Springer Tracts in Natural

Philosophy

Volume 35

Edited by C. Truesdell
Springer Tracts in Natural Philosophy
Vol. 1 Gundersen: Linearized Analysis of One-Dimensional Magnetohydrodynamic Flows.
With 10 figures. X, 119 pages. 1964.

Vol. 2 Walter: Differential- und Integral-Ungleichungen und ihre Anwendung bei Abschatzungs-
lind Eindeutigkeitsproblemen
Mit 18 Abbildungen. XIV, 269 Seiten, 1964.

Vol. 3 Gaier: Konstruktive Methoden der konformen Abbildung

Mit 20 Abbildungen und 28 Tabellen. XIV, 294 Seiten. 1964.

Vol. 4 Meinardus; Approximation von Funktionen und ihre numerische Behand1ung

Mit 21 Abbildungen. VIII, 180 Seiten, 1964.

Vol. 5 Coleman, Markovitz, Noll: Viscometric Flows of Non-Newtonian Fluids.

Theory and Experiment
With 37 figures. XII, 130 pages. 1966.

Vol. 6 Eckhaus: Studies in Non-Linear Stability Theory

With 12 figures. VIII, 117 pages. 1965.

Vol. 7 Leimanis: The General Problem of the Motion of Coupled Rigid Bodies About a Fixed
Point
With 66 figures, XVI, 337 pages. 1965.

Vol. 8 Roseau: Vibrations non lineaires et theorie de la stabilite

Avec 7 figures. XII, 254 pages. 1966.

Vol. 9 Brown: Magnetoelastic Interactions

With 13 figures. VIII, 155 pages. 1966.

Vol. 10 Bunge: Foundations of Physics

With 5 figures. XII, 312 pages. 1967.

Vol. II Lavrentiev: Some Improperly Posed Problems of Mathematical Physics

With 1 figure. VIII, 72 pages. 1967.

Vol. 12 Kronmuller: Nachwirkung in Ferromagnetika

Mit 92 Abbildungen. XIV, 329 Seiten. 1968.

Vol. 13 Meinardus: Approximation of Functions: Theory and Numerical Methods

With 21 figures. VIII, 198 pages. 1967.

Vol. 14 Bell: The Physics of Large Deformation of Crystalline Solids

With 166 figures. X, 253 pages. 1968.

Vol. 15 Buchholz: The Confluent Hypergeometric Function with Special Emphasis on its
Applications
XVIII, 238 pages. 1969.

Vol. 16 Slepian: Mathematical Foundations of Network Analysis

XI, 195 pages. 1968.

Vol. 17 Gavalas: Nonlinear Differential Equations of Chemically Reacting Systems

With 10 figures. IX, 107 pages. 1968.

Vol. 18 Marti: Introduction to the Theory of Bases

XII, 149 pages. 1969.
 G. Capriz

Continua
with Microstructure

Springer-Verlag
New York Berlin Heidelberg
London Paris Tokyo
 Gianfranco Capriz
Dipartimento di Matematica
Universita
Pisa 56100, Italy

Mathematics Subject Classification (1980): 73B25, 73S99

Library of Congress Cataloging-in-Publication Data

Capriz, Gianfranco.
Continua with microstructure/Gianfranco Capriz.
p. cm.-(Springer tracts in natural phi}osophy; v. 35)
"Lectures given in 1986 at the Scuola Estiva di Fisica Matematica
in Ravello (ltaly),,-Pref.
Bibliography: p.
ISBN-13: 978-1-4612-8166-5
1. Continuum mechanics. 1. Title. II. Series.
QA808.2.C33 1989
531-dc19 88-8483

Printed on acid-free paper.

© 1989 by Springer-Verlag New York Inc.

Softcover reprint of the hardcover 1st edition 1989
All rights reserved. This work may not be translated or copied in whole or in part
without the written permission of the publisher (Springer-Verlag, 175 Fifth Avenue,
New York, NY 10010, USA), except for brief excerpts in connection with reviews or
scholarly analysis. Use in connection with any form of information storage and re-
trieval, electronic adaptation, computer software, or by similar or dissimilar method-
ology now known or hereafter developed is forbidden.
The use of general descriptive names, trade names, trademarks, etc. in this publication,
even if the former are not especially identified, is not to be taken as a sign that
such names, as understood by the Trade Marks and Merchandise Marks Act, may
accordingly be used freely by anyone.

Typeset by Asco Trade Typesetting Ltd., Hong Kong.

98 76 54 32 1

ISBN-13: 978-1-4612-8166-5 e-ISBN-13: 978-1-4612-3584-2

DOl: 10.1007/978-1-4612-3584-2
To Barbara
 Preface

This book proposes a new general setting for theories of bodies with
microstructure when they are described within the scheme of the con-
tinuum: besides the usual fields of classical thermomechanics (dis-
placement, stress, temperature, etc.) some new fields enter the picture
(order parameters, microstress, etc.). The book can be used in a
semester course for students who have already followed lectures on
the classical theory of continua and is intended as an introduction to
special topics: materials with voids, liquid crystals, meromorphic con-
tinua. In fact, the content is essentially that of a series of lectures
given in 1986 at the Scuola Estiva di Fisica Matematica in Ravello
(Italy).
I would like to thank the Scientific Committee of the Gruppo di
Fisica Matematica of the Italian National Council of Research (CNR)
for the invitation to teach in the School. I also thank the Committee
for Mathematics of CNR and the National Science Foundation: they
have supported my research over many years and given me the
opportunity to study the topics presented in this book, in particular
through a USA-Italy program initiated by Professor Clifford A.
Truesdell.
My interest in the field dates back to a period of collaboration
with Paolo Podio-Guidugli and some of the basic ideas came up
during our discussions.
Successive versions of the text of the lectures, in Italian, were
circulated among friends and colleagues, who have offered welcome
viii Preface

comment and criticism; I am grateful in particular to Paolo Podio-

Guidugli, Piero Villaggio an~ Epifanio Virga. My thanks are due also
to my secretary Mrs. Tao Pei Lin, who has helped me with her usual
competence and dedication.
 Contents

Preface vii

§1 IntrodQction

Part I General Properties 5

§2 The Model for the Microstructure 5
§3 The Notion of Observer 8
§4 Continua with Microstructure 10
§5 Invariance Properties 13
§6 Conservation of Mass: Kinetic Energy 15
§7 Inertia 18
§8 Dynamic Equations of Balance 21
§9 Balance of Moment of Momentum 23
§1O Boundary Conditions: Change of Variables 25
§1l The Conservative Case in Statics 27
§12 Perfect Fluids with Microstructure 30
§13 Rules of Invariance and the Balance of Moment of
Momentum: Variational Principles in Dynamics 34
§14 Internal Constraints: Continua with Latent Microstructure 36

Part II Special Theories 42

§15 Continua with One-Dimensional Microstructure:
Continua with Voids 42
§16 Liquids with Bubbles 44
§17 Dilatant Granular Materials 46
x Contents

§18 The Perfect Korteweg Fluid 47

§19 Continua with Vectorial Microstructure 49
§20 Uniaxial Liquid Crystals 54
§21 Continua with Affine Microstructure 57
§22 Micromorphic Elastic Continua: Bodies with Continuous
Distribution of Dislocations 59
§23 The Continua of Cosserat 62
§24 Biaxial Nematic Liquid Crystals 64

Part III Thermodynamics 67

§25 Balance Equations 67
§26 Interpretation of the Equations of Balance 69
§27 Thermodynamics of Continua with Latent Microstructure 71
§28 Comparison with the Traditional Class of
Hyperelastic Bodies 74

Part IV Mathematical Problems Posed by the Theory 77

§29 The Influence of the Topological Properties of the
Manifold Jt 77
§30 Further Remarks on the Topological Theory of Defects 81
§31 Existence of Singular Solutions in Statics 82
§32 Phase Transitions 84
§33 Droplets of Perfect Liquids with Microstructure 87

Appendix
90
 §1. Introduction
The continuum with microstructure is a refined mathematical model
for a wide class of material bodies endowed with some sort of local
microscopic order, a model that preserves the well established advan-
tages accruing from the classical scheme of the continuum.
Actually, the variety of physical phenomena observed and the wealth
of specific mathematical tools invoked to represent them seem at first
to deny the possibility for a global approach, in some way similar to
that which in the classical context precedes the study of special
theories (of fluids, of hyperelastic bodies, of perfect gases, etc.) and
puts in evidence common properties. But recent work shows that such
a global approach is possible. The first part of this book describes the
proposed route and expands on general axioms and theorems. Special
properties valid in particular contexts (e.g., media with voids, liquid
crystals, Cosserat continua) are either proposed or derived in the
second part. Thermodynamic questions are discussed in the third part.
In Part IV some mathematical problems are stated which arise within
the theory.
The notation, as far as possible, is standard. IR is the set of real
numbers, tff the three-dimensional Euclidean space, "f/ the translation
space of tff (the space of three-component vectors), Lin the space of
the linear mappings of "f/ into itself (the space of second-order ten-
sors), and Sym (Skw, Sph, Dev) the subspace of symmetric (skew,
2 §1. Introduction

spherical, traceless) tensors in Lin. Orth is the group of orthogonal

tensors in Lin; Lin+ (Orth+) the group of (orthogonal) tensors with
positive determinant.
Greek letters are used for scalars, lowercase latin letters for vectors,
capital letters for second-order tensors, boldface lowercase and capital
letters for tensors of the third and fourth order, respectively; in parti-
cular, I is the identity tensor and e is Ricci's permutation tensor. The
same letters with indices are used to indicate orthogonal cartesian
components of vectors and tensors; indices are generally lowercase,
but when double vectors are involved capital letters are also used.
To bow to tradition, the only exception regarding components is
made for the unit tensor, for which the Kronecker notation bij is pre-
ferred. The convention that repeated indic~s are summed is adopteq
throughout.
The scalar, vector, tensor products of two vectors u and v are
denoted by u· v, u x v, u ® v, respectively. For other operations in-
volving tensors, in particular tensors of different order, the following
conventions are adopted:
(i) The result of the transformation of a vector v by a tensor A is
denoted by Av .or, using components, Aijvj . There is a similar
convention for the result W of the transformation of a second-
order tensor V by a fourth-order tensor A, W = AV; or, in com-
ponents, W;j = A ijhk l'hk·
(ii) The notion of scalar and tensor product applies also to tensors; in
particular, one has, for instance, A· B = AijBij and (u ® A)ijk =
uiAjk ·
(iii) The composition AB of two second-order tensors A, B is the
tensor with components Cij = AihBhj .
(iv) For the compositions of tensors of different order it is agreed that
the lower-order tensor is on the right and all its indices are
saturated. For instance: (ev);j = eijkvk; (eA); = eijkAjk (note the order
of saturation); (e(u ® v)); = e;jkUjVk = (u x v);.
The linear operators sym, skw, sph, and dev when applied to a
second-order tensor lead, respectively, to its symmetric, skew, spheri-
cal and traceless part; the operator tr extracts from a second-order
tensor its scalar trace. The operator det extracts its determinant. The
exponent T indicates transposition. On tensors of the third order one
can operate a major transposition (exponent T) or a minor right or
left transposition (exponent t) with the following properties:
 §1. Introduction 3

«aa)b)c = «aTc)b)a,
(aa)b = (atb)a, «aa)b)c = «taa)c)b, 'Va, b, c E "Y.
Frequent use will be made of the identities
ejr.eipq = brpbsq - brqb.p, (1.1)
eirsejpqApq = 2(skw A)r., (1.2)
and
(1.3)
We recall also the relations that apply between any proper ortho-
gonal tensor Q and the vector q of the rotation associated with Q,
that is, the vector that has the direction of the axis of rotation, the
appropriate orientation and modulus equal to the angle of rotation (J
(Iql = (J). The main relation is
Q=I + sin (J W + (1 - cos (J)W2, (1.4)
where
W = -ec, with q = (Jc.

The verification of (1.4) is trivial, if components are evaluated on

a frame with an axis parallel to c. If one notices further that
W 2 = c ® c - I, W 3 = - W, W 4 = - W 2, etc. (1.5)
and introduces for sin (J and cos (J the corresponding developments in
power series of (J, one is led formally to the exponential series in - eq
Q = I - eq + !(eq)(eq) - ... ; (1.6)
this fact suggests the compact notation
Q = e-eq, (1.7)
where e is the basis of natural logarithms.
As we shall see, an essential role is played in this book by a
connected differential manifold Jt of finite dimension m; bold greek
letters such as v, are used for the elements of Jt. On a local chart,
coordinates are denoted by a superscript greek letter as, for instance,
vlX • The same notation is used for vectors in the tangent space ff"Jt,
whereas for components of vectors of the cotangent space fTy* Jt the
index is a subscript, as, for instance, ~. The notation for linear opera-
tions, products, and contractions involving these vectors is modeled
4 §1. Introduction

on the choices already made for vectors of.y and tensors of Lin. For
instance, if v(.) is a smooth mapping of (an interval of) ~ into Jt,
then dv/d. is a vector of [T"JI which is denoted by v (with com-
ponents \1") If €p(v) is a smooth mapping of (a subset of) JI into ~,
then ~:= o€p/ov is a vector of !Tv· JI and its components are denoted
by ~IZ. Then the derivative of the composition cP(.):= €p(v(.» can be
written as the scalar product
r,b = ~ . v = ~ \1".
We recall also that, given a vector field J1(v) on JI with J1(v) e
!TvJt, the "flow generated by the field" is defined as the set of integral
curves v(.) of the differential equation·
dv ~
d. = J1(v), v(O) = vo, Vo e JI.

The flow is said to represent the action on JI of the group of real

numbers, which has J1(v) as infinitesinial generator.
Given a continuous group !l of transformations Q on Jt, a defini-
tion (convenient for later developments) of the infinitesimal generator
of the group action is as follows. Suppose that each Q of !l be in
one-to-one correspondence with a finite set of real numbers qt at least
in the neighborhood of the unit element of !l, so that such element
corresponds to the null value of all qt. Then the infinitesimal generator
is the operator (dv(q)/dq)lq=o, if v(q) is the element of JI obtained by
transformation of v through Q.
In the following paragraphs we need often to express rules of
invariance against changes ensuing from a group of transformations.
A class of theorems is available, of which the simplest version is given
below.
Let €p(v) be a real-valued function defined over JI; suppose that €I'
is invariant under !l; that is, suppose that
Vv e Jt, VQ e .fl. (1.8)
Then the following relation applies:

( d€p lZ dv(q») = O. VveJl; (1.9)

dv dq q=O '
The reverse is also true: when (1.9) holds over the whole of Jt, the
invariance property (1.8) applies.
The notation hints at the special case, of paramount interest here,
when .fl is isomorphic with the group of proper orthogonal tensors.
 PART I
General Properties

§2. The Model for the Microstructure

In the classical theory of continua a body mis thought of as a set of

"material elements" x; each element has a distinct identity and occu-
pies at each instant 't" an exclusive place x within the closure of an
open set of the euclidean space Iff, so that one can identify x also by
assigning its place x* in a placement of m, conventionally chosen as
reference. It is implied that nothing which could be geometrically
interesting would be perceived by a finer observation of the element:
details, if any, are chaotic.
For a large class of bodies these preconceptions are justified; but
there are also cases, those we want to consider here, when a closer
look at the element reveals at least a partial microscopic order. Then
it is natural to associate with each x a certain number (say, m) of
"order parameters" v'" (a = 1,2, ... , m), so as to characterize the ob-
served microstructure. Liquid crystals are an emblematic case; they
are fluids made up of molecules which are rodlike in shape and rather
rigid, so that locally a preferred direction is in evidence; then there
are two order parameters, azimuth and elevation of the preferred
direction.
Examination of many special cases shows that the parameters v'"
are best interpreted as coordinates (in a local chart of an atlas) of an
element v in an appropriate manifold J{ of dimension m.
In this Part general axioms are stated and basic properties are
6 I. General Properties

derived, valid for any choice of .;/I. However, before we begin, we list a
number of important examples.

(i) Continuum with voids (more precisely: continuum with finely

dispersed spherical voids). The dimension of ..I( is 1; actually, for
Jt, an appropriate interval [0, v) of real numbers may be taken,
v < 1. Then v, which is now a scalar, can be interpreted as the
void fraction and v as the critical value of v for which the
continuum becomes disconnected (for instance, v = (j3/8)n, for a
compact array of spherical voids).
(ii) Liquid with nondifTusing gas bubbles. As above, though now it is
appropriate to choose for ..I( the closed interval [0, 1], to com-
prise also single phase conditions.
(iii) Continuum with planar spin. The microstructure is specified by a
unit vector in a material surface; the order parameter can be an
angle in [0, 2n) formed by the vector with a material line tangent
to the surface. Because of periodicity, ..I( is now isomorphic with
a circle.
(iv) Liquid crystals. They are characterized locally by a preferred
direction which can be specified by a unit vector d. Because
orientation is irrelevant, more properly the set of diads d ® d can
be taken for the manifold. ..I( is isomorphic with the projective
plane, or with a spherical surface where the antipodes are identi-
fied.
(v) Continuum with ordinary spin. v is an arbitrary unit vector, of
which, in contrast with liquid crystals, also the orientation is
relevant. ..I( is the complete spherical surface.
(vi) Cosserat continuum. The microstructure is described by a triad of
oriented orthogonal axes; hence ..I( can be taken to coincide with
the set Orth + of proper orthogonal tensors R (R being the tensor
which sends a fixed orthogonal triad onto the triad which repre-
sents the microstructure). Sometimes one takes for ..I( a ball of
radius n, for which the antipodes on the boundary are identified:
the vector r with origin at the center of the ball and head at any
point of the ball is the rotation vector associated with R; if
Irl = n, then the effects of the rotations rand -r are the same.
Actually an isomorphism could be stated also with the projective
space. We remark finally that one could use, as order parameters,
couples (u, v) of orthogonal unit vectors (lui = Ivl = 1, u"v = 0)
which specify two of the three axes of the triad, and hence the
whole triad.
 §2. The Model for the Microstructure 7

(vii) Biaxial nematic. The microstructure has the symmetries of a

parallelopiped with unequal edges and can be described by a
symmetric tensor having three ftxed positive eigenvalues, differ-
ent one from the other. vii can be obtained from the manifold
unGer (vi) with appropriate identiftcations.
(viii) Continuum with vector microstructure. vii coincides with the
whole "1""; this case is interesting, at least from a theoretical point
of view, because of the possible comparison and contrast with
cases (vi) and (vii).
(ix) Micromorphic continuum. The microstructure is described by a
nondegenerate ellipsoid with oriented axes; one can choose for
vIIthe set Lin+ of second-order tensors with positive determinant.
(x) Bodies with continuous distribution of dislocations. The disloca-
tion density is locally deftned by a tensor B of Lin. On any
plane of unit normal n, the Burgers vector b is given by b = Bn.
B may be null and is often of low rank.
There are continua for which the microscopic ordering has kinetic
rather than geometric character; mixtures for instance. In that case a
ftner observation of a mixture element shows the existence of separate
components each with its own peculiar velocity, so that phenomena of
diffusion are observed. Then order parameters may be used to record
diffusion velocities; note also that the observation has local rather
than material charaCter. Thus, notwithstanding certain formal analo-
gies, there are substantial differences that strongly suggest a separate
study. We observe that, in the liquids with bubbles quoted under (ii),
diffusion was explicitly ruled out.
Even more delicate problems arise in the study of the behaviour of
liquid Helium II, i.e. of superfluid liquid helium; then order para-
meters are used to specify velocity potentials. Naturally, if interest is
limited to a study of the topological properties of the manifold vH,
there may be no formal differences from examples specifted above.
Therefore, also for their special interest in the topological theory of
defects, we state here two further examples
(xi) Normal superfluid helium (i.e., 4He). It is formally identical with
example (iii): the order parameter is a complex scalar fteld '" of
constant modulus and arbitrary phase v.
(xii) Superfluid helium isotope (i.e., 3He). It is a very complex sub-
stance with different phases. From the point of view of the
properties of vii what is called the dipole-locked A phase can be
considered a Cosserat continuum as in (vi) above. Instead, the
8 I. General Properties

dipole-free A phase requires also the assignment of a direction

(that of a unit vector w, say), but with the following condition. If
(u, v) is the ordered couple indicated under case (vi), then only
the diads u ® wand v ® w count, a simultaneous change of sign
in all three vectors being irrelevant. The dimension of .H is five;
.H can be considered as the product of a unit sphere (to represent
w) by the ball of radius 11:, but with all the required identifica-
tions. As can be guessed, the topological properties of .H are
complex with physically very interesting consequences.

