INTERNATIONAL CENTRE FOR MECHANICAL SCIENCES

COURSES AND LECTURES - No. 292

APPLICATION S OF TENSOR FUNCTIONS

IN SOLID MECHANICS

EDITED BY

J.P. BOEHLER
UNIVERSITY OF GRENOBLE

SPRINGER-VERLAG WIEN GMBH

Le spese di stampa di questo volume sono in parte coperte da contributi
del Consiglio Nazionale delle Ricerche.

This volume contains 36 illustrations.

This work is subject to copyright.

All rights are reserved,
whether the whole or part of the material is concerned
specifically those of translation, reprinting, re-use of illustrations,
broadcasting, reproduction by photocopying machine
or similar means, and storage in data banks.
© 1987 by Springer-Verlag Wien
Originally published by Springer Verlag Wien-New York in 1987

ISBN 978-3-211-81975-3 ISBN 978-3-7091-2810-7 (eBook)

DOI 10.1007/978-3-7091-2810-7
 PREFACE

The mechanical behavior of materials with oriented internal struc-

tures, produced by formation processes and manufacturing procedures
(crystal arrangements, stratification, fibrosity, porosity, etc.) or in-
duced by permanent deformation (anisotropic hardening, softening, cree~

internal darnage growth, etc.) requires a suitable mathematical modelllng.

The properties of tensor valued functions of tensor variables constltute
a rational basis for a consistent mathematical modelling of complex mate-
rial behavior.
This book, which contains lectures presented at a CISM Advanced
School, presents the principles, methods and results of applications in
solid mechanics of the general laws governing tensor functions. The prin-
ciples of mathematical techniques employed to derive representations of
tensor functions are explained. The rules of specifying irreducible sets
of tensor invariants and tensor generators for various classes of mate-
rial symmetries are discussed. Representations of isotropic and anisotro-
pic tensor functions arederived, in order to develop the general inva-
riant forms of non-linear constitutive laws in mechanics of solids.
Within this approach, the mathematical modellization of the mate-
rials' mechanical response is explained and specific models are presented
in elasticity, plasticity, hardening, internal darnage and failure, for
materials such as metals, composites, stratified rocks, consolidated soils
and granular materials. The approach specifies a rational way to develop
approximate theories and gives the necessary precision as to the number
and the type of independent variables entering the mechanical laws to be
used in engineering applications.
 Experimental justifications as to the pertinence of the approach are
given on examples of composite materials, rolled sheet-steel and strati-
fied rocks. Information concerning proper experimental setting of tests
for materials with oriented internal structures is developed.
This book is addressed to specialists in solid mechanics, both theo-
retical and applied, material scientists concerned with metals and compo-
sites, specialists in soil and rock mechanics and to structural engineers
facing problems involving anisotropic and inelastic solids at various
environments, nonlinearity and couplings.
I wish here to pay homage to the memory of my deeply missed friend
Professor Antoni Sawczuk, co-coordinator of the CISM Advanced School. His
untimely death did not allow him to see the final fulfilment of our sha-
red project ; nevertheless, his help was invaluable in the preparatory
phase of the Advanced School.
I would like to take this opportunity to thank CISM for having pro-
vided lecturers and participants with a chance to work tagether on Appli-
cations of Tensor Functions in Solid Mechanics. I am indebted to Profes-
sor Giovanni Bianchi, Secretary General of CISM, for his help with the
organization of the School and to Miss Elsa Venir for her kindness and
efficiency.

Jean-Paul Boehler
 CONTENTS

ehapter 1 PHYSieAL MOTIVATION

by 1. P. Boehler

1. Introduetion
2. Different domains of meehaniea1 anisotropy =
3. Essential features of the anisotropie meehanieal
behavior of rolled sheet-steel 6

ehapter 2 INTRODUeTION TO THE INVARIANT FORMULATION OF

ANISOTROPie eONSTITUTIVE EQUATIONS
b_v 1. P. Hoehler

1. Introduetion 13
2. Prineiple of Isotropy of Spaee 14
3. Isotropie materials 16
4. Anisotropie materials 18
5. Orthotropie materials 21
6. Representation of the funetion F 23
7. eonclusions 29

ehapter 3 REPRESENTATIONS FüR ISOTROPie AND ANISOTROPie

NON-POLYNOMIAL TENSOR FUNCTIONS
by 1. P. Boehler

1. Introduction 31
2. Representations for isotropic scalar and tensor
funetions 35
 3. Representations for non-polynomial anisotropic
scalar and tensor functions
3.1. Method based on the introduction of
structural tensors 40
3.2. Generalization of the Rivlin-Ericksen
method 41
3.3. General anisotropy 42
3.4. Orthotropy 44
3.5. Transverse isotropy 45
3.6. eomparison with representations for
polynomial anisotropic tensor functions 47
4. Representations for non-polynomial isotropic and
orthotropic tensor functions in a two-dimensional
space 49

ehapter 4 ANISOTROPie LINEAR ELASTieiTY

by J. P. Boehler
1. Introduction 55
2. Transverse isotropy invariant and classical
formulations 57
3. Orthotropy invariant and classical formula-
tions 60
4. Isotropy 64

ehapter 5 YIELDING AND FAlLURE OF TRANSVERSELY ISOTROPie

SOLIDS
by J. P. Boehler

1. Introduction 67
2. General theory 68
3. Plastic deformations in uniaxial and triaxial
tests
4. Failure criteria for glass/epoxy composites
under confining pressure
4.1. Introduction 80
4.2. General form of the yield criterion for
triaxial tests 80
 4.3. Failure modes and directional strengths 81
4.4. Proposed failure criteria 84
4.5. Comparison with experimental results 87
5. Simplified theory
5.1. Plastic behavior 89
5.2. Yield eriteria 93

Chapter 6 ON A RATIONAL FORMULATION OF ISOTROPIC AND

ANISOTROPie HARDENING
by J. P. Boehler
1. Introduetion 99
2. Classical formulation of isotropic, kinematic
and anisotropie hardening
2.1. Isotropie hardening 101
2.2. Kinematie hardening 102
2.3. Anisotropie hardening 104
2.4. Hardening phenomena whieh eannot be
deseribed by the elassical formulations 105
2.5. Conclusions 107
3. General formu1ation of isotropie and anisotro-
pie hardening
3.1. Proposed general eoneept 108
3.2. Inf1uenee of the p1astie strain on the
hardening ru1e 108
3.3. Isotropie hardening 109
3.4. Anisotropie hardening 112
4. Examples
4.1. Introduetion 113
4.2. Proposed genera1 isotropic hardening rule 113
4.3. Proposed general anisotropie hardening
ru1e 116
5. Conelusions 120
ehapter 7 ANISOTROPie HARDENING OF ROLLED SHEET-STEEL
by J. P. Boehler
1. Introduetion 123
2. eonstitutive relation
2.1. General invariant form of the eonstitutive
relation 124
2.2. Anisotropie hardening 126
3. Plastie behavior
3.1. General invariant forms of the flow law
and the yield eriterion 130
3.2. Proposed eriterion 131
4. Experimental behavior of rolled sheet-steel
4.1. Experimental proeedure 132
4.2. Experimental results and eomparison with
theoretieal predietions 133

ehapter 8 ISOTROPie POLYNOMIAL INVARIANTS AND TENSOR

FUNeTIONS
by A. J. M. Spencer
1. Introduction ; notations and definitions 141
2. Results from classieal theory 145
3. Orthogonal transformation groups 146
4. Integrity bases for veetors 149
5. Isotropie tensors 150
6. Isotropie invariants of veetors and seeond
order tensors - General form 151
7. Traees of matrix produets and matrix
polynomials 153
8. Invariants of symmetrie seeond-order tensors 157
9. Invariants of seeond-order tensors and veetors
proper orthogonal group 159
10. Invariants of seeond-order tensors and veetors
full orthogonal group 162
11. Isotropie tensor po1ynomial funetions of
veetors and tensors 164
ehapter 9 ANISOTROPie INVARIANTS AND ADDITIONAL RESULTS
FOR INVARIANT AND TENSOR REPRESENTATIONS
by A. 1. M. Spencer

1. Transverse isotropy 171

2. Orthotropic symmetry 174
3. erystal symmetries 175
4. Tensors of third and higher order 176
5. Reduction of a general tensor to a sum of
traceless symmetric tensors 178
6. Linearly independent invariants - Generating
functions 181
7. Minimality of an integrity basis 185

ehapter 10 KINEMATie eONSTRAINTS, eONSTITUTIVE EQUATIONS

AND FAILURE RULES FOR ANISOTROPie MATERIALS
by A. 1. M. Spencer

1. Kinematic constraints 187

2. Linear elasticity 191
3. Finite elasticity 193
4. Plasticity - Yield conditions 194
5. Plasticity- Flow rules 197
6. Plasticity - Hardening rules 197

ehapter 11 INVARIANTS OF FOURTH-ORDER TENSORS

by 1. Betten

1. Introduction 203
2. Integrity basis for a second-order tensor 204
3. Simplified characteristic polynomial 207
4. The Hamilton-eayley theorem 210
5. eonstruction of simultaneaus invariants 212
6. eonstruction of invariants by the polarization
process 214
7. Extended characteristic polynomial 215
 8. The Lagrange multiplier method 221

Chapter 12 FORMULATION OF ANISOTROPie CONSTITUTIVE EQUATIONS

by J. Betten

1. Introduction 227
2. The damage state in a continuum 229
3. Stresses in a damaged continuum 236
4. Constitutive equations involving damage and
initial anisotropy 240

Chapter 13 INTERPOLATION METHODS FOR TENSOR FUNCTIONS

by J. Betten

1. Introduction 251
2. Tensor function of one variable 252
3. Tensor function of two variables 256
4. Interpolation at coincident points 260
5. Polynomials of second-order and fourth-order
tensors 262
6. Simple examples 263
7. Tensorial generalization of Norton's creep law 265
8. Separation of tensor variables 274

Chapter 14 TENSOR FUNCTION THEORY AND CLASSICAL PLASTIC

POTENTIAL
by J. Betten

1. Introduction 279
2. Isotropy 280
3. Oriented solids 283
4. Modification of the classical flow rule 288
5. Anisotropy expressed through a fourth-rank
tensor 295
 LIST OF AUTHORS

J. BETTEN Rheinisch - Westfälische Technische Hochschule Aache~

Templergraben 55
D-5100 AACHEN
Federal Republic of Germany

J. P. BOEHLER Universite Scientifique Technologique et Medicale de

Grenoble
Institut de Mecanique de Grenoble
B.P. n° 68
38402 SAINT MARTIN D'HERES CEDEX
Franc-e

A. J. M. SPENCER The University of Nottingham

Department of Theoretical Mechanics
NOTTINGHAM NG7 2RD, England
United Kingdom
APPLICATIONS OF TENSOR FUNCTIONS

IN SOLID MECHANICS
 Chapter 1

PHYSICAL MOTIVATION

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

Theorems of representations for tensor functlons are valuable for

modelling non-linear constitutive laws, particularly when the mechanical
response of the material depends on more than one tensor agency. It is
an approach that leads to the general invariant forms of the non-linear
constitutive equations andgives the number and type of the scalar varia-
bles involved.
These representations for tensor functions have proved to be even
more pertinent in attempts to model the mechanical behavior of anisotropic
materials, since here invariance conditions predominate and the number of
independent scalar variables cannot be found by simple arguments.
In this Chapter, we present experimental evidence of anisotropic
response of materials and show the complexity of phenomena observed,
which indicates the need for a rational and unified formulation of aniso-
tropic constitutive laws.
4 J.P. Boehler

E (GPa)
225 R0 (MPa)
142
~ ,001

220 L ca lcula ted urv~

v
139 \
? ~

215 I 136
,
~~

I / ~ ...
133 L
' "
I

I
I
210 ,;ex peri men tal curv e

! 130
../ i
205 I 127 -
124
200
0 30 60
0 30 60 90

Fig.l - Anisotropy of the elastic Fig.2 - Anisotropy of the elastic

modulus of rolled sheet- limit of rolled sheet-
steel (after [1]). steel (after [2]).

!
[!oo
f R 0 (MPa)
R E (MPa)
10

a) Diatomite

0,2
5

Fig.3 - Anisotropy of the elastic Fig.4 -. Strength in simple compres-

modulus and compressive sion for uniform and perio-
strength of a natural clay dically non-homogeneaus
(after [3]). structure of diatomite
( after [ 4] ) .
Physical Motivation 5

2. DIFFERENT DOMAINS OF MECHANICAL ANISOTROPY

Oriented internal structures of solids, such as oriented crystallo-

graphic axes, grains, particles, fissuration, cracks, cavities, etc ... ,
result on the macroscopic level in a directional mechanical response to
applied agencies. Different domains of the mechanical behavior can be
influenced.
Fig.l shows the variation of the elastic modulus E of rolled sheet-
steel with respect to the angle 8 between the direction of the tensile
stress and the rolling direction (after [1]), whereasthe anisotropy of the
elastic limit R is presented in Fig.2 (after [2]). Anisotropy of the
elastic modulus E and strength R of a consolidated clay subjected to
oriented compressions is given in Fig.3 (after [3]).
According to the type of the oriented internal structure, the varia-
tion of mechanical characteristics, with respect to the orientation of
the material, may be continuous or discontinuous. This is shown in Fig.4
(after [4]), where the uniaxial compressive strength of diatomite, a
stratified soft rock, is traced versus the orientation 8 of the specimens
with respect to the normal of the strata. It is seen that the standard
compressive strength possesses a minimum within the range of inclination
of the strata. For a diatomite with marked layers of weaker strata, there
appears a sudden drop in the strengthfor inclinations ranging between 30°
and 45°.
In Fig.S, results regarding the strengthofdifferent consolidated
clays in compression are presented (after [5, 6]) : a) range of variation
for London clay ; b) Little Belt clay ; c) Vienna clay ; d) Welland clay
e) experimental points for Grenoble clay. It is seen that the variation
of strength with the orientation of the privileged direction of transverse
isotropy is quite irregular. The strength either decreases, increases or
passes through an extremum when the inclination of the privileged direct-
ion changes with respect to the principal stress direction. For triaxial
tests on shales, Fig.6, the strength is plotted against this inclination
angle for several values of the confining pressure (after [7]). Two
remarks are appropriate in connection with the experimental results pre-
6 J.P. Boehler

sented. In the first place a diversity of mechanical response with the

orientation of the privileged plane of transverse isotropy can be seen
in the figures. Secondly,a quite irregular dependence of the strength on
the hydrostatic pressure is noticed. This indicates that proposed crite-
ria of failure should possess a sufficient flexibility to permit an expla-
nation of experimentally observed behavior of solids with directional
mechanical properties. In view of the complexity of mechanical behavior,
a study of yielding and failure of oriented solids requires an adequate
and unified approach.

10 5 ,---------t---~----~---t---~-----,
psi
12

fO

r
9 0.6

0.4

0.2
6n t
0 '----~~~~3~0~~4~5--~60~--~~~~goo 0 '----1-;j:5:------;;3~0-----:;4~5---6;:;0;-------:;7::;:5--~90°
ß-- ß-

Fig.S - Directional strength Fig.6 - Directional strength of

of clays (after [5, 6]). shales under confining
pressure (after [7]).

3. ESSENTIAL FEATURES OF THE ANISOTROPie MECHANICAL BEHAVIOR OF ROLLED

SHEET-STEEL

In this Section, we present some essential features of the anisotro-

pic mechanical behavior of rolled sheet-steel, as well as the evolution
of the anisotropic behavior during irreversible deformations (anisotropic
hardening).Three questions are considered.
Physical Motivation 7

>'< First guestion : what is mechanical anisotropy ?

The definition is : "Variation of the mechanical response with the

orientation of the agencies with respect to the material". This is sket-
ched in Fig.7, which presents stress-strain curves for different orient-
ed specimens of rolled soft steel when subjected to simple tensile tests;
L is the rolling direction and 8 the angle between L and the direction of
the applied tensile stress. The figure shows the generally obtained
stress-strain curves in the rolling direction (8 = 0), the transverse
direction (8 = 90°) and at 45° from the rolling direction. The essential
features are:
• The yield stress in the transverse direction is higher thaL in
the rolling direction; at 45°, it is higher than in the trans-
verse direction.
• The deformations corresponding to the different yield stresses
vary with the orientation of the specimens.
• For the elastic behavior, we remark that the material is stiffer
for the orientations of lower yield stress.

We can already notice that the description of these anisotropic

phenomena is not simple: there exists no a priori rule for the variation
of the stiffness, the yield stress and the corresponding deformation;
moreover, these three variations are generally independent.
Fig.8 shows the evolution of the yield stress with the orientation
of the specimens. The more the curve moves away from the horizontal line,
the more the degree of anisotropy increases, In the case of isotropy, we
have a horizontal straight line.

* Secend guestion : how does the anisotropic mechanical behavior evolve

during cold rolling ?

Fig.9 represents experimental results [8] concerning the yield cri-

terion in simple tensile tests performed on oriented specimens of the
same steel initially subjected to three different rolling reductions:
r = 5 %, 10 % and 15 %. We observe that the anisotropy increases with the
rolling reduction, but no simple relation exists between the three dif-
8 J.P. Boehler

400
'LIR45
390 17 LI

~
~ -c:::J·
380 R90 1

370
LL V L1
'
///
360
RoD
350 •I
Fig.7 -Anisotropie stress-strain
relations. 340 ~V
330
_.!:_ /ß"
Re ;:l V

'~·
320
5 10 15

anisotropy
Fig.lO - Evolution.of oriented
I
yield stresses with the
:-isotropy
rolling reduetion.
eo
0 15 30 45 60 75 90 R9 (MPa>
420
IL_R4 R45
Fig.8 - Anisotropie yield 410 L.
eriterion in simple
tension. 400 iL VRsjo
/V
Re(MPa) Series
390 l Series C
r=15% V :L
Rgo

390 - ""'"~ r--. c 380

w L_RO / /
V
/
15%
L / /
/
I
........-j_ Ro

V
380 370

370 I
L ll
360
1/
L_ / v ~
l
I

I~ I/V
~ I
Series A I
I r=s·;.
I
360 350 '

_L ~
-
I

........
I
~a
350 / II
II '-...... 340
..... ~ .... B 10%
I/I_
II

I
34 0 I 330
I II
,...-
I

r-- r---. E%
33 0
=~ A
90
320
0 2 4 6 8 10 12 14 16

32 0
0 15 30 45 60 75 90 Fig.ll - Evolution of oriented
yield stresses of rolled
Fig.9 - Evolution of the yield sheet-steel with further
eriterion with the rolling applied irreversible
reduetion. prestrains.
Physical Motivation 9

ferent yield curves. The evolution of the criterion is non-linear. This is

also the case for the evolution of the orientation corresponding to the
maximum yield stress.
Another means of stuyding the evolution of the anisotropy of the cri-
terion is presented in Fig.lO, which shows the evolution of the yield
stress in the rolling direction, transverse direction and at 45° with
respect to the rolling direction. In the isotropic case, the three curves
would coincide. We observe that in the anisotropic case, the curves are
strongly different and that there exists no simple relation between them.

* Third guestion : how does the anisotropic mechanical behavior of rolled

sheet-steel evolve when subjected to further irreversible deformatio~

(prestrainings) ?

Fig.ll represents the evolution of the yield stress in the rolling and
transverse directions and at 45°, with respect to the prestrainings €

applied to steel initially rolled at r = 5 i. and 15 i. reduction. We can

observed two facts:
• For a given rolling reduction, the three curves are not parallel;
thus, the anisotropy changes with the prestrain: this is the phe-
nomenon of anisotropic hardening.
• The respective evolutions of the three curves are not the same for
two different rolling reductions.
So we see here further the complexity of anisotropic phenomena.

So far, we were concerned with the anisotropy of the yield criterion.

We shall now analyse the anisotropy of the flow law, not directly, but
through one of its consequences, which is the fact that the two necking
lines are no longer symmetric with respect to the axis of the specimens,
as it is the case for isotropic steel. Fig.l2 (after [8]) shows the evo-
lution of the orientation of the two necking lines with the orientation
of the specimens. For each value of the irreversible prestrain, the two
curves ~l and ~2 together form the shape of an eight. We observe that the
figures evolve in a complex manner with the prestrain.
10 ].P. Boehler

58
E =0,020

~-----r--_____",,-------".,-. isotropy
90
0 15 30 45 60 75 90 0 15 30 45 60 75 90

62
necking
I ine 2 necking
line 1

R
a
0 15 30 45 60 75 90

Fig .12 - Evolution of the orientation of the necking lines (after [8]).

From these experimental facts, we can conclude that the anisotropic

mechanical behavior concerns very different phenomena, which are very
complex and strongly non-linear. There appears to be a need to develop
an objective and unified theory able to describe these different pheno-
mena. In such a development, two main problems arise:
• What is the general invariant form of a constitutive law able
to describe the anisotropic phenomena ?
• What are the essential variables, i.e. the nurober and type of
variables which have to be measured during experiments in order
to identify the specific constitutive equations ?

We shall see in the next Chapters that the theorems of representa-

tions for anisotropic tensor functions allow these two requirements to
be fulfilled.
Physical Motivation 11

REFERENCES

1. PARNIERE, P., Relations entre Textures et Comportement Mecanique des

Metaux, in: Mechanical Behavior of Anisotropie Solids, Proc. Euro-
mech Colloquium 115 - Colloque Internatioral 295 du CNRS, J.P.BOEHLER
ed., Editions du CNRS (Paris) and Martinus Nijhoff Publishers (The
Hague), 1982, 303-332.
2. BUNGE, H.J. and ROBERT, W.T., J. Applied Cryst., 9 (1969) : 116.
3. BOEHLER, J.P., Gontribution a l'Etude de l'Equilibre Limite des Sols
Anisotropes, These de Doctorat de Specialite, Grenoble, 1968.
4. ALLIROT, D., BOEHLER, J.P. and SAWCZUK, A., Pressure-Induced Evolution
of Anisotropies in Stratified Rock, Studia Geotechnica et Mechanica,
Vol. III, n°2-4, 1981, 59-73.
5. DUNCAN, J.M. and SHEED, H.B., Anisotropy and Stress-Reorientation in
Clay, Proc. ASCE 92, SM5, 1956, 21-50.
6. BOEHLER, J.P., Gontributions Theoriques et Experimentales a l'Etude
des Milieux Plastiques Anisotropes, These de Doctorat es Sciences,
Grenoble, 1975.
7. GRAY, K.E., Some Rock Mechanics Problems of Petroleum Engineering,
Proc. 9th. Symp. Rock Mechanics, The Colorade School of Mines, 1967,
405-433.
8. BOEHLER, J.P. and RACLIN, J., Anisotropie Hardening of Prestrained
Rolled Sheet-Steel, in: Current Advances in Mechanical Design and
Production, Proc. Second Cairo University MDP Conference, Dpt. of
MDP, Faculty of Engineering, Cairo University, 1182, 483-492.
 Chapter 2

INTRODUCTION TO THEINVARIANT FORMULATION

OF ANISOTROPie CONSTITUTIVE EQUATIONS

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

The aim of this Chapter is to present some elementary notion for

non-specialists in the invariant formulation of anisotropic constitutive
equations. Much of this Chapter is taken from [1].
We consider the very simple case of a two-dimensional anisotropic
material of constitutive equation specifying a symmetric 2nd order tensor
T, which is a function of two symmetric second order tensors D and ~:

T F(D, 0 ( 1)

In (1), ~ is the structural tensor, taking into account the symmetries of

the material internal structure, D is the mechanical agency and T the
mechanical response of the material. We use the absolute notation, not to
be related to coordinates at this point. For definiteness, T is the stress
tensor, D is the strain tensor in the case of elastic behavior or the
14 J.P. Boehler

rate of strain tensor in the case of plastic behavior. In genera1, the

principal directions (e ' e ) of the response T do not coincide with the
- 1 - 2
principal directions (E ' E ) of the agency ~·
-1 - 2
The constitutive equation (1) is sketched in the Figure below,
where the material, in its initial configuration, is represented by a
circle and the attached orthonormal frame (v , v ); the tensorsTand D
-1 -2 -
are represented by their associated ellipses. Similar sketches will be
used throughout this Chapter.

Initial configuration
F
...
of the material
Agency D Response T
the material
F depends on ~
the behavior

Constitutive relation

2. PRINCIPLE OF ISOTROPY OF SPACE

2.1 Physical concept

Constitutive equations are subjected to the invariance requirements

of the Principle of Isotropy of Space (or Principle of Material Indif-
ference [2]). We shall apply it here to the simple situation involved in
the equation (1).
Introduction to the Invariant Formulation 15

A consequence of this Principle is that an arbitrary transformation

Q of the orthogonal group 0 and applied to both the material and the
agency D, results in the same orthogonal transformation of the material
response T. A naive statement of this is that the orientation of the
material in the space has no effect on its constitutive relation. This
is sketched in the Figure below.

D T

} =<> ara'

Consequence of the Principle of Isotropy of Space

2.2 Mathematical concept

The transformation of the body

and the transformation of the agency D =>

result in the same transformation of the response: T =>

Finally we obtain:

VQ E 0 ~(9~9t' 9~9t) = 9~(~, pgt

or: VQ E 0 F(D, ~) = !<~. ~) (2)

Relation (2) indicates the invariance conditions to which the constitu-

tive equation (1) is subjected by the application of the Principle of
Isotropy of Space.
16 J.P. Boehler

The tensor-valued function F, subjected to the invariance conditions (2),

is called isotropic with respect to its two arguments (D, ~).

3. ISOTROPIC MATERIALS

3.1 Physical concept

A material is called isotropic if an arbitrary transformation Q of

the orthogonal group 0 and applied to the material, but not to the agency
D, results in the same response T. This is sketched in the Figure below.

E1 e1
Qv2~ Ov2\
\
Qv1
__.,.
- --av 1

-- v1
~
v1

===t>T

Definition of an isotropic material

A direct consequence of this definition is that the principal directions

E , E ) of the agency -D and the principal directions (e , e ) of the
-1 -2 -1 -2

response T are the same. To see that, we apply the reflection S with res-
pect to the direction ~1 to both the material and the agency. We obtain
obviously:

D => D D ( 3)

The Principle of Isotropy of Space requires that:

Introduction to the Invariant Formulation 17

T => T (4)

But, in view of (3), the agency is unchanged by the transformation S.

Thus, the isotropy of the material requires that:

T = T (5)

Introducing (5) in (4), we obtain:

STSt = T (6)

The relation (6) implies that E is a principal direction of T; finally,

-1

the principal directions of T and D are the same.

Thus, the mechanical behavior of an isotropic material can be sketch-
ed by the following Figure, where the isotropic material is represented by
a circle and the agency ~ and the response ~ by their associated
ellipses, which admit the same principal axes.

Isotropie material

3.2 Mathematical concept

The isotropy of the material implies that if we:

transform the body

and keep the same agency D => D

then the response is the same: T => T

18 J.P. Boehler

Finally, we obtain:

VQ E 0 F(D, ~) (7)

It is easy to see that the condition (7) means that the structural tensor
~ is an isotropic tensor:

~=AI

Then,for isotropic materials, the constitutive equation (1) reduces to:

T = F(D, AI) = F(D) (8)

- -
If we now apply the Principle of Isotropy of Space, we obtain:

VQ E 0 : F(D) = F(D) (9)

which means that for isotropic materials, the constitutive function F is

isotropic with respect to the agency D.

4. ANISOTROPie MATERIAL

4.1 Physical concept

A material is called anisotropic if when we transform the material

by an arbitrary orthogonal transformation 9 and keep the same agency D,
then, in general, the response !* is different from the response !
obtained before the transformation. This is sketched in the Figure below.
The following question arises immediately: are there orthogonal transfor-
mations Q such that the response T is the same:

39
-lc
~ ?
E 0 : ! = ( 10)
If only the transformations (I, -I) satisfy (10), then the material is
called "general anisotropic material". If other orthogonal transformations
9 satisfy (10), then they constitute a subgroup S of the orthogonal group
0. The subgroup S is called the material symmetry group.
For examp1e, the material symmetry group for a two-dimensional ortho-
tropic material, with privileged directions (v , v ), is given by:
-1 -2
Introduction to the Invariant Formulation 19

s = ~ ± I, S , S
- -1 -2~
l (11)

where S and S are the reflections with respect to the directions

-1 -2
v and v
-1 -2

v1

D,v1 ,v 2 ===(> T V' \

0 ,Qv1 ,Qv2 ~ r•=F T

Anisotropie material

4.2 Mathematical concept

Suppose that s, the symmetry group of the material, is also the

"invariance group" of the structural tensor ~:

QE S <=) ~ (12)

It follows that:
Q E S =) F(D, ~) F(D, 0 (13)

The Principle of Isotropy of Space requires that:

VQ E 0 : F(D, ü F(D, n (14)

Employing (14) in ( 13), we arrive at:

Q E S =) F(D, n F(D, n (15)

20 J.P. Boehler

Conversely, suppose that the relation (15) holds for a given transforma-
*
tion Q . Employing the Principle of Isotropy of Space (14), we obtain:

3 9*e: 0 : F(D, 0 = F(D, ü (16)

for an arbitrary agency D. It is easy to see that for a general function F,

the condition (16) leads to:
>~
~ = ~ => Q e: s (17)

Finally, considering the relations (13), (16) and (17), we obtain the
following converse implication:

Q e: s <=> F(D, ~) F(D, ~) (18)

The implication (18) means that, with respect to ~· the function ~ is

invariant in the group S and only in this group. Thus, the function F
is anisotropic with respect to ~· the type of anisotropy being charac-
teristized by the invariance group S of the structural tensor ~. This
leads to the following fundamental result :

T = F(D, ~) + Isotropy of Space

,g,
F isotropic I (D, 0
,g,
F anisotropic I (D)
the material symmetry group s
being the invariance group of the structural tensor ~

Finally, for an anisotropic material, the invariance group of the struc-

tural tensor ~ involved in the constitutive equation (1) is identical to
the group of transformations characterizing the material symmetries of
the body. The problern now is to find a structural tensor for each type
of anistropy.
In the constitutive equation (1) for the considered two-dimensional
anisotropic material, we supposed that the internal structure can be
Introduction to the Invariant Formularion 21

taken into account by a single symmetric secend order tensor ; this is

indeed the case for a two-dimensional orthotropic material (cf. Section 5
below). For other types of anisotropy and in the three-dimensional case,
several structural tensors may be necessary. For example, the internal
structure of a three-dimensional orthotropic material is represented by
three symmetric second order tensors ([3] ; seealso Chapter 3, Section
3-4) ; in this case, the group S is the invariance group of the set of
structural tensors.

5. ORTHOTROPIC MATERIALS

Consider a two-dimensional orthotropic material, with privileged

direction (v , v ). The material symmetry group s is given by (ll).We
-1 -2
introduce the following tensor M:

M (19)

The tensor M is a very particular symmetric second order tensor, which

transforms the unit vectors n into their projections upon the v axis:
-a -l

M
cosa. cosa.
n => d
-a. sina. -a. 0
M -1 +1 v,

Thus, the tensor M transforms the unit circle into the segment [ -1, +1] lying
on the axis. The principal directions of M are the privileged directions
~1

(v , v ) of the orthotropic material. In the (v , v ) axes and in the

-1 -2 -l -2
(~, ~) axes deduced from (v , v ) by a rotation of angle 6, the tensor
-l -2
M is expressed by :
22 J.P. Boehler

M
cos 2 8 s~n8cos81 ( 2 0)
sin8cos8 s1n 28
(x,y)

It is easy to see that the tensor M verifies the following converse impli-
cation:

(21)

Thus, the invariance group of M is the material symmetry group (11), i.e.
M is the structural tensor for a two-dimensional orthotropic material.
We now apply the fundamental result of Section 4. Consider the cons-
titutive equation:

T F(D, M) (22)

The Principle of Isotropy of Space implies that the function F is isotro-

pic with respect to (~, ~), thus anisotropic with respect to (D), the
type of anisotropy being characterized by the invariance group of the
structural tensor M. Finally, the function F is orthotropic with respect
to D:

Isotropy of Space => VQ E Q F(D, M) F(D, M) (23)

VQ E S : F(D, M) F(D, M) (24)

+
F orthotropic I (D)
- -
Finally, for orthotropic materials, the constitutive equation (1)
takes the particular form (22) which is subjected to the invariance condi-
Introduction to the Invariant Formulation 23

dition (23) given by the Principle of Isotropy of Space. Equations (22)

and (23) constitute the general form of a two-dimensional orthotropic
constitutive law relating two symmetric secend order tensors.

6. REPRESENTATION OF THE FUNCTION F

6.1 Introduction

In Mechanics of Anisotropie Solids, it appears necessary to establish

invariant forms of the constitutive laws, i.e. forms which verify automa-
tically the material symmetries of the considered body and are thus, auto-
matically invariant under the material symmetry group. This is not the
case of the constitutive equation (22), because it is subjected to the
invariance condition (23). It is also necessary to derive the essential
variables, i.e. the type and the minimal number of independent scalar
variables which must appear in the constitutive law and which constitute
the set of independent anisotropic mechanical variables.
The theorems of representationsfor anisotropic tensor functions al-
low canonical forms which fulfil these two requirements to be developed.

6.2 Theorem of representation

The theorem of representation for orthotropic tensor functions ([3])

indicates that the relation (22), tagether with the invariance condition
(23), admits the following irreducible canonical form:

T a I
0-
+ a 1-M + a 2-D (25)
a. a.(trD, trD 2 , trMD)
1 1 - - --

In the representation (25), the response ~ appears as a linear combination

of the following three well determined tensors:

I, M, D (26)
- - -
which constitute the generating set of the representation. More precisely
the generators (26) are known s - invariant symmetric second order ten-
sors, where s is the material symmetry group (11) for two-dimensional
24 J.P. Boehler

orthotropy. The coefficients a. are arbitrary scalar-valued functions,

1
of arguments:

trD, trD 2 , trMD (27)

which are called "orthotropic invariants" of D. They constitute the

"functional basis" of D for the considered two-dimensional orthotropic
group of transformations. Precise definitions of the terms introduced
above are given in Chapter 3.

6.3 Orthotropic invariants of ~

We introduce the expressions of the agency ~ in its principal frame

(E , E ) and in the privileged frame (v , v ) of the material:
-1 -2 -l -2

E E
11 12
D (28)

The first two invariants (27) admit the following expressions

trD E
1
+ E
2
E
11
+ E
22
tr(QDQt) , VQ E o

t
(29)
E2
1
+ E2
2
= E 11 + E
22
+ 2E 2
12
= tr(QDQ
---
)2 , VQ E 0

Thus, trD and trD 2 admit the same values for arbitrary orthogonal trans-
formations of D: they are called isotropic invariants of D. As they are
invariant under the full orthogonal group O, they are invariant under the
group S a fortiori ; thus, they constitute also orthotropic invariants
of D.
Introduction to the Invariant Formularion 25

The expressions of the last invariant (27) are given by:

trMD E
l l
e (v ,
-l
E
-l
) (30)

Thus, trMD is equal to the component E 11 of D in the privileged frame

(~ 1 , ~ 2 ). If we transform both ~ and ~ by an arbitrary orthogonal trans-
formation Q, the value of this component remains constant: trMD is an
isotropic invariant of the set (D, M). If we transform D, but not M, the
value of the component E
l l
is modified, except if, and only if, the trans-
formation Q belongs to the orthotropic group s given by (11) ; thus,
trMD is an orthotropic invariant of D. Finally, we have:

VQ E 0 ::::> s trD trD trD 2 trMD trMD (31)

VQ E s trMD = trMD (32)

6.4 Fundamental resu1t

The representation (25) verifies automatically the consequence(2)of

the Principle of Isotropy of Space. The proof of this property is obtained
in considering that, in view of (31), we have:

VQ E 0 a.(trD, trD 2 , trMD) (33)

l -

VQ E o: F(D, M) = a I
0-
+ a l-M + a 2-D

Q(a I
- 0-
+ a lM
-
+ a 2-D)Q- t (34)

= F(D, M)

Relation (34) signifies that the representation (25) is automatically

isotropic with respect to the set of arguments (D, M) ; thus, it is auto-
matically orthotropic with respect tothemechanical agency D. The essen-
tial variables are the arguments (trD, trD 2 , trMD)of the arbitrary scalar-
26 J.P. Boehler

valued functions a. ; they are the independent variables which must be

1
observed in experimental investigations, in order to be able to specify
the particular forms of the functionsa., foreachtype of material and
1
each type of mechanical behavior. Finally, both requirements mentioned
in Section 6.1 for the invariant formulation of the constitutive equation
(22) for a two-dimensional orthotropic material are fulfilled by the
representation (25).
In representation (25):
• The structural tensor M with principal directions
(v , v ) characterizes the orthotropy ; i.e. it specifies
-1 -2
the privileged directions of the material.

• The invariant trMD = E l cos 2 8 + E

2
sin 2 8 characterizes the
"degree" of the orthotropy, i. e. i t takes into account
the influence of the orientation 8 of the agency D with
respect to the privileged directions (v , v ) onto the
-1 -2
mechanical response T of the orthotropic material.

6.5 Characteristic property

If the privileged directions (v , v ) are not the principal direc-

-l -2
tions of the agency D, then the principal directions of D and of the
response T do not coincide in general. This property is obvious, in view
of the linear combination:

T a I + a M + a. D ( 35)
0- l- 2-

It is also clear in the sketch below

y2 E1
v2 E1

<F
v1
+
~ v1 +

T -- o0 1 + o1 M + o2 D
Introduction to the Invariant F ormulation 27

It can be proved in considering the expression of (35) in the principal

frame (E , E ) of D:
- 1 - 2 -

a
11
a
1 2
1 0 sin8cos8 E:
1
0
Cl
0
+ Cl
l
+ Cl
2
(36)
a1 2 a2 2 0 1 sin8cos8 0 E:
2

where

8 (v , E )
-1 -1

If we exclude the special case a 1 - 0, then the component:

a a 1 sin8cos8 (37)
12

of T vanishes if and only if: 8 = 0, n/2.

6.6 Structural interpretation of M

In Section 5, we remarked that the tensor M transforms the unit

circle into the segment [-1, +1] on the ~1 axis. The unit circle repre-
sents the isotropic distribution of the material internal structure
(crystallographic axes, grains, particles, crystals, fissuration, cracks,
cavities, etc ... ). The segment [-1, +1] represents the ideal orthotropic
distribution of the internal structure, parallel to the v axis. Thus,
l

M is the tensor of an ideal orthotropic distribution. As real material

orthotropic structures are seldom ideal, the following question arises
immediately: does the representation (25), in which the involved struc-
tural tensor is only the ideal tensor M, allow an arbitrary orthotropic
distribution to be taken into account ? First, we show that the represen-
tation (25) takes into account an elliptic orthotropic distribution. From
(25), we can construct the structural tensor:

M
-2
= I - M = v-2 e v
-2
(38)
28 J.P. Boehler

The tensor M is a very particular symmetric second order tensor, which

-2
transforms the unit vectors n-a into their projections upon the v-2 axis:

v2
+1
M
-2
cosa 0
n => f M2:
-Cl
sina
-Cl
sina v,

-1

Thus, the tensor ~2 transforms the unit circle into the segment [-1, +1]
1ying on the ~2 axis. Consider now the structural tensor:

aM + bM-2 bl + (a-b)M (39)

V )
-2

The tensor ~. which is involved in the representation (25), transforms

the uni t circle into the ellipse with principal directions ( v- l , v-2 ) and
with semi-axes a and b.

M·
2'

-1 +1 V
1
v, -1 +1 V
1

-1

Thus, the tensor ~ is the structural tensor for an elliptic orthotropic

distribution.
In fact, representation (25) can account for an arbitrary orthotropic
distribution. Indeed, the representation involves the arguments of the
Introduction to the Invariant Formulation 29

scalar-valued functions ai, which are the invariants:

trD E + E trMD (40)

1 2

where E and E are the principal values of D and 8 is the angle (v , E ),

1 2 -1 -1
Let us suppose thatthe values of the three invariants (40) are given.
The expressions of trD and trD 2 allow thedetermination of the values of
E1 and E2 • Introducing these values in the expression of tr~~· one
obtains a single value for 8 in the interval [0, TI/2], the three other
solutions in the interval [0, 2TI] being obtained by the symmetries with
respect to the v and v directions. Now, for a given arbitrary ortho-
-l -2
tropic distribution in the interval [0, TI/2], the distribution in the
interval [0, 2TI] is obtained by the same symmetries. Finally, from the
arbitrary scalar-valued functions a., arbitrary scalar-valued orthotropic
1
functions a(S) and b(8) can be constructed. Thus, from representation (25),
the following structural tensor can be constructed:

~
-
= a(S)M- + b(S)M
-2
b(S)I + [a(S) - b(S)]M (41)

which is the polar representation for an arbitrary orthotropic structural

tensor.

7. CONCLUSIONS

The theory of representations for tensor functions is a powerful and

efficient tool for the invariant formulation of non-linear anisotropic
constitutive equations. This theory specifies the invariant tensorial
generators which are involved in the general form of a non-linear consti-
tutive equation, as well as the type and the nurober of the independent
scalar variables, which are mixt invariants of the mechanical arguments
and of the structural tensors.
In this elementary introduction to the invariant formulation of
anisotropic constitutive equations, we considered the very simple case
of a two-dimensional equation specifying the stress tensor as a function
of a single mechanical argument ; in this case, the invariant general
30 J.P. Boehler

form of the constitutive equation is given by the representation (25). In

the three-dimensional case, representations of tensor functions of an arbi-
trary number of mechanical arguments are given in Chapter 3 for general
anisotropic, orthotropic and transversely isotropic materials.
In the invariant general forms of the anisotropic non-linear consti-
tutive equations so obtained, a certain number of arbitrary scalar-valued
functions appear, the arguments of which are the well defined essential
variables. It is the particular forms of these arbitrary functions which
specify the invariant forms of the constitutive equations for each type
of anisotropic material. The specification of the arbitrary scalar-valued
functions needs coherent and well-organized experimental investigations,
which really allow the observation and measurement of the evolution of
the essential variables. The experimental study of the mechanical beha-
vior of anisotropic materials involves specific difficulties, which are
mentioned in Chapters 5 and 7.

REFERENCES

1. BOEHLER, J.P., Anisotropie et Comportement Rheologique des Materiaux,

Conference Generale, 19eme Colloque du Groupe Fran9ais de Rheologie,
Novembre 1984 (in print).
2. TRUESDELL, C. and NOLL, W., The Non-Linear Field Theory of Mechanics,
Handbuch der Physik, III/3, Springer-Verlag, Berlin, 1965.
3. BOEHLER, a.P., Lois de Comportement Anisotrope des Milieux Continus,
Journal de Mecanique, 17,2 (1978) : 153-190.
 ehapter 3

REPRESENTATIONS FOR ISOTROPie AND ANISOTROPie

NON-POL YNOMIAL TENSOR FUNeTIONS

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

1.1 Invariant scalar functions and form-invariant tensor functions

Material symmetries of a continuuro impose definite restrictions on

the form of constitutive relations. The restrictions are specified in the
representations of isotropic and anisotropic tensor functions and indi-
cate the type and the nurober of independent variables involved in a cons-
titutive relation. Thus, in a properly written constitutive equation, the
material symmetries are automatically verified.
In this Chapter, "tensors" and "vectors" mean second order tensors
and vectors in a three-dimensional space. We restriet ourselves to consti-
tutive functions of the arguroents:

A
_l
• A
_2
• . .. A
_a • w
_l
•w
_2
• • .• Wb.
_
V •
_l
V •
_2
• •• V
_c (1 )

where theA., W. and vk are respectively an arbitrary nurober of symmetric

-1 -J -
tensors, skew-symmetric tensors and vectors.
32 J.P. Boehler

Consider first a constitutive law specifying a scalar A

A = f(A., W., vk) (2)

-1 -J -

If S is the group of transformations characterizing the material symme-

tries of the medium, the function f is subjected to the conditions:

VQ e: s (3)

The scalar A is called scalar invariant under the group s.

Consider now a constitutive equation specifying a symmetric second
order tensor T:

T = F(A., W., ~k) (4)

- - -1 -J

Similarly, the function F is subjected to the conditions:

The function F is then called form-invariant under the group S and T is a

tensorial invariant.

1.2 Representations for polynomial scalar and tensor functions

In this Section, we suppose that the function f and the components in

a reference frame of the function F are polynomials in the components of
the arguments (1). The values A and T are then polynomial invariants under
the group s.
The problern of the representation for the scalar-valued function f is
to determine a basic set of polynomial scalar invariants (I , I , ... I ),
1 2 p
such that an arbitrary polynomial scalar invariant of the same arguments
can be expressed as a polynomial in the basic invariants. Such a set is
called "integrity basis" for the considered list of arguments (1) and the
group of transformations s. An integritybasis is termed irreducible if
none of its subsets constitutes a complete representation by itself, i.e.
Representations for Tensor Functions 33

if no element of it can be expressed as a polynomial in the remainder.

Even if an integrity basis is irreducible, polynomial relations may exist
between its elements; but these relations do not enable any one of it to
be expressed as apolynomial of the others. Such relations are called
syzygies.
The problern of the representation for the tensor-valued function F
is to determine a generating set of tensors G., thus symmetric secend
-1
order tensors which are invariant under the group s, such that the tensor
T can be expressed as the linear combination:

l
T = a..G.
1-1

a..= a..(I , I 2 , ••• I ) (6)

1 1 1 p

where the a.. are arbitrary polynomial scalar functions of the invariants
1
of the integrity basis. The values of the functions a.. are thus polynomial
1
scalar invariants of the arguments (1) under the group s. The representa-
tion (6) is irreducible if the integrity basis is irreducible and if none
of the generators G. can be expressedas a linear combination of the others,
-1
with coefficients a. ..
1
Explicit representations for isotropic and anisotropic polynomial
scalar-valued and tensor-valued functions are presented in Chapters 8 and
9 by Professor SPENCER.

1.3 Representations for non-polynomial scalar and tensor functions

In this Section, we suppose that the function f and the components

of the function F are general functions, not necessarily polynomials, of
the components, in a reference frame, of the arguments (1). The values A
and T are then (general) invariants under the group s.
The representation for the scalar-valued function f is a basic set of
scalar invariants (I , I , •.. I ), suchthat an arbitrary scalar invariant
1 2 n
of the arguments (1) can be expressed as a single-valued function of the
basic invariants. Such a set is called "functional basis" for the consi-
dered list of arguments and the group S. A functional basis is termed
irreducible if none of its elements can be expressed as a single-valued
34 J.P. Boehler

function of the others. Even if a functional basis is irreducible, func-

tional relations may exist between its elements; but these relations do
not enable any one of it to be expressed as a single-valued function of
the remainder. Although they represent all functions, irreducible func-
tional bases contain, in general, fewer terms than do the corresponding
integrity bases, because here reductions are obtained with the help of
general functions and not only polynomials.
A criterion characterizing functional bases is given by WINEMAN and
PIPKIN [1]. We call equivalent set, under the group s, of the erdered list
of arguments (1) any erdered list (A, A, ... A , W , W , ... Wb' v , v , ...
-1 -2 -a -1 -2 - -1 -2
; ) whose elements are the transformed, by a same orthogonal transformation,
-c
of the corresponding elements of the list (1). The set of lists equivalent
to a given list (1) is called the orbit of this list in the group S. A
functional basis characterizes the orbits: a set of invariants constitutes
a functional basis if, and only if, their values are the same for two lists
of arguments lying on the same orbit and if, and only if, at least one of
them, called orbit-seperator, takes different values for two lists of argu-
ments lying on two different orbits.
The representation for the tensor-valued function F admits the same
form as in the polynomial case:

T a..G.
1-1

a.. a..(I , 1 2 , ••• I ) (7)

1 1 1 n

but, here, the scalar valued-functions a.. are general functions of the in-
1
variants of the functional basis of the arguments (1). The representation
(7) is irreducible if the functional basis is irreducible and if none of
the generators G. can be expressed as a linear combination of the others,
-1
formed with general functions a. .. As is the case for the bases for scalar
1
functions, irreducible generating sets for general tensor functions con-
tain, in general, fewer elements than do the corresponding generating sets
for polynomial tensor functions, because here reductions are obtained with
the help of linear combinations, the coefficients of which are general
scalar-valued functions and not only polynomials.
Representations for Tensor Functions 35

WINEMAN and PIPKIN [1,2] showed that a representation for the

polynomial case constitutes a complete representation for the non-polyno-
mial case as well. The scalar functions a. are then not necessarily poly-
~

nomials. However, in the non-polynomial case these representations are not

irreducible in general (SPENCER [3], WANG[S]). Conversely, irreducible
representations for non-polynomial functions are also representations for
polynomial functions, but they are not necessarily polynomials.
Finally, representations for general functions present a double ad-
vantage : on one hand, they allow restrictive assumptions for the consi-
dered functions to be avoided ; on the other hand, they contain, in gene-
ral, a fewer number of basic elements and thus allow a more concise formu-
lation for the considered constitutive equations to be obtained.

2. REPRESENTATIONS FOR ISOTROPIC SCALAR AND TENSOR FUNCTIONS

In this Section, we suppose that the functions f and F are isotropic,

i.e. the relations (3) and (5) are verified for all orthogonal transfor-
mations Q. The values A and T are then invariant under the full orthogonal
group o.
For polynomial isotropic functions of the arguments (1), the integri-
ty bases for the scalar functions f have been established by SMITH [7] and
the generating sets for the tensor functions F by SPENCER [3]. The lists
of isotropic invariants and tensor generators are too long to be reproduced
here. The methods used to derive them are presented in Chapter 8 by
Professor SPENCER.
The representations for general isotropic scalar-valued and tensor-
valued functions of arguments (1) were studied separately by WANG [4-6]
and by SMITH [8,9]. For the construction of the functional bases, WANG
applies the characteristic criterion presented in Section 1.3. In order to
establish the representations for the tensor functions F, WANG decomposes
the list of arguments (1) into independent sublists. Irreducihle generating
sets associated with each sublist are obtained by geometrical reasoning.
The irreducible generating set associated with the complete list (1) is
then the union of the generating sets associated with all independent
36 J.P. Boehler

sublists.
The procedure used by SMITH in order to construct the functional
bases for general isotropic scalar functions f constitutes an extension
of the method developed by RIVLIN and ERICKSEN in [10], applied to a par-
ticular case of the arguments (1). The method consists in establishing a
set of basic invariants such that once the values of these invariants are
known, the components of the arguments (1) can be determined uniquely in
a reference frame fully specified beforehand (orientation and sense uf the
Coordinates are specifjed). Such a set of invariants constitutes a func-
tional basis since any invariant is uniquely expressed in terms of these
components. The required isotropy is assured by the preliminary choice of
the reference system, which must be defined from the arguments (1) them-
selves. In order to establish generating sets for the general isotropic
tensor functions F, SMITH employs WANG 1 s procedure.
The representations obtained by WANG and by SMITH are not identical.
Moreover, the functional bases derived by SMITH contain redundant elements.
This has been analysed in [11], where after suitable amendments of the
previously established bases, an irreducible functional basis is obtained.
Reasonings similar to those employed in [11] allow conclusions to be drawn
for the generating sets as well. The complete and irreducible representa-
tions so obtained are given in the Tables I and II below.
For example, the irreducible representation for an isotropic tensor
function of two symmetric second order tensors :

T F(A, B) (8)

involves eight tensor generators

(9)

and ten invariants

trA, trA 2 , trA 3 , trB, trB 2 , trB 3

(10)
trAB, trA 2 B, trAB 2 , trA 2 B2
Representations for Tensor Functions 37

Variables Invariants

A trA, trA 2 , trA 3

- - -
w trW 2
- -
V v·v
- - -
A trA A , trA 2 A , trA A22 , trA- 21-A22
- 1' - 2
A - 1- 2 -1-2 - 1-
A, w trAW 2 , trA 2 W2 , trA 2 W2 AW
- - ---- - - --
A, v·Av, v·A 2 v
- -
V
- -- - - -
w' w
-l -2
trW
-l- 2
w
W, v•W 2 v
- - V
- --
V V ·v
- 1'
V - 2 - 1 - 2
A trA A A
A -2 ' - 3
-1 ' A -1-2-3
w A2 W, trA W2 A w
trA A W, trA 21-A2-W, trA- 1-2-
A -2 '
-1 ' A - - 1- 2- - -1- -2-
A, w ' w trAW W , trAW 2 W , trAW W2
- -1 -2 --1- 2 --1-2 --1-2

w ' w w
-2 ' -3
trW ww
-1-2-3
- 1
A v•A
- 1 ' ~2'
V
A V
- -1- -2-
A, V ' ·Av , •A 2 v
- - 1
V
- <
V
-1 -- 2
V
-1 - -2
w v•W v, W2 v·W v, w v•W 2 v
-l ' w
w -2 ' -
V
- l - -2- -1- -2- -1- -2-
w, V- 1 ' V-2 V ·Wv V •W 2 v
-
- - 1 .... -2' - l - 2
A, W, Av·Wv, A2v•Wv, AWv•W 2 v
- - - -- --- - -
V
-- --
A ,A , V A V ·A - -A1-V 2 ·A- 2-V 1
V
-1 -2 - 1'
V -2 -1-1 -2-2
A, w, Av •Wv - Av •Wv
- -
V
-l '
V -2 -- 1 --2 --1
--2
w w V w V ·W V - w V ·W V
-1 -2' -1 '
V -2 -1-1 -2-2 -1-2 -2-1

The complete list of invariants for a set of variables (A- l ,

A, W , ... Wb' v, ... v) is obtained considering all the
-a -1 - -1 -c
(unordered ) combinatio ns of one, two, three and four variables
given in the Table.

