Вы находитесь на странице: 1из 262

This page

intentionally left
blank
Copyright © 2006, New Age International (P) Ltd., Publishers
Published by New Age International (P) Ltd., Publishers

All rights reserved.


No part of this ebook may be reproduced in any form, by photostat, microfilm,
xerography, or any other means, or incorporated into any information retrieval
system, electronic or mechanical, without the written permission of the publisher.
All inquiries should be emailed to rights@newagepublishers.com

ISBN (13) : 978-81-224-2700-4

PUBLISHING FOR ONE WORLD


NEW AGE INTERNATIONAL (P) LIMITED, PUBLISHERS
4835/24, Ansari Road, Daryaganj, New Delhi - 110002
Visit us at www.newagepublishers.com
To
My parents
This page
intentionally left
blank
FOREWORD

It gives me great pleasure to write the foreword to Dr. Nazrul Islam’s book entitled “Tensors and Their
Applications. I know the author as a research scholar who has worked with me for several years. This
book is a humble step of efforts made by him to prove him to be a dedicated and striving teacher who
has worked relentlessly in this field.
This book fills the gap as methodology has been explained in a simple manner to enable students
to understand easily. This book will prove to be a complete book for the students in this field.

Ram Nivas
Professor,
Department of Mathematics and Astronomy,
Lucknow University,
Lucknow
This page
intentionally left
blank
PREFACE

‘Tensors’ were introduced by Professor Gregorio Ricci of University of Padua (Italy) in 1887
primarily as extension of vectors. A quantity having magnitude only is called Scalar and a quantity with
magnitude and direction both, called Vector. But certain quantities are associated with two or more
directions, such a quantity is called Tensor. The stress at a point of an elastic solid is an example of a
Tensor which depends on two directions one normal to the area and other that of the force on it.
Tensors have their applications to Riemannian Geometry, Mechanics, Elasticity, Theory of Relativity,
Electromagnetic Theory and many other disciplines of Science and Engineering.
This book has been presented in such a clear and easy way that the students will have no difficulty
in understanding it. The definitions, proofs of theorems, notes have been given in details.
The subject is taught at graduate/postgraduate level in almost all universities.
In the end, I wish to thank the publisher and the printer for their full co-operation in bringing out
the book in the present nice form.
Suggestions for further improvement of the book will be gratefully acknowledged.

Dr. Nazrul Islam


This page
intentionally left
blank
CONTENTS

Foreword ............................................................................................................. vii


Preface ................................................................................................................ ix

Chapter–1 Preliminaries ........................................................................................... 1-5


1.1. n-dimensional Space ............................................................................................... 1
1.2. Superscript and Subscript ....................................................................................... 1
1.3. The Einstein's Summation Convention ...................................................................... 1
1.4. Dummy Index ....................................................................................................... 1
1.5. Free Index ............................................................................................................. 2
1.6. Krönecker Delta ..................................................................................................... 2
Exercises ............................................................................................................... 5

Chapter–2 Tensor Algebra ..................................................................................... 6-30


2.1. Introduction .......................................................................................................... 6
2.2. Transformation of Coordinates ................................................................................ 6
2.3. Covariant and Contravariant Vectors ......................................................................... 7
2.4. Contravariant Tensor of Rank Two .......................................................................... 9
2.5. Covariant Tensor of Rank Two ................................................................................ 9
2.6. Mixed Tensor of Rank Two..................................................................................... 9
2.7. Tensor of Higher Order ......................................................................................... 14
2.8. Scalar or Invariant ................................................................................................ 15
2.9. Addition and Subtraction of Tensors ....................................................................... 15
2.10. Multiplication of Tensors (Outer Product of Tensors) ............................................... 16
2.11. Contraction of a Tensor ........................................................................................ 18
2.12. Inner Product of Two Tensors .............................................................................. 18
2.13. Symmetric Tensors .............................................................................................. 20
2.14. Skew-symmetric Tensor ....................................................................................... 20
2.15. Quotient Law ....................................................................................................... 24
xii Tensors and Their Applications

2.16. Conjugate (or Reciprocal) Symmetric Tensor .......................................................... 25


2.17. Relative Tensor .................................................................................................... 26
Examples ............................................................................................................ 26
Exercises ............................................................................................................. 29

Chapter–3 Metric Tensor and Riemannian Metric ............................................ 31-54


3.1. The Metric Tensor ............................................................................................... 31
3.2. Conjugate Metric Tensor (Contravariant Tensor) ...................................................... 34
3.3. Length of a Curve ................................................................................................ 42
3.4. Associated Tensor ................................................................................................ 43
3.5. Magnitude of Vector ............................................................................................. 43
3.6. Scalar Product of Two Vectors .............................................................................. 44
3.7. Angle Between Two Vectors .................................................................................. 45
3.8. Angle Between Two Coordinate Curves .................................................................. 47
3.9. Hypersurface ....................................................................................................... 48
3.10. Angle Between Two Coordinate Hyper surface ........................................................ 48
3.11. n-Ply Orthogonal System of Hypersurfaces ............................................................. 49
3.12. Congruence of Curves .......................................................................................... 49
3.13. Orthogonal Ennuple .............................................................................................. 49
Examples ............................................................................................................ 52
Exercises ............................................................................................................. 54

Chapter–4 Christoffel’s Symbols and Covariant Differentiation ................................ 55-84


4.1. Christoffel’s Symbol............................................................................................. 55
4.2. Transformtion of Christoffel’s Symbols .................................................................. 64
4.3. Covariant Differentiation of a Covariant Vector ........................................................ 67
4.4. Covariant Differentiation of a Contravariant Vector ................................................... 68
4.5. Covariant Differentiation of Tensors ....................................................................... 69
4.6. Ricci’s Theorem .................................................................................................. 71
4.7. Gradient, Divergence and Curl ............................................................................... 75
4.8. The Laplacian Operator ......................................................................................... 80
Exercises ............................................................................................................. 83

Chapter–5 Riemann-Christoffel Tensor ............................................................ 85-110


5.1. Riemann-Christoffel Tensor ................................................................................... 85
5.2. Ricci Tensor ........................................................................................................ 88
5.3. Covariant Riemann-Christoffel Tensor .................................................................... 89
5.4. Properties of Riemann-Christoffel Tensors of First Kind Ri j k l ................................... 91
5.5. Bianchi Identity .................................................................................................... 94
5.6. Einstein Tensor .................................................................................................... 95
5.7. Riemannian Curvature of V n .................................................................................. 96
Contents xiii

5.8. Formula For Riemannian Curvature in Terms of Covariant


Curvature Tensor of V n ......................................................................................... 98
5.9. Schur’s Theorem ............................................................................................... 100
5.10. Mean Curvature ................................................................................................. 101
5.11. Ricci Principal Directions .................................................................................... 102
5.12. Einstein Space ................................................................................................... 103
5.13. Weyl Tensor or Projective Curvature Tensor ......................................................... 104
Examples .......................................................................................................... 106
Exercises ........................................................................................................... 109

Chapter–6 The e-systems and the Generalized Krönecker Deltas ................ 111-115
6.1. Completely Symmetric ......................................................................................... 111
6.2. Completely Skew-symmetric ................................................................................ 111
6.3. e-system ........................................................................................................... 112
6.4. Generalized Krönecker Delta ................................................................................ 112
6.5. Contraction of ä iá jkâ ã ............................................................................................ 114
Exercises ........................................................................................................... 115

Chapter–7 Geometry ........................................................................................ 116-141


7.1. Length of Arc .................................................................................................... 116
7.2. Curvilinear Coordinates in E 3 .............................................................................. 120
7.3. Reciprocal Base System Covariant and Contravariant Vectors .................................. 122
7.4. On The Meaning of Covariant Derivatives ............................................................. 127
7.5. Intrinsic Differentiation ....................................................................................... 131
7.6. Parallel Vector Fields ........................................................................................... 134
7.7. Geometry of Space Curves ................................................................................. 134
7.8. Serret-Frenet Formulae ....................................................................................... 138
7.9. Equations of A Straight Line ................................................................................ 140
Exercises ........................................................................................................... 141

Chapter–8 Analytical Mechanics ..................................................................... 142-169


8.1. Introduction ...................................................................................................... 142
8.2. Newtonian Laws ................................................................................................ 142
8.3. Equations of Motion of Particle ............................................................................ 143
8.4. Conservative Force Field ..................................................................................... 144
8.5. Lagrangean Equation of Motion ........................................................................... 146
8.6. Applications of Lagrangean Equations ................................................................... 152
8.7. Hamilton’s Principle ............................................................................................ 153
8.8. Integral Energy .................................................................................................. 155
8.9. Principle of Least Action ..................................................................................... 156
xiv Tensors and Their Applications

8.10. Generalized Coordinates ...................................................................................... 157


8.11. Lagrangean Equation of Generalized Coordinates ................................................... 158
8.12. Divergence Theorem, Green’s Theorem, Laplacian Operator and Stoke’s
Theorem in Tensor Notation ................................................................................ 161
8.13. Gauss’s Theorem ............................................................................................... 164
8.14. Poisson’s Equation ............................................................................................. 166
8.15. Solution of Poisson’s Equation ............................................................................. 167
Exercises ........................................................................................................... 169

Chapter–9 Curvature of a Curve, Geodesic .................................................... 170-187


9.1. Curvature of Curve, Principal Normal................................................................... 170
9.2. Geodesics ......................................................................................................... 171
9.3. Euler’s Condition ............................................................................................... 171
9.4. Differential Equations of Geodesics ...................................................................... 173
9.5. Geodesic Coordinates ......................................................................................... 175
9.6. Riemannian Coordinates ...................................................................................... 177
9.7. Geodesic Form of a Line Element ........................................................................ 178
9.8. Geodesics in Euclidean Space .............................................................................. 181
Examples .......................................................................................................... 182
Exercises ........................................................................................................... 186

Chapter–10 Parallelism of Vectors ................................................................. 188-204


10.1. Parallelism of a Vector of Constant Magnitude (Levi-Civita’s Concept) ..................... 188
10.2. Parallelism of a Vector of Variable Magnitude ......................................................... 191
10.3. Subspace of Riemannian Manifold ........................................................................ 193
10.4. Parallelism in a Subspace .................................................................................... 196
10.5. Fundamental Theorem of Riemannian Geometry Statement ..................................... 199
Examples .......................................................................................................... 200
Exercises ........................................................................................................... 203

Chapter–11 Ricci’s Coefficients of Rotation and Congruence ....................... 205-217


11.1. Ricci’s Coefficient of Rotation ............................................................................. 205
11.2. Reason for the Name “Coefficients of Rotation” .................................................... 206
11.3. Curvature of Congruence .................................................................................... 207
11.4. Geodesic Congruence ......................................................................................... 208
11.5. Normal Congruence ........................................................................................... 209
11.6. Curl of Congruence ............................................................................................ 211
11.7. Canonical Congruence ........................................................................................ 213
Examples .......................................................................................................... 215
Exercises ........................................................................................................... 217
Contents xv

Chapter–12 Hypersurfaces .............................................................................. 218-242


12.1. Introduction ...................................................................................................... 218
12.2. Generalized Covariant Differentiation .................................................................... 219
12.3. Laws of Tensor Differentiation ............................................................................ 220
12.4. Gauss’s Formula ................................................................................................ 222
12.5. Curvature of a Curve in a Hypersurface and Normal Curvature, Meunier’s Theorem,
Dupin’s Theorem ............................................................................................... 224
12.6. Definitions ......................................................................................................... 227
12.7. Euler’s Theorem ................................................................................................ 228
12.8. Conjugate Directions and Asymptotic Directions in a Hypersurface.......................... 229
12.9. Tensor Derivative of Unit Normal......................................................................... 230
12.10. The Equation of Gauss and Codazzi ..................................................................... 233
12.11. Hypersurfaces with Indeterminate Lines of Curvature ............................................ 234
12.12. Central Quadratic Hypersurfaces .......................................................................... 235
12.13. Polar Hyperplane ................................................................................................ 236
12.14. Evolute of a Hypersurface in an Euclidean Space ................................................... 237
12.15. Hypersphere ......................................................................................................238
Exercises ........................................................................................................... 241
Index .................................................................................................................... 243-245
This page
intentionally left
blank
CHAPTER – 1

PRELIMINARIES

1.1 n-DIMENSIONAL SPACE


In three dimensional rectangular space, the coordinates of a point are (x, y, z). It is convenient to write
(x1, x2, x3) for (x, y, z). The coordinates of a point in four dimensional space are given by (x1, x2, x3, x4).
In general, the coordinates of a point in n-dimensional space are given by (x1, x2, x3,...., xn) such n-
dimensional space is denoted by V n.

1.2 SUPERSCRIPT AND SUBSCRIPT


In the symbol Aklij , the indices i, j written in the upper position are called superscripts and k, l written
in the lower position are called subscripts.

1.3 THE EINSTEIN'S SUMMATION CONVENTION


n
Consider the sum of the series S = a1 x1 + a2 x 2 + ... + a n x n = Σ a i x i . By using summation convention,
i =1
drop the sigma sign and write convention as
n
Σ ai x i = a i x i
i =1

This convention is called Einstein’s Summation Convention and stated as


“If a suffix occurs twice in a term, once in the lower position and once in the upper position then
that suffix implies sum over defined range.”
If the range is not given, then assume that the range is from 1 to n.

1.4 DUMMY INDEX


Any index which is repeated in a given term is called a dummy index or dummy suffix. This is also called
Umbral or Dextral Index.
e.g. Consider the expression ai xi where i is dummy index; then
ai xi = a1 x1 + a 2 x 2 + ⋅ ⋅ ⋅ + an x n
2 Tensors and Their Applications

and a jx j = a1x1 + a2 x2 + ⋅ ⋅ ⋅ + an x n
These two equations prove that
a ix i = a j x j
So, any dummy index can be replaced by any other index ranging the same numbers.

1.5 FREE INDEX


Any index occurring only once in a given term is called a Free Index.
e.g. Consider the expression aij x i where j is free index.

1.6 KRÖNECKER DELTA

The symbol δij , called Krönecker Delta (a German mathematician Leopold Krönecker, 1823-91 A.D.)
is defined by
1 if i = j
δij = 
0 if i ≠ j
Similarly δij and δij are defined as
1 if i = j
δ ij = 0 if i ≠ j

1 if i = j
and δ ij = 
0 if i ≠ j
Properties
1. If x1, x2, ... xn are independent coordinates, then
∂x i
= 0 if i ≠ j
∂x j
∂x i
= 1 if i = j
∂x j
This implies that
∂x i
= δij
∂x j
∂x i ∂x k
It is also written as = δ ij .
∂x k ∂x j
2. δ ii = δ11 + δ 22 + δ 33 + ⋅ ⋅ ⋅ + δ nn (by summation convention)
δ ii = 1 + 1+ 1 + ⋅ ⋅ ⋅ + 1
δ ii = n
3. a ij δkj = a ik

a 3 j δ 2j = a δ2 + a δ2 + a δ2 + ⋅ ⋅ ⋅ + a δ 2
31 1 32 2 33 3 3n n
Since (as j is dummy index)
Preliminaries 3

= a32 (as ä 12 = ä 32 = ⋅ ⋅ ⋅ = ä n2 = 0 and ä 22 = 1)


In general,
a ijδ kj = a i 1δ1k + a i 2 δ2k + a i 3 δ 3k + ⋅ ⋅ ⋅ + aik δkk + ⋅ ⋅ ⋅ + a in δ nk
ik
a ij δkj = a (as δ1k = δ2k = ⋅ ⋅ ⋅ = δnk = 0 and δ kk = 1)
i j
4. δ j δk = δik

δ ij δkj = δ1i δ1k + δi2 δ 2k + δi3 δ3k + ⋅ ⋅ ⋅ + δ ii δ ik + ⋅ ⋅ ⋅ + δ in δ nk

= δik (as ä1i = δi2 = δi3 = ⋅ ⋅ ⋅ = δin = 0 and δ ii = 1)

EXAMPLE 1
dφ ∂φ dx1 ∂φ dx 2 ∂φ dx n
Write = 1 + 2 + ⋅⋅⋅ + n using summation convention.
dt ∂x dt ∂x dt ∂x dt
Solution
dφ ∂φ dx 1 ∂φ dx 2 ∂φ dx n
= + + ⋅ ⋅ ⋅ +
dt ∂x 1 dt ∂x 2 dt ∂x n dt
dφ ∂φ dx i
= i
dt ∂x dt

EXAMPLE 2
Expand: (i) aij xixj; (ii) glm gmp
Solution
aij x i x j = a1 j x x + a2 j x x + ⋅ ⋅ ⋅ + a nj x x
1 j 2 j n j
(i)
= a11 x1 x1 + a22 x 2 x 2 + ⋅ ⋅ ⋅ + ann x n x n
aij x i x j = a11 ( x1 )2 + a22 ( x 2 ) 2 + ⋅ ⋅ ⋅ + ann (x n )2
(as i and j are dummy indices)

(ii) g lm g mp = g l 1 g 1 p + g l 2 g 2 p + ⋅ ⋅ ⋅ + g ln g np , as m is dummy index.

EXAMPLE 3
If aij are constant and aij = aji, calculate:
∂ ∂
(i) ∂x (aij x i x j ) (ii) ∂x ∂x (aij xi x j )
k k l

Solution
∂ ∂
(i) (aij x i x j ) = a ij ( xi x j )
∂x k ∂xk
4 Tensors and Their Applications

∂x j ∂x
= a ij xi ∂ x + a ij x j ∂ x
i

k k
∂x j
= a ij xi δ jk + aij x j δ ik , as = δ jk
∂x k
= (aij δ jk ) xi + (aij δ ik ) x j
= aik xi + akj xj as aij δjk = aik
= aik xi + aki xi as j is dummy index
∂(aij x i x j )
∂ xk = 2aik xi as given aik = aki

∂(aij x i x j )
(ii) = 2aikxi
∂x k
Differentiating it w.r.t. xl :
∂ 2 (aij xi x j ) ∂xi
= 2 aik
∂xk ∂xl ∂xl
= 2 aik δ il

∂ 2 (aij xi x j )
= 2alk as a ik δil = alk.
∂xk ∂xl

EXAMPLE 4
If aij xi xj= 0
where aij are constant then show that
aij + aji = 0
Solution
Given
aijx ix j= 0
⇒ almxlxm= 0 since i and j are dummy indices
Differentiating it w.r.t. xi partially,

(alm x l x m ) = 0
∂xi
∂ l m
a lm (x x ) = 0
∂xi

∂x l m ∂x m l
a lm x + alm x =0
∂xi ∂x i

∂x l ∂x m
Since = δ li and = δmi
∂xi ∂xi
Preliminaries 5

almδ il x m + alm δ im x l = 0
aim x m + ali x l = 0
as alm δli = aim and alm δmi = ali .
Differentiating it w.r.t. xj partially
∂x m ∂x l
aim + ali
∂x j ∂x j = 0
aim δmj + ali δ lj = 0
aij + aji= 0 Proved.

EXERCISES

1. Write the following using the summation convention.


(i) (x 1 )2 + (x2 )2 + (x3 )2 + . . . + (xn )2
(ii) ds2 = g 11 (dx1 )2 + g 22 (dx2 )2 + . . . + g nn (dxn )2
(iii) a 1 x1 x3 + a 2 x2 x3 + . . . + a n xn x3
2. Expand the following:
∂ i
(i) a ij xj (ii) i
( ga ) (iii) Aik Bi
∂x
3. Evaluate:
(i) x j δ ij (ii) δ ij δkj δkl (iii) δ ij δij

4. Express b i j y i y j in the terms of x variables where yi = cij xj and b ij cik = δ ik .

ANSWERS

1. (i) xi xj (ii) ds2 = g ij dxi dxj (iii) a i x i x 3 .


2. (i) a i1 x + a i2 x2 + a i3 x3 + . . . + a in xn
1

∂ 1 ∂ 2 ∂ n
(ii) 1
( ga ) + 2
( g a ) + ⋅⋅⋅ + n
( ga )
∂x ∂x ∂x

(iii) A1k B1 + A2k B2 + ... + Ank Bn

3. (i) xi (ii) δ il (iii) n

4. Cij xi xj
CHAPTER – 2

TENSOR ALGEBRA

2.1 INTRODUCTION
A scalar (density, pressure, temperature, etc.) is a quantity whose specification (in any coordinate
system) requires just one number. On the other hand, a vector (displacement, acceleration, force, etc.)
is a quantity whose specification requires three numbers, namely its components with respect to some
basis. Scalers and vectors are both special cases of a more general object called a tensor of order
n whose specification in any coordinate system requires 3n numbers, called the components of tensor.
In fact, scalars are tensors of order zero with 3° = 1 component. Vectors are tensors of order one with
31 = 3 components.

2.2 TRANSFORMATION OF COORDINATES


In three dimensional rectangular space, the coordinates of a point are (x, y, z) where x, y, z are real
numbers. It is convenient to write (x1, x2, x3) for ( x, y , z ) or simply xi where i = 1, 2, 3. Similarly in
n- dimensional space, the coordinate of a point are n-independent variables (x1, x2,..., x n) in X-coordinate
system. Let ( x 1 , x 2 ,..., x n ) be coordinate of the same point in Y-coordinate system.
Let x 1 , x 2 , …, x n be independent single valued function of x1, x2,...., xn, so that,
x 1 = x 1 (x 1 , x 2 , ...., x n )

x 2 = x 2 ( x1 , x 2 , ...., x n )

x 3 = x 3 (x 1 , x 2 , ...., x n )

M M
x n = x n (x1 , x 2 , ..., x n )
or
x i = x i ( x1 , x 2 , ..., x n ) ; i = 1, 2, …, n …(1)
Tensor Algebra 7

Solving these equations and expressing x i as functions of x 1 , x 2 ,..., x n , so that

xi = x i ( x1 , x 2 ,..., x n ); i = 1, 2, ..., n
The equations (1) and (2) are said to be a transformation of the coordinates from one coordinate
system to another

2.3 COVARIANT AND CONTRAVARIANT VECTORS (TENSOR OF RANK ONE)


Let (x1, x2, ..., xn) or x i be coordinates of a point in X-coordinate system and ( x 1 , x 2 ,..., x n ) or x i be
coordinates of the same point in the Y-coordinate system.
Let A i, i = 1, 2, ..., n (or A 1, A 2, ..., A n) be n functions of coordinates x 1, x2, ..., x n
in X-coordinate system. If the quantities A i are transformed to A i in Y-coordinate system then according
to the law of transformation

∂x i j ∂x j
Ai = A or A j
= i
∂x j ∂x i A
Then A i are called components of contravariant vector.
Let Ai , i = 1, 2,..., n (or A 1, A 2, …, A n) be n functions of the coordinates x1, x2, ..., x n
in X-coordinate system. If the quantities Ai are transformed to Ai in Y-coordinate system then
according to the law of transformation

∂x j ∂x i
Ai =
i
Aj or Aj = Ai
∂x ∂x j
Then A i are called components of covariant vector.
The contravariant (or covariant) vector is also called a contravariant (or covariant) tensor of rank
one.
Note: A superscript is always used to indicate contravariant component and a subscript is always used to indicate
covariant component.

EXAMPLE 1
If xi be the coordinate of a point in n-dimensional space show that dxi are component of a
contravariant vector.
Solution
1 2 n
Let x1, x2, ..., xn or xi are coordinates in X-coordinate system and x , x ,..., x or x i are
coordinates in Y-coordinate system.
If
x i = x i (x1 , x 2 ,..., x n )

dx i = ∂x dx 1 + ∂ x dx 2 + ⋅ ⋅ ⋅ + ∂ x dx n
i i i

1
∂x 2
∂x n
∂x
8 Tensors and Their Applications

∂x i
d xi = dx j
∂x j
It is law of transformation of contravariant vector. So, dx i are components of a contravariant
vector.

EXAMPLE 2

∂φ
Show that is a covariant vector where φ is a scalar function.
∂x i

Solution

Let x1, x2, ..., xn or x i are coordinates in X-coordinate system and x 1 , x 2 , ..., x n or x i are
coordinates in Y-coordinate system.
Consider φ( x 1 , x 2 ,..., x n ) = φ( x1 , x 2 ,..., x n )

∂φ 1 ∂φ 2 ∂φ
∂φ = ∂x + 2 ∂x + ⋅ ⋅ ⋅ + n ∂x n
∂x 1
∂x ∂x

∂φ ∂φ ∂x1 ∂φ ∂x 2 ∂φ ∂x n
= + + ⋅ ⋅ ⋅ +
∂x i
∂x 1 ∂x i ∂x 2 ∂x i ∂x n ∂x i

∂φ ∂φ ∂x j
=
∂x i ∂x j ∂x i

∂φ ∂x j ∂φ
or =
∂x i ∂x i ∂x j
∂φ
It is law of transformation of component of covariant vector. So, is component of covariant
∂x i
vector.

EXAMPLE 3
Show that the velocity of fluid at any point is a component of contravariant vector
or
Show that the component of tangent vector on the curve in n-dimensional space are component
of contravariant vector.

Solution
dx1 dx 2 dx n
Let , , …, be the component of the tangent vector of the point ( x1 , x 2 ,..., x n ) i.e.,
dt dt dt
dx i
be the component of the tangent vector in X-coordinate system. Let the component of tangent
dt
Tensor Algebra 9

dx i
vector of the point ( x 1 , x 2 ,..., x n ) in Y-coordinate system are . Then x 1 , x 2 , ..., x n or x i being a
dt
function of x1 , x 2 , ..., x n which is a function of t. So,

d xi ∂x i dx1 ∂x i dx 2 ∂x i dx n
= + + ⋅ ⋅ ⋅ +
dt dt1 dt dx 2 dt dx n dt

d xi ∂x i dx j
=
dt dx j dt
dx i
It is law of transformation of component of contravariant vector. So, is component of
dt
contravariant vector.
i.e. the component of tangent vector on the curve in n-dimensional space are component of
contravariant vector.

2.4 CONTRAVARIANT TENSOR OF RANK TWO


Let A ij (i, j = 1, 2, ..., n) be n2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the
quantities A ij are transformed to A ij in Y-coordinate system having coordinates x 1 , x 2 , ..., x n . Then
according to the law of transformation
∂x i ∂x j kl
A ij = A
∂x k ∂x l
ij
Then A are called components of Contravariant Tensor of rank two.

2.5 COVARIANT TENSOR OF RANK TWO


Let A ij (i, j = 1, 2, ..., n) be n 2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the
quantities Aij are transformed to Aij in Y-coordinate system having coordinates x 1 , x 2 ,...,x n , then
according to the law of transformation,
∂x k ∂x l
Aij = Akl
∂x i ∂x j
Then A ij called components of covariant tensor of rank two.

2.6 MIXED TENSOR OF RANK TWO


Let Aij (i, j = 1, 2, ..., n) be n2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the
quantities Aij are transformed to A ji in Y-coordinate system having coordinates x 1 , x 2 ,..., x n , then
according to the law of transformation
∂x i ∂x l k
= A ji Al
∂x k ∂x
Then Aij are called components of mixed tensor of rank two.
10 Tensors and Their Applications

Note: (i) The rank of the tensor is defined as the total number of indices per component.
(ii) Instead of saying that “ A ij are the components of a tensor of rank two” we shall often say “ Aij is a tensor
of rank two.”

THEOREM 2.1 To show that the Krönecker delta is a mixed tensor of rank two.
Solution
Let X and Y be two coordinate systems. Let the component of Kronecker delta in X-coordinate
system δij and component of Krönecker delta in Y-coordinate be δij , then according to the law of
transformation
∂x i ∂x i ∂x l ∂x k
δij = =
∂x j ∂x k ∂x j ∂x l
∂x i ∂x l k
δij = δl
∂x k ∂x j
This shows that Krönecker δij is mixed tensor of rank two.

EXAMPLE 4
∂Ai
If Ai is a covariant tensor, then prove that do not form a tensor..
∂x j
Solution
Let X and Y be two coordinate systems. As given Ai is a covariant tensor. Then

∂x k
Ai = Ak
∂x i
Differentiating it w.r.t. x j
∂Ai ∂  ∂x k 
 
∂x j =
∂x j  ∂x i Ak 
 
∂Ai ∂x k ∂Ak ∂ 2 xk
= + A …(1)
∂x j ∂x i ∂x j
k
∂x i ∂x j
∂Ai
It is not any law of transformation of tensor due to presence of second term. So, is not a
∂x j
tensor.

THEOREM 2.2 To show that δij is an invariant i.e., it has same components in every coordinate
system.
Proof: Since δij is a mixed tensor of rank two, then

∂x i ∂x l k
δij = δl
∂x k ∂x j
Tensor Algebra 11

∂x i  ∂x l k 
 
=
∂x k  ∂x j δ l 
 
∂ x i ∂x k ∂x l k ∂x k
= , as δl =
∂x ∂ x
k j
∂x j ∂x j
∂x i ∂x i
δij = = δ i
, as = δij
∂x j
j
∂x j
So, δij is an invariant.

THEOREM 2.3 Prove that the transformation of a contravariant vector is transitive.


or
Prove that the transformation of a contravariant vector form a group.
Proof: Let Ai be a contravariant vector in a coordinate system x i (i = 1, 2,...,n) . Let the coordinates

x i be transformed to the coordinate system x i and x i be transformed to x i .


When coordinate x i be transformed to x i , the law of transformation of a contravariant vector is
∂x p q
Ap = A ... (1)
∂x q
When coordinate x i be transformed to x i , the law of transformation of contravariant vector is
∂x i p
Ai = A
∂x p
∂x i ∂x p q
A i = ∂x i ∂x q A from (1)
∂x i q
A i = A
∂x q
This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of contravariant vectors is transitive or form a group.

THEOREM 2.4 Prove that the transformation of a covariant vector is transitive.


or
Prove that the transformation of a covariant vector form a group.
Proof: Let Ai be a covariant vector in a coordinate system x i (i = 1, 2, ..., n) . Let the coordinates x i be
transformed to the coordinate system x i and x i be transformed to x i .
When coordinate x i be transformed to x i , the law of transformation of a covariant vector is
∂x q
Ap = Aq ... (1)
∂x p
When coordinate x i be transformed to x i , the law of transformation of a covariant vector is
∂x p
Ai = Ap
∂x i
12 Tensors and Their Applications

∂x p ∂x q
Ai = Aq
∂x i ∂x p
∂x q
Ai = Aq
∂x i
This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of covariant vectors is transitive or form a group.

THEOREM 2.5 Prove that the transformations of tensors form a group


or
Prove that the equations of transformation a tensor (Mixed tensor) posses the group property.
Proof: Let Aij be a mixed tensor of rank two in a coordinate system x i (i = 1, 2,...,n) . Let the coordinates
x i be transformed to the coordinate system x i and x i be transformed to x i .
When coordinate x i be transformed to x i , the transformation of a mixed tensor of rank two is
∂x p ∂x s r
Aqp = As ... (1)
∂x r ∂x q
When coordinate x i be transformed to x i , the law of transformation of a mixed tensor of rank
two is
∂x i ∂x q p
A ji = Aq
∂x p ∂x j

∂x ∂x ∂x ∂x
i q p s
r
= p j r q
As from (1)
∂x ∂x ∂x ∂x

∂x i ∂x s r
A ji = As
∂x r ∂x j

This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of tensors form a group.
THEOREM 2.6 There is no distinction between contravariant and covariant vectors when we restrict
ourselves to rectangular Cartesian transformation of coordinates.
Proof: Let P(x, y) be a point with respect to the rectangular Cartesian axes X and Y. Let ( x , y ) be the
coordinate of the same point P in another rectangular cartesian axes X and Y , Let (l1, m1) and (l2, m2)
be the direction cosines of the axes X , Y respectively. Then the transformation relations are given by
x = l1 x + m1 y 
 ...(1)
y = l2 x + m2 y 
and solving these equations, we have
x = l1 x + l 2 y 
 ...(2)
y = m1 x + m2 y 
put x = x1 , y = x 2 , x = x 1 , y = x 2
Tensor Algebra 13

Consider the contravariant transformation


∂x i j
A i = A ; j = 1,2
∂x j
∂x i 1 ∂x i 2
A i = A + 2A
∂x1 ∂x
for i = 1, 2 .
∂x 1 1 ∂x 1 2
A 1 = A + 2A
∂x1 ∂x
∂x 1 ∂x 2 2
2

A 2 = A + 2 A
∂x1 ∂x
∂x
= l , but x = x1 , y = x , x = x 1 , y = x
2 2
From (1)
∂x 1
Then
∂x ∂x 1
= =l .
∂x ∂x 1 1
Similarly,
∂x ∂x 1 
= m1 = 2 ; 
∂y ∂x 

∂y ∂x 2
∂y ∂x
= m2 = 2 
2 ...(3)
= l2 = 1 ;
∂x ∂x ∂y ∂x 
So, we have
A 1 = l1 A1 + m1 A 2 
 ..(4)
A 2 = l2 A1 + m 2 A 2 
Consider the covariant transformation
∂x j
Ai = A j ; j = 1,2
∂x i
∂x1 ∂x 2
Ai = A + A2
∂x i
1
∂x1
for i = 1, 2 .
∂x1 ∂x 2
A1 = A + A2
∂x 1
1
∂x 1
∂x 1 ∂x 2
A2 = A + A2
∂x 2
1
∂x 2
From (3)
A1 = l1 A1 + m1 A2 
 ...(5)
A2 = l2 A1 + m2 A2 
14 Tensors and Their Applications

So, from (4) and (5), we have


A 1 = A1 and A 2 = A2
Hence the theorem is proved.

2.7 TENSORS OF HIGHER ORDER


(a) Contravariant tensor of rank r
i i ...i
Let Ai1i 2 ...ir be nr function of coordinates x1, x2, ..., xn in X-coordinates system. If the quantities A i 2 r
are transformed to A i1i 2 ...i r in Y-coordinate system having coordinates x 1 , x 2 , ..., x n . Then according
to the law of transformation

∂x i1 ∂x i2 ∂x i r
A i1i 2 ...i r
= K
p1 p 2 ...p r
∂x p1 ∂x p 2 ∂x p r A
Then Ai ii 2 ...i r are called components of contravariant tensor of rank r.
(b) Covariant tensor of rank s

Let A j1 j 2 ... j s be n s functions of coordinates x1, x2, ..., xn in X-coordinate system. If the quantities
A j1 j 2 ... j s are transformed to A j1 j 2 ... j s in Y- coordinate system having coordinates x 1 , x 2 ,...,x n . Then
according to the law of transformation

∂x q1 ∂x q 2 ∂x q s
A j1 j 2 ... j s = ⋅ ⋅ ⋅ Aq1 ,q 2 ...,q s
∂x j1 ∂x j 2 ∂x j s
Then A j1 j 2 ... j s are called the components of covariant tensor of rank s.

(c) Mixed tensor of rank r + s


i i ...i
Let A j11 2j 2 ...rj s be nr+s functions of coordinates x1, x2, ..., xn in X-coordinate system. If the quantities
Aij11i 2j 2......i rj s are transformed to A ij11ij22......i rj s in Y-coordinate system having coordinates x 1 , x 2 , … , x n . Then
according to the law of transformation
∂x i1 ∂x i 2 ∂x ir ∂x q1 ∂x q 2 ∂x q s
A ij11ij22......i rj s = ⋅ ⋅ ⋅ Aqp11qp22......q ps r
∂x p1 ∂x p2 ∂x p r ∂x ∂x
j1 j2
∂x js

i i ...i
Then A j11 2j 2 ...rj s are called component of mixed tensor of rank r + s .

A tensor of type Ai1j1i 2j...


2 ... j s
ir is known as tensor of type (r, s ) , In (r,s), the first component r
indicates the rank of contravariant tensor and the second component s indicates the rank of covariant
tensor.
ij
Thus the tensors Aij and A are type (0, 2) and (2, 0) respectively while tensor Aij is type (1, 1).
Tensor Algebra 15

EXAMPLE
ijk
Alm is a mixed tensor of type (3, 2) in which contravariant tensor of rank three and covariant
tensor of rank two. Then according to the law of transformation

∂x i ∂x j ∂x k ∂x a ∂x b αβγ
Almijk = Aab
∂x α ∂x β ∂xγ ∂x l ∂x m

2.8 SCALAR OR INVARIANT

A function φ ( x1 , x 2 , ..., x n ) is called Scalar or an invariant if its original value does not change upon
transform ation of coordinates from x1, x2, ..., xn to x 1 , x 2 , ..., x n . i.e.

φ( x1 , x 2 ,..., x n ) = φ( x 1 , x 2 ,..., x n )
Scalar is also called tensor of rank zero.
For example, Ai Bi is scalar..

2.9 ADDITION AND SUBTRACTION OF TENSORS

THEOREM 2.7 The sum (or difference) of two tensors which have same number of covariant and the
same contravariant indices is again a tensor of the same rank and type as the given tensors.
Proof: Consider two tensors Aij11i 2j 2......i rj s and B ij11i 2j 2......i rj s of the same rank and type (i.e., covariant tensor of
rank s and contravariant tensor of rank r.). Then according to the law of transformation
∂x i1 ∂x i 2 ∂x ir ∂x q1 ∂x q 2 ∂x q s
A ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Aqp11qp22......q ps r
∂x ∂x
p1 p2
∂x ∂x ∂x
pr j1 j2
∂x js

and
∂x i1 ∂x i 2 ∂x ir ∂x q1 ∂x q 2 ∂x q s
B ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Bqp11qp22......q ps r
∂x p1 ∂x p2 ∂x p r ∂x j1 ∂x j2 ∂x j s
Then

A ji11 ij22......i rj s ± B ij11ij22......i rj s =


∂x i1 ∂x i 2
∂x p1 ∂x p2
⋅ ⋅ ⋅
∂x i r ∂x q1 ∂x q 2 … ∂x q s
∂x p r ∂x j1 ∂x j 2 ∂x j s
(A p1 p2 ...p r
q1q 2 ...q s
± B qp11qp2 2......qps r )
If
A ij11ij22......i rj s ± B ji11ij22......irj s = C ij11ij22......i rj s
and
Aqp11qp2 2......q ps r ± Bqp11qp2 2......qps r = C qp11qp2 2......qps r
So,
∂x i1 ∂x i 2 ∂x ir ∂x q1 ∂x q 2 ∂x q s
⋅ ⋅ ⋅
C ij11ij22......i rj s = ⋅ ⋅ ⋅ Cqp11,,qp22,...,
,..., pr

∂x p1 ∂x p2 ∂x p r ∂x j1 ∂x j 2 ∂x j s qs

i , i ,..., i
This is law of transformation of a mixed tensor of rank r+s. So, C j11 , j22 ,..., jrs is a mixed tensor of
rank r+s or of type (r, s).
16 Tensors and Their Applications

EXAMPLE 5

If Akij and Bnlm are tensors then their sum and difference are tensors of the same rank and type.

Solution
As given Akij and Bkij are tensors. Then according to the law of transformation
∂x i ∂x j ∂x r pq
Akij = Ar
∂x p ∂x q ∂x k
and
∂x i ∂x j ∂x r pq
Bkij = Br
∂x p ∂x q ∂x k
then

Akij ± Bkij =
∂x i ∂x j ∂x r pq
∂x ∂x ∂x
p q k
(
Ar ± Brpq )
If
Akij ± B kij = C kij and Arpq ± B rpq = C rpq
So,
∂x i ∂x j ∂x r pq
Cr
C kij =
∂x p ∂x q ∂x k
The shows that C kij is a tensor of same rank and type as Akij and Bkij .

2.10 MULTIPLICATION OF TENSORS (OUTER PRODUCT OF TENSOR)


THEOREM 2.8 The multiplication of two tensors is a tensor whose rank is the sum of the ranks of
two tensors.
i i ...i
Proof: Consider two tensors A j11 2j 2 ...rj s (which is covariant tensor of rank s and contravariant tensor of
k k2 ...km
rank r) and Bl11l 2 ...
ln (which is covariant tensor of rank m and contravariant tensor of rank n). Then
according to the law of transformation.
∂x i1 ∂x i 2 ∂x ir ∂x q1 ∂x q 2 ∂x q r
A ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Aqp11qp22......q ps r
∂x p1 ∂x p2 ∂x p r ∂x j1 ∂x j 2 ∂x j s
and
∂x k1 ∂x k 2 ∂x km ∂xβ1 ∂x β2 ∂x βn
Bl1kl12k...2 ...km
= ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Bβα11βα2 2......βαn m
∂x α1 ∂x α 2 ∂x α m ∂x l1 ∂x l 2 ∂x l n
ln

Then their product is

∂x i1 ∂x i r ∂x q1 ∂ x q s ∂ x k1 ∂ x km ∂ x β 1 ∂x βn
A ij11ij22......i rj s Blk1l12k...2 ...l nkm = ⋅⋅⋅ ⋅⋅⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅
∂x p1 ∂x pr ∂x j1 ∂x j s ∂x α1 ∂x α m ∂x l1 ∂x l m

Aqp11qp22......q ps r Bβα11βα2 2......βαn m


Tensor Algebra 17

If
C ij1ij2 ......i rjk1l kl 2 ...
...km i i ...i k k ...k
1 2 s 1 2 ln
= A j11 j22 ...rjs Bl1l12 ...2 ln n
and
C qp11qp2 2......qpsβr α1β12α...2β...nαm = Aqp11qp22...,
...,p r
qs
Bβα11βα2 2......βαn m
So,
∂x i1 ∂x ir ∂x q 1 ∂x q s ∂x k1
C ji11ij22......irj ksl11kl 22 ...
...km
ln =
⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅
∂x p1 ∂x p r ∂x j1 ∂x j s ∂x α1
∂x k m ∂x β1 ∂ x βn
⋅ ⋅ ⋅ α ⋅ l ⋅ ⋅ ⋅ l C qp11qp2 2......q ps βrα1 β1α2 2......βαn h
∂x m ∂x 1 ∂x n
i i ...i ...k
This is law of transformation of a mixed tensor of rank r + m + s + n. . So, C j11 j22 ...rj ks1lk12l 2 ...lmn is a

mixed tensor of rank r + m + s + n . or of type (r + m, s + n ) . Such product is called outer product or


open proudct of two tensors.

THEOREM 2.9 If A i and B j are the components of a contravariant and covariant tensors of rank one
then prove that A iB j are components of a mixed tensor of rank two.
Proof: As Ai is contravariant tensor of rank one and B j is covariant tensor of rank one. Then
according to the law of transformation
∂x i k
Ai = A ...(1)
∂x k
and
∂x l
Bj = Bl ...(2)
∂x j
Multiply (1) and (2), we get
∂x i ∂x l k
Ai B j = A Bl
∂x k ∂x j
This is law of transformation of tensor of rank two. So, Ai B j are mixed tensor of rank two.
Such product is called outer product of two tensors.

EXAMPLE 6

Show that the product of two tensors Aij and B mkl is a tensor of rank five.

Solution
As Aij and B mkl are tensors. Then by law of transformation

∂x i ∂x q p ∂x k ∂x l ∂x t rs
A ji = Aq and Bmkl = Bt
∂x p ∂x j ∂x r ∂x s ∂x m
18 Tensors and Their Applications

Multiplying these, we get

∂x i ∂x q ∂x k ∂x l ∂x t p rs
A ji Bmkl = A B
∂x p ∂x j ∂x r ∂x s ∂x m q t
This is law of transformation of tensor of rank five. So, Aij B mkl is a tensor of rank five.

2.11 CONTRACTION OF A TENSOR


The process of getting a tensor of lower order (reduced by 2) by putting a covariant index equal to a
contravariant index and performing the summation indicated is known as Contraction.
In other words, if in a tensor we put one contravariant and one covariant indices equal, the
process is called contraction of a tensor.
ijk
For example, consider a mixed tensor Alm of order five. Then by law of transformation,

∂x i ∂x j ∂x k ∂x s ∂x t pqr
Almijk = Ast
∂x p ∂x q ∂x r ∂x l ∂x m
Put the covariant index l = contravariant index i, so that

∂x i ∂x j ∂x k ∂x s ∂x t pqr
Aimijk = Ast
∂x p ∂x q ∂x r ∂x i ∂x m

∂x j ∂x k ∂x s ∂xt pqr
= Ast
∂x q ∂x r ∂x p ∂x m

∂x j ∂x k ∂x t s pqr ∂x s
= δ p Ast Since = δ sp
∂x q ∂x r ∂x m ∂x p

∂x j ∂x k ∂x t pqr
Aimijk = A pt
∂x q ∂x r ∂x m
ijk
This is law of transformation of tensor of rank 3. So, Aim is a tensor of rank 3 and type (1, 2)
ijk
while Alm is a tensor of rank 5 and type (2, 3). It means that contraction reduces rank of tensor by
two.

