to obtain the value of G(r) for
r = 0, which plays a significant role in the theory of point defects:

(5.3.26)

The dimensionless multiplier (3 depends on the function w(k) and is ulti-

mately determined by the coefficient of r 4 in the expansion offer). In the general
case

(5.3.27)

For the Debye model and for the approximation (5.18) we have, respec-
tively,

{3 = 1, (3 = 30 ~511 r . (5.3.28)

The asymptotics of the Green's tensor, for large r or, equivalently, as a -+ 0,

is of interest. It is not difficult to show that the essential contribution to the
asymptotics is made only by the first two terms of, the expansions of the type
(5.3.15) of (5.3.18). The terms of higher orders give rapidly decaying oscilla-
tions with periods of the order of a, and must be excluded from the asymptotic
expansion. The asymptotics of G is determined by the asymptotics of the
functionf(r). The latter has the form

(5.3.29)

The substitution into (5.3.23) yields

G(r) ~ ~~ fl (n) + p.o;sa2 h(n) , (5.3.30)

where flo f2 are dimensionless tensor functions of the unit vector na = xa/r
136 5. Internal Stress and Point Defects

and the Poisson coefficient ))0. For the Debye model, the second term in (5.3.30)
is absent and the asymptotics coincides with the zeroth approximation (with
a = 0). In the general case, the second term in (5.3.30) accounts for the devia-
tion of w(k) from the linear law and contains approximate information about
the presence of microstructure in the medium.
In conclusion, let us note that if the Green's tensor for an isotropic medium
is known, an explicit expression for the Green's tensor of a medium with weak
anisotropy can be constructed.
Let the operator of elastic compliance B have the form

B = Bo - BI . (5.3.31)

where Bo is the operator of the isotropic medium, BI is the operator of perturba-

tion. Then, taking into account (5.3.4), by standard methods of the perturbation
theory, we obtain the expression for the Green's tensor in the first approxima-
tion

(5.3.32)

5.4 Isolated Point Defect

The point defects like vacancies, interstitial atoms, substitutional impurity

atoms, etc, are naturally considered both as sources of internal stress and as
local inhomogeneities.
As a characteristic of a point defect as a source of internal stress, it is con-
venient to take a dimensionless symmetric tensor MJ.I' which is a moment of the
defect. It characterizes the relative volume or shear distortion of an elementary
cell, caused by the defect. In particular, in the case of a spherically symmetric
defect (a center of dilatation) MJ.I' = MoJ.1" where M is the ratio of the change
of volume of the defect 11, to the volume of an elementary cell v; more precisely
11 v / v = 3M. To a single defect at the point Xo there corresponds the density of
moments

(5.4.1)

The change of force constants in a neighborhood of the point defect, as dis-

tinct from Sect. 4.2, is now conveniently described with the help of the kernel
b(x, x') of the operator of elastic compliance B

b(x, x') = bo(x - x') - vbIo(x - xo)o(x' - xo), (5.4.2)

where bo(x - x') is the kernel of the operator Bo of the homogeneous medium,
bi is a constant tensor, which has the symmetry and dimension of the tensor of
elastic compliance.
 5.4 Isolated Point Defect 137

Let us denote by Go(x - x') the Green's tensor of the homogeneous medium,
which corresponds to bo(x - x') and by G(x, x') the Green's tensor of the
medium with the defect, relating to b(x, x'). Omitting calculations, which are
analogous to those, carried out in Sect. 4.2, let us write down the expression
for G(x, x') in terms of Go(x - x')

The constant tensor P 11m has the symmetry of the tensor of elastic constants
and is the inverse of the tensor

(5.4.4)

where Cl is the inverse of the tensor b1•

The internal stress caused by the defect, according to (5.3.1), has the form

aa,8(x) = vMA"Ga{JA,,(x, xo)

= vM A,,[Ga{JA,,(x - xo) - G3{JvP(x - xo)PvpK~G8dl'(O)] (5.4.5)
= vM~P3{JA"(X - xo),

where M)" is the effective moment of the defect, which depends on the dis-
persion law and the ratio of elastic constant of the defect and that of the medium.
Taking into account (5.4.4), after simple transformations one can obtain the
expression for M ;,,:

(5.4.6)

where El is defined by (5.3.8).

From (5.3.6), we find the self-energy of the defect

2(/) = v 2M a(JGa(JA,,(xo, XO)MA" = v 2M a(JGa(JA"(O)M)."

- v2Ma(JG3{JvP(O)PVpK~G8d"(O)M)." . (5.4.7)

Taking into account (5.4.4), this expression can be transformed into

(5.4.8)

where

(5.4.9)

Let us consider the interaction between the defect and another field of stress.
We denote by -r(x) the unperturbed stress, i.e. the stress which would exist in the
medium in the absence of the defect. If -r(x) is caused by external forces, then
the defect behaves only as a local inhomogeneity. This case was considered in
138 5. Internal Stress and Point Defects

Sect. 4.2. Therefore, let us assume that -rex) is an internal stress, caused by some
distant sources with the moment density m (x). The resultant stress field will now
consist of the defect field O'(x), the field -rex) and a perturbed field rex). For the
latter, taking into account (4.3), it is easy to obtain

(5.4.10)

where

(5.4.11)

Thus, the perturbed field coincides with the field of a point source of internal
stress with the moment M.
For the energy of interaction Wint of the field -rex) with the field of the defect,
we have

(5.4.12)

where M~{3 is an effective moment of the defect. Hence, the force, which acts
on the defect, is equal to

(5.4.13)

Let us proceed to the case of an isotropic medium. The kernel of the operator
Bo is conveniently represented in the form 5

(5.4.14)

Here, h, 12 is an orthonormalized basis for isotropic tensors

(5.4.15)

The operators Xo(Ll) and Po(Ll) are the analogs of the modulus of volume
compression Xo and the shear modulus PO' Assuming, as in Sect. 2.3, that the
Poisson coefficient ))0 is a constant, we have

PO(Ll) = Pop(Ll) ,
(5.4.16)
2Po(1 + )) 0) - (Ll) - (Ll)
Xo(Ll) = 3(1 _ 2))0) P = XoP ,

where P(Ll) is determined by the dispersion law. Recall (Sect. 2.3) that, for the

SAs above, in the case of an isotropic medium we introduce the notations rand k for the
position-vectors of points and r, k for their moduli.
 5.4 Isolated Point Defect 139

Debye model, P(6.) = 1 and, for the model with a polynomial approximation
of the dispersion curve, the operator p(6.) is given by (5.3.21).
Set now B = Bo - B I , where BI is the operator of perturbed elastic com-
pliance. For better physical clarity it is expedient to express the kernel of the
operator BI in terms of the kernel of the operator of perturbed elastic moduli
C I . Confining ourselves to the case of a spherically symmetric defect, we obtain

hI (r, r') = v( 3Xo(XoX+ lXl) II + 2 ( P I ) I z ) o(r - ro) o(r' - ro) ,

Po Po + PI
(5.4.17)

where Xl> PI is the perturbation of the elastic moduli in a neighborhood of

the defect.
From (5.4.4), (5.3.26) and (5.4.17), in view of the properties of the basis
(5.4.15), we obtain the expression for the tensor P

P = PIll + pzlz,
-1 _ 2po(1+ ))0) (2~(1 - 2))0) _ Xo + Xl )
PI - v(1 - 2))0) 3(1 - ))0) Xl'
(5.4.18)
P21 = 2po ( ~(7 - 5))0) _ Po + PI) .
v 15(1 - ))0) PI

The formulae (5.4.3), (5.3.24) and (5.4.18) give an explicit expression for
the Green's tensor G(r, r') of an isotropic medium with a point defect.
Let the defect be a centre of dilatation, i.e. MAl' = M AI" Then, for the a
stress caused by the defect, we find from (4.15) (for convenience of notation
we assume r 0 = 0)

qa/l(r) = v(1 + ))o)M' rotav roti(6.) Si Kr , (5.4.19)

')tz(1 - ))0) r

where the effective moment M', according to (5.4.6), is determined by the

expression

M' = (1 _ 2~(1 - 2))0) Xl )-IM (5.4.20)

3(1 - ))0) Xo + Xl .

Let us consider two limiting cases. When Kr ::> 1, we have

(5.4.21)

To the other limiting case there corresponds the value r = 0 and

(5.4.22)
140 5. Internal Stress and Point Defects

The expressions (5.4.21, 22) admit a simple interpretation. Suppose that

in a homogeneous isotropic medium with the moduli f10, Xo there is a spherical
cavity of volume v and let a sphere of volume v + !:J..v with different elastic
moduli f1' = f10 + f1b X' = Xo + Xl be inserted into this cavity. Then the
stresses inside and outside the sphere exactly coincide with (5.4.21, 22), if
we set f3 = 1, which corresponds to the Debye model. Thus, (5.4.19) contains,
as a particular limiting case, a solution of the classical problem of an elastic
inclusion of the type of a center of dilatation [5.10].
In the case of rigid inclusion (Xl -+ 00) the relation between M' and M,
according to (5.4.20), is the following:

(5.4.23)

When Xl = 0, the moments M' and M coincide. Finally, the case Xl = - Xo

corresponds to a cavity for which the moment M and the stress a vanish.
In connection with this, there arises a question about an adequate model
of a vacancy. From physical reasonings, it is obvious that we should set f11 =
- f10, Xl = - Xo, i.e. a vacancy will behave as a cavity with respect to an ex-
ternal field. The components of the tensor P, according to (5.4.18), will have
in this case the form

3v(l - ))0) 15v(l - ))0)

(5.4.24)
PI = 4f3f1o(l + ))0)' pz = 2f3 f10(7 - 5))0)

However, at the same time, a vacancy is a source of internal stress, its asymp-
totics being of the form (5.4.21), and, as r -+ 0, the condition 0'(0) = 0 should
also be satisfied. The corresponding model in an elastic medium is a hollow
sphere, which is inserted with stress into a cavity.
It is easy to show that all the above conditions can be satisfied, if one
assumes that to the vacancy there corresponds the moment density

(5.4.25)

Then the stress caused by the vacancy takes the form (for the Debye model)

(5.4.26)

The asymptotics of O'(r) coincides with (5.4.21) and the coefficient a is

found from the condition 0'(0) = O. Calculation yields a = 5/3. It is natural
that the effective moment M' of the vacancy is no longer determined by (5.4.20),
it should be regarded as a known quantity.
Let us also present the expression for the self-energy (/J of the defect of
the type of a dilatation center. From (5.4.8), for the Debye model, we obtain
 5.5 System of Point Defects 141

'" = 2v(1 + ))0)/1oM2 (1

1
_ 2(1 - 2))0)
3(1)
Xl )-1 . (5.4.27)
IJ.I
- ))0 -))0 Xo + Xl
It can be shown that this expression coincides exactly with the energy of the
field of a sphere inserted with stress into the elastic medium.
For interaction of the field of stress z-(r) with a defect of the type of a center
of dilation, we obtain from (5.4.l2, 13)

r/)int = vM' Tr{z-(O)}, (5.4.28)

f = -vM' grad Tr{z-(O)}, (5.4.29)

where Tr{ z-} = z-~, and the moment M' is determined by (5.4.20).

5.5 System of Point Defects

Let us consider a system of point sources of internal stress with the moment
density

(5.5.1)

In the general case, with the sources of internal stress there is associated a
local change of elastic constants, and hence, for the kernel of the operator
B we should set

bafJ ).l'(x, x') = b~fJ).l'(x - x') - v ~ (5.5.2)

, b~fJ).I'0(x - Xi)O(X' - x,),
where the constant tensors bi have dimension and symmetry of the tensor of
elastic compliance.
For the Green's tensor G(x, x') of the medium with defects, it is possible
to obtain the expression, see (4.4.2),

GafJ).I"(x, x') = GfffJ).I'(X - x') - r: Gffj3<n(x -

'J
Xi)P~{""G~v).I'(Xj-x'). (5.5.3)

The matrix pjj with tensor components is the inverse of the matrix

(5.5.4)

where Cj = bil (note that Ci do not coincide with Ci in (4.4.1».

In the case of two defects, the components p j j are expressed in terms of
the components Rij according to (9.4.5).
The internal stress caused by a system of defects with the moment density
(5.4.l), according to (5.3.1), has the form
142 5. Internal Stress and Point Defects

aaf3(X) = V ~. M~paf3).I'(x, Xj) • (5.5.5)

•
In the presence of local inhomogeneities, this expression cannot be, of course,
obtained by superposing the fields of separate point sources of internal stress.
For the elastic energy of the system of defects from (5.3.6) we find

(5.5.6)

The above expressions give a solution of the problem of elastic fields,

self-energy, energy of interaction and consequently of the forces of interaction
in a system of point defects.
Let us consider in more detail the case of two defects of the type of a centre
of dilatation in an isotropic medium. Making use of a special tensor basis
(Appendix A4) enables us, in this case, to write down the final results in an
explicit form. Since the general formulae are rather involved, let us confine
ourselves to the first terms of the expansion of the elastic energy in the inverse
powers of the distance p between the defects

(5.5.7)

The first term 1/)0, as it should be, is equal to the sum of the self-energies of
each defect, which are given by (5.4.27). For the coefficients 1/)1 and 1/)2, we
have

(5.5.8)

where

mj=
(5.5.9)
, fl.j , Xj
fl.j= Xj= --'--'-'--
fl.o + fl.j , Xo + Xj

The expression for the interaction force is obtained by differentiating (5.5.7)

with respect to p. At sufficiently large distances, defects of the same sign repel
and defects of different signs always attract each other. With decreasing distance
between the defects, the interaction forces can change sign, provided the limiting
value of the dispersion curve exceeds the Debye frequency. In this case, a
generation of stable pairs of defects of different signs is possible, i.e. the so-
called dipoles of dilatation.
 5.6 Notes 143

In the paper by Lifshitz and Tanatarov [5.11], the first term of the
expansion of the elastic energy of interaction for two identical centers of
dilatation in an ordinary isotropic elastic medium was calculated. Comparison
shows that it coincides with the term f/JIp-6 shown above for the case of identical
defects in the Debye model.

5.6 Notes

Internal stress in the local theory of elasticity is discussed, for example, in

[B6.5-7, 11, 15-18, 35]. Internal stress in the nonlocal theory of elasticity
was considered in [B3.21]. Point defects as sources of internal stress in Debye's
quasicontinuum were investigated by Brailsford [B6.2] and Kosilova, Kunin
and Sosnina [B6.14].
6. Dislocations

There is an extensive literature devoted to crystal lattice dislocations and to the

continuum theory of dislocations based on the local theory of elasticity. The
purpose of this chapter is to consider the theory of dislocations on an inter-
mediate level-in the scope of non-local elasticity. The main focus is on the
specific non-local effects which cannot be described in principle in the local
elasticity.

6.1 Elements of the Continuum Theory of Dislocations

We suppose that the reader is familiar with the basic physical notion for dis-
locations as elementary carriers of plastic deformation in crystals. At the pre-
sent, elements of the theory of dislocations are standard tools for mechanical
engineers and stress analysts. Descriptions of the theory of dislocations, which
are intended for different groups of readers, can be found, for example, in
[6.1,2]. In this section we present some facts from the continuum theory of dislo-
cations, which are necessary for what follows.
Recall that a dislocation in a crystal is a linear defect of the crystal lattice
of a special type. In Fig. 6.1 an edge dislocation is schematically shown, its line

B/

Fig. 6.1 Edge dislocation Fig. 6.2. Screw dislocation

coinciding with edge of an extra atomic semi-plane ABCD. The segment DC

belongs to the line of dislocation. In Fig. 6.2, a screw dislocation is schematically
 6.1 Elements of the Continuum Theory of Dislocations 145

shown, this being formed by the displacements of the surfaces of the cut ABCD
in the direction AB. Here, the segment DC belongs to the line of the dislocation.
The general case of a curvilinear dislocation can be considered as some sort of
combination of edge and screw components.
In Fig. 6.3, the edge dislocation L is surrounded by a closed contour r which
is called the Burgers circuit. Its image F' in a crystal without dislocation has a
discontinuity. The vector b required to close the contour, is called the Burgers
vector. For an edge dislocation the vector b is perpendicular to the dislocation
line L, while for a screw dislocation it is parallel to the latter.

12 11 10 8 8 7 1211 109 8 7
1.1 6 1.1 B

-c-" -
j..-- ~-
1"15 [j lq 5

18 .1
1718 19 1 2
I-- r
r'!-- f- -
15
15 b 0"
17 18 19 r- 1 Z
Fig. 6.3.
Burgers circuits

In the general case, the Burgers vector has an arbitrary angle with respect
to the dislocation line, this angle being able to change along the line for a
linear dislocation, but the Burgers vector remains constant. It is the existence
of a Burgers circuit with a non-zero constant vector b that is the definition of a
dislocation in a crystal. From the definition, it follows that the dislocation line
is always either closed or ends at the edge of the crystal (or goes to infinity
in the case of an unbounded crystal).
The plane, which passes through the Burgers vector and is tangent to the
line of dislocation, is called the slip plane. Dislocations move mostly along these
planes.
In the continuum theory of dislocations a real crystal is replaced by a con-
tinuous elastic medium and the dislocation is modelled by a singular line
L, which possesses the property that the integral

(6.1.1)

has a constant value when taken over any closed contour r enclosing L. Here
up. are displacements in the medium, and rand b are called the Burgers circuit
and the Burgers vector, respectively.
Besides an individual dislocation, there are also dislocations, which are
continuously distributed over a surface or a region. In such a case, the dis-
146 6. Dislocations

locations density a!'v(x) is defined as a total density of the Burgers-vector flux

at the point x. It is convenient to introduce a formalism, which would enable us
to consider in a unified manner arbitrary distributions of dislocations, in par-
ticular the singular ones [6.3].
Let L be a sufficiently smooth curve. The generalized function o(L) on the
space of indefinitely differentiable test functions g>(x) with bounded support,
is defined by the relation

So(L)g>(x) dx= Sg>(xL)dL. (6.1.2)

L

Analogously, for a surface S and region Vwe have O(S) and o(V) according to

So(S)g>(x) dx = Sg>(xs) dS ,
s
(6.1.3)
So(V)g>(x) dx = Sg>(xv) dV .
v

It is clear that o( V) coincides with the characteristic function of the region. The
above-introduced notation is due to the convenience in writing down the
following formulae.
Concerning L, S, and V it is assumed only that the corresponding integrals
in (6.1.2) and (6.1.3) exist.
The following representations are valid

o(L) = So(x - XL) dL, o(S) = So(x - xs) dS,

L S
(6.1.4)
o(V) = So(x - xv) dV ,
v

and they are to be understood in the sense that

Sdx g>(x) So(x - XL) dL = SdL So(x - xL)g>(x)dx = Sg>(XL) dL. (6.1.5)
L L L

For bounded L, S, and V

JO(L) dx = I, So(S) dx = s, JO(V) dx = v, (6.1.6)

where I, s and v are length, area and volume respectively.

Besides the scalar o-functions, we shall also consider vector ones, which
for oriented Land S are defined by
 6.1 Elements of the Continuum Theory of Dislocations 147

Jo(L)g>(x) dx = Jqy(XL) dL,

L
(6.1.7)
J0(8)qy(x) dx = Jqy(xs) d8 ,
s

where dL and d8 are the elements of the curve and the surface.
From Gauss' and Stockes' formulae, important relations for the vector 0-
functions follow.
Let S be the boundary of a region V. Then

0(8) = - grad o( V) , (6.1.8)

whence it follows that, for a closed surface S,

rot 0(8) = o. (6.1.9)

Let S be now a surface and L be its boundary; then

o(L) = rot 0(8) (6.1.10)

and consequently for a closed contour

div o(L) = o. (6.1.11)

In order to prove, for example, (6.1.8), let us write down the relations

(6.1.12)

where we have used the definitions of o-functions, the usual rules of operation
with derivatives of generalized functions and the analog of the Gauss theorem
(recall that we are working in an affine coordinate system). Then (6.1.12)
immediately implies (6.1.8).

Problem 6.1.1. Prove (6.1.10).

Let us now consider an oriented contour L and a scalar or tensor function

f(XL) defined on it. Then the o-function with weightf(xL) is defined by

(6.1.13)
148 6. Dislocations

or

(6.1.14)

Analogously, other a-functions with weight can also be defined. The deri-
vatives of a-functions with weight are defined, as usually, by a transposition of
the operation to the test function. In the cases when no confusion can arise,
we shall omit the brackets in the expressions for the a-functions with weight
and write, for example, rotl'.l.M.l.(xL)a(L), where the operator rot is applied to
the entire a-function with weight.
It can be easily seen that the following relations are valid:

(6.1.15)

where [.l. and n.l. are the unit tangent vector and the normal, respectively.
Note that formulae of the type (6.1.8) and (6.1.10) are applicable only to
a-functions with weight of the special form (6.1.15). More detailed information
on a-functions concentrated on lines and surfaces, may be found in [6.4, 5] .
Let us now proceed to the consideration of characteristics of a distribution
of dislocations. Let L be a closed (or going to infinity) line, which is the bound-
ary of the surface S, and let bv be a constant vector. The expression

(6.1.16)

due to (6.1.11), satisfies the condition

(6.1.17)

and, according to (5.1.19), can be considered as a density of a distribution of

dislocations. Let us draw an arbitrary surface F, which intersects the line L
at one point and let us write the expression for the flux o;l'V through this surface.
Taking into account the properties of a(V), we obtain

Jo;/.w(XF) dFI' = bv Ja(v) dFI' = ± bv , (6.1.18)

F F

where the signs + or - are taken depending on whether the orientations of

L and Fare in conformity or not.
Since this flux does not depend on the choice of the surface F which inter-
sects the line L and is equal to zero if F does not interest L, we come to the con-
clusion that (6.1.16) is the density, which corresponds to an individual disloca-
tion with the line L and Burgers vector bv• On the other hand, taking into
account (6.1.10) we have
 6.1 Elements of the Continuum Theory of Dislocations 149

A comparison with (5.2.17) shows that this dislocation density can be as-
sociated with the density of moments

(6.1.20)

where

(6.1.21)

is the surface density of moments.

Here, the degree of non-uniqueness of the moment density mp"(x) as a
characteristic of a distribution of dislocation, becomes evident. The surface S
in (6.1.20) can be replaced by any other surface S', which has the same line L
as a boundary. The expression (6.1.20) can be also interpreted as a distribution
of dislocations on the surface S with a constant Burgers vector b" or, equiv-
alently, with the two-dimensional density of moments (6.1.21). According to
(5.2.17), this distribution is equivalent to an individual dislocation with the
line L and with the same Burgers vector.
We see that there is a far-reaching analogy between a density of dislocations
and a density of currents (or vortices). In fact, if we ignore the additional index
)), then al'"(x) can be considered either as a linear current with the line L, or as
currents distributed with constant density over the surface S. Both distributions
of currents are equivalent, since the currents on the surface S compensate each
other, except for the boundary current. Generalizing this analogy, we note that
the static continuum theory of dislocations can be compared with the magne-
tostatics, and to the usual static elasticity there corresponds the electrostatics.
From a mathematical standpoint the only important distinction between these
theories is the higher tensor dimension of quantities in the former theories in
comparison with their analogs in the latter ones.
From the preceding, it becomes clear that a distribution of dislocations in a
region can be prescribed in three equivalent ways

mp"(x) = Mp"(xv)o(V) ,
al'"(x) = rot~p Mp"(xv)o(V) , (6.1.22)
7}AI'(X) = Rot~~12 Mp"(xv)o(V) ,
where Mp"(xv) is the volume density of moments of dislociations. If the dis-
locations are distributed over a surface S or a contour L, then o(V) is to be
replaced by o(S) or o(L), and Mp"(xv) by a surface moment density Mp"(xs)
or by a linear moment density Mp"(XL), respectively.
In the limiting case of an elementary dislocation loop, which can be obtained,
for example, by contracting the surface Sin (6.1.19)to a point Xo or, equivalent-
150 6. Dislocations

ly, by moving away far enough the point of observation x, the expression for the
density of moments takes the form

(6.1.23)

where Mp"(xo) = bps" is the moment of an elementary dislocation, sP is the

normal vector of the surface element s.
A comparison with (5.4. 1) shows than an elementary dislocation loop is
equivalent to a point source of internal stress.
Together with the results of the previous chapter, we have now all the
necessary formalism to compute fields and interactions of dislocations.

6.2 Some Three-Dimensional Problems

The computation of static fields and the interaction of dislocations within the
framework of the simplest model of an isotropic elastic continuum has been
dealt with in many papers, their results are given in the monographs and reviews
mentioned above. Computations, for specific crystals, carried out on com-
puters, showed that, in many cases, formulae of the continuum theory of dis-
locations remain valid even in the field where they seem to be inapplicable,
namely, for distances of the order of a few interatomic distances. However, at
the same time, it was also found that a number of effects essentially depend
on the discrete structure of the crystal, and for their theoretical description one
has to refute the continuum model.
An intermediate place between the crystal lattice and the elastic continuum
is occupied by the above investigated Debye's model of the quasi-continuum
with the space dispersion taken into account or not. As we know, this model
gives a qualitatively correct description of the discrete structure and it is not
too complicated for theoretical analysis. In Sections 6.4-6 within the framework
of this model, a few problems will be considered, in which typical effects of
microstructure appear. In particular, taking into account a change of force
constants in the core of the dislocation is one of such problems.
Let the operator of elastic compliance B for an isolated dislocation with
the line L, has the form

(6.2.1)

where Bo is the operator of a homogeneous medium, BL is the operator, which

describes the local change of elastic properties in the core of the dislocation.
The simplest expression for the kernel of the operator BL has the form
(s ~ a2)
 6.2 Some Three-Dimensional Problems 151

= so(X - X')[B'(XL)O(L)] , (6.2.2)

where B'(XL) is a tensor, which has dimension and symmetry of the tensor of
elastic compliance. It is obvious that B'(XL) must depend on the direction of
the Burgers vector and on the tangent vector to the dislocation line.
The construction of the Green's tensor G, which corresponds to (6.2.1),
the Green's tensor Go of the homogeneous medium being given, leads, in the
general case, to the solution of an integral equation on the dislocation line L.
If the perturbation of the force constants in the core' of the dislocation is small,
then the integral equation can be. effectively solved by the method of successive
approximations. In the first appoximation, we have

(6.2.3)

or, in a more detailed description,

(6.2.4)

Similarly, for the Green's tensor H, we obtain

H(x, x')

= Ho(x - x') + J
s Ho(x - xL)RotB'(XL)RotHo(XL - x') dL .
L
(6.2.5)

The expression for the energy qi of a dislocation with line L and a

Burgers vector b, taking into account (5.3.6) and (5.1.16), can be represented in
the form

(6.2.6)

where the tensor qiap(L) depends only on the line of the dislocation

qi(L) = JJrot rot H(XL' xl) dL dL' • (6.2.7)

LL

The stress produced by the dislocation, according to (5.3.1) and (6.1.20),

is given by

(6.2.8)
152 6. Dislocations

Let us now consider the interaction of the dislocation with a stress field
o-(x) , which is caused by sufficiently remote sources of stress. The total force
q, which acts on the dislocation, can be represented as the sum of two com-
ponents

q = qo + q', (6.2.9)

where qo depends linearly on the Burgers vector b of the dislocation and on the
external field (J, and q' is quadratic with respect to the external field and does not
depend on b. Omitting not difficult but cumbersome calculations, we present
the final result.
The force qo can be written in a form, which coincides with the known ex-
pression of Peach-Koeller [6.1]

(6.2.10)

but with effective stress "C1'"' In the case of a small change of the force constants
in the core of the dislocation,

"Ca,B(X) = (Ja,B(x) + s JGff,BJ.I'(X - XL) B'J.u"p(XL)(J"P(XL) dL . (6.2.11)

L

Under the same assumption,

(6.2.12)

In an analogous manner, one can consider systems of dislocations or systems

of dislocations and point defects.

6.3 Two-Dimensional Problems

Beginning with this section, we consider the straight line dislocations, which
are parallel to the z-axis. In this case the stress and the strain depend only on
the two-dimensional position-vector r(x, y).
Let us decompose the 3-dimensional stress tensor (Jjk (i, k = 1, 2, 3) into a
two-dimensional tensor (Ja,B, vector aa and scalar (J

(Ja,B 1 (Ja )
(Jik = ( ..........:.......... , a, f3 = 1, 2. (6.3.1)
(Ja i (J
 6.3 Two-Dimensional Problems 153

The equations of equilibrium imply the representability of ~afJ and ~a in

the form

(6.3.2)

where lfJ, ¢ are scalars, and the operators Rot and rot are defined by the ex-
pressions

(6.3.3)

where eafJ is the two-dimensional unit antisymmetric pseudotensor.

In an analogous manner, let us decompose the three-dimensional strain
tensor Cik into two-dimensional tensor cafJ" vector Ca and a scalar c.
The distribution of dislocations in the case under consideration is com-
pletely characterized by the density of Burgers vector b,{r). Let us represent
bi in the form of a two-dimensional vector ba and a scalar b = b 3• It is obvious
that the vector ba describes the distribution of edge dislocations and the scalar
b describes the distribution of screw dislocations.
Taking into account the relation (5.2.16) between the incompatability and
the dislocations density, we find that (5.2.13) takes now the form

(6.3.4)

From the last equation, taking into account the boundary conditions at infinity,
it follows that c = 0.
The operator form of the Hooke's law can be now written in the form

CafJ= BafJ)."~)." + 2B afJ ).(J). + BafJ(J,

Ca = BdfJ;' (J)." + 2B~;. (J). + Ba(J , (6.3.5)
c = Bf;, (J)." + 2Bt (J). + B(J .

The two-dimensional operators appearing here are connected with the

three-dimensional operator B ijkl in an obvious way, the cross denoting the
Hermitian conjugation.
In view of (6.3.2, 4 and 6.5), taking into account that C = 0, we find the
equations for lfJ and ¢

RotafJ !JBafJ)." RotA" lfJ + RotafJ fJO afJ). rot). ¢ = - rotab a ,

(6.3.6)
rota!JB;t;." Rot)." lfJ + rotafJO a). rot).¢ = - b.

Here

!JBa{J)." = Ba{J)." - BafJB-IBf;" !JBafJ ). = 2(BafJ). - BafJB-1Bt) ,

(6.3.7)
fJOa). = 4(B~;. - BaB-1Bt).
154 6. Dislocations

If the operator ($af>). is equal to zero, the system (6.3.6) is uncoupled. This
will take place in all cases, when, due to the symmetry conditions in the plane
x, y, there do not exist two-dimensional tensorial material characteristics of a
medium of an odd order. In particular, these conditions are satisfied in the
presence of central symmetry, cubic symmetry or in the isotropic case.
Let us represent the solution of the system (6.3.6) in the form

¢(r) = - SGn(r, r')rotabir') dr' + SGdr, r')b(r')dr' ,

(6.3.8)
cj;(r) = - S G21 (r, r')rotaba(r') dr' + SGzz(r, r')b(r')dr' ,

where G is the matrix Green's operator. Substitution in (6.3.2) yields an ex-

pression for the stress.
It can be shown that the elastic energy per unit of length of z-axis (/J is
presented in terms of the Green's operator in the form

(6.3.9)

Here, the two-dimenstional brackets are defined by

<uIGlv) = SSu(r)G(r, r')v(r') dr dr' = <vIG+lu) . (6.3.10)

When the system (6.3.6) is uncoupled, the Green's operator is diagonal and
hence, according to (6.3.9), in the case of a homogeneous medium, edge dis-
locations do not interact with the screw ones. If the change of force constants
in the dislocation core is taken into account, then the above degeneration is
removed, since the diagonal components of the Green's operator depend on
the changes of force constants of all dislocations'!
Let us now proceed to a separate consideration of screw and edge dis-
locations.

