The Mathematics

of Continuity

From General Relativity
to Classical Dynamics
Albert Tarantola
WARNING!
This is an unnished text
It was a project for a book, but,
as my interests shifted towards a dierent project,
I decided to stop this, and start writing my
Elements for Physics.
The occasional reader is invited to have a look
at the Preface and Table of Contents.
It may well be that she/he nds
useful some parts of this text.
The Mathematics of Continuity:
from General Relativity
to Classical Dynamics
Albert Tarantola
Institut de Physique du Globe de Paris
Xxxx Publications
Copyright c _ 1995, by Xxxx Publications.
All rights reserved. Published simultaneously in Jupiter.
Reproduction or translation of any part of this work beyond that permitted by Sections
107 or 108 of the United States Copyright Act without the permission of the copyright
owner is unlawful. Requests for permission or further information should be addressed to
the Permissions Department, Xxxx Publications.
Library of Congress Cataloging in Publication Data:
Tarantola, Albert.
The mathematics of continuity: from General Relativity to Classical Dynamics.
Includes bibliographies.
1. General Relativity.
2. Gravitation.
3. Curvature and torsion.
4. Spin.
5. Dynamics of Continuous media.
XXX.XXX XXX.XX XX-XXXXX
ISBN 0-XXX-XXXXXX
Printed in xxxx
For information contact:
Albert Tarantola
Institut de Physique du Globe de Paris
4, place Jussieu
F-75252 Paris Cedex 05
Phone: +33 1 44 27 38 98
Fax: +33 1 44 27 48 94
E-mail: tarantola@ipgp.jussieu.fr
5
Preface
The theory of continuous media is not of universal validity but it is applicable in numerous domains.
Not only we can analyze the dynamics of uids or solids, but we know that even a good model of the
space-time itself can be based on the theory.
The notions necessary to describe general coordinate systems in Euclidean spaces or curved
surfaces embedded in them are the same as those needed to describe general (non Euclidean) spaces,
and are the object of the dierential geometry. As our physical space is not Euclidean, the study of
the dierential geometry is essential. General spaces may have curvature and torsion, and the tensors
needed to describe them satisfy some important identities: the Bianchi identities.
As demonstrated by Einstein, the geometrical analysis of space-time introduces some tensors that,
when identied with material properties, give because of mathematical properties valid in any
space not only the fundamental equations of conservation of mass, linear and angular momentum,
but also describe the gravitational interactions.
The equations so obtained describe the dynamics of continuous media, but leave some degrees of
freedom: those that are locked by the constitutive equations that allow the conservation equations to
describe dierent types of physical media with dierent rheologies.
Einsteins gravitation theory identies mass with space-time curvature. But, originally, the theory
assumed that space-time had no torsion. This is because when the theory was developed (1915) [note:
give here a historically accurate data] it was not yet realized that, in addition to mass, particles have
another intrinsic property: spin (Stern and Gerlachs paper dates from 1922). There has been so
many failed attemps to geometrize other particle properties, like the electric charge, that I see as a
social mystery the fact that it is not widely recognized how the simple dropping of the assumption no
torsion leads to a consistent theory where, while mass produces curvature of space-time, spin produces
space-time torsion. This theory is not only more general, but more equilibrated that the traditional
one (treating in an equivalent way curvature and torsion). An interesting result is that if mass and
space-time curvature are related through a fundamental constant (the gravitational constant G ),
spin and space-time torsion are related through the same constant. This is enough for trying to put
this theory forward.
An interesting by-product of the theory is that it naturally leads, in the classical limit, to the
dynamics of media where it is not assumed that stresses are symmetric, and where the microtorques
existing, for instance, in magnetized media, appear naturally.
Why the theory presented in this book is not widely accepted as the gravitation theory (or,
equivalently, as the theory of continuous media)? This is possibly due to the hiatus existing between
4-D and 3-D (plus time) formulations. It is true that exact 3-D formulations are so cumbersome that
one may give up any development before the understanding of the results, but it is also true that only
3-D formulations allow the introduction of the notion of force, that gives an intuitive sense to the
gravitational interactions. Also, how could one realize the importance of the fact mentioned above
(the same constant G couples mass to curvature and spin to torsion) if, as in usual formulations, a
system of units is taken where G = c = 1 ?
This book will start describing the geometry of a general manifold (any dimension, with curvature
and torsion, with or without metric). Then the 4-D physics of space-time will be introduced. It
will then be acknowledged that one of these 4 dimensions is special (the time), and an exact 3-D
formulation will be deduced. We will then see how complex can be the gravitational interactions: tens
of dierent forces exist in a continuous medium submitted to gravitational interactions only. No one,
to our knowledge, has explored in depth this part of the theory.
6
I accept in full the viewpoint expressed by John Maddox in a leading article of the journal Nature
(October 22, 1992):
A model is an approximate description of reality [. . . ] and a good model is a model which,
on the one hand, is comprehensible in the sense of providing an image for the mind and
which, at the same time, is calculable. [. . . ] It is dicult to think [. . . ] of a more generally
applicable model of the real world unless it is the tautologous continuum model of three-
dimensional reality. Formally, at least, you can calculate almost anything that way. The
deformation of a solid object under the inuence of an external eld of force? No problem.
[. . . ] But a little reection will show that these elegant ways of talking, which brought great
fame to late-Victorian Cambridge, are not much more than ways of doing calculus in three
dimensions (whence the term tautologous). They have little to say about physics.
In this book, I take more seriosly the denition and interpretation of the geometry (what is curvature,
torsion, parallel transportation, true derivatives) than the denition of mechanical quantities, that
will be introduced by analogy to classical physics. Our problem here will not be to analyze the
experimental limits of the theory, but rather to give a complete formal development of it.
I have been helped in this work by the nice environment existing at the Institut de Physique
du Globe de Paris. Also, my friend Bartolome Coll has always had time to listen to my (repetitive)
questions, and has always given me the good advice for a safe advance through the conceptual jungle of
non directly intuitive space-times. My wife, Maria Zamora, has always actively protected my working
hours against the assaults of the exterior world.
Paris, November 1995
A.T.
Contents
I First Part: Geometry 11
1 Tensor elds 13
1.1 Chapters overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 A small introduction to tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Tensors, in general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.1 Dierentiable manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.2 Tangent Linear Space. Tensors. . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3.3 Vectors and Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3.4 Natural Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3.5 Change of Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.6 Tensors, Densities, and Capacities . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.4 Tensors in metric spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.4.1 The metric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.4.2 The scalar product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.5 Integration, estimation of densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.6 The Levi-Civitas and the Kroneckers tensors . . . . . . . . . . . . . . . . . . . . . . . 39
1.6.1 The Levi-Civitas tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.6.2 Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.6.3 The Kroneckers tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.6.4 The Kroneckers determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2 Derivatives of tensors 45
2.1 Parallel transportation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.1.1 Example of parallel transportation in a metric space . . . . . . . . . . . . . . . 46
2.1.2 Example of parallel transportation in a space without metric . . . . . . . . . . 47
2.1.3 Important Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1.4 Parallel transport of the basis vectors . . . . . . . . . . . . . . . . . . . . . . . 52
2.1.5 Parallel transport of vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.1.6 Parallel transport of densities and capacities . . . . . . . . . . . . . . . . . . . 59
2.1.7 The Connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.1.8 Parallel transport of the metric tensor . . . . . . . . . . . . . . . . . . . . . . . 61
2.2 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.2.1 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.2.2 Second derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.2.3 Derivatives of the metric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.2.4 Old text: Ordinary derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.2.5 Stars and Crosses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.3 The nite volume method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.4 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7
8 CONTENTS
2.4.1 Appendix: Parallel transportation . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.4.2 Appendix: Old text: Parallel transportation . . . . . . . . . . . . . . . . . . . . 81
2.4.3 Appendix: Taylors series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.4.4 Appendix: Some usual formulas of vector analysis . . . . . . . . . . . . . . . . 86
2.4.5 Appendix: Transporting the basis along Meridians and Parallels . . . . . . . . 88
2.4.6 Appendix: Turning around the Pole . . . . . . . . . . . . . . . . . . . . . . . . 89
2.4.7 Appendix: Derivatives of the basis vectors . . . . . . . . . . . . . . . . . . . . . 90
2.4.8 Appendix: Parallel transport at the surface of a sphere . . . . . . . . . . . . . 92
2.4.9 Appendix: Cartesian coordinates: Metric, Connection. . . . . . . . . . . . . . . 93
2.4.10 Appendix: Spherical coordinates: Metric, Connection. . . . . . . . . . . . . . . . 94
2.4.11 Appendix: Cylindrical coordinates: Metric, Connection. . . . . . . . . . . . . . . 96
2.4.12 Appendix: Gradient, Divergence and Curl in usual coordinate systems . . . . . 97
2.4.13 Appendix: Some boxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
2.4.14 Appendix: Computing in polar coordinates . . . . . . . . . . . . . . . . . . . . 103
3 Curvature and Torsion 105
3.1 Curvature and Torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.2 The Bianchi Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.2.1 The Contracted Bianchi identities . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.2.2 The Contracted Bianchi identities: global form . . . . . . . . . . . . . . . . . . 107
3.3 Derivatives of the Levi-Civitas tensors? . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4 More tensors 109
4.1 Totally Antisymmetric tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.2 Dual tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.3 Exterior Product of tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.3.1 Particular cases: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.4 Exterior Derivative of tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5 Integration theory, in short 117
5.1 The volume element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.2 The Stokes theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
II Second part: Gravitation and Dynamics 121
6 The 4-D space-time 123
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.2 Chapters overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3 Measuring proper time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.4 Measuring (improper) distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.5 Dening space-time coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.6 The space-time metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.7 The space-time torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7 Matter 131
7.1 Recall: classical dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
7.2 Mass and spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
CONTENTS 9
8 From geometry to physics 135
8.1 Media at rest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.1.1 Fundamental constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.1.2 All denitions together . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
8.1.3 Space-time connection and curvature . . . . . . . . . . . . . . . . . . . . . . . . 138
8.1.4 Recall: Bianchi identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
8.1.5 4-D eld equations (Einstein-Cartan Equations) . . . . . . . . . . . . . . . . . 139
8.1.6 Expressing the matter properties . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.1.7 The conservation equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.1.8 The simplest model with spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.1.9 The gravitational forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.1.10 3-D evolution equations for a medium locally at rest . . . . . . . . . . . . . . . 144
8.2 Moving media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.2.1 Measuring (improper) velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.2.2 Dening the 4-velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.2.3 Lorentz transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8.2.4 3-D evolution equations for moving media (low velocities) . . . . . . . . . . . . 147
8.2.5 Global conservation equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2.6 Global conservation equations in Euclideanspaces with Cartesian coordinates . 150
8.2.7 Ordinary matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.2.8 Interpretation of the conserved quantities . . . . . . . . . . . . . . . . . . . . . 151
8.3 Classical limit (low velocities and weak elds) . . . . . . . . . . . . . . . . . . . . . . . 153
8.4 Classical elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8.4.1 Elasticity with nonsymmetric stresses . . . . . . . . . . . . . . . . . . . . . . . 155
8.4.2 Elasticity with symmetric stresses . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.5 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.5.1 Appendix: Model of an electron . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.5.2 Appendix: The Einstein-Cartan equations in 3-D form . . . . . . . . . . . . . . 160
8.5.3 Appendix: 3-D evolution equations for a medium locally at rest . . . . . . . . . 162
8.5.4 Appendix: Strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.5.5 Appendix: The deformation as a function of the displacement vector . . . . . . 168
8.5.6 Appendix: Physical dimensions of tensors . . . . . . . . . . . . . . . . . . . . . 170
8.5.7 Appendix: Shearing, stretching, bending and twisting . . . . . . . . . . . . . . 171
8.5.8 Appendix: Tolos letter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.5.9 Appendix: Maxwells equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
8.5.10 Appendix: Demonstration: Time derivative and space integration (Euclidean
spaces) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.5.11 Appendix: Demonstration: Time derivative and space integration (general case) 181
8.5.12 Appendix: Demonstration: Classical conservation equations . . . . . . . . . . . 182
9 Bibliography 185
9 Index 600
10 CONTENTS
Part I
First Part: Geometry
11
Chapter 1
Tensor elds
The rst part of this book recalls some of the mathematical tools developed to describe the geometric
properties of a space. By geometric properties one understands those properties that Pythagoras
(6th century B.C.) or Euclid (3rd century B.C.) were interested on. The only major conceptual progress
since those times has been the recognition that the physical space may not be Euclidean, but may
have curvature and torsion, and that the behaviour of clocks depends on their space displacements.
Still these representations of the space accept the notion of continuity (or, equivalently, of dier-
entiability). New theories are being developed dropping that condition (e.g. Nottale, 1993). They
will not be examined here.
A mathematical structure can describe very dierent physical phenomena. For instance, the struc-
ture 3-D vector space may describe the combination of forces being applied to a particle, as well
as the combination of colors. The same holds for the mathematical structure dierential manifold.
It may describe the 3-D physical space, any 2-D surface, or, more importantly, the 4-dimensional
space-time space brought into physics by Minkowski and Einstein. The same theorem, when applied
to the physical 3-D space, will have a geometrical interpretation (stricto sensu), while when applied
to the 4-D space-time will have a dynamical interpretation.
The aim of this rst chapter is to introduce the fundamental concepts necessary to describe geo-
metrical properties: those of tensor calculus. Many books on tensor calculus exist. Then, why this
chapter here? Essentially because no uniform system of notations exist (indices at dierent places,
dierent signs. . . ). It is then not possible to start any serious work without xig the notations rst.
This chapter does not aim to give a complete discussion on tensor calculus. Among the many books
that do that, the best are (of course) in French, and Brillouin (1960) is the best among them. Many
other books contain introductory discussions on tensor calculus. Weinberg (1972) is particularly lu-
cid. I do not pretend to give a complete set of demonstrations, but to give a complete description of
interesting properties, some of which are not easily found elsewhere.
Perhaps original is a notation proposed to distinguish between densities and capacities. While the
trick of using indices in upper or lower position to distinguish between tensors or forms (or, in metric
spaces, to distinguish between contravariant or covariant components) makes formulas intuitive,
I propose to use a bar (in upper or lower position) to distinguish between densities (like a probability
density) or capacities (like a volume element), this also leading to intuitive results. In particular
the bijection existing between these objects in metric spaces becomes as natural as the one just
mentioned between contravariant and covariant components.
13
14 CHAPTER 1. TENSOR FIELDS
1.1 Chapters overview
A vector at a point of an space can intuitively be imagined as an arrow. As soon as we can introduce
vectors, we can introduce other objects, the forms. A form at a point of an space can intuitively be
imagined as a series of parallel planes. . . At any point of a space we may have tensors, of which the
vectors of elementary texts are a particular case. Those tensors may describe the properties of the
space itself (metric, curvature, torsion. . . ) or the properties of something that the space contains,
like the stress at a point of a continuous medium.
If the space into consideration has a metric (i.e., if the notion of distance between two points has
a sense), only tensors have to be considered. If there is not a metric, then, we have to simultaneously
consider tensors and forms.
It is well known that in a transformation of coordinates, the value of a probability density f at any
point of the space is multiplied by the Jacobian of the transformation. In fact, a probability density is
a scalar eld that has well dened tensor properties. This suggests to introduce two dierent notions
where sometimes only one is found: for instance, in addition to the notion of mass density, , we will
also consider the notion of volumetric mass , identical to the former only in Cartesian coordinates.
If (x) is a mass density, and v
i
(x) a true vector, like a velocity. Their product p
i
(x) = (x) v
i
(x)
will not transform like a true vector: there will be an extra multiplication by the Jacobian. p
i
(x) is
a density too (of linear momentum).
In addition to tensors and to densities, the concept of capacity will be introduced. Under a
transformation of coordinates, a capacity is divided by the Jacobian of the trasformation. An example
is the capacity element dV = dx
0
dx
1
. . . , not to be assimilated to the volume element dV . The
product of a capacity by a density gives a true scalar, like in dM = dV .
It is well known that if there is a metric, we can dene a bijection between forms and vectors (we
can raise and lower indices) through V
i
= g
ij
V
j
. The square root of the determinant of g
ij

will be denoted g and we will see that it denes a natural bijection between capacities, tensors, and
densities, like in p
i
= g p
i
, so, in addition to the rules concerning the indices, we will have rules
concerning the bars.
Without a clear understanding of the concept of densities and capacities, some properties remain
obscure. We can, for instance, easily introduce a Levi-Civita capacity
ijk...
, or a Levi-Civita density
(the components of both take only the values -1, +1 or 0). A Levi-Civita pure tensor can be dened,
but it does not have that simple property. The lack of clear understanding of the need to work
simultaneously with densities, pure tensors, and capacities, forces some authors to juggle with pseudo-
things like the pseudo-vector corresponding to the vector product of two vectors, or to the curl of a
vector eld.
Many of the properties of tensor spaces arte not dependent on the fact that the space may have
a metric (i.e., a notion of distance). We will only assume that we have a metric when the property
to be demonstrated will require it. In particular, the denition of covariant derivative, in the next
chapter, will not depend on that assumption.
In this chapter, the dimension of the dierentiable manifold (i.e., space) into consideration, is
arbitrary (but nite). We will use Latin indices i, j, k, . . . to denote the components of tensors.
In the second part of the book, as we will specically deal with the physical space and space-time,
the Latin indices i, j, k, . . . will be reserved for the 3-D physical space, while the Greek indices
, , , . . . will be reserved for the 4-D space-time.
1.2 A small introduction to tensors
This small introduction is for readers not well acquainted with tensor language. It is independent of
the rest of the book, and can be skipped without harm.
1.2. A SMALL INTRODUCTION TO TENSORS 15
The velocity of the wind at the top of Eiels tower, at a given moment, can be represented by a
vector v with components, in some local, given, basis, v
i
(i = 1, 2, 3) . The velocity of the wind
is dened at any point x of the atmosphere at any time t : we have a vector eld v
i
(x, t) .
The waters temperature at some point in the ocean, at a given moment, can be represented by a
scalar T . The eld T(x, t) is a scalar eld.
The state of stress at a given point of the Earths crust, at a given moment, is represented by
a second order tensor with components
ij
(i = 1, 2, 3; j = 1, 2, 3) . In a general model of
continuous media, where it is not assumed that the stress tensor is symmetric, this means that we need
9 scalar quantities to characterize the state of stress. In more particular models, the stress tensor is
symmetric,
ij
=
ji
, and only six scalar quantities are needed. The stress eld
ij
(x, t) is a second
order tensor eld. Another example of second order tensor arises in the analysis of deformation of an
elastic medium. If u
i
(x, t) is the displacement eld (displacement at time t of the point that, in
the undeformed state, was at point x ), the strain eld is dened by

ij
=
1
2
(
i
u
j
+
j
u
i

i
u
k

j
u
k
) (i = 1, 2, 3; j = 1, 2, 3) , (1.1)
where
i
means true derivative along the i-th direction (the dierence between true or covariant
derivatives and partial derivatives will be seen below).
Tensor elds can be combined, to give other elds. One example has just been given, that denes
the deformation eld
ij
as a function of the displacement eld u
i
(x, t) . Another example is the
scalar eld
c(x, t) =
1
2
3

i=1
3

j=1

ij
(x, t)
ij
(x, t) =
1
2

ij
(x, t)
ij
(x, t) , (1.2)
representing the elastic energy density stored at a point of an elastic medium, due to the elastic
deformation. Also, if n
i
is a unit vector considered at a point inside a medium, the vector

i
(x, t) =
3

j=1

ij
(x, t) n
j
(x) =
ij
(x, t) n
j
(x) (i = 1, 2, 3) (1.3)
represents the traction that the medium at one side of the surface dened by the normal n
i
exerts
the medium at the other side, at the considered point.
As a further example, if the deformations of an elastic solid are small enough, the stress tensor
is related linearly to the strain tensor (Hookes law). A linear relation between two second order
tensors means that each component of one tensor can be computed as a linear combination of all the
components of the other tensor:

ij
(x, t) =
3

k=1
3

=1
c
ijk
(x)
k
(x, t) = c
ijk
(x)
k
(x, t) (i = 1, 2, 3; j = 1, 2, 3) . (1.4)
The fourth order tensor c
ijkl
represents a property of an elastic medium: its elastic stiness. As
each index takes 3 values, there are 3 3 3 3 = 81 scalars to dene the elastic stiness of a solid
at a point (assuming some symmetries we may reduce this number to 21, and asuming isotropy of the
medium, to 2).
We are not here interested in the physical meaning of equations 1.1 to 1.4, but in their structure.
First, tensor notations are such that they are independent on the coordinates being used. This is
not obvious, as changing the coordinates implies changing the local basis where the components of
vectors and tensors are expressed. That the equalities 1.1 to 1.4 hold for any coordinate system, means
that all the components of all tensors will change if we change the coordinate system being used (for
instance, from Cartesian to spherical coordinates), but still the two sides of the expression will take
equal values.
16 CHAPTER 1. TENSOR FIELDS
The mechanics of the notation, once understood, are such that it is only possible to write expres-
sions that make sense (see a list of rules at the end of this section).
For reasons about to be discussed, indices may come in upper or lower positions, like in v
i
, f
i
or T
i
j
. The denitions will be such that in all tensor expression (i.e., in all expressions that will be
valid for all coordinate systems), the sums over indices will always concern an index in lower position
an one index on upper position. For instance, we may encounter the expressions =

3
i=1
A
i
B
i
or
A
i
=

3
j=1

3
k=1
D
ijk
E
jk
. These expressions are simplied by not writing explicitly the sums, like in
= A
i
B
i
and A
i
= D
ijk
E
jk
. This notation is useful as one easily forgets that one is dealing with
sums, and that it happens that, with respect to the usual tensor operations (sum with another tensor
eld, multiplication with another tensor eld, and derivation), a sum of such terms is handled as one
single term of the sum could be handled.
In an expression like A
i
= D
ijk
E
jk
it is said that the indices j and k have been contracted (or
are dummy indices), while the index i is a free index. A tensor equation is assumed to hold for all
possible values of the free indices.
In some spaces, like our physical 3-D space, it is posible to dene the distance between two points.
These are called metric spaces. A mathematically convenient manner to introduce a metric is by
dening the length of an arc by S =
_

ds , where, for instance, in Cartesian coordinates,

ds
2
= dx
2
+dy
2
+dz
2
or, in spherical coordinates, ds
2
= dr
2
+r
2
d
2
+r
2
sin
2
d
2
. In general, we
write ds
2
= g
ij
dx
i
dx
j
, and we call g
ij
(x) the metric eld or, simply, the metric.
The components of a vector v are associated to a given basis (the vector will have dierent
components on dierent basis). If a basis e
i
is given, then, the components v
i
are dened through
v = v
i
e
i
(implicit sum). The dual basis of the basis e
i
is denoted e
i
and is dened by the
equation e
i
e
j
=
i
j
(equal to 1 if i are the same index and to 0 if not). When there is a metric,
this equation can be interpreted as a scalar vector product, and the dual basis is just another basis
(identical to the rst one when working with Cartesian coordinates in Euclidena spaces, but dierent
in general). The properties of the dual basis will be analyzed later in the chapter. Here we just need
to recall that if v
i
are the components of the vector v on the basis e
i
(remember the expression
v = v
i
e
i
), we will denote by v
i
are the components of the vector v on the basis e
i
: v = v
i
e
i
.
In that case (metric spaces) the components on the two basis are related by v
i
= g
ij
v
i
: It is said
that the metric tensor ascends (or descends) the indices.
Here is a list with some rules helping to recognize tensor equations:
A tensor expression must have the same free indices, at the top and at the bottom, of the two
sides of an equality. For instance, the expressions
= A
i
B
i
(1.5)
= g
ij
B
i
C
j
(1.6)
A
i
= D
ijk
E
jk
(1.7)
D
ijk
=
i
F
jk
(1.8)
are valid, but the expressions
A
i
= F
ij
B
i
(1.9)
B
i
= A
j
C
j
(1.10)
A
i
= B
i
(1.11)
are not valid.
1.3. TENSORS, IN GENERAL 17
Sum and multiplication of tensors (with eventual contraction of indices) gives tensors. For
instance, if D
ijk
, G
ijk
and H
i
j
are tensors,
J
ijk
= D
ijk
+G
ijk
(1.12)
K
ijk
m
= D
ijk
H

m
(1.13)
and
L
ik
= D
ijk
H

j
(1.14)
are tensors.
True (or covariant) derivatives of tensor felds give tensor elds. For instance, if E
ij
is a
tensor eld,
M
i
jk
=
i
E
jk
(1.15)
and
B
j
=
i
E
ij
(1.16)
are tensor elds. But partial derivatives of tensors do not dene, in general, tensors. For instance,
if E
ij
is a tensor eld,
M
i
jk
=
i
V
jk
(1.17)
and
B
j
=
i
V
ij
(1.18)
are not tensors, in general.
All objects with indices that are normally introduced are tensors, with four notable exceptions.
The rst exception are the coordinates x
i
(to see that it makes no sense to add coordinates,
think, for instance, in adding the spherical coordinates of two points). But the dierentials
dx
i
appearing in an expression like ds
2
= g
ij
dx
i
dx
j
do correspond to the components on a
vector dr = dx
i
e
i
. Another notable exception is the symbol
i
mentioned above. The third
exception is the connection
ij
k
to be introduced later in the chapter. In fact, it is because
both of the symbols
i
and
ij
k
are not tensors than an expression like

i
V
j
=
i
V
j
+
ik
j
V
k
(1.19)
can have a tensorial sense: if one of the terms at right was a tensor and not the other, their sum
could never give a tensor. The objects
i
and
ij
k
are both non tensors, and what one term
misses, the other term has. The fourth and last case of objects with indices which are not
tensors are the Jacobian matrices arising in coordinate changes
J
i
I
=
x
I
x
i
. (1.20)
That this is not a tensor is obvious when considering that, contrarily to a tensor, the Jacobian
matrix is not dened per se, but it is only dened when two dierent coordinate systems have
been chosen. A tensor exists even if no coordinate system at all has been dened.
1.3 Tensors, in general
This section introduces the very basic tensor concepts (vectors, forms, bases, changes of coordinates. . . ).
It also introduces the notions of densities and capacities.
It is not assumed that the manifold under consideration has a metric. The particular properties
of metric spaces are analyzed in section 1.4.
18 CHAPTER 1. TENSOR FIELDS
1.3.1 Dierentiable manifolds
A manifold is a continuous space of points. In an n-dimensional manifold it is always possible to
draw coordinate lines in such a way that to any point T of the manifold correspond coordinates
x
1
, x
2
, . . . , x
n
and vice versa.
Saying that the manifold is a continuous space of points is equivalent to say that the coordinates
themselves are continuous, i.e., if they are, in fact, a part of
n
. On such manifolds we dene
physical elds, and the continuity of the manifold will allow to dene the derivatives of the consid-
ered elds. When derivatives of elds on a manifold can be dened, the manifold is then called a
dierentiable manifold.
Obvious examples of dierentiable manifolds are the lines and surfaces of ordinary geometry.
Our 3-D physical space (with, possibly, curvature and torsion) is also represented by a dierentiable
manifold. The space-time of general relativity is a four dimensional dierentiable manifold.
A coordinate system may not cover all the manifold. For instance, the poles of a sphere are as
ordinary as any other point in the sphere, but the coordinates are singular there (the coordinate
is not dened). Changing the coordinate system around the poles will make any problem related to
the coordinate choice to vanish there. A more serious diculty appears when at some point, not the
coordinates, but the manifold itself is singular (the linear tangent space is not dened at this point).
Those ane named essential singularities. No eort will be made on this book to classify them.
1.3.2 Tangent Linear Space. Tensors.
Consider, for instance, in classical dynamics, a trajectory x
i
(t) on a space which may not be at, as
the surface of a sphere. The trajectory is on the sphere. If we dene now the velocity at some point,
v
i
=
dx
i
dt
, (1.21)
we get a vector which is not on the sphere, but tangent to it. It belongs to what is called the tangent
linear space to the considered point. At that point, we will have a basis for vectors. At another point,
we will another tangent linear space, and another vector basis.
More generally, at every point of a dierential manifold, we can consider dierent vector or tensor
quantities, like the forces, velocities, or stresses of mechanics of continuous media. As suggested by
gure 1.1, those tensorial objects do not belong to the nonlinear manifold, but to the tangent linear
space to the manifold at the considered point (that will only be introduced intuitively here).
At every point of an space, tensors can be added, multiplied by scalars, contracted, etc. This
means that at every point of the manifold we have to consider a dierent vector space (in general,
a tensor space). It is important to understand that two tensors at two dierent points of the space
belong to two dierent tangent spaces, and can not be added as such (see gure 1.1). This is why we
will later need to introduce the concept of parallel transport of tensors.
All through this book, the two names linear space and vector space will be used as completely
equivalent.
The structure of vector space is too narrow to be of any use in physics. What is needed is the
structure where equations like
= R
i
S
i
T
j
= U
i
V
ij
+ W
j
X
ij
= Y
i
Z
j
(1.22)
make sense. This structure is that of a tensor space. In short, a tensor space is a collection of vector
spaces and rules of multiplication and dierentiation that use elements of the vector spaces considered
to get other elements of other vector spaces.
1.3. TENSORS, IN GENERAL 19
Figure 1.1: Surface with two planes tangent at two points, and a vector drawn at each point. As the
vectors belong to two dierent vector spaces, their sum is not dened. Should we need to add them,
for instance, to dene true (or covariant) derivatives of the vector eld, then, we would need to
transport them (by parallel transportation) to a common point.
1.3.3 Vectors and Forms
When we introduce some vector space, with elements denoted, for instance, V, V

. . . , it often
happens that a new, dierent, vector space is needed, with elements denoted, for instance F, F

. . . ,
and such that when taking an element of each space, we can multiply them and get a scalar,
= F, V ) . (1.23)
In terms of components, this will be written
= F
i
V
i
. (1.24)
The product in 1.231.24, is called a duality product, and it has to be clearly distinguished from an
inner (or scalar) product: in an inner product, we multiply two elements of a vector space; in a duality
product, we multiply an element of a vector space by an element of a dual space.
This operation can always be dened, including the case where the do not have a metric (and,
therefore, a scalar product). As an example, imagine that we work with pieces of metal and we need
to consider the two parameters electric conductivity and temperature T . We may need to
consider some (possibly nonlinear) function of and T , say S(, T) . For instance, S(, T) may
represent a mist function on the (, T) space of those encountered when solving inverse problems
in physics if we are measuring the parameters and T using indirect means. In this case, S
is adimensional
1
. We may wish to know by which amont will S change when passing from point
(
0
, T
0
) to a neighbouring point (
0
+ , T
0
+ T) . Writing only the rst order term, and using
matrix notations,
S(
0
+ , T
0
+ T) = S(
0
, T
0
) +
_
S

S
T
_
T
_

T
_
+. . . , (1.25)
where the partial derivatives are taken at point (
0
, T
0
) . Using tensor notations, setting x =
(x
1
, x
2
) = (, T) , we can write
S(x + x) = S(x) +

i
S
x
i
x
i
1
For instance, one could have the simple expression S(, T) =
|
0
|
s
P
+
|TT
0
|
s
T
, where sP and sT are standard
deviations (or mean deviations) of some probability distribution.
20 CHAPTER 1. TENSOR FIELDS
Box 1.1 Denition of vectors
Consider the 3-D physical space, with coordinates x
i
= x
1
, x
2
, x
3
. In classical mechanics, the
trajectory of a particle is described by the three functions of time x
i
(t) . Obviously the three values
x
1
, x
2
, x
3
are not the components of a vector, as an expression like x
i
(t) = x
i
I
(t) + x
i
II
(t) has, in
general, no sense (think, for instance, in the case where we use spherical coordinates).
Dene now the velocity of the particle at time t
0
:
v
i
(t
0
) =
_
dx
i
dt
_
t=t
0
.
If two particles coincide at some point of the space x
1
0
, x
2
0
, x
3
0
, it makes sense to dene, for instance,
their relative velocity by v
i
(x
1
0
, x
2
0
, x
3
0
, t
0
) = v
i
I
(x
1
0
, x
2
0
, x
3
0
, t
0
) v
i
II
(x
1
0
, x
2
0
, x
3
0
, t
0
) . The v
i
are the
components of a vector.
If we change coordinates, x
I
= x
I
(x
j
) , then the velocity is dened, in the new coordinate
system, v
I
= dx
I
/dt , and we have v
I
= dx
I
/dt = x
I
/x
i
dx
i
/dt , i.e.,
v
I
=
x
I
x
i
v
i
,
which is the standard rule for transformation of the components of a vector when the coordinates
(and, so, the natural basis) change.
Objects with upper or lower indices not always are tensors. The four classical objects which do
not have necessarily tensorial character are:
the coordinates x
i
,
the partial dierential operator
i
,
the Connection Coecients
ij
k
,
the elements of the Jacobian matrix J
i
I
= x
I
/x
i
.
1.3. TENSORS, IN GENERAL 21
= S(x) +
i
x
i
(1.26)
= S(x) + , x) ,
where the notation introduced in equations 1.231.24 is used. As above, the partial derivatives are
taken at point x
0
= (x
1
0
, x
2
0
) = (
0
, T
0
) .
Note: say that gure 1.2 illustrates the denition of gradient as a tangent linear application. Say
that the mille-feuilles are the level-lines of that tangent linear application.
Note: I have to explain somewhere the reason for putting an index in lower position to represent
/x
i
, i.e., to use the notation

i
=

x
i
.
Note: I have also to explain in spite of the fact that we have here partial derivatives, we have
dened a tensorial object: the partial derivative of a scalar equals its true (covariant) derivative.
It is important that we realize that there is no scalar product involved in equations 1.26. Here
are the arguments:
The components of
i
are not the components of a vector in the (, T) space. This can directly
be seen by an inspection of their physical dimensions. As the function S is adimensional (see
footnote 1), the components of have as dimensions the inverse of the physical dimensions
of the components of the vector x = (x
1
, x
2
) = (, T) . This clearly means that x
and are objects that do not belong to the same space.
If equations 1.26 involved a scalar product we could dene the norm of x , the norm of and
the angle between x and . But these norms and angle are not dened. For instance, what
could be the norm of x = (, T) ? Should we choose an L
2
norm? Or, as suggested by
footnote 1, an L
1
norm? And, in any case, how could we make consistent such a denition of a
norm with a change of variables where, instead of electric conductivity we use electric resistivity?
(Note: make an appendix where the solution to this problem is given).
The product in equations 1.26 is not a scalar product (i.e., it is not the product of two elements
belonging to the same space): it is a duality product, multiplying an element of a vector space and
one element of a dual space.
Why this discussion is needed? Because of the tendency of imagining the gradient of a function
S(, T) as a vector (an arrow) in the S(, T) space. If the gradient is not an arrow, then, what
it is? Note: say here that gures 1.3 and 1.4 answer this by showing that an element of a dual space
can be represented as a mille-feuilles.
Up to here we have only considered a vector space and its dual. But the notion generalizes to more
general tensor spaces, i.e., to the case where we have more than one index. For instance, instead of
equation 1.24 we could use an equation like
= F
ij
k
V
ij
k
(1.27)
to dene scalars, consider that we are doing a duality product, and also use the notation of equa-
tion 1.23 to denote it. But this is not very useful, as, from a given tensor F
ij
k
we can obtain scalar
by operations like
= F
ij
k
V
i
W
j
k
. (1.28)
It is better, in general, to just write explicitly the indices to indocate which sort of product we
consider.
Sometimes (like in quantum mechanics), a bra-ket notation is used, where the name stands for
the bra [ and the ket [ ). Then, instead of = F , V ) one writes
= F [ V ) = F
i
V
i
. (1.29)
22 CHAPTER 1. TENSOR FIELDS
Then, the bra-ket notation is also used for the expression
= V [ H [ W ) = H
ij
V
i
W
j
. (1.30)
Note: say that the general rules for the change of component values in a change of coordinates,
allow us to talk about tensors for generalized vectors as well as for generalized forms.
The number of indices that have to be used to represent the components of a tensor is called
the rank, or the order of the tensor. Thus the tensors F and V just introduced are second rank,
or second order. A tensor object with components R
ijk

could be called, in all rigor, a (third-rank-

form)-(rst-rank-vector) will we wil not try to usew this heavy terminology, the simple writing of the
indices being explicit.
Note: say that if there is a metric, there is a trivial identication between a vector space and
its dual, through equations like F
i
= g
ij
V
j
, or S
ijk

= g
ip
g
jq
g
kr
g
s
R
pqr
s
, and in that case, the
same letter is used to designate one vector and its dual element, as in V
i
= g
ij
V
j
, and R
ijk

=
g
ip
g
jq
g
kr
g
s
R
pqr
s
. But in non metric spaces (i.e., spaces without metric), there is usually a big
dierence between an space and its dual.
Gradient and Hessian Explain somewhere that if (x) is a scalar function, the Taylor develop-
ment
(x + x) = (x) + g [x) +
1
2!
x[ H[ x) (1.31)
denes the gradien g and the Hessian H .
Old text We may want the gradient to be perpendicular at the level lines of at O , but there
is no natural way to dene a scalar product in the P, T space, so we can not naturally dene what
perpendicularity is. That there is no natural way to dene a scalar product does not mean that we
can not dene one: we can dene many. For any symmetric, positive-denite matrix with the right
physical dimensions (i.e., for any covariance matrix), the expression
__
P
1
T
1
_
,
_
P
2
T
2
__
=
_
P
1
T
1
_
T
_
C
PP
C
PT
C
TP
C
TT
_
1
_
P
2
T
2
_
denes a scalar product. By an appropriate choice of the covariance matrix, we can make any of
the two lines in gure 1.3 (or any other line) to be perpendicular to the level lines at the considered
point: the gradient at a given point is something univocally dened, even in the absence of any scalar
product; the direction of steepest descent is not, and there are as many as we may choose dierent
scalar products. The gradient is not an arrow, i.e, it is not a vector. So, then, how to draw the
gradient? Roughly speaking, the gradient is the linear tangent application at the considered point.
It is represented in gure 1.4. As, by denition, it is a linear application, the level lines are straight
lines, and the spacing of the level lines in the tangent linear application corresponds to the spacing
of the level lines in the original function around the point where the gradient is computed. Speaking
more technically, it is the development
(x +x) = (x) + g , x ) +. . .
= (x) +g
i
x
i
+. . . ,
when limited to its rst order, that denes the tangent linear application. The gradient of is then
g . The gradient g = g
i
at O allows to associate a scalar to any vector V = V
i
(also at O ):
= g
i
V
i
= g , V ) . This scalar is the dierence of the values at the top and the bottom of the
arrow representing the vector V on the local tangent linear application to at O . The index on
the gradient can be a lower index, as the gradient is not a vector.
1.3. TENSORS, IN GENERAL 23
x
0
f(x)
x
y
f(x,y)
Figure 1.2: The gradient of a function (i.e., of an application) at a point x
0
is the tangent linear
application at the given point. Let x f(x) represent the original (possibly nonlinear) application.
The tangent linear application could be considered as mapping x into the values given by the linearized
approximation fo f(x) : x F(x) = + x . (Note: explain better). Rather, it is mathematically
simpler to consider that the gradient maps increments of the independent variable x , x = x x
0
into increments of the linearized dependent variable: y = y f(x
0
) : x y = x . (Note:
explain this MUCH better).
Note: say that gure 1.5 illustrates the fact that an element of the dual space can be represented
as a mille-feuilles in the primal space or a an arrow un the dual space. And reciprocally.
Note: say that gure 1.6 illustrates the sum of arrows and the sum of mille-feuilles.
Note: say that gure 1.7 illutrstes the sum of mille-feuilles in 3-D.
1.3.4 Natural Basis
A coordinate system associates to any point of the space, its coordinates. Each individual coordinate
can be seen as a function associating, to any point of the space, the particular coordinate. We can
dene the gradient of this scalar function. We will have as many gradients f
i
as coordinates x
i
. As
a gradient, we have seen, is a form, we will have as many forms as coordinates. The usual requirements
that coordinate systems have to fulll (dierent points of the space have dierent coordinates, and
vice versa) gives n linearly independent forms (we can not obtain one of them by linear combination
of the others), i.e., a basis for the forms.
If we have a basis f
i
of forms, then we can introduce a basis e
i
of vectors, through
f
i
, e
j
) =
j
i
. (1.32)
This is illustrated in gure 1.8
If we dene the components V
i
of a vector V by
V = V
i
e
i
, (1.33)
then, we can compute the components V
i
by the formula
V
i
= f
i
, V) , (1.34)
as we have
f
i
, V) = f
i
, V
j
e
j
) = f
i
,

j
V
j
e
j
) =

i
V
j
f
i
, e
j
) =

i
V
j

i
j
= V
j

i
j
= V
i
. (1.35)
24 CHAPTER 1. TENSOR FIELDS
A A
B
B
T/K T/K
P/(N m
-2
) P/(dyne m
-2
)
Figure 1.3: A scalar function (P, T) depends on pressure and temperature. From a given point, two
directions in the P, T space are drawn. Which one corresponds to the gradient of (P, T) ? In the
gure at left, the pressure is indicated in International Units (m, kg, s), while in the gure at right,
the c.g.s. units (cm, g, s) are used (remember that 1 Pa = 10 dyne/cm
2
). From the left gure, we
may think that the gradient is direction A , while from the gure at right we may think it is B . It
is none: the right denition of gradient (see text) only allows, as graphic representation, the result
shown in gure 1.4.
T/K T/K
P/(N m
-2
) P/(dyne m
-2
)
Figure 1.4: Gradient of the function displayed in gure 1.3, at the considered point. As the gradient
is the linear tangent application at the given point, it is a linear application, and its level lines are
stright lines. The value of the gradient at the considered point equals the value of the original function
at that point. The spacing of the level lines in the gradient corresponds to the spacing of the level
lines in the original function around the point where the gradient is computed. The two gures shown
here are perectly equivalent, as it should.
1.3. TENSORS, IN GENERAL 25
f = +1
f = 0
f = 1
f = +2
f = +3
f = +4
v = 0 v = +3 v = +6 v = 3
V
F
"Primal" space Dual space
F,V = 2
Figure 1.5: A point, at the left of the gure, may serve as the origin point for any vector we may want
to represent. As usual, we may represent a vector V by an arrow. Then, a form F is represented by
an oriented pattern of lines (or by an oriented pattern of surfaces in 3-D) with the line of zero value
passing through the origin point. Each line has a value, that is the number that the form associates
to any vector whose end point is on the line. Here, V and F are such that F, V) = 2 . But a
form is an element of the dual space, wich is also a linear space. In the dual space, then, the form F
can be represented by an arrow (gure at right). In turn, V is represented, in the dual space, by a
pattern of lines.
26 CHAPTER 1. TENSOR FIELDS
f = +1
f = 0
g = +1 g = 0
f+g = 0
f+g = +1
V
W
V+W
The sum of two vectors Two vectors
The sum of two forms Two forms
Figure 1.6: When representing vectors by arrows, the sum of two vectors is given by the main diagonal
of the parallelogram drawn by two arrows. Then, a form is represented by a pattern of lines. The
sum of two forms can be geometrically obtained using the parallelogram dened by the principal
lozenge (containing the origin and with positive sense for both forms): the secondary diagonal of the
lozenge is a line of the sum of the two forms. Note: explain this better.
1.3. TENSORS, IN GENERAL 27
f = -1
f = 0
f = +1
g = -1 g = 0 g = +1
g = +1 g = 0 g = -1
f = -1
f = 0
f = +1
f+g = -1
f+g = +1
f+g = 0
f+g = 0
f+g = +1
f+g = -1
Figure 1.7: Sum of two forms, like in the previous gure, but here in 3-D. Note: explain that this
gure can be sheared an one wants (we do not need to have a metric). Note: explain this better.
28 CHAPTER 1. TENSOR FIELDS
Note that the computation of the components of a vector does not involve a scalar product, but a
duality product.
To nd the equivalent of equations 1.33 and 1.34 for forms, one denes the components F
i
of a
form F by
F = F
i
f
i
, (1.36)
and one easily gets
F
i
= F, e
i
) . (1.37)
The notation e
i
for the basis of vectors is quite universal. Although the notation f
i
seems well
adapted for a basis of forms, it is quite common to use the same letter for the basis of forms and for
the basis of vectors. In what follows, we will use the notation
e
i
f
i
. (1.38)
whose dangerousness vanishes only if we have a metric, i.e., when we can give sense to an expression
like e
i
= g
ij
e
j
. Using this notation the expressions
V = V
i
e
i
V
i
= f
i
, V) ; F = F
i
f
i
, F
i
= F, e
i
) (1.39)
become
V = V
i
e
i
V
i
= e
i
, V) ; F = F
i
e
i
, F
i
= F, e
i
) . (1.40)
We have now basis for vectors and forms, so wa can write expressions like V = V
i
e
i
and
F = F
i
e
i
. We need basis for objects with more than one index, so we can write expressions like
B = B
ij
e
ij
; C = C
ij
e
ij
; D = C
i
j
e
i
j
; E = E
ijk...
mn...
e
ijk...
mn...
(1.41)
The introduction of these basis raises a diculty. While we have an immediate intuitive represen-
tation for vectors (as arrows) and for forms (as millefeuilles), tensor objects of higher rank are
more dicult to represent. If a symmetric 2-tensor, like the stress tensor
ij
of mechanics, can be
viewed as an ellipsoid, how could we view a tensor T
ijk
m
? It is the power of mathematics to suggest
analogies, so we can work even without geometric interpretations. But this absence of intuitive inter-
preation of high-rank tensors tells us that we will have to introduce the basis for these objects in a
non-intuitive way. Essentially, what we want is that the basis for high rank tensors is not independent
for the basis of vectors and forms. We want, in fact, more than this. Given two vectors U
i
and
V
i
, we understand what we mean when we dene a 2-tensor W by W
ij
= U
i
V
j
. The basis for
2-tensors is perfectly dened by the condition that we wish that the components of W are precisely
U
i
V
j
and not, for instance, the values obtained after some rotation or change of coordinates.
This is enough, and we could directly use the notations introduced by equations 1.42. Instead,
common mathematical developments introduce the notion of tensor product, and, instead of nota-
tions like e
ij
, e
ij
, e
i
j
, or e
ijk...
mn...
, introduce the notations e
i
e
i
, e
i
e
j
, e
i
e
j
, or
e
i
e
j
e
k
. . . e

e
m
e
n
. . . . Then, equations 1.42 are written
B = B
ij
e
i
e
i
; C = C
ij
e
i
e
j
; D = C
i
j
e
i
e
j
E = E
ijk...
mn...
e
i
e
j
e
k
. . . e

e
m
e
n
. . . . (1.42)
What follows is an old text, to be updated.
In section 1.4.1 we will properly introduce the metric tensor. Let us show here that if the space
into consideration has a scalar product, then, the metric can be computed. Here, the scalar product
of two vectors V and W is denoted V W . Then, dening
dr = dx
i
e
i
(1.43)
1.3. TENSORS, IN GENERAL 29
and
ds
2
= dr dr (1.44)
gives
ds
2
= dr dr = (dx
i
e
i
) (dx
j
e
j
) = (e
i
e
j
) dx
i
dx
j
. (1.45)
Dening the metric tensor
g
ij
= e
i
e
j
(1.46)
gives then
ds
2
= g
ij
dx
i
dx
j
. (1.47)
To emphasize that at every point of the manifold we have a dierent tensor space, and a dierent
basis, we can always write explicitly the dependence of the basis vectors on the coordinates, as in
e
i
(x) . Equation 1.33 is then just a short notation for
V(x) = V
i
(x) e
i
(x) , (1.48)
while equation 1.36 is a short notation for
F(x) = F
i
(x) e
i
(x) . (1.49)
Here and in most places of the book, the notation x is a short-cut notation for x
1
, x
2
, . . . .
The reader should just remember that x represents a point in the space, but it is not a vector.
It is important to realize that, when dealing with tensor mathematics, a single basis is a basis for
all the vector spaces at the considered point. For instance, the vector V may be a velocity, and the
vector E may be an electric eld. The two vectors belong to dierent vector spaces, but the are
obtained as linear combinations of the same basis vectors:
V = V
i
e
i
E = E
i
e
i
,
(1.50)
but, of course, the components are not pure real numbers: they have dimensions. Box 1.2 recalls what
the dimensions of components are.
Let us examine the components of the basis vectors (on the basis they dene). Obviously,
(e
i
)
j
=
i
j
(e
j
)
i
=
i
j
, (1.51)
or, explicitly,
e
1
=
_
_
_
_
_
1
0
0
.
.
.
_
_
_
_
_
e
2
=
_
_
_
_
_
0
1
0
.
.
.
_
_
_
_
_
. . . . (1.52)
Equivalently, for the basis of 2-tensors we have
(e
i
e
j
)
kl
=
i
k

j
l
(1.53)
e
1
e
1
=
_
_
_
_
_
1 0 0
0 0 0
0 0 0

.
.
.
_
_
_
_
_
e
1
e
2
=
_
_
_
_
_
0 1 0
0 0 0
0 0 0

.
.
.
_
_
_
_
_
. . .
e
2
e
1
=
_
_
_
_
_
0 0 0
1 0 0
0 0 0

.
.
.
_
_
_
_
_
e
2
e
2
=
_
_
_
_
_
0 0 0
0 1 0
0 0 0

.
.
.
_
_
_
_
_
. . . (1.54)
30 CHAPTER 1. TENSOR FIELDS
. . . . . . . . .
and similar formular for other basis.
Note: say somewhere that the denition of basis vectors given above imposes that the vectors of the
natural basis are, at any point, tangent to the coordinate lines at that point. The notion of tangency
is independent of the existence, or not, of a metric, i.e., of the possibility of measuring distances in
the space. This is not so for the notion of perpendicularity, that makes sense only if we can measure
distances (and, therefore, angles). In, general, then, the vectors of the natural basis are tangent to the
coordinate lines. When a metric has been introduced, the vectors in the natural basis at a given point
will be mutually perpendicular only if the coordinate lines themselves are mutually perpendicular at
that point. Ordinary coordinates in the Euclidean 3-D space (Cartesian, cylindrical, spherical,. . . )
dene coordinate lines that are orthogonal at every point. Then, the vectors of the natural basis will
also be mutually orthogonal at all points. But the vectors of the natural basis are not, in general,
normed to 1. For instance, gure XXX illustrates the fact that the norm of the vectors of the natural
basis in polar coordinates are, at point (r, ) , |e
r
| = 1 and |e

| = r .
1.3.5 Change of Coordinates
Let us consider two dierent coordinate systems, x
i
and x
I
. The two following matrices are
introduced, called Jacobian matrices,
J

I
i
(x
J
) =
x
i
x
I
(x
J
) , (1.55)
J
i
I
(x
j
) =
x
I
x
i
(x
j
) . (1.56)
For short, we simply write
J

I
i
=
x
i
x
I
(1.57)
and
J
i
I
=
x
I
x
i
. (1.58)
Sometimes, the same letter J is used to designate both J

I
i
and J
i
I
, as the position of indices
indicates exactly the matrix we consider. We do not follow that convention here, as it complicates
some expressions we will nd below.
To remember which matrix has an index, think that, for instance, J

I
i
is obtained taking the
partial derivative of the functions x
i
(x
I
) , and, as such, we obtain a function of x
I
, from where the
index.
As we will see in section 1.3.6, the partial derivatives J

I
i
and J
i
I
, are important because they
allow, in a coordinate change, to obtain the components of tensors in the new coordinate system.
For instance, if we consider, at a given point in the space, a tensor with components T
ij...
k...
in
the natural basis associated to some coordinate system x
i
, and we change to another coordinate
system x
I
, then, the components of the same tensor in the natural basis associated to the new
coordinate system will be
T

IJ...
KL...
= J

I
i
J

J
j
. . . T
ij...
k...
J
k
K
J

L
. . . . (1.59)
The determinants of the Jacobian matrices, are called Jacobian determinants, or, simply, Jaco-
bians,

= det(J

I
i
) =
1
n!