2.12 INNER PRODUCT OF TWO TENSORS


Consider the tensors Akij and Bmn
l
if we first form their outer product Akij B mn
l
and contract this by
putting l = k then the result is Akij B mn
k
which is also a tensor, called the inner product of the given
tensors.
Hence the inner product of two tensors is obtained by first taking outer product and then
contracting it.

EXAMPLE 7
If A i and B i are the components of a contravariant and covariant tensors of rank are respectively
then prove that A iB i is scalar or invariant.
Tensor Algebra 19

Solution
As A i and B i are the components of a contravariant and covariant tensor of rank one respectively,
then according to the law of the transformation
∂x i p ∂x q
Ai= A and Bi = Bq
∂x p ∂x i
Multiplying these, we get
∂x i ∂x q p
ABi= i A Bq
∂x p ∂x i
∂x q p ∂x q
= A B , since = δ qp
∂x p
q
∂x p
= δ qp A p Bq
p
A i Bi = A B p
This shows that A iB i is scalar or Invariant.

EXAMPLE 8

If Aij is mixed tensor of rank 2 and B mkl is mixed tensor of rank 3. Prove that Aij B mjl is a mixed
tensor of rank 3.

Solution
As Aij is mixed tensor of rank 2 and B mkl is mixed tensor of rank 3. Then by law of transformation
∂x i ∂x q p ∂x k ∂x l ∂x t rs
A ji = A kl
q and Bm = Bt ...(1)
∂x p ∂x j ∂x r ∂x s ∂x m
Put k = j then
∂x j ∂x l ∂x t rs
Bmjl = Bt ...(2)
∂x r ∂x s ∂x m
Multiplying (1) & (2) we get
∂x i ∂x q ∂x j ∂x l ∂x t p rs
A ij B mjl = Aq Bt
∂x p ∂x j ∂x r ∂x s ∂x m
∂x i ∂x l ∂x t q p rs ∂x q ∂x j ∂x q
= δr Aq Bt since = = δ qr
∂x p ∂x s ∂x m ∂x j ∂x r ∂x r

∂x i ∂x l ∂x t p qs
A ji Bmjl = Aq B t since δ qr Btrs = Btqs
∂x p ∂x s ∂x m

This is the law of transformation of a mixed tensor of rank three. Hence Aij B mjl is a mixed tensor
of rank three.
20 Tensors and Their Applications

2.13 SYMMETRIC TENSORS


A tensor is said to be symmetric with respect to two contravariant (or two covariant) indices if its
components remain unchanged on an interchange of the two indices.

EXAMPLE
(1) The tensor A ij is symmetric if A ij = A ji
ijk
(2) The tensor Alm ijk
is symmetric if Alm = Almjik

1
THEOREM 2.10 A symmetric tensor of rank two has only n(n + 1) different components in n-
2
dimensional space.
Proof: Let A ij be a symmetric tensor of rank two. So that Aij = A ji .

 A11 A12 A13 L A1n 


 21 
A A 22 A 23 L A2n 
The component of A ij are  A31 A32 A33 L A 3n 
 
 M M M L M 
 A n1
 An2 An3 L A nn 

i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn are
different. Thus remaining components are (n2– n). In which A 12 = A 21, A 23 = A 32 etc. due to symmetry.
1 2
So, the remaining different components are (n − n) . Hence the total number of different
2
components
1 2 1
= n+ (n − n) = n(n + 1)
2 2

2.14 SKEW-SYMMETRIC TENSOR


A tensor is said to be skew-symmetric with respect to two contravariant (or two covariant) indices if
its components change sign on interchange of the two indices.

EXAMPLE

(i) The tensor A ij is Skew-symmetric of A ij = − A ji


ijk
(ii) The tensor Alm ijk
is Skew-symmetric if Alm = − Almjik
1
THEOREM 2.11 A Skew symmetric tensor of second order has only n(n − 1) different non-zeroo
2
components.
Proof: Let A ij be a skew-symmetric tensor of order two. Then Aij = − A ji .
Tensor Algebra 21

 0 A12 A13 L A1n 


 21 
A 0 A 23 L A2n 
The components of A are  A 31
ij
A 32 0 L A3n 
 
 M M M L M 
 A n1
 An 2 An3 L 0 

[Since A ii
= − Aii ⇒ 2 A ii = 0 ⇒ Aii = 0 ⇒ A11 = A 22 = ⋅ ⋅ ⋅ A nn = 0]
i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn
are zero. Omitting there, then the remaining components are n2 –n. In which A 12 = – A 21, A 13 = – A 31
1 2
etc. Ignoring the sign. Their remaining the different components are (n − n) .
2
1
Hence the total number of different non-zero components = n(n − 1)
2
Note: Skew-symmetric tensor is also called anti-symmetric tensor.

THEOREM 2.12 A covariant or contravariant tensor of rank two say Aij can always be written as the
sum of a symmetric and skew-symmetric tensor.
Proof: Consider a covariant tensor A ij. We can write A ij as
1 1
Aij = ( A + A ji ) + ( Aij − A ji )
2 ij 2
Aij = Sij + Tij
1 1
where Sij = ( Aij + A ji ) and Tij = ( Aij − A ji )
2 2
Now,
1
S ji = ( A + Aij )
2 ji
S ji = Sij
So, Sij is symmetric tensor..
and
1
Tij = ( A + A ji )
2 ij
1
T ji = ( A − Aij )
2 ji
1
= − ( Aij − A ji )
2
T ji = −Tij

or Tij = −T ji
So, Tij is Skew-symmetric Tensor..
22 Tensors and Their Applications

EXAMPLE 9

If φ = a jk A A . Show that we can always write φ = b jk A j Ak where b jk is symmetric.


j k

Solution
As given
φ = a jk A j A k ...(1)
Interchange the indices i and j
φ = ak j A k A j ...(2)

Adding (1) and (2),

2φ = (a jk + akj ) A j A k
1
φ = (a jk + akj ) A j A k
2
φ = b jk A j A k
1
where b jk = (a + akj )
2 jk
To show that b jk is symmetric.

Since
1
b jk = (a + akj )
2 jk
1
bkj = (a kj + a jk )
2
1
= (a jk + akj )
2
bkj = b jk

So, b jk is Symmetric.

EXAMPLE 10
 ∂Ti ∂T j 
If Ti be the component of a covariant vector show that  j − i  are component of a Skew-
 ∂x ∂x 
symmetric covariant tensor of rank two.

Solution
As Ti is covariant vector. Then by the law of transformation
∂x k
Ti = Tk
∂x i
Tensor Algebra 23

Differentiating it w.r.t. to x j partially,,

∂Ti ∂  ∂x k 
 
j = ∂x j  ∂x i Tk 
∂x  

∂2 x k ∂x k ∂Tk
= T +
∂x j ∂x i
k
∂x i ∂x j
∂T j ∂ 2 xk ∂x k ∂x l ∂Tk
= T + ...(1)
∂x j ∂x j ∂x i
k
∂x i ∂x j ∂x l
Similarly,
∂T j
∂ 2 xk ∂x k ∂x l ∂Tk
= T +
∂x i ∂x i ∂x j
k
∂x j ∂x i ∂x l
Interchanging the dummy indices k & l
∂T j ∂2 x k ∂x k ∂x l ∂Tl
= T + ...(2)
∂x i ∂x i ∂x j
k
∂x i ∂x j ∂x k
Substituting (1) and (2), we get

∂Ti ∂T j ∂x k ∂x l  ∂Tk ∂Tl 


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

∂Ti ∂T j
This is law of transformation of covariant tensor of rank two. So, − are component of
∂x j ∂x i
a covariant tensor of rank two.
∂Ti ∂T j
To show that − is Skew-symmetric tensor..
∂x j ∂x i
Let
∂Ti ∂T j
Aij = −
∂x j ∂x i
∂T j ∂Ti
A ji = −
∂xi ∂x j
 ∂T ∂T 
= −  ∂ x j − ∂x i 
i j

 
A ji = − Aij
or Aij = − A ji
∂Ti ∂T j
So, Aij = − is Skew-symmetric.
∂x j ∂x i
∂Ti ∂T j
So, − are component of a Skew-symmetric covariant tensor of rank two.
∂x j ∂x i
24 Tensors and Their Applications

2.15 QUOTIENT LAW


By this law, we can test a given quantity is a tensor or not. Suppose given quantity be A and we do not
know that A is a tensor or not. To test A, we take inner product of A with an arbitrary tensor, if this
inner product is a tensor then A is also a tensor.
Statement
If the inner product of a set of functions with an atbitrary tensor is a tensor then these set of
functions are the components of a tensor.
The proof of this law is given by the following examples.

EXAMPLE 11
Show that the expression A(i,j,k) is a covariant tensor of rank three if A(i,j,k)B k is covariant
tensor of rank two and B k is contravariant vector

Solution
Let X and Y be two coordinate systems.
As given A (i, j, k)B k is covariant tensor of rank two then

∂x p ∂x q
A ( i, j, k ) B k = A( p, q, r )B r ...(1)
∂x i ∂x j

Since B k is contravariant vector. Then


∂x k ∂x r
Bk = r Br or Br = k B k
∂x ∂x
So, from (1)
∂x p ∂x q ∂x r
A ( i, j, k ) B k = A ( p , q, r ) Bk
∂x i ∂x j ∂x k
∂x p ∂x q ∂x r
A ( i, j, k ) B k = A( p, q , r )B k
∂x i ∂x j ∂x k
∂x p ∂x q ∂ x r
A ( i, j , k ) = A( p , q , r )
∂ x i ∂x i ∂x k
As B k is arbitrary..
So, A(i , j , k ) is covariant tensor of rank three.

EXAMPLE 12
If A (i, j, k)A iB jCk is a scalar for arbitrary vectors A i, B j, Ck. Show that A(i, j, k) is a tensor of
type (1, 2).

Solution
Let X and Y be two coordinate systems. As given A(i , j, k ) A i B j C k is scalar. Then
i j p q
A (i, j, k ) A B C k = A( p, q, r ) A B C r ...(1)
Tensor Algebra 25

Since Ai , B i and C k are vectors. Then


∂x i p ∂x p i
Ai = A or Ap = A
∂x p ∂x i
∂x j q ∂x q j
Bj = B or Bq = B
∂x q ∂x j
∂x k ∂x r k
C k = r Cr or Cr = C
∂x ∂x k
So, from (1)
i j ∂x p ∂xq ∂x k
A (i, j , k ) A B C k = A( p, q, r) Ai B jCk
∂x i ∂x j ∂xr
As A i , B j , C k are arbitrary..
Then

∂x p ∂xq ∂x k
A ( i, j , k ) = A( p, q, r )
∂xi ∂x j ∂xr
So, A(i, j, k) is tensor of type (1, 2).

2.16 CONJUGATE (OR RECIPROCAL) SYMMETRIC TENSOR


Consider a covariant symmetric tensor Aij of rank two. Let d denote the determinant Aij with the
elements Aij i.e., d = Aij and d ≠ 0 .
Now, define A ij by
Cofactor of Aij is the determinan t Aij
A ij =
d
A is a contravariant symmetric tensor of rank two which is called conjugate (or Reciprocal) tensor
ij

of Aij .

THEOREM 2.13 If Bij is the cofactor of Aij in the determinant d = |Aij| ≠ 0 and Aij defined as
Bij
A ij =
d
Then prove that Aij A = . δ ki
kj

Proof: From the properties of the determinants, we have two results.


(i) Aij Bij = d
Bij
⇒ Aij =1
d
Bij
Aij Aij = 1, given Aij =
d
26 Tensors and Their Applications

(ii) Aij B kj = 0
B kj
Aij = 0, d ≠0
d
Aij Akj = 0 if i ≠ k
from (i) & (ii)
1 if i = k
Aij Akj = 0 if i ≠ k

i.e., Aij Akj = δik

2.17 RELATIVE TENSOR


i i ...i
If the components of a tensor A j11 2j 2 ...rj s transform according to the equation
ω
∂x ∂x k1 ∂x k2 ∂x k r ∂x j1 ∂x j2 ∂x js
Alk11l 2k...
2 ...k r
= Aij11i 2j 2......i rj s ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅
∂x
ls
∂x i1 ∂x i 2 ∂xi r ∂x l r ∂x l 2 ∂x l s
∂x
Hence A j11 2j 2 ...rj r is called a relative tensor of weight ω, where
i i ...i
is the Jacobian of transformation. If
∂x
ω = 1, the relative tensor is called a tensor density. If w = 0 then tensor is said to be absolute.

MISCELLANEOUS EXAMPLES
1. Show that there is no distinction between contravariant and covariant vectors when we
restrict ourselves to transformation of the type
x i = am x + b ;
i m i

where a's and b's are constants such that


a ir ami = δ rm
Solution
Given that
x i = am x + b
i m i
...(1)
or a im x m = x i − bi ...(2)
Multiplying both sides (2) by a ir , we get
a ir ami x m = a ir x i − bi a ir
δ rm x m = a ir x i − bi a ir as given a ir ami = δ rm
x r = a r x − b a r as δ rm x m = x r
i i i i

xs = as x − b as
i i i i
or
Differentiating Partially it w.r.t. to x i
∂x s
= a is ..(3)
∂x i
Tensor Algebra 27

Now, from (1)


x i = as x + b
i s i

∂x i
a is
s = ...(4)
∂x
The contravariant transformation is
∂x i s
A = A = a is A s
i ...(5)
∂x s
The covariant transformation is
∂x s
As = a is As
Ai = ...(6)
∂x i
Thus from (5) and (6), it shows that there is no distinction between contravariant and
covariant tensor law of transformation
2. If the tensors aij and g ij are symmetric and u i , v i are components of contravariant vectors
satisfying the equations
(aij − kgij )u i = 0, i, j = 1, 2,...,n
( aij − k'gij )v i = 0, k ≠ k ′ .
Prove that gij u i v j = 0, a ij u i v j = 0.

Solution
The equations are

(aij − kgij )u i = 0 ...(1)

(aij − k ′gij )v i = 0 ...(2)


j j
Multiplying (1) and (2) by u and v respectively and subtracting, we get
aij u i v j − aij v i u j − kgij u i v j + k ′gij u j v i = 0
Interchanging i and j in the second and fourth terms,

a ij u i v j − a ji v j u i − kgij u i v j + k ′g ji u i v j = 0

As aij and g ij is symmetric i.e., a ij = a ji & g ij = g ji

− kgij v j u i + k ′gij u i v j = 0

(k ′ − k ) g ij u v = 0
i j

g ij u i v j = 0 since k ≠ k ′ ⇒ k − k ′ ≠ 0
Multiplying (1) by v j , we get
a ij v j u i − kgij u i v j = 0
a ij u i v j = 0 as g ij u i v j = 0. Proved.
28 Tensors and Their Applications

3. If Aij is a Skew-Symmetric tensor prove that


i k i k
(δ j δl + δl δ j ) Aik = 0

Solution
Given Aij is a Skew-symmetric tensor then Aij = − A ji .
Now,
i k i k
(δ j δl + δl δ j ) Aik = δ ij δ kl Aik + δil δ kj Aik

= δ ij Ail + δil Aij

= A jl + Alj
i k i k
(δ j δl + δl δ j ) Aik = 0 as A jl = − Alj
4. If aij is symmetric tensor and bi is a vector and a ij bk + a jk bi + a kib j = 0 then prove that
a ij = 0 or bk = 0 .

Solution
The equation is
aij bk + a jk bi + akib j = 0

⇒ aij bk + a jk bi + akib j = 0
By tensor law of transformation, we have

∂x p ∂x q ∂x r ∂x p ∂x q ∂x r ∂x p ∂x q ∂x r
a pq b + a b + a b =0
∂x i ∂x j ∂x k
r pq
∂x j ∂x k
r
∂x i
pq
∂x k ∂x i
r
∂x j

 ∂x p ∂x q ∂x r ∂x p ∂x q ∂x r ∂x p ∂x q ∂x r 
a pq br  i + j +  = 0
 ∂x ∂x ∂x ∂x ∂x k ∂x i ∂x k ∂x i ∂x j 
j k

⇒ a pq br = 0 ⇒ a pq = 0 or br = 0
⇒ a ij = 0 or bk = 0
5. If a mn x x = bmn x x for arbitrary values of x r , show that a(mn) = b(mn) i.e.,
m n m n

a mn + a nm = bmn + bnm
If a mn and bmn are symmetric tensors then further show the a mn = bmn .

Solution
Given
a mn x m x n = bmn x m x n

(a mn − bmn )x m x n = 0
Tensor Algebra 29

Differentiating w.r.t. x i partially

(ain − bin )x n + (ami − bmi ) x m = 0


Differentiating again w.r.t. x j partially
(aij − bij ) + (a ji − b ji ) = 0
a ij + a ji = bij + b ji

or a mn + anm = bmn + bnm or a (mn ) = b( mn )

Also, since a mn and bmn are symmetric then a mn = a nm , bmn = bnm .


So,
2 amn = 2bmn
a mn = bmn

EXERCISES

1. Write down the law of transformation for the tensors


(i) Aij

(ii) Bkij
ijk
(iii) Clm

2. If Arpq and Bts are tensors then prove that Arpq Bts is also a tensor..

3. If Aij is a contravariant tensor and Bi is covariant vector then prove that Aij Bk is a tensor of rank
three and Aij Bj is a tensor of rank one.
4. If Ai is an arbitrary contravariant vector and Cij Ai Aj is an invariant show that Cij + Cji is a covariant
tensor of the second order.
5. Show that every tensor can be expressed in the terms of symmetric and skew-symmetric tensor.
n n
6. Prove that in n-dimensional space, symmetric and skew-symmetric tensor have (n + 1) and (n − 1)
2 2
independent components respectively.

7. If U ij ≠ 0 are components of a tensor of the type (0, 2) and if the equation fUij + gU ji = 0 holds

w.r.t to a basis then prove that either f = g and U ij is skew-symmetric or f = –g and U ij is symmetric.

8. If Aij is skew-symmetric then ( B j Bl + B l B j ) Aik = 0 .


i k i k

ij i
9. Explain the process of contraction of tensors. Show that aij a = δ j .
30 Tensors and Their Applications

10. If Arpq is a tensor of rank three. Show that Arpr is a contravariant tensor of rank one.

11. If a k λi µ j γ is a scalar or invariant, λi , µ j , γ are vectors then akij is a mixed tensor of type (2, 1).
ij k k

12. Show that if a hijkλ µ λ µ = 0 where λi and µ i are components of two arbitrary vectors then
h i h k

ahijk + ahkji + a jihk + a jkhi = 0

13. Prove that Aij Bi C j is invariant if Bi and C j are vector and Aij is tensor of rank two.
14. If A(r, s, t) be a function of the coordinates in n-dimensional space such that for an arbitrary vector
Br of the type indicated by the index a A(r, s, t)Br is equal to the component Cst of a contravariant
tensor of order two. Prove that A(r, s, t) are the components of a tensor of the form Arst .
15. If Aij and Aij are components of symmetric relative tensors of weight w. show that
w− 2 w+ 2
ij ij ∂x ∂x
A = A and Aij = Aij
∂x ∂x
16. Prove that the scalar product of a relative covariant vector of weight w and a relative contravariant
vector of weight w′ is a relative scalar of weight w + w ′ .
CHAPTER – 3

METRIC TENSOR AND RIEMANNIAN METRIC

3.1 THE METRIC TENSOR


In rectangular cartesian coordinates, the distance between two neighbouring point are (x, y, z) and
( x + dx, y + dy, z + dz ) is given by ds 2 = dx 2 + dy 2 + dz 2 .
In n-dimensional space, Riemann defined the distance ds between two neighbouring points x i
and x i + dx i (i = 1, 2,...n) by quadratic differential form

ds 2 = g11 (dx 1 ) 2 + g12 dx 1dx 2 + ⋅ ⋅ ⋅ + g1n dx1 dx n

+ g12 (dx 2 )dx1 + g 22 (dx 2 ) 2 + ⋅ ⋅ ⋅ + g 2 n dx 2 dx n


. . . . . . . . .. . . . . . . . . . . . . . . . .
+ +
+ g n1 dx n dx1 + g n 2 dx n dx 2 + ⋅ ⋅ ⋅ ⋅ + g nn (dx n )2
ds 2 = g ij dx i dx j (i, j = 1,2,...n) ...(1)
using summation convention.
Where g ij are the functions of the coordinates x i such that

g = g ij ≠ 0
The quadratic differential form (1) is called the Riemannian Metric or Metric or line element for n-
dimensional space and such n-dimensional space is called Riemannian space and denoted by Vn and
g ij is called Metric Tensor or Fundamental tensor..
The geometry based on Riemannian Metric is called the Riemannian Geometry.

THEOREM 3.1 The Metric tensor g ij is a covariant symmetry tensor of rank two.
Proof: The metric is given by
i j
ds 2 = g ij dx dx ...(1)
32 Tensors and Their Applications

Let x i be the coordinates in X-coordinate system and x i be the coordinates in Y-coordinate


system. Then metric ds2 = gij dxidxj transforms to ds 2 = g ij d x i dx j .
Since distance being scalar quantity.

ds 2 = g ij dx dx = gij dx dx
i j i j
So, ...(2)
The theorem will be proved in three steps.
(i) To show that dxi is a contravariant vector.
If x i = x i (x1, x2 ,...x n )

∂x i 1 ∂x i 2 ∂x i n
dxi = dx + dx + ⋅ ⋅ ⋅ + dx
∂x i ∂x 2 ∂x n
∂x i k
dxi = dx
∂x k
It is law of transformation of contravariant vector. So, dx i is contravariant vector..
(ii) To show that g ij is a covariant tensor of rank two. Since
∂x i k ∂x j l
dxi = dx and d x j
= ∂x
∂x k ∂x l
from equation (2)
∂x i ∂x j l
g ij dx i dx j = g ij dxk dx
∂x k ∂x l
∂x i ∂x j k l
g ij dx i dx j = g ij k ∂x dx
∂x ∂x l
∂x i ∂x j k l
k l
g kl dx dx = ij kg dx dx
∂x ∂x l
Since g ij dx i dx j = g kl dx k dx l (i, j are dummy indices).

  k l
 g kl − g ij ∂x ∂x
i j
dx dx = 0
 ∂x k ∂x l 
 

∂x i ∂x j
or g kl − gij = 0 as dx k and dx l are arbitrary..
∂x k ∂x l
∂x j ∂x j
g kl = g ij
∂x k ∂x l
∂x k ∂x l
or g ij = g kl
∂x i ∂x j

So, g ij is covariant tensor of rank two.


Metric Tensor and Riemannian Metric 33

(iii) To show that g ij is symmetric. Then g ij can be written as


1 1
g ij = (g + g ji ) + ( gij − g ji )
2 ij 2
g ij = Aij + Bij
1 g +g
where Aij = ( ij ji ) = symmetric
2
1
Bij = (g ij − g ji ) = Skew-symmetric
2
Now, g ij dx i dx j = ( Aij + Bij )dxi dx j from (3)

( gij − Aij )dx i dx j = Bij dxi dx j (4)

Interchanging the dummy indices in Bij dx i dx j , we have

Bij dxi dx j = B ji dx i dx j
Bij dxi dx j = − B ij dx i dx j

Since Bij is Skew-symmetric i.e., Bij = −B ji

Bij dx i dx j + Bij dx i dx j = 0

2 Bij dx i dx j = 0

⇒ Bij dxi dx j = 0
So, from (4),

( gij − Aij )dx i dx j = 0

⇒ g ij = Aij as dx i , dx j are arbitrary..

So, g ij is symmetric since Aij is symmetric. Hence g ij is a covariant symmetric tensor of rank
two. This is called fundamental Covariant Tensor.

THEOREM 3.2 To show that g ij dx i dx j is an invariant.

Proof: Let x i be coordinates of a point in X-coordinate system and x i be coordinates of a same point
in Y-coordinate system.
Since g ij is a Covariant tensor of rank two.

∂x k ∂x 1
Then, g ij = g kl
∂x i ∂x j
34 Tensors and Their Applications

∂x k ∂x l
⇒ g ij − g kl =0
∂x i ∂x j
  i j
 gij − g kl ∂x ∂x
k l
dx dx = 0
 ∂x i ∂x j 
 

i j ∂x k ∂x l i j
( g ij dx dx ) = g kl dx dx
∂x i ∂x j
∂x k i ∂x l
= g kl dx dx j
∂x i ∂x j

g ij dx i dx j = g kl dx k dx l

So, g ij dx i dx j is an ivariant.

3.2 CONJUGATE METRIC TENSOR: (CONTRAVARIANT TENSOR)

The conjugate Metric Tensor to g ij , which is written as g ij , is defined by


Bij
g ij = g (by Art.2.16, Chapter 2)

where Bij is the cofactor of g ij in the determinant g = g ij ≠ 0 .


By theorem on page 26

Aij Akj = δik

So, g ij g kj = δik
Note (i) Tensors g ij and g ij are Metric Tensor or Fundamental Tensors.
(ii) g ij is called first fundamental Tensor and g ij second fundamental Tensors.

EXAMPLE 1
Find the Metric and component of first and second fundamental tensor is cylindrical coordinates.
Solution
Let (x1, x2, x3) be the Cartesian coordinates and (x1, x2 , x3 ) be the cylindrical coordinates of a
point. The cylindrical coordinates are given by
x = r cosθ , y = r sin θ, z = z
So that
x1 = x, x 2 = y, x 3 = z and x 1 = r , x 2 = θ, x 3 = z ...(1)

Let g ij and g ij be the metric tensors in Cartesian coordinates and cylindrical coordinates
respectively.
Metric Tensor and Riemannian Metric 35

The metric in Cartesian coordinate is given by

ds 2 = dx + dy + dz
2 2 2

ds 2 = (dx ) + (dx ) +(dx )


12 2 2 3 2 ...(2)
i j
But ds 2 = g ij dx dx

= g11(dx1 )2 + g12 dx1dx2 + g13 dx 1 dx 3 + g 21 dx 2 dx 1

+ g22(dx2)2 + g 23 dx 2 dx 3 + g 31dx 3 dx 1

+ g 32 dx 3 dx 2 + g33(dx3)3 ...(3)
Comparing (2) and (3), we have
g11 = g 22 = g 33 = 1 and g12 = g13 = g 21 = g 23 = g31 = g 32 = 0
On transformation
∂x i ∂x j
g ij = g ij , since g ij is Covariant Tensor of rank two. (i, j = 1, 2, 3)
∂x i ∂x j
for i = j = 1.
2 2
 ∂x1   ∂x 2   ∂x 3 
g11 = g11 1  + g 22  1  + g 33  1 
 ∂x   ∂x   ∂x 
since g12 = g13 = ⋅ ⋅ ⋅ = g32 = 0 .
2 2 2
 ∂x   ∂y   ∂z 
g11 = g11  + g 22   + g 33  
 ∂r   ∂r   ∂r 
Since x = r cos θ, y = r sin θ, z = z
∂x ∂y ∂z
= cosθ, = sin θ, =0
∂r ∂r ∂r
and g11 = g 22 = g 33 = 1.
g11 = cos 2 θ + sin 2 θ + 0
g11 = 1
Put i = j = 2.
2 2 2
 ∂x 1   2   3
g 22 = g11  2  + g 22  ∂x  + g 33  ∂x 
  ∂x 2   ∂x 2 
 ∂x     

 ∂x 
2
 ∂y 
2
 ∂z 
2
g 22 = g11   + g 22   + g 33  
 ∂θ   ∂θ   ∂θ 
36 Tensors and Their Applications

Since g11 = g 22 = g 33 = 1
∂x ∂y ∂z
= −r sin θ, = r cosθ, =0
∂θ ∂θ ∂θ
g 22 = (− r sin θ )2 + (r cos θ)2 + 0

= r 2 sin 2 θ + r 2 cos 2 θ
g 22 = r 2
Put i = j = 3.
2
 ∂x1   ∂x 2   ∂x 3 
g 
= 11  3   + g 
22 
 + g  
g 33 3 33  3
 ∂x   ∂x   ∂x 
 ∂x 
2
 ∂y   ∂z 
= g11   + g 22   + g 33  
 ∂z   ∂z   ∂z 
∂x ∂y ∂z
Since = 0, = 0, = 1 . So, g 33 = 1 .
∂z ∂z ∂z
So, g11 = 1, g 22 = r 2 , g 33 = 1

and g12 = g13 = g 21 = g 23 = g31 = g 32 = 0


(i) The metric in cylindrical coordinates
i j
ds 2 = g ij dx dx i, j = 1, 2,3.

ds 2 = g11 dx ( )
1 2
( )
+ g 22 dx 2
2
( )
+ g 33 dx 3
2

since g12 = g13 = ⋅ ⋅ ⋅ = g32 = 0

ds 2 = dr + r (dθ ) + dφ
2 2 2 2

(ii) The first fundamental tensor is

 g11 g12 g13  1 0 0 


   2 
g ij =  g 21 g 22 g 23  = 0 r 0
 g 31 g32 g 33  0 0 1 
 

1 0 0
since g = g ij = 0 r2 0
0 0 1
2
g= r
Metric Tensor and Riemannian Metric 37

(iii) The cofactor of g are given by


B11 = r 2 , B22 = 1, B33 = r 2
and B12 = B21 = B13 = B 23 = B31 = B32 = 0
Bij
The second fundamental tensor or conjugate tensor is g =
ij
.
g

cofactor of g11 in g
g 11 =
g

B11 r 2
g 11 = = 2 =1
g r

B12 1
g 22 = = 2
g r

B33 r 2
g 33 = = 2 =1
g r
and g 12 = g13 = g 21 = g 23 = g 31 = g 32 = 0

1 0 0
 1 
Hence the second fundamental tensor in matrix form is 0 0 .
 r2 
0 0 1

EXAMPLE 2
Find the matrix and component of first and second fundamental tensors in spherical coordinates.
Solution

Let ( x1 , x 2 , x 3 ) be the cartesian coordinates and ( x 1 , x 2 , x 3 ) be the spherical coordinates of a


point. The spherical coordinates are given by
x = r sin θ cos φ, y = r sin θ sin φ, z = r cos θ

So that x1 = x, x 2 = y , x 3 = z and x 1 = r , x 2 = θ , x 3 = φ

Let g ij and g ij be the metric tensors in cartesian and spherical coordinates respectively..
The metric in cartesian coordinates is given by
ds 2 = dx + dy + dz
2 2 2

ds 2 = dx1( ) + (dx ) + (dx )


2 2 2 3 2

ds 2 = g ij dx dx ; (i, j = 1, 2, 3)
i j
But
38 Tensors and Their Applications

⇒ g11 = g 22 = g 33 = 1 and g12 = g 23 = g13 = g 21 = g31 = g 32 = 0


On transformation

∂x i ∂x j
g ij = g ij
∂x i ∂x j
(since g ij is covariant tensor of rank two) (where i, j = 1,2,3).

∂x1 ∂x 1 ∂x 2 ∂x 2 ∂x 3 ∂x 3
g ij = g11 + g + g
∂x i ∂x j
22
∂x i ∂x j
33
∂x i ∂x 1
since i, j are dummy indices.
Put i = j = 1
2 2 2
 ∂x 1   ∂x 2   ∂x 3 
g 
= 11  1   + g 
22 
 + g  
g11 1 33  
 ∂x   ∂x   ∂x
1

 ∂x 
2
 ∂y 
2
 ∂z 
2

g11 = g11   + g 22   + g 33  
 ∂r   ∂r   ∂r 
Since x = r sin θ cos φ, y = r sin θ sin φ , z = r cos θ
∂x ∂y ∂z
= sin θ cos φ, = sin θ sin φ, = cos θ
∂r ∂r ∂r

and g11 = g 22 = g 33 = 1 .
So,
g11 = (sin θ cos φ )2 + (sin θ sin φ )2 + cos2 θ

g11 = 1
put i = j = 2
2 2
 ∂x 1   2   3 
= g11  2  + g 22  ∂x  + g 33  ∂x 
g 22   ∂x 2   ∂x 2 
 ∂x     

 ∂x 
2
 ∂y 
2
 ∂z 
2

g 22 = g11   + g 22   + g33  
 ∂θ   ∂θ   ∂θ 
since g11 = g 22 = g 33 = 1

∂x ∂y ∂z
= r cos θ cos φ, = r cos θ sin φ, = −r sin θ
∂θ ∂θ ∂θ

g 22 = (r cos θ cos φ )2 + (r cos θ sin φ)2 + (− r sin θ)2

g 22 = r 2
Metric Tensor and Riemannian Metric 39

Put i = j = 3
2 2 2
 ∂x1   ∂x 2   ∂x 3 
g 
= 11  3   + g 
22 
 + g  
g 33 3 33  
 ∂x   ∂x   ∂x
3

2 2 2
 ∂x   ∂y   ∂z 
g 33 = g11   + g 22   + g33  
 ∂φ   ∂φ   ∂φ 

since g11 = g 22 = g 33 = 1
∂x ∂y ∂z
and = −r sin θ sin φ, = r sin θ cos φ, =0
∂φ ∂φ ∂φ
g 33 = (− r sin θ sin φ)2 + (r sin θ cos φ )2 + 0

g 33 = r 2 sin 2 θ
So, we have
g11 = 1, g 22 = r 2 , g 33 = r 2 sin 2 θ

and g12 = g13 = g 21 = g 23 = g31 = g 32 = 0


(i) The Metric in spherical coordinates is

ds 2 = g ij dx dx ; i, j = 1, 2, 3
i j

ds 2 = g11 dx ( )
1 2
( )
+ g 22 dx 2
2
( )
+ g 33 dx 3
2

ds 2 = dr 2 + r 2 dθ 2 + r 2 sin 2 θdφ 2
(ii) The Metric tensor or first fundamental tensor is

 g11 g12 g13  1 0 0 


 g g g  = 0 r 2 
g ij =  21 22 23   0 
 g 31 g32 g 33  0 0 r 2 sin 2 θ
 
and

1 0 0
g = g ij =0 r2 0 = r 4 sin 2 θ
0 0 r 2 sin 2 θ
(iii) The cofactor of g are given by B11 = 1, B22 = r 2 , B33 = r 2 sin 2 θ and B12 = B21 =
B31 = B13 = B23 = B32 = 0

Bij
The second fundamental tensor or conjugate tensor is g ij = .
g
40 Tensors and Their Applications

cofactor of g11 in g B11


g 11 = =
g g
r 4 sin 2 θ
=
r 4 sin 2 θ
g 11 =1
B 22 r 2 sin 2 θ
g 22
= = 4 2
g r sin θ
1
g 22 =
r2
B33 r2
g 33
= =
g r 4 sin 2 θ
1
g 33 =
r sin 2 θ
2

and g = g = g = g = g = 0
12 13 21 31 32

Hence the fundamental tensor in matrix form is

 
0 0 0 
g 11
g 12
g 
13
 1 
 21 23  0 0
g ij
= g g 22 g  =  r2 
 g 31  
 g 32 g 33  0 0
1 
 r 2 sin 2 θ 
 
EXAMPLE 3
If the metric is given by

( ) ( )
ds 2 = 5 dx 1 + 3 dx 2
2 2
( )
+ 4 dx 3
2
− 6dx 1dx 2 + 4dx 2 dx 3

Evaluate (i) g and (ii) g ij .

Solution
i j
The metric is ds 2 = g ij dx dx ; (i, j = 1, 2, 3)

ds 2 = g11 dx
1
( ) 2
+ g12 dx 1dx 2 + g13 dx1 dx 3 + g 21dx 2 dx1

+ g 22 (dx 2 ) 2 + g 23 dx 2 dx 3 + g31 dx 3 dx1 + g 32 dx 3 dx 2 + g 33 (dx 3 )2

Since g ij is symmetric ⇒ g ij = g ji

i.e., g12 = g 21, g 23 = g32 , g13 = g 31


Metric Tensor and Riemannian Metric 41

ds 2 = g11 (dx ) + g 22 (dx ) + g33 (dx ) + 2 g12 dx dx


1 2 2 2 3 2 1 2
So,

+ 2 g 23 dx 2 dx 3 + 2 g13 dx1 dx 3 ...(1)


Now, the given metric is

ds 2 = 5(dx ) + 3(dx ) + 4(dx ) − 6dx dx + 4dx dx


1 2 2 2 3 2 1 2 2 3
...(2)
Comparing (1) and (2) we have
g11 = 5, g 22 = 3, g 33 = 4, 2g12 = −6 ⇒ g12 = −3 = g 21
2 g 23 = 4 ⇒ g 23 = 2 = g 32 , g13 = 0 = g 31

g11 g12 g13 5 −3 0


g = g ij = g 21 g 22 g 23 = − 3 3 2 = 4
g 31 g 32 g33 0 2 4

(ii) Let Bij be the cofactor of g ij in g.


Then
3 2
B11 = Cofactor of g11 = =8
2 4
5 0
B22 = Cofactor of g 22 = = 20
04

5 −3
B33 = Cofactor of g 33 = =6
−3 3
−3 2
B12 = Cofactor of g12 = − = 12 = B 21
0 4

−3 3
B13 = Cofactor of g13 = = −6 = B31
0 2
5 −3
B 23 = Cofactor of g 23 = – = −10 = B 32
0 2

Bij
Since g ij =
g
We have
B 8
g 11 = 11 = = 2; g 22 = 5, g 33 = 3 , g 12 = g 21 = 3, g 13 = g 31 = − 3 , g 23 = g 32 = − 5
g 4 2 2 2
42 Tensors and Their Applications

Hence,

 3
 2 3 − 
2

=  5 − 
5
g ij 3
2
 
− 3 5 3

 2 2 2 
3.3 LENGTH OF A CURVE
Consider a continuous curve in a Riemannian Vn i.e., a curve such that the coordinate x i of any
current point on it are expressible as functions of some parameter, say t.
The equation of such curve can be expressed as
x i = xi (t)
The length ds of the arc between the points whose coordinate s are x i and x i + dx i given by
i j
ds 2 = g ij dx dx

If s be arc length of the curve between the points P1 and P2 on the curve which correspond to
the two values t1 and t 2 of the parameter t.
12
t2  i j 
 gij dx dx  dt
∫ ∫
P2
ds =  dt dt 
s =

P1 t1

NULL CURVE
dx i dx j
If g ij = 0 along a curve. Then s = 0. Then the points P1 and P2 are at zero distance, despite
dt dt
of the fact that they are not coincident. Such a curve is called minimal curve or null curve.

EXAMPLE 4
A curve is in spherical coordinate xi is given by
−1  1 
x1 = t, x = sin   and x = 2 t − 1
2 3 2

 
t
Find length of arc 1 ≤ t ≤ 2.
Solution
In spherical coordinate, the metric is given by
ds 2 = (dx ) + ( x ) (dx ) + ( x sin x ) (dx )
1 2 1 2 2 2 1 2 2 3 2
Metric Tensor and Riemannian Metric 43

−1 1
x1 = t, x = sin , x3 = 2 t 2 − 1
2
given
t

dx 1 = dt , dx 2 =
1  1 1
(
 − 2  dt , dx 3 = 2 ⋅ t 2 − 1
−1 2
)2t dt
 1  t 
2 2
1− 
t

dt 2t
dx 2 = − , dx 3 = dt
t t −1 2
t 2 −1

2 2
 dt     2t 
2
−1 1  
ds = (dt )
2
+t − + t sin sin dt 
2
 t t 2 − 1   
2  
  t   t 2 − 1 

dt 2 4t 2
ds =2dt 2
+ + (dt )2
t −1 t −1
2 2

5t 2 2
ds 2 = t 2 − 1 dt

t
ds = 5 dt
t −12

Now, the length of arc, 1 ≤ t ≤ 2, is


2
t 5  t2 −1

2


t2
ds = 5 dt =   =
2  12  15 units
t1 1
t2 −1  1

3.4 ASSOCIATED TENSOR


i i ... i
A tensor obtained by the process of inner product of any tensor A j11 2j 2 ... rj s with either of the fundamental
tensor g ij or g ij is called associated tensor of given tensor..
e.g. Consider a tensor Aijk and form the following inner product

g αi Aijk = Aαjk ; g αj Aijk = Aikα ; g αk Aijk = Aijα

All these tensors are called Associated tensor of Aijk .


Associated Vector
Consider a covariant vector Ai . Then g ik Ai = A k is called associated vector of Ai . Consider a
contravariant vector A j . Then g jk A j = Ak is called associated vector of A j .
44 Tensors and Their Applications

3.5 MAGNITUDE OF VECTOR


i
The magnitude or length A of contravariant vector A . Then A is defined by

A= g ij Ai A j
i j
or A 2 = g ij A A
Also, A2 = A j A j as g ij A = A j
i

i.e., square of the magnitude is equal to scalar product of the vector and its associate.
The magnitude or length A of covariant vector Ai . Then A is defined by

A= g ij Ai A j

or A 2 = g ij Ai A j
A vector of magnitude one is called Unit vector. A vector of magnitude zero is called zero vector
or Null vector.

3.6 rSCALARr PRODUCT OF TWO VECTORS r r


Let A and B be two vectors. Their scalar product is written as A ⋅ B and defined by
r r
A ⋅ B = Ai Bi
r r
A ⋅ B = A B i = g ij A B since Bi = g ij B
i i j j
Also,
r r
A ⋅ B = Ai B = g Ai B j since B = g B j
i ij i ij

Thus
r r
A ⋅ A = A Ai = g ij A A = A
i i j 2

r
i.e., A = A = g ij Ai A j
Angle
r between two vectors
r
Let A and B be two vectors. Then
r r r r
A ⋅ B = A B cos θ
r r
A⋅B g ij A i B j
⇒ cos θ = r r =
AB g ij A i A j g ij B i B j
r r
since A = g ij Ai A j ; B = g ij B i B j
This is required formula for cos θ .
Definition
r r
The inner product of two contravariant vectors A (or A i ) and B (or Bi ) associated with a symmetric
tensor g ij is defined as g ij A i B j . It is denoted by
r r i j
g ( A , B ) = g ij A B
Metric Tensor and Riemannian Metric 45

r r
THEOREM 3.3 The necessary and sufficient condition that the two vectors A and B at 0 be orthogonal
r r
if g ( A, B ) = 0
r r
Proof: Let θ be angle between the vectors A and B then
r r r r
A ⋅ B = A B cos θ
r r
or A ⋅ B = AB cos θ
g ij A i B j = AB cos θ
g ij Ai B j
⇒ cos θ = ...(1)
AB
r r
If A and B are orthogonal then θ = π ⇒ cos θ = 0 then from (1)
2
i j
g ij A B = 0
r r r r
⇒ ( ) ( ) i j
g A, B = 0 since g A, B = g ij A B
Conversely if g ij A B = 0 then from (1)
i j

cos θ = 0 ⇒ θ = π .
r r 2
So, two vectors A & B are orthogonal. Proved.
r r r r
Note: (i) If A and B be unit vectors. Then A = B = 1. Then
r r
cos θ = A ⋅ B = gij A i B j
r r π π
(ii) Two vectors A and B will be orthogonal if angle between them is i.e., θ = then
2 2
π
cos θ = cos θ = = 0
2

3.7 ANGLE BETWEEN TWO VECTORS

THEOREM 3.4 To show that the definition of the angle between two vectors is consistent with the
requirement cos 2θ ≤ 1.
OR
To justify the definition of the angle between two vectors.
OR
To show that the angle between the contravariant vectors is real when the Riemannian Metric is
positive definition.
r r
Proof: Let θ be the angle between unit vectors A and B then
cos θ = g ij A B = A j B = A j B Bi = g A j Bi = A Bi
i j j ij ij i
46 Tensors and Their Applications

To show that θ is real i.e., |cosθ| ≤ 1.


Consider the vector lAi + mB i when l and m are scalars. The square of the magnitude of

lAi + mB i = g ij (lA + mB ) (lA + mB )


i i j j

= g ij l A A + g ij lmA B + m lg ij B A + m g ij B B
2 i j i j i j 2 i j

= l 2 + 2lm cos θ + m 2
Since
g ij A i A j = A 2 = 1; g ij B B = B = 1.
i j 2

and
r r r
g ij A i B j = cosθ; as A & B are unit vector i.e., A = 1 ⇒ A 2 = 1 .
Since square of magnitude of any vector ≥ 0.
So, the square of the magnitude of lAi + mB i ≥ 0.
or l 2 + 2lm cos θ + m 2 ≥ 0
(l + m cos θ )2 + m 2 − m 2 cos 2 θ ≥0
(l + m cosθ) 2 + m 2 (1 − cos2 θ) ≥ 0
This inequality holds for the real values of l & m.
if 1 − cos2 θ ≥ 0
⇒ cos 2 θ ≤ 1
cos θ ≤ 1
Proved.
THEOREM 3.5 The magnitude of two associated vectors are equal.
Proof: Let A and B be magnitudes of associate vectors A i and A i respectively. Then
i j
A 2 = g ij A A ...(1)
and
ij
B 2 = g Ai A j ...(2)
From equation (1)
A 2 = (gij Ai ) A j
A2 = Aj A j ...(3)

since g ij A i = A j (Associate vector)


From equation (2)
ij
B 2 = (g Ai )A j
j
B2 = A Aj ...(4)
Metric Tensor and Riemannian Metric 47

since g ij Ai = A j
from (3) and (4)
2
A2 = B
⇒ A=B
i
So, magnitude of Ai and A are equal.

