Tensor Analysis and Continuum Mechanics

Tensor Analysis and

Continuum Mechanics

by

Yves R. Talpaert
Faculties of Science and Schools of Engineering at
Algiers University, Algeria;
Brussels University, Belgium;
Bujumbura University, Burundi;
Libreville University, Gabon;
Lome University, Togo;
Lubumbashi University, Zaire and
Ouagadougou University, Burkina Faso

Springer-Science+Business Media, B.Y.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-90-481-6190-4 ISBN 978-94-015-9988-7 (eBook)

DOI 10.1007/978-94-015-9988-7

Printed on acid-free paper

All Rights Reserved

© 2002 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 2002.
Softcover reprint of the hardcover 1st edition 2002
No part of this work may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work.
CONTENTS

PREFACE ........................................................... xv

Chapter 1. TENSORS 1

1. FIRST STEPS WITH TENSORS ......................... .

1.1 Multilinear forms .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I

Linear mapping .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Multilinear form ........................................... 2
1.2 Dual space, vectors and covectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Dual space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 3
Expression of a covector .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Einstein summation convention. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Change of basis and cobasis '" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Tensors and tensor product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. to
Tensor product of multilinear forms. . . . . . . . . . . . . . . . . . . . . . . . . . to
Tensor of type (?) ......................................... II
Tensor of type (~) ......................................... 12
Tensor of type (g) ......................................... 14
Tensor of type m......................................... 16
Tensor of type (:) ......................................... 18
Tensor of type (j,) ........................................ 20
Symmetric and antisymmetric tensors 22

2. OPERATIONS ON TENSORS ........ . . . . . . . . . . . .. .. . . . . . 25

2.1 Tensor algebra ........................................... 25
Addition of tensors ....................................... , 25
Multiplication of a tensor by a scalar ..... . . . . . . . . . . . . . . . . . . . .. 25
Tensor multiplication ...................................... 26
2.2 Contraction and tensor criteria ..... . . . . . . . . . . . . . . . . . . . . . . . .. 27
Contraction ............................................... 27
Tensor criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 31

vii
viii Contents

3. EUCLIDEAN VECTOR SPACE ............................ 33

3.1 Pre-Euclidean vector space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 33
Scalar multiplication and pre-Euclidean space ................... 33
Fundamental tensor .. , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 34
3.2 Canonical isomorphism and conjugate tensor . . . . . . . . . . . . . . . .. 35
Canonical isomorphism ..................................... 35
Conjugate tensor and reciprocal basis .... . . . . . . . . . . . . . . . . . . . . .. 37
Covariant and contravariant representations of vectors ............ 40
Representation of tensors of order 2 and contracted products ....... 42
3.3 Euclidean vector spaces ............................... . . . . . 45

4. EXTERIOR ALGEBRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 49
4.1 p-forms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 49
Definition of a p-form ...................................... 49
Exterior product of I-forms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 50
Expression of a p-form ..................................... 52
Exterior product of p-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 56
Exterior algebra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 57
4.2 q-vectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59

5. POINT SPACES ......................................... 63

5.1 Point space and natural frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 63
Point space ............................................... 63
Coordinate system and frame of reference ...................... 64
Natural frame ............................................. 66
5.2 Tensor fields and metric element .. . . . . . . . . . . . . . . . . . . . . . . . . .. 69
Transformations of curvilinear coordinates ..................... 69
Tensor fields .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 73
Metric element .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 74
5.3 Christoffel symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 75
Definition of Christoffel symbols. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 75
Ricci identities and Christoffel formulae ........................ 78
5.4 Absolute differential, Covariant derivative, Geodesic .. . . . . . . . . 80
Absolute differential of a vector, covariant derivatives ............ 80
Absolute differential of a tensor, covariant derivatives . . . . . . . . . . .. 83
Geodesic and Euler's equations ..................... . . . . . . . . 85
Absolute derivative of a vector (along a curve) .................. 86
5.5 Volume form and adjoint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Volume form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Adjoint. .. .. . .. .... . .. .. . . ... . . .. .. . .. . ... . .. ..... .. .. . 91
5.6 Differential operators .................................... 92
Gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Divergence .............................................. 99
Curl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 101
 Contents ix

Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 103

EXERCISES .................................................. 106

Chapter 2 LAGRANGIAN AND EULERIAN DESCRIPTIONS 147

1. LAGRANGIAN DESCRIPTION .......................... 147

1.1 Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 147
1.2 Deformation and Lagrangian Description ................... 148
1.3 Flow and hypotheses of continuity .... . . . . . . . . . . . . . . . . . . . . . 152
1.4 Trajectories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 153
1.5 Streakline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 154
1.6 Velocity and acceleration of a particle . . . . . . . . . . . . . . . . . . . . . .. 155
1.7 Abstract configuration .................................... 156

2. EULERIAN DESCRIPTION ... . . . . . . . . . . . . . . . . . . . . . . . . . .. 157

2.1 Definition; Comparison between L- and E-descriptions ..... . .. 157
2.2 Trajectory and velocity ................... . . . . . . . . . . . . . . .. 158
2.3 Streamline .............................................. 161
2.4 Steady motion ........................................... 163

EXERCISES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 165

Chapter 3 DEFORMATIONS 171

1. HOMOGENEOUS TRANSFORMATION .......... . . . . . . .. 172

1.1 Definition of homogeneous transformations ...... . . . . . . . . . .. 172
1.2 Convective transport .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
Convective transport of a vector ............................ 174
Convective transport of a volume . . . . . . . . . . . . . . . . . . . . . . . . . . .. 174
Simple shear .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 176
1.3 Cauchy-Green deformation tensor and stretch ............... 177
(Right) Cauchy-Green deformation tensor. . . . . . . . . . . . . . . . . . . . 177
Stretch ................................................. 180
Shear angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 181
x Contents

Principal stretches ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 182

1.4 Finite strain tensor .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 184
1.5 Polar decomposition ..................................... 186
Pure stretch and rotation .................................. 186
Euler-Almansi strain tensor .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 190
1.6 Rigid body transformation ................................ 191

2. TANGENTIAL HOMOGENEOUS TRANSFORMATION. . . . 193

2.1 Deformation gradient .................................... 193
2.2 Homogeneous transformations of elements .................. 196
Transport of vectors, volume deformation, and area deformation ... 196
Stretches ........ ....................................... 199
Strain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 201
2.3 Displacement and gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 203
Material displacement gradient .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 204
Spatial displacement gradient .... . . . . . . . . . . . . . . . . . . . . . . . . . .. 206
Curvilinear coordinate system .............................. 208

3. INFINITESIMAL TRANSFORMATION . . . . . . . . . . . . . . . . . .. 2 I 0
3.1 Tensor notions relating to infinitesimal transformations ........ 2 I I
3.2 Compatibility conditions .................................. 216
3.3 Rigid body transformation ....... ......................... 220

EXERCISES ................................................... 222

Chapter 4 KINEMATICS OF CONTINUA 263

1. LAGRANGIAN KINEMATICS 263

1.1 Homogeneous transformation motion ...................... 264
1.2 General motion and gradient 266

2. EULERIAN KINEMATICS ............................... 268

2.1 Homogeneous transformation motion ....................... 268
Velocity field ............................................. 268
Material derivative of a vector .... . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Material derivative of a volume .............................. 269
Eulerian rates ......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 271
2.2 General motion and velocity gradient . . . . . . . . . . . . . . . . . . . . . .. 274
Velocity gradient tensor and Eulerian rates ......... . . . . . . . . . .. 274
 Contents xi

Lagrangian and Eulerian strain tensors ... . . . . . . . . . . . . . . . . . . . .. 279

Rate of rotation .......................................... 280
Decomposition of motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 285
2.3 Rigid body motion ....................................... 286

3. MATERIAL DERIVATIVES OF CmCULATION, FLUX, AND

VOLUME .............................................. 287
3.1 About the particle derivative ... . . . . . . . . . . . . . . . . . . . . . . . . . .. 287
Physical quantity ......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Vector field ....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 290
Tensor field ...... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 292
3.2 Material derivative of circulation ........................... 293
3.3 Material derivative of flux .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 296

3.4 Material derivative of volume integral ..... . . . . . . . . . . . . . . . .. 299

Lagrangian and Eulerian approaches ......... . . . . . . . . . . . . . . . .. 299
Proper motion case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 303

EXERCISES .................................................. 305

Chapter 5 FUNDAMENTAL LAWS; PRINCIPLE OF VIRTUAL WORK 315

1. CONSERVATION OF MASS AND CONTINUITY EQUATION 315

1.1 Axiom of mass conservation .............................. 315

1.2 Continuity equation ...................................... 316
Continuity equation in the Lagrangian description ........ . . . . . .. 316
Continuity equation in the Eulerian description ..... . . . . . . . . . . .. 317
Mass flow rate ........................................... 319
1.3 Material derivative of integral of mass density . . . . . . . . . . . . . .. 320
1.4 Isochoric motion, steady and irrotational flows .............. 322
Isochoric motion ..... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 322
Steady flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 326
Steady isochoric flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 327
Irrotational flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 328
Isochoric irrotational flow ....... . . . . . . . . . . . . . . . . . . . . . . . . . .. 329

2. FUNDAMENTAL LAWS OF DYNAMICS ................. 330

2.1 Body forces and surface forces .. . . . . . . . . . . . . . . . . . . . . . . . . . . 330
2.2 Principles of linear momentum and moment of momentum. . . .. 333
2.3 Cauchy's stress tensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 337
xii Contents

2.4 Cauchy's stress tensor and principles of dynamics ............ 342

Linear momentum principle and equilibrium equations .. . . . . . . . .. 342
Moment of momentum principle ............................ 344
The generalized Cauchy's theorem. . . . . . . . . . . . . . . . . . . . . . . . .. 346
Poisson's theorem ...... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 347

3. THEOREM OF KINETIC ENERGY ....................... 348

3.1 Theorem of kinetic energy in the Eulerian description ......... 348
3.2 Theorem of kinetic energy in the Lagrangian description 350

4. STUDY OF STRESSES.................................. 356

4.1 Reciprocity of stresses .... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 356
4.2 Principal stresses ......................................... 358
4.3 Stress invariants; deviator .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 364
4.4 Stress quadric of Cauchy and Lame stress ellipsoid 370

4.5 Geometrical constructions and Mohr's circles ................ 374

(Mohr's) stress plane ...................................... 374
Stress vector and plane of Mohr . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 375
Description of Mohr's circles ............................... 378
Particular stresses ......................................... 380

5. PRINCIPLE OF VIRTUAL WORK ........................ 384

5.1 Preliminary recalls ....................................... 384
5.2 Rigid body motion ....................................... 386
5.3 Expressions of virtual power (and virtual work) .............. 389
5.4 Principle of virtual work ................................. , 391

6. THERMOMECHANICS AND BALANCE EQUATIONS .... 393

6.1 Balance equation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Proper motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 393
Material domain .......................................... 398
Fixed domain ............................................ 401
6.2 First principle of thermodynamics .......................... 402
Principle ....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 402
Balance equations and local forms . . . . . . . . . . . . . . . . . . . . . . . . . .. 404
Potential energy of body forces . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 406
Internal energy and balance equation ......................... 407
6.3 Second principle of thermodynamics ........... . . . . . . . . . . .. 409
Principle ............................................... 409
Clausius-Duhem inequality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 411
Dissipation and reversibility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 412
 Contents xiii

6.4 Conclusion and constitutive equations ....................... 414

EXERCISES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 417

Chapter 6 LINEAR ELASTICITY 455

1. ELASTICITY AND TESTS ............................... 455

2. GENERALIZED HOOKE'S LAW IN LINEAR ELASTICITY . 458

2.1 Generalized Hooke's law ............ . . . . . . . . . . . . . . . . . . . .. 458
2.2 Quadratic forms and strain energy function .................. 459
2.3 Isotropic material and Lame coefficients ..................... 463
Constitutive equations ...................................... 463
Young's modulus and Poisson's ratio ......................... 466
Bulk modulus ........................................... 469
Shear modulus ........................................... 471
Hooke's law's expression in a general coordinate system .......... 472
Navier's equations of motion ................................ 475

3. EQUATIONS AND PRINCIPLES IN ELASTOSTATICS ..... 477

3.1 Navier's equation; the Beltrami equations of compatibility. . . . 478
3.2 Principle ofsuperposition .............. . . . . . . . . . . . . . . . . .. 481
3.3 Saint-Venant's principle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 484

4. CLASSICAL PROBLEMS .......... .. . . .. . .. . .. .. . .. . . ... 485

4.1 Plane problems .......................................... 485
Plane stress problems ................ . . . . . . . . . . . . . . . . . . . . .. 485
Plane strain problems ...................................... 488
4.2 Classical problems in elastostatics .......................... 491
Uniaxial stresses .......................................... 491
Torsion of a circular cylinder body ............................ 493
Torsion of cylindrical shafts ................................. 496

EXERCISES ....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 501

SUMMARY OF FORMULAE ............................................. 541

xiv Contents

BIBLIOGRAPHY ......................................................... 575

GLOSSARY OF SYMBOLS ............................................... 577

INDEX .................................................................... 585

PREFACE

This book is designed for students in engineering, physics and mathematics. The material
can be taught from the beginning of the third academic year. It could also be used for self-
study, given its pedagogical structure and the numerous solved problems which prepare
for modem physics and technology.

One of the original aspects of this work is the development together of the basic theory of
tensors and the foundations of continuum mechanics.

Why two books in one?

Firstly, Tensor Analysis provides a thorough introduction of intrinsic mathematical

entities, called tensors, which is essential for continuum mechanics. This way of
proceeding greatly unifies the various subjects. Only some basic knowledge of linear
algebra is necessary to start out on the topic of tensors. The essence of the mathematical
foundations is introduced in a practical way.

Tensor developments are often too abstract, since they are either aimed at algebraists only,
or too quickly applied to physicists and engineers. Here a good balance has been found
which allows these extremes to be brought closer together.
Though the exposition of tensor theory forms a subject in itself, it is viewed not only as an
autonomous mathematical discipline, but as a preparation for theories of physics and
engineering. More specifically, because this part of the work deals with tensors in general
coordinates and not solely in Cartesian coordinates, it will greatly help with many
different disciplines such as differential geometry, analytical mechanics, continuum
mechanics, special relativity, general relativity, cosmology, electromagnetism, quantum
mechanics, etc ..

Secondly, the foundations of Continuum Mechanics constitute the most important part of
this work. It involves chapters on the Lagrangian and Eulerian descriptions, deformations,
kinematics of continua, fundamental laws, the principle of virtual work, and linear
elasticity. These chapters lay the groundwork for other or more technical subjects as fluid
mechanics, strength of materials, plasticity and viscoelasticity, thermoelasticity and
thermodynamics, nonlinear continuum mechanics, finite element methods in continuum
mechanics, etc ..

xv
xvi Preface

The reader will quickly discover the importance of these chapters and, given their
controlled and logical progression as well as their role of introducing the above mentioned
disciplines, will eagerly take up the challenge.

Unlike other authors who denote tensors by (several) lines above and below bold letters, I
simply represent both general tensors and vectors by bold letters. This simplification is
possible since the reader gradually gains experience in dealing with tensors and because
tensors of orders higher than 2 are not frequently used in continuum mechanics. In
addition, the difference in level of operation dot '.' or '.' has no mathematical meaning, it
has only a pedagogical value, namely: bringing down dots specifies that the corresponding
operation results are real numbers. This stipulated simplification leads the reader to think
carefully about operations between tensors and about the types of different tensors. Of
course, scalars are not designated by bold letters.
According to usage the vertical brackets completely enclose the elements of matrices,
whereas they partly enclose the normal mathematical expressions.

Terms of the continuum mechanics terminology are sometimes translated into French for
French speaking readers.

The important propositions and the formulae to be framed are shown by W and~.

The summary of formulae and glossary of symbols should make the assimilation of
notions easier.

All the proofs and the 95 solved exercises are described in detail.

Acknowledgements. Many thanks are due to my former students who let me expound a
part of the material that resulted in this book.
I wish to express my gratitude to Kluwer Academic Publishers for their cooperation.
I would particularly appreciate it if readers would let me know of any errors or further
suggestions.

Any faculty of science or engineering school, interested in my analysis of continuum

mechanics and in the way differential geometry and theoretical mechanics have been
developed in my previous books, can contact me for a possible collaboration.

Yves R. Talpaert

[email protected]
CHAPTER 1

TENSORS

The idea of tensor took form at the end of the 19th century when it became
necessary to express pressure forces in continua. But the first important developments of
the notion of tensor date back to the very beginning of the 20th century; they are generally
owed to Ricci, Levi-Civita, E. Cartan, ....
The name tensor, introduced by the physicist Voigt, is reminiscent of tension in fluids,
elastic solids, .... This terminology designates intrinsic mathematical entities which are
suitable for the expression of the laws of mechanics regardless of the choice of coordinate
system.
The reader has already encountered tensors, since vectors and linear forms are examples of
them.
It is pointless insisting on the considerable importance that tensors have gained through
the developments of exact and applied sciences in the 20th century, more especially in
continuum mechanics, special relativity, general relativity, quantum mechanics,
differential geometry, Riemannian geometry, analytical mechanics, fluid dynamics,
cosmology, electromagnetism and so on I.

1. FIRST STEPS WITH TENSORS

1.1 MUL TILINEAR FORMS

Let E and F be finite-dimensional real vector spaces.

I This chapter. is based on our books: Mecanique Analytique vol.2 (1982), Differential Geometry with
Applications to Mechanics and Physics (2000), where tensors are considered in the manifold context, and
Mechanics, Tensors and Virtual Works (2002), where tensors are applied to theoretical mechanics.

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
2 Chapter 1

1.1.1 Linear Mapping

D A mapping
g: E ---+ F: x H g(x)
is linear if Vx,y E E, Vk E R:

g(x + y) = g(x) + g(y) , g(kx) = k g(x).

Let us denote by L( E; F) the set of (continuous) linear mappings of E to F.

D The addition in L(E; F) is the mapping

L(E;F)x L(E;F) ---+ L(E; F): (g,h) H g +h
such that the sum g + h is the linear mapping defined by
E ---+ F : x H (g + h)(x) = g(x) + h(x).

D The multiplication of a linear mapping g of E into F by a scalar k is the mapping

Rx L(E;F) ---+ L(E;F): (k,g) H kg
such that the product kg (of g by k) is the linear mapping defined by
E ---+ F : (k, g)(x) = kg(x).

We know that L(E; F) provided with the two previous laws of addition and multiplication
has the structure of a vector space.

1.1.2 Multilinear Form

In mechanics we particularize F by choosing this vector space to be R. So we will

consider the vector space L( E; R) later on.

D A linear form on E is a mapping

f : E ---+ R: x H f(x)
such that Vx,y E E, Vk E R:
f(x + y) = f(x) + f(y), f(kx) = kf(x).
A linear form on E is also called a one-jorm or covector.

Let E(I) , ... , E(p) be p vector spaces.

 Tensors 3

D A p-linear form defined on the Cartesian product of p spaces E(l) x ... x E(p) is a
mapping

which is linear with respect to each vector, that is,

\fx(l)'Y(l) EE(l), ... , \fx(p),Y(p) EE(p), \fkER:

f(x(l) + Y(l),X(2)""'x(p») = f(X(l),X(2)""'x(p») + f(Y(l),X(2)""'x(p»)

Provided with laws of addition and multiplication by a scalar defined as before, the space
L p (E; R) of p-linear forms on E has the structure of a vector space.

1.2 DUAL SPACE, VECTORS AND COVECTORS

1.2.1 Dual Space

D The vector space of linear forms defined on E is called the dual space of E.

It is denoted by E* .

So, the dual space is a vector space the elements of which, called covectors, are linear
functions E ~ R . It is a space of functions.

Example 1. The row vectors are covectors (or l-forms). With the multiplication of
matrices, a row vector (linearly) associates a real to each column vector.
For instance:

Example 2. In quantum mechanics the I-forms called bras and denoted (¢ I linearly
associate complexes (¢, 'If) to vectors called kets and denoted I'If) .

1.2.2 Expression of a Covector

Let E be a real vector space of dimension n.

4 Chapter 1

A covector on E is a linear mapping f: E ~ R which associates a real f(x) to each

vector xEE.
We denote by xl, ... ,xn the components ofx with respect to a basis (el, ... ,en ) of E.

The real f(x) is written:

f(x) = f(xle l + ... + xne n )

= xlf(e l ) + ... + xn f(e n ).
By letting

we have
n
f(x) = IJ; xi.
i=l

We mention that the image of x under f is sometimes called the value of the form (for x).

We are now going to express the covector f with respect to the dual basis.
The dual basis (e*l, ... ,e*n) of the basis (el, ... ,en ) is such that:

(1-1)

where 15~ is the Kronecker delta l and the n linear forms making up the dual basis are

e*i :E~R:xHe*i(x)=xi.

Thus \;:fx E E :
n
f(x) = IJ; e*i (x),
i=l

which leads to the expression of the covector

n
f = IJ; e*i. (1-2)
i=l

Remark. The reader will compare the previous expression with that of a vector:
n
x=Ixie i ·
i=l

I The Kronecker delta is the symbol

I if i = j,
15; = 15ij = 15ij = { 0 if
i *- j.
 Tensors 5

So according to usage the components of vectors show an upper index and the
components of covectors a lower index.

Notation. Generally we will represent the covectors (or linear forms) by Greek
characters, and since they are the elements of a vector space, namely E* , we have decided
to write them in bold characters.

1.2.3 Einstein Summation Convention

The Einstein summation convention consists in removing the summation sign ~,

more precisely:
Summation is implied when an index is repeated on upper and lower levels 1•

For example, we denote:

n n n
aUkxiyizk = I I I aljkx'ylzk
i=l i=l k=l

On the one hand, any repeated index of summation is called a dummy index because it
does not matter what the letter is; for instance:

b ijX i
x i=
brsbllb12
rsX X = llX X + 12X X +

In particular, the Kronecker symbol is such that

0/1 = 011 + ° + ...
22 = 1+ 1+ ...

Attention! In the summation convention we emphasize that an index of summation shall

never be repeated more than once.

1 The Einstein convention will also be used with indices at the same height in the frame of the usual
Euclidean space when considering orthononnal bases.
6 Chapter 1

So there is no question of writing:

aJbx j (never!),
, J

biiCjX Y i J
(never!).

In the same manner, given

Q = aijxiyj
with
yj =b(x',
if we want to express Q as a function of the coordinates Xi only, then there is no question
of writing:
(never!),

but we must firstly change the dummy index in yJ , for instance:

yj = b£x k
in order to obtain the following right expression:
Q -_ aij bkX
j i k
x .

In the same manner, if we want to multiply aix i and biyi we must first rename the
dummy index of one term of the product and write aib jX i yj .

On the other hand, there is another type of index. An index which appears once in each
expression is called afree index.
So, for instance, the equation
i = 1,2; j = 1,2,3,4

represents the following system of equations:

Y
I
= blk X k ,
y2 = b;x'
that is, explicitly:
l
Y I = bI XI + b2X2 + bl3 X3 + bl4X4 ,
l

y2 =b12XI +bix 2 +bix 3 +b1x4.

In this example the index i is free and the indexj is dummy.

Remark 1. The reader must take care to write the equations correctly. Any free introduced
index must appear in every term.
So the following equations are
 Tensors 7

Meaningless: meaningful:
(never!) Zj=Xj+Yj'

(never!) a pq -
-
bpC;.
' q

Remark 2. Factoring is possible with the Kronecker symbol.

For instance, given an orthonormal basis of the Euclidean space, by viewing n; = liijnj
(with summation over}) the equation

is written:

1.2.4 Change of Basis and Cobasis

Let (e 1 ,e2) be a basis of a 2-dimensional vector space called the unprimed basis
and (e;, e;) be another basis called the primed basis to make matters simpler.
The following change of basis

e; == a~e1 +a 12e 2

e; == a;e 1 + a~e2

is written in a condensed manner:

where a = (a~) and {3 = ({3,k ) are inverse matrices and thus a~{3; = Ii;.
The equalities

imply by comparison:
I I,] I /2
X ==alx +a 2x ,
2 2,] 2 /2
X ==alx +a 2 x

In the same manner the equalities

,] / /2 / I 2
X==X e l +x e 2 =X e l +X e 2
= X 1({3I'
lei + {32/)
I e2 + x
2({31' {32/)
2e l + 2 e 2
8 Chapter 1

imply by comparison

We generalize, and in an n-dimensional space E we say:

PRI The matrix associated with the expression of unprimed components as functions of
those primed is the transpose of the matrix associated with the expression of
primed basis vectors as functions of the unprimed.

Proof The matrices (al) and (p~) being inverse, we denote

(1-3)
(1-4)

The equations

imply
(1-5)

By comparing the following explicit expressions

I I,] 1,2
X =alx +a 2x + ... ,
and so on, the proposition is thus proved.

PR2 The matrix associated with the expression of primed components as a function of
the unprimed components is the transpose of the matrix associated with the
expression of unprimed basis vectors as functions of the primed.
It is the inverse and transpose of the matrix associated with the expression of
primed basis vectors as functions of the unprimed.

Proof Since

the equalities
x=x'Je }'. =x'e.1 =xiRJe'.
I-'I}

imply

x'} = Plx ' . (1-6)

By comparing the following explicit expressions
,] 1/1 I 1/1 2
X = 1-'1 X + 1-'2 X + ... ,
1/ 1 , 1/ 2 ,
el = 1-'1 e l + 1-'1 e 2 + ...
 Tensors 9

and so on, the proposition is thus proved.

Weare now going to show the formulae of the change of dual bases (a dual basis is also
called a cobasis).
From every

because e'*J(ek)=oj we deduce:

e'*J (x) = x'J = f3! Xi = f3( e*i (x)

which implies
e'*J =f3!e*i (1-7)

and

(1-8)

The reader will easily say the propositions which refer to (1-7) and (1-8).

Example. Primed axes being obtained from a 90° direct rotation! about the e 2 -axis,
express the vector 4e 3 with respect to the primed basis ( e;).

The 'primed' basis vectors e~ = a~el are explicitly

and the 'unprimed' vectors e i = f3!e~ are explicitly:

[e
l
] [ 0 0 1][e; 1
:: = ~ 1 ~ ~ :~ .

From PR2 we deduce:

1 A direct rotation is also called a counterclockwise rotation.

10 Chapter 1

The given vector (such that x3 = 4) is written - 4ei since xrl = -4, x,2 = X,3 = o.
This result is obvious because e3 = -ei.

1.3 TENSORS AND TENSOR PRODUCT

Let E(I)' ... ,E(p), ... , E(p+q) (or simply E) be finite-dimensional vector spaces.

1.3.1 Tensor Product of Multilinear Forms

Let / be a p-linear form defined by

E(I)x ... x E(p) ~ R: (x(l)' ... 'x(P» H f(x(I)' ... 'x(P»'

let h be a q-linear form defined by

D r:F The tensor product of a p-linear form/ and a q-linear form h is the (p+q)-linear
form denoted / ® h :

E(I) x ... x E(p+q) ~ R : (x(l) ,... , x(p+q» H / ® h (x(l) ,... , x(p+q»

such that

Example 1. If we consider two linear forms/and h defined on E, namely:

/:E ~ R:XH lex),
h :E ~ R : y H hey),
then the tensor product of linear forms/and h is the bilinear form
/®h:ExE ~R:(X,Y)H /®h(x,y) =/(x) hey).

Given a basis (e;) of E, the linear forms/and h have the respective values:

lex) = J; Xi = J; e Oi (X),
hey) = hj yj = hj e Oj (y)

and the corresponding value of the tensor product is

/ ® h(x,y) = lex) hey) = J; hj Xi yj .
 Tensors \1

Example 2. The tensor product of two bilinear forms

f : E(l)x E(2) ~ R: (X(l)'X(2)) H f(X(I),X(2)) '

h: E(3) X E(4) ~ R : (X(3),X(4)) H h(X(3)'X(4))

is the following quadrilinear form

1.3.2 Tensor of Type (~)

D A tensor of type (~) or covector is a linear form defined on E,

It is an element of the vector space E' ,

According to usage the covectors are generally denoted by Greek letters; for instance:
wEE',

Thus the definition of a covector or tensor of type (~) is expressed as follows:

w: E ~ R:XHW(X)

with \::fa,b E R, \::fx,y E E:

w(ax + by) = aw(x) + bw(y) ,

The co vector expressed as (1-2) is written:

where

The image of any vector x under w is the real

w(x) = OJie*i (xie i) = OJixie*i (e i)

this value being also denoted by

(w,X) = w(X) , (1-10)

12 Chapter 1

Change of basis

We recall that a linear form m behaves towards any vector x in the following way:
w(x) = w(x' e,) = w(x'j <)
(1-11)

where w: = w(e~).
This obvious requirement allows testing of the 'tensor character'. Let us use it in order to
obtain the formulae of transformation of components of w .
By recalling (1-5) and (1-6):
i
X = akx
i,k
, x
,P
=
fJP n
n X ,

yi = a! y", yfS = fJ~ym,

the condition (1-11) that is

implies
(1-12)
and conversely
(1-13)

1.3.3 Tensor of Type (~)

D A tensor of type (~) or vector is a linear form defined on E* .

So, the definition of a tensor of type (~) or vector x is expressed as follows:

x: E' ~ R: w H x(w)

such that Va,b E R, Vw,)I E E' :

x( am + b)l) = ax(w) + bxC)l) .

A linear form x defining a tensor of type (~) is obviously written with respect to a basis
(e i ) of E as follows:
 Tensors 13

where

So the image of any covector OJ under the linear form x is the real
f.) =xe
x~w
i
i ( OJje OJ) =XOJjUi
i s:j i
=XOJ i
which is written:
x(w) = (x,w).

In conclusion, we have obtained the following equality

(x,w) = (w,x) (1-14)

and this important result expresses the duality between covectors and vectors.

Remark 1. In fact, we may identify E" = L(E'; R) with E:

E" = E.
By referring to the formulae (1-3) and (1-4) for change of basis in E on the one hand, and
to their corresponding (1-7) and (1-8) in E' on the other hand, we immediately see that
the relevant vectors e;o of EO' and e i of E are transformed according to the same rule.

To each vector expressed with respect to a basis (e i ) of E there corresponds a vector with
the same components with respect to the corresponding basis (e;O) of E" and conversely,
such that to the sum of any two vectors of E corresponds the sum of two corresponding
vectors of E", to the product of a vector of E by a scalar corresponds the product of the
corresponding element of E" by this scalar.
Since there is no reason to distinguish the elements of E" from those of E, we have the
right to identify these vector spaces.
Algebra courses deal with this question, and the existence of an isomorphism between the
finite-dimensional vector spaces E and E*' is easily proved.

Remark 2. Following from the duality expressed by (1-1), we point out that the covector
e *i of the dual basis associates with x the ith component Xi :

(e Oi ,X) = e'i(x) = eOi(xje) = Xi.

Terminology and notation. We note that the law (1-12) of the change of components of
any covector is that of change of basis vectors (1-3).
It is not the case for a vector: the matrix is inverse! That is the reason why, initially, every
vector (element of E) was called a contravariant vector and every covector (element of
EO) was called a covariant vector.
14 Chapter 1

This terminology is logically given up because vectors and covectors exist as their own
entities regardless of any basis change.
But later it could well be that we say 'indices of contravariance' and 'indices of
covariance' .
According to convention the components of vectors show an upper index and the
components of covectors a lower index.
Recall that the Einstein summation convention about indices is systematically used and
that any basis vector e; is distinguished by a lower index whilst any covector e*; of the
dual basis is characterized by an upper index.
We are going to consider all of that with tensors of higher order.

1.3.4 Tensor of Type (~)

The inertia tensor is a well known example of this tensor type.

D A tensor of type (~) is a bilinear form defined on Ex E .1

D The vector space of bilinear forms defined on Ex E is called the tensor product
space of two spaces E* .

It is denoted
E* ®E*.

So any tensor of type (~) is an element of E * ® E * and we denote such a tensor as

t E E* ® E*.

Tensor expression. Given a basis (e; ) of E we say:

PR3 A tensor of type (~) is expressed as

t =I ue *;.0. e *} ,
'CI (1-15)
where
lij = t(e;,e)

and (e*; ® eO}) is a basis of E* ® E*.

I Sometimes called a covariant tensor of order 2.

 Tensors 15

. . I
Proof Let x = x' e; be any vector of E where the various x' are defined by
e*; : E ~ R : x H e*; (x) = x; .

Let us fonn the n 2 tensor products

e*; ®e*} : Ex E ~ R: (x,y) H e*; ®e*} (x,y)

such that2

Let us prove that the n 2 tensors e*; ® e*} of type (g) compose a basis of E* ® E*.
First, every tensor t of type (g)
is a linear combination of the different elements e *; ® e *} .
Indeed, its value for any (x,y) is

t(x,y) = t(x;e;,y}e) = x; y}t(e;,e) = tijx; y}

= tij (e*; ® e*})(x,y).

So every tensor t of type (g) is written:

its components being the reals t ij = t( e;, e} ) .

Second, the different elements e *; ® e *} are linearly independent. Indeed, for every pair
of vectors (e"e s ) we have:
e*; ® e* j (er,e..) =e*; (e r ) e* J (e s ) =8:8/ .

Therefore we have Vw, # E E*:

w®# = w;Jl} (e*; ® e*}) = 0

=> OJ;Jl/e*; ® e*})(e"e s ) = 0

=> OJrJl s = 0 Vr,s E {l, ... ,n}.

Remark. As a vector x is sometimes and excessively referred to its components x;, a

tensor t of type (g) can be referred to its components t ij (and so for higher order tensors).

I
We aIso denote \/ e *; , x ) = x ; .

2 We also denote (e x)(e

*; , *} , y) = x; y} .
16 Chapter 1

Change of basis
We recall that any tensor is an 'intrinsic mathemathical entitiy'; that is,
independent of the choice of basis; in other words, each real defined by a bilinear form t is
not 'altered' by a change of basis.
Given a change of basis defined by ej = a~e i ' the components t ij of a tensor t E E* ® E*
are transformed as follows:
V(x,y),(x',y') E Ex E: t(x,y) = t(x',y')
<=> t( x i ei,y Je J) -- t( x "e"y
, 'S')
es
i j _ I ,r fS
tijx y - t rs x y . (1-16)

Such a (general) requirement of tensor theory allows testing the 'tensor character'. So, let
us use it in order to obtain the formulae of transformation of components of t.
We recall [see (1-5) and (1-6)]:
i
X = akx
i,k
x
,P
=
fJPn X n ,

yi = a/y" , ylS = fJ~ym .

The condition (1-16), namely:

implies
(1-17)

Rule. We will notice the presence of elements of two matrices a in (1-17) and of one
matrix a in (1-12); that is, to each covariance index corresponds one matrix a. It is the
reason why every tensor of type (~) is sometimes called a second order covariant tensor.

Conversely we have:
r fJs ,
t lj = fJ i } t rs • (1-18)

1.3.5 Tensor of Type (6)

D A tensor of type (6) is a bilinear form defined on E* x E* .

D The vector space of bilinear forms defined on E* x E* is called the tensor product
space of two spaces E.
 Tensors 17

This space being denoted E ® E , any tensor of type (~) is such that
IE E®E.

Tensor expression. The reader can transpose the previous developments from tensors of
type to tensors of type (~).
(g)
He will define n linear forms on E* :
ej : E* ~ R :(JH·~ ej(m) = mj

and then will consider the n 2 tensor products

e; ® ej .- E'x E' ~ R: (m,p) H e; ® ej(m,p)
such that

Therefore he will be able to state:

PR4 A tensor of type (~) is expressed as

(1-19)
where

and (ej®e) is a basis of E®E.

For that the reader will prove that if (e j ) and (e j ) are bases of E, then the different
e j ® e j are linearly independent.
He will also prove that every tensor tEE ® E is expressed as (1-19).

Change of basis
It is proved (see Exercise 4) that the components of a tensor t of type (~) are
transformed as follows:
(1-20)

Rule. We will notice the presence of elements of two matrices fJ in (1-20) and of one
matrix fJ in (1-6); that is, to each contravariance index corresponds one matrix fJ (inverse
of the matrix of basis change). It is the reason why every tensor of type (~) is sometimes
called a second order contravariant tensor.
18 Chapter 1

Conversely we have:
ti] = a pi a qj t f pq . (1-21)

Remark. Given two vector spaces Eq and E r of respective dimensions q and r, the
corresponding tensor product space is Eq ® E r of dimension qr. It is the set of tensor
products x® y of any x E Eq and any y E Er.
Bases (e i ) and (e~) of respective spaces Eq and E r imply that (e i ® e~) is a basis of
the qr-dimensional space Eq ® E r .

1.3.6 Tensor of Type (:)

D A tensor o/type (:) is a bilinear form l defined either on Ex E* or on E* x E .

Thus a tensor of type (:) is either an element of the tensor product space E* ® E or an
element of the tensor product space E ® E* •

From the covectors of the dual basis:

e*i : E ~ R: x H (e*i ,x) = Xi

and the vectors of the basis of E:

ej :E* ~R:wH(ej,W)=OJj'
the reader will define the n2 tensor products:

e*i ®e j :ExE* ~R:(x,ro)He*i ®ej(x,ro)

such that
(1-22)

that is

As before, the reader will establish that the various tensors e *i ® e j form a basis of the
vector space E* ® E (likewise for E ® E* ).

I Sometimes called mixed tensor of order 2.

 Tensors 19

From

it will be deduced:
t=t~e,®e'jEE®E'
and also
u=u/e*;®e j EE*®E.

In addition, the reader will easily establish the formulae of transformation of the previous
tensors:

Ur
'S_ ipSj
- ar U;
j
,

the rule being: To every covariance index corresponds one matrix a and to every
contravariance index one matrix p.

Given t = tje; ®e*j E E®E* we say:

D The transposed tensor of t is the tensor of E* ® E ,denoted t t , such that

VepEE, Ve,qEE':
't (e p,e'q) = t(e,q ,e p)' (l-23a)
that is:
(l-23b)

In other words:
V t = t iJ e i to.
'CI
e' j •• ,
t = ( 't) j i e,J,o.
'OJ ei
i
=t )e
'),0.
'OJ ei • (l-23c)

This last result actually verifies (l-23a).

Example. Given OJ = Wi e" and x = xi ei' the tensor product

,0.
W'OJ X = w;x i e *;,0.
'OJ ei

leads to

In the same manner the reader will define the transposed tensor of u E E* ® E .
20 Chapter 1

1.3.7 Tensor of Type (~)

D The vector space of p-linear fonus defined on Ex ... x E (p spaces E) is called the
tensor product space ofp identical vector spaces E* .

It is denoted

and has dimension n P •

In the same manner we say:

D The vector space of q-linear fonus defined on E* x··· x E* (q spaces E*) is called
the tensor product space of q identical vector spaces E.

It is denoted

and has dimension nq .

D The (~)-tensor space associated with E is the vector space of (p+q)-linear fonus
defined on the Cartesian product (x P E) x (xq E*) of p spaces E and q spaces E* .

This n p +q -dimensional space is denoted l

D A tensor o/type (;) associated with E is an element2 of the (; )-tensor space r:.
We denote this (p+q)-linear fonu by

t Er:.

The reader will easily verify the following proposition.

I To simplify the presentation, we have fust chosen p spaces E and next q spaces E*. The order of

successive spaces must be specified.

2 Also called a p-order covariant and q-order contravariant mixed tensor.
 Tensors 21

PR5 CJr A tensor of type (~) is expressed as

(1-24)
where
t ,- j
t( e
JI···jq -
- ,e ,e °JI , ... ,e *Jq)
j , ••• j
/ ... p i p

and were
h thedl'f'"~erent e*il,o,,o,
'CJ ••• 'CJ e
*ip,o,
'CJ eiI
,0, ,0,
'CJ ••• 'CJ ejq constItute
•
a b'
aSls 0f Tqp

=(®P E*)®(®q E).

Change of basis
Every transfonnation of components of a tensor of type (~) associated with a basis
change (1-3) is immediately obtained by considering the rule:
[ 3 covariance index => 3 matrix a]
and
[3 contravariance index => 3 matrix.8] .

From this rule it is easy to express any transfonnation of components.

For example, let
tijk, e*i ®e*j ®e k ®e*'

be a tensor of type (;).

To express the 'primed' components in function of the 'unprimed' we simply proceed as
follows.
Having written

where the various a and .8 follow from the rule, we immediately replace the dots by
successive indices of t ij k i , that is:

Remark 1. According to usage, it is necessary and useful to consider tensors of type (~).
They are the scalars (independent of basis choice!).

Remark 2. We recall that the previously introduced Kronecker symbol is only a symbol
(and not a tensor).

But we can introduce:

22 Chapter 1

D The Kronecker tensor t5 is a tensor of type (:) whose components are

gi = I if i=j,
J 0 if i*j.

It is a very helpful tensor because its components are unaltered under any change of basis:

D The zero tensor is a zero multilinear form denoted by o.

1.3.8 Symmetric and Antisymmetric Tensors

Let us consider for instance elements of E' ® E' [resp. E ® E].

D A tensor of type (~) [resp. oftype (6 h is symmetric if

Vx,YEE: t(x,y) = t(y,x)

[resp. Vw,p E E' : t(w,p) = t(P,w)].

This definition is equivalent to the following

Indeed, we have:

Of course, the property of symmetry is intrinsic; that is, independent of the basis. Given a
change of basis ej = a~ei we have:

[likewise: t'ij = t' ji] •

The previous definition may be generalized to higher order tensors.

 Tensors 23

D A tensor of type (~) is partially symmetric if it is symmetric with respect to

pair(s) of corresponding indices;
in other words:
If there are symmetries following from every transposition of two indices of same
variance.

D A tensor of type (~) or type (Z) is completely symmetric if every transposition of

indices changes the corresponding component into itself.

Remark. Given an n-dimensional vector space, every symmetric tensor of order 2 has
n(n + 1)/2 independent components.

Now we consider tensors which play an important role in mathematics and physics: the
anti symmetric tensors.

D A tensor t of type (~) [resp. of type (~)] is antisymmetric if

"Ifx,y E E: t(x,y) = -t(y,x)

[resp. "IfW,fl E E· : t(W,Jl) = -t(u ,w) ].

This definition is equivalent to the following

Indeed, we have:
t(W,Jl) = -t(u ,w)

Of course, the property of anti symmetry is intrinsic; that is, independent of the basis.
Given a change of basis ej = a~ei we have:

t'i} = pipq
pitpq = _pipq
pitqp = _pipi
qp t
qp = _t'ii
,

[likewise tij = -tji].

We deduce
t(x,X) = 0
24 Chapter 1

or
t;; = t ii = 0 .

The previous definition may be generalized to higher order tensors.

D A tensor of type (~) is partially antisymmetric if it is antisymmetric with respect

to pair(s) of corresponding indices;
equivalently,
if there are anti symmetries following from every transposition of two indices of
same variance.

D A tensor of type (~) or of type (g) is completely antisymmetric if every

transposition of indices changes the corresponding component into its opposite.

Remark. Given an n-dimensional vector space, every anti symmetric tensor of order 2 has
n(n -1)/2 independent components.

PR6 Every tensor of type (~) [or of type (~)J can always be decomposed, in a unique
manner, into the sum of a symmetric tensor t s and an antisymmetric tensor t A :
t=tS+tA·

Proof We have, Vx,y E E:

ts(x,y) = t(t(x,y)+t(y,x»,
tA(X,y) = t(t(x,y)-t(y,x»

and the proof follows.

Given any basis (e; ) of E and its dual basis (e *;), the previous proposition is also proved
from
t
(ts ) ij = (t ij + (j;)'
(tA)ij =t(tij -(j;)'

[likewise with tensors of type (~)].

 Tensors 25

2. OPERATIONS ON TENSORS

Before defining operations on tensors, we say:

D Tensors are equal if they are the same element of a same tensor space.

Example. The following tensors

. • . k
t = t' jk e i ® e J ® e' ,
u= uP qr e P ® e *q ® e *r

of the same space E ® E * ® E * are equal if all their corresponding components are equal

for every index value.

2.1 TENSOR ALGEBRA

2.1.1 Addition of Tensors

An inner law, namely the addition, can be defined on the set of same type tensors.
We say:

D The sum of two tensors whose n p + q components are respectively t},'jp and

uJ"'q is the tensor of type

, ... J p
Cp ) the components of which are til iq
J1··Jp
il . iq
+ U JI···Jp •

The addition of two tensors of type (~) is

Tq
P
x Tq
P
~ Tq .
p"
(t u) H t + u

where t + u is the tensor sum.

2.1.2 Multiplication of a Tensor by a Scalar

D The product of a tensor with components t},'jp by a scalar k is the tensor whose
components are k t },\ .

The multiplication of a tensor of type (~) by a scalar k is

RxT; ~T; :(k,t)Hkt,
where k t is the product of t by k.
26 Chapter 1

Example. Given two tensors t, U E T1, we have Vkl , k2 E R:

k It + k 2 U = k It i jk ei to.
'DI e *j to.
'DI e *k + k 2 Uijkt O
e i 'DI.e' j to.
'DI e *k
= (k/ jk + k2U i jk )e i ® e*j ® e*k.

2.1.3 Tensor Multiplication

D The tensor multiplication of any tensor I of type <;) and any tensor u of type G) is
the mapping

® :T: xT/ T::: :~ (/,u) ~ I®u,

the tensor product t ® u being such that

I ® u(w(l) , ... ,w(q),w(q+l) , ... ,w(q+s)' x(l) , ... , x(p) ,X(p+l) , ... , x(p+r))
= t(w(l) , ... ,w(q), x(l) , ... , x(p)) U(W(q+l) , ... ,w(q+s)' X(p+l) , ... , x(p+r))· (1-25)

Example. Vt E Tl, Vu E TIl:

I® U = (tijk e i ®e*j ®e*k) ® (u/ e*P ®e q )

=tiku
J p
qe®e*j®e*k®e*P®e
I q.

It is a tensor of Tl.
This law ® verifies the following properties:
P1. The tensor multiplication is bilinear:

VI T: ' VU(l), U(2) E T/ :I ® (u(1) + U(2)) = I ® u(1) + I ® U(2) ,

E

Vt(1),/(2) E T:, Vu E T/ : (/(l) +/(2))® U = 1(1) ® U+/(2) ® u,

Vk E R, VI E T;, VUE T/: k(/®u) = kl®u =I®ku.
These properties are immediately checked from the definition of tensor product.

P2. The tensor multiplication is associative:

VI,u,s E T: : (/® u)®s = I®(u®s) = I ® u®s.

P3. The tensor multiplication is not commutative.

Let us give the following obvious counter-example of tensor product of vectors:
x®y=y®x
i j to. ji to.
<=> xyei'DIej=yxej'DIe i
 Tensors 27

~ xiyf =/Xf
Xi Xf
~ ----,-=-. =(k), kER
y' yl
that is iff the two vectors are parallel. Thus the tensor multiplication is generally not
commutative.

D r;r The tensor algebra is the infinite-dimensional vector space:

T R e Ee Eoe TOe 1: 2 e 1:1e ... e rqe ...
=
2 ° I p'

direct sum of vector spaces the dimensions of which are higher and higher, and
where R represents the tensors of type (~) (also called scalars).

This space is provided with a bilinear inner law: the tensor multiplication.

Therefore we can express:

PR7 The tensor algebra T is associative, non-commutative, and of infinite dimension.

2.2 CONTRACTION AND TENSOR CRITERIA

We consider tensors of type (~) such that p,q?l.

2.2.1 Contraction

D The contraction of a tensor is the operation which consists in choosing a

contravariance index and a covariance index, in equalling these indices, and in
summing with respect to the repeated index.

For example, let us consider the tensor

t mp k em 'OIe
to. to. ok
p 'Ole
'1'2
E 1\ .

J
Contracting p and k we obtain a tensor of T whose components are

um = Lt mp p = o;t mp k = t
m }} + tm2 2 + ....
p

It is a tensor having lost one contravariance and one covariance because the component
change is such that:
_ s:kt,mp - s:k pmpPa r tif - pmpka r (if _ pm s:rtif
U ,m -up k-u p i f k r- i f k r - i uf r

=Pimui •
28 Chapter 1

This is in accordance with the rule for the change of vector components, and thus the
contraction of t E r/ leads to u E rd .
We can express:

PR8 Every contraction of a tensor removes one contravariance and one covariance.

PR9 After q contractions a tensor of type (:) is reduced to a tensor of type (~) (in
principle q! in number).

Example 1. Given a 3-dimensional space E, the contraction of

leads to the scalar

Example 2. Given a 2-dimensional space E, two successive contractions of

kr *i to. • j to. to.
t = tij e 'Ole 'OIe k 'OIe r

lead to the following scalars:

kr 11 12 21 22
t kr = tll +t12 +t 2I +t22 '
kr 11 21 12 22
trk = til +t12 +t 21 +t 22

D The contracted multiplication is the tensor multiplication with contraction.

Example. The contracted multiplication

E' ® E ~ R: (w,x) H (w,x)

is such that
(1-26)

Amongst the different contractions of the tensor product t ® u of two tensors we

emphasize the following:

Notation. The contraction with respect to the last index of t and the first index of u is
denoted byl
t·u.

I If this conventional notation is allowed.

 Tensors 29

Example. Given two tensors

the contraction of

with respect to indices k and q leads to

Amongst the possible double contractions of a tensor product of two tensors I ®u we

emphasize the following:

Notation. The contraction with respect to the last index of I and the first index of u
followed by the contraction with respect to the penultimate index of I and the second index
of u is denoted by
I: u.

Remark. The dot between two tensors corresponds to the previous type of contraction. If
the contraction concerns other indices, then this must be specified by letting both the
indices of contraction between brackets.
For instance, given 1= tije'i ® e'} and x = xk e k , we have the following covector

I· x = t;kxk e';

= (tIl XI+tI2 x2+ ... )e'l +(t21XI+t22x2+ ... )e'2 + ... ,

which is different from the (1,1) contraction:

tk}X k e ' j -_ (
til I
X +t2I x 2
+ ...) e 'I + ( I
t12X+t22x 2
+ ...) e '2 + ....

Following from the properties of the tensor multiplication, we immediately have:

- the associative property:
(I(l) ./(2))·/(3) = I(l) . (/(2) ./(3)) = I(l) '/(2) ·/(3)
and

- the following distributive property:

(t(1) + 1(2))' 1(3) = 1(\) . 1(3) + 1(2) . 1(3)'
30 Chapter 1

Example 1. Given x==xie i EE, y==yJ eJ EE and t==tpqe*P®e*q EE*®E', we

immediately have:
X·I·y==tijX'yl

== I(x,y)
Indeed, we have
x·t·y==xie Ipq
·(t e,p®e,q)·y1e 1 ==xie·t
Ipq e'P(jqyi
1
== Xi e ·tPi. yl e'P == xiyit pj. (jP == txiyl
I lj f ,

which is really I(x,y).

Example 2. Given x == xie i E E and 1 == t / e*P ®e q E E' ® E, we have:

x·1 == Xi e i . (t Pq e'P ® e q ) == Xi t Pq (jt e q

== X' ti q e q E E
and also

q I
== t, X eq .
Thus, we have
x.t== I t · X . (1-27)

Example 3. The double contraction of any tensors 1 E E ® E and u E E' ® E' is a scalar
since, given
t==tiJe,®e j , U==U pq e*P®e*q ,
we have:

It is commutative:
t:u==u:/.

Example 4. Given arbitrary second order tensors I, u and v, verify that

I: (u· v) == u: (v· I) == v: (I· u). (1-28)

For instance, given tensors of type (:), we have

t·• (u· v) == t pq (u· v)Pq == tq(uPv i ) == uPv i t q == uP(v· t)'

P I q I qp I P

== u: (v ·t)
 Tensors 31

Remark 1. The double contraction of tensors of order 2 decomposed into symmetric and
anti symmetric parts is such that:
l:u=l s :u s +IA :u A·

Indeed, we have VI E E ® E, Vu E E* ® E*:

Is :u A = (tSri(UA)ji =_(/S)ji(UA)ij =-I s :u A =0

and

Remark 2. The contraction 0·1 between the Kronecker tensor and 1= t~ er®e*k is the
(: )-tensor

Thus the double contraction is the following real

s: •
u.
I -uJ
- s:k tJk -- t kk ' (1-29)

with summation over k.

This is the so called trace of the second order tensor I and this scalar associated with I is
said to be an invariant because its value is the same in all coordinate systems

In particular, we mention that the (: )-tensor 0 . 0 has the following components

(1-30)

In a 3-dimensional space the double contraction is the following real value:

0;0/ = oj = oi + 0; + 0: = 3.
We also have
(1-31 )

2.2.2 Tensor Criteria

Until now we have been able to recognize tensors either from the definition
directly, or from the transformation of components through basis changes. We are going to
show a very useful criterion ensuring the tensor character of given mathematical entities; it
will be based on contractions.
32 Chapter 1

(i) Special case.

We know that given two vectors x = x'e i and y = yi e i' if aijxi yi is a tensor of type (g)
(scalar) then the various aij are the components of a tensor.
Conversely, if the various aij are the components of a tensor a then the expression
aijxi yi which is the result of two contractions, namely between the tensor a and the

tensor x ® y of type (~) is a scalar.

(ii) Special case.

With respect to a basis (e*i ® e*i ® ek) we consider n 3 expressions ailk of which one
may wonder if they are tensor components.
First, given arbitrary following vectors and covector x = Xi ei' Y = yie i' m = wke*k , if
the different aij k are the components of a tensor then aij kXi yi W k is an (intrinsic) scalar.

Conversely, if the previous a/xi yi OJ k is an (intrinsic) scalar then we are going to prove
that the different a/ are the components of a tensor.
Indeed, let x' P ,y,q and OJ~ be primed components of respective arbitrary vectors and
covector with respect to a basis (ej) = (a~ei).
The scalar is such that

and thus, given the arbitrariness of the vectors and of the covector, we have:
I r _ k i ifJr
a pq -aij apaq k.

This proves that the various a/ are the components of a tensor.

The previous example may be generalized; n p +q expressions are the components of a
tensor of type (~) iff the complete contraction with respect to p arbitrary vectors and q
arbitrary covectors is an (intrinsic) scalar.

(iii) General tensor criterion.

First, in particular, we consider n 3 expressions aij k and an arbitrary covector m = w pe*P .

We know that if the various aij k are the components of a tensor, then the various
k
aij OJ k (= Aij) are the components of a tensor.

Conversely, if the various Ail are the components of a tensor; then given arbitrary vectors
x andy, Aijxiyi is an (intrinsic) scalar. This scalar a/wkxiyi means [see (ii)] that the

various aij k are the components of a tensor.

 Tensors 33

So with respect to a basis (e*; ® e*; ® ek) the n 3 expressions a/ are the components of
a tensor iff, for every arbitrary covector w = OJke*k the n3 expressions a/ OJ k are the
components of a tensor.

The general tensor criterion is expressed as follows:

If the contracted multiplication of a mathematical entity and an arbitrary tensor is a tensor,
then the mathematical entity is a tensor.

Criterion. If the contracted multiplication (k times) of a mathematical entity and a tensor

of type (~) leads to a tensor of type (:), then the mathematical entity is a tensor of type
(::t~ ).
In the previous example the tensor (of components) aij k is really of type (~::=~) = (~).

Example. The work done by a force f applied to a particle whose infinitesimal

displacement is represented by the vector dx is expressed as
dW = J; dx'.
The contracted multiplication of the entity of components J; and the vector of
components dx; leads to a scalar. So the entity of components J; is a tensor of type
( 0+1-1
0+1-0
)= (0)
l'

So the force appears to be a covector.

3. EUCLIDEAN VECTOR SPACE

Let E be an n-dimensional vector space that we are going to provide with a metric
from the definition of a new law: the scalar multiplication.

3.1 PRE-EUCLIDEAN VECTOR SPACE

3.1.1 Scalar Multiplication and Pre-Euclidean Vector Space

We are going to recall the scalar multiplication which immediately introduces a

very important tensor on E.
34 Chapter 1

D The scalar multiplication on E is a mapping

Ex E ~ R: (x,y) f-7 x.y

such that, \ix,y,z E E, \ik E R:

PI. x.y= y.x , (commutativity)

P2. (x + y).z = x.y + y.z ,
x.(y + z) = x.y + x.z , (distributivity)
P3. (kx).y = k(x.y) = kx.y , (mixed associativity)
P4. [\iYEE: x.y=O] => x=O.
The real x.y is called the scalar product of x andy.

D A vector space E provided with the scalar multiplication is called a pre-Euclidean

vector space.

We note that the first three properties show a symmetric bilinear form since it is linear
with respect to every element of any pair of vectors and is commutative. The fourth
property shows this form is nondegenerate.

3.1.2 Fundamental Tensor

PRIO The scalar multiplication on E defines the nondegenerate symmetric bilinear form
g: ExE ~ R: (x,y) f-7 g(x,y) = x.y,

that is, a tensor of type (~).

D <iF The tensor g is called thefundamental tensor on E.

It is such that \ix, y E E:

g(x,y) = X· g. y = x.y = (x,y). (1-32)

Now, we express the tensor g by considering a basis (e i ) of E.

Given vectors x = Xi ei and y = yJ eJ ' the properties P2 and P3 imply:

(1-33)

In a natural manner we denote the components of g by

(1-34a)

So we have:
 Tensors 35

(1-34b)

and thus (1-33) is written:

(1-35)

Thus the expression of g is

(1-36)

Remark. The property PI means that the scalar product is symmetric; that is, the tensor g
is symmetric:

The property P4 is interpreted as follows:

[Vyi: gijXiyi = 0] ~ [Vi: Xi = 0]
or

So the system of n equations g ijXi =0 only admits the trivial solution, and thus

g = det (g ij ) *0.
Conversely, the reader will immediately verify that if n 2 reals gij are such that gij = gii

and det(g ij ) * 0, then the scalar product fulfills the four properties of the scalar
multiplication.

To conclude this section we say:

D Any two vectors x andy are orthogonal if their scalar product is zero:
xl.y if (x,y)=O.

3.2 CANONICAL ISOMORPHISM AND CONJUGATE TENSOR

3.2.1 Canonical Isomorphism

Let us establish the canonical isomorphism existing between E and E*. How?
Given a bilinear form
g: Ex E ~ R :(x,y)~ g(x,Y)

we express:
36 Chapter 1

PR11 The bilinear fonn g being nondegenerate, there is a (canonical) isomorphism

between E and E' defined by the flat mapping 1:

b : E ~ E' : x H Xb = g(x, )

such that with every vector x there is associated a covector defined by

g(x, ):E~R:YHg(X,y)=x.y.

We observe that the real x.y is

g(x,y) = xb(y) = (x,YJ (1-37)

We note that the covector x b, that is g(x, ), is expressed with respect to a basis (e';) of
E' as
x,) =gijx j e ';
g(
(1-38a)
=g·x

This last equality shows the contraction between g = g ije ,; ® e' j and x = Xk e k .
We really have Vv E E:
j ';
) =gijx e .v e =gijx v k j k s:;
g(
X,V k Uk

In particular, with any basis vector e k E E there is associated the covector of E' :

(1-39)

Now let us prove the proposition.

The flat mapping is linear since immediately 'ix, Y E E:

(x + Y)b = g(x + y, ) = g(x, ) + g(y, )

It is injective (,one-to-one') since

Xb = Yb
=> (x-Yh =g(x-y, )=0
=> x = y (since the bilinear fonn is nondegenerate).

I Also called 'lowering mapping' (or 'application bemol' in French).

 Tensors 37

It is surjective ('onto') because we are going to prove that with every covector WEE' is
associated one vector x such that W = x b .
Indeed, we have
W = OJe'}
}

where

In addition, given the bilinear fonn of the components gij the vector which fits the
question has components Xi such that

Therefore, there is a (unique) solution since the matrix (gij) is nonsingular.

The proposition is so proved.

The isomorphism provides E' with a pre-Euclidean structure in a natural manner.

3.2.2 Conjugate Tensor and Reciprocal Basis

D The inverse mapping of a flat mapping is called a sharp, or raising, mapping l :

#: E' ~ E:w HW#.

Symbolically:
(#rl = b.

Let us show the components of the vector w# dual of x b by introducing the 'conjugate
tensor'.
Under the previously defined isomorphism, to the components Xi of x there correspond
the following components of the covector xb:

Xi =xb(e,) =g(x,e,)=\x J eJ,e,)=gijx J (1-40)

and thus (I-38a) is written:

(I-38b)

Owing to the canonical isomorphism, we have written Xi instead of OJi , as specified later
on.

1 Called 'diese' in French.

38 Chapter 1

Let us denote the inverse matrix of (g ij) by (g ij ); it exists because det( g ij ):;to O. It is
written without rigor as
i cofactor of gij
gY= _ _ _ _-----'-
det(gij)

From the equalities

G'c/' g ij g;k=uk
"i
(1-41)
we deduce
(1-42)

D (iF The gil are the n 2 components ofa tensor of type (~) called the conjugate tensor
ofg.

It is denoted

and is expressed as
(l-43)

The bilinearity of g-I is certain because this latter is a function of covectors g(x,) and
g(y, ) such that:
g-l(g(X, ),g(y, » = g(x,y).
In an explicit manner, given
g-I : E* x E* ~ R: (g(x, ),g(y, »H g-l(g(X, ),g(y, »,
we have
g -I( g ( x, ) ,g( y, » = g -I( xie *i ,yje OJ) = g -I(e,e
*i OJ) xiYj

and
_ r s_ ir js _ ;r s: j
g ( x,y ) - grsx Y - grsg x;g Yj - g U r xiYj

= g ij x;Yj"

Remark. The tensor character of g-I is also revealed from a criterion of tensor calculus.
Indeed, the formula (1-42) expresses a contraction between the tensor of components gij

and the tensor of components Xi of type (~). Since the contraction leads to a (~)-tensor
of components xj, we know that the gil are the components of a tensor of type (tll-=-~)'
 Tensors 39

The components of the conjugate tensor g-I are

g ij = g -I(e *i ,e OJ) (1-44)

since

Given any co vector OJ E E*, we notice that the vector OJ # = g -I (OJ, ) is expressed with
respect to a basis (e i ) of E as
g -I ( OJ, ) _
- g ij OJ) e,.

This is also written:

In particular:
(1-45a)

We say:

D QF The image of every vector of the dual basis (e *i) under the raising mapping, that is
the vector of E
ek =(e*k)# =g-I(e*k, ) (1-45b)

is called the reciprocal vector of e k .

Thus we have

(1-45c)
and consequently

(1-46)

In addition we have:
ei.e) = gU,
(1-47)
e,.e) = J/.
Indeed,
e i . e) = gig ego g)r er = gig g)r ggr = giq 15; = gU ,
e,.e) =e,.g)k ek =g)k g'k =0/.

D The reciprocal basis is the basis of E composed of vectors e' .

40 Chapter 1

Of course, the vectors of a reciprocal basis are linearly independent; that is, the linear
combination

implies

Indeed, from

we deduce

We note that every vector of the reciprocal basis, for instance e k , is orthogonal to each of
n - I vectors of the original basis (e j ) , except for i = k where the scalar product equals I.

Classic example. In the usual 3-dimensional space the reciprocal basis of a basis made up
of normed vectors (e,) is composed of three vectors

. e) /\e k
e' = --'------
ej.(e) /\ek)

where e j .(e) /\ e k ) is the volume ofthe parallelepiped constructed on the basis vectors.

3.2.3 Covariant and Contravariant Representations of Vectors

Fundamental remark

Given a basis (e;) of E and the dual basis (e*)) the formulae (l-40) and (1-42):

Xj = gijx ) , x j = gijxJ.

show the correspondence between any vector x E E and its dual: the covector Xb E E*.

CiF The scalar multiplication lets us identifY the isomorphic spaces E and E*, the
reciprocal basis (e j) and the dual basis (e *j) are identified.
In an explicit manner we recall "Ix E E :

which is also denoted:

 Tensors 41

Thus the scalar multiplication (or fundamental tensor g) allows the following
representation of every vector x of E:

D (iF The covariant representation of a vector x is the expression

x=xie i
where the components

are called the covariant components of x.

This terminology follows because the components Xi are the components of the covector
Xb in the dual basis (e*J) of E* since, given the covariant representation x = xJe J , we
have:

We also have

<iF In general, the (usual) contravariant components x' are different from the
covariant components as illustrated with the following example; but in the case of
orthonormal bases we have (e ' ) == (eJ and the corresponding contravariant and
covariant components are equal.

Example. Let us consider a basis composed ofnormed vectors e 1 and e 2 .

", ,

,
,

,,
,
,,
Fig. 1
42 Chapter 1

The detennination of the vector e l of the reciprocal basis follows from the following
conditions:
ell. e 2 and el.e l = 1.

[likewise for e 2 ].

The covariant components of x are

Xl = x.e l = Ilxll cosa,
x2 = x.e z = Ilxllcos p,

whilst the contravariant components are the well known Xl and xZ.

Let us conclude:

(jF With the scalar multiplication we can consider x and Xb = g(x,) as a unique
tensor of order 1 for which two representations Xi e i and xie i exist.

Contracted product and scalar product

In this new vision we consider again the contracted product of any two tensors of
order 1; namely Vx,y E E:

X'Yb =XiYi =x'gijyJ =xJyJ =xb·y (1-48)

It is nothing else but the scalar product

i J
X.y = gijx Y
sInce

3.2.4 Representations of Tensors of Order 2 and Contracted Products

Given the canonical isomorphism between E and E*, we know that Ex E ,

Ex E*, E* x E and E* x E* are isomorphic and so are E* ® E*, E* ® E, E ® E* and
E®E.

(i) First, in particular, given vectors u and w we consider the possible tensor products of
corresponding vectors and covectors:
 Tensors 43

where
Ub =g(u,) and Wb =g(w,).

We are going to make explicit these particular tensors of order 2; that is, Vx,y E E:

x·/\ . Y = 1\ (x,y) = u b 18> wb(x,y) = uiwje*i 18> e*j (x,y)

= UiWjXi yj = (ub,X)(Wb,Y) = (u b ' x)(w b . Y),

x·/ 2 . Yb = 12 (X'Yb) = Ub 18> W(X'Yb) = u i w j e*i 18> e j(X'Yb)

= UiW)XiYj = (ub,X)(W'Yb) = (U b ·x)(w· Yb)'

Xb . 13 , Y = 13 (Xb,y) = U 18> Wb(Xb,Y) = uiwje i 18> e*j (Xb,y)

=UiWjXiy j =(u,Xb)(Wb'Y)=(U'Xb)(W b .y),

Xb ·/4' Yb = 14(X b'Yb) = ul8> W(Xb'Yb) = uiwje i l8>e j (x b'Yb)

= uiw j XiYj = (U'X b)(W'Yb) = (u· Xb)(W' Yb)'

Since we have u b . x = U· Xb and w b ' Y = W· Yb' the four previous reals are equal and we
denote the common value by
t(x,y)=x·t·y,
that is,
(1-49)

where u b and x are tensors of order 1 of opposite variances, obviously, and so are Wb and
y.

If we now consider a tensor of type (:), for example U 18> WbEE 18> E* , we have Vy E E :

(I-50)

sInce

[likewise for u b ® W E E* ® E].

From the definition of the contracted multiplication and if we consider vectors and
co vectors a, b, c, d such that the following operations are possible, then the reader will
immediately prove:
(a ® b) . (c 18> d) = (b.c) a 18> d (I-51 )
and the double contraction:
(a ® b) : (c ® d) = (b.c)(a.d) . (I-52)
44 Chapter 1

(ii) Second, we view the general tensors of order 2 by considering the canonical
isomorphism between E and E*. The generalized expressions of the previous tensors I\,
12, 13 and 14 are obviously:

'<Ye *j
I I -- t ije *i!O. E E* '!O.E*
<Y ,

12 =t/e*i ®e j E E* ®E ,

13 =tijei®e*j EE®E*,

14 = tij e i ® e j E E®E.

By referring to the contravariant- contravariant representation of a general tensor of

order 2, namely:

we obtain the other representations as follows.

From

we deduce the contravariant-covariant representation:

the covariant - contravariant representation:

and the covariant-covariant representation:

The four previous representations are the ones of the second-order tensor I.

We conclude this section with the respective representations of the transposed tensor /1:
/1 = (/ l)iJ e ® e . = t ii e ® e .
I J I ] '

tl=(/I)le®e j =tie®e j
) I } I '

I We specity that repeated contractions withg allow making tensors of types (6) and (~) correspond
canonically, so we have:
 Tensors 45

I
t -_(It) if e ltv.
'CI e
j_
- t ji e itv. j
'CI e .

So the various representations of a symmetric tensor t = It are such that t if = t ji, t if = t ji

and
I I
t j = tj .

We note that the symmetry of any (~)-tensor t , for instance, is expressed as follows:
VX,YEE: x·t·y=y·t·x,
since
x·t· y = Ct·x).y == y.(t·x);
that is, explicitly:

3.3 EUCLIDEAN VECTOR SPACES

Let E be a pre-Euclidean vector space.

D (iF The vector space E is said to be Euclidean if the symmetric bilinear form g
defining the scalar multiplication is positive-definite;
that is, for every nonzero x E E:
g(x,x) > o. (1-53a)

In other words, by Euclidean vector space we mean pre-Euclidean vector space such that
given any basis (e i ):

(1-53b)

The previous definition can be considered from the following signature notion.
Given a basis change e~ = a~ e i we know that the formula of component change of a (~)
tensor is

If (g) designates the matrix (gij) we denote these equalities in the matrix context as
follows:
(g') == la (g) a
where la == (a;) is the transpose of a = (an.
46 Chapter 1

In order to present (g') as a diagonal matrix, we view a as a = OD where 0 IS

some orthogonal matrix and D some diagonal matrix:

Thus,

=> (g') = D 0- 1 (g) 0 D.

We can transform the matrix (g) into diagonal form by a good choice of orthogonal matrix
0; that is:

[g" .

Thus we have:
d ll
(g') = [

Ifwe choose d ii = !glY, for every i, then (g') is a diagonal matrix with elements + 1 or
-1, the nonsingularity preventing any zero.
Note that the choice of d ii cannot alter the signs of diagonal elements.
For instance, the matrix 0 can be selected such that the +1 are in front.

Thus, the fundamental tensor can be written:

g = e'! ® e'! + ... + e' P ® e' P - e' P+! ® e' P+! -'" -e' p + q ® e' p +q

withp+q=n.
Several definitions of the signature of g have been given. Let us make the following
choice.

D The signature of g is the pair (p,q) of natural numbers such that p + q = n.

This is independent of the basis choice.

If q = 0 then the fundamental tensor is positive-definite, the signature shows positive

signs only. It characterizes every Euclidean vector space.
 Tensors 47

D If the signature shows positive and negative signs, then it is said to be indefinite
and the corresponding pre-Euclidean vector space is called a pseudo-Euclidean
vector space.

D If the fundamental tensor g is positive-definite, it is called the metric tensor of the

Euclidean vector space.

D The norm of any vector x of a Euclidean vector space is the real

Ilxll = ~(x,x). (1-54)

We note that if the space is pseudo-Euclidean then ~(x,x) is not a norm in the usual

sense because it can be a positive real, an imaginary number or zero, and ~(x, x) is said
to be a pseudo-norm.

D A basis of a Euclidean vector space is said to be orthonormal if the basis vectors

are normed and orthogonal; that is:
(1-55)

or in an equivalent manner:

or

PRl2 The covariant and contravariant components of every vector x with respect to any
orthonormal basis of a Euclidean vector space are such that
Xi =Xi,

the variance being without significance.

Proof We have:

Notation. The metric tensor appearing not any more, the Einstein summation convention
is not applicable any longer and
- either it is necessary to introduce the summation sign:
I xl1 = L (Xi)2 ,
2
48 Chapter 1

- or we adapt the Einstein convention, namely, if an index is repeated once then it is a

dummy index indicating a summation with the index running through the appropriate
integers (i = 1, ... , n ).
For example:

Remark 1. In every orthonormal basis the various representations of any tensor coincide.
Indeed, it is obvious that

and for instance:

k _ _ ijkl_
t ij I - t ijkl - t - ....

Remark 2. In the same manner the formulae of transformation of components are for
example:
t''l""q. =a·
't)l
... a 'q}q. t )l···Jq..

Example. The metric tensor g and its conjugate g-I are immediately expressed in
relation to the cylindrical coordinate basis.
Indeed, since any position vector is written
p = rcosO Ix + rsinO Iy + Z I z
with respect to the Cartesian orthonormal basis (1 x,I y,l z) of the usual 3-dimensional
space, we deduce the vectors of the cylindrical basis, namely:

e r = 8p =(cosO,sinO,O),
8r
eo = 8p = (-r sinO, r cosO, 0),
ao
8p
ez =-=(0,0,1).
8z
We note this cylindrical basis is not normed (but orthogonal!) and the only nonzero gij

are

The matrices of the metric tensor and its conjugate are respectively:

o
-2
r
o
 Tensors 49

4. EXTERIOR ALGEBRA

The works of H. Poincare and E. Cartan were at the root of exterior calculus. This
theory has been essential for developing the 20th century physics. Weare going to consider
completely antisymmetric tensors.

4.1 p-FORMS

Let E be an n-dimensional vector space,

t be a p-linear form defined on x P E.

4.1.1 Definition of a p-Form

D A p-linear form is called skew-symmetric or completely antisymmetric if, for any

permutation CT making (I, ... ,p) ~ (CT(1), ... ,CT(p», we have "itx\ ,... ,x pEE:

(1-56)

where &" (or sign CT ) is +1 or -1 according to CT being an even or an odd

permutation.

D The alternation mapping or antisymmetrization is a (linear) mapping

Ap : T~ ~ T: : t ~ Apt
such that "itx\ , ... , x pEE:
I
Apt(x\, ... ,xp ) = -; ~>O"t(xO"(I)' ... 'xO"(P»' (I-57)
p. 0"

where
CT is the permutation (l, ... ,p) ~ (CT(l), ... ,CT(p»),
L is the sum over all the permutations of the sequence (l, ... ,p).
0"

Remark 1. This sum is also denoted L

(jES p
where S p is the symmetric group (order p!).

We note the presence of the conventional factor ~ and the sum over all p! elements of
p!

Remark 2. The linearity of the antisymmetrization is obvious.

50 Chapter I

D Cir A p-form (or exterior form of degree p) is the image of a p-linear form by
antisymmetrization.
In other words:
A p-form is a completely antisymmetric tensor of type (~ ).

Notation. The p-form Apt being well defined, we denote it by a Greek letter:
W= Apt. (1-58)

The set of p-forms is a vector subspace of T~ denoted by QP (E) or simply by QP. Thus
we write:
WE QP.

Example 1. "dt E T20 , "dx,y E E:

w(x,y) = AzI(x,y) = Ht(x,y) -t(y,x)).

Example 2. \;ft E T20 , "de;,e J E E (basis vectors):

w(epe) = A2 t(e i ,e j ) = t(tij - I ji )·

We evidently have the skew-symmetry property, namely:

w(e;,e J ) = -w(eJ,e;),
also denoted

Example 3. \;ft E T30, "dX\,X2,X 3 E E:

1
w(XI'X 2,X3 ) = A t(XI'X 2,X3 ) = -[t(Xp X2,X3 ) + t(X 2,X3 ,X1 ) + t(X 3 ,X1 ,X2)
J
3!
- t(X1,X3 ,X2) - t(X3 ,X2,X1 ) -t(X2,XI'X3 )],

4.1.2 Exterior Product of I-Forms

Given the I-forms e'; making up a basis of 1)0, we have the successive and parallel
equalities:

A 2(e'; @e*J)(x,y)

=t(e"@e*j(x,y)-e*i @e*j(y,x»
 Tensors 51

= t(e' i ®e') -e') ®e'i)(x,y)

= toY.,e· 1 ® e' J (x,y)

where the symbols oZ are: where the symbols 0;::::' are:

o if (/J) is not a pennutation of OJ) o if (II".!p) not a pennutation of 01 .. i p) ,

1 if (/J) is an even pennutation of (ij) 1 if (II .. .!p) an even pennutation of (il .. i p),
-1 if (/J) is an odd pennutation of OJ). -I if (II".! p) an odd pennutation of 01 ...i p).

In conclusion we have:

(l-59a)

Notation. From now on, according to convention, we denote the I-fonns e Oi making up a
basis of 1)0 by (J i •

So the previous expression is written:

(1-59b)

Remark.
c5Z (JI ® (JJ (x,y) = «(Ji ® (Ji _(Ji ® (Ji)(X,y)
= (Ji (x) (Ji (y) - (Ji (x) (Ji(y)

D r:r The exterior product ofp linear fonns (J ik is the p-fonn I

(JiJ. (Jip =Oil .. .ip (Jh ® ... ®(Jl p (1-60)

/\ ... /\ 11··'!p •

1 The symbol 1\ can be called 'wedge' or 'hat'.

52 Chapter 1

1
We note that the factor - is not present in this definition. So we have conventionally
p!
chosen:

Several other conventions exist, but, as the reader will see afterwards, we have adopted the
one which eliminates the most constants.

Examples. 031\0' =03 ®O'-O' ®e 3 =--fJ'I\03,

Oi I\O} I\Ok = aY!JK o1 ®OJ ®OK
=Oi®OJ®Ok+OJ®Ok®Oi+Ok®Oi®OJ
-oj ®Oi ®Ok _Oi ®Ok ®OJ _Ok ®OJ ®Oi.
These examples show the importance of the order of exterior product terms.

Exercise 22 proves that the exterior product of two I-forms Q = aiO' and P = fJ}O} is

Ql\p= i)aA -a;fJ;)O' 1\0i =~(a,fJj -a,fJJO' 1\0i

,~ 2
= aifJJOi I\O).

4.1.3 Expression of a p-Form

(i) First, we consider the case of any 2-form WE 0 2.

PRI3 The C~ products ()' 1\ (}i (i < j) form a basis of the vector space 0 2.

Proof On the one hand, every 2-form is a linear combination of C~ = n! products

2!(n - 2)!
of I-forms °
i I\O} (i < j).
Indeed, let us express the value of W on vectors x and y of E, more precisely the image of
(x,y) by w; namely:
w(x,y) = w(x' e"y) e j ) = x'y) w(ei,e)
= Wi}Xi y} [letting wi} =w(ei,e}) = A2t(e i ,e})]
=W'2X 'Y2 +WZ,x 2Y ' +W13 X'Y3 +W 3,X 3
Y' +... SInce W;; = 0 ]
[.

=W 12 (x'y2 -x 2y')+W 13 (x'y3 _x 3y')+... [since wi} =-wji]

= IWi}(Xiy} -x}/)
i<j
 Tensors 53

i<j

= LOJijOi /\OJ(x,y).
i<J

Thus the products 0' /\ OJ (i < }) generate every 2-form.

On the other hand, these products are linearly independent, namely:
LOJijOi/\OJ=O => Vi,}E{I, ... ,n}: OJij=O,
i<j

because Vr,s E {I, ... ,n}:

0= LOJijOi /\OJ(er,e,) = LOJij(Oi(er)OJ(es)-OJ(er)Oi(e s))
i<J i<j

= 20J rs (r < s).

Therefore, the expression of a 2-form relative to the basis (0 i /\ 0 J) i<J is

i<)

' . C n!
Th IS sum contams 2n = terms.
2!(n - 2)!
The components of a 2-form OJ relative to the basis (0 i /\ 0 J) i<J are called the strict
components and are denoted OJ(ij)'
SO the expression of a 2-form relative to (0 i /\ 0 J) i<J is

OJ = OJCij)O i /\0 J,
by knowing that the sum is only over i < } .
Of course, we have
(J) -
- -wij
I oj /\ OJ .
2!

Example. Let E be a 3-dimensional vector space.

The exterior product of two I-forms a,p E QI is a 2-form OJ with C~ strict components,
namely:

these being the components relative to the basis (0 i /\0 J)i<J of Q2 .

(U) Second, let us view the following.

Example. We consider any 3-form OJ E Q3.

We recall that a tensor of type (~) OJ = OJljk 0' ® OJ ® Ok is completely anti symmetric if
54 Chapter 1

The value of w for x, y, Z E E is

w(x,y,z) = x I y j z kw(e"ej,e k ) = OJijkx I y j Z k

123
= OJ 123 X Y z + OJ 231 X 231 312
Y z + OJ 312 X Y Z
132 321 213
+ OJ 132 X Y Z + OJ 321 X Y Z + OJ 213 X Y Z
+ ...
and the anti symmetry of various OJ Uk implies:

w(x,y,z) = OJ123(X1/Z3 + X2/Z1+ X3/ Z 2_ Xl/Z2_ X 3y 2z 1_ X2/Z3)

+ ...
x' Xi Xk

= IOJUk y' yj yk
i<j<k
Z' Zi Zk

where

and so on.
Therefore the expression of OJ is immediately:
w= I OJ uk 0' /\Oi /\Ok
i<j<k

by letting

The strict components lead to

(iii) Finally, we consider the case of any p-form.

D A p-form OJ is decomposable if there existsp I-forms a , such that

PR14 The C~ decomposable forms Oil /\ ... /\O'p (il <···<i p ) make up a basis of the
vector space Q P •
 Tensors 55

Proof This proposition can be proved as before or in a rather different manner as shown
in Exercise 23.

We specify that:
fOil Oip)( )_t5il",iPfOit®"'®OIp)( )_t5 il .. ,ipt5 it t5 Ip
~ /\ ... /\ eil,···,e ip - Il .. .I p ~ eil,···,e ip - It .. .Ip il ... i p

=t5 il .. .iP •
}l .. ,}p

As previously we can obtain the expression of a p-form, that is

l!) - - L OJ,','
l"'p
Oil /\ ... /\ Oip , (l-6la)
~<"'<ip

where il"'" i p E { 1, ... , n }.

We also have:
1
l!) = -OJ i OiJ /\ ... /\ O'p • (1-6lb)
p! IJ .. P

It is essential to distinguish the n P components of a tensor of type (~) from the C:

components of the p-form obtained from anti symmetrization. Consequently we define:

D The components of a p-form l!) relative to a basis (9 ~ /\ ... /\0 ip )~< <ip of QP are
called the strict components and denoted by

So the expression of a p-form is

il Oip
l!) = OJ(" ,') O /\ ••• /\ (1-62)
l"'p

by knowing that the sum is only over il < ... < ip •

Remark 1. Expression of a p-form relative to the basis (0 ~ ® ... ®O ip) of T~ .

Let

be a tensor of T~ .

We have:
56 Chapter 1

The next to last equality allows to write:

= ~t5il 'Pt (1-63)
OJil ·ip p! JI .. Jp 'I .. ip •

Remark 2. According to usage we call O-form any scalar; that is an element of QO (of
dimension C~ = 1).

PR15 If the degree of a p-form is higher than the dimension n of the vector space E then
the form is necessarily zero, that is, Vp > n: QP = {O}.

Proof If p > n then two indices are necessarily equal and the p-form is thus zero.

4.1.4 Exterior Product of p-Forms

D The exterior product of a p-form W and a q-form 11 is a (p + q) - form defined

by
(p+ q)!
W/\Il = --Ap+q(w® 11), (1-64)
p!q!

that is an element ofthe vector space Qp+q having C~+q dimensions.

Problem. What is the image of p + q vectors Xl , ... , X p+q by the exterior product W /\11 ?
From the definition of anti symmetrization we have VXl , ... , X p+q E E :
 Tensors 57

_ 1 ~ I, . . i p +q
- - - L. 01 p+q OJ(X, "",Xi ) P(Xi "",Xi ).
p!q! 'I ... i p +
q
." 1 P p+1 p+q

Remark. In particular, we find again the definition of exterior product of two l-forms:

(Ji /\(J) =~A «(Ji 0(JJ) =~(fJ' 00 J _(JJ 0(Ji).

l!l! 2 2!

4.1.5 Exterior Algebra

Let us show that there are vector spaces whose direct sum is an algebra of finite
dimension.
Let us recall that the following (internal) law can be defined in the set of exterior forms of
same type.

D The addition of two p-forms is the law

QPxQP ~QP :(OJ,p)HOJ+P

where OJ + P is the p- form sum,

the sum of two p-forms with respective (strict) components OJ..
(""'p)
and 11
,....O,··.,p)

being the p-form with the C~ following components:

We can also define a second law:

D The multiplication ofap-form by a scalar k is an (external) law:

RxQP ~QP :(k,m)Hkm

where km is the product of OJ by k,

the product of the p-form with (strict) components mOl ip) by the scalar k being
the p-form with the C~ components:
k OJO" p )'

D The exterior product space of vector spaces QP and n q is the vector space of
(p+q)-forms; that is,

Now we can define a third law that is bilinear.

58 Chapter 1

D The exterior multiplication of a p-form OJ and a q-form,., is the mapping

/\ : n p x n q ~ n p+q : (OJ,Jl) H OJ /\}1 ,

the exterior product OJ /\,., being such that VXl , •.• , x p+q E E:

The following properties are verified:

(i) The exterior multiplication is bilinear:

VOJ E n p , V}I,V E n q :OJ /\ c,., +v) =OJ /\}1 +OJ /\ v

VOJ,Jl En p , Vv En q : (OJ+}I)/\v =OJ/\V +}I/\v

Vk E R, VOJ E np , VI' E n q : k(OJ/\}I) = kOJ /\}1 =OJ /\ k}l .

Proof The first equality (the others being demonstrated in the same way) is immediate:
OJ/\c,., +v) = 2!A 2(OJ®c,., +v)) = 2!(OJ®}1 +OJ®v)
= 2!A2 (OJ ®}I) + 2!A2 (OJ®v) =OJ /\}1 + OJ /\ V.

(ii) The exterior multiplication is associative:

VOJ E n p , VI' E nq , Vv E n r :
(1-65)
(OJ /\}I) /\ V =OJ /\ c,., /\ v) =OJ /\}I/\ V
Proof See, for instance, Talpaert (2002).