§3. The Notion of Observer

As in the classical theory, the assumption will be accepted here

that the set fJi of the places x occupied in tS at a given time -r by
the elements of a body ~ be the closure of an open connected set with
smooth boundary ofJi. Over fJi the field v is assigned; for this assign-
ment it is advisable, especially when following time-dependent pro-
cesses, to have in mind a fixed copy of .H and perhaps even a
particular atlas on .H. Then one can imagine that
v(x, -r), x E fJi, -r E [0, 'f], (3.1)
be also a compact notation for the m-tuple of real-valued functions
VIZ (x, -r).

Often it is convenient to assign the values of v for each material

element, rather than at a given place; and each element, as already
mentioned, can be singled out through its place x* in a reference
placement fJi* of the body
~(x*' -r), x* E fJi*, -r E [0, f]. (3.2)
The choice of the copy of .H must be guided also by the need to
give a significance, absolute in a certain sense, to the measure of the
time rate of change v = o~/o-r. Furthermore the notion of privileged
observer must be introduced not only for the macromotion but also
for microstructural processes and the hypothesis must be accepted
that only such observers are able to achieve an absolute measure of
the order parameters, so that, as we shall see, certain simple expres-
sions in terms of v apply for the density of kinetic energy and of the
power of internal actions. Thus, the concepts of framing and of
inertial frame of reference must be extended to comprise microstruc-
tural events.
 §3. The Notion of Observer 9

Any two observers can be brought to coincide at any given time

through a rigid body displacement. It is assumed here throughout that
a translation leaves the measure of v unaffected, whereas two observers
differing by a rotation (of vector q) may read different values v and
v(q) of the'order parameters. Here, for consistency, q is interpreted as

the vector of the rotation which reduces the second observer (reading
the value v(q») to the first.
Consider the development

v(q) = v + ( dq I
dV(q») q=O [q] + o(q), (3.3)

where the derivative (dv(q)/dq)lq=o acts as an operator on elements of

"f/ into the tangent space f/"Jt, and introduce the notation

d(v):= (dV(q»)
dq
Iq=O . (3.4)

d(v) is the infinitesimal generator of the local action on Jt of the

group f!l of the proper orthogonal tensors Q = e-eq. With the use of a
cartesian reference for "f/ and a local chart for Jt, the components
da. j of d may be defined. These components may be used also to
introduce an m-tuple of vectors a(a.) of "f/
a!a.) = da. j • (3.5)
When v(q) belongs to the same chart as v (a situation which certainly
occurs when q is sufficiently small), then
v(q) = va. + a(a.)· q + o(q). (3.6)

Remark. Let us refer back to the examples of Section 2, and deter-

mine in some cases the particular form taken by the operator d. In
examples (i) and (ii) it is obvious that all observers read the same
value for v; hence d vanishes identically.
In case (viii), if it is assumed that the absolute orientation of the
director d be of the essence, then the operator d coincides with the
second-order tensor
A = -ed,
because
d(q) = d + q x d + o(q).
In case (ix), v is interpreted as a double vector G with det G > 0;
then Jt coincides with Lin+ and each 9;Jt with Lin. In general it is
10 I. General Properties

assumed that G(q) be simply QG, so that, to the first order in q

G(q) = G + aq + o(q),
where a is the following third-order tensor

In the case (iv) of liquid crystals there are interesting alternatives. If

one chooses for J( the set of unit vectors c, but with the identification
of c with -c, the tangent space at ±c is the subspace of l ' that
consists in the vectors orthogonal to c. Then d is the skew tensor
-ec and
c(q) = c + q x c + o(q).
If, on the contrary, one chooses for J( the set of diads c ® c, then
d is the third-order tensor
(ec) ® c + c ® (ec),
which transforms any vector of l ' into a second-order tensor the null
space of which is the set of all vectors parallel to c.

§4. Continua with Microstructure

We can proceed now to give a precise definition of continuum with
microstructure. A body ~, made up of material elements x, is said to
be a continuum with microstructure if the following properties hold:
(PI) A smooth manifold J( of finite dimension m is assigned, with
elements veach of which represents a different microstructural
condition. On J( a group of transformations is defined, which
renders explicit the effect of rigid rotations; for any v E J( and for
any Q E Orth+ an element v(q) E J( (Q = e-eq) is uniquely deter-
mined and the operation v -+ v(q) has the group properties men-
tioned in Section 3.
(P2) A class ~ of mappings of ~ into tf x J( (called here complete
placements) is assigned
x = x(x), v = v(x), x E tf, v E Jt, X E ~,
such that:
(a) the apparent placement x = x(x) is a one-to-one mapping of ~
into tf and the codomain f1l is the closure of an open connected
set with regular boundary iJf1l, as is usually assumed for ordinary
continua;
 §4. Continua with Microstructure 11

(b) each couple of apparent placements x'(.I}, x"(.I) is such that the
induced bijection of fJI' = x'(~} onto fJI" = x"(~) is smooth,
again as for ordinary continua; some specific exceptions to the
rule of regularity are treated separately;
(c) each -complete placement (x, v) E ((j is such that the mapping
v 0 x -1 of fJI = x(~) in vi{ is also smooth; again exceptions to the
condition of smoothness are considered separately;
(d) if (x, v) E ((j and Q is any element of Orth+, then also (x(q), v(q)
belongs to ((j; here v(q) is as defined under (PI) and
x(q)(.I) = x(.I) + c + Q(x(.I) - x(.I'»;
where c is any arbitrary constant vector of "f/" and .I' is a fixed
element of ~. Here the subscript (q) is used to iq~ntify a rotation
of the body rather than of an observer (as was done in Section
3); nevertheless there is no inconsistency with the notation of that
section.

Remark. The extent of the class ((j is irrelevant for most of the
developments which follow, provided property (d) be assured; how-
ever, for simplicity, it is convenient to have in mind bodies 'without
constraints', i.e., bodies for which all placements are accessible, which
can be obtained through smooth bijections as far as the apparent
placement is concerned and through arbitrary smooth variations of
the field v. Actually reference to such bodies is obligatory in some of
the developments of Sections 7, 11, and 12. Constrained bodies are the
topic of Sections 14 and 27; also anholonomic constraints are con-
sidered there.

Properties (PI) and (P2) suggest also that, whenever convenient, we

introduce for ~ a reference placement. An asterisk will be used to
denote quantities in that placement: for instance, as we have already
agreed, the place occupied by an element is x* (.I), the set of x* is fJI*,
and so on.
Use will also be made of traditional notation for the deformation
gradient and for tensors deduced from it:
ox
F:= ox*;

F = RU, R E Orth+, U E Sym+; C :=FTF = U 2 •

Standard differential operators on functions of x are written with
lowercase initial letters
12 I. General Properties

Of{) . oai oak

grad f{) = T' dIva = -;-, (rot a)i = eijk-;-' (4.1)
uX uXi uXj

A capital initial letter denotes similar operations, but on the ref-

erence placement; for instance,

Of{)
Grad f{) = -;--, .
DIva = (oa i
-~-) . (4.2)
vX* uX*i

We can also give now a precise definition of a motion of duration r

(r > 0) for~: it is a mapping of ~ x [0, rJ on the class C(f of complete
placements
x = X(I, -r), v = V(I, -r).
If a reference placement is available, the motion can be described
also in terms of x* and -r
x = x(x*, -r), v = v(x*, -r).
During a motion the velocity v and the rate of change of the
microstructure v (which will be called, briefly, microvelocity) are given
by
v := X(I, -r), v:= V(I, -r).
Velocity and micro velocity at a given time -r are often thought as
fields on P4,:= x(~, -r). Use will also be made here of the connected
concepts of virtual velocity and virtual microvelocity; notice that, in
the absence of constraints, any smooth vector field on P4, can be
taken as a field of virtual velocities, whereas the choice of a field of
virtual microvelocities is conditioned by previous assignment of
the field v(x, -r) and the consequent determination of the set of
tangent spaces: V(I, -r) must belong to the tangent space ffy(x.,).H in
V(I, -r).
A veloci~~distribution is rigid, with translation velocity c(-r) and
angular velocity w(-r) if it is given by
V(R)(I, -r) = c(-r) + w(-r) x (X(I, -r) - X(I', -r»,
(4.3)
V(R)(I, -r) = d(V(I, -r»w(-r),
where d is the operator defined in (3.4).
With the use of a local chart and of the notation (3.5) the IX-th
component ViR) of V(R) can be written
ViR) = a(a)(V(I, -r»' w.
 §5. Invariance Properties 13

§5. Invariance Properties

Knowledge of the infinitesimal generator of the rotation group on ..H
allows us to state various conditions of invariance both static and
kinematic: for instance, conditions of invariance under the galilean
group or conditions of frame indifference.
Suppose that qJ be a function only of v, and suppose also that qJ be
a scalar; that is, condition (1.8) applies with 1l the rotation group.
Then qJ must satisfy condition (1.9), which, in terms of (3.4), (3.5) can
be written
(5.1)
or, more compactly
(5.2)
or, finally

( OqJ)a(") = o. (5.3)
OV" '
besides, as was remarked in the Introduction, anyone of these condi-
tions, if intended valid for all v, implies the property of invariance
required of qJ.
We have already remarked that the derivative ~:= (oqJ/ov) belongs
to the cotangent space !/v*..H of ..H at v; now, if v(r), .. E [0, 'f]
represents the time-evolution of the microstructure on a material
element, one C~lD evaluate the time rate of change of qJ through the
scalar product

~ is itself a scalar and we can ask how this property can be assured
also when ~ is not obtained from a potential. The property derives
from an appropriate transformation law for cotangent vectors, ~ -+ ~q),
such that, if q is any time-constant rotation vector, the relation

applies identically, and one has, in particular,

(O~(q»)1
oq q=O
·V + ~.(OV(q»)
oq
I
q=O
= o.
Remark that, on the other hand,
14 I. General Properties

(O.a.)O
V(q) v +
= .:.a. ov ll V°II q; + 0 (q).
(Oda.;) (5.4)

Hence

(O~(q)a.)1
oq; q=O
= _(Od:;)~II'
ov
and, finally,
11
~(q)a. =~- ( 0.91 .)
0V« I ~lIq; + o(q). (5.5)

a formula which, in a sense, is the dual of (5.4).

If one imagines Q as itself time-depen~ent, the formula which
expresses (v(q»" to the first order is more complex

(v(q»" = (<5; + 0;:1Ia.; q) vII + da.;r; + o(q), (5.6)

where r is the vector associated with the skew tensor QQT

r = _!e(QQT). (5.7)
Let us now formulate with the help of (5.6) an invariance condition
for situations more complex than that envisaged by (1.8). Consider
first the case of a scalar cp which depends on v and grad v only; then
the invariance condition is
cp(V(q), grad' v(q» = cp(v, grad v), 'fIv E.R, 'fIQ E Orth+, (5.8)
where the inverted comma affixed to the operator grad is there to
indicate that the gradient is taken on the rotated body. From (5.8) it
follows that

Ocp) a(a.) + (grad a(a.) _ e(grad

( 0V« va.»(o(grad
ocp ) = 0,
v«)
(5.9)

and, by a repeatedly invoked theorem, the inverse property applies;

that is (5.9), if valid without restrictions, implies (5.8). The last formula,
as we shall see, has a direct bearing on the dynamics of continua with
microstructure.
The case to which (5.6) is relevant is when cp depends on v and v
only. Then the condition of invariance

cp( v + dq + "', v + e(:Vq») v + dr + ... ) = cp(v, v) (5.10)

 §6. Conservation of Mass: Kinetic Energy 15

requires that

(5.11)

and, vice versa, (5.11) implies (5.10).

When cp is invariant under the galilean group only, rather than
under the full euclidean group, condition (5.l1h drops out.

§6. Conservation of Mass: Kinetic Energy

A complete kinematic picture of our model for the microstructure
depends still on a number of choices. The ones indicated below are at
a level of generality sufficient for our purposes; alternative choices
could be made, of course, as will be pointed out later in passing.
The first step is the choice of an explicit expression of the principle
of mass conservation. Here we are largely conditioned by the ac-
ceptance of property (P2) of Section 4; phenomena of diffusion, such
as those occurring in mixtures, are excluded there. The material ele-
ment is supposed to be identifiable at all times and to carry all its
mass in its motion. Phenomena involving creation of mass or vacancy
(through radiation, for instance) could be envisaged, but we exclude
them here.
Thus we assume, as in the classical theory, that the mass of ~ is
the total of a mass density p and is conserved, so that, if P. is the

t.
value of p in the reference placement, the equality applies

fa P = P., (6.1)

together with similar equalities for any subbody. Along sufficiently

regular motions and for smooth density distributions, well-known
local relations follow:
pi = P., 1:= det F, (6.2)
p + p div x = 0, (6.3)

Remark 1. The condition of incompressibility, when relevant, can be

expressed through the constraint
16 I. General Properties

div i = o. (6.4)
In porous media where the void fraction is v and the density of the
matrix material Pm' the global density P is given by the product
P = Pm(1 - v)
and a condition of incompressibility of the matrix material, if relevant,
can be derived from

the condition is
- v + (1 - v) div i = O. (6.5)

A second set of hypotheses leads to an expression for the kinetic

energy of m. We speak here of kinetic energy before we mention
momentum and inertia, because it appears, in each particular case,
easier to conceive an appropriate expression of the former quantity
rather than of the latter. Some authors think it is even more suitable
to assign directly some global function (for instance, a lagrangian
function) and avoid a hypothesis of additive separation of the kinetic
and potential energies for the microstructure. However, such separa-
tion offers the advantage that it is then possible to apply to the
potential energy and the related concepts of stress and microstress the
powerful conditions of frame indifference.
To cover most situations studied so far, it is sufficient to restrict
our attention even more and to consider only the case when the
density per unit mass of the kinetic energy is the sum of two terms,
the classical one, and one related to the microstructure.
ti 2 + K(V, v) (6.6)
with K a nonnegative function with the properties
82 K
K(V, 0) = 0, 8v 2 #- O.

One notable advantage of this subcase is that, as we shall see, it leads

to the usual expression of macroscopic inertia, that is -pi. There are a
few examples in the literature of expressions for the kinetic energy still
with the additive property (6.6) but with K dependent also on grad v;
we avoid such complexities here (see the Remark in Section 17).

Remark 2. Here, for the first time, the discussion of Section 3 regard-
ing the best choice of the copy for ./It and the notion of observer find
 §6. Conservation of Mass: Kinetic Energy 17

justification. Within a purely geometric context, to think of the values

of the order parameters as observer dependent may appear inappro-
priate. Only within a dynamic context the need appears for such a
dependence: for instance, to accept the validity of the decomposition
(6.6), one- needs to be sure of the absolute meaning of v. Liquid
crystals are again a good example; the microstructure could be specified
through the unit vector c = F-1dIF-1dl-1, rather than through d, and
c is observer independent. But, as a consequence, the microstructural
kinetic energy, which is thought to be proportional to d2 , would not
be expressible in terms of c and c only, but would involve also ft, as is
easy to check. In several cases an objective measure of the micro-
velocity is, nevertheless, of use; one can refer then to a co rotational
derivative
v-dr, (6.7)
where r is the spin vector of macromotion
r = -te(ftF- 1 ). (6.8)
The choice (6.6) for the kinetic energy can be further specialized,
by the assumption that K be a quadratic form in v
(6.9)
or, more compactly,
K = tv·nv. (6.10)
In any case K must have the same value for all observers at rest,
hence it must satisfy the condition
~ (al OK Oa(a l op _ 0
ova a + ova OVp V - ,
(6.11)

a condition which, under hypothesis (6.9), becomes

( an ) (oa(Yl) (oa(Yl)
ov~p a(vl + nay ovp + npy ova = O. (6.12)

For instance, in case (iii) of Section 2, when v coincides with a

vector d and d with the tensor ed, condition (6.11) reads
OK OK.
ad x d + ad x d = 0,

a condition which is satisfied if K depends on d, d only through the

products d 2 , d· d, d2 •
18 1. General Properties

Remark 3. When both (6.6) and (6.9) apply, the total kinetic energy of
a body ~ in a rigid rotation with angular speed w can be expressed
~s follows
!W'(J + M)w, (6.13)
where

J:= fa p(X2 I - x ® x) (6.14)

is the classical inertia tensor, whereas

M:= f6f pfl«pa(<<) ® alP) (6.15)

is an additional inertia tensor due to the microstructure. This result

puts in evidence the decisive role of the choice of the dependence of
v on the observer. When the measure of the order parameters is
observer independent, the tensor M vanishes identically.

§7. Inertia

A delicate matter in the dynamics of continua with microstructure is

the connection between the chosen expression for the kinetic energy
density and an appropriate consequent choice of the density of the
microstructural inertia force. It is important that the choice be such as
to assure the validity, in the context, of two fundamental results of the
classical theory:
(i) the validity of a kinetic energy theorem, which implies that the
power for unit mass of the inertia forces be the opposite of the
time-derivative of the kinetic energy per unit mass;
(ii) the validity of a variational principle similar to the d'Alembert
principle; this condition requires that the virtual power of in-
ertial forces coincide with the first variation of an appropriate
functional.
If one calls - m and - (x, respectively, the densities of inertia per
unit mass in the macromotion and micromotion and requires that (X
be an element of ffv* Jt, the first condition reads
<!X2 + K(V, v»" = m'x + (X·v.
It is easy to check that the condition is satisfied if
 §7. Inertia 19

"
m = x, CE =
(ax)· ax
av - av' (7.1)

where X is the density of kinetic coenergy, that is, the function which
is related t~ K by the rule

K=G~}V-X; (7.2)

K and X coincide if K is a quadratic form in v

To render the second condition explicit, consider any motion of ~
x = x(x*' r), v = v(x*' r), r E [0, r], x* E fII*,
and a class (depending on a parameter e) of varied motions
x.(x*' r) = x(x*' r), + e ox(x*' r),
v.(x*, r) = v(x*' r), + eov(x*, r),
with the property
ox(x*,O) = ox(x*, r) =0, ov(x*,O) = ov(x*' r) =0,
t
With if;(x, v, v) an appropriate function, the second condition requires

o (J: dr pif; ) = J: dr fa, (m'ox + CE' oV), (7.3)

where the variation of the integral in the left-hand side must be

interpreted as follows:

I p. if;(x + eox, v+ e ov, v+ eov»)J.=0 ;

[:e (JoIf dr Ja,
here fII. is the domain occupied by the body in the varied motion and
p. is the ensuing varied value of p which assures mass conservation: in
terms of (6.2)
p. det(Grad x.) = p det(Grad x).