3.8 ANGLE BETWEEN TWO COORDINATE CURVES


Let a Vn referred to coordinate x i , (i = 1, 2, ... n) . For a coordinate curve of parameter x l, the coordinate
x l alone varies. Thus the coordinate curve of parameter x l is defined as
x i = c i , ∀i except i = l ...(1)
i,
where C s are constants.
Differentiating it, we get
dx i = 0, ∀i , except i = l and dx l ≠ 0
Let Ai and B i be the tangent vectors to a coordinate curve of parameters x p and x q respectively..
Then
Ai = dx i = (0,...0, x p , 0...0) ...(2)

B i = dx i = (0,...0, x q , 0...0) ...(3)


If θ is required angle then

g ij A i B j
cos θ =
g ij A i A j g ij B i B j

g pq A p B q
=
g pp A p A p g qq B q B q

g pq A p B q
=
g pp g qq A p B q

g pq
cos θ = ...(4)
g pp g qq

which is required formula for θ .


The angle wij between the coordinate curves of parameters x i and x j is given by

g ij
cos wij =
g ii g jj
48 Tensors and Their Applications

If these curves are orthogonal then


π
cos wij = cos =0
2
⇒ g ij = 0

Hence the x i coordinate curve and x j coordinate curve are orthogonal if g ij = 0 .

3.9 HYPERSURFACE
The n equations xi = xi (u1) represent a subspace of Vn . If we eliminate the parameter u1, we get
(n –1) equations in x j, s which represent one dimensional curve.
Similarly the n equations xi = xi (u1,u2) represent two dimensional subspace of V n. If we eliminating
the parameters u1, u2, we get n –2 equations in xi,s which represent two dimensional curve V n. This
two dimensional curve define a subspace, denoted by V 2 of V n.
Then n equations xi = xi (u1, u2, ... un–1) represent n – 1 dimensional subspace V n–1 of V n. If we
eliminating the parameters u1, u2, ...un–1, we get only one equation in x i, s which represent n –1
dimensional curve in V n. This particular curve is called hypersurface of V n.
Let φ be a scalar function of coordinates x i . Then φ (x ) = constant determines a family of
i

hypersurface of V n.

3.10 ANGLE BETWEEN TWO COORDINATE HYPERSURFACE


Let
φ (x i ) = constant ...(1)

and ψ( x i ) = constant ...(2)


represents two families of hypersurfaces.
Differentiating equation (1), we get
∂φ i
dx = 0 ...(3)
∂x i
∂φ ∂φ
This shows that is orthogonal to dx i . Hence is normal to φ = constant, since dx i is
∂x i
∂x i
tangential to hypersurface (1).
∂ψ
Similarly is normal to the hypersurface (2). If ω is the angle between the hypersurface (1)
∂x i
and (2) then ω is also defined as the angle between their respective normals. Hence required angle ω
is given by

∂φ ∂ψ
g ij
∂x i ∂x j
cos ω = ...(4)
∂φ ∂φ ∂ψ ∂ψ
g ij i g ij i
∂x ∂x ∂x ∂x j
j
Metric Tensor and Riemannian Metric 49

If we take
φ = x p = constant ...(5)
and ψ = x q = constant ...(6)
The angle ω between (5) and (6) is given by

∂x p ∂x q
g ij
∂x i ∂x j
cos ω =
∂x p ∂x p ij ∂x ∂x
q q
g ij i g
∂x ∂x j ∂x i ∂x j

g ij δip δ qj
=
g ij δip δ qj g ij δqi δ qj

g pq
cos ω = ...(7)
g pp g qq

The angle ωij between the coordinate hypersurfaces of parameters x i and x j is given by

g ij
cos ωij = ...(8)
g ii g jj

If the coordinate hypersurfaces of parameters x i and x j are orthogonal then


π
ωij =
2
⇒ cos ωij = 0
from (8), we have g = 0 .
ij

Hence the coordinate hypersurfaces of parameters x i and x j are orthogonal if g ij = 0 .

3.11 n-PLY ORTHOGONAL SYSTEM OF HYPERSURFACES


If in a Vn there are n families of hypersurfaces such that, at every point, each hypersurface is orthogonal
to the n − 1 hypersurface of the other families which pass through that point, they are said to form as
n-ply orthogonal system of hypersurfaces.

3.12 CONGRUENCE OF CURVES


A family of curves one of which passes through each point of Vn is called a congruence of curves.

3.13 ORTHOGONAL ENNUPLE


An orthogonal ennuple in a Riemannian Vn consists of n mutually orthogonal congruence of curves.
50 Tensors and Their Applications

THEOREM 3.6 To find the fundamental tensors g ij and g ij in terms of the components of the unit
tangent e h ( h = 1, 2,... n ) to an orthogonal ennuple.
Proof: Consider n unit tangents e ih (h = 1, 2,...n) to conguence e h ( h = 1, 2,... n ) of an orthogonal ennuple
in a Riemannian Vn . The subscript h followed by an upright bar simply distinguishes one congruence
from other. It does not denote tensor suffix.
The contravariant and covariant components of eh | are denoted by eh | and eh |i respectively..
Suppose any two congruences of orthogonal ennuple are eh | and ek | so that

g ij ehi |ekj| = δ hk ...(1)

ehi |ek |i = δ hk
from (1),
g ij ehi | ekj| = 0

and g ij ehi | ehj| = 1


We define

cofactorof eh |i in determinant eh |i
ehi | =
e h|i
Also, from the determinant property, we get
n

∑e e
h =1
i
h| h | j = δij ...(2)

Multiplying by e jk
n

∑e e
h =1
i
h| h| j g j k = δ ij g jk

or ∑e e
h =1
i k
h | h| = g ik ...(3)

Again multiplying (2) by g ik .


n

∑e e
h =1
i
h | h | j g ik = δ j g ik
i

or g jk = ∑e h |k eh | j ...(4)
from (3) and (4)
n
g ij = ∑e
h =1
h |i e h| j ...(5)
Metric Tensor and Riemannian Metric 51

n
g ij = ∑e
h =1
i
h| e hj| ...(6)

This is the required results.


Corollary: To find the magnitude of any vector u is zero if the projections of u on eh | are all zero.
Proof: Let
n
i
u = ∑C e
h =1
i
h h| ...(7)
Then
n n
u i ek |i = ∑C e e
h =1
i
h h | k |i = ∑C δ
h =1
h
h
k = Ck

or C k = u i ek |i ...(8)

i.e., C k = projection of u i on ek |i
Using (8), equation (7) becomes
n
ui = ∑u e
h =1
j i
h| j ek |

Now,

  
u 2 = u iu i = 
 ∑ C e   ∑ C e
i
h h| k k |i
 from (7)

 h  k 

= ∑C C e e
h ,k
h
i
k h | k| i

= ∑C C δ
h ,k
h k
h
k

= ∑C C
h
h h

u2 = ∑ (C
h =1
h )2

This implies that u = 0 iff u 2 = 0 iff C h = 0 .


Hence the magnitude of a vector u is zero iff all the projections of u (i.e. of ui) on n mutually
orthogonal directions ehi | are zero.
52 Tensors and Their Applications

Miscellaneous Examples
1. If p and q are orthogonal unit vectors, show that
( g hj g ik − g hk gij ) p h q i p j q k = 1
Solution
Since p and q are orthogonal unit vectors. Then
g ij pi q j = 0, p 2 = q 2 = 1 .
Now,
( g hj g ik − g hk gij ) p h q i p j q k = g hj g ik p p q q − g hk g ij p q q p
h j i k h k i j

= ( g hi p h p j ) ( gik q i q k ) − (g hk p j q k ) ( gij q i p j )
= p2.q2 – 0.0
=1.1
= 1 (since g hi p h p j = 1 & g hk p h q k = 0 )
2. If θ is the inclination of two vectors A and B show that

(g hi g ik − g hk g ij ) A h A j B i B k
sin θ =
2
g hj gik A h A j B j B k

Solution
If θ be the angle between the vectors A and B then

g ij A j B i
cos θ =
gij A i A j g ik B i B k
But sin 2 θ = 1 − cos 2 θ

( gij B i A j ) (g hk A h B k )
sin 2 θ = 1 −
( g hj A h A j ) ( g ik B i B k )

(g hj g ik − g hk g ij ) A h A j B i B k
=
g hj g ik A h A j B i B k

3. If X ij are components of a symmetric covariant tensor and u, v are unit orthogonal to w and
satisfying the relations
( X ij − αg ij )u i + γw j = 0

( X ij − βg ij ) v i + δw j = 0
where α ≠ β prove that u and v are orthogonal and that
Metric Tensor and Riemannian Metric 53

X ij u i v j = 0

Solution

Suppose X ij is a symmetric tensor. Since u i , v j are orthogonal to wi then


u i wi = 0 ...(1)
v i wi = 0 ...(2)

given ( X ij − αgij )u i + γw j = 0 ...(3)

( X ij − βgij )vi + δw j = 0 ...(4)


where α ≠ β.
Multiply (3) & (4) by v j , u j respectively and using (1) and (2), we have
( X ij − αg ij )u iv j = 0 ...(5)
( X ij − βg ij )v iu j = 0 ...(6)
Interchanging the suffixes i & j in the equation (6) and since g ij , X ij are symmetric, we get
( X ij − αg ij )u iv j = 0 ... (7)
Subtract (6) & (7) we get
(β − α)g iju i v j = 0
Since β ≠ α and β − α ≠ 0.
Hence,
g ij u i v j = 0 ...(8)
So, u and v are orthogonal.
Using (8) in equation (5) & (6), we get
X ij u i v i = 0 Proved.
4. Prove the invariance of the expression g dx1 dx 2 ...dx n for the element volume.
Solution

Since g ij is a symmetric tensor of rank two. Then


∂x k ∂x l
g ij = g kl
∂x i ∂x j
Taking determinant of both sides
∂x k ∂x l
g ij = g kl
∂x i ∂x j
∂x
Since = J (Jacobian)
∂x
g kl = g & g ij = g
54 Tensors and Their Applications

So,
g = gJ 2
or
g
J= g
Now, the transformation of coordinates from x l to x i , we get
∂x 1 2 n
dx 1 dx 2 ... dx n = ∂x d x d x ... dx

= Jd x 1d x 2 ... d x n
g 1 2
dx 1dx 2 ... dx n = d x d x ... d x n
g

g dx1 dx2 ... dx n = g d x1d x 2... d x n

So, the volume element dv = g dx1 dx 2 ... dx n is invariant.

EXERCISES

1. For the Metric tensor g ij defined g kl and prove that it is a contravariant tensor.
2. Calculate the quantities g i j for a V3 whose fundamental form in coordinates u, v, w, is
2 2 2
adu + bdv + cdw + 2 fdvdw + 2 gdwdu + 2 hdudv
3. Show that for an orthogonal coordinate system
1 1 1
g 11 = , g 22 = , g 33 =
g11 g 22 g 33
4. For a V2 in which g11 = E , g12 = F , g 21 = G prove that

g = EG − F 2 , g 11 = G g , g 12 = − F g , g 22 = E g
1
5. Prove that the number of independent components of the metric g ij cannot exceed n( n + 1) .
2
6. If vectors u i , vi are defined by u i = g ij u j , v i = g ij v j show that u i = g ij u j , u i vi = u i v i and u i g ij u j = u i g ij u j
7. Define magnitude of a unit vector. prove that the relation of a vector and its associate vector is
reciprocal.

8. If θ is the angle between the two vectors Ai and B i at a point, prove that

(g hi gik − g hk gij ) A h Ai B j B k
2
sin θ = ghi g jk Ah Ai B j B k

9. Show that the angle between two contravariant vectors is real when the Riemannian metric is positive
definite.
CHAPTER – 4

CHRISTOFFEL'S SYMBOLS AND COVARIANT


DIFFERENTIATION

4.1 CHRISTOFFEL'S SYMBOLS


The German Mathematician Elwin Bruno Christoffel defined symbols
1  ∂gi k ∂g j k ∂g i j 
[ij, k ] =

2  ∂x j
+
∂x i
− k  , (i, j, k = 1, 2,...n )
∂x 
...(1)
called Christoffel 3-index symbols of the first kind.

k
and   = g k l [ij, l ] ...(2)
i j 
called Christoffel 3-index symbols of second kind, where g i j are the components of the metric Tensor
or fundamental Tensor.
There are n distinct Christoffel symbols of each kind for each independent g i j . Since g i j is
1
symmetric tensor of rank two and has n (n + 1) independent components. So, the number of
2
independent components of Christoffel’s symbols are n ⋅ n(n + 1) = n 2 (n + 1) .
1 1
2 2
k
THEOREM 4.1 The Christoffel's symbols [ij, k ] and   are symmetric with respect to the indices i
i j 
and j.
Proof: By Christoffel’s symbols of first kind
1  ∂g ik ∂g jk ∂g ij 
[ij, k ] =  j +
2  ∂x
∂ − k 
∂x i
∂x 
Interchanging i and j, we get
1  ∂g jk ∂gik ∂g ji 
[ ji, k ] =  + − k 
2  ∂x i ∂x j ∂x 
56 Tensors and Their Applications

1  ∂g i k ∂g j k ∂g i j 
 + − k  since g ij = g ji
2  ∂x j
=
∂x i ∂x 

[ ji, k ] = [ ij , k ]
Also, by Christoffel symbol of second kind
 k
  = g k l [ij, l ]
i j
= g k l [ ji, l ] since [ij , l ] = [ ji, l ]

k k
  =  j i
i j   
Proved.
THEOREM 4.2 To prove that
 k 
(i) [ij, m] = g km  
i j 
∂g ij
(ii) [ik , j ] + [ jk , i] =
∂x k
∂g ij  i   j 
(iii) = − g jl   − g im  
∂x k
 
l k m k 
Proof: (i) By Christoffel’s symbol of second kind
k
  = g k l [ij, l ]
i j 
Multiplying this equation by g k m , we get
 k 
g k m   = g k m g k l [ij, l ]
i j 

= δ lm [ij , l ] as g k m g k l = δ lm
k
g km   = [ij, m]
i j 
(ii) By Christoffel’s symbol of first kind
1  ∂g k j ∂g i j ∂g i k 
[ik , j ] =  + −
2  ∂x i ∂x k ∂x j 
 ...(1)

1  ∂g ki ∂g ji ∂g jk 
and [ jk ,i ] =  + −  ...(2)
2  ∂x j ∂x k ∂x i 
adding (1) and (2),
1  ∂ g i j ∂g ji 
[ik , j ] + [ jk , i ] =  +  g = g ji
2  ∂ x k ∂x k  since ij
Christoffel's Symbols and Covariant Differentiation 57

1 ∂g ij ∂g ij
[ik , j ] + [ jk , i] = ⋅2 =
2 ∂x k ∂x k

(iii) Since g ij g lj = δ li .

Differentiating it w.r.t. to x k , we get

∂glj ∂g ij
g ij + glj =0
∂x k ∂x k
Multiplying this equation by g lm , we get

∂g lj ∂g ij
g ij g lm + g lm g lj =0
∂x k ∂x k

∂g ij ∂g
g lm glj k =
− g ij g lm ljk
∂x ∂x

∂g ij ∂g lj
δ mj = − g ij lm
g { [lk , j ] + [ jk , l ] } since = [lk , j ] + [ jk , l ] .
∂x k ∂x k

∂g im { } {
= − g g [lk , j ] − g g [ jk , l ]
lm ij ij lm
}
∂x k

∂g im  i   m
= − g lm   − g ij  
∂x k
l k   j k
Interchanging m and j, we get

∂g ij  i 
lj 
 j 
im 
= − g   − g  
∂x k  l k   m k 

∂g ij  i   j 
or k =
− g ij   − g im   as g lj = g jl
∂x l k  m k 
Proved.

 i  ∂ log g
THEOREM 4.3 To show that i j  =
( )
  ∂x j
Proof: The matrix form of g ik is

 g11 g12 ... g1n 


g 
 21 g 22 ... G2 n 
g ik =
 M 
 
 g n1 g n 2 ... g nn 
58 Tensors and Their Applications

g11 g12 ... g1n


g 21 g 22 ... g 2 n
and g = gik =
M
g n1 g n 2 ... gnn
But g ik g = δlk
il

Take l = k.
g ik g ik = δ kk = 1

g ik = [g ik ]−1 =
Gik
⇒ (Theorem 2.13 , Pg 25)
g

where Gik is cofactor of g ik in the determinant g ik


⇒ g = g ik Gik ...(1)

Differentiating w.r.t. g ik partially


∂g ∂g ik
= Gik since =1
∂g ik ∂g ik
Now,
∂g ∂g ∂g ik
j =
∂x ∂g ik ∂x j
∂gik
= Gik
∂x j

But Gik = gg ik

∂g ik ∂gik
= gg
∂x j
∂x j

1 ∂g ∂g
j =
g ik ikj
g ∂x ∂x

1 ∂g ∂g ik
= g ik { [ jk , i] + [ij, k ] } as = [ jk , i] + [ij, k ]
g ∂x j ∂x j

= g [ jk , i] + g [ij , k ]
ik ik

 k  i
=  + 
 j k  i j 
Christoffel's Symbols and Covariant Differentiation 59

1 ∂g  i   i 
j =   +   as k is dummy indices
g ∂x  j i  j i

1 ∂g  i 
= 2 
g ∂x j
 j i
1 ∂g  i 
=  
2 g ∂x j
 j i

∂ log( g )  i 
=   Proved.
∂x j
 j i 

EXAMPLE 1

If g ij ≠ 0 show that

 β ∂  β 
  = ∂x j [ik , α] − i k 
g αβ
∂ ( [βj, α] + [αj , β] )
∂x j i k   
Solution
By Christoffel’s symbol of second kind
 β
  = g [ik , α]
βα

 
i k
Multiplying it by g αβ , we get

β 
g αβ   = g αβ g [ik , α ]
βα

 
i k
 β 
g αβ   = [ik , α ] as g αβ g = 1
βα

i k 

Differentiating it w.r.t. to x j partially

∂   β  ∂
j  αβ  
g = [ik , α]
∂x  i k  ∂x j

∂  β   β  ∂g α β ∂
g αβ  +   j =
[
ik,α ]
∂ x j i k   i k  ∂ x ∂x j
∂g αβ
since = [αj, β] + [βj, α]
∂x j
∂  β  β  ∂
g αβ j   +   ([αj, β] + [βj , α]) = [ik , α]
∂x    
i k i k ∂x j
60 Tensors and Their Applications

 β ∂ β 
  = ∂x j [ik , α] − i k  ( [αi, β] + [βj, α ] )

g αβ Solved.
∂x j i k   

EXAMPLE 2
 k
Show that if g ij = 0 for i ≠ j then (i)   = 0 whenever i, j and k are distinct.
i j 
 i  1 ∂ log gii  i  1 ∂ log g ii  i  1 ∂g jj
(ii)   = (iii)   = (iv)   = −
i i  2 ∂x i j  2 ∂x 2 gii ∂x i
i j
 j j
Solution
The Christoffel’s symbols of first kind
1  ∂g jk ∂g ik ∂g ij 
[ij, k ] =  + −  ...(1)
2  ∂x i ∂x j ∂x k 
(a) If i = j = k
The equation (1) becomes
1 ∂gii
[ii, i] =
2 ∂x i
(b) If i = j ≠ k
The equation (1) becomes
1  ∂g i k ∂gi k ∂g i i 
[ii, k ] = 2  ∂x i
+ i – k
∂x ∂x 
Since g ik = 0 as i ≠ k (given)
1 ∂g ii 1 ∂g jj
[ii, k ] = − or [ jj, i] = −
2 ∂x k
2 ∂xi
(c) i = k ≠ j

1  ∂g ji ∂g ii ∂g ij 
[ij,i ] =  + j − i
2  ∂x i ∂x ∂x 
1 ∂g ii
= , as g ij = 0, i ≠ j
2 ∂x j
(d) i ≠ j ≠ k
[ij, k ] = 0 as g ij = 0 , g j k = 0 , i ≠ j ≠ k

(i) as i, j, k are distinct i.e., i ≠ j ≠ k

 k 
  = g kl [ij, l ] since g k l = 0, k ≠ l
i j 
Christoffel's Symbols and Covariant Differentiation 61

k 
  =0
i j 
i
(ii)   = g ii [ii, i ]
i i 
1 ∂gii
= g ii ⋅ from (a)
2 ∂x i
1 ∂g ii g ii =
1
= i as
2 gii ∂x gii
i 1 ∂ log g ii
  =
i i  2 ∂x i
(iii) i=k≠ j

i 
  = g [ij, i]
ii

i j 

= g [ij, i] as g =
1 ii 1
ii gii
 i  1 ∂ log gii
  =
i j  2 ∂x j
(iv) j=k≠i

 i 
  = g [ jj, i]
ii

 j j 
1 1 ∂g j j
= − from (b)
g ii 2 ∂x i
 i  − 1 ∂g j j
  = 2g Solved.
i i ∂x
i
 j j

EXAMPLE 3

If ds 2 = dr 2 + r 2 dθ 2 + r 2 sin 2 θdφ 2 , find the values of

 1  3 
(i) [22, 1] and [13, 3], (ii)   and  
2 2 1 3
Solution

The given metric is metric in spherical coordinates, x1 = r , x 2 = θ, x 3 = φ.


Clearly,
g11 = 1, g 22 = r 2 , g 33 = r 2 sin 2 θ and gi j = 0 for i ≠ j
62 Tensors and Their Applications

1 g 33 = 1
g 11 = 1, g =
22
Also, ,
r2 r sin 2 θ
2

(See Ex. 2, Pg. 39, and g ij = 0, for i ≠ j. )


(i) Christoffel Symbols of first kind are given by

1  ∂g jk ∂g ik ∂g ij 
[ij, k ] = 
2  ∂x i
+ − , i, j, k = 1, 2,3 ...(1)
∂x j ∂x k 
Taking i = j = 2 and k = 1 in (1)

1  ∂g 21 ∂g 21 ∂g 22 
[22, 1] = + 2 − 1  Since g21 = 0.
2  ∂x 2 ∂x ∂x 

1  ∂0 ∂0 ∂r 2 
=  2 + 2 − 1
2  ∂x ∂x ∂x 

1 ∂r 2
= − = −r
2 ∂r
Taking i = 1, j = k = 3 in (1)

1  ∂g33 ∂g13 ∂g13 


[13,3] = + 3 − 3
2  ∂x1 ∂x ∂x 

1 ∂r 2 sin 2 θ
= since g13 = 0
2 ∂r
[13, 3] =
r sin 2 θ
(ii) Christoffel symbols of the second kind are given by

k
  = g kl [ij, l] = g k 1 [ij ,1] + g k 2 [ ij ,2] + g k 3 [ ij,3]
i j 
Taking k = 1, i = j = 2.

 1 
  = g 11 [22, 1] + g 12 [22, 2 ] + g13 [22, 3]
 2 2 

 1 
  = 1[22, 1] + 0[22, 2 ]+ 0[22, 3] Since g 12 = g 13 = 0
 2 2 

 1 
  = −r
2 2
Christoffel's Symbols and Covariant Differentiation 63

and

3 
  = g [13, 1] + g [13, 2] + g [13, 3]
31 32 33

 
1 3

=
1
[13, 3] Since g 31 = g 32 = 0
r sin θ
2 2

 3  1 1
 = 2 2 ⋅ r sin θ =
2

1 3  r sin θ r

4.2 TRANSFORMATION OF CHRISTOFFEL'S SYMBOLS


The fundamental tensors g ij and g ij are functions of coordinates x i and [ij, k ] is also function of

coordinates x i . Let g ij , g − ij and [ ij , k ] in another coordinate system x i .

(i ) Law of Transformation of Christoffel's Symbol for First Kind


Let [i j , k ] is a function of coordinate x i and [ ij , k ] in another coordinate system x i . Then

1  ∂gik ∂g ik ∂g ij 
[ ij , k ] =  + −  ...(1)
2  ∂x i ∂x j ∂x k 

Since g ij is a covariant tensor of rank two. Then

∂x p ∂x 2
g ij = g pq ...(2)
∂x i ∂x j

Differentiating it w.r.t. to x k , we get

∂g ij ∂  ∂x p ∂x q 
 
∂x k
=
∂x k  ∂x i ∂x j g pq 
 

∂  ∂x p ∂x q  q ∂g
 g pq + ∂x ∂x
p
 pq
=
∂x k  ∂x i ∂x j  ∂x ∂x ∂x k
i j
 

∂g i j ∂ 2 x p ∂x q ∂x p ∂ 2 x q  q ∂g
 g pq + ∂x ∂x pq ∂x
p r

∂xk
=  ∂x k ∂x i ∂x j + ∂x i ∂x k ∂x j  ∂x i ∂x j ∂x r ∂x k
...(3)
 
Interchanging i, k and also interchanging p, r in the last term in equation (3)

∂g kj  ∂ 2 x p ∂x q ∂x p ∂ 2 x q  q ∂g
 g pq + ∂x ∂x rq ∂x
r p
=  i k + k  ...(4)
∂x i  ∂x ∂x ∂x ∂x ∂x i ∂x j ∂x k ∂x j ∂x p ∂x i
j

64 Tensors and Their Applications

and interchanging j, k and also interchange q, r in the last term of equation (3)

∂ g ik  ∂ 2 x p ∂x q ∂x p ∂ 2 x q  q ∂g
 g pq + ∂x ∂x ∂x
p r
=  j i + i
pr
 ...(5)
∂x j  ∂x ∂x ∂x
i
∂x ∂x k ∂x j  ∂ x i
∂x k
∂ x j
∂ x q

Substituting the values of equations (3), (4) and (5) in equation (1), we get

1  ∂ 2 x p ∂x q ∂x p ∂x q ∂x r  ∂g rp ∂g qr ∂g pq 
[ij, k ] =  2 i j k g pq + i  q + − 
2  ∂x ∂x ∂x ∂x ∂x j ∂x k  ∂x ∂x p ∂x r 

∂ 2 x p ∂x q ∂x p ∂x q ∂x r 1  ∂g rp ∂g qr ∂g pq 
[ij, k ] = g +  + p − 
∂x i ∂x i ∂x k 2  ∂x q
pq
∂x i ∂x j ∂x k ∂x ∂x r 

∂ 2 x p ∂x q ∂x p ∂x q ∂x r
[ij, k ] = g + [ pq, r] ...(6)
∂x i ∂x j ∂x k
pq
∂x i ∂x i ∂x k
It is law of transformation of Christoffel's symbol of the first kind. But it is not the law of
transformation of any tensor due to presence of the first term of equation (6).
So, Christoffel's symbol of first kind is not a tensor.
(ii) Law of Transformation of Christoffel's Symbol of the Second Kind
 k   k 
Let g kl
[ij , l ] =   is function of coordinates x i and g [ij, l ] =
kl
  in another coordinate system
i j  i j 
i
x . Then
∂ 2 x p ∂x q ∂x p ∂x q ∂x r
[ij, l ] = g pq + i [ pq, r ] from (6)
∂x ∂x ∂x
i j l
∂x ∂x j ∂x l
kl
As g is contravariant tensor of rank two.

∂x k ∂x l st
g kl = g
∂x s ∂x t
Now

∂x k ∂x l st ∂ 2 x p ∂x q ∂x k ∂x l st ∂x p ∂x q ∂x r
g kl [ij, l ] = g g + g [pq, r]
∂x s ∂x t ∂x i x j ∂x l
pq
∂x s ∂x t ∂x i ∂x j ∂x l

∂x k  ∂x l ∂x q  st ∂ 2 x p ∂x k  ∂x l ∂x r  ∂x p ∂x q st
 ∂x t ∂x l  ∂x i ∂x j g [ pq, r ]
 g g +  
=
∂x s  ∂x t ∂x l  ∂x i x j pq
∂x s
   

∂x k 2 st ∂ 2 x p ∂x k r ∂x p ∂x q st ∂x l ∂x q
= δ g g + δ g [ pq , r ] as = δqt
∂x
t
∂x x
pq
∂x
t
s i j s
∂x ∂x
i j
∂x ∂x
t l

∂x k ∂ 2 x p sq ∂x k ∂x p ∂x q sr
= g g + g [ pq, r ] as δ rt g st = g sr
∂x s ∂x i ∂x j
pq
∂x s ∂x i ∂x j
Christoffel's Symbols and Covariant Differentiation 65

∂x p ∂ 2 x p s ∂x k ∂x p ∂x 2  s 
g kl [ij, l ] = δp + s  
∂x s ∂x i ∂x j ∂x ∂x i ∂x j  p q 
 s 
= δ sp and g [ pq, r ] =  p q 
sr
Since g sq g pq
 
 k  ∂x k ∂ 2 x s ∂x k ∂x p ∂x q  s 
  = +   ....(7)
i j  ∂x s ∂x i ∂x j ∂x s ∂x i ∂x j  p q
It is law of transformation of Christoffel’s symbol of the second kind. But it is not the law of
transformation of any tensor. So, Christoffel’s symbol of the second kind is not a tensor.

∂x s
Also, multiply (7) by , we get
∂x k

∂x s  k  ∂x s ∂x k ∂ 2 x s ∂x s ∂x k ∂x p ∂x q  s 
  = +  
dx k i j  ∂x k ∂x s ∂x i ∂x j ∂x k ∂x s ∂x i ∂x j p q

∂x s ∂x k
Since = δ ss = 1
∂x k ∂x s

∂x s  k  ∂ 2 xs ∂x p ∂x q  s 
  = +  
∂x k i j  ∂x i ∂x j ∂x i ∂x j  p q

∂ 2 xs ∂x s  k  ∂x p ∂x q  s 
= i j  − i j   ...(8)
∂x i ∂x j ∂x k   ∂x ∂x  p q 

It is second derivative of x s with respect to x ’s in the terms of Christoffel’s symbol of second


kind and first derivatives.
THEOREM 4.4 Prove that the transformation of Christoffel’s Symbols form a group i.e., possess the
transitive property.
Proof: Let the coordinates x i be transformed to the coordinate system x i and x j be transformed to x i .
When coordinate x i be transformed to x i , the law of transformation of Christoffel’s symbols of
second kind (equation (7)) is
 k  ∂x k ∂ 2 x s ∂x k ∂ x p ∂ x q  s 
  = +   ...(1)
i j  ∂ x s ∂ x i ∂x j ∂x s ∂x i ∂ x j  p q 

When coordinate x i be transformed to x i . Then

 r   k  ∂x i ∂x j ∂x r ∂ 2 x k ∂x r
  =   +
i j  ∂x ∂x ∂x ∂x u ∂x v ∂x k
u v k
u v
66 Tensors and Their Applications

 s  ∂x k ∂x p ∂x q ∂x i ∂x j ∂x r ∂ 2 x s ∂x k ∂x i ∂x j ∂x r

= p q  +
  ∂x s ∂x i ∂x j ∂x u ∂x v ∂x k ∂x i ∂x j ∂x s ∂x u ∂x v ∂x k

∂ 2 x k ∂x r
+
∂x u ∂x v ∂x k

 r   s  ∂x p ∂x q ∂x r 2 k
∂ x ∂x
r
 =   + +
u v  p q  ∂x ∂x ∂x
u v s u v k
∂x ∂x ∂x

∂ 2 x s ∂x r ∂x i ∂x j
...(2)
∂x i ∂x j ∂x s ∂x u ∂x v

∂x
p
∂x
i
∂x
p

as i u = u
∂x ∂x ∂x
Since we know that

∂x s ∂x i ∂x s
= ...(3)
∂x i ∂x u ∂x u

Differentiating (3) w.r.t. to x v , we get

∂  ∂x s  ∂x i ∂x s ∂  ∂x i 
= ∂ x
2 s
  
∂x v  ∂x i  ∂x u + ∂x i ∂x v  ∂x u 
    ∂x u ∂x v

∂ 2 x s ∂x j ∂x i ∂x s ∂ 2 x i ∂ 2 xs
+ = ...(4)
∂x i ∂x j ∂x v ∂x u ∂x i ∂x u ∂x v ∂x u ∂x v

∂x r
Mutiply (5) by .
∂x s

∂ 2 x s ∂x j ∂x i ∂x r ∂ 2 x i ∂x s ∂x r ∂ 2 x s ∂x r
+ =
∂x i ∂x j ∂x v ∂x u ∂x s ∂x u ∂x v ∂x s ∂x s ∂x u ∂x v ∂x s
Replace dummy index i by k in second term on L.H.S.

∂ 2 x s ∂x j ∂x i ∂x r ∂ 2 x k ∂x r ∂ 2 x s ∂x r
+ = ...(5)
∂x i ∂x j ∂x v ∂x u ∂x s ∂x u ∂x v ∂x k ∂x u ∂x v ∂x s
Using (5) in equation (2), we get

 r   s  ∂x p ∂x q ∂x r ∂ 2 x s ∂x r
  =   u v + ...(6)
 p q ∂x ∂x ∂x ∂x u ∂x v ∂x s
s
u v
Christoffel's Symbols and Covariant Differentiation 67

The equation (6) is same as the equation (1). This shows that if we make direct transformation
from x i to x i we get same law of transformation. This property is called that transformation of
Christoffel's symbols form a group.

4.3 COVARIANT DIFFERENTIATION OF A COVARIANT VECTOR


Let Ai and Ai be the components of a covariant vector in coordinate systems x i and x i respectively..
Then
∂x p
Ai = Ap ...(1)
∂x i
Differentiating (1) partially w.r.t. to x j ,

∂Ai ∂2 x p ∂x p ∂A p
= A +
∂x j ∂x j ∂x i
p
∂x i ∂x j

∂Ai ∂2 x p ∂x p ∂A p ∂x q
= A + ...(2)
∂x j ∂x j ∂x i
p
∂x i ∂x q ∂x j
It is not a tensor due to presence of the first term on the R.H.S. of equation (2).
Now, replace dum m y index p by s in the first term on R.H.S. of (2), we have

∂Ai ∂ 2 xs ∂x p ∂A p ∂x q
j =
As + i ...(3)
∂x ∂x ∂x
j i
∂x ∂x q ∂x j
Since we know that from equation (8), page 65,

∂ 2 xs ∂x s  k  ∂x p ∂x q  s 
=  − i j  
∂x i ∂x j ∂x k i j  ∂x ∂x  p q 

∂ 2 xs
Substituting the value of in equation (3), we have
∂x i ∂x j

∂Ai  ∂x s  k  ∂x p ∂x q  s   ∂x p ∂A p ∂x q
=  k   −   s A +
∂x j  ∂x i j  ∂x i ∂x j  p q   ∂x i ∂x q ∂x j

 k  ∂x s ∂x p ∂x q  s  ∂x p ∂x q ∂A p
=   k As − i   As + i
i j  ∂x ∂x ∂x j  p q ∂x ∂x j ∂x q

∂Ai  k  ∂x p ∂x q  ∂A p  s 
  A +  q − As   
∂x j
= k
∂x i ∂x j  ∂x
i j    p q
68 Tensors and Their Applications

∂Ai  k  ∂x p ∂x q  ∂A p  s 
 
− Ak  =
∂x i ∂x j  ∂x q − As  p q   ...(4)
∂x j i j   
Now, we introduce the comma notation

∂Ai k 
Ai , j = − Ak   ...(5)
∂x j
i j
Using (5), the equation (4) can be expressed as

∂x p ∂x q
Ai , j = Ap, q ..(6)
∂x i ∂x j
It is law of transformation of a covariant tensor of rank two. Thus, Ai , j is a covariant tensor of
rank two.
So, Ai , j is called covariant derivative of Ai with respect to x j .

4.4 COVARIANT DIFFERENTIATION OF A CONTRAVARIANT VECT OR


Let Ai and A i be the component of contravariant vector in coordinate systems x i and x i respectively..
Then
∂x i s
Ai = A
∂x s
∂x s i
or As = A
∂x i
Differentiating it partially w.r.t. to x j , we get

∂A s ∂ 2 xs ∂x s ∂A i
= A i
+ ...(1)
∂x j ∂x j ∂x i ∂x i ∂x j
Since from equation (8) on page 65,

∂ 2 xs ∂ x s  k  i ∂ x p ∂x q  s 
=  A −  
∂x j ∂x i ∂ x k i j  ∂ x i ∂x j  p q 

∂ 2 xs
substituting the value of in the equation (1), we get
∂x j ∂x i

∂A s ∂x s  k  i ∂x p ∂x q  s  i ∂x s ∂A i
=  A − i  A + i
∂A j ∂x k i j  ∂x ∂x j  p q  ∂x ∂x j

∂A s ∂A q ∂x s  k  i ∂x p i ∂x q  s  ∂x s ∂A i
=  A − i A  + i
∂A q ∂x j ∂x k i j ∂x ∂x j  p q  ∂x ∂x
j
Christoffel's Symbols and Covariant Differentiation 69

∂x p i
Interchanging the dummy indices i and k in the first term on R.H.S. and put A = A p we get
∂x i
∂A s ∂x q ∂x s  i  k ∂x q p  s  ∂x s ∂A i
=  A − j A  + i
∂x q ∂x j ∂x i k j ∂x  p q  ∂x ∂x
j

∂x q  ∂A s p  s 
 ∂x s   i  k ∂A i 
  
∂x j  ∂x q + A  p q  = ∂x i  k A + j 
  j  ∂x
 

 i   ∂A s 
∂A i ∂x i ∂x q  p  s 
+ Ak   = + A   ...(2)
 ∂x
∂x ∂x s ∂x j  p q 
j
k j  q

Now, we introduce the comma notation

∂Ai  i 
+ Ak 
A,i j =  ...(3)
∂x j
k j 
Using (3), the equation (2) can be expressed as

∂x p ∂x q
A,ij = Ap, q ...(4)
∂x i ∂x j
It is law of transformation of a mixed tensor of rank two. Thus, Ai , j is a mixed tensor of rank
two. Ai , j is called covariant derivative of Ai with respect to x j .

4.5 COVARIANT DIFFERENTIATION OF TENSORS


Covariant derivative of a covariant tensor of rank two.
Let Ai j and Ai j be the components of a covariant tensor of rank two in coordinate system x i and x i
respectively then

∂x p ∂x q
Aij = A pq ...(1)
∂x i ∂x j

Differentiating (1) partially w.r.t. to x k

∂Aij ∂x p ∂x q ∂A pq ∂  ∂x p ∂x q 
+ k   A pq
∂x k
=
∂x ∂x ∂x
i j k
∂x  ∂x i ∂x j 
 

∂Aij ∂x p ∂x q ∂Apq ∂x r ∂ 2 x p ∂x q ∂x p ∂ 2 x q
= + A + A pq ...(2)
∂x k ∂x i ∂x j ∂x r ∂x k ∂x k ∂x i ∂x j
pq
∂x i ∂x k ∂x j

∂A pq ∂A pq ∂x r
as = (since A pq components in x i coordinate)
∂x k ∂x r ∂x k
70 Tensors and Their Applications

∂ 2 x p ∂x q ∂ 2 x l ∂x q
A pq = A
∂x i ∂x k ∂x j lq
∂x i ∂x k ∂x j

∂ 2 x p ∂x q ∂x q  h  ∂x l  l  ∂x p ∂x r 
A pq = A   h −   i 
∂x i ∂x k ∂x j lq
∂x j i k  ∂x  p r  ∂x ∂x 
k

Since we know that from equation (8) on page 65.

∂ 2 xl  h  ∂x l  l  ∂x p ∂x r
k =   h −  i
∂x ∂x
i
i k  ∂x  p r  ∂x ∂x
k

 h  ∂x l ∂x 2  l  ∂ x p ∂ x q ∂x r
∂ 2 x p ∂x q −
A pq 
= i k  Alq   Alq
∂x i ∂x k ∂x j   ∂x h ∂ x j  p r  ∂x i ∂x j ∂x k

∂ 2 x p ∂x q  h   l  ∂x p ∂x q ∂x r
A pq j =   Ahj −   Alq i ...(3)
∂x ∂x ∂x
i k
i k  p r ∂x ∂x j ∂x k

∂x l ∂x q
as Ahj = Alq by equation (1)
∂x h ∂x j
and

∂x p ∂ 2 x q ∂x p ∂ 2 x l
Apq = A
∂x i ∂x j ∂x k pl
∂x i ∂x j ∂x k

∂x p  h  ∂x l  l  ∂x q ∂x r 
= Apl   h −   j 
∂x i  j k  ∂x q r  ∂x ∂x 
k

 h  ∂x p ∂x l  l  ∂x q ∂x r ∂x p
=   pl
A −  Apl
 j k  ∂x i ∂x h q r  ∂x j ∂x k ∂x i

∂x p ∂ 2 x q  h   l  ∂x p ∂x q ∂x r
Apq =   A −   i A pl ...(4)
∂x i ∂x j ∂x k
ih
 j k  q r  ∂x ∂x ∂x
j k

Substituting the value of equations (3) and (4) in equation (2) we get,

∂Aij  ∂A pq  l   l  ∂x p ∂x q ∂x r  h   h 
=  r − Alq  p r  − A pl q r  +   Ahj +   Aih
∂x k  ∂x     ∂x ∂x ∂x
i j k
i k   j k 

∂Aij  h   h   ∂A pq  l   l  ∂x p ∂x q ∂x r
−   Aih −   Ahj =  − Alq   − A pl  
∂x k  j k  i k   ∂x
r
 p r q r  ∂x ∂x ∂x
i j k
Christoffel's Symbols and Covariant Differentiation 71

∂Aij  h h 
Aij , k = −   Aih −   Ahj , then
∂x k
j k i k 

∂x p ∂x q ∂x r
Aij , k = Apq , r
∂x i ∂x j ∂x k
It is law of transformation of a covariant tensor of rank three. Thus, A ij,k is a covariant tensor
of rank three.
So, Aij , k is called covariant derivative of Aij w.r.t. to x k .
Similarly we define the covariant derivation xk of a tensors A ij and Aij by the formula
∂A ij  i   j
ij
A ,k = + A lj   + A il  
∂x k
l k  l k 
∂A ij  i   l 
and Aij ,k = + A lj   − Ali  
∂x k
l k   j k
In general, we define the covariant deriavative xk of a mixed tensor Aab
ij ... l
... c
by the formula

∂Aab
ij ...l
pj ... l  i  ip... l  j  ij... p  l 
ij ... l
Aab = ...c
+ Aab ...c   + Aab... c   + ⋅ ⋅ ⋅ + Aab... c  
∂x
... c ,k k
p k  p k p k
 p  ij... l  p  ij ... l  p 
− A ijpb......l c   − Aap... c   − ⋅ ⋅ ⋅ − Aab ... p  
a k  b k  c k 
Note: Ai , k is also written as Ai , k = ∇ k Ai .

4.6 RICCI'S THEOREM


The covariant derivative of Kronecker delta and the fundamental tensors gij and gij is zero.
Proof: The covariant derivative xp of Kronecker delta is

∂δ ij  i   l 
δ ij,k = + δ lj   − δ il  
∂x k
l k   j k

 i   i 
= 0+  − 
 j k  j k 

∂δ ij  i   i 
δ ij,k = 0 as = 0; δlj   =  
∂x k
l k   j k 
Also, consider first the tensor gij and the covariant derivative of gij is

∂g ij m  m
g ij , k = − g mj   − gim  
∂x k
i k  j k
72 Tensors and Their Applications

∂g ij m
g ij , k = − [ik , j ]− [ jk , i] as g mj   = [ik , j ]
∂x k
i k 
∂g ij
But = [ik , j ]+ [ jk, i]
∂x k

So, g ij , k = ∂g ij − ∂g ij
∂x k ∂x k
g ij , k = 0
W e can perform a sim ilar calculation for the tensor gij.
Since we know that g im g mj = δ ij . Similarly taking covariant derivative, we get

k g mj + g g mj , k = δ j , k
g ,im im i

gmj,k = 0 and δ ij , k = 0. So, g , k = 0 as g mj ≠ 0


im
But

EXAMPLE 4
Prove that if A ij is a symmetric tensor then

Ai j, j =
1 ∂
( 1
Ai j g − A jk )
∂g jk
g ∂x ∂x i
j
2

Solution

Given that A ij be a symmetric tensor. Then


A ij = A ji ...(1)

We know that

∂Ai j  j  l 
Ai j, k = + Ail   − Al j  
∂x k
l k  i k 
Put k = j, we get

∂Ai j  j  l 
Ai j, j = + Ail   − Al j   ...(2)
∂x j
l j  i j 

=
∂Ai j
+ Ail
(
∂ log g )
− Al j g hl [ ij, h]
∂x j
∂x l

∂Ai j Ai j ∂ g
= + − A jh [ij, h] since A ij is symmetric.
∂x j g ∂x j

Ai j, j =
(
1 ∂ Ai j g )
− A jk [ij, k ] ...(3)
g ∂x j
Christoffel's Symbols and Covariant Differentiation 73

But

1 jk  ∂g jk ∂g ki ∂g ij 
A jk [ij , k ] = A  i + − 
2  ∂x ∂x j ∂x k 

1  jk ∂g jk jk ∂g ki
∂g ij 
A jk [ij , k ] =  A + − 
jk
A A
2 ∂x i ∂x j ∂x k 

∂g ki ∂g
A jk
j =
A kj kji ...(4)
∂x ∂x
On Interchanging the dummy indices j & k.