6.4 Screw Dislocations

Suppose there are only screw dislocations in the medium, i.e. ba = O. Let us
further assume that the conditions of symmetry are satisfied, ensuring the
uncoupling of the system (6.3.6). Then the elastic field of screw dislocations is
described by a scalar function cj;(r), which satisfies the equation

(6.4.1)

lIn connection with the divergence of the self-energy of a dislocation in a two-dimensional

problem, as r -+ 00, (6.3.9) determines the energy with logarithmic accuracy.
 6.4 Screw Dislocations 155

Let us introduce a Green's function G(r, r') defined by

Srota~aJ3(r, r") rotJ3G(r", r') dr" = - oCr - r') . (6.4.2)

In the case under consideration, only the components (Ta(r) of the stress differ
from zero, and for them, according to (6.3.2), we have

(Ta(r) = rota SG(r, r')b(r')dr' . (6.4.3)

The energy f/), as it follows from (6.3.9), has the form

2f/) = <bIGlb) . (6.4.4)

Thus, the problem of determining elastic fields and of investigating the inter-
action between screw dislocations, is reduced to constructing the Green's
function G. In this section we consider the case of a homogeneous isotropic
medium without taking into account the change in the force constants in the
core of the dislocation. The kernel of the operator ~aJ3 and the Green's function
G are then of the difference type.
For an isotropic medium, taking into account (5.3.9), we obtain

(6.4.5)

where p is an operator with the kernel p(k). (6.4.2) takes the form

p-l6.G(r) = - oCr) . (6.4.6)

From the above discussion it follows that G(r) is the inverse Fourier trans-
form of p(k)k- 2, where k = Ikl. It is not difficult to prove that, in the two-
dimensional case, the inverse Fourier transform of the function k- 2 (k :::;; K) is

g(r) = -.l.
2rc
(7 Jo(,") -
0 '"
1 d," + c), (6.4.7)

where J o(,") is the Bessel function of the zeroth order, r = Irl. The constant c
should be chosen from the condition that, at large distances (Kr ~ 1), the func-
tion g(r) is to transform into the classical Green's function of the Laplace
operator

1 r
g(r) ~ - 2rc ln R' (6.4.8)

where the quantity R ~ a has the meaning of a characteristic dimension of the

156 6. Dislocations

body. Taking into account the asymptotic behaviour of (6.4.7), we find that
c = In "R. The solution of (6.4.6) may now be represented in the form

G(r) = flg(r) . (6.4.9)

Iffor fl(k) the approximation (5.3.18) is taken, then

(6.4.10)

where

(6.4.11)

and J 1("r) is the Bessel function of the first order. When" ..... 00 (a ..... 0), the
function air), becomes the ordinary two-dimensional a-function.
For the Debye's model

G(r) = flog(r) . (6.4.12)

The asymptotics of G(r), omitting the rapidly oscillating and decaying

terms, has the form (independent of the model accepted)

G(r) ~ - i~ In ~ , (6.4.13)

i.e., in the case of screw dislocations the space dispersion makes no additional
contribution to the asymptotics of G(r).
For a single screw dislocation at the point r = 0 with the Burgers vector b,
taking into account (6.4.3) we have

lTir) = boyG(r), lTy(r) = - bo"G(r). (6.4.14)

In particular, for the Debye's model,

(6.4.15)

which coincides with the expressions obtained by Brailsford [6.6] When", » 1,

(6.4.15) becomes the well-known solution of the continuum theory of dis-
locations [6.1].
The self-energy of the dislocation is given by

(6.4.16)
 6.4 Screw Dislocations 157

wl;tere the constant G(O) for the model (5.3.18) and for the Debye's model is
equal to

G(O) = ;~ In ICR , (6.4.17)

respectively.
Consider now a system of screw dislocations with the Burgers vectors bj'
the dislocations being located at points r j' Then, for the stress we have

(6.4.18)

The elastic energy is equal to

(6.4.19)

Thus, in the case of two dislocations, we find for the energy of interaction

(6.4.20)

The force of interaction q is directed in this case along the vector r = rl -

r2 and for the model (5.3.18) and for the Debye's model is equal to

(6.4.21)

1'\
\ 1'=-7
~

~ f5
~ !/ /'
-7~'" .L
=--
5
~
o1 pia 7
/ J

-OJ!
\) Fig. 6.4. Interaction force q as
function of distance p
158 6. Dislocations

(6.4.22)

respectively.
The asymptotics of (6.4.21, 22) coincide with the expression for the force of
interaction, which is known in the continuum theory of dislocations.
Since /JO(Kp)/ :s; 1, it follows from (6.4.22) that, in the Debye's model, the
sign of the force cannot change with decreasing distance between the dislocations.
At the same time, an investigation of (6.4.21) shows that, when the space dis-
persion is taken into account, the sign of the force can change, provided the
boundary value W(K) of the dispersion curve for transverse oscillations is greater
than the corresponding Debye frequency. The graphs of q(p) for different values
of r are presented in the Fig. 6.4. The dotted curve corresponds to the Debye
model.

6.5 Influence of Change of the Force Constants in Cores of

Screw Dislocations

In view of large relative displacements of atoms in the core of a dislocation, the

force constants, generally speaking, differ from the force constants of a perfect
crystal.
In the isotropic case, for screw dislocations, (6.4.5) is still satisfied, but the
operator p-l is no longer a difference operator if the change in the elastic prop-
erties of the core are taken into account. Let us denote B = p-l and represent
the kernel of the operator B for a system of screw dislocations in the form

B(r, r') = Bo(r - r') - S 1:i Bio(r - ri)o(r' - r i ) , (6.5.1)

where Bo(r, r') is the kernel of the operator of the homogeneous medium, the
B/s are constants which have dimensions of POl. For convenience of inter-
pretation of the constants B i , let us consider single screw dislocation within the
framework of the Debye model. The kernel of the operator p in this case can
be represented in the form

(6.5.2)

The kernel of the inverse operator B has the form

(6.5.3)

where

(6.5.4)
 6.5 Influence of Change of the Force Constants in Cores of Screw Dislocations 159

An analogous representation for B is conveniently preserved also in the gen-

eral case of (6.5.1). From (6.5.4) it follows that to the values Bj > 0 there cor-
respond more rigid elastic bonds in the dislocation core and to the values Bj < 0
there corresponds weakening of the elastic bonds.
Assuming that parallel screw dislocations differ from each other only in the
sign of the Burgers vector b as in a crystal, it is natural to set Bj = B1. Eq. (6.4.1)
in the case under consideration takes the form

div B grad sb = - b. (6.5.5)

Let us denote by G (r - r') the Green's function for the homogeneous med-
ium, which was constructed in the preceeding section. Then, for the Green's
function G(r, r') for the operator (6.5.1), we can derive the expression

G(r, r') = Go(r - r') - r;., oaGo(r - rj)P/JPor,Go(rj - r'), (6.5.6)

where the matrix Pr,fl is inverse to the matrix

Rijp = oaopGo(rj - rj) + s-IB-l(J.)Jap • (6.5.7)

For stress and energy of a system of dislocations we have

(1a(r) = ~ bj rota G(r, rj) , (6.5.8)

j

(6.5.9)

Taking into account that oaGO(O) = 0, we find from (6.5.6, 8 and 9) that the
change of the force constants in the core of an isolated screw dislocation does
not contribute to the stress and the elastic self-energy.
Let us consider two screw dislocations with the Burgers vectors b l = b,
b2 = ± b. Assume also that BI = B2 = B. The elastic energy in this case has the
form

(6.5.10)

The determination of the Green's function G(r, r') is reduced to an inversion of

the four-dimensional matrix R. Computations yield (p = Irl - r2D

a[Go(p)]2
G(rb rl) = G(r2' rz) = Go(O) + a2 _ [G;;(p)]2 ,
(6.5.11)
_ [Go(p)]2G;;(p)
G(rb r2) - Go(p) + a2-[G;;(p)]2 '

where
160 6. Dislocations

a = s-lB-l + [lp GO(p)]p=o . (6.5.12)

The value Go(O) is given by (6.4.17), and the second term in (6.5.12), for the
Debye's model and the model (4.3.18), respectively, has the form

f - fJ.o,,2
[ 1 G'()]
poP p=o
_
-1- 81t'
(1 _2.8 T) .
fJ.o,,2
-9"1t
(6.5.13)

Let us write down the energy (/) in the form

(6.5.14)

where (/)0 is the self-energy of the dislocations, (/)int is the interaction energy.
For the latter, taking into account (6.5.10,11), we obtain

'" () _ b2 ( [G o(p)2]) (6.5.15)

YJint p - ± Go( P) + a =+ G~(p) .

The force of interaction is equal, by definition, to

d (6.5.16)
q(p) = - dp (/)int(P) .

Taking into account (6.4.13), for the asymptotics of q(p), we obtain

q(p) ~ fJ.o b2
27rp
(± 1 + ~).
7rap2
(6.5.17)

From this expression, it follows that the change of the sign of the force of inter-
action at large distances, is possible only under simultaneous fulfillment of
two conditions: the values of the force constants in the core of the dislocation
increase, and the limiting frequency we,,) is higher than the Debye's frequency.
An analysis of (6.5.16) in the general case, shows that change of the sign of
the force at small distances (of the order offew interatomic distances) is possible
only if the spatial dispersion is taken into account.

6.6 Edge Dislocations

Suppose that there are only edge dislocations in the medium, their lines being
parallel to the z-axis and their distribution being given by a density of the
vector baCr). As in the case of screw dislocations, we assume the fulfillment
 6.6 Edge Dislocations 161

of symmetry conditions, which ensure the uncoupling of the system (6.3.6).

Then the elastic field of the dislocation is described by a scalar function q>(r),
which satisfies the equation

(6.6.1)

Let us introduce the Green's function G(r, r'), which is a solution of

Rot a j3 !!8aj31i' Rot1i'G = oCr - r') . (6.6.2)

In this case, for non-zero components of stress, taking into account (6.3.2,
5), we have

(Tall(r) = - SRotaIlG(r, r')rot1bir') dr' ,

(6.6.3)
(T(r) = - SB-IB:rp aj3 •
The energy <!J according to (6.3.9), can be written in the form

(6.6.4)

where

(6.6.5)

For a system of N dislocations with the Burgers vectors b~·

(i = 1, 2, ... , N)

2<!J = r.; b~/JfJGall(r

'J
j, r) . (6.6.6)

Let us first consider the case of a homogeneous isotropic medium without

taking into account the change of force constants in the dislocation core.
With (5.3.9) and assuming that the Poisson coefficient Vo = const, we transform
(6.6.2) for the Green's function G(r - r') into

1 -; Vo f1.-16.2G(r) = oCr) . (6.6.7)

Hence we have for G(k)

G(k) = 1 2 f1.(k) (6.6.8)

-vo~'

It is not difficult to prove that the function her), which is the inverse Fourier
transform of k- 4, has the form
162 6. Dislocations

(6.6.9)

where g(r) is defined by (6.4.7).

The Green's function G(r) = G(r) is then represented in the form

2
G(r) = -1--- ph(r) . (6.6.10)
- ))0

For the Debye's model and for the model (5.3.18) we have

2po
G(r) = T=-))o h(r) ,
(6.6.11)
G(r) = 1 2!0))0 [her) - p;;-2g(r) + 2r;- 1 ,,-40/ r )J,
respectively.
In the case of a single edge dislocation, the substitution of (6.6.11) into
(6.6.3) enables us to obtain closed expressions for the stress. For the Debye
model they coincide with the corresponding expressions obtained by Brailsford
[6.6].
Consider now two parallel dislocations at the points rl and r2 with the
Burgers vectors bI = (b, 0), b2 = (± b, 0) (the slip planes are parallel). Then,
for the energy of interaction I/>int' according to (6.6.6), we have

(6.6.12)

The projection of the force of interaction onto the slip plane for the model
(5.3.18) is equal to

(6.6.13)

For the Debye's model the last two terms in the above expression should be
omitted.
In the polar coordinate system the asymptotics of qx has the form

~ + POb2 (COSOCOS20 _ 4 ,,-2 cos 0(1 - 4 sin 2 0) (6.6.14)

qx - 2n-(1 - ))0) p r p3 .

The first term coincides with the known expression for qx in the continuum
theory of dislocations. As distinct from the case of a screw dislocation, the
dispersion makes here the additional contribution of the order of p-3 to the
asymptotics. For the Debye model the asymptotics is given by the first term
only.
 6.7 Notes 163

From (6.6.14) it follows that the equilibrium of a dislocation is possible

either when () = 1C/2, or on some curve with the asymptotics () = 1C/4. In the
continuum theory of dislocations, equilibria are possible only when () = 1C/2
and () = 1C/4. The qualitative correspondence between signs of the Burgers
vectors and the stability of states of equilibrium are also preserved when the
spatial dispersion is taken into account.
At small distances of the order of a few interatomic distances, the ap-
pearance of new equilibria is possible, but it happens only if the boundary
frequency w(,.) is larger, than the Debye frequency.
Let us now consider in the general form the influence of the change of
the force constants in the cores of edge dislocations, on the Green's function.
From physical considerations it is clear that the change ofthe force constants
in the core of an edge dislocation is different on different sides of the slip
plane. On the side of the extra-semi plane, the atoms of the core come closer
to each other and the rigidity of bonds increases. On the other side of the
slip plane, the rigidity of the bonds decreases. Thus the perturbation of elastic
bonds in the core has the nature of a dipole perpendicular to the Burgers
vector.
For a system of parallel edge dislocations, let us represent the kernel of
the operator (6.6.1) in the form

(6.6.15)

Here, Vi = nro« is a scalar operator, n~ is the unit vector perpendicular to

the Burgers vector bi and directed to the side of the extra semiplane.
Let us denote by Go(r - r') the Green's function of the operator 1!IfJ. Then,
for the Green's function G(r, r') of the operator (}l, we have

G(r, r') = Go(r - r') - r;ViG3.8(r - r'.)P~~J.,..VPtt(rj - r'). (6.6.16)

IJ

The matrix po is inverse to the matrix

(6.6.17)

where

(6.6.18)

6.7 Notes

Basic physical notions of dislocation theory can be found, for example, in

[B6.4, 9-11]. Mathematical aspects of the continuum theory of dislocations
are considered in [B5.6, B6.1, 5-7, 15-20, 30, 32-35, 38, 41, 44].
164 6. Dislocations

Dislocations in the scope of models of elastic media with spatial dispersion

and of the Debye's quasicontinuum have been investigated in works of Rogula
[B6.36], Kosevich and Natsik [B6.13], Vdovin and Kunin [B6.40] and Brailsford
[B6.2, 3].
7. Elastic Medium with Random Fields of Inhomogeneities 1

In this chapter the method of the effective field is applied to solve problems for
composites and cracked solids. Under the assumption of a random change of
the effective field from one particle to another the formulae for the first and
second moments of random stress-strain fields are presented.

7.1 Background

The problem of a random field of inhomogeneities in elastic medium has

implications in a number of important applications for example, to composite
materials in engineering. Such materials, which consist of essentially different
components, facilitate a construction with extremely high strength and defor-
mational properties. In calculating properties of such designs with respect to
strength there is a problem of choice of an adequate mechanical model for the
composites. In a number of cases their behavior is well described by the model
of homogeneous elastic medium (matrix), which contains a random field of
inclusions with elastic properties.
An other field of applications of the theory of inhomogeneous medium is
the mechanics of fracture. The description of the process of microcracks, voids
and other defects growing in solids, is connected with the problem of many-
particle interactions of inhomogeneities in an elastic medium.
As a rule the sizes of inclusions and the distances between them are much
smaller than size of a body and a typical scale for the change of an external
field. That is why here an infinite homogeneous elastic medium containing a
homogeneous random set of inclusions in a constant external field of stress is
considered. The problem is to describe random stress and strain fields in such a
medium.
It is well known [7.1] that the complete information about a random field is
provided by its characteristic functional or an infinite set of its many points
statistical moments. Unfortunately, the difficulties which arise in constructing
all these objects for stress-strain fields in composits, can hardly be overcome.
But for applications it is not necessary to have complete information about
a random field. As a rule only several first statistical moments can be of interest.

IThis chapter is written by S. K. Kanaun.

166 7. Elastic Medium with Random Fields ofInhomogeneities

One of the most important characteristics of an inhomogeneous medium

is its effective elastic moduli tensor. In order to determine this tensor we should
have the first moment of the statistical solution or the mean values of the stress
and strain fields. Most of the investigations in mechanics of an inhomogeneous
medium are devoted to that very subject.
The fluctuations of the elastic fields are important for a description of
phenomena depending on the fine structure of the composite, for example
fracture and the beginning of the plastic shear. The numerical characteristics
of the fluctuations (dispersions, correlation radii) can be found from the second
statistical moments of the solution. The problem of the second moment is more
complicated than that of calculating the overall moduli; the difficulties increase
here substantially.
Most probably the first two statistical moments give us the most valuable
information about the random stress-strain fields in composites. This informa-
tion is necessary for the right description of the elastic-plastic deformation and
fracture.
In many cases we can distinguish a typical element or a particle of the
heterogeneous medium. An inclusion in a composite, a grain in a polycrystal,
a crack in a cracked solid can be an example of a particle of this kind. The
solution of the problem is reduced here to the description of many-particle
interactions.
If the concentration of particles in a homogeneous medium is small and we
can neglect their interactions, then the problem can be solved exactly, but the
solution for an isolated particle must be known. In the cases of large con-
centrations, when the interaction is essential, as a rule, an approximate solution
can only be obtained.
There are several methods to account for the interactions between particles
in an elastic heterogeneous medium. A survey of these methods can be found in
[7.2], for example.
Here we shall consider the method for evaluating the statistical moments of
the random elastic fields in composities. The idea behind this method is closely
related to the self-consistent method, well known in physics.
The self-consistent method (SCM) is one of the most powerful methods to
solve the many-particle problem. In the quantum theory of the atom (the
Hartree-Fock approach), or the description of phase transitions (Weiss method,
Landau theory) SCM enables one to obtain good approximations in many
important cases. These cases are well known. The external field for every particle
must weakly depend on the concrete configuration of the surrounding particles
and must be approximately equal to the combined field of all interacting
particles.
In the theory of an inhomogeneous medium the self-consisted method is
based on the following assumption. Every particle is considered isolated and
contained in a homogeneous medium, the elastic moduli of which are equal to
the unknown effective elastic moduli of the heterogeneous material. The external
 7.2. Formulation of the Problem 167

field for every particle, which is assumed equal to the external field, applies to
the whole body. Let the solution of the last problem be known and therefore
the function describing the state of every isolated particle can be expressed in
terms of its properties, the effective parameters of the medium and an external
field. After the summing of perturbations from all the particles an overall stress
and strain may be found and an effective elastic moduli tensor of the composite
is expressed in this way through the particle states functions.
As a result this way leads to a system of algebraic equations for the effective
parameters of elasticity (Sect. 7.3). Such a modification of SCM will be called
the method of an effective medium.
For the case of a composite medium another avenue for solving this problem
can be suggested. Every particle may be considered isolated in a homogeneous
medium-the matrix of the composite. The presence of surrounding particles is
accounted for through an effective external field applying to every particle. The
scheme which is based on such an assumption will be called the method of the
effective field.
It is commonly accepted that the effective field is constant (and the same
for all the particles). In this case the method of the effective field coincides
with a modification of SCM, which is widely used in quantum mechanics,
statistical physics and also in the mechanics of heterogeneous medium. The
assumption about the constant effective field can be confirmed by calculating
the effective parameters of an inhomogeneous medium. But in order to describe
the fluctuations of elastic fields in a composite we have to account for the
fluctuations of the effective field.
In this chapter the method of the effective field is applied to solve problems
for composites and cracked solids. Under the assumption of a random
change of the effective field from one particle to another the formulae for the
first and second moments of random stress-strain fields are presented. The
general scheme of the method is considered in Sect. 7.2-5. Sections 7.6-9 are
devoted to the first approximation of the method, coinciding with a modifica-
tion of SCM. Formulae for the effective constants of composites are obtained.
The construction of the second moment is described in Sect. 7.11 for the case
of point defects. The results are compared with the exact solutions for regular
composites and experiments. Estimates for the accuracy of the method is
given. Finally, the use of the results obtained for composite fracture is
shown in Sect. 7.12.

7.2 Formulation of the Problem

Let us consider an infinite elastic medium the elastic moduli tensor of which
is

C(x) = Co + C1(X) , (7.2.1)

168 7. Elastic Medium with Random Fields ofInhomogeneities

where Co is the elastic moduli tensor of the homogeneous medium, C1(x)

is a perturbation caused by the inhomogeneities. In the case of a composite
medium C1(x) can be represented in the form

(7.2.2)

where V,(x) is the characteristic function of the region occupied by the i-th
inclusion, the tensor Cli is constant within the i-th inclusion but, generally
speaking, varies from one inclusion to another.
The elastic compliance tensor of the inhomogeneous medium B = C-l
is represented in the same way as (7.2.1, 2),

B(x) = Bo + B1(x), (7.2.3)

where Bo = COl, Bli are the constant tensors within each inclusion.
If the fields of stress o-t(x) and strain ct(x) inside the inclusions are known,
the tensors I1(X) and c(x) at any point of the medium are represented in the
form (Sect. 4.8)

c(x) = co - JKo(x - x')C1c+(x') dx' ,

(7.2.4)
o{x) = 110 - JSo(x - x')B l l1+(x') dx' ,

where co and 110 are the external strain and stress fields, respectively. Here
we denote

(7.2.5)
B l l1+(x) = 1:; Blil1t(X) V,{x) .
i

The kernels of the integral operators Ko and So in (7.2.4) are defined via
the Green's tensor for displacements of the homogeneous medium (Sect. 4.6).
It will be assumed further that the regions Vi occupied by inclusions consti-
tute a homogeneous random field and all the tensors Cli are independent ran-
dom variables having the same known density functions. Such a model involves
the most interesting composite materials. If an external field 110 (or co) is constant,
the functions C1c+(x) and B l l1+(x) for each realization of the random distribu-
tion of the regions Vi in space may be represented at the sum of an almost per-
iodic function and a function with bounded support. The operators Ko and So
are defined upon such a class of functions in Sect. 4.8 and Appendix A3. It is
essential that a unique definition of these operators is possible, if it is known
what kind of external field is given: the stress field 110 or the strain field co.
 7.2 Formulation of the Problem 169

Further, for definiteness we assume that the external stress field 0'0 is always
fixed. Then the operators Ko and So are defined on constants via (Sect. 4.8)

JKo(x - x') dx' = Bo, JSo(x - x') dx' = O. (7.2.6)

The problem for a fixed external strain field can be treated similarly.
Let us consider separately the important case of a homogeneous medium
containing isolated cavities or cracks. For cavities C li = - Co, and Bli =
(Co + Cli)-l - COl tends to infinity while C li ~ - Co. Since

(7.2.7)

then BliO't ~ e+ while C li ~ - Co.

The relations (2.4) can be rewritten in this case as

e(X) = eo + JKo(x - X')Coe+(X') dx' ,

(7.2.8)
O'(x) = 0'0 - JSo(x - X')e+(X') dx' .

Note that the right-hand sides of (7.2.4 and 8) may be interpreted as the stress
and strain fields in a homogeneous medium, containing, in the regions V k ,
sources of dislocations moments with the density

(7.2.9)

in the case of inclusions and

(7.2.10)

in the case of cavities (cf. Sect. 5.3). It is essential that these dislocation moments
are induced in the medium by applying the external field 0'0 because the fields
O'+(x) and e+(X) are linear functions of 0'0'
Let us consider now the case in which defects are cracks or cuts along the
smooth, oriented surfaces Ok' If we pass to the limit from a cavity Vk to the
crack Ok, as it was shown in Sect. 4.11, we have et(X) ~ Mk(x)o(D k)
where

(7.2.11)

nk(x) is the normal to the surface Ok, b,.(x) is a vector field at Ok which can be
interpreted as a jump of displacements across the crack, 0(0.) is a delta-func-
tion concentrated at Ok (Sect. 6.1).
170 7. Elastic Medium with Random Fields of Inhomogeneities

As it follows from (7.2.6), the stress and strain fields in a cracked medium
are represented in the form

c(x) = co + JKo(x - x')CoM(x')o(O) dx' ,

(7.2.12)
O'(x) = 0'0 - JSo(x - x')M(x')o(O) dx' ,

where we denote

(7.2.13)

The equations for the vectors bix) follow from the boundary conditions
on the surfaces of the cut and have the form

4:, Jn,,(x)So(x - x')n,{x')b;(x')o(Oj) dx' = nk(x) 0'0 ,

with x E Ok, k = 1,2, .... (7.2.14)

For the case of a single crack a similar equation is considered in Sect. 4.11.

7.3 The Effective Field

Suppose that isolated defects constitute a spatially homogeneous, random

field. Let us fix one of its typical realizations and consider an arbitrary defect
with number i. If an external field is applied to the medium, this defect is sub-
ject to the field O';(x); on the basis of previous considerations O',{x) can be re-
presented in the form

X E Vj , (7.3.1)

where m,,(x) is defined by (7.2.9 or 10) for the case of inclusions or cavities and
by (7.2.11) for the case of cracks. The field O';(x) defined by this relation in the
region V j may be interpreted as the external field of the i-th defect. In the field
O',{x) this defect behaves as an isolated one.
Sometimes it is more convenient to consider an external strain field e,{x)
of the i-th defect. The connection between cSj(x) and O',{x) is represented by the
obvious relation

cSj(x) = BoO',{x) . (7.3.2)

Let the solution of the elastic problem for isolated defects in an arbitrary
external field be known. This means that we have explicit expressions for the
 7.3 The Effective Field 171

functions mk(x) in (7.3.1) henceforth denoted by mk(x, if k). Thus the system
of equations governing the fields O'k(X) for each of the interacting defects,
is obtained from (7.3.1) in the form

(7.3.3)

If the solution of this system has been found, the stress and strain fields in an
inhomogeneous medium are determined by

(7.3.4)

which are the consequences of (7.2.4, 8). Thus the fields ifk(X) can be considered
as the main unknowns of the problem.
If inhomogeneities constitute a random field, ifk(x) are random functions.
For the purpose of constructing the statistical moments of ifk(x) let us introduce
the following simplifying assumptions about the structure of O'k(X) (hypothesis
of the effective field method) :
HI) The field O'k(X) is practically constant inside the region Vk (or Ok) but,
generally speaking, these fields are different for the different defects.
H2) The random variables 0' k do not depend statistically on the geometric
characteristics and elastic constants of the defect Vk •
Note that the constants if k may be regarded as the average of the external
field O'k(X) over the volume Vk.
The field O'k(X) defined in the region V = UV k and equal to the constant
O'k in Vk will be called the effective field. Note that in the classical scheme of the
self-consisted method if is assumed constant and equal for each particle [7.3, 4].
In the present modification of the method this field is considered as a random
one and as it will be shown below, it enables us to describe more precisely the
interaction between defects in the various stochastic fields of inhomogeneities
The picture for the interactions corresponding to the hypotheses HI and H2
may be described qualitatively in the following way. For the typical defect the
external field O',.(x) which is represented as a sum of the external field 0"0 and the
fields caused by all the surrounding inhomogeneities, is approximately constant
in the region V .. occupied by this defect. Furthermore the contribution of every
individual defect to this field (0' .. ) is negligible. Without considering the realm
of application of these hypotheses, we proceed now to an analysis of the formal
consequences.
Further we shall assume that defects are inclusions of ellipsoidal shape, or
plane elliptical cuts (cracks). As follows from the solution for an isolated el-
lipsoidal inhomogeneity in a constant external field, the density of the disloca-
tion moment mk(x, O'k) induced in the region V k is represented in the form
172 7. Elastic Medium with Random Fields of Inhomogeneities

(7.3.5)

where

(7.3.6)

Here Ak is the singular part of the operator Ko which in the three-dimensional

case has the form (Appendix A3)

(7.3.7)

The linear transformation a,/ transforms V k into the unit sphere, Ko(k) is the
k-representation of the Green's function for strains, ill is the surface of the unit
sphere in k-space.
In the case of elliptical cracks the density mix, iJ k ) in (7.3.3, 4) has the form

(7.3.8)

where the expression for P k(X) is found from the solution of the problem for an
elliptical crack in the constant external field iJ k (Sect. 4.12) and has the form

(7.3.9)

Here the function hk(x) is defined on the surface Ok by the relation

( l 2)
h kX,X _ !Ii
- b
k
J (XI)2
1- -
ak
- (- X2 )2 ,
bk
(7.3.10)

where the axes Xl> X2 are connected with the main axes of the crack. Here ak and
bk are the half axes of the ellipse Ok, ak 2 b k. The constant tensor P k in (7.3.9)
is defined by

(7.3.11)

where nk is the unit normal to Ok and the tensor To has the form of(4.12.8).
In the case of an isotropic medium TOI is

T-l - 2a~(1 - vo) d 5:

(7.3.12)
Oaf3 - bk/l-O aUa(J

where the scalar coefficients d a are defined by (4.12.8), if we substitute al and a2

into ak and b k , respectively.
 7.3 The Effective Field 173

The relations (7.3.4) for cracks may be rewritten as

(7.3.13)

Let us fix now a point Xo and introduce the region Vxo by the relation

if Xo E V
(7.3.14)

where Vj is the region occupied by j-th inclusion. Vxo denotes the complement
of Vxo to V. Evidently Vxo is the region Vj in which the point Xo is situated.
The characteristic functions (of argument x) of the regions Vxo and Vxo are
denoted by V(xo; x) and V(xo; x), respectively: hence

Vex) = V(xo; x) + V(xo; x). (7.3.16)

Let P(x) be an arbitrary continuous tensor field which coincides with Pi'
see (7.3.6), in the regions Vi. Let the field O'(x) be defined in the region V by

O'(x) = 0"0 + JSo(x - x')P(x')O'(x')V(x; x') dx', x E V. (7.3.17)

If the hypothesis HI is valid, the field O'(x) coincides with O'i in the regions
Vi and consequently with the effective field in V.
For the case of cracks let us introduce the region 0 of all surfaces of the
cracks and the region Oxo which is

(7.3.18)

Thus the equation governing the effective field in the region 0 is represented
in the form

O'(x) = 0"0 + JSo(x - x')P(x')O'(x')o(Ox) dx', (7.3.19)

where P(x) is an arbitrary continuous tensor field coinciding with P k(X) , see
(7.3.9), on the surfaces Ok.
Equations (7.3.17) and (7.3.19) represent the grounds for constructing
the statistical moments of the effective field 0'. It will be necessary for the
solution of this problem to calculate the various averages of the random
fields which have to be considered here. In the following section we shall
discuss in details the problem of construction of these averages.
174 7. Elastic Medium with Random Fields of Inhomogeneities

7.4 Several Mean Values of Homogeneous Random Fields

In what follows, it will be assumed that V(x) is the characteristic function

of a random set of ellipsoidal regions, which are distributed homogeneously
in the three-dimensional space. Fields of the type of the stress tensor O'(x),
the strain tensor e(x) as well as the effective field O'(x) are functionals of the
random function V(x). For a homogeneous external stress field, all these
random fields are homogeneous, since their statistical characteristics must
be invariant with respect to the group of translations.
Letf(x, V) be one of the functions under consideration. The mean (f(x)
with respect to the ensemble of realizations is the integral in the functional
space [7.5]

(f(x) = Jj(x, V)df.l(V) , (7.4.1)

where f.l(V) is a measure in the functional space. Analogously, the correlation

function of the field f(x) is defined as

(f(XI)f(X2) = Sf(XI, V)f(X2, V)df.l(V) , (7.4.2)

and the higher moments of f(x) are defined in the same manner.
In what follows, an important role will be played by the means of f(x)
calculated under the condition that the region V contains a fixed point Xl' For
this conditional mean, we introduce the commonly accepted notation 2 (f(x)lxI)'
By definition [7.5],

(f(x)lxI) = (V(x)-l J!(x)V(xI)df.l(V) . (7.4.3)

Analogously, the mean of the field f(x) under the condition that the region
V contains the points Xh ... , Xn has the form

n n
(f(x)lxh X2, ... , x n ) = (IT V(X;)-l S/(XI) IT V(x,-}df.l(V)' (7.4.4)
;=1 ;=1

Let now X E V, Xl E V x ' The mean of the fieldf(x) under these conditions
will be denoted by (f(x)Ix ; Xl), and the expression for this mean has the
form

(f(x)lx; Xl) = (V(x)V(x; XI)-l Sf(x) V(x) V(x; xI)df.l(V). (7.4.5)

2The symbol <I> is nowhere used in this chapter to denote the scalar product, therefore
no confusion should arise.
 7.4 Several Mean Values of Homogeneous Random Fields 175

Note that here the equality

(7.4.6)

is valid.
In the general case, the sign";" separates variables which belong to the
region V x , from variables which belong to the region Vx • Thus the mean of
!(XI)!(X2) under the condition Xl E V, X2 E V:r' E VX1 is written in the form

(!(XI)!(X2)I Xl> X2; X3, X4)

= (V(XI)V(XI; X2)V(XI; X3)V(XI; X4»-1
U(XI)!(X2)V(XI)V(XI; X2)V(XI; X3)V(XI; xJdf.t(V). (7.4.7)

Generalization of this formula to any number of functions and variables is

obvious.
If the value of! at a point X does not depend statistically on values of the
function V(x) (though! can be a functional of V), then the following relations
are valid:

(!(X)V(X» = (!(x)lx) (V(x» , (7.4.8)

(7.4.9)

and so on.
Let us note the equality which is a consequence of (7.3.16)

(7.4.10)

For a homogeneous field of inclusions, both terms on the right-hand side

are functions of the difference X - Xl. If the dimensions of inclusions are
bound, the second term is essentially non-zero only in a finite neighborhood
of the origin. The first term vanishes when X = Xl> since V(x; x) == 0 in view
of (7.3.14) (Appendix A4).
All homogeneous random fields we shall use will be assumed to be ergodic:
the means with respect to the ensemble of realizations V(x) coincide with the
corresponding means with respect to the whole three-space of a fixed typical
realization. Let to such a realization V*(x) correspond a field f*(x). Then

(7.4.11)

(7.4.12)
176 7. Elastic Medium with Random Fields of Inhomogeneities

where on the left-hand sides we have the means with respect to the ensemble
of realizations of the random function V(x), G is a region of the volume v, which
occupies the whole space at the limit.
Analogously, for conditional means, the ergodic property enables us to
write

<J(x)lx) = <V(x»-llim -.L Sf*(x)V*(x) dx, (7.4.13)

v--+oo V G

where

<V(x» = lim -.L SV*(x) dx. (7.4.14)

v->oo V G

An example of a more complicated conditional mean is the following:

<!(XI)!(XI + x2)lx l> Xl + X2;)

= <V(XI)V(XI; Xl + X2»-1
lim -.L Sf*(XI)!*(XI
v->oo V G
+ X2) V*(XI) V*(XI; Xl + X2) dXI. (7.4.15)

Here

(7.4.16)

The asterisk for a fixed realization will later be omitted.