It is easy to verify, by use of the classical transport theorem and

assuming the motions sufficiently smooth, that, with the choice (7.1) of
m and CE, (7.3) is satisfied by
if; = tx 2 + X(v, v).
20 I. General Properties

As the remarks of this section and others which we will make later
all recommend the expression (7.1), we will adhere to this choice in
the rest of this book. Note that in (7.lh the only term that contains v
is (02X/OV 2)v, whereas the other terms contain at most v and v.
txamples. In the cases (i) and (ii) of Section 2 (continua with voids; v
a scalar, which, as already indicated, we will denote with the same
letter in normal print) it is traditional to think of the elements of !B as
spheres containing a bubble (a spherical concentric void) and therefore
each having a volume equal to the reciprocal of the number density v
of the bubbles. It is assumed that each element can (not only translate
as a whole with velocity x but also) expand or contract. If the matrix
material is incompressible, changes of volume of the element may
derive only from changes in the size of each bubble; the kinetic
energy associated with each element in expansions/contractions is
that of a spherical crust of external radius (3/4nv) 1/3 , of constant
volume (1 - v)v-l, constant density Pm and with a rate of change
of the external radius which can be expressed in terms of y by
(36nvt1/3(1 - V)-l y. The kinetic energy of each element is then
JPm(48n 2vv 5 )-1/3(1 - v)-2(1 _ V 1/ 3 )Y2.
As the global density P equals Pm(1 - v), the total kinetic energy per
unit mass is
1 .2 1 y2 (1 - V1/3 )(1 - V*)2/3
(7.4)
'2x + '2 (48n2v;v)1/3 (1 _ V)11/3

where the values v*, v* of v, v in a reference placement have been

introduced. When v is small one obtains for K a formula due to
Rayleigh
(7.5)

In this special case the expression for (X becomes

Y(V- 1/3 V - f,V- 4/3 y2).
An admittedly artificial example of a function K which is not a
quadratic form in v could apply to a subcase of example (viii) in
Section 2 when

where y and XO are positive constants; in such a case

X = Xo(1 - (y2 - J2)1/2 y -1).
 §8. Dynamic Equations of Balance 21

These expressions for K and X could possibly apply to a medium,

such as a suspension of rod-like particles in a compressible fluid,
where the microstructure accounts also for entrainment effects and
one would like to put in evidence the existence of an intrinsic velocity
of propagation for perturbations in the fluid, which limits the range of
values that ltil may take.

§8. Dynamic Equations of Balance

In this section we suggest a plausible form for the equations of

balance of momentum and micromomentum, a form which is based
on the remarks in previous sections and on analogies with standard
relations valid either for continuous bodies or for discrete point sets.
The consistence of the suggestion with certain variational principles
will be examined in later sections. The developments of Part II will
confirm the adequacy of the proposals when confronted with special
theories.
For the balance of macromomentum, the equation of Cauchy is
preserved in the usual form: for any subbody with current apparent
placement 6, the balance is expressed by

(L Pi)" LpI
= + fM t, (8.1)

where I is the external body force per unit mass and t is the surface
traction per unit area. From (8.1), under standard conditions, the
tetrahedron theorem follows: a tensor T exists such that t = Tn, with
n the unit vector of the exterior normal to 06. Furthermore, wherever
i and T are sufficiently regular the local equation of balance
applies
px = pI + div T. (8.2)

For the balance of micromomentum we propose directly a local

equation that has analogies with (8.2), but derives from our view of
body elements as Lagrangian systems:

p( (;!} - G~)) = pp - ~ + div f/, (8.3)

here:
22 I. General Properties

(i) the left-hand side is written so as to ensure that the density of

microinertia will conform with the expression proposed in Section
7;
(ii) P and ~ are elements of the cotangent space ffv*.H, whose compo-
nents will be denoted by Pa, ~a;
(iii) P is the density per unit mass of the external actions on the
microstructure (for instance electric actions on the molecules of a
liquid crystal);
(iv) -~ is the density per unit volume of the internal microstructural
actions which do not necessarily sum to zero as internal forces
do; the choice of sign here is purely conventional;
(v) ff is a linear operator with domain "Y and codomain ffv*.${ such
that, on a surface element where n is the unit vector of the
exterior normal, the vector 0" of ffv*.${

0" = ffn, (8.4)

represents the actions exerted on the microstructure through the

element.

In terms of components on a local chart, (8.3) may be written

formally as follows

p( (::a)' - (::a)) = pPa - ~a + div s(a» (8.5)

where sIal is the member of 1"" such that

O"a = S(a)· n.
whereas, in general, the quantities ~a do not coincide with the compo-
nents ~a of ~.
Later developments required equations of balance valid in the
reference placement are. One must introduce, together with the
Piola tensor
P := IT(F- 1 f, (8.6)
similar quantities related to the microstructure
:}:= l~,
(8.7)
[!J' := lff(F- 1 f,
and also the operator Div (see (4.2)2), so that, in particular,
1 div ff = Div [!J'.
 §9. Balance of Moment of Momentum 23

In all these relations the notation I for det F is used. The referential
versions of (8.2) and (8.4) are
p*x = p*! + Div P,
and (8.8)

§9. Balance of Moment of Momentum

From (6.2), (8.2), and (8.3), along regular motions, a theorem of kinetic
energy can be inferred in a standard way, that is, taking the scalar
product of both sides of (8.2) by x, operating similarly with v on (8.3),
integrating both sides over a subbody 6 by parts where possible,
taking account of (6.3) and (7.2) and summing finally term by term:

! (1 p(!x 2+ K») = 1 p(f' x + P' v) + t~ (t· x + 0"' v)

-1 (T' grad x + ~'V + y. grad v). (9.1)

The identification of terms in this equation is as follows. On the

left-hand side the time-rate of change of the total kinetic energy
appears; on the right-hand side the power of external body actions
and the power of external surface actions appear in succession. Thus
the last integral must be interpreted as the power of internal actions;
its density is
-(T' grad x + ~. v + y. grad v). (9.2)
It is now possible to render explicit a mechanical condition of
balance (the condition of balance of moment of momentum) which, in
a sense, translates, within the present context, the action-reaction
principle: it is required of T, ~, and Y to render the quantity (9.2)
identically zero along any rigid virtual motion.
In explicit terms we must require that
T· grad vIR) + ~ .U(R) + y. grad U(R) = O. (9.3)
for all VIR)' U(R) of the form (4.3) and for all choices of w in these
formulae. The following condition ensues
(9.4)
24 I. General Properties

To avoid ambiguities of interpretation we rewrite (9.4) also in terms

of components:
(9.5)
or
(9.6)
Still another version of Eq. (9.4) is as follows:
skw T = te(~a:da) + (grad da)s(a)' (9.7)
A referential form of (9.4) can be given easily:
e(PFT) = d T3 + (Grad dT)r!J>. (9.8)

Remark 1. An important consequence of,.(9.4) is the symmetry of the

Cauchy stress tensor not only in the classical theory when the micro-
structure is absent (or, more precisely, the microstress vanishes), but
also when the operator d vanishes. Summing up: when the order
parameters va which lead to an expression of densities such as (6.6) for
the kinetic energy, (7.1h for the microinertia and (9.2) for the power
of internal actions have the same value for all observers, then the
Cauchy stress is necessarily symmetric.

Remark 2. With the use of (9.4) one can obtain a version of (9.2)
which puts in clear evidence the independence of the power of internal
actions from the observer. Recall first a standard decomposition of the
velocity gradient
grad i = D - er, (9.9)
where r is given by (6.8) and
D := sym grad i.
Now note that
-(T' grad i + ~'v + [1'. grad v)
= -T'D + skw T'(er) - ~'v - [1'. grad v (9.10)
= - T· D - ~. (v - dr) - [1'. grad(v - dr) + (d T[1'). grad r;
here, in the last expression, D, v - dr, and grad r are all frame
indifferent.

Remark 3. In some papers the proposed equation of balance for the

moment of momentum seems to take a form more akin to (8.3) than
 §1O. Boundary and Jump Conditions: Change of Variables 25

to (9.4). To clarify this point remark that, operating on both sides of

(S.3) with d T, one obtains a vectorial equation from which dT~ can
be eliminated by use of (9.4). Taking also account of the invariance of
X under changes among stationary observers (see (5.11)d one is led to
the equation

p( dT:~} = pdT~ - eT + div(d T 9"). (9.11)

When m = 3 and d T is nondegenerate, this elimination of ~ does

not lead to information loss; then (9.11) is the only relevant equation
to be added to the balance equation of Cauchy and in fact it can take
up the role of equation of balance of moment of momentum. If, more-
over, microinertia and body actions on the microstructure can be
disregarded, as happens often, then (9.11) reduces to
(9.12)
In this case it is enough that d T 9" have null divergence for T to be
symmetric.

§1O. Boundary Conditions: Change of Variables

Together with the equations of balance (S.2), (S.3) analytically accept-
able boundary conditions for "traction" problems must be displayed:
if the vectors t and (1 are assigned over of1B, then the boundary condi-
tions read
Tn = t, 9"n = (1, on of!J; (10.1)
they do not involve ~ at all.
These boundary conditions can be written also in terms of referen-
tial stress (S.6) and microstress (S.7); it need only be recalled that the
unit vectors nand n* normal to of1B and of1B* respectively at corre-
sponding points are related by
In* = '1FT n (10.2)
where 'I is the coefficient of surface expansion
'1:= lIF-Tn*1 = 1/IFTnl. (10.3)
The referential versions of (10.1) are
Pn* = 'It, ~n* = 'I (1. (10.4)
26 I. General Properties

From an experimental point of view, however, the second condition

(10.1) seems often hard to impose. The experiments in special cases
(on liquid crystals, for instance) suggest that "conditions of place" are
more significant; even more appropriate at times are consequences of
constitutive boundary conditions. Such conditions depend not only on
the type of medium under consideration but also on the material
which constitutes the boundary and on its particular physical treat-
ment. Analytically the conditions appear as relations involving F, v,
grad v on the one hand and quantities characterizing the boundary,
such as n and the mean curvature.
We conclude this section with some remarks on the formal con-
sequences of a coordinate change on Jt. Suppose we have introduced,
besides the coordinates v", a new set -vP which is in a one-to-one
correspondence with v" (a correspondence which is independent of
time and place)
(10.5)

so that the linear relation between microvelocity components is that

of a contravariant vector
av"
v." = r" P-'-P
v , r"-
P- a-vP' (10.6)

Then it is easy to verify that, as a consequence, the left-hand side of

(8.4) behaves like a covariant vector,

(10.7)

-1 -1
here r/ is the inverse operator defined by the property r" y r/ =
(j;. It is also immediately observed that the operator d changes in
accordance with a contravariant rule:
- ; - r"':;;p
.A" -
P-;'
Finally, the question remains as to choice of the rule of change for the
microstress variables which leads to the same formal expression for
the power of microstress;
I' ."
~"v + J,,;V ,i = r~"v-'-" + J,,;V
('0' " ('0 -'-"
,i'

to the same formal expression for the right-hand side of (9.5):

 §11. The Conservative Case in Statics 27

and to a covariant rule of change for the right-hand side of (8.5):

-1 P A _
+ ~i,i - r., + Y'pi.i)·
_ A _

~., (-~p

It is easy to verify that the rule is as follows

r _ -1 p? -lp_
"'., - r., "'P + ra ,iY'pi'
-1 _ (l0.8)
~i= r/Y'pi'
Thus, ~a does not behave as a vector with respect to coordinate
changes on .H.

§11. The Conservative Case in Statics

In the static and conservative case, the balance equations proposed in

Sections 8 and 9 can be deduced also from a variational principle.
Let us suppose that the internal actions on I.E derive from a
potential qJ per unit mass and that qJ be a function of the state of I.E
through F, v, Grad v only. As for the external actions on I.E let us
assume, as already done in earlier Sections, that body actions f, JJ and
contact actions t, (1 through the boundary are present. As we wish to
express totals on the reference placement, we use the vectors
(11.1)
where 1] is the coefficient of surface expansion introduced in (10.3).
Suppose we have chosen the fields x(x*), v(x*) on f!I* as our
candidates to represent an equilibrium state of I.E under the given
body and surface actions.
Let 15x(x*) be an arbitrary vector field on f!I* and 15v(x*) a field
such that its value at x* is an arbitrary vector of !!I~(x.).H. We define
the global virtual work of the external actions as the quantity

15f£':= r
JilB*
p*(f' 15x + JJ. 15 v) +
oilB.
f
(t* . [)x + (1* • 15v). (11.2)

Then we require that x(x*), v(x*) satisfy for all allowed choices of [)x,
15v the relation
(11.3)

where the operator 15 on the left-hand side is defined as follows:

28 I. General Properties

:= [~
uB
r
Jill'.
p*(cp(Grad(x + B(jX), v + B(jV, Grad(v + B(jV))]
£=0
.

It is easy to see that

(t. P*Cp) foill'> {(P* :; n*)-

(j = (jx + (P* O(G~:d v) n*)- (jV}
+ t. {(P* !: - (P* O(G~:d Div V)) )- (jv - Div (P* :; )- (jX},
where the integrands must be computed f()r the fields X, v. Because of
the arbitrariness of (jx and the freedom of choice for (jv, the following
relations ensue:
Div P + p*b = 0, -3 + Div [!P + p*p = 0, in fJ6*, (1104)
and
(11.5)
if the notation

P= p*(:;).
(11.6)

3= P* (!:). [!P = P* (O(G~:d V)).

is introduced and it is agreed that the right-hand sides of (11.6) are
expressed as functions of x* through the fields x, v.
Thus we have obtained again the static balance equations in the
referential version (compare (1104) with (8.8)) and also the boundary
conditions in the form (1004). There is an alternative formulation of
the principle of virtual work, which allows one to confront the critical
remarks made in Section 10 regarding boundary conditions. The
formulation is as follows

Here one excludes a direct assignment of boundary actions and

accepts the existence of a surface potential, its density per unit area of
ofJ6*being y. With regard to this quantity it is presumed below that it
 §11. The Conservative Case in Statics 29

depends only on the present values of v and n on the boundary; more

complex situations (for instance a dependence on F) could be accom-
modated but are excluded for simplicity.
To deduce the consequences of (11.7) one must compute the varia-
tion of y; and, to this end, evaluate bn in terms of bx. Referring back
to (10.2) and using ~he notation bF:= o(bx)/ox*, one obtains to
within terms of higher order,
n + bn = (F + bF)-Tn*/I(F + bFrTn*1
,..., (n - F-T(bF)Tn)/ln - F-T(bF)Tnl
,..., (n - F-T(bF)T n)(l + n· F-T(bF)T n)
,..., n - (grad bxfn + (n· (grad b.X)Tn)n,
and hence
bn = (n ® n - /)(grad bX)Tn. (11.8)
To proceed to the evaluation of the last addendum in (11.7), we
remark that, by (11.8), one has

b faill. y= faill. (oy.

OV
bv + oY .bn)
on

where the index s is affixed to indicate that the component tangent to

of!l* of the vector within brackets must be taken.
We recall now a divergence theorem, which is valid on closed
surfaces and which we apply to of!l*: for any tensor V which has n*
as null direction, Vn* = 0, and for any vector v the equality holds

f aill*
V·Grad v = -f aill.
(Div. V)·v, (11.9)

where the surface divergence operator Div. is defined by the property

(Div. V)· k = (1- n ® n)· Grad (V Tk), Vk E "Y. (11.10)
We use this theorem taking for v the vector bx and for V the
tensor
(11.11)
30 I. General Properties

which, in fact, satisfies the condition Un* = 0, so that

fa~. (n ® G~})- grad ~x = L~* (n ® F- (:~)). Grad ~x

1

= - L~. Div.(n®F-IG~))·
This result allows us to reach the desired expression for ~ Ja~* y:

~ fa~* y= fa~* (::.~V+~X.Div.(n®F-l(:~))).

Introducing this expression in (11.7) and repeating the transforma-
tions indicated earlier for the left-hand side, we are led again to (11.4),
(11.6) and to the new boundary conditions
Pn* = Div.(n ® F-1(oy/on).)), ~n* = oy/ov. (11.12)
in place of (11.5).

§12. Perfect Fluids with Microstructure

We treat here separately the case of perfect fluids with microstructure

for its own intrinsic interest and also because it offers the occasion to
examine the consequences of a variational principle of local, rather
than material type (as was done in the previous Section) and to
discuss other matters of general interest.
In fact, to invoke for fluids a reference placement does not seem
proper; more appropriate seem developments based only on local
quantities and their variations. As we wish to include in our consider-
ations compressible fluids, we need a measure of volume expansion
and we choose to this end the scalar l; the choice is equivalent to the
inclusion of the density p among state variables. A formula due to
Euler establishes the relation between the variation ~x and the varia-
tion of l, as we shall see. As for the variations ~x, ~v, they are thought
as fields on f1I, rather than on 81* as intended in Sections 7 and 9.
Naturally, the variation ~X of the place of an element implies varia-
tion of the domain f1I, of its boundary and of the other fields defined
on 81. Thus these fields (such as v(x)) are subject to two variations: a
direct arbitrary variation (~v in the example) and the variation con-
sequent on the virtual displacement ~x. We will indicate by 8 the total
variation, for instance
 §12. Perfect Fluids with Microstructure 31

8v := c5v + (grad v)· c5x, (12.1)

Notice though that the total variations may be subject to restrictions;
the Euler formula for differentiation of a determinant requires 8, to
satisfy the condition
8, = , div c5x, (12.2)

and conservation of mass restricts the total variation of density:

8p + P div c5x = O.
One can express this last condition also in terms of c5p:
c5p + div pc5x = o. (12.3)
The potential qJ will be assumed now to be a function of i, v, and
of the local gradient grad v (rather than on the referential gradient, as
in Section 11). The dependence of qJ on , rather than on the whole
tensor F reflects the restriction of our present analysis to fluids.
As for the virtual work of external actions, formula (11.2) is ac-
cepted again but it is written using the present placement as domain;
also, the total variation for v is now involved:

c52= r P(f·c5x+~.8v)+f[)£i (t·c5x + a·8v).

J£i (12.4)

The variational principle which is assumed valid for any complete

equilibrium placement of rJB is

(12.5)

To deduce the local consequences of (12.5) observe first that the

variation in the left-hand side is understood in the sense made explicit
by the transport theorem:

8 (t tPqJ) = c5(pqJ) + L£i pqJn' c5x.