(iii) The exterior multiplication is not commutative and VOJ E n p , VI' E n q :

01 /\}1 = (-l)pq }I /\01. (1-66)
Proof On the one hand, we have:

and on the other hand:

 Tensors 59

The pennutation (i p+' •••j p+q i, ... j p ) ~ (i, .. .i p i p+' •••i p+q) corresponds to pq transpositions;
thus we have:

Special cases

If WE o.P and ifp is odd then w/\w = o.

If w"u E 0. 1 then W /\)1 = --# /\w and W /\w = 0 .

D The exterior algebra is the vector space

o.(E) = 0.0 E9 0.' E9 ... E9 o.n = ffio.i ,
i=O
the direct sum of vector spaces, which is necessarily finite.

This last precision follows from the fact there is no exterior fonn of degree p higher than
n.

The previous vector space is provided with a bilinear law: the exterior multiplication. We
can express:

PR16 The exterior algebra 0. is associative, noncommutative, and of finite dimension.

The exterior algebra is a vector subspace of the tensor algebra T whose dimension is

dimo. o +···+dimo." =C~ +C~ + ... +C: = 2".

Let us specify that the product of an element of degree p with an element of degree q is of
degree p + q (graduation!).

4.2 q-VECTORS

The previous developments about p-fonns are applicable to q-vectors by 'replacing'

E by E' and thus the reader will easily transpose the main notions about p-fonns into the
context of q-vectors.

D A q-vector is the image of a q-linear form (defined on x q E') by

anti symmetrization.
In other words:
A q-vector is a completely anti symmetric tensor of type (g).
60 Chapter 1

D The exterior product of q vectors e ik is the q-vector

_8 ilh......i 1q e ® ... ® e ·
e il /\ ... /\e iq - (1-67)
q h 1q

The C~ decomposable q-vectors e iI /\ .•• /\ e iq (i, < ... < iq) make up a basis of the vector
space of q-vectors.

The expression of any q-vector is a q-linear form:

h · .
werel'<"'<lqE {I , ... ,n } an d wil···lq -we
_ (*il , ... ,e *iq) .

D The components of a q-vector relative to the previous basis are called the strict
components W(il.iq)

We denote
(1-68)

Example 1. Let us consider any 2-vector w. We have successively:

i<j i?i

i<) is)

(since w!i = _w ii , wii = 0)

i<} i<j

i<}

In particular, the exterior product of two vectors x, Y E E:

x /\ Y = x ® Y - Y®x = (Xi yi - xi yi)e i ® e i

is such that

and thus
 Tensors 61

This completely anti symmetric tensor has C~ strict components

(i < J).

More particularly, concerning the usual 3-dimensional space the C~ = 3 strict components
are

Example 2. Let us consider the exterior product of three vectors x, y, Z E E. We have:

Xi xi xk

xl\yl\Z=
L
i<i<k Zi
y' yl yk e i I\e I\e k ·
i
zi Zk

This particular 3-vector has C~ strict components represented by the determinants.

More particularly, concerning the usual 3-dimensional space E, the C~ =1 strict

component is

It is the well-known mixed product:

X.yI\Z=Xl\y.Z

defining the volume (with sign) of parallelepiped having x, y, Z as sides.

Example 3. Let us consider the exterior product of q vectors:

t=XI\YI\ .. ·I\Z= L
it <··-<iq

where x, y, ... , Z are q vectors of an n-dimensional vector space E.

It is a completely anti symmetric tensor of type (z) having C~ strict components.

In particular, if the number of vectors is equal to the dimension n of the space E then there
is only one strict component and we have clearly:
t =t(12 .. n)e 1 I\e 2 1\ ... l\e n .
62 Chapter 1

So the components of t relative to a basis of the space of tensors of type (~) are such that

where
=0 if two indices are equal,
e'I'n = 1 if the number of transpositions with respect to (il ... i n ) is even,
= -1 if the number of transpositions with respect to (il .. .in> is odd.

More particularly, conceminf a 3-dimensional space the components of a completely

anti symmetric tensor of type ~ ~) are such that

where t is the strict component.

Basis change

The reader having carried on with the comparison between p-forrns and q-vectors
can now conclude this section with the notion of change of basis.
For example, given a 2-dimensional space E, we know that a tensor of type (~) is such
that

but it is also
t = t,(l2) e; /\ e;
and thus

If E is an n-dimensional space, we have:

t(iJ)e i /\e j = t(iJ)(fJte~) /\ (PJe~)
=t(ij)(p/e; + ... + Pte~)/\(p~e; + ... + p;e~)
= t(ij)[CP/p; -Pi2p~)e; /\e; +CP/p; -p?p;)e; /\e; + ...]

=t(ij) [P; p; - Pt P;] e~ /\ e~

but it is also
t,(rs) e~ /\ e~
 Tensors 63

and thus

In general, given a p-vector t, a change of basis of an n-dimensional space E, leads to

/3: /3t ... /3in

t,(rsn) = t(ij ..k) /3; /3} ... /3] (1-69)

/3; /3% ... /3:

In the following paragraph we are going to consider affine spaces, also called point
spaces. [See, for instance, Talpaert (2002)].

5. POINT SPACES

In mechanics, the consideration of the point space associated with a vector space is
essential, it allows us to introduce the notion of tensor fields for instance.

5.1 POINT SPACE AND NATURAL FRAME

Let E be a (real) vector space of dimension n which we will assume to be pre-

Euclidean.

5.1.1 Point Space

We know that there exists an n-dimensional point space E associated with E and a
mapping

such that:
(i) \:;Ia,b E E: ab =-ba,
(ii) \:;Ia,b,c E E: ab+bc =ac,
(iii) \:;10 E E; \:;Iv E E, ::I! x E E: ox=v.
64 Chapter 1

Remark 1. It follows from (i) and (ii) that

ab = ao+ob,
that is
ab = ob-oa.

Remark 2. Given an arbitrary point 0 E E, condition (iii) establishes a bijection between

E and E; that is, given 0 the knowledge of point x implies the knowledge of v and
conversely.

5.1.2 Coordinate System and Frame of Reference

Let us recall the following notions.

D A frame of E is the set {o;e i } composed of a point 0 E E and vectors of a basis
(eJ. The point 0 is the origin of the frame.

D The coordinates of any point x E E relative to a frame {o;e i } are the components
of ox E E with respect to the basis (e i ).

Therefore the coordinates Xi of x E E are such that

We say that the coordinates Xi (i = l, ... ,n) form a system of coordinates (Xi).

Given points a and b of respective coordinates a i and b l , from

ab = ob-oa = (b i _a i ) e i
we deduce that the components of ab E E relative to the basis (e i ) are the differences
between coordinates of respective points band a.

Change of coordinate system

Let {o;e i } and {o/;ej} be frames of reference.

Let us establish the relations between the coordinates of a given point x E E with respect
to both frames.
Given
o/x = x/ie;,
 Tensors 65

and since the relations between basis vectors are

, j
ei=aie j ,

then we have:

Therefore we deduce by identification that

In the same manner, by expressing o'x with respect to the basis (e ri ) we immediately
obtain

Coordinate systems

D The correspondence between the points of E and n-tuples of reals u i (called the
coordinates) defines a coordinate system on E.
Every coordinate line is the locus of points whose only one coordinate varies.

For instance, u 1 is varying while the other coordinates are invariable (u 2 , U 3 , ••• ).
Another coordinate curve can be defined by u 2 = k (k constant) and so on.
So we have a mesh of n coordinate curves defining a coordinate system.

We know that any straight line of E is a one-dimensional subspace of E determined for

instance by a fixed point a E E and a fixed vector wEE; it is defined by
ax=kw,
where k is a real parameter.
Given a fixed frame of reference {o;e~} of E, we know that the straight line equation is
ox=oa+ax.
From oa = a i e~ and W = Wi e~ we deduce that the straight line equation is

ox = (a i + kwi)e~.

The position vector of x being ox = Xi e~ , the straight line is defined by the n following
equations:

The coordinates Xi of the running point x being first degree functions of the parameter k,
we say:
66 Chapter 1

D The coordinate lines being straight, the coordinates are called rectilinear
coordinates with respect to a given fixed frame.

Now, given a fixed frame {o;e~} of E, we consider the rectilinear coordinates Xi of

X E E as arbitrary functions of n arbitrary real coordinates ui , namely:

i,j=l, ... ,n.

We assume the following:

The functions ¢/ are (at least) of class C 2 , the Jacobian such that
J = 8(x 1 , ••• ,x n ) "* 0
8(u 1, ••• ,u n )
on a given domain of (u i ), and we can express:
u i =/fJ(x') i,j=l, ... ,n.

In general, with these new coordinates u i the coordinate lines are not straight lines. Thus
we say:

D The coordinates corresponding to (general case) coordinate curves are called

curvilinear coordinates.

Remark. The curvilinear coordinates u i are rectilinear iff the rectilinear coordinates Xi

are first degree functions of u i ; that is iff the various ax i

are constant.
8u J

5.1.3 Natural Frame

We consider the rectilinear coordinates Xi of a point X E E relative to a fixed

frame {o;e~}; in other words, we define:
. 0
ox = xJe i .

We know that there are n coordinate lines through the point x and let u i be the curvilinear
coordinates.
The Jacobian being different from zero, we denote the tangent vectors to different
coordinate lines by
80x
e·=--. (1-70)
, 8u i

Their expressions with respect to the basis (e~) are

 Tensors 67

We note that the derivatives of ox do not depend on point 0 since for every point 0' E E
different from 0 we have:
o'x = ox-oo' = ox+c (constant vector c)
and thus

So we write naturally:
(1-71)

These n vectors which are tangent to n coordinate lines through x are linearly independent
since the Jacobian J is nonzero and we say:

D The n vectors e i = aix constitute a basis called a natural basis or a (natural)

coordinate basis.

D A frame composed of any point x E E and n basis vectors through x is called a

naturalframe.
It is denoted
{x,aix }.

D The differential of ox is the vector

ax .
dox=-. du f , (l-72a)
auf
also denoted
(l-72b)

where the (contravariant) components of the differential vector are du i •

Example 1. Let us show the local frame at point x of the classical Euclidean plane in the
case of polar coordinates (differential presentation).

Let xl, x 2 be the (Cartesian) rectilinear coordinates,

r, () be the (curvilinear) polar coordinates.

The Cartesian frame being {o;e? ,e~], we write:

68 Chapter 1

dx = dx ' e,0 + dx z e z0
= (cosO dr - r sinO)ef + (sinO dr + rcosOdO)e~
but

and thus
e, = cosO e~ + sinO e~ ,
e z = -rsinO ef +rcosO e~.

So the natural frame is made up of point x and two orthogonal vectors e, and e z tangent
to respective coordinate lines corresponding to constant values of 0 on the one hand, and
to constant values of r on the other hand. Their respective norms are 1 and r.

A natural frame is shown in the following figure.

O~------------L---
Fig. 2

Example 2. Let us show the natural frame at point x of the classical 3-dimensional
Euclidean space in the case of spherical coordinates (derivative presentation).

We know that the Cartesian coordinates and the (curvilinear) spherical coordinates are
related by
x' = rsinO cos¢, x Z = rsinO sin¢, x 3 = rcosO,
where r, 0 and ¢ are respectively the radial distance, the colatitude, and the longitude.

Given the Cartesian frame { o;ef ,e~ ,e~ }, the various vectors of a natural basis are:
 Tensors 69

e] = 8]x = sinB cos¢ e~ + sinBsin¢ e~ + cosB e~,

e 2 = 8 2 x = rcosBcos¢ e~ + rcosBsin¢ e~ - rsinB e~,
e 3 = 8 3 x = -r sinBsin ¢ e~ + r sin Bcos¢ e~ .

o~ ___-\-__ e;

Fig. 3

The natural basis vector e] has the radial direction of ox (¢ and B being constant), e 2 is
tangent to the local meridian (¢ and r being constant) and e3 is tangent to the local parallel
(r and B being constant).
The orientation senses are induced by the ones ofr, Band ¢respectively.
So the natural frame is made up of the point x and three orthogonal vectors of respective
norms 1, rand r sinB .

5.2 TENSOR FIELDS AND METRIC ELEMENT

We are going to define one tensor at each point of the point space E, but first we
consider curvilinear coordinate transformations.

5.2.1 Transformations of Curvilinear Coordinates

We consider a transformation of curvilinear coordinates

70 Chapter 1

such that the functions f} , defined by

U" =fl(u1, ...u n ),

are ofc1ass C q (q 22).

·
The Jaco b Ian a(u1,
1
... ,u n ) .IS assumed
to eb · h·mg.
non-vanls
a(u' ,... ,u rn )

Conversely we have:
j,k = I, ... ,n.

Whenever there is such a coordinate transformation, there is a corresponding change of

frame at a given point.
More explicitly, we consider a point x and a correspondence between two natural bases of
respective vectors

ej
ax
=-. ,
,
e=--.·
ax (1-73)
au' } au'}

Let us seek the formulae of transformation of tensor components given the previous
coordinate transfonnation simply denoted
u'} = u'} (u') i,j = l, ... ,n.

First, we say:

PR17 Natural frames {x;e;} and {x;e~} are respectively associated with any systems of
curvilinear coordinates (u;) and (u'}) such that the formulae of change of natural
bases are

(1-74)

(1-75)

By recalling previous notations of basis changes, we note that in this case we have:
. au'}
PI =--
..
au'

Proof From the differentiable composite mapping theorem, we obviously have

e
, = -ax- . au'
--.
} au' au'} ,
 Tensors 71

ax au'}
e· = - - . - -
, au') au i '

which proves the proposition.

Remark. We specify that the fixed Cartesian frame {o;e~}, from which the various e i
have been introduced, does not appear in this context of curvilinear coordinate
transformations. So the correspondence between natural bases associated with different
coordinate systems is only a correspondence between curvilinear coordinate systems at a
given point.

PRI8 The set of natural bases at x E E is identical to the set of bases of the vector space
E associated with the point space E.

Proof At a given point x E E, by considering the set of frames {x;e i } which correspond
to various curvilinear coordinate systems, we know that each natural basis (e i ) is a basis
of the vector space E associated with E.
Conversely, any basis (ej) of E is a natural basis at x E E by considering the formulae of
coordinate transformations as

I au'
e·) =--e
au'} ,

D Any tensor of the vector space E is said to be a tensor on the point space E
(associated with E).

This is well defined from the previous proposition. Any tensor t on E can be expressed
with respect to a local basis (e i) associated with the curvilinear coordinate system (u i) .

Transformation of tensor components

We are going to find the formulae of transformation of tensor components under a

curvilinear coordinate transformation at x E E.
For a tensor t of type (~), we have:

i au'} I
t = ('}e ' . = (e. = ( --. e).
) , au'
and thus
U
aI}
i i _ au
i I)
(
I)·
=--( (---.( . (1-76)
aui ' au')
72 Chapter 1

For a tensor t of type (~) we obtain in the same manner:

au,j
t· = - - t ' (1-77)
, au i )

Likewise we have:
'pq _ au'p au,q ij ij _ au' au) 'pq
t ---.--. t , ( -----(
au' au) au'p au,q

a 'p a )
t'P =_u_ _u_(. i
t . -----. t
aU i "'-.Iq
uu 'p
q au i au,q ), ) - au' P au) q
and so on.
For example:
a
, pqr I _
a a " a
U 'P U ,q U "'-.II
U m uu ijk n
t s ---.--.- - - - - - t m •
au' au) auk au/S au n

Thus we say:

Rule. Each position of index of primed components of any tensor at x E E is clearly

imposed: To any upper index corresponds one au' and to any lower index corresponds
au
au
one-.
au'
(Inversely for unprimed components).

Example. Gradient of a real-valuedfunction.

We consider the function

f: R n ~ R: (u! ,... ,u n ) ~ feu! ,... ,u n )
and a coordinate transformation defined in a simplified writing as

i,j = l, ... ,n.

The differential

df= af du i = 8/. du,j

au' au')
af au') i
=----du
au') au'
implies
 Tensors 73

a1 au'}
----
8f

So the various a1 are the components of a tensor of type U).

au'
This tensor is called the gradient off

Its components with respect to any frame { x, e i } are a1 .

au'
Its Cartesian coordinates are the successive a1 .
ax'

5.2.2 Tensor Fields

From previous developments we are going to assign only one tensor to each point
XE E.

If we denote the set of tensors of type (~) at x E E by Tx ~ we say:

D (if> A tensor field of type (~) on E is the mapping

t: E~ U( {x} x Tx~): x H t(x) = (x,t x ),

XEE

where t x is a (~)-tensor assigned to x.

We know that at each point x E E a tensor can be expressed with respect to a natural basis
(e i ) of E associated with a coordinate system (u i ).

The operations on tensors assigned to the same point pose no problem since the tensors are
relative to natural frames at this point (for which changes of natural bases have been
defined). For example, the sum of two tensors is clearly defined from their components
with respect to different natural bases at the same point, because the tensors can be
expressed with respect to a same natural basis.

However, difficulties arise for tensors assigned to different points! How can we compare
tensors related to natural frames at different origins? It is necessary to study the change of
the natural frame between two infinitely neighboring points by referring to the natural
frame at one of the points.

We are going to consider pre-Euclidean spaces; more precisely, we recall:

74 Chapter 1

D A point space E is pre-Euclidean (resp. Euclidean) if the associated vector space

E is pre-Euclidean (resp. Euclidean).

5.2.3 Metric Element

D The metric element of the point space E (also called line element) is the square of
the distance between two infinitely neighboring points of E.

Given any two infinitely neighboring points x and x + dx, we know that their distance is
the norm of the vector defined by this pair of points, i.e. Ildxll.
We denote the metric element by

(i) First, we consider a rectilinear coordinate system (Xi) on E.

The coordinates of points x and x + dx are respectively x' and Xi + dx i with respect to
e; }.
the rectilinear frame { 0;

From

we express the metric element as

or
(1-78)

where we have obviously let g~ = e;. e; .

Remark 1. The metric element of E and thus the components gij of the fundamental
tensor g with respect to the rectilinear basis (e7) completely define the metric of E.
We also note that the various gij are not dependent on x, they are constant.

Remark 2. In relation to an orthonormal basis of a Euclidean space frame, the metric

element of the Euclidean space is immediately:
 Tensors 75

(ii) Second, we consider a system of curvilinear coordinates (u i ) such that:

i,j = l, ... ,n.

Given a natural frame {x;e i } as defined before, the expression of the metric element of
the point space is
(1-79)

where the various gy = e;.e are the components of the fundamental tensor g.
j

Remark. The metric element of E and thus the components g ij of g with respect to the
natural basis (e i) completely define the metric of E. But here we note that the various g u
are dependent on x and are obviously not constant.
A metric tensor is assigned at each point of E and is an example of tensor field.

Example. The metric element relative to spherical coordinates r,O,¢ such that
x = r sin 0 cos ¢ , y = r sin 0 sin ¢ , z = r cos 0
is the sum
ds 2 = dr 2 + r 2d0 2 + r2 sin 2 0 d¢2

of the following radial dr 2 , meridian (rdO)2, and longitudinal (rsinO d¢)2 squared
displacements.

5.3 CHRISTOFFEL SYMBOLS

5.3.1 Definition of Christoffel Symbols

Given a coordinate system (u i ) on E, the essential problem is to determine at any

point x + dx a natural frame in comparison with the natural frame at an infinitely
neighboring point x.

Relating to a natural frame { x; aiX } we have:

The natural frame {x; e; } is deformed from x to x + dx and becomes {x + dx; e; + de; }.

The problem is to determine the increments de; of natural basis vectors.

76 Chapter 1

The differentials of vectors of the natural frame {x;e i } are the following vectors
expressed with respect to the natural basis (e i ) :

dej = OJ~e, (1-80)

where the n 2 matrix elements OJ( are such that

dej.e k =OJ~gik =OJ jk ·

We will later give a physical interpretation of these elements.

Remark. From the relation between reciprocal basis vectors ei.e J = J/, we deduce:

and thus

So the (covariant) components of dej are - OJ( and the covariant representation of
vector differentials is
(1-81)

We can now introduce the Christoffel symbols.

The vectors of the natural basis (e i ) depending on coordinates uk, we have obviously:

(1-82)

From (1-80) and (1-82), we deduce that the vectors ake j are expressed with respect to the
natural basis (e i ) so that we say:

D The Christoffel symbols r~k are n 3 functions l of curvilinear coordinates Uk that

are the components of the vectors ake j :

(1-83)

We also note that

(1-84)

I Sometimes called Christoffel symbols ofsecond kind and denoted L~}.

 Tensors 77

We have immediately

dei =r~kduk e i (1-85)

and we point out the differential fonns

i i k
OJ i =rik du . (1-86)

PR19 The Christoffel symbols are not the components of a tensor (in general).

Proof Given a curvilinear coordinate transfonnation u' P = U' P (u i ), we have:

au' au' au'

de' = d(-- e) = d(--)e +- de
P au'p' au'p' au'p ,
but
de~ =r~~ du" e~ ,
thus
,q d"
r pr U e q, -- d( -aui-) e i + -
au'-riik dUk e i
aU'p au'p
a i a,q G' a ,q
= d(_u_)~e' + rJ _u_ dU k _U_e,q
au'p au' q ,k GU'p aui
a u' a,q
2 a i a ,q a k
u du" e' + r J _u__u_ _ u_ du " e'
au,pau" au' q ,k au'p aui au" q
and so

(1-87)

The second tenn of the right hand member means that the Christoffel symbols are not the
components of a (~)-tensor.

Remark 1. The lateral position of the upper index with respect to the lower indices does
not matter. In addition we note that the symbols OJ~ are also not tensors.

Remark 2. We note that the Christoffel symbols define tensors if the affine (or linear)
changes of curvilinear coordinates are such that
2ui
---=0
a
au' P au"
for every i,p and r.
78 Chapter 1

D The Christoffel symbols of first kind are n 3 functions of curvilinear coordinates

uk such that
rijk = gihr;k (1-88)

where the first index is the descending one!.

We have

that is
(1-89)

We can also write :

(1-90)

Property of symmetry. The lower indices are interchangeable.

Indeed, from the (necessarily supposed) equality of second derivative vectors

o /OiX) = 0 jei =r; e k

and

we deduce

In the same manner we have

r kij =rkji

and the last two indices are interchangeable.

·
GIven th·IS property 0 f symmetry, th
ese i ·
re abons n + 1 In
are nn-- . numb er.
2

5.3.2 Ricci Identities and Christoffel Formulae

(i) From the n(n + 1)/2 functions gij = ei.e j of the curvilinear coordinates uk we
deduce:

1 The Christoffel symbol offrrst kind is also denoted [i, jk].

There are other defmitions such that r ijk = g jFit = [ik, j).
 Tensors 79

that is,

So by considering the coordinates Uk, we obtain the Ricci identities

(1-91)

Since for any value kthere are n(n+1)/2 tenns i\gij there are then n 2 (n+1)/2 Ricci
identities.

(ii) The Christoffel fonnulae are obtained from the Ricci identities written as follows:
(1)

and, after cyclic pennutations of indices and by using the symmetries, they are:
(2)
and
(3)

From (1) + (3) - (2) we obtain the first Christoffel formula:

(1-92)

The second Christoffel formula is

(1-93)

Examples. The Christoffel fonnulas let calculate the Christoffel symbols given a metric
(and thus given the various gij):

In polar coordinates:
r 212 = r221 = r, r l22 =-r,

the other symbols being zero.

In spherical coordinates:

ri2 = -r, rf2 =r~3 = cot() ,

r 33l • 2 ()
= -rSln , r;3 = -sin() cos() , rr3 =rfl = l/r,
the other symbols being zero.
80 Chapter 1

In cylindrical coordinates:
T221 =T212 = r , T122 = -r,

T122 =T;1 = l/r , Ti2 = -r,

the other symbols being zero.

Remark. As long as the tensor g does not exist (no pre-Euclidean structure), the only
method to calculate the Christoffel symbols consists in going back to rectilinear
coordinates by expressing

and thus

[See also Exercise 27].

5.4 ABSOLUTE DIFFERENTIAL, COVARIANT DERIVATIVES, GEODESIC

5.4.1 Absolute Differential of a Vector, Covariant Derivatives

We consider a vector field

v:E~U ({x}xTd):x~v(x)=(x,vx)·
XEE

The field vector at point x is Vi e i with respect to the basis of the natural frame {x;e i }.

From any x to a neighboring point x + dx , the components of any field vector (as well as
the basis) change.
The corresponding variation of v is:
dv = dv i e i + Vi de i
j
= d V e i + V OJ i
i i
ej
and thus we say:

D w The absolute differential of a vector field is a vector field defined by

dv = (dv i + v k OJDe i (1-94)

 Tensors 81

So the corresponding components with respect to (e i) are not dv i apart from when the
coordinates are rectilinear since the various rjk , and thus de;, vanish.

D The (contravariant) components (dV)i of the absolute differential dv with respect to

(e i) are denoted
Vv i = dv i +vkmi
(1-95)
= (8 j v i +vkr~)duj

and are also called the absolute differentials of contravariant components v; .

PR20 The various

(1-96a)

are the components of a tensor of type (:) called the (: )-covariant derivative of v.

Proof Since the various du j are the components of an arbitrary vector with respect to
(e i ) and since Vv i are the components of the vector dv with respect to this basis (e i ),
then a tensor criterion relative to the contracted multiplication lets us assert that the
various components 8 j Vi +rjk v k are of a tensor of type (:) also denoted by

(1-96b)

[See also Exercise 31].

If we now consider the reciprocal basis (e i ) and thus the covariant representation

then we have the following absolute differential of v:

dv = dV i e i + vide i
= (dvj -vimj)e~

D The covariant components (dv) j of the absolute differential dv with respect to

(e i ) are denoted
VVj = dVj -vimj
(1-97)
i k
= (8 k vj -Virjk)du

and are also called the absolute differentials of the covariant components vj •
82 Chapter 1

PR21 In general, the various components

a)Vi -rtvk

are the ones of a tensor of type (n denoted by

V)Vi = a )Vi -rtvk
(1-98)
= Vi,) -r~Vk

and are called the covariant components of the (~)-covariant derivative of v.

Proof Given a transformation of coordinates
u') =u')(u i ) i,j = l,oo.,n,
we know that the primed components of v = vie i are expressed in function of unprimed
components as follows:
, au i
v· =--v
} au') ,
and thus

but from
P au_
a __
r" =r' _u " _
au_+
q a 2u i a"
u
}k pq au'} au' au,k au')au,k au i '
that is,

we deduce:
8v' a_
"-. i _U
_}_=_U_V m_
aU_+r"v'
i
_ri _Ua p___
aU_q
v
aU,k aU m au,k au') }k r pq au') au,k "

Thus, we have obtained:

8v' "-.
-}--r"v' = (_U_Vi -rsv ) _u_ _u_
am a i

au,k }k r aum m, s au,k au')

 Tensors 83

5.4.2 Absolute Differential of a Tensor, Covariant Derivatives

We consider a tensor field

t :E ~ U({x} x T; ) : x H t(x) =(x,t x) .
xEE

From any x to a neighboring point x + dx, the components of any field tensor (as well as
the basis) change.

We are going to determine the differential of any tensor of the field.

Without restricting the theory we choose tensors of order 3, for instance in order to
simplify the notation:

t=t i j k ei'CIe
"'" *j 'CIe
"'" k .

Let us determine the components Vt i / of the absolute differential tensor with respect to
the basis of the natural frame { x; ei }, this tensor being denoted

dt = Vt i / e i ®e*j ®e k .

Given a fixed basis (e i ), for every vector v = Vi e i and everyone-form fJ = !lie'i of class
C', we have:

In particular, since e *i (e j) = t5S we have:

From

we deduce
de*i = -OJ~ e*j. (1-99)

Now, from (1-80) and (1-99) we can determine the components Vt i / of dt . Indeed,

dt = dt i / e i ® e*j ® e k + t i l d(e i ®e*j ®ek)

= dti/e i ®e*j ®ek +t'/(OJ;e h ®e*j ®e k -e i ®OJ/,e'h ®ek +e i ®e*j ®OJte h )

= dI i j k e i 'CIe
"'" *j"",
'CIek +t hJk OJhe
i "'" *j"", i k .0. h *j.o. i h "'" *j"", k
i 'CIe 'CIe k -t h e i 'CIOJje 'CIe k +t J e i 'CIe 'lYOJhek

= (dt}ik ihk hik kih) .o.*j.o.

+OJht J -OJ}I h +OJht J ei'lYe 'lYe k
84 Chapter 1

So given a natural basis (e i ) we say:

D The absolute differential of a (~)-tensor field t is a (; )-tensor field dt whose

components are

Vt i/ =dt i/ +liJ~th/ -liJ;t i/ +liJ!t i /

=(8/? +rhpth/ -r;pti/ +r!pti/)du P
. k
= V pt' i du p .

PR22 The various

(1-100)

are the components of a tensor of type (;) called the covariant derivative of t.

Proof Since the various du p are the components of an arbitrary vector with respect to
(e i ) and since Vt i / are the components of the tensor dt with respect to the same basis,
then a tensor criterion relative to the contracted multiplication lets us assert that the
various components V/ / are those of a tensor of type (;). The proposition is thus
proved.

The reader can now immediately write the covariant derivative of any tensor.
For example, the components of the covariant derivative of t = t ii ei ® ei are

V q t ii = 8 q t ii +rihq t hi +rihtqih ' (1-101)

Likewise, we have:
(1-102)

More particularly, from the Ricci identities [Eqs.(1-91)] we immediately deduce:

Vqg ii =0.

So the following Ricci theorem is stated:

PR23 The absolute differential of the fundamental tensor g is zero:

dg=O.
 Tensors 85

5.4.3 Geodesic and Euler's Equations

Consider the following curve c on a Euclidean point space:

[a,b] ~ E: t H c(t)
such that, given the subdivision a = to < tl < ... < t p = b of [a, b), the continuous function
t H c(t) is differentiable on each [ti ,ti+1 ] for every i E [O,p -I].

The curve being defined by the parametric equation u i = u i (I), we say:

D The length of arc c is the real

I rh
= du i du i
e Ja g--dt. (1-103a)
lj dt dt

The length of c is also denoted in terms of the curvilinear parameter s as follows:

(1-103b)

The geodesic equations can be found from the calculus of variations, which consists in
obtaining extremal curves; in this context the geodesic distance between two points is the
lower bound of lengths (if it exists) of curves joining the points, the geodesic arc is the arc
corresponding to the lower bound.

Let us note that a geodesic does not necessarily lead to a minimum of length. Furthermore,
the example of the sphere proves that there are several geodesics joining two diametrically
opposite points (an infinity).

Let us now recall a fundamental result of the variational calculus and let us consider (class
C 2 ) curves joining two points.
With each curve is associated a previously defined length:
du'
Ie = r
Je
F(u i ,-,t)dt = r ds,
dt Je
where

- , t)_~
i
F( u i ,du - giju·i·j
u i,j = I, ... ,n.
dt
86 Chapter 1

A geodesic being a curve such that the variation

8 iSh

Sa
ds = 0,

the variational calculus teaches us that a (necessary) condition of extremum of Ie is

obtained if the system of Euler equations

~(8F)_ 8F =0 (1-104)
dt 8u' 8u'
is satisfied, that is

Substituting the curvilinear parameter s for t, the geodesic equations are

d du i 1 1 8 du i du k _ 0
dt (gu dt ds) - 2ds ;(gik)dtdt-

d 2ui du k du i 1 du i du k
<=> g - - + 8 g . - - - - 8 g - - - 0
if ds2 k i] ds ds 2 i jk ds ds -

d 2u i 1 du k du i du i du k 1 du i du k
g --+-(8 g - - + 8 g - - ) - - 8 g k - - - 0
i] ds 2 2 k if ds ds i ik ds ds 2 i i ds ds - .

Multiplying by gfi we obtain the following system of geodesic equations

(1-105)

5.4.4 Absolute Derivative of a Vector (Along a Curve)

Let (u i) be a curvilinear coordinate system in a point space E.

Let x be a vector field such that to each value of any parameter t corresponds only one
vector of the field and we denote
x = x(t).
By considering
8x
e· =-=8x
,i 8u '

we recall that the differential vector of x is

 Tensors 87

dx
If the parameter t is the time in particular, then is called the velocity of x and is
dt
denoted by v.
The components of the velocity v with respect to (e;) are

Vi
. = __
du ' .
(1-106)
dt
This generalizes, in curvilinear coordinates, the classical velocity relative to (Cartesian)
rectilinear coordinates.

Now, from the notion of absolute differential of a vector field v and by considering an
increment dt of parameter t making pass (along a curve) from point x to a neighboring
point x + dx, we define:

D The absolute derivative of vector field v = Vi e i along a curve is the vector field

If the parameter t is the time, dv is called the acceleration of x.

dt

The components of the acceleration with respect to (e i) are

(1-107)

Remark. Geodesics are curves followed by points of zero acceleration.

Indeed, if the acceleration vanishes then, with a system of rectilinear coordinates (x'), we
have:

i = 1, ... ,n

~ x' =k't+K' (ki, Ki : constants),

88 Chapter 1

and for a system of curvilinear coordinates (u i ) we have:

5.5 VOLUME FORM AND ADJOINT

Let us consider a coordinate transformation defined by

i,j=I, ... ,n.

5.5.1 Volume Form

First we consider the example of a completely skew-symmetric tensor of type (~).

In the case of a 3-dimensional space we recall that such a tensor has one strict component;
let us denote this by D.
The components of the (n-tensor relative to the basis (e*' 0 e* J 0 e*k ) are

Dijk =£ ijk D

where the Levi-Civita symbol is

o if at least two indices are equal,
[; ijk = 1 if (ijk) is an even permutation of (123),
-I if (ijk) is an odd permutation of (123).
Every change of component is clearly denoted by
au i au} auk
D' = - - - - - - [ ; .. D
pqr au' p au, q au" IJk .
Since

the change of strict component is thus

D' = det[ au i ) D. (1-108)

au'}

This formula can be generalized to completely anti symmetric tensors of type (~) on n-
dimensional spaces.
 Tensors 89

Physicists call D a tensor density.

The components of a completely anti symmetric (~)-tensor relative to (e'" ® ... ® e"")
are

Exercise. The reader will immediately verify that a completely antisymmetric tensor C of
type (~) on an n-dimensional space shows the following change of strict component

C' == (det( au' ))-1 c.

au')
It will be proved that the components of such a tensor relative to (e,1 ® ... ® ein ) are

Now, let us define the volume form on a Euclidean space.

Under any basis change defined by

au i
or e'. ==--e
} au')'
we have:

or

By denoting
det g == det(g ij ) ,
the rule of product of determinants is applied, namely:

detg' == (deta)2 detg, detg' == (det ( -au'

- . ) rdetg.
au')

We assume det g and det g' are positive; so we consider direct basis changes (on the
oriented space). We have:

Jdetg' == deta Jdetg, aui.) Jdetg.

Jdetg' == det ( -
au'}

So Jdetg is a tensor density, it is the strict component of a completely anti symmetric

tensor of type (~).

Hence we express:
90 Chapter 1

D The volume form on the Euclidean space IS the following completely

anti symmetric n-form

Ii = v~
uet g e *1 /\ ... /\ e
*n
. (1-109)

The components of Ii relative to the basis (e*;1 ® ... ® e*;n) are

f-l 'I···l. = s '1··.I.

n n
~det g .
(1-110)

Remark 1. The volume spanned by vectors el, ... ,e n is

In particular, in 3-dimensional Euclidean space we have:

f-lyk =Ii(e;,ej,e k ) = e; /\ej.e k

and Ii is the tensor associated with the trilinear mapping making a correspondence from
any (x,y,z) to the mixed product (x /\ y. z).
In this example we note that the various f-lijk are the components of a tensor since we
consider bases of same orientation only; but it would not be a tensor any more for bases of
different orientations.

Remark 2. The following change of strict component

_1_ = (det( au'

~detg' au')
)tl _1_
~detg
(1-111)

lets the introduction of a completely anti symmetric tensor of type (~), dual of volume
form and having the following components relative to (e;1 ® ... ® e;n)

(1-112)

Remark 3. We note that

since the previous sum contains n! equal terms as IIILn I"".n

I""
1112 = 11 213 ... n 1""213
I""
II ... n = ... = 1 •
 Tensors 91

5.5.2 Adjoint

From the p-form notion, duality has allowed us to introduce the notion of q-vector.
The reader will easily transpose the developments about p-forms; for example, he will be
convinced that the q-vectors are elements of a C~ -dimensional vector space.

Let t be a completely anti symmetric tensor of type (~).

The corresponding q-vector of components t il .iq is such that

(1-113)

D The adjoint of the q-vector t il " q relative to the volume form f..L is the (n - q)-
form denoted * t such that

(*t) lq+l
. . =
... l
n
2.,r'l .. in t
1/. il .. iq (1-114)
q.

So the operator * defines an (n - q) -form from a q-vector and conversely.

We also denote the following contravariant components:

In the same manner we say:

D Theadjointofthep-form OJ il " p relative to 11 is the (n-p)-vector,denoted *OJ,

such that

(1-115)

Remark 1. The adjoint ofa completely anti symmetric tensor of type (~) or of type (~) is
a scalar.

Remark 2. The adjoint of a completely anti symmetric tensor of type (n~l) [resp. (n~l) ]
is a covector [resp. vector].

This notion lets us find again such mathematical notions of elementary geometry on R3 as
curl and divergence.
92 Chapter 1

Example. Given an orthonormal basis of classical 3-dimensional Euclidean space, the

vector product of two vectors is the adjoint of the exterior product of two vectors.
Indeed, let us consider
1 = X /\ Y = Xi yl ei /\ eJ = (x' yl - xi y') ei ® ei
= L (X' yJ_xiy')e /\er i
'<l

The adjoint of this (~)-tensor is the covector of components

(*/)k = 2!I Ji"k(X'yl - Xl y')

= t~detg Cijk(X i yl - X' y')

= ~detg C(y)k (Xiyl - Xl y').

or
(1-116)
that is,
(*/)1 = ~detg (x 2/ - x 3y2),
(*/h =~detg(x3yl-xly3), (l-1l7)

(*/h = ~detg (x 1y2 _x2yl).

By considering an orthonormal basis of Euclidean space (detg = I), we know that the
covariance and contravariance are indistinguishable, and we find again the components of
the vector product x /\ y :

5.6 DIFFERENTIAL OPERATORS

We are going to give various presentations of differential operators.

5.6.1 Gradient

5.6.1a Gradient of a Function

Let f: R n ~ R be a differentiable real-valued function of curvilinear coordinates

i
U.
 Tensors 93

Weare going to consider the differential of I denoted

dl= af du i ,
au'
where the various af behave like components of a covector field called the gradient of f
au'
Let us recall the notions of differential and directional derivatives of f

D The differential of I at point x is the linear mapping

dlx : R n ~ R: h H dlx(h)
such that

I(x + h) - I(x) = dlx(h) + iihii1J(h)

lim 1J(h) = O.
h--+O

Let us make this mapping explicit.

Let (e i ) be the natural basis.

Ifwe consider the special vector hi e of R

l
n we have:

I(u l + hi ,u 2 , ... ,u n ) - I(u l ,u 2 , •.. ,u n ) = hldlx(e l ) + Ihlliielii1J(hlel) .

But the following

lim 1J(h l e l ) = 0
hl--+O

implies

Consequently at x the function I possesses a partial derivative with respect to u 1 :

In general we have:

(1-118)
94 Chapter 1

If we now consider any vector h = hi ei of R n , the differential of f at x is the (linear)

mapping

and we write:

dfx(h) = 8f (x)h i , j = l, ... ,n. (1-119)

au)

We also recall the following notion.

D The directional derivative of fat x in the direction of a (unit) vector v is the limit
(if existence):
Dvf(x) = lim f(x + lev) - f(x) . (1-120)
k--+O k

In particular, the partial derivative with respect to u i is the directional derivative in the
direction of the vector ei of the natural basis (e J :

DeJ(x) = af (x). (1-121)

J au}

Now we say:

D The gradient of a function f : R n ~ R, at x, is the covector

gradJ: R n ~ R: hI---? gradxf(h)

such that
gradJ(h) = dfx(h)
that is explicitly:
. af .
(gradJ); hi =- . (x) h} .
au}

So, the components of grad xf with respect to the natural basis (e;) are:

(gradJ); = aJ(x). (1-122)

In conclusion, the covector

gradxf = aJ(x) e*; (1-123)
is such that
 Tensors 95

gradxl.h = o;/(x)e*i .hiei

= o;/(x) hi.

The corresponding gradient field is the covector field denoted by

gradl·

In differential terms, by letting

h =dx =duiei

gradxl.dx = (gradJ)i du i = o;/(x)du i

(1-124)
=dlx ·

Remark 1. The gradient of a function is obviously determined from the knowledge of the
tensor g.
We recall that the natural basis associated with curvilinear coordinates is made up of
vectors ei = ax , where x is the position vector of a point x; each ei is tangent to the ui
au'
coordinate curve at x.
By denoting the unit vector in each tangent direction as e i, we write:

From the differential vector dx = e i du i also written

dx = k(l)du l el + k(2)du z ez + k(3)du 3e3,

we deduce for orthogonal curvilinear coordinates
ds z = dx.dx = k(~)(dul)Z +k(~)(duz)z +k[3/du3 )z.
Therefore the gradient of I, which is
lolA lolA lolA
gradx l =-k - I e l +-k --Z e z +-k - 3 e3 , (1-125)
(I) au (2) au (3) au

is easily determined by knowing that the various k(i) are such that:

(no summation!).

I For a nice and complete presentation see section 2.3 of Lecture 0 of our book « Differential Geometry with
Applications to Mechanics and Physics».
96 Chapter 1

For example, in spherical coordinates r,f),t/J the gradient field IS expressed m the
orthononnal basis (e r ' eo, e¢ ) as
of I Of I of
gradf =-e r + - - eo +---e¢.
A A A

or r of) r sinf) of)

Remark 2. By considering orthononnal bases with curvilinear coordinates Eq. (1-124) is

simply denoted (for any point):
df = oJ du' = (gradf); du'
or
df = (gradf); e*'. (1-126)

Thus given spherical coordinates r,f),t/J we have with respect to the orthononnal basis
(e"eo,e¢) :

and the corresponding I-fonns e*; are

In the same manner, given cylindrical coordinates r,f),z we have with respect to the
,e
orthononnal basis (e r , eo z ) :

and the corresponding 1-fonns e*; are

We are going to generalize the previous notions.

 Tensors 97

S.6.1h Gradient of a Tensor

Let t(x) be a tensor field whose tensor components are differentiable with respect
to the curvilinear coordinates u i of the position vector x of x E E.

D The directional derivative of a tensor field t( x) in the direction of a (unit) vector v

is the tensor field such that Vx E E:

() I " t(x+kv)-t(x) "

D vtx=lm (1-127)
k-->O k

In particular, the partial derivative with respect to u j is the directional derivative in the
direction of the vector e j ofthe natural basis (e i) :

at
Det(x) = - . (x)
J au} =a}.t(x) " (1-128)

In conformity with the gradient of a function we say:

D The gradient ofa tensor field t(x) of type (~) is the tensor field of type C~J such
that:
grad t(x) = a jt(x)®e*j
(1-129)
= D ej t(x)®e*j

Remark 1. The symbol ® does not appear in (1-123) where the tensor is of type (~)"

Remark 2. We note that the linear directional derivative operator in the direction of h,
namely:

lets us deduce the following:

Dh t(x) = gradt(x)"h

since
Dh t(x) = D h i e, t(x) = h' ait(x) = grad t(x)"h"
98 Chapter 1

In differential tenns we have:

(1-130)

Example 1. Given a vector field v = Vi e i let us calculate grad x v at any point x.

We know that the gradient of a tensor field v of type (~) is such that

and the components are

. av i
(grad xv)' j = - -..
au}

The equality
dv x = gradxv· dx
is explicitly
. av'
dv' = - - . du}.
au}

In orthononnal rectilinear coordinates Xi we have:

Example 2. Given a tensor field I = I v e i ® e j let us express grad xl at any point x.

The gradient of a tensor field I of type (;) is such that

I} ali] *k
gra d x ( I e i
,0. ) _ ,0. ,0.
'61 ej - --k ei '61 ej '61 e
au
and the components are

(grad xt)i] k = a kli] .

In a system of orthogonal Cartesian coordinates Xi we have:

all}
(gradxt)i]k = ax .
k
 Tensors 99

5.6.2 Divergence

5.6.2a Divergence of a Vector Field

D The divergence of a vector field v is the (real-valued) function following from the
contraction of the (:) covariant derivative of v.

In coordinates we denote:

(1-131)

Make explicit this notion if a metric is introduced.

From
V'jv i =8 j v i +r~k v k =8 j v i +gih rhjk vk
-- 8 jV i 1 ih(8
+2 g jghk + 8 kghj -
8 hgjk ) V k ,

we deduce that a contraction leads to the following (~)-tensor field:

n
Y iV
i
= 8 iV i 1 ih (8
+2 g ighk + 8 kghi -
8 ) k
hgik V .

By changing the names of summation indices we obtain:

g
ih 8 ighk V
k
= g
hi 8 hgik V
k
= g
ih 8 hgik V k

and thus

Denoting the cofactor of gih by G ih we have:

g ih = - _ .
G 'h
detg
From determinant theory we know that
8 k detg = a h 8 k gih

8 k detg _ ih 8
detg - g k gih

n i _ 8 i ~ k 8 k det g .
Y iV - iV + 2 V ,
detg
but we have l :

I We consider det g > O.

100 Chapter 1

o;~detg
-1- o.det g = --'--;~-
2detg' ~detg

and thus

or
divv = ~ o;(v' ~detg). (1-132)
.ydetg

Given orthogonal Cartesian coordinates x; we find the usual definition of the divergence:
ov
divv=-' , i = l, ... ,n,
ox;
that is tr(grad v) or the contraction of grad v.

Example. In a pre-Euclidean n-dimensional space we have for any real-valued function f

and any vector field v:
div(f v) = f div v + v . grad f (1-133)
since

S.6.2b Divergence of a Tensor Field

We consider tensor fields whose last index is contravariant and we say:

D The divergence of a tensor field t of type (~) is the tensor field of type (q;!)
resulting from the contraction of the gradient of t in the last two indices; that is, at
any point x:
divt(x)=Okt(x)·e'k =Dek t(x)·e'k (1-134)
<=>
(1-135)

Example. Given a tensor field t = tUe; ® e j let us express div t at any point x.
We have:
 Tensors 101

It is also
dZ'vt(x) -- a k t Ue i iO.
'CJ
e} iO.
'CJ
e*k .• (e P iO.
'CJ
e*P) -- a k tUe i Us:kpUs:P
}

ip
-vi
_'<7
ei ·

By recalling [Eq. (1-101)]:

the contraction (in q andj) leads to the following expression for the divergence:
V q t iq = a q t iq +rihq t hq q t ih
+rh q' (1-l36)

Given a system of orthogonal Cartesian coordinates Xi we write:

at
(divt). =~.
, ax p '
for example:
at at at
(divt)x = ;: + ;; + a;z.

5.6.3 Curl

5.6.3a Curl of a Covector Field

Given any covector field a of components a /u 1, ••• ,u n ), we wonder if there exists

a tensor field whose components with respect to a curvilinear coordinate system (u i ) are
aa} .
au i

Given a change of coordinates u'} = u'} (u i ), i, j = 1, ... , n,

we have:

and thus
aa} au,k au,m aa~ a 2u,k ,
--=-------+--.-.a.
au i au} au i au,m au} au' k

They would be components of a tensor if there was no second term in the right-hand
member; but we may eliminate this right hand member term causing difficulty.
102 Chapter 1

Indeed, the differences

aa } aa i _ au.Ik au 1m (a
a k a am) I I

au i - auj - au} au i au 'm - au 'k

constitute the components of a tensor.

D The curl of a covector field a = a j e' j is a tensor of type (~) such that
curia = (aPj -ajai)e' i ®e· j
= I (aia j -ajaJ(e' i ®e· j -e· j ®e' i ) (1-137)
I<j

= I (aia j -ajai)e' i /\ e· j .
I<j

This curl is obviously expressed as:

1 .'
curla=-(ap)-aJaJe /\ e *' J
2! (1-138)
= a,aj e'! /\ e').

The reader can make explicit all the previous expressions of curia, which is an element
of the exterior algebra but which may be considered as an element of the tensor algebra.

Remark. From the symmetry property of the Christoffel symbols we note that the
components of curia are

We are now going to find the adjoint of the curl on a 3-dimensional space and find again
the curl's expression in elementary geometry.

The adjoint of curia in a 3-dimensional space is a tensor of type (~) whose components
are

that is, explicitly:

 Tensors 103

and so on.

Remark. The definition of curia is also written from the well known (~}-tensor p,
namely:
curia =p : grada, (1-139)
since
(curl a)k =.L2r
IIkU (aa
1 J
. - aa)
J 1
= .L2 rIIkij a.aJ. _.L2rII kji a.a j
1 1

= pkij aia j .

In orthonormal rectilinear coordinates Xi, we write:

aa
(curlah = Gkii _J ;
" 8:/
that is, for the x-component:
aa aa y
(curia) =__ z _ __
x ay az'
and so on.

5.6.3b Curl of a Tensor Field

We can also define the curl of second order tensors. For instance, the components
of a tensor of second order t with respect to an orthonormal rectilinear system are

(L curlt)ij = Girs artSj

or

5.6.4 Laplacian

5.6.4a Laplacian of a Function

Given a real-valued function fwe say:

D The Laplacian of f is the divergence of the (vector field) gradient of f

I':.f = div(grad f) . (1-140)
104 Chapter 1

So in local coordinates the divergence of the tensor field of contravariant components

gY ajl is the Laplacian

(1-141)

This formula lets us find again the classical expressions of the Laplacian in curvilinear
coordinates.

In rectilinear coordinates, det g being constant, we have:

and in an orthonormal rectilinear coordinate system we have:

Example. Let us calculate the Laplacian of a function 1 in cylindrical coordinates

e
u 1 = r, u 2 = and u 3 = z.
The components of grad 1 are

a21-- ae
al '

From the following conjugate tensor

we deduce the corresponding contravariant components gij ajl:

8f
grad 11 = - ,
1
grad 2 1 = - - ,
al al .
grad 3 1 = -
& r 2 ae &
So, since
Jdetg=r,
we obtain:

N la al aral a al
= -[-(r-) + - ( - - ) +-(r-)]
r ar ar ae r2 ae az az
a21 1 8f 1 a21 a21
=--+--+---+--.
ar2 r ar r2 ae 2 az 2
 Tensors 105

5.6.4b Laplacian of a Vector Field

Given a vector field v we say:

D The Laplacian ofv is the divergence of the «( ~ )-tensor) gradient of v:

Av = div(grad v). (1-142)

Since

the components of the Laplacian are the following [See Final Remark]:
(AvY = 8 p (gPl 8) Vi). (1-143)

In orthonormal rectilinear coordinates we have:

We also denote:
(1-144)

For example, the x-component is

(Av) = 8 2 Vx + 8 2 Vx + 8 2 Vx
x 8x 2 8/ 8z 2 '

and so on.

Final remark. Roughly speaking, we know that the gradient operates as follows:

(~)~(P:I)'
whilst the divergence operates as follows:

But when we define the Laplacian of a (~)-tensor f we introduce the contravariant

components of grad f; so we consider the divergence of a (~)-tensor (not of a (~)­
tensor).
Likewise, when we define the Laplacian of a (~)-tensor v, we consider the divergence of a
( ~) -tensor (not of a ( : )-tensor).

So we note that the Laplacian 'preserves the tensor type'.

106 Chapter 1

6. EXERCISES

Exercise 1.

Given an n-dimensional vector space solve the equation

'" i j k '" j k i-O
uijxxx -ujkXXX - .

Answer. In an explicit manner the equation

L (Xi)2 Xk - L (Xl)2 Xi =0
j

Thus the solution is

since L (XI)2 = 0 is equivalent to the particular solution Xl = X2 = ... = xn = O.

Exercise 2.

We consider a tensor t of type (~) of components

(1]=1, (]2=0, 121 =0, 122=2

with respect to a basis (e] ,e 2 ) •

Given a change of basis ej = a~ ei defined by

express the 'new' components oft in function of the 'old' ones.

Answer. We have:

= (1)(1)(1) + (-1)(-1)2 = 3 ,
 Tensors 107

Exercise 3.

We consider two frames of right handed unit vectors {o;e\,e2,e3} and

{ o;e; , e;, e; } such that the angle between horizontal vectors e; and e\ is (e;, e\) = + 1l/3,
e; has the horizontal bisector direction of the angle (e\,e 2 ) = lr/3 and e3 has the bisector
direction of the angle (e;,e;) = +lr/2 in the vertical plane {e;,e;}.

(i) Express the matrix (/3!) of change of basis ei = /3! e~ .

(ii) Given any vector x = xie i express the primed components x 'j in function of the
unprimed components x' .
(iii) Express the primed components in function of the unprimed components.

(iv) Given a covector W = OJ,e" express the primed components of w in function of

unprimed components and prove that
OJiX i = OJ~x'j = (w,x) =w(x).

- - --
,
,,
, , -, __e2 \
\
\
I
\I
,
I e'l.
/
/

I
e,
------
Fig. 4

Answer. (i) We immediately have

e3 =.fi
2 2
e' + .fi
2 3
e'

and thus
108 Chapter 1

-fj

(P!{!
T
-fj
T

T
-fi :1
-fi
T

(ii) The matrix (a~) such that e~ = a~ei is the inverse of the previous matrix:

-\ 0

(a~)= 1
-fj -fj
\ 0

-I -\ 2
-fj -fj-fi

and we have:

(iii) The matrix associated with the expression for primed components as functions of
unprimed components being the transpose of the matrix associated with the unprimed
basis vectors in function of primed, we have:
XII = lXI _lx 2 X'2 = .J3 Xl +.J3 x 2 +..fi x3 X ,3 _ ..fi x3
2 2' 2 2 2' -2

(iv) We know that the primed components of OJ are expressed as functions of unprimed
components as

that is, explicitly:

and we immediately verify the following

Exercise 4.

Given a change of basis e~ = a~ei express the primed components of a tensor t of

type (~) in function of unprimed components.

Answer. For any covectors OJ and f.J we have

 Tensors 109

I(w,p) = I(W',p')

but

and thus we have by comparison:

t'rs = [3: [35 tij .

Exercise S.

For every (~)-tensor of components aij and vector c of components c i the real
r = aijcic J can be viewed as r = byc'c), where the various bij are the components of a
symmetric tensor.

Answer. From

we deduce

and thus
r = bijc ic)
where the various
bij =±(aij +aji)=bji

are actually the components of a symmetric tensor.

Exercise 6.

Given a two-dimensional vector space E, we consider the following tensor of type

U -- uij e *1,0..
'Of e
*J
E E*,o.
'Of
E* •

Find the condition in order that u is the tensor product of two given covectors a = G;e*i

and b = be*J
) .
110 Chapter 1

Answer. The condition u = a @ b is satisfied iff

U ll = alb l , ul2 = a lb2 , u 21 = a 2bl , u22 = a 2b2 .
If all of uiJ are different from zero (and thus all of a; and bj are so), the previous system
is equivalent to

and the required condition is

This condition, namely a lbla 2b2 = alb2a2bl' IS necessarily satisfied even if any uiJ
vanishes.
In this last case the required condition is also sufficient. For instance, if U ll = 0 then
u l2 = 0 or u2l = 0 .
If Ul2 = 0 then a l = 0 necessarily; if U21 = 0 then bl = 0 necessarily.
The reader can write u = a @b in these particular cases (and also for U 22 = 0).

Exercise 7.

Prove that the following components

if i = k, j = I, i *- j,
if i=l,j=k,i*-j,
o otherwise
are such that:
(i) (1-145)

(ii) Show they are the components of a (; )-tensor called the extended Kronecker tensor
which is anti symmetric with respect to its covariant indices and its contravariant
indices.

Answer. (i) In the case {i *- j, i *- k} (~ O;k = 0) we have:

if j = k, i = I,
okl = -1
lj 0 otherwise.

In the case { i *- j, i = k } (~ 0; = 0 ) we have:

1 if j = I,
okl =
lj 0 otherwise.
 Tensors 111

In the case i = j we obviously have:

ojl = O.

Thus (i) is proved.

(ii) With usual notations we have:

t5'rs
pq
= otY 1515
P q
_ t5 qtY 1515
P
_ ; fir s:k jfiS s:1 jfir s:k ; fis d (1)
- a p ku;a q IUj-a q kUja p IU;

; jfirfiS(s:ks:1 s:ks:l)
=apa q k I u; Uj -Ui U; .
_ ; ifirfis s:kl
-apaqklui}

and thus the 'extended Kronecker delta' is a tensor of type 0).

From (1) we emphasize that
s:tYs k fir I fis kfir I fis s:r s:s s:r s:S
u pq =a p kaq I - a q kap I =upu q -uqu p

= t5~~ .
Clearly, the previous tensor is skew-symmetric with respect to its contravariant indices
and its covariant indices.
The reader will immediately define the generalized Kronecker tensor of components
15:'1··.:'1' •
'] ,. J"

Exercise 8.

Prove that the sum 0 f ( I, ) -tensors t = t i} k e; to.

'01 ej to.
'01 e *k and u = u ij k e; '01
to.
ej to.
'01 e *k
is a tensor of the same type (~).

Answer. From
('i}k =fi~fi:a;(pqr,

u'i) k = fi~fi: a; u pq r
we deduce:
(t'i) k + u,ij k ) = fi~fi: a; (t pq r + u pq r)

and thus t + u is actually a tensor of type (~).

112 Chapter 1

Exercise 9.

Given the tensors /i and t with respective components /i! and t pq r, express the
contracted product /i. t .

Answer. Since /if are components of the Kronecker tensor the contraction leads to the
( ~ )- tensor of components

Exercise 10.

Prove that the contraction of the tensor product of t = t P e P and u = u q e*q IS a

tensor of type (~).

Answer. By considering
t®u= t P Uq e P ® e *q ,
a change of basis leads to

and the contraction implies:

=tPu p.

Exercise 11.

By contracting a tensor with components tPqrsh in p and h, and next in q and s,

prove that the result is a covector.

Answer. First, by summing over the repeated index p = h , we have

and so we have obtained a tensor of type (;), denoted by

 Tensors 113

Second, by summing over the repeated index n = b, we have:

,}.
U Ig = j3}n aka}
a b n s:b a n
U ab = una k U ab

= a ka U b ab.
We have actually obtained a covector.

Exercise 12.

Given two tensors t = tij e*; Q9 e*} and u = uk! e k Q9 e! prove that

t :u = tr(t· u) = u : t
and
t :I = tr(t).

Answer. The contracted product t· u is the following second-order tensor

The only supplementary contraction leads to the scalar

but the trace of the tensor t· u is obviously the scalar following from the contraction in
the indices i and I; that is:
tr(t· u) = ("U Ji = t :u .
We immediately have
u .• t = uij ( JI.. = (u
1J
JI

= t: u
In addition we have:

Exercise 13.

Given a change of basis of a Euclidean vector space

express the components gij as functions of the various gij'

114 Chapter 1

Answer. The change of basis e~ = a~ei is such that

a~ = -2,
and thus
, I I I 2 2 I 2 2
gil =alalg ll +ala l g12 +a l a l g 21 +a l a l g22
= 4g11 - 4g 12 + g22 '

g;2 = g;1 = -2g ll - 3g 12 + 2g 22 ,

g;2 = gIl + 4g12 + 4g22 .

These results can be verified from the definition g~ = e;.< ; for instance:
g;2 = (-2e l +e 2 )·(e l +2e 2 )
=-2g I1 +g21 -4g 12 +2g 22 ·

Exercise 14.

Prove that Vro,p E E' :

Answer. We have:
g -I (ro, ) . g . g -I (p., )
= (gij OJ je i )· (grse,r ® e'S). (gpq f.1 qe p)
_ ij sq _ s:q ij _ qj
- g OJ jgisg f.1 q - Vi g OJ jf.1 q - g OJ jf.1 q

= OJ .gjSe .f.1 e'q

J s q

Exercise 15.

Given a vector a = a i ei and a basis (e~) such that

(i) Find the contravariant and covariant components of a with respect to (e~).

(ii) Given a = e l + e2 + e 3 and b = 2e l + e 2 - e 3 , from the contravariant and covariant

components of a and b with respect to (e~) calculate the scalar product a.b.
 Tensors 115

Answer. (i) From

a = Aie; = Aiel +A2(e l +e 2 )+A 3(e 2 +e3)
=(A I +A2)e l +(A 2 +A3)e 2 +A3 e3

we deduce
a l =AI +A2, a 2 =A2 +A 3 , a 3 =A 3
and thus the contravariant of a with respect to (e~) are

The covariant components of a with respect to (e~) are

A I = a .e l' = a i e i .e l = a I ,
' =ae
A 2=a.e 2 i (
i .e l +e2 ) =a I +a,
2

A3=a 2 +a.
3

They are also obtained as follows:

with
gl2 =g21 =1, gl3 = g31 = 0,
g23 = g32 = 1, g33 = 2.
So, we find again
AI = gliA' = Al + A2 = a l ,

A 2 = g2i A ' = Al + 2 A2 + A 3 = a I + a 2 ,
A3 = g3i A ' = A 2 + 2A 3 = a 2 + a 3 .

(ii) From
Al =1 , A2 =0, A 3 = 1,
B I =0, B2 = 2, B3 =-1
and
AI =1, A2 =2, A3 =2,
BI =2, B2 =3, B3 = 0,
we deduce

Of course this result can be directly found.

116 Chapter 1

Exercise 16.

Given a two-dimensional Euclidean vector space let t = t/ e*i ® e j ® e k be a

tensor of type (n such that
tIll = 1, t/ 2 = 1, t/ I = 1, tl22 = 2,
t/ I = 1, t212 =2, t 2 21 -- 0, t/ 2 = -1.
(i) Calculate the contraction over first and second indices.
(ii) Given the fundamental tensor g such that

determine the components tijk .

(iii) Calculate the various components t ij k •

Answer. (i) The contraction leads to a tensor of type (~) of components uk = t/k which
are:
U
I _ t II
- I
+ t 2 21 -1
- ,

U
2 _ t 12
- I
+ t 2 22 -
-
0.

Thus, this tensor of type (~) is e I .

(ii) Since the conjugate tensor has the following components

we have:
t lll- I l t l l + 12 t ll_3
-g I g 2 - ,

t l12 -g
_ lIt 12
I
+ g 12 t 2 12_5
-
and so on:

(iii) From

we deduce:
 Tensors 117

and so on:
] 1 1 2 2 5
to]
-
=--,
3 t22 = -, t22 =-
3 3

Exercise 17.

By considering a transformation of curvilinear coordinates

R n ~ R n : (u i ) H (u'i)

and the following tensors

a=ai'ke®e*i®e*k
J , ,

calculate
a:b.

Answer. The change of components of a ® b is such that:

a' P r blS t a ' P _u

= _u_, a i a k a 's a m
u __u_ _u_ a i k b'm.
___
q au' au,q au" QU' QU,t J

A first contraction in rand s leads to

Qu 'P Qu i QU m s:k i, bl
a , P qr b" t -
- - - , - - - - v , a Jk m.
QU' QU,q QU,t

A contraction in p and t leads to

a , Pqr b" P QUi s:m s:k i bl

=- - a i,Jk bk i .
QUi
- v i v I a jk m = -
Qu,q au,q

This proves that a : b is a tensor of type (~).

Exercise 18.

Given the fundamental tensor g as well as two vectors x = x' e i and y = yJ ei , find
the expression of the double contraction (x ® y) : (e k ® el ).
118 Chapter 1

Answer. From

we deduce:
(x®y):(e k ®ez)=xiyj(e i ®e):(ek ®ez)
= Xi yj (e j"ek)(ei.e Z)
i j
=gjkgi/ X Y
=xZYk
= (x.e z)(y.ek)·

In particular, if x =ei and y = ej we have:

(e i ®e): (e k ®e z) = (ei.e Z)(e j"e k ) = giZgjk.

Exercise 19.

Prove that

Answer. We have:
ijk _ Ijk 2jk 3jk
£: Cli C2j C3k-£: CllC 2j C3k+£: CI2C2jC3k+£: CnC2jC3k
= Cll (C 22 C33 - C23 C32 ) + Cl2 (C 23 C31 - C21C33) + Cn (C 21 C32 - C22 C31)

that is

Exercise 20.

Given x = Xi ei and y = yje j of a Euclidean vector space show that the expression
gijxiyj

~x\~yjYj
is a tensor of type (~) and give its interpretation.

Answer. From
 Tensors 119

and
X'Py'P =aifJP i xy/ =X.y/
xy / =t5JI
PJI J'
we deduce that g ijXi y/ is an intrinsic scalar that is x. y ; and so are

~XiXi =.Jx.x =llxll

and
~y/ y/ =Ilyll·
The introduced expression is the cosine of angle (x, y) .

Exercise 21.

Given two vectors y = yi e i and z = z/ e / of a Euclidean space E, we consider the

tensor of type (~):

that is,
yi
Yl\z=L Ii
i</ Z

(i) By introducing the tensor of components tij give an interpretation of 1tijtij with
respect to the paraileiogram of sides corresponding to vectors y and z.
(ii) Determine the contracted product tim xm between y 1\ Z and any x E E.

Answer. (i) We know that

tij =giPg/qt Pq =gipg/q(Y Pzq - yq zP) =YiZ / - Y/Zi

and, of course, we have the double contracted product:
1tijtij = 1(YiZJ - yjz)(y'Z j - yJ Zi)
= t (y). t(iJ)

that is, a tensor of type (~).

If we consider an orthonormal basis, this scalar (being independent of the basis) IS
expressed as
twiij) = L
(tij)2 = L
(yi z/ - yJ Zi)2 ,
i<j 1<)

which is the square of the area of the parallelogram having adjacent sides corresponding to
y and z.
120 Chapter 1

(ii) For any x = e the contracted product

Xi i ,

== (z.x)y - (y.x)z
defines a vector.
In the usual Euclidean space, it is the well-known formula of the triple product:
x A (y A z) == (x.z)y - (x.y)z.

Exercise 22.

Given two I-forms a = aiO i and P = PjO j show various expressions of a AP .

Answer. The tensor product

a ®p = aiP/)i ®oj
isa (~)-tensorsuchthatthe n 2 components relative to (Oi ®OJ)of T20 are aiP j .

The exterior product a AP is an antisymmetric tensor of T20:

a AP ==aiPjO i A()j =aiPjO i ®()j -aiPj()j ®Oi

= (aiP j -ajPi)Oi ®O j

i<j f?:.j i<j i75,j

To sum up:

i<j

Exercise 23.

h C p pro ducts () it
Prove t h at ten A •.. A °ip (.'I < ... < I•P ) £orm a baSls
· 0f AP .
u
 Tensors 121

Answer. We have previously seen that to eachp-form was associated ap-form 'generated'
by C~ elements (0 il A .•. AO ip ) such that i l < ... < i p .
Thus it is sufficient to show that these C~ elements are linearly independent, that is:

~liJil ... ip O'l A ... A lip = 0 OJ .

'\""p
. =0
i\< .. <lp

Let (jl < ... < jp) be a sequence which is completed in order that (jp ... ,jP,jp+I, •.• ,jn) be
(1, ... , n). The order of terms will be changed if necessary.
The assumption implies
"
£..." -0.
liJil .. .ipOil A .•. A Oip A o Jp+1 A ..• A oJn -
it <..<ip

In this sum only

OJI A ••. AO Jp AO Jp +1 A ..• AO Jn

is nonzero, the other terms having necessarily two equal factors.

The previous sum is thus reduced to only one term, namely:
liJJ' J' OJI A ••• AO Jp AO Jp+1 A ..• AO Jn
I··· p

which is necessarily zero.

Thus we conclude:

Exercise 24.

Given the 3-dimensional Euclidean vector space E, prove that

(i) 8 yk 8kqr = OjqOjr -OjrOjq'

(ii) 8tjk8kjr =-20 jr ,

(iii) 8ijk8ijk = 6,

(iv) Vx,y,z E E: x A (y A z) = (x.z)y - (X.Y)Z.

Answer. (i) Let us give the following interesting demonstration by introducing the
determinants:

and thus
122 Chapter 1

a jp a,q air alp a 1q

a jp a jq a jr = E:Uk a 2p a 2q

a kp akq a kr a 3p a 3q

We deduce the following special result:

Ojp Ojq Ojr

E: ijkE: pqr = 0jp 0jq °jr .

°kp °kq °kr

Thus the summation over the index k = p in the previous interesting result leads to

(ii)

(iii)

(iv)

=(x.z)y-(x.y)z.

Exercise 25.

In classical Euclidean plane geometry we consider a system of elliptic coordinates

(u l ,u 2 )== (A,B) which are connected to Cartesian coordinates as follows:

Xl = a cosh A cosB, x 2 = a sinh A sinB

(i) Show the natural frame at any point; that is, {x; e l ,e 2 }, or simply { x; e i }.

(ii) Express the differential vectors de I and de 2 in function of natural basis vectors.

Answer. (i) By denoting Xl the Cartesian coordinates of x relative to a fixed frame

{o;er }, the position vector ofx is
 Tensors 123

= a(coshA cosO ef + sinhA sinO en.

So, there are two coordinate lines through x which are:
- an hyperbola corresponding to a given value of 0,
- an ellipse corresponding to a given value of A.

The natural basis vectors are

e1 = ax = a(sinhA cosO e: + coshA sinO e;),

aA
e2 = ax = a( -cosh A sinO e: + sinhA cosO e;).
ao

e1

----;-~~--~O+-----+--+--~-X1

Fig. 5

We note that the Jacobian

SinhA cosO COSA sinO I 2 2
J =I = sinh A + sin 0
- coshA sinO sinhA cosO

is different from zero except when (XI,X2) is (-a,O) or (a,O).

(ii) We have:

de = ae1 dA + ae1 dO
1 aA ao
= a(coshA cosO e: + sinhA sinO e;)dA + a( -sinhA sinO e: + cosh A cosO e;)dO
=a(COShA cosO dA - sinhA sinO dO)et + a(sinhA sinO dA + COShA cosO dO)e;
and in the same manner:
de 2 = -a(sinhA sinO dA + cosh A cosO dO)e: + a(coshA cosO dA - sinhA sinO dO)e; .
124 Chapter 1

But from
ae~ =7(sinhlcosOe 1 -coshlsinOe 2 ),

ae~ =~(coshlsinOel +sinhlcosOe 2)

J
we deduce:

del = ~ sinhl coshl (dl e l + dO e 2 ) + ~ sinO cosO (dO e l - dl e 2 ),

J J

Exercise 26.

Let us consider a co vector of Cartesian components

I 2
t3 =X X

Express these components as functions of spherical coordinates.

Answer. We know that

axi
t~ = - - k t,
au'
with
ud = r, U,2 = 0 and U,3 = ¢ such that:
Xl = u d sin U,2 cos U,3,
x2 = u d sin U,2 sin u d ,

x3 =udcosu,2.

Thus, for instance, we have

t'
axl
=--/ +--( +--(
ax 2 ax 3
I auf! I auf! 2 auf! 3

= (sinu,2cosu,3)xlx3 +(sinu'2 sinu'3)2(x 3)2 +COSU'2XIX 2

= r 2sin 20 cos 2 ¢ cosO + 2r 2sinO sin¢ cos 2 0 + r 2 cosO sin 20 cos¢ sin¢

Exercise 27.

Find the Christoffel symbols for spherical coordinates,

(i) from the second Christoffel formula,
(ii) from the expressions de i •
 Tensors 125

Answer. (i) From the expressions for the natural basis vectors e l ,e2 and e3 [see Example
2 of Section 5.1.3], with spherical coordinates u l = r, u 2 = e, and u 3 = ¢, we deduce:

and

So we have:

and

The conjugate tensor is

o 0

o o
o o

We have successively:

r;3 = -sinB cosB,

126 Chapter 1

(ii) Without introducing the fundamental tensor, from the following expressions

del = (cose cos¢ ef + cose sin¢ eg - sine en de

+ (-sine sin¢ elo + sine cos¢ en d¢,

de 2 = (cos e cos ¢ e f + cos e sin ¢ e g- sin e en dr

+ r( -sine cos¢ ef - sine sin¢ eg - cose en de
+ r(-cose sin ¢ ef + case cos¢ en d¢,

de3 = (-sine sin¢ ef + sine cos¢ en dr

+ r(-cose sin¢ ef + case cos¢ en de
+ r( -sine cos¢ ef - sine sin¢ en d¢,
we deduce:

de 3 = dr e3 +cot/J de e3 -rsin 2 e d¢ e l -sine case d¢ e 2 •

r

But we know that

and by identification we obtain, for instance:

and thus
2 1 3 1
TI2 =- T\3 =-
r r
I
Til I
=T12 =T\3 = T2II = T213 =TII3 =T12
I 3
=0

and so on.
 Tensors 127

Exercise 28.

From the first Christoffel formula prove that

(i)

in the case of systems of orthogonal coordinates.

(ii) In particular, again find r13 and rJ3 in spherical coordinates.

Answer. (i) We have

(ii) We have:
3 3 1 1 2.2
r 23 =r32 = - 8 2 lng33 = - 8(J In(r sm 0) = cotB,
2 2

rJ3 = _ _
1_ 8 2g 33 = -~ 8(J(r2 sin 2 0) = -sinO cosO.
2g 22 2r

Exercise 29.

A particle follows a circle of radius R with the constant angular speed iJ = mo .

From the notion of absolute derivative find the particle acceleration.

Answer. By considering the polar coordinates u l = r, u 2 = () the only nonzero Christoffel

symbols are
ri2 =-R,
The acceleration components
, d 2 u' ,du i du k
a =--+r'k - - - -
dt 2 J dt dt
are, for u I = r :
du 2 du 2
I
a =r22
I
---- = -RO• 2 = -Rmo2
dt dt
and for u 2 = 0 :
128 Chapter 1

Thus the acceleration is directed towards the circle center and its norm is proportional to
the distance from the center.

Exercise 30.

Prove that any geodesic (in a surface of classical Euclidean space) is such that at
each of its points the osculating plane is orthogonal to the surface.

Answer. We know that every osculating plane equation is determined by the first and
second derivative vectors, and that the geodesic equations are
d 2u J du k du J 1 du J du k
g - - + 8 g. - - - - 8 g . - - - 0
ij ds 2 k ij ds ds 2 i Jk ds ds -

But we have:

8 k(e"e J )-~8i(eJ,ek) = (ei'8 ke J )+(e J ,8 ke i ) -~(eJ,8iek )-~(ek,8ieJ)

2 2 2

= (e i ,8 ke J ) +~(eJ,8iek )-~(ek,8ieJ)
2 2

and thus the geodesic equations are written:

(1)

Let us consider the following vectors

dx du J
-=--e
ds ds J
and
 Tensors 129

The geodesic equations (1) mean that at each point of the geodesic the second derivative
2
vector -d2x- IS
. perpend'lCUIar to vectors e;.
ds
Thus the osculating plane is orthogonal to the surface.

Exercise 31.

Let v be a vector field on a point space.

(i) Prove that the expressions
(dv); = dv; +VkW~

are the components of a tensor of type (~) (called the absolute differential of v).
(ii) By considering a system of curvilinear coordinates (u i ) prove that the expressions
i i j k
V'jv =v,,+rjkv

are the components of a tensor of type (:) (called the covariant derivative of v ).

Answer. (i) Given a change of basis e~ = aje; we consider

The components of the absolute differential with respect to (e~) are

In order to express w~q in function of £ok we consider

d e'p-- d (ape
i )i --d ape;
i i j ej -(d'
+apwi - ki) ei
a p +apwk
= (d a pi + a pkWi) aq e'q
k P,

but
de'p = w,q
p
e'q

and thus
130 Chapter 1

Therefore we have:

From

we deduce:

Cdv),q =dC/3?vi)+IJ!:vna~CPiqm~ -dP%)

=dC/3?v')+v k C/3?mL -dP%)
= P? dv i + Vi dPiq - VkdPZ + Vk PiqmL

= P?Cdv i +vkmD
= /3?Cdvr·

(ii) In

let us make r~~ explicit by considering the system of curvilinear coordinates.

From (i) we know that

a ,q a i a k a k a,q a ,q
m,q =_u_Cd_u_+_U_mi)=_U_C_u_. m ' _d_u_ )
P au i au'p au'p k au'p au ' k auk
and thus
a ka ,q a 2 ,q
r,q dulY = m,q = _u_c_u_r _ U )du i
pr P au' P au i }k au} auk
",-.k a i a,q a 2 ,q
r,q = _u_u_u_c_u_ r _ U )
pr au' P au" au i }k au} auk .
Thus, we have
a ka a,q a i a,q a 2 ,q a"
V' v,q =_u_[ __ C_U_Vi)+_U_c_u_ri U )_u_ vn ]
P au' P auk au i au" au }k au} auk au n
i

auk a 2 u,q . au,q avi au,q· a 2 u,q .

= au'p [aukau i v' + aui auk +C aui rjk - aukaui)v}]
a ka,q
= _u
::>..i
__u_c_u_v+r vi)
au'p au i auk }k
ka
a __u_ V' v'
,q
= _u
au'p au i k

and they are really the components of a tensor of type (:).

 Tensors 131

Exercise 32.

(i) Given a 3-dimensional pre-Euclidean vector space and given any two covectors
x = xie*i and Y = yje*j, if we consider the adjoints of x /\ y, curlx and curly prove
the formula
div(x /\ y) = curl x .y - x .curly .
(ii) Given any two vectors v and w of the 3-dimensional vector space E, directly obtain
the formula, namely:
\tv, w: div(v /\ w) = curlv.w - v.curlw.

Answer. (i) The adjoint of the (n-tensor X/\Y of components xiYj-XjYi is the vector
of components

[;(ij)k
= (xy -xy)
~Idetgl I } } I

and thus we have:

d· ( ) 1 8 ( (ij)k ( »)
IV x/\Y = ~Idetgl k [; xiYj -XjYi

1
= ~ [8 1 (X2Y3 -x3Y2)-8 2(x 1Y3 -x3Yl)+8 3(x 1Y2 -X 2Yl)]
"Idetgl
I
= ~ [Yl(8 2x 3 -8 3x 2 )+ Y2(8 3x 1 -8 1X3)+ Y3(8 1x 2 -8 2 x 1)
"Idetgl
-x 1(8 2Y3 -8 3Y2)-x 2(8 3Yl -8 1Y3)-x 3(8 1Yl -8 2 Yl)]'

This is really equal to

curlx.y - x.curly
since the following

implies
132 Chapter 1

and in the same manner:

(ii) Given any two vectors v and W of the Euclidean vector space E, the divergence of
their vector product

is written:
div(v /\ w) = &uk(VjWk),1
=&Uk(Vj,iWk +VjWk)
= &kijVj,iWk -Vj&jik Wk,i
= (curlv)k Wk - Vj (curlw)i'

Exercise 33.

Given the 3-dimensional Euclidean vector space E prove that the gradient of the
product of functions f : E ~ Rand g : E ~ R is
grad(fg) =! gradg + g grad!.

Answer. It is obvious because

Exercise 34.

Given the 3-dimensional Euclidean vector space E prove that the divergence and
curl of the product of f: E ~ R and any vector v are such that
(i) div(fv) = f divv + v.grad!,
(ii) curl(fv) = gradf/\ v + ! curlv.

Answer. It is obvious because

(i)

(ii)
 Tensors 133

Exercise 35.

Given any vector v of the 3-dimensional Euclidean vector space E prove that
div curl v = O.
Answer. From
curl v = E:Yk vk.} e i ,
we deduce
divcurlv = (E:Yk vk,)" = E:ijk Vk,}i' (1)

Since this is also equal to

(2)

and because E:"k = -E:}ik the comparison between (1) and (2) leads to the expected result.

Exercise 36.

Given the 3-dimensional Euclidean vector space E prove that:

curl(gradf) = 0,
for every function f: E ~ R.

Answer. Since

is also written:
(2)

Because E:yk = -E:ik} the comparison between (l) and (2) leads to the expected result.

Exercise 37.

Given a vector field v in the 3-dimensional Euclidean vector space E prove that
Av = grad divv -curl curl v .

Answer. From
divv = v},} ,
we deduce that
134 Chapter 1

From
curl V = B pqr Vr,q e p'

we deduce that
curl(curlv) = Bijk (curlv)k,j e ,

=Bijk(Bkqr Vr,q),jei
= BljkBkqr Vr,q) e i
=(t5iq t5jr - t5ir t5;q)vr ,qj e i

Thus we have

which is actually

Exercise 38.

In the 3-dimensional Euclidean space E let

2 1
0
.16 jj

[:} -.16

-.16
1

1
jj

jj
1

1
--J2
1
1

-J2
[::1
be a change of orthogonal Cartesian coordinate systems.
(i) Given the function
f: E ~ R: (x.) ~ x]2-xi,

calculate the gradient of f with respect to both coordinate systems.

(ii) Verify that the image of gradf under the previously defined change of coordinate
systems is actually the gradient of the function with respect to the primed coordinate
system.

Answer. (i) Since the change of coordinates x' =a x is such that a-I = t a, we
immediately have
 Tensors 135

The gradient offis immediately

gradf = 2XI el -2X3 e3 ·

Since the given function is expressed in the primed coordinate system as

f ' : E ~ R : (Xl'X
, ,2 ,X3, ) H 1 (2'
-
,
Xl -x 2 -X3
, )2 1 (-x ,2 +X3, )2 ,
--
6 2
that is:

then the gradient of the function with respect the primed basis is the following vector

grad'f'=

(ii) Since a grad f is equal to

4 , 1
-x
3 1 J3
1 1
-J6 J3
1 1
-J6 J3
this image of the gradient under the change of coordinate system defined by the matrix a
is actually the gradient of the function with respect to the primed basis:

4 , 2 , 2 ,
-x --x --x
3 I 3 2 3 3
2 , 2 , 4 ,
--x --x +-x
3 I 3 2 3 3
2 , 4 , 2 ,
--x +-x --x
3 I 3 3 3 3
136 Chapter 1

Exercise 39.
Calculate the Laplacian of a real-valued function I in spherical coordinates u l = r,
u2 = 19 and u3 =¢.

Answer. Since
detg = r 4 sin 219,
r2 '

Eq. (1-141) is written as follows:

M = -2-1. - [a l (gllall r2 sine) + a 2(g22ad r2 sine) + a 3 (g 33 ad r2 sinB)]

r sme
= _2_1-.- [a,(aJ r2 sinB) + ae(ael sinB) + a¢(a¢1 _._1_)]
r smB smB
=...!..-~(r2 at)+_I-~(sinB al)+ 1 a 21
r2 ar ar r2 sinB aB aB r2 sin 2 B a¢2
a 2I 2 at 1 a 2I coW at 1 a 2I
=--+--+---+----+ --.
ar2 r ar r2 ae 2 r2 aB r2 sin 2 B a¢2

Exercise 40.

Given a vector field v find the expressions of div v :

(i) with respect to the natural basis (e r ' ee ,e¢) associated with the spherical coordinates,
(ii) with respect to the corresponding orthononnal basis (1,,Ie,I¢).

Answer. (i) Given

we know that
,
divv = a;v' + 2 ~ 2 a;(r2 sinB)
r sm 19

(ii) We know that

1
1 =--e
¢ rsinB ¢
and
 Tensors 137

Thus we have:
I I
v = vr 1r + - vele + - . - vI/> 11/>.
r rsmO
Because the corresponding components of contravariance and covariance are identical
with respect to this (orthonormal) basis (lr,Ie,Il/>), we deduce they are:

_ _I/> 1
vI/> =V = - . - vI/>
rsmO
and so the well known corresponding expression of the divergence ofv is
· I a-I a - cotO_
dIV v = arV- 2-
r + - BVe + - . - I/> v I/> + - v r + - - Ve
r rsmO r r

Exercise 41.

Find the gradient of a function f: R2 ~ R m the case of elliptic coordinates

ul = A, u2 = 0 and deduce the Laplacian !1f.

Answer. We know that the natural basis vectors associated with elliptic coordinates A
and 0 are
e j = a(sinhA cosO e; + coshA sinO e~),
e 2 = a( -coshA sinO e; + sinhA cosO e~)
and we have:

From
(gradf)1 - 1
- a 2(Sl·nh2 A1 + sm
(gradf) - 1 af
• 20) I -
a 2(Sl·nh2 A1 + sm
. 20) i

we deduce:
~detg (gradfr = aJ
138 Chapter 1

and thus

Exercise 42.

Calculate the components of the gradient of a real-valued function f and the

corresponding contravariant components (under l)
(i) relative to the natural basis (er,eo,e z ) of cylindrical coordinates,
(ii) relative to the basis of unit vectors (1r,1 0 ,1 z ).

Answer. (i) With respect to the natural basis (e"eo,eJ of cylindrical coordinates, the
components of covector grad fare

aJ, aof, aJ

and since gIl = g33 = 1 and g22 = 1/ r2 , we obtain:

(gradf)1 =glla1f=aJ,
2 22 1
(gradf) aJ = "2 aof ,
= g
r
(grad f)3 = g33 a3f = a J.

(ii) We know that the natural basis (e r ,eo ,e z ) with u l = r, u 2 = (), u 3 = z is such that

er =cos()ef+sin()e~, eo =-rsin()ef+rcos()e~, ez =e~

and
1
1r = e r , 10 =-eo ,
r

The components (grad f) i relative to (1 r ,10 ,1 z) are respectively:

1
aJ, -aof, aJ.
r
Before obtaining these components we emphasize that
df = (gradf).dx.
 Tensors 139

With respect to the natural basis we have:

dx=dre r +dBeo +dze z

and
df = 8 J du i = 8 J dr + 8 of dB + 8 J dz .

dx = dr 1r +rdB 10 +dzl z
and thus the corresponding components (grad f), are respectively:
1
8J, - 80 f, 8J.
r

Since (i,,1 0 ,lz ) is an orthonormal basis, they are the components (gradfY too.