IJK...
J

I
i
J

J
j
J

K
k
. . .
ijk...
(1.60)
1.3. TENSORS, IN GENERAL 31
Box 1.2 Which dimensions have the components of a vector?
Contrarily to the basis of elementary calculus, the vectors dening the natural basis are not normed
to one. Rather, it follows from g
ij
= e
i
e
j
that the length (i.e., the norm) of the basis vector e
i
is
|e
i
| =

g
ii
.
For instance, if in the Euclidean 3-D space with Cartesian coordinates
|e
x
| = |e
y
| = |e
z
| = 1 ,
the use of spherical coordinates gives
|e
r
| = 1 |e

| = r |e

| = r sin .
Denoting by [|V|] the physical dimension of (the norm of) a vector, this gives
[|e
i
|] = [

g
ii
] .
For instance, in Cartesian coordinates,
[|e
x
|] = [|e
y
|] = [|e
z
|] = 1 ,
and in spherical coordinates,
[|e
r
|] = 1 [|e

|] = L [|e

|] = L,
where L represents the dimension of a length. A vector V = V
i
e
i
has components with dimensions
_
V
i
_
=
[|V|]
[|e
i
|]
=
[|V|]
_

g
ii
.
For instance, in Cartesian coordinates,
[V
x
] = [V
y
] = [V
y
] = [|V|]
and in spherical coordinates,
[V
r
] = [|V|]
_
V

_
=
[|V|]
L
[V

] =
[|V|]
L
.
In general, the physical dimension of the component T
ij...
k...
of a tensor T is
_
T
ij...
k...
_
= [|T|] [|e
i
|] [|e
j
|] . . .
1
[|e
k
|]
1
[|e

|]
. . .
= [|T|] [

g
ii
]
_

g
jj

. . .
1
_

g
kk

1
_

. . . .
32 CHAPTER 1. TENSOR FIELDS
x
1
=7 x
1
=8
x
1
=6
x
1
=5
x
2
=3
x
2
=4
x
2
=5
x
2
=2
f
2
=0
f
2
=1
f
2
=2
f
2
=-2
f
2
=-1
f
1
=-2 f
1
=-1 f
1
=0 f
1
=1
Figure 1.8: A system of coordinates, at left, and their gradients, at right. These gradient are forms.
When in an n-dimensional space we have n forms, we can dene n associate vertors by f
i
, e
j
) =
j
i
.
= det(J
i
I
) =
1
n!

ijk...
J
i
I
J
j
J
J
k
K
. . .
IJK...
, (1.61)
where the Levi-Civitas symbols
ijk...
take the value +1 if i, j, k, . . . is an even permutation
of 1, 2, 3, . . . , the value 1 if i, j, k, . . . is an odd permutation of 1, 2, 3, . . . , and the value
0 if some indices are identical. The Levi-Civitas tensors will properly be dened in Section 1.6.1.
It is easy to see that the matrices J

I
i
and J
i
I
are mutually inverses:
J
i
J
J

J
k
=
i
k
, (1.62)
and
J

I
j
J
j
K
=
I
K
, (1.63)
which implies that the Jacobians are also mutually inverses:
=
1

. (1.64)
The second order partial derivatives will also be useful. As there is no risk of confusion, as they
have three indices, the same symbols J

and J can be used:

J

JK
i
(x
L
) =

2
x
i
x
J
x
K
(x
L
) , (1.65)
and
J
jk
I
(x
l
) =

2
x
I
x
j
x
k
(x
l
) , (1.66)
or, for short,
J

JK
i
=

2
x
i
x
J
x
K
, (1.67)
and
J
jk
I
=

2
x
I
x
j
x
k
. (1.68)
In Euclidean spaces, the connection coecients (see below) vanish if we use Cartesian coordinates
y
I
. In any other coordinate system x
i
, they are given by (see demonstration below)

ij
k
= J
ij
S
J

S
k
, (1.69)
1.3. TENSORS, IN GENERAL 33
i.e.,

ij
k
=

2
y
S
x
i
x
j
x
k
y
S
. (1.70)
When the coordinate system x
i
is called old, and the coordinate system x
I
is called
new, then, what is usually called the Jacobian determinant is that of J

I
i
.
1.3.6 Tensors, Densities, and Capacities
We are familiar with two sorts of scalar objects (i.e., objects without indices): one class has its
values at every point of the space unchanged when we change coordinates, like a temperature eld; the
other class has its values multiplied by the Jacobian of the transformation, like a probability density.
It is often less obvious but the Jacobian of the transformation also appears when transforming
the components of some tensor objects (i.e., objects with indices): when it is said that a vector
product of two vectors is a vector, excepted that it changes sign if we reverse the sense of one of the
coordinate lines, we are, in fact, multiplying the components of the vector by the Jacobian of the
transformation.
Let us start with the crude, general denition. An object with components Q
ij...
kl...
is called a
density of order p if, in a change of coordinates
x
I
= x
I
(x
j
) , (1.71)
the components transform as
Q

IJ...
KL...
= (

)
p
J

I
i
J

J
j
. . . Q
ij...
k...
J
k
K
J

L
. . . (1.72)
where J

I
i
and J
i
I
are the Jacobian matrices
J

I
i
(x
J
) =
x
i
x
I
(x
J
) , (1.73)
and
J
i
I
(x
j
) =
x
I
x
i
(x
j
) , (1.74)
and

is the Jacobian determinant, i.e, the determinant of the matrix J

I
i
.
If p = 0 , that is, if the Jacobian is not in equation 1.72, we say that Q
ij...
kl...
is a (pure) tensor,
and we use the simple notation Q
ij...
kl...
.
If p = 1 we say that Q
ij...
kl...
is a (tensor) density, and we use the notation Q
ij...
kl...
.
If p = 1 we say that Q
ij...
kl...
is a (tensor) capacity, and we use the notation Q
ij...
kl...
.
To be explicit, let us write how change the components of tensors with rank 0 , 1 , and 2 .
Comment: Say somewhere that the bars have the same sort of rules as the indices: one upper bar
multiplied by a lower bar gives no bar, the total number of bars at each side has to be homogeneous,
etc.
As we are still dealing with spaces that may not have a metric, we cannot transform forms into
vectors, or densities and capacities into pure tensors. This will be done in Section 1.4.1. Comment:
But the structure of the book has changed, and now I have to give all this here.
Comment: I have to say somewhere that, as in a change of variables, a density is multiplied by
the Jacobian determinant, this, gives, when applied to a change of the fundamental density g , from
Cartesian coordinates (where it takes the numerical value of 1), to an arbitrary system of coordinates,
that g equals the value of the Jacobian determinant (between the given coordinates and the Cartesian
ones). I have checked this in dimension 2. It is dicult to give a valid expression of the property, as the
fundamental density is a density and the Jacobian determinant a pure scalar. Let us try nevertheless.
34 CHAPTER 1. TENSOR FIELDS
The Jacobian is
J
I
i
=
x
i
X
I
, (1.75)
and the Jacobian determinant is
=
1
n!

ijk...
J
i
I
J
j
J
J
k
K
. . .
IJK...
. (1.76)
We see that is a true scalar. If the coordinates X
I
are Cartesian coordinates in an Euclidean
space, we have the property
g = . (1.77)
But how can we have an identity between a density and a true scalar? Assuredly, this is not a tensor
equation. For it makes only sense in the particular case when the X
I
are Cartesian coordinates.
(Comment: more fundamentally, g is a tensor eld, and is not). In that case, the distinction
between true scalars, densities and capacities, disappears and, for instance, we have

IJK...
=
IJK...
=
IJK...
. (1.78)
Then, we can write
=
1
n!

ijk...
J
i
I
J
j
J
J
k
K
. . .
IJK...
, (1.79)
which shows that, in the special circumstance that the X
I
are Cartesian coordinates, the Jacobian
determinant can be considered a density as well.
Comment: what precedes seems bizarre. . .
Comment: Give somewhere the formula
i
g = g
i
. It can be justied by the fact that, for any
density, s ,
k
s =
k
s
k
s , and the result follows by using s = g and remembering that
k
g = 0 .
1.4 Tensors in metric spaces
1.4.1 The metric tensor
Comment: explain here that it is possible to give a lot of structure to a manifold (tangent linear space,
(covariant) derivation, etc.) without the need of a metric. It is introduced here to simplify the text,
as, if not, we would have need to come bak to most of the results to add the particular properties
arising when there is a metric. But, in all rigor, it would be preferable to introduce the metric after,
for instance, the denition of covariant dierentiaition, that does not need it.
Having a metric in a dierential manifold means being able to dene the length of a line. This
will then imply that we can dene a scalar product at every local tangent linear space (and, thus, the
angle between two crossing lines).
The metric will also allow to dene a natural bijection between vectors and forms, and between
tensors densities and capacities.
A metric is dened when a second rank symetric form g with components g
ij
is given. The
length L of a line x
i
() is then dened by the line integral
L =
_

ds , (1.80)
where
ds
2
= g
ij
dx
i
dx
j
. (1.81)
Once we have a metric, it is possible to dene a bijection between forms and vectors. For, to the
vector V with components V
i
we can associate the form F with components
F
i
= g
ij
V
j
. (1.82)
1.4. TENSORS IN METRIC SPACES 35
Then, it is customary to use the same letter to designate a vector and a form that are linked by this
natural bijection, as in
V
i
= g
ij
V
j
. (1.83)
The inverse of the previous equation is written
V
i
= g
ij
V
j
, (1.84)
where
g
ij
g
jk
=
i
k
. (1.85)
The reader will easily give sense to the expression
e
i
= g
ij
e
j
. (1.86)
The equations above, and equations like
T
ij...
kl...
= g
ip
g
jq
. . . g
kr
g
ls
. . . T
pq...
rs...
, (1.87)
are summarized by saying that the metric tensor allows to raise and lower indices.
The value of the metric at a particular point of the manifold allows to dene a scalar product for
the vectors in the local tangent linear space. Denoting the scalar product of two vectors V and W
by V W , we can use any of the denitions
V W = g
ij
V
i
W
j
= V
i
W
j
= V
i
W
j
. (1.88)
To dene parallel transportation of tensors, we have introduced a connection
ij
k
. Now that we
have a metric we may wonder if when parallel-transporting a vector, it conserves constant length. It is
easy to show (see demonstration in [Comment: where?]) that this is true if we have the compatibility
condition

i
g
jk
= 0 , (1.89)
i.e.,

i
g
jk
= g
sk

ij
s
+g
js

ik
s
. (1.90)
The compatibility condition 1.89 implies that the metric tensor and the nabla symbol commute:

i
(g
jk
T
pq...
rs...
) = g
jk
(
i
T
pq...
rs...
), (1.91)
which, in fact, means that it is equivalent to take a covariant derivative, then raise or lower an index,
or rst raise or lower an index, then take the covariant derivative.
Note: introduce somewhere the notation

ijk
= g
ks

ij
s
, (1.92)
warn the reader that this is just a notation: the connection coecients are not the components of
a tensor. and say that if the condition 1.89 holds, then, it is possible to compute the connection
coecients from the metric and the torsion:

ijk
=
1
2
(
i
g
jk
+
j
g
ik

k
g
ij
) +
1
2
(S
ijk
+S
kij
+S
kji
) . (1.93)
As the basis vectors have components
(e
i
)
j
=
i
j
, (1.94)
36 CHAPTER 1. TENSOR FIELDS
we have
e
i
e
j
= g
ij
. (1.95)
Dening
dr = dx
i
e
i
(1.96)
gives then
dr dr = ds
2
. (1.97)
We have seen that the metric can be used to dene a natural bijection between forms and vectors.
Let us now see that it can also be used to dene a natural bijection between tensors, densities, and
capacities.
We denote by g the determinant of g
ij
:
g = det(g
ij
) =
1
n!

ijk...

pqr...
g
ip
g
jq
g
kr
. . . . (1.98)
The two upper bars recall that g is a second order density, as there is the product of two densities
at the right-hand side.
For a reason that will become obvious soon, the square root of g is denoted g :
g = g g . (1.99)
In (Comment: where?) we demonstrate that we have

i
g = g
is
s
. (1.100)
Using expression (Comment: which one?) for the (covariant) derivative of a scalar density, this simply
gives

i
g =
i
g g
is
s
= 0 , (1.101)
which is consistent with the fact that

i
g
jk
= 0 . (1.102)
We can also dene the determinant of g
ij
:
g = det(g
ij
) =
1
n!

ijk...

pqr...
g
ip
g
jq
g
kr
. . . , (1.103)
and its square root g :
g = g g . (1.104)
As the matrices g
ij
and g
ij
are mutually inverses, we have
g g = 1 . (1.105)
Using the scalar density g and the scalar capacity g we can associate tensor densities, pure
tensors, and tensor capacities. Using the same letter to designate the objects related through this
natural bijection, we will write expressions like
= g , (1.106)
V
i
= g V
i
, (1.107)
or
T
ij...
kl...
= g T
ij...
kl...
. (1.108)
1.5. INTEGRATION, ESTIMATION OF DENSITIES 37
So, if g
ij
and g
ij
can be used to lower and raise indices, g and g can be used to put and
remove bars.
Comment: say somewhere that g is the density of volumetric content, as the volume element of
a metric space is given by
dV = g d , (1.109)
where d is the capacity element dened in (Comment: where?), and which, when we take an element
along the coordinate lines, equals dx
1
dx
2
dx
3
. . . .
Comment: Say that we can demonstrate that, in an Euclidean space, the matrix representing the
metric equals the product of the Jacobian matrix times the transposed matrix:
g
ij
=
_
_
_
g
11
g
12
. . .
g
21
g
22
. . .
.
.
.
.
.
.
.
.
.
_
_
_ =
_
_
_
_
X
1
x
1
X
1
x
2
. . .
X
2
x
1
X
2
x
2
. . .
.
.
.
.
.
.
.
.
.
_
_
_
_

_
_
_
_
X
1
x
1
X
2
x
1
. . .
X
1
x
2
X
2
x
2
. . .
.
.
.
.
.
.
.
.
.
_
_
_
_
. (1.110)
In short,
g
ij
=

K
X
K
x
i
X
K
x
j
. (1.111)
This follows directly from the general equation
g
ij
=
X
I
x
i
X
J
x
j
g
IJ
(1.112)
using the fact that, if the X
I
are Cartesian coordinates,
g
IJ
=
_
_
_
g
11
g
12
. . .
g
21
g
22
. . .
.
.
.
.
.
.
.
.
.
_
_
_ =
_
_
1 0 . . .
0 1 . . .
0 0 . . .
_
_
. (1.113)
1.4.2 The scalar product
Comment: explain here that the metric introduces a bijection between forms and vectors:
V
i
= g
ij
V
j
. (1.114)
Comment: introduce here the notation
(V, W) = g
ij
V
i
W
j
= V
i
W
i
= W
i
V
i
. (1.115)
1.5 Integration, estimation of densities
Comment: explain here what the capacity elementis. Explain that, in polar coordinates, it is given
by drd , to be compared with the surface element rdrd . Comment gure 1.9.
Bijection between densities, tensors, and capacities
Comment: the text below has already been written in the section where the metric was introduced.
We have seen that the metric can be used to dene a natural bijection between forms and vectors.
Let us now see that it can also be used to dene a natural bijection between tensors, densities, and
capacities.
38 CHAPTER 1. TENSOR FIELDS
-1 -0.5 0 0.5 1
-1
-0.5
0
0.5
1
-4 -2 0 2 4
-4
-2
0
2
4
=
= /2
= 3/2
= 0
= 0
=
= 2
r = 0 r = 1 r = 1/2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
. .
.
.
.
.
.
. . . .
.
.
.. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 1.9: We consider, in an Euclidean space, a cylinder with a circular basis of radius 1, and
cylindrical coordinates (r, , z) . Only a section of the cylinder is represented in the gure, with
all its thickness, dz , projected on the drawing plane. At left, we have represented a map of the
corresponding circle, and, at right, the coordinate lines on the circle itself. All the cells at left have
the same capacity dV = drddz , while the cells at right have the volume dV (r, , z) = rdrddz . The
points represent particles with given masses. If, at left, at point with coordinates (r, , z) the sum of
all the masses inside the local cell is denoted, dM , then, the mass density at this point is estimated by
(r, , z) = dM/dV , i.e., (r, ) = dM/(drddz) . If, at right, at point (r, , z) the total mass inside
the local cell is dM , the volumetric mass at this point is estimated by (r, , z) = dM/dV (r, , z) , i.e.,
(r, , z) = dM/(rdrddz) . By denition, then, the total mass inside a volume V will be found by
M =
_
V
dV (r, , z) =
_
V
drddz(r, , z) or by M =
_
V
dV (r, , z)(r, , z) =
_
V
rdrddz(r, , z) .
We denote by g the determinant of g
ij
:
g = det(g
ij
) =
1
n!

ijk...

pqr...
g
ip
g
jq
g
kr
. . . . (1.116)
The two upper bars recall that g is a second order density, as there is the product of two densities
at the right-hand side.
For a reason that will become obvious soon, the square root of g is denoted g :
g = g g . (1.117)
In (Comment: where?) we demonstrate that we have

i
g = g
is
s
. (1.118)
Using expression (Comment: which one?) for the (covariant) derivative of a scalar density, this simply
gives

i
g =
i
g g
is
s
= 0 , (1.119)
which is consistent with the fact that

i
g
jk
= 0 . (1.120)
We can also dene the determinant of g
ij
:
g = det(g
ij
) =
1
n!

ijk...

pqr...
g
ip
g
jq
g
kr
. . . , (1.121)
and its square root g :
g = g g . (1.122)
1.6. THE LEVI-CIVITAS AND THE KRONECKERS TENSORS 39
capacity tensor density
0-rank s

= s s

= s s

s
1-form F

I
= J

I
i
F
i
F

I
= J

I
i
F
i
F

I
=

I
i
F
i
1-vector V
I
= V
i
J
i
I
V
I
= V
i
J
i
I
V
I
=

V
i
J
i
I
2-form Q

IJ
= J

I
i
J

J
j
Q
ij
Q

IJ
= J

I
i
J

J
j
Q
ij
Q

IJ
=

I
i
J

J
j
Q
ij
(1-form)-(1-vector) R

I
J
= J

I
i
R
i
j
J
j
J
R

I
J
= J

I
i
R
i
j
J
j
J
R

I
J
=

I
i
R
i
j
J
j
J
(1-vector)-(1-form) S
I
J
= J
i
I
S
i
j
J

J
j
S
I
J
= J
i
I
S
i
j
J

J
j
S
I
J
=

J
i
I
S
i
j
J

J
j
2-vector T
IJ
= T
ij
J
i
I
J
j
J
T
IJ
= T
ij
J
i
I
J
j
J
T
IJ
=

T
ij
J
i
I
J
j
J
.
.
.
.
.
.
.
.
.
.
.
.
Table 1.1: Comment: write here the legend
As the matrices g
ij
and g
ij
are mutually inverses, we have
g g = 1 . (1.123)
Using the scalar density g and the scalar capacity g we can associate tensor densities, pure
tensors, and tensor capacities. Using the same letter to designate the objects related through this
natural bijection, we will write expressions like
= g , (1.124)
V
i
= g V
i
, (1.125)
or
T
ij...
kl...
= g T
ij...
kl...
. (1.126)
So, if g
ij
and g
ij
can be used to lower and raise indices, g and g can be used to put and
remove bars.
Comment: say somewhere that g is the density of volumetric content, as the volume element of
a metric space is given by
dV = g d , (1.127)
where d is the capacity element dened in (Comment: where?), and which, when we take an element
along the coordinate lines, equals dx
1
dx
2
dx
3
. . . .
1.6 The Levi-Civitas and the Kroneckers tensors
1.6.1 The Levi-Civitas tensor
There are two dierent Levi-Civitas symbols:
ijk...
and
ijk...
. The number of indices they contain
equals the dimension of the space. For instance, in spaces of dimension 1, 2, 3, and 4 , one of the
Levi-Civitas tensors will be written, respectively,
i
,
ij
,
ijk
, and
ijkl
.
They are dened by:

ijk...
=
_

_
+ 1 if ijk . . . is an even permutation of 12 . . . n
0 if some indices are identical
1 if ijk . . . is an odd permutation of 12 . . . n,
(1.128)
40 CHAPTER 1. TENSOR FIELDS
and, equivalently,

ijk...
=
_

_
+ 1 if ijk . . . is an even permutation of 12 . . . n
0 if some indices are identical
1 if ijk . . . is an odd permutation of 12 . . . n.
(1.129)
In fact,
ijk...
and
ijk...
are more than symbols: they are respectively a capacity and a
density, in the sense that, if when changing the coordinates, we compute the new components of
the Levi-Civitas capacity and density using the rules applying to all capacities and densities, the
properties 1.1281.129 remain satised.
Comment: be more explicit. Say that if
ijk...
satises the property 1.128, then, in a change of
coordinates, the transformed object

IJK...
=

I
i
J

J
j
J

K
k
. . .
ijk...
(1.130)
also satises 1.128.
Comment: say that an exercice gives the direct demonstration in 2-D.
Commment: give somewhere the general demonstration.
Notice that the Levi-Civitas capacity
ijk...
has been dened with lower indices, and the Levi-
Civitas density
ijk...
with upper indices. Raising the indices of
ijk...
gives
ijk...
, but does not give

ijk...
: there are two dierent objects.
We have just dened the Levi-Civitas density and capacity. Both are simple in what they take only
the values +1, 0, 1 in any coordinate system. A pure tensor can be dened through the canonical
bijection:

ijk...
= g
ijk...
. (1.131)
The reader should remember that this pure tensor, instead of the values +1, 0, 1 it takes the
values +g, 0, g .
1.6.2 Determinants
The Levi-Civitas tensors can be used to dene determinants. For instance, the determinants of the
tensors Q
ij
, R
i
j
, S
i
j
, and T
ij
are dened by
Q =
1
n!

ijk...

mnr...
Q
im
Q
jn
Q
kr
. . . , (1.132)
R =
1
n!

ijk...

mnr...
R
i
m
R
j
n
R
k
r
. . . ,
=
1
n!

ijk...

mnr...
R
i
m
R
j
n
R
k
r
. . . , (1.133)
S =
1
n!

ijk...

mnr...
S
i
m
S
j
n
S
k
r
. . . ,
=
1
n!

ijk...

mnr...
S
i
m
S
j
n
S
k
r
. . . , (1.134)
and
T =
1
n!

ijk...

mnr...
T
im
T
jn
T
kr
. . . , (1.135)
where the Levi-Civitas tensors
ijk...
,
ijk...
,
ijk...
and
ijk...
have as many indices as the space
under consideration has dimensions.
1.6. THE LEVI-CIVITAS AND THE KRONECKERS TENSORS 41
1.6.3 The Kroneckers tensor
There are two Kroneckers symbols, g
i
j
and g
i
j
. They are dened similarly:
g
i
j
=
_
1 if i and j are the same index
0 if i and j are dierent indices ,
(1.136)
and
g
i
j
=
_
1 if i and j are the same index
0 if i and j are dierent indices .
(1.137)
Comment: I should be avoid this last notation.
It can easily be seen
(Comment: how?)
that g
i
j
are more than symbols: they are tensors, in the sense that, if when changing the
coordinates, we compute the new components of the Kroneckers tensors using the rules applying to
all tensors, the property (Comment: which equation?) remains satised.
The Kroneckers tensors are dened even if the space has not a metric dened on it. Note that,
sometimes, instead of using the symbols g
i
j
and g
j
j
to represent the Kroneckers tensors, the
symbols
i
j
and
j
j
are used. But then, using the metric g
ij
to lower an index of
i
j
gives

ij
= g
jk

i
k
= g
ij
, (1.138)
which means that, if the space has a metric, the Kroneckers tensor and the metric tensor are the same
object. Why, then, use a dierent symbol? The use of the symbol
i
j
may lead, by inadvertence,
after lowering an index, to assing to
ij
the value 1 when i and j are the same index. This is
obviously wrong: if there is not a metric,
ij
is not dened, and if there is a metric,
ij
equals g
ij
,
which is only 1 in Euclidean spaces using Cartesian coordinates.
There is only one Kroneckers tensor, and g
i
j
and g
i
j
can be deduced one from the other raising
and lowering indices. But, even in that case, we dislike the notation g
i
j
, where the place of each index
is not indicated, and we will not use it sistematically.
Warning: a common error in beginners is to give the value 1 to the symbol g
i
i
(or to g
i
i
) .
In fact, the right value is n , the dimension of the space, as there is an implicit sum assumed:
g
i
i
= g
0
0
+g
1
1
+. . . = 1 + 1 +. . . = n .
1.6.4 The Kroneckers determinants
Let us denote by n the dimension of the space into consideration. The Levi-Civitas tensor has then
n indices. For any (non-negative) integer p satisfying p n , consider the integer q such that
p +q = n . The following property holds:

i
1
...ips
1
...sq

j
1
...jps
1
...sq
= q! det
_
_
_
_
_
_

j
1
i
1

j
2
i
1
. . .
jp
i
1

j
1
i
2

j
2
i
2
. . .
jp
i
2
.
.
.
.
.
.
.
.
.
.
.
.

j
1
ip

j
2
ip
. . .
jp
ip
_
_
_
_
_
_
, (1.139)
where
i
j
stands for the Kroneckers tensor. The determinant at the right-hand side is called the
Kroneckers determinant, and is denoted
j
1
j
2
...jp
i
1
i
2
...ip
:

j
1
j
2
...jp
i
1
i
2
...ip
= det
_
_
_
_
_
_

j
1
i
1

j
2
i
1
. . .
jp
i
1

j
1
i
2

j
2
i
2
. . .
jp
i
2
.
.
.
.
.
.
.
.
.
.
.
.

j
1
ip

j
2
ip
. . .
jp
ip
_
_
_
_
_
_
. (1.140)
42 CHAPTER 1. TENSOR FIELDS
As the Kroneckers determinant is dened as a product of Levi-Civitas tensors, it is itself a tensor. It
generalizes the denition of the Kroneckers tensor
i
j
, as it has the properties

j
1
j
2
...jm
i
1
i
2
...im
=
_

_
+1 if (j
1
, j
2
, . . . , j
m
) is an even permutation of (i
1
, i
2
, . . . , i
m
)
1 if (j
1
, j
2
, . . . , j
m
) is an odd permutation of (i
1
, i
2
, . . . , i
m
)
0 if two of the i

s or two of the j

s are the same index

0 if (i
1
, i
2
, . . . , i
m
) and (j
1
, j
2
, . . . , j
m
) are dierent sets of indices .
(1.141)
As applying the same permutation to the indices of the two Levi-Civitas tensors of equation 1.139
will not change the total sign of the expression, we have

i
1
...ips
1
...sq

j
1
...jps
1
...sq
=

s
1
...sqi
1
...ip

s
1
...sqj
1
...jp
= q!
j
1
j
2
...jp
i
1
i
2
...ip
, (1.142)
but we only perform a permutation in one of the Levi-Civitas tensors, then we must care about the
sign of the permutation, and we obtain

i
1
...ips
1
...sq

s
1
...sqj
1
...jp
=

s
1
...sqi
1
...ip

j
1
...jps
1
...sq
= (1)
pq
q!
j
1
j
2
...jp
i
1
i
2
...ip
. (1.143)
This possible change of sign has only eect in spaces with even dimension (n = 2, 4, . . .) , as in spaces
with odd dimension (n = 3, 5, . . .) the condition p +q = n implies that pq is an even number, and
(1)
pq
= +1 .
Remark that a multiplication and a division by g will not change the value of an expression, so
that, instead of using Levi-Civitas density and capacity we can use Levi-Civitas true tensors. For
instance,

i
1
...ips
1
...sq

j
1
...jps
1
...sq
=
i
1
...ips
1
...sq

j
1
...jps
1
...sq
. (1.144)
Comment: explain better.
In Boxes 1.3 to 1.5 we give special formulas to spaces with dimension 2 , 3 , and 4 . As shown in
appendix XXX, these formulas replace more elementary identities between grad, div, rot, . . .
As an example, a well known identity like
a (b c) = b (c a) = c (a b) (1.145)
is obvious, as the three formulas correspond to the expression
ijk
a
i
b
j
c
k
. The identity
a (b c) = (a c) b (a b) c (1.146)
is easily demonstrated, as
a (b c) =
ijk
a
j
(b c)
k
=
ijk
a
j

km
b

c
m
, (1.147)
which, using 1.157, gives
a (b c) = (a
m
c
m
)b
i
(a
m
b
m
)c
i
= (a c) b (a b) c . (1.148)
Comment: I should clearly say here that we have the identity

ijk...

mn...
=
ijk...

mn...
. (1.149)
Comment: say somewhere that if B
i
1
...ip
is a totally antisymmetric tensor, then
1
p!

1
...p
i
1
...ip
B

1
...p
= B
i
1
...ip
(1.150)
1.6. THE LEVI-CIVITAS AND THE KRONECKERS TENSORS 43
Box 1.3 The Kroneckers determinants in 2-D

k
ij
= (1/0!)
ij

k
=
k
i

k
j
(1.153)

k
j
= (1/1!)
ij

ik
=
k
j
(1.154)
= (1/2!)
ij

ij
= 1 (1.155)
Box 1.4 The Kroneckers determinants in 3-D

mn
ijk
= (1/0!)
ijk

mn
=

m
j

n
k
+
m
i

n
j

k
+
n
i

m
k

n
j

m
k

n
i

m
j

k

m
i

n
k
(1.156)

m
jk
= (1/1!)
ijk

im
=

m
k

m
j

k
(1.157)

k
= (1/2!)
ijk

ij
=

k
(1.158)
= (1/3!)
ijk

ijk
= 1 (1.159)
Comment: give somewhere the property
1
q!

k
1
...kp
1
...q
i
1
...ipj
1
...jq

j
1
...jq
m
1
...mq
=
k
1
...kp
1
...q
i
1
...ipm
1
...mq
. (1.151)
Comment: give somewhere the property
1
q!

i
1
...ipj
1
...jq

j
1
...jq
k
1
...kq
=
i
1
...ipk
1
...kq
. (1.152)
Note: Check if there are not factors (1)
pq
missing.
44 CHAPTER 1. TENSOR FIELDS
Box 1.5 The Kroneckers determinants in 4-D

mnpq
ijk
= (1/0!)
ijk

mnpq
= +
m
i

n
j

p
k

+
m
i

p
j

q
k

+
m
i

q
j

n
k

+
n
i

q
j

p
k

+
n
i

p
j

m
k

q

+
n
i

m
j

q
k

+
p
i

q
j

m
k

n

+
p
i

m
j

n
k

+
p
i

n
j

q
k

+
q
i

m
j

p
k

+
q
i

n
j

m
k

p

+
q
i

p
j

n
k

m
i

n
j

q
k

m
i

p
j

n
k

m
i

q
j

p
k

n
i

p
j

q
k

n
i

q
j

m
k

p

n
i

m
j

p
k

p
i

q
j

n
k

p
i

m
j

q
k

p
i

n
j

m
k

q

q
i

m
j

n
k

q
i

n
j

p
k

q
i

p
j

m
k

n

(1.160)

mnp
jk
= (1/1!)
ijk

imnp
=
m
j

n
k

+
n
j

p
k

+
p
j

m
k

n

m
j

p
k

n
j

m
k

p

p
j

n
k

(1.161)

mn
k
= (1/2!)
ijk

ijmn
= (
m
k

n

n
k

) (1.162)

= (1/3!)
ijk

ijkm
=
m

(1.163)
= (1/4!)
ijk

ijk
= 1 (1.164)
Chapter 2
Derivatives of tensors
Note: rewrite this text. . .
Besides the necessary introduction of the notations, the aim of the chapter is to demonstrate
that the tensors describing curvature and torsion satisfy, by denition, certain identitiesthe Bianchi
identities. In the 4-D space-time, after postulating a relationship between the tensors describing
space-time curvature and torsion and the tensors describing mass and spinthe Einstein-Cartan
equations, the Bianchi identities lead to the dierential equations governing the dynamics of con-
tinuous media.
45
46 CHAPTER 2. DERIVATIVES OF TENSORS
2.1 Parallel transportation
Transporting a vector parallel to itself is trivial in an Euclidean space, but not so in a general space.
The disturbing property of non-Euclidean spaces is that the vector we obtain by parallel transportation
between two points depends on the path followed. This is obvious, for instance, when transporting
vectors at the surface of a sphere (gure 2.1 shows an example).
2.1.1 Example of parallel transportation in a metric space
In a metric space, i.e., in a space where the notion of length of a line exists, it is not dicult to dene
the notion of parallel transport of a vector.
First, we need to introduce the notion of geodesic line: a line joigning two points is a geodesic
line (for short: a geodesic) if all other neighbouring lines joining the two points have greater length.
For instance, the great circle joigning two points in a sphere is the shortest of all neighbouring lines
joining the two points: great circles on a sphere are geodesics.
Second, we need to remark that if the notion of length of a line exists, then we also have the notion
of angle, as the angle (in radians) betwen two directions is just a ratio of two lengths. We will later
see more precise characterizations of lengths and angles.
Then, the transportation of a vector parallel to itself along a geodesic gives, by denition, the
vector with constant length that makes a constant angle with the geodesic.
We will see below how the parallel transportation of a vector along a line which is not a geodesic
may be dened. If the path is not specied, the parallel transportation of a vector between two points
is made along the geodesic joining the two points.
In a space with curvature, the result of parallel transport of a vector between two points depends
on the path followed: in a sphere, transporting a vector parallel to itself from the Equator to the
Pole, then from the Pole to another point in the Equator, does not give the same vector as when
transporting the original vector directly along the Equator (note that a path made by two geodesics
that do not join straightly is not a geodesic path).
Let us familiarize more with the parallel transport of a vector at the surface of a sphere.
Consider a vector at the North geographic pole of the Earth, and let us transport it, parallel to
itself, to all other points of the Earth. As no particular path is specied, let us transport it along the
geodesics of the sphere (the the great circles). The great circles joining the North pole to any other
point of the sphere are the Meridians. If the given vector at the North pole points towards, say, the
Meridian with longitude = 180

(the line of change of date), it makes with any Meridian (at the
Pole) an angle equal to the longitude of the Meridian. Then, the vector obtained by parallel transport
along any Meridian will conserve this angle. This gives the vector eld represented in gure 2.2. All the
vectors of the gure are tangent to the sphere and have the same length (the orthographic projection
used here gives, of course, dierent projected lengths). Notice that the vector obtained on a sphere by
parallel transport of a vector at a given point on the sphere is uniquely dened everywhere excepted
at the antipodal point: it is possible to go from one point to the antipodal point along dierent great
circles, each one giving a dierent parallel-transported vector.
The components of this vector eld are, in the natural basis,
V

(, ) =
V
0
R
cos V

(, ) =
V
0
R
sin
sin
. (2.1)
It is easy to verify (using the notion of metric tensor to be introduced below) that this corresponds to
a vector eld with constant norm |V| = V
0
, and making with any Meridian an angle equal to the
longitude of the Meridian.
2.1. PARALLEL TRANSPORTATION 47
Figure 2.1: An illustration of a parallel transport of vectors. Left: a vector at the Equator, pointing
to the North Pole, is transported, parallel to itself, to the Pole, then it is transported, parallel to
itself, to the equator. Right: the same vector is transported, parallel to itself, to the same point,
following a dierent path. The two vectors so obtained are not parallel to each other: the result of
parallel transportation between two points depends, in general, on the path followed. Here, if the
two Meridians followed during the parallel transportation are distant of 90 degrees, the two vectors
obtained by parallel transportation of the same vector are perpendicular to each other.
Even if this eld is dened by paralel transportation on the sphere from a single vector and, this
should not be called a constant eld. For its divergence (to be dened below) is nonzero,
divV =
V
0
R
cos
2 sin
2
2
sin
, (2.2)
excepted at the pole itself (as it should, as very close to the pole we can replace the sphere by the
tangent plane, and the parallel transport of a vector on an Euclidean plane gives a constant vector
eld).
The denition of parallel transportat of a vector along an arbitrary line, not necessarily a geodesic,
requires geodetic notions, i.e., the notions that schools of physical geography teach to geodesists.
Consider that we have two directions dened at a point, i.e., two lines (not necessarily geodesics)
intersecting at a point. Point O in gure 2.3 represents the intersecting point, and points A and
B are points along each of the lines (the limit will be taken when these points tend to point O ). If
ACB is a geodesic with the length of AC equal to the length of CB , and OCD is a geodesic with
the length of OC equal to the length of CD , then, by denition, in the limit where points A and
B tend to point O , the direction BD is parallel to the direction OA , and the direction AD is
parallel to the direction OB .
Comment: explain why I have drawn small straight lines in gure 2.3.
Comment: gures 2.4 and ?? just happen to be there.
We will see below that if we know how to transport a vector, then we know how to transport a
tensor of any order.
So far, for the paralle transport of a vector in metric spaces. Let us now turn to an example of
parallel transport in spaces where the notion of distance is not introduced.
2.1.2 Example of parallel transportation in a space without metric
If the manifold where we work is the physical space (or a surface embedded on the physical space), we
may introduce the notion of metric (only very nonintuitive theories of the space-time do not assume
the possibility of measuring distances or durations).
But we may need to develop methematics in a space which is not the physical space, and in which
the notion of distance may not be dened. Let us see an example.
Assume an electrical conductor whose state we characterize by two parameters: its electrical
conductivity, c , and the velocity v of some acoustic waves traveling inside the conductor. Assume
48 CHAPTER 2. DERIVATIVES OF TENSORS
. .
Figure 2.2: A vector at the North geographic pole of the Earth is transported, parallel to itself, to
all other points of the Earth. Unless otherwise specied, the parallel transportation from one point
to another point is made along a geodesic (here a great circle). The geodesics joining the North
pole to any other point are the Meridians. The given vector at the North pole pointing towards the
Meridian with longitude = 180

, it makes, at the North pole, with any Meridian an angle equal

to the longitude of the Meridian. Then, the vector obtained by parallel transport along any Meridian
will conserve this angle, and this gives the vector eld represented in the gure. All the vectors are
tangent to the sphere and have the same length (the orthographic projection used here gives, of course,
dierent projected lengths). Notice that the vector obtained on a sphere by parallel transport of a
vector at a given point on the sphere is uniquely dened everywhere excepted at the antipodal point:
it is possible to go from one point to the antipodal point along dierent great circles, each one giving
a dierent parallel-transported vector.
2.1. PARALLEL TRANSPORTATION 49
A
B
O
C
D
A
B
O
C
D
A
B
O
C
A
B
O
Figure 2.3: Consider three points O, A, B. In the limit where points A and B tend to point O, the
lines OA and OB dene two directions. We wish to transport the direction OA, parallel to itself, from
point O to point B. This is accomplished by, rst, locating the unique point C such that the distance
AC equals the distance CB and that the line ACB is geodesic (i.e., no other point would give a smaller
value for the distance AC plus the distance CB). Once that point is located, the unique point D is
dened similarly: the distance OC equals the distance CD and the line OCD is geodesic. In the limit
whwre all the distances tend to zero, the direction BD is, by denition, the direction obtained by
parallel transport of the direction OA (and the direction AD is the direction line obtained by parallel
transport of the direction OB). We have simply built a parallelogram. (Comment: is the length BD
equal to the length OA? If yes, we do not need the point C and we can locate directly the point D)
A
O
B
D
Figure 2.4: The method dened in 2.3 can be iterated to transport a direction along a nite arbitrary
line OB. Comment: how can I demonstrate that the paralle transport along a geodesic conserves the
angles?
50 CHAPTER 2. DERIVATIVES OF TENSORS
that all the other parameters we are interested in are simply functions of the state of the system, i.e.,
functions of c , and v .
For instance, when the system is at point (c
0
, v
0
) , we may measure the rate of increase of
conductivity, c , and the rate of increase of the velocity of the acoustic waves, v . The values ( c, v)
can be seen as the components of a vector at point (c
0
, v
0
) , of the conductivity-velocity space (see
gure 2.5).
If we want to develop some mathematics in the conductivity-velocity space, we may need to dene
the (covariant) derivatives of some parameter with respect to the conductivity or the velocity.
But, as we will see below, there is no proper denition of derivatives without denition of parallel
transportation of vectors. What then may be the vector obtained at any point by parallel transport
of the vector ( c, v) at point (c
0
, v
0
) ? The question is graphically illustrated at the bottom of
gure 2.5. The suggested answer, that the vector translated keeps its components constant, can not
be the answer. Let us see why.
There are in physics many physical parameters with this double property:
1. The parameter may only take positive values.
2. One indierently uses the parameter or its inverse.
Here are some examples:
Conductivity c and resistivity = 1/c .
Velocity v and slowness s = 1/v .
Period (of a periodical phenomenon) T and frequency = 1/T .
Half-life of a radioactive nucleus, T and the desintegration rate = . . . .
Mass density and mass lightness = 1/ .
Temperature T and thermodynamic parameter = 1/kT ( k is the Boltzmanns constant).
Comment: mention here Jaynes (1968).
Thus, if we have chosen as parameters the conductivity and the velocity (of acoustic waves),
another physicist may choose the resistivity and the slowness, and there is no way of deciding which
choice is more natural.
But the denition of parallel transportation guessed at the bottom of gure 2.5 is not consistent
with that freedom of choice for the parameters: Should the physicist working with resistivity and
slowness also attempt to dene the parallel transportation of a vector by keeping the components
constant, she/he would afound another vector at the nal point.
Let us be more precise. The top of gure 2.6 shows some vectors (with constant components) on
the plane conductivity-velocity. Using the relation between resistivity and conductivity, = 1/c , and
between slowness and velocity, s = 1/v , we can map this gure into the gure at bottom. Should
one be tempted to consider that the vectors at the top gure are all parallel, this would not be true
for the vectors at the bottom gure.
Of course, all this is reciprocal for the alternate choice of parameters. The assumption that a
set of vectors in the resistivity-slowness plane with constant components is a set of parallel vectors
(top of gure 2.7) is not consistent with the gure obtained by coordinate transformation(bottom of
gure 2.7).
Comment: explain here the Jaynes argument. It leads to the conclusion that the set of positive
parameters described above have some fundamental properties. For instance, the null information
2.1. PARALLEL TRANSPORTATION 51
probability density is not constant, but of the form 1/x , they do not usually show normal probability
densities but log-normal ones, etc.
Much simpler, then, that those parameters, are their logarithms. While, for instance, the null
information probability density for a period T is
f(T) =
const.
T
, (2.3)
the null information probability density for the logarithmic period, T

= log(T/T
0
) ( T
0
arbitrary),
is (as deduced from the rules of change of variables in probability theory) constant:
f

(T

) = const. (2.4)
Also, if a period shows often log-normal normal probability densities, the logarithmic period shows
normal ones.
Comment: say somewhere that the form of a probability density is not changed when passing from,
say, a logarithmic period, to a logarithmic frequency.
The simplicity of the logarithmic parameters (for positive parameters) with respect to probability
density functions is kept for the problem of parallel trabsport.
If physicist A, who prefers to work with conductivity and velocity, and physicist B, who prefers
to work with resisitivity and slowness, both agree that parallel transportation can only be dened
with simplicity in logarithmic parameters, then all problems disappear. Figure 2.8 illustrates this:
The set of (truly) paralle vectors at top, in the logarithmic plane conductivity-velocity, maps, through
the transformation =
1
c
, and s =
1
v
, into the set of (truly) parallel vectors at bottom, in the
logarithmic plane resistivity-slowness, as the logarithmic parameters are related by

= c

and
s

= v

. Comment: explain all this much better.