∂g ki ∂g ji
A jk jk
= A since Aij = A ji
∂x j ∂x k

∂g ki ∂g
⇒ A jk − A jk ijk = 0 as g ij = g ji ...(5)
∂x j
∂x
Using (5), equation (4) becomes

1 jk ∂g jk
A jk [ij , k ] = A
2 ∂x i

Put the value of A jk [ij , k ] in equation (3), we get

Ai j, j =
(
1 ∂ Ai j g 1)
− A jk
∂g jk
Proved.
g ∂x j
2 ∂x i

EXAMPLE 5
 k  k  k  k
Prove that   −   are components of a tensor of rank three where   and  
i j a i j b i j a i j b
are the Christoffel symbols formed from the symmetric tensors aij and bij .
Solution
Since we know that from equation (8), page 65.
∂ 2 xs ∂x s  k  ∂x p ∂x q  s 
=  − i j  
∂x i ∂x j ∂x k i j  ∂x ∂x  p q 

∂x s  k  ∂ 2 xs ∂x p ∂x q  s 
  = +  
∂x k i j  ∂x i ∂x j ∂x i ∂x j  p q

 k   ∂ 2 xs ∂x p ∂x q  s  ∂x k
or   =  i j +   s
i j   ∂x ∂x ∂x i ∂x j  p q  ∂x
74 Tensors and Their Applications

Using this equation, we can write

 k   ∂ 2 xs ∂x p ∂x q  s   ∂x k
  =  +    s
 ∂x ∂x ∂x i ∂x j  p q  a  ∂x
i j
i j  a

 k   ∂ 2 xs ∂x p ∂x q  s   ∂x k
and   =  i j+ i    s
i j b  ∂x ∂x ∂x ∂x j  p q b  ∂x
Subtracting, we obtain

 k   k   s   s   ∂x p ∂x q ∂x k
    =−    −    i
 p q a  p q b  ∂x ∂x ∂x
j s
i j a i j b
Put

 s   s 
  −  = A pq
s

 p q a  p q b
Then above equation can written as

s ∂x p ∂x q ∂x k
Aijk = Apq
∂x i ∂x j ∂x s
It is law of transformation of tensor of rank three.
 k  k
So,   −   are components of a tensor of rank three.
i j a i j b

EXAMPLE 6
If a specified point, the derivatives of gij w.r.t. to xk are all zero. Prove that the components of
covariant derivatives at that point are the same as ordinary derivatives.
Solution
Given that
∂g ii
= 0, ∀ i , j, k at P 0 ...(1)
∂x k
Let Aij be tensor..
∂Aij
Now, we have to prove that Aij , k = at P0 .
∂x k
∂A ij  i  i  α 
Aij ,k = + A αj   − Aα   ...(2)
∂x α k 
k
j k
Christoffel's Symbols and Covariant Differentiation 75

 k ∂g ij
Since   and [ij,.k ] both contain terms of the type and using equation (1) we get
i j  ∂x k
 k
  = 0 = [ij, k] at P0 .
i j 
So, equation (2) becomes
∂A ij
Aij ,k = at P0
∂x k

4.7 GRADIENT, DIVERGENCE AND CURL


(a) Gradient
If φ be a scalar function of the coordinates, then the gradient of φ is denoted by
∂φ
grad φ =
∂x i
which is a covariant vector.
(b) Divergence
The divergence of the contravariant vector Ai is defined by
∂A i  i 
div A i = + Ak  
∂x i
 k i
It is also written as A,ii
The divergence of the covariant vector Ai is defined by
div Ai = g ik Aik

EXAMPLE 7

Prove that div A =


i (
1 ∂ g Ak )
g ∂x k
Solution:

If Ai be components of contravariant vector then


∂A i  i 
div A i = A, i = + Ak  
i

∂x i
k i 
Since
 i 
  =

∂x
log g = (
1 ∂ g
g ∂x k
)
k i 
k

So,
∂A i 1 ∂ g k
div A i = + A
∂x i
g ∂x k
76 Tensors and Their Applications

Since i is dummy index. Then put i = k , we get

∂A k 1 ∂ g k
div A i = + A
∂x k
g ∂x k

div A i =
1 ∂ g Ak ( ) ...(1)
g ∂x k
Proved
(c) Curl
Let Ai be a covariant vector then
∂Ai  k
Ai , j = − Ak  
∂x j
i j 
∂A j  k
and Aj, i = − Ak  
∂x i
 j i
are covariant tensor.
∂Ai ∂A j
So, Ai , j − A j , i = − is covariant tensor of second order, which is called curl of Ai .
∂x j ∂x i
Thus
curl Ai = Ai , j − A j ,i
Note: curl A i is a skew-symmetric tensor.
Since
A j, i –A i,j = – (A i,j –A j,i)

EXAMPLE 8
If Aij be a skew-symmetric tensor of rank two. Show that
∂Aij ∂A jk ∂Aki
Aij , k + A jk ,i + Aki, j = + +
∂x k
∂x i
∂x j
Solution
Since we know that
∂Aij  l   l 
Aij , k = − Alj   − Ail  
∂x k
 
i k  j k

∂A jk h  l 
A jk ,i = − Alk   − A jl  
∂x i
 j i  j k

∂Aki  l   l 
Aki, j = − Ali   − Akl  
∂x j
k j i j 
Christoffel's Symbols and Covariant Differentiation 77

Adding these, we get

∂Aij ∂A jk ∂Aki   l   l 
Aij , k + A jk ,i + Aki, j = + + − A   + A   
∂x j  i k 
lj jl
∂x k ∂x i k i 
  l   l    l   l 
−  Ali   + Ail    −  Akl   + Alk   
 k j   j k    i j   j i

 l  l   l 
Since   is symmetric i.e.,   =   etc.
i k  i k  k i 

∂Aij ∂A jk ∂Aki  l 
= + + −   ( Alj + A jl )
∂x k
∂x i
∂x j i k 
 l   l 
−  ( Ali + Ail ) −   ( Akl + Alk )
k j i j
Since Aij is skew-symmetric. Then Alj = − A jl ⇒ Alj + A jl = 0 . Similarly,,
Ali + Ail = 0 and Akl + Alk = 0
So,
∂Aij ∂A jk ∂Aki
Aij , k + A jk ,i + Aki, j = + +
∂x k
∂x i
∂x j
THEOREM 4.5 A necessary and sufficient condition that the curl of a vector field vanishes is that
the vector field be gradient.
Proof: Suppose that the curl of a vector Ai vanish so that
curl Ai = Ai , j − A j , i = 0 ...(1)
To prove that Ai = ∇φ, φ is scalar..
Since from (1),
Ai , j − A j ,i = 0
∂Ai ∂A j
⇒ − =0
∂x j ∂xi
∂Ai ∂A j
⇒ =
∂x j
∂x i
∂Ai ∂A j j

j
dx = dx
∂x j ∂xi

⇒ dAi =

∂x i
A j dx j ( )
Integrating it we get

Ai = ∫

∂x i
(
A j dx j )
78 Tensors and Their Applications


∂x
=i ∫
A j dx j

∂φ
Ai =
∂x i
, where φ = ∫ A j dx j

or Ai = ∇φ.
Conversely suppose that a vector Ai is such that
Ai = ∇φ, φ is scalar..
To prove curl Ai = 0
Now,
∂φ
Ai = ∇φ =
∂x i
∂Ai ∂ 2φ
=
∂x j ∂x j ∂xi
∂A j ∂ 2φ
and =
∂x i ∂x i ∂x j
∂Ai ∂A j
So, − =0
∂x j ∂xi
∂Ai ∂A j
So, curl Ai = Ai , j − A j ,i = − =0
∂x j ∂x i

So, curl Ai = 0 Proved.

THEOREM 4.6 Let φ and ψ be scalar functions of coordinates xi. Let A be an arbitrary vector then
(i) div (φA) = φ div A + A ⋅ ∇φ
(ii) ∇(φψ) = φ∇ψ + ψ∇φ

(iii) ∇ 2 (φψ) = φ∇ 2 ψ + ψ∇ 2 ψ + 2∇φ ⋅ ∇ψ

(iv) div (ψ∇φ) = ψ∇ 2 φ + ∇φ ⋅ ∇ψ

Proof: (i) Since we know that

div A i =
(
1 ∂ g Ai ) ...(1)
g ∂x i

replace A i by φA i , we get

div( φA ) =
i
(
1 ∂ g φAi )
g ∂x i
Christoffel's Symbols and Covariant Differentiation 79

1  ∂ ( g Ai ) i ∂φ

=  φ + g A ⋅ 
 ∂x i ∂x 
i
g

1 ∂( g A )
i
∂φ
⋅A +φ
i
=
∂x
i
∂x
i
g
= ∇φ ⋅ A i + φ div A i
Thus
div( φA) = ∇φ ⋅ A + φ div A ...(2)
(ii) By definition of gradient,
∂(φψ )
∇(φψ) =
∂x i

∂ψ ∂φ
= φ +ψ i
∂x i
∂x
Thus ∇(φψ) = φ∇ψ + ψ∇φ ...(3)
(iii) Taking divergence of both sides in equation (3), we get
div (∇φψ) = div [φ∇ψ + ψ∇φ ]

∇ 2 (φψ) = div (φ∇ψ) + div (ψ∇φ)


= ∇φ ⋅ ∇ψ + φ div (∇ψ) + ∇ψ ⋅ ∇φ + ψdiv (∇φ)

∇ 2 (φψ) = φdiv (∇ψ) + ψdiv (∇φ) + 2∇φ ⋅ ∇ψ


Thus,

∇ 2 (φψ) = φ∇ 2 ψ + ψ∇ 2 φ + 2∇φ ⋅ ∇ψ
(iv) Replace A by ∇ψ in equation (3), we get
div (φ∇ψ) = ∇φ ⋅ ∇ψ + φ div (∇ψ)

div (φ∇ψ) = ∇φ ⋅ ∇ψ + φ∇ 2 ψ

THEOREM 4.7 Let Ai be a covariant vector and φ a scalar function. Then


(i) curl (φA) = A × ∇φ + φ curlA
(ii) curl (ψ∇φ) = ∇φ × ∇ψ
Proof: (i) Let Ai be a covariant vector then
curl A = curl Ai = Ai , j − A j ,i
Replacing Ai by φAi , we get
( )
curl (φAi ) = (φAi ), j − φA j , i
80 Tensors and Their Applications

= φ, jAi + φAi , j − φ, i A j − φA j , i

= ( Ai φ, j − A j φ, i) + φ( Ai , j − A j ,i )

= Ai × ∇φ + φcurl Ai
So,
curl (φA) = A × ∇φ + φcurlA ...(1)
(ii) Replacing A by ∇ψ in equation (1), we get
curl (φ∇ψ) = φcurl (∇ψ) + ∇ψ × ∇φ
Interchange of φ and ψ , we get
curl (ψ∇φ) = ψcurl (∇φ) + ∇φ × ∇ψ.
Since curl (∇φ ) = 0.
So,
curl (ψ∇φ) = ∇φ × ∇ψ. Proved.

4.8 THE LAPLACIAN OPERATOR


The operator ∇ 2 is called Laplacian operator read as "del square".

THEOREM 4.8 If φ is a scalar function of coordinates x i then


1 ∂  ∂φ 
∇ 2φ =  g g kr r 
g ∂x  ∂x 
k

Proof: Since
∇ 2 φ = div grad φ ...(1)
and
∂φ
grad φ = ,
∂x r
which is covariant vector.
But we know that any contravariant vector A k associated with Ar (covariant vector) is
A k = g kr Ar (Sec Art. 3.4, Pg 43)
∂φ
Now, the contravariant vector A k associated with (Covariant vector) is
∂x r
∂φ
A k = g kr
∂x r
Since

div A i =
1 ∂ g Ak( )
, (Sec Ex. 7, Pg 75)
g ∂x k
Christoffel's Symbols and Covariant Differentiation 81

So, from (1)


 ∂φ 
∂  g g kr r 
 ∂φ  1  ∂ x 
∇ 2 φ = div  g kr r  = Proved.
 ∂x  g ∂x k

EXAMPLE 9
Show that, in the cylindrical coordinates,
1 ∂  ∂V  1 ∂ 2V ∂ 2V
∇ 2V = r + +
r ∂r  ∂r  r 2 ∂θ 2 ∂z 2
by Tensor method
Solution
The cylindrical coordinates are (r , θ, z ). If V is a scalar function of (r , θ, z ).
Now,
∇ 2V = ∇ ⋅ ∇V
Since
∂V ∂V ∂V
∇V = i +j +k
∂r ∂θ ∂z
Let
∂V ∂V ∂V
A1 =, A2 = , A3 = ...(1)
∂r ∂θ ∂z
Then ∇V = iA1i + jA2 j + kA3 k , since ∇V is covariant tensor. The metric in cylindrical coordinates
is
ds 2 = dr 2 + r 2 dθ2 + dz 2

here, x1 = r , x 2 = θ, x 3 = z .
Since ds 2 = g ij dx i dx j
g11 = 1, g 22 = r 2 , g 33 = 1
and others are zero.

g11 g12 g13 1 0 0


g = g ij = g 21 g 22 g 23 = 0 r 2 0 = r 2
g 31 g 32 g33 0 0 1

g =r (See Pg. 34, Example 1)


Now,

div (∇V ) = div Ai =


(
1 ∂ gA
k
)
g ∂x k
82 Tensors and Their Applications

1 ∂( g A )
1
1 ∂( g A )
2
∂( g A )
3

= + +
∂x
1
∂x
2
∂x
3
g g

1  ∂ ( rA1 ) ∂ ( rA 2 ) ∂ ( rA3 ) 
div (∇V ) =  + +  ...(2)
r  ∂r ∂θ ∂z 
We can write
A k = g kq Aq (Associated tensor)
A k = g k1 A1 + g k 2 A2 + g k 3 A3
Put k = 1
A1 = g A1 + g A2 + g A3
11 12 13

A1 = g 11 A1 as g 12 = g 13 = 0
Similarly,
A 2 = g 22 A2
A3 = g 33 A3
and
Cofactor of g11 in g r 2
g 11 = = 2 =1
g r

Cofactor of g 22 in g 1
g 22 = = 2
g r

Cofactor of g 33 in g r 2
g 33 = = 2 = 1 (See. Pg. 34, Ex.1)
g r
So,
A1 = g A1 = A1
11

1
A 2 = g A2 = 2 A2
22

r
A3 = g A3 = A3 .
33

or A1 = A1, A 2 = 1 A2 , A3 = A3 .
r2
from (1), we get
∂V 1 ∂V ∂V
A1 = ∂r , A = r 2 ∂θ , A = ∂z
2 3

from (2),
Christoffel's Symbols and Covariant Differentiation 83

So,

1  ∂  ∂V  ∂  1 ∂V  ∂  ∂V 
∇ 2V = div A = r  ∂r  r ∂r  + ∂θ  r r 2 ∂θ  + ∂ z  r ∂z 
i

      

1  ∂  ∂V  ∂  1 ∂V  ∂  ∂V  
div (∇V ) = r +   + r 
r  ∂r  ∂r  ∂θ  r ∂θ  ∂z  ∂z  

1  ∂  ∂V  1 ∂ 2 V ∂ 2V 
=  r + + r 
r  ∂r  ∂r  r ∂θ 2 ∂z 2 

1 ∂  ∂V  1 ∂ 2V ∂ 2V
div (∇V ) = r + +
r ∂r  ∂r  r 2 ∂θ 2 ∂z 2

1 ∂  ∂V  1 ∂ 2V ∂ 2V
∇ 2V = r + +
r ∂r  ∂r  r 2 ∂θ 2 ∂z 2

EXERCISES

1. Prove that the expressions are tensors


∂Aij α   α 
(a) Aij , l = −   Aαj −   Ai α
∂x i l  jl 
l

∂Air jk α r α α  r  Aα
Air jk ,l = −  A −   Ar −   Ar  
l  αjk  j l  i αk k l  ijα + α l  ijk
(b)
∂x l
i
2. Prove that
j
1 ∂ ( Ai g) j k 
Ai j, j = j
− Ak  
g ∂x i j

1 ∂
3. If A ijk is a skew-symmetric tensor show that ( g Ai jk ) is a tensor..
g ∂xk
4. Prove that the necessary and sufficient condition that all the Christoffel symbols vanish at a point is
that g ij are constant.
5. Evaluate the Christoffel symbols in cylindrical coordinates.
6. Define covariant differentiation of a tensor w.r. to the fundamental tensor gij . Show that the covariant
differentiation of sums and products of tensors obey the same result as ordinary differentiation.
7. Let contravariant and covariant components of the same vector A be Ai and Ai respectively then
prove that
i
div A = div Ai
84 Tensors and Their Applications

8. If Aij is the curl of a covariant vector. Prove that


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

9. To prove that u ⋅∇u = – ucurl u if u a vector of constant magnitude.


10. A necessary and sufficient condition that the covariant derivative vector be symmetric is that the
vector must be gradient.
11. Show that, in spherical coordinates

1 ∂  2 ∂V  1 ∂  ∂V  1 ∂ 2V
2
∇ V = r + 2  sin θ + 2 2
r ∂r  ∂ r  r sin θ ∂ θ
2
∂ θ  r sin θ ∂ φ2

by tensor method.
CHAPTER – 5

RIEMANN-CHRISTOFFEL TENSOR

5.1 RIEMANN-CHRISTOFFEL TENSOR


If A i is a covariant tensor then the covariant x j derivative of A i is given by
∂Ai  α 
A i,j = −   Aα ...(1)
∂x j i j 
Differentiating covariantly the equation (1) w.r. to x k , we get
∂Ai , j  α  α 
Ai , jk = −   Aα, j −   Ai , α
∂x k
i k   j k 
∂  ∂A  α    α   ∂A  β  
 i −   A  −    α − 
=  ∂x α
   Aβ 
∂x k j
  
i j   i k   ∂ x j

α j  

 α   ∂A  γ  
−    αi −   Aγ 
i k   ∂x i k  

 α 
∂ 
∂ Ai
2
i j   α  ∂ Aα  α  ∂ Aα
Ai , jk = − Aα −   − 
∂x k ∂x ∂x i k  ∂x  i k  ∂ x
j k k j

 α   β   α  ∂Ai  α   γ 
+    Aβ −   α +   Aγ ...(2)
i k  α j   j k  ∂x  j k  i α 
Interchanging j and k in equation (2), we get
α 
∂ 
∂ Ai
2
i k   α  ∂A  α  ∂A
Ai , kj = − Aα −   αj −   αk
∂x j ∂x k ∂x j i k  ∂x i j  ∂x
86 Tensors and Their Applications

 α   β   α  ∂Ai  α   γ 
` +    Aβ −   α +     Aγ ...(3)
i j  α k  k i  ∂x k j  i α
Subtract equation (3) from (2), we get
 α 
∂  α
∂ 
α   β  i j   α   β 
Ai , jk − Ai , kj =     Aβ − Aα −     Aβ + i k  A
α
i k  α j  ∂x k
i j  α k  ∂x j

Interchanging of α and β in the first and third term of above equation, we get

 
  α   α  
∂   ∂  
 i k  − i j   β   α   β   α  Aα
Ai , jk − Ai , kj =  +   −    
β k 
...(4)
∂x j ∂x k i k  β j  i j 

 

Ai , jk − Ai , kj = Aα Riαjk ...(5)
where

 α   α 
∂  ∂ 
i k   i j   β   α   β   α 
Riαjk = − +   −    ...(6)
∂x j ∂xk  i k  β j   i j  β k 

Since A i is an arbitrary covariant tensor of rank one and difference of two tensors A i, jk – Ai, kj is
a covariant tensor of rank three. Hence it follows from quotient law that Riαjk is a mixed tensor of rank
four. The tensor Riαjk is called Riemann Christoffel tensor or Curvature tensor for the metric g ij dx i dx j .
The symbol Riαjk is called Riemann’s symbol of second kind.
Now, if the left hand side of equation (4) is to vanish i.e., if the order of covariant differentiation
is to be immaterial then
Riαjk = 0
Since A α is arbitrary. In general Riαjk ≠ 0, so that the order of covariant differentiation is not
immaterial, It is clear from the equation (4) that “a necessary and sufficient condition for the validity of
inversion of the order of covariant differentiation is that the tensor Riαjk vanishes identically..
Remark
The tensor

∂ ∂  i   i 
   
∂x k ∂x α k  α l 
l
i
=  i  + ...(7)
Rikl  i   α α 
       
j k  
j l j k  j l
Riemann-Christoffel Tensor 87

THEOREM 5.1 Curvature tensor Riαjk is anti symmetric w.r.t. indices j and k.
Proof: We know that from (7) curvature tensor

∂ ∂  α   α 
   
∂x j ∂ xk β j  β k 
Riαjk =  α   α  +
     β   β 
       
i j  i k   i j   i k 

 α   α 
∂  ∂ 
i k  i j   α   β   α   β 
Riαjk = − +   −   
∂x j ∂x k β j  i k  β k  i j 
Interchanging j and k, we get

α α
∂  ∂  
i j  − i k  +  α   β  −  α   β 
R iαkj = β k  i j  β j i k 
∂x k ∂x j      

  α   α  
∂   ∂  
 i k  i j   α   β  α   β 
= − − +  −   
 ∂x ∂x k β j  i k  β k  i j 
j

 
 

Rikα j = – Riαjk

So, Riαjk antisymmetric w.r.t. indices j and k.

Theorem 5.2 To prove that


Riαjk + R αjk i + R kiα j = 0
Proof: Since we know that

∂ ∂  α   α 
   
∂x j
∂x k β j  β k 
+
=  α   α 
α
Rijk  β   β 
       
i j  i k  i j  i k 

 α   α 
∂  ∂ 
  i j  α   β   β   α 
Rijαk = i k  − +   −    ...(1)
∂x j ∂x k β j i k  i j β k 
88 Tensors and Their Applications

Similarly

α α
∂  ∂  
α   β   β   α
=  k  −  i  + 
j i j k
R αj ki    −   ...(2)
∂x ∂x β k   j i  j k  β i
and

 α   α 
∂  ∂ 
k j  k i   α   β   β   α 
Rkiα = − +   −    ...(3)
∂ xi ∂x j
j
β i  k j k i  β j

On adding (1), (2) and (3), we get

Riαj k + R αj k i + Rkαi j = 0
This is called cyclic property.

5.2 RICCI TENSOR

The curvature tensor Riαjk can be contracted in three ways with respect to the index α and any one of
its lower indices

Rααj k , Riαα k , Riαj α

Now, from equation (7), art. 5.1,


∂ ∂  α   α 
   
∂x j ∂x k β j  β k 
+
=  α   α 
α
Rαjk  β   β 
       
α j  α k  α j  α k 
 α   α 
∂  ∂ 
α k  α j  α   β   α   β 
= − +   −   
∂x j ∂x k β j α k  β k  α j 

∂ 2 log g ∂ 2 log g
= –
∂x j ∂ x k ∂ xk∂ x j
 α  ∂ log g
Since α k  = and α and β are free indices Rααj k = 0 .
  ∂x k
Also for Riαj α .

Write R i j for R αi j α
Riemann-Christoffel Tensor 89

∂ ∂  α  α 
   
β j  β α
α
∂x j
∂x
Rij = Rijαα = +
 α  α  β  β
      
i j  i α i j  i α

α α
∂  ∂ 
α  j +  α   β  −  α   β 
Rij =  j  −
i i
β j  i α β α i j 
∂x ∂x α      

α
∂  
∂ 2 log g α  β   α  β 
−  α  + 
i j
Rij =  −    ...(1)
∂x j ∂x i ∂x  j i α β αi j 
β

Interchanging the indices i and j we get


α α
∂ ∂ 
α  i +  α  β  −  α   β 
R ji =  i  −
j j
β i  j α  β α  j i 
∂x ∂x α      

α
∂  
∂ 2 log g β α  α β 
−  α  +   
i j
R ji = −   ...(2)
∂x i ∂x j ∂x i α β j  β α  i j 
(Since α and β are dummy indices in third term).
Comparing (1) and (2), we get
Rij = R ji
Thus Rij is a symmetric Tensor and is called Ricci Tensor..
For Riααk :

Riααk = − Rikαα = −R ik

5.3 COVARIANT RIEMANN-CHRISTOFFEL TENSOR


The associated tensor
Ri j k l = g iα R αjkl ...(1)
is known as the covariant Riemann-Christoffel tensor or the Riemann-Christoffel tensor of the first
kind.
Expression for Rijkl
Rijkl = gi α Rαjkl
90 Tensors and Their Applications

 ∂  α  ∂  α   β   α   β   α 
= g iα  k  j l  − l  j k  +  j l  β k  −  j k  β l   …(2)
 ∂x   ∂x         
Now,
∂  α ∂   α   α  ∂g i α
g iα   =
∂x k  j l   gi α  j l  −  j l  ∂x k
∂x k  

∂[ jl , i ]  α  ∂g i α
= −  k ...(3)
∂x k  j l  ∂x
Similarly,
∂  α  ∂[ jk , i ]  α  ∂gi α
g iα l   = −  l ...(4)
∂x  j k  ∂x l  j k  ∂x
 udv duv vdu 
By the formula dx = dx − dx 
 
Using (3) and (4) in equation (2), we get

∂[ jl , i]  α  ∂gi α ∂[ jk , i ]  α  ∂g i α  β   α   β   α 
Ri j k l = −  k − +   l + gi α     − g iα    
∂x k  j l  ∂ x ∂ xl  j k  ∂x  j l  β k   j k  β l 

∂[ jl , i ] ∂[ jk , i ]  α  ∂g iα  α  ∂g iα  βα   β  α
= − +  l −  j l  k + g i α  j l  β k  − g iα  j k  β l 
∂x k
∂x l
 j k  ∂x   ∂x      

∂[ jl , i ] ∂ [ jk , i]  α   α
Ri j k l = − + ( [il, α ] + [αl , i ] ) −  ( [ik , α] + [αk , i] )
∂x k
∂x l
 j k  j l

 β  β 
+  [βk , i] −  [βl , i ]
 j l  j k

∂[ jl , i ] ∂ [ jk , i]  α   α
Ri j k l = − + [il, α] −  [ik , α ] ...(5)
∂x k
∂x l
 j k  j l
It is also written as

∂ ∂  α  α
   
Ri j k l = ∂x k
∂x k +  j k  j l ...(6)
[ jk , i ] [ jl, i] [ik ,α ] [il, α ]
But we know that

1  ∂gli ∂g ji ∂g jl 
[ jl, i] =  + − i  ...(7)
2  ∂x j ∂x l ∂x 
Riemann-Christoffel Tensor 91

1  ∂g ki ∂g ji ∂g ik 
and [ jk ,i ] =  + − i  ...(8)
2  ∂x j ∂x k ∂x 
Using (7) and (8), equation (5) becomes

1 ∂  ∂g li ∂g ji ∂g jl  1 ∂  ∂g ki ∂g ji ∂g jk 
Rijkl =  j + −  −  j + k − 
2 ∂x k  ∂x ∂ xl ∂ xi  2 ∂x
l
 ∂x ∂x ∂x i 
 α α 
+ [il, α ] −  [ik , α ]
 j k  j l

1  ∂ g il 
2 2
∂ g jk ∂ g jl
2 2
∂ g ik +
= 2 j k + − −
Rijkl 
 ∂x ∂x ∂x ∂x
i l
∂x ∂x ∂x ∂x
j l i k

α  α 
 [il , α] −  [ik , α ] ...(9)
 j k  j l
Since

 α  α
 = g αβ [ jk , β] and  j l  = g [ jl , β]
αβ

 j k   

1  ∂ 2 g il ∂ 2 g jk ∂ 2 g ik ∂ 2 g jl 

Ri j k l = + − −
2  ∂x j ∂x k ∂x i ∂x l ∂x j ∂x l ∂x i ∂x l 

+ g αβ [ jk , β][il, α ] − g αβ [ jl , β][ik , α ] ...(10)

This is expression for Ri j k l .


The equation (9) can also be written as
1  ∂ 2 gil ∂ 2 g jk ∂ 2 gik ∂ 2 g jl 

+ − −
= 2  ∂x j ∂x k ∂x i ∂x l ∂x j ∂x l ∂x i ∂x k
Ri j k l 
 
 α   β   α   β 
αβ 
+ g αβ     − g   
 j k  i l   j l  i k 

5.4 PROPERTIES OF RIEMANN-CHRISTOFFEL TENSORS OF FIRST KIND Ri j k l

(i) R jikl = − Rijkl

(ii) Rijlk = −Rijkl


(iii) Rklij = Rijkl
(iv) Rijkl + Riklj + Riljk = 0
92 Tensors and Their Applications

Proof: We know that from equation (9), Pg. 91.

1  ∂ 2 gil ∂ 2 g jk ∂ 2 g ik ∂ 2 g jl 

Rijkl = + − −
2  ∂x j ∂x k ∂x i ∂x l ∂ 2 x j ∂x l ∂x i ∂x k 
 

 α α 
+ [il, α ] −  [ik , α ] ...(1)
 j k  j l
(i) Interchanging of i and j in (1), we get

1  ∂ g jl ∂ 2 g jk  α 
2
∂ 2 gik ∂ 2 gil
R jikl =  i k = − − +
 i k 
[ jl, α]− iαl  [ jk , α]
2 ∂x ∂x ∂x j ∂x l ∂x i ∂x l ∂x j ∂x k  
 

1  ∂ 2 gil ∂ 2 g jk ∂ 2 g jl ∂ 2 gik   α 
 +   [ jl , α ] −  α 
= − + − −   [ jk , α ]
2  ∂x j ∂x k ∂x i ∂x l ∂x i ∂x k ∂x j ∂x l   i k  i l 

R jikl = − Rijkl

or Rijkl = − R jikl
(ii) Interchange l and k in equation (1) and Proceed as in (i)
(iii) Interchange i and k in (1), we get

1  ∂ 2 g kl ∂ 2 g ji ∂ 2 g ki ∂ 2 g jl  α 
 +  [kl, α] −  α [ki, α]
R kjil = + − −
2  ∂x j ∂x i ∂x k ∂x l ∂x j ∂x l ∂x k ∂x i   j i  j l

Now interchange j and l, we get

1  ∂ g kj ∂ 2 g lj  α
 +  [kj, α] −  α [ki, α ]
2
∂ 2 g li ∂ 2 g ki
R klij =  l i+ k j − l j − k i  l i
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x l j 

For
 α  α  αβ  β 
 [kj, α ] =   g  
l i   l i   k j 

 β  αβ  α 
=  g  
k j  l i 

 β 
=  [li, β]
k j

α α
 [li, β] = k j [li, α]
l i   
Riemann-Christoffel Tensor 93

So,

1  ∂ g kj ∂ 2 g lj   α  α
2
∂ 2 gli ∂ 2 g ki
R klij = + − −  +  [li, α] −  [ki, α]
2  ∂x l ∂xi ∂x k ∂x j ∂x l ∂x j ∂x k ∂x i  k j  l j

R klij = Rijkl , from (1)


from equation (1)

1  ∂ g il ∂ 2 g jk ∂ 2 g jk ∂ 2 g jl   α 
2

Rijkl = 2 j k + i l − j l − j k
+
  j k 
[il , α] −  jα l [ik , α ]
 ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x  

1  ∂ g ij ∂ 2 g kj  α 
 +  [ij, α] −  α [il , α]
2
∂ 2 g kl ∂ 2 g il
Riklj =  k l + i j − k j − i l  k l 
2  ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x k j 

1  ∂ g ik ∂ 2 g lj ∂ 2 gij ∂ 2 g lk  α 
2

Riljk = + − − +
 l j 
[ik , α] − lαk [ij, α ]
2  ∂x l ∂x j ∂x i ∂x k ∂x l ∂x k ∂x i ∂x j  
 
On adding these equations, we get
Rijkl + Riklj + R iljk = 0

This property of Rijkl is called cyclic property..

Theorem 5.3 Show that the number of not necessarily independent components of curvature tensor
1 2 2
does not exceed n (n − 1) .
12
Or
Show that number of distinct non-vanishing components of curvature tensor does not exceed
1 2 2
n (n − 1) .
12
Proof: The distinct non-vanishing components of Rijkl of three types.

(i) Symbols with two distinct indices i.e., Rijij . In this case total number of distinct non-vanishing
1
components of Rijkl are n(n − 1) .
2
(ii) Symbols with three distinct indices i.e., Rijik . In this case, total number of distinct non-
1
vanishing components Rijkl are n(n − 1) (n − 2) .
2
(iii) Symbols Rijkl with four distinct indices. In this case, total number of distinct non-vanishing
n (n − 1)
2 2

components of Rijkl is .
12
94 Tensors and Their Applications

Hence the number of distinct non-vanishing components of the curvature tensor Rijkl does not
1 2 2
exceed n (n − 1) .
12
Remark
When Rijkl is of the form Riiii i.e., all indices are same
In this case, Riiii has no components.

5.5 BIANCHI IDENTITY


It states that
R ijkl, m + R ijlm, k + R ijmk ,l = 0
and Rhjkl, m + R hjlm, k + Rhjmk ,l = 0
Proof: Introducing geodesic coordinate1 in which Christoffel symbols are constant with the pole at P0 .
Since we know that

∂ ∂  i  i 
   
∂x ∂x
k l
mk  ml 
R ijkl = +
 ii   m  m
      
  
jk jl  jk   jl 
i i 
∂  ∂ 
 jl  −  jk  +  i  m  −  m  i 
R ijkl =      
∂x k ∂x l mk  jl   jk ml 

 i  i 
∂2   ∂2  
 j l −  j k 
R ijkl, m = ...(1)
∂x m ∂x k ∂x m ∂x l
 i   m
Since  j k ,  j l  etc. are constant at pole.
   
So, their derivatives are zero.

∂ ∂  i   i 
   
∂x l ∂x m α l  α m
+
R ijlm =  i   i   α   α 
       
 j l   j m  j l   j m

1 Details of geodesic coordinate given in chapter curvature in curve . Geodesic.


Riemann-Christoffel Tensor 95

 i   i 
∂  ∂ 
     i   α   i   α 
=  j m −  j l  +    −   
∂x ∂ xm
l
α l   j m α m   j l 

 i  2 i 
∂2   ∂  
 j m −  j l 
R ijlm, k = ...(2)
∂x k ∂x l ∂x k ∂x m
and

∂ ∂  i   i 
   
∂x ∂x β m β k 
m k
R ijmk = +
 i   i   β   β 
       
 j m  j k   j m  j k

 i   i 
∂  ∂ 
 j k  −  j m +  i   β  −  i   α 
R ijmk =      
∂x m ∂x k β m   j k  β k   j m 

 i  2 i 
∂2   ∂  
 j k  j m
R ijmk = − ...(3)
∂x m ∂x l ∂x k ∂x l
On adding (1), (2) and (3), we get
R ijkl, m + R ijlm, k + R ijmk ,l = 0 ...(4)

Multiplying R ijkl, m by g hi i.e.,


g hi R ijkl, m = Rhjkl, m
Then equation (4) becomes
Rhjkl, m + R hjlm, k + Rhjmk ,l = 0 ...(5)
Since every term of equation (4) and (5) is a tensor. So, equation (4) and (5) are tensor equations
and therefore hold in every coordinate system. Further, P0 is an arbitrary point of Vn . Thus there hold
throughout Vn . Hence equation (4) or (5) is called Bianchi identity..

5.6 EINSTEIN TENSOR


1 i
Theorem 5.4 To prove the tensor R j − δ j R is divergence free.
i
2
Proof: We know that from equation (5)
96 Tensors and Their Applications

Rhjkl, m + R hjlm, k + R hjmk,l = 0

Multiply it by g hl g jk , we get
g hl g jk R hjkl,m + g hl g jk R hjlm, k + g hl g jk Rhjmk, l = 0

g jk R jk, m + g hl g jk (−R hjml, k ) + g hl g jk (− R jhmk, l ) = 0

Since Rhjlm = −R hjml & Rhjmk = − R jhmk

g jk R jk, m − g jk R jm , k − g hl R hm, l = 0

R , m − Rmk , k − R ml ,l = 0 Since g jk R jk = R

R , m − R mk , k − Rmk , k = 0

R , m − 2 Rmk , k = 0

1
Rmk , k − R, m = 0
2
1
Rmk , k − δ km R , k = 0 since R , m = δkm R, k
2
 k 1 k 
 Rm − δ m R  = 0
 2 , k
1
Rmk − δ km R is divergence free.
2
1 1 i
The tensor Rmk − δ km R = Gmk or R j − δ j R is known as Einstein Tensor..
i

2 2

5.7 RIEMANN CURVATURE OF A V n


Consider two unit vectors pi and qi at a point P 0 of V n. These vectors at P 0 determine a pencil of
directions deferred by ti = αpi + βqi. α and β being parameters. One and only one geodesic will pass
through P 0 in the direction of p i . Similarly one and only one geodesic will pass through in the direction
qi. These two geodesics through P 0 determined by the orientation of the unit vectors pi and qi. Let this
surface is denoted by S.
The Gaussian curvature of S at P 0 is defined to be the Riemannian Curvature of Vn at P0 for the
orientation determined by pi and qi.
Let the coordinates y i of Vn are Riemannian coordinates with origin at P 0. The equation of
surfaces S in given by
y i = ( p i α + q iβ)s ...(1)
i.e., y i = p iu 1 + q i u 2 ...(2)
Riemann-Christoffel Tensor 97

where αs = u1 and βs = u 2 , three parameters namely α, β, s can be reduced to two parameters


1
u and u 2 . Here u1 and u 2 are coordinates of any current point on S.
Let ds 2 = bαβ du α du β be the metric for the surface S. where

∂y i ∂ y j
bαβ = g ij (α, β = 1, 2)
∂uα ∂uβ

 k  γ 
Let   and   be the Christoffel symbols corresponding to the coordinates y i and u α .
 g
i j  α β 
Let αβγδ and R hijk be curvature tensor corresponding to the metrices b αβ du αdu β and g ij dy i dy j .
R
Since the Greek letters α, β, γ, δ take values 1,2 and so that the number of independent non-
1
vanishing components of Rαβγδ is n2 (n 2 − 1) for n = 2, i.e., they are
12
1 2 2
12
(
⋅ 2 2 − 1 = 1. Let us )
transform the coordinate system u α to u ′ α and suppose that the corresponding value of R1212 are
'
R1212 .
Then

∂u α ∂u β ∂u γ ∂u δ
'
R1212 = Rαβγδ
∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2

∂u1 ∂u β ∂u γ ∂u δ ∂u 2 ∂u β ∂u γ ∂u δ
= R1βγδ + R βγδ
∂u ′1 ∂u′ 2 ∂u ′1 ∂u′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2
2

∂u 1 ∂u 2 ∂u γ ∂u δ ∂u 2 ∂u 1 ∂u γ ∂u δ
= R12 γδ + R 21γδ
∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2

∂u1 ∂u 2 ∂u1 ∂u 2 ∂u1 ∂u 2 ∂u 2 ∂u1


= R1212 + R
∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2
1221

∂u 2 ∂u1 ∂u1 ∂u 2 ∂u 2 ∂u1 ∂u 2 ∂u 1


+ R 2112 + R
∂u11 ∂u ′ 2 ∂u′ 2 ∂u ′ 2 ∂u ′1 ∂u′ 2 ∂u ′1 ∂u′ 2
2121

 ∂u 1 ∂u 2 ∂u1 ∂u 2 ∂u 2 ∂u ′  ∂u 2 ∂u1  
2

R  − 2 +
= 2112  ∂u′1 ∂u′ 2 ∂u ′1 ∂u ′ 2 ∂u′ 2 ∂u′ 2  ∂u ′1 ∂u ′ 2  
 

∂u1 ∂u1
∂u ′1 ∂u ′ 2 ∂u
= R1212 J 2 where J = =
∂u 2 ∂u 2 ∂u′

so, ∂u ′1 ∂u ′ 2
′ = R1212 J 2
R1212 ...(3)
98 Tensors and Their Applications

Again
∂u γ ∂u δ
b′αβ = bαβ
∂u ′ α ∂u ′β

∂u γ ∂u δ
⇒ b′αβ = bαβ
∂u′ α ∂u′ β

or b′ = bJ 2 ...(4)
from (3) and (4), we get

= 1212 = K (say)
R1212 R
...(5)
b′ b
This shows that the quantity K is an invariant for transformation of coordinates. The invariant K
is defined to be the Guasian curvature of S. Hence K is the Riemannian Curvature of S at P0 .
Since Riemannian Coordinates yi with the origin at P0 . We have as geodesic coordinates with the
pole at P0 .
Therefore

 k  γ 
  ,   = 0 at P0
i j  g α βb
Then
 ∂[ij , h]g ∂[ik , h] g 
Rhijk =  − +  at P0 .
∂y k ∂y j 
 
and

 ∂[βγ, α] b ∂[βδ, α] b 
Rαβγδ =  − +  at P0 .
 ∂u δ ∂u γ 

∂[21,1] b ∂[22,1] b
R1212 = − + at P0 ...(6)
∂u 2
∂u1
from (5) we get

1  ∂[21, 1] b ∂[22, 1]b 


K = b  − + 
∂u 1 
...(7)
 ∂u 2

This is required expression for Riemannian curvature at P0 .

5.8 FORMULA FOR RIEMANNIAN CURVATURE IN THE TERMS OF COVARIANT


CURVATURE TENSOR OF V n

 k γ 
Let   and   be the Christoffel symbols of second kind relative to the metrices bαβdu αdu β
i j  g α β 
Riemann-Christoffel Tensor 99

and g ij dy i dy j respectively. We have

∂y i ∂y j ∂y k
[αβ, γ]b = [ij, k ]g ...(8)
∂u α ∂u β ∂u γ

∂y i ∂y j ∂y k
[21, 1]b = [ij, k ]g
∂u 2 ∂u1 ∂u1
= q i p j p k [ij , k ] g using (2)
Now,
∂[21, 1] b ∂[ij, k ]g
= qi p j p k
∂u 2
∂u 2

i j k
∂[ij, k ]g ∂y h
= q p p
∂y h ∂u 2

i j k h
∂[ij, k ]g
= q p p q
∂y h
Interchanging h and k, we get

∂[21, 1] b i k j h
∂[ij, h ]g
= qq p p ...(9)
∂u 2 ∂y k
Similarly,

∂[22, 1] b i k j h
∂[ij, h ]g
= qq p p ...(10)
∂u 1
∂y k
Using (9) and (10), equation (6) becomes

 ∂[ij, h ]g ∂[ik , h]g 


R1212 = p q p q  − ∂y k + ∂y j
h i j k
 at P0
 
h i j k
R1212 = p q p q Rhijk at P0 ...(11)
Since

∂y i ∂y j
bαβ = g ij
∂u α ∂u β

∂y i ∂y j
b11 = g ij = g ij p i p j
∂u ∂u
1 1

b11 = g hj p h p j
100 Tensors and Their Applications

Similarly

b22 = g ij q i q j = g ik q i q k

b12 = g ij p i q j = g ji p j q i = g hk p h q k

b11 b12
b = b b = b11b22 − b12 b21
21 22

[
b = p h qi p j qk g hig ik − gij ghk ] ...(12)
Dividing (11) and (12), we get
R1212 p h q i q k Rhijk
K= b = ...(13)
p h qi p j q k ( gik g hj − g ij g hk )
This is formula for Riemannian Curvature of Vn at P0 determined by the orientation of Unit vectors p i
and q i at P0 .

5.9 SCHUR’S THEOREM


If at each point, the Riemannian curvature of a space is independent of the orientation choosen then it
is constant throughout the space.
Proof: If K is the Riemannian curvature of Vn at P for the orientation determined by unit vectors p i
and q i then it is given by

p h q i p j q k R hijk
K= ...(1)
(g ik g hj − g ij g hk ) p h qi p j q k
Let K be independent of the orientation choosen. Then equation (1) becomes
R hijk
K=
g ik g hj − gij g hk
Rhijk = K (g ik g hj − g ij g hk ) ...(2)
We have to prove that K is constant throughout the space Vn .
If N = 2, the orientation is the same at every point. So, consider the case of Vn when n > 2.