Table I: Irreducihle isotropic functional bases

38 J.P. Boehler

Variables Generators

I
-
A A, A2
- -
w wz
- -
V V
-
V
Qll
- -
A A A A + A A , A2 A + A A2 , A A2+ A2 A
-l ' -2 -l-2 -2-l -1-2 -2-1 -l-2 -2-l

- w
A,
- AW-WA, WAW, A2 W-WA 2 , WAWLW 2AW
---- --- - - -- --- - --
A2 v + A2 v
- -- + -- - -
A, V Av Av v, V
- - -- -- -
V Qll Qll Qll Qll V

w w
-l ' -2
w w+ w
-l-2 w' w
-2-l
W2 + W
-1-2
2W '
-2-1
W2 W+
-l-2
w wz
-2-l

w, Wv Wv, Wv + Wv v, Wv W2v + W2 v Wv
V
-- -- V- -- -- --
Qll Qll Qll Qll Qll
- - - -- - - --
V
-l '
V
-2
V
-1
Qll V
-2
+ V-2 1111 V
-1

A, V (v ® Av + Av Qll V ) - (v ® Av + Av Qll V )
- V
-1 ' -2 -1 --2 --2 -1 -2 --1 --1 -2

\:!. '::\' V
-2
(v
- 1
® Wv
--2
+ Wv
-- 2
Qll V
- 1
) - (v- 2 ® Wv
-- l
+ Wv
-- 1
Qll V
- 2
)

An irreducible representation for an isotropic function T F(A , ..•

- -1

A, W, ... Wb'
-a -1 -
v,
-l
... v)
-c
is given by : F = a.G.,
-1-1
where the a. 's are arbitrary scalar-valued functions of the invariants
1
of the functional bases given in Table I and where the G. 's are the
-1
generators given in Table II, considering all (unordered) combinations
of zero, one, two and three variables.

Table II: Irreducihle isotropic generating sets

Representations for Tensor Functions 39

An irreducible functional basis is not unique. The nurnber of its ele-

ments depends on the type of the invariants chosen for its construction.
As pointed out by WANG [6], an irreducible functional basis should contain
a single element. Suppose that a set of p scalar invariants, the values
of which belong toffiP, whereffi is the set of real nurnbers, constitutes a
functional basis. A one-to-one correspondence between the set of basic in-
variants andffi is a scalar invariant which constitutes by itself a new
functional basis. As ffi and ffi p have the same cardinali ty, such a corres-
pondence exists. However, as these two spaces are not homeomorphic, such
a correspondence is discontinuous almost everywhere and, thus, rather un-
wieldy in practical applications. The invariants in Table I are simple
polynomial functions of the argurnents (1), more precisely they are traces
of products of the argurnents (1) ; thus, general isotropic scalar functions
are represented by polynomial invariants. Another choice of polynomial in-
variants can lead to irreducible functional bases containing fewer terms.
The minimal nurnber of polynomial invariants necessary to constitute a com-
plete functional basis is at present a problern for which only partial
results are known. Similar comments hold for irreducible sets of tensor
generators.
As pointed out in Section 1.3, the representations for general func-
tions contain, in general, fewer elements then the corresponding represen-
tations for polynomial functions. Consider, for example, the tensor func-
tion (8). The irreducible generating set in the polynomial case contains
nine tensors :

I, A, A2 , B, B2 , AB+BA,
- ... - - - ( 11)

whereas in the general non-polynomial case, the irreducible generating set

contains the eight tensors {9), the last generator (11)being then redundant.
Consider the list of an arbitrary nurnber of symmetric second order
tensors:

A , A ••. A (12)
-l -2 -a
40 J.P. Boehler

The integrity basis for the list (12) contains the traces of products of
up to six tensors [7], whereas the functional basis contains the traces
only up to the order three (Table I). Similarly, the generating set in
the polynomial case contains products of up to five tensors [31, whereas
in the general case, it contains products up to only two tensors (Table
II). The complete analysis for the reductions introduced in the represen-
tations when considering general functions instead of polynomial functions
is presented in [12].

3. REPRESENTATIONS FOR NON-POLYNOMIAL ANISOTROPie SCALARAND TENSOR

FUNCTIONS

3.1 Method based on the introduction of structural tensors

The proposed method is an extension of the procedure developed in

Chapter ·2 concerning a two-dimensional constitutive equation relating
two symmetric secend order tensors. Here, we consider an anisotropic me-
dium of constitutive equation specifying a symmetric secend order tensor
T which is a function of the arguments (1), i.e. of the mechanical agen-
cies. In order to take into account the oriented internal structure of the
material, we introduce additional argument tensors ~ , ~ , ... ~ , which
-1 -2 -p
are the structural tensors:

T ~ •..• ~ )
• w_2 . . . . w_b' v_l. V-2 • • • • v-c • ~-1 • -2 (13)
F(A
- -1
• A -a • w
-2 •... A -1 -p

The relation (13) can be a constitutive equation provided it satisfies the

Principle of Isotropy of Space. One of the consequences of this principle
is that an arbi trary transformation Q of the orthogonal group 0 and applied to
both the medium and the agencies, produces the same orthogonal transforma-
tion of the material response:

VQ e: 0 F(A.,
- -1
W.,
-J
v-k' ~-")
"' FÜ.,
- -1
W.,
-J
V-k' ~- "'n) (14)

where F, ~i' ~j' ~k' ~2.are the transformed of !• ~i' ~j' ~k' h by an orthogo-
nal transformation 9·
The condition (14) means that F is an isotropic func-
Representations for Tensor Functions 41

tion with respect to its a + b + c + p argurnents.

We introduce now the invariance group S of the structural tensors ~~:

QE S (==) ~~ = ~~ 1,2, ... p. (15)

Taking into account the relation (15) into (14), we arrive at

VQ E s F(A.,
- -1
W.,
- J
'!_k'
-
~0 )
·-Tv
F(A.,
- -1
W., '!_k' ~ 0 )
- J - -Tv
(16)

For a general function ~· it can be shown that (16) constitutes a converse

inp1ication. Thus, (16) signifies that ~ is an isotropic function of the
agencies A., W., '!k' the type of anisotropy being characterized by the i~
-1 -J
variance group S of the structural tensors h. It follows that the repL·e-
sentation of the isotropic function (14) is also a representation fo~ r
considered as an anisotropic function of the agencies. For a given group
s, the set of structural tensors is not unique. The representations for
anisotropic functions so obtained are necessarily complete, but not neces-
sarily irreducible, even if the representations employed for the isotropic
functions (14) are irreducible.
Thus, two main problems arise: find the ~tructural tensors associated
with each type of anisotropy; reduce the obtained representations in apply-
ing the proposed method. This procedure is applied in Sections 3.3 to 3.5
below for the derivation of the irreducible representations for respecti-
vely general anisotropic, orthotropic and transversely isotropic tensor func-
tions of an arbitrary nurober of symmetric secend order tensors. The repre~

sentations for tensor valued functions are composed of the functional bases
and the generating sets of tensors. The functional bases constitute the
representations for the corresponding scalar-valued functions.

3.2 Generalization of the Rivlin-Ericksen method

The irreducible functional bases for general orthotropic and trans-

verse isotropic scalar-valued functions f of the complete list of argurnents
(1) have been established in [14], employing a generalization for aniso-
tropic functions of the method developed by RIVLIN and ERICKSEN (cf. Sec-
42 J.P. Boehler

tion 2.) for isotropic functions. The list of invariants obtained is too
long to be reproduced here.

3.3 General anisotropy

In the case of general anisotropy, the material symmetry group s is

reduced to:

( 17)

Consider the six linearly independent tensors:

M
-11
V
-1 • V
-1
M
-22
V
-2 • V
-2
M
-33
V
-3 • V
-3

M
-23
+ M
-32
V
-2
• V
-3
+ V
-3
• V
-2
M
-31
+ M
-13
V
-3 • V
-1
+ V
-1 • V
-3
(18)

-2 •
M
-12
+M
-21
V
-l
• V
-2
+ V V
-1

where (v , V ' V ) are three orthonormal vectors. In the (v , V

V )
- 1 -2 - 3 - 1 -2 ' -3
frame, the expressions of the tensors (18) are given by:

1 0 0 0 0 0 0 0 0
M 0 0 0 0 1 0 M 0 0 0
11 3 3
0 0 0 0 0 0 0 0 1
(19)
0 0 0 0 0 1 0 1 0
M
23
+ M32 0 0 1 M31 + M
13
0 0 0 M12 + M21 1 0 0
0 1 0 1 0 0 0 0 0

It is easy to show that the group (17) is the invariance group of the ten-
sors (18), which constitute a basis in the space of symmetric secend order
tensors. The tensors (18) are tt~ structural tensors for general anisotro-
PY·
Thus, the representation for the isotropic function

obtained from Table I and II, is a representation for F considered as a

general anisotropic function of the agencies A , A, ... A . Among the in-
-1 -2 -a
Representations for Tensor Functions 43

variants of this representation the following 6a invariants appear:

trM A. trM A. trM A.

-11-1 -22-1 -33-1
(21)
trM A. trM A. trM A.•
-23-1 -31-1 -12-1

The analysis of the explicit expressions show that the invariants (21)
are just the independent components in the (v , v , v ) frame of the agen-
(") -1 -2 -3
cies A. : trMk 0 A.
-1 - JC-1
= Ak!"' . Any scalar invariant of the A. can be expressed
-1
uniquely with respect to the invariants (21). Thus, the invariants (21)
constitute a complete functional basis. As they are functionally indepen-
dent, this functional basis is irreducible a fortiori.
Among the generators for the isotropic function (20) appear the six
tensors (18), which constitute a complete set of tensor generators. As
the tensors (18) are linearly independent, this generatingset is irreduci-
ble.
Finally, an irreducible representation for the general anisotropic
tensor function F of the agencies A. is given by:
-1

T ~ M + u M + u M + u (M + M ) +
1-11 2-22 3-33 4 -23 -32

(22)

~. ~.(trM A., trM A., trM A., trM A., trM A., trM A.)
1 1 -11-1 -22-1 -33-1 -23-1 -31-1 -12-1

i=1,2, ... a.

This result could have been established directly. Indeed, the tensor
T can always be expressed as a linear combination of the basic tensors
(18), with the components Tmn of T as coefficients. As the function F does
not verify any symmetry, the components T are arbitrary functions of the
(') mn
6a independent components Ak~ of the agencies ~i'
44 J.P. Boehler

3.4 Orthotropy

In the case of orthotropy, with privileged directions (v-1 , v-2 , v-3 ),

the material symmetry group S is given by :

s ?~±I- , S , S , S
-1 -2 -3~
l (23)

where S , S , S are the reflections with respect to the basic planes

-1 -2 -3
(v , v ), (v , v ), (v , v ) of the orthonormal privileged frame
-2 -3 -3 -1 -l -2
(v , V , V ).
-1 -2 -3
Consider the three tensors:

M
-1
V
-1
~ V
-1
M
-2
= V ~ V
-2 -2
M
-3
V
-3
~ V
-3
(24)

The expressions of the tensors (24) are given by (19). It is easy to show
that the group ( 23) is the invariance group of the tensors ( 24) :

QE S <==> QM.Qt
--1-
= M.
-1
i 1,2,3 . (25)

Thus, the tensors (24) are the structural tensors for orthotropy.
The representation for the isotropic function

T = F(A, ... A, M , M, M) (26)

- - -1 -a -1 -2 -3

obtained from Table I and II, is a complete representation for F consider-

ed as a general orthotropic function of the agencies A , ... A. But the
-a -1

representation so obtained is not irreducible. The redundant elements have

been analysed in [13,15]. The reductions are mainly based on the following
relations, resulting from the normality and orthogonality of the vectors
(v , V , V ):
-1 -2 -3

MI?
-1
= M.
-1
trM.
-1
1 i = 1,2,3
M.M. = 0
-1-J
i +j i,j = 1,2,3
(27)
(M-1-A+AM ) + (M-2-A+AM
--2
) + (M-3-A+AM
--3
) 2A
--1
trM A + trM-2-A + trM-3-A trA
-1-
Representations for Tensor Functions 45

where A is an arbitrary symmetric second order tensor.

Finally, an irreducible functional basis is composed of the invariants
given in Table III, in considering all unordered combinations of one, two
and three variables. An irreducible generating set is composed of the ten-
sor generators given in Table III, in considering all unordered combina-
tions of zero, one and two variables.
For example, the irreducible representation for T, considered as an
orthotropic function of two symmetric tensors A and B, is given by:

T et 1-M1 + et 2-M2 + et 3-M3 + et (M A+AM ) + et (M A+AM ) +

~ -1- --1 5 -2- --2

+ et 6 (M-3-A+AM
--3
) + a 7 (M-1-B+BM
--1
) + et 8 (M-2-B+BM
--2
) +

+ et 9 (M-3-B+BM ) + ll10~2 + ll Bz + et (AB+BA) ,

--3 11- 12 -- --

(28)

et. et.(trM A, trM A, trM A, trM B, trM B, trM B,

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

trM AB, trM AB, trM AB, trA 2 B, trAB 2 , trA 3 , trB 3 )
-1-- -2-- -3--

3.5 Transverse isotropy

In the case of transverse isotropy, with privileged direction ~3 , the

material symmetry group S is given by :

s (29)

where ~1 ,3 have the same definitions as in Section 3.4 and

~2 , ~ ~8 are
all rotations about the v axis.
-3
Consider the tensor:

M (30)

It is easy to show that the invariance group of M is the characteristic

group (29):
46 J.P. Boehler

Variables Functional bases Integrity bases

trM A, trM A2 , trA 3 , trM A, trM A2 , trA 3 ,

-A -1- -1- - -1- -1- -
trM A, trM A2 , trM A, trM A2 ,
-2- -2- -1- -2-
trM· A, trM A2. trM A, trM A~
-3- -3- -3- -3-

A A trM A A , trA 2A , trA A2 trM A A , trM A2A , trM AA 2

-1 ' -2 -1 -1 -2 -1 -2 -1 -2 -1-1-2 -1-1-2 -1-1-2
trM A A , 2
trM A A , trM A A , trMAA 2 ,
-2-1-2 -2-1-2 -2-1-2 -2-1-2
trM A A • A~A 2 , trM A A2 .
tr~3~1~2• trM-3-
-3-1-2 - -3-1-2

A A trA A A trM A A A , trM A A A ,

- 1' A
-2 ' -3 -1-2-3 -1-1-2-3 -1-2-1-3
trM A A A , trM A A A ,
-2-1-~-3 -2-2-1-3
trM A A A , trM A A A
-3-1-2-3 -3-2-1-3.

Generators for Generators for

Variables general functions polynomial functions

M
-1 ' M
M M
-2 ' -3 M
-1 ' M
-2 ' -3

A M A+AM , A2, M A+AM , M A2+A 2M

- -1- --1 - -1- --1 -1- - -1'
M A+AM , M A+AM , M A2+A 2M
-2- --2 -2- --2 -2- - -2'
M A+AM M A+AM , M A2+A 2M
-3- --3 -3- --3 -3- - -3

A A A A +A A M A A +A A M , M A A +A A M
-1 ' -2 -1-2 -2-1 -1-1-2 -2-1-1 -1-2-1 -1-2-1
M A A +A A M , M A A +A A M
-2-l-2 -2-1-2 -2-2-1 -1-2-2
M A A +A A M , M A A +A A M
-3-1-2 -2-1-3 -3-2-1 -l-2-3

Table III Irreducihle representations for orthotropic functions

Representations for Tensor Functions 47

QE s <==> QMQt M (31)

Thus, M is the single structural tensor for transverse isotropy.

The representation for the isotropic function

T F(A , ... A , M) (32)

- -1 -a

obtained from Table I and II, is a complete representation for ~ consider-

ed as a general transversely isotropic function of the agencies A , ... A .
-1 -a
After some reductions ([13], [15]), the irreducible representations ob-
tained are listed in Table IV.
For example, the irreducible representation of !• considered as a
transversely isotropic function of two symmetric tensors A and B, is given
by:

T ~ I + ~ M+ ~ A+ ~ B+ ~ (MA+AM) + ~ (MB+BM) +
l- 2- 3- 4- 5 -- -- 6 -- --

+ ~ (AB+BA) + ~ (A 2 B+BA 2 ) + ~ (AB 2 +B 2 A) (33)

ll -- -- 12 - - -- 13 -- - -

~. ~.(trA, trA 2 , trA 3 , trB, trB 2 , trB 3 , trAB, trA 2 B,

1 1 - - - - - - -- - -

trAB 2 , trMA, trMB,trMA 2 , trMB 2 , trMAB.

3.6 Comparison with representations for polynomial anisotropic tensor

functions

In the case of general anisotropy, the representations for polynomial

tensor functions can be derived directly by the same arguments as those
developed in Section 3.3. Thus, the irreducible representations for poly-
nomial functions are similar to those for general functions, the scalar
functions ~. in (22) being then polynomials in their arguments.
1
In the case of orthotropy and transverse isotropy, the integrity ba-
ses and generating sets for polynomial functions are given in Tables III and
48 J.P. Boehler

Variables Functional bases Integrity bases

trMA, trMA 2 . trMA, trMA 2 .

A ,A trA A , trMA A
-1 -2 -1-2 --l-2

A ,A ,A
-1 -2 -3

Generators for
Variables general Generators for
functions polynomial functions

I, M. I, M.

A A, MA+AM, A, MA+AM,
-
A2, MA 2+A 2M.
- -- --
A2 , MA 2+A 2M.

A , A
-1 -2
AA + AA , MAA + A A M, MAA
_2_1_ __ 2_1 + A A M,
_1_2 _
-1-2 -2-1 --l-2

A2A + A A2 ,
_2_1
MA 2A + A
__ 1_2 A2M,
_2_1 _
-1-2

A A2 + A2A. MA A2 + A
__ 1_2
2A M.
_2_1 _
-l-2 -2-1

A , A , A MAA A + A
__ 1_2_3 A A M, MA AA + A
__ 1_3_2 A A M,
-1 -2 -3 _3_2_1_ -2-3_1 _

MA AA + A
__ 2_1_3 A A M,
_3_1_2 _ MA AA + A
__ 2_3_1 A A M,
_1_3_2 _

MA AA + A
__ 3_1_2 A A M,
_2_1_3 _ AMAA
_1 __ 2_3 + A A MA
_3_2 __ 1

Table IV: Irreducihle representations for transversely isotropic func-

tions.
Representations for Tensor Functions 49

IV against the corresponding functional bases and generating setsfor gene-

ral functions. The invariants of the integrity bases, given here in direct
tensor notation, are equivalent to those given by ADKINS [16,17] in tensor
component notation. We have derived the generating sets by the method indi-
cated in the Chapters by Professor SPENCER (see also [12]). In the ortho-
tropic case, the functional bases and the integrity bases are similar for
a single tensor variable. But, from two tensor arguments, the integrity
bases contain an increasingly greater number of elements. For the generat-
ing sets, this difference already appears for a single tensor argument. In
the case of transverse isotropy, the two types of representations are si-
milar for a single tensor argument. From two tensor arguments, the repre-
sentations for polynomial functions contain a markedly greater number of
elements then the corresponding representations for general functions.

4. REPRESENTATIONS FOR NON-POLYNOMIAL ISOTROPIC AND ORTHOTROPIC TENSOR

FUNCTIONS IN A TWO-DIMENSIONAL SPACE

4.1 Isotropie functions

Consider the isotropic tensor function:

T = F(A , A , •. . A ) (34)
- - -1 -2 -a

where T and the A. are symmetric second order tensors in a two-dimensio-

-1
nal space. A complete representation for the function (34) is obtained
from Table I and II for the three-dimensional case:

T = ~0 I
-
+ ~.G.
1-1
(35)

where I is the two-dimensional identity tensor, the G. 's are the genera-
-1
tors:

. ...l
A., A~,
~k~~ + ~~~k' A-k-~
2A + A Ak'
-J -J -~-
(36)
~~ Ak + A A2
-k-~
k < ~ j, k, ~ = 1, 2,
50 J.P. Boehler

and ~0 , ~i are arbitrary scalar-valued functions of the invariants:

trA., trA~, trA~, trAkA 0 , trAk2 A0 ,

-J -J -J - _,_ - _,_
(37)

Obviously, therepresentation (35) is not irreducible. It can be reduced by

the application ofthe generalization fortwo-dimensional second orderten-
sors of the Cayley-Hamilton theorem, due to RIVLIN [18].
The Cayley-Hamilton theorem for a two-dimensional second order tensor
is given by:

0 (38)

If in (38) we replace ~ by ~+ A~, whereA is an arbitrary scalar, and

equate the coefficient of A to zero, we obtain:

AB + BA - (trA)B - (trB)A + (trA trB - trAB)I 0 (39)

From the theorem (39), we obtain the following relations for the gene-
rators (36):

1
A~ = -(trA~- tr 2 A.)I + (trA.)A.,
-J 2 -J -J- -J-J

(tr~k)~~ + (tr~~)~k +
(40)

- ~tr~~tr~k - [tr~~tr~~~k- t tr~(tr 2 ~~ - trA~)]~~·

(tr~~)~k + (tr~k)~~ +

- ~tr~ktr~~ - [tr~ktr~k~~ - ttr~~(tr 2 ~k - tr~k]~!

and the following relations for the invariants (37):
Representations for Tensor Functions 51

3
trA~
-J 2 trA.
-J
trA:
-J

(41)

Taking into account the relations (40) and (41), we obtain from (36)
and (37) the following irreducible representation for the isotropic func-
tion (34):

~.
~
= ~.(trA.,
~ -J
trA:, trAkA 0
-J - ~
), (42)

k < R, j,k,R. = 1,2, ... a.

It is worthwhile to point out that the representation (42) obtained is qua-

si-linear.

4.2 Orthotropic tensor functions

In the case of two-dimensional orthotropy, with privileged directions

(v , v ), the material symmetry group of transformations is given by:
- 2 - 3

S = ~±I,
( - -2 -3)
S , S l (43)

where ~2 and ~3 are the reflections with respect to the

3 axes. It ~2 and ~

is easy to show that S is the invariance group of the structural tensor:

M (44)

Thus, the representations of the isotropic function:

52 J.P. Boehler

T F(A , A , .. . A , M) (45)
- -l -2 -a -

is a representation for F, eonsidered as an orthotropie funetion of the

ageneies A .•
-1
Taking into aeeount the relations:

trM 1 (46)

we obtain direetly from (42) the following irredueible representation:

T + Ql a+z-a
A

Ql.(trA., trA:, trMA., tr~k~ 0 ), (47)

1 -J -J --J "

k < 1', j 'k, 1', 1,2, ... a

whieh is a quasi-linear form.

REFERENCES

1. WINEMAN, A.S. and PIPKIN, A.C., Material Symmetry Restrietions on

Constitutive Equations, Arch. Rat. Mech. An., 17 (1964) : 184-214.

2. PIPKIN, A.C. and WINEMAN, A.S., Material Symmetry Restrietions on

Non-Polynomial Constitutive Equations, Arch. Rat. Mech. An., 12
(1963) : 420-426.

3. SPENCER, A.J.M., Theory of Invariants, in Continuum Physics, ed. by

C. Eringen, Aeademie Press, (1971) : 239-353.

4. WANG, C.C., On Representations for Isotropie Funetions, Part I and II,

Arch. Rat. Mech. An., 33 (1969) : 249-287.

5. WANG, C.C., A New Representation Theorem for Isotropie Funetions,

Part I and II, Arch. Rat. Mech. An., 36 (1970) : 166-223.

6. WANG, C.C., Corrigendum, Arch. Rat. Mech. An., 43 (1971) : 392-395.

7. SMITH, G.F., On Isotropie Integrity Bases, Arch. Rat. Mech. An., 18

(1965) : 282-292.
Representations for Tensor Functions 53

8. SMITH, G.F., On a Fundamental Error in Two Papers of C.C. WANG, Arch.

Rat. Mech. An., 36 (1970) : 161-165.

9. SMITH, G.F., On Isotropie Functions of Symmetrie Tensors, Skew-

Symmetric Tensorsand Vectors, Int.J.Engng. Sei., 19 (1971) : 899-916.

10.RIVLIN, R.S. and ERICKSEN, J.L., Stress-Deformation Relations for

Isotropie Materials, J. Rat. Mech. An., 4 (1955) : 323-425.

11.BOEHLER, J.P., On Irreducihle Representations for Isotropie Scalar

Functions, ZAMM, 57 (1977) : 323-327.

12.BOEHLER, J.P.,Lois de Comportement Anisotrope des Milieux Continus,

Journal de Mecanique, 17,2 (1978) : 153-190.

13.BOEHLER, J.P., A Simple Derivation of Representations for Non-

Polynomial Constitutive Equations in Some Cases of Anisotropy, ZAMM,
59 (1979) : 157-167.

14.BOEHLER, J.P.,Functional Bases for Orthotropie and Transverse Isotropie

Non-Polynomial Funetions, (in preparation).

15.BOEHLER, J.P. and RACLIN, J.,Representations Irreduetibles des Fonetions

Tensorielles Anisotropes Non-Polynomiales de Deux Tenseurs Symetriques,
Arch. Mech. Stos., 19,3 (1977) : 431-444.

16.ADKINS, J.E., Further Symmetry Relations for Transversely Isotropie

Materials, Arch. Rat. Mech. An., 5 (1960) : 263-274.

17.ADKINS, J.E., Symmetry Relations for Orthotropie and Transversely Iso-

tropie Materials, Arch. Rat. Mech. An., 4 (1960) : 193-213.

18.RIVLIN, R.S., Further Remarks on the Stress-Deformation Relations for

Isotropie Materials, J. Rat. Mech. An., 4 (19~) : 681-702.
 Chapter 4

ANISOTROPie LINEAR ELASTICITY

J.P. Boehler
University of Grenoble

1. INTRODUCTION

Consider a symmetric second order tensor T which is a function F of

a symmetric secend order tensor D. If F is a transversely isotropic
- -
function of D, its irreducible representation is obtained from Tab1e IV
of Chapter 3:

T a I + a M + a D + a (MD+DM) + a D2 + a (MD 2 +D 2 M)
0- l- 2- 3 -- -- 4- 5 -- - -
(1)
a.
1
= a.(trD,
1 -
trD 2 , trD 3 , trMD, trMD 2 )

where M is the structural tensor:

M V ® V (2)
-3 -3

~3 being the privileged axis of transverse isotropy.

56 J.P. Boehler

If F is an orthotropic function, its irreducible representation is

obtained from Table III of Chapter 3:

T a M + a M + a M + a (M D+DM ) + a (M D+DM ) +
1-1 2-2 3-3 4 -1- --1 5 -2- --2

+ a (M D+DM ) + a D2 • (3)
6 -3- --3 7-

a. a.(trM D, trM D, trM D, trM D2 , trM D2 , trM D2 , trD 3 )

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

where ~1 , ~2 , ~3 are the structural tensors:

M V ® V M M V ® V (4)
-1 -1 -1 -2 -3 -3 -3

(v , v , v ) being the privileged orthonormal frame of orthotropy.

-1 -2 -3
Equations (1) and (3) can describe any material response which is
expressible by an explicit relation between two symmetric second order
tensors. Let T be the stress tensor and D the deformation tensor. The
representations (1) and (3) are then the most general invariant forms of
non-linear elasticity for the corresponding types of anisotropy.
The aim of this Chapter is to present the application of the theory
of representations for tensor functions to the invariant formulation of
anisotropic linear constitutive equations relating two symmetric second
order tensors; for this class of mechanical behaviors, a potential func-
tion for the stress tensor may or may not exist. Thus, in the case of
elasticity, without entering into the question concerning the existence
of a strain-energy function for all elastic materials, we first analyze
elastic materials (of Cauchy-elastic materials), for which the existence
of a strain-energy function is not assumed, and then hyperelastic mate-
rials (or Green-elastic materials), for which a strain-energy function
exists. The obtained invariant formulations involve only second order ten-
sors in linear combination. For hyperelastic materials, the transitions
from the invariant formulations to the classical formulations will be
given explicitly in the cases of transverse isotropy and orthotropy
(cf. [ 1], [2]); the case of general anisotropy is analyzed in [3]. Kine-
Anisotropie Linear Elasticity 57

matic constraints result in further restrictions on the constitutive

equations for anisotropic linear elasticity; such restrictions are dis-
cussed in Chapter 10, Section 2, by Professor SPENCER.
In the Sections below, we use the following derivatives:

atrn atrn 2 atrn 3

I 2n 3n 2
an an an
(5)
atrM.n atrM.n 2
-1- -1-
M. M.n+nM. i 1,2,3
an -1 an -1- --1

2. TRANSVERSE ISOTROPY

2.1 Invariant formulation

Consider a linear elastic solid, which is transversely isotropic

whith respect to an undistorted state, which will be taken as the refe-
rence configuration. The linear restriction of the representation (1) is
given by:

T a I + a M + a n + a (Mn+nM)
0- 1- 2- 3 -- --

a a + b trn + c trMn
0 0 0 0
(6)
a1 a 1 + b 1 trn + c 1 trMn

a
2
a2 a
3
a3

Thus, the linear elastic behavior of a transverse1y isotropic solid

depends on eight material constants. In an undistorted state, the elas-
tic solid is subjected to the following transversely isotropic stress:

T
-u
a I
o-
+ a M
1-
(7)

Suppose that the undistorted state is a natural state, i.e. the

stress T is zero:
-U
58 J.P. Boehler

D - 0 => T
-U
a I + a M
0- 1-
0 (8)

From relation (8), we obtain:

trT
-u = 3a +a
0 1
0
=> a a1 0 (9)
trMT
--u
= a 0 +a 1 0
0

Thus, when the solid admits a natural state, the transversely isotropic
linear elastic behavior depends on six material constants.
Suppose now that the elastic solid is a hyperelastic material, i.e.
its mechanical response T is the derivative of a scalar-valued strain-
energy function W:

aw
T (10)
an
The representation of the strain-energy function is obtained from Table IV
of Chapter 3:

w W(trD, trMD, trD 2 , trMD 2 , trD 3 ) ( 11)

Introducing (11) in (10) and using the expressions (5), we obtain

aw + ~ M aw
T atrD ! atrMD - + 2 atrD 2 ~ +
(12)

Identifying (12) with the linear form (6) of the representation (1), we
obtain:

aw b trD + c trMD aw aw
atrD = o o 0
atrD 2 atrD 3

(13)
aw
a
3
Anisotropie Linear Elasticity 59

The derivatives (13) are subjected to the following integrability condi-

tion:

a aw => c b . (14)
atrD <atrMD) 0 1

Thus, the linear hyperelastic behavior of a transversely isotropic solid

depends on five independent material constants.
Finally, the invariant form of transverse isotropic linear hyper-
elasticity is given by [2]:

T = (b trD+b trMD)I + (b trD+c trMD)M + a D + a (MD+DM) (15)

- 0 - 1 -- - 1 - 1 -- - 2- 3 ~- --

Integrating the partial derivatives (13) and taking into account

relation (14), we obtain the following invariant form for the strain-
energy function:

w l2 bo tr 2 D + 1 c tr 2 MD + b trMDtrD + 1 a trD 2 + a trMD 2

21 1 22 3
(16)

It is worthwhile to point out that the expression (15) furnishes a

formulation of transverse isotropic hyperelasticity which involves only
second order tensors in linear combination. The transition to non-linear
cases (1) is direct and does not necessitate expansions of the strain-
energy function or introduction of higher order tensors, as it is the
case in the classical formulation ([4], [5]).

2.2 Classical formulation

In the classical formulation, the law of transverse isotropic linear

hyperlasticity is written as follows:

T A D (17)

where A is a fourth order tensor, specifically the tensor of transversely

::
isotropic constants in the linear case. The expression of A in any
::
 60 J.P. Boehler

(•. •, v-3 ) privileged orthonormal frame is given by:

a f d 0 0 0
f a d 0 0 0
d d c 0 0 0
A (18)
0 0 0 g 0 0
0 0 0 0 g 0
0 0 0 0 0
1
-(a-f)
2

where a, c, d, f, g, are five independentmaterial constants.

The transition from the invariant formulation (15) to the classical
formulation (17)-(18) has been established in [2]:

b f b d-f c c+a-2d-4g
0 1 1
(19)
a a-f a3 2g-a+f
2

Thus, an alternative invariant form of transverse isotropic linear hyper-

elasticity, employing the material constants of the matrix (18), is given
by:

- = [f trD+(d-f)trMD]I
T
... -- - + [ (d-f)trD+(c+a-2d-4g)trMD]M +
- .........
(20)
+ (a-f)D + (2g-a+f)(~~+~~)

3. ORTHOTROPY

3.1 Invariant formulation

Consider a linear elastic solid, which is orthotropic with respect

to an undistorted state. The linear restriction of the representation (3)
is given by:
T a. M + a. M +
2- 2
a. M + a. (M D+DM ) + a. (M D+DM ) + a. 6 (M_3 D+DM
___ 3 ) ;
1- 1 3- 3 4 -1- --1 5 -2- --2
Cl.
1
a 1 + b 1 trM D + c 1 trM D + d 1 trM D
- 1- - 2- - 3-
Cl. = a + b trM D + c trM D + d trM D (21)
2 2 2 2
- 1- 2 -2- -3-
Cl. a 3 + b trM D + c trM D + d trM D
3 3 3- 1- 3 - 2- - 3-
Cl. a4 Cl. a5 Cl. a6
4 5 6
Anisotropie Linear Elasticity 61

Thus, the linear elastic behavior of an orthotropic solid depends on 15

material constants. In an undistorted state, the elastic solid is sub-
jected to the following orthotropic stress:

T
-u
a M
1-1
+ a 2-2
M +a M
3-3
(22)

If the undistroted state is a natural state, we obtain:

D - 0 => T
-u
M + a 2-2
a 1-1 M + a 3-3
M 0 (23)

The relation (23) leads to:

trM T a 0 trM T a 0 trM T a 0 . (24)

-1-U 1 -2-U 2 -3-U 3

Thus, when the solid admits a natural state, the orthotropic linear elas-
tic behavior depends on 12 material constants.
In the case of an orthotropic hyperelastic material, the represen-
tation of the scalar-valued strain-energy function is obtained from
Table III of Chapter 3:

W = W(trM D, trM D, trM D, trM D2 , trM D2 , trM D2 , trD 3 ) (25)

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

Introducing (25) in (10) and employing the expressions (5), we obtain:

T aw M + aw M + ~- M aw ( )
3trM D -1 3trM D -2 tltrM D -3 + tltrM D2 ~1~+~~1 +
-1- -2- 3- -1-
(26)

Identifying (26) with the linear form (21) of the representation (3), we
obtain:
62 J.P. Boehler

aw aw
b 1 trM D + c 1 trM D + d 1 trM D a4
atrM D -1- -2- -3- atrM D2
-1- -1-

aw aw
= b 2 trM D + c trM D + d 2 trM D a5 (27)
atrM D -1- 2 -2- -3- atrM D2
-2- -2-

aw aw aw
b 3 trM D + c 3 trM D + d 3 trM D a6 atrD 3 = 0.
atrM D -1- -2- -3- atrM D2
-3- -3-

The derivatives (27) are subjected to the following three integrability

conditions:

aw aw aw aw
atrM D <atrM-2-D) atrM D <atrM D)
=> c1 b2
- 1- -2- - 1-

aw aw aw aw
atrM D <atrM D) atrM D <atrM D)
=> b3 d1 (28)
-1- - 3- - 3- - 1-

aw aw aw aw
atrM D ( atrM D) atrM D <atrM D) => d
2
c
3
- 2- - 3- - 3- - 2-

Thus, the linear hyperelastic behavior of an orthotropic solid depends

on nine independent material constants.
Finally, theinvariant formulation of orthotropiclinear hyperelasti-
city is given by [1]):

T = (b trM D + b trM D + d trM D)M + (b trM D+c trM D+c trM D)M +
1 -1- 2 -2- 1 -3- -1 2 -1- 2 -2- 3 -3- -2

+ (d trM D+c trM D+d trM D)M + a 4 (M-1-D+DM

--1
) + (29)
1 -1- 3 -2- 3 -3- -3

+ a (M D+DM ) + a (M D+DM )
5 -2- --2 6 -3- --3

The integration of the partial derivatives (27) results in :

W= 21 h 1 tr 2 ~ 1 ~ + 21 c 2 tr 2 ~ 2 ~ + 21 d 3 tr 2 ~ 3 ~ + c 3 tr~ 2 ~tr~ 3 ~ + d 1 tr~ 3 ~tr~ 1 ~ +

(30)
+ b 2 trM-1-DtrM-2-D + a4 trM-1-D2 + a5 trM-2-D2 + a6 trM-3-D2
Anisotropie Linear Elasticity 63

Comments on the invariant forms (29) and (30) are similar to those given
in Section 2 .1.

3.2 Classical formulation

In the classical formulation, the expression of the hyperelastic

fourth order tensor A in the privileged orthonormal frame (v , v , v )
:: -1 -2 -3
is given by:

a f e 0 0 0
f b d 0 0 0
e d c 0 0 0
T AD A (31)
- 0 0 0 g 0 0
0 0 0 0 h 0
0 0 0 0 0 j

where a, b, c, d, e, f, g, h, j, are nine independentmaterial constants.

The transition from the invariantformulation (29) to the classical
formulation (31) has been established in [1], [3]:

b
1
a+2g-2h-2j c
2
b+2h-2j-2g

d c+2j-2g-2h a4 h+j-g a j+g-h (32)

3 5

a
6
g+h-j c
3
d d1 = e b
2
f

Thus, an alternative invariant form of orthotropic linear hyperelasticity

is given by:

T [(a+2g-2h-2j)trM D+f trM D+e trM D]M +

-1- -2- -3- -1

+ [f trM D+(b+2h-2j-2g)trM D+d trM D]M

-1- -2- -3- -2
+ (33)

+ ~ trM D+d trM D+(c+2j-2g-2h)trM D]M +

-1- -2- -3- -3

+ (h+j-g)(M D+DM ) + (j+g-h)(M D+DM ) + (g+h-j)(M D+DM ) .

-1- --1 -2- --2 -3- --3
64 J.P. Boehler

4. ISOTROPY

Consider an elastic solid, which is isotropic with respect to an

undistorted state. The most general invariant form of non-linear isotro-
pic elasticity is given by the following irreducible representation
obtained from Table I of Chapter 3:

T a I
0-
+ a D + a D2
1- 2-
(34)
a. a.(trD, trD 2 , trD 3 )
1 1 - - -

The linear restriction of the representation (34) is given by:

T a I
0-
+ a 1-D
(35)
a
0
a
0
+ b0 trD
a
1
a1.

Thus, isotropic linear elasticity depends on three material constants.

In an undistorted state, the elastic solid is subjected to the following
isotropic stress:

T a I (36)
-U 0-

If the undistorted state is a natural state, we obtain:

D - 0 => T
-U
a I
0-
0 => a
0
0 (37)

Thus, if the solid admits a natural state, the isotropic linear elasti-
city depends on two material constants:

T (b trD)I
0
+ a 1-D (38)

The invariant form (38) is equivalent to the classical form:

T (AtrD)I + 211D (39)

Anisotropie Linear Elasticity 65

where A and ~ are Lame's coefficients.

If the solid is a hyperelastic material, the strain-energy function
is given by:

w l2 b tr 2 D + 21 a 1 trD 2
0 -
(40)

In the isotropic case, the assumption of the existence of a strain-energy

function introduces no integrability conditions.

REFERENCES

1. BOEHLER, J.P., A Simple Derivation of Representations for Non-

Polynomial Constitutive Equations in Some Cases of Anisotropy,
ZAMM, 59 (1979): 157-167.
2. BOEHLER, J.P., Sur les Formes Invariantesdans le Sous-Graupe
Orthotrope de Revolution ... , ZAMM, 55 (1975): 609-611.
3. BOEHLER, J.P., Contributions Theoriques et Experimentales a l'Etude
des Milieux Plastiques Anisotropes, These de Doctorat es Sciences,
Grenob1e, 1975.
4. GOLDENBLAT, I.I., Same Problems of the Mechanics of Deformable Media,
Noordhoff, Groningen, 1962.
5. LEHMANN, Th., Anisotrope Plastische Formänderungen, Rheol. Acta,
3 (1964): 281-285.
 Chapter 5

VIELDING AND FAlLURE

OF TRANSVERSEL Y ISOTROPIC SOLIDS

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

A suitable framework allowing to describe yielding and failure of

anisotropic materials with the required generality and pertinence is fur-
nished by the theory of representations for tensor functions. In this
Chapter, invariant formulations of the plastic behavior and failure of
transversely isotropic solids are developed. For other material symme-
tries, the corresponding constitutive relations can be derived in a simi-
lar manner.
Employing the results of Chapter 3, a general theory of the rate
independent flow of transversely isotropic materials is developed. Gene-
ral invariant forms of both the flow law and the yield criterion are
obtained. The type and the number of form-invariant tensor generators and
of scalar invariants entering the flow law, as well as the rules for cons-
tructing failure criteria as scalar-valued functions of invariants of the
stress and structural tensors are established. The restrictions imposed
by the material symmetries are then automatically satisfied.
68 J.P. Boehler

The essential features of the anisotropic flow law and their conse-
quences in experimental investigations of the mechanical behavior of aniso-
tropic solids are analyzed in Section 3. Specific forms of failure crite-
ria are proposed for glass-fiber reinforced composites and compared with
experimental data (Section 4).
A simplified theory employing a fourth order tensor of plastic ani-
sotropy and a tensorially linear description of the material response is
discussed in Section 5, in order to arrive at a warkable form of the cons-
titutive relation and at simple forms of the failure criteria for trans-
versely isotropic solids.

2. GENERAL THEORY

Consider a transversely isotropic solid with privileged direction

~3 • For further investigations of the plastic behavior, we restriet our
attention to the general invariant form of the constitutive relation for
a rate type material. Let T denote the stress tensor and D the rate of
deformation tensor. In the case of transverse isotropy, the irreducible
representation of T considered as a tensor function of D is obtained from
Table IV of Chapter 3:

T a I + a M + a D + a (MD+DM) + a D2 + a (MD 2 +D 2 M)
0- 1- 2- 3 -- -- 4- 5 -- - -
(1)
a. a. (trD, trD 2 , trD 3 , trMD, trMD 2 )
1 1

where M is the structural tensor for transverse isotropy:

M V ® V (2)
-3 -3

Representation (1) can describe perfectly plastic behavior when it

satisfies the condition of homogeneity of order zero with respect to time
([1], [2], [3]). The only time involving variable being the rate of de-
formation tensor D, the constitutive equation (1) is necessarily homoge-
neaus of order zero with respect to D
Vielding and Failure of Transversely Isotropie Solids 69

aT
o if a~ t= ~ ( 3)

The homogeneity condition (3) results in the following equations of Euler

for the scalar functions u. in (1):
1

au.
1
E. + nu. 0
aE. 1 1
J

~~ ~0, 1
where: n = when i 2, 3 j 1,2,3,4,5, (4)
4, 5

In (4), the following notation is used for the arguments of the functions
u.:
1

l I 2 l I 3
El trD E2 tr D2 E3 tr D3,
l I 2 (5)
E4 trMD E5 tr MD 2 ,

The solutions of equations (4) are:

1
u P(p,q,r,s) ul Q(p,q,r,s) u2 R(p,q,r,s,)
0 E2
( 6)
1 1 1
u3
E2
s(p,q,r,s); u 4 E2 T(p,q,r,s,);u 5 E2 u(p,q,r,s,)
2 2

where P, Q, R, s, T and U are arbitrary scalar-valued functions of the

following four dimensionless kinematic variables:

El E3 E4 E5
p q r = s = (7)
E E E2 E2
2 2

Introducing the solutions (6) into the representation (1), we obtain

the general invariant form for the plastic constitutive law of a trans-
versely isotropic solid:
(8)
ß2 ß3 ß4 ß
T ß0~ + ß /_I + E D + E (MD+DM) + E2 D2 + E~ (~~ 2 +~ 2 ~) ß. =ß.(p,q,r,s)
1 1
2 2 2 2
70 J.P. Boehler

The tensor form (8) is equivalent to the system of five scalar rela-
tions which specify the following five invariants of the functional basis
of T in the transversely isotropic case:

trT trT 2 , trT 3 , trMT (9)

These relations are established in calculating the tensors T 2 , T3 , MT+TM

_.... - -
and MT 2 +T 2 M from the tensor relation (8) and then taking the traces of
the obtained expressions. Some reductions are obtained by the application
of the generalization for 3 three-dimensional second order tensors of the
Cayley-Hamilton theorem, due to RIVLIN [4].
The Cayley-Hamilton theorem for a three-dimensional second order
tensor is given by:

0 (10)

If in (10) we replace ~ by ~+A~+~~. wher A and ~ are arbitrary scalars,

and equate to zero the coefficient of A~, we obtain the folloqing genera-
lization of (10):

ABC + ACB + CAB + BAC + BCA + CBA +

- (trBC-trBtrC)A - (trCA-trCtrA)B - (trAB-trAtrB)C +

(11)
- trA(BC+CB) - trB(CA+AC) - trC(AB+BA) +

(trAtrBtrC-trAtrBC-trCtrAB-trBtrAC+trABC+trCBA)I - 0

The method allowing the expressions of the invariants (9) to be

reduced by application of the identity (11) is presented in [1]. Finally,
after reduction, we obtain the following system of scalar relations:
trT g 1 (p,q,r,s )
trT 2 g 2 (p,q,r,s )
trT 3 g 3 (p,q,r,s ) (12)
trMT g 4 (p,q,r,s )
trMT 2 g 5 (p,q,r,s )
Vielding and Failure of Transversely Isotropie Solids 71

When the four dimensionless kinematic variables (7) are independent,

their elimination from the five relations (12) results in the most general
invariant form of the yield criterion for a transversely isotropic solid:

0 (13)

Singular cases of yield loci correspond to the case where the dimension-
less kinematic variables (7) arenot independent, i.e. when kinematic
constraints are imposed on the motion of the solid.
In the general form (13) of the yield condition, the independent
variables involve, in addition to the isotropic invariants trT, trT 2 ,

trT 3 , two mixed invariants tr~~· tr~~ 2 ,which account for the anisotropic
character of the material behavior.They specify the orientation of the
principal directions of the stress tensor with respect to the privilegcd
axis of the transversely isotropic material. In an arbitrary privileged
orthonormal frame (·, v ), the mixed invariants are expressed by:
-3

trMT T (14)
33

The general invariant form (13) is of importance in the sense that it

specifies the minimal number and the type of independent variables to be
observed in experiments.
In order to derive the general invariant form of the flow law, we
invert the plastic constitutive law (8). The method consists in calculat-
ing from (8) the expressions of the tensors:

(15)

The above-mentioned generalization (11) of the Cayley-Hamilton theorem

allows the obtained expressions to be reduced; the employed method is
presented in [1]. After reduction, the tensors (15) are expressed with
respect to the tensors I, M, D, D2 , MD+DM and MD 2 +D 2 M, which are the ten-
sor generators of the constitutive equations (8). The elimination of
D2 , MD+DM and MD 2 +D 2 M between the reduced expressions of the tensors (15)
results in the constitutive equations (8) solved with respect to D:
72 J.P. Boehler

D
~ I + ~ M+ ~ T + ~ (MT+TM) + ~ T2 + ~ (MT 2 +T 2 M)
0- 1- 2- 3 -- -- 4 5 -- - -
(16)
~.
1
= ~.(p,q,r,s)
1

In (16), the scalar-valued functions ~. can also be expressed as functions

1
of the basic stress invariants (9), since the dimensionless kinematic
variables p,q,r, and s are expressed with respect to the invariants (9),
when eliminating them from relations (12). However, the invariants (9) are
no langer independent variables, as they must verify the yield criterion
(13), which is the condition for plastic motion. Thus, in (16) the scalar-
valued functions ~. can be expressed as:
1

~. = ~.(trT, trT 2 , trT 3 , trMT, trMT 2 )

1 1 - - - --

with the restriction (13): ( 17)

Relation (16) constitutes the most general invariant form of the

plastic flow law for a transversely isotropic solid. By (16), the rate of
deformation tensor is specified only to within a scalar multiplier. The
presence in (16) of tensor generators involving the structural tensor M
results, in general, in non-coincidence of the principal directions of
T and D, exceptwhen the principal directions of D coincide with that of
- -
M, i.e. with (·, ·, v ). This property is further analyzed in Section 3.
-3
It is worthwhile to point out that once the scalar-valued functions
ßi of the invariant form (8) for the plastic constitutive law are known,
both the generalform (13) of the yield condition and the generalform
(16) of the flow law are specified. Thus, the flow law and the yield
condition are interrelated. However, in this generalapproach by the
theory of representations for tensor functions to the formulation of the
plastic behavior of solids, the flow law is not necessarily associated
with the yield condition by the rule of plastic potential. Theories em-
ploying theplastic potential result in restricted forms of the flow law,
Vielding and Failure of Transversely Isotropie Solids 73

whatever generalflow potential is assumed. This point is further discus-

sed in Chapter 14 by Professor BETTEN.
Further details, as well as the analysis of incompressibility and
plane strain, are given in [1-3].
A quasi-linear form of the representation (1) is obtained by neglect-
ing the generators involving D2 , hence:

T
(18)
a.
1

A similar procedure leads to the same general invariant form (13) for
the yield criterion and to the following invariant quasi-linear form f~r

the flow law:

D
= ~0-I + ~
1-
M+ ~
2-
T + ~
3
(MT+TM)
-- --

(19)
~. ~.(p,q,r,s
1 1

~. = ~.(trT, trT 2 , trT 3 , trMT, trMT 2 )

1 1 - - - --

f( trT, trT 2 , trT 3 , trMT, trMT 2 ) = 0

In this case, the explicit calculation necessary to derive (13) and (19)
from the quasi-linear form (1) and the homogeneity condition (3) is given
in [3].