Thus we have seen a set of nontrivial rules dening the parallel transportation of vectors in a space
where the notion of metric does not exist (for what is the distance between two points?).
Comment: mention somewhere that all this is compatible with the choice of metric
_
g
cc
g
cv
g
vc
g
vv
_
=
_
1/c
2
0
0 1/v
2
_
. (2.5)
2.1.3 Important Notation
Comment: say what follows elsewhere.
In this section we will pave the way for the denition of covariant derivatives, i.e., proper
denitions of derivatives that, when applied to tensor elds, give tensor elds. The fundamental
concept at the base of the denition of a tensor derivative is that of parallel transport of vectors.
I will introduce very explicit notations, as there are some technical details which are nontrivial. In
particular, we will see that, when manipulating tensor elds, we may wish to introduce some elds
that greatly simplify calculations, but which are not tensor elds (the connection is one of them).
Comment: in one of the previous sections, I should make explicit that, to any point of a curve, we
can associate the tangent geodesic. Then, when I consider to neighbouring points along a curve, it is
unimportant to consider that the points are along the given curve or along the tangent geodesic.
Let x V(x) be a tensor eld associating to any point x of the space the vector V(x) . In
this section, we will not simplify the discourse to just say the vector eld V(x) . (Or yes?).
As discussed above (section XXX), unless another choice is explicitly stated, the parallel trans-
portation of the vector V from point x
A
to point x
B
will mean along the geodesic joining the
points x
A
and x
B
.
The following notation is not common, and is very important for the denition of derivatives. Let
x V(x) be a vector eld, and let y be one particular point of the space. The value of the vector
52 CHAPTER 2. DERIVATIVES OF TENSORS
V
e
l
o
c
i
t
y
Conductivity (c)
(v)
c
0
v
0
v
.
c
.
V
e
l
o
c
i
t
y
Conductivity (c)
(v)
c
0
v
0
v
.
c
.
?
Figure 2.5: To be written.
eld at point y is V(y) . When transporting the vector V(y) , parallel to itself, from point y to
another point x , we will, by denition, obtain a vector at point x , and there is no reason that the
vector so obtained at point x equals V(x) , unless the vector eld is constant. This is illustrated in
gure 2.9. The vector obtained at point x by parallel transport from point y to point x of the
vector V(y) will be denoted V(x | y) .
For given point y , the expression x V(x | y) denes a vector eld which depends on the
value of the vector eld x V(x) at the single point y , but that contains no information about
the values of this vector eld at other points (comment: very confusing...). See gure 2.9
2.1.4 Parallel transport of the basis vectors
The (natural) basis vectors at any point x have been dened in section XXX, where they have been
denoted by e
i
(x) . In the preceding section we have introduced the notation V(x | y) to denote
the vector obtained at point x by parallel transport of the vector V(y) . Then, by e
i
(x | y) we
will denote the vectors obtained at point x by parallel transport of the basis e
i
(y) . Of course, the
vectors obtained by transporting the basis from point y to point x , e
i
(x | y) , will not, in general,
equal the basis vectors at x , e
i
(x) .
At point x , the transported basis vectors, e
i
(x | y) , will have some components on the local
basis e
i
(x) , that will be denoted
i
j
(x[y) :
e
i
(x | y) =
i
j
(x[y) e
j
(x) . (2.6)
The coecients for the parallel transportation of the dual basis will be denoted
i
j
(x[y) :
e
i
(x | y) = e
j
(x)
j
i
(x[y) . (2.7)
2.1. PARALLEL TRANSPORTATION 53
1 2 3 4 5 6 0
1
2
3
4
0
A B
C D
0.5 1.0 1.5 0
0.5
1.0
1.5
0
A B
C D
Conductivity
Resistivity
V
e
l
o
c
i
t
y
S
l
o
w
n
e
s
s
Figure 2.6: To be written. Comment: explain here what are the points A, B, C, and D. Comment:
explain that there are units, not written.
54 CHAPTER 2. DERIVATIVES OF TENSORS
1 2 3 4 5 6 0
1
2
3
4
0
A B
C D
0.5 1.0 1.5 0
0.5
1.0
1.5
0
A B
C D
Resistivity
V
e
l
o
c
i
t
y
S
l
o
w
n
e
s
s
Conductivity
Figure 2.7: To be written.
2.1. PARALLEL TRANSPORTATION 55
A B
C D
L
o
g
a
r
i
t
h
m
i
c
S
l
o
w
n
e
s
s
A B
C D
0
1
-1
0
1
-1
1 2 3 0 -1
-3 -2 -1 0 1
Logarithmic Conductivity
V
e
l
o
c
i
t
y
L
o
g
a
r
i
t
h
m
i
c
Logarithmic Resistivity
Figure 2.8: To be written.
V(x)
V(y)
V(x)
V(y)
V(x||y)
Figure 2.9: The notation described here is not common, and is very important for the denition of
derivatives. Let x V(x) be a vector eld, and let y be one particular point of the space. The
value of the vector eld at point y is V(y) . When transporting the vector V(y) , parallel to itself,
from point y to another point x , we will, by denition, obtain a vector at point x , and there is
no reason that the vector so obtained at point x equals V(x) , unless the vector eld is constant.
The vector obtained at point x by parallel transport from point y to point x of the vector V(y)
will be denoted V(x | y) . For a given point y , the expression x V(x | y) denes a vector eld
which depends on the value of the vector eld x V(x) at the single point y , but that contains
no information about the values of this vector eld at other points.
56 CHAPTER 2. DERIVATIVES OF TENSORS
As transporting a vector or a form from point x to point x along the geodesic means no
transportation at all, we have

i
j
(x[x) = g
i
j
=
i
j

i
j
(x[x) = g
i
j
=
i
j
. (2.8)
As the transportation of a basis and of the dual basis must preserve their duality character, we
will have
e
i
(x | y) e
j
(x | y) = g
i
j
=
i
j
, (2.9)
i.e.,

i
j
(x[y)
j
k
(x[y) = g
i
j
=
i
j

i
j
(x[y)
j
k
(x[y) = g
i
j
=
i
j
(2.10)
(the matrix representing
i
j
and the matrix representing
i
j
are mutually inverses). Comment:
say somewhere that, in fact, these equations simply mean that that the parallel transportation of the
Kroneckers tensor gives the Kroneckers tensor.
Comment: I have to nd how to espress
i
j
(x[y) as a function of the metric, when there is one.
Comment: Old demonstration. Using
e
i
(x | y) e
k
(x | y) =
j
i
(x[y)
k
m
(x[y) e
j
(x) e
m
(x) , (2.11)
e
i
(x | y) e
k
(x | y) = g
i
k
(2.12)
and
e
j
(x) e
m
(x) = g
j
m
, (2.13)
we deduce

k
j
(x | y)
j
i
(x | y) = g
k
i
. (2.14)
In fact, we have simply transported the Kroneckers tensor.
Example: In an Euclidean space with Cartesian coordinates, transporting a vector parallel to itself
gives, on the local basis at the nal point, the same components than on the local basis at the original
point. Then,

i
j
(x[y) = g
i
j
=
i
j
. (2.15)

Example: In polar coordinates, let e

r
(r
0
,
0
), e

(r
0
,
0
) denote the natural basis at point
(r
0
,
0
) , and let e
r
(r, | r
0
,
0
), e

(r, | r
0
,
0
) denote the vectors obtained by parallel trans-
port of e
r
(r
0
,
0
), e

(r
0
,
0
) from point (r
0
,
0
) to point (r, ) . It is quite easy to express the
transported vectors on the local basis, as there is only a rotation and a change of length of the vector
e

. We have (see gure 2.10)

_
e
r
(r, | r
0
,
0
)
e

(r, | r
0
,
0
)
_
=
_
cos(
0
)
1
r
sin(
0
)
r
0
sin(
0
)
r
0
r
cos(
0
)
_ _
e
r
(r, )
e

(r, )
_
. (2.16)
This gives, for the coecients
i
j
,
_

r
r
(r, [r
0
,
0
)
r

(r, [r
0
,
0
)

r
(r, [r
0
,
0
)

(r, [r
0
,
0
)
_
=
_
cos(
0
)
1
r
sin(
0
)
r
0
sin(
0
)
r
0
r
cos(
0
)
_
, (2.17)
and the coecients
i
j
are obtained by inverting the matrix:
_

r
r
(r, [r
0
,
0
)
r

(r, [r
0
,
0
)

r
(r, [r
0
,
0
)

(r, [r
0
,
0
)
_
=
_
cos(
0
)
1
r
0
sin(
0
)
r sin(
0
)
r
r
0
cos(
0
)
_
. (2.18)

Note: we have there

j
i
(x[y) =
j
i
(y[x) . (2.19)
Check if this is a general property and, if yes, give the corresponding theorem.
2.1. PARALLEL TRANSPORTATION 57
Example: The surface of the sphere being a curved (2-D) space, the result of parallel transportation
of a vector between two points depends on the path followed (see for instance gure 2.1). The
transportation of vectors along geodesics (great circles) has been illustrated by gure 2.2. Let us now
examine the parallel transportation along the coordinate lines: if the Meridians are geodesics, the
Parallels are not. We show in appendix XXX that, for the parallel transport along a Meridian, we
have
e

(, |

, ) = e

(, )
e

(, |

, ) =
sin

sin
e

(, ) , (2.20)
i.e.,
_

(, [

, )

(, [

, )

(, [

, )

(, [

, )
_
=
_
1 0
0 sin

/ sin
_
, (2.21)
while, for the parallel transportation along a Parallel, we have
e

(, | ,

) = cos
__

_
cos

(, )
1
sin
sin
__

_
cos

(, )
e

(, | ,

) = sin sin
__

_
cos

(, ) + cos
__

_
cos

(, ) , (2.22)
i.e.,
_

(, [,

(, [,

(, [,

(, [,

)
_
=
_
cos [(

) cos ] sin [(

) cos ] / sin
sin sin [(

) cos ] cos ([

) cos ]
_
.
(2.23)

2.1.5 Parallel transport of vectors

Let us naturally denote V
i
(x | y) the components on the local basis at point x of the vector
V(x | y) :
V(x | y) = V
i
(x | y) e
i
(x) . (2.24)
How can we express them as a function of the original components of the vector?
At point y we have
V(y) = V
i
(y) e
i
(y) . (2.25)
If we transport to point x both the vector V(y) and the basis e
i
(y) , the components of the
transported vector on the transported basis will remain the same they were at the original point:
V(x | y) = V
i
(y) e
i
(x | y) , (2.26)
or, using equation 2.6,
V(x | y) = V
i
(y)
i
j
(x[y) e
j
(x) , (2.27)
i.e.,
V
j
(x | y) = V
i
(y)
i
j
(x[y) . (2.28)
This important equation solves formally the problem of parallel transport of vectors: if the functions

i
j
(x[y) are given, then we know the result of parallel transportation of any vector from any point
to any other point.
For a form (or a covariant vector) we will have
F(x | y) = e
i
(x | y) F
i
(y) = e
j
(x)
j
i
(x[y) F
i
(y) , (2.29)
58 CHAPTER 2. DERIVATIVES OF TENSORS
(r
0
,
0
)
(r
0
,
0
)
(r,)
(r,)
Figure 2.10: Top: the local basis at two points. Bottom: in addition, the vectors obtained at point
XXX by parallel transport of the basis at XXX.
2.1. PARALLEL TRANSPORTATION 59
i.e.,
F
j
(x | y) =
j
i
(x[y) F
i
(y) . (2.30)
For a general tensor
T(y) = T
k...
ij...
(y) e
k
(y) e

(y) . . . e
i
(y) e
j
(y) . . . (2.31)
we obtain
T
pq...
mn...
(x | y) = T
k...
ij...
(y)
p
k
(x[y)
q

(x[y) . . .
i
m
(x[y)
j
n
(x[y) . . . . (2.32)
Let us look with some detail at equation 2.27. It is a perfectly valid tensor equation at point x ,
in the sense that we have one tensor at each side of the equality sign, and that the two tensors are
equal in the standard sense. But, besides the eld variable x , the equation has, at both sides, the
variable y . We have to interpret properly the dierence between the variables x and y in this
equation. For any value of y , we have an ordinary tensor equation at point x , so, in this sense, x
is a eld variable, but y is not. Comment: explain here that the variable y plays here a role very
similar to the variable y in a conditional probability density f(x[y) : while x is a random variable,
y is not.
2.1.6 Parallel transport of densities and capacities
We have just seen that during the parallel transportation of a vector, its components on the local
basis may change (because the local basis may be dierent at every point).
We have introduced above, three kinds of scalar elds: true scalars, (x) , densities (x) and
capacities (x) . As these are tensor elds, they must have their parallel transportation dened.
We are going to justify here the following result: The parallel transportation of a true scalar (y) ,
from point y to point x , leaves its value unchanged:
(x | y) = (y) . (2.33)
The parallel transportation of a scalar capacity (y) gives
(x | y) =
g(y)
g(x)
(y) . (2.34)
The parallel transportation of a scalar density (y) gives
(x | y) =
g(x)
g(y)
(y) . (2.35)
Here, g(x) denotes the value at point x of the fundamental density, i.e, the square root of
the determinant of the metric tensor (see section xxx). Comment: say that, in non metric spaces, a
function g(x) has to be given that, even it is not related to a metric, denes, in an ad-hoc way, the
parallel transportation of densities and capacities.
We have
T
p
m
(x | y) = T
k
i
(y)
p
k
(x[y)
i
m
(x[y) (2.36)
which gives
T
s
s
(x | y) = T
k
i
(y)
s
k
(x[y)
i
s
(x[y) (2.37)
and using

s
k
(x[y)
i
s
(x[y) =
i
k
, (2.38)
T
s
s
(x | y) = T
s
s
(y) . (2.39)
As the trace of a (true) tensor is a true scalar, this justies the denition 2.33. Equations 2.34 and 2.35
follow from it when considering that the product (resp. the ratio) of a capacity (resp. a density) eld
by the fundamental density gives a pure tensor eld.
60 CHAPTER 2. DERIVATIVES OF TENSORS
2.1.7 The Connection
In what follows we will need the denition

ki
j
(x[y) =

i
j
x
k
(x[y) . (2.40)
where the loose notations indicate derivation with respect to the second variable in the function

i
j
(x[y) . As there is no risk of confusion, and to economize symbols, we will also use the letter
to denote the value of the expression 2.40 for y = x :

ki
j
(x) =
ki
j
(x[x) . (2.41)
Here we are taking derivatives with respect the variable y which, as mentioned in section 2.1.4
is not a eld variable. In addition, as we take partial derivatives, and not covariant derivatives,
the index k in
ki
j
is not a tensor index:
ki
j
is not a tensor.
Similarly, for the coecients dening the parallel transport of forms we dene
L
ij
k
(x[y) =

j
k
y
i
(x[y) (2.42)
and

ki
j
(x) =
ki
j
(x[x) . (2.43)
Taking the derivative of equation 2.14 gives

k
j
y
m
(x[y)
j
i
(x[y) +
k
j
(x[y)

j
i
y
m
(x[y) = 0 , (2.44)
and, making y = x ,

mk
j
(x[x)g
j
i
+g
k
j
L
mj
i
(x[x) = 0 , (2.45)
i.e.,

mk
i
(x) = L
mk
i
(x) . (2.46)
Comment: this is a very important equation.
We have introduced the coecients
i
j
(x[y) and
k
ij
(x) , corresponding the vector basis, and
the coecients
i
j
(x[y) and L
k
ij
(x) , corresponding the dual basis. We could well introduce dierent
names for those elds (that have all in common of not being ordinary tensor elds). Rather we will
simply say the connection coecients
i
j
(x[y) , or the connection coecients
ij
k
(x) , the
symbol indicating the eld being used.
Example: Polar coordinates. We compute here the connection directly from the coecients
dening the parallel transportation of tensors, without using the metric. We have seen alsewhere that
the coecients are given, for polar coordinates, by
_

r
r
(r
0
,
0
[r, )
r

(r
0
,
0
[r, )

r
(r
0
,
0
[r, )

(r
0
,
0
[r, )
_
=
_
cos(
0
)
1
r
0
sin(
0
)
r sin(
0
)
r
r
0
cos(
0
)
_
. (2.47)
The denition 2.40 then gives
_

rr
r
(r
0
,
0
[r, )
rr

(r
0
,
0
[r, )

r
r
(r
0
,
0
[r, )
r

(r
0
,
0
[r, )
_
=
_
0 0
sin(
0
)
1
r
0
cos(
0
)
_
(2.48)
_

r
r
(r
0
,
0
[r, )
r

(r
0
,
0
[r, )

r
(r
0
,
0
[r, )

(r
0
,
0
[r, )
_
=
_
sin(
0
)
1
r
0
cos(
0
)
r cos(
0
)
r
r
0
sin(
0
)
_
. (2.49)
2.1. PARALLEL TRANSPORTATION 61
Finally, denition 2.41 gives
_

rr
r
(r, )
rr

(r, )

r
r
(r, )
r

(r, )
_
=
_
0 0
0
1
r
_
(2.50)
_

r
r
(r, )
r

(r, )

r
(r, )

(r, )
_
=
_
0
1
r
r 0
_
. (2.51)

2.1.8 Parallel transport of the metric tensor

The notions of parallel transport dened above need rules that may not be based on any notion of
metric (see the example conductivity-velocity). If there is a metric, it is also possible to dene the
parallel transport, as if we can measure distances and angles, we can perform parallel transportation.
Then, if the have the symbols
i
j
(x | y) dening the parallel transportation and we have a
metric g
ij
(x) , there must be some relations between them that ensure compatibility (comment:
explain better).
Comment: I have to explain here why the parallel transportation of the metric gives the metric.
In any case, this writes
g(x | y) = g(x) , (2.52)
i.e.,
g
ij
(x | y) = g
ij
(x) . (2.53)
The general rule of transportation of tensors (equation 2.32) gives
g
ij
(x | y) =
i
p
(x[y)
j
q
(x[y) g
pq
(y) , (2.54)
or, using 2.53,
g
ij
(x) =
i
p
(x[y)
j
q
(x[y) g
pq
(y) . (2.55)
This is the constraint linking the notion of parallel transport and the metric (when it exists). (Com-
ment: yes, but how can I compute the coecients
i
j
(x | y) from the metric g
ij
(x) ?)
Taking partial derivatives in equation 2.55 with respect to the (non-eld) variable y and using
the denition 2.42 of the connection gives
L
mi
p
(x[y)
j
q
(x[y) g
pq
(y) +
i
p
(x[y) L
mj
q
(x[y) g
pq
(y) +
i
p
(x[y)
j
q
(x[y)
m
g
pq
(y) = 0 . (2.56)
Taking there x = y , and using the notation 2.43 gives
L
mi
p
(x)
j
q
g
pq
(x) +
i
p
L
mj
q
(x) g
pq
(x) +
i
p

j
q

m
g
pq
(x) = 0 , (2.57)
i.e.,
L
mij
(x) + L
mji
(x) +
m
g
ij
(x) = 0 , (2.58)
where I have introduced the notation
L
ijk
(x) = g
ks
(x) L
ij
s
(x) (2.59)
that toguether with

ijk
(x) = g
ks
(x)
ij
s
(x) (2.60)
are common, but we should not forget thet the connections
ij
k
and L
ij
k
are not tensors. Us-
ing nally the property 2.46, equation 2.58 gives a simple and important relationship between the
connection and the metric:

k
g
ij
=
kij
+
kji
. (2.61)
62 CHAPTER 2. DERIVATIVES OF TENSORS
The torsion tensor will be dened below as
S
ij
k
=
ij
k

ji
k
. (2.62)
With this denition, a simple substitution shows that equation 2.61 implies the property

ijk
=
1
2
(
i
g
jk
+
j
g
ik

k
g
ij
) +
1
2
(S
ijk
+S
kij
+S
kji
) . (2.63)
It shows that if the metric and the torsion are given, then we can compute the connection. This, in
particular, shows that the metric alone is not enough to compute the connection.
Comment: I have a problem here. Apparently, the metric perfectly denes the parallel trans-
portation, so it should dene the coecients
i
j
(x[y) and, thus, the connection. This implies that
my previous denition of parallel transportation from the metric is only valid for spaces
without torsion.
Comment: how can I compute
i
j
(x | y) from g
ij
(x) ?
Example: In the Euclidean plane with polar coordinates, the connection coecients
i
j
are given
by (equation 2.18)
_

r
r
(r, [r
0
,
0
)
r

(r, [r
0
,
0
)

r
(r, [r
0
,
0
)

(r, [r
0
,
0
)
_
=
_
cos(
0
)
1
r
0
sin(
0
)
r sin(
0
)
r
r
0
cos(
0
)
_
. (2.64)
Them the parallel transportation of a tensor F
ij
(r
0
,
0
) from the point (r
0
,
0
) to a point (r, )
is given, using equation 2.32, by
F
ij
(x | x
0
) =
i
p
(x[x
0
)
j
q
(x[x
0
) F
pq
(x
0
) . (2.65)
As an example, this gives, for the component :
F

(r, | r
0
,
0
) =

r
(r, [r
0
,
0
)

r
(r, [r
0
,
0
) F
rr
(r
0
,
0
)
+

r
(r, [r
0
,
0
)

(r, [r
0
,
0
) F
r
(r
0
,
0
)
+

(r, [r
0
,
0
)

r
(r, [r
0
,
0
) F
r
(r
0
,
0
)
+

(r, [r
0
,
0
)

(r, [r
0
,
0
) F

(r
0
,
0
) , (2.66)
i.e.,
F

(r, | r
0
,
0
) = r
2
sin
2
(
0
) F
rr
(r
0
,
0
)
(r
2
/r
0
) sin(
0
) cos(
0
) F
r
(r
0
,
0
)
(r
2
/r
0
) sin(
0
) cos(
0
) F
r
(r
0
,
0
)
+ (r
2
/r
2
0
) cos
2
(
0
) F

(r
0
,
0
) . (2.67)
So far for a general tensor. The metric tensor in polar coordinates is, at any point (r, ) ,
_
g
rr
(r, ) g
r
(r, )
g
r
(r, ) g

(r, )
_
=
_
1 0
0 r
2
_
. (2.68)
As it is symmetric, equation 2.67 simplies rst to
g

(r, | r
0
,
0
) = r
2
sin
2
(
0
) g
rr
(r
0
,
0
)
+ (r
2
/r
2
0
) cos
2
(
0
) g

(r
0
,
0
) , (2.69)
and, using the values g
rr
(r
0
,
0
) = 1 and g

(r
0
,
0
) = r
2
0
, nally gives
g

(r, | r
0
,
0
) = r
2
_
sin
2
(
0
) + cos
2
(
0
)
_
= r
2
, (2.70)
which is simply the value of the metric at the new point (r, ) : the parallel transportation of the
metric from point x
0
to point x gives the metric at point x .
2.2. DERIVATIVES 63
2.2 Derivatives
2.2.1 Derivatives
Let us rst consider an ordinary function x f(x) of the ordinary variable x (here ordinary
means that we do not introduce tensorial concepts). The derivative of the function f() at point x
is the function f

() dened by the condition that the rst order development

f(x +x) = f(x) +f

(x) x (2.71)
must hold for any x .
For a vector eld x V(x) , the partial derivatives
j
V
i
of the vector eld at point x are
dened by the condition that the rst order development
V
i
(x +x) = V
i
(x) +
_

j
V
i
_
(x) x
j
(2.72)
must hold for any x . Although equation 2.72 makes perfect sense, it is not a tensor equation.
For we know (see section XXX) that tensors dened at a point of a manifold belong to the linear
space tangent to the manifold at that point. At a given point, we can add and multiply tensors, but
equation 2.72 has, at his left hand side, a vector at point x+x and, at his right hand side, a vector
at point x . So, even if the functions
_

j
V
i
_
(x) x
j
= V
i
(x +x) V
i
(x) (2.73)
are dened, they are not the components of a tensor eld. This implies that, in a change of coor-
dinates in the space, these functions will not transform according to the rule XXX that tansforms
the components of a tensor eld. The transformation rules of these functions are complicated and
uninteresting. (Comment: say that these functions describe how the components of the tensor eld
change when we change the point but not the basis).
By opposition to what precedes the rst order development
V
i
(x | x +x) = V
i
(x) +
_

j
V
i
_
(x) x
j
(2.74)
is a perfectly valid tensor equation. The expression in the left hand side describes the vector at point
x obtained by parallel transportation of the vector V(x + x) from point x + x to point x .
We have at both sides tensors at the same point of the space. Then,
_

j
V
i
_
(x) is a tensor eld,
named the tensor derivative of the eld V(x) . Comment: say that equivalent names are covariant
derivative or, simply, derivative.
Note: we can write equation 2.74 as
V
i
(x | x x) = V
i
(x)
_

j
V
i
_
(x) x
j
, (2.75)
and the diference between equations 2.74 and 2.75 gives a symmetric denition for the covariant
derivative:
V
i
(x | x +x) V
i
(x | x x) = 2
_

j
V
i
_
(x) x
j
. (2.76)
Taylors series: We know the properties
f(x +x) = f(x) +f

(x) x +
1
2!
f

(x) (x)
2
+. . . . (2.77)
and
V
i
(x +x) = V
i
(x) + (
j
V
i
)(x) x
j
+
1
2!
(
k

j
V
i
)(x) x
j
x
k
+. . . . (2.78)
64 CHAPTER 2. DERIVATIVES OF TENSORS
I conjecture that the development
V
i
(x | x +x) = V
i
(x) + (
j
V
i
)(x) x
j
+
1
2!
(
k

j
V
i
)(x) x
j
x
k
+. . . , (2.79)
holds, but I have yet to prove it. Comment: nd the proof.
From
V(x | y) = V
i
(y) e
i
(x | y) (2.80)
and
e
i
(x | y) =
i
j
(x[y) e
j
(x) (2.81)
we have
V(x | x +x) = V
i
(x +x)
i
j
(x[x +x) e
j
(x)
=
_
V
i
(x) +
V
i
x
k
(x) x
k
+. . .
_
_

i
j
(x[x) +
ki
j
(x[x) x
k
+. . .
_
e
j
(x)
=
_
V
i
(x) +
V
i
x
k
(x) x
k
+. . .
_
_
g
i
j
+
ki
j
(x) x
k
+. . .
_
e
j
(x)
=
_
V
j
(x) +
_
V
j
x
k
(x) +
ki
j
(x) V
i
(x)
_
x
k
+. . .
_
e
j
(x) . (2.82)
Then, from the denition
V
j
(x | x +x) = V
j
(x) +
_

k
V
j
_
(x) x
k
+. . . (2.83)
it follows
_

k
V
j
_
(x) =
_

k
V
j
_
(x) +
ki
j
(x) V
i
(x) . (2.84)
Comment: say, in the equations above, which symbols = and + correspond the tensor equations
and which do not.
The general formula for the covariant derivative of a tensorial object (density of order p) is

i
T
jk...
lm...
=
i
T
jk...
lm...
p
is
s
T
jk...
lm...
+
is
j
T
sk...
lm...
+
is
k
T
js...
lm...
+. . .

il
s
T
jk...
sm...

im
s
T
jk...
ls...
. . .
(2.85)
As the term
is
s
appears quite often, the following notation is useful:

i
=
is
s
. (2.86)
(Comment: say somewhere that, in all rigor, if V denotes the covariant derivative in intrinsic
notation, we should write (V)
i
j
instead of
i
V
j
but the last notation is much easier to handle.)
To be explicit, we write below the covariant derivatives of tensor objects with rank 0 , 1 , and 2 .
Derivatives in Euclidean spaces
We consider here an Euclidean space, and explain how covariant derivatives can be computed in an
arbitrary system of coordinates.
The equation
_

k
V
j
_
(x) =
_

k
V
j
_
(x) +
ki
j
(x) V
i
(x) , (2.87)
2.2. DERIVATIVES 65
taken au pied de la lettre seems to imply that covariant dierentiation will impose, when using a
nite-dierence computational grid, that we consider at the same point x , a vector V(x) and its
derivative (V)(x) . This is not so.
Let us see, for instance, how a practical computation should proceed in an Euclidean space with
arbitrary coordinates.
In what follows, X
I
denotes a Cartesian system of coordinates, while x
i
denotes another,
arbitrary, system of coordinates.
Using the Cartesian coordinates, we can compute the covariant derivative of a tensor eld using
simple partial derivatives:

M
T
IJ...
KL...
=
T
IJ...
KL...
X
M
. (2.88)
From this we can deduce the components of the same tensor in the general system of coordinates:

m
T
ij...
kl...
=
X
M
x
m
X
I
x
i
X
J
x
j
. . .
x
k
X
K
x

X
L
. . .
M
T
IJ...
KL...
=
X
M
x
m
X
I
x
i
X
J
x
j
. . .
x
k
X
K
x

X
L
. . .
T
IJ...
KL...
X
M
. (2.89)
Using
T
IJ...
KL...
X
M
=
x
n
X
M
T
IJ...
KL...
x
n
(2.90)
and
X
M
x
m
x
n
X
M
=
m
n
(2.91)
gives

m
T
ij...
kl...
=
X
I
x
i
X
J
x
j
. . .
x
k
X
K
x

X
L
. . .
T
IJ...
KL...
x
m
. (2.92)
Expresing the Cartesian components T
IJ...
KL...
in terms of the working coordinates,
T
IJ...
KL...
=
x
p
X
I
x
q
X
J
. . .
X
K
x
r
X
L
x
s
. . . T
pq...
rs...
, (2.93)
nally gives

m
T
ij...
kl...
=
X
I
x
i
X
J
x
j
. . .
x
k
X
K
x

X
L
. . .

x
m
_
x
p
X
I
x
q
X
J
. . .
X
K
x
r
X
L
x
s
. . . T
pq...
rs...
_
. (2.94)
In particular this gives, for a vector eld,

m
T
k
=
x
k
X
K

x
m
_
X
K
x
r
T
r
_
. (2.95)
These equations have two important uses. First, they give compact analytic formulas. For instance,
when computing the divergence of a vector eld using spherical coordinates, it directly gives the
compact expression

i
V
i
=
1
r
2

r
(r
2
V
r
) +
1
sin

(sin V

) +
V

(2.96)
and not an expression with all derivatives developed, as we would have obtained from the formula
_

k
V
j
_
(x) =
_

k
V
j
_
(x) +
ki
j
(x) V
i
(x) . (2.97)
66 CHAPTER 2. DERIVATIVES OF TENSORS
Figure 2.11: Finite-dierence (centered) approximation to covariant derivatives. In the limit when
x 0 , this denes the covariant derivative through the equation V
i
(x | x+x) V
i
(x | xx) =
2 (
k
V
i
)(x) x
k
or, to speak properly, V(x | x + x) V(x | x x) = 2 (V)
k
i
(x) x
k
e
i
(x) .
(Comment: explain that the thin lines are geodesics crossing at their mid points). (Comment: explain
that if N is the dimension of the space, we need N dierent directions x
1
, x
2
, . . . , x
N
to
compute all the values
i
V
k
).
Figure 2.12: Finite-dierence (non-centered) approximation to covariant derivatives. In the limit
when x 0 , this denes the covariant derivative through the equation V
i
(x | x +x) V
i
(x) =
(
k
V
i
)(x) x
k
or, to speak properly, V(x | x +x) V(x) = (V)
k
i
(x) x
k
e
i
(x) .
Second, it can be directly applied to a numerical computation of covariant derivatives. If the partial
derivatives x
i
/X
I
are known (analytically or numerically), it is a very convenient formula to use
(see chapter XXX) for more details).
Comment: Explain somewhere that equations 2.94 and 2.95 are, of course compatible with the
expression giving the connection in terms of Cartesian coordinates. For instance, equation 2.95 gives

m
T
k
=
x
k
X
K

x
m
_
X
K
x
r
T
r
_
=
x
k
X
K
_
X
K
x
r
T
r
x
m
+

2
X
K
x
m
x
r
T
r
_
=
x
k
X
K
X
K
x
r
T
r
x
m
+
x
k
X
K

2
X
K
x
m
x
r
T
r
=
k
r
T
r
x
m
+
mr
k
T
r
=
T
k
x
m
+
mr
k
T
r
, (2.98)
where we have used the expression (demonstrated somewhere)
x
k
X
K

2
X
K
x
m
x
r
=
mr
k
. (2.99)
Comment: The following text seems too big to be considered as a caption for the previous table.
Equations 1 give the components of the transported basis on the local basis at the arriving point.
Equation 2 expresses that the matrix with the coecients
i
k
is the inverse of the matrix with the
2.2. DERIVATIVES 67
Basic formulas for parallel transportation
e
i
(x | y) =
i
j
(x[y) e
j
(x) e
i
(x | y) = e
j
(x)
j
i
(x[y) (1)

i
j
(x[y)
j
k
(x[y) =
i
j
(x[y)
j
k
(x[y) =
i
k
(2)

i
j
(x[x) =
i
j

i
j
(x[x) =
i
j
(3)

i
j
(x[y)
j
k
(y[x) =
i
k

i
j
(x[y)
j
k
(y[x) =
i
k
(4)

i
j
(x[y) =
i
j
(y[x) (5)
V
i
(x | y) = V
j
(y)
j
i
(x[y) F
i
(x | y) =
i
j
(x[y) F
j
(y) (6)
T
ij...
k...
(x | y) =
i
p
(x[y)
j
q
(x[y) . . . T
pq...
mn...
(y)
m
k
(x[y)
n

(x[y) . . . (7)

ij
k
(x[y) =

j
k
y
i
(x[y) L
ij
k
(x[y) =

j
k
y
i
(x[y) (8)

ij
k
(x) =
ij
k
(x[x) L
ij
k
(x) = L
ij
k
(x[x) (9)
L
ij
k
(x) =
ij
k
(x) (10)
Denition of covariant derivatives
T
ij...
k...
(x | x +x) = T
ij...
k...
(x) +
_

p
T
ij...
k...
_
(x) x
p
+. . . (11)
Expression of covariant derivatives

p
T
ij...
k...
=
p
T
ij...
k...
+ L
pi
s
T
sj...
k...
+ L
pj
s
T
is...
k...
+. . . +
ps
k
T
ij...
s...
+
ps

T
ij...
ks...
+. . . (12)
Metric spaces
g
ij
(x) =
i
p
(x[y)
j
q
(x[y) g
pq
(y) (13)

k
g
ij
=
kij
+
kji
(14)

ijk
=
1
2
(
i
g
jk
+
j
g
ik

k
g
ij
) +
1
2
(S
ijk
+S
kij
+S
kji
)
_
S
ij
k
=
ij
k

ji
k
_
(15)
Euclidean spaces

ij
k
=

2
X
K
x
i
x
j
x
k
X
K
=
X
I
x
i
X
J
x
j

2
x
k
X
I
X
J
(16)
Table 2.1: Intended text too large (see below).
68 CHAPTER 2. DERIVATIVES OF TENSORS
V(x
B
|x
A
)
V(x
B
)
V(x
A
)
Figure 2.13: Parallel transportation of a vector. This gure schematically represents a vector eld
V(x) . The two vectors V(x
A
) and V(x
B
) represent the values of the vector elds at points
x
A
and x
A
respectively. Also indicated is the vector V(x
B
[x
A
) , the vector at point x
B
obtained by parallel transport of V(x
A
) . If, as in left of gure XXX, the space is Euclidean and we
introduce, as an intermediary, Cartesian coordinates, then, we have the simple relationship (see text):
V
i
(x
B
[x
A
) =
x
i
y
K
(y
B
)
y
K
x
j
(x
A
) V
j
(x
A
) . If the space is not assumed to be Euclidean, Cartesian
coordinates may not exist. The only way then to perform parallel transportation of a vector is by
assuming the connection coecients
ij
k
given and to use the rst order approximation (see text)
V
j
(x+x [ x) = V
j
(x)
ki
j
(x) x
k
V
i
to . . . at neighbouring points (or at distant points through
analytic or numerical integration).
coecients
i
j
. Equations 3 say that transporting a tensor from a point x to the same point (along
a geodesic) means no transportation at all. Equations 4 express the fact that transporting a tensor
from point y to point x , then back to point y (always along the geodesic) gives the original
tensor. Equation 5 says that the coecients transporting the basis from x to y are the same that
the coecients transporting the dual basis from y to x . Equations 6 expresses how the components
of a vector or a form change in a parallel transportation, while equation 7 gives the general result for
an arbitrary tensor. Equations 8 dene the two point connection, while equations 9 dene the one
point connection. Equation 10 expresses that the connection associated to vectors is opposite to the
connection associated to forms. Equation 11 gives the denition of covariant derivatives (the . . .
stand for terms of high order [proportional to x
p
x
q
, to x
p
x
q
x
r
, etc.]. Equation 12 gives the
general formula for covariant derivatives using the connection. Equation 13 gives the relation between
the coecients dening the parallel transportation and the metric (wheren there is one): the parallel
transportation of the metric from point y to point x gives the metric at point x . Equation 14 gives
the relation between the metric and the connection, from which it is possible to obtain (equation 15)
the formula expressing the connection from the metric and the torsion (the torsion is to be introduced
in section XXX). Equation 16 shows that, in Euclidean spaces, in addition to arbitrary working
coordinates x
i
, one can always introduce Cartesian coordinates X
I
, and then the connection
coecients can be computed from the partial derivatives of one coordinate system with respect to the
other.
2.2.2 Second derivatives
For an arbitrary tensor eld T
jk...
pq...
(x) , we have dened its derivative (
i
T
jk...
pq...
) (x) , by the rst
order development
T
jk...
pq...
(x | x +x) = T
jk...
pq...
(x) + (
i
T
jk...
pq...
) (x) x
i
. (2.100)
If the tensor eld T
jk...
pq...
(x) , is itself the derivative of some other eld,
T
jk...
pq...
=
j
S
k...
pq...
, (2.101)
2.2. DERIVATIVES 69
Capacity Tensor Density

k
s =
k
s +
k
s
k
s =
k
s
k
s =
k
s
k
s

k
F
i
=
k
F
i
+
k
F
i

ki
s
F
s

k
F
i
=
k
F
i

ki
s
F
s

k
F
i
=
k
F
i

k
F
i

ki
s
F
s

k
V
i
=
k
V
i
+
k
V
i
+
ks
i
V
s

k
V
i
=
k
V
i
+
ks
i
V
s

k
V
i
=
k
V
i

k
V
i
+
ks
i
V
s

k
Q
ij
=
k
Q
ij
+
k
Q
ij

ki
s
Q
sj

kj
s
Q
is

k
Q
ij
=
k
Q
ij

ki
s
Q
sj

kj
s
Q
is

k
Q
ij
=
k
Q
ij

k
Q
ij

ki
s
Q
sj

kj
s
Q
is

k
R
i
j
=
k
R
i
j
+
k
R
i
j

ki
s
R
s
j
+
ks
j
R
i
s

k
R
i
j
=
k
R
i
j

ki
s
R
s
j
+
ks
j
R
i
s

k
R
i
j
=
k
R
i
j

k
R
i
j

ki
s
R
s
j
+
ks
j
R
i
s

k
S
i
j
=
k
S
i
j
+
k
S
i
j
+
ks
i
S
s
j

kj
s
S
i
s

k
S
i
j
=
k
S
i
j
+
ks
i
S
s
j

kj
s
S
i
s

k
S
i
j
=
k
S
i
j

k
S
i
j
+
ks
i
S
s
j

kj
s
S
i
s

k
T
ij
=
k
T
ij
+
k
T
ij
+
ks
i
T
sj
+
ks
j
T
is

k
T
ij
=
k
T
ij
+
ks
i
T
sj
+
ks
j
T
is

k
T
ij
=
k
T
ij

k
T
ij
+
ks
i
T
sj
+
ks
j
T
is
.
.
.
.
.
.
.
.
.
Table 2.2: Note: write here the caption.
Divergence of densities

k
V
k
=
k
V
k

k
T
ik
=
k
T
ik
+
ks
i
T
sk

k
T
i
k
=
k
T
i
k

ki
s
T
s
k
If T
ij
= T
ji
and
ij
k
=
ji
k
, then,
k
T
ik
=
k
T
ik
.
Table 2.3: Note: explain that the divergence of densities has one less term than the divergence of a
tensor, as two of the terms of the previous table compensate.
70 CHAPTER 2. DERIVATIVES OF TENSORS
equation 2.100 gives the rst order development
(
j
S
k...
pq...
)(x | x +x) = (
j
S
k...
pq...
)(x) + (
i

j
S
k...
pq...
) (x) x
i
, (2.102)
that denes the second order derivative of a tensor eld.
In section xxx we will see that the operator
i

j
is not identical to
j

i
: contrarily to the
partial derivatives, the covariant derivatives do not conmute.
Figure 2.14 gives a pictorial representation of the nite-dierence computation of (noncentered)
second derivatives.
Comment: all this is probably false. I should only say that the derivative of a tensor eld is a
tensor eld, so we can take derivatives of it.
2.2. DERIVATIVES 71
x+x
A
x+x
A
x+x
B
x+x
B
x x
B
A
C
B
A
A
D'
D'
C'
C'
E=D'-B
F=C'-A
E
F
E'
G=E'-F
D
C
B
B
B'
D'
D'
D
C
H'
H
I=B'-A
I
H=D'-C
A
J = H'-I
Figure 2.14: See caption here below (L
a
T
E
X error when put here).
Caption of the gure above: Comment: this gure has been redrawn, and the caption has to
be rewritten accordingly. At a point x of the space, consider the two coordinate perturbations
x
A
and x
B
. Consider also a vector eld V
i
(x) dened at all points of the space. Let us
describe the column at left. Using parallel transportation, from the expressions V
i
(x + x
B
|
x + x
B
+ x
A
) = V
i
(x + x
B
) +
_

j
V
i
_
(x + x
B
) x
j
A
and V
i
(x | x + x
A
) = V
i
(x) +
_

j
V
i
_
(x) x
j
A
we can easily deduce the vectors
_

j
V
i
_
(x + x
B
) x
j
A
and
_

j
V
i
_
(x) x
j
A
,
while from the expression
_

j
V
i
_
(x | x + x
B
) x
j
A
=
_

j
V
i
_
(x) x
j
A
+
_

j
V
i
_
(x) x
j
A
x
k
B
we easily deduce the vector G =
_

j
V
i
_
(x) x
j
A
x
k
B
. Similarly, in the right column, we de-
duce the vector J =
_

j
V
i
_
(x) x
j
B
x
k
A
. The dierence between these two vectors, namely
_

j

j

k
V
i
_
(x) x
j
A
x
k
B
, is, in general, not zero, if the space has curvature. Comment: explain
better that, in the left column, we move rst left, then down, while in the right column, we rst move
down, then left.
72 CHAPTER 2. DERIVATIVES OF TENSORS
2.2.3 Derivatives of the metric tensor
To be written.
2.2.4 Old text: Ordinary derivatives
Let f(x) be a real function of the real variable x . The expression
f(x +x) = f(x) +f

(x) x (2.103)
denes the secant of the function f(x) at points x and x + x as the linear function taking the
value f(x) at point x and with slope f

(x) . The secant is, of course, a function of x . The limit

of the secant when x 0 is called the tangent of f(x) at point x . The slope of the tangent is
the derivative.
Let now f(x) = f(x, y) be a real function of the two-dimensional variable x = x, y . Setting
x
1
=
_
x
1
y
1
_
x
2
=
_
x
2
y
2
_
, (2.104)
the secant of the function f(x) at points x , x x
1
and x x
2
is dened as being the linear
application taking the value f(x, y) at point (x, y) and with slopes (along each axis)
f
x
(x, y) and
f
y
(x, y) dened the the set of two equations with two unknowns
f(x +x
1
, y +y
1
) = f(x, y) +
f
x
(x, y) x
1
+
f
y
(x, y) y
1
f(x +x
2
, y +y
2
) = f(x, y) +
f
x
(x, y) x
2
+
f
y
(x, y) y
2
. (2.105)
Again, the limit of the secant when x
1
0 and x
2
0 is the tangent to f(x, y) at (x, y) .
If in the 1-D case, equation 2.103 can be solved for x , and the limit dening the derivative can
be explicitly written as
f

(x) = lim
x0
f(x +x) f(x)
x
, (2.106)
equation 2.105 involves the solution of a linear system, and the explicit expression is less simple,
although this is not important. Of course, when choosing the perturbed points x + x
1
and
x +x
2
along the coordinate lines,
x
1
=
_
x
0
_
x
2
=
_
0
y
_
, (2.107)
then, equation 2.105 becomes
f(x +x, y) = f(x, y) +
f
x
(x, y) x
f(x, y +y) = f(x, y) +
f
y
(x, y) y , (2.108)
and, in the limit, we can write
f
x

