Tensor Calculus and

­Applications
 Mathematics and Its Applications:
Modelling, Engineering, and Social
Sciences
Series Editor:
Hemen Dutta

Discrete Mathematical Structures: A Succinct Foundation

Beri Venkatachalapathy Senthil Kumar and Hemen Dutta
Concise Introduction to Logic and Set Theory
Iqbal H. Jebril and Hemen Dutta
Tensor Calculus and Applications: Simplified Tools and Techniques
Bhaben Chandra Kalita

For more information on this series, please visit: www.crcpress.com/

Mathematics-and-its-applications/book-series/MES
Tensor Calculus and
­Applications
Simplified Tools and Techniques

Bhaben Chandra Kalita

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2019 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-0-367-13806-6 (Hardback)

This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors
and publishers have attempted to trace the copyright holders of all material reproduced in this
publication and apologize to copyright holders if permission to publish in this form has not been
obtained. If any copyright material has not been acknowledged, please write and let us know so we
may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.­
copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC),
222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that
provides licenses and registration for a variety of users. For organizations that have been granted a
photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and
are used only for identification and explanation without intent to infringe.

Library of Congress Cataloging‑in‑Publication Data

Names: Kalita, Bharat Chandra, 1937- author.

Title: Tensor calculus and applications : simplified tools and techniques / authored
by Bhaben Chandra Kalita.
Description: Boca Raton : Taylor & Francis, [2019] | Includes bibliographical
references.
Identifiers: LCCN 2018052527 | ISBN 9780367138066 (hardback :acid-free paper) |
ISBN 9780429028670 (e-book)
Subjects: LCSH: Calculus of tensors. | Geometry, Differential.
Classification: LCC QA433 .K345 2019 | DDC 515/.63—dc23
LC record available at https://lccn.loc.gov/2018052527

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com
Contents

Preface.......................................................................................................................ix
About the Book........................................................................................................xi
Author.................................................................................................................... xiii

Part I  Formalism of Tensor Calculus

1. Prerequisites for Tensors...............................................................................3

1.1 Ideas of Coordinate Systems................................................................3
1.2 Curvilinear Coordinates and Contravariant and Covariant
Components of a Vector (the Entity)��������������������������������������������������3
1.3 Quadratic Forms, Properties, and Classifications.............................7
1.4 Quadratic Differential Forms and Metric of a Space in the
Form of Quadratic Differentials���������������������������������������������������������9
Exercises........................................................................................................... 11

2. Concept of Tensors........................................................................................ 13
2.1 Some Useful Definitions..................................................................... 13
2.2 Transformation of Coordinates.......................................................... 14
2.3 Second and Higher Order Tensors.................................................... 17
2.4 Operations on Tensors......................................................................... 18
2.5 Symmetric and Antisymmetric (or Skew-Symmetric) Tensors..... 20
2.6 Quotient Law........................................................................................ 25
Exercises........................................................................................................... 26

3. Riemannian Metric and Fundamental Tensors...................................... 29

3.1 Riemannian Metric.............................................................................. 29
3.2 Cartesian Coordinate System and Orthogonal Coordinate
System��������������������������������������������������������������������������������������������������� 29
3.3 Euclidean Space of n Dimensions, Euclidean Co-Ordinates,
and Euclidean Geometry�������������������������������������������������������������������30
3.4 The Metric Functions gij Are Second-Order Covariant
Symmetric Tensors������������������������������������������������������������������������������30
3.5 The Function gij Is a Contravariant Second-Order
Symmetric Tensor������������������������������������������������������������������������������� 33
3.6 Scalar Product and Magnitude of Vectors........................................ 38
3.7 Angle Between Two Vectors and Orthogonal Condition............... 38
Exercises........................................................................................................... 39

v
vi Contents

4. Christoffel Three-Index Symbols (Brackets) and Covariant

Differentiation............................................................................................... 41
4.1 Christoffel Symbols (or Brackets) of the First and
Second Kinds����������������������������������������������������������������������������������� 41
4.2 Two Standard Applicable Results of Christoffel Symbols.............42
4.3 Evolutionary Basis of Christoffel Symbols (Brackets).....................43
4.4 Use of Symmetry Condition for the Ultimate Result...................... 49
4.5 Coordinate Transformations of Christoffel Symbols...................... 50
4.5.1 Transformation of the First Kind ij , k .................................. 50
i
4.5.2 T  ransformation of the Second Kind jk............................... 51
4.6 Covariant Derivative of Covariant Tensor of Rank One................ 55
4.7 Covariant Derivative of Contravariant Tensor of Rank One......... 56
4.8 Covariant Derivative of Covariant Tensor of Rank Two................ 57
4.9 Covariant Derivative of Contravariant Tensor of Rank Two......... 59
4.10 Covariant Derivative of Mixed Tensor of Rank Two...................... 60
4.10.1 Generalization......................................................................... 61
4.11 Covariant Derivatives of g ij ′ g ij and also g ij........................................ 62
4.12 Covariant Differentiations of Sum (or Difference) and
Product of Tensors������������������������������������������������������������������������������63
4.13 Gradient of an Invariant Function..................................................... 66
4.14 Curl of a Vector..................................................................................... 67
4.15 Divergence of a Vector......................................................................... 68
4.16 Laplacian of a Scalar Invariant........................................................... 69

4.17 Intrinsic Derivative or Derived Vector of v...................................... 71
4.18 Definition: Parallel Displacement of Vectors................................... 72
4.18.1 When Magnitude Is Constant............................................... 72
4.18.2 Parallel Displacement When a Vector Is of Variable
Magnitude������������������������������������������������������������������������������� 73
Exercises........................................................................................................... 76

5. Properties of Curves in Vn and Geodesics................................................77

5.1 The First Curvature of a Curve..........................................................77
5.2 Geodesics............................................................................................... 78
5.3 Derivation of Differential Equations of Geodesics......................... 78
5.4 Aliter: Differential Equations of Geodesics as Stationary
Length���������������������������������������������������������������������������������������������������80
5.5 Geodesic Is an Autoparallel Curve.................................................... 82
5.6 Integral Curve of Geodesic Equations..............................................85
5.7 Riemannian and Geodesic Coordinates, and Conditions for
Riemannian and Geodesic Coordinates������������������������������������������ 86
5.7.1 Another Form of Condition for Geodesic Coordinates....... 88
5.8 If a Curve Is a Geodesic of a Space (Vm), It Is also a Geodesic
of Any Space Vn in Which It Lies (Vn a Subspace)�������������������������� 89
Exercises........................................................................................................... 91
Contents vii

6. Riemann Symbols (Curvature Tensors).................................................... 93

6.1 Introduction.......................................................................................... 93
6.2 Riemannian Tensors (Curvature Tensors)........................................ 93
6.3 Derivation of the Transformation Law of Riemannian
α
Tensor Rabc ��������������������������������������������������������������������������������������������� 95
α
6.4 Properties of the Curvature Tensor Rijk ............................................. 97
6.5 Covariant Curvature Tensor............................................................... 99
6.6 Properties of the Curvature Tensor Rhijk of the First Kind........... 100
6.7 Bianchi Identity.................................................................................. 101
6.8 Einstein Tensor Is Divergence Free................................................. 102
6.9 Isometric Surfaces.............................................................................. 103
6.10 Three-Dimensional Orthogonal Cartesian Coordinate
Metric and Two-Dimensional Curvilinear Coordinate
Surface Metric Imbedded in It�������������������������������������������������������� 103
6.11 Gaussian Curvature of the Surface S immersed in E3.................. 104
Exercises......................................................................................................... 108

Part II  Application of Tensors

7. Application of Tensors in General Theory of Relativity.................... 113

7.1 Introduction........................................................................................ 113
7.2 Curvature of a Riemannian Space................................................... 114
7.3 Flat Space and Condition for Flat Space......................................... 118
7.4 Covariant Differential of a Vector................................................... 118
7.5 Motion of Free Particle in a Curvilinear Co-Ordinate
System for Curved Space����������������������������������������������������������������� 119
7.6 Necessity of Ricci Tensor in Einstein’s Gravitational Field
Equation��������������������������������������������������������������������������������������������� 120

8. Tensors in Continuum Mechanics........................................................... 123

8.1 Continuum Concept.......................................................................... 123
8.2 Mathematical Tools Required for Continuum Mechanics........... 123
8.3 Stress at a Point and the Stress Tensor............................................ 125
8.4 Deformation and Displacement Gradients.................................... 126
8.5 Deformation Tensors and Finite Strain Tensors............................ 127
8.6 Linear Rotation Tensor and Rotation Vector in Relation to
Relative Displacement���������������������������������������������������������������������� 130

9. Tensors in Geology..................................................................................... 133

9.1 Introduction........................................................................................ 133
9.2 Equation for the Determination of Shearing Stresses
on Any Plane Surface����������������������������������������������������������������������� 135
viii Contents

9.3  eneral Transformation and Maximum and Minimum

G
Longitudinal Strains������������������������������������������������������������������������� 137
9.4 Determination of the Two Principal Strains in a Plane................ 140

10. Tensors in Fluid Dynamics....................................................................... 143

10.1 Introduction........................................................................................ 143
10.2 Equations of Motion for Newtonian Fluid..................................... 143
10.3 Navier–Stokes Equations for the Motion of Viscous Fluids........ 145

Appendix.............................................................................................................. 149
Remarks................................................................................................................ 153
Bibliography......................................................................................................... 155
Index...................................................................................................................... 157
Preface

There is a great demand from students for a book on tensors with simple
and conceivable presentations. The theoretical development of the subject
“tensor calculus” is critical to be understood by students because of its com-
plex nature of uses of the subscripts and superscripts. The simplification in
the working process with repeated/nonrepeated indices of a mixed tensor
makes it rather more complex for readers if some special clues are not men-
tioned. Moreover, the fields of application, namely, non-isotropic media and
the exact situation of deformation of bodies, cannot be identified easily in
the true sense. Through concrete citation of applicable media and physical
bodies, the subject can be made conceivable. For example, in elastic media
such as motion of viscous fluids or dirty water, the application of tensors
is inevitable. Research on viscous media use of Navier–Stokes equation
governed by tensors is one of the primary prerequisites. The investigation
causing deformation with elasticity in physics needs application of tensors.
However, a clear concept to make use of the different classes of tensors in
such fields is of paramount importance. Only then correct results of investi-
gation can be unearthed. Eventually, calculus of tensors can be considered as
the most appropriate tool to know the physical field theories. Hence, applied
mathematicians, physicists, engineering scientists, and geologists cannot
excel without the knowledge of tensors. Emphasis is given primarily on the
subject, and only to motivate the readers, some physical fields are described
for real interest.
During the past 37 years of teaching this subject at the MSc level, I could
clearly read the minds of students on why they found tensor calculus dif-
ficult to understand. For the skillful teaching arts adopted in the class lec-
tures, hundreds of students requested that I write a book on tensors for
the benefit of students. Students’ feedback and suggestions from many col-
leagues inspired me to undertake this venture of writing this book. With
the teaching arts based on some individual special techniques of changing
the indices of tensors, I attempted to place a book on tensors in the hands of
students. The complexity that arises out of the use of shorthand notations
was removed so that readers can easily understand them. If the students
­capitalize the techniques provided in the book, in addition to its elegant
­presentation and simple language, my belief will become a reality.
I will be highly obliged if this book renders at least some ­service to students
and researchers so that they can understand its ­tremendous ­importance in
research gates such as relativity, physics, continuum ­mechanics, and geology.
Overall, this book is designed to cater to the needs of students of mathemat-
ics, physics, engineering, and geology from all universities. Further, I will
acknowledge the readers with sincere gratitude if they point out mistakes in

ix
x Preface

the formative stage of the book on tensors, which is very difficult to publish
without any printing ­mistakes.

Dr. Bhaben Chandra Kalita

Professor Emeritus
Department of Mathematics
Gauhati University
Guwahati-781014
India
About the Book

The book Tensor Calculus and Applications is not elementary in nature; rather,
it is physically motivated in the sense of application. Theoretically, the subject
“tensor calculus” is critical for students to understand the ­complex nature of
using subscripts and superscripts. Besides the lack of knowledge about the
fields of application in non-isotropic media and of identifying deformation
situations of bodies poses rather more difficulty to earn the concepts. The
elegant nature of description of the theory with specific style of changing
suffixes and prefixes and reasons to recover meaningful results of the subse-
quent fields can only make the subject easier. With this objective in mind, the
author was inspired to write the book for the benefit of readers. In the opinion
of the author, the old books written by L. P. Eisenhart and C. E. Weatherburn
could not serve this purpose though they are of ­fundamental nature from a
theoretical standpoint but not available in the market and ­conceivable at the
same time. The experience derived from teaching the ­subject for more than
37 years to the postgraduate students and the psychology gathered from the
feedback of students are the intense feeling of the author to write a book on
tensors using special techniques.
The techniques adopted in the book with directions will definitely encour-
age the students to read the book to develop concepts and make use of them
in appropriate geometrical fields and space. For example, the curvature of
space (a geometrical entity) is the manifestation of gravity, and hence, ­tensor
calculus becomes the fundamental tool as discovered by Einstein for ­general
theory of relativity. The book is designed to discuss the fundamental ingre-
dients such as Riemannian tensors, which are essential to enter into the
threshold of research in general theory of relativity. Tensor being an intrinsic
concept independent of any referential systems, different from Newtonian
mechanics, is the essence of invariance for physical laws. Besides, in non-
isotropic media such as viscous fluids, elastic media, deformation of bodies
similar to structural geology, uses of tensors are essential ingredients. To
amplify the uses of tensors in these fields, some relevant ideas are included
in the book. In this context, the preface is written to manifest its suitability
and necessary background why the author has written the book.
The book consists of 10 chapters. Chapter 1 is devoted to giving some
­prerequisites of the subjects. Chapter 2 deals with the fundamental concepts
of quadratic forms and their properties. Chapter 3 discusses the essential
concept of generating space of any dimensions and corresponding geom-
etry like the Riemannian metric inherent in fundamental tensors. Chapter 4
is devoted to developing the subject with the use of shorthand symbols
called Christoffel symbols and the important tensorial operation covariant
differentiations theoretically. Chapter 5 includes the geometrical concept

xi
xii About the Book

“geodesics” primarily required for dynamical scenario. Chapter 6 discusses

the curvature tensors or Riemannian tensors, the fundamental ingredients of
general theory of relativity along with properties. Chapters 7–10 narrate the
applications of tensors in general theory of relativity, continuum mechanics,
geology, and fluid dynamics, respectively for students.
Author

Dr. Bhaben Chandra Kalita has been first class throughout his career. He
has served 37 years in the Department of Mathematics, Gauhati University,
in capacity of assistant and associate professors and professor and head of
the department since 1978. Dr. Kalita was granted the prestigious award
“Professor Emeritus” by the University Grants Commission, Government of
India on September 2015. He has published more than 50 papers in Physics
of Fluids, Physics of Plasmas, Astrophysics and Space Science, Journal of Plasma
Physics, Physical Society of Japan, Canadian Journal of Physics, IEEE Transaction on
Plasma science, Communication in Theoretical Physics, and Plasma Physics Reports,
besides some other papers of relativity and graph theory. He has acted as an
invited speaker on astrophysics and particle physics in Dallas (2016) and San
Antonio (2017). He has presented papers in Granada (Spain), Pissa (Italy), and
Swansea (UK), and acted as a speaker in many local ­universities and insti-
tutions. He has also authored several textbooks on advanced ­mathematics
and reference books of higher secondary level. Recently, he has served as
‘Keynote Speaker’ in Astrophysics and Particle Physics conference (2018) held
in Chicago, Illinois, USA.

xiii
 Part I

Formalism of Tensor
Calculus
1
Prerequisites for Tensors

1.1 Ideas of Coordinate Systems

Geometric ideas and entities can be well defined in various forms with
­reference to coordinate systems. In most of the cases for convenience, rect-
angular coordinate system is taken into account, but it is not applicable in
all fields of physical system. Deviating from it, we may think of curvilin-
ear coordinate system in the lowest level. To consider the geometry of space
for dynamical scenario as “dynamics deals with the geometry of motion,”
coordinate system should be suitably selected. For an essential entity in this
sense, the author is tempted to give a brief idea of curvilinear coordinate
system.

1.2 Curvilinear Coordinates and Contravariant and

Covariant Components of a Vector (the Entity)
Consider the rectangular Cartesian coordinates (x, y, z) of any point P with

position vector r . For the coordinates (x, y, z) of P, a correspondence can be
made with (u1, u2, u3) as

x = x(u1 , u2 , u3 ), y = y(u1 , u2 , u3 ), z = z(u1 , u2 , u3 ).

If these functions are single valued and have continuous partial derivatives,
they can be solved as

u1 = u1  ( x , y , z ) , u2 = u2  ( x , y , z ) , u3 = u3 ( x , y , z ) .

Here, u1, u2, u3 are called the curvilinear coordinates of the point P.
Consequently, the position vector r = ix + jy + kz can be expressed as r =
r (u1, u2, u3).

3
4 Tensor Calculus and Applications

^t2
u2 n^2
n^3
^t1

^t3
P
n^1 u1
u3

FIGURE 1.1
Vector reprersentations in Curvilinear system.

∴ A unit tangent vector tˆ1 to the curve u1 (i.e., the curve of intersection of
u2 = c2, u3 = c3) at P (Figure 1.1) is
∂r
∂ u1
tˆ1 = .
∂r
∂u1

Similarly, the unit tangent vectors tˆ2 and tˆ3 along u2 and u3 curves, respec-
tively, can be written as
∂r ∂r
∂u2 ∂u3
tˆ2 = and tˆ3 =
∂r ∂r
∂u2 ∂u3

Again the normal vectors to the surfaces u1 = c1 , u2 = c2 , u3 = c3 are given

by the vectors ∇u1 , ∇u2 , ∇u3 , respectively. Hence, the unit normal vectors
nˆ 1 , nˆ 2 , nˆ 3 to the direction of the vectors are
∇u1 ∇u2 ∇u3
nˆ 1 = , nˆ 2 = , nˆ 3 = .
∇ u1 ∇ u2 ∇ u3

Thus, at each point P of a curvilinear coordinate system, there exist two sets
of unit vectors: (i) tˆ1 , tˆ2 , tˆ3 tangent to the coordinate curves and (ii) nˆ 1 , nˆ 2 , nˆ 3
normal to the coordinate surfaces. Of course, the two sets become identical
if and only if the curvilinear coordinate system is orthogonal. In this case,
the sets are similar to i, j, k of rectangular coordinate system but differ in the
sense of changing directions  from point to point.
Eventually, any vector A can be expressed in terms of the base vectors t̂i
and n̂ j (i, j = 1, 2, 3) as

A = A1tˆ1 + A2 tˆ2 + A3tˆ3
    ∂ r (1.2.1)
= a1 α 1 + a2 α 2 + a3 α 3 , where α j = .
∂u j
Prerequisites for Tensors 5

X2
S

N P
A

X1
O M R

FIGURE 1.2
Vector representations in oblique Cartesian system.

These are not necessarily unit tangent base vectors.

Also,

A = B1nˆ 1 + B2 nˆ 2 + B3 nˆ 3
    (1.2.2)
= b1 β 1 + b 2 β 2 + b3 β 3 , where β i = ∇ui .

These are not unit normal vectors.

The quantities a1 , a2 , a3 are called the contravariant components,
 and
b1 , b2 , b3 are called the covariant components of the same vector A .
Thus, based on the representative “base vectors,” a vector can have
­contravariant or covariant components.
To illustrate this, let us consider the oblique Cartesian coordinate lines X1
and X2 (not rectangular) in two dimensions in a plane.  
Let us consider the components OM and ON of any vector A = OP
­measured parallel to the coordinate lines  OX and OX . They are called the
1 2

contravariant components of the vector A (Figure 1.2).

Consider the perpendicular projections PR on OX1 and PSon  OX . OR and
2

OS are called the covariant components of the same vector A . Obviously, if

the coordinate lines are perpendicular, then there will not be any distinction
between the contravariant and covariant components.
Depending upon the parallel and perpendicular projections from the
point P upon the coordinate axes, we assume the coordinates of P as (x1, x2)
and (x1, x2), respectively.
The contravariant components xi (i = 1, 2) of the vector A and its covariant
components xi (i = 1, 2) are connected by the relations

x1 = x 1 + x 2 cos α

x2 = x 2 + x 1 cos α

where α is the inclination between the coordinate axes OX1 and OX2.

 x1   1 cos α   x 1 
∴  =  2  (1.2.3)
 x 2   cos α 1  x 
6 Tensor Calculus and Applications

 
If e i and e j are the unit vectors stipulated along the coordinate axes, then
1−   
A = x e1 + x 2 − e2 = x i ei so that the length OP is

(x )
2 1− 2
A = A⋅A = e1 + x 2 − e2

     
( ) ( )
2 2
= x1 e1 ⋅ e1 + 2 x 1 x 2 e1 ⋅ e 2 + x 2 e2 ⋅ e2  .

= g ij x i x j

Defining

   1 cos α 
g ij = ei ⋅ e j =   = g ji
 cos α 1

∴ g ij = sin 2 α .

1  1 − cos α 
∴ The inverse of g ij =   .
sin 2 α  − cos α 1 
1  1 − cos α 
∴ If we denote the inverse of g ij as g ij , then g ij =   .
sin 2 α  − cos α 1
∴ Equation (1.2.3) can be written as xi = g ij x j
Otherwise, x i = g ij x j
 1 0 
so that g ij g jk =  , and hence
 0 1 

2
A = g ij x i x j

= xi x i

= x i g ij x j

= g ij x i x j .

Definition

Basis: A non-empty subset S = {α 1 , α 2 ,  , α n } of a vector space V(F) is said

to be its basis if

i. S is linearly independent of V. 
S generates V, i.e., if every vector α ∈V is expressible in terms of
ii.
the basis set {α i }. But S is not unique for if S = {α 1 , α 2 , , α n } is the
Prerequisites for Tensors 7

basis, then {cα 1 , α 2 , … , α n } is also the basis. For classical example,


any vector d in three dimensions can be expressed in terms of three
   
noncoplanar vectors a, b , c as d = λ a + µb + υ c.

Symbolically, if b( i ) are three noncoplanar
 vectors, they can be taken as the

basis of vector a so that a = ai b( i ) , where ai ’s are the components.

Definition

Coordinate basis: The coordinates of a point in n-dimensional space are

identified with reference to a set of axes, and every point xi (i = 1, 2,…, n) can
be correlated with the position vector r (= xi eˆi ) = x1eˆ1 + x2 eˆ2 +  + xn eˆn ) where
each term represents the displacement (vector) in the direction of the respec-
tive axis. The set {eˆi } is called the coordinate basis. For example, in the three-
 ˆ ˆ ˆ is the position vector of the
dimensional Euclidean space, E3 , r = ix + jy + kz
point (x, y, z) with the coordinate
  bases î , ĵ, and k̂.
Orthogonality: Let α and β be any two vectors in an inner product  (or

dot product)
 space. The vector α is said to be orthogonal  to  the  vector β

   
if α ⋅ β (=<  α , β  >) = 0.
 If for any two vectors α i , α j (i ≠ j) of a set of vec-
   
tors S = {α 1 , α 2 ,… , α n } , α i ⋅ α j = 0(= < α i , α j >), then the set is called the
orthogonal set. 
Orthonormal set: A set S = {αˆ 1 , αˆ 2 ,  , αˆ n } of vectors V ( F ) is said to be
orthonormal if
   
α i ⋅ α j = 0 (= < α i , α j >), if i ≠ j

= 1, if i = j .

Norm: For any inner product space V, the norm (or length or magnitude)
   
of any vector α ∈V is defined by α = < α , α > = α ⋅ α α ≠ 0, and it is a
­nonnegative value. Multiplying each of the vectors of S by the reciprocal of
its norm or length, S can be transformed to an orthonormal set.
N.B.: The coordinate bases may not be orthogonal or orthonormal.

1.3 Quadratic Forms, Properties, and Classifications

Definition

A homogeneous second-degree polynomial in n variables (in general)

x 1 , x 2 ,  , x n is called a quadratic form and is expressed as

ai j x i x j , (1.3.1)
8 Tensor Calculus and Applications

with double sum i, j = 1, 2,  , n.

The form is said to be real if all the coefficients ai j are real, and it is called
nonsingular if the corresponding determinant

a11 a12  a1n

aij = a21 a22  a2 n ≠ 0,

an1 an2  ann

Otherwise, it is called singular.

Further, the rank of the square matrix ( ai j ) represents the rank of the
­quadratic form.
Again with the help of a nonsingular linear transformation x i = α ij y j ,
the quadratic form can be reduced to the form cij y i y j (i, j = 1, 2,  , n) in terms
of the variables y i subject to the well-established result: rank of the matrix
( aij ) = rank of the matrix (cij ).

Theorem

If r is the rank of a real quadratic form (cij ) y i y j , there exists a nonsingular

linear transformation of the variables which can reduce it to the form cr ( x r )2
where none of cr ’s is zero.
The quadratic form can be expressed as

c1 ( x 1 )2 + c2 ( x 2 )2 +  + c r ( x r )2 . (1.3.2)

Here, cr can be a positive or negative nonzero constant.

Signature: The difference between the number of positive and negative
coefficients or the excess of positive coefficients over the negative coefficients
of (1.3.2) is called the signature of the real quadratic forms.
Of course, according to the “Sylvester’s law of inertia,” the number of
­positive coefficients is invariant (remains unchanged).
Normal form: If the real quadratic form cr ( x r )2 is transformable by means
of a nonsingular linear transformation to a form in which the coefficients cr ’s
take up values +1 and −1, it is called the normal form of it.
If p is the number of positive coefficients (+1), q is the number of nega-
tive  coefficients (−1), and S is the signature, then the normal form can be
written as

(x ) + (x ) ( ) − (x ) − (x ) ( )
1 2 2 2 2 p+1 2 p+2 2 2
+ + xp −  − xr , (1.3.3)

1
where p + q = r and p − q = s, so that p = (r + s).
2
Prerequisites for Tensors 9

Definite and indefinite quadratic forms: If all the signs of the quadratic
normal form (1.3.3) are the same (positive or negative) or different, then the
quadratic form is called definite or indefinite.
Positive definite and negative definite quadratic forms: If all the signs
of the normal quadratic form (1.3.3) are positive, it is called positive definite,
and if all of them are negative, it is called negative definite.

1.4 Quadratic Differential Forms and Metric of a

Space in the Form of Quadratic Differentials
Definition

A second-degree homogeneous polynomial of the differentials dxi of the vari-

ables x 1 , x 2 ,  , x n is called a quadratic differential form, e.g., Q( a) = aij dx i dx j
is a quadratic differential form where aij’s are the functions of x i’s or may be
constant. A quadratic differential form is of paramount importance, which
will be demonstrated in the next few chapters.
Considering n independent functions y i = y i ( x 1 , x 2 ,  , x n ) instead of the
variables xi, we can get a homogeneous linear transformation of differentials:

∂y i j
dy i = dx . (1.4.1)
∂x j

∂xi
This can also be written as dx i = dy j, since for nonsingular linear trans-
∂y j
∂y i
formation of quadratic forms of the same rank ≠ 0.
∂x j

∴ The quadratic differential form aij dx i dx j can be transformed to

(1.4.2)
Q(b) = bij dy i dy j .

Noticeably at a given point or for a given x i , the relation (1.4.1) stands for a
linear transformation of differentials with constant coefficient. Hence, it is
analogous to quadratic forms discussed in Section 1.3. Contextually, (1.4.2)
can also be reduced to the similar normal form:

( dy ) + ( dy ) ( ) − ( dy ) − ( dy ) ( )
1 2 2 2 2 p+1 2 p+2 2 2
+  + dy p −  − dy r , (1.4.3)

where r is the rank of (bij).

The notions called positive definite, negative definite, and signature of
the normal form of this quadratic differential are exactly similar to those as
­discussed in Section 1.3.
10 Tensor Calculus and Applications

For simplicity, let us consider the position vector r of a point P (as in

Figure  1.1) in curvilinear coordinates u1 , u2 , u3 as r = r (u1 , u2 , u3 ) in three
dimensions (Figure 1.2).

∂r ∂r ∂r
∴ dr = du1 + du2 + du3.
∂u1 ∂u2 ∂u3

If ds is the element (length) between the adjacent points P( r ) and Q( r + dr ),

then (in the limit)

dr ⋅ dr = ds2

 ∂r ∂r ∂r   ∂r ∂r ∂r 
= du1 + du2 + du3  .  du1 + du2 + du3 
 ∂u1 ∂u2 ∂u3   ∂u1 ∂u2 ∂u3 
2
 ∂r   ∂r ∂r   ∂r ∂r 
= du21 + 2  . du1 du2 + 2  . du1 du3
 ∂u1   ∂u1 ∂u2   ∂u1 ∂u3 
2 2
 ∂r   ∂r ∂r   ∂r 
+  du22 + 2  .  du2 du3 +  du32
 ∂u2   ∂u2 ∂u3   ∂u3 

= a11 du12 + 2 a12 du1 du2 + 2 a13 du1 du3 + a22 du22 + 2 a23 du2 du3 + a33 du32
3 3
2
∴ ds = ∑ ∑a
i=1 j=1
ij dui du j ,

where

∂r ∂r
aij = . ≠0
∂ui ∂u j

and

ds2 = aij dui du j (1.4.4)

by dropping summation sign for summation convention (defined in Section

2.1).
This is the quadratic differential form representing the elementary distance
between two adjacent points in the oblique curvilinear coordinate system.
This is called the metric or line element of the system. Of course, in oblique
curvilinear coordinates (u, v , w), (1.4.4) can be written in explicit form as

ds2 = adu2 + bdv 2 + cdw 2 + 2 h dudv + 2 gdudw + 2 fdvdw.

Prerequisites for Tensors 11

But in rectangular Cartesian coordinates of Euclidean space of three dimen-

sions, the elementary distance between the adjacent points ( x , y , z) and
( x + dx , y + dy , z + dz) is given by

ds2 = dx 2 + dy 2 + dz 2,

where

∂r ∂r
a ij = . = 0 for i ≠ j
∂ui ∂u j
= 1 for i = j

u1 = x , u2 = y , u3 = z .

To include almost all physical spaces, Riemann has generalized this concept
(notion) to n dimensions. This will be discussed in detail in Chapter 3.

Exercises
1. If the quadratic form aijxixj transforms to a quadratic form bijyiyj,
write down the corresponding form of nonsingular linear transfor-
mation of it and what is the rank of |bij| if the rank of |aij| is 5 under
suitable transformation in the range of i and j.
2. If the quadratic form aijxixj (i, j = 1, 2, …, n) reduces to the form
bi ( y i )2 , bi ≠ 0 ∀ i by means of a nonsingular linear transformation,
what are the nonzero values of b’s if r (<n) is the rank of the latter.
3. With the help of an example, show that the signature of a real
­quadratic form remains the same.
4. Show that a definite quadratic form cannot be singular.
5. What are the loci (if possible) represented by the quadratic form
­aixixi  = 1, where N = 2, 3 where xi = 1, 2,…, N are rectangular
coordinates.
2
Concept of Tensors

2.1 Some Useful Definitions

In mathematics, symbolic representations of some concepts are essential for
writing expressions in concise form for convenience. For example,
Summation convention: It means the drop of summation sign ( Σ) in an
n

expression of the type ∑ a x = a x . It’s explicit form is a x = a x + a x

i=1
i
i
i
i
i
i
1
i
2
2
++

an x n, a linear expression in terms of n variables x 1 , x 2 , , x n; the index does

not mean power rather designates variable (position).
Dummy index: An index that occurs twice in a term once in the upper
position (prefix) and once in the lower position (subscripts) is called a dummy
index, e.g., in ai x i, i is the dummy index which indicates the summation of the
desired range, say, from 1 to n (if not mentioned).
Illustration: In order to write a system of linear equations of n
unknowns,  we  can use aαi x i = bα , where α = 1, 2, , n, i = 1, 2, , n. Actually,
it stands for

a11x 1 + a21 x 2 +  + an1 x n = b1

a12 x 1 + a22 x 2 +  + an2 x n = b 2



a1n x 1 + a2n x 2 +  + ann x n = b n .

These n linear equations in n unknowns are concisely written as aαi x i = bα .

This is definitely a stylistic representation of mathematics.
Real index: An index that occurs only once in a term like α in aαi x i = bα is
called a real index.
N.B.: A dummy index can be replaced by any other index not occurring
in that term, e.g., ai x i = a j x j indicating the same expression. But a real index
cannot be replaced except on either sides.

13
14 Tensor Calculus and Applications

Kronecker delta: Let xi (i = 1, 2, …, n) be a set of n independent variables.

Then,

∂xi
= 1, when i = j
∂x j
= 0, when i ≠ j

It is denoted by the symbol δ ji called the Kronecker delta.