Exercise 43.

Given the cylindrical coordinates r,B,z:

(i) Find the following Christoffel symbols with respect to the corresponding (orthonormal)
basis (lolo,1 z) also denoted (e "eo,e z):

-0 1 -r 1
rrO =-, roo =--
r r
and prove, for example:
-0
rOr = O.
(ii) Given (e r ,eo, e z), calculate the divergence of a tensor t of type (~).

Answer. (i) Weare going to calculate the previous Christoffel symbols with respect to
(l"lo,1z) from the known Christoffel symbols with respect to the natural basis
(er,eo,e z )·
Since in general
r~k =(8 e ,e*i),
k j

we have the following results for the particular basis (e r , eo, e z) [see Sect. 5.6.1 a]:

t:~ = (; 8 ir,(e* /) = (8 oer ,dB) = (8 oer,dB) = rr~ = ~,

I
140 Chapter 1

r:r = (8 reo,(e*)O; = (8J~ eo), rdB) = (- r\ eo + ~8reo, rdB)

= -~ (eo,dB) + (8 reo,dB) = -~ + r:r = O.
r r

(ii) From (1-136) expressing the divergence of a (~)-tensor we obtain the following
components relative to (e" ee , ez ) :
for i = r:

for i = B :

for i = z:
(divt) z = 8 rt zr +~8 e t Ze +8 Z
t ZZ +rree t zr
r
=8,tzr +~8etze +8/ zZ +~tzr.
r r

Exercise 44.

(i) Find the expression of the gradient of a vector field v with respect to the
orthonormal basis (1 r,J o,J J of cylindrical coordinates.
(ii) Deduce the divergence of v directly from gradv and next by considering the various
~~ associated with the orthonormal basis.

Answer. (i) We know that the gradient of a vector field v defined at each point (r,B,z) is
the tensor field gradv which associates the differential of v with the position vector
element dx as follows:
 Tensors 141

dv = gradv ·dx,
where
dx=dr1 r +rdBlo +dz1,.

Given an orthononnal basis (i,j,k) we recall that

Jr = -ax = cos BI +' Sin

· B'J,
ar
1o = -I -ax = . e'
-Sin 1 + COS
B'
J,
r aB
1z =k
and thus, we have:

-aIr = -Sin
. B'I + cos B'J = J0'
ar

-a10 = -cos B ' · B' 1

I - Sin J = - "
aB

all other partial derivative vectors being zero.

Thus the differential of v = vr 1 r + vol0 + v, 1, is explicitly:

(-aVr dr+-
aVr dB + av-z dz)J + (aVO
- dr+- aVe dB + avo
- dz ) 1
araB az ' ar aB az 0
av av av
+ (_l dr+ _ z dB + _l dz)l. + vr dB 10 - Vo dB 1,.
ar aB Dz 0

1
If we denote dx by the ",Iuron-vocto, [,;, the previou, definition dv ~ grod v . dx i,

explicitly written with respect to the orthononnal basis (1r,1o,lJ as follows:

av r .!.(avr -v ) av r
-aVr dr+ (av
_r -Ve )dB + - dZ
aVr
Dr aB Dz ar r aB 8 az

E:l
avo I ave ave
-aVe dr+ (aVO
-+v )dB + aVe
- dz
Dr DB r Dz Dr -;:(aB +Vr) az
aV' dr+_o
_ 0
avo dB +_0
DV' dz av z I av z av z
ar aB Dz Dr r aB Dz

where the tensor gradv clearly appears in the second member.

142 Chapter 1

(ii) The divergence ofv is tr(gradv) and thus immediately it is:

.
dIVV Ovr 1 Ov o vr Ov z
=-+--+-+-.
or r 00 r oz
We may also refer to Section 5.4 and consider (1-95).
The components dv' with respect to the orthononnal basis (1, ,1 0 ,1:), namely:

(Vi,k +~~vj)dUk
lead to the following tensor gradv since the only nonzero Christoffel symbols are
- 1 -0 1 I 2 3
F:e = - -, FrO = - and since du = dr , du = r dO , du = dz :
r r
o
-v
1 0 1
- - v --v
o
-v
or r r 00 r r 0 oz r

o
-v
1 0 1
- - v +-v
o
-v
or 0 r ao 0 r r oz 0
o
-v
1 a
--v
a
-v
or z roO z oz z

Exercise 45.

Given the spherical coordinates r, 0, rjJ and the corresponding orthononnal basis
(er,e(J,e¢) :

(i) Prove that the only nonzero Christoffel symbols are

-¢ _ -(J _ cotO
T(J¢ --F¢¢ - - -
r
(ii) Given (e"e(J ,e¢) calculate the divergence of a tensor of type (~).

Answer. (i) Since in general

T~k =(akej,e*i),
we have [see Sect. 5.6.la] the following results:
 Tensors 143

and so on.

(ii) From (1-136) expressing the divergence of a (~)-tensor, we obtain the following
components with respect to ( e" e e¢ ) :
(J ,

fori=r:

for i = B :

(divt) =a tllr+!a tllli+_l_a to;+rlltrll+rllt~+rlltllr+r¢tllr+r¢tllli

II r r II r sinB ¢ rll # rll r¢ 0;

1 1 1
= a r tllr +-a ll t lili +-.-a to; +_(3t rlJ +(t lili -t~}coW);
r rsmB ¢ r

for i = ¢:

(div t) = a t¢r +!a t¢li + _1_a t# + r¢ t r¢+ r¢ to; + rll t¢r + r¢ t¢r + r¢ t¢li
¢ r r IJ r sinB ¢ r¢ 0; rlJ r¢ 0;

1 1 1
= a r t¢r +-a lJ t¢li +-.-t~+-(3tr¢+2tO; coW}.
r rsmB r

Exercise 46.

(i) Find the expression of the gradient of a vector field v with respect to the
orthonormal basis {lr,llJ,1¢} of spherical coordinates.
(ii) Deduce the divergence of v directly from gradv and next by considering the various
r;k associated with the orthonormal basis.

Answer. (i) We know that the gradient of a vector field v at each point (r,B,¢) is the
tensor field grad v which associates the differential of v with the position vector elements
dx as follows:
144 Chapter 1

dv = grad v , dx,
where
dx = dr lr +rdB1o +rsinBd¢I¢,
Given an orthonormal basis (i,j,k) , we recall that

lr = ax = sinBcos¢ ; + sinBsin¢ j + cosB k ,

ar
Ie =~ ax =cosBcos¢i+cosBsin¢j-sinBk,
r aB

1 ax '''' ,' +cosI"J

1¢ =-,--=-sml" '" '
rsmB a¢
and thus, we have

ale =-1
aB r'

aIr 'eJ ¢' al¢ ,

-=Sln -=-smBl -cosBl
a¢ a¢ r e'

all other partial derivative vectors being zero,

Thus the differential of v = vr 1r + ve Ie + v¢ 1¢ is explicitly:

(aV, dr+ av, dB+ 8v, #)1 + (ave dr+ aVe dB+ aVe #)1 + (av¢ dr+ 8v¢ de+ av¢ d¢)1
ar ae 8¢ r ar ae a¢ e ar aB a¢ ¢

+ V,del e + v, sine#l¢ - vedel, + ve cosed¢l¢ - v¢ sinBd¢I, - v¢ cosBd¢le'

If we denote dx by the column-vector [ ~B 1

r the recalled definition dv = gradv ,dx
rsinBd¢
is explicitly written with respect to the orthonormal basis (I, ,Ie ,1¢) as follows:

aV r dr+ aV r dB+ aV r d¢-v de-v sined¢

ar aB a¢ e ¢

av av av
_e dr+_e dB+_e #+vrdB-v¢cosB#
ar ae a¢
av av av
_ ¢ dr + _¢ dB + _¢ d¢ + vecosB d¢ + V rsinB d¢
ar ae a¢
 Tensors 145

;(~~ -Vo) ~(_l_Ovr -v )

I
Ov r
Or r sinO at; ; dr
Ov o 1 Ov o 1 1 avo
ar -;:(ao +vr) -(---coWv)
r sinO at; " ,dO]
Ov
-¢dr ~ Ov" 1 1 av"
-(----+v coW+v )
rsinO dt; ,
ar r ao r sinO at; 0 r

where the tensor gradv clearly appears in the second member.

(ii) The divergence ofv is tr(gradv) and thus it is immediately:

. Ov r 1 Ovo 1 av¢ cotO 2

d,vv = - + - - + - - - - + - - vo +-Vr '
Or r ao rsinO at; r r

We may also refer to Section 5.4 and consider (1-95).

The components dv' with respect to the orthonormal basis (1r,l o ,l;) , namely:

(dvy = (Vi,k+~~ vJ)du k ,

lead to the following tensor

a
-v
1 a
- - v --v
1 1 a
- - - v --v
1
ar r r ao r r 0 rsinOat; r r;
gradv = - v
a 1 a 1
- - v +-v
1
- - - v ---v
a cotO
ar 0 r ao 0 r r rsinO at; 0 r ;
a
-v
1
--v
a 1 a
- - - v +-v +--v
1 coW
ar " r ao ; r sinO at;; r r r 0

- 1 - 1 -0 1
since the only nonzero Christoffel symbols are F:O =--, F r =-- FrO =-,
r ;; r' r
-0 coW -; 1 -; coW I 2 3
F;" = - - - , Fr; =-, Fe; = - - and since du =dr, du =rdO, du =rsinOdt;.
r r r
CHAPTER 2

LAGRANGIAN AND EULERIAN DESCRIPTIONS

In this study every set of material points (called particles) which is capable of
changing in shape and move in a Euclidean space is called a continuum; but sometimes
this restrictive frame of reference will be widespread in subsequent chapters.
By choosing an orthonormal basis we can identify this Euclidean space with R3.

1. LAGRANGIAN DESCRIPTION

1.1 CONFIGURATION

Let S be a system of particles.

D The configuration of a material system S at a time t is the state of S with respect to

a frame of reference at time t.
It is denoted by C, .

The following notion will be considered more particularly.

D The geometrical configuration of a material system S at a time t designates the set

of positions of particles of S at the time t with respect to a frame of reference.

However, the previous term geometrical will be omitted later.

147

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
148 Chapter 2

At a time t, the position p of any particle in a frame of reference 90 = {o;el ,e2,e3} IS

determined by the position vector op with respect to 90:

op=x1ei ,
where the three components Xi are the coordinates locating p at time t.

Within some neighborhood of any particle there are infinitely many particles and it is
impossible to follow them, unlike in the kinematics of a single particle.

At this stage of the study it is necessary to distinguish two configurations:

(i) A reference configuration at a time to is generally chosen at to = 0 according to
usage. It is the reason why it is called the initial configuration. It is denoted by Co.

(ii) The configuration at a time t, called the current configuration, describes the set of
positions of particles of the material system at time t. It is denoted by Ct .

The parameter t is the index of the current configuration.

1.2 DEFORMATION AND LAGRANGIAN DESCRIPTION

Given a material system S we say:

D A material domain is a domain of a continuum whose points represent the positions

of particles at a given time.

A material domain of a continuum is denoted by Do at the initial time to and by Dt at a

later time t.
The boundary of a material domain will be denoted by aDo and aDt •

Unless otherwise specified, we decide to choose the same frame of reference at the initial
time to = 0 and the current time t.
This unconstrained simplification is illustrated in Fig. 6.

At the initial time let us consider a particle which occupies a position Po (in Do)
determined in the space by the position vector
oPo=X=Xie i ·
 Lagrangian and Eulerian Descriptions 149

At a later time t the particle occupies a position p (in DD determined in the space by the
position vector
op=x =x' e;.

e,
Fig. 6

Notation. We decide to use upper case letters when referred to the initial configuration
and lower case letters when referred to the current configuration.

A natural process consists in making the connection between the configurations Co and
Ct , thus between the position vectors x and X.

To progress we introduce the following definitions.

D The Lagrangian description is the way which consists in expressing, in the current
configuration, every physical quantity as a function of the particle concerned by the
quantity and identified in the initial configuration, given t.

D QF A deformation between an initial configuration Co and a given subsequent

configuration Ct is a change in the shape of the continuum between Co and C1 .
It is also called a transformation between Co and C1 .

Mathematically it is expressed as

l/J1 :Do ~Dt cR 3 :XHX=l/Jt(X),

this mapping being assumed at least of class Cl.

Thus in the configuration C, the position of the particle initially identified by X is

determined by the vector
(2-1)
150 Chapter 2

Of course, in the initial configuration the particle position is determined by

The basis (e i) lets us write (2-1) in the equivalent form

Xi = (/)/ (X) (2-1 ')

or explicitly:

D The components Xi of the vector X locating the particle in Co and the time tare
called the Lagrangian variables.

The connection between the initial and current configurations must satisfy regularity
conditions.
First, we postulate that every interpenetration is impossible; that is, two particles having
different initial positions cannot occupy the same position in the current configuration.
In other words, every mapping of Do into D/ defined by

is assumed to be injective.
Knowing that (/)/ is surjective ('onto'), namely:

we deduce that (/)/ is a bijection.

Therefore the existence of (/)/-1 lets us denote

At a given time t we consider the Jacobian matrix

(aa~ J
also denoted by

i,j = 1,2,3.
 Lagrangian and Eulerian Descriptions 151

D The Jacobian of the defonnation l/Jt is the detenninant of the Jacobian matrix:

PRI The Jacobian is such that

(2-2)

Proof Given the regularity condition on l/Jt and l/Jt- 1 , the Jacobian matrix is invertible
and the continuous Jacobian is different from zero and finite. Its sign remains constant,
that is, positive, since initially the Jacobian on Do is Jo(X) = 1.

Remark. The Jacobian plays a fundamental role in continuum mechanics, notably in the
study of defonnations.
Indeed, given the volume of an elementary material domain at Po of Co ' denoted by

given the corresponding volume following from the defonnation at p of Ct ' denoted by

then there is the well known relationship

Thus we say:

PR2 The Jacobian of the transfonnation between the initial and current configurations is
the ratio

that is the volume per unit original volume in the neighborhood of some particle.

D CiT' The displacement of a particle from the position Po in Co to the position p in Ct is

the vector
U(X,t) = PoP
=x(X,t)-X. (2-3)

Remark. This vector connects two positions at different instants in a given frame of
reference. It is the reason why we introduce a special notation that is U.
152 Chapter 2

1.3 FLOW AND HYPOTHESES OF CONTINUITY

If we consider the continuous sequence of deformations between an initial

configuration Co and a current configuration C/ , we can introduce the following notion.

D ClF A flow, that is the continuing state of motion of a continuum, also simply called a
motion, is the mapping
f/J: Dox R ~ R3 : (X,t) H x = f/J(X,t)
such that for any t of a given time interval I cR
VtE!: f/J(X,t) =f/J/ (X).

By assuming that the bijection f/J/ (or f/J/-I) exists for every t of a given time interval then
f/J is a bijection for every X and t, and we write:
x = f/J(X,t) <=> X = f/J-I(X,t) .

In a general way, the consideration of velocities and accelerations suggests that f/J and
f/J-I are to be of class C2 with respect to t.

For the same set of particles a material domain shows new posltlons of particles at
different instants. It is for this material domain, which is 'transported by the motion', that
we are going to introduce the following hypotheses of regularity:
(i) Two particles occupying 'infinitely neighboring' positions at the initial time to
occupy 'infinitely neighboring' positions in any other configuration.
(ii) Particles which occupy an initial connected domain occupy a connected domain of the
same type (either a curve, or a surface, or a spatial domain) in any other configuration.
(iii) The points of every material domain which is delimited by a closed surface Eo at to
remain inside the material domain delimited by the transformed surface E .
(iv) The particles which occupy positions on the boundary aDo are assumed to occupy
positions on the transported boundary aD. It is said that the boundary of any space
material domain is a material surface.

Remark. Certain situations do not let us consider regularity conditions as drastic! For
instance, two 'infinitely neighboring' particles of a fluid can be separated by an obstacle in
the course of time. The situations of shocks, fissures, and so on introduce discontinuities
and lead to giving up the continuity (the existence of discontinuity surfaces, for instance).
 Lagrangian and Eulerian Descriptions 153

1.4 TRAJECTORIES

The vector function such that

x=([J(X,t)

lets us follow the successive positions PI,P2, ... of a particle located by X at an initial time
in the frame of reference 90.

In the Lagrangian description we say:

D Cir The trajectory of a particle identified by X at to = 0 is the set of positions occupied

by the particle in the course of time in the frame of reference 90.

It is the curve defined by the vector equation

x =([J(X,t) ,
where the given vector X locates the initial position of a particle.
This equation of the trajectory is equivalent to the following parametric equations denoted
according to usage by
Xl =xl(t) ,
where t is the parameter.

The following figure shows two trajectories

Fig. 7

To each particle identified by a vector in the initial configuration corresponds a trajectory.

Each trajectory is 'indexed' by a vector defined in Co.
In a practical manner we can 'mark' each particle and observe the corresponding
individual trajectory (followed in function of t).

For example, taking a photo by a time exposure allows to observe the (relative) trajectory
of an object.
154 Chapter 2

1.5 STREAKLINE

We are going to consider another curve, the so called streakline l •

Let q be a point of the frame of reference 90. For example, this point locates the end of a
chimney in the distance.
Each particle through the point q is 'marked' at an instant; we observe the streak of the
smoke at a given time T, for instance.
Let p (1) be the position at the time T of the particle 1 which passed through q at the
II

time II' and so on.

Let p (1) be the position at the time T of the particle n which passed through q at the
I ..

time In.
Of course, the position of the particle at the point q at the time T is p (T).
10

We denote by pr(T) the position at time T of the particle which passed through q at the
time 'i E [to,T].

Fig. 8

The position of this particle is located on its trajectory at time T by the following position
vector
OPr(T) = t/J(X,T) ,
where X is the vector indicating the particle which occupied the initial position q at time
'i .

I In French: "Ligne d'emission."

 Lagrangian and Eulerian Descriptions 155

In particular, we have:

Therefore the streakline equation is obtained in function of the parameter r as follows:

D The streakline through a point q at a given time T later than an initial instant t is the
set of positions at time T of particles which passed (or will pass) through q at any
time 1: E [to,T] .

More simply, the streakline 'indexed' by fixed Xq and T is defined by the following
function:
x = If/ (r I xq,T) . (2-4)

Photography allows us to illustrate the difference between the notions of trajectory and
streakline. A trajectory, or path line, is obtained from some time exposure of a reflecting
particle and a streakline is obtained from a snapshot.

1.6 VELOCITY AND ACCELERATION OF A PARTICLE

In a frame of reference c"l0, we consider a particle which at to occupies a given

position Po determined by 0Po = X in the initial configuration.

D At time t the velocity of the particle identified by X (that is, occupying the position
Po at to) in 90 is the tangent vector to the trajectory of the particle at point
x = l/J (X, t) , namely:
V(X,t) = -al/JI eX,t). (2-5)
at x

It is the partial derivative vector with respect to t, given a fixed X.

Likewise we say:

D At time t the acceleration of the particle identified by X in the initial configuration

is in 90 the vector

, I (X,t).
a l/J2
reX,t) = - (2-6)
ar x
156 Chapter 2

Example. A particle identified by the initial vector X = (Xl ,X2 ,X3 ) occupies at any time t
the position defined by
Xl =Xl , x2 =X2 + tX3 , x3 =X3 + tX3 ·
(iJ The velocity of the particle which occupied the position (0,1,2) at the initial time is
such that

and the velocity vector is at any time:

V = 2e 2 + 2e 3 •
(iiJ The velocity of the particle which occupies the position (0,1,2) at time t=l is such that

Since Xl =0, X 2 =0, and X3 =1 we have:

1.7 ABSTRACT CONFIGURATION

The Lagrangian description of motion consists in identifying each particle with the
help of a vector X in the initial configuration; that is, with the aid of three coordinates of
the position initially occupied by the particle.

Nevertheless, we note that every particle must not be necessarily marked in this manner. It
can be judicious to choose three other parameters denoted, for instance, by ai (i=1,2,3). In
this case the configuration of reference is not labelled as the initial configuration but as the
abstract configuration of reference, denoted by Ca'
The various a i are called the coordinates of the particle in the configuration Ca'

In a manner full of imagery, let us consider the New York marathon. It would be
inappropriate to refer to initial coordinates of starting runners, it is more advisable to
identify each runner with a number.

The notions of previous sections may be introduced again.

For instance, the position of the particle identified at time to = ° by the vector

(/l(a' ,0) i = 1,2,3

is determined at any time t by

 Lagrangian and Eulerian Descriptions 157

where the various ai are the Lagrangian variables.

We note that the sign of the Jacobian

. D(x i )
J(al,t) = D(a i )

remains constant during the motion, but it is not necessarily positive.

2. EULERIAN DESCRIPTION

2.1 DEFINITION; COMPARISON BETWEEN LAGRANGIAN AND EULERIAN

DESCRIPTIONS

The Lagrangian description of any continuum motion refers to an initial

configuration; in contrast, the Eulerian description does not require the knowledge of the
initial state.
The classic image illustrating this new description is given by an angler; he is only
interested in the neighborhood of the float at each instant t, no matter what the past water
behavior around the float was like.
Within this new framework of study the part of the material system is not considered, only
the study of physical quantities at given points is important.
We say:

D The Eulerian description is the way which consists in expressing every physical
quantity at each considered time and at every point viewed as fixed in the frame of
reference.
It only refers to the current configuration, more precisely between instants t and
t+dt.

D The coordinates Xi of the vector x locating the point p where the particle is (in the
current configuration) and time t are called the Eulerian coordinates.
The Eulerian variables Xl are also called the spatial coordinates, whereas the
Lagrangian variables Xi are also called the material coordinates.

This last terminology can be justified because the Lagrangian coordinates are connected to
a particle; in a way, they constitute the 'identity card' of the particle. In contrast, the
Eulerian coordinates specify every point where any particle is, this last having no 'identity
card'.
158 Chapter 2

To sum up we say that the Lagrangian description identifies each particle and leads so
quickly to the individual trajectory, in which attention is fixed on a specific particle.
In contrast, the Eulerian description has a meaning at the instant considered, at each given
point, the description being concerned with a particular spatial domain occupied by the
continuum.

The Lagrangian description corresponds to a certain time exposure, the Eulerian

description corresponds to a sequence of snapshots.

Notation. In general, the expression of every physical quantity as a function of

Lagrangian variables must be distinguishable from the corresponding expression in
Eulerian variables.
According to usage every physical quantity expressed in Lagrangian variables is denoted
by an upper case letter, for example
Q(X,t)
and the same physical quantity expressed in Eulerian variables is denoted by a lower case
letter:
q(x,t) .
Thus we write:
Q(X,t) =q(x(X,t),t)

q(x,t) =Q(X(x,t),t) .

2.2 TRAJECTORY AND VELOCITY

(i) Exceptionally, the previous notation will not be used for a particular physical
quantity: the velocity.
If we denote the velocity of the particle passing through the position corresponding to x at
time t by V(E) (x, t), that is, the velocity expressed in Eulerian variables, then there is the
following relation between the velocity expressions in the two descriptions:
V(F)(X,t) = V(l:) (tP(X,t),t) = V(X,t),

in accordance with the law of motion x = tP (X, t) .

In a natural way we choose the same notation V for the two expressions of the velocity.

Example. Given a frame of reference { o;e1 ,eZ,e3 }, the motion of a particle of a continuum
is defined by the following spatial coordinates in function of the material coordinates of
the particle:
 Lagrangian and Eulerian Descriptions 159

where c is a constant of dimension [V T-l]

We note that
Xl2 +X22 = X2I + X22·

The components of the velocity vector are in the Lagrangian description:

c ct . ct
Vl(X,t)= 2 2 (Xlcos( 2 2)-X 2 sm( 2 2»'
Xl +X2 X +X Xl +X2

The components ofthe velocity vector are in the Eulerian description:

In addition, in the initial configuration we consider the particles belonging to the straight
line of equations
Xl =1,

and such that X 2 E [0,00) .

Since we have

. ( ct ) ct
sm 2 2 - COS(-2- - 2) 0
Xl +X2 Xl +X2
ct . ( ct )
cos( 2
Xl +X2
) sm- 2--
Xl +X2
2 0
[::1
0 0 1 X3
160 Chapter 2

the spatial coordinates of the particles are such that

. ct ct
sm(-2--2 )XI -COS(-2--2 )X2 = 1, x3 =0,
XI +X2 XI + x 2

with

Remark. We emphasize that the Lagrangian description allows us to obtain the Eulerian
description immediately, for example:
VeX,!) = V(cP-I(x,t),t).

Conversely, the Eulerian description allows us to obtain the Lagrangian description if the
initial configuration is known, for example:
V(X,t) = V(cP(X,t),t) ,
that is,
acPl
- (X,t) = V(cP(X,t),t).
at x

From the moment that X is fixed, that is, a particle is well determined by the initial
condition x(O)=X, then we note that the partial derivative ~I may be replaced by a
at x
derivative ~ since t is the only variable.

Thus we must solve the equation

dx = V(x,t) dt (2-7)
with the following initial condition
x(O)=X.

If V satisfies the regularity conditions, then Eq. (2-7) represents a system of 3 equations,
with 3 unknowns x' (t) , having a unique solution x which determines the trajectory of the
particle.

Likewise, every physical quantity q in the Eulerian description leads to its corresponding
expression in Lagrangian coordinates:
Q(X,t) = q( cP(X,t),t) .
 Lagrangian and Eulerian Descriptions 161

(iiJ From the velocity field V(x,t) we deduce that the trajectories are defined by

dx l dx 2 dx 3
--= = (=dt)
VI(X,t) V 2 {x,t) V\x,t) ,

where t is an integration variable.

Every trajectory is 'indexed' by three parameters which are the coordinates of each initial
vector X, Y, ...
The following figure represents trajectories and velocities.

Fig. 9

2.3 STREAMLINE

D At a given time T a curve whose tangent at every point has the direction of the
velocity at that point and time T is called a streamline .

We consider a curve of parametric equations

where the parameter s is the curvilinear abscissa for instance.

dx'
Let us express that the tangent vector (of components - ) at any point is collinear with
ds
the velocity at time T; that is:
dx' .
- = kV'(x,T) kER, i=1,2,3.
ds

I Called in French: 'Ligne de courant'.

162 Chapter 2

So the streamlines are made explicit by a system of two differential equations:

{/xl {/x 2 (/x 3
(2-8)
VI(x,T) V 2 (x,T) V\x,T)'

where fixed T is not an integration variable!

The streamlines in en (at time T) form a collection of curves 'indexed' by two (arbitrary)
constants, for instance a and jJ. In other words, the general solution of the previous
system of equations is expressed as function of two parameters.
Of course, two variables are functions ofthe third variable x 3 , for example; that is:

Xl =F(a,jJ,x 3,T), x 2 =G(a,jJ,x3,T).

The particular streamline through the point (x1,xp,x~) is determined from the values of c
and d which are obtained by solving the following system of equations
x1 = F( a,jJ,x~,T), xp = G( a,jJ,x~,T) .

To sum up, a trajectory represents the set of positions occupied by every particle when t
varies; a streamline is the envelope of the velocity field at a given time.
The obvious difference between the two notions is illustrated in the following figure; the
streamline at a point defined by Xl has the velocity at this point at time T as tangent
vector, and at a point defined by X2 has the velocity at this point at time T as tangent
vector. Of course, in general this last vector is different from the velocity tangent to the
trajectory at this point.

-=~~;r---===-- t raj.

Fig. 10
 Lagrangian and Eulerian Descriptions 163

Example. We consider the velocity field of components

VI =Asin(e(x2-kt»sinh(exI)'
V2 = Acos(e(x2-kt»cosh(exl)'
~ =0,

where the dimensions of the constants A, c and kare [Lrl], [L- l], and [Lrl] respectively,
Let us express the equations of streamlines in function of X2 at given time T.
Answer. The differential equations of streamlines are

A sin( e( x 2- kT» sinh( eXI ) Acos(e(x2- kT» cosh(exl )

The first equation is easily integrable, namely:

Jcoth(ex l ) dxl = Jtan(e(x2 - kT»dx2+ K (KeR),

that is
Insinh(ex I ) = -lncos(c(x2 - kT» + K
or
In(sinh(exI) cos(e(x2- kT» = K.
Thus we have obtained:
1 'nh- I (
XI =-Sl eK ) .
C cos(e(x2-kT»

Finally, the second equation leads to

where C is the second arbitrary constant.

2.4 STEADY MOTION

Let us mathematically translate the situation of the angler who continuously observes
the same stream velocity in the neighborhood of the float.

D A motion is said to be a steady motion l in ~ if the velocity field (in Eulerian

variables) is independent of time, that is,
av
-(x,t)=O. (7-9)
at

I Called in French: «Mouvement stationnaire» or «Mouvement permanent.»

164 Chapter 2

In other words, the velocity at any fixed point in 9-0, that is, V(X) , is invariable during a
given period oftime.

PR3 In the case of. every steady motion corresponding trajectories, streaklines, and
streamlines coincide (geometrically but not conceptually).

Proof The common differential system is

dx' dx 2 dx 3
(7-lO)
V'(x) = V2(X) = V\x) .

PR4 Every steady motion is invariant under any time translation

Proof For steady motions the following equation

dx = V(x)dt
shows that x does not vary for a time translation t' = Hc (cER).

The position vector of the particle at point p at time t and located at point q at time r is a
function of (- r (and of q), namely:

x=!(xq,t-r).

This is the equation ofthe trajectory passing through q, the time parameter being t- r .
 Lagrangian and Eulerian Descriptions 165

EXERCISES

Exercise 1.

A velocity field has the following components

VI==XI' V2 ==~
2t+3' V3=='
0

with respect to a basis of the usual Euclidean space and where t is the time parameter.
(i) Find the parametric equations of the trajectory of a particle which was at (X] ,X2 ,X3)
at the initial time to==O .
(ii) Determine the streakline at given time T made up of the particles which passed
through the point (a,b,c) at a variable time TE [to,T].

Answer. (i) In order to find the trajectory of the particle we must solve the system

dx3 ==0
dt
with the initial conditions

First, we have:

that is
(1)

Second, we have:

lx,x' -;;
du
==
r' dt
Jo 2t+ 3
~ ln~ == In-J2t+ 3 -In.J3
X2
that is

(2)

Finally, we have:
X3 == X3 . (3)
Eqs. (1), (2) and (3) give the Lagrangian description of the motion.

(ii) At a given time T the coordinates of the particle must verify the following equations
of the trajectory
166 Chapter 2

Since at some time r « T) the particle was at (a,b,c) we have:

By eliminating Xl, X2, X3 between the previous equations we obtain the parametric
equations of the streakline:

Xl =ae
T-r
,
b~
X2= ~
'\'1Hl '
where r is the parameter.

Exercise 2.

We consider the motion of a continuum defined by

Xl = Xl' X2 = 2f X 3 , X3 = X3 +I
(i) Give the Lagrangian description of the physical quantity defined by

(ii) Express the velocity in Lagarangian and Eulerian descriptions.

(iii) Calculate the rate of change of the physical quantity.

Answer. (i) The Lagrangian description of q is immediately

Q(X,t) = 2X l + (2t-1)X3 .

(ii) The velocity ofa particle identified by X is the vector V(X,f) = ox I

of x
of components

~=O, VZ =2X3 , V3=0.

By considering the motion we obtain the velocity V(x,f) in the Eulerian description,
namely:
v; =0, ~ =0.

(iii) The rate of change of the physical quantity is

oQI _
YtX- 2X3 ,
that is, also
 Lagrangian and Eulerian Descriptions 167

Exercise 3.

In a frame of reference 90 we consider the motion defined by the following

orthogonal Cartesian coordinates

(i) Determine the velocity components in Lagrangian and Eulerian descriptions.

(ii) Give the Cartesian equations of the trajectory.
(iii) Find the equations of the streamline through the point (a, b, c).

Answer. (i) First, we note that the motion is defined for f < 2 since the Jacobian

° t/2° =1--
1
t2
4
° t/2 1

is actually positive.
In Lagrangian form the velocity components are
1
VI (X,I) = 0, V2 (X,I) = aX21 = .!..X3 , V3 (X,t) = -X2 .
at x 2 2

From

which implies:
t I
x 2 --x
2 3 x3 --x
2 2
X 2 =--=---
J J
we deduce the following velocity components in Eulerian form:
t I
x3 --x
2 2
x 2 --x
2 3
VI (x,f) = 0, V2 (x,f) = 2 V3 (x,f) = 2
t t
2--- 2--
2 2

(ii) The trajectory is obtained by eliminating t. Thus, its equations are

168 Chapter 2

(iii) At time T the equations of streamlines are

dx 2 = dx 3
v3 '
where a is an arbitrary constant.
This parameter is determined and equal to a for the streamline passing through the point
(a,b,c), and the equations of this streamline are

also written:

Exercise 4.

We consider the velocity field of components

V=~
2 2t+ 3 '
Find the streamline which passes through the point (a,b,c) at given time T.

Answer. From
dx2=k~ dx3 =O
ds 2T+3 ' ds '

we deduce the following system of equations of streamlines

dxl = (2T+3) dx2 ,

XI x2
and thus
x\ du =(2T+3)I x 2 dv ,
I
a U b V

that is
X X
In-.L = (2T+3) In....l
a b
or
2T+3
( x2 ) = ~.
b a
 Lagrangian and Eulerian Descriptions 169

We note that we obtain the same result as follows:

CHAPTER 3

DEFORMATIONS

Weare going to introduce the notion of deformation of a continuum on the basis of

a comparison between two configurations such that the first one is the initial configuration
Co and the other is the current configuration Ct. We emphasize that no attention is paid to
intermediate states.
In this context of taking into account an initial configuration the Lagrangian description is
imperative, and it is perfectly suited to the study of geometrical deformations of solids, for
instance.
We consider again a frame of reference 90 with an orthonormal basis (e i ).
We recall that a transformation undergone by a material system has been geometrically
defined by the vector function (/Jt such that
X =(/Jt(X) ,
where X locates the position of a particle of the material system in the configuration Co
and where t is the given instant corresponding to the final configuration Ct.
Given X and t, the vector function (/JI is sometimes called the convective transport of the
particle located by X between the configurations Co and Ct.

By using a comparison with photography, in this section we are going to consider two
snapshots to study the deformation showing a change in the shape of the medium between
the two configurations Co and Ct.
Weare going to use notions which we have previously introduced in the Lagrangian
description. Let us begin by studying finite transformations before considering
infinitesimal transformations, which latter will playa major role in continuum mechanics.

171

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
i72 Chapter 3

1. HOMOGENEOUS TRANSFORMATION

1.1 DEFINITION OF HOMOGENEOUS TRANSFORMATIONS

D rJF A transformation of a system between two configurations Co and C1 is

homogeneous if the corresponding vector function l/J 1 is such that
x = l/JI(X) = FI . X + k l , (3-1)

where F; is a tensor whose components only depend on the choice of t;

equivalently, if, given t there is an affine relation between the coordinates, namely:
i,j = 1,2,3. (3-1 ')

We emphasize that FI is independent of X.

The tensor FI defining a linear mapping on the reference space (into itself) is expressed as
follows:

whilst the vector kl is explicitly written:

kl = kiei .

Remark 1. We note that we have introduced a contracted multiplication. The contracted

product FrX between a tensor of type (:) and a tensor of type (~) is a tensor of type (~).
Of course, in the matrix context this vector is a column matrix following from the usual
rule of the 'row by column'.

Remark 2. We specify that the function l/J 1 must satisfy the regularity conditions
mentioned in Chapter 2. For example, the nonsingular character of the matrix (Fj) is
essential and thus J = det(F~) cannot vanish.

Remark 3. For every affine transformation we know that a straight line corresponds to
any straight line, a plane corresponds to any plane, the parallelism is preserved, a
parallelogram corresponds to a parallelogram, a parallelepiped corresponds to a
parallelepiped, the segment subdivision ratio is preserved.
In short, all the properties of an affine transformation can be found from the most natural
example given by the solar projection (the Sun being viewed as being at infinity).
 Deformations 173

Fig. 11

Example 1. Given an orthonormal basis the transformation defined by

x = (X\)2e\ +X2X3e2+X\X3e3

is not homogeneous.

Example 2. Let us consider the homogeneous transformation defined by the following

rectangular Cartesian coordinates

In the initial configuration we consider two particles a and b of respective position vectors
A=e\+e2 and B=2e\+2e2+e3.
Let us express the vector connecting the particles in the current configuration and let us
show that initially parallel vectors remain parallel after transformation.
We have
v = B-A =e\ +e2+e3.
Since the tensor F is explicitly

121] ,
[020
102

we deduce that the position vectors of the respective particles in the current configuration
are

and thus in this configuration the vector joining the particles is

v = b-a = 4e\ + 2e2+ 3e3.
Any vector parallel to V, for example 2e\ + 2e2 + 2e3' becomes after transformation the
vector 8e\ + 4e2+ 6e3 which is actually parallel to v.
174 Chapter 3

1.2 CONVECTIVE TRANSPORT

1.2.1 Convective Transport of a Vector

A pair (Po,qo) in the frame of reference defines the vector V =Poqo in the initial
configuration Co. Let v = pq be the corresponding vector in the final configuration Ct.

Fig. 12

Since the parameter t is fixed in Ct , the vector k t in Eq. (3-1) is a constant vector.
Thus we have:

Consequently we say:

D Given a homogeneous transformation the convective transport of a vector V (in

the initial configuration) is the nonsingular linear mapping which associates to V
the vector v in the final configuration, such that
v=F;Y. (3-2)
The vector v is called the transported vector of V (by convection).

With respect to a basis the components of v are expressed as follows:

1.2.2 Convective Transport of a Volume

Let U, V, W be vectors in the initial configuration Co,

u, v, w be the respective transported vectors in the configuration Ct.

We recall that the homogeneous transformation changes the parallelepiped with sides
corresponding to U, V, W into the parallelepiped spanned by the transported vectors u, v,
w.
 Deformations 175

We know that the volume value (with sign) of the parallelepiped defined by U, V, W is the
mixed product U "V.W , in which we can interchange the dot and vector products.
We also know that the adjoint of the exterior product
U "V = (UiVJ-UJVi)ei ®e J
is the covector whose components are

The previous mixed product is the contracted product of the adjoint of U "v and vector
W, namely:
(h = 1,2,3).

In an orthonormal basis, since ~detg = 1 we find again the classical expression of the
mixed product:
U' U2 U3
V' V2 V3
W'W2 W3

By coming again to the notion of transported vectors, we explicitly denote:

and thus

We deduce that
u" v.w = detF; (U " V.W).
Thus we say:

D Given a homogeneous transformation the convective transport of a volume flo of

a parallelepiped (in Co) is the nonsingular linear mapping which associates the
volume fll (in CI ) such that
fl, = flo detF; . (3-3)

The transported volume of flo (by convection) is the deformed volume fll.

Therefore given a homogeneous transformation between configurations Co and CI we say:

PR1 The ratio !!.J..., that is, the volume per unit original volume, is J(t) = detF; .
flo
176 Chapter 3

1.2.3 Simple Shear

Given an orthononnal basis (e" el , e3) of a Cartesian frame of reference 90, let us
consider the classical example of the simple shear defonnation', which is defined by the
homogeneous transfonnation
x t =(/Jt(X)=X+ktXle"
that is explicitly:
(3-4)

Let us describe this defonnation and find the tensor F t •

A particle located at the origin 0 in the initial configuration will remain at this position,
since the material coordinates are X, = Xl = X3 = O.
Likewise, a particle located at any point A(X"O,O) in the initial configuration will remain
at this position because x, = X" Xl = 0 and X3 = O. Thus the material coordinate line oX, is
fixed.
A particle located at any point B( O,Xl'O) in the initial configuration will be translated to
the point b( kt,Xl'O) at a distance kt parallel to ox" since x, = X,+ ktXl , Xl = Xl, X3 = X 3.
Likewise, a particle located at a point C( X, ,Xl'O) in the initial configuration will be
translated to c(X, + kt,Xl,O).
Let us represent this situation in the following figure.

b ( c

~o~------------------~----------x,
A

Fig. 13

In general X3 = X3 is different from zero and the simple shear defonnation is in the plane
of the equation X3 = X 3 .

I Called in French: 'Cisaillement simple'.

 Deformations 177

The expressions

lead to
Fj=l, F~= let, F~=O,
and so on.

The tensor F/ of a simple shear deformation is explicitly

[ ~o ~ ~l·
0 1

We note that the volume per unit original volume is

J(t) = detE; = 1,
and there is 'no dilation' in this case.

1.3 CAUCHY-GREEN DEFORMATION TENSOR AND STRETCH

We are going to make explicit the geometrical deformation of continua in the case
of homogeneous transformations between initial and final configurations.
We recall that direct isometries are obtained as translation and rotation products for which
there is no deformation between the configurations Co and Ct. The distances of every pair
of particle positions are invariant and each angle between every pair of directions is
preserved. In kinematics this corresponds to rigid displacements.
On the other hand, the study of medium deformations between Co and C/ requires us to
evaluate the changes of distances and angles between directions. The scalar product is the
mathematical notion which takes account of both distances and angles. Therefore to make
explicit some deformation between two configurations Co and C/o we are going to compare
the scalar product of any two vectors between the configurations.

1.3.1 (Right) Cauchy-Green Deformation Tensor

Given two vectors V and W in Co we consider the scalar product v.w of

transported vectors in Ct.
To express v.w in function of Vand W we introduce the following notion.
178 Chapter 3

D CiF The (right) Cauchy-Green deformation tensor 1 at a point Po between Co and Ct

is the tensor of type (g) defined by

C = t£.g·F. (3-5)

It is explicitly denoted by

e(t) = Clj e i 0e i ,

where its components, which depend on t, are dimensionless (as those of Fl

Let us express the components Cij as functions of those of F.

We have
C = « tF)/ ei 0ep )' grs e r0e s . (Fj eq 0e J)
= (IF)f gpq Fj e i 0e 1

= g pq FP' Fq
1 e'0e
i .

Thus the components of Care

(3-6)

Let us now consider the scalar product of v and w, and let us prove the following
proposition.

PR2 The tensor C is the bilinear form for which the image of every pair of vectors
(V, W) is the real

(3-7)

that is, the scalar product of the transported vectors v and w.

Proof We have
v.W = gijviwi = gijFirFls vrw s
= Crs vrw s .

We have obtained:
v.W =V.C·W. (3-7')

I Called in French: 'Tenseur des dilatations de Cauchy-Green (11 droite) 'or 'Tenseur des dilatations '.
2 To simplifY the notation we remove the time reference t from the written form of tensors in C,.
 Deformations 179

In every orthonormal basis we have gpq =opq, the higher or lower level of indices do not
matter much and we denote indiscriminately:

Cij = opqF~Fj = F'qi Fqj

3
=2: FqiFqj
i=1
or simply
(3-8)

Since

we denote in a natural manner:

C= 'F·F (3-9)

and
v.w = (F-V).(F-W)
= v.(1F-F)·W.

Properties.

Pl. The tensor C is symmetric:

fC=C.

This property Cij = Cji is obvious from (3-6).

Of course we also denote:

P2. The tensor C is nonsingular.

Indeed, we have
0< detC < 00
since
detC = (detF)2 .

P3. The bilinear form associated with C is positive-definite.

Indeed, for any V in the initial configuration Co we have

ceV,V) = CijViV j = V.v = II V 112 ~O,

ceV,V) =0 ¢::} V= 0 .
180 Chapter 3

1.3.2 Stretch

From C we are going to study the changes of distances from one configuration to
another. More precisely, we are going to compare the norm of any vector V in Co with the
norm of the transported vector v in Ct.

D c:Jr The stretch in the direction 1 of V is the ratio of norms of the transported vector v to
the corresponding vector V in Co, that is the positive real

A(v)=M (3-10)
I Vii ,
or equivalently:
A(V) _..[V:c.V - JWY) . (3-10')
- IIVII - I Vii

Remark 1. The stretch in the direction of a unit vector ei of any basis is obviously

Remark 2. The terminology 'stretch' is used even if A(V)<1 (contraction).

Remark 3. The local 'character' of C lets us determine the stretch in every direction, in
the neighborhood of a given point.

We can also consider the change oflength per unit original length and we say:

D The unit elongation or unit extension 2 relative to V is at a time t the real

b(V) - 11vII-IIVI = A(V)-l . (3-11)

- I Vii
Thus we have
(3-12)

Now let us take up the changes in angle; more precisely, the changes in angle between two
line elements which were originally at right angles to one another.

I Called in French: 'Dilatation dans la direction '.

2 Called in French: 'Allongement unitaire '.
 Deformations 181

1.3.3 Shear Angle

From pairs of originally orthogonal vectors we are going to make explicit changes
in angle through a convective transport.

Let Vand W be orthogonal vectors at Po in Co.

The angle between the respective transported vectors v and w being !!..- B, we have:
2
v.w = 1/vIIllwil COS(li - B)
2
= 1/vIIllwil sinB .

L
Po V p
v

Fig. 14

Thus we express:

D The shear angle at time t for a pair of orthogonal directions l of initial vectors V
and W is the angle B such that

. B C(V,W)
(3-13)
Sin = -..;r=C=(V=,=V=-)";"C=(=W=,W=)

Remark. In particular, for vectors ei and ej of an orthonormal basis, we have:

. B Clj
Sin = . (3-14)
rc:~
"Lit v"-'j}

1 Called in French: 'Glissement dans un couple de directions orthogonales "

182 Chapter 3

1.3.4 Principal Stretches

We recall that the notions of eigenvalues and eigenvectors let us define principal
directions of tensors.
Let us consider the Cauchy-Green tensor C, which is symmetric and real. We know that a
real symmetric tensor has three principal directions in the 3-dimensional space which are
mutually perpendicular.

Let (EI ,E2,E3) be a basis of unit vectors in the principal directions and called the
principal basis (or the basis ofprincipal vectors).
Let Yi (i = 1,2,3) be the eigenvalues of C.

The Cauchy-Green tensor C is explicitly written:

C=~~®~+~~®~+~~®~. (3-15)

PR3 The three (real) values Yi of every symmetric tensor C are (strictly) positive.

Proof The definition of every principal vector

(no sum!)
implies

Since the bilinear form associated with C is positive-definite we conclude that each Yi is
(strictly) positive.

D The stretch in a principal direction is said to be a principal stretch.

Given a principal vector Ei we denote the corresponding principal stretch by

(3-16)

PR4 Every eigenvalue Yi of C is the square of the corresponding principal stretch:

(3-17)

Proof We have:
y, = C(Ei,E,) = eii = (A,(E,»)2.
 Deformations 183

PR5 Given a convective transport the principal directions of C in the configuration Co

remain orthogonal in the configuration Ct.
In other words:
The shear angle for any pair of principal directions is zero.

Proof The transported vectors of a principal basis (EI ,E2,E3 ) are

( i = 1,2,3).

The scalar product of any two vectors of the basis (e;) is

e;.ej = (F.E;).(F-E j ) = E;.(F-F).E j =E;.C.E j =E;1j E j

and since the principal basis is orthogonal in the initial configuration, we have:
e;.ej = Yj t5ij'
The proof is also immediate from (3-14) since Cij = 0 and thus sinO = 0 .

Remark 1. The (orthogonal) transported vectors of the vectors of the principal basis are
not necessarily normed since

Ile;11 =IIF-E;I!= A(EJIIE;II =,1;.

Of course, it is also written:

Remark 2. A frame defined from a principal basis in Co and the frame corresponding to
the transported vectors of this principal basis in C/ have the same orientation since
J(t) = detF> O.

PR6 The volume per unit original volume J(t) is the product of the three principal
stretches.

Proof The determinant associated with a given tensor does not depend on the chosen
basis. By choosing the principal basis in Co, we have for C:
J 2(t) = (det('F·F»2 = detC = YIY2Y3
= (~~~)2

and since J(t) is strictly positive we have:

J(t) = ~~~. (3-18)

184 Chapter 3

1.4 FINITE STRAIN TENSOR

Let us study the defonnations by considering the scalar product. More precisely,
we are going to compare the scalar product of any two vectors V and W in the initial
configuration with the scalar product v. w of the transported vectors, namely:
v.w-V.W.

In (general) curvilinear coordinates this difference has the following fonn:

V.c·w - v.w = (Cij- gij)V'W 1 .

This leads to introduce the following notion.

D Cff' Thejinite strain tensor 1 is the symmetric tensor

L = t<C-g). (3-19)

Explicitly we have:
(3-20)

Thus the comparison between the previous scalar products is

v.w-v.w = 2V.L·W. (3-21)

In an orthonormal basis the finite strain tensor is defined as follows:

D Cff' The Green-Lagrange strain tensor 2 is the symmetric tensor

1
L=-(C-I). (3-22)
2

Its components are

(3-23)

PR7 The Cauchy-Green defonnation tensor C and the Green-Lagrange strain tensor L
have the same principal directions.

Proof. It is obvious since L = ~(C - I).

2

I Called in French: . Tenseur des deformations jinies '.

2 Called in French: . Tenseur des deformations de Green-Lagrange '.
 Deformations 185

D The eigenvalues of the Green-Lagrange strain tensor L are called the principal
strains.

The principal strains are denoted by L;.

Given a principal basis (£1,1£2'£3)' the expression of the Green-Lagrange strain tensor is
(3-24)

The principal strains are connected to the principal stretches as follows:

1
L; = 2(y;-I)

= ~(A?-l)
2 1
(3-25)

(which implies that Li > - ~ ).

2
Thus given a principal basis the strain tensor L is explicitly:

*<Al-I) 0 0
2
0 l(~_I) 0
2
0 0 l(~_I)
2

Remark. If there is no deformation then we have:

C=I <=> L=O.

Let us consider the shear angle for a pair of orthogonal directions of two initial vectors V
and W.
Since V.W=O we have:
C(V,W) = 2L(V,W)
and thus the shear angle at time t is such that
2L(V,W)
sinO = (3-26)
~I + 2L(V,v) ~I + 2L(W,W)

For vectors ej and ej of an orthonormal basis we have:

2Lij
sinO = - ; = = = = - " ; = = = (3-27)
'1+2[" i l + 2 L
'\j II " JJ
186 Chapter 3

To conclude this section we introduce the following notion.

D The unit elongation for an element in a principal direction IS said to be the

principal unit elongation.

For the principal vector if:; it is denoted by

8, =8(Ei)'

The principal unit elongation is connected with

(i) the principal stretch: 8 i = Ai - 1 ; (3-28)
(ii) the eigenvalue of C: 8i =h -I; (3-29)
(iii) the principal strain: 8, =~1+2L,-l. (3-30)

1.5 POLAR DECOMPOSITION

1.5.1 Pure Stretch and Rotation

Let us consider the principal directions of C in the initial configuration Co' We

recall that the transported vectors of the various vectors Ei of the principal basis are
ei=F.Ei ·

Given a convective transport we know that the transported vectors e,

of the principal
basis are mutually orthogonal and the orientation of the basis (e;) is preserved, but the
basis vectors are generally no longer normed, namely:

We know that the tensor F characterizes the change from the frame {Po;Ej,EbEJ } to the
,e
frame {p;e; 2 ,eJ }. Given the principal basis (Ei) at point Po let us express F as a
product of two tensors, which of course define linear mappings.

The tensor associated with the first linear mapping, called the pure stretch tensor Set) and
denoted simply by S, is aiming at multiplying the unit vectors Ei by the respective
principal stretches Ii., to obtain the norms of respective vectors e, . The tensor S stretches
the principal vectors Ei • It preserves the principal directions and is a positive symmetric
tensor.
The tensor associated with the second linear mapping, called the rotation tensor R(t) and
denoted simply by R, is aiming to bring the principal directions of vectors AiEi to the
(orthogonal) directions of homologous vectors e,. It is an orthogonal tensor.
 Deformations 187

Moreover, we recall that the orientation is preserved.

I
--... I
R-' I /
,I
,I

Fig. 15

Thus we express:

D rJr The product of the pure stretch tensor S with the rotation tensor R, denoted by
F=R·S, (3-31 )
is called the polar decomposition of F.
The order of the linear mappings so defined lets call S the right pure stretch tensor

Explicitly we denote:

We sum up:

The decomposition can be obtained by starting with a rotation which is followed with a
pure stretch, that is, in terms of tensors:
F=T·R, (3-32)
where T=F-R-i is called the left pure stretch tensor.

We sum up:

An obvious relationship between the pure stretch tensors Sand Tis

T= R·S·R-i. (3-33)
188 Chapter 3

We note that the 'eigendirections' of T are relative to the basis vectors 'i; =F.ji; ; in other
words the matrix of F is diagonal with respect to the basis ('i;), which is not principal in
general.

The right pure stretch tensor is diagonal and its nonzero elements are the various Ai >0.
Thus given the principal basis the expression of the right pure stretch tensor is
(3-34)

We emphasize that S and T have the same respective eigenvalues, whereas it is not true
for the 'eigendirections'.

PR8 The polar decomposition of F is unique.

Proof (i) Let us prove the following

F=R'·S'=R"·S" { R'=R"
s'=s"
We successively have:
I(R'· S') = I(R"·S")
~ S,·IR'= S".IR"
~ S'· IR'·R'·S' = S"· IR"·R"·S"

and since S' and S" are positive-definite we conclude that S'= S"= S.
In addition, from
F = R'·S = R"·S ,
we deduce that R' = R"
(ii) Likewise, it is proved that the decomposition F = T·R is unique.
The knowledge of F implies that of T (because T2 = T . R· I R· T = F- I F and the various
eigenvalues of T are strictly positive) and R is determined by T~l·F .

(iii) The orthogonal rotation for both decompositions is unique, namely:

F= R·S = T·R' ~ R=R'.
Indeed, we have:
F = R·S = T-R' = R'·R,-I·T ·R' = R'· (R,-I . T· R)
=R'·S'
and by referring to (i) we deduce R=R'.
 Deformations 189

PR9 The squared right pure stretch tensor is the Cauchy-Green deformation tensor:
S2=C (3-35)
and the Lagrange-Green strain tensor is expressed as follows:

(3-36)

Proof We have:

Next, the above expression of L is obvious.

Remark 1. From the previous proposition we deduce the general expression of

components of C:
(3-37)

Remark 2. The nonzero elements of S2 are the diagonal elements

i= 1,2,3.

Remark 3. The positive-definite tensor S can be calculated from the expression

(3-38)

and
(3-39)

Remark 4. The well known property R-I = IR of the (orthogonal) rotation tensor R is
obviously verified since we have
S-I. IF· F· S-I = S-I . S2. S-I = 1
¢::> t(F-S-I)·F-S-1 =1

¢::> tR·R=I.

Important Remark. The following example shows that a pure stretch deformation does
not necessarily preserve the direction of a vector; nevertheless, if the initial vector
direction is principal then it is preserved.
Let us consider the deformation defined at t = 1 by

X2= 2X2 ,

The Cauchy-Green deformation tensor C is

190 Chapter 3

1 00]
[ 040 .
009

The vectors of the principal basis are

El = el , E2 = e2 , E3 = e3.
For instance, the point (0,1,1), initially belonging to the bisector of the angle (ox 2x\
becomes the point (0,2,3) in the final configuration, and thus the direction of the deformed
bisector is defined by tan -I (3/2) in the plane OX2X 3 •

1.5.2 Euler-Almansi Strain Tensor

First, we say:

D The left Cauchy-Green deformation tensor is

C L =T2. (3-40)

PRIO The left Cauchy-Green deformation tensor is expressed as

CL=Y'F. (3-41 )

Proof From the implication

F= T·R => tF- l = tT-l.tR- l =T-l·R ,
we deduce

Thus we have obtained

C~I = 'F-1.F- 1,
which proves the proposition.

PRII The left Cauchy-Green deformation tensor is

(and C = 'R·C L ·R) (3-42)

Proof We have:
C L =F·'F=R·S·'S·'R=R·S 2 ·'R=R·C.'R
and
 Deformations 191

Notation. Logically a lower case letter denotes C~', namely:

c =C~' =(£. 'Ff'. (3-43)

D The Euler-Almansi strain tensor is in Ct the tensor

(3-44)

In an orthonormal basis, its components are

e =..I..(8- aXr aXr) (3-45)

I} 2 I} a a
Xi Xj
.

1.6 RIGID BODY TRANSFORMATION

We are going to consider an important type of transformation, that is, preserving

the distances and angles. In particular, the rigid bodies are media which illustrate this
characteristic, but there are others. Let us particularize a general transformation
x = fPt (X) between an initial configuration Co and a given configuration Ct , as follows:

D (ff' A rigid body trans/ormation l of a system between Co and Ct is a homogeneous

transformation defined by
(3-46)
where R denotes an orthogonal tensor of rotation at time t and k is a translation
vector.

Oy-_ __

Fig. 16

I Called in French: 'Transformation rigidifiante '.

192 Chapter 3

PR12 A rigid body transfonnation between Co and C1 implies

C=I (orL=O).

Proof In this case F is a tensor of rotation R and thus we have:

IF=F-l.

The matrix defined by F is orthogonal and such that det F = + 1 , thus we have:

Remark. As was to be expected, there is no defonnation in the case of rigid body

transfonnations and thus v.w - V.W = O. So a rigid body transfonnation expresses the
invariance of the scalar product V. W of any pair of vectors and thus the invariance of
distances and angles.

We can introduce the converse of the previous proposition in the manner:

PR13 If the Green-Lagrange strain tensor L vanishes at some point in Co, then the
system undergoes a rigid body transfonnation between Co and Ct.

Proof From L = 0 or C= I we deduce:

S2=I
and thus
S=I
because S is positive-definite.
In conclusion we have:
F=R
with
detR=+l.

Likewise we have at time t :

T=I
and
e=O.

Remark. We note that the proof of the previous converse requires the homogeneity of the
transfonnation; that is, F only depends on t and not on X.
It can be proved that R is actually independent of X. This is intuitively justified by
splitting the system in Co into infinitesimal parallelepipeds. These transfonned
parallelepipeds fonn in Ct a system of rotating and continuously adjoining parallelepipeds;
so R must be the same for every parallelepiped, that is, independent of X.
 Deformations 193

Training. The reader will be convinced that every rigid body transformation is such that:
J = detF = detR = 1,
the stretch in the direction of every unit vector Iv is

the unit elongation in this direction is

8(V) = .-t(V)-l = 0,

the shear angle () for a pair of orthogonal directions of initial unit vectors Iu and Iv IS

such that

the Euler-Almansi strain tensor vanishes:

2. TANGENTIAL HOMOGENEOUS TRANSFORMATION

Let us consider a continuum undergoing a transformation between two

configurations Co and C, defined by the already introduced vector function
(/), : Do~ Dt cR3 :XH (/),(X).

We are going to use a procedure which consists in approaching locally at each point in Co
a general transformation by a homogeneous transformation which is tangent to it. So we
will be able to use the developments of previous sections; but the local meaning cannot be
too strongly emphasized.

2.1 DEFORMATION GRADIENT

Given a frame of reference 90 = {0;e"e2,e3} we consider two particles which

occupy neighboring positions Po and qo in Co and defined by

X=OPo,
194 Chapter 3

In the configuration Ct these particles occupy the respective positions

x=op, y=oq.

Thus in Co a pair of neighboring points (Po,qo) determines a vector Po% which becomes
in Ct :
pq= y-x
= ([Jt ( Y) - ([Jt (X) .

Let us consider the point Po such that oPo=X =Xiei and the point qo in the neighborood
of Po such that oqo =Y =Y;ei .

Given the function ([Jt of class C l , if Xi and yi denote the coordinates of the
corresponding points p and q in Ct then the comparison between these transformed points
is expressed as follows:
yi_xi = <Pi(yl,f2,y3 )_<Pi(XI,X2,X3), i = 1,2,3,

where the index t is omitted.

We consider the truncated expansions (i = 1,2,3 ):

(3-47)

where
lim lJ'(Po,qo)=O.
qo--"po

Consequently the derivability of ([Jt (X) is expressed through the linear function defined
by the matrix

a<Pi).
[ aX}

Notation. In coordinate form the coordinates Xi of X and t are no longer denoted.

Thus given the coordinates Xi of Po and the coordinates X, of its corresponding point p in
Ct, from the three scalar equations (3-47) written as follows:
a<pi. 8<Pi.
yi = --.
aX}
f1 + (Xi ---. X}) + IIPoqolllJi(po,qo)
ax}
(3-48)

we then deduce that there is an 'affine correspondence' between the coordinates Y; and
yi of any particle in Co and Ct successively.
 Deformations 195

By considering the tensor notation Fpreviously introduced in the context of homogeneous

transformations, each scalar Eq. (3-48) is then written:

(3-49)
where the various

(3-50)

are the components of an important tensor which we are going to define by considering
the three scalar equations (3-47) in a vector form.

D r:r The deformation gradient of f/J t (X) is the tensor! at time t:

1'; (X) = V~ (X)

such that
y-x = Vf/Jt(X),(Y-X) + IIY-XII7Jt(Y-X). (3-51)

This (: )-tensor is explicitly denoted by

afPi .
Vf/Jt(X) = ax J e/i)e J • (3-52)

We note that Eq. (3-51) is written equivalently as:

pq = 1'; (Po)'Poqo + jjPoqojj 'It(Poqo)· (3-51 ')

D The linear mapping associated with the deformation gradient V f/JI (X) is called the
homogeneous transformation tangent to the transformation f/J t, at Po (or given
X).

The terminology 'tangent' follows from the limit of the last term being zero:

lim 11,(Poqo)=O.
qo .... po

The effect of the tangential homogeneous transformation is schematized in Fig. 17.

I Also called the' Material deformation gradient' to be distinguished from the 'Spatial deformation

gradient' which will later be introduced with partial derivatives with respect to spatial coordinates Xk •
196 Chapter 3

v~t

Fig. 17

Remark. The tangential homogeneous transfonnation at Po is a local notion (in the

neighborhood of Po). It is the reason why we are going to introduce the differential
notation. This notion is rigorously studied by considering tangent vector spaces at
points Po and P of respective configurations Co and Ct.

2.2 HOMOGENEOUS TRANSFORMATIONS OF ELEMENTS

In this section devoted to local defonnations we introduce the differential concept;

more precisely, any pair of neighboring points (Po,qo) in Co defines a vector element
denoted by

2.2.1 Transport of Vectors, Volume Deformation, and Area Deformation

(i) First, let us define the transported vector of a vector element.

D qr The transported vector of Poqo in the configuration Ct is the vector

F; (Po)· Poqo
denoted by
dx = F-dX = Vl/J·dX (3-53)

where, to simplify the writing as previously, the reference position Po of the

particle as well as t will be omitted.
 Deformations 197

Eq. (3-51 ') is written:

pq == F ·dX + IldX11 7l(dX)
== V fP ·dX + IldX11 7l(dX) , (3-54)

where the reference position Po and time t are not any more denoted.
The transported vector of dX is expressed with respect to a (general) curvilinear basis as
follows:

where
(3-55)

(ii) The volume of the parallelepiped of sides corresponding to vector elements

dU, dV and dW in Co is
dpo == dUl\dV.dW.
Since the components of the transported vectors du== F-dU, dv ==F-dV and dw==F·dW
are respectively
dv i ==FidV J dw i ==Fi·dW J
} ' } '
we have:

By considering the determinant of each matrix we obtain the expression of the transported
volume element, namely:
dp == detF dpo
or
(3-56)
where J is the ratio of volumes, which is equal to 1 if the volume is preserved between the
configurations Co and Ct.

(iii) Now let us consider the convective transport of an area element defined as follows.

Given a material element of a surface at Po determined by two tangent vectors denoted by

d l X and d 2 X, we recall that the bivector d l X 1\ d 2 X defines an element of area
formed from these vectors.
198 Chapter 3

Explicitly, the area element corresponds to a tensor of type ( 6) whose components with
respect to a system of (general) curvilinear coordinates (qi) are

dSij = d]qid2qJ-d]qid2qi

d] qi d] qJ
i,j = 1,2,3.
d 2 q' d 2q J

We introduce the adjoint of this tensor, that is, a covector whose components are

dS k = (Jlg)u k dS u = ~detg eijk dSij . (3-57)

In an orthonormal Cartesian basis we again find the classic notion of the area vector d4 of
components indifferently denoted by dS k or dS k .
Its squared norm is

da
dA
N
--
---
- --

Fig. 18

We denote the area vector d] X /\ d 2 X as follows:

that is, the vector orthogonal to the surface element whose norm dA is the area of the
parallelogram of sides corresponding to d 1 X and d 2 X (called the undeformed area).
Given a convective transport, to this area vector dA corresponds the homologous area
vector in Ct :

da = dain,
that is, orthogonal to the surface element and whose norm da is the area of the
homologous parallelogram of sides corresponding to d 1 x and d 2 x, called the deformed
area.
 Deformations 199

The volume generated by the previous bivector element and an arbitrary vector V in Co is
,uo(dA,V) = dA.v

and the homologous volume obtained by the convective transport in Ct is

,u(da,v) = da.v.
We suppose that the orientation of surface elements indicated by the unit normal vectors is
preserved.
The well known expression of the convective transport of a volume, that is:
,u(d a,v) = detF ,uo(d A,V)
is written:
da.v = J dA.v
or
F .V. da = J V. dA .

Thus the convective transport of the area vector is expressed as

(3-58)

2.2.2 Stretches

Let dV, dW be vectors at Po in Co ,

dv, dw be the transported vectors in Ct.

(i) We have:
dv.dw = F-dV.F-dW = dV. t F-F-dW
and thus
dv.dw = dV.C.dW . (3-59)

Let us make explicit the (right) Cauchy-Green deformation tensor C.

We know that

is written with respect to an orthonormal Cartesian basis as follows:

C - 6 8xP 8x q _ 8xq 8x q
if - pq ax' ax} - ax' ax}
(3-60)

since the position of indices is indifferent.

200 Chapter 3

(ii) Likewise, the components of the left Cauchy-Green deformation tensor C L = Po' F
are

and, more particularly with respect to an orthonormal Cartesian basis, they are written:

( L ).
By considering the inverse tensor which we recall to be
c = (£0 1£)-1,
we have:
dV.dW = F- 1·dv.F-1·dw = dv.'F-l ·F-1·dw = dV.C~1 ·dw
= dv.c·dw. (3-61)

The tensor c is more particularly used in the current configuration (as the Euler-Almansi
tensor) and its components are
ax ax
cpq -- --'--' (3-62)
axp aX q

(iii) The stretch in the direction of dV is the positive real

A(dV) = Ildvll = ~C(dV,dV)

IldVll IldVll

By considering
dV = dso lv, dv = dslv ,
the previous ratio of norms becomes:

A(dV) =~
dso
or

_ IIF ·dVll _IIF ·dso lvii_II . II

A(dV) - IldVll - ds o - F Iv .

The stretch in the direction of lv, equal to that in the direction of dV, is written:

(3-63)
 Deformations 201

(iv) The shear angle for a pair of orthogonal directions of initial vectors dVand dW is such
that
sine = C(dV,dW) C(lv,l w )
(3-64)
~C(dV,dV) ~C(dW,dW) A(lv ) A(1w )

since, given dV = dS olv, dW = dS61w , we have:

C(dV,dW) = dsodS61v·C-1w, C(dV,dV) = (ds O)2 C(lv,lv) ,

C(dW,dW) = ( ds6F C(1w,lw)·

2.2.3 Strain

Let dV, dW be vectors at Po in Co,

dv, dw be the transported vectors in Ct.
The Green-Lagrange strain tensor L is such that
dv.dw-dVdW = dV.(C-/)·dW,
that is
dv.dw-dVdW = 2dVL·dW. (3-65)

Explicitly this difference is equal to 2 Li} dV' dW J.

From
dv = dv Iv,
and by recalling
ds o = II dVll '
we deduce:
ds 2 -ds5 = dv.dv-dV.dV
= 2dV.L·dV. (3-66)

This is explicitly written:

(3-66')

We also have
(3-67)

where the real L(1v ,Iv) is the strain in the direction of Iv.
202 Chapter 3

The relationship between the strain in the direction of lv and the corresponding stretch
},(1v) is

(3-68)

and thus
(3-69)

The unit elongation in the direction of lv is

(3-70)

The Euler-Almansi strain tensor is introduced by using Eq. (3-61) as follows:

dv.dw -dV.dW = dv.(l-c)·dw,
that is,
dv.dw- dV.dW = 2 dv.e·dw. (3-71 )

PR14 The relations between the Green-Lagrange tensor and the Euler-Almansi tensor
are
L= tF·e·F,
e = tF-I.L·F-'. (3-72)

Proof The comparison between Eqs. (3-65) and (3-71) leads to

dV.L·dW = dv.e·dw = dV.tF-e·F-dW,
that is,
L(dV,dW) = tF-e·F(dV,dW).

Since this equality holds for every pair of vectors dVand dW the proposition is proved.

Given an orthonormal basis, we have:

( I)
q
(3-73)

and
e -1(t5- oXq oXq )
I) - 2 I) ax ax·
I )
I ).
q
(3-74)
 Deformations 203

We immediately have:

(3-75)

- ax, -a
eij--a axs L rs· (3-76)
X; Xj

Training. Given the vectors

dV = ds olv, dW = dsb lw
and the respective transported vectors
dv=dslv , dw=ds'lw,
we then have:

We also have:
dv.dv-dV.dV = 2dv.e·dv
dS2-ds~
--2=--= 2Iv ·e·lv = 2e(lv,lv)' (3-77)
ds

2.3 DISPLACEMENT AND GRADIENT

Expressions of previous tensors can be viewed by considering the displacement

U(X,t) of a particle located by X at Po in Co and at p in the final configuration Ct.
We recall that this displacement is the vector
U(X,t) = PoP = x-X (2-3)

which connects two positions, at different instants, in the frame of reference ~.

In Co each displacement is a function of the initial position of the particle and thus a field
of such vectors can be considered.
A particle at a neighboring position located at X + dX occupies a position x + dx in Ct such
that
x+dx = X+dX+U(X+dX,t). (3-78)

From Eqs. (2-3) and (3-78) we deduce:

dx = dX + U(X+dX, t) - U(X,t) .
204 Chapter 3

The following figure shows previous vectors.

Fig. 19

2.3.1 Material Displacement Gradient

D CiF The material displacement gradient is the vector VU (X,I) such that

U(X + dX, I) - U(X,I) = VU (X,I) . dX (3-79)

This is equivalent to
dx - dX = VU(X,I)· dX,
that is:
dU(X,I) = VU(X,t)·dX. (3-80)

Since
dx = dX + VU . dX , (3-81 )
explicitly written:
(3-82)
and since
dx=F·dX,
we deduce the following expression of the deformation tensor F =V f/J (for X and time t):
F= /+ VU. (3-83)

Thus the material displacement gradient is expressed as follows:

GV' VU= F-/. (3-84)
 Deformations 205

We know that every tensor may be decomposed into a symmetric part and an
anti symmetric part.
Thus the displacement gradient is written:

VU =8 +£0 (3-85)
where the following tensors are defined:

D The linearized strain tensor I is the symmetric tensor

(3-86)

D The linearized rotation tensor 2 is the anti symmetric tensor

£0 = ~(VU- tVU). (3-87)

2

Thus we have
F = 1+8+£0. (3-88)

The tensors C and L are expressed in function of V U as follows:

C = t(I+ VU)·(l+VU)
=1+2e+ tVU·VU (3-89)

and
I t
L =8 + - VU·VU
2

= ~(VU + tvu + tVU· VU). (3-90)

2

Remark. The linearized strain tensor 8 plays a fundamental role in continuum

mechanics, notably in linear elasticity. Its very simple expression in the Green-Lagrange
strain tensor L justifies the choice of the factor 12 in the definition of L.

From Eq. (3-82), by considering orthonormal (Cartesian) bases we deduce that the
components of the material displacement gradient are written

au
(VU)i} = a; }

= ax, -5. (3-91)

ax }
I)'

I Called in French: Tenseur des deformations limiarise '.

2 Called in French: 'Tenseur de rotation limfarise '.
206 Chapter 3

By considering the linearized tensors & and OJ, we also have:

(3-92)

where

&
1 au
- -(--'+--)
au}
(3-93)
if - 2 axJ ax,

(3-94)

(L ) (3-95)

The last expression shows the non-linearity between displacements and deformation
fields.

The reader will make explicit the previous expressions; for instance:

au! ~(au! + au2) 1(au!

--+-)
au)
ax! 2 aX2 ax! 2 ax) ax!
&=
1 au!
-(-+-)
aU 2 aU 2 1(au 2
--+-)
au)
2 aX2 ax! aX2 2 ax) aX2
~(au! + au)) ~(au2 + au)) aU 3
2 ax) ax! 2 ax) aX2 ax)
and

and so on.

2.3.2 Spatial Displacement Gradient

Let us introduce the spatial displacement gradient with a new notation which
specifies partial derivatives with respect to Eulerian variables.

D (iF In the configuration C1 the spatial displacement gradient is the vector gradU (x,t)
such that
U(X +dX,t)-U(X,t) = gradU(x,t)·dx. (3-96)
 Deformations 207

This definition is well founded. Indeed, we know that

U(X+dX,t)-U(X,t) = VU(X,t)·dX
= gradU(x,t)·Vf/J·dX.

This last equality, which immediately leads to Eq. (3-96), is explicitly obvious since
au' aufaxk
ax) = axk ax) .

To conclude this section let us consider the components of the spatial displacement
gradient given an orthonormal basis, namely:

(gradU)ij
au I
ax
= - a = liij--a I .

Xj xi

By introducing the spatial deformation gradient gradX with components ~:I we have:
J
gradU = I-gradX. (3-97)

We can now deduce the components of the Euler-Almansi tensor, namely:

eij ="2I (Jij - aX

ax aX -"2I [Ii (Ii aU r )(Iir) --a
aU r )]
r r ) _
ax lj - rt - -a
I J Xi Xl

These components are those of the following Euler-Almansi strain tensor

I
e = -(gradU +1 gradU - IgradU.gradU). (3-98)
2

By analogy with the linearized tensor & we can introduce the following tensor

"& = ~(gradU + IgradU) (3-99)

2
and thus
e ="& -~ IgradU·gradU. (3-100)
2
208 Chapter 3

2.3.3 Curvilinear Coordinate System

Given a system of (general) curvilinear coordinates (q'), the partial derivatives of

the displacement vector U = x - X are
au ax ax

By letting
ax ax
ei =-·, e,'
aq' aq'
we have:
auo

ei = e, + aqi '

The corresponding metric tensors are

0 au) (0 au)
gil = ei .e l = ( ei +--.. ej +--.
aq' aqJ
o 0 0 au 0 au au au
== e.e + e . - + e · - + - .. - -
I J aqJ I1 aq' aq' aqJ
and
o 0 0au ) (( au )
gij == ei .eJ == e, - aqi . ej - aql

au au au au
==e,.e -ei·---e . - . +-.'--'
J aqJ J aq' aq' aqJ

We have:

Since
a a (Ur 0) _ aUr 0 aer Ur _ ( au r rr US) 0 0

- U. -_ - e - --e + - - --+ e
aq' aq' r aq' r aq' aq' r IS

== V;U r e;,
where we recall that the various V; Ur are the components of the (! )-covariant derivative
of U, then we have

2"1 (glj - gij0) == 2"1( ei0·VoUr 0 0 DoUr 0 voUr 0DOUS 0)

J er + eJ . vier + , er · V J e,

1 ( ° DoUr
== - g ir V j + g jr DoUr
V i + g rs DoUr
0
V i
DOUS)
V j
2
 Deformations 209

From

and since at point (defined by X) of local coordinates q; we have

we deduce:
(3-101)

Given an orthonormal Cartesian basis we again find:

Li) = 2.( au; + aU j + au s aU S ).

(3-102)
2 aX j ax; ax; aX j

We note that
1 0 1
-(gi)- gi/) = -[e; .ej -(e; - - . ).(ej
au au
- -)]
2 2 aq' aqJ
1 au au
= -(e; .--. + e . - . - -
au.. _-)
au
2 aqJ J aq' aq' aqJ

We have obtained the components

eij = 2.(VPi + ViU; - V,U s VPS). (3-103)

2

of the Euler-Almansi tensor which are expressed with respect to an orthonormal

Cartesian basis as follows:

(3-104)
210 Chapter 3

3. INFINITESIMAL TRANSFORMATION

The measure of deformation is obtained from the well known difference ds 2 - ds~
which is expressed in terms of L and thus as a function of VU as shown by Eq. (3-90).
Our goal consists in neglecting the nonlinear terms which follow from the presence of
t VU. VU in the expressions of C and L.

The 'small deformation theory' of continuum mechanics requires that the displacement
gradients be small compared to unity. More precisely, we say:

D A transformation between configurations Co and Ct , in a frame of reference en, is

said to be an infinitesimal transformation if the norm of the (dimensionless)
displacement gradient is very small compared to unity, which is written:
IIVUII « 1,

for every X in Co .

In particular, given a Cartesian coordinate system the terms

are small compared to unity, thus the products of such terms will be negligible and
omitted.

Notation. The equalities following from the 'process of linearization' defined above will
be denoted by the symbol == .

Before showing the consequences of this 'linearization' we note the following remarks.

Remark 1. Every infinitesimal transformation implies that the corresponding deformation

is infinitesimal. Indeed, the hypothesis

implies that the components Lij of L = ~(VU + t VU + t VU . VU) are infinitesimal.

2

The example of rigid body transformations proves that the converse is false.

Remark 2. The reader will take care to distinguish clearly the notions of infinitesimal
deformation and infinitesimal displacement.
 Deformations 211

The hypothesis of small displacements, mathematically called 'infinitesimal

displacements' means that the norm of such any displacement is very small with respect to
a reference length 10 in the configuration Co, that is:

sup
x
M«
10
I.

For instance, the reference length 10 may be the measure of the minimum displacement for
which the medium is not elastic any longer.

Remark 3. On the one hand, we emphasize that infinitesimal displacements do not

necessarily imply infinitesimal transformations (and thus infinitesimal deformations) since
the derivatives aPi are not necessarily infinitesimal, whereas the various Ui of U are
infinitesimal.
On the other hand, if the conditions of infinitesimal displacements, infinitesimal
transformations, and infinitesimal deformations simultaneously satisfy a problem, then it
is said that the problem satisfies the 'hypothesis of small perturbations '.
Of course, very small displacements and very small deformations are not always
'equivalent'. For example, problems of fissures show important deformations for very
small displacements; conversely, a bending of beams can show important displacements
for small deformations.

3.1 TENSOR NOTIONS RELATING TO INFINITESIMAL TRANSFORMATIONS

PRI5 Material and spatial displacement gradients are 'very nearly equal' for every
infinitesimal transformation.

Proof We have:
VU(X,t) = gradU (x,t)·VlPt(X)
= gradU(x,t)·F; (X)
= grad U (x,t)· (I + VU (X,t)).
Since V U is very small compared to unity for infinitesimal transformations, we conclude
that
V U (X,t) == grad U (x,t). (3-105)

PR 16 The Green-Lagrange strain tensor L and the linearized strain tensor Ii are 'very
nearly equal' for every infinitesimal transformation.

Proof From (3-90) we deduce

212 Chapter 3

L == ~(VU+tvU),
2
that is
(3-106)

Dw The linearized strain tensor 6 is called the infinitesimal strain tensor.

It is the symmetric part of the displacement gradient and by considering orthonormal

Cartesian bases its components are

Likewise we say:

Dw The linearized rotation tensor (j) is called the infinitesimal rotation tensor.

PR17 The Euler-Almansi tensor e and the infinitesimal strain tensor 6 are 'very nearly
equal' for every infinitesimal transformation.

Proof From (3-98) we deduce:

1
e == -(gradU+ t gradU)
2
and thus from (3-105) we conclude that
(3-107)

Likewise, for every infinitesimal transformation we have the following proposition:

PRl8 The Cauchy-Green deformation tensor is such that

C == /+26. (3-108)

Proof It is obvious because we have

26 == 2L = C-/.

Given a (general) curvilinear coordinate system (q\ the components of the infinitesimal
strain tensor 6 are

(3-109a)
 Deformations 213

and are 'very nearly equal' to

I
eij = 2(Y'P;+Y';U j ). (3-109b)

By considering orthonormal Cartesian bases we can compare

(3-110a)

with

(3-110b)

by knowing that the components

au; and
au; of respective material and spatial
aXj aXj
displacement gradients are 'very nearly equal'.

Let us now consider the notions of stretch and unit elongation in the context of
infinitesimal transformations.

The stretch in the direction of lv

is such that
(3-111)

The unit elongation in the direction of V is such that

(3-112)

and the unit elongation in the direction of a basis vector e; IS

(no summation). (3-113)

PR19 If G; denotes the eigenvalue of G associated with the principal axis defined by it;,
then the corresponding unit elongation is such that
(3-114)

Proof The principal stretch corresponding to it; is

214 Chapter 3

and thus the principal unit elongation is

b; = b(E;) =c;,
that is

This result is confirmed by the property

L =c.

PR20 The ratio of volumes J = det F = det (I + VU) is such that

J=I+tre
= 1+ trVU. (3-115)

Proof From PR6 we know that J is the product of three principal stretches

and thus
J = 1+ LC; = 1+ tre = 1+ trVU,
i

by recalling that
trVU = tree + w) = tre.
More particularly, given an orthonormal basis this is written:

J=det(I+VU)=I+--
l + __
2+ __
au au
3 =1+trVU,
au
aXl aX2 aX3
where trVU is the divergence of U(X,t) with respect to Xi, denoted by DivU.

Let us introduce the following definition.

D The change in volume per unit original volume is called the dilatation, that is:
dp-dpo
dpo

PR21 The dilatation is connected to the displacement and to the infinitesimal strain
tensor as

dp-dpo - D' U
dpo = IV
=tre. (3-116)

Proof It is obvious from the well known equality dJ.1- dJ.1o = J -1 and from (3-115).
dJ.1o
 Deformations 215

PR22 The shear angle for a pair of directions of orthonormal basis e; and ej is such that

(3-117)

Proof Eq. (3-64) is written with initially orthonormal vectors e; and ej as follows:
C(e ,e)
sinO = I J .
2(e; )2(e)

Since for orthonormal vectors we have

2L(e;,e) = 2e;Lej = e;.(C-l)·e j = C(e;,e)
and by taking Eq. (3-69) into account we find again Eq. (3-27):
. 2L(e; ,e j )
smO = -----;======-r='=====
~ 1+ 2L(e; ,e;) ~ 1+ 2L(e j ,e j )
2Lij

Thus for infinitesimal transformations we have

that is
o =: 2&(e;,e) = 2&ij.

PR23 (i) The infinitesimal strain tensor & is connected with the pure stretch tensor S as
follows:
S =: 1+&. (3-118)
(ii) The infinitesimal rotation tensor {jJ is connected with the rotation tensor R as
follows:
R =: 1+{jJ. (3-119)
(iii) The rotation and pure stretch tensors commute:
R-S=:S·R (andT·R =: RoT). (3-120)

Proof (i) The relation (3-118) is obvious since

S2=C =: 1+2&.

(ii) The relationship (3-119) is satisfied if we prove that the tensor q such that R =: 1 +q
is {jJ. It is true because we have VS :
216 Chapter 3

R·S =(I + 0·(1 + s) ==d +s+';,

R·S = F= I+VU =I+s+m,
and thus
R =I+m.
(iii) We have:
S·R =(/+6)·(/+m) =I+s+m = R·S.
We note that
T = F· R- 1 =(I + 6 + m)· (I - m) =1 + =S .
S

3.2 COMPATIBILITY CONDITIONS

Eq. (3-90) allows us to obtain the Green-Lagrange strain tensor L from the
displacement field U(X,/). In other words, the three displacement functions Uk determine
the six components Lij of the (symmetric) strain tensor L.

Conversely, given a (symmetric) tensor field L in Co does a corresponding displacement

field U(X,/) exist?
We restrict our consideration within the context of infinitesimal transformations; this
restriction is of no consequence and this context is the most interesting, notably in linear
elasticity. Thus if the six components of the infinitesimal strain tensor s are given
arbitrarily, do three displacement functions Uk exist?

Explicitly, the question is whether the three components of U satisfy the six conditions (3-
93), namely:

This system of equations is over-determined, that is, there are more equations than
unknowns Ui , and there is generally no solution for an arbitrary choice of strain
components 6ij'
So, for instance, given

we have
 Deformations 217

and thus

arp alj/ 2 _
aX3 (X] ,X3) + aX2 (X] ,X2 ) + 3(X2 ) X3 - o.
Since the type of the last tenu of this equation cannot be that of the first two, we conclude
there is no displacement field U for the chosen Bij.

In general the displacement field components U; do not satisfy the six conditions (3-93)!
We must impose further conditions upon strain components Bij called the compatibility
conditions, which are often called the Saint-Venant compatibility equations. They are
valid for simply connected domains (which are such that every closed curve remaining
within the domain can continuously be shrunk to a point).