Since g ij are constants with respect to covariant differentation, therefore covariant differentation
of (2) gives
Rhijk ,l = ( g ik g hj − g ij g hk ) K ,l ...(3)
where K, l is the partial derivative of K.
Taking the sum of (3) and two similar equations obtained by cyclic permutation of the suffices,
j, k and l.
Riemann-Christoffel Tensor 101

( g hj g ik − g hk g ij ) K ,i +( g hk g il − g hl g ik )K , j +( g hl g ij − g hj g il )K , k = Rhijk, l + Rhikl, j + Rhilj,k ...(4)

Here n > 2 therefore three or more distinct values to indices j, k, m can be given .
Multiplying (4) by g hj and using g hj g hl = δ lj , we get

(ng ik − δ ih g hk )k , l +(g il δ kj − gik δ l j ) K , j +(δ ih g h l − ngil )K , k = 0


or,
(n − 1) g ik K , l + 0 + (1 − n) gil K , k = 0 for δ ij = 0, i ≠ j
g ik K ,l − g il K , k = 0
Multiplying by g ik and using

g ik δik = n, g il g ik = δ kl ,
we get
nK,l −δ kl K ,k = 0 or (n − 1)K ,l = 0
or K, k = 0 as (n − 1)K ,l = 0
∂K
or = 0.
∂x l
integrating it, we get K = constant. This proves that the partial derivatives of K w.r.t. to x’s are all zero.
Consequently K is constant at P. But P is an arbitrary point of Vn . Hence K is constant throughout Vn .

5.10 MEAN CURVATURE


The sum of mean curvatures of a Vn for a mutually orthogonal directions at a point, is independent of
the ennuple choosen. Obtain the value of this sum.
Or
Prove that the mean curvature (or Riccian Curvature) in the direction ei at a point of a Vn is the
sum of n – 1 Riemmanian curvatures along the direction pairs consisting of the direction and n – 1
other directions forming with this directions an orthogonal frame.
i
Proof: Let e hi be the components of unit vector in a given direction at a point P of a Vn . Let e k be the
components of unit vector forming an orthogonal ennuple.
Let the Riemannian curvature at P of Vn for the orientation determined by l hi and e ik (h ≠ k ) be
denoted by K hk and given by
ehp e qk λ rk λ sk R pqrs
K hk =
ehp e qk e rh eks ( g pr q qs − g ps g qr )

e hp1 e qk1 e rh1 e ks1 R pqrs


= (e p e r g ) (e q e s g ) − (e p e s g ) (e q e r g ) ...(1)
h h pr k k qs h k ps k h qr
102 Tensors and Their Applications

Since unit vectors e hi , λ ik are orthogonal. Therefore


e hp e hr g pr = 1
p s
and e h eh g ps = 0 etc.
Using these in equation (1), then equation (1) becomes
e hp ekq e rh e ks R pqrs
K hk =
1×1 − 0 × 0
K hk = e hp e kq ehr e ks R pqrs ...(2)
n n

∑K
k =1
hk = ∑e p q r s
e e e
h k h k
R pqrs
k =1
n

Put ∑K
k =1
hk = M h . Then

n
p r
M h = eh eh ∑e
k =1
q s
e
k k
R pqrs

p r qs
= e h e h g R pqrs

= − e h e h g R qprs
p r qs

M h = − eh e h R pr
p r
...(3)
This shows that M h is independent of (n – 1) orthogonal direction choosen to complete an orthogonal
ennuple. Here M h is defined as mean curvature or Riccian curvature of Vn for the direction e hp1 .
Summing the equation (1) from h = 1 to h = n, we get
n

∑M
h =1
h = − eh |e h| R pr
p r

= − g pr R pr
= –R
n

or ∑M
h =1
h = − g pr R pr == −R
This proves that the sum of mean curvatures for n mutual orthogonal directions is independent of the
directions chosen to complete an orthogonal ennuple and has the value − R.

5.11 RICCI’S PRINCIPAL DIRECTIONS


Let e ih1 is not a unit vector and the mean curvature M h is given by
Rij e hi e hj
Mh =
g ij e hi ehj
Riemann-Christoffel Tensor 103

⇒ ( Rij + M h g ij ) e ih ehj = 0

Differentiating it w.r. to e ih1 , we get

∂M h
g jk ehj e hk + 2 ( Rij + Mg ij ) e hj
∂ ehi = 0 ...(1)

For maximum and minimum value of M h .


∂M h
=0
∂e ih
Then equation (1) becomes
( Rij + Mg ij )ehi = 0
These are called Ricci's Principal direction of the space as they are principal directions of Ricci
tensor Rij .

5.12 EINSTEIN SPACE


A space, which is homogeneous relative to the Ricci tensor Rij is called Einstein space.
If space is homogeneous then we have
Rij = λg ij ...(1)

Inner multiplication by g ij , we get

R = λn since Rij g ij = R and g ij gij = n


1
⇒ λ = nR
from (1)
R
Rij = g
n ij
R
Hence a space is an Einstein space if Rij = g at every point of the space.
n ij

Theorem 5.5 To show that a space of constant Curvature is an Einstein space.


Proof: Let the Riemannian curvature K at P of Vn for the orientation determined by p i and q i , is
given by

p h q i p j q k R hijk
K = p hq i p j q k ( g g − g g )
ik hj ij hk

Since K is constant and independent of the orientaion.


104 Tensors and Their Applications

Rhijk
K = (g g − g g )
ik hj ij hk

Rhijk = K (g ik g hj − g ij g hk )

Multiplying by g hk

Kg hk ( g ik g hj − g ij g hk ) = g hk Rhijk

K (δhi g hj − ng ij ) = Rij

Since g hk g ik = δ hi ; g hk g hk = n and g hk Rhi jk = Rij

K (g ij − ng ij ) = Rij

K (1 − n) g ij = Rij ...(1)
ij
Multiplying by g , we get
Kn(1 − n) = R as g ij Rij = R ...(2)
from (1) & (2)
R 1
Rij = (1 − n) gij ⋅ = Rgij
n(1 − n ) n
R
⇒ g
Rij =
n ij
This is necessary and sufficient condition for the space Vn to be Einstein space.

5.13 WEYL TENSOR OR PROJECTIVE CURVATURE TENSOR


Weyl Tensor denoted as Whijk and defined by
1
Whijk = Rhijk + ( g R − g kh g ij )
1 − n ki hj

Theorem 5.6 A necessary and sufficient condition for a Riemannian V n (n > 3) to be of constant
curvature to that the Weyl tensor vanishes identically throughout Vn .

Proof: Necessary Condition:


Let K be Riemannian Curvature of Vn . Let K = constant.
We have to prove that Whijk = 0
Since we know that

p h q i p j q k R hijk
K= = constant
(g hj g ik − g ij g hk ) p h qi p j q k
Riemann-Christoffel Tensor 105

Since K be independent of the orientation determined by the vector p i and q i .


Then
R hijk
K= ...(1)
g hj gik − gij g hk

Multiplying by g hk , we get

g hk Rhijk = g hk K (g hj g ik − g ij g hk )

Rij = K (δkj gik − ng ij )

Rij = K (1 − n) g ij ...(2)

Multiplying by g ij again, we get

g ij Rij = K (1 − n) g ij g ij

R = K (1 − n )n ...(3)
Putting the value of K from (3) in (2), we get
R
Rij = g ...(4)
n ij
The equation (3) shows that R is constant since K is constant.
Now, the W tensor is given by
1
Whijk = Rhijk + [ g R − g hk g ij ]
1 − n ik hj
from (5), we get
1  R R 
Whijk = Rhijk +  g ik g hj − g hk g ij 
1−n  n n 
R
= Rhijk + n(1 − n) [ gik g hj − g hk gij ]

R Rhijk
= Rhijk + ⋅ , by. eqn. (1)
n(1 − n) K

Rhijk
Whijk = Rhijk + K , by equation (3)
K
Whijk = 2 R hijk

Since K is constant. The equation (5) shows that Whijk = 0 .


This proves necessary condition.
106 Tensors and Their Applications

Sufficient Condition
Let Whijk = 0. Then we have to prove that K is constant.
Now, Whijk = 0.
1
⇒ Rhijk + [ g R − g hk Rij ] = 0
1 − n ik hj
Multiplying by g hk , we get
1
Rij + [g hk g ik Rhj − g hk g hk Rij ] =0
1− n
1
Rij + [δh R − nRij ] =0
1 − n i hj
R
Rij + ij (1 − n) =0
1− n
⇒ 2 R ij =0
⇒ Rij = 0

Since Rij = 0 ⇒ g hk Rhijk = 0.


⇒ g = 0 or Rhijk = 0
hk

If Rhijk = 0, then clearly K = 0. So, K is constant.


If g hk = 0 then

R hijk p h qi p j q k
K=
(g hj g ik − 0 g ij ) p h qi p j q k

R hijk p h q i p j q k R hijk p h q i p j q k
K= =
( p h p j g hj ) (q i q k g ik ) p2 ⋅ q 2

h i j k
K = Rhijk p q p q since p 2 = 1, q 2 = 1
K = constant as Rij = 0
This proves sufficient condition.

EXAMPLE 1
For a V2 referred to an orthogonal system of parametric curves (g 12 = 0) show that
R12 = 0, R11 g 22 = g 22 g11 = R1221

2R1221
R = g Rij =
ij

g11 g 22
Consequently
1
Rij = Rg .
2 ij
Riemann-Christoffel Tensor 107

Solution

Given that g12 = 0 so that g 12 = 0.


Also,
1 1 1
& g =g .
22
g ij = ⇒ g 11 =
g ij g11 22

The metric of V2 is given by


i j
ds 2 = g ij dx dx ; (i, j = 1, 2)

ds 2 = g11 (dx ) + g 22 (dx ) Since g12 = 0.


1 2 2 2

We know that g1hk Rhijk = Rij .

g11 g12 g 0
and g = g ij = = 11 = g11 g 22
g 21 g 22 0 g 22
(i) To prove R12 = 0
R12 = g Rh12 k = g Rh 2 as Rh 122 = 0
hk h1

= g 21 R2121 as R1121 = 0
R12 =0
(ii) To prove R11 g 22 = R22 g11 = R1221
R11 = g hk Rh 11k = g 2 k R 211k
R 2112
R11 = g R 2112 = g
22

22
So,
R 2112
R11 = ...(1)
g 22
and R22 = g hk Rh 22k = g 1k R122k
R1221
= g R1221 =
11

g11
R1221
So, R22 = ...(2)
g11
from (1) and (2)
R 11g22 = R 1221=R 22g11 … (3)
(iii) To prove
2R1221
R=
g11 g 22

R = g Rij = g Ri 1 + g Ri 2
ij i1 i2
108 Tensors and Their Applications

= g 11 R11 + g 22 R22 for g 12 = 0


R 11 R 22
= +
g 11 g 22
R1221
= g g +g g
R1221
[Q by eqn. (3)]
22 11 11 22
2R1221
R=
g11 g 22
1
(iv) To prove Rij = Rg
2 ij
2R1221
R=
g11 g 22
2R1221
R= as g = g11 g 22
g
1
R1221 = Rg
2
The eqn (3) expressed as
1
R11 g 22 = Rg = R 22 g11 .
2
it becomes
Rg Rg g Rg
R11 = = 11 22 = 11
2 g 22 2g 22 2
Rg Rg11 g 22 Rg22
R22 = = =
2 g11 2g11 2
So,
1 1
R11 = Rg11 & R22 = g 22
2 2
1
R12 = g12 as R12 = 0 = g12
2
1
This prove that Rij = R .
2 ij
EXAMPLE 2
The metric of the V 2 formed by the surface of a sphere of radius r is ds 2 = r 2 dθ 2 + r 2 sin 2 θdφ 2
1
in spherical polar coordinates. Show that the surface of a sphere is a surface of constant curvature 2 .
r
Solution
Given that
ds 2 = r 2 d θ 2 + r 2 sin 2 θd φ 2
Since r is radius of curvature then r is constant.
Riemann-Christoffel Tensor 109

g11 = r 2 , g 22 = r 2 sin 2 θ, g12 = 0, g = g11 g 22 = r 4 sin 2 θ.

We can prove
R 1221 = r2 sin 2 θ
Now, the Riemannian curvature K of V n is given by

p hq i p jq k
K= h i j k .
p q p q ( g hj gik − g hk gij )

At any point of V2 there exists only two independent vectors.


Consider two vectors whose components are (1, 0) and (0, 1) respectively in V2 . Then
R 1212 R
K= = 1212
g11 g 22 g

r 2 sin 2 θ 1
K= = = constant.
r 4 sin 2 θ r 2

EXERCISES

1. Show that
∂[ jl, i] ∂[ jk , i]  α   α 
Rijkl = − +   [il , α ] −   [ik , α]
∂xk ∂xl  j k   j l
2. Show that

1  ∂ g il 
2 2
2
∂ g jk ∂ 2 g ik ∂ g jl 
 + g ( [ jk , β][il , α] − [ jl , β][ik , α ] ) .
Rijkl = + − − αβ
2  ∂x j ∂xk ∂xi ∂xl ∂ x j ∂xl ∂xi ∂ xk
 
3. Using the formula of the problem 2. Show that
Rijkl = − R jikl = − Rijlk = Rklij and Rijkl + Riklj + R iljk = 0

4. Show that the curvature tensor of a four dimensional Riemannian space has at the most 20 distinct
non-vanishing components.
h
5. (a) If prove that the process of contraction applied to the tensor Rijk generates only one new tensor
Rij which is symmetric in i and j.

1
(b) If ds2 = g11 ( dx1 ) 2 + g 22 (dx 2 )2 + g 33 ( dx3 )2 . Prove that Rij =
Rihhj , (h, i, j being unequal).
g hh
6. Show that when in aV3 the coordinates can be chosen so that the components of a tensor g ij are zero
when i, j, k are unequal then
1
(i) Rhj = Rhiij
g ii
110 Tensors and Their Applications

1 1
(ii) Rhh = Rhiih + Rhjjh
g ii g jj
7. Prove that if
α 1 ∂R
Ri = g Rij then Ri , α = 2 i
α αj
∂x
and hence deduce that when n > 2 then scalar curvature of an Einstein space is constant.
8. If the Riemannian curvature K of Vn at every point of a neighbourhood U of Vn is independent of the
direction chosen, show that K is constant throughout the neighbourhood U. Provided n > 2.
9. Show that a space of constant curvature K0 is an Einstein space and that R = K 0 n (1 –n ).
10. Show that the necessary and sufficient condition that Vn be locally flat in the neighbourhood of 0 is
that Riemannian Christoffel tensor is zero.
11. Show that every V2 is an Einstein space.
12. For two dimensional manifold prove that
R
K= −
2
13. Show that if Riemann-Christoffel curvature tensor vanishes then order of covariant differentiation is
commutative.
CHAPTER – 6

THE e-SYSTEMS AND THE GENERALIZED


KRÖNECKER DELTAS

The concept of symmetry and skew-symmetry with respect to pairs of indices can be extended to
cover to pairs of indices can be extended to cover the sets of quantities that are symmetric or skew-
symmetric with respect to more than two indices. Now, consider the sets of quantities A i ... i k or
Ai l ... i k depending on k indices written as subscripts or superscripts, although the quantities A may not
represent tensor.

6.1 COMPLETELY SYMMETRIC


The system of quantities A i1 ... i k (or Ai1... i k ) depending on k indices, is said to be completely symmetric
if the value of the symbol A is unchanged by any permutation of the indices.

6.2 COMPLETELY SKEW-SYMMETRIC


The systems Ai1 ... ik or ( Ai1... i k ) depending on k indices, is said to be completely skew-symmetric if the
value of the symbol A is unchanged by any even permutation of the indices and A merely changes the
sign after an odd permutation of the indices.
Any permutation of n distinct objects say a permutation of n distinct integers, can be accomplished
by a finite number of interchanges of pairs of these objects and that the number of interchanges
required to bring about a given permutation form a perscribed order is always even or always odd.
In any skew-symmetric system, the term containing two like indices is necessarily zero. Thus if
one has a skew-symmetric system of quantities Aijk where i, j, k assume value 1, 2, 3. Then
A122 = A112 = 0
A123 = − A213 , A312 = A123 etc.

In general, the components Aijk of a skew-symmetric system satisfy the relations.

Aijk = − Aikj = − A jik

Aijk = A jki = Akij


112 Tensors and Their Applications

6.3 e-SYSTEM

Consider a skew-symmetric system of quantities ei1 ... i n ( or ei1 ,..., i n ) in which the indices i1 ... in assume

values 1,2,...n. The system ei1 ... i n ( or ei1... i n ) is said to be the e-system if

= +1; when i1 , i2 ,..., in



 an even permutation of number1, 2, ..., n


ei1 ... in (or ei1 ... in ) = −1; when i1 , i2 ,..., in

 an odd permutation of number1, 2, ..., n

= 0 in all other cases

EXAMPLE 1
Find the components of system eij when i, j takes the value 1,2.

Solution
The components of system eij are
e11 , e12 , e21 , e 22 .
By definition of e-system, we have
e11 = 0, indices are same
e12 = 1, since i j has even permutation of 12
e 21 = − e12 = −1 since i j has odd permutation of 12
e22 = 0, indices are same

EXAMPLE 2

Find the components of the system ei jk .

Solution
By the definition of e-system,
e123 = e 231 = e321 = 1
e213 = e132 = e321 = −1
eijk = 0 if any two indices are same.

6.4 GENERALISED KRÖNECKER DELTA

A symbol δi1j1...... ikjk depending on k superscripts and k subscripts each of which take values from 1 to n,
is called a generalised Krönecker delta provided that
(a) it is completely skew-symmetric in superscripts and subscripts
The e-Systems and the Generalized Krönecker Deltas 113

(b) if the superscripts are distinct from each other and the subscripts are the same set of
numbers as the superscripts.
The value of symbol
 = 1; an even number of transposition is required to arrange the
 superscripts in the same order as subscripts.

δ j ... j  = −1; where odd number of transpositions arrange the superscripts
i1 ... i k
1 k

 in the same order as subscripts
 = 0, in all other cases the value of the symbolis zero

EXAMPLE 3
Find the values of δ ijkl .
Solution

By definition of generalised Kronecker Delta, δ ijkl = 0 if i = j or k = l or if the set. ij is not the set kl.

δ11
pq = δ pq = δ13 = ⋅ ⋅ ⋅ = 0
22 23
i.e.,
δ ijkl = 1 if kl is an even permutation of ij
i.e., δ12
12
= δ 21
21
= δ13
13
= δ31
31
= δ 23
23
= ⋅⋅⋅ =1

and δ ijkl = −1 if kl is an odd permutation of ij.


i.e., δ12
21
= δ13
31
= δ13
31 = δ12 = ⋅ ⋅ ⋅ = − 1
21

i i ... i
Theorem 6.1 To prove that the direct product e 1 2 n e j1 j2 ... jn of two systems e i1 ... in and e j1 j 2 ... jn is the
generalized Krönecker delta.
i i ... i n
Proof: By definition of generalized Krönecker delta, the product e 1 2 ej
1 j2 ... jn
has the following values.
(i) Zero if two or more subscripts or superscripts are same.
(ii) +1, if the difference in the number of transpositions of i1 , i2 ,...,in and j1 , j2 ,..., jn from
1,2,...n is an even number.
(iii) –1, if the difference in the number of transpositions of i1, i2,...,in and j1, j2, ...jn from 1, 2,..n
an odd number.
Thus we can write

= δ 1j1 2j 2 ... nj n
i i ... i
e i 1i 2 ... i n e j1 j 2 ... jn

THEOREM 6.2 To prove that

(i) e i 1 i 2 ... i n = δ 1j1 2j 2 ... nj n


i i ... i

(ii) ei1i 2 ... i n = δ 1j12j 2 ...n jn


i i ... i
114 Tensors and Their Applications

Proof: By Definition of e-system, e i 1 i 2 ... i n ( or e i1 i 2 ... i n ) has the following values.

(i) +1; if i1 , i2 ,...,in is an even permutation of numbers 1,2,...n.


(ii) -1; if i1 , i2 ,...,in is an odd permutation of numbers 1,2,...n
(iii) 0; in all other cases
Hence by Definition of generalized krönecker delta, we can write
(1) e 1 2 n = δ11 22... nn
i i ... i i i ... i

and
(2) ei1i2 ... in = δi1i 2 ... in
1 2 ... n

6.5 CONTRACTION OF ä iáâãjk

Let us contract δiαβγ on k and γ . For n = 3, the result is


jk

δ ijk
αβγ = δ αβ1 + δ αβ2 + δαβ 3 = δαβ
ij1 ij 2 ij 3 ij

This expression vanishes if i and j are equal or if α and β are equal.

If i = 1, and j = 2, we get δ123


α β 3.

Hence

+ 1; if αβis an even permutatio n of 12



− 1; if αβ is an odd permutatio n of 12
δ12
αβ = 
 0; if αβ is not permutatio n of 12

Similarly results hold for all values of α and β selected from the set of numbers 1, 2, 3.
Hence

+ 1; if i j is an even permutatio n of αβ

− 1; if i j is an odd permutatio n of αβ

δ αβ =  0; if two of the subscripts or superscripts are equal or when the
ij


 subscripts and superscripts are not formed from the same numbers.

1
If we contract δ ijαβ . To contract δ ijαβ first contract it and the multiply the result by . We obtain
2
a system depending on two indices
1 ij 1 i 1
δ iα = 2 δαj = 2 (δ α1 + δ α2 + δα 3 )
i2 i3
The e-Systems and the Generalized Krönecker Deltas 115

1 12
It i = 1 in δiα then we get δ α = δα 2 + δ α3
1
2
13
( )
This vanishes unless α = 1 and if α = 1 then δ11 = 1.

Similar result can be obtained by setting i = 2 or i = 3. Thus δ iα has the values.

(i) 0 if i ≠ α, (α, i = 1, 2, 3, )
(ii) 1 if i = α .
By counting the number of terms appearing in the sums. In general we have
1 ij
δ iα = n −1 δ αj and δ ij = n(n − 1)
ij
...(1)
We can also deduce that
(n − k )! i1i2 ... ir i r −1... i k
δi1j1i2j...2 ...i r jr = (n − r )! δ j1 j2 ... j r j r −1... j k ...(2)

and
n!
δi1j1i2j...2 ...i r jr = n (n −1) (n − 2) ⋅ ⋅ ⋅ (n − r + 1) = ...(3)
n − r!
or

e i1 i 2 ... i n e i1 i 2 ... i n = n! ...(4)


and from (2) we deduce the relation

e i1i 2 ... iri r+1... i n e j1 j2 ... j r jr +1 ... i n = n! ...(5)

EXERCISE

1. Expand for n = 3
i α 12 i j αβ i j ij
(a) δ αδ j (b) δ ij x x (c) δ ij x y (d) δij
2. Expand for n = 2
(c) e aα aβ = e a .
ij 1 2 ij 2 1 ij i j ij
(a) e ai a j (b) e ai a j

3. Show that δ ijk = 3 ! if i, j, k = 1, 2, 3.


ijk

4. If a set of quantities Ai1i 2 ... ik is skew-symmetric in the subscripts (K in number) then


i ... i
δ 1j1... jk Ai1 ... ik = k! A j1 ... j k
k

5. Prove that εijk g is a covariant tensor of rank three where where ε ijk is the usual permutation
symbol.
CHAPTER – 7

GEOMETRY

7.1 LENGTH OF ARC


Consider the n-dimensional space R be covered by a coordinate system X and a curve C so that
C : x i = x (t ), (i = 1,2..., n)
i
...(1)
which is one-dimensional subspace of R. Where t is a real parameter varying continuously in the
interval t1 ≤ t ≤ t 2 . The one dimensional manifold C is called arc of a curve.
 1 2 n dx dx
1 2
dx n 
Let F  x , x ,..., x , dt dt ,..., dt  be a continuous function in the interval t1 ≤ t ≤ t 2 . Wee
 
 dx  i
assume that F  x,  > 0, unless every dx = 0 and that for every positive number k
 dt  dt
 dx 1
dx 2
dx n   1 2 1 2 n 
F  x1 , x 2 ,..., x n , k ,k ,..., k kF  x , x ,..., x n , dx , dx ,... dx  .
dt   dt 
=
 dt dt  dt dt

The integral
 dx 

t2
s= F  x,  dt ...(2)
t1  dt 
is called the length of C and the space R is said to be metrized by equation (2).

Different choices of functions F  x,  lead to different metric geometrices.


dx
 dt 
If one chooses to define the length of arc by the formula
dx p dx q

t2
s= g pq ( x)dt, (p,q = 1, 2, ..., n) ...(3)
t1 dt dt
dx p dxq dx p
where g pq ( x) is a positive definite quadratic form in the variable , then the resulting
dt dt dt
geometry is the Riemannian geometry and space R metrized in this way is the Riemannian n-dimensional
space R n.
Geometry 117

Consider the coordinate transformation T : x i = x i ( x1 ,..., x n ) such that the square of the element
of arc ds,
p q
ds 2 = g pq dx dx ...(4)
can be reduced to the form
ds 2 = dx i dx i ...(5)
Then the Riemannian manifold R n is said to reduce to an n-dimensional Euclidean manifold E n.
The Y-coordinate system in which the element of arc of C in E n is given by the equation (5) is
called an orthogonal cartesian coordinate system. Obviously, E n is a generalization of the Euclidean
plane determined by the totality of pairs of real values ( x 1 , x 2 ). If these values ( x 1 , x 2 ) are associated
with the points of the plane referred to a pair of orthogonal Cartesian axes then the square of the
element of arc ds assumes the familiar form
ds 2 = (d x1)2 + (d x 2 )2 .
 dx   dx   dx 
THEOREM 7.1 A function F  x,  satisfying the condition F  x, k  = kF  x ,  for every
 dt   dt   dt 
k > 0. This condition is both necessary and sufficient to ensure independence of the value of the
 dx 

t2
integral s = t F  x,  dt of a particular mode of parametrization of C. Thus if t in C : x i = xi (t)
1  dt 
is replaced by some function t = φ(s ) and we denote x i [φ(s )] by ξ i (λ) . so that x i (t ) ξ i (s ) we have
equality
 dx 

t2


s2
F  x ,  dt = F (ξ, ξ′) ds
t1  dt  s1

ds i
where ξ′ =
i
and t1 = φ (s1 ) and t2 = φ (s 2 ).
ds
Proof: Suppose that k is an arbitary positive number and put t = ks so that t1=ks 1 and t2=ks 2. Then
i
C : x i = x (t ) becomes
C : x i (ks) = ξ i (t )
dx i (ks ) dx i (ks)
and ξ′ i (s ) = =k
ds dt
 dx 

t2
Substituting these values in s= F  x ,  dt we get
t1  dt 
 dx (ks )
∫ F  x (ks),
s2
kds
dt 
s= s1 

F [ξ(s ),ξ′( s )]ds



s2
or s= s1

 dx   dx 
We must have the relation F (ξ, ξ′ ) = F  x, k  = kF  x, . Conversely, if this relation is true
 dt   dt 
118 Tensors and Their Applications

for every line element of C and each k > 0 then the equality of integrals is assumed for every choice of
parameter t = φ ( s), φ1 (s ) > 0, s1 ≤ s ≤ s 2 with and t2 = φ (s 2 ).
i
dx
Note: (i) Here take those curves for which x i (t ) and are continuous functions in t1 ≤ t ≤ t 2.
dt
 dx   dx   dx 
(ii) A function F  x,  satisfying the condition F  x, k  = kF  x,  for every k > 0 is called
 dt   dt   dt 
i
dx
positively homogeneous of degree 1 in the .
dt

EXAMPLE 1
What is meant, consider a sphere S of radius a, immersed in a three-dimensional Euclidean
manifold E 3 , with centre at the origin (0, 0, 0) of the set of orthogonal cartesian axes O − X 1 X 2 X 3 .
Solution
Let T be a plane tangent to S at (0, 0,–a) and the points of this plane be referred to a set of orthogonal
cartesian axes O′ –Y 1Y 2 as shown in figure. If we draw from O (0, 0, 0) a radial line OP , interesting
the sphere S at P ( x1 , x 2 , x 3 ) and plane T at Q( x 1 , x 2 , −a) then the points P on the lower half of the
sphere S are in one-to-one correspondence with points ( x 1 , x 2 ) of the tangent plane T..
3
X

2
X
O
P2
P1
C 2
O´ Y
1
X
Q2
1 K
Y Q1

Fig. 7.1

If P ( x1 , x 2 , x 3 ) is any point on the radial line OP,, then symmetric equations of this line is
x1 − 0 x2 − 0 x3 − 0
= 2 = =λ
x1 − 0 x − 0 −a− 0
or
x1 = λx 1 , x 2 = λx 2 , x 3 = −λa ...(6)
Since the images Q of points P lying on S, the variables x i satisfy the equation of S,
( x1 )2 + (x 2 )2 + (x 3 ) 2 = a 2

or ( ) ( )
λ2  x 1 + x 2

2 2
+ a2  = a2

Geometry 119

Solving for λ and substituting in equation (6), we get

ax1 ax 2
x =
1 , x2 =
( x 1 ) 2 + ( x 2 )2 + a 2 ( x 1 )2 + (x 2 )2 + (x 3 ) 2

− a2
and x =
3 ...(7)
( x 1) 2 + ( x 2 ) 2 + ( x 3 ) 2
These are the equations giving the analytical one-to-one correspondence of the points Q on T and
points P on the portion of S under consideration.
Let P1 ( x1 , x 2 , x 3 ) and P2 ( x1 + dx1 , x 2 + dx 2 , x 3 + dx 3 ) be two close points on some curve C lying
on S. The Euclidean distance P1 P2 along C, is given by the formula
ds 2 = dx i dx i , (i = 1, 2, 3) ...(8)
Since
∂x i p
dx = ∂x p d x , (P =1, 2)
i

Thus equation (8) becomes


∂ xi ∂ xi
ds 2 = d x pd x q
∂x ∂ x
p q

= g pq ( x )d x pd x q , (p,q = 1, 2)
∂ xi ∂ x i .
where g pq (x ) are functions of x i and g pq =
∂ x p ∂x q
If the image K of C on T is given by the equations
 x 1 = x 1 ( t )
K:
 x 2 = x 2 (t ), t1 ≤ t ≤ t2

then the length of C can be computed from the integral


dx p dxq

t2
s= g pq dt
t1 dt dt

A straight forward calculation gives

1 1 2
(d x 1)2 + (d x 2) 2 + (x d x − x 2 d x 1)2
a2

{ }
2
ds2 =  1 2 2  ...(9)
1 +
 a 2 ( x 1 2
) + ( x ) 
and
120 Tensors and Their Applications

2 2 2
 d x1   d x 2   2 1

  +  + 12  x1 d x − x 2 d x 

t 2  dt      dt 
 dt  a  dt
s= t ∫
1 1
{
1 + 2 (x ) + ( x )
1 2 2 2
}
dt

a
So, the resulting formulas refer to a two-dimensional manifold determined by the variables ( x 1 , x 2 ) in
the cartesian plane T and that the geometry of the surface of the sphere imbeded in a three-dimensional
Euclidean manifold can be visualized on a two-dimensional manifold R 2 with metric given by equation (9).
1
If the radius of S is very large then in equation (9) the terms involving 2 can be neglected. Then
a
equation (9) becomes
ds 2 = (d x ) + (d x ) .
1 2 2 2
...(10)
Thus for large values of a, metric properties of the sphere S are indistinguishable from those of
the Euclidean plane.
The chief point of this example is to indicate that the geometry of sphere imbedded in a Euclidean
3-space, with the element of arc in the form equation (8), is indistinguishable from the Riemannian
geometry of a two-dimensional manifold R 2 with metric (9). the latter manifold, although referred to
a cartesian coordinate system Y, is not Euclidean since equation (9) cannot be reduced by an admissible
transformation to equation (10).

7.2 CURVILINEAR COORDINATES IN E 3


Let P (x ) be the point, in an Euclidean 3-space E 3 , referred to a set of orthogonal Cartesian coordinates Y..
Consider a coordinate transformation
T : x i = x i (x 1 , x 2 , x 3 ), (i = 1, 2, 3)

∂ xi
Such that J = ∂ x j ≠ 0 in some region R of E 3 . The inverse coordinate transformation

T −1 : x i = x ( x , x , x ), (i = 1, 2, 3)
i 1 2 3

will be single values and the transformations T and T −1 establish one-to-one correspondence
between the sets of values ( x1 , x 2 , x 3 ) and ( x 1 , x 2 , x 3 ).
The triplets of numbers ( x1 , x 2 , x 3 ) is called curvilinear coordinates of the points P in R.
1 2 3
If one of the coordinates x , x , x is held fixed and the other two allowed to vary then the point
P traces out a surface, called coordinate surface.
If we set x1 = constant in T then

x1 ( x 1 , x 2 , x 3 ) = constant ...(1)
defines a surface. If constant is allowed to assume different values, we get a one-parameter family of
surfaces. Similarly, x 2 (x 1 , x 2 , x 3 ) = constant and x 3 ( x 1 , x 2 , x 3 ) = constant define two families of
surfaces.
The surfaces
Geometry 121

x1 = c1 , x = c2 , x = c3
2 3
… (2)

3 3
Y X
2
X

P 1
X

2
Y

1
Y

Fig. 7.2

intersect in one and only one point. The surfaces defined by equation (2) the coordinate surfaces
and intersection of coordinate surface pair-by pair are the coordinate lines. Thus the line of intersection
of x1 = c1 and x 2 = c2 is the x 3 − coordinate line because along this the line the variable x 3 is the only
one that is changing.

EXAMPLE 2
Consider a coordinate system defined by the transformation
3
Y

1
X
X2

2
O Y
3
X

Y1
Fig. 7.3

x 1 = x1 sin x 2 cos x 3 ...(3)

x 2 = x1 sin x 2 sin x 3 ...(4)


x 3 = x1 cos x 2 ...(5)
122 Tensors and Their Applications

The surfaces x1 = constant are spheres, x 2 = constant are circluar cones and x3 = constant are
planes passing through the Y 3-axis (Fig. 7.3).
The squaring and adding equations (3), (4) and (5) we get,
( x 1 )2 + (x 2 )2 + (x 3 ) 2 = ( x1 sin x 2 cos x 3 ) 2 + (x1 sin x 2 sin x 3 )2 + (x 1 cos x 2 ) 2
On solving

( x 1 )2 + (x 2 )2 + (x 3 ) 2 = (x1 )2

x1 = (x 1 ) 2 + ( x 2 ) 2 + ( x 3 )2 ...(6)
Now, squaring and adding equations (3) and (4), we get
( x 1 )2 + (x 2 )2 = ( x1 sin x 2 sin x 3 )2 + (x1 sin x 2 sin x 3 )2

( x 1 )2 + (x 2 )2 = ( x1 )2 (sin x 2 ) 2

x1 sin x 2 = (x 1 ) 2 + ( x 2 ) 2 ...(7)
Divide (7) and (5), we get
(x1 )2 + (x 2 )2
tan x 2 =
x3
 2 2
−1  ( x ) + ( x ) 
1 2
2 tan
or x =  x3  ...(8)
 
Divide (3) and (4), we get

x2
1 =
tan x 3
x

−1  x 
2
⇒ x 3 = tan  1 
 ...(9)
x 
So, the inverse transformation is given by the equations (6), (8) and (9).
If x1 > 0, 0 < x 2 < π, 0 ≤ x 3 < 2π. This is the familiar spherical coordinate system.

7.3 RECIPROCAL BASE SYSTEMS


Covariant and Contravariant Vectors
r r r
Let a cartesian coordinate system be determined by a set of orthogonal base vectors b1 , b2 ,b3 then the
position vector rr of any point P (x 1 , x 2 , x 3 ) can be expressed as
r r
r = bi x i (i = 1, 2, 3) ...(1)
Geometry 123

r
Since the base vectors bi are independent of the position of the point P (x 1 , x 2 , x 3 ) . Then from (1),
r r
dr = bi dx i ...(2)

If P (x 1 , x 2 , x 3 ) and Q (x 1 + dx 1 , x 2 + d x 2, x 3 + d x 3 ) be two closed point. The square of the


element of arc ds between two points is
r r
ds 2 = dr ⋅ dr
from equation (2),
r i r
ds 2 = bi dx ⋅ b j dx
j

r r
= bi ⋅ b j d x id x j
r r
ds 2 = δij d x d x ; since bi ⋅ b j = δ ij
i j

3 3
Y X
2
a3 a2 X

a1 1
X
P
r
b3

O b2 2
Y
b1

1
Y

Fig. 7.4.
r r r r r r r
(b1 , b2 , b3 areorthogonal base vector i.e.,b1 ⋅ b1 = 1 & b1 ⋅ b2 = 0) .
as äij = 1, i = j
ds 2 = d x id x i;
= 0, i ≠ j
a familiar expression for the square of element of arc in orthogonal cartesian coordinates.
Consider the coordinate transformation
i 1 2 3
x i = x (x , x , x ), (i = 1, 2, 3)
r
define a curvilinear coordinate system X. The position vector r is a function of coordinates x i .
i.e.,
r r
r = r (x i ), (i = 1, 2, 3)
Then
r
r ∂r i
dr = dx ...(3)
∂x i
124 Tensors and Their Applications

and
ds 2 = d rr ⋅ d rr
r r
∂r ∂r
= ⋅ d x id x j
∂ xi ∂ x j

ds 2 = g ij dx i dx j
where
r r
∂r ∂r
g ij = ⋅ ...(4)
∂ xi ∂ x j
r
∂r
The vector is a base vector directed tangentially to X i - coordinate curve.
∂x i
Put
r
∂r r
= ai ...(5)
∂x i

Then from (3) and (4)


r r r r
dr = ai d x i and g ij = a i ⋅ a j ...(6)
Now, from equations (2) and (6), we get
r r
a j dx j = bi d x i

r r ∂x i j
a j dx j = bi dx
∂x j

r r ∂ xi
⇒ a j = bi , as d x j are arbitrary
∂x j
r
So, the base vectors a j transform according to the law for transformation of components of
covariant vectors.
r
The components of base vectors a i , when referred to X-coordinate system, are
r r r
a1 : (a1 , 0,0), a 2 : (0, a2 , 0), a3 : (0,0, a3 ). ...(7)
and they are not necessarily unit vectors.
In general,
r r r r
g11 = ar1 ⋅ ar1 ≠ 1, g 22 = a 2 ⋅ a2 ≠ 1, g 33 = a3 ⋅ a3 ≠ 1. ...(8)
If the curvilinear coordinate system X is orthogonal. Then
r r r r
g ij = a i ⋅ a j = ai a j cos θij = 0, if i ≠ j. ...(9)
r r
Any vector A are can be written in the form A = k d rr where k is a scalar..
Geometry 125

r
r ∂r i
Since d r = d x we have
∂ xi
r
r ∂r i
A = ∂x i (kdx )
r r i
A = ai A
r
where Ai = kdxi . The numbers Ai are the contravariant components of the vector A
Consider three non-coplanar vectors
r r r r r r
r1 a 2 × a3 ar 2 = a3 × a1 , ar 3 = a1 × a2
a = rr r ,
[a1a 2a3 ] [ar1ar2 ar3 ] [ar1ar2ar3 ] ...(10)
r r r r rr r
where a 2 × a3 , etc. denote the vector product of a 2 and a3 and [a1 a2 a3 ] is the triple scalar
r r r
prduct a1 ⋅ a2 × a3 .
Now,
r r r rr r
a2 × a3 ⋅ a1 [a1 a2 a3 ]
r1 r = = 1.
a ⋅ a1 = [ar ar ar ] [ar1ar2ar3 ]
1 2 3

r r r r r r
r1 r a2 × a3 ⋅ a2 [a2 a2 a3 ]
a ⋅ a 2 = [ar ar ar ] = = 0.
1 2 3 [ar1ar2 ar3 ]
r r r
Since [a2 a2 a3 ] = 0.
Similarly,
r r r r
a 1 ⋅ a3 = a 2 ⋅ a1 = ⋅ ⋅ ⋅ = 0
r r r r
a 2 ⋅ a2 = 1, a 3 ⋅ a 3 = 1
Then we can write
r r
a i ⋅ a j = δij

EXAMPLE 3
rr r
To show that [a1 a2 a3 ] = [ ]
g and ar1ar 2 ar 3 = 1 where g = g ij .
g
Solution
The components of base vectors ai are
r r r
a1 : (a1 , 0,0), a 2 : (0, a2 , 0) and a3 : (0, 0, a3 )

Then
a1 0 0
rr r
[a1a2a3 ] = 0 a2 0 = a1a2a3 ...(11)
0 0 a3
126 Tensors and Their Applications

and

g11 g12 g13


g= g ij = g 21 g 22 g 23
g 31 g 32 g33

from equations (8) and (9), we have


g11 = ar1 ⋅ ar1 ⇒ a12 = g11
Similarly
a22 = g 22 , a 32 = g 33
and
g12 = ar1 ⋅ ar2 = 0, g13 = ar1 ⋅ ar3 = 0 etc.
So,

a12 0 0
g= 0 a22 0 = a12 a22 a32
0 0 a32

g = a1 a2 a3 ...(12)
from eqn. (11) and (12), we have
[ar1ar2ar3 ] = g
r r r 1
Since the triple products [a 1 a 2 a 3 ] = . Moreover,,
g
r r r r r r
r a2 × a3 r a 3 × a1 r a1 × a 2
a1 = r1 r 2 r 3 , a 2 = r r r , a 3 = r r r
[a a a ] [a1 a 2 a 3 ] [a1 a 2 a 3 ]
The system of vectors ar1 , ar 2 , ar3 is called the reciprocal base system.
r r r
Hence if the vectors a 1 , a 2 , a 3 are unit vectors associated with an orthogonal cartesian coordinates
then the reciprocal system of vector defines the same system of coordinates. Solved.
r r ri
The differential of a vector r in the reciprocal base system is d r = a d xi .
where dxi are the components of drr. Then
r r
ds 2 = dr ⋅ dr
r r
= (a i dxi ) ⋅ (a j dx j )
r r
= a i ⋅ a j dxi dx j
ij
ds 2 = g dxi dx j
r r
where g ij = a i ⋅ a j = g ji ...(13)
Geometry 127

The system of base vectors determined by equation (10) can be used to represent an arbitrary
r r r
vector A in the form A = a i Ai , where Ai are the covariant components of A.
r r
Taking scalar product of vector Ai a i with the base vector a j , we get
r r r r
Ai a i ⋅ a j = Ai δij = A j as a i ⋅ a j = δ ij .

7.4 ON THE MEANING OF COVARIANT DERIVATIVES


r
r ∂A r
THEOREM 7.2 If A is a vector along the curve in E 3 . Prove that = A,αj aα
∂x j

∂A i r
Also, prove that A,i j = j . Where Ai are component of A.
∂x
r r
Proof: A vector A can be expressed in the terms of base vectors ai as
r r
A = Ai ai
r
r ∂r r
where a i = i and Ai are components of A.
∂x
r
The partial derivative of A with respect to x j is
r r
∂A ∂Ai r i ∂ai
= a + A ...(1)
∂x j ∂x j
i
∂x j
r r
Since g ij = a i ⋅ a j .
Differentiating partially it w.r.t. x k , we have
r r
∂g ij ∂a i r ∂a j r
= ⋅a j + k ⋅ ai
∂x k ∂x k ∂x
Similarly,
r r
∂g jk ∂a j r ∂ak r
= ⋅ a + ⋅ aj
∂x i k
∂xri ∂xri
∂g ik ∂ai r ∂a r
and j =
⋅ ak + kj ⋅ ai
∂x ∂x j
∂x
r
Since A can be written as
r r i ri
A = a i A = a Ai
r
Taking scalar product with a j , we have
r r r r
a i ⋅ a j A i = a i ⋅ a j Ai

⇒ g ij A i = δ ij Ai = A j
r r r r
As ai ⋅ a j = g ij , a i ⋅ a j = δ ij and δ ij Ai = A j .
We see that the vector obtained by lowering the index in Ai is precisely the covariant vector Ai .
128 Tensors and Their Applications

r
The two sets of quantities Ai and Ai are represent the same vector A referred to two different
base systems.