 ∂xi 
∴  j =  δ ji = 1, when i = j
 ∂x 

= 0, when i ≠ j,

2.2 Transformation of Coordinates
Let us consider the set of n real independent variables x i (i = 1, 2, , n) to denote
the coordinates of a point P with respect to a system xi in an ­n-dimensional
space Vn. Otherwise, every n tuple of real variables corresponds to a point
in a space Vn. Let x ′ i (i = 1, 2, , n) be the coordinates of the same point P but
with respect to another system x ′ i (say). Since the point is the same, the two
sets of variables must be functionally connected by means of a transforma-
tion as x ′ i = f ( x i ).
From differential calculus, we can get

∂f ∂f ∂f
dx ′ i = dx 1 + 2 dx 2 +  + n dx n
∂x1 ∂x ∂x
n

∑ ∂∂xf dx = ∑ ∂∂xx′ dx
i
α α
= α α
α =1

or

∂x′i α
dx ′ i = dx (2.2.1)
∂ xα

∂x′i
by summation convention, where is the nonzero Jacobian of the
transformation. ∂ xα
The infinitesimal displacement dx i for the coordinate system xi determines
the direction at P in Vn, and they are called the components of a contravariant
Concept of Tensors 15

vector; dx ′ i are the corresponding contravariant vector components in the

other system x ′ i.
Defining dxα = Aα , (2.2.1) can be written as

∂x′i α
A′ i = A . (2.2.2)
∂ xα

Thus, if the set of n functions Aα (α = 1, 2, , n) of the x i coordinate system

transforms to the set of n functions A′ i (i = 1, 2 , n) in another system x ′ i by
means of the transformation law (2.2.2), it is called the components of a con-
travariant vector.
∂xβ
Multiplying (2.2.2) by and summing over i, we can get
∂x′i

∂xβ i ∂xβ ∂x′i α ∂xβ α

A′ = A = α A = δ αβ Aα
∂x′i ∂ x ′ i ∂ xα ∂x

∂xβ i
∴ Aβ = A′ , which is the reverse transformation law of (2.2.2).
∂x′i
Again from the transformation x ′ i = f ( x i ), we can solve for xi in terms x ′ i
∂x′i
of another function φ ( x ′ i ) as the Jacobian of the transformation ≠ 0. In
∂x j
this sense, x = φ ( x ′ ).
i i

Let the function φ ( x ′ i ) be invariant subject to coordinate transformation

(in general) or in restricted sense be a scalar invariant. Therefore, its partial
∂φ
derivatives with respect to the coordinates x ′ i, namely , admit the coor-
∂x′ j
dinate transformation according to the rule:

∂φ ∂φ ∂ x 1 ∂φ ∂ x 2 ∂φ ∂ x n
= . + +  +
∂x ′ i ∂x1 ∂x ′ i ∂x 2 ∂x ′ i ∂xn ∂x′i

∂φ ∂ x j
= ( j = 1, 2,… , n),
∂x j ∂x′i

by summation convention.
∂φ
Putting j = A j , it can be thrown to the form:
∂x

∂x j
Ai′ = A j (2.2.3)
∂x′i

∂φ
where A j ( j = 1, 2, , n) =
; the functions of the coordinates xj are called the
∂x j
components of a covariant vector, and A′j are its corresponding components
16 Tensor Calculus and Applications

in the x ′ i coordinate system. Otherwise, any set of n functions A j ( j = 1, 2, , n)

of the x i coordinate system is transformed to another set of n functions
Ai′(i = 1, 2, , n) of the x ′ i coordinate system according to the transformation
law (2.2.3) is called the components of a covariant vector.
∂φ
The partial derivatives are the components of grad φ vector and if
∂x j
φ ( x i ) = const(i = 1, 2, 3 say) like invariant, then ∇φ represents a vector normal
to the surface φ = c in three dimensions.
N.B.: The readers are advised to delve in coining the names “contravariant
vector” and “covariant vector” in the above definitions (2.2.2) and (2.2.3) with
the earlier uses of the definitions discussed in Section 1.2.
Cartesian tensors: If the components of a tensor are expressed relative to
Cartesian coordinates and the operative transformation is from Cartesian
to Cartesian, they are called Cartesian tensors.
Theory of continuum mechanics is developed on the basis of Cartesian
tensors.

Definition

Relative and absolute tensors: If the components of a tensor (may be of any

ijk
order, see Section 2.3) Aαβ are transformed according to the law,

w α
 ∂x  ijk ∂ x ′ ∂ x ′ ∂ x ′ ∂ x
m n r
∂xβ
Ast′ mnr =   Aαβ
 ∂x′  i j
∂x ∂x ∂x ∂x′ ∂x′t
k s

∂x
where is the Jacobian of the transformation associated with the factor
∂x′
w
 ∂x 
  , then the tensors are called relative tensors of weight w.
∂x′ 
When w = 0, the tensor is called absolute tensor, and when w = 1, the ­relative
tensor is called a tensor density.

Definition

Permutation (or pseudo) tensor: The absolute tensor ∈jik is defined as follows:

∈jik = 0, if any two of the indices i, j, k repeat

= 1, if i, j, k is an even permutation of 1, 2, 3
= −1, if i, j, k is an odd permutation of 1, 2, 3 is called permutation or
pseudo tensor.

Observation: According to the transformation law (2.2.2) for the two sets of
functions Aα and A′ i , we can write
Concept of Tensors 17

∂x′i α j ∂x′ j β
A′ i = α A and B′ = B ,
∂x ∂xβ

so that

∂x′i ∂x′ j β α β ∂x′ ∂x′

i j
A′ i B′ j = Aα B = ( A B )
∂ xα ∂ x β ∂ xα ∂ x β

∂x′i ∂x′ j
C ′ ij = Cαβ if we recognize Aα Bβ = Cαβ .
∂ xα ∂ x β

This prompts us to define the product of two vectors.

2.3 Second and Higher Order Tensors

Any set of n2 quantities A ij(i, j = 1, 2, , n) is called a second-order (rank)
­contravariant tensor if it satisfies the transformation law:

∂ x ′ i ∂ x ′ j ab
A′ ij = A . (2.3.1)
∂x a ∂xb

A set of n2 quantities A ij (i, j = 1,2,…,n) in the xi coordinate system transformed

to another set of n2 quantities A′ ij in the x ′ i coordinate system according to
the transformation law (2.3.1) is called the contravariant tensor of second
order (or of rank two).
Similarly, if a set of n2 quantities Bij (i, j = 1, 2, …, n) in the xi coordinate sys-
tem transforms to another set of n2 quantities Bij′ in the x ′ i coordinate system
according to the transformation law,

∂x a ∂xb
Bij′ = Bab (2.3.2)
∂x′i ∂x′ j

is called the covariant tensor of second order (or of rank two).

Also, if a tensor of the type A ij covariant character by one and contravariant
character by one in a coordinate system xi transforms to A′j i in the x ′ i coordi-
nate system according to the law of transformation,

∂x ′ i ∂xb a
A′j i = Ab
∂x a ∂x′ j

is called a mixed tensor of second order.

18 Tensor Calculus and Applications

Definition

Order or rank of a tensor: The number of designated real index in a tensor is

called the rank or order of a tensor. A′ ij , Aijk are the tensors of rank two and
three, respectively, of contravariant and covariant characters.

i. A vector is a tensor of rank one with 31 (or n1 ) components in three

dimensions (n dimensions).
ii. A stress α ij is a tensor of rank two with 32 = 9(or n2 ) components in
three dimensions (n dimensions).
iii. An invariant is also called a tensor of rank (or order) zero.
iv. The upper (superscript) index is reserved to represent contravariant
character and lower (subscript) to represent covariant character.

2.4 Operations on Tensors
Definition

Outer or open product of tensors: The components of a tensor of any rank

(or order) and character can be multiplied with the components of a tensor
of any rank and character, and the rank (order) of the resulting tensor is the
sum of their individual ranks.
For example, the product of the second-order mixed tensor A ij with the
third-order contravariant tensor Bmnr is Cjimnr, a fifth-order mixed tensor.
This is called outer or open product of two tensors.
Contraction: The process of equating a contravariant index with a covari-
ant index in a mixed tensor implicating summation with respect to it is
called a contraction.
ijk
For example, the tensor Aab may be subjected to several contractions,
namely,

Aibijk , Aaiijk , A ijk ijk ijk ijk

jb , Aaj , Akb , Aak .

In all these tensors, the number of real indices transforms to three instead of
five as in the original tensor. Can it not be guessed the following?

Theorem

Every contraction reduces the rank of a tensor by two.

ijk
Proof: Let us consider a fifth-order mixed tensor Aqp in the xi system.
mnr i
By transformation law Ast′ in the x ′ system, we can write
Concept of Tensors 19

∂ x ′ m ∂ x ′ n ∂ x ′ r ∂ x q ∂ x p ijk
Ast′ mnr = Aqp
∂xi ∂x j ∂x k ∂x′ s ∂x′t

∂ x ′ m ∂ x ′ n ∂ x ′ r ∂ x q ∂ x p ijk
Asr′ mnr = Aqp
∂xi ∂x j ∂x k ∂x′ s ∂x′r

taking r = t, i.e., assuming a contraction

∂ x ′ m ∂ x ′ n ∂ x p ∂ x q ijk
Asr′ mnr = Aqp
∂xi ∂x j ∂x k ∂x′ s
∂ x ′ m ∂ x ′ n p ∂ x q ijk
= δk Aqp
∂xi ∂x j ∂x′ s
∂x′m ∂x′n ∂xq ijk
Asr′ mnr = 1 ⋅ Aqk
∂xi ∂x j ∂x′ s

which is the transformation law of a third-order mixed tensor as the dummy

index r on the left and k on the right are not countable in the rank.
Hence, proved.
Inner product of tensors: A multiplication associated with a contraction in
tensors is called an inner product of tensors.
For example, Aij B jkr = C ikr is a third-order tensor instead of the fifth-order
tensor due to one contraction.

Aij B jir = C r

is the first-order tensor (vector) due to two successive contractions.

Higher order tensors: After formulation of second-order tensors and mul-
tiplication (outer or open product) of tensors, we can define tensors of any
order.
The mth-order contravariant tensor A r1 r2 ,, rm connecting the nm components
in the xi system with the corresponding nm components in the x ′ i system can
be defined as

∂ x ′ s1 ∂ x ′ s2 ∂ x ′ sm r1 , r2 ,…, rm
A′ s1 , s2 ,…, sm =  rm A
∂ x r1 ∂ x r2 ∂x

Similarly, a kth-order covariant tensor Aabc…s of nk components in the xi

­system can be defined by means of the transformation to the corresponding
nk ­components in the x ′ i system as

∂x a ∂xb ∂x c ∂x s
′ …u =
Apqr  Aabc…s ,
∂x′ p ∂x′q ∂x′r ∂x′u
20 Tensor Calculus and Applications

where (p, q, r, …, u) or (a, b, c,…, s) is the k number of indices. Any mixed

­tensor of any order can similarly be defined.
Addition (subtraction) of tensors: Two tensors of the same rank and simi-
lar character (covariant or contravariant) can be added (components), and the
components of one can be subtracted from the other. Otherwise, the sum or
difference of two tensors of the same rank and character is a tensor of the
same type. For example,
Let A ijk and Bijk be the two third-order mixed tensors of the same character
so that

A ijk + Bijk = C ijk , ( say ) where C ijk is doubtful. (2.4.1)

In the x ′ i system, its replica is

∂x′m ∂x j ∂x k i ∂x′m ∂x j ∂x k i
′ m = Anr
Cnr ′ m + Bnr
′m = A jk + Bjk
∂xi ∂x′n ∂x′r ∂xi ∂x′n ∂x′r

∂x j ∂x k ∂x′m ∂x j ∂x k
) ∂∂xx′
m
(
= A ijk + Bijk i
∂x′ ∂x′
n r
= C ijk
∂xi ∂x′n ∂x′r
,

(using 2.4.1) which is the transformation law of a third-order mixed tensor of

the same character.
Hence, C ijk (the doubtful one) is also a tensor of the same kind.
Similarly, the difference of two tensors of the same type like
A ijk − Bijk (or A ij − Bij ) can also be proved to be the tensor of the same type.

2.5 Symmetric and Antisymmetric

(or Skew-Symmetric) Tensors
When the relative position of two indices contravariant or covariant in the
components is interchangeable in a tensor, it is called a symmetric ten-
sor  with respect to these indices, i.e., A ijk (or A ij) is symmetric if A ijk = A ikj
(or A ij = A ji ).
ij
Similarly, Aij or A is said to be antisymmetric (or skew-symmetric) if

Aij = − A ji   or A ij = − A ji .

Theorem

The symmetric (or antisymmetric) property of tensors remains unchanged

(invariant) due to transformation of coordinates.
Concept of Tensors 21

Proof: Let Aij be the components of a symmetric tensor in the xi coordinate

system so that Aij = A ji .
′ are the corresponding components of the tensor in the x ′ i coordinate
If Amn
system, then it is to satisfy

∂xi ∂x j
′ =
Amn Aij
∂x′m ∂x′n
∂x j ∂xi
= A ji .
∂x′n ∂x′m
∴ Aij = A ji

′ = Anm
Amn ′

Hence, proved.
Similarly, it can also be proved that

A′ ij = − A′ ji if A ij = − A ji.

Example 1
Prove that every second-order tensor covariant (or contravariant) can be
expressed as the sum of symmetric and antisymmetric tensors.
Let Aij be any second-order covariant tensor. It can be written as

1 1
Aij =
2
( ) (
Aij + A ji + Aij − A ji
2
)
∴ Aij = Bij + Cij , (i)

where

1

2
( )
Aij + A ji = Bij (ii)

1

2
( )
Aij − A ji = Cij (iii)

as they are conformable for addition and subtraction of tensors.

In view of writing (ii),

1 1
Bji =
2
( )
A ji + A ij = Aij + A ji
2
( )

∴ Bij = Bji .

Hence, Bij is a symmetric tensor.

22 Tensor Calculus and Applications

Also, in view of writing (iii),

1 1
C ji =
2
( ) ( )
A ji − Aij = − Aij − A ji = −Cij .
2
Hence, Cji is an antisymmetric tensor.
From (i), Aij is expressed as the sum of symmetric and antisymmetric
tensors.
Similarly, it can be proved for second-order contravariant tensor Aij also.
N.B.: This result is similar to the result of a square matrix, namely,
every square matrix can be expressed as the sum of symmetric and
­antisymmetric matrices.

Example 2
If Ai are the components of any contravariant vector in the sum AiBi
which is a scalar invariant, prove that Bi are also the components of a
covariant vector.
Given AiBi is a scalar invariant.
∴ For a transformation of coordinates from the xi system to the x ′ i
system,

A ′ i Bi′ = A i Bi ∴ i is dummy

∂x ′ i α
A Bi′ = Aα Bα (i → α )
∂ xα
 ∂x′ i 
∴ Aα  Bα − α Bi′ = 0
 ∂x 

∂x′ i
Bα = Bi′ .
∂ xα

 Aα is any contravariant vector.

This is the transformation law of covariant vector.
Hence, B𝛼 or Bi is also a covariant vector.
N.B.: It will be seen later that AiBi is the scalar product of two vectors
Ai and Bi.

Example 3
The number of independent components of a second-order symmetric
1
tensor in n dimensions is n(n + 1).
2
The total number of components of a second-order covariant tenser Aij
(say) is n2. But the number of components with repeated suffix Aii is n.
Hence, the number of components with distinct indices Aij (i ≠ j) is
n2 − n. But for symmetric property Aij = Aji, the actual number of indepen-
1
dent components out of these is (n2 − n).
2
Concept of Tensors 23

∴ Total number of independent components is

1 2 1 1 1
= (n – n) + n = n2 + n = n ( n + 1)
2 2 2 2

Hence, proved.

Example 4
If n3 quantities Bijk are connected with n3 quantities A ijk by the relations,
h ∂y ∂x
i m
∂xl i ∂x
l
Bijk = Aml , show that B i
ik = Aij .
∂x h ∂y j ∂y k ∂y k
h ∂y ∂x
i m
∂xl
Given Bijk = Aml j .
∂x ∂y ∂y k
h

Putting i = j, i.e., making a contraction with respect to i and j,

h  ∂y ∂x  ∂x
i
h ∂x ∂xl
m l m
Biki = Aml  ∂ x h ∂ y i  ∂ y k = Aml ∂ x h ∂ y k

∂xl h ∂x
l
i ∂x
l
= Aml
h
δ hm = Ahl = Ail ( h → i).
∂y k ∂y k ∂y k

Hence, proved.

Example 5
If aij and bij are the two symmetric tensors, and ui, vi are the components
of contravariant vectors satisfying the equations

(a ij   ) ( )
– kbij ui = 0 and aij   – k ′bij v i = 0,

i, j = 1, 2,  , n k ≠ k ′

prove that bij uivj = 0 and aij uivj = 0.

Given aij = aji, bij = bji and

( aij  –  kbij )ui = 0 (i)

and

( aij – k ′bij )v i = 0

( a ji – k ′b ji )v j = 0 . (ii)

Multiplying (i) by vj and summing over j and multiplying (ii) by ui and

summing over i from i, j = 1, 2,…, n, we get

( aij – kbij )ui v j = 0 (iii)

24 Tensor Calculus and Applications

( a ji – k ′b ji )ui v j = 0 . (iv)

(iii)–(iv) gives

( k – k ′)bij ui v j = 0 ∴ aij = a ji and bij = b ji

∴ bij   ui v j = 0 k ′ ≠ k.

Also (iii) × k′ and (iv) × k gives

( k ′ – k)aij   ui v j = 0 ∴ aij   ui v j = 0 k ′ ≠ k

Hence, proved.

Theorem

Show that the tensor law of transformation possesses the group property.
Proof: Consider the contravariant (may be covariant also) components Ai
and A′ i of the same vector in xi and x ′ i systems, respectively.

∂x′i α
∴ A′ i = A (i)
∂ xα

Let us consider a third coordinate system x ′′ j with components A′′ j of the

same vector.

∂ x ′′ j i ∂ x ′′ j ∂ x ′ i α
∴ A′′ j = A′ = A ,
∂x′i ∂ x ′ i ∂ xα

using (i) and summing over i.

∂ x ′′ j α
A′′ j = A
∂ xα

which is the direct transformation of tensor (here vector) components from

the xi system to the x ′′ j system. Thus,

i. Two successive tensor law of transformations give rise to another

tensor law of transformation, i.e., identical to closure property of
group.
ii. Compositions of tensors are always conformable with respect to
addition and multiplication and so associatively is obvious.
iii. Setting x ′ i = x𝛼 in (i), we can recover the same tensor, namely, A′ i = A𝛼,
which is equivalent to the existence of identity.
Concept of Tensors 25

∂x′ j
iv. For nonzero value of the Jacobian ≠ 0, every tensor law of trans-
∂ xα
∂ xα i
formation admits the “inverse” transformation (like Aα = A′ )
∂x′i
law.

Hence, tensor law of transformations are said to possess the group property.

2.6 Quotient Law
If the functions uaijkα in the xi system and the functions ua′αijk in the x ′ i system are
such that uaijkα vα and ua′αijk v ′α are the components of a tensor, where vα and v′α
are the components of arbitrary vectors in the systems, then the given func-
tions must be the components of a tensor.
Proof: Given

uaijkα vα = a tensor Aaijk (say, a tensor in the x i system) (i)

(
ua′αijk v ′α = a tensor Aa′ ijk in the x ′ i system (ii) )
∂x′i ∂x′ j ∂x′ k ∂xm
∴ ua′αijk v ′α = Ambcd , from ( ii )
∂xb ∂x c ∂x d ∂x′ a
∂x′i ∂x′ j ∂x′ k ∂xm
= v , using ( i )
bcd n
, umn
∂xb ∂x c ∂x d ∂x′ a
∂x′i ∂x′ j ∂x′ k ∂xm bcd ∂xn α
= umn v′
∂xb ∂x c ∂x d ∂x ′ a ∂ x ′α

 vn (or v𝛼) is a tensor.

or

 ijk ∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ x m ∂ x n bcd  α
 ua′α − ∂ x b ∂ x c ∂ x d ∂ x ′ a ∂ x ′α umn  v ′ = 0

∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ x m ∂ x n bcd
∴ ua′αijk = umn .
∂ x b ∂ x c ∂ x d ∂ x ′ a ∂ x ′α

 v𝛼 (or v′α ) is arbitrary.

26 Tensor Calculus and Applications

This is the transformation law of a fifth-order mixed tensor. Hence, the

given functions are also the components of a tensor.
Hence, proved.
N.B.: Some authors state the quotient law as “If the inner product of a class
of functions with an arbitrary vector is a tensor, then the class of functions
is also a tensor.” Students are advised to explore the fallacy in the statement.

Exercises
1. Show that subject to rectangular Cartesian coordinate transforma-
tion, there is no distinction between contravariant and covariant
vectors (or tensors).
2. From the transformation law of second-order contravariant and
j
∂x k kl ∂ x ′ ∂x′i ∂xl
covariant tensors, show that A′ ij = A and A ′
ij = Akl .
∂x′i ∂xl ∂x k ∂x′ j
[Hint: They can be obtained from the tensor law of transformations.]
3. If the equation fAij + gAji = 0 is satisfied by the nonzero second-order
tensor Aij with respect to a basis, prove that
i. f = g when Aij = −Aji
ii. f = − g when Aij = Aji.
4. If the components of a tensor (of any rank) vanish in one coordinate
system, prove that they identically vanish in any other coordinate
system. [Hint: Take Aij = 0 and write the tensor law of transforma-
tion in the x ′ i system.]
5. If the relation aijbk + ajkbi + akibj = 0 is satisfied by the symmetric tensor
aij and the arbitrary tensor bk, prove that either aij = 0 or bk = 0.
6. If yi are n independent functions of the variables xi, and zi are n inde-
∂xi ∂y i ∂y j
pendent functions of the yi so that ui = v j j , v i = w j j , U i = Vj ,
∂y ∂z ∂xi
∂z j
and Vi = Wj i , show that
∂y

uiU i = v iVi = w iWi

w−2
∂x
7. Show that A′ ij = A ij , where Aij and Aij are the components of
∂x′
symmetric relative contravariant and covariant tensors, respectively,
of weight w.
Concept of Tensors 27

8. If aij are the components of a covariant tensor, show that the cofactors
of the elements aij in |aij| are the components of a relative contravari-
ant tensor of weight 2.
9. Show that the Kronecker delta is a mixed tensor of rank two.
[Hint: Use quotient law, taking an arbitrary vector.]
10. If x , y ; x, y are the components of velocity and acceleration,
­respectively, of a fluid element in Cartesian coordinates, find the
­corresponding quantities in plane polar coordinates.
11. Defining the conjugate component of a symmetric covariant t­ ensor
of second order, prove that the conjugate tensor fields aij and aij ­satisfy
aij aij = n in an n-dimensional Riemannian space.
12. Prove that the velocity of a fluid element is a contravariant vector
of rank one, but acceleration is not a tensor in general. How do you
define the acceleration so that it is a tensor?
3
Riemannian Metric and
Fundamental Tensors

We have already introduced the concept of metric in Section 1.4 as the

­infinitesimal distance ds between two adjacent points in terms of a quadratic
differential form.

3.1 Riemannian Metric
If xi and xi + dxi (i = 1, 2,…, n) are the two adjacent points of a coordinate
­system of n dimensions, then the infinitesimal distance ds between the
points is defined by Riemann as

ds2   = g ij   dx i   dx j, (3.1.1)

where g ij ’s are the functions of the coordinates xi subject to the restric-

tion that the determinant g = g ij ≠ 0 . At this stage, g ij’s are termed as
metric ­f unctions. For real quadratic differential, it is assumed to be posi-
tive definite (as d ­ iscussed in Section 1.3). But it needs to mention that
the metric may not always be definite as in the case of general theory of
relativity which may be indefinite also. The metric defined by Riemann
above is an extension of ­c lassical metric to n dimensions and is called the
Riemannian metric.
The geometry based on a Riemannian metric is called the Riemannian
geometry, and the space whose geometry is developed on the basis of such a
metric is called the Riemannian space.

3.2 Cartesian Coordinate System and

Orthogonal Coordinate System
If the metric (3.1.1) is reducible to a particular coordinate system yi by means
of a transformation so that

29
30 Tensor Calculus and Applications

ds2 = aij dy i dy j (3.2.1)

where aij are constants, the system is called a Cartesian coordinate system. On
the other hand, if aij = 0 for i ≠ j, the coordinate system is called o
­ rthogonal
coordinate system.

3.3 Euclidean Space of n dimensions, Euclidean

Co-ordinates, and Euclidean Geometry
If the metric (3.2.1) in terms of the coordinates yi reduces in particular to the
form

n n

ds2 = ∑
i=1
δ ji dy i dy j = ∑( dy ) (3.3.1)
i=1
i 2

for aij = δ ji , it is called the Euclidean metric, and the corresponding space is
called the Euclidean space. The geometry developed on the basis of Euclidean
metric is called the Euclidean geometry. Here, yi’’s (when expressed in (3.3.1)
form) are called the Euclidean coordinates.

Definition

Element of length: The element of length ds (in conformity with the concept
of metric) is defined as ds2 = ∈ g ij dx i dx j , where ∈ is ±1 so as to make the right
hand side positive. Therefore, the element of length “ds” is the magnitude
of a contravariant vector, dx i  = λ i (say)  so that λ 2 = ∈ g ij λ i λ j (will be seen
later). It represents the magnitude of any contravariant vector characterized
by the vector components λi.
Therefore, ds2 from its nature being the intrinsic concept of a space and
­correlating with the magnitude of a vector must be invariant.

3.4 The Metric Functions g ij Are Second-Order

Covariant Symmetric Tensors
The metric ds2 = g ij dx i dx j in Riemannian space is invariant.
So,
Riemannian Metric and Fundamental Tensors 31

α
′ dx′ dx′ β = g ij dx i dx j
(ds2 =) gαβ in x′α and x i systems

∂xi ∂x j
= g ij α
dx′α dx′ β
∂ x′ ∂ x′ β

 ∂xi ∂x j  α β
 g ′ − g dx′ dx′ = 0
∂ x′α ∂ x′ β 
αβ ij

∂xi ∂x j
′ = g ij
gαβ ( dx′α is arbitrary),
∂ x ′α ∂ x ′ β

which is the transformation law of a covariant second-order tensor.

Hence, gij is a second-order covariant tensor.
Subject to the conformability of addition and subtraction of tensors, gij can
be written as

1 1
g ij =
2
( )
g ij + g ji + g ij − g ji
2
( )

= Aij + Bij ,

1 1
putting
2
( ) (
g ij + g ji = Aij and g ij − g ji = Bij .
2
)
Interchanging i and j in both Aij and Bij, we get

1
A ji =
2
( )
g ji + g ij = Aij .

1 1
Hence, it is symmetric, and Bji = ( g ji − g ij ) = − ( g ij − g ji ) = − Bij ; hence, Bij is
antisymmetric. 2 2

( )
∴ g ij dx i dx j = Aij + Bij dx i dx j = Aij dx i dx j + Bij dx i dx j
(3.4.1)
= Aij dx i dx j + 0.

For

Bij dx i dx j = Bji dx j dx i ( i ↔ j )

= − Bij dx i dx j .

 Bij is antisymmetric
or 2 Bij dx i dx j = 0
or Bij dx i dx j = 0
32 Tensor Calculus and Applications

( )
Hence, from (3.4.1), g ij − Aij dx i dx j = 0.

∴ g ij = Aij ,

which is a symmetric tensor,  dx i is arbitrary.

∴ The metric functions gij are the second-order symmetric covariant
­tensors. They are also called fundamental covariant tensors.
Note: The readers are advised to explore the use of the nomenclature
­“fundamental” for the tensor gij.
For the evolution of “tensor calculus,” the covariant and contravariant
tensors (yielding mixed tensor) of some kind it should go together hand in
hand. So, it demands to define a contravariant tensor as the counterpart of gij
to supplement the development of the branch.
Some applicable results:

1. If g = g ij ≠ 0 and Gjk are the cofactors of gkj in g, then from the proper-
ties of determinant connecting cofactors, we have

g ij G jk = g when i = k

= 0 when i ≠ k.

Otherwise, g ij G ik = gδ jk .
2. If λij is a second-order covariant tensor and u j an arbitrary vector,
then their inner product λij u j is a covariant vector vi , i.e.,
λij u j = vi but g ij u j = u i.
The difference is to be noted down for conception (!). In this case,
u i is called the associate vector to u j by means of the fundamental
tensor g ij.
Similarly, if g ij is the second fundamental tensor (it will be proved
in Section 3.5) and u j (associate to ui by means of g ij) is an arbitrary
vector, then gijuj=ui, a contravariant vector. Thus,
3. The inner multiplication of any tensors (vectors) contravariant and
covariant by gij and gij, respectively, is to be treated as a means to
lower down or raise the indices in tensors.

Definition

The set of n2 functions gij is defined as

cofactor of g ij in the determinant g

g ij = ,
g

where
Riemannian Metric and Fundamental Tensors 33

g = g ij ≠ 0

G ji
= , G ji ,
g

written conventionally to denote the cofactor of gij in g.

It can be proved that the class of functions gij is a second-order symmetric
contravariant tensor.

3.5 The Function g ij Is a Contravariant

Second-Order Symmetric Tensor
Let us consider the covariant vector (tensor) uj by means of its associate
­ ector uk so that uj=gkj uk, where uk is an arbitrary vector.
v
Now,

G ji G ji uk
u j g ij = u j
g
= g kj uk
g
( )
= g kjG ji
g

uk
= gδ ki = ui ,
g

which is a vector (i.e., a tensor rank one).

∴ The inner product of the function gij with the arbitrary vector uj is a
tensor.
Hence, by quotient law of tensors, gij is a contravariant tensor of rank
(order) two. It is also symmetric.
For gij is symmetric, ⇒ G ji is also symmetric.
G ji
⇒ is also symmetric, since the value of g = g ij ≠ 0 will not change the
g
symmetric character on division except the numerical value.
∴ g ij is also a second-order symmetric contravariant tensor, and it is also
called “fundamental tensor.”
N.B.: In conformity with gij, gij a mixed-order tensor g ij is defined as g ij = δ ji,
it is also a tensor because δ ji is a mixed second-order tensor.

Example 1
Show that

( )
g hj g ik − g hk g ij g hj = ( n − 1) g ik
i.
34 Tensor Calculus and Applications

∂K ∂K ∂K
ii. j ( g hk g il − g hl g ik ) g hj = k g il − l g ik
∂x ∂x ∂x

(
i. L.H.S. = ng ik − g hkδ ih ) ( g hj g hj = δ hh = n )
= ng ik − g ik

= ( n − 1) g ik = R.H.S.
Hence, proved.

∂K
ii. L.H.S. =
∂x j
( ghk gil − g hl gik ) g hj
∂K ∂K

=
∂x
(
j g hk g
hj
)
g il − j g hl g hj g ik
∂x
( )
∂K j ∂K
= δ k g il − j δ lj g ik
∂x j ∂x
∂K ∂K
= g il − l g ik = R.H.S.
∂x k ∂x
Hence, proved.

Example 2
Calculate gij for the coordinates xi, given that

y 1 = x 1 x 2 cos x 3
y 2 = x 1 x 2 sin x 3
1 1
( ) + (x ) 
2 2 2
y3 = x
2  

where x 1 ≥ 0, x 2 ≥ 0, 0 ≤ x 3 ≤ 2π.
yi is a Cartesian orthogonal coordinate system.
For a Cartesian coordinate system yi, the metric is of the form
ds 2 = gαβ′ dy ′α dy ′ β , where gα′ β are constants (for convenience of tensor,
y ′  – dash is used in the metric).
But if it is orthogonal too, then

′ = 1 if α = β
gαβ

= 0 if α ≠ β .

∴It can be expressed in the form:

( )
2
ds 2 = g ii′ dy ′ i , ′ =1 if α = β .
gαβ

To make use of tensor notation, the given Cartesian orthogonal coordi-

nate system yi is replaced by x ′ i so that x′ 1 = y 1 , x′ 2 = y 2 , x′ 3 = y 3
Riemannian Metric and Fundamental Tensors 35

∂ x ′α ∂ x ′ β
g ij = ′
gαβ
∂xi ∂x j

∂ x ′α ∂ x ′α
g ij = ′
gαα
∂xi ∂x j
′ =1
 gαβ

exists when α = β .

∂ x′ 1 ∂ x′ 1 ∂ x′ 2 ∂ x′ 2 ∂ x′ 3 ∂ x′ 3
∴ g11 = ′ +
g11 ′ +
g 22 ′
g 33
1
∂x ∂x 1 1
∂x ∂x 1
∂x1 ∂x1

( ) 1 + (x ) ( )
2 2 2
= x 2 cos x 3 2
sin x 3 1 + x1 ⋅1

( y i
= x′ ) i

( ) + (x )
2 1 2
= x2

∂ x′ 1 ∂ x′ 1 ∂ x′ 2 ∂ x′ 2 ∂ x′ 3 ∂ x′ 3
g 22 = ′ +
g11 ′ +
g 22 ′
g 33
2
∂x ∂x 2 2
∂x ∂x 2
∂x 2 ∂x 2

( ) ( ) ( )
2 2 2
= x 1 cos x 3 ⋅ 1 + x 1 sin x 3 ⋅ 1 + x2 ⋅1

( ) + (x )
2 2 2
= x1

∂x′1 ∂x′1 ∂x′ 2 ∂x′ 2 ∂x′ 3 ∂x′ 3

g 33 = ′ +
g11 3 3 ′ +
g 22 ′
g 33
3
∂x ∂x 3
∂x ∂x ∂x 3 ∂x 3

( ) ( )
2 2
= − x 1 x 2 sin x 3 1 + x 1 x 2 cos x 3 ⋅1+ 0

( )
2
= x1x 2 .

But

∂ x′ 1 ∂ x′ 1 ∂ x′ 2 ∂ x′ 2 ∂ x′ 3 ∂ x′ 3
g13 = ′ +
g11 ′ +
g 22 ′
g 33
1
∂x ∂x 3 1
∂x ∂x 3
∂x1 ∂x 3

( )( ) (
= x 2 cos x 3 − x 1 x 2 sin x 3 + x 2 sin x 3 x 1 x 2 cos x 3 ⋅ 1 + x 1 × 0 )( )

( ) ( )
2 2
= − x1 x 2 cos x 3 sin x 3 + x 1 x 2 cos x 3 sin x 3

=0

 0 ≤ x 3 ≤ 2 π and x 1 ≥ 0

and
36 Tensor Calculus and Applications

∂ x′ 1 ∂ x′ 1 ∂ x′ 2 ∂ x′ 2 ∂ x′ 3 ∂ x′ 3
g 23 = g ′
11 + g ′
22 + ′
g 33
∂x 2 ∂x 3 ∂x2 ∂x3 ∂x2 ∂x3

( ) (
= x 1 cos x 3 − x 1 x 2 sin x 3 ⋅ 1 + x 1 sin x 3 x 1 x 2 cos x 3 + 0 )
=− x ( )1 2
x 2 cos x 3 sin x 3 + x ( )1 2
x 2 cos x 3 sin x 3

 0 ≤ x 3 ≤ 2π

= 0.

( ) + (x ) ( )
2 2 2 2
Thus, g11 = g 22 = x 1 , g12 = 2 x 1 x 2 , g 33 = x 1 x 2 , and g13 = g 23 = 0 .

Example 3
If the metric in V2 is such that ds 2 = Edu2 + 2 Fdudv + Gdv 2 , then find the
values of gij (i, j = 1, 2).
The given metric in V2 is ds 2 = Edu2 + Fdudv + Fdvdu + Gdv 2

E F
∴ g ij = g = = EG − F 2 ≠ 0,
F G
where

g11 = E, g12 = g 21 = F , and g 22 = G.