We recall that the displacement gradient

V U(X,t) = B(X,t) +aJ(X,t)

is explicitly written:

More particularly, given an orthononual Cartesian basis we denote:

1
B ij = l(U,,} + U j,')

1
OJlj = l(U,.] - U j,').

The displacement field U is obtained from VU by integrating the following differential

fonus

provided that

or equivalently:

How shall we choose the various Bij so that these conditions define the corresponding OJij?
218 Chapter 3

The partial derivatives of OJi} :

I a2 u _a_2~uJ,-._) _ ~(U . -U )
i

OJy,k =2(axax
} k
ax axk - 2 I,;k j,ikI

are also written:

I
OJI),k = 2(Ui,;k + Uk,1) - Uk,y - U),ik)
I
= 2[(Ui,k),) +(Uk,,),) -(Uk,),i -(Uj,k),J,

that is
(3-121)

These equations connect the partial derivatives of the tensors OJ and Ii , they determine the
partial derivatives of OJi} •

Requested conditions are such that (l:ik,) - l:)k, i) dXk are the differentials of various OJi}' In
other words, necessary and sufficient conditions of integrability of the system of Eqs. (3-
121) are

(3-122)

Given these 34 = 81 equations of which six are different, then the system (3-121) leads to
the components of OJ from those of l: •

To summarize, given the symmetric infinitesimal strain tensor Ii such that the
compatibility conditions are satisfied, then the differential of every component of the
infinitesimal rotation tensor OJ , namely:

dOJ y = OJij,k dXk

is integrable and thus the various components au i of V U = Ii + liJ are determined,
ax}
We specify that the knowledge of VU allows us to obtain the (unknown) components of
U because the system

is integrable since
Ui,;m = l:y,m + OJy,m = l:y,m +l:im,; -l:;m,i

= l:irn,) +l:y,m -linlj,i = l:lm,) +OJlm,j

= Ui,m;.
 Deformations 219

Thus the displacement field U is obtained from V U without new condition and from the
integration of the differential forms
(3-123)

Remark 1. The compatibility equations (3-122) are necessary and sufficient so that 8 is
the symmetric part of VU within a simply connected domain. This is a system of 6
equations with 6 unknowns, but these equations are not independent.
By denoting

this system is written:

t:1I ,22 + t:22 ,11 == 2t:12 ,12 (3-124a)

822,33 + 833,22 == 2£23,23 (3-124b)
8 33,11 +811 ,33 == 2£31,31 (3-124c)
-£23,11 +£31,12 +£12,13 == t:11 ,23 (3-124d)
-t:3 l,22 +£12,23 +823,21 == t:22 ,31 (3-1 24e)
-£12,33 +£23,31 +£31,32 == £33,12' (3-124t)

Remark 2. We note that only second order derivatives are present in the system of linear
and homogeneous partial differential equations (3-122). Thus these compatibility
conditions are automatically satisfied if the components of 8 are affine functions of
variables Xi,

Remark3. The reader will verify that, given a tensor field of components Tij such that

then the compatibility conditions are simply written with the infinitesimal strain tensor 8
as follows:
curl(curl'£) == 0== curl (' curl 18) == curl ('curI8)

and similarly:
curl'(curI8) == 0 == curl'(curl 18) == curl'('curI'8) .
220 Chapter 3

3.3 RIGID BODY TRANSFORMATION

Let us consider a pair of neighboring points defining a vector Poqo in Co.

First, given an arbitrary infinitesimal transformation let us prove the following.

PR24 The displacement qos joining the corresponding positions qo (in Co) and s (in CD
is composed of
- a translation of vector k = qoq = U(po,t),
- a pure stretch deformation generated by the stretch tensor S,
- a rotation defined by m = ±curl U(po,t).

Proof We have
U(qo,t) = qoq+qr+rs
= k + (pr-pq) + (ps-pr)
= k + (S-l)·pq + (R-l)·pr
and thus

Fig. 20

Since for small deformations we have:

S-I='6,
(R-I)-S =. m·(I +6) =. m,
we deduce that
(3-125)

We conclude that the displacement U(qo,t) is composed of a translation, a pure stretch

deformation, and a rotation.
 Deformations 221

In connection with this rotation, we note that the adjoint of OJ is a covector whose first
Cartesian component is
1 1 au; 8U) 1
OJ I = -elU - ( - - - - - ) = -[eI23 (8 2 U3 - 8 3 U2 )+el32 (8 3U 2 - 8 2 U3 )]
2 2 ax) 8X; 4

and so on.

Thus the kth component is expressed as

OJ k = t( curl U)k . (3-126)

It represents the vector of rotation in 3-dimensional Euclidean space.

Secondly, in the special case of rigid body transformations we have

e == L = O.

In this homogeneous transformation the displacement is characterized as follows:

(3-127)
222 Chapter 3

EXERCISES

Exercise l.

Let

be a transfonnation between two configurations Co and CJ •

Given an orthononnal basis (ei) of a frame of reference, we denote by Xi and Xi the
Cartesian coordinates of a particle respectively with respect to Co and CI.
(i) Prove that this defonnation is an example of a homogeneous transfonnation.
(ii) Detennine the components of the displacement U in Lagrangian variables (or
material coordinates) and Eulerian variables (or spatial coordinates).
(iii) Locate in C1 the set of particles which initially fonned the material circle of
equations
X3=O.
(iv) Express the Cauchy-Green defonnation tensor C and calculate the principal
stretches Yi.
(v) Find the Green-Lagrange strain tensor L and the principal strains L i •
au i
(vi) Prove that the tensor VU of components - - . verifies the following equality
aX]

L = 2.(VU + 'VU+ 'VU· VU) .

2
(vii) Detennine the shear angle () for a pair of orthononnal vectors el and e 2 • Verify the
result by considering the transported vectors of these vectors.
(viii) Given the vector X = Xl el + X2 e2 + X3 e3 , prove that the transported vector x of X
verifies the following fonnulae
x.x =CrsX rX s
x.x- X.X = 2Lrs X r X s •

Answer. (i) It is a homogeneous transfonnation because we actually have a relation of type

Xi = F}X J, that is, explicitly:
 Deformations 223

(ii) The components of the displacement U=X - X with respect to the Lagrangian
variables are
u' =X' +.fi. X2_X' =.fi. X2
2 2'

U2 = .fi. X' +X2 _X2 = .fi. X'

2 2 '

Thus the displacement is expressed as follows:

U = .fi. X2 e l + fi Xl e2 .
2 2
To find the components of the displacement with respect to the Eulerian variables, let us
express the Lagrangian variables in function of the Eulerian variables, that is:
X' = 2x'- fiX2
X2 =2X2_fix'
X3 =x3
and thus the components of the displacement U with respect to the Eulerian variables are
U' = Xl_Xl = fix 2 -x',
U 2 =x 2 _X 2 = fix'-x 2,
U3 = x 3 _ X 3 = o.
Thus the displacement is written:
u=(fix 2 -x')e, +(fix'-x 2 )e 2 ·

(iii) Any particle keeps the same material coordinates Xi for the transformation; in other
words, the equations
(X')2 + (X2)2 = 2 ,
are preserved after the transformation.
These equations are expressed in Eulerian variables as follows:

They define an ellipse having its center at the origin of the frame of reference and the
principal directions of which are obtained from the following equation
3-A
det [ -2fi]= O.
-2fi 3-A
224 Chapter 3

For the eigenvalue A =3-2..fi we deduce that

3 -2..fij [XI)_ (3-2..fi) [Xl)

[-2..fi 2 - 2 '
3 X X

that is

It is the first bisector.

For the eigenvalue A = 3+ 2..fi we deduce that

3 -2..fij
[-2..fi 3
(Xl J
X2
= r;; (Xl ,
(3+2,,2)
x2
J
that is

It is the second bisector.

For the ellipse axis of equation Xl =X2 the semi-major axis is equal to (6-4.firI/ 2
(because 6(XI)2 - 4.fi(XI)2 =1).
For the ellipse axis of equation Xl = - X2 the semi-minor axis is equal to (6 + 4.fi) -1/2
(because 6(XI)2 + 4.fi(XI)2 = 1).

(iv) Since the tensor F is explicitly the following

then the Cauchy-Green deformation tensor C is expressed as

3/2 .fi

C=
[
~ 3/2
o

The principal directions are obtained from the following eigenvalues

A'=t+..fi, A"=t-..fi, A"'= 1.

For the value A' we have

(- ~-.fi)c
2
=0
3'
 Deformations 225

and the corresponding vector of the principal basis is

~ 1 1
EI = .fi e l + .fi e 2 ·

Thus the squared stretch in the principal direction of EI is

~(,,)' ~ E, cE, ~(J2 J, lr~ l [I/.fi:

.fi
o
Y, 0 3/2
o ~J I/~
= t+.n·
For the value A" the vector of the principal basis is
~ I I
E2 = .fi e l - .fi e 2

and the squared stretch in the corresponding principal direction of E2 is

For the value A'" the vector of the principal basis is

E3 =e3

and the corresponding squared stretch is

Y3=(A 3 )2=1,

which is obvious since there is no deformation in this last principal direction.

(v) The Green-Lagrange strain tensor is

l
that is explicitly:

V4 .n/2 OJ
L= .n/2 1/4 0 .
o 0 0

From

we deduce the following values of principal strains:

T
~
=~_.n
4 2'
T
~
=0.
226 Chapter 3

We note that these values actually verify the characteristic equation:

A[(~-A)2_~]=O.
4 2

(vi) From

we deduce:
-fi12
1/4

°
(vii) The required shear angle B is such that

sinB = 2LI2 = 'l..-fi.

~1+2LII ~1+2L22 3

Since the transported vector of e, is the vector

and since the transported vector of e2 is the vector

[fil2I fil21 OJ0· [0]1 = [fiI2]

1 ,

° °1 ° °
then the angle between these vectors, denoted by ~ -B , is such that

J( -fi 2 r::
cos(--B)= hh ~=-v2.
2 ,>/3/2,>/3/2 3

This is actually the previous sinB.

 Deformations 227

(viii) On the one hand, the vector

20
1 Ji/ 1 [Xl]
x = [ Ji/2 1 O· X2 =
o 0 1 X2

leads to

On the other hand, we have

C,sX' X s =X.C·X

~XI +fix 2
2

= (XI X2 X3). fi Xl +~X2 ,

2

which is actually X.x.

We also have

and we deduce that

is actually the previous result and thus the second formula of (viii) is established.
228 Chapter 3

Exercise 2.

Given the transformation of a system between two configurations Co and Ct defined

at time t=l, by

calculate the tensors F, S, R, T, the principal stretches, principal strains and principal unit
elongations.

Answer. We have

[0-2 0]
F= 3 0 0 ,
o0 I

9 0 0'
040 3 0 0]
S2 = C= [ => S=020,[
001 001

R = FS-l = I 0 0 , [0-I 0]
001

The vectors of the principal basis are

By considering C (= S2) we deduce that the squared principal stretches are

(~F =9,

Of course, these values correspond to the following

Thus the principal stretches are

~=3, ~=2, A:J=1.

The principal strains Li = teA; -I) are the following eigenvalues of L = t(C - /):
 Deformations 229

The principal unit elongations are

Exercise 3.

Let us consider the transformation defined by

XI =X2 , X 2 =-XI , X3 =2X3 .

(i) Calculate the deformed volume of a cube with unit edges (10-2 m) in the first octant.
(ii) Find the deformed area vectors of the left, front and lower faces.
(iii) Express and characterize the right pure stretch tensor S and the rotation tensor R
which decompose F.

Answer. (i) From

[ 0 1 OJ
F= -1 0 0 ,
o 0 2
we deduce that
J=detF=2.
Thus the deformed volume is equal to

J.1 = 2J.1o = 2 ( 10-6 m3 ).

(ii) The unit vector corresponding to the face of the coordinate plane OXI X3 is -e2 and
SInce

F-I= 1 0 [0-1 0: 0 ,
o 0 112

then by recalling (3-58) the area vector of this face is expressed as

a = a (-e 2 ) = 2[~1 ~ ~]A [~11 -2e = l (l0-4 m2),

o 0 V2 0

since J = 2 and A =1.

Likewise we find that the area vector of the front face (corresponding to el ) is a = -2 e2
and that of the lower face (corresponding to -e3 ) is a = -e3 .
230 Chapter 3

(iii) By calculating

S2=C= IF-F= 0 1 0 , [100]

o0 4

we remark that there is a pure stretch Il.:J =2 in the direction of e3 •

The tensor

R=F-S-I = [-10010]0
o0 1

is associated with a rotation of 90° about e3 •

Exercise 4.

We consider the following transformation between two configurations Co and CI

referred to the same frame of orthonormal basis (ei ):

where the dimensions of the constants k and / are respectively [L-1r l ] and [L].

(i) Find the positions of particles initially located at a o (0,0,0), bo (0,0,-/), Co (0,0,-1/2)
and do (0, I ,-1/2), at given time t.
Determine the positions of the straight segments Clabo and codo at time t.

(U) Let us consider a pair of neighboring points (Po,qo) such that Po is (0,X2,X3) and qo
is (0,X2+dX 2,X3+dX3). In Co this pair of points defines the vector
dX = dX 2 e 2 + dX 3 e3. Calculate the transported vector dx at time t and deduce the
squared norm ds 2 •
Verify this last result by using the Cauchy-Green deformation tensor.

(iii) In a configuration Cr let us denote by qi the coordinates of a point of

coordinates Xi in the initial configuration Co.

Express the coordinates of this point at time t by referring to the configuration Cr.
 Deformations 231

Answer. (i) The positions of the particles at Go and bo remain unaltered while the particle
located at Co occupies the position c(O,kI 2t/4,-1/2) at time t and the particle at do
occupies the position d(O,l + kI 2t/4,-1/2).

The material segment Gobo is the set of particles whose Lagrangian variables are

We know that the material coordinates Xi of a particle remain unchanged (unlike the
spatial coordinates x}

Since

the segment Gb, which is the transformed segment of Gobo at time t, is then defined by
V x3 E [-1,0]: XI=O, x 2 +ktx3(x 3+l)=0.

Thus the segment Gobo becomes at time t this delimited parabola in the plane Xl =0.

Likewise the material segment codo is the set of particles whose Lagrangian variables are

By introducing the spatial coordinates as previously and since 0 S;X2 S; 1is equivalent to
kl 2t/4 S; x 2S; 1+k12 t/4, then the segment cd, which is the transformed segment of codo at
time t, is defined by

Thus the segment codo becomes at time t the straight segment cd of unit length along the
straight of equations Xl =0, x3 =-t; this is a translation.

(ii) The deformation gradient F is defined by

o
232 Chapter 3

The components of the transported vector dx = F ·dX are expressed as dx ' = F; dX J ,

that is:

[~
Thus the squared norm of this vector is

The Cauchy-Green deformation tensor C = IFF is explicitly

The squared norm of dx is

2 .
ds = dx.dx = Cij dX ' dX J = dX.C ·dX ,

that is explicitly:

which is actually the obtained value of ds 2 .

(iii) From

we deduce:

and thus
Xl =X1 =';1,
X2 = X2 - ktX\X 3 +/) = e +k(T - t).; \.; 3 +/),
X3 = X3 =';3.
 Deformations 233

Exercise 5.

Given an orthononnal Cartesian basis let us consider the following transfonnation

between two configurations Co and CI and where the dimension of a ER+ is [L-Il-
(i) Depict the defonnation_
(ii) Find the stretches in the directions of coordinate axes.
(iii) Find the infinitesimal strain tensor6 and the unit elongations for the elements which
were initially in the directions of basis vectors.
(iv) Give an estimate of the shear angle between orthogonal directions defined by e2 and
e3' at point (0,0,2) originally, if a =10- 5 •

Answer. (i) The displacement components are

There is no displacement in the directions of the axes oXI and oX3 .

Every point of the axis oX2 is fixed, but any material straight interval of the axis oX3 IS

transfonned into a parabolic line segment, that is, { XI = 0, X 2= a (X3)2 }.

Also, a material line {XI = 0, X 2 = a} becomes the parabola {XI =0, X 2= a+a (X3)2 }.

'>-_---_...1.-_ _ _ _ _ X2
e2
el
Xl
Fig. 21
234 Chapter 3

(ii) The displacement gradient is

vu=[~
0

2U:' 1
0
0

and, from F= 1+ V U , we deduce:

2~X' J[~' J= e, ,
0

dx, =FdX, =f: &,

0

dx, =FdX, =f:

0
2a:,lE}&,e"

dx, =FdX, = l: 0

0
2a:, Hl, 1=(m x, e, +e,)&,.
Thus the stretches are

(iii) The infinitesimal strain tensor is

The unit elongation for the element which was in the direction of e] is

and likewise
 Deformations 235

Of course, for very small deformations we have

o(e3) == 2a 2(X3)2

which is confirmed by

(iv) The required shear angle is

() == 2e(e2 ,e3 ) = 2623 = 2a X 3 •
For instance, if a=10-5 and X3 =2 then we obtain

() == 4xlO-5 •

Exercise 6.

Given an orthonormal Cartesian basis let us consider the following transformation

between two configurations Co and Ct.

(i) Express the tensor F of this homogeneous transformation.
(ii) Determine the transported vector of any vector, the transport of a volume, and the
transport of an area vector.
(iii) Find the principal directions of the Cauchy-Green deformation tensor and the
principal stretches.
(iv) Find the principal directions of the Green-Lagrange strain tensor and the principal
strains.
(v) Give the condition of existence of infinitesimal transformations and express the
infinitesimal strain tensor.

Answer. (i) The tensor F is expressed as follows:

F = (1 + kt)e 1 ®e1 +e 2 ®e 2 +e 3 ®e 3 •

(ii) The transported vector of V = Vi e) + V2 e2+ VJ e3 is

l+kt o
v=F·V= f ~
o
236 Chapter 3

The convective transport of a volume Po is

PI =(l+kt)po 0

The transported vector of the unit area vector IN = Nl e 1 + N2 e 2+ N 3 e 3 is the vector

~lEl
o

o
which leads to

Since the norm of the transported area vector n is the previous square root, we deduce that
every area A is transformed into the following

a=A~ NI2+(l+kt)2(N;+N;) 0

(iii) The Cauchy-Green deformation tensor is expressed as follows:

C= IF of= (l+kt)2e) ®e, +e 2 ®e 2 +e3®e 3 0

The eigenvalues of

(l+ktf 0 01
c= r 0
o
1 0
0 1

An eigenvector W corresponding to the first eigenvalue is such that

C oW=(l+kt)2W,
thus a principal direction is defined bye) 0

For the double eigenvalue we recall that every direction perpendicular to e1 is principal.
From

we deduce that the principal stretches are

 Deformations 237

(iv) The Green-Lagrange strain tensor is expressed as follows:

L
1 et
= -(C -J) = (kt+-)e ®e
2
j j •
2 2
The principal directions of L are those of C.

The principal strains are

~] =kt+ k2t2
2
[ that is, ~] = t (C
ll -1) ]
and

(v) The required condition [[vult «1 is equivalent to the following

[kt[« 1
and thus, by considering L, we obtain:
e=kte1 ®e1 •

Exercise 7.

In the classic example of the simple shear deformation between two configurations
Co and C2 defined by

[see (3-4)] we know that the deformation gradient is such that:

12k 0
F= [ 0 I 0 , 1
o 0 I

also written:
F= e j 0e 1 + 2ke 1 ®e 2 + e 2 ®e 2 + e3 ®e 3 •
We also know that the volume per unit original volume is
J =detF=l.
(i) Is the simple shear a homogeneous transformation?

(ii) Find the transported vectors ofthe vectors of the orthonormal Cartesian basis (e;).
Show in a figure the transformation of material lines along the respective vectors e]
and e2 whose origin Po belongs to the plane ox] x 2 •
238 Chapter 3

(iii) Determine the Cauchy-Green deformation tensor C.

(iv) What are the stretches which were in the directions of basis vectors and in the
direction of the bisector of the angle X I OX2 • Determine the corresponding unit
elongations.
(v) Calculate the shear angle for initial orthogonal directions defined by el and e2'
(vi) Find the Green-Lagrange strain tensor L and the strains along the directions of the
basis vectors.
(vii) Find the principal stretches, the principal strains, and the transported vectors of the
eigenvectors E, of L. Determine the principal basis (if;) and the direct principal
basis (i().
(viii) From the right pure stretch tensor S with respect to the principal basis (if;) express
the rotation tensor R with respect to (e i ).
If k = 1 calculate the rotation tensor and the transported vectors of the direct
principal basis (X;) .
Characterize the polar decomposition.
(ix) Calculate the left Cauchy-Green deformation tensor eLand the Euler-Almansi
strain tensor e. How does e22 (for instance) characterize a material line segment?
(x) In the case of very small deformations, that is, for k << 1, show that L ='6.
Find the principal basis associated with 6 and show that it is the 'linear
approximation' of the principal basis associated with L. Determine the unit
elongations along the principal elongations.

Answer. (i) It is a homogeneous transformation since F is not dependent on X.

(ii) The transported vectors of the basis (e; ) are

We note that F is independent of the choice of Po in the initial configuration since the
transformation is homogeneous.
 Deformations 239

I
Po I
I
I
/

Xf/2K
X2__________ p I e 2
9 !
e' ................ :.,
1 e e2 ..........
X, 1 .....
Fig. 22

(iii) The Cauchy-Green deformation tensor is

that is, explicitly:

C==ej ®e j +2k(ej ®e2+e2®ej)+(4e+l)e2®e2+e3®e3
== I+2k(e j ®e2+e2®ej)+4ee2®e2·

(iv) Since the stretch in the direction of some unit basis vector is /l,(e i ) ==..JC: ' we have

These results are confirmed by

Since the unit vector in the bisector direction is

and since
240 Chapter 3

we deduce the following expression from (3-63):

}.(1b) = -he+2k+ I.

The required unit elongations are respectively

o(eJ = -J4e+I-I,

(v) From (3-14) the shear angle between the initial directions of e 1 and e 2 is obtained as
follows:
sinO = C l2 = 2k
JC:: JC;; -J4e+1

This is illustrated in the following figure.

2ket " ...... -- .............. : E2

et

Fig. 23

(vi) From

we deduce that the strains in the directions of basis vectors e are j

Lll = -1 ( ).2 (eJ-I)=O,

2

L22 = ~(}.2 (e 2 ) -I) = 2 e ,

2

L33 = ~(}.2(e3)-I) = O.
2
 Deformations 241

(vii) Since the principal directions indifferently correspond to eigenvectors of Cor L we

prefer to look for those of L.
Thus from
-L k o
k 2e-L o = L(-L2+2k2 L+e) = 0,
o 0 -L
we deduce the following eigenvalues or principal strains:

LI = e + k -Jk2;i ,
L 2 =e-kJe+l '

The eigenvector EI is obtained as follows:

k
e-kJe+l ° 1.[a] = [0]°
-(k2+k~k2+1)
b
co'
°
that is

and likewise:
E2 =e l +(k-~e+l)e2'
E3 = e 3 •

From (3-25) we can obtain the squared principal stretches

(AY = 2k2+ 2k..[k2;1+1 ,

(A2)2 =2e-2k~e+l+1,
(A3)2 = 1,
that are the eigenvalues Yi of C.

The Cauchy-Green deformation tensor with respect to the principal axes is written:

2e +2k~+1 ° °
C= 2e-2k~+1 °
° °
242 Chapter 3

Thus the principal stretches are

A, =..Je+l+k, ~=l.

We find again that

The transported vectors of vectors E are the following orthogonal vectors:

j

F·E I =(l+2e+2k..Je+l)e l + (k+..Je+l)e 2 ,

F·E 2 = (l+2e-2k..Je+l)e l + (k-..Je+1)e 2 ,
F·E 3 =e 3 •

Let us introduce a nice presentation of principal basis vectors.

By letting
1 ~
tana= 1 =-=k+vk-+l
"1 -1.
2
'
we obtain:

Since
1
IIEIII = cosa [cosa >0],

the corresponding (normed) vector of the principal basis is

Likewise, since
-1.2 = cota = ..J k 2 + I - k ,
we have

Since
[sina > 0],

the corresponding normed vector is

£2 =sinael-cosae 2·
The right-handed (or direct) principal basis is formed of the following unit vectors
Xl = cosae l +sinae 2 , X 2 = -£2 = -sinae l + cosae 2 , X3 = e3 •
 Deformations 243

Fig. 24

(viii) With respect to the principal basis (EJ of C we have

that is

The inverse tensor is

=F

n
The rotation tensor R with respect to the basis (eJ is defined by R . S-l (e,).

Since

(w,a
~](:J
sma

:32 = si~a -cosa

0

the following expression

(WW
~]
sina sin a
S(~~) = si~a -cosa 0]~ S(-~,) ('Osi~aW -cosa
0 0
244 Chapter 3

immediately leads to the expression of R with respect to (e,) .

For the value k = 1, we have

E] = e] + (1 + -J2)e 2 "" e] + 2.4142 e 2
=> E] "" 0.3827 e] + 0.923ge 2
and
E2 = e] + (l-.J2)e 2 "" e] -0.4142e 2

=> E2 "" 0.923ge] - 0.3827 e 2 .

We also have

° O~l
0.4142

and thus
°
-I
StEil ~
~ [0.4142
° °
2.4142

° °
We note that

and of course
cosa "" 0.3827, sina "" 0.9239.

Finally, given these values and taking account of the previous expression s(i,) , we obtain
2.121 -0.707 001'
S(e~) [
"" - 0;07 0.707

°
and thus
0.707 0.707
R = F . S(e~) "" [ -.0707 0.707
o 0

which actually represents the 45° rotation tensor.

The transported vectors of the direct principal basis (Ii;) are

 Deformations 245

b2 = F .il 2 "" -0.16e l + 0.38e 2 ,

bj = e j •

The polar decomposition of F exhibits:

- a rotation of 45° ,
-a right pure stretch such that

Al = IIblll "" 2.4142 (expansion along ill)'

~ = IIb2 0.4142 (contraction along il 2 ).
11 ""

(ix) The left Cauchy-Green deformation tensor is

From

we deduce that the Euler-Almansi strain tensor is expressed as follows:

Given a material line segment pq in the e 2 direction in the final configuration, that is,

we have from (3-77):

[see Fig. 25].

246 Chapter 3

l-----\-t
eLl p e,

Fig. 25

Thus we have obtained

(x) The displacement gradient

VU " F - h [ ~ i ~1
2

is made up of a symmetric part, namely:

0 k 0]
Ii= [k ° ° ,
also written:
°°°
Ii = keel ®e 2 +e2 ®el ),
and an anti symmetric part, namely:

Ok 0]
[
OJ=-k ° 0,

also written:
° °°
If Ikl « 1, we have:
 Deformations 247

The principal basis associated with & is obtained as follows.

From
-£ k 0
k -£ 0 =0,
o 0 -£

we deduce that the eigenvalues £i of the infinitesimal strain tensor & are

and the corresponding unit eigenvectors are immediately:

Thus the principal basis is made up of unit vectors in the directions of two bisectors of the
angle between e 1 and e z , and in the e 3 direction.
Thus & is expressed with respect to this principal basis as follows:

k 0 0]
[ o -k 0,
000

that is:

The following (not normed) basis vectors

Kl =E1 =e1 +(k+~e+1)e2'

K z =-E z =-e1 -(k-~kz+1)ez'
K3 = E3 =e 3
are approximated as follows:
Kl == e 1 +ez,
K Z ==ez-e1 ,
K3 = e3 ·

From
At =k+~e+1, A:J =1,
we deduce the following unit elongations

°1 = k+ ~ e + 1 -1, 0z = ~ k Z + 1 - k-l, °3 = O.
248 Chapter 3

Since k is small with respect to 1 in the infinitesimal approximation, we deduce that the
unit elongations are such that:

We have found again the opposite values k and -k which are the nonzero elements of &
(with respect to the principal basis).

Exercise 8.

A displacement field with respect to an orthonormal basis is defined as follows:

U(X) =,u f(X;+ Xi) e l + ,uX2 e 2 +,u X3 e 3

where f is a given function of X; + xi and ,u a real.

(i) Calculate the Green-Lagrange strain tensor and the Euler-Almansi strain tensor.
(ii) If ,u is a real sufficiently small to be able to neglect the powers from the order 2,
prove that the components of the infinitesimal displacement vector in the respective
Lagrangian and Eulerian formulations are equal (in first approximation).
(iii) Show that the infinitesimal strain tensors & and & in respective Lagrangian and
Eulerian descriptions are equal in first approximation at arbitrary same point.
(iv) Check that the functions &ij(X) and &ij(x) are the same in first approximation.

(v) Verify that for infinitesimal transformations we have the following equalities in first
approximation:
e=&.

Answer. (i) From

Xl-Xl = Ul =,uf(x;+xiL
x 2-X2 = U2 = ,uX2'
x3- X 3 =U 3 =,u X 3'
we deduce the Lagrangian variables Xi in function of Eulerian variables Xi:

1
X 2 =--X2 ,
1+,u
1
X3 =--X3 ·
1+,u

Thus in the Eulerian formulation the components of U are

 Deformations 249

Firstly, the material gradient VU whose components are au; is explicitly the following
ax}
2Ji Xd'(X; + Xi) 2Ji X3J'(X; + Xi)]
Ji o .
o Ji

The Green-Lagrange strain tensor is

1 1
L =-(VU + tvu + tVU.VU) = e + - tVU·VU.
2 2

Thus from the linearized strain tensor

JiX 2 J'(X;+ Xi)

Ji
o
and since
o
4Ji2 X;J'2(x;+xi)+Ji 2
4Ji2 X 2Xd,2(X;+xi)

we deduce the following expression of the Green-Lagrange strain tensor

JiXd'(X;+X;)
Ji+ t Ji2 + 2Ji2 X; 1'2 (Xi +X;)
JiXd'(xi+x;)
2Ji2 X 2 Xd,2(xi+ x ;)
1 .
2Ji2 X2Xd'2(X~+X;) J1+ t J12 + 2Ji2 XU,2 (xi +X;)

Secondly, the spatial displacement gradient gradU whose components are

au; IS
ax}
explicitly:
250 Chapter 3

2 22
o pX2 X2/,(X2+ X3)
(1 + p)2 (1 + p)2
gradU= 0 -L
1+ P
o o

The Euler-Almansi strain tensor is

e = ..!..(gradU + tgradU - tgradU· gradU) = & - ..!.. tgradU· gradU.

2 2

Thus, from the linearized strain tensor

2 2 2 2
o PX2 /,(X2+X3) P X3 /'( X2 + X3 )
(1 + p)2 (1 + p)2 (1 + p)2 (1 + p)2

&=
P o
1+ Jl
o Jl
1+ p

and

o o o
I gradU. gradU = 0

we deduce the following expression of the Euler-Almansi strain tensor:

0 P x2 /' Jl X3 /'
(1 + p)2 (1 + p)2
222 2
e= P X2 /' -L_ 1 P _ 2p X2 /,2 - 2p X2X3 /,2
(1 + p)2 1+ P 2 (1 + p)2 (1 + p)4 (1 + p)4
2 2 2
P X3 /' - 2p X2X3 /,2 -L_ 1 Jl2 2p X3 /,2
(1 + p)2 (1 + p)4 1 + P 2 (1+Jl)2 (1 + Jl)4

X2+X2
where /'( 2 32 ) is simply denoted by /'.
(1 + p)
 Deformations 251

(ii) We assume that the function/can be expanded about (xi+xi), that is

We deduce that

We also have

In conclusion we have obtained

U\ =Ji/(xi+xi)+o(Ji 2)=U\ +o(u 2 ),

U2 = Ji x2 + O(Ji2) = U 2 + O(Ji2),
UJ = JixJ+o(Ji2) = UJ +o(Ji2).

Thus we have proved that the components of U in the Lagrangian and Eulerian
formulations are equal (in first approximation).

(iii) Since we know that

x\ =X\+Ji/(Xi+xi), x 2 =(1+Ji)X 2 ,
we can express the components of & in function of coordinates Xi to obtain the following
tensor:

o Ji(l + Ji)X2 f'(X 2+ X2)

(1+Ji)2 2 J

Ji o
1+ Ji

o Ji
1 + Ji

but since

L- = Ji +O(Ji2)
1 + Ji

the previous tensor is written:

252 Chapter 3

f.lXd'(X;+ xi) + O(f.l2) f.lXJ'(X;+ Xi) + O(f.l 2)]

[
f.lXd'(X; +0Xi) + O(f.l2) f.l + O(f.l2)

o
° ,
f.lXd'(X;+ Xi) + O(f.l2) f.l + o (f.l 2 )

whose components are actually those of & in first approximation.

(iv) Since

and finally

then the linearized strain tensor & is the infinitesimal strain tensor, that is:

0 f.lxd'(xi+xi) + o(f.l2) f.lxd'(xi+xi) + O(f.l 2)1

& = [ f.l X2 f'(xi + xi) + O(f.l2) f.l + O(f.l2) O.
f.lX 3 f'(xi+ xi) + O(f.l2) 0 f.l+O(f.l2)

The comparison between this tensor & and the following infinitesimal strain tensor

f.lXd'(xi+xi)
f.l
o
answers the question.

(v) First, the various components of the following tensor

L- & = .lIVU·
2
VU

contain the factor f.l2 . Thus we have

Lij- &ij = O(f.l2) .

Secondly, the various components of the following tensor

e - &= -t IgradU. gradU

 Deformations 253

contain the factor J-L2 . Thus we have

eij- Bij = O(J-L2).

Exercise 9.

Given a simply connected domain let us consider the following infinitesimal strain
tensor

where J-L is a 'sufficiently small' real.

Let us assume that the infinitesimal rotation and displacement vector vanish at the origin.
(i) Determine the infinitesimal rotation tensor.
(ii) Deduce the displacement vector.

Answer. (i) First, we note that the compatibility equations are satisfied since the
components
of li are functions of first degree in Xi'
Let us express the differentials of components of the infinitesimal rotation tensor, namely:
dOJ 12 = (£]k,2 -£2k,])dXk

= (£ll.2 - £2l,! )dX] + (£12,2 - £22,1)dX 2 + (£]3,2 - £23,] )dX3

= (-J-L-J-L)dX 3 = d(-2J-L X 3) ,
d OJ]3 = (£]k,3 - £3k,])dXk

= (£ll,3 -£3],])dX]+(£]2,3 -£32,l)dX 2 +(£\3,3 -£33,])dX3

= -J-LdX 2 = d(-J-LX 2 ),

d OJ23 = (£2k,3 - £3k,2)dXk

= (£2],3 - £3],2 )dX] + (£22,3 - £32,2) dX 2 + (£23,3 - £33,2) dX 3

=J-LdX] =d(J-LX]).

Since OJij(O,O,O) = °we conclude that

OJ12 = -2J-L X3 ,

and thus we have obtained:

254 Chapter 3

(ii) From

we deduce:

[~l=[2
ax
J
flOx3
0

Thus the various differential forms

dU; =(sij +OJij)dXj

are successively written:

dU, =-2flX3dX2-2flX2dX3 =d(-2flX2X3)'
dU 2 = 2flX3 dX, +2flX, dX3 = d(2flX, X 3),
dU 3 = O.

Since U;(O,O,O) = 0 we conclude that the components of the displacement vector are

Exercise 10.

We consider the following infinitesimal strain tensor

0 0 0
X 2X 3 X2
s= 0 fl-[2- 3
-fl-
[2
X2 X 2X 3
0 3
-fl- -fl--
[2 [2

where fl is a 'sufficiently small' real and [ is a constant of dimension [L].

We assume that the infinitesimal rotation and displacement vanish at the origin.
(i) Determine the infinitesimal rotation tensor.
(ii) Deduce the displacement vector.

Answers. (i) We immediately have

Since OJlj (0,0,0) = 0 we deduce that the infinitesimal rotation tensor is such that:
 Deformations 255

° 0 0

(j)= 0 0 L(X2_X2)
2/2 2 3

0 _L(X2_X2) 0
2/2 2 3

(ii) From

we deduce:

o
° °
[:iJ= °
o

Thus the various differential forms dU j can be integrated, and from U(O,O,O) = 0 we
obtain the following components of the displacement vector:
U\ =0,

U2 =4(x/x
2/
-X/) , 3

U3 = :Z2 (_i-X 23_X2 X 32 ).

Exercise 11.

A cylindrical horizontal beam of X2 -axis, of length / and radius R is supported at its

ends. The infinitesimal strain tensor is explicitly
- 2J1(x2-/)x3 -3J1xjx3
E: = [ -3J1xjx 3 8J1(x2-I)x3
o J1(3.5(R 2- X32) - O.5Xj2)

where J1 is a 'sufficiently small' positive real.

(i) Determine the unit elongations at any point of the X 2 -axis in the directions of the
following orthogonal vectors

e 2, f2 (e 2 +e 3 ) '
= -
2
256 Chapter 3

(ii) Given the change of basis ej =a ji e i such that e; = e l ' make explicit the transformed
components e~s of the infinitesimal strain tensor.
Check that the unit elongations in the directions of normed vectors e; and e; are
really e;2 and ei3 respectively.
(iii) Find the infinitesimal rotation tensor (j) which vanishes at point 0 and deduce the
displacement vector vanishing at this point.

e'2

Fig. 26

Answers. (i) Since the equations of X2 -axis are

the unit elongation in the direction of e; is the following positive real

and the unit elongation in the direction of e; is the following negative real
0 0
c5(e;) = (0 - ~ ~} 0 0 [
o 3.5pR 2

(ii) Given the change of basis ej = a jiei we know that

that is, in matrix form:

 Deformations 257

C;, J (a"
a l2 a 21
a,,]
C'T"
&;2

aT"
"12
(&2[
C:' &;2 &~3 = a 21 a 22 a 23 &21 &22 &23 a l2 a 22 a 32

&; I &;2 &33 a 31 a 32 a 33 &31 "32 &33 a l3 a 23 a 33

=a(&;)la.

[Io -fi./2
0 -fi./2
o - -fi./2
0 r
In this question the previous result is written:

0
-fi./2 0 3.5JlR 2
0
0
35~R'l: -;,+(~
0
-fi./2
-fi./2 -fi./2 0
0
3.5Jl R2
0 -3L,]
and thus we obviously have

&;2 = 3.5Jl R2 = J(e;), &;3 = -3.5Jl R2 = J(e;).

(iii) From

we deduce:
d OJij = (&;k.j - & jk.;) dx k

= (a&il _ a&jl)dxl +(a&;2 _ a&j2)dx2 +(a&i3 _ a&j3)dx3 .

aXj ax; aX j ax; aXj ax;

First, we have

and thus

Secondly, we have

and thus

Finally, we have

and thus
258 Chapter 3

Now, from

-l
which is equal to

2Jl(X2 -I) X3
-4Jlxt x 3
- 2Jlxt x 3
8Jl(x2-/)x3
-2/1xt (x2-/)
/1(3.5R 2-6x/-2x/+4x2(x 2-2/)) ,
1
2Jlxt (x 2-/) /1(3.5 R2 - x/ + x t 2- 4x2(X 2- 2/)) -2/1(x2-/)x3

we deduce that

and thus

Likewise we have

dU l = -4Jl Xl X3 dx l + 8Jl(Xl -/)X3 dx l + /1(3.5 Rl - 6X 3l - 2x ll + 4Xl (Xl - 2/))dx 3

and to find Ul we consider

Ul = -4Jl fX I x3 dx l + ¢(xl ,x3) = -2/1 x/ X3 + ¢(Xl ,x3 ),

but
aU l
-=8Jl(x2-/)x3 = -
a¢ =>
1
¢=/1(4x2 x 3 -8Ix l X3 )+'I'(X3 )
aX l aX 2

and thus

But the following,

aU 2 1 1
-=Jl(3.5R -6X3 -2XI +4x l (x 2 -2/))
1
aX3

= Jl(-2x/ + 4x} - 8lx l ) + 'I"(x 3 ),

implies

Finally, we have

Likewise we obtain:
 Deformations 259

Exercise 12.

A solid whose surface is an elliptic cylinder is deformed in accordance with the small
deformation theory.
Given a system of rectangular Cartesian axes OX l X 2 x 3 corresponding to a frame of
reference {o; e l , e 2, e3 }, the equations of the ellipse of each cylinder basis
2 2
~ 2 + ~-l
2 - , X3 = c
a b
are such that
a =0.2, b = 0.1, c=O
and
a = 0.2, b = 0.1, c = 0.2.
The infinitesimal strain tensor is explicitly the following
0
&= [ 0
- 0.8,uX2

with ,u = 10-2 .
(i) Knowing that the infinitesimal rotation tensor and the displacement vector vanish at
the origin, find the displacement field.
(ii) We consider two fibers along the respective straight lines defined by

and
Xl = 0.06, X2 =0.03.

Determine the shear angle for this pair of orthogonal fibers at their intersection point.
(iii) Calculate the unit elongation in the direction of the unit vector of the first fiber and in
1
the direction of the unit vector lw = ..j3(e l +e 2 +e3) at the previous intersection

point.

Answer. (i) First, we note that the compatibility equations are satisfied since the
components of & are functions of first degree in Xi'
Let us express the differentials of the infinitesimal strain tensor components, namely:
d())l2 = (&lk,2 -&2k,l)dx k = (-0.8-0.2),udx) =-,udx),
d())13 = (&lk,) -&3k,I)dx k =-0.2,udx 2 ,
d())23 = (C 2k,3 -C3k,2)dxk =0.8,udxl ,
260 Chapter 3

and thus since liJu(O,O,O) = 0 we deduce that

The displacement field U is obtained from the integration of the following differential
forms

which are
dU, =(£12 + liJ'2) dx 2+ (£13 + liJ13)dx3 =-Jl x3dx2-Jlx 2 dx 3,
dU 2 = (£21 + liJ 21 )dx l + (£23 + liJ 23 )dx3 = Jl X3 dx l + Jl XI dx 3 ,
dU 3 =(£3' + liJ3,)dx, + (£32 +liJ32 )dx2 =-0.6Jl X2dx,-0.6JlX,dx2 ·

Thus we have

since the various components Ui(O,O,O) = O.

(ii) The unit vector along the first straight line (in the plane x3 = 0) is lv = Js e, + ~ e2
and, of course, the unit vector along the other straight line is e 3 •

At intersection point i(0.06, 0.03, 0) the infinitesimal strain tensor is expressed as

o
£ = 10- 2 [ ~ o
-0.024]
0.012
-0.024 0.012 o
and thus the shear angle for the pair of orthogonal fibers has the following value at point i

~ - 7.2
-J5 x 10-4 (d.
ra lans ) .

Thus we have obtained

 Deformations 261

(iii) The unit elongation in the direction of Iv at point i is

and the unit elongation in the direction of 1w at point i is

6(Jwl'd w c lw ~10'[~ J, ~lU4

== -O.8x 10-4 •
CHAPTER 4

KINEMATICS OF CONTINUA

In the previous chapter we took the care to consider initial and final configurations
Co and C1 only. Now we are going to follow the evolution of various physical quantities
which characterize the continuous change in the configuration of a continuum in the
course of time.
First, in the Lagrangian description considered we are going to compare notions at each
time t while referring to an initial configuration Co. Next, this last constraint will be
removed in the Eulerian description.

1. LAGRANGIAN KINEMATICS

The study of continuum kinematics requires the introduction of the material

derivative, that is, a notion which refers to a particle or a system of particles whose motion
will be considered.

Let E be a (real) vector space,

en be a frame of reference of the associated point space E,
1= [O,T] be a time interval,
X be a position vector locating a particle.

Given a particle identified by X let us introduce the notion of material derivative of a

physical quantity Q(X,t), that is, the time rate of change of Q, while X is fixed and only t
vanes.

263

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
264 Chapter 4

In the Lagrangian description the material derivative of a physical quantity is an

intuitively well known notion because it consists in deriving this quantity with respect to t
by 'following' the particle (to which this quantity is attached).
Thus let us introduce the following essential definition.

D The material derivative 1 of a physical quantity Q(X,t) IS the following time

derivative

.
Q- dQI (4-1)
dt X(fixed)

where X is fixed.
It is equivalently expressed as

Q= aQI . (4-1 ')

at x

1.1 HOMOGENEOUS TRANSFORMATION MOTION

By adapting the notion of homogeneous transformation to kinematics we set:

D w A vector function
I ~ E :tH X = F(t)· X + k(t) (4-2)
defines a homogeneous transformation motion on the time interval I, where F(t)
and k(t) respectively correspond to F t and k t in the case of homogeneous
transformations between two given configurations.

D The velocity at time t of the particle identified by X is

V(X,t) = p(t). X + k(t), (4-3)

h
were F'·IS the matena
. Id ' . -aF I .
envatIve
at x

The notion of convective transport of some vector V in Co, which defines the transported
vector v in a given configuration Ct , can be immediately generalized at every time tEl.
Thus we introduce the following definitions.

1 It is also called the 'convective derivative' and in French the 'derivee particulaire '. It is sometimes denoted
. DQ
by Q=_.
Dt
 Kinematics of Continua 265

D The transported vector of V by the homogeneous transfonnation motion is the

vector
v = F(t)·V,
defined for every tEl .

D W' The material derivative of the transported vector v along the motion is
VtEI: v=F(t)·V. (4-4)

Likewise let us introduce the following definitions.

D The transported volume of a parallelepiped volume Po is the volume

defined for every t E ! .

D The material derivative of the transported volume along the motion is

'litE!: P, = j(t)po' (4-5)

The reader will transpose the concept of homogeneous transfonnations of Chapter 3 into
this new context of motion.

For example, let us recall that the Cauchy-Green defonnation tensor C is such that
'litE!: v.w=V.C·W.

By introducing the following definition:

D W' The (Lagrangian) rate ofstretch tensor I is

c= 'F.F+'F·F, (4-6)

we obtain:
(v.w)· = V.('F·F+ 'F·F).W,

=V.C·W. (4-7)

Likewise, let us recall that the Green-Lagrange strain tensor L is such that
'litE!: v.w=V.(I+2L)·W.

1 Called in French: 'Taux de dilatation lagrangien '.

266 Chapter 4

Let us introduce the following definition.

D <iF The Lagrangian rate of strain tensor 1 is

L=~c. (4-8)
2

Thus we deduce:

(v.w) =2v.L·w. (4-9)

1.2 GENERAL MOTION AND GRADIENT

We recall that the notion of continuum motion was introduced in Section 1.3 of
Chapter 2.
Let us consider the motion defined by the function
tP: Ex! ~ E: (X,t) ~ tP(X,t),
which expresses that a particle identified by X occupies successive positions such that
\:ItE!: x=tP(X,t).

The deformation gradient V tPt (X) at X and time t was introduced in Chapter 3. We say:

D <iF The deformation gradient at X is the tensor field such that

\:ItE!: VtP(X,t)=VtPt(X)=~(X). (4-10)

It is also denoted by
VtP(X,t) = F(X,t).

We recall that the linear mapping associated with every deformation gradient, given any
time t and a point determined by X, is said to be the homogeneous transformation tangent
to tP given X and t.

Given the local meaning of the following notions, let us introduce the differential notation.
Any vector dX and its transported vector dx are connected at every time t as follows:
dx = F(X,t) ·dX.

1 Called in French: 'Taux de de/ormation lagrangien '.

 Kinematics of Continua 267

In the Lagrangian description each particle is followed during the motion and the material
derivative is the time partial derivative. Thus we have

dx' = F(X,t)· dX . (4-11)

D QF The Lagrangian velocity gradient is the tensor field V V such that

\;j tEl: F(X,t) = VV(X,t). (4-12)

This notion is well defined because we have

.
F(X,t) =-
a VfIJ(X,t)
at
and by recalling that the velocity field is

-afIJl (X,t) = V(X,t),

at x

we actually obtain (4-12).

Given a basis (e i ) the components of the velocity gradient tensor are

.. a a Xi I a a Xi I
(F)~ = at ax) x = ax) at x
avi
ax)
and we denote

VV(X,t)
av'
= --. e i Qge}.
.
(4-13)
aX}

The material derivative of dx is expressed from the notion of velocity gradient tensor as
follows:

dx' = VV(X,t) ·dX. (4-14)

Moreover, the material derivative of a volume element df-ll is expressed as follows:

(4-15)

Of course, the material derivative of the scalar product of two vectors dv and dw is
written:
(dv.dw)' = 2 dV.L dW , (4-16)
268 Chapter 4

where L(X,/) is the Lagrangian rate of strain tensor expressed as follows:

. 1 .
L(X,t) = -C(X,/)
2

= 2.( IVV(X,/). V tlJ(X,/) + IV tlJ(X,/)· VV(X,t))

2
also written:

L(X,t) = 2. eVV ·F+ IF· VV). (4-17)

2

2. EULERIAN KINEMATICS

We recall that the Lagrangian description of motion amounts to comparing

configurations at two closely neighboring instants by referring to the initial configuration;
this reference is abandoned in the Eulerian description.
In this last description we will consider the material derivative in the current configuration
at each instant tEl and compare quantities in infinitely neighboring configurations
without referring to the initial configuration.

2.1 HOMOGENEOUS TRANSFORMATION MOTION

2.1.1 Velocity Field

From (4-3) we deduce the expression of the velocity field V(x,/) in the Eulerian
description of a homogeneous transformation motion during I, namely:
V tEl: V(x,t) = F(t)· [F-l (t)· x - F- 1(t)· k(t)] + k(t) .

Since F(t)· F- 1(I) refers to each current configuration, but not to an initial configuration,
we denote:
G(t) = F(t). F- 1(t). (4-18)

PRI In the current configuration the velocity field of a homogeneous transformation

motion is expressed as an affine function of coordinates.

Proof From the previous expression of V , that is:

 Kinematics of Continua 269

V(x,t) = G(t)· x - G(t)· k(t) + ic(t),

and by letting K(t) = -G(t)· k(t) + ic(t), we actually have

V(x,t) = G(t)· x + K(t). (4-19)

2.1.2 Material Derivative of a Vector

Let us consider a pair of points p and q of respective position vectors x and y. In a

current configuration it defines a vector
v=pq=y-x.

The material derivative of v is

v = pq. = V(y,t) - V(x,t)

=G(t)·(y-x),
that is,
v = G(t)· v. (4-20)

We note that the Lagrangian description leads to this result as follows:

v = F(t). V = F(t)·(F-l(t) ·v)
= G(t)· v,
but no reference is made to an initial configuration in the Eulerian description!

2.1.3 Material Derivative of a Volume

Let us consider the material derivative of a volume element. By proceeding as

previously we obtain from (4-5) the material derivative of the volume element dJl(:

dJi(· =j(t)r1(t) dJi( ,

where J(t) = det F(t) refers to the initial configuration.

Weare going to make explicit the ratio

(4-21 )
270 Chapter 4

PR2 The previous ratio is the divergence of the velocity field:

d .
~=divV. (4-22)
dILt

Proof Given a basis, the ratio of volumes is expressed as

J(t) = detF(t) = E:ijk FliF2jF3k
and thus

·1
F1 ·1
F2 F3
·1
F'1 F' 2 F13 F1 1 F12 FI3
F21 F22 F23 +
. 2 • 2
F1 F2 F3
•2
+ F21 F22 F23 .
F31 F32 F33 F31 F32 F33 1<3
1
. 3
F 2
. 3
F 3

By recalling (4-12) it is also written:

av 1 av 1 av 1 ax 1 ax 1 ax 1 ax 1 ax 1 axl
ax 1 ax2 ax 3 ax 1 ax2 ax 3 ax l ax 2 ax 3
jet) =
ax 2 ax 2 ax 2 av 2 av 2 av 2 ax 2 ax 2 ax 2
+ +
ax 1 ax 2 ax 3 ax 1 ax 2 ax 3 ax 1 ax2 ax 3
ax 3 ax 3 ax 3 ax 3 ax 3 ax 3 av 3 av 3 av 3
ax 1 ax2 ax 3 ax1 ax2 ax 3 axl ax2 ax 3

but
rl (I) = a(x' ,X2 ,X3)
a(x l,x 2,x3)
and thus we have

that is, explicitly:

 Kinematics of Continua 271

av 1 av 1 av 1 0 0 0 0
axl ax 2 ax 3
aV 2 aV 2 av 2
j(t) rl(t) = 0 0 + + 0 0
axl ax 2 ax 3
av 3 av 3 av 3
o o o o axl ax 2 ax 3

avl av2 av 3
=--+--+--
axl ax 2 ax 3
= divV 0

2.1.4 Eulerian Rates

The material derivative of the scalar product of vectors v and w is written:

(vow)" =v.w+v.w
= G(t) v. w + v. G(t) . w
0

= v.CG(t)+G(t))· w.

Let us introduce the following important notion.

D (8" The Eulerian strain rate tensor, also called the rate of deformation tensor l, is the
symmetric part of the tensor G(t) :

d(t) = ~(G(t) + 'G(t))

2
= ~(F(t). F-l(t) + IF-l(t) oIF(t)). (4-23)
2

Thus the material derivative of the previous scalar product is written:

(v.w)" =2v.d(t)·wo (4-24)

The strain rate tensor is explicitly denoted by

d(t) = dj e; ®e i 0 (4-25)

1 Called in French the 'tenseur de de/ormation euierien '.

272 Chapter 4

PR3 The strain rate tensor is the symmetric bilinear form such that the image of every
pair of vectors v = Vi e i and w = w J e j in the current configuration is the real

(4-26)

Proof From

we deduce:

PR4 Given a homogeneous transformation motion between an initial configuration Co

and a current configuration Ct there is the following relation between the
Lagrangian and Eulerian rates of strain:
(4-27)

Proof We have
d(t) = ~(t . F- I + tF- I . tF)
2
=~CF-I.tF.F ·F- I + IF-I.tF·F .F-I )
2
= tF-I·i·F- I .

In this context of homogeneous transformation motion let us introduce the following rates
defined in the current configuration Ct •

D The rate of extension I in the direction of v is the rate of change of length per unit
length in this direction, that is:

PR5 The rate of extension is expressed as

[lvil'
M = 1v .d(t) ·1v · (4-28)

Proof We have

I Called in French: 'Taux d'allongement unitaire' ou 'Taux d'extension unitaire '.

 Kinematics of Continua 273

211 villi vii' ( 11vII 2 )"

211v11 2 211vi1 2
which is really equal to
1v ·d·1v ·

In particular, the rate of extension in the direction of a basis vector e i is

(4-29)

Let us introduce the shearing as follows.

The angle between orthogonal vectors v and w in a configuration Ct is assumed to become

:r _ e in a neighboring configuration Ct ,.
2

The material derivative

(v. w)' = I villi w I [cos(!: - e)]" = I villi w I iJ cose

2
is written:
2v.d(t)· w = Ilv 1IIIwil iJ,
since e is zero at time t.
Thus we set the following:

D The rate of shear or shearing' for a pair of orthogonal unit vectors 1v and 1w is
the rate of decrease of angle (from :r/2):

(4-30)

also written, given t, as follows:

In particular, the rate of shear for a pair of orthogonal basis vectors e i and e j is

(4-31 )

I Called in French: 'Taux de glissement'.

274 Chapter 4

More particularly, let us consider the three (mutually orthogonal) principal directions of
the (symmetric) Eulerian strain rate tensor d(t) in the current configuration Ct.
In a natural manner we say:

D The (real) eigenvalues of the strain rate tensor d(t) are said to be the principal
rates of extension, also called the stretchings.

They are denoted by

d; (i=1,2,3).

Thus given the (orthonormal) principal basis (if;) we denote

d(t) = d] it] ® it] + d 2it2 ® it2 + d 3it3 ® it3. (4-32)

Remark. Since this form of the strain rate tensor is diagonal, the rate of shear for every
pair of principal directions is then zero.
Such orthogonal directions define the principal directions of d(/).

2.2 GENERAL MOTION AND VELOCITY GRADIENT

Let us follow the process of Section 1.2, but this time the current configuration Ct
will play the role of 'initial configuration of reference'.

2.2.1 Velocity Gradient Tensor and Eulerian Rates

(i) By referring to (4-14) we are going to introduce the velocity field, but this time
in the Eulerian context, that is, from any vector element dx expressed in the current
configuration Ct.

D CJr The Eulerian velocity gradient for x in Ct is the tensor gradV(x,t) such that
dx' = gradV(x,t)· dx. (4-33)

Given a basis (e;) we express this gradient as follows:

(4-34)
 Kinematics of Continua 275

PR6 Relations between VV(X,t) and gradV(X,t) are

gradV(x,t) = VV(X,t) of-l(X,t) , (4-35)

VV(X,t) = gradV(x,t) V t/J(X,t) .
0
(4-36)

Proof The material derivative of the transported vector dx of an arbitrary vector element
dX at time t is written:

dx· =VV·dx=VV·r1·dx,

but it is also
dx· = grad V . dx

and thus (4-35) is proved.

The second relation of the proposition follows from (4-35) since F(X,t) = Vt/J(X,t).

Remark. In explicit form (4-35) and (4-36) are successively written:

av,] [~].[axr],
[ax} ax r ax}
=

[:~~ ]=[:~~l[:;Sj J
Likewise the vector dx· is explicitly written:

PR7 In the context of general motion tangent (at x) to some homogeneous

transformation motion, there is the following relation
F(X,t). F-l(X,t) = gradV(x,t). (4-37)

Proof This relation is immediate since we have:

F(X,t)· F- l (X,t) = VV(X,t)· F- l (X,t)
= gradV(x,t)· F(X,t)· F- l (X,t)
= gradV(x,t).
276 Chapter 4

Thus in this context of a general motion tangent to some homogeneous transformation

motion, we find a relation analogous to (4-18) which was written for homogeneous
transformation motions and which we express as follows:
F(X,t)· F- 1 (X,t) == gradV(x,t)
== G(x,t).

(ii) We may use the notions of material derivatives, introduced in Section 2.l, which will
be rewritten in the context of differentials.

First, we let:

D Cir The rate of change of volume per unit volume is

(dfl)' == divV(x,t), (4-38)

dfl
which is also equal to trgradV(x,t).

Next, we introduce the following definition.

D Cir The (Eulerian) strain rate tensor 1, also called the rate of deformation tensor, is
the symmetric part of the velocity gradient:

d(x,t) == ~(gradV(x,t) + 'gradV(x,t)) . (4-39)

2

PR8 Given any two vectors dv and dw in the current configuration C1 , then the
material derivative of their scalar product is

(dv. dw)" == 2d(x,t)(dv,dw). (4-40a)

Proof We have
(dv.dw)" == gradV ·dv.dw + dv.gradV· dw
== dv.( 'grad V + gradV)· dw
== 2dv.d(x,t) ·dw. (4-40b)

In terms of components the previous proposition is expressed as

(4-41 )

I Called in French: '(Tenseur des) Taw: de de/ormation (eulerien)',

 Kinematics of Continua 277

More particularly, given an orthononnal Cartesian basis the components of the Eulerian
strain rate tensor are

d .. -_ ~ ( av; + av) ), (4-42a)

y 2 ax) ax;
also written:
1
d.y =-(V
2 /,J. +V).
J,/
(4-42b)

(iii) Let us consider the rate of extension and the rate ofshear.

D r:jF The rate of extension in the direction of dv = ds 1. is the rate of change of length
per unit length in this direction I, that is:

Ildvll' ds' (4-43)

Ildvll =-ds .
PR9 The rate of extension is expressed as

(4-44)

Proof See PR5.

In particular, the rate of extension in the direction of a basis vector e; is

d ii = ej.d(x,t)·ej. (4-45)
Some such diagonal element is also known as stretching.

D r:jF The rate of shear for a pair of orthogonal unit vectors 1. and 1", is the rate of
decrease of angle (from 1r12 i:
iJ = 2 Iv .d(x,t) ·1"" (4-46)
also written, given x and t, as follows:
iJ = 2d(1• .1",).
In particular, the rate of shear for a pair of orthogonal basis vectors e; and e) is
iJ = 2 dij(x,t). (4-47)

Some such rate of decrease of angle is also known as a shearing.

1 Called in French: 'Taux d'allongement unitaire dans une direction'.

2 Called in French: 'Taux de glissement de deux directions orthogonales '.
278 Chapter 4

Finally, with respect to the principal basis (EJ which corresponds to eigenvectors of d,
the Eulerian strain rate tensor is expressed as follows:
(4-48)

Remark. Since this form of strain rate tensor is diagonal, the rate of shear for every pair
of principal directions is then zero. Therefore during a motion defined by the velocity field
V(x,t), the three principal directions of d(x,t) continuously define three mutually
orthogonal axes.

Example. The Eulerian description of a continuum motion is given by

The Lagrangian and Eulerian forms of the velocity vector are respectively

VeX , t) = _~e-1/5
5
X e + ~ ellS X e
I I 5 2 2

and

This is a steady motion.

The components of the strain rate tensor are dij

1 av av
= -(--' + __J) and thus we obtain:
2 aXj aXi

d= [-1/50 1/50 0]0 .

o 0 0

Given two small orthogonal fibers along the bisectors of 0 XI X 2 , the rate of shear for a pair
of unit vectors along these fibers is

B 0
2( ~ ~ ol[T 1~5 :lf~104
There is no extension along the fibers, for instance the rate of extension in the direction of
the first fiber is

fi
2
 Kinematics of Continua 279

2.2.2 Lagrangian and Eulerian Strain tensors

PRI0 The (Eulerian) strain rate tensor d(x,f) and the Lagrangian rate of strain tensor
L(X,f) are connected as follows:

(4-49)

Proof From (4-40) and (4-16) which are

(dv.dw)" = 2dv.d(x,t)·dw,
(dv.dw)" = 2dV.L(X,t)·dW,
we deduce, by comparison, that
d(x,l) =«dVr1.r1)®(dV.L(X,I)·dW)«dWr1.F-1)
= tF-' .L(X,t).F-'.

Of course, the strain rate tensor d is not the material derivative L of the Green-Lagrange
strain tensor, but they are connected as follows:
L(X,I) = tF(X,I) ·d(x,f)·F(X,I). (4-50)

PR 11 In the case of infinitesimal motions the strain rate tensor d is the material derivative
of the infinitesimal strain tensor 6.

Proof We know that

L='6,
where we recall that the infinitesimal strain tensor 6 is the symmetric part of the
displacement gradient vu .
Since this gradient is small compared to unity we deduce that
F=I+ VU =. I
and thus we have
d=.L=.C. (4-51)

Remark 1. We emphasize that the previous comparison between d(x,t) and 6(X,t)
poses no problem since Lagrangian and Eulerian variables can be used indiscriminately
for such tensors in the case of infinitesimal motions.

Remark 2. Given a tensor field T, in the case of infinitesimal motions we have:

VT(X,I) =. gradT(x,t) , (4-52)
280 Chapter 4

because
VT(X,t) = gradT(x,t)· Vf/J(X,t)
= grad T(x,t)· (I + VU(X,t)).

2.2.3 Rate of Rotation

Let us be careful with the anti symmetric part of the velocity gradient gradV .

D <:iF The rate of rotation tensor I is the skew-symmetric part of gradV(x,t):

1
[}(x,f) = -(gradV(x,t) - 'gradV(x,t)). (4-53)
2

The following proposition is obvious.

PR12 The velocity gradient tensor is decomposed into a symmetric part which is the
strain rate tensor d and an anti symmetric part which is the rate of rotation tensor
[}:
gradV(x,t) = d(x,t) + [}(x,t). (4-54)

In general coordinates, instead of considering the mixed components of the tensor [},
namely:
[}l =~(aV) _ av') (4-55)
, 2 ax' ax}'
we introduce the covariant components

(4-56)

which are such that

Now let us consider the adjoint of the tensor of covariant components [}l;'

It is a vector denoted by [} whose components are

ilk = ~2! J..l kij ~(V

2 j,1
- V )=
I,j
~2 ~ ~(V
~detg 2 j,l
- V )
I,j'

I Called in French: Tenseur de taux de rotation '.

 Kinematics of Continua 281

also written [see Chapter 1]:

Qk =~(curlV)k .
2
Thus we say:

D The rate of rotation vector 1, also called the vorticity vector 2, is

- 1
Q =-curIV. (4-57)
2

The components of the vorticity vector with respect to an orthonormal Cartesian basis are
written:

that is:

Remark 1. The rate of rotation vector Q can be obtained from the adjoint of the tensor
with components Q ji = -Qij'
Indeed, given a 3-dimensional space we know that the adjoint * (J) of some antisymmetric
(g)-tensor (J) of components (J)ij is the vector of components

( )
k 1 ijk 1 £ijk

*(J) = 2! fl (J)ij = 2! ~detg (J)ij'

Given an orthonormal basis, the previous result applied to the tensor of components
Q ji = -Qij leads to the following components of the rate of rotation vector:

(4-58)

Thus the components of the vorticity vector Q are respectively:

-1 -2 -3
GV' Q = Q32' Q = Ql3 ' Q = Q 21 ' (4-59)
since we have, for instance:
-1 1 231 321
Q = --(£ Q23 +£ Q32) = Q32'
2

I Several authors, in fluid mechanics notably, define the vorticity vector as curlV.
2 Called in French: 'Vecteur tourbillon (du champ des vitesses) '.
282 Chapter 4

We observe that the sense of the vector n depends on the orientation of the space given
the presence of pijk. Of course it is not the case for the tensor n.
More precisely, every vector associated with any antisyrnmetric tensor is called an axial
vector, or a pseudovector.
In the present context we shall not be concerned with this distinction and we will use the
tenn vector regarding {} with full knowledge of the mentioned precision.

Remark 2. Given an arbitrary vector v we emphasize that

{}·V={}l\v. (4-60)

With respect to an orthononnal basis (e;) the previous vector is written:

el e1 e3
= QI {}l Q 3 ={}I\v.
VI v1 v3

Remark 3. Let us note the analogy between the displacement U and the velocity vector V,
and thus between quantities resulting from these vectors, as, for example:
E B d,
ro B {},
VU = E+ro B gradV=d+Q,

(VU) .. = au; B (gradV)ij

av
=~ .
lj ax
J Xj

Remark 4. We recall that

u= (x-X)' =x,
and in the context of infinitesimal motions we have:

(VU)· = & + tV
== (gradU)' = grad V = d + n .
 Kinematics of Continua 283

Classical example. Given a simple shear, that is,

where k E R+ , we recall that

kt 01
1 0
o 1

and thus

and

0 kt 0]
VU [ o.
= 0 0
000

The Lagrangian rate of stretch tensor is

C~[~ +n
The Lagrangian rate of strain tensor is
k/2

o
The velocity field is

VeX,!) = -af/JI (X,t) = (kX2'0,0),

at x

which is also

The Lagrangian velocity gradient is

VV(X,t) ~ F ~ [~ ~ ~]
284 Chapter 4

and the Eulerian velocity gradient is

gradV(x,l)~vv.r' ~f'·r' ~[~ ~ H

since
-kt

o
We actually have
gradV = VV.

The rate of deformation tensor is

~+(F.r' + 'r' .'F) ~[+

k/2
d(t) o
o
The reader will note the simple form of this tensor.

The rate of extension in the direction of any basis vector e j is zero:

d,.II = e . d . eI = 0
I
(no summation!).

The rate of shear for the pair of orthonormal basis vectors e l and e 2 is

B=2d12 =k.

The rate of rotation tensor is

~ ±(g,adV - 'g,adV) ~ [ - ~2
k/2
n o
o
and of course we have
d+Q =gradV.

The rate of rotation vector is

- 1 k
Q = -curlV = --e 3 •
2 2
Its only component is really Q2l'

It is an example of motion such that the trajectories of particles are straight lines while the
instantaneous rotation vector Q does not vanish.
 Kinematics of Continua 285

2.2.4 Decomposition of Motion

Let p and q be two neighboring points in the current configuration Ct •

In general the velocity field is written:

Vq(t) = V/t) + gradV/t). pq + IIpqll o(pq,t) (4-61)

or equivalently:
(pq)' = gradV(p,t)· pq + IIpqll o(pq,t). (4-62)

Let x and y be the respective position vectors ofp and q.

By considering the limit as q tends to p and by denoting pq = dx we observe that the
previous equation becomes Eq. (4-33), namely:

dx' =gradV(x,t)·dx.

The velocity field is also expressed as

V(y,t) = V(x,t) + gradV(x,t)· (y - x) + IIy - xII o(y - x,t)

or equivalently as
V(y,t) = V(x,t) + d(x,t)· (y - x) + .Q(x,t)· (y - x) + IIy - xII o(y - x,t).

It is explicitly written as follows:

Vi (y,t) = Vi(X,t) + d~ (yi - xi) + il~(yk - Xk) + IIy - xII Oi (y - x,t). (4-63)

The previous expressions mean that

PR13 In the Eulerian context the particle motion can be decomposed into
- a translation defined by V(x,t) ,
- a deformation defined by the strain rate tensor d(x,t) ,
- a rotation of instantaneous rotation vector .Q .

Remark 1. Of course the term of rotation, namely:

il~(yk _Xk) = gij iljk(yk _xk),

leads to the following components of the vector .Q(x,t) 1\ (y - x) in the 3-dimensional

Euclidean vector space:
gil (il12 (y2 _ x 2) + il13 (i _x 3» = .0 2(y3 _ x 3) _ .o\y2 _ x 2),
286 Chapter 4

g22(.o21 (y' _XI)+.o23(y3 _X 3)=.o3(y' _X I )_.oI(y3 _X3),

g33(.o31 (y' _ Xl) + (.032 (y2 _ X2) = .0 1 (y2 _ X2) _ .0 2 (y' _ Xl).

Remark 2. In general the deformation defined by d(x,t) contributes to the rate of change
in direction of any material element (for instance dx) in addition to the change of length,
whilst the rotation does not contribute this change of length.
On the other hand, if we consider three material vectors in the principal directions of
d(x,t) , then these directions undergoing a rotation remain mutually orthogonal. Indeed,
d(x,t) leads to rates of extension d;(x,t) which are for any vector dv such that
d(x,t)· dv = d; dv .

2.3 RIGID BODY MOTION

Given any two points p and q in C{ we recall:

D The velocity field of a rigid body motion is the vector field

V(q,t) = V(p,t) + qp A .o(t) ,
also written:
V(y,t) = V(x,t) + .o(t) A (y - x). (4-64)

Of course, in this special case the previous expressions are not approximations. The
deformation term is obviously missing and the vector .o(t) is independent of any
reference point.

PR14 A material system follows a rigid body motion during a time interval! iff
'\ItE!: d(x,t)=O.

Proof The necessary condition is obvious since the velocity field of every rigid body
motion is of type
V(y,t) = V(x,t) + .o(x,t)· (y - x)
and thus the velocity gradient has no symmetric part: d(x,t) = O.

Conversely, if d(x,t) = 0 for every x of a simply connected domain in C{, then the
material system is rigid body moving.
Indeed, we note that the method of 'integration' of the strain rate tensor d(x,t) to obtain
the velocity field V(x,t) is similar to the previous method of 'integration' of the
infinitesimal strain tensor &(X,t) leading to the displacement field U.
 Kinematics of Continua 287

The above mentioned analogy leads to the following relations [see Eq. (3-121)]:

Since every spatial derivative of various Qij is zero we can denote the corresponding
tensor by Q(t) whose components are

) av;
Qij(t = - .
ax j

By integrating we immediately obtain the general vector equation

V(x,t) = Q(t)· x + V,
where V is an arbitrary vector.
By introducing again the adjoint notion Q(t)· x becomes Q(t) /\ X and it is actually a
matter of rigid body motion.

Remark. The above mentioned analogy leads to the following compatibility conditions:
d;k,jm - d jk,;m = d;m,jk - d jm,;k (4-65)

3. MATERIAL DERIVATIVES OF CIRCULATION, FLUX, AND

VOLUME

The aim of this section consists in defining material derivatives which will be
essential for numerous subjects and exercises of the next chapter.
Let us consider a particle or a system of particles.

3.1 ABOUT THE PARTICLE DERIVATIVE

3.1.1 Physical Quantity

In the Lagrangian description every quantity Q is defined by a (scalar, vector, or

tensor) function of initial position vector(s) X and time t.
For a particle the quantity is clearly denoted by
Q(X,t)

and for a set of particles by

Q(X(k),X(I),oo.,t) .
288 Chapter 4

The material derivative is the partial derivative with respect to t, namely:

. aQI
Q-- (4-66)
- at x'

In the Eulerian description every quantity is defined by a (scalar, vector, or tensor)

function of
- time t because the quantity can change (even if each particle occupies a fixed position
in the course of time!),
- the position vector x of the considered particle during the motion.

The quantity Q is denoted by q in the Eulerian description and the connection is expressed
as
Q(X,t) = q(x(X,t),t).

In the Eulerian description the material derivative is not expressed as easily as in the
Lagrangian description.
Indeed, (4-66) is written in the Eulerian context as follows:

q. = -aql x + gradq. -aXI x .

at at

The symbol I x (which can be removed) emphasizes that x is fixed.

D The term aq Ix expressing the rate of change at a given position x is called the
at
local rate of change.

The term grad q . -aXI x expressing the quantity change owed to the particle motion
at
is called the convective rate of change l .

Thus the material derivative dq (x,t) in the Eulerian description is expressed as follows:
dt

q(x,t) = aq (x,t) + gradq(x,t). V(x,t) , (4-67)

at
where the terms of the sum respectively express the nonsteady aspect of q and the
convection.

I Called in French: 'Terme de convection de fa deriwfe particufaire' or 'Taux de variation par convection '.
 Kinematics of Continua 289

In an orthononnal Cartesian basis we have:

q==-+--,
. oq oq
at at ox;
Ox;
(4-68a)

also written:
(4-68b)

Examples.

(i) Given a system of orthogonal Cartesian axes, in a spatial description we consider the
continuum motion of a particle whose position at time t is defined in function of the
material coordinates X; as follows:

where c is a constant with dimension [L2 T- 1].

The time rate of change of the distance between the particle and a fixed point (0,1,0) IS

the material derivative of I == ~ Xj2 + (X2 - 1)2 + x; ,that is:

since we have

(ii) In the orthononnal basis (1r' 10 ,1 z) associated with the cylindrical coordinates, we
know that the gradient of a real-valued function q is

Vq== :~lr+7::10+:!lz
290 Chapter 4

and thus

(4-69)

(iii) In the orthonormal basis (J,,18,1¢) associated with the spherical coordinates, we
know that the gradient of a real-valued function q is

'\Iq=- aq1 +~aq1 +_1_aq1

ar r r ao (} rsinO a¢ ¢
and thus
q =- aq + oq V + oq V(} + oq ~ . (4-70)
at ar r 00 r o¢ rsinO

3.1.2 Vector Field

The preceding section can be particularized in case of vector fields.

The material derivative of a vector field q is immediately

dq (x,t) = oq (x,t) + gradq(x,t) . V(x,t) , (4-71 )

dt at
whose components with respect to a Cartesian basis are

dq; = aq; + aq; V ,

dt at Ox; J
also denoted by
dq; _
--atq
dt 1
+qV
I,J J
. (4-72)

More particularly, the velocity vector may be defined from the displacement U = x - X .

In the Lagrangian description the velocity vector is expressed as

V(X,t) = -OXj x(X,t) = - aUj x(X,t). (4-73)

at at
 Kinematics of Continua 291

In the Eulerian description the velocity vector is expressed as

V(x,t) = U(x,t)

=
au
-(x,t) + gradU(x,t)· V(x,t) ,
(4-74)
at
whose components with respect to a Cartesian basis are

v = aU i + au, v
, at ax) j

= a1u i +Ui,jVj . (4-75)

In the Lagrangian description the acceleration vector is simply expressed as

aVI x(X,t) =a-22

F(X,t)=- xI x(X,t). (4-76)
at at

In the Eulerian description the acceleration vector is expressed as

dV
r(x,t) = -(x,t)
dt
av
= -(x,t) + gradV(x,t)· V(X,t). (4-77)
at

With respect to a Cartesian basis the components of the acceleration are

= al0 + al0 V
ri at ax J
J

= a/vi + Vi.} V}' (4-78)

PR15 The acceleration is expressed in function of the vorticity vector n = ~curlV as

2
av V2
r(x,t) = -(x,t) + (curl V) 1\ V + grad- . (4-79)
at 2
With respect to an orthonormal Cartesian basis the components of r are

av - 1 a
r 1
=_1 +2c
at Ij
k nvk +--(V V).
2 ax
} } }
(4-80)
i

Proof From
gradV = gradV - 'gradV + 'gradV = 2 n + 'gradV
we deduce:
292 Chapter 4

gradV . V = 2fl· V + 'gradV· V

= 2fl /\ V + 'grad V . V
V2
= curlV /\ V + grad- ,
2
such that the components of the last vector term are explicitly:
a 1 aV j
- ( - VjV) = - Vj .
ax, 2 ax,
Thus Eq. (4-79) is proved as well as Eq. (4-80) because
(curl V /\V)i =&ijk(curIV)jVk =2&ijk fl j Vk •

3.1.3 Tensor field

The material derivative of any tensor field T(x,t) is expressed as follows:

.
T(x,t)
aT
= -(x,l) + gradT(x,t)· V(x,t). (4-81)
at
For instance, the material derivative of a second order tensor of components Tij is the
second order tensor of components:

where the last term expresses the contraction between the third order tensor gradT and
V.

We complete this paragraph by considering the density notion.

(i) In the Lagrangian description let us introduce the following notions.

The integral ofa mass density (or mass per unit volume) P(X,t) is

J(Jlo,t) = f1'0
P(X,t) dJ.1.

The integral of a mass per unit surface area .E(X,t) is

J(So,t) = Jsr .E(X,t) dS.

o

The integral of a mass per unit length A(X,t) is

J(so,t) = io A(X,t) ds .
 Kinematics of Continua 293

If Q denotes any previous mass density and Do any previous integration domain, we say:

D In the Lagrangian configuration the material derivative of the integral of a density

Q is
aI=C aQdt. (4-82)
at JDo at

(ii) In the Eulerian description it is a little less simple. In this case, given a configuration
C, , let us introduce the following notions.

The integral of a mass density (or mass per unit volume) p(x,t) is

i(Jlf't) = i
1',
p(x,t) dJl.

The integral of a mass per unit surface area (j(x,t) is

The integral of a mass per unit length A,(x,t) is

i(s"t) = I, A,(x,t) ds.

We will have to consider derivatives of the type di in the next sections.
dt

Remark. We could 'translate' the previous integrals from the Eulerian context to the
Lagrangian one, next we could calculate the time derivative in this Lagrangian context and
finally 'translate' the result so-obtained to the Eulerian description.

3.2 MATERIAL DERIVATIVE OF CIRCULATION

Let c( be a curve in the current configuration C(,

s be the curvilinear abscissa of the particle on c(,
lr(x,t) be the unit vector tangent to c(.

D Cff' The circulation of a vector field v(x,t) along c( is the integral

c = ic, v(x,t).1r(x,t) ds. (4-83)

294 Chapter 4

In the Eulerian context we say:

D The arc element

is the transported element of dX = ds o1'0 ' where So and 1'0 are the respective
curvilinear abscissa and unit vector tangent to the curve in the initial configuration
Co·

First, by considering the Lagrangian variables we know that the vector field v(x(X,t),t)
is V(x,t).
By introducing the convective transport of dso 1'0 ' namely:

ds 1,(x,t) = F(X,t)· ds o1'0 (X),

the circulation is then written:

c= Jfco V(X,t).F(X,t) ·1,0 (X)ds o
and thus in this Lagrangian context the material derivative of the circulation is expressed
as
. i (-(X,t).F(X,t)
C= 8V
at
Co
8F
·1, + V(X,t). -(X,t)
at 0
·1, )dso .
0
(4-84)

Secondly, by knowing that

1'0 ds o = F- 1 ·1, ds
and thus
. ·-1
F .1'0 ds o = F· F ·1, ds
=gradV(x,t) .1, ds,
we can return to the Eulerian description and so Eq. (4-84) is written:
C= Jc,
f (v(x,t).1, (x,t) + v(x,t).gradV(x,t) ·l,(x,t))ds,

which is denoted briefly

c= ic, (v.1, + v.gradV·1,)ds. (4-85)

Since
(v.1 r ds)" = (v.dx)" = v.dx + v.(dx)" = v.dx+ v.gradV·dx,
then Eq. (4-85) leads to the following proposition:
 Kinematics of Continua 295

PR16 The material derivative of a circulation is the integral of the derivative of the
expression under the integral sign; that is:

c. = -dtd i" v(x,t).i,(x,t) ds

(4-86)
= Je,f (v(x,!).1, (x,!) ds)' .

Now let us carryon with C by writing Eq. (4-85) as follows:

C= Je,f (v.i, + 'gradV·v.i,)ds (4-87)
or
C= Je,f (a,v+gradv·V + 'gradV·v).i,ds. (4-88)

Since
t gradV.v = grad(vV) - tgradv· V,
then the material derivative is written:
C= Je,f (at v + gradv· V - tgradv· V + grad(vV)).l, ds.
But the expression
gradv - tgradv

is twice the anti symmetric part of gradv whose adjoint is the vector !..curlv .
2
Thus we have
1
C= , (a,V + curlvi'V + grad(vV)).i,ds.

From this result we can say:

PRI7 The material derivative of the circulation of a vector field v(x,t) along a curve Ct
from a and b is
(4-89)

For a closed curve the material derivative of the circulation is

C= f (a,v + curlv /\ V).i, ds. (4-90)

PRI8 The material derivative of the velocity field along a closed curve is the circulation
field along the curve; that is:

~J V.i, ds = f y,i, ds. (4-91 )

dt 'j
296 Chapter 4

Proof From Eq. (4-85) we deduce in particular:

~ f V .1, ds = f (r· 1, + V. gradV·1,) ds

and since the second term of this last integral is explicitly written:
avax.
V.gradV·1 ds=V-' _ J ds=VdV
r 'a Xj aS "
=d(~2 ),
we have
d V2 a
-AV.1, ds
~1
= 1A r.1, ds + [-] ,
2 a

where the last term is obviously zero.

Remark. The previous expressions concerning vector fields can be generalized to tensor
fields of higher order, for instance t(x,t) = tlj e; ® ej •
By analogy with Eq. (4-88), we would calculate the material derivative of the following
circulation
i
C = c, t·1,ds

and we would find the vector result

c= i , (a,t+gradt·V+ 'gradV·t)·l,ds,

where grad t is a tensor of order 3.

3.3 MATERIAL DERIVATIVE OF FLUX

(i) The derivative of the well known convective transport of the area element vector
da = J(X,t) 'F-1(X,t) ·dA is

(day = j 'rl·dA + J('F-Iy ·dA.

First, we recall that

j=JdivV.

Next, let us calculate CF- 1 ) as follows.

 Kinematics of Continua 297

From F· F- 1 = I we deduce that

F'(F-Ir =-F·r l =-gradV,
and thus

that is,

We conclude that

that is,
(dar = divV da - 'grad V ·da. (4-92)

(iiJ Let S( be a surface in the current configuration C(,

in(x,t) be the unit nonnal vector at each point of S('
v(x,t) be a vector field through S( .

D Cff' Theflux of v(x,t) through St is the integral

<D r v(x,t).da,
= Js{ (4-93)

where da = da in (x,t) is the area element vector.

The well known expression of the convective transport of an area element vector lets us
express the flux in the Lagrangian description as follows:
<D = I
So
V(X,t). J tF- 1 • dA,

where So names the surface in the configuration Co and where v(x(X,t),t) is V(X,t)

In the Lagrangian description, where ~ and ~I are indistinguishable, the material

dt at x
derivative of <D is written:
(4-94)

Since
j = divV J
and since
298 Chapter 4

[because ~('F-l.t F) = 0]
dt

we deduce:

Thus in the Eulerian description the material derivative of the flux <i> = ~ J v(x,t). da is
dt 'St
written:
<i> = J(v. da + v. div V da -
'St
V. tgrad V . da) (4-95)

= f (v.da + v.(da)")
Js,
= Jfs (v.daY .
I

This result is expressed as follows:

PR 19 The material derivative of a flux is the integral of the derivative of the expression
under the integral sign; that is:

<D. = -d
dt s,
Jv(x,t).da = fs, (v(x,t).da).. (4-96)

By writing Eq. (4-95) as follows:

d> = JSf (atv + gradv . V + vdivV - grad V . v).da (4-97)
t

and since we have

curl (v /\ V) = v divV - V divv + gradv· V - grad V . v
we deduce:
d> = f (atv + curl(v /\ V) + V divv).da.
'St
(4-98)

Remark. We can transpose Eq. (4-97) to the case of tensor fields of any order.
For instance, given a tensor field t of order 2 we can consider the vector flux
<D = JSf t·1n da = JSft· da '
I I

whose material derivative d> = ~ r t· da generalizes Eq. (4-97) as follows:

dt J'St
 Kinematics of Continua 299

<i> =
Jrs, (a,t + gradt . V + tdivV - gradV· t)· da, (4-99)

where grad t is a tensor of order 3 and of course grad t . V is a tensor of order 2 as well
as gradV·t.

3.4 MATERIAL DERIVATIVE OF VOLUME INTEGRAL

In the current configuration Ct we consider a material domain D t whose particles

occupy a volume Pt.

Let Do denote the domain considered whose particles occupy a volume Po in the initial
configuration.
Let q(x,t) be a real-valued function representing a given physical quantity.

To obtain the material derivative of the following volume integral

I(p"t) = r
JD,
q(x,t) dpt ,

it is necessary to consider the boundary aDt of D t •

We assume that the conditions of regularity are satisfied and thus we assume in the proofs
that the functions are continuously differentiable (class C 1 ).

3.4.1 Lagrangian and Eulerian Approaches

First, we are going to obtain the material derivative of the volume integral in the
Lagrangian description, and next we will come back to the Eulerian description. Secondly,
there is another method dealing directly with this material derivative in the Eulerian
description.

(i) By recalling that Q(X,t) = q(x(X,t),t)

and by referring to the initial volume Po, the
previous volume integral is immediately written:
I(Po,t) = r
JDo
Q(X,t) J(X,t) dpo .

By knowing that!!'" and.!!.-I are indistinguishable in the Lagrangian description the

dt at x
material derivative of I is written:
.i
1= aQ .
(-J+QJ)dpo. (4-100)
Do at
300 Chapter 4

From

we deduce that the material derivative (4-100), that is, in the Eulerian description:
. d r
1= - Ji q(x,t) dpl ,
dt D,
is written:
i =!D, (dq (x,t) + q(x,t) divV) dpl .
dt
(4-101)

We note that the symbols !!.-.

dt
and r
JD,
are not permutable, apart from when divV = o.
This interesting case will later be studied.

In conclusion, since
(q dpJ = q dp, + q (dp,)"
= qdpi + qdivVdpl'
we have proved the following proposition:

PR20 The material derivative of a volume integral is the integral of the derivative of the
expression under the integral sign, that is:

. dtd!v, q(x,t) dp, = !

1= -
D,
.
(q(x,t) dp,) . (4-102)

The material derivative (4-101) is also written:

r
i = JD, (a l q(x,t) + gradq(x,t). V + q(x,t) divV) dpl' (4-103a)

or equivalently:

r
i= JD, (a,q+div(qV»dp,. (4-103b)

Remark. We can transpose Eq. (4-103) to the case of tensor fields of any order.
For instance, given a vector field q(x,t), by knowing that
div(q ® V) = (gradq)· V + qdivV,
where div(q ® V) is obviously a vector field, we obtain the material derivative

r (a1q+div(q®V»)dpl'
i= JD,
and thus from the divergence theorem:
 Kinematics of Continua 301

(4-104)

Eq. (4-104) is composed of

- a first term which shows the non-steady condition of q or the amount of q created within
a volume Jlt considered as fixed (x fixed),
- a second term which shows the convection flux created with the motion of D t .