When considering random fields of cracks, the role of the region V will be
assumed by the set Q of points of crack surfaces, and, instead of the char-
acteristic function V(X) , the generalized function o(Q) will be used (see the
definition in Sect. 6.1). In order to simplify our formulae, let us introduce the
notation

o(Q) == Q(x). (7.4.17)

Then the mean of !(XI) under the condition that Xl> X2, ... , Xn ED, becomes

(7.4.18)

Here fJ-(D) is a measure in the functional space, TI7=ID(xj) is a generalized func-

tion on the direct product of Euclidean spaces.
Let now Qx be the surface, defined by (7.3.18), and Dx be the complement
to Dx in D. The generalized functions o(Qx) and o(D x) will be denoted as follows:
 7.5 General Scheme for Constructing First Statistical Moments of the Solution 177

o(D,,) = D(x; x'), o(Q,,) = Q(x; X'). (7.4.19)

The mean of the field f(x) under the condition that xED, Xl> xz, ... , xn E
Q,,; x n+l> ... , xm ED" is represented in the form

n m m
= <D(x) IT Q(x; x,) IT D(x; Xk)-l S!(xI)D(x) IT D(x; xk)dp(x).
1=1 k=n+1 k=n+1
(7.4.20)

Note that it will always be clear from the context, which field (of cracks
or of inclusions) is considered; therefore the left-hand sides of (7.4.4 and 18)
have the same notations.
For ergodic fields of cracks, calculations of their means with respect to
the ensemble of realization can be substituted by averaging throughout the
space for a fixed typical realization D(x). In this case, the mean <f(XI)lxI;
Xl + xz) for example, can be found from the relation, see (7.4.15),

<f(XI)lxI; Xl + xz)
= <D(XI)D(XI; Xl + xz)-llim.l
V G
v-->oo
S!(XI)Q(XI)D(XI; Xl + Xz) dXI, (7.4.21)

where

(7.4.22)

In conclusion, let us note the formula for the mean <D(x)lxI), which is
analogous to (7.4.10),

(7.4.23)

For homogeneous fields of cracks, both terms on the right-hand side are
functions of the difference X - Xl, the first term being equal to zero when
X = Xl> and the support of the second one is concentrated in a finite neigh-
borhood of the origin, if the dimensions of the cracks are bound.
The construction of means of the type (7.4.22, 23) for specific stochastic
fields of cracks is considered in Appendix A4.

7.5 General Scheme for Constructing First Statistical Moments

of the Solution

Let us begin with constructing the main statistical characteristics of the effective
field O'(x). We denote by O'n(Xl> ... , xn) an n-tuple moment of the effective field,
i.e. the mean of the tensor product O'(XI) ..... O'(xn) under the condition that the
178 7. Elastic Medium with Random Fields of Inhomogeneities

points XI. ... , Xn belong to V(or ()). In particular, the expectation and the two-
point moment of the effective field are conditional means of the form

(7.5.1)

In order to obtain an expression for ii', let us average both sides of(7.3.17)
with respect to the ensemble of realizations of the random field of inhomogen-
eities under the condition x E V

(ii(x)lx) = 0'0 + JSo(x - x') (P(x')ii(X') Vex; x')lx) dx' . (7.5.2)

Since x is a fixed point here, So(x - x') is a determined kernel. Using the
hypothesis H2 (Sect. 7.3) about the statistical independence of the field ii(X)
at the point x of the elastic properties, the form and the dimension of the in-
clusion located at this point, let us represent the mean in the integrand in(7.5.2)
in the form of a product of means, analogously to (7.4.9). Then the expression
for iiI becomes

iiI = 0'0 + JSo(x - x') (P(x')V(x; x')lx) (ii(x')lx'; x) dx' , (7.5.3)

where the mean (ii(x')lx'; x) is calculated under the condition that x' E V and
XE Vx "
In order to construct the mean (ii(x')lx'; x) in the integrand of (7.5.3),
let us again consider (7.3.17). Averaging both sides of the equation under the
condition x E V, Xl E Vx and using the hypothesis H2 (Sect. 7.3) once more,
we have

(ii(x)lx; Xl) = 0'0

+ JSo(x - x')(P(x')V(x; x')lx; x')(ii(x')lx; xI. x') dx' . (7.5.4)

The equations (7.5.3, 4) are not closed, since their right-hand sides contain
the mean values of the effective field, which are calculated under conditions,
different from those in the left-hand sides. Thus, a chain of equations arises
here, and to close it, one has to introduce additional assumptions about the
means in (7.5.3,4) or in analogous relations which connect more complicated
conditional means.
As it was pointed out in Sect. 7.3, the classical scheme of the method of self-
consistent field consists of assuming ii(x) to be constant and equal for all par-
ticles. Then ii' and the conditional mean (ii(X') lx' ; x) coincide with each other
and with ii, and the constant ii is given by (7.5.3).
In the scope of the modification (of the method) developed here, the equiv-
alent assumption which enables one to obtain equations for expectation of the
 7.5 General Scheme for Constructing First Statistical Moments of the Solution 179

field 0'1, is that 3

(O'(x')lx'; x) = (O'(x')lx') = 0'1. (7.5.5)

Substituting this relation into (7.5.3) we arrive at the equation with respect
to 0'1

0'1 = 110 + f SO(X - x') (P(X')V(X; x')lx) dx' 0'1. (7.5.6)

Thus, the problem is reduced to constructing the mean by calculating the

integral on the right-hand side of (7.5.6).
The next, more precise, approximation for 0'1 can be obtained, if one assumes
in (7.5.4) that

(O'(x')lx; Xl> x') = (O'(x')lx'; Xl). (7.5.7)

For a homogeneous random field of defects, the mean on the right-hand

side depends on the difference x' - Xl only:

(O'(x') lx' ; Xl) = €p(x' - Xl) . (7.5.8)

The function €p(x' - Xl) is the mean value of the effective field at the point x'
under the condition that there is a defect at the point Xl> and this characterizes
the pair intt.raction in the system of interacting defects. It is obvious that,
as lx' - xli --+ 00, this function tends to a mean value of the effective field 0'1.
The equation for €p(x) follows from (7.5.4, 7) and has the form

€p(x) = 110 + f So(x - x') (P(x')V(x; x')lx; x', 0) qJ(x') dx' . (7.5.9)

To obtain the second moment 0'2(X1 - X2) of the effective field, let us take
a product of values of the field O'(x) (the relation (7.3.17) at different points
Xl and xz) and average the result under the condition Xl> X2 E V

0'2(X1 - X2) = 110(0'(x2)lxl> X2)

+ f SO(X1 - x') (P(X')V(X1; x')0'(x')0'(x2)lxl> X2) dx' . (7.5.10)

The equation for the function 0'2(X1 - X2) will be obtained after splitting
the mean in the integrand with the help of the hypothesis H2 (Sect. 7.3)

SHere, the influence of a defect at the point x on the mean value of the effective field at
the point x' is not taken into account.
180 7. Elastic Medium with Random Fields of Inhomogeneities

(P(x') V(X1; X')0'(x')tr(x2)lxl> X2)

= (P(X') V(X1; x')IXl> X2) (0'(X')0'(x2)lx', xl> X2) (7.5.11)

and assuming, see (7.5.5, 7), that

(7.5.12)

From here and from (7.5.10) we obtain

0'2(X1 - X2) = 0'0 g>(X2 - Xl)

+ JSO(X1 - x') (P(X')V(X1; x')IXl> X2)0'2(X' - X2) dx' , (7.5.13)

where g>(x) is a solution of (7.5.9).

The way of constructing subsequent approximations for 0'2(X) and g>(x)
in the scope of the scheme given, is obvious. In what follows, we confine our-
selves to the consideration of (7.5.6, 9 and 13). Solutions constructed on the
basis of these equations apparently describe a number of experimental data
and, in many cases interesting for application, give a good agreement with exact
solutions (Sects. 7.7, 9).
Let us proceed to derive the first two statistical moments of the strain and
stress fields in a composite medium. Let us substitute the expressions for
mk(x. O'k) in the form of (7.3.5) into (7.3.4) for O'(x) and e(x) and average the
result with respect to the ensemble of realizations of the random field of inclu-
sions. Using the hypotheses HI and H2 (Sect. 7.3) about the structure of the
effective field O'(x), we have

(O'(x) = 0'0 + JSo(x - x') (P(x') V(x') (O'(x')lx') dx' ,

(7.5.14)
(e(x) = eo - JKo(x - x')Co (P(x')V(x') (O'(x')lx') dx' .

For homogeneous random fields of inclusions, the means (P(x)V(x) and

(O'(x)lx) = 0'1 are constants. Since the operators So and Ko act on constants in
accordance with (7.2.6), then, from (7.5.14), we obtain

(O'(x) = 0'0, (e(x) = eo - (P(x) V(X)O'l . (7.5.15)

The tensor of effective elastic compliance B* is introduced in a natural way

(7.5.16)

Then, from (7.5.15) and (7.5.6) it follows that

B* = Bo - (P(x)V(x) A, (7.5.17)
 7.5 General Scheme for Constructing First Statistical Moments of the Solution 181

where the fourth rank tensor A connects the external field with the first moment
jj1 of the effective field

jjl = AO'o (7.5.18)

and is determined by the relation

A= [1 - JSo(x - x') (P(x')V(x; x')Jx) dX'JI . (7.5.19)

In order to construct the second statistical moments of the solution, let us

consider the product of the right-hand side of the expressions (7.3.4) for stress
(strain) taken at different points Xl and X2 and average the result with respect
to the ensemble of realizations of fields of inclusions. Using the main hypothesis
of the method of the self-consistent field and (7.2.6) which defines the action of
the operators Ko and So on constants, we obtain

(0'(XI)0'(X2» = 0'5
+ JSO(XI J
- x') dx' SO(X2 - x") (P(x')P(x") Vex') V(x") )jj2(X' - x") dx" ,
(E(XI)e(X2» = E5 + EO(P(X) V(x» jjl + (P(x) V(X»jjIEo
+ JKO(XI - x')Codx' KO(X2J
-x")Co(P(x')P(x")V(x')V(X"»jj2(X' - x") dx" . (7.5.20)

Analogously, one can write down the expression for mean values of the
energy density (rJJ) = (O'E)/2, of a composite medium, of the density (rJJint )
of defects interaction, etc.
When considering a random field of cracks, one has to substitute the func-
tions P(x') Vex') and P(x') Vex; x') into the preceding relations for P(X')Q(x')
and P(x')Q(x; x'), respectively, where now an arbitrary continuous function
P(x) coincides with the functions Pix), (7.3.9), on the surfaces Qk' Thus, to
construct the first two moments of the random fields O'(x) and E(X), one has
to find the statistical moments jj1 and jj2 of the effective field and then to con-
struct the means and to compute the integrals appearing in (7.5.17, 20).
In conclusion, let us compare the method of the self-consistent field with
another self-consistent method of constructing effective elastic constants,
namely, with the method of the effective medium. It was pointed out in Sect.
7.1 that the main assumption, on which the second is based, is that each inclu-
sion in a composite behaves as an isolated one in a homogeneous medium,
properties of the latter coinciding with the effective properties of the whole com-
posite. The external field for each inclusion is assumed to be 0'0'
Using these assumptions, we can transform (7.3.4) for stress and strain fields,
in the case of ellipsoidal inclusions, to the form
182 7. Elastic Medium with Random Fields of Inhomogeneities

O'(x) = 0'0 + JSo(x - x') 1:k Pk* Vix') dx' 0'0,

(7.5.21)
C(X) = co - JKo(x - xo)Co 1: P k* Vk(X') dx' 0'0
k

where the tensors P k * are expressed through yet unknown effective elastic pa-
rameters B* of the composite

(7.5.22)

Here Clk * is a perturbation of the elastic modulus for the k-th inclusion with
respect to the effective tensor C* = B;l

(7.5.23)

A tensor A k * is defined by (7.3.7), where the function Ko(a,/k) is to be replaced

by K*(k), being k-representation of the Green's tensor for the effective medium
strain.
The equation for the unknown tensor B* is obtained from the condition of
self-consistency, which coincides with (7.5.16)

(7.5.16)

Having averaged(7.5.1 l)for the stress and strain tensors and acting then on
the constant, which is the mean of the sum in the integrand, by the operators
Ko and So, we obtain from (7.5.16)

(7.5.24)

where

(7.5.25)

The relation (7.5.24) which connects the constants of the tensor C* with
each other, is to be considered an equation with respect to effective elastic
constants of the composite. The latter are contained on the right-hand side
through the tensor Clk *, (7.5.23), and A k *.
In the case of elliptic cracks, assumptions of the method of the effective
medium lead to expressions for stress and strain tensors, which coincide with
(7.3.13), if one sets if k = 0'0 and substitutes Pk(x) by Pk*(x), while Pk*(x) (an
analog to Pix) in (7.3.8)) is to be found from the solution of the problem
about an elliptic crack in a homogeneous medium with the elastic moduli
tensor C*. Using (7.5.16), one can obtain equations for the effective tensor
B* for a medium with cracks in the form
 7.6 Random Field of Ellipsoidal Inhomogeneities 183

B* = Bo - (P *h(x)Q(x» , (7.5.26)

where

(7.5.27)

Solving (7.5.24 and 26) with respect to the components of B* is possible if

there are explicit expressions for the tensors A* and P,,* in terms of the elastic
constants of the effective medium. However, a construction of these expres-
sions in the case of an arbitrary anisotropy of the tensor B* is connected with
serious technical difficulties (Sect. 4.8, 9). Besides, if the number of unknowns
is large (21 in the general case), then one can obtain the solution only numerical-
ly, which is not always convenient for applications. Thus, the method of the
effective medium is usually used when the composite is macroisotropic. A
comparison of numerical values of effective elastic constants, calculated with
the help of the method of the effective medium and of the method of the effec-
tive field, is carried out in the following sections.

7.6 Random Field of Ellipsoidal Inhomogeneities

Let us now consider in more detail the method of the effective field for a medium,
which contains inclusions of the ellipsoidal form. In this section, the ex-
pectation 0'1 of the effective field and the tensor B* of the effective elastic
compliance in the case of composite and polycrystal media will be derived.
From (7.5.17, 18), it follows that the problem of determining the tensors
0'1 and B* is reduced to constructing explicit expressions for the means

(7.6.1)

?J!(x - x') = (P(x')V(x; x')lx) (7.6.2)

and calculating the integral which represents the action of the generalized
functions So(x) on the function ?J!(x) defined by the last equality. The tensor
P" in (7.6.1) has the form of (7.3.6).
Let us start with constructing the means (7.6.1, 2). Due to the ergodic
property assumed, the mean (7.6.1) for homogeneous fields of nonintersecting
inclusions, has the form

(7.6.3)

where v" is the volume of the k-th inclusion, N is the number of inclusions
184 7. Elastic Medium with Random Fields of Inhomogeneities

in the region G. It is taken into account here that P k is a constant tensor inside
each of the inclusions.
If one considers the ensemble of realizations of the homogeneous random
field P(x) Vex), then the tensors PkVk will be stochastic variables with one and
the same distribution function for any k. Therefore, if both sides of (7.6.3)
are averaged with respect to the ensemble of realizations once more, then

<P(x)V(x» = lim ~<v(a)P(a» =

v-->oo V
<v(a)
Va
P(a» , (7.6.4)

where Va is the mean volume per an inclusion, v(a) is the volume of an ellipsoid
with semi axes aI, a2, a3, pea) is a tensor determined by the form of the ellipsoid
and by the elastic constants of the latter, see (7.3.6). On the right-hand side
of (7.6.4), we have already a mean value of the random variable v(a)P(a) ,
the distribution function of which is determir.ed by the statistical properties
of the ensemble of inclusions and is assumed to be known.
The problem of constructing the mean (7.6.2) for a homogeneous random
field of elliptic regions is considered in Appendix A4. Let us point out the
main properties of the function 1f!(x). This is a continuous, bound function,
its values being tensors of rank four. If the dimensions of all inclusions are
bound, then 1f!(x) can be represented as a sum of a constant component <1f!),
an oscilating function with the zero mean value and a function with bounded
<
support. The constant 1f!) is easily calculated, if one takes into account that
the constant components of the functions 1f!(x) and <P(x') V(x')lx) coincide,
since, for inclusions of bound dimensions, these functions differ from each
other in a finite neighborhood of the origin only. Using the ergodic property
and averaging later with respect to the ensemble, we obtain

< 1f!) = <V~X » <Jim V~ GSdx' GSP(x') Vex') Vex) dx)

v-->oo
= <v(a)
Va
P(a» , (7.6.5)

where the definition of the conditional mean is taken into account (Sect. 7.4):

<P(x') V(x')lx) = <P(x') V(x')V(x» (7.6.6)

<V(x»

The function 1f!(x) characterizes the averaged density of defects, which

surround a typical inclusion with its center at the origin. The form of /f/(x)
is determined by a specific structure of the random field of inhomogeneities.
If the latter possesses some symmetry (in the statistical sense), then that results
in the corresponding symmetry of the function 1f!(x). For example, if the regions
Vk form an isotropic field, then the function 1f!(x) is spherically symmetric,
i.e. 1f!(x) = 1f!(lxl).
A break of spherical symmetry of 1f!(x) can result in the appearance of
texture. By texture we understand the symmetry of the elastic constants tensor
 7.6 Random Field of Ellipsoidal Inhomogeneities 185

of the matrix being different from the symmetry of the effective constants
tensor of the inhomogeneous medium. For many stochastic composites, the
symmetry of the textUle connected with geometrical properties of the field
of inhomogeneities is not too low and this can be described with the help of
a second order tensor a. In such a simple case, a determines a linear transform,
which transforms 1f!(x) into a spherically symmetric function [1f!(ax) = 1f!'(lx'I)].
In the general case, such a linear transformation can not be found.
The function 1f!(x) is contained in (7.5.17, 18) for 0'1 and B* due to the
integral, which represents the action of the generalized function So(x) on 1f!(x).
Using properties of this generalized function (Appendix A3 and Sect. 4.6) and
the definition of the operator So acting on constants (7.2.6), we have

sSo(x - x') 1f!(x - x') dx' = SSo(x)[1f!(x) - <1f!)] dx

= -D<1f!) + fSo(ax)[1f!(ax) - <1f!)]detadx, (7.6.7)

where

D = 4~ SSo(a-1k) d{J) . (7.6.8)

1tw

Here the tensor a determines an arbitrary nondegenerate transformation of

the x-space, (J) is the surface of the unit sphere in the k-space, and 1f!) the <
mean value of 1f!(x) of the form of (7.6.5). The last integral in (7.6.7) is under-
stood in the sense of a principal value and converges at the origin and at
infinity.
Let us consider an important particular case. If 1f!(x) is a spherically sym-
metric function or there exists a linear transformation a, for which 1f!(ax)
is spherically symmetric, then due to properties of integrals in the sense of
a principal value, the last integral in (7.6.7) vanishes and the equality assumes
the highly simple form

SSo(x - x')1f!(x - x') dx' = -D<1f!). (7.6.9)

Substituting this relations into (7.5.17, 18) and taking into account (7.6.4),
we obtain expressions for the tensors 0'1 and B* in the form

(7.6.10)

B* = Bo - <v~:) p(a)[I <

+ D v~:) pea)~ J1 , (7.6.11)
186 7. Elastic Medium with Random Fields ofInhomogeneities

where

(7.6.12)

and the tensor A(a) is defined by (7.3.7) with ak = a.

Let us sum up. If, from a statistical investigation of a composite, one
can obtain an expression for the function W(x) and find a mean value of the
stochastic variable v(a)P(a), then the problem of constructing the tensors ii l
and B* in the scope of the first approximation in the method of the effective
field is reduced to a calculation of the converging integral (7.6.7). In the case
of spherical symmetry of the function or its simplest analog, this integral is
calculated in a finite form and the required tensors, take the form of (7.6.10,
11). A more complicated case of a regular lattice of inclusions will be con-
sidered in Sect. 7.7, numerical methods being then necessary even for construct-
ing the integral (7.6.7).
Note one important peculiarity of the solution obtained. Let all the in-
clusions have the form of a ball, the same elastic properties and be distributed
in the space isotropically. Then the function W(x) is spherically symmetric and
the expressions for the tensors B* and C* = B;l take the form

B* = Bo - pBOCl [Co + pCl + (1 - p)COACl]-l , (7.6.13)

C* = Co + pCl [I + (1 - p)ACl]-l , (7.6.14)

where A is defined by (7.3.7) with dif3 = o«f3. It is easy to verify that the formal,
limiting transition, as p -+ 1, leads to a physically reasonable result: B* -+
Bo + Bl . Thus, the results obtained are noncontradictory in the entire region
of the concentration of inclusions, though, while p is close to 1, the hypotheses
of the method of an effective field lose sense, since one can not speak of isolated
inhomogeneities in this case.
Note one more important case. In the scope of the method presented, one
can consider not only effective elastic constants, but also more delicate char-
acteristics of a stress state in a composite. Indeed, by virtue of the hypothesis
HI (Sect. 7.3), each inclusion in the composite behaves as an isolated one
in a homogeneous matrix under a constant field iik, k being the number of
the inclusion. From the solution for an isolated ellipsoidal inhomogeneity
(Sect. 4.8), an expression follows for the stress tensor oJ inside each inclusion

(7.6.15)

In the first approximation, one can replace the tensor by its mean value iil,
namely (7.6.10). From here, it is easy to evaluate the stress concentration in
the matrix on the boundaries of inclusions, making use of (4.7.16) [7.6].
Let us apply the method of an effective field to consider themost important
class of inhomogeneous media, namely to one-phase polycrystals. In compar-
 7.6 Random Field of Ellipsoidal Inhomogeneities 187

ison with composites, polycrystals possess a pecularity: one cannot distinguish

explicitly a basic medium (matrix). Let us take as Bo the mean value of B(x)
with respect to the ensemble

B(x) = Bo + BI(x), Bo = <B(x». (7.6.16)

The tensor BI(x) has a constant value inside each crystallite, but changes
from one crystallite to another in accordance with a change of crystallographic
axes.
Following the scope of the method of an effective field, let us assume that
each crystallite k behaves as an isolated ellipsoidal inhomogeneity with the
elastic properties Bo + Blk in a homogeneous medium Bo and an external
field if k' the latter being the sum of an external field if k and a perturbation
caused by all other cristallites. In the first approximation, expressions for
ifl and the tensor of effective compliance B* still have the form of (7.5.18, 17).
Since now the inclusions (crystallites) fill the whole space,

(7.6.17)

where Vk(x) is the characteristic function of the region occupied by the k-th
crystallite. From here, there follow the equalities

<P(x)V(x» = <P(x')V(x')V(x» = <P(a» , (7.6.18)

where, on the right-hand side we have the mean of the stochastic variable
Pea), the latter still being of the form (7.6.12). It is determined by the distribu-
tion function of the orientation of the crystallites anisotropy axes and by the
form of the crystallites.
The properties of the function 7Jf(x - x') = <P(x') Vex, x')lx) coincide,
in the case under consideration, with all the properties of the same function
for composites, and its symmetry depends on the geometry of the random
field of crystallites, which form the polycrystal.
If there exists a linear transformation a, which converts the function 7Jf(x)
into a symmetric one 7Jf(ax) = 7Jf'(lxl), then, as before (7.5.17, 18), emply
the equalities

ifl = [I + D<P(a»]-IO'o, (7.6.19)

B* = Bo - <P(a»[1 + D<P(a»]-I, (7.6.20)

where the tensor D has the form of (7.6.8).

The relation (7.6.15), if one uses ifl in the form of (7.6.19) instead of ifk'
enables one to evaluate the microstress 0'1; in the individual crystallites. It
is easily proved that the values of 0'1; and of the tensor of the effective elastic
compliance B* in the form of (7.6.20) do not depend on the way of decom-
188 7. Elastic Medium with Random Fields of Inhomogeneities

posing the tensor B(x) into the constant Bo and the perturbation B I (x), see
(7.6.16); however, the value ifi depends on this decomposition.
In conclusion, let us dwell on (7.5.24) for the effective elastic compliance,
obtained with the help of the method of an effective medium. We consider,
as above, the case of composite and polycrystalline media. Carrying out the
averaging in (7.5.24) according to the scheme presented in this section, we
have

(7.6.21)

As it was already pointed out, the last relation can be considered an equa-
tion with respect to effective elastic parameters of the medium. The right-
hand side contains components of the tensor B* through the tensors C I *
and A *(a) (Sect. 7.5). In the case of polycrystals, one has to assume additionally,
that v(a)Jvo = 1.
The equation (7.6.21) was investigated in [7.7-10]. Values of the effective
elastic compliance, obtained from its solution, satisfie rather delicate variation-
al extimates of Hashin-Shtrikman [7.11], and, in some cases, coincide well
with experimental data. However, for media, which contain cavities (pores),
the method of an effective medium leads to negative values of the effective
elastic moduli, when the concentration of pores is larger than in the case of
close packing. The method of an effective field has an explicit advantage in
this case: as it was pointed out, (7.6.11) for B* is physically non-contradictorary
up to a concentration of inclusions, which is equal to unity.

7.7 Regular Structures

Let us apply the method of an effective field for calculating the elastic macro-
constants of composites. This problem is of particular interest, since in the
two-dimensional case, for a number of regular structures there are explicit
solutions, which enable us to estimate the error of the method.
Let identical ellipsoidal inclusions form a regular lattice in a homogeneous
matrix. The tensor of effective compliance of the composite is defined as before,
by

(7.7.1)

while averaging of the fields e(x) and o'(x) is performed here with respect to
an elementary cell of the lattice, that is obviously equivalent to averaging
with respect to the whole space. The means of e(x) and o{x) in turn coincide
with the means at a fixed point x under all possible translations of the regular
 7.7 Regular Structures 189

lattice. This means that averaging in (7.7.1) can be understood as averaging

with respect to the ensemble of stochastic functions Vex), which is given by

Vex) = vix + r) , (7.7.2)

where Vf(x) is the characteristic function of the region, which is occupied

by an arbitrary fixed lattice of inclusions, r is a stochastic vector, homogeneously
distributed throughout the space. Note that, when averaging with respect to
the ensemble (7.7.2), the ergodic property is fulfilled automatically. Thus,
the general scheme of the method of an effective field, which was developed
in preceeding sections, can be applied to the case of regular structures without
any changes.
Let us consider (7.6.1, 2). Since all the inclusions are now identical, then
P k = pea) = const, where pea) has the form of (7.6.12), and the tensor a-l
determines a linear transformation, which transforms each inclusion into a
unit ball. The means (7.6.1, 2) become

<P(x)V(x) = pP(a) , (7.7.3)

1lf(x - x') = <P(x')V(x; x')lx) = P(a)<V(x; x')lx) , (7.7.4)

where p is stiJI the inclusions density.

As it is shown in Appendix A4 for regular lattices of ellipsoidal regions
in the space,

<Vex; x')lx) = I;' l(x - x' - m), (7.7.5)

m

where m = maea is the vector of the lattice formed by the centers of inclusions
(Sect. 2.1), the prime attached to the summation sign means excluding the
item with m = 0, and the function lex) is defined by the relation

lex) = { (1 - la-~ xl y(1 + la-l xl) with la- l xl ::;; 2 (7.7.6)

° with la- l xl > 2 .

From (7.7.4, 5), it follows that the function 1lf(x) is the sum of a periodic
function with the mean value pP(a) and a function -l(x)P(a) with bounded
support. The action of the generalized function on such functions is defined
by (7.6.7), from where it follows that

f So(x - x')1lf(x - x')dx' = pQP(a). (7.7.7)

The tensor Q is represented in the form of a converging series

190 7. Elastic Medium with Random Fields of Inhomogeneities

(7.7.8)

where the tensor D is defined by (7.6.8).

Substituting this result into (7.5.17, 18), for the mean value iT l of the effec-
tive field and the tensor of effective compliance B*, we obtain

iT l = [I - pQP(a)]-1 0'0 , (7.7.9)

B* = Bo - pP(a)[I - pQP(a)]-1 . (7.7.10)

Let us consider the example of an isotropic medium, in which spherical

isotropic inclusions form a cubic lattice. The tensor B* becomes

(7.7.11)

where the tensors A and r are defined by

(7.7.12)

(7.7.13)

Here Po and ))0 are the shear modulus and the Poisson ratio of the matrix,
r
the components of the tensor are written in the basis ea of the lattice vectors,
and the tensors Ei are defined in Appendix AI. The scalar function a(p) in
(7.7.11) is illustrated by its graph if Fig. 7.1, which was obtained by numerical
summation of a converging series of integrals.

0.08

0.06

0.04

0.02

OL-____ ~ ____- L_ _ _ _ ~~~ Fig. 7.1. Function a(p)

0.2 P

Let us proceed to the two-dimensional problem. In this case, all construc-

tions are carried out analogously to those of the three-dimensional case.
The expression for the tensor of effective compliance has still the form of
 7.7 Regular Structures 191

(7.6.11) in the case of stochastic fields of inclusions and of (7.7.10) for regular
structures. All tensors in these relations are to be replaced by their two-dimen-
sional analogs. The two-dimensional problem was considered in more details
in [7.12].
Let isotropic circular inhomogeneities form a regular lattice in an isotropic
(two-dimensational) matrix. The following function is the two-dimensional
analog of the function J(x), (7.7.6),

J(x) = f tan-1(~~I ~ 1 - t~) - ~~~ ~1 - (t~;2 with Ixl ~ 2To

1 0 with Ixl > 2To (7.7.14)

where TO is the radius of an inclusion.

Let us consider a regular triangular lattice of inclusions. One can show
[7.12] that, in this case, B* has the form

(7.7.15)

which coincides with the expression for B* in the case of an isotropic field of
inclusions, see (7.6.13). Then the tensor A for a two-dimensional stress state
if defined by

A = 16~0 [8E1 - (1 + ],)0) (E2 + 2E1)] , (7.7.16)

where E1, E2 are the basis tensors, defined in Appendix A4, but for the two-
dimensional case.
If inclusions form a square lattice, then the tensor B* is defined by a rela-
tion of the form of (7.7.11), where A is the tensor (7.7.16), and

(7.7.17)

Here the Greek indexes take values 1, 2.

The function a(p) in (7.7.11), for the two-dimensional case, was also found
numerically, and a graph has been presented in [7.12].
In [7.13] the problem of the stress state of plane lattices of inclusions in a
hbmogeneous external field was solved with the help of a series expansion in
bi-periodic functions. The values of the effective elastice constants (of different
lattices) presented there can be considered exact ones, since in the correspond-
ing series a sufficient number of terms was conserved.
In Figs. 7.2 and 7.3, a comparison is given of exact values of effective
elastic moduli with those obtained with the help of the method of the effective
field for a regular triangular lattice. Here Eo and fl.o are the Young modulus
and the shear modulus of the matrix, Eo + El and fl.o + fl.l are the same quan-
192 7. Elastic Medium with Random Fields ofInhomogeneities

+t
6 6
~ ~+E1,'
I
5 !..*
Eo ~ -.
"-0 5 l!...*
fto ~'.
~+E1=O
td
"

kt
4 4
b 10.1 b
3 3 I

2 2

0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0.2 0.4 0.6 0.8 b/a 0.2 0.4 0.6 0.8 b/a
Fig. 7.2. Comparison of exact and effective Fig. 7.3. Comparison of exact and effective
Young's moduli shear moduli

tities for inclusions, E*, and f1.* are the values of the effective elastic modulus.
The solid curves represent the exact solutions, the dotted line with small circles
corresponds to (7.7.15)
For this case, let us consider the error !1 which arises when evaluating ef-
fective constants with the help of the method of the effective field

(7.7.18)

Here EI and E* are the exact values of the effective Young modulus and the one
calculated from (7.7.15), respectively. Calculations show that !1 is a function
of the dimensionless parameter

(7.7.19)

where Co is a characteristic value of the elastic modulus of the matrix, Cl is the

same for the perturbation of the modulus in inclusions, r is a dimensionless
scalar parameter, which depends on the form of inclusions and is equal to a
characteristic value of components of the tensor ACo. For circular inclusions,
r = 0.66. The dependence !1(X) is exhibited in Fig. 7.4, where the scale on the
absciss axis is In (2 + X).
The parameter X can vary between the limits - 1 :s; X :s; 00, while to ne-
gative values of X there correspond inclusions more compliant than the matrix,
and to positive X the opposite. As it is seen from Fig. 7.4, an error higher than
10% arises in the regions -1 < X < -0.8 and 4 < X < 6. The continuous
part of the CUlve in the Fig. 7.4 is obtained with the help of the results of [7.13].
 7.8 Fields of Elliptic Cracks 193

Fig. 7.4. Function t:J(X)

The dotted line is an extrapolated value of the error. The fact that, as p -+ I,
(7.7.15) gives a correct result for all Ch is taken into account.
The relative error for the effective shear modulus practically coincides with
A(X) for all values of the parameter X.
A comparison of the effective field approximation with exact solutions for a
square lattice of circular inclusions was carried out in [7.12]. The form of the
error of calculating the effective elastic constants is here the same as in the case
of a triangular lattice.