3. PLASTIC DEFORMATIONS IN UNIAXIAL AND TRIAXIAL TESTS

In this Section, we analyze the essential feature of the anisotropic

flow law (16), i.e. the non-coincidence of the principal directions of
the stress and rate of deformation tensors in "off-axis" specimens, as
well as the consequences on the plastic deformations produced in uniaxial
and triaxial tests.
We consider first triaxial tests consisting of axial compressions or
74 J.P. Boehler

tractions a on oriented specimens that are subjected to confining pres-

n
sures p. We denote by 8 the angle between the axis (e ) of the specimen
- 3
and the privileged direction v of the material (Fig.1).
-3

0
p

p p p p
~ p

e:v

Fig.1 - Triaxial tests on transversely isotropic materials

If we assume that stresses are prescribed during the test, the stress
tensor T is defined in the axes of the specimens by:

p 0 oI
T 0 p
I
0 0 00 n (v ,e ,e )
(20)

-1 -2 -3

For the stress (20), we obtain the following components for the rate of
deformation tensor, as obtained from the law (16) and expressed in the
principal frame (v ,e ,e ) of the stress tensor:
-1 -2 -3

D
11

D2 2
- - - = '!' + '!' 1 sin 2 8 + ('!' 2 +2'!' 3 sin 2 8)p + ('!' +2'!' sin 2 8)p 2
0 4 5
/trD 2
(21)
D
33
- - - = '!' + '!' cos 2 8 + ('!' +2'!' cos 2 8)o + ('!' +2'!' cos 2 8)o 2
0 1 2 3 n 4 5 n
/trD 2

D
23
['!'
1
+'!' (a +p) + '!' (o 2 +p 2 )]sin8cos8
3 n 5 n
D
3 1
D
l 2
o.
Vielding and Failure of Transversely Isotropie Solids 75

One can conclude that the principal directions of the stress tensor T
and the rate of deformation tensor D coincide if, and only if:

[1 1 + ' 3
(o n +p) + ' s (o n2 +p 2 )]sin8cos8 0 (22)

Thus, for the so-called "in-axis" tests, i.e. when the privileged direct-
ion v ofthe material coincide with one of the principal directions of
-3
the stress tensor (8 = 0 or TI/2), the principal directions of T and D
are the same. For "off-axis" tests (8 f. 0 and TI/2), the principal direct-
ions of T and D cannot coincide in general.
In fact, in a standard compression test under confining pressure,
mixed boundary conditions are prescribed, since D rather than T = o
33 33 n
is enforced in the axial direction e of the specimens. Thus, on the ends
-3
of the specimens, the principal directions of the imposed rate of defor-
mation tensor are the axes of the specimens. On the lateral sides, the
principal stresses T 11 = T 22 =p are prescribed, without shear stress. Thus,
on the lateral sides of the specimens, the principal directions of the im-
posed stress tensor are also the axes of the specimens. Now, if the deve-
loped stress and kinematic fields were homogeneous, the axes (v-1 , e-2 , e-3 )
of the specimens would be the principal diructions of both the stress ten-
sor and the rate of deformation tensor. As this is impossible in general
for off-axis tests, the developed stress and kinematic fields in specimens
of transversely isotropic material for 0 < 8 < TI/2 are heterogeneous; in
fact, a shear stress T 2 3 f. 0 appears on the contact of the specimens with
the platens and the specimens tend to assume a S-shape.
Experimental evidence of the S-shaped deformation of off-axis speci-
mens of transversely isotropic materials subjected to standard compres-
sions under confining pressure is given in ([2], [5], [6]) for a strati-
fied rock (diatomite). Fig.2 shows an undeformed specimen,which was drill-
ed out at the orientation 8 = 60°; Fig.2b shows the deformed shape of
the specimen after axial compression (o 13.2 MPa) under a confining
n
pressure p = 6 MPa.
The same reservations apply to other tests and particularly to the
standard uniaxial tensile tests, where the ends of the specimens are
76 J.P. Boehler

a) b)
Fig.2 - Nonhomogeneity of deformation in compression tests
under confining pressure on a stratified rock (after
[2]); a) specimen before loading ; b) specimen after
unloading.

rigidly clamped and subjected to a translation parallel to the axis of

the specimens, without allowing rotation. The standard uniaxial test on
anisotropic materials is analyzed in ([7], [8]). The S-shaped deformation
of off-axis specimens is studied numerically in ([9], [10]) for composite
materials and analyzed experimentally in ([8], [9]) for rolled sheet-
steel.
As standard tests on anisotropic materials are generally interpreted
on the assumption of homogeneity for the developed stress and kinematic
fields, the experimental data obtained for "off-axis" specimens are erro-
neous. This indicates a need to imagine new experimental techniques, sui-
table for testing the mechanical behavior of anisotropic solids.
In order to avoid the S-shape effect in standard uniaxial tensile
tests on off-axis specimens, a new testing procedure has been proposed
in [8], [10], [11]. The numerical analysis for composite materials ([9],
[10]) and the experimental results obtained for rolled sheet-steel ([8],
[9], [10]) indicate that the new testing device allows homogeneaus kine-
matic fields to be obtained in off-axis specimens and thus leads to reliable
measurement of the mechanical properties of anisotropic solids (see also
Yielding and Failure of Transversely Isotropie Solids 77

Chapter 7, Section 4.1).

Consider now a hydrostatic pressure test:

T pl (23)

The non-vanishing components of the rate of deformation tensor, as ob-

tained from the law (16) and expressed in a privileged frame (·, ·, v)
-3
of the material, are the following:

Dl l D2 2
'jl + 'jl p + 'jl p2'
0 2 4
ltrD 2 ltrD 2
(24)
D3 3
'jl + 'jl l + ('ji 2 +2'jl 3 )p + ('ji 4 +2'jl 5 )p 2
0
/trD 2

D2 3 D3 l Dl 2
0

The rate of volume change is given by:

trD
3'jl + 'jl + (3'j1 +2'jl )p + (3'j1 +2'jl )p 2 (25)
0 l 2 3 4 5

One can conclude that the principal directions of the rate of deformation
tensor D coincide with the privileged directions (·, ·, v) of the trans-
- 3
versely isotropic material. Since D3 3 'i Dl l D22 , the deviatoric part of
D is different from zero; this constitutes an essential difference from
the behavior under hydrostatic pressure of isotropic materials.
An experimental study has been developed in [12] in order to show
that yieldingand hardening of a soft rock occur under hydrostatic pressure
and that the directional properties of an anisotropic material result in
a non-isotropic deformation under isotropic pressures. The material used
was diatomite, a soft stratified rock. Cylindrical specimens were drilled
out from blocks in seven different directions, oriented at 8 0°, 15°,
30°, 45°, 60°, 75o and 90° with respect to the privileged axis v of the
- 3
78 J.P. Boehler

material, which is the normal to the strata.

In a first series of experiments, virgin specimens were subjected
to monotically increasing isotropic stress, up to the required intensity
and then unloaded. In Fig.3, the irreversible deformations öL./1. in each
1 1
v. direction are plotted against the hydrostatic pressure. Marked perma-
-1
nent deformation occurs between 10 and 15 MPa; in the isotropic plane
(v , v ), permanent deformations are much smaller. For higher pressures,
-1 -2
significant hardening occurs ~nd the pressure-volume change relation
tends toward a locking behavior when no volume changes occur.

MPar------.------,------,------,
"10
~*~
L1 L2
_j
30

0.1 0.2 0.3 0.4

~­
L

Fig.3 - Deformations of diatomite under isotropic pressure

(after [12]).

The directional character of deformation under isotropic pressures

was studied in the second series of tests for the seven values of the
orientation angle 8. The pressure went up to 100 MPa. A general view of
deformed specimens is given in Fig.4. The off-axis cylindrical specimens
(0 < 8 < n/2) were deformed into elliptic inc1ined cylinders; this con-
firms the previsions of the theory as to the orientation of the strain
principal directions ( .' .' v ) with respect to the oriented cylinders,
-3
as well as the regularity of the material response. In Fig.5, the deform-
Yielding and Failure of Transversely Isotropie Solids 79

ed and undeformed cross-sections are superimposed. Fig.Sa shows the

permanent deformation in the plane of isotropy (orientation 8 = 0);
Fig . Sb (orientation 8 = 90°) furnishes information as to the strain
ellipsoid for a transversely isotropic material subjected t6 an isotropic
pressure.

Fig.4 - View of oriented cylinders of diatomite deformed by

isotropic stress (after [12]).

•l b)

Fig.S - Permanent deformation of cylindrical specimens of

diatomite under isotropic pressure (after [12]).
a): 8 = 0° ; b): 8 = 90°.
80 J.P. Boehler

4. FAlLURE CRITERIA FOR GLASS/EPOXY COMPOSITES

UNDER CONFINING PRESSURE

4.1 Introduction

The generalinvariant form (13) of the yield criterion for transver-

sely isotropic materials requires specifications for each class of mate-
rials. Such specifications can be obtained from experimental investiga-
tions. In this Section, we consider failure of glass-fiber reinforced
composites subjected to simple tensile tests and compression tests under
different confining pressures. An experimental study on glass-fiber mat
laminates discloses complex phenomena in the failure range of such mate-
rials: pronounced anisotropy of the discontinuous type associated with
the appearance of different failure modes, marked stress-sign sensitivity
and hydrostatic stress effects. We propose an elaborate failure condition,
which accounts for the different observed phenomena and furnishes correct
predictions for both directional strengths and corresponding failure
modes. Much of this Section is taken from [13].

4.2 General form of the yield criterion for triaxial tests

An equivalent form of the general invariant form (13) of the yield

criterion for transversely isotropic materials is given by:

0 (26)

where S is the stress deviator:

s T - (.!.
3
trT)I
- -
(27)

We consider triaxial tests consisting of axial compressions or trac-

tions On on specimens oriented at 8° to the privileged axis ~3 of the
material and subjected to confining pressures p (cf. Section 3 and Fig.l).
For such stress conditions, the arguments of the criterion (26) can be
expressed in terms of the three variables on' p and 8:
Yielding and Failure of Transversely Isotropie Solids 81

2
-(o -p)2 2
-(o -p)3
trT 2p+o
n 3 n 9 n
(28)
1
trMS -(o -p)(3cos 2 8-l)
3 n

Thus, if restricted to tests under confining pressure, the five arguments

of the general form (26) are no langer independent; they are related by:

1 l I2
-trS 2 + ~ trMS tr S2 E =1 if 0 -p ) Ü
3 - 16 n

3f2 where (29)

2
=E tr S2
= -1 if -p ( Ü
316 E 0
n

Only three invariants remain independent and the general form of the
yield condition for triaxial tests is reduced to:

h(trT, trS 2 , trMS) 0 (30)

Further details, as well as the visualization of the yield criterion (30)

in the space of the basic invariants, are given in [14].

4.3 Failure modes and directional strengths

We performed a number of tests on a chopped strand mat laminate

(glass fiber/epoxy resin). Suchmaterialsare transversely isotropic, the
privileged axis v being the normal to the mats' plane. Specimens with a
-3
parallelepipedie shape were cut in seven different orientations 8 to the
v
-3
axis (8 = 0, 15, 30, 45, 60, 75 and 90°) and subjected to axial com-
pressions under four different confining pressures: p = 0, 25, 50 and
75 MPa. Simple tensile tests were performed on specimens with a special
shape ( [ 13]).
The compression tests under confining pressures disclose two diffe-
rent modes of failure. For thespecimens inclined at 8 = 45, 60 and 75°,
whatever the confining pressure, failure occurs by sliding with friction
on a plane parallel to the mats; this well-defined failure mode will be
called "parallel mode". For the other orientations (8 = 0, 15, 30 and
82 J.P. Boehler

p= 0

P= 25 MPa

p= 50 MPa

p=75MPa

Fig.6 - Failure modes in compression under confining pressures

(after [13]).

Fig.7 - Failure modes in simple tension (aft er[13]).

 Yielding and Failure of Transversely Isotropie Solids 83

and 90°), failure occurs in planes crossing the mats; this mode of failure
will be called "across mode". Simple compression parallel to the mats
(8 = 90°, p = 0) is a particular case, where failure is produced by an
extension parallel to the mats; when the confining pressure increases,
this particular failure mode is replaced by the "across mode". These
phenomena are visible in Fig.6, where one deformed specimen, after fai-
lure, is presented for each orientation and each confining pressure; the
visible sides are those with inclined mat planes.
Failure modes in simple tensile tests are quite different from that
observed in compression(Fig.7). For the orientations 8 = 0 to 60°, fai-
lure occurs by an extension parallel to the mats, which is progressively
accompanied by a shear stress parallel to the strata when 8 increases
from 1S to 60°. For 8 = 90°, failure occurs in several planes crossing
the mats' planes. At 8 = 7S 0 , mixed phenomena are observed.
For the compression tests under confining pressures, the obtained
values for the limit over-stresses a -p are presented in Table I. What-
n
ever the confining pressure, a marked drop in the directional strength
is observed at the transitions from 8 = 30° to 8 = 4S 0 and from 8= 90°
to 8 = 7S 0 , which correspond to the transitions from the "across mode"
to the "parallel mode" of failure. Thus, the anisotropy of the triaxial
strengths is of the discontinuous type [1]. In simple tensile tests, the

p(MPa) 8 = 0 8 = 1S 0
8 = 30° 8 = 4S 0
8 = 60° 8 = 7S 0
8 = 90°

0 17 21 20 38 50 84 230 traction
0 480 410 340 160 1SS 208 325 compression
2S SOS 4S8 37S 180 16S 22S 383 II

so S2S 476 40S 220 180 268 414 II

7S sso 484 41S 318 198 290 449 II

Table I - Directional limit over-stresses a -p (MPa) in compression

under confining pressures and innsimple tension
(after [13]).

directional strengths increase strongly for 8 > 4S 0

•
84 J.P. Boehler

The ratios r of the directional strengths in simple compression

with respect to that in simple tension (Table II) decrease when the
orientation angle 8 increases. The values of the ratios rare very high for
8 = oo to 8 = 30°and a marked drop is observed between the orientations
8 = 30° and 8 = 45°, corresponding to the transition from the "across
mode" to the "parallel mode" of failure in compression. These results
disclose the strong sensitivity of the material to the sign of the normal
stress, as well as the importance of the effects due to the orientation
of the applied stress tensor.

8 oo 15° 30° 45° 60° 75° 90o

r 27,6 19,3 16,9 4,5 3,1 2,5 1,4

Table II - Ratios r of the simple compressive strengths

with respect to the simple tensile strengths
(after [ 13]).

4.4 Proposed failure criteria

The criterion of Tsai and Wu [15], which is widely used in engineer-

ing, takes into account sensitivity of composite materials to the sign
of normal stresses, as well as hydrostatic pressure effects. Nevertheless,
the comparison between the experimental values of the directional
strengths and the predictions of the Tsai and Wu criterion shows that
this criterion does not adequately predict the actual behavior of mat
laminates in compression under confining pressure [13]; in fact, this
criterion underestimates strongly directional strengths for the orienta-
tions 8 =15, 30 and 45°. Thus, it is necessary to develop a more elabo-
rate yield condition.
The analysis of the directional strengths in compression under con-
fining pressures shows that for the tested composite the anisotropy of
the strength is of the discontinuous type and that failure occurs with
two 'different well-defined modes, according to the orientation of the
specimen. As a reliable criterion must account for this phenomenon, we
Vielding and Failure of Transversely Isotropie Solids 85

propose a failure condition composed of two continuouslyderivable bran-

ches, each branch characterizing a failure mode. A similar approach has
been proposed by Hashin [16] in his study on failure criteria for unidi-
rectional fiber-reinforced composites under tensile stress.
For compressive stress states under confining pressures and the
"parallel mode" of failure, a generalization of Coulomb's law of friction
is proposed:

l o 23 I= (l+ap+ap 2 )o 33 tan<jl 0 + (l+ap+ap

0 1 2 3
2 )c
0
(31)

where o is the stress tangential to the failure plane, o 3 3 is the stress

23
normal to this plane and p is the confining pressure (Fig.8). The cons-
tant!l <P and c are the friction angle for p = 0 and the shear strength in
0 0
the mat's plane respectively. The material constants a. account for the
1
influance of the confining pressure.

P-

L -------
lan
--------
'""_j
Fig.8 - Failure in "parallel mode" (after [13]).

Expressin8 the o 3 3 and o 2 3 stresses in terms of the axial over-stress

on-p , the confining pressure p and the orientation 8, and employing
86 J.P. Boehler

relations (28), we obtain:

io -picos8sin8 = ltrMS 2 - tr 2 MS
n -- --
l
0 (o -p)cos 2 8 + p trMS + -trT (32)
3 3 n 3

3p trT - I .\rs 2
2

From relations (32), one can conclude that the proposed criterion (31)
is a particular form of the general criterion (30). Fina1ly, the proposed
criterion (31) admits the following form:

( o -p) cos8sin8 (1+a p+a p 2 )(o cos 2 8 + psin 2 8)tan~ +

n o 1 n o
(33)
+ (1+a p +a p 2 )c
2 3 0

From relation (33), we notice that oriented simple compression tests

allow the values of the constants ~ and c to be determined. The four
0 0
other constants a. specify the evolution of the directional strenghts
1
with the confining pressure.
For compressive stress states under confinig pressure and the "across
mode" of failure, a criterion is proposed that relates the axial over-
stress o -p , the confining pressure p and the orientation 8 by the fol-
n
lowing expression:

(1+a p+a p 2 ) (o -p) 2 + (b +b p+b p 2 ) (o -p) 2 cos 2 8 +

12 n 012 n
(34)

where a., b. and c. arematerial constants and k is the constant of the

1 1 1
criterion. Employing relations (28), we obtain:

trMS
0 < on-p /2.trS 2 cos 2 8 l+
3
I ~trS 2
2

2
(35)
Jp trT /2.trS
2
2 •
Vielding and Failure of Transversely Isotropie Solids 87

From relation (35), one can conclude that the proposed criterion (34)
is a particular form of the general criterion (30). Griented simple com-
pression tests allow the values of the material constants b and c and
0 0
the constant k of the criterion to be determined.The six other material
constants account for the influence of the confining pressure.
When applied to simple tensile tests, the criterion (34) admits the
reduced form:

(J (36)
n

This form is more general than the corresponding reduced form of the
criterion (33) and will be used for the prediction of the directional
simple tensile strengths. The three constants k, b and c must be recFti-
o 0
culated, because the failure modes in simple tensile tests are different
from the "across mode".

4.5 Comparison with experimental results

The values of the material constants involved in the proposed cri-

teria (33), (34) and (36) were calculated so as to obtain the best theo-
retical representation for the experimental data; the values obtained
are given in [13].
In Figure 9, experimental data, as well as the theoretical curves
obtained in application of the proposed criteria are presented. Whatever
the confining pressure p and the orientation 8, the theoretical predict~

ions fit the experimental results well.

Observe also that the criter~a (33) and (34) proposed for failure
under compression with confining pressures are compatible. Indeed, what-
ever the confining pressure, for the orientations corresponding to fai-
lure in "across mode", the criterion (33) specifying the failure strengths
for the "parallel mode" predicts limit-stresses which are higher than
those predicted by the criterion (34) specific for failure in "across
mode"; conversely, whatever the confining pressure, for the orientations
corresponding to failure in "parallel mode", the criterion (34) speci-
88 J .P. Boehler

u.--••25
350r---~--~~~hr~----~~~~~~~

250

100

50

Fig.9 - Comparison between experimental data and the propofed

failure criteria (after [13]).

fying the failure strengths for the "across mode" predicts limit-stresses
which are higher than those predicted by the criterion (33) specific for
failure in "parallel mode". Finally, the lower envelope of the two pro-
posed criteria furnishes correct predictions for the values of failure
strengths, as well as for the orientations corresponding to a given
failure mode.
Vielding and Failure of Transversely Isotropie So Iids 89

5. SIMPLILFIED THEORY

5.1 Plastic behavior

In order to describe the plastic behavior of an anisotropic solid

in a simple manner, as well as to be able to employ the methods already
developed for isotropic solids, a simplified theory was proposed in
[17], [18]. The underlying idea is to replace the anisotropic material
by an equivalent isotropic material by considering, instead of the stress
tensor T, the follow'ing transformed tensor involving anisotropic effects:

T AT

where A is a fourth order tensor, specifically the tensor of plastic d,.i-

:::
sotropy coefficients. In the case of transverse isotropy, the expression
of A in any (•, •, v ) privileged orthonormal frame is given by:
::: -3

A diag.(y, y, a, ß/2, ß/2, y/2) (38)

where a, ß, y are three independent material constants.

In the simplified theory, a quasi-linear isotropic relation between
the transformed stress tensor T and the rate of deformation tensor D is
used. The corresponding irreducible representation isgiven by:

_1 _1
T <P
o-I +
<P D
1-
or T= <P A I +
0::;
<P A D
1:::
(39)

In this case, the homogeneity condition (3) results in the following

equations of Euler:

0 (40)
90 J.P. Boehler

where: when i = ~ ~ j = 1,2,3

E trD E
l 2

The solutions of equations (40) are:

1
cj>o = A(p,q) cj>l =E B(p,q) (41)
2

where p and q are the following two dimensionless kinematic variables:

E E
l 3
p
E
q
E
(42)
2 2

Introducing the solutions (42) into the representation (39), the inva-
riant form for the plastic constitutive law of a transversely isotropic
solid is obtained:

T =Al + JL D
E2 -
(43)
A A(p,q) B B(p,q)

The tensor form (43) is equivalent to the system of scalar relations

between the independent isotropic invariants of T and D:

trT 3A + pB

(44)

When the two dimensionless kinematic variables (42) are independent,

their elimination from (44) results in the following expression of the
yield criterion:
0 (45)
Yielding and Failure of Transversely Isotropie Solids 91

which is a particular form of the criterion (13) of the general theory,

since, [19]:

T = A T [(a+y-2ß)trMT]M + yT + (ß-y)(MT+TM) (46)

where M v ® v .
-3 -3

The relations (39) of the simplified theory can be written as a

particular case of the quasi -linear form ( 18), [ 19]:

T
(47)
+ l $ D + (l - l)$ (MD+DM),
y 1- ß y 1 -- --

where $ and $ are specified in (41). The flow law is obtained by inver-
o 1
sion of (47); this is equivalent to the inversion of (43) and the intro-
duction of (46):

D (a+y-2ß)trMT
1 I 2
A I + M+ 1 T + b (MT+TM) . (48)
B - B B - B -- --
tr D2

The flow law (48) of the simplified theory camplies with the general
theory in its quasi-linear form (18). Finally, the simplified theory
(39) is a particular case of (1).
In any (•, •, v ) privileged orthonormal frame, the component form
- 3
of the flow law (48) is given by:

-A+yT 11 yT12 ßT 1 3
B
D yT 12 -A+yT 2 2 ßT23 (49)
tr1r2D2-
ßT 13 ßT 2 3 -A+aT
33

It is easy to see that the flow law (49) of the simplified theory pos-
sesses the same essential features as the flow law (16) of the general
theory, i.e.:
92 J.P. Boehler

• non-coincidence of the principal directions of the stress

and rate of deformation tensors in "off-axis" tests.

• non-isotropic deformation under hydrostatic pressure.

• non-association with the yield criterion (45) by the rule

of plastic potential, which appears as a special case.
If the flow is incompressible, the following restriction results
from (48):

-3A +(a-r)trMT + rtrT 0 (50)

The components of D in the privileged (•, •,v ) frame become:

-3

A.
D11 3(2rT 11 rT 2 2 - aT 33 ) D2 3 A.ßT 23

A (51)
D22 -( -rT 11 + 2rT 22 - aT 3 3 ) D31 AßT 31
3

A -rT
D33 -( 2 2 + 2aT 3 3 ) D1 2 AaT 1 2
3 11 - rT

tr1/ZD2
where: A
B

When further restrictions of plane flow is made so that

D D D 0 (52)
11 31 12

the reactive stresses express as fo1lows:

T
11
1 (T
2 22
+ ~r T
33
) T
31
T
12
= 0 (53)

and the components (51) of the rate of deformation tensor D become:

A
D33 = -2 (-rT 22 + aT 33 ) D
1 2
AßT 1 2 . (54)
Yielding and Failure of Transversely Isotropie Solids 93

Relations (53) will be used in further developments concerning the ana-

lysis of specific yield criteria in the case of plane flow.

5.2 Yield criteria

In the simplified theory, isotropic yield criteria are generalized

for transversely isotropic materials in a Straightforward manner by the
Substitution of the stress tensor T for the transformed stress tensor T.
Thus, for cohesive materials, the generalization of the Von Mises crite-
rion for transversely isotropic solids takes the form, [17], [18]:

(55)

where S is the deviator of the transformed stress tensor T:

s T - (l trT)I (56)
3 - -

From (56) and (46), we obtain:

(57)

Thus, the criterion (55) of the simplified theory appears as a particular

case of the yi•ld condition (13) of the general theory.
We shall first check whether the form (55) is versatile enough to
allow an interpretation of experimental results concerning uniaxial
traction and compression of transversely isotropic materials. In Figure
10, (e , e ) are the principal directions of the applied stress tensor,
-2 -3
~3 is the privileged direction of the transversely isotropic solid and
8 is the angle (v , e ). The principal values of the applied stress ten-
-3 -3
sor are denoted by on and ot in the case of simple traction or com-
pression, ot is equal to zero. If R8 denotes the yield point in simple
traction or compression in the direction e , the following expression
- 3
is obtained from (55):
94 J.P. Boehler

R 2 (58)
8

The variation of R8 passes through an extremum for:

2a 2+ ar - 3ß 2 (59)
2y2 + ay - 3ß2

In the case where y > a, depending upon the values of the material plastic
anisotropy constants a, ß and y, the strength variation is of the type
shown in Figure 11. The strength variation:

1
a) admits a maximum for 8
max < .:!!.4 if ß2 < -(2a
3
2 + ay)

1
b) admits a minimum for 8
min > .:!!.4 if ß2 > -(2y2
3
+ ay)
1 1
c) decreases monot ically if -(2a 2 + ay) < ß2 < -(2y2 + ay)'
3 3

Fig.lO - Transversely isotropic Fig.ll - Theoretical directional

material in strength tests. strength variation in sim-
ple compression or traction
of cohesive transversely
isotropic materials (after
[ 2 J).
Vielding and Failure of Transversely Isotropie Solids 95

We remark that the curves of Figure 11 allow for an appropriate interpre-

tation of the experimental facts presented in Chapter 1 for clays and
rolled sheet-steel.
We consiJer now simple shear tests, for which a n - crt (cf. Figure
10). If Ce denotes the yield point in simple shear, the following expres-
sion is obtained from (55) for the variation of Ce:

(60)

which is symmetric with respect to the axis e = n/4.

If we let e = 0 and e = i in (58) and e =*in (60), we obtain the
following meaning for the three material plastic anisotropy constants:

ß (61)

In the case of plane flow, taking into account relations (53), the
yield condition (55) results in:

(62)

where an and crt are the principal stresses in the plane of deformation.
The analysis of the curves (62), as well as the analysis of the criterion
(55) in plane strain and axisymmetric three-dimensiona!. stresses is given
in [1], [20].
For materials with internal friction, a simple criterion can be
developed starting with an invariant form of the Drucker-Prager crite-
rion for isotropic solids. In the simplified theory, the generalization
for transversely isotropic materials of the Drucker-Prager criterion
takes the form, [17], [18]:

=c (63)
96 J.P. Boehler

where trS 2 is given by (57) and trT expresses as follows:

trT (a-y)trMT + ytrT (64)

In (63), the material constant a is associated with internal friction.

When applied to uniaxial compression or traction, the criterion (63)
results in the following variation of R8 :

c
(65)

where the + sign before the constant a holds for the stren~th in simple
compression and the - sign holds for the strengthin simple traction.
Similar curves to thoseof Figure 11 are obtained in this case, except
that their form is now influenced by the constant a. The respective dis-
cussion, as well as the analysis of the criterion (63) in plane flow,
plane stress and axisymmetric three-dimensional stres~es are given in
[1], [20]. Camparisans with experiments are discussed in [2].

REFERENCES

1. BOEHLER, J.P., Contributions Theoriques et Experimentales a l'Etude

des Milieux Plastiques Anisotropes, These de Doctorat es Sciences,
Grenoble, 1975.
2. BOEHLER, J.P. and SAWCZUK, A., On Yielding of Griented Solids,
Acta Mechanica, 27 (1977): 185-206.
3. BOEHLER, J.P., Lois de Comportement Anisotrope des Milieux Continus,
Journal de Mecanique, 17, 2 (1978): 153-190.
4. RIVLIN, R.S., Further Remarks on the Stress-Deformation Relations
for Isotropie Materials, J. Rat. Mech. An., 4 (1955): 681-702.
5. ALLIROT, D. and BOEHLER, J.P., Evolution des Proprietes Mecaniques
d'une Rache Stratifiee saus Pression de Confinement, Proc. 4th Int.
Congress on Rock Mechanics, A.A. Balkema, Rotterdam, Vol.1, 1979,
15-22.
Vielding and Failure of Transversely Isotropie Solids 97

6. ALLIROT, D., BOEHLER, J.P. and SAWGZUK, A., Pressure-Induced Evolut-

ion of Anisotropies in Stratified Rock, Studia Geoteehniea et Meeha-
niea, III, 2-4 (1981): 59-73.
7. BOEHLER, J.P., Anisotropie et Gomportement Rheologique des Materiaux,
Gonference Generale, 19eme Colloque du Graupe Fran9ais de Rheologie
(in print).
8. BOEHLER, J.P., EL AOUFI, L. and RAGLIN, J., On Experimental Testing
Methods for Anisotropie Materials, Res Mechaniea (in print).

9. EL AOUFI, L., Gontributions a l'Etude Experimentale et Numerique du

Gomportement Mecanique des Materiaux Anisotropes, These de Doetorat
de Speeialite, Grenoble, 1984.
10. BOEHLER, J.P. and EL AOUFI, L., Heterogeneite des Ghamps de Gontrain-
tes et de Deformations dans les Materiaux Anisotropes, 19eme Collo-
que du Graupe Fran9ais de Rheologie (in print).
11. RAGLIN, J., Gontributions Theoriques et Experimentales a l'Etude de
la Plasticite, de l'Ecrouissage et de la Rupture des Solides Aniso-
tropes, These de Doetorat es Seienees, Grenoble, 1984.
12. ALLIROT, D., BOEHLER, J.P. and SAWGZUK, A., Irreversible Deformat-
ions of an Anisotropie Rock under Hydrostatic Pressure, Int. J.
Rock Meeh. Min. Sei. & Geomeeh. Abstr., 14 (1977): 77-83.
13. BOEHLER, J.P. and RAGLIN, J., Failure Griteria for Glass-Fiber
Reinforced Gomposites under Gonfining Pressure, J. Struet. Meeh.,
13, 3 & 4 (1985): 371-393.
14. BOEHLER, J.P. and DELAFIN, M., Failure Griteria for Unidirectional
Fiber-Reinforced Gomposites, in: Meehanieal Behavior of Anisotropie
Solids, Proe. Euromeeh Coll. 115 - Colloque International 295 du
CNRS, edited by J.P. Boehler, Editions du GNRS (Paris) - Martinus
Nijhoff (The Hague), 1982, 449-470.
15. TSAI, S.W. and WU, E.M., A General Theory of Strength for Anisotropie
Materials, J. Composites Materials, 5 (1971): 58-80.
16. HASHIN, Z., Failure Griteria for Unidirectional Fiber Gomposites,
J. Appl. Meeh., 102,2 (1980): 329-334.
17. BOEHLER, J.P., Gontribution a l'Etude de l'Equilibre Limite des Sols
Anisotropes, These de Doetorat de Speeialite, Grenoble 1968.
18. BOEHLER, J.P. and SAWGZUK, A., Equilibre Limite des Sols Anisotropes,
J. de Meeanique, 9, 1 (1970): 5-33.
19. BOEHLER, J.P., Sur les Formes Invariantesdans le Sous-Groupe Ortho-
trope deRevolutiondes Transformations Orthogonales ... , ZAMM, 55
(1975): 609-611.
20. BOEHLER, J.P. and SAWGZUK, A., Analyse Geometrique des Griteres
d'Ecoulement Plastique Anisotrope, in: Problemes de Rheologie, PWN,
Varsovie, 1977, 69-75.
 Chapter 6

ON A RATIONAL FORMULATION OF ISOTROPIC AND

ANISOTROPIC HARDENING

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

The aim of this work is to develop a unified and rational formulation

of isotropic and anisotropic hardening with.~n the framewerk of the tensor
functions representation theory. We consider the evolution of the yield
criteria during plastic deformations and we assume that this evolution
depends on the strain history only through the present value of the plastic
strain. For more complicated situations, the proposed concepts can be
developed in a Straightforward manner.
The classical concepts of isotropic, kinematic and anisotropic harde-
ning are analysed in detail. It is first shown that kinematic hardening
can be split into specific isotropic and anisotropic hardening, the con-
tribution of the anisotropic part being of a very restrictive orthotropic
type. Secondly, it is shown that certain materials, when subjected to iso-
tropic plastic strain, undergo an isotropic evolution of their internal
structure, which results in a modification of the shape of the initial
100 J.P. Boehler

yield condition, without appearance of anisotropic properties. Examples

are given for this phenomenon, which, obviously, cannot be taken into ac-
count by the classical formulation of isotropic or kinematic hardening.
Finally, the classical concepts of isotropic, kinematic and anisotropic
hardening are not disjointed and do not cover the full range of hardening
phenomena. Consequently, there appears a need for a more general concept
and a rational formulation of isotropic and anisotropic hardening.
In our analysis of the classical formulations, we show that the pro-
posed rules are equivalent tc very restrictive transformations of the
stress tensor. We define the generalized concepts of isotropic and aniso-
tropic hardening by, respectively, general isotropic and anisotropic trans-
formations of the stress tensor. Physically, such a formulation is based
on the assumption that the modifications of the materials' internal struc-
ture during plastic deformations result in an evolution of the mechanical
properties, which can be taken into account by specific transformations of
the stress tensor. By application of the therorems of representations for
tensor functions, it is shown that the proposed procedure enables the trans-
formation of an arbitrary initial yield criterion into an arbitrary isotro-
pic or anisotropic subsequent criterion.
Within the proposed concept, general isotropic and anisotropic harde-
ning rules are developed, in considering first-orderpolynomial isotropic
or anisotropic transformations of the stress tensor and applying the pro-
cedure to materials obeying the von Mises and linear criterion. Although
the considered stress transformation and yield criteria are very simple,
the proposed general rules constitute a framework, which is already wide
enough to be able to account for the observed mechanical phenomena and to
include the classical concepts of isotropic, kinematic and anisotropic har-
dening as special cases.
The essential results of the proposed rational formulation are present-
ed in this Chapter, which was originally published under the same title
in: Plasticity Today, edited by A. Sawczuk and G. Bianchi, Elsevier Applied
Science Publishers, London, 1985: 483-502.
Further details and developments on the subject will be included in
forthcoming papers [6], [7].
Formularion of Isotropie and Anisotropie Hardening 101

2. CLASSICAL FORMULATION OF ISOTROPIC, KINEMATIC AND ANISOTROPie

HARDENING

2.1 Isotropie hardening

Consider first an isotropic material. The general invariant form of

the initial yield criterion is given by:

1
f(trT, /trS 2 , ltrS 3 ) 0 s T-(- trT)I ( 1)
- 3 - -

In the classical concept of isotropic hardening [3], it is assumed that

the yield surface (1) undergoes a homothetic transformation of the ratio
A(P) in the stress space, where P is the plastic strain tensor. This
transformation is equivalent to a homothetic transformation of the stress
tensor T, with the ratio 1/A(P). Thus, the classical isotropic hardening
rule can be written in the form:

f(trT, /trS 2 , ltrS 3 ) 0

(2)
where

A(P) > 0
; { A(~) )
1: hardening

A(P) < 1: softening

If the material is anisotropic, the general form of the initial cri-

terion is given by:

0 ( 3)

where the ~. are structural tensors, taking into account the anisotropic
-1
material behavior [4], [5]. In order to derive the invariant form of (3),
the theory of representation of anisotropic tensor functions [4], [21] is
very useful, for it indicates the type and the minimal number of basic
invariants involved. By application of this theory, we obtain the general
invariant form of the initial criterion:
102 J.P. Boehler

-- 3 __
f(trT, /trS 2 , /trS 3 , J 1 , J 2 , • •• , Jn) 0 (4)

where the Jk are mixed invariants of the stress deviator S and the struc-
tural tensors ~. and arehomogeneaus of order 1 with respect toS. The
-1
type of anisotropy of the material is specified by the invariance group
of the Jk [4].
For example, the initial yield criterion of an orthotropic material
is given by:

f(trT, /trS 2 , ftrS 3 , trM S, trM S, /trM S 2 , /trM S 2 ) 0

-l- -2- -l- -2-
(5)
M V ® V M V ® V
-l -l -l - 2 -2 -2

where (v , v ) are two of the three privileged orthogonal directions of

-l -2
the medium.
Applying the classical concept of isotropic hardening, the subsequent
criterion of an initially anisotropic material is given by:

~ 3~ ( 1 1 r:-;;-:; 1 3r.-;;-;- 1 )
8 3 IJk
f ( trT, vtrs~, vtrs~, Jk
)
= 0 => f Itr!, ~tr~ 4 • ~tr~~, 0 (6)

Such a transformation does not modify the invariance group of the Jk. Thus,
the initial type of anisotropy is preserved in the subsequent criterion.
Finally, for initia1ly isotropic or anisotropic materials, the cla-
ssical concept of isotropic hardening is equivalent to a homothetic trans-
formation of the stress tensor:

T => T _l_T (7)

,\(P)-

For this type of isotropic hardening rule, we suggest the name "homothetic
isotropic hardening" (see also Section 2.5 below).

2.2 Kinematic Hardening

In order to account for the Bauschinger effect, different kinematic har-

Formularion of Isotropie and Anisotropie Hardening 103

dening rules were proposed [18], [25]. The classical concept of kinematic
hardening assumes that the yield surface undergoes a rigid translation
C(P) in the stress space, where C(P) is a symmetric second order tensor.
This transformation is equivalent to the translation -C(P) ofthe stress
tensor:

T => T T - C(P) (8)

We consider first an isotropic material, with initial criterion (1).

The translation tensor C(P) can be split into isotropic and deviatoric
parts:

C(P) = l~(P)I + C'(P) trC'(P) 0 (9)

3 - -

1-
By the isotropic part 3c(~)~, which induces a translation of T parallel
to the isotropic axis of the stress space, the initial criterion (1) is
transformed into:

f(trT- ~(P), /trS 2 , ltrS 3 ) 0 (10)

Thus, the subsequent criterion remains isotropic.

By the deviatoric part C'(P), which induces a translationorthogonal
to the isotropic subspace of the full six-dimensional stress space, the
initial isotropic criterion (1) is transformed into:

f(trT, ftrS 2 , ~) = 0 s - 1 -
T -(-trT)I
- 3 - -
(11)

where the isotropic invariants of the transformed stress tensor T are

related to that of T by:

trT = trT

trS 2 -2(2c'+c')trM S- 2(c'+2c')trM S +2(c' 2 +c'2+c'c' )

- l 2 _l_ l 2 _2_ l 2 l 2
(12)
trS 3 +3(c'+c')trS 2 -3(2c'+c')trM S 2 -3(c'+2c')trM S 2
- l 2 - l 2 -1- l 2 -2-

-3c'(2c'+c')trM S- 3 c'(c'+2c')trM S +3c'c'(c'+c')

2 1 2 -1- 1 1 2 -2- 1 2 1 2
104 J.P. Boehler

where c 11 , c 21 J c' 3 are the principal values of C'(P) and:

M = e-1 • e-1 M e • e (13)

-1 -2 -2 -2

(e, e) being the two first principal directions of C'(P).

-1 -2
Comparing (11) and (12) with (5), we can conclude that the subsequent
criterion is that of an orthotropic medium, with privileged directions
e , e , e . It is worthwhile to point out that the induced orthotropy is
-1 -2 -3
of a very restrictive type: for a general orthotropic medium, seven inde-
pendent invariants are involved in the yield criterion (5), whereas in
(11) only the three combinations (12) are involved.
Finally, kinematic hardening can be split into isotropic and ortho-
tropic parts. The isotropic part transforms an initial isotropic criterion
into a subsequent isotropic criterion by a translation of the stress
tensor parallel to the isotropic axis of the stress space. For this type
of hardening rule, we suggest the name "translational isotropic hardening"
The orthotropic part transforms an initial isotropic criterion into a
subsequent orthotropic criterion of a very restrictive form. For this type
of anisotropic hardening rule, we suggest the name "translational ortho-
tropic hardening".
If the material is initially anisotropic, the translational isotropic
hardening rule leaves the type of anisotropy unchanged, whereas the trans-
lational orthotropic hardening rule modifies the type of the initial ani-
sotropy. Further details will be given in ref. 6.

2.3 Anisotropie hardenins

The analysis of Section 2.2 shows that the anisotropic part of kine-
matic hardening is a very restrictive type of orthotropic hardening. In
fact, kinematic hardening cannot take into account most observed phenomena
of anisotropic hardening [9]. Different, more elaborate anisotropic har-
dening rules have been proposed [1,2,22,24], in order to account for change
in shape and size and for rotation and translation of the initial yield
criterion. The aforementioned rules are based on anisotropic linear trans-
formations of the stress tensor. A more general approach, based on the
Formulation of Isotropie and Anisotropie Hardening 105

theory of representation for tensor functions, has been developed in ref.

17 for initially isotropic and in ref. 8 and 9 for initially anisotropic
material. More information will be given in ref. 6.

2.4 Hardening phenomena which cannot be described by the classical

formulations

Several important hardening phenomena cannot be taken into account

by the classical approach of isotropic, kinematic and anisotropic harden-
ing. We shall present here two examples for two different initally iso-
tropic materials subjected to isotropic plastic strains.
The first example concerns granular materials without cohesion, obe~ ,tg
the isotropic linear criterion:

trT - a/trS 2 0 (14)

If such materials are subjected to isotropic compressionsP =EI, they

remain isotropic, but the density increases with E. The increase of densi-
ty induces an increase in the internal friction, without appearance of
cohesion. Such phenomena have been observed experimentally on sands, for
example [23]. Thus, the subsequent criterion is given by:

trT - ~( E)/trS 2 0, with ~(0) = a da < 0 (15)

dE

By the homothetic isotropic hardening rule, the initial criterion is

unchanged; by translational isotropic hardening, the initial friction is
unchanged and there appears non-zero cohesion (Fig.l). Thus, the subse-
quent criterion (15) cannot be obtained by the classical concepts of iso-
tropic hardening.
The secend example concerns rolled sheet-steel obeying the von Mises
criterion:

(16)
106 J.P. Boehler

translational
isotropic
hardening
h ;::::_ J initial criterion and
I homothetic isotropic hardening

tr T

Fig.1 - Granularmaterials subjected to isotropic plastic strain

For orthotropic ro1led sheet-stee1, a generalization of the von Mises cri-

terion has been proposed,taking into account the initial anisotropy and
the anisotropy induced by the p1astic strain [8]; the theoretical predict-
ions are we11 in agreement with experimental resu1ts [9]. For an initia1ly
isotropic steel subjected to a plane isotropic plastic strain P = EI,
the subsequent criterion, developed with respect to the principal plane
stresses, is given by:

(17)

wher a, band c arematerial constants. In the general case (a+2b ~ 0),

equation (17) represents in the (o 1 , o 2 ) plane an ellipse with the same
principal axes as the von Mises ellipse, but with a different size and
shape.
By the homothetic isotropic hardening rule, the initial von Mises
ellipse is transformed into a homothetic ellipse without change of shape;
by the translational isotropic hardening rule, the initial criterion (16)
is unchanged (Fig.2). Thus, the subsequent criterion (17) cannot be obtain-
ed by the classical concepts of isotropic hardening.
In these two examples, an initial isotropic criterion is transformed
into a subsequent isotropic criterion with change in size or (and) change
in shape. For such hardening phenomena, we suggest the name "isotropic
Formularion of Isotropie and Anisotropie Hardening 107

/
' _jsubsequent
....----1
criterion

homothetic
isotropic hardening

initial criterion
and
translational isotropic
hardening

/
Fig.2 - Von Mises materials subjected to isotropic plastic plane strain

hardening" in a more general sense that the classical one.

That is why we suggest the name "homothetic isotropic hardening" for the
classical concept of isotropic hardening and "translational isotropic
hardening" for the isotropic part of the classical concept of kinematic
hardening, in order to underline that the classical concepts are special
cases of a more general formulation of isotropic hardening.

2.5 Conclusions

The classical concepts of isotropic, kinematic and anisotropic har-

dening:

(1) arenot disjointed. Indeed, kinematic hardening is a combination

of a special isotropic hardening (translational isotropic har-
dening) and a special anisotropic hardening (translational ortho-
tropic hardening).
(2) give restrictive forms for the subsequent criteria.
(3) cannot take into account several important hardening phenomena.
108 J.P. Boehler

Thus, there appears a need for a more general concept and a rational
formulation of isotropic and anisotropic hardening.

3. GENERAL FORMULATION OF ISOTROPie AND ANISOTROPie HARDENING

3.1 Proposed generalconcept

In our analysis of the classical formulation of isotropic, kinematic

and anisotropic hardening, we have noticed that the proposed rules can be
interpreted by very restrictive transformations of the stress tensor.
A general concept of isotropic and anisotropic hardening can be based
on the assumptionthat the evolution of mechanical properties, due to the
modifications of the material internal structure, can be taken into account
by specific transformations of the stress tensor involved in the constitu-
tive law. A similar process, in a more restrictive manner, has been pro-
posed for the modelling of darnage growth [11, 14, 15, 16].
Within theproposedscheme, the generalized concepts of isotropic and
anisotropic hardening are defined by, respectively, general isotropic and
anisotropic transformations of the stress tensor. The invariant forms of
these general transformations will be specified by application of the
theory of representation of tensor functions. The main object of Section
3 is to set up the mathematical tool involved in the proposed general
concept of isotropic and anisotropic hardening. How to derive specific
transformation rules of the stress tensor from microscopic observations
or phenomenological considerations will not be treated here. Proposed
general isotropic and anisotropic hardening rules will be presented in
Section 4.

3.2 Influence of the plastic strain on the hardening rule

The influence of the plastic strain P has been mathematically analys-

ed for initially isotropic [17] and anisotropic materials [8], by the
application of the theorems of representation for tensor functions. The
results of these analyses are: if the initial isotropic or anisotropic
criterion undergoes an isotropic transformation, the involved plastic
Formulation of Isotropie and Anisotropie Hardening 109

strain is necessarily isotropic; a non-zero plastic strain deviator induces

an anisotropic transformation of the yield criterion.
Physically, these theoretical results can be interpreted in the
following manner:

(a) A non-isotropic plastic strain induces the appearance of a pre-

ferred orientation of the material internal structure, which
results in anisotropic hardening on the macroscale. For example,
an initially isotropic criterion is transformed into a subse-
quent anisotropic criterion.
(b) Possibly there may exist materials for which a non-isotropic
plastic strain induces only isotropic modifications of the in-
ternal structure. But, in this case, only the isotropic part of
P is involved in the lawof transformation.

Finally, if the hardening is isotropic, either the plastic strain is

isotropic, or only its isotropic part is involved in the hardening rule.
If the hardening is anisotropic, the deviatoric part of the plastic strain
is necessarily different from zero. Further details will be developed in
ref. 6.

3.3 Isotropie hardening

3.3.1 Initially isotropic materials

Consider first an initially isotropic material. The invariant form of

the initial criterion is given by (1). The general concept of isotropic
hardening is introduced by the general isotropic transformation of the
stress tensor T, involving the plastic strain EI:

T => T 1ü, EI) (18)

The invariant form of (18) is given by the theory of representation for

tensor functions:
110 J.P. Boehler

T a I + a S + a S2
o- 1- 2-

1
(19)
a. a.(trT, /trS 2 , ltrS 3 , E)
1 1 - -

Ernploying the Cayley-Hamilton theorem, the tensorial relation (19) can be

written in the form of relations between the basic isotropic invariants
of T and those of the transformed tensor T:

3 __

trT = kE(trT,
1 -
/trS 2 , /trS 3
3 __
/trS2 kE(trT, /trS 2 , ltrS 3 ) (20)
2 -
3__ 3 __
/trS 3 kE(trT, /trS 2 , /trS 3 )
3 -

where the superscripts denotes the dependence of the functions k. on the

1
plastic strain.
Finally, the general rule of isotropic hardening for an initial1y
isotropic material is given by:

f(trT, /trS 2 , ltrS 3 ) 0 (21)

The versatility of the proposed concept is manifest, when considering the

following properties of rule (21):

(1) Homothetic and translational isotropic hardening rules are

obviously involved as very special cases.
(2) The hardening phenomena which cannot be described by the clas-
sical formulations are taken into account (see Section 4.2
below).
(3) By rule (21), a given initial isotropic criterion can be trans-
formed into an arbitrary subsequent isotropic criterion.

The last property shows that the proposed concept of isotropic har-
dening is the most general. The demonstration of this property will be
Formulation of Isotropie and Anisotropie Hardening 111

given in ref.6.

3.3.2 Initially anisotropic materials

Consider now an initially anisotropic material. The invariant form

of the initial criterion is given by (4). The general concept of isotropic
hardening is introduced by a similar procedure: the stress tensor T is
subjected to a general isotropic transformation by the invariant form (19).
Using Rivlin's generalization of the Cayley-Hamilton theorem [20],
the three isotropic invariants of T and the mixed (anisotropy) invariants
Jk of ! and the structural tensors §i' can be expressed in terms of the
corresponding invariants of the stress tensor T. Such a procedure has
been developed in ref. 3.
Finally, the general rule of isotropic hardening for an initially
anisotropic material is given by:

0 (22)

Further detailswill be given in ref.6. We simply mention here the main

properties of the general rule (22):

(1) The transformation (19) of the stress tensor being isotropic,

the transformed mixed invariants Jk admit the same invariance
group as the initial mixed invariants Jk; thus, the type of ani-
sotropy of the initial criterion is preserved for the subsequent
criterion.
( 2) The "degree of anisotropy" varies in general. In fact, there
exists a problern for a rea li stic def ini tion of the "degree of
anisotropy". We propose the following definition: this degree
is specified by the shape, the position with respect to the
deviatoric subspace and the orientation of the yield surface
in the stress space. Thus, the degree is preserved if and only
if the transformation of the stress tensor is that of homothetic
or translational isotropic hardening rules.
(3) By rule (22), a given initial anisotropic criterion can be trans-
 112 J.P. Boehler

formed into an arbitrary subsequent anisotropic criterion of the

same type of anisotropy.

3.4 Anisotropie hardening

Within the proposed concept, the anisotropic modifications of the

materials' internal structure, which are induced by non-isotropic plastic
strains, can be taken into account by the following transformation of
the stress tensor:

1
T => T T(T, ~ .• P); P-(-trP)I E :f. 0 (23)
-1 - - 3 - -

In (23), the tensor function T is isotropic with respect to its arguments

(T, ~., P), where T is the stress tensor, P the plastic strain and ~. the
- -1 - - -1
initial structural tensors of the material. It can be shown [4] that T is
an anisotropic function with respect toT, the type of anisotropy being
specified by the invariance group of (~ .• P), which is obvious1y diffe-
-1 -
rent from the initial invariance group of the material.
Following a similar procedure to that for isotropic hardening, we use
the theorems of representations for tensor functions and the generalized
Cayley-Hamilton theorem, in order to obtain the general ru1e of anisotropic
hardening:

f(trT, ltrS 2 , /trS 3 , Jk) = 0 => f(trT, ~. ~. K~) = 0 (24)

where the K~ are mixed invariants of !• ~i and P, which take into account
the initial anisotropy, the induced anisotropy and the coupling between
them [8].
Furtherdetails and developements will be presented in ref. 6. The
main properties of the general rule (24) are the following:

(1) A given initial anisotropic criterion can be transformed into an

arbitrary subsequent anisotropic criterion, of arbitrary type
and degree.
(2) The hypothesis that the hardening rule depends only on the present
Formulation of Isotropie and Anisotropie Hardening 113

plastic strain results in the fact that for an initially isotro-

pic material, the subsequent criterion is necessarily orthotropic
[8]. But, if strain history is taken into account, considering
for example a sequence of small plastic strains, it can be shown
that the initial isotropic condition can be transformed into an
arbitrary subsequent anisotropic criterion.

4. EXAMPLES

4.1 Introduction

In developing examples of hardening rules, different specific trans-

formations of the stress tensor may be considered. We propese a general
isotropic (respectively anisotropic) hardening rule, in considering a
very simple transformation of the stress tensor: the isotropic (respecti-
vely anisotropic) first order polynomial transformation.
It is worthwhile to point out that this procedure furnishes a frame-
werk which is already large enough to include the classical isotropic,
kinematic and proposed anisotropic hardening rules as special cases and
to account for phenomena which cannot be described by the classical for-
mulations.

4.2 Proposed general isotropic hardening rule

When restricted to a first order polynomial transformation, the

general invariant form (19) can be written in the following form:

(25)

where the material coefficients A, A, C, D, are scalar-valued functions

of the plastic strain E.