EXAMPLE 4
Show that g iα g jα = δ ij .
Solution
Since we know that
g i α = ari ⋅ arα and g jα = ar j ⋅ ar α
Then
r r r r
g i α g jα = (ai ⋅ aα ) (a j ⋅ a α )
r r r r
= (ai ⋅ a j ) (aα ⋅ a α )
r r
= δ ij δ αα as ai ⋅ a j = δ ij

g i α g jα = δij as δ αα = 1.
But
r
r ∂r
ai =
r ∂x i r
r r
∂ai ∂ 2r ∂ 2r ∂a j
= = =
∂x j ∂x j dx i ∂x i ∂x j ∂xi
So,
r r
∂ai ∂a j
=
∂x j ∂x i
Now,
1  ∂g ik ∂g jk ∂g ij 
 + − k  , Christoffel’s symbol
2  ∂x j
[ij, k ] =
∂x i ∂x 
∂g ik ∂g jk ∂g
Substituting the value of , i and ijk , we get
∂x j
∂x ∂x
r
1 ∂a i r
[ij,k] = ⋅ 2 j ⋅ a k
2 ∂x
r
∂ai r
[ij,k] = ⋅ ak
∂x j
r
∂ai rk 1 rk
or j = [ ij, k ]a , as r = a
∂x ak
Hence
r
∂a i r α r r
⋅ a = [ij, k ]a k ⋅ a α
∂x j
Geometry 129

r r
= [ij, k ]g kα , Since g kα = a k ⋅ a α
r α 
∂ai r α α  [ i j , k ] g kα =  
. a =  , as
∂x j i j  i j 

r  α  r
∂ai
=   aα ...(2)
∂x j i j 
r
∂ai
Substituting the values of in equation (1), we get
∂x j
r
∂A ∂Ai r  α  i r
= ai +   A a α
∂x j ∂x j i j 
∂A r α
α  r
= a +   A i aα
j α
∂x i j 
r
∂A  ∂A α  α  i  r
=  j +   A  aα
∂x j  ∂x i j  
r
∂A α r α ∂A α  α  i
= A , a α since A = +  A
∂x j j ,j
∂x j i j 

Thus, the covariant derivative A,αj of the vector A α is a vector whose components are the
r
∂A r
components of j referred to the base system ai .
∂x
r
α  ∂ai
If the Christoffel symbols vanish identically i.e.,   = 0 the = 0, from (2).
i j  ∂x j
Substituting this value in equation (1), we get
r
∂A ∂Ai r
= ai
∂x j ∂x j
∂Ai  i  β
But A,i j = + A
∂x j β j 
∂Ai  i 
A,i j = as   = 0.
∂x j β j 
Proved.
r r
THEOREM 7.3 If A is a vector along the curve in E 3 . Prove that A j , k a j wheree A j are components
r
of A. .
r
Proof: If A can be expressed in the form
r r
A = Ai a i
130 Tensors and Their Applications

r
where Ai are components of A.
r
The partial derivative of A with respect to x k is
r r
∂A ∂Ai ri ∂a i
= a + Ai k ...(1)
∂x k ∂x k ∂x
r r
Since a i ⋅ a j = δij , we have,

Differentiating it partially w.r. to x k , we get


r r
∂a i r r i ∂a j
⋅ a + a ⋅ =0
∂x k
j
∂x k
r r
∂a i r r i ∂a j
⋅a j = − a ⋅ k
∂x k ∂x
r
r r α  ∂a j  α r
= − ai ⋅ aα  , Since =  a α
j k ∂x k
 j k
r r
But a i ⋅ aα = δiα . Then
r
∂a i r α 
⋅ a j = − δiα  
∂x k  j k
r
∂a i r  i 
⋅a j = −  
∂x k  j k
r 1 rj
∂a i  i r
= −  a j , as ar = a
∂x k  j k j

r
∂a i
substituting the value of in equation (1), we get
∂x k
r
∂A ∂Ai ri  i r j
a − Ai  a
k = ∂x k
∂x  j k
∂Ai r j  i rj
= a − Ai  a
∂x k
 j k
r
∂A  ∂Ai  i  r
k = 
− Ai   a j
∂x  ∂x
k
 j k 
r
∂A rj ∂A j  i 
= A a , Since A j, k = − Ai  
∂x k j, k
∂x k
 j k
Proved.
Geometry 131

7.5 INTRINSIC DIFFERENTIATION


r
Let a vector field A(x ) and

C : x i = x (t ), t1 ≤ t ≤ t 2
i

r
be a curve in some region of E 3 . The vector A(x ) depend on the parameter t and if A(x ) is a
differentiable vector then
r r
dA ∂A dx j
= ⋅
dt ∂x j dt
r
α r dx
j
dA
= A, j aα
dt dt
Since we know
r  ∂A α  α  r
∂A α r
= A, j a α =  j +   aα (See Pg. 127, Theo. 7.2)
∂x j  ∂x i j

So,
r
dA  ∂A α  α  i  r dx j
=  j + i j  A  aα dt
dt  ∂x   
r
dA  dAα  α  i dx j  r
=  +  A  aα
dt  dt i j dt 

dAα  α  i dx j
The formula +  A is called the absolute or Intrinsic derivative of A α with respect
dt  
i j dt
δA α
to parameter t and denoted by .
δt
δA α dAα  α  i dx j δA δA
So, δt = dt + i j A dt is contravariant vector. If A is a scalar then, obviously, = .
  δt δt
Some Results
(i) If Ai be covariant vector
δAi dAi  α  dx β
= dt − i β Aα dt
δt  
δA ij δAij  i  αj dx β  j  iα dx β
(ii) = + A + A
δt dt α β dt α β dt
δA ij δA ij
 i  α dx β  α  i dx β
(iii) = + A −  A
δt dt α β j dt i β α dt
132 Tensors and Their Applications

δA ijk δA ijk
 i  α dx β  α  i dx β  α  i dx β
(iv) = +  A jk − A − A
δt dt α β  dt  j β αk dt k β j α dt

EXAMPLE 5
δg ij
If g ij be components of metric tensor, show that = 0.
δt
Solution

The intrinsic derivative of g ij is

δg ij dg ij α dx β  α  dx β
= −   g αj −  g iα
δt dt i β dt β j  dt

∂g ij dx β  α  dx β  α  dx β
= −   g α −   g α
∂x β dt i β dt β j 
j i
dt

 ∂g ij  α   α   dx β
 − 
=  ∂x β    g α –  g iα 
β j   dt
j
  
i β

δg ij  ∂g ij  dx β
=  β − [iβ, j ] − [βj, i]
δt  ∂x  dt

α   α 
as   gαj = [ i β, j ] and   g i α = [β j , i ] .
i β  β j 

∂g ij
But = [iβ, j ] + [ jβ, i].
∂x β
δg ij
So, = 0.
δt

EXAMPLE 6
Prove that
d ( g ij A i A j ) δA i
= 2 g ij A i
dt δt
Solution
Since g ij Ai a j is scalar..
Then
d ( g ij A i A j ) δ(g ij A i A j )
=
dt δt
Geometry 133

δ( A i A j )
= g ij , since g ij is independent of t.
δt
 δA i j δA j 
= g ij  A + Ai 
 δt δt 
Interchange i and j in first term, we get

d ( g ij A i A j )  i δA j δA j 
= g ij  A + Ai  as g ij is symmetric.
dt  δt δt 

d ( g ij A i A j ) δA j
= 2 g ij Ai
dt δt
Proved.

EXAMPLE 7
Prove that if A is the magnitude of Ai then
Ai , j Ai
A, j =
A
Solution
Given that A is magnitude of Ai . Then
Since
g ik A i A k = Ai A i

g ik A i A k = A 2
Taking covariant derivative w.r. to x j , we get
g ik A,i j A k + g ik A i A,kj = 2 AA, j
Interchange the dummy index in first term, we get

g ki A,kj A i + g ik A i A,kj = 2 AA, j

2 g ik Ai A,kj = 2 AA, j

g ik A i A,kj = AA, j

( )
Ai g ik A,kj = AA, j

Ai Ai , j = AA, j since g ik A,kj = Ai , j

Ai , j Ai
A, j =
A
Proved.
134 Tensors and Their Applications

7.6 PARALLEL VECTOR FIELDS


Consider a curve
C : x i = x (t ), t1 ≤ t ≤ t 2 , (i = 1, 2, 3)
i

r
in some region of E 3 and a vector A localized at point P of C. If we construct at every point of C a
vector equal to A in magnitude and parallel to it in direction, we obtain a parallel field of vector along the
curve C.
3
3 X
Y 2
X
C
1
X
P

O 2
Y

1
Y

r r
if A is a parallel field along C then the vector A do not change along the curve and we can write
r
dA r
= 0. It follows that the components Ai of A satisfy a set of simultaneous differential equations
dt
∂A i
= 0 or
∂t
dAi  i  α dxβ
+ A =0
dt α β dt
This is required condition for the vector field Ai is parallel.

7.7 GEOMETRY OF SPACE CURVES


Let the parametric equations of the curve C in E 3 be
C : x i = x i (t ), t1 ≤ t ≤ t 2 (i = 1, 2, 3).
The square of the length of an element of C is given by
i j
ds 2 = g ij dx dx ...(1)
and the length of arc s of C is defined by the integral

dxi dx j

t2
s= gij dt ...(2)
t1 dt dt
from (1), we have
dx i dx j
g ij =1 ...(3)
ds ds
Geometry 135

dx i
Put = λi . Then equation (3) becomes
ds
g ij λi λj = 1 ...(4)
r r
The vector λ , with components λi , is a unit vector. Moreover,, λ is tangent to C, since its
dx i
components λi , when the curve C is referred to a rectangular Cartesian coordinate Y,, becomes λi = .
ds
These are precisely the direction cosines
r of the tangent vector to the curve C.
r
r
Consider a pair of unit vectors λ and µ (with components λi and µ i respectively) at any point
P of C. Let λ is tangent to C at P Fig. (7.6).
λ
λ + dλ

P(x) (Q(x + dx)


µ
C
r + dr
r

O
Fig. 7.6
r r
The cosine of the angle θ between λ and µ is given by the formula

cos θ = g ij λ λ
i j
...(5)
r r
and if λ and µ are orthogonal, then equation (5) becomes

g ij λi µ j = 0 ...(6)
r
Any vector µ satisfying equation (6) is said to be normal to C at P..
Now, differentiating intrinsically, with respect to the are parameter s, equation (4), we get

δλi j δλj i
g ij λ + gij λ =0 …(7)
δs δs

as g ij is constant with respect to s.


Interchange indices i and j in second term of equation (7) we get

δλi j δλi j
g ij λ + gij λ =0
δs δs
Since g ij is symmetric. Then

δλj
2 g ij λi =0
δs
136 Tensors and Their Applications

δλ j
⇒ g ij λi =0
δs
δλj
we see that the vector either vanishes or is normal to C and if does not vanish
δs
δλj
we denote the unit vector co-directional with by µ j and write
δs

1 δλ j K = δλ
j

µj = ,
δs ...(8)
K δs

where K > 0 is so chosen as to make µ j a unit vector..


The vector µ j is called the Principal normal vector to the curve C at the point P and K is the
curvature of C.
r r
The plane determined by the tangent vector λ and the principal normal vector µ is called the
osculating plane to the curve C at P.
r
Since µ is unit vector

g ij µi µ j = 1 ...(9)
Also, differentiating intrinsically with respect to s to equation (6), we get
δλj j δλ j
g ij µ + g ij λi =0
δs δs
or

δµ j δλi j
g ij λi = − gij µ
δs δs
δλi
= − Kg ij µ i µ j Since = K µi
δs
δµ j
g ij λi = − K , since g ij µ i µ j = 1.
δs
δµ j
g ij λi +K =0
δs
δµ j
g ij λi + g ij λi λ j K = 0 as g ij λi λj = 1
δs
 δµ j 
g ij λi  + Kλ j  = 0
 δs 
δµ j
This shows that the vector + Kλ j is orthogonal to λi .
δs
Geometry 137

r
Now, we define a unit vector v , with components v j , by the formula

1  δµ j 
v i = τ  δs + Kλ 
j
...(10)
 
δµi r r r
where τ = + Kλi the vector v will be orthogonal to both λ and µ .
δs
To choose the sign of τ in such a way that
g eijk λi µ j v k = 1 ...(11)
r r
so that the triad of unit vectors λ , µ and νr forms a right handed system of axes.
2
∂ xi
Since eijk is a relative tensor of weight −1 and g = it follows that ε ijk = g eijk is an
∂xj
obsolute tensor and hence left hand side of equation (11) is an invariant v k in equation (11) is determined
by the formula

v k = ε λi µ j
ijk
...(12)

α 1 ijk
where λ i and µ j are the associated vectors g iα λα and g iα µ and ε ijk = e is an absolute tensor..
g
The number τ appearing in equation (10) is called the torsion of C at P and the vector vr is the
binormal.
We have already proved that in Theorem 7.2, Pg. 127.
r
∂A r
= A,αi a α
∂x i
r
if the vector field A is defined along C, we can write
r
∂A ∂x i α ∂x r
i
= A a ...(13)
∂x i ∂s ∂s α
, i

Using definition of intrinsic derivative,


δA α α dx
i
= A,i
δs ds
Then equation (13) becomes
r r r
dA δA α r ∂A dx i dA
= a ⋅ ⋅ ⋅ ; as = ...(14)
ds δs α ∂x i ds ds
r r
Let r be the position vector of the point P on C then the tangent vector λ is determined by
r
dr r r
= λi ai = λ
ds
from equation (14), we get
r r
d 2r dλ δλα r r
⇒ 2 = ds
= aα = c ...(15)
ds δs
138 Tensors and Their Applications

r r
where c is a vector perpendicular to λ.
r r
With each point P of C we can associate a constant K , such that c K = µ is a unit vector..
Since
r
c r
= µ
K
r 1 δλα r
µ = a , from (15)
K δs α
from equation (8), we get
r r α
µ = µα aα , since µ α = 1 δλ
K δs

7.8 SERRET-FRENET FORMULA


The serret-frenet formulas are given by

1 δλi δλi δλ i
= Kµ , K > 0 whereK =
i
(i) µi = or
K δs δs δs
1  δµ i 
i δµi δµi
(ii) ν = τ  δs + K λ = τν – Kλ where τ = δs + K λ
i
i  or i i
  δs
k
δν
(iii) = – τµ k
δs
First two formulas have already been derived in article (7.7), equation (8) and (10).
Proof of (iii)
From equation (12), article (7.7), we have

ε ijk λ i µ j = ν k

where λi , µ i , ν k are mutually orthogonal.


Taking intrinsic derivative with respect to s, we get

δλi δµ j δνk
ε ijkµ j + ε ijk λ i =
δs δs δs
From formulas (i) and (ii), we get

δνk
ε ijk Kµ iµ j + εijk λ i (τν j – Kλ j ) =
δs
δνk
ε ijk λ i (τν j – Kλ j ) = , Since ε ijk µ i µ j = 0
δs

δνk
ε ijk λ iν j τ – Kε ijk λ j λi =
δs
Geometry 139

Since ε ijk λ i λ j = 0

δνk
ε ijk λi ν j τ =
δs
Since ε ijk λ i ν j = µ k , but ε ijk are skew-symmetric.
Then
ε ijk λ i ν j = – µ k
So,
δνk
– µkτ =
δs
δνk
= – τµ
k
or
δs
δν i
⇒ = – τµi
δs
This is the proof of third Serret-Frenet Formula
Expanded form of Serret-Frenet Formula.

dλi  i  j dx k d 2 xi  i  dx j dx k
(i) + λ i +  = Kµ i
ds  j k  ds = Kµ or
ds 2  j k  ds ds

dµi  i  j dx k
+  µ
ds = τν – Kλ
(ii) i i
ds  jk 
dv i  i  j dx k
+  ν
ds = – τµ
(iii) i
ds  jk 

EXAMPLE 8
Consider a curve defined in cylindrical coordinates by equation
 x1 = a
 2
 x = θ( s )
 x3 = 0

This curve is a circle of radius a.
The square of the element of arc in cylindrical coordinates is
ds 2 = (dx ) + ( x ) (dx ) + (dx )
1 2 1 2 2 2 3 2

so that g11 = 1, g22 = (x1 )2 , g33 = 1, gij = 0, i≠ j


It is easy to verify that the non-vanishing Christoffel symbols are (see Example 3, Page 61)
 1 2 2 1
  = x1 ,   =   = 1 .
 
22  
12  
21 x
140 Tensors and Their Applications

dx i dθ
The components of the tangent vector λ to the circle C are λi = so that λ1 = 0, λ2 = ,
ds ds
λ3 = 0.
Since λ is a unit vector, gijλiλj= 1 at all points of C and this requires that

(x1 )2  ddsθ 


2 2
2  dθ 
= a   =1
 ds 
2

So,   = 2 and by Serret-Frenet first formula (expanded form), we get
1
 ds  a
dλ1  1  j dx k  1  2 dx 2 1
Kµ = ds  j k  ds =  2 2λ ds = –
+1 λ
    a

dλ2  2  j dx k  2  2 dx1
Kµ2 = ds +  j k λ ds =  2 1λ ds = 0
   

dλ3  3  j dx k
Kµ = ds +  j k λ ds = 0
3
 

1
Since µ is unit vector, g ij µi µ j = 1 and it follows that K = , 1 = –1, µ 2 = 0, µ 3 = 0
a µ

Similarly we can shows that τ = 0 and ν1 = 0, ν 2 = 0, ν 3 = 1.

7.9 EQUATIONS OF A STRAIGHT LINE


Let Ai be a vector field defined along a curve C in E 3 such that
i
C : x i = x (s ) . s1 ≤ s ≤ s 2, (i = 1, 2, 3).
s being the arc parameter.
If the vector field Ai is parallel then from article 7.6 we have

δA i
=0
δs

dAi  i  α dx β
or + A
ds = 0 (1)
ds α β
We shall make use of equation (1) to obtain the equations of a straight line in general curvilinear
r
coordinates. The characteristic property of straight lines is the tangent vector λ to a straight line is
r
directed along the straight line. So that the totality of the tangent vectors λ forms a parallel vector field.
Geometry 141

dxi
Thus the field of tangent vector λ =
i
must satisfy equation (1), we have
ds
δλi d 2 xi  i  dx α dx β
= +  = 0
δs ds 2 α β ds ds

d 2 xi  i  d x α d x β
The equation +  = 0 is the differential equation of the straight line.
ds 2 α β ds d s

EXERCISE

d (gij AiB j ) δA i j δB i
1. Show that = g ij B + g ij A i
dt δt δt

∂ Ai ∂A j
2. Show that Ai , j – A j , i = –
∂x j ∂ xi

3. If Ai = g ij A j show that Ai , k = giα Aα,k

d ( gij Ai B j )
= Ai,k B + Bi,k A
i i
4. Show that
dxk
5. Show that

δ2 λi dK i
= µ – K ( τνi – Kλi )
ds 2 ds

δ 2µi dτ i dK i
= ν – ( K 2 + τ2 )µi – λ
δs2 ds ds

δ 2νi dτ
= τ( Kλ – τv ) – µ
i i i
δs2 ds
6. Find the curvature and tension at any point of the circular helix C whose equations in cylindrical
coordinates are

C : x1 = a, x2 = θ, x 3 = θ
Show that the tangent vector λ at every point of C makes a constant angle with the direction of X 3 -
axis. Consider C also in the form y1 = a cosθ, y2 = a sinθ, y3 = θ. Where the coordinates yi are
rectangular Cartesion.
CHAPTER – 8

ANALYTICAL MECHANICS

8.1 INTRODUCTION
Analytical mechanics is concerned with a mathematical description of motion of material bodies subjected
to the action of forces. A material body is assumed to consist of a large number of minute bits of matter
connected in some way with one another. The attention is first focused on a single particle, which is
assumed to be free of constraints and its behaviour is analyzed when it is subjected to the action of
external forces. The resulting body of knowledge constitutes the mechanics of a particle. To pass from
mechanics of a single particle to mechanics of aggregates of particles composing a material body, one
introduces the principle of superposition of effects and makes specific assumptions concerning the
nature of constraining forces, depending on whether the body under consideration is rigid, elastic,
plastic, fluid and so on.

8.2 NEWTONIAN LAWS


1. Every body continues in its state of rest or of uniform motion in a straight line, except in so
far as it is compelled by impressed forces to change that state.
2. The change of motion is proportional to the impressed motive force and takes place in the
direction of the straight line in which that force is impressed.
3. To every action there is always an equal and contrary reaction; or the mutual actions of two
bodies are always and oppositely directed along the same straight line.
The first law depends for its meaning upon the dynamical concept of force and on the kinematical
idea of uniform rectilinear motion.
The second law of motion intorduces the kinematical concept of motion and the dynamical idea
of force. To understand its meaning it should be noted that Newton uses the term motion in the sense
of momentum, i.e., the product of mass by velocity, this, "change of motion" means the time of change
of momentum.
In vector notation, the second law can be stated as
r
r d (mv )
F = dt … (1)
Analytical Mechanics 143

If we postulate the invariance of mass then equation (1) can be written as


r r
F = ma … (2)
r
r d (mv )
from (1) if F = 0 then dt = 0.
So that
r
m v = constant.
hence vr is constant vector..
Thus the first law is a consequence of the second.
The third law of motion states that accelerations always occur in pairs. In term of force we may
say that if a force acts on a given body, the body itself exerts an equal and oppositely directed force on
some other body. Newton called the two aspects of the force of action and reaction.

8.3 EQUATIONS OF MOTION OF A PARTICLE


THEOREM 8.1 The work done in displacing a particle along its trajectory is equal to the change in
the kinetic energy of particle.
Proof: Let the equation of path C of the particle in E 3 be
i
C : x i = x (t ) … (1)
and the curve C the trajectory of the particle. Let at time t, particle is at P {x i (t )}.
If v i be the component of velocity of moving particle then
dx i
vi = … (2)
dt
and if ai be the component of acceleration of moving particle then

δv i dv i  i  j d x k
ai = δt = dt +  j k v d t
 

d 2 x i  i  dx j dx k
ai = +  … (3)
dt 2  j k  dt dt

δv i  i 
where is the intrinsic derivative and the   are the Christoffel symbols calculated from the
δt  jk
metric tensor gij
If m be the mass of particle. Then by Newton’s second law of motion
δv i
Fi = m = ma i … (4)
δt
r
We define the element of work done by the force F in producing a displacement drr by invariant
r
dW = F .drr .
r
Since the components of F and drr are F i and dxi respectively..
144 Tensors and Their Applications

Then
dW = gij F i dxj … (5)
j i
= F j dx where F j = gij F
The work done in displacing a particle along the trajectory C, joining a pair of points P 1 and P 2, is
line integral


P2
W= Fi dx i … (6)
P1
using equation (4) then equation (6) becomes
δv i

P2
W= m g ij dx j
P1 δt
δv i dx j

P2
= m g ij dt
P1 δt dt
δv i j

P2
W= m g ij v dt … (7)
P1 δt
Since g ij v i v j is an invariant then
δ (g ij vi v j ) d
= ( g v iv j )
δt dt ij
d δvi j
or ( g ij v i v j ) = 2 g ij v
dt δt
δvi j 1 d
⇒ g ij v = (g ij v i v j )
δt 2 dt
using this result in equation (7), we get
P2 m d
W= ∫P1
( g v i v j ) dt
2 dt ij
m
[g v i v j ]PP12
W=
2 ij
Let T2 and T1 is kinetic energy at P 2 and P 1 respectively.
W = T2 – T1
m m v2
where T = gij v i v j = is kinetic energy of particle.
2 2
We have the result that the work done by force F i in displacing the particle from the point P 1 to
1
the point P 2 is equal to the difference of the values of the quantity T = mv 2 at the end and the
2
beginning of the displacement.

8.4 CONSERVATIVE FORCE FIELD


P2
The force field F i is such that the integral W = Fi dx i is independent of the path.
P1
Therefore the integrand F i dxi is an exact differential
dW = F i dxi … (8)
Analytical Mechanics 145

of the work function W. The negative of the work function W is called the force potential or
potential energy.
We conclude from equation (8) that
∂V
Fi = – i … (9)
∂x
i
where potential energy V is a function of coordinates x . Hence, the fields of force are called conservative
∂V
if F i = – .
∂x i
THEOREM 8.2 A necessary and sufficient condition that a force field Fi, defined in a simply
connected region, be conservative is that Fi,j = Fj,i.
∂V
Proof: Suppose that F i conservative. Then F i = –
∂x i
Now,
∂Fi  k 
F i,j = −   Fk
∂x j i j 
∂V
∂  − i 
 ∂x  −  k F
F i,j =   k
∂x j i j 

∂ 2V k
=− −  Fk … (1)
∂x ∂x  j i 
j i

and
∂F j k
F j,i = −   Fk
∂x  j i
i

Similarly,
∂ 2V k
F j,i = – −  Fk … (2)
∂x ∂x
i j
 ji
From equation (1) and (2), we get
F i,j = F j,i
conversely, suppose that F i,j = F j,i
Then
∂Fi  k  ∂F j  k 
−   Fk = −   Fk
∂x j
i j  ∂x i  j i

∂Fi ∂F j  k 
as   due to symmetry..
j =
∂x ∂x i i j
146 Tensors and Their Applications

∂V
Take F i = −
∂x i
Then
∂Fi ∂ 2V
j =
− j i
∂x ∂x ∂x
∂ 2V
= − i j
∂x ∂x
∂  ∂V 
= − 
∂x i  ∂x j 
∂Fi ∂F j
j =
∂x ∂x i

∂V
So, we can take Fi = − .
∂x i
Hence, F i is conservative.

8.5 LAGRANGEAN EQUATIONS OF MOTION


Consider a particle moving on the curve
C : x i = x i (t )
At time t, let particle is at point P (xi).
1
The kinetic energy T = mv 2 can be written as
2
1
T = m g ij x& x&
i j
2
Since x& i = v i.
1
or T= m g jk x& j x& k … (1)
2
Differentiating it with respect to x& i , we get
∂T 1  ∂x& j k &k 
j ∂x
m g  &
x + &
x 
= 2  ∂x& i & i 
∂x& i
jk
 ∂x 

=
1
2
(
m g jk δij x& k + δki x& j )
1 1
= m g jk δ ij x& k + m g jk δ ki x& j
2 2

=
1
2
(
m gik x& k + g ji x& j )
1
= m (gij x& + gij x& ) gij = g ji
j j
as
2
Analytical Mechanics 147

∂T
= m g ij x& j
∂x& i
∂T
or = m g ik x& k …(2)
∂x& i
Differentiating equation (2) with respect to t, we get
d  ∂T 
  = m
dt  ∂x& i 
d
g x& k
dt ik
( )
d k
= m  g ik x& + g ik &x& 
k

 dt 

 ∂g ik dx j k 
=  ∂x j dt x& + g ik &x& 
k
m
 

d ∂T ∂g
= m ikj x& x& + m g ik &x&
j k k
… (3)
dt ∂x& i ∂x
1
Since T = m g jk x& j x&k
2
Differentiating it with respect to xi, we get
∂T 1 ∂g jk j k
i =
m x& x& … (4)
∂x 2 ∂x i
Now,
d  ∂T  ∂T ∂g 1 ∂g
  − i = m ikj x& j x&k + m g ik&x&k − m jki x& j x& k
dt  ∂x& i  ∂x ∂x 2 ∂x

1 gik j k 1 1 ∂g
= m g ik &x& + m j x& x& + m g ik x& j x& k − m jki x& j x& k
k
2 ∂x 2 2 ∂x

1 g ik j k 1 ∂g ij k j 1 ∂g jk j k
= m g ik &x& + x& x& + m k x& x& − m i x& x&
k
m
2 ∂x j 2 ∂x 2 ∂x

1  ∂gik ∂g ij ∂g jk  j k
= m g ik &x& + 2 m  j + k −  x& x&
k

 ∂x ∂x ∂x i 

= m g ik &x&k + m [ jk , i ] x& j x& k

= m g ik &x&k + m g il gil [ jk , i ]x& j x& k

= m g il &x&l + m g il gil [ jk , i]x& j x& k


148 Tensors and Their Applications

[
= m g il x&&l + g il [ jk , i ]x& j x& k ]
  l   l 
= m g il  &x&l +  x& j x& k  , as g il [ jk , i] =  
  jk   j k

d  ∂T  ∂T  l  j k
Since a = x&& +  j k x& x&
l l
 − = m g il a l ,
dt  ∂x& i  ∂x i  
where al is component of acceleration
d  ∂T  ∂T
or  − = m ai
dt  ∂x& i  ∂x i

d  ∂T  ∂T
 − = Fi … (5)
dt  ∂x& i  ∂x i
where F i = m ai, component of force field. The equation (5) is Lagrangean equation of Motion.
∂V
For a conservative system, F i = – . Then equation (5) becomes
∂x i
d  ∂T  ∂T ∂V
 − i = − i
dt  ∂x& i  ∂x ∂x

d  ∂T  ∂(T − V )
or  − =0 … (6)
dt  ∂x& i  ∂x i
Since the potential V is a function of the coordinates xi alone. If we introduce the Lagrangean
function
L=T–V
Then equation (6) becomes
d  ∂L  ∂L
 − =0 … (7)
dt  ∂x& i  ∂x i

EXAMPLE 1
Show that the covariant components of the acceleration vector in a spherical coordinate system
with
ds 2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2 are

a1 = &x&1 − x1 (x& 2 )2 − x1 (x& 3 sin x 2 ) 2

a2 =
d
dt
[ ]
( x1 )2 x& 2 − (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2

and a3 =
d
dt
[
(x1 sin x 2 ) 2 x& 3 ]
Analytical Mechanics 149

Solution
In spherical coordinate system, the metric is given by
ds2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2
If v is velocity of the paiticle then
2 2 2
 ds   dx1   2  3
2
1 2  dx  2 2  dx 
2   = + ( x ) + ( x1
sin x )
 dt   dt 
v =  dt   dt 
   
v 2 = ( x&1 )2 + (x1 ) 2 ( x& 2 ) 2 + ( x1 sin x 2 )2 (x& 3 ) 2
If T be kinetic energy then
1 2
T= mv
2
1
2
[
T = m (x& ) + ( x ) (x& ) + ( x sin x ) (x& )
1 2 1 2 2 2 1 2 2 3 2
] … (1)
By Lagrangean equation of motion
d  ∂T  ∂T
 − = F i and m ai = F i
dt  ∂x& i  ∂x i
where F i and ai are covariant component of force field and acceleration vector respectively.
So,
d  ∂T  ∂T
 − = m ai … (2)
dt  ∂x& i  ∂x i
Take i = 1,
d  ∂T  ∂T
m a1 =  −
dt  ∂x&1  ∂x 1
from (1), we get

m a1 =
1 d m
[
m (2x&1 ) − 2x 1 ( x& 2 )2 + 2 x1 (sin x 2 ) 2 ( x& 3 )2
2 dt 2
]

a1 =
dx&1
dt
[
− x1 ( x& 2 )2 + x1 (sin x 2 ) 2 ( x& 3 )2 ]
a1 = &x&1 − x1 (x& 2 )2 − x1 (x& 3 sin x 2 ) 2
Take i = 2,
d  ∂T  ∂T
m a2 =  − 2
dt  ∂x& 2  ∂x

m a2 =
1 d 1 2 2 1
[ ]
m ( x ) 2 x& − m 2( x1 )2 sin x 2 cos x 2 (x&3 ) 2
2 dt 2

a2 =
d
dt
[ ]
( x1 )2 x& 2 − (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2
150 Tensors and Their Applications

Take i = 3
d  ∂T  ∂T
m a3 =  −
dt  ∂x&3  ∂ x 3

m a3 =
1 d
[ ]
m 2( x& 3 ) ( x1 sin x 2 )2 − 0
2 dt

a3 =
dt
[
d 3 1 2
x& ( x ) (sin x 2 ) 2 ]
EXAMPLE 2
Use Lagrangean equations to show that, if a particle is not subjected to the action of forces then
its trajectory is given by yi = ait + bi where ai and bi are constants and the yi are orthogonal cartesian
coordinates.
Solution
If v is the velocity of particle. Then we know that,

v 2 = g ij y& i y& j
where yi are orthogonal cartesian coordinates.
Since
gij = 0, i ≠ j
gij = 1, i = j
So,
v 2 = ( y& i ) 2
But,
1 2
T=mv , T is kinetic energy..
2
1
T = m( y& )
i 2
2
The Lagrangean equation of motion is

d  ∂T  ∂T
 −
dt  ∂y& i  ∂yi
= Fi

Since particle is not subjected to the action of forces.


So, F i = 0
d 1 
Then  m 2 y& i  – 0 = 0
dt  2 

dy& i
m =0
dt
Analytical Mechanics 151

dy& i
or =0
dt

⇒ y& i = ai
⇒ yi = ait + bi
i i
where a and b are constant.

EXAMPLE 3
Prove that if a particle moves so that its velocity is constant in magnitude then its acceleration
vector is either orthogonal to the velocity or it is zero.
Solution
If v i be the component of velocity of moving particle then
dx i
vi = or v i = x& i
dt
given |v| = constant.
Since
g ij vi v j = | v |2 = constant
Taking intrinsic derivative with respect to t, we get
δ
(g v i v j ) = 0
δt ij

 δv i j δv j 
g ij  v + vi  =0

 δt δt 

δv i j δv j
g ij v + g ij v i =0
δt δt

δv i j δv i
g ij v + g ji v j = 0, (Interchange dummy index i and j in second term)
δt δt

δv i j
2 g ij v = 0 as gij = gji
δt

δvi j
g ij v =0
δt
δv i δv i
This shows that acceleration vector is either orthogonal to v i or zero i.e., = 0.
δt δt
152 Tensors and Their Applications

8.6 APPLICATIONS OF LAGRANGEAN EQUATIONS


(i) Free-Moving Particle
If a particle is not subjected to the action of forces, the right hand side of equation (5), 148, vanishes.
Then we have
d  ∂T  ∂T
 − =0 … (1)
dt  ∂x& i  ∂x i
1
If xi be rectangular coordinate system, then T = m y& i y& i .
2
Hence, the equation (1) becomes m &y&i = 0. Integrating it we get yi = ait + bi, which represents
a straight line.
(ii) Simple Pendulum
Let a pendulum bob of mass m be supported by an extensible string. In spherical coordinates, the
metric is given by
ds2 = dr 2 + r 2 d φ2 + r 2 sin 2 φ d θ 2
If T be the kinetic energy, then

mv = m (r&2 + r 2 φ& 2 + r 2 sin 2 φ θ& 2 )


1 2 1
T= … (1)
2 2

O
Y2

φ
R
mg cos φ
Y1 θ P
mg sin φ
Y3 mg

Fig. 8.1.

from Lagrangean equation of motion


d  ∂T  ∂T
 − = Fi i = 1, 2, 3
dt  ∂x& i  ∂x i

x1 = r , x 2 = φ, x 3 = θ.

So, take x1 = r
Analytical Mechanics 153

d  ∂T  ∂T
 − = mg cos φ − R
dt  ∂r&  ∂r
from (1), we have
R
&r& − rφ&2 − r sin 2 φ θ&2 = g cos φ − … (2)
m
Take x2 = φ, we have

r &φ& + 2r&φ& − r sin φ cos φ θ& 2 = – g sin φ … (3)


and take x3 = θ, we have
d 2& 2
(r θ sin φ) = 0 … (4)
dt
If the motion is in one plane, we obtain from equations (2), (3), and (4), by taking θ& = 0.
R
r&& − r φ& 2 = g cos φ − m

r &φ& + 2r&φ& = – g sin φ

If r& = 0, we get, &φ& = −  sinφ which is equation of simple pendulum supported by an
g
r
inextensible string. For small angles of oscillation the vibration is simple harmonic. For large vibration
the solution is given in the term of elliptic functions.

8.7 HAMILTON’S PRINCIPLE


If a particle is at the point P 1 at the time t1 and at the point P 2 at the time t2, then the motion of the
particle takes place in such a away that


t2
(δT + Fi δxi ) dt = 0
t1

where xi = xi (t) are the coordinates of the particle along the trajectory and xi + δx i are the coordinates
along a varied path beginning at P 1 at time t1 and ending at P 2 at time t2.
Proof: Consider a particle moving on the curve

C : x i = x (t ),
i
t1 ≤ t ≤ t 2
At time t, let particle is at P(xi). If T is kinetic energy. Then
1
T= m g ij x& i x& j
2
or T = T ( x i , x& i ) i.e, T is a function of xi and x& i . Let C ' be another curve, joining t1 and t2 close
to be C is
C' : x i (ε, t ) = x i (t ) + δx i (t)
154 Tensors and Their Applications

At t1 and t2
xi = x i = x i + ε δ x i

⇒ δx i (t1 ) = 0 and δx i (t 2 ) = 0
But T = T ( x i , x& i ).
If δT be small variation in T.. Then
∂T i ∂T i
δT = ∂x&i δx& + ∂x i δx

Now,
∂T ∂T
∫ {(δT + F )δx }dt = ∫
t2  
 i δx i + i δx& i + Fi δx i  dt
t2
i
t1 i t1  ∂x ∂x& 
∂T i ∂T i
∫ ∫ ∫
t2 t2 t2
= δx dt + δx& dt + Fi δx i dt
t1 ∂x i t1 ∂x&i t1

∂T
Integrating second term by taking as 1st term
∂x& i

∂T i  ∂T i  2
t
d  ∂T  i
∫ ∫ ∫
t2 t2 t2
= δx dt +  ∂x& i δx  −  δx dt + Fi δxi dt
t1 ∂x i   t1 t1 dt  ∂x&i  t1

Since δx i (t1 ) = 0, δx i (t 2 ) = 0.

 ∂T i  2
t

then  & i δx  = 0.
 ∂x  t1
So,

∫ (δT + F δx )dt = ∫
∂T d  ∂T  i
∫t
t2 t2 i t2
i
δ x dt −  i δ x dt
 ∂ x&
i i
t1 t1
∂x 1 dt 


t2
+ Fi δx i dt
t1

t2  ∂T  i
∫ (δT + F δx )dt = ∫
t2 d  ∂T 
 i −  i  + Fi δx dt
i

 ∂x dt  ∂x& 
i
t1 t1 
since particle satisfies the Lagrangean equation of motion. Then
d  ∂T  ∂T
 − = Fi
dt  ∂x& i  ∂x i

∂T d  ∂T 
or −   = – Fi
∂x i dt  ∂x& i 
Analytical Mechanics 155

So,

∫ ∫
t2 t2
(δT + Fi δx i ) dt = [−Fi + Fi ] δxi dt
t1 t1


t2
(δT + Fi δx i ) dt = 0 Proved.
t1

8.8 INTEGRAL OF ENERGY


THEOREM 8.3 The motion of a particle in a conservative field of force is such that the sum of its
kinetic and potential energies is a constant.
Proof: Consider a particle moving on the curve
C : x i = x i (t ), t1 ≤ t ≤ t 2
i
At time t, let particle is at P (x ). If T is kinetic energy. Then
1
T= m g ij x& i x& j
2
1
or T= m g ij v i v j
2
As T is invariant. Then
Taking intrinsic derivative with respect to t, we get
dT δT
=
dt δt
δ 1 
=  m gij v i v j 
δt  2 

1  δv i j δv j 
= m g ij  v + vi
2  δt δt 

1  δv i j δv j i 
= m  gij v + g ij v 
2  δt δt 
1  δvi j δvi j 
= 2 m  gij δt v + g ji δt v  , Since i and j are dummy indices.
 

1 δv i j
= m 2 g v as g ij = g ji
δt
ij
2

dT δv i j
= m gij v
dt δt
dT δv j i
or = m g v
δt
ij
dt
156 Tensors and Their Applications

δv j
j i
= m g ij a v as = aj
δt
dT
= m ai v i, since gij aj = ai
dt
dT
= F i v i,
dt

Since F i = m ai is a covariant component of force field.


But given F i is conservative, then
∂V
Fi = – , where V is potential energy..
∂x i
So,
dT ∂V i
=– v
dt ∂x i
∂V dx i
= −
∂x i dt
dT dV
= −
dt dt
dT dV d
+ = 0 ⇒ (T + V ) = 0
dt dt dt
⇒ T + V = h, where h is constant.
Proved.

8.9 PRINCIPLE OF LEAST ACTION


Let us consider the integral


P2
A= mv.ds (1)
p1

evaluated over the path

C : x i = x (t ), t1 ≤ t ≤ t 2
i

where C is the trajectory of the particle of mass m moving in a conservative field of force.
In the three dimensional space with curvilinear coordinates, the integral (1) can be written as
dx i

P2
A= m g ij dx j
p1 dt

dx i dx j

t ( P2 )
= m gij dt
t ( P1 ) dt dt
Analytical Mechanics 157

1 dx i dx j
Since T = m g ij , we have
2 dt dt


t ( P2 )
A= 2T dt
t ( P1 )

This integral has a physical meaning only when evaluated over the trajectory C, but its value can
be computed along any varied path joining the points P 1 and P 2.
Let us consider a particular set of admissible paths C' along which the function T + V, for each
value of parameter t, has the same constant value h. The integral A is called the action integral.
The principle of least action stated as “of all curves C' passing through P 1 and P 2 in the
neighbourhood of the trajectory C, which are traversed at a rate such that, for each C', for every value
of t, T + V = h, that one for which the action integral A is stationary is the trajectory of the particle.”

8.10 GENERALIZED COORDINATES


In the solution of most of the mechanical problems it is more convenient to use some other set of
coordinates instead of cartesian coordinates. For example, in the case of a particle moving on the
surface of a sphere, the correct coordinates are spherical coordinates r, θ, φ where θ and φ are only
two variable quantities.
Let there be a particle or system of n particles moving under possible constraints. For example, a
point mass of the simple pendulum or a rigid body moving along an inclined plane. Then there will be
a minimum number of independent coordinates required to specify the motion of particle or system of
particles. The set of independent coordinates sufficient in number to specify unambiguously the system
configuration is called generalized coordinates and are denoted by q 1 , q 2 , ... q n where n is the total
number of generalized coordinates or degree of freedom.
Let there be N particles composing a system and let x(i α) , (i = 1, 2,3), (α = 1, 2,...N ) be the positional
coordinates of these particles referred to some convenient reference frame in E 3. The system of N free
particles is described by 3N parameters. If the particles are constrained in some way, there will be
certain relations among the coordinates x(i α) and suppose that there are r such independent relations,

f i ( x1(1) , x(21) , x 3(1 ) ; x1( 2 ) , x(22 ) , x(32 ) ;...x1( N ) x(2N ) x(3N ) ) = 0, (i = 1, 2, ..., r) … (1)
By using these r equations of constraints (1), we can solve for some r coordinates in terms of the
remaining 3N – r coordinates and regard the latter as the independent generalized coordinates qi. It is
more convenient to assume that each of the 3N coordinates is expressed in terms of 3N – r = n
independent variables qi and write 3N equations.
x(i α) = x(i α) (q1 , q 2 ,...,q n , t ) … (2)
where we introduced the time parameter t which may enter in the problem explicitly if one deals with
moving constraints. If t does not enter explicitly in equation (2), the dynamical system is called a
natural system.
The velocity of the particles are given by differentiating equations (2) with respect to time. Thus
∂x i(α ) ∂x i(α )
x&(i α) = q& j + … (3)
∂q j ∂t
158 Tensors and Their Applications

The time derivatives q& i of generalized coordinates qi the generalized velocities.


For symmetry reasons, it is desirable to introduced a number of superfluous coordinates qi and
describe the system with the aid of k > n coordinates q1, q2,..., qk. In this event there will exist certain
relations of the form
f j (q1 ....,q k , t ) = 0 … (4)
Differentiating it we get
∂f j i ∂f j
q& +
∂t = 0 … (5)
∂q i
It is clear that they are integrable, so that one can deduce from them equations (4) and use them
to eliminate the superfluous coordinates .
In some problems, functional relations of the type
F j (q1 , q 2 ,...,q k ; q&1 ,...,q& k , t ) = 0, ( j = 1, 2, 3, ..., m) … (6)
arise which are non-integrable. If non-integrable relations (6) occurs in the problems we shall say that
the given system has k – m degrees of freedom, where m is the number or independent non-integrable
relations (6) and k is the number of independent coordinates. The dynamical systems involving non-
integrable relations (6) are called non-holonomic to distinguish them from holonomic systems in which
the number of degrees of freedom is equal to the number of independent generalized coordinates.
In other words, a holonomic system is one in which there are no non-integrable relations involving
the generalized velocities.

8.11 LAGRANGEAN EQUATIONS IN GENERALIZED COORDINATES


Let there be a system of particle which requires n independent generalized coordinates or degree of
freedom to specify the states of its particle.
The position vectors xr are expressed as the function of generalized coordinates q i , (i = 1, 2,...,n)
and the time t i.e.,
1 2 n
xr = xr (q , q ,...,q , t ); (r = 1, 2, 3, )

The velocity x& r of any point of the body is given by


∂x r dq j ∂x r
x& = ∂q j dt + ∂t
r

∂x r j ∂x r
= q& + ,
∂q j ∂t ( j = 1, 2, ..., n)

where q& j are the generalized velocities.