But

cofactor of g ij in g
g ij =
g
( g ≠ 0)

cofactor of g11 in g G
∴ g 11 = =
g g
cofactor of g12 or g 21 in g
g 12 = g 21 =
g
−F
=
g
and

cofactor of g 22 in g E
g 22 = =
g g
G −F
∴ g 11 = , g 12 = g 21 =
g g
and

E
g 22 = ,
g
g = EG − F 2 ≠ 0.
Riemannian Metric and Fundamental Tensors 37

Example 4
Determine the conjugate metric tensor of gij in the spherical coordinate
system.
In the spherical coordinate system, the metric is given by

ds 2 = dr 2 + r 2 dθ 2 + r 2 sin 2 θ dφ 2 .

Comparing it with ds 2 = g ij dx i dx j = g11 dx 1 dx 1 + g 22 dx 2 dx 2 + g 33 dx 3 dx 3, we

can write

( ) ( ) ( )
2 2 2
ds 2 = dx 1 + r 2 dx 2 + r 2 sin 2 θ dx 3

in the given metric, where x 1 = r , x 2 = θ , x 3 = φ .

∴ g11 = 1, g 22 = r 2 , g 33 = r 2 sin 2 θ , and g12 = g 23 = g13 = 0

1 0 0
∴g = 0 r2 0 = r 4 sin 2 θ ≠ 0.
0 0 r2 sin 2 θ

Now, by definition,

cofactor of g11 in g 1 r2 0
g 11 = = 4 ×
g r sin 2 θ 0 r 2 sin 2 θ

r 4 sin 2 θ
= =1
r 4 sin 2 θ

cofactor of g 22 in g 1 1 0
g 22 = = 4 ×
g r sin 2 θ 0 r 2 sin 2 θ

r 2 sin 2 θ 1
= =
r 4 sin 2 θ r 2

cofactor of g 33 in g 1 1 0
g 33 = = 4 ×
g r sin 2 θ 0 r2

r2 1
= =
r sin 2 θ r 2 sin 2 θ
4

and gij = 0 for i ≠ j.

38 Tensor Calculus and Applications

3.6 Scalar Product and Magnitude of Vectors

 
If a(ai ) and b (bi ) are the two non-null vectors, the scalar product of the two
vectors is defined as
i
a ⋅ b = gij a b j
(3.6.1)
= a ibi = a j b j
in terms of associate vectors of ai and bj.
If ai and bi are their covariant components, the scalar product of a and b can
also be defined as

( )
a ⋅ b = g ij a i b j a jb j or a jb j .

On the other hand, if the scalar product of (say) a is considered in itself, then

a ⋅ a = g ij a a j = a j a j = a 2 (3.6.2)
i


which is the square of the magnitude of the vector a .
dx i
If we recall the metric in a Vn, namely, ds2 = g ij dx i dx j and t i =
are the
ds
components of a unit tangent at a point P to a curve C of Vn, the metric can
be simplified to give

dx i dx j
1 = g ij
ds ds
= g ijt it j ,

which confirms the unit magnitude (defined above) of ti.

dx i dx j
N.B.: If for a curve or for a portion of a curve g ij = 0 that is of length
ds ds
(magnitude) zero or minimal, the lines of zero length are identified as the
world lines of light in space–time continuum of general relativity.

3.7 Angle between Two Vectors and Orthogonal Condition

If a and b are the magnitudes of the two non-null and
 nonparallel vectors
  
a b
a  and b, respectively, with inclination θ, then and are the unit vectors to
represent their directions. a b
Riemannian Metric and Fundamental Tensors 39

The cosine of the angle between these two unit vectors (in analogy with
Euclidean space of three dimensions E3) in a Riemannian Vn is defined as
 
a b a ⋅b g ij a ib j
cosθ = ⋅ = = (3.7.1)
a b ab g ij a i a j g ijb ib j

using (3.6.1) and (3.6.2).

For orthogonality of the vectors cosθ = 0 which gives the mathematical
expression g ij a ib j = 0; otherwise, a j b j = 0 when the two vectors are orthogonal.
If â and b̂ are the unit vectors so that

g ij a i a j = 1 = g ijb ib j , then cosθ = g ij a ib j = a jb j .

But for a real positive definite fundamental form of a Riemannian space Vn,
the numerical value of cos θ must not be greater than unity. This can be
proved as follows:
The square of the vector maˆ + nbˆ determined by the pencils ai and bi can be
written as

λ 2 = g ij (ma i + nb i )(ma j + nb j ) [= g ij λ i λ j ]
= g ij a i a j m2 + g ij a ib j mn + g ij b i a j mn + n2 g ijb ib j
(i ↔ j)

i j
2
= m + 2 g ij a b mn + n 2

 n  n 
2
= m2 1 + 2 g ij a ib j +   
 m  m 

 n 2
2

(
= m2  + g ij a ib j  + 1 − g ij a ib j  .
  )
 m 

n
( )
2
Since for all , this value should be positive, so g ij a ib j ≤ 1.
m
∴ cos 2θ ≤ 1.

∴ The angle θ defined above is real for a positive definite fundamental form
of Riemannian Vn.

Exercises
1. If u(ui ) and v ( vi ) are the two vectors, then show that the quantities
Aij  = uivj − ujvi are the components of a skew-symmetric covariant
40 Tensor Calculus and Applications

tensor of order two. What is the connection of it with the cross prod-
uct of the two vectors? [Hint: Interchanges i and j.]
2. Prove that gdx 1dx 2 dx n = g ′ dx′ 1dx′ 2  dx′ n for an element of
­volume in Vn.
3. Determine the components of the fundamental contravariant tensor
gij for
i. ds2 = dr2 + r2 d𝜃2 + dz2
ii. ds2 = dr2 + r2 d𝜃2 + r2sin2 𝜃d𝜙 2.
4. Prove that the cosine of angle between two coordinate paramet-
g ij
ric curves xi = constant and xj = constant is . [Hint: Angle
g ii g jj
between the curves is the angle between the normal vectors ∇x i , ∇x j.]
5. Show that the angle between two contravariant vectors is real when
the Riemannian metric is positive definite.
4
Christoffel Three-Index Symbols (Brackets)
and Covariant Differentiation

Notations and symbols are some fundamental ingredients for strong and
beautiful representation of mathematical concepts. Tensor is the embodi-
ment of such shorthand symbols and notations, which can concisely and
forcefully uphold mathematical expressions as described in this chapter.

4.1 Christoffel Symbols (or Brackets) of

the First and Second Kinds
If gij and gij be the first and second kinds of fundamental tensors, respec-
tively, then the Christoffel symbols (or brackets) of the first and second kinds
are, respectively, defined as

 ∂g ∂ g jk ∂ g ij 
 ij, k  = = 12  ikj + − (4.1.1)
∂ x i ∂ x k 
ij , k
 ∂x

and

{ }=
i
jk jk
i
= g iα jk , α (4.1.2)

Since gij is symmetric, ij , k is also symmetric, which is obvious from its nature
of occurrence in (4.1.1) with respect to i and j indices. Also, as g iα and jk ,α are
i
both symmetric, jk is also symmetric with respect to j and k indices.
i
Observation: From the nature of definition of ik in (4.1.2), we can promptly
i
verify that g iβ jk = g iβ g iα jk ,α = δ βα jk ,α = jk , β which is the first kind.
i
N.B.: The author will make use of the symbols ij , k and jk throughout this
book instead of the following old symbols:  ij, k  and { }.
i
jk

41
42 Tensor Calculus and Applications

4.2 Two Standard Applicable Results of Christoffel Symbols

∂ g ik
ij , k +
i. jk , i = .
∂x j
∂
ii. ij =
i

∂x j
( )
log ± g ( g > 0, g < 0).

Proof of (i)
1  ∂ g ik ∂ g jk ∂ g ij  1  ∂ g ji ∂ g ki ∂ g jk 
+ = + − + + −
2  ∂ x j ∂ x i ∂ x k  2  ∂ x k ∂ x j ∂ x i 
ij , k jk , i

1 ∂g
i. = × 2 ikj , g ij = g ji , etc.
2 ∂x
∂ g ik
= .
∂x j

Hence, proved.
Proof of (ii)
Differentiating partially the elements of the determinant g = g ij row wise
with respect to x j , and expanding each of the n determinants in terms of the
elements of that particular differentiated row, it can be precisely written as
a whole:

∂ g ∂ g ik ki
= G ,
∂x j ∂x j

where Gki are the cofactors of gik in g (i, k = 1, 2,…, n).

∂ g ik
= g g ik
∂x j

1 ∂g
= g ik  ij , k + jk , i  , ( g ≠ 0)

g ∂x j

using the result (i) above.

= g ik ij , k + g ki jk , i g ik = g ki
i k i
= ij + jk =2 ij
( k → i)
Christoffel Three-Index Symbols 43

1 ∂g ∂
ij
i
=
2 g ∂x j
=
∂x j
(
log g ) if g > 0

∂
=
∂x j
(
log − g ) if g < 0.

Hence, proved.

4.3 Evolutionary Basis of Christoffel Symbols (Brackets)

In the Cartesian coordinate system, the law of parallel displacement takes
the form:

ai , k δ x k = 0
∂ai α , (4.3.1)
i.e., δ xk = 0  ik =0
∂x k

where δxk represents the infinitesimal displacement. Let us transform the

vector components ai to a new coordinate system (ξi). Then, we have

∂ξ r
ai = ar′
∂xi
∂ ai ∂  ∂ξ r  ∂ξ s
∴ = s  i ar′  k
∂x k
∂ξ  ∂ x  ∂ x

∂ξ s ∂ξ r ∂ ar′ ∂ξ s ∂ x l ∂2 ξ r
= + ar′ (4.3.2)
∂ x k ∂ x i ∂ξ s ∂ x k ∂ξ s ∂ x l ∂ x i

∂ ai  ∂ξ r ∂ a ′ ∂ x l ∂2 ξ r  ∂ξ s
∴0 = δ x k =  i rs + s l i ar′  k δ x k
∂x k
 ∂ x ∂ξ ∂ξ ∂ x ∂ x  ∂ x

 ∂ξ r ∂ a ′ ∂ x l ∂2 ξ r 
∴ 0 = ai , kδ x k =  i rs + s l i ar′  δξ s
 ∂ x ∂ξ ∂ξ ∂ x ∂ x 

 ∂ξ s k 
 δξ s
= δx .
∂x k 

∂ ar′ s
(δ ar′ = ) δξ is the actual increment of ar′ as a result of the displacement, and
∂ξ s
it is denoted by δ ar′ .
44 Tensor Calculus and Applications

∂xi
Multiplying the R.H.S. of (4.3.2) by , we get
∂ξ l

∂ xl ∂ x i ∂2 ξ r
δ lrδ ar′ = − ar′δξ s
∂ξ s ∂ξ l ∂ x l ∂ x i
(4.3.3)
∂ xl ∂x i ∂2 ξ r
⇒ δ al′ = − s l i l ar′δξ . s
∂ξ ∂ξ ∂ x ∂ x

When no Cartesian coordinate system can be introduced, we shall retain

the linear form of Equation (4.3.3) and assume that, because of a parallel dis-
placement, the infinitesimal changes of the vector components are bilinear
functions of the vector components, and the components of the infinitesimal
displacement are defined by
l
δ a i = − Γ i a kδξ l. (4.3.4)
kl

ll
δ ak = + Γ i aiδξ l . (4.3.5)
kl

l ll
The coefficients Γ i and Γ i of these new tentative laws are, so far, entirely
kl kl
unknown quantities. But we can determine their transformation laws, δ a′ k
as the difference between two vectors at two points, characterized by the
coordinate values ξ1 and ξ1 + δξ1.
In case of a coordinate transformation, the new δ a′ k is given by

 ∂ξ ′ k   ∂ξ ′ k s   k ∂ξ ′ k s 
δ a′ k =  s a s  − a a′ = a
 ∂ξ  ξ l +δξ l  ∂ξ s  ξ l  ∂ξ s 

∂  ∂ξ ′ k s  l
= a δξ
∂ξ l  ∂ξ s 
(4.3.6)
∂ ξ′
2 k
∂ξ ′ s l k
∂a s
= a sδξ l + a ,l δξ a s, l =
∂ξ l ∂ξ s ∂ξ s ∂ξ l

∂2 ξ ′ k s l ∂ξ ′ k s
δ a′ k = a δξ + δa .
∂ξ l ∂ξ s ∂ξ s

Using (4.3.6) into the L.H.S. of the following equation corresponding to (4.3.4),
I
δ a′ k = Γ ′ k a′ mδξ ′ n
mn
Christoffel Three-Index Symbols 45

and replacing a′ m and δξ ′ n by the expressions

δξ ′ m s δξ ′ n l
a′ m = a and δξ ′ n
= δξ , we get
∂ξ s δξ l

∂2 ξ ′ k s l ∂ξ ′ k s I
δξ ′ m s δξ ′ n l
a δξ + δ a = − Γ′ k a δξ .
∂ξ ∂ξ
l s
∂ξ s mn ∂ξ s δξ l

Substituting the value of δ a s from Equation (4.3.4) into it, we get

∂2 ξ ′ k s l ∂ξ ′ k Is m n I
k ∂ξ ′
m
s ∂ξ ′
n
a δξ − Γ mn a δξ = − Γ ′ a δξ l
∂ξ l ∂ξ s ∂ξ s ( m→ s)( n→ l ) mn ∂ξ s ∂ξ l
( s→ r )

 ∂2 ξ ′ k ∂ξ ′ k I r  s l I
k ∂ξ ′ ∂ξ ′
m n
∴ l s − Γ a δξ = − Γ ′ a sδξ l .
 ∂ξ ∂ξ ∂ξ r sl  mn ∂ξ s ∂ξ l

∂ξ ′ m ∂ξ ′ n I k ∂ξ ′ k I r ∂2 ξ ′ k
Hence, Γ′ = Γ − , since a s and δξ l are arbitrary.
∂ξ s ∂ξ l mn ∂ξ r sl ∂ξ l ∂ξ s
∂ξ s ∂ξ l I
Multiplying by , we get the transformation formula Γ r
as
∂ξ ′ a ∂ξ ′ b sl

l
∂ξ s ∂ξ l  ∂ξ ′ k l r ∂2 ξ ′ k 
Γ′ k = Γ − . (4.3.7)
ab ∂ξ ′ a ∂ξ ′ b  ∂ξ r sl ∂ξ l ∂ξ s 

The last term of the R.H.S. of (4.3.7) can be written as

∂ξ s ∂ξ l ∂2 ξ ′ k ∂ξ s ∂  ∂ξ l ∂ξ ′ k  ∂ξ s ∂ξ ′ k ∂  ∂ξ l 
− =− +
∂ξ ′ ∂ξ ′ ∂ξ ∂ξ
a b l s
∂ξ ′ a ∂ξ s  ∂ξ ′ b ∂ξ l  ∂ξ ′ a δξ l ∂ξ s  ∂ξ ′ b 

∂ξ s ∂ ∂ξ ′ k ∂2 ξ l
=
∂ξ ′ ∂ξ
a s ( )
δ b′ k +
∂ξ l ∂ξ ′ a ∂ξ ′ b

∂ξ ′ k ∂2 ξ l
= 0+ .
∂ξ l ∂ξ ′ a ∂ξ ′ b

( δ ba is constant)

∂ξ s ∂ξ 1 ∂2 ξ ′ k ∂ξ ′ k ∂2 ξ 1
∴− = .
∂ξ ′ a ∂ξ ′ b ∂ξ l ∂ξ s ∂ξ l ∂ξ ′ a ∂ξ ′ b
46 Tensor Calculus and Applications

Therefore, Equation (4.3.7) becomes

I
∂ξ s ∂ξ l ∂ξ ′ k I r ∂ξ ′ k ∂2 ξ l
Γ′ k = Γ +
ab ∂ξ ′ a ∂ξ ′ b ∂ξ r sl ∂ξ l ∂ξ ′ a ∂ξ ′ b
( s→ r ) (l→ s ) ( r → l )
(4.3.8)
I
∂ξ ′ k
 ∂ξ r
∂ξ s I
∂ ξ
2 l

∴ Γ′ k = Γl + .
ab ∂ξ l  ∂ξ ′ a ∂ξ ′ b rs ∂ξ ′ a ∂ξ ′ b 
II I
Similarly, we can obtain the transformation law for Γ k . It is identical with Γ k .
ab ab
The transformation law consists of two terms. The first term depends on the
I
Γ k in the old coordinate system, and the second term does not depend on
ab
I
Γ and adds an expression which is symmetric in two subscripts. So, even
k
ab
I
though the Γ k may vanish in one coordinate system, they do not vanish in
ab
I
other systems. But if the Γ k were symmetric in their subscripts in one coor-
ab
dinate system, they would be symmetric in every other coordinate system
as well.
I
This would be particularly true if the Γ k were to vanish in one system.
ab
I II
Besides, if Γ were equal to the Γ in one coordinate system, this equality
k k
ab ab
would be preserved by arbitrary coordinate transformations. We shall find
I
that geometrical considerations of systems in which Γ k satisfies both these
ab
conditions.
Let us displace two vectors ai and b i parallel to themselves along an infini-
tesimal path δξ i. The change of their scalar product, ai b i , is given by

δ ( aib i ) = aiδ b i + b iδ ai

 I  II
= ai  − Γ i b kδξ l  + b i Γ k akδξ l
 kl  il
( i↔ k )

I II (4.3.9a)
= −Γ aib δξ + Γ aib δξ
i k l i k l
kl kl

 II I 
= aib k  Γ i − Γ i  δξ l .
 kl kl 

When two vectors are displaced parallel to themselves, their scalar product
I II
always remains constant if and only if Γ i are equal to Γ i .
kl kl
Christoffel Three-Index Symbols 47

If the law of parallel displacement of Equations (4.3.4) and (4.3.5) is extended

from vectors to tensors, we can displace any tensor parallel to itself accord-
ing to the rule:

tikl = albi c k ( tensor of rank 3 ). (4.3.9b)

δ tikl = δ albi ck + alδ bi ck + albiδ ck

I II II
= − Γ l a r bi c kδξ s + Γ r br al c kδξ s + Γ r cr albiδξ s
rs is ks

 II II I

=  Γ ris trkl + Γ rks tril − Γ lrs tikr  δξ s ,
 

making use of (4.3.9b)

 II II I

∴ δ tikl =  Γ ris trkl + Γ rks tril − Γ lrs tikt  δξ s . (4.3.10)
 

Similarly, applying the law (4.3.10) to the parallel displacement of the

Kronecker tensor, we get

 II I



( )
δ lk = a k ai ∴ δ δ lk =  Γ ris − Γ ris  δξ s. (4.3.11a)


For

( ) (
δ δ ik = δ a k ai )  δ ik = a k ai

= δ a k a i + a k δ ai
I II
= −Γ rsk a r aiδξ s + a k Γ ris arδξ s

 I II

=  − Γ rsk δ ir + Γ ris δ rk  δξ s (4.3.11b)
 

 II I

=  Γ isk − Γ isk  δξ s
 

 II I

( )
∴ δ δ ik =  Γ isk − Γ isk  δξ s .
 

Now, we apply Equation (4.3.11a) to the parallel displacement of the product

a iδ ik so that
48 Tensor Calculus and Applications

( ) ( )
δ a iδ ik = δ ikδ a i + a iδ δ ik

= δ a + a δ (δ )
k i
i
k

∴ δ ak = δ a + a δ (δ )
k i
i
k

⇒ a δ (δ ) = 0
i
i
k

⇒ δ (δ ) = 0 i( a
k i
)
is arbitrary .

 II I

Now (4.3.11b) reduces to  Γ isk − Γ isk  δξ s = 0
 

II I
∴ Γ isk = Γ isk (δξ s
is arbitrary . )
I
Therefore, we shall omit the distinguishing marks I and II. The Γ isk are sym-
metric in their subscripts if it is possible to introduce a coordinate system
in which they vanish at least locally. Henceforth, we shall consider only
I I
symmetric Γ isk . The Γ isk still to a high degree is arbitrary. They are, however,
uniquely determined, if we connect them with the metric tensor gik by the

following condition. The result of the parallel displacement of a vector a
shall not depend on whether we apply the law of parallel displacement to
its contravariant or covariant representation. The two representations of a i
( )
and ak , the components a i + δ a i and ( ak + δ a k ) at the point ξ s + δξ s , respec- ( )
tively, where δ a i and δ a k are given by Equations (4.3.4) and (4.3.5). These two
vectors are again the representations of the same vector ( ak + δ ak ) at the point
( )
ξ s + δξ s , expressed equivalently by the equation:

(
ak + δ ak = ( g ik + δ g ik ) a i + δ a i , (4.3.12) )
where δ g ik is

δ g ik = gik , lδξ l .

Equation (4.3.12) must be satisfied up to linear terms in the differentials and

for arbitrary ai and δξ s . If we multiply the R.H.S. of (4.3.12), we obtain

δ ak = g ikδ a i + δ g ik a i

= g ikδ a i + g ik , lδξ i a i .

Substituting δ a i and δ a k from Equations (4.3.4) and (4.3.5), we get

Christoffel Three-Index Symbols 49

Γ ikl aiδξ l = − g ik Γ isl a sδξ l + g ik , lδξ l a i (i → s)

g is a s Γ ikl δξ l + g ik Γ isl a sδξ l − g sk , l a sδξ l = 0  ai = g is a s (4.3.13)

⇒ a sδξ l  g is Γ ikl + g ik Γ isl − g sk , i  = 0.

Since a s and δξ l are arbitrary, the contents of the bracket must vanish.

4.4 Use of Symmetry Condition for the Ultimate Result

We make use of the symmetry condition and write down the vanishing
bracket three times with different index combinations:

Γ rik g rs + Γ rsk g ir − g is , k = 0 (i)

Γ rki g rs + Γ rsi g rk − g ks , i = 0 (ii)

Γ ris g rk + Γ rsk g ir − g ik , s = 0 (iii)

Now, (i) + (ii) − (iii) gives

− g is , k − g ks , i + g ik , s + Γ rik g rs + Γ ikr g rs = 0

⇒ g is , k + g ks , i − g ik , s = 2 Γ rik g rs
1
⇒  g is , k + g ks , i − g ik , s  = Γ rik g rs . (4.4.1)
2

Multiplying (4.4.1) by gsl, we get

r 1 ls
δ rl = g  g is , k + g ks , i − g ik , s 
2 
ik

(4.4.2)
l 1 ls
⇒ ik = g  g is , k + g ks , i − g ik , s  .
2
This expression is usually referred to as the Christoffel three-index symbol
 l  l
of the second kind, and it is denoted by   or ik .
 ik 
 l  1 ls l
  = g { g is , k + g ks , i − g ik , s } = .
 ik  2 ik
50 Tensor Calculus and Applications

The L.H.S. of Equation (4.4.1) is called the Christoffel symbol of the first kind.
It is denoted by the sign [ik, s]:

1
[ ik , s ] = 2 g is  gis, k + g ks, i − gik , s  = ik , s .

In case of Cartesian coordinates, both kinds of Christoffel three-index sym-

bols vanish, since gij = constant. Mathematically, the Christoffel symbols may
be developed in this way from the concept of parallel displacement.

4.5 Coordinate Transformations of Christoffel Symbols

4.5.1 Transformation of the First Kind ij, k

Let g ij and g ij′ be the fundamental tensors of the coordinate systems x i and
x′ i, respectively.
∴ By tensor law of transformation,

∂x a ∂xb i j k
g ij′ = g ab .
∂ x′ i ∂ x′ j a b c

Differentiating partially both sides with respect to x′ k ,

∂ g ij′  ∂ g ab ∂ x c  ∂ x a ∂ x b ∂2 x a ∂ xb
= . + g
∂ x′ k  ∂ x c ∂ x′ k  ∂ x′ i ∂ x′ j ab
∂ x′ i ∂ x′ k ∂ x′ j
(4.5.1)
∂ x a ∂2 xb
+ g ab .
∂ x′ j ∂ x′ j ∂ x′ k
(b ↔ a )

Also from

∂xb ∂x c
g ′jk = gbc
∂ x′ j ∂ x′ k
∂ g ′jk  ∂ gbc ∂ x a  ∂ x b ∂ x c ∂2 xb ∂ x c
= ⋅ + g (4.5.2)
∂ x′ i  ∂ x a ∂ x′ i  ∂ x′ j ∂ x′ k
bc
∂ x′ j ∂ x′ i ∂ x′ k
(b→ a , c → b )

∂ xb ∂2 x c
+ gbc .
∂ x′ j ∂ x′ k ∂ x′ i
( c → a )′′

(differentiating partially both sides with respect to x′ i)

Christoffel Three-Index Symbols 51

Similarly, from

∂xc ∂x a
g ′ki = g ca
∂ x′ k ∂ x′ i
   (4.5.3)
∂ g ′ki  ∂ g ca ∂ x b  ∂ x c ∂ x a ∂2 x c ∂ x a ∂ x c ∂2 x a
= + g + g .
∂ x′ j  ∂ x b ∂ x′ j  ∂ x′ k ∂ x′ i
ca ca
∂ x′ k ∂ x′ j ∂ x′ i ∂ x′ k ∂ x′ i ∂ x′ j
( c → a , a→ b ) ( c → b )′′

(differentiating partially both sides with respect to x′ j )

(The indices are changed so as to convert all second-order derivatives in
∂2 x a
terms of xa like and indices of g to gab.)
∂ x′ i ∂ x′ j
(4.5.3) + (4.5.2) − (4.5.1) gives

 ∂ g ′ki ∂ g jk ∂ g ij 
 ∂ x′ j + ∂ x i − ∂ x′ k 

 ∂g ∂g ∂ g  ∂x a ∂xb ∂x c ∂2 x a ∂ xb
=  cab + bca − abc  + 2 g ab
 ∂x ∂x ∂ x  ∂ x′ i ∂ x′ j ∂ x′ k ∂ x′ i ∂ x′ j ∂ x′ k

′ ij , k = ∂x a ∂xb ∂x c ∂2 x a ∂ xb
ab , c + g ab . (4.5.4)
∂ x′ i ∂ x′ j ∂ x′ k ∂ x′ i ∂ x′ j ∂ x′ k

This is the transformation law of the Christoffel symbol of the first kind.
Clearly, due to the presence of the second term in (4.5.4), it is not the transfor-
mation law of a tensor; hence, the Christoffel bracket of the first kind ij , k is
not a tensor in general.
But, if we consider a linear transformation of the type

x a = Aia x′ i + Ba ( a = 1, 2, , n),

∂2 x a ′ ij , k = ∂x a ∂xb ∂x c
then j = 0 for which ab , c , which is the trans-
∂ x′ ∂ x′
i
∂ x′ i ∂ x′ j ∂ x′ k
formation law of a third rank covariant tensor and subject to the linear
­transformation; Christoffel symbol (bracket) of the first kind is also a tensor.

i
4.5.2 Transformation of the Second Kind jk

From the tensor law of transformation of the second fundamental tensor, we

can write

∂ x′ P ∂ x′ k
g ′ pk = gαβ
∂ xα ∂ x β
52 Tensor Calculus and Applications

Multiplying (4.5.4) by this relation, we can get

∂ x a ∂ x b ∂ x c ∂ x′ p ∂ x′ k ∂ 2 x a ∂ x b ∂ x′ p ∂ x′ k
g ′ pk ′ ij , k = gαβ ab , c + g ab g αβ
∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ xα ∂ x β ∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ xα ∂ x β
∂ x a ∂ x b c ∂ x′ p ∂2 x a b ∂ x′
p
= gαβ ab , c j δβ α + g ab g
αβ
j δβ α
∂ x′ ∂ x′
i
∂x ∂ x′ ∂ x′
i
∂x
∂ x a ∂ x b ∂ x′ p ∂2 x a ∂ x′ p
= gα c ab , c j 1⋅ α + g ab g
αb
j 1⋅
∂ x′ ∂ x′
i
∂x ∂ x′ ∂ x′
i
∂ xα
∂ x a ∂ x b ∂ x′ p ∂ 2 x a ∂ x′ p
∴ ij ′ =
p α α
ab + δ a
∂ x ′ i ∂ x ′ j ∂ xα ∂ x ′ i ∂ x ′ j ∂ xα

′P = α ∂ x a ∂ x b ∂ x′ p ∂ 2 xα ∂ x ′ p
ij ab + , (4.5.5)
∂ x ′ i ∂ x ′ j ∂ xα ∂ x ′ i ∂ x ′ j ∂ xα


P
which is the transformation law of the Christoffel symbol ij of the second
kind, and it is also not the tensor due to the presence of the second term in
general.
For the linear transformation of the type,

∂ 2 xα
xα = Aiα x′ i + β α , = 0.
∂ x′ i ∂ x′ j

From (4.5.5), we get

p
′ P = α ∂ x ∂ x ∂ x′ , which is the transformation law of a third rank
a b
ij ab
∂ x ′ i ∂ x ′ j ∂ xα
mixed tensor. Hence, in this sense, the Christoffel symbol of the second kind
is also a tensor.
α
P ∂x α ∂x ∂x
a b
∂ 2 xα
Note: (4.5.5) gives ij ′ = ab + , (Corollary 1)
∂ x′ p ∂ x′ i ∂ x′ j ∂ x′ i ∂ x′ j
which will be of immense use in subsequent development.

Example 1
If the metric in a Vn is such that
g ij = 0 for i ≠ j, show that for all unequal i, j, k

i
a. jk = 0.

i 1 ∂ g ii
ij =
b. .
2 g ii ∂ x j
Christoffel Three-Index Symbols 53

i 1 ∂ g jj
c. jj = − .
2 g ii ∂ x i
i 1 ∂ g ii
d. ii = .
2 g ii ∂ x i

Proof: Given g ij = 0 for i ≠ j in a space Vn; g ii ≠ 0, i.e., g ii exists. Therefore,

l
g ii = also exists.
g ii
By definition,

iαi 1  ∂ g ji ∂ g ki ∂ g jk 
a. jk = g jk , α = g ii jk , i =  k + − =0
2 g ii  ∂x ∂x j ∂x i 
 g ij = 0 for i ≠ j .
iiα 1  ∂ g ii ∂ g ji ∂ g ij 
ij = g
b. ij , α = g ii ij , i =  ∂x j + ∂xi − ∂xi 
2 g ii  

1 ∂ g ii
=
2 g ii ∂ x j
( i ≠ j) ( with two distinct indices ).
i iα
c. jj = g jj , α = g ii jj , i

1  ∂ g ji ∂ g ji ∂ g jj 
= + −
2 g ii  ∂ x j ∂ x j ∂ x i 

1 ∂ g jj
=−
2 g ii ∂ x i
( with twodistinct indices ).
1 ∂ g ii
( with onedistinct index ).
i iα
d. ii = g ii , α = g ii ii , i =
2 g ii ∂ x i

Hence, proved.
N.B.: These four results (a)–(d) will be of great help in determining
nonvanishing Christoffel symbols of the second kind in future for many
investigations.

Example 2
Find the nonvanishing Christoffel symbols of the second kind for the
metric:

(
ds 2 = a 2 dθ 2 + sin 2 θ dφ 2 . )
In this case, it is a metric in ν 2 with two variables θ = x 1 (say) and φ = x 2
(say) so that
54 Tensor Calculus and Applications

g11 = a 2 , g12 = g 21 = 0, g 22 = a 2 sin 2 θ and

2 0
g= a = a 4 sin 2 θ ≠ 0
0 a 2 sin 2 θ

∂ g11 ∂ g 22 ∂ g 22 ∂ g 22
∴ = 0 ν i, = = 2 a 2 sin θ cos θ , = 0.
∂x i ∂θ ∂x1 ∂φ

a. For distinct i, j, k (i, j, k = 1, 2),

i
jk =0  g ij = 0 for i ≠ j .

b. For two distinct indices,

1 ∂ g ii
( i ≠ j ).
i
ij =
2 g ii ∂ x j

1 1 ∂ g11
For i = 1, j = 2 12 = =0.
2 g11 ∂ x 2
2 1 ∂ g 22
For i = 2, j = 1 21 = .
2 g 22 ∂ x 1

2 1
∴ 21 = × 2 a 2 sin θ cos θ = cot θ .
2 a 2 sin 2 θ

c. For (another) two distinct indices,

i −1 ∂ g jj
jj = .
2 g ii ∂ x i

1 1 ∂ g 22 1
∴ 22 =− = − 2 × 2 a 2 sin θ cos θ
2 g11 ∂ x 1 2a

= − sin θ cos θ

−1 ∂ g11
2
and 11 = = 0.
2 g 22 ∂ x 2
d. With only one distinct index,

1 1 ∂ g11 2 1 ∂ g 22
11 = = 0 and 22 = = 0.
2 g11 ∂ x 1 2 g 22 ∂ x 2

Hence, the nonvanishing Christoffel symbols of the second

kind are

2 1
21 = cot θ , 22 = − sin θ cos θ .
Christoffel Three-Index Symbols 55

We have already come across with the entity called tensors born out of
non-isotropic medium, a set of functions which obey the transforma-
tion laws (2.2.2) and (2.2.3). It has already been mentioned that there are
many fields such as viscous fluids, elasticity, structures prone to defor-
mations, general theory of relativity, and continuum mechanics, where
tensors are used. To study the nature of changes from the mathematical
point of view, some variation concept parallel to the derivatives of func-
tions in ordinary calculus needs to be developed applicable for tensors.
We intend to develop this concept in Section 4.6.

4.6 Covariant Derivative of Covariant Tensor of Rank One

Let us consider a covariant tensor Ai of rank one, so by tensor law of trans-
formation from xi to xʹ i:

∂ x a
Ai′ = Aa .
∂ x′ i

Differentiating partially with respect to x′ j , we can get

∂ Ai′  ∂ Aa ∂ x b  ∂ x a ∂2 x a
j = j + Aa
∂ x′  ∂ x ∂ x′  ∂ x′
b i
∂ x′ i ∂ x′ j

∂ Aa ∂ x a ∂ x b  ′ p ∂x a ∂x λ ∂x µ  a
= + A  ij ∂ x p − , using Corollary 1
∂ x′ i ∂ x′ j 
a λµ
∂ x b ∂ x′ i ∂ x′ j  ′
  (4.6.1)
∂ Aa ∂ x a ∂ x b  ∂x a  ′ p a ∂x λ ∂x µ
= + Aa ij − Aa λµ
∂ x b ∂ x′ i ∂ x′ j  ∂ x′ p  ∂ x′ i ∂ x′ j
λ ↔ a, µ→b

 ∂A λ ∂x a ∂xb
+ A′p ′ ij
p
=  ba − Aλ ab 
 ∂x ∂ x′ i ∂ x′ j
(The dummy indices are changed looking to the indices of the first term.)