Consider first a material obeying in its initial state the von Mises
criterion. In this case, the coefficients involved in (25) are subjected
to the initial conditions: C(O) = A(O) = 0, A(O) = 1, so thatfor P = 0, we
have T(O) S. Employing (25) and rule (21) of isotropic hardening, the
114 J.P. Boehler

subsequent criterion is given by:

(26)

In (26), the coefficient D(E) and A(E) account for the change of size of
the criterion, C(E) describes the translation parallel to the isotropic
axis and A(E) stands for the change in shape. Thus, the very simple stress
transformation rule (25) and the application of the von Mises criterion
provide a hardening rule involving the essential features of our general
concept of isotropic hardening: the initial criterion can change in size,
shape and position along the isotropic axis.
The functions A(E), C(E), D(E) and A(E) must be further specified
from physical observations or (and) phenomenological considerations. The
purpose of this paper is to show the possibilities of the proposed rule.
Thus, we consider the following simple case:

E ~ 0 /3 c
(27)
/(a+2d)E /l+(a-d)E
1
where a, c, d, ~ arematerial constants. For E 0, we obtain the von Mises
criterion:

E = 0 => trT 2 (28)

Employing (27) in (26) and developingwith respect to the principal

stresses, the subsequent criterion takes finally the form:

(29)
+ 2(- -31 + dE)(o l o 2 +o 2 o 3 +o 3 o l ) = 2k 2 (l+~E)

In plane stress, it is easy to see that in the special case c = 0, the

initial von Mises ellipse is transformed into a subsequent ellipse with
different size and shape. Thus, the second example of isotropic hardening
Formulation of Isotropie and Anisotropie Hardening ns

phenomena described in Section 2.4 is taken into account. If c # 0, the

von Mises ellipse undergoes in addition a translation parallel to the
isotropic axis.
Consider now a material obeying in its initial state the linear cri-
terion:

trT - a/trS 2 = c (30)

Employing (25) and rule (21) of isotropic hardening, the subsequent cri-
terion, when developed with respect to the principal stresses, is given
by:

(3A+C) (a +a +a ) + a/{3C 2 + 2AC(o +a +a ) +

l 2 3 1 2 3
(31)

Unfortunately, there is a lack of experimental results for testing the

proposed rule (31) in its general form. The increase of internal friction
of granular materials, when subjected to isotropic strain E, can be des-
cribed by introducing the following forms of the coefficients involved in
the stress transformation (25):

,\ = 1 c - 0 A= D=k(E),
(32)
with 0 < k(E) ~ 1, k(O) = 1

Introducing the coefficients (32) in (25) and employing rule (21) of

isotropic hardening for the criterion (30), we obtain for cohesionless
granular materials the following subsequent criterion:

trT - ;/trS 2 0 ' with a = k( da (33)

Thus, the first example of isotropic hardening phenomena described in

Section 2.4 is taken into account.
116 ].P. Boehler

4.3 Proposed generalanisotropic hardening rule

4.3.1 Initially isotropic materials

Consider an initially isotropic material subjected to a plastic

strain P with a non-zero deviatoric partE. In this case, the subsequent
criterion is necessarily orthotropic [8].
The proposed general hardening rule in the case of a three-dimensional
stress-tensor will be developed in ref.7. We simply mention here the pro-
cedure employed and the main results obtained. Using the general invariant
form of an orthotropic first-order polynomial tensor transformation, we
derive the stress transformation rule, involving 15 material coefficients,
which are scalar-valued functions of P. Introducing the transformed stress
into the isotropic von Mises criterion, we obtain a general three-dimen-
sional hardening rule, of which the rules proposed by Yoshimura [24] and
Svensson [22] are special cases.
We develop here a similar procedure in the case of plane stress. In
this case, it can be shown that the representation of the transformation
T T(T, P) takes the form:

}
T IJI M + IJI M + IJI T
1-1 2-2 3-
(34)
'f'. 'f'. ( trM T, trM T, trT 2 , trM P, trM P)
1 1 -1- -2- -1- -2-

where ~1 = :1® : 1 ' ~2 = :2® :2; (v-1 , v-2 ) being the principal directions
of P in the plane of stress.
When restricted to a first-order polynomial transformation, the in-
variant form (34) can be written in the form:

AT (C +B trM T + D trM T)M +

1 1 -1- 1 -2- -l
(35)
+ (C + B trM T + D trM T)M + FT
2 2 -1- 2 -2- -2

valued functions of trM P E and trM P E ' E and E2 are the princi-
-1- 1 -2- 2 ' 1
pal values of P.
Formulation of Isotropie and Anisotropie Hardening 117

Employing the stress transformation (35), different initial yield

conditions may be considered. For a material obeying in its initial
state the von Mises criterion, the subsequent criterion is given by:

(C2 + C2) + 2[C (B +F) + C B ]trM T +

1 2 1 1 2 2 -l-

+ 2[C 1 D1 + C2 (D 2 +F)]trM-2-T +

+ (B 21 +B 22 +2FB 1 )tr 2M T+
-1-
(36)
+ (D 21 +D 22 +2FD 2 )tr 2M T+
-2-

+ 2[D 1 (B 1 +F) + B2 (D 2 +F)]trM-1-T trM-2-T +

+ Fz trT 2 =2 k2,)_2

where the initial values of the material coefficients are:

p 0 => c1 c 2
B
l
D
l
0 B
2
F 1 (37)

so that for P 0, we obtain the initial isotropic von Mises criterion:

p 0 => trT 2 (38)

4.3.2 Prager's kinematic hardening rule

We consider the following very simple case for the stress transfor-
mation (35):

>-(P) - F(P) - 1 B (P) - B (P) - D (P) - D (P) - 0

1 - 2 - 1 - 2 -

C (P)
1 -
CE
1
c ( p)
2 -
CE
2
=> T CE M +
1- 1
CE M + T
2- 2
} (39)

where c is a material constant.

Employing (39) in (36) and developing with respect to the principal
stresses, the subsequent criterion takes the particular form:
118 J.P. Boehler

(40)

where 8 is the angle between the principal directions of T and P. Thus,

we obtain Prager's kinematic hardening rule [18], which appears as a very
simple particular case of the proposed general rule. We recall that by
the kinematic hardening rule, the initially isotropic criterion is trans-
formed into a particular orthotropic subsequent criterion (see Section
2.2).
Similar results are obtained for Ziegler's modification of Prager's
hardening rule [25]. The Prager-Hodge hardening rule [13] is simply ob-
tained by introducing Ä = Ä(E l +E 2 ) in therestrictions (39), in order to
add homothetic isotropic hardening to the kinematic hardening.

4.3.3 Baltov-Sawczuk hardening rule

Severalproposedanisotropic hardening rules appear as special cases

of thegeneralrule (36) in plane stress. Consider for example the Baltov-
Sawczuk hardening rule [2], which can be written in the following inva-
riant three-dimensional form:

tr(S - cE) 2 + A tr 2 [E(S-cE)] (41)

where S and E are respectively the stress and plastic strain deviators; c
and A are material constants.
The rule (41) involves a kinematic part, which is taken into account
by the following stress transformation:

s => s S - cE (42)

With respect to S, the rule (41) takes the invariant form:

(43)

Introducing the plastic strain principal values

Formularion of Isotropie and Anisotropie Hardenig 119

M + M - (t: +t: )M

}
E E E
1-1 2-2 1 2 - 3

(44)
M V ® V M V ® V
-2
M V ®V
- 1 - 1 - l - 2 -2 - 3 - 3 - 3

where (v , v ,v ) are the principal directions of E, we obtain finally:

-l -2 -3

2A(2t: +t: )(t: +2t: )trM S trM S +

1 2 l 2 -1- ---
(45)
+ A(2t: +t: ) 2 tr 2M
l 2 -1-
s + A(t: +2t: ) 2 trM S + trS 2
l 2 -2-

When applied toS, the proposed anisotropic hardening rule (36), with
C1 - C2 = 0, can be written in the form:

2K trM S trM S + K tr 2M S + K tr 2M S + K trS 2 2k2,\2 ( 4t)

1 -1- 2- 2 -1- 3 -2- 4

Identifying (45) with (46), we obtain:

,\ - K - 1 K (2t: +t: )2 K (t: +2t: )2 K2l - K2 K3 (47)

4 2 l 2 3 l 2

Thus, the Baltov-Sawczuk hardening rule appears as a special case of our

proposed general rule in plane stress. The Baltov-Sawczuk rule is three-
dimensional, but the involved transformation of stress has the tensorial
form af a particular two-dimensional first-order polynomial transformation.
A similar result is obtained for the anisotropic hardening rule proposed
by Backhaus [1]. In the sense of the tensorial form of the involved stress
transformation, the anisotropic hardening rules proposed by Yoshimura [24]
and by Svensson [22] are more general.

4.3.4 Initially anisotropic material

Considering the procedure presented in Section 3.4, a general aniso-

tropic hardening rule for initially anisotropic materials can be derived
ina Straightforward manner. In the case of an initially orthotropic mater-
ial, such a procedure has already been developed and the hardening rule
obtained has been tested by experimental results [10, 9]. Furtherdetails
and other examples will be developed in ref. 7.
120 J.P. Boehler

5. CONCLUSIONS

The proposed general concept results in a rational formulation of

isotropic and anisotropic hardening. In the developed scheme, the generaliz-
ed sense of isotropic hardening is clearly defined: a given initial iso-
tropic criterion can be transformed into an arbitrary subsequent isotropic
criterion, change of size, shape and position along the isotropic stress
axis being possible for the yield surface ; likewise, a given initial
anisotropic criterion can be transformed into an arbitrary subsequent
anisotropic criterion, but of the same type of anisotropy. All other har-
dening processes are involved in our general formulation of anisotropic
hardening. Thus, the proposed concepts are definitely disjointed.
In the presented examples, the proposed general isotropic and aniso-
tropic hardening rules are based on first-order polynomial isotropic and
anisotropic transformations of the stress tensor and on the application
of the developed procedure to the von Mises and linear criteria. Although
these stress transformations and the applied criteria are very simple,
the rules obtained are already general enough to be able to include the
classical hardening rules and to generalize them for the description of
hardening phenomena which they cannot take into account.
More general stress transformation rules and more elaborate criteria
can be introduced in the proposed concept. But even for the simple scheme
developed in this work, there exists a great lack of experimental results
for testing fully the proposed isotropic and anisotropic hardening rules.
Well organized multiaxial experiments, guided by a sound theoretical basis
are needed.
The work presented shows the usefulness and the capacity of the theory
of representation for tensor functions in the developement of a rational
and unified description of complex mechanical phenomena.
Formularion of Isotropie and Anisotropie Hardening 121

REFERENCES

1. BACKHAUS, G., Zur Fliessgrenze bei Allgemeiner Verfestigung, ZAMM,

48 (1968) : 99-108.
2. BALTOV, A. and SAWCZUK A., A Rule of Anisotropie Hardening, Acta
Mech., 1 (1965) : 81-92.
3. BOEHLER, J.P., Gontributions Theoriques et Experimentales a l'Etude
des Milieux Plastiques Anisotropes, These de Doctorat es-Sciences,
Grenoble, 1975.
4. BOEHLER, J.P., Lois de Comportement Anisotrope des Milieux Continus,
J. de Mecanique, 17 (2), (1978) : 153-190.
5. BOEHLER, J.P., A Simple Derivation of Representations for Non-Poly-
nomial Constituti ve Equations in Some Cases of Anisotropy, ZAMM, 59
(1979) : 157-167.
6. BOEHLER, J.P., On a General Concept of Isotropie and Anisotropie
Hardening. Part I: Theoretical (in preparation).
7. BOEHLER, J.P., On a General Concept of Isotropie and Anisotropie
Hardening. Part II: Examples (in preparation).
8. BOEHLER, J.P. and RACLIN J., Ecrouissage Anisotrope des Materiaux
Orthotropes Predeformes, J. Mecanique Theorique et Appliquee, Numero
Special, 1982, 23-44.
9. BOEHLER, J.P. and RACLIN J., Anisotropie Hardening of Prestrained
Rolled Sheet-Steel. In: Current Advances in Mechanical Design and
Production, Proc. 2 nd Cairo Un. MDP Conf., Cairo University, 1982,
483-492.
10. BOEHLER, J.P. and RACLIN J., A Unified Theoretical and Experimental
Study of Anisotropie Hardening, Trans. 6th Int. Conf. on SMIRT,
North-Holland, Amsterdam, 1982, Invited Lecture 13/2.
11. CHABOCHE, J.L., Le Concept de Cantrainte Effective Applique a l'Elas-
ticite et a la Viscosite en Presence d'un Endommagement Anisotrope.
In: Mechanical Behavior of Anisotropie Solids, Proc. Euromech Coll.
115-CNRS Int. Coll. 295 (Villard-de-Lans, June 1979), (Ed. J.P.
Boehler), Editions du CNRS (Paris) and Martinus Nijhoff, The Hague.
1982, 737-760.
12. Hill, R., The Mathematical Theory of Plasticity, Oxford University
Press, Oxford, 1950.
13. HODGE, P.G., Discussion on W. Frager Paper "A New Method of Analyz-
ing Stresses and Strains in Work-Hardening Plastic Solids"J J. Appl.
Mech., 24 (1957) : 482-483.
14. KACHANOV, L. M., Time of Rupture Process under Creep Conditions,
Izv. Akad. Nauk SSR Otd. Tekh. Nauk, 8 (1958), 26-31 (in Russian).
15. LEMAITRE J. and CHABOCHE J.L., Aspect Phenomenologique de la Rupture
par Endommagement, J. Mecanique Appliquee, 2 (3), 1978, 317-365.
122 J.P. Boehler

16. MURAKAMI S. and OHNO N., A Continuum Theory of Creep and Creep Damage.
In: Creep in Structures, Proc. IUTAM Symp. (Leicester, Sept. 1980),
(Ed. A.R.S. Panter and D.R. Hayhurst), Springer, Berlin, 1981, 422-
444.
17. MURAKAMI, S. and SAWCZUK A., On Description of Rate-Independent Beha-
vior for Prestrained Solids, Arch. Mech., 32 (1979) : 251-264.
18. FRAGER, W., A New Method of Analyzing StressesandStrains in Work-
hardening Plastic Solids, J. Appl. Mech., 23 (4) (1956) : 493-496.
19. REES, D.W.A., A Survey of Hardening in Metallic Materials. In: Failure
Criteria of Structured Media, Proc. CNRS Int. Coll. 351 (Villard-de-
Lans, June 1983), in press.
20. RIVLIN, R.S., Further Remarks on the Stress-Deformation Relations for
Isotropie Materials, J. Rat. Mech. Anal. 4 (1955) : 681-702.
21. SPENCER, A.J.M., The Formulation of Constitutive Equations for Ani-
sotropie Solids. In: Mechanical Behavior of Anisotropie Solids,
Proc. Euromech. Coll. 115-CNRS Int. Coll. 295 (Villard-de-Lans, June
1979), (Ed. J.P. Boehler), Editions du CNRS (Paris) and Martinus
Nijhoff, The Hague, 1982, 3-26.
22. SVENSSON, J.L., Anisotropy and the Bauschinger Effect in Cold Rolled
Aluminium, J. Mech. Engng Sei., 8 (1966) : 162-172.
23. TOUATI A., Comportement Mecanique des Sols Pulverulents sous Fortes
Contraintes, These de Docteur-Ingenieur, Ecole Nationale des Ponts
et Chaussees, Paris, 1982.
24. YOSHIMURA, Y., Hypothetical Theory of Anisotropy and the Bauschinger
Effet Due to PlasticStrain History, Aero. Res. Inst.,Tokyo Univer-
sity, 349 (1959) : 224-246.
25. ZIEGLER, H., A Modification to Prager's Hardening Rule, Q.J. Appl.
Math., 17 (55), (1959) : 55-65.
 Chapter 7

ANISOTROPIC HARDENING OF ROLLED SHEET-STEEL

J.P. Boehler
University of Grenoble, France

1. INTRODUCTION

It is well known that during manufacturing processes, the internal

structure of polycrystals become strongly oriented, thus anisotropic [1].
On the macroscopic level, the oriented internal structure results in the
anisotropy of the mecahnical behavior. For rolled sheet-steel, the inter-
nal structure generally presents three orthogonal planes of symmetry ; the
macroscopic behavior is thus orthotropic. When sheet-steel is subjected to
further irreversible processes, the anisotropy of the internal structure
is modified. The modifications affect not only the degree of initial ortho-
tropy, but also the type of anisotropy, according to the orientation of
the principal directions of the irreversible deformation with respect to
the rolling direction: this is the phenomenon of anisotropic hardening.
The observed experimental facts show the complexity of anisotropic
phenomena and reveal the need to develop a consistent formulation of cons-
titutive relations regarding anisotropic hardening. The aim of this Chapter
124 J.P. Boehler

is to provide a unified theoretical and experimental description of pheno-

menological macroscopic anisotropic behavior and of its evolution during
irreversible prestrains. The proper description and the reliable assess-
ment of anisotropic hardening of prestrained sheet-steel are of importance
when developing materials with adequate mechanical properties and when
designing engineering structures.
In Non-Linear Continuum Mechanics, an objective and rational formu-
lation of the constitutive relations for anisotropic hardening materials
can be developed within the framewerk of the theory of representations
for tensor functions [2-4]. This approach has already proved fruitful
in the study of the plastic behavior of anisotropic materials [4] and
of prestrained initially isotropic materials [5], [6]. The same approach
is developed here for initially orthotropic metals subjected to irrever-_
sible deformations. Much of this Chapter is taken from [7].

2. CONSTITUTIVE RELATION

2.1 General invariant form of the constitutive relation

We consider a plane orthotropic sheet-steel with privileged directions

y 1 , v-2 , v-3 . We define v-1 as the rolling, v-2 the transverse, and v-3 the
normal direction. We suppose that the sheet-metal has been subjected to
an irreversible prestraint P_ with principal directions (E-1 , E
-2
, v-3 ): ~

is the angle between the directions v and E (Fig.l). A plane stressT

-1 -1 -
is then applied to the prestrained sheet. The principal directions of
! are <: 1, : 2, ~ 3 ); 8 is the angle between the directions ~ 1 , and : 1
(Fig.2).
We restriet our analysis to the case where the mechanical behavior of
the material can be described in considering the stress tensor T as a
tensor-valued function of a kinematic tensor D. We suppose further that
the constitutive law depends on the history of the irreversible prestrain-
ing only through the present value of the Green-Lagrange prestrain ten-
sor P. The tensors T and D are introduced with respect to the prestrained
configuration. In order to derive the constitutive law in a correct manner,
we consider the two following tensors:
Anisotropie Hardening of Rolled Sheet-Steel 125

v1
e2
e1
01

v2

Fig.l. Irreversible prestrain Fig.2. Plane stress

T D (1)

where R is the rotation tensor of the canonical polar decomposition of the

prestrain gradient tensor. It can then be shown [8] thatthe constitutive
law of the prestrained sheet-steel takes the following form:

T F(D, P, M) (2)

where T, D and P are the plane parts of the stress, the kinematic and
the prestrain tensors respectively. The structural tensor M
accounts for the initial orthotropy of the material.
As a consequence of the Principle of Isotropy of Space, F is an
isotropic function with respect to its three arguments D, P and M. The
analysis of form (2) shows that the stress tensor T is an orthotropic
function with respect to the two mechanical arguments D and P, [9] (see
also Chapter 3).
An irreducible representation of the constitutive relation (2) is
given by (see Chapter 3):

T=a.I+a.M
1-
+ a.D
2-
+ a.P
3-
0- (3)
a.. = a..(trD, trMD, trP, trMP, trDP, trD 2 , trP 2 )
1 1 -
126 J.P. Boehler

The generating tensors of the representation and the scalar invariants

of the functional basis are entirely determined. According to the mechani-
cal significance ofthe kinematic tensor D, theinvariant canonical form
(6) constitutes the most general form of, for example, non-linear elasti-
city or non-linear plasticity of the prestrained sheet-steel.

2.2 Anisotropie hardening

For a given prestrain ~, the relation between T and D involved in

( 3) is anisotropic. If P = 0, the relation is orthotropic ; if P i= 0,
the initial orthotropy of the material is modified and the anisotropy does
not necessarily remain orthotropic: this is the phenomenon of anisotropic
hardening ([7], [13]). In this Section, we analyze various modes of aniso-.
tropic hardening for initially isotropic and orthotropic materials, taking
into account the orientation of the principal directions (E , E ) of the
-1 -2
prestrain tensor P with respect to the privileged directions(v_1 , v_2 ) of
the material.
In order to allow a proper analysis of the different representations
obtained from (3) for the various cases considered, we derive an invariant
expression for the prestrain tensor P. Introducing the following two
secend order tensors:

M E GD E M E GD E
- 1 -1 -1 -2 -2 -2
(4)
M
1 I~ ~I(E -1
'E) -2
M
2

we obtain the relations

M
- 1
+ M
- 2
= I

trMP E trM P = E (5)

- 1- 1 - 2- 2

trM P + trM P = trP

- 1- -2-

where E1 and E2 are the principal values of P in the (~ 1 , ~ 2 ) plane (cf.

Fig.2).
Anisotropie Hardening of Rolled Sheet-Steel 127

From relations (5), the following invariant expression for P can be

established:

P
-
= (2trM P- trP)M
-l- --l
+ (trP - trM P)
- -l--
I (6)

Employing expression (6), the invariants trDP and trP 2 involved in the
representation (3) can be expressed as single-valued functions of the
invariants trP, trD, tr M P and trMD:
-l- -l-

trDP (2trM P - trP)trM D + (trP - trMP)trD

}
-l- - - 1- -1- -
(7)
trP 2 2(trMP
- 1-
- tr P)trM P
- - 1-
+ tr 2 P

Now, we consider first an initially isotropic material, i.e. a mate-

rial for which the structural tensor M is identical to the isotropic ten-
sor ~.Introducing ~ =~ and employing relations (6) and (7), the constitu-
tive law (3) reduces to:

T <PI+<PM +<PD

}
0- 1 - 1 2-
(8)

Consider now an initially orthotropic material without prestrain. An

irreducible representation of the constitutive relation for such a mate-
rial is obtained from (3) with the condition P = o:

+ a. M +

}
T a. I a. D
0- l- 2-
(9)
a.. a.. ( trD, trMD, trD 2 )
1 1 -

Comparing the representations (8) and (9), we conclude that for the
initially isotropic material the prestrain induced anisotropy is an ortho-
tropy of privileged directions E , E . If the prestrain tensor P evolves
-1 -2 -
without rotation of its principal directions, the induced anisotropy re-
mains an orthotropy of privileged directions ~1 , ~2 ; the degree of the
128 J.P. Boehler

induced orthotropy changes with the values of the invariants trP, trMP.
- 1-
We consider now an initially orthotropic material and we first suppose
that the principal directions of ~ coincide with the initially privileged
directions (v , v ) of the material. In this case, the secend order tensor
- 1 - 2
M is identical to the structural tensor M. Introducing the conditions
-1
M
-1
=M and employing relations (6) and (7), the constitutive law (3)
reduces to:

T ß0 I- + ß 1-M + ß 2-D
I (10)
ßi ß.(trD, trMD, trD 2 , trP, tr~~)
1 - ~
Comparing representation (10) with thatfor a non-prestrained orthotro-
pic material (8), we conclude that the orthotropy with privileged direc-
tions (v , v ) is preserved ; the degree of this orthotropy changes with
-1 -2
the values of the invariants trP and trMP.
Consider now the general case, when the principal directions (E , E )
-1 -2
of the prestrain P do not coincide with the initially privileged direc-
tions (v , v ) of the material. Introducing the expressions (6) for P
- 1 - 2
and (7) for the invariants trDP and trP 2 in the representation (3), we
obtain the following equivalent representation:

a) T y I + y 1-M + y 2~1 + y 3-
D
0-
( 11)
b) yi = y.(trD,
1 -
trD 2 , trMD, trM D,
-1-
trP, trMP, trM P)
-1-

The relation between ~ and ~ involved in the representation (11) is ani-

sotropic. In the general case, when the arbitrary functions y 1 and y 2 are
independent, the type of this anisotropy is characterized by the invariance
group s of the tensors M and M, which is reduced to the set {I, -I} ; the
- 1
relation between T and D is then of the general anisotropic type. This
signifies that in the general case, the Superposition of an initial ortho-
tropy with a prestrain induced orthotropy results in general anisotropy.
However, the arbitrary functions y, having to be specified from expe-
1
rimental investigations for each material and each class of behavior, some
Anisotropie Hardening of Rolled Sheet-Steel 129

special cases can occur, where the resultant anisotropy reduces to an

orthotropy with specific symmetry directions. Such a special case is ana-
lyzed below.
Consider the case where the ratio of the functions y 1 and y 2 is in-
dependent of D:

y = k(P)y (12)
2 - 1

where k(P) is an arbitrary scalar-valued function of P. Then, the tenso-

rial relation (11a) reduces to:

T y I + y 1 [M- + k(P)M ] +y D (13)

0- - - 1 3-

In (13), we define the tensor m(P) by:

m(P) M + k(P)M
- 1
(14)
-

The principal directions of the tensor m makes an angle 6(P) with the ini-
tially privileged directions (v , v-2 ), such that:
-1

k(P)
tan[26(P)] sin 2'i' (15)
1-k(P)

where 'i' is the angle between the principal directions of the prestrain
tensor P and the initially priveleged directions (v-1 , v-2 ). We further
suppose that the scalar-valued functions y.1 satisfy the indentities:

y.(trD, trD 2 , trMD, trMD, trP, trMP, trMP)::::

1 - - -- -1- - -- -1-
(16)
:: ~.(trD, trD 2 , trrnD, trP, trrnP) i = 0,1,2,3.
1 - - -- -

On these hypotheses, the representation (11) reduces to the following

form:

(17)
~i ~.(trD, trmD, trD 2 , trP, trrnP)
1 -
130 J.P. Boehler

Comparing (17) with the representation (8) for a non-prestrained

orthotropic material, we conclude that the relation between T and D in-
volved in (17) is orthotropic, the privileged directions of the orthotropy
being the principal directions of the tensor m. When the prestrain tensor
P evolves, both the privileged directions and the degree of the resultant
orthotropy vary.

3. PLASTIC BEHAVIOR

3.1 General invariant forms of the flow law and the yield criterion

Representation (3) can describe plastic behavior, provided that D

stands for the rate of deformation tensor and that T is homogeneaus of
order zero with respect to time or, which is equivalent, with respect to
the rate of deformation ([4], [9]):

aT
0 if - f. 0 (18)
aD

Employing condition (18) in representation (3), we obtain the most general

invariant forms of both the flow law:

1J2 'I'I+'I'M+'I'T+'I'p
o- 1- 2- 3-
tr D2
(19)
'1'. = 'l'.(trT, trMT, trP, trMP, trTP, trT 2 , trp 2 )
1 1 -

and the yield condition:

f(trT, trMT, trP, trMP, trTP, trT 2 , trP 2 ) 0 (20)

The general invariant forms (19) and (20) in the plane stress case
are derived by a similar procedure than the corresponding forms in the
three-dimensional case (see Chapter 5).
The flow law (19) specifies the rate of deformation tensor only to
within a scalar multiplier. The relation between D and T is anisotropic.
Anisotropie Hardening of Rolled Sheet-Steel 131

In general, the principal directions of the tensors ~ and T do not coin-

cide, their relative orientation depending on the relative orientations of
the mechanical tensors D and P and the structural tensor M. It is worth-
wile to point out that the flow law (19) and the yield condition (20) are
not necessarily related by the rule of plastic potential. Relation (20)
indicates the type and the minimal nurober of the basic invariants involved
in the most general form of the yield condition. The essential variables
to be observed in experiments are thus specified.

3.2 Proposed criterion

The general invariant form (20) of the yield criterion ~as to be

specified for each class of orthotropic materials through further hypo-
theses. For prestrained rolled sheet-metal, we have proposed [8] a new
generalization of the Von Mises condition in developing a homogeneaus
form of order two in stress and employing the simplest combinations of
the basic invariants involved in (20):

A tr 2 T
1
+ A trT 2 + (A trMT + A trT)trMT + (A trT + A trMT)trTP
2 3 -- 4 - -- 5 - 6 -- --
= 2A 7 k 2 (21)

where the A. are polynomials of order two wit~ respect to the invariants
1/ 2 1
trP, tr P2 and trMP. In criterion (21), the terms A1 tr 2 T and A2 trT 2
account for the isotropic part of the initial criterion and the isotropic
hardening induced by ~· the terms A3 tr 2 MT and A4 trTtrMT account for the
orthotropic part of the initial criterion and the evolution of its degree
with P, the term A5 trTtrTP describes the new anisotropy inrluced by P and
the term A6 trMTtrTP stands for the coupling between thB initial aniso-
tropy and the induced anisotropy.
For the non-prestrainedmaterial (P = 0), criterion (21) reduces to
the yield condition for orthotropic materials submitted to plane stress
[4]. Thus, the proposed criterion constitutes a double generalization of
the isotropic Von Mises criterion generalization for initially orthotro-
pic materials generalization for the description of deformation induced
anisotropy.
132 J.P. Boehler

The proposed criterion (21) describes both isotropic and anisotropic

hardening induced by P. In order to give a simple examp1e of anisotropic
hardening,we consider an initially isotropic material submitted to a
Prestraining of principal directions v-1 , v-2 , v-3 , (~ = 0, cf. Figure 1).
Before prestraining, the material obeys the isotropic Von Mises condition.
After prestraining, the criterion is orthotropic and is represented in the
principal stress plane by a family of ellipses parametrized by the orien-
tation angle 8 (cf. Fig.2). For a prestraining P- suchthat E = -E = 0.1,
1 2

the obtained family of e]lispes is represented in Figure 3.

9=0°

2 a1
v'3k

/ '
,/ ' -2
'1-- ''
Fig. 3. - Anisotropie hardening of an initially isotropic metal

4. EXPERIMENTAL BEHAVIOR OF ROLLED SHEET-STEEL

4.1 Experimental procedure

A nurober of tests were performed on large sheets of mild steel obtain-

ed by cold rolling with three different reductions in thickness. For each
rolling reduction, the large sheets were submitted to four different uni-
axial prestrainings in the rolling direction, namely the zero prestrain,
a prestrain the closest possible to the limiting curve of appearance of
Anisotropie Hardening of Rolled Sheet-Steel 133

localized deformation and two intermediate prestrains (Fig.4).

For each rolling reduction and each prestrain, small specimens were
cut out from the large sheets in seven different orientations 8 wi th res-
pect to the principal axis of prestrain: 8 = 0°, 15°, 30°, 45°, 60°, 75°
and 90°. The small specimens obtained in this waywere then testedunder
simple tensile stress. The size and the shape of the specimens are pre-
sented in Fig.5.
A preliminary experimental study was developed in order to improve
the interpretation of the off-axis tests. For such tests, the principal
directions of the mechanical sollicitation do not coincide with the mate-
rial symmetry axes of the medium. In the classical testing procedure, with
rigidly clamped heads, non homogeneaus states of strain and stress develop
during the axial straining ([14], [15]) and the specimens tend to assume
a S-shape (Fig. 6a). This effect is characteristic for anisotropic solids
(see also Chapter 5). In order to avoid such effects, the ideal procedure
should assureauniform stress distribution on the endsof the specimens
andnon-constrained deformation (Fig. 6b). Such an ideal procedure cannot
be achieved practically. In order to approach the ideal case, we propese a
new testing device consisting of hinged fixtures with knife-edges permet-
ting the ends of the specimens to rotate. The optimate position for the
knife-edges is at the ends of the specimens, wich results in a quasi-
homogeneaus state of strain in the working length (Fig.7).
At different stages of the axial straining, the developed displacement
fields in the specimens were measured by the method of stereo-photogramme-
try with deformation parallaxes [11]. Examples of measurements ([15],
[16]) are given in Fig. 8, which shows the axial and transverse isodispla-
cement lines of specimens oriented at 8 = 60° to the direction of prestrain-
ing and subjected to tensile axial deformation up to 1 %. Fig.Sa shows the
S-shaped deformation obtained with the classical testing device. Fig. Sb
shows that the new testing device proposed allows a homogeneaus state of
strain in nearly all of the working length to be obtained.

4.2 Experimental results and comparison with theoretical predictions

In order to apply the proposed criterion to the performed experimen-

134 J.P. Boehler

Dimensions in mm
v,
.;16r----,--- -----,-------- .
-.... 0
~ t t
360

~12r----r-v~----T-------~

. .....
4 ~--4~-------m=+----~~
..
.
..
.. u
...
u
C/1 C/1 C/1 r"
0 5 10 15 \ rolling direction
ro11ing reductlon
v,) principal axis of prestrain

Fig. 4 - Applied irreversible Fig. 5 - Shape and size of specimens

prestrains

Fig. 6 - a) Classical testing device b) Ideal testing device

.----
.I
I

GI
Cl)
"C
GI
I
~
1:

""
a a
Fig. 7 - New testing device for "off-axis"
anisotropic specimens
Anisotropie Hardening of Rolled Sheet-Steel 135

X X

- U = 500

-- - V: 15

- U = 400
·-V=20

---V: 25

u=Cst u = Cst
v=Cst v = Cst
Step : S ... m
S tep : 5 ... m
10 m m
t Om m
A 8

Fig. 8 - Anisotropie steel subjected to "off-axis" simple tension ( v1 =

rolling direction ; 9 = 60°) axial and transverse isodispla-
cement lines.
A : classical testing device B proposed new testing device
136 J .P. Boehler

tal programm, we first analyze the flow law for the non-prestrained mate-
rial (P = 0). In the case of the associated flow law (normality rule),
criterion (21) results in the following expression:

D
2 + 2T
3 trT + d 0 trMT)I +(d 0 trT + c 0 trMT)M
(-- (22)
-- -

where c and d are constants. In our tests, the irreversible prestrain-

o 0
ings were applied in the ~ 1 direction. The stress and rate of deformation
tensors are thus expressed in the (v , v-2 ) plane by:
-1

(23)

Introducing the expression (23) in the flow law (22), one obtains:

3d - 2
0
E: AE: 1 with (24)
2 2(3d
0
+ 3c 0 + 2)

Relation (24) is only valid for incipient flow. The evolution of this
relation during the prestraining can be obtained from the flow law asso-
ciated with criterion (21) through the normality rule. However, the pre-
strains applied in our tests are small ; thus, in a first approximation,
we can admit that relation (24) is valid during the entire prestraining
process. In the case of an incompressible material, the principal values of
the prestraining tensor P are then given by:

E
1
=E E
2
AE E
3
-(1 + A)E (25)

Making use of the expressions (25), the application of the proposed cri-
terion (21) to oriented simple tensile tests results in:

(26)
R~(E)
(~ + A E + B E + B E 2 ) cos 2 8 + ( c + c E + c E 2 ) cos 4 8
+ A2 E 2 ) + ( B0
3 1 1 2 01 2

where Ra is the simple tensile strength in the direction 8 and the coeffi-
Anisotropie Hardening of Rolled Sheet-Steel 137

ficients H, L, A., B., C. are constants related to isotropic and anisotro-

1 1 1
pic hardening.
The eleven constants appearing in (26)were calculated from our expe-
rimental data for the three rolling conditions. The numerical values are
given in [12]. The obtained results are presented in Fig.9, where the limit
stresses in tensile axial straining are given with respect to the orien-
tation angle 8, for each rolling reduction r. The points indicate the
tests' data ; the curves parametrized by the prestrain € represent the
theoretical predictions obtained from the proposed criterion.
Considering that the further the yield curve moves awav from a hori-
zontal straigth line (corresponding to the isotropic case), the greater
the increase in the degree of anisotropy, we observe two principal phen.-
mena: for a given rolling reduction, the initial anisotropy of the shae~­

steel increases with the irreversible prestrain; the initial anisotrop~·,

as well as the anisotropic hardeningduring irreversible prestrains, are

markedly affected by the rolling reduction.
For each rolling reduction, we have four different prestrains and
seven different orientations, thus 28 experimental data. Eleven data are
needed to calculate the constants of criterion (26); thus, seventeen data
test the reliability of the proposed theory for each rolling reduction. In
spite of the complexity of the observed phenomena, the agreement between
theoretical predictions and experimental data is very good.

Ra Series A Re Series 8 R Series C

r=5% r=10/

:#~
42 hba 42 hba

(:13,25%

40 40

~~"%,.
36
~~2,00%
36 36 r.% : 1 - 0
2 - 0,75
34~ ~ 34 34
3 - 1,35
4 - 2,60

32
~ e' 32 e' 32 e'
L0~~15-7.30~45~67
0~7~5~9~0~ L0~~15~30~45~6~0~7~5-9~0~ L0~~15~3-0-4~5-6~0~75~9L0~

Fig.9 - Evolution of the yield criterion with respect to rolling

reduction, irreversible prestrainand orientation of the
specimens (after [7]).
138 J.P. Boehler

REFERENCES

1. PARNIERE, P., Relations entre Textureset Comportement Mecanique

Anisotrope des Metaux, in: Mechanical Behavior of Anis0tropic Solids,
Proc. Euromech Colloquium - Colloque International 295 du CNRS,
J.P. BOEHLER ed.,Editions du CNRS (Paris) and Martinus Nijhoff
Publishers (The Hague), 1982 : 303-332.
2. ADKINS, J.E., Symmetry Relations for Orthotropic and Transversely
Isotropie Materials, Arch. Rat. Mech. An., 4 (1960) : 192-213.
3. SPENCER, A.J.M., Theory of Invariants, in: Continuum Physics,
C. ERINGEN ed., Academic Press, New-York, (1971) : 239-353.
4. BOEHLER, J.P., Lais de Comportement Anisotrope des Milieux Continus,
Journal de Mecanique, 17,2 (1978) : 183-190.
5. RACLIN, J., Remarques sur les Conditions de Plasticite des Milieux
Ecrouissables, Troisieme Congres Fran9ais de Mecanique, Grenoble,
(1977).
6. MURAKAMI, S. and SAWCZUK, A., On Description of Rate-Independent
Behaviourfor Prestrained Solids, Archives of Mechanics, 32 (1979)
251-264.
7. BOEHLER, J.P. and RACLIN, J., Anisotropie Hardening of Prestrained
Sheet-Steel, in: Current Advances in Mechanical Design and Production,
Proc. Second Cairo University MDP Conference, Dpt. of MDP, Faculty
of Engineering, Cairo University, (1982) : 483-492.
8. BOEHLER, J.P. and RACLIN, J., A Unified Theoretical and Experimental
Study of Anisotropie Hardening, 6th Int. Conference on SMIRT, North-
Holland, Amsterdam, (1981), Invited Lecture L 3/2.
9. BOEHLER, J.P. and RACLIN, J., Ecrouissage Anisotrope des Materiaux
Orthotropes Predeformes, Journal de Mecanique Theorique et Appliquee,
Numero Special (1982) : 23-44.
10. BOEHLER, J.P. and SAWCZUK, A., On Yielding of Griented Solids, Acta
Mechanica, 27 (1977) : 185-204.
11. BENEFICE,P.and BOEHLER, J.P., The Applications of Stereophotogram-
metry to the Analysis of Displacement Fields in Solid Mechanics,
Proc. IUTAM Symposium Optical Methods in Mechanics of Solids,
A. LAGARDE ed., Sijthoff and Noordhoff, Alphen aan den Rijn, (1981)
657-668.
12. BOEHLER, J.P., RACLIN, J. and BENEFICE, P., Etude des Champs Cinema-
tiques Non Homogenes, de l'Ecrouissage et de la Fissuration des Milieux
Anisotropes, Rapport Cantrat ATP 3669 CNRS, Grenoble (1981).
13. RACLIN, J., Gontributions Theoriques et Experimentales a l'Etude de
la Plasticite, de 1 1 Ecrouissage et de la Rupture des Solides Aniso-
tropes, These de Doctorat es-Sciences, Grenoble, 1984.
Anisotropie Hardening of Rolled Sheet-Steel 139

14. BOEHLER, J. P. , Anisotropie et Comportement Rheologique des Materiaux,

Conference Generale, 19eme Colloque du Groupe Fran<;ais de Rheologie,
Novembre 1984 (in print).
15. BOEHLER, J.P., EL AOUFI, L. and RACLIN, J., On Experimental Testing
Methods for Anisotropie Materials, Res Mechanica (in print).
16. EL AOUFI, L., Gontributions a l'Etude Experimentale et Numerique du
Comportement Mecanique des Materiaux Anisotropes, These de Doctorat
de Specialite, Grenoble, 1984.
 Chapter 8

ISOTROPIC POL YNOMIAL INV ARIANTS

AND TENSOR FUNCTIONS

A.J .M. Spencer

The University of Nottingham, England

1. INTRODUCTION. NOTATION AND DEFINITIONS

Some of the circumstances in which invariance problems arise in

continuum mechanics are described elsewhere in this book. In ChaJ?ters 8
and 9 we consider the purely algebraic problern of determining systems of
polynomial invariants and tensor polynomial functions for a given set of
vectors and tensors, for some of the transformation groups which are of
importance in continuum mechanics.

For the most part we consider vectors and tensors in three-

dimensional Euclidian space; occasionally it will be useful to refer to
results in two-dimensional space. Unless otherwise stated, 'tensor' means
a second-order tensor. Reetangular cartesian coordinates are used
throughout. Results may, if desired, be expressed in terms of curvilinear
coordinate systems by standard techniques of tensor analysis. Index
notation and summation convention are employed; lower case Latin indices
take e1e values 1,2,3 unless otherwise stated.
142 AoJ.Mo Spencer

General vectors are denoted by u, v, w or u and their components in

- - - -R
a fixed reetangular cartesian coordinate system Ox 1 x 2 x 3 by ui,vi,wi,. o•
(R)
or ui The symbols ~1 , ~2 , ~3 will be reserved for Special vectors
which represent directions associated with material anisotropy. For some
purposes it is convenient to associate with a vector u (say) an anti-
symmetric tensor U, with components Uo 0, where
l.J
(1)

and eo o· are components of the third-order alternating tensor. Then

l.JK
u. (2)
1.

Components in the Ox 1 x 2 x 3 coordinate system of general (not

necessarily symmetric or anti-symmetric) second-order tensors P, Q, ... ,
(R) f d- d
or ~R' are denoted by Pij' Qij' · .. , or P o
symmetr1.c secon -or er
0 -

0 • ;

l.J (R)
tensors A, B, C, ... , or A , by Ao ., B C .. , ... , or AiJ. ; of anti-
l.J
0 • ,

-R l.J l.J
symmetric tensors X, Y, ~, or ~, by X .. , Y .. , zo ., or
With these x(R).
l.J ij l.J l.J
second-order tensors we associate 3 x 3 matrices, also denoted by P, Q,
~R; A, B, C, A ; X, Y, Z, X , where, for example, P = (Po.). The use of
- - - - -R - - - -R - l.J
the same notation to denote tensors and the associated matrices of tensor
components should not generally cause confusion. When the meaning is not
clear from the context we shall state explicitly whether matrix or tensor
notation is being employed.

We shall use both tensor component notation and direct tensor

notation as convenient, with a bias towards direct notation. Thus, for
example, u.v denotes the scalar product uivi; ~ and ~~ are vectors with
components uo P and P u. respectively; PQ is the second-order tensor
1. l.J
0 • 0 •

l.J J
with components P .. Q.k and PQT is the second-order tensor with components
l.J J
Pij"'kj,
,) o and the dyadic product u~v is the second-order tensor with
components uov .. The calculation of the components of such products is
1. J
most easily performed by matrix multiplications. Thus if u, v are
interpreted as 3 x 1 column matrices and P, Q as 3 x 3 matrices of tensor
-
components, then the scalar u.v is the single element of the matrix
T
product u v; the components of the vectors uP and Pu are the elements of
Isotropie Invariants and Tensor Functions 143

the matrix products ~T~ and ~~ respectively; the components of the tensor
P>2 are the elements of the matrix product PQ; and the components of u0v
T
are elements of the matrix product uv

The transformation

X. M .. x. or (in matrix notation) x = Mx ( 3)

~ ~J J

determines a new set of reetangular cartesian coordinates Ox 1 x 2 x 3 if

or MMT I. (4)

Here eS .. is the Kronecker delta symbol, I is the unit tensor or matrix (so
~J
that (cSij) = I) and MT is the transpose of M. Then also

eS , , 1 I and det M ±l . (5)

~J

The orthogonal matrix M defines an orthogonal transformation of

coordinates. The set of orthogonal transformations form a group (the
full orthogonal group in three dimensions) . The set of matrices M is a
matrix representation of this group.

Under an orthogonal transformation, an absolute vector u with

components ui in the Ox 1 x 2 x 3 system has components Üi in the ox 1 x2 x3
system where
U, M .. u. or u Mu (6)
~ lJ J

The components of an axial vector u transform according to the rule

u. (det M) !1 .. u. or (matrix notation) u (det !1) Mu. (7)

l ~J J

The components of an (absolute) second-order tensor P transform according

to the rule

P .. M. M. P or (matrix notation) P (8)

lJ lr JS rs

If
-
u .. u .. (9)
lJ lJ
144 A.J.M. Spencer

and U.. transform as a second-order tensor, so that

~]

U .. = M. M. U
~J ~r JS rs

then we find that

u. (10}
~

so that u. are components of an axial

~ .
vector. It is therefore possible,
and often convenient, to identify an axial vector with an anti-symmetric
tensor, by (9), and it is not then necessary to consider axial vectors as
separate entities. Henceforth we shall do this and, in the case of the
full orthogonal group, deal only with absolute vectors and second-order
tensors.

If attention is confined to pPopeP aPthogona~ tPansfoPmations (i.e.

rotations of the coordinate system) for which det M = +1, then it is not
necessary to distinguish between absolute and axial vectors.

A function f(u. ,v. , ... ,P .. ,Q .. , ... ) of the components of a set of

~ ~ ~] ~]

vectors and tensors in an inVaPiant under a given group of transformations

if

p - - - -
f ( U. , V. , . . . , P .. , Q .. , . . . ) = (det M) f (u. , V. , . . . , P .. , Q .. , . . . )
~ ~ ~J ~J - ~ ~ ~J ~J

for every transformation M.. of the group. Since, in our case,

~J
det !;1 = ±1, we may take p = 0 or p = l. If p = 0, then f is an abso~ute

or even invariant; if p 1, then f is a Pe~ative or odd invariant. We

suppose invariants to be absolute unless otherwise stated.

The following are three simple and well-known examples of invariants

under the full orthogonal group.
(a) The scalar product u.v of two absolute (or two axial) vectors:

u.v. = M. U
p
M. V
q
= 6pq Up V q U V u.v.
~ ~ ~p ~q
PP ~ ~

(b) The trace tr P of the matrix P:

P .. M. M. P 6 p p P ..
~~ ~p ~q pq pq pq PP ~~
Isotropie Invariants and Tensor Functions 145

(c) The determinant of P:

det P (det M) 2 det P det P

The central problern in invariant theory is as follows. For a given

set of vectors and tensors and a given transformation group, determine
a set of invariants (a basis) from which all other invariants can be
generated. For polynomial invariants the basis is called an integrity
basis. A polynomial invariant is reducib~e if it can be expressed as a
polynomial in other polynomial invariants, otherwise it is irreducib~e.
A basis which contains only irreducible invariants is minima~.
Polynomial relations between invariants which do not permit any one
invariant to be expressed as a polynomial in the others are syzygies.

2. RESULTS FROM CLASSICAL THEORY

Classical invariant theory, as described in, for example, Grace and

Young [1], Elliott [2~, Turnbull [3], Weyl [4] and Gurevich [5], was
mainly motivated by projective geometry and is usually concerned with
invariance under the full linear group of transformations. Consequently
its numerous results are often not directly applicable to invariance
under orthogonal transformations. However, there are important general
theorems.
(a) HiZbert's Theorem. For any finite set or vectors and tensors of any
order, there exists a finite integrity basis. (Proof in Gurevich
[ 5]) .
(b) Peano's Theorem. Consider a set of R tensors A (of any order) all
-p
of which are similar in the sense that they have the same order and
have the same index symmetries. Suppose each has v distinct
components which, arranged in some definite order, are denoted A (P)
i
(i = l,2, ... ,v; P = 1,2, ... ,R). The poZarization operator DPQ is
defined by
146 A.J .M. Spencer

Then Peano's theorem states that 1 with the possible exception of the
vth order determinants in which a typical column is comprised of
the elements A~P) 1 every polynomial invariant of the tensors can be
~

expressed as a polynomial in invariants of v-1 of the tensors and

invariants formed from them by polarization. It is therefore
sufficient to consider v-1 of the R tensors. A proof of the theorem
is given (under the name Pascal's Theorem) in Weyl [4].
(c) Every aLgebraic invariant is a solution of an algebraic equation
whose coefficients are rationaL invariants.
(d) Every rationaL invariant is a quotient of two poLynomiaL invariants
(which may be relative invariants) .
(e) Every poLynomiaL invariant is a sum of homogeneaus poLynomiaL
in varian ts.

The results (c) 1 {d) and (e) (proofs in Gurevich [5] and Spencer
[6]) show that the study of algebraic invariants reduces to that of
homogeneaus polynomial invariants.

3. ORTHOGONAL TRANSFORMATION GROUPS

Orthogonal transformations in three dimensions comprise rotations

and reflections. A rotation 8 about an axis defined by a unit vector v 1
'th components v;( 1 ) .~s c h aracter~ze
w~ . M = (M.( 8. )) 1 wh ere -
. dby th e matr~x
~ -8 ~J

M(S)
ij
o~J
.. cos8+e l.J
.. kvk( 1 ) sinS+ (l-cos8)v~ 1 )v~ 1 )
~ J

In particular 1 if ~1 coincides with the base vector : 1 of a system of

reetangular cartesian coordinates ox 1 x 2 x 3 1 then

0 0

cos e sin e
-sin 8 cos e
A reflection in the planes normal to ~1 is characterized by the
( l)
matrix ~1 (Rij ) 1 where
Isotropie Invariantsand Tensor Functions 147

( 1)
R ..
~J

If ~1 coincides with : 1, then

-1 0 0

~I 0 1 0
0 0 1

For the remainder of this section we choose ~ 1 , ~2 , ~3 tobe a

right-handed set of mutually orthogonal unit vectors, and define ~2 and
~3 in the analogaus manner to ~1 •

Particular orthogonal transformations of special interest are those

represented by the following matrices:

I (the identity transformation);

c which represents a central inversion;

M which represents a rotation 'TT about the ~1

-'TT
axis. If ~1 , ~2 , ~3 coincide with
~~, ~2 , ~3, then

0 0

-1 0
~I
0 -1

The matrices J?z and J? 3 are defined similarly.

Material symmetries require invariance under certain groups of

orthogonal transformations. The material symmetries oi concern to us are
the following.
(a) Isotropy with a centre of symmetry. In this case the isotropy group
is the full orthogonal group, which consists of all orthogonal
transformations, represented by the orthogonal matrices M with
det M = ±l.
(b) Isotropy without a centre of symmetry. In this case the isotropy
group is the proper orthogonal group, or rotation group, which group
148 A.J.M. Spencer

is represented by the proper orthogonal matrices M with detM = +1.

Haterial isotropy with no centre of symmetry requires invariance
under this group.
(c) Transverse isotropy. Material with a single preferred direction is
transversely isotropic. The constitutive equations are invariant
under rotations about the preferred direction. If the ~ 1 direction

ls chosen as the preferred direction, we have invariance under the

rotations ~e· There are five cases, whose transformation groups
consist of the following matrices:

{i) (iiiJ ±~8 , ~ 1 , c;

(vJ ±~e' ~~, ~2' ~3' ~2' ~3' c.

Clearly, all the groups include ~0 = ~ and ~TI=~~·

The groups {a), (b), (c) are continuous groups, with an infinite
number of elements. Finite sub-groups of the full orthogonal group are
also of interest. In particular crystal symmetries are represented by
thirty-two groups arranged in seven systems. These groups are known as

the crystal classes. The transformations which form these groups are not
listed here. They can be found in Green and Adkins [7], Spencer [6] or in
texts on crystal physics. We consider only two cases, which may also be
regarded as symmetries appropriate to fibre-reinforced materials
reinforced by two families of continuously distributed fibres.
{d) Prismatic symmetry. The transformation group consists of the
matrices

This is the appropriate symmetry group for a material reinforced by

two families of non-orthogonal fibres lying in planes x 1 = constant,
the two families having distinct mechanical properties (e.g. an
unbalanced angle-ply laminatel .
(e) Orthogonal (rhombic) symmetry. The transformation group consists of

This is the symmetry group for an ortnotropic material, which has

symmetry under reflections in each of the coordinate planes. It is
Isotropie lnvariants and Tensor Functions 149

appropriate for a material reinforced by two families of orthogonal

fibres, with distinct mechanical properties, directed parallel to

the x 2 and x 3 axes (e.g. a cross-ply laminatel, or for a material

reinforced by two mechanically equivalent families of fibres whose
bisectors are parallel to the x 2 and x 3 axes (e.g. a balanced angle-
ply laminatel .

4. INTEGRITY BASES FOR VECTORS

Consider a set of M absolute vectors u (?l (P = 1, 2, ... , Ml in three

dimensions. Then it is easily verified that the following scalar

products are invariant under the full orthogonal group

(Pl (Ql
U, U, u .u (P,Q 1,2, ... ,Ml. (lll
l l -P -Q

It can be shown that this set is complete, and that (lll form an integrity
basis under the full orthogonal group for the set of vectors. The result
is due to Cauchy; a proof based on Peano' s theorem is gi ven in \~ey l [ 4]
and Spencer [6].

For the proper orthogonal group, there are additional invariants of

the form

(P f. Q f. R f. P) (l2l

and the integrity basis comprises the invariants (lll dnd (12). The
expressions (12) are relative invariants for the full orthogonal group.
We note the syzygies

u .u u .u u .u
-P -S -P -T -P -U

I I u .u ':Q. ':T u .u (l3l

PQR STU -Q -S -Q -U

u .u u .u
-R -S -R -T '::R·':u

If the set of vectors contains axial vectors then some modification

150 A.J.M. Spencer

is required. However, as remarked in Section l, it is usually preferable

to identify an axial vector with an anti-symmetric tensor.

5. ISOTROPIC TENSORS

An isotropic tensor is defined to be a tensor whose components are

unchanged by any proper orthogonal transformation of coordinates.
Isotropie tensors which are unchanged under both proper and improper
orthogonal transformations can be obtained by deleting from the list of
isotropic tensors those which are not invariant under the central
inversion C. The following method of deriving isotropic tensors is due
to Smith and Rivlin [8]. We consider isotropic tensors in three
dimensions.

Let a. . . be an isotropic tensor of order ~· Then for any

l l l 2 ••• lll
orthogonal matrix M

M. . M. . . . • 14 . . a. . . (14)
llJl l2J2 lllJll JlJ2 . . ·]lJ

Consider the scalar expression

( l) (2) (~)
J = a. , . U, U, ••• U, (15)
l 1 l 2 ..• l~ l 1 l 2 lll

where u~P)
l are componen t s o f ab so 1 u t e vec t ors. Th en

a. . . (16)
l 1 l 2 ••• l 11

It is easily verified that J is invariant under orthogonal transformations.