(x, y) = lim
x0
f(x +x, y) f(x)
x
f
y

(x, y) = lim
y0
f(x, y +y) f(x)
y
. (2.109)
It is geometrically clear that the tangent function so dened is independent, in the limit, of the
particular vectors x
1
and x
2
chosen.
2.2. DERIVATIVES 73
2.2.5 Stars and Crosses
Here we introduce the practical? geometrical? aspects of the computation of covariant derivatives.
We will not use explicitly the connection coecients Comment: but I must also explain how the
computation is to be made if we use them).
Comment: explain that all this will be ready for a nite dierence computation.
Comment: I have to write somewhere the three equations
T
ij...
k...
(x | x +x) T
ij...
k...
(x) =
_

m
T
ij...
k...
_
(x) x
m
(2.110)
T
ij...
k...
(x | x +x) T
ij...
k...
(x | x x) = 2
_

m
T
ij...
k...
_
(x) x
m
(2.111)
and
T
ij...
k...
(x | x +x
A
) T
ij...
k...
(x | x +x
B
) =
_

m
T
ij...
k...
_
(x) (x
m
A
x
m
B
) , (2.112)
and say that, independently of any star or cross, one of these equations denes the covariant derivative
(if for any. . . in the limit. . . ).
There are two fundamental ways to compute nite dierences. For instance, in 2-D one may take
two point along each direction or, alternatively, one can take a Mercedes star. Comment: explain
better.
Stars:
Let us rst examine the special cases in 2 and 3 dimensions, then, the general, n-dimensional case.
2-D: Consider the points indicated in gure 2.15, and assume that a vector eld V
i
(x) has been
dened on the space. Using the denition of tensorial (or covariant) derivative (equation XXX), we
can write, at rst order, the set of equations
V
i
(x | x +x
1
) V
i
(x | x +x
2
) =
_

j
V
i
_
(x)
_
x
j
1
x
j
2
_
V
i
(x | x +x
2
) V
i
(x | x +x
3
) =
_

j
V
i
_
(x)
_
x
j
2
x
j
3
_
V
i
(x | x +x
3
) V
i
(x | x +x
1
) =
_

j
V
i
_
(x)
_
x
j
3
x
j
1
_
. (2.113)
These three equations are not independent (we can obtain any of them as [minus] the sum of the other
two): it is enough to take two of them.
Each of the two selected equations being valid for i = 1 and i = 2 we have, in all, four equations,
exactly the number needed to estimate the four components of
j
V
i
(x) . For any nite value of x
1
,
x
2
, and x
3
, this will give a nite-dierence approximation to the covariant derivative of the vector
eld. In the limit when x
1
, x
2
, and x
3
, tend to zero, we obtain the exact value of the covariant
derivative.
For short, equation 2.113 can be written
V(x | x +x
1
) V(x | x +x
2
) = (
j
V) (x)
_
x
j
1
x
j
2
_
V(x | x +x
2
) V(x | x +x
3
) = (
j
V) (x)
_
x
j
2
x
j
3
_
V(x | x +x
3
) V(x | x +x
1
) = (
j
V) (x)
_
x
j
3
x
j
1
_
. (2.114)
More generally, if a tensor eld
T(x) = T
ij...
k...
(x) e
i
(x) e
j
(x) . . . e
k
(x) e

(x) . . . (2.115)
74 CHAPTER 2. DERIVATIVES OF TENSORS
has been dened on the space the following set of equations perfectly denes the nite-dierence
approximation the the covariant derivatives of the tensor eld:
T(x | x +x
1
) T(x | x +x
2
) = (
j
T) (x)
_
x
j
1
x
j
2
_
T(x | x +x
2
) T(x | x +x
3
) = (
j
T) (x)
_
x
j
2
x
j
3
_
T(x | x +x
3
) T(x | x +x
1
) = (
j
T) (x)
_
x
j
3
x
j
1
_
. (2.116)
Example: Consider the Euclidean plane, with Cartesian coordinates, and let us evaluate the
gradient of a scalar eld.
Figure 2.16 shows the Mercedes star chosen to evaluate derivatives. We have
_
x
1
1
x
2
1
_
=
_

0
_ _
x
1
2
x
2
2
_
=
_
/2
+

3/2
_ _
x
1
3
x
2
3
_
=
_
/2

3/2
_
. (2.117)
Applying equation 2.116 to a scalar eld (x, y) gives
(x +, y) (x
1
2
, y +

3
2
) = (
x
) (x, y) (
3
2
) + (
y
) (x, y) (

3
2
)
(x
1
2
, y +

3
2
) (x
1
2
, y

3
2
) = (
x
) (x, y) (0) + (
y
) (x, y) (

3)
(x
1
2
, y

3
2
) (x +, y) = (
x
) (x, y) (
3
2
) + (
y
) (x, y) (

3
2
) (2.118)
From these three (nonindependent) equations we get
(
y
) (x, y) =
(x
1
2
, y +

3
2
) (x +, y)

3
(2.119)
(
x
) (x, y) =
2
3
_
(x +, y)
1
2
_
(x
1
2
, y +

3
2
) + (x
1
2
, y

3
2
)
__
. (2.120)
Comment: these notations are too complicated. I should name the points P , P
1
, P
2
, and P
3
,
and give the equations
(
y
) (P) =
(P
2
) (P
1
)

3
(2.121)
(
x
) (P) =
2
3
_
(P
1
)
1
2
((P
2
) + (P
3
))
_
. (2.122)
Comment: say somewhere that for the derivative along y this gives the simple dierence between the
values of at points P
2
and P
3
, while for the derivative along x this gives the simple dierence
between the value of at point P
1
and the average of the values of at points P
2
, and P
3
.
Comment: say also somewhere the this generalizes to arbitrary dimension, and give the corresponding
formulas.
3-D: Comment: say here that we have a tetrahedron. For a vector eld in 3-D (see gure 2.17),
equation XXX gives, at rst order,
V
i
(x | x +x
1
) V
i
(x | x +x
2
) =
_

j
V
i
_
(x)
_
x
j
1
x
j
2
_
V
i
(x | x +x
2
) V
i
(x | x +x
3
) =
_

j
V
i
_
(x)
_
x
j
2
x
j
3
_
V
i
(x | x +x
3
) V
i
(x | x +x
4
) =
_

j
V
i
_
(x)
_
x
j
3
x
j
4
_
V
i
(x | x +x
4
) V
i
(x | x +x
1
) =
_

j
V
i
_
(x)
_
x
j
4
x
j
1
_
, (2.123)
2.2. DERIVATIVES 75
x
x+x
1
x+x
2
x+x
3
x
3
x
2
x
1
Figure 2.15: To be written.
and, again, any set of three equations among the four can be taken as the set used to dene the
components of the covariant derivatives of the vector eld (the three equation split, for the three
values of the index i = 1, 2, 3 ) into nine equations, which is the number of components to be
evaluated for
i
V
j
(x) in three dimensions.
For a general tensor,
T(x | x +x
1
) T(x | x +x
2
) = (
j
T) (x)
_
x
j
1
x
j
2
_
T(x | x +x
2
) T(x | x +x
3
) = (
j
T) (x)
_
x
j
2
x
j
3
_
T(x | x +x
3
) T(x | x +x
4
) = (
j
T) (x)
_
x
j
3
x
j
4
_
T(x | x +x
4
) T(x | x +x
1
) = (
j
T) (x)
_
x
j
4
x
j
1
_
. (2.124)
N-D: Comment: say here that we take N+1 points. We have
T(x | x +x
1
) T(x | x +x
2
) = (
j
T) (x)
_
x
j
1
x
j
2
_
T(x | x +x
2
) T(x | x +x
3
) = (
j
T) (x)
_
x
j
2
x
j
3
_
. . . = . . .
T(x | x +x
n
) T(x | x +x
n+1
) = (
j
T) (x)
_
x
j
n
x
j
n+1
_
T(x | x +x
n+1
) T(x | x +x
1
) = (
j
T) (x)
_
x
j
n+1
x
j
1
_
. (2.125)
From these n+1 equations, we can arbitrarily take n of them. Whatever the number of components
of the tensor T(bfx) may be we will have three times as many equations, and this is the number of
components of the gradient of the tensor.
1-D: Give some sense to what follows. See gure 2.18.
T
11...
11...
(x | x +x
1
) T
11...
11...
(x | x +x
2
) =
_

1
T
11...
11...
_
(x)
_
x
j
1
x
j
2
_
T
11...
11...
(x | x +x
2
) T
11...
11...
(x | x +x
1
) =
_

1
T
11...
11...
_
(x)
_
x
j
2
x
j
1
_
. (2.126)
This is the same equation.
Comment: what is a tensor in 1-D (only the index i = 1 is possible)?
Crosses:
Let us rst examine the special cases in 2 and 3 dimensions, then, the general, n-dimensional case.
76 CHAPTER 2. DERIVATIVES OF TENSORS
Figure 2.16: To be written.
x
3
x
2
x
1
x
4
x
Figure 2.17: To be written.
x
2
x
1
x
Figure 2.18: To be written.
2.2. DERIVATIVES 77
2-D: Consider the points indicated in gure 2.19, and assume that a vector eld V
i
(x) has been
dened on the space. Using the denition of tensorial (or covariant) derivative (equation XXX), we
can write, at rst order, the set of equations
V
i
(x | x +x
1
/2) V
i
(x | x x
1
/2) =
_

j
V
i
_
(x) x
j
1
V
i
(x | x +x
2
/2) V
i
(x | x x
2
/2) =
_

j
V
i
_
(x) x
j
2
. (2.127)
Each of the two equations being valid for i = 1 and i = 2 we have, in all, four equations, exactly
the number needed to estimate the four components of
_

j
V
i
_
(x) . For any nite value of x
1
, and
x
2
, this will give a nite-dierence approximation to the covariant derivative of the vector eld. In
the limit when x
1
and x
2
tend to zero, we obtain the exact value of the covariant derivative.
For a general tensor eld, equation 2.127 becomes
T
ij...
k...
(x | x +x
1
/2) T
ij...
k...
(x | x x
1
/2) =
_

m
T
ij...
k...
_
(x) x
m
1
T
ij...
k...
(x | x +x
2
/2) T
ij...
k...
(x | x x
2
/2) =
_

m
T
ij...
k...
_
(x) x
m
2
. (2.128)
and, again, we have as much equations as unknowns to solve for the nite-dierence approximation
to the covariant derivative and, in the limit, to the derivative itself.
Comment: I have to say somewhere why I can get the exact derivative having selected to take the
limit along two specially selected directlions only.
Example: Consider the Euclidean plane, with Cartesian coordinates, and let us evaluate the
gradient of a scalar eld.
Figure 2.20 shows the star chosen to evaluate derivatives. Applying equation 2.128 to a scalar eld
(x, y) gives
(x +/2, y) (x /2, y) = (
x
) (x, y)
(x, y +/2) (x, y /2) = (
y
) (x, y) , (2.129)
i.e.,
(
x
) (x, y) =
(x +/2, y) (x /2, y)

(2.130)
(
y
) (x, y) =
(x, y +/2) (x, y /2)

. (2.131)
3-D: For a vector tensor in 3-D (see gure 2.21), equation XXX gives, at rst order,
T
ij...
k...
(x | x +x
1
/2) T
ij...
k...
(x | x x
1
/2) =
_

m
T
ij...
k...
_
(x) x
m
1
T
ij...
k...
(x | x +x
2
/2) T
ij...
k...
(x | x x
2
/2) =
_

m
T
ij...
k...
_
(x) x
m
2
T
ij...
k...
(x | x +x
3
/2) T
ij...
k...
(x | x x
3
/2) =
_

m
T
ij...
k...
_
(x) x
m
3
. (2.132)
and, again, we have as many equations as unknowns.
N-D: Comment: say here that we take N+1 points. We have
T
ij...
k...
(x | x +x
1
/2) T
ij...
k...
(x | x x
1
/2) =
_

j
T
ij...
k...
_
(x) x
j
1
. . . = . . .
T
ij...
k...
(x | x +x
n
/2) T
ij...
k...
(x | x x
n
/2) =
_

j
T
ij...
k...
_
(x) x
j
n
. (2.133)
We have as many equations as unknowns.
78 CHAPTER 2. DERIVATIVES OF TENSORS
x
2
x
2
x+
2
x
2
x-
2
x
1
x+
2
x
1
x-
Figure 2.19: To be written.
( x , y ) ( x + /2 , y ) ( x - /2 , y )
( x , y + /2 )
( x , y - /2 )
Figure 2.20: To be written.
1-D: Comment: say here that we take N+1=2 points. See gure 2.22. We have
T
11...
11...
(x | x +x/2) T
11...
11...
(x | x x/2) =
_

j
T
ij...
k...
_
(x) x
j
. (2.134)
We have one equation and one unknown.
2.3 The nite volume method
This section is to become an important section of the chapter, for this is the main numerical menthod
we are going to use to estimate (covariant) derivatives.
Assume we want to numerically estimate the divergence of a vector eld
i
at a given point. We
consider a volume around the point, and write the divergence theorem (see section 5.2)
_
3D
dV
i

i
=
_
2D
dS
i

i
, (2.135)
If the vector eld is smooth enough, and the volume small enough, we can write

i
=
1
V
_
2D
dS
i

i
. (2.136)
Dividing the surface into small pieces on each of which the eld can be approximately taken as constant
gives

i
=
1
V

All facets
S
i

i
. (2.137)
Figure 2.23 illustrates this in a 2-D context.
Note: explain here how to compute all types of derivatives for all types of tensors. For instance,
if
ij
is the stress tensor of dynamics of continuous media, how to estimate
i

ij
?
-x
1
/2
x
1
/2
-x
2
/2
x
2
/2
-x
3
/2
x
3
/2
Figure 2.21: To be written.
2.3. THE FINITE VOLUME METHOD 79
x
x/2 x/2
Figure 2.22: To be written.
Figure 2.23: Top: The gradient of a tensor eld at the black point can be estimated, through the use of
the divergence theorem, by integrating the eld along the boundary, i.e, by estimating the ux of the
tensor eld. Bottom left: Usually, a polygon is chose as nite volume. Bottom right: Approximating
the values of the tensor eld at the boundary by a few points (here, the three white points), the
gradient of the tensor eld is obtained as a discrete sum (here with three terms). Note: This has to
be much better explained, as we can integrate only scalars. We have to use test vector elds.
80 CHAPTER 2. DERIVATIVES OF TENSORS
2.4 Appendices
2.4.1 Appendix: Parallel transportation
We consider here an Euclidean space with cartesian coordinates X
I
, and a working system of
coordinates x
i
that may not be cartesian (think, for instance, in the spherical coordinates for the
3-D physical space).
In Section xxx we have denoted by V(x
B
| x
A
) the vector (or tensor) at point x
B
obtained by
parallel transportation of V(x
A
) from point A to point B .
Using the working coordinates x
i
there is no obvious way of performing that parallel trans-
portation. Instead, let us transform, at point A , the components of V(x
A
) into the components in
the basis associated with the cartesian coordinates:
V
I
(X
A
) =
X
I
x
i
(x
A
) V
i
(x
A
) . (2.138)
Now, in cartesian coordinates, the parallel transportation is trivial, as tensors keep constant compo-
nents:
V
I
(X
B
| X
A
) = V
I
(X
A
) . (2.139)
We can now transform back, at point B , the components V
I
(X
B
| X
A
) into the components
corresponding to the working coordinates x
i
:
V
i
(x
B
| x
A
) =
x
i
X
I
(X
B
) V
I
(X
B
| X
A
) . (2.140)
This gives the nal result:
V
i
(x
B
| x
A
) =
x
i
X
M
(X
B
)
X
M
x
j
(x
A
) V
j
(x
A
) . (2.141)
As the coecients of parallel transportation were dened (see XXX) by
V
i
(x
B
| x
A
) =
j
i
(x
B
[x
A
)V
j
(x
A
) , (2.142)
this gives

j
i
(x
B
[x
A
) =
x
i
X
M
(X(x
B
))
X
M
x
j
(x
A
) .
(2.143)
As

kj
i
(x
B
[x
A
) =

j
i
x
k
A
(x
B
[x
A
) (2.144)
(the index A in the derivative is to recall that it concerns the second variable), this gives

kj
i
(x
B
[x
A
) =
x
i
X
M
(X(x
B
))

2
X
M
x
j
x
k
(x
A
) , (2.145)
i.e., recalling that we use the notation
kj
i
(x) for
kj
i
(x[x) ,

kj
i
(x) =
x
i
X
M
(X(x))

2
X
M
x
j
x
k
(x) .
(2.146)
2.4. APPENDICES 81
Example: We have somewhere calculated the coecients for the parallel transportation of vectors
in the Euclidean plane with polar coordinates using direct geometrical considerations. We could also
have used the result here above.
En vrac:
x
2
, x
2
= r,
X
1
, X
2
= x, y
r
2
= x
2
+y
2
tg =
y
x
x = r cos
y = r sin
_
x/r x/
y/r y/
_
=
_
cos r sin
sin r cos
_
_
r/x r/y
/x /y
_
=
_
cos sin

1
r
sin
1
r
cos
_
For instance, we have

r
r
(r
A
,
A
[r
B
,
B
) =
r
x
(r
A
,
A
)
x
r
(r
B
,
B
) +
r
y
(r
A
,
A
)
y
r
(r
B
,
B
)
= cos
A
cos
B
+ sin
A
sin
B
= cos(
B

A
) .
In total:
_

r
r
(r
A
,
A
[r
B
,
B
)
r

(r
A
,
A
[r
B
,
B
)

r
(r
A
,
A
[r
B
,
B
)

(r
A
,
A
[r
B
,
B
)
_
=
_
cos(
B

A
)
1
r
A
sin(
B

A
)
r
B
sin(
B

A
)
r
B
r
A
cos(
B

A
)
_
From this, the connection coecients can be obtained by a simple derivation.
2.4.2 Appendix: Old text: Parallel transportation
We consider here an euclidean space with cartesian coordinates y
I
, and a working system of
coordinates x
i
that may not be cartesian (think, for instance, in the spherical coordinates for the
3-D physical space).
In Section xxx we have denoted by V(x
B
| x
A
) the vector (or tensor) at point x
B
obtained by
parallel transportation of V(x
A
) from point A to point B .
Using the working coordinates x
i
there is no obvious way of performing that parallel trans-
portation. Instead, let us transform, at point A , the components of V(x
A
) into the components in
the basis associated with the cartesian coordinates:
V
I
(y
A
) =
y
I
x
i
(x
A
) V
i
(x
A
) . (2.147)
Now, in cartesian coordinates, the parallel transportation is trivial, as tensors keep constant compo-
nents:
V
I
(y
B
| y
A
) = V
I
(y
A
) . (2.148)
We can now transform back, at point B , the components V
I
(y
B
| y
A
) into the components
corresponding to the working coordinates x
i
:
V
i
(x
B
| x
A
) =
x
i
y
I
(y
B
) V
I
(y
B
| y
A
) . (2.149)
82 CHAPTER 2. DERIVATIVES OF TENSORS
This gives the nal result:
V
i
(x
B
| x
A
) =
x
i
y
I
(y
B
)
y
I
x
j
(x
A
) V
j
(x
A
) .
(2.150)
In Section xxx the partial derivatives of a vector eld where dened by the rst order development
V
i
(x +x) = V
i
(x) + (
j
V
i
)(x) x
j
+. . . , (2.151)
while the (covariant) derivatives where dened by
V
i
(x | x +x) = V
i
(x) + (
j
V
i
)(x) x
j
+. . . . (2.152)
Using equation 2.141 we have
V
i
(x | x +x) =
x
i
y
I
(y(x))
y
I
x
j
(x +x) V
j
(x +x) . (2.153)
Using equation 2.151 and the rst order development
y
I
x
j
(x +x) =
y
I
x
j
(x) +

2
y
I
x
j
x
k
(x) x
k
, (2.154)
the last equation can be written, up to the rst order,
V
i
(x | x +x) =
x
i
y
I
(y(x))
_
y
I
x
j
(x) +

2
y
I
x
j
x
k
(x) x
k
_
_
V
i
(x) + (
j
V
i
)(x) x
j
_
, (2.155)
or, using
x
i
y
I
(y(x))
y
I
x
j
(x) =
i
j
, (2.156)
we get, up to the rst order,
V
i
(x | x +x) = V
i
(x) +
_
(
k
V
i
)(x) +
x
i
y
I
(y(x))

2
y
I
x
j
x
k
(x) V
j
(x)
_
x
k
+. . . (2.157)
A direct comparison with 2.152 directly gives
(
k
V
i
)(x) = (
k
V
i
)(x) +
x
i
y
I
(y(x))

2
y
I
x
j
x
k
(x) V
j
(x) , (2.158)
or, for short,

k
V
i
=
k
V
i
+
kj
i
V
j
, (2.159)
where the connection coecients
jk
i
are dened by

jk
i
(x) =
x
i
y
I
(y(x))

2
y
I
x
j
x
k
(x) . (2.160)
For a density, equation 2.141 becomes
V
i
(x
B
| x
A
) =
1
(y
B
)
x
i
y
I
(y
B
) (x
A
)
y
I
x
j
(x
A
) V
j
(x
A
) , (2.161)
and, developing the demonstration along the same lines as before (WARNING: This has to be done),
we obtain, for the derivative of a vector density,

k
V
i
=
k
V
i
+
kj
i
V
j
+
kj
j
V
i
. (2.162)
2.4. APPENDICES 83
2.4.3 Appendix: Taylors series
The Taylors series:
Note: this appendix is, for the moment being, just an essay (i.e., a set of conjectures, to be worked
out).
V
i
(x | x +x) = V
i
(x)
+
_

j
V
i
_
(x) x
j
+
1
2
_

k
V
i
_
(x) x
j
x
k
+
1
3!
_

V
i
_
(x) x
j
x
k
x

+ . . . (2.163)
The two step Taylors series:
V
i
(x | x +x
A
| x +x
A
+x
B
) =
= V
i
(x)
+
_

j
V
i
_
(x)
_
x
j
A
+x
j
B
_
+
1
2
__

j
V
i
_
(x)
_
x
j
A
+x
j
B
_ _
x
k
A
+x
k
B
_
+
_

j

j

k
V
i
_
(x) x
j
B
x
k
A
_
+
1
3!
__

j
V
i
_
(x)
_
x
j
A
+x
j
B
_ _
x
k
A
+x
k
B
_ _
x

A
+x

B
_
+
_
(

j

j

k
+

j

j

) V
i
_
(x) x
j
B
x
k
A
x

A
+
_
(

k

k

j
+

j

j

) V
i
_
(x) x
j
B
x
k
B
x

A
_
+ . . . (2.164)
Going from one point to the other by two dierent paths:
V
i
(x | x +x
A
| x +x
A
+x
B
) V
i
(x | x +x
B
| x +x
B
+x
A
) =
=
_

j

j

k
V
i
_
(x) x
j
B
x
k
A
+
1
3!
__
(

j

j

k
+
k

j

j

+ 2

j

j

) V
i
_
(x) x
j
B
x
k
A
x

A
+
_
(
j

k

k

k

k

j
+ 2

j

j

) V
i
_
(x) x
j
B
x
k
B
x

A
_
+ . . . (2.165)
Demonstration: Taylors series
To demonstrate equation 2.163, we will assume that a development
V
i
(x | x +x) = V
i
(x) +
_
A
1
j
V
i
_
(x) x
j
+
_
A
2
jk
V
i
_
(x) x
j
x
k
+
_
A
3
jk
V
i
_
(x) x
j
x
k
x

+. . .
(2.166)
exists, where the operators A
n
are linear, and then show that it implies
A
2
jk
=
1
2
A
1
j
A
1
k
A
3
jk
=
1
3!
A
1
j
A
1
k
A
1

. . . A
n
i
1
i
2
...i
n1
i
n =
1
n!
A
1
i
1A
1
i
2 . . . A
1
i
n1A
1
i
n . (2.167)
84 CHAPTER 2. DERIVATIVES OF TENSORS
As we know that, at rst order,
V
i
(x | x +x) = V
i
(x) +
_

j
V
i
_
(x) x
j
+. . . , (2.168)
then
A
1
j
=
j
, (2.169)
and equation 2.163 will be demonstrated.
First, let us write equation 2.166 using the variable y :
V
i
(y | y +y) = V
i
(y) +
_
A
1
j
V
i
_
(y) y
j
+
_
A
2
jk
V
i
_
(y) y
j
y
k
+
_
A
3
jk
V
i
_
(y) y
j
y
k
y

+. . . .
(2.170)
Setting there y = x +x/2 and y = x/2 gives
V
i
(x +
x
2
| x +x) = V
i
(x +
x
2
)
+
_
A
1
j
V
i
_
(x +
x
2
)
x
j
2
+
_
A
2
jk
V
i
_
(x +
x
2
)
x
j
2
x
k
2
+
_
A
3
jk
V
i
_
(x +
x
2
)
x
j
2
x
k
2
x

2
+ . . . . (2.171)
Transporting this equality from point x +x/2 to point x gives
V
i
(x | x +
x
2
| x +x) = V
i
(x | x +
x
2
)
+
_
A
1
j
V
i
_
(x | x +
x
2
)
x
j
2
+
_
A
2
jk
V
i
_
(x | x +
x
2
)
x
j
2
x
k
2
+
_
A
3
jk
V
i
_
(x | x +
x
2
)
x
j
2
x
k
2
x

2
+ . . . , (2.172)
i.e., as V
i
(x | x +x/2 | x +x) = V
i
(x | x +x) (transportation along a geodesic),
V
i
(x | x +x) = V
i
(x | x +
x
2
)
+
_
A
1
j
V
i
_
(x | x +
x
2
)
x
j
2
+
_
A
2
jk
V
i
_
(x | x +
x
2
)
x
j
2
x
k
2
+
_
A
3
jk
V
i
_
(x | x +
x
2
)
x
j
2
x
k
2
x

2
+ . . . . (2.173)
Developing each term gives
V
i
(x | x +x) = V
i
(x) +
_
A
1
j
V
i
_
(x)
x
j
2
+
_
A
2
jk
V
i
_
(x)
x
j
2
x
k
2
+
_
A
3
jk
V
i
_
(x)
x
j
2
x
k
2
x

2
+. . .
2.4. APPENDICES 85
+ A
1
j
_
V
i
(x) +
_
A
1
k
V
i
_
(x)
x
k
2
+
_
A
2
k
V
i
_
(x)
x
k
2
x

2
+. . .
_
x
j
2
+ A
2
jk
_
V
i
(x) +
_
A
1

V
i
_
(x)
x

2
+. . .
_
x
j
2
x
k
2
+ A
3
jk
_
V
i
(x) +. . .
_
x
j
2
x
k
2
x

2
+ . . . , (2.174)
i.e., grouping the terms of same order,
V
i
(x | x +x) = V
i
(x)
+
1
2
_
A
1
j
+A
1
j
_
V
i
(x) x
j
+
1
4
_
A
2
jk
+A
1
j
A
1
k
+A
2
jk
_
V
i
(x) x
j
x
k
+
1
8
_
A
3
jk
+A
1
j
A
2
k
+A
2
jk
A
1

+A
3
jk
_
V
i
(x) x
j
x
k
x

+ . . . . (2.175)
A comparison of this with equation 2.166 directly leads to equation 2.167.
Demonstration: The two step Taylors series
Comment: In what follows, I do not think there is any need in having the norm |x| dened, but
then, I do not know which notation should I use to replace O(|x|
n
) .
We start with
V
i
(x | x +x) = V
i
(x) +
_

j
V
i
_
(x) x
j
+
1
2
_

j
V
i
_
(x) x
j
x
k
+O(|x|
3
) (2.176)
and
_

j
V
i
_
(x | x +x) =
_

j
V
i
_
(x) +
_

j
V
i
_
(x) x
k
+O(|x|
2
) . (2.177)
Replacing x by x +x
A
and x by x
B
in equation 2.176 gives
V
i
(x +x
A
| x +x
A
+x
B
) =
V
i
(x +x
A
) +
_

j
V
i
_
(x +x
A
) x
j
B
+
1
2
_

j
V
i
_
(x +x
A
) x
j
B
x
k
B
+O(|x|
3
) , (2.178)
and a parallel transportation of this equation to point x gives
V
i
(x | x +x
A
| x +x
A
+x
B
) =
V
i
(x | x+x
A
)+
_

j
V
i
_
(x | x+x
A
) x
j
B
+
1
2
_

j
V
i
_
(x | x+x
A
) x
j
B
x
k
B
+O(|x|
3
) . (2.179)
Replacing there the developments 2.176 and 2.177 gives
V
i
(x | x +x
A
| x +x
A
+x
B
) = V
i
(x) +
_

j
V
i
_
(x) x
j
A
+
1
2
_

j
V
i
_
(x) x
j
A
x
k
A
+O(|x|
3
)
+
_

j
V
i
_
(x)
j
B
+
_

j
V
i
_
(x) x
j
B
x
k
A
+O(|x|
3
)
+
1
2
_

j
V
i
_
(x) x
j
B
x
k
B
+O(|x|
3
)
+ O(|x|
3
) , (2.180)
86 CHAPTER 2. DERIVATIVES OF TENSORS
Figure 2.24: A vector is transported, parallel to itself, from one point to another point, follow-
ing two dierent paths. At the rst order of approximation, the two vectors thus obtained are
equal. At the seond order of approximation, the dierence between these two vectors equals
= (
k

j

j

k
) V
i
x
j
B
x
k
A
. Note: See text for an explanation. Note: explain better.
which can be written, after reordering, as
V
i
(x | x +x
A
| x +x
A
+x
B
) = V
i
(x)
+
_

j
V
i
_
(x)
_
x
j
A
+
j
B
_
+
1
2
__

j
V
i
_
(x)
_
x
j
A
+
j
B
_ _
x
k
A
+
k
B
_
+
_

j

j

k
V
i
_
(x) x
j
B
x
k
A
_
+ O(|x|
3
) . (2.181)
This third order terms in 2.164 can be found similarly, by just keeping them in the developments.
Equation 2.165 is a direct consequence of 2.164 and can be demonstrated by direct substitution.
2.4.4 Appendix: Some usual formulas of vector analysis
Let be a , b , and c vector elds, a scalar eld, and a the vector Laplacian (the Laplacian
applied to each component of the vector). The following list of identities holds:
div rot a = 0 (2.182)
rot grad = 0 (2.183)
div(a) = (grad) a +(div a) (2.184)
rot(a) = (grad) a +(rot a) (2.185)
2.4. APPENDICES 87
grad(a b) = (a )b + (b )a +a (rot b) +b (rot a) (2.186)
div(a b) = b (rot a) a (rot b) (2.187)
rot(a b) = a(div b) b(div a) + (b )a (a )b (2.188)
rot rot a = grad(div a) a . (2.189)
Using the nabla symbol everywhere, these equations become:
(a) = () a +( a) (2.190)
(a) = () a +(a) (2.191)
(a b) = (a )b + (b )a +a (b) +b (a) (2.192)
(a b) = b (a) a (b) (2.193)
(a b) = a( b) b( a) + (b )a (a )b (2.194)
(a) = ( a) a . (2.195)
The following three vector equations are also often useful:
a (b c) = b (c a) = c (a b) (2.196)
a (b c) = (a c) b (a b) c (2.197)
(a b) (c d) = a [b (c d)] = (a c)(b d) (a d)(b c) (2.198)
As, in tensor notations, the scalar product of two vectors is a b = a
i
b
i
, and the vector product
has components (a b)
i
=
ijk
a
j
b
k
(see section XXX), the identities 2.1962.198 correspond
respectively to:

ijk

j
a
k
= 0 (2.199)

ijk

k
= 0 (2.200)
a
i

ijk
b
j
c
k
= b
i

ijk
c
j
a
k
= c
i

ijk
a
j
b
k
(2.201)

ijk
a
j
(
km
b

c
m
) = (a
j
c
j
) b
i
(a
j
b
j
) c
i
(2.202)
(
ijk
a
j
b
k
)(
im
c

d
m
) = a
i
(
ijk
b
j
(
km
c

d
m
)) , (2.203)
88 CHAPTER 2. DERIVATIVES OF TENSORS
while the identities 2.1902.195 correspond respectively to

i
(a
i
) = (
i
)a
i
+(
i
a
i
) (2.204)

ijk

j
(a
k
) =
ijk
(
j
)a
k
+
ijk

j
a
k
(2.205)

i
(a
j
b
j
) = (a
j

j
)b
i
+ (b
j

j
)a
i
+
ijk
a
j
(
km

b
m
) +
ijk
b
j
(
km

a
m
) (2.206)

i
(
ijk
a
j
b
k
) = b
k

kij

i
a
j
a
j

jik

i
b
k
(2.207)

ijk

j
(
km
a

b
m
)a
i

j
b
j
b
i

j
a
j
+b
j

j
a
i
a
j

j
b
i
(2.208)

ijk

j
(
km

a
m
) =
i
(
j
a
j
)
j

j
a
i
, (2.209)
where the (inelegant) notation
i
represents g
ij

j
.
The truth of the set of equations 2.1992.209, when not obvious, is easily demonstrated by the
simple use of the property (see section XXX)

ijk

km
=
i

j
m

i
m

(2.210)
Comment: I am assuming that the nabla symbol and the
ijk
symbol conmute. Is that true?
2.4.5 Appendix: Transporting the basis along Meridians and Parallels
Let us nd here the formulas for the parallel transportation of the basis vectors at the surface of the
sphere.
The metric:
_
g

(, ) g

(, )
g

(, ) g

(, )
_
=
_
R
2
0
0 R
2
sin
2

_
(2.211)
Notice thet the length of the basis vectors is, at point (, ) , |e

| = R and |e

| = R sin .
The connection:
_

(, )

(, )

(, )

(, )
_
=
_
0 0
0 sin cos
_
(2.212)
_

(, )

(, )

(, )

(, )
_
=
_
0 cotg
cotg 0
_
(2.213)
General rst order transportation formula:
e
i
(x | x +x) = e
i
(x) +
ki
j
(x) x
k
e
j
(x) +. . . (2.214)
Special case for the sphere:
e

(, | +, +) = e

(, ) + cotg e

(, ) (2.215)
e

(, | +, +) = e

(, ) + cotg e

(, ) sin cos e

(, ) (2.216)
2.4. APPENDICES 89
Along a Meridian:
e

(, | +, ) = e

(, ) (2.217)
e

(, | +, ) = e

(, ) + cotg e

(, ) (2.218)
This integrates to
e

(, |

, ) = e

(, ) (2.219)
e

(, |

, ) =
sin

sin
e

(, ) (2.220)
An easy computation shows that the length of the transported vectors is
|e

(, |

, )| = R
|e

(, |

, )| = R sin

,
i.e., the length of the basis vectors at the original point.
Along a Parallel:
e

(, | , +) = e

(, ) + cotg e

(, ) (2.221)
e

(, | , +) = e

(, ) sin cos e

(, ) (2.222)
This integrates to
e

(, | ,

) = cos
__

_
cos

(, )
1
sin
sin
__

_
cos

(, )
e

(, | ,

) = sin sin
__

_
cos

(, ) + cos
__

_
cos

(, ) . (2.223)
An easy computation shows that the length of the transported vectors is
|e

(, | ,

)| = R
|e

(, | ,

)| = R sin ,
i.e., the length of the basis vectors at the original point.
2.4.6 Appendix: Turning around the Pole
Let us use the results of the previous appendix to transport a basis a whole turn around the pole,
following a parallel (gure 2.25).
Using equation 2.223 we obtain
e

(, | , 2) = cos (2 cos ) e

(, )
1
sin
sin (2 cos ) e

(, )
e

(, | , 2) = sin sin (2 cos ) e

(, ) + cos (2 cos ) e

(, ) . (2.224)
This shows that the vectors of the transported basis turn, with respect to the local natural basis,
with period 2 cos . Near the Pole, cos 1 , and the period is close to 2 . But if the transported
natural basis turns with respect to the local basis by approximately one turn per turn, this is mainly
due to the fact that the natural basis itself turns, not so the transported basis (see gure 2.26. That
the parallel transportation of the basis vectors around the Pole will only produce, at most, a small
turn, is understandable, as near the Pole, we can assimilate the surface of the sphere to its tangent
90 CHAPTER 2. DERIVATIVES OF TENSORS
Figure 2.25: To be written.
Figure 2.26: Left: The natural basis at some points near the Pole. Right: The parallel transport of
a basis around the Pole, along a Parallel. Near the Pole, the basis only performs a slow turn (in the
limit, for a Parallel tending to the Pole, no turn at all). See text.
(Euclidean) plane, and that the parallel transport of vectors in an Euclidean space does not provoke
any turn.
Comment: too much turn, probably not well used.
If the local basis turns the angle 2 , and the transported basis turns the angle 2 cos with
respect to the local basis, the absolute turn of the transported basis is
= 2(1 cos ) (2.225)
( is the colatitude of the Parallel). A Taylors development gives
=
_

4
12
+. . .
_
. (2.226)
Near the Pole, 0 , and 0 . Let be the second order approximation to . We have
=
2
. (2.227)
As the radius of the small circle around the Pole is = R ,
=

2
R
2
. (2.228)
This rotation undergone (?) by the basis vectors (or by any vector) in a trip around the Pole (or
around any point at the surface of a sphere) attests the curvature of the space. It is a second order
eect.
2.4.7 Appendix: Derivatives of the basis vectors
By denition of dual basis,
e
i
e
j
=
i
j
. (2.229)
Therefore, the components of the basis vectors are
(e
i
)
j
= e
i
e
j
=
i
j
. (2.230)
2.4. APPENDICES 91
The parallel transport of the Kroneckers tensor gives

i
k
(x | y) =
i
p
(x[y)
p
m
(y)
m
k
(x[y)
=
i
p
(x[y)
p
m

m
k
(x[y)
=
i
j
(x[y)
j
k
(x[y)
=
i
k
. (2.231)
The denition of derivative then leads to

j
k
= 0 . (2.232)
As
j
k
is the kth component of e
j
, we can write

i
(e
j
)
k
= 0 (2.233)
or, for short,

i
e
j
= 0 . (2.234)
This equation is consistent with the introduction of a metric, as, then
g
ij
= e
i
e
j
, (2.235)
and the standard derivation rules lead to

i
g
jk
=
i
(e
j
e
k
) = e
j
(
i
e
k
) + (
i
e
j
)e
k
= 0 . (2.236)
It is also consistent with the blind application of the derivation formula (demonstrated for components
of tensors only) to a vector:

i
e
j
=
i
e
j

ij
k
e
k
, (2.237)
as

i
e
j
=
ij
k
e
k
. (2.238)
The danger comes from the following fallacious argument. Equation
(e
i
)
j
(x | x +x) = (e
i
)
j
(x) +
_

m
(e
i
)
j
_
(x) x
m
+. . . (2.239)
if interpreted as dening the derivative of the Kroneckers tensor,

i
j
(x | x +x) =
i
j
(x) +
_

i
j
_
(x) x
m
+. . . (2.240)
(which is zero) is perfectly valid, as the parallel transport of the Kroneckers tensor gives the Kro-
neckers tensor (see just above). If interpreted as if we could drop the component j :
e
i
(x | x +x) = e
i
(x) + (
m
e
i
) (x) x
m
+. . . (2.241)
makes no sense. For, the parallel transport of a basis vector does not give the basis vector.
92 CHAPTER 2. DERIVATIVES OF TENSORS
A small demonstration
Comment: I have not found a direct demonstration of
i
e
j
=
ij
k
e
k
. Rather, I can demonstrate
that if an expression of the type
i
e
j
= A
ij
k
e
k
exists, then A
ij
k
=
ij
k
.
For from

m
g
ij
=
mij
+
mji
(2.242)
and
g
ij
= e
i
e
j
(2.243)
it follows

m
(e
i
e
j
) =
mij
+
mji
, (2.244)
e
j

m
e
i
+e
i

m
e
j
=
mij
+
mji
, (2.245)
e
j
A
mi
k
e
k
+e
i
A
mj
k
e
k
=
mij
+
mji
, (2.246)
g
jk
A
mi
k
+g
ik
A
mj
k
=
mij
+
mji
, (2.247)
A
mij
+A
mji
=
mij
+
mji
. (2.248)
Warning: this only demonstrates the equality between the symmetric part of A
ij
k
and that of
ij
k
.
We still need to demonstrate the equality of the antisymmetric parts. Comment: The antisymetric
part of
ij
k
is a tensor: the torsion tensor.
2.4.8 Appendix: Parallel transport at the surface of a sphere
Consider the surface of a sphere, with the spherical coordinates , , and let V(, ) be a vector
eld dened on it.
Demonstrate that the vector V( + , + | , ) , i.e., the vector obtained at point ( +
, +) by parallel transport of the vector V(, ) , has components
V

( +, + | , ) = V

(, ) + sin cos V

(, ) (2.249)
and
V

( +, + | , ) = V

(, ) cotg
_
V

(, ) +V

(, )
_
. (2.250)
Demonstrate also that, along a meridian (,
0
) , these nite-dierence expressions give, for a
nite displacement from point (
0
,
0
) ,
V

(,
0
|
0
,
0
) = V

(
0
,
0
) (2.251)
and
V

(,
0
|
0
,
0
) =
sin
0
sin
V

(
0
,
0
) . (2.252)
Demonstrate nally that, along a parallel (
0
, ) , this gives
V

(
0
, |
0
,
0
) = A
0
sin (cos
0
(
0
)) (2.253)
V

(
0
, |
0
,
0
) =
A
0
sin
0
cos (cos
0
(
0
)) , (2.254)
where
A
0
= sin
V

(
0
,
0
)
sin
_
tg
1
_
1
sin
0
V

(
0
,
0
)
V

(
0
,
0
)
__ (2.255)
2.4. APPENDICES 93
and

0
=
0

1
cos
0
tg
1
_
1
sin
0
V

(
0
,
0
)
V

(
0
,
0
)
_
. (2.256)
(Comment: Say somewhere that this means that for a parallel transport along a parallel of the
sphere, a vector transported along the equator does not turn, while a vector transported along a
parallel of the sphere near the poles makes one turn per turn. For any other parallel of the sphere,
the vector turns with a velocity between 0 and one turn per turn.
Solution. (Comment: this demonstration uses the connection coecients. I should make one with-
out them).
As we have seen in xxx, the parallel transport of a vector is dened by
V
i
(x +x | x) = V
i
(x)
kj
i
(x) V
j
(x) x. (2.257)
We have also seen in equations 2.329 that the only non-vanishing connection coecients are

= cotg ;

= cotg ;

= sin cos . (2.258)

Equations 2.2492.250 follow directly from this.
Obtaining equations 2.2512.256 is just a matter of very classical integration of dierential equa-
tions. Its truth can simply be demonstrated by direct dierentiation and verication that, at point
(
0
,
0
) , the expressions degenerate into V

(
0
,
0
) and V

(
0
,
0
) .
(Comment: formulas checked February 19, 1993).
2.4.9 Appendix: Cartesian coordinates: Metric, Connection. . .
Line element
ds
2
= dx
2
+dy
2
+dz
2
(2.259)
Metric
_
_
g
xx
g
xy
g
xz
g
yx
g
yy
g
yz
g
zx
g
zy
g
zz
_
_
=
_
_
1 0 0
0 1 0
0 0 1
_
_
(2.260)
Fundamental density
g = 1 (2.261)
Connection
_
_

xx
x

xy
x

xz
x

yx
x

yy
x

yz
x

zx
x

zy
x

zz
x
_
_
=
_
_
0 0 0
0 0 0
0 0 0
_
_
_
_

xx
y

xy
y

xz
y

yx
y

yy
y

yz
y

zx
y

zy
y

zz
y
_
_
=
_
_
0 0 0
0 0 0
0 0 0
_
_
_
_

xx
z

xy
z

xz
z

yx
z

yy
z

yz
z

zx
z

zy
z

zz
z
_
_
=
_
_
0 0 0
0 0 0
0 0 0
_
_
(2.262)
94 CHAPTER 2. DERIVATIVES OF TENSORS
Contracted connection
_
_

z
_
_
=
_
_
0
0
0
_
_
(2.263)
Relationship between covariant and contravariant components for rst order tensors
_
_
V
x
V
y
V
z
_
_
=
_
_
V
x
V
y
V
z
_
_
(2.264)
Relationship between covariant and contravariant components for second order tensors
_
_
T
xx
T
xy
T
xz
T
yx
T
yy
T
yz
T
zx
T
zy
T
zz
_
_
=
_
_
T
x
x
T
x
y
T
x
z
T
y
x
T
y
y
T
y
z
T
z
x
T
z
y
T
z
z
_
_
=
_
_
T
xx
T
xy
T
xz
T
yx
T
yy
T
yz
T
zx
T
zy
T
zz
_
_
(2.265)
Norm of the vectors of the natural basis
|e
x
| = |e
y
| = |e
z
| = 1 (2.266)
Norm of the vectors of the normed basis
| e
x
| = | e
y
| = | e
z
| = 1 (2.267)
Missing Comment: give also the norms of the vectors of the dual basis.
Relations between components on the natural and the normed basis for rst order tensors
_
_
V
x
V
y
V
z
_
_
=
_
_

V
x

V
y

V
z
_
_
;
_
_
V
x
V
y
V
z
_
_
=
_
_

V
x

V
y

V
z
_
_
(2.268)
Relations between components on the natural and the normed basis for second order
tensors
_
_
T
xx
T
xy
T
xz
T
yx
T
yy
T
yz
T
zx
T
zy
T
zz
_
_
=
_
_

T
xx

T
xy

T
xz

T
yx

T
yy

T
yz

T
zx

T
zy

T
zz
_
_
_
_
T
x
x
T
x
y
T
x
z
T
y
x
T
y
y
T
y
z
T
z
x
T
z
y
T
z
z
_
_
=
_
_

T
x
x
T
x
y
T
x
z

T
y
x
T
y
y
T
y
z

T
z
x
T
z
y
T
z
z
_
_
_
_
T
xx
T
xy
T
xz
T
yx
T
yy
T
yz
T
zx
T
zy
T
zz
_
_
=
_
_

T
xx
T
xy
T
xz

T
yx
T
yy
T
yz

T
zx
T
zy
T
zz
_
_
(2.269)
2.4.10 Appendix: Spherical coordinates: Metric, Connection. . .
Line element
ds
2
= dr
2
+r
2
d
2
+r
2
sin
2
d
2
(2.270)
2.4. APPENDICES 95
Metric
_
_
g
rr
g
r
g
r
g
r
g

g
r
g

_
_
=
_
_
1 0 0
0 r
2
0
0 0 r
2
sin
2

_
_
(2.271)
Fundamental density
g = r
2
sin (2.272)
Connection
_
_

rr
r

r
r

r
r

r
r

r
r

r
_
_
=
_
_
0 0 0
0 r 0
0 0 r sin
2

_
_
_
_

rr

_
_
=
_
_
0 1/r 0
1/r 0 0
0 0 sin cos
_
_
_
_

rr

_
_
=
_
_
0 0 1/r
0 0 cotg
1/r cotg 0
_
_
(2.273)
Contracted connection
_
_

_
_
=
_
_
2/r
cotg
0
_
_
(2.274)
Relationship between covariant and contravariant components for rst order tensors
_
_
V
r
V