(ii) In the Eulerian description we consider the current configuration as the configuration
of reference.
We are going to express the material derivative of the volume integral by comparing two
states of the material system at close instants t and t+ dt .
Let D( be a set of positions of particles at time t,
aDt be the boundary of D( ,
x be the position vector of a particle of D(,
x' = x + V /':;.t be the position of the considered particle at time t+ /':;.t ,
In be the outward unit normal to the boundary aD(.

The set D( becomes the set Dt+!'J.t at t+ /':;.t , which is schematized as follows:

n
o\>\~
/'
- - --- .......
"-

,.
n
1J

Fig. 27
.....
oD, - --
0\>'1

Let fj be the set of points common to D( and Dt+t-. t ,

D' be the set of points of D( which do not belong to D/+f',('
D" be the set of points of D(+f..( which did not belong to D(.

Thus we denote
DtM ( = fj + D" = D( + D" - D' .
302 Chapter 4

Let us split up the surface aDt into a part aDl for which Vo1n < 0 and a part aD2 for
which Vo1n > 0 0

Thus we denote:

To calculate the material derivative of a volume let us consider

jet) = lim _1 (r q(x', t+ M) df.1t - r q(x,t) df.1t) ,

~->O ~ t JDt+,t JDt
that is,

jet) = lim _1 [r (q(x',t+ M) - q(x,t))df.1t + r q(x' ,t+ M)df.1t - r q(x',t+ M)df.1t]

~t JDt JD" JD ,
0

tlHO

By considering the expansion

q(x' ,t+ M) = q(x,t) + M aq (x,t) + oeM)

at
and by taking the limit as M ~ 0 the first integral then becomes:
r
J,
aq
-(x,t) df.1, 0

D, at

With regard to the second and third integrals, first we note that every volume element of
D" is constructed from an element of aD whose area is da and 'length' is V ~t ; that is:
df.1t =V M 0 da = VA M da,

where the various nt are the direction cosines of 1n 0

Consequently we have:

lim _I r (q+ L1t aq )df.1t = lim _1 r (q+ L1t a q ) V L1t da

0

LiHO L1t JD" at LiHO L1t JaD2 at

r
= JaD
2
qV.dao

Secondly, by considering D' we have

df.1t = -V, nj M da,

and thus

lim-l
LiHO
r
L1t JD'
(q+L1t aq )df.1t
at
=- r
JaLlj
qVodao

In conclusion, since aD, = aDl U aD2 , we have obtained:

 Kinematics of Continua 303

j = r aq
at dPI
JD,
+ faD, q V . da . (4-105)

The first tenn of the sum is owed to the time variation of q(x,t) while DI does not
change. The second tenn represents the convection following from the change of DI
during the time.
It is the result obtained from the Lagrangian context, since in the particular case of a real-
valued function q(x,t) , Eq. (4-104) is actually Eq. (4-105).

Remark 1. The Lagrangian approach is mathematically rigorous, whereas the Eulerian

approach is approximate but explicit and vivid. Of course, both approaches lead to the
same result.

Remark 2. In the case of a vector field v( x, t), from the equality

(v ® V) ·In da = (V.ln)v da,
which is explicitly
Vi vj nj da,

we deduce the following vector material derivative

(4-106)

3.4.2 Proper Motion Case

Let us consider a (not necessarily material) domain 151 whose proper motion is in
general different from the material medium motion. For instance, it corresponds to a wave.
Let us introduce a velocity field W defined at t and at every point of aDI •

The time derivative of a real-valued function q(x,t) 'along' the proper motion is written
with the velocity field Was follows:

~q = aq +W.gradq
dt at
and thus

(4-107)
304 Chapter 4

As usual, denoting the velocity field of particles of the material medium by V we express:

D The relative velocity of the domain D, which moves in a proper motion of velocity
Wis
Vr=V-W. (4-108)

By subtracting the following expressions

dq = aq + V.gradq, dq aq
-=-= -+W.gradq,
dt at dt at
we obtain the relationship between the material derivative and the time derivative 'along'
the proper motion of velocity W:

dq dq
- =-=-+ Vr·gradq, (4-109)
dt dt
as well as

~
dt JDt
~ JOtr_ qdfl, + JaD,
r qdfl, = dt r_ qVr.da, (4-110)

where the material domain D, coincides with the domain D, at time t.

Remark. In the case of a material domain D" that is:

we have
dq dq
dt = dt '
and more particularly, if the domain D is at rest (W = 0 ) we have
dq aq
dt at
and
~ r d
dt JD q fl
= r aq d
JD at fl· (4-111 )
 Kinematics of Continua 305

EXERCISES

Exercise 1.

For 'sufficiently small' displacements and displacement gradients, if we consider

the infinitesimal strain tensor as a function of displacements, prove explicitly that the rate
of deformation tensor, also called the strain rate tensor d is approximated by the material
derivative of the infinitesimal strain tensor, namely:
d =&.

Answer. From

we deduce:
d 1 a dU a dU j
d/;ij = 2(ax, d/ + ax, Tt)
= ~(av: + av;)
ax; ax!
2
1 av; aXk av) axk
=-(--+--)
2 ax k ax) aX k ax;
1 av; a av) a
= - ( - - ( X k +Uk)+--(Xk +U k ))
2 ax k ax) ax k ax;
= ~(av; (0 + aUk) + av) (0 + aUk))
2 axk Jk ax) ax k ,k ax;
1 av av)
=-(-' + - ) ,
2 ax) ax;
which is actually dij'

Exercise 2.

The vibration motion of a body about its equilibrium position is given by

XI =x l ,
. (I X3
x 2 = X 2 + a sm27T - - -),
r I
1 X3
X3
'.
= X3 + b COS27T(- - - ) ,
r I
306 Chapter 4

where the dimension of each constants a, b and I is [L] and that of the constant r is [T].
(i) Determine the trajectory, the velocity, and the acceleration of any particle.
(ii) Find the Lagrangian velocity gradient tensor VV(X,t).
(iii) Express the Eulerian velocity gradient tensor d + n as a function of the deformation
gradient Vf/J(X,t) ,
(iv) Make explicit the vorticity vector n.
Answer. (i) The trajectory of a particle is obtained by eliminating t; that is:

(X2 - X 2 ) 2 + (3 - X3) 2 = 1.
a b
These equations define an ellipse in the coordinate plane XI = XI whose center is the
point (XI' X 2' X 3 ) and axes are parallel to the x 2 and x3 coordinate axes.
The velocity of a particle is the material derivative

af/JI ,
V(X,t)=-
at x

whose components with respect to the orthonormal Cartesian basis are

aXIl
VI =& x =0,

aX21 27ra X3
= at x = -r- cos27r( -; -
t
V2 -[-) ,

aX31 27rb. t X3
V3 = fit x= - -r- sm27r(-; - -[-) ,

and the components of the acceleration are

ax
2
r; = a/ Ix = 0,
a 2X21 47r 2 a. t X3
F2 =7 x = - -----;:z sm27r(-; - -[-) ,

a2X31 47r 2 b t X3
r; = 7 x = - -----;:z cos27r(-; - -[-) .

(ii) From the deformation gradient F(X,t) = Vf/J(X,t) , which is represented as follows:
 Kinematics of Continua 307

o
21Ta t X3
F= 0 --I-cOS21T(-;--I-) ,
21Th. t X3
o o 1+ - - sm21T(- - - )
I r I
we deduce the following Lagrangian velocity gradient tensor:

0 0 0
41T2 a. t X3
VV=F= 0 0 - - sm21T(- - - )
Ir r I
41T2h t X3
0 0 - - COS21T( - - - )
Ir r I

(iii) Let us calculate the rate of deformation tensor d.

Since

o o

21Ta t X3
--COS21T(-----)
I r I
21Th. t X3
I + ----- sm21T( - - - )
I r I

o o
21Tb. t Xl
1+ - sm21T( - - - )
I r I
and thus

o o o

2. t X3
41T a sm21T( - - - )
o r I
t X '
I r + 21T h rsin21T( ___3 )
r I
2 t X3
41T b COs21T( -- - - )
o o r I
Ir+21Thrsin21T('--- Xl)
r I
308 Chapter 4

the rate of deformation tensor

is explicitly:

o o o

27r 2 asm27r
• X3
(t- - --)
d = 0 o T I
IT + 27r hi sin27r( ~ _ X3)
T I
2. t X3 2b t X3)
27r asm27r(---) 47r cos27r( - - -
o T I T I
IT + 27r b T sin27r( ~ _ X3 )
T I

The rate of rotation tensor is

o o o

27r 2 asm27r
• Xl)
(t- - -
o o T I
t X '
IT + 27r b T sin27r( ___3 )
T I
2. t X3
27r asm27r( - - - )
o T I o
IT+27rbTsin(~- X3)
T I

which is obviously antisymmetric.

Of course,

is the Eulerian velocity gradient tensor expressed in function of Lagrange coordinates.

 Kinematics of Continua 309

(iv) The components of the rate of rotation vector n are

2. t X3
27r a sm27r( - - - )
T I
t X '
IT + 27r bTsin27r( ___3)
T I

Exercise 3.

Given the following continuum motion

determine with respect to an orthonormal Cartesian basis:

(i) The velocity components in both Lagrangian and Eulerian descriptions,
(ii) The acceleration components.

Answer. (i) In the Lagrangian description we have from (4-73):

In the Eulerian description the velocity components can be obtained in two ways.
First, since

the various components Vi obtained in the Lagrangian description are then written:

Secondly, since
UI = XI - XI = (1- 2e- t )x I - e- t x 2 '
U2 = x 2 - X 2 = _e t XI + (1- e t )x2 ,
U3 = x3 - X3 = 0,

we deduce from (4-75):

VI = e- t (2x I + x 2 ) + (1- 2e- t )VI - e-V2 '

V2 =-et(xI +x2)-e t VI +(1-e t )V2'
V3 = O.
310 Chapter 4

By solving this system of equations we again find

(ii) In the Lagrangian description the components of the acceleration are immediately:

~ =av,!
- (X,t)=e I
X,-e
- I
X 2 (=x 1),
at x

r;=av2 ! (X,t)=-e'X,+2e- I X 2 (=X2)'

at x
I3 =0
and in the Eulerian description Eq. (4-78) implies:
aV1 aV1
Yl = - + - Vj =3V1 +2V2 =x 1 ,
at ax j
Y2=-4VI -3V2 =x2 , y)=O.

Exercise 4.

The Eulerian description of the velocity field of a fluid in steady motion is

where c and k are constants of respective dimensions [eT- 1] and [T- 1 ].

(i) By letting r = ~ x~ + xi , in the Eulerian form prove that the material derivatives of r
and of the third coordinate x) of any particle vanish. Deduce the trajectory of any
particle.
(ii) Let us consider two cylindrical surfaces of the x) -axis of respective radii 0.5 (m) and

1 (m), which respectively rotate about this axis at 180 rev/min clockwise and at 240
rev/min counterclockwise.
First, observe that the norm of any velocity vector only depends on the distance r.
Secondly, determine c and k if the particles in contact with the respective surfaces
revolve with the velocities of points of the concerned surfaces.

Answer. (i) The material derivative of r in the Eulerian description is

r=ar+~v
at ax i I
 Kinematics of Continua 311

and the material derivative of X3 is

Since ~xf + xi and X3 are constant we deduce that every particle follows a circle of
center belonging to the X3 -axis and parallel to the plane of coordinates Xl' X 2 •

(ii) First, since

we deduce that the nonn depends on r as follows:

Secondly, the angular velocity ofthe first surface is

JSO(2n)
OJ] = = -67r (rad/s).
60

The components of the velocity vector of the particle at point (0.5, 0, x 3 ) of this surface
are
V3 = 0 (m/s),

since at this point the velocity vector is parallel to the X2 -axis.

Thus we have a first equation for detennining c and k:
C
-0.5(-+k)=-37r.
0.25

Likewise, for the other surface the components of the velocity vector of the particle at
point (1, 0, X3) are

V3 =0 (m/s).

Thus we have a second equation to detennine c and k:

- ( c + k) = 87r .
The two constants are
312 Chapter 4

Exercise 5.

Given the orthononnal basis (lr,llJ,lJ of cylindrical coordinates we consider the

motion defined by the velocity field
c
V(r,8,z,t) = -11J'
2n r
where c is a constant of dimensions [L2 T- 1 ].
(i) Detennine the trajectory of the particle occupying the position (ro' 8 0 , zo) at t = 0
and characterize the trajectories.
(ii) Express the Eulerian velocity gradient gradV(x,t).

(iii) How are the rate of rotation tensor and the vorticity vector?
Detennine the (Eulerian) strain rate tensor d.
(iv) Calculate the principal rates of extension (or stretchings) and the corresponding
principal directions of d.

(v) Find the rate of change of volume per unit volume (dJl)' .
dJl
(vi) Calculate the circulation of V along any closed curve r around the z-axis and verify
that the material derivative of this circulation vanishes.

Answer. (i) Since the velocity field is

and thus
. c
r =0, 8=--
2nr2 '
z = 0,

we deduce that the trajectory of the particle at the initial position (ro' 8 0 , zo) is defined by
c
r = ro , 8= --2t +80 , z = Zo'
2nro
Each trajectory is a circle in the plane of the equation z = k (k E R) and with center on
the z-axis.

(ii) The Eulerian velocity gradient is immediately

because it is explicitly:
 Kinematics of Continua 313

av, ~(av, -V ) av, c

or r aB e az 0 0
21fr2
aVe ~(ave +V) aVe c
0 0
or r aB ' az 21fr2

avz 1 aVe avz

0 0 0
or r aB az

(iii) Since gradV is a symmetric tensor we know that {} = 0 and the vorticity vector is
zero. Of course, the (Eulerian) strain rate tensor d is grad V .

(iv) From the characteristic equation

2
3 c
-A +A-2- 4 =0,
41f r

-c c
we deduce that the principal rates of extension d i are equal to --2 ' --2 and O.
21fr 21fr

The stretching - c/21fr 2 corresponds to the principal direction along the bisector of
(1,,1 e ), the stretching c/21fr 2 corresponds to the principal direction of the other bisector
and there is no stretching along the principal z-direction.

(v) The rate of change of volume per unit volume is indiscriminately equal to

(dp) = divV = av, + ~ aVe + V, I avz

dp or r aB r az
= tr(gradV) = tr d = 0 .

There is no change of volume and this type of motion will later be labelled as an isochoric
motion.

(vi) Since the circulation is

we deduce that:
c=O,
which is obvious because a,v = 0 and curl V = 2il = O.
CHAPTER 5

FUNDAMENTAL LAWS;

THE PRINCIPLE OF VIRTUAL WORK

In this chapter we state the fundamental laws of continuum mechanics that are the
conservation of mass, the principle of linear momentum, the principle of the moment of
momentum, and the principle of conservation of energy.
Of course, the stress tensor notion will be considered.

1. CONSERVATION OF MASS AND CONTINUITY EQUATION

1.1 AXIOM OF MASS CONSERVATION

Let us consider a 3-dimensional material system in the configuration C/ which

occupies a domain D/ whose spatial volume is denoted as follows:

f-l(D/) = fD I
df-lt,

where the notation of the integration variable is mnemonic.

It is important to specify the following
D A material domainl is the set made up of the same particles during the motion of
the material.

1 Also called a 'material volume '.

315

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
316 Chapter 5

We recall that any material system (such as particles, rigid bodies, etc.) is characterized by
a real number called the mass of the system.
By assuming that the measure of the mass is continuous with respect to the volume the
mass element 11 m of any volume element 11Pt introduces the notion of (mass) density,
that is, the following mass per unit volume
dm . 11m
p(x,t) = - = hm - - , (5-1 )
dll, ,111,->0 11 p,

where x locates any particle of 11 Pt.

The mass of a system which occupies a material domain D t at time t is the following
positive real number
m(D,) = r
JD/
p(x,t) dll, . (5-2)

The basic axiom of classical mechanics requires that the mass of any material system
considered does not change during time; in other words, the motion of particles of the
system does not modify the mass of the system; and we state:

PRI The mass of any definite material system remains constant during the motion of the
system.

1.2 CONTINUITY EQUATION

1.2.1 Continuity Equation in the Lagrangian Description

First, in the Lagrangian description we denote the density as follows:

P(X,t) = p(x(X,t),t).

The axiom of conservation of mass, namely:

is expressed as
r
JDo
Pa(X,O) dpo = r
JIJ 1
p(x,t) dp"

where Po(X,O) is the initial density. It is independent of time.

Since this last integral is equal to
r
JDo
P(X,t) l(X,t) dpo '

where the notation of the integration variable is mnemonic, we deduce that

 Fundamental Laws; Principle of Virtual Work 317

r
JDo
(Po(X,O) - P(X,t)J(X,t)) dJio = O.

Since it is the integral of a continuous function and since this equation is verified for any
Do, the continuity equation in the Lagrangian description is therefore

P(X,t) J(X,t) = Po (X,O) , (5-3)

Since it is independent of time we deduce the following proposition:

PR2 The Lagrangian differential form of the continuity equation is

~(P(X,t) J(X,t)) = 0, (5-4)

dt
where X is fixed.

Remark. The term J(X,t) clearly shows the 'dilatation' of the volume. Indeed, from

dm = Po (X,O) dJio = P(X,t) dJit,

we deduce from (5-3):

J(X,t) = Po(X,O) = d Jit .

P(X,t) d Jio

1.2.2 Continuity Equations in the Eulerian Description

In the Eulerian description the axiom of conservation of mass is written mathematically:

d d
-m(D,)=-
dt dt v,
1
p(x,t)dJi,=O. (5-5)

This material derivative leads to the various global expressions of mass conservation,
which are:

r
JD, dt
(dp + P divV) dJit =0 (5-6)

(5-7)

I
D,
(a p + div(pV)) dJit = O.
at (5-8)
318 Chapter 5

From the divergence theorem (or Gauss' theorem) this equation may be written:

f ap df.1, + f p V.da = o. (5-9)

JIJ, at JoD,

This allows the further interpretation of the mass conservation, that is:

PR3 The rate of change of mass inside a (moving) material domain is the negative of
the rate of outward flow of mass across the boundary.

Since any previous integrand is assumed to be continuous and since the previous equations
are verified for any boundary aD" we deduce the following proposition showing various
expressions of the continuity equation and whose equivalence is obvious.

PR4 In the Eulerian description, the continuity equation is expressed as

dp + pdivV = 0 (5-10)
dt

ap + gradp.V + pdivV = 0 (5-11 )

at
ap + div(pV) = o. (5-12)
at
For instance, Eq. (5-12) is explicitly written:
ap
-+a(pV) =0. (5-13)
at ' ,
We note that Eq. (5-11) is explicitly written:
- In Cartesian coordinates (XI' x 2, x3) :

- in cylindrical coordinates (r, e, z) :

alP + Vr ap + Vo ap + Vz ap + p(aVr +! avo + Vr + aVZ) = 0,

or r ae az ar r ae r az

- in spherical coordinates (r,e,rjJ):

a p+ V -ap Ve ap V¢ ap (avr 1 aVe Vr J av¢ Ve cot()

+--+---+p -+---+2-+----+ )=0.
, r ar r ae rsine arjJ ar r ae r rsine arjJ r
 Fundamental Laws; Principle of Virtual Work 319

Remark 1. Eq. (5-10) leads to

divV = _ p = + (df.1t>" ,
P df.1t

since p is the ratio dm of elements of mass and volume.

df.1t
This result confirms that divV is the rate of change of volume per unit volume.
[See (4-38)].

Remark 2. The continuity equation contains four unknown functions: the (mass) density
p and the component of the velocity V.

1.2.3 Mass Flow Rate

Let us introduce an important notion: the rate of flow of mass, simply called the
mass flow rate.
Let us consider a general domain D, which is not necessarily material and which has a
proper motion of velocity W.

D The mass flow rate l through the boundary aD, of a domain having a proper
motion of velocity W is the integral
f _ pVr.da,
.lJD,
where Vr = V - W is the relative velocity of Dt .

PR5 The only exchanges of mass through the boundary aD, are opposed to the mass
flow rate.

Proof By considering Eq. (4-110), the conservation of mass is expressed as

~ 1-
dt D,
p df.1,+ f_ pVr·da=O,
.lJD,
(5-14)

which proves the proposition.

I Called in French: 'Debit-masse',

320 Chapter 5

By recalling Eq. (4-107), the previous equation is also written:

f_ apdf-l,+ f_ pV.da=O.
JDt at JaDt

Remark 1. The outward nonnal unit vector In being chosen (according to convention in
mechanics), the flow is outward if the mass flow rate is positive. It is inward if the mass
flow rate is negative.

Remark 2. The particular case Vr = 0, that is, W = V , is that of the material domain
(whose particles are unchanged).
In this case, the mass flow rate vanishes:

1.3 THE MATERIAL DERIVATIVE OF THE INTEGRAL OF MASS DENSITY

Before introducing the material derivative of integral of mass density, let us

demonstrate useful fonnulae which concern the material derivative of a function of type
q(x,t) and which take the continuity equation into account.

From
p dq = p(aq + gradq.V)
dt at
aq ap .
= p(-+ gradq.V)+q(-+dlV(pV))
at at
= ~(pq) + qdiv(pV) + gradq.pV,
at
we deduce:
p dq = ~(pq) + div(pqV). (5-15)
dt at

This is also written:

 Fundamental Laws; Principle of Virtual Work 321

p dq = .i(pq) + grad(pq). V + pq divV

dt at
= !!....(pq) + pq divV. (5-16)
dt

The previous fonnulae are widespread amongst vector functions. So we have, for instance:

p dq =.i (pq) + div(pq ® V) . (5-17)

dt at

Now let us consider the integral ofa function q(x,t). From

dm = p(x,t) dp,
we obtain the following integral (in the Stieltjes' sense):
r
JMD,
qdm= r
JD,
pqdp, (5-18)

whose material derivative is

!!.... r q dm = !!.... r pq dp
dt JM D, dt JD,

= r
JD, at
(.i(pq)+div(pqV»dp [see (4-103b)]

_r dq d [because (5-15)].
- JD, P dt P

In conclusion we have proved the following equality:

-d
dt
I M D, q
dm= I'M D ,
dq
-dm
dt '
(5-19a)

that is

!!....r
dt JD,
d=r dqd
pq P JD, p dt p. (5-19b)

This result can be widespread in the case of tensors and we say:

PR6 The mass of any domain of a continuum remaining constant when following the
particles, then the material derivative of the (Stieltjes' sense) integral of a tensor q
is equal to the integral of the time derivative of the tensor:

-d
dt
f M D, q
dm= fM D,
dq
-dm.
dt
(5-20)
322 Chapter 5

Proof We have:
~ r
dt JM
q lj.. dm r
= JM q lj... dm = JDr q lj ... pdfl
DI D, t

or more explicitly:

; LDI qijdm= JD, C!(pqijJ+8 k(pqij,Yk))dfl

= JD, Ca; qij. + P :t qij , + 8k(pVk)qij + pVk8kqijJ d fl

= JD, P(~ij , + Vk8kqij )dfl

= L D,
qijdm.

1.4 ISOCHORIC MOTION; STEADY AND IRROTATIONAL FLOWS

1.4.1 Isochoric Motion

D A motion is called isochoric if the volume of every material domain remains

constant during the motion.
The continuum characterized by an isochoric motion is said to be incompressible.

This means that the measure of the volume remains constant while the form of the
material domain changes generally.

Remark. The incompressibility is not an intrinsic property of the medium; that is, a given
continuum may be incompressible or compressible according to the circumstances. In
addition we specify that the incompressibility is an idealization; for instance, no fluid is
absolutely incompressible, liquids may be considered as practically incompressible on
many occasions and sometimes gases too.

PR7 An isochoric motion is such that at any point and all time:
divV(x,t) = O. (5-21)

Proof The volume of any material domain D, is expressed like the following

By letting q = 1 in Eq. (4-101) we obtain:

 Fundamental Laws; Principle of Virtual Work 323

This last integral necessarily vanishes for every domain following an isochoric motion.
Thus we deduce that divV vanishes at any point and every time.

Consequently, in the case of an isochoric motion the continuity equation becomes:

dp
-(x,t) =0, (5-22)
dt
which expresses that the (mass) density p of an element of volume remains constant
when following the motion, it being understood that p may be different for any other
element of volume.

We are going to clarify PR7.

PR8 In the course of an isochoric motion the Jacobian remains equal to 1.

Proof From
j(t)dfio = (dfi,)'
= divV dfi,
= divV J(t) dfio'

we deduce:
jet) = divV J(t).
Thus in case of isochoric motion we have

Since initially

we conclude that the constant Jacobian equals 1.

By taking this proposition into account in the Lagrangian description the isochoric motion
is then clearly described by the following equality:
P(X,t) = Po (X,O) , (5-23)

since J = 1.
We conclude again that the (mass) density remains constant by following any particle.

More particularly we say:

324 Chapter 5

D If the (mass) density is constant at any time and at every point of a continuum, then
it is said to be a homogeneous incompressible continuum.

Example 1. A fluid moves around a sphere of radius R with the velocity field V whose
components are in spherical coordinates:
R3 3R R3 3R
Vr =C(---+l)cosB, Ve =C(-3 +--l)sinB,
2r 3 2r 4r 4r
where C is a positive constant of dimension [LI-I] .

The motion is isochoric because we have

divV= 8Vr +2. 8Ve +_1_8V¢ + cotBV +~V

8r r 8B rsinB 8¢ r B r r

= C( -~ R 3r- 4 + ~ Rr- 2 )cosB + C(~r-4 + ~ Rr- 2 - r- 1 )cosB

2 2 4 4
J 3
+ C( -R r -4 + -3 Rr -2 - r -1)cos B + 2c(R
- r -4 - -3 Rr -2 + r -1) cos B
4 4 2 2
= o.
Since divV vanishes at any point and every time the continuity equation dp = 0 means
dt
that the (mass) density remains constant along each trajectory.

If we assume that limp is a constant Po, then the (mass) density is constant at any time
r->",

and everywhere, that is:

p(x,t) = Po'

Example 2. The plane isochoric motion.

In the case of a plane motion let us recall that the velocity field has no component along
the direction orthogonal to the plane at every time and does not depend on the variable
associated with this direction.
In concrete form, we consider for instance a motion parallel to the plane {o;e l ,e 2 }; that
IS:
 Fundamental Laws; Principle of Virtual Work 325

The existence condition of a corresponding isochoric motion is

divV = aVI + aV2 = 0 .

aX I ax 2
We know that a differential form is exact iff its integral along any closed curve r of a
simply connected domain D is zero.
In this context Stokes' theorem leads to

Sf av; aV
ax + -ax) dx
JI ( -
D l
2
2
l /\ dx2 = 1.Jr v, dx
\I
2-
V
2 dx l =0

and thus there is a function lfI(x, ,x2,t) of dimension [erl], called the stream/unction,
such that:

that is:

In other words, the circulation of gradlfl along r is zero:

We observe that the problem with two unknowns VI and V2 is reduced to a problem with
one unknown IfI .
We also note that
v = curl(lfIe3) = IfI curl e3 + gradlfl /\ e3
=gradlfl /\ e3 .

The consideration of IfI leads immediately to streamlines.

Indeed, first we mention that V is orthogonal to grad IfI .
Secondly, if we consider at a given time T a curve c of equation IfI(X" x 2) = k , then if the
curvilinear abscissa is denoted by s we have

dlfl = alfl dx, + alfl dx 2 = O.

ds ax l ds ax 2 ds

This means that gradlfl is orthogonal to the unit tangent vector 1( to c, since the previous
result is nothing else than grad lfI.l( = O.

Thus V and 1( are parallel, and we conclude that, at time T, lfI(x I , X2) = k is the equation
of a streamline.
326 Chapter 5

1.4.2 Steady Flow

First, we let

D The flow of a continuum is steadyl if all quantities associated with the continuum
(velocity, acceleration, density, temperature, ... ) are constant in time at any fixed
point in space.

This is equivalent to stating that every partial derivative with respect to time of these
quantities vanishes.

Of course, the time rate of change ~ of any quantity relative to a moving point will
dt
generally change as illustrated as follows:

Example. If the components of a velocity field of a steady flow are

we actually have OV = 0 but of course there is an acceleration field r whose

at '
components are
r, = oVI + V oVI =0
1 at 'ox;

Now let us consider Eq. (5-9) of the mass conservation which becomes in the case of a
steady flow:
r pV.inda=O.
.laD,
(5-24)

At this stage it is useful to introduce the following notion.

D A tube of flow is a volume which is laterally delimited by streamlines passing

through a closed curve and bounded by two cross sections.

Thus at time t let us consider a closed curve c and two cross sections SI and S2
(perpendicular to the streamlines).
On the lateral walls there is no matter that leaves or enters the tube since V.in = 0, and
thus Eq. (5-24) is reduced to

(5-25)

where VI and V2 are the velocity vectors on the respective cross-sections S1 and S2.

I A steady flow is called in French: "Ecoulement permanent. "

 Fundamental Laws; Principle of Virtual Work 327

Fig. 28

Let us consider an infinitely thin tube offlow.

Since for an entry section dS I we have V1.ln1 = -"VIII and for an escape section element
dS 2 we have V 2 .1 nz = 1IV211, then Eq. (5-25) leads to the equality

(5-26)

where da 1 and da 2 denote the respective area elements.

In conclusion we state:

PR9 The quantity p11V11 da is constant along any current line of steady flows.

1.4.3 Steady Isochoric Flow

PRIO The vectors Vand gradp are perpendicular for every steady isochoric flow.

Proof In this particular case, the continuity equation (5-11) is simply written:
gradp.V = o.

PRII Velocity is inversely proportional to cross section area for every steady isochoric
flow.

Proof Eq. (5-26) which describes every steady flow becomes for an isochoric flow:
(5-27)

because the mass density p of the element of volume remains constant, namely: dp = 0,
dt
during the motion.
328 ChapterS

1.4.4 Irrotational Flow

D A flow for which the rate of rotation tensor n vanishes everywhere is called an
irrotational flow.
In particular, a flow such that the components of the vorticity vector vanish

is said to be an irrotationalflow.

Let us prove the following necessary and sufficient condition.

PR12 A flow is irrotational iff the velocity field is conservative.

In other words iff there is a potential rp such that:

n =!....curlV
2
=0 v = gradrp. (5-28)

Proof First, within a simply connected domain we consider two curves C1 and C2
connecting a fixed point p with an arbitrary point x.
Stokes' theorem states that the surface integral of the normal components of curlV taken
over any surface S which has the closed curve C = c1 U c 2 as a boundary is equal to the line
integral of the tangential component of V around c:

LcurlV.ln da ={ V.I, ds.

Therefore the hypothesis curl V = 0 implies the following equality

r
.Ie, (p-H)
V.I, ds = r
.le2 (p-+x)
V.I, ds

and thus this common integral is only a function of x and is denoted by

rp(x) = f: V.ds.

Now we are going to prove that V is the gradient of rp.

We consider any pair of neighboring points (x, x') such that

xx' = ds = ds lxx"
where lxx, denotes the unit vector in the direction of xx' and the previously introduced
notation ds is mnemonic. This last remark will be used below.

We have explicitly:
 Fundamental Laws; Principle of Virtual Work 329

d ¢ = o¢ dx i ds.
ox l ds
But
¢(x') - ¢(x) = rv. ds

implies

where the various l xx" Ii are the direction cosines of the arbitrary vector xx', which are
dx'/ds'
From the comparison between the expressions of d¢ we deduce that

(V - o¢) dx ' = 0
I ox
l ds
and the arbitrariness of the direction cosines allows the conclusion V = grad ¢ .

Conversely, let us prove that V = grad¢ implies that n = O.

From

we deduce that

or, in orthogonal Cartesian coordinates:

(curlVh = Okij o;Vj = 0kij oM j ¢) = 0,
the cancellation of previous components resulting from both equalities 0kij = -Okji and
o,D j = Djo i •
The proposition is completely proved.

1.4.5 Isochoric Irrotational Flow

PR 13 The velocity potential is a harmonic function for every isochoric irrotational flow:
iJ¢ = O. (5-29)

Proof From the previous section we know that

V= grad¢
330 Chapter 5

and since the flow is isochoric the continuity equation

dp +pdivV=O
dt
is reduced to
p div(grad¢) = 0,
that is
LI¢ = O.

Remark. We note this result can be obtained for any isochoric motion since, in general,
every vector field may be expressed as
V = curl 'I' + grad¢
which implies that
divV = div(curl'l') + div(grad¢) = LI¢.
Thus given an isochoric motion, this last result leads to a harmonic function ¢ too, since
we have
o= div V = LI¢ .
In this case there is no requirement on 'I' , but if in addition the motion is irrotational then
curl'l' = o.

2. FUNDAMENTAL LAWS OF DYNAMICS

Until now only the kinematical aspects have been taken into consideration. But we
know that the forces bring about motion and deformation of a continuum.

2.1 BODY FORCES AND SURFACE FORCES

We are going to analyze the forces which act on the elements of a continuum.

Given a material system S let us consider any part D, which we assume to be enclosed
within a surface S at time t.

Two types of forces are considered in continuum mechanics.

First, the body forces act throughout S by a long range interaction. More precisely and
particularly, they act on each element inside D, and these forces are also said to act 'at a
distance' because they act with matter or charge at a distance without contact on S.
 Fundamental Laws; Principle of Virtual Work 331

The most common examples are the gravity and inertia forces.
Let us define in D, a body force per unit volume as follows.

Let dp be the volume of any element of D, about a point P E D"

dm = pdp be the corresponding element of mass,
f dp be the resultant of body forces which act on the matter corresponding to dp.
We observe that f is the resultant of body forces per unit volume.
For example, the force of gravity per unit volume is
dm
f = dp g=pg,
where g is the acceleration owed to gravity.

The body force is f dm, while f is the body force per unit mass.
p p
In general we will assume that the force per unit volume f and the mass density p are
continuous functions of space coordinates and time.
We note that long range couples also appear, for instance with magnetic and electric
fields.

Secondly, the surface forces are the forces acting on any surface element and are owed to
material adjacent to this surface element.

Fig. 29

We admit that the following has a local significance.

Let us consider a point q of any element LI S of the surface S. By introducing a unit
normal vector ;, at q of LI S we split the space in the neighborhood of LI S into two
regions, namely: S+ which is the part corresponding to ;, and the part S- corresponding
to -;,.
332 Chapter 5

Let us denote by LlF the resultant of surface forces acting on the surface element L1 S .
Let us take care! The notation L1 is introduced to emphasize that LlF is dependent upon
the choice of L1S (and upon in)'
Let L1 a denote the area of L1 S .
Before introducing the Cauchy stress principle and in accordance with this one, let us
recall the following notion.

D ClF The stress vector l (simply called stress) at a point q of L1 S is the surface force per
unit area
0) M'
t(n = lim-. (5-30)
Lla->O ~a

The Cauchy stress principle is stated as follows:

PR14 At a given point (and given time) the stress vector is the same for material
elements having a common tangent plane at the point and for the same side of the
surface.

Thus the stress vector is only dependent upon the point q and the normal vector at a given
time. Of course, for different surface elements associated with different normal vectors
there are different corresponding stress vectors at q. This is the reason why we denote the
index (n), where n is the corresponding unit normal vector.

Remark. We recall that the dimension of stresses is [ML- l T- 2 ] and the units are the
pascal Pa (Nm-2 ), the kilopascal kPa, the megapascal MPa (10 6 Pa), or the gigapascal
OPa (10 9 Pa).
Another unit for stress which has been accepted is the bar, which is equal to 10 5 Pa , i.e.
10 2 kPa.

The following proposition is obvious.

PR 15 The resultant of surface forces exerted across the surface element of area da by the
material of S+ upon the material of S- is t(iI) da.
The resultant of surface forces exerted across the surface element of area da by the
material of S- upon the material of S+ is t(-il) da .

Remark 1. From the law of action and reaction we have:

t(-il) = _t(iI) .

I In French: 'Vecteur de contrainte' or 'Vecteur desJorces de contact',

 Fundamental Laws; Principle of Virtual Work 333

Remark 2. The actions exerted on L1 S can show couple-stresses as in particular in the

elasticity theory, but they are not considered in this context.

We conclude this section with important definitions.

D (iF A traction is a stress (vector) such that the angle between t and n is acute (positive
normal stress).
A compression is a stress (vector) such that the angle between t and n is obtuse
(negative normal stress).
A shear stress (vector)' is a force exerted parallel to the surface.
It is denoted by •.
A normal stress (vector) is a force exerted perpendicular to the surface.
It is denoted by CT .

The following proposition is obvious.

PR16 The stress is the sum of the shear and normal stresses:
t=.+CT. (5-31 )

2.2 PRINCIPLES OF LINEAR MOMENTUM AND MOMENT OF MOMENTUM

Let x be the position vector of an arbitrary point of a material domain D t ,

V denote the velocity V(x,t),
p denote the density p(x,t),
f be the resultant of all body forces per unit volume.
Let us introduce the notions of 'kinetic dynam' and 'dynamic dynam' which are
developed in theoretical mechanics [see for instance Talpaert (2002)].

D The kinetic dynam at time t is the dynam defined by its following reduction
elements:
- the kinetic resultant or (total) linear momentum:

~ =! M lJ,
V dm =i I
pV dfJ, (5-32)

- the kinetic moment or (total) moment of momentum about a reference point 0:

6V" mo = JMDt
r x A V dm r
= 1t x A pV dfJ . (5-33)

I Called in French: 'Cisaillement',

334 Chapter 5

The kinetic dynam about 0 so defined is denoted by

The acceleration r(x,t) = dV (x,t) , simply denoted r, allows us to introduce the

dt
following definition.

D The dynamic dynam at time t is the dynam defined by its following reduction
elements:
- the dynamic resultant:

G\/' R r r dm = hir
= JMD
I
P r dl1, (5-34)

- the dynamic moment about a reference point 0:

G\/' Mo r x/\rdm = .br

= JMD
1 t
X/\P r dl1. (5-35)

The dynamic dynam about 0 so defined is denoted by

We can state two fundamental principles of continuum dynamics.

PRl7 The time derivative of the total linear momentum of any material domain of a
continuum is equal to the resultant of all external forces acting on the domain
considered.

PRl8 The time derivative of the total moment of momentum of any material domain of a
continuum (about a fixed point 0) is equal to the sum of all moments of external
forces (about 0) acting on the domain considered.

Proof We are going to prove these propositions simultaneously. For that let us introduce
two lemmas.
The first lemma is

Indeed by recalling (5-20) we have

d
-9{,
dt
=d-
dt
i
M D,
V dm =!r p-dl1
),
dV
dt
= 1, p r dl1 = R
 Fundamental Laws; Principle of Virtual Work 335

and

.
SIllce -d x = V .
dt

Given an inertial frame l and an absolute time, the second lemma, which follows from the
law of action and reaction, is written:

[MR] = [R(e)]
M(e)'
o 0

where the previous dynam of external forces is such that

R(e) = r f df.1 + r ten) da,
JOt JaD 1

M~e) = r
.hI
x /\ f df.1 + r
JaD,
x /\ in) da .

In conclusion, both the previous propositions are proved because we have demonstrated
that

d
dt
[90]
em
[R(e) ]
0 = M(e) 0
(5-36)

PR19 The mathematical expressions of the two fundamental principles of continuum

dynamics so obtained, namely:

!!....
dt .6,
r pV df.1 =r
.6,
f df.1 + r
JaD,
in) da, (5-37)

!!....
dt .6,
r x /\ pV df.1 r
= .b, x /\ f df.1 + r
JaD,
x /\ten) da (5-38)

lead to the Euler formulation

r ~(pV)df.1 + r p(V.lnW da = rf df.1 + r ten) da, (5-39)

~& ~ ~ ~

r x/\~(pV)df.1+
.ID, at r x/\p(V.l.Nda = .6,r x/\fdf.1+ JaD,
JaD,
r x/\t(fr)da. (5-40)

Proof Eq. (5-39) follows from Eq. (4-106) which is written:

1 Transport and Coriolis forces must be considered in the case of non inertial frames.
336 Chapter 5

~ r pV dp = r ~(pV)dp + r (V.In)PV da,

dt ob, ob, at JoD,
where we recall that
(pV ® V) ·In da = (V.In)PV da.

Likewise we have

~ r x /\ pVdp = .10,r x /\ ~(pV)dp

dt ob, at
+ r x /\ (V.In)PV da,
JoJ),
which proves Eq. (5-40).

Remark. From the well known equalities

~~=~
dt
1 pVdp= i , p r
dt,
dp,

~mo =~ r x/\pVdp= r x/\P r dp,

dt dt .b, .b,
we deduce that the two fundamental principles of continuum mechanics are also expressed
as follows:
r
.b,
p r dp = r f dp
.b,
+!oD, t(it) da, (5-41 )

r
Jot
x /\ p r dp = r x 1\ f dp + r x /\ t(it) da.
Jot .taD
(5-42)
l

Let us conclude this section by considering a fixed 'control domain', that is, a domain,
fixed in space, which instantaneously coincides with the moving material domain D t •

PR20 Given a fixed 'control domain' Dc and any tensor field q(x,t) we have

~ i qdpt =
dt ,
i c
(q+qdivV)dp. (5-43)

Proof We successively have

~i
d
r,--Dc ~(d)
q d p, -- .10 dt q P
t ,
r qdp + JDc
= .be r q(dpy
= 1c (q + q div V) d p .
 Fundamental Laws; Principle of Virtual Work 337

In particular, if q is pV then this proposition takes the following fonn:

!!.-
dt 1,
f pV d}1, = i «pV)" + pV divV)dJ.1.
c

Thus for a fixed 'control domain' the Euler equation (5-39) is written:

2.3 CAUCHY'S STRESS TENSOR

At this stage it is easy to justify the following property of the stress.

PR2l For every surface element of unit nonnal vector N we have

t(-IV) = _t(IV) .

Proof Inside a domain Dt we consider a point p which is the origin of a unit vector N.

Fig. 30

The tangent plane at p to a surface orthogonal to N detennines a partition of D t , that is,

and a partition of the boundary, that is,

aD, = aD; U8D;.
Let us denote by St the part of the plane which is included in D, .
The linear momentum principle is written:
forD t : f (py-/)d}1= f t(iJ)da, (a)
JD, Jan,
338 Chapter 5

for D;: f (pr-f)dp= f t(n)da+ f t(-N)da, (b)

~ ~ ~
(c)

By adding (b) and (c), and by comparing the result with (a), we deduce that
f_ (t(-N) + t(N» da = o.
J.~t

Since this result is valid for any SI the property is established.

Now let us introduce the stress tensor 'in accordance' with the Cauchy vision.
At instant t, at point P of a continuum we consider a small tetrahedron P PI P2 P3 as
shown in the following figure where the system of Cartesian coordinates Xl' Xl' X3 is the
frame of reference of orthonormal basis (eI,e 2 ,e3 ).

Fig. 31

Let D be the domain delimited by the tetrahedron,

S be the inclined face of the tetrahedron,
Si be the face perpendicular to each Xi -axis.

If In (or ;') denotes the outward unit vector through P and perpendicular to the inclined
face, we have
In = nj ej,
with the following direction cosines
nj = In.ej = cos(1n,e j).

The point h of the inclined face is such that

 Fundamental Laws; Principle of Virtual Work 339

By considering this figure we have ni > 0 and we observe that

and thus

Let us prove the Cauchy theorem (1822-1823).

PR22 The stress vector t(ilj on the inclined face of the small tetrahedron is linearly
dependent on the inclined face direction.

More precisely, this proposition lets introduce the definition of a fundamental tensor:
D r:tr The Cauchy stress tensor, also simply called the stress tensor, is the (~}-tensor u
such that
(5-45)

Proof of PR22. The linear momentum principle expresses that the resultant of the external
forces acting on the tetrahedron, that are body forces and surface forces, balances the
inertia forces whose orthogonal Cartesian components are
Yj dm = P Yj dp .

Thus, this is mathematically rendered as

or

By considering the formula of the mean this equation is written:

3
a tjiJ)(r) = V(PYj - ~)(q) - La; tY'j)(rJ,
i=1

where qED, rES, ri E Si' where a and a i are the areas of the respective surfaces S and
Si' and where V is the volume of D.

By dividing by a and by making tend k to zero we know that Via, IIphll and Ilppill all
together tend to zero.

Thus we obtain at p by continuity:

340 Chapter 5

From the following obvious relation between the areas

aj =Injla

and knowing that nj = -e j for each face Sj, we have at p (with nj > 0):

tjit) = In,,;e,).
3

l=!

Thus the stress vector t(it) = t(n,e,) on the inclined face is expressed from the stress vectors
t(ei ) as follows:
3
t(it) = I nj t(e;) . (5-46)
j~l

We set
t(e')=a .. e j' (5-47)
Jl

where aJi represents the jth component of the stress vector on the face whose outward
normal vector is e j , that is:
t j(e i ) =a ji' (5-48)

Thus (5-46) is written:

with summation over i and j.

Thus the components <it) with respect to the basis (e k ) are such that for every j:

(5-49)

and are explicitly denoted as follows:

(5-50)

This is the explicit expression of the Cauchy theorem:

t(it) =a.n.
This means that, to each direction, the stress tensor associates a vector (the stress) as a
linear combination of the direction cosines that are the -components. n
 Fundamental Laws; Principle of Virtual Work 341

The Cauchy theorem can be widespread to any curvilinear coordinate system.

In this case the fundamental relation between the stress vector and the stress tensor is
explicitly written:
(5-51 )

We summarize in this general context:

t(iI) = IJ(iI) e .
J
Ji
=a ni e J
=a Jl (n.eJ eJ •

Remark 1. The Cauchy relation (5-49) shows the nine components a J, of the stress
tensor. The index j represents the direction of the jth coordinate axis, which is the
direction of the projection l;'i) of the stress vector t(ei ). The index i of this projection aJi
recalls that it concerns the stress vector t(e,) associated with the coordinate plane
perpendicular to e i •
For instance, the next figure shows
t (e2 ) = a l2 e l + a 22 e 2 + a 32 e 3 ,

where a l2 is the component of the stress vector in the el -direction which acts on the
plane whose outward normal is in the e2 -direction, and so on.
X3
...
~2 "

X2

Fig. 32

In this example a l2 and a 32 are shear stress components in the plane of outward normal
e 2 , a 22 is the normal stress component.
The 6 shear stresses a ij (i ~ j) and the 3 normal stresses a ij (i = j) are represented in the
following figure.
342 Chapter 5

~-~-X2

Fig. 33

Remark 2. We note that some authors define

tIe,) = (j ..
lj
e j'
and thus

where

The Cauchy stress tensor is transposed in this representation. However, from the moment
of momentum principle we will see that the stress tensor is (generally) symmetric, and
thus the choice of definition is not important.

2.4 CAUCHY'S STRESS TENSOR AND THE PRINCIPLES OF DYNAMICS

The two fundamental principles of dynamics can be expressed form the consideration
of the stress tensor.
The notations have previously been introduced.

2.4.1 Linear Momentum Principle and Equilibrium Equations

Eq. (5-37) is immediately written:

.!!...- r pVdJl=
dt JD, D,
I
fdJl+ r (T·ln da
JaD,
(5-52a)

(5-52b)
 Fundamental Laws; Principle of Virtual Work 343

iv, (p r - 1 - diva) dp = O. (5-52c)

Since this equation must be satisfied for any material domain Dt , we deduce the local
expression of the motion equation, also called the Cauchy equation 01 motion:
p r=l+diva (5-53)

p (av + gradV . V) =1 + diva

at
av V2
p ( - + (curIV) /\ V + grad - ) = 1 + diva. [because (4-79)]
at 2

From Eq. (5-17), it is also written:

~ (p V) + div{pV ® V) =1 + diva.
at
In orthogonal Cartesian coordinates the motion equation (5-52b) is written:

1, pr; dp = fD, /; dp + LD, O'ij nj da. (5-54)

Locally we have the following motion equations

i=1.2.3, (5-55a)
also written:
(5-55b)

Thus we have obtained the following system:

p r x = Ix + axO'xx + ayO'xy + azO' xz'

pry = Iy + axO' yx + ayO' y.y + azO' yz' (5-56)
p r z = Iz + axO'zx + ayO' zy + azO' zz .

In general curvilinear coordinates the equations of motion are written:

(5-57)

Indeed, the introduction of the covariant derivative leads to the invariance of the ~otion
equations under any change of coordinate system; in addition, these equations are those of
the motion in orthogonal Cartesian coordinates.
344 Chapter 5

From the well known expressions of the divergence of a (~)-tensor with respect to the
orthonormal basis (I r ,10,1 z ), we obtain the equations of motion in cylindrical
coordinates:

(5-58)

From the well known expressions of the divergence of a (~)-tensor with respect to the
orthonormal basis (Ir,1o ,1¢), we obtain the equations of motion in spherical coordinates:

(5-59)

Equilibrium equations

If there is no acceleration, for instance at equilibrium, we obtain the general equations:

(5-60)

In particular, in orthogonal Cartesian coordinates the equilibrium equations are

;; + O'ik,k = o. (5-61)

2.4.2 Moment of Momentum Principle

From Eq. (5-38) we deduce that the moment of momentum principle is expressed
in integral form as follows:

~ f X/\PVdfl = f x/\fdfl
dt JD, JD,
+ f
JaD,
x/\O'.1n da, (5-62)

also written:
f X/\P
JD,
r dfl = JD,
f x/\fdfl+ f x/\O'·1n da.
JaD,
(5-63)
 Fundamental Laws; Principle of Virtual Work 345

PR23 The moment of momentum principle implies the symmetry of the Cauchy stress
tensor.

Proof Since the Cartesian components of the vector x /\ u·1n are

we deduce that

We note that the previous result leads to

r
JoD,
(x /\ U ·1n)da = r
Jill
(p: U + x /\ divu) dp. (5-64)

and we have locally obtained:

div(x /\ u ·1n ) = x /\ divu + p: u , (5-65)

where p is the well known orientation tensor of components Pijk = E: Uk in orthogonal

Cartesian coordinates.

Let us go on with the proof.

Eq. (5-63) is written in terms of components:

1)1 E:ijk(Xj(PYk - fk -Uks,,)-uk)dp=O.

By taking account of the linear momentum principle this equation is reduced to

r E:ijk Uk} dp = O.
JD I

Since the integration domain is arbitrary we deduce:

E:Uk Uk} = O. (5-66a)

This result means that

(5-66b)
346 Chapter 5

We have explicitly obtained

0'32 -0'23 =0 (for i = 1),
0'31 - 0'13 = 0 (for i = 2),
0'12-0'21=0 (for i = 3 ),
namely:
Vi, j E { 1,2,3 } : 0' ij = 0' ji .

We have proved that the stress tensor is symmetric:

(5-67)

Remark. If there is a couple C per unit volume then the local form of the equation of
moment of momentum principle is written:
X 1\ P r = x 1\ f +X 1\ divO' + p : 0' + C (5-68)
or explicitly:

Therefore instead of
P:O'=O,

we have

and thus it is impossible to conclude that 0' is symmetric any longer.

2.4.3 The Generalized Cauchy's Theorem

Keeping in mind the previous remark, we summarize the generalized Cauchy's

theorem:

PR24 The general equations of continuum dynamics are

p yi = r + V' kO'ik ,
 Fundamental Laws; Principle of Virtual Work 347

2.4.4 Poisson's Theorem

The reciprocal of the Cauchy theorem is the Poisson theorem (1829), which is
expressed in tensor form as follows:

PR25 A symmetric tensor field a such that for every unit normal n
t(ii) = a.n (5-45)
and
p r=f + diva (5-53)
implies that the equations of the linear momentum and moment of momentum
principles are verified.

Proof First, Eq. (5-53) leads to

r p r dp = r f dp + r diva dp
~ ~ ~
or equivalently:
r
Jot
p r dp = r f
JDt
dp + r
JaDt
tin) da .

This equation really expresses the linear momentum principle.

Secondly, by multiplying both members of Eq. (5-55) [which makes (5-53) explicit] by
£: pqi Xq and integrating, we obtain in orthogonal Cartesian coordinates:

b£:pqiXqpridp=
, b£:pqiXqJ;dp+
, b£:pqiXqaik,kdp
,
r r 8
= .b, £:pqixqJ;dp + Jt'D, £:pqi(-(Xqaik)-aik
8xq
-)dp
m k 8x k

= r £:pqixqaiknk da + JD,r £:pqi(XqJ; -aikOqk)dp.

JiJD,

The last term £: pqPiq vanishes since it represents terms which cancel two by two, for
instance (p = 1):

and so on.

Thus the first part of the hypothesis lets us write

which is actually the expression of the moment of momentum principle.

Until now the internal forces have not gone into action, it will not be the case any more
with the consideration of kinetic energy and power, as viewed in the following sections.
348 Chapter 5

3. THEOREM OF KINETIC ENERGY

For once, let us begin with the Eulerian description.

3.1 THEOREM OF KINETIC ENERGY IN THE EULERIAN DESCRIPTION

Let us prove the following theorem of kinetic energy.

PR26 The time rate of change (or material derivative) of the kinetic energy related to a
material domain of a continuum is equal to the sum of the powers of all external
and internal forces.

Proof The scalar product of terms of each member of the motion equation by the velocity
vector V immediately leads to

~p£.(V.v) == f.V + V.diva. (5-69)

2 dt

Let us prove that

V.diva == div(a . V) - a: d , (5-70)
where we recall that d is the symmetric part of gradV.
Indeed, we have
div(a· V) == div fa. V + a: gradV (5-71)
since, given an orthonormal Cartesian basis, we know that
(a· Vh == aki Vi
and thus
div(a· V) == ~(aji V,)
ox)

oCa)ij oV;
== Vi + aji
ox) ox)
==(divfa)i Vi +aji(gradV)ij
== div I a . V + a: gradV .

Since a is a symmetric tensor Eq. (5-71) is written:

V. diva == dive a . V) - a : gradV .

Since the anti symmetric part n of gradV does not contribute to the double contracted
product (because a is symmetric), then Eq. (5-70) is proved.
 Fundamental Laws; Principle of Virtual Work 349

From this result Eq. (5-69) becomes

1 dV 2
- p - = f. V + div(u· V) - u: d . (5-72)
2 dt
It is a local expression of the theorem a/kinetic energy.

By integrating over the domain D/ occupied by the material system at t, we obtain

r -1 pdV
2
- d p - r f.Vdp+ r div(u.V)dp- r u:ddp
JD, 2 dt JD, Jo, Jo,
= r
Jo,
f.Vdp + r
JoD,
(u.V).lnda - r
Jo,
u:d dp

Since

which is immediately proved by using coordinates and the symmetry ofu , the previous
equation becomes
1 dV 2
r -p-dp = r
ob. 2 dt JD.
f.v dp + r
JoD.
t(n).v da - r
Jo,
u: d dp. (5-73)

At this stage of the proof let us introduce the following important definitions.

D The power of external forces is the power owed to body forces and owed to
surface forces
G't,/' ~e) -1
-
VI
f. V dp + L t . V da .
Dt
(n) (5-74)

The power of internal forces is

G't,/' ~i) = - r u: d dp. (5-75)

JD.

By introducing the following

D The kinetic energy of the material system is
r
T = Jo, "21 PV 2 dp, (5-76)

and since
r -.!. dV 2 d -!!..- r -.!. V 2 d
JD, 2 P dt p - dt JD. 2 P P,
Eq. (5-73) clearly expresses the theorem o/kinetic energy, namely:
dT
-=~e)+~i) . (5-77)
dt
350 Chapter 5

The theorem of kinetic energy is explicitly:

-dJ 1
-pVY,dfl
dt ), 2
= 1!,v,dfl + 1 v,t,
D, aD,
(Ii) da- 1v, O'ydlj dfl· (5-78)

Remark 1. The internal power g>(i) is zero for any rigid body motion since d = 0 at every
point and each time.

Remark 2. The theorem of kinetic energy follows from the motion equation (linear
momentum principle) and does not constitute a further fundamental law, which means that
there are 7 fundamental equations (six equations of principles and one of continuity) with
13 unknowns (p, 11" a y). At this stage 6 equations are lacking.
We note that the fundamental equations are 'universal' whereas the six additional
equations, which will later be called the constitutive equations, will necessarily be
dependent on the physical properties of the continuum.

3.2 THEOREM OF KINETIC ENERGY IN THE LAGRANGIAN DESCRIPTION

Let us consider the theorem of kinetic energy by referring to the initial

configuration.

(i) First, let us tum our interest to the power of internal forces.
Given the motion defined by x = (/J(X,t) let us consider the convective transport of the
following integral

1 u(x): d(x,t) dfl·

D,

From the following equalities with second order tensors [see (1-28)]:
t : (u . v) = u : (v . t) = v : (t . u) ,
we deduce that
a : d = u : (' F- Ii . F- • I )

= (F-I ·(a· 'F- I »: i (5-79)

This result may also be obtained by considering a coordinate system.

Thus, since
 Fundamental Laws; Principle of Virtual Work 351

the convective transport of the introduced integral into the initial configuration is
expressed as

An interesting tensor appears:

D The Piola-Kirchhoff stress tensor is

7r(X,t) = J(X,t)F- I (X,t)· u(x)· IF- I (X,t), (5-80a)
memorized as follows:

~=F-I. U . IF-I. (5-80b)

Po P

PR27 The power of internal forces is expressed in the following purely Lagrangian form
~i) =- r 7r(X,t):i(X,t) df.1o. (5-81)
JDo

Proof From Eq. (5-79)

u:d= (F-I·u·IF-I):i,

which is also written:

I .
u:d= -7r:L, (5-82)
J
we deduce the expected result:

~i) = - JD,
r u: d d f.1 = - JDo
r 7r: i df.1o .

From this result let us observe the following equality:

r
JI
u r
-:d dm = JI
7r .
- : L dm. (5-83)
M D, P Mf)o Po

PR28 The Piola-Kirchhoff stress tensor is symmetric.

Proof It is obvious because

17r = J F- I . I U . I F- I = J F- I . U. I F- I = 7r .
352 Chapter 5

Training. Let us establish Eq. (5-82) in orthogonal Cartesian coordinates.

From
Try = J(F-1);p O"pq (F-1)qj
= J (F-1);p O"pq (F-1)jq
_ ax; ax)
-J-O"pq--'
fu p fuq

we deduce:

and thus

Since t F . d . F = i we actually have:

I . I .
u:d =-Trmk(L)km =-H:L.
J J

PR29 The Piola-Kirchhoff tensor which acts on the area element vector dA (in the initial
configuration) is connected to the element of force df = t(ll) da exerted on the
transported surface element of area da as follows
H(X,t)· dA = F- 1 (X,t)· df . (5-84)

Proof The force exerted on a surface of area da and oriented by the orthogonal unit vector
1" is denoted
df = ten) da = u ·In da,
that is
df = 0" ·da, (5-85)
where we recall that u, as well as the stress vector ten) = 0" . In' have the dimension
[ML- 1T-2 ] of a force per unit area.

But the area element vector dA is transported into the current configuration, which leads to
da =J tF- 1 ·dA.
Thus we have

and finally:
 Fundamental Laws; Principle of Virtual Work 353

F- 1 ·df = J F- 1 ·u· IF- 1 ·dA

= 7r·dA
= 7r·l N dA,

which is actually Eq. (5-84).

n" df
da

Fig. 34

Remark. Given the nature of each member of this relationship, we can say that the
Piola-Kirchhofftensor shows a 'purely Lagrangian character'.

Let us recall that our goal is to express the theorem of kinetic energy in Lagrangian form.
After settling the term of internal power let us consider the external power.

(ii) Secondly, let us deal with the following term of external power
r t(n).V da
l3DI
also written:

= r
JaDI
u.V.da

= r
1/)0
u .V.(J tP-'·dA)

= r
l3Do
V. J u. 1 F- 1 • dA . (5-86a)

Before continuing let us introduce another tensor.

D The Piola-Lagrange stress tensor, or Boussinesq tensor, is

B(X,t) = J(X,t) u(x) . IF-'(X,t), (5-87a)
or equivalently,
B(X,t) = F(X,t)· 7r(X,t). (5-87b)
354 Chapter 5

We immediately deduce:
7r(X,t) = F-1(X,t)· B(X,t) (5-88)

Eq. (5-86a) is now written:

r
JoD,
t(n).Vda= r
JoDo
V.B.dA= r
JoDo
IB.V.dA. (5-86b)

Remark 1. Unlike the Piola-Kirchhofftensor this new tensor is not a 'purely Lagrangian'
tensor.
Indeed, by acting on a area element vector, the Piola-Lagrange stress tensor leads to a
force which is exerted on a surface element in the current configuration, namely:
B·dA =F ·7r·dA = df.

Remark 2. In general the Piola-Lagrange tensor is not symmetric.

Indeed, its components

are generally different from Bji .

(iii) Next, the other term to be considered in the power of external forces is

(iv) Finally, since the first member ofEq. (5-77) expressing the theorem of kinetic energy
is written:

-d! -1 P V 2 dp = ! 1 dV 2
- - - dm =
! I dV 2
- Po - - dpo ,
dt D, 2 MD, 2 dt Do 2 dt
this theorem of kinetic energy has the following expression in the Lagrangian description:

fDo
1 dV 2
"2 Po -----;;t dpo = t J f· V dJ.1o + LDo lB· V. dA - fDo 7r: i dpo· (5-89)

Denoting the body force per unit mass (acceleration!) by 1= f ,the local expression of
p
the theorem of kinetic energy is then
1 dV 2 - . I .
- Po - - = Po f· V + dzv ( B . V) - 7r : L. (5-90)
2 dt
 Fundamental Laws; Principle of Virtual Work 355

Return to the motion equation

To conclude this paragraph let us consider the motion equation in the Lagrangian
description. But before stating the equation of the linear momentum principle let us recall
that any tensor quantity q(x,t) is expressed in Lagrangian variables as follows:
q(x(X,t),t) = Q(X,t).

We take the liberty of denoting this quantity by the same notation q.

Thus we successively denote:

f p(x,t)q(x,t) df-l = f q(x,t) dm

JD, JM D ,
= JDo
f Po(X) q(x(X ,t),t) df-lo.

By omitting the variables, this integral is written:

f pqdf-l = f pq J df-lo = f Poq df-lo.

Jot JDo Joo

Let us use this simplified writing for the acceleration:

f p r df-l = f prJ dJ10

JD, JDo
= .100
f por dJ1o.

Thus the equation of the linear momentum principle, namely:

JD, 1
f p r df-l = D, f df-l + JoD,
f (T ·da

is written in the Lagrangian description:

f Poydf-lo= f Jfdf-lo+ f B.dA. (5-91)

JDo JDo JaDo

By introducing the divergence in Lagrangian variables and denoting by j the body force
per unit mass as above, the local expression of the motion equation is then

Po Y = Po j + div B , (5-92)

that is, given an orthonormal basis of a Cartesian frame:

- aBy
Po ri = Po.f + ax. . (5-93)
,
356 Chapter 5

4. STUDY OF STRESSES

We are going to study the properties of the stress tensor from the Cauchy's stress
vector in the current configuration.

We know that the stress tensor (J and the stress vector t(n), where " is the unit normal
vector to an infinitesimal surface element l , are connected by the relationship
lil) = (J . "

and the element of force acting on this surface element of area da is

tin) da = (J'" da = (J ·da.
We also recall that
tie,) = (J·e = (Je
I Jt j'

where (J ji is the Jth component of the stress vector on the plane whose outward normal is
in the e i -direction, that is:
_tie,) =t(e,)
(5-94)
O'ji- .e; ; .

Let us extend the stress notion.

4.1 RECIPROCITY OF STRESSES

D CF The normal stress relative to the infinitesimal surface element of unit normal " is
the component of tin) in the direction of ".
It is denoted by (J nn .

Thus it is such that

(J nn = tin) . "
=«(J.").,,.
Thus we have obtained:
(Jnn=n.(J·n. (5-95)

I In mechanics the vector" is chosen outward and the surface element is automatically oriented.
This choice being made, we recall that a traction is a vector such that the angle between it and " is acute, a
compression is a stress vector such that the angle between it and " is obtuse.
 Fundamental Laws; Principle of Virtual Work 357

In an orthonormal Cartesian basis it is:

From (5-49) we deduce:

(5-96)
which is actually the explicit form of (5-95).

D 'iF The shear stress' relative to the infinitesimal surface element of unit normal ;, IS

the component of t(it) tangent to this element.

It is denoted by , .

Fig. 35

Of course, the squared shear stress is

,2 = (0" .;,)2 -(O"nn)2 = (0" .;,)2 -(;'.0" .;,)2. (5-97)

In orthonormal Cartesian basis it is written:

Let us consider two faces of respective unit normal vectors;' and ;".

The property of the reciprocity of stresses is the following

PR30 The projection on ;" of the stress vector acting on the face of normal ;, is equal to
the projection on ;, of the stress vector acting on the face of normal ;".

I In French: 'Contrainte de cisaillement' or 'Cission '.

358 Chapter 5

Proof From the symmetry of a , we deduce that

a . ii . ii' = ii' . a . ii = a . ii' . ii ,
which in orthogonal Cartesian coordinates is actually:

This property is illustrated in the following figure by the distance equality

pa= ph.

Fig. 36

In the interesting case of orthogonal faces we obviously have:

(Yn'n = (Inn'·

4.2 PRINCIPAL STRESSES

The following developments are usable in every context where a symmetric tensor
of second order exists, which is the case for the Cauchy's stress tensor a .
The problem consists in finding the eigenvalues and eigenvectors associated with a. In
other words, let us look for a vector, for instance ii, such that the stress vector t(il) = a . ii
and ii are collinear.
If a denotes an eigenvalue the well known requirement is rendered as follows:
t(il) = a.ii
=an,

or explicitly:
t(il) =an
1 I'

that is:
 Fundamental Laws; Principle of Virtual Work 359

or equivalently:
(5-98)

Given any symmetric tensor, (T for instance, let us recall that there are at least three
mutually orthogonal eigenvectors; that is, there exists at least three principal directions
and we say:

D rJF The principal directions associated with (T are called the principal stress
directions. The eigenvalues of (T are called the principal stresses.

These values are obtained from the homogeneous system (5-98) provided that the
characteristic equation be verified, namely:
(5-99)

which, in an orthogonal Cartesian coordinate system oxyz, is written

(Txx-(T (Txy (T xz

(T yx (Tw-(T (T yz =0. (5-100)

(Tzx (Tzy (Tzz-(T

The following three properties are valid for general symmetric second order tensors, [see,
e.g., Talpaert (2002)].

PR31 The eigenvalues of the symmetric tensor (T are real.

Proof Let (Tt be an eigenvalue and ,,(I) the corresponding eigenvector; that is:
(I) _ (I)
(Tijnj -(Ttni •

Let

be a change of basis, where the various a ji are the direction cosines of vectors of the
primed basis with respect to the vectors of the unprimed basis.
We choose the change of basis such that e; and ,,(I) are collinear. Thus for the eigenvalue
at the stress vector has the direction of e; == ,,(l) whose components are clearly

Every surface element perpendicular to a principal direction has no shear stress!

360 Chapter 5

In the primed basis the characteristic equation becomes

0"1 - 0" 0 0
0 0";2 - 0" 0";3 =0.
0 0"~2 0"~3 - 0"

For 0" *- 0"1 the problem is reduced to taking the following equation into consideration:

whose discriminant
(0";2 -0"~3)2 +4(0";3)2

is always more than or equal to zero. We conclude that the two other eigenvalues are also
real.

PR32 The eigenvectors of the symmetric tensor 0" which correspond to different
eigenvalues are orthogonal.

Proof Given the eigenvalue 0"1 we have

(O"y - 0"15y) njl) = O.

Given another eigenvalue 0"2' we have

(O"y - 0"25y) n;2) = o.

By multiplying the first equation by n?), the second by n?), and subtracting the results
we obtain:

which implies

and thus the principal directions are orthogonal.

PR33 If two eigenvalues are equal then there exists a plane of eigenvectors.

Proof Let 0"1 = 0"2 be two eigenvalues equal to (f .

There are two eigenvectors such that
- s: ) (I) -
( O"ij-O"uij nj -
0,

(O"ij -(f5ij)nY) =0.

 Fundamental Laws; Principle of Virtual Work 361

By multiplying the first equation by an arbitrary real k and the second by an arbitrary real
K, and adding the results we obtain:
V k,K E R: (O"y -iYoij)(knY) + Kn;2») = 0,

which means that every linear combination of orthogonal unit vectors ,,(1) and ,,(2) is also
an eigenvector.
Thus every vector of the plane defined by two eigenvectors is also an eigenvector.

Notation. Let us choose an orthonormal basis (El' E2 ,it)) whose vectors have principal
stress directions.

Given this principal basis and denoting the principal eigenvalues by 0"1' 0"2 and 0"), the
matrix of (F is diagonal and the stress tensor is written:

(5-101)

It is also written:
0" = 0"1 EI ® EI + O"n E2 ® E2 + O"III E3 ® E3, (5-102)

where the subscripts I, II, III specify that the principal stresses are ordered, namely:

In this case the stress tensor is explicitly denoted by

o
oo ] .
O"Ill

The stress vector t(iI) = 0" ." on any face has the following expression with respect to a
principal basis:

o
n = [0"1~]
0]o .[nIl
2 n0"2 2 •

0"3 n3 0"3 ~

At this stage the following important proposition is obvious.

PR34 The stress vector on a face perpendicular to a principal direction of 0" is

orthogonal to this face and its component equals the corresponding principal stress.

The stress vector and the principal direction are collinear in this case. There is no shear
stress!
362 Chapter 5

PR35 The normal component of the stress vector is extremum in the principal directions.

Proof The problem consists in obtaining the n-directions making

t(ii). n = t?i)n i = (jij njni
an extremum.
We are looking for an extremum with the constraint condition
n l2 + n 22 + n32 = njnj = I .

We recall that the Lagrange multiplier method consists in introducing Lagrange

multipliers; in this case it is one arbitrary real A whose value has no interest in this
context.
n
The -directions which are suitable for the question are those making an extremum the
following expression:

that is:

((jkj -Aokj)n j =0.

This equation really leads to the n-principal directions as shown by Eq. (5-98).

PR36 The maximum shear stress is equal to one half of the difference between the largest
and smallest principal stresses and takes place in the plane which bisects the right
angle between the directions of the maximum and minimum principal stresses.

Proof We know that the components of the stress vector t(ii) are such that

Thus the normal component of t(ii) is

(jnn = t(ii).n = tfn)n i = (j] n~ + (ju ni + (jIll ni· (5-103)
 Fundamental Laws, Principle of Virtual Work 363

Consequently the squared shear stress is

(5-104)

The Lagrangian multiplier method leads to the three equations

8 2
-8 [, - A(nin i -1)] = 0
ni
which are
n, (o} - 2a, (a, n~ + an n~ + alii nff) - A) = 0,

n 2 (a,~ - 2 an (a, n~ + a n n~ + a III nff) - A) = 0,

By taking into account nin i = 1 we can solve the previous system and find particular
solutions (n" n 2 , n3 ) that are the unit vectors
(±l,O,Q) , (O,±l,O), (O,O,±l) .

For each of these solutions we deduce from (5-104) that ,=

O. These solutions
correspond to the case where the shear components vanish on every plane orthogonal to a
principal direction (called 'principal plane'). On such planes ,2 is minimum (zero!).

Other solutions are

for which ,2 =~(a,-an)2,
4

for which ,2 = ~(a, - a nd 2 ,

4

for which ,2 =~(an -a nd 2 .

4

The expression ~(a, - alii) gives the maximum shear stress value, which is equal to half
2
the difference between the largest and smallest principal stresses.

The maximum shear stress takes place in the plane which bisects the right angle between
the directions of the maximum and minimum principal stresses.

This is illustrated in the following figure where the maximum shear is obtained on the
plane element which is parallel to the principal direction associated with an and bisects
the right angle of two other principal directions.
364 Chapter 5

Fig. 37

The following particular stress tensors are very interesting.

D A stress tensor is said to be a plane stress tensor if its components, for a given
indexj, are such that a
aij =0

whatever the index i ;

in other words, if one and only one of the principal stresses vanishes.

D A stress tensor is said to be an axisymmetric stress tensor (about a principal

direction) if only two eigenvalues are equal.

For example, if a, = a 2 7= a 3 , the stress tensor is axisymmetric about £3 -axis.

D A stress tensor is said to be a spherical stress tensor or a uniform stress tensor if

the three principal eigenvalues are equal.

In this last case every direction is automatically principal.

4.3 STRESS INVARIANTS; DEVIATOR

Let us recall the important notion of invariant.

D Every function of components of a tensor is said to be an invariant if it is

independent of the coordinate system.
 Fundamental Laws, Principle of Virtual Work 365

(i) First, let us define invariants from the characteristic equation. It is well known that its
coefficients are invariants.

PR37 The characteristic equation of u, that is

det(O'Ij -O'0ij)=O,

is expressed as follows:
3 2 1
0' -O'uO' +'2(O';;O'j/-O'ijO'j;)O'-det(O'ij)=O. (5-105)

Proof First, let us establish the following lemma.

The expression of det( 0'ij) is

1
D = - Ii pqrliijkO' pi 0'qjO' rk •
6

Indeed, by recalling that the mixed product of three vectors with respect to an orthonormal
basis is
UAV.W = U.VAW= Ui(VAW)i

UI U2 U3

= VI V2 V3

WI W2 W3

and by considering

we obtain

and by permuting two lines this determinant becomes:

The previous results are rendered as follows:

and thus the lemma is proved because Ii pqrli pqr = 6.

From this lemma we deduce the following expression of the characteristic equation of 0':
Ii pqrliijk (0' pi - 0'0 pi )(0' qj - O'Oqj )(O'rk - O'0rk) =0
366 Chapter 5

The coefficient of a 3 is

The coefficient of a 2 is made up of three terms of the following type

and so on.

Finally, the characteristic equation wanted is actually

3 2 1
a - ali a + - (alia jj - a ij a ji )a - det( a ij ) = 0 .
2

The previous proposition has established that the coefficients of the characteristic equation
are invariants, which are the next ones given any orthonormal basis (not necessarily
principal):
- ajirst degree invariant: (5-106)

- a second degree invariant: (5-107)

- a third degree invariant: 13 = deta . (5-108)

We note that a 2 = a· a is the square of the stress tensor. Its contraction is actually:

In conclusion, the expression of the characteristic equation is

(5-109)

(ii) It can be proved (Cayley-Hamilton) that all the invariants of degrees higher than 3 are
expressed as functions of invariants of the first, second, and third degrees. Let us illustrate
this in the case of a .
Let us consider the stress tensor referred to principal axes:

The nth power of a has the following diagonal form

 Fundamental Laws, Principle of Virtual Work 367

~·1·
o

0'3

Thus 0' and all its integer powers have the same principal directions.

Each of the principal values 0'1' 0'2' and 0'3 must satisfy Eq. (5-109) and thus 0'3 is
written:

1]0']2- 12 0']+13 0

0'3= [ 0 1]0' 22- 12 0'2+ 1 3

Hence we have:
0' 4 =1]0' 3 - I 20' 2 + I 30'
= 1](/] 0'2 -1 2 0'+13 1)-12 0'2 +130'
= (/]2 _ 1 2 ) 0'2 + (/3 - I] 1 2 )0' + I] 13 J .

Thus 0'4 is a linear combination of 0'2, 0', and J, the coefficients of which are
polynomials in I], 12 and 13 ,

In addition, the matrix multiplication of 0'4 by 0' leads to the conclusion that 0'5 IS a
linear combination of 0'2, 0', and J; etc ..

By defining

then the Cayley-Hamilton property is satisfied since the traces of 0'2 and 0' are functions
of I] and 12 ,

(iii) Let us introduce two definitions which allow to decompose 0'.

D The value 0'5 = ~ trO' is called the mean normal stress:

3

(5-110)
368 Chapter 5

Let us introduce the following spherical stress tensor:

(5-111)
o
It is obvious that
(5-112)

D The deviator of a is the tensor

(5-113)

It is obvious that

and
(5-114)
smce

From the previous definitions it is evident that:

PR38 The stress tensor is decomposable into a spherical stress tensor and a deviator
stress tensor:
(5-115)

PR39 The tensors a and aD have the same principal directions.

Proof It is obvious since a and aD are different from as.

Remark 1. If the deviator vanishes then the stress tensor a is spherical, and thus every
direction is automatically principal.
In this case the stress is necessarily normal.
The choice of the name deviator is justified because the deviator gives the deviation of the
constraint from the normal direction.
 Fuudameutal Laws, Principle of Virtual Work 369

Remark 2. The principal values of (T and (TD are such that

(T
I
= (Ts + (TD
I

and thus
(5-116)

Consequently, given aprincipal basis of (T, we have

o o
o o
o o

Of course, if the basis is not principal we have

D S
(Til = (Til - (T

1
= -(2(Tll - (Tn - (T3J) ,
3
and so on.

Remark 3. We can give an interesting interpretation of the mean normal stress (Ts and of
tr( (T D) 2 by considering the octahedral plane. What is it about?
It is the plane which makes equal angles with the principal stress directions. Let us
consider the following unit vector perpendicular to this inclined octahedral face with
respect to the principal basis of (T :

- - - "7
, I
I

Fig. 38
370 Chapter 5

The stress vector on the octahedral face is immediately decomposed into two parts:
t(ri) = u.}n = us.}n + u D .}n'

First, we observe that the first term of the sum is normal to the face and is such that

Next, let us prove that the second term of the sum is tangent to the inclined face. Indeed,
we successively have
D f3 D ~ ~ ~
U ·}n =-u ·(EI +E2 +E3 )·
3
fj D~ D~ D~
= -(UI E] +u 2 E2 +u3 E 3 ) ,
3
since u D has the principal directions of u .
Thus
D 1 D~ D~ D~ ~ ~ ~
U ·}n.}n =-(u] E] +u2 E2 +u3 E3)·(E] +E2 +E3)
3
1 D
=-tru = O.
3

This really proves that uD·}n belongs to the face and this tangential part of t(ii) is called
~ }
the octahedral shear stress, while the normal part u·}n = --L}
3 n
is called the octahedral

normal stress.

To conclude the promised interpretation the norm of the octahedral shear stress is

(5-117)

4.4 STRESS QUADRIC OF CAUCHY AND LAME STRESS ELLIPSOID

Let px\ x 2 X3 be a system of orthogonal Cartesian axes at a point p of a continuum.

(i) The components u ij of the stress tensor u define a quadric whose equation is

uijX,Xj =K (K E R),

the real character of this surface imposing the sign of K. Weare going to study this
quadric.
 Fundamental Laws, Principle of Virtual Work 371

Let q be an arbitrary point belonging to this quadric. The vector position of q is

pq=r=rn,

where, at p, n is the unit vector orthogonal to a surface element of the continuum.

The coordinates of q are
Xi =rn i ,

where the various ni are the direction cosines of r.

The components of the stress vector t(iI) at p with respect to an orthonormal basis of the
local Cartesian frame are
(Ii)
ti = (7ij n j •

The normal component of the stress vector t(iI) is

XiX j K
(7 nn
(iI) - (iI)
= t . n = Ii n i = (7 ij nin j = (7 ij - 2 -
r
= 2 .
r

Thus we say:

D The locus of points of position vectors rn whose coordinates Xi verify

(5-118)

is called the stress quadric of Cauchy.

x,
Fig. 39

In particular, if we consider the locus of points of position vectors

n
r=---
~1(7nnl '
372 Chapter 5

that is, the points which are 1/~IO"nnl away from p in different directions of ;" then the
equation of the stress quadric of Cauchy is
(5-119)

Of course, this equation is also obtained from

with
X=~'
I v'FnJ
PR40 If the Cauchy stress quadric is known then the stress vector t(n) , which acts on a
surface element at point p orthogonal to the position vector r of any point q of the
quadric, is parallel to the normal to the quadric at q.

This is illustrated in the previous figure.

Proof. From the equation of the Cauchy stress quadric

<I> =) O"ijx;Xj - K=0
we deduce:

that is
t(iJ) = _1_V<I> .
2r

(ii) Besides the previous geometrical representation obtained from the normal component
of tin) we can consider the Cartesian components:
(n)
t; = O"ij n j ,

and thus
n; = (0"
-I) ij I (n) .
j

Consequently from

we deduce:
 Fundamental Laws, Principle of Virtual Work 373

In particular, if we consider the locus of points which are Ilt(ti) II away from p in all the
directions of t(ti) , in other words, whose coordinates are

then we obtain the Lame stress ellipsoid defined by

(5-120)

More particularly, if we express the stress vector in the principal basis of (1, namely:

then we have
X 1 -- t(ti)
1 -
- (1
1
nl'

Thus the Lame stress ellipsoid is defined by

(5-121)

This ellipsoid will be connected with the forthcoming study of Mohr's circles.

Example 1. For a uniform tensor of traction or compression, whose spherical stress tensor
IS

we have

and thus the Cauchy stress quadric is the sphere of equation

2 2 2 K
Xl + X 2 + X3 = - .
(1

Example 2. For a simple shear of stress tensor such that

0 T 0]
[TOO ,
000
374 Chapter 5

we have

and thus the Cauchy stress quadric is the hyperbolic cylinder parallel to the X3 -axis of
equation

4.5 GEOMETRICAL CONSTRUCTIONS AND MOHR'S CIRCLES

The graphic method has lost much of its interest in comparison with the analytical
method, but it keeps a geometrical significance allowing us to visualize the stresses and to
have a first knowledge of the sign of the shear stresses.

4.5.1 (Mohr's) Stress Plane

We know that every stress vector t(n) on a surface element (or face) is the sum of a
normal stress vector and a shear stress vector:
t(ii) = (J
nn it + ri ,

where i is the unit vector along the projection of t(ii) onto the face.
There are two possible senses of orientation for t(ii) .

The orientation in the plane defined by the face plane associated with t(iJ) and it is given
by the angle (it,i) = +1[/2, but, of course, we might have chosen the opposite sense for i.

~-
A

O"nn ft

q'
Fig. 40
 Fundamental Laws; Principle of Virtual Work 375

D The plane {o,n,i } is called the stress plane or plane of Mohr.

Notation. In this study we denote the normal component ann of the stress vector by a.
Thus the stress vector is the vector sum
(5-122)
where r denotes the tangential (or shear) component of the stress vector.
The point q of coordinates (a, r) in the plane of Mohr locates the stress vector tin) which
can rotate about the origin o.
In this plane the vectors oq and oq' correspond to a compression, whilst oq" represents a
traction.

We note that some studies are restricted to a half of the symmetric stress plane and
consider the points (a, Irl)
since the choice of i is arbitrary and the sign of the shear
stress is not of fundamental importance.

4.5.2 Stress Vector and Plane of Mohr

With respect to a principal basis (if1 ,if2' itJ) , the stress vector

has the following components

where the direction cosines satisfy the well known relation

222
n] + n 2 + nJ = 1. (5-123)

Since
in) A _ ~ ~ ~ ~ ~ ~

t .n - (a]~ E] + a 2 n2 E2 + a 3 n3 E3)'(~ E] + n2 E2 + n3 E 3)
we deduce that the normal component a of t(ii) with respect to the principal basis is
(5-124)

In addition we have

that is,
(5-125)
376 ChapterS

Problem. Given the stress tensor U is there a unit nonnal it direction such that the values
U and r are the respective nonnal and shear components for this direction?

It is important to recall that each point (u, r) in the stress plane represents a stress vector.
The system of Eqs. (5-123), (5-124) and (5-125), where the squares of the direction
cosines
(u - U 2 )(u - (3) + r2
(5-126a)
(u 1 - u 2 )(u, - (3)

2 (U-CT 3)(u-u1)+r 2
n2 = (5-126b)
(CT2 - CT3)(CT2 - u,)

2 (U-CT,)(u-u 2 )+r 2
n3 = (5-126c)
(CT 3 -U')(U3 - CT2)
are unknown, can be solved provided that the principal stresses are distinct.
The three principal stresses are assumed to be distinct and ordered (without restriction)
according to

Since the second members of Eqs. (5-126) are not negative, we have the following
relations

(CT-U n )(U-CT m )+r 2 ~O,

(CT-U m )(CT-CT,)+r 2 ::;;0,
(u-u,)(u-u n )+r 2 ~O.

These expressions, which are obvious given the chosen order of principal stresses, are
equivalent to the following inequalities:

In the stress plane these inequalities delimit the allowed values of components CT and r
of the stress vector.
The allowed domain of the stress plane (a, r) is

- exterior to the circle C, of center on the u -axis at CTn + u m and of radius

an - am
2 2 '
- interior to the circle C 2 of center on the a -axis at a I + am and of radius a I - am
2 2 '
 Fundamental Laws; Principle of Virtual Work 377

exterior to the circle C3 of center on the (j -axis at (jI + (jll and of radius (jI - (jll
2 2

--.--+------~~-- __--~--~-CT
OJ,,

Fig. 41

Of course, every point of C I has coordinates such that n I = 0 ; that is, such that the stress
vector has no component along the principal direction corresponding to E]. That means
that when;' rotates about the originp of the stress vector t(n) in the plane {p;E 2 ,E3 }
perpendicular to EI -principal axis, the point q of the stress plane, which locates the stress
vector, describes the circle CI .
The reader will conclude for C 2 and C3 in the same manner.
For every direction of the unit nonnal ;" the point q locating t(n) belongs to the hatched
part in the plane of Mohr.

Conversely, given any point q of the hatched part in the stress plane, that is such that the
values of ( j and r verify the previous inequalities, then the unknowns nl, n~ and n~
can be detennined from Eqs. (5-126a,b,c). Given (j and r there are 2 3 different
directions for ;,.