It can therefore be expressed as a polynomial in expressions of the types
(11) and (12). It is also multilinear in the components of each of the
vectors ~P (P = 1,2, ... ,IJ). Thus a typical term of J is

(17)

where S is a numerical coefficient and A, B, C, D, ... , G, H, K, ... is a

Isotropie Invariants and Tensor Functions 151

permutation of 1,2, ... ,~. Since

"3 ( (G) (H) (K))

a e u u u
pqr p q r
e
dU~A) dU(B) dU ( G) dU (H) dU (K) stu
J .K s t u

it follows from (16) and (17) that a. . . can be expressed as a

1.11.2···1.~
linear combination of terms each of which is an outer product of
Kroneckerdeltas and third order alternating tensors. Hence:

Every isotropic tensor is a linear combination

of outer products of Kronecker deltas 6 .. and
l.J
alternating tensors e
stu

Since

6.l.r 6.l.S

e l.J
.. ke rst 6. 6. (18)
Jr JS
6kr 6ks

a. . . can be expressed so that each of its terms contains at most

1.!1.2···1.~
one alternating tensor.

To obtain the isotropic tensors for the full orthogonal group, we

delete the tensors of the form e Hence for the full orthogonal group
stu
every isotropic tensor is a linear combination of products of Kronecker
deltas.

6. ISOTROPIC INVARIANTS OF VECTORS AND SECOND ORDER TENSORS - GENERAL FORM

Let I be a homogeneaus polynomial invariant of V vectors u and A

-R
tensors ~s in three dimensions. Thus

(19)
152 A.J.M. Spencer

where R 1 , R2 , ••• , ~ are integers (not necessarily different) chosen

from 1,2, .•. ,V; S 1 , 8 2 , ••• , SM are integers (not necessarily different)
from 1,2 .•. ,A, and the ßs are numerical coefficients. Under an
orthogonal transformation M, the components of u and P transform to
-R -s
- (R)
u M .u.
(R) p(S) =M .M p~S) (20)
p p~ ~ qr qJ rk Jk

and since is invariant, it is also equal to (19) with each u. and Pjk
I
-
replaced by the corresponding u. and pjk
~

Hence, using (20) and eq1.1Ai:ing

~

coefficients

Thus the ßs are components of an isotropic tensor and so can be expressed

as polynomials in the isotropic tensors o..
~J
and e. 'k (with indices chosen
~J
from i 1 ,i 2 , ••• ,iN,j 1 ,j 2 , ••• ,jM and k 1 ,k 2 , ••• ,kM). On substituting such a
polynomial in (19) , it follows that I can be expressed as a polynomial in
expressions of the following types.

( 1)
(a) 7Ti i = tr 2: 1 , (2la)

(b) (2lb)

(o) (22a)

(5) (R1l {6) (Rz) (7) (R3)

(d) e .. k7T. u TI. u 7Tk u , (22b)
~J ~q q Jr r s s

( 1) ( 7)
where 7T. . , •.. , 7T. . are expressions of the form
~J ~J

7T .. or (23)
~J

or expressions obtained from (23) by replacing some or all of P(S) by

ij
Isotropie lnvariants and Tensor Functions 153

p(S) In each term of I at most one expression of type (c) or (d) need
ji .
appear. R 11 R21 R 3 are integers (not necessarily different) chosen from
1121 ••• 1 V 1 and S 11 S 21 ••• are integers (not necessarily different)
chosen from l 1 2 1 ••• 1 A. Expressions of the form (21) areabsolute
invariants for the full and proper orthogonal groups; expressions of the
form (22) are absolute invariants for the proper orthogonal group and
relative invariants for the full orthogonal group.

The expressions (21) and (22) give an integrity basis for u and P
-R -S
but the basis is not minimal or even finite. It is therefore necessary
to eliminate the redundant members.

Since any tensor can be expressed as the sum of a symmetric tenscr

and an anti-symmetric tensor 1 it is sufficient to consider invariants of
symmetric tensors 1 anti-symmetric tensors 1 absolute vectors and axial
vectors. Since an anti-symmetric tensor can be associated with an axial
vector 1 it is sufficient to consider invariants of symmetric tensors 1
anti-symmetric tensorsandabsolute vectors. For invariance under the
proper orthogonal group and its sub-groups 1 it is not necessary to
distinguish between absolute and axial vectocs.

7. TRACES OF MATRIX PRODUCTS AND MATRIX POLYNOMIALS

In this section 1 all matrices are 3 x 3 matrices. A matrix product

of degree N formed from a set of matrices ~1 1 ~2 1 ~Ais a product

TI (24)

or 1 in components 1

TI ..
l]

where S 11 8 2 1 ••• 1 SN are integers (not necessarily different) chosen from

the integers l 1 2 1 ••• 1 A. The trace of TI is denoted tr TI and is a scalar:
154 A.J.M. Spencer

tr'TT tr P P .•• P (25)

-S1-S2 -SN

Note that the trace of a matrix product is the trace of its transpese

T T T T
tr TI tr TI , tr PQ ... S = tr S ... Q P (26)

and the trace of a matrix product is unchanged by cyclic interchange of

its factors

tr PQ ... RS = tr Q••• RSP = • • • = tr SPQ ... R . (27)

A matrix po~ynomia~ is a sum of matrix products with coefficients which

are polynomials in traces of matrix products.

We use the notation

tr 'TT :: 0, (28)

to indicate that tr ~ can be expressed as a polynomial in traces of matrix

products of degree lower than the degree of 'TT. ~! and ~2 are of the
If

same degree, and tr<:; 1-:; 2 ) = o, we say that !;I and !;2 are equiva~ent.

The Cayley-Hamilton theorem for a 3 x 3 matrix is the identity

o. (29)

Hence

0 (30)

so that det P can be expressed as a polynomial in tr P, tr P 2 and tr P 3 •

If in (29) we replace ~ by A~+~g+v~, where A, ~ and v are arbitrary

scalars, and equate to zero the coefficient of A~V, we obtain an
important identity due to Rivlin [9], namely
Isotropie lnvariants and Tensor Functons 155

PQR + PRQ + QPR + QRP + RPQ + RQP

- (QR+RQ)
......... ............
tr P- (RP+PR)
.... ..........
tr Q- (PQ+QP)
.... ......... ............
tr R
.....--
- P (tr QR- tr Q tr R) - Q (tr RP- tr. .R tr P) - R(tr PQ- tr P
.......................
tr Q)
......... --

= 0. (31)

The identity (29) is recovered from (31) by setting P Q R.

By multiplying (29) on the right by S and taking the trace, we

obtain

..... ..... -
tr P 3 S- tr P 2 S tr P + ~ tr PS{ (tr P)
..... ..... -- 2 - tr P 2 }
......
- tr s det P
.....
0 ,

or,

tr P 3 S ::: 0. (32)

Similarly, from (31),

tr (PQR+PRQ+QPR+QRP+RPQ+RQP) S - 0 .
.................. ................ ................. .................. --- --- .....
(33)

In (33) set P = Q. This gives

- --- -- -
tr (P 2 R+PRP+RP 2 ) S ::: 0 .
-
(34)

Replace ~ in (34) by ~~, and we obtain

(35)

Next consider tr PQPRQS. Then from relations of the type (34)

tr(PQP)RQS - - tr(P 2 Q+QP 2 )RQS

--- ---

- - - --
- tr P 2 (Q 2 R+RQ 2 ) S + tr (Q 2 P 2 R+P 2 RQ 2 ) S .
- .............. - -- (36)

Also, from relations of the form (34),

 156 A.J.M. Spencer

tr P (QPRQ) S - - tr P (Q 2 PR+PRQ 2 ) S

( 37)

Comparing (36) and (37) gives

( 38)

By replacing P in turn by P+T and P-T, there follows

tr (PT+TP) RQ 2 S ::: 0 , ( 39)

"'"' .....,...... ""'"" ......

and, in a similar manner,

tr(PT+TP)R(QK+KQ)S ::: 0. (40)

"""' """" ...... ............. """" ......

Now, in turn, in (40), replace S by LS and Q by QL

"""' "' ............

Hence, by subtracting these two relations

tr (PT+TP) RQ (KL-LK) S ::: 0 . ( 41)

...., ....... "'"" ""'"" ........ ....., ......, ............

Also, in (40), replace R by RQ and Q by L

tr(PT+TP)RQ(LK+KL)S ::: 0. (42)

"'...... ...... ...... "'"' -.,

From (41) and (42)

tr ( PT+:E~) ~9~~~ - 0 .

It follows that the trace of a matrix product of degree seven &s

equivalent to the negative of the trace of the matrix product obtained
from it by interchanging two adjacent factors.

Now, in (33) replace ~, g, ~, s by PQ, ~~, TK, L respectively. Then

( 43)

Each term in (43) can be derived from tr PQRSTKL by an even number of

................. ------ ........ "'
interchanges of adjacent factors. Hence each term is equivalent to

...... ......, .................... _

tr PQRSTKL,
,.."
and hence

(44)
Isotropie Invariants and Tensor Functions 157

Thus the trace of any matrix product of degree seven (or more) can be
expressed as a polynomial in traces of matrix products of degree less than
seven.

To summarise the main results of this section we have the following

theorem:

The trace of any matrix polynomial in 3 x 3 matrices P can be

~
-R
expressed as a polynomial in traces of matrix products, each of which
products satisfies:
( a) it is a product of factors of the form ~R' ~~, ~!;
(b) the first and last factors are not powers of the same matrix;
( c) if p 3 is a factor, there are no other factors;
-R
(d) no two factors are the same;

(e) ~R precedes ~~ in any product with both ~R and ~~ as factors;

(f) if ~~ and ~~, ~ f S, are both factors, they are consecutive factors;
(g) no product has total degree greater than six.

8. INVARIANTS OF SYMMETRIC SECOND-ORDER TENSORS

For invariants of second-order tensors only, it is not necessary to

distinguish between the full and the proper orthogonal groups. From
Section 6, equation (2la), any polynomial invariant of a set of symmetric
second-order tensors is a polynomial in expressions of the form tr TI,
where TI is a matrix product in the matrices of the tensor components.
For a finite set of matrices there is only a finite nurober of matrix
products which satisfy the theorem proved in Section 7. Hence the matrix
products which satisfy the theorem yield a finite integrity basis.
However, in general this basis is not minimal.

An invariant of the basis cannot be of degree greater than six, and

so cannot involve more than six distinct tensors. It is therefore
sufficient to consider invariants of up to six distinct tensors; an
158 A.J.M. Spencer

integrity basis for more than six tensors is obtained by combining the
bases for the tensors taken six at a time, in all possible combinations.

The procedure is to list the matrix products which satisfy the

theorem, for one to six matrices in turn, and express as many as possible
of these invariants in terms of the remainder. We eventually arrive thus
at a minimal basis; proof of minimality is discussed briefly in Section 7
of Chapter 9. An alternative approach to the construction of minimal
integrity bases, which is based on group-theoretical rnethods, is described
by Smith [10, 11].

In addition to the results of Section 7, we make extensive use of

the fact that the trace of a matrix product formed from symmetria
matrices is equal to the trace of the matrix product with its factors
written in reverse order; that is, if A, B, • • • I F are symmetric matrices

tr AB ... F = tr F ... BA (45)

The results for up to three tensors are as follows.

On.e matrix A
A minimal integrity basis is:

tr A, tr A2 , tr A3 •

TWo matriaes A, B
The following invariants satisfy the theorem:

and this invariant can be omitted from the basis.

Three matriaes A, B, c
- - -
We have to consider

and invariants derived from these by permuting A, B, c.

By cyclic interchanges and transposition

trABC trBCA trCAB tr CBA trACB tr BAC,

Isotropie lnvariants and Tensor Functions 159

so of these we retain only tr ABC.

Similarly trA 2 BC = trA 2 CB, etc., sowe retain

By similar arguments, we retain

The remaining invariants in (46) are reducible, as follows.

Replacing B by B 2 gives tr AB 2 A2 C :: 0. Finally

As the number of tensors increases, the number of invariants needing

consideration and the number of relations between them grow rapidly, and
we omit the rather lengthy details. The final results are shown in
Table l. The most difficult case is that of six tensors, where we have
ini tially to consider 6! = 720 invarian ts of the type tr ABCDEF. These
reduce to 60 by using the results that the invariants are unchanged by
factor reversal and cyclic interchange. Between these 60 invariants
there are various relations which can be reduced to (a) 60 relations of
the type

tr ABC (DEF+DFE+EDF+EFD+FDE+FED) - 0 ,

and (b) 90 relations of the type

trAB{ (CD)EF+(CD)FE+E (CD)F+EF(CD)+F(CD)E+FE(CD)} - 0.

Of these lSO relations, it can be shown that SO are independent; they may
be used to express SO of the invariants in terms of the remaining ten,
for example the ten listed in Table l.

9. INVARIANTS OF SECOND-ORDER TENSORS AND VECTORS; PROPER ORTHOGONAL GROUP

For the proper orthogonal group it is not necessary to distinguish

160 A.J.M. Spencer

between absolute and axial vectors. With each vector u we can associate
an anti-symmetric tensor U by the relations

(47)

It is therefore sufficient to consider invariants of symmetric and anti-

symmetric second-order tensors. Hence the forms of invariants we have to

consider are, from (2la),

tr TI ,

where TI is a matrix product formed from symmetric matrices A, B, ... , and

anti-symmetric matrices U, V, X, y,

Equivalently, we recall that the required invariants are of the

forms (2la,b) and (22a,b). By making Substitutions of the forms (47) and
employing the identity (18) it can be shown that each of the invariants
(2lb), (22a) and (22b) can be expressed as a polynomial in traces of
matrix products formed from symmetric and anti-symmetric matrix products
(details are given in Spencer and Rivlin [12] and Spencer [6, 13]). Hence
it is sufficient to consider the invariants tr TI whicb are traces of
these matrix products.

The same methods as were used to reduce the integrity basis for
symmetric tensors can now be used for symmetric and anti-symmetric tensors.
In addition to the results used earlier based on the Cayley-Hamilton
theorem for general matrices we have available the following results for
traces of matrix products involving anti-symmetric matrices X, Y, Z and
symmetric matrices A, B, C, ... (which may be the unit matrix I):

tr X = 0, tr XA 0, (48)

tr XAB ... C -tr C ••• BAX, (49)

tr XA ... YB ... C tr C ••• BY ... AX, (50)

tr XA ... YB ... ZC ... D -tr D ... CZ ... BY ... AX, (51)

where the dots represent symmetric matrices. In addition we have the

Isotropie lnvariants and Tensor Functions 161

TABLE l
INVARIANTS OF VECTORS AND TENSORS - PROPER ORTHOGONAL GROUP

MATRICES MATRIX PRODUCTS

A A; A2 ; A3
A1 B AB; AB 2 *; A2 B 2
A1B 1C ABC; A2 BC*; A2 B 2 C*
A 1 B 1 C1 D ABCD 1 ABDC; A2 BCD* 1 A2 BDC*; A2 B 2 CD 1 A2 C2 BD 1 A2 D2 BC 1 B 2 C2 AD 1
B 2 D2 AC 1 C 2 D2 AB; A2 BACD*
A 1 B 1 C 1 D1 E ABCDE 1 ABDEC 1 ABECD 1 ACBED 1 ACDBE 1 ADBCE; A2 BCDE* 1 A2 BCED* 1
A2 CBDE* 1 A2 BEDC*
A 1 B 1 C 1 D1 E 1 F ACFEBD 1 ADCBFE 1 ADCFBE 1 ADFBCE 1 ADFCBE 1 AEBDCF 1 AECBDF 1
AECDBF 1 AEDBCF 1 AEDCBF
X xz
X1 A X2 A; X2 A 2 ; X2 AXA 2
X1 A 1 B XAB; XA 2 B*; XA 2 B 2 ; XA 2 BA*; XA 2 B 2 A* 1 X 2 AB; X2 A2 B*; X 2 AXB;
X2 AXB 2 *
X1 A 1 B 1 C XABC 1 XACB 1 XBAC; XA 2 BC* 1 XA 2 CB* 1 XBA 2 C*; XA 2 BCA*; XA 2 B 2 C* 1
XB 2 A2 C; X2 ABC 1 X2 BCA; X2 A2 BC*; X2 AXBC
X1 A 1 B 1 C1 D XABCD 1 XABDC 1 XACDB 1 XBACD 1 XBADC 1 XCABD; XBA 2 CD* 1 XCA 2 DB* 1
XDA 2 BC*; X2 ACDB 1 X2 ABDC 1 X2 ABCD
X 1 A 1 B 1 C 1 D1 E XBDACE 1 XACBDE 1 XABCED 1 XABDCE 1 XACEBD
X1 Y XY
X1 Y 1 A ~~A; XYA 2 ; X2 YAt 1 X2 YA 2 t; X2AYA 2 t
X1 Y 1 A 1 B XYAB1 XYBA; XYA 2 B* 1 XYBA 2 *; XYA 2 B 2 ; XYA 2 BA*; X2 YAB't;
~;~~Bt~-~;BY~;;t-
X1 Y 1A 1 B 1C XYABC1 XYACB 1 XYBAC 1 XYBCA 1 XYCAB; XYA 2 BC* 1 XYA 2 CB* 1
----- ----- _f __ _
XYBCA 2 *; X2AYBC
X1 Y 1 A 1 B 1 C 1 D XYACDB 1 XYABDC 1 XYABCD 1 XYBCDA 1 XYBADC 1 XYCBDA
X1 Y 1 Z XYZ
X1 Y 1 Z 1 A XYZA 1 XZYA; XYZA 2 1 XZYA 2 ; XYAZA 2
X1 Y 1 Z 1 A 1 B XYZAB 1 YZXAB 1 XYAZB; XYBZA 2 *
X1 Y 1 Z 1 A 1 B 1 C XYAZBC
162 A.J.M. Spencer

Speneer-Rivlin identity [12]

T T T T
XPY + P XY + XYP = ~ tr ~ + ;~ tr ~ + ~ ( tr ~~ - ; tr ~ tr ~) (52)

where P is a general 3 x 3 matrix. This identity is independent of the

Cayley-Hamilton theorem, although it reduees to the Cayley-Hamilton theorem
for anti-symmetrie 3 x 3 matriees when X = Y = P. From (52) it follows that

for any matrix Q. (53)

Rivlin and Smith [14] have proved that there are no other distinet matrix
identities exeept ( 31) and (52) for symmetrie and anti-symmetrie 3 x 3
matriees.

We omit details of the reduetions, whieh are similar to those

employed in the ease of symmetrie tensors only. The results are given in
Table 1. In Table 1, the invariants are the traees of the tabulated
matrix produets, where A, B, C, D, E, F are symmetrie matriees assoeiated
- - - - -
with symmetrie tensors, and X, Y, z are anti-symmetrie matriees assoeiated
with anti-symmetrie tensors or with veetors (by (47)). The integrity
basis for a given set of not more than six veetors and tensors is obtained
by adding to the invariants tabulated for the given set the invariants for
all the subsets of the given set. The integrity bases for more than six
veetors and tensors is the sum of the integrity bases for the veetors and
tensors taken six at a time in all eombinations. An asterisk (*) indieates
that in addition to the listed invariant, the basis also ineludes
invariants formed by eyelie permutation of symmetrie matriees, and a
dagger (t) that it also ineludes invariants formed by eyelie permutation
of skew-symmetrie matriees.

10. INVARIANTS OF SECOND-ORDER TENSORS AND VECTORS; FULL ORTHOGONAL GROUP

For the full orthogonal group it is neeessary to distinguish axial

from absolute veetors. We assoeiate axial veetors with seeond-order
anti-symmetrie tensors (whieh transform in the eorreet manner; see (9),
(10)) but do not make this assoeiation for absolute veetors. Invariants
Isotropie Invariantsand Tensor Functions 163

of the forms (2la,b) are absolute invariants under the full group, and
those of the forms (22a,b) are only relative invariants under the full
( 1) ( 2) ( 7) .
group, where now TI. . , TI .. , ... , TI. . are componen ts of ma tr1.x products
l.J l.J l.J
in symmetric matrices and in anti-symmetric matrices associated with
anti-symmetric tensors and axial vectors. General tensors can be dealt
with by decomposing them into sums of symmetric and anti-symmetric
tensors.

To obtain an integrity basis for the full group it is not sufficient

to eliminate odd invariants from a basis for the proper group, because
some invariants which are reducible when regarded as invariants of the
proper group are irreducible when regarded as invariants of the full
group.

We have to consider invariants of the forms (2la) and (2lb).

Invariants of the form (2la) are as for the proper orthogonal group. For
( 2)
invariants of the form (2lb), that is U, TI, , V. or, as a special case,
l. l.J J
( 2)
u.TI .. u., we set
l l.J J
T
A (u. u.), c ( U. V . +u . V, ) , X -X ~(u.v.-u.v.).
l J l J J l l J J l

(54)

Then
(2) ( 2)
U, TI, . V, = tr (~+~) ~ 2 , u.TI .. u. = trATI 2 (55)
l. l.J J - - - l. l.J J --

so that it is sufficient to consider invariants of the forms

tr ATI , tr CTI , tr XTI , (56)

where A and C are symmetric, and X is anti-symmetric. However, invariants

of the forms (56) were considered in Section 9, and listed in Table 1.
llence the required results can be read off from Table l. The complete
list of invariants is long; it is given in Smith [15], but is not
reproduced here.

The integrity bases derived in Sections 8, 9 and 10 by reduction of

traces of matrix products, using matrix identities based on the Cayley-
164 A.J.M. Spencer

Barnilton theorem and the Spencer-Rivlin identity, are in fact minimal

bases in the sense that none of their members is reducible. However, we
have so far given no proof of this minimality property. We refer to
this problern in Section 7 of Chapter 9.

11. ISOTROPie TENSOR POLYNOMIAL FUNCTIONS OF VECTORS AND TENSORS

A problern which arises frequently in continuum physics is the

following. Let F. . . be components of a tensor of order r, which
1 11 2 .. · 1 r ( 1) (2) (V)
are functions of the components u. . , u .. , ... , u. . of a set of V
(1) 1.~2) l.J ( A) l.J
vectors and the components Pmn, Pmn, ... , P of a set of A second-
mn
order tensors. For example, in the cases r 1 and r = 2 we have

f. F .. <!> •. (u~p) ,P(q))

1. l.J l.J J mn

(p = 1,2, ... v; q = 1,2, ... A). (57)

Under a coordinate transformation x Mx, the components of the vectors

and tensors become

f. :E', . M. M. F
1. l.J 1.r JS rs

-(p) - (q)
u. p .. M !1 P(q)
1. l.J im jm mn

We say that f., F .. are form-invariant under a group of transformations M

1. l.J
if, for each M of the group,

(58)
.t-1. M. <!> (u ~p) ,P (q))
1.r JS rs J mn

The extension to tensors of higher order is evident. If the

transformation group is the full or proper orthogonal group, we say that
f. and F .. are isotropic vector and tensor functions respectively of
1. 1.]
their arguments. We restriet ourselves to polynomial functions. If f is
a scalar, then the equivalent statement to (58) is that f is an invariant
Isotropie lnvariants and Tensor Functions 165

of u~p)
and P(q); the concept of a form-invariant tensor is therefore a
J mn
generalization of that of a scalar invariant.

To determine canonical forms for form-invariant tensor polynomials

we reduce the problern to the problern of determining integrity bases 1 which
has already been considered. The method was first described by Pipkin and
Rivlin [16]. Let vi 1 wi be components of arbitrary vectors 1 and form

J 1 =v.f. 1 (59)
~ ~

Evidently

f. F .. (60)
~ ~]

It is easily verified that J1 is an invariant Of V, 1 ~p) an d p mn

(q)
I and
~

linear in v. ; also that J2 is an invariant of V, I W, I ~p) and P (q) 1 and

~ ~ mn ~

bilinear in V. and w .. Hence J1 and J 2 can be expressed in the forms

~ ~

J (2)
J1 I I J ( 1)
Cl. Cl.
J2 I I
Cl. Cl.
Cl. Cl.

where Ia are polynomials in invariants of the vectors ~p) and P~); J ( 1)

Cl.
are invariants of Vi 1 ~p) and P~~) which are linear in vi; and Jci 2) are
invariants of vi 1 wi 1 ~p) and P~~) which are bilinear in vi and wi.
Hence 1 from (59) and (60) 1

f.
~
\' I 3J < 1 l ;av
L. Cl. Cl. i F ..
~J
=I (61)
Cl. Cl.

This gives canonical forms· for fi and Fij although 1 because of the
possible existence of syzygies 1 the expressions may contain redundant
terms even if J(a. 1 ) and J(a. 2 ) are e 1 emen t s o f an ~rre
· d uc~'bl e ~n
· t egr~' ty

basis. The extension to tensors of higher order is obvious. The use of

the differentiation operation is only for convenience; the process may be
described easily in purely algebraic terms.

The method needs modification if F .. (or a tensor of higher order)

~]

has index symmetries. If 1 for example 1 F .. is symmetric 1 there are

~]

several possible ways to proceed.

 166 A.J.M. Spencer

(a) Construct F .. as in (61); then interchange i and j to form F .. , and

l.J Jl.
take

F .. ~(F .. +F .. )
l.J l.J Jl.

This expression will often contain redundant terms.

(b) Instead of forming J2 as in (59), form

J'2 V. v.F ..
l. J l.J
(p) p (q)
Then J'2 is an invariant of V.' ~, and quadratic in V.' and
l. mn' l.
F .. (62)
l.J

and F .. is symmetric with respect to interchanges of i and j.

l.J
(c) Insteady of forming J 2 as in (59), form

J"2 K .. F ..
l.J l.J

where K .. are components of an arbitrary symmetric tensor, and J2 is

l.J
expressed as a symmetric function of K .. and K ... Then J2 is an
l.J Jl.
invariant of K Q(p) P(q) and linear in K ..• Hence
ij' k ' mn ' l.J
F .. 8J;/K .. = 8J~/8K .. (63)
l.J l.J Jl.

so F .. has the required index symmetry.

l.J

Methods analogous to (a) and (c), but not to (b), can be exployed
when F .. is anti-symmetric.
l.J

As a simple example let us use method {c) above to find the form-
invariant symmetric tensor function of a single symmetric tensor A. From
the tables, the invariants of A and K are

trA, trA 2 , trA 3 , trAK, trA 2 K, trK, trK 2 ,

-
trK 3 , trAK 2 , trA 2 K2
- •

Since we require J2 to be linear in K, we retain only the first six of

these. Hence J2 has the form

J~ I 1 tr ~ + I 2 tr ~ + I 3 tr ~ 2 ~ ,

where I 1, I 2 and I 3 are polynomials in tr A, tr A2 and tr A 3 • Using (63)

Isotropie lnvariants and Tensor Functions 167

then gives

or

(64)

which is a well-known result.

In a similar manner we can form symmetric tensor functions of any

nurober of symmetric tensors. Full results are given in [6]. For two
symmetric tensors A and B the result is

(65)

where I are invariants of A and B. This result was obtained directly by

0.
using results in matrix polynomial theory by Rivlin [9], and its extension
for an arbitrary nurober of tensors by Spencer and Rivlin [17, 18] and
Spencer [19].

As a further example, consider an anti-·symmetric tensor function X

of two symmetric tensors A and B, for the proper orthogonal group. We
form

K 2X .. W.. X .. (W .. -W .. )
~J ~J ~J ~J J~

where W.. are components of an arbitrary anti-symmetric tensor and K is

~J

expressed as a symmetric function of W.. and -W.. . Then, from tables of

~J J~

invariants of W, A and B, K is a linear combination of the invariants

with coefficients I which are invariants of A and B. Hence X has the

0.
form
168 A.J.M. Spencer

X 11 (~-~~) + 12 (~2~-~~2) + 13 (~2~_~2) + 14 (~2~ 2 _~2~2)

+ 15 (~2~~-~~2) + 16 (~2~-~~2) + 17 (~2~2~_~2~2)

+ 18 (~2~2~-~~2~2) • (66)

To obtain a vector function u of A and B we simply associate u with the

anti-symmetric tensor X, and set

Several other examples of symmetric tensor, anti-symmetric tensor and

vector functions of symmetric tensors, anti-symmetric tensors and vectors,
under both the full and the proper orthogonal groups, are given in [6].

HEFERENCES

1. GRACE, J.H. and YOUNG, A., The Algebra of Invariants, Cambridge

University Press, 1903
2. ELLIOTT, E.B., Algebra of Quantics, Oxford University Press, 1913
3. TURNBULL; H.W., The Theory of Determinants~ Matrices and Invariants,
3rd edition, Dover Pub1ications, 1960
4. WEYL, H., The Classical Groups, Princeton University Press, 1939
5. GUREVICH, G.B., Foundations of the Theory of Algebraic Invariants
(Eng1ish trans1ation), Noordhoff, 1964
6. SPENCER, A.J.M., Theory of Invariants, in: Continuum Physics~ Vol.I
(ed. A.C. Eringen), Academic Press., 1971: 239-353
7. GREEN, A.E. and ADKINS, J.E., Large Elastic Deformations, Oxford
University Press, 1960
8. SMITH, G.F. and RIVLIN, R.S., The anisotropic tensors, Q.Appl.Math.
15 (1957) : 309-314
9. RIVLIN, R.S., Further remarks on the Stress-deformationrelations for
isotropic materials, J.Rat.Mech.Anal. 4 (1955): 681-701
10. S~UTH, G.F., On the generation of integrity bases, Atti Accad.Naz.
Lincei, Ser.8,9 (1968): 51-101
11. SMITH, G.F., The generation of integrity bases, in: Non-linear
ContinuUm Theories in Mechanics and Physics and their Applications
(Coordinator R. S. Riv1in), Edizioni Cremonese, Roma, 1970: 311-351
U. SPENCER, A.J.M. and RIVLIN, R.S., Isotropie integrity bases for vectors
and second-order tensors, Part I, Arch.Rat.Mech.Anal. 9 (1962): 45-63
Isotropie Invariants and Tensor Functions 169

13. SPENCER, A.J.H., Isotropie integrity bases for vectors and second-
order tensors, Part II, A~ch.Rat.Mech.Anal. 18 (1965): 51-82
14. RIVLIN, R.S. and SMITH, G.F., On identities for 3 x 3 matrices, Rend.
di Matematica Se~ie VI, 8 (1975): 345-353
15. SMITH, G.F., On isotropic integrity bases, A~ch.Rat.Mech.Anal. 18
(1965) : 282-292
16. PIPKIN, A.C. and RIVLIN, R.S., The formu1ation oi constitutive
equations in continuum physics I, A~ch.Rat.Mech.Anal. 4 (1969): 129-144
17. SPENCER, A.J.M. and RIVLIN, R.S., The theory of matrix po1ynomia1s
and its app1ication to the mechanics of isotropic continua, A~ch.Rat.
Mech.Anal. 2 (1959): 309-336
18. SPENCER, A.J.l1. and RIVLIN, R.S., Further resu1ts in the theory of
matrix po1ynomia1s, A~ch.Rat.Mech.Anal. 4 (1960): 214~230
19. SPENCER, A.J .H., The invariants of six symmetric 3 x 3 matrices, A~ch.
Rat.Mech.Anal. 7 (1961): 64-77
 Chapter 9

ANISOTROPIC INV ARIANTS AND ADDITIONAL RESUL TS FOR

INVARIANT AND TENSOR REPRESENT ATIONS

A.J.M. Spencer
The University of Nottingham, England

l. TRANSVERSE ISOTROPY

The transformation groups which characterize transverse isotropy

were listed in (c) of Section 3 of Chapter 8. For simplicity, we consider
Case (ii), which is invariance under the group generated by the rotations
~e and the reflection ~3 • If only second-order tensors are considered, i t
is not necessary to distinguish between the five cases listed under (c) of
Section 3 of Chapter 8 .

The isotropic tensors olJ

.. and e .. k played an important part in
lJ
establishing isotropic integrity bases. Smith and Rivlin [l] constructed
the corresponding anisot~opic tenso~s for other groups of transformations.
For transverse isotropy a. satisfies (14) of Chapter 8, but now M is not a
general orthogonal matrix, but is ~3 or any matrix ~6 which represents a
rotation through 8 about the axis ~~· By the same argument as was used in
Section 3 of Chapter 8, the tensor a. is given by (16) of Chapter 8, where
now J is a transversely isotropic invariant of the vectors u , and is
-P
172 A.J.M. Spenccr

multilinear in their components. The transversely isotropic integrity

basis for a set of vectors is known; it is essentially the integrity basis
for two-dimensional vectors under rotations and reflections in two
dimensions 1 and consists of the scalar products u .u tagether with the
-P -Q
resolved components ~ 1 -~P of the vectors ~P in the direction ~ 1 . Hence a
typical term of J is

and it follows that the components of a are of the form

\ ( 1) ( 1) ( 1)
L ß Ö ~ Ö ~ •• • ~ V V ••• V
n n k1 1 k2 2 kq q m1 m2 mr

where ßn are numerical coefficients and k 1 1 ~ 1 1 k 21 ~2 1 k

q
I ~
q
1 mjl

m21 ... 1 m are permutations of i 1 1 i 2 1 • • • I i .

r p

Now let I be a homogeneaus polynomial transversely isotropic

invariant of V vectors ~ and A tensors ~s 1 so that I has the form (19)
of Chapter 8. Then by arguments similar to those of Section 6 of Chapter
8 it follows that the ßs are now components of a transversely isotropic
tensor. Consequently I can be expressed as a polynomial in expressions of
the following types:

(a) (b)

(1)
( 1) ( 4) ( 1)
( c) (d) V. 1T .. V,
~ ~] J

where 1T~~) are expressions of the form (23) of Chapter 8. However 1 (1)
~]

simply states that I is a polynomial in the members of an isotropic

integrity basis for u 1 P and v 1 ; thus an isotropic integrity basis for
-R -S -
~R~ ~S and ~1 is a transversely isotropic intergrity basis for ~R and ~s·

It must be noted 1 however 1 that because of the possible existence of

syzygies 1 and because ~1 is required tobe a unit vector 1 an irreducible
isotropic integrity basis for ~R 1 ~s and ~1 is not necessarily irreducible
when it is regarded as a transversely isotropic integrity basis.
Anisotropie Invariantsand Additional Results 173

As an examp1e we determine the transverse1y isotropic integrity

basis for a single symmetric tensor A. This is the isotropic integrity
basis for the tensor ~ and the vector ~ 11 name1y
2
tr A1 tr A2 1
~l~ ~l I
(2)

and this basis is irreducible.

An alternative derivation of (1) can be based on physica1 arguments.

F'or definiteness 1 suppose I is a transversely isotropic po1ynomia1
invariant of a tensor A. Then I must be a po1ynomial in the components
of ~ and ~1 . But I must be unchanged if it is referred to a new
coordinate system x Mx. Thus

That is 1 I is an isotropic invariant of A and ~1 . This approach is

developed further in Chapter 10 •

A complete list of invariants of an arbitrary nurober of vectors 1

symmetric tensors and anti-symmetric tensors 1 for each of the five
transformation groups defined in (c) of Section 3 of Chapter 81 has been
given by Smith [2]. His method of derivation differs from the one
described here.

Tensor polynomia1 functions of vectors and tensors 1 which are form-

invariant under the transformations which describe transverse isotropy 1
may be derived by methods analogous to those used in Section 11 of
Chapter 8. For example 1 todeterminesuch a form-invariant symmetric
tensor function F of a symmetric tensor A1 we write down the most general
isotropic invariant of ~~ ~1 and a symmetric tensor K 1 which is linear in
K. This invariant (written as an invariant which is symmetric in K.. and
~J
174 A.J .M. Spenccr

where Ia are invariants of A and ~1 • Then F .. 3J/3K. . which gives

lJ lJ
F I1~+I2~+I3~2+It.::'1@::'1

+ I 5 (:: 1 0 ~::' 1 +:: 1~ 0:: 1 ) + I 6 (:: 1 0 ~ 2 :: 1+;: 1~ 2 0:: 1 ) · ( 3)

2. ORTHOTROPIC SYMMETRY

The symmetry group for orthotropic symmetry with respect to three

orthogonal planes normal to the three unit vectors ;: 1 , ;: 2 , ;: 3 is given in

(e) of Section 3 of Chapter 8. By arguments similar to those used in
Section 1, the anisotropic tensors for this symmetry group are generated
by the tensors

(4)

Since ;: 1 , ;: 2 and ;: 3 are mutually orthogonal, we have

and so we may discard one of the set ( 4) , say ;: 3 0;: 3 , in favour of the
unit tensor I and take the basic set of anisotropic tensors to be

It follows by arguments analogaus to those used in Section 1 that any

polynomial invariant, under this transformation group, of a set of

vectors ~R and tensors ~s' is an isotropic invariant of ~R' ~s' ~ 1 and

;: 2 , and of even degree in the components of ~ 1 andin the components of
::z· Thus in this case also the tables of isotropic invariants can be
used to generate integrity bases for the symmetry group under
consideration. For example, for a single symmetric tensor A, an integrity
basis consists of

trA, tr A 3 , (5)

tagether with squares and products in pairs of

However, since ;: 1 and :: 2 are orthogonal, we have ::~·::z 0, and it can be

Anisotropie Invariantsand Additional Results 175

2 2 2 2
shown that (~ 1 ~~ 2 ) , (~ 1 ~ ~ 2 ) and (~ 1 ~~ 2 ) (~ 1 ~ ~ 2 ) are reducible, and so
(5) constitutes the required integrity basis.

Ta obtain the most general tensor function F of a symmetric tensor A

which is form-invariant under the transformations which characterize
orthotropic symmetry, we first write down the most general isotropic
invariant of A, ~ 1 €>~ 1 , ~ 2 €>~ 2 and and arbitrary symmetric tensor K, which
is linear in K. This (written as an invariant which is symmetric in K ..
~J
and K .. ) is
J~

2
J = I 1 tr ~ + I 2 tr ~ + I 3 tr ~ ~ +I 4 ~ 1~~ 1

+ I5~1 (~+~)::1 + I6::1 (~ 2 ~+~ 2 )::1

(6)

where I 1 , I 2 , .•. , I 9 are polynomials in the invariants (5). In

constructing (6) it has been taken into account that ~1 and ~2 are
orthogonal unit vectors. Then, as before, F .. = 3J/3K .. , and hence
~J ~J

+I6(~1@~ 2 ~1+~1~ 2 @~1) +I7~2€)~2

+ Ia (~2 @~~2+~2~€)~2) + Ig (~2 @~ 2 ~2+~2~ 2 @~2) (7)

Same other examples of the construction of tensor polynomial

functions appropriate for transversely isotropic or orthotropic symmetry
will be described in Chapter 10.

3. CRYSTAL SY~~TRIES

The same general procedures can be used to determine integrity bases

under the symmetries which correspond to the thirty-two crystal classes.
Extensive results have been derived by G. F. Smith, Rivlin and
M. M. Smith. The number of cases which require attention for a full
treatment is very large, and in many of these cases the results are very
176 A.J.M. Spencer
complicated 1 so no attempt is made to summarize them here. References
to results obtained prior to 1968 are given in Spencer [3].

For a given crystal classl the first task is to establish an

integrity basis for an arbitrary number of vectors. This Wds achieved
for all the crystal classes by Smith and Rivlin [4 1 5]. To do this it is
necessary to consider polynomials 1 in the components of the vectors 1
which remain invariant under the transformations of the group under
consideration. The process is very lengthy 1 and involves the use of
several theorems in the theory of symmetric polynomials.

Having established integrity bases for vectors 1 it is then possible

to construct the anisotropic tensors for each group [1 1 4 1 5]. Knowledge
of the anisotropic tensors then enables general forms of invariants of
vectors and tensors to be written down. In special cases 1 it is then
possible to reduce these to minimal integrity bases. Full results are
known for vectors [4 1 5] 1 a single symmetric tensor (Smith and Rivlin
[6]) 1 a single symmetric tensor and a single vector (Smith 1 Smith & Rivlin
[7]) 1 and for N symmetric tensors (Smith and Kiral [8]). Further results
have been derived by Kiraland Smith [9].

In principle 1 a knowledge of the appropriate integrity bases leads

to the determination of form invariant tensor functions by the procedure
described in Section ll of Chapter 8 and also used in Sections l and 2 of
this chapter.

4. TENSORS OF THIRD AND HIGHER ORDER

Relatively little is known about integrity bases for tensors of order

higher than two 1 for any of the transformation groups of interest in
continuum mechanics. An exception is the set of results presented by
Betten in Chapter 11. Apart from these 1 there appears tobe no thorough
systematic treatment 1 although some isolated results appear in the
literature.
Anisotropie Invariantsand Additional Results 177

F'or definiteness let us consider the full orthogonal group, for

which all isotropic tensors can be expressed in terms of the Kronecker
delta ol.J
... By procedures analogous to those described in Section 6 of
Chapter 8 , all invariants can be obtained by forming inner products of
Kronecker deltas with the tensors under consideration. For example, for
a fourth-order tensor A, with components Aijk~' we may form the following
invariants of degree one in the components of A:

A ....
l.JJl.

Thus invariants are (in this case) obtained by contracting indices in

pairs until a scalar is obtained. Similarly, the invariants which are
bilinear in the components of two third-order tensors A and B, with
components Aijk and Bijk' are

AiijBjkk ' Al.l.J

... Bk J'k' AiijBkkj ,

AijiBjkk' AijiBkjk ' AijiBkkj ,

AijjBikk ' AijkBijk ' AijkBikj '

Aijlkik' AijkBjik ' AijkBkij '

Aijj 8 kki ' Aijk 8 jki ' Aijk 8 kji

Clearly, the existence of any index symmetries will reduce the number of
independent invariants.

As in the case of vectors and second-order tensors, the operation of

inner multiplication by isotropic tensors yields an infinite number of
invariants. By Hilbert's theorem, only a finite number of invariants can
be independent. Presumably the reduction to a finite and even a minimal
set of polynomial invariants may in principle be effected by application
of identities of the form
178 A.J.M. Spencer

6. 6. 6. 6,
l.p J.q J.r l.S
6. 6. 6. 6.
JP ]q ]r JS
0 ,
6kp 6kq 6kr 6ks

69,p 69,q 69,r 69,s

but the details are certain to be algebraically complicated.

In a similar manner, infinite integrity bases can be generated by

using the appropriate isotropic or anisotropic tensors, for invariance
under any other group of orthogonal transformations. The extension to
the construction of form-invariant tensor functions is also Straight-
forward in principle, but sure to be complicated in practice.

5. REDUCTION OF A GENERAL TENSOR TO A SUM OF TRACELESS SYMMETRIC TENSORS

In constructing integrity bases andin related problems, it is often

desirable to express tensor quantities in their simplest form, that is
with the smallest possible number of independent components. Thus it is
best to decompose a general second-order tensor, with 9 components, into
a symmetric tensor (6 components) and an anti-symmetric tensor (3
components). Since the trace of a symmetric tensor is a scalar, a
symmetric tensor can be reduced to a scalar and a traceless (deviatoric)
tensor with 5 independent components. Also, an anti-symmetric tensor can
be associa ted wi th a vector. Thus, for the componen ts P. . of a general
l.J
tensor P, we may write

P .. A .. +e .. kuk+w6 .. ,
l.J l.J l.J l.J

where

A .. (5 independent components) ,
l.J

le P ( 3 componen ts) ,
2 rsk rs

w (l scalar),
Anisotropie Invariantsand Additional Results 179

whieh express P in terms of the traeeless symmetrie tensor A, the veetor

u, and the sealar w.

This proeess of redueing tensors to their simplest forms is even

more desirable when dealing with tensors of higher order. An elementary

method of redueing tensors of any order to sums of traeeless symmetrie

tensors is deseribed in Speneer [10]. A related problern is treated in a
more general way by Hannabuss [11].

We first express the tensor as a sum of fully symmetrie tensors,

i.e. tensors whieh are symmetrie with respeet to interehange of any pair
of indiees. Veetors and sealars are regarded as fully symmetrie tensors
of orders one and zero. We use braeketed indiees to denote the index

symmetrization operation, so that, for example,

1
A(;J'k) = 6(A .. k+A.k.+A .. k+A.k.+A. .. +A. .. ) .
~ ~ J ~JJ~ J ~ k~J kJ~

Then A(ij)k is symmetrie for interehanges of (ij) and A(ijk) is symmetrie

for any permutation of (ijk) .

Suppose A. . are eomponents of a tensor of order n, and

... kipq ... s
~J
that R, is the (r-l)th index and p the rth index. Then

A(ij ... ki)pq ... s . . ... k x-p

rA ( ~J 0 )
q ... s - A ( J. ... k x-p)~q
0 •
... s

-A. n. - ••• - A . . n
(L .. kx-p)Jq ... s (~J ... kp)x-q ... s

Now add (r-l)A(. . k") to eaeh side, and note that for any pair
... ~J X, pq ... s
of indiees t and u

e e A.
Ai ... t ... u ... s- Ai ... u ... t ... s tuv xyv ~ ... x ... y ... s

We also denote

= e . . k) vq ... s
B.. k
~J ... q ... sw uvwA( U~J...

Then B.. are eomponents of a tensor B of order n-1 whieh is symmetrie

~J ... w
180 A.J .M. Spencer

in its first (r-2) indiees. Henee

rA(ij ... kt)pq .•. s rA(.. "'P q •.. s +e.1pwB.] ... knJVq ... sw
1] ... kn)

+e. B. n + ... +e 0 B ..
JPW 1 ... kJVq ... sw "'PW 1] ... q ..• sw

Thus the tensor A symmetrized in its first r-1 indiees is the sum of A
symmetrized in its first r indiees, and tensors formed by eontraeting B
with the third order a1ternating tensor. By repeated app1ieation of this
proeess, A ean be expressed entire1y in terms of symmetrie tensors of
order n or 1ess, and isotropie tensors.

For examp1e, the eomponents Aijk of a third-order tensor ean be

expressed in this way as

12A. 'k = 12A(. 'k) +4e.k c(.) +4e.k c(.) +6e .. B(k)
1J 1J 1 p JP J p 1p 1JP p

+3(6.ku.-6. u.) +6.kv. +6.kv. -26 .. vk,

J 1 1k J J 1 1 J 1]

where

e A
prs rsk
, c.1p = e A
rsp (ri)s
0) ,

u e B A -A v = e C A -A •
p rsp rs prr rpr p rsp rs (pr)r rrp

In the most genera1 ease, Aijk has 27 independent eomponents, A(ijk) has
10, B(ij) has 6, C(ij) has 5, ui has 3, vi has 3. The deeomposition is
not unique.

A symmetrie tensor of any order is traee1ess if the va1ue zero is

obtained as a resu1t of eontra~.l:ing on any pair of indiees. Any
symmetrie tensor may be deeompos~d into traee1ess symmetrie tensors. For
examp1e, for a third-order symmetrie tensor ~

A. 'k =
1J
A~ .k+~(A .6.k+A . k.+A k6 .. )
1J pp1 J PPJ 1 PP 1J

This defines the tensor A' whieh has the property 'k = 0. Then A' has
A~
11
seven independent eomponents. In genera1 a traee1ess symmetrie tensor of
ordern has 2n+1 independent eomponents.
Anisotropie Invariants and Additional Results 181

These decompositions can be extended to tensors of any order.

Detailsare given in [10].

6. LINEARLY INDEPENDENT INVARIANTS - GENERATING FUNCTIONS

The procedure followed in Sections 4-10 of Chapter 8 was to construct

integrity bases which usually contain redundant elements 1 and then by
examination of particular cases to reduce as many as possible of the
elements. In this way we hope to arrive at a minimal basis but (except
in the siruplest cases) the procedure gives no assurance that the basis is
minimal. For the integrity bases derived in these lectures 1 minimality
can be established by methods of the theory of group representations.
Space does not permit a detailed description of these methods 1 but we
state some of the principal results. For details we refer to Smith [12 1
13].

Consider first a single traceless symmetric tensor ~ of order n under

the proper orthogonal transformation group in three dimensions. The
tensor has 2n+l = p independent components. These components in a
coordinate system ~~ arranged in some definite order 1 are regarded as
elements of a p-dimensional vector s-n .
Suppose a proper orthogonal transformation of coordinates

X = Mx

induces in the vector s

-n
the transformation

~
-n
= L
-n-n
s I

where ~~ = !:_1 and :::n is a p x p matrix.

Now form all the distinct products of degr~e m in the components of

A1 and arrange these in some definite order as components of a vector
§~m) . The coordinate transformation induces a transformation of ~~m) 1

which is of the form

182 A.J.M. Spencer

~ (m) = L (m) E;, (m) .

-n -n -n

Then it is known that the number P(m) of linearly independent invariants

of degree m in the components of A is given by

1 r21T
j
(m)
P(m) 2 1T tr!::'n (l-cos8) d8,
0

where e is the rotation angle represented by the matrix !-1.

tr L (m) is the coefficient of am in the formal

It is also known that
-n i8
expansion in positive powers of a of ~(a,e ,n), where

-n+l
= {(1-at-n )(1-at
-1
~(a,t,n) ) .•. (1-at )(1-a)(l-at) ...

n }-1
... (1-atn-1 ) (1-at)
n
II {(1-ats)- 1 } .
s=-n

Hence

00

L tr L (m) am <P (a,e

i8
,n) •
-n
m=O

It follows that

1 r2 n
00
i8
L P(m)am 2 1TJ 0 ~(a,e ,n)(l-cos8)d8
m=O

n
= _!_ J21T II {(l-aeisB)- 1 }(1-cos8) d8,
21T 0 s=-n

and hence the integral on the right-hand side above is a generating

function for the number P(m) of linearly independent invariants of A
which have degree m.

These results extend to sets of tensors. If A 1 ,A 2 , ••• ,A are a set

- - -11
of traceless symmetric tensors of orders n 1 ,n 2 , ••• ,n, then the number of
jl
their linearly independent invariants of degree m1 in the components of
Anisotropie lnvariants and Additional Results 183

~ 11 m2 in the components of ~ 21 and so on 1 is

P ( m1 1m 2 1 ••• 1m ) = __!_ J2TI

].1 2TI
0

and it fo11ows that

1 r2n J.1 i8
= 2TI j0 TI
r=1
{<j)(a
r
1 e 1 n ) } (1- cos 8) d8
r

d8.

Some examp1es of these generating functions are as fo11ows:

(a) One vector u. Let A = u. Then

00

I P ( m) u
m = -
1 jr2n (1-cos8) d8
-----'-:;::------=-=-=---='-'----==---:-~
2TI i8 -i8
0 ' 0 (1-ue ) (1-u) (1-ue )

(b) One trace~ess symmetric second-order tensor A. Then

00
2 (1 8) d8
\ P(m)am = __!_ J 'TT --~~--~--~c~o~s~~~~------~~
L. 2TI 2i8 i8 -i8 -2i8
0 0 (1-ae ) (1-ae ) (1-a) (1-ae ) (1-ae )

(c) Thlo vectors u1 v. Let ~1 U1 ~2 = v. Then

= J.:_
2TI
J2TI --~~----(~1_-~c_o_s_8~)~d~8~-----~
i8 -i8 i8 -i8
0 0 (1-ue ) (1-u) (1-ue ) (1-ve ) (1-v) (1-ve )
184 A.J .M. Spencer

(d) A vector u = ~ 1 and a tracetess symmetric tensor A ~2 • Then

1 r2TT ( 1- cos 8) d8
2 Tl J i8 -i8 2i8 i8 -i8 -2i8
0 (1-ue ) (1-u) (1-ue ) (1-ae ) (1-ae ) (1-a) (1-ae ) (l-ae )

Generating functions are particularly useful because they indicate

the existing of syzygies. For example, for three vectors in two
dimensions, the generating function can be evaluated as

The denominator indicates the existence of invariants of degrees (2,0,0),

(0,2,0), (0,0,2), (1,1,0), (1,0,1) and (0,1,1) (these are the six scalar
products). The numerator reveals a syzygy of degree (2,2,2), which is in
fact (for vectors in two dimensions)

u.u u.v u.w

v.u v.v v.w 0 .

w.u w.v w.w

The natural method of evaluating the integral generating functions

is by contour integration, but except in the simplest cases the evaluation
and summation of the residues is very tedious. Many further results are
available because of an analogy between the linear group in two dimensions
and the orthogonal group in three dimensions. Classical invariant theory
yields generating functions for numbers of invariants under the two
dimensional linear group and many cases were worked out, at great length,
by Sylvester and Franklin (for references, see Spencer [14]). These
generating functions are algebraic, and involve no integration, but the
effort needed to apply them is similar to that required to use the
integral generating function. However, the known results can be
interpreted in terms of the three-dimensional orthogonal groups. The
connection between tl1e types of generating functions is explored in [14].
Anisotropie lnvariants and Additional Results 185

7. MINHlALITY OF AN INTEGRITY BASIS

For a given set of vectors and tensors 1 and for invariants of

specified degrees in these vectors and tensors 1 let

k number of linearly independent invariants;

S number of invariants in a minimal integrity basis;

ß 1 number of invariants in an integrity basis under consideration;

e number of invariants which can be contructed from invariants

of lower degree from a minimal basis;

n number of syzygies between the e invariants;

nl number of known syzygies between the e invariants.