Consider the relation, with n degree of freedom,
1 2 n
xr = xr (q , q ,...,q ) … (1)
involve n independent parameters qi. The velocities x& r in this case are given by
∂x r j
x&r= q& , (r = 1, 2, 3; j = 1, 2, ..., n) …(2)
∂q j
Analytical Mechanics 159

where q& j transform under any admissible transformation,


q k = q k (q1 ,...,q n ), (k = 1, 2, ..., n) … (3)
in accordance with the contravariant law.
The kinetic energy of the system is given by the expression of the form
1
T= Σ m g rs x& r x& s , (r,s = 1,2,3,) … (4)
2

where m is the mass of the particle located at the point xr. The grs are the components of the metric
tensor.
Substituting the value of x& r from equation (2), then equation (4) becomes

1 ∂x r ∂x s i j
T = 2 Σ m g rs i q& q&
∂q ∂q j

1
T= a q&i q& j … (5)
2 ij
∂x r ∂r s
where aij = Σ m g rs , (r, s = 1, 2, 3), (i, j = 1, ..., n)
∂q i ∂q j
1
Since T = a q&i q& j is an invariant and the quantities aij are symmetric, we conclude that the aij
2 ij
are components of a covariant tensor of rank two with respect to the transformations (3) of generalized
coordinates.
Since the kinetic energy T is a positive definite form in the velocities q& i , | aij |> 0. Then we
construct the reciprocal tensor aij .
Now, from art. 8.5, Pg. 146, by using the expression for the kinetic energy in the form (5), we
obtain the formula,

d  ∂T  ∂T  l  l  j k
 ∂qi = a il  &q& +  j k q& q& 
−  
dt  ∂q& i (6)
    
 l 
where the Christoffel symbol   are constructed from the tensor akl.
 jk
Put
l 
q&&l +  q& j q& k = Q l
jk
so, the equation (6), becomes

d  ∂T  ∂T
− l
dt  ∂q& i  ∂qi = ail Q

= Qi (i = 1, 2, ..., n) … (7)
160 Tensors and Their Applications

∂x& r ∂x r ∂x& r ∂ 2 xr j ∂x&r d  ∂x r 


Now, from the realtions & j = , = &
q =
and qi dt  q i  and using equations
∂q ∂q j ∂qi ∂x i ∂q j ∂ ∂ 
(2) and (4).
Then by straightforward calculation, left hand member of equation (7) becomes

d  ∂T  ∂T ∂x r
−
dt  ∂q& i
 ∂qi

= ∑
m a r
∂q i
… (8)

in which a j = g ij a i is acceleration of the point P.


Also, Newton's second law gives
m ar = F r … (9)
where Fr's
are the components of force F acting on the particle located at the point P..
From the equation (9), we have

∂x r ∂x r
∑ m ar
∂q i
= ∑ Fr
∂q i
and equation (8) can be written as

d  ∂T  ∂T ∂x r
dt  ∂q& i
−
 ∂qi = ∑ Fr
∂q i
… (10)

comparing (7) with (8), we conclude that

∂x r
Qi = ∑ Fr
∂q i
where vector Qi is called generalized force.
The equations

d  ∂T  ∂T
−
dt  ∂q& i  ∂qi = Qi … (11)

are known as Lagrangean equations in generalized coordinates.
They give a system of n second order ordinary differential equations for the generalized coordinates qi.
The solutions of these equations in the form
C : q i = qi (t)
Represent the dynamical trajectory of the system.
If there exists a functions V (q1 , q 2 ,...,q" ) such that the system is said to be conservative and for
such systems, equation (11) assume the form

d  ∂L
 ∂L
−
 ∂qi = 0
dt  ∂q& i
… (12)

where L = T – V is the kinetic potential.
Analytical Mechanics 161

Since L (q, q& ) is a function of both the generalized coordinates and velocities.
∂L i ∂L i
= & i q&& + i q&
dL
… (13)
dt ∂q ∂q

∂L d  ∂L 
.
dt  ∂q& i 
from (12), we have i =
∂q 
Then equation (13), becomes
dL ∂L i d  ∂L  i
= q&& +  i  q&

dt ∂q& i dt  ∂q& 
d  ∂L i 
q&
dt  ∂q& i 
= … (14)

since L = T – V but the potential energy V is not a function of the q& i


∂L i ∂T i
q& = q& = 2T
&
∂q i
∂q& i
1
since T= a q& i q& j .
2 ij
Thus, the equation (14) can be written in the form
d ( L − 2T ) d (T + V )
= =0
dt dt
which implies that T + V = h (constant).
Thus, along the dynamical trajectory, the sum of the kinetic and potential energies is a constant.

8.12 DIVERGENCE THEOREM, GREEN'S THEOREM, LAPLACIAN OPERATOR AND


STOKE'S THEOREM IN TENSOR NOTATION

(i) Divergence Theorem


r
Let F be a vector point function in a closed region V bounded by the regular surface S. Then
r r
∫ ∫
div F = F ⋅ nˆ ds … (1)
V S

where n̂ is outward unit normal to S.


Briefly the theorem states that the integral with subscript V is evaluated over the volume V while
r
the integral in the right hand side of (1) measures the flux of the vector quantity F over the surface S.
r
In orthogonal cartesian coordinates, the divergence of F is given by the formula

r ∂F 1 ∂F 2 ∂F 3
div F = + + … (2)
∂x 1 ∂x 2 ∂x 3
162 Tensors and Their Applications

r
If the components of F relative to an arbitrary curvilinear coordinate system X are denoted by
F i then the covariant derivative of F i is

∂F i  i  k
F ,ij = +  F
∂x j k j 
r
The invariant F, ij in cartesian coordinates represents the divergence of the vector field F .
Also,
r
F ⋅ nˆ = g ij F n = F ni since g ij n = ni
i j i j

Hence we can rewrite equation (1) in the form

∫ F dV ∫ F n dS
i i
,i i
= … (3)
V S

(ii) Symmetrical form of Green's Theorem


Let φ ( x1 , x 2 , x3 ) and ψ (x1 , x 2 , x3 ) be two scalar function in V.. Let φ i and ψ i be the gradients of φ
and ψ respectively, so that
∂φ ∂ψ
and ∇φ = φ i = ∇ψ = ψ i =
∂xi ∂x i
Put F i = φψi and from the divergence of we get

F,i j = g ij Fi , j = g ij (φψi , j + ψi φ j )
Substituting this in equation (3), we get

∫g (φψi , j + ψ i φ j )dV ∫ φψ n dS
ij i
= i … (4)
V S

Since ∇ψ = ψi , then
g ij ψi , j = ∇ 2 ψ … (5)

Also, the inner product g ij ψi φ j can be written as

g ij ψi φ j = ∇φ.∇ψ
where ∇ denote the gradient and ∇ 2 denote the Laplacian operator..
Hence the formula (4) can be written in the form

∫(g φψi , j + g ij ψi .φ J ) dV ∫ φ n̂ ⋅ ∇ψ dS
ij
=
V S

∫ (φ∇ ψ + ∇φ ⋅ ∇ψ) dV = ∫ φ n̂ ⋅ ∇ψ dS
2

V S

∫ φ ∇ ψ dV = ∫ φ n̂ ⋅ ∇ ψ − ∫ ∇φ ⋅ ∇ψ dV
2
… (6)
V S V
∂ψ .
where nˆ ⋅ ∇ψ = ψi n =
i
∂n
Analytical Mechanics 163

Interchanging φ and ψ in equation (5), we get

∫ ψ∇ φ dV ∫ ψn̂ ⋅ ∇φ − ∫ ∇ψ ⋅ ∇φ dV
2
= … (7)
V S V
Subtracting equation (5) from equation (6), we get

 ∂ψ ∂φ 
∫ (φ∇ ψ − ψ∇ φ) dV ∫  φ ∂n − ψ ∂n  dS
2 2
= … (8)
V S

This result is called a symmetric form of Green's theorem.


(iii ) Expansion form of the Laplacian Operator
The Laplacian of ψ is given by
∇ 2 ψ = g ψi , j from (5)
ij

when written in the terms of the christoffel symbols associated with the curvilinear coordinates
i
x covering E 3,
 ∂2 ψ  k  ∂ψ 
∇ 2 ψ = g  i j −   k 
ij
 … (9)
 ∂x ∂x i j  ∂x 
i
and the divergence of the vector F is

∂F i  i 
F,ii = +  F j
…(10)
∂x i  j i 

i ∂
But we know that   = log g
 
j i ∂ xj
The equation (10) becomes

∂F i  ∂ 
F,ii = +  j log g  F j
∂x  ∂x 
i

1 ∂( g F i )
or F,ii = … (11)
g ∂x i
∂ψ
If putting F i = g
ij
= g ij ψ j in equation (11), we get
∂x j
 ∂ψ 
∂  g g ij j 
1  ∂x 
g ij ψ j , i = … (12)
g ∂x i

But from equation (5), we know that


∇ 2 ψ = g ψ j, i
ij
164 Tensors and Their Applications

Hence equation (12) becomes


 ∂ψ 
∂ g g ij j 
1  ∂x 
∇ 2 ψ = g ij ψ j ,i =
∂x i
g
It is expansion form of Laplacian operator.
(iv) Stoke's Theorem
r
Let a portion of regular surface S be bounded by a closed regular curve C and let F be any vector point
function defined on S and on C. The theorem of Stokes states that
r r

nˆ. curl F ds = F .λ̂ ds ∫ … (13)
S C
r
where λ is the unit tangent vector to C and curl F is the vector whose components in orthogonal
cartesian coordinates are determined from

e1 e2 e3
∂ ∂ ∂ r
r
curl F = ∂ x1 ∂x 2 ∂x 3 = ∇ × F … (14)
F1 F2 F3
where ei being the unit base vectors in a cartesian frame.
We consider the covariant derivative F i,j of the vector F i and form a contravariant vector
Gi = – ε ijk F j , k … (15)
r
we define the vector G to be the curl of F .
r dx i
Since n̂ . curl F = ni G = −ε F j ,k ni and the components of the unit tangent vector λ and
i ijk
.
ds
Then equation (13) may be written as


− εijk F j , k ni ds
= ∫ Fi
dx i
ds
ds … (16)
S C

r
∫ F dx
i
The integral i is called the circulation of F along the contour C.
c

8.13 GAUSS'S THEOREM


The integral of the normal component of the gravitational flux computed over a regular surface S
containing gravitating masses within it is equal to 4 πm where m is the total mass enclosed by S.
Proof: According to Newton's Law of gravitation, a particle P of mass m exerts on a particle Q of unit
m
mass located at a distancer r from P. Then a force of magnitude F = 2 .
r
Consider a closed regular surface S drawn around the point P and let θ be the angle between the
unit outward normal to n̂ to S and the axis of a cone with its vertex at P.. This cone subtends an
element of surface dS.
Analytical Mechanics 165

The flux of the gravitational field produced by m is


r m cos θ r 2 dw
∫ F .nˆ dS = ∫ r 2 cos θ
S S

2
r dw
where dS = and dw is the solid angle subtended by dS.
cos θ
Thus, we have,
r

F .nˆ dS
= ∫
m dw = 4 πm
… (1)
S S

n
F
θ

dS

P
m

Fig. 8.2.

If there are n discrete particles of masses mi located within S, then


r
n
mi cos θ i
F .nˆ = ∑
i =1 ri 2
and total flux is

r n

∫ F .nˆ dS = 4 π ∑m
i =1
i … (2)
S
The result (2) can be easily generalized to continuous distributions of matter whenever such
distribution no where melt the surface S.
The contribution to the flux integration from the mass element ρ dV contained within V, is
r cos θρ dV
∫ F .nˆ dS = ∫ r2
dS
S S

and the contribution from all masses contained easily within S is


 cos θ ρ dV 
r   dS
∫ F .nˆ dS = ∫ ∫  r2  … (3)
S V 
S
166 Tensors and Their Applications

where ∫ denotes the volume integral over all bodies interior to S. Since all masses are assumed to be
V
interior to S,r never vanishes. So that the integrand in equation (3) is continuous and one can interchange
to order of integration to obtain

r  cos θ dS 
∫ F .nˆ dS = ∫ ρ  ∫
s r2 
 dV … (4)
V
S

cos θ dS
But ∫r2
= 4π. Since it represents the flux due to a unit mass contained within S.
S
Hence
r
∫ F.nˆ dS = ∫
4 π ρ dV = 4 πm
… (5)
V
S
where m denotes the total mass contained within S. Proved.
Gauss's theorem may be extended to cases where the regular surface S cuts the masses, provided
that the density S is piecewise continuous.
Let S cut some masses. Let S' and S" be two nearby surfaces, the first of which lies wholly within
S and the other envelopes S. Now apply Gauss's theorem to calculate the total flux over S" produced by
the distribution of masses enclosed by S since S" does not intersect them.
We have
r
∫( F .nˆ )i dS = 4 πm
S"
r
where the subscript i on F ⋅ nˆ refers to the flux due to the masses located inside S and m is the total
mass within S. On the other hand, the net flux over S' due to the masses outside S, by Gauss's theorem
is
r

( F ⋅ nˆ)o ds = 0
S'
r
where the subscript o on F ⋅ nˆ refers to the flux due to the masses located outside S.
Now if we S' and S" approach S, we obtain the same formula (5) because the contribution to the
r

total flux from the integral (F ⋅ nˆ)o dS is zero.
S'

8.14 POISSON'S EQUATION


By divergence theorem, we have
r r
∫ F .nˆ ds = ∫
V
div F dV
S
and by Gauss's Theorem,
r
∫ F.nˆ ds = 4 π∫ ρ dVV
S
from these, we have
Analytical Mechanics 167

r
∫ (div F − 4πρ ) dV = 0
v
Since this relation is true for an arbitrary V and the integrand is piecewise continuous, then
r
div F = 4 πρ
By the definition of potential function V, we have
r
F = −∇V
and div ∇V = ∇ 2V
So,
r
div F = 4 πρ
div ( −∇V ) = 4 πρ

∇ 2V = – 4 πρ
which is equation of poisson.
If the point P is not occupied by the mass, then ρ = 0. Hence at all points of space free of matter
the potential function V satisfies Laplace's equation
∇ 2V = 0

8.15 SOLUTION OF POISSON’S EQUATION


We find the solution of Poisson’s Equation by using Green’s symmetrical formula. We know that
Green's symmetrical formula
 ∂ψ ∂φ 
∫ ∫
(φ∇ 2 ψ − ψ∇ 2 φ) dV =  φ ∂n − ψ ∂n  dS … (1)
S
V
where V is volume enclosed by S and φ and ψ are scalar point functions.
1
Put φ = where r is the distance between the points P ( x1 , x 2 , x3 ) and Q ( y1 , y 2 , y 3 ) and V is
r
the gravitational potential.
n

Q (y )

r
S
P(x)

Fig. 8.3.
168 Tensors and Their Applications

1
Since has a discontinuity at x i = y i , delete the point P(x) from region of integration by
r
surrounding it with a sphere of radius ε and volume V'. Apply Green’s symmetrical formula to the
1
region V – V' within which and V possess the desired properties of continuity..
r
2 1
In region V − V', ∇ φ = ∇ = 0.
2
r
Then equation (1) becomes

1 2  1 ∂ψ ∂ 1r   1 ∂ψ ∂ 1r 
∫ r
∇ ψ dV = ∫ 
 r ∂n

− ψ
∂n 
 dS +
∫ 
 r ∂n

− ψ
∂n 
 ds … (2)
V −V' S S'

where n̂ is the unit outward normal to the surface S + S' bounding V – V' . S' being the surface of the
∂ ∂
sphere of radius ε and =− .
∂n ∂r
Now

 1 ∂ψ ∂ 1r   1 ∂ψ ∂ 1r 
∫ 
 r ∂n

− ψ
∂n
 dS =


∫ −
 r ∂r

− ψ
∂r
 dS


S' S'

 1 ∂ψ ψ 2
= ∫  − r ∂r − r 2 

r dw
S'

 ∂ψ 
= − ∫  r ∂r + ψ  dw
S'

 1 ∂ψ ∂( 1r )  ∂ψ 
 = − ε 
∫ 
 r ∂n

− ψ
∂n   ∫
 dw − 4πψ
 ∂r  r = ε … (3)
S' S'

where ψ is the mean value of V over the sphere S' and w denote the solid angle.

Let ψ( x , x , x ) = ψ (P ) as r → 0 then as ε → 0 from (3), we have


1 2 3

 1 ∂ψ ∂ 1r 
∫ 
 r ∂n

− ψ 
∂n  = −4πψ (P )
S'

Then equation (2) becomes

1 2  1 ∂ψ ∂ 1r 
∫ r
∇ ψ dV = ∫ 
 r ∂n

− ψ
∂n 
 dS − 4πψ (P )
V S

1
∫ r ∇ ψ dV
2
Since ε → 0 then = 0.
V'
Analytical Mechanics 169

1 1 2 1 1 ∂ψ 1 ∂ 1r

ψ(P ) = 4 π r
V
∇ ψ dV∫+
4π S r ∂n
dS − ∫ψ
4π S ∂n
dS ∫ … (4)

This gives the solution of Poisson's equation at the origin.


If ψ is regular at infinity, i.e., for sufficiently large value of r, ψ is such that
m ∂ψ m
(ψ) ≤ and ≤ … (5)
r ∂r r 2
where m is constant.
If integration in equation (4) is extended over all space, so that r → ∞ . Then, using equation (5),
equation (4) becomes

1 ∇ 2ψ
ψ(P ) = − ∫
4π ∞ r
dV … (6)

But ψ is a potential function satisfying the Poisson's equation i.e. ∇ 2 ψ = −4πρ.


Hence, from (6), we get

dV
ψ(P ) = ∫ρ r

This solution is Unique.

EXERCISES

1. Find, with aid of Lagrangian equations, the trajectory of a particle moving in a uniform gravitational
field.
2. A particle is constrained to move under gravity along the line yi = ci s (i = 1, 2, 3). Discuss the motion.
3. Deduce from Newtonian equations the equation of energy T + V = h, where h is constant.
4. Prove that

∫ ψ n dS ∫ ∇ ψ dV
i 2
i =
S V

∂ψ
where ψ i = .
∂x i
5. Prove that the curl of a gradient vector vanishes identically.
CHAPTER – 9

CURVATURE OF CURVE, GEODESIC

9.1 CURVATURE OF CURVE: PRINCIPAL NORMAL


Let C be a curve in a given Vn and let the coordinates x i of the current point on the curve expressed
as functions of the arc length s. Then the unit tangent t to the curve the contravariant components
dxi
t i = ds ...(1)
The intrinsic derivative (or desired vector) of t i along the curve C is called the first curvaturee
r r
vector of curve C relative to Vn and is denoted by p . The magnitude of curvature vector p is called
first curvature of C relative Vn and is denoted by K:
So,

K= g ij p i p j
r
where P i are contravariant components of p so that
i dx j
P = t , j ds
i

 ∂t i  i
α  i   dx

=  j + t   
 ∂x  α j   ds

∂t i dx j α dx  i 
j

= + t  
∂x j ds ds  α j

dt i dxα dx j  i 
= ds + ds ds α j
 

d 2 xi dx j dx k  i 
= 2
+  , Replacing dummy index α by k
ds ds ds  k j
Curvature of Curve, Geodesic 171

d 2 xi dx j dx k  i   i   i 
pi = ds 2 + ds ds  j k  as  j k  = k j 
     
r
If n̂ is a unit vector in the direction of p , then we have
r
p = k nˆ
The vector n̂ is called the Unit principal normal.

9.2 GEODESICS
Geodesics on a surface in Euclidean three dimensional space may be defined as the curve along which
lies the shortest distance measured along the surface between any two points in its plane.
But when the problem of find the shortest distance between any two given points on a surface is
treated properly, it becomes very complicated and therefore we define the geodesics in V3 as follows:
(i) Geodesic in a surface is defined as the curve of stationary length on a surface between any
two points in its plane.
(ii) In V3 geodesic is also defined as the curve whose curvature relative to the surface is
everywhere zero.
By generalising these definitions we can define geodesic in Riemannian Vn as
(i) Geodesic in a Riemannian Vn is defined as the curve of minimum (or maximum) length
joining two points on it.
(ii) Geodesic is the curve whose first curvature relative to Vn is zero at all points.

9.3 EULER'S CONDITION


THEOREM 9.1 The Euler condition for the integral


t1
f (x i , x&i ) dt
to

to be staionary are
∂f d  ∂f 
−   =0
∂x i dt  ∂ x& i 
dxi
where x& = dt i = 1, 2, 3,...
i

Proof: Let C be a curve in a Vn and A, B two fixed points on it. The coordinates x i of the current point
P on C are functions of a single parameter t. Let t0 and t1 be the values of the parameter for the points
A and B respectively.
To find the condition for the integral


t1
f (x i , x&i ) dt ...(1)
t0

to be stationary.
Let the curve suffer an infinitesimal deformation to C ′, the points A and B remaining fixed while
the current points P(xi) is displaced to P' (xi + ηi) such that ηi = 0 at A and B both.
172 Tensors and Their Applications

P' C’

A B

P C

Fig. 9.1

In this case the value of integral (1) becomes I ′


So,

∫ F (x )
t1
I′ =
i
& i dt
+ ηi , x& i + η
t0

By Taylor's theorem

 ∂f ∂f 
F ( x + h , y + k ) = f ( x, y ) +  h + k  + ⋅ ⋅ ⋅ ⋅
 ∂x ∂y 
Then
t1   ∂F i ∂F i  
∫  F ( x , x& ) +  i η + i η&  + ⋅ ⋅ ⋅ dt
i i
I′ = t0  ∂x ∂x  
t1  ∂F ∂F i 
∫ ∫
t1
I′ = F ( x i , x& i ) dt +  ηi + η&  dt
t0 t 0  ∂x i ∂x& i 

(Neglecting higher order terms in small quantities ηi )


t1  ∂F ∂f 
I′ = I + ∫  i η i + i η& i  dt
t0  ∂ x ∂ x& 

t1  ∂F ∂F i 
δI = I ′ − I = ∫ 
t0  ∂ x i
ηi + η&  dt
∂ x&i 
...(2)

∂ zi & j
where η& i = x
∂x j
Now,

 ∂F i  1
t
t1 ∂ F d  ∂F  i

t1

∫ &
η =  ∂ x&i η  –  η dt
i
dt
&i
t0 ∂ x t0
t0 dt  ∂ x&i 

d  ∂F  i

t1
= −   η dt ...(3)
t0 dt  ∂ x&i 
Curvature of Curve, Geodesic 173

  ∂F
t
1 
since  i ηi (t ) = 0, η i (t1 ) = ηi (t0 ) = 0
  ∂x& t0 

Then equation (2) becomes


ti  ∂F d  ∂F   i
δI = ∫ 
t0  ∂ x i
−   η dt
dt  ∂ x&i   ...(4)
The integral I is stationary if δI = 0 .
t1  ∂F d  ∂F   i
i.e., if ∫ 
t0  ∂ x i
−   η dt = 0
dt  ∂ x&i  

Since ηi are arbitrary and hence the integrand of the last integral vanishes, so that

∂F d  ∂F 
−   = 0, (i = 1, 2,..., n) … (5)
∂ xi dt  ∂ x&i 
Hence the necessary and sufficient condition for the integral (1) to be stationary are

∂F d  ∂F 
–   = 0, (i = 1, 2, …, n)
∂x i dt  dx i 
These are called Euler's conditions for the integral I to be stationary.

9.4 DIFFERENTIAL EQUATIONS OF GEODESICS


To obtain the differential equations of a geodesic in a Vn , using the property that it is a path of minimum
(or maximum) length joining two points A and B on it.
Proof: Consider a curve C in Vn joining two fixed points A and B on it and x i (t ) be the coordinates of
point P on it.
The length of curve C is
dxi d x j

B
s= gij dt ...(1)
A dt dt
ds dxi d x j
= gij
dt dt dt
Put
ds d xi d x j
= gij = F (say) ...(2)
dt dt dt
or s& = g ij x& i x& j = F
Then equation (1) becomes


B
s= F dt ...(3)
A
174 Tensors and Their Applications

Since curve C is geodesic, then the integral (3) should be stationary, we have from Euler's
condition
∂F d  ∂F 
−   =0 ...(4)
∂x i dt  ∂ x& i 
Differentiating equation (2) with respect to x k and x& k we get,

∂F 1 ∂g ij i j
= x& x&
∂x k
2 s& ∂ x k

∂F  1  1
and k =
2 gik x& i  = g ik x&i
∂x&  2s&  s&

d  ∂F  1 1 ∂ g ik j i 1
 = − 2 &s& g ik x& + x& x& + g ik &x& i
i

dt  ∂ x& k  s& s& ∂x j s&
Putting these values in equation (4), we get

1 ∂g ij i j  1 1 ∂ g ik j i 1 
x& x& − − 2 &s&g ik x& i + x& x& + g ik &x&i  = 0
2 s& ∂ x k  s &
s ∂ x j &
s 

&s&  ∂g 1 ∂ g ij  i j
g ik &x&i − g ik x& i +  ikj −  x& x& = 0
s&  ∂x 2 ∂x k 

&s&
g ik &x&i − g x& i + [k , ij ] x&i x& j = 0
&s ik
multiplying it by g km , we get

&s& km
g km gik &x&i − g g ik x&i + g km [k , ij ] x& i x& j = 0
s&

m
g km gik = δim and g [k , ij] = i j
km
But
 

&s& m  m  i j
x&&m − x& +   x& x& = 0
s& i j 

d 2xm &s& dx m  m  dx j dxk  Replacing dummy 


− +  = 0  index i by k  ...(5)
dt 2
s& dt  j k  dt dt  
This is the differential equation for the geodesic in parameter t.
Taking s = t , s& = 1, &s& = 0. Then equation (5) becomes

d 2xm  m  dx j dx k
+  = 0 ...(6)
ds 2  j k  ds ds
Curvature of Curve, Geodesic 175

which may also written as

dx k  dx m 
 
ds  ds  =0
  ,k
Then the intrinsic derivative (or derived vector) of the unit tangent to a geodesic in the direction of the
curve is everywhere zero. In otherwords, a geodesic of Vn is a line whose first curvature relative to
Vn is identically zero.

THEOREM 9.2 To prove that one and only one geodesic passes through two specified points lying in
a neighbourhood of a point O of a Vn .
OR
To prove that one and only one geodesic passes through a specified point O of Vn in a prescribed
direction.
Proof: The differential equations of a geodesic curve in a Vn are

d 2xm  m  dx j dx k
+  =0
ds 2  j k  ds ds
These equations are n differential equations of the second order. Their complete integral involves 2n
arbitrary constants. These may be determined by the n coordinates of a point P on the curve and the n
components of the unit vector in the direction of the curve at P. Thus, in general, one and only one
geodesic passes through a given point in a given direction.

9.5 GEODESIC COORDINATES


A cartesian coordinate system is one relative to which the coefficients of the fundamental form are
constants. Coordinates of this nature do not exists for an arbitrary Riemannian V n.. It is, however,
possible to choose a coordinate system relative to which the quantities gij are locally constant in the
neighbourhood of an arbitrary point P 0 of V n. Such a cartesian coordinate system is known as geodesic
coordinate system with the pole at P 0.
The quantities g ij are said to be locally constants in the neighbourhood of a point P0 if
∂g ij
= 0 at P0
∂ xk
∂g ij
and ≠ 0 elsewhere
∂ xk
k 
This shows that [ij, k ] = 0 , i j = 0 at P0 .
 
Since the covariant derivative of A ij with respect to xk is written as

∂Aij  h   h 
Aij , k = ∂x k −   Aih −   A h j , see pg. 71
 j k  i k 
176 Tensors and Their Applications

The covariant derivative of Aij at P0 with respect to x k reduces to the corresponding ordinary
derivatives. Hence
∂Aij
Aij , k = at P0
∂x k
THEOREM 9.3 The necessary and sufficient condition that a system of coordintes be geodesic with
pole at P0 are that their second covariant derivatives with respect to the metric of the space all vanish
at P0.
Proof: We know that (equation 8, Pg. 65)

∂2xs ∂x s  k  ∂x p ∂x q  s 
=  −  
∂x j ∂ x j ∂x k i j  ∂x i ∂x j  p q 

∂  ∂x s  ∂x s  k  ∂x p ∂x q  s 
or   =  −   ...(1)
∂x j  ∂x i 
 ∂x k i j ∂x i ∂x j  p q

Interchanging the coordinate system x i and x i in equation (1), we get

∂  ∂x s  ∂x s  k  ∂x p ∂x q  s 
  =  − i j  
∂x j  ∂x i 
 ∂x k i j ∂x ∂ x  p q 

∂x p ∂x q  s  ∂  ∂x s  ∂x s k 
− j   =  −  
∂x ∂x  p q
i
∂x j  ∂x i  ∂x k
  i j

=
∂x

j
(x )− x
s
,i
s
,k
k  ∂x s
  since k = x, k at P0

s

 
i j x
k 
= ( x ,is ), j since  i j = 0 at P0
 
Thus,

∂x p ∂x q  s 
x,sij = −   ...(2)
∂x i ∂x j  p q
Necessary Condition
Let x s be a geodesic coordinate system with the pole at P0 so that
 s 
  = 0 at P0
 p q
Hence from (2), we have
x,sij = 0 at P0
Sufficient Condition
Conversely suppose that x,sij = 0 at P0 .
Curvature of Curve, Geodesic 177

Then equation (2) becomes


 s  ∂x p ∂x q
  i j = 0
 p q ∂x ∂x
 s  ∂x p ∂x q
⇒   = 0 at P0 , as ≠ 0 and ≠ 0 at P0
 p q ∂x i ∂x j
So, x s is a geodesic coordinate system with the pole at P 0.

9.6 RIEMANNIAN COORDINATES


A particular type of geodesic coordinates introduced by Riemann and known as Riemannian coordinates.
Let C be any geodesic through a given point P0 , s the length of the curve measured from P0 and ξ i the
quantities defined by
 dxi 
 
ξ =  ds
i
 ...(1)
 o
the subscript zero indicating as usual that the function is to be evaluated at P0 . The quantities ξi
represents that only one geodesic will pass through P 0 in the direction of ξi in V n. Let yi be the
coordinates of a point P on the geodesic C such that
yi = sξi ...(2)
where s is the arc length of the curve from P0 to P . The coordinates y i are called Riemannian
coordinates.
The differential equation of geodesic C in terms of coordinates y i relative to Vn is given by

d 2 y i  i  dy i dy j
+  =0 ...(3)
ds 2  j k  ds ds
 i 
where   is a christoffel symbol relative to the coordinates y i .
 j k
The differential equation (3) will be satisfied by (2), we have,

 i  i j i
0+  ξ ξ = 0 since dy = ξ i
 j k  ds

 i  i j
or  ξ ξ = 0 ...(4)
 j k
using equation (2), equation (4) becomes

 i  yi y j yi
  = 0 as = ξi
 j k s s s

 i  i j
or  y y = 0 ...(5)
 j k
The equation (5) hold throughout the Riemannian Vn .
178 Tensors and Their Applications

Since y i ≠ 0, y j ≠ 0, from (5) we get


 i 
  = 0 at P0
 j k
Hence the Riemannian coordinates are geodesic coordinate with the pole at P0 .
THEOREM 9.4 The necessary and sufficient condition that the coordinates y i be Riemannian coordinates
 i  i j
is that   y y = 0 hold throughout the Riemannian Vn .
 j k
 i  i j
Proof: If y are Riemannian coordinates then the condition 
i  y y = 0 (from equation 5) throughout
 j k 

the Riemannian Vn .

 i  i j d 2 y  i  dy i dy j
Conversely if  j k  y y = 0 hold then ds 2 +   = 0 are saitsfied by y i = s ξi .
   j k  ds ds
Hence yi are Riemannian coordinates.

9.7 GEODESIC FORM OF A LINE ELEMENT


Let φ be a scalar invariant whose gradiant is not zero. Let the hypersurface φ = 0 be taken as coordinates
hypersurface x1 = 0 and the geodesics which cut this hypersurface orthogonally as the coordinate
lines of parameter x1 , this parameter measuring the length of arc along a geodesic from the hypersurface
x1 = 0 .
Since dx1 is the length of the vector µi is given by
i j
u 2 = g ij u u
i.e., (dx )
1 2 1 1
= g11dx dx
⇒ g 11 = 1 ...(1)
Now, if v i is the tangent vector to the hypersurface x1 = 0 then we have
2 3 n
v i = ( 0, dx , dx ,..., dx )
since the vectors ui and v i are orthogonal vectors.
Then,
g ij u i v j = 0

⇒ g 1 j u 1 v j = 0, [u i = 0, i = 2, 3, ..., n ]

⇒ g 1 j v j = 0, as u 1 ≠ 0.

⇒ g1 j = 0, for j = 2, 3, …, n. ...(2)
Again the coordinate curves of parameter x1 are geodesics. Then s = x1.
If ti is unit tangent vector to a geodesic at any point then
t1 = 1 and t i = 0, for i = 2, 3, …, n.
Curvature of Curve, Geodesic 179

Now,

d xi d x i
ti = = 1
ds dx
dx1 dxi
⇒ = 1 and = 0 for i ≠ 1
dx1 dx1
d 2xi d 2xi
and = = 0 for i = 1, 2,..., n
ds 2 dx12
Also, the differential equation of geodesic is

d 2 x i  i  dx j dx k
+  =0
ds 2  j k  ds ds
using above results, we have

 i  dx 1 dx1
  =0
11 ds ds
i
⇒   =0
1 1
⇒ g ij [11, j ] = 0
⇒ [11, j ] = 0 as g ≠ 0
ij

1  2 ∂g 1 j ∂ g 11 
−  = 0, since g 11 = i ⇒ ∂g 11 = 0
2  ∂ x i ∂x j 
 ∂x j
So,
∂g 1 j
= 0 for j ≠ 1 ...(3)
∂x1

from equations (1), (2), and (3), we have


∂ g1 j
g 11 = 1, g1j = 0; ( j = 2, 3, …, n), = 0, ( j = 2, 3, …, n)
∂ x1
The line element is given by
i j
ds 2 = gij dx dx

ds 2 = g 11dx dx + g jk dx dx
1 1 j k

ds 2 = ( dx ) + g jk dx dx ; ( j = 2, 3, …, n, k = 2, 3, …, n)
1 2 j k
...(4)
The line element (4) is called geodesic form of the line element.
Note 1: We note that the coordinate curves with parameter x 1 are orthogonal to the coordinate curve
x i = c i (i = 1, 2, ..., n) at all points and hence to the hypersurfaces x 1 = c at each point.
Note 2: The existence of geodesic form of the line element proves that the hypersurfaces φ= x1 = constant form a
system of parallels i.e., the hypersurfaces φ = x 1 = constant are geodesically parallel hypersurfaces.
180 Tensors and Their Applications

THEOREM 9.5 The necessary and sufficient condition that the hypersurfaces φ = constant form a
system of parallel is that (∇φ)2 = 1.
Proof: Necessary Condition
Suppose that hypersurface φ = constant form a system of parallels then prove that (∇φ 2) = 1.
Let us take the hypersurface φ = 0 as the coordinate hypersurface x1 = 0. Let the geodesics
cutting this hypersurface orthogonally, be taken as coordinate lines of parameter x1. Then the parameters
x1 measures are length along these geodesics from the hypersurface x1 = 0. This implies the existence
of geodesic form of the line element namely
ds 2 = ( dx ) + g ij dx dx
1 2 i j
...(1)
where i, j = 2, 3,..., n.
From (1), we have
g 11 = 1, g 1i = 0 for i ≠ 1.
from these values, it follows that
g11 = 1, g1 i = 0, for i ≠ 1
Now,
∂φ ∂φ
(∇φ)2 = ∇φ ⋅ ∇φ = g
ij

∂x i dx j

∂ x1 ∂x1
= g
ij
= g ij δ1i δ1j
∂ x ∂x
i j

so (∇φ)2 = g11 = 1
(∇φ)2 = 1
Sufficient Condition
Suppose that (∇φ)2 = 1 then prove that the hypersurface φ = constant from a system of parallels.
Let us taken φ = x1 and orthogonal trajectories of the hypersurfaces φ = x1 constant as the
coordinate lines of parameter x1. Then the hypersurfaces
x1 = constant
xi = constant (i ≠ 1) are orthogonal to each other. The condition for this g1i = 0 for i ≠ 1 .
Now, given that (∇φ)2 = 1

∂φ ∂φ
⇒ g ij =1
∂x i ∂x j

∂x1 ∂x1
⇒ g ij =1
∂x i ∂x j

⇒ g ij δ1i δ1j = 1

⇒ g11 = 1
Curvature of Curve, Geodesic 181

Thus
g11 = 1 and g1 i = 0 for i ≠ 1.
Consequently
g 11 = 1, g = 0 , for i ≠ 1.
1i
Therefore, the line element
i j
ds 2 = g ij dx dx
is given by
ds 2 = (dx1 ) 2 + g ik dx i dx k ; (i, k = 2, 3, …, n)

which is geodesic form of the line element. It means that the hypersurfaces φ = x1 = constant
form a system of parallels.

9.8 GEODESICS IN EUCLIDEAN SPACE


Consider an Euclidean space S n for n-dimensions. Let y i be the Euclidean coordinates. The differential
equation of geodesics in Euclidean space is given by
d 2 y i  i  dy j dy k
+  =0 ...(1)
ds 2  j k  ds ds

In case of Euclidean coordinates the fundamental tensor g ij is denoted by aij and

g ij = a ij = δij = 1, if i = j
0, if i ≠ j
∂g ij ∂a ij
= =0
∂x k
∂x k

This implies that   = 0, [ij, k ] = 0 relative to S n .


k
 j
i
Then equation (1) becomes
d 2 yi
=0
ds 2
Integrating it, we get
dyi
= ai, where ai is constant of integration.
ds
Again Integrating, we get
y i = ais + bi, where bi is constant of integration ...(2)
The equation (2) is of the form y = mx + c.
Hence equation (2) represents a straight line. Since equation (2) is a solution of equation (1) and
therefore the geodesic relative to S n are given by equation (2). Hence geodesic curves in Euclidean
space S n are straight lines.
182 Tensors and Their Applications

THEOREM 9.6 Prove that the distance l between two points P (y i) and Q (y' i ) in Sn is given by

∑ (y ′ − y )
n
i i 2
l=
i =1

Proof: We know that geodesics in S n are straight line. Then equation of straight line in S n may be
taken as
yi = ais + bi ...(1)
Let P ( y i ) and Q ( y' i) lie on equation (1). Then
yi = ais + bi, y′ i = a i s′ + bi
y′ i − y i = a i (s ′ − s ) ...(2)
Then equation (2) becomes
y′ i − y i = a i l
n n
l2 ∑
i =1
(ai ) 2 = ∑ ( y′
i =1
i
− y i )2

But a i is the unit tangent vector to the geodesics. Then


n

∑ (a )
i =1
i 2
=1
So,
n

l 2 = ∑ ( y′
i =1
i
− y i )2

l= ∑( y′ i
− y i )2
i =1

EXAMPLE 1
Prove that Pythagoras theorem holds in S n .
Solution
Consider a triangle ABC right angled at A i.e., ∠BAC = 90 o .

C (y i3)

A ( y1i ) B ( y2i )
Fig. 9.2
Curvature of Curve, Geodesic 183

Then the lines AB and AC are orthogonal to each other. So,


AB ⋅ AC = 0

or a ij ( y i2 − y1i ) ( y i3 − y1i ) = 0

or ∑ (y
i =1
i
2 − y1i ) ( y 3i − y1i ) = 0 ...(1)

By distance formula, we have


n

(AB) = 2
∑ (y
i =1
i
2 − y1i ) 2 ...(2)

(AC) = 2
∑ (y
i =1
i
3 − y1i )2 ...(3)

(BC) = 2
∑ (y
i =1
i
3 − y i2 )2 ...(4)

Now, equation (4) can be written as


n

(BC) = 2
∑ [( y
i =1
i
3 − y1i ) + ( y1i − y i2 )]2

= ∑ [( y
i =1
i
3 − y1i )2 + ( y1i − y i2 ) 2 + 2( y3i − y1i ) ( y1i − y i2 ) ]

n n
= ∑
i =1
( y i3 − y1i )2 + ∑ (y
i =1
i
1 − y i2 ) 2 + 2 × 0, [from (1)]

n n

= ∑
i =1
( y i3 − y1i )2 + ∑( y
i =1
i
1 − y2i )2

(BC)2 = ( AC)2 + ( AB)2


Hence Pythagoras theorem holds in S n .

EXAMPLE 2
Prove that if θ is any solution of the differential equation (∇θ)2 = f (θ) then the hypersurfaces θ =
constant constitute a system of parallels.

Solution
Given that
(∇θ)2 = f (θ) ...(1)
184 Tensors and Their Applications

Then prove that the hypersurfaces θ = constant form a system of parallel.


Suppose
dθ dθ
φ= ∫ f ( θ)
, Then, dφ =
f (θ)

dφ 1
or =
dθ f ( θ)

∂φ ∂φ ∂θ 1
Now, ∇φ = i = = ∇θ
∂x ∂θ ∂x i f ( θ)

2
 1 
(∇φ)2 =  ∇θ 

 f ( θ) 
1 1
= (∇ θ) 2 = f (θ); from (1)
f ( θ) f ( θ)
(∇φ)2 = 1
This proves that the hypersurfaces φ = constant form a system of parallels and therefore the
hypersurfaces θ = constant.

EXAMPLE 3
Show that it is always possible to choose a geodesic coordinates system for any Vn with an
arbitrary pole P 0.
Solution
Let P0 be an arbitrary pole in a Vn . Let us consider general coordinate system x i . suppose the
value of x i and P0 are denoted by x0i . Now consider a new coordinate system x j defined by the
equation.
1  h  i
= a m ( x − x0 ) + a h   ( x − x0 ) ( x − x0 )
j j m m j l m m
x ...(1)
2 l m 
The coefficients amj being constants and as such that their determinant do not vanish.
Now we shall prove that this new system of coordinated x j defined by equation (1) is a geodesic
coordinate system with pole at P0 i.e., second covariant derivative of x j vanishes at P..
Differentiating equation (1) with respect to x m , we get
∂x j 1 h 
= a m + ah   2 ( x − x0 )
j j i l
...(2)
∂x m 2  l m 

 ∂x j 
  = a j at P ...(3)
 ∂x m  m 0
 0
Now, the Jacobian determinant
Curvature of Curve, Geodesic 185

 ∂x j 
 m = a j ≠ 0
∂x  m
 0
and therefore the transformation given by equation (1) is permissible in the neighbourhood of P0 .
Differentiating equation (2) with respect to x j , we get

 ∂2 x j 
 j m  = a j  h 
 ∂x ∂x  h ...(4)
 0 l m  0
But we know that

 ∂2x  h   ∂x j 
(x )
j
j
=  l  −
  
 ∂x h


 ∂x ∂x  0  l m  0
,lm 0 m
 0

h  h  j
 −   a h , (from (3) and (4))
j
= ah 
l m  0 lm 

(x , )
j
lm 0 =0
Hence equation (1) is a geodesic coordinate system with pole at P0 .