Remark 1. The representation of the so called Mohr's circles confinns previous analytical
results, as for example:
- The maximum nonnal stress is (j] ; it is obtained for the surface element perpendicular
to the principal direction associated with (j] .
- The maximum shear stress is (ji - (jm and thus the normal stress is equal to (ji + (jm .
2 2

In the stress plane the points m and m' show this last situation. They belong to the Mohr's
circle C 2 and correspond to the direction cosines such that

n]
2
= n32 =-1
2
378 ChapterS

We again find the case in which the maximum shear component acts in the plane which
bisects the right angle between the £] and £3 principal directions which correspond to
the respective maximum and minimum principal stresses.

Remark 2. If two principal stresses are equal, for instance CT[ = CT u "* CTUI' then the circles
C] and C 2 coincide with the circle with center (0, CT] + CT 1II) and of radius equal to
2

CT] - CTlII ,while C3 is reduced to a point.

2
If the three principal stresses are equal, in other words if CT is spherical, then the three
Mohr's circles are reduced to a point CT] = CT n = CTm = a on the CT -axis. The stress vector
has no shear component. It is the well known case in which every direction is principal.
If a is positive the stress tensor defines a uniform traction; if a is negative the stress
tensor defines a uniform compression.

4.S.3 Description of Mohr's Circles

Let us carry out a short study of Mohr's circles.

For example, let us consider the great circle every point of which has coordinates such that
n2 = 0, that is, the stress vector has no component along the principal £2 -direction.
From the principal directions, let us form the direct (i.e., 'right handed') basis
(£]'£3'£2)·

The intersection of the plane { 0; £] , £3 } and the Lame ellipsoid is the ellipse of equation

With respect to the basis (£]'£3)of the plane orthogonal to £2' the stress vector
t(ir) = pq has the following components

that is, actually

with
n[ = cos¢, n3 = sin¢,

where ¢ is the oriented angle (£],n) such that the basis (£],n'£2)isdirect.
 Fundamental Laws; Principle of Virtual Work 379

Let us choose the sense of orientation of i such that the basis (ii, i, E2) is direct, this unit
tangent vector is written:

i = -sin¢ £, + cos¢ £3 .
.
t
,/

Fig. 42

In the stress plane the normal and tangential components of t(Ii) = oq are immediately:
a =t(n). n
= a, cos 2¢ + am sin 2 ¢
= a, +am + a, -am cos2¢
2 2
and
r = I(Ii). i
=-a, cos¢ sin¢ + am sin¢ cos¢
= _,
a -am sin2¢.
2

These components of t(Ii) satisfy the equation of the great circle C 2 and thus q belongs to
this circle.

Fig. 43
380 Chapter 5

When the small face turns about the it2 -axis; that is, as the unit normal ;, rotates in the
plane {p; it1, it3 } through an angle ¢, then the extremity q of the stress vector describes
the circle C2 in the opposite orientation sense, through an angle - 2¢ .

In the case in which ¢ = 0, that is, a = a, and r = 0, then ten) and ;, are collinear. This
is the case in which the small face is perpendicular to the principal direction associated
with a,.

In the case in which ¢ = ,,/2, that is, a = am and r = 0, then ten) and;' are collinear.
This is the case in which the small face is perpendicular to the principal direction
associated with a Ill .
All the previous results are transposable to the Mohr's circles C 1 and C3 by taking care
that orientation maintains an angle (;" i) = +"/ 2 .

In any case, the geometric support leads to a better understanding of the following
particular situations.

4.5.4 Particular Stresses

(i) Plane stress

D A stress is said to be a plane stress if one and only one of the principal stresses is
zero.

Given a principal basis three situations are possible, namely:

[~' ~l [~
0 0
all 0 o ,1
0 0 a",

with the corresponding Mohr's stress circles

Fig. 44
 Fundamental Laws; Principle of Virtual Work 381

The first case is illustrated by the following situation. At an unloaded point on the body's
surface which is free from any stress, the stress vector is tin) = 0 along the direction of
the unit normal ir, that is, in the principal direction associated with the principal stress
O"IlI = 0 :

In Chapter 6 plane stresses will be introduced in the case of stresses which are not
principal.

(ii) Uniaxial stress

D A stress is said to be a uniaxial stress if one and only one of the principal stresses
is nonzero.
It is a uniaxial traction if the principal stress is positive.
It is a uniaxial compression if the principal stress is negative] .

Given a principal basis let us choose, for instance, the principal stress 0"] different from
zero. Thus we have

0"]
0 " [= 0 0 0 .
0 0]
o 0 0

The direction of the stress vector is principal and if if] denotes a unit vector along this
direction we have

sInce
~ n -
t 3in) -- '-'3 0.
3 -

Of course, the stress vector vanishes for every small face parallel to the principal direction.

The normal component O"nn = O"ij n, nj of the stress vector is simply written:

and by considering the shear stress T we have

2 2
T + 0" = 0"]2 n]2 ,

1 Also called a simple traction and a simple compression.

382 Chapter 5

that is,
,2 + a 2 - aa j =0 .
Thus in the stress plane the extremity of the stress vector only belongs to the circle of
equation

C=------~~--_cr

Fig. 45

In other words, the extremity of t(ill describes the circle C 2 = C 3 as it varies.

(iii) Simple shear

The notion of simple shear has already been introduced. For instance, let us view it for
two orthogonal e j - and e3 -directions.
Given

[~ ~ a~3l'
a l3 0 0
the characteristic equation immediately leads to the following principal values:

and the principal directions are defined by

~ 1
Ej = fi(e j +e 3 ),

t
m

crIll - + - - - - - - : I = - - - - ! I - - - -

Fig. 46
 Fundamental Laws; Principle of Virtual Work 383

Let us choose the (direct) principal basis (E;), the stress tensor of a simple shear is of
type:

and we say:

D A stress is said to be a simple shear stress if one of the principal stresses vanishes
while the other two ones are opposed.

This situation is illustrated in the stress plane as follows:

Fig. 47

From Remark 1 of Section 4.5.2 we deduce that the maximum of the absolute value of the
shear stress is 11'1 = IO'( I, whilst the normal stress is zero.
This situation is obtained at points m and m' of C2 which correspond to the direction
cosines such that

This is the case in which the unit normal ;, on the small face is inclined at an angle - 1r/4

with respect to E(, that is l' = 0'(; this is also the case in which the unit normal ;, is
inclined at + 1r/4 with respect to E(. In other words, it is obtained on the face parallel to
the principal direction associated with O'n = 0 and which bisects the right angle between
other principal directions.

(iv) Triaxial stress oj revolution

D A stress is said to be a triaxial stress oj revolution if two principal stresses are

equal and different from zero.

Given a principal basis let us choose, for instance, the tensor

384 Chapter 5

[~ o
o 1.
alII

We know that in this case every principal direction associated with a l = all is
perpendicular to if3 .
This situation is represented in the stress plane as follows:

o~~-------------+~

Fig. 48

In the stress plane the extremity of the stress vector only belongs to the circle C] of
equation

The circle C3 is reduced to the point a] = all of a -axis.

5. PRINCIPLE OF VIRTUAL WORK

The method of virtual work is essential in both theoretical] and continuum

mechanics. It consists in introducing arbitrary vectors a priori; such vectors will next be
called virtual displacements 2 or virtual velocities according to the considered context.

5.1 PRELIMINARY RECALLS

Let us recall basic notions.

I For example, concerning this method applied to analytical mechanics the reader could refer to Mechanics,
Tensors & Virtual Works (Talpaert, 2002).
2 This concept was introduced by Jean Bernoulli in the 18th century, of course under another formulation.
 Fundamental Laws; Principle of Virtual Work 385

D The generalized coordinates are the independent coordinates required to specify

the position of each particle of a system.

They are denoted by

qj (j = l, ... ,n).

A virtual displacement of a particle p is an association between two neighboring positions

at a same instant. More precisely, it is a vector that not necessarily represents a
displacement! A fixed time is imposed at any virtual displacement.
The positions of the particle are given by r: (t,ql ,... ,qn) H r(t,ql , ... ,qn) and let us
denote a virtual displacement of p by or.

Arbitrary increments oql of generalized coordinates lead to arbitrary increments of the

components of the position vector:

i = l, ... ,n.

Thus the expression of a virtual displacement ofp is

that is
n ar .
or= L - j oql. (5-127)
j=l aq

Thus we say:

D A virtual displacement of a point p is a vector which associates with p a

neighboring point from the variation of the only generalized coordinates.

Besides this notion let us recall the virtual velocity of a particle.

The expression of the velocity of p is

V = ar + ar 'l+~ '2+~.3
at aql q aq2 q aq 3 q

As for virtual displacements, we consider a fixed time and define:

386 ChapterS

D A virtual velocity ofp is an arbitrary

V' = ~ ~ "j
L..,.q,
j=1 aql

where the various q' j are arbitrary reals.

These arbitrary reals will be denoted by q' j and thus a virtual velocity of p is expressed
as
_ ~ ar >j
V • -L..,.-.q. (5-128)
j=1 oql

5.2 RIGID BODY MOTION

Let 9{, be an inertial frame of reference,

D/ be a set of particles of a continuum,
aD, be the boundary of D/.

Let us recall that the dynam [F(e)] of external forces is defined by the following elements
R(e) = r fdp+ r t(ii)da,
.hI JaD,

By denoting a virtual velocity field by V' and a field of virtual displacements by 8r we

introduce:

D The virtual power of external forces exerted on D/ is

.9l:) = 1,
r f. V· dp + JoD,
r t(ir). V· da. (5-129)

D The virtual work done by external forces on D t is

OT(e) = i ,
f.8r dp + LaD, t (n)
.8r da. (5-130)

(i) Firstly, let us express the virtual power of the external forces as the product of
dynams of external forces and virtual displacements.
Given an arbitrary reference point 0 'fixed' in a rigid body, the well known expression of
the virtual velocity field for a rigid body motion 1 (e.g. Taipaert, 2002):

1 Called in French: 'Mouvement rigidijiant'.

 Fundamental Laws; Principle of Virtual Work 387

v; = V~ + (J)' 1\ Op

leads to
g>(:) = .br f. V; dfJ + r
JaD
t(n). V; da
t I

= i t
f.( V0• + (J) • I\Op) d/1 + 1 t . (V' +(J)' 0'P) da
aDt
(n) 0 1\

=v;.(r fdfJ+ r t(n)da)+{J)'.(r 0pl\fd/1+ r OpAl(n)da)

~ ~ ~ ~
= v~.R(e) + (J)·.M~),

that is:

where (J)' is an (arbitrary) vector o/virtual rotation.

PR41 The virtual power developed by a system of external forces for any virtual rigid
body motion is zero if.fthe dynam of external forces is zero (equilibrium).

Proof First, if g>(:) = 0 for every virtual rigid body motion, that means that

In particular, this equation must be verified for (J)' = 0 and an arbitrary V~ (that is, a
translation), which implies R(e) = 0 .
But (J)'. M~) = 0 must be satisfied for an arbitrary vector of virtual rotation (J)', which
implies M~) = 0 .
Thus the dynam of external forces vanishes.

Conversely, the assumption

implies that g>(:) = 0 for every virtual rigid body motion. Thus the proof is complete.

By analogy with the virtual velocity expression of a rigid body motion we consider
(infinitesimal) virtual displacements of type
orp = oro +oOI\Op,

where the arbitrary vector 00 represents a virtual angular displacement.

388 Chapter 5

The corresponding virtual work is

&(e) = or.R(e) + oO.Mg)

= [R(e)] .
M(e) or
[00] .
o 0

From this result we obtain in the same manner as before:

PR42 The virtual work done by a system of external forces for any virtual rigid body
motion is zero iff the dynam of external forces is zero (equilibrium).

(iiJ Secondly, let us consider the general explicit expression of ~(:) , namely:

~:) = 10r1 !y* df.l f t?')

+ JoDI "'* da. (5-131)

It is also

that is

D . tensor d • 0 f components d •
The symmetnc = -21 (V·!,J. . + V·J,I.) is called the virtual
I}

strain rate tensorl.

The antisymmetric tensor .0* of components .o~ = ~ (Vi~j - Vj~i) is called the rate
of virtual rotation tensor2.

Thus the virtual power developed by a system of external forces is written:

~:) = 1 (f + diva). V'

t
df.l + 1ta: .
d' + fa:.o'
b
df.l
1
df.l . (5-132)

We know that the virtual power developed by a system of forces in equilibrium is zero for
every virtual rigid body motion.
More particularly, for a motion of translation (such that every velocity field is constant)
the various derivatives "'~j vanish.

I Called in French: 'Taux de de/ormation virtuelle'.

2 Called in French: 'Taux de rotation virtuelle'.
 Fundamental Laws; Principle of Virtual Work 389

In this case the virtual power is reduced to

.b (J; + ay,)V,* dp = O.
I

Since this equation is verified for every virtual translation and every volume, we obtain
the local equations o/translation equilibrium:
J; +aij,j = 0

or equivalently
f + diva =0. (5-133)

In addition, since for every virtual rigid body motion the virtual strain rate is zero (d* =0)
the third integral of g>(:) must vanish:

(5-134)

Since this condition is fulfilled for any domain D, we deduce that

Denoting by aWl the components of the anti symmetric part of a the previous equation is
equivalent to

Since the virtual rotation is arbitrary we conclude that

that is

These equations I are sometimes called the local equations 0/ the rotation equilibrium and
are equivalent to
(5-135)

5.3 EXPRESSIONS OF VIRTUAL POWER (AND VIRTUAL WORK)

From the considerations of the previous section, at equilibrium, Eq. (5-132) is

reduced to the following expression of the virtual power ofal! external forces:
~:) = f a : d* dp .
.b,

1 The couples are not considered.

390 Chapter 5

In this section let us consider arbitrary virtual velocity fields; from the previous expression
we immediately define:

D The virtual power of internalforces exerted on D is

i
~:) = - , u: d' dJi . (5-136)

Of course, for every rigid body motion we have:

g;(;) = o.

Now let us calculate the virtual power of all forces in a general way.
The virtual power of external and internal forces:

is written from Eqs. (5-129) and (5-136) as follows:

g;'= f j.V'dJi+ f t(fi).V'da- f u:d'dJi (5-137)

.lvt JaDt .lvt
or explicitly:
(5-138)

In the last integral which represents the virtual power of internal forces, we equivalently
obtain:
u:d*

= u: (gradV' -n')

= div(u, V*) - V*. div t u - u : n* ,

this last equality directly follows from

div(u· V*) = V'. div t u + u : gradV' .

Therefore the power of external and internal forces is written:

g;'= f (J;+uy,)V,'dJi+ f
.6, JaD, (t;n)-uyn)V,*da+ .bf uJ2;dJi
t
(5-139)

that is
g;*= J~ U+div'u).V*dJi+ ~f (t(fi)-u·n).V'da+ ~f u:n'dJi. (5-140)
 Fundamental Laws; Principle of Virtual Work 391

Remark. The general expressions of the virtual work are analogically:

(5-141)
or

5.4 PRINCIPLE OF VIRTUAL WORK

First, let us define the virtual power owed to acceleration.

D The virtual power owed to acceleration is

~:CC) = i I
pro V' dp. (5-143)

Let us state the principle o/virtual work.

PR43 At every time and for every domain of a mechanical system of a continuum, there
exists (at least) an inertial frame such that:
(i) The virtual power of external forces ~(:)' that of internal forces ~(;)' and that
due to acceleration ~:CC) verify the fundamental relation:

(5-144)
(ii) The virtual power of internal forces is zero for every rigid body motion.

Remark. For every rigid body motion the virtual power of external forces is equal to the
virtual power due to acceleration.

Training. The reader will state the analogous form of the virtual work principle :
(5-145)

PR44 The principle of virtual work for any velocity field leads to the following
equations:
pr =/+divtu
t(iI) = u.;'

u= tu
and conversely.
392 Chapter 5

Proof (i) For every domain we assume that 9>(:) + 9>(~) - ~:CC) = 0, namely:

.6
f (f+ div 'a'). V' dJl + f (t(n)-a.n).V'da+ f a:il' dJl-
1 -'<)Df .bt
J P'Y.V'dJl=O .
Vi

(5-146)

Firstly, let us choose V' = 0 on D, and aD, except on a domain D' where V' is constant.
The second integral of (5-146) vanishes since V' = 0 on aD .
The third integral also vanishes because V' = 0 implies V/j = 0 and thus the various Q;
are zero.
Thus 9>' is reduced to

t (/ + div a - P r). V' dJl = O.

t

Since the domain D' is arbitrary and V' is an arbitrary constant vector we obtain the well
known fundamental equation of motion

P r =/ + div t a (a)
or explicitly:

Secondly, let us choose V' = 0 on D and on aD except on aD" c aD where V' IS

constant.
The third integral vanishes since V' = 0 on D.
Knowing that
P r = / + div t a ,
we conclude that the second integral of(5-146) is such that

1D" (tUI) - a· n). V' da = O.

Since aD" is arbitrary and V' is an arbitrary constant vector we obtain:

(b)

Now, from the previous results (a) and (b), the hypothesis (5-146) implies that the third
integral is zero for every domain and thus

where the various a[ji] denote the components of the anti symmetric part of a.
 Fundamental Laws; Principle of Virtual Work 393

The virtual rotation being arbitrary, we conclude that

a[ji] = 0,
that is,

Conversely, from
pr=/+divta,
we immediately deduce that

which completes the proof.

PR45 At equilibrium the virtual power of all external and internal forces is zero for every
virtual velocity field and conversely.

Proof By considering the previous demonstration it is proved that

divta+ / =0,

t(ii) = a ·n.

6. THERMOMECHANICS AND BALANCE EQUATIONS

To conclude this fundamental chapter of continuum mechanics we are going to

search for a type of equation capable of taking into account both Eq. (5-5) of the axiom of
conservation of mass, Eq. (5-37) of the linear momentum principle and Eq. (5-73) of the
kinetic energy theorem notably. Such a type of equation will provide a balance equation
every time connected with a fundamental law of continuum mechanics.

6.1 BALANCE EQUATION

6.1.1 Proper Motion

Let us consider a domain Dt included in a material system.

We assume that Dt has a proper motion such that there is a velocity field W at every
point of aDt •
394 ChapterS

As before, V denotes the velocity field of particles of the material medium.

The comparison between Vand W leads to the relative velocity field: Vr = V - W .

(i) Scalar quantity

We assume there is a scalar quantity defined at time t by

Q(t) = L-D, p(x,t) q(x,t) dp ,

where p is the usual density (mass per unit volume) and q is the (local) quantity per unit
mass.
First, the change of Q can be caused by the motion of the domain Dr with respect to the
material medium. Indeed, exchanges between Dr and the external medium exist through
the boundary aDr • This phenomenon is called the convection.

D The convectionjlux of Q through the boundary aDt is the surface integral

(5-147)

where rpc = P q Vr is said to be the convection jlux vector l .

We note that this notion is well known. Indeed, from t to t+ dt the quantity passing
through a surface element of area da corresponds to the volume r • dt da of a cylinder IV iii
of basis area equal to da; thus this quantity is per unit time:

pq IVr."1 da .
In the definition the choice of the relative velocity Vr is appropriate because it concerns
the particles passing through the boundary, this phenomenon following from the motion of
the medium with respect to Dr (actually: V - W = Vr).

Moreover, in the case of a material domain Dr (Vr = 0) the domain is made up of the
same particles during the time; that is actually: <Dc = 0 .

By denoting the time derivative 'along' the proper motion by ~ ,the rate of change of Q
dt
owed to the convection is written:

~ .b pqdp = -<Dc
dt '
=- r _ p q Vr . da .
.lJD,

1 In French <Dc is called the flux de convection and rpc is the vecteur densite surfacique de flux convectif
 Fundamental Laws; Principle of Virtual Work 395

The minus sign is correct since for a convection flux such that Vr . > 0 the flux is n
outward, which leads to a decrease of Q (negative derivative).

Second, the change of Q can be caused by particles passing through the boundary on both
sides independently of the motion of jj(. This phenomenon at the microscopic scale is
called the diffusion.

D The dijJusion flux of Q through the boundary ajj( is the surface integral

<l>d r_ (hnda,
= ,l,Dt (5-148)

where rpd(X,t) is said to be the dijJusionflux vector l .

The change of Q owed to the diffusion is expressed with the minus sign before <l> d' since
for a diffusion flux such that rpd' n> 0 the flux is outward, which leads to a decrease of Q.
Third, besides the convection and diffusion, a change of Q can be caused by a source of Q
within jj( which can produce or remove a part of Q with regard to jj( .
The gained or lost part of Q due to the source is denoted by

Qs = 1- pq s df1.
D,
(5-149)

In conclusion, this primary analysis leads to the next balance equation:

(5-150a)

that is

~ r pqdf1=-r_ pqVr.da- rpd.da+ r_

pq s df1. r (5-150b)
d t !s, JoD, ,l,Dt !s,

Remark. If we put together the convection and diffusion terms to define the total flux
vector

rp=rpc+rpd'
then the balance equation is written:

~ r
d t !s,
pqdf1 =-
JoD,
r_ rp.da + !s,r pq s df1. (5-151)

I In French <l> d is called the flux de diffusion and rpd is the vecteur densite surfacique de flux diffusif.
396 Chapter 5

Example. In the particular case in which the quantity Q is the mass M of D" it is
expressed as follows:
M = 1, pqdJl,
which means that p q is the element of mass per unit volume or density p, thus q = 1 .

The mass conservation law expresses that the mass of 15, remains constant during the
proper motion and there is no source within 15" that is:

qs = O.

In consequence the balance equation (5-150b) is reduced to Eq. (5-14):

~ r pdJl+ JaD,
dt 1,
r_ pVr.da=O,
which proves that the exchanges of mass through the boundary a15t are opposed to the
mass flow rate (PR5).

(ii) Vector quantity

We assume there exists a vector quantity associated with 15, and defined, at time t,
by
Q(t) = 1 p(x,t)q(x,t) dJi.
I[),

First, the change of Q can be caused by the motion of the domain 15, with respect to the
material medium.

In this case of convection we say:

D The convection jlux of Q through the boundary a15t is the vector surface integral

(5-152)

where 'Pc = pq ® Vr is said to be the convectionjlux tensor.

This convection flux tensor is also written:

(5-153)

Thus the rate of change of Q owed to convection is written:

 Fundamental Laws; Principle of Virtual Work 397

~ .b
dt I
pq dp = -cI> c

=- r _ pq (V,. da) .
.l3D,

Second, the change of Q can be caused by particles passing through the boundary on both
sides independently of the motion of D[.
In this case of diffusion we say:

D The diffusion flux of Q through the boundary aD[ is the vector surface integral

(5-154)

where fPd(x,t) is said to be the diffusionflux tensor.

Third, besides the convection and diffusion a change of Q can be caused by a source of Q
within Dt which can produce or remove a part of Q with regard to D[.
The gained or lost part due to the source is denoted by
(5-155)

In conclusion, this primary analysis leads to the next balance equation:

d
=-Q = -cI>c -cI>d + Qs, (5-156a)
dt
that is

~ .v,r pq dp = - .l3D,
dt
r_ pq(V,. da) - .l3D,
r_ fPd' da + ob,r pqs dp. (5-156b)

Remark. If we put together the convection and diffusion terms to define the total flux
tensor

fP = fPc +fPd'
then the balance equation is written:

~ r
dt .fB,
pqdp=- r_
.l3D,
fP.da+ r
.fB,
pqsdp. (5-157)
398 Chapter 5

Example. In the particular case in which the quantity Q IS the linear momentum
associated with 15(, it is expressed as follows:

en=1, pVdp,

which means that the vector quantity q is V.

Thus the convection flux of the linear momentum en is

(5-158)

6.1.2 Material Domain

In the case of a material domain D( we know that Vr = 0 (V = W) and the

· · =-
denvatlve d·IS the matena
·ld · · -d .
envatlve
dt dt

(i) Scalar quantity

From above we deduce that the balance equation (5-150b) becomes in the case of
a material domain:

(5-159a)

It is also [because (5-19b)]:

1,pddqt dp =-1 D,
'Pd· da + 1 pqs dp
,
(5-159b)

and [because (4-67)]:

i , p(aatq + V.grad q)dp = - ./aD"f 'Pd· da + 1 pqs dp (5-159c)

and [because (4-101 )]:

f (!!...-(pq)+pqdivV)dp =- f 'Pd. da + f pqs dp (5-159d)

1, dt JaD, 1,

and [because (4-105)]:

f
1,
~(pq)dp+
at f pqV.da=_f 'Pd. da + f pqsdp.
./aD, ./aD, 1,
(5-15ge)
 Fundamental Laws; Principle of Virtual Work 399

Remark. The second tenn 1aD, pqV.da does not represent the convection flux of Q
through the boundary since Vr = 0 implies that the convection flux vanishes; but we can
say this second tenn represents the convection flux through the boundary of a fixed
domain D which coincides with D t at time t.

Example 1. If Q is the mass of a material domain we find again the law of mass
conservation whose expression is a balance equation where
q = I, qJd = 0, qs = 0,
that is:
!!....lp
dt ,
dfl=o,
or
r (d p + pdivV) dfl = 0,
1, dt
or

Example 2. We recall that the theorem of kinetic energy has the following expression:
d V2
-dt ob,r p2- dfl = JD,r f. V dfl + r
JaD,
(17' V).n da - r
b,
17: d dfl.

In the particular case in which Q is the kinetic energy T, the previous equation is actually a
balance equation of type (S-IS9a), where
1
qs =-(f.V-u:d),
p
but we note that the source proceeds from internal forces and body forces too.

(ii) Vector quantity

For a material domain Dt , the (vector) balance equation (5-IS6b) becomes

!!....
dt
J p q d fl = - JoD,r qJd • da
If),
+ r
.b,
p qs d fl , (S-160a)

or

(5-160b)
400 Chapter 5

or

(5-l60c)

or

r (~(pq) + pqdivV) df-L = - LD" 'I'd' da + 1 pq, df-L,

Jo, dt
(5-160d)

or

r ~(pq)df-L+
at
Jo,
r pq(V.n)da =- JiJD,
lJD,
r 'I'd ·da + Jo,r pq,df-L. (5-l60e)

Example. We recall that the equation of the linear momentum principle is

~ 1pVdf-L=
dt ,
L u·nda+ 1fdf-L.
D, ,

In the particular case in which the vector quantity Q is the linear momentum ~, q is equal
to V and the previous equation is really a balance equation of type (5-l60a) where

q=V, q, f
=-.
p

(iii) Local formulation

We can formulate a balance equation locally in various ways. Indeed, from various
Eqs. (5-159) and by converting the surface integrals into volume integrals for any material
domain, we obtain:
dq .
p - = - dlVrpd + P q s ' (5-l60a)
dt

p(a q + V.gradq) = - divrpd + pqs' (5-l60b)

at
~(pq)+pqdivV=-divrpd+ pqs' (5-l60c)
dt

~(pq)+div(pqV) = -divrpd + pqs' (5-l60d)

at
Likewise, from various Eqs. (5-160), we obtain:

dq d' (5-l6la)
Pdi=- IVrpd+Pqs'
 Fundamental Laws; Principle of Virtual Work 401

p(a q + gradq ·V) = -div'Pd + pqs' (5-161b)

at

~(pq) + pqdivV = -div'Pd + pqs ' (5-161c)

dt

~(pq)+div(pq®V) = -div'Pd + pqs. (5-161d)

at

6.1.3 Fixed Domain

For any fixed domain D we know that W = 0 [See Section 3.4.2 of Chapter 4].

(i) Scalar quantity

Since the domain is independent of time and if the quantities under the integral
signs fulfill conditions which allow the permutation of derivative and integral signs, the
balance equation (5-150b) becomes:

1:/pq)dP = - LDpqV.da - LD 'Pd· da+ 1pqs dp. (5-162)

Unlike the case of any material domain, for a fixed domain the term r
JeD
pqV.da really
represents the convection flux through aD.

(ii) Vector quantity

In the same manner, given a fixed domain D the vector balance equation (5-156b)
becomes:

f ~(pq)dp = - r pq(V.da) - r 'Pd ·da+ r pqsdp, (5-163a)

.b at JeD JeD JD

or equivalently

(5-163b)

Example. The mass conservation, the linear momentum principle, and the theorem of
kinetic energy can be expressed from a balance equation given a fixed clomain.

With regard to the mass conservation, the balance equation (5-162), such that
q = 1, 'Pd = 0, qs = 0,
402 Chapter 5

becomes
rapdfl+r pV.da=O,
.In at JaD

which actually expresses the mass conservation for a fixed domain D and where the
second integral is the flow rate through aD .

As regards the linear momentum principle, the balance equation (5-163a) such that

q=V, 'Pd = -rr,

becomes
r ~(pV)dfl=- ~
~&
r pV(Vfz)da+ ~
r rr·fzda+ .Inr fdfl.
This is actually the Euler formulation (5-39) of the linear momentum principle.

Finally, with regards to the theorem of kinetic energy, the balance equation (5-162), such
that
1
'Pd = -rr' V, q, = -(f. V - rr : d),
p
becomes
a v2
r -(p-)dfl v2
.Inat 2 = - JraD p-V.da
2
+ r (rr· V).da + r f. V dfl - r rr: d dfl .
JiJD JD JD

6.2 FIRST PRINCIPLE OF THERMODYNAMICS

Until now only mechanical quantities have been considered. But we know that
mechanical changes bring about thermal changes and conversely (chemical and
electromagnetic energies are not considered in this book). In other words, the theorem of
kinetic energy involves mechanical quantities, but we know that non-mechanical energies
must be considered.
More precisely, we are going to obtain a balance equation for a continuum studied in
thermo mechanics .
Let D t be a material domain included in a material system.

6.2.1 Principle

The first principle of thermodynamics postulates the convertibility between

mechanics and thermal energies, more precisely it brings into play:
 Fundamental Laws; Principle of Virtual Work 403

(i) The kinetic energy of the material inside D[:

V2
T= r p-dJl,
JD, 2
(5-76)

(ii) The (mechanical) power of externalforces exerted on D[:

9>(e) = 1 f.V dJl+ L t . V da,

I D,
(n)
(5-74)

(iii) The rate of increase of total heat l into D[ :

Q=- r
!3D,
q.da+ r
ob,
rdJl, (5-164)

(the minus sign because;' is outward).

This so called rate is more exactly the rate of change of heat into D t •

The first term making up Q is the part of the rate owed to surface actions. It is defined by
q , which is called the heat flux vector per unit area (at t) by conduction following from
diffusion processes through aD[ or simply:

D The vector q is said to be the heat flux vecto?

The quantity - q . Ii under the integral sign is the rate of increase of heat per unit area, at t,
by conduction (short range actions)

The second term making up Q is due to long range actions gained by D[ , it is defined by r.

D The quantity r is called the radiant heat constant per unit volume (at t).3

A last term must be considered in the first principle. Indeed, besides external supplies
which may alter the kinetic energy (as the motion of particles of a boiling liquid), there are
those which can modify the energy of the material system without changing the kinetic
energy. Thus we introduce:

(iv) The internal energy of D[:

E(i) = i I
pu dJl (5-165a)

= L'MDI
udm. (5-165b)

I Called in French: 'Taux de chaleur reryue " more exactly: 'Taux de chaleur echangee'.
2 Called in French: 'Vecteur densite surfacique du taux de chaleur reryue' or 'Vecteur courant de chaleur'.
3 Called in French: 'Densite volumique de taux de chaleur fournie a D, par des sources exterieures '.
404 Chapter 5

D The internal energy per unit volume pu is called the strain energy function l •
The internal energy per unit mass u is called the specific internal energy or the
strain energy per unit mas;'.

The balance equation of energy or first principle of thermodynamics is the following

PR46 At every time, the material derivative of the (kinetic and internal) energy for the
material domain Dr included in a material system is the sum of the power of
external forces exerted on Dr and the rate of increase of total heat:

d
-(T+ E(i)) = ~e) +Q. (5-166)
dt

Remark. Before continuing, let us give dimensions and units:

[Q] =[ML2r3] (W), [q]=[Mr 3 ] (Wm-2 ), [r]=[ML- l r 3] (Wm-3 ),

[E(il] =[Mer2] (J), [u]=[er 2 ] (Jkg- l ), [pu] = [ML- l r 2] (Jm-3 ).

By considering Eq. (5-164), that is

Q= 1, (r- divq) dj.J,
we say:

D The evolution of Dt is adiabatic if there is no heat loss or gain for Dt , that is, if
Q=O.

In particular, an evolution such that q =0 and r =0 is adiabatic.

6.2.2 Balance Equations and Local Forms

The first principle of thermodynamics is expressed as

8V' -1
d
dt,
V2
p(-+u)dj.J=-L
2 D,
q.da+ L t(nl.Vda+
D,
1f. Vdj.J+ 1rdj.J.
, ,
(5-167)

I Called in French: 'Energie interne volumique '.

2 Called in French: 'Energie interne massique '.
 Fundamental Laws; Principle of Virtual Work 405

Of course, it is a balance equation such that

V2 1
q=-+u 'Pd =q-O"'V, qs =-(f. V +r ),
2 ' p

since, given the symmetry of 0" we recall that

(0"' n).v = (0"' V).n

The time rate in the previous principle equation can be obviously written:
d V2
.b, p-(-+u)dp
dt 2
or
av 2
f p(- ( - + u) + V.grad(- + u)) dp
v2

.h, at 2 2
or
a v v 2 2
J~ -(p(-+u))dp+
& 2
f p(-+u)V.da.
~ 2

Remark. As before, if we consider any domain 151 moving with a proper motion we
immediately obtain:
J V2 V2
-=- f_ p(_+ u) dJ1 = - f _ p(-+ u)Vr.da - f _ (q-O"' V).da + f_ (f.V + r) dJ1.
dt Jb, 2 JaD, 2 JaD, Jb,

In the same manner we obtain for a fixed domain:

-
d V2 V2
f p(-+u)dJ1=_f p(-+u)V.da- f (q-a·V).da+ f (f.V+r)df.1.
dt JJ 2 JaD 2 JaD .b

Local forms

The various local formulations of the balance equation (5-160) lead to

d V2
P - ( - + u) = dive 0" . V - q) + f. V + r , (5-168a)
dt 2
a v2 V2
p(-(-+u) + V.grad(-+u)) = div(O"· V -q)+ f. V +r, (5-168b)
at 2 2

d V2 V2
-(p(-+u)) +p(-+u) divV = div(O"· V -q)+ I.V +r, (5-168c)
dt 2 2
406 ChapterS

a V2 V2
-(p(-+u)) + div(p(-+u) V) = div(u· V -q)+ f. V +r, (5-168d)
at 2 2

Training. The reader will express every local fonn given an orthononnal Cartesian basis,
for instance:

6.2.3 Potential Energy of Body Forces

In the case in which the body force f is conservative there is a potential function
V: x H V(x) such that
f = -gradV.
The power developed by f is

r f.V dJ.1 = - 1,r gradY. V dJ.1 = - 1,r dV dJ.1 = -~ r V dJ.1.

1, dt dt 1,
By letting
CV 1
= , V(x) dJi ,
the power developed by f is written:

1, f. V dJi = _ dCV
dt
.

In this special case the first principle of thennodynamics is expressed as

~(T+E(i)+CV)= r (u·n.V-q.n)da+ r rdJl. (5-169)

~ ~ ~
It is a balance equation for the energy T + E(i) + CV such that
V2 r
q=-+u+V, IJ'd=q-u·V, qs =-.
2 p

Remark. In the case of an adiabatic evolution of a material domain Dt , if the surface

forces exerted on aDt are negligible then the first principle of thennodynamics expresses
a special conservation law, namely:
d
-(T+E(jJ +CV)=O. (5-170)
dt
 Fundamental Laws; Principle of Virtual Work 407

6.2.4 Internal Energy and Balance Equation

By subtracting the fundamental equations of first principle of thermodynamics and

theorem of kinetic energy:
d dT
-(T + E(i» = 9{e) +0, - = gJ(e) + gJ(i) ,
dt dt
we obtain the important equation
dE(i)
Tt=O-gJ(i)' (5-171)

which shows that the power gJ(i) of internal forces has opposite effects on the time rates of
T and E(i) respectively; for instance, a positive gJ(i) results in an increase of T and a
decrease ofE(i) .
Eq. (5-171) is written:

.!!.- r
dt ob,
pu dp = r
ob,
(J': d dp - r
JoD,
q.da + r rdp.
ob,
(5-172)

It is a balance equation (where the quantity Q is E(i» such that

I
q=u, 'Pd=q, qs=-(O':d+r).
p

Training. The reader will immediately express this equation in the cases of proper
motions and fixed domains.

Local forms. By taking the last results into account, the various local forms (5-160) of the
balance equation lead to the following expressions of the energy equation:

du .
p - = (J':d -d,vq+r, (5-173a)
dt

p(au + V.gradu) = (J': d -divq +r, (5-173b)

at

.!!.-(pu) + pu divV = 0' : d - divq + r , (5-173c)

dt

~(pu)+ div(puV) = (J': d -divq+ r. (5-173d)

at
408 Chapter 5

Eq. (l73b) is interesting since it holds the different terms due to heat, namely those of
- heat convection V. grad u ,
- heat conduction - div q,
- heat radiation r.

In addition this equation shows a non-steady term of internal energy p au and a term of
at
power per unit volume owed to (mechanical) internal forces (Y : d .
Let us specify that this power per unit volume owed to internal forces is put in the second
member of different equations, though this term does not derive from external forces.

Eq. (1 73a) is written as follows:

du 1 1 .
W' -=-(Y:d+-(r-dlvq). (5-174)
dt p p

This equation is rendered into the following proposition:

PR47 The rate of change of specific internal energy is the sum of the specific power (that
is, per unit mass) of internal forces plus the specific rate of increase of total heat.

Given an orthonormal Cartesian basis, this is expressed as

du 1 1
- = -(Yydij +-(r -q;J. (5-175)
dt p p ,

Remark. The first principle of thermodynamics is not concerned with the reversible or
irreversible character of the process of energy conversion.
We simply mention that the stress tensor can be expressed as a sum of a reversible part
(YR and an irreversible part (YI :
(Y = (YR + (YI .

By recalling Section 6.2.3 we can also view the part of reversible power

~~ =- i (YR: d dp
I

to be
R dZ
~(i) =-di'
where Z plays the role of a deformation energy.
Thus Eq. (5-171) leads to the following time rate

d
-(E(j)-Z)=Q+
dt
i (Y:d dp.
I
I
 Fundamental Laws; Principle of Virtual Work 409

6.3 SECOND PRINCIPLE OF THERMODYNAMICS

Let us consider the second principle of thermodynamics which deals with the type
of evolution of the material domain D t included in a material system. Let us study this
evolution by introducing two essential notions developed in thermodynamics.

First, the absolute temperature T is a positive function of the empirical temperature and is
defined at every point of the material domain and at every time.

Second, the entropy is a quantity which estimates the 'energy degradation' and shows the
'disorderly state' of the material system. It is denoted by

S= 1, ps dJ.1, (5-176)

where s is the specific entropy (per unit mass) 1•

We recall that the entropy in any material system is the sum of the entropies of its parts.

6.3.1 Principle

By referring to the rate of increase of total heat, the second principle of

thermodynamics introduces the following notion.

D The external entropy supply rate 2 is composed of a part of external heat supplied
to D t through aDt by conduction and a part gained by D t owed to external body
sources, namely:
/(e)=-1
D,
~q.da+l~dJ.1.
T ,T
(5-177)

Let us state the second principle of thermodynamics:

PR48 The material derivative of the entropy associated with a material domain Dt IS

never less than the external entropy supply rate:

dS
-:::::r(e) ,
dt
that is:
dS > r ~ dJ.1 -
dt - JD, T
faD,
~ q. da .
T
(5-178)

I Called in French: 'Entropie massique'.

2 Called in French: 'Taux d'apport exterieur d'entropie'.
410 ChapterS

We note that

We mention that the equality in the principle expression holds for reversible processes and
the inequality for irreversible processes.

For the general irreversible case (in which the reversible hypothesis cannot be considered),
the time rate of change of entropy dS is strictly greater than /(e) .
dt

The difference between this time rate of change of entropy and /(e) characterizes the
internal production of entropy and we say:

D The internal entropy production rate l is the non-negative quantity

dS
/(i) == di -/(e) . (5-179)

In this manner dS == /(e) + /(i) can be viewed as a 'generalized balance equation' where
dt
the quantity Q is the entropy S, namely:

!!.-1 ps d,u == - L ~q.da + 1~d,u + /(i), (5-180)

dtt DtT tT
and such that
q r
q ==s, qJd == T' q, == pT .

But there is a complementary term I(i) ~ 0.

We mention that

Remark. Let us observe that for an adiabatic evolution characterized by q == 0 and r == 0,

the external entropy supply rate vanishes, and thus the following consequence
dS
di== /(i) ~O

means that the entropy cannot decrease, even without any external supply.

1 Called in French: 'Taux de production interne d'entropie '.

 Fundamental Laws; Principle of Virtual Work 411

6.3.2 Clausius-Duhem Inequality

The inequality of the second principle has the following local forms:

ds . q r
p dt :2: -dzv T + T ' (5-181a)

as q r
(5-181b)
p(8i+V.grads):2:-diVT+f'

:r (ps) +ps divV :2: - div ~ +; , (5-181c)

~(ps)+div(psV) :2:-div~+~ . (5-181d)

at T T

Eq. (5-18Ia) is also written:

ds I. 1 r
p-:2: --dzvq-q.grad-+-
dt T T T

or equivalently [since T > 0 and grad ~ = -~ gradT]:

T T
ds 1
PT - :2: - div q + - q. grad T + r .
dt T

By recalling Eq. (5-173a), that is:

·
d zvq-r = u: d - pdu
- ,
dt
the last inequality becomes the Clausius-Duhem inequality:
ds du 1
u: d + peT - - - ) - -q.gradT :2: O. (5-182)
dt dt T

By introducing the function

If/ = u- Ts,

the Clausius-Duhem inequality is written:

dlf/ dT 1
u: d - p(-+ s - ) - -q.gradT :2: O. (5-183)
dt dt T
412 Chapter 5

6.3.3 Dissipation and Reversibility

Let us introduce the following natural notion.

D The internal entropy production rate per unit volume rei) IS the density
associated with
f(i) = 1 rei) dJl.
I
(5-184)

The first local form of the 'generalized balance equation' (5-180) becomes
ds . q r
rei) = p dt +dlVy- T . (5-185)

Thus the local form (5-181a) of the second principle of thermodynamics is simply written:
(5-186)

that is:
ds . 1
rei) T = pT -+ dlvq --q.gradT -r ~ o. (5-187)
dt T

The internal entropy production is connected to the dissipation so defined:

D The dissipation per unit volume is the non-negative quantity

rp = r(i)T, (5-188)

the dissipation is the integral

(/J = 1rp dJl .

I

We emphasize that the second principle of thermodynamics expresses the condition

rp~O

everywhere in D t and at every time.

It is profitable to split rp into two parts according to

rp = rp) + rp2 ,
whose definitions are the following.
 Fundamental Laws; Principle of Virtual Work 413

D The intrinsic dissipation per unit volume is the quantity

qJ] = P T -ds + d·Ivq - r , (5-189)

dt
the intrinsic dissipation is

(/J] .6
= , qJ] dJ.l .

This intrinsic dissipation results from the dissipation of mechanical energy into heat.
In other words, qJ] such that

P T -ds = - d·Ivq + r + qJl

dt
may be considered as an internal heat source.

We immediately have
ds du
qJl = a:d + peT - - - )
dt dt
dlf dT
=a:d - p ( - + s - ) .
dt dt

D The thermal dissipation per unit volume is the quantity

1
qJ2 =- T q.gradT, (5-190)

the thermal dissipation is

(/J2 =
.rb, qJ2 dJ.l .

The thermal dissipation results from heat exchanges by conduction.

Remark. The thermal dissipation per unit volume vanishes if the evolution of Dt IS

adiabatic, but also if its evolution is isothermal, since in this case gradT = 0 .

D The evolution of a domain Dt is said to be non-dissipative if the dissipation per

unit volume qJ is negligible everywhere in D t and at any time.

D The evolution of a domain Dt is said to be reversible if qJl = qJ2 = 0 everywhere in

D t and at any time.
It is said to be irreversible if qJ is positive.
414 Chapter 5

The case of (ideal) reversible evolution corresponds to the equality in the expression of the
second principle of thermodynamics.

PR49 For every reversible evolution there is no internal entropy production rate, unlike
irreversible evolutions.

Proof Since any reversible evolution is non-dissipative, that is:

(jJ
Y(i) = T =0,

we conclude that the internal entropy production rate I(i) vanishes.

To end this paragraph let us compare the various balance equations.

QorQ q orq (jJd or (jJd qs or qs

Mass m 0 0

f
Linear momentum ~ V -(1'
P

V2 1
Kinetic energy T -(1'·V -(f.v - (1': d)
2 P
1
Internal energy E(i) u q -((1' : d + r)
p

V2 q-(1"V
1
Energy T+E(i) -+u -(f.v +r)
2 p

q r
Entropy S s
T pT

6.4 CONCLUSIONS AND CONSTITUTIVE EQUATIONS

Given a continuum in which only the mechanical and thermal phenomena are
considered, let us draw up the list of independent balance equations and corresponding
unknown functions of time and positions.

We recall that the body forces f (per unit volume) and the radiant heat source r are
supposed to be known.
 Fundamental Laws; Principle of Virtual Work 415

(i) The equation of continuity:

ap + div(pV) == 0,
at
that is: I scalar equation and 4 unknowns p, Vi'

(ii) The equation of motion (balance equation oflinear momentum):

dV f + d·Iva,
p--==
dt
that is: 3 scalar equations and 9 unknowns a ij .

But the moment of momentum principle implies the symmetry of a, and thus there are
only 6 independent unknowns a ij .

Thus the total comes to 1 + 3 == 4 equations and 4 + 9 - 3 == 10 unknowns.

(iii) The energy equation

d V2
P - ( - + u) == dive a . V - q) + f.v +r ,
dt 2
that is: 1 scalar equation and 4 unknowns u, qi'

Thus, finally, there are 5 equations and 14 unknowns which are the density p, the three
velocity components Vi' the six independent stress components a ij , the three components
of the heat flux vector qi' and the specific internal energy u.

(iv) Last but not least, the inequation of the entropy

ds q r
p-~-div-+-
dt T T
introduces two unknowns: the entropy s and the absolute temperature T .

There are 16 unknowns and 11 equations are missing! We must find eleven equations in
order to hope the system of partial differential equations be solved ('to hope' because
boundary and initial conditions must be considered in addition!).

Unlike the five previous equations which are not dependent on the medium (and thus
called 'universal'), six ofthe additional equations will characterize the physical properties
of the continuum studied. These supplementary equations are called the constitutive
equations.
The remaining additional equations are three temperature-heat conduction relations and
two thermodynamics equations of state.
416 Chapter 5

Remark 1. Constitutive equations cannot take the complex properties of any particular
material into account, but they are used for defining 'ideal materials' as shown in the next
chapter which deals with ideal elastic solids.

Remark 2. For numerous problems of continuum mechanics the 'uncoupled thermoelastic

theory' can be considered; that is, the interaction between mechanical and thermal
processes can be neglected. It is the case for particular evolutions such as isothermal or
adiabatic evolutions, but also if the process nature allows the above mentioned
approximation; for example, when the heat production leads to slight variations of
temperature. In the 'uncoupled thermoelastic theory' it is assumed that the temperature is
known or determined from non-mechanical information.
Of course, the temperature field influences p and various physical constants such as
elastic or viscosity coefficients.
In the above mentioned approximation the mechanical processes are only governed by the
continuity equation and the motion equations, while there are ten unknowns p, Vi and
a ij , therefore six constitutive equations are required.

We are going to have a closer look at the previous considerations in the next chapter.
 Fundamental Laws; Principle of Virtual Work 417

EXERCISES

Exercise 1.

The velocity field of a gas moving through a 5 meter pipe of x 2 -axis is defined by

When entering (x 2 = 0) the gas has a (mass) density p equal to 1.5 kg/m 3 . Find p at
the end (X2 = 5) in the case of a steady motion.

Answer. The continuity equation (5-13) is simply

a
-(pV2 )=0.
&2
Since pV2 does not explicitly depend on X2' we immediately have

(pV2 )1 (X],X2,X,) = (pV2 )1 (X],O,X,) = 0.075 (kgm-2 S-l).

From
_ 0.075 (kglm 3 )
p - 0.02X2 + 0.05 '

we deduce the following (mass) density for x 2 = 5 (m):

pIX2~S =0.5 (kglm3 ).

Exercise 2.

An infinitesimal deformation is specified by the displacement field of components

u 1 =A(x3 -!)X1X 2 +Bx2'

u2 = C( x 3 -l)xi + DX2'
u3 = Ex~ +FX2'
where the constants A, B, C, D, E, F and I are sufficiently small and have the required
dimensions.
(i) Find the infinitesimal strain tensor.
(ii) From the continuity equation, determine the ratio of the (mass) density in the initial
configuration to that after deformation.
(iii) Find relationships between constants if the continuum is incompressible.
418 Chapter 5

Answer. (i) Since

we know that e is the symmetric part of [ ~; ].

Thus
A(X3 -1)Xl + B
.] [A(X 3 -1)x2
[ au, = 0 2C(x3 -1)x2 + D
ax ] 0 3Ex~ +F

leads to
A
A(X3 -1)x2 A (x -l)x + B
2 3 I 2 2 X1X2
A B 3E+C 2 F
e = -(X3 -1)Xl +- 2C(x3 -1)x2 + D - - - x +-
2 2
2 2 2
A 3E+C 2 F
- - - x +- o
2 X1X2 222

(ii) The continuity equation in Lagrangian description, namely:

P(X,t) J(X,t) = Po(X,O),
is briefly written:
Po =J.
P
But for infinitesimal deformations we know that
J == 1 + tre
and thus
Po
- == 1+(A +2C)(X3 -1)x2 + D.
P

(iii) In the case of an incompressible medium, since J = 1 the infinitesimal deformation is

such that
tre = 0,

that is
(A +2C)(x3 -1)x2 + D = 0
and thus
D=O, A+2C =0.
 Fundamental Laws; Principle of Virtual Work 419

Exercise 3.

In spherical coordinates r, B, rp we recall that every axisymmetric motion is such

that the components of the velocity field are only meridian and independent of the position
of the meridian plane; that is:
V=Vr (r,B)1r +Ve(r,B)Ie ·
If the motion is steady determine the components of the velocity field as well as the
corresponding vector expression as functions of a stream function.

Answer. The continuity equation for a steady axisymmetric motion is written:

. 1 8 1 8 .
d/v(pV) = 2-(pr 2Vr ) + - .--(pVosmB) = o.
r 8r rsmB 8B
We know that a differential form is exact iff its integral along any closed curve r of a
simply connected domain D is zero.
By multiplying the previous expression by r2 sinB and considering the Stokes' theorem
(Green-Riemann formula), we obtain:

II [~(pr2sinB
8r
'lJ
Vr ) + ~(prVo sin B)] dr /\ dB
8B
= H) [d(pr 2sinB Vr ) /\ dB - d(prVo sinB) /\ dr]

= 1r (pr 2sinB Vr dB - prVo sinB dr) = o.

There is a so called stream function ljI(r, B) of dimension [L3T- 1] such that

pr2 sinB Vr dB - prVo sinB dr = dljl.

We deduce that
pV =_~81j1
o rsinB 8r '

where Pc is a constant of (mass) density dimension.

Since in this problem we simply have
81j1 181j1
gradljl = - I r + --Ie,
8r rae
we immediately deduce the corresponding vector expression

pV = ~ gradljl /\ I¢.
r smB
420 Chapter 5

Exercise 4.

Given a frame of reference {o;e l ,e 2 ,e3 }, the edges of length 0.4 (m) of an elastic
cube are parallel to the Cartesian coordinates axes passing through the center 0 of the
cube. The forces to be considered are the gravitational attraction and surface forces whose
stress field is given by the tensor
0.5xj + 1.3 -0.25x2 -0.25(Xj +X 3)]
0' = 10 5 X [ - 0.25x2 1.1 - 0.85x2 (Pa).
- 0.25(xj + x 3) - 0.85x2 1.884x3 +1.2

We assume that the (mass) density of the elastic cube is constant and equal to 8xl0 3
(kg m ~3) and the gravitational acceleration equals 9.8 (m s ~2 ).
(i) By considering the previous given forces (per unit volume and per unit area) prove
that the equilibrium equations are verified.
(ii) Determine the stress vector on the plane at the center 0 whose unit normal vector is
1
Jj(e l +e2 +e3)'

(iii) What are the resultant and the sum of all moments about 0 of the stresses acting on
the face in the plane Xl = -0.2 (m).

Answer. (i) The force of gravity per unit volume has the following components

/3 =-pg=-0.784xl0 5

The equilibrium equations O'ik,k + It = 0 are the following identities expressed in 10 5 Pa:

a a
-(-0.25(XI +X3» + -(-0.85X2) + -(1. 884x3 + 1.2) - 0.784 = O.
a
Oxl oX 2 oX 3

(ii) The stress vector in) = 0' . n, at 0, is explicitly:

 Fundamental Laws; Principle of Virtual Work 421

(iii) For the plane with equation xI = -0.2 the outward unit vector is - el , and thus
n l =-1.

The stress vector at any point (-0.2, x 2' x 3 ) of this face is explicitly:

tl] [ -0.2SX2
1.3-0.1 0.OS-0.2SX 3] [-1]
[t2 = -0.2SX2 1.1 -0.8SX2 . 0
t3 0.OS-0.2Sx3 - 0.8SX2 l. 884x3 + 1.2 0

-1.2 ]
=[ 0.2Sx 2 (bars).
0.2Sx 3 - O.OS

The resultant of the stresses exerted on the domain D of the face of equation Xl = -0.2 is
the following vector:

- 1.2 10.21°2 dx 2 dx 3
0.2 0.2
0.2
0.25 1 x 2 dx 2
1°·2 dx3
0.2 0.2
0.25 L 2 1°·2
0..2 dx 2 0.2'X3 dx, - 0.05 1°·21°·2 dx 2 dx 3
0.2 0.2

-19200:
= [ 0 (N).
-800

The sum of all moments of the stresses about 0 is

M
o
= r
JD
x 1\ t(iI) da
'

where x denotes the position vector of any point of the face in the plane of equation
XI = -0.2.

This sum of moments is explicitly:

422 Chapter 5

£
0.21°. 2
0.2 0.2
(-0.05X2 + 1.2x 2) dx 2 dx 3

Exercise 5.

At a given point we consider the stress tensor

0' [1 1 2]1 (MPa).

== 1 0
2 1 1

(i) Determine the stress vector on the plane through the given point and of equation
x + y + 2z - 4 == 0
and find its normal component.

(ii) Given e; == ~ (e l +e2 + e3 ) and e; == Js (2e 2 + e3 ), express the shear component 0';3'

Answer. (i) Since the unit normal vector to the plane is

, 1
n== j6(e, +e 2 +2e3)'

the stress vector is

and
 Fundamental Laws; Principle of Virtual Work 423

(ii) We have
I _ t,(e;) - e' t,(ej)
0"13 - 1 - I'

that is,
0";3 = e;.(O" ·e;) = ~ '" 2.06559 (MPa).
-vIS

Exercise 6.

A fluid moves around a sphere of radius R. The components of the velocity field
are, in spherical coordinates:
v =~ sinO
¢ T r2
where T is a positive constant of dimension [T] .
The stress tensor is
3cR 3 sin¢
0 0 -----
T r3
0"= 0 0 0
_ 3cR sin¢
3
0 0
T r3
where c is a positive constant of dimension [ML-1T- 1].
The body forces are neglected.

(i) Find the trajectories of particles.

(ii) Give the Lagrangian description of the motion.
(iii) Is the motion isochoric?
(iv) Determine the dynam about the origin of surface forces exerted on the sphere.
(v) In the Eulerian description express the time rate of change of the distance between the
particle at point (r, 0, ¢) at time t and the fixed plane of equation ¢ = 0 .
(vi) Given two very small colored lines in the respective directions 1, and 1,p' find the
rate of shear (or shearing), and do the fibers move closer or away from each other
(angular velocity!)?

Answer. (i) It is a steady motion, the trajectories and streamlines coincide.

Their equations are
424 Chapter 5

dr dB d¢
Vr Vo R3 sinB
T r2
They are circles defined by constant values of the radial distance r and colatitude B_

(ii) Let us choose such constants respectively equal to ro and Bo-

Since the radius of the circular trajectory is ro sinBo, then the velocity of the
corresponding particle is the constant V¢ = (ro sinBo h.} and thus

V¢ t R3 t
¢=----+¢o =--1 +¢o-
rosmBo Tro
The Lagrangian description of the motion is obtained by referring to the initial constants
ro, Bo and ¢o, that is:

(iii) The motion is isochoric because

divV = 8Vr +~ 8Vo +_1_8V¢ + cotBVo +~Vr = 0_

8r r 8B rsinB 8¢ r r

(iv) Given the basis (l,,1e,1¢)' the stress vector t=u-;' at any point of the sphere is
explicitly written:

o
o o0 - ~ [1]l
0SinB] -0= 00 1
- ~ sinB 0 0 0 - 3 ; sinB

= -3~ sinB 1¢,

T
because;' = lr -
The moment of t about 0, namely M 0 = R 1r /\ t , is the following vector

lr 10 1¢
R 0 0 =3~R sinBlo_
T
0 0 -3~sinB
T

Thus the required dynam is

 Fundamental Laws; Principle of Virtual Work 425

(v) The distance between the point (r, 0, ¢) and the plane of equation ¢ = 0 is
y = r sinO sin¢ .
The time rate of change of y is

y= ay +V.grady,
at
where the gradient of y is explicitly

ay
ar
1 ay
grad y =
r ao
1 ay
r sinO a¢

Thus the distance sought is

y = Vr sinOsin¢ + VecosO sin¢ + V¢ cos¢
R3 sinO cos¢
T r2

(vi) In spherical coordinates we have the following tensor

o o
grad V = o o
o
and since
2d(x,t) = grad V + tgradV,
we deduce:
426 Chapter 5

o 0 -3!f sinB
T r3
2d(x,t) = 0 0 0

-3!f sinB 0 0
T r3

. d rate 0 f shear IS
Thus the reqUire . -
R3- 3
3- sinB
-'
T r
Except for B = 0 and e = ff, this rate is always negative. Thus the angular velocity
between the fibers is positive and the fibers move away from each other.

Exercise 7.

We consider a continuum for which every particle is viewed as a microstructure.

The motion of the system is defined by a vector field V which is the velocity of the mass
center of the microstructure and by the angular velocity vector m of the microstructure (at
x, at time t). For a mass element dm we introduce the spherical inertia tensor e
I dm ,
where k is the radius of gyration of the considered microstructure.
In this problem the couple C of body forces per unit volume and the couple-stress vector
M per unit area are taken into consideration. Furthermore, it can be proved that there is a
tensor m for every x and every time t such that at a point of a surface we have M = m . n,
n
where is the unit normal to the surface.
State the equations of the moment of momentum principle in integral and local forms by
using the rigid body theory.

Answer. Since the moment of momentum associated with a mass element dm of a

material domain D t is
(x 1\ V + em)dm,
then the equation of the moment of momentum principle is

~!
dt 'D,
(xl\pV+pk2m)dfl= r (xl\j+C)dfl+ r
obi kD,
(xl\u·n+M)da

or, equivalently, by using Eq. (5-64):

~l(XI\Pv+pem)dfl=
dt I
r (xl\j+C+xl\divu+f1:u+divm)dfl.
1,

We deduce the local form:

 Fundamental Laws; Principle of Virtual Work 427

X /\ r
P + pe dO) = x /\ f + C + X /\ diva + p : a + div m ,
dt
where we recall that p is the orientation tensor (of order 3).
It is equivalent to

X /\ (p r - f - diva) + pe dO) = C + p:a + div m

dt
and, given the equation of the linear momentum principle, this is reduced to

pk 2 dO) _- C + p:a + d·IV m.

dt
Thus the vector p:a is different from zero in general and the Cauchy stress tensor is not
symmetric any more.
Weare going to make p: a explicit by decomposing a into symmetric and
anti symmetric parts:

First, we know that the adjoint of the anti symmetric tensor of components a~k is the
covector with components
(*a A)i -_ -1 Pyk a AJk _
-
1 'd="= a AJk •
-liiJk'<Juetg
2 2

In classical Euclidean space the adjoint of the tensor a A is the vector

_ _ 1 Jk _ 1 .
a A - - Eyk a A - - P .a A•
2 2

Secondly, we have p: as = 0 .
Indeed, every component

(p : as) I = -21 li Ij
k as}
k

is equal to
1 jk _ 1 kj _ 1 jk
-lEi/gaS --2, liikjaS --2,liyk a S .

Thus the double contraction is written:

p:a=2ii A •

Finally, the equation of the moment of momentum principle is locally written:

P k 2 -dO) = C + 2 a- A + d"IV m .
dt
428 Chapter 5

Exercise 8.

Given the deformed configuration

and the following Cauchy stress tensor of a body:

u = [00 100 0]0 (MPa),

000
(i) Find the Piola-Kirchhoff stress tensor 1r .

(ii) Determine the Piola-Lagrange stress tensor (or Boussinesq tensor) B.

(iii) Calculate the vector 1r . IV on the e 2 -plane in the deformed state.

Answer. (i) From

0]
o ,
1/2
and since
J =detF= 4,

we deduce that the Piola-Kirchhoff stress tensor is

~ ~] ~Pa)
(ii) The Piola-Lagrange stress tensor is

B~FK+~O ooo 0] 0 (MPa).

0

(iii) To the unit area vector for the e 2 -plane in the deformed state (that is da = e2 )
corresponds the undeformed area vector
0
1 t 1[
dA=-; F.da=4 ~
 Fundamental Laws; Principle of Virtual Work 429

and thus

We immediately have
1{. IV = 40e] (MPa),
whose direction is not that of the Cauchy stress vector.

Exercise 9.

An elastic cube whose edges of length 1 are along the coordinate axes is m
equilibrium under the action of surface forces only. The stress tensor is

2
-X 2

°
°
where the units of the constant a are N m -4.

The infinitesimal strain tensor is

where the units of the sufficiently small constant Jl are m -2 .

(i) Knowing that the infinitesimal rotation and infinitesimal displacement vanish at
origin, find the displacement field.
(ii) An inclined plane through the points (0,0,1), (1,0,1), (1,1,0) and (0,1,0) splits the cube
into a 'left' domain and a 'right' domain.
At any point (x 1 ,x2 ,x3 ) of this plane section determine the stress vector t(iil arising
from the action across this section of the material of the right domain upon the
material of the left domain of the cube.
(iii) Determine the total moment about (0,0,0) the previous surface forces exerted on the
plane section domain D.
(iv) Express the unit elongation in the direction of the edge defined by the points (0,1,0)
and (1,1,0). Give the change oflength (that is the elongation) of this edge.
(v) Calculate the maximum value of the maximum shear stress.
430 Chapter 5

Answer. (i) The following differentials of the components of the infinitesimal rotation
tensor

dOJI3 =(S]k,3 - S3k,]) dx k = 0,

dOJ23 = (e 2k ,3 - &3k,2) dx k = 0

lead to the following expression of the infinitesimal rotation tensor

3 2 2
0 '2,u(x] - x 2 ) 0
3 2 2
OJ= --flex
2 ] -x2 ) 0 0

0 0 0

The various differential forms

dU i =(&ij+OJy)dx j

are successively
3 2 7 2
dU] =3f.1X]X2dx] +,u(-x] --x2)dx 2 ,
2 2
3 2 1 2
dU 2 =-fl(-X] +-x2)dx]-fl X]x 2 dx 2 ,
2 2
dU 3 = O.

Thus the components of the displacement field are

3 2 7 3
U] = '2 flX] X2 - '6 flX 2 ,

1 3 1 2
U2 = - - flX] - - fl XI X2 ,
2 2

(iiJ The outward unit normal at (XI' x 2 ,x3 ) of the domain D on which the forces are
exerted is

The required stress vector is explicitly:

 Fundamental Laws; Principle of Virtual Work 431

that is:

(iii) Since the equation of the inclined plane is

X2 + x3 =1
I I '
we deduce that the position vector x of any point of the plane section is (XI' x 2 , 1- x 2 ) .

Thus the moment of the stress at this point about (0,0,0) is

X I\. (iI) =

-(afi/2)X~ ° (afi/4)x I X2

.fi 2.fi 2 2 .fi3 I

= -ax l x 2 e l --a(xi x 2 +2x2 (/-x 2 »)e 2 +-ax2 e 3 (Nm-).
4 4 2
The total moment about 0 is
Mo = HD x I\. (iI) dS
and since dS = .fi dxl dx 2 it is equal to

15 15 15
M o =a-el -a-e 2 +a-e3 (Nm).
12 6 4

(iv) The unit elongation sought is

o(eJ ) = &11 = 3pxJ I ,
since x 2 =1.
The corresponding elongation is equal to

3pl io' XJ dxJ = -3 pi 3 .

2

(v) The principal stresses are obtained from the characteristic equation:

2axJx2 - -ax;
det((J'ij - (J'oJ= -ax22
(J'

° =0,
°
-(J'

O.5axJx 2 -

that is,
° ° (J'
432 ChapterS

and thus

The maximum shear stress is

and its maximum value is lal1 2 fi.

Exercise 10.

A fluid moves around a sphere of radius R and center 0(0,0,0). The expression of
the velocity field V with respect to the spherical basis (J,,l(j,lt/!) is

V = a( R -1) cosB 1, + a(R + 1) sinB Ie'

r r

where the dimension of the constant a is [L T- 1 ] •

The stress tensor is expressed in spherical coordinates as follows:

u-
_ ~[SinB
0
2
0 0]
0 0 ,
r
000

where the dimension of the constant c is that of a force [ML T -2] .

(i) Characterize the velocity from a very great distance of the origin o.
(ii) Prove that the fluid does not go across the sphere and slides without friction along the
sphere.
(iii) In a spatial description express the time rate of change of the distance between the
particle at point (r,B,t/!) at time t and the fixed plane of equation t/! = O.
(iv) Prove that the rates of shear of two small colored lines in the directions of 1, and l¢,
next in the directions of 1(j and It/!, are zero.

Answer. (i) At the limit where r tends towards infinity we have

lim V = -acosB 1, + asinB Ie
,->00
 Fundamental Laws; Principle of Virtual Work 433

and, since
e3 =cosB1 r -sinB1e,
we deduce that in this case the velocity is constant, namely:
V=-ae 3 •

(ii) The normal component of V, at any point (R,B,t/J) of the sphere, is the radial
component, which vanishes:
R
a(- -l)cosB = 0,
R
thus the fluid does not go across the sphere.
At any point (R,B,tjJ) the stress vector is written:

C . B
=-2 sm 1r'
R
since iI == 1r •

In addition, the normal stress corresponding to the component of I(ir) in the direction of
ii, namely:
_ (ir) ' _ (
(Tnn - I .n- (T·n' ).n,
'

is expressed at any point (R,B,t/J) as

. B
(Tnn == Ji2
C
sln .

Thus the shear stress vanishes because

T
2 ' 2 2 c2 • 2 c2 • 2
== «(T . n) - (T nn == - 4 sm B - - 4 sm B ==
R R
°.
(iii) The distance between (r, B, tjJ) and the plane of equation t/J == ° is
y = rsinB sint/J.
Since the material derivative of this distance is

dy = ay + grady. V
dt at
and since
434 Chapter 5

8y 18y 1 8y
grady=-1 +--10 + - - - 1 ;
8r' r 8() rsinO 8fjJ

= sinO sinfjJ 1, + cosOsin¢Jlo + cosfjJ1;,

we deduce that

dy = a(R -1) cosO sinO sinfjJ + a(R + l)sinO cosO sinfjJ

dt r r

= 2a R sinOcosOsinfjJ,
r

(iv) In spherical coordinates, the expression of gradV is

R
--cosO
r2
.!.(_(R -1)sinO-(R + 1) sinO) o
r r r
1 R R
R '0
a --sm
r2
-«(- + l)cosO + (- -1)cosO) o
r r r
o o .!.«(R +1)cotB sinO+(R -1)cosO)
r r r
and thus the rate of deformation tensor is written:
R 3R , 0
--cosO
r2
--sm
2r2
o
2R
- 2 cosO o
r
2R
o o - 2 cosO
r

In conclusion, the rates of shear 2d,,p and 2de,p are actually zero,

Exercise 11.

A fluid moves around the cylinder with Cartesian equation x 2 + y2 = 9, The

motion is defined in Lagrangian description by
r=R, O=@-aRt, Z = Z - b In(2R) t ,

where R, @, Z are the cylindrical Lagrangian variables, a is a positive constant of

dimensions [L-1T-1] and b is a positive constant of dimensions [LT- 1],
The stress tensor is written in cylindrical Eulerian variables r,O,z as follows:
 Fundamental Laws; Principle of Virtual Work 435

aa 2r2 0
f3b
2r
CT= 0 aa 2r2 0

f3b 0 aa 2r2
2r

where the dimensions of the positive constant a are [ML- 1] and those of the positive
constants f3 are [ML- 1T- 1 ].
(i) Find the velocity components in both Lagrangian and Eulerian forms.
(ii) Determine the trajectory.
(iii) In the Eulerian description (at time t) express the time rate of change of the distance
between the particle at point (r,B,z) and the fixed point (3,Jr/2,O). Express the result
in Lagrangian form and find this result directly in the Lagrangian description.
(iv) Are the fluid forces exerted on the cylinder surface tensile stresses or compressive
forces?
(v) Find the resultant of all fluid stresses exerted on the cylinder h meters in height.

Answer. (i) In the Lagrangian description the velocity of a particle identified by (R, 8, Z)
is calculated as follows

v -8,
_ atP, 1
(R.e,Z)

a
=-(r
at 1, +zl z )I(Rez)
' ,
=(ar 1 +r aB 1 +az 1 )1
at' at 8 at z (R,e,Z)
and thus

We immediately deduce the velocity in Eulerian form

V = -ar2 18 - b In(2r)l z •

(ii) The angle between the velocity V , which is obviously tangent to the trajectory, and
the fixed z-axis is constant because
V.1 z = -bln(2R).
The trajectory is thus a helix of equations

r = R, z = b In(2R) (B - 8) + Z .
aR
436 ChapterS

(iii) The squared distance between (r,B,z) and (3,1[/2,0) is immediately:

12 = (rcosB)2 +(rsinB-3)2 +Z2

and thus
1 = ~r 2 - 6r sin B + z 2 + 9 .

Since
01 =0
at
and
01 1 01 01
g rad 1 =-1 +--18 +-1
Or r roB az Z

r-3sinB 1
.Jr2-6rsinB+z2+9 '

we deduce the time rate of change of I:

dl 3ar 2 cosB b In(2r)z
- = V.gradl = --;=======
dt .Jr2-6rsinB+z2+9

It is in Lagrangian form:
dl 3aR 2cos(e-aRt) bln(2R)(Z -bln(2R) t)
dt ~ R2 -6Rsin(e-aRt)+(Z-ln(2R)t)2+ 9 ~ R2 - 6R sin(e-aRt)+(Z -b In(2R)t)2+ 9

We can obtain this result directly in Lagrangian description. Indeed, from

1= ~R2 -6Rsin(B-aRt)+(Z -bln(2R)t)2 +9

we deduce the previous result as follows:

all _ - 6R(-aR) cos(B - aRt) + 2( -b In(2R»(Z - b In(2R)t)

(ReZ) - ,
at " 2~R2 -6Rsin(B-aRt)+(Z -bln(2R)t)2 +9

which confirms the previous result.

(iv) The cylindrical surface is 'in traction' under the influence of tensile forces which are
such that the (positive) normal stress is
t(n) . ;, = t U,) . 1r = (J'
rr = aa 2 R2 ,

where;' = 1, is the outward unit normal vector.

 Fundamental Laws; Principle of Virtual Work 437

(v) The stress vector is expressed as

t U,l=u·1
r
=u.[~l=aa2R21 +fi b 1 .r 2R z
o
The resultant of fluid stresses exerted on the domain D of the cylindrical surface is

R = fIv t(1,) dS = HD (t,1, + t z 1J dS.

But the r-component of this resultant vanishes since there are only pairs of equal and
opposite vectors tr 1r . Thus we obtain:

fib
R=- f2" fh rdBdz1 z =7rfibh1z •
2R 0 0

Exercise 12.

With respect to the rectangular Cartesian axes OX j X2 X3 an elastic rectangular

parallelepiped is in equilibrium under only the action of stresses exerted on the faces of
respective equations Xl = 0, Xl = a, x 2 = 0, x 2 = b, X3 = 0 and x3 = c (a,b,c E R+).
The expression of the stress tensor is
a Q(X2 -b)(2x3 -c)
U= [ 0
o
such that the dimension of the positive constant a and that of the constant k are
[ML- 3 r 2 ], while the real Q belongs to ]0,1[.
This problem of small displacements is characterized by the following infinitesimal strain
tensor

where f.1 is a positive 'sufficiently small' constant of dimensions [L-2 ].

(i) Give the value of k to maintain the parallelepiped in equilibrium.
(ii) Find the maximum of the maximum shear stress relating to the face in the plane
X2 =b.
438 Chapter 5

(iii) Calculate the resultant of all stresses exerted on this face.

(iv) Express the infinitesimal rotation tensor (j) which vanishes at (a,b,c/2) and find the
displacement field which vanishes at this point.
(v) From the continuity equation calculate the ratio of the (mass) density after
deformation to that in the initial configuration.
(vi) Determine the domain of the parallelepiped where small fibers parallel to x 2 -axis
lengthen.

Answer. (i) The equilibrium equations a ij,j == 0 are explicitly written:

alj,j == 0 ,
a 2j ,j == a(2x3 - c) + k(2x 3 - c) == 0,
a 3j ,j ==0.
Thus there is equilibrium if k == -a .

(ii) For x2 == band k = -a the stress tensor becomes

0 0
a = [0 0
o -ax 3 (x 3 -c)

The principal stresses which follow from

_a 3 +aa 2 xi(x3 _c)2 == 0
are respectively
a I == -ax3 (X3 - c), a 11 ==O,

The maximum shear stress value is

T == -ax3(x3 -c)

and its maximum is equal to a c 2 /4 .

(iii) Since the unit normal to the face concerned is ;, == IX2 ' the corresponding stress vector
is
 Fundamental Laws; Principle of Virtual Work 439

and the resultant of all stresses on this face is

R= Ioadx[ IC tdx 3 =-aacl

0
1
6
3
z
2
•

(iv) First, from

and since
OJ12 (a,b,c/2) = 0,

we deduce that

Second, from

and since
OJ[3 (a,b,c/2) = 0,
we deduce that

Finally, from
dOJ 23 = (C 2k ,3 -C3k ,2)dxk
= 2p(q-l)(x2 -b)dx2 + P(2X3 -c-q(2x3 -c»dx3
= p(q -1)[2(x2 - b)dx 2 - (2x 3 - c) dx 3J

and since

we deduce that
2 C 2
OJ23 =p(q-1)«x2 -b) -(X3--»'
2

The various differential forms

are successively:
dU J =0,

dU 2 = p(q -1)(x2 - b)(2x3 - c)dx 2 + [,ux3 (X3 - c) + p(q -1)«x2 - b)2 - (X3 - E/
2
)]dx3 ,
440 ChapterS

First, the condition U(a,b,c/2) = 0 implies

U\ =0.

Secondly, we have

From aU 2 we deduce the derivative !'(x3 ) and thus

ax3
C 3
(x 2 -b)2 2 X3 c (X 3 -"2) c3
u2 = lI(q-l)
r 2
(2x -c)+ IIX ( - - - ) - lI(q-l)
3 r 3 3 2 r 3
+ 11-,
r 12

expression to be simplified and where the constant c 3 /12 is determined from the condition
U(a,b,c/2) = O.

Finally, we have
C 2
U3 =J.lq(x2 -b)(x 3 - - ) +g(x 2 )·
2
From aU3 we deduce the derivative g'(x 2 ), and thus
ax2
C 2 c 2b
II(X
r 3 --)
2
+
X2 r
11- ,
4

2
expression to be simplified and where the constant J.l c b is determined from the
4
condition U(a,b,c/2) = o.

(v) The ratio sought is

J- 1 :: I-tre
::1-J.l(2q-l)(x 2 -b)(2x3 -c).

(vi) Fibers parallel to x 2-axis get longer if e2 2 is positive, that is:

J.l(q-l)(x2 -b)(2x3 -c»O.

Since
J.l > 0, q -1 < 0, x 2 - b ~ 0 ,
the required domain is
{(X\,X2,X3):~~X3 ~c, x 2 :;o!:b}.
2
 Fundamental Laws; Principle of Virtual Work 441

Exercise 13.

Given an orthogonal Cartesian system OXIX 2X3 the state of stress throughout a
body is

where the positive constant k has the required dimensions.

(i) Determine the equilibrium conditions.
(ii) Find at the point (2,2,4) the stress vector on the plane of equation 2xI - x 2 + 2X3 = 10
and that on the ellipsoid of equation 2X2 + y2 + z2 = 28.
(iii) Calculate the maximum shear stress value at (2,2,4).

Answer. (i) The well known equilibrium equations

1; +O'ik,k =0

are reduced to
J.. = 0,

which implies that the body is in equilibrium if there is no body force.

(ii) The unit normal to the given plane is immediately

A 2 I 2
n = -e l --e2 +-e3'
3 3 3
Thus the stress vector on the plane at (2,2,4) is

t(n)= [ 00 02k·-1/3
0 2k] [2/3]
2k 2k 0 2/3
442
=-ke l +-ke 2 +-ke3 ·
333
Since the equation of the plane tangent to the ellipsoid at (2,2,4) is
8(x - 2) + 4(y - 2) + 8(z - 4) = 0,
then the unit normal to the ellipsoid at this point is
A 2 I 2
n = -e l +-e2 +-e 3 .
3 3 3
Thus the stress vector on the ellipsoid at (2,2,4) is
442 Chapter 5

o
o
2k
~~l·[~j:l
0 2/3

(iii) From the following characteristic equation

-a 0 2k
o -a 2k =0,
2k 2k -a

we deduce the principal stress values:

0'1 = 2fik, an = 0, am = -2fik

and the maximum shear stress is

Exercise 14.

Given a plane stress tensor defined by

4 O-JIO]
0'=
r-JIO 00 0I ,
0

find the principal stresses, the principal stress directions, and the maximum shear stress.

Answer. The characteristic equation is

0'(0'2 - 50' - 6) =0
and the principal stress values are 0'1 = 6, an = 0 and am = -1 .

The eigenvectors are obtained from the well known equation (a ij - m)ij ) n j =0 .
For an = 0, from

we deduce that
 Fundamental Laws; Principle of Virtual Work 443

Thus the principal stress direction is given by the unit eigenvector "n =e 2•

For a] = 6, from
-6n2 = 0,
we deduce that

Thus the principal stress direction is given by the unit eigenvector ( b, 0, V35
v'35
flO). This
vector is also written:

such that

For am = -1, we have

Thus the principal stress direction is given by the unit eigenvector ( ~,O,- II). This
v'14 V"7
vector, of course orthogonal to "1 and "n, is also written:
such that

Finally, the maximum shear stress, whose value is

a] -am 7
2 2
acts in the plane which bisects the right angle between the directions of the maximum and
minimum principal stresses.
444 Chapter 5

Exercise 15.
With respect to an orthonormal basis (e l ,e2 ,e3 ) of the Cartesian frame of
reference, we consider the following velocity field of particles between two horizontal
planes of equations X3 = 0 and X3 = h :

v = ccosJr(~
2 h 2
_~) e-"' at lh 2
'

where the dimensions of the constants c and a are respectively [LT- 1 ] and [L2 T- 1 ] .

The (mass) density of the fluid is the constant p of dimensions [ML-3 ] and the force of
gravity per unit volume is

The stress tensor is explicitly

o
x 1 2 I 2
BsinJr(2--)e-" al h
h 2 '
. x 1 _ 2 Ilh'
o CSlllJr(2--)e " a A(x3 - h)
h 2

where the dimension of the constant A is [ML-2 T -2] and those of the constants Band C
are [ML- 1r 2 ].

(i) Find the values of A, B and C from the linear momentum and moment of momentum
principles.
(ii) At t =0 calculate the stress vectors on the horizontal planes of respective equations
X3 = 0 and X3 =h .
Answer. (i) We know that the linear momentum principle leads to the following motion
equations
pyi = P + Y' kcy ik ,
which are simply in this context:
PYi =;; + CYik,k ,
where the various 1; are the components of the body forces per unit volume; namely:

13 =-pg,
of course the dimension of pg is [ML-2 T-2 ].

The only nonzero component of the acceleration r is

 Fundamental Laws; Principle of Virtual Work 445

Thus the motion equations are

0=0,
ac x I
- p7r 2 -COS7r(--.l--)e- Ir
2
at
/h2 7r X I 2 /h 2
= 0+ B-COS7r(--.l--)e- Ir at
h2 h 2 h h 2 '
O=-pg+A,
from which we deduce the following values
ac
A = pg, B = -7r p- .
h
We know that the moment of momentum principle implies that (]' ij = (]' ji ' that is:
. --.l
B Slll7r( x _ _1 )e- Ir 2at/h2 = C Slll7r(
. --.l x __I )e-Ir 2 at/h 2
h 2 h 2
and thus, we obtain:
ac
C=-7rP-·
h

(ii) At t = 0 the stress tensor is written

o
ac. X3 I
-np-Slll7r(---)
h h 2
ac. X3 I
o -np-sllln(---) pg(X3- h)
h h 2

The stress vector on the plane of equation x3 =0 is explicitly

- ;gh
7rpac I h
7rp~Clhl'[ ~
- p gh -1
] = [-7rP:Clh],
P gh
that is:
ac
t = -7rp-e 2 + pgh e 3 •
h

The stress vector on the plane of equation X3 = h is explicitly

[ ::] =
t3
[~ ~ -7rpOaclhl'[~] =
0 -7rpaclh 0 1
pO
[_7r aclh] '
0
446 Chapter 5

that is:

Exercise 16.

We consider the stress tensor

[
a = 0
2 0
1
-7]
25 .
-7 25 24

(i) Decompose this tensor into its spherical and deviatoric parts.
(ii) Calculate the principal stresses of the deviator aD and deduce the principal stresses
of the stress tensor a .
(iii) Determine the principal directions of a from those of aD .

Answer. (i) Since the mean normal stress is

as = 27 = 9
3 '
the required decomposition is
o
900][-7
a= [ 0 9 0 + 0 -8 ~;].
o 0 9 -7 25 15

The ordered principal deviator stresses are

aiD = 31. 916705, alf = -7.070553, al~ = -24.846152

and the corresponding principal stresses are

a l = a/J + as = 40.916705, an = 1.929447, a III =-15.846152.

The principal directions of aD are given by the following unit vectors

-.150700] -.214107]
"I = [ .524732 ,
.965115]
"II = [ .261645 , "III = [ 0.810061 .
-.545856
.837822 .009727

These vectors, which make up a (direct) principal basis associated with aD, obviously
give the principal directions of the stress tensor a.
 Fundamental Laws; Principle of Virtual Work 447

Exercise 17.

The state of stress at a given point of a body is characterized by the stress tensor

0 6 8]
(1'= [ 6 0 0 .
800

(i) What is the deviator stress tensor?

(ii) Determine the principal stress values and principal stress directions.
Express the principal basis in function of the orthonormal basis (e 1,e2,e3 ) of the
frame of reference.
(iii) Calculate the maximum shear stress.
(iv) Sketch the three Mohr's circles.
(v) Find the stress vector, at the given point, on a surface element of the outward unit
'
normaI n=O.75e J6
j +O.25e 2 - - e 3 •
4
Express the normal and shear components respectively denoted by (1' and T .
Prove that the corresponding point ((1', T) belongs to the allowed domain.
(vi) Express the unit normal ;, with respect to the principal basis.
(vii) Draw the frame of principal axes with respect to the frame of reference.

Answer. (i) From (1's = 0 we deduce that

(ii) The principal stresses are obtained from the characteristic equation

(1' 6 8
6 -(1' 0 =0.
8 0 -(1'

Thus the ordered principal stresses are

(1'n=O, (1'm =-10.

For (1'1 = 10 the well know equations

are expressed as
448 ChapterS

and since
1 = n l2 + n 22 + nj2 = n l2 + -9 2
nl
16 2
+ - n l = 2nl
2
25 25
we deduce that a first principal direction is determined by
, J2 3J2 2J2
nl =- e l + --e 2 + --e j •
2 10 5

For O"n = 0 we have

and since
2 9 2 25 2
1 = n2 + - n2 = - n2
16 16
we deduce that another principal direction is determined by
,43
nil =-e 2 --e j •
5 5

For 0"1ll = -10 we have

10n i + 6n2 + Sn j = 0,
and since

we deduce that the third principal direction is determined by

, J2 3J2 2J2
n lll = - - e l + --e 2 + --e j ,
2 10 5

where we have chosen nl = - J2 so that (npnlI,nnJ has a direct orientation, that is:
2

J2 3J2 2J2
2 10 5
4 3
ni' (nIl /\ nIll ) = 0 =+1.
5 5
J2 3J2 2J2
2 10 5

Of course, the vectors of the principal basis are expressed as functions of those of the
reference basis through the following notation:
 Fundamental Laws; Principle of Virtual Work 449

-J2
2
o

(iii) The maximum shear is equal to 10.

(iv) In the stress plane let us represent the normal component of ten) as the abscissa
(along the ann -axis, simply called the a -axis) and the tangential component of ten) as the
ordinate (along the T -axis).
It is a well known case where all = 0 and a, = -alII.

The circle of center on the a -axis at a, + alii = 0 and of radius a,- alii = lOis labeled
2 2

The circle of center on the a -axis at all + alii = - 5 and of radius all - alii == 5 is
2 2
labeled C" the circle of center on the a -axis at a, + a" = 5 and of radius a, - a" = 5
2 2
is labeled C3 .