_
_
=
_
_
V
r
r
2
V

r
2
sin
2
V

_
_
(2.275)
Relationship between covariant and contravariant components for second order tensors
_
_
_
T
rr
1
r
2
T
r
1
r
2
sin
2

T
r
T
r
1
r
2
T

1
r
2
sin
2

T
r
1
r
2
T

1
r
2
sin
2

_
_
_ =
_
_
T
r
r
T
r

T
r

r
T

r
T

_
_
=
_
_
T
rr
T
r
T
r
r
2
T
r
r
2
T

r
2
T

r
2
sin
2
T
r
r
2
sin
2
T

r
2
sin
2
T

_
_
(2.276)
Norm of the vectors of the natural basis
|e
r
| = 1 ; |e

| = r ; |e

| = r sin (2.277)
Norm of the vectors of the normed basis
| e
r
| = | e

| = | e

| = 1 (2.278)
Missing Comment: give also the norms of the vectors of the dual basis.
Relations between components on the natural and the normed basis for rst order tensors
_
_
V
r
V

_
_
=
_
_

V
r
r

V

r sin

V

_
_
;
_
_
V
r
V

_
_
=
_
_

V
r
1
r

1
r sin

_
_
(2.279)
96 CHAPTER 2. DERIVATIVES OF TENSORS
Relations between components on the natural and the normed basis for second order
tensors
_
_
T
rr
T
r
T
r
T
r
T

T
r
T

_
_
=
_
_

T
rr
r

T
r
r sin

T
r
r

T
r
r
2
T

r
2
sin

T

r sin

T
r
r
2
sin

T

r
2
sin
2

_
_
_
_
T
r
r
T
r

T
r

r
T

r
T

_
_
=
_
_
_

T
r
r 1
r

T
r
1
r sin

T
r

r
T

1
sin

r sin

T

r
sin

T

_
_
_
_
_
T
rr
T
r
T
r
T
r
T

T
r
T

_
_
=
_
_
_

T
rr 1
r

T
r 1
r sin

T
r
1
r

T
r 1
r
2

T
1
r
2
sin

1
r sin

T
r 1
r
2
sin

T
1
r
2
sin
2

_
_
_ (2.280)
Note: say somewhere in this appendix that the two following formulas are quite useful in deriving
the formulas above.
1
r
n

r
(r
n
) =

r
+
n
r
(2.281)
1
sin
n

(sin
n
) =

+ncotg . (2.282)
2.4.11 Appendix: Cylindrical coordinates: Metric, Connection. . .
Line element
ds
2
= dr
2
+r
2
d
2
+dz
2
(2.283)
Metric
_
_
g
rr
g
r
g
rz
g
r
g

g
z
g
zr
g
z
g
zz
_
_
=
_
_
1 0 0
0 r
2
0
0 0 1
_
_
(2.284)
Fundamental density
g = r (2.285)
Connection
_
_

rr
r

r
r

rz
r

r
r

z
r

zr
r

z
r

zz
r
_
_
=
_
_
0 0 0
0 r 0
0 0 0
_
_
_
_

rr

rz

zr

zz

_
_
=
_
_
0 1/r 0
1/r 0 0
0 0 0
_
_
_
_

rr
z

r
z

rz
z

r
z

z
z

zr
z

z
z

zz
z
_
_
=
_
_
0 0 0
0 0 0
0 0 0
_
_
(2.286)
Contracted connection
_
_

z
_
_
=
_
_
1/r
0
0
_
_
(2.287)
2.4. APPENDICES 97
Relationship between covariant and contravariant components for rst order tensors
_
_
V
r
V

V
z
_
_
=
_
_
V
r
r
2
V

V
z
_
_
(2.288)
Relationship between covariant and contravariant components for second order tensors
_
_
T
rr
1
r
2
T
r
T
rz
T
r
1
r
2
T

T
z
T
zr
1
r
2
T
z
T
zz
_
_
=
_
_
T
r
r
T
r

T
r
z
T

r
T

z
T
z
r
T
z

T
z
z
_
_
=
_
_
T
rr
T
r
T
rz
r
2
T
r
r
2
T

r
2
T
z
T
zr
T
z
T
zz
_
_
(2.289)
Norm of the vectors of the natural basis
|e
r
| = 1 ; |e

| = r ; |e
z
| = 1 (2.290)
Norm of the vectors of the normed basis
| e
r
| = | e

| = | e
z
| = 1 (2.291)
Missing Comment: give also the norms of the vectors of the dual basis.
Relations between components on the natural and the normed basis for rst order tensors
_
_
V
r
V

V
z
_
_
=
_
_

V
r
r

V

V
z
_
_
;
_
_
V
r
V

V
z
_
_
=
_
_

V
r
1
r

V
z
_
_
(2.292)
Relations between components on the natural and the normed basis for second order
tensors
_
_
T
rr
T
r
T
rz
T
r
T

T
z
T
zr
T
z
T
zz
_
_
=
_
_

T
rr
r

T
r

T
rz
r

T
r
r
2
T

T
z

T
zr
r

T
z

T
zz
_
_
_
_
T
r
r
T
r

T
r
z
T

r
T

z
T
z
r
T
z

T
z
z
_
_
=
_
_

T
r
r 1
r

T
r

T
r
z
r

r
T

T
z
r 1
r

T
z

T
z
z
_
_
_
_
T
rr
T
r
T
rz
T
r
T

T
z
T
zr
T
z
T
zz
_
_
=
_
_
_

T
rr 1
r

T
r
T
rz
1
r

T
r 1
r
2

T
1
r

T
z

T
zr 1
r

T
z
T
zz
_
_
_ (2.293)
2.4.12 Appendix: Gradient, Divergence and Curl in usual coordinate systems
Here we analyze the 3-D Euclidean space, using Cartesian, spherical or cylindrical coordinates. The
words scalar, vector, and tensor mean true scalars, vectors and tensors, respectively. The scalar
densities, vector densities and tensor densities (see section XXX) are named explicitly.
98 CHAPTER 2. DERIVATIVES OF TENSORS
Denitions
If x (x) is a scalar eld, its gradient is the form dened by
G
i
=
i
. (2.294)
If x V
i
(x) is a vector density eld, its divergence is the scalar density dened by
D =
i
V
i
. (2.295)
If x F
i
(x) is a form eld, its curl (or rotational ) is the vector density dened by
R
i
=
ijk

j
F
k
. (2.296)
Properties
These denitions are such that we can replace everywhere true (covariant) derivatives by partial
derivatives (see exercise XXX). This gives, for the gradient of a density,
G
i
=
i
=
i
, (2.297)
for the divergence of a vector density,
D =
i
V
i
=
i
V
i
, (2.298)
and for the curl of a form,
R
i
=
ijk

j
F
k
=
ijk

j
F
k
(2.299)
[this equation is only valid for spaces without torsion; the general formula is R
i
=
ijk

j
F
k
=

ijk
(
j
F
k

1
2
S
jk

) ].
These equations lead to particularly simple expressions. For instance, the following table shows
that the explicit expressions have the same form for Cartesian, spherical and cylindrical coordinates
(or for whatever coordinate system).
Cartesian Spherical Cylindrical
G
x
=
x
G
r
=
r
G
r
=
r

Gradient G
y
=
y
G

G
z
=
z
G

G
z
=
z

Divergence D =
x
V
x
+
y
V
y
+
z
V
z
D =
r
V
r
+

D =
r
V
r
+

+
z
V
z
R
x
=
y
F
z

z
F
y
R
r
=

R
r
=

F
z

z
F

Curl R
y
=
z
F
x

x
F
z
R

F
r

r
F

=
z
F
r

r
F
z
R
z
=
x
F
y

y
F
x
R

=
r
F

F
r
R
z
=
r
F

F
r
Remarks
Although we have only dened the gradient of a true scalar, the divergence of a vector density, and
the curl of a form, the denitions can be immediately be extended by putting bars on and taking
bars o (see section XXX).
As an example, from equation 2.294, we can immediately write the denition of the gradient of a
scalar density,
G
i
=
i
, (2.300)
2.4. APPENDICES 99
from equation 2.295 we can write the denition of the divergence of a (true) vector eld,
D =
i
V
i
, (2.301)
and from equation 2.296 we can write the denition of the curl of a form as a true vector,
R
i
=
ijk

j
F
k
, (2.302)
or a true form,
R

= g
i

ijk

j
F
k
. (2.303)
Although equation 2.301 seems well adapted to the practical computation of the divergence of a
true vector, it is better to use 2.298 instead. For we have successively
D =
i
V
i
g D =
i
(g V
i
) D =
1
g

i
(g V
i
) . (2.304)
This last expression provides directly compact expressions for the divergence of a vector. For instance,
as the fundamental density g takes, in Cartesian, spherical and cylindrical coordinates, respectively
the values 1 , r
2
sin and r , this leads to the results of the following table.
Divergence, Cartesian coordinates : D =
V
x
x
+
V
y
y
+
V
z
z
(2.305)
Divergence, Spherical coordinates : D =
1
r
2
(r
2
V
r
)
r
+
1
sin
(sin V

+
V

(2.306)
Divergence, Cylindrical coordinates : D =
1
r
(rV
r
)
r
+
V

+
V
z
z
(2.307)
Replacing the components on the natural basis by the components on the normed basis (see section
XXX) gives
Divergence, Cartesian coordinates : D =

V
x
x
+

V
y
y
+

V
z
z
(2.308)
Divergence, Spherical coordinates : D =
1
r
2
(r
2
V
r
)
r
+
1
r sin
(sin

V

+
1
r sin

(2.309)
Divergence, Cylindrical coordinates : D =
1
r
(r

V
r
)
r
+
1
r

V
z
z
(2.310)
These are the formulas given in elementary texts (not using tensor concepts).
Similarly, although 2.303 seems well adapted to a practical computation of the curl, it is better to
go back to equation 2.299. We have, successively,
R
i
=
ijk

j
F
k
g R
i
=
ijk

j
F
k
R
i
=
1
g

ijk

j
F
k
R

=
1
g
g
i

ijk

j
F
k
.
(2.311)
This last expression provides directly compact expressions for the curl. For instance, as the fun-
damental density g takes, in Cartesian, spherical and cylindrical coordinates, respectively the values
1 , r
2
sin and r , this leads to the results of the following table.
R
x
=
y
F
z

z
F
y
Curl, Cartesian coordinates : R
y
=
z
F
x

x
F
z
(2.312)
100 CHAPTER 2. DERIVATIVES OF TENSORS
R
z
=
x
F
y

y
F
x
R
r
=
1
r
2
sin
(

)
Curl, Spherical coordinates : R

=
1
sin
(

F
r

r
F

) (2.313)
R

= sin (
r
F

F
r
)
R
r
=
1
r
(

F
z

z
F

)
Curl, Cylindrical coordinates : R

= r(
z
F
r

r
F
z
) (2.314)
R
z
=
1
r
(
r
F

F
r
)
Replacing the components on the natural basis by the components on the normed basis (see section
XXX) gives

R
x
=
y

F
z

z

F
y
Curl, Cartesian coordinates :

R
y
=
z

F
x

F
z
(2.315)

R
z
=
x

F
y

y

F
x

R
r
=
1
r sin
_
(sin

_
Curl, Spherical coordinates :

R

=
1
r
_
1
sin

F
r

(r

)
r
_
(2.316)

=
1
r
_
(r

)
r



F
r

R
r
=
1
r
_

F
z

(r

)
z
_
Curl, Cylindrical coordinates :

R

=


F
r
z



F
z
r
(2.317)

R
z
=
1
r
_
(r

)
r



F
r

_
These are the formulas given in elementary texts (not using tensor concepts).
Comment: I should remember not to put this back in a table, as it is not very readable:
2.4. APPENDICES 101
Box 2.1 Polar coordinates
(Two-dimensional Euclidean space with non-Cartesian coordinates).
ds
2
= r
2
+r
2
d
2
(2.320)

= 1/r ;
r

= 1/r ;

r
= r ; (the others vanish) (2.321)
R
ij
= 0 (2.322)

i
V
i
=
1
r

r
(rV
r
) +
V

(2.323)
Curl

R
x
=
y

F
z

z

F
y
Cartesian

R
y
=
z

F
x

F
z

R
z
=
x

F
y

y

F
x

R
r
=
1
r sin
_
sin

_
Spherical

R

=
1
r
_
1
sin

Fr

F
r
_

=
1
r
(
r

R
r
=

Fz

F
z
Cylindrical

R

Fr
z

Fz
r

R
z
=
1
r
_
r

F
r

Fr

_
Comment: What follows is not very interesting and should be suppresed.
From 2.300 we can write
g G
i
=
i
( g ) , (2.318)
which leads to the formula
G
i
=
1
g

i
( g ) . (2.319)
For instance, as the fundamental density g takes, in Cartesian, spherical and cylindrical coordinates,
respectively the values 1 , r
2
sin and r , this leads to the results of the following table.
Cartesian Spherical Cylindrical

G
x
=

x

G
r
= r
2
r
_
1
r
2

_

G
r
= r

r
_
1
r

_
Gradient

G
y
=

y

= sin

_
1
sin

_

G

G
z
=

z

G
z
=

z
2.4.13 Appendix: Some boxes
(Comment: mention here the boxes with dierent coordinate systems).
102 CHAPTER 2. DERIVATIVES OF TENSORS
Box 2.2 Cylindrical coordinates
(Three-dimensional Euclidean space with non-Cartesian coordinates).
ds
2
= r
2
+r
2
d
2
+dz
2
(2.324)

= 1/r ;
r

= 1/r ;

r
= r ; (the others vanish) (2.325)
R
ij
= 0 (2.326)

i
V
i
=
1
r

r
(rV
r
) +
V

+
V
z
z
(2.327)
Box 2.3 Geographical coordinates
(Two-dimensional non-Euclidean space).
ds
2
= R
2
(d
2
+ sin
2
d
2
) (2.328)

= cotg ;

= cotg ;

= sin cos ; (the others vanish) (2.329)

R

= 1/R
2
; R

= 1/R
2
; (the others vanish) ; R = 2/R
2
(2.330)

i
V
i
=
1
sin

(sin V

) +
V

(2.331)
Box 2.4 Spherical coordinates
(Three-dimensional Euclidean space).
ds
2
= dr
2
+r
2
d
2
+r
2
sin
2
d
2
(2.332)

= 1/r ;
r

= 1/r ;
r

= 1/r ;

r
= r ;

= cotg ;
r

= 1/r ;

= cotg ;

r
= r sin
2
;

= sin cos ;
(the others vanish)
(2.333)
R
ij
= 0 (2.334)

i
V
i
=
1
r
2

r
(r
2
V
r
) +
1
sin

(sin V

) +
V

(2.335)
2.4. APPENDICES 103
2.4.14 Appendix: Computing in polar coordinates
General formula
Simple-minded computation From
div V =
1
r

r
(rV
r
) +
V

, (2.336)
we obtain, using a simple-minded discretisation, at
(div V)(r, ) =
1
r
(r +r)V
r
(r +r, ) (r r)V
r
(r r, )
2 r
+
V

(r, +) V

(r, )
2
. (2.337)
Computation through parallel transport The notion of parallel transport leads to
(div V)(r, ) =
V
r
(r, | r +r, ) V
r
(r, | r r, )
2 r
+
V

(r, | r, +) V

(r, | r, )
2
, (2.338)
which gives
(div V)(r, ) =
V
r
(r +r, ) V
r
(r r, )
2 r
+cos()
V

(r, +) V

(r, )
2
+
sin()

1
r
V
r
(r +r, ) +V
r
(r r, )
2
. (2.339)
Note: Natural basis and normed basis The components on the natural basis V
r
et V

are related with the components on the normed basis

V
r
and

V

through

V
r
= V
r
(2.340)
and

= r V

. (2.341)
Divergence of a constant eld
A constant vector eld (oriented as the x axis) has components
V
r
(r, ) = k cos (2.342)
and
V

(r, ) =
k
r
sin . (2.343)
Simple-minded computation An exact evaluation of approximation 2.337 gives
(div V)(r, ) =
k
r
cos
_
1
sin()

_
, (2.344)
expression with an error of order ()
2
.
104 CHAPTER 2. DERIVATIVES OF TENSORS
Computation through parallel transport An exact evaluation of approximation 2.339 gives
(div V)(r, ) = 0 , (2.345)
as it should.
Chapter 3
Curvature and Torsion
Note: rewrite this text. . .
We will be interested in spaces with curvature and torsion. All geometrical properties of such
spaces are perfectly dened when a torsion S
ij
k
and a metric g
ij
are given. From them, the
connection coecients
ij
k
can be computed that allow the parallel transportation of the
basis vectors:
j
e
k
= e
i

jk
i
.
105
106 CHAPTER 3. CURVATURE AND TORSION
3.1 Curvature and Torsion
Covariant derivatives do not commute in general. Rather, for an arbitrary vector eld with components
V
i
a direct computation shows that we have
(
i

j

j

i
)V
l
= R
ijk
l
V
k
(
k
V
l
) S
ij
k
, (3.1)
where
R
ijk
l
=
i

jk
l

ik
l
+
is
l

jk
s

js
l

ik
s
(3.2)
and
S
ij
k
=
ij
k

ji
k
. (3.3)
The tensor R
ijk
l
so dened is termed the curvature (Riemann) tensor. The contraction dened
by
R
ij
= R
sij
s
(3.4)
is named the Ricci tensor.
As we have not yet introduced a metric, we can not yet raise and lower indices of tensors. When
this is done, it is possible to show that, in addition to the obvious symmetry
R
ijkl
= R
jikl
, (3.5)
the curvature tensor has also the symmetries (Hehl, 19XX)
R
ijkl
= R
ijlk
(3.6)
and
R
ijkl
= R
klij
. (3.7)
Then, the curvature scalar is dened by
R = R
i
i
. (3.8)
(Comment: give here some examples illustrating the notion of curvature.)
The tensor S
ij
k
dened above is termed the torsion tensor. In spite of the fact that
ij
k
are not
the components of a tensor, those of S
ij
k
are .
(Comment: demonstrate that.)
The following contraction is introduced:
S
i
= S
si
i
. (3.9)
Notice that, by denition, if curvature and torsion vanish, then covariant derivatives commute.
(Comment: give here some examples illustrating the notion of torsion.)
3.2 The Bianchi Identities
It is easy to show (see problem [Comment: which problem?]) that, by denition, the curvature and
the torsion tensors, satisfy the following identities, termed the Bianchi identities:

[i
R
jk]q
p
= S
[ij
s
R
k]sq
p
(3.10)
and

[i
S
jk]
p
= R
[ijk]
p
+S
[ij
s
S
k]s
p
. (3.11)
where [ijk] means a sum with circular permutation: ijk +jki +kij . Explicitly,

i
R
jkq
p
+
j
R
kiq
p
+
k
R
ijq
p
= S
ij
s
R
ksq
p
+S
jk
s
R
isq
p
+S
ki
s
R
jsq
p
(3.12)
and

i
S
jk
p
+
j
S
ki
p
+
k
S
ij
p
= R
ijk
p
+R
jki
p
+R
kij
p
+S
ij
s
S
ks
p
+S
jk
s
S
is
p
+S
ki
s
S
js
p
. (3.13)
3.2. THE BIANCHI IDENTITIES 107
3.2.1 The Contracted Bianchi identities
If we have introduced a metric we can dene the contraction of the Ricci tensor,
R = g
ij
R
ij
= R
i
i
, (3.14)
called the curvature.
In the Bianchi identities 3.103.11 we can now contract the index of the covariant derivatives with
the indexes in upper position, in order to get an expression involving the divergence of some tensors.
This gives (see problem [Comment: which problem?]):

k
E
j
k
= F
j
, (3.15)
and

i
C
jk
i
= H
jk
, (3.16)
where
E
j
k
= R
j
k

1
2

j
k
R (3.17)
is called the Einstein Tensor,
C
ij
k
= S
ij
k
+S
i

j
k
S
j

i
k
(3.18)
is called the Cartan tensor, and where
F
j
= S
lr
s
(
1
2
R
js
lr
+
j
l
R
s
r
) (3.19)
and
H
jk
= (R
jk
R
kj
) +S
s
S
jk
s
. (3.20)
We will see in the next chapter that the contracted Bianchi identities are fundamental to make
the link between space-time geometry and physics.
3.2.2 The Contracted Bianchi identities: global form
An integration per parts easily shows that the equality
_
V
4
dV
4
_

j
_
E
j
k
+
_
V
4
dV
4

j
F
j
=
_
V
3
dV
3

j
A
j
(3.21)
is equivalent to
_
V
4
dV
4

j
_
F
j

k
E
j
k
_
=
_
V
3
dV
3

j
_
A
j
n
k
E
j
k
_
. (3.22)
If the condition 3.21 is satised for any eld
j
(x) , then, for any point x 1
4
,

k
E
j
k
= F
j
, (3.23)
and, for any point x 1
3
,
n
k
E
j
k
= A
j
. (3.24)
Note: explain what this means and why this is useful.
Similarly, an integration per parts easily shows that the equality
_
V
4
dV
4
_

jk
_
C
jk
i
+
_
V
4
dV
4

jk
H
jk
=
_
V
3
dV
3

jk
B
jk
(3.25)
108 CHAPTER 3. CURVATURE AND TORSION
is equivalent to
_
V
4
dV
4

jk
_
H
jk

i
C
jk
i
_
=
_
V
3
dV
3

jk
_
B
jk
n
i
C
jk
i
_
. (3.26)
If the condition 3.25 is satised for any eld
jk
(x) , then, for any point x 1
4
,

i
C
jk
i
= H
jk
, (3.27)
and, for any point x 1
3
,
n
i
C
jk
i
= B
jk
. (3.28)
Note: explain what this means and why this is useful.
Note: the symbols used here above ( and H ) may not be clever. Change?
Note: explain here the sense of these global equations, and explain that they can be used numeri-
cally to force some fuzzy knowledge to satisfy the theoretical constraints.
Note: explain here to what correspond these integrals in the 4-D space-time (actions?).
3.3 Derivatives of the Levi-Civitas tensors?
To be written?
Chapter 4
More tensors
4.1 Totally Antisymmetric tensors
A tensor is completely antisymmetric if any even permutation of indices does not change the value of
the components, and if any odd permutation of indices changes the sign of the value of the components:
t
pqr...
=
_
+ t
ijk...
if ijk . . . is an even permutation of pqr . . .
t
ijk...
if ijk . . . is an odd permutation of pqr . . . ,
(4.1)
For instance, a fourth rank tensor t
ijkl
is totally antisymmetric if
t
ijkl
= t
iklj
= t
iljk
= t
jilk
= t
jkil
= t
jlki
= t
kijl
= t
kjli
= t
klij
= t
likj
= t
ljik
= t
lkij
= t
ijlk
= t
ikjl
= t
ilkj
= t
jikl
= t
jkli
= t
jlik
= t
kilj
= t
kjil
= t
klji
= t
lijk
= t
ljki
= t
lkij
,
(4.2)
a third rank tensor t
ijk
is totally antisymmetric if
t
ijk
= t
jki
= t
kji
= t
ikj
= t
jik
= t
kji
, (4.3)
a second rank tensor t
ij
is totally antisymmetric if
t
ij
= t
ji
, (4.4)
and a rst rank tensor t
i
can always be considered totally antisymmetric.
Well known examples of totally antisymmetric tensors are the Levi-Civitas tensors of any rank,
the rank-two electromagnetic tensors, the vector product of two vectors:
c
ij
= a
i
b
j
a
j
b
i
, (4.5)
etc.
Comment: say somewhere that the Kroneckers tensors and determinants are totally antisymmet-
ric.
4.2 Dual tensors
In a space with n dimensions, let p and q be two (nonnegative) integers such that p +q = n . To
any totally antisymmetric tensor of rank p , B
i
1
...ip
, we can associate a totally antisymmetric tensor
of rank q , b
i
1
...iq
, dened by
b
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
B
j
1
...jp
. (4.6)
109
110 CHAPTER 4. MORE TENSORS
The tensor b is called the dual of B , and we write
b = Dual[B] (4.7)
or
b =

B (4.8)
From the properties of the product of Levi-Civitas tensors it follows that the dual of the dual
gives the original tensor, excepted for a sign:

B) = Dual[Dual[B]] = (1)
p(np)
B. (4.9)
For spaces with odd dimension (n = 1, 3, 5, . . .) , the product p(n p) is even, and

B) = B (spaces with odd dimension) . (4.10)

For spaces with even dimension (n = 2, 4, 6, . . .) , we have

B) = (1)
p
B (spaces with even dimension) . (4.11)
Although denition 4.6 has been written for pure tensors, it can obviously be written for densities
and capacities,
b
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
B
j
1
...jp
b
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
B
j
1
...jp
, (4.12)
or for tensor where covariant and contravariant indices have replaced each other:
d
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
D
j
1
...jp
d
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
D
j
1
...jp
d
i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
D
j
1
...jp
, (4.13)
Boxes 4.1 to 4.3 give explicitly the dual tensor relations in spaces with 2, 3, and 4 dimensions.
Example: Consider an antisymmetric tensor E
ij
in three dimensions. It has components
_
_
_
E
11
E
12
E
13
E
21
E
12
E
23
E
31
E
32
E
33
_
_
_ =
_
_
_
0 E
12
E
13
E
21
0 E
23
E
31
E
32
0
_
_
_ , (4.14)
with E
ij
= E
ji
. The denition
e
i
=
1
2!

ijk
E
jk
(4.15)
gives
_
_
_
0 E
12
E
13
E
21
0 E
23
E
31
E
32
0
_
_
_ =
_
_
_
0 e
3
e
2
e
3
0 e
1
e
2
e
1
0
_
_
_ , (4.16)
which is the classical relation between the three independent components of a 3-D antisymmetric
tensor and the components of a vector density.
4.3. EXTERIOR PRODUCT OF TENSORS 111
Box 4.1 Dual tensors in 2-D
In 2-D, we may need to take the following duals of contravariant (antisymmetric) tensors:

B
ij
=
1
0!

ij
B

B
ij
=
1
0!

ij
B

B
ij
=
1
0!

ij
B (4.19)

B
i
=
1
1!

ij
B
j
B
i
=
1
1!

ij
B
j

B
i
=
1
1!

ij
B
j
(4.20)

B =
1
2!

ij
B
ij
B =
1
2!

ij
B
ij

B =
1
2!

ij
B
ij
(4.21)
We may also need to take duals of covariant tensors:

B
ij
=
1
0!

ij
B

B
ij
=
1
0!

ij
B

B
ij
=
1
0!

ij
B (4.22)

B
i
=
1
1!

ij
B
j

B
i
=
1
1!

ij
B
j

B
i
=
1
1!

ij
B
j
(4.23)

B =
1
2!

ij
B
ij

B =
1
2!

ij
B
ij

B =
1
2!

ij
B
ij
(4.24)
As in a space with an even number of dimensions the dual of the dual of a tensor of rank
p equals (1)
p
the original tensor (see text), we have, in 2-D, that for a tensor with 0
or 2 indices,

(

B) = B , while for a tensor with 1 index,

(

B) = B .
Example: The vector product of two vectors U
i
and V
i
can be either dened as the antisymmetric
tensor
W
ij
= U
i
V
j
V
j
U
i
, (4.17)
or as the vector density
w
i
=
1
2!

ijk
U
j
V
k
. (4.18)
The two denitions are equivalent, as W
ij
and w
i
are mutually duals.
Denition 4.18 shows that the vector product of two vectors is not a pure vector, but a vector
density. Changing the sense of one axis gives a Jacobian equal to 1 , thus changing the sign of the
vector product w
i
.
4.3 Exterior Product of tensors
In a space of dimension n , let A
i
1
i
2
...ip
and B
i
1
i
2
...iq
, be two totally antisymmetric tensors with
ranks p and q such that p + q n . Note: check that total antisymmetry has been dened. The
exterior product of the two tensors is denoted
C = A B (4.43)
and is the totally antisymmetric tensor of rank p +q dened by
C
i
1
...ipj
1
...jq
=
1
(p +q)!

k
1
...kp
1
...q
i
1
...ipj
1
...jq
A
k
1
i
2
...kp
B

1
i
2
...q
. (4.44)
112 CHAPTER 4. MORE TENSORS
Box 4.2 Dual tensors in 3-D
In 3-D, we may need to take the following duals of contravariant (totally antisymmetric)
tensors:

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B (4.25)

B
ij
=
1
1!

ijk
B
k
B
ij
=
1
1!

ijk
B
k

B
ij
=
1
1!

ijk
B
k
(4.26)

B
i
=
1
2!

ijk
B
jk
B
i
=
1
2!

ijk
B
jk

B
i
=
1
2!

ijk
B
jk
(4.27)

B =
1
3!

ijk
B
ijk
B =
1
3!

ijk
B
ijk

B =
1
3!

ijk
B
ijk
(4.28)
We may also need to take duals of covariant tensors:

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B (4.29)

B
ij
=
1
1!

ijk
B
k

B
ij
=
1
1!

ijk
B
k

B
ij
=
1
1!

ijk
B
k
(4.30)

B
i
=
1
2!

ijk
B
jk

B
i
=
1
2!

ijk
B
jk

B
i
=
1
2!

ijk
B
jk
(4.31)

B =
1
3!

ijk
B
ijk

B =
1
3!

ijk
B
ijk

B =
1
3!

ijk
B
ijk
(4.32)
As in a space with an odd number of dimensions the dual of the dual of a tensor always
equals the original tensor (see text), we have, in 3-D, that for all tensors above,

(

B) = B.
4.3. EXTERIOR PRODUCT OF TENSORS 113
Box 4.3 Dual tensors in 4-D
In 4-D, we may need to take the following duals of contravariant (totally antisymmetric)
tensors:

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B (4.33)

B
ijk
=
1
1!

ijk
B

B
ijk
=
1
1!

ijk
B

B
ijk
=
1
1!

ijk
B

(4.34)

B
ij
=
1
2!

ijk
B
k
B
ij
=
1
2!

ijk
B
k

B
ij
=
1
2!

ijk
B
k
(4.35)

B
i
=
1
3!

ijk
B
jk
B
i
=
1
3!

ijk
B
jk

B
i
=
1
3!

ijk
B
jk
(4.36)

B =
1
4!

ijk
B
ijk
B =
1
4!

ijk
B
ijk

B =
1
4!

ijk
B
ijk
(4.37)
We may also need to take duals of covariant tensors:

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B

B
ijk
=
1
0!

ijk
B (4.38)

B
ijk
=
1
1!

ijk
B

B
ijk
=
1
1!

ijk
B

B
ijk
=
1
1!

ijk
B

(4.39)

B
ij
=
1
2!

ijk
B
k

B
ij
=
1
2!

ijk
B
k

B
ij
=
1
2!

ijk
B
k
(4.40)

B
i
=
1
3!

ijk
B
jk

B
i
=
1
3!

ijk
B
jk

B
i
=
1
3!

ijk
B
jk
(4.41)

B =
1
4!

ijk
B
ijk

B =
1
4!

ijk
B
ijk

B =
1
4!

ijk
B
ijk
(4.42)
As in a space with an even number of dimensions the dual of the dual of a tensor of rank
p equals (1)
p
the original tensor (see text), we have, in 4-D, that for a tensor with 0 ,
2 or 4 indices,

(

B) = B , while for a tensor with 1 or 3 indices,

(

B) = B .
114 CHAPTER 4. MORE TENSORS
Permuting the set of indices k
1
. . . k
p
by the set
1
. . .
q
in the above denition gives the
property
(A B) = (1)
pq
(B A) . (4.45)
It is also easy to see that the associativity property holds:
A (B C) = (A B) C. (4.46)
Comment: say that
j
1
j
2
...
i
1
i
2
...
are the components of the Kroneckers determinant dened in Sec-
tion 1.6.4.
Say that it equation 1.151 gives the property
(A1 A2 . . . AP)
i
1
i
2
...ip
=
1
p!

j
1
j
2
...jp
i
1
i
2
...ip
A1
j
1
A2
j
2
. . . AP
jp
. (4.47)
4.3.1 Particular cases:
It follows from equation 1.150 that the exterior product of a tensor of rank zero (a scalar) by a totally
antisymmetric tensor of any order is the simple product of the scalar by the tensor:
(A , B
i
1
...iq
) (A B)
i
1
...iq
= AB
i
1
...iq
. (4.48)
For the exterior product of two vectors we easily obtain (independently of the dimension of the
space into consideration)
(A
i
, B
i
) (A B)
ij
=
1
2
(A
i
B
j
A
j
B
i
) . (4.49)
The exterior product of a vector by a second rank (antisymmetric) tensor gives
(A
i
, B
ij
) (A B)
ijk
=
1
3
(A
i
B
jk
+A
j
B
ki
+A
k
B
ij
) . (4.50)
Finally, it can be seen that the exterior product of three vectors gives
(A
i
, B
i
, C
i
) (4.51)
(A B C)
ijk
=
1
6
(A
i
(B
j
C
k
B
k
C
j
) +A
j
(B
k
C
i
B
i
C
k
) +A
k
(B
i
C
j
B
j
C
i
))
=
1
6
(B
i
(C
j
A
k
C
k
A
j
) +B
j
(C
k
A
i
C
i
A
k
) +B
k
(C
i
A
j
C
j
A
i
))
=
1
6
(C
i
(A
j
B
k
A
k
B
j
) +C
j
(A
k
B
i
A
i
B
k
) +C
k
(A
i
B
j
A
j
B
i
)) .
Let us examine with more detail the formulas above in the special case of a 3-D space.
The dual of the exterior product of two vectors (equation 4.49) gives

(a b)
i
=
1
2

ijk
a
j
b
k
, (4.52)
i.e., one half the usual vector product of the two vectors:

(a b) =
1
2
(a b) . (4.53)
The dual of the exterior product of a vector by a second rank (antisymmetric) tensor (equation 4.50)
is

(a b) =
1
3
a
i
_
1
2!

ijk
b
jk
_
, (4.54)
4.4. EXTERIOR DERIVATIVE OF TENSORS 115
or, introducing the vector

b
i
, dual of the tensor b
ij
,

(a b) =
1
3
a
i

b
i
. (4.55)
This shows that the exterior product contains, via the duals, the contraction of a form and a vector.
Finally, the dual of the exterior product of three vectors (equation 4.51) is

(a b c) =
1
3!

ijk
a
i
b
j
c
k
, (4.56)
i.e., one sixth of the triple product of the three vectors.
Comment: explain that the triple product of three vectors is a (b c) = b (c a) = c (a b) .
4.4 Exterior Derivative of tensors
Let T be a totally antisymmetric tensor with components T
i
1
i
2
...ip
. The exterior product of nabla
with T is called the exterior derivative of T , and is denoted T :
( T)
ij
1
j
2
...jp
=
k
1

2
...p
ij
1
j
2
...jp

k
T

2
...p
. (4.57)
Here,
i
T
jk...
denotes the covariant derivative dened in section XXX.
The nabla notation allows to use direclty the formulas developed for the exterior product of
a vector by a tensor to obtain formulas for exterior derivatives. For instance, from equation 4.49 it
follows the denition of the exterior derivative of a vector
( b)
ij
=
1
2
(
i
b
j

j
b
i
) , (4.58)
or, if we use the dual (equations 4.524.53),

( b)
i
=
1
2

ijk

j
b
k
, (4.59)
i.e.,

( b) =
1
2
(b) . (4.60)
The exterior derivative of a vector equals one-half the rotational (curl) of the vector.
The exterior derivative of a second rank (antisymmetric) tensor is directly obtained from equa-
tion 4.50:
( b)
ijk
=
1
3
(
i
b
jk
+
j
b
ki
+
k
b
ij
) . (4.61)
Taking the dual of the expression and introducing the vector

b
i
, dual of the tensor b
ij
, gives (see
equation 4.55)

( b) =
1
3

i

b
i
, (4.62)
which shows that the dual of the exterior derivative of a second rank (antisymmetric) tensor equals
one-third of the divergence of the dual of the tensor. The exterior derivative contains, via the duals,
the divergence of a vector.
116 CHAPTER 4. MORE TENSORS
Chapter 5
Integration theory, in short
5.1 The volume element
Consider, in a space with n dimensions, p linearly independent vectors dr
1
, dr
2
, . . . , dr
p
. As
they are linear independent, p n .
We dene the dierential element
d
(p)
= p! (dr
1
dr
2
. . . dr
p
) . Note : should be bold (5.1)
Using equation 4.47 (Note: in fact this equation with indices changed of place) gives the components
d
(p)

i
1
...ip
=
i
1
...ip
j
1
...jp
dr
i
1
1
dr
i
2
2
. . . dr
ip
p
. (5.2)
In a space with n dimensions, the dual of the dierential element of dimension p will have q
indices, with p +q = n . The general denition of dual (equation 4.12) gives

d
(p)

i
1
...iq
=
1
p!

i
1
...iqj
1
...jp
d
(p)

j
1
...jp
(5.3)
The denition 5.2 and the property 1.152 give

d
(p)

i
1
...iq
=
i
1
...iqj
1
...jp
dr
j
1
1
dr
j
2
2
. . . dr
jp
p
. (5.4)
In order to simplify subsequent notations, it is better not to keep the notation. Instead, we will
write

d
(p)

i
1
...iq
= d
(p)

i
1
...iq
(5.5)
For reasons to be developed below, d
(p)

i
1
...iq
will be called the capacity element.
We can easily see, for instance, that the dierential elements of dimensions 0, 1, 2 and 3 have
components
d
0
= 1 (5.6)
d
1

i
= dr
i
1
(5.7)
d
2

ij
= dr
i
1
dr
j
2
dr
j
1
dr
i
2
(5.8)
d
3

ijk
= dr
i
1
(dr
j
2
dr
k
3
dr
k
2
dr
j
3
) +dr
j
1
(dr
k
2
dr
i
3
dr
i
2
dr
k
3
) +dr
k
1
(dr
i
2
dr
j
3
dr
j
2
dr
i
3
)
= dr
i
2
(dr
j
3
dr
k
1
dr
k
3
dr
j
1
) +dr
j
2
(dr
k
3
dr
i
1
dr
i
3
dr
k
1
) +dr
k
2
(dr
i
3
dr
j
1
dr
j
3
dr
i
1
)
= dr
i
3
(dr
j
1
dr
k
2
dr
k
1
dr
j
2
) +dr
j
3
(dr
k
1
dr
i
2
dr
i
1
dr
k
2
) +dr
k
3
(dr
i
1
dr
j
2
dr
j
1
dr
i
2
) . (5.9)
117
118 CHAPTER 5. INTEGRATION THEORY, IN SHORT
dr
1
dr
2
dr
3
dr
1
dr
1
dr
2
Figure 5.1: From vectors in a three-dimensional space we dene the one-dimensional capacity element
d
1

ij
=
ijk
dr
k
1
, the two-dimensional capacity element d
2

i
=
ijk
dr
j
1
dr
k
2
and the three-dimensional
capacity element d
3
=
ijk
dr
i
1
dr
j
2
dr
k
3
. In a metric space, the rank-two form d
1

ij
denes a surface
perpendicular to dr
1
and with a surface magnitude equal to the length of dr
1
. The rank-one
form d
2

i
denes a vector perpendicular to the surface dened by dr
1
and dr
2
and with length
representing the surface magnitude (the vector product of the two vectors). The rank-zero form d
3

is a scalar representing the volume dened by the three vectors dr

1
, dr
2
and dr
3
(the triple
product of the vectors). Note: clarify all this.
For a given dimemsion of the dierential element, the number of indices of the capacity elements
depends on the dimension of the space. In a three-dimensional space, for instance, we have
d
0

ijk
=
ijk
(5.10)
d
1

ij
=
ijk
dr
k
1
(5.11)
d
2

i
=
ijk
dr
j
1
dr
k
2
(5.12)
d
3
=
ijk
dr
i
1
dr
j
2
dr
k
3
. (5.13)
Note: explain that I use the notation d
(p)
but d
1
, d
2
, . . . in order not to suggest that p is a tensor
index and, at the same time, for not using too heavy notations..
Note: refer here to gure 5.1, and explain that we have, in fact, vector products of vectors and
triple products of vectors.
5.2 The Stokes theorem
Comment: I must explain here rst what integration means.
Let, in a space with n dimensions, (T) be a totally antisymmetric tensor of rank p , with
(p < n) . The Stokes theorem
_
(p+1)D
d
(p+1)

i
1
...i
p+1
( T)
i
1
...i
p+1
=
_
pD
d
(p)

i
1
...ip
T
i
1
...ip
(5.14)
holds. Here, the symbol
_
(p+1)D
d
(p+1)
stands for an integral over a p+1)-dimensional volume,
(embedded in an space of dimension n ), and
_
pD
d
(p)
for the integral over the p-dimensional
boundary of the volume.
This fundamental theorem contains, as special cases, the divergence theorem of Gauss-Ostrogradsky,
and the rotational theorem of Stokes (stricto sensu). Rather than deriving it here, we will explore its
consequences. For a demonstration, see, for instance, Von Westenholz (1981).
In a three-dimensional space (n = 3) , we may have p respectively equal to 2 , 1 and 0 . This
gives the three theorems
_
3D
d
3

ijk
( T)
ijk
=
_
2D
d
2

ij
T
ij
(5.15)
_
2D
d
2

ij
( T)
ij
=
_
1D
d
1

i
T
i
(5.16)
5.2. THE STOKES THEOREM 119
_
1D
d
1

i
( T)
i
=
_
0D
d
0
T . (5.17)
It is easy to see (appendix XXX) that these equation can be written
1
0!
_
3D
d
3

_
1
2!

ijk

i
T
jk
_
=
1
1!
_
2D
d
2

i
_
1
2!

ijk
T
jk
_
(5.18)
1
1!
_
2D
d
2

i
_
1
1!

ijk

j
T
k
_
=
1
2!
_
1D
d
1

ij
_
1
1!

ijk
T
k
_
(5.19)
1
2!
_
1D
d
1

ij
_
1
0!

ijk

k
T
_
=
1
3!
_
0D
d
0

ijk
_
1
0!

ijk
T
_
. (5.20)
Simplifying equation 5.18 and introducing the vector density t
i
, dual to the tensor T
ij
, ( i.e.,
t
i
=
1
2!

ijk
T
jk
), gives
_
3D
d
3

i
t
i
=
_
2D
d
2

i
t
i
. (5.21)
This corresponds to the divergence theorem of Gauss-Ostrogradsky: The integral over a (3-D) volume
of the divergence of a vector equals the ux of the vector across the surface bounding the volume.
It is worth to mention here that expression 5.21 has been derived without any mention to a metric
in the space. We have sen elsewhere that densities and capacities can be dened even if there is no
notion of distance. If there is a metric, then from the capacity element d
3
we can introduce the
volume element d
3
using the standard rule for putting on and taking o bars
d
3
= g d
3
, (5.22)
as well as the surface element
d
2

i
= g d
2

i
. (5.23)
d
3
is now the familiar volume inside a prism, and d
2

i
the vector (if we raise the index with the
metric) representing the surface inside a lozenge.
Equation 5.21 then gives
_
3D
d
3

i
t
i
=
_
2D
d
2

i
t
i
, (5.24)
which is the familiar form for the divergence theorem.
Keeping the compact expression for the capacity element in the lefthand side of equation 5.19, but
introducing its explicit expression in the right hand side gives, after simplication,
_
2D
d
2

i
(
ijk

j
T
k
) =
_
1D
dr
i
1
T
i
, (5.25)
which corresponds to the rotational theorem (theorem of Stokes stricto sensu): the integral of the
rotational (curl) of a vector on a surface equals the circulation of the vector along the line bounding
the surface.
Finally, introducing explicit expressions for the capacity elements at both sides of equation 5.20
gives
_
1D
dr
i
1

i
T =
_
0D
T . (5.26)
Writing this in the more familiar form gives
_
b
a
dr
i

i
T = T(b) T(a) , (5.27)
which corresponds the fundamental theorem of integral calculus: the integral over a line of the gradient
of a scalar equals the dierence of the values of the scalar at the two end-points.
120 CHAPTER 5. INTEGRATION THEORY, IN SHORT
Note: the demonstration that follows is to be put in an appendix:
In a three-dimensional space (n = 3) , we may have p respectively equal to 2 , 1 and 0 . This
gives the three theorems
_
3D
d
3

ijk
( T)
ijk
=
_
2D
d
2

ij
T
ij
(5.28)
_
2D
d
2

ij
( T)
ij
=
_
1D
d
1

i
T
i
(5.29)
_
1D
d
1

i
( T)
i
=
_
0D
d
0
T . (5.30)
Explicitly, using the results of sections 4.3 and 4.4, this gives
_
3D
d
3

ijk
1
3
(
i
T
jk
+
j
T
ki
+
k
T
ij
) =
_
2D
d
2

ij
T
ij
(5.31)
_
2D
d
2

ij
1
2
(
i
T
j

j
T
i
) =
_
1D
d
1

i
T
i
(5.32)
_
1D
d
1

i
T =
_
0D
d
0
T , (5.33)
or, we use the antisymmetry of the tensors,
_
3D
d
3

ijk

i
T
jk
=
_
2D
d
2

ij
T
ij
(5.34)
_
2D
d
2

ij

i
T
j
=
_
1D
d
1

i
T
i
(5.35)
_
1D
d
1

i
T =
_
0D
d
0
T . (5.36)
We can now introduce the capacity elements instead of the dierential elements:
1
0!
_
3D
d
3

_
1
2!

ijk

i
T
jk
_
=
1
1!
_
2D
d
2

i
_
1
2!

ijk
T
jk
_
(5.37)
1
1!
_
2D
d
2

i
_
1
1!

ijk

j
T
k
_
=
1
2!
_
1D
d
1

ij
_
1
1!

ijk
T
k
_
(5.38)
1
2!
_
1D
d
1

ij
_
1
0!

ijk

k
T
_
=
1
3!
_
0D
d
0

ijk
_
1
0!

ijk
T
_
. (5.39)
Introducing explicit expressions for the capacity elements gives
_
3D
(
jk
dr
j
1
dr
k
2
dr

3
)
i
t
i
=
_
2D
(
ijk
dr
j
1
dr
k
2
) t
i
(5.40)
_
2D
(
im
dr

1
dr
m
2
) (
ijk

j
T
k
) =
_
1D
dr
i
1
T
i
(5.41)
_
1D
dr
i
1

i
T =
_
0D
T , (5.42)
where, in equation 5.40, t
i
stands for the vector dual to the tensor T
ij
, i.e., t
i
=
1
2!