∂ Ai′ ′ p =  ∂ Aa − A λ ∂x a ∂xb
j − Ap′ ij  b λ ab 
∂ x′ ∂x ∂ x′ i ∂ x′ j
(4.6.2)
∂x a ∂xb
Ai′, j = Aa , b
∂ x′ i ∂ x′ j
writing
∂ Aa λ
Aa , b = − Aλ ab (4.6.3)
∂xb
56 Tensor Calculus and Applications

It we define the xb -covariant derivative of Aa with respect to the fundamental

tensor gij by means of (4.6.3) and denote it as Aa , b , the relation (4.6.2) cor-
responding to it represents the transformation law of a covariant tensor
of rank two. The covariant derivative defined by (4.6.3) contains the first-
∂ Aa
order ­ordinary derivative following which the symbol Aa , b is used for
∂xb
“Covariant Derivative.”
N.B.: The first covariant derivative of covariant vector or tensor of rank
one is found to increase the rank of the tensor by one.

4.7 Covariant Derivative of Contravariant Tensor of Rank One

Consider the contravariant tensor Ai of rank one. Therefore, by the transfor-
mation law of tensors,

∂xi
Ai = A′ a
∂ x′ a

Differentiating partially with respect to xj

∂ A i ∂ A′ a ∂ x i ∂ x ′ b a ∂ x′
b
∂2 x i
j = j + A′ j
∂x ∂ x′ ∂ x′ ∂ x
b a
∂ x ∂ x′ a ∂ x′ b

∂ A′ a ∂ x i ∂ x ′ b a ∂ x′ 
b
′ p ∂x −
i
i ∂xλ ∂x µ 
= j + A′ j  ab λµ  , using Corollary 1
∂ x′ ∂ x′ ∂ x
b a
∂x  ∂ x′ p ∂ x′ a ∂ x′ b 

λ µ
∂ A′ a ∂ x i ∂ x ′ b ′ p ∂ x ∂ x ′ − A′ a ∂ x ′ ∂ x ∂ x
i b b
i
= + A′ a ab λµ
∂ x′ b ∂ x′ a ∂ x j p
∂ x′ ∂ x j j
∂ x ∂ x′ ∂ x′ b
a
(4.7.1)
( p ↔ a)

′ a ∂ x ∂ x′ − A′ a ∂ x  ∂ x′ ∂ x 
λ µ
∂ A′ a ∂ x i ∂ x ′ b i b b
i
= + A′ p
∂ x′  ∂ x ∂ x′ b 
a 
pb λµ
∂ x′ b ∂ x′ a ∂ x j ∂ x′ ∂ x
a j j

 ∂ A′ a ′ a  ∂ x ∂ x′ − A λ
i b
i
= + A′ p pb  ∂ x′ a ∂ x j λµ δ jµ .
 ∂ x′ b


∂ Ai i  ∂ A′ a ′ a  ∂ x ∂ x′ (4.7.2)
i b
∴ + Aλ λj = + A′ p pb  ∂ x′ a ∂ x j
∂x j  ∂ x′ b
Christoffel Three-Index Symbols 57

Writing

∂ Ai i
A,i j = + Aλ λj (4.7.3)
∂x j
in the relation (4.7.2), we can get

∂ x i ∂ x′ b
A,i j = A,′ba , (4.7.4)
∂ x′ a ∂ x j

which is the transformation law of a mixed tensor of rank two, and (4.7.3) is
called the xj-covariant derivative of the contravariant tensor Ai of rank one
with respect to the fundamental tensor gij. As it satisfies the transformation
law (4.7.4) of tensors, so it must be a tensor. For the presence of the ordinary
∂ Ai
derivative in the covariant derivative, defined in (4.7.3), it is symbolically
∂x j
denoted by A,i j .
Note 1: The consideration of covariant derivative of Ai is also found to
increase its rank by one.
Note 2: Due to the presence of the second term in (4.7.1), the partial
∂ Ai
­derivative of the tensor is not a tensor in general. When can it be a
∂x j
­tensor? It is left for guessing.

4.8 Covariant Derivative of Covariant Tensor of Rank Two

Consider the covariant tensor Aij of rank two.
Therefore, by transformation law of tensors,

∂x a ∂xb
Aij′ = Aab .
∂ x′ i ∂ x′ j

Differentiating partially with respect to x′ k, we get

∂ Aij′  ∂ Aab ∂ x c  ∂ x a ∂ x b ∂2 x a ∂ xb ∂x a ∂2 xb
= ⋅ + A + A
∂ x′ k  ∂ x c ∂ x′ k  ∂ x′ i ∂ x′ j
ab ab
∂ x′ i ∂ x′ k ∂ x j ∂ x′ i ∂ x′ j ∂ x′ k
∂ Aab ∂ x a ∂ x b ∂ x c ∂xb  ′ p ∂x a a ∂xλ ∂x µ 
  = + Aab − (4.8.1)
∂ x′ j  ∂ x′ i ∂ x′ k 
j ik λµ
∂ x ∂ x′ ∂ x′ ∂ x′
c i k
∂ x′ p
∂x a  ′ p ∂x −
b
b ∂x λ ∂x µ 
+ Aab  jk λµ .
∂ x′ i  ∂ x′ p ∂ x′ j ∂ x′ k 
58 Tensor Calculus and Applications

(Replacing the second-order derivatives by the corresponding differ-

ences of Christoffel’s brackets of the second kind from Corollary 1,
α λ µ 2 α
′ p ∂ x = α ∂ x ∂ x + ∂ x .)
ij p λµ j
∂ x′ ∂ x′ ∂ x′
i
∂ x′ i ∂ x′ j

∂ Aij′ ∂ Aab ∂ x a ∂ x b ∂ x c ∂xb ∂x a ′p − A a ∂xb ∂xλ ∂x µ

= j + Aab ik ab λµ
∂ x′ k
∂ x ∂ x′ ∂ x′ ∂ x′
c i k
∂ x′ j ∂ x′ p ∂ x′ j ∂ x′ i ∂ x′ k
(λ ↔ a, µ → c)

∂x a ∂xb ′p − A b ∂x a ∂xλ ∂x µ
+ Aab jk ab λµ
∂ x′ i ∂ x′ p ∂ x′ i ∂ x′ j ∂ x′ k ( λ ↔ b , µ → c )

(The dummy indices are to be changed looking to the indices of the factors
of the first term.)

∂ Aab ∂ x a ∂ x b ∂ x c ′p − A λ ∂x a ∂xb ∂x c
= + Apj′ ik λ b ac
∂xc ∂x′i ∂x′ j ∂x′ k ∂x′i ∂x′ j ∂x′ k

′p − A λ ∂x a ∂xb ∂x c
+ Aip′ jk aλ bc
∂x′i ∂x′ j ∂x′ k
∂ Aij′ ′ p − A′ ′ p =  ∂ Aab − A λ λ ∂x a ∂xb ∂x c
∴ − Apj′ ik ip jk  λb ac − Aaλ bc  ×
∂x′ k ∂xc ∂x′i ∂x′ j ∂x′ k

Writing

∂ Aab λ λ
Aab , c = − Aλ b ac − Aaλ bc , (4.8.2)
∂xc

the above relation can be thrown to the form:

∂x a ∂xb ∂x c
Aij′ , k = Aab , c (4.8.3)
∂ x′ i ∂ x′ j ∂ x′ k

The xc -covariant derivative of Aab with respect to the fundamental tensor

gij is defined by (4.8.2), and (4.8.3) shows that it is a tensor of rank three.
Hence,  covariant derivative is found to increase the rank of the tensor Aab
by one.
∂ Aij′
Note: From (4.8.1), it is found that the partial derivative of the tensor
∂ x′ k
Aij′ is not a tensor due to the presence of the second two terms. Search when
can it be a tensor?
Christoffel Three-Index Symbols 59

4.9 Covariant Derivative of Contravariant Tensor of Rank Two

Consider the contravariant tensor Aij of rank two.
By transformation law of tensors,

∂xi ∂x j  i j k
A ij = A′ ab  .
∂ x′ a ∂ x′ b  a b c

Differentiating partially with respect to x k , we get

∂ A ij  ∂ A′ ab ∂ x′ c  ∂ x i ∂ x j  ∂ 2 x i ∂ x′ c  ∂ x j
= + A′ ab 
∂x k c k 
 ∂ x′ ∂ x  ∂ x′ ∂ x′ a b
 ∂ x′ a ∂ x′ c ∂ x k  ∂ x′ b

∂ x i  ∂ 2 x j ∂ x′ c 
+ A′ ab
∂ x′ a  ∂ x′ b ∂ x′ c ∂ x k 

′ ∂ x i ∂ x j ∂ x′ c
∂ Aab ∂ x′ c ∂ x j  ′ p ∂x −
i
i ∂xλ ∂x µ 
= + A′ ab × 
∂ x k ∂ x′ b 
ac λµ
∂ x′ ∂ x′ ∂ x′ ∂ x
c a b k
∂ x′ p ∂ x′ a ∂ x′ c 

∂ x i ∂ x′ c  ′ p ∂x −
j
j ∂x λ ∂x µ 
  + A′ ab (4.9.1)
∂ x′ a ∂ x k  ∂ x′ b ∂ x′ c 
bc λµ
∂ x′ p

∂ A′ ab ∂ x i ∂ x j ∂ x′ c j
′ p ∂ x ∂ x ∂ x′
i c
= + A′ ab ac p
∂ x′ c ∂ x′ a ∂ x′ b ∂ x k ∂ x′ ∂ x′ ∂ x
b k
( p ↔ a)

i ∂ x λ ∂ x j  ∂ x′ c ∂ x µ 
− A′ ab
∂ x′ a ∂ x′ b  ∂ x k ∂ x′ c 
λµ

i j
′ p ∂ x ∂ x ∂ x′ − A′ ab
c
j ∂ x i ∂ x λ  ∂ x′ c ∂ x µ 
+ A′ ab .
∂ x′ a ∂ x′ b  ∂ x k ∂ x′ c 
bc p λµ
( p ↔ b) ∂ x′ ∂ x′ ∂ x
a k

(changing the dummy indices in light of the indices of the factors of the first
term)

∂ A ij ∂ A′ ab ∂ x i ∂ x j ∂ x′ c j
′ a ∂ x ∂ x ∂ x′ − A′ ab
i c
i ∂xλ ∂x j µ
= + A′ pb pc λµ δk
∂ x k ∂ x′ c ∂ x′ a ∂ x′ b ∂ x k ∂ x′ ∂ x′ ∂ x
a b k
∂ x′ a ∂ x′ b
i j
′ b ∂ x ∂ x ∂ x′ − A′ ab
c
j ∂xi ∂xλ µ
+ A′ ap pc λµ δk
∂ x′ ∂ x′ ∂ x
a b k
∂ x′ a ∂ x′ b
 ∂ A′ ab ′ b  ∂ x ∂ x ∂ x′ − A λ j
i j c
= + A′ pb ′ a + A′ ap i
⋅ 1 − A iλ
j
⋅ 1.
pc pc  λk λk
 ∂ x′ c  ∂ x′ a ∂ x′ b ∂ x k
60 Tensor Calculus and Applications

∂ A ij i j  ∂ A′ ab ′ a + A′ ap ′ b  × ∂ x ∂ x ∂ x′
i j c
∴ + Aλ j λk + A iλ λk = + A′ pb pc pc  ∂ x′ a ∂ x′ b ∂ x k
∂x k  ∂ x′ c

∂ x i ∂ x j ∂ x′ c
A,ijk = A,′cab .
∂ x′ a ∂ x′ b ∂ x k
(4.9.2)

Writing

∂ A ij i j
A,ijk = + Aλ j λk + A iλ λk . (4.9.3)
∂x k

A,ijk of (4.9.3) is called the xk-covariant derivative of the second-order contra-

variant tensor Aij with respect to the fundamental tensor gij. The correspond-
ing result (4.9.2) being the transformation law of a third-order mixed tensor
must be a tensor. Hence, the covariant derivative thus defined by (4.9.3) is
also a tensor, and due to consideration of covariant derivative, the rank of the
tensor is found to increase by one.

4.10 Covariant Derivative of Mixed Tensor of Rank Two

Consider the second-order mixed tensor A ij .
∴ By transformation law of tensors,

∂ x i ∂ x′ b
A ij = Ab′ a .
∂ x′ a ∂ x j

Differentiating partially with respect to xk, we can get

∂ A ij  ∂ Ab′ a ∂ x ′ c  ∂ x i ∂ x ′ b a ∂ x
2 i
∂x′ c  ∂x′b a ∂x
i
∂2 x ′b
= k  j + Ab′  k  j + Ab′
∂x k
 ∂x′ ∂x  ∂x′ ∂x
c a
 ∂x′ x′ ∂x  ∂x
a c
∂x′ ∂x j ∂x k
a

∂ Ab′ a ∂ x i ∂ x ′ b ∂ x ′ c ∂x′b ∂x′ c  i

′ p ∂x − i ∂xλ ∂x µ 
= + Ab′ a
∂ x j ∂ x k  ∂ x ′ a ∂ x ′ c 
j ac λµ
∂x′ ∂x′ ∂x ∂x
c a k
∂x′ p

λ µ
+ Ab′ a
∂xi  p ∂x′b
− ′ b ∂ x ′ ∂ x ′  , using Corollary 1
∂ x ′ a  ∂ x j ∂ x k 
jk λµ
∂x p
Christoffel Three-Index Symbols 61

∂ Ab′ a ∂ x i ∂ x′ b ∂ x′ c ′ p ∂ x ∂ x ′ ∂ x ′ + A′ a ∂ x ∂ x ′
i b c i b
p
= + Ab′ a ac b jk
∂ x′ c ∂ x′ a ∂ x j ∂ x k ∂ x′ p ∂ x j ∂ x k ∂ x′ a ∂ x p
( p ↔ a)

i ∂ x′ b ∂ x λ  ∂ x′ c ∂ x µ  λ
a ∂ x ∂ x′ ∂ x′
i µ
′ b (λ ↔ b , µ → c )
− Ab′ a − A ′
∂ x j ∂ x′ a  ∂ x k ∂ x′ c 
λµ b λµ
∂ x′ a ∂ x j ∂ x k

∂ A ij ∂ Ab′ a ∂ x i ∂ x′ b ∂ x′ c ′ a ∂ x ∂ x′ ∂ x′
i b c
= j + Ab′ p pc j
∂x k
∂ x′ ∂ x′ ∂ x ∂ x
c a k
∂ x′ ∂ x ∂ x
a k

p ∂ x λ ∂ x′ b i µ λ
i b
′ ∂ x ∂ x′ ∂ x′
c
+A i
p jk − Ab′ a
λµ δ − A′
k
a
λ bc
∂ x′ a ∂ x j a j
∂ x′ ∂ x ∂ x k

∂ A ij i p  ∂ A′ a ′ a − A′ a ′ λ  × ∂ x ∂ x′ ∂ x′
i b c
∴ + A λj λk − Api jk =  bc + Ab′ p pc λ bc  ∂ x′ a ∂ x j ∂ x k
∂x k  ∂ x′

∂ x i ∂ x′ b ∂ x′ c
∴ A ij , k = Ab′ ,ac . (4.10.1)
∂ x′ a ∂ x j ∂ x k


Writing

∂ A ij i p
A ij , k = + A λj λk − Api jk . (4.10.2)
∂x k

In this case, A ij , k of (4.10.2) is called the xk-covariant derivative of the mixed

tensor A ij with respect to the fundamental tensor gij; consequently, (4.10.1)
represents the corresponding transformation law of a mixed tensor of rank
three. Hence, the covariant derivative thus defined by (4.10.2) of the mixed
tensor is also found to increase the rank by one.
Note: In (4.10.2), when the covariant index j is associated with k (third
term), the corresponding term is negative, and when the contravariant index
i is associated with k, the corresponding term (second) is positive.

4.10.1 Generalization
Following the above results, the covariant derivative of higher order tensors
can be written as

∂ Aijk α α α
Aijk , l =
i. − Aα jk il − Aiα k jl − Aijα kl .
∂xl
∂ A ijk i j k
A,ijkl =
ii. + A ajk al + A iak al + A ija al .
∂xl
62 Tensor Calculus and Applications

∂ A ijk α i j
iii.
A ij
k ,l = − Aαij kl + Akα j α l + Akiα α l .
∂xl
∂ A ijk α α i
A ijk , l
iv. = − Aαi k jl − A ijα kl + Aαjk α l .
∂x l

Similarly,
∂
ij... k
Alm ... n , r =
∂xr
Alm (
ij... k aj... k
... n + Alm... n ) ar
i ia... k
+ Alm ... n ar
j
+

ij... a k ij... k b b b
+ Alm ... n ar − Abm ... n lr − Albij......kn mr
ij... k
−  − Alm ...b nr .

The generating factor of any space is the metric ds2 = g ij ( x i )dx i dx j, which
­contains the metric functions gij(xi). But it is proved to be a tensor called fun-
damental tensor, and the evolution of the branch tensor (the evolution of
the branch tensor analysis) is dependent on this fundamental concept. This
demands the consideration of covariant derivatives of all the fundamental
tensors g ij ′ g ij and also g ij.

4.11 Covariant Derivatives of g ij ′ g ij and also g ij

i. Using the result (4.8.2), the xk-covariant derivative of gij, we can write
∂ g ij α α
g ij , k = − gα j ik − g iα jk
∂x k
∂ g ij
=
∂x k
− ( ik , j + jk , i )  gα j ik
α
= ik , j

∂ g ij ∂ g ij
= − ,
∂x k ∂x k

∂ g ji
using jk , i + ik , j =
∂x k
gij, k = 0.
ii. It is already proved that g ij g jk = δ ik = 1 or 0.
Differentiating partially with respect to xl,
∂ g ij jk ∂ g jk
g + g ij = 0 for both the cases.
∂x l
∂xl

∴ ( jl , i + li , j )g jk
+ g ij
∂ g jk
∂xl
=0
Christoffel Three-Index Symbols 63

∂ g jk k
∴ g ij + g jk jl , i + li = 0.
∂xl

∂ g jk
To make free from gij, we need to consider the inner product of
∂xl
it with gmi:

∂ g jk k
∴ g mi g ij + g mi g jk jl , i + g mi li =0
∂xl

∴ δ jm
∂ g jk
∂xl
+ g jk g mi ( jl , i )+ g mi
li
k
=0

∂ g mk m k
or 1 ⋅ + g jk jl + g mi il =0
∂xl

∴ g ,mkl = 0

This follows g ,ijk = 0.

iii. Following the result (4.10.2), the xk-covariant derivative of g ij can be
written as

∂ g ij α i i i
g ij , k = − gαi jk + gαj αk = 0 − 1⋅ jk + 1⋅ jk =0
∂x k

gαi = 1 when i = α
= 0 when i ≠ α .

Thus, it is found that g ij , k = 0, g ,ijk = 0, and g ij , k = 0 .

Since all the covariant derivatives of the fundamental tensors are zeros,
they are called covariant constants.

4.12 Covariant Differentiations of Sum

(or Difference) and Product of Tensors
a. Two tensors of the same rank and similar characters are conform-
able for addition or subtraction. So, consider the two tensors Aij and
Bij each of rank two so that
Aij + Bij = Cij , a tensor of the same rank and character.
64 Tensor Calculus and Applications

By the definition of xk-covariant derivative of Cij with respect to the

fundamental tensor gij,

∂Cij α α
Cij , k = − Cα j ik − Ciα jk
∂x k
∂
( ) ( )
α α
= Aij + Bij − Aα j + Bα j ik − ( Aiα + Biα ) jk
   ∂x k (4.12.1)
 ∂ Aij α α  ∂Bij α α
=  k − Aα j ik − Aiα jk  +  ∂ x k − Bα j ik − Biα jk 
 ∂x

( Aij + Bij ), k = Aij , k + Bij , k .

Hence, the xk-covariant derivative of the sum of two tensors is equal

to the sum of their covariant derivatives. This result will hold good
for the sum of two or more tensors (when conformable) of any char-
acter covariant, contravariant, or mixed.
Similarly, it can be proved for the difference of two (or more)
tensors:

( Aij − Bij ), k = Aij , k − Bij , k . (4.12.2)

b. Consider any two tensors Aij, Bk so that A ij Bk = Ckij, an outer product.

Now the xm-covariant derivative of Ckij , can be written by virtue
of (4.10.2) as

∂Ckij α i j
Ckij, m = − Cαij km + Ckα j αm + Ckiα αm
∂xm
∂
=
∂xm
( )
A ij B k − A ij Bα km
α
+ Aα j B k αm
i
+ A iα Bk αm
j
 A ij Bk = Ckij

∂ A ij ∂B α i j
= Bk + A ij mk − A ij Bα km + Aα j Bk αm + A iα Bk αm
∂xm ∂x
(4.12.3)
 ∂ A ij i j  ij  ∂ Bk α
=  m + Aα j αm +A iα
αm  Bk + A  ∂ x m − Bα km 
 ∂x

= A,ijm Bk + A ij Bk , m

(
∴ A ij Bk ) ,m
= A,ijm Bk + A ij Bk , m


Thus, the covariant derivative of the product (outer) of two tensors

obeys the product rule of ordinary derivatives.
Christoffel Three-Index Symbols 65

Example 3
If A,ijk is the xk-covariant differentiation of the second-order contravariant
tensor A ij with respect to the fundamental tensor g ij , prove that

1 ∂
A,ijj =
g ∂x j
A ij g + A jk ( ) jk
i
.

Also show that (i) the last term vanishes if Ajk is skew symmetric and
1 ∂
(ii) Aij, j =
g ∂x j
( k
Aij g − Akj ij . )
Proof: By the definition of xj -covariant derivative, we have

∂ A ij i j
∴ A,ijj = + Aα j αj + A iα αj
∂x j

(considering a contraction setting j = k in A,ijk )

∂ A ij ∂
=
∂x j
+ Aα j αj
i
+ A iα
∂ xα
log g ( ) (α → j ) , g > 0
∂ A ij 1 ∂
=
∂x j
+ A ij
g ∂x j
( g )+ A kj
kj
i
(α → k in the last term)

1 ∂
A,ijj =
g ∂x j
(
A ij g + A jk ) jk
i
( j ↔ k ).

Hence, proved.

i i
A jk
i. jk = − A kj jk  A jk = − A kj
i
= − A jk kj ( j ↔ k)

= − A jk jk
i
( jk
i
= kj
i
)
i i
∴ 2 A jk jk = 0 ∴ A jk jk =0
ii. Also, from the definition of xk-covariant derivative,

∂
Aij, k =
∂x k
( )
Aij − Aαj ik
α
+ Aiα αk
j
.

Considering a contraction with respect to j and k, i.e., setting j = k, we get

∂
Aij, j =
∂x j
Aij − Aαj( ) ij
α
+ Aiα αj
j

∂ ∂
=
∂x
j j
( )α α
j Ai − Aα ij + Ai
(α → k )
(log g , g > 0
∂ xα (α → j)
( )
66 Tensor Calculus and Applications

∂ 1 ∂
∴ Aij, j =
∂x j
( )
Aij − Akj ij
k
+ Aij
g ∂x j
g

1 ∂
=
g ∂x j
(
Aij g − Akj ) ij
k
.

Hence, proved.

Example 4
If, at a specified point, the derivatives of the gij ’s with respect to the coor-
dinates are all zero, the components of covariant derivatives of a vector
at the point are the same as ordinary derivatives.
∂ g ij
Given that k = 0 at a point P (say) ∀ i, j, k.
∂x
If A i is any covariant (may be contravariant also) vector, then its
xj -covariant derivative is given by

∂ Ai α ∂ Ai
Ai , j = − Aα ij = − A α g αβ ij , β
∂x j ∂x j
∂ Ai αβ 1  ∂ g iβ ∂ g jβ ∂ g ij 
= j − Aα g  j + −
∂x 2  ∂x ∂ x i ∂ x β 

∂ Ai ∂ g ij
Ai , j =  = 0 ∀ i, j, k at the point P.
∂x j ∂x β

4.13 Gradient of an Invariant Function

The partial derivatives of an invariant function φ is defined as the compo-
nents of a vector called grad φ or ∇φ .

Theorem of Invariant Function

If ϕ is invariant function, show that its covariant derivative is equal to its

ordinary derivatives.
Proof: Let Ai be an arbitrary vector so that (ϕAi) is also an arbitrary vector
since ϕ is invariant.
∴ Its xj-covariant derivative with respect to gij can be written as

∂ α
(φ Ai ), j = (φ Ai ) − (φ A )α ij
∂x j

∂A i ∂φ α
=φ + Ai j − φ Aα ij
∂x j ∂x
Christoffel Three-Index Symbols 67

 ∂A α ∂φ
φ, j Ai + φ Ai , j = φ  ji − Aα ij  + Ai j
 ∂x ∂x

∂φ
= φ Ai , j + Ai j .
∂x

∂φ
∴φ, j Ai − Ai =0
∂x j

 ∂φ 
∴ Ai  φ, j − j  = 0
 ∂x 

∂φ
∴ φ, j =  Ai is arbitrary.
∂x j
∴ The xj-covariant derivative of ϕ, namely, ϕj, is equal to its ordinary par-
∂φ
tial derivatives j .
∂x
∂φ
Hence, these partial derivatives of the invariant function ϕ are the
∂x j
components of grade ϕ vector. Otherwise, the covariant derivative ϕj of the
invariant function is a vector which is nothing but the (∇ϕ) vector.

4.14 Curl of a Vector
The xj-covariant derivative of the covariant vector Ai with respect to the fun-
damental tensor gij is given by
∂ Ai α
Ai , j = j − Aα ij , which is a second-order covariant tensor.
∂x
The components Ai,j is interpreted as obtained from Ai due to the operation
“covariant differentiation” with respect to the fundamental tensor gij.
The curl of the vector A( Ai ) is defined as
Curl A = Ai , j − A j , i .

 ∂A α  ∂ Aj α
=  ji − Aα ij  −  i − Aα ji 
 ∂x ∂x

∂ Ai ∂ A j α α
= −  ij = ji
∂x j ∂xi
∂ Ai ∂ A j
Hence, Curl A ( = Curl Ai ) = Ai , j − A j , i = − .
∂x j ∂xi
68 Tensor Calculus and Applications

Theorem: The Necessary and Sufficient Condition That the First

Covariant Derivative of a Covariant Vector Will Be Symmetric
If the Vector Is Gradient of Some Invariant Function

Proof: By the definition of curl of a vector A, we have

Curl A = Ai , j − A j , i = 0 if Ai , j = A j , i (symmetric).

But curl ∇ϕ = 0, where ϕ is some scalar invariant.

Otherwise, if A = ∇φ ,  Curl A = 0,
∴ Curl ∇φ = 0.

∴ Ai , j − A j , i = 0.

∴ Ai , j = A j , i , i.e., symmetric.

Hence, proved.

4.15 Divergence of a Vector
( )
The divergence of a contravariant vector uα ui is defined as the result of
contraction with respect to its covariant derivative.
∂ ui i
In, u,i j = j + uα α j
∂x
Let us allow the contraction setting i = j, so that

∂ ui ∂ ui ∂
u,ii =
∂xi
+ uα αi
i
=
∂xi ∂x
(
+ uα α log g , ( g > 0 ) )
(α → i )

∂ ui ∂
=
∂xi
(
+ ui i log g
∂x
)
∂ ui 1 ∂
=
∂x i
+ ui
g ∂xi
( g)
1 ∂
=
g ∂xi
(
ui g )
1 ∂
(
∴ div u or div ui = u,ii = ) g ∂xi
( )
gui ,

which is a scalar invariant.

Christoffel Three-Index Symbols 69

4.16 Laplacian of a Scalar Invariant

From the expression of divergence of the contravariant vector u ui ( )
(Section 4.15), we have

1 ∂
(
∴ div u = u,ii = ) g ∂xi
( gui )

1 ∂
( )
∴ div u = u,ii =
g ∂xi
( g g ij u j . ) (i)

∂φ
But we have already proved that φ, j = (Section 2.13), which is the covari-
∂x j
ant derivative of the scalar invariant ϕ and is the component of ∇ϕ; (i) can be
written as

1 ∂
div ( ∇φ ) =
g ∂xi
( g g ijφ, j )
or

1 ∂
∇ 2φ =
g ∂xi
( )
g g ijφ, j . (ii)

Hence, the divergence of the vector ∇ϕ, symbolically ∇ 2ϕ, is defined as the
Laplacian of ϕ, and (ii) is its explicit expression.
In deriving div ui from (i), we have replaced ui as

∂φ
ui = g ij u j = g ijφ, j  grad φ = φ, j =
∂x j

to recover the Laplacian of ϕ.

ui = g ijφ, j , since g ij is symmetric.
Therefore, div (∇ϕ) can also be interpreted as the outcome of contraction of
the covariant derivative of gij𝜙,j:

∴∇ 2φ = g ijφ, ij

∂φ
 φ, i = and div ui = u,ii = ( g ijφ, j ), i
∂xi

 ∂2 φ ∂φ k 
∇ 2φ = g ij  i j − k ij 
 ∂ x ∂ x ∂ x 
70 Tensor Calculus and Applications

∂  ∂φ  ∂φ k
φ, ij =  − ij  g ,iji = 0
∂x j  ∂xi  ∂x k

and

div ui = u,ii = ( g ij u j ), i

= ( g ijφ, j ), i = g ijφ, ij ,

which is another form of the Laplacian of ϕ.

Example 5
If Aij is the curl of a covariant vector, prove that

Aij , k + A jk , i + Aki , j = 0; otherwise,

∂ Aij ∂ A jk ∂ Aki
+ + = 0.
∂x k ∂x i ∂x j

Let Bi be the covariant vector so that

∂Bi ∂Bj
Aij = Bi , j − Bj , i = − . (i)
∂x j ∂xi

A ji = Bj , i − Bi , j interchanging i and j

(
= − Bi , j − Bj , i )
= − Aij .

Hence, Aij is an antisymmetric tensor.

∴ Aij + A ji = 0. (ii)

Now,

∂ Aij α α
Aij , k + A jk , i + Aki , j = − Aα j ik − Aiα jk
∂x k
∂ A jk α α ∂ Aki α α
+ − Aα k ji − A jα ki + − Aα i kj − Akα ij
∂xi ∂x j
∂ Aij ∂ A jk ∂ Aki
=
∂x k
+
∂xi
+
∂x j
− Aα j + A jα ( ) ki
α

α α
− ( Aiα + Aα i ) jk − ( Aα k + Akα ) ij

∂ Aij ∂ A jk ∂ Aki
= + +
∂x k ∂x i ∂x j
Christoffel Three-Index Symbols 71

α α
using (ii) and ij = ji .
Also,

∂ Aij ∂ A jk ∂ Aki ∂  ∂Bi ∂Bj  ∂  ∂Bj ∂Bk 

+ + = k j − i+ i k −
∂x k ∂x i ∂x j ∂x  ∂x ∂x  ∂x  ∂x ∂ x j 

∂  ∂Bk ∂Bi 
+  − , using ( i )
∂x j  ∂x i ∂x k 

∂ 2 Bi ∂2 Bj ∂2 Bj ∂ 2 Bk
= j
− k i + i k − i j
∂x ∂x
k
∂x ∂x ∂x ∂x ∂x ∂x
∂ 2 Bk ∂ 2 Bi
+ j
− j k = 0.
∂x ∂x i
∂x ∂x

Hence, proved.


4.17 Intrinsic Derivative or Derived Vector of v

If â is any vector representing a direction and v is any vector with covariant
components vi (or contravariant components vi), the intrinsic derivative or

derived vector of v in the direction of â is defined by means of the covariant
components vi , k a k or by contravariant components v,i k a k :

   


(
 For vi , k a k = δ ji v j ) ,k  ,k
(
a k =  g iα g jα v j  a k g iα vα ) ,K
a k = v,i k a k  .


It is usually denoted by the symbol a .∇v , which is nothing but the generaliza-
tion of the derivative of a vector in Euclidean space E3 in the direction of a.

Theorem: Show That a Vector of Constant Magnitude

Is Orthogonal to Its Intrinsic Derivative

Proof: Let u ( ui ) be a covariant vector of constant magnitude so that



u2 = g ij ui u j.

Considering the xk-covariant differentiation with respect to the fundamental

tensor gij, we can get

(
0 = g ij ui , k u j + ui u j , k )  g ,ijk = 0
72 Tensor Calculus and Applications

( g u )u
ij
j i, k ( )
+ g ij u j ui , k = 0 (i ↔ j in the second term)

ui ui , k + ui ui , k = 0 ∴ 2 ui ui , k = 0

ui ui , k a k = 0 (considering the inner product with a k )

( )
∴ ui ui , k a k = 0, which is of the form ui ui = 0.
This shows that the intrinsic derivative of ui , k in the direction of aˆ a k is ( )
orthogonal to u itself.
Hence, proved.

4.18 Definition: Parallel Displacement of Vectors

4.18.1 When Magnitude Is Constant

( )
Let the unit vector tˆ t k represent the direction at any point of a vector field u.
The vector u of constant magnitude (need not necessarily be of constant mag-
nitude) is said to be parallel along a curve C with respect to a Riemannian Vn,
if its intrinsic derivative (or derived vector) in the direction of the curve at all
points of C vanishes; mathematically,

dx k
u,i k t k = 0 or ui , k = 0.
ds

This can be thrown to the form:

 ∂ ui α i  dx k
 ∂ x k + u αk
 ds
=0

or

∂ui dx k i dx k
+ uα αk = 0,
∂ x k ds ds

so that

dui i dx k
+ uα αk = 0 . (4.18.1)
ds ds

Otherwise, the vector u is said to undergo parallel displacement (according

to Levi-Civita) along the curve C of Vn if (4.18.1) is satisfied.
Christoffel Three-Index Symbols 73

Increment of u:
dui i dx
k
From (4.18.1), + uα α k = 0.
ds ds
This can be said as the arc rate of change of the contravariant component
ui. This can be put to the form:
i
dui = − uα αk dx k , which is called the increment of ui due to the displace-
ment dxk.