(v) The stress vector on the plane of unit normal n is explicitly

and thus
(n) 3 Ir 9
t =(--2-v6)e, +-e 2 +6e3 •
2 2

The normal component of the stress vector is

a=(a )=t(") n=tn

M . I I
=~-3.J6"'-5098
4 ..

From

and thus
T '" 6.466,
450 Chapter 5

we deduce that the corresponding point ((Y, r) really belongs to the allowed domain, that
is, inside C2 and outside C] and C3 •

(vi) The vectors of the basis (e i) are expressed as functions of those of the principal basis
as follows:

[::1=[3~ ; ~kl[;':l'
e3
2.[i
-5-
3 2.[i
5 -5-
,
nm

We know that the components of the unit normal with respect to the principal basis are
obtained from the transposed matrix of the previous one, namely:

9.[i J3
20-5
1 3../6
5+20
3.[i JU.
-10-10

and thus these components are approximately:

0.289986]
[ 0.567423 .
- 0.770674

(vii)

Fig. 49
 Fundamental Laws; Principle of Virtual Work 451

Exercise 18.

The state of stress throughout a spherical body of radius R is given by the stress
tensor expressed in spherical coordinates (r,B,¢;) as

2r(3R - 2r)cosB r(3R - 2r)sinB

U [
= a r(3R - 2r) sinB 2r( 6R - Sr) cosB o 1,
o 0 2r(6R - Sr)cosB

where the dimension of the constant a is [ML-3 T- 2 ].

The body forces are neglected.
(i) Prove the body is in equilibrium.
(ii) Calculate the stress vector at any point (R,B,¢;) of the body surface.
(iii) Express the direct orthonormal principal basis at this arbitrary point.
(iv) Find the maximum shear stress on the body surface.
(v) Given R = 1, draw the Mohr's circles if B = 17/4 and B = 1<'

(vi) Determine the deviator stress tensor u D on the body surface, the principal stress
values of u D , and the corresponding principal directions.
(vii) By calculating the invariants of this deviator stress tensor find again the characteristic
equation of u D .

Answer. (i) In spherical coordinates the equilibrium equations are

1 1 1
8 rUrr + -8 oU rO + -.-8 ¢u r¢ + -(2u rr - u eo - u # + u rO cot B) = 0,
r rsmB r
1 1 1
8 rU Or + -8 oU00 + - . - 8 ¢Uo¢ + -(3u rO + (u 00 - U# )cotB) = 0,
r rsmB r
1 1 1
8 rU ¢r + -8 oU ¢O + - . - 8 ¢U # + -(3u r¢ + 2u o¢ cote) = 0,
r rsmB r
smce
/r =/e =/¢ =0.

The equilibrium equations are actually verified because we respectively have:

a(2(3R - 4r)cosB + (3R - 2r)cosB + 4(3R - 2r)cosB - 2(6R - Sr)cosB
- 2(6R - Sr)cosB + (3R - 2r)cosB) = 0,
a«3R - 4r) sinB - 2(6R - Sr)sinB + 3(3R - 2r)sinB + 2(6R - Sr)cosBcotB
- 2(6R - Sr)cosBcote) = 0,
0= o.
452 Chapter 5

(ii) On the surface (r = R) the outward unit normal is n= 1 r and the stress vector is
explicitly:

sinB
2cosB
o
and thus

(iii) The characteristic equation

2aR 2cosB - (J' aR 2sinB o

aR 2sinB 2aR 2cosB - (J' o = 0,
o o
that is
(2aR 2cosB - (J')«2aR 2cosB - (J')2 - (aR 2sinB)2) = 0
leads to the following ordered principal stress values
(J'l = aR2(2cosB + sinB), (J'n =2aR 2cosB, (J'm =aR 2(2cosB-sinB).

The principal direction associated with (J'" determined by (n[, nf, nt), is obtained from
the following equations
- aR 2 sinB n; + aR 2 sinB n~ = 0,
aR 2 sinBn; -aR2sinBn~ =0,
- aR 2 sinB nt = 0,
whose solution is
nt = 0
and thus the corresponding unit vector of the principal basis is

n = fi lr + fi IIJ.
l
2 2

The principal direction associated with (J'", determined by (n;l'nfr,ntr) , is obtained from
the following equations
0= 0,
whose solution is
 Fundamental Laws; Principle of Virtual Work 453

and thus the corresponding unit vector of the principal basis is

nIl =1(>.
The principal direction associated with O"IIl' determined by (n;n,nfn,n~I)' is obtained from
the following equations
aR2 sinO n~I + aR2 sinO n~I = 0,
aR2 sinO n~I + aR2 sinO n~I = 0,
aR2 sinO n~I = 0,

whose solution is

and thus the corresponding unit vector of the principal basis, such that this basis is direct,
IS

(iv) The maximum shear stress value is equal to aR 2 sinO.

(v) For R = 1 and 0 = 7r/4, the principal stress values are

O"m
J2
=-a.
2

For R = 1 and 0 = 7r, the principal stress values are

The ensuing Mohr's circles are schematized as follows:

-2Q

Fig. 50
454 ChapterS

(vi) Since the mean normal stress exerted on the body surface is
I
US = -tru = 2aR 2 cosO,
3
the deviator stress tensor is explicitly

u D "U-U' "lDR'~ine °
°
The connection between the principal stress values of u D and u leads to the following
principal stress values u D :
UID = aR2 (2cosO + sinO) - 2aR 2cosO = aR 2sinO,
ug =2aR 2cosO-2aR 2cosO=0,

UI~ = aR2 (2cosO - sinO) - 2aR2 cosO = -aR2 sinO.

These values are really the eigenvalues of u D, whose characteristic equation is

u(u 2 _a 2R 4 sin 20) = 0.
Of course, the principal directions of u D are those of the stress tensor u.

(vii) The invariants of u Dare

II =tru O =0,
100 DO I DO DO 24'2
12 =-(uiiu jj -uij u j , ) =-(-OI2U21 -U 21 ( 12 )=-a R Sill 0
2 2
13 = detu = 0.
D

Thus the characteristic equation of u D is actually

u 3 _ (a 2 R 4 sin 2 0) u = 0.
CHAPTER 6

LINEAR ELASTICITY

We know that the principles of conservation of mass, conservation of linear

momentum, conservation of moment of momentum, and of conservation of energy hold
for every continuum, whatever material we consider in this study. However, the notions
introduced until now, notably these principles, do not let us determine the behavior of a
specific material.
On the one hand, experience shows that the responses of different materials under the
same loading conditions are extremely varied; on the other hand, it shows that the
behavior of a given material changes greatly under various loads.
Of course, the introduction of constitutive equations cannot allow us to describe all the
observed phenomena, but these supplementary equations are useful for the definition of
idealized continua.

1. ELASTICITY AND TESTS

In particular, the deformation of certain materials completely vanishes when the

loads are removed and it is said that the material has an 'elastic' behavior. The initial
'natural' state remains in the elastic material memory. In other words, an elastic material
'remembers' its sole unstressed state but 'has forgotten' all the intermediate states.
Thus, more precisely, we say:

D A material or a body is elastic if the stress tensor u only depends on the

deformation between the initial unstressed state and the final stressed state.

455

Y. R. Talpaert, Tensor Analysis and Continuum Mechanics

© Springer Science+Business Media Dordrecht 2002
456 Chapter 6

So, the elastic behavior is independent of the 'motion' between the two states mentioned,
only the strain effect is to be considered.
On the contrary, beyond a given level of loading, continua may permanently be deformed
and fracture can even occur.
To know the mechanical behavior of materials it is useful to test specimens of those whose
shapes are very simple as slender circular cylinders, slender right angled parallelepipeds
for instance.

, .... - ....
F F
.... .,

Fig. 51

The most frequent test consists in applying tensile forces axially to the material!. The
connection between the magnitude of the force and the elongation of the body is
illustrated in the following figure.

Ci~_ . . . _ . . . ________ ~_~_~ _______

L..-_ _ _ _ _ _ _ _ _ _ _ _ _ E
Fig. 52

Let us consider the initial length Zof the cylinder. The stress a = FIA (N m -2), where F is
the axial tensile force exerted at the extremities of the specimen and A the initial area of
the cross-section, is represented as a function of the axial strain E: = LiZ I Z, where LiZ is the
elongation. By choosing the previous ratios referring to the cross-section and length of the
test body we have obtained a representation that does not depend on specimen dimensions.

I In French a traction test is called: 'Essai de traction '.

 Linear elasticity 457

For values corresponding to the line oa the continuum has an elastic behavior and reverts
to its initial state as soon as the load is removed. For higher tensile forces leading to stress
values greater than an elastic limit a e there is a permanent deformation, which means that
,11 (and thus [;) is different from zero, whereas the load vanishes. Finally, for increasing
tensile forces the material can break.
The slope of the curve oa is a fundamental constant in linear elasticity called Young's
modulus E = a / [;. It characterizes each material such that F / A < a e and will later be
intensively used.
For example, the Young's modulus ofa steel specimen is approximately equal to 2xl0 5
(MPa), and by considering a cylinder of radius equal to 0.03 (m) which carries a load of
5 xl 0 5 (N) we obtain:

and thus
,11 "",0.883x1O-3 •
[

So the axial strain [;, that is, the (dimensionless) relative elongation ,1[/[ , is of the order
of 10-3 . Given this smallness, the infinitesimal strain theory is applicable to this steel
deformation.

Of course, the axial strain

1 F
6=--
EA
is all the more large since the force F is large.
When the material elongates it shrinks perpendicularly to the elongation direction. This
lateral contraction is proportional to the increase of the tensile force.
By considering conditions such that an initial circular cross-section of diameter d remains
circular after deformation, the relative change of diameter is then denoted by the following
lateral strain

6
, = -M
-.
d
The ratio between lateral and axial strains is another important constant characterizing the
material if the strains are small. It is called Poisson's ratio and is defined in this particular
case as follows:
6'
v=-_·
6

Experience reveals that this real number is always positive but smaller than 0.5. For steel
the value of v is of the order ofO.3 and in general between 0.2 and 0.3 for metals.
458 Chapter 6

In the following sections, unless otherwise specified, the thermal effects will not be taken
into account, isothermal and adiabatic processes will not be considered. In other words,
the temperature is constant and no heat variation is viewed.
The isothermal hypothesis is valid for very slow evolutions of materials which conduct
heat quickly or for equilibrium states in the cases of uniform and constant temperatures.
The adiabatic hypothesis fits vibrations or waves which have no time for heat exchanges.

After introducing linear elasticity in general, we will study the properties of classical
elasticity, namely those of homogeneous and isotropic materials. In brief, a homogeneous
continuum is such that its material properties do not differ at different neighborhoods and
an isotropic continuum is such that its material properties do not depend on direction.

2. THE GENERALIZED HOOKE'S LAW IN LINEAR ELASTICITY

2.1 THE GENERALIZED HOOKE'S LAW

We assume that the displacement gradients are 'sufficiently small' and

displacements too, so that the Lagrangian and Eulerian descriptions of vectors and tensors
can be considered as indistinguishable.
Thus at any point and every time the material behavior is known from u in function of 6,
by knowing that the 'natural' state is described as 'no strain-no stress'.

From the previous experience and ensuing comments we state that the stress tensor is
expressed as a function u of the infinitesimal strain tensor 6 such that
u(O)=O,
that is, explicitly:
urs(O) =0.

Let us assume these functions eyrs can be expanded in Taylor's series of components of
6, that is, explicitly:

urs = Ersy 6 JI.. + ...

- ErsIl
-
6
11
+ E rs12 6
21
+ E rs21 "12
co + ... ,

where the coefficients Ersil of dimension [ML-IT-2 ] are to be specified.

A priori these coefficients are functions of position vectors, but we will assume these are
constants; that is, the material is said to be homogeneous.
 Linear elasticity 459

In linear elasticity we assume that the various ars are linearly depending on 8ij' Thus
given 'sufficiently small' deformations, the constitutive equations for a linear elastic body
are the nine equations

r,s,i,j E {l,2,3}, (6-1)

which relate the stress tensor u and the infinitesimal strain tensor &.

The equivalent tensor relation

a=E:& (6-2)
is called the generalized Hooke's law.
It expresses the stress tensor as the specified double contraction.

D The tensor if of 81 constant components ErSij is called the elasticity tensor.

It is really a tensor of fourth order since the double contraction between if and the
infinitesimal strain tensor & leads to the stress tensor a .

The symmetry of both stress and infinitesimal strain tensors implies there are at most 36
independent components of if .
Indeed, the symmetry of & implies that

and thus

These 27 equalities reduce the number of independent elastic components to 54.

By considering the symmetry of u for a linear elastic body, it then follows that we have
18 relations of symmetry, namely:

and the number of independent elastic components is 36 at the most.

2.2 QUADRATIC FORMS AND STRAIN ENERGY FUNCTION

To make progress let us introduce an important positive-definite quadratic form l which

characterizes linear elasticity, namely:

I We recal\ that U can only vanish if all the strain components are zero. The quadratic fonn U is a convex
function (positive second derivatives of U), this is equivalent to stating that the quadratic fonn is positive-
defmite.
460 Chapter 6

1 -
U(E) = - (E: E: E) (6-3)
2
or
1
U(E) = - (E : u). (6-4)
2

Given an orthonormal basis it is explicitly written:

1 - ..
U(E) = - [; E rslJ [; .
2 sr jI

or (6-5)

PRI The positive definite quadratic form U is symmetric in pairs of indices of E and
the independent components of E are at most 21 in number.

Proof From the expression of the quadratic form U we deduce the differential

By changing the dummy indices this expression is written:

Thus by comparison we conclude that

£,ijkl = £,klij.

Since
£,ijkl = £, jikl = £, jllk = £,ijlk

we conclude that the various independent components of E are

£,1111, £,1112, £,1ll3, £,1122, £,1123, £,ll33,
£,1212, £,1213, £,1222, £,1223, £,1233,

£,1313. £,1322, £,1323, £,1333,

£,2222, £,2223, £,2233,

£,2323, £,2333,
£,3333.
 Linear elasticity 461

PR2 The partial derivative of the quadratic fonn U with respect to & is the stress tensor
a.

Proof We have
au a 1 - 1 - --
- = - ( - & : E :&)=-(E:& + &:E) = E:&
a& a& 2 2
=0'.

We can also prove this proposition in explicit fonn. Indeed, we have

dU = ~(Eklij
2
& d& + Eijkl& de)
M, M,

= Eijkl&kl d&y

=0' ijd&ij

and thus, by considering the various &ij as independent variables, we have actually
au .
--=0". (6-6)
a&ij

PR3 In the infinitesimal context the quadratic fonn U is the strain energy function.

Proof The energy equation in local form

du
p-= a:d-divq+r,
dt
which gives the rate of change of internal energy, becomes
du
p-=a:d,
dt
when the thermal effects are ignored.
This particular balance equation is explicitly written:

P il=aiJd ,
According to usage we denote this last relation by

P du = a' d&,
and we have obtained the following result:
ij au
a =p--'
a&ij
462 Chapter 6

The quadratic form is really the strain energy function U = pu.

We are now going to introduce the dual quadratic form U' as follows.
The quadratic form
1 - 1
U(c) =-(c: E :c) = -csra rs
2 2
is also written:
1 rs
-a c sr '
2

Since U is a positive-definite form its discriminant is nonzero and Eq. (6-1) can be
inverted to obtain the strain components C if in terms of the stress components a rs ,
namely:

[c=C:a] (6-7)

and U(c) is successively written:

_ 1 rs _ 1 rs - qp _ 1 . C- .
U(c) - -a c sr - -a C srpq a - -a. .a.
2 2 2

Thus if we introduce the following quadratic form

1 -
U'(a)=-a:C :a, (6-8)
2
we have obtained the duality equation, namely:

U(c) = U*(a). (6-9)

PR4 The partial derivative of the dual quadratic form U* with respect to a is the
infinitesimal strain tensor c .

Proof We have
au' a 1 - 1 - --
- = - ( - a : C :a)=-(C :a+a:C)=C:a
aa aa 2 2
=c.

Thus we have obtained in explicit form:

au'
aa if = c if ' (6-10)

which is the dual expression ofEq. (6-6).

 Linear elasticity 463

2.3 ISOTROPIC MATERIAL AND LAME COEFFICIENTS

We know that the number of elastic coefficients ltijkl is reduced to 21 when a

strain energy function exists.
The more a material has material symmetries, the less it has elastic coefficients in its
behavior constitutive equations. We are going to consider the simple case of isotropic
material.

D CJr A material is said to be isotropic if its mechanical properties are independent of

direction.
On the contrary, it is said to be anisotropic if its mechanical properties are
described with reference to direction.

2.3.1 Constitutive Equations

Since every isotropic material is equivalent in all directions, the corresponding

definite quadratic form (6-3) is a quadratic invariant of the infinitesimal strain tensor &.
By denoting

invariants of &, the quadratic form is necessarily a linear combination of second order
invariants 1? = E:iiE: jj and 12 . Thus we explicitly denote:

(6-11)

that is:

By considering double contractions, we equivalently denote:

A
U(&) = - (I: &)2 + Jl &: &
2
(6-11 ')
A
U(&) = - (tr&)2 + Jl & : & .
2

At this stage it is interesting to consider the sum of the spherical strain tensor &s and the
deviator strain tensor 8 D :
464 Chapter 6

We recall that every strain deviator tensor is associated with a shear deformation for which
the dilatation vanishes (trc D = 0); on the other hand, every spherical strain tensor
represents a uniform deformation. In addition we note that spaces of shear and uniform
deformations are supplementary.

To make progress we observe that

CUC jj = (buC s + Ci~)(bjjCS + cf) = 9(C S )2,

because

We also have

Thus the quadratic form (6-11) is expressed as

U(c) = -1 (9,1,(C S )2 + 6j.J(C S )2 + 2j.Jcffc~)

2
and, by letting
2
K=,1,+-j.J,
3
it is written:
(6-12)

Since it is a sum of squared terms the quadratic form is positive-definite iff

3K = 3,1, + 2j.J > 0, j.J>o.

D The constants A, and j.J are called the Lame coefficients.

Their dimensions are those of It, namely: [ML-IT-2 ].

PR5 Hooke's law's equation u = It : C is expressed as a fimction of Lame coefficients

as follows:
u=,1,(trc)I+2j.Je. (6-13)

au
Proof We recall that the components of the stress tensor u = - are
ac
 Linear elasticity 465

aU
.. = __
au
a£ij ,
and, given an orthononnal basis, they are expressed as
au = At5ij£kk + 2,u£ji' (6-14)

In conclusion, the constitutive equations for a linear isotropic elastic material are
respectively
all =A(£ll +£22 +£33)+2,u£1l '
a22 = A(£" + £22 + £33) + 2,u£22 ,
a 33 =A(£11 +£22 +£33)+2,u£33 ' (6-15)
aI2 = 2,u£12 '
a l3 = 2,u£l3 ,
a 23 = 2,u£23 .

Each Eq. (6-14) can be inverted to express the strain components in function of the stress
components as follows:
1
£u =-(aij -At5ij£kk)'
2,u
Since the addition of first three Eqs. (6-15) is
a kk = 3A£kk + 2,u£kk = (3,1 + 2,u)£kk' (6-16)
we deduce the following strain components
1 A
£ .. = -(a - t5a kk) (6-17)
U 2,u U 3,1 + 2,u U

and thus
1 A
Ei=-(a- (tru) 1). (6-18)
2,u 3A + 2,u

PR6 The stress and strain tensors are spherical together; that is:

£ff = O.
Proof The comparison between

dU = a ij dEi ij = 3a s dEi s + aff dEiff

and the differential of the expression (6-12) of U, that is,

dU = 9K£s dEi s + 2,uEiff d&ff,
466 Chapter 6

implies the following relationship between spherical tensors as well as deviator tensors:
(6-19)
D D
aij =2peij. (6-20)

This proves the proposition.

Remark 1. For a uniform compressive stress as < 0 there is a uniform compressive

strain (or uniform contraction) e S < 0 since K is positive.

Remark 2. We emphasize that Eq. (6-16) gives the sum of the first three Eqs. (6-15), that
is:
tra = (3A.+ 2p)ekk = 3Ktre. (6-21)

This result also follows from as = 3K e S since we successively have

tra = tra S = 3a s = 9K e S = 3Ktre.

2.3.2 Young's Modulus and Poisson's Ratio

We recall that a stress is said to be uniaxial if there is one and only one nonzero
normal stress component.
The uniaxial stress type approximates the stress in the cylindrical bar considered in the
introductory test specimen.
For instance, let us choose the principal stress in axial direction.
Thus Eq. (6-17) gives
1 A A+p
ell = 2p (all - 3A. + 2p all) = p(3A. + 2p) all' (6-22)

A
-----a (6-23)
2p(3A. + 2p) II'

all the other strain components are zero.

We note that
(6-24)

Two important constants are to be introduced.

 Linear elasticity 467

D CF The constant a" / £11 of dimension [ML-'T-2 ] is called the Young's modulus.
It is denoted by

(6-25)

D The real number - £22 =- £33 is called Poisson's ratio.

£11 £11

It is denoted by

v=--- (6-26)
2(1 + ji)

Of course, we obtain a relation between E, ji, and v by eliminating 1 between Eqs. (6-25)
and (6-26), namely:
E
ji= . (6-27)
2(1 + v)

Since

the constitutive equation (6-18) is written:

I+v v
£ =- -a - - (tra)] (6-28)
E E '
that is, explicitly:
(6-29)

Of course, we can obtain 1 in function of v and E by eliminating ji between Eqs. (6-25)

and (6-26), namely:
vE
1=----- (6-30)
(1- 2v)(! + v)

Thus Eq. (6-13) is written:

E vE
a = -- e + (tre)] , (6-31 )
I+v (l-2v)(l+v)

that is explicitly:

(6-32)
468 Chapter 6

We also have
A E
2(A+.u)=-=---- (6-33)
v (1-2v)(J+v)
and
1 2 (l-v)E
'" + .u = --'-------'--- (6-34)
(I - 2v)(I + v)

Now let us make explicit the constitutive equations in terms of v and E for a linear
isotropic elastic material.
So Eq. (6-29) leads to
1
cll =-(O'll -v(0'22 +0'33»'
E
1
c22 = -(an - v(0'33 + all»,
E
1
c33 =-(0'33 -V(O'II +0'22», (6-35)
E
1+ V
cl2 =£0'12'

1+ V
c[3 = --0'[3 ,
E
I +V
c23 = --0'23 '
E

and Eq. (6-32) leads to

E vE
+
all = - - c l l (cll +c22 +c33),
l+v (I-2v)(J+v)
E vE
0'22 = --c22 + (cll + c22 + C33)'
l+v (l-2v)(l+v)
E vE
0'33 = --c33 + (cll + c22 + c33)' (6-36)
l+v (1-2v)(l+v)
E
0'12 = --c12 '
l+v
E
0'[3 = --c[3'
I +v
E
0'23 = --c23 .
l+v
 Linear Elasticity 469

2.3.3 Bulk Modulus

Another stress state corresponding to a uniform (normal) tension is defined by the

following stress tensor
u=uI.
It is sometimes called a hydrostatic stress state.
Eq. (6-16) becomes
3u = (3,1, + 2J1.) trs.
Thus we have
2 u
K=A+-J1.=- . (6-37)
3 tre

D The ratio K of the hydrostatic stress u to the dilatation IS called the bulk
modulus l .

Of course, given a positive value of u, K is all the more great as the dilatation is small.

Let us express K otherwise

From (6-25) we know that
A+J1.
3K =3A+2J1. = - - E
J1.
A
=(-+l)E. (6-38)
J1.

Let us express AI J1. as a function of v .

Since the Lame parameter J1. is such that
E
-=2J1.
l+v
(6-30) may be written:

,1,= 2J1.v .
1-2v

Thus by replacing AI J1. with 2v/(1- 2v) in Eq. (6-38) we obtain:

K= E (6-39)
3(1- 2v)

1 Called in French: 'Module de rigidite ala dilatation·.

470 Chapter 6

Since K and p are positive we necessarily have

1
E>O, -l<v<- .
2

Remark 1. The previous conditions can be found from the dual quadratic form written as
follows [by recalling (6-29) too]:
• 1 1 I+v v
U =-G"ij£ji =-(--G"ijG"ji --G"""G"jj).
2 2 E E
Indeed, since

and since

we deduce:

U • -_~(31-2v( G" S)2 + I+v G"G"

D D)...
2 E E Ij ]1

Since it is a sum of squared terms, this positive definite form implies that
1
E>O, -l<v<- .
2

Remark 2. We note that v tends to 1/2 when K/ E tends to infinity. It is the case where
the material is incompressible (trs ~ 0) because
G" 3(l-2v)G"
trs =-= .
K E

Remark 3. Experience reveals that v is always positive.

Summary of conversion of constants:

vE 2pv p(E-2p) 3Kv

,1,=) - - - -
(1 + v)(l- 2v) , 1- 2v' 3p- E ' 1 + v '

E 3K(l-2v)
W=)
2(1+v) 2(l+v)

E =) '-----'----'--
p(3A + 2p)
, 2p(l + v), 3K(I- 2v),
A+P
 Linear Elasticity 471

A E
v =) , - -1 ,
2,u
2(A + ,u)

2 E 2,u(1 + v) ,uE
K=) A+-,u
3 ' 3(1- 2v) 3(1- 2v) 3(3,u - E)

Examples. Values of E and v for elastic materials (GPa = 10 3 MPa) :

Aluminum: E = 69 GPa, v = 0.32 ,

Brass: E=lOOGPa, v=0.33,
Copper: E = 118GPa, v = 0.33,
Plexiglas: E = 2.9 GPa, v = 0.4 ,
Rubber: E = 2 MPa, v = 0.5,
Stainless steel: E = 200 GPa, v = 0.30,
Titanium: E = 106 GPa, v = 0.34 .

2.3.4 Shear Modulus

A special stress state is given by a simple shear stress tensor, corresponding to only
one pair of nonzero shear stresses, for example a l2 = a 21 = T.
In this case, since

we have

D The ratio of the shear stress T to the shear angle 2cl2 for a pair of orthogonal el-

and e 2 -directions is called the shear modulus l .

It is denoted by

G=_T_ . (6-40)
2C l2

In fact, it is the positive Lame coefficient ,u .

The shear angle is the smaller the greater is G.

I Called in French: 'Module de rigidite au glissement'.

472 Chapter 6

2.3.5 Hooke's Law's Expression in a General Coordinate System

Given a principal basis (Ei) the generalized Hooke's law (6-14) is expressed as

CJ 1 = 1(e1 + e2 + e3) + 2,ue 1'

CJ 2 =1(e1 +e2 +e3)+2,ue 2'
CJ3 = 1(el + e2 + e3) + 2,u e 3'
where the various e, are the principal strains.

For the corresponding coordinate system exi), these equations lead to

CJ 1 (dX 1 )2 + CJ 2 (dX 2 )2 + CJ3(dX 3)2

= 1 It[(dX')2 + (dX 2/ + (dX 3 )2] + 2,u[ e, (dX')2 + e2 (dX 2)2 + e3(dX 3 )2],

where It is the first invariant tre.

We point out two well known quadratic forms respectively associated with CJ and E; they
are invariable through any change of coordinates.

In a general coordinate system (rgi) the previous invariant relation has the following form:

or equivalently:

which implies the generalized Hooke's law in general coordinates:

(6-41 )

Hooke's law in cylindrical and spherical coordinates

The expression of a in cylindrical and spherical coordinates are immediately

obtained from the corresponding expression of
1
E = - (gradU + I gradU) .
2

Firstly, given cylindrical coordinates r,e,z, Exercise 44 ofCh. 1 leads to

 Linear Elasticity 473

aUr ~(aUr -U e ) aUr aUr aUe oU z

or r aB az ar or or

&=-
1 oUe ~(oUe +U ) oUe +-1 ~(oUr -U ) ~(oUe +U ) loU z
2 ar r oB r oz 2 r oB e r aB r r oB
oU z ~ au z oU z aUr oUe au z
or r oB az oz oz az

oUr ~(~ oUr _ Ue + aUe) ~(oUr + oU Z)

ar 2 r oB r or 2 oz or

~(~ aUr _ Ue + aUe) ~(oUe +U ) ~(aUe +.!. au Z)

2 r aB r or r oB r 2 az r oB
oU z
oz

and thus we have

oU
a rr = A tre + 2J1_r ,
ar
1 au
alJO =.11, tre+2J1-(_8 +U r ) ,
r oB
oU
au = A tre + 2J1_z ,
oz

are = J1(~ aUr _ Ue + oU e ),

r oB r or

a = lI(au r + aU Z )
rz roz or'

a = (oU e +~ aU Z )
ez J1 az r oB '
where

Secondly, given spherical coordinates r, B, ¢ , Exercise 46 of Ch. 1 leads to

474 Chapter 6

aU r ~(aUr -U ) !(_l_aU r -U )
ar r ao e r sinO a¢ ¢
1 aU e
6=- ~(aUe +U ) 1 1 aU e
- ( - - - U cotO)
2 ar r ao r r sinO a¢ ¢
au¢ ~ au¢ 1 1 au¢
-(----+U cot(}+U )
ar r ao r sinO a¢ e r

aU r aU e au¢
ar ar ar
1
+- ~(8Ur -U ) ~(aUe +U ) I au¢
2 r ao e r ao r r ao
1 1 aU e
!(_l_aUr -U ) - 1 1 au¢
( - - - U cot(}) -(----+U cot(}+U )
r sinO a¢ ¢ r sinO a¢ ¢ r sinO a¢ e r

aU r 2.(~ aU r + aUe _ Ue )
au + _au¢_ _ U¢ )
1 _ I_ _ _
_( r

ar 2 r ao ar r ar 2 rsine a¢
r
~(~ aU r + aU e _ Ue ) ~(aUe +U ) 1 1 aU e au¢
- ( - - + - - U coW)
2 r ao ar r r ao r 2r sine a¢ ao ¢
1 1 au
_( _ _ _ _ au¢
r + _ _ _)U¢ 1 laUe au¢ I 1 au¢
- ( - - + - - U cote) -(-.- - - + Uecot() +U r )
2 rsine a¢ 8r r 2r sine a¢ ao ¢ r sme a¢

and thus we have

where
aU r laUe U 1 au¢ Ue
tre =- + - - + 2-r + - - - - + -cotO.
ar r ao r rsinO a¢ r
 Linear Elasticity 475

2.3.6 Navier's Equations of Motion

Let us state the equations of motion of an elastic body in accordance with the
generalized Hooke's law.

We know that the motion equations are in general coordinates:

py' = f' + V JaY •

By recalling that the components of the strain tensor Ii are

lipq =~(VpUq +VqUp)'

2
then the components of the stress tensor a are written:
aiJ =).,Ng iJ + J-lgiPgjq(VpU q +VqU p ).

Thus Navier's equations of motion are

py' = r + ).,giJV /t + J-lV j(gi gjq (V pUq + V qU p»
P

=r + ).,giJV /divU) + J-lgiPV jV puj + J-lgjqV j V qUi

=r +)., Vi (divU) + J-lV'V jUi + J-lvqv qUi.

This result, namely:

(6-42)
is written in vector notation as follows:
pr = pO = f + (A + j1)grad(divU) + J-lL1U. (6-43)

- In orthogonal Cartesian coordinates Navier's equations of motion are written:

476 Chapter 6

- In cylindrical coordinates, by denoting U=Url r +Uele +Uzl z Navier's equations of

motion are

(6-45a)

(6-45b)

PY z
~
= f z + ("+J1 )a(aUr Ur laUe au
- -+-+---+--
Z )

az ar r r aB az
+J1(o2 UZ +~ au z +~ a 2u z + a2 u Z ).
(6-45c)
ar2 r ar r2 aB 2 az 2

- In spherical coordinates, by denoting U=Url r +Uele +U¢l¢ Navier's equations of

motion are

PY,
o a au, 1oU
=/, + (2+JL)-(-+--+-. ---+-U
I au~ 2 Uo
+-cotB)
r
ar ar r aB rsmB arp r r
1 a ( 2 au r) I . B-au, )+ -
a (sm 1 a2 u,
+JL [- - r - +- -- ---
r2 ar ar 2
r sinB aB aB r sinB arp2
2

2 I
--(U +--(U sinB)+--)]
a I au~
(6 - 46a)
r2 r sinB aB 0 sinB arp

I a aUr I au o I au¢ 2 Uo
PY =J, +(2+J1)--(-+--+----+-U +-cotB)
o 0 r aB Or r aB rsinB arp r' r
[ I a 2-au o I a . au o I a 2u o
+J1--(r )+---(smB-)+----
r2 ar ar r 2 sinB aB aB r 2sinB arp2
+~(aUr _~_ cosB au¢)]
r2 aB 2sin 2B sin 2 B 8rp

also written:
 Linear Elasticity 477

(6-46b)

1 8 aUr 1 8U e 1 au¢ 2 Ue
pr =/, +(A,+,u)---(-+--+----+-U +-cotB)
¢ ¢ rsinO 8rp 8r r ao rsinO 8rp r r r
1 a 2 au¢ 1 a . au¢ 1 8 2 U¢
+ ,u[z-8 (r -a-)+-2-'-0-a-0(smO--)+-2-.-2-0 8,;..2
aO
r r r r sm r sm Of'

2 aUr aU e U¢]
+-r2-s-in-0 (-arp-+cotB-arp---2-si-n-0) , (6 - 46c)

also written:
1 a aUr 1 8U e I au¢ 2 Ue
pr =/, + ( A , + , u ) - - - ( - + - - + - - - - + - U +-cotB)
¢ ¢ rsinO arp ar r ao rsinO 8rp r r r
I a 2 au¢ I a 1 a I a 2 u¢
+,u[ ~ ar (r 8r) + r 2 ao (sinO ao (U¢sinO») + r2 sin 20 arp2

+_2_(8U r +cotB aU e )].

r 2 sinO 8rp 8rp

3. EQUATIONS AND PRINCIPLES IN ELASTOSTATICS

Elastostatics consists in studying the equilibrium of any elastic material system

under certain conditions. It deals with the research of stress tensor and displacement fields
from a reference configuration.
More precisely, in the frame of classical elasticity of homogeneous and isotropic material
systems, we have 15 equations at our disposal, namely:
(i) Equilibrium equations:

au.} + J; = 0,
(iiJ Infinitesimal strain-displacement relation:
I
&u = "2(U i ,) +Uj,i)'
478 Chapter 6

(iii) Hooke's constitutive equations:

There are 15 unknown functions Ui ,(Jij , and e ij which must verify the previous equations
at any point of the system. Of course, unknowns may be eliminated between equations.
In addition, stresses and displacements must fulfill the required conditions on the
boundary.

3.1 NA VIER'S EQUATION; THE BELTRAMI EQUATION OF COMPATIBILITY

Let us consider Navier's equation taking the equilibrium equations into account.

PR7 A displacement field U may be the displacement field observed in an elastic system
ijfthe equilibrium equations and Navier's equation, that is:
P LlU + (2 + p) grad(divU) + f = 0, (6-47)
are verified at every point of the system.
This last equation is explicitly
pUi,jj +(2+p)U j ,ji + 1; =0 (6-48)

Proof By considering the infinitesimal strain-displacement relations, the Hooke's

constitutive equations become:
(Jij = 2Uk,koij + p(Ui,j +U j ) .
Substituting the following partial derivatives
(J ij,j = 2U k,ki + p(U i,jj + U j,u)
into the equilibrium equations we obtain after obvious manipulations of indices the
expected equations
pUi,jj + (2 + p)U j,ji + 1; = O.

Of course, they immediately follow from (6-42).

Navier's equation (6-47) is also written:

(2 + 2p) grad(divU) - p curl(curIU) + f = 0, (6-49)
since we recall the well-known formula
LlU = grad(divU) - curl(curlU).
 Linear Elasticity 479

PR8 With conservative body forces an irrotational field of displacements U may be the
displacement field observed in an elastostatic system iff the dilatation is expressed
as
. V
tre = dlVU = ---+c,
J. + 2J1
where V is the potential and c an arbitrary constant.

Proof The irrotational property ofU simplifies the Navier's equation (6-49) as follows:
f = -(J. + 2J1) grad(divU)
= -grad V ,
where the potential V is such that

divU =_V __ + c .
J. + 2J1

PR9 The Laplacian of dilatation is expressed in function of body forces as follows:

Lt(tre) = Lt(divU) = - _l-div f = - (1 + v)(1- 2v) div f , (6-50)

J. + 2J1 E(l- 2v)
and in the absence of body forces the dilatation is a harmonic function.

Proof It is obvious since

div f = -(J. + 2J1) div(grad(divU».

Of course, the previous Laplacian does not mean that U is the solution of an elastostatic
problem.

Let us introduce the Beltrami equations of compatibility.

PRIO Given a simply connected domain of a material system, a stress tensor field may be
the stress field observed in the elastostatic system iff the equilibrium equations and
the Beltrami equations, that is:
1 v
Lta ii + - - (tra) ,i + J, j + fl·' + --fkk ai' = 0, (6-51 )
"l+v ., . 'I-v"
are verified at every point of the domain.

Proof By substituting
480 Chapter 6

into the six independent compatibility equations

we obtain:
I+v v
E " ' + a rs ,"ii
--(a'i rs - air jS - a jS ir) - -(a"" rsOi'
, , E ," + a kk ,"iiOrs - akk jSOir - a kk irOjS) =
' , O.

By setting r = s these equations become

v
", + a rr ,",ii - air ' ji - a jr ,ir) - -
(a,ii rr
" + a kk ,"iiOrr - akk ,jrOir - a kk ,irOjr) = O.
1+-v( akk ,rrOii

By recalling that orr =3 and by taking account of the equilibrium equations air,r + 1; =0
we obtain:
v
(ai'rr + a rr i' + 1; j + f j i) - --(akk rrOii + akk Ii) = O. (6-52)
"' ,", , 1+ v '" '"
But Eq. (6-21) implies
(1 + v)(1- 2v)
L1(tra) = 3K L1(tr&) = - 3 K div f
E(1-v)
and since
3K=~
1-2v '
we deduce:
L1(tra) = _I+v divf.
I-v

In addition, the sum of the second and last terms ofEq. (6-52) is
1
- - (tra)".
I+v ,"

Thus Eq. (6-52) is really the expected Beltrami equation since it is written:
1 v 1+v
L1a ii + - - (tra) ii + h j + fj· i - - - ( - - - ) f k k 0ii = O.
"I+v " , 'I+v I-v ' "

Remark 1. In the case of conservative body forces (with potential V) the Beltrami
equations are
I v
L1a i, + - - (tra)" + 2V'i + - - V kk 0i' = O.
" 1+ v " "I - v' "

Remark 2. In the absence of body forces the Beltrami equations are simply
(l+v)L1aij + (tra),ij =0. (6-53)
 Linear Elasticity 481

To conclude this paragraph let us sum up the process for solving boundary problems of
elastostatics (with existence and uniqueness of solution).

(i) If the displacement field U is given on the boundary, the solution U of the problem has
to verify Navier's equations throughout the material system (obtained from Hooke's
equations and equilibrium equations) and such that U fulfills the boundary conditions.

n
(ii) If the stress vector t(iI) = IT • is known at every point of the boundary, the solution
IT of the problem has to verify the Beltrami equations throughout the material system
(obtained from the combination of Hooke's, compatibility, and equilibrium equations) and
n
such that IT satisfies t = IT • on the boundary.

(iii) If the displacement field U is given on a part of the boundary and the stress field
n
t = IT • is known over the remaining part, then the solution of the 15 equations at our
disposal gives the displacement and stress fields throughout the material system with the
respective boundary conditions.

3.2 PRINCIPLE OF SUPERPOSITION

The principle of superposition lets us obtain solutions from those already known,
since the equations of linear elasticity, including boundary conditions, are linear with
respect to the unknowns.

(i) Concerning the existence of solutions we state:

PRll If a first problem of 15 elastostatic equations, given body force components 1/'),
has a solution throughout the material system expressed as lTJI), UP), which
satifies the boundary conditions;
if for the same material system a second problem, given body forces components
/;(2) , has a solution throughout the system expressed as lTJ2) , uj2) , which satisfies
the boundary conditions;
then for the same material system the stress and displacement fields defined by
U = aU(I)
I I
+ bU(2)
I
, (a,b E R)

represent a solution of a problem with body forces such that

Ii = a/;(I) + bJ;(2)

and which satisfies a similar combination of two previous boundary conditions.

482 Chapter 6

(ii) Concerning the uniqueness of the solution we prove the following Work Theorem:

PR12 In an elastic material system in equilibrium under external forces the work done by
the external forces acting along a displacement field is twice the elastic energy (or
internal energy):
(6-54)

also written:

i j.Udll+ faD I.Uda= i a-rs8srdll.

D
(n)
D

Proof In linear elasticity we recall that at any point of a material domain the constitutive
equations lead to the quadratic form
• 1
U(8) = U (a-) = pu = 2a-rs 8".

We assume that the following integrals exist

fD U(8) dll = fD U*(a-) dll = fD !a- rs 8 sr dll·

Now let us consider

faD I(n). U da = fD a-ynp, da = fD (a-ijUJJ dll

= L(a-ij,JUi + a-yUi,) dll '

But we know that

and

since the double contraction between the symmetric tensor a- and the anti symmetric
tensor Ii vanishes.
Thus the work theorem is proved since we have obtained:
r I (n) .Uda + .10r JU, dll
JaD r
= .10 a-ij&lj dll· (6-55)

Remark. We note that the work theorem is valid within the context oflinear elasticity and
not only in classical elasticity.

(iii) Now let us prove the uniqueness of the elastostatic solution of a linear elastic
material system,
 Linear Elasticity 483

Let us assume there are two solutions, on the one hand , 0'(1)
y' U(1)
r for body force
components ;;(1) and, on the other hand, 0'~2), U~2) for body force components /;(2).

We are going to prove that these solutions cannot be different.

The principle of superposition implies that
_ (I) (2)
O'ij -O'ij -O'ij

and
u I
= U(I) - U(2)
I I

would also represent a solution for body forces of the problem such that
;; = ;;(1) _ /;(2) = O.

Thus the work theorem means that

faD
t(il).Uda=2r U(&)dJ1.
JD

The left member vanishes because both tfir) =0'~I)nrO'~2)nj and Ui=U~I)-U)2) satisfy
the boundary conditions.
Since
JD U(&) dJ1 = 0
and since U (&) is a non-negative and continuous function on D we deduce that U (&) = 0
at every point of D. Knowing that U(&) = t(&: If : &) is a positive-definite quadratic form
we conclude that the various 6ij = 6~1) - 6~2) vanish at every point of D.

From the equalities 6~1) = 6~2) the generalized Hooke's law O'rs = Ersi}6j ; leads to the
various equalities O'~I) = O'?) at every point of the material system and the displacements
are equal to within a rigid body displacement.

Reciprocity Theorem

PR13 If an elastic system is in equilibrium under the action of two external forces given
respective displacement fields U(I) and U(2), then the work done by the first
external force along U(2) is equal to the work done by the second external force
along u(l).

Proof In fact, by recalling Eq. (6-55), we must prove that

r 0'(1)6(2) dll = .br 0'(2)6(1) dll
.b lj lj ,... lj lj ,....
484 Chapter 6

In other words, we must prove that

This equality is obvious because the integrands are well known symmetric quadratic forms
(U is symmetric in pairs of indices of It), which proves the theorem.

The following principle is very useful, as is demonstrated by future exercises.

3.3 SAINT-VENANT'S PRINCIPLE

Two points are to be emphasized.

The first concerns the mathematical difficulties generated by the system of partial
differential equations with boundary conditions.
The second is a matter for experimental physics. Experiments do not permit the
verification of possible mathematical results with sufficient precision; for instance,
experimentally it is very difficult to exert faithful surface forces as accurately as they
should be, in view of assumed known forces on the boundary.
These aspects of the elastic theory which appears as 'too sharp a theory' are
counterbalanced by the so called Saint-Venant principle whose only experimental
character does not allow a mathematical proof.

This principle concerns locations inside the material system considered and which are
sufficiently remote from the area of application of two equivalent sets of surface forces;
that is, sets having the same dynamo

PR 14 If some set of surface forces acting on a part of the boundary of the material system
is replaced by another equivalent set of forces acting on the same part of the
boundary, while the boundary conditions on the complementary part of this
boundary are unchanged, then the fields of stresses, strains and displacements
practically remain unaltered within every area sufficiently remote from the region
of application of the forces.

Without any mathematical foundation this principle is in good agreement with numerous
experiments.

The accuracy of this principle is all the better that the area of application of surface forces
is small and that both inner and surface regions are far.
 Linear Elasticity 485

4. CLASSICAL PROBLEMS

4.1 PLANE PROBLEMS

4.1.1 Plane Stress Problems

Let us consider isothermal, isotropic, and elastic materials.

A material system is considered to be in a state of plane stress if at any point of the system
the stress tensor is a plane stress tensor such that the nonzero stress component can only
depend on the coordinates of the plane.

Thus we say:

D A material system is said to be in a state of plane stress parallel to the plane

x3 = 0 if at any point of the system the stress tensor is written:

(6-56)

where the nonzero components depend only on XI and x 2 ; namely, O'ij(X I ,X 2 ),

For example, a state of plane stress is obtained with forces uniformly applied over the
(lateral) thickness of the thin plate and parallel to the plane of the plate.
Let us assume that the body forces are conservative; that is, there is a potential V(X I ,X2 )
such that
(f3 = 0).

(i) First, we are going to prove that the plane stress problems do not consist in finding 4
unknown functions 0' ij' but only one function ¢(XI ,X2)' which is simpler!

PR15 There is a function ¢(X I ,X 2 ) called the Airy function such that

0'11 = ¢,22 + V, 0'12 = -¢,12' 0' 22 = ¢,II + V . (6-57)

Proof The equilibrium equations are simply

0'111 +0'122 --=0,

av
, 'ax I

while the third equation is automatically satisfied.

486 Chapter 6

The first equilibriwn equation leads to

(O"Il,1 + 0"12,2 - V,l )dx l /\ dx 2 = 0,

Thus there is a function f(xl ,X2) + c, where c is an additive constant, such that

0"11 dx 2 - V dx 2 - 0"!2 dx l = df,

that is:

Likewise, the second equilibriwn equation leads to the existence of a function

g(x l , X2) + k , where k is an additive constant, such that

-0"21dx2 +O"22dxl-Vdxl =dg,

that is:
0"22 =g,l +V.

From the equality

1,1 = g,2
we deduce the existence of a function qJ(x l , x2) such that

f = qJ,2 , g = qJ,I'

with
qJ,21 = 1,1 = g,2 = qJ,12 .

The proposition is proved because we have obtained the following results:

0"11 = 1,2 + V = qJ,22 +V ,
0"12 = -g,2 = -qJ,12 '

0"22 =g,l +V=qJ,Il +V.

We note that the Airy function qJ belongs to a set offunctions which differ from a (linear)
affine function arising from the presence of constants c and k.

(ii) Now let us carry on with the infinitesimal strain tensor

l+v v
6 = - - 0" - - (trO")/
E E
whose components are
l+v v
&ij =TO"Ij - EO"kk 6 ij (i,j = 1,2),
v
&33 =--O"kk'
E
 Linear Elasticity 487

It is explicitly denoted by

(6-58)

o
We note that all components of & are independent of x3 , but the strain tensor is not
'plane'.

We also denote

(»,22 - V(»,ll + (1- v)V - (1 + V)(»,12 o

1
&=- - (1 + V)(»,12 (»,11 - v(»,22 + (1- v)V o (6-59)
E
o o - v(Lt(» + 2V)

Given a thin plate let us consider the functions (» and V as only depending on XI and x 2 •
Thus amongst the six compatibility Eqs. (3-123) let us pay special attention to the only
equation without partial derivatives with respect to X3' namely:

that is

Therefore this compatibility equation is written:

LtLt(» + (1- v) LtV = o. (6-60)

In the absence of body forces the compatibility equation is reduced to

M(» =0,
that is,
488 Chapter 6

(6-61)

The solutions of such an equation are said to be biharmonic functions.

4.1.2 Plane Strain Problems

A material system is considered to be in a state of plane strain if the displacement

field is a plane field and its two nonzero components are functions of the only plane
coordinates.
Thus we say:
D A material system is said to be in a state of plane strain parallel to the plane
x3 = 0 if at every point ofthe system the strain tensor has the following expression

(6-62)

where the nonzero [; ij do not depend on X3 •

The state of plane strain is represented, for example, by an X3 -axis cylinder of arbitrary
cross-section and of isothermal, isotropic and elastic material, with forces acting
uniformly on the lateral surface and parallel to the (x], x 2 ) -plane. Likewise, the points of
the end faces may only move parallel to this plane.

In this situation the deformation field is the same in all planes perpendicular to the x3 -axis
and we say it is a plane deformation problem]. The corresponding displacement field is
such that

From
I au; au)
[; .. = - ( - + - )
I} 2 ax. ax '
J I

we deduce that the only nonzero strain components are

I In Exercise 5, by considering a dam, similar developments are introduced for a vertical (x 2 , x 3 ) -plane.
 Linear Elasticity 489

Since

the stress tensor has the following expression

(6-63)

where
0"1l =2(c ll + C22) + 2jlCIl ,
0"22 = 2(cll + C22) + 2jlC 22 •
In addition, from

we deduce:

that is
0"33 = V(O"ll + 0"22)'
We note that the components of 0" only depend on XI and X2 •

Let us assume that the body forces are conservative; that is, there is a potential
V(x l ,x2)such that

( /3 = 0).

Let us introduce the Airy function t/J( XI' X2) such that

O"ll =t/J,22 + V , 0"12 = -t/J,12 ' 0"22 =t/J,ll + V . (6-64)

The equilibrium equations

are written

a a2 t/J a a2 t/J
- ( - - - ) + - ( - 2 +V)--=o,
av
&1 &1&2 &2 &1 &2
while the third equation is automatically satisfied.
490 Chapter 6

If the cross-section is a simply connected domain the equilibrium equations are verified iff
the Airy function exists.

We note that the strain components deduced from the stress components are not
necessarily compatible. For this problem of compatibility, let us first consider Eqs. (6-35),
that is:

The following result,

confinns that

Therefore we denote the strain tensor by

V I
0"11 ---0"22 --0"12 0
I-v I-v

8
[eu
= 8~I
8 12

8 22
0]0 =I-V'
--
I
--0"12
V
0"22 - - - 0 " 0
I-v I-v 11
0 o E
0 o 0

and by letting
v - E
v=--, E=--2 '
I-v I-v
we have obtained the following result taking the generalized Hooke's law into account:

(6-65)
 Linear Elasticity 491

We conclude that
¢,22 - V¢,ll ~ (I - v)V -(1 + V)¢,12

E =[ - (I + V)¢,12 ¢,ll -V¢,22 +(I-V)V (6-66)

o o

To ensure the compatibility we must consider only one equation, which is not
automatically satisfied, namely:

8 2Ell 8 2E22 8 2EJ2

--+--=2--
m; m 2
J 8x J 8x 2
and by proceeding as for a plane stress problem, we obtain the compatibility equation
L1Ll¢ + (1- v)LlV = 0, (6-67)
where
LlLI¢ = ¢,J JJJ + 2¢,J J22 + ¢,2222 .

In the absence of body forces the solutions of the compatibility equations are biharmonic
functions.

4.2 CLASSICAL PROBLEMS IN ELASTOSTATICS

4.2.1 Uniaxial Stresses (or Simple Extension)

Equal and opposite forces uniformly act on the end faces of an elastic, isotropic
and homogeneous cylindrical body of arbitrary cross-section and with X3 -axis, whereas
the lateral surface is free from any surface force. We assume that no body force is to be
taken into account, for instance the gravity forces are neglected.
We know by experience that the state of stress is the same at every point of this body, this
state is independent of the length and thickness of the cylindrical body. Thus the stress
tensor is explicitly:

0'=
a
[ 000,
0 0]
o 0 0
where a is constant.

This type of tensor actually fulfills the following requirements.

The equilibrium equations are identically verified because
a if,} = a,} = 0 .
492 Chapter 6

Next, the lateral surface is really free from any surface force since the stress vector is
actually

Finally, Hooke's law leads to

1 a
ell =-(all -v(a22 +a 33 ))=-,
E E

Since the strain components are constant the compatibility equations are always verified.
There is a displacement field for these strains which is actually a solution of a uniaxial
stress problem:
(6-68)

with the frame of reference {o;e l ,e 2 ,e3 } such that e l is along the cylindrical body's axis.

This displacement field corresponds to an axial traction if a is positive and to an uniaxial

compression if a is negative.

If the area of the cross-section is equal to the constant a then the surface force exerted on
any end surface gives rise to a resultant force, through 0, of norm equal to aa, and the
stress tensor is written

(6-69)

The maximum normal stress equals F / a and the maximum shear stress is F /2a.

If the length of the undeformed cylindrical body is [ and the corresponding elongation is
LJ[ , we have:

and thus
 Linear Elasticity 493

f F
LJ.f = fli tt = - -
Ea

Likewise, if d represents a distance along a transverse line of a cross-section of the

undefonned cylindrical body, then the corresponding elongation M is such that
vF
M=C 22 d = - d - .
Ea

In the case of traction, F is positive and M is obviously negative; but in the case of
compression, F is negative and LJ.d is obviously positive.

Saint-Venant principle enables us to conclude that (6-67) and (6-68) define displacement
and stress fields every time the previous conditions about the forces (notably through the
centroid of the cross-section area considered) are satisfied and for regions which are
sufficiently far from the end faces.

4.2.2 Torsion of a Circular Cylindrical Body

A vertical cylindrical bar of circular cross-section of radius R and with x3 -axis has
its end faces .Eo and .EI in the planes of the respective equations X3 =0 and x3 =I . A
couple of moment M e 3 acts on the upper end face .EI and an 'equal and opposite' couple
of moment - M e3 is exerted on the lower face .Eo. The elastic, isotropic, and
homogeneous body is twisted in this way.

We assume that no body force is to be taken into account and the lateral surface is free
from any surface force.

~I

Fig. 53 X1
494 Chapter 6

From the material's symmetry it is expected that the motion of each cross-section owed to
the end moments is a rigid body rotation about the x3 -axis with respect to the basis, for
instance .Eo' and proportional to the distance from this basis.
So every plane section rotates with a very slightly different angle e from an adjoining
section.

Let us determine the corresponding stress tensor in the context oflinear elasticity.
The displacement vector of any particle located by

IS
u=ee 3 /\op
= e(xl e 2 -x2 e 1 )·

Since e is only a function of x3 we obtain the following nonzero strains

From Hooke's law we deduce the nonzero stress components:

0"l3 = 0"31 = -flx2 -

de ,
dx 3
0"23 = 0"32 = flx 1 -
de •
dx 3
To obtain a possible state of stress from this, let us introduce the equilibrium equations
0" ij,} = o.
The last equation, that is, 0"3},j = 0, is automatically satisfied.
The first two equations, namely:
d 2e
0"13,3 = - flX 2 dx 2 = 0,
3

lead to the constant variation in angular rotation:

de -_ e'0
dx 3
and thus this constant variation of angle per unit length (or twist per unit length) lets
generate a possible stress field.
The displacement field is written:
U = e(-x2 e 1 +Xl e2) = e~(-X2X3 e 1 +X 1 X3 e z )
 Linear Elasticity 495

and thus its components are

U1 = -Bo X 2 X3 , (6-70)

Likewise, the nonzero strain components are

1",
£13 = £31 = --uo X 2 , (6-71 )
2
and the nonzero stress components are
0"13 = 0"31 = -pBo X 2 ' (6-72)

On the lateral surface the outward unit normal is

and the stress vector on this surface is

that is
pB~
I=--(-xjx +xjx 2 )e3 =0
R 2

and thus the lateral surface is actually free from any surface force.

On the upper end face II' where the outward unit normal is ;, = e 3 , the stress vector is
expressed as follows:

that is,

Therefore the resultant of stresses acting on the upper end face is

R=-pB~ JI,
r x 2 dSe j +pB~ JI,r xjdSe 2 =0+0=0.
496 Chapter 6

If we denote the position vector of any particle of I, by r then the total moment of
stresses is
M = r r /\I dS ,
JI,
that is:

Of course, M has no component along e l and e 2 •

By introducing polar coordinates in the plane of I, we obtain:
R R4
M = ,lLl/~ 2Jr Lr2 rdre 3 = p(}~ Jr-:; e 3 ,

that is:

where I = Jr ~ is the moment of inertia of the circular cross-section area (of unit mass
2
density) about the x} -axis.

On the lower end face Io there is a counter-balancing couple of moment equal to

- M = -p(}~ I e3 •

By considering the Saint-Venant principle and given the previous couples such that
M = p(}~ I and thus ()~ = ~ , the stress tensor is then written:
pI
M
0 0 --x
1 2
M
U= 0 0 -xI (6-73)
1
M M
--x -x
1
0
1 2 I

4.2.3 Torsion of Cylindrical Shafts

Given a system of coordinates OXlx2X3 let us generalize the previous problem with
any cylindrical body of x} -axis. The end faces are Io and I, in the planes of respective
equations x3 = 0 and X3 =I .
 Linear Elasticity 497

We assume that no body force is to be taken into account and the lateral surface L is free
from any surface force.
The total moment of forces acting on each end face will be in the x3 -axis direction when
the stress component 13 is zero at any point of La and L,.

Each cross-section of the general cylindrical body rotates about the X3 -axis at a very
slightly different angle B from an adjoining section, but in addition an axial displacement
may exist for every cross-section other than circular.
Given the resultant of body forces per unit volume f = 0, let us consider the following
problem (where the constant Bb will later be specified):

I(L) = 0,

UI (LO)=U 2 (LO)=O, (6-74)

UI (L,) = -Bb IX2' U2 (L[) = Bb I XI'

To solve this problem let us prove the following.

PR16 Given a function X(x 1 , X2) of class C 2 on any cross-section D which is zero on the
boundary aD, the stress field of components
0'11 = 0'22 = 0'33 = 0'12 = 0'21 = 0,
0'13 = 0'31 = X,2 ' (6-75)
0'23 = 0'32 = - X,I ,

satisfies the equilibrium equations, the condition on the lateral surface I(L) = 0
and the conditions on the end faces t 3 (La) = 13 (L, ) = 0 .
In addition, the stresses exerted on La and L, are couples of respective moments
equal to -Me3 and Me3' where M = 21xdS.

Proof The equilibrium equations 0' ij,j = 0 are automatically verified SInce X IS
independent of X3 •

The condition

r~
o
I(L) = o
lX.2 -X.l
is satisfied since X is constant on aD.

We also have
498 Chapter 6

since the vectors (1' . (-e3) and (1' • e 3 have no X3 -component.

In addition, the resultant of stresses acting on Io or on II is zero. For instance,

r
J1'{
r X 2 dx l /\ dx 2 el -
t dS = J1'{' fX
1'{ ,
I dx l /\ dx 2 e2

= - LD Xdxl e l - LD X dx 2 e 2 = 0,

The total moments of stresses on II and Io are respectively M e 3 and - M e 3 with

M = 1(XI 0'23 - X2(1'n)dS = -1 (XI x,! + X2X,2) dxl /\ dx2

and since

we have

The proposition is proved since this result is reduced to 2 .bXdS .

Now we have to find a function x(X] , X2) such that there is a displacement field U
satisfying all the conditions (6-74) as well as (6-75).

By considering a priori the displacement field of components

we have

The constitutive equations for a linear isotropic elastic material imply:

0=0'33 = (A + 2,u)&33
and thus
(/),3 = 0,
We deduce that
 Linear Elasticity 499

The nonzero stresses and strains are related as follows:

%,2 =0"\3 =0"31 =2/-1e13 =/-IB~«/),I-X2)' (6-76)

- X,I = 0"23 = 0"32 = 2 /-Ie 23 = /-IB~ «/),2 + XI ) . (6-77)

By letting
_ 1
X=-B'X
/-I 0
and
If' = X- + -l( XI2 + X22) ,
2
Eqs. (6-76) and (6-77) are written:
1
1f'2 =--X2 +X2 =(/)1'
, /-IB~' ,
-1
-1f',1 = /-IB~ X,I - XI = (/),2 .

The first and second equilibrium equations are satisfied.

The third equilibrium equation
0"31,1 + 0"32,2 = %,21 - X,12 = /-IB~ «/),11 + (/),22) = 0
requires that If' and (/) are conjugate harmonic functions,

In addition the function i must verify the following equation

Lli +2 = O. (6-78)

In conclusion, the problem (6-74) amounts to the search for the stress field which satisfies
Eqs. (6-75), Eqs. (6-78) and which vanishes on the boundary aD.

The constant B~ is such that

Remark. If we express the stress vector on the lateral surface .I;:

o
1=0"';'=[ ~ o
X,2 -X,\

that is,
500 Chapter 6

t = (f1()~ (- n1x 2 + n2 x 1 ) + f1()~ (nl (/J,l + n2 (/J,J)e 3

= (f1()~ (-n 1x 2 + n 2 x 1 ) + f1()~ V (/J, n) e 3

and since it is zero, we deduce that the function (/J must fulfill the following condition

(6-79)

It is a classic boundary value problem for the harmonic function (/J, namely (/J must
satisfY L1 (/J = 0 with the condition (6-79), and which admits an exact solution.
 Linear Elasticity 501

EXERCISES

Exercise 1.

A gas moves in a vertical cylindrical tube fixed at its ends, and whose axis
equation is r = 0, given cylindrical coordinates r, e,
Z, and whose inside and outside
diameters are respectively equal to 0.2 m and 0.22 m. The normal stress (or pressure)
inside the pipe is equal to 7 xl 05 Pa and 10 5 Pa outside.
The constitutive equations of this pipe are those of an isothermal, isotropic, and elastic
solid whose Young's modulus is E = 1.8xlO s MPa and Poisson's ratio is v = 0.28.
Deformations owed to gravity and gas friction on the inner surface are neglected.
Determine the displacement field of type
Ur =f(r),

as well as the stress tensor 0" and the infinitesimal strain tensor & •

Answer. The infinitesimal strain tensor is expressed in cylindrical coordinates as follows:

df(r)
o o
dr
f(r)
&= o o
r

o o o
and thus
tr& = f'(r) + f(r) .
r
The corresponding stress tensor is written:

A(f' + f) + 2f.1f' o o
r
0"= 0 A(f'+ f) + 2f.1 f
r r
o o
In the absence of body forces the equilibrium Navier equations are

(A+ f.1) ~(f' + f) + f.1(f" + f' _L) = 0,

or r r r2
0=0,
0= o.
502 Chapter 6

Thus we have to consider

that is
(r2f')' - (rf)' =0 .
By introducing the following constant
C=r 2f'-rf,
we have
rf' - f _ ( f )' _ C
-r-2- - ---; - ~

and thus
f
-=---+c
1C
(c E R),
r 2 r2

that is
k
f(r)=cr+- ,
r
where c and k are constants to be determined from the limit conditions.

The stress tensor is written

o o
k
(J"= o 2(A+,u)c+2,u2" 0
r
o o 2Ac

Since the outward unit normal is expressed at the outer surface in cylindrical coordinates
as

the stress vector (for r = 0.11) is

tr]_ [1]- [2(...1, + ,u)c - 2,uk/O.0121]

[to - (J". 0 - 0 .
tz 0 0
 Linear Elasticity 503

Since the external pressure is equal to 1 x 105 Pa, we deduce a first equation with c and k:
k
2(A, + p)c - 2 p - - = -1 X 105 Pa.
0.0121
Since the inward unit normal is expressed at the inner surface in cylindrical coordinates as

the stress vector (for r = O. 1 ) is

tr]
(to = a·
(-1] =
0
(-2(A,+P)C+2 flk /O.Ol]
0 .
tz 0 0

Since the internal pressure is equal to 7 x 105 Pa, we deduce a second equation with c and
k:
k
- 2(A, + p)c + 2 p - = 7x 105 Pa.
0.01

The equations so obtained let us determine the constants c and k. Indeed, we have

7.26 105 Pam,

2P k =--x 2
21
579
2(A,+p)c=-xl0 5 Pa.
21
But, we know that
E
2p = - - = 1.40625 X 105 MPa,
l+v
E
2(A,+p) = =3.19602xl0 5MPa
(1- 2v)(1 + v)
and thus
k = 0.24584 X 10-6 m 2 , c = 8.6268 X 10-6 •

The stress tensor is

579 7.26
---- 0 0
21 21r2
579 7.26
a= 0 -+-- 0 10 5 Pa.
21 21r2

0 0 15.44
504 Chapter 6

The infinitesimal strain tensor is

k 0.2458
c-- 0 0 8.6268 - --2- 0 0
r2 r
k 0.2458 1O~6 .
6= 0 c+- 0 0 8.6268 + - - 2- 0
r2 r
0 0 0 0 0
0

The displacement field is

U=(8.6268r+ 0.24584)x10- 6 1r (m).

r

Exercise 2.

The pressure inside a submerged spherical body in equilibrium is equal to 10 5 Pa,

whilst the external pressure equals 5 MPa. The inside and outside diameters of the
spherical surface are respectively 3 (m) and 2R (m).
The constitutive equations are those of an isothermal, isotropic, and elastic brass solid
whose Young's modulus is E = 1X 10 5 MPa (100 GPa) and Poisson's ratio is v = 0.33.
The body forces are neglected.
(i) Calculate the Lame coefficients A, and J.l .
(ii) Determine the displacement field of type
Ur =f(r),

in spherical coordinates, as well as the stress tensor a and the strain tensor 6 .

(iii) In the case in which the outside diameter is equal to 3.2 m, find the principal stress
and characterize the Mohr's circles.
Calculate the principal stresses on the inner and outer surfaces respectively. Draw the
corresponding Mohr's circles.
(iv) Find the smallest value of the radius R of the outer surface for which the maximum
shear stress value is not higher than 900 x 10 5 Pa.

Answer. (i) The Lame coefficients are immediately

vE
A, = = 0.730 X 10 5 MPa (73.0 GPa),
(1- 2v)(1 + v)

J.l= E =0.376xl0 5 MPa (37.6GPa).

2(1 + v)
 Linear Elasticity 505

(iO The infinitesimal strain tensor is explicitly in spherical coordinates:

f'(r) o o
f(r)
8= 0 o
r
f(r)
o o
r
Since
tr8 = f'(r) + 2 f(r)
r
we deduce:

o o
U= 0 A(f' + 2 f) + 2Jl f o
r r
o o A(f' + 2 f) + 2Jl f
r r

In the absence of body forces, equilibrium Navier's equations are

0= 0,
0=0.

Thus we have to consider

(A + 2Jl)f" + 2A(f)' + 4 Jl (f' - f) = 0
r r r

(A + 2Jl)f" + 2(A + 2Jl)(f), =0

r

(aER),

that is,

Thus the required functionfis such that

k
f(r)=cr+ 2 '
r
where the constants c and k are to be determined from the limit conditions.
506 Chapter 6

The stress tensor is written:

(H + 2,u)c - 4 ,u; o o
r
U= o (3A + 2,u)c + 2 ,u; o
r
o o (H + 2,u)c + 2 ,u;
r

Since the outward unit normal is expressed at the outer surface in spherical coordinates as

then the stress vector is written:

Since the external pressure for r =R is equal to 5 MPa we deduce a first equation for
obtaining c and k:
(3A + 2,u)c - 4 ,u; = -50 (10 5 Pa).
R
Since the inward unit normal is expressed at the inner surface in spherical coordinates as

then the stress vector is written:

tr 1 [-1] [-(H
[ te = U· 0 =
+ 2,u)c +
0
4~l .
t~ 0 0

Since the internal pressure for r =1. 5 m is equal to 1x 10 5 Pa, we deduce a second
equation for obtaining c and k:

- (3A + 2,u)c + 4 ,u~ = 1 (l05 Pa).

1.5
 Linear Elasticity 507

The equations so obtained let us determine the constants c and k. Indeed, we have

4 Jlk _ 4 Jlk = -49 (l05 Pa)

1.5 3 R3
1 1
2(3A + 2Jl)c - 4Jlk(-3 + -3) = -51 (10 5 Pa).
R 1.5
From
3A + 2Jl = 294.2 (GPa) , 4Jl = 150.4 (GPa) ,
we deduce:
- 50R 3 + 3.375
c = -----:-----
2942000(R 3 -3.375) ,

k= -165.375 R3 m3
1504000(R 3 -3.375) ( ),

which determines the stress tensor 0' as well as the infinitesimal strain tensor

2k
c-- 0 0
r3
k
&= 0 c+- 0
r3
k
0 0 c+-
r3

and the displacement field

k
U=cr+ 2 .
r

(iii) For R = 1.6 m, the stress tensor is written:

_ 279.37 + 93~.5 o o
r
0'= o _ 279.37 _ 469;75 o
r
o o _ 279.37 _ 469;75
r

The principal stresses are equal to

939.5 5
0'[ =-279.37+-3- (10 Pa),
r
469.75 5
all = alII = -279.37 - - - 3- (10 Pa).
r
508 Chapter 6

Since two principal stresses are equal (an = am) there is only one Mohr's circle

(C 2 =C3 ) of radius equal to at - a l l , while the Mohr's circle C t is reduced to a point.

2

On the inner surface (r = 1.5 m), we obtain:

at = -1 (l05 Pa), all = am = -418.55 (10 5 Pa)
and on the outer surface (r = 1.6 m), we obtain:
at = -50 (10 5 Pa), an = am = -394.05 (10 5 Pa).

The possible shear and normal stresses are represented by points of plane (a, r) which are
located between the Mohr's circles determined from the previous values, the greatest
circle corresponding to the inner surface and the smallest circle corresponding to the outer
surface.

Fig. 54

(iv) Since we know that

4pk = 165.375R 3 (l05 Pa)

R3 -3.375
and
(3,1, + 2p)c = - 50R 3 + 3.375 (10 5 Pa),
R3 -3.375
the stress tensor a is written:

t 3 165.375R 3
3 (-50R +3.375 + 3 ) o o
R -3.375 r
1 3 82.687
o 3 ( 50R +3.375---
3 -) o
R -3.375 r
1 3 82.687
o o 3 ( 50R +3.375--1- )
R -3.375 r
 Linear Elasticity 509

At the radial distance r, the principal stresses are

0",= 3 1 (_50R3+3.375+165.3;5R3)XI05pa,
R -3.375 r
3
0"I1 = 0" III = 3 1 (-50R 3 +3.375- 82.68;5R ) Xl05 Pa.
R -3.375 r

Of course, the maximum shear stress value will be obtained for the largest Mohr's circle,
that is, for r = 1. 5 m, that is:
1 1 3 165.375R 3
2(0", - 0"I11 )r=15 = 2(R 3 _ 3.375) (2 1.5 3 )

and thus
3 165.375R 3 = 900 (105 Pa)
4 (R 3 - 3.375)1.5 3 '
that is:
R = 1.521 (m).

Exercise 3.

A steel disk of radius of 0.25 m, of thickness equal to 0.01 m and density equal to
p = 7.84 x 10 3 (10 3 kg/m 3 ) rotates about its vertical axis oz at a constant angular velocity
OJ = 5000 rev/min. This high speed lets neglect the deformation due to gravity with
respect to that owed to rotation.
The Young's modulus, Poisson's ratio, and density of the disk are respectively E = 2 X105
MPa, v = 0.26 and p = 7.84x 103kg/m 3.
The problem will be studied with respect to a frame fixed in the disk and not with respect
to a frame which is fixed in space.
Since the disk is not fixed, but by considering the problem with respect to a frame fixed in
the disk (with cylindrical coordinates r, e, z), we must take the transport and Coriolis
forces' into account. Of course, the Coriolis forces are zero because the relative velocities
vanish.
(i) Express the potential whose transport force derives.
(ii) The small thickness of the disk lets us refer to a plane stress problem. By considering
the polar coordinates rand e find the compatibility equation satisfied by the Airy
function¢ and give its general solution.
(iii) Express the stress components 0" rr ' 0" rB and 0" BB •

(iv) If the stresses are finite at the disk center and if there is no stress on the disk
boundary, calculate the radial displacement on the disk boundary.
I The transport forces are called: 'Forces d'entrainement' in French.
510 Chapter 6

Answer. (i) The transport force is the centrifugal force pOJ2 r 1r which is derived from a
potential V, that is:
2 8V 1 8V 8V
POJ rJ r =-gradV =--J ---J --J
8 r r r 8g B 8 z Z '

which implies:
8V 2
a,=-POJ r,

and thus the potential, such that V(O) = 0 , is

VCr) = - P OJ2 r 2 .
2

(ii) We recall that deformations are compatible if

LlLI ¢ + (1 - v) L1 V = 0 .
The symmetry of the body leads us to search for Airy functions independent of g. Thus
for functions such that ¢(r) and V (r) the last term of the following Laplacian

82 1 8 1 82
-+--+---
8r 2 r ar r2 8g 2
is not to be taken into account.
The compatibility equation is written:
d2 1 d d2 1 d d2 1 d
( dr 2 + -; dr)( dr 2 + -; dr) ¢ + (l - v )( dr 2 + -; dr) V = 0

d2 1 d d2 1 d
(-+--)[(-+--)¢ + (l-v)V] = O.
dr 2 r dr dr 2 r dr
By letting
d 2¢ 1 d¢
IjI = -+--+(l-v)V
dr 2 r dr
the compatibility equation is written:

d 2 1j1 + ~ dljl = 0
dr 2 r dr '
or simply:
~(r dljl) = 0,
dr dr
whose integration leads to
IjI = cllnr + c 2
where c 1 and C2 are integration constants.
 Linear Elasticity 511

By considering the expression of V(r) , the last result is written:

d 2rp 1 drp P 2 2
--+---(l-v)-m r =c j lnr+c 2 •
dr 2 r dr 2
To integrate this equation, we write it as follows:

~(r dr/J) = cjr lnr + c 2r + (1- v) P m2r 3 , (1)

dr dr 2
where rlnr is such that
d r2 d r2 r2
rlnr = - ( - ) lnr = -(-lnr--).
dr2 dr2 4
From this last result, the integration of Eq. (1), namely:
drp r2 r2 r2 P 2 4
r-=c (-lnr--)+c -+(1-v)-m r +c
dr j 2 4 2 2 8 3

drp c j c2 cj P 2 3 C3
-=-rlnr+(---)r+(1-v)-m r +-,
dr 2 2 4 8 r
leads to

(iii) Let us introduce the Airy stress function rp(r,B) such that

Given the symmetry, these components are simply

I drp cj c2 c j P 2 2 C3 P 2 2
a =--+V =-lnr+---+(1-v)-m r +---m r
rr r dr 2 2 4 8 r2 2 '
d 2rp cj c2 cj 3 2 2 C3 P 2 2
a = - - + V = -lnr +-+-+-(1-v)pm r --'---m r ,
eo dr 2 2 2 4 8 r2 2

(iv) At the disk boundary the stress vector is expressed in polar coordinates as
512 Chapter 6

Thus, the condition tr = 0 for r = 0.25 m leads to

2= p (3+v)ah2
2 8
= 7.84 (103)(3.26)(5000)2 (27r )2 (0.25)2 (N / m 2 )
8 60
~ 547.42xI0 5 (Pa).

We have
1 c2 3P 2 2 2
8 =-(u -vu )=-(I-v)---(I-v )m r
rr E rr /J() 2E 8E '

1 c2 P 2 2 2
8 = -(u -VU ) = -(I-v) - - ( l - v )m r .
/J() E /J(} rr 2E 8E
Since the components of the displacement field are
Ur = Ur(r) ,
we have
6 rr
au,
=-=--,
dU r 1 au o Ur
8/J(} = - - - + - = -
Ur
ar dr r ae r r

and thus
U = r8/J(} = 5....(1- v)r - L(l- v 2)m 2r 3
r 2E 8E
= L(3 + v)(l- v)m 2 r 3 _L(l- v 2 )m 2 r 3
8E 8E
=L(l-v)m 2 r 3 •
4E
For r = 0.25 m we obtain:

U (0.25)= 7.8410 3 0.74(5000)2(27r)2(0.25)3

r 4(2.101\) 60
~3IxIO-6(m).

Exercise 4.

A vertical flat rectangular plate with a 0.45 m base and 0.3 m height has a thickness
of 0.002 m and is made of copper. In this plane stress problem the Young's modulus and
Poisson's ratio are respectively E = 1.2 X 105 MPa and v = 0.34.
The two faces are free from any surface force as well as the two 0.45 m sides, but the two
other edges are under the action of stresses parallel to the plate.
 Linear Elasticity 513

Let { 0; e, ,e2 } be a frame of reference in the median plane of the plate such that 0 is in the
left and lower comer of the plate, ox, is along the 0.45 m base, and oX 2 is along the left
side of the plate.
The resultant of stress forces acting on the right boundary (x, = 0.45) is 2 e2 (N). We note
that normal stresses are to be exerted on the left boundary (x, = 0) to maintain the plate in
equilibrium.
(i) Express the compatibility condition by introducing the Airy function as a general
fourth degree polynomial.
(ii) Determine the coefficients of the polynomial from the stress conditions on the
boundaries corresponding to X 2 = 0, X 2 = 0.3 and x, = 0.45.
(iii) Calculate the components of the infinitesimal strain tensor &.

(iv) Find the component OJ 12 of the infinitesimal rotation tensor OJ such that OJl2 (0,0) = 0.
(v) Determine the components of the displacement U such that U(O,O) = O.

Answer. (i) The general fourth degree polynomial as an Airy function is

- 4 + b- X3 X +ex,x
-1.( x"x ) =ax, - 2 2 d- 3 - 4 f- 3
'f' 2 1 2 2 + x,x 2 +ex 2 + X1
+ g x~ X2 + h XIX; + I X~ + JXl2 + k x,x 2 + l x; .

Since the gravity forces are negligible the compatibility equation governing the Airy
function r(J is

which implies

and thus
24a + 4c + 24e = °
_ _ c
e =-a - - .
6

(ii) The stresses are expressed as follows:

all = a2
--2
r(J - 2
= 2ex d-
1 + 6 x,x 2 -\ a + -c) x 22 + 2 h-x, + 61X
2(- "0 1-
2+2 ,
Ox 2 6

a r(J
2 - 2 -
0'12 =----=-3bx1 -4ex,x 2 -3dx 2 -2gx, -2hx 2 -k,
- 2 _ - -

<7x,<7x2

= -a- 2rp = 12-aX12 + 6b-X,X 2 + 2-eX 22 + 6f-x, + 2-gX2 + 2-;J.

2
0'22
aX1

At the lower boundary (X2 = °)the stress vector is expressed as

514 Chapter 6

Since this boundary is free from any stress force we deduce that 0'12 and 0'22 are
identically zero for X 2 = 0; that is:

At the upper boundary (x 2 = 0.3) the stress vector is expressed as

Since this boundary is free from any stress force we deduce that 0'22 and 0'12 are
identically zero for X 2 = 0.3; that is:
2c(O.3)2 =0 => c=O,
- 3d (0.3)2 - 2h (0.3) = 0 => h = -0.45 d.
At the boundary such that Xl = 0.45 the stress vector is expressed as

The component all of this vector is the normal component and 0'12 is the tangential
component. By hypothesis all is zero for Xl = 0.45 , that is:
6d (0.45)x 2 + 2h (0.45) + 61 x 2 + 2[ = 0
=> 1= -0.45d, [ = -0.45 h = 0.2025 d.
Finally, since the thickness of the plate is equal to 0.002 m, the resultant of all stress
vectors at Xl = 0.45 (m) is expressed from 0'12 as follows:

2 = 0.002 S:3 0'12 dx 2 = 0.002 S:3 (-3dxi +0.9dx2 ) dx 2

= (0.002) 0.0135 d
and thus
d =74074.07, h =-33333.33, 1=-33333.33, [=14999.99.

(iii) Since
 Linear Elasticity 515

and
0"11 = 444444xI X2 - 66666. 6x I -199999. 8x 2 + 29999.97,
0"12 = -222222x; + 66666.6x 2,
0"22=0,
we obtain:

[;11 = i. 0"11 = (3.7037 XI x 2 - 0.555555x l -1. 666665x 1 + 0.25) xl 0. 6 ,

E

[;]2 =i.(1+V)0"12 = (-2.481479x; +0.744444x 2)xI0-6 ,

E
v ~
[;22 =--0"11 = (-1.259258xh +0.188889x l +0.566666x 1 -0.085)xI0 ,
E
v
[;33 =--0"11 = (-1.259258x IX2 +O.188889xl +0.566666x2 -0.085)xlO ~ .
E

(iv) Since

that is:
{O]2,1 = (3. 7037x I -1.666665) x 10-6 ,

{O12,2 = (-3.7037x2 +0.55555)xI0--6 ,

by integrating

(with (O]2(O,O) = 0) we obtain:

{On = (1. 85185x l2 -1. 666665xI -1. 85185x; + O. 555554x 2) X 10-6 •

(v) From

we deduce

UI = f [;11 dx l + tp(x 1 )
= (1. 85185x~ x 2 - 0.277778x I1 -1.666666x IX 2 + 0.25x l ) xlO--6 + tp(x 2).

By calculating aU I / ax 2 , which is also [;12 + (Oil' we obtain tp'(x 2), and finally:

UI = (1.85185x 12x 2 - 0.277778x I2 -1.666666x I X1 + 0.25xl-1.444444x; + 0.65x;) x 10--6 (m).

516 Chapter 6

Likewise we have

U2 = f &22 dx 2 + ljI(x l )

= (-0. 629629x l x; + 0.188889xlx2 + 0.283333xi - 0.085x 2) X 10-6 + ljI(x l ).

By calculating au 21axl , which is also &12 - OJ 12 ' we obtain ljI'(x l ), and finally:

U 2 =(-0. 629629x 1xi +0. 188889x 1x 2+0.28333x;- 0.085x 2- 0.617283x; +0.83333xI2) x 10-6
(m).

Exercise 5.

A homogeneous, isotropic, and elastic dam, built across a river, is delimited by the
horizontal bottom SI with equation X3 = -h, by a vertical plane S2 with equation x 2 = 0
in contact with the water, and by an inclined plane S3 with equation X3 = 2x 2 , as shown in
the following figure.
We are going to examine the equilibrium of the central part of the barrage between the
vertical sections S -d and Sd of planes of respective equations Xl = -d and Xl = d .

R IV E R
F

,
Fig. 55

The water exerts a hydrostatic force all over S 2 , defined by

such that the positive constantp denotes the pressure at any point (xl'0,-h) of the bottom,
more precisely, where F = - p e 2 ; of course F = 0 at any point (Xl' 0,0) of the top of the
dam.
 Linear Elasticity 517

The body forces are neglected. We assume that S3 is free from any force. We also assume
there is no tangential force on S -d and S d which have no normal displacement too.
This dam in equilibrium is considered as an example of the plane deformation problem for
which the displacement field is such that
U = U 2 (x 2 ,x 3 )e 2 + U3 (x 2 ,x3 )e 3 •

(i) Express the stresses as functions of second order derivatives of an Airy function
¢( X 2 ' x 3 ) and characterize ¢.
(ii) Since any third degree polynomial function is a biharmonic function determine such a
function ¢(X 2,X 3) which defines a stress field if fulfilling the boundary conditions
and express if.

(iii) Find the general field Ii of possible corresponding displacements.

Answer. (i) In this problem of plane strain parallel to the plane XI = 0 the infinitesimal
strain tensor is explicitly

where the components & lj (X2 , x 3 ) do not depend on XI .

We know that the stress tensor is expressed as

o

where

that is:
a 22 = A,(&22 + (33) + 2,li&22'
a 33 = A,(&22 + (33) + 2,li&3}"
From

we deduce:
518 Chapter 6

The stress components must verify the equilibrium equations (ji/,j + 1; = O. In the absence
of body forces these equations are simply
(j 22,2 + (j 23,3 = 0,
(j32,2 + (j33,3 = O.
From the first equation there is a function f(x 2 , x 3 ) such that

(j22 = 1,3 ,

Likewise, from the second equation there is a function g(X 2,X3) such that

The symmetry of (j implies that

1,2 = -(j23 = -(j32 = g,3'

There is an Airy function ¢( x 2 ' x 3 ) such that

g =¢,2 '

and which lets us express the general solution of the equilibrium equations as follows:

In addition we immediately deduce:

(jll = v( (j 22 + (j33) = v L\¢ .
In the absence of body forces, the Beltrami equations (6-53), which are
(1 + v)L\(jij + (tr(j),ij =0

lead to the following equation

LILI¢ = 0,
as expected from Eq. (6-67); in explicit form:

L\L\¢ = ¢,2222 + 2¢,2233 + ¢,3333 = 0 .