ijk
T
jk
.
Equations 5.37 and 5.40 correspond to the divergence theorem of Gauss-Ostrogradsky, equa-
tions 5.38 and 5.39 correspond to the rotational theorem of Stokes (stricto sensu), and equation 5.42,
when written in its more familiar form
_
b
a
dr
i

i
T = T(b) T(a) (5.43)
corresponds the fundamental theorem of integral calculus.
Part II
Second part: Gravitation and
Dynamics
121
Chapter 6
The 4-D space-time
Summary
The tensors describing curvature and torsion in an arbitray manifold satisfy two identities the
contracted Bianchi identities . When considering the four-dimensional space-time as a manifold
with curvature and torsion, the fundamental eld equations the Einstein-Cartan equations are
obtained by identifying the conserved quantities in the Bianchi identities with the mass tensor
and the spin tensor. Without any extra assumption, this gives a theory describing the dynamics
of continuous media (including general gravitational interactions). Simplifying hypothesis give then
classical limits (e.g., non relativistic, elastic).
6.1 Introduction
Any mathematical structure may be used to represent dierent physical objects. For instance, a 3-D
linear vector space may represent the space of forces acting on a particle, as well as the space of colors:
each vector of the basis represents then one fundamental color (e.g., yellow, cyan, magenta) with some
standard intensity.
The major Einsteins discovery was that the formalism of dierentiable manifolds (i.e., the geom-
etry of curved surfaces and spaces) may also be used to describe the 4-D space-time, and that, then,
gravitational forces can be interpreted as space-time curvature. One corollary of the theory was that
the space itself has curvature (i.e., is not euclidean).
The fact that two things as dierent as the geometry of the space, and the dynamics of matter
can be described in a single formalism was surprising at the time. Of course, it is futile to ask if this
is well understood or not. Any moderately intelligent human being can understand enough of the
formalism as to be able to make predictions on the outcome of experiments.
The claim that in the universe there are no forces, but only space-time curvature, gives a biased
description of physics. The 4-D space-time of general relativity has two very dierent sort of dimen-
sions: three of the dimensions are spatial dimensions, and one dimension is a time dimension. Vey
precisely, this means that if a suitable choice of space-time coordinates makes the space-time metric
diagonal, we will necessarily have something like ds
2
= (d x
0
)
2
+ (d x
1
)
2
+ (d x
2
)
2
+ (d x
3
)
2
: we can
say that the coordinate x
0
is the time coordinate.
At any moment we can switch between a representation of the universe as a four-dimensional space-
time continuum, with space-time curvature and torsion, where all instants of time are simultaneously
considered, and where the physical laws are 4-D geometrical laws, and a representation of the universe
as a three-dimensional space continuum (with curvature and torsion), evolving in time, and where the
physical laws invoke forces.
123
124 CHAPTER 6. THE 4-D SPACE-TIME
Those two representations are both perfectly valid. The advantage of the rst is its great compacity
of formulation. The second has the advantage of corresponding better to our perception of beings
evolving with time, and perceiving forces. We will see that, when considering all the gravitational
interactions in a continuous medium, it is the second representation that will give us the best insight
into the phenomenology involved.
The original Einsteins theory of gravitation related gravitational forces to the curvature of the
space-time, while space-time torsion was assumed to vanish. This approximation is not valid if in
addition to mass, we consider that matter possesses spin. The coupling of spin to space-time torsion
leads to the so-called Einstein-Cartan theory of gravitation, which predicts gravitational forces related
to the distribution of spin. Inside a continuous medium, the predicted spin-spin gravitational force
density is proportional to the gradient of the squared spin density, thus tending to concentrate spin. It
is then for instance possible to imagine a spherical distribution of matter where repulsive electrostatical
forces are compensated by the attractive spin-spin gravitational forces.
In the previous chapter, we have used an underbar to represent a tensor density. In this chapter
we only use true tensors (i.e., tensor densities of weight 0). The underbar will then be reserved here
to 4-D tensors, like in T

. The space components of the tensor, T

i
j
, when considered as the
components of a 3-D tensor will be denoted without the underbar: T
i
j
.
6.2 Chapters overview
Our physical 3-D space is not euclidean: it may have (time-dependent) curvature and torsion. If we
denote by g
ij
the space metric and by S
ij
k
the space torsion, then all other space properties can
be computed, as, for instance, the space connection
ij
k
or the space Riemann tensor R
ijk

.
If instead of being interested in the geometry of space, we are interested in the geometry of the
space-time, in addition to g
ij
and S
ij
k
we need a scalar eld , a vector eld I
i
, an antisymmetric
tensor eld

J
ij
, and a tensor eld K
i
j
(all these being time-dependent) in order to build the space-
time metric g

:
_
g
00
g
0j
g
i0
g
ij
_
=
_

2
0
0 g
ij
_
(6.1)
and the space-time torsion S

:
_
S
00
0
S
0j
0
S
i0
0
S
ij
0
_
=
_
0 c
2
I
j
c
2
I
i
c

J
ij
/
_
(6.2)
_
S
00
k
S
0j
k
S
i0
k
S
ij
k
_
=
_
0 cK
j
k

cK
i
k
S
ij
k
_
. (6.3)
We can then compute all the other properties of the space-time, as, for instance, the space-time
connection

or the space-time Riemann tensor R

: we have a model of the Universe (if

we are interested in cosmology) or, more simply, a model of a continuous medium. Dierent choices
for , g
ij
, I
i
,

J
ij
, K
i
j
, and S
ij
k
, give dierent models of continuous media.
The material properties of the universe are introduced through the mass (stress-energy) tensor
_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
cp
i
/ T
i
j
_
(6.4)
and the spin tensor
_
M
00
0
M
0j
0
M
i0
0
M
ij
0
_
=
_
0 c
2
t
j
c
2
t
i
c
ij
/
_
(6.5)
_
M
00
k
M
0j
k
M
i0
k
M
ij
k
_
=
_
0 c
j
k

c
i
k
M
ij
k
_
. (6.6)
6.3. MEASURING PROPER TIME 125
Here , p
i
, q
i
, and T
i
j
are respectively the mass density, the linear momentum density, the
mass ux, and the stress, while
ij
, M
ij
k
, t
i
, and
i
j
are respectively the spin density, the
moment stress, the unbalance, and the unbalance ux.
The Einstein-Cartan equations
T

=
1

_
R

1
2

R
_
, (6.7)
and
M

=
1

_
S

+S

_
, (6.8)
link then these material properties to the geometrical quantities , g
ij
, I
i
,

J
ij
, K
i
j
, and S
ij
k
,
introduced above.
In particular, this gives the conservation equations for the material properties:

+
i
q
i
= , (6.9)
p
i

+
j
T
i
j
= f
i
, (6.10)
t
i

+
j

i
j
=
i
, (6.11)

ij

+
k
M
ij
k
=
ij
, (6.12)
with the corresponding expressions for , f
i
,
i
, and
ij
.
The term f
i
is important, as it represents the force density acting on the medium. In addition to
the Newtonian gravitational force density g
i
, responsible of apples fall, it contains, a spin-spin grav-
itational force density (G/c
2
)
i
(
jk

jk
) that may have important implications for understanding
the microscopical structure of matter.
The gravitational eld g
i
will appear as follows: the eld will be the (squared root of the) ratio
between proper time and coordinate time. The scalar eld U dened (up to an additive constant)
by
= exp(U U
0
) , (6.13)
is the gravitational potential , and exactly corresponds to the classical Newtonian potential (times
c
2
). Its (negative) gradient
g
i
=
1
c
2

i
U (6.14)
is the Newtonian gravitational eld (times c
4
).
From the conservation equations for the material properties it is easy to obtain the nonrelativistic
limit. The elastic hypothesis then leads to the equations governing the propagation of elastic waves in
solids, but for a general medium, where the stress tensor is not necessarily symmetric (called Cosserats
or micro polar medium).
6.3 Measuring proper time
For long, humans thought that there was an absolute way of dening the length of a line joining two
points in space.
Today, that hypothesis has been dropped, and replaced by the hypothesis that the only absolute
possible denition is the proper time associated to every particle of the universe.
126 CHAPTER 6. THE 4-D SPACE-TIME
That our space-time has some property that is close to what we call the proper time of particles is
suggested by the consistency we get in the description of physical processes with regular evolution,
like radioactive decay of particles, pulsar rotation, assumed periodicity of electromagnetic radiation,
etc.
The only way for an observer to measure his proper time is by observing near him one of those
processes with regular evolution, and use it as a clock. The more regular the process, the more accurate
will be his estimation of proper time.
Discovering more and more regular clocks (pulsars, cesium clocks, . . . ) is an important task of
experimental physics.
Of course, we cannot exclude that a future, more precise, vision of the universe will force us to
admit that the proper time of a particle depends on some extra property of the universe in a way not
yet understood. But, for the time being, as we do not know of any experimental facts contradicting
the hypothesis, we assume the the concept of proper time of a particle has an absolute sense, and that
the proper time of the particle can be measured as precisely as we wish using clocks whose regularity
is checked by the internal consistency of the physics so obtained.
When this book is printed, physicists choose as unit of proper time the second, dened as the
duration of 9 192 631 770 periods of the electromagnetic radiation corresponding to the transition
between the two hyperne levels of the fundamental state of an atom of cesium (Comment: check that
denition when the book goes to print).
6.4 Measuring (improper) distances
For long, the denition of the unit of distance has been independent of that of time. It was made by
selecting a given object, accessible to all, and by deciding what was, by denition, its length (a given
rod in Paris measured, by denition, one meter).
Since then, a major discovery was made: that if two observers, moving one relatively to the other,
measure the speed of electromagnetic waves in the vacuum, they obtain the same value, denoted by
c .
Should we be able to dene the absolute motion of an observer, physics as seen by an abserver
absolutely at rest should be dierent to physics as seen by any other observer. Our experiments
show that this is not the case (in particular, we know that the equations describing the electromagnetic
and similar elds are the same for all observers, irrespectively of their relative velocity, as they have
to predict the same velocity of waves, c ).
In fact, there is a way of dening an absolute motion in the universe, as the 3 K background
radiation observed is interpreted as a fossil radiation from the early universe. We can say that an
observer is absolutely at rest if that radiation appears isotropic (i.e., if there is no Doppler shift
eect on the radiation).
Thus, we postulate that no signal can propagate faster than the electromagnetic waves, and cur-
rent theories predict that other sort of waves (gravitational, . . . ) and particles (with zero mass) also
propagate at that speed.
Again, we cannot exclude that, some day, we will discover ner facts (for instance, c at a given
point could weakly depend on the direction of propagation), but, for the time being, the postulate of
the existence of that limiting velocity gives a consistent physical vision of the universe.
This being so, we have two possibilities: we can continue to dene the unit of distance (say, the
meter) independently of the velocity of light, and try to made ever-more-precise measurements of c ,
or we can x c , at an arbitrary value, and try to made ever-more-precise realizations of the meter.
It is the second way that has been chosen by physicists now, by xing
c = 299 792 458 meter/second . (6.15)
6.5. DEFINING SPACE-TIME COORDINATES 127
Then, a meter is the distance traveled in the vacuum by an electromagnetic wave in 1/299 792 458
seconds.
An experimental realization of this denition shows that the line here below measures approxi-
mately one tenth of a meter (Comment: in the nal print, some reduction may be applied; check and
correct the length of the line).
This is 1/10 of a meter
If instead of keeping the book at rest in front of you, you look at it while walking (assuredly, you
should have to walk quite rapidly. . . ), you may nd that the above rod measures less than announced.
This is the so-called Lorentz contraction of length, not discussed here.
Notice that, instead of the meter, we could have chosen the second-light as unit of distance, or, if
the qualicative light is dropped (as the radian when measuring angles), we could have chosen
a time (the second) to measure distances. In the rst case, the value of c would be c = 1
second-light/second , while with the second choice it would be c = 1 , without any unit.
6.5 Dening space-time coordinates
Let us call a clock an instrument that emits signals at some predetermined time intervals.
To dene a system of space-time coordinates in some region of the space-time, we drop in the
space many clocks, in a conguration that is sort of continuous and arbitrary deformation of
a 3-D regular grid. Each clock will then carry three integers i, j, k identifying it, and which, by
denition, will denote its position. If we label the signals each clock emits by the integer n , then, the
collection of all the events n, i, j, k dene a space-time grid of points. By interpolation between
the points of the grid we dene then the space-time coordinates x
0
, x
1
, x
2
, x
3
of any event in the
space-time continuum.
By no means we need to assume that the clocks are xed or follow a particular movement, or
that the clocks emit their proper time.
In that way, we have perfectly dened a system of coordinates in a region of the space-time.
Obviously, that system of coordinates is (quite) arbitrary. Nevertheless, one coordinate (x
0
) is very
dierent from the others: the coordinate line (x
0
) can be the space-time trajectory of a particle,
while the other coordinate lines cannot (this would imply velocities higher than c .)
Later on, we will encounter an expression like g

dx

dx

, where every term in the sum has a

dimension of a squared length. As we will choose g
00
adimensional, then x
0
will have the dimension
of a length, and will dene the coordinate time, t , by
t =
x
0
c
. (6.16)
Obviously, t has a dimension of time. The dimensions of the x
i
are variable (think, for instance, that
a choice of spherical coordinates gives dierent dimensions for x
1
= r and for x
2
, x
3
= , ).
We have dened our coordinate system x
0
, x
1
, x
2
, x
3
by sprinkling the space with clocks.
Another observer may decide to use his own clocks, in his own way, to dene another coordinate
system x
0
, x
1
, x
2
, x
3
.
Pamameterizing the trajectory of a particle by an arbitrary parameter will give, in our coor-
dinate system, x

() , and, in the coordinate system of the other observer, x

() .
The proper time of the particle in the trajectory from, say,
A
to
B
, is an invariant (evaluating
it when the trajectory is followed from the rst or the second coordinate system has to give identical
results). This is very similar to the property of a distance, and means that we can introducing a metric
128 CHAPTER 6. THE 4-D SPACE-TIME
tensor in the space-time, with components g

(x
0
, x
1
, x
2
, x
3
) , in such a way that the proper time
can be computed, in each coordinate system, by the expresions

B

A
=
_

B

A
_

1
c
2
g

dx

dx

=
_

B

A
_

1
c
2
g

dx

dx

, (6.17)
or, equivalently, by

B

A
=
_

B

A
d , (6.18)
where
c
2
d
2
= g

dx

dx

= g

dx

dx

. (6.19)
6.6 The space-time metric
The component g
00
of the metric can esaily be interpreted as the (square root of the) ratio between
proper time and coordinate time (see below). The components g
ij
correspond to the metric of the
3-D space. What about the components g
0i
? A suitable (and very reasonable) choice of space-time
coordinates just makes them to vanish: we show in appendix that if our choice of space-time coordi-
nates is such that a light signal leaving a clock at (coordinate) time t
0
and reaching a neighbouring
clock at t
0
+ t , comes reected back at t
0
+ 2t , then g
0i
= 0 .
As there is no physical restriction in such a choice, we always will assume it. Then, the following
notations can be used:
_
g
00
g
0j
g
i0
g
ij
_
=
_

2
0
0 g
ij
_
. (6.20)
Obviously, g
ij
is the metric of the 3-D subspace of the 4-D space-time dened by a section x
0
= ct =
const. , i.e., the metric of the 3-D space in the classical sense. As we will later see, that space our
space may have curvature and torsion.
Using the general arguments of previous chapter, it is possible to see that the space components
of space-time tensors are 3-D tensors, and that we can raise and lower indices of those tensors with
the space metric g
ij
.
For a particle following the time coordinate line, dx
i
= 0 . Using the expression 6.19 for the time
element and the decomposition 6.20 of the 4-D metric in 3-D elds, we found that the proper time of
the particle satises then
c
2
d
2
= g

dx

dx

= g
00
dx
0
dx
0
=
2
dx
0
dx
0
=
2
c
2
dt
2
, (6.21)
i.e.,
d = dt . (6.22)
This gives a physical interpretation for as the ratio between the proper time and the coordinate
time for a particle describing a time coordinate line, i.e., for a particle at rest in the coordinate
system.
6.7. THE SPACE-TIME TORSION 129
6.7 The space-time torsion
Let us denote by S

the space-time torsion. From this 4-D tensor eld we can introduce the space
torsion S
ij
k
and three 3-D elds I
i
,

J
ij
, and K
i
j
:
_
S
00
0
S
0j
0
S
i0
0
S
ij
0
_
=
_
0 I
j
c
2
I
i
c
2
J
ij
c/
_
_
S
00
k
S
0j
k
S
i0
k
S
ij
k
_
=
_
0 K
j
k
c
K
i
k
c S
ij
k
_
. (6.23)
The space torsion S
ij
k
has the standard interpretation of the 3-D torsion of a constant-time
section of the 4-D space-time. We will later see that this torsion of the physical 3-D space is essentially
proportional to the moment stress tensor of the matter lling the space.
The 3-D eld

J
ij
will be proportional to the matter spin density, and the 3-D elds I
i
and K
i
j
will essentially be proportional to two other properties of matter: the unbalance, and the unbalance
ux.
In that sense, the introduction of space-time torsion (i.e., of spin) in the theory, does not compli-
cate the equations. The new quantities introduced are simply algeabrically related to the quantites
describing the matter properties: we do not have new eld (dierential) equations.
130 CHAPTER 6. THE 4-D SPACE-TIME
Chapter 7
Matter
Mass and spin are obvious properties of matter that are easily accommodated within the framework
of the general relativistic theory developed here. Other fundamental properties (like electric charge)
can only be introduced in an ad-hoc way.
Precisely, this means that, while mass and spin curve and twist the space-time (and the space
itself), electric charge does not. It is not forbidden by the theory that such properties may exist, but
they are not geometrized. In that sense, they are extraneous to the theory.
The theory will predict that mass and spin are conserved, or, more precisely, that they satisfyme
equations traditionally called conservation equations.
The theory will also predict that two other quantities are conserved. One will easily be identied
with the linear momentum. The other (the unbalance) will not have (to our knowledge) a classical
equivalent.
Of special interest for us will be the equation of conservation of linear momentum, as the source
term will describe the whole set of forces acting in a medium submitted to gravitational interaction.
When discussing the properties of space-time (Section 6.4), we saw that we could either choose to
arbitrarily x the unit of distance (the meter) and made ever-more-precise measurements of c , or
arbitrarily xe the value of c and made ever-more-precise realizations of the meter. Equivalently,
we can either choose to arbitrarily x the unit of mass (the kilogram) and made ever-more-precise
measurements of the gravitational constant G , or arbitrarily xe the value of G and made ever-
more-precise realizations of the kilogram. For the time being, it is the rst of these options that has
been selected, and the kilogram is the mass of a given iridioplatinum cylinder deposited at the Pavillon
de Breteuil , in S`evres, near Paris.
Choosing to x the value of G , would lead to two options: xing it to a value close to its
actual estimate, and with its present units, or xing it at G = 1 , without any unit, in which case,
the kilogram would not be an independent unit, but derived from the second and the meter. If, as
discussed before, also the value of c was xed to c = 1 , then we would measure time intervals,
distances, and masses, all in seconds.
To be complete, we should made here a complete operational denition of all matter properties
(linear momentum, spin, . . . ), but will not attempt such a task.
131
132 CHAPTER 7. MATTER
7.1 Recall: classical dynamics
The fundamental equations of the classical dynamics of continuous media are the equations describing
conservation of mass, linear momentum and angular momentum:

t
+
j
(v
j
) = 0 , (7.1)
p
i
t
+
j
(p
i
v
j

i
j
) = f
i
, (7.2)

ij
t
+
k
(
ij
v
k
m
ij
k
) =
ij
+ (
ji

ij
) . (7.3)
Here, is the mass density, v
i
the velocity of the medium, p
i
= v
i
the linear momentum density,
and
ij
the stress. When considering an arbitrary interface inside the medium, with unit normal
n
i
, then
i
=
i
j
n
j
represents the traction, i.e., the force per unit area that one side of the surface
exerts on the other side. The antisymmetric tensor
ij
represents the spin density (or proper angular
moment density). The tensor m
ij
k
is antisymmetric in its two lower indices, and when considering
an arbitrary interface inside the medium, with unit normal n
i
, then
ij
= m
ij
k
n
k
represents the
surface density of torque that one side of the surface exerts on the other side. We call m
k
ij
the
moment stress or couple stress.
The term f
i
represents all the external densities of force acting on the medium. These, for instance,
can be electromagnetic forces (see section XXX). We are specially interested in the gravitational forces,
as they are going to be naturally predicted by the theory to be developed below. The Newtons force
density is
f
i
= g
i
, (7.4)
where the gravitational eld g relates to the mass density distribution by

i
g
i
= 4G (7.5)
and is curl-free, i.e., derives from a potential U , called the gravity potential:
g
i
=
i
U . (7.6)
In equation 7.5, G is the Newtonian constant of gravitation with the experimental value
G = 6.672 59 10
11
m
3
kg
1
sec
2
. (7.7)
The term
ij
represents all the external torque densities acting on the medium, for instance,
when an electromagnetic eld acts on a medium with magnetic properties.
To better understand equations 7.17.3, let us consider a volume inside the medium that follows
the matter on his movement and call it V (t) . The surface of the volume will be denoted S(t) . We
dene, for that volume, the following quantities:
Mass
M(t) =
_
V (t)
dV (x) (x, t) (7.8)
Linear momentum
P
i
(t) =
_
V (t)
dV (x) p
i
(x, t) (7.9)
7.2. MASS AND SPIN 133
Force
F
i
(t) =
_
V (t)
dV (x) f
i
(x, t) +
_
S(t)
dS(x)
i
(x, t) (7.10)
(as stated above,
i
=
i
j
n
j
are the surface tractions)
Angular momentum with respect to the origin of coordinates

ij
(t) =
_
V (t)
dV (x)
ij
(x, t) + (r
i
(x) p
j
(x, t) r
j
(x) p
i
(x, t)) (7.11)
( r(x) is the position vector: as we are only interested here in at (Euclidean) spaces [with
possibly non Cartesian coordinates] the introduction of this vector is easy, which is not the case
in curved spaces).
Torque with respect to the origin of coordinates

ij
(t) =
_
V (t)
dV (x)
ij
(x, t) + (r
i
(x) f
j
(x, t) r
j
(x) f
i
(x, t))
+
_
S(t)
dS(x)
ij
(x, t) + (r
i
(x)
j
(x, t) r
j
(x)
i
(x, t)) (7.12)
(as stated above,
ij
= m
ij
k
n
k
represents the surface density of torque))
It is easy to show (see appendix XXX) that the set of local conservation equations 7.17.3 are
equivalent to the set of (global) conservation equations
dM
dt
(t) = 0 (7.13)
dP
i
dt
(t) = F
i
(t) (7.14)
d
ij
dt
(t) =
ij
(t) . (7.15)
Comment: explain somewhere how important are these equations.
Comment: the term p
i
v
j
p
j
v
i
accounts for the eect discussed by Sedov (transfer from intrinsic
angular momentum into extrinsic angular momentum).
Comment: explain that equation 7.3, setting the conservation of angular momentum is not a
consequence of the other two. We have seen that we have intrinsic as well as extrinsic angular
momentum density (
ij
versus r
i
p
j
r
j
p
i
), and only the conservation of the extrinsic (orbital)
angular momentum density is a consequence of the conservation of the linear momentum.
Comment: explain that the purpose of this chapter is to demonstrate the equations 7.17.3 from
very general principles.
7.2 Mass and spin
In our general relativistic theory, matter possesses two fundamental properties: mass and spin, de-
scribed respectively by the 4-D tensors T

and M

, the last being antisymmetric in its two

lower indices:
M

= M

. (7.16)
134 CHAPTER 7. MATTER
Introducing right now a general system of space-time coordinates will complicate our goal (to
obtain the 3-D equations governing the dynamics of continuous media, valid for arbitrary reference
systems). For didactical purposes, it is better to consider rst a very special reference system, where
the space coordinate lines are attached to the medium, and move with it. In that reference system,
the medium is at rest.
Once the corresponding 3-D dynamical equations will be found and they already are going to
contain many terms , it is possible to change the reference system to a general one, to get the general
equations.
So, from now, and up to the section on relative velocities (section 8.2), we consider that we, local
observer, see a continuous medium around us that evolves, but has zero local velocity.
From the four-dimensional mass-tensor T

we can introduce the following three-dimensional elds:

_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
p
i
c/ T
i
j
_
, (7.17)
where is the mass density and T
i
j
the stress (what we call the stress here will be related with
the classical stress
j
i
through T
i
j
=
j
i
). Both and
j
i
have been mentioned in section 7.1.
We will see below that p
i
represents that part of linear momentum density that is not of convective
origin i.e., apart from v
i
. Also, q
i
is the mass ux that is not of convective origin like
the Poynting vector if the medium is electromagnetically charged . Finally, from the spin tensor
M

we can introduce the three-dimensional spin density

ij
and the moment stress M
ij
k
, (what
we call the moment stress here will be related with the classical moment stress m
k
ij
through
M
ij
k
= m
k
ij
). Both,
ij
and m
k
ij
have been already seen in section 7.1. We can also introduce
two new tensors, t
i
and

, to be analised below:
_
M
00
0
M
0j
0
M
i0
0
M
ij
0
_
=
_
0 t
j
c
2
t
i
c
2

ij
c/
_
(7.18)
_
M
00
k
M
0j
k
M
i0
k
M
ij
k
_
=
_
0
j
k
c

i
k
c M
ij
k
_
. (7.19)
Chapter 8
From geometry to physics
8.1 Media at rest
8.1.1 Fundamental constraints
Assume that, for all times, a space metric g
ij
and a space torsion S
ij
k
are given, satisfying
g
ij
= g
ji
, (8.1)
and
S
ij
k
= S
ji
k
. (8.2)
If the constraint

i
g
jk
= 0 (8.3)
is postulated, then the connection can be computed through

ijk
=
1
2
(
i
g
jk
+
j
g
ik

k
g
ij
) +
1
2
(S
ijk
+S
kij
+S
kji
) , (8.4)
and the curvature (Riemann) tensor
R
ijk

=
i

jk

ik

+
is

jk
s

js

ik
s
(8.5)
is antisymmetric in its two last indices:
R
ij
k
= R
ij
k
(8.6)
(the antisymmety in the rst two indices is obvious).
We are going to show that if an arbitrary scalar eld , an arbitrary vector eld I
i
, an arbitrary
antisymmetric tensor eld

J
ij
, and an arbitrary tensor eld K
i
j
are given, then the space-time
metric
_
g
00
g
0j
g
i0
g
ij
_
=
_

2
0
0 g
ij
_
, (8.7)
and the space-time torsion
_
S
00
0
S
0j
0
S
i0
0
S
ij
0
_
=
_
0 c
2
I
j
c
2
I
i
c

J
ij
/
_
(8.8)
_
S
00
k
S
0j
k
S
i0
k
S
ij
k
_
=
_
0 cK
j
k

cK
i
k
S
ij
k
_
(8.9)
are consistent, i.e., if from these, we compute the space-time connection through the usual expression

=
1
2
(

) +
1
2
(S

+S

+S

) (8.10)
135
136 CHAPTER 8. FROM GEOMETRY TO PHYSICS
and when dening covariant space-time derivatives through usual expressions, like

, (8.11)
we satisfy

= 0 . (8.12)
Then, the space-time Riemann tensor can be computed through the usual expression
R

, (8.13)
and we have the antisymmetry property
R

= R

. (8.14)
We have then a model of the Universe. Dierent choices for , g
ij
, I
i
,

J
ij
, K
i
j
, and
S
ij
k
, give dierent universe models. We will express the material properties of the universe (mass
density, spin density,. . . ) as a function of these elds, and will obtain the corresponding (3-D) evolution
equations.
The hat over

J
i
j
will later mean that this tensor is the antisymmetric part of a tensor J
i
j
to
be dened.
8.1.2 All denitions together
From now on, given an arbitrary tensor A
ij
, we will dene its symmetric part by

A
ij
= A
ij
+A
ji
, (8.15)
and its antisymmetric part by

A
ij
= A
ij
A
ji
. (8.16)
Then, we have
A
ij
=
1
2
(

A
ij
+

A
ij
) . (8.17)
Instead of considering partial derivatives with respect to the time coordinate, much simpler ex-
pressions will be obtained is we consider partial derivatives with respect to the proper time of the
observer (dened below). If dt represents a time interval, then, as shown in Section 6.6, the the
proper time of the observer is d = dt , where is the eld introduced in the space-time metric
(equation 8.7). Then, we write
_
dx
0
dx
i
_
=
_
c dt
dx
i
_
=
_
c d/
dx
i
_
, (8.18)
and use the notation

0


x
0


c

. (8.19)
The eld will not have any simple physical interpretation. Instead, the scalar eld U dened
(up to an additive constant) by
= exp(U U
0
) , (8.20)
i.e.,
U = U
0
+ Log , (8.21)
8.1. MEDIA AT REST 137
will be the gravitational potential, and will exactly correspond to the classical Newtonian potential
(times c
2
). In particular, its (negative) gradient
g
i
=
1
c
2

i
U (8.22)
will be the gravitational eld, and, as shown later, the product g
i
will give the gravitational force
density (times c
4
).
In the equations below, we will also use the time derivative of the potential:
=
1
c
4
U

. (8.23)
Dening
G
ij
=
1
2
_
1
c
2
g
ij

+

J
ij
+

K
ij
_
, (8.24)
gives

G
ij
=
1
c
2
g
ij

, (8.25)
and

G
ij
=

J
ij
+

K
ij
. (8.26)
Dening
J
ij
= G
ij
K
ij
, (8.27)
i.e.,
J
ij
=
1
2
_
1
c
2
g
ij

+

J
ij


K
ij
_
, (8.28)
gives

J
ij
=

G
ij


K
ij
, (8.29)
and the antisymmetric part of J
ij
corresponds, as it should, to the antisymmetric tensor

J
ij
introduced in equation 8.8.
The following denition is going to be useful for expressing the space-time connection:
h
i
= g
i
I
i
, (8.30)
while the following ones are going to be useful to express the curvature:

i
j
=
1
c
2
J
i
j

+
i
h
j
c
2
g
i
h
j
+J
i
k
G
k
j
, (8.31)

ij
k
=
i
J
j
k

j
J
i
k
+S
ij
l
J
l
k
, (8.32)

ij
k
=
1
c
2

ij
k

i
G
j
k
+c
2
(g
i
G
j
k
+J
i
k
h
j
J
ij
h
k
) , (8.33)
and
r
ij
k
= R
ij
k
+c
2
(J
i
l
J
j
k
J
j
l
J
i
k
) , (8.34)
where R
ij
k
has been dened in equation 8.5.
It is not obvious, but it is possible to show that

i
jk
=
i
kj
. (8.35)
138 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.1.3 Space-time connection and curvature
When computing the space-time connection using equation 8.10, we obtain
_

00
0

0j
0

i0
0

ij
0
_
=
_
c
3
c
2
h
j
c
2
g
i
cJ
ij
/
_
(8.36)
_

00
k

0j
k

i0
k

ij
k
_
=
_
c
2
h
k

2
cG
j
k

cJ
i
k

ij
k
_
, (8.37)
where
ij
k
was dened in equation 8.4.
When computing the components of the Riemann tensor using equation 8.13, we obtain
_
R
00
00
R
0j
00
R
i0
00
R
ij
00
_
=
_
0 0
0 0
_
, (8.38)
_
R
00
0
R
0j
0
R
i0
0
R
ij
0
_
=
_
0 c
2

j
l
c
2

i
l
c
ij
l
/
_
, (8.39)
_
R
00
k0
R
0j
k0
R
i0
k0
R
ij
k0
_
=
_
0 c
2

j
k
c
2

i
k
c
ij
k
/
_
, (8.40)
_
R
00
k
R
0j
k
R
i0
k
R
ij
k
_
=
_
0 c
j
k

c
i
k
r
ij
k
_
. (8.41)
The Ricci tensor R

= R

is given by
_
R
0
0
R
0
j
R
i
0
R
i
j
_
=
_
c
2

k
k
c
k
jk

c
ki
k
/ c
2

i
j
+r
ki
jk
_
, (8.42)
and the scalar curvature R = R

by
R = 2c
2

i
i
+r
ji
ij
. (8.43)
The contracted torsion
S

= S

(8.44)
is given by
_
S
0
S
i
_
=
_
cK
s
s

c
2
I
i
+S
si
s
_
. (8.45)
8.1.4 Recall: Bianchi identities
We have seen in the previous chapter how to dene, for an arbitrary dierentiable manifold (i.e.,
for any geometrical space), the curvature and the torsion. In general relativity, as we consider the
geometry of the space-time, we also have to introduce its curvature and torsion. Let us denote them
by R

and S

respectively.
The contracted curvature is dened by
R

= R

, (8.46)
and is called the Ricci tensor. The scalar
R = g

= R

, (8.47)
8.1. MEDIA AT REST 139
is called the scalar curvature. The torsion tensor is antisymmetric,
S

= S

, (8.48)
the contracted torsion is dened by
S

= S

, (8.49)
and does not receive any particular name.
We have seen in the previous chapter that curvature and torsion satisfy, by their very denition,
the contracted Bianchi identities

_
R

1
2

R
_
= S

_
1
2
R

_
, (8.50)
and

_
S

+S

_
=
_
R

_
+S

. (8.51)
8.1.5 4-D eld equations (Einstein-Cartan Equations)
The form of the contracted Bianchi identities 8.508.51 suggests to introduce the Einstein tensor
E

= R

1
2

R (8.52)
and the Cartan tensor
C

= S

+S

. (8.53)
Then, the contracted Bianchi identities can be written

= S

_
1
2
R

_
, (8.54)
and

=
_
R

_
+S

. (8.55)
Einstein demonstrated that a consistent gravitation theory can be obtained by identifying the
tensor E

with the mass (i.e., stress-energy) tensor:

E

= T

, (8.56)
where the proportionality constant is
=
8G
c
4
. (8.57)
As when Einstein developed his gravitation theory, the spin was not yet discovered, the theory did
not account for spin. The Einstein equation 8.56 couples mass with space-time curvature. Today we
understand that spin couples with space-time torsion:
C

= M

. (8.58)
We will later see (Section 8.3) that in order to get, in the classical limit, the right equations of
conservation of angular momentum, the proportionality constant in the Cartan equation 8.58 must be
the same than in the Einstein equation 8.56, i.e., the dened by equation 8.57. So, even if this
Einstein-Cartan theory has two fundamental equations, there is only one fundamental constant.
140 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.1.6 Expressing the matter properties
Recall that from the space-time metric g

we have introduced the space metric g

ij
and the scalar
eld , while from the space-time torsion S

we have introduced the space torsion S

ij
k
and
the 3-D tensor elds I
i
,

J
ij
, and K
i
j
.
From those, we can express all the components of the space-time curvature R

and of the
space-time Einstein tensor E

, and all the components of the space-time torsion S

and of the
space-time Cartan tensor C

. We have then expressions for the left-hand side of the Einstein-

Cartan equations 8.56 and 8.58.
The Einstein equation
T

=
1

_
R

1
2

R
_
, (8.59)
allows to compute the material properties , p
i
, q
i
, and T
i
j
(respectively mass density, linear
momentum density, mass ux, and stress). This gives
_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
cp
i
/ T
i
j
_
, (8.60)
where
=
1
2c
2
r
k
k
, (8.61)
p
i
=
1

ki
k
, (8.62)
q
j
= +
1

k
jk
, (8.63)
and
T
i
j
=
1

_
c
2
(
i
j

i
j

k
k
) + (r
ki
jk

1
2

i
j
r
k
k
)
_
. (8.64)
The Cartan equation
M

=
1

_
S

+S

_
, (8.65)
allows to compute the material properties
ij
, M
ij
k
, t
i
, and
i
j
(respectively spin density,
moment stress, unbalance, unbalance ux). This gives
_
M
00
0
M
0j
0
M
i0
0
M
ij
0
_
=
_
0 c
2
t
j
c
2
t
i
c
ij
/
_
, (8.66)
_
M
00
k
M
0j
k
M
i0
k
M
ij
k
_
=
_
0 c
j
k

c
i
k
M
ij
k
_
, (8.67)
where
t
i
=
1
c
2
S
ji
j
, (8.68)

ij
=
1

J
ij
, (8.69)

i
j
=
1

_
K
i
j
K
k
k

i
j
_
, (8.70)
and
M
ij
k
=
1

_
S
ij
k
+S
i
l

j
k
S
j
l

i
k
+c
2
(I
i

j
k
I
j

i
k
)
_
. (8.71)
8.1. MEDIA AT REST 141
8.1.7 The conservation equations
From the previous expressions for the material properties of the continuous medium, , p
i
, t
i
, and

ij
, we can try to derive the corresponding conservation equations. It is much simpler to directly
use the contracted Bianchi identities

_
R

1
2

R
_
= S

_
1
2
R

_
, (8.72)
and

_
S

+S

_
=
_
R

_
+S

, (8.73)
which directly give

+
i
q
i
= , (8.74)
p
i

+
j
T
i
j
= f
i
, (8.75)
t
i

+
j

i
j
= p
i
q
i
+
i
, (8.76)

ij

+
k
M
ij
k
= T
ij
T
ji
+
ij
, (8.77)
where
= c
2
_
J
i
i
h
i
(p
i
+q
i
)
_
T
i
j
J
j
i
+
1

_
K
i
j
r
kj
ik

1
2
S
ij
k

k
ij
+c
2
I
i

j
ij
_
, (8.78)
f
i
= +c
4
h
i
c
2
_
h
j
T
i
j
+J
ji
q
j
G
i
j
p
j
+J
j
j
p
i
_
+
1

_
c
4
_

i
j
I
j

j
j
I
i
_
+c
2
_

1
2

i
jk

J
jk


J
ij

k
jk

ij
k
K
k
j
+K
i
j

kj
k
_
+
1
2
r
ij
k
S
k
j
+S
ij
k
(c
2

k
j
+r
k
j
)
_
, (8.79)

i
= c
2
_
J
j
j
t
i
G
i
j
t
j

ij
h
j

i
j
h
j
_
M
ij
k
J
k
j
+
1

_
c
2
K
j
j
I
i
+K
i
j
(c
2
I
j
+S
kj
k
)
_
, (8.80)
and

ij
= +c
4
(t
i
h
j
t
j
h
i
) c
2
_
(J
ki

j
k
J
kj

i
k
) (
ik
G
j
k

jk
G
i
k
)
_
c
2
_
J
k
k

ij
M
ij
k
h
k
_
+
1

_
c
2
K
k
k

J
ij
+S
ij
k
(c
2
I
k
+S
k
l
)
_
. (8.81)
The term corresponds to the rate of creation of mass density. The terms f
i
and
ij
are
respectively the force density and the torque density acting on the medium (they can also be named
the rate of creation of linear momentum and of angular momentum respectively). See below for an
interpretation of
i
(and of t
i
).
142 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.1.8 The simplest model with spin
We have seen that, to dene a model of the universe, we can freely choose the elds , g
ij
, I
i
,

J
ij
,
K
i
j
, and S
ij
k
. In standard gravitation theory (i.e., in the theory without spin), the fundamental
elds are and g
ij
, while the elds I
i
,

J
ij
, K
i
j
, and S
ij
k
identically vanish. This corresponds
to a universe with mass and without spin. Let us be interested in the opposite case, where is
constant (so the gravitational potential U is constant), and the space metric g
ij
is that of an
euclidean space, independent on time. The simplest model is then obtained when only the eld

J
ij
is non vanishing, while the three elds I
i
, K
i
j
, and S
ij
k
vanish.
Formula 8.69 shows that the eld

J
ij
is then just proportional to the spin density of the universe:

J
ij
=
ij
. (8.82)
From now on, let us use, instead of , its expression as a function of the Newtonian Gravitational
constant G (as given by equation 8.57). Then, the previous equation writes

J
ij
=
8G
c
4

ij
. (8.83)
An easy use of the formulas of section 8.1.6 allows then to express all the material properties of
the universe as a function of the spin density:
=
G
c
4

ij

ij
, (8.84)
p
i
= q
i
=
1
2

i
j
, (8.85)
and
T
i
j
=
1
2

i
j
+
G
c
2

i
j

k
, (8.86)
while the material elds t
i
,
i
j
, and M
ij
k
, all vanish.
Of particular interest are the right-hand side of the evolution equations 8.74 8.77:
=
G
c
4

(
ij

ij
) , (8.87)
f
i
=
G
c
2

i
(
jk

jk
) , (8.88)

i
= 0 , (8.89)
and

ij
=

ij
. (8.90)
In particular, we see that within a uid with spin density
ij
there is a force density f
i
proportional to the squared spin density (equation 8.88). Like the Newtonian gravitational force, the
gravitational force due to the spin density tends to concentrate matter, as is directed towards regions
of high spin density.
If, like the mass-mass force, the spin-spin gravitational force is an attractive force (it tends to
concentrate spin), unlike the mass-mass force it has vanishing range: it only depends on the local
value of the spin density.
At the molecular (or larger) scale, where matter is concentrated in atoms surrounded by vacuo,
the spin-spin gravitational interaction should not play any signicant role.
On the contrary, such a force could play a major role in accounting for the stability of elementary
particles. For instance, it is well known that simple models of an electron, can not explain its stability,
8.1. MEDIA AT REST 143
as the electrostatic repulsive forces dominate, by orders of magnitude, the attractive gravitational
forces. This is only true if we consider only the mass-mass gravitational forces. It is possible to
imagine matter distributions where the spin gravitational force equilibrates other repulsive forces, like
Coulombs electrostatic repulsion in elementary models of an electron as a continuous medium. For
instance, appendix 8.5.1 shows a simple model of an electron where the electrostatic repulsive forces
are exactly compensated by the attractive spin-spin gravitational force just described.
Let us mention that Kerlick [1975], and OConnell [1977] give a discussion on the spin-spin gravi-
tational interaction discussed here, but based on Lagrangian arguments.
8.1.9 The gravitational forces
The gravitational force density acting in a uid has been given by equation 8.79, with reference to
equation 8.75.
We see that 8.79 has many terms: there are many diferent gravitational forces acting on a contin-
uous medium.
Let us focus in the rst term,
f
i
= c
4
h
i
. (8.91)
Equation 8.30 gives the denition of h
i
:
h
i
= g
i
I
i
, (8.92)
equation 8.31 gives the denition of
i
j
:

i
j
=
1
c
2
J
i
j

+
i
h
j
c
2
g
i
h
j
+J
i
k
G
k
j
, (8.93)
and one of the reversed form of the Einstein equations (equation 8.248) gives

s
s
=

2
_
+
1
c
2
T
s
s
_
, (8.94)
i.e.,

s
s
=
4G
c
4
_
+
1
c
2
T
s
s
_
. (8.95)
For ordinary matter, the term T
s
s
/c
2
can be neglected in front of , the eld I
i
can be neglected
in front of g
i
, and the time derivative of J
i
j
and the quadratic terms can be neglected. This gives
for the gravitational force density
f
i
= c
4
g
i
, (8.96)
where

i
g
i
=
4G
c
4
, (8.97)
and we should remember that g
i
was dened by (equation 6.14)
g
i
=
1
c
2

i
U . (8.98)
The force density 8.96 is the standard Newtonian gravitational force density. The eld g
i
is
the Newtonian gravitational eld (times c
4
), satisfying the Poissons equation 8.97, and U is the
classical Newtonian potential (times c
2
).
So far, for the gravitational force density in ordinary matter, where only the massive properties
of the medium are taken into account.
144 CHAPTER 8. FROM GEOMETRY TO PHYSICS
The previous section has shown that when taking into account only the spin properties of the
medium, we have the gravitational force density
f
i
=
G
c
2

i
(
jk

jk
) . (8.99)
This means that, among the many forces acting in a uid in gravitational interaction, two are
particularly simple, depending respectively on the mass and spin only:
f
i
= c
4
g
i
+
G
c
2

i
(
jk

jk
) . (8.100)
8.1.10 3-D evolution equations for a medium locally at rest
What we would have liked to solve is the following problem of initial conditions: given , p
i
, t
i
,

ij
, g
ij
(and some other elds) at t = t
0
, and assuming given some constitutive equations that
allow us to compute q
i
, T
i
j
,
i
j
, and M
ij
k
at any time as a function of the history of the
medium, then, it is possible to extrapolate the values of , p
i
, t
i
,
ij
, g
ij
(and some other
elds) at all times t > t
0
.
Instead, what we have solved is: given g
ij
,

G
ij
,
ij
, and t
i
, at t = t
0
, and assuming given
some constitutive equations that allow us to compute p
i
p
i
q
i
, T
i
j
,
i
j
, and M
ij
k
at any
time as a function of the history of the medium, as well as assuming given U at all times, then, it
is possible to extrapolate the values of g
ij
,

G
ij
,
ij
, and t
i
, at all times t > t
0
.
This is shown in Appendix 8.5.3.
8.2 Moving media
The equations seen so far are valid in a coordinate system attached to the continuous medium, with
respect to which the medium is at rest. Usually we want to measure forces in the laboratory
reference system, with respect to which the medium is moving. This introduces many more terms in
the equations (convective terms) and many more forces similar to the magnetic force in a moving
electric medium.
Let us start by introducing the velocity.
8.2.1 Measuring (improper) velocities
Assume we have, in the laboratory, a rod of length and a clock, both at rest.
If a particle passes by, and it takes a time (in the laboratory clock) to the particle to travel
between the two extremities of the rod, then, by denition, the velocity of the particle is
v = lim
0

. (8.101)
How small has to be in practice for an accurate enough estimation of the velocity will depend of
the regularity of the displacement of the particle.
Should the particle travel at the speed of light, then, by the operational denition of distance,
= c , and v = c . For any other particle, v < c .
8.2.2 Dening the 4-velocity
Consider, again, an arbitrary particle. In section 6.5, we have parameterized the space-time trajectory
of a particle by an arbitrary parameter . Instead, we can now choose the particles proper time:
8.2. MOVING MEDIA 145
x

() . The tangent to the trajectory is called the 4-velocity:

U

=
dx

d
. (8.102)
It satises
g

= g

dx

d
dx

d
=
g

dx

dx

d
2
=
c
2
d
2
d
2
, (8.103)
i.e.,
g

= c
2
. (8.104)
It is important to realize that the denition of four velocity uses the proper time of the particle,
while the 3-D denition of (improper) velocity in section 8.2.1 used the proper time of the observer.
Let us consider a particle that in the time interval dt has made the space displacement dx
i
, and
let us denote by d
obs
and d
part
the proper time elapsed as measured by the particle itself or by
the observer respectively.
The proper time for the observer is, from equation 6.22,
d
obs
= dt , (8.105)
while the proper time for the particle is, from equation 6.19,
d
part
=
_