Then s< s I I n I <; n I k = s+ e- n. Hence

s~ ~ S k-8+n ?k-8+nl .

Now ß 1 and n 1 are known 1 k can be calculated by the methods outlined in

Section 6 of Chapter 9 and 1 working by ascending degrees 1 8 can be
calcu1ated. Thus if

s~ k-8+n 1

then SI s and nl = nl and the minima1ity of the basis is proved for

invariants of the specified degrees. The process is app1ied successive1y
to invariants of increasing degrees. For a finite basis the process
eventual1y terminates.

Minimality of the integrity basis 1isted in Table 1 has been proved

by Smith [12 1 13] in this manner.

REFERENCES

1. SMITH 1 G.F. and RIVLIN 1 R.S. 1 The anisotropic tensors 1 Q.App~.Math. 1

15 (1957) : 309-314
2. SMITH 1 G.F. 1 On transverse1y isotropic functions of vectors 1 symmetric
186 A.J .M. Spencer

second-order tensors and skew-symmetric second-order tensors, Q.App~.

Math. 39 (1982) : 509-516
3. SPENCER, A.J.M., T.heory of Invariants, in Continuum Physics, Vo~.I
(ed. A.C. Eringen), Academic Press, 1971
4. SMITH, G.F. and RIVLIN, R.S., Integrity bases for vectors- the
crystal classes, Arch.Rat.Mech.Ana~. 15 (1964): 169-221
5. SMITH, G.F., Tensor and integrity bases for the gyroidal crystal
classes, Q.App~.Math. 25 (1967)_: 218-221
6. SMIT8, G.F. anq RIVLIN, R.S., T.he strain-energy function for
anisotropic elastic materials, Trans.Amer.Math.Soc. 88 (1958): 175-193
7. SMITH, G.F., SMITH, M.M. and RIVLIN, R.S., Integrity bases for a
symmetric tensor and a vector - the crystal classes, Arch.Rat.Mech.
Anal. 12 (1963): 93-133
8. SMITH, G.F. and KIRAL, E., Integrity bases for N symmetric secend-
erder tensors - the crystal classes, Rend.Circ.Mat.Pa~ermo II, Ser.l8
(1969): 5-22
9. KIRAL, E. and SMITH, G.F., On the constitutive relations for
anisotropic materials - triclinic, monoclinic, rhombic, tetragonal and
hexagonal crystal systems, Int.J.Engng.Sci. 12 (1974) :471-490
10. SPENCER, A.J.M., A note on the decomposition of tensors into traceless
symmetric tensors, Int.J.Engng. Sei. 8 (1970) :489-505
11. HANNABUSS, K.C., T.he irreducible components of homogeneaus functions
and symmetric tensors, J.Inst.Math.App~ics. 14 (1974): 83-88
12. SMITH, G.F., On isotropic integrity bases, Arch.Rat.Mech.Ana~. 18
(1965): 282-292
13. SMITH, G.F., On the minimality of integrity bases for symmetric 3 x 3
matrices, Arch.Rat.Mech.Ana~. 5 (1960): 382-389
14. SPENCER, A.J.M., On generating functions for the number of invariants
of orthogonal tensors, Mathematika 17 (1970): 275-286
 Chapter 10

KINEMATIC CONSTRAINTS, CONSTITUTIVE EQUATIONS

AND FAlLURE RULES FOR ANISOTROPIC MATERIALS

A.J.M. Spencer
The University of Nottingham, England

1. KINEI-1ATIC CONSTRAINTS

It is common in many branches of continuum mechanics to treat

material as though it is incompressible. Although no material is truly
incompressible, there are many materials in which the ability to resist
volume changes greatly exceeds the ability to resist shearing
deformations; examples are liquids with low viscosity, like water, and
some natural and artificial rubbers. For such materials, the assumption
of incompressibility is a good approximation in many circumstances, and
often greatly simplifies the solution of specific problems. It should
be noted, though, that there are occasions when even a small degree of
compressibility may produce a major effect; an example is the propagation
of sound waves in water.

Incompressibility is an example of a kinematic constraint; it

restricts the range of admissible deformations. Another example is the
constraint of inextensibility in specified directions. Some highly
188 A.J.M. Spencer

anisotropic materials exhibit strong resistance to extension in particular

directions, compared to their shear resistance and resistance to extension
in other directions. Obvious examples are fibre composite materials
composed of strong stiff aligned fibres reinforcing a relatively soft
matrix. Materials of this kind may, approximately, be treated as
inextensible in the fibre direction, and analysis of their behaviour is
often greatly simplified by making this approximation. As in the case of
incompressibility, some caution is needed, because slight inextensibility
can produce large effects. However, the approximation is often useful,
and results derived from i t can be used as a basis on which to construct
more accurate solutions. The mechanics of these ideal fibre-reinforced
materials is described in [1].

We refer quantities to a fixed reetangular coordinate system. A

typical particle has position vector X and coordinates X in its reference
R
configuration at time t = 0. At a subsequent time t the same particle
occupies the position x with coordinates x.. The deformation is
l
described by the dependence of x on X and t, thus

X= x(X,t), or X.
l

The deformation gradient tensor F has components FiR' where

F.
1R
= ax. ;ax
1 R
.

We employ the finite strain tensors C and B, with components CRS and B ..
l]
respectively, where
T
c F F, B FFT

dX. dX.
l
dX. dX.
-- l
-- ___2:. _ 2
CRS FiRFiS B .. F. F.
axR axs l] lR JR axR axR

and also the infinitesimal strain tensor E, with components E .. , where

l]

ax. Clx
l +_2 -eS
.!. ( __
.J
E E,.
lJ 2 ax.J ax.l ij ·

The velocity v is regarded as a function of x and t. The rate-of-

Kinernarie Constraints and Constitutive Equations 189

deformation tensor has components Do 0' where

~J

- l. (-~- ___2
avo av OJ
Do~J 0 - 2 "oxo +"OXo .
J ~

The condition that the material is incompressible is that, for all

possible deformations,

det F = 1 , or det B = 1 , or det C = 1 •

In terms of the rate-of deformation tensor, the incompressibility

condition is

Do 0 tr D 0 .
~~

Within the approximation of infinitesimal theory, the incompressibility

condition is

Eo
~~
0 trE o.

Let ~0 be a unit vector field in the reference configuration. A

material line element with direction ~0 in the reference configuration has
the direction of a unit vector a in the deformP.d configuration, where

( 0) ( 0)
A.a or, in components, A.ao a 3xo /3X ,
~~0 ' ~
aR FiR R ~ R

where A. is the stretch of the line element. It follows that

( 0) ( 0)
a a C
R S RS

If the material is inextensible in the direction ~0 in the reference

configuration, then A. = 1, and

( 0)
a 3xo/3X , 1 .
R ~ R

In terms of D, the inextensibility condition is

aoaoD.. 0.
~ J ~J

Within the approximation of infinitesimal theory, it is not necessary to

190 AoJ.Mo Spencer

distinguish a and ~ 01 and the inextensibility condition is

aoa OEO 0.
J l.J
0

1.

The mechanical effect of a kinematic constraint is to produce a

reaction stress. Thus in a constrained body the stress T can be expressed
as the sum of a reaction Stress R and an extra-stress S 1 so that

T S+R 1 or 1 in components 1 To So +Ro

l.J l.J l.J
0 0 0

For an incompressible material the reaction is an arbitrary hydrostatic

stress of the form

R -p~ Ro
l.J
0 = -po l.J
0 0 •

The reaction to an inextensibility constraint is an arbitrary tension T

in the inextensible direction (for convenience we shall in future call
this direction a fibre direction and its trajectories fibres) 1 so in this
case

R Ta®a 1 Ro Taoao
l.J
0

1. J

The reaction stress does no work in a deformation which satisfies

the constraint. The rate of working of the reaction stress is Ro ODO 0.
l.J l.J
In an incompressible material

Ro ODO 0
l.J l.J
0 1

and in a material with an inextensible direction a

Ro ODO o Taoa ODO 0.

l.J l.J J l.J
0

1.

If a material is subject to two or more constraints the total

reaction stress is the sum of the reactions of each of the constraints.
F'or examplel in a material which is incompressible and inextensible in
two directions defined by vectors a and b 1
Kinernarie Constraints and Constitutive Equations 191

where Ta and Tb represent arbitrary tensions.

The reaction pressure p and tension T are arbitrary in the sense

that they are not given by constitutive equations but are determined by
equilibrium or momentum equations and boundary conditions. In an
incompressible material we may, without loss of generality, require that
S.]_]_. = tr S = 0, and in a material wi th an inextensible direction a that
a.a.S .. = 0. The remainder of the extra-stress S requires a constitutive
]_ J l.J
equation. Generally speaking, the constitutive equation becomes
relatively less significant as the number of constraints increases. The
extreme case is a rigid body, which is totally constrained ~d in which
the total stress is an indeterminate reaction stress.

2. LINEAR ELASTICITY

In a linear elastic material the strain-energy W is a quadratic

function of E. For a transversely isotropic material with preferred
direction a (whose sense is immaterial) we require

W(E, a€>a)

for allorthogonal M. Hence, by the results of Section 8 of Chapter 8,

W has the form

where A, ~T' ~L' a and ß are elastic constants. If the material is

inaompressible, then tr ~ = 0, and

W = ~ trE 2 +2(~ -~ )aE 2 a+-21 ß(aEa) 2 ,

T L T -- - ---

and only three elastic constants ~T' ~L and ß are required. If the
material is inextensible in the direction a, then aEa = 0, and

In this case the material is characterized by three elastic constants A,

192 A.J .M. Spencer

~L and ~T. If the material is incompressible and inextensible in the

fibre direction 1

W = ~ tr E 2 + 2 (~ -~ ) aE 2 a 1
T ~ L T ~~ ~

and only the shear moduli ~T and ~L are required to specify the material.

In each case the extra-stress S .. is given by

l.J
s. . ClW/ClE .. I
l.J l.J

so we have the following results.

Incompressible material
S = 2~ E+ß(aEa)a@a+2(~ -~) (a0aE+Ea0a)
~ T- ~~~ ~ ~ L T - ~~ ~~ ~

Material which is inextensible in direction a

S
~
= ,\ItrE+2~ E+2(~ -~)
~ ~ T~ L T
(a0aE+Ea0a).
- -~ -~ ~

Material which is incompressible and inextensible in direction a

S 2~ E + 2 (~ -~ ) (a0aE+Ea0a) .
T~ L T ~ -- ~~ -

In each case the appropriate reaction stress must be added to give the
total stress.

Similar results apply in the case of a material reinforced by two

families of fibres 1 whose dire~tions are specified by two unit vectors a
and b. In this case W is a function of

tr E 1 aEa 1

bEb I (a.b) 2 cos 24; aEb

where cos 24; is the angle between the two fibre directions. !t is then
Straightforward to write down the most general function W which is a
quadratic in E formed from the above list. If the material is
Kinematic Constraints and Constitutive Equations 193

incompressible, then tr E = 0, and i f i t is inextensible in the directions

a and b then aEa = 0 and bEb = 0, and the expression for W is simplified
accordingly. Orthotropic symmetry obtains if either (a) the fibres are
orthogonal, so that a.b = 0, or (b) the families of fibres are
mechanically equivalent, and then W is a symmetric function of a and b.

3. FINITE ELASTICITY

For a finite elastic material, W W(CRS), and the extra-stress is

given by

s .. (1)
lJ

where p 0 and p denote densities in the reference and deformed

configurations respectively.

For a transversely isotropic material, W is a function of the

invariants of ~0 and C, namely

tr C , detC

If the material is incompressible, then I 3 = 1. If it is inextensible in

the initial direction ~0 , then I 4 = 1. This leads to the following
results. In each case Wa = 3W/dia.

Incompressib~e materia~

Materia~ inextensib~e in initia~ direction a 0

S 2I~~{ (I 2 W2 +I 3 W3 ) _:+W 1 ~-I 3 W 2 ~- 1 +W 5 (~0~~+~~0~)}

194 A.J.M. Spencer

Material incompressibZe and inextensibZe ~n initiaZ direction ~0

s 2{w 1 ~-W 2 ~- 1 +W 5 (~0~~+~0~)}

For a material reinforced by two farnilies of fibres, defined by unit

vectors ~0 and ~0 in the reference configurations, W is a function of

I 3 '

cos 21P ~0~~0 and

If the material is incompressible, then I 3 = lo If it is

inextensible in the two fibre directions, then I 4 =1 and I 6 = lo If the
two fibre families are mechanically equivalent, then W is a symmetric
function of ~0 and ~0 o

For example, for a material which is incompressible and inextensible

in the two mechanically equivalent fibre directions,

In the case of an elastic material, the reaction stress may be

derived from the strain-energy function by regarding p and T as Lagrangian
multiplierso Thus for an incompressible material, with I 3 1, we replace
W by w-tp(I 3 -l), and the appropriate results then follow from (1), with
S .. replaced by T .. o Sirnilarly, for a material which is inextensible in
~] ~]

the direction ~ 0 , we replace W by W+tT(I 4 -l) and proceed in the same wayo

4o PLASTICITY - YIELD CONDITIONS

Most theories of plasticity assume the existence of a yield

conditiono We postulate a yield function f(T .. ) , suchthat
.
in admissible
~)

stress states f ~ 0, with f = 0 when deformation is taking placeo In

addition, f may depend on one or more parameters, which in turn may
depend on the deformation historyo
Kinematic Constraints and Constitutive Equations 195

If the material is isotropic, then f can be expressed as a function

of the stress invariants tr T, tr T2 and tr T 3 • In isotropic metal
plasticity, it is observed experimentally that for many materials yielding
is effectively independent of a superposed hydrostatic pressure. This is
incorporated in the theory by restricting f to depend on the deviatoric
stress S, where

s T- .!r tr T .
- 3_ -

Then tr S 0 and f can be expressed as a function of tr s 2 and tr s 3 • We

note that S is the extra-stress for an incompressible material.

For a material reinforced by a single family of fibres, with

direction a, f is a function of the invariants of T and a0a. For a
metal reinforced by inextensible fibres it is reasonable to expect that
yielding is not affected by a superposed hydrostatic pressure, or by a
superposed tension in the fibre direction, since such a tension produces
no stress in the matrix. These conditions can be incorporated by
assuming f to depend on T only through the extra-stress S, where

T R+S 1 R -pi + Ta0a.

The indeterminacy in S is removed by imposing the conditions

tr S Ü I aSa = 0 1

and then

The invariants of S and a®a are

J 2 = aS 2 a 1 J 3 = tr ~ 3

and f is a function of J 1 1 J 2 and J 3 • Two particular formsoff which

have been found to give good agreement with experiment are

(2)

and
196 A.J.M. Spencer

( 3)

Here kT and kL represent shear yield stresses for shear in directions

normal and parallel to a respectively. The above forms are natural
generalizations to transversely isotropic material of von Mises' and
Tresca's yield conditions respectively. For a perfectly plastic material
kT and kL are constants; for a hardening material they depend on the
deformation history.

For a material reinforced by two families of fibres, characterized

by unit vectors a and b, the corresponding results are

T R+S R = -pi +T a0a+T b0b,

~ a~ ~ b~ ~

with

tr S 0 1 aSa 0, bSb 0.

Then

S T+ (l+3cos 2 21jl)- 1 [{aTa+bTb-(l+cos 2 21jl)trT}I

,......, "" ............. """'"' ,......, ......

+{trT- (2 cosec 2 21jl)aTa- (cosec 2 2<P- 3cot 2 21jl)bTb}a0a

+ { tr T- (2 cosec 2 21jl) bTb- (cosec 2 2<P- 3 cot 2 21jl) aTa}b0b] ,

where cos 21jl a.b. Then f is a function of

J 1 I J 3 I

and

The most general yield function which is quadratic in the stress is (for
the case in which the fibre families are mechanically equivalent)

where c 1 , c 2 , c 3 have the dimensions of stress, are functions of cos 2 21j!

and, in general, depend on the deformation history.
Kinematic Constraints and Constitutive Equations 197

5. PLASTICITY - FLOW RULES

A common procedure in plasticity theory is to assume that the yield

function is a plastic potential for the plastic strain-rate D 1 so that
-p

D~. ~Clf/ClT .. I (4)

:LJ :LJ

where ~ is a scalar multiplier (not a material constant). In a rigid-

plastic theory D is the total strain-rate D; in an elastic-plastic
-P
theory the classical procedure is to decompose D as

D D +D
-p -e

where the elastic strain-rate ~e is given by linear elasticity theoryl

as described in Section 2. For the yield function (2) 1 equation (4) gives

D
-p
= "{1 (1 11
~ --yS + --y---zj (a®aS+Sa0a)
kT - kL kT - -- -- -
} (5)

and for the yield function (3)

(
0 (S-a 0as-sa 0 a) Jl k2
TI J2 < k2LI

D
-p " { V (a0aS+Sa0a)

)JS + (v-0J (a0as+sa0a)

1 Jl < k2T

Jl k2
I J2

J2
k2
LI
k2
'"" "'"'"' ""'"' '""
T I LI

where )l and V are scalar mul tipliers.

The D obtained as above automatically satisfy the constraints of

-p
incompressibility and fibre inextensibility.

6. PLASTICITY - HARDENING RULES

In isotropic plasticity theory 1 the simplest hypothesis to describe

strain-hardening or work-hardening is called isotropic hardening. Since
this term is confusing when applied to anisotropic materials 1 we coin the
term "proportional hardening" for the corresponding theory for anisotropic
plasticity.
198 A.J.M. Spencer

In proportional hardening the yield function retains its form as a

function of the invariants Ja' but the parameters (such as kL and kT in

(2) and (3)) which determine the yield surface depend on the deformation
history. Hence, as the yield surface evolves, it expands (not necessarily
uniformly in all directions) but successive yield surfaces are concentric
(in isotropic hardening of an isotropic material, successive yield
surfaces are similar as well as concentric, their scale being determined
by the change of a single parameter, such as the shear yield stress).

For definiteness, we consider the yield function (2) and its

associated flow rule. Other yield functions and flow rules may be treated
in a similar way. In isotropic theory i t is assumed that the yield stress
is a function of an "equivalent strain" E. By analogy, we postulate that
kT and kL depend on parameters t: 1 , t: 2 , ••• , suchthat s1, t: 2 , ••• are
scalar invariants of D and a, with dimensions (time)- 1• With the
-p
plastic incompressibility and inextensibility conditions

tr D = 0, aD a = 0,
-p -~p~

it follows that there are three such independent parameters, which may be
taken tobe ET, EL and E 3 , where

(6)

We neglect dependence on t: 3 , and assume that

Then

Clk Clk
- -TE• + - -TE• ( 7)
Clt:T T dEL LI

Now, from the flow rule (5), and equation (6),

(8)

Hence, when f 0, from (2) and (8)

Kinematic Constraints and Constitutive Equations 199

Also, during plastic loading, when f 0, (2) gives

and it follows from (7) and (8) that

This determines A in terms of the current stress, the current values of

kT and kL, the hardening parameters akT/asT, akT/asL, akL/asT, akL/asL

and the rate of increase of the stress invariants J 1 and J 2 •

There are several possibilities for simplification. One is that kT

and k depend on the plastic work W . We have
L p

w = S .. D~. 2~ (k- 2 J +k- 2 J ) = 2A.

p lJ lJ T 1 L 2

Therefore, when k
T
and k
L
are functions of wp only,

k. T = 2~ak ;awp ,
T
kL = 2~dk /dW
L p
(10)

and i t follows from (9) and (10) that

Another possibility is to assume that the mechanisms for hardening

in shear in directions parallel and normal to the fibres are independent,
and then

k
L

Kinematic hardening. Kinematic strain-hardening is characterized by a

tensor a termed the "back stress" or "shift tensor". Then, in the yield
200 A.J.M. Spencer

condition and flow rule 1 s is replaced by s-a so that the yield surface
1

is translated. The tensor a must satisfy the same constraints as S 1

namely 1 in the case of a single family of fibresl

tra = o 1 aaa 0.

Evolution equations are required for a. A possible generalization

to transverse isotropy of the hardening rule proposed by Prager [2] for
isotropic materials is

ä = g 1 (E: 1 E: ) D + g2 (E: 1 E: ) (a0aD +D a0a) 1

- T L -P T L - --p -p- -

where a is an objective time-rate.

Elastic-plastic materials. In the small elastic-plastic deformation

range composite materials show intense kinematic hardening. At the
microstructure level this is due to interactions between fibre and matrix 1
leading to residual stress in both. This phenomenon has been termed
"constraint hardening".

For small elastic-plastic deformations 1 it is no langer valid to

assume that the yield function depends only on S1 and it is necessary to
allow dependence on all the stress components. A suitable set of
invariants is

J 3 I tr T- aTa.

The most general quadratic yield function based on these is

Now Y1 and Y2 are yield stresses in tension in the fibre direction 1 and
equibiaxial tension in planes normal to ~~ respectively. In practice
they are much larger than the shear yield stresses k and k . To admit
T L
constraint hardening f is modified to take the form

f 1.
Kinernarie Constraints and Constitutive Equations 201

The parameters a 1 and a 2 represent a back-stress and may be expected to

depend on the history of aE a and tr E . The very high rates of
--p- -p
hardening associated with constraint hardening mean that the fibre strain
and the di1atation are restricted to sma11 magnitudes which may be
neg1ected in the 1arge deformation theory.

Further discussion and app1ications of p1asticity theory for fibre-

reinforced materia1s, with references, is given in [1, 3, 4].

REFERENCES

1. SPENCER, A.J.M., Deformations of Fibre-reinforced Materials, Oxford

University Press, 1972
2. PRAGER, W., A new method of ana1yzing stresses and strains in work-
hardening p1astic solids, J.Appl. Mech. 23 (1956): 493-496
3. SPENCER, A.J.M., The formu1ation of constitutive equations for
anisotropic so1ids, in Mechanical Behaviour of Anisotropie Solids
(ed. J. P. Boeh1er), Editions du CNRS, Paris and M. Nijhoff, The
Hague, 1982: 2-26
4. SPENCER, A.J.M., Yie1d conditions and hardening ru1es for fibre-
reinforced materia1s with p1astic response, in Failure Criteria of
Structured Media (ed. J. P. Boeh1er), A. A. Ba1kema, 1986
 Chapter 11

INVARfANTS OF FOURTH-ORDER TENSORS

J. Betten
Technical University Aachen, F .R. Germany

I . INTRODUCTION

In solid mechanics representing scalar-valued tensor functions or

second-order tensor-valued tensor functions is of major concern. For ln-
stance, the plastic potential is scalar-valued, whereas constitutive
equations are tensor-valued.
Many scientists, for instance, PIPKIN, RIVLIN, SMITH, SPENCER, WI-
NEMAN, to name just a few, have had an ongoing about the response of an-
isotropic materials. This discussion was based upon constitutive ex-
pressions invariant under the group of transformations which defines the
symmetry of material.
In this course, however, anisotropic effects are characterized by
additional material tensors of rank two and four. Therefore, scalar-
valued and tensor-valued functions involving both second-order and
fourth-order argument tensors are represented. Under any orthogonal
transformation these functions have to fulfil the conditions of inva-
riance and form-invariance, respectively.
204 ]. Betten

Many mathematicians have studied the theory of algebraic invariants

~n detail. The results have ben published, for instance, by GRACE and
YOUNG [I], GUREVICH [z], RIVLIN [3], SCHUR [4], WEITZENBÖCK [5], WEYL
[ 6]. Very extensive contributions to algebraic invariant theory with
respect to its application in modern continuum mechanics have been pre-
sented, for example, by SPENCER [7], TRUESDELL and NOLL [8].
Emploing the theory of invariants to the mechanics of isotropic and
anisotropic materials has proven tobe of practical worth [ 9, 10, II]. In
the theocy of algebraic invariants the central problern is: For a given
set of tensors which are not necessarily of the same order, and a given
group of transformations, one must find an integrity basis, whose ele-
ments are algebraic invariants. An integrity basis is a set of polyno-
mials, each invariant under the group of transformations, so that any
polynomial function invariant under the group is expressible as a poly-
nomial in elements of the integrity basis [3,4,6,7, 12].

2. INTEGRITY BASIS FüR A SECOND-ORDER TENSOR

Let F = F(~) be a scalar-valued function of a second-order tensor,

for instance CAUCHY's stress tensor o. This function is said tobe iso-
tropic if the condition of invariance

F(a. a. o ) _ F(o .. ) ( I)
~p Jq pq ~J

is fulfilled under any orthogonal transformation (a.ka.k = 6 .. ), where

l J 1 J
the summation convention is utilized, and ~ represents KRONECKER's ten-
sor.
For all purposes the function c in (I) could be the plastic po-
tential. Then, from the theory of isotropic tensor functions [ 7,9, 13],
it is evident that for an isotropic medium the plastic potential F might
be expressed as a single-valued function of the irreducible basic inva-
riants

sV _ tr o V V I, 2, 3 ( 2)
Invariants of Fourth-Order Tensors 205

or, alternatively, of the irreducible principal invariants

J I - o (3a,b,c)
ll

of the stress tensor 2• that is,

F F[S (o)] or F F[J (o) ]; V I ,2 ,3 , (4a, b)

V - V -

respectively. In (3), i.e., (-1) 3-v Jv = oi 1[i 1] ... oiv[iv]' the operation

of alternation is used. This process is indicated by placing square

brackets araund those indices to which it applies, that is, the v
bracketed indices i ... k are permutated 1n all possible ways, whereas ln-
dices which are excluded from the alternation are not bracketed. They
keep their position. Thus, we obtain v! terms. The terms corresponding
to even permutations are given a plus sign, and those which correspond
to odd permutations a minus sign, and then they are added and divided by
v!. Comparing (2) and (3), we find the following relations:

Jl SI ' (Sa,b,c)

The irreducible (basic or principal) invariants in (2) or, alternatively,

in (3) are the elements of the integrity basis for the stress tensor o
under the proper orthogonal group, i.e. Ia .. I = + I, and this integrity
lJ
basisalso forms a functional basis. One can say, that a set of in-
variants constitutes a functional basis, if any invariant can be expres-
sedas a function of them [7]. It has been shown by PIPKIN and RIVLIN
[ 14] and PIPKIN and WINEMANN [ 15] that an integrity basis will also form
a functional basis which may not necessarily be irreducible; then all
invariants can be expressed as functions of the invariants of an inte-
grity basis.
It is evident, that the principal values o 1 , o 11 , o 111 of the
stress tensor o are independent of the choice of the coordinate system,
i.e., the principal values can be regarded as invariants of the tensor.
For isotropic behaviour the plastic potential F must be a symmetric
206 J. Betten

scalar-valued function of the principal values, and an irreducible re-

presentation is given in terms of the elementary symmetric functions

J I' (6a,b)

(6c)

which are identical to the principal invariants (3a,b,c). Thus, the re-
presentations (4a,b) imply isotropy.
The elementary symmetric functions (6a,b,c) are the coefficients of
the characteristic polynomial P3 (A) of the indeterminate A

(7)

whose roots are the principal values o 1 , o 11 , o 111 . All symmetric

functions can be expressed in terms of the elementary symmetric
functions (6a,b,c). In other words: the elementary symmetric functions
form a functional basis for the symmetric functions. For instance, the
. . 4 4 4 . f
symmetr1c funct1on s4 = o 1 + o 11 + o 111 can be expressed 1n the ol-
lowing way:

( 8)

that is, in terms of the elementary symmetric functions [ 16].

As we have already seen, the representations (4a,b) imply isotropy.
In the anisotropic case the restriction (I) on the plastic potential F
is less severe. Then, the function F is invariant under the group of
transformations (s.ks'k = 8 . . ) associated with the symmetry properties
l J l J
of the material [ 13], where ~ is a subgroup of the orthogonal group ~·
In other words, the symmetry properties of the material impose re-
strictions upon the manner in which the function F depends on the stress
components [ 17].
For a particular crystal class [ 18] the plastic potential F in (I)
may be represented as a polynomial which is invariant under the sub-
Invariants of Fourth-Order Tensors 207

group ~ of transformations associated with the symmetry properties of

the crystal class considered. The function F is then expressible as a
polynomial in these invariants, which form a functional basis.

3. SIMPLIFIED CHARACTERISTIC POLYNOMIAL

Instead of the plastic potential being represented by an integrity

basis under a subgroup it can be expressed in the following way:

F ( 9)

where Aijkl are the components of a fourth-order constitutive tensor

characterizing the anisotropic properties of the material. Then, by
analogy to (I), we have the invariance condition

F ( a. a . o ; a . a . ak a 1 A ) =: F ( o .. ; A1. J. k 1 ) · ( I 0)
1p Jq pq 1p Jq r s pqrs lJ

In this case the central problern is: to construct an irreducible

integrity basis for the tensors o and A. Tagether with the invariants of
the single argument tensors, like (2) or (3), we have to consider the
system of simultaneaus or joint invariants.
Let us first construct a set of irreducible principal invariants of
the fourth-order tensor Aijkl' which may be a linear operator, i.e.,

y .. or y ( !Ia, b)
lJ a

where i,j,k,l = 1,2,3 or a,ß = 1,2, ... ,9, respectively. In(!!) two pos-
sible indicial notations are used. The transformation laws of the Car-
tesian tensor components can be written as

Y~'. a. a. Y or Y* (12a,b)
lJ lp Jq pq a 3 aßyß

and
208 ]. Betten

a. a. ak a 1 A or: A* a a A , (13a,b)
1p Jq r s pqrs aß a~ ßn ~Tl

with the orthonormal conditions:

aitajt 6 .. or: aac;;aßc;; 6aß ( 14a,b)

lJ

where:

aac;; a. a.
lp Jq
( 15)
A A

} =' aac;;aßc;; = a.lp a.Jq ak pa 1q 6 ik 6 jl 6aß

aßc;; akpalq

In ( 12b) the operator A defines a linear transformation on a 9-di-

mensional vector space v9 . This transformation is a correspondence which
assigns to every vector ~ in v9 a vector AX in v9 , in such a way that

identically in the vectors and the scalars a 1, a 2 .

~I' ~ 2

Let Xa; a = 1,2, ... ,9; be the components of an arbitrary vector of

unit magnitude which we shall call a direction vector or simply, a di-
rection. We then ask: For what direction X does the linear transfor-
mation A yield a vector Y according to ( II) which is in the same di-
rection as X? That is,

0 .. or 0 ( 16a,b)
lJ a

A(O) (0)
where A. is a real scalar to be determined, and Aijkl - 6 ik6 j 1 or
aß -
- 6 Ao , whereas 0 .. or 0 are
aß are the components of the unit tensor lJ a
~

the components of the zero tensor 0.

For a nontrivial solution of ( 16a,b) we must have

0 or 0 (17a,b)
Invariants of Fourth-Order Tensors 209

~n order to determine the principal or proper values of the linear

transformation A.
In order to construct a set of principal invariants of a fourth-or-
der tensor f::, we shall note that the determinant ( 17) is an invariant,
and we therefore consider the characteristic polynomial

n
(0)
p (;\) - det(Aijkl - AAijkl)
n I J (A)A n-v
- ( 18)
v=O \)

~n which, as we can see from ( 17a,b), the first indexpair (ij) =a cha-
racterizes the rows and the second one (kl) =ß the columns of a n x n
matrix ~' generally n = 9.
Analogaus to (3), the principal invariants J ~n ( 18) can be deter-
v
mined by performing the operation of alternation [ 19]:

where (-l)n J 0 =I. The right hand side in (19) is equal to the sum of

a ll ( n}
v = v.'( nn_! V )'• · · 1 m~nors
pr~nc~pa · o f order \! ~ n, wh ere \! = I and
v = n lead to trA and det~, respectively.
Assuming we have the usual symmetry conditions

(20a)

or alternatively written as

a,ß I ,2, ... ,6 (20b)

then we can express the zero power tensor of fourth-order in ( 18) in the
following way:

A(-I) A ( 2 I)
ijpq pqkl

as we can see from (I I) for Y ~X.

210 ]. Betten

I f we expand ( 19) with n = 6, then we can find a set of irreducible

principal invariants of a fourth-order tensor:

( s 2 -52)/2!, (53 - 35152 + 253)/3! ,

Jl - SI, J2 -
I J3 - I

2
J4 - (54 + 85153
I 65251 + 35 22 - 654)/4!

J5 - <s5 - 305154 + 15 sls22 - 20 s s - 10 s2s 3I + 20 s 3 s~ + 25 s 5 )/5!,

I 2 3

(22)

where

SV - trA V - ( 23)

are irreducible basic invariants found by forming irreducible traces of

tensor powers. We see, that the principal invariants (22) of a fourth-
order tensor can be determined uniquely by polynomial relations from the
irreducible basic invariants (23).
The set of the six quantities (22) or, alternatively, the six quan-
tities (23) are irreducible invariants of the fourth-order symmetric
tensor (20) under the orthogonal group. However, the systems (22) and
.... ,
(23) are not complete, because some irreducible invariants like AllJ J
A.. A .. , A... A. A etc. cannot be expressed through (23). Thus,
11pq pqJJ lJlp pJqr rsqs
the characteristic polynomial ( 18) must be generalized. Proposals are
made in sections 7 and 8.

4. THE HAMILTON-CAYLEY THEOREM

In the theory of invariants the HAMILTON-CAYLEY theorem plays an

important role [ 7, 12,20]. It states that
lnvariants of Fourth-Order Tensors 211

..
J2A lj J3 ..
.. _ olJ
olJ (24)

where (A .. ) is any 3 x 3 matrix. This theorem can be applied to the

lJ
fourth-order tensor (20):

I
v=O
J
V
~ n-v)
b)'
:: 0
ßy
ß,y 1, 2, ... ,n ( 25)

as has been pointed out in detail in [ 21]. Thus, a second-order tensor

Aij or a fourth-order tensor Aijkl satisfies its own characteristic
equation (7) or ( 18), respectively. From (25) we find, that a tensor of
power n and all higher powers can be expressed in terms of powers lower
than n:

n
pQ A( n-V)
I n-v ßy n '
( 26)
V= I

where n = 6, if Ais a symmetric fourth-order tensor (20), or where

n = 3, if A is a symmetric second-order tensor. The coefficients Pq
n-v
in (26) are scalar polynomials of degree p - n + v in the principal In-
variants J 1, ... ,JV. For the symmetric fourth-order tensor (20) with
n = 6 we have listed the polynomials Pq 6 in table I.
-v

Table I Scalar Polynomials Pq in ( 26) for a symmetric fourth-order

tensor (n = 6) n-v

r>z 6 7

2
8

1 - JI - J2 JI - J3 - J3
+ + 21 11 2 I
2 - J2 - J3 + J 11 2 - J4 + J 11 3 + J2(J2 - J2)
I
3 - J3 - J4 + J 11 3 - J5 + J 11 4 + J/J2 - J2)
I
4 - J
4 - J5 + J 11 4 - J6 + JIJ5 + J4(J2 - J2)
1
5 - J
5
- J
6 + J1J5 JIJ6 + Js(Jz - J2)
I
6 - J6 1 11 6 J6(J2 - J2)
I
212 ]. Betten

From (22), (26) and table I, we see, that all invariants formed by
traces of tensor powers can be expressed by scalar polynomials in the
irreducible basic invariants s 1, ... ,s 6 defined in (23).
The scalar coefficients in (26) can be determined by using the re-
cursion relation [22,23]

p-n
(-l)n-I(J + I p-)lQ J) ( 27)
p-n+V ].1= 1 n-v ].1

which is valid for the arbitrary dimension n. For the symmetric tensor
(20) of rank four (n = 6) we can use (27) to find the scalar polynomials
listed in table I, and for a symmetric second-order tensor (n = 3) we
can calculate polynomials listed in [ 16].

5. CONSTRUCTION OF SIMULTANEOUS INVARIANTS

In the cases where we have several differenttensorvariables (9),

besides the irreducible invariants (3), ( 19), (22), we must take simul-
taneaus or joint invariants into consideration. In order to construct a
set of simultaneaus invariants of the stress tensor o and the fourth-or-
der constitutive tensor A we could begin by applying the following theo-
rem:
A scalar-valued function f(v,T) of one n-dimensional vector v and one
symmetric second-order tensor T is an orthogonal invariant, i.e. inva-
riant under the orthogonal group, if and only if it can be expressed as
a function of the 2n special invariants

2 n-1
v , "'v•Tv,
,.......,,......,
... ,v•T
,....., ,.....,
v (28)

This theorem is valid for the arbitrary dimension n [ 8] . By using the

HAMILTON-CAYLEY theorem (25), which states that Tn and all higher powers
1' n+h can be expres sed in terms o f ;::.s ,_! ,_! 2 , ... ,_! n-l accord ing to ( 26) , and
by assuming the symmetry conditions (20a,b), we can find the set of 15
simultaneaus invariants [ 19,24] from (28):
Invariants of Fourth-Order Tensors 213

11~ v] -
(29)
V "' I ,2, ... ,5,

where V in a squarebracket denotes v's several invariants, whereas v in

parenthesis is an exponent.
Some of these simultaneaus invariants can be constructed in the
following way. If we take the second-order tensor B _ o .. A •. into
pq lJ lJpq
account, then, for instance, the basic invariant B B is identical to
[2] pq qp
the simultaneaus invariant 11 2 In a different example we can consider

t h e secon d -order tensor C - o.( 2. )A . . and f.1nd, f or 1nstance,

. .
the ln-
pq lJ lJpq
variant c c = 11[ 2 J. Further examples could be:
pq qp 4

B 0 11[ I J
pq qp - 2
B 0
(2)
pq qp
cpq 0 qp = 11[ I J
3
cpq 0(2)
qp - rri I J.
The isotropic special case can be expressed by the isotropic con-
stitutive tensor

( v)
a V 6 lJ
.. okl + bV (o.ko"l (30)
Aijkl 1 J + o.lo.k)
1 J

of power V"' 1,2, . . . . Th~n the simultaneaus invariants IT~v] in (29) are
equal to the principal invariant J 2 (~) for av "' - 1/2, bv "' 1/4 and
equal to the basic invariant s 2 (o) for a "' 0, b "' 1/2. Similarly, the
[v] - V 3 V
invariants 11 3 in (29) are equal to J 3 (~) - J 1 (~)/6 for av "' - 1/6,
bv "' 1/6 and equal to the cubic basic invariant s3 (o) for a "' 0, b
- V V
. . [v] _ (v) [v] _
1/2. Furthermore, the 1nvar1ants 11 1 = oijAijklokl or ITI* =
(v) (2)
- oijAijklokl are equal to s 1 (~) or s 2 (~), respectively, if 3av + 2bv
I.
It must be mentioned, that the irreducible invariants 11[lv] and 11[v]
I*
arenot contained in (29). Furthermore, a vector v of arbitrary dimen-
sion n has only one invariant, namely v 2 in (28).-However, in cantrast
to ~· a second-order tensor in three dimension has three invariants (3a,
214 J. Betten

b,c). Thus, the system (22) of irreducible principal invariants or the

irreducible set (29) of simultaneaus invariants cannot be complete. This
fact must be emphazised and to taken into account, if we use the no-
tation (20b) for the fourth-order tensor (20a) or the notation oa, a =
= I,2, ... ,6, for the symmetric second-order tensor o .. = o .. with i,j =
1 J J1
I, 2, 3.

6. CONSTRUCTION OF INVARIANTS BY THE POLARIZATION PROCESS

In the case of further tensor variables in the scarlar function

(9), for instance F = F(o .. , B.. , A. 'kl)' we can find new invariants
1J 1J 1J
from the systems (2), (23) or (29) by using the polarization process.
Some examples are listed below:

[v] B .. A~~) o( 2 ) +
( v)
B .. arr 3 /Clo .. 2B .. o .kAk. o - tr BC + 2tr BOC
1J 1J 1J 1Jpq pq 1J J 1pq pq

2A .... - 2J I(~)
1J1J

2ll[vJ.
2 '

Clll[v]
3
B .. - - - = B.. A<.v.) o(2) +
2B A(v)
.. o J'k k.1pq 0 pq _ tr BC + 2tr BoD
1 J Clo .. 1J 1Jpq pq 1J
1J

where C.. - A~~) o( 2 ) and D.. -

1J 1Jpq pq 1J

where 1 .. _
1J
lnvariants of Fourth-Order Tensors 215

(3)
B .. o .. okk
l J J l

Some applications of the theory of invariants to the formulation of

yield criteria and failure criteria are discussed in [25,26], where
theoretical results have been compared with experimental data for sin-
tered powder materials and several polymers.

7. EXTENDED CHARACTERISTIC POLYNOMIAL

The characteristic polynomial ( 18) in section 3 should be admitted

as a simplified one. An extended form of the characteristic polynomial
( 18) can be achieved by using the isotropic fourth-order tensor

(31)

(0) (0)
instead of the spherical tensor AAijkl' where Aijkl 1s the zeropower
tensor defined in (21). Thus we take into consideration the characte-
ristic polynomial

P(A,W) = det(Aijkl- Iijkl) =0 , (32)

i.e., we formulate the eigenvalue problern

(33)

instead of ( 16a,b). The isotropic tensor (31) yield an image dyad

y .. AX c5 .• + 2WX .. (34)
lJ rr lJ lJ

which 1s coaxial with the dyad X.

In continuum mechanics equation (34) is known as the constitutive
equation for an isotropic linear elastic solid. Such a material is cha-
216 ]. Betten

racterized by the two elastic constants A. and lJ, which are called the
LAME constants. Using the notation Ya = IaßXß, a,ß = 1,2, ... ,6, we see
that the isotropic tensor I in (34) can be represented in the matrix
form

2).J + A. A. A. 0 0 0
A. 2).J + A. A. 0 0 0
A. A. 2).J + A. 0 0 0
(35)
1 ijkl -
0 0 0 2).J 0 0
0 0 0 0 2).J 0
0 0 0 0 0 2lJ

which is quasi-diagonal of the struct~re {3, I, I, 1}. The inverse tensor

(-I)
of (31) can be found fromX .. Iijklykl in connection with (34):
lJ
(-I)
1 ijkl (36)

In order to control this result we prove:

I(-I)I I I(-l) (0)

( 37)
ijpq pqkl ijpq pqkl - Aijkl

where the zero power tensor has the diagonal form

(0) (38)
{ I, I, I, I, I, I} .
Aijkl

We see, in the special case (,\ = 0, ).J = 1/2) the isotropic tensor (35)
tends to the unit tensor (38). Then we find from (32) the simplified
characteristic polynomial ( 18). The determinant of (35) is

5
det(Iijkl) = 32(2).J + 3A)lJ ,

(0)
which tends to det(Aijkl) in the special case (,\ 0, ).J I I 2).

In the following we should calculate irreducible invariants as

lnvariants of Fourth-Order Tensors 217

coefficients in the characteristic polynomial (32). Because of (lla,b),

(20a,b) and (35) the characteristic polynomial (32) can be written as:

All li-~ AII22-A AII33-A 2A1112 2AII23 2A1131

A2211-A A2222-~ A2233- A 2A2212 2A2223 2A2231
A33 II- A A3322-A A3333-~ 2A3312 2A3323 2A3331
0
A1211 A1222 Al233 2A12 12- 2 lJ 2AI223 2AI231
A2311 A2322 A2333 2A2312 2A2323- 2 1J 2A2331
A3111 A3122 A3133 2A3112 2A3123 2A3131- 2 1J

(39J

where the abbreviation ~ :: 2]J + A is used. The non-symmetric determinant

in (39) is a polynomial of degree 3 in A and degree 6 in ]J:

,p q
P( A, JJ) c( p,q )"JJ, p 0,1,2,3; q 0, I, ... ,6, ( 40)
p,q p + q ~ 6

with 22 coefficients C(p,q)' However, the characteristic equation

P(A, JJ) = 0 can we devided by the coefficient C(0, 6 )' so that we could
find no more than 21 invariants. Now, the main problern is: to expand the
determinant (39) and to find out if all 21 coefficients are irreducible
invariants [27].
In order to avoid the lengthy expansion of the determinant (39) we
propose the following way. If we consider a second-order tensor we can
start from the diagonal form

( 4 I)

Then, the determinant det (o .. - Ao .. ) is identical to the characte-

lJ lJ
ristic polynomial (7). Similarly, we could consider the orthopropic case
with

0.
218 ]. Betten

Then, the dyads ~ and Xin (I Ia) are coaxial and, instead of (39), the
following determinant is to expand:

AI-( 2Jl+A) B -A BII-A 0 0 0

I
B -A All- ( 2\..!+A ) BIII-A 0 0 0
I
BII-A BIII-A AIII-( 2J.1+A) 0 0 0
0 0 0
0 AIV-2].1 0 0
0 0 0 0 A -2\..1 0
V
0 0 0 0 0 AVI-2\..1
(42)

From (42) we find the characteristic polynomial:

(43)

where the 15 invariants I 1, ... ,I 15 are listed below:

I = KM
6 - I 3 '
Invariants of Fourth-Order Tensors 219

I I I - - I 6 [ 2 ( K I - L I ) + 3M I ] ,

(44)

In the system (44) the following abbreviations are used:

(45)

We see, in the system (44) the invariants r 1 and r 15 are the trace and
determinant of the fourth-order tensor, respectively.
In the special case (A = 0) only the six invariants r 1, ... ,1 5 , r 15
are relevant, and they are identical to the six invariants in (22), if
furthermore B1 =
B11 = B111 = 0.
In the two dimensional case (i,j,k,l = 1,2) the linear transfor-
mation (I Ia) can be represented in the following matrix notation
220 J. Betten

ZAIII2)
C") (All II
C)
A1122
y22 A2211 A2222 2A2212 x22 (46a)
yl2 A1211 A1222 2A1212 x12

y ~I X (46b)

where the matrix 21 is non-symmetric.

The isotropic case (34) can now be written as

C)- C°
y22
yl2
+;>., ,, ;>., +

0
;>.,

2]J o0) Cl)

21J
x22
xl2 .
(4 7)

Considering (46a,b) and (47), the eigenvalue problern (33) leads to the
characteristic polynomial

AI I I I - O + 2 IJ) A I 122 - A 2A1112

P( il, IJ) A2211- A A2222- O + 2 1J) 2A2212 (48)
Al211 A1222 2 A1212- 2 ~

from which we find the characteristic equation P(A,IJ) 0:

with the following 2 irreducible invariants:

lnvariants of Fourth-Order Tensors 221

(SO)

The orthotropi~ special case is characterized by A 1112 = A1222 = 0.

Then, the matrix '~( in (46a,b) is symmetric, and the invariants in (SO)
are reduced to simpler forms. However, all 5 invariants arerelevant 1n
this special case, too. But the invariants K 1 and K 2 in (49) are not re-
levant, if we take a simplified characteristic equation with A = 0 into
consideration. Then, the number of 5 invariants in (50) is reduced to 3.

8. THE LAGRANGE MULTIPLIER METHOD

In order to find the characteristic eqpation of a second-order ten-

sor one can utilize the LAGRANGEmultiplier method, i.e., we vary the
orthogonal transformation ~ in the tensor law

A~. a. a. A ( 5 I)
lJ lp Jq pq

1n such a way that the tensor components Aj 1, A2 2 , A~ 3 will be extremal,

where the "auxiliary" or "subsidiary" condition

M .. - a. a. 6 0 .. (52)
lJ 1p JP ij lJ

1s to fulfil. This is the condition, which follows from the property

(-1)
aiJ. = a .. of an orthogonal matrix. From (SI) and (52) we form the ten-
Jl
sor
222 ]. Betten

<P .. a.a.A -1-(a.a. -6 .. ) (53)

1J l p ]q pq 1p J p 1J

where the undetermined factor A 1s called a LAGRANGE multiplier. Now,

the tensor components Aj 1, A2 2 , A) 3 are extremal, if we require

(54)

From (54) we find

( A. . - ,\ ( ) 6 .. ) a( ) . 0 . (55)
1J r lJ r J r1

( I)
with 3 eigenvalues \I):::: >- 1 , ... "\( 3 ) - >- 111 and 3 eigenvectors ni
I (3) III
- ni'' .. ,ni - ni where

(r)
n. (56)
1

1s the r-th row of the orthogonal matrix ~· Note, that there 1s no sum
on the bracketed index r in (54) and (55). The condition for (55) to
have non-trivial solutions (56) yields the characteristic equation
de t( A. . - ,\ 6 .. ) = 0.
1J 1J
We can also start from the quadratic form

F A .. x.x. (57)
1J 1 J

with the "auxiliary" or "secondary" condition

2
M 6 .. x.x. - X (58)
1J l J

where the invariant x - ~ 1s the magnitude of the vector x. Now,

1 1
the LAGRANGE multiplier method requires

3<P/3x. - d(F - AM)/dx. 0. (59)

1 1 l
Invariants of Fourth-Order Tensors 223

and yields

(A .. - \6 .. h. 0. (60)
l J l J J l

From (60) we again read the characteristic equation.