7.8 Fields of Elliptic Cracks

Let us proceed to a consideration of a random field of elliptic cracks in a homo-

geneous elastic medium. In what follows, it is assumed, for simplicity, that all
cracks are opened, when the medium is loaded. This imposes certain restrictions
on the external field 110.
The expressions (7.5.17, 18) for the tensors ifl and B*, in the case under
consideration, have the form

(7.8.1)

B* = Bo - (P(x)Q(x)A, (7.8.2)

where

A = [/ - JSo(x - x') (P(x')Q(x; x')lx)dx']-l . (7.8.3)

Here the functions Q(x) and Q(x; x') are defined by (7.4.17, 19), the continuous
tensor field P(x) coincides with P,,(x), (7.3.9), on the surfaces of the cracks.
To construct the means in (7.8.2,8) let us use the same procedure as in the
determination of the means (7.6.1, 2) in the case of ellipsoidal inclusions. Using
194 7. Elastic Medium with Random Fields ofInhomogeneities

the ergodic property with the subsequent averaging with respect to the ensemble,
we obtain

(P(x)Q(x) = <lim .l JP"hk(X)Qk(X)dx) = 231t <~ pea, b), (7.8.4)

v->oo V G Vo

where Qix) is a generalized function concentrated on the crack surface Qk,

hk(x) has the form of (7.3.10), the tensors P k are given by (7.3.11), Vo is as before
the mean volume per crack. On the right-hand side of this relation, there ap-
pears the mean of the stochastic tensor a3pea, b), where pea, b) has the same form
as P k , see (7.3.11), and is determined by the stochastic values a and b(a ~ b) of
the semiaxes of an elliptic crack and by the orientation n of the plane of the latter.
Let us consider the conditional mean in (7.8.3)

1ff(x - x') = (P(x')Q(x; x')lx) . (7.8.5)

The constant components (1ff) of the functions 1ff(x) and 1ff(x - x') =
(P(x')Q(x')lx) coincide, since their difference is a function, with bounded sup-
port, if the dimensions of the cracks are bound (Appendix A4). By analogy
with (7.6.5), we have

(1ff) = (Q~x) <!~ V\ bdx' bP(x')Q(x')Q(X)dx)

(7.8.6)
= 3t~) <~: P(a, b).

Here, the relations

(P(x')Q(x')lx) = (P(x')Q(x')Q(x) (7.8.7)

(Q(x) ,

(Q(x) = 1t~b) , (7.8.8)

are taken into account, the latter being easily obtained, as (7.8.4) was. For cracks
of identical dimensions

21t a 3 (7.8.9)
(1ff) = -3 ~ (P(a, b),
Vo

where the mean is calculated only with respect to cracks orientations.

As in the case of ellipsoidal inclusions, the function 1ff(x) for homogeneous
fields of cracks is represented in the form of a sum of a periodic (generally
speaking, generalized) or continuous oscillating function with the mean value
equal to (1ff) and of a function with bounded support. At the same time the
integral in (7.8.3) admits regularization, which is analogous to (7.6.7)
 7.8 Fields of Elliptic Cracks 195

JSo(x - x')?P'(x - x') dx' = JSo(x)[?P'(x) - <?P')] dx

= D[?P'(O) - <?P')] + f So(ax)[?P'(ax) - <?P')] det a dx , (7.8.10)

where D is defined by (7.6.8), the tensor a being an arbitrary nondegenerate

linear transformation of coordinates. The function ?P'(x) is assumed to be con-
tinuous at the origin; this is true for most of the interesting distributions of
cracks.
Let us consider some specific examples of stochastic fields of cracks.
1) The Possionfield of cracks. Let, in a bounded volume v of the three-
space, there be N cracks, dimensions and orientations of the latter being stoch-
astic quantities with known distribution functions. the position of each crack
does not depend on positions of others and is homogeneously distributed in v.
Letting v and N tend to infinity so that (v/N) = Vo < 00, we obtain a homo-
geneous field of cracks, which will be called the Poisson field. In this case
(Appendix A4)

?P'(x) = <P(x')Q(x; x')lx) = <?P'), (7.8.11)

i.e. ?P'(x) = const. By virtue of the definition (7.2.6) of the action of the operator
So on constants

JSo(x - x')<?P')dx' = 0, (7.8.12)

the expression (7.8.2) for B* becomes

(7.8.13)

For circular cracks of the stochastic radius a in an isotropic medium we

then have

(7.8.14)

where the tensors Ei are defined by (AI) and the averaging of these tensors
is carried out with respect to all possible orientations of cracks. If the distribu-
tion with respect to orientations is homogeneous, then

<a3 )
B* = Bo - - 8 (1 - ].10) --[].loB 2
- 2(5 - ].I0)E1], (7.8.15)
45 llo(2 - ].10) Vo

and, in particular, for the effective shear modulus Il* and the Poisson coeffi-
cient ].1*, we have
196 7. Elastic Medium with Random Fields oflnhomogeneities

_
fl* - flo
(1 + 32 (a 3 ) (1 - ).10) (5 - ).10)
45 -v;;- (2 - ).10)
)-1 , (7.8.16)

).1* = ).10 fl* (1 + ~ (a 3 ) (1 - ).I~») (7.8.17)

1 + ).1* 1 +).10 flo 45 Vo (2 - ).10) .

2) The model with a restriction, concerning the intersections of cracks.

Let, around each crack, there be a neighborhood with a small probability of
containing other cracks. This means that lff(x) ~ 0 in a neighborhood of the
origin analogously to the case of ellipsoidal inclusions. If, moreover, lff(x) is
spherically symmetric, then the integral (7.8.10) takes the extremely simple form

JSo(x - x')lff(x - x') dx' = - D(lff) . (7.8.18)

At the same time, the tensor B*, due to (7.8.2-4), is

(7.8.19)

Note that the expression for B* will have the same form, if there exists a
linear transformation a, which makes lff(x) a spherically symmetric function
lff'(lxD = lff(ax). Then the tensor D depends on a and is defined by the relation
(7.6.8).
In the case of a spherically symmetric lff and of a homogeneous distribution
with respect to orientations, B* is an isotropic tensor

(7.8.20)

where

{31 = ~ (a 3 ) (7 - 5).10) (5 - ).10)

675 Vo (2 - ).10) ,
(7.8.21)
{32 = ~ (a 3 ) (5 - ).10 + 10).10).
675 Vo 2 - ).10

Here all cracks are circular and the radius a is stochastic.

3) Regular lattices of cracks in the three-space. Let us first consider a simple
lattice, in which all cracks have identical dimensions and orientations. In this
case the means (7.8.4, 5) become

21t a3 (7.8.22)
(P(x)Q(x» = (lff) = -3 - P(a, b) ,
Vo
 7.8 Fields of Elliptic Cracks 197

1/f(x - x') = P(a, b) (h(x')Q(x; x')lx) , (7.8.23)

where P(a, b) is a constant tensor of the form of (7.3.11) and the function h(x)
has the form of (7.3.10) at each crack. In these relations, averaging is carried
out with respect to all translations, i.e. analogously to the case of regular
lattices of inclusions (Section 7.7).
As it is shown in AppendixA4, the mean (7.8.23) for regular lattices of elliptic
cracks is represented by

1/f(x) = ~b (1/f) I:' J(x - m)Q;"(x) , (7.8.24)

1ta m

where m is a vector of the lattice formed by the centers of the cracks, the primed
summation sign means excluding the term with m = 0, Q;"(x) is the a-function
concentrated on a flat surface with its center at the point m and parallel to
all cracks. The form of the surface is an ellips with semiaxes, which are equal to
doubled semiaxes of the crack under consideration. The function J(x) is defined
by the relation (A 4.47)

3 J1 (~- e) .Je(2~ - e)' + ~{ sin-1 ( 1 - ~) + ~ ]

J(x) = 2VO S .J 2 ~d~ with e~ 2
1t )le 1- ~
l2 0 with e> 2 (7.8.25)

where

(7.8.26)

and the xL and x 2-axes are directed along the principal axes of the elliptic cracks.
From (7.8.10) we can now find the result of the action of the generalized
function So(x) on 1/f(x) in the form of (7.8.24)

S So(x - x')1/f(x - x') dx' = Q( 1/f) . (7.8.27)

Here, the tensor Q has the form of a converging series of integrals

(7.8.28)

where D is defined by (7.6.8) with a«~ = aa~. Substituting (7.8.27) into (7.8.1,
2) for the tensors 0"1 and B*, we obtain

(7.8.29)
198 7. Elastic Medium with Random Fields of Inhomogeneities

B* = Bo - (W) (J - Q (W) )-1 , (7.8.30)

where (W) has the form of (7.8.22). Note that, for simple lattices, iiI is the
real external field o{x) averaged over the surface of any crack.
If cracks form a complex lattice, then the expression for W(x) consists of
sums of the type of (7.8.24) over each sublattices. In this case, one can make
the method of solution more precise by assuming that the value iiI of the
effective field is different for different lattices. Equations for the set of quan-
tities iii (the values of iiI on the k-th sublattice) follow from an analog of
(7.5.6) for the case of cracks and become
n
iii + 2: Q}kl (Wi) iii = 0"0, k = 1, 2, ... , n , (7.8.31)
i=1

where

Q,(kl = - D + ~ f So(X) [ 1t~ibj J(x - mi - ~ik) Q~i(X - ~ik)-1 ] dx,

(7.8.32)

and, with i = k, the expression for Q)!'l has the form of (7.8.28), where the
quantities a, band m concern the k-th sublattice only. In these relations, m..
is a vector of the i-th sublattice, ai, bi are semi axes of elliptic cracks in the
i-th sublattice, ~ik is a vector, which connects knots of i-th and k-th sublattices
in an elementary cell, the tensor (Wi) is determined by the orientation and
the dimensions of cracks in the i-th sublattice and has the form, analogous
to (7.8.22).
Solving the system (7.8.31) with respect to the tensors ii} and substituting
the result into the expressions for the mean values of stress and strain, analogous
to (7.5.14), using then the definitions (7.5.16) for the tensor B* we obtain
n
B* = Bo - 2:
i=1
(Wi) Ai , (7.8.33)

where the tensors Ai connect ii} with the external field 0"0

(7.8.34)

~nd are determined from the system (7.8.31).

Since we have assumed that each crack behaves as an isolated one in the
field iiI, then, in the scope of our method, one can consider a value of the
averaged stress intensity factor on the crack contour (for the definition of the
stress intensity factors see Sect. 7.8.11). To achieve this, one has to use the
solution of the problem for an isolated crack in a constant external stress field
iiI (Sect. 7.8.12).
 7.8 Fields of Elliptic Cracks 199

Let us consider the example of an isotropic medium, which contains circular

cracks of identical orientations, the cracks forming a lattice with three or-
thogonal planes of symmetry. In the case of cracks with identical orientations,
the openings of the cracks are uniquely determined by the convolution of the
tensor jjl with the normal to the plane of the cracks. It can be shown that for
the considered lattices jjln has the form

(J-lafJ nfJ = (1 _1 a) (Sla

UfJ + 1 _ a(3 _ {3 n anfJ) (Jon
fly
v, (7.8.35)

the tensor B* is determined by the relation

I ~[2E5(n) + (2{3(2-v o) - vo) E6(n)],

(I-a) Vo l-a-{3
(7.8.36)

where

a = (2 -4 vo) [P./"0 - (2 - Vo)ao,

] P. _ 4(vo - 3) p.
/" - (2 _ vo) /,,0' (7.8.37)

and the radius of all cracks is a. The dimensionless scalar coefficients ao and
(30 depend only on the geometry of the lattice and are presented in the form
of converging series of the type of (7.8.28). Fig. 7.5, displays the graphs of

0.20

0.15

0.10

0.05

Fig. 7.S. Functions a.(q) and ~.(q)

ao and {30 as functions of the parameter q = (a3)lvo for a regular cubic lattice
of cracks, all planes of the cracks being parallel to a fixed face of the cube.
The stress intensity factors on the contour of a crack are found from the
solution of the problem for a circular isolated crack in the constant field jji
of the form of (7.8.35). In the case of pure tension in the direction of the normal
200 7. Elastic Medium with Random Fields of Inhomogeneities

to the plane of cracks, lTff,B = lTonanf3, and the stress intensity factor K J has
the form [7.8]

(7.8.38)

This expression is to be considered as the mean stress intensity factor of the

crack.
Let us consider a lattice of cracks located in one plane. This case can be
obtained from the spatial one, if one of the parameters of the lattice tends to
infinity. The problem of a square lattice of circular cracks on a plane was
solved in [7.14]. In the Fig. 7.6, the value of the parameter

1.2

Fig. 7.6. Function kI(A)

(7.8.39)

calculated with the help of the method of effective field (dotted curve) is com-
pared with the maximun value of kJ on the edge of the cracks from [7.14]
(continuous curve).
In conclusion, let us consider the expression for the tensor of the effective
elastic compliance of a medium with cracks, to which the method of effective
medium leads. Eq. (7.5.26) for the effective compliance is represented by

(7.8.40)

where P * is determined from the solution for an elliptical crack in the medium
with the tensor B* of elastic compliance and has the form, which is analogous
to that of Ph in (7.3.8) (for more details, see the end of Sect. 7.5).
 7.9 Two-Dimensional Systems of Rectilinear Cuts 201

In the case of an isotropic medium, which contains a homogeneous field

of circular cracks with homogeneously distributed orientations, the effective
medium is also macroisotropic, and (7.8.40) becomes

B* = Bo - 2'Jt<
3 Vo a3 ) p * , where (7.8.41)

P * -_ 4 (1 - v*) [ E2 2(5 ) E1] (7.8.42)

- -15
'Jt f1.* (2 - v*) v* - - v* .

From here, the relations follow

(7.8.43)

(7.8.44)

The latter can be considered as an equation with respect to v*.

The authors of [7.15] who obtained the result, pointed out that, with
<a3)/vo~ 9/16, the value of the shear modulus f1.* calculated according to (7.8.43,
44) becomes nonpositive and the formulae lose their physical sense. For com-
parison, note that (7.8.16,17), to which the method of an effective field leads
in the case of a Poisson set of cracks, are free of this defect: here the effective
moduli are always positive. For the model with a restriction on cracks inter-
sections [Eq. (7.8.20) for B*l, the effective moduli vanish for q = <a3)/vo much
larger than 906 (for q ~ 2).
In the next section a comparison will be carried out of expressions for
effective constants with the results of experiments and the exact solutions in
the two-dimensional case; this will enable us to estimate the limits of ap-
plicability of both methods of self-consistency.

7.9 Two-Dimensional Systems of Rectilinear Cuts

In the two-dimensional case, a rectilinear cut is the analog of an elliptic crack.

The formalism of the method of effective field is transfered without any
important changes to the two-dimensional case. In this section, we shall present
mostly the final results, outlining only the derivation. The planar problem
is interesting due to the fact that there exist a number of exact solutions and
of experimental observations, which enable one to find out the limits of ap-
plicability of different approximate approaches.
Let us start with an individual rectilinear cut. The a-function, concentrated
on L and defined by (6.1.2) will be denoted by L(x)
202 7. Elastic Medium with Random Fields of Inhomogeneities

a(L) = L(x). (7.9.1)

When applying the stress if, one can interpret the cut at infinity as a line on
which dislocation moments of some density M(x 1) are induced. In the case of
a constant external field, from the solution of the problem about an isolated
rectilinear cut, it follows that

M(x l ) = P(n)if -//2 - (X 1)2, (7.9.2)

where the xLaxis of the cartesian coordinate system with the origin at the
center of the cut is directed along L; I is the half length of the cut. The constant
fourth order tensor P has the form

P a{J).p = - n (aTji/().ntl) , (7.9.3)

(7.9.4)

Here SoCk) is the k-representation of the Green's function for stress in the
two-dimensional problem, n is the normal to L, JI(k l ) is the Bessel function
of the first order, the Greek subscripts in this section take values 1, 2. For
an isotropic medium and for a plane state of stress

(7.9.5)

Let us proceed to the case of a set L of rectilinear cuts on a plane and

introduce the following notations. Let L(x) be now the a-function, concentrated
on all cuts, and the function L(x; x') be defined by the relation

with x E L
(7.9.6)
with x E Li ,

where Lk(x) is the a-function, concentrated on the k-th cut L k.

In the case of a system of cracks in a plane, an analog of (7.3.17) for the
effective field if(X) has the form

if(X) = 110 + JSo(x - x')P(x')if(x')L(x; x') dx', x E L , (7.9.7)

where the continuous function P(X)if(X) = M(x) has the form of (7.9.2) on
each of the cuts L k. The expressions for the tensors l1(x) and c(x) are represented
by
 7.9 Two-Dimensional Systems of Rectilinear Cuts 203

q(x) = qo + JSo(x - x')P(x')q(x')L(x') dx' , (7.9.8)

e(x) = eo - JKo(x - x')CoP(x')q(x')L(x') dx' . (7.9.9)

From (7.9.7), using hypotheses HI and H2 of the method of an effective

field (Sect. 7.3), analogously to (7.5.17, 18), we have

(7.9.10)

B* = Bo - (P(x)L(x»A, (7.9.11)

where

A = [I - JSo(x - x')(P(x')L(x; x')lx)J 1 • (7.9.12)

The mean in (7.9.11) is represented by

(P(x)L(x» = i (:02p), (7.9.13)

where c:vo is an average area per crack, the tensor P has the form of (7.9.3) and
depends on a stochastic orientation n of the crack. The mean in the integrand
in (7.9.12)

UJ(x - x') = (P(x')L(x; x')lx) (7.9.14)

possesses all the properties of the mean (7.8.5) of the three-dimensional case.
The constant component of the function UJ(x) is equal to

(UJ) = ~ (/3P) ~
2c:vo ( I ) 2
1..(1tc:vo12) (P), (7.9.15)

while the last equality occurs if the dimensions of all the cracks are approxi-
mately identical.
A regularization of the integral in(7.9.12)is defined by a relation analogous
to (7.8.10). In the two-dimensional case, the tensor D becomes (Appendix A3)

D = 2~fSo(k) dL, (7.9.16)

where the integration is carried out over the unit circle in the two-dimensional
k-space. For an isotropic medium, the tensor D takes the form
204 7. Elastic Medium with Random Fields of Inhomogeneities

D = _ lio(1 + ))0) (E2 + 2El) . (7.9.17)

4

Here El and E2 are the two-dimensional analogs of the basis tensors, defined
by (AI).
Let us proceed to consider specific random fields of cracks on the plane.
1) A Poisson field of rectilinear cuts is defined analogously to a Poisson field
of elliptic cracks in the three-dimensional case (Sect. 7.8), and here

<W(x» = <W) = const. (7.9.18)

Then the integral in (7.9.12) vanishes, and (7.9.10,11) become

0'1 = (To, (7.9.19)

B* = Bo - -1
2
< 12
'It - P ) .
-
(vo
(7.9.20)

In the case of an isotropic medium

(7.9.21)

If the distribution of the cracks in orientations is homogeneous and the dis-

tribution of dimensions does not depend on orientation, then

(7.9.22)

where q is a dimensionless parameter of the type of a concentration

_ <'It /2)
q---. (7.9.23)
(VO

This implies

(7.9.24)

where Eo and E* are the Young moduli of the real and effective media, respec-
tively.
2) A model with a restriction on the intersections of cracks (Sect. 7.8). In
this case, the mean of W(x) defined by (7.9.14) is a continuous function, which
is equal to zero at x = O. If W(x) is also spherically symmetric, thus

sSo(x - x')W(x - x')dx' = - D <W) , (7.9.25)

 7.9 Two-Dimensional Systems of Rectilinear Cuts 205

<
where D has the form of (7.9.16), and 1Jf) has the form of (7.9.15). In this case,
the tensors 0'1 and B* are represented by

0'1 = (/ - D<IJf) )-10"0 , (7.9.26)

B* = Bo - <1Jf) (l - D< 1Jf) )-1 . (7.9.27)

Let all the cracks be circular, their distribution in orientations be homo-

geneous, and the matrix be isotropic. The preceeding relations imply the equali-
ties

(7.9.28)

B* = Bo + 2 (1
Po
1
+ Vo) (1 - Tq
1 )q( 1 - 41)q [E1 + ~ q(E2 - E1) l
(7.9.29)

Hence, for the relative elastic modulus E*/ Eo and the Poisson coefficient v*/vo,
we have

(7.9.30)

(7.9.31)

where q has the form of (7.9.20).

In Fig. 7.7, the dependences (7.9.24) (curve 1), (7.9.30) (curve 2) and (7.9.31)

Fig. 7.7. Comparison of experimental

and calculated elastic moduli
206 7. Elastic Medium with Random Fields ofInhomogeneities

(curve 3) are compared with experimental data [7.16]. The experiments were
carried out using thin rubber sheets, containing random fields of rectilinear
cuts (vo = 0.5). Results of the experiments are exhibited in Fig. 7.7, by the dot
line with small circles (E*/ Eo) and by the dot-and-dash line (v*/vo).
The statistical analysis shows that, to the field of cracks investigated in [7.16],
there corresponds the model with the restriction on cracks intersection. The
experimental data are described by (7.9.30, 31), obtained for this model, in the
best way. Therefore one can state that the approximation of an effective field
enables one to take into account correctly interaction between cracks in the
situation under consideration.
Let us present here expressions for E* and v*, which are given by the meth-
od of an effective medium in the case of an isotropic tensor of effective elastic
compliance. The analog of (7.8.41) in the two-dimensional case has the form

(7.9.32)

Thus, we have

E* = ~ = 1 _ q. (7.9.33)
Eo Vo

To this dependence, there corresponds the straight line 4 in Fig. 7.7.

3) Regular lattices of cracks in the plane. In the case of cracks with identical
orientations, the expression for l/f(x) becomes

l/f(x) = 21, 1:'

m
l(x - m)L;"(x) , (7.9.34)

where m is a vector of the lattice formed by the centers of cracks, the primed
summation sign means the exclusion of the term with m = 0, I is half the length
of each crack, L;"(x) is the a-function concentrated on the segment of the length
41 with its center in the m-th knot and parallel to all the cracks. The function
lex) has the form

lex) =
2~ {(I - M) /1&(2 - 1&) + sin- (1-1&)
I Y ' I I
+~
2
1 if Ixl :::;; 21
11: 0 if Ixl > 21
(7.9.35)

The action of the generalized function So(x) on l/f(x) in the fOlm of (7.9.34)
is represented as in (7.8.27),

sSo(x - x') l/f(x - x')dx' = Q< l/f) , (7.9.36)

 7.9 Two-Dimensional Systems of Rectilinear Cuts 207

where the tensor Q is represented in the form of a converging series

Q= - D + ~'f SO(x)[l1 J(x - m)L~(x) - IJdx. (7.9.37)

Here D has the form of(7.9.16) and the mean value of1ff(x) is

7C1 2
(1ff) = -2-P(n). (7.9.38)
000

Expressions for the tensors q1 and B* have in this case the form which is an-
alogous to (7.8.29, 30).
Let us consider a lattice of cracks of one orientation, which has two or-
thogonal symmetry axes. The expression for the vector q 1n (which, here, deter-
mines the state of all cracks) has the form

-Up
0' np = (1 _1 (3) Up (l:.t + 1_ a .l)
a _ {3 n np e..tnv •
0'0 (7.9.39)

The expression for the tensor of effective elastic compliance B* in terms of the
basis tensors Ei defined in Appendix AI, is

(7.9.40)

The dimensionless coefficients a and {3 have the form of converging double

series, and their explicit expressions were presented in [7.17].
The stress intensity factors K J and Kn on cracks (the first is under tension
along the direction of the normal, the second is under pure shear) are de-
termined by

(7.9.41)

where 0'0 and 't"o are the values of external tensile stress and shear stress, re-
spectively,

1
kJ = 1 - a - (3 , k n = 1 - a (7.9.42)

Let us consider a few particular cases.

a) Regular triangular lattice of cracks. Fig. 7.8, illustrates E*/Eo (curve 1)
and ))*/))0 (curve 2) as functions of the parameter.it = 21/p of the lattice, where
p is the distance between the centers of cracks, these functions being obtained
by the method of an effective field. The curves for the intensity factors kJ and k n
are shown in Fig. 7.9. The continuous curves in this figure correspond to the
exact solutions obtained in [7.18].
208 7. Elastic Medium with Random Fields of Inhomogeneities

1.0 2.0,--------------,0--,

1.5

OL-----~~------~

Fig. 7.8. Elastic moduli for triangle Fig. 7.9. Stress intensity factors for
lattice of cracks triangle lattice of cracks

b) A system of collinear cracks on one straight line. The values of the co-
efficients a and (3 in (7.9.42), which correspond to this case, can be obtained from
the solution for a rectangular lattice when one of the characteristic dimensions
of the latter tends to infinity. Then a and (3 are expressed as simple series [7.17].
The value of the coefficient kJ is in agreement with its exact value, which has
the form [7.19]

2 sin ~ A (7.9.43)
k9=-~~=~
.y' ATe sin TeA

(c) A row of parallel cracks. In Fig. 7.10, the values of the intensity factors
kI and kIl' obtained by the method of an effective field (dotted line with
small circles), are compared with results of numerical solution of the corres-
ponding integral equations [7.20] (solid curves).

kll k J
2.0 1.0
1.8
1.6
1.4
1.2 Fig. 7.10. Stress intensity factors for a row of
parallel cracks
0

As it can be seen from the above examples, in the two-dimensional case, the
first approximation of the method of an effective field gives good agreement
 7.10 Random Field of Point Defects 209

with the results of more precise calculations of elastic constants and of in-
tensity factors, when the lengths of the cracks are smaller than the distances
between their centers. This statement appears to remain valid for other types of
regular systems of cracks as well.
When the centers of cracks in a lattice become closer (see case c), the ex-
ternal field q(x), in which each crack is contained, begins to differ more and
more from a constant one. Therefore, the error which is given here by the
method of the effective field, is connected with violating the hypothesis HI
(Sect. 7.3). However, a way of making the computational part of the method
more precise can be suggested.
Let the effective field q(x) be not constant and have the form of a polynomial.
The solution for an isolated inhomogeneity can be found, using results of
Sect. 4.9, 12. This will enable one to express the corresponding induced density
m(x) of dislocation moments on defects in terms of the coefficients of the
polynomial q(x).
Equations for these coefficients can be obtained, substituting q(x) and ex-
plicit expressions for m(x) into (7.3.3) for the effective field and averaging a
necessary number of times the expression obtained, over the region occupied
by defects. As weights for the averaging, a system of linearly independent func-
tions (polynomials, for example) is to be chosen. An analogous modification of
the method can be used also in the case of stochastic fields of ellipsoidal in-
clusions and cracks.

7.10 Random Field of Point Defects

In the previous sections the effective field method was applied only to the
determination of the first moments of stress-strain fields in composite materials.
The calculation of the second moments is technically a more complicated
problem. Finding the averages in (7.5.9, 13), determined by the second mo-
ments of q(x), and the solution of the system of these equations is only possible
by using numerical methods, even for simple cases. However, many technical
difficulties disappear if we pass to the model of point inhomogeneities, where
inclusions of finite sizes are replaced by isolated point defects.
In this and the following section the general scheme of the effective field
method is applied to the description of the elastic medium containing a homo-
geneous random field of point defects. Note that in this case the application
of the method has a better foundation than for inhomogeneities of finite
sizes. As a matter of fact the main hypothesis HI (Sect. 7.3) about the constancy
of the effective field q(x) in the region occupied by an arbitrary inclusion,
turns into a trivial fact in case of point defects.
Let us begin with the problem of replacing an isolated finite inclusion by
a point defect. The basis for such a substitution is the following convention.
210 7. Elastic Medium with Random Fields of Inhomogeneities

The perturbed fields of the isolated inclusion and of the corresponding point
defect should have the same asymptotics at infinity. Thus in order to pass to
the point-defects model from the model of the composite with inclusions of
finite sizes, we should substitute the functions PkVk(X) and Pk(X)Ok(X) in the
previous relations by their first terms in the multipole series expansions

(7.10.1)

where ~k is the center of the region Vk • Note that PkVlx) characterizes the
density of dislocation moments induced in the region V k by applying an ex-
ternal field. The tensor P2 in (7.10.1) has the form

(7.10.2)

where Vk is the volume of the region V k ; (7.3.6) being used here.

In case of cracks we deal with the function PiX)Ok(X) which is also re-
presented in the form of (7.10.1), where

(7.10.3)

Here Pk(x) and P k are defined by (7.3.9, 11), ak is the larger half-axis of the
ellips Ok'
Let X be a set of points ~k with the point defects specified by the set Xxo
and defined by

X
xo
= l x=
U ~i
i*k
U~i
'
by Xo E X
(7.10.4)

by the analogy with V Xo ' see (7.3.14). We consider the generalized functions

X(x) = ~ o(x - ~i) , (7.10.5)

ei EX

X(x; x') = ~ o(x' - ~i)' (7.10.6)

ei EX•

concentrated on the sets X and Xx, respectively.

By the transition to the point defects the governing equation (7.3.17 or 19)
for the effective field O'(x) is reduced to

O'(x) = 0"0 + SSo(x - x') PO(x')O'(x') X(x; x') dx', x EX, (7.10.7)
 7.10 Random Field of Point Defects 211

where Po(x) is a continuous function coinciding with P2, see (7.10.2, 3), at the
point x = ~k'
For the determination of the effective field's first moment 6'1 both parts
of (7.10.7) should be averaged with the convention x E X. Then the hypothesis
H2 (Sect. 7.3) about the statistical independence of O'(x) from PO(x')X(x; x')
at the point x together with an assumption similar to (7.5.5), should be used.
As a result the equation for 0'1 is presented by

0'1 = (To + f So(x - x') (PO(x')X(x; x')lx)dx'O'l, (7.10.8)

where the operation of the average with the conventions x E X and Xl> •.. ,
Xn E Xx, is denoted by (Ix; Xl, ... , x n ), see also (Sect. 7.4).
In case of the point-defects model, (7.5.17) for the effective compliance
tensor B* is reduced to

B* = Bo - (PO(x)X(x» A , (7.10.9)

where the tensor A, defined by (7.5.19) in case of inclusions of finite size, is

represented in the form

A= [I - f So(x - x') (PO(x')X(x; x')lx>dx'J 1 , (7.10.10)

which is a consequence of (7.10.8).