By the divergence theorem and (12.3) we obtain

8 (t PqJ) = t (c5(pqJ) + div(pqJ c5x» = t P 8qJ

= f£i P {~~ 8, + ~: . 8v + O(g~:d v)' 8(grad V)}

32 I. General Properties

and

b (L P~ L {~~
) = P 1 div bx + ~: . bV

+ O~d) . (grad bV -
o(gra v
(grad v)(grad bX))},
where (12.1) and (12.2) have been invoked.
U sing the divergence theorem again we get

b (L P~ ) = foal {p ~~ lbx' n- P (O(g::d v) n) .((grad v) bx)

+ p(O(g::d v) n}bv} +" fal {p ~: 'bv

- (grad (Pl ~~)). bx - bV' div (p o(g::d V))
- bx' div((grad vf P o(g::d V))}·

Now applying this result to (12.5) and taking into account the
arbitrariness of bx and bv we arrive at the local relations
div T + pf = 0, div 9' + p~ - ~ = 0, in!JI (12.6)
Tn = t, 9'n = cr, on ofll, (12.7)

where we have introduced the notation

T = pl(o~/ol)I - (grad vfff, (12.8)
9' = p(o~/o(grad v)), ~ = p(o~/ov). (12.9)

Equations (12.6) are exactly the local balance relations for statics
(compare them with (8.2), (8.3)). It is interesting also to compare (12.8),
(12.9) with (11.6): the second term in the expression of the Cauchy
stress which appears in (12.8) is absent from (11.6)1; it derives here
from our having assumed ~ to depend on grad v rather than Grad v.
With a terminology borrowed from the theory of liquid crystals that
term in T could be called Ericksen stress.
As in the preceding Section we can invoke here a modified varia-
tional principle obtained from (12.5) by replacing the surface integral
 §12. Perfect Fluids with Microstructure 33

in the expression for virtual work by

8 r
JiJ~
y(v, n),

where now y is the density of the surface potential per unit area of ofJl
If we denote by 1 + 811 the coefficient of surface expansion for the
varied configuration when compared with the present one, defined by
adapting (10.3)
1 + 811 = det(grad(x + c;x))I(grad(x + C;X)tTnl
(12.10)
= det(/ + grad c;x)I(I + (grad c;xtT)nl,
observe that
C;11 = (I - n ® n)' grad c;x,
and recall also (11.8), we conclude that

8 r
JiJ~
y= r {Oy .8v + oy ·8n + y811 }
JiJ~ OV on

= fiJ~ {;:.8v+(y(I-n®n)-n®G~).)·gradc;x}.
Recourse must be made now to the special divergence theorem on a
closed surface invoked in the preceding section (see (11.9)); the theo-
rem is applied here to ofJl and under the condition Vn = 0

r
JiJ~
V·grad v = - r
JiJ~
(divs V)·v.

The appropriate identifications are v = c;x and

V = y(I - n ® n) - n ® G~},
the latter being consistent with the condition required of V.
Hence we have finally

8 L~y= L~ {;:.8v+C;X.divs(n®G~} -y(I-n®n))};

and we conclude that (12.7) must be replaced by
Tn = divs(n ® (oy/on)s - y(I - n ® n)), [/'n = oy/ov. (12.11)
34 I. General Properties

In all these formulae the subscript s indicates a component in the

tangent plane to ofJI and the surface operator div. is defined by
analogy with (11.10).
It is interesting to observe that, in the standard theory, when a
surface tension is present but the microstructure is absent (y =
constant), as a consequence of known formulae from differential geo-
metry, the right-hand side of (12.11)1 reduces to 2ycon, where co is the
mean curvature of ofJI.

§13. Rules of Invariance and the Balance of Moment of

Momentum: Variational Principles in Dynamics
Though admittedly under special conditions, the developments of the
two preceding sections have led again, through a different route, to
the same balance equations (8.8) and to associated boundary condi-
tions. So far absent from the conclusions is the balance equation for
the moment of momentum (9.8). We show here that that equation is a
necessary consequence of a property of the potential qJ which we have
not yet mentioned: qJ is frame indifferent; that is, its value must not
change if the observer is changed, although the variables F, v, and
Grad v (or grad v) on which qJ depends do change. For all choices of
the proper orthogonal tensor Q (and the associated vector of rotation
q) the following functional equation must hold in the case examined in
Section 11
qJ(QF, v(q), Grad v(q» = qJ(F, v, Grad v), (13.1)

Conditions of this type have been examined in Section 5; we have

mentioned there that (13.1) can be replaced by
oqJ(QF, v(q), Grad v(q»joq/q=o = 0, Vq E 1'.
If one observes that
(o(QmnFnN)joq.)/Q=I = emn.FnN ,
one concludes that the following equation must hold
-e((oqJjoF)FT) + dT(oqJjov) + (Grad dT)(oqJjo(Grad v» = 0, (13.2)
and hence, recalling the notation (11.6), also eq. (9.8). Arguments
mentioned in Section 5 show that vice versa (13.2) implies (13.1).
 §13. Rules of Invariance and the Balance of Moment of Momentum 35

In view of some important differences in detail it is worth repeating

the analysis also for the case treated in Section 12. The condition
required now is
Ocp(l, v(qj' grad' v(qj)/oqlq=o = 0,
where the inverted comma affixed to the operator grad is there to
indicate that the operation must be carried out on the rotated body,
as remarked for similar developments in Section 5. Actually the situa-
tion is almost the same as one expressed by formula (5.8), in view of
the fact that 1 is not affected by rotations. The conclusion is that (5.9)
must apply, where now the notation (12.9) can be introduced. In term
of components one has
(13.3)

but, from (12.8), we have

(13.4)
hence (13.3) coincides with the balance equation for moment of mo-
mentum (9.5).
An interesting remark follows from (12.8), (13.4): in perfect fluids
with microstructure, even at rest, the Cauchy tensor need not be
spherical. Hence the properties of fluids with microstructure may be
drastically different from those of ordinary fluids.

The developments and results of Sections 11 and 12 are confined to

statics; but one can seek a variational principle covering also dyna-
mics. At least formally there is no difficulty, for instance, in extending
the case considered in Section 11; one needs only invoke the varied
synchronous motions introduced in Section 7 and require that, for an

i
arbitrary choice of these motions, the Hamiltonian principle holds:

(i r d7:
t

Jo ~*
p*(tx 2 + X(v, v) - cp) + r (iff d7:
t

Jo
= 0, (13.5)

with the specification of cp and (iff stipulated in Section 11. Develop-

ments strictly akin to those of Section 7 and 11 lead to the dynamic
equations in material form (8.8).
In a substantially similar way one can proceed for the case of
perfect fluids. The varied fields must now be defined at each instant

°
on a domain f!4 which may itself depend on time; the virtual displace-
ment (ix is a function of time which vanishes for 7: = and 7: = T. For
36 I. General Properties

the other variations one must distinguish between the total ones
denoted by b and the partial ones denoted by ~, as in Section 12;
finally one assumes valid a Hamiltonian principle in local form:

b J: dr f.w p(!x 2 + X(v, v) - qJ) + J: ~fi' dr = 0,

with qJ and ~fi' specified again as in Section 12. On the strength of

formula (12.3), one observes that

b f.w p(!x 2 + X - qJ) = f.w Pb(tX2 + X - qJ),

and one can proceed along a familiar line of thought to obtain the
dynamic equations in local form (8.2), (8.3);·
The conservative case is of interest mainly in statics, however.
Under dynamic conditions viscous effects cannot in general be
disregarded.

§14. Internal Constraints: Continua with

Latent Microstructure

In the preceding three sections we have tacitly assumed that, at least

in principle, the variables x and v could take, for each element of ~,
arbitrary values and, as a consequence, the variations ~x in "Y and
~v(x) in ffv(x).H could be arbitrary, subject only to sufficient regularity
that justifies use of the classic arguments of calculus of variations.
Also for velocity distributions X, v no restriction was envisaged.
One of the consequences of this tacit assumption was, in conserva-
tive cases, the specification of the constitutive relations for T, ~, [/' (see
(11.6) and (12.8), (12.9»; once the potential qJ was known, the depen-
dence of T, ~, [/' on F, v, Grad v or grad v could be assigned. An as-
signment of this type, even in nonconservative cases, is always deemed
possible, perhaps with a wider range of independent variables, for
example,
T = f(F, D, v, v, grad v).
These constitutive relations must be chosen case by case; some ex-
amples will be illustrated in Part II.
However, materials exist which can be appropriately thought of as
continua with microstructure, the complete placements or the veloci-
 §14. Internal Constraints: Continua with Latent Microstructure 37

ties of which are somehow restricted; one need only think of liquids,
traditionally modeled as incompressible continua. As we shall see, it
is possible to give a full dynamical description for a wide class of
such constrained continua: it is the class of continua with perfect
internal constraints, a class which we now proceed to define.
In the definition the expression of the power per unit volume of the
internal actions has an essential role; we recall it in the form (see
(9.10))
-(sym T)'D + ~'(v - dr) + 9"(grad v - (grad d)'r), (14.1)

where, as usual, the exponent t denotes a minor transposition as

defined in Section l.
For the purpose of this section it is of the essence to decide if,
when an element of m has reached a certain state, all kinematic
options remain open to it or not. More precisely if it is possible to
choose arbitrarily for each element the values of D (in the space
of symmetric tensors), of v - dr (in the tangent space !iy.H), of
grad v - (grad d)'r (in the space of linear operators from "Y into
!iy.H). If the choice is restricted, we will say that the body is subject
to an internal constraint.
Next comes an important hypothesis: we suppose, as is done in
classical theories, that, in the presence of internal constraints, stress
and micros tress are sums of two components, one called active and
the other called reactive
(14.2)
and that constitutive relations can be assigned only for the active
components. As for the reactive components no information can be
given without inspection of the physical mechanism which causes the
constraint. However, as we have anticipated above, there is a wide
class of constrained bodies for which a general characterization can be
proposed: it is the class of bodies with perfect constraints. The con-
straint is called perfect if the reactive stresses do no work, more
precisely if they satisfy identically the condition
r r r
-(sym T)' D + ~. (v - dr) + 9" (grad v- (grad d)'r) = 0, (14.3)
for all allowed velocity distributions.
We will draw many consequences from hypotheses (14.2), (14.3) in a
variety of contexts in later Sections. Here we restrict our attention to
three special cases.
38 I. General Properties

Suppose we constrain the micromotion to be locally of the type of

a rigid rotation, without a global rigidity condition. In other words
suppose that v(x, t) depends on the choice of an arbitrary vector field
h(x, t) as follows:
v= dh(x, r), (14.4)
so that condition (14.3) becomes
r, r r
-(sym T)' D + ~ 'd(h - r)+[/"«grad dY(h - r» + ([/'T d)' grad h = 0.
As the constraint leaves totally free the local choice of values of the
symmetric tensor D, of the vector r - h and of the second order
tensor grad h, we infer that
r
sym T= 0,
dT~ + (grad d T)9> = 0, (14.5)

When account is taken of(14.5)1' (14.5h and of (9.7) we have

f= -skw T + te(dTe + (grad d T ):}),

and finally
(14.6)
Thus, T is totally known through constitutive assignments and the
Cauchy equation can be considered as a 'pure' equation, i.e. an
equation from which all traces of the unknown reactions due to the
constraint are eliminated.
We still need another pure equation for h; we can obtain it from
(8.3) by following developments proposed in Remark 3 of Section 9,
that is, by applying to both sides the operator d T and taking ad-
vantage of (14.5h and (14.5h.

pdT ((~;)" - G~)) = pdT~ - dTe + (grad d T)9> + dT(div :})

+ dT(div 9»

here we need only express v in terms of h through (14.4) to achieve

the desired result.
 §14. Internal Constraints: Continua with Latent Microstructure 39

Actually if we recall that X, as /C, must be invariant under change

of stationary observer and thus must satisfy the condition (6.11) to
which /C is subject, we conclude as in Section 9 that the left-hand side
of the above equation can be written in a more compact self-adjoint
form:

(14.7)

Remark. The constraint (14.4) is, in general, anholonomic; hence, in

general, the present value of v on an element of ~ may depend on the
whole history of h at that element, i.e. on the values of h(x(x., 0), 0)
for (j E [0, t). Thus equation (14.7) in h may contain history terms.

If we require further that h coincide everywhere with r, so that the

micro velocity field is totally determined
r
by the macrovelocity field,
condition (14.5h drops out, leaving ~ totally arbitrary, whereas condi-
tions (14.5)1 and (14.5h still apply. Also Eq. (14.6) must be replaced
by
(14.8)
On the other hand, Cauchy's equation would be now sufficient to
study the evolution of the body, could we manage to render it "pure".
For that purpose we need to eliminate from T the reactive component;
we eliminate ~ from (14.8) by use of (8.4)

T=sym T+te( dT(P~+diV y-p((:!} - :~)+(grad dT)Y)) ,

and then take into account (14.5h:

T = sym T+ te(PdT~ - p(d T :!} + diV(dTY)} (14.9)

finally we need only express v in terms of r.

In conclusion only an indirect trace of the microstructure remains;
Cauchy's equation is modified by the addition of terms which may
involve gradients of the acceleration and even the history of the
motion

pi + t rot p( d T :!} = pf + t rot(pdT~)

+ div(T+ te(div dTy». (14.10)
40 I. General Properties

We have only one field to determine: the apparent placement

x(x*, r). We will say, in such cases, that the microstructure is latent.

Let us consider a third interesting special case, when the micro-

structure is determined by one scalar quantity v only, and 11 is bound
to the macrodeformation through a constraint expressed analytically
by
11 = co(z). (14.11)
The value of v is unaffected by changes of observer, as occurs in
continua with voids. Then the operator d is null. At the same time ~
is a scalar , and [/' a vector s.
Because the constraint is perfect,
(sym +)·D + tv + s·gnld v= 0, (14.12)
for arbitrary choices of grad X, the symmetric part of which is D,
whereas the trace determines v:

. = (dCO)
11 dz i = (dCO).
dz 1 dlV X = (dCO)
dz z tr(grad x). (14.13)

Putting (14.13) into (14.12) we get

r
(dev sym T)·dev D + (r
ttr T +,r (dCO)
dz z) div x

+ s· grad ( (~~) div x) = 0.

As it is possible to choose arbitrarily in this relation the values of

dey D, div x and grad(div x) we conclude that

dey sym Tr = 0, sr = 0, t tr Tr +, (dCO)

r
dz 1 = 0. (14.14)

On the other hand the balance of moment of momentum (9.7)

requires that T be symmetric:
r a
skw(T + T) = 0, (14.15)
whereas from (8.4) we deduce

,r = p{J - ,a - p ((ax)"
ov - ax)
ov + . s. dlV
a

Exploiting these results, for the Cauchy stress tensor we obtain

successively
 §14. Internal Constraints: Continua with Latent Microstructure 41

a r r r
T = T + skw T + dey sym T + t(tr T)J
= sym Ta- (dW)
dl (p{3 -
I ,a- p ((ax)"
av - ax)
av + div sa) 1. (14.16)

Hence the pure equation which determines the motion is as follows:

= pf - grad(~7 P*{3) + div(sym f + «( - div S)(~7}J). (14.17)

Thus, in the present case, the constraint has a holonomic character,

Eq. (14.17) has purely differential character, but, of course, the order is
higher than in the classical case.
 PART II
Special Theories

§15. Continua with One-Dimensional Microstructure:

Continua with Voids

In this second part we review a number of special theories, beginning

with the simplest class, which is characterized by the existence of only
one order" parameter, so that vH is one-dimensional (e.g., an interval of
the real axis or a closed curve), ~ and p are scalars which we denote
by the same letter but not bold, d and f/ are vectors denoted by a
and s, respectively. In summary the balance equations take the form
p + P div x = 0,
p(i - f) = div T,

(:~)' - G~) )= p p -
(15.1)
p( , + div s,
skw T = !e('a + (grad a)s).
Each particular material is distinguished by constitutive choices for
X, a, " s, and T. Stress T and microstress " s must be thought of as
functions of geometric and kinematic variables (e.g., F, D, v, grad v),
functions which satisfy rules of frame-indifference.
If v, for instance, is the cosine of the angle between a fixed direction
in space, of unit vector c, and a privileged microstructural direction, of
unit vector d, then
v = c·d,
 §15. Continua with One-Dimensional Structure: Continua with Voids 43

and one finds easily that

a = d x c,
and, as a consequence, the balance equation for moments (15.1)4 takes
the form
skw{(T - (Cd + (grad d)s) ® c} = o.

Actually we are interested below in cases when the measure of v is

not affected by rigid rotations, so that a vanishes and
skw T=O. (15.2)
For the other constitutive choices it is necessary to explore in detail,
case by case, the behaviour of the material element:- We show below
how one can express X(v, v) in the case one wants to study continua
with voids, using v to measure the void fraction, when the lacunae are
so finely dispersed that the continuum model still makes sense. Then v
is a real number belonging to a subinterval of [0, 1J and its measure
is obviously not influenced by rotations of the body.
The element of a continum is imagined as a small ball with a
concentric spherical cavity as in the first example in Section 7; as
remarked there each element has volume equal to the reciprocal of the
number density v of the lacunae, and the volume of each lacuna is vivo
No diffusion of lacunae through the matrix is envisaged and the only
motion allowed within the element is an expansion/contraction of the
lacuna and corresponding radial motions of the spherical crust. Thus
the model of the element is very simple; notwithstanding all these
restrictions the motion of the element could scarcely be described in
general by one parameter only: there are infinitely many models of
radial oscillation of the crust. One adds therefore the hypothesis that
the matrix material be incompressible, a constraint which we will
later discuss further and which can be expressed by the condition
(exploited in Section 7)
1-v 1-v
--=--*; (15.3)
v v*
where the asterisk indicates values in a reference configuration. In
conclusion one accepts all premisses which lead to the expression (7.4)
of the kinetic energy density.
By Remark 1 of Section 6, the constraint (15.3) can be written also
in the form
(1 - v)z = 1 - v* (15.4)
44 II. Special Theories

where 1 is put equal to 1 in the reference placement. Hence the

microstructure can be described through knowledge of the apparent
placement only; the continuum we are envisaging has latent micro-
structure and we can apply the results of the second part of Section
14. The function OJ is given by
OJ = 1 _ (1 - v*); (15.5)
1

and, as we have repeatedly observed, the vector a is null and so is

skw T. Furthermore, as
dOJ (1 - v*) 1- v
(15.6)
dl
we have by (14.16)

T = sym Ta- (1 - v) ((
p f3 - (ax)·
ov - ax)
ov + div s-,a) I. (15.7)

and this expression leads to a simplified special version of the

equation of evolution (14.17).

§16. Liquids with Bubbles

Special cases of continua with voids correspond to appropriate special

versions of the constitutive equations for t s, t
and We propose here
a choice aimed at the description of the behaviour of an incompressible
fluid in which bubbles of gas are finely dispersed.
In such a case it seems appropriate to exclude direct external
actions on the bubbles, and so we take f3 = O. External surface
actions on the bubbles are also excluded; this is a more suspect
hypothesis, especially in the presence of high gradients of v, but it
will
a
be accepted
a
s
here for simplicity: = O. We must still choose
a
T and , or, more precisely, as appears from (15.7), dev T and
a a a
t tr T + (1 - vK As for dev T we accept a gross hypothesis: that the
mixture behaves as an ordinary linear viscous compressible fluid, with
viscosity IJ, which could depend on v:
dev f = 21J dev D. (16.1)
One may expect IJ to be slightly less than the viscosity lJe of the
liquid in absence
a
of bubbles,
a
and decreasing as v increases. As for
the sum (t tr T + (1 - v)O, one can advance a relatively simple
 §16. Liquids with Bubbles 45

expression which has the merit of consistency with the model of the
element proposed in the previous Section from the point of view of
the mechanical power expended during a motion. Actually we think
here each lacuna full of gas, rather than void; in view of the smallness
of the ratio of gas density to liquid density, this change does not
substantially affect the evaluation of the kinetic energy. On the other
hand we can attribute to the gas in each lacuna a non negligible
pressure ro, depending on v as follows
ro = iij«v.(l - v)/v(l - v.))y,
where iij and yare appropriate constants. It seems also reasonable
to assume, within the limits of the model. that the expansions or
contractions of the gas in the bubble be uniform, SQ that the power
per unit volume connected with this motion be -roD- 1 V or (see (15.3»

1~ v= iij(:(:I_-v:~Y C~ v)-
Volume changes of the element cause also dissipation of energy in
the liquid crust. The corresponding power per unit mass is

-111e C(/~ v»).