2
dt
2

1
c
2
dx
i
dx
i
= dt

1
1
c
2

2
dx
i
dt
dx
i
dt
= dt

1
1
c
2
dx
i
d
obs
dx
i
d
obs
.
The 3-velocity was dened in section 8.2.1 as
v
i
=
dx
i
d
obs
. (8.106)
Then,
d
part
= dt

1
v
i
v
i
c
2
, (8.107)
or, if we introduce the standard notations

i
=
v
i
c

2
= v
i
v
i
, (8.108)
we obtain the equation giving the relation between the proper time of a particle, and the coordinate
and proper times of the observer:
d
part
= dt
_
1
2
= d
obs
_
1
2
, (8.109)
where we have used 6.22.
146 CHAPTER 8. FROM GEOMETRY TO PHYSICS
If, as it is customary, we introduce the gamma factor
=
1
_
1
2
, (8.110)
then the previous equation gives
d
part
= dt = d
obs
. (8.111)
As, obviously, > 1 , equations 8.109 and 8.111 show that the proper time for the particle is
always smaller that the proper time of the observer: this is the relativistic contraction of time, seen
here in the context of general relativity.
Using the previous equations we easily obtain the components of the particles 4-velocity in terms
of the 3-velocity measured by the observer:
_
U
0
U
i
_
=
_
c/
v
i
_
. (8.112)
8.2.3 Lorentz transformation
When two observers cross each other, they may want to compare the components of the tensors in their
respective reference systems. In particular, the material properties (mass,. . . )are usually measured
in a reference system where the matter is at rest, and from those measured values we may need to
predict the values that should be measured in reference systems where the matter is in movement.
It is the Lorentz transformation that performs such a task. If the components of a space-time
vector are V

in some reference system, they will be

V

, (8.113)
in another reference system whose relative velocity with respect to the rst reference system is v
i
=
c
i
. Here, the

are the components of the Lorentz transformation. It is given by

_

0
0

0
j

i
0

i
j
_
=
_

j
/

i

i
j
+
(1)

2

i

j
_
. (8.114)
It can be introduced in many ways. For instance, it may be said that it is a space-time rotation:
the length of any vector has to remain unchanged. Equivalently me may say that it has to leave the
components of the metric tensor invariant.
Notice that, although the Lorentz transformation is used in the theory of special relativity, the
expression given above is fully valid within the framework of the general relativistic theory developed
here.
It is easily seen that the inverse transformation is obtained by simply reversing the velocity (
i

i
).
Note: Check if I say somewhere that the product of the Lorentz transformation timest its transpose
gives the Kronecker tensor.
Note: Check if I say somewhere that the determinant of the Lorentz transformation equals -1.
For the sake of completeness, let us mention explicitly that the components of second and third
order tensors change respectively according to
T

(8.115)
and
M

. (8.116)
8.2. MOVING MEDIA 147
For instance, if in the proper reference system, the mass tensor (stress-energy tensor) is given by
_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
cp
i
/ T
i
j
_
, (8.117)
then, in a system moving with velocity v
i
with respect to it, it is given by
_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
c p
i
/

T
i
j
_
, (8.118)
where
=
2
_
+
1
c
2
(v
k
p
k
+v
l
q
l
) +
1
c
4
v
k
v
l
T
k
_
,
p
i
=
_
p
i
+
1
c
2
v
l
T
i
+
_
+

c
2
v
l
q
l
+
1
v
2
v
k
p
k
+
1
c
2
1
v
2
v
k
v
l
T
k
_
v
i
_
,
q
j
=
_
q
j
+
1
c
2
v
k
T
kj
+
_
+

c
2
v
k
p
k
+
1
v
2
v
l
q
l
+
1
c
2
1
v
2
v
k
v
l
T
k
_
v
j
_
,

T
ij
= T
ij
+
_
1
v
2
v
l
T
i
+p
i
_
v
j
+
_
1
v
2
v
k
T
kj
+q
j
_
v
i
+
_
+
1
v
2
(v
k
p
k
+v
l
q
l
) +
( 1)
2
v
4
v
k
v
l
T
k
_
v
i
v
j
.
Analogous (but more lengthy) expressions are found for the spintensor.
Later on, we are going to be interested in the limit v
i
c . This obviously gives
= ,
p
i
= p
i
+v
i
,
q
j
= q
j
+v
j
,

T
ij
= T
ij
+p
i
v
j
+q
j
v
i
+v
i
v
j
.
Similarly, in the limit v
i
c we obtain for the spin tensor

t
i
= t
i
,

ij
=
ij
+t
i
v
j
,

ij
=
ij
+t
i
v
j
t
j
v
i
,

M
ijk
= M
ijk
+ (
ik
v
j

jk
v
i
) + (
ij
+t
i
v
j
t
j
v
i
)v
k
,
and similar expressions for all other tensors involved.
8.2.4 3-D evolution equations for moving media (low velocities)
After a Lorentz transformation of all the elds appearing in the evolution equations of section 8.1.10,
we obtain exact 3-D evolution equations, valid for any value of the velocity of the medium.
The equations are easy to obtain, but so cumbersome, that we prefer not to explicitly list them
here. Thats where the advantage of 4-D notations becomes obvious. But the advantage of the the 3-D
notations is that while the 3-D space has curvature and torsion, the geometry of the 4-D space-time
is no more into consideration: space-time curvature and torsion have been replaced by ordinary 3-D
forces. We emphasize that we are talking about exact expressions.
148 CHAPTER 8. FROM GEOMETRY TO PHYSICS
There is a second reason why we have chosen not to explicitly list the exact 3-D equations here.
At the right-hand side of the conservation equations 8.748.77, there are the source terms ,
f
i
,
ij
, and
i
, given in equations 8.788.81. It is very interesting to look at the expressions
of these terms after a Loretz transformation has added all the terms depending on the velocity of
the medium. In particular, the force density term f
i
shows all the gravitational forces acting in a
continuous medium under gravitational interaction and there are many . For instance, besides the
classical Newtonian force density, there is the spin-spin gravitational force density already discussed
in section 8.1.9. Adding the terms depending on the velocity of the medium will add a lot of terms,
like in electromagnetism, where if the medium is at rest, there is only the electric force, while for a
moving medium the magnetic force appears.
The second reason why we have chosen not to explicitly list the exact 3-D equations is that we would
have liked to give the expression of these forces (in particular) as a function of some fundamental
quantities, instead as giving them as a function of the quatities appearing in equations 8.788.81. Our
hope was to discover what those fundamental quantities could be when solving the problem of initial
conditions, where, in addition to the material quantities , p
i
, t
i
, and
ij
, some other quatities
should have been given at t = t
0
to have uniqueness of the evolution of the system.
The evolution theorem in Appendix 8.5.3 is not as fundamental as we have been looking for. We
have the inpression of having missed to recognize those fundamental quantities, and, therefore, we have
not made the eort of writing down explicitly all the source terms in the 3-D conservation equations
and interpreting them physically.
Let us now turn to the simplest problem where we are only interested in the limit v
i
c . The
conservation equations 8.788.81 become (replacing the proper time by the Newtonian time t ),

t
+
i
(q
i
+v
i
) = , (8.119)

t
(p
i
+v
i
) +
j
_
(v
i
+p
i
) v
j
+v
i
q
j
+T
i
j
_
= f
i
, (8.120)
t
i
t
+
j
(t
i
v
j
+
i
j
) = (p
i
q
i
) +
i
, (8.121)

t
(
ij
+t
i
v
j
t
j
v
i
) +
k
_
(
ij
+t
i
v
j
t
j
v
i
) v
k
+
i
k
v
j

j
k
v
i
+M
ij
k
_
= (T
ij
T
ji
) +
ij
, (8.122)
where the source terms are, again, long to express, but easy to obtain (Comment: say here that the
source terms are, in classical physics given in an ad-hoc way).
8.2.5 Global conservation equations
Comment: say here that the local conservation equations 8.122 can, equivalently, be written as global
conservation equations. Considering a volume 1(t) attached to the medium (comment: explain),
with surface o(t) these equations can be written
M
t
+Q(t) = K(t)
P
t
+S(t) = F(t)
T
t
+A(t) = Z(t)
T
t
+((t) = ((t) , (8.123)
8.2. MOVING MEDIA 149
where I have dened the four fundamental conserved quantities, the mass
M(t) =
_
V(t)
dV (x) (x, t) , (8.124)
the linear momentum
P(t) =
_
V(t)
dV (x)
_
p
i
(x, t) +(x, t) v
i
(x, t)
_
e
i
(x) , (8.125)
the unbalance
T(t) =
_
V(t)
dV (x) t
i
(x, t) e
i
(x) , (8.126)
and the angular momentum
T(t) =
_
V(t)
dV (x)
_

ij
(x, t) +
_
t
i
(x, t)v
j
(x, t) t
j
(x, t)v
i
(x, t)
__
e
i
(x) e
j
(x) , (8.127)
where I have also dened four external sources, the rate of creation of mass
K(t) =
_
V(t)
dV (x) (x, t) , (8.128)
the force (or rate of creation of linear momentum)
F(t) =
_
V(t)
dV (x) f
i
(x, t) e
i
(x) , (8.129)
the rate of creation of unbalance
Z(t) =
_
V(t)
dV (x)
__
p
i
(x, t) q
i
(x, t)
_
+
i
(x, t)
_
e
i
(x) , (8.130)
and the torque (or rate of creation of angular momentum)
((t) =
_
V(t)
dV (x)
__
T
ij
(x, t) T
ji
(x, t)
_
+
ij
(x, t)
_
e
i
(x) e
j
(x) , (8.131)
and where, nally, I have dened four internal sources (note that these are surface integrals), the
energy(?) heat(?) mass(?) ux
Q(t) =
_
S(t)
dS(x) (x, t) , (8.132)
the traction (stress ux(?))
S(t) =
_
S(t)
dS(x)
_
(x, t) v
i
(x, t) +s
i
(x, t)
_
e
i
(x) , (8.133)
the unbalance ux (?)
A(t) =
_
V(t)
dV (x) a
i
(x, t) e
i
(x) , (8.134)
and the angular momentum ux (?)
((t) =
_
V(t)
dV (x)
__
a
i
(x, t)v
j
(x, t) a
j
(x, t)v
i
(x, t)
_
+d
ij
(x, t)
_
e
i
(x) e
j
(x) , (8.135)
150 CHAPTER 8. FROM GEOMETRY TO PHYSICS
where I have introduced four uxes (?), the xxx
(x, t) = q
i
(x, t) n
i
(x, t) , (8.136)
the traction
s
i
(x, t) = T
i
j
(x, t) n
j
(x, t) , (8.137)
the xxx
a
i
(x, t) =
i
j
(x, t) n
j
(x, t) , (8.138)
and the xxx
d
ij
(x, t) = M
ij
k
(x, t) n
k
(x, t) . (8.139)
8.2.6 Global conservation equations in Euclideanspaces with Cartesian coordi-
nates
Comment: This is certainly not the good place for this section, but I have to give somewhere the
equivalent of the equations of the previous section for Euclidean spaces with Cartesian coordinates:
M
t
+Q(t) = K(t)
P
i
t
+S
i
(t) = F
i
(t)
T
i
t
+A
i
(t) = Z
i
(t)
D
ij
t
+G
ij
(t) = C
ij
(t) , (8.140)
M(t) =
_
V(t)
dV (x) (x, t) , (8.141)
P
i
(t) =
_
V(t)
dV (x)
_
p
i
(x, t) +(x, t) v
i
(x, t)
_
, (8.142)
T
i
(t) =
_
V(t)
dV (x) t
i
(x, t) , (8.143)
D
ij
(t) =
_
V(t)
dV (x)
_

ij
(x, t) +
_
t
i
(x, t)v
j
(x, t) t
j
(x, t)v
i
(x, t)
__
, (8.144)
K(t) =
_
V(t)
dV (x) (x, t) , (8.145)
F
i
(t) =
_
V(t)
dV (x) f
i
(x, t) , (8.146)
Z
i
(t) =
_
V(t)
dV (x)
__
p
i
(x, t) q
i
(x, t)
_
+
i
(x, t)
_
, (8.147)
C
ij
(t) =
_
V(t)
dV (x)
__
T
ij
(x, t) T
ji
(x, t)
_
+
ij
(x, t)
_
, (8.148)
Q(t) =
_
S(t)
dS(x) (x, t) , (8.149)
S
i
(t) =
_
S(t)
dS(x)
_
(x, t) v
i
(x, t) +s
i
(x, t)
_
, (8.150)
8.2. MOVING MEDIA 151
A
i
(t) =
_
V(t)
dV (x) a
i
(x, t) , (8.151)
G
ij
(t) =
_
V(t)
dV (x)
__
a
i
(x, t)v
j
(x, t) a
j
(x, t)v
i
(x, t)
_
+d
ij
(x, t)
_
, (8.152)
(x, t) = q
i
(x, t) n
i
(x, t) , (8.153)
s
i
(x, t) = T
i
j
(x, t) n
j
(x, t) , (8.154)
a
i
(x, t) =
i
j
(x, t) n
j
(x, t) , (8.155)
d
ij
(x, t) = M
ij
k
(x, t) n
k
(x, t) . (8.156)
8.2.7 Ordinary matter
A simple, but quite general, model of matter is obtained when p
i
= q
i
= 0 , t
i
= 0 , and
i
j
= 0 .
We talk about ordinary matter. For instance, p
i
= 0 means that the only linear momentum
density in the medium is that of convective origin, v
i
, and that there are no currents (like in an
electromagnetic medium).
Then, the equations above simplify to

t
+
i
(v
i
) = , (8.157)

t
(v
i
) +
j
(v
i
v
j
+T
i
j
) = f
i
, (8.158)

ij
t
+
k
(
ij
v
k
+M
ij
k
) = (T
ij
T
ji
) +
ij
. (8.159)
Here, to match classical notations, we have to dene

j
i
= T
i
j
(8.160)
and
m
k
ij
= M
ij
k
. (8.161)
The equations above then give

t
+
i
(v
i
) = , (8.162)

t
(v
i
) +
j
(v
i
v
j

j
i
) = f
i
, (8.163)

ij
t
+
k
(
ij
v
k
m
k
ij
) = (
ji

ij
) +
ij
. (8.164)
8.2.8 Interpretation of the conserved quantities
The rst conservation equation (8.119) corresponds to the conservation of mass (or, equivalently, of
energy). being the mass density, the vector v
i
gives, when integrated on a surface, the mass
crossing the surface per unit of time due to the convective motion of the medium. Traditionally, this
vector v
i
is called the (convective) mass (or energy) ux. The vector q
i
is then the intrinsic mass
ux, i.e, the mass ux not due to convection. For instance, if there is an electromagnetic eld, then,
independently of the motion of the medium, there will be an electromagnetic mass (or energy) ux
(the Poynting vector), and the total mass ux will be the sum v
i
+q
i
.
The second conservation equation (8.120) corresponds to the conservation of linear momentum.
While v
i
is the linear momentum density due to convection (i.e, due to the motion of the medium),
152 CHAPTER 8. FROM GEOMETRY TO PHYSICS
p
i
is the intrinsic linear momentum density, an example being obtained when considering a medium
with electric currents. The term T
i
j
is the stress tensor, and is sometimes called the (intrinsic) linear
momentum ux.
Lets skip for a moment equation 8.121 and give the interpretation of equation 8.122. It obviously
corresponds to the conservation of angular momentum. While the antisymmetric tensor
ij
represents
the spin density, the tensor M
ij
k
is the moment stress (comment: explain). The source of spin is
(among others) the antisymmetric part of the stress. Comment: say that this is classical for micropolar
media. We see that the total spin density at a point of a continuous medium equals the intrinsic spin
density,
ij
, and a convective part, t
i
v
j
t
j
v
i
. As this is non classical, let us examine it with some
care.
If instead of using the antisymmetric tensor
ij
to represent the spin density, we use its dual
vector
s
i
=
1
2

ijk

jk
, (8.165)
the total spin density is given by the vector
S = s +t v , (8.166)
i.e, the total spin vector equals the intrinsic spin density plus a convective part. This convective part
equals the vector product of the material property described by the vector t by the velocity v .
A very imperfect analogy of this material property corresponds to a wheel attached to an axle but
able to run and glide on a plane. If we associate to the wheel a vector perpendicular to the plane (and
to the axle), and with magnitude equal to the ratio (moment of inertia/radius of the wheel), then,
the kinetic moment of the wheel, when running at velocity v , equals the vector product of the two
vectors.
Comment: make a drawing here.
Equation 8.121 says that this property, described by the vector t , is conserved. Thus, we have four
conserved quantities: the mass, the linear and angular momentum, and the vector t . Although it is
not common that a theory predicts the conservation of a quantity not yet interpreted, this conservation
equation has not received much attention so far. Comment: mention here Halbwachhhhsss (balourd,
unbalance) and the personnal communication with Vigier.
Equation 8.121 also says that the source of t is the dierence between the linear momentum
density p anf the energy ux q . For instance, as shown in appendix xxx, an electromagnetic eld
has a linear momentum density
p = DB (8.167)
and an energy ux
q =
1
c
2
EH. (8.168)
In vacuo,
=
0
(8.169)
and
=
0
, (8.170)
and, as

0
=
1
c
2
, (8.171)
the linear momentum density equals the energy ux:
p = q. (8.172)
In a general medium, the linear momentum density is not equal to the energy ux, and there is a
source for the quantity t .
8.3. CLASSICAL LIMIT (LOW VELOCITIES AND WEAK FIELDS) 153
8.3 Classical limit (low velocities and weak elds)
In the equations of the previous section, the elds were not assumed to be weak. If they are, then we
obtain the classical limit, where the physical 3-D space is Euclidean, the time is Newtonian, and from
the many gravitational forces acting in a continuous medium, only a few are assumed to be signicant.
We say that the gravitational elds are weak when the equations 8.1628.164 of previous section
can be approximated by

t
+
j
(v
j
) = 0 , (8.173)
(v
i
)
t
+
j
(v
i
v
j

j
i
) = c
4
g
i
+
G
c
2

i
(
jk

jk
) , (8.174)
and

ij
t
+
k
(
ij
v
k
m
k
ij
) =
ij

ji
, (8.175)
where we have kept in the right-hand terms only the classical Newtonian gravitational force density,
end the spin-spin force density analysed above.
Note that, should we have taken, instead of equations 8.56 and 8.58:
E

= T

, (8.56 again)
and
C

= M

, (8.58 again)
a dierent coupling constant:
E

= T

, (8.176)
and
C

, (8.177)
then, equation 8.175 would have been, instead,

ij
t
+
k
(
ij
v
k
m
k
ij
) =

(
ij

ji
) . (8.178)
If we want to obtain, in this classical limit, the classical equation 7.3, we are forced to take
=

, (8.179)
i.e., to take the same coupling constant between space-time curvature and mass (Einstein equation)
that between space-time torsion and spin (Cartan equation).
8.4 Classical elasticity
All classical (i.e., non relativistic) media are assumed to satisfy the equations of the dynamics of
continous media 8.1738.175.
The stress
j
i
and the moment stress m
k
ij
are dened ad-hoc, as functions of all the past
history of the medium. These are the constitutive equations that dene the rheology of the
medium. Dierent choices of rheology allow the equations describing the dynamics of continous media
to describe, for instance, uids, plastics, or solids.
The elastic approximation consists in assuming that the stress T
i
j
and the moment stress M
ij
k
depend, at any time t , only on the state of deformation of the solid at time t .
154 CHAPTER 8. FROM GEOMETRY TO PHYSICS
To dene deformation, we have to refer to an undeformed state of the solid, where, by denition,
stress and moment stress vanish. Usually, the undeformed state is assumed to be the state of the
medium at some initial time t
0
.
So far, we have used, as fundamental variable for describing the movement of the particles of the
continuous medium, the velocity v
i
(x, t) . Now, instead, we consider the displacement eld u
i
(x, t)
as giving the total displacement of the particle that is at point (x) at time t with respect to the
position the particle had at time t
0
.
The relation
v
i
(x, t) =
u
i
t
(x, t) (8.180)
is not valid in general: While v
i
(x, t) represents the velocity at time t of the particle that is at
point x , (u
i
/t)(x, t) represents the velocity at time t of the particle that was at point x
at time t
0
(and that, at time t is not more there, having been displaced by an amount u
i
(x, t) ).
Usually, one is interested (as in here) in small deformations. Then equation 8.180 is considered a valid
approximation.
We are interested in describing elastic waves. Usually, the change in density at a point when a
wave passes through is small and can be neglected. Then, the equation describing the conservation of
mass 8.173 can be dropped, and the density term in the equation describing the conservation of linear
momentum 8.174 can be taken out from the time derivative term, thus leading to

v
i
t

j

j
i
= 0 , (8.181)
or, if we use the approximation 8.180,

2
u
i
t
2

j

ji
= 0 . (8.182)
Let us now turn to the equation describing the conservation of angular momentum 8.175.
The spin density
ij
is antisymmetric. It can be represented by a vector s
i
(the dual tensor
of the previous chapter):
s
i
=
1
2

ijk

jk
;
ij
=
ijk
s
k
. (8.183)
Similarly, the moment stress m
k
ij
can also be represented by the tensor
ij
:

i
=
1
2

ijk
m

jk
; m

ij
=
ijk

k
. (8.184)
Then, equation 8.175 writes
s
i
t

j

ji
=
ijk

jk
. (8.185)
The spin density s
i
may have dierent interpretations in dierent models of continuous media.
In an elastic medium, it can be directly linked with the rotations of the particles of the medium. If

i
is the local rotation vector, then
s
i
= J
ij

j
t
, (8.186)
where J
ij
the local moment of inertia. As an example, consider a body with large molecules, like a
polymer, or a granular body, like an ordinary rock. In addition to the ordinary displacements, points
may obviously undergo rotations.
Equation 8.185 becomes then
J
ij

j
t
2

j

ji
=
ijk

jk
. (8.187)
8.4. CLASSICAL ELASTICITY 155
Equations 8.182 and 8.187 are the fundamental equations for elastic media. The stress
ij
and the
moment stress
ij
are functions of the deformation, which is, in turn, function of the displacements
u
i
and rotations
i
.
A nal modication will lead to the classical form of the equations. The theory developed here
ignores electromagnetic, chemical, and other phenomena. It is usually assumed that those only appear
as new right-hand (source) terms in the dynamical equations. Calling those respectively f
i
and
i
gives

2
u
i
t
2

j

ji
= f
i
(8.188)
and
J
ij

j
t
2

j

ji
=
ijk

jk
+
i
. (8.189)
Given the sources f
i
and
i
of the eld, the parameters describing the medium, and J
ij
,
and the particular relation between stresses and deformations (see for instance the linearized theory
of the next section), the set of equations 8.188 8.189 have an unique solution if we prescribe initial
conditions,
u
i
(x, t
0
)
u
i
t
(x, t
0
)
i
(x, t
0
)

i
t
(x, t
0
) , (8.190)
and conditions at the space boundaries. For instance, we can impose surface tractions and moment
tractions:

ji
(x, t)n
j
(x)
ji
(x, t)n
j
(x) for x at the surface . (8.191)
8.4.1 Elasticity with nonsymmetric stresses
A medium with non symmetric stresses is called a Cosserats micropolar medium. Early references to
media with non symmetric stress can be found in Cosserat and Cosserat (1896, 1907, 1909). Truesdell
and Toupin (1960) also mention in their general exposition, the basic equations of Cosserats media.
Nowacki (1986) gives a modern vision of the theory, and discusses the physical properties of micropolar
media.
The requirement that the expressions describing the conservation of mechanical energy should be
invariant by space translation and rotation imply (Nowacki, 1986; see also the Appendix)

ij

ij
t
+
ij

ij
t
= 0 , (8.192)
where

ij
=
i
u
j

ijk

k
(8.193)
and

ij
=
i

j
. (8.194)
Note: say somewhere that the physical dimensions of the tensors above are as follows:
[] = L
1
(8.195)
[] = 1 (8.196)
[u] = L (8.197)
[] = 1 . (8.198)
That the conservation of mechanical energy has a simple expression as a function of
ij
and
ij
suggests that those are good measures of deformation. The (non symmetric) tensor
ij
is called the
156 CHAPTER 8. FROM GEOMETRY TO PHYSICS
translational deformation or, simply, the deformation. The antisymmetric tensor
ij
is called the
rotational deformation or, also, the torsion.
In an elastic medium, we can express stress and moment stress as functions of the deformation. If
deformations are small enough, a rst order development can be used:

ij
= c
ijk

k
+b
ijk

k
(8.199)
and

ij
= b
ijk

k
+a
ijk

k
. (8.200)
Note: say somewhere that the physical dimensions of the tensors above are as follows:
[c] = [] (8.201)
[b] = []L (8.202)
[a] = []L
2
. (8.203)
The form of equations 8.1998.200, and, in particular, the fact that the same tensor b
ijk
appears in
both equations comes from a thermodynamic simplifying hypothesis, that the internal energy density
is quadratic on the deformations:
U =
1
2
c
ijk

ij

k
+
1
2
a
ijk

ij

k
+b
ijk

ij

k
. (8.204)
Then,
c
ijk
=
_

2
U

ij

k
_
, a
ijk
=
_

2
U

ij

k
_
, b
ijk
=
_

2
U

ij

k
_
. (8.205)
The expression of the internal energy density (equation 8.204) implies the symmetries
c
ijk
= c
kij
(8.206)
and
a
ijk
= a
kij
, (8.207)
while the tensor b
ijk
does not have any particular symmetry. This makes a total of 45+45+81 = 171
independent elastic coecients.
The asymmetric stresses play a role in situation of strong stress gradients occurring for instance
in the vicinity of cracks, or, more generally, in highly heterogeneous media. It could, for instance, well
happen that the understanding of the process of seismic rupture in elastic solids requires to drop the
usual hypothesis that the stress tensor is symmetric.
For isotropic media (Nowacki, 1986),
c
ijk
=
ij

k
+(
ik

j
+
i

jk
) +(
ik

jk
) , (8.208)
a
ijk
=
ij

k
+(
ik

j
+
i

jk
) +(
ik

jk
) , (8.209)
and
b
ijkl
= 0 . (8.210)
This gives

ij
=
k
k

ij
+(
ij
+
ji
) +(
ij

ji
) (8.211)
and

ij
=
k
k

ij
+(
ij
+
ji
) +(
ij

ji
) . (8.212)
8.4. CLASSICAL ELASTICITY 157
Note: say somewhere that the physical dimensions of the tensors above are as follows:
[] = [] = [] = [] . (8.213)
In addition to the Lames parameters and there are four other elastic parameters: , ,
, and . While , , and have the physical dimension of a pressure, , , and have
the physical dimension of a force.
As the tensors c
ijkl
and a
ijkl
have to be positive denite, the elastic parameters satisfy some
inequalities:
> 0 > 0 > 0 > 0 3 + 2 > 0 3 + 2 > 0 . (8.214)
Example (torsional waves): Replacing 8.211 and 8.193 in the dynamical equation 8.188 gives the
dierential equation for the displacement eld u
i
. Applying the operator
i
to the equation so
obtained, and dening
=
i
u
i
, (8.215)
leads to

t
2
( + 2)
2
=
i
f
i
, (8.216)
which demonstrates that the dilatation obeys the wave equation, with the (compressional wave)
velocity
c =

+ 2

. (8.217)
This is a clasical wave. Let us, instead, be interested in purely torsional waves. The variable
i
represents the angle of rotation of every point of the medium. We dene its divergence,
=
i

i
, (8.218)
and assume isotropy of local moment of inertia:
J
i
j
= J
i
j
. (8.219)
Equation 8.189 then leads to
1
c
2

t
2

2
+
1
L
2
=
i

i
, (8.220)
where
c =

+ 2
J
(8.221)
has the dimension of a speed, and where
L =

+ 2
4
(8.222)
has the dimension of a length. This is the well known Klein-Gordon equation, whose solution is
analyzed in Appendix ??.
This wave is missing in the theory where the stress is assumed symmetric. When this wave
arrives to a region of the space, all points experience a rotation, without any displacement.
Few experiments have been performed to test the theory of elasticity with non symmetric stresses.
On the other side, it is well known that beam models, where a regular collection of beams is
assembled to form a 3-D structure (like in concrete buildings), have as limit such a Cosserat medium
(Herrmann, 1989; Herrmann et al., 1989; see also Kaliski, 1963, and Askar and Cakmak, 1968).
158 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.4.2 Elasticity with symmetric stresses
If local rotations and moment stresses may be neglected, we have the standard theory of elastic media.
Then equation 8.187 simply tells us that

ij
=
ji
, (8.223)
i.e., the stress tensor is symmetric.
We are then left with

2
u
i
t
2

j

ji
= f
i
, (8.224)

ij
= c
ijk

k
(8.225)
and

ij
=
1
2
(
j
u
i
+
i
u
j
) . (8.226)
This makes a total of 21 independent elastic parameters.
For isotropic media,
c
ijk
=
ij

k
+(
ik

j
+
i

jk
) . (8.227)
This gives

ij
=
k
k

ij
+ 2
ij
. (8.228)
Note: I have to explain that the tensor
ij
in equation 8.227 is the metric, g
ij
. Thus, I should
rather write
c
ijk
= g
ij
g
k
+(g
ik
g
j
+g
i
g
jk
) (8.229)
and

ij
= g
ij

k
k
+ 2
ij
. (8.230)
I should also correct equations 8.2088.209.
8.5. APPENDICES 159
8.5 Appendices
8.5.1 Appendix: Model of an electron
Let us rst dene the constants
A
s
=
e
2
c
2
8
3
G
0
h
, (8.231)
A
e
=
e
3
c
2
4
3
G
0
h
2
, (8.232)
and
L =
h
ec
_
G
0
, (8.233)
where e is the electric charge of an electron and
0
is the dielectric constant of vacuo.
Numerically, we have
A
s
= 1.5 10
32
kg s
1
, (8.234)
A
e
= 4.5 10
47
Cm
2
, (8.235)
and
L = 9.5 10
35
m . (8.236)
Let k
i
be an unit (constant) vector pointing upwards, and r a radial variable.
Then, the radial distribution of spin density is (we use here denition 8.183)
s
i
= k
i
A
s
1
r
sin
r
L
(8.237)
and the radial distribution of electric charge

e
= A
e
1
r
sin
r
L
(8.238)
have the following properties (note that L is the rst zero of the sinc function):
The total spin inside r < L is h/2 ,
The total electric charge inside r < L is e ,
At every point, the electrostatic (repulsive) force density exactly equilibrates the (attractive)
gravitational spin-spin force density,
The (attractive) gravitational mass-mass force density is orders of magnitude smaller.
I.e., we have a model of an electron.
The spin-spin force density is computed from
f
i
=
G
c
2

i
(
jk

jk
) =
2G
c
2

i
(s
j
s
j
) , (8.239)
while the Coulombs force density is computed from Coulombs law:
f
i
=
1
4
0

e
(r)Q(r)
r
2
, (8.240)
where Q(r) is the total electric charge inside the sphere of radius r .
We obtain for both force densities the expression
[f
i
[ =
e
5
c
3
16
1/2
G
3/2

3/2
h
3
1
r
2
sin
r
L
cos
r
L
_
1
L
r
tg
r
L
_
, (8.241)
the gravitational spin-spin force density pointing inward and the electrostatic one pointing outward.
The choice r < L for the radius of the electron is quite arbitrary: Any nite value of r
would have done as well.
This shows that simple models of an electron exist where the Coulombs electrostatic force is
exactly compensated by the gravitational (spin-spin) force.
160 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.5.2 Appendix: The Einstein-Cartan equations in 3-D form
The Einstein equations
The Einstein equation give the stress-energy tensor as a function of the Ricci tensor:
T

=
1

_
R

1
2

R
_
. (8.242)
It is easy to obtain the inverse relation:
R

=
_
T

1
2

T
_
. (8.243)
In 3-D notations, using the denitions introduced in sections 8.1.2, 8.1.3, and 8.1.6 gives, for the
direct Einstein equation,
=
1
2c
2
r
s
s
, (8.244)
p
i
=
1

si
s
, (8.245)
q
j
= +
1

s
js
, (8.246)
and
T
i
j
=
1

_
c
2
_

i
j

i
j

k
k
_
+
_
r
i
j

1
2

i
j
r
k
k
__
, (8.247)
and, for the inverse Einstein equation, the (equivalent) expressions

s
s
=

2
_
+
1
c
2
T
s
s
_
, (8.248)

si
s
= p
i
, (8.249)

s
js
= +q
j
, (8.250)
and
r
i
j
+
i
j
c
2
=
_
T
i
j
+
1
2

i
j
_
c
2
T
k
k
_
_
. (8.251)
The Cartan equations
The Cartan equation gives the spin tensor as a function of the torsion tensor:
M

=
1

_
S

+S

_
. (8.252)
The inverse relation is
S

=
_
M

+
1
2
_
M

_
_
. (8.253)
In 3-D notations, using the denitions introduced in sections 8.1.1, 8.1.2, 8.1.3, and 8.1.6 gives, for
the direct Cartan equation,

ij
=
1

J
ij
, (8.254)

i
j
=
1

_
K
i
j
K
k
k

i
j
_
, (8.255)
8.5. APPENDICES 161
t
i
=
1
c
2
S
ji
j
, (8.256)
and
M
ij
k
=
1

_
S
ij
k
+S
i

j
k
S
j

i
k
+c
2
(I
i

j
k
I
j

i
k
)
_
(8.257)
and, for the inverse Cartan equation, the (equivalent) expressions

J
ij
=
ij
, (8.258)
K
i
j
=
_

i
j

1
2

k
k

i
j
_
, (8.259)
S
ji
j
= c
2
t
i
, (8.260)
S
ij
k
=
_
M
ij
k
+
1
2
_
M
i

j
k
M
j

i
j
_
+
c
2
2
_
t
i

j
k
t
j

i
k
_
_
. (8.261)
The Cartan equations are somewhat simplied if, from the spin tensor M
ij
k
and the torsion
tensor S
ij
k
, we introduce their traces
M
i
= M
ji
j
(8.262)
S
i
= S
ji
j
(8.263)
and the traceless tensors

M
ij
k
= M
ij
k
+
1
2
_
M
i

j
k
M
j

i
k
_
(8.264)
and

S
ij
k
= S
ij
k
+
1
2
_
S
i

j
k
S
j

i
k
_
. (8.265)
Then, obviously,
M
ij
k
=

M
ij
k

1
2
_
M
i

j
k
M
j

i
k
_
(8.266)
and
S
ij
k
=

S
ij
k

1
2
_
S
i

j
k
S
j

i
k
_
. (8.267)
With those denitions, the whole set of direct Cartan equations are now written

ij
=
1

J
ij
, (8.268)

i
j
=
1

_
K
i
j
K
k
k

i
j
_
, (8.269)
t
i
=
1
c
2
S
i
, (8.270)
M
i
=
1

_
S
i
+ 2c
2
I
i
_
, (8.271)
and

M
ij
k
=
1

S
ij
k
, (8.272)
while the inverse Cartan set of equations is written

J
ij
=
ij
, (8.273)
162 CHAPTER 8. FROM GEOMETRY TO PHYSICS
K
i
j
=
_

i
j

1
2

k
k

i
j
_
, (8.274)
S
i
= c
2
t
i
, (8.275)
I
i
=

2
_
t
i

1
c
2
M
i
_
, (8.276)
and

S
ij
k
=

M
ij
k
. (8.277)
8.5.3 Appendix: 3-D evolution equations for a medium locally at rest
We show here that, given g
ij
,

G
ij
,
ij
, and t
i
, at t = t
0
, and assuming given some constitutive
equations that allow us to compute p
i
p
i
q
i
, T
i
j
,
i
j
, and M
ij
k
at any time as a function
of the history of the medium, as well as assuming given U at all times, then, it is possible to
extrapolate the values of g
ij
,

G
ij
,
ij
, and t
i
, at all times t > t
0
.
In section 8.1.6, we have seen the Einstein-Cartan equations. Rather, we will use here the more
convenient form of these equations derived in appendix 8.5.2, equations 8.248 and 8.251:

s
s
=

2
_
+
1
c
2
T
s
s
_
, (8.278)
r
i
j
+
i
j
c
2
=
_
T
i
j
+
1
2

i
j
_
c
2
T
k
k
_
_
, (8.279)
and equations 8.2758.277:
S
i
= c
2
t
i
, (8.280)
I
i
=

2
_
t
i

1
c
2
M
i
_
, (8.281)
and

S
ij
k
=

M
ij
k
. (8.282)
Let us rst see what we can compute at any t .
As U is given for all times, using 6.13 we compute :
= exp(U U
0
) (8.283)
( U
0
being an arbitrary constant). From U we can get the gravitational eld g
i
(equation 6.14):
g
i
=
1
c
2

i
U . (8.284)
As we assume tha the constitutive equations will give us at any time the moment stresses, we will be
able to compute, using 8.264,

M
ij
k
= M
ij
k
+
1
2
_
M
i

j
k
M
j

i
k
_
(8.285)
and, using 8.277,

S
ij
k
=

M
ij
k
. (8.286)
The constitutive equations giving also
i
j
, using 8.274we can obtain
K
i
j
=
_

i
j

1
2

i
j

k
k
_
. (8.287)
8.5. APPENDICES 163
The denitions of symmetric and antisymmetric part of a tensor (equations 8.15 and 8.16) give then

K
ij
= K
ij
+K
ji
(8.288)
and

K
ij
= K
ij
K
ji
. (8.289)
Now, at t = t
0
we can successively computeusing equations all written in the text
S
i
= c
2
t
i
(8.290)
S
ij
k
=

S
ij
k

1
2
_
S
i

j
k
S
j

i
k
_
(8.291)
I
i
=

2
_
t
i

1
c
2
M
i
_
(8.292)
h
i
= g
i
I
i
(8.293)

J
ij
=
ij
(8.294)

ijk
=
1
2
(
i
g
jk
+
j
g
ik

k
g
ij
) +
1
2
(S
ijk
+S
kij
+S
kji
) (8.295)
R
ijk
l
=
i

jk
l

ik
l
+
is
l

jk
s

js
l

ik
s
(8.296)

G
ij
=

J
ij
+

K
ij
(8.297)
G
ij
=
1
2
_

G
ij
+

G
ij
_
(8.298)

J
ij
=

G
ij


K
ij
(8.299)
J
ij
=
1
2
_

J
ij
+

J
ij
_
(8.300)
r
ij
=

R
ij
+c
2
J
k
k

J
ij
c
2
(J
i
k
J
kj
+J
j
k
J
ki
) (8.301)

R
ij
=

R
ij
+c
2
J
k
k

J
ij
c
2
(J
i
k
J
kj
J
j
k
J
ki
) (8.302)
=
1
2c
2
r
k
k
(8.303)

i
=
k
J
i
k
+
i
J
k
k
+S
ki
l
J
l
k
(8.304)
p
i
=
1

i
(8.305)

i
= q
i
= (p
i
+ p
i
) (8.306)
c
2

i
j
= r
i
j
+(

T
i
j
+
i
j
(c
2
T
k
k
)) (8.307)
and
c
2

i
j
= r
i
j
+

T
i
j
. (8.308)
Then, at t = t
0
we can deduce

g
ij
= c
2

G
ij
(8.309)

G
ij
=
ij
+

K
ij
(
i
h
j
+
j
h
I
) + (g
i
h
j
+g
j
h
i
) + (J
i
s
G
js
+J
j
s
G
is
) (8.310)

ij
=
ij
(
i
h
j

j
h
I
) + (g
i
h
j
g
j
h
i
) + (J
i
s
G
js
J
j
s
G
is
) (8.311)
and

t
i
=
i

1
2

i
G
k
k
+
k
G
i
k
g
k
G
i
k
J
k
k
h
i
+J
ik
h
k
. (8.312)
As these are the quantities given at t = t
0
, we can extrapolate them for t > t
0
, and the problem of
extrapolation of initial conditions is solved.
164 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.5.4 Appendix: Strain tensor
Homogeneous 1-D strain
Assume, in a 1-D world, an object of length which, in an homogeneous deformation, elongates by
d . The incremental strain is dened as the relative elongation:
d =
d

. (8.313)
The integration =
_

0
d

denes then the strain:

= log

0
(8.314)
Equivalently, we can write
=
0
e

. (8.315)
General 1-D strain
A general (i.e., inhomogeneous) deformation of a 1-D medium can be described, for instance, by giving
the function s(s
0
) dening the distance s to some reference point, after deformation, of the material
point whose distance to the same reference point was s
0
before deformation
1
. When applied to this
situation, formula 8.314 gives (see gure 8.1)
= log
_
lim
s
0
0
s
s
0
_
(8.316)
i.e.,
(s
0
) = log
_
ds
ds
0
(s
0
)
_
, (8.317)
which is the main denition of strain.
An integration gives the position as a function of the strain:
s(s
0
) =
_
s
0
0
ds

e
(s

)
. (8.318)
We have
s(s
0
+ s
0
) =
_
s
0
+s
0
0
ds

e
(s

)
=
_
s
0
0
ds

e
(s

)
+
_
s
0
+s
0
s
0
ds

e
(s

)
(8.319)
= s(s
0
) +
_
s
0
+s
0
s
0
ds

e
(s

)
.
Assume that s
0
is small enough, and use the notation s
0
. We can then use the rst order
approximation
s(s
0
+s
0
) = s(s
0
) +e
(s
0
)
s
0
, (8.320)
i.e,
s(s
0
+s
0
) s(s
0
) = e
(s
0
)
s
0
, (8.321)
1
These points are material points, i.e., points attached to the medium undergoing deformation.
8.5. APPENDICES 165
which can be written
s = e

s
0
(8.322)
or
s
2
= e
2
s
2
0
, (8.323)
i.e.,
s
2
s
2
0
=
_
e
2
1
_
s
2
0
. (8.324)
This is a very important equation. It relates the change in length of a small segment at a given point
to the strain at that point. Notice that although the segment is assumed small, the deformation may
arbitrarily large: this equation may be used to dene a nite strain.
For a small deformation (i.e., for a small value of ),
s
2
s
2
0
2 s
2
0
2 s
2
. (8.325)
3-D deformation
Consider, in an arbitrary space furnished with an arbitrary coordinate system, a deformable medium
and, on it, two neighbouring points x
0
and x
0
+x
0
with respective coordinates x
i
0
and x
i
0
+x
i
0
.
They are separated by a distance s
0
that can be computed using the metric tensor:
s
2
0
= g
ij
(x
0
) x
i
0
x
i
0
. (8.326)
If the medium undergoes (inhomogeneous) deformation, the two points will have new coordinates x
i
and x
i
+x
i
, and the new distance between the two points, s can be computed from the expression
s
2
= g
ij
(x) x
i
x
i
. (8.327)
Equation 8.324 suggests to dene the strain eld by the expression
2
.
s
2
s
2
0
=
_
e
2
ij
g
ij
_
x
i
0
x
j
0
. (8.328)
(Note: I should explain here why the coordinate increments x
i
have replaced the length increment
s of equation 8.324.)
For a small deformation (i.e., for a small value of ),
s
2
s
2
0
= 2
ij
(x
0
) x
i
0
x
j
0
(8.329)
or, equivalently,
s
2
s
2
0
= 2
ij
(x) x
i
x
j
, (8.330)
the two expressions corresponding respectively to a Lagrangian or Eulerian description of the defor-
mation.
In a space of dimension N , to compute the strain tensor at a given point we may, for instance,
consider at the given point point N(N + 1)/2 small noncolinear vectors. The comparison of their
lengths before and after deformation through equations 8.329 or 8.330 provides the necessary conditions
to obtain the N(N + 1)/2 independent components of the strain tensor. Practically, choosing
N(N + 1)/2 small noncolinear vectors may mean, for instance, considering the three sides of a small
triangle in 2-D, of the six edges of a small tetrahedron in 3-D.
2
The exponential of a tensor i
j
should, in all rigor, be written (e

)i
j
, and is dened by e

i
j
= i
j
+ i
j
+
1
2!
i
k

k
j
+
1
3!
i
k

j
+ . . .. Remember that the metric tensor gij appearing in the equation is identical to the
Kroneckers tensor: gi
j
= i
j
(see chapter 1).
166 CHAPTER 8. FROM GEOMETRY TO PHYSICS
s
0
s(s
0
)
s
s
0
Figure 8.1: Deformation af a 1-D medium.
Example: Homogeneous deformation in polar coordinates
Consider the Euclidean plane. Using rst Cartesian coodinates, it is well known that a homogeneous
deformation is described by a constant deformation eld, with components
_

xx

xy

yx

yy
_
=
_
0
0
_
, (8.331)
where is a dimensionless constant taking any positive or negative small value. Two points separated
by a distance s
0
before deformation, will be separated, after deformation, by a distance s =
(1 +) s
0
.
The components of this deformation tensor when using polar coordinates may be obtained from
equation 8.331 using, for instance, the general rules for transformation of the components of a tensor
under a coordinate change (see section XXX). This gives
_

rr

r

_
=
_
0
0 r
2
_
. (8.332)
Let us obtain this result from a direct use of equations 8.329 and 8.330.
In the left (resp. right) column, we will make use of equation 8.329 (resp. 8.330).
double colonne
8.5. APPENDICES 167
The radial segment shown in gure 8.3 has
a length, before deformation,
s
0
= r
0
, (8.333)
and, after deformation,
s = r = (1 +) r
0
. (8.334)
This gives
s
2
s
2
0
=
_
(1 +)
2
1
_
r
2
0
=
_
2 +
2
_
r
2
0
2 r
2
0
. (8.335)
The length of the tangential segment is, be-
fore deformation,
s
0
= r
0

0
(8.336)
and, after deformation,
s = r = (1 +) r
0

0
. (8.337)
This gives
s
2
s
2
0
=
_
(1 +)
2
1
_
r
2
0

2
0
=
_
2 +
2
_
r
2
0

2
0
2 r
2
0

2
0
. (8.338)
The strain tensor was dened by equa-
tion 8.329:
s
2
s
2
0
= 2
ij
(x
0
) x
i
0
x
j
0
.
Using the results 8.335 and 8.338 gives then

rr
(r
0
,
0
) = (8.339)
and

(r
0
,
0
) = r
2
0
(8.340)
(demonstrating that
r
= 0 would require
the consideration of a third vector).
The radial segment shown in gure 8.3 has
a length, before deformation,
s
0
= r
0
=
1
1 +
r , (8.341)
and, after deformation,
s = r . (8.342)
This gives
s
2
s
2
0
=
_
1
1
(1 +)
2
_
r
2
=
2 +
2
(1 +)
2
r
2
2 r
2
. (8.343)
The length of the tangential segment is, be-
fore deformation,
s
0
= r
0