In order to find the characteristic equation of a fourth-order ten-
sor we can also apply the LAGRANGE multiplier method, and, in extending
(51), (52), (53), we have the transformation law

A~ 'kl = a.1p a.Jq a kr a 1 s Apqrs ( 6 I)

1J

the 2 "auxiliary" conditions

( 62a)

(62b)

and the fourth-order "modified" tensor

( 63)
<P ijkl

respectively. Analogously with (54), we require

! !
Cl<P( 1111
.... )/Cla(.) 0. or: Cl<P('''')/Cla(') 0 .. (64)
1 u lU ljlj 1 u lJU

Cno sum on the bracketed indices i or j). Going on 1n this way, we find
more details in [27].
The other possibility to apply the LAGRANGE multiplier method is
illustrated in the following. Similarly to (57), we start from the sca-
lar-valued function

F (65)

which may be an analogaus function with the elastic potential (strain-

224 ]. Betten

energy function). Because the second-order tensor X in (65) has 3 irre-

ducible invariants s 1 = tr ~·· .. ,s 3 = tr ~ 3 , being-the elements of the
integrity basis, in cantrast to the vector x in (57) having only one in-
variant, namely its magnitude x in (58), we take into consideration the
following l "auxiliary" conditions:

L 0 ' (66a)

M (66b)

N = ( 6 J. k6 p l + 6 . 1 6 k) X. X.. Xk l - 2S3
J p lp lJ
(66c)

Thus, the quadratic form (65) of the dyad X 1s modified to the form

F - AL - wM - vN ( 6 7)

which 1s tobe made stationary, i.e., similarly to (59), we require

CltP/dX ! 0 (68)
rs rs

From (68) we find the system of non-linear equations 1n the dyad ~.

where Iijkl are the components of the fourth-order isotropic tensor de-
fined in (31). In the special case when V= 0 we find from (69) the el-
genvalue problern (33) which yields the characteristic equation (32).
Some more details concerning the system (69) can be found in [27]. Fur-
thermore, applications of the LAGRANGE multiplier method to tensors of
order six are discussed in [27], too.
lnvariants of Fourth-Order Tensors 225

REFERENCES

I. GRACE, J.H. and YOUNG, A., The Algebra of Invariants, Carnbridge

Univ. Press. London and New York 1903.
2. GUREVICH, G.B., Foundations of the Theory of Algebraic Invariants,
P. Noordhoff, Groningen 1964.
3. RIVLIN, R.S., An Introduction to Non-linear Continuum Mechanics, in:
Non-linear Continuurn Theories in Mechanics and Physics and their ap-
plications (Ed. R.S. Rivlin), Edizioni Crernonese, Rorne 1970.
4. SCHUR, I., Vorlesungen über Invariantentheorie, bearbeitet und
herausgegeben von H. Grunsky, Die Grundlehren der mathematischen
Wissenschaften in Einzeldarstellungen, Bd. 143, Springer-Verlag,
Berlin/Heidelberg/New York 1968.
5. WEITZENBÖCK, R., Invariantentheorie, P. Noordhoff, Groningen 1923.
6. WEYL, H., The Classical Groups, Their Invariants and Representation,
Princeton Univ. Press, Princeton and New Jersey 1946.
7. SPENCER, A.J.M., Theory of Invariants, in: Continuum Physics (ed.
A.C. Eringen), Academic Press, New York 1971, 239-353.
8. TRUESDELL, C. and NOLL, W., The Non-Linear Field Theories of Mecha-
nics, in: Handbuch der Physik (Ed. S. Flügge), Vol. III/3, Springer-
Verlag, Berlin/Heidelberg/New York 1965.
9. BETTEN; J., Ein Beitrag zur Invariantentheorie in der Plastemechanik
anisotroper Stoffe, Z. Angew. Math. Mech. (ZAMM), 56 ( 1976), 557-
559.
10. BETTEN, J., Theory of Invariants in Creep Mechanics of Anisotropie
Solids (Ed. J.P. Boehler), Martinus Nijhoff Publishers, The Hague/
Boston/London 1982, 65-80.
I I. BETTEN, J., Creep Theory of Anisotropie Solids, Journal of Rheology,
25 (1981), 565-581.
12. SPENCER, A.J.M. and RIVLIN, R.S., The Theory of Matrix Polynomials
and its Application to the Mechanics of Isotropie Continua, Arch.
Rational Mech. Anal. 2 ( 1958/59), 309-336.
13. SMITH, G.F., On the Yield Condition for Anisotropie Materials,
Quart. Appl. Math. 20 ( 1962), 241-247.
14. PIPKIN, A.C. and RIVLIN, R.S., The Forrnulation of Constitutive
Equations in Continuum Physics, Arch. Rational Mech. Anal. 4 ( 1959),
129-144.
15. PIPKIN, A.C. and WINEMAN, A.S., Material Symmetry Restrietions on
Non-Polynomial Constitutive Equations, Arch. Rational Mech. Anal. 12
( 1963)' 420-426.
16. BETTEN, J., Elementare Tensorrechnung für Ingenieure, Vieweg-Verlag,
Braunschweig 1977.
226 ]. Betten

17. WINEMAN, A.S. and PIPKIN, A.C., Material Symmetry Restrietions on

Constitutive Equations, Arch. Rational Mech. Anal. 17 ( 1964), 184-
214.
18. SMITH, G.F., SMITH, M.M. and RIVLIN, R.S., Integrity Bases for a
Symmetrie Tensor and a Vector - The Crystal Classes, Arch. Rational
Mech. Anal. 12 (1963), 93-133.
19. BETTEN, J., Integrity Basis for a Second-Order and a Fourth-Order
Tensor, Internat. J. Math. & Math. Sei. 5 (1982), 87-96.
20. RIVLIN, R.S. and SMITH, G.F., Orthogonal Integrity Basis for N Sym-
metrie Matrices, in: Contributions to Mechanics (Ed. D. Abir), Per-
gamon Press, Oxford/ ... /Braunschweig 1969, 121-141.
21. BETTEN, J., Zur Anstellung von Stoffgleichungen in der Kriechmecha-
nik anisotroper Körper, Rheol. Acta 20 ( 1981), 527-535.
22. BETTEN, J., Elastizitäts- und Plastizitätslehre, Vieweg-Verlag,
Braunschweig/Wiesbaden 1985.
23. BETTEN, J., Irreducihle Invariants of Fourth-Order Tensors, Fifth
International Conference on Mathematical Modelling, University of
California, Berkeley 1985, to be published in the proceedings: Ma-
thematical Modelling in Science and Technology (eds. X.J.R. Avula,
G. Leitmann, C.D. Mote, Jr. and E.Y. Rodin), Pergarnon Press, New
York 1986.
24. BETTEN, J., Representation of Constitutive Equations in Creep Mecha-
nics of Isotropie and Anisotropie Materials, in: Creep in Structures
(Ed. A.R.S. Ponter and D.R. Hayhurst), Springer-Ve'rlag, Berlin/Hei-
delberg/New York 1981, 179-201.
25. BETTEN, J., Pressure-dependent Yield Behaviour of Isotropie and An-
isotropie Materials, IUTAM-Symposium on Deformation and Failure of
Granular Materials, Delft 1982, published in the proceedings (eds.
P.A. Vermeer and H.J. Luger), A.A. Balkema, Rotterdam 1982, 81-89.
26. BETTEN, J., Formulation of Failure Criteria for Anisotropie Mate-
rials under Multi-Axial States of Stress, Colloque International du
CNRS n° 351 on Failure Criteria of Structured Media, Grenoble 1983,
tobe published in the proceedings (ed. J.P. Boehler).
27. BETTEN, J., The Eigenvalue Problem of a Fourth-Order Tensor, 1n pre-
paration.
 Chapter 12

FORMULATION OF ANISOTROPie
CONSTITUTIVE EQUATIONS

J. Betten
Technical University Aachen, F.R. Germany

I . INTRODUCTION

This chapter is concerned with the formulation of constitutive ex-

pressions of the form

d .. f. .(o ,w ,A ) ( I)
lJ lJ pq pq pqrs

where d .. = (v . . + v . . )/2 are the cartesian components of the so

lJ l,J J,l
called "rate-of-deformation tensor" d. Other common names are the "rate-=-
of-strain" or "strain-rate-tensor". Note that d .. is linear in the velo-
lJ
city components vi' and that this linearity is exact and no approxi-
mation has been made in deriving it. Furthermore, the tensor ~ is not to
confuse with the "material time derivative" ~ of the infinitesimal
strain tensor c. (u . . + u . . )/2, because we have [I]:
lJ l 'J J 'l

E:.. d .. - (u. V . + U. V .)/2

lJ lJ l,p p,J J,p p,l
228 ]. Betten

Only in the case of small displacement gradients ( u. . _ au. lax.) and

l,J ,l J
small Velocitygradient tensors (v . . :: dV./dx.) We have: E •• "d ...
l,J l J lJ lJ
The symmetric strain-rate tensor (d .. = d .. ) is represented as a
lJ Jl
second-order tensor-valued function (I) of two symmetric argument ten-
sors of rank two and one symmetric argument tensor of rank four, being
the CAUCHY stress tensor ~· the darnage tensor ~· and a fourth-order con-
stitutive tensor ~ characterizing the anisotropy from, perhaps rolling,
i.e. the anisotropy of the material in its undamaged state. Constitutive
equations of the form (I) are applied in creep mechanics [ 1,2,3,4,5],
because the creep behaviour of isotropic and anisotropic materials in
the tertiary stage is characterized by a darnage tensor.
The general representation of (I) is given through a linear combi-
nation

d L: cp G (2)
0. ~o.
0.

where the G's are symmetric tensor generators of rank two involving the
argument tensors o, w, A. The coefficients cp 0. in (2) are scalar-valued
~ ~ ~

functions of the integrity basis associated with the representation (2).

They must also contain experimental data measured 1n uni-axial creep
tests [5]. The main problems are: to construct an irreducible set of
tensor generators and to determine the scalar coefficients involving ex-
perimental data.
A further a1m is to represent the constitutive equations (2) 1n ca-
nonical form

d .. ( 3)
lJ

0 I 2
where Hijkl' Hijkl' Hijkl are the components of fourth-order tensor-

valued functions 0 ~. 1 ~. 2 ~. depending on the darnage tensor ~ and the

anisotropy tensor ~· The canonical form (3) is a representation in three
terms, which are the contributions of zero, first, and second orders in
the stress tensor ~· influenced by the functions 0 ~, 1 ~, and 2 ~, respec-
tively.
Formularion of Anisotropie Constitutive Equations 229

2. THE DAMAGE STATE IN A CONTINUUM

Before formulating constitutive equations of the form ( 1), we

should describe the darnage state in a continuum.
In a uni-axial tension specimen material deterioration can be des-
cribed by introducing an additional variable w or, alternatively,
W= I - w into constitutive equations, i.e. the strain rate d can be ex-
pressed in the form d = f(O,w), or d = f(O, W), where 0 is the uni-axial
stress [6,7]. The materialparameterswand Wdescribe the current da-
mage state and the "continuity" of the material, respectively. The para-
meter of "continuity", w. represents that fraction of the cross-sectio-
nal area which is not occupied by either voids or internal fissures. The
net stress acting over the cross-section of the uni-axial tension speci-
men is then o = o/w. When W = I the material is in its virgin undamaged
state, and when w= 0 the material can no longer sustain any load. In
the latter case the constitutive equation would be required in order to
approach an infinite strain rate. Furthermore, it is assumed that the
darnage rate w or, alternatively, the rate of continuity change
.
Wis also
governed by the uni-axial stress and by the current state of continuity,
i.e. w = g(o,w), or ~ g(o,w).
For example, the functions f and g may have the forms

(4a,b)

where n, m, , d 0 , w0 , and o 0 are constants. Then, the undamaged case

v,~

(w = 0) leads to NORTON's power law, which is assumed tobe valid for

the secondary creep stage, while the strain rate d approaches infinity
as w approaches I.
Because of its microscopic nature, darnage generally has an aniso-
tropic character even if the material was originally isotropic. The fis-
sure orientation and length cause anisotropic macroscopic behaviour.
Therefore, darnage in an isotropic or anisotropic material, which is in a
state of multi-axial stress, can only be described in tensorial form.
When generalizing the above uni-axial concept, constitutive equations
230 J. Betten

and anisotropic growth equations are expressed as tensor-valued

functions

d .. = f. . ( o ,w) w..
0
g .. ( o, w) , (Sa,b)
1J 1J - - 1J 1J - -

respectively, where ( 0 ) denotes the JAUMANN derivative and w is an ap-

propiately defined darnage tensor.
Darnage tensors are constructed, for instance, in [8]. Furthermore,
we refer to the work of MURAKAMI and OHNO [9]. They assumed that darnage
accumulating in the process of creep can be expressed through a symme-
tric tensor of rank two.
RABOTNOV [ 10] also introduced a symmetric second-order tensor of
darnage and defined a symmetric net stress tensor o by way of a linear
transformation

0 .. (6)
1J

where the fourth-order tensor ~ is assumed tobe symmetric.

However, in [I I] is has been pointed out that the fourth-order ten-
sor in (6) is only symmetric when corresponding to the first index pair
ij, but not to the second one kl. Thus, the net stress tensor o is not
symmetric in the anisotropic darnage case. The net stress tensor can be
decomposed into a symmetric part and into an antisymmetric one, where
only the symmetric part is equal to the net stress tensor introduced by
RABOTNOV [10], as shown 1n [II].
A tensor of continuity and an anisotropic darnage tensor can be con-
structed in the following way.
In three-dimensional space a parallelogram formed by the vectors v
and w can be represented by

s.1 _(vxw). (7a)

- 1

or 1n the dual form

Formularion of Anisotropie Constitutive Equations 231

s .. Ob)
1J

where E. 'k is the third-order permutation tensor. From the relations

1J
(7a,b) we immediately find the decomposition:

V. V.
s .. 1 J ( 8)
1J w. w.
1 J

Such an alternating product of two vectors ~ and ~ is called a simple

bi-vector ~, which has the three nonvanishing essential components s 12 ,
s 23 , s 31 . In rectilinear components in three-dimensional space, we see
that the absolute values of the components s 12 , s 23 , s31 are projections
of the area of the parallelegram (7a) on the coordinate planes. Thus
S .. , according to (8), represents an area vector in three-dimensional
1J
space and has an orientation fixed by the cross-product (7a).
Now we consider at a point "0" in an undamaged continuum a diffe-
rential tetrahedron, as shown in Fig. Ia.

undomoged conflgurollon on1sotrop1c domoged conf f1CI111ous undomoged conf

x,

ol bl cI

Fig. Anisotropie Tensor of Continuity

Such a tetrahedron can be characterized by a system of bi-vectors:

232 ]. Betten

d 1S. -E .. k(dx 2 ). (dx 3 \!2

1 1J J

d 2 S. - E .. k ( dx 3 ) . (dx 1\!2
1 1J J
(9)
d 3S. - E .. k ( dx 1). (dx 2 \!2
1 1J J

d4 S = E .. k [ ( dx 1) . - (dx 3 ) j] [(dx 2 )k- (dx 3 )k]/2

1 1J J '

where the surn 1S equal to the zero vector:

2 3 4
d 1S. + d s.1 + d s.1 + d s.1 0. ( 10)
1 1

The sarne tetrahedron in a darnaged state (Fig. 1b) can be characte-

rized by the following systern of bi-vectors:

d1 s.1 a . .k(dx2). (dx 3 \!2

1J J
- ad Is.
1

ct 2 s.1 ß .. k(dx 3 ). (dx 1\!2

1J J
- ßd 2S.
1
( 11)
d3 s.1 -Y .. k(dx 1). (dx 2 \/2 :: Yd 3 S.
1J J 1

d4 s.1 = li 1..Jk [ ( dx 1) J. - (dx 3 ) j] [ (dx 2 \ - (dx 3 )k]/2 = lid 4 Si '

where aijk- a Eijk' ßijk =ß Eijk' etc. are total skew-symmetric tensor
of rank three, which have the essential cornponents a 1 ~ 3 = a, B123 = ß,
etc., respectively. We see, the vectors d J~Si, ... , d 4 ~s.1 in Fig. lb dif-
fer frorn the corresponding vectors of the undarnaged configuration (Fig.
Ja) only in lengths. It should also be noted that the vector surn is not
equal to the zero vector:
~ ~

d Js~. 2~ J~
+ d S. + d S. + d S.
4~ 1
~ 0. ( 12)
1 1 1 1 1

unless there 1s no darnage (a = ß = y = li = 1) according to ( 10) or the

darnage is only isotropic (a = ß = y = lt) .
Formulation of Anisotropie Constitutive Equations 233

Next we should consider a fictitious undamaged configuration (Fig.

lc) characterized by the following system of bi-vectors:
A
d1 s. 1
E:.
1J
1
"k (d~2). (d~3\12 - d 1
J
s.
d2 s.
1
E:.
1J
"k (d~3). (d~ 1 \l2 - d 2
J
s.
1
( 13)
d3 s.
1
E ••
1J
k(d~l). (d~ 2 \12
J
- d3 s.
1

where, by analogy to ( 10), the vector sum is equal to the zero vector:

d jsA. + d 2As. + d 3As. + d4As. 0 .. ( 14)

1 1 1 1 1

The three area vectors d 1Si' ... , d 3 Si in ( 13) are identical to the cor-
4A
responding vectors of the damaged configuration. The fourth vector d S.,
4 1
4~
having the same magnitude as d S., differs from d S.1 not only in length,
1
4A 4
but also in its direction. Therefore, the vectors d Si and d Si are con-
nected by a linear operator ~ of rank two (second-order tensor):

( 15)

By comparing the systems of bi-vectors (9), (I 1), ( 13) and by using

the equations ( 10), ( 14) in connection with the transformation ( 15), we
find the following relation:

where the transvection W.l r ErJ"k leads to a third-order tensor of conti-

nuity:
234 ]. Betten

\)!, E • ( 17)
lr qk

which is skew-symmetric with respect to the bracketed index pair [jk].

Because of a. 'k =a E. 'k' etc. the terms on the right-hand side of
1J
( 16) are vectors with magnitudes of ld
1J I
sl = aljk(dx 2 )j(dx 3 )k/2, etc.
I 2 3
and with the same directions as the base vectors e. , e. , e. be-
1 1 1
longing to the Cartesian coordinate system. Therefore, in connection
with ( 17), the relation ( 16) can be written in the form of:

( 18)

in which we can immediately see the decomposition:

,~, I a ( 19)
'+'ijk = ei ljk +

or because of aljk- a Eljk' etc., we can write:

w1J
.. k ( 20)

From ( 17) we can find the dual relation

\)J .. k : ;: \)!, [ 'k] = E 'k \)!, <=>\)!. = E 'k\)J .. k/ 2 ' ( 2 I)

1J 1 J J r 1r 1r rJ 1J

which is similar tothat in (7b). Finally, we can use (21) to deduce the
diagonal form:

\)! .. \)!. E. /2 diag {a, ß, y} . ( 22a)

1J 1pq Jpq

Inserting (20) into (22a) and replacing o1 j ::: 1ej, etc. we see, that the
secend rank tensor of continuity can be decomposed in terms of dyadics
formed from the base vectors:

I I 2 2 3 3
\)! .. a( .!: ® e) .. + ß( e ® e) .. + y( ~. 0 e) .. ( 22b)
1J - 1J - lJ - - 1J
Formulation of Anisotropie Constitutive Equations 235

The relations (21) and (22a) are illustrated 1n Fig. 2.

anisotropic darnage (a * ß*y * a)

0 y 0/
/
/
0/ 0
/ 0 0 -ß
/ /
/ 0 a 0 0
/
/
0 0/
/
0 /
"I i =31
/
0 0 /0 / 0 ß 0
/ -/
0 / 1i =zl
/
0 /0 a / 0 0 y
/ /
/ /
/0 -a 0 Y'li=ll
Fig. 2 Dual Tensor of Continuity

Especially, frorn Fig. 2 we can see the skew-syrnrnetric character of

the third order tensor of continuity indicated in ( 17) and its three es-
sential cornponents a, ß, y. These values are fractions which represent
the net cross-section elernents perpendicular to the coordinate axes x 1,
x 2 , x 3 (Fig. lb) and which can be rneasured in tests on specirnens cut
along three rnutually perpendicular directions x 1, x 2 , x 3 .
The darnage rnay sornetirnes develop isotropically, as observed by
JOHNSON [ 12] for R.R. 59 alurniniurn alloy. In this special case (a ß =

= y = w) the second rank tensor of continuity (21) is a spherical tensor:

W·lJ'k = W E.k Ö.
J r 1r
= W E.lJ'k <=>w.
1r
= W Er J.kE lJ
.. k/2 = \),' Ö.
1r
(23)

and the third-order tensor lS now total skew-syrnrnetric (Wijk- W[ijk])'

in cantrast to (21).
Instead of the continuity tensor ~tobe found in (21) we can use
the darnage tensor w defined as:

w.. -
lJ
eS .• -
lJ
wlJ
.. (24a,b)
236 ]. Betten

and characterized by the dual relation:

E W <=> W E .kw .. k/2 . (25)

jkr ir 1r rJ 1J

3. STRESSES IN A DAMAGED CONTINUUM

In the undamaged continuum (Fig. 3a) CAUCHY's formula

p. = O .. n. ( 26)
1 J1 J

was derived from the principle of equilibrium, where p., n. are the com-
1 1
ponents of the stress vector p and the unit vector n, respectively.

Pi= Pi 4>1nl

CAUCHY 's tensor crii net - stress tensor 0'··IJ

o) b)
Fig. 3 Stress Tensor regarding: a) an undamaged, b) a damaged continuum
Formulation of Anisotropie Constitutive Equations 237

In the same way we can also find the corresponding relation for a
damaged continuum:

( 27)

where wjk are the components of the continuity tensor ~ according to

(22a,b).
The surface elements dS and dS 1n Fig. 3 are subjected to the same
force vector:

dP.
1
p.dS - p.ds
1 1
dP.1 (28)

Thus, after comparing (26) to (27) we can finally find the actual net-
stress tensor 2• which is simply a transformation from CAUCHY's tensor:

0 ..
1]
w.1r &rJ. 0 ..
]1
<= > o ..
1]
~

0 ..
]1
(29a,b)

(-I) ~ (-1) (-1)

By suitable transvections we find o .. o .k ij 0 jk = wik . = W.k and 0
1] J 1
As indicated in (29a,b) the actual net-stress tensor &is non-symmetric,
unless we have isotropic darnage expressed as in (23).
Because of the symmetry O ..
1]
= (o 1]
.. + o .. )/2 of CAUCHY's stressten-
]1
sor o we can write the following representations:

0 ..
1] <w.1po.]q + o.1qw.JP )o pq /2 -
cpijpqopq (29c)

0 .. (\1!~-l)c;_ + w~-l)c;_ )o 12 - <P 0 (29d)

1] 1p ]q 1q JP pq ijpq pq

from ( 2 9a, b) . We see, the fourth-order tensors ~ and <P de f ined in ( 29c, d)
are only symmetric with respect to two indices:

cpijpq = cpjipq , <P •• <P •• (30a, b)

1]pq l]qp

so that, for instance, the tensor ~ can be decomposed into a symmetric

and an antisymmetric part [I 1]:
238 ]. Betten

<p ••
lJpq
(I(; .
1p Jq
o. + o1q
. I(; JP
. + I(; . o.
1q JP
+ o1p
. wJq
. )I 4 +

+ <w . o.
1p Jq
+ o1q
. I(; JP
. - I(; . o . - o1p
1q JP
. wJq
. ) 14 ( 3 I)

where the symmetric part is identical to the tensor ~ in (6) introduced

by RABOTNOV [10].
Furthermore, from (29) and (30) one can see, that the actual net-stress
tensor o is non-symmetric in the anisotropic darnage case:

yl ß , o/y . (32)

This fact is disadvantageous, and therefore it is awkward to use the

actual net-stress tensor § in constitutive equations with a symmetric
strain rate tensor d. A transformed net-stress tensor t is introduced 1n
[sJ, which is symmetric, and thusly more advantageous. This tensor is
defined through the following operation:

t .. (&ik 1(;(~1)
kj
(33)
lJ

Inserting (29d) into (33), one can find:

(-1)
t .. c.. 0 => 0 .. c.. t (34)
lJ ljpq pq lJ lJpq pq

where

(-I)
clJpq
.. (I(;. I(; .
1p Jq
+ I(; . I(; . ) I 2 wi th C ..
1q JP lJpq
( 35)

is a symmetric fourth-order tensor of continuity having the following

properties:

clJpq
.. cJlpq
.. clJqp
.. c ..
pqlJ
(36)

Because of (22a,b) and (36) the fourth-order tensor of continuity 1n

Formularion of Anisotropie Constitutive Equations 239

(35) can be represented by a 6 x 6 matrix, which has the diagonal form:

(37a)

.
d 1ag la2 , ß2 , l, aß/2, ßr/2, ra/2l , (37b)

that is, the components of the "pseudo-net-stress" tensor! are given as

0 /0.2 o 12 ;(aß) o 1/(aß)

tII II tl2 tl3

t 2I t22 0 ;ß 2 t23 02/(ßy) (38)

t 12 ' 22

2
t31 = t 13 ' t32 = t23 ' t33 = 03/1

An interpretation of the introduced "pseudo-net-stress" tensor in (33)

can be given in the following way. An alternative form of CAUCHY's for-
mula (26) is:

dP. 0 .. dS. (39)

l Jl J

where dP. is the actual force vector (28). According to ( 15) we can
1

write:

(-I)
dP. o . . \(!. dS
A

(40a)
l J1 Jr r

or substituting o .. C.. t we find the relat ion:

J1 J1pq pq

dP.
1
w.1p t pr ds r (40b)

. h can b e mu l t1p
wh 1c . l.1ed b y '~ki I) 1' ( -
so that we have Wk(~l)dP.
l l
tk r dS r , or
after changing the indices:

dP.
1
t ..
Jl
ds.J (4 I)
240 ]. Betten

Comparing (39) to (4 I) we see, that (41) can be interpreted as CAUCHY's

formula for the fictitious undamaged configuration (Fig. lc), which is
subjected to the pseudo-force dPi = w~~I)dPk instead of the actual force
dP .. The "pseudo-force" vector dP. is equal to the actual force vector
1 1
dP., if the continuum is in an undamaged state (W .. = 6 .. ). In the total
1 1J l J
damaged state (w .. = 0 .. ), however, dP. is approaching infinity.
1J l J l

4. CONSTITUTIVE EQUATIONS INVOLVING DAMAGE AND INITIAL ANISOTROPY

One must consider the fact that the symmetric tensor-valued functions
(Sa,b) arevalid for an isotropic material in an anisotropic damaged
state. Furthermore, one must differentiate between the anisotropic da-
mage growth and the initial anisotropy resulting from a forming process,
for instance rolling. Then constitutive equations can be found in ex-
pressions such as those in (I).
In formulating constitutive equations like (I) one has to take the fol-
lowing into account: the undamaged case (w ~ O) immediately leads to
(Sa), while the strain rate tensor ~ approaches infinity as w approaches
6. With respect to polynomial representations of constitutive equations
+)
it is convenient to use the tensor

(-I) (-I)
D .. _ ( 6 ..
1J 1J
w .. )
1J
- wlJ
.. (42)

as an argument tensor instead of the tensorial darnage variable w. Thus,

expressions such as

d .. f .. (o , D , A ) (43)
1J 1J pq pq pqrs

must be taken into considerat ion, i. e. , where f .. = f.. are the compo-
1J Jl
nents of a second-order tensor-valued function of two symmetric tensors
of rank two and one symmetric tensor of rank four. The general represen-

+)
Then, the pseudo-force vector (41) 1s defined as dP. - D.. P., for 1n-
stance. 1 1J J
Formulation of Anisotropie Constitutive Equations 241

tation is given through a linear cornbination (2), where the g's are syrn-
rnetric tensor generatorsofrank two involving the argurnent tensors ~.

p, ~· Such generators, for instance, are given by rnatrix products of ty-

pes [ 14] :

J ,\ V \! ,\} (,\) (V) (V) (,\)

JO D + D o .. - 0 ik Dkj + Dik 0 kj (44)
~ ~ lJ

or [ 8] :

\) V ,\} (,\) (w) (V) (w) (V) (,\)

• D Do .. + A (45)
~ ~ lJ - o 1. k AkJpq
. D
pq
D
ikpq pq
ok .
J

Then, the condition of forrn-invariance

a.ka. 1 fk 1 (o , ... ,A
1 J pq pqrs
) = f .. (a a o , ... ,a a a a A
lJ pt qu tu pt qu rv sw tuvw
) (46)

1s fulfilled under the transforrnation of the orthogonal group ~·

In finding an irreducible set of tensor generators we note, that, ac-

cording to the HAMILTON-CAYLEY theorern [ 14] none of the exponents ,\,v
of the second-order tensors ~. D need be larger than 2, and note that a
syrnrnetric fourth-order tensor of power six and all higher powers can be
expressed in terrns of lower powers, that is W ~ 5, as shown in [I, 15,
I6J.
The coefficients ~a in the linear cornbination (2) are scalar-valued
functions of the integrity basis the elernents of which are the irre-
ducible invariants. In finding the integrity basis associated with the
representation (2) of the constitutive equations (43) we rnust take the
theorern into account [ 17], which states that the trace of any product
forrned frorn rnatrices and their transposes is a polynornial invariant un-
der the full or proper orthogonal group. Thus we form the following
traces:

l
tr d 0 tr L: ~a Qa 0 - L: ~ a tr Qa 0
a a
(4 7)
tr d D tr ~ A•D tr d A•D - d .. A.. D
lJ lJpq pq
242 J. Betten

3 f . .
.
Alternat1vely, we can form tr Q, tr Q2 , and tr d rom the l1near combl-
nation (2). The sets of invariants so formed contain reducible and equi-
valent elements, i.e. redundant elements. In writing down the possible
irreducible invariants, we can use some results from the theory of in-
variants, for instance the following LEMMAs [ 17]:
- If Q is a reducible generator, then tr XG and tr GX are redu-
cible, where ~ is an arbitrary tensor.
- The trace of a matrix product is unaltered by cyclic permutations
of its factors.
- The trace of a matrix product is equal to the trace of the trans-
pose of the product.
- The transpose of a matrix product TI 1s the matrix product formed
by writing down the transposes of the factors of TI in reverse or-
der, for instance, tr ~Qf = tr f Q ~
1 1 1 •

Further deliberations concerning reducibility and equivalence and con-

cerning the relevant consequences of the HAMILTON-CAYLEY theorem are gi-
ven in [13,14,17,18].
In the following some examples of constitutive equations (43) re-
presented through a linear combination (2) or in a canonical form (3)
are discussed.

4.1 Representation by Superposition

First, we assume that a representation of the tensor-valued

function (43) in the three tensor variables~. ~. A of rank two, two,
four, respectively,is given by "superposition":

( 48)

Then a tensor polynomial in~. ~. h is made up by a linear combination

(2) which includes the tensor generator

I [ A, v] (M~~,)J,V] + i~,)J,V))/2,
G ..
lJ lJ Jl
(49a,b)
Formulation of Anisotropie Constitutive Equations 243

3[A)JV] - (N~~,JJ,V]
G .. ' ' \,v 0,1,2; )J 0, I, ... ,5, (49c)
lJ lJ

where the following matrix products are introduced:

[\,V] (A) (V)

L .. - 0 ik Dkj (50a)
lJ

[\,)J,V] CA) (JJ) (v) [\,)J,V] CA) (JJ) (v)

M .. - o A . o N .. - Dl.k Ak.Jpq Dpq (50b,c)
lJ ik kJpq pq ' lJ

The representation of (43) 1s then given 1n the form:

1 [\,v] 3cp 3 [\,)J,V]

f .. G .. + ... + L: G.. , ( 5 I)
lJ lJ (A,JJ,V) lJ
A,)J,V

which contains 99 irreducible tensor generators: 9 from the set (49a)

and 45 generators each from the sets (49b,c). Note, that in (50a,b,c)
those values of A,)J,V which are bracketed by parentheses are exponents,
whereas a squarebracket gives any set \,vor A,)J,V the character of a
label which indicates several matrix products.
In the undamaged case (~ Q) the tensor 2 defined in (42) is
identical to KRONECKER's tensor ~. and the expressions (49a,c) are then
the generators of the single argument tensors ~. ~. only. This special
case leads to constitutive equations appropiate for the description of
the response of an anisotropic material in its undamaged state.
In the special case of an undamaged (D ~ 6) and isotropic (A)J ~ A0 )
. . CA) 2 [ \, )J, v] ~ CA +v)
1 [ \ ~ v] ~
cont1nuum we flnd from (49a,b,c) that G.. ~ 0 .. , G.. ~ o ..
3 [\,)J,V] ~ .r . lJ. ( l)J :J lJ.
and G.. u .. , 1.e., the representat1on 51 conta1ns the constl-
lJ lJ
tutive equation

(2)
d .. cpoo·. + cplo .. + cp2o .. (52)
lJ lJ lJ lJ

as a special case.
The representation (51) may be extented by using matrix products of
the types:
244 J. Betten

P~~.w.v] _ C\) A (w) D(v) Q~~.w.v] _ (53a,b)

lJ 0 ik kjpq pq ' lJ

that is by considering further tensor generators:

P [A,]J,V] Q [A,w,v] (54a,b)

- (ij) - (ij)

each of which are 45 irreducible generators. Instead of (51) we then

find a representation with 139 irreducible tensor generators.
The scalar-valued functions 1 ~(A v)' 2 ~(A )' etc. in (51) are
' 'W, V
functions in the elements of the integrity basis, which can be found by
forming the following transvections:

(2) (2)
f .. o .. , f. .D .. , f. .A.. o , f. .A.. o , f. .A.. D , f. .A.. D ,
lJ Jl lJ Jl lJ Jlpq pq lJ Jlpq pq lJ Jlpq pq lJ Jlpq pq

(55)

or, alternatively, by considering the basis invariants f .. , f .. f .. ,

ll lJ Jl
f. .f.kfk. of the tensorial expression (51), as was pointed out in detail
lJ J l
after (47).
We see, that the representation (51) with 99 irreducible generators
or its extended form with 139 irreducible elements are too complicated.
Therefore, we have to search constitutive equations for more practical
use.

4.2 Simplified Representations

The effect of anisotropy characterized by a fourth-order tensor A

can be expressed through a symmetric "image" tensor T of rank two, i.e.
through a linear transformation:

T •• T •• (56)
lJ Jl

Then we have to consider the constitutive expression:

Formulation of Anisotropie Constitutive Equations 245

d .. (57)
lJ

being a tensor-valued function of three symmetric tensor variables~. Q,

1· As was pointed out by SPENCER and RIVLIN [ 13, 18] in detail, the re-
presentation of the tensor function (57) is characterized by 46 irredu-
cible tensor generators and an integrity basis with 28 irreducible in-
variants. The representation of a non-polynomial function (57), which was
discussed by WANG[21 ], SMITH[J9]and BOEHLER[22], involves 16 tensor gene-
rators and only 22 invariants. In[J I] the author of this chapter has con-
sidered also 19 tensor generators, but the associated integrity basis
consits of 28 irreducible elements. It is remarkable that these 28 ele-
ments are identical to those given by RIVLIN and SPENCER [13, 14, 17, 18],
although these authors used a more extended representation than the re-
presentation discussed in [I I]. On the one hand, the former with 46 ten-
sor generators and, on the other hand, the latter with 19. Furthermore,
in [II Ja canonical form (3) for (57) is found by constructing fourth-
order tensor-valued functions 0 ~, 1 ~, 2H of the argument tensors Q and
1·
A further simplification in representing constitutive equations is
shown in [I I] by considering the tensor-valued function

(58)
lJ.(&pq '
d .. f.
lJ
1
pq )

in two second-rank tensor variables each of which is a linear transfor-

mation (29b) and (56) from CAUCHY's stress tensor ~. where the linear
operators ~ and A of rank four are determined from the effect of darnage
(~ = ~ - ~) and from the anisotropy of the material in its undamaged
state.
The only inconvenience in finding a set of irreducible tensor ge-
nerators for (58) is the fact, that the net-stress tensor ~ is non-sym-
metric. The list of tensor generators given in [I J] contains 13 symme-
tric tensor generators, and the associated integrity basis has 20 ir-
reducible invariants. A canonical form for (58) can also be found in
[ II].
246 ]. Betten

Another way of simplifying the form (43) 1s the tensor function

d .. f. .(t , A ) (59)
lJ lJ pq pqrs

in the symmetric "pseudo-net-stress" tensor t defined in (34) and the

symmetric anisotropy tensor ~ of rank four. The representation (2) for
(59) is characterized by 48 tensor generators, i.e. we have:

d .. (60)
lJ

(A,v = 0, 1,2; ~ = 0, 1, ... ,5), where the ~·s are scalar-valued functions
of the integrity basis.
Finally, instead of (58), we should consider the function

d .. f .. (t , T ) ( 6 I)
lJ lJ pq pq

1n the two symmetric second-rank tensors each of which is a linear

transformation (34) and (56) from CAUCHY's stress tensor ~· where the
-I
symmetric linear Operators ~ and ~ of rank four are determined by the
effect of darnage and by the anisotropy of the material in its undamaged
state.
It is well known [ 14], that the representation (2) of the tensor
function (61) involves nine generators, i.e., we have:

2 ( (A) (v) (v) (A))

d .. (62)
lJ 2 A ~=O~(A,v) tik 'kj + 'ik tkj
'
where ~(A,v) are scalar-valued functions of the integrity basis asso-
ciated with the representation (62), which can be found by forming tr dt
and tr dT. In this way we find the invariants

(hl) A+l v (A) (V+ I ) I. V+ I

t.. T .. - tr t T and t .. T .. -trtT (63a,b)
lJ Jl lJ Jl

some of which are reducible or equivalent. In seperating the redundant

elements, as was pointed out in detail after (47), we find from (63a,b)
Formulation of Anisotropie Constitutive Equations 247

the integrity basis:

l
2 3
tr _:::, tr T tr T
( 64)
2 2 2 2
tr ~_:::, tr tT , tr ~ _:::, tr t T

which is associated with the representation (62).

We can write the constitutive equation (62), for instance, in a ca-
nonical form, like (3), if we introduce [20] the following identities
for tensor generators:
Let X and Y be two symmetric second-order tensors, i.~. X.. X..
lJ J1
and Y.. Y .. , respectively. Then, the identities
lJ Jl

,[>.] Y(v) (65)

- "ijkl kl

[/.]
are valid for arbitrary values A and v, where the symbols ~ijkl and
[v]
nijkl are A and \! several fourth-order tensors defined in the following
way:

[ >.] (A) (A) (A) ' (A))/4

<xik 0j 1 + xil 0jk + 0ikxjl + '\lxjk (66a)
~ijkl -

[v] ( \)) ( \)) ( \)) ( \))

nijkl - (Yik 0jl + Yil 0jk + 0ikyjl + 0i1Yjk )/ 4 (66b)

Similar identities for non-symmetric tensors were introduced in [I I].

4.3 Separation of Tensor Variables

In the following a simple method will be suggested [3] where one

can find representations for tensor-valued functions by separating ten-
sor variables. For instance, the function (61) may be represented by se-
parating the two variables t and T in the following way:
248 J. Betten

d .. f .. (t,T) [o.k 0xk.CT) + 0x.k(T)6k.]/2 +

1J 1J ~ ~ 1 J ~ 1 ~ J

+ [t.kiXk.(T) + 1X.k(T)tk. ]/2 +

1 J ~ 1 ~ J

(2)2 2 (2)]
+ [ t .k xk.(T) + x.k(T)t,. 12 , ( 6 7a)
1 J ~ 1 ~ l<J

I 2 [ (A) ). 1 0.)]
d .. -2 L: t.k Xk.(T) + X.k(T)tk. . (67b)
1J \=0 1 J ~ 1 ~ J

This is a representation consisting of three parts (A = 0, 1,2), which

may be interpreted as the contributions of the zero, first, and second
orders found in the variable t. The three functions 0 ~, 1 ~, 2x are iso-
tropic tensor functions of the other variable T:

2
>x 1J.. ( T) L: cp(\,v) T ••
( v)
1J
(68)
~
v=O
Inserting (68) in (67b), the usual polynomial (62) is obtained.
In a similar way one can represent the symmetric tensor-valued
function (57) of three symmetric argument tensors:

I 2 [ (A) \ \ 0.) J
d .. L:
2 \=0 O.k Xk.(D,T) + X.k(D,T)Ok. +
1J 1 J ~ ~ 1 ~ ~ J

+ l ~ [D(\) Ay (T o) + Ay (T o)D(;\)J +
1'k k'J ~ ' ~ (69)
{_,
2 \=0 1'k ~ ' ~ k'J

I 2 [ 0.) \ (A) J
\ z.k(o,D)Tk.
+ -2 L: T.k zk.(o,D) + ,
\=0 1 J ~ ~ 1 ~ ~ J

where nine isotropic tensor functions each containing two variables are
introduced:

\
X .• (D,T)
2 [ (~) (v) (v) (~)] 0, I, 2. (70)
1J 2 L: Cfl( ' ) D. k Tk . + T. k Dk . ; A
J J
v= 0 "'~'V
~ ~
1 1
~.

The other functions>-!, Az in (69) have been defined in a similar manner.

Formularion of Anisotropie Constitutive Equations 249

Thus, the representation (69) consists of 81 terms, where some of them

are equivalent. After selecting the equivalent elements, one can find a
set of 43 irreducible terms [I].
The obove method of "separation of tensor variables" can be used in
order to find tensor-valued functions for several argument tensors.

REFERENCES

I. BETTEN, J., Elasticitäts- und Plastizitätslehre, Vieweg-Verlag, Braun-

schweig/Wiesbaden 1985.
2. BETTEN, J., Zur Aufstellung von Stoffgleichungen in der Kriechmecha~ik
anisotroper Körper, Rheol. Acta 20 ( 1981): 527-535.
3. BETTEN, J., Constitutive Equations of Isotropie and Anisotropie Ma
terials in the Secondary and Tertiary Creep Stage, in: Creep and Frac-
ture of Engineering Materials and Structures, Part II, (eds. B. Wil-
shire and D.R.J. Owen), Pineridge Press, Swansea 1984, 1291-1305.
4. BETTEN, J., Materialgleichungen zur Beschreibung des sekundären und
tertiären Kriechverhaltens anisotroper Stoffe, Z. Angew. Math. Mech.
(ZAMM) 64 (1984): 211-220.

5. BETTEN, J., Applications of Tensor Functions to the Formulation of

Constitutive Equations involving Darnage and Initial Anisotropy, TUTAM-
Symposium on Mechanics of Darnage and Fatigue, Haifa and Tel Aviv 1985,
tobe published in the proceedings (eds. S.R. Bodner and Z. Hashin).
6. KACHANOV, L.M., On the Time to Failure under Creep Conditions (in
Russian), Izv. Ak. Nauk USSR Otdel. Tekh. Nauk, 8 ( 1958): 26-31.
7. RABOTNOV, Y.N., Creep Problems in Structural Members, North-Holland
1969.
8. BETTEN, J., Darnage Tensors in Continuum Mechanics, Euromech Colloquium
147 on "Damage Mechanics", Paris VI, Cachan 1981, published in Journal
de Mecanique the6rique et appliquee 2 ( 1983): 13-32.
9. MURAKAMI, S. and OHNO, N., A Continuum Theory of Creep and Creep Da-
mage, in: Creep in Structures (eds. A.R.S. Ponter and D.R. Hayhurst),
Springer-Verlag, Berlin/Heidelberg/New York 1981, 422-444.
10. RABOTNOV, Yu. N., Creep Rupture, in: Applied Mechanics Conference
(eds. M. Hetenyi and H. Vincenti), Stanford University 1968, 342-349.
I I. BETTEN, J., Net-Stress Analysis in Creep Mechanics, Ingenieur-Archiv 52
( 1982): 405-419.
12. JOHNSON, A.E., Complex-Stress Creep of Metals, Metallurgical Reviews 5
( 1960): 447-506.
250 ]. Betten

13. SPENCER, A.J.M. and Rivlin, R.S., Finite Integrity Bases for Five or
Fewer Symmetrie 3 x 3 Matrices, Arch. Rational Mech. Anal. 2 ( 1958/59):
435-446.
14. SPENCER, A.J.M., Theory of Invariants, in: Continuum Physics. Vol. I,
(ed. A.C. Eringen), Academic Press, New York 1971, 239-353.
15. BETTEN, J., Integrity Basis for a Second-Order and a Fourth-Order Ten-
sor, International J. Math. & Math. Sei. 5 ( 1982): 87-96.
16. BETTEN, J., Irreducihle Invariants of Fourth-Order Tensors, Fifth
International Conference on Mathematical Modelling, University of Cali-
fornia, Berkeley 1985, to be published in the proceedings: Mathematical
Modelling in Science and Technology (eds. X.J.R. Avula, G. Leitmann,
C.D. Mate, Jr. and E.Y. Rodin), Pergarnon Press, New York 1986.
17. RIVLIN, R.S. and SMITH, G.F., Orthogonal Integrity Basis for N Symme-
trie Matrices, in: Gontributions to Mechanic (ed. D. Abir), Pergarnon
Press, Oxford/ ... /Braunschweig 1969, 121-141.
18. SPENCER, A.J.M. and RIVLIN, R.S., The Theory of Matrix Polynomials and
its Application to the Mechanics of Isotropie Continua, Arch. Rational
Mech. Anal. 2 ( 1958/59): 309-336.
19. SMITH, G.F., On Isotropie Function~ of Symmetrie Tensors, Skew-Symme-
tric Tensors, and Vectors, Int. J. Engng. Sc. 9 ( 1971): 899-916.
20. BETTEN, J., On the Representation of the Plastic Potential of Aniso-
tropie Solids, Colloque International du CNRS n° 319, Comportement
plastique des solides anisotropes, Grenoble 1981, published in the
the proceedings: Plastic Behavior of Anisotropie Solids (ed. J.P.
Boehler), CNRS, Paris 1985, 213-228.

21. WANG, C.C., On Representations for Isotropie Functions, Part I,

Arch. Rational Mech. Anal. 33 ( 1969): 249-267.

22. BOEHLER, J. P., On Irreducihle Representations for Isotropie Scalar

Functions, Z. Angew. Math. Mech. (ZAMM) 57 (1977) : 323-327.
 Chapter 13

INTERPOLATION METHOS FOR

TENSOR FUNCTIONS

J. Betten
Technical University Aachen, F.R. Germany

I . INTRODUCTION

In this chapter polynomial interpolation is considered and extended

to tensor-valued functions of one and two atgument tensors. Let x
a
(a = I ,2, ... ,n) be dist inct points and y
a
corresponding values. The po-
lynomial of degree n-1

n
y f(x) = L La(x)ya +Rn ( I)
a =1

1s called "LAGRANGE interpolation formula", where the polynomials

n
L (x)
a TI ( x - x 6 ) I ( xa - xß ) , a = I ,2, ... ,n (2)
ß=I
s+a
are introduced. lt 1s clear that La(x 8 ) is equal to one for a ß and
equal to zero for a ~ ß. The remainder in (I) is given by
252 J. Betten

R (3)
n-1

where min (x, x 1, ... , xn) < ~ < max (x, x 1, ... , xn).
Since the above interpolation has a unique solution, all other re-
presentations must coincide with the LAGRANGE polynomial (I). But the
LAGRANGE formula is disadvantageous, if we pass from a space of dimen-
sion n to a space of one higher dimension. Then the "NEWTON formula"

(4)

is to prefer, where a 0 = f(x 1). The coefficients a 1, a 2 , ... , an are

called the "devided differences" of the first, second, ... , n-th order,
respectively.

2. TENSOR FUNCTION OF ONE VARIABLE

In extending the LAGRANGE interpolation method to a tensor-valued

function

y .. f .. (X) Y.. i,j I ,2,3 ( 5)

lJ lJ - Jl

of one symmetric argument tensor (X .. = X.. ), we consider the principal

lJ J1
values AI= x1 , All= x11 , All!= x111 of the tensor ~ as interpolating
points and find, by analogy to (I) and (2), the tensorial representation
[ IJ

Ill
Y •.
lJ
f. . (X)
lJ -
I Cl.
f"' L .. (XJ + R .. (X)
.... lJ - lJ -
(6)
Cl.= I

where the values fa = Ya, a = I,II,Ill, correspond to the principal va-

a
lues ~· The tensors ~· a I,II,Ill, are defined by

(Xik- A(a+I) 0 ik)(Xkj - A(a+ll) 0 kj)

(7)
- (A(a) - A(a+I))(A(a) - A(a+II))
Interpolation Methods for Tensor Functions 253

where the summation convention is utilized, and ~ represents the KRON-

ECKER tensor. The tensor-valued remainder in (6) is given by

R .. (Ba)
~J

and can be expressed by

R .. J_ (X(3) JIX~~) - d 3 tOP (8b)

3'. .. - J2X .. J30 .. )
lJ
dX 3
~J ~J ~J ~J

with the principal invariants (elementary symmetric functions)

(9a)

(9b)

(9c)

which are the elements of the integrity basis.

In (8b) we see the HAMILTON-CAYLEY equation

( 10)

i.e., the tensor-valued remainder (8a,b) is always equal to the zero

tensor.
The tensors (7) can be written in the form

<). .. a o
<po ij +
am X
'~'I ij +
am x\~)
'~'2 lJ ( I I)
~J

if we introduce the abbreviations

a = >.. >.. I ( 12a)

cpO - (~I) (~II) N(a) '

( 12b)
254 ]. Betten

( 12c)

with

( 13)

Thus, we can represent the tensorial interpolation formula (6) as

(2)
y.. = .. + <ll IX.lJ. + <ll2Xl. J.
<ll 0° lJ ' ( 14)
lJ

where the scalars <ll0 , <ll 1, <ll 2 are given by

III
<ll 'J =L 'J 0, I ,2 . ( I 5)
O:=l

The coefficients ( 15) can also be found by solving the system of linear
equations

( 16)

which is obtained by inserting the distinct principal values AI - XI

etc. into equation ( 14).
The tensor-valued function ( 14) is an "isotropic tensor function" be-
cause the relation

a. a. Y f. .(a. a. X ) ( 17)
lp Jq pq lJ lp Jq pq

is satisfied under any orthogonal transformation (aikajk = akiakj =

=eS •. ). According to the definition ( 17), we can say: A tensor-valued
1J
function (5) of one tensor variable ~ is isotropic if and only if it has
Interpolation Methods for Tensor Functions 255

a representation of the form ( 14), where the ~

V
are invariants of X and
~

hence can be expressed as functions of the principal invariants of X.

The determination of the ~ is given by ( 15).
V
As an alternate approach, we find, by extending the NEWTON formula (4),
the tensorial representation

y .. f. . (X)
lJ lJ ~

( I 8)

Further terms are not possible, because the second rank tensor ~ has on-
ly three principal values. Moreover, the last term in ( 18) vanishes be-
cause of the HAMILTON-CAYLEY theorem. Thus we have the formula

y .. ( I 9)
lJ

where the coefficients can be found by inserting the principal values:

(20a,b)

( 20c)

The interpolation formula ( 19) can be written as

(2)
y ..
lJ
= lj!Oo .. + \)!IX .. + I)J2X. .
lJ lJ lJ
, ( 2 I)

where the scalars are defined by

a - (22a)
0

(22b,c)

The tensor-valued function (21) 1s an isotropic tensor function and must

256 ]. Betten

coincide with (14), i.e., the scalar coefficients (15) and (22a,b,c) are
identical:

v=0,1,2, (23)

as we can easily prove.

3. TENSOR FUNCTION OF TWO VARIABLES

The problem, to construct a scalar polynomial

z = f(x,y) (24)

of two scalar variables sastisfying the mn interpolation conditions

f(x ,y ) = z a I ,2, ... ,m

( 25)
a o ao 0 I ,2, ... ,n

is solved by the interpolation formula

m n
z = I I L (x)L (y)f(x ,y ) + R(x,y)
a 0 a 0
(26)
a= 1 o= 1

which may be regarded as the generalization of the LAGRANGE formula (I)

to two variables. Thus the polynomials L (x) and L (y) are given in the
a 0
form (2), while the remainder R(x,y) can be expressed by

R(x,y) = Rm_ 1(x) + Rn_ 1(y) - M(xy) , ( 27)

where the first and second term can be read from

(x - x 1) (x - X ) amt<~,,y)
m
R (28a)
m-1 m! axm

n
(y - y I) (y - y ) a f(x,n 1)
n (28b)
R
n-1 n! ayn
Interpolation Methods for Tensor Functions 257

and the third term in (27) 1s given by

M(xy)

(28c)

The points 1,n 1) and C~ 2 ,n 2 ) are lying in the interpolation region.

(~

Extensions to any number of variables will follow in a similar fashion.

An extension to a tensor-valued function

Z .. =
lJ
[f. .(X,Y) + f .. (X,Y)]/2 =
lJ~- Jl~~
Z ..
Jl
(29)

of two symmetric tensor variables (X .. X .. , Y .. Y.. ) 1s achieved by

lJ Jl lJ Jl
the formula

III
f .. (X,Y) =
lJ ~ ~
I aL .k(x) 0 Lk.(_Y)f(a
a ,p=I 1 - J - ,p
) + R .. (X,Y)
lJ - -
, (30)

using the following tensor polynomials:

(X.-,\(
1p a+ I) 6 1p
.)(Xk-,\(
p a+ II) 6 pk) ( 3 I a)

( yk g - )J ( p + I ) 6 k q ) ( y qJ. - )J ( p+ I I ) 6 <.LL
.)
PLk. (Y) - (31b)
J ~
(JJ(p) - JJ(p+I))(JJ(p) - u(p+II))

where ,\(a)' a = I,II,III, and )J(p)' p = I,II,III, are the principal va-
lues of the argument tensors X and I• respectively. The remainder in
(30) is given by

R .. (X, Y) = R .. (X) + R .. ( Y) - M .. ( XY) (32)

lJ~~ lJ~ lJ~ lJ~~

and is a tensorial extension of (27). Because of the HAMILTON-CAYLEY

equation ( 10), the tensor-valued remainder (32) vanishes like (8a,b).
258 ]. Betten

The generalization of the formula ( 19) to two tensor variables can be

expressed by

f. .(X,Y)
lJ - -

(33)

where the abbreviations

I I
A .. - (X .. AI 6 .. ) B .. - (Y .. - lli 6 .. ) ( 34a, b)
lJ lJ lJ ' lJ lJ lJ '
2
A .. - (Xik AI 6ik)(Xkj AII 6kj) (3Sa)
lJ
2
B .. = (\k- lli 6ik)(Ykj- llii 6kj) ( 3Sb)
lJ

are used. The nine coefficients a 0 , a 1, ... , c 4 in formula (33) are de-
termined by the nine interpolation conditions

a, p = I ,II , III . (36)

For example, we find a 0 by substituting x 11 =XI = \• YII = YI = lli and

t 1 1 :: f(I,I)" The coefficient c 2 can be determined by substituting
XII :: XII = AII' y II :: YIII = )JIII and fii(XII, YIII) :: f(II,III). In
this way we can express the coefficients in (33) by the given values
(36):

ao - f(I,I)

f(I, I) - f ( II, I) f(I,I) - f(I,II)

al - bI -
AI All )JI -]JII

III III
a2 - L f(a,I/M(a) ' b2 - L f(I,P/N(P)
a=I P=I
Interpolation Methods for Tensor Functions 259

III
c2 - L (f(I,p) - f(II,p))/[(AI- AII)N(p)J
P=I

III
c3 - L ( f(a, I) - f(a,II))/[(~I- ~II)M(a)J
Ct=l

III
L f(a,p)(A(a+I) - A(a+II))(~(p+I) - ~(p+II))/
a,p=I

where the abbreviations

are used.
The tensor-valued function (30) or, alternatively, the approach (33) can
be represented by the same minimum polynomial

2
( ) 1 ,1, l~t)Y(v) ( 37)
f. . X' y = L 'I' (1t ,v) ik kJ' '
lJ~~ lt,V=O

where the system lji(~t,V) of scalar coefficients is given by:

260 J. Betten

By analogy to ( 14) the polynomial (37) of two tensor variables is an

isotropic tensor function because, in extension of ( 17), the relation

a. a. f (X,Y) = f. .Ca.~p a.Jq Xpq , a.~p a.Jq Ypq ) (38)

~p Jq pq ~ ~ ~J

is satisfied under any orthogonal transformation.

The polynomial (37) with the above determined scalar coefficients ~( lt,V )
may be regarded as the generalization of the simple case ( 14) to two ar-
gument tensors. Extensions to any number of tensor variables can follow
in a similar fashion.