4.18.2 Parallel Displacement When a Vector Is of Variable Magnitude

If the direction of two vectors is the same or if their corresponding compo-
nents are
 proportional, they are said to be parallel. 
Let B be a vector of variable magnitude parallel to a vector A at each point
of the curve C in Vn so that

Bi = λ ( s ) A i. (4.18.2)

( )
If A A i is assumed to be avector of constant magnitude parallel to the curve
( )
C with respect to Vn, then B Bi must also be parallel with respect to Vn along
the curve but of variable magnitude.
As Ai undergoes parallel displacement,

dx j
A,i j = 0. (4.18.3)
ds

Now,

dx j dx j dx j dx j
B,i j
ds
(
= λ Ai ) ,j ds
= λ, j A i
ds
+ λ A,i j
ds
∂λ dx j i dλ B i dλ
= Ai = A = , using ( 4.18.3 ),
∂ x j ds ds λ ds
dx j
∴ B,i j = Bi f ( s) (4.18.4)
ds

d
where f ( s ) =  log λ ( s)  .
ds
 i
( )
The vector B B of variable magnitude should be expressed in the form
(4.18.4), if it undergoes parallel displacement with respect to Vn along the
curve C. It can be promptly concluded that the intrinsic derivative of B
(L.H.S. of 4.18.4) at each point on the curve C has the same direction as that
( )
of B Bi (R.H.S.). Otherwise, B is parallel to itself along the curve C subject to
the condition (4.18.4).
74 Tensor Calculus and Applications

Writing Ai = Bi μ(s), we can get

dx j  i ∂ µ  dx j
A,i j =  µB, j + Bi j 
ds  ∂ x  ds

 dµ 
=  µBi f ( s) + Bi , using ( 4.18.4 ).
 ds 

 dµ 
= Bi  + µ f ( s) 
 ds 

dµ
We can choose + µ f ( s) = 0 so that
ds
dx j
A,i j = 0, which is the condition of parallel displacement of the vector
ds
( )
A A i of constant magnitude along the curve C in Vn.
( ) ( )
Hence, B Bi also undergoes  A i = µBi parallel displacement along the
curve C in Vn.
Hence, proved.
Note: From the condition of parallel displacement (4.18.4),

dx j
B,i j = Bi f ( s)
ds

dx j
Bk B,i j
ds
= Bk Bi f ( s) ( multiplying by B ) k

dx j
Bi B,kj = Bi Bk f ( s) ( k ↔ i ).
ds

j
(
Subtracting, Bi B,kj − Bk B,i j ) dxds = 0, (or) eliminating f ( s).
This is a modified form of the condition of parallelism of the vector B Bi ( )
of variable magnitude.

Theorem: If Two Vectors of Constant Magnitude Undergo Parallel

Displacement along a Given Curve, They Incline at a Constant Angle

( ) ( )
Proof: Let aˆ a k and bˆ b i be the two unit vectors of constant magnitudes,
and θ be the angle between them:
Christoffel Three-Index Symbols 75

∴ cosθ = g ij a ib j

d d ∂ dx k
∴
ds
(cosθ ) =
ds
( )
g ij a ib j = k g ij a ib j
∂x ds
( )
dx k
(
= g ij a ib j ) ,k ds
 g ij a ib j is invariant

dx k
( )
=  g ij a,i k b j + g ij a ib,jk 
  ds ( )  g ij , k = 0

 dx k   j dx k 
= bi  a,i k + a j  b, k
 ds   ds 

( )
= bi a,i k t k + a j b,jk t k . ( )
If â and b̂ undergo parallel displacements along a curve C given by the direc-
dx k
tion t k = , then
ds

a,i k t k = 0 = b,jk t k .

∴ The above relation reduces to

dθ
− sin θ =0
ds
dθ
∴ =0
ds
sin θ ≠ 0.

Hence, the vectors â and b̂ incline at a constant angle.

Hence, proved.

Theorem: If a Vector u Undergoes Parallel Displacement

along a Given Curve, It Must Be of Constant Magnitude

Proof: By definition, u2 = g ij ui u j = u j u j

∂ dx k
∴
d 2
ds
( )
u =
d
ds
(
uju j = k uju j
∂x ds
) ( )

 u j u j is invariant
76 Tensor Calculus and Applications

 dx k   dx k 
= u j  uj, k  + u j  u,j k
 ds   ds 

d 2
⇒
ds
( )
u = 0.

∴ For parallel displacement of u along the curve C of direction,

dx k dx k dx k
= t k , uj, k = 0 = u,jk .
ds ds ds

u2 = constant, i.e., u = constant.

Hence, proved.

Exercises
1. If â and b̂ are the two unit vectors, ϕ is a scalar invariant and the
derivative of ϕ in the direction of â is ϕ, i ai,, then show that the deriv-
ative of this quantity in the direction of b̂ is (ϕi ai),j bj = (ϕ,i ai,j + ai,ij)bj.
2. Find the nonvanishing Christoffel symbols for the metrics:
i. ds2 = a2 dr2 + sin2𝜃d𝜃2.
ii. ds2 = dr2 + r2 d𝜃2 + dz2.
iii. ds2 = dr2 + r2 d𝜃2 + r2 sin2 𝜃 dϕ 2.
iv. ds2 = e−2kt (dx2 + dy2 + dz2) − dt2.
3. If ϕ is a scalar and f(ϕ) is a function of ϕ, show that

∇ 2φ = f ′′(φ )(∇φ )2 + f ′(φ )∇ 2φ .

5
Properties of Curves in Vn and Geodesics

Definition

Curvature: The arc rate at which the tangent to a curve at a point P changes
the direction as P moves along the curve is called the curvature (κ) of the
curve.

5.1 The First Curvature of a Curve

If the coordinates xi of the current point P of a curve C in a Vn is taken as
function of the parameter arc length s, then the unit tangent vector tˆ t i to

( )
dx i
the curve is defined as t i = . The first curvature vector P( p i ) of the curve
ds
C relative to Vn is the derived vector of t̂ along the curve which is defined as

dx k
p i = t,i k . (5.1.1)
ds

The first curvature vector is of magnitude κ = g ij p i p j .

From the definition (5.1.1),

 ∂t i i dx j  dx k ∂t i dx k i dx j dx k
pi =  k + = +
ds  ds
jk jk
 ∂x ∂ x k ds ds ds
(5.1.2)
dt i i dx j dx k d 2 x i i dx j dx k
= + jk = 2 + jk .
ds ds ds ds ds ds

If the first curvature of the curve pi vanishes, then

j
d2 xi i dx dx
k
(
pi =) ds 2
+ jk
ds ds
= 0 will represent a particular class of curves called
geodesics, which will be discussed in Section 5.2. 
Principal normal: The direction of the first curvature vector P( p i ) is called
the principal normal, and the unit vector n̂ in the direction of p is called the
principal unit normal so that P = κ n̂.

77
78 Tensor Calculus and Applications

5.2 Geodesics
Definition

A geodesic on a surface in Vn is a curve or path of extremum (or stationary)

length joining any two given points on it.
For example,

i. A line joining two points in a plane is the shortest distance called

straight line which is geodesic.
ii. The great circular arc joining two points on the surface of a sphere is
the shortest distance between two points. Therefore, it is geodesic in
three-dimensional spherical surface.

5.3 Derivation of Differential Equations of Geodesics

Let A and B be the two points on a surface in an n-dimensional space Vn.
The arc length joining the two points on the surface of Vn is given by ds2 = gij
dxidxj.
For stationary (or extremum) length, let us consider a small variation so
that
∂ g ij k i j
2 dsδ ( ds ) =
∂x k
( ) ( )
δ x dx dx + g ij δ dx i dx j + gij dx iδ dx j

∂ g ij k i j
=
∂x k
( ) ( )
δ x dx dx + g ij d δ x i dx j + gij dx i d δ x j i↔ j

∂ g ij k i j
=
∂x k
( )
δ x dx dx + 2 g ij d δ x i dx j ∴ gij = g jt

 1 ∂ gij k dx i dx j d dx j 
∴ δ (ds) =  δx + gij δ xi ( )
ds 
ds.
 2 ∂x
k
ds ds ds

Applying variational principle for extremum values of the arc length join-
ing the two arbitrarily chosen points A and B on Vn, we are to make use of
B

∫ A
δ ( ds ) = 0.

B
 1 ∂ g ij k dx i dx j d i dx 
j
∴
∫ ( )
 2 ∂ x k δ x ds ds + g ij ds δ x ds  ds = 0 (5.3.1)
A  
Properties of Curves in Vn and Geodesics 79

Let us consider the second term of (5.3.1):

B
B
 d dx j   dx j  B  d  dx j  
∫  g ij δ xi( )  ds =  g ij ⋅δ xi  −
∫   g ij  ⋅ δ x i  ds
A  ds ds   ds A A  ds  ds  
B  d  dx j  i 
=−
∫ A
  g ij
 ds  ds 
δ x  ds.


 For small variation, δ x i = 0 at both the ends A and B.
∴ Equation (5.3.1) reduces to

  j 
 1 ∂ g ij dx dx δ x k − d  g ij dx  δ x i  ds = 0
B i j

∫ A  2 ∂ x k ds ds  ds 
  ids 
→k 



or

B  1 ∂ g ij dx i dx j  ∂ g jk dx i dx j d2 x j 

∫ A
 − i
 2 ∂ x ds ds  ∂ x ds ds
k
+ g kj 2   δ x k ds = 0.
ds  

But for all δ x k, i.e., arbitrary value of δ x k if the above result is to hold good,
then we must have

1 ∂ g ij dx i dx j ∂ g jk dx i dx j d2 x j
− 1 ⋅ − g jk =0
2 ∂ x k ds ds ∂ x i ds ds ds 2

or

d 2 x j  1 ∂ g jk dx i dx j 1 dg ik dx j dx i  1 ∂ g ij dx i dx j
g jk + + − =0
ds2  2 ∂ x i ds ds 2 dx j ds ds  2 ∂ x k ds ds

(interchanging i and j in one term within the bracket).

Considering the inner product of it with gmk and summing over k, we get

d2 x j
{ } dxds dxds
i j
g mk g jk + g mk ij , k =0
ds 2

or
80 Tensor Calculus and Applications

d2 x j m dx i dx j
δ jm + ij =0 g mk g jk = δ jm
ds2 ds ds

or

d2 xm m dx i dx j
+ ij =0
ds2 ds ds
 δ jm = 1 for m = j

= 0 for m ≠ j.

This can be written as

d2 xi i dx j dx k
+ jk = 0 (5.3.2)
ds2 ds ds

(changing m → i and i → k).

The preceding equation is the differential equation of the curve called
geodesis.
These are the second-order differential equations in terms of n variables xi
(i = 1, 2,…, n); each solution must have got two arbitrary constants. Therefore,
second-order n differential equations should contain 2n arbitrary constants
in their general solutions. Hence, 2n given conditions are necessary to know
the complete solutions.
The 2n coordinates of the points A and B are sufficient to determine the
2n arbitrary constants occurring in the solutions. Hence, the geodesic can be
determined uniquely. Otherwise, through the two given points A and B on
the surface in a Vn, one and only one geodesic can pass.
Again, besides the coordinates A(xi), if the n components of the unit tan-
 dx i 
gent vector tˆ  t i = at the point A are known, then the geodesics can be
 ds 
determined uniquely.

5.4 Aliter: Differential Equations of

Geodesics as Stationary Length
Let A and B be the two fixed points on a curve C of Vn, and t0 and t1 be the
parametric values of A and B, respectively.
Properties of Curves in Vn and Geodesics 81

∴ The length of the arc joining A to B is given by

t1
dx i dx j
∫
t0
g ij
dt dt
dt

( )
∵ ds 2 = g ij ( x i )dx i dx j . (5.4.1)
t1

=
∫
t0
g ij x i x j dt

If the curve C is to be geodesic, the above length should be stationary (or

extremum):
t1

But the Euler’s equation for extremum value of the integral I =

states that it must satisfy the differential equation:
∫ f ( x x ) dt
t0
i j

∂f d  ∂f 
−  i  = 0 . (5.4.2)
∂ x dt  ∂ x 
i

But for (5.4.1),

ds
f = g ij x i x j = = s
dt

∂f 1 ∂ g jk j k 1 ∂ g jk j k
∴ = x x = x x (i → k ).
∂x i
2 g jk x x ∂ x
k j i
2 s ∂ x i

∂f 1
Also = g ij x j .
∂ x i s
∴ Equation (5.4.2) reduces to

1 ∂ g jk j k d  1 
x x −  g ij x j  = 0
2 s ∂ x i dt  s 

1 ∂ g jk j k  s 1 ∂ g ij k j 1 
x x −  − 2 g ij x j + k x x +
  g ij x j  = 0
2 s ∂ x i  
s 
s ∂ x 
s 
∂ g ij j k 1 ∂ g jk j k s
g ij x j + x x − x x − g ij x j = 0
∂x k 2 ∂xi s
 1 ∂ g ij j k 1 ∂ g ik k j  1 ∂ g jk j k s
g ij x j +  x x + x x  − x x − g ij x j = 0.
 2 ∂x k 2 ∂x j  2 ∂xi s
82 Tensor Calculus and Applications

s
g ij x j + kj , i x j x k − g ij x j = 0
s
s
g iα g ij x j + g iα kj , i x j x k − g ij g iα x j
s
=0 ( taking the inner product with g )
iα

α s
xα + jk x j x k − x α =0
s

If the usual arc length parameter s is chosen instead of t, then the equation
transforms to

d2 xi i dx j dx k
∴ + jk =0
ds 2 ds ds

(∵ s = 1, s = 0) .
This is the differential equation of geodesics obtained from the notion of
stationary arc length.

5.5 Geodesic Is an Autoparallel Curve

Let t̂ be the unit tangent vector to a geodesic curve. The differential equations
of geodesics are

d2 xi i dx j dx k
+ jk =0
ds2 ds ds

or

d  dx i  i dx j dx k
+ =0
ds  ds 
jk
ds ds

dt i i j k dx i
∴ + jk t t = 0 ti = .
ds ds
It can be written as

∂t i dx k i j k
+ jk tt =0
∂ x k ds

 ∂t i j i  k
 ∂ x k + t  t = 0 ∴ t, k t = 0,
i k
jk
Properties of Curves in Vn and Geodesics 83

which is the condition of parallel displacement of the unit tangent vector ti to

the curve geodesic but in the direction of itself.
Hence, geodesics are autoparallel curves.

Example 1
Determine the differential equations of geodesis in a space given by the
metric

( )
ds 2 = − e 2 kt dx 2 + dy 2 + dz 2 + dt 2 .

For the given metric, g11 = g 22 = g 33 = − e 2 kt , g 44 = 1.

But g ij = 0 for i ≠ j

∂ g11 ∂ g 22 ∂ g 33
∴ = = = − 2 ke 2 kt (i)
∂x 4 ∂x 4 ∂x 4

∂ g 44
and = 0 for all “ i”.
∂x i

1
Also for the metric, gii exists only; therefore, g ii = .
g ii
Now the differential equations of geodesics are

d2 xi i dx j dx k
+ jk = 0. (ii)
ds 2 ds ds
i
∴ We are to determine nonvanishing Christoffel symbols jk of the
­second kind:

1 ∂ g jj
jj = g
1.
i
ii
jj , i =−
2 g ii ∂ x i
=−
1
2.1
( )
−2 ke 2 kt = ke 2 kt

due to i = 4 only and j = 1, 2, 3(i ≠ j)

4 4 4
∴ 11 = 22 = = ke 2 kt
33
i 1 ∂ g ii
2.
ij = g
ii
ij , i = .
2 g ii ∂ x j
1 2 3
∴ 14 , 24 , 34 exists only due to (i) for i = 1, 2, 3, and j = 4 only.

1 ∂ g11
) ( )
1 1
∴ = = × −2 ke 2 kt = k .
14
2 g11 ∂ x 4 2 − e 2 kt (
2 3
Similarly, = 34 = k. 24
i 1 ∂ g ii ∂ g 44
3. But ii = = 0 ∴ 4 = 0.
2 g ii ∂ x i ∂x
84 Tensor Calculus and Applications

Now for i = 1, the differential equation (2) of the geodesic

takes the form:

d 2 x1 1 dg j dg k
+ jk =0
ds 2 ds ds
or

d 2 x1 1 dx 1 dx 4 1 dx 4 dx 1
+ 14 + 41 =0
ds 2 ds ds ds ds
or

d2 x dx dt 1 1
+ 2k =0 ∴ 14 = 41 =k
ds 2 ds ds
or

d2 x
ds 2 = −2 k dt .
dx ds
ds

Integrating, log   = −2 kt + log a.

dx
 ds 
This can be written as

dx
= ae −2 kt . (iii)
ds
Similarly, putting i = 2 and 3, respectively, we can get

dy
= +be −2 kt . (iv)
ds
dz
= ce −2 kt . (v)
ds

Again putting i = 4, Equation (ii) can be simplified as

d2 x4 4 dx j dx k
+ jk =0
ds 2 ds ds
or

d2 x4 4 dx 1 dx 1 4 dx 2 dx 2 4 dx 3 dx 3
+ 11 + 22 + 33 =0
ds 2 ds ds ds ds ds ds

or

d2t 
2 kt  dx 
2
 dy 
2
 dz  
2
+ ke    +   +  =0
ds 2
 ds  ds   ds  

Properties of Curves in Vn and Geodesics 85

or
d2t
ds 2
( )
+ ke 2 kt × a 2 + b 2 + c 2 e −4 kt = 0 using (iii), (iv), and (v)

d2t
∴ + kα 2 e −2 kt = 0 (vi)
ds 2

(
putting α 2 = a 2 + b 2 + c 2 . )
Hence, Equations (iii)–(vi) are the required differential equations for
the geodesics.

5.6 Integral Curve of Geodesic Equations

The equations of geodesics are (Equation 5.3.2)

d2 xi i dx j dx k
+ jk = 0. (5.6.1)
ds2 ds ds

If C is the curve geodesic through a point P and s is the arc length, then by
Taylor’s theorem, we can write

 dx i  1  d2 xi  2 1  d3xi  3
x i = x0i +  s + s +  3  s +  . (5.6.2)
 ds  0 2  ds 2  0 3  ds  0

The second and higher derivatives of xi with respect to s can be determined

from the equations of geodesics (5.6.1) as follows:

d2 xi i dx j dx k
=− jk (5.6.3)
ds2 ds ds

d
( ) dxds dxds + dx j d 2 x k 
j
d3 xi i
k
i d 2 x j dx k i
= − +
ds ds2 
jk jk jk
ds3  ds ds2 ds

 
= − l
∂
 ∂x ( ) jk
i dx j dx k dxl
ds ds ds
− jk
i
αβ
j dxα dx β dx k
ds ds ds
− jk
i
αβ
k dxα dx β dx j 
ds ds ds 
 (α →l , β → j , j→α ) (α →l , β → k , k →α ) 

using ( 5.6.3 )

( ) dxdsdxds dx j dx k dx l 
j
 ∂ i dx k l
i α dx j dx k dx l i α
=−  l − −
ds ds ds 
jk αk lj jα lk
 ∂x ds
 ( j↔ k ) ds
 dsds
   
86 Tensor Calculus and Applications

d3xi i dx j dx k dx l
∴ + jkl = 0, (5.6.4)
ds3 ds ds ds
where

jkl
i
=
1  ∂
p
3  ∂ x l ( )− jk
i
αj
i
lk
α
− jα
i
lk
α 


=
1  ∂
p
3  ∂ x l ( )− 2 jk
i
αj
i
kl
α 


and P indicates the sum of the terms obtained by permuting the subscripts
cyclicly.*
Using (5.6.3) and (5.6.4) in (5.6.2) for P(xo), we get

 dx i 
( )  dxds   dxds  s − 13 ( )
j k
1 i 2 i
x i = xoi +  s−
 ds  o
jk jkl
2 o
o
o

 dx j   dx k   dxl  3
 ds   ds   ds  s + .
o o o

 dx i 
If we define ξ i =  , it can be thrown to the form:
 ds  o

x i = xoi + ξ i s −
1
2 ( ) ξ ξ s − 13 ( ) ξ ξ ξ s +  (5.6.5)
jk
i

o
j k 2
jkl
i

o
j k l 3

i
The convergence of the series is dependent on ξ i and gij (for jk ). This, of
course, represents an integral (solution) curve of the equations of geodesics
(5.6.1) for small values of the parameter s.

5.7 Riemannian and Geodesic Coordinates, and Conditions

for Riemannian and Geodesic Coordinates
Riemannian coordinates: Consider yi = ξ is for the particular geode-
i.
sic* (5.6.5) which passes through the point P(xo). It takes the form:

x i = xoi + y i −
1
2 ( ) y y − 13 ( ) y y y +  . (5.7.1)
jk
i

o
j k
jkl
i

o
j k l

* [2, p. 52].
Properties of Curves in Vn and Geodesics 87

It represents all geodesics passing through Po(xo) given by various

 dx i 
directions ξ j =  . Also the equations yi = ξ is define a curve
 ds  o
otherwise geodesic for the given set of values of ξ i in terms of new
coordinates yi. These coordinates yi ’s are called Riemannian coordi-
nates as adopted by Riemann. Of course, these coordinates have got
their own domain about Po.
Let the fundamental form, Christoffel symbols in terms of the
Riemannian coordinates yi, be represented by*
i
g ij ( y i )dy i dy j , ij , k and jk , respectively.

∴ The equations of geodesics are

d2 y i i dy j dy k
+ jk = 0 (5.7.2)
ds 2 ds ds

i
∴ jk ξ jξ k = 0  y i = ξ i s
.
i
j
∴ jk y y =0
k

Clearly, subject to these conditions, the above equations of geodesics

are satisfied, and hence, yi ’s are Riemannian coordinates.
Following the method adopted in Section 5.6, but for Equation
(5.7.2) in terms of the Riemannian coordinates yi, we can get

1 i  j k 2
yi = ξ is −   ξ ξ s +  , ( y )o = 0 for s = 0.
i
2
jk
o

 i

 = 0 if it is to reduce to y = ξ s .
i i
This gives 
 jk
o

∴ 
i
 = 0 ⇒  ∂ g ij  = 0
 ij , k   ∂ y k  ( i, j, k = 1, 2, , n)
o

 ∂ g ij  α α
 =  ∂ g ij  = 0.
g ij , k =  k  −  g α j ik + gαi jk
 o  ∂ y k 
 ∂y  o  o

∴The first covariant derivative of the components of the fundamen-

tal tensor in these coordinates at the origin must vanish. This is the
condition of yi ’s to be Riemannian coordinates.
88 Tensor Calculus and Applications

ii. Geodesic coordinates:

In an arbitrary Riemannian Vn, it is not possible to choose the
Cartesian coordinate system, where the metric function gij ’s are con-
stants. But it is possible to choose the coordinate system where gij ’s
are locally constants so that

∂ g ij
=0 at the point called the pole,
∂x k
≠0 elsewhere.
i
∴ = 0 at the pole Po ; such a system of coordinates are called
jk
∂ Ai
geodesic coordinates with pole Po . Hence, for any tensor Ai , j =
∂x j
which is a condition for geodesic coordinates.

5.7.1 Another Form of Condition for Geodesic Coordinates

From the transformation law (Section 4.4), the resulting equation (4.4.5) of
Christoffel symbol of the second kind, we have
α
′ p ∂x = α ∂xλ ∂x µ ∂ 2 xα
ij λµ j +
∂x′ p ∂x′ ∂x′ ∂x′ i ∂x′ j
i

λ µ α
2
′α ∂ x ′ ∂ x ′ = ∂ x ′ − ∂ x ′α
∴− λµ
∂xi ∂x j ∂xi ∂x j
ij
p

∂x p
(
interchanging x i and x ′ i systems )

∂  ∂ x ′α  p  ∂ x ′α  (5.7.3)
= −
∂ x i  ∂ x j   ∂ x p 
ji

 ∂ x ′α 
= .
  ∂ x j  , i

Taking a fixed value of x′α out of n independent values, we can entrust

∂ x ′α
= x,′αj  x′α is a scalar invariant for the fixed value.
∂x j
∴ Equation (5.7.3) can be written as
λ µ
− ′α ∂ x′ ∂ x′ = x′α , (5.7.4)
λµ , ij
∂xi ∂x j
i.e., a second covariant derivative with respect to the metric of Vn. But if
x′α  are geodesic coordinates, then ′ = 0.
α
λµ

∴ x,′ijα = 0.
Properties of Curves in Vn and Geodesics 89

∴ The second covariant derivatives of the geodesic coordinates x′α must

vanish.
Otherwise, if x′α = 0 , then from (5.7.4), ′ = 0 at pole.
α
, ij λµ
Hence, if a system of coordinates are geodesic coordinates with pole, then
their second covariant derivatives with respect to the metric of the space
must vanish at that point. From the above interpretation, it can be concluded
that it is a necessary and sufficient condition.

5.8 If a Curve Is a Geodesic of a Space (Vm), It Is also a

Geodesic of Any Space Vn in Which It Lies (Vn a Subspace)
Let C(xi) be any non-minimal curve in a Vn at points xi(s) with fundamental
dx i
form ds2 = gij dxi dxj. If λ i = are the components of a unit vector field of Vn,
ds
then the derived vector of λ i in the direction of C is given by

dx k
λ ,ik = η i (say). (5.8.1)
ds

But at points of C, if the vectors are parallel in the direction of C, then

dx k
ηi = λ ,ik = 0.
ds
Let the space Vn be immersed in a space Vm of coordinates yα and funda-
mental form ds2 = aαβ dyα dy β so that yα = f α ( x i ) (i = 1, 2, , n; α = 1, 2, , m)

∂xi ∂x j ∂ yα ∂ y β
∴ aαβ = g ij or g ij = aαβ . (5.8.2)
∂ yα ∂ y β ∂xi ∂x j

Let the components in terms of y’s of the vector field in Vm be ξ α , and λ i’s be
the corresponding components in x’s.

dyα ∂ yα dx i ∂ yα i
∴ = ∴ξα = λ
ds ∂ x i ds ∂xi
(5.8.3)
dξ α dλ i ∂ yα ∂2 yα dx j
∴ = + λi i j
ds ds ∂ x i
∂ x ∂ x ds

dyα  ∂ξ β
( )a  dyds ,
α
β
∴ ξ ,αβ β
i =η =  α +ξ
γ
γα
ds  ∂ y
90 Tensor Calculus and Applications

where ( ) γα
β

a
denotes the Christoffel (bracket) symbol of the second kind
with respect to the fundamental tensor aαβ .

ηβ =
∴
∂ξ β dyα
∂ yα ds
+ ( ) γα
β

a
ξγ
dyα dξ β
ds
=
ds
+ ( )γα
β

a
λi
∂ yγ ∂ yα dx j
∂ x i ∂ x j ds
, using (5.8.3)

dλ i ∂ y β
( )
2 β
j ∂ y dx i β ∂ yγ ∂ yα dx j
∴η β = + λ + γα λi . (5.8.4)
ds ∂ x i ∂ x i ∂ x j ds a ∂ x i ∂ x j ds
( i→ j ) (i↔ j)

Now,

1  ∂ gik ∂ g jk ∂ gij 
( )
ij , k
g
= + −
2  ∂ x j ∂ x i ∂ x k 

1  ∂ aαγ ∂ y β ∂ yα ∂ yγ ∂2 yα ∂yγ ∂yα ∂yγ

=  β + aαγ  g ik = aαγ , etc.
2  ∂y ∂x ∂x ∂x
j i k i j
∂x ∂x ∂x k
∂x i ∂x k

∂yα ∂2 yγ ∂ a βγ ∂ y α ∂ y β ∂ y γ ∂2 y β ∂yγ
+ aαγ + + a
∂yα ∂x i ∂x j ∂x k
βγ
∂x i ∂x j ∂x k ∂x i ∂x j ∂x k
(β →α )

 i j k 
 
 
 α β γ 
 

∂ aαβ ∂ y γ ∂ y α ∂ y β ∂2 yα ∂ y β ∂yα ∂2 y β 
− ⋅ − aαβ − aαβ
∂yγ ∂x k ∂x i ∂x j ∂x i ∂x k ∂x j ∂x i ∂x j ∂x k 
(α →γ ) ( β →γ ) 

1  ∂ aαγ ∂ aβγ ∂ aαβ  ∂ y α ∂ y β ∂ y γ ∂2 yα ∂ yγ 

=  β + α − + 2 a 
∂ y γ  ∂ x i ∂ x j ∂ x k
αγ
2  ∂ y ∂y ∂ x i ∂ x j ∂ x k 

∂yα ∂y β ∂yγ ∂2 yα ∂yγ

( ) =(
ij , k
g
αβ ,γ ) a i j
∂x ∂x ∂x k + aαγ
∂x i ∂x j ∂x k
(α → c )

( ) ∂∂yx
α
c ∂y β ∂yγ ∂2 y c ∂yγ
= acγ αβ j k + acγ
a
i
∂x ∂x ∂x i ∂x j ∂x k
 ∂ yc
( ) ∂yα ∂y β  ∂yγ
2
c
= acγ  i j + αβ 
 ∂x ∂x a ∂x i ∂x j  ∂x k

 ( )
ij , k
g
 ∂2 x β
= aβγ  i j + α c
 ∂x ∂x
β

( c ↔γ )
∂y c ∂yα  ∂yγ
j i 
a ∂x ∂x  ∂x
k
. ( ) (5.8.5)
Properties of Curves in Vn and Geodesics 91

∂y c
Multiplying (5.8.4) by aβ c ,
∂x k

∂ y c β dλ j ∂y c ∂y β
aβ c η = a β c
∂x k ds ∂
 xj
x k∂

+λ j
dx i
ds
∂y c  ∂2 x β
aβ c k  i j +
∂x  ∂x ∂x ( ) αγ
β

a
∂yα ∂yγ 

∂x i ∂x j 

= g jk
dλ j
ds
+ λj
dx i
ds
( ) ij , k
g
using ( 5.8.5 )

= g jk
∂λ j dx i
∂ x ds
i
+ λj
dx i
ds
( ) ij , k
g

dx i ∂λ j
( )
i
j dx l
= g jk + λ glk ij
ds ∂ x i ds g
( j→ l)

= glk
dx i
ds
 ∂λ l
 ∂x i + λ

j
( ) 
ij
l

 dx i l 
= glk  λ, i  .
 ds 

dyα
If ξ β is parallel with respect to the curve C in Vm, then ξ ,βα = η β = 0;
dx i ds
­therefore, λ,l i = 0.
ds
∴ The vectors that are parallel to a curve C in Vm, they are also parallel with
respect to Vn, a subspace of Vm.
If λ i is a unit tangent vector to the curve C, then parallelism of the vectors
along C means that the curve is a geodesic.
Hence, if a curve is a geodesic in a space (here Vm), it is also a geodesic in
the subspace Vn as it immerses in Vm.
Hence, proved.

Exercises
1. Determine the differential equations of geodesics for the metric:
i. ds2 = dr2 + r2 dθ 2 + r2sin2θdϕ 2.
ii. ds2 = dr2 + r2 dθ 2 + dz2.
92 Tensor Calculus and Applications

2. Find the differential equations of geodesics for the metric:

1
ds2 = f ( x ) dx 2 + dy 2 + dz 2 + dt 2.
f ( x)
3. Applying variational principles, derive the differential equations of
geodesic:

d2 xi i dx j dx k
+ jk = 0.
ds2 ds ds
4. If the coordinates xi of points on a geodesic are the functions of s, the
arc length, show that
d rφ dx i dx j dx p
r
= φ, ij…p  , where ϕ is any scalar function of x’s.
ds ds ds ds
5. Obtain the differential equations of geodesic, if
dt 2 1
ds2 = − (dx 2 + dy 2 + dz 2 ), k = constant. Hence,
(1 − kx) c 2 (1 − kx)2
prove that along a geodesic, V2 − v2 = kc2x, where V is constant and
2 2 2
 dx   dy   dz 
v2 =   +   +   .
 dt   dt   dt 
6. Find the differential equations of geodesic for the metric:
ds2 = − dx2 − dy2− dz2 + f(x, y, z) dt2.
7. Obtain the differential equations of geodesic as a means of stationary
arc length.
8. Obtain the differential equations of geodesic for the metric:

 g 2t 2 
ds2 = dx 2 + dy 2 + dz 2 + 2 gt dx dt – c 2  1 − 2  dt 2 .
 c 
6
Riemann Symbols (Curvature Tensors)

6.1 Introduction
The importance of intrinsic multidimensional differential geometry has
found its place in the study of general theory of relativity. The general the-
ory of relativity is inherently related to the geometry of curved space due
to the effects of gravity. For this reason, Einstein needs to deal with four-­
dimensional space–time continuum in support of “absolute differential
calculus” or tensor calculus. The Riemannian geometry associated with
covariant differentiation through the fundamental tensor gij characterizing
the space is properly suited for the curved space of general theory of rela-
tivity. This demands, though not general, an important notion (or concept),
namely, Riemannian symbols or curvature tensors.

6.2 Riemannian Tensors (Curvature Tensors)

The xj-covariant differentiation of the covariant vector Ai with respect to the
fundamental tensor gij is given by

∂ Ai α
Ai , j = − Aα ij , (6.2.1)
∂x j

which is again a second-order covariant tensor. Hence, considering its xk-

covariant differentiation again, it can be written as

∂ Ai , j
( treating A )
a a
Ai , jk = − Aa , j ik − Ai , a jk ij = Ai , j
∂x k
∂  ∂ Ai α   ∂ Aa α a  ∂A α a
= − Aα  −  ∂ x j − Aα  −  ai − Aα  using ( 6.2.1)
∂ x k  ∂ x j
ij aj ik ia jk
 ∂x

93
94 Tensor Calculus and Applications

=
∂2 Ai
∂x k ∂x j
− ij
α ∂ Aα
∂x k
∂
− k
∂x ( ) A − ∂∂Ax
ij
α
α
a
j ik
a
+ Aα aj
α
ik
a
−
∂ Ai
∂x a
jk
a
+ Aα ia
α
jk
a

 ∂2 Ai ∂ Aα α ∂ Aa a ∂ Ai a α a 
Ai , jk =  k j − ij − ik − jk + Aα ia jk 
 ∂x ∂x ∂x k ∂x j ∂x a 

−
∂
∂x k ( )A
ij
α
α + Aα aj
α
ik
a
.