Finally, surface tension effects need not be negligible; they involve
an energy proportional (through a coefficient e) to the surface of the
bubble; an elementary calculation leads to the following expression of
the corresponding power per unit volume:
_ 2e (( 4n) ((1 - v.)v ))1/3 v.
3D. (1 -
V)2

On the other hand, the power per unit volume for the continuum
model is
a a
-(! tr T div i + ev),
where div i = (l/(1 - v.»v.
In conclusion a reasonable choice for e and
a a
tr T which ensures
consistency is as follows:

1
"3
4 (1 - v) d. .
tr Ta = "311e -v- IV X.
46 II. Special Theories

The choice is attractive also because it attributes to the mixture an

effective volume viscosity, namely,
1-
1'1e ( -v- ,
v)
which is the value often accepted in gross theories.

§17. Dilatant Granular Materials

Some of the properties of these continua (which are models of sus-

pensions of rigid spheres in a compressible fluid) are in a sense com-
plementary to those of liquids with bu.bbles. The fluid density is
normally considered negligible compared with the proper constant
density 15 of the the suspended particles. If v is now the volume
fraction of the particles, mass conservation requires, in the absence of
diffusion,
15vl = 15v*. (17.1)
Equation (17.1) can be interpreted as a constraint which renders
the microstructure latent. Equations (15.1), (15.2) apply again but with
new choices for the function X and for the stress (15.7).
As for X an expression of the form
(17.2)
has been suggested, with an appropriate choice of the constants y
(y > 0) andIX; trivially one could even take IX = o.
As a consequence, Eq. (15.7) takes the form
T = f + 15v;l-2(P + YV,!+«,-«/2)(,-2-«/2i)"I + v*r 1 (div S- ,)1.
Because diffusion phenomena are excluded, so must also be
collisions between suspended particles. The theory obviously applies
at most
a
to dilute suspensions; then one can adopt the expression (16.1)
for T and take there for '1 Einstein's formula
'1 = '1e(1 + ~v),
where '1e is the proper viscosity of the fluid.
More appropriately another a
term proportional to the diad
(grad v) ® (grad v) is added to T to account for Ericksen's stress (see
our observations in Section 12).
 §18. The Perfect Korteweg Fluid 47

To study suspensions at higher concentration the Coulomb model

is used sometimes, with the following constitutive rule for the Cauchy
stress
T = (Po.- Pi v2 + ao(grad V)2 + 2a i v L\v)I - 2a2(grad v) ® (grad v)
+ A.(div i)I + 2"D,
where the choice of the coefficients is variously justified.
In conclusion the latent microstructure seems to be a round-
about device to introduce higher-order gradients in the constitutive
equation. In the present context this devious route may appear
artificial; however, in thermodynamics, it may help to avoid certain
inconsistencies, as we shall see.
Naturally, continua with scalar microstructure need not reduce to
continua with latent microstructure. As an example we cite a proposal
of Goodman and Cowin for materials of the type discussed above in
this Section. They exclude the constraint of incompressibility for the
suspended particles and assume that the particles compress/expand
only uniformly; thus the microstructural behaviour is still described by
one parameter only. The balance equations are (IS.lh, (IS.lh, (IS.lh,
and (15.2); p can be expressed again as the product of pv, but P is now
variable; the constitutive prescriptions involve separately T, (, and s.

Remark. Actually there is an important additional hypothesis which is

proposed in some work on granular materials and which appears to
be at variance with the general principles proposed in this tract but
could, in fact, be interpreted as the consequence of a partially latent
deeper microstructure: the kinetic energy density is supposed to
contain a term proportional to (grad V)2. Other proposals involve the
concepts of a "fabric tensor" and of a "granular temperature".

§18. The Perfect Korteweg Fluid

This model of a fluid was introduced to help in the description of
phenomena of surface tension and capillarity; it belongs to a class
of continua with constitutive equations involving higher-order deriva-
tives. We pursue here its interpretation as a continuum with latent
microstructure thus offering a further example of application of the
developments in Section 14.
We begin by recalling the general properties of a perfect fluid with
48 II. Special Theories

one scalar order parameter; the balance equations are eqs. (15.lh,
(15.1h, (15.1h and (15.2). Also, in the absence of internal constraints,
the constitutive relations (12.8), (12.9) apply in the special form
T = p,(aqJ/a,)I - p(grad v) ® (aqJ/a(grad v»,
(18.1)
s = p(aqJ/a(grad v», , = p(aqJ/av).
The question may be raised as to whether (15.2) is satisfied by
(18.1)1. To check that the answer is in the affirmative one could refer
to the general developments of Section 12; but here it is enough to
remark that the function qJ depends on two scalars (1 and v) and one
vector (grad v). For reasons of invariance, qJ may depend on grad v
only through its modulus; hence
T = p,(aqJ/a,)I - 2p(aqJ/a(lgrad vt)2}(grad v ® grad v). (18;2)
What we said so far applies when the choice of the value of v is
free, whereas we want to examine here a case where a constraint
is present. Then recourse must be made to results of Section 14,
preserving, however, for the active components of stress and micro-
stress t" ~ the hypothesis of their dependence on a potential. The
constraint we want to introduce is the coincidence of the values of v
and 1, so that the function W(l) of Section 14 is the trivial one W(l) = 1.
Then, by (14.16), we recognize that it must be

T = sym Ta - 1 [P(P- (ax)"

aV + ax)
aV + div sa - , a] 1. (18.3)

We come now to the constitutive prescription of the active

components. The potential qJ is a function of 1 and grad 1 (which is
exactly the property assumed in Section 12, when we adapt it to the
present circumstances) and thus the following relation must apply

pcp = p(~~ )i + p(a(g~:d I)} (grad It = T· D + (i + ~·(grad i).

Here we must take into account two kinematic identities, of which we
have already repeatedly used the first
i = 1 div X,
(grad It = grad I - (grad xV grad "
and decompose T and D into the sums

T= -f111 + dev t D = ~(DI + dev D.

 §19. Continua with Vectorial Microstructure 49

We get then
pcp = dev(T + grad I ® s)· dev D + «( - t111- 1 + ,-1 (grad I)· s)i
+ s·(grad If
and conclude with
a (a<p)
s = P a(grad I) , (18.4)

(18.5)

The Cauchy stress tensor is finally obtained by substitution of

(18.5) in (18.3), using also (18.4)
a
T= [p,(a<p/a,) - I div(p(a<p/a(grad I))) + pI«ax/ai), - aXia, - (J)]/
- p(grad I) ® (a<p/a(grad I». (18.6)
Leaving aside terms containing X, this is the constitutive equation
prescribed usually for a perfect Korteweg fluid; actually also second
order gradients of I could be easily involved leading to more complex
constitutive formulae at the cost of heavier developments.

§19. Continua with Vectorial Microstructure

In a second wide class of continua with microstructure, Jt is the space

l ' of vectors, or a subset of 1'. Then, we will call d an element of Jt,
vectors are also p and ~ (and we will denote them by b and z,
respectively); finally .!II and !/ are second-order tensors (and for them
we use the symbols A and S, respectively). The balance equations (8.3)
and (9.7) take the form

p( G~} - (~~)) = pb - z + div S, (19.1)

skw T = te(AT z + (grad AT)S). (19.2)

As usual, the choice of the expression for A, which depends on
the physical significance one wants to assign to d, has a decisive
importance for the ensuing theory. We mention below two very
different choices, both significant.
For an ample class of continua, comprising liquid crystals, d is used
to identify a privileged material direction, marking the alignment of
50 II. Special Theories

rod-like molecules; in another class d is interpreted as a displacement

(case viii of Section 2). In both cases, in a rigid motion with angular
velocity w, d is equal to w x d; hence, as we have remarked more
than once,
A =ed. (19.3)
As a consequence, Eq. (19.2) becomes
skw(TT + d ® z + (grad d)ST) = O. (19.4)
Alternatively d may be used to specify a rigid rotation, associated
with the orthogonal tensor R = e-ed. The material element is then
viewed as a rigid body capable of rotations which are independent of
the local rotation in the macromotion (as occurs in Cosserat continua
or in biaxial liquid crystals); the microstructural orientation· is
obtained from a ftxed one through R. Unfortunately a compact
formula is not available which expresses the vector associated with the
tensor R(q) which is product of two orthogonal tensors R(q) = QR
(Q = e-eq). A possible way out is to interpret the vector not as the
product d = Oe (0, angle of rotation, 0::::;; 0 ::::;; n; e unit vector) as
suggested above but rather as k = e tg 012.
If we imagine d and k issuing from the same origin, the change
d --+ k sends the ball of radius n (on the boundary of which antipodes
are identifted) onto the projective space (see remarks under case vi in
Section 2).
Then, if h is the vector related to q in the same way as k is
related to d, a known formula gives the following expression for the
transformed vector k(q)
k _k+h+hxk
(q) - 1 - h·k (19.5)

Putting in evidence ftrst order terms in h, or in q, we have

k(q) = k + (ek + I + k ® k)h + o(h)
= k + !(ek + I + k ® k)q + o(q),
so that we obtain the expression for A
A= !(I + ek + k ® k).
It follows in particular, by (19.2), that
skw T = ie(z + (z· k)k +k xz + (S· grad k)k + (grad k)(STk»,
a formula obviously quite distinct from (19.4).
 §19. Continua with Vectorial Microstructure 51

We return now to the case when d is interpreted as a displacement.

The simplest image one can conjure of a material element is that of a
cell of a centered cubic crystal with the central atom belonging to a
different species from the corner ones and free to move, within the
cell, without participating exactly in the mean motion.
Then one might want to know not only the mean motion, but also
the relative displacement which is thus taken as the microstructural
variable d. Elementary considerations suggest for the density of kinetic
energy the expression
tPi 2 + tp~(1 - ~)d2,
where ~ and 1 - ~ are the concentrations of the two atomic species.
Although these premises seem to imply the possiqP.ity of diffusion,
such phenomenon is excluded here: Idl is supposed to remain confined
to less than the atomic spacing. Then ~ must be taken as a constant,
the balance equations are (8.2), (19.1), (19.4) and there z represents the
interaction force per unit volume between species. The interest centres
usually on the conservative case: then the potential q; is taken to be a
function of F, d and Grad d; an immediate application of eqs. (11.6)
gives
oq; oq; T
z = P od' S = P o(Grad d)F
A number of further results are available, in particular for the
linear approximation.

An interesting subclass of continua with vectorial microstructure is

characterized by the existence of a potential w for the field d. In that
class the following constraint applies
d = grad w. (19.6)
To obtain the evolution equations in this case, one starts again
from the balance equation of momentum and from eqs. (19.1), (19.2),
assuming that T, z, S be the sum of active and reactive components
and that the property of vanishing power for the reactive components
applies
r r. r •
T· grad i + z· d + S· grad d = 0 (19.7)
for all velocity distributions allowed by the constraint (19.6), i.e. for all
d of the type
d = grad w- (grad if grad w. (19.8)
52 II. Special Theories

Introducing this expression of il in (19.7) we get

(T - d ® f - (grad d)ST). grad x + f· grad dJ
+ S·grad 2 dJ - (d®S)·grad 2 x = 0,
and hence, the properties
r r r
S E Skw, f = 0, T = (grad 2 W)ST. (19.9)
As (19.7) applies for all allowable velocity distributions, it is valid
in particular for rigid body velocities; so, we can write (19.2) in terms
of active components only
a Ta T a
skw T = te(A z + (grad A )S). (19.10)
Taking into account the first two conditions (19.9) we deduce from
(19.1) the pure equation

diV(P(:~} - :~) - pb + z- div s) = O. (19.11)

From (19.1), with the help of the second and third equations of (19.9)
we deduce also

T= dived ® S) - d® {p ((:~} - :~ - b) + z- div s};

recourse to (19.9)1 leads finally to the pure version of equation of
momentum balance

pi + diV(d ® p( (:~} - :~)) = pf + div(pd ® b)

a a a
+ div(T - d ® (z - div S».
(19.12)
When (19.3) applies, (19.10) becomes
a a a
skw T = skw(d ® z + (grad d)ST)
and the last term in (19.12) can be written
a a
div(sym T + dived ® S». (19.13)

We pass on now to some developments which are a necessary

premise to the study of uniaxial liquid crystals; precisely we examine
the consequences of the introduction of the constraint Idl = 1 for d.
The constraint implies that micromotions belong locally to the class
of rigid rotations. Hence we can adapt to the special case certain
 §19. Continua with Vectorial Microstructure 53

results obtained in Section 14, in particular those expressed by

formulae (14.6) and (14.7). First of all from (14.6) we observe that, in
view of (19.3), in our case the Cauchy stress is expressed again by the
sum
a a a
T = sym T + skw(d ® z + (grad d)ST). (19.14)

The pure equation of evolution for the microstructure reads

dx [p (G~) - b) + z- div sJ = 0; (19.15)

hence the vector in square parenthesis must be parallel to d. Here in

expressing the inertia term use was made of a property of invariance
of X (see the remarks regarding the function K in Section 6); X
a
depends on d, at most through lal 2, because in the "present case d· a
is null and Idl 2 constant; usually one thinks of X as quadratic in
lal 2 : X = taa 2•

The preoccupation to frame the theory of liquid crystal within a

general scheme is absent in specific studies. Then, d is taken to be
a unit vector at the outset and the pure equation of evolution is
proposed directly. Also, in general, rather than expressing that
equation as a vector product as in (19.15) the equivalent form is
preferred
p(ad - b) + z - div S = Ad, (19.16)
where, however, a trace of the implicit contraint remains in the
presence of an unknown Lagrange multiplier A.
An alternative form of Eqs. (19.14) and (19.16) can be obtained by
introducing, as a new kinematic variable, the angular speed h associ-
ated with a;
without essential restriction and with the advantage of
uniqueness, one may think of h as orthogonal to d
a= (ed)h, h = -(ed)a. (19.17)
These formulae can be considered as special case of an anholonomic
version of the change of variables discussed in Section 10 (see formula
(10.7) and the following ones). Here the tensor ed takes the place of
the operator r which appears in (10.7); as a consequence, by (10.8)
(but also by direct calculation), one obtains in particular

dx G~} = (~~)'.
Furthermore, formulae (10.10) suggest introduction of the microstress
variables
54 II. Special Theories

2: = d x z + e«grad d)ST), S= -(ed)S, (19.18)

whence
-2: + div S= d x (-z + div S),
with the final consequence that (19.15) is replaced by

p G~)" = ph - 2: + div S (19.19)

if h = d x b, whereas (19.14) becomes

T= sym T+ e2:. (19.20)

§20. Uniaxial Liquid"Crystals

As mentioned above these liquid crystals are a subclass of the

continua with vectorial microstructure for which the condition that d
be a unit vector applies and furthermore special choices are proposed
for the constitutive functions T, " s. For completeness also the
function X should be specified; in practice microstructural inertia
effects are hard to record. In any case we will mention here, for the
sake of brevity, only a few results in statics and for that purpose not
only is the form of X irrelevant, but also the liquid crystal can be
modeled, without essential loss, as a perfect incompressible fluid
and so we can invoke the general results of Section 12. The condition
of incompressibility changes those results only in so far as I disappears
from the variables which influence stress and microstress and, at the
same time, a contribution appears to the spherical component of the
Cauchy stress which has reactive character.
If '(JJ is the pressure due to the constraint of incompressibility, the
constitutive relations (12.8) and (12.9) can be adapted to the present
case as follows

T - - I - ( d d)T o((J
- '(JJ P gra o(grad d)'

o((J o((J
S = P o(grad d)' z = p od'
It remains for us to choose the potential function ((J(d, grad d),
subject to properties of frame indifference and for d in the class of unit
vectors. Frame indifference requires that ((J satisfy the condition
 §20. Uniaxial Liquid Crystals 55

skw(d ® (acp/ad) + (grad d)(acp/a(grad dW) = 0, (20.1)

for all d for which
d Z = 1, (grad df d = O. (20.2)
Besides, a change in the orientation of d should not affect cpo
Condition (20.1) is satisfied if and only if cp is a function of d and
grad d (with properties (20.2» through their scalar invariants which
are not affected by the change of d into - d. The search for the
appropriate invariant base, i.e. the smallest set of physically significant
independent invariants, suggests the choice of the five invariants
(div df, (d, rot d)Z, I(grad d)dI Z,
(20.3)
(grad d)' (grad df, (det grad d)z.
The last of these is of the sixth degree in the components of
grad d; for reasons for simplicity it is traditional to drop this invariant
from the list and to propose an expression of cp which is of degree two
in grad d.
For the class of nematic liquid crystals one requires moreover that
cp vanish on uniform director fields; hence one takes
2cp = ("1 - "2 - '(4)(div df + "zid' rot dl z
+ "3 I(grad d)dl z + ("z + '(4)(grad d)' (grad df; (20.4)
Here "1' "z, "3' "4 are appropriate constants; the use here of such
curious combinations as coefficients leads below to simpler formulae.
The interesting identity
(grad d)' (grad d)T = (div d)Z + div«grad d)d - d div d).
can now be called upon to modify not only formally the last term in
(20.4). In fact to change cp by a term having the form of the divergence
of a vector is equivalent to an appropriate related change in the
surface potential energy rather than to an essential change of the
volume potential energy of the body. In the present context, where we
pursue the expression of cp that leads to the constitutive prescriptions
for the stress, such a term becomes irrelevant and can be crossed out.
Notice also that
(grad d)d = «grad d) - (grad df)d = -(e(rot d»d = -d x rot d.
Thus we can write (20.4) in the final form (Frank's potential)
2cp = "1 (div d)Z + "z(d' rot d)Z + "31d x rot dl z , (20.5)
56 II. Special Theories

where only three constants appear, as coefficients of quantities which

are physically significant measures of the distortion in the liquid
crystal, that is, of the discrepancy of the field d from uniformity. The
scalar quantities div d and d· rot d are called, respectively. the splay
and the twist of the field, whereas the vector d x rot d is called the
bend.
The explicit form of the constitutive equations is
z = p("2(d·rot d) rot d + 2"3«rot d)2d - (d· rot d)d».
S = p("l(div d)I + "2(d·rot d)ed + "3d ® «grad d)d), (20.6)
T = -rnI - (grad dfS.
They are still too complex to lead to closed solutions even of
simple problems, so that it is often supposed (though in contrast with
experimental evidence) that "1 = "2 = "3; then qJ reduces to
qJ = !"l(grad d)· (grad d), (20.7)
the microstress becomes
z = 0, S = P"l grad d, (20.8)
whereas the Cauchy tensor is
T = -rnI - p"l(grad d)T grad d,
with the conclusion that, in the absence of external actions, the
microstructural balance equation requires that d satisfy a simple linear
elliptic partial differential equation
div(p"l grad d) = A.d. (20.9)
This conclusion is interesting because it is possible to take
advantage of many well-known results valid for this type of equation
in the solution of special problems arising in the statics of liquid
crystals.
We remark that in the case of plane fields the constant "2 has no
influence (because d· rot d = 0 identically) on z and S; then to obtain
"1
(20.9) it is sufficient to assume that = "3.
In cholesteric liquid crystals, because the molecules look like helical
springs, the natural microstructural state is not uniform; rather it is of
the type d = (cos Te, sin Te, 0) if one chooses appropriately the e-axis.
For the potential qJ the following expression is advanced
2qJ = ", (div d)2 + "2(d· rot d + T)2 + "31d x rot d1 2.
where T is the so-called chiral constant.
 §21. Continua with Affine Microstructure 57

We quote finally the class of smectic liquid crystals; they are

characterized by the existence of a potential for the field d. For these
crystals both constraints Idl = 1 and (19.6) apply; torsion and bend
are both identically zero.