0
=
1
1 +
r (8.344)
and, after deformation,
s = r . (8.345)
This gives
s
2
s
2
0
=
_
1
1
(1 +)
2
_
r
2

2
=
2 +
2
(1 +)
2
r
2

2
2 r
2

2
. (8.346)
The strain tensor was dened by equa-
tion 8.330:
s
2
s
2
0
= 2
ij
(x) x
i
x
j
.
Using the results 8.343 and 8.346 gives then

rr
(r, ) = (8.347)
and

(r, ) = r
2
(8.348)
(demonstrating that
r
= 0 would require
the consideration of a third vector).
168 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.5.5 Appendix: The deformation as a function of the displacement vector
It should rst be understood that small deformations is not equivalent to small displacements. For
consider that a solid translation gives no deformation but arbitrarily large displacements.
Second, it is important to realize that the notion of displacement vector makes only general
sense in Euclidean spaces, or in general spaces only if displacements are so small that we can consider
that they belong to the tangent space.
In the sections before we have seen that the direct denition of the strain tensor has operational
meaning, and that no other expression is needed. But as it is common usage to introduce a displace-
ment vector and to use it for dening the strain, we are going to obtain the corresponding formulas
here.
We face here one notational problem. The equations below, when developed in a general space,
need an extensive use of the parallel transport of vectors and tensors, and become quite intricate. To
avoid this complication, we are going to work here only with Euclidean spaces, and using Cartesian
coordinates. The nal formulas will then be generalized to arbitrary coordinate systems (but still for
Euclidean spaces).
In what follows we will dene the deformation of a medium at some time t with respect to an
undeformed state at time t
0
.
8.5. APPENDICES 169
As we did before, we will develop at left (resp. at right) the formulas correspondig to the denition
of the strain tensor given by equation 8.329 (resp. 8.330), corresponding respectively to a Lagrangian
or Eulerian description.
Let us dene the deformation of the
medium by the functions
x
i
= x
i
(x
0
, t) (8.349)
giving the coordinates at time t of the mate-
rial point whose coordinates at time t
0
were
x
i
0
. Two points that, before deformation, were
separated by a (squared) distance
s
2
(t
0
) = g
ij
x
i
0
x
j
0
(8.350)
will be separated, after deformation, by the
(squared) distance
s
2
(t) = g
k
x
k
x

(8.351)
= g
k
x
k
x
i
0
(x
0
, t)x
i
0
x

x
j
0
(x
0
, t)x
j
0
.
This easily gives
s
2
(t) s
2
(t
0
) = (8.352)
=
_
g
k
x
k
x
i
0
(x
0
, t)
x

x
j
0
(x
0
, t) g
ij
_
x
i
0
x
j
0
,
and, using the denition 8.329 of the strain ten-
sor,

ij
(x
0
, t) =
1
2
_
g
k
x
k
x
i
0
(x
0
, t)
x

x
j
0
(x
0
, t) g
ij
_
.
(8.353)
If a displacement vector u
i
(x
0
, t) is intro-
duced through
x
i
(x
0
, t) = x
i
0
+u
i
(x
0
, t) (8.354)
then, developing the expression 8.353 we obtain
(dropping the variables (x
0
, t) everywhere)

ij
=
1
2
_
u
i
x
j
0
+
u
j
x
i
0
+
u
k
x
i
0
u
k
x
j
0
_
. (8.355)
In a general coordinate system, this equa-
tion generalizes into

ij
=
1
2
_

i
u
j
+
j
u
i
+
i
u
k

j
u
k
_
. (8.356)
Let us dene the deformation of the
medium by the functions
x
i
0
= x
i
0
(x, t) (8.357)
giving the coordinates at time t
0
of the ma-
terial point whose coordinates at time t are
x
i
. Two points that, after deformation, are
separated by a (squared) distance
s
2
(t) = g
ij
x
i
x
j
(8.358)
were separated, before deformation, by the
(squared) distance
s
2
(t
0
) = g
k
x
k
0
x

0
(8.359)
= g
k
x
k
0
x
i
(x, t)x
i
x

0
x
j
(x, t)x
j
.
This easily gives
s
2
(t) s
2
(t
0
) = (8.360)
=
_
g
ij
g
k
x
k
0
x
i
(x, t)
x

0
x
j
(x, t)
_
x
i
x
j
,
and, using the denition 8.330 of the strain ten-
sor,

ij
(x
0
, t) =
1
2
_
g
ij
g
k
x
k
0
x
i
(x, t)
x

0
x
j
(x, t)
_
.
(8.361)
If a displacement vector v
i
(x, t) is intro-
duced through
x
i
0
(x, t) = x
i
v
i
(x, t) , (8.362)
then, developing the expression 8.361 we obtain
(dropping the variables (x, t) everywhere)

ij
=
1
2
_
v
i
x
j
+
v
j
x
i

v
k
x
i
v
k
x
j
_
. (8.363)
In a general coordinate system, this equa-
tion generalizes into

ij
=
1
2
_

i
v
j
+
j
v
i

i
v
k

j
v
k
_
. (8.364)
170 CHAPTER 8. FROM GEOMETRY TO PHYSICS
t
0
s(t
0
)
s(t)
x+x
x
x
0
x
0
+x
0
t
Figure 8.2: Two neighbouring points x
0
and x
0
+x
0
are separated by a distance s
0
that can be
computed using the metric tensor: s
2
0
= g
ij
(x
0
) x
i
0
x
i
0
. After a deformation, the two points will
have new coordinates x
i
and x
i
+ x
i
, and the new distance between the two points, s can be
computed from the expression s
2
= g
ij
(x) x
i
x
i
. The strain eld can be dened by the expression
s
2
s
2
0
= 2
ij
(x
0
) x
i
0
x
j
0
or, equivalently, by s
2
s
2
0
= 2
ij
(x) x
i
x
j
. .
Figure 8.3: Under an homogeneous deformation, the two small vectors dened, in polar coordinates, by
the components (r
0
, 0) and (0,
0
) , transform into vectors whose components are ((1 +) r
0
, 0)
and (0,
0
) . The denition s
2
s
2
0
= 2
ij
(x
0
) x
i
0
x
j
0
and/or the denition s
2
s
2
0
=
2
ij
(x) x
i
x
j
allow a direct calculation of the strain tensor (see text).
8.5.6 Appendix: Physical dimensions of tensors
Let us be specic about the physical dimension of all elds considered. Mass, lenght and time dimen-
sions will be respectively denoted by M , L , and T .
By a language abuse, we will write something like [g
i
] = L
3
T
2
, which, in all rigor, would mean
that the components of the vector g have as dimension L
3
T
2
. This, of course would not be
correct, as the physical dimension of the components of any tensor depend on the coordinates chosen
(think, for instance, that the physical dimension of the components of a tensor change if we replace,
say, cartesian by spherical coordinates). So when we write [g
i
] = L
3
T
2
, we mean that it is the norm
of the vector g that has such a dimension.
The space-time vector dx
0
, dx
i
= c dt, dx
i
is assumed to have the physical dimension of a
length.
Then, the main tensors considered in this chapter have the following physical dimensions:
[U] = [] = [g
ij
] = 1
[S
ij
k
] = [
ij
k
] = L
1
[J
ij
] = [K
i
j
] = [A
ij
] = [B
ij
] = L
2
T
[I
i
] = [h
i
] = [g
i
] = L
3
T
2
[] = L
4
T
3
[r
ij
k
] = L
2
[
i
jk
] = [
ij
k
] = L
3
T
[
i
j
] = L
4
T
2
8.5. APPENDICES 171
[T
i
j
] = ML
1
T
2
[p
i
] = [q
i
] = ML
2
T
1
[] = ML
3
[M
ij
k
] = MT
2
[
i
j
] = [
ij
] = ML
1
T
1
and
[t
i
] = ML
2
.
Note: Say somewhere that, in the section on torsional waves, while , , and have the
physical dimension of a pressure, , , and have the physical dimension of a force.
8.5.7 Appendix: Shearing, stretching, bending and twisting
Consider a point on a surface inside a continuous medium, and let n
i
be the normal to the surface
at the considered point. Let T
i
j
be the stress at that point, and M
ij
k
the moment stress.
The vector

i
= T
i
j
n
j
(8.365)
is called the traction, and the antisymmetric tensor

ij
= M
ij
k
n
k
(8.366)
is called the moment traction. Instead of the antisymmetric tensor
ij
, we can use its dual vector

i
=
1
2

ijk

jk
. (8.367)
The decompositions

..
= (n ) n
. .
+ (n ) n
. .
..
traction =
..
normal traction +
..
shearing traction
(8.368)
and

..
= (n ) n
. .
+ (n ) n
. .
..
couple (s. d.) =
..
twisting couple (s. d.) +
..
bending couple (s. d.)
(8.369)
project and respectively onto the normal and onto the considered plane.
The letters (s. d.) stand there because we do not have couples, but surface densities of couples.
A surface density of a force is called a traction, but there is no common name for the surface density
of a couple.
To understand how these dierent eorts act, we may rst imagine a crystal, and a plane separating
two atomic planes (see gure). It is easy to imagine which sort of forces between atoms correspond to
normal tension and shearing stress. An homogeneous twisting couple in the plane should correspond
to a twisting couple of each atom facing the plane on the atom at the other side of the plane. The
reader should realize that a global twist of the crystal (obtained by some external forces) does not
generate this sort of (microscopical) twisting couples: the microscopical eects of a global twist in an
ordinary elastic homogeneous medium generate only tractions, and not moment tractions.
172 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.5.8 Appendix: Tolos letter
Primer punto de vista (que no me gusta)
Usando un sistema de coordenadas arrastradas por el cuerpo en su deformaci on, todo punto material
tiene coordenadas constantes. El tensor metrico depende del tiempo. En un punto de coordenadas
x
i
, un peque no vector de componentes dx
i
tiene, at instante t , una longitud ds(t) denida por
la expresi on
ds
2
(t) = g
ij
(x, t) dx
i
dx
j
. (8.370)
Tenemos
ds
2
(t) ds
2
(t
0
) = (g
ij
(x, t) g
ij
(x, t
0
)) dx
i
dx
j
, (8.371)
y se dene el tensor de deformaci on
2 e
ij
(x; t, t
0
) = g
ij
(x, t) g
ij
(x, t
0
) . (8.372)
Entonces
ds
2
(t) ds
2
(t
0
) = 2 e
ij
(x; t, t
0
) dx
i
dx
j
. (8.373)
Segundo punto de vista (que me gusta mas)
El sistema de coordenadas esta denido independientemente del movimiento del cuerpo. El tensor
metrico no depende del tiempo. Todo punto material se representa por las coordenadas x
i
que
tenia en cierto instante t
0
, antes de la deformacion. La deformaci on del cuerpo estar a perfectamente
denida si se conocen las coordenadas y
i
de todo punto en todo tiempo t . Estas coordenadas
pueden estar dadas, por ejemplo, por las funciones
y
i
= y
i
(x, t) . (8.374)
Por hyp otesis,
y
i
(x, t
0
) = x
i
. (8.375)
Sea, en el instante t
0
y en el punto de coordenadas x
i
, un peque no vector de componentes dx
i
,
y sea ds(t
0
) su longitud. En el instante t el vector estara en el punto de coordenadas y
i
(x, t) ,
tendr a componentes dy
i
(t) y longitud ds(t) .
Tenemos
ds
2
(t
0
) = g
ij
(y(x, t
0
)) dy
i
(t
0
) dy
j
(t
0
) = g
ij
(x) dx
i
dx
j
(8.376)
y
ds
2
(t) = g
ij
(y(x, t)) dy
i
(t) dy
j
(t) = g
ij
(y(x, t))
y
i
x
k
(x, t) dx
k
y
j
x

(x, t) dx

. (8.377)
Deniendo
2 e
ij
(x; t, t
0
) = g
k
(y(x, t))
y
k
x
i
(x, t)
y

x
j
(x, t) g
ij
(y(x, t
0
))
= g
k
(y(x, t))
y
k
x
i
(x, t)
y

x
j
(x, t) g
ij
(x) , (8.378)
se tiene
ds
2
(t) ds
2
(t
0
) = 2 e
ij
(x; t, t
0
) dx
i
dx
j
. (8.379)
La deformaci on as denida no ha sido linearizada. Por consiguente, es v alida incluso para defor-
maciones nitas.
A n de cuentas la denici on 8.378 es identica a 8.372, si interpretamos los y
i
como un nuevo
sistema de coordenadas.
8.5. APPENDICES 173
Medios elasticos (por ejemplo)
La deformaci on introducida arriba es la que entra en la denici on de un cuerpo el astico, donde el
tensor de los esfuerzos,
ij
en el punto x y en el instante t , depende solo de la deformaci on en
este punto y en este instante:

ij
(x, t) = f
ij
(e(x, t)) . (8.380)
Para un cuerpo el astico lineal (ley de Hooke),

ij
(x, t) =
(0)
ij
(x) +c
ij
k
(x) e
k
(x, t) . (8.381)
Medio micropolar
Nota: aqu, de momento, se olvida todo rigor, y se hace todo a la americana, quick and dirty.
Consider a body with large molecules, like a polymer, or a granular body, like an ordinary rock.
In addition to the ordinary displacements u
i
, points may undergo rotations
i
.
We have the equations of conservation of linear and angular momentum:

2
u
i
t
2

j

ji
= 0 (8.382)
and
J
ij

j
t
2

j

ji
=
ijk

jk
. (8.383)
Aqu, J
ij
representa la densidad de momento de inercia, y
ij
el tensor momento de esfuerzos.
Un tal medio, con un tensor de esfuerzos
ij
no simetrico, y un tensor
ij
no nulo es un medio
micropolar de Cosserat.
Las deformacion ordinaria est a denida por

ij
=
i
u
j

ijk

k
(8.384)
y la deformaci on angular por

ij
=
i

j
. (8.385)
If deformations are small enough, a rst order development can be used:

ij
= c
ijk

k
+b
ijk

k
(8.386)
and

ij
= b
ijk

k
+a
ijk

k
. (8.387)
Arguments based on the internal energy density imply the symmetries
c
ijk
= c
kij
(8.388)
and
a
ijk
= a
kij
, (8.389)
while the tensor b
ijk
does not have any particular symmetry. This makes a total of 45+45+81 = 171
independent elastic coecients.
For isotropic media (Nowacki, 1986),
c
ijk
=
ij

k
+(
ik

j
+
i

jk
) +(
ik

jk
) , (8.390)
a
ijk
=
ij

k
+(
ik

j
+
i

jk
) +(
ik

jk
) , (8.391)
174 CHAPTER 8. FROM GEOMETRY TO PHYSICS
and
b
ijkl
= 0 . (8.392)
This gives

ij
=
k
k

ij
+(
ij
+
ji
) +(
ij

ji
) (8.393)
and

ij
=
k
k

ij
+(
ij
+
ji
) +(
ij

ji
) . (8.394)
La pregunta
La deformaci on
ij
utilizada para el medio micropolar se dene correctamente como en las sec-
ciones 8.5.88.5.8 arriba. C omo se dene correctamente la deformacion angular
ij
?
Mi impresi on es que si la deformacion ordinaria es simplemente la diferencia entre el tensor metrico
nal e inicial, la deformaci on angular deber ser la diferencia entre el tensor de torsi on nal e inicial.
Si la torsi on se escribe, en funcion de la conexi on como
S
ij
k
=
ij
k

ji
k
, (8.395)
se puede introducir el tensor
s
i
=
ijk
S
jk

. (8.396)
Entonces, la deformaci on angular seria

ij
(x; t, t
0
) = s
ij
(x, t) s
ij
(x, t
0
) . (8.397)
Pero no veo como justicar esto. Crees tu que es cierto? Estando t u convencido como estas de que
la torsi on del espacio es digna de estudio, no seria interesante tener para la torsi on la misma analogia
que para la metrica con la deformaci on de un cuerpo el astico?
8.5. APPENDICES 175
8.5.9 Appendix: Maxwells equations
Comment: for the time being, this section is written for special relativity (generalize?), and for media
at rest (generalize?).
Note: I can not underline here the , that may represent the Levi-Civita capacity.
The rst electromagnetic tensor: The rst electromagnetic tensor will be here denoted F

.
We will see that its 3-D components correspond to the elds E and B . The dual of this tensor will
be denoted f

. We have then the equivalent equations

F

=
1
2!

=
1
2!

. (8.398)
In terms of 3-D elds we have (note: given the 4-velocity, the introduction of the 3-D can be done
rigorously [to be done?]), equivalently,
_
F
00
F
0j
F
i0
F
ij
_
=
_
0 B
j
c
B
i
c e
ij
_

_
f
00
f
0j
f
i0
f
ij
_
=
_
0 E
j
E
i
b
ij
c
_
, (8.399)
where we have the 3-D duality relations
e
ij
=
1
1!

ijk
E
k
E
i
=
1
2!

ijk
e
jk
(8.400)
B
i
=
1
2!

ijk
b
jk
b
ij
=
1
1!

ijk
B
k
. (8.401)
The rst Maxwell equation: The rst Maxwell equation states that the tensor F

is divergence
free. This gives (writing also the equivalent dual equation)

= 0

= 0 . (8.402)
The second electromagnetic tensor: The second electromagnetic tensor will be here denoted
K

. We will see that its 3-D components correspond to the elds H and D . The dual of this
tensor will be denoted k

. We have then the equivalent equations

K

=
1
2!

=
1
2!

. (8.403)
In terms of 3-D elds we have, equivalently,
_
K
00
K
0j
K
i0
K
ij
_
=
_
0 D
j
c
D
i
c h
ij
_

_
k
00
k
0j
k
i0
k
ij
_
=
_
0 H
j
H
i
d
ij
c
_
. (8.404)
where we have the 3-D duality relations
h
ij
=
1
1!

ijk
H
k
H
i
=
1
2!

ijk
h
jk
(8.405)
D
i
=
1
2!

ijk
d
jk
d
ij
=
1
1!

ijk
D
k
. (8.406)
176 CHAPTER 8. FROM GEOMETRY TO PHYSICS
The electric current: The electric current is described using the vector J

, or the dual tensor

j

. We have the equivalent equations

J

=
1
3!

=
1
1!

. (8.407)
In terms of 3-D elds we have, equivalently (CHECK SIGNS HERE!),
_
J
0
J
i
_
=
_
c
j
i
_

j
123
j
230
j
301
j
012
_

_
=
_

_
c
j
1
j
2
j
3
, (8.408)
where, the other components not shown at the right hand side are zero or are obtained by simple
permutaion of indices.
The second Maxwell equation: The second Maxwell equation states that the divergence of the
tensor K

equals the electric current vector, i.,e., (writing also the equivalent dual equation)

= J

+j

= 0 (8.409)
3-D form of Maxwell equations: The rst 4-D Maxwell equation can be written, in 3-D, using
any of the two equivalent set of expressions

i
B
i
= 0

j
e
ij
+
B
i
t
= 0
_

_

i
b
jk
+
j
b
ki
+
k
b
ij
= 0
(
i
E
j

j
E
i
) +
b
ij
t
= 0
, (8.410)
although the standard (pseudo-vectorial) form of the equations is
div B = 0 (8.411)
rot E+
B
t
= 0 . (8.412)
The second 4-D Maxwell equation can be written, in 3-D, using any of the two equivalent set of
expressions (note: sign of the last equation not checked)

i
D
i
=

j
h
ij

D
i
t
= j
i
_

_

i
d
jk
+
j
d
ki
+
k
d
ij
=
(
i
H
j

j
H
i
)
d
ij
t
=
ijk
j
k
, (8.413)
although the standard (pseudo-vectorial) form of the equations is
div D = (8.414)
rot H
D
t
= j . (8.415)
The polarization tensor: The polarization tensor T

is dened by
T

= f

. (8.416)
Writing
_
T
00
T
0j
T
i0
T
ij
_
=
_
0 P
j
c
P
i
c m
ij
_
(8.417)
gives
P
i
+E
i
= c D
i
; m
ij
+h
ij
= c b
ij
. (8.418)
8.5. APPENDICES 177
Minkowskis energy-momentum tensor: It is dened by
T

=
1
2c
_
f

_
. (8.419)
Using the denition 8.60
_
T
0
0
T
0
j
T
i
0
T
i
j
_
=
_
c
2
q
j
c
cp
i
T
i
j
_
(8.420)
gives
=
1
2c
2
(E D+H B) (8.421)
p = DB (8.422)
q =
1
c
2
EH (8.423)
T
i
j
=
1
2
(E D+H B)
i
j

_
E
i
D
j
+H
i
B
j
_
. (8.424)
Here, represents the energy density of the electromagnetic eld. T
i
j
is the Maxwells stress tensor.
p is the linear momentum density and q is the energy ux (the Poynting vector). (Comment: names
from Mller, denitions from Minkowski). We see that, in general,
p ,= q, (8.425)
i.e., the linear momentum density is not identical to the energy ux.
Antisymmetric part of the energy-momentum tensor: We have seen that the antisymmetric
part of the energy-momentum tensor

= T

(8.426)
is the source term for the continuity equations concerning the spin. The antisymmetric tensor
ij
is equivalent to a 3-D vector and to an antisymmetric 3-D tensor.
The 3-D vector is p q . The antisymmetric 3-D tensor is

ij
= T
ij
T
ji
= (B
i
H
j
B
j
H
i
) + (D
i
E
j
D
j
E
i
) . (8.427)
Associating to
ij
the vector

i
=
1
2!

ijk

jk
(8.428)
we can write
= BH+DE . (8.429)
Using equations 8.418, i.,e.,
D =
1
c
(P+E) ; B =
1
c
(M+H) . (8.430)
gives
=
1
c
(MH+PE) , (8.431)
which is the standard expression for the couple acting on a medium magnetized and electrically
polarized.
Note: emphasize here that the nonsymmetry of the Minkowsky tensor gives this couple. It is too
bad to see most books on electromagnetism overlook this important property of the Minkowsky tensor.
There has been considerable debate in the literature, and some have proposed symmetrized versions
178 CHAPTER 8. FROM GEOMETRY TO PHYSICS
of it (see for instance Abraham, 1909; and Abraham and Becker, 1933). Brevik (1970) showed that
the Minkowskis tensor more conveniently describes some experimental results. The reader may refer
to the short discussion in Mller (1972).
Note: say that for a linear and isotropic medium,
D = E (8.432)
and
B = H, (8.433)
and the Maxwells stress is symmetric:
T
ij
= T
ji
. (8.434)
In vacuo,
=
0
(8.435)
and
=
0
, (8.436)
and, as

0
=
1
c
2
, (8.437)
the linear momentum density equals the energy ux:
p = q. (8.438)
The Lorentz density of force:

=
1
c

(8.439)
Comment: try to justify why the force density plus the divergence of the stress should be zero.
This gives:

=
1
c
_
f

+
1
4
_
f

_
_
(8.440)
The space components of the force density are

i
= E
i
+
ik
j
k
B

+
1
2
_
B
k

i
H
k
H
k

i
B
k
+D
k

i
E
k
E
k

i
D
k
.
_
(8.441)
If the medium is linear and isotropic ( D = E ; B = H ) and homogeneous (
i
= 0 ;
i
=
0 ), then the Lorentz Force density reduces to the usual expression
= E+j B. (8.442)
Comment: Say somewhere that Sedov mentions the moment stress tensor M
ij
k
for the electro-
magnetic eld, but that it does not give any expression for it.
Physical dimensions:
[q] = Q
[E] = Q
1
MLT
2
[D] = QL
2
[H] = QL
1
T
1
[B] = Q
1
MT
1
[] = Q
2
M
1
L
3
T
2
[] = Q
2
ML
[c] = LT
1
(8.443)
8.5. APPENDICES 179
Values of physical constants:
c = 299 792 458 m s
1
(8.444)
Comment: this is an exact value, the unit of length being a derived quantity.

0
= 410
7
C
2
kg m (8.445)
Comment: exact value.

0
=
1

0
c
2
= 8.854 187 817 . . . C
2
kg
1
m
3
s
2
(8.446)
Comment: exact value.
Comment: C stands for the Coulomb, the unit of charge. The Amp`ere is A = C s
1
.
Note: for the values of the physical constants, see for instance Cohen and Taylor (1991).
References: Abraham, M., 1903, Annln Phys., 10, 105.
Abraham, M., and Becker, R., 1933, Theorie der Elektrizit at, vol. 2, 6th edn., Teubner, Leipzig.
Brevik, I., 1970, Mat. Fys. Medd. Dan. Vid. Selsk. 37, no. 13.
Cohen, E.R., and Taylor, B.N., The fundamental physical constants, Physics Today, August 1991.
Minkowski, H., 1908, Nachr. Ges. Wiss. G ottingen 53.
Minkowski, H., 1910, Math. Annln 68, 472.
Mller, C., The theory of relativity, Oxford University Press, New Delhi, 1972.
Sedov, L., 1975, Mecanique des Milieux Continus, Mir, Moscou.
180 CHAPTER 8. FROM GEOMETRY TO PHYSICS
8.5.10 Appendix: Demonstration: Time derivative and space integration (Eu-
clidean spaces)
This appendix demonstrates that if we dene the vector
P(t) =
_
V (t)
dV (x) p(x, t)
=
_
V (t)
dV (x) p
i
(x, t) e
i
(x) , (8.447)
then,
dP
dt
(t) =
_
V (t)
dV (x)
_
p
i
t
(x, t) +
j
_
v
j
(x, t) p
i
(x, t)
_
_
e
i
(x) . (8.448)
Here, the notation V (t) indicates that the considered volume follows the matter in his movement
(i.e., the relative velocity of the matter with respect to the surface of the volume is zero).
As a corollary, if the condition
dP
dt
(t) = 0 (8.449)
holds for any volume inside the medium, then, at any point,
p
i
t
(x, t) +
j
_
v
j
(x, t) p
i
(x, t)
_
= 0 . (8.450)
The demonstration for a general tensor is procedes along the same lines.
Warning: the integrals here are not using the concept of parallel transport. Therefore they are
only dened in Euclidean spaces.
Demonstration:
dP
dt
(t) =
d
dt
_
V (t)
dV (x) p
i
(x, t) e
i
(x) (8.451)
= lim
t0
1
t
_
_
V (t+t)
dV (x) p
i
(x, t + t) e
i
(x)
_
V (t)
dV (x) p
i
(x, t) e
i
(x)
_
(8.452)
= lim
t0
1
t
_
_
V (t+t)
dV (x)
_
p
i
(x, t) +
p
i
t
(x, t)t
_
e
i
(x)
_
V (t)
dV (x) p
i
(x, t) e
i
(x)
_
(8.453)
= lim
t0
1
t
_
_
V (t)
dV
_
p
i
+
p
i
t
t
_
e
i
+
_
V
dV
_
p
i
+
p
i
t
t
_
e
i

_
V (t)
dV p
i
e
i
_
(8.454)
= lim
t0
1
t
_
t
_
V (t)
dV
p
i
t
e
i
+
_
V
dV
_
p
i
+
p
i
t
t
_
e
i
_
(8.455)
= lim
t0
1
t
_
t
_
V (t)
dV
p
i
t
e
i
+
_
S(t)
dS n
j
v
j
t
_
p
i
+
p
i
t
t
_
e
i
_
(8.456)
=
_
V (t)
dV
p
i
t
e
i
+
_
S(t)
dS n
j
v
j
p
i
e
i
(8.457)
=
_
V (t)
dV
p
i
t
e
i
+
_
V (t)
dV
j
_
v
j
p
i
e
i
_
(8.458)
=
_
V (t)
dV
p
i
t
e
i
+
_
V (t)
dV
j
_
v
j
p
i
_
e
i
(8.459)
=
_
V (t)
dV
_
p
i
t
+
j
_
v
j
p
i
_
_
e
i
. (8.460)
8.5. APPENDICES 181
Equation 8.451 is just the original denition. Equation 8.452 explicits the time derivative. Note that,
as the considered volume follows the matter movement, at time t + t we have to integrate over
a volume that, in general, will be dierent from V (t) . Equation 8.453 takes the rst order of a
Taylors development (higher orders are not needed). Equation 8.454 says that the volume at time
t + t will equal the volume at time t plus a certain volume change V to be evaluated below.
Equation 8.455 is just a simplication of the previous one. Equation 8.456 says that the volume change
V can be obtained by integrating on the surface S(t) the normal component of the velocity (times
t ). Equation 8.457 takes the limit t 0 , which amounts to dropping a second order term and
simplifying. Equation 8.458 uses the Greens theorem (the surface integral of a ux equals the volume
integral of the divergence). Equation 8.459 uses the formal property
i
e
j
= 0 (see section XXX).
This directly leads to equation 8.460.
8.5.11 Appendix: Demonstration: Time derivative and space integration (general
case)
This appendix demonstrates the formula
d
dt
_
V(t)
dV (x)
i
(x) p
i
(x, t) = (8.461)
=
_
V(t)
dV (x)
i
(x)
_
p
i
t
+
j
_
p
i
v
j
_
_
(x, t) +
_
V(t)
dV (x) p
i
(x, t) v
j
(x, t)
_

i
_
(x, t) .
Here, the notation 1(t) indicates that the considered volume follows the matter in his movement
(i.e., the relative velocity of the matter with respect to the surface of the volume is zero).
Demonstration:
d
dt
_
V(t)
dV (x)
i
(x) p
i
(x, t) =
= lim
t0
1
t
_
_
V(t+t)
dV (x)
i
(x) p
i
(x, t + t)
_
V(t)
dV (x)
i
(x) p
i
(x, t)
_
(8.462)
= lim
t0
1
t
_
_
V(t+t)
dV (x)
i
(x)
_
p
i
(x, t) +
p
i
t
(x, t)t
_

_
V(t)
dV (x)
i
(x) p
i
(x, t)
_
(8.463)
= lim
t0
1
t
_
_
V(t)
dV
i
_
p
i
+
p
i
t
t
_
+
_
V
dV
i
_
p
i
+
p
i
t
t
_

_
V(t)
dV
i
p
i
_
(8.464)
= lim
t0
1
t
_
t
_
V(t)
dV
i
p
i
t
+
_
V
dV
i
_
p
i
+
p
i
t
t
_
_
(8.465)
= lim
t0
1
t
_
t
_
V(t)
dV
i
p
i
t
+
_
S(t)
dS n
j
v
j
t
i
_
p
i
+
p
i
t
t
_
_
(8.466)
=
_
V(t)
dV
i
p
i
t
+
_
S(t)
dS n
j
v
j

i
p
i
(8.467)
=
_
V(t)
dV
i
p
i
t
+
_
V(t)
dV
j
_
v
j

i
p
i
_
(8.468)
=
_
V(t)
dV
i
p
i
t
+
_
V(t)
dV
i

j
_
v
j
p
i
_
+
_
V(t)
dV p
i
v
j

i
(8.469)
=
_
V(t)
dV
i
_
p
i
t
+
j
_
v
j
p
i
_
_
+
_
V(t)
dV p
i
v
j

i
. (8.470)
182 CHAPTER 8. FROM GEOMETRY TO PHYSICS
Equation 8.462 explicits the time derivative. Note that, as the considered volume follows the matter
movement, at time t +t we have to integrate over a volume that, in general, will be dierent from
V (t) . Equation 8.463 takes the rst order of a Taylors development (higher orders are not needed).
Equation 8.464 says that the volume at time t + t will equal the volume at time t plus a certain
volume change V to be evaluated below. Equation 8.465 is just a simplication of the previous
one. Equation 8.466 says that the volume change V can be obtained by integrating on the surface
S(t) the normal component of the velocity (times t ). Equation 8.467 takes the limit t 0 ,
which amounts to dropping a second order term and simplifying. Equation 8.468 uses the Greens
theorem (the surface integral of a ux equals the volume integral of the divergence). Equation 8.469
uses the formal property
i
e
j
= 0 (see section XXX). This directly leads to equation 8.470.
8.5.12 Appendix: Demonstration: Classical conservation equations
Consider a continuous medium from the viewpoint of nonrelativistic mechanics, and let, at a point, x
and time t , (x, t) denote the density of mass,
ij
(x, t) the density of intrinsic angular momentum
(spin), and v
i
(x, t) the velocity with respect a Galilean (i.e., nonaccelerated) frame. The space is
here assumed Euclidean, but the coordinates are not necessarily chosen Cartesian.
We consider a volume inside the medium that follows the matter on his movement and call it
V (t) . The surface of the volume will be denoted S(t) .
The mass inside the volume is:
M(t) =
_
V (t)
dV (x) (x, t) . (8.471)
Denoting p
i
= v
i
, the linear momentum is:
P
i
(t) =
_
V (t)
dV (x) p
i
(x, t) . (8.472)
The total force acting on that portion of matter equals the sum of the forces acting inside the
volume plus the sum of the forces acting on the surface of the volume:
F
i
(t) =
_
V (t)
dV (x) f
i
(x, t) +
_
S(t)
dS(x)
i
(x, t) . (8.473)
Here, f
i
is the force density of external origin,
i
denotes the surface traction that the exterior
medium exerts on the surface, related to the stress
i
j
through
i
=
i
j
n
j
. (Comment explain the
normal).
The total angular momentum (with respect to the origin of coordinates) is obtained as the sum of
the intrinsic angular momentum density (or spin) plus the extrinsic (or orbital) angular momentum
density:

ij
(t) =
_
V (t)
dV (x)
ij
(x, t) + (r
i
(x) p
j
(x, t) r
j
(x) p
i
(x, t)) , (8.474)
where r(x) is the position vector of the point x . As we are only interested here in at (Euclidean)
spaces the introduction of this vector poses no problem.
The total torque (with respect to the origin of coordinates) is obtained as the sum of four terms:
The result of the external torque density,
ij
, the couple resulting from the external force density f
i
,
and the two surface contributions: the result of the surface torque density that the exterior medium
exerts on the surface,
ij
, and the couple resulting from the surface traction
i
:

ij
(t) =
_
V (t)
dV (x)
ij
(x, t) + (r
i
(x) f
j
(x, t) r
j
(x) f
i
(x, t))
+
_
S(t)
dS(x)
ij
(x, t) + (r
i
(x)
j
(x, t) r
j
(x)
i
(x, t)) . (8.475)
8.5. APPENDICES 183
The surface torque density
ij
is related to the couple stress m
ij
k
through
ij
= m
ij
k
n
k
.
We are going to show that if the (global) conservation equations
dM
dt
(t) = 0 (8.476)
dP
i
dt
(t) = F
i
(t) (8.477)
d
ij
dt
(t) =
ij
(t) (8.478)
hold for any volume inside the medium, then, they are equivalent to the set of local conservation
equations

t
+
j
(v
j
) = 0 (8.479)
p
i
t
+
j
(p
i
v
j

i
j
) = f
i
(8.480)

ij
t
+
k
(
ij
v
k
m
ij
k
) =
ij
+ (
ji

ij
) + (p
i
v
j
p
j
v
i
) . (8.481)
Demonstrations
The mass: Equation 8.479 follows directly from equation 8.476 using the results of the previous
annex.
The linear momentum: We remark that from
i
=
i
j
n
j
it follows, using Greens theorem,
F
i
(t) =
_
V (t)
dV (x) f
i
(x, t) +
_
S(t)
dS(x)
i
(x, t)
=
_
V (t)
dV (x) f
i
(x, t) +
_
S(t)
dS(x) n
j
(x)
i
j
(x, t)
=
_
V (t)
dV (x) f
i
(x, t) +
_
V (t)
dV (x) (
j

i
j
)(x, t)
=
_
V (t)
dV (x)
_
f
i
(x, t) + (
j

i
j
)(x, t)
_
(8.482)
Then, equation 8.480 follows directly from the previous appendix.
The angular momentum: We have dened (equation 8.475):

ij
(t) =
(V)
ij
(t) +
(S)
ij
(t) , (8.483)
where

(V)
ij
(t) =
_
V (t)
dV (x)
ij
(x, t) + (r
i
(x) f
j
(x, t) r
j
(x) f
i
(x, t)) (8.484)
and

(S)
ij
(t) =
_
S(t)
dS(x)
ij
(x, t) + (r
i
(x)
j
(x, t) r
j
(x)
i
(x, t)) . (8.485)
184 CHAPTER 8. FROM GEOMETRY TO PHYSICS
Introducing
ij
= m
ij
k
n
k
and
i
=
i
j
n
j
, using the Greens theorem, and using
i
r
j
=
i
j
,
gives

(S)
ij
(t) =
_
S(t)
dS n
k
_
m
ij
k
+
_
r
i

j
k
r
j

i
k
__
=
_
V (t)
dV
k
_
m
ij
k
+
_
r
i

j
k
r
j

i
k
__
=
_
V (t)
dV
_

k
m
ij
k
+g
ki

j
k
+r
i

j
k
g
kj

i
k
r
j

i
k
_
=
_
V (t)
dV
_

k
m
ij
k
+ (
ji

ij
) +
_
r
i

j
k
r
j

i
k
__
. (8.486)
This gives, for the total torque,

ij
(t) =
_
V (t)
dV
_

k
m
ij
k
+ (
ji

ij
) +
ij
+
_
r
i
_

j
k
+f
j
_
r
j
_

i
k
+f
i
___
. (8.487)
Using equation 8.474 and the result of the previous appendix, we have
d
ij
dt
(t) =
_
V (t)
dV (x)
_

t
(
ij
+r
i
p
j
r
j
p
i
) +
k
_
v
k
(
ij
+r
i
p
j
r
j
p
i
)
_
_
=
_
V (t)
dV (x)
_

ij
t
+r
i
p
j
t
r
j
p
i
t
+
k
_
v
k

ij
_
+r
i

k
_
v
k
p
j
_
r
j

k
_
v
k
p
i
_
+g
ki
v
k
p
j
g
kj
v
k
p
i
_
=
_
V (t)
dV (x)
_

ij
t
+
k
_
v
k

ij
_
+r
i
_
p
j
t
+
k
_
v
k
p
j
_
_
r
j
_
p
i
t
+
k
_
v
k
p
i
_
_
+ (v
i
p
j
v
j
p
i
)
_
. (8.488)
Then, the condition that equation 8.478 must hold for any volume is equivalent to the condition that,
at any point,

ij
t
+
k
_
v
k

ij
m
ij
k
_
+r
i
_
p
j
t
+
k
_
v
k
p
j

j
k
f
j
_
_
r
j
_
p
i
t
+
k
_
v
k
p
i

i
k
f
i
_
_
=
=
ij
+ (
ji

ij
) + (p
i
v
j
p
j
v
i
) . (8.489)
If the conservation of linear momentum holds (equation 8.480), then this equation simplies to equa-
tion 8.481.
Chapter 9
Bibliography
Ander, M.E., Goldman, T., Hughes, R.J., and Nieto, M.M., 1988, Possible resolution of the Brookhaven
and Washington E otv os experiments, Phys. Rev. Lett., Vol. 60, No. 13, 12251228.
Askar, A., and Cakmak, A.S., 1968. A structural model of a micropolar continuum, Int. J. Eng. Sci.,
6, 10, 583.
Balakrishnan, A.V., 1976. Applied functional analysis, Springer-Verlag.
Belinfante, XXX, Physica, 7, 887.
Bender, C.M., and Orszag, S.A., 1978. Advanced mathematical methods for scientists and engineers,
McGraw-Hill.
Bland, D.R., 1960. The theory of linear viscoelasticity, Pergamon Press.
Born, M., and Huang, K., 1954, Dynamical theory of crystal lattices, Oxford Clarendon Press.
Brillouin, L., 1953, Wave propagation in Periodic Structures, Dover Publications.
Burnett, D.S., 1987, Finite element analysis, from concepts to applications, Addison-Wesley, Reading,
Ma, USA.
Cartan, E. Comptes Rendus Acad. Sci. 174, 437439 (1922).
Cartan, E. Comptes Rendus Acad. Sci. 174, 593595 (1922).
Cartan, E., Sur les varietes `a connexion ane et la theorie de la relativite generalisee (Premi`ere partie),
Annales Scientiques de lEcole Normale Superieure 58, 40, 325412 (1923).
Cartan, E. Sur les varietes `a connexion ane et la theorie de la relativite generalisee (Premi`ere partie,
suite), Annales Scientiques de lEcole Normale Superieure, 59, 41, 125 (1924).
Cartan, E., Sur les varietes `a connexion ane et la theorie de la relativite generalisee (Deuxi`eme
partie), Annales Scientiques de lEcole Normale Superieure, 60, 42, 1788 (1925).
Choquet-Bruhat, Y., DeWitt-Morette, C., Dillard-Bleick, M., 1982, Analysis, Manifolds and Physics,
North-Holland Publishing Co.
Cosserat, E., and Cosserat, F., 1896. Sur la theorie de lelasticite, Ann. de l

Ecole Normale de
Toulouse, 10, 1, 1.
Cosserat, E., and Cosserat, F., 1907. Sur la mecanique generale, C. Rend. hebd. des seances de
lAcad. des Sci., 145, 1139.
Cosserat, E., and Cosserat, F., 1909. Theorie des corps deformables, A. Herman et Fils, Paris.
Courant, R., and Hilbert, D., 1953. Methods of mathematical physics, Interscience Publishers.
Dautray, R., and Lions, J.L., 1984 and 1985. Analyse mathematique et calcul numerique pour les
sciences et les techniques (3 volumes), Masson, Paris.
de Broglie, L., 1952, La theorie des particules de spin 1/2, Gauthier Villars, Paris.
de Broglie, L., 1954, Theorie generale des particules `a spin, Gauthier-Villars, Paris.
Eckhardt, D. XXXXX XX, XXXXXX (XXXX).
Einstein. . .
Halbwachs, F., 1960, Theorie relativiste des uides `a spin, Gauthier Villars, Paris.
185
186 CHAPTER 9. BIBLIOGRAPHY
Hehl, F.W., 1973, Spin and torsion in general relativity: I. Foundations, General relativity and gravi-
tation, Vol. 4, No. 4, 333-349.
Hehl, F.W., 1974, Spin and torsion in general relativity: II. Geometry and eld equations, General
relativity and gravitation, Vol. 5, No. 5, 491-516.
Hehl, F.W., 1980, Four lectures on Poincare Gauge Field Theory, in: Cosmology and Gravitation,
edited by P.G. Bergmann and V. De Sabbata, Plenum Press, New York.
Hehl, F.W. Foundations of Physics 15, 451471 (1985).
Hehl, F.W., 1985, On the kinematics of the torsion of space-time, Foundations of Physics, Vol. 15,
No. 4, 451-471.
Hehl, F.W., and McCrea, J.D., 1986, Bianchi identities and the automatic conservation of energy-
momentum and angular momentum in general-relativistic eld theories, Foundations of Physics,
Vol. 16, No. 3, 267293.
Houard, J.C., 1981, Sur la description intrins`eque des grandeurs dimensionnelles, Ann. Inst. Henri
Poincare, Vol. XXXV, n0 3, 225-252.
Jobert, G., 1975, Propagator and Green matrices for body force and dislocation. Geophys. J. Roy.
astr. Soc., 42, 755-762.
Kaliski, S., 1963. On a model of a continuum with an essentially non-symmetric tensor of mechanical
stress, Arch. Mech. Stos., 15, 1, 33.
Kerlick, G.D. Phys. Rev. 12, 3005 (1975).
Kibble, T.W.B., 1961, Lorentz invariance and the gravitational eld, J. Math. Phys. 2, 212221.
Kittel, C., 1976, Introduction to solid state physics, Wiley.
Levi-Civita, T., 1926, The absolute dierential calculus, Blackie & Son, Glasgow. (republished by
Dover, New York, in 1977).
Lichnerowicz, A., 1960. Elements de Calcul Tensoriel, Armand Collin, Paris.
Mase, G.E., 1970. Theory and problems of continuum mechanics, McGraw-Hill.
Milne, R.D., 1980. Applied functional Analysis, Pitman Advanced Publishing Program, Boston.
Misner, Ch.W., Thorne, K.S., and Wheeler, J.A., 1973. Gravitation, Freeman.
Morse, P.M., and Feshbach, H., 1953. Methods of theoretical physics, McGraw Hill.
Nottale, L., 1993, Fractal space-time and microphysics; towards a theory of scale relativity, World
scientic, [note: which city?].
Nowacki, W., 1986. Theory of asymmetric elasticity, Pergamon Press.
OConnell, R.F. Phys. Rev. D 16, 12471247 (1977).
Ray, J.R., and Smalley, L.L., 1982, Improved perfect-uid energy-momentum tensor with spin in
Einstein-Cartan space-time, Phys. Rev. Lett., Vol. 49, No.15, 10591061.
Ray, J.R., and Smalley, L.L., 1982, Spinning uids in the Einstein-Cartan theory, Physical Review D,
Vol 27, No. 6, 13831385.
Rosenfeld, XXX, 1940, Acad, Roy. Belgique, 18, 6.
Runcorn, XXX. XXXXX XX, XXXXXX (XXXX).
Schwartz, L., 1965. Methodes mathematiques pour les sciences physiques, Hermann, Paris.
Schwartz, L., 1966. Theorie des distributions, Hermann, Paris.
Schwartz, L., 1970. Analyse (topologie generale et analyse fontionelle), Hermann, Paris.
Sedov, L., 1975, Mecanique des milieux continus, Mir, Moscou.
Stacey, F.D., Tuck, G.J., Moore, G.I., Holding, S.C., Goodwin, B.D. & Zhou, R., 1987, Geophysics
and the law of gravity, Rev. Mod. Phys. 59, No. 1, 157174.
Stacey, F.D., Tuck, G.J. & Moore, G.I. J. Geophys. Res. 93, 10 57510 587 (1988).
Stern, XXX., & Gerlach, XXX. XXXXX XX, XXXXXX (1922).
Tarantola, A., 1987. Inverse problem theory; methods for data tting and model parameter estimation,
Elsevier.
Tarantola, A., 2003 (?). Elements for Physics (in preparation).
187
Taylor, A.E., and Lay, D.C., 1980. Introduction to functional analysis, Wiley.
Thirring, W., 1979, A course in mathematical physics, Springer-Verlag, New-York.
Trautman, A. in Symposia Mathematica, Vol XII, (Academic Press, London & New York, 1973).
Trautman, A., Pirani, F.A.E., and Bondi, H., 1965, Lectures on general relativity, Vol. I, Prentice-Hall
Inc.
Truesdell, C., 1980, The tragicomical history of thermodynamics 1822-1854, Springer-Verlag.
Truesdell, C., and Toupin, R.A., 1960. The classical eld theories, Encyclopedia of physics, vol. III/1,
Springer Verlag, Berlin.
Von Westenholz, C., 1981, Dierential forms in mathematical physics, North-Holland publishing com-
pany, Amsterdam.
Weinberg, S., 1972, Gravitation and cosmology, John Wiley & Sons.
Ziman, J.M., 1965, Principles of the theory of solids, Cambridge University Press.