As above, we assume that random sizes and elastic moduli of inclusions
do not depend on their position in space. Thus the functions PO(x) and X(x)
are statistically independent and consequently

(PO(x)X(x» = P(X(x» , (7.10.11)

(PO(x')X(x; x')lx) = P(X(x; x')lx) , (7.10.12)

where P = (PO) is the average over the ensemble of random variables (7.10.2
or 3).
Let us consider two stochastic point-detects models in space (a Poisson
field and a random point lattice) and determine the tensors 6'1 and B* in these
cases.
1) The Poisson point field. An example is the Poisson field of the set of
the centers of the cracks described in Sect. 7.8.
In case of the Poisson field the averages on the right-hand sides of (7.10.11,
12) have the following forms, see (A4.6, 17),

(X(x» = (X(x; x')lx) = ~ = const. (7.10.13)

Vo
212 7. Elastic Medium with Random Fields of Inhomogeneities

As a consequence of this relation the integral in (7.10.10) vanishes and

when using (7.10.9) we obtain

1 (7.10.14)
B* = Bo - -P.
Vo

If the point-defect model describes a random field of cracks and therefore

P has the form of (7.10.3), the last relation coincides with (7.8.19) which was
presented in the case of the finite crack field in space.
As it was emphasized in Sect. 4.8 and 4.13, caution is necessary in employing
the point-defect model within the framework of the local theory of elasticity.
Strictly speaking, if inclusions are replaced by the point defects, the character-
istic length 1 is introduced in the continuous medium. The length 1 is approxi-
mately equal to the mean size of the defects and if the distances between the
neighboring point defects become smaller than I, phenomena show up, which
have no analogy in the case of finite inhomogeneities (Sect. 7.11). It turns
out that the point-detects model cannot describe the interaction between
finite inclusions for such distances even qualitatively.
On the basis of such considerations let us complete the model of the Poisson
point field. It will be assumed hereafter that there is a neighborhood (spherical,
for example) with the diameter ~ 1 surrounding every point, such that the
other points of the field cannot enter it.
For this new model the properties of the average W(x - x') =
<X(x; x')/x) do not differ from the properties of <V(x; x')/x) where V(x) is the
characteristic function of the random field of finite regions in space. In this
case W(x) is a continuous function which is equal to zero at x = O. The asymp-
totic value of this function at infinity is V-I.
Because this model is isotropic, W(x) has a spherical symmetry if the density
of defects is not too large. As a consequence of these properties of W(x) , the
integral in (7.1 0.10) is represented in the form (Sect. 7.6)

sSo(x - x')P<X(x;x')/x) dx' = - .l

Vo
DP, (7.10.15)

where the tensor D is defined by (7.6.5) and aa(3 = oa(3. This relation together
with (7.10.9) yields

B* = Bo - -1 P ( I
Vo
-1 DP
+ Vo )-1 . (7.10.16)

It coincides with (7.6.11) in the case of inclusions and with (7.8.19) in the
case of cracks.
Thus, the point-defects model with the necessary restrictions yields the
same expressions for the effective constants tensors as the model for finite
sizes of inhomogeneities.
 7.10 Random Field of Point Defects 213

2) Random lattice of point defects. Let point defects constitute a random

lattice in space. The set of the lattice nodes are described by the random vector
~m

~m = m + Pm + r, (7.10.17)

where m is the vector of the fixed deterministic lattice, Pm are independent

random vectors with the same density functions f(x) and zero expectations,
r is a random vector the same for all m and uniformly distributed in the whole
space. Obviously, if f(x) = o(x) , then (7.1O.17) defines a deterministic lattice
to within random translations in space. If all P~ are distributed uniformly in
space we get a Poisson point field.
The averages (7.10.11, 12) for random point fields are considered in Ap-
pendix A4. The relations (7.7.8, 29) yield

(X(x) = v10 ' (7.10.18)

lff(x - x') = (X(x; x')lx) = 2:' g(x - x' - m) , (7.10.19)

m

where the primed summation sign means the omission of the term m = O.
The functions g(x) and f(x) are connected with each other by

g(x) = Sf(x - x')f( -x') dx' . (7.10.20)

The integral entering (7.10.10) for B* has the form

S So(x - x') (PO(x') X(x; x')lx) dx'

= .l S So(x - x') [volff(x) - 1] dx P. (7.10.21)

Vo

Note that (lff) = l/v and this equality may be proved as (7.6.5). Thus
(7.10.21) is a consequence of (7.10.19) and (7.2.6).
In order to calculate the integral (7.10.21) let us pass to the k-representation
of the functions in the integrand. Using Parseval's formula we obtain

1
Q = S So(x)[volff(x) - 1] dx = (21t)3 S So(k)[volff(k) - (21tpo(k)] dk.

(7.10.22)

Let f(x) be the Gaussian density with the dispersion (j

f(x) = -v'21t (j exp( - x 2/2(j2) . (7.10.23)

214 7. Elastic Medium with Random Fields of Inhomogeneities

It may be shown for this case that

vo1/f(k) - (2rc)3 0(k) = exp( - J a2k 2)[(2rc)3 ~ o(k - 2rcm') -1], (7.10.24)

where m' is the vector of the reciprocal lattice.

If the basic medium is isotropic the tensor SoCk) is defined by (4.6.28, 16).
After substituting the k-representation of SoCk) the integral (7.10.22) is ex-
pressed in terms of the following three integrals:

10 = (2rc)-3 S [v o1/f(k) - (2rc)3 o(k) dk ,

It~ = (2rc)-3 S k~~~ [v o1/f(k) - (2rc)30(k)] dk,

It 8AI' = (2rc)-3 Sk(Xk~~Akl' [v o1/f(k) - (2rc)3o(k)] dk .

If the expectation of the random point field is the cubical lattice these
integrals become

(7.10.25)

where the components of the tensor r are defined in the system of the basic
vectors of the cubical lattice by

3
r(X~).1' = .t...J
'" u,uJ:qJ:l$o~ol!
t t t· (7.10.26)
;=1

The coefficients ao, al> a2 have the form of the convergent series (1 is the
lattice parameter)

a1 = t
;,j,k=-oo
(i2 + J('~~ (P
l
.i k 2) })~2+ i2) exp[-2( rcla )(i2 + p + k 2)]
+} + k
(7.10.27)

Taking into consideration the previous relations, (7.10.22) for Q may be

represented by

1 Q _ ( voao + a1 ) E2 _ (a o + 2a1 )E1 _ a2 r

2/10 - - 3(1 - 2))0) 1 - Vo 1 - Vo 1 - Vo .
(7.10.28)
 7.10 Random Field of Point Defects 215

It should be noted that in the case of regular lattices of point defects

(if -+ 0), it is more convenient to calculate the integral (7.10.22) using the x-
representations of the integrand functions.
The effective moduli tensor for the random cubical lattice of defects in
the isotropic medium is given by

B* = Bo - -1 P ( I - -QP
Vo
1
Vo
)-1 (7.10.29)

where Q is defined by (10.28).

Note that in the case of regular lattices of identical point defects, the ex-
pression for B* obtained by the effective field method is the exact solution of the
problem. In fact, in the case of simple lattices all hypotheses of the method
become exact statements. If we consider a complex lattice consisting of several
simple lattices, the exact solutions may be found in much the same way as in the
case of regular lattices of finite inclusions (see the end of Sect. 7.8).
It is interesting to estimate the error due to the substitution of finite inclu-
sions for the point defects by the calculation of effective constants. Let us con-
sider a cubical lattice of isotropic spheres embedded in an isotropic matrix. It
may be shown that the point-defects model corresponding to this case gives the
expression for the effective moduli tensor which is similar to (7.7.1 I) but the
coefficient a(p) should be replaced by its value at zero, i.e. a(O). The analogous
fact is encountered in the plane case for the square lattice of circular inclusions.
In Fig. 7.11 the curves for the relative shear modulus f1.*/ f1.o in the case of
square lattices of circular holes (curve 1) and circular rigid inclusions (curve 2)
are represented. The continuous curves correspond to the exact solution of
the problem [7.13], the broken curves are obtained from the corresponding
point-defects model and the broken curves with ringlets represent the effective-
field approximation in the case of finite inclusions (Sect. 7.7).
Analyzing the curves in the Fig. 7.11 we arrive at the following conclusion.
The substitution of inclusions for the point defects gives an error less than 10%
if the distances between the centers of inclusions are larger by 1.5 + 2 times
than the diameters of the inclusions.

6
5 1-'-*
4 fLo
3
2
1
0.8
0.6
0.4
0.2 Fig. 7.11. Shear modulus for square lattices of
holes (1) and rigid inclusions (2)
0
216 7. Elastic Medium with Random Fields of Inhomogeneities

In the case of a regular triangular lattice the transition to the point-defects

model gives the same values for effective constants as by an application of the
effective field method to the lattice of finite inclusions.
Thus the point-defects model yields good results by the calculation of the
effective constants of composites and simplifies simultaneously the process of
constructing the solution.

7.11 Correlation Functions in the Approximation by Point

Defects

Let us proceed to constructing correlation functions of fields of stress and strain

in a medium with point inhomogeneities. Since these correlation functions are
expressed in terms of the second moment jjZ of the effective field, see (7.5.20),
the problem is reduced to constructing this very quantity.
When approximating inclusions by point inhomogeneities (7.5.10) for jjZ(x)
becomes

jjZ(XI - xz) = jjo<P(xz - Xl)

+ JSO(Xl - X') (PO(X')X(XI; x')jj(x')jj(xz)lxl> xz) dx' , (7.11.1)

where the mean

(7.11.2)

by virtue of (7.10.7) is represented in the form

The function

(7.11.4)

is the mean under the condition that Xl> X2 E X, where X is the set of points, at
which isolated defects are located. Obviously, if Xl i= X2, then in Xl and Xz
only two different defects can occur. Therefore, jj2(XI - X2) pertains to the
pair interaction in a random field of point defects.
Along with the two-point moment of the effective field (7.11.4), let us intro-
duce the mean of jj2 at a point Xl under the condition that there is a defect at X2

(7.11.5)
 7.11 Correlation Functions in the Approximation by Point Defects 217

The limit of d(x), as Ixl -+ 00, will be denoted by doo- Since at large distances
the dependence on X2 in (7.11.5) disappears, we have

doo = (O'(x)O'(x)lx>. (7.11.6)

The expression for d(x) can be obtained analogously to (7.11.1) and has the
form

d(XI - X2) = IJoql(XI - X2)

+ S SO(XI - x') (PO(X')X(XI; x')O'(x')O'(xI)lxI; X2> dx' . (7.11.7)

Our next aim is to construct closed equations for the functions 0'2(X), d(x)
and q;(x). Here we shall make use of (7.11.1, 2 and 7), uncoupling the means in
the integrands with the help of statistical hypotheses of the type of H2 (Sect.
7.3) and some additional assumptions concerning the structure of conditional
means.
Let X X1X2 be X with points ~j, which appeared at Xl or X2. If

(7.11.8)

then, by the virtue of the definition (7.10.6) of the function X(xt. x'), we have
the equality

X(XI; x') = X(Xh X2; x') + o(x' - X2) with X2 E X. (7.11.9)

We consider the conditional mean in the integrand in (7.11.1). Taking into

account (7.11.9), we have

(PO(X')X(XI; x')0'(x')0'(x2)lxI; X2> = (PO(X')X(Xh X2; x')0'(x')0'(x2)lxI; X2>

+ o(x' - X2)(PO(X2)0'(X2)0'(X2)lxI; X2> . (7.11.10)

Using the hypothesis H2 (Sect. 7.3) and the assumption (7.5.12)

(0'(x')0'(x2)lx', Xh X2> = (0'(x')0'(x2)lx', X2> , (7.11.11)

the expression for each term on the right-hand side of (7.11.10) can be repre-
sented in the form

<pO(X')X(Xh X2; x')0'(x')0'(x2)lxI; X2>

= P<X(Xh X2; X')IXI; X2> 0'2(X' - X2) , (7.11.12)

(7.11.13)
218 7. Elastic Medium with Random Fields of Inhomogeneities

where P is the mean with respect to the ensemble of the stochastic variables
(7.10.2,3).
Substituting these expressions into (7.11.10) and then the result into (7.11.1),
we find

q2(XI - X2) = O'otP(X2 - Xl) + SO(XI - x2)Pd(X2 - Xl)

+ JSO(XI - X') P q 2(X' - X2) F(x', Xl> X2) dx' , (7.11.14)

where

(7.11.15)

The equation for the function 9'(x) follows from (7.11.3), if one represents
the mean in the integrand in the form

(PO(X') X(XI; x')q(X')IXI; X2) = P(X(Xl> X2; X')IXI; x2)(q(x')lx', Xl> X2)
+ o(x' - X2)P(q(X2)lxI; X2). (7.11.16)

Here the hypothesis H2 was used again.

Substituting this result into (7.11.2) with the assumption (7.5.2) taken into
account, we obtain

9'(XI - X2) = 0'0 + SO(XI - X2) P9'(X2 - Xl)

+ JSO(XI - X') P9'(x' - X2) F(x', Xl> X2) dx' . (7.11.17)

Analogously, transforming (7.11.7) for d(XI - X2), we have

d(XI - X2) = O'otP(XI - X2) + SO(XI - X2) P q 2(XI - X2)

+ JSO(XI - x') P q 2(X' - xI)F(x', Xl> X2) dx' . (7.11.18)

The equations (7.11.14,17 and 18) form a closed system with respect to the
required three functions. A specific structure of the random field of defects is
contained in these equations through the function F(x', Xl> X2), which is de-
fied by (7.11.15). Let us consider this function in more detail.
From (7.11.15) and the definition of the conditional mean (Sect. 7.4), we
obtain

(7.11.19)

This expression for the function F is the point of departure for the construction
 7.11 Correlation Functions in the Approximation by Point Defects 219

of the explicit expressions for random point fields of different types. Specific
examples of such constructions are carried out in Appendix A4.
In the case of a Poisson point field, the function F turns out to be equal to
the constant

F(x', Xl> X 2) = 1....

Vo
(7.11.20)

If the set X is a random spatial lattice (Sect. 7.10), then F is determined from
(A4.30) and has the form

1:' g(x' - Xl - m) 1:' g(X2 - Xl - n)

F(x', Xl> X2) = -<m"-----_ _--.=;_-,--------""""*!!!m_ _,---_ __ (7.11.21)
1:'
m
g(XI - X2 - m)

Here m, n are vectors of a fixed regular lattice (the expectation of the random
one under consideration), the primed summation signs mean exclusion of the
terms with m = 0 and n = 0, respectively, the function g(x) is defined by
(7.10.20).
The system of equations (7.11.14, 17 and 18) will be solved for the ex-
ample of the two-dimensional linear chain of defects. Let a system of parallel
rectilinear cuts of a stochastic length 21 located on one straight line be modeled
by point defects. The coordinates of the centers of cuts form a onedimensional
homogeneous random point field. Let us assume that the external stress 0"0 is
a tension along the normal to the line of cuts, i.e.

(7.11.22)

where 0"0 is a scalar.

Note that in the given problem, it is not the field O"(x) itself, that is of interest,
but nii(x), the latter determining completely the state of each crack. From the
symmetry properties it follows that

qaP(x)np = q(x)n a , (7.11.23)

where q(x) is a scalar.

In the case of an isotropic medium, the tensors P and nSo(x)n take the forms
(x is the coordinate along the straight line of the defects)

(7.11.24)

na
s,a(J).p( )
0 X np = Po 1... !lpJ.
21t(1 _ vo) x 2 u , (7.11.25)
220 7. Elastic Medium with Random Fields of Inhomogeneities

where l/x2 is a generalized function, the inverse Fourier transform of which

is -lllki.
Multiplying (7.11.14, 17 and 18) by the normal n and taking into account
(7.11.24, 25), we arrive at the system of equations (b 2 = <f2)/2)

f) 0'2(X) = O'o\D(X) + b2 d~~) + b2 I (x _1 X')2 F(x', x, O)O'2(X') dx' ,

d(x ) -- 0'0\£>()
X + b2 ----.xz
0'2(x) + b2 SOO D('
(x _.1 X')2 L"' X ,x, 0)0'-2(X') d x,
'

r
-00

\D(x) = 0'0 + b2 \D(x) + b2 1 F(x', x, O)\D(x') dx' (7.11.26)

x2 -00 (x - X')2

with respect to the three scalar functions

0'2(X) = <O'(x)O'(O)lx, 0), d(x) = <O'(x)O'(x)lx, 0) ,

(7.11.27)
\D(x) = <O'(x)lx, 0) .

If the point-defects density tends to zero, then the integral terms in these
equations vanish, and the equations describe interaction between two isolate
defects on the straight line. At the same time, the solution of the system (7.11.26)
has the form

(7.11.28)

Here the expression for \D(x) is the normal component of the stress field, in which
there are located two point defects, the distance between them being x. It is
obvious, that \D(x) is the asymptotics, as x > b, ofthe solution for the interaction
of two rectilinear cracks, which lie on the same straight line. If x :::;; b, the solu-
tion (7.11.28) has no physical sense.
As pointed out in Sect. 7.10, in the case of finite defects, there is no analog
to the point field, in which defects can occur at arbitrary small distances. For
example, in the case under consideration, the centers of cracks must be not
closer to each other, than the sum of their half lengths. This circumstance can
be partially taken into account, if one makes use of the model of a one-dimen-
sional point field, which is analogous to a random lattice of defects in the space.
Let Xk be a coordinate of the k-th defect and the differences XHI - Xk be
independent stochastic variables (for different k's), which have the same normal
distribution

1 [ ( x - 10)2 ] (7.11.29)
f(x) = .v21t0' exp - 20'2 .
 7.11 Correlation Functions in the Approximation by Point Defects 221

Here 10 is the average distance between defects, 0 2 is the dispersion. If 0 -+ 0,

then we have a regular chain of defects, and if 0 -+ 00, then this is a Poisson
point field on the straight line. To restrict the probability of defects coming
too close to each other, let us assume that the dispersion of stochastic variables
Xk+l - Xk is sufficiently small. Since the quantity (Xk+l - Xk), distributed ac-
cording to the law (7.11.29), lies in the interval (/0 - 30, 10 + 30)with a prob-
ability practically equal to 1, let us assume x == 10/0 ~ 3.
The function F(x', Xl> X2) in the integrands of (7.11.26) has, in the case
under consideration, the form
00 00

1:' fk(X' - Xl) 1:" f ..(XI - X2)

F(x', Xl> X2) = --"k==--'OO"::-_-----,ooco----"k==__
-oo=--_ _ __ (7.11.30)
1:' fk(XI
k=-oo
- X2)

where the primed summation sign means exclusion of the term with k = 0,
and the double prime means the same for n = 0 and n = k,

3 [ ( x - xlo)2] (7.11.31)
fix) = .v21tlklo exp - 202 .

Let us substitute (7.11.30) into (7.11.26) and search for the solution of the
latter in the form

X2
9'(x) = (bI(X) X2 _ b2 . (7.11.32)

This choice of the structure of the solution is due to the following reasons.
The functions 9', ff2 and d characterize an interaction between two defects in a
random field of point defects. In the first approximation one can assume that
their form coincides with that of the corresponding functions for two isolated
defects (7.11.28), and the presence of other defects can be taken into account
by correcting the external field. Thus we immediately arrive at (7.11.32).
Numerical computations show that the functions ffi(x), dl(x) and 9'I(X) are
well approximated by constants, the values of the latter depending on the para-
meters p = 2b/lo and x = I/o. Let us denote these constants by ffMp, ,.), doo(p,,.)
and 9'oo(p, ,.), respectively; one can prove that

(7.11.33)

The dependence of 9'oo(p, ,.) and doo(p, ,.) - 9'!(p,,.) on the parameters is shown
in Figs 7.12 and 13.
Since 9'00 = ffl the mean value of the effective field, as it is seen from Fig.
7.12, attains its maximum for large relative dispersion of distance between
222 7. Elastic Medium with Random Fields of Inhomogeneities

cr 1
2.5
d=-\p~
~2-
0.8 0
0.6
0.4
P=l 0.2
0
0.8
-0.2
0.6
0.4 -0.4
10 'X
-0.6

Fig. 7.12. Dependence u,/q. on" andp Fig. 7.13. Dependence (d~ - Y~)M on "
andp

defects (I> = 2.5 ..;- 3.5). With I> = 5 ..;- 6, the quantity 0'1 no more coincides
with the value, which corresponds to a regular chain of defects (I> = (0) .
The difference doo - rp! is equal to the dispersion of the effective field 0'.
From Fig. 7.13, it is seen that the dispersion has its maximum for I> = 4 ..;- 5,
and for I> = 9 ..;- 10, is practically zero, which corresponds to the transition
to a regular structure. For I> < 3, the quantity doo - rp! becomes negative.
This loss of physical meaning is connected with an increase of the probability
of the defects approaching each other at distances less than b for small 1>.
As it is seen from (7.11.32) for 0'2(X), the correlation radius of the effective
field has the order of the average value of the dimension b of a defect and
weakly depends on the relative dispersion of the distance 1>-2 between defects.
Let us now proceed to calculating the second moment of the stress field
o-(x) in a medium with point defects. The relation (7.5.20) becomes

(0"(XVo"(X2» = o-ij
+ JSO(X1 - x')dx' JSO(X2 - x")(PO(x')PO(x")X(x')X(x")O'(x')O'(x"» dx'.
(7.11.34)

Taking into account the equality analogous to (7.11.8), namely,

X(x') = X(x"; x') + o(x' - x") with x" E X (7.11.35)

and the hypothesis H2 about the statistical independence (Sect. 7.3), we repre-
sent the mean in the integrand in (7.11.34) by

(PO(x')PO( x")X (x')X(x")O'( x')O'( x") )

= (X(x", x')X(x"»p0'2(X' - x")P + ~o o(x' - x")(P2)doo , (7.11.36)

 7.11 Correlation Functions in the Approximation by Point Defects 223

where doo has the form of{7.11.6).4

Substituting (7.11.36) into (7.11.34) we obtain

(O-(x)O'{O» = 0'5 + .1. H{x)doo + Vo JH{x - x')¢{x')a2{x') dx', {7. I 1.37)

Vo

where

H{x) = JSo{x - x')PSo{x') Pdx' , (7.11.38)

¢(x - x') = (X{x; x')lx) . (7.11.39)

Let us return to the one-dimensional field of point defects and calculate

the second moment of the normal component p(x) of the stress tensor, when
x is a point on the line of the defects

(7.11.40)

where

(7.11.41)

From (7.11.37), for the given case, we obtain

d
0'5 + /1t(x) J 1t{x -
00

p(x) = + 10 x')¢(x')a 2(x') dx' , (7.11.42)

o -00

where doo == d( 00) and a 2(x) have the form of (7.11.32), the function ¢(x) is
represented by

00

¢(x) = I;' fk(X). (7.11.43)

k=-oo

The primed summation sign means exclusion of the term with k = 0, and
fk{X) has the form of (7.11.31). The function 1t(x) is an analog of H(x) from
(7.11.37) and is defined by the relation

1t(x) = b4 -L (x _1 x')2 . X~2 dx' = -b4o"(x). (7.11.44)

From here and from (7.11.39), we finally obtain

4It is taken into account here that at one and the same point there cannot be two different
defects; therefore, we have <a(x')a(x")lx, x") = <a(x')a(x')lx') = d« if x' = x".
224 7. Elastic Medium with Random Fields of Inhomogeneities

p(x) = 0'5 - d [dI; o(x) + sb(x)a (x)] .

2
b4 dx 2- 2 (7.11.45)

The graph of the continuous part of the function p(x) - O'~ is presented in
the Fig. 7.14. The presence of a singular part and of a singularity at x = bin

O~~k-~~~~~~~~~~
4 5 x/21
-1
-2
-3
Fig. 7.14. Dependence of correlation radius p on ,,/21

the correlation function O'nn(x) of the random field on the line of defects is con-
nected with replacing real defects by point inhomogeneities. For a random
field of defects with finite dimensions, the correlation function is smooth,
bound and has the minimum correlation radius of the order of the average
dimension of a defect. When the random field of defects tends to a regular
structure, the correlation radius of the stress field increases; this is easily seen
from Fig. 7.14.

7.12 Conclusions

Let us now estimate the accuracy of the method of an effective field and discuss
a possibility of using the first moments of the solution of the stochastic elastic
problem, for strength calculations of structures made of composites.
It was pointed out above that in the general case it is impossible to obtain
an explicit solution of the elastic problem for stochastically inhomogeneous
medium. Each statistical moment of the solution is expressed in terms of a set
of moments of higher orders, and to truncate the chain of equations, one has
to introduce additional assumptions, usually without rigourous proofs of their
validity.
There exists a number of approaches to solving such problems; surveys
can be found, for example, in [7.2, 21-24]. The method of an effective field
considered in this chapter, stands somewhere between self-consistent schemes
and methods of the type of smoothing approximation or stochastic phases
approximation [7.21, 27].
 7.12 Conclusions 225

Introducing the local external field if and assuming if to be constant in a

neighborhood of each particle (hypothesis HI of Sect. 7.3)5 are in agreement
with the method of the self-consistent field (MSF). The distinction from the
classical scheme of MSF is that we consider the field if as stochastic and, when
constructing the closed equations for statistical moments if, the procedure of
decomposing complex means is used, this being typical for the method of
smoothing approximation.
Note that the first approximation of the method of an effective field (Sect.
7.6-9) yields the same results as MSF does, when applied to computations of
effective parameters of inhomogeneous media [7.12, 25, 26]. However, this
does not exhaust all possibilities of the method of an effective field. The method
enables one to obtain expressions, generally speaking, for any statistical mo-
ment of the solution, while it appears to be possible to assign physical sense to
hypotheses and corrections of the employed calculating scheme. This is the
advantage of the method of an effective field in comparison with, for example,
an approximate summation of the formal series of perturbation theory [7.2, 24].
An estimate of the accuracy of the method of an effective field is important.
It is known [7.27] that when describing a system of point particles in quantum
and stochastic physics in the region where interaction is essential, the slower
the potential of an individual particle decays at infinity and the larger the density
of particles, the better the classical MSF works. For particles with a potential
of the Coulomb type, the self-consistent solution gives a good coincidence with
experimental data, though rigorous estimates cannot be obtained. Such esti-
mates are usually deduced on the basis of physical reasonings.
Particles of composite media considered in this chapter have the following
peculiarities. First, their dimensions are finite and, secondly, the state of each
particle is described by its own field which depends on the local external field
that contains the particle. The method of an effective field is based on hy-
potheses which concern the approximation of a state of each particle in a
deformable composite. The range of the variation of the parameters of inhomo-
geneities, in which these hypotheses are precise statements, is small: this
covers low concentrations or absence of interaction. In fact, in this case each
particle is contained in a constant external field (hypothesis HI of Sect. 7.3),
which statistically does not depend on stochastic dimensions and elastic con-
stants of the particle (hypothesis H2 of Sect. 7.3). Hence, all the results of the
preceeding sections become precise if the concentration of defects is low enough.
With increasing the concentration, the validity of the hypotheses HI and
H2 is, generally speaking, broken. The hypothesis HI becomes invalid, if the
effective field essentially differs from a constant one in regions occupied by

5The latter is, generally speaking, not obligatory. One can assume only an identical be-
havior of the field if in a neighborhood of each particle; see the end of Sect. 7.9.
226 7. Elastic Medium with Random Fidds of Inhomogeneities

the particles. This can occur when particles are situated close enough to each
other, i.e. when the concentration is high. On the other hand, the more the
defects affect each other, the more reasonable the hypothesis H2 is.
However, invalidity of the hypotheses HI and H2 can weakly influence
the first and the second moments of the solution, i.e. the quantities of most
interest for applications. These moments are extremely rough statistical char-
acteristics of a solution, and a part of the information about detailed behavior
of the particles is here lost. Therefore, a detailed consideration of the state
of each particle is not reasonable. Of course, this argument can be confirmed
only by comparing the derived results with exact solutions or experiments.
Since exact solutions of stochastic problems are unknown, comparisons
with experiments play an important role. Fr9m the results of Sect. 7.7 and 9,
it follows that the method of an effective field enables one to describe suffi-
ciently the results of experimental measurements of effective elastic constants
in the two-dimensional case, and, for regular composites, yields a fair agree-
ment with exact solutions.
In Sect. 7.7. a dimensionless scalar parameter X was introduced, with the
help of which the range of applicability of the method can be described. The
size of the region of values X, where the method gives an error more than 10 %,
is relatively small. This region contains, for example, X for the cases of circular
holes or of absolutely rigid inclusions at concentrations of p < 0.4. The esti-
mate seems to be valid not only in the case of a regular triangular lattice of
inclusions, but also for other types of structures, in which distributions of
inclusions differ weakly from the isotropic one. Analogous estimates for
stochastic fields of cracks follow from the results of Sect. 7.9.
In the spatial case, the method of an effective field has qualitatively the
same region of applicability. However, in the three-dimensional medium,
the potential of a particle decays more rapidly than in the planar case ( ,...., r- 1
and ,...., r- 2 , respectively). Another peculiarity of the three-dimensional problem
is, that here more particles can interact directly. All this can exhibit quantita-
tive boundaries for the applicability of the method.
Unfortunately, the absence of sufficient data in the literature, concerning
the determination of correlation functions of elastic fields in an inhomogeneous
medium, does not allow us to estimate the error of the method when calculating
the second moments of the solution.
In conclusion, let us dwell briefly upon the scheme of using the obtained
results for strength computations of structures made of composites. Such a
computation can be divided into two steps: computing the macro-stress and
estimating the level of micro-stress.
When computing the macro-stress, the composite is considered as a homo-
geneous medium, the elastic properties of which are determined by effective
constants. Here one can use the classical elasticity and consider the macro-
stress obtained as an external one when estimating the level of micro-stress
connected with an inhomogeneity of the material.
 7.13 Notes 227

Even the first approximation of the method of an effective field enables

one to calculate effective constants and to estimate the stress concentration
on defects under the assumption that each of them is contained in a constant
external field ql (Sect. 7.6-9).
More precise results follow by taking into account fluctuations of the
effective field (Sect. 7.5, 11). Knowledge of the values of dispersion and of
correlation radii of the effective field q{x) and knowledge of the real micro-
stress field l1{x) enables one to calculate the probability of fracture of the
structure and the value of critical loadings. When using a criterion of fracture,
for example, through the value of maximum stress, one has, first of all, to
determine the validity of the inequality l1{x) > 11*, where 11* is a critical level
of stress, and to estimate the set of points, where this condition is valid. If
one assumes that l1{x) is a Gauss random field, then the last problem is solved
by standard methods of probability theory.
When the condition of fracture is fulfilled at a point, a whole neighborhood
of the point will become fractured, its dimension being of the order of the
correlation radius of the random field of micro-stress. In such a way dimen-
sions and concentration of fractured regions are determined, the level of macro-
stress being given. Replacing the material of the fractured regions by defects
with appropriate properties, one can repeat the computation and find the
probability of fracture of the whole structure.
An analogous scheme can be used for estimating the probability of trans-
forming the composite into the plastic state.

7.13 Notes

The method of an effective medium as a tool for calculating macro-constants

of heterogeneous media has a long history. With the help of this method,
effective coefficients of electroconductivity, and the dielectric and magnetic
conductivities of inhomogeneous materials of various structures were found.
A listing of corresponding works can be found, for example, in the surveys
[B7.5, 17] where other approaches to these problems are also discussed. The
work of Hori [B7 .11] in which practically all the existing methods of construct-
ing effective constants are considered, is of particular interest.
A number of basic concepts, concerning the calculation of elastic macro-
constants of polycrystal metals were suggested by Kroner [B7.19, 20]. The
method of an effective medium was treated by Kroner in [B7,I9] (see also the
work of Hershey [B7.9]). In the paper of Hill [B7.IO], effective constants of
composites with ellipsoidal inclusions were found with the help of this method,
textured polycrystals were considered by Kneer [B7.18] and Morris [B7.27],
media with cracks were studied by Budjansky and O'Connel [B7.3]. A sub-
sequent development of static and dynamic problems is due to Willis [B7.33,34].
228 7. Elastic Medium with Random Fields of Inhomogeneities

Application of the method of self-consistent fields to solving multi-particle

problems of quantum mechanics (the Hartree-Fock method) and to a de-
scription of phase transitions (Weiss method) can be found in standard courses
of quantum mechanics and statistical physics. Some estimates of the applica-
bility of the method were considered by Kac [B7.12].
The notion of a constant self-consistent field, in which each particle is
situated, was used in a number of works for solving the problem of a wave
propagating in an inhomogeneous medium. The calculation of effective char-
acteristics of propagation and attenuation of sound in a medium with bubbles
was described in the monograph of Morse and Feshbach [B8.22], an analogous
problem for liquid suspensions was solved by Chaban by the method of a
self-consistent field [B7.4].
The method of a self-consistent field was also used implicitly in works
of Walpole [B7.32] and Levin [B7.24] for the calculation of effective elastic
constants of composites, which contain an isotropic random field of ellipsoidal
inclusions. A generalization to arbitrary fields of random ellipsoidal inhomo-
geneities (in particular, on regular structures) was obtained by Kanaun [B7.14,
15].
An application of the method to the description of stochastic and regular
systems of cracks in the space and in the plane was considered by Kanaun
[B7.13], and Kanaun and Iablokova [B7.16] (see also [B7.24]).
Appendices

At Fourth-Order Tensors of Special Structure

When constructing Green's operators for isotropic media and in a number of

other cases, one has to deal with tensors of a special type, which depend on the
Kronecker delta and on a unit vector. Here, the formulae are presented, which
enable one to simplify the operation with these tensors, and in particular, to
obtain explicit expressions for inverse tensors.
Let us consider tensors of the forth order made up of Kronecker deltas and
the unit vector n, which are symmetric in the first and the second pairs of indices,
but which are, generally speaking, not symmetric with respect to permutation
of the pairs. It is easy to show that all tensors having such a structure can be
represented as linear combinations of the following six linearly independent
tensors (these thus play the role of a basis):

1
E!{3).!' = T(Oa),0fl!' + oa!'ofl),), E';fl).!' = oafl0),!',

E~fl).!' = oafln),nl" E!fl).f' = nanfl0).f' '

(A 1.1)
E~(3).!, = ! (nan)'0fl!' + nan!'0fl). + nfln),oa!, + nflnf'0a),) ,
E~fl).!' = nanfln),nf' .
Let us define a multiplication-convolution of two tensors

Eik = Ei.Ek, (A 1.2)

where

(A 1. 3)

It is easy to verify that the tensors Eik can be expressed as linear combinations
of the basis elements Ei.
The linear tensor space, spanned by the basis Ei is closed with respect to the
product introduced, and hence forms an algebra which we denote A(n). We set
Ei.p = a~~jEj,

where a~~j is a matrix defined by the table

230 Appendices

1 E1 1 E21 E31 E41 E5 1 E6

E1 1 E1 1 E21 E31 E41 E5 1 E6
E2 1 E213E21 3E3 1 E21 E3 1 E3
E3 1 E31 E21 E31 E21 E3 1 E3 (A1.4)

E4 1 E413E413E61 E41 E6 1 E6
E5 1 E51 E41 E6 1 E41 J(E5 + E6) 1 E6
E6 1 E61 E41 E61 E41 E6 1 E6

However, for practical calculations, it is more convenient to change the basis

to the set Ii which is connected with the former by the relation

Ii = A~Ek, (A 1.5)

where A is the following matrix (i is the number of a row):

0 1 1 1 0 3
2 -2 -2 2
0 1 1 1 0 1
-2 2 2 2
0 0 -1 1 0 3 (A 1.6)
A~ = -2 2
0 0 -1 1 0 1
2 2
1 1 1 1 -2 1
-2 2 2 2
The inverse matrix a~ has the form

1 0 0 0 1 1
3 1 3 1 0 0
2 -2 -2 2
1 1 1 1 0 0
2 2 -2 -2
a~ = (A 1.7)
1 1 - 1 1 0 0
2 2
1 1 0 0 1 0
2 2 2
1 1 0 0 0 0
2 2
 Al Fourth-Order Tensors of Special Structure 231

The multiplication table for the basis tensors Ji has a simpler form

I
J1 I J3 I J4 I J5 I J6
I
J2

JI JI J2 I J3 I J4 I 0 0
I I I
J2 J2 JI I-J4 I- J 3 I 0 0
I I I
J3 J3 J4 I Jl I J2 I 0 0 (A1.8)
I I I
J4
I
J4
I
J3 I-J2 1-J1 I 0 I
0
J5 0 0 0 0 J5 0
I I I I I I
J6 0 0 I 0 0 0 J6
I I I I I I

It is seen from the table that the algebra A(n) is split into the direct sum of two
subalgebras with the basis J5, J6 and JI, ... , J4, the second subalgebra being
isomorphic to a real form of the complex quaternion algebra (the usual real
quaternion algebra will be obtained if one substitutes J2 -+ J ~ = iJ2, J3 -+ J~ =
iJ3).
A tensors Qa~).1' is inverse to a tensor P a~).1' if these tensors satisfy the relation

(A 1.9)

Let a tensor P = Pk JR, and a tensor Q = q~k be inverse to P. Then, for the
components of the tensor Q, we have

PI _l!:L
( qb ... , q6) = ( T' 11 '
_l!L _l!.L
11 ' 11' is' ;6), (A 1. 10)

where

(A 1. 11)

It is easily seen that qh ... , q4 are connected with Ph ... , P4, as in the qua-
ternion case, and only difference being that "modulus squared" is not of con-
stant sign. The conditions of existence of the inverse tensor are: 11 =F 0, P5 =F 0,
P6 = o.
A special case, when the tensor P a~).1' is symmetric with respect to permuta-
tion of the pairs a{3 and Af1. is of interest, since the tensor of elastic constants
has this symmetry. One can show that in this case, the inverse tensor Qa~).1'
possesses the same symmetry.
232 Appendices

As usual, one can define a scalar product of two elements of the algebra
A(n) as the complete contraction of the two tensors. Then, the scalar product
of the unit tensor and an aribtrary one coincides with the trace of the latter. In
particular, for the basis elements Jk, we have

J1 I J2 I J3 I J41 J5
(AU2)
Tr Jk I 2 I 0 I 0 I 0 I 2

The algebra A(n) and its applications to elasticity is considered in more

detail in [B6.25].