§21. Continua with Affine Microstructure

The affine microstructure has been studied by theorists much longer
than all other types of microstructure; contributions to the theory
date from the beginning of the century and even earlier, in restricted
contexts. Many authors in fact consider the class as exhaustive
and propose the special relations and properties "listed below as
appartaining to the general theory. They argue that, as the micro-
structure must account for local kinematic discrepancies from the
macromotion, a linear description of such a discrepancy must suffice;
however, the example of bubbly liquids (and others that could be
quoted) negates the general validity of the argument.
Within the class, v is an element, denoted here by G, of the space
Lin of second-order tensors or of a subset of Lin, such as Lin +. II and
~ are also second-order tensors (which will be denoted here by B and
Z, respectively), whereas d and f/ are third-order tensors (and for
them the symbols a and s will be used). The microstructural balance
equation and the balance equation for moment of momentum take
the form
p( (:~)" - :~) = pB - Z + div s, (2l.1)

skw T = te(aZ + grad(aT)ts}. (21.2)

To proceed one must attach a physical significance to G. Most
commonly it is presumed that each element of the continuum is
capable of an affine deformation distinct from and independent of
the local affine deformation ensuing from the macromotion and
characterized, locally, by the tensor F. From a geometric point of
view one can think G as having a role similar to that of the tensor
that specifies the local permanent deformation in a plastic body.
Then G must belong to Lin +; more precisely G can be thought of
as a double vector, and, as a consequence, we will denote its cartesian
components with one small and one capital index: Gu . A rigid
rotation of the whole body, t;haracterized by the orthogonal tensor Q,
58 n. Special Theories

changes the values of G everywhere into QG; with our usual notation
we will write
G(q) = QG. (21.3)

One deduces for a the expression

aiJk = elkm GmJ (21.4)
so that (21.2) becomes
skw( - I i j + G1LZjL + G IL ,.SjL3) = o.
Ij

Actually, it is not immediate to recognize in this equation and in

(21.1) the balance equations which appear in current analyses of the
dynamics of continua with affine microstructure (or of micromorphic
continua as they are called by some authors). There are differences,
but not only of notation; first of all the function X is always assumed
to be quadratic in G
X = tG'(JG), (21.5)
and the fourth-order tensor J is taken to be constant, so that it must
have the form
J iJhK = ~lhJJK'
where the second-order tensor J is constant, symmetric and positive
definite. As a consequence
X = t(GJ)' G, (21.6)
and the left-hand side of (21.1) reduces to pGJ.
Furthermore the microstructural stress variables are also usually
defined differently. In particular, use is made of the symmetric tensor
(21.7)
a tensor which can be introduced in (21.1) after multiplication of both
sides byG, with the following result:
pGJ(JT = pB + TT + Y + div b, (21.8)
where two new tensors are introduced:
B:= GB T and bjjk = GIHSjHk'

It is traditional finally to measure the microstructural inertia through

a new tensor
(21.9)
which is in one-to-one correspondence with J, but is not constant. It
satisfies rather the equation
 §22. Micromorphic Elastic Continua 59

(21.10)
where a new kinematic tensor appears
W:= GG- 1• (21.11)
Some authors consider (21.10) as a new fundamental equation of
conservation, parallel, in some sense, to the equation of continuity for
the macromotion. Here, however, the equation appears as a trivial
consequence of the defmition (21.9); note also that we could easily
imagine, at least in principle, cases in which J depends on F and G
with radical changes in the consequences. Leaving this question aside,
we conclude that, when the left-hand side is expressed in terms of the
tensors J and W, Eq. (21.8) takes the form that usually appears in the
literature; that is,
pJ(W + w2f = pjj + TT + Y + div h, (21.12)
because

Remark. The transformation of (21.1) into the form (21.12) can be

interpreted also as the consequence of a change of variables, similar to
those envisaged in Section 10 but now anholonomic, in accordance
with the rules

Also the analogy with the final remarks in Section 19 is very strict.

§22. Micromorphic Elastic Continua: Bodies with

Continuous Distribution of Dislocations
Eringen and Grioli have studied with particular attention the case
of hyperelastic micromorphic continua, a case when the constitutive
prescriptions for T, Z, and s derive from a potential cp. For that
purpose it is appropriate to determine first a group of geometric
variables which have the necessary invariance properties and can be
adopted as deformation characteristics because they vanish if and only
if the complete displacement of ~ from the reference placement is
rigid.
As G transforms under rigid rotations in accordance with t21.3), a
convenient set seems to be the following couple of tensors:
(22.1)
60 II. Special Theories

It is very easy to verify that the displacement of ~ is rigid if and only

if E and M vanish simultaneously. Besides the values of E and M on
any element of ~ are both unaffected by a rigid rotation of the body.
Hence a good choice for cp is a function of E and M. Reference to
(11.6) shows that the following constitutive equations ensue:
acp p. acp
P = p.F aE + 2 G aM'
(22.2)
acp)T
Z=!p.F ( aM ' s= 0,

or, alternatively, using also definition (21.7)

acp T 1 .acp T
T=pFaEF + "2 PG aM F ,

acp
Y=-pFaEF T
-sym (alP
pFaMG T) , (22.3)

s=O.
Both these prescriptions render the moment equation (21.2) auto-
matically satisfied.
More exacting circumstances occur when a dependence of lP also
on Grad G must be envisaged. A set of appropriate variables from
which cp might depend (at least as far as the present requirements are
concerned) are E, M and the third-order tensor
(22.4)
the time derivative of which is actually in a simple relation with the
gradient of the kinematic tensor W introduced in preceding section
f8n = Giit Wij.L Gi]·
As for the constitutive relations that follow from the hypothesis
that lP depends on F, M, and f, we remark only that (22.2)1 and
(22.2}z remain formally unchanged whereas s turns out to be different
from zero.

We conclude this section with some remarks which open up other

fields in which the concept of tensorial microstructure may be applied.
In the theory of continuous distributions of dislocations a tensor such
as G is used to identify the triad of edges dU )' d(2)' d(3) of a crystal
lattice (at least in some "averaged" sense) though the rule
d(R)i = GiR •
 §22. Micromorphic Elastic Continua 61

The lattice, even in deformed placements, is thought to be perfect if

one can associate the directors d(R)i with a global placement, possibly
distinct from the apparent placement. Contrariwise the presence of
defects is put in relation with denial of the rule of congruence
(Grad G)t #- Grad G.
More precisely, for reasons that would be too long to justify here, an
appropriate measure of defectivity of the lattice is the tensor density of
dislocations B
3
B := LR d(R) ® rot d(R),
1

where d(R) (R = 1, 2, 3) is the dual triad of directors defined by the

property

-1
easy developments show that in terms of G and G the tensor B is
given by the formulae
-1 -1
Bij = GipejmnGpn,m = eABcGAjGiB,C'

Thus, B vanishes if and only if the congruence property is satisfied.

Furthermore, this tensor has the property that, given locally a plane
of unit normal n, then
b = Bn {22.5)
is the so-called Burgers vector relative to that plane.
In the most interesting cases B is a diad: if, locally, the density
of dislocation is purely of the screw type with axis along the unit
vector c, with corresponding Burgers vector equal to tXc, then
B = !Xc ® c;
if, locally, the density of dislocation is purely of the edge type with the
edge along the unit vector c and Burgers vector equal to !Xc (Icl = 1,
C'c = 0) then
B = tXc® c.
A complete dynamical theory is not yet available; one must expect
that an appropriate expression for X be a function of B and that a
potential for the microstress might possibly exist and be expressible in
terms of B and grad B, though the directors, or G, could also be
involved. If it were possible to use only B as microstructural variable,
62 II. Special Theories

the balance equation for the microstructure would still formally be

(21.1) with B in the place of G; but the form of the equation of
balance of moment of momentum could not be (21.2), because the rule
of change of B under rigid rotation of the body is not that valid for G.
If G were also involved together with B as microstructural variable
then the balance equations would be far more complex.

A last connection of the theory of the preceding section can be

stated, at least for certain aspects, with what is called theory of
thirteen moments. It is a theory applicable mainly to certain fluids
with microstructure in which the unknown fields are exactly thirteen,
i.e., with our notation: the density p, the macrovelocity field v and the
nine components of the microvelocity tensor W. The name comes
from the fact that the equations of balance usually are derived from
the kinetic theory of gases through evaluation of global moments of
the molecular equations of motion.

§23. The Continua of Cosserat

At this point in our exposition the Cosserat continua can be seen

from many different, but substantially equivalent, points of view. They
could be viewed as continua with vectorial microstructure (q, a vector
of rotation) on the basis of certain remarks in Section 19; alternatively
they could be classed among continua with atfme microstructure, but
subject to the constraints
det G = 1,
which require of G to be a proper orthogonal tensor. Here we
prefer to adapt certain developments of Section 14; in fact, as the
microstructure is now determined by the assignment of R e Orth +, the
microvelocity is locally of the type occurring in rigid microrotations,
namely,
R=WR, WeSkw
and (14.4) also applies if only one represents W in terms of a vector
h: W = eh. In components

RIH = ~jkhkRjH'
The third-order tensor 8iHk = eijkRjH here takes the place of the
operator d.
 §23. The Continua of Cosserat 63

For the microstructural actions we adopt the notation introduced

in general in Section 21, and hence we write (14.7) in the form

skw {p G~ RT)" - pBRT + ZR T - (div S)R T} = 0, (23.1)

where we have abandoned the superscript a for Z and s.

It proves convenient to introduce the vectors band z and the
tensor S as follows:

Zi = teihk(ZhHRkH + ShHjRkH,), (23.2)

Sij = teihkShHjRkH,
and to remark that the following relations apply:
skw BRT = eb,
(23.3)
skw(ZRT - (div S)RT) = e(z - div S),
It is also appropriate to recall that, as X must be invariant under
changes among stationary observers, its dependence on Rand R
can occur only through the product RTR, and that this tensor is
skew so that it can be expressed in terms of one vector. We choose
the vector

for which

and so

:~RT = e( R:~). (23.4)

As a consequence of (23.3), (23.4) the equation of balance (23.1)

becomes

p(R :~)" = pb - z + div S. (23.5)

If X depends on u only through its modulus, then X can be thought

of also as a function of w = Ru and, as a consequence, the left-hand
member in (23.5) can be written p(OX/owr.
We must adapt to the present circumstances also (14.6); the result
64 II. Special Theories

is simple because of the property expressed by (23.2h

T = sym T + ez. (23.6)
In conclusion the pure evolution equations for the Cosserat con-
tinua are the Cauchy equation (with a generally nonsymmetric stress
tensor) and a vectorial equation of evolution for the microstructure

p(R:~} = pb -ieT + div S. (23.7)

If, as for micromorphic continua, one assumes that the internal

actions derive from a potential cp which is a function of F and R, one
can obtain constitutive relations akin to (22.2). However, under the
present circumstances, one tensor, that is" . M = i(RTF - I), suffices to
characterize the deformation; in fact it is easy to check that E can be
expressed in terms of M
E = 2MTM - MT - M. (23.8)
In conclusion, cp depends on F and R through M only and the
constitutive prescriptions (22.3) become simpler
1 ocp T
T = 2PR oMF , s = 0,
(23.9)
z = tpe(F :;RT).
Note that these results can be used also in direct formulations of
the theories of plates and shells.

Remark. One can repeat here a comment already expressed within a

slightly different context in the final Remark of Section 21. The
kinematic variables which exchange their role here are Rand u; there
is a formal identity between (19.19), (19.20) on the one hand and
(23.6), (23.7) on the other, although in the first two relations all
vectors must be taken to be orthogonal to d.

§24. Biaxial Nematic Liquid Crystals

These materials are incompressible fluids with a Cosserat type of

microstructure, so that the mechanical balance equations are the
Cauchy equation and (23.7). If we restrict our attention to static
phenomena we can imagine the fluid perfect. The open question is the
 §24. Biaxial Nematic Liquid Crystals 65

appropriate choice of the potential ffJ of internal actions; ffJ may

depend now on R and its gradient. Actually as we intend to represent
the behaviour of a fluid, we must imagine all constitutive quantities
as functions of variables involving the present placement only; to
this end we introduce a new proper orthogonal tensor Q such that
R = QR, where R is the value of R in a fixed place of !1IJ and we think
of ffJ as a function of Q and grad Q. Recourse to (12.8), (12.9), (23.2h
and (23.2h with the introduction of the appropriate changes leads to
affJ
Tij = -rnDij - Qmn,iPa-Q.'
mn,]

Zi = teihkP (aaQffJ Qkm + Qkm,j aQa ffJ .),._ (24.1)

hm hm,j
1 affJ
Sij = 2e ihkP a-Q_Qkm'
hm,j
Leaving aside the question of the most general function admissible
to represent ffJ, we deduce here only the consequences of the direct
hypothesis that ffJ depend on Q and grad Q, only through the product
QhiQhj,kQkm' a third-order tensor which is skew in the indices i, j and
thus can be represented through a second order-tensor, as follows:

with
Anm = teijnQhiQhj,kQkm'
Then we derive from (24.1) the following constitutive relations;
affJ
T = -rnl - pQAT aA QT,
(24.2)

For instance, in the very simple case when

ffJ = tK1A' A = tKl(grad Q)·(grad Q)
one gets in particular

Z =0, (24.3)
66 II. Special Theories

The static balance equation for the microstructure can be derived

from (23.7):
Kl skw((L\Q)QT) + 2eb = 0;
when external actions are absent the tensor (L\Q)QT must be
symmetric, and, in view of the properties of orthogonal tensors,
necessarily
(Qih,kQjh),k = O.
Ordinary nematic liquid crystals can be immediately included as
a special case by adding the hypothesis that qJ depends on Q and
grad Q only through a unit vector Qd and a tensor grad(QJ), where J
is a constant unit vector.
A particularly simple case, still nearer to the case treated in Section
20 occurs when, with the same meaning of d and d, one has
qJ =!K1A·(I -d®d)A =!K1(A·A -IATdF) = !K1(grad d)· (grad d).
Then
T = -mI - pK 1QAT(I - d ® d)AQT = -mI - pKdgrad d)T grad d,
z = 0,
We could not expect the above expression for S and that given by
(20.8)2 to be the same because here the microstructure is characterized
by Q rather than by d. The two expressions are related to one another
through an appropriate adaptation of (10.10)2, where the operator r
must be derived from the formula
d= (I ® d)Q.
 PART III

Thermodynamics

§25. Balance Equations

We proceed now to suggest plausible thermodynamic properties for

continua with microstructure. Actually we should pause to consider
first of all whether there could be, in thermal phenomena, an inherent
complexity parallel to that of geometric and mechanical character
which we have attributed throughout this tract to each element of the
medium, in other words if some form of thermal microstructure exists.
But that question is far too delicate and controversial; we will leave it
totally open. We will rather conjecture that, even in the presence of
microstructural phenomena, it be possible to measure unambiguously
on each element at each instant along any process a density of
internal energy I> per unit mass, a heat flux vector q, a rate of heat
generation .Ie per unit mass due to radiation or other sources, an
absolute temperature () and a density of entropy per unit mass 1'/. We
concede that this conjecture is here far fetched (it has been criticized
even within the context of ordinary continua), but we accept it
nevertheless and show that it leads to some not totally unwarranted
con seq uences.
We will also accept the validity of a balance equation for energy,
an equation modeled on the one which applies in the classical context,
but where account is taken of the additional kinetic energy due to the
micromotion and of the internal and surface power expended on the
microstructure. Specifically we postulate the validity, over the region 6
68 III. Thermodynamics

occupied by anyone of the subbodies of f!A, of the following equation

of balance

(f,/(e + ti + K)} = Lp(f. i

2 + P' v + A.) + fad (t· i + 't . V - q' n).
(25.1)
Along processes which are sufficiently regular to assure the validity
of the kinetic energy theorem, Eq. (25.1) can be substituted by

(L L
pe} = (T·grad i + ~'v + [/'·grad v) + L p..1. - fad q'n;
and, when account is taken of the arbitrariness of ~ and the classic
argument of localization is repeated, the--following relation can be
deduced:
pe = T· grad i + ~ .v + [/'. grad v - div q + pA.. (25.2)
Along a regular surface of discontinuity which propagates with
normal velocity oc, again a classical argument leads to the jump
condition
[p(e + ti 2 + K)]OC + [-q + TX + [/'T v]· n = O. (25.3)
A referential version of (25.2) can also be given where the Piola
type of stress and microstress appear (as defined by (8.6), (8.7)
together with the referential heat flux vector
k:= 1F- 1 q.
That version is as follows
p*e = p. p + :.. v + ~. Grad v - Div k + p*..1., (25.4)
to obtain which the identity
div q = ,-1 DiV(lF- 1q)
was exploited.
In a similar vein we admit also that the entropy inequality still
applies in exactly the classical Clausius-Duhem form

(1dP11)"a d-f e~ 1p~,e

2': +
d
(25.5)

with the local version, also standard, valid along sufficiently regular
processes
 §26. Interpretation of the Equations of Balance 69

. d' q
PI1 ~ - IVe + O·
p).
(25.6)

In a "reduced" version of this inequality the Helmholtz free-energy

density I/! appears, namely,
(25.7)
the reduced inequality is
. .
P(I/!+110)--T'gradx-~'v-9"gradv+
(q'gradO)
0 ::;0 (25.8)

or, in a referential form,

. " k·GradO
P*(I/!+110)-P'P-3'v-&"Gradv+ 0 ::;0. (25.9)

§26. Interpretation of the Equations of Balance

For the equations of balance, Coleman and Noll have proposed an

interpretation which has given rise to much debate, but has been
widely accepted, because it leads to some new, interesting results and
confirms and generalizes some old statements, known before to hold
only under restricted circumstances. Here we illustrate the interpreta-
tion within the context of continua with microstructure.
Substantially Coleman and Noll attribute to (9.4) and (25.8) (which
imply the validity of a balance, or imbalance, relation to be realized
totally within the material element without influence of actions at a
distance) a differ~nt analytical character than that they attribute to
(8.2), (8.4) and (25.4), where there is an influence of actions at a
distance, represented by pf, plJ and p)., an influence that, at least in
principle, could be arbitrarily adjusted in a different way element by
element and instant by instant. Consequently they require that the
constitutive properties of each material be such as to render (9.4),
(25.8) satisfied identically, that is, along any thermodynamic process,
and vice versa assume that in each material element any thermo-
dynamic process can be realized provided that the actions at a
distance be appropriately regulated. No other restriction is imposed in
principle to the class of admissible processes by the equations of
balance (8.2), (8.3), (25.4); naturally each specific process in a body is
consequent to a different assignment of the actions represented by j, IJ
70 III. Thermodynamics

and A (and of appropriate initial and boundary conditions), so that

(8.2), (8.4), (25.4) take on the role of evolution equations.
The material properties of the element are assigned through consti-
tutive hypotheses of kinematic and thermomechanic character. The
former involve the functions d(v) and K(V, \I); the latter will be ren-
dered explicit below only for a special class of materials, which is the
simplest and most direct generalization of the class within which
Coleman and Noll originally tested their proposal. The class com-
prises the materials for which free energy 1/1, entropy rf, stress P,
microstress 3 and [!P and the heat flux k can be expressed in terms of
the local values of deformation gradient F, of the microstructural
variable v and its referential gradient Grad v, of the absolute tempera-
ture e and its gradient Grad e:
1/1 = tjJ(F, v, Grad v, e, Grad e), etc. (26.1)
Let us check the compatibility of these constitutive prescnptlOns
with the Clausius-Duhem inequality in the reduced version (25.9).
We require that (25.9) be valid for all values of F, v, e and their
derivatives. When the terms are appropriately ordered the inequality
reads

otjJ - P)..F + (otjJ

( p* of p* ov - 3).
. v + (otjJ
p* 0 Grad v - [!P
) . Grad v.

otjJ ) . otjJ .
+ p* ( oe + rf e + p* o(Grad e) . Grad e- (kle)· Grad e : : ; o.
Because the left-hand member is linear in ft, \I, Grad \I, Grad e, e
and in view of the fact that one can imagine, for each material
element, processes along which at a given instant these quantities take
up arbitrary values, the coefficients in the linear expressions must all
vanish, and hence

otjJ otjJ
3 = p* ov' [!P = p* 0 Grad v' (26.2)

otjJ otjJ
rf = - oe' o Grad e = o.
Thus, the constitutive equations for P, 3, [!P, and rf are determined
as soon as the constitutive equation for 1/1 is known; besides tjJ (and as
 §27. Thermodynamics of Continua with Latent Microstrcture 71

a consequence also P, 3, fJJ, and 1/) is independent of Grad e. Finally

the prescription for k must satisfy identically the inequality
k· Grad () :s 0. (26.3)
Vice versa, once the functions ~ and k have been assigned (the
latter so as to satisfy (26.3)) and once the relations (26.2) are con-
sequently accepted, then the Clausius-Duhem inequality is auto-
matically satisfied along any process.
We must now assure compatibility with (9.8), which can also be
written, using (26.2), as follows:

(a~
-e aF F
T)
+d
T a~
av
a~
+ (Grad d) a(Grad v) =
t
0. (26.4)

One need only recall here the developments in Section 5 and in

Section 13 (in particular (13.2)) to recognize that (26.4) is equivalent to
d ,
dq !/I (QF, v(q), Grad v(q), ())Iq=o = 0, (26.5)

where Q is, as usual, the proper orthogonal tensor e-eq. Thus (26.4),
when understood to be valid for all choices of F, v, Grad v, and e,
expresses the condition of frame-indifference for ~, namely,
~(QF, v(q)' Grad v(q), ()) = ~(F, v, Grad v, ()), VQ E Orth+.
We conclude the section by giving the local, rather than referential,
version of (26.2):
T = p(a~/aF)FT,
~ = p(ahav), (26.6)
ff = p(a~/a(grad v)).
These relations give ~ the role of potential for stress and micro-
stress. The setting, however, is here wider than that envisaged for qJ in
Section 11, so that one manages to derive also relation (26.2)4.