(6.2.2)

As covariant differentiation is not commutative, we can write (interchang-

ing j and k)

 2 
∂ A ∂ Aα ∂ Aa ∂ Ai
Ai , kj =  j i k − 
α a a α a
ik − ij − kj + Aα ia kj
 ∂x ∂x ∂x j ∂x k ∂x a 
 (α → a ) ( a→α )  (6.2.3)

−
∂
∂x j ( )A ik
α
α + Aα ak
α
ij
a
.

Subtracting (6.2.3) from (6.2.2), we get

Ai , jk − Ai , kj = − Aα ( )+ A ∂
∂x k
+
∂x (
∂
)A ij
α
α aj
α
ik
a
j ik
α
α − Aα ak
α
ij
a

 ∂
=A 
 ∂x
(
α ) −
∂x (
∂
)+ −
j ik
α
k ij
α
aj
α
ik
a
ak
α
ij
a 
 (6.2.4)

Ai , jk − Ai , kj = Aα Rαijk ,

where

α
Rijk =
∂
∂x j ( ) − ∂∂x ( ) +
ik
α
k ij
α
aj
α
ik
a
− ak
α
ij
a
(6.2.5)

The left-hand side of (6.2.4) being the difference of two third-order covariant
tensors must also be a third-order covariant tensor with real covariant indi-
ces, i, j, k. Hence, the right-hand side must also be a third-order covariant ten-
sor with real indices i, j, k. But already there appears the first-order covariant
tensor Aα on the right-hand side of (6.2.4). This necessitates the use of real
α
covariant indices i, j, k in the symbol Rijk with dummy index α in place of the
bracket to yield a covariant tensor only.
α
∴ By quotient law, Rijk must be a fourth-order mixed tensor with its explicit
form:
Riemann Symbols 95

α
Rijk =
∂
∂x j ( ) − ∂∂x ( ) +
ik
α
k ij
α
aj
α
ik
a
− ak
α
ij
a
.

This is called the curvature tensor for the Riemannian metric ds 2 = g ij dx i dx j.

α
Also Rijk is referred to as the Riemannian symbols of the second kind.

6.3 Derivation of the Transformation Law

α
of Riemannian Tensor Rabc
From the transformation law of Christoffel bracket (symbol) (4.4.5) of the sec-
ond kind, we get
α
′ p ∂x = α ∂x a ∂xb ∂ 2 xα
ij ab j + . (6.3.1)
∂x′ p ∂x′ ∂x′ ∂x′ i ∂x′ j
i

 i j k 
 .
 a b c 

Differentiating partially with respect to x ′ k ,

∂  ′ p  ∂ xα ′p ∂ x
2 α

 ij  + ij
∂ x′ k  ∂ x′ p ∂ x′ k ∂ x′ p

∂
( ) ∂∂xx′ ∂x a ∂xb ∂2 x a ∂ xb
c
α α
= ab + ab (6.3.2)
∂xc k
∂ x′ i ∂ x′ j ∂ x′ i ∂ x′ k ∂ x′ j
α ∂ x a ∂2 xb ∂ 3 xα
+ ab + .
∂ x′ i ∂ x′ j ∂ x′ k ∂ x′ i ∂ x′ j ∂ x′ k

Interchanging j and k in the above expression,

α
∂  ′ p  ∂x + ′p ∂ 2 xα
 ik  ik
∂ x′ j  ∂ x′ p ∂ x′ j ∂ x′ p

∂  α  ∂x c ∂x a ∂xb α ∂2 x a ∂xb
= + (6.3.3)
∂ x c  ( c ↔ b ) ∂ x′ j ∂ x′ i ∂ x′ k
ab ab
∂ x′ i ∂ x′ j ∂ x′ k
(b→ c )

α ∂x a ∂2 xb ∂ 3 xα
+ ab + .
∂ x′ i ∂ x′ k ∂ x′ j ∂ x′ i ∂ x′ k ∂ x′ j

∂ 3 xα
Eliminating from (6.3.3) and (6.3.2),
∂x′ ∂x′ j ∂x′ k
i
96 Tensor Calculus and Applications

∂ xα  ∂  ′p  − ∂  ′p   + ′p ∂ 2 xα ′p ∂ x
2 α
 j  ik  ∂ x′ k  ij   ik p − ij
∂ x′ p  ∂ x′ 
j
 ∂ x′ ∂ x′ ∂ x′ k ∂ x′ p

=
∂
∂xb ( ) ac
α ∂x a ∂xb ∂x c
j
∂ x′ ∂ x′ ∂ x′
i k
∂
− c
∂x ( ) ab
α ∂x a ∂xb ∂x c
∂ x′ i ∂ x′ j ∂ x′ k

α ∂2 x a ∂x c α ∂2 x a ∂xb α ∂x a ∂2 xb α ∂x a ∂2 xb
+ ac − ab + ab − ab
∂ x′ i ∂ x′ j ∂ x′ k ∂ x′ i ∂ x′ k ∂ x′ j ∂ x′ i ∂ x′ k ∂ x′ j ∂ x′ i ∂ x′ j ∂ x′ k

 
∂ xα  ∂  ′p  ∂  ′p   ′p  m ∂ xα α ∂xλ ∂x µ 
∴  ∂ x′ j  ik  − ij   + −
 ∂ x′ k 
ik jp λµ
∂ x′ p    ∂ x′ m ∂ x′ j ∂ x′ p 
 ( p ↔ m) (λ →b) ( µ → a) 

 α 
′m ∂x − ∂xλ ∂x µ
− ij ′ 
α
p
kp λµ

 ∂ x′ m ∂ x′ k ∂ x′ p 
 ( p ↔ m) (λ → c ) ( µ → a) 

 
 ∂
( ) ∂
( )  ∂x ∂x ∂x ∂xc  ′m ∂x a ∂xλ ∂x µ 
a b c
α α α a
= b ac − c ab  ∂ x′ ∂ x′ ∂ x′ k + ac ij − λµ
 ∂x ∂x i j
∂ x′ k  ∂ x′ m ∂ x′ i ∂ x′ j 
 ( m→ p ) 

 
α ∂ xb  ′ m ∂ x a a ∂xλ ∂x µ 
− ab ik − λµ
∂ x′ j  ∂ x′ m ∂ x′ i ∂ x′ k 
 ( m→ p ) 
α
 ∂  ′p  − ∂  ′p  + ′m ′ p − ′m ′ p  ∂x
 ∂ x′ j  ik  ∂ x′ k  ij  ik jm ij km  ∂ x′ p
 

( ) ( ) ∂xλ ∂x µ
a b c
 ∂ α ∂ α  ∂x ∂x ∂x α ∂xc a
= b ac − ab  ∂ x′ i ∂ x′ j ∂ x′ k − ac λµ
 ∂x ∂xc ∂ x′ k ∂ x′ i ∂ x′ j
(λ ↔ a , µ →b)

α ∂ xb a ∂xλ ∂xµ
+ ab λµ
∂ x′ j ∂ x′ i ∂ x′ k
(λ ↔ a , µ → c)

∂ xα  ∂
( ) − ∂∂x ( ) −
a b c
α α α λ α λ  ∂x ∂x ∂x
′p
∴ Rijk = + ×
∂ x′ p  ∂ xb  ∂ x′ ∂ x′ ∂ x′ k
ac c ab λc ab λb ac i j

∂ xα α ∂x
α
∂ xb ∂ x c
′p
Rijk = R abc ,
∂ x′ p ∂ x′ i ∂ x′ j ∂ x′ k

α
where Rabc =
∂
∂ xb ( ) − ∂∂x ( ) +
ac
α
c ab
α
λb
α
ac
λ
− λc
α
ab
λ
. (6.3.4)
Riemann Symbols 97

∂x′ h
Multiplying both sides by , we get
∂ xα

∂ xα ∂ x ′ h α ∂x
a
∂ xb ∂ x c ∂ x′ h
′p
Rijk = R abc
∂ x ′ ∂ xα
p
∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ xα

or

∂x a ∂ xb ∂ x c ∂ x′ h
′ p δ p′ h = Rαabc
Rijk
∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ xα

∂x a ∂ xb ∂ x c ∂ x′ h
′ h = Rαabc
∴ Rijk ,
∂ x ′ i ∂ x ′ j ∂ x ′ k ∂ xα

which is the transformation law of the fourth-order mixed tensor.

Rαabc in (6.3.4) is the similar expression for the Riemannian tensor or curva-
ture tensor of the second kind.

α
6.4 Properties of the Curvature Tensor Rijk
α α
i. Rijk = − Rikj (antisymmetric with respect to the second pair of indices).
α α α
ii. Rijk + R jki + Rkij = 0.
iii. Contraction in two different ways: (a) Rijk i
= 0 and (b) Rijαα , a tensor
called Ricci tensor.

Proof of (i)
By definition,

α ∂  α
− ∂  α
+ α b α b
Rijk = −
∂x j   ∂x k  
ik ij bj ik bk ij

and

α ∂  α
− ∂  α
+ α b α b
Rikj = − .
∂x k   ∂x j  
ij ik bk ij bj ik

α α
On comparison, it is clear that Rijk = − Rikj , i.e., antisymmetric with respect to
j and k.
Hence, proved.
Proof of (ii)
98 Tensor Calculus and Applications

By definition, it can be written as

α ∂  α
− ∂  α
+ α b α b
Rijk + Rαjki + Rαkij = −
∂x j   ∂x k  
ik ij bj ik bk ij

∂  α
− ∂  α
+ α b α b ∂  α

+ − +
∂x k   ∂xi   ∂xi  
ji jk bk ji bi jk kj

∂  α
+ α b α b
− −
∂x j  
ki bi kj bj ki

b b
= 0 since ij = ji .

It is called cyclic property.

Hence, proved.
Proof of (iii)
α
a. Considering a contraction with the first index in Rijk , we get (i.e.,
α  = i)

∂  i
− ∂  i
+ i b i b
i
Rijk = − (b ↔ i )
∂x j   ∂x k  
ik ij bj ik bk ij

∂  ∂ ∂  ∂
= j 
∂x  ∂x k

( 
log g  − k  j log g  , g > 0
 ∂x  ∂x 
) ( )

∂2 ∂2
= j k log g − k j log g = 0
∂x ∂x ∂x ∂x
( ) ( )
∂
( )
i
 ij = log ± g .
∂x j
α
b. Considering a contraction with the third index in Rijk , we get (i.e.,
α = k)

∂  α
− ∂  α
+ α b α b
Rijαα = iα iα − bα
∂x j   ∂ xα  
ij bj ij

∂  ∂ ∂ ∂

( ) ( )
α α
log g  − α  + b b
= j  iα − log g , when g > 0
 ∂x  
ij bj ij
∂x  ∂x i
∂xb

∂2 ∂ ∂
( ) ( )
α α
log g − α  + b b
∴ Rij = iα − log g .
∂x  
ij bj ij
∂xi ∂x j ∂xb

This contracted tensor is called the Ricci tensor. In future, it will

have a great deal of application in the field equation of general the-
ory of relativity.
It can also be easily seen that Rij = Rji.
Riemann Symbols 99

6.5 Covariant Curvature Tensor

α
The fourth-order covariant curvature tensor (by virtue of definition Rijk ) is
defined as

Rhijk = g ha Rijk
a
. (6.5.1)

The symbols Rhijk are known as the Riemannian symbols of the first kind.
Now,

 ∂
Rhijk = g ha  j
 ∂x
( ) − ∂∂x ( ) +
ik
a
k ij
a
bj
a
ik
b
− bk
a
ij
b 


=
∂
∂x j (
g ha ik
a
) − ∂∂x ( g ) − ∂∂gx
k ha ij
a ha
j ik
a
+
∂ g ha
∂x k
ij
a
+ g ha bj
a
ik
b
− g ha bk
a
ij
b

=
∂
∂x j
( ) − ∂∂x ( ) − (
ik , h k ij , h hj , a + aj , h ) ik
a
+ ( hk , a + ak , h ) ij
a
+ bj , h ik
b
− bk , h ij
b

1 ∂  ∂ g ih ∂ g kh ∂ g ik  1 ∂  ∂ g ih ∂ g jh ∂ g ij 
Rhijk = + − h − + −
2 ∂ x j  ∂ x k ∂xi ∂ x  2 ∂ x k  ∂ x j ∂ x i ∂ x h 

α a α a α a α a
− gα a hj ik − gα h aj ik + gα a hk ij + gα h ak ij

α b α b
+ gα h bj ik − gα h bk ij
(b→ a ) (b→ a )

1  ∂2 g ∂2 g ∂2 g ∂2 g ∂2 g jh ∂2 g ij 
=  j ih k + j kh i − j ik h − k ih j − k i + k h 
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 

α a α a
+ gα a hk ij − gα a hj ik

1  ∂2 g ij ∂2 g kh ∂2 gik ∂2 g jh 
=  h k + i j − j h − i k
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 

α a α a
+ gα a hk ij − gα a hj ik , (6.5.2)

which is the expression of the curvature tensor of the first kind.

100 Tensor Calculus and Applications

6.6 Properties of the Curvature Tensor Rhijk of the First Kind

i. Rhijk = − Rihjk .
ii. Rhijk = −Rhikj (antisymmetric with respect to the second pair of
indices).
iii.
Rhijk = Rjkhi (symmetric with respect to the first and second pairs of
indices).
iv. Riijk = Rhikk = 0.
v. Rhijk + Rhjki +Rhkij = 0.

By definition (6.5.2), we get

1  ∂ g ij 
2
∂2 g hk ∂2 g ik ∂2 g hj α a α a
Rhijk =  + − −  + gα a hk ij − gα a ik hj .
2  ∂x h ∂x k ∂xi ∂x j ∂x h ∂x j ∂xi ∂x k 

Interchanging h and i,

1  ∂2 g hj ∂2 g ik ∂2 g hk ∂2 g ij 
Rihjk =  i k + h j − i j − h k
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 
α a α a
+ gα a ik hj − gα a ij hk (α ↔ a ).
Comparison gives the following result:

i. Rhijk = −Rihjk, i.e., antisymmetric with respect to first pair of indices.

ii. Similarly, it can be proved Rhijk =−Rhikj, i.e., antisymmetric with
respect to second pair of indices.
iii.
Rhijk = Rjkhi, i.e., symmetric with respect to the first and second pair of
indices.
Proof
From definition (6.5.2), it follows

1  ∂ g ij ∂2 g hj 
2
∂2 g hk ∂2 g ik α a α a
Rhijk =  h k + i j − h j − i k  + gα a hk ij − gα a hj ik
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 

and

1  ∂2 g kh ∂2 g ji ∂2 g ki ∂2 g jh  α a
R jkhi = + − − + gα a
2  ∂ x ∂ x ∂ x ∂ x ∂ x k ∂ x i 
j j ij kh
i k h
∂x ∂x h

α a
− gα a ki jh (α ↔ a).
Riemann Symbols 101

α ∂2 g ik ∂2 g kiα
Clearly, Rhijk = Rjkhi ∴ ij = j = ji and
, i.e., symmetric
∂x ∂x
h
∂x j ∂x h
with respect to the two pairs of indices.
Hence, proved.
iv.
Riijk = Rhikk = 0.
Replacing k by i and j by k, it can easily be proved from definition.
v. Rhijk + Rhjki +Rhkij = 0
α
Using the definition of Rijk , it is proved in Section 6.4 (ii) that
α α α
Rijk + R jki + Rkij = 0, the cyclic property.

Multiplying it by g hα and summing over α ,

α
g hα Rijk + g hα Rαjki + g hα Rαkij = 0

∴ Rhijk + Rhjki + Rhkij = 0.

Hence, proved.

6.7 Bianchi Identity
Let Po be the pole of a geodesic coordinates xi for which
i ∂ Ai
jk = 0 = ij , k and so Ai , j = , Ai is any tensor (vector).
∂x j
Now from definition of curvature tensor of the second kind, we have

a
Rijk =
∂
∂x j ( ) − ∂∂x ( ) +
ik
a
k ij
a
αj
a
ik
α
− αk
a
ij
α
. (6.7.1)

Considering xl-covariant differentiation of (6.7.1) with respect to gij, we get

∂2
( ) − ∂x∂∂x ( )
2
a a
,l =
a
Rijk ik ij at pole Po
∂xl ∂x j l k

Similarly,

∂2
( ) − ∂x∂ ∂x ( )
2
a a
,j =
a
Rikl il ik at pole Po
∂x j ∂x k j l

and

∂2
( ) − ∂ x∂ ∂ x ( )
2
a a
Rilja , k = ij j il at pole Po .
∂x k ∂xl k
102 Tensor Calculus and Applications

Addition of all these three relations gives

, l + Rikl , j + Rilj , k = 0 at pole Po . (6.7.2)

a a a
Rijk

Each of the terms of the equation is a tensor, so it holds for all coordinate
systems and at all points. Hence, it is an identity instead of an equation. It is
known as Bianchi identity.
The inner product of (6.7.2) with gha (summing over a) gives

(g ha
a
Rijk ) + (g
,l
ha
a
Rikl ) + (g
,j
ha Rilja ) ,k
= 0  g ha , k = 0, etc.

Rhijk ,l + Rhikl , j + Rhilj , k = 0 .

It is an alternative form of the Bianchi identity.

6.8 Einstein Tensor Is Divergence Free

, l + Rikl , j + Rilj , k = 0 can be written as
a a a
The Bianchi identity Rijk

a
Rijk a a a
(
, l − Rilk , j + Rilj , k = 0 Rikl , j = − Rilk , j .
a
)
Considering a contraction with respect to a and k,

, l − Rila , j + Rilj , a = 0
a a a
Rija

(R a
ija = Rij , Rii is Ricci tensor . )
An inner multiplication of it with gil gives

(g R ) − (g R ) + (g R )
il
ij
,l
il
il
,j
il a
ilj
,a
=0

 g il is covariant constant

Rlj , l − R, j + R aj , a = 0.

It can be written as 2R ij , i − R, j = 0, changing l → i and a → i.

 1 
∴  R ij − δ ji R  = 0 .
 2  ,i
Riemann Symbols 103

1
Hence, G ij , i = 0, where G ij = R ij − δ ji R is the Einstein tensor.
2
∴ The Einstein tensor is divergence free ( contraction is with respect to
covariant derivative index which is required for definition of divergence).

6.9 Isometric Surfaces
The intrinsic geometry of a surface is based on the corresponding funda-
mental quadratic form or metric ds2 = aαβ duα uβ with surface coordinates
uα = uα (u1 , u2 ). The intrinsic properties such as the lengths of curves and
the angle between two intersecting curves primarily depend on the metric
tensor of the surface and its derivatives. If there exists a coordinate system
which characterizes the linear element of two surfaces S1 and S2 by the same
metric aαβ , then they are called isometric. The corresponding parameters for
the transformation is known as isometry. From the Euclidean plane, surfaces
of cylinder and cone can be constructed by means of rolling without chang-
ing arc length, areas, and measurement of angles. Hence, they are isometric
with the Euclidean plane.

6.10 Three-Dimensional Orthogonal Cartesian

Coordinate Metric and Two-Dimensional Curvilinear
Coordinate Surface Metric Imbedded in It
In order to enter into the threshold of geometry of surfaces in a surrounding
space, we need to consider two distinct coordinate systems. Let uα (u1 , u2 )
be the two curvilinear coordinates of the surface S imbedded in a three
orthogonal Cartesian coordinate systems in E3. But the intrinsic property of
geometry of a space is characterized by the metric or quadratic differential
form.
Let xi = xi (y1, y2, y3), i = 1, 2, 3, be the orthogonal Cartesian coordinates cov-
ering the space E3 and xi = xi (u1, u2) be the Gaussian surface coordinates of S
imbedded in E3. The line element in E3 is given by ds 2 = g ij dx i dx j, where

∂xi α ∂y k ∂y k
dx i = du and g ij =
∂uα ∂xi ∂x j

∂xi α ∂x j β
= g ij du du
∂uα ∂ uβ
104 Tensor Calculus and Applications

∂xi ∂x j α β
ds2 = g ij du du = aαβ duα duβ ,
∂uα ∂uβ
where (6.10.1)

∂xi ∂x j
aαβ = g i j ,
∂uα ∂uβ

so that a = gJ2,|aαβ |= a, | g i j |= g.
∂xi α
Looking at dx i = du , it can be concluded that dx i is a space vector and
∂uα
is surface invariant, and duα is a surface vector and is space invariant.

6.11 Gaussian Curvature of the Surface S immersed in E 3

( )
If the functions aαβ and bαβ = g ij in (6.10.1) related to some surface, then xi is
to satisfy the condition of integrability:

∂2 xαi ∂2 xαi
β γ = , (6.11.1)
∂u ∂u ∂uγ ∂uβ

∂xi
where xαi = is continuously differentiable of degree 2.
∂uα
∂xi
Now denoting α by tαi which is tangent to the surface and considering
∂u
the covariant (surface) derivative of it, we can get

∂tαi i j γ ∂2 x i i j γ
tαi , β = + t tβk −
jk α
i
αβ tγ = + t tβk −
jk α
i
αβ tγ .
∂ uβ ∂uα ∂uβ

Since tαi is tangent to the surface, its partial derivative tαi , β is normal to the sur-
face; otherwise, it is proportional to the normal ni of the surface. Therefore,
tαi , β = bαβ ni
or

 ∂2 x i i γ 
bαβ = tαi , β ni =  α β + jk tαj tβk − αβ tγi  ni. (6.11.2)
 ∂u ∂u 

But for the use of Cartesian coordinates and geodesic surface coordinates,
the Christoffel symbols can be made to zero at a particular point. Of course,
the derivatives of the Christoffel symbols in space
i
jk will vanish but ( )
γ
not the symbols with surface αβ . ( )
Riemann Symbols 105

∂3 x i ∂  ∈  i
Hence, tαi , βγ = − αβ t∈.
∂u ∂uβ ∂uγ ∂uγ 
α 
∂3 x i ∂  ∈  i
Similarly, tα ,γβ = α γ β − β γα t∈.
i
∂u ∂u ∂u ∂u  

∈
∴ tαi , βγ − tαi , γβ = Rαβγ t∈i . (6.11.3)

∈
Using (6.11.1) and (6.11.2) and making use of definition for Rαβγ
From tα ,  β = bαβ n , we can get
i i

tαi ,  βγ = bαβ , γ ni + bαβ ni , γ

= bαβ , γ ni − bαβ bγδ aδε t i ε .

( ni , α =   −   bαβ a βγ t iγ  from Weingarten ′s formula due to the linear 

combination of nαi  and tβi , namely, n, i α = Cαβ tβi )

∴ tαi ,  βγ − tαi , γβ = ( bαβ , γ − bαγ ,  β ) ni + ( bαγ bβδ − bαβ bγδ ) aδε tεi .

ε
∴ Rαβγ tεi = ( bαβ , γ − bαγ ,  β ) ni + ( bαγ bβδ − bαβ bγδ ) aδε tεi , using (6.11.3)

∴ bαβ , γ − bαγ ,  β = 0, (6.11.4)

multiplying by ni and using nitεi = 0.

This is known as Codazzi equation of surface. Putting this value, the
above equation can be reduced to

ε
Rαβγ tεi = ( bαγ bβδ − bαβ bγδ ) aδε tεi .

Therefore, for arbitrary tεi , it can be transformed to

Rλαβγ = bαγ bβλ − bαβ bγλ . (6.11.5)

 aλ ∈aδ ∈ = δ λδ .

This is known as Gauss equation of surface.

But for Riemannian curvature tensor Rλαβγ , Rααβγ = Rλαββ = 0, the only sur-
viving component is R1212 = − R2112 = − R1221 = R2121.
( )
Since the surface x i = x i u1 ,  u2 in two-dimensional curvilinear coordi-
nates is immersed in E3 with the metric aαβ  or  aαβ = a, the quantity defined
R
by κ = 1212 is called the total curvature or the Gaussian curvature.
a
2
But R1212 = b11b22 − b12 b21 = b11b22 − b12 = b, from (6.11.5) so that
106 Tensor Calculus and Applications

2
b11b22 − b12  b
κ=   = 
a  a
2
b11b22 − b12
= 2
a11 a22 − a12
is the Gaussian curvature of two-dimensional surface.

Example 1
Prove that the differential equation Ai,j = 0 is integrable only when the
Riemann Christoffel tensor vanishes.
It needs to show that Rijka
= 0, if the differential equation Ai,j = 0 is
integrable.
Now,
∂ Ai a
Ai , j = 0 gives − Aa ij = 0. (i)
∂x j

∂ Ai j
∫A
a a
dx = Aa ij dx j ∴ Ai = a ij dx j . (ii)
∂x j
This shows that the right-hand side of (ii) must be integrable, and hence,
it should be a perfect differential of some function, say Bi, so that
a
Aa ij dx j = dBi

∂Bi j a
∴ dx = Aa ij dx j
∂x j

 ∂Bi a  j
 j − Aa ij  dx = 0.
∂x
∂Bi a
It gives = Aa ij  dx j is arbitrary.
∂x j
Differentiating partially with respect to xk, we get

∂2 B
=
∂ Aa
∂x j ∂x k ∂x k
ij
a
+ Aa
∂
∂x k ( ). (iii)
ij
a

Interchanging j and k, it can be written as

∂2 B
=
∂ Aa
∂x k ∂x j ∂x j
ik
a
+ Aa
∂
∂x j ( ). (iv)
ik
a

From (iii) and (iv),

∂ Aa a  ∂ a  ∂ Aa a  ∂ a 
ik + Aa  j ik  − k ij − Aa  k ij  = 0
∂x j  ∂x ∂x  ∂x

Ab
b
aj ik
( a ↔ b)
a
+ Aa
∂
∂x j ( )− A
ik
a b
b ak ij
( a ↔ b)
a
− Aa
∂
∂x k ( ) = 0.
ij
a
Riemann Symbols 107

(making use of (i))

 ∂
∴ Aa  j
 ∂x
( ) − ∂∂x ( ) +
ik
a
k ij
a
bj
a
ik
b
− bk
a
ij
b 
 = 0.

a
Aa Rijk = 0 ∴ Rijk
a
= 0 Aa is arbitrary in the inner product.
Hence, proved.

Example 2
Show that the number of independent components of the Riemannian
curvature tensor of the first kind Rhijk in n-dimensional space Vn is
1 2 2
2
(
n n −1 . )
In general, the number of independent components of the fourth-order
tensor Rhijk in a Riemannian space Vn is n4. But due to the following prop-
erties, the number of independent components will be reduced from n4.

i Rhijk = − Rihjk (antisymmetric property with respect to h and i).

ii Rhijk = − Rhikj (antisymmetric property with respect to j and k).
iii Rhijk = R jkhi (symmetric with respect to the first and second pairs).
iv Rhijk + Rhjki + Rhkij = 0 (cyclic property).

Case I: When there is only one distinct index of the type Rhhhh

Rhhhh = − Rhhhh ( due to (i)) ∴ Rhhhh = 0.

Hence, there is no component of Rhijk with one distinct index.

Case II: When there are two distinct indices of the type Rhihi ( h ≠ i )
The two distinct indices from n values can be selected in n(n − 1) ways
which correspond to n(n − 1) numbers of independent components of
Rhijk.
But Rhihi = − Rihhi = Rihih, due to (i) and (ii).
After interchanging h and i in the first, the last one can be recovered.
1 1
∴ The number n(n − 1) is reduced only by so that it becomes n(n − 1).
2 2
Also, Rhihi = Rhihi (due to (iii)), so there is no reduction due to this
property.
Interestingly, the cyclic property

Rhihi + Rhhii + Rhiih = Rhihi − Rhihi = 0

is identically satisfied due to (ii).

Hence, there is no reduction due to the cyclic property.
1
∴ The number of independent components in this case is n(n − 1).
2
Case III: When there are three distinct indices of the type Rhihj
∴ The values of h, i, j can be selected from n values in n(n − 1)(n − 2)
ways, and hence, the number of independent components of Rhijk in this
case is n(n − 1)(n − 2). But this number will be reduced for the properties
(i)–(iv).
108 Tensor Calculus and Applications

Now, Rhihj = Rhjhi (due to (iii)) which can be recovered by merely inter-

changing i and j
 = −Rjhhi (due to (i))
 = −Rhijh (due to (iii)) which is nothing but (ii).
Hence, due to symmetric properties, the number of independent com-
1 1
ponents is finally reduced by only to yield n(n − 1)(n − 2).
2 2
The cyclic property Rhihj + Rhhji + Rhjih = Rhihj + 0 + Rihhj = Rhihj − Rhihj = 0
is identically satisfied, and hence, there is no reduction of the above
number.
Case IV: When all the four indices are distinct of the type R hijk
Clearly, the values of h, i, j, k can be selected in n(n − 1)(n − 2)(n − 3)
ways, and hence, the number of independent components in this case is
n(n − 1)(n − 2)(n − 3).
But due to the three symmetric properties, (i)–(iii), it is reduced to
1 1 1
× × n(n − 1)(n − 2)(n − 3).
2 2 2
But from (iv), we get Rhijk + Rhjki = −Rhkij.
This shows that, knowing two components, the third can be readily
determined. Hence, due to (iv), the number of independent components
2
is reduced by .
3
∴ In this case, the number of independent components of R hijk is
1 2
× n(n − 1)(n − 2)(n − 3).
8 3
Hence, the total number of independent components of the Rieman­
nian curvature tensor R hijk is

1 1 1
0+ n(n − 1) + n(n − 1)(n − 2) + n(n − 1)(n − 2)(n − 3)
2 2 12
1
= n(n − 1) [ 6 + 6(n − 2) + (n − 2)(n − 3)]
12

1
= n(n − 1) 6n − 6 + n2 − 5n + 6 
12
1 1 2 2
= n(n − 1)n(n + 1) = n (n − 1).
12 12

Hence, proved.

Exercises
1
1. Show that the divergence of the tensor R ij − δ ji R is identically zero.
2
2. If the metric of V2 formed by the surface of a sphere of radius r is
ds2 = r2 (dθ 2 + sin2θ dϕ 2) in spherical polar coordinates, show that
R1212 = r2sin2 θ.
Riemann Symbols 109

a
3. Derive the expression of the curvature tensor of the second kind Rijk .
4. Derive the expression of the curvature tensor of the first kind Rhijk .
5. Discuss the properties of Rhijk .
6. For a V3 referred to a triply orthogonal coordinate system, prove that
1
Rij = Rikkj (where i ≠ j ≠ k)
g kk
1 1
and Rhh = Rhiih + Rhjjh .
g ii g jj
7. Calculate the Ricci tensor Rij for the metric on the sphere ds2 = a2
(dθ 2 + sin2 θ dϕ 2), where i, j = 1, 2, x1 = θ and x2 = ϕ, and a is constant.
α i ∂
[Hint: Find ij and use ij = j (log g ).]
∂x
8. Show that on a two-dimensional surface, the curvature tensor is
completely defined by a single component, say R1212.
9. For a surface with the metric ds2 = (du)2 + λ 2 (dv)2 , show that the
1 ∂2 λ
Gaussian curvature is − .
λ ∂u2
( )( ) ( )( )
2 2 2 2
10. Show that the surface with the metric ds2 = u2 du1 + u1 du2
is isometric or developable.
[Hint: Show κ  = 0.]
11. Find the conditions that the surfaces S1 : y 1 = v 1 cos v 2 ,
 v1 
y 2 = v 1 sin v 2 , y 3 = a cosh −1   ; S2 : y 1 = u1 cos u2, y 2 = u1 sin u2 , y 3 = au2
 a
are isometric.
[Hint: Show that the two metrics are the same subject to some
conditions.]
 Part II

Application of Tensors
7
Application of Tensors in General
Theory of Relativity

As mentioned in Chapter 1, the knowledge of the geometry of space is

important for Newtonian (classical) as well as Einstein’s relativistic mechan-
ics which is reflected in the statement “Dynamics deals with the geometry of
motion.” To develop geometry of a space, the paramount importance is the
assumption of coordinate systems to suitably describe the space concerned
based on the corresponding metric. Deformation is an essential notion to
invite the concept of “tensors” in non-isotropic medium from an applicable
point of view in mathematical science. Tensors being independent of any
coordinate system possess the intrinsic property of the geometry of space.

7.1 Introduction
The general theory of relativity is known as the theory of gravitation. For
applicability of principle of relativity [5, p. 17] to preserve fully the privileged
position among all conceivable frames of reference, the concept of special
theory of relativity (STR) based on Lorentz transformation [5, p. 39] is devel-
oped. To develop the new theory of gravitation, the idea of privileged posi-
tion, namely, inertial frames is destroyed to include the most general form
of transformation applicable to any positive integral number of dimensions.
Setting aside the Galileo’s view of “law of inertia” and the amount of gravi-
tational action of one mass point on another (great mass) of Newton (Kepler),
Einstein gave a different interpretation. He has concluded that gravity is not
a force (as Newton believed), but the curvature of space–time, and the matter
is the source for it and material objects create the gravitational field, which
distorts (deforms) or curves the surrounding space–time as the magnet sets
up the magnetic field. So, the generating space due to the presence of mate-
rial objects demands the use of non-Euclidean geometry of curved space for
its true description. Otherwise, curvilinear coordinates are essential at large
as rectilinear coordinates cannot be set up in the curved region of space–
time. Hence, curvature of the space–time continuum is the fundamental
ingredient for characteristic representation of gravitational theory or the
general theory of relativity. Eventually, tensors applicable to all coordinate

113
114 Tensor Calculus and Applications

systems developed in Riemannian space of n dimensions (a manifold) are

the best tool (as detected by Einstein) to study general theory of relativity
[5,6]. There is no gravity and no curvature, so free particles follow geode-
sics (shortest path), i.e., straight lines, when the space–time is flat in special
theory of relativity.
Therefore, to enter into the threshold of general theory of relativity, it is
essential to know the curvature of the space–time, a four-dimensional mani-
fold which is a clear deviation from flat space of special theory of relativity.