A2 Green's Operators of Elasticity

Here, the Green's operators of static elasticity will be considered in the k-

representation. The elastic medium is assumed to be unbound and homogene-
ous. The tensors of the elastic moduli C and of the elastic compliance B = C-l
are assumed to be given. In a local theory, C and B are constants, while in a
nonlocal theory they are functions of the vector k. For the sake of simplicity
the dependence of C and B on k is not pointed out explicitly.
As a preliminary, let us consider the operator Rot in the k-representation.
According to (5.3.11), the Fourier transform R(k) of the operator Rot can be
represented in the form

R(k) = -k211l(n) , (A2.I)

where n« = k«/ I k I and

(A2.2)

It is easily seen that Ell(n) belongs to the class of tensors considered in Appendix
AI. In fact, one can directly verify the identity

c«").c fJpp = o«fJo"Po).p + o"fJo).poap + o).{3o«po"P - o«{3o).po"P

- o).{3o"Po«p - o"{3o«po).p (A2.3)

and taking into account the formulae (AU) for the basis Ei, we find

Ell(n) = _El + E2 - E3 - E4 + 2E5. (A2.4)

With the help of the matrix a~ given by the table (A1.7), we can rewrite
(A2.4) in the basis Ji of the algebra A(n)

men) = -.l
2
(Jl - J2) - J6 . (A2.5)
 A2 Green's Operators of Elasticity 233

Note the following important identities for &len), being easily verified with
the help of the Table (A 1. 8) :

(A2.6)

From this, it follows, in particular, that&l2(n) is a projection operator. Let us

denote this projector by eo
and introduce the complementary projector = no
/ - eo (/ no eo
== E 1). It is easily seen that the projections and are self-adjoint
and their representation in the basis Ji has the form

no = i (Jl + J2) + J5, eo = i (Jl - J2) + J6. (A2.7)

Let ,(x) be a second-order tensor field, which vanishes at infinity, and let
,(k) be its Fourier transform. Then, as is easily verified, the following equiv-
alence relations

=0
div , '¢!> no, = 0 , (A2.8)
Rot, = 0 '¢!> eo, = 0
hold.
Let us proceed to the Green's operators of elasticity. The Green's operator
for displacements G(k) is defined by (4.6.4, 5). Its explicit expression in the
isotropic case is given by (2.8.6), where;' and f.i are to be taken as constants.
From the results of Sect. 4.6, it follows that the Green's operator for the
strain K(k) is connected with G(k) by the relation

(A2.9)

it satisfies the equations

&lK = 0, k·CKC = k·C (A2.10)

and the identity

KCK= K. (A2.H)

In the isotropic case, its explicit expression is given by (4.6.16).

The Green's operator for internal stress S(k) is connected with K(k) by

S = C - CKC, K = B - BSB, (A2.12)

satisfies

k·S = 0, RBS= R (A2.13)

and the identity

234 Appendices

SBS = S. (A2.14)

In the isotropic case, its expression is given by (5.3.13).

The relations given above express K and S in terms of G. Conversely, one
can accept K and S as the main Green's operators and express G in terms of
them. A complete system of equations, which determines the operators K and
S, has the form

Rot K = 0, div S = 0, (A2.I5)

S =C- CKC or K =B - BSB. (A2.I6)

Taking into account (A2.8), (A2.I5) can be rewritten in an equivalent form

8 0K = 0, DoS = O. (A2.I7)

In order to show the existence and uniqueness of the solution ofthe algebraic
(in the k-representation) system of equations (A2.I6, 17), let us introduce some
useful notations. Let L be an arbitrary linear operator, L1 and L2 its restrictions
to the subspaces of the projections Do and eo, respectively, and

Lu = DoLi, L2i = eOLi (i = 1,2). (A2.I8)

Then the operator L can be represented in the form of a two-by-two matrix

with operator components Lij (i, j = 1, 2).
Let us write down all the operators in the matrix form and substitute them
into (A2.I6, 17). Elementary computations show that the only components of
K and S which differ from zero, are Kll and S22, respectively, and

(A2.I9)

where Cll and B22 are the inverses in the corresponding subspaces. Finally, the
existence of the inverse operators follows from the positive-definiteness of the
operators C and B. The Green's tensor G(k) for displacement is, in tum, ex-
pressed in terms of K or S (or D and e) with the help of simple algebraic opera-
tions. At the same time, obtaining explicit formulae in the x-representation is
possible only in the simplest cases, in particular for the isotropic or hexagonal
symmetries.
The Green's tensors and projection operators for the isotropic medium were
obtained in [B6.25]. In particular, it is shown that the projections Do and 8 0
admit an additional invariant decomposition

Do = m+ m, 8 0 = 80+8 ~, (A2.20)

where D~ and 8~ are projections onto subspaces of traceless tensors.

 A3 Green's Operators K and S in x-Representation 235

A3 Green's Operators K and S in x-Representation 1

For a homogeneous medium, the action of the Green's operators K and S in

the x-representation is reduced to convolution with the corresponding kernels
K(x) and S(x). Therefore we shall begin with a consideration ofthese kernels.
Generalized functions K(x) and S(x) can be defined as the inverse Fourier
transforms of the functions K(k) and S(k), or, according to (4.6.13) and (4.6.29),
as the second derivatives of G(x). Let us base our argument on the second
approach and consider K(x).
We shall use a known form of regularization, which is applicable to gener-
alized functions of the type of the second derivatives of the Green's functions
[BS]. In the case given, for K(x), we have in shorthand notation

K(x) = - VG(x) V = K(x) + Ao(x) , (A3.1)

where K(x) is a regular functional and A is a constant. This representation of

K(x) in the form of a sum of regular and singular components depends on the
choice of a special form of regularization, although K(x) itself does not depend
on the latter.
Let V be a region with a smooth boundary 0 and x = 0 be an internal
point of V. Then K(x) acts on the test functions 9>(x) according to the rule

(K, 9» = SK(x)[9>(x) - 9>(0)] dx + SK(x)9>(x) dx, (A3.2)

v v
where K(x) is, according to (4.6.13) a formal second derivative of G(x) and V-
is the complement of V in three-space. The constant A is given by the expression

A = - S VG(x)n(x) dO, (A3.3)

D

where n(x) is the normal to 0 at a point x.

Our first task is to obtain a more convenient expression for A directly in
terms of K(k). For this purpose, not that, making use of Gauss' theorem, one
can rewrite (A3.3) in the form

A = (VG(x), VV(x)) , (A3.4)

where V(x) is the characteristic function of the region V. Further, taking into
account Parceval's equality (2.1.6), we find

1
A = (2n)3 SkG(k)kV(k) dk, (A3.5)

lAppendix A3 was written in collaboration with S. K. Kanaun.

236 Appendices

or, taking into account (9.6.13),

1
A = (21t)3 JK(k) V(k) dk . (A3.6)

Let us assume now that {) is an ellipsoid given by (4.9.13) and perform the
substitution x = ay. Then V'(y) = V(ay) is the characteristic function of the
unit sphere and V(k) = det{a}V'(ak). Substituting into (A3.6) and replacing
k by a-1k, we obtain

A = (2~)3 JK(a-1k)V'(k)dk. (A3.7)

Let k = Ikl Q). Then K(a-1k) = K(a-1Q), since K(k) is a homogeneous

function of the zeroth degree. On the other hand, it is obvious that V'(k) =
V'(lkl). This enables us to carry out the integration in (A3.7) with respect to
Ikl and the unit vector Q) separately. Using formulae from Appendix Al for the
Fourier transform of spherically symmetric functions, we find

(A3.8)

and, hence,

(A3.9)

where the integration is carried out over the unit sphere Q)2 = 1.
Note that the constant A does not depend on the absolute dimensions of
the ellipsoid {), since K(k) is a homogeneous function of the zeroth degree.
Therefore the functional defined by (A3.2) does not depend on absolute dimen-
sions of the ellipsoid. This enables one to pass to the limit by contracting the
ellipsoid to a point; for definiteness, let this point be x = 0). Then the integral
over V vanishes, since K(x) - IxP, and the second integral is expressed as an
integral in the sense of principal value, which exists since (K, rp) is well defined.
Thus, finally,

K(x) = K(x) + Ao(x) , (A3.10)

where A is given by (A3.9) and the regular generalized function is defined by

the principal value of the integral
K, rp = det {a} f K(ax)rp(ax) dx. (A3.H)
Here and in (A3.9), the tensor a can be considered as an arbitrary nondegenerate
linear transformation of x-space.
 A3 Green's Operators K and S in x-Representation 237

Quite analogously, for the generalized function Sex) we have

Sex) = Sex) + Do(x) , (A3.12)

where

(S, 'P) =det{a} f S(ax) 'P (ax) dx, (A3.13)

D = l1t JS(a-1l1) dll) , (A3.14)

while, according to (4.6.9),

D = C - CAC. (A3.15)

The regularization of K(x) and Sex) can be carried over to the plane case
in an obvious way, while

A = I
21t JK(tr 1l1) dll), D = I
21t JSea-ll1) dll), (A3.16)

where the integration is performed over the unit circle lI)2 = 1.

Let us now consider a generalized function F(x) = VV(/)(x), where (/)(x) is a
homogeneous function of degree -m + I, m being the dimension of the space.
It is obvious that F(x) is a homogeneous function of degree -m - I and that
its Fourier transform is a homogeneous function of degree I. The regularization
formula for F(x) can be obtained analogously to (A3.1, 2) and has the form

(F, 'P) = - JV(/) (x) [V'P(x) - V'P(O)] dx + JV(/)(x)['P(x) - qJ(O)] n(x)dD

v {}
- J(/)(x)n(x)dDV'P(O) + JF(x) ['P(x) - 'P(O)] dx . (A3.17)
{} v

Let V be a sphere of radius e. If e ~ 0, then the first integral vanishes, since

V(/) '" Ixl- m , the integrals over D are reduced to integrals of the function
F(k)k over the unit sphere in k-space, and the last integral tends to an integral
in the sense of a principal value and exists since (F, 'P) is well defined.
If F( -x) = F(x), then the integrals over D in (A3.17) vanish also, and
the regularization becomes

(F, 'P) = f F(x) ['P(x) - 'P(O)] dx . (A3.18)

Let Xl, X2, x 3 be cartesian coordinates in the three-dimensional x-space,
and let n be a normal to the plane x 3 = O. We consider a generalized function
T over the plane Xl, x 2 which is generated by the function Sex),
238 Appendices

(A3.19)
It is easily verified that the k-representation of the function T has the form

(A3.20)

Using the explicit expression for the k-representation of the function S

(Appendix A2), one can show that this integral converges absolutely and
determines an even, homogeneous function of degree 1. Since, in the x-repre-
sentation, regularization of such functions has the form of (A3.6, 18), the
functional Tacts on a (two-space) test function q>, according to the rule

(T, q» = f T(x) [q>(x) - q>(O)] dx , (A3.21)

where the integral over the plane is understood in the sense of principal value
and exists.
It is not difficult to verify that in the Fourier representation,

(T, q» = (2;)2 JT(k)q>(k) dk , (A3.22)

corresponds to the regularization (A3.21). The integral converges absolutely.

Let D be a smooth, closed surface with the normal n(x). We consider the
integral

lex) = Jn(x)S(x - x')n(x')b(x')o(D) dx' , (A3.23)

where b(x)o(D) is a o-function with a smooth weight b(x) (Sect. 6.1). The func-
tion lex) is defined for all x, other than xED, since, in the latter case, the in-
tegral (A3.23) formally diverges. Let us find a regular representation of the
integral lex) for x E D. For this purpose, we rewrite lex) in the form

lex) = Jn(x)S(x - x') n(x')[b(x') - b(x)]o(D) dx'

+ Jn(x)S (x - x')n(x')o(D)dx'b(x) . (A3.24)

From Gauss' formula and from the property div S = 0 of the function Sex),
follows the equality

Jsex - x') n(x')o(D)dx' = Sdiv Sex - x') dx' = 0, (A3.25)

v

where V is the region bounded by D. Thus, the second integral in (A3.24) is

equal to zero.
 A4 Calculation of Certain Conditional Means 239

Let us consider a sequence of finite functions <p.-, which converges to 0(0),

as i -+ 00. Replacing 0(0) in (A3.12) for Sex) and then take the limit as i -+ 00.
The final expression for lex) is

lex) = f n(x)S(x - x')n(x')[b(x') - b(x)]o(O) dx' . (A3.26)

This representation is valid for all x, while, for XED the existence of the integral
in the sense of a principal value follows from (A3.21).
Now let 0 in (A3.23) be a smooth, not closed, simply-connected surface
bound by a smooth contour r.This case is reduced to the previous one, if we
consider 0 to be a part of the closed surface Or. and hex) to be zero for all x E
{) = 0 1\0. In this case, the integral (A3.26) is represented by

lex) = f n(x)S(x - x')n(x')[b(x') - b(x)] dO'

D

- f n(x)S(x - x')n(x')dO'b(x) . (A3.27)

lJ

With the help of Stokes' formula, the second integral is transformed into an
integral along the boundary r.
Here one has to take into account that, due to
(5.3.3)1

Sex) = Rot Z+(x) , (A3.28)

where Z is the Green's tensor for internal stress (Sect. 5.3).

Finally, the regularization of the integral (A3.23) becomes

Jn(x)S(x - x')n(x')b(x')o(O) dx'

= f n(x)S(x - x')n(x')[b(x') - b(x)]dO' -

fJ
Jn(x)rot Z+(x -
r
x')drb(x),
(A3.29)
where dr is a vector element of length on r. For the integrals on the right-
hand side to exist, it is sufficient that b(x) be continuously differentiable on D.

A4 Calculation of Certain Conditional Means

We consider a composite medium which is a homogeneous matrix contaning a

homogeneously distributed set of ellipsoidal inhomogeneities. Let Vex) be the
characteristic function of the whole region occupied by the inclusions. When
using the method of an effective field in the case of composites, the problem of

'Recall that, in Chaps. 5 and 6, the Green's function S is denoted by G.

240 Appendices

constructing different conditional means of stochastic functions, which are

analogous to Vex), arises. The one-dimensional analog of the problem is con-
sidered in describing stochastic pulse flows in radio engineering [B7. 17]; a num-
ber of general results was obtained in papers on geometrical probabilities
[B8.19, 26].
In a number of cases, it is convenient to approximate inclusions in the com-
posite by point defects (Sect. 7.10). Therefore, we start with the the simplest
case of a homogeneous point field in three-space. Let X be a random point field
of ~i' XXI ... Xn reduces to X after excluding from the latter those ~; which
coincided with one of the points X; (i = I, ... , n) in the space; X(x) and
X(Xh ... , Xn; x) are the following generalized functions, concentrated on X
XXI ... n ' respectively:
X

X(x) = 1: o(x - ~,), (A4.1)

ei EX

X(Xh X2, ... , Xm x) = 1: o(k - ~,) . (A4.2)

~iEXX1X2 .•. Xn

Let us consider the following averages of the functions X(x) and X(XI; x)
with respect to the ensemble of realizations of the point field X:
<X(X» = <eiEX
1: o(x - ~,.) , (A4.3)

(A4.4)

(A4.5)

(with regard to the interpretation of these quantities as conditional means,

Sect. 7.4).
In what follows, all random fields are assumed to be ergodic. The standard
method of constructing the means of the type (A4.3-5), which we shall use, con-
sists of using the ergodic property with the subsequent averaging with respect
to the ensemble, if necessary. For example,
. 1 N 1
<X(x» = hm -
v-->oo
f 1:
V G ;=1
N
o(x -
.
~i) dx = hm - = - .
v-->oo V Vo
(A4.6)

Here G is the region of averaging, its limit coinciding with the whole space,
v is the volume of G, N is a number of points of a fixed realization of the point
fieId, which are contained in G, and Vo is an average volume per point.
Analogously,

-I 1'f X "# Xl,

<X(XI' x» = lim 1-
V
f 1:
Gei=XI
o(x - r;;)dx = { Vo (A4.7)
o if x
v-->oo
= Xl,
 A4 Calculation of Certain Conditional Means 241

where it is taken into account that, by definition X(XI; Xl) = O.

Hence

(X(X» = (X(XI; x», (A4.8)

to within a null-set. Note that all the equalities will be understood in this sense.
Let us proceed to the more complicated problem of constructing the two-
point moment of the function X(x). For ergodic fields

(X(X)X(X + Xl» = lim

v--+oo
~ f L; o(x -
V G i,j
g,)o(x + Xl - gj) dx (A4.9)

We introduce a stochastic vector g"j = g. - gj and let its distribution density

be g ../x). We then average (A4.9) once more with respect to the ensemble of
realizations. The means of spearate terms in (A4.9) are

(A4.1O)

where it is taken into account that g. i = O.

Separating on the right-hand side of (A4.9) the terms with i = j, we have

(X(X)X(X + Xl» = ~ O(XI) + lim ~ L; giixV . (A4.11)

vo v-->oo V ",j
U*j)

From here we find the expression for the numerator in the conditional mean
(A4.4). By virtue of the definition (A4.2) of the function X(XI; x), the equality

X(X) = o(x - Xl) + X(XI; X) with Xl E X (A4.12)

is valid, from which it follows that

(X(X)X(x + xV) = ~ O(XI) +

Vo
(X(x; X + XI)X(X» . (A4.13)

Comparing this result with (A4.11), we obtain

(A4.14)

Let us consider two examples of random point fields in space.

1) Poisson's field (Sect. 7.8, 10). To Poisson's point field in space, there cor-
responds in (A4.14) the density gij = l/v with any finite v. It is easy to verify that
242 Appendices

then

(A4.15)

and the conditional mean (A4.4) becomes

(X(Xl; x) I x) = -L . (A4.16)
Vo

In order to construct the conditional mean (A4.5), note that, due to the
absence of correlation between positions of the points (Poisson's field I), the
equalities

(X(Xl; X)X(Xl)X(Xl; X2» = (X(Xl» (X(Xl; X)X(Xl; X2», (A4.17)

(X(Xl; X)X(Xl» = (X(Xl» (X(Xl; x» (A4.18)

hold.
Calculating the right means analogously to the preceding and substituting
the result into (A4.5), we obtain

(A4.19)

2) Random spatial lattice. Let us consider a random point field, to each

element f;j of which there corresponds a stochastic vector

f;m = m + pm + r , (A4.20)

where m is a vector of a fixed regular lattice, Pm are independent stochastic

vectors with zero mathematical expectation, r is a stochastic vector with a homo-
geneous distribution.
The characteristic function g mn(k) of the stochastic vector

(A4.21)

has the form

gmn(k) = f(k)f( _k)e-ik(m-n) , (A4.22)

wheref(k) is the characteristic function of the stochastic vector pm'

If one introduces

J
g(X) = (21t)-3 f(k)f( - k)eikx dk , (A4.23)
 A4 Calculation of Certain Conditional Means 243

then the function gmix) is represented in the form

gmn(x) = g(x - m + n) . (A4.24)

Let us consider the double sum in (A4.14) and proceed from two-index
quantities gmn to one-index ones, and rewrite this double sum as a single one.
The vectors m - n, with m, nEG, take on N2 values, some of which may be
identical. Each of these vectors coincides with one of the vectors of the fixed
lattice. The number of vectors, which are equal to a fixed one m, depends on
the relations between the length Iml of m and dimensions of G. If the dimen-
sions of G are much larger than Iml, then this number is '" N. Taking this
into account, (A4.14) can be represented by

<X(Xl; X)X(Xl» lim.l 1: g(x

= v-->oo - Xl - m + n) = .l 1:' g(x - m) ,
V m, nEG Vo m
(m",n)
(A4.25)

where the primed summation sign means exclusion of the term with m = O.
From here and from (A4.6) follows the desired expression for the mean
(A4.4)

<X(Xl; x) I Xl) = 1:' g(x - Xl - m) . (A4.26)

m

Let us proceed to constructing the mean (A4.5) for the random lattice.
The mean in the numerator of (A4.5) has the form

(A.4.27)

where €ij is given by (A4.21).

Let us average both sides of (A4.27) with respect to the ensemble of real-
izations of the point field (A4.20). Since €ji and €ki withj =F k are independant
stochastic vectors, then their joint distribution function is gji(X)gki(X). Averag-
ing with respect to the ensemble of the separate terms in (A4.27) yields

(A4.28)
244 Appendices

(O(X2 - ~i.)O(X3 - ~ji) = JO(X2 - X)O(X3 - x)gji(x)dx

= O(X2 - X3)gji(X3) if j = k. (A4.29)

We substitute this result into (A4.27) and transform the double sums to
single ones. Using (A4.25), we find the expression for the mean (A4.5) in
the form

(A4.30)

where the primed summation sign means exclusion of the terms with m =
o or n = o.
We complete the consideration of random point fields and proceed to the
construction of analogous means for homogeneous random fields of noninter-
secting ellipsoids. Let us confine ourselves to the problem of constructing
the means of the characteristic functions Vex) and V(XI; x) of the form

(V(x) = (L: V,{x) , (A4.31)

•
(A4.32)

(A4.33)

where the quantities appearing here are defined in Sects. 7.3, 4.

For ergodic fields, the means (A 4.31, 32) are most simply calculated
according to

(V(x) = (V(XI; x) = lim 1..- JL: Vk(x) dx = p, (A4.34)

.-+00 V G k

where p is the concentration of inclusions.

Let us consider the second moment of the function Vex). By definition,

(V(x)V(x + Xl) = lim vI

tJ~oo
JG,r; V,(x)V,{x + Xl) + ~
J,l
V.(x)Vix + Xl] dx.
('*i)
(A4.35)

Note that JV,{x) V.(x + Xl) dx is the volume of intersection of two identical
ellipsoids, the centers of which differ by the vector Xl. From here, we have

(A4.36)
 A4 Calculation of Certain Conditional Means 245

where

J; (x) = j(o1_ /a i1 x/)2

2
(1 + /a i1 x/) if /ai 1x/ ::;; 2
4
if /ai1x/ > 2
(A4.37)

The tensor ail determines a linear transformation which transforms the

ellipsoid Vi to the unit sphere.
For homogeneous fields of inclusions, the stochastic variables v;Ji(x) with
any fixed x have the same distributions for all i. Therefore, the first term in
(A4.35) is equal to the mean value of this stochastic variable, divided by the
volume Vo per inclusion.
The mean (V(Xb X)V(Xl» coincides with the mean (A4.35), if one does
not consider those realizations, for which the points x and Xl occur inside the
same inclusion. It is obvious that this mean is equal to the second item in
(A4.35). From here and from (A4.33, 34) we find

W(X) = (V(Xl; Xl + X / Xl» = ~ !~~ ! b~V,{xl)Vixl

(#j)
+x) dXl. (A4.38)

Starting with this expression, one can show that W(x) is a continuous
function with the mean value p. Since the different V; do not intersect W(O) =
O. For each realization it follows from the definition of the function V(x; Xl)
that V(x; x) = O.
The form of the function W(x) depends on specific structure of the random
field of inclusions. In particular, for regular lattices of identical inclusions,
the mean (V(x) V(Xl» is a periodic function of the difference X - Xl. In a
neighborhood of the origin, the form of this function is determined by the
first term in (A4.35). From here, it follows that W(x) can be represented in
the form

W(X) = p 1:'
m
J(x - m) , (A4.39)

where m is a vector of the regular lattice, formed by the centers of inclusions,

the primed summation sign means that the term with m = 0 is omitted, and
the function J(x) is defined by (A4.37).
In conclusion, we consider the means which appear in the problem of a
random field of cracks in an elastic medium (Sect. 7.8).
Let Q(x) be the sum of o-functions concentrated on the set of flat elliptic
surfaces, the latter forming a random field in space (Sect. 7.4). For ergodic
fields, we have

(O(x» = lim .l
v-+oo
J1: 0; (x)dx =
V G ;
lim.l
v-+oo
t 1ta;b;,
V ;=1
(A4.40)
246 Appendices

where the aj, h j are semiaxes of the ellipse OJ. Since, for a homogeneous field,
all stochastic variables aj, h j have identical distributions with the same mean
value of <a h), then

<0) = 1C<ah) . (A4,41)

Vo

Further, let us consider the mean

<h(X)O(X)O(XI}) = <1;; h;(x)O;(X)Oj(XI», (A4,42)

IJ

where the functions hj(x) are given on OJ by the relations

hj(x I, x2) -_ J (a:xl )2 - ( Ii;

1- X2 )2 . (A4,43)

The axes xl, x 2 coincide with the principal axes of the ellipse OJ.
Analogously to (A4.35), let us write (A4,42) in the form

<h(x)O(x)O(x + Xl»

= lim .l..
v-+ooV.
2: JG h;(x)O,{x)Oj(x + xI)dx + <~) U
?IJ'(XI) , (A4.44)

where

(A4,45)

In order to calculate each of the integrals in the first term in (A 4.44),

let us choose a cartesian coordinate system on OJ, by taking the x 3-axis along
the normal to OJ and the x 2 and x 3-axes along the principal axes of the ellipse.
Integrating first with respect to x 3 and then with respect to Xl and X2, we obtain

Jhj(x')Oj(x')Oj(x' + x)dx' = 1Cajb,.J;(xl, x2)o(x 3) , (A4,46)

where
 A4 Calculation of Certain Conditional Means 247

Substituting this result into (A4.44), we have

(A4.48)

where (ahJ(x)D(x» is the mean value of the stochastic variables

a,biJi(X) D'(x, ai' bi), D'(x, ai' bi) is the a-function, concentrated on the flat
elliptic surface, whose normal coincides with that of Di , the values of the
semiaxes are 2ai , 2bi and the center is located at the origin.
The function 7Jf(x) depends on a specific model of a random field of cracks.
For a regular lattice of identically oriented surfaces, 7Jf(x) has the form

7Jf(x) = 7tOb 1:' J(x - m)D'(x - m) , (A4.49)

Vo m

where m is a vector of the lattice, formed by the centers of the ellipses Dm,
the primed summation sign denoting omission of the term with m = O.
In Chap. 7, the simplified notation

D'(x - m) = a;,,(x) (A4.50)

was introduced.
References

Chapter 1
1.1 E. Cosserat, F. Cosserat: Theorie des Corps Deformables (Hermann, Paris
1909)
1.2 C. Truesdell, R.A. Toupin: "The Classical Field Theories", in: Principles
of Classical Mechanics and Field Theory, Encyclopedia of Physics, Vol.
III/I (Springer, Berlin, Gottingen, Heidelberg 1960) pp. 226-793
1.3 Cz. WoZniak: Dynamic models of certain bodies with discrete-con-
tinuous structure. Arch. Mech. Stosow. 21, 707 - 724 (1969)
1.4 G.N. Savin, Iu. N. Nemish: Investigations on stress concentration in
couple-stress elasticity (survey). Prikl. Mekh. 6, 1-17 (1968) [in Rus-
sian]
1.5 A.A. Il'jushin, V.A. Lomakin: "Couple-Stress Theories in the Mechanics
of Solid Deformable Bodies", in Prochnost i plastichnost (Strength and
Plasticity), ed. by A. Il'jushin (Nauka, Moscow 1971) pp. 54-61 [in
Russian]
1.6 M. Misicu: Mecanica Mediilor Deformabile. Fundamentele Eleaticitatii
Structurale (Continuum Mechanics. Foundations of Elasticity) (Acad.
R.S.R., Bucharest 1967) [in Rumanian]
1.7 W. Nowacki: Theory of Micropolar Elasticity (Wydawn. uczeln. poli-
techn., Poznan 1970) [in Polish]
W. Nowacki: Theory of Nonsymmetric Elasticity (Zaklad narod. im
Ossolinskich PAN, Wroclaw 1970) [in Polish]
1.8 E. Kroner (ed.): Mechanics of Generalized Continua. Proc IUTAM
Symposium on the Generalized Co sse rat Continuum and the Continuum
Theory of Dislocations with Applications, Freudenstadt and Stuttgart
1967 (Springer, Berlin, Heidelberg, New York 1968)
1.9 LA. Kunin, V.G. Kosilova (eds): Teoria Sred s Mikrostrukturoi (Theory
of Media with Microstructure, Bibliography) (Sib. Otdelenie AN SSSR,
Institut teplofiziki, Novosibirsk 1976)
1 .1 0 E. Kroner: On the physical reality of torque stresses in continuum
mechanics. Int. 1. Eng. Sci. 1 ,261-278 (1963)
1.11 LA. Kunin: Teoria Uprugikh Sred c Mikrostrukturoi (Theory of Elastic
Media with Microstructure) (Nauka, Moscow 1975)
l.12 LA. Kunin: Elastic Media with Microstructure lOne-Dimensional
Models, Springer Series in Solid-State Sciences, Vol. 26 (Springer,
Berlin, Heidelberg, New York 1982)
1.13 D.G.B. Edelen: Irreversible thermodynamics of nonlocal systems. Int. 1.
Eng. Sci. 12,607 -631 (1974)
250 References

Chapter 2
2.1 l.M. Gel'fand, G.E. Shilov: Generalized Functions (Academic, New York
1964)
2.2 M.Born, K. Huang: Dynamical Theory of Crystal Lattices (Clarendon
Press, Oxford 1954)
2.3 E.L. Aero, E.V. Kuvshinskii: Fundamental equations of the theory of
elastic media with rotationally interacting particles. SOy. Phys. Solid
State 2, 1272-1281 (1960)
2.4 G. Grio1i: Elasticita asimmetrica. Annali di Matematica Pura e Applicata
4,389-418 (1960)
2.5 R.D. Mindlin, H.E. Tiersten: Effects of couple-stresses in linear elasticity.
Arch. Rat. Mech. Anal. 11 ,415-448 (1962)
2.6 G. Leibfried: "Gittertheorie der mechanischen und thermischen Eigen-
schaften der Kristalle ", in Encyclopedia of Physics, Vol.7, Part 1
(Springer, Berlin 1955)
2.7 1.M. Ziman: Electrons and Phonons (University Press, Oxford 1960)

Chapter 3
3.1 V.S. Oskotskii, A.L. Efros: On the theory of crystal lattices with non-
central interatomic interaction. SOy. Phys.-Solid State 3, 448-457
(1961)
3.2 V.E.Vdovin, LA. Kunin: A theory of elasticity with spatial dispersion.
Three-dimensional complex structure. 1. Appl. Math. Mech. 30, 1272-
1281 (1966)
3.3 l.M. Gel'fand, S.V. Fomin: Calculus of Variations (Prentice-Hall, Engle-
wood Cliffs, NJ 1963)
3.4 V.E. Vdovin: On energy relations in the theory of media with micro-
structure. Dinamika Sploshnoi Sredy 7, 94-104 (1971)
3.5 M.Born, K. Huang: Dynamical Theory of Crystal Lattices (Clarendon
Press, Oxford 1954)
3.6 K. Huang: On the atomic theory of elasticity. Proc. R. Soc. A203, No.
1072, 178-194 (1950)
3.7 A.O. Gel'fond: Calcul des Differences Finies (Dunod, Paris 1963)
3.8 V.A. Pal'mov: Fundamental equations of the theory of asymmetric
elasticity. Prikl. Math. and Mech. 28, 401-408 (1964)