§27. Thermodynamics of Continua with

Latent Microstructure

The results of the preceding section depend totally on the possibility

of an arbitrary choice, on any material element, of the values of ft, it,
Grad it, e, grad (), and grad e.
If constraints were present such as to
72 III. Thermodynamics

condition and limit that choice, then the proofs would fail. The results
are replaced by others, which cannot have general character, however,
but change with the choice of the constraint. In this section we
explore the matter with reference to an extreme condition: a case of
latent microstructure when the order parameters are totally determined
by the macrodeformation gradient and by the temperature
v = ro(F, 0). (27.1)
We are interested in the case when (27.1) expresses an internal
constraint. Thus, the function ro must be such as to ensure the
objective character of the relation: for all proper orthogonal tensors Q
v(q) = ro(QF, 0)
and this condition implies and is implied (in accordance with the
results of Section 5) by

d«j = -ejpq(:~:)FpA' (27.2)

It is necessary at this point to be precise as to the reactive

capabilities of the constraint, which is now of thermomechanic rather
than purely mechanical type, and to attribute in this wider context
a meaning to the notion of perfect constraint. Both questions are
controversial; on these matters we accept here as valid the following
general assertions:
(i) All quantities, which, in the absence of constraints, are ruled by a
constitutive prescription (i.e. ifJ, 17, P, 8, &', k) are now the sum of
one active and one reactive component
a r
ifJ = ifJ + ifJ, etc.
(ii) Only the active components are bound through constitutive rules
to the thermokinetic variables (here the set: F, v, Grad v, 0,
Grad 0).
(iii) The contribution of reactive components to the left member of
(25.9) is identically zero for all processes allowed by the constraint
r
r r' r. r r k·GradO
p*(ifJ + 170) - p. F - 3· v-&' . Grad v + 0 = O. (27.3)

If we go back to (27.1) and observe that the constraint leaves locally

e,
the choice of F, Grad F, Grad eand Grad 0 totally free we deduce
from (27.3) the relations
 §27. Thermodynamics of Continua with Latent Microstructure 73

(27.4)

r
k = O.
Recourse to (25.9) leads instead to
a
01/1 a OIDa a a (oIDa)
-P*oF
,H
+ Pm + of,H Sa + f!J>aK of
,H ,K
= 0,
a
01/1 a a OIDa a (oIDa)
- P* 00 - P*1J + Sa 00 + f!J>aK 00 ,K = 0,

-P* ( ~
a
01/1
,H,K
a
01/1)
+ ~ + f!J>aK
a

,K,H
IDa a IDa
oF + f!J>aH oF = 0,
° ,H
° ,K
(27.5)

OIDa
a
01/1 a
-P*M + f!J>aK ()() = 0,
,K
a
kHO,H :$; O.
By adding the left-hand members of (27.4)1 and (27.5)1' and using
immediate consequences of (27.4h, and (27.5h, together with (8.Sh,
we obtain

(27.6)

with

(27.7)

where, for brevity, the notation (7.1)2 is introduced. The term within
brackets in the right-hand side of (27.6) has null divergence; in con-
clusion the Cauchy equation (8.8)1' where for P we put P given by
(27.7), is a pure equation which rules the mechanical evolution of the
body.
74 III, Thermodynamics

Similarly, by adding the left-hand members of (27.4h and (27.5h

and using immediate consequences of (27.4)4, (27.5)4' one obtains a
pure expression for '1

0'" + ae'(P 0"')

a ' a

'1 = - oe oro - ((X)v=m) + p*1 DIV

' (P*o Grad e . (27.8)

It is also immediate to remark that, by (27.4), (27.5), the balance

equation for energy can be written
p*e = p.(~ + '10) - Div k + p*A..
Recalling (25.7) one gets finally
p*e~ = - Div k + P.A., (27.9)
where there is no trace, in view of (27.8) and' (27.4)6' of effects due to
the constraint; thus we have obtained a pure equation of evolution for
the temperature.
Note that, if ro is not temperature dependent, then (27.5)4 excludes
a dependence of '" on Grad e. Moreover the usual expression of '1 in
terms of", would ensue in view of (27.8).

§28. Comparison with the Traditional Class of

Hyperelastic Bodies
Continua with latent microstructure (in particular those which belong
to the class discussed in the preceding section) share some fundamental
properties with classical continua (in particular with hyperelastic
media): their thermomechanic evolution is determined by two fields
only, displacement and temperature, and these fields satisfy the
balance equations of momentum (8.8)1 and of energy (27.9).
There are, however, radical differences in the constitutive equations.
To put in evidence in detail those differences allowed essentially
by the presence, or absence, of a microstructure, albeit latent, a
comparison of the class discussed in Section 27 with the class of
hyperelastic bodies is in order. For the task in hand, significant
properties of media in the latter class (properties which can be read
also off (27.5) when terms involving ro are everywhere cancelled) are
the following ones:
(i) the free energy '" is independent of Grad F and grad e,
(ii) those two gradients do not enter the constitutive equations for P
 §28. Comparison with the Traditional Class of Hyperelastic Bodies 75

and '1 because

• 8~ (28.1)
'1 = - 80 .

Thus, within the classic context no constitutive equation can be

proposed for a hyperelastic body involving the second gradient of F,
as occurs for the so called elastic continua of second grade or for
perfect fluids of Korteweg (where grad I is involved and hence, in an
appropriate combination, F and Grad F).
Continua of second grade become acceptable either if they are
thought of as continua with latent microstructure or, equivalently, at
the cost of some modification in the balance equations; for instance,
with the introduction, in the balance of energy, of an extra term,
called the interstitial working. Notice, in fact, that, from a purely
formal point of view, the microstructural actions represented by ~ in
(25.2) could be eliminated by use of (8.3) and the balance equation for
energy could be written again as follows
p(e + K)" = T· grad x + div(9'T V + q) + p(A + ~. v), (28.2)
an equation which appears to be very near the classical one; to bring
the two into coincidence one need only modify the definition of e, q
and A by the addition of microstructural terms. In particular when, as
done tacitly by some authors, one disregards all microstructural mass
densities (putting X, or K, and ~ equal to zero) and supposes besides
that d = 0, then (8.5) reduces to
~ = div 9' (28.3)
and can be interpreted as an equation for the evolution of internal
parameters (or even as a relation which defines v explicitly in terms of
F and 0), the balance of moment of momentum is ensured by the
symmetry of T, the Cauchy equation is formally not altered, whereas
the equation of balance for energy contains, all reductive assumptions
not withstanding, an additional term, div (9' T v), which is then in-
terpreted as due to the interstitial working.
If one does not want to disregard microstructural body densities,
so that, for instance, in the conservative case studied in the previous
section, one needs to accept for P the complete expression (27.7), then,
always from a formal point of view, also the equation of balance of
momentum must be modified by the addition of appropriate terms.
Precisely, with the introduction of the following three tensors
76 III. Thermodynamics

(i) a tensor of partial stress

_ o~
P := p* of -
.( o~)
DIV p* o(Grad F) , (28.4)

(ji) a tensor of micromomentum flux

MiH := P* : ; : ( (::~)" - :~~) ~=~, (28.5)

(iii) a flux of external actions tensor

oro~

NiH := P* OFiH Pa' (28.6)

the Cauchy equation (always in the case studied in the preceding

section) can be given the following form
p*x - Div M = p*f - Div N + Div P, (28.7)
so that fluxes of different origin are clearly in evidence, in particular
those of inertial origin are put in the left-hand side; in a sense each
subbody behaves as a body with variable mass with an appropriate
surface momentum flux. In the special case when ro depends on F
only through l, (28.7) is reduced to (14.17).
 PART IV
Mathematical Problems Posed by the Theory

§29. The Influence of the Topological Properties of the

Manifold Jt
The characteristic features of all special problems in the theory of
continua with microstructure stem from the basic fact that knowledge
of a complete placement of a body !8 requires the assignment of a
field v(x) whose values, in general, are not in a Euclidean space or in
a linear space but in a smooth manifold, with sometimes complex
topological properties; this fact by itself creates new types of mathe-
matical problems. Other interesting questions derive from the higher
order of the differential systems which describe the evolution of !8 or
from higher-order derivatives appearing in the balance equation of
momentum itself, when bodies with latent microstructure are envis-
aged. Numerous other problems are generated by the wealth of possi-
bilities arising from constitutive equations which involve more vari-
ables than in the classical case. Here, in Part IV, we quote some
special examples, beginning with some elementary aspects of the
topological theory of defects.
For simplicity we intend to consider below only the field v(x), on
fJI; v(x) is either an instantaneous value or, perhaps, the value in an
equilibrium placement; in fact, we will not worry at all as to how the
field was generated: we are only interested in its geometric properties.
Actually we will even take fJI to be the whole space g and suppose the
field continuous on g with the exclusion of isolated points or isolated
78 IV. Mathematical Problems Posed by the Theory

regular curves and surfaces; these singular sets we will call dejects of
the field. Now, the significant fact emerges that the physically most
interesting set of defects is that for which a classification is possible
by the methods of algebraic topology; it is the set of defects which
cannot be eliminated by 'local surgery' and the existence of which
can be determined by a study of the field in a neighbourhood of the
singularity.
To be specific we will refer to the case of a line deject: we suppose
that v(x) be regular except for the points of a line t of 8. We explore
the defect by examining the behavior of v along a loop c of 8 which
"surrounds" t; that is, which is the edge of a piece of surface that
intersects t. We consider the loop y which is the image of c on vii
through the mapping x -+ v(x), x E c and the family (y) of loops
obtained in vii while we deform c continuously, but otherwise arbi-
trarily, with only one condition, that of avoiding t. (y) is a subset of
the family {y} of all loops belonging to vii, which can be obtained
from y by free homotopy, that is, by a continuous deformation of y
on vii.
The type of singularity associated with the line t depends on the
properties of the family {y}, in particular, for instance, on whether {y}
contains or not a degenerate loop reduced to a point; the properties
of {y} can be studied in general with topological methods. A couple of
elementary examples may serve to introduce the topic.
Consider the case of liquid crystals, a case where each v is a
direction. Consider a plane field of directions on 8, with a singular
straight line t perpendicular to the plane; call e the angle in [0, n)
formed locally by the preferred direction with a fixed straight line in
the plane and suppose that, as the point x moves around a circle c of
the plane, the angle e increases continuously from 0 to n. On vii
(a half sphere of unit radius, on whose boundary circle antipodal
points are identified) the corresponding loop y is a half meridian
circle which is closed by the identification of equatorial antipodes. If
the circle c is continuously deformed on the plane and is shrunk
towards the trace of t on the plane (but without ever crossing
that trace), the corresponding loop on vii might change but it will
be impossible to wrench it away so as to avoid the "exchange"
of antipodal points. A continuous change of the original field of
directions in the neighborhood of t, even abandoning the condition of
planarity, will not change matters substantially. The result is bound
with the topological properties of vii, in particular with the existence
of "irreducible" loops on vii.
 §29. The Influence of the Topological Properties of the Manifold.l{ 79

In an alternative example suppose that, other things being equal, it

is observed that, on moving x once around c, e increases from 0 to
2n; then the corresponding loop y on .A consists of two superposed
half meridian circles described successively in opposite directions; this
new y is" reducible to a point on .A by continuous deformation.
Correspondingly, by a continuous nonplane change of the direction
field in the neighborhood of t all directions could be brought to
coincide.
Consider also the case of ordinary spins (example (v) of Section 2)
and a plane field of unit vectors; the angle e may now take any value
in [0,2n). Let us suppose again that there is a singular straight line t
normal to the plane, c being a closed curve encircling the trace of t
on the plane. Suppose further that, when the poi!!t x moves once
around c, e increases continuously from 0 to 2n (notice that now e
can change only by a multiple of 2n); the corresponding loop y on .A
(itself a unit sphere) is a meridian. When c is shrunk against the trace
of t on the plane, the corresponding loop y on .A might not change at
all if the field of unit vectors on S were appropriately defined, so that
t would be really a singular line. But the singularity now has a
different character; in particular it could be eliminated by a con-
tinuous change of the field in the neighbourhood of t, such as to
make all unit vectors form a locally uniform vector field parallel to t.
This circumstance reflects the property that any loop on the unit
sphere can be shrunk to a point by continuous deformation.
In short, and without any pretence of precision or completeness,
the topological questions involved in the classification of singularities
are as follows.
One begins with the observation that one can introduce a group
structure on the set of classes of homotopic loops of .A. The structure
is first built up for classes of homotopic loops all passing through a
fixed point of .A (or, as is usually said, all based on that point),
employing the following definitions: (1) if y(l) and y<2) are two loops of
.A both going through a point v, the product yO) 0 y( 2 ) is the loop that
is obtained by describing, starting from v, first y(l) and then y(2); (2) if
y is a loop, y-l is the same loop described from v in the opposite
sense; (3) a cycle that reduces to the point v is the unit loop 8.
One proceeds, then, to define a product for classes of homotopic
loops based on v through the following rules:
{y(l)} 0 {y(2)} := {y(l) 0 y(2)},

{y}-l:= {y-l},
80 IV. Mathematical Problems Posed by the Theory

and to tl'e remark that these properties hold:

(i) the associative property
({y(1)} 0 {y(2)}) 0 {y(3)} = {y(l)} 0 ({y(2)} 0 {y(3)}),
(ii) the property of the unit element
{y} 0 {e} = {y},
(iii) the property of the inverse loop
{y}-l 0 {y} = {e}.
Thus we have constructed on J( a homotopy group based on v:
1tl(~ v).
Given two groups, one based on v(1) and the other based on V(2),
one can establish an isomorphism between 1tl(~ vll) and 1tl(~ vl'2)
as follows. One chooses a path -r on J( joining v(1) and V(2), denotes
by -r- l the same path but followed in an opposite sense and one
associates with any loop y based on V(2) the loop based on v(1) which,
with an obvious interpretation of the notation, can be described
by -ry-r- l . One proceeds to verify that, if y and yare two homotopic
loops based on V(2), then -ry-r- l and -ry-r- l are two homotopic loops
based on vel). Thus one can define a mapping -r({y}) from the class
{y} based on V(2) on the class {-ry-r- l } based on v(1); it is a mapping
which preserves the algebraic multiplication structure between classes.
Hence an isomorphism is defined between the groups 1tl (~ v(1»
and 1tl (~V(2» which in general depends on the choice of the
path -r.
As a consequence one can think of a single abstract group 1tl (Jt)
of which each group based on a particular element v is an isomorphic
copy.
One verifies that 1tl(Jt) is abelian if and only if the mapping -r({y})
of 1tl(~ V(2» on 1tl(~ vel»~ is independent of -r. In fact, if 1tl(..H) is
abelian so is 1tl (~ V(2»; if -r(l), -r(2) are two arbitrary paths joining vel)
to V(2) and y is any loop based on V(2), then
{-r(1)y-r(l)-'} = {(-r(2)-r(2)-')-r(1)y-r(1)-'}.
But -r(2)-1-r(1) is a loop based on V(2) and, because 1tl (~V(2» is
abelian,
{-r(2)(-r( W1 -r(1»y-r(1)-1} = {(-r(2)y(-r(W'-r(1»-r(l)-'} = {-r(2)y-r(2)-'};
thus one concludes that the classes {-r(1)y-r(1)-l} and {-r(2)y-r(2)-1}
coincide.
 §30. Further Remarks on the Topological Theory of Defects 81

Vice versa, let y and b be two arbitrary loops based on v(1) and let
a be a path joining a point v of b with V(2) and b(1), b(2) the two paths

into which b is cut by v. Then

{y} 0 {b} = {yb} = {y(b(1)b(2))} = {y(b(1)aa- I b(Z)}
= {( (b(1) a)(b(1) ar i )y«b(1) a)(a- 1 b(Z»))}.
Let us put ,(1) = b(1)a and ,(Z) = b(2)-l a and suppose that the
isomorphism between n1(~ v(1») and n1(~ VIZ») be independent of
the path used to join v(1) with vIZ); then
{,(1)-l y,(1)} = {,(Z)-l y,(Z)}
and so
{y} 0 {b} = {(b(1)a)(a- 1 b(2»)y(a- 1 b(2»-1 (a- i b(Z»)}
= {by} = {b} 0 {y}.
We conclude that, if the fundamental group is abelian, it is possible
to establish a one-to-one correspondence between classes of freely
homotopic loops and the elements of the group; finally the classifica-
tion of defects is linked to the list of elements of the group.