7.2 Curvature of a Riemannian Space

Riemann adopted the Gaussian curvature (Section 6.11) of a geodesic surface
 
S at a point P determined by the orientation of the unit vectors p( p i ) and q(q i )
at P as the definition of (Riemannian) curvature of Vn at that point. The pen-
cil of directions can be expressed as

( )
y i = α p i + β q i s (7.2.1)

 dy i 
with unit tangent ti =  of geodesics at P, where α and β given by α s = u1
 ds 
and β s = u2 are the current coordinates of points on S generated through

y i = u1 p i + u2 q i . (7.2.2)

If the metric of Vn is denoted by gij dyi dyj and that of the surface S by
aαβ duα duβ (α, β = 1, 2 ) , then

dy i dy j
aαβ = g ij . (7.2.3)
duα duβ
∴ Using Equation (5.8.5), we can write

( )
ij , k
g
 ∂2 y β
= aβ c  i j +
 ∂x ∂x
( )
αγ
β

a
∂ yα ∂ yγ  ∂ y c

∂xi ∂x j  ∂x k

∂y c ∂2 y β ∂ yα ∂ yγ ∂ y c
= aβ c
∂ u k ∂ ui ∂ u j
+ ( ) αγ , c
a ∂ ui ∂ u j ∂ u k
Changing variables of the present systems; in this assumption αγ , c for aαβ (uα)
and ij , k for gij(yi).
∂ yα ∂ yγ ∂ y c
( ) ( )
∴ ij , k = αγ , c
g a ∂ ui ∂ u j ∂ u k
, using (7.2.2) where pi, qi are constants.
Application of Tensors 115

Hence,

( αβ ,γ ) g
= g kl
∂y i ∂y j ∂y k
∂uα ∂uβ ∂ yγ ( ) (7.2.4)
ij
l

i ↔α

γ → β↔ j
changing on both sides.
c → γ ↔k

But yi ’s are the Riemannian coordinates of geodesics of Vn for the metric

gij dyi dyj and Christoffel symbol αβ ,γ = 0. ( )
( )
g
l
∴ ij = 0 at the origin P, from (7.2.4).
a
Hence, the Riemannian curvature tensor Rh𝛼𝛽𝛾 for the surface S can
take nonzero values R1212 for two values α,β = 1,2 subject to the properties
Rhαβγ = − Rα hβγ , Rhαβγ = − Rhαγβ , Rhαβγ = Rβγ hα  Rhhβγ = Rhαββ = 0.
∴ The nonvanishing Riemannian curvature tensors are

R1212 , − R2112 , − R1221 , and R2121

 In two-dimensional space, the number of independent components is

1
 N 2 ( N 2 − 1)  = 1 only, namely, R1212.
12 
∴ From transformation laws of tensors,

∂uh ∂uα ∂uβ ∂uγ

′ = Rhαβγ
R1212
∂u′ 1 ∂u′ 2 ∂u′ 1 ∂u′ 2
2 2
 ∂u1   ∂u2  ∂u2 ∂u1 ∂u1 ∂u2
= R1212  − R
 ∂u′ 1   ∂u′ 2 
1212
∂u′ 1 ∂u′ 2 ∂u′ 1 ∂u′ 2

∂u1 ∂u2 ∂u2 ∂u1

− R1221 (7.2.5)
∂u′ 1 ∂u′ 2 ∂u′ 1 ∂u′ 2
∂u2 ∂u1 ∂u2 ∂u1
+ R2121
∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2
2
 ∂u1 ∂u2 ∂u1 ∂u2 
∴ R1212
′ = R1212  − .
1
 ∂u′ ∂u′ 2
∂u′ 2 ∂u′1 

The Gaussian curvature κ (Section 6.11) which is invariant for coordinate

transformations is defined as
116 Tensor Calculus and Applications

′
R1212 R1212 a11 a12
κ= = where a = aαβ =
a a′ a21 a22

R1212 ′
R1212
= =  a′ = a J 2 ,
a (
a a11a22 − a 212 )
∂u
where J = and a = aαβ , a′ = aαβ
′ .
∂u′
Now, by definition,

Rhijk = g ha Rijk
a  ∂
= g ha  j
 ∂x
( ) − ∂∂x ( ) +
ik
a
k ij
a
bj
a
ik
b
− bk
a
ij
b 

(7.2.6)
∴ R1212
∂
= 1
∂u
( )22,1
a
∂
− 2
∂u
( ). 21,1
a

 ( αβ ,γ ) a
= 0 at the origin P for Riemannian coordinates.
If L, M, N are the second-order and E, F, G are the first-order magnitudes
of Gaussian surface* characterized by ds 2 = Ldu2 + 2 Mdudv + Ndv 2 and
ds 2 = Edu2 + 2 Fdudv + Gdv 2, respectively, then
R1212 LN − M 2
=
a EG − F 2
a = EG − F 2
= a11a 22 − a 212 ,
where E = a11, F = a12 = a21, G = a22.
cofactor of a11 in aαβ = a a
But a11 = = 22 .
a a
12 a12 22 a11
Similarly, a = − ,a =
a a
and

∂y i ∂y j
a11 = g ij = g ij p i p j = g hj p h p j
∂u1 ∂u1
∂y i ∂y j
a22 = g ij = g ij q i q j = g ik q i q k
∂ u2 ∂ u2
∂y i ∂y j (7.2.7)
a12 = g ij = g ij p i q j = g hk q k p h = g ji p j q i
∂u1 ∂u2
∴ a = a11a22 − a 212 = g hj g ik p h p j q i q k − g hk g ij p h p j q i q k

= p h p j q i q k ( g hj g ik − g hk g ij ).

* Willmore [4].
Application of Tensors 117

Now, from (7.2.4),

( 22,1 ) a
= g hl
∂y i ∂y j ∂y h
∂u2 ∂u2 ∂u1 ( ) l
ij
g
( k → h)

= g hl q i q k p h ( ) l
ik
g
( j → k)
(7.2.8)
∂
∴ 1
∂u
( 22,1 ) a
= g hl q q p
∂
i k
∂y j
h
( ) l
ik
g
∂y j
∂u1

= g hl q i q k p h p j
∂
∂y j ( ). l
ik
g

Also,

( 21,1 ) a
= g hl
∂y i ∂y j ∂y h
∂u2 ∂u1 ∂u1 ( ) l
ij
g

= g hl q i p j p h ( ) l
ij
g
(7.2.9)
∂
∴ 2
∂u
( 21,1 ) a
= g hl q p p
i ∂
j
∂y k
h
( ) l
ij
g
∂y k
∂ u2

= g hl q i p j p h q k
∂
∂y k ( ). l
ij
g

Hence, (7.2.6) can be written as (using 7.2.8 and 7.2.9)

 ∂
R1212 = g hl p h p j q i q k  j
 ∂y
( ) − ∂∂y ( )
l
ik k
l
ij

= p h p j q i q k g hl Rijk
l
= p h p j q i q k Rhijk .

∴ The curvature

R1212 p h p j q i q k Rh ijk
κ= = h j i k (7.3.10)
a p p q q g hj g ik − g hk g ij ( )
is invariant.
This is the mathematical expression of curvature of the Riemannian space
Vn with Rhijk. This ascertains the nomenclature “covariant curvature tensor”
for Rhijk. Hence, Rhijk (or Rhijk) characterizes the behavioral properties of the
curve space of general theory of relativity.
118 Tensor Calculus and Applications

7.3 Flat Space and Condition for Flat Space

Definition

If the curvature κ of the Riemannian space Vn, namely Rhijk vanishes, it is

called flat space.

Condition for Flat Space

From the definition of curvature shown in (7.3.10), where Rhijk is the covariant
curvature tensor, and pi, qj are the unit vectors showing orientation at the
origin P of geodesics.

p h p j q i q k Rhijk
κ= j i k .
p p q q ( g hj g ik − g ij g hk )
h

For flat space κ = 0 ⟹ Rhijk = 0, the bracket in the denominator must not be
zero.
( a
)
∴ If Rhijk = g ha Rijk = 0, or if all the components of Rhijk or Rijk
a
are zero, then
the curvature will be zero, and the space will be flat. This is the required
condition for flat space.
1  ∂2 g ij ∂2 g ∂2 g ∂2 g hj  a b a b
But Rhijk =  h k + i hk j − h ik j − i k  + g ab hk ij − g ab hj ik .
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x 
b
If gij ’s are constants, ij , k = ij = 0, ∴ Rhijk = 0.
Hence, gij = constants are basically the conditions for flat space.
Note: STR is restricted to flat space only.

7.4 Covariant Differential of a Vector

Let Ai be the components of a vector at the point xi and Ai + dAi be the compo-
nents at a neighboring point xi + dxi in a vector field. Therefore, the difference
dAi of the two vectors Ai + dAi and Ai being ordinary differential is a vector.
∂ x′ i k
The transformation of the coordinates gives dx′ i = dx , otherwise,
∂ x′ i k ∂x k
A′ =
i
A stands for the transformation of a vector.
∂x k
But

∂ x′ i ∂ 2 x′ i
dA′ i = dA k
+ dx j A k , (7.4.1)
∂x k ∂x j ∂x k
Application of Tensors 119

which is not a vector for the presence of the second term. Of course, for
linear transformation belonging to rectilinear coordinate systems, this
­transformation characteristically behaves like a vector since the second term
vanishes in that case. On the other hand, the difference of two vectors needs
to be a vector in a general coordinate system. Hence, to get this difference to
be a vector in a curvilinear coordinate system, it necessitates to translate a
vector Ai at xi to the location of xi + dxi of the other vector Ai + dAi so that they
are located at the same point. This translation is related to parallel transla-
tion to itself. In a general curvilinear coordinate system, this translation by
itself changes the components of the vector, and the changes are denoted by
δAi different from ordinary differential dAi. From (7.4.1), it is observed that
this change δAi should be proportionate to both the vectors Ak (i.e., Ai) and
 ∂ 2 x′ i 
the displacement dxj  if j k ≠ 0 .
 ∂x ∂x 
Hence, the difference between the original vector Ai + dAi and the trans-
ported vector Ai + δAi at the point (xi + dxi) is given by

( DA ) = dA
i i
– δ Ai . (7.4.2)

∴ In consistent with the above changes in the components Ai of the vector

due to translation and the components of the displacement dxi, the changes
δAi in (7.4.2) can be written as†
i
δ Ai = − jk A k dx j . (7.4.3)
i
( are some functions of coordinates xi)
jk
Thus, (7.4.2) is called the covariant differential of the given vector Ai.

7.5 Motion of Free Particle in a Curvilinear

Co-Ordinate System for Curved Space
dx i
If ui = (i = 1, 2, 3, 4) are the four vectors (velocity) tangential to the time
ds
track of a free particle, it is represented in a rectilinear coordinate system by
the equations dui = 0.
This gives xi = As + B which is the intrinsic equation of a straight line in a
four-dimensional continuum. Eventually, dui = 0 characterizes the uniform
rectilinear motion in three-dimensional physical space also. But to write the
corresponding equations of motion in a curvilinear coordinate system, we

† Section 6.4 of Ref. [5].

120 Tensor Calculus and Applications

are to make use of covariant differential instead of ordinary differential.

Hence, the equations of motion of free particle in the curvilinear coordinate
system are given by

( Du = ) du – δ u
i i i
= 0 (7.5.1)

i
i.e., dui + jk uk dx j = 0 from Equations (7.4.2) and (7.4.3),

d2 xi i dx j dx k
i.e. , + jk = 0 (7.5.2)
ds2 ds ds

which are the differential equations of geodesics, and contextually, they

characterize the world line (track) of free particle in the general coordinate
system.
j
d2 xi i dx dx
k
If we compare Equation (7.5.2) or 2 = − jk with Newtonian grav-
ds ds ds
2 i
d x dφ i
itational equation with potential ϕ, namely, 2
= −δ ij j , then jk can be
dt dx i
attributed to “forces” that arise inherently in the system. Hence, jk ≠ 0 or
g ij ≠ constant is the generating factor of the potential (ϕ) and responsible for
the curvature of the path of free particle caused by gravitational field or mass
energy of matter in curved space.

7.6 Necessity of Ricci Tensor in Einstein’s

Gravitational Field Equation
In classical (Newtonian) mechanics, the field equation in the presence of
matter according to Newton’s law of gravitation is described by Poisson’s
equation:

∇ 2φ = 4πγρ ,
∂2 φ ∂2 φ ∂2 φ (7.6.1)
i.e. , 2
+ 2 + 2 = 4πγρ ,
∂x ∂y ∂z

where ϕ is the gravitational potential, ρ is the density of distributed matter,

and γ is the gravitational constant. So, to get its general relativistic analogue,
we need to correlate these quantities suitably under physical system intro-
ducing general relativistic concepts.
Application of Tensors 121

1. It is discussed is Section 7.1 that the material energy creates the grav-
itational field to turn the space–time continuum into a curved space.
Therefore, the counterpart of Equation (7.6.1) should be made appli-
cable to curved space.
2. For the “principle of covariance,”‡ an essence of general relativity, all
natural laws must be expressed in tensor forms (Covariant) for their
validity in all coordinate systems including non-inertial frames
of curved space required for general theory of relativity. Hence,
Equation (7.6.1) needs to be expressed completely in tensor form.
3. It is shown in Section 7.5 that the equations of motions of free fall of
a particle in curvilinear coordinates of space–time are characterized
j
d2 xi i dx dx
k
by the geodesic equations 2
+ jk = 0 , which is reducible
ds ds ds
2 i
d x ∂φ
to Newton’s equation of motions 2 = − i (= −∇φ ), where ϕ is the
gravitational potential. dt dx
It has already been mentioned in Section 7.5 that gij (≠ constant)
 2φ 
is responsible to generate gravitational potential φ  g 44 = 1 ± 2 
 c 
[Section 7.3 of Ref. (5, p. 191)]. Hence, in relativistic theory of gravi-
tation, ϕ is to be replaced by the metric tensor g µυ (≠ constant) or
by some relation with g µυ . Moreover, the left-hand side (L.H.S.) of
Equation (7.6.1) does not involve the derivatives of ϕ higher than
two (second), so the replacement tensor (for covariant form) must
also contain the second-order derivatives of g µυ , but the Ricci
tensor Rµυ is the tensor with second-order derivatives of g µυ as
∂
( )
α ∂ α
( )
Rµυ = υ µα − α µυ + βυ µα − βα µυ
∂x ∂x
α β α β

α 1  ∂g ∂g ∂g 
with µυ = gα a µυ , a = gα a  µυa + υµa − µυa 
2  ∂x ∂x ∂x 
∂  α  ∂2
and υ µα = υ µ (log ± g ).
∂x   ∂x ∂x
Hence, the suitable relativistic analogue of the L.H.S. of Equation
(7.6.1) is the Ricci tensor, if it satisfies some other conditions in con-
formity with the right-hand side (R.H.S.).
4. In agreement with the L.H.S. of (7.6.1), we need a second-order tensor
for the R.H.S. in place of density ρ of matter.
The gravitational field is the outcome (effect) of mass distribution
or mass–energy distribution of matter, and energy momentum ten-
sor Tµυ or T µυ characterizes the cause of the distributions. Hence,
the density ρ of material particles generating the force of gravity is

‡ Section 6.1 of Ref. [5, p. 155].

122 Tensor Calculus and Applications

to be represented by the second-order “energy momentum” tensor

T µυ >(or Tµυ ).
Also, the energy momentum tensor (T µυ or Tµυ ) of a closed system
(like one comprising the material distribution and the force, or the
field together) requires to be bounded by the conservation laws
( )
Tυµ, µ = 0 or gαυ Tυµ = 0 or T,αµµ = 0,
,µ
which is divergence free and T αµ = T µα , i.e., symmetric.
5. In light of conservation laws Tυµ, µ = 0 as satisfied by the energy
momentum tensor, it is desirable to make it applicable to the Ricci
 1 
tensor Rµυ also. But the Einstein tensor Gυµ =  Rυµ − δ υµ R  in terms of
 2 
µ  µ 1 µ 
Ricci tensor is divergence free, i.e., Gυ , µ =  Rυ − δ υ R  = 0.
 2  ,µ
 1   1 
Otherwise,  gαυ Rυµ − gαυδ υµ R  = 0 or  Rαµ − gαµ R  = 0 .
 2  ,µ  2  ,µ
µα αµ
Also, R = R , which is symmetric.
Hence, the Ricci tensor R µα (or Rµυ ) also satisfies the required other
conditions like T µυ .
∴ Based on the aforesaid requirements (1)–(5), Einstein adopted the
1
field equations for general theory of relativity as R µυ − g µυ R = − KT µυ
1 2
or Rµυ − g µυ R = − KTµυ .
2
These are the essential field equations for general theory of relativity,
which were expanded subsequently after various cosmological consider-
ations. This is purely a subject matter of “general theory of relativity,” and it
is beyond the scope of the book, to discuss it completely.
Of course, in the absence of matter (or gravitational field),

T µυ = T µυ = 0, g µυ = constant.

The field equations reduce to R µυ = 0 = Rµυ for empty space which is identical
to ∇ 2φ = 0, which is Newton’s vacuum equation.
N.B.: The importance of Ricci tensor in general theory of relativity is
beyond description. In this pursuit, it is amplified only the use of tensors
(primarily the Ricci tensor) in general theory of relativity.
8
Tensors in Continuum Mechanics

8.1 Continuum Concept
In the investigation of material behavior of a body or medium, the bulk of
the matter is considered as a whole but not the individual molecule. For this
reason, the observed macroscopic behavior is counted in general by assum-
ing that the material is continuously distributed throughout its volume and
completely fills the space it occupies instead of considering its molecular dis-
tributions. This continuum concept of matter is the fundamental postulate
of continuum mechanics.

8.2 Mathematical Tools Required for Continuum Mechanics

The physical quantities related to continuum mechanics for its in-depth
description are independent of any particular coordinate system for refer-
ence. Generally, from a mathematical point of view, these physical quan-
tities conveniently need to describe by means of referring them to some
coordinate system. Eventually, tensors which are independent of any par-
ticular coordinate system are the appropriate tools to adopt in this con-
sideration. Hence, physical laws of continuum mechanics are expressed in
terms of tensor equations. Usually, tensor transformations are linear and
homogeneous, and if they (laws of continuum mechanics) are expressed in
the form of tensor equations in one coordinate system, they remain valid
in any other coordinate system. This invariance of tensor equations under
coordinate transformations is one of the principal reasons (similar to gen-
eral theory of relativity) for the utility of tensors in continuum mechan-
ics. Of course, Cartesian tensors are sufficient to deal with continuum
mechanics, and hence, it can be developed with reference to Cartesian
tensors.
In Euclidean space of three dimensions, the number of components of a
tensor of order n is 3n. A vector is a tensor of order one with 31 components.

123
124 Tensor Calculus and Applications

The stress and strain are the second-order tensors having 32 = 9 components
in general. The physical quantities (mentioned earlier) involved in the study
of continuum mechanics are the stresses and strains, which are invariably
related to the deformation of media and bodies. Second-order tensors are
also known as “dyadics,” and the quantities in continuum mechanics are
represented by dyadics [11].

Definition

Stress: The forces occurring in a bulk of the material medium proportional

to the mass of the substance (e.g., gravity, magnetic force, centrifugal force)
are known as body forces, which are measured per unit volume.
The forces acting over the volume of bounding surface of a body and mea-
sured in units of force per unit area are called surface forces. This force per
unit area is called stress which gives the measure of the intensity of the
reaction of the material lying on one side of the material or that which lies on
the other side. For this inherent property, the tendency of deformation of the
state of a body or medium is bound to be acted by stress. Stress being a force
per unit area must have two directions those of the force and normal to the
area associated with it.
The stress tensor components perpendicular to the surface (or plane face)
are called normal stress (σ) and tangential to the surface (or plane face) are
called shearing stress (τ).

Definition

Strain: The deformation caused by stress which may be dilated resulting in

change in volume or distortion with changes in form or both is called strain.
Otherwise, a distortion, deformation of change in the position of particles
relative to each other, is known as strain.
The change in confining pressure can change in volume but in shape for
isotropic bodies where mechanical properties are uniform in all directions.
With increasing/decreasing confining pressure, the volume of the body
decreases/increases, and dilation is negative/positive accordingly.
Mathematically, if Δfi (fi body force) is the resultant force exerted across the
surface element Δs of S enclosing a volume V, then the Cauchy’s stress prin-
∆f
ciple states that the average force per unit area on Δs, namely, i , tends to
∆s
dfi
a finite limit as Δs → 0 at a point P of the surface. Symbolically, the stress
ds
 ∆f df
vector is written as t( nˆ )i = lim i = i when the moment of Δfi at the point P
∆s → 0 ∆s ds
vanishes in the limiting process. The stress vectors are different at the same
time for different surfaces containing the same point P. The stress principle
Tensors in Continuum Mechanics 125

is necessary to know the state of stress at a point in a medium in motion or a

body subjected to deformation.

8.3 Stress at a Point and the Stress Tensor


In a continuum, the Cauchy’s stress principle associates a stress vector t( nˆ i )
with each unit normal nˆ i representing the orientation of an infinitesimal sur-
face element having an arbitrary internal point P. 
The totality of all possible pairs of such combined vectors t( nˆ i ) and n̂
at P defines the state of stress at that point. Every pair of stresses and
normal vectors are essential to represent the state of stress at that point.
Eventually, the stress vectors on each of the three mutually perpendicular
planes (for Cartesian system) through P can give the state of stress. Of
course, coordinate transformation equations can serve to relate the stress
vector on any other plane at that point with the given three planes. The
state of stress is perfectly homogeneous under the application of a force
without rotation.
The stress vectors in the coordinate plane surfaces of a cubic element with
mutually perpendicular axes 1(X1), 2(X2), 3(X3) can be written as

 
t( nˆ 1 ) = t( nˆ 1 ) nˆ 1 + t( nˆ 1 ) nˆ 2 + t( nˆ 1 ) nˆ 3
1 2 3

= t( nˆ 1 ) j nˆ j

 
Similarly, t( nˆ 2 ) = t( nˆ 2 ) j nˆ j and t( nˆ 3 ) = t( nˆ 3 ) j nˆ j
The nine components are t( nˆ i ) j ≡ σ ij or τ ij which is called the second-order
(Cartesian) stress tensor. Here, t( nˆ i ) is the stress vector in the direction nˆ i per-
pendicular to the surface, and t( nˆ i ) j is its resolved part in the jth direction.
The stress tensor can be expressed in matrix form as

 σ 11 σ 12 σ 13   σ xx σ xy σ xz 
   
σ ij =  σ 21 σ 22 σ 23  or  σ yx σ yy σ yz 
 σ 31 σ 32 σ 33   σ zx σ zy

σ zz 


with reference to the coordinate planes (shown in Figure 8.1). The compo-
nents in the diagonal σ 11 = σxx, σ 22 = σyy, σ 33 = σzz are called normal stresses
perpendicular to planes, and the components σij (i ≠ j) or σxy, σyx, σxz, σzx, σyz,
σzy tangential to the coordinate planes are called shear stresses (τij) in other
notations.
126 Tensor Calculus and Applications

3(X3)
σ33

τ32 τ31
τ23

τ13

σ22 τ21

τ12 σ11
2(X2)

1(X1)

FIGURE 8.1
Stress components on the plane surfaces of a cubic element (for simplicity).

8.4 Deformation and Displacement Gradients

Let (X1, X2, X3) be the material coordinates of the point Po in undeformed
configuration of a material continuum at t = 0, with respect to O−X1X2X3,
and P(x1, x2, x3) be the corresponding position in the deformed configu-
ration
 at t = t1 with respect to the  Cartesian system O′ − x1x2 x3 so that
( ) ( 
)
OP o = X = X 1Iˆ1 + X 2 Iˆ2 + X 3 Iˆ3 and O′P = x = x1eˆ1 + x2 eˆ2 + x3 eˆ3.
The particles of the continuum undergoing deformation can move along
different paths in space. If the particle Po initially at t = 0 is assumed to move
to the position P at t = t1, then functionally, it can be represented by xi = xi(X1,
X2, X3, t) = xi (X, t) or conversely Xi = Xi(x1, x2, x3, t) = Xi (x, t).

∂ xi
∴ dxi = dX j (8.4.1)
∂X j

∂ xi
where the tensor is called the material deformation gradient tensor,
∂X j
and t is the absolute time. Similarly, from Xi = Xi(x, t),

∂Xi j
dX i = dx (8.4.2)
∂x j

∂Xi
It can be seen that is also a tensor and is called spatial deformation
∂x j
gradient tensor.
Definitely, the material and spatial deformation tensors are interrelated by
means of the chain rule of partial differential:
Tensors in Continuum Mechanics 127

x3
X3
P(x1,x2,x3)

Î3 u
x
) ê3
,X 2,X 3 x2
X P o(X 1

b ê2
O/
X2
Î1 O ê1 t = t1
Î2
t=0
X1
x1

FIGURE 8.2
Deformation graph.

∂ xi ∂ X j ∂Xi ∂x j
= δ ik (8.4.3)
∂X j ∂x k ∂x j ∂X k

From the displacement vector ui = xi − Xi + bi (Figure 8.2), the partial differen-
tiation with respect to coordinates gives the material displacement gradient
∂ui ∂u
and the material displacement i as
∂X j ∂x j

∂ui ∂ xi ∂ X i ∂ xi
= − = − δ ij (8.4.4)
∂X j ∂X j ∂X j ∂X j

and

∂ui ∂ xi ∂X i ∂X i
= − = δ ij − (8.4.5)
∂x j ∂x j ∂x j ∂x j

8.5 Deformation Tensors and Finite Strain Tensors

Let us consider two superimposed rectangular Cartesian coordinate sys-
tems O–X1X2X3 for initial configuration and O–x1x2x3 for final configuration
128 Tensor Calculus and Applications

after deformation of a material continuum. To extract the characteristic dif-

ference of the measure of material deformation, it is justified to refer to a dif-
ferent coordinate system (instead of referring to the same system) but from
the same position. The neighboring particles Po and Qo before deformation
are supposed to move to the points P and Q, respectively, in the deformed
configuration (Figure 8.3).
The square of the differential element of length between Po and Qo

dX 2 = dX i dX i = δ ij dX i dX j (8.5.1)

∂X i
where dX i = dx j (from 8.4.2).
∂x j

∂X k ∂X k
( dX )2 = dX k dX k = dxi dx j
∂ xi ∂ x j (8.5.2)
= Cij dxi dx j

∂X k ∂X k
in which the second-order tensor Cij = is called the Cauchy’s defor-
mation tensor. ∂xi ∂x j
Also, the square of the differential element of length between P and Q for
the deformed configuration is

( dx )2 = dxi dxi = δ ij dxi dx j , (8.5.3)

∂ xi
where dxi = dX j (from 8.4.1).
∂X j

X3 (x3)
Q

u dX
u+d
Qo
dX
u P
dX

Po
X+

x
x
O
X2 (x2)

X1 (x1)

FIGURE 8.3
Graph of deformation tensor and strain tensor.
Tensors in Continuum Mechanics 129

It can be written as

∂ xk ∂ xk
( dx )2 = dX i dX j
∂X i ∂X j (8.5.4)
= Gij dX i dX j ,

∂ xk ∂ xk
where the second-order tensor Gij = is called the Green’s deforma-
∂X i ∂X j
tion tensor.
The difference ( dx ) − ( dX ) corresponding to two neighboring particles
2 2

between the initial and final configurations of a material continuum is taken

as the measure of deformation. For all neighboring particles, if this differ-
ence vanishes for a continuum, the corresponding displacement is called
rigid displacement.
Now, (dx)2 − (dX)2 = (Gij − δij)dXidXj = 2LijdXidXj, using (8.5.4) and (8.5.1)
where

1 1  ∂ x ∂ xk 
Lij =
2
( )
Gij − δ ij =  k
2  ∂X i ∂X j
− δ ij  (8.5.5)


which is called the Lagrangian (or Green’s) finite strain tensor.

Also, (dx)2 − (dX)2 = (δij − Cij)dxidxj using (8.5.3) and (8.5.2) = 2Eijdxidxj, using
(8.5.3) and (8.5.2)
where

1 1 ∂X k ∂X k 
Eij =
2
( )
δ ij − Cij =  δ ij −
2 ∂ xi ∂ x j 
(8.5.6)

which is called the Eulerian finite strain tensor.

Making use of the relation of displacement vector,

ui   = xi – X i + bi

Lij can be written as

1  ∂uk   ∂u  
Lij =  + δ ki   k + δ kj  − δ ij 
2  ∂X i   ∂X j  

1  ∂uk ∂uk ∂u j ∂ui 

=  + + + δ ij − δ ij  (8.5.7)
2  ∂X i ∂X j ∂X i ∂X j 

1  ∂ui ∂u j ∂uk ∂uk 

=  + + 
2  ∂X j ∂X i ∂X i ∂X j 
130 Tensor Calculus and Applications

Similarly,

1  ∂ui ∂u j ∂uk ∂uk 

Eij =  + −  (8.5.8)
2  ∂ x j ∂ xi ∂ xi ∂ x j 

Now, if we impose the condition for small deformation theory on the dis-
placement gradients, which are very small compared to unity, then the prod-
uct term in (8.5.7) can be ignored. Hence, the expression corresponding to Lij
can be written as

1  ∂ui ∂u j 
lij = + (8.5.9)
2  ∂X j ∂X i 

It is called the Lagrangian infinitesimal strain tensor.

∂uk
Also, for  1, ignoring the product term, (8.5.8) can be equivalently
∂ xi
written as

1  ∂ui ∂u j 
eij = + (8.5.10)
2  ∂ x j ∂ xi 

which is called the Eulerian infinitesimal strain tensor.

8.6 Linear Rotation Tensor and Rotation Vector

in Relation to Relative Displacement
 
Let u( Po ) and u(Qo ) represent the displacement vectors of the two neighboring
particles in a material medium.

 
∴ du = u (Qo ) − u( Po )

∴ dui = ui (Qo ) − ui( Po )

which is the measure of relative displacement vector of the particle origi-

nally at Qo with respect to the particle originally at Po. Subject to continuous
displacement, ui( Po ) the expansion about by Taylor’s series, it can be written
 ∂u 
as dui =  i  dX j (neglecting higher order terms for small displacement).
 ∂X j  Po
Tensors in Continuum Mechanics 131

∂ui
The material displacement can be decomposed into a symmetric part
∂X j
and an antisymmetric part so that

 1  ∂u ∂u j  1  ∂ui ∂u j  
dui =   i + +  −   dX j
 2  ∂X j ∂X i  2  ∂X j ∂X i   Po
(8.6.1)
∂u
∴ i = lij + Wij
∂X j
( using 8.59) ,

where

1  ∂ui ∂u j 
Wij = − (8.6.2)
2  ∂X j ∂X i 

which is called the linear Lagrangian rotation tensor (Figure 8.4).

For infinitesimal rigid body rotation, corresponding to relative displace-
ment at point Po, the Lagrangian strain tensor lij vanishes. Hence, the infini-
tesimal rotation vector is thus given by

1
wi = ∈ijk Wkj (8.6.3)
2

du
Qo

dX

Po
u

FIGURE 8.4
Graph for rotation tensor and vector.
132 Tensor Calculus and Applications

Again, according to Eulerian description of the relative displacement vector,

it can be written as

∂ui
dui = dx j
∂x j

so that

∂ui  1  ∂ui ∂u j  1  ∂ui ∂u j  

= + + − 
∂ x j  2  ∂ x j ∂ xi  2  ∂ x j ∂ xi   (8.6.4)

= eij + wij ,

where

1  ∂ui ∂u j 
wij = − (8.6.5)
2  ∂ x j ∂ xi 

which is called the Eulerian rotation tensor with Eulerian rotation vector

1
wi = ∈ijk wkj . (8.6.6)
2

It has already been discussed that deformation is connected with displace-

ments. Therefore, motion of any continuum such as fluid or gas requires the
use of tensors (or vectors), namely, stress–strain tensors, rotation tensor, and
displacement vector. Ironically, there is no alternative but to use tensors in
non-isotropic medium to arrive at accurate results of an investigation.
9
Tensors in Geology

9.1 Introduction
The structural investigations of spasmodic deformation [it means the change
in shape of a body from the initial (undeformed) configuration to a subse-
quent (deformed) configuration] caused by nature and origin of some forces
need application of some mathematical concepts such as tensors. The math-
ematical entities such as stress and strain (Section 8.2) are largely responsible
to study any kind of deformation. A second-order tensor called “stress” is
the essence of structural geology.
Rheology is closely related to the study of deformation of the structures
of the earth and any kind of material structure ranging from the order of
seconds (seismic-wave propagation) to hundreds of millions of years (geody-
namics). The aforesaid “stress and strain” are the fundamental ingredients
to deal with the analysis of continuum mechanics of deformation (discussed
in Section 8.2) of extended bodies in the context of rheology. Newtonian and
non-Newtonian viscosity, linear rheological bodies, plasticity, and brittle fail-
ure can be investigated with these mathematical entities—the second-order
tensors. From a rheological standpoint, properties of lithosphere and the
mantle, temperature distribution of lithosphere, thermal convection in the
mantle, flexure of the lithosphere, stresses on it, and viscosity of the ­mantle
from surface loading data are some of the features within this domain.
Of course, atomic basic deformation and flow in polycrystalline m ­ aterials
­covering hydrolytic weakening dynamic recrystallization and pressure
and temperature effects may attract the involvement of both geologists and
­geophysicists [Ref. 7].
The displacement gradient tensor (one of many tensors related to strain)
relates the position vector of a point to the displacement of the point dur-
ing a displacement. Knowing displacement gradient tensor, it is possible
to calculate how all points within a body are displaced as a function of
position during deformation. If we know the stress tensor, we can calcu-
late the stress vector on a plane of any orientation within a body, which is

133
134 Tensor Calculus and Applications

largely important in any study of earthquakes, induced seismicity, fault-

ing, etc.
There are three stages of deformation (mentioned earlier) when direct
forces act on a body extending to a short period of time, minutes, or hour.
They are as follows:

i. Elastic
ii. Plastic
iii. Rupture

Elastic: If the body returns to the original shape and size after the with-
drawal of the stress, the deformation is called elastic. If the body does not
return to the original shape when the stress exceeds a certain stage, it is
called elastic limit.
The strain is proportional to stress when it remains always less than the
elastic limit, and the deformation obeys Hooke’s law.
Plastic: The deformation is said to be plastic if the stress exceeds the elastic
limit.
N.B.: The difference between the external force applied to a body and
the corresponding outcomes of internal actions and reactions generates
stress.
Rupture: If the specimen is subjected to continuous increase of stress, one
or more fractures based on several factors can develop, which eventually
fails by rupture. Rupture is responsible for “brittle” in substances before
certain stage of plastic deformation.
From a geometrical point of view, strain causing distortion of a body can
also be classified as homogeneous and inhomogeneous.
Homogeneous deformation: After deformation,

i. Straight lines remain straight lines.

ii. Parallel lines remain parallel.
iii. In the strained body, all lines in the same direction have constant
value of e (extension—change in unit in length), λ (quadratic elonga-
tion), ψ (angular strain), and γ (shear strain).