Chapter 4
4.1 l.M. Lifshits, L.N. Rosentsveig: Green's tensor for anisotropic unbounded
elastic medium. Zh. Exsp. Teor. Fiz.17, 783-791 (1947)
4.2 LM. Gel'fand, G.E. Shilov: Generalized Functions (Academic Press, New
York 1964)
4.3 G.l. Eskin: Boundary- Value Problems for Elliptic Pseudo-Differential
Equations (Math. Soc. of USA, N. Y. 1981)
 References 251

4.4 J.R. Willis: The stress field around an elliptical crack in an anisotropic
elastic medium. Int. J. Eng. Sci. 6, 253-263 (1968)
4.5 J.D. Eshelby: "Elastic inclusions and inhomogeneities", in Prog. Solid
Mech., Vol. 2, ed. by LN. Sneddon, R. Hill (1961) pp. 88-140
4.6 J.D. Eshelby: The elastic field outside an ellipsoidal inclusion. Proc. R.
Soc. A252, 561-569 (1959)
4.7 J.D. Eshelby: The determination of the elastic field of an ellipsoidal
inclusion, and related problems. Proc. R. Soc. A241, 376-396 (1957)
4.8 I.M. Lifshifts, L.V. Tanatarov: On the elastic interaction of impurity
atoms in crystal. J. Phys. Metals Metallography 12, 331-338 (1961)

Chapter 5
5.1 E. Kroner: Kontinuumstheorie der Versetzungen und Eigenspannungen,
(Springer, Berlin, Gottingen, Heidelberg 1958)
5.2 B.A. Bilby: "Geometry and continuum mechanics", in [Ref. B. 2.1, pp.
180-199]
5.3 E. Kroner: Allgemeine Kontinuumstheorie der Versetzungen und Eigen-
spannungen, Arch. Rat. Mech. Anal. 4 , 273-334 (1960)
5.4 I.A. Kunin: "Methods of tensor analysis in the theory of dislocations".
A supplement to the Russian translation of J .A. Schouten: Tensor Ana-
lysis for Physicists (Nauka, Moscow 1965), pp. 374-443. (The supple-
ment is available in English from the U.S. Department of Commerce,
Clearing house for Federal Sci. and Techn. Information, Springfield,
VA 2215})
5.5 RAA G Memoirs of the Unifying Study of Basic Problems in Engineering
and Physical Sciences by Means of Geometry, ed. by K. Kondo (Gaku-
jutsu Bunken Fukyu-Kai, Tokyo, V.I. 1955, V. II 1958; V. III 1962,
V. IV 1968)
5.6 C.c. Wang: "On the geometric structure of simple bodies, a mathematical
fo'undation for the theories of continuous distributions of dislocations",
in [Ref. B2.1, pp. 247-250J
5.7 R. de Wit: "Differential geometry of a nonlinear continuum theory of dis-
locations", in [Ref. B2.1, pp. 251-261]
5.8 I.A. Kunin: Internal stresses in media with microstructure. J. Appl.
Math. Mech. 31, 898-906 (1967)
5.9 LA. Kunin: An algebra of tensor operators and its applications to elasti-
city. Int. J. Eng. Sci., 19, 1551-1561 (1981)
5.10 A.I. Lur'e: Three-dimensional Problems of the Theory of Elasticity (In-
terscience, New York 1964)
5.11 I.M. Lifshits, L.V. Tanatarov: On the elastic interaction of impurity
atoms in crystals. J. Phys. of Metals and Metallography 12, 331-338
(1961)
252 References

Chapter 6
6.1 J. Friedel: LesDislocations (Gauthier-Villars, Paris 1956)
6.2 E. Kroner: Kontinuumstheorie der Versetzungen und Eigenspannungen,
(Springer, Berlin, Gottingen, Heidelberg 1958)
6.3 LA. Kunin: Fields due to arbitrary distributions of dislocations in an ani-
sotropic elastic medium. J. Appl. Mech. Tech. Phys. 6, 76 (1955)
6.4 LA. Kunin: "Methods of tensor analysis in the theory of dislocations". A
supplement to the Russian translation of J.A. Schouten. Tensor Analysis
for Physicists (Nauka, Moscow 1965), pp. 374-443. (The supplement is
available in English from the U.S. Department of Commerce, Clearing
house for Federal Sci. and Techn. Information, Springfield, VA 22151)
6.5 LM. Gel'fand, G.E. Shilov: Generalized Functions (Academic Press, New
York 1964)
6.6 A.D. Brailsford: Stress field of a dislocation. Phys. Rev. 142,383-392
(1966)

Chapter 7
7.l 1.1. Gikhman, A.V. Skorokhod: Theory of Random Processes, Vol. 1,
(Nauka, Moscow 1971) [in Russian]
7.2 T.D. Shermergor: Theory of Elastic Inhomogeneous Media (Nauka,
Moskow 1977) [in Russian]
7.3 LA. Chaban: Self-consistent field approach to calculation of effective
parameters of microinhomogeneous media. Sov. Phys.-Acoust. 10, 298-
304 (1964)
7.4 E. Kroner: Bounds for effective elastic moduli of disordered materials. J.
Mech. and Phys. Solids, 25, 137 -15 5 (1977)
7.5 1.1. Gikhman, A.V. Skorokhod: Theory of Random Processes, Vol.2
(Nauka, Moscow 1971) [in Russian]
7.6 V.M. Levin: On the stress concentration in inclusions in composite
materials. J. Appl. Math. and Mech. 41, 753-761 (1977)
7.7 A.V. Hershey: The elasticity of an isotropic aggregate of anisotropic
cubic crystals. J. Appl. Mech. 21,236 (1954)
7.8 G. Kneer: Dber die Berechnung der Elastizitatmoduls vielkristaller Ag-
gregate mit Textur. Phys. Stat. Sol. 9, 825--838 (1965)
7.9 E. Kroner: Berechnung der elastischen Konstanten des Vielkristalls
aus den Konstanten des Einkristalls, Z. Phys.151, 504 (1958)
7.l 0 P.R. Morris: Elastic constants of polycrystals. Int. J. Eng. Sci. 8, 49
(1970)
7.ll Z. Hashin, S. Shtrikman: On some variational principles in anisotropic
and nonhomogeneous elasticity; J. Mech. Phys. Solids 10, 343 (1962)
7.12 S.K. Kanaun: Self-consistent field approximation for an elastic compo-
site medium. J. Appl. Mech. Tech. Phys. 18, 274-282 (1977)
7.13 E.L Grigoliuk, L.A. Fil'shtinskii: Perforated Plates and Shells (Nauka,
Moscow 1970) [in Russian]
 References 253

7.14 A.E. Andreikiv, V.V. Panasiuk, M.M. Stadnik: Fracture of prismatic bars
with cracks. Problemy Prochnosti, No. 10. 10-16 (1972)
7.15 B. Budjanski, R.J. O'Connel: Elastic moduli of a cracked solid. Int. J.
Solids Struct. 12, (1976)
7.16 A.S. Vavakin, R.L. Salganik: On effective characteristics of inhomogenous
media with local inhomogeneities. Izv. Akad. Nauk SSSR, MTT No.3
(1975)
7.17 S.K. Kanaun, G.I. Iablokova: "Self-consistent field approximation in
plane problem for systems of interacting cracks", in Mekhanika Sterzhne-
vykh Sistem i Sploshnykh Sred, No.9, (Leningr. Inshenemostroitel'nyi
Institut, Leningrad 1976) pp. 194-203
7.18 L.A. Fil 'shtinskii: Interaction of a doubly-periodic system of rectilinear
cracks in an isotropic medium. J. Appl. Math. and Mech. 38 ,853-861
(1974)
7.19 P.e. Daris, G.C.M. Sih: "Stress analysis of cracks" in Fracture Toughness
Testing and its Applications (Am. Soc. for Testing and Materials,
1965) pp. 30-83
7.20 V.V.Panasiuk, M. Savruk, A.P. Datsishin: Stress Distribution Around
Cracks in Plates and Shells (Naukova Dumka, Kiev 1976)
7.21 G. Adomian: Linear random operator equations in mathematical physics.
J. Math. Phys. 12, 1944 (1971)
7.22 R.J. Elliott, J.A. Krumhansl, P.L. Leath: The theory and properties of
randomly disordered crystals. Rev. Mod. Phys. 46,465 (1974)
7.23 M. Hori: Statistical theory of effective electrical, thermal and magnetic
properties of random heterogeneous materials. J. Math. Phys. 14, 514-
523; 1942-1948 (1973); 15,2177-2185 (1974); 16,352-364; 1772-
1775 (1975); 18,487-501 (1977)
7.24 A.H. Nayfeh: Pertubation Methods (Wiley, New York 1970)
7.25 V.M. Levin: On calculation of effective moduli of composites. Dokl.
Akad. Nauk SSSR 220, No.5 (1975)
7.26 LT. Walpole: On bounds for the overall elastic moduli of inhomoge-
neous systems - I, II. J. Mech. Phys. Solids 14, 151 and 289 (1966)
7.27 M. Kac: "Mathematical mechanisms of phase transitions" in Statistical
Physics. Phase Transitions and Superf!uidity, Vol. 1 ed. by M. Chretien,
(Gordon & Breach, New York 1968) pp. 241-305
Bibliography

BI General Topics in Continuum Mechanics

Bl.1 M.Y. Beran: Statistical Continuum Theories (Interscience, New York,

1968)
B 1.2 I.M. Lifshits, L.N. Rosentsveig: Green's tensor for anisotropic unbounded
elastic medium. Zh. Exsp. Teor. Fiz.17, 783-791 (1947)
B 1.3 A.I. Lur'e: Three-dimensional Problems of the Theory of Elasticity (Inter-
science , New York 1964)
B 1.4 N.I. Muskhelishvili: Singular Integral Equations: Boundary Problems of
Functions Theory and their Applications to Mathematical Physics,
(Wolters-Noordhoff, Groningen 1972)
BI.S S. Timoshenko, D.H. Young, W. Weaver: Vibration Problems in Engineer-
ing, (Wiley, New York 1974)
B l.6 C. Truesdell, R.A. Toupin: "The Classical Field Theories", in Handbuch
der Physik, Vol. III/l, ed. by S. Fliigge (Springer, Berlin, Gottingen, Hei-
delberg 1960)

B2 Couple-Stress Theories
B2.l E. Kroner (ed.): Mechanics of Generalized Continua. Proc. IUTAM
Symposium on the Generalized Cosserat Continuum and the Continuum
Theory of Dislocations with Applications, Freudenstadt and Stuttgart
1967 (Springer, Berlin, Heidelberg, New York 1968)
B2.2 I.A. Kunin, V.G. Kosilova (eds.): Theory of Media with Microstructure.
Bibliography (Sib. Otdelenie AN SSSR, Institut Teplofiziki, Novosibirsk
1976)
B2.3 E.L. Aero, E.V. Kuvshinskii: Continuum theory of asymmetric elastici-
ty. The problem of internal rotation. SOy. Phys.-Solid State 5, 1892-
1897 (1963)
B2.4 E.L. Aero, E.V. Kuvshinskii: Continuum theory of asymmetric elastici-
ty. Equilibrium of an isotropic body. SOy. Phys.-Solid State 6,2141-
2148 (1964)
B2.S E.L. Aero, E.V. Kuvshinskii: Fundamental equations of the theory of
elastic media with rotationally interacting particles. SOy. Phys.-Solid
State 2, 1272-1281 (1960)
B2.6 A. Askar: Molecular crystals and the polar theories of the continua. Ex-
perimental values of material coefficients for KN0 3 . Int. J. Eng. Sci.
10,233-300 (1972)
256 Bibliography

B2.7 E. Cosserat, F. Cosserat: Theorie des Corps Deformables (Hermann,

Paris 1909)
B2.8 A.C. Eringen: Balance laws of micromorphic mechanics. Int. J. Eng.
Sci. 8, 819-828 (1970)
B2.9 A.C. Eringen: Linear theory of micropolar elasticity. J. Math. anlt
Mech. 15, 909-924 (1966)
B2.10 A.C. Eringen: "Mechanics of micromorphic continua", in [Ref. B2.1,
pp. 18-35]
B2.11 A.C. Eringen, E.S. Suhubi: Nonlinear theory of simple micro-elastic
solids. Int. J. Eng. Sci. 2,189-203 and 389-404 (1964)
B2.12 G. Herrmann, J.D. Achenbach: Applications of theories of generalized
Co sse rat continua to the dynamics of composite materials, in [Ref.
B2.1, pp. 69-79]
B2.13 A.E. Green: Micro-materials and multipolar continuum mechanics. Int.
J. Eng. Sci. 3,533-537 (1965)
B2.14 A.E. Green, P.M. Naghdi: "The Cosserat surface", in [Ref. B2.1, pp.
36-48]
B2.15 A.E. Green, R.S. Rivlin: Multipolar continuum mechanics. Arch. Rat.
Mech. Anal. 17, 113-147 (1964)
B2.16 G. Grioli: Elasticita asimmetrica. Annali di matematica pura e applicata
4,389-418 (1960)
B2.17 W. GUnter: Zur Statik und Kinematik des Cosseratschen Kontinuum.
Abh. Braunschw. Wiss. Ges.10, 195-213 (1958)
B2.18 A.A. Il'jushin, V.A. Lomakin: "Couple-stress theories in the mechanics
of solid deformable bodies", in "Prochnost' i plastichnost'" (Nauka,
Moscow 1971) pp. 54-61
B2.19 W.T. Koiter: Couple-stresses in the theory of elasticity. Proc. K. Ned.
Acad. Wet. Ser. B6 7, 17 (1964)
B2.20 M.R. Korotkina: Note about couple-stresses in discrete media. Vestn.
Mosk. Univ.-Math. Mech. No.5, 103-109 (1969)
B2.21 R.D. Mindlin: Micro-structure in linear elasticity, Arch. Rat. Mech.
Anal. 16,51-78 (1964)
B2.22 R.D. Mindlin, H.E. Tiersten: Effects of couple-stresses in linear elastici-
ty. Arch. Rat. Mech. Anal. 11 ,415-448 (1962)
B2.23 M. Misicu: Mecanica Medulor Deformabile Fundamentele Elasticitatii
Structurale (Acad. R. S. R., Bucuresti 1967) [in Roumanian]
B2.24 W. Nowacki: Theory of Micropolar Elasticity (Wydawn. uczeln. poli-
techno Poznan 1970) [in Polish]
B2.25 W. Nowacki: Theory of Nonsymmetric Elasticity (Zaklad narod. im Os-
solinskich FAN, Wroc1aw 1970) [in Polish]
B2.26 V.A. Pal'mov: Fundamental equations of the theory of asymmetric
elasticity. Prikl. Math. and Mech. 28, 401-408 (1964)
B2.27 V.A. Pal'mov: On a model of a medium with complex structure. J.
Appl. Math. and Mechanics 33,747-753 (1969)
B2.28 R.S. Rivlin: "Generalized Mechanics of continuous media", in [Ref.
B2.1, pp. 1-17]
 Bibliography 257

B2.29 D. Rogula: Moment stresses and the symmetry of the stress tensor in
bodies with non-local structure. Bull Acad. Pol. Sci. Ser. Sci. Techn.18,
159-164 (1970)
B2.30 G.N. Savin, lu. N. Nemish: Investigations on stress concentration in
couple-stress elasticity (survey). Prikl. Mekhanika 6,1-17 (1968)
B2.31 H. Schaefer: The Cosserat continuum. Z. Angew. Math. und Mech. 47,
485-498 (1967)
B2.32 la. M. Shiriaev: Investigation of the influence of a scale factor on stress
concentrations around holes. Mekhanika Polimerov, No.3, 565 (Akad.
Nauk Latv. SSR, 1970)
B2.33 E. Sternberg: "Couple stresses and singular stress concentrations in
elastic solids", in [Ref. B2.1, pp. 95-108]
B2.34 R. Stojanovic: "Dislocations in the generalized elastic Cosserat con-
tinuum", in [Ref. B2.1, pp. 152-155]
B2.35 R.A. Toupin: Theories of elasticity with couple-stresses. Arch. Rat.
Mech. Anal. 17, 85-112 (1964)
B2.36 R.A. Toupin: "Dislocated and oriented media", in [Ref. B2.1, pp.
126-140]
B2.37 K. Wilmanski, Cz. Wozniak: On geometry of continuous medium with
microstructure. Arch. Mech. Stosowanej 19, 715 - 723 (1967)
B2.38 Cz. Wozniak: Dynamic models of certain bodies with discrete-contin-
uous structure. Arch. Mech. Stosowanej 21, 707-724 (1969)

B3 Nonlocal Theories
B3.1 D.G.B. Edelen, A.E. Green, N. Laws: Nonlocal continuum mechanics.
Arch. Rat. Mech. Anal. 43, 36-44 (1971)
B3.2 A.C. Eringen, D.G.B. Edelen: On non-local elasticity. Int. J. Engng. Sci.
10,233-248 (1972)
B3.3 A.C. Eringen: Nonlocal polar elastic continua. Int. J. Eng. ScilO, 1-6
(1972)
B3.4 A.C. Eringen: Linear theory of nonlocal microelasticity and dispersion
of plane waves. Lett. Appl. Sci. 1, 129-146 (1973)
B3.5 A.C. Eringen, C.G. Speziale, B.S. Kim: Crack-tip problem in nonlocal
elasticity. J. Mech. Phys. Solid. 25, 339-355 (1977)
B3.6 A.C. Eringen: "Nonlocal continuum mechanics and some applications",
in Nonlinear Equations in Physics and Mathematics, ed. by A.O. Barut
(Reidel, Dordrecht, Holland 1978) pp. 271-318
B3.7 F. Hehl, E. Kroner: Zum Materialgesetz eines elastichen Mediums mit
Momentenspannungen. Z. Naturforsch. 20, 336-350 (1965)
B3.8 S. Kaliski, C. Rymarz: Surface waves in continua with nonlocal interac-
tion. Biul. WAT J. Dabrowskiego 20,17-39 and 25-37 (1971)
B3.9 D. Kessel, E. Kroner: Nichtlokale Elastizitatstheorie der lonenkristalle
vom Typ A+Ir. Z. Naturforsch. 25. 1046-1053 (1970)
B3.1O V.G. Kosilova, I.A. Kunin: Dynamics of generalized Cosserat models.
Dinamika Sploshnoi Sredy No.4, 73-82 (1970)
258 Bibliography

B3.11 V.G. Kosilova, LA. Kunin: On a scattering problem for media with
dispersion. Chislennye Metody Mekhaniki Sploshnoi Sredy, 5, No.2,
51~72 (1974)
B3.12 V.G. Kosilova, LA. Kunin: Wave packets propagation in one-dimen-
sional medium with space dispersion. Dinamika Sploshnoi Sredy, No.6,
12'7~142 (1970)
B3.13 E. Kroner: On the physical reality of torque stresses in continuum
mechanics. Int. J. Eng. Sci. 1, 261~278 (1963)
B3.14 E. Kroner: Elasticity theory of materials with long range cohesive
forces. Int. J. Structures, 3, 731 ~ 742 (1967)
B3.15 E. Kroner: "Interrelations between various branches of continuum
mechanics", in [Ref. B2.1, pp. 330~340]
B3.16 E. Kroner: " Ober die Symmetrie-Eigenschaften elastischer Materialien
mit Kohasionskraften endlicher Reichweite," in "Problemy Gidrodina-
miki i Mekhaniki Sploshnoi Sredy" (Nauka, Moskva 1969) pp. 293~300
B3.17 E. Kroner, B.K. Datta: Nichtlokale Elastostatik: Ableitung aus der
Gittertheorie. Z. Physik 196. 203~211 (1966)
B3.18 E. Kroner, B.K. Datta: "Non-local theory of elasticity for a finite
inhomogeneous medium - a derivation from lattice theory", in Funda-
mental Aspects of Dislocation Theory, Vol. 2, ed. by J. Simmons and
R. de Wit (National Bureau of Standards, Washington 1970) pp. 737 ~
746
B3.19 J.A. Krumhansl: "Some considerations of the relation between solid
state physics and generalized continuum mechanics", in [Ref. B2.l,
pp. 298~311]
B3.20 LA. Kunin: Inhomogeneous elastic medium with nonlocal interaction.
J. App!. Mech. Tech. Phys. 8. 41 ~44 (1967)
B3.21 I.A. Kunin: Internal stresses in media with microstructure. J. Appl.
Math. Mech. 31, 898~906 (1967)
B3.22 LA. Kunin: Model of an elastic medium of simple structure with three-
dimensional dispersion. J. App!. Math. Mech. 30. 642~652 (1966)
B3.23 LA. Kunin: Theory of elasticity with spatial dispersion. One-dimen-
sional complex structure. J. App!. Math. Mech. 30, I 025~1 034 (1966)
B3.24 LA. Kunin: The Green's tensor for an elastic isotropic medium with
spatial dispersion. J. App!. Mech. Tech. Phys. 8, 103 (1967)
B3.25 LA. Kunin: Theories of elastic media with microstructure. Proc. Vibr.
Probl. 9, 323~336 (1968)
B3.26 LA. Kunin: "The theories of elastic media with microstructure and the
theory of dislocations", in [Ref. B2.1 , pp. 321 ~3 29]
B3.27 LA. Kunin: "The theory of elastic media with microstructure", in
Prochnost'i Plastichnost' (Nauka, Moskva 1971) pp. 65~ 70
B3.28 LA. Kunin, A.M. Vaisman: "On problems of the non-local theory of
elasticity", in: Fundamental Aspects of Dislocation Theory, Vo!. 2,
ed. by J. Simmons and R. de Wit (National Bureau of Standards,
Washington 1970) pp. 747~759
 Bibliography 259

B3.29 R.D. Mindlin: "Theories of elastic continua and crystallattice theories",

in [Ref. B2.1, pp. 312-320]
B3.30 E. Soos: One model of pseudo-continuum for ionic crystal lattices.
Proc. Vibrat. Probl. Pol. Acad. Sci. 11, 251-265 (1970)
B3.31 A.M. Vaisman, LA. Kunin: Boundary value problems in nonlocal
theory of elasticity. J. Appl. Math. Mech. 33, 765-777 (1969)
B3.32 A.M. Vaisman, A.P. Putintseva: Certain effects of a nonlocal nature in
a chain involving long-range interaction, J. Appl. Mech. Tech. Phys. 11,
621-626 (1970)
B3.33 V.E. Vdovin: On energy relations in the theory of media with micro-
structure. Dinamika Sploshnoi Sredy No.7, 94-104 (1971)
B3.34 V.E. Vdovin, LA. Kunin: A theory of elasticity with spatial dispersion.
Three-<limensional complex structure. J. Appl. Math. Mech. 30, 1272-
1281 (1966)

B4 Nonlinear Waves in Media with Dispersion

B4.l A Discussion on Nonlinear Theory of Wave Propagation in Dispersive
Systems, Proc. R. Soc.A 299, No. 1456 (1967)
B4.2 MJ. Ablowitz, D.S. Kaup, A.C. Newel, H. Segur: Method for solving
the Sine-Gordon equation. Phys. Rev. Lett. 30, 1262 (1973)
B4.3 MJ. Ablowitz, D.S. Kaup, A.S. Newell, H. Segur: The inverse scattering
transform - Fourier analysis for nonlinear problems. Studies in Appl.
Math. 53, 249 (1974)
B4.4 Backlund Tranjormations, the Inverse Scattering Method, Solitons and
their Applications, ed. by R.M. Miura, Lecture Notes Math., Vol. 515
(Springer, Berlin, Heidelberg, New York 1976)
B4.5 A.A. Belavin, V.E. Zakharov: Multi-<limensional method of the inverse
scattering problem and duality equations for the Yang-Mills fields.
JETP Lett. 25, 567-570 (1977)
B4.6 V.A. Belinskii, V.E. Zakharov: Integration of the Einstein equations by
means of the inverse scattering problem technique and construction of
exact soliton solutions. Sov. Phys. JETP 48, 985-994 (1978)
B4.7 A.S. Budagov: "A completely integrable model of a classical field
theory with non-trivial particles interaction in two-dimensional space-
time", in Voprosy Kvantovoi Teorii Polia i Statisticheskoi Fiziki
(Nauka, Leningrad 1978)
B4.8 F. Calogero, A. Degasperis: Nonlinear evaluation equations solvable by
the inverse spectral transform. Nuovo Cimento 32B, 201 (1976); 39B
1 (1977)
B4.9 J. Corones: An illustration of the Lie group framework for soliton
equations: generalization of the Lund-Regge model. J. Math. Phys.
19, 2431-2436 (1978)
B4.l 0 J. Corones: Solitons and simple pseudopotentials. J. Math. Phys. 17,
756-759 (1976)
260 Bibliography

B4.11 J. Corones: Solitons, pseudopotentials and certain Lie algebras. J. Math.

Phys.18, 163-164 (1977)
B4.12 1. Corones, B.L. Markovski, VA. Risov: A Lie group framework for
soliton equations. I. Path independent case. 1. Math. Phys. 18, 2207-
2213 (1977)
B4.13 J. Corones, B.L. Markovski, V A. Risov: Bilocal Lie groups and solitons.
Phys. Lett. 61A, 439-440 (1977)
B4.14 M. Crampin: Solitons and SL(2,R).Phys.Lett.66A, 17.0-172(1978)
B4.l5 M. Crampin, FA.E. Pirani, D.C. Robinson: The soliton connection.
Lett. Math. Phys.2, 15-19(1977)
B4.16 B.A. Dubrovin: Inverse problem for periodic finite-zones potentials in
the theory of scattering. Funct. Anal. Appl. 9, 61-62 (1975)
B4.17 B.A. Dubrovin, V.B. Matveev, S.P. Novikov: Non-linear equations of
Korteweg-de Vries type, finite-zone linear operations, and Abelian
varieties, Russian Math. Surveys 31, 59-146 (1976)
B4.18 BA. Dubrovin, S.P. Novikov: A periodicity problem for the Korteweg-
de Vries and Sturm-Liouville equations. Their connection with algebraic
geometry. Soviet Math. Dok1.15, 1597-1601 (1974)
B4.19 G. Eilenberger: Solitons, Springer Ser. Solid-State Sci., Vol. 19
(Springer, Berlin, Heidelberg, New York 1981)
B4.20 Exact Treatment of Nonlinear Waves, Prog. Theor. Phys. (Suppl.) 59,
1-161 (1976)
B4.21 C.S. Gardner, J.M. Green, M. Kruskal, R.M. Miura: Method for solving
the Korteweg-de Vries equation. Phys. Rev. Lett. 19, 1095-1097 (1967)
B4.22 I.M. Gel 'fand, L.A. Dikii: A Lie Algebra structure in a formal variational
calculus. Funct. Anal. App1.10, 16-22 (1976)
B4.23 I.M. Gel'fand, LA. Dikii: Asymptotic behavior of the resolvent of
Sturm-Lioville equations and the algebra of the Korteweg-de Vries
equations. Russian Math. Surveys 30, No.5, 77-113 (1975)
B4.24 I.M. Gel 'fand, L.A. Dikii: The calculus of jets and nonlinear Hamilto-
nian systems. Funct. Anal. Appl. 12, 81-94 (1978)
B4.25 I.M. Gel'fand, LA. Dikii: The resolvent and Hamiltonian systems.
Funct. Anal. Appl.ll, 93-105 (1977)
B4.26 R. Hermann: Pseudopotentials of Estabrook and Wahlquist, the geom-
etry of solitons and the theory of connections. Phys. Rev. Lett. 37,
835-836 (1976)
B4.27 R. Hirota: Exact N-soliton solutions of the wave equation oflong waves
in shallow-water and in nonlinear lattices. J. Math. Phys. 14, 810-814
(1973)
B4.28 B.B. Kadomtsev, V.I. Karpman: Nonlinear waves. Sov. Phys.-Uspekhi
14,40-60 (1971)
B4.29 V.I. Karpman: Non-linear Waves in Dispersive Media (Pergamon, Ox-
ford, New York 1975)
B4.30 I. Kay, U.E. Moses: Reflectionless transmission through dielectrics and
scattering potentials. J. Appl. Phys. 27, 1503-1508 (1956)
 Bibliography 261

B4.31 B.G. Konopelchenko: Completely integrable equations: dynamical

groups and their nonlinear realizations. J. Phys.A12, 1937-1949 (1979)
B4.32 I.M. Krichever: A1gebro-geometric construction of the Zakharov-Shabat
equations and their periodic solution. Sov. Math. Dokl. 17, 394-397
(1976)
B4.33 I.M. Krichever: Methods of algebraic geometry in the theory of non-
linear equations. Russian Math. Surveys 32, No.6, 185-213 (1977)
B4.34 E.A. Kuznetsov, A.V. Mikhailov: On the complete integrability of the
two-dimensional classical Thirring model. Theor. Math. Phys. 30, 193-
200 (1977)
B4.35 M. Lakshmanan: Geometric interpretation of solitons. Phys. Lett. 64A,
354-356 (1978)
B4.36 P.D. Lax: Integrals of nonlinear equations of evolution and solitary
waves. Commun. Pure Appl. Math. 21, 467 -490 (1968)
B4.37 P. Lax: Periodic solutions of the Korteweg-de Vries equations. Commun
Pure Appl. Math. 28,141-188 (1975)
B4.38 P. Lax: Periodic solutions of the Korteweg-de Vries equations. Lect.
Appl. Math. 15, 85-95 (1974)
B4.39 S.C. Lowell: Wave propagation in monoatomic lattices with anharmonic
potential. Proc. R. Soc. A318, No. 1532,93-106 (1970)
B4.40 F. Lund: Example of a relativistic completely integrable Hamiltonian
system. Phys. Rev. Lett. 38, 1178 (1977)
B4.41 F. Lund: Classically solvable field theory model. Ann. Phys. 115, 251-
268 (1978)
B4.42 Iu. I. Manin: Algebraic Aspects of Nonlinear Differential Equations
(VINITI, Moskva 1978)
B4.43 S.V. Manakov: Complete integrability and stochastization of discrete
dynamical systems. Sov. Phys.-JETP 40, 269-274 (1974)
B4.44 S.V. Manakov: Nonlinear Fraunhofer diffraction. SOy. Phys.-JETP 38,
693-696 (1974)
B4.45 S.V. Manakov: On the theory of two-dimensional stationary se1f-
focusing of electromagnetic waves. Sov. Phys.-JETP 38, 248-253
(1974)
B4.46 V.A. Marchenko: Spectral Theory of Sturm-Liouville Operators (Nau-
kova Dumka, Kiev 1972)
B4.47 V.A. Marchenko: The periodic Korteweg-de Vries problem. Math. of
the USSR - Sbomik 24,319-344 (1974)
B4.48 M.S. Marinov: Nonlinear wave equations and differential geometry.
Sov. J. Nucl. Phys. 28, 125-132 (1978)
B4.49 R.M. Miura: The Korteweg-de Vries equation: a survey of the results.
SIAM Rev. 18, 412-459 (1976)
B4.50 R.M. Miura, C.S. Gardner, M.D. Kruska1: Korteweg-de Vries equation
and generalization. II. Existence of conservation laws and constants
of motion. J. Math. Phys. 9,1204-1209 (1968)
262 Bibliography