4. INTERPOLATION AT COINCIDENT POINTS

In formulating the problern of polynomial interpolation for tensor

functions, we have assumed that the principal values of the argument
tensors, i.e. the interpolating points, are distinct. In the case of
Interpolation Methods for Tensor Functions 261

coincident points, we need the derivatives of the tensor function (5),

which can be obtained from ( 19) or (21):

t 1•
lJ
• - . 1ax PJ.
ay 1p \)! I o.. +
lJ
21);2X l. J. ' (39)

el J 1• _ at~
lq
1axqj. = 2w 2 o..
lJ
(40)

For example, 1n the case of AI = AIII + All' we find from ( 19) and (39):

(4la,b)

( 4 lc)

if we substitute Y1 = t 1 1(X 1 I = X1 A1 ), Y11 = t 22 <x 22 - XII

and tj 1<x 11 = A1 ) = fi· In a similar way we obtain

ao YI ' aI fl (42a,b)
I '

a2 [fI -
(YI - YIII)I(AI- AIII)]I(AI- AIII) (42c)
I

for the case AI All ~ AIII and

ao YI a I = (YI - Yll)I(AI - All) (43a,b)

'

a2 ( a I - fl I) I (AI - All) (43c)

'

if the case of AI ~ All = All! is considered.

Finally, if all principal values coincide, AI AIII' we determine

fl a2 = f~ I 12 , (44a,b,c)
I

But in this special case the argument tensor is a spherical one, X.. =
lJ
A1 6 .. , and therefore the formula ( 19) reduces to the trivial result:
lJ
262 ]. Betten

f .. = fiO ... Note, that the interpolation formulas for scalar functions
lJ lJ
y f(x) approaches the TAYLOR expansion for f(x) at x 0 , if we make xa'
a = 1,2, ... ,n, coincide at x 0 .
In extending the interpolation at coincident points to tensor
functions of two variables we need derivatives of the function (29) with
respect to both the variable X and Y. For example, we have to use the
derivatives

z ~ - caz. ;ax + az. ;ax . ) I 2 (45a)

lJ lp pj pl
0

JP
.
z - caz. ;ay + az.JP ;aY pl.)/2 (45b)
lJ lp pj
0 0

az ~ az'. az.
. .
az.
z' -
]
4
(~ + ________lE + ~ +
aY . aY
pi
ax
pj
~)
pl
( 45c)
PJ

and so on.

5. POLYNOMIALS OF SECOND-ORDER AND FOURTH-ORDER TENSORS

In the following we consider a tensor-valued function

zlJ 0 0 f. .(X
lJ pq
,Y
pq
,A
pqrs
) (46)

of two second-order and one fourth-order tensor the general represen-

tation of which is given by a linear combination

z (4 7)

where the G's are symmetric tensor generators of rank two involving the
argument tensors ~. ~. 6· Such generators, for instance, are given by
matrix products of types
Interpolation Methods for Tensor Functions 263

1 ~ v v ~} _ (~) (v) Y(v)x(~)

1~ X + X~ ij = xik ykj + ik kj

In finding an irreducible set of tensor generators we note, that, accor-

ding to the HAMILTON-CAYLEY theorem, none of the exponents ~.v of the
second-rank tensors ~·X needs be larger than two. In generalizing the
HAMILTON-CAYLEY theorem it is pointed out in detail in [2] that a sym-
metric fourth-order tensor of power six and all higher powers can be ex-
pressed in terms of lower powers, that is P ~ 5. The coefficients ~a in
the linear combination (47) are scalar-valued functions of the integritv
basis the elements of which are the irreducible invariants. Tagether
with the invariants of the single argument tensors we have to consider
the set of simultaneaus or joint invariants.
Tensor valued functions like (47) are used in [3,4,5] in order to
represent constitutive equations of creep involving CAUCHY's stress ten-
sor ~· a darnage tensor Y of rank two, and a fourth-order tensor A cha-
racterizing the initial anisotropy or the anisotropy of the material in
its undamaged state.

6. SIMPLE EXAMPLES

Given the second-rank tensor

X •. (48)
lJ

we will find the third root

y .. X~ ~/3). (49)
lJ lJ

The principal values of (49) are AI 2. From (43a,b,c)

264 ]. Betten

we calculate:

a = 41/3
0

and find the coefficients (22a,b,c) of the polynornial (21):

41/3 - ~ 2-2/3 ( 50a)

3

(50b)

( 50c)

Thus, the third root (49) of the tensor (48) is given by

21/3 + 21/3 -
0
22/3 i13
X~ ~/3) 0 21/3 0 ( 5 I)
lJ
21/3- 21/3 +
0
22/3 22/3

This result was obtained by BETTEN [6] in another way.

In the theory of finite deforrnation the tensorial HENCKY rneasure of
strain and strain rate plays an irnportant role [5,7] because it can be
decornposed into a surn of an isochoric distortion and a volurne change [5,
8,9, 10, I I, 12]. The problern to represent the logarithrnic function

y ln X (52)

as an isotropic tensor function ( 14) is solved by deterrnining the scalar

functions in ( 15):
Interpolation Methods for Tensor Functions 265

(53)

Especially for the given tensor (48) we find the result

ln ~ ln 2 (3~2 0
0 ) (54)
0 3/2

by considering the relations (21), (22a,b,c), and (43a,b,c).

Other examples are X= exp ~. X= sin ~ etc. which can be treated in the
same way.
Finally, the elastic-plastic behaviour of solids loaded under uni-
axial stress a may be expressed by the stress-strain relations [ 13]

[ tanh(EE/OF ) n] 1/n , ( 55a)

(55b)

where E is YOUNG's modulus, and oF represents the yield stress determi-

ned in an uni-axial tension test. The exponent n regulates the elastic-
plastic transition. For instance, an elastic-perfectly plastic behaviour
is characterized by n ~ 00 • For engineering applications it is important
to generalize the relations (55a,b) to the multi-axial state of stress.
This can be achieved by an isotropic tensor function (21), where the
scalars (22a,b,c) are to express by the empirical relations (55a,b).
In the next paragraph a similar example is discussed in detail.

7. TENSORIAL GENERALIZATION OF NORTON's CREEP LAW

The NORTON creep law

or d Ka n , (56a,b)
266 ]. Betten

which 1s often used in order to describe the secondary creep behaviour

under uni-axial states of stress (o) should be generalized to multi-
axial states of stress [ 14], i .e., we consider an isotropic tensor
function

f . . ( o ) = cp*ru. . + cp'"'o. . + cp *o.( 2) (57)

d ..
lJ lJ - 0 lJ 1 lJ 2 lJ. ,

where d .. are the cartesian components of the "rate-of-deformation ten-

lJ
* cp"' cp* 1n (57) should be determined as
sor". The scalar coefficients cpo,
I' 2
functions of the experimental data (K, n) 1n (56a,b).
Alternatively, we can represent the constitutive equation 1n the
form

d .. f. .(o') cpou . . + cp I o. . + cp2o .. 2)

5; ' ' (
(58)
lJ lJ - lJ lJ lJ

where oij = oij - okk6ij/3 are the components of the stress deviator o'.
For the special case of incompressible behaviour (dkk = 0) we find from
(58):

(59)

with J; = oikoki/2, so that the constitutive equation (58) is reduced to

the form

d .. cp lo!. (60)
lJ lJ + cp2o!lJ'.

containing the traceless tensors

and o! '. - aJ3' /ao .. (6la,b)

lJ lJ

Wl· th ' I3.

J'3 -=Oijojkoki
' '
The uni-axial equivalent state of stress (index V) is characterized
through the tensor variables
Interpolation Methods for Tensor Functions 267

. {o, O, O) , ( oij
( oij ) V = d 1ag , ) V= d1ag
. {2 I
)Ü• - )Ü• I ) ,
- )Ü (62a,b)

( d .. ) V = d iag {d, -vd, -vd) , (62c)

lJ

where v 1s the transverse contraction ratio.

In the following the diagonal elements in (62a,b,c) are considered
as interpolating points, where two points coincide. Because the two
coincident points in (62a) are zero, it may be more convenient to de-
termine the coefficients ~O' ~I' ~ 2 in the constitutive equation (58)
instead of (57). Thus, we use the uni-axial creep law

d ( 63)

instead of (56b), where o' = 2o/3. Because of (62b), i.e. XII

=- o/3, and (63), we find from (43a) the coefficient

( 64a)

Furthermore, because of (62b), (64a), and Y11 - vd - vKon, we find

from (43b) the coefficient

n-1
a1 = (I + v)Ko . ( 64b)

The derivative fii at the coincident points XII XIII can be determined
in the following way. From (63) we derive:

f' = 3d/ao' = n(3/2)n(o')n-l = nd/o' ( 65a)

or

(65b)

From (62b,c) we read

268 ]. Betten

a' - 0/3 and

II

so that (65b) can be written as

' = 3Vn KOn-l .

f II (65c)

Considering (62b) and (64b), we calculate from (43c) the coefficient:

n-2 ( 64c)
a 2 = ( I + v - 3vn)Ko

Inserting (64a,b,c) in (22a,b,c), we finally determine the scalar

functions (~ = ~):

m
't'Q
= l9 ( I - Sv + 6vn)Kon (66a)

32 ( I + v + 23 vn)Kon-1 (66b)

n-2 · ( 66c)
~2 = ( I + v - 3vn)Ko .

In order to control the result. (66a,b,c) we make the following tests.

For the uni-axial case (i = j = I) the constitutive equation (58)
leads to

d ( 67)

and this equation immediately leads to NORTON's power law (56b), if we

insert the result (66a,b,c).
The transverse contraction ratio defined as

(68)

can be calculated from (58):

Interpolation Methods for Tensor Functions 269

v = - (cp 0 - J...cp o + lm...i)/(

3 I 9 TZ
m...
TU
+ lcp
3 I
o+ !!_m...i)
9 Tz
(69)

With the result (66a,b,c) we find the identity V = V.

The cond1tion (59) for imcompresible behaviour leads with (62b) and
(66a,b,c) to

( I - 8v + 6vn)Kon - 2( I + v - 3vn)Kon ,

from which we immediately read: v = 1/2.

The volume change is obtained from (58) by forming the trace
(i j = k):

( 70a)

For uni-axial loading (62b) we find with (66a,c) the trivial result:

( I - 2v)Kon _ ( I - 2v)d (70b)

which is identical with the trace of (62c).

The rate of dissipation of creep energy

D
. o . . d .. (71)
lJ Jl

can be expressed through

(72)

if the constitutive equation (58) is inserted into (7 I). For the uni-
axial case (62a,b,c) with J 1 : : o, J' : : o2 /3, J) :::: 2o2/27 we find from
2
(72) and (66a,b,c)

Kon+ I =
_ d
o (73)
270 ]. Betten

Thus, the hypothesis

o . . d .. _ od (74)
lJ J 1

can be used in order to determine the equivalent stress o as a function

of the stress invariants [see equations (79) to (82)].
For incompressible behaviour (59) the rate of dissipation of creep
energy (72) is reduced to the simple relation

.
D (75)

1n which only deviator invariants are contained. Uni-axial load (62b)

leads from (75) to (73), if we insert (66b,c) with v = 1/2 into (75).
Assuming incompressibility (59) and neglecting tensorial nonli-
nearity (~ 2 = 0 => a 2 = 0, ~O = 0, and ~I= a 1), we find from (58) the
simplified constitutive equation

d .. a 1 o~ . ( 76a)
lJ lJ

which can be written as

3 n-1 ,
d .. -2 Ko o .. (76b)
lJ lJ

if we use equation (64b) with V = 1/2. The result (76b) is identical to

the constitutive equation proposed by LECKIE and HAYHURST in [ 15]. If we
insert the MISES equivalent stress o =~ into (76b), we can find the
constitutive equation

n-1
d .. l2 K ( 3J ' ) 2 o ~ .
lJ
(76c)
lJ 2

used by ODQUIST and HULT in [ 16].

Inserting the deviator o~. o .. - J 1o.. /3 and its square
lJ lJ lJ
Interpolation Methods for Tensor Functions 271

2)
0.
1 (
0 <. 2.) - 2J 1o.. /3 + J 21u-" .. /9 ( 77)
lJ
0

lJ lJ lJ

into equation (58), one can find a constitutive equation of the form
(57) with the scalar coefficients

2
cp5 cpo - J 1cp/ 3 + J lcpzl 9 ( 78a)

cp* cpl- 2Jicpzl 3 ' (78b)

cp* - cp2 0 (78c)

As mentioned above one can determine the equivalent stress o in

(66a,b,c) as a function of the stress invariants, if we use the hypothe-
sis (74) in connection with (58), (72), (66a,b,c), and (56b). The result
is:

0 (79)

where the abbreviations

A _- (I- 8v + 6vn) J 1/9 , (80a)

B _ - 4( I + V + 3vn/2) Jz/3 , (80b)

C _ - ( I + V - 3vn) (3J J+ 2J IJ z /3) (80c)

are used. Thus, the skalar coefficients (66a,b,c) or alternatively (78a,

b,c) are functions of irreducible invariants and of experimental data:

cpo cpo <JI,Jz,J);K,n,v), ···• cp2

This statement is compatible with the representation theory of tensor-

valued functions (47) in which the coefficients cp are scalar-valued
a
272 ]. Betten

functions of the integrity basis.

For incompressible behaviour equation (79) 1s reduced to

0 (82)

because of (75) and V = 1/2. In this case the equivalent stress o 1n

(82) 1s influenced by the deviator invariants J;, J; and the parameter
n.
At the end of this paragraph two further examples should be men-
tioned. First, the tensorial generalization of the uni-axial stress-
strain relation (SSa) can be found in the alternative forms:

o.lJ. o(·l·o*o
'~-'
.. + •I·*E:
lJ
.. + ·I·*P~
'~-'I lJ '~-'2 lJ
) ,

o .. = o<woo .. + 1/! 1€! . + 1/!2€! \ 2)) ,

lJ lJ lJ lJ

where [similar to (78a,b,c)] the relations

with rl - E:kk and the scalar functions

Wo ( I - 2121)!2)/3 ,

1/!1 [C + 2/( I + v)]/2€ ,

2 2
1/! 2 = [ 1/( I + v) - C]/( I + v) E:

with 21.2 _ €!lJ.( Jl

'.. and
Interpolation Methods for Tensor Functions 273

C 19+4v 1 - 8 ( 1 + v) J
-z(l+v)2 (9 + 4v) 2

are used.
Another example discussed in [ 17] is the tensorial generalization
of the RAMBERG-OSGOOD stress-strain relation [ 18]

E = o/E + k(o/E)n .

By using the interpolation method we can find the tensorial represen-

tation

1 ( 2)
(..
~J
= cpocS ~J
.. +cp1o~. + cp2o. .
~J ~J
'

which ~s similar t 0 (58) and where the scalar functions

cpo 3
1 1-n)-
-( 0
+
E
t< 1-8V+6Vn)[o/E+k(u/E)n] '

have been derived ~n [ 17]. The alternative representation

E.. = cp*;:
0 u.. + cp*o.. + cp*2o (2)
..
~J ~J 1 ~J ~J

can be achieved, if we make use of the relations (78a,b,c).

274 J. Betten

8. SEPARATION OF TENSOR VARIABLES

In the following the uni-axial stress-strain-rate relation

d with D ·- 1/( 1-w) ( 83)

should be generalized to multi-axial states of stress, where w is the

parameter of darnage [3,4]. This can be done by representing (83) as a
tensor-valued function

1 ~ [ (v) (~) (~) (v)]

d ..
1J
f .. ( o, D) = 2: L \)!( ) o. k Dk . + D. k ok . . (84)
1J ~ - V' ~= O V, ~ 1 J 1 J

Now, the ma1n problern is: to determine the nine scalar coefficients

V,]J ) as functions of the integrity basis and experimental data. To

\)!(
solve this problern we suggest the following method which may be useful
for practical applications. A representation with the same tensor ge-
nerators as contained 1n (84) can be found by seperating the two ten-
sor variables o and D 1n the following way:

I
d .. = f. .(o,D) 2(Xikykj + Yikxkj) ( 85)
1J 1J - -

where the isotropic tensor functions

l
X.. X.. (o) = cp86 .. + cp'f 0 . . + cp2*0 (. .2)
1J 1J ~
1J 1J 1J
(86)
,\
ct<Ctro ) ~( 0 I' 0 11' 0 III)
~ V ~

l
(2)
y ..
1J
Y.. (D) =
1J ~
<Po o1J
.. + <P ID .. +
1J
<P2D ..
1J
( 87)

<P <P (trD,\) = <P]J(DI,DII'DIII)

]J ~

(]J,V = 0, 1,2 and A = 1,2,3) are used.

Thus, we find the representation (84) with the scalar coefficients
Interpolation Methods for Tensor Functions 275

lJ V 0, I, 2, (88)
'

where the scalars cp'~ are determined in (78a,b,c). Similarly to ( 15) the
V
coefficients ~ can be found by solving the system of linear equations:
lJ
m
2 (D ) I
~0 + ~IDI + ~2DI I

2 m
~0 + ~IDII + ~2DII ( D ) II
II (89)

m
~ 2 (D ) III
~0 + ~IDIII + 2DIII III

The exponents mi, ... ,miii 1n (89) are determined by using the creep law
(83) in tests on specimens cut along the mutually perpendicular direc-
tions x 1 , x 2 , x 3 .
(-I)
Because of D. . :=
lJ
w.. )
lJ
( ol..J -
and
w.. = d iag {( I - a) , (I - ß), (I - y)}, the principal values in (89) can
lJ
be expressed through [ 14]:

DII - 1/ß , (90)

where a,ß,y are fractions which represent the net cross-section elements
of CAUCHY 1 s tetrahedron perpendicular to the coordinate axes [3].
In the case of two equal parameters, for instance a 1 ß = y, the scalars
~ can be determined by using the interpolation method described 1n
lJ
section 4.
Instead of (86) we can use the isotropic tensor function

.t I 1 ( 2)
X .. X .. (o 1 ) cpou .. + cplo .. + cp2o .. ( 9 I)
lJ lJ - lJ lJ lJ

and find the representation

I D()J) l(v)J
d .. ( 92)
lJ 2 + ik 0 kj '

where the scalar coefficients cp are determined 1n (66a,b,c) and the

V
276 J. Betten

~ are taken from (89).

)J
The scalar coefficients ~(~,]J) _ ~V~)J 1n (92) must be functions of
the integrity basis

JI - J' - a~.a~./2 J' - ..

a lJ ' /3
' a '.kok.
0 kk 2 lJ Jl 3 J l '
D(2) D(3) 11' o! .D .. ( 93)
LI - Dkk ' 12 - kk 13 -
kk ' I
-
lJ Jl

11' - o~~ 2 )D .. 11' - o~.n\~). 11' - o~~ 2 )n\~),

2 lJ Jl 3 lJ Jl 4 lJ Jl

and experimental data. To show this we can start from (74) and find sl-
milarly to (79) the cubic equation

0 ' (94)

if we insert equations (85), (87), (91), (66a,b,c) into (74). In (94)

the following abbreviations are used:

(9Sa)

(95b)

(95c)

D- (95d,e)

We see, that the elements of the integrity basis (93) and experimental
Interpolation Methods for Tensor Functions 277

data are contained in (95a,b,c,d,e). Thus, the coefficients \jJ (v ,]J) =

:: ~ ~ in (92) are scalar functions of the integrity basis and experimen-
v jJ
tal data:

(96)

In the case of darnage (w .. ) and initial anisotropy (A. "kl) the con-
lJ lJ
stitutive equations discussed in [3,5, 14] are tensor-valued functions
like (46) which can be represented as (47), where the scalar functions ~a

can be determined in a similar way as described above.

REFERENCES

I. BETTEN, J., Interpolation Methods of Tensor Functions, Fourth Inter-

national Conference on Mathematical Modelling, ETH-Zürich (Switzer-
land) 1983, published in the proceedings: Mathematical Modelling in
Science and Technology (eds. X.J.R. Avula, R.E. Kalman, A.I. Liapis
and E.Y. Rodin), Pergarnon Press, New York / ... /Frankfurt 1984, 52-
57.
2. BETTEN, J., Integrity Basis for a Second-Order and a Fourth-Order
Tensor, International J. Math. & Math. Sei. 5 ( 1982): 87-96.
3. BETTEN, J., Darnage Tensors in Continuum Mechanics, Euromech Col-
loquium 147 on "Damage Mechanics", Paris VI, Cachan 1981, published
in: Journal de Mecanique theorique et appliquee 2 ( 1983): 13-32.
4. BETTEN, J., Net-Stress Analysis in Creep Mechanics, Ingenieur-Archiv
52 ( 1982): 405-419.
5. BETTEN, J., Elastizitäts- und Plastizitätslehre, Vieweg-Verlag,
Braunschweig/Wiesbaden 1985.
6. BETTEN, J., Elementare Tensorrechnung für Ingenieure, Vieweg-Verlag,
Braunschweig 1977.
7. FITZGERALD, J.E., A tensorial Hencky Measure of Strain and Strain
Rate for finite Deformations, J. Appl. Phys. 51 ( 1980): 5111-5115.
278 ]. Betten

8. RICHTER, H., Verzerrungstensor, Verzerrungsdeviator und Spannungs-

tensor bei endlichen Formänderungen, Z. Angew. Math. Mech. (ZAMM) 29
( 1949): 65-75.
9. SEDOV, L.I., Foundations of the Non-Linear Mechanics of Continua,
Pergarnon Press, Oxford / ... /Frankfurt 1966.
10. LEHMANN, Th., Einige Bemerkungen zu einer allgemeinen Klasse von
Stoffgesetzen für große elaste-plastische Formänderungen, Inge-
nieur-Archiv 41 ( 1972): 297-310.
I I. BETTEN, J., On the Creep-Behaviour of an elastic-plastic thick-wal-
led circular cylindrical Tube subjected to internal Pressure, Inter-
national Symposium on Mechanics of inelastic Media and Structures,
Warsaw 1978, published in: Mechanics of inelastic Media and Structu-
res (eds. 0. Mahrenholtz and A. Sawczuk), Polish Academy of Science,
Warszawa/Poznan 1982, 51-72.
12. BETTEN, J., Zur Kriechaufweitung zylindrischer Hochdruckbehälter,
Rheol. Acta 19 ( 1980): 517-524.
13. BETTEN, J., Zum Traglastverfahren bei nichtlinearem Stoffgesetz,
Ingenieur-Archiv 44 ( 1975): 199-207.
14. BETTEN, J., Applications of Tensor Functions to the Formularion of
Constitutive Equations involving Darnage and Initial Anisotropy,
TUTAM-Symposium on Mechanics of Darnage and Fatigue, Haifa and Tel
Aviv 1985, tobe published in the proceedings (eds. S.R. Bodner and
Z. Hashin).
15. LECKIE; F.A. and HAYHURST, D.R., Constitutive Equations for Creep
Rupture, Acta Metallurgica 25 ( 1977): 1059-1070.
16. ODQUIST, F.K.G. and HULT, J., Kriechfestigkeit metallischer Werk-
stoffe, Springer-Verlag, Berlin/Göttingen/Heidelberg 1962.
17. BETTEN, J., Tensorielle Verallgemeinerung einachsiger Stoffgesetze,
Z. Angew. Math. Mech. (ZAMM) 66 ( 1986), in press.
18. RAMBERG, W. and OSGOOD, W.R., Description of Stress-Strain Curves by
three Parameters, NACA Technical Note No. 902, July 1943.
 Chapter 14

TENSOR FUNCTION THEORY AND

CLASSICAL PLASTIC POTENTIAL

J. Betten
Technical University Aachen, F.R. Germany

I . INTRODUCTION

In the theory of elasticity an elastic ~otential (strain-energy

function W) is assumed, from which the constitutive equations can be
derived by using the relation 0,. = dW/ dE, . , where 0 and E are appro-
1] 1] - -
priately defined stress and strain tensors, respectively. In the iso-
tropic special case, when the elastic constitutive equation can be re-
presented as an isotropic tensor function

(2)
0 .. cp06 .. + cpiE .. + cp2E .. ' ( I)
1] 1J 1J 1J

the elastic potential is a scalar-valued function only of the strain

tensor and can be represented in the form W = W(S 1, s 2, s3), where s 1,
s2' s3 are the basic invariants of the strain tensor (finite or infini-
tesimal strain tensor). In [I] it has been shown in detail that the sca-
lar coefficients in (I) can be expressed through the elastic potential:
280 ]. Betten

(2a,b,c)

Eliminating the elastic potential, one can find the following sufficient
and necessary conditions

i.e., the elastic potential is "compatible" with the tensor function

theory (I), if the conditions (3) have been fulfiled [1].
In the classical theory of plasticity a plastic potential F is as-
sumed, and the flow rule

di\(ClF/Clo .. ) (4)
~J

is used ~n order to find the constitutive equations, where PE .. are the

~J

components of the plastic part of the strain tensor E, and the factor dA
~s LAGRANGE's multiplier.
In the following the classical plastic potential theory is compared
with the tensor function theory. It will be shown, that the former theo-
ry is compatible with the latter, if the material is isotropic, and if
additional conditions are fulfiled.
However, for anisotropic materials the plastic potential theory on-
ly furnishes restricted forms of constitutive equations, even if a gene-
ral plastic potential has been assumed [2]. Consequently, the classical
flow rule (4) must be modified for anisotropic solids. In the following,
appropiate modifications are discussed and resulting conditions of "com-
patibility" are derived.

2. ISOTROPY

For isotropic materials the plastic potential F = F(a . . ) can be ex-

~J

pressedas a single-valued function F = F[S (o)] of the irreducible in-

V ~

variants SV = tr ov,
~
V= 1,2,3, of the stress tensor. Thus, considering

the form F = F(s 1,s 2 ,s 3 ) and using the classical flow rule (4), one im-
Tensor Function Theory and Classical Plastic Potential 281

mediately obtains the constitutive equation

d P c. . = [< 2JF 1 C!s 1) eS. • + 2 < 2JF 1 as 2 ) o. . + 3 < 3F 1 C!s 3 ) o .< ~ ) ]d i\ ( 5)

1J 1J 1J 1J

Instead of ( 5), one can represent the tensor dp E .. as a symmetric ten-

1J
sor-valued function of one second-order argument tensor:

( 2)
t. . ( o ) = cp 0 cS. . + cp 1o. . + cp 2 o. . , (6)
1J pq lJ lJ 1J

where cp 0 , cp 1, cp 2 are scalar-valued functions of the integrity basis s 1,

s2, s3.
Comparing the "minimum polynomial representation" (6) with the re-
sult (5) based upon the plastic potential theory, one can find the fol-
lowing identities

By eliminating the plastic potential F 1n (7) we can find ad-

ditional restrictions imposed on the scalar functions cp0 , cp 1 , and cp2 , if
the existence of a plastic potential is assumed:

( 8)

Thus, in the isotropic special case the plastic potential theory is com-
patible with the tensor function theory, if the conditions of "compati-
bility" in (8) have been fulfiled.
For example, the requirements 1n (8) can be satisfied, if one assu-
mes:

( 9)

Then, because of (7), the plastic potential can be represented as

282 ]. Betten

F ( I 0)

where g 1, g 2 , and g 3 are arbitrary functions of the invariants s 1, s 2 ,

and s3' respectively.
The conditions of "compatibility" in (8) can be found in the following
way. The identities (7) can be written in the form

( I I)

where

( 12)

is normally a scalar-valued function of the integrity basis. The plastic

potential F in (I I) can be eliminated, if we consider the(~}equalities

( 13)

i.e., assuming the second partial derivatives of the plastic potential

are continuous. Hence, from (I I) and ( 13) we find a system of linear
equations in the partial derivatives 3L/3SA = L,A with A = 1,2,3:

( 14)

On the right hand side in ( 14) the abbreviations

(~ 0 ), 2 = 3~ 0 /3s 2 , ... ,(~ 2 ), 2 = 3~ 2 /3S 2 are used. Because the coefficient
determinant in ( 14) vanishes,
Tensor Function Theory and Classical Plastic Potential 283

cpl 0

0 - 3cp 0 , ( 15)
0

0 - ]cp
I

one can assume, that the right hand side of the system ( 14) is also
equal to zero. It is possible to find the three conditions of "compa-
tibility" in (8). Note, that these conditions are only sufficient. But
if we use CRAMER's rule, we find the following sufficient and necessary
condition

( 16)

which contains (8) as a special case. As has already been mentioned be-
fore, the conditions in (8) are only sufficient, whereas the similar
conditions (3) in elasticity are both sufficient and necessary.
For more complicated examples than the above mentioned the plastic
potential theory is not compatible with the tensor function theory. For
instance, MURAKAMI and SAWCZUK found in [3], that for the model of
prestrained plastic solids the flow theory of classical plasticity only
furnishes restricted forms of constitutive equations. In [4] LEHMANN
proposed an extended form of the classical flow rule (4) with respect to
experimental results and based upon thermodynamical considerations. In
[5] oriented solids are considered and an appropiate modification of the
classical flow rule (4) is discussed.

3. GRIENTED SOLIDS

In order to describe yielding and failure of oriented solids BOEH-

LER and SAWCZUK [6,7] use the tensor generator A =v ® ~· where the
vector v specifies a privileged direction. Then the constitutive
284 J. Betten

equation can be represented by the minimum polynomial

2
dpE:. . f. .(o )
I
E
( [A,v] M~~·vJ)
lJ lJ pq' pq
A
2 cp(A , V ) MlJ
.. +
Jl
( 17)
A, v=O
where [A,v] several symmetric tensor generators are formed by matrix
products of the forms

O (A)A(v) A,V 0, I ,2 . ( 18)

ik kj

The cp(A;V) 's in ( 17) are nine scalar functions of the integrity basis,
the elements of which can be found by forming all the irreducible traces
of the matrix products ( 18):

[A,v] (A) (v) A 0 =>V= 1,2,3 ( 19 a)

M - a A where A,v I ,2 and for
rr pq qp ' V 0=>A 1,2,3,

which can be written as:

SA tr aA T tr Av ; A,V 1,2,3
-
V-
}
~

( 19b)
tr aß Q2 tr OA 2 - tr Ao2 tr o2A2
Ql ~~ ' Q3
-
- ~
Q4 -

From the ten invariants in ( 19) only the seven stress-dependent lnva-
riants are essential for the plastic potential:

F F(M [A,v]) , where

,
/\,V I ,2 but for v 0 => A I ,2, 3 . (20)
rr

Thus, by using the flow rule (4), one can find the constitutive equation

( 21)

where A,V I ,2 and for v 0 => A I ,2,3 .

In (21) the A several fourth-rank tensors g[A] are defined as

Tensor Function Theory and Classical Plastic Potential 285

:\-I
[ :\]
Qpqij -
ao (:\ 7ao.. ~
{_,
[ 0 (a) (:\-I-a)
. 0 . + 0
( a)
. 0
CA-
.
I -a) J 12
(22)
pq lJ a=O pl qJ PJ ql

and have the following properties of symmetry:

[:\] [:\] [ AJ [:\]

(23)
Qpqij Qqpij Qpqji Qijpq

One can see, that because of (22) the value A cannot be equal to zero.
Therefore, only the seven stress-dependent invariants of the integrity
basis ( I9) are relevant for the plastic potential (20). Furthermore, by
comparing the constitutive equations ( I7) and (2I), one can see, that
the minimum polynomial ( I7) consists of nine tensor generators, whereas
(2I) has only seven terms, i.e., the plastic potential theory furnishes
only restricted forms of constitutive equations, even if a general
plastic potential is assumed in an anisotropic case.
It may be more useful for practical applications to represent the
constitutive equation ( I7) in a canonical form

(24)

where the fourth-order tensor-valued functions 0 ~. I~. and 2 ~ are defi-

ned 1n the following way:

0 (0) [2]
Hijkl - Cflco,o)mijkl + cp( 0, I )mijkl + cp(0,2)mijkl (2Sa)

I (0) [ 2J
Hijkl -
cp( I , 0) mi j k l + cp( I, I)mijkl + cp( I ,2)mijkl (2Sb)

2 (0) [2]
Hijkl -
cp(2, O)mijkl + cp(2, I )mijkl + cp(2,2)mijkl (2Sc)

[v]
having the symmetric tensors m \! = 0, I,2, of rank four:

[v]
mijkl (26)

Especially, for \! = 0 one finds in (26) the zero power tensor of rank
four:
286 J. Betten

[0] (0) (-I) (- I)

mijkl - mijkl - m1..J pq mpq kl = (o.ko'l
1 J
+ o.lo.k)/2
1 J mijpqmpqkl
( 2 7)

The representation found in the canonical form (24) consists of three

terms, which are the contributions of zero, first, and second-orders in
the Stress tensor ~' influenced by the functions O~' ~~, and 2 ~, respec-
tively. In finding these fourth-order tensor-valued functions, the iden-
tities

v Al _ (A) [A] (v)

l-fcll) +YX)
~ ~
.. =n 1J
1J
[v]
.. klxkl -
~ijklykl ,

[ A] 0) 0) 0) (A)
~ ijkl -
cxik 0 jr + xi 1 6 jk + 6 ikxj 1 + oilxjk )/4

[v] ( v) ( v) 6 Y( v) ( v)
n ijkl -
(Yik 0 jr + y i l 0 jk + 1k jl + 0 iryjk )/ 4

have been used [5], where X and Y are two symmetric second-order ten-
sors. Similar identities for non-symmetric tensors were introduced in
[ sJ.
Similarly to (24) one can represent the constitut!ve equation (21),
derived from the plastic potential theory, 1n a canonical form:

dp€ .. (28)
1J

where the fourth-order tensor-valued functions 0......._,

h 1........h , and 2h are defi-
ned by using (26) in the following way:

0
hijkl - (29a)

Ihijkl = 2[(aF/as2)m~~~l + (aF;a~3)mijkl + (aF/a~4)m[i~L]df\ (29b)

2hijkl = 3[(aF;as 3 )m~~~ 1 + 0 + 0 ]df\ . (29c)

Without loss of generality in the case of incipient motion, the

vector v can be regarded as an unit vector in the reference configuration
Tensor Function Theory and Classical Plastic Potential 287

[6]. Therefore, only v = 0, I in ( 18) has tobe taken into account. The
number of tensor generators in the constitutive equation ( 17) is reduced
from nine to six, i.e., the fourth-order tensor-valued functions (25a,b,
c) can be simplified into:

0 (0) (30a)
Hijkl - <p(O,O)mijkl + <p(O, l)mijkl

I ( 0) (30b)
Hijkl - <p( I ,O)mijkl + <p( I, l)mijkl

2 ( 0) (30c)
Hijkl - <p(2,0)mijkl + <p(2,1)mijkl

Furthermore, the number of irreducible invariants ( 19a,b) is reduced

from ten to five,

1,2,3, ~I _ tr ~~· ~ 3 _ tr Ao 2 ( 31)

if we regard the vector ~ as a unit vector.

The fourth-order tensor-valued functions (29a,b,c) resultung from
the plastic potential theory, can be simplified into:

0 [ ( 0) (32a)
hijkl - <aF/as 1)mijkl + (aF/a~ 1 )mijkl]dA

I [ ( 0) (32b)
hijkl - 2 (aF;as 2 )mijkl + <aF/a~ 3 )mijkl]dA

2 [ ( 0) 02c)
hijkl - 3 (aF/as 3 )mijkl + 0 ]dA .

Comparing the fourth-rank tensors (29a,b,c) with the corresponding ten-

sors (25a,b,c) or, alternatively, (32a,b,c) with (30a,b,c), one can see
that the scalar functions cp( 2 , I) and cp( 2 , 2 ) in (25c) or cp( 2 , I) in (30c)
cannot be expressed through the plastic potential (20), i.e., in the
anisotropic case the plastic potential theory with its classical flow
rule (4) only furnishes restricted forms of constitutive equations even
where a general plastic potential F = F(o .. ,A .. ) is assumed. Further-
lJ lJ
more, if one considers the functions (29a,b,c) or, alternatively (32a,b,
288 ]. Betten

[2]
c) one can see, that any terms containing mijkl and mijkl do not appear
1n (29c) or in (32c), i.e., the "second-order effect" in (28), characte-
rized through (29c) or (32c), is not influenced by the anisotropy (26)
of the material. Consequently, the flow rule (4) of the classical flow
theory of plasticity should be modified in the anisotropic case.
The results mentioned above can also be applied to perforated ma-
terials or damaged materials, if the anisotropy tensor A.. in (26) is
lJ
substituted with a perforation tensor [9] or a darnage tensor [8, 10].

4. MODIFICATION OF THE CLASSICAL FLOW RULE

By taking the representation theory into account we can justify the

following expansion of the flow rule (4):

dPE .. = cti\( 'dFI'do .. + am. "kl'dFI'dAkl) (33)

lJ lJ lJ

Assuming that the plastic potential F = F(o .. ,A .. ) is a scalar-valued

lJ 1J
function of the integrity basis ( 19a,b) and by using the modified flow
rule (33), we can find a constitutive equation which is compatible with
the represent at ion theory o f tensor funct ions ( 17) , ( 24) , ( 25a, b, c) tha t
is: all nine scalar-valued functions ~(\,v)' \,v = 0, 1,2, 1n ( 17) or
(25a,b,c) can be expressed through the plastic potential F:

( 34a)

( 34b)

( 34c)

~(I ,O) - 2di\ 'dFI'dS 2 , (34d,e)

~( 1, 2 ) - d [\( 'dF I 'd\24 + a'dF I 'd\22 ) , ( 34 f)

Tensor Function Theory and Classical Plastic Potential 289

a: 3F
- 3d/\ 3F
cp(2, I) - 7 d/1 "'dTIJ - a: d/\ ~~4 ' (34g,h,i)
dS3
where the abbreviations

(3Sa,b,c)

have been used. In this way the second-order effect 1n (28) is in-
fluenced through the material tensor ~ because of a: f 0 found in (34h,
i). This influence is not at all possible, if we use the tensor-valued
function (29c) which results from the classical flow rule (4) which does
not contain the parameter a:. Thus, the plastic potential theory is com-
patible with the tensor function theory, if the modified flow rule (33)
is used instead of (4). But among the scalar functions (34a, ... ,i) we
can find additional relationships ("conditions of compatibility") by
eliminating the plastic potential F. This can be done by closely adhe-
ring to the following procedure. First, one must find the partial deri-
vatives 3F/3S 1, ... ,3F/3~ 4 of the plastic potential F from the system of
linear equations (34a, ... ,i). Second, one must apply the {~0)equalities

3 2F 32F 32F 32F 32F 32F

(36)
3s 13s 2 as 2as 1 , ... , as 3ar2 4 a~
4 3
as , ... , 3 4
3~ 3r2 3r2
4 3~ 3

to every pair of the obove mentioned ten derivatives. Thus, in this way
we have eliminated the plastic potential F and found 45 conditions of
compatibility.
If the vector v 1s regarded as a unit vector, then it is possible-
to find the following identities by first taking (30a,b,c), (31) into
consideration and by then using the modified flow rule (33):

cp(O,O) - ( 3 F /3 S I ) dJ\ cp(O, I) - (aF/3~ l)d/\ (37a,b)

'

cp(I,O) - 2(3F/3S 2 )d/\ ' cp ( I , I) - 4cp ( 2, I) /a: + a: cp ( 0, I ) ( 37c ,d)

cp(2,0) - 3(3F/3 s 3 )d/\ cp(2, I) - ~(3F/3~ 3 )d/\ (37e,f)

'
290 ]. Betten

The results of (37a, ... ,f) can be used to find the partial derivatives
of the plastic potential:

ap;as 1 - <p(0,0) 1 •
ap;as 2 - <p< 1,o)L/2, aF/as 3 - <p(2,0)L/3 '
(38)
aF;an 1 - <p(O, I)L' aF;an 3 - 2<p( 2 ,I)L/a. f
In (38)

(39)

is a scalar-valued function of the integrity basis (31). The reader may

note the analogy between (39) and ( 12). The plastic potential F 1n ( 38)
can be eliminated in the following way. By regarding every pair of the
derivatives (38), and by using the {;) equalities

a2F a2F a2F a2F a2F a2F

as 1as2 as2as 1 ' ... ' as 2an 3 an 3as2 ' ... ' an l;:m3 an 3an 1 '
(40)

the following ten linear equations 1n the partial derivatives are

yielded:
Tensor Function Theory and Classical Plastic Potential 291

cp( I ,0) 1 • I - 2cp(0,0) 1 '2 + 0 + 0 + 0 R 11

cp(2,0) 1 • I + 0 - 3 cp(0,0) 1 '3 + 0 + 0 R21

cp(0,1) 1 '1 + 0 + 0 - cp(0,0) 1 '4 + 0 R31

a
cp(2, 1) 1 • I + 0 + 0 + 0 - Jil(0,0) 1 '5
R4 1

0 + 2 cp(2,0) 1 '2 - 3 cp ( I , 0) 1 ' 3 + 0 + 0 R51

0 + 2 cp ( 0 , I ) 1 ' 2 + 0 - cp( I ,0) 1 '4 + 0 R61

0:
0 + 2 cp(2, 1) 1 '2 + 0 + 0
- Jil( I ,0) 1 '5
R71

0 + 0 + 3 cp(O, 1) 1 '3 - cp(2,0) 1 '4 + 0 R81

a
0 + 0 + 3 cp ( 2, I) 1 ' 3 + 0 - Jil(2 ,0) 1 '5
R91

a
0 + 0 + 0 + cp ( 2 ' I ) T~ ' 4 - 2'-Pco, I) 1 •s Rl01

( 4 I)

where the abbreviations

(42)

and
292 J. Betten

Rl - 2acp<o,o/as 2 acp(l,o/as 1, R2 = 3acp(o,o/as 3 acp(2,o/asl,

R3 - acp<o,o/a~ 1 acp<o. l);as 1, R4 = %acp<o,o/a~3 acp(2,1/asl,

Rs = 3acp< 1,o/as 3 2acp< 2 , 0 /as 2 , R6 - acp< 1,o/a~ 1 2acp(O,I/as 2 ,

R7 = %acp< 1,o/a~3 - 2acp< 2 • 1/as 2 , R8 :: acp(2,o/a~l - 33cp<o, 1/as3'

Rg = %acp(2,o)/a~3 - 33cp(2, l);as3, RIO = %acp<o. l/a~3 - acp(2, l);a~l

(43)

are introduced.
From the equations found in (41) one takes (~) systems each of
which consists of 3 linear equations. In other words one starts from
(38) and combines (~) identities. Then, by using the (i} equalities
(40), one can find 10 systems each with 3 linear equations, for in-
stance:

(44)

In another example we could find:

Tensor Function Theory and Classical Plastic Potential 293

F,2 - cp(l,0)1/2 2cp(O, 1) 1 '2 - cp( I ,0) 1 '4 + 0 R61

a
=> 2cp ( 2 , I ) 1 ' 2 + 0 -zcr(I,0) 1 •5 R71
F,4 - cp(O,I) 1

a
F,5 - 2cp( 2 , !)1/a 0 + cp(2, 1) 1 '4 -zcr<o. 1) 1 '5 Rl01.

(45)

Note, that the coefficient determinants of the systems (44), (45), and
all other systems, which have been found in the same way, are equal to
zero:

0 2cp ( 0, I ) -cp(I,O) 0
cp(2 ,0) - 3cp(O,O)

a a
0 2cp(2,1) 0 - ~(I ,0) 0.
cp(2, I) - Jil(O,O)

a a
0 3cp(2, I) - Ji>(2 ,0) 0 cp(2, I) -~(0,1)

(46)

Thus, if one requires that all systems be homogeneaus (R I = R2 = ...

RIO = 0), then 10 conditions of "compatibility" can be derived:

acp(O,O) acp < I 0) acp<o o) acp(2,o) acp{O,I2 acp{2,I2

2 3 as 1 , ... ' a
as 2 as 1 ' as 3 a~3 a~,

(4 7)

These conditions are fulfiled, for instance, if the following as-

sumptions are made:
294 J. Betten

Then the plastic potential is as follows:

F (49)

where g 1, ... ,g 5 are arbitrary functions of only one invariant. The rea-
der may note the analogy between (49) and ( 10).
Note, that the conditions (47) are only sufficient, but not neces-
sary. In order to find sufficient and necessary "conditions of compati-
bility" CRAMER's rule should be used. After applying this rule on (44)
and (45) and after taking (46) into account, one can find the following
sufficient and necessary "conditions of compatibility":

0 (50)

and

( 5 I)

Continuing on in this way, the following complete set of ten sufficient

and necessary "conditions of compatibility" is obtained:

R Icp(2 ,0) -R2cp(l,O) + RScp(O,O) 0

Rlcp(O, I) - R3cp( I ,0) + R6cp(O,O) 0

Rlcp(2, I) - R4cp( I ,0) + R7cp(O,O) 0

R2cp(O, I) - R3cp(2,0) + R8cp(O,O) 0

R2cp ( 2, I) - R4cp(2,0) + R9cp(O,O) 0

(52)
R3cp ( 2, I) - R4cp(O, I) + R!Ocp(O,C) 0

RScp(O,l) - R6cp(2,0) + R8cp( I ,0) 0

R5cp ( 2, I) - R7cp(2,0) + R9cp( I ,0) 0

R6cp(2, I) - R7cp ( 0, I) +R!Ocp(l,O) 0

R8cp(2, I) - R9cp ( 0, I) + R!Ocp(2,0) 0

In the isotropic special case the ten conditions (52) are reduced to the
first one of the ten equations in (52), which is identical to ( 16).
Tensor Function Theory and Classical Plastic Potential 295

The results (37a, ... ,f) and (52) point out, that for oriented so-
lids the plastic potential theory can be made compatible with the repre-
sentation theory of tensor functions, if the classical flow rule (4) is
modified according to (33) and if, in addition, the "conditions of com-
patibility" in (52) have been fulfiled.
Finally, the special case of incompressibility should be consi-
dered. In such instances it is more practical to use the stress deviator
~· in the integrity basis ( 19a,b) than to use the stress tensor Q·
Furthermore, the modified flow rule

(53)

should be used instead of (33), where the fourth-order tensor

(54)

~s deviatoric with respect to the free indices {ij).

5. ANISOTROPY EXPRESSED THROUGH A FOURTH-RANK TENSOR

A more general case than that in ( 17) is described by using a mate-

rial tensor of rank four with the usual conditions of symmetry:

(55)

Then the constitutive equation ~s a symmetric tensor-valued function

f. .(o ,A ) (56)
~J pq pqrs

which has to fulfill the condition of form-invariance

a.~ ka J. 1 fk 1 ( opq , Apqrs ) - f.~J.(a pt a qu otu' a pt a qu a rv a swAtuvw ) (57)

296 J. Betten

under the transformation of the orthogonal group (a.ka.k = 6 .. ). Thus, a

J lJ 1
suitable representation of the constitutive equation (56) is given
through a linear combination

= G[;\,w,vl (58)
'A,w,v cro,w,v) ij

with [;\,w,v] several symmetric tensor generators

(59)

formed by matrix products of the following types

( 60)

which fulfill the condition (57) of form-invariance.

In finding an irreducible set of tensor generators note, that, ac-
cording to the HAMILTON-CAYLEY theorem, none of the exponents A,V of the
second-order tensor o needs tobe larger than 2. In generalizing the HA-
MILTON-CAYLEY theorem one can find in [I I] that a symmetric fourth-order
tensor (55) of power six and all higher powers can be expressed in terms
of lower powers, that is w~ 5. For w 0 the tensor A in (60) is a
spherical one. Then, on the one hand, the combinations [;\,O,v] are sym-
metric with respect to A,V, and on the other hand the restriction
;\ + V ~ 2 follows. Thus, 48 irreducible matrix products of the forms
(60) can be found, which are listed in Table I.
In formulating the plastic potential the integrity basis is used,
which can be derived by forming al' irreducible traces of the matrix
products (60):

M[;\,w,v] _ where A, V = 0, I, 2 for W = I, 2, ... , 5

tt ( 6 I)
but for w = 0 = > ;\ + v ~ 3.

These invariants are listed 1n Table 2.

Tensor Function Theory and Classical Plastic Potential 297

Table I Matrix Products of the Fonns (60) for Isotropy (1-l 0) and
Anisotropy <1-1 ~ 0)

11 =0 '. M.IJ1.x.o.vl with A. + v ;lS Z ll =1, Z, .. . ,5; M.l.A.!l . VI

r- IJ

~ 0 z ~ 0 z
~~------~--------+---------~
d.m A_t.l-ll 1!-ll A.l.l-ll diZI
0 d..
IJ IJ
0 IJ r r Aijpq dPQ IJPQ PQ
II-II II-II 11-ll 12)
d.k
1 Ak )rr
. dik Akjpq dPQ dik Akjpq dPQ

diZI A l1-1l
ik kjrr
diZI A11-ll d dm A 1 ~ 1 d 121
1k kJPQ PQ 1k kJPQ PQ

Table 2 System (61) of irreducible Invariants

lA [A..IJ.. V] = d(A.) A(IJ.) d(v) 12 5

tt
1' 1 pq pqr s rs ; fl = ' ' ·· ·'

0 z
(IJ.)
0 A..
IIJJ

d A(IJ.l o' (21

pq pqrs rs
-

z symmetric

One can see, that the combinations [A,~, v ] are symmetric with
respect to A,V for all ~. i.e. in eqn (61) or in Table 2 we see 33 In-
.
var1ants . I n ad d.1t1on,
. t h e 1nvar1ant
. . A(. 6. ).. must be cons1dered.
· Thus a
llJJ
set of 34 irreducible invariants is found.
298 J. Betten

The plastic potential 1s given 1n the form

F ( 62)

and by using the classical flow rule (4) one obtains the constitutive
equation

d 1\ L: aF [Q[A] A (f.J) (v) + Q[v] A (f.J) (A) J ( 63)

A,)J,V
aM[A f.l v] ijpq pqrs 0 rs ijpq pqrs 0 rs
tt
where the A several fourth-rank tensors g[A] defined as (22) are deriva-
tives of tensor powers. One sees that only 28 invariants in the set (61)
are stress-dependent. Therefore, the constitutive equation (63) contains
only 28 terms, contrary to the tensor polynomial function (58) with 48
terms, that is: the classical flow rule (4) only furnishes restricted
forms of the constitutive equations, even if a general plastic potential
has been used.
The constitutive equations (17) or (21) are special cases and can
be dealt with by inserting the special fourth-rank tensor

(64)

having the tensor power

A ( JJ)= ( A ( IJ )A ( f1) A ( \J )A ( \J )) / 2 ( 65)

ijkl - ik j 1 + il jk

1n equations (58) or (63), respectively.

Finally, further material influences, which are of practical impor-
tance, should be considered, e.g. the effect of darnage Cp) on materials
with inborn anisotropy (~). Then, we have to consider a symmetric second
order tensor-valued function

f. .(o ,D ,A ) (66)
lJ pq pq pqrs

in three argument-tensor ~' Q, and A of rank 2, 2, and 4, respectively.

Tensor Function Theory and Classical Plastic Potential 299

Different representations of (66) are discussed in [8, 10].

REFERENCES

I. BETTEN, J., Elastizitäts- und Plastizitätslehre, Vieweg-Verlag,

Braunschweig/Wiesbaden 1985.
2. BETTEN, J., The Classical Plastic Potential Theory in Camparisan
with the Tensor Function Theory, International Symposium PLASTICITY
TODAY on Current Trends and Results in Plasticity, Udine 1983,
published in a special Olszak Memorial volume of the Engineering
Fracture Mechanics 21 (1985): 641-652.
3. MURAKAMI, S. and SAWCZUK, A., On Description of rate-independent Be·
haviour for prestrained Solids, Archives of Mechanics 31 ( 1979):
251-264.
4. LEHMANN, Th., Einige Bemerkungen zu einer allgemeinen Klasse von
Stoffgesetzen für große elasta-plastische Formänderungen, Ingenieur-
Archiv 41 ( 1972): 297-310.
5. BETTEN, J., On the Representation of the Plastic Potential of Aniso-
tropie Solids, Colloque International du CNRS n° 319, Comportement
plastique des solides anisotropes, Grenoble 1981, published in the
proceedings: Plastic Behavior of Anisotropie Solids (ed. J.P. Boeh-
ler) CNRS, Paris 1985, 213-228.
6. BOEHLER, J.P. and SAWCZUK, A., Application of Representation Theo-
rems to describe Yielding of transversely isotropic Solids, Mech.
Res. Comm. 3 ( 1976): 277-283.
7. BOEHLER, R.P. and SAWCZUK, A., On Yielding of Griented Solids, Acta
Mechanica 27 ( 1977): 185-206.
8. BETTEN, J., Net-Stress Analysis in Creep Mechanics, Ingenieur-Archiv
52 ( 1982): 405-419.
9. LITEWKA, A. and SAWCZUK, A., A Yield Criterion for Perforated
Sheets, Ingenieur-Archiv 50 ( 1981): 393-400.
10. BETTEN, J., Darnage Tensors in Continuum Mechanics, Euromech Col-
loquium 147 on "Damage Mechanics", Paris VI, Cachan 1981, published
in Journal de Mecanique theorique et appliquee 2 ( 1983): 13-32.
II. BETTEN, J., Integrity Basis for a Second-Order and a Fourth-Order
Tensor, International J. Math. & Math. Sei. 5 ( 1982): 87-96.