Inhomogeneous deformation: After deformation,

i. Straight lines change to curves.

ii. Parallel lines turn nonparallel.
iii. The values of the above four parameters e, λ, ψ, and 𝛾 in any one
given direction of the body become variable, not constant.
Tensors in Geology 135

9.2 Equation for the Determination of Shearing

Stresses on Any Plane Surface
Let the coordinate axes x, y, z be rotated in the directions of three mutually
perpendicular axes of principal stress, σ 1, σ 2, σ 3 acting on any plane sur-
face. Let ABC be the plane of unit area which inclines to these coordinate
axes. The normal stress σ is supposed to act along the normal to this plane
having direction cosines l, m, n and shearing stress τ on it. If S (Sx, Sy, Sz) is
the resultant stress on the plane ABC (Figure 9.1) considered at the point
P, then

S2   = σ 2 + τ 2 , (9.2.1)

where Sx = σ 1l, Sy = σ 2m, Sz = σ 3n

so that

S2 = σ 12l 2 + σ 22 m2 + σ 32 n2 . (9.2.2)

Also, the measure of the normal stress σ in the directions l, m, n is given by σ.

σ = Sxl + Sy m + Sz n
(9.2.3)
σ = σ 1l 2 + σ 2 m2 + σ 3 n2 .

C Sz σ

σ3
S

P
Sy
σ2 Sx
O
B
σ1
Y
A
X

FIGURE 9.1
Normal stress σ and shearing stress τ in any plane ABC of unit area.
136 Tensor Calculus and Applications

Using (9.2.2) and (9.2.3) in (9.2.1), we get

( )
2
τ 2 = σ 12l 2 + σ 22 m2 + σ 32 n2 − σ 1l 2 + σ 2 m2 + σ 3 n2

( ) (
= σ 12l 2 1 − l 2 + σ 22 m2 1 − m2 + σ 32 ) n ( 1 − n ) − 2σ σ l m
2 2
1 2
2 2
− 2σ 2σ 3 m2 n2

− 2σ 1σ 3 l 2 n2
( ) (
= σ 12l 2 m2 + n2 + σ 22 m2 l 2 + n2 + σ 32 n2 m2 + l 2 ) ( )
(∴ l 2
+ m 2 + n2 = 1 )
−2σ 1σ 2l 2 m2 − 2σ 2σ 3 m2 n2 − 2σ 1σ 3l 2 n2
(9.2.4)
∴ τ 2 = l 2 m2 (σ 1 − σ 2 ) + m2 n2(σ 2 − σ 3 ) + l 2 n2 (σ 3 − σ 1 ) .
2 2 2

This is the required equation for shearing stress τ in terms of principal stress
in the direction of the normal stress σ to the plane surface ABC.
To determine the maximum and minimum stresses, we are to study the
stationary values of τ from this equation replacing n2 by n2 = 1 − l2 − m2.

(
∴ τ 2   = l 2 m2 (σ 1 – σ 2 ) + m2 1 – l 2 – m2 (σ 2 – σ 3 )
2
) 2

( )
+ l 2 1 – m2 – l 2 (σ 3 – σ 1 ) .
2

Differentiating it partially with respect to l and m successively, we get

∂τ
2τ
∂l
2
(
= 2lm2 (σ 1 − σ 2 ) − 2lm2 (σ 2 − σ 3 ) + 2l l − l 2 − m2 (σ 3 − σ 1 ) − 2l 3 (σ 3 − σ 1 )
2 2 2
)
= 2lm2 (σ 1 − σ 2 ) − (σ 2 − σ 3 )  − 2l (σ 3 − σ 1 ) 1 − l 2 − m2 − l 2 
2 2 2

= 2lm2 (σ 1 − σ 3 )(σ 1 − 2σ 2 + σ 3 ) − 2l (σ 3 − σ 1 ) 1 − 2l 2 − m2
2
( )
∂τ
τ
∂l
= l (σ 1 − σ 3 )[m2 (σ 1 − 2σ 2 + σ 3 ) + (σ 1 − σ 3 ) 1 − 2l 2 − m2 , ( ) (9.2.5)

and similarly,

∂τ
τ
∂m
= m (σ 2 − σ 3 ) l 2 (σ 2 − 2σ 1 + σ 3 ) + (σ 2 − σ 3 ) 1 − 2 m2 − l 2  . (9.2.6) ( )
∂τ ∂τ
But for stationary values, = = 0 simultaneously.
∂l ∂m
Tensors in Geology 137

Hence, from (9.2.5), l = 0 for unequal values of σ 1, σ 2, σ 3 (in general) gives
when substituted in (9.2.6),
1
m = 0 and m = ± .
Thus, 2

l = 0, m=0 gives n = 1
1 1
l = 0, m=± gives n = ± (9.2.7)
2 2
1 1
l = 0, m=± gives n =  .
2 2

Similarly, eliminating l and m successively from (9.2.4), we get for other

­stationary values

l = 0, m = 1, n=0
1 1
=± , m = 0, n= ± (9.2.8)
2 2
1 1
l= , m = 0, n= ± .
2 2

l = 0, m = 1, n=0
1 1
=± , m=± , n=0 (9.2.9)
2 2
1 1
=± , m= , n = 0.
2 2

The first three stationary values of the three sets of (9.2.7)–(9.2.9) are related
to principal planes which correspond to minimum values of the shearing
stress. The remaining stationary values represent the maximum shearing
stresses with respective planes orienting to contain one of the principal axes
of stress, which makes 45° angle with the other two axes [7, p. 38] in the plane.

9.3 General Transformation and Maximum

and Minimum Longitudinal Strains
Let P(x, y) be any point before deformation which is displaced to the position
Q(x1, y1) with the general displacement.

x1 = ax + by and y1 = cx + dy .
138 Tensor Calculus and Applications

Solving for x and y (a > 0, d > 0),

dx1 − by1 ay − cx1

x= , y= 1 . (9.3.1)
−
ad bc ad − bc

Let the rectangle joining the points O(0, 0), M(x, 0), P(x, y), and N(0, y) be
deformed with the same point P into the parallelogram ORQS with points
(0,0), (ax, cx), (ax + by, cx + dy), and (by, dy). It is to be noted that a and d rep-
resent the components of longitudinal strains parallel to the x and y coordi-
nates, respectively. Moreover, b and c represent the part shear components
characterizing the angular displacements of the initial sides of the rectangle
so that

ax = l cosθ , cx = l sinθ ,

c = tan θ
a

c = a tanθ

and

dy = l cosφ , by = l sinφ

b
= tan φ
d
b = d tanφ .

Now let us investigate what will be the change in a general line y = mx + p
after deformation subject to the above general displacement with x, y given
by (9.3.1).
∴ Putting the values of x, y in the equation of the line, we get

ay1 − cx1 dx − by1

=m 1 +p
ad − bc ad − bc

y1 ( a + bm ) = ( c + dm ) x1 + p ( ad – bc )

c + dm p( ad − bc)
∴ y1 = x1 + , (9.3.2)
a + bm a + bm

which is also a straight line.

Hence, the general transformation representing Figure 9.2 characterizes
the homogeneous strain.
Tensors in Geology 139

Y
Q(x1, y1)

N P(x, y)
(o,y)

.dy
.Θφ
R

.Θθ
O(o,o) M(x,o) X

.ax

FIGURE 9.2
General strain translation.

Again the points lying on the circle (inscribed rectangle) x2 + y2 = 1 are sub-
jected to change due to the general transformation, and the circle takes the
form:
2 2
 dx1 − by1   ay1 − cx1 
  +  =1
ad − bc   ad − bc 

(c 2 + d 2 )x12 − 2(bd + ac)x1y1 + ( a 2 + b 2 )y12 = ( ad − bc)2, (9.3.3)

which is an ellipse called strain ellipse.

The major and minor axes of this strain ellipse identify the positions of
maximal and minimal longitudinal strains. By means of a rotation ψ of axes,
the term with x1y1 of the equation of strain ellipse (9.3.3) can be removed to
reduce it to a standard form, and ψ is given by

1  −2( ac + bd) 
ψ = tan −1  2 2 2 2 
. (9.3.4)
2  (c + d ) − ( a + b ) 

The lengths of the semimajor axis λ1 and minor axis λ2 or maximum and
minimum strains are given by 1 + e1 = λ1 and 1 + e2 = λ2 , respectively,
x2 y 2
where the strain ellipse is + = 1.
λ1 λ 2
140 Tensor Calculus and Applications

9.4 Determination of the Two Principal Strains in a Plane

Let P(x, y) be a point with origin O and OP be one of the principal axes of
strains so that OP = 1. Let θ be the angle made by OP with x-axis. Q (x1, y1) is
the deformed position of P without rotation, and hence, for the strain ellipse,
the length of one principal strain is OQ = 1 + e (Figure 9.3).
ax + by bx + dy
Now from the Figure 9.3, cosθ = x = and sin θ = y = .
( for irrotational strain c = b) 1+ e 1+ e

∴ x  a – ( 1 + e )  + by = 0 (9.4.1)

and

y  d – ( 1 + e )  + bx = 0. (9.4.2)

b a − (1 + e)
Dividing, =+
d − (1 + e) b

∴  a – ( 1 + e )   d – ( 1 + e )  – b 2 = 0

(1 + e )2 – ( 1 + e )( a + d ) + (ad – b 2 ) = 0, (9.4.3)
which gives the two principal irrotational strains (the two roots of it), (1 + e1)
and (1 + e2). From (9.4.1) and (9.4.2), replacing x by cos θ and y by sin θ, we can
get

Q(x1, y1)

. bx by

)
P (x , y
.
.dy φ
1
.θ
X
O .ax

FIGURE 9.3
Graph for principal strains.
Tensors in Geology 141

a – ( 1 + e ) + b tan θ = 0 and d – ( 1 + e ) + b cot θ = 0

so that

a−d
tan 2 θ + tan θ − 1 = 0, (9.4.4)
b

which is quadratic in tanθ. This gives the directions of the two mutually
perpendicular axes of strain ellipse.
Hence, with reference to the chosen reference axes, the finite displace-
ments (two dimensional) can be instrumental to determine the principal
elongation 1 + e1 = λ1 and 1 + e2 = λ2 with lengths of principal axes λ1, λ2 of
x2 y 2
the strain ellipse + = 1 derived from the deformation of the initial unit
λ1 λ 2
circle x2 + y2 = 1. To speak the truth, the measures of the two principal strains
(magnitudes of strain tensor) can be completely determined from Equations
(9.4.3) and (9.4.4), and the angle of rotation ψ by the lines turning to axes of
the ellipse.
Finally, to determine the displacements of the points in space resulting
in the strain ellipse, significantly no particular coordinate system is cho-
sen in the deformation state. There could have been the same distortions
or strains for the displacements with reference to different coordinate sys-
tems (Cartesian, of course, in this case) though the equations would have
been slightly different. It is discussed in the beginning of the book that the
independence of reference frame to study some mathematical concepts is
essentially required to use tensors. Quantities such as stresses and strains
related to geological context are some second-order tensors or general ten-
sors. For three-dimensional cases, the mathematical entities such as stress
will require (32 − 3) = 6 components (for a symmetric tensor (σxy = σyx)) and
32 = 9 components (for anti-symmetric case). Of course, there are some quan-
tities which remain invariant irrespective of reference frames but need full
knowledge of tensor calculus.
The geometrical processes of determination of the measures of stress and
strain are discussed here (instead of using truly tensors but magnitudes) to
help in its geological context.
N.B.: Though actually stress and strain are three dimensional, it is beyond
the scope of the book to discuss the general case, and only classical result is
reported.
10
Tensors in Fluid Dynamics

10.1 Introduction
Modern scientists and physicists believe that matter is composed of elemen-
tary particles, and in most of the scientific fields, it is not looked into the
individual molecules, which is regarded as an entity of small but infinite
dimensions interacting with its fellows according to certain laws. So matter
is not continuous but discrete, and its gross properties are taken as averages
over a large number of molecules. The equations of fluid motion have been
formulated from this viewpoint, though they are considered at first sight as
much more fundamental one. The average velocity of the molecules is taken
in the neighborhood of a point, but how large this neighborhood should be
is a questionable one.
In the investigation of ordinary fluid motion, the variations in the medium
is considered isotropic (uniform in all directions), and hence, pressure (can
be said as classical stress) is a constant quantity. But in case of viscous fluids,
it is to be replaced by stress tensor. Eventually, the corresponding body and
the surface forces will occur in the system in appropriate forms. The Navier–
Stokes equations can suitably meet all the requirements to study the motion
of fluids when viscosity is present in the medium.

10.2 Equations of Motion for Newtonian Fluid

Let us consider an arbitrary volume V within the fluid medium enclosed by
a surface S.
( )
If ρ is the density of the fluid particles moving with velocity ν ν i , then by
Reynolds’ transport theorem,

143
144 Tensor Calculus and Applications

δ  δρ 
δt ∫∫∫ ρ dV = ∫∫∫  δ t + ρv  dV i
,i

 ∂ρ 
=
∫∫∫  ∂t + ρ v + ρv  dV
,i
i i
,i

 ∂ρ 
=
∫∫∫  ∂t + (ρv )  dV. i
,i

But for conservation of mass within the arbitrary volume dV,

∂ρ
+ ( ρ v i ), i = 0. (10.2.1)
∂t

This is the equation of continuity characterizing the conservation of mass.

If T ij is the contravariant stress tensor on the element ds of the surface with
normal nj, then the force on this element is Tijnjds.
Also let the contravariant vector f i represent the external body forces per
unit mass of the fluid.
∴ The total net force on the arbitrary volume V enclosed by the surface S
when resolved in any direction li (say) to it is

∫ ∫∫ ρ f l dV + ∫∫ T n l ds = ∫ ∫∫ ρ f l dV + ∫∫ ∫ T l dV
i ij i ij
i j i i ,j i
V S V V

=
∫ ∫∫ (ρ fV
i
+ T,ijj )li dV

using Gauss’s divergence theorem to the second term.

But the rate of change of linear momentum is equal to the net external
force in the direction li.
δ
Hence,
δt ∫ ∫∫ρ v ili dV =
V
∫ ∫∫
( ρ f i + T,ijj )li dV
V

δ vi
=
∫ ∫∫
V
ρ
δt
li dV =
∫ ∫∫ (ρ f
V
i
+ T,ijj )li dV .

But for arbitrary volume V and direction li it holds, only when


(for the velocity vector (tensor) v , v,i j means covariant derivative)

 ∂vi   ∂vi 
 ρ ∂t = ρ a
i
= ρ  + v i v,i j  = ρ f i + T,ijj
  ∂t 
Tensors in Fluid Dynamics 145

so that
ij
∂vi T, j
(a i =) + v i v,i j = f i + , (10.2.2)
∂t ρ

which is the equation for conservation of linear momentum.

Thus, the equations of motion of Newtonian fluids are given by Equations
(10.2.1) and (10.2.2).

10.3 Navier–Stokes Equations for the Motion of Viscous Fluids

In the case of ordinary fluids, the stress tensor Tij corresponding to ­hydrostatic
pressure p is Tij = − pgij, and the fluid medium is isotropic. But in the presence
of viscosity in the medium, it is anisotropic and is related to deformation.
Let vi and v i + dv i be the velocities of two neighboring points xj and xj + dxj.

∂vi j
dv i = dx = v,i j dx j .
∂x j
But

1 1
v,kj = g ik vi , j = g ik  vi , j + v j , i + vi , j − v j , i 
( ) ( )
2 2 

 1  ∂v ∂v j  1  ∂v ∂v j  
= g ik   ij + i  +  ij − i   (10.3.1)
 2  ∂x ∂x  2  ∂x ∂x  

= g ik  (eij + wij ),

where

1  ∂ vi ∂ v j 
eij = + (10.3.2)
2  ∂ x j ∂ x i 

which is the deformation symmetric strain tensor (8.6.4) and

1  ∂ vi ∂ v j 
wij = − (10.3.3)
2  ∂ x j ∂ x i 

which is the rotation tensor indicating rigid body rotation of the element.
Now the equation of motion with acceleration ai, stress tensor Tij, and body
force fi is (from Equation (10.2.2))
146 Tensor Calculus and Applications

ρ a i = ρ f i + T,ijj . (10.3.4)

But the stress tensor Tij = −pgij for viscous fluids is to be supplemented by the
viscous stress tensor pij to result for deformation

T ij = – pg ij   + p ij .

This relationship must be linear and isotropic (same for all coordinate
­systems) when Newtonian fluid is isotropic.
∴ For the isotropic fourth-order symmetric tensor Gijmn with respect to i, j
and m, n, pij can be written as pij = Gijmn emn, in terms of the deformation strain
tensor emn.
But for generalization of isotropic fourth-order symmetric tensors, it must
be a linear combination of gijgmn and (gimgjn + gingjm)
so that

p ij   = λ g ij g mnemn   + µ( g im g jn + g in g jm )emn

= λ g ij emm   + 2 µe ij

T ij = (– p + λ emm )g ij  + 2 µeij . (10.3.5)

In the equations of motion (10.3.4), it is necessary to employ stress and strain

relation for viscous fluid motion. This requires to express Tij in the following
form:

T,ijj = [(− p + λ emm ) g ij + 2 µe ij ], j

= (− p, j + λ emm, j ) g ij + 2 µe ij, j

= (− p, j + λ emm, j ) g ij + 2 µ( g ik e kj ), j = (− p, j + λ emm, j ) g ij + 2 µ( g ik g jm e km ), j

1
= (− p, j + λ emm, j ) g ij + 2 µ  g ik g jm ( vk , m + vm , k ) 
2 , j

(
= (− p, j + λ emm, j ) g ij + µ g jm v,imj + g ik v,jkj ) ( j ↔ k)

= (− p, j + λ emm, j ) g ij + µ g jm v,imj + µ g ij v,kjk

T,ijj = (− p, j + λ emm, j + µ v,kjk ) g ij + µ g jm v,imj .

But emi = g ki ekm =

1 ki
2
1
(
g ( vk , m + vm , k ) = v,im + g ki vm , k
2
)
Tensors in Fluid Dynamics 147

1 m 1
∴ emm = ( v, m + g km vm , k ) = ( v,mm + v,kk ) = v,mm ( k → m).
2 2
Hence,

T,ijj = (− p, j + λ v,mmj + µ v,mmj ) g ij + µ g jm v,imj

= [− p, j + (λ + µ )v,mmj ] g ij + µ g jm v,imj .

Equation (10.3.4) for viscous fluids takes the form:

ρ a i = ρ f i − g ij p, j + (λ + µ )v,kkj g ij + µ g jk v,i jk
(10.3.6)
( v, jk = v, kj ),

which is the Navier–Stokes equations for viscous fluids.

Appendix
Some Standard Integrals in
Connection with Applications

In many investigations of dynamical scenarios, different kinds of integrals

appear to characterize the development of physical situations inviting trans-
formations. The transformation from a volume integral of certain physical
property to an integral over bounding surface is a veryimportant frequent
measure of classical or tensor analysis. For example, if F is the flux of some
 
physical quantity such as flux of fluids or charges, then the integral

F ⋅ n ds
∫∫
over the whole surface S, where n is the outward drawn normal to it, is equal
to the total flux out of the closed volume. The Green’s theorem (or Gauss’s
divergence theorem) in connection with such entity can be mathematically
stated as follows:

Green’s Theorem

Statement: If F is any continuously differentiable vector field in volume V
and bounded closed surfaces which may have piecewise smooth boundary
with outward drawn normal n, then
  

∫∫∫ ∇ ⋅ F dv = ∫∫ F ⋅ n ds,
V s
where ∇ is the vector operator .


In tensor form,
∫∫∫ F, ii dv =
∫∫ ( )
F i ni ds, for contravariant vector F F i .
v s
For covariant formalism,

∫∫∫ g
v
ij
⋅ Fi , j dv =
∫∫ g F n ds = ∫∫ F n ds.
s
ij
j i
s
j
j

The partial derivatives in Cartesian coordinates are replaced by covariant

derivatives, since they are identical.

Stoke’s Theorem

Statement: If F is any continuous vector field with continuous partial
­derivatives, then for any two-sided piecewise smooth surface S spanning

149
150 Appendix

  
a closed curve C,
∫
c
F ⋅ t ds =
∫∫ curl F ⋅ n dS, where t is the tangent vector to
s

the curve C and n is the normal right handedly orienting the direction of the
curve C.
But in tensor form,

∫∫ e
s
ijk
Fk , j ni dS =
∫ F t
c
k
k
ds, where ∈ijk is the permutation tensor.

Reynold’s Transport Theorem

If f ( x , t) is any function and V(t) is a closed volume with the fluid consisting
of the same fluid particles, then

 
d  ∂ f 

dt 

∫∫∫
v(t )
f ( x , t) dv  =


∫∫∫  ∂t + ∇ ⋅ ( f ν ) dv
v(t )
x = x ( x1 , x2 , x3 ).

dF
Let F (t) =
∫∫∫ f (x , t) dv, so dt
v(t )
is the material derivative which needs to be

determined.

dF (t) d
∴
dt
=
dt ∫∫∫ f (x , t) dv.
v(t )

Since V(t) is variable volume, the differentiation with respect to time c­ annot
be taken under the integral sign. But if we consider the integration with
respect to volume in ξ − space, ξ being material Cartesian coordinate, and
change ξ (ξ1 , ξ 2 , ξ 3 ) coordinates to x ( x1 , x2 , x3 ), so that
∂( x1 , x2 , x3 )
dV = dV0 = JV0 with dV0 = dξ1 .dξ 2 .dξ 3 at t = 0,
∂(ξ1 , ξ 2 , ξ 3 )
then differentiation and integration can be interchanged.
In this case, V(t) can be taken as the moving material volume coming from
fixed initial volume V0 at t = 0 due to the transformation
( )
x = x ξ , t and ξ can be treated as constant.

   
d  d
∴
dt 

∫∫∫
v(t )
f ( x , t) dv  = 
 dt 
 
∫∫∫ v0
f { x ( ξ , t), t} J dv0 



 df dJ 
=
∫∫∫  dt J + f dt  dv
v0
0
Appendix 151

 df 
=
∫∫∫  dt + f (∇ ⋅ v) J dv
v0
0

dJ ∂ J
 = + ( v ⋅∇) J
dt ∂t
 df 
=
∫∫∫  dt + f ∇ ⋅ v  dv
v(t )

Associating the gradient terms after making use of material derivatives, it

can be arranged to the form:

 
d  ∂ f 

dt 

∫∫∫
v(t )
f ( x , t) dv  =


∫∫∫  ∂t + ∇ ⋅ ( f v) dv,
v(t )

which is the Reynold’s transport theorem. The function f can be some scalar
or component of tensor.

Corollary 1

Now
∫∫∫ ∇ ⋅ ( f v) dv = ∫∫ f v ⋅ nˆ ds,
v(t ) s
(applying Green’s theorem)
where S is the surface enclosing the volume V(t).
∴ The Reynolds’ transport theorem can be put to the form:
d ∂f

dt ∫∫∫ f ( x , t) dv = ∫∫∫ ∂t dv + ∫∫ f v ⋅ nˆ ds.
v(t ) v(t ) s( t )

Corollary 2

The Jacobian of the transformation from material to spatial coordinates in

dv = J dv0 is a scalar, and the divergence of the velocity field v,ii is also scalar.
δ J dJ
∴ (The intrinsic derivative =) = = v,ii J .
δ t dt
∴ The Reynolds’ transport theorem in tensor form can be written as

 
δ  δ f 
δt 

∫∫∫
v(t )
f ( x , t) dv  =


∫∫∫  δ t + f v  dv
v(t )
i
,i

∂ f 
=
∫∫∫  ∂t + f
v(t )
,i v i + f v,ii  dv.

152 Appendix

δ f ∂f
 Intrinsic derivative of scalar invariant = + f, i v i
δt ∂t

∂ f 
=
∫∫∫  ∂t + ( fv )  dv (i)
v(t )
i
,i

∂f
=
∫∫∫ ∂t dv + ∫∫ fv ⋅ n ds,
v(t ) s
i
i

where f is a scalar or component of tensor. Replacing f by ρ (density), the

equation of continuity can be recovered from (i).
Remarks

In Chapters 7–10, importance is laid on the theoretical aspects only, showing

the use of tensors in the respective branches and that too in restricted senses.
Only the fundamental aspects required for initial approach and to pave the
way for studying each of the branches which are included in this book are
discussed. For this reason, problems are not discussed which would be pos-
sible for books of individual subject. The fundamental objective of the author
in the book is to report only the glimpses of applications of tensors in some
known branches so that readers can in reality enter into the threshold of
these branches.

153
Bibliography

1. P. G. Bergmann, Introduction to the Theory of Relativity. Prentice Hall, New Delhi

(1962).
2. L. P. Eisenhart, Riemannian Geometry. Princeton University Press/Oxford
University Press, Princeton, NJ/London (1949).
3. C. E. Weatherburn, An Introduction to Riemannian Geometry and the Tensor
Calculus. Cambridge University Press, Cambridge (1963).
4. T. J. Willmore, An Introduction to Differential Geometry. Oxford University Press,
Oxford (1959).
5. R. K. Pathria, The Theory of Relativity. Dover Publications, New York (2003).
6. J. B. Hartle, Gravity: An Introduction to Einstein’s General Relativity. Pearson,
New Delhi (2007).
7. J. G. Ramsay, Folding and Fracturing of Rocks. McGraw-Hill, New York (1967).
8. G. Ranalli, Rheology of the Earth. Allen & Unwin, Boston, MA (1982).
9. M. P. Billings, Structural Geology. Prentice Hall, New Delhi (1987).
10. M. R. Spiegel, Schaum’s Outline of Theory and Problems of Vector Analysis and an
Introduction to Tensor Analysis, SI edition. McGraw-Hill, New York (1959).
11. G. E. Mase, Continuum Mechanics. Tata McGraw-Hill, New York (2005).
12. R. Aris, Vectors, Tensors, and the Basic Equations of Fluid Mechanics. Dover
Publications, New York (1989).

155
Index

A symmetric tensors 30
tensor, 31, 55
Absolute tensor, 16
Curvature, 77, 114, 120
Angular Strain, 134
tensors, 93, 95, 99, 100, 109
Anisotropic, 145
Curved space, 93, 113, 119–121
Antisymmetric, 20, 21, 22, 31, 70, 100
Curvilinear coordinantes 3, 4, 103, 113, 119
Associate vector, 32, 33
Cyclic property, 98
Autoparallel curve, 82, 83

B D

Bianchi Identity, 103,104 Definite quadratic form, 9

Body force, 124, 143, 145 Deformation, 113, 124–125, 133, 134, 145,
Bounding surface, 124 146
Brittle, 134 configuration, 126
failure, 133 strain tensor, 146
tensor, 127
Dialation (distrotion), 124
C Discrete, 143
Cartesian coordinate, 3, 22, 27, 29, 30, Displacement gradients, 126, 133
43, 104 Distorts, 113
basis, 6 Divergence of a vector, 68, 108
orthogonal coordinates, 34 Dummy index, 13
tensors, 16, 123 Dynamical scenario, 3
Cauchy’s deformation tensor, 128
stress principle, 124, 125
E
Centrifugal force, 124
Christoffel symbols (brackets), 41–43, 50, Elastic, 134
51, 53, 58, 87, 88, 90, 104, 115 limit, 134
Codazzi equation, 105 Einstein, 93, 113, 114
Conjugate, 27 tensor, 122
Conservation of linear momentum 145 Empty space, 122
mass, 144 Energy momentum, 121, 122
Continuum concept, 123 Entity, 3
mechanics, 16, 123, 133 Equation of continuity, 144
Contraction, 18, 98 Euclidean coordinates, 30
Contravariant, 3–5, 14–15,17–18, 21–22, geometry, 30
24, 27, 32, 33, 56, 59, 60, 68, 70, 71 metric, 30
stress tensor, 144 plane, 103
Covariant constants, 63 space, 11, 30, 71
curvature tensor, 99, 117 Eulerian finite strain tensor, 129
derivatives, 55, 57–61, 64, 67, 68 infinitesemal strain tensor 130
differential, 118, 119, 120 rotation tensor, 132
differentiations, 63, 65, 71, 93, Evolutionary basis, 43

157
158 Index

F Induced secismicity, 134

Infinitesemal, 14
Faulting, 134
Inherent property, 124
Finite strain tensor, 127, 129
Inner product, 7, 19
First curvature of a curve, 77
Integral curve of geodesic equation 85
Flat, 114
Intensity, 124
space, 118
Intrinsic concept, 30, 93
Flexure, 133
derivative, 71, 151
Fundamental tensor, 33, 56–58, 60–62,
equation, 119
65, 67, 71,87, 90, 93
geometry, 103
contravariant tensor, 40
property, 113
covariant tensor, 32
Invariance, 123
ingradients, 133
Invariant, 8, 15, 16
quadratic form, 103
Irrotational strain, 140
Isometric, 109
G surfaces, 103
Isometry, 103
Galileo’s view, 113 Isotropic, 143, 145, 146
Gauss equation, 105 bodies, 124
Gaussian curvature, 104, 105, 109, 114, 115
surface, 103, 116
divergence theorem, 144, 149 J
General theory of relativity, 22, 93, 113, Jacobian, 14–16
114, 117, 122
Generating factor, 120
Geodesics, 77–78, 80, 82–83, 91–92, 114 K
coordinates, 86, 88 Kepler, 113
Geodynamics, 133 Kronecker delta, 14, 27
Geometry of motion, 113
Gravitational field equation, 120
potential, 120 L
Green’s deformation tensor, 129 Lagrangian (Green’s) finite strain tensor
theorem, 149, 151 129
infinite strain tensor, 130
H Laplacian, 69
Law of inertia, 113
Homogeneous deformation, 134 Levi-Civita, 72
strain, 138 Linear Lagrangian rotation tensor 130,
Hooke’s law, 134 131
Hydrolytic, 133, 138 strain tensor, 131
Hydrostatic pressure, 145 Linear rheological bodies, 133
transformation, 9, 119
Linearly independent, 3
I
Lithosphere, 133
Imbedded, 103 Loci, 11, 7
Immersed, 89 Longitudinal strains, 137
Increment, 73 Lorentz transformation, 113
Indefinite quadratic forms, 9 Lower down, 32
Index 159

M Orthogonality, 39
Order, 18
Macroscopic, 123
Orthonormal, 7
Magnetic field, 113
Orthonormal condition, 32
force, 124
coordinate system, 29, 30
Mantle, 133
set, 7
Material cartesian coordinates, 150
Outer or open product, 18
deformation, 128
deformaton gradient tensor, 126
derivative, 151 P
volume, 150
Parallel displacement, 43, 72–76, 83
Measure of deformation, 129
translation, 119
Metric, 9, 10
Permutation or pseudo tensor, 16
functions, 30
Plastic, 133, 134
space, 9, 83, 92, 103
Plasticity, 133
Mixed tensor, 60
Poission’s equation, 120
Molecular distribution, 123
Positive definite, 9, 29
fundamental form, 39
N Principal elongation, 141
irrotational strain, 140
Navier’s stokes, 143, 145 normal, 77
Negative definite, 9 strain, 141
Newton, 113 Principle of covariance, 121
Newton’s law of gravitation, 120 relativity, 113
vacuum equation, 122
Newtonian, 113
fluid, 143, 145, 146 Q
gravitational equation, 120 Quadratic differential forms, 9, 103
Non-Euclidean geometry, 113 elongation, 134
Non-inertial frames, 121 forms, 7
Non-Newtonian viscosity, 133 Quotient law, 25, 27, 33
Non-singular, 8 Questionable one, 143
Non-isotropic medium, 113
Norm, 7
Normal form, 8, 9 R
stress, 124, 125 Raise the indices, 32
Notion of stationary arc length, 82 Rank, 8, 18
N-tuple, 14 Real index, 13
Recrystallisation, 133
Rectangular cartesian coordinates, 3,
O
11, 26
Oblique cartesian coordinates, 5 coordinates, 45
curvilinear coordinates, 10 Rectilinear coordinates, 119
Orientation, 133 motion 119
Orthogonal, 4, 7 Relative displacement, 130
cartesian coordinates, 103 tensor, 116
condition, 32 vector, 130
coordinate system, 29, 30 Reynold’s transport theorem, 143, 151
160 Index

Rheology, 133 Strain ellipse, 139, 141

Rheological standpoint, 133 Stress and strain, 124, 133, 134
Ricci tensor, 98, 109, 120, 122 tensor, 125, 133, 145, 146
Riemann, 11 vectors, 124
Riemannian, 39, 87 vector on a plane, 133
Christoffel tensor, 106 Structural geology, 133
coordinates, 86, 87, 115, 116 Stylistic, 13
curvature tensor, 107, 108, 115 Subspace, 89
geometry, 29, 93 Supplement, 25
metric, 29, 40 Summation convention, 13, 15
space, 27, 29, 39, 114 Surface invariant, 104
symbols, 93, 95, 99 Sylvester law of inertia, 8
Rotation tensor, 145 Symmetric, 20, 22, 100
vector, 130 stain tensor, 145
Rupture, 134 tensors, 20, 21, 23, 26, 31, 32, 46

S T
Scalar invariant, 69 Tangential, 124
product, 31 Tensor density, 16
Seismic wave propagation, 133 Transformation of coordinates, 14
Seismicity, 134
Shear strain, 134
V
Shearing stress, 124, 125, 135
Signature, 8, 9, 11 Variable magnitude, 73, 74
Single value, 3 principle, 78, 92
Singular, 8 Variation, 79
Skew symmetric, 39 Vector of constant magnietude 73, 75
Space invariant, 104 Viscosity, 143
Space time continuum, 113, 114, 121 Viscous fluids, 143
Spasmodic deformation, 133 stress tensor, 146
Spatial deformation gradient tensor
126
W
Special theory of relativity, 113, 114
State of stress, 125, 133 Weight, 27
Stationary length, 78, 80, 81 Weight of tensors, 16
Stipulated, 6 World line (track), 120
Stoke’s theorem, 149 lines of light, 38