B4.51 H.C. Morris: A prolongation structure for the AKNS system and its
generalization. J. Math. Phys. 18, 533-536 (1977)
B4.52 H.C. Morris: Prolongation structures and a generalized inverse scat-
tering problem. J. Math. Phys.17, 1867-1869 (1976)
B4.53 S.P. Novikov: The periodic problem for the Korteweg-de Vries equa-
tion. Funct. Anal. Appl. 8,236-246 (1974)
B4.54 N. Ooyama, N. Saito: On the stability of lattice solitons. Supplement
of the Progress of Theoretical Physics No. 45,201-208 (1970)
B4.55 A.C. Scott, F.Y. Chu, D.W. Laughlin: The soliton: a new concept in
applied science. Proc. IEEE 61, 1443-1483 (1973)
B4.56 A.B. Shabat: On non-scattering potentials in the Sturm-Liouville
equation. Dinamika Sploshnoi Sredy, No.5, 130 (1970)
B4.57 A.B. Shabat: On the Korteweg-de Vries equation. Sov. Math. Dokl.
14, 1266-1270 (1973)
B4.58 Solitons, ed. by R. Bullough, P. Caudrey, Topics Current Phys. Vol. 17
(Springer, Berlin, Heidelberg, New York 1980)
B4.59 L.A. Takhtadzhyan: Exact theory of propagation of ultrashort optical
pulses in two-level media. Sov. Phys.-JETP 39, 228-233 (1974)
B4.60 LA. Takhtadzhyan, L.D. Faddeev: Essentially nonlinear one-dimen-
sional model of classical field theory. Theor. Math. Phys. 21, 1046-
1057 (1974)
B4.61 S. Tanaka: On the N-tuple wave solutions of the Korteweg-de Vries
equation. Publ. Res. Inst. Math. Sci. Kyoto University 8, 419-427
(1972/1973)
B4.62 Theory of Solitons, ed. by S.P. Novikov (Nauka, Moscow 1980) [in
Russian]
B4.63 M. Toda: Waves in nonlinear lattice. Supplement of the Progress of
Theoretical Physics No. 45,174-200 (1970)
B4.64 M. Toda: Theory of Nonlinear Lattices, Springer Ser. Solid-State Sci.,
Vol. 20 (Springer, Berlin, Heidelberg, New York 1981)
B4.65 M. Wadafi, H. Sanuki, K. Kondo: Relationships among inverse methods,
Blicklung transformation and an infinite number of conservation laws.
Progr. Theor. Phys. 53,419-436 (1975)
B4.66 M. Wadati, M. Toda: The exact N-solution of the Korteweg-de Vries
equation. J. Phys. Soc. Jpn. 32 (1972)
B4.67 H.D. Wahlquist, F.B. Estabrook: Prolongation structures of nonlinear
evolution equations. J. Math. Phys. 16, 1-7 (1975); 17, 1293-1297
(1976)
B4.68 Y. Watanabe: Note on gauge invariance and conservation laws for a
class of nonlinear partial differential equations. J. Math. Phys. 15, 453-
457 (1974)
B4.69 N.J. Zabusky: Solitons and bound states of the time-independent
SchrOdinger equation. Phys. Rev. 168, 124-128 (1968)
B4.70 V.E. Zakharov: Kinetic equation for solitons. Sov. Phys.-JETP 33,
538-541 (1971)
 Bibliography 263

B4.7l V.E. Zakharov: On stochastization of one-dimensional chains of non-

linear oscillators. SOy. Phys.-JETP 38, 108-110 (1974)
B4.72 V.E. Zakharov: The Method of Inverse Scattering Problem, Lecture
Notes in Mathematics (Springer, Berlin, Heidelberg, New York 1978)
B4.73 V.E. Zakharov, L.D. Faddeev: Korteweg-de Vries equation: a comple-
tely integrable Hamiltonian system. Funct. Anal. Appl. 5, 280-287
(1971 )
B4.74 V.E. Zakharov, S.V. Manakov: Resonant interaction of wave packets in
nonlinear media. JETP Lett. 18,243-245 (1973)
B4.75 V.E. Zakharov, S.V. Manakov: The theory of resonance interaction of
wave packets in nonlinear media. SOy. Phys. JETP 42, 842-851 (1975)
B4.76 V.E. Zakharov, A.B. Shabat: A scheme for integrating the nonlinear
equations of mathematical physics by the method of the inverse scat-
tering problem 1. Funct. Anal. Appl. 8, 226-235 (1975)
B4.77 V.E. Zakharov, A.B. Shabat: Exact theory of two-dimensional self-
focusing and one-dimensional self-modulation of waves in nonlinear
media. SOy. Phys.-JETP 34, 62-69 (1972)
B4.78 V.E. Zakharov, L.A. Takhtadzhyan: Equivalence of the nonlinear
Schrodinger equation and the equation of a Heisenberg ferromagnetics.
Theor. Math. Phys. 38, 17-23 (1979)
B4.79 V.E. Zakharov, L.A. Takhtadzhyan, L.D. Faddeev: Complete descrip-
tion of solutions of Sine-Gordon equation. Dokl. Akad. Nauk SSSR
219, 1334-1337 (1974)
B4.80 A. Seeger: Solitons in crystals, in Continuum Models of Discrete Sys-
tems, ed. by E. Kroner, K.-H. Anthony and H.H.E. Leipholz (Univer-
y

sity of Waterloo Press, Waterloo 1980) pp. 253-327

BS Crystal Lattice
B5.l V.M. Agranovich, V.L. Ginzburg: Crystal-Optics with Spatial Disper-
sion and the Theory of Excitons. Springer Ser. Solid-State Sci., Vol. 42
(Springer, Berlin, Heidelberg, New York 1983)
B5.2 M. Born, K. Huang: Dynamical Theory of Crystal Lattices (Clarendon
Press, Oxford 1954)
B5.3 L. Brillouin, M. Parodi: Propagation des On des dans les Milieux Perio-
diques (Masson, Paris 1956)
B5.4 K. Huang: On the atomic theory of elasticity. Proc. R. Soc. A203,
No. 1072, 178-194 (1950)
B5.5 P.N. Keating: On the sufficiency of the Born-Huang relations. Phys.
Lett. A25, 496-497 (1967)
B5.6 A.M. Kosevich: Fundamentals of Crystal Lattice Mechanics (Nauka,
Moscow 1972)
B5.7 G. Leibfried: "Gittertheorie der mechanischen und thermischen Eigen-
schaften der Kristalle", in Encyclopedia of Physics, Vol. 7, Part 1
(Springer, Berlin 1955)
B5.8 A.N. Luzin: Some problems in the dynamics of crystal lattices and of
the theory of elasticity. SOY. Phys.-JETP 24,615-620 (1966)
264 Bibliography

B5.9 H.J. Maris, C. Elbaum: Validity of the Born-Huang relations in lattice

dynamics. Phys. Lett. A25, 96-97 (1967)
B5.10 A.A. Maradudin, E.M. Montroll, G.H. Weiss, I.P. Ipatova: Theory
of Lattice Dynamics in the Harmonic Approximation, Solid State
Physics, Suppl. 3, 2nd ed. (Academic, New York 1971)
B5.11 V.S. Oskotskii, A.L. Efros: On the theory of crystal lattices with
noncentraI. interatomic interaction. SOy. Phys.-Solid State 3, 448-457
(1961)
B5.12 J.M. Ziman: Electrons and Phonons (University Press, Oxford 1960)

B6 Dislocations and Local Defects

B6.1 B.A. Bilby: "Geometry and continuum mechanics", in [Ref. B2.1, pp.
180-199]
B6.2 A.D. Brailsford: Interaction between localized defects in an isotropic
elastic medium. J. Appl. Phys. 40, 3087-3088 (1969)
B6.3 A.D. Brailsford: Stress field of a dislocation. Phys. Rev. 142,383-392
(1966)
B6.4 A.H. Cottrell: Theory of Crystal Dislocations (Gordon and Breach, New
York 1964)
B6.5 J.D. Eshelby: "Elastic inclusions and inhomogeneities", in Prog. Solid
Mech., Vol. 2, ed. by LN. Sneddon, R. Hill (1961) pp. 88-140
B6.6 J.D. Eshelby: The elastic field outside an ellipsoidal inclusion. Proc.
R. Soc. A252, 561-569 (1959)
B6.7 J.D. Eshelby: The determination of the elastic field of an ellipsoidal
inclusion, and related problems. Proc. R. Soc. A241, 376-396 (1957)
B6.8 Fracture: an Advanced Treatise, Vol. 1-7, ed. by H. Liebowitz (Aca-
demic, New York 1968)
B6.9 J. Friedel: Les Dislocations (Gauthier-Villars, Paris 1956)
B6.10 V.L. Indenbom, A.N. Orlov: "Physical foundations of dislocation
theory", in [Ref. B2.1, pp. 166-179]
B6.11 V.L. Indenbom, A.N. Orlov: Physical theory of plasticity and strength.
SOy. Phys.-Uspekhi5, 272-291 (1962)
B6.12 N. Kinoshita, T. Mura: Elastic fields of inclusions in anisotropic media.
Phys. Stat. Sol. (A) 5, 759-768 (1971)
B6.13 A.M. Kosevich, V.D. Natsik: Dislocation damping in a medium having
dispersion of the elastic moduli. SOy. Phys.-Solid State 8, 993-999
(1966)
B6.14 V.S. Kosilova, LA. Kunin, E.G. Sosnina: Interaction of point defects
taking into account spatial dispersion. SOy. Phys.-Solid State 10, 291-
296 (1968)
B6.15 E. Kroner: Kontinuumstheorie der Versetzungen und Eigenspannungen
(Springer, Berlin, Gottingen, Heidelberg, 1958)
B6.16 E. Kroner: Allgemeine Kontinuumstheorie der Versetzungen und Ei-
genspannungen, Arch. Rat. Mech. Anal. 4,273-334 (1960)
B6.17 I.A. Kunin: Green's tensor for anisotropic elastic medium with internal
stress sources. Dok!. Akad. Nauk SSSR 157, 1319-1320 (1964)
 Bibliography 265

B6.l8 I.A. Kunin: Internal stress in an anisotropic elastic medium. J. Appl.

Math. Mech. 28, 612 (1964)
B6.19 I.A. Kunin: Fields due to arbitrary distributions of dislocations in an
anisotropic elastic medium. J. Appl. Mech. Tech. Phys. 6. 76 (1955)
B6.2U LA. Kunin: "Methods of tensor analysis in the theory of dislocations".
A supplement to the Russian translation of J.A. Schouten: Tensor Ana-
lysis for Physicists (Nauka, Moscow 1965) pp. 374-443. (The supple-
ment is available in English from the U.S. Department of Commerce,
Clearing house for Federal Sci. and Techn. Information, Springfield,
VA 22151)
B6.21 LA. Kunin, G.N. Mirenkova, E.G. Sosnina: An ellipsoidal crack and
needle in an anisotropic elastic medium. J. Appl. Math. Mech. 37, 501-
508 (1973)
B6.22 LA. Kunin, E.G. Sosnina: An ellipsoidal inhomogeneity in an elastic
medium. Dokl. Akad Nauk SSSR 199,571-574 (1971)
B6.23 LA. Kunin, E.G. Sosnina: Local inhomogeneities in an elastic medium.
J. Appl. Math. Mech. 34, 399-405 (1970)
B6.24 LA. Kunin, E.G. Sosnina: Stress concentration on an ellipsoidal in-
homogeneity in an anisotropic elastic medium. J. Appl. Math. Mech.
37, 287 -305 (1973)
B6.25 LA. Kunin: An algebra of tensor operators and its applications to elasti-
city. Int. J. Eng. Sci., 19,1551-1561 (1981)
B6.26 I.A. Kunin: "Projection operator method in continuum mechanics", in
Continuum Models of Discrete Systems 4 (North-Holland, Amsterdam
1981) pp. 179-187
B6.27 G. Leibfried, N. Breuer: Point Defects in Metals I, Springer Tracts Mod.
Phys., Vol. 81 (Springer, Berlin, Heidelberg, New York 1978)
B6.28 LM. Lifshifts, L.V. Tanatarov: On the elastic interaction of impurity
atoms in crystal. J. Phys. Metals Metallography 12, 331-338 (1961)
B6.29 Mechanics of Fracture, Vols. 1-4, ed. by G.C. Sih (Nordhoff, Leyden
1973)
B6.30 T. Mura: Continuum theory of plasticity and dislocations. Int. J. Eng.
Sci. 5,341-351 (1967)
B6.31 T. Mura, P.C. Cheng: The elastic field outside an ellipsoidal inclusion.
J. Appl. Mech. 44,591-594 (1977)
B6.32 T. Mura, T. Mori: Elastic fields produced by dislocations in anisotropic
media. Phil. Mag. 33, 1021-1027 (1976)
B6.33 F.R.V. Nabarro: The mathematical theory of stationary dislocations.
Adv. Phys. 1, 271-395 (1952)
B6.34 W. Noll: "Inhomogeneities in materially uniform simple bodies", in
[Ref. B2.l, pp. 239-246]
B6.35 RAAG Memoirs of the Unifying Study of Basic Problems in Engineer-
ing and Physical Sciences by Means of Geometry, ed. by K. Kondo
(Gakujutsu Bunken Fukyu-Kai, Tokyo, V.L 1955, V.II 1958, V.III
1962, V.IV 1968)
266 Bibliography

B6.36 D. Rogula: Influence of spatial acoustic dispersion on dynamical pro-

perties of dislocation. Bull. Akad. Pol. Sci., Ser. Sci. Techn. 13, 337-
585(1965)
B6.37 M.A. Sadowsky, E. Sternberg: Stress concentration around a triaxial
ellipsoidal cavity. J. Appl. Mech.16, 149-157 (1949)
B6.38 A.M. Vaisman, LA. Kunin: On boundary-value problems of the con-
tinuum theory of dislocations. Dokl. Akad. Nauk SSSR 173, 1024-
1027 (1967)
B6.39 H.G. van Bueren: Imperfections in Crystals (North-Holland, Amster-
dam 1960)
B6.40 V.E. Vdovin, LA. Kunin: Dislocation interactions taking into account
spatial dispersion. Sov. Phys.-Solid State 10, 297 -303 (1968)
B6.41 C.c. Wang: "On the geometric structure of simple bodies, a mathe-
matical foundation for the theories of continuous distributions of
dislocations", in [Ref. B2.1, pp. 247-250]
B6.42 L.T. Wheeler, LA. Kunin: On voids of minimum stress concentration.
Int. J. Sol. Struct. 18, 85-89 (1982)
B6.43 J.R. Willis: The stress field around an elliptical crack in an anisotropic
elastic medium. Int. J. Eng. Sci. 6, 253-263 (1968)
B6.44 R. de Wit: "Differential geometry of a nonlinear continuum theory of
dislocations" in [Ref. B2.1, pp. 251-261]
B6.45 H. Zorski: Statistical theory of dislocations. Int. J. Solids Struct. 4,
959-974 (1968)

B7 Random Fields of Inhomogeneities

B7.l G. Adomian: Linear random operator equations in mathematical phys-
ics. J. Math. Phys.12, 1944 (1971)
B7.2 A.E. Andreikiv, V.V. Panasiuk, M.M. Stadnik: Fracture of prismatic
bars with cracks. Problemy Prochnosti 10,10-16 (1972)
B7.3 B. Budjansky, R.J. O'Connel: Elastic moduli of a cracked solid. Int.
J. Solids Struct.12, No.2 (1976)
B7.4 LA. Chaban: Self-consistent field approach to calculation of effective
parameters of micro inhomogeneous media. Sov. Phys.-Acoust. 10,
298-304 (1964)
B7.5 R.J. Elliott, J.A. Krurnhansl, P.L. Leath: The theory and properties of
randomly disordered crystals. Rev. Mod. Phys. 46, 465 (1974)
B7.6 L.A. Fil'shtinskii: Interaction of a doubly-periodic system of recti-
linear cracks in an isotropic medium. J. Appl. Math. and Mech. 38,
853-861 (1974)
B7.7 E.L Grigoliuk, L.A. Fil'shtinskii: Perforated Plates and Shells (Nauka,
Moskow 1970)
B7.8 Z. Hashin, S. Shtrikman: On some variational principles in anisotropic
and nonhomogeneous elasticity. J. Mech. Phys. Solids 10,343 (1962)
B7.9 A.V. Hershey: The elasticity of an isotropic aggregate of anisotropic
cubic crystals. J. Appl. Mech. 21,236 (1954)
 Bibliography 267

B7.l0 R. Hill: A self-consistent mechanics of composite materials. J. Mech;

Phys. Solids 13, 213 (1965)
B7.11 M. Hori: Statistical theory of effective electrical, thermal and magnetic
properties of random heterogeneous materials. J. Math. Phys. 14, 514-
523; 1942-1948 (1973); 15,2177-2185 (1974); 16, 352-364; 1772-
1775 (1975); 18, 487-501 (1977)
B 7 .12 M. Kac: "Mathematical mechanisms of phase transitions", in Statistical
Physics. Phase Transitions and Superfluidity, Vol. 1, ed. by M. Chretien,
(Gordon & Breach, New York 1968) pp. 241-305
B7.13 S.K. Kanaun: "Random field of cracks in elastic solid medium", in:
"Issledovaniia po uprugosti i plastichnosti", No. 10 (Izd. Leningr. Univ.
Leningrad 1974)
B7.l4 S.K. Kanaun: Self-consistent field approximation for an elastic compo-
site medium. J. Appl. Mech. Tech. Phys. 18, 274-282 (1977)
B7.15 S.K. Kanaun: Self-consistent field method in the problem of effective
properties of an elastic composite. J. Appl. Mech. Tech. Phys. 16,
649-657 (1975)
B7.16 S.K. Kanaun, G.!. Iablokova: "Self-consistent field approximation in
plane problem for systems of interacting cracks", in: Mekhanika
sterzhnevykh sistem u sploshnykh sred No.9, (Leningr. Inshenerno-
stroitel'nyi Institut, Leningrad 1976) pp. 194-203
B7.17 S. Kirkpatrick: Percolation and conductivity. Rev. Mod. Phys. 45,574.
(1973)
B7 .18 G. Kneer: Dber die Berechnung des Elastizitatsmodules vielkristaller
Aggregate mit Textur. Phys. Stat. Sol. 9, 825-838 (1965)
B7.19 E. Kroner: Berechnung der elastischen Konstanten des Vielkristalls
aus den Konstanten des Einkristalls, Z. Physik 151,504 (1958)
D7.20 E. Kroner: Elastic moduli of perfectly disordered composite materials.
J. Mech. Phys. Solids 15,319 (1967)
B7.21 V.M. Levin: On calculation of effective moduli of composites. Dokl.
Akad. Nauk SSSR 220, No.5 (1975)
B7.22 E. Kroner: Bounds for effective elastic moduli of disordered materials.
J. Mech. Phys. Solids 25, 137-155 (1977)
B7.23 E. Kroner: Graded and perfect disorder in random media elasticity.
ASCE J. Eng. Mech. Div.l06, 889-914 (1980)
B7.24 V.M. Levin: "On calculation of microstress in polycrystals", in: Issle-
dovaniia po uprugosti i plastichnosti, No. 10 (Izd. Leningr. Univ.,
Leningrad 1974)
B7.25 V.M. Levin: On the stress concentration in inclusions in composite
materials. J. Appl. Math. Mech. 41, 753-761 (1977)
B7.26 B.R. Levin: Theoretical Foundations of Statistical Radioengineering,
(Sovetskoe Radio, Moskow 1969)
B7.27 P.R. Morris: Elastic constants of polycrystals. Int. J. Eng. Sci. 8,49
(1970)
268 Bibliography

B7.28 V.V. Panasiuk, M. Savruk, A.P. Datsishin: Stress Distribution Around

Cracks in Plates and Shells, (Naukova Dumka, Kiev 1976)
B7.29 P.e. Daris, G.C.M. Sih: "Stress analysis of cracks", in: Fracture Tough-
ness Testing and its Applications (Am. Soc. for Testing and Materials,
New York 1965) pp. 30-83
B7.30 T.D. Shermergor: Theory of Elastic Inhomogeneous Media (Nauka,
Moskow 1977)
B7.31 A.S. Vavakin, R.L. Salganik: On effective characteristics of inhomo-
geneous media with local inhomogeneities. lzv. Akad. Nauk SSSR,
MTT No.3 (1975)
B 7.3 2 L.T. Walpole: On bounds for the overall elastic moduli of inhomo-
geneous systems I, II. J. Mech. Phys. Solids 14,151 and 289 (1966)
B7.33 J.R. Willis: Bounds and self- consistent estimates for the overall proper-
ties of anisotropic composites. J. Mech. Phys. Solids 25, 185-202
(1977)
B7.34 J.R. Willis: A polarization approach to the scattering of elastic waves
I, II. J. Mech. Phys. Solids 28,287-305 and 307-327 (1980)

B8 References of General Nature

B8.1 N.I. Achieser, I.M. Glasmann: Theorie der linearen Operatoren im
Hilbert Raum, (Adademie, Berlin 1968)
B8.2 D. Bohm: Quantum Theory (Prentice-Hall, New-York 1952)
B8.3 H. Bremermann: Distributions, Complex Variables, and Fourier Trans-
forms (Addison-Wesley, Reading, MA 1965)
B8.4 E.T. Copson: Asymptotic Expansions (University Press, Cambridge
1965)
B8.5 A. Erdelyi: Asymptotic Expansions (Dover Publications, New York
1956)
B8.6 G.I. Eskin: Boundary- Value Problems for Elliptic Pseudo-Differential
Equations, (Math. Soc. of USA, N.Y. 1981)
B8.7 I.M. Gel'fand, S.V. Fomin: Calculus of Variations (Prentice-Hall,
Englewood Cliffs, NJ 1963)
B8.8 I.M. Gel'fand, G.E. Shilov: Generalized Functions (Academic, New
York 1964)
B8.9 A.O. Gel'fond: Calcul des Differences Finies (Dunod, Paris 1963)
B8.1O U. Gikhman, A.V. Skorokhod: Theory of Random Processes, Vol. 1,
(Nauka, Moscow 1971) [in Russian]
B8.11 I.S. Gradshtein, LM. Ryzhik: Table of Integrals, Series and Products,
4th ed. (Academic, New York 1965)
B8.12 D. Jackson: Fourier Series and Orthogonal Polynomials (The Mathema-
tical Association of America, Oberlin 1941)
B8.13 M.G. Kendall, P.A.P. Moran: Geometrical Probability (Hafner, New
York 1963)
B8.14 la. I. Khurgin, V.P. lakovlev: Localized Functions in Physics and
Engineering (Nauka, Moscow 1971) [in Russian]
 Bibliography 269

BS.15 M.G. Krein: Integral equations on semiaxis with kernels depending on

the difference of arguments. Usp. Mat. Nauk. 13, 3-120 (195S)
BS.16 I.A. Kunin: Quantum mechanical formalism in classical wave propaga-
tion problems. Int. J. Eng. Sci. 20, 271-2S1 (19S2)
BS.17 L.D. Landau, E.M. Lifshitz: Quantum Mechanics: Nonrelativistic
Theory, 3rd ed. (Pergamon, Oxford, New York 1977)
BS.IS P.D. Lax, R.S. Phillips: Scattering Theory (Academic, New York,
London 1967)
BS.19 B.Iak. Levin: Distribution of Zeros of Entire Functions (Am. Math.
Soc., Providence, RI 1964)
BS.20 G.Ia. Lubarski: Groups Theory and its Application in Physics (Fizmat-
giz, Moscow 1955) [in Russian]
BS.21 G. Matheron: Random Sets and Integral Geometry (Wiley, New York
1975)
BS.22 M. Morse, H. Feshbach: Methods of Theoretical Physics Vol. I, 2
(McGraw-Hill, New York 1953)
BS.23 A.H. Nayfeh: Perturbation Methods (Wiley, New York 1970)
BS.24 R.G. Newton: Scattering Theory of Waves and Particles (McGraw-Hill,
New York 1966)
BS.25 J .A. Schouten: Tensor Analysis for Physicists (Clarendon Press, Oxford
1951 )
BS.26 c.E. Shannon: The Mathematical Theory of Communication (Univer-
sity of Illinois Press, Urbana 1949)
BS.27 M.M. Vainberg: Variational Methods for the Study of Non-Linear
Operators (Holden-Day, San Francisco 1964)
BS.2S I.A. Kunin: "Lie group deformations and quantizations", in Trends in
Applications of Pure Mathematics to Mechanics (Pitman, London 1975)
Vol. 2, pp. 171-178
BS.29 LA. Kunin: "Group-theoretical foundations for interrelations between
physical models", in Continuum Models of Discrete Systems 3, ed. by
E. Kroner, K.-H. Anthony (University of Waterloo Press, Ontario 19S0)
pp.43-5S
Subject Index

Acoustical mode of vibration 57 Dislocation 144

Approximation curve 145
local 14 edge 145,160
long wavelength 54 screw 145,154
zeroth 28 Dislocation loop 149
Dispersion equation 57
Binary interaction 54 Dispersions
Born-Karman model 134 spatial and time 61
Boundary layer 28 Distortion
Boundary problems 41, 86 elastic 123
Burgers circuits 145 elastic external 123
Burgers vector 145 elastic internal 123, 124

Centre of dilatation 142 Effective elastic moduli tensor 166

Composites 165 Elastic energy 84
Concentration coefficient 101 Elastic moduli 74
Conservation laws 56 Ellipsoidal needle 104
Cosserat continuum model 62 Energy
Couple-stress theories 54 dislocation 17
Crack III interaction 119
ellipsoidal 104 Ergodic property 176
elliptic 115, 118
elliptic fields 193 Fluctuations of elastic fields 166
Cracked solid 165 Force constants 12,45
Cubic lattice 30
Four-order tensor of special structure
229
Debye model 11
Fourier transform (image) 7
Debye quasicontinuum 37, 73
Defect 88
local 68 Green's operator of elasticity 232
point 70, 96, 122, 136 for displacement 80, 233
point random field 209 for strain 82, 233
o-function 146 for stress 84, 233
Density Gyrotropic media 29
dislocations 146
energy 17 Hermiticity 13,19,49,132
external forces 81 Heterogeneous medium 166
force moments 82 Homogeneity 15
moments of dislocations 149 Hooke's law
quasidislocations 126 nonlocal 127
Discontinuity 86, 92 operator from 21
272 Subject Index

Impurity atom in a lattice 70 Parseval equality 8

Inhomogeneity 95, 104 Periodic functions 92
elliptical 96, 118 Point defect 70
random field 183 Poisson coefficient 134
Inhomogeneity of medium Poisson field
local 77,94 cracks 195
point 96 point defects 211, 241
Integral equations 89 Poly crystal 166
InteractIOn
energy 119 Quasicontinuum 6,53
many-particle 186 Radom lattice of defects 213,242
Internal degrees of freedom 57 Regular lattice of cracks 196
Invariance Relative elastic modulus 205
with respect to rotation 14 Rotation of the polarization plane 29
with respect to translation 14
Saint-Venant compatibility con-
Kernel of operator 9 ditions 128
Self-consistent method (SCM) 166
Lagrangian 13,48,62 Stability 14
Lame constants 74 Statistical moments 166, 177
Lattice knots 6 Stochastic fields of cracks 195
Layer Strain 20
double 90 Stress
single 90 external 127
Line of dislocations 145 internal 122
Local defects 68 Stress concentration 102
Stress intensity factor 114, 200
Media System of collinear cracks 208
homogeneous 22,80 System of point defects 75, 141
with binary interaction 15,42
with sources of internal stress 127 Tensor
Method of effective field 167 elastic compliance 90
Metric external strain 128
external 128 internal strain 130
internal 128 internal stress 122
Microdeformation 48 Green's 16,35
Micromoments 48 Green's dynamic 16, 71
Microrotations 48 Green's for internal stress 131
Green's static 16,35
Nonlocal theory 122 strain 20
stress 17
Operator Theory
elastic compliance 127 continuum dislocations 144
elastic energy 13 couple-stress 54
elastic moduli 21 elasticity, classic 56
energy 49 elasticity with constrained rota-
Hooke's law 29 tions 21
projection 35, 83, 94 Trinary interaction 54
Operator's kernel 9
Optical mode of vibrations 61 Unit antisymmetric pseudo tensor 62
Solitons and Condensed Solitons
Matter Physics Editors: R.K.Bullough, P.J.Caudrey
Proceedings of the Symposium on Non- 1980.20 figures. XViII, 389 pages
linear (Soliton) Structure and Dynamics in (Topics in Current Physics, Volume 17)
Condensed Matter, Oxford, England, June ISBN 3-540-09962-X
27-29,1978 Contents: RKBullough, P.J.Caudrey: The
Editors: A.R.Bishop, T.Schneider Soliton and Its History. - G.L.Lamb Jr.,
D. W.McLaughlin: Aspects of Soliton
Revised 2nd printing. 1981. 120 figures. Physics. - RKBullough, P.J.Caudrey,
XI, 342 pages. (Springer Series in Solid- H.M. Gibbs: The Double Sine-Gordon
State Sciences, Volume 8) Equations: A Physically Applicable System
ISBN 3-540-09138-6 of Equations. - M. Toda: On a Nonlinear
"....It is of an unusually high quality. The Lattice (The Toda Lattice). - RHirota:
contributing authors are clearly authorities Direct Methods in Soliton Theory. -
in their fields, and try to convey new infor- A C. Newell: The Inverse Scattering Trans-
mation and insight in the psychology and form. - V.E.Zakharov: The Inverse Scatte-
sociology of solitons instead of repeating a ring Method. - M. Wadati: Generalized
well-trodden script under the milky gaze of Matrix Form of the Inverse Scattering
a sleep audience..... one has the feeling that Method. - F.Calogero: ADegasperis:
the authors tried to put their own speciality Nonlinear Evolution Equations Solvable by
in the language or the context of condensed the Inverse Spectral Transform Associated
matter physics, and that (with some home- with the Matrix SchrMinger Equation. -
work by the reader) genuine transmission of S.P.Novikov: A Method of Solving the
information has been achieved. Periodic Problem for the KdV Equation and
This is sufficiently rare in proceedings of Its Generalizations. - L.D.Faddeev: A
meetings as to be commended, and this is a Hamiltonian Interpretation of the Inverse
good book to have in one's departmental Scattering Method. - AH.Luther:
library." Contemporary Physics Quantum Solitons in Statistical Physics. -
Further Remarks on John Scott Russel and
on the Early History of His Solitary Wave. -
G. Eilenberger Note Added in Proof. - Additional
References with Titles. - Subject Index.
Solitons
Mathematical Methods for Physicists Structural Stability
1981. 31 figures. VIII, 192 pages
(Springer Series in Solid-State Sciences, in Physics
Volume 19). ISBN 3-540-10223-X Proceedings of Two International Symposia
Contents: Introduction. - The Korteweg-de on Applications of Catastrophe Theory and
Vries Equation (KdV-Equation). - The Topological Concepts in Physics. Tiibingen,
Inverse Scattering Transformation (1ST) as Federal Republic of Germany, May 2-6 and
Illustrated with the KdY. - Inverse Scatte- December 11-14, 1978
ring Theory for Other Evolution Equations. Editors: W.Giittinger, H.Eikemeier
- The Classical Sine-Gordon Equation
(SGE). - Statistical Mechanics of the Sine- 1979.108 figures, 8 tables. VIII, 311 pages
Gordon System. - Difference Equations: (Springer Series in Synergetics, Volume 4)
The Toda Lattice. - Appendix: Mathema- ISBN 3-540-09463-6
tical Details. - References. - Subject Index. " ....It shows that, in physics, Thorn's ideas
are far from trivial, and that to express them
rigorously in a novel context is a difficult
problem whose solution is no mean achie-
vement. The book should be in every
library, and should be put in every hand .... "
Optica Acta

Springer-Verlag Berlin Heidelberg New York Tokyo

Positrons in Solids D.C.Mattis

Editor: P.Hautojarvi The Theory

1979. 66 figures, 25 tables. XIII, 255 pages
(Topics in Current Physics, Volume 12) of Magnetism I
ISBN 3-540-09271-4
Statics and Dynamics
Contents: P. Hautojiirvi, A. Vehanen: 1981. 58 figures. XV, 300 pages
Introduction to Positron Annihilation. - (Springer Series in Solid-State Scienes,
P.E. Mijnarends: Electron Momentum Densi- Volume 17)
ties in Metals and Alloys. - RN. West: Posi- ISBN 3-540-10611-1
tron Studies of Lattice Defects in Metals. -
RM. Nieminen, M.J. Manninen: Positrons in "... Mattis's relentless approach will not be to
Imperfect Solids: Theory. - A. Dupasquier: everyone's taste, but his book will be a useful
Positrons in Ionic Solids. addition to the library of anyone deeply
interested in the origins of magnetism and the
careful study of mathematical models. The
statistical mechanician, or the particle theorist
looking for hints on how to solve the lattice
M.Toda gauge theory problem, may, however, prefer
to wait for the second volume which will cover
Theory of thermodynamics and statistical mechanics.

Nonlinear Lattices Finally, praise must be given for the introduc-

tory chapter, 38 pages long, which spells out
1981. 38 figures. X, 205 pages. (Springer the history of magnetism from the earliest
Series in Solid-State Sciences, Volume 20) days to the present, places it in the perspective
ISBN 3-540-10224-8 of the general evolution of physics and the
development of Westem thought, and is
Contents: Introduction. - The Lattice with backed up by marvellous quotations and an
Exponential Interaction. - The Spectrum and impressive bibliography. This chapter can be
Construction of Solutions. - Periodic Systems. strongly recommended to anyone interested
- Application of the Hamilton-Jacobi Theory. in the history fo science and, almost alone,
- Appendices A-J. - Simplified Answers to would justify purchase ofthe book." Nature
Main Problems. - References. - Biblio-
graphy. - Subject Index. - List of Authors
Cited in Text.
The Structure and
Properties of Matter
Editor: T.Matsubara
With contribution by numerous experts
1982. 229 figures. XI, 446 pages
(Springer Series in Solid-State Sciences,
Volume 28)
ISBN 3-540-11 098-4
Springer-Verlag Contents: Atoms as Constituents of Matter. -
Berlin System of Protons and Electrons. - Helium. -
Superfiuid Helium 3. - Matals. - Nonmetals.
Heidelberg - Localized Electron Approximation. -
Magnetism. - Magnetic Properties of Dilute
New York Alloys - the Kondo Effect. - Random

Tokyo
Systems. - Coherent Protential Approxima-
tion (CPA). - References. - Subject Index.