Thus the Airy function is biharmonic,

(ii) Since the stress field is obtained from second order derivatives of the Airy function ¢,
the following general polynomial is to be considered:
¢(X 2 ,X3) = Ax~ + Bxix3 + Cx 2 xi + Dx; + Exi + FX 2 X 3 +Gxi
and thus
 Linear Elasticity 519

a22 =1.33 =2Cx2 + 6Dx3 +2G,

a23 = -r/J,23 = -2Bx2 - 2Cx 3 - F,
a33 =1.22 =6Ax 2 + 2Bx3 +2E,
0';1 = V L1¢ = 2v[(3A + C)X2 + (B + 3D)X3 + E + G],

Now let us introduce the boundary conditions,

First, the stress vector on S2 is expressed as follows:

X3
(j,e z = p;;e 2 •

Since the equation of the plane of S2 is x 2 = ° this condition is simply written:

that is:
l:'
°
6Dx3 +2G
-2Cx3 -F
_2~'-F]n+;
° °
2Bx3 +2E
l
X3
6Dx .1 +2G = ph- , -2CX3 -F = °
and thus
C=G=F=O, D=l!..... ,
6h

Second, the stress vector on S3 of equation X3 = 2X2 is expressed as follows:

(j,n=O,
where
, 2 1
n =- j5e2 + Fse3'

Thus by taking account of the previous results this condition becomes

° °
P2x
h 2
-2BX2
-2Bx2
6Ax 2 +4Bx 2 +2E
1[ 0] [0]°
' -2 =
1 0'

that is,

E=O,
520 Chapter 6

In conclusion, the polynomial is written:

- p8 3 2 13
¢(X 2 ,X3 ) =h(3X2 -2X2 X3 +6 X3 )
and
o

(iii) We know that

1 +v - - -
- =--u
6
v (tru
-)J .
E E
Since
v v p
-(tra)J = --[(v+ 1)(l6x2 - 3X3)]J
E Eh
we immediately deduce:

e=~E.[~E h
o
The stress field a satisfying the compatibility conditions, we know that there IS a
displacement field IT which is obtained as follows:
- 1 - - l+vp
U2 ,2 =2(U22 +U2,2)=6 22 =Eh[X 3 -v(16X2 -3x3)]
and thus

Likewise, we have
- _ l+v P
U33 =611 =---[16X2 - 4X3 -v(l6x2 -3x3 )]
, .. E h
and thus

In addition we have
- - _ l+v P
U23, + U3,2 = 26 21. = -E--h 8X 2

and since
 Linear Elasticity 521

we deduce that

By considering an arbitrary constant c, we have

qJ'(x3 )=-16(l-v)x3 +C,
Consequently the components of the general displacement field are
- l+vp 2 2
U2 =---[X2 X3 -V(8X 2 -3X2X3 )+8(v-l)xd+k1x 3 +k2'
E h

while U\ = o.
Of course, the displacement field U could be well determined from further suitable
boundary conditions.

Exercise 6.
A long cylindrical girder of radius R = 0.4 m and density p = 7.8 x 10 3 kg / m 3 is
in equilibrium on a horizontal rectangular surface of very small width I and large length L
of an elastic, homogeneous, isotropic, and isothermal floor whose Young's modulus and
Poisson's ratio are respectively E = 10 4 MPa and v = 0.15 .
A system of cylindrical coordinates is chosen such that the origin 0 is located in the
middle of the rectangular surface, the z-axis is along the corresponding generatrix parallel
to the long sides of the rectangle, and B is measured from the inward normal to the
horizontal surface of the floor.
We assume that the stress tensor fulfills conditions on this surface that are widespread to
the whole floor.
Within the region of the floor such that
I L L L
-«r«- --«z«-
2 2 ' 2 2
and
522 Chapter 6

() = -1[/2 and () = + 1[/2 characterizing half-planes of the surface, express the tensors of
strain and stress due to the girder weight.

Fig. 56 z
Answer. Given the long length of L and the previous conditions for r and z relative to L/2,
this problem can be considered as a plane strain problem, that is such that Uz = O.
Since the viewed area is such that r» 1/2 Saint-Venant's principle is applicable to this
problem, and thus the system of forces exerted by the girder on the floor can be replaced
by the equivalent system of forces concentrated along the z-axis.
We know that the infinitesimal and strain tensors have the following forms:

n l~" :J
E: r{} a r (}
E:
[""
= E:~(} E:(}(} a= a~{} a{}{}
0

whose components do not depend on z.

The outward unit normal on the half-plane of equation () = 1[/2 is explicitly

and thus the stress vector is

 Linear Elasticity 523

The outward unit nonnal on the half-plane of equation () = -tr/2 is explicitly

and thus the stress vector is

For () = tr/2 and () = - tr/2 the stress vectors are zero and thus

We are going to seek infinitesimal strain and stress tensors such that these conditions are
everywhere satisfied, that is, for -tr/2 :0:; (}:o:; tr/2 .
By introducing the Airy function ¢(r.()) such that

we deduce that

~~ = f((}), .!.r 8¢ = ((})

8(} g ,

thus by letting
¢(r,(}) = f((})r + h((})
we obtain:
g((})=~::(r,(})=f'((})+hr;(}) ,

which implies that h is a constant.

The previous components of (T are expressed from only derivatives of ¢ and thus we can
choose h = 0 without restriction.

The compatibility equation L1L1¢ = 0 is written:

524 Chapter 6

f4) (0) + 21"(0) + f(O) = o.

The general solution ofthis equation leads to
¢(r,O) = ArcosO+ B r sinO +C rOcosB+ DrBsinO,

where A,B,C and D are constants to be determined.

But the stress tensor has the form

0'= [
0'
0
rr 0
0
0
0
1
o 0 O'zz
where
1 8¢ 1 82¢
0' =--+---
rr r 8r r2 80 2

=_2 C sinO+2 D cosO

r r
and

Thus it only remains to determine C and D. For that we consider a cylindrical surface:
{(r,O,z): r = k, -7r/2 '.5. 0 '.5. 7r/2},
where k is a constant.
The girder exerts a concentrated force per unit length along oz equal to 7r R 2 P g. This
force per unit length 7rR 2 P g i must be balanced by contact forces per unit length exerted
on the cylindrical surface of outward unit normal 1r •
Let us calculate the resultant of these forces by considering, at any point of this cylindrical
surface, the stress vector

tr]_ [O'rr 0 0
( te - 0 0 0 ·1r
1 =
[O'rr
0 0
0 0
0
j. [1]0 =
(O'rr]
0 ,
tz 0 0 O'zz 0 0 O'zz 0 0
 Linear Elasticity 525

that is:

The resultant of stresses per unit length on the cylindrical surface is

;r ;r

Ll trd(} = il (-2Csin()+ 2 Dcos(})lr d()

2 2

il (- C sin() + D cos(})(cos(} i + sin(} j) d(}

;r

= 2
2

;r

= il [(-Csin2() + D(l+cos2(}»i + (-C(I-cos2(}) + Dsin2(})j]d(}

2

=lLDi-lLCj.

Thus, the equilibrium condition

lL R2 p g i + (lL Di -lLC j) = 0
implies
C=O,

We conclude that
2 cos(} 2 cos(}
eY rr = - 2R p g - - , eY zz =-2 v R p g - - .
r r

The infinitesimal strain tensor is

o o
1 [eY rr
E= -= 0 -v
E
o o o
where
- E 4
E =--2 =1.023018xl0 MPa,
I-v
_ v
v =--=0.17647,
I-v
o
&" - 2393483 00:0 w-,[~
x -0.17647
o
526 Chapter 6

The stress tensor is

U=_2R2pgCO;B 0 0 0 [1 0 0]
o 0 v

~
1 0
= -24485 co;B [ 0 0 ] Pa.
o 0 0.15

Exercise 7.

A circular cylinder 0.8 m long and 0.03 m diameter is compressed at its end faces
by equal and opposite forces of 1.5 x 10 4 N.
By knowing that E =7.4 x 10 4 MPa and v = 0.34 , determine the maximum normal stress,
the maximum shear stress, the negative elongation, and the increase in the lateral
dimension.

Answer. The maximum normal stress is

The maximum shear stress is

1
-umax = 10.61 MPa.
2

The negative elongation is

,dl = (0.8)(21.22) = -229 x 1O-6 m.

7.410 4

The diameter will increase by an amount

M = (0.34)(0.03)~21.22) = 2.9xl0-6 m.
7.410
 Linear Elasticity 527

Exercise 8.

A vertical cylindrical bar of circular cross-sections of radius R = 0.3 m and with

X3 -axis is characterized by a shear modulus Ji = 40 X 10 9 Pa. Two equal and opposite

moments of magnitude M = 2000 Nm are applied at the plane circular ends Io and II'

(i) Calculate the constant variation of angle per unit length (twist per unit length ()~) and
find the displacement vectors of respective points (0.3,0,2) and (0,0.3,2).
(ii) Show that the shear and normal stresses are maximum on the lateral surface I and
calculate them.
(iii) Determine the principal direction at any point of the vertical line defined by x 1 = R
and x 2 = 0.
Answer. (i) The constant variation of angle per unit length ()~ IS obtained from
M = Ji()~I; that is:

()' = M = 2M = 4x 10 3 = 3.93 xlO- 6 m-1.

o JiI Ji7r R4 47r(0.3)4 X 1010
The components of the displacement vector of the point (0.3,0,2) are

and of the point (0,0.3,2) are

U1 =-()~(0.3)2=2.358xlO-6, U2 =0, U3 =0 (m).

(ii) The principal stresses are obtained from the characteristic equation
M
-a --x
° M
I 2
-a - x =0
°
--x
M M
-x
I 1
-a
I 2 I 1

Thus the principal values are

or, by introducing the radial distance r (that is from the X3 -axis), they are
M M
0'1 = -r , an = 0, am = - - r .
I I
528 Chapter 6

Thus the maximum shear stress is

<max = M R = 2M = 47157Pa
I JrR 3
and the maximum normal stress is obviously the previous value.

(iii) For the principal value MR at any point (R,0,X3) we have

I
M
--R o o
I
M MR
0 --R
I I
MR M
0 --R
I I

which leads to the following eigenvector

, J2( 2 +e 3 )·
n=-e
2

Exercise 9.

In the study of the torsion of a cylinder, we consider elliptic cross-sections defined

by

The cylindrical bar is twisted by equal and opposite moments, that is M e3 for the upper
end face and - M e 3 for the lower end face.
The conditions are those of the torsion of general cylinders.
(i) Prove that the function ¢ which verifies the following boundary value problem

L1¢ = 0,
has the solution
¢=CX j X 2 ,

where C is a constant.
(ii) Express the corresponding displacement field and the nonzero stresses.
(iii) A couple of moment M e 3 acting on the upper end face I/, find the corresponding
variation in angular rotation ()~ .
 Linear Elasticity 529

(iv) Given an equal and opposite couple applied on 1:0 , express the stress tensor in
function of M.
(v) Determine the maximum normal stress and the maximal shear stress.

Answer. (i) First, we immediately observe that ¢ satisfies iJ¢ =0 .

Next we consider the boundary whose unit normal is the vector
A Vf 2 Xl X2

n= IIVfl1 = IIV/II (7 el +[;2e 2).

Thus the boundary condition
A 8¢ 8¢
V¢.n =-nl +-n 2 =nlx 2 -n 2x I
ax l ax 2

is written:
2X I X 2 2X 2 X I

IIV/lla IIV/llb
2 2

or
8¢ 2 8¢ 2 2 2
-b Xl +-a X 2 =x1x 2 (b -a).
ax l 8x 2

The function ¢ = CX1X 2 actually satisfies this condition with the constant C = b: - a: .
a +b

(ii) The components of the displacement field are

The nonzero stresses are

(iii) On the upper end face the stress vector is the following

o
t=a.n=[ ~ o
a 31

and the resultant of stress forces on the elliptical upper face end is zero.
530 Chapter 6

By denoting the position vector of any particle of I/ by r, the total moment of stress
forces is
M = L, rl\ldS
= L, (x}e} +x 2 e 2 )I\(a!3e} +a 23 e 2)dS

= L, (x}a 23 -x2 a!3)dSe 3

=---
2J1()~ (b 2 lr--+a
a 3b 2 b 3 a) e
lr-- 3
a 2 +b 2 4 4

J1()~ 3 b3
=-2--2 lra e 3"
a +b
The moment has only one X3 -component M and we deduce the following constant
variation in angular rotation
2 b2
()~ = a + 3 M"
J1lra 3 b

(iv) The stress tensor is expressed in function of the twisting moment as follows:

_ 2Mx2
0 0
lrab 3
2Mx}
a= 0 0
lra 3b
2Mx2 2Mx}
0
lrab 3 lra 3 b

(v) The characteristic equation

a3 _ (2M / (X}2 + x~)a =0

lrab a4 b4

leads to the ordered principal values

2M
am = - - -
lrab

Thus the maximum shear stress and the maximum normal stress are equal to

2M X}2 x~
- - -4+ -4 "
lrab a b
 Linear Elasticity 531

Exercise 10.

A circular aluminum girder of 0.2 m diameter is horizontal and is bent under

mechanical actions exerted on its end faces with respective equations x 2 = 0 and
x2=2(m).
It is assumed that the lateral surface is free from any stress force and the body force of
gravity is neglected.
The resultant of forces acting on the face L'(X2~O) is F = 25 e3 (N) and there is also a total
moment equal to M = 50e. (Nm). The resultant of all forces acting on the other end face
L'(X2~2) is - F .

F
,
,... - 1- - - - - - - - - -- --+-X 2
" 0 ,
I

-F

Fig. 57

We are going to determine the stress field having no lateral component, as well as the
corresponding strains.
More explicitly:
(i) Express the stress vector.
(ii) Prove that
O"I2 X • +0"23X3 =0
on the lateral surface.
(iii) Find the equilibrium equations, strain-stress relations and compatibility equations.
(iv) Deduce the expression of 0"22 in function of x., x 2 and x 3 •
Express the second equilibrium equation from the 0"22 so obtained.
Prove that the compatibility equations are identically satisfied except for the two
compatibility equations containing &.2 and &23 to have to integrate.
532 Chapter 6

Show that the bending problem has been reduced to determine 4 functions
0"12'0"23,e I2 ,e 23 and 7 arbitrary constants.

(v) To find the seven arbitrary constants which have to fulfill conditions of the given
problem, calculate the components of the resultant of all surface forces acting on
1:(X2=0) , next the first and third components of the total moment M a of these surface

forces about 0, as well as the second component of the resultant of stress vectors over
1:(x,=2) .

From the expression of the second component of M 0' deduce the value of the
integration constant introduced in (iv).
(vi) Determine 0"12'0"23,e I2 and e 23 in function of Xl and x 3 ,aswellas 0"22,e ll ,e 23 and

Answer. (i) We note that the system is in equilibrium and the girder is not twisted
Since there is no lateral stress the stress tensor is explicitly

(1 )

and the stress vector t is explicitly

(2)

(ii) The lateral surface is the following set

{(XpX2,X3):XI2 +xi =0.01, 0~X2 ~2}.

By choosing a point (X p O,X3) on the lateral surface we may determine the outward unit
normal to this surface:

which implies that any outward unit normal to the lateral surface is such that

Since the lateral surface is free from any force we necessarily have for any point
(X p X 2 ,X3 ) such that X12+ xi = 0.01, 0 ~ x 2 ~ 2:
 Linear Elasticity 533

that is,
(3)

(iii) The equilibrium equations are immediately:

0"12,2 = 0, (4)

(5)

0"32,2 = O. (6)

We deduce that 0"12 and 0"23 do not depend on x 2 •

By recalling the various Eqs. (6-35), the strain-stress relations are

v
c l1 = --0"22' (7)
E
1
C 22 = -0"22' (8)
E
V
C 33 = --0"22' (9)
E
l+v
C 12 = £ 0 " 1 2 ' (10)

c 13 = 0, (11)
l+v
C 23 = --0"23' (12)
E
Given the following results which imply that C 12 and C 23 do not depend on x 2

l+v
(13)
C 122
, =--0"122
E ' =0,
l+v
C 232 = - - 0 "
, E 23,2
=0 (14)

and since c 13 = 0, the compatibility equations are written:

C I1 ,22 + C 22 ,I1 = 0, (15)
C 22 ,33 +C33 ,22 =0, (16)
+ C ll ,33 = 0,
C 33 ,I1 (17)
(-C 23,1 + C 12 ,3),1 = C II ,23 , (18)
C 22 ,13 = 0, (19)
(-C12,3 + C 23,1),3 = C 33,12 . (20)

(iv) First, Eq. (5) leads to

0"21,12 + 0"22,22 + 0"23,32 = 0,

534 Chapter 6

that is,
0"22,22 =0, (21)

since 0"21 and 0"23 do not depend on x 2 •

Thus Eqs. (7) and (9) imply

8 ll ,22 = 0, 8 33 ,22 = 0

and the compatibility equations (IS) and (16) become

8 22 ,11 = 0, (22)

8 22 ,33 =o. (23)

From these equations and Eq. (8) we deduce

0"22,11 = 0, (24)

0"22.33 = O. (2S)

Likewise Eqs. (19) and (8) lead to

0"22.13 = 0, (26)

From Eqs. (21), (24), and (2S) we deduce that every partial derivative of 0"22 of order
higher than 2 is zero, thus 0"22 is a polynomial of second degree in XI' x 2 and x 3 •
By taking account of Eqs. (21), (24), (2S) and (26), this polynomial is expressed as
(27)
where the constants a, b, c, e, g and h must be determined.
Thus, the second equilibrium equation (S) is written:

0"21,1 + 0" 23,3 + c + gxl + hx3 = 0. (28)

Next, we immediately have

1
8 22 = -(a + bXI + CX2 + eX3 + gxl x 2 + hX2 X3 ).
E
Thus Eqs. (IS), (16), (17), (19) are automatically satisfied, whereas Eqs. (18) and (20) are
written:
V
(8123 - 8 231 )1
, "
=--h,
E
V
(8123 - 8 231 )3 =+-g.
, " E
 Linear Elasticity 535

In addition Eqs. (13) and (14) imply

(&12,3 -&23,1),2 = o.
Therefore we have:
d(&I2,3 - &23,1) = (&12,3 - &23,1 ),1 dxl + (&12,3 - &23,1 ),2 dx2 + (&12,3 - 8 23,1 ),3 dx3

which implies

(29)

where k is an arbitrary constant to be determined, adding to the six other constants.

(v) The outward unit normal to the end face .E(X2~0) is ;, = -e2 and thus the corresponding
stress vector t on this face is explicitly

and the corresponding moment about 0 is

Mo =, At = CT 22 X3el + (0'23 X I -O'12 x 3 )e2 - CT22 XI e 3·

The resultant of all stress vectors over .E(X2~0) has the following components

where .Eo denotes .E(X2~0)'

The components of the total moment about 0 are

M I= tCT22X3dxldx3, M 2= t(CT23XI-O'12X3)dxldx3' M3=-LoCT22xldxldx3'

The following kinds of integrals over the end faces will be introduced:
Ldxldx 3 = 7l' X 10-2 ,

LXI dxldx3 = L X3 dx ldx 3 = LXIX3 dx ldx 3 = Lx; dx ldx 3

= Lx; dx 1dx 3 = LXIx; dxldx3 = LX~X3 dxldx3 = 0,

536 Chapter 6

From

we deduce that

R j = L) (-Xj O"I2),j + Xj O"j2,j ] dx jdx 3

= L)(-XjO"I2),j +Xj(-0"23,3 -c-gxj -hx3)]dxjdx 3

=- L)(XjO"I2),j +(Xj0"23),3]dxjdx 3 -g t x j2dx jdx 3

because r is a circle on the lateral surface such that O"l2nj + 0"23n3 = 0, where nj and n3
are the direction cosines of the outward unit normal to r .

Since x 2 =0 over the left end face, we also have

R2 = - La 0"22 dx jdx 3 = - t (a + bXj + ex 3)dxjdx 3 = -a7r x 10-2

and since

we have

Since x 2 = 0 over the left end face we deduce that the first component of the total
moment is

and the third component is

M3 =- r 0"22 Xj dx jdx 3 = - r bx j2 dx jdx 3 = -b 7r4 x 10- 4 •

.lro .lro

The outward unit normal to the end face L'C 2=2) is ir = e 2 , and since x 2 = 2 the resultant
X

of all stress vectors over this face has the following second component
 Linear Elasticity 537

Thus we have obtained the following values

g=O [because RI = 0],
a=O [because R2 = 0],

h=-~ (Nm-4) [because R3 = 25 ],

7r
2 _
e=-xI0 6 (Nm 3) [because MI = 50 ],
7r
b=O [because M3 = 0],
c=O [because ][2 = 0].
The unused equations of the overabundant system confirm the previous results.

Now, let us express the component M 2 for Io'

M2 = f (a 23 x l -a12 x 3 )dxl dx 3 = ~ f (C 23 X I -c12 x3 )dxl dx 3

10 I + v JEo

=-
E
1+ v
L[(c
0 23
Xl
-
2
) -(c
2,1 12,3
V
+-hx
E I
V
--gx
E 3
Xl
-k)-
2
2

2 2
X3 V V X3
- (C 12 -) 3 + (C23] --hXI +- gX3 + k)-]dx l dx 3
2' 'E E 2
538 Chapter 6

= -EkxlO-2 r dx dx3 +
1
Ektr xl0-4
2(1 + v) 1:0 4(1 + v)

Since M 2 = 0 we deduce that k =0 .

(vi) We are going to seek functions 0'12,0'23'&12'&23 which verify the strain-stress
relations (10) and (12), the equilibrium equation (28), the integrated compatibility
equation (29) and the condition (3) on the lateral surface; namely:
(30)

(31)
and
(32)

l+v
&23 =£0'23' (33)

The equilibrium equation (28) becomes

L+M +c+(2P+N + g)xI +(N +2Q+h)X3 = 0

and Eq. (29) leads to

(I + v)M -(I + v)L-kE +[(1+ v)N -(I +v)2P +vh]xI +[(1+v)2Q- (1+ v)N -vg]X3 = O.

These equations imply:

L=-M-c=-M, (34a)
1 - I -
P = -2(N + g) = -2 N , (34b)

N=-2Q-h, (34c)

-
M=L + -kE = L- , (34d)
l+v
- I vh - h vh - h
P=-N+--=-Q--+ =-Q- , (34e)
2 2(1+ v) 2 2(1+v) 2(1+v)
Q=!N+~=!N. (34f)
2 2(I+v) 2
 Linear Elasticity 539

Finally, the expression a'2 x, + a 23 X3 is a polynomial of third degree in x, and X3 which

vanishes if X,2 +x; -0.01=0.

Thus this polynomial is constructed as follows:

(X,2 + xi - 0.01)(ax, + f3x3 + y)

= ax: + ax,xi - 0.01ax, + f3X,2X3 + f3xi - 0.01f3x3 + yx,2 + yx; - 0.01y,

but this polynomial which is also written:

3 - 2 - 2 -3 - 2-2
=Px, + (Q+ N)x'X 3 + Kx, + (P + N)x, X3 + QX3 + Kx3 + Lx, + Mx3 + Mx,x 3
implies:

1. P = Q+N =-100K,
which is equivalently written:
1 - 3-
--N=-N=-IOOK (35)
2 2 '
2. P + N = Q = -lOOK,
which is equivalently written:

Q=-3Q- h(3+2v) =-lOOK, (36)

2(1 +v)

3.
which is equivalently written:

-M=M, (37)
4. M=O. (38)

From the four previous equations, we deduce:

N=O, K=O,
- __ !!.. 3+2v
Q K=~ 3+2v
- ,
8 l+v 800 l+v '
M=O, L=O,
M=O.
In addition, the various Eqs. (34) lead to
p=O,
540 Chapter 6

N=_~1+2v ,
4 1+v
L =0,
~ h 2v-1
P=---,
8 1+v
Q=O.
Since Poisson's ratio is
v = 0.34
and

we obtain:
0"12 = 99.76877 x,x 3 (10 3Pa) ,

0"23 = -1.09271 + 9.50179x,2 + 109.27056x; (10 3 Pa),

and since E = 0.71 x 10 5 MPa we have

1+ V -6
5 12 =--0"12 =1.883x1O XIX3'
E
5 xi.
23 = -0.0206 x 10-6 + 0.1793 X 10-6 X,2 + 2.0623 X 10-6

Since e=~x106(Nm-3)wehave
7r

and thus
SUMMARY OF FORMULAE

Chapter 1.

Section 1.

e" (e) = 0;,

x =x'e"

e, = fJ i e k ,
k ,
e'*J = /3/e*i ,

x i -aix,k
- k , m~ =a~m"
mj =w(ej) ,

(w,x) =w(x) = m,x' = m>'J ,

'-'
V t = t 'j t
e jo .'j
VY e:
,
t = ( 't) J e ,1 to.
j
VY e, =t 'J.e' J to. e
VY j ,

'(m®x)=x®m,

oj = 1 if i = j,
J 0 if i 7= j,

ts(x,y) = t(t(x,y) + t(y,x)) ,

t A (x, y) = Ht(x,y) - t(y, x)) ,

541
542 Summary of Formulae

Section 2.

+: T%xT% ~T%:(t,U)I-H+U,

. : R x T: ~ T: : (k,t) H k t ,

®:T% xT: ~T::: :(t,U)Ht®U,

I ® U(W(I) ,... ,W(q) ,W(q+l) ,... ,W(q+s)' X(I) ,... , X(p), X(p+l) , ... , X(p+r»)
=t(W(I),· .. ,W(q),X(1)""'X(P») U( W(q+l),···,W(q+S),X(P+l),···,X(p+r»)'
x·l· Y = tijXiyJ = I(x,y),
x.l= I t · X ,

t : (U . V) = U : (V . t) = v : (t . U) E R,

Section 3.

g: ExE ~ R:(X,Y)H g(x,y) = x.y

g(x,y) = X· g. Y = x.y =(X,y) = g;/yJ,

g(x, ): E ~ R :YH g(X,y) = X.y,

(ek)b = g(ek, ) = gkie*i,

#: E* ~E:WHW#, (#rl=b,
 Summary of Formulae 543

E'

x .Yb = Xb . Y = X.y ,
(u® Wb)' Y = (Wb,YJU,

(a ®b) ·(c®d) = (b.c) a ®d,

(a ® b): (c ® d) = (b.c)(a.d) ,

IIXII = ~(x,xJ'

Section 4.

(0'1 /\ ... /\ O'p )(e iJ , ••• , eip) = 15;: Jp '

1
OJ = -OJ i Oil /\ ... /\ O'p ,
p! II· p
544 Summary of Formulae

(p+ q)!
WAP = --Ap+q(w® p),
p!q!

A: QP x Qq ~ Qp+q : (W,p) HWAP,

n
Q(E) = $Q'.
1",0

Section 5.

dx =e i du i ,
I OU ' au'} ,
e, =--e ei = - - , ej ,
} ou,j " au'
I,j
a ,j
=_u_ t i , ou i ,j
t =--1
ou i ou,j

'" ' P uU
, pqr / _ uU '" ,q uU '" m uU
'" rr uU '" ,/ ijk n
t s - --,---,- - - - - - 1 m ,
au' au} auk ours OU n

t: E~ U( {x} x Tx~): x H t(x) = (X,tJ,

XEE

i i k
wj=rjkdu,

a ou,q a k 02U ' ou ,q

r q =rJ _ u___. ~+---
I

pr Ik ou'P ouJ au" au'pou" OU '

r ijk -- g ih rhjk' r ,Jk' = gJh r hlk ,
 Summary of Formulae 545

r ijk =21 (a jgik + a kgij - a igjk'

) rijk -g
- ihrhjk --21 g ih (a jghk + a kghj - a hgjk ) ,

dv = (dv i + vkmDe i ,
Vv j = (ajv' + vkr~)duj, VVj = (akv j - vjr~k)duk ,

i du i
V =-
dt '
~ e '1 /\ ... /\e 'n ,
II = -..juetg

f.l '}··.I.n = e 'l···I.n ~det g '

1 . .
(*t) . = - f.l . t '1 ... lq
lq+l···ln q! 't·o.ln '

af
dfx(e) = - . (x),
au}

(gradJ)i = aJ(x), gradJ = aJ(x)e' i , gradxf· dx = dfx ,

df = (gradf)j e",
grad t(x) = al(x)®e'j ,

gradxt ·dx = dt x '

546 Summary of Formulae

div v = d::-:: a (v' ~detg),

"detg
i

div(fv) = f div v + v.gradf,

divt(x) = akt(x)· e'k = gradl: (e p ® e'P),

. alip
(dlVt). = -
I a'Xp '

I<}

curia =11 : grada,

Av = div(gradv) ,

Cartesian coordinates (Xl ,x 2 ,x 3 ): Orthonormal Cartesian coordinates (x,y,z):

dx = dx ' e i , dx = dxe l + dye 2 + dze 3 = dx ' e i ,

ds 2 = g°dx
lj ' dx} , ds 2 = oi} dx dx} ,
i

F;f =0,
g radf= af e '
ax' '

Given v = Vi ei :
 Summary of Formulae 547

av' av av av
divv=-. , divv=-X + _ Y + _ z
ax' ax ay az'
A _
~V - g
]k a2v'
- - - k ei ,
a vi_e
Il.v =__
2

ax] ax ax] ax] i '

Given t = tV e i ®e]:
at U k
gradt = - k ei ®e j ®e ,
ax

Cylindrical coordinates (r, e, z) :

Nonzero components of g: g" =1, goo = r2 , g zz =1 ,

Nonzero components of g-l: g" = 1, gOO = r- 2 , gZZ =1 ,
detg = r2 ,
In the orthonormal basis (1,,J 0 ,Jz ) :
- 1 -0 1
Nonzero: F:e = --, F,o = - ,
r r

1 = ax 1 = ~ ax 1 = ax
, ar' 0 r ae' z az'

. aI, _ a10 _
Nonzero. --10' - - - 1 ,
ae ae'

g radf = a f 1 + ~ a f 1 + af 1
ar ' r ae 0 a z Z ,

Given dx = dr 1, + r de 10 + dz (, v = v, 1, + ve Ie + V z l z :
548 Summary of Formulae

~(Ovr -v )
r Be 0
1 Ov o
grad v = ;(8ti+ Vr )
1 Ov z
r Be
. Ovr 1 Ov Vr Ov z
dlVV = - + - -o+ - + - ,
Br r Be r Bz

Given t = tlj e i ®e}:

1 t rr - tOO 1 1'8
divt = (B r t rr + -Bo f'O + Bz t + - - ) 1 , + (B, t 8 , + -Be teo + Bz t OZ + 2-)10
TZ

r r r r
1 t ZT
+(B,tzr+-BetZO+oztzz+-)lz'
r r

Spherical coordinates (r,e,¢):

• 2e
Nonzero components of g: g# = r 2 sm ,

Nonzero components of g-l: grr = 1, gOO = r- 2 ,

detg = r 4 sin 2e,

In the orthononnal basis (1, ,Ie ,1¢ ) :

- 1
Nonzero: 1:0 = --,
r
dx = drl, + rde 18 +rsined¢l¢ ,

ds 2 =gljduidu} =dr 2 +r 2 de 2 +r 2 sin 2ed¢2,

1 = Bx 1 = ~ Bx 1 = _1_ Bx
, Or' 0 r Be ' ¢ r sine B¢ ,
Blr 818 Bl,. B1e Bl¢
Nonzero:- = 10 , - = -1" - = smel¢, - = cosel¢, - = -sinel, - coselo ,
Be ae
B¢ B¢ o¢

gradf = Bf 1 +~ Bf 1 +_I_of 1
or ' r Be 0 r sine B¢ ¢ '
 Summary of Formulae 549

-v
a 1 a
- - v --v
1 1 a 1
- - - v --v
ar ' r ao' r IJ rsinO a¢' r ¢
gradv =
a
-v
1 a
1
- - v +-v
1 a
- - - v ---v
cot()
ar IJ r ao IJ r ' rsinO a¢ IJ r ¢
a
-v
1
--v
a 1 a 1
- - - v +-v +--v
cotO
ar ¢ r ao ¢ - rsinO a¢ ¢ r ' r 0

. av, 1 avlJ 1 av¢ cot() 2

dzvv =- +--+---+--v +-v
ar r ao rsinO a¢ r 0 r"
2 1 a . 1 av¢ 2 av, Vo cosO av¢
Jv=[Llv --(v +---(v smO) + - - ) ] 1 +[.1v + - ( - - - - - - - - - ) ] 1
, r2 , sinO ao 0 sinO a ¢ ' IJ r2 ao 2sin 20 sin 20 a¢ 0
2 av, avo v¢
+[Llv +--(-+cot()----)]1
¢ r2 sinO a¢ a¢ 2sinO ¢,

Given t = tlj e, ®e J :

divt = [a, t rr + ~ao t'O + -~-a t r¢+ ~(2trr - tOO - t# + t'O cot())]1,
r rsmO ¢ r
1 1 1
+ [a r to, + -a o tOO + - . - a t lJ¢+ -(3t'0 + (too - t#)cot())]lo
r rsmO ¢ r
1 1 1
+ [a, t¢' + -a o t¢O + -.-t¢¢ + -(3t'¢ + 2t O¢ cotO)]1 ,
r rsmO r ¢
550 Summary of Formulae

Chapter 2.

Section 1.

tPI :Do ~D, cR 3 :XHX=tPI(X),

Xi = tP/ (X),

0< J (X)= D(X I ,X 2 ,X 3 ) <00

- t D(XI ,X 2 ,X3) - ,

dill =dxl dx 2 dx 3 , dllo =dJ{1 dJ{2 dJ{3 ,

J = dill
I '
dll o

U(X,t) = PoP = x(X,t)-X ,

V(X,t) = -atP I (X,t),

at x

r(X,t)
2
=-2 tP a
at
Ix (X,f).

Section 2.

Steady motion:
av
-(x,f) =0,
at
dx l dx 2 dx 3
VI(x,T) V2(x,T) V\x,T)·
 Summary of Formulae 551

Chapter 3.

Section 1.

Xi = FiXJ
J
+ ki '

v=F;·V,

~=detF, =J(t),
flo
Xl = Xl + kt X 2 , X2 =X 2 , X3 = X3 (simple shear),
C = 'Yg·F, C(t)= Cij ei®e J ,

C(V,W) = CrsV'Ws, V.w =V.C·W,

c= 'F· F , Cij = F'qi Fqj ,

_ IIvII _.[i7:c.V _ ~ cev,v)

A(V) - M- Ilvll - WII '
rc:
A(e) = -V '--'if'
I

6(V) = Ilvii~rll = A(V)-l,

IJvII =A(V) IIVII = (6(V)+ 1) I Vii,

.e cev,W) . () Cij
SIn = '~==C==(V==,V~).J"rc==('=w==,w===) , SIn = Kt.jC;
Ce- ,
II 1)

c=~~®~+~~®~+~~®~,
Ai =A(EJ,

YI· = (J.)2
I
= CIi.. = C(EI ' E)I ,
552 Summary of Formulae

L=1(C-g), Lij=1(Cij-gij),

v.w-V.W = 2V.r.W,

L=1(C-I), Lij = 1(Cij-Jij) ,

L=~~®~+~~®~+~~®~,
1 1 2
L1 = -(y
2 1
-1) = -(X
2 1
-I),

sinO = 2L(V, W) , . 2Lij

smO= ,
~I+2L(V,v) ~I+2L(W,W) ~1+2L,. ~1+2L
II ))

F=R·S, F= T·R, T= R·S·R-l,

S = A, EI ® EI + ~ E2 ® E2 + ~ E3 ® E3 ,

L= 1(S2_1) ,

S=~/F-F, R=F·S-l,

C L=T 2 , CL=F.'F, CL=R·C·'R, C= 'R·CL·R,

C=C~I=(F-'Frl,

Rigid body transformation:

x=tJ)/(X)= R·X + k,
C=I, L=O, S=T=I, e=O.

Section 2.
 Summary of Formulae 553

y' = F; yJ + k i + IIPoqoll7J'(Po,qO),

Fi = M>i = (VtPy. = aXi.

) aX} J ax}'

pq = V tP· dX + !!dX!!7J(dX),

dx = F·dX = VtP·dX,
df.l=Jdf.lo,

dSij = d l qi d 2 qi - dl qi d 2 qi. dS k = (f.lg)ijk dSij = ~detgeijk dSij,

dv.dw = dV.C·dW ,

dV.dW= dv.c·dw, cpq=-'-'

ax ax
axp aXq ,

sinO = C(dV.dW) = C(1v,lw) ,

~C(dV.dV) ~C(dW,dW) J(1v) J(1w)

dv.dw-dV.dW = 2dV.L·dW,

ds 2 -ds; = 2 dV.L·dV , ds 2 -ds o2 = 2Lij dV i dV j ,

I 2
L(1v,lv) = -(J (lv)-l),
2

J(lv) = ~ 1+ 2L(lv,Iv ) ,
554 Summary of Formulae

dv.dw- dV.dW = 2 dv.e·dw,

L = IF·e·F, L - aXr axs rs

ij - ax ax. e ,
I )

_ aXr axs L rs,

eY -a- -a
Xi Xj

_ 1 (<5 aXq aXq )

eij -"2 ij - ax ax . '
I )

U(X + dX, t) - U(X,t) = VU (X,t) . dX ,

dU(X,t) = VU(X,t)·dX, dx= dX+ VU·dX,

F=I+VU,
VU=6+OJ,

1 I
OJ = -(VU- VU),
2

F = 1+6 +OJ,

C=I+26+ 'VU·VU,

( VU) .. = aU i = ax; -0
Ij ax) ax }
ij'

U(X +dX,t)-U(X,t) = gradU(x,t)·dx,

gradU = l-gradX,

"6 = ~(gradU + IgradU),

2
 Summary of Formulae 555

1
e = -(gradU +1 gradU -lgradU·gradU), e= 8 -!.. IgradU·gradU,
2 2

Section 3.

v U (X,f) == grad U (x,f),

1
L == & == e == -(gradU+ IgradU),
2

c == 1+2&,

&.
I
= -2 (ax.
au/.
--+--)
au}. ==_ e
I
= _(_I
au+_)
au).
Ij "'x.
] U I
Y :::,
2 uX j :::,
uX;,

A(lv) == 1+ & (1v,lv ),

6(1v) == &(1v,lv) = Iv·& ·Iv,

6(e;) == t:ii (no summation),

6; = 6(£;) == &; == L; •
J == l+tr& == 1+ trVU,

dp;;dp o == DivU == tr&,

Po

S==I+&, R==I+m, R·S==S·R, T·R==R·T,

556 Summary of Formulae

8 11 ,22 + 8 22 .1 I = 28]2,12 ,

8 22,33 +833 ,22 = 2&23,23'

8 33 ,11 +&11,33 = 2831.31'
- 8 23 ,11 +8 3 1,12 +8]2,13 = 8 11 ,23'
- 8 31 ,22 +&]2,23 +8 23 ,21 = 8 22 ,31'

-8]2,33 +8 23 ,31 +8 31 ,32 = 8 33 ,12'

Rigid body transformation:

U(qo,t) == U(po,t) + 8(PO,t)'Poqo + w(Po,t)'Poqo

U(qo,!) == U(po,t) + w(t) 1\ Poqo .

 Summary of Formulae 557

Chapter 4.

Section 1.

Q= dQi
dt X (fixed)
= aQi
at x '

I ~ E :t H X = F(t)· X + k(t),

V(X,t) = F(t). x + k(t),

v = F(t) . V , V = F(t) . v ,

fit = j (t) j.Jo ,

c= fF·F+fF·F,
. 1·
L=-C
2 '

(v.w)" =V.C·W =2v.L·w,

Vf/J(X,t) = Vf/Jf(X) = Fr(X) = F(X,t),

dx' = F(X,t). dX = VV(X,t)· dX,

dj.Jf· = j(X,t)dj.Jo ,

(dv.dw)' = 2dV.L·dW,
. 1 . 1 f
L(X,t) = -C(X,t) = - ('VV· F + F· VV).
2 2

Section 2.

G(t) = F(t)· F- 1(t),

V = G(t)·v,

d j.Jf· = jet) rl(t) = divV,

d j.Jf
1
d(t) = -( G(t) + IG(t))
2
558 Summary of Formulae

(v.w)' = 2 v .d(t)· w,

gradV(x,t) = VV(X,t)· F- 1(X,t) = F(X,t). F- 1(X,t),

VV(X,t) = gradV(x,t)· V f/J(X,t) ,

gradV
av'
= - . e, ®e),
.
ax)

dx' = gradV(x,t)·dx,

(dl1->" = divV(x,t) = trgradV(x,t) ,

dJi
1
d(x,t) = -(gradV(x,t) + tgradV(x,t» ,
2

(dv. dw)' = 2d(x,t)(dv,dw) = 2dv.d(x,t)· dw, (dVi dw i )' = 2dij dv i dw j ,

1 aVi aVj 1
dij = - ( - + - ) =-(Vi ,) +Vj,J,
2 aX j aX i 2

iidvil' ds'
Ildvil = ds = Iv' d(x,t) ·lv '

e= 2 dij(x,t) ,
d(x,t) = d1£, ® £1 + d2 £2 ® £2 + d3 £3 ® £3'
d(x,t) = tF-1(X,t)·i(X,t)·y-l(X,t), i(X,t) = tF(X,t)·d(x,t)·F(X,t) ,
d;:i;:&,

VT(X,t) ;: gradT(x,t) ,

Q(x,t) = ~(gradV(x,t) - tgradV(x,t» ,

2
 Summary of Formulae 559

1
nJ' =-(V
2 J,'
-V.)
',J'

gradV(x,t) = d(x,t) + n(x,t),

- 1
n =-curlV,
2
kij
nk=~ kij~(V_V)=~_S_~(V_V)
2! J.l 2 ),1 ),1 2 ~detg 2 ),1 I,j'

(VU)· = t + OJ ;; (gradU)' = grad V = d + n,

Vq (t) = V/t) + gradV/t) . pq + IIpqll o(pq,t) ,
V(y,t) = V(x,t) + n(t) /\ (y - x) (rigid body motion),

d jk.jm - d jk.jm = d jm.jk - d jm.ik (j"* i, m"* k).

Section 3.

q(x,t) = aq (x,t) + gradq(x,t). V(x,t) ,

at
aq aq
q = - + - - ' =a,q+qjVj ,
ax.
at at ax; '

Vq = aq 1 +~ aq 1(J + aq 1
ar r r ao az Z ,

aq 1 aq 1 aq
Vq = - 1 + - - 1 + - - - - 1 q = aq + aq V + aq V(J + aq l
at r r ao (J r sinO ar/J ' , at ar r ao r ar/J rsinO '

dq
dq (x,t) = aq (x,t) + gradq(x,t) . V(x,t) , - ' =a/q; +q;J.VJ. ,
dt at dt '

V(X,t) = -aXI (X,t) = -

at x at x
(X,t) , aUI
V(x,t)
. au
= U(x,t) = -(x,t) + gradU(x,t)· V(X,t), V = a,u i +Uj,jVj ,
j
at
560 Summary of Formulae

aVI (X,t) =a-22

r(X,t)=- xI (X,t) ,
at x at x
av
r(x,t) = -(x,t) + gradV(x,t)· V(x,t) ,
at
av V2
r(x,t) = -(x,t) + (curlV) A V + grad- ,
av
r , =-'
- 1 a
at 2 af +2&·k.aV
Y ]
k +--(V. V),
2 Ox; ] ]

. aT
T(X,f) = -(x,t) + gradT(x,t)· V(x,f) ,
d
-T
a
at dt Y =-T
at Y +TkVk
Y, ,

c = Jetr v(x, t).1r (x,t) ds,

c= i , (v.lr + v. gradV ·IT)ds,

dt , ,
i
C=!!...i v(x,t).IT(x,t)ds= (v(x,t).IT(x,t)ds),

i, (a,v + gradv·V + 'gradV.v).ITds,

=

C= i, (a, v + curlvI'V). IT ds + [v.V]~ ~ f (a,v+curlvAV).ltds,

! fV.ltds = f y.lt ds ,
c= Je,r I·ltds,

c= L, (a,t+gradt·V+ 'gradV./).lt ds ,
(da)' = divV da - 'gradV ·da,
<1>= Js,r v(x,t).da,

<i>=!!... r v(x,t).da = Js,r (v.da)' = Js,r (v.da + v.divVda - v.'gradV·da) ,

dt Js,

r (a, v + gradv . V + vdivV - gradV . v).da,

<i> = Js,

<i> = Is, (atv + curl(v A V) + V divv).da,

<1> = Jrs, I·]n da = Js,

r t· da ,
<i> = Js,r (a,1 + gradt . V + tdivV - gradV . I) . da,
 Summary of Formulae 561

I(pl't) = r
JD,
q(x,t) dpl ,

i=~ r
dt JD,
q(x,t) dpt = r
JD,
(q(x,t)d pJ = r (dq (x,t) + q(x,t) divV) dpl'
JD, dt
r
i = JD, (a l q(x,t) + gradq(x,t). V + q(x,t) divV) dpl'

r
i= JD, (atq+div(qV»dpt= r aq dp, +
JD, at
r
JaD,
qV.da,

V,=V-W,

dq = aq + V.gradq, dq aq
-=-= -+ W.gradq,
dt at dt at

dq dq
- - - (material domain),
dt - dt
dq aq
--;::::-=-, ~q r
dt JD
q dp = r aq dp
JD at
(material domain at rest).
dt at
562 Summary of Formulae

Chapter 5.

Section 1.

dm . LIm
p(x,t) = - = hm - ,
df.ll .1 )1, ....0 LI III

m(Dt ) = 1 p(x,t) df.lt ,

D,

P(X,t) J(X,t) = Po (X,O) ,

~(P(X,t) J(X,t» =0,

dt

d
-m(D
dt
t ) =- dJ
dt v,
p(x,t) dill = 0,

f (dp + P divV) dill =0, f (a p + gradp.V + pdivV) dill = 0,

JD, dt JD, at
f (a p + div(pV») dll t = 0, f ap dill + f P V. da = 0 ,
JD, at JD, at kD,

dp + pdivV = 0, ap + gradp. V + pdivV = 0, ap + div(pV) = 0,

dt at at

ap + ai(pvJ = 0,
at

. (d)·
div V =- p = +---.!:!:.L
P dill'

~
dt Jo,
f_ pdlll + f_ pVr.da=O,
.laD,
f_ ap dill + f_ pW.da+ f_ pVr.da=O, f_ ap dill + r _ pV.da = 0,
Jo, ot kD, kD, Jo, at .laD,

p dq =~(pq)+div(pqV) =~(pq)+ pq divV ,

dt at dt
 Summary of Formulae 563

p dq = ~(pq) + div(pq ® V),

dt at
r
JMo,
q dm = r
JD,
pq dll ,

-d
dt
L
M D,
qdm= 1 M D,
dq
-dm,
dt
dp
divV(x,t) = 0, - ( x 1)=0 P(X,t) = Po (X,O) (isochoric motion),
dt' ,

r
kD,
pV. da =0 (steady flow),

n- = -21 curlV = 0 ¢:> V = grad¢' ( irrotational flow),

,1¢' = 0 (isochoric irrotational flow).

Section 2.

ten) = lim LJF

Lla--.O ,1a '

t=1:+(1',
~= r V dm r
= ob, pV dll , '»lto = JrM0, x 1\ V dm r
= ob, x 1\ pV dll ,
JM D,
R= r
JM Dt
ydm= r
h,
py dll, Mo= r
JM D,
xl\ydm= r
.vI
xl\py dll.

[MR] = [R(e)]
M(e)
0 0'

~r pVdll= r fdll+ r t(ir)da,

dt ob, ob, JaD,
r
hi
p y dll = r
.In,
f dll + r
.faD,
ten) da,

r ~(pV)dll+ r p(V.lnJVda= r fdll+ r t(n)da,

~& ~ ~ ~

~ r
dt ob,
x 1\ pV dll = r
ob,
x 1\ f dll + r
JaD,
x 1\ ten) da ,
564 Summary of Formulae

r
Jot
x /\ P r dp = .br x /\ f l
dp + r
JaD,
x /\ t(Ii) da,

r x /\ ~(pV)dp
obi ot + r x /\ p(V.1n)V da = r x /\ f
JaD obi I
dp + r
JaD I
x /\ten) da,

~ lqdpt
dt I
= 1(q+qdivV)dp,
,

(n) _
t(Ii) =u.1 tJ - Up ni ,
n'

tee,) = U
J )i'

ten) =" 3
~
n.t(e,)
I
=(j.
)l
ne
I J'
i=1

1 (p r
D,
- f - divu) dp = 0,

p r=f + divu,

p (oV + gradV. V) =f +divu,

ot
oV V2
P ( - + (cur/V) /\ V + grad - ) = f + divu ,
ot 2

~(p V) + div(pV @ V) = f + divu,

ot

1 urr-u ee
PYr =fr +orurr +-oeure +ozu rz + ,
r r
1 2
PYo = fo + 0ruOr + -oou
r
oe + ozuoz +-UOr '
r
1 1
PYz =fz +oruzr +-oeuze +ozu zz +-uzr'
r r
 Summary of Formulae 565

1 1 1
P Yr = Ir + 8 rrr rr + -8 err rB + -.-8¢rrr¢ + -(2rrrr - rree - a ¢¢ + rrrO cot B) ,
r rsmB r
1 1 1
P Ye = Ie + 8 rrrer + -8 erree + - . - 8 ¢ae¢ + -(3a re + (rree - rr¢¢)cotB),
r rsmB r
1 1 1
P Y¢ = I¢ + 8 rrr¢r + -8 orr¢o + - . - 8 ¢a¢¢ + -(3a r¢ + 2a B¢ cotB),
r rsmB r

;; + rrik,k = 0,
r x/\pydp=~r
JLJ, dt JD,
x/\pVdp= r x/\fdp+ r x/\rr']n da ,
JD, JoD,
div(x /\ rr ·1n ) = x /\ divrr + p: rr ,

x /\ P Y = X /\ f + X /\ divrr + p : rr (+ C) (if couple C per unit volume).

Section 3.
1 dV 2
- p-- = f. V + dive rr . V) - rr: d ,
2 dt

~e) -i f. V dp + i
-
Dt Dt
t(n), V da ,

~i) = - JLJ,
r rr: d d p ,
r 1 2
T = Ji - pV dp,
D, 2

dT
-=~e)+~i) ,
dt

1f(X,t) = J(X,t)F-1(X,t) .rr(x). tF-1(X,t), ~ = F-1 . rr . tF-',

Po P

rr:d= (F-1·rr·tF-'):L = ~1f:L,

J
~i) = - JDo
r 1f(X,t): L(X,t) dpo '
rrr
J,-:d dm =
r
J,
1f.
- : L dm,
M D, P MDo Po
df = rr . da , 1f(X,t)· dA = F-' (X,t)· df ,
566 Summary of Formulae

B(X,t) = J(X,t) a(x) . tF-'(X,t),

B(X,!) = F(X,t) '1r(X,t), 1r(X,t) = F-' (X,t)· B(X,t),

dV dflo = 1 J I·V dflo + 1

1 -Po-- l 'L
2
I t B·V. dA - 1r: dflo,
dt
Do2 Do Do Do

I dV 2 - • t .
- Po - - = Po I· V + dlv( B . V) - 1r : L
2 dt

r
JDo
Po r dflo = r
JDo
J I dflo + f
JaDo
B. dA ,

Po r = Po 1 + div B ,

Section 4.

ann=t (Ii), ( ')' , ,

.n=a·n.n=n.a·n,

det( a ij - a 0 ij ) = 0 ,
3 2 I 2
a -ail a +-(aiiall-aijajl)a-det(aij) =0, a 3 - Il a +I2a-I3=0,
2
I, =aji = tra,

12 = 7.I (a;p jj - a;pji) = 7.I((tra) 2 -tra 2) ,

13 = deta,

aD =a - as, tru D = 0,

a=a S +a D ,
 Summary of Formulae 567

(principal values),

Section 5.

n ar .
or=L-joqJ,
j=1 aq

~
V* = L. ~ q *j ,
j=1 aqJ

Or(e) = i t
f·or dl1 + 1t
aD,
(Ii)
.or da,

~:) .b (f + divu). V* dl1 + .b u: d' dl1 + .b u: Q' dl1 ,

=
t t t

~~) = -.b u: d* dl1 , ~~) = - 1,

t
r u d; dl1, If

<;p'= r
1,
j.V'dl1+ r
JaD,
t(Ii).V'da- r
Jot
u:d'dl1,

<;p' = r
.b,
(f +divtu).V· dl1 + r
JaD,
(t(ia) -u·n).V' da +
.b,
r 0": Q' dl1,

_ r U; + ulj,) oU dl1 + r
Or - JD, j JaDt ((
(ia)
-
r 1
O"ljn)oU,da + 1, ulj z(OU j ,) -ou)) dl1,

i
~:CC) = , pro V' dl1,

* +
<;pee) * AI'
<;P(i) - J'(acc) =0 ,
568 Summary of Formulae

Section 6.

<l>c = kD,
f _ pqVr.iida,

~ .v,f pqdp=_f_
dt kD,
pqVr.da- f_ rpd.da+ f pqsdp,
kD, !:i,

Cl>c = kDI
f_ p(q®Vr)·iida= f_ pq(Vr.ii)da,
kDI

Qs = 1 pqs dp ,
I

~ f pq dp =- kD,
dt ,v,
f _ pq(Vr. da )- f _ 'I'd ·da+ f pqsdp,
kD, ob,

1, p(8atq + V.grad q)dp =- kD,.h,

f 'I'd' da + f pq s dp,

1(~(Pq)+pqdivV)dp=-L
I dt
rpd·da+ 1 pqsdp,
D, I

f ~(pq)dp+ f pqV.da=_f rpd.da+ f pqsdp,

.h, at kD, kD, .h,

~i pqdp= f pdq dp=_f rpd,da + 1 pqsdp,

dt I .h, dt kD, I

1 p(8atq + gradq' V)dp = - kD,f 'I'd' da + 1 pqs dp.

I I

rpd,da+ 1 pqsdp,
.v,f (~(pq)+pqdivV)dp=_f
dt kD, I

f
.h,
~(pq)dp+
at f pq(V.ii)da=_f rpd·da+ f
kD, kD,.h,
pqsdp,
 Summary of Formulae 569

dq . dq d'IV(jJd+Pqs'
p-= -dlV(jJd + pqs' p-=-
dt dt
p(oq + V.gradq) = -div(jJd + pqs' p(oq + gradq.V) = -div(jJd + pqs'
at at
d
-(pq) + pqdivV = - div(jJd + pqs • 3..... (pq) + pqdivV = -div(jJd + pqs'
dt dt
a
-(pq) + div(pqV) = -div(jJd + pqs' .i.(pq)+div(pq®V) = -div(jJd + pqs'
at at
f i.(pq)dJi=- f pqV.da - f (jJd.da+ f pq,dJi,
.6 at kD JaD .6

f i.(pq)dJi=- f pq(V.da) - f (jJd·da+ f pqsdJi,

~fu ~ k ~

Q= - f q.da + f r dJi ,
.l3D t .b
t

E(i) = i t
pu dJi = 1MDt
u dm ,

d
-(T + E(i») = 9>(e) +Q,
dt
d V2
P - ( - + u) = dive u . V - q) + f. v + r ,
dt 2
a v2 V2
p(-(-+u) + V.grad(-+ u)) = div(u· V -q) + f. V +r,
at 2 2
d v2 V2
-(p(-+u)) +p(-+u) divV = div(u· V -q)+ f.V +r,
dt 2 2
a V2 v2
-(p(-+u)) + div(p(-+u) V) = div(u· V -q)+ f. v +r,
at 2 2

3.....(T+E(i)+9!)= f (u·n.V-q.n)da+ f rd/1,

~ ~ ~
dE(i)
--=Q-~i)'
dt

3..... f pudJi= f u:ddj1- f q.da+ f rdJi,

dt JD, ~, JaD, ob,
du
p - = u:d -divq+r,
dt
p(ou + V.gradu) = u: d -divq + r,
at
570 Summary of Formulae

~(pu)+ pudivV = u: d -divq + r,

dt
~ (pu) + div(pu V) = u : d - div q + r ,
at
du 1 1
-=-ud +-(r-q)
dt p lj lj P 1,1 '

S 1
= , ps dJl,

~e) =_rJaD, ~q.da+ r ~dJl,

T !., T

-dS ?
dt
!.r -r dJl -
,T
!D,T-1 q. da ,
dS
~i) =--~e)'
dt

~lpSdJl=-L
dt,
~q.da+1~dJl+rei)'
D,T ,T
ds . q r
p dt ? -dzv T + T '
as .q r
p( 8(+ V.grads)? -dzv T + T '

~(ps)+ psdivV ?-div~+~ ,

dt T T
~(ps)+diV(psV) ?-div~+~,
at T T
ds du 1
u:d + peT - - - ) --q.gradT ? 0,
dt dt T
u:d - p(dlf +s dT) - ~q.gradT ? 0 (If = u- Ts) ,
dt dt T

1
~i) = , Y(i) dJl ,

ds . q r
Y(i) = P dt + dzv T - T '

Yei) T = pT ds + divq -~q.gradT -r? 0 ,

dt T
({J = Y(il,
({Jl ds d'zvq-r,
=p T -+
dt
 Summary of Formulae 571

Chapter 6.

Section 2.

ars = lt rsij e )..I '

1 - 1
U(e) = - (6: E : 6), U(6) = - (6 : a) ,
2 2

U( 6 ) -_1- e sr E-rsij
e ji '
2

, 1-
U(a)=-a:C:a,
2
au'
-- .. =e·
aal) zj'

U(6) = U'(a),

U(6) = -A- (I: 6)2 + P 6: 6 = -A- (tr6)2 + p6: 6,

2 2
2
(K=A-+-p),
3

1 A-
6=-(a- (tru)),
2p 3A- + 2p

tra = 3K tr6 ,

E vE E v
a = -- 6 + (tr6) 1 , a·· = --(e +--ekk 15)
l+v (l-2v)(l+v) lj 1+ v l) 1- 2v lj'
572 Summary of Formulae

A =) vE 2pv p(E - 2p) 3K v

(1+v)(1-2v) , 1-2v' 3p-E 'l+v'
E 3K(I- 2v)
p=) - -
2(1 + v) 2(1 + v)
E =) p(3A + 2p) , 2p(1 + v), 3K(I- 2v),
A+P
A E
--1
V=)2(A+p)' 2p ,

2 E 2p(l + v) pE
K=) A+-P
3 ' 3(1-2v) , 3(1- 2v) , 3(3p- E)

G--'-
- 2£]2 ,

In cylindrical coordinates :
au 1 o
= A tre+2p-(--+U r ),
au au
azz = A tre + 2p_z ,
O"rr=Atre+2p a:' 0"00
r ae az
0" = p(~ aUr _ Ue + auB) _ (aU r aU Z )
O"rz -p + , 0" = (au B +~ au Z )

rB r ae r ar' az ar Bz P az r ae '

In spherical coordinates:
au 1 au
O"rr = A tre+ 2p_r, O"ee = A tre+ 2p_(_O +UJ,
ar r ae

O"M
1 1 au
= Atre+2p-(-.--¢ +UoCOt()+U r ), a = p(~ aUr + aU e _ UB )
r sme 8¢ rB r ae ar r'
1 aUr au¢ U¢ 1 1 au o au¢
O"r¢ = p(----:--e 8'" +-a --), 0" = p - ( - - - - + - - U cot()
rsm 'f' r r o¢ r sine a¢ ae ¢ ,

aUr 1 auo Ur 1 8U¢ Uo

tre =-+--+2-+---+-cot()
ar r 8e r rsine 8¢ r '

P'Y =pU=/+(A+J1)grad(divU)+ pAU,

pi = pOi = f' + (A+ p)V'divU+ p,dU'.

 Summary of Formulae 573

In cylindrical coordinates:

pyz = f z + (A-+p
1 ) a
(au r
--+- u+ r 1 au o au Z )
- --+-
azar r r aB az
+p(a u Z +~ au z +~ a u z + a u Z ) .
2 2 2
ar2 r ar r2 aB 2 az 2

In spherical coordinates:
a au, 1 aU
e 1 au¢ 2 Ue
py, = f, + (A, + p) - ( - + - - + - .- - - + -u, + -cotB)
ar ar r aB rsmB a¢ r r

+ p[~~(r2 au,) +-I-~(sinB au,) + _1_ a 2 u,

r2 ar ar r 2 sinB aB aB r 2 sinB a¢2
2 1 a 1 au¢
--(U + - - ( U sinB)+--)]
r2 ' sinB aB e sinB a¢ ,

1 a au, 1 aU e 1 au¢ 2 ue
Py =f +(..1,+p)---(-+--+----+-u +-cotB)
¢ ¢ rsinB a¢ ar r aB rsinB a¢ r ' r
1 a 2 au¢ 1 a . au¢ 1 a 2 u¢
+ 11[--(r -)+---(smB-)+--:------:--
r r2 ar ar r 2sinB aB aB r 2sin 2B a¢2
2 au, aU e U¢
+--(-+cotB----)].
r sinB a¢
2 a¢ 2sinB
574 Summary of Formulae

Section 3.

lJAU+ (J. + p)grad(divU) + f = 0, pU;,jj + (J. + p)U j,j; + J; = 0,

Ll(tre) = Ll(divU) = _ _l-div f = (l + v)(l- 2v) div f ,

J. + 2p E(l- 2v)
1 v
LlO"y. + - - (tro") u· + J;, + f, i + --fkk t5u·· = 0.
l+v ' , 'I-v'

Section 4.

Plane stress problems:

O"ll = rp,22 + V, 0"\2 = -rp,12' 0"22 = rp,ll + V ,

O"ll -V0"22 (I+v)O"12

8 = ~ [ (1 + V)0"12 0"22 - VO"l I

°
LlLlrp + (1- v) LlV = 0,
°
a4 rp 4 rp a
4 rp
LlLlrp = - + 2 - - + - = 0.
a
ax: ax axi ax;
2
1

Plane strain problems:

0"11 = rp,22 + V, 0"12 = -rp,12' 0"22 = rp,ll +V ,

0"11 - V0"22 (1 + V)0"12 oOol

v -
~ [ (1 + V)0"12
E
8 = 0"22 - VO"l1 (v=-, E = - - 2 ),
I-v I-v
° °
LlLlrp + (1- v)LlV = O.
BIBLIOGRAPHY

This list, by no means complete, is made up of books expounding the theoretical

foundations or setting out solved problems about tensors and continuum mechanics.

BOWEN, R.M., Introduction to Continuum Mechanics for Engineers, Kluwer Academic/

Plenum Publishers, DordrechtlNew York, 1989.

BRILLOUIN, L., Les Tenseurs en Mecanique et en Electricite, Masson, Paris, 1960.

DUVAUT, G., Mecanique des Milieux Continus, Masson, Paris, 1990.

ERINGEN, A.C., Mechanics of Continua, John Wiley and Sons, New York, 1967.

FUNG, Y.C., Foundations of Solid Mechanics, Prentice Hall, London, 1965.

GERMAIN, P., Cours de Mecanique des Milieux Continus, vol. 1, Masson, Paris, 1973.

GERMAIN, P., Mecanique, Ecole Polytechnique, Ellipses, Paris, 1986.

GERMAIN, P. & MULLER, P., Introduction a la Mecanique des Milieux Continus, Masson,
Paris, 1994.

HUNTER, S.c., Mechanics of Continuous Media, John Wiley and Sons, New York, 1967.

LAI, W.M., RUBIN, D. & KREMPL, E., Introduction to Continuum Mechanics, Butterworth-
Heinemann, Oxford, 1993.

LICHNEROWICZ, A., Elements de Calcul Tensoriel, A. Colin, Paris, 1964.

LOVELOCK, D. & RUND, H., Tensors, Differential Forms and Variational Principles, John
Wiley and Sons, New York, 1975.

MALVERN, L.E., Introduction to the Mechanics of a Continuous Medium, Prentice-Hall,

Englewood Cliffs, New Jersey, 1969.

MASE, G.E. & MASE, G.T., Continuum Mechanics for Engineers, CRC Press, Boca Raton,
1992.

SALEN~ON, J., Mecanique du Continu, vol. 1 and 2, Ellipses, Paris, 1995.

SEDOV, L., A Course in Continuum Mechanics, vol. 1 and 2, Kluwer Academic Publishers,
Dordrecht, 1987.

575
576 Bibliograpby

SEGEL, L.A. & HADELMANN, G.H., Mathematics Applied to Continuum Mechanics,

Macmillan Co., New York, 1977.

SMITH, D.R., An Introduction to Continuum Mechanics, Kluwer Academic Publishers,

Dordrecht, 1993.

TALPAERT, Y., M(!canique Analytique, vol. 2, Talpaert, Bruxelles, 1982.

TALPAERT, Y., Cours et Applications de Mecanique Generale et Analytique, Ellipses, Paris,

1987.

TALPAERT, Y., Differential Geometry with Applications to Mechanics and Physics, Dekker
Inc., New York, 2000.

TALPAERT, Y., Mechanics, Tensors and Virtual Works, Cambridge International Science
Publishing, Cambridge, 2002.

TRUESDELL, c., The Elements of Continuum Mechanics, Springer-Verlag, Inc., New York,
1966.
GLOSSARY OF SYMBOLS

This list, which is not intended to be complete, is made up of essential symbols.

Scalar quantities are designated in the text by italics, whereas vectors and tensors are
designated by boldface italic letters. In general, every quantity referred to the initial
configuration is represented by an upper case letter and every quantity referred to the
current configuration is denoted by a lower case letter. A dot over a letter indicates
differentiation with respect to time.

Latin Letters

A surface area
Ap anti symmetrization
B Boussinesq tensor or Piola-Lagrange tensor
b flat (or lowering) mapping
b- 1 = # sharp (or raising) mapping
C circulation of vector field
C circulation of tensor field
C (right) Cauchy-Green deformation tensor
t (Lagrangian) rate of stretch tensor
CL left Cauchy-Green deformation tensor
Co initial configuration
Ca abstract configuration
C, current configuration or configuration at time t
C inverse of tensor C L
curve at initial time
c,' c curve at time t
curl I curl (or rotational) of tensor field I
D fixed material domain
Do material domain at initial time
Dc control domain
D, material domain at time t
D, proper motion domain

dynamic dynam

577
578 Glossary of Symbols

dynam of external forces

DivU dilatation
d rate of deformation tensor or Eulerian strain rate tensor
d* virtual strain rate tensor
d j principal rates of extension or stretchings
dA area element in initial configuration
dA area element vector in initial configuration
da area element in current configuration
da area element vector in current configuration
df element of force in current configuration
divv divergence of vector field v
divV(x,t) rate of change of volume per unit volume
dfx differential of fat x
dm mass element
dSij tensor of area element
dS k adjoint of area element tensor
ds 2 metric element
dv absolute differential of vector field v
dv
absolute derivative of vector field v
dt
d time derivative 'along' proper motion
dt
dX vector element in initial configuration
dx transported vector of dX
dpo volume element in initial configuration
dpi volume element in current configuration
E,F, ... (finite-dimensional real) vector spaces
E Young's modulus
E point space
E elasticity tensor
E(i) internal energy
E* dual space of E
E®E®··· tensor product space (of n vector spaces E)
E* ®E* ® ... tensor product space (of n vector spaces E*)
(EJ principal basis
e Euler-Almansi strain tensor
(eJ basis of vector space, orthonormal Cartesian basis
 Glossary of Symbols 579

(e*'), (8') dual basis of (e,)

e j transported vector of if,
F; (X)=V l/J, (X) deformation gradient at X (at t)
F(X,t) = VV Lagrangian velocity gradient
f real-valued function
f resultant of body forces per unit volume
G shear modulus
g acceleration of gravity
g fundamental tensor on (pre)-Euc1idean vector space
g-I conjugate tensor of g
gradxf gradient (covector) of real-valued function/, at x
gradf gradient (covector) field of f
gradt gradient of tensor field t
gradU spatial displacement gradient
gradV(x,t) Eulerian velocity gradient
I volume integral of real-valued function
I identity tensor
I volume integral of tensor field
II first degree invariant
12 second degree invariant
13 third degree invariant
J Jacobian, volume per unit original volume
K bulk modulus

[:1
L
kinetic dynam

Green-Lagrange strain tensor or finite strain tensor

i Lagrangian rate of strain tensor
L, principal strains
L(E,F) set of (continuous) linear mappings of E to F
L(E,R) space oflinear forms on E
Lp(E,R) space ofp-linear forms on E
Mo (dynamic) moment about 0
mo kinetic moment about 0
ii, In unit normal vector
n, direction cosines (or components of iI)
P(X,t) (mass) density in Lagrangian description
Po(X,O) initial (mass) density in Lagrangian description
9>(e) power of external forces
580 Glossary of Symbols

gJ(i) power of internal forces

9{:CC) virtual power owed to acceleration
9{;) virtual power of external forces
9{~ virtual power of internal forces
p pressure
(~) type of tensor (covariant of order p and contravariant of order q)
Q(X,t), Q(X,t) quantity in Lagrangian variables
Q rate of increase of total heat
Qs change of scalar quantity Q caused by source
Qs change of vector quantity Q caused by source
Q(X,t) material derivative of Q(X,t)
q heat flux vector (or heat flux vector per unit area by conduction)
q(x,t), q(x,t) quantity in Eulerian variables
q, (local) change of quantity per unit mass caused by source
qs (local) change of vector quantity per unit mass caused by source
R field of real numbers
R rotation tensor
R (dynamic) resultant
9<v frame of reference
9?, kinetic resultant (or linear momentum)
r radiant heat constant per unit volume
S entropy
S surrace
S right pure stretch tensor
3 material system
s specific entropy (per unit mass)
T kinetic energy
T absolute temperature
T left pure stretch tensor
T; space of (~)-tensors
t time
I, U, ... tensors
I unit vector along the shear stress
I(n) stress vector (or stress)
tr trace
tI transposed tensor of I
I· U contraction in (last I index-first u index)
I :u double contraction in (last I index-I st u index) and (penultimate t index-
Sd u index)
I® u tensor product of tensors
 Glossary of Symbols 581

U displacement
U(e) positive definite quadratic form in linear elasticity
U' (u) dual quadratic form in linear elasticity
u specific internal energy or strain energy per unit mass
V potential energy
V vector of initial configuration
V velocity
V' virtual velocity
Vr relative velocity
V transported vector of V
W velocity of proper motion
X position vector in initial configuration
X' Lagrangian variables or material coordinates
x position vector in current configuration
x,y, Z, ... E E vectors
x h = g(X, ) = OJ covector defined by flat mapping
x' Eulerian variables or spatial coordinates

Greek Letters

a,p,OJ, ... E E' covectors (or I-forms)

r acceleration
I(e) external entropy supply rate
I(i) internal entropy production rate
ijk Christoffel symbols
Yi eigenvalues of Cauchy-Green deformation tensor C
Y(i) internal entropy production rate per unit volume
r(x,t) acceleration in Eulerian description
LJj Laplacian of real-valued function j
Llv Laplacian of vector field v
o Kronecker tensor
0, principal unit elongations
generalized Kronecker symbol
unit elongation or unit extension relative to V
arbitrary increments of generalized coordinates
virtual displacement
virtual angular displacement
virtual work owed to acceleration
virtual work of external forces
582 Glossary of Symbols

virtual work of internal forces

linearized strain tensor, infinitesimal strain tensor
deviator strain tensor
spherical strain tensor
eigenvalues of 6 (principal unit elongations given infinitesimal transf.)
Levi-Civita symbol or permutation symbol
signature of permutation a
() shear angle
iJ rate of shear or shearing
()~ constant variation in angular rotation
A Lame coefficient
A(V) stretch in direction of V
Ai principal stretch
JI Lame coefficient
JI volume form,
Jlo initial volume
Jlt deformed volume or transported volume of Jlo
V Poisson's ratio
;r(X,t) Piola-Kirchhoff stress tensor
p(x,t) (mass) density in Eulerian description
pu strain energy function
normal stress
normal stress (vector)
Cauchy stress tensor or stress tensor
deviator of a
mean normal stress
spherical stress tensor
reversible part of a
irreversible part of a
a, principal stresses
shear stresses
normal stresses
a] > all> alII ordered principal stresses
T shear stress (relative to the surface element of unit normal it
T shear stress (vector)
(/J dissipation
(/J] intrinsic dissipation
(/J2 thermal dissipation
(/J flow or motion
 Glossary of Symbols 583

defonnation (or transfonnation) between configurations

flux of vector field
convection flux of quantity Q
diffusion flux of quantity Q
flux of tensor field
convective flux of quantity Q
diffusion flux of quantity Q
dissipation per unit volume
Airy's function
rpl intrinsic dissipation per unit volume
rp2 thennal dissipation per unit volume
rpc convection flux vector, convection flux tensor
rpd diffusion flux vector, diffusion flux tensor
Ij/ stream function
Q(E) exterior algebra
.0 rate of rotation (tensor)
.0' rate of virtual rotation (tensor)
.0 rate of rotation vector (vorticity vector)
.oP space ofp-fonns
m linearized rotation tensor
m,J.l, ... E .oP p-fonns
m# dual (vector) of Xb

vector of virtual rotation

strict component of m

Other Symbols

® tensor multiplication
(X,y) = X.y scalar product of vectors

Ilxll nonn of x
/\ exterior multiplication
{o;e, } frame of point space E
{x;8,x}= {x;e,} natural frame
*1 adjoint of tensor 1
Vv' absolute differential of contravariant component v'
Vv) absolute differential of covariant component v)
V)v' components of (: )-covariant derivative of v
VU(X,t) material displacement gradient
584 Glossary of Symbols

VtP,(X) deformation gradient

aD, boundary of D,
aD, boundary of D,
1v =v unit vector (or normed vector) defined by v
unit vectors tangent to respective curves Co and c,
'equality' following from the process oflinearization
INDEX

A c
absolute derivative, 87 calculus of variations, 85
canonical isomorphism, 35
absolute differential of (~) components, 81
Cauchy equation of motion, 343-344
absolute differential of (~) components, 81 Cauchy-Green deformation tensor, 178
absolute differential of a vector field, 80, 129 Cauchy stress principle, 332
absolute temperature, 409 Cauchy stress tensor, 339
abstract configuration, 156 Cauchy theorem, 339-341
acceleration, 87, 155,291 change of basis, 7, 62
addition in L( E, F) , 2 change of cobasis, 9
addition ofp-forms, 57 change of natural basis, 70
addition of tensors, 25 characteristic equation, 365-366
adiabatic evolution, 404, 410 Christoffel formulae, 79
adjoint of exterior product, 175 Christoffel symbols, 76
adjoint of p-form, 91 Christoffel symbols of first kind, 78
adjoint of q-vector, 91 circulation, 293
alternation mapping, 49 classical elasticity, 458
anisotropic material, 463 Clausius-Duhem inequality, 411
anti symmetric tensor, 23 compatibility conditions, 216-219
antisymmetrization, 49 completely antisymmetric p-linear form, 49
arc element, 294 completely antisymmetric tensor, 23-24
area deformation, 197-199 completely symmetric tensor, 23
axisymmetric stress tensor, 364 compression, 333, 356
Airy function, 485, 489 configuration, 147
conjugate tensor, 37-38
constitutive equation, 350, 414-416
continuity equation, 316-319
B continuum, 147
contracted multiplication, 28
contracted product, 42, 172
balance equation, 393-406 contraction, 27
Beltrami equations, 472-473 contravariant components, 41-42
biharmonic function, 488, 491 contravariant-contravariant representation, 44
body force, 330
contravariant-covariant representation, 44
Boussinesq tensor, 353
contravariant vector, 13
bras, 3
control domain, 336
bulk modulus, 469
convection, 394

585
586 Index

convection flux, 394, 396 directional derivative of function, 94

convection flux tensor, 396 directional derivative of tensor field, 97
convection flux vector, 394 displacement, 151, 203-209
convective derivative, 264 (material) displacement gradient, 204
convective rate of change, 288 (spatial) displacement gradient, 206-207
convective transport, 174 dissipation, 412
convective transport of vector, 174 dissipation per unit volume, 412
convective transport of volume, 175 divergence
coordinate basis, 67 of displacement, 214
coordinate line, 65 of tensor field, \ 00
coordinate system, 65 of vector field, 99
coordinates, 64 of velocity field, 270
covariant component, 41 double contraction, 29-3\
covariant-contravariant representation, 44 dual basis, 4
covariant-covariant representation, 44 dual space, 3
(: )-covariant derivative, 81 dummy index,S
dynamic dynam, 333
(~ )-covariant derivative, 82 dynamic moment, 333
covariant derivative of t, 84 dynamic resultant, 333
covariant representation
of vector, 41
of vector differential, 76
covariant vector, 13
covector, 2, 11 E
curl of covector field, 10 I-I 03 Einstein summation convention,S
curl of second order tensor, 103 elastic material, 455
current configuration, 148 elasticity tensor, 459
curvilinear coordinates, 66 elastostatic problems, 479, 491
(unit) elongation, 180
equal tensors, 25
energy equation, 407
D entropy, 409
Euclidean vector space, 45
decomposable p-form, 54 Euler-Almansi strain tensor, 191,202-203,
decomposition of motion, 285 207,209
deformation, 149 Euler's equations, 86
deformation gradient, 195, 266 Eulerian coordinates, 157
deformation area, 197-198 Eulerian description, 157
deformed volume, 175 Eulerian strain rate tensor, 271, 276
(mass) density, 316 Eulerian velocity gradient, 274
determinant, 89-90, 365 extended Kronecker tensor, 110
deviator, 368 (unit) extension, 180
differential exterior algebra, 59
of function, 93 exterior form of degree p, 50
of vector, 67 exterior multiplication, 58
diffusion, 395 exterior product of I-forms, 51
diffusion flux, 395, 397 exterior product ofp-forms, 56-58
diffusion flux vector, 395 exterior product of q-vectors, 60
diffusion flux tensor, 397 exterior product space, 57
dilatation, 214, 317 external entropy supply rate, 409
 Index 587

F (generalized) Hookes's law, 458-459

hydrostatic stress, 469
finite strain tensor, 184 hypothesis of small perturbations, 211
first principle ofthennodynamics, 402-408
flat mapping, 36
flow, 152
flux, 297 I
(total) flux tensor, 397
(total) flux vector, 395 incompressible continuum, 322, 324
p-fonn,50 indefinite signature, 47
frame of reference, 64 infinitely thin tube of flow, 327
free index, 6 infinitesimal displacement, 211
fundamental principles of c-dynamics, 335 infinitesimal rotation tensor, 212
fundamental tensor, 34 infinitesimal strain tensor, 212
infinitesimal transfonnation, 210
initial configuration, 148
integral of mass density, 292-293
G internal energy, 403
internal entropy production rate, 410
generalized balance equation, 410 internal entropy production rate per unit
generalized Cauchy's theorem, 346 volume, 412
generalized coordinates, 385 intrinsic dissipation, 413
generalized Hooke's law, 458-459 intrinsic dissipation per unit volume, 413
- in cylindrical coordinates, 472-473 invariant, 364-366
- in general coordinates, 472 irreversible evolution, 413
- in spherical coordinates, 473-474 irrotational flow, 328
generalized Kronecker tensor, 111 isochoric irrotational flow, 329
isochoric motion, 322-325
geodesic, 85-86
geometrical configuration, 147 isotropic material, 463
(plane) isochoric motion, 324
gradient of function, 72-73, 93-96
isothennal evolution, 413
gradient of tensor field, 97-98
graduation, 59
Green-Lagrange strain tensor, 184, 199-202

H Jacobian, 151

hat, 51
heat conduction, 408
heat convection, 408 K
heat flux vector, 403
heat radiation, 408 kets, 3
homogeneous material, 458 kinetic dynam, 333
homogeneous transfonnation, 172 kinetic energy, 349, 403
homogeneous transfonnation motion, 264, kinetic energy theorem, 349-350, 354
268 kinetic moment, 333
homogeneous transfonnation tangent to (/It' kinetic resultant, 333
195-196 Kronecker tensor, 22, 112
588 Index

L maximum shear stress, 363

mean normal stress, 367
Lagrange multiplier method, 361 metric element, 74
Lagrangian coordinates, 150, 157 metric tensor, 47
Lagrange description, 149 mixed tensor, 18, 20
Lagrangian rate of strain tensor, 266 Mohr's circle, 374-380
Lagrangian rate of stretch tensor, 265 moment of momentum, 333
Lagrangian variables, 150 moment of momentum principle, 343-346
Lagrangian velocity gradient, 267 motion, 152
Lame coefficients, 464, 471 motion equations, 343-344, 355
Lame stress ellipsoid, 373 multiplication
Laplacian of function, 103 of p-form by scalar, 57
Laplacian of vector field, 105 of linear mapping by scalar, 2
left Cauchy-Green deformation tensor, 190 of tensor by scalar, 25
left pure stretch tensor, 187, 215-216
length of arc, 85
Levi-Civita symbol, 88
line element, 74 N
linear elasticity, 459
linear form, 2 natural basis, 67
linear mapping, 2 natural frame, 67
linear momentum, 333 Navier's equation, 478-479
linear momentum principle, 342 Navier's equations of motion, 475-477
linearized rotation tensor, 205 nondissipative evolution, 413
linearized strain tensor, 205 norm, 47
local rate of change, 288 normal stress, 356
lowering mapping, 36 normal stress vector, 333

o
M
octahedral normal stress, 370
mass, 316 octahedral plane, 369
mass conservation, 316-319 octahedral shear stress, 370
mass density, 316 one-form, 2
mass flow rate, 3 19 orthogonal vectors, 35
material coordinates, 157 orthonormal basis, 47
material deformation gradient, 195
material derivative, 263-264
material derivative
p
of circulation, 294-296
of flux, 297-298
of integral of density, 293 partially antisymmetric tensor, 24
of tensor, 292 partially symmetric tensor, 23
of vector, 269 Piola-Kirchhoff stress tensor, 35\-353
of vector flux, 300-301 Piola-Lagrange stress tensor, 353-354
of volume, 269 plane deformation problem, 488
material displacement gradient, 204, 21 I plane isochoric motion, 324
material domain, 148, 3 15 plane of Mohr, 375
material surface, 152 plane strain problems, 488-491
 Index 589

plane strain state, 488 rate

plane stress, 380 of change of volume per unit volume, 276
plane stress problems, 485-488 of deformation tensor, 271, 276
plane stress state, 485 of extension, 272, 277
plane stress tensor, 364 of increase oftotal heat, 403
point space, 63 of rotation tensor, 280
Poisson's ratio, 457, 467 of rotation vector, 281
Poisson's theorem, 347 of shear, 273, 277
polar decomposition, 187 of virtual rotation, 388
power (Lagrangian) rate of strain tensor, 266
of external forces, 349 (Lagrangian) rate of stretch tensor, 265
of internal forces, 349 reciprocal basis, 39
pre-Euclidean vector space, 34 reciprocal vector, 39
principal basis, 182 reciprocity of stresses, 356-358
principal rate of extension, 274 reciprocity theorem, 483-484
principal strain, 185 rectilinear coordinates, 66, 74
principal stress, 359 reference configuration, 148
principal stress basis, 361 relative velocity, 306
principle reversible evolution, 413-414
of Saint-Venant, 484 Ricci identities, 79
of superposition, 481-484 Ricci theorem, 84
of virtual work, 391 right pure stretch tensor, 187-189, 215-216
principle of thermodynamics rigid body transformation, 191-193, 221
first - , 402-408 rigid body motion, 286-287
second - , 409-413 rotation equilibrium, 389
principal stretch, 182 rotation tensor, 186
principal unit elongation, 186
product
ofp-form by scalar, 57
of tensor by scalar, 25
S
proper motion, 305-306
pseudo-Eucl idean vector space, 47 Saint-Venant compatibility equations,
pseudo-norm, 47 217-219
pseudovector, 282 Saint-Venant's principle, 484
pure stretch tensor, 187,215-216 scalar multiplication, 34, 40
scalar product, 34, 42
second order contravariant tensor, 17
second order covariant tensor, 16
Q second principle of thermodynamics,
409-414
quadratic form U, 459-462, 482-483 sharp mapping, 37
(dual) quadratic form U', 462, 482 shear angle, 181,201,215
shear modulus, 471
shear stress, 357
shear stress component, 357, 376
R shear stress vector, 333
shearing, 273, 277
radiant heat constant per unit volume, 403 signature, 46
raising mapping, 37 simple compression, 381
simple extension, 491
590 Index

simple shear, 176-177 of type (~), 14

simple shear stress, 383, 471
simple traction, 381 of type (~), 16
skew-symmetric p-1inear form, 49 of type C), 18
spatial coordinates, 157
spatial deformation gradient, 195,207 of type (~), 20
spatial displacement gradient, 206, 211 tensor on point space, 71
specific entropy, 409 tensor product, 10, 26
specific internal energy, 404 tensor product space, 14, 16,20
spherical stress tensor, 364, 368 tensor space, 20
steady flow, 326 test specimen, 455-457
steady isochoric flow, 327 tetrahedron, 337
steady motion, 163 thermal dissipation, 413
strain, 201-202 thermal dissipation per unit volume, 413
strain energy function, 404, 461 torsion of circular cylindrical body, 493-496
strain energy per unit mass, 404 torsion of cylindrical shafts, 496-500
(finite) strain tensor, 184 total flux tensor, 397
(Green-Lagrange) strain tensor, 184 trace, 31, 214
streakline, 154-155 traction, 333, 356
streamline, 161-162 traction test, 456
stream function, 325 trajectory, 153, 161
stress plane, 375 transformation between configurations, 149
stress quadric of Cauchy, 371 transformation of tensor components, 71-72
stress tensor, 339 translation equilibrium, 389
stress vector, 332 transported vector, 174, 196,265
stretch, 180, 199-200, 213 transported volume, 265
stretching, 274, 277 transposed tensor, 19
strict components, 53, 55, 60, 88-90 triaxial stress of revolution, 383
source, 395 tube of flow, 326
sum of p-forms, 57
sum of tensors, 25
surface force, 331
symmetric tensor, 22 u
symmetry of stress tensor, 346, 389
undeformed area, 198
uniaxial compression, 381
uniaxial stress, 381, 466, 491-493
T uniaxial traction, 381
uniform compression, 378
tangential homogeneous transformation, 193- uniform stress tensor, 364
196 uniform tension, 469
tensor algebra, 27 uniform traction, 378
tensor criterion, 31-33 unit elongation, 180,213
tensor density, 89 unit extension, 180
tensor field, 73
tensor multiplication, 26
tensor
of type (n, 11 v
of type (~), 12 vector, 12,91
 Index 591

q-vector, 59 w
vector flux, 298
velocity, 87,158-161,264,290 wedge, 51
velocity field, 268 work theorem, 482
velocity of rigid body motion, 286
(Eulerian) velocity gradient, 274
(Lagrangian) velocity gradient, 267
virtual angular displacement, 387
y
virtual displacement, 385
virtual power
owed to acceleration, 391 Young's modulus, 457, 467
of external forces, 386
of internal forces, 390
virtual rigid body motion, 388
virtual rotation vector, 387
virtual strain rate tensor, 388
z
virtual velocity, 386 zero form, 56
virtual work done by external forces, 386 zero tensor, 22
volume deformation, 197
volume form, 90
vorticity vector, 281