Examples. Planar spins: the fundamental group is abelian and

isomorphic with the group of additive integers; the homotopy classes
are specified by the winding number.
Ordinary spins: the fundamental group contains only the identity_

§30. Further Remarks on the Topological Theory

of Defects

The types of microstructure for which the fundamental group of At is

abelian are the simplest types; for them recourse to deep results of
algebraic topology is unnecessary. Thus one must attack the non-
abelian case. In that case the isomorphism of the two groups, one
based on v(1), the other based on v(2), may depend on the choice of
the path used to join V(I) and V(2). One notices, however, that the two
isomorphisms differ at most by an automorphism; indeed, let ,(1) and
,(2) be two paths joining v(1) with V(2) and let y be a loop based on
V(2); the following relations apply

,(1)y(,(1»)-l = ,(1)(,(2»)-1 ,(Z)y(,(Z)r 1 ,(2)(,(1))-1 = a,(2)y(,(Z)r 1 a-I,

82 IV. Mathematical Problems Posed by the Theory

where
a = 1'(1)(,[(2»)-1

is the loop based on V(l) which is used to establish the automorphism

{y} ~ {a} 0 {Y} 0 {a- 1 }
on n 1 (./1(, v(1»).
Two loops y, g based on v(l) are said to be conjugate if there is a
loop a such that
g = aya- 1
and the set of loops obtained from y through the operation aya- 1 by
changing a continuously is said to be a conjugate class of y.
It is not too difficult to prove that n 1 (J{, v(1») can be partitioned
into distinct conjugate classes; it is also possible to state a one-to-one
correspondence between the classes of freely homotopic loops on At
and the conjugate classes of n 1(Jt).
This result opens the road to application of penetrating results of
algebraic topology to the classification of line defects in cases (such as
the case of the dipole-free A phase of He 3 ), where inti uti on scarcely
helps.
We quote finally the fact that similar developments and the intro-
duction of the so-called second homotopy group n2(Jt) serve in the
classification of point defects.

§31. Existence of Singular Solutions in Statics

In the two previous sections a classification of some singularities

which can occur in a body with microstructure was pursued, ex-
ploiting properties of the manifold At. The effective existence (for
instance in statics under appropriate loads or even under no external
actions) of solutions of the balance equations which may represent
fields of the types mentioned offers, in general, a very difficult question;
for instance, one cannot exclude cross effects between microstresses
and macrostresses, microdeformations and macrodeformations.
To reduce the complexity of problems, some authors, encouraged
by experimental observations, have begun to explore situations where
the macroscopic deformation vanishes, or has no influence on micro-
stress, or is otherwise irrelevant, whereas the field v(x) is nontrivial.
We refer below for definiteness to the case of a perfect incompressible
 §31. Existence of Singular Solutions in Statics 83

fluid with microstructure under static circumstances. Then one must

study (12.6), where !7 and ~ are given by (12.9) whereas for T the
expression (12.8) must be taken with a reaction pressure -fDl re-
placing the first addendum in the right-hand side. It is easy to recog-
nize that, jn the absence of body actions (f = 0, p = 0), the Cauchy
equation requires
a~ a~
grad fD + t(grad 2 v)p a(grad v) + (grad V)T p av = 0

or
grad(fD + ~) = 0
and hence the coincidence of fD with -~, an additive constant apart.
At the same time the field v must satisfy the equation

.( a~ ) a~
dlV p a(grad v) - p av = 0, (31.1)

As we have repeatedly remarked, the boundary conditions to be

associated with (31.1) may be complex, in general, but here we fix
attention on the case when f!I is the whole 8, in line with our analysis
in the last two Sections. Then periodicity conditions, or conditions at
infinity may be important.
The difficulties that must be met to find solutions of (31.1) may be
appreciated already in a special case. Suppose that
(31.2)

where the coefficients Ji",p, to insure the properties of invariance ex-

pected of ~, must satisfy the conditions (see (5.9))

..A'"
- i
aJipa p a
av'" V ,jV ,j
+ (-..A'"
i,k -
"') P
eikmV ,m Jip",V,k -
- 0•

Then (31.1) can be written

Y /j (aJiPY 1 aJi/jY) _ 0
Ji V ,i V ,i av/j -"2 av P -
+ -l",p
'" (31.3)
V ,ii ,

where the quantities "lap have been introduced with the property
-Jil,.YJi
YP = bIZp'
84 IV. Mathematical Problems Posed by the Theory

The strong nonlinearity present, in general, in (31.3) is the main

obstacle in the search for solutions. A degenerate case of limpid
simplicity occurs when the coefficients of the quadratic terms in
grad v in (31.3) vanish. Then our problem is to fmd appropriate
harmonic functions defined in 8 and having values in Ji; here many
results from harmonic analysis are at hand.
A still more special case was mentioned already at the end of
Section 20 with reference to nematic liquid crystals. Let us recall that
case, which offers opportunity for an explicit example. In (20.8), (20.9)
we take P and K1 constant and d parallel to a plane of the coordinate
axes '1; '2:
d == (cos (), sin (),O),
Then the function () must itself be harmonic and the Cauchy stress
tensor is given by
T11 = !P K1(():2 - ():d, T22 = !pKt(():t - ():2)'
(31.4)
T33 =!P K1(():1 + ():2)' T12 = T21 = -P K t O,1(),2'
A line defect where t is a straight line normal to the (,?, ,g) plane
through the point (,?, ,g, 0) is represented by Frank's solution, that is,
by the harmonic function

() =
v
2" arctan
("2'1 -- ,g)
,? ' (31.5)

where the integer v specifies the winding number (and hence the
homotopy class). Note that even this simple defect generates a rather
complex state of macroscopic stress.

§32. Phase Transitions

We have already mentioned that one basic reason for the wealth and
complexity of situations which may arise in continua with micro-
structure is the possible presence, in the balance equations, of higher-
order derivatives.
As a sample in this direction we explore a static phenomenon of
simple compression in a Korteweg fluid, hence taking for the Cauchy
stress the expression (18.6) and assuming that the measure of volume
compression 1 be a function of only one space variable ,. For the
potential ({) we make also a special choice, as follows:
 §32. Phase Transitions 85

qJ{l, grad l) = -a{l) + tp(l) Igrad zl2

so that the stress is given more explicitly by

T =- p* ( da
dl - (P1- 2.1 dl
dp) Igrad II 2
+ PAl ) 1- pp grad 1® grad I.

In the one-dimensional case we want to consider, the component

1{~ is supposed to be equal to a negative constant -t1J different from
zero; all shear components are null. Thus the following equality
applies:

(32.1)

The response of a classic fluid is included in (32.1) if we put 13 == 0

in it and take a as a concave function of land da/dl as strictly
decreasing from +00 to zero as 1 increases from zero. Then, (32.1) can
be solved uniquely in terms of l; the solution is independent of and e
decreasing with t1J (remember that fD is a pressure).
Going back now to the general case, we observe that (32.1) can be
written in the form

( fD 1 _ a - t{1 .~)2 13) = o. (32.2)

p* .~
If the hypotheses of the standard case are maintained for a, and
furthermore one assumes that p(l) be negative, then this equation
admits not only the classic constant solution but also an interesting
set of periodic solutions. Precisely, let us call 'm the unique solution of
the equation in 1

y a constant with the property

t1J
y> -a{lm) + -lm'
p*
while l1 and l2 (II < lm < 12) are the two roots of the other equation
in 1
fD
a{l) = - l - y.
p*
86 IV. Mathematical Problems Posed by the Theory

The solutions in question are periodic functions with period equal

1
to

2
'2 ( - p
*
13(/) )112 dl
I, 2(p*0:(/) + yp* - tin) ,
which can be obtained by successive inversion of one and the other of
the equations

e= +
-
1 I,
'2 (
*
- P
2(P*0:(1) + yp*
13(/)
- Till)
)112 dl. (32.3)

Even more interesting situations arise when the function of 0:(1) has
the more complex properties presumed in certain studies of phase
transition: do:jdl strictly decreasing up to I = Ib' then strictly increasing
up to 1 = Id and finally strictly decreasing £0 zero.
Given those properties it is possible to choose in a unique manner
la and Ie (Ia < Ib < Id < Ie) by the "Maxwell rule", that is, so that the
two following properties are satisfied:

and to imagine W chosen so that

-dO:)
- * ( dl
w-p
1=1 0

(actually a third value Ie of I exists where do:jdl takes the value wjp*)
and the constant y such that the equation in I
W
0:(/) - -I +y= 0
p*
has the two solutions 1 = la' and I = Ie.
Suppose now that 13(/) be strictly positive. Then the function I(e)
obtained by inversion of the relation

e- I I (

Ie 2(p*0:(/)
- P 13(1)
*
+ yp* - WI)
)1/2 dl (32.4)

is a solution of (32.1). It is a function again defined over (-00, +(0)

always increasing from the value la to Ie; actually I(e) increases very
slowly at first, and then, in a neighbourhood of the origin, grows
very rapidly from values still near la to values near Ie and tends
e
asymptotically to this value as goes to +00. The solution represents
 §33. Droplets of Perfect Liquids with Microstructure 87

a kink, a sudden change from the phase with 1 = la to the phase with
1 = leo A similar kink, but decreasing, is obtained by choosing the
negative sign for the root in (32.4).

§33. Droplets of Perfect Liquids with Microstructure

In the general part of this book (Sections 11 and 12) we have men-
tioned certain variational properties as providing a tool to ascer-
tain the form of balance equations in statics. One can invoke those
properties also for the solution of special problems, always in
statics, through recourse to the direct methods of. the calculus of
variations.
In this Section we mention a class of isoperimetric problems that
arise from the variational formulation (under rather special circum-
stances) of the problem of determining the form taken by a droplet of
a perfect incompressible fluid with microstructure, not subject to body
actions and immersed in an ambiance at constant pressure.
We refer to the developments of Section 12 and use the same
notation. We seek the shape of the domain fJI that renders the
functional

t&"[fJI, v] = r
J@
PqJ + f
8@
y,

a minimum; the potential qJ is assumed to be a given function of v

and grad v; the potential y a function of v and of the unit vector n of
the external normal to afJI. Furthermore fJI has a given fixed volume
v; hence the constraint applies

vol fJI = v, (33.1 )

with va preassigned positive constant.
The constraint could be absorbed, through recourse to a modified
functional
t&"*[fJI, v] = t&"[fJI, v] + ro vol(fJI),
where a Lagrange multiplier ro appears, which has here the meaning
of a pressure. Naturally, if the problem C* = min has to make sense,
appropriate qualitative and quantitative properties must be required
of the potentials qJ and y; the class of admissible domains fJI and that
of the admissible fields v on each fJI must also be chosen.
88 IV. Mathematical Problems Posed by the Theory

In any case, with the well-known formal developments mentioned

in Section 12, Euler's equations for the problem
8*[~, v] = min (33.2)
can be deduced; they are (12.6), (12.11) with the meaning of symbols
specified in (12.8), (12.9) provided that in (12.8) the first addendum in
the right-hand side be replaced by - mI, with m the reactive pressure,
which compensates for the incompressibility constraint.
Both the direct problem (33.2) and the boundary-value problem
(12.6), (12.11) are of great complexity. Therefore, it has been proposed
to consider a simplified problem, obtained from (33.2) through a sort
of singular perturbation.
To begin with, some assumptions must be accepted for qJ:
(i) qJ is nonnegative and vanishes if and only if grad v = 0;
(ii) in addition,
qJ(V, (X grad v) = (X2qJ(V, grad v), (33.3)
for all constants (X.
A potential which satisfies these properties is, for instance, that
given by (31.2) if the coefficients llafJ form a positive-definite matrix.
Then one introduces the change of variables y = v- 1[3x, which
transforms any region ~v of volume v into a region ~1 of unit
volume. One defines finally the field v(y) = v(v 1/3 y) and the functional
8[~1' v] = 8[~v' v].
By (33.3) one gets

8[~1' v] = v2/3 [ ~3
v J~l
r qJ(v, grad v) + Ja~.
r y(v, n)].
The minimum of 8 is achieved where also the functional

~[~1> v] = -:73
v J~l
r qJ(v, grad v) + Ja~l
r y(v, n)
is a minimum.
Let us introduce now the hypothesis that the drop be small; that is,
the parameter v be small in (33.1). For v -+ 0+, ~ reaches a minimum
only if

r qJ(v, grad v) = 0,
J~l
that is, only if v is constant, say v= Vo on ~1. Hence, in a rough
 §33. Droplets of Perfect Liquids with Microstructure 89

approximation and a change of scale apart, the equilibrium form of a

droplet is such as to satisfy the following variational problem: find the
domain of unit volume which renders the functional fo~Y(vo, n) a
minimum.
Theorems are available which assure the existence of this minimum
and also provide an explicit rule to find it. The minimizing set is the
set of unit volume which is homothetic to what is called Wulff's set
'1Y, which is specified by the rule
11/"= {x E tflx· n ~ y(v o , n), for all unit vectors n}.
To obtain 11/" one can proceed as follows: fix a point 0 in space,
draw from 0 in any direction of unit vector n a segment of length
y(v o, n); send through the end of the segment a plane orthogonal to
n, thus defining the half space, which contains o. Wulff's set is the
intersection of all half-spaces thus determined when n is arbitrarily
varied.
 Appendix

The theory of continua with microstructure has a long history,

although the contributions in the past have been few. Love in his
treatise cites Poisson, who "proposed to regard the molecules ...
as little rigid bodies capable of rotation as well as translatory
displacement" (Mem. Acad., 18 (1842». The suggestion of Poisson was
worked out by W. Voigt (Abh. Ges. Wiss. Gottingen, 34 (1887»: the
question was discussed also by P. Duhem in the second chapter of his
memoir on the thermodynamic potential (Ann. Ecole Norm. Sup. 10
(1893), 187-230). A complete and satisfactory tract on the theory of
continua with rigid microstructure is due to the brothers Eugene and
Fran~ois Cosserat (Sur la tbeorie des corps deformables, Paris, 1909).
Their work was generally forgotten for fifty years; the only inter-
mediate contribution based on it is that of Sudria (Mem. Sci. Paris, 29
(1935».
Clifford Truesdell in the second of his Six Lectures on Modem
Natural Philosophy (Springer, 1966), a lecture which is devoted to
polar and oriented media, expresses regret that, in early stages of the
revival (due in the USA to J. L. Ericksen and R. A. Toupin, to Aero
and Kuvshiskii in the URSS and to Grioli in Italy), the contribution
of the Cosserats was ignored. However, a contemporary paper of
W. Gunther (Abh. Braunschw. Wiss. Ges. 10 (1958), 195-213) refers to
the Cosserat in the title.
In the 1960s so many new results were published that the IUTAM
found it appropriate to organize a Symposium, which was held
 Appendix 91

in Stuttgart and Freudenstadt in 1967. The proceedings of the

Symposium (Mechanics of Generalized Continua, E. Kroner (ed.),
Springer, 1968) provide an updated overview of the topic. The
prevailing interest is in continua with affine microstructure and multi-
polar continua, but there are also contributions, for instance, to the
theory of continua with continuous distribution of dislocations.
In the following year there was a meeting on the theory of polar
continua at the Istituto di Alta Matematica in Rome; the contributions
appear in the second part of the first volume of the series Symposia
Mathematica (Academic Press, 1969). Also the International Centre
for Mechanical Sciences in Udine organized a Course on Polar
Continua in 1969; the lectures of R. Stojanovic (Mechanics of Polar
Continua, CISM Lectures 2, (1969)), mainly devoted ..to the theory of
continua with affine microstructure, contain also an ample list of
references and report on contributions in slavonic languages otherwise
not so easily accessible.
The literature in the field has become later so ample that I have
abandoned the idea of giving specific references, though I find it
impossible not to cite explicitly at least the work of Green and Rivlin;
many important papers of theirs can be found in the volumes of the
Archive for Rational Mechanics and Analysis. The later progress in
the field is reflected, for instance, in the Proceedings of the Symposia
on Continuum Models of Discrete Structures held in Jodlowy and Dwor
(Poland) in 1975, Mount Gabriel (Canada) in 1977, Freudenstadt
(West Germany) in 1979, Stockholm in 1981 and Nottingham in 1985.
Studies on continua with microstructure find often concrete
applications when couched within a much wider context than that
envisaged in these lectures, that is, the context where also electro-
mechanical and magnetomechanical actions are envisaged. This fact
was already in the minds of early Authors; for the sixties the relevant
reference is a paper by R. Toupin (Arch. Rat. Mech. Anal. S (1960),
440-452). An ample memoir by G. A. Maugin (Acta Mech. 3S (1980),
1-70) must also be cited and there are number of more recent con-
tributions by the latter Author.
The compilation of a fair list of contributions to special theories, in
particular to the theories of continua with voids and of liquid crystals,
would be an immense task. Obvious quotations are the books by
P. G. De Gennes (The Physics of Liquid Crystals, Clarendon Press,
1974) and S. Chandrasekhar (Liquid Crystals, Cambridge University
Press, 1977), the series of volumes on the Advances in Liquid Crystals
published by Academic Press, review papers by L. van Wijngaarden
92 Appendix

and A. Prosperetti on liquids with bubbles (e.g., Ann. Rev. Fluid

Mech. 9 (1977), 145-185), papers by S. C. Cowin, J. L. Ericksen,
A. C. Eringen, M. A. Goodman, G. Grioli, J. T. Jenkins, F. M. Leslie,
J. W. Nunziato, S. L. Passman and others (many of which have
appeared in the Archive for Rational Mechanics and Analysis).
For the background to the development in Part IlIon thermo-
dynamics the reference is the second edition of C. Truesdell's book on
Rational Thermodynamics (Springer, 1984). For the topological theory
of defects an excellent review article is N. D. Mermin's in Reviews of
Modern Physics 51 (1979), 591-648.
As I have mentioned in the Preface the point of view taken in these
lectures is that originally proposed in a paper written in collaboration
with P. Podio-Guidugli (Ann. Mat. Pura Appl. 135 (1983), 1-25).
Vol. 20 Edelen/Wilson: Relativity and the Question of Discretiztion in Astronomy
With 34 figures. XII, 186186 pages. 1970.

Vol. 21 ¥cBride: Obtaining Generating Functions

XIII, 100 pages. 1971.

Vol. 22 Day: The Thermodynamics of Simple Materials with Fading Memory

With 8 figures. X, 134 pages. 1972.

Vol. 23 Stetter: Analysis of Discretization Methods for Ordinary Differential Equations

With 12 figures. XVI, 388 pages. 1973.

Vol. 24 Strieder/ Aris: Variational Methods Applied to Problems of Diffusion and Reaction
With 12 figures, IX, 109 pages. 1973.

Vol. 25 Bohl: Momotonie: Losbarkeit und Numerik bei Operatorgleichungen

Mit 9 Abbildungen. IX, 255 Seiten. 1974.

Vol. 26 Romanov: Integral Geometry and Inverse Problems for Hyperbolic Equations
With 21 figures. VI, 152 pages. 1974.

Vol. 27 Joseph: Stability of Fluid Motions I

With 57 figures. XIII, 282 pages. 1976.

Vol. 28 Joseph: Stability of Fluid Motions II

With 39 figures. XIV, 274 pages. 1976.

Vol. 29 Bressan: Relativistic Theories of Materials

XIV, 290 pages. 1978.

Vol. 30 Day: Heat Conduction within Linear Thermoelasticity

VII, 82 pages. 1985.

Vol. 31 Valent: Boundary Value Problems of Finite Elasticity

XIII, 191 pages. 1988.

Vol. 32 Day: A Commentary on Thermodynamics

IX, 96 pages. 1988.

Vol.33 CohenjMuncaster: The Theory of Pseudo-rigid Bodies

X, 180 pages. 1988.

Vol. 34 Angeles: Rational Kinematics

XII, 190 pages. 1989.

Vol. 35 Capriz: Continua with Microstructure

X, 192 pages. 1989.