EXAMPLE 4
If the coordinates xi of points on a geodesic are functions of arc lengths s and φ is any scalar
function of the x's show that
d pφ d xi d x j dxl
= φ, ij... l ⋅⋅ ⋅ ...(1)
ds p ds ds ds
Solution
Since the coordinates x i lie on a geodesic. Then

d 2 xi  i  dx j dx k
+  =0 ...(2)
ds 2  j k  ds ds
Here the number of suffices i j...l is p.
We shall prove the theorem by mathematical induction method.
Since x′s are functions of s and φ is a scalar function of x′ s , we have

dφ ∂φ dxi dφ dx i
= or = φ, i ...(3)
ds ∂x i ds ds ds
d 2φ ∂φ, i dx j d 2xi
= + φ ,i
ds2 ∂ x j ds ds 2
∂φ ,i dx i dx j  i  d x j d x k
= ∂x j ds ds − φ ,i   , from (2)
 j k  ds ds
186 Tensors and Their Applications

∂φ ,i dx i dx j  m  dx j dx k
= ∂x i ds ds − φ, m  
 j k  ds ds

∂φ, i dx i dx j  m  dxi dx j
= − φ , m   (adjusting the dummy index.)
∂ x j ds ds i j  ds ds
 ∂φ, i  m   dx i dx j

=  j − φ,m  
i j  ds ds
...(4)
 ∂x
Equations (3) and (4) imply that the equation (1) holds for p = 1 and p = 2.
Suppose that the equation (1) holds for p indices r1 , r2 ,...,r p so that
r
d pφ dx r1 dx r1 dx p
= φ, r1 , r2 , ...,
⋅ ⋅ ⋅ rp ...(5)
ds p ds ds ds
Differentiating the equation (5) with respect to s, we get

d p +1 φ ∂φ, r1 r2 ... r p dx r1 r r
dx p dx p +1 d 2 x r1 dx r2
r
dx p
= ⋅ ⋅ ⋅ + φ, ⋅ ⋅ ⋅ + ⋅ ⋅⋅
ds p +1
r r1 r2 ... rp
∂ x p +1 ds ds ds ds 2 ds ds
r
dx r1 d 2x p
+ ⋅ ⋅ ⋅φ, r1 r2 ... rp ⋅ ⋅⋅ ...(6)
ds ds 2
d 2 x r1
substituting value of etc. from (2) in (6) and adjusting dummy indices, we have
ds 2
 ∂ φ, r r ... r m  m  
d p +1 φ  − φ ⋅ ⋅ ⋅ φ 
   
1 2 p

p +1 =  ∂x p +1
r , m r ... r ,r r ... m
   
2 p 1 2
r r
 1 p +1  r r
 p p +1  
ds 
r
dx r1 dx r2 dx p +1
⋅⋅⋅
ds ds ds
r
d x r1 d x r2 d x p +1
= φ ,r1 r2 .. r p rp +1
⋅⋅ ⋅
ds ds ds
This shows that the equation (1) holds for next values of p. But equation (1) holds for p = 1, 2,
... Hence equation (1) holds for all values of p.

EXERCISES

1. Prove that at the pole of a geodesic coordinate system, the components of first covariant derivatives
are ordinary derivatives.
2. If x i are geodesic coordinates in the neighbourhood of a point if they are subjected to the
transformation
1 i
xi = x +
i
c x j xk xl
6 jkl
Curvature of Curve, Geodesic 187

where C′s are constants then show that xi are geodesic coordinates in the neighbourhood of O.
3. Show that the principal normal vector vanishes identically when the given curve is geodesic.
4. Show that the coordinate system x i defined by
1 i  j k
x i = x + 2  j k x x
i

 
is geodesic coordinate system with the pole at the origin.
5. Obtain the equations of geodesics for the metric
ds2 = e−2 kt ( dx2 + dy2 + dz2 ) + dt 2
6. Obtain the differential equations of geodesics for the metric
1
ds2 = f ( x) dx + dy + dz + f ( x) dt
2 2 2 2

 d2x 1 d  dx 
2
1 d dt d2y d2z d 2t d (log f ) dx dt 
Ans : 2 − (log f )   + (log f ) = 0; = 0; = 0; − = 0
 ds 2 dx  ds  2 f 2 dx ds ds2 ds2 ds2 dx ds ds 

7. Find the differential equations for the geodesics in a cylindrical and spherical coordinates.
8. Find the rate of divergence of a given curve C from the geodesic which touches it at a given point.
CHAPTER – 10

PARALLELISM OF VECTORS

10.1 PARALLELISM OF A VECTOR OF CONSTANT MAGNITUDE (LEVI-CIVITA’S


CONCEPT)
Consider a vector field whose direction at any point is that of the Unit Vector ti. In ordinary space, the
field is said to be parallel if the derivative of ti vanishes for all directions ui (say) and at every point of
the field i.e.,
∂t i j
u =0
∂x j
Similarly in a Riemannian V n the field is said to be parallel if the derived vector of ti vanishes at
each point for every direction ui at each point of V n. i.e.,
t,i j = u j = 0
It can be shown that it is not possible for an arbitrary V n. Consequently we define parallelism of
vectors with respect to a given curve C in a V n.
A vector ui of constant magnitude is parallel with respect to V n along the curve C if its derived
vector in the direction of the curve is zero at all points of C i.e.,
dx j i
u =0 … (1)
ds , j
where s is arc-length of curve C.
The equation (1) can be written in expansion form as
 ∂ui 
m  i  dx
j

 j + u m j =0
∂ x   ds

∂ui d x j m i  dx
j
+ u   =0
∂ x j ds m j ds
du i  i  dx j
+ um  
j  ds = 0 … (2)
ds m
Parallelism of Vectors 189

This concept of parallelism is due to Levi-Civita. The vector ui is satisfying the equation (1) is
said to a parallel displacement along the curve
Now, multiplying equation (1) by gil, we get
 dx j 
g il  u,i j 
 ds  = 0

dx j
( gil u,i j ) =0
ds

dx j
( g il u i ) , j =0
ds

dx j
ul , j
ds = 0

dx j
or u i, j =0
ds

 ∂ui  m  dx j
 j − u m   =0
 ∂x i j  ds

dui  m  dx j
− um   =0 … (3)
ds i j  ds
The equation (2) and (3) can be also written as
m i 
dui = − u m j  d x
j
… (4)
 
m  j
and dui = um i j  d x … (5)
 
The equation (4) and (5) give the increment in the components ui and ui respectively due to
displacement dxj along C.
THEOREM 10.1 If two vectors of constant magnitudes undergo parallel displacements along a given
curve then they are inclined at a constant angle.
Proof: Let the vectors ui and v i be of constant magnitudes and undergo parallel displacement along a
curve C, we have (from equation (1), Pg. 188.)
dx j 
u,i j = 0
ds 
j 
d x … (1)
vi, j = 0
ds 
at each point of C.
Multiplying (1) by gil, we get
dx j
( g il u,i j ) =0
ds
190 Tensors and Their Applications

dx j
ul , j =0
ds

dx j
or u i, j =0 … (2)
ds
Similarly,
dx j
vi , j =0 … (3)
ds
Let φ be the angle between ui and v i then
ui.v i = uv cos θ
Differentiating it with respect to arc length s, we get
d d (u i vi )
(uv cos θ) =
ds ds

i dx j
= (u v i ), j
ds

dθ i dx
j
dx j
− uv sin θ = u, j vi + u i vi , j … (4)
ds ds ds
Using equation (1) and (3), then equation (4) becomes

− uv sin θ =0
ds

⇒ sin θ = 0, as u ≠ 0, v ≠ 0
ds

⇒ Either sinθ = 0 or =0
ds

⇒ Either θ = 0 or θ = constant.
⇒ θ is constant. Since 0 is also a constant.
THEOREM 10.2 A geodesic is an auto-parallel curve.
Proof: The differential equation of the geodesic is given by (See Pg. 174, eqn. 6)

d 2 x m  m  dx j dx k
+  =0
ds 2  j k  ds ds

d  d x m   m  d x j d x k
+ 
ds  ds   j k  ds ds = 0

∂  dx m  d x j  m  d x j d xk
 
∂x j  ds  ds +  j k  ds ds = 0
   
Parallelism of Vectors 191

 ∂  dx m   m  dx k  dx j
 j  +   =0
 ds 
 ∂x    j k  ds  ds

 d xm  dx j dx j
  =0
 ds  ds = 0 or t,mj
 , j ds
dx m
This shows that the unit tangent vector suffer a parallel displacement along a geodesic curve.This
ds
confirms that geodesic is an auto-parallel curve. Proved.

10.2 PARALLELISM OF A VECTOR OF VARIABLE MAGNITUDE


Two vectors at a point are said to be parallel or to have the same direction if their corresponding
components are proportional. Consequently the vector v i will be parallel to ui at each point of curve C
provided
v i = φui … (1)
where φ is a function of arc length s.
If ui is parallel with respect to Riemannian V n along the curve C. Then,
dx j
u ,ij =0 … (2)
ds
The equation (1) shows that v i is of variable constant and parallel with respect to Riemannian V n
so that

dx j dxj
v ,ij i
= (φ u ), j
ds ds

dx j
= (φ, j u + φu, j )
i i
ds

dx j i dxj
= φ, j u + φ u,i j
ds ds

dx j i dx j
= φ, j u Since u,i j =0
ds ds

∂φ d x j i
= u
∂x j ds

dx j dφ i
v,i j = u
ds ds

d φ vi
= ds from (1)
φ
192 Tensors and Their Applications

i d (log φ)
= v
ds
dx j d (log φ )
v,i j = v i f (s) where f (s) = … (3)
ds ds
Hence a vector v i of variable magnitude will be parallel with respect to V n if equation (3) is satisfied.
Conversely suppose that a vector v i of variable magnitude such that
dx j
v,ij = = v i f (s )
ds
to show that v i is parallel, with respect to V n .
Take
ui = v i ψ (s) … (5)
Then
dx j dx j
u ,ij = (v ψ), j
i
ds ds
j
i dx dx j i
= v, j ψ + ψ, j v
ds ds
∂ψ dx j i
= v i
f ( s ) ψ + v
∂x j ds
dx j i dψ 
u,i j = v  ψf (s ) +
ds 
… (6)
ds 

Select ψ such that ψf ( s ) + = 0.
ds
Then equation (6) becomes
dx j
u,i j =0
ds
This equation shows that the vector u i is of constant magnitude and suffers a parallel displacement
along curve C. The equation (5) shows that v i is parallel along C.
Hence necessary and sufficient condition that a vector v i of variable magnitude suffers a parallel
displacement along a curve C is that
dx j
v,i j = v i f (s ).
ds
EXAMPLE 1
Show that the vector v i of variable magnitude suffers a parallel displacement along a curve C if
and only if
dx k
(v i v i,k − v i v l,k ) = 0, i = 1, 2, ..., n.
ds
Solution
From equation (4), we have
dx j
v ,ij = v i f (s ) … (1)
ds
Parallelism of Vectors 193

Multiplying by v l, we get
dx j
v l v,i j
= v l v i f (s )
ds
Interchange the indices l and i, we get
dx j
v i v,l j
= v i v l f (s) … (2)
ds
Subtract (1) and (2), we get
dx k
(v l u,ik − v i v l,k ) = 0 by interchanging dummy indices j and k.
ds

10.3 SUBSPACES OF A RIEMANNIAN MANIFOLD

Let V n be Riemannian space of n dimensions referred to coordinates xi and having the metric
ds2 = gij dxi dxj. Let V m be Riemannian space of m dimensions referred to coordinates y α and having
the metric ds2 = axβ dyα dyβ, where m > n. Let Greek letters α , β, γ take the values 1, 2, ..., m and Latin
indices i, j, k ... take the values 1, 2, … n.
If the n independent variables xi are such that the coordinates ( y α ) of points in V m are expressed
as a function of xi then V n is immersed in V m i.e. V n is a subspace of V m. Also V m is called enveloping
space of V n.
Since the length ds of the element of arc connecting the two points is the same with respect to V n
or V m. it follows that
gij dxi dxj = a αβdy α dy β
∂y α ∂y β i j
gij dxi dxj = a αβ dx dx
∂x i ∂x j
∂y α ∂y β
⇒ a
gij = αβ i … (1)
∂x ∂x j
As dxi and dxj are arbitrary.
This gives relation between gij and aαβ .

THEOREM 10.3 To show that the angle between any two vectors is the same whether it is calculated
with respect to Vm or Vn.
Proof: Consider two vectors dx i and δx j defined at any point of V n and suppose that the same vectors
in V m are represented by dy α and δy α respectively. If θ is the angle between dxi and δx i then

g ij dx i δx j
cos θ = … (1)
g ij dx i dx j g ij δx i δx j
If φ is the angle between the vectors dyα and δyα then

aαβ dy αδy β
cos φ =
aαβ dy α dy β aαβ δy αδy β
194 Tensors and Their Applications

∂ y α i ∂ yβ
aαβ dx dx j
∂x i ∂x j
=
∂ yα i ∂ yβ j ∂ yα i ∂ yβ j
aαβ d x d x a αβ δx δx .
∂x i ∂x j ∂ xi ∂x j

∂y α ∂y β i j
aαβ dx dx
∂x i ∂x j
=
∂y α ∂y β i j ∂y α ∂y β i j
aαβ dx dx a αβ δx δx .
∂xi ∂x j ∂x i ∂x j

gij dx i dx j
cos φ = … (2)
g ij dx i dx j g ij δx i δx j

∂y α ∂y β
Since g ij = aαβ (from equation (1), art. 10.3)
∂x i ∂x j
from (1) & (2)
cos θ = cos φ ⇒ θ = φ. Proved.
THEOREM 10.4 If Uα and ui denote the components of the same vector in Riemannian Vm and Vn
respectively then to show that
α
i ∂y
α = u
U ∂x i
Proof: Let the given vector be unit vector at any point P of V n. Let the component of the same vector
x's and y's be a i and A α respectively. Let C be curve passing through s in the direction of the given
vector then
∂y α ∂y α dx i ∂y α i
A =α = i = i a … (1)
∂s ∂x ds ∂x
But the components of a vector of magnitude a are a times the corresponding components of the
Unit vector in the same direction. Then
α
U α = aA , u = aa
i i … (2)
Multiplying (1) by a, we get
∂y α
aA α = (aa i )
∂x i
∂y α i
Or α = u , using (2) Proved.
U ∂x i
THEOREM 10.5 To show that there are m-n linearly vector fields normal to a surface V n immersed in
a Riemannian V m.
Proof: Since V n is immersed in V m, the coordinates y α of points in V m are expressible as functions of
coordinates xi in V n.
Parallelism of Vectors 195

Now,
dy α ∂y α dx i
= ...(1)
ds ∂x i ds
i
for the curve x = s, we have
dy α ∂y α
= ...(2)
ds ∂x i
dy α ∂ yα
Since is a vector tangential to the curve in V m. Then from equation (2) it follows that
ds ∂ xi
i α
in V m is tangential to the coordinate curve of transmeter x in V n. Let the Unit vectors N in V m be
normal to each of the above vector fields of V n then
∂y α β
a αβ N =0
∂xi
∂y α
⇒ (aαβ N β ) i = 0 ⇒ a αβ N β y ,αi = 0 ...(3)
∂x
∂y α
⇒ N α i = 0, i = 1, 2, …, n & α = 1, 2, …, m
∂x
 ∂y α 
The equation (3) are n equations in m unknowns N α (m > n). The coefficient matrix  i  is

 ∂x 
of order m × n and the rank of this matrix is n.
It means that there will be only m – n linearly independent solution of N α . This shows that there
are m – n linearly independent normals to V n to V m.

EXAMPLE 2
Show that
∂y α ∂y β ∂y γ ∂ 2 y α ∂x β
[ij, k] = [αβ, γ ] + a αβ
∂x i ∂x j ∂x k ∂xi ∂x j ∂x k
where [αβ, γ] and [i j, k ] are the Christoffel’s symbols of first kind relative to metrics a αβ dy a dy β and
g ij dx i dx j .
Solution

Since relation between a αβ and g ij is given by


∂y α ∂y β
g ij = a αβ
∂x i ∂x j
Differentiating it with respect to xk, we get
∂g ij ∂aαβ ∂y α ∂y β ∂y γ ∂ 2 y α ∂y γ ∂y α ∂ 2 y β
= + a αβ
+ a αβ … (1)
∂x k ∂x γ ∂x i ∂x j ∂x k ∂x i ∂x k ∂x j ∂x i ∂x j ∂x k
196 Tensors and Their Applications

Similarly
∂g jk ∂aβγ ∂y α ∂y β ∂y γ ∂ 2 y α ∂y β ∂y α ∂ 2 yβ
= + a αβ + a αβ … (2)
∂x i ∂y α ∂x i ∂x j ∂x k ∂x j ∂x i ∂x k ∂x j ∂x k ∂x i
and
∂g ki ∂a γα ∂y α ∂y β ∂y γ ∂ 2 y α ∂y β ∂y α ∂ 2 y β
= + a αβ
+ a αβ … (3)
∂x j ∂x β ∂x i ∂x j ∂x k ∂x k ∂x j ∂x i ∂x k ∂x i ∂x j
But we know that
1  ∂g jk ∂g ki ∂g ij 
[ij, k ] =  + −  … (4)
2  ∂x i ∂x j ∂x k 
1  ∂gβγ ∂g γα ∂g αβ 
[αβ, γ] = 2  α + −
∂y γ 
and, … (5)
 ∂y ∂y β
Substituting the value of (1), (2), (3) in equation (4) and using (5) we get,
∂y α ∂y β ∂ y γ ∂ 2 yα ∂ yβ
[ij, k ] = [αβ, γ ] + a αβ .
∂x i ∂x j ∂ x k ∂ xi ∂ x j ∂ x k

10.4 PARALLELISM IN A SUBSPACE

THEOREM 10.6 Let T α and t i be the components of the same vector t relative to Vm and Vn
respectively. Let the vector t be defined along a curve C in V n. If p i and qα are derived vectors of t
along C relative to Vn and Vm respectively. Then
∂y α
qα = pi
∂xi
Proof: Since from equation (1), theorem, (10.4), Pg. 194 we have
∂y α
i
T α = t … (1)
∂x i
dT α dt i ∂y α 2 α
i ∂ y dx j
= + t … (2)
ds ds ∂x i ∂x i ∂x j ds
Now,
i dx j
p i = t, j … (3)
ds
β
α dy
qα = T,β
ds
 ∂T α  β
γ  α   dy

=  β + T  
 ∂x γβ  ds
∂T α dy β γ  α  dy
β

= + T  
∂xβ ds γβ ds
dT α γ  α  dy
β

q α
= ds + T   … (4)
 γβ ds
Parallelism of Vectors 197

dT α
Putting the value of and T γ from (2) and (1) in equation (4), we get
ds
dt i ∂y α i ∂y dy  α 
2 α γ β
i ∂ y dx j
qα = + t + t  
ds ∂x i ∂x i ∂x j ds dx i ds γβ

∂t i ∂y α dx j i ∂y dy  α 
2 α γ β
i ∂ y dx j
= + t + t  
∂x j ∂x i ds ∂x i ∂x j ds ∂x i ds  γβ

∂t i ∂y α dx j i ∂y ∂y dx  α 
2 α γ β
i ∂ y dx j j

= + t + t  
∂x j ∂x i ds ∂x i ∂x j ds ∂x i ∂ x. j ds γβ

∂t i ∂ y α d x j i d x j  ∂ 2 y α ∂y β ∂y γ  α 
α + t + +  
ds  ∂ x i ∂x j ∂x j ∂x i γβ
q = … (5)
∂x j ∂x i ds
But we know that

∂y α ∂y β ∂y γ ∂ 2 y α ∂y γ
[ij, k ] = [αβ, γ] + a αγ
∂x i ∂x j ∂x k ∂x i ∂x j ∂x k

∂y β ∂y γ ∂y δ ∂ 2 y α ∂y δ
[ ji, k ] = [βγ, δ] + a αδ
∂x j ∂x i ∂x k ∂x i ∂x j ∂x k

 α ∂y β ∂y γ ∂y δ ∂ 2 y α ∂y δ
= βγ αδ j
a + a αδ
  ∂x ∂x i ∂x k ∂x i ∂x j ∂x k

∂y δ   α  ∂y β ∂y γ 2 α 
a   j i + ∂ iy j 
= αδ ∂x k  βγ ∂x ∂x ∂x ∂x  … (6)
  
∂y δ
Multiplying (5) by a αδ , we get, using (6),
∂x k
∂y δ α ∂t i dx j ∂y α ∂y δ i dx
j
aαδ q = a αδ + t [ij, k ]
∂x k ∂x j ds ∂x i ∂x k ds

∂y δ dx j  ∂t i  p  ∂y α ∂y β
Or qδ k =  g + t i  g pk  ,
j ik since g ij = aαβ
∂x ds  ∂x i j   ∂x i ∂x j

 ∂t i 
 j gik + t a   g ik 
dx j i
= ds  ∂x 
 a j  

dx j i
= g ik t,
ds j
198 Tensors and Their Applications

i dx j
= ( gik t ), j
ds
dx j
= tk, j
ds
∂y α dx j
qα = t
∂x k
k, j
ds
α
∂y
or q α k = pk …(7)
∂x

Proved.
Properties of Vm
(i) If a curve C lies in a subspace V n of V m and a vector field in V n is parallel along the curve C
with regard to V m, then to show that it is also a parallel with regard to V n.
Proof: If a vector field t is a parallel along C with respect to V m. then its derived vector qα
vanishes i.e.,
qα = 0, α = 1, 2,... m.
Hence equation (7) becomes pk = 0, k = 1, 2, ..., n. This shows that the vector field t is also
parallel along C with respect to V n.
(ii) To show that if a curve C is a geodesic in a V n it is a geodesic in any subspace V n of V m.
α ∂y α
Proff : Science, q = pi
∂x i
Let t be unit tangent vector to the curve C them T α is unit tangent vector to C relative to V m
and ti is unit tangent vector to C relative to V n.
Now, pi = 0 ⇒ curve C is a geodesic in V n
α
q =0 ⇒ curve C is a geodesic in V m
Also, qα = 0, ∀α ⇒ pi = 0, ∀i
This shows that if a curve C in V n is a geodesic relative to V m then the same curve is also a geodesic
relative to V n.
(iii) A necessary and sufficient condition that a vector of constant magnitude be parallel with
respect to V n along C in that subspace, is that its derived vector relative to V m for the
direction of the curve be normal to V n.
Proof: Let t be a vector of constant magnitude. This vector t is parallel along C relative to Vn iff pi =
0, ∀i
∂y α
iff q α =0
∂x i
∂ yα
But qα = 0 implies that qα is normal to V n.
∂x i
∂ yα
For = y,αi lying in V m is tangential to a coordinate curve of parameter xi in V n.
∂ xi
Parallelism of Vectors 199

These statements prove that a necessary and sufficient condition that a vector of constant
magnitude be parallel along C relative to V n is that its derived vector i.e., qα along C relative
to V m be normal to V n.
(iv) A necessary and sufficient condition that a curve be a geodesic in V n is that its principal
subnormal relative to V m the enveloping space be normal to V n at all points of the curve.
Proof: In particular let the vector t be unit tangent vector to the curve C. In this case qα is
called principal normal the curve C. Also pi = 0, ∀i implies that the curve C is a geodesic
relative to V n.
Using result (iii), we get at once the result (iv).
(v) To prove that the tendency of a vector is the same whether it is calculated with respect to
V m.
Proof: Since, we have
∂y α
q α i = pi
∂x
∂y α dx i dxi
qα = p
∂x i ds
i
ds
dy α dxi
qα = pi
ds ds
β α
dy dy dx j dx i
Tα, β = ti, j
ds ds ds ds
i.e., tendency of Tα along C = tendency of ti along C.
i.e., tendency of t along C relative to V m = tendency of t along C relative to V n.

10.5 THE FUNDAMENTAL THEOREM OF RIEMANNIAN GEOMETRY


STATEMENT
With a given Riemannian metric (or fundamental tensor) of a Riemannian manifold there is associated
a symmetric affine connection with the property that parallel displacement (or transport) preserves
scalar product.
Proof: Let C be a curve in V n. Let pi and qi be two unit vectors defined along C. Suppose that the unit
vectors pi and qi suffer parallel displacement along the curve C in V n, then we have
dx j
p,i j =0 … (1)
ds
dx j
and q ,ij =0 … (2)
ds
Let gij be the given fundamental tensor of a Riemannian manifold. Hence, the scalar product of
vectors p i and q i is g ij p i q j.
Now,

dx k
( gij p i q j ),k =0
ds
200 Tensors and Their Applications

 dx k  j  k 
i  j dx 
 dx k  i j
g ij  p,ik q + g p + 
 ,k ds   ij ,k ds  P q = 0
q g
ds 
ij … (3)
    
Using equation (1) and (2), the equation (3) becomes,
 k 
 gij , k dx  p i q j = 0
 ds 

gij, k = 0
dx k
pi qi
(Since and are unit vectors and ≠ 0)
ds
∂g ij m  m
− g mj   − gim   = 0
∂x k
i k   j k
∂g ij
− [ik , j ] − [ jk , i ] = 0
∂x k
∂g ij
= [ik , j] + [ jk , i ] … (4)
∂x k
Now, using equation (4), we have
∂g jk ∂g ki ∂g ij
+ − = [ ji, k ] + [ki, j ] + [kj, i ] + [ij, k ] – ([ik , i] + [ jk , i ])
∂x i ∂x j ∂x k
since [ij, k ] = [ ji, k ].
So,
∂g jk ∂g ki ∂g ij
+ − = 2[ij, k ]
∂x i ∂x j ∂x k
1  ∂g jk ∂g ki ∂g ij 
[ij,k] =  + −  … (5)
2  ∂x i ∂x j ∂x k 
k
But we know that i j = g [ij, l ]
lk

 
from (5), we have
k 1 lk  ∂g jl ∂g li ∂gij 
  = g  i + j − l  Proved.
i j 2  ∂x ∂x ∂x 
EXAMPLE 3
If t i and T α are contravariant components in x's and y's respectively, of n vector field in V n
immersed in V m. Show that
T, αj = y,αij t i + y ,αi t,i j
Solution
Since we know that
i ∂y α
T α = t
∂x i
Parallelism of Vectors 201

i α
T α = t y, j .
Taking covariant differentiation of both sides, we get

T, αj = (t i y α,i ), j

= t,i j y,αi + t i ( y ,αi ), j

T, αj = t i y ,αij + t,i j y ,αi


EXAMPLE 4
dx i dx j
Show that g ij remains constant along a geodesic.
ds ds
Solution
dx i
Let t =
i
. Then
ds
dx i dx j
gij = g ij t i t j = t 2
ds ds
Since we know that geodesics are autoparallel curves. Then
dx j
t,i j =0
ds

or t,i j t j = 0 … (1)
Now,
dt 2 d d
= ( g ij t i t j ) = (ti t i )
ds ds ds
dx j
i
= ( ti t ) , j = (t i t i ), j t j
ds
dt 2
= (ti , j t i )t i + (t ,i j t i )ti = 0, from (1)
ds
Integrating it we get
t2 = constant.
dx i dx j
So, g ij remains constant along a geodesic.
ds ds

EXAMPLE 5
If ti are the contravariant components of the unit tangent vector to a congruence of geodesics.
Show that
t i (ti , j + t j ,i ) = 0
and also show that | ti , j + t j ,i |= 0.
202 Tensors and Their Applications

Solution
Let t i denote unit tangent vector to a congruence of geodesic so that
t,i j t j = 0 … (1)
Since geodesics are auto-parallel curves. Then to prove that
(i) t i (ti , j + t j ,i ) = 0
(ii) | ti , j + t j ,i |= 0
Since
t i ti = t 2 = 1
t i ti = 1
Taking covariant derivative of both sides, we get
(t i t i ), j = 0
or t,i j t i + t i ti , j = 0
Since t is a free index. Then we have
t,i j t i + ti , j t i = 0
2t i, j t i = 0

⇒ ti , j t i = 0 … (2)
from (1)
g ik t,i j t j = 0
or tk, j t j = 0

⇒ t k ,i t i = 0 ⇒ t j ,i t i = 0 ⇒ t i t j, i = 0

Thus t i t j, i = 0 … (3)
Adding (2) and (3), we get
t i (ti , j + t j ,i ) = 0
Also, since ti ≠ 0, ∀i.
Taking determinants of both sides we get
| t i (ti , j + t j , i ) | = 0
⇒ | ti , j + t j ,i | = 0
Solved.
EXAMPLE 6
When the coordinates of a V 2 are chosen so that the fundamental forms is
(dx&' ) + 2 g12dx1 dx 2 + (dx 2 )2 , prove that the tangent to either family of coordinate curves suffers a
parallel displacement along a curve of the other family.
Parallelism of Vectors 203

Solution
The metric is given by
i j
ds 2 = g ij dx dx i,j = 1,2.
Comparing it with
ds 2 = (dx ) + 2 g12 dx dx + (dx )
1 2 1 2 2 2
(1)
We have,
g11 = 1, g22 = 1.
In this case, we have coordinate curves of parameters x1 and x2 respectively. The coordinate x1
curve is defined by
xi = ci, ∀i, except i = 1. (2)
2
and the coordinate x curve is defined by
x i = d i, ∀i, except i = 2 (3)
i i
where c and d are constants.
Let p i and q i be the components of tangents vectors to the curves (2) and (3) respectively. Then
we have
pi = dxi = 0, ∀i, except i = 1
and qi = dxi = 0, ∀i, except i = 2
So,
pi = (dx i, 0) and qi = (0,dx i)
Let t be the unit tangent vector to the curve (2).
Hence
dx i pi
= = (1,0)
i
= t where p = dxi
ds p
So, we have
dx j
p,i j =0
ds
Hence the tangent to the family of coordinates curves (2) suffers a parallel displacement along a
curve of the family of curves (3).

EXERCISES

1. Explain Levi-Civita's concept of parallelism of vectors and prove that any vector which undergoes a
parallel displacement along a geodesic is inclined at a constant angle to the curve.
2. Show that the geodesics is a Riemannian space are given by
d 2 x m  m  dx i dx k
+  =0
ds 2 i k  ds ds
Hence prove that geodesics are auto-parallel curves.
204 Tensors and Their Applications

3. Establish the equivalence of the following definitions of a geodesic.


(i) It is an auto–parallel curve.
(ii) It is a line whose first curvature relative to Vn identically zero.
(iii) It is the path extremum length between two points on it.
4. If u and v are orthogonal vector fields in a V n , prove that the projection on υ of the desired vector of
u in its own direction is equal to minus the tendency of v in the direction of u.
5. If the derived vector of a vector u i is zero then to show that vector u i has a constant magnitude along
curve.
6. Prove that any vector which undergoes a parallel displacement along a geodesic is inclined at a
constant angle to the curve.
7. Prove that there are m – n linearly independent vector fields normal to a surface Vn immersed in a
Riemannian Vm and they may be chosen in a multiply infinite number of ways. But there is only one
vector field normal to the hyperface.
8. Show that the principal normal vector vanishes identically when the given curve in geodesic.
9. Show that if a curve is a geodesic of a space it is a geodesic of any subspace in which it lies.
10. Define parallelism in a subspace of Riemannian manifold. If a curve C lies in a subspace Vn of Vm and
a vector field in Vn is parallel along C with respect to V m . Then show that it is also parallel with
respect to V n .
CHAPTER – 11

RICCI'S COEFFICIENTS OF ROTATION AND


CONGRUENCE

11.1 RICCI'S COEFFICIENTS OF ROTATION


Let e ih| (h=1, 2, …n) be the unit tangents to the n congruences eh , of an orthogonal ennuple in a
e
Riemannian V n, . The desired vector of el|i in the direction of e k | has components el|i , j ekj| and the

projection of this vector on e ih| is a scalar invariant, denoted by γlhk, so that

γlhk = el|i, j eih|ekj| …(1)


The invariants γlhk are Ricci's Coefficients of Rotation.
Since i being a dummy index, has freedom of movement. Then equation (1) may be written as

γlhk = el |i , j eh|i ekj| …(2)


The indices l, h and k are not tensor indices. But these indices in γlhk are arranged in proper way,
the first index l indicates the congruence whose unit tangent is considered, the second h indicates the
direction of projection and the third k is used for differentiation.
THEOREM 11.1 To prove thast the Riccie's coefficients of rotation are skew-symmetric in the first
two indices i.e.,
γlhk = –γlhk

Proof: If e ih| (h = 1, 2, …, n) be n unit tangents to n congruences e h| of an orthgonal ennuple in a Vn


then
eh|i eli| = 0
convariant differentiation with respect to xj, we get

eh|i eli| ,j = 0

eh |i , j eli| + eli|, j eh|i = 0


206 Tensors and Their Applications

multiplying by ekj| and summing for j, we get

eh |i , j eli|eki | + eii|, j eh|i e kj| = 0


γhlk + γlhk = 0
or γhlk = –γlhk …(1)
Note: Put l = h in equation (1), we get
γllk = –γllk
or 2γllk = 0
or γllk = 0

THEOREM 11.2 To prove that

∑ γ lhk eh|i = el|i , j ekj|


h
Proof: since we know that

γlhk = el|i, j eih|ekj|


Multiplying by eh|m and summing for h.

∑ γ lhk eh|m = ∑ el|i , j eih|ekj|eh|m


h h
j
= el |i , j e k| ∑ (e e )
h
i
h| h| m

= el |i , j ekj|δim since ∑ ehi | eh|m = δ i


m
h
= (e l |i , j δ m ) e kj |
i

∑γh
lhk eh |m = el |m , j ekj |

Replacing m by i, we get

∑ γ lhkei|m = el |i , j e kj|
h

11.2 REASON FOR THE NAME "COEFFICIENTS OF ROTATION"


Let Cm be a definite curve of the congruence whose unit tangent is emi | and P 0 a fidxed point on it. Let
u1 be a unit vector which coincides with the vector eli| at P 0 and undergoes a parallel displayment along
the curve Cm.
Thus
u i = eli| at P 0 … (1)

and u,ij ehj| = 0 … (2)


Ricci's Coefficients of Rotation and Congruence 207

If θ is the angle between the vectors ui and e ih| , we have

cos θ = ui eh|i
Differentiating it with respect to arc length sm along Cm, we get

– sin θ i j
ds m = (u e h|i ), j em |

= (u i e h|i , j + u i, j eh |i ) emj |

– sin θ u i e h |i , j e mj | + u ,i j e mj |
ds m = …(3)

π
at the point P 0,θ = , we have
2

– u i eh|i , j emj |
ds m =
= u i eh|i , j emj |

= eli|eh|i, j emj | ; from (1)

= e h|i , j eli| emj |



ds m = γhlm


⇒ ds m = – γhlm … (4)


In Eucliden space of three dimensions, ds is the arc-rate of rotation of the vector e ih| about the
m
curve Cm. Hence the quantities γhlmare called coefficients of rotation of the ennuple. Since it was
discovered by Ricci and hence it is called Ricci's coefficient of rotation.

11.3 CURVATURE OF CONGRUENCE


The first curvature vector pl| of curve of the congruence is the derived vector of e ih| in its own
direction. Where e ih| (h = 1, 2, …, n) be unit tangents to n congruence eh| of an orthogonal ennuple in
a Vn .
If phi | be the contravariant component of first curvature vector pl|. Then by definition, we have

phi | = ehi , j ehj|


from theorem (11.2), we have

phi | = ∑ γ hlh eil| … (1)


l
208 Tensors and Their Applications

The magnitude of p ih| is called curvature of the curve of congruence e h| and denoted by K h | .
Now,
K h2| = g ij phi | p hj|

   
= g ij  ∑γ i 
hlh e l | 

 ∑γ i
hmh em |


 l   m 

= ∑∑ g ij eli| emi |γ hlh γ hmh


l m

= ∑∑ δ
l m
m γ hlm
l
γ hmh , Since g ei e i = l
ij l| m| δm

= ∑ γ hmhγ hmh
m

K h2| = ∑ (γ
m
hmh )
2

This is the required formula for Kh|.

11.4 GEODESIC CONGRUENCE


If all the curves of a congruence are geodesics then the congruence is called a geodesic congruence.
THEOREM 11.3 A necessary and sufficient condition that congruence C of an orthogonal ennuple
be a geodesic congruence is that the tendencies of all the other congruences of the ennuple in the
direction of C vanish indentically.
Or
To obtain necessary and sufficient conditions that a congruence be a geodesic congruence.
Proof: From equation (1), Pg. 207, we have
phi | = ∑ γ hlh eil|
l
Since eli| ≠ 0 , ∀h and hence phi | = 0 iff γhlh = 0, ∀h . But pli| = 0 iff the congruence C is geodesic
congruence.
Hence C is a geodesic congruence iff γhlh = 0, ∀h .
But
γhlh = – γlhh
So, C is a geodesic congruence iff – γlhh = 0, –∀h .
or C is a geodesic congruence iff γlhh = 0.
Hence γlhh = 0 are the necessary and sufficient conditions that congruence with unit tangents ehi |
be geodesic congruence. Again γlhh is the tendency of the vector eli| in the direction vector e ih| .
Ricci's Coefficients of Rotation and Congruence 209

Thus a congruence C of an orthogonal ennuple is a geodesic congruence iff the tendency of all
other congruences in the direction of C vanish identically.

11.5 NORMAL CONGRUENCE


A normal congruence is one which intersects orthogonally a family of hypersurfaces.

THEOREM 11.4 Necessary and sufficient conditions that the congruence eh | of an orthogonal ennuple
be normal is that
γnqp = γnpq
Proof: Consider a congruence C of curves in a V n. Let φ( x1 , x 2 , …, x n ) = constant be a family of
hypersurfaces. To determines a normal congruence whose tangent vector is grad φ or ∇ φ.
Let ti be the convariant components of unit tangent vector to C. The congruence C is a normal
congruence to a family of hypersurface φ( x1 , x 2 , …, x n ) = constant if
φ,1 φ, 2 φ,
= = … n = y (say) … (1)
t1 t2 tn
In order that (n – 1) differential equations given by equation (1) admit a solution which is not
constant, these must constitute a complete system.
From (1)
φ,i
= y or φ, i = yti
ti
∂φ
⇒ yti =
∂x i
Differentiating it with respect to x j , we get

∂y ∂t i ∂ 2φ
ti + y = … (2)
∂x j ∂x j
∂x i ∂x j
Interchanging indices i and j, we get

∂y ∂t j ∂ 2φ
t + y = … (3)
∂x i
j
∂x i ∂x j ∂x i

Subtracting (2) and (3) we get


 ∂y ∂t   ∂y ∂t 
 j ti + y i j  –  i t j + y i j  = 0
 dx ∂x   ∂x dx 

 ∂t ∂t j   ∂y ∂y 
y ij – i  +  t i j – t j i  = 0
 ∂x ∂x   ∂x ∂x 
210 Tensors and Their Applications

Multiplying by tk,

 ∂ti ∂t j  ∂y ∂y
ytk  j – i  + ti tk – ik ti i = 0 …(4)
 dx ∂x  ∂x j
dx
By cyclic permutation of i, j, k in (4), we get

 ∂t j ∂t  ∂y ∂y
yti  k – kj  + t j t i k – i k ti =0 … (5)
 ∂x ∂x  ∂x ∂x j
and

 ∂t ∂t  ∂y ∂y
yti  ki – ik  + t k t j i – ti t j k = 0 … (6)
 ∂x ∂x  ∂x ∂x
On adding (4) , (5) and (6) we get

  ∂t ∂t j   ∂t j ∂t  
y t k  ij – i  + ti  k – kj  + t j  ∂tk – ∂ti  = 0

  ∂x ∂x   ∂x ∂x   ∂x i ∂x k  

or t k (ti , j – t j ,i ) + ti (t j , k – t k , j ) + t j (t k , i – ti ,k ) = 0 … (7)

as y ≠ 0 , where i, j, k = 1, 2, …, n.
These are the necessary and sufficient conditions that the given congruence be a normal congruence.
Now suppose that the congruence is one of an orthogonal ennuple in V n. Let en|i be the unit
tangents of given congruence C so that ti = en|i
Then equation (7) becomes
en|k (en| i,j – en| j,i) + en| i (en| j,k – en| k,j) + en| j (en|k,i – en|i,k) = 0 … (8)

Now, multiplying equation (8) by e ip| eqR| , we get

e n|k (en |i , j – en | j ,i ) e ip| eqk| + en |i (en | j , k − en| k , j ) e ip| eqk + en | j (en |k ,i – en |i , k ) e ip |eqk|
where p and q are two new indices chosen from 1, 2, …, n – 1. i.e., p,q and n are unequal.

But e n |k eqk| = δ q = 0, q ≠ n
h

en |i e kp| = δ hp = 0, p ≠ n
so, we have

e n| j (en |k ,i – en | j , k ) eip | eqk| = 0

e n| j (en |k , i e iq |e kp | – en | j , k eip |e kq| ) = 0


en|j (γnqp – γnpq) = 0

Since en |k , i eip | eqk| = γnpq


Ricci's Coefficients of Rotation and Congruence 211

⇒ γnqp – γnpq = 0, en| j ≠ 0


⇒ γnpq = γnqp, … (9)
Conversely if equation (9) in true then we get equation (8). Which implies that equation (7) are
satisfied bty en|i Hence en| is a normal congruence.
Thus necessary and sufficient conditions that the congruence en| of an orthogonal ennuple be a
normal congruence are that
γnqp = γnpq (p, q = 1, 2, …, n – 1 such that p ≠ q)
THEOREM 11.5 Necessary and sufficient conditions that all the congruences of an orthogonal ennuple
be normal.
Proof: If all the congruences of a orthogonal ennuple are normal. Then
γnqp = γnpq (p, q = 1, 2, …, n – 1 such that p ≠ q)
If the indices h, k, l and unequal then
γhkl = γhlk … (1)
But due to skew-symmetric property i.e., γhlk = – γlhk.
So,
γhkl = γhlk = – γlhk = – γlkh, from (1)
= γklh (skew-symmetric property).
= γkhl, from (1)
γhkl = – γhkl , (skew-symmetric property).
γhkl + γhkl = 0
⇒ 2 γhkl = 0
⇒ γhkl = 0
where (l, h, k = 1, 2, …, n such that h, k, kl are unequal).

11.6 CURL OF CONGRUENCE


The curl of the unit tangent to a congruence of curves is called the curl of congruence.
If en| is a given congruence of curves then
Curl en|i = en|i, j – en| j,i
If curl en| i = 0 then congruence is irrotational.

THEOREM 11.6 If a congruence of curves satisfy two of the following conditions it will also satisfy
the third
(a) that it be a satisfy the third
(b) that it be a geodesic congruence
(c) that it be irrotational.
Proof: Consider an orthogonal ennuple and e ih| (h = 1, 2, …, n) be n unit tangent to n congruences of
this orthogonal ennnuple.
From theorem (11.2), we have

∑ γ lhkeh|i = el|i , j ekj|


h
212 Tensors and Their Applications

Putting l = n and j = m, we have


n

∑γ
h =1
nhk eh |i
m
= e n |i , m e k |

Now, multiplying by ek| j and summing with respect to k from 1 to n, we have


n
∑ γ nhk eh|i ek | j = en|i ,me mk| ek| j
h,k =1
= en |i ,mδ mj Since ekm| ek| j = δmj
n

∑γ
h, k =1
nhk eh|i e k | j = e n|i, j … (2)

By definition of curl of congruence, we have


curl en|i = en| i, j – en| j,i
n n

= ∑ γ nhke h|i ek| j – ∑ γ nhkeh| j ek |i ; from (2)


h,k =1 n ,k =1

n n

curl en| i = ∑
h ,k =1
γ nhk eh |i ek | j – ∑γ
h , k =1
nkh ek | j e h|i

curl en|i = ∑
h ,k =1
(γnhk – γnkh) eh| iek| j … (3)

This double sum may be separated into two sums as follows.


(i) Let h and k take the values 1, 2, …, n – 1.
(ii) Either h = n or k = n or h = k = n.
Now, the equation (3) becomes
n −1 n –1

curl en| i = ∑
h ,k =1
(γ nhk – γ nkh ) eh |i ek | j + ∑
h =1
(γ nhn – γ nnh ) e h|i en | j

+ ∑ k =1
( γ nnk – γ nkn ) en |i e n| j + (γ nnn – γ nnn ) en |i en | j

Since we know that γnnk = γnnh = γnnn= 0.


So,
n –1 n –1 n–1

curl en|i = ∑
h , k =1
(γ nhk – γ nkh ) e h |i e k | j + ∑
h =1
γ nhn e h|i en | j – ∑γ
k =1
nkn en |i ek| j

n–1 n–1

curl en|i = ∑h ,k
(γ nhk – γ nkh ) eh|i ek | j + ∑γ
h =1
nhn ( eh |i en | j – en |i ek | j ) … (4)
Ricci's Coefficients of Rotation and Congruence 213

The first term on R.H.S. of equation (4) vanishes


if γnhk – γnkh = 0
i.e.,
if γnhk = γnkh.
i.e., if the congruence e n| is normal.
Again the second term of R.H.S of equation (4) vanishes
if γ nhn = 0 i.e., if γ hnn = 0
i.e., if the congruence e n| is a geodesic congruence. Further, if first and second term on right
hand side of equation (4) both vanishes then
curl en|i = 0

Hence we have proved that if the congruence e n| satisfies any two of the following conditions
then it will also satisfy the third.
(a) e n| is a normal congruence
(b) e n| is irrotational
(c) e n| is a geodesic congruence

11.7 CANONICAL CONGRUENCE


It has been shown that given a congruence of curves, it is possible to choose, in a multiply infinite
number of ways, n – 1 other congruences forming with the given congruence an orthogonal ennuple.
Consider the system of n – 1 congruence discovered by Ricci, and known as the system canonical with
respect to the given congruence.
THEOREM 11.7 Necessary and sufficient conditions that the n – 1 congruences eh| of an orthogonal
ennuple be canonical with respect to en| are
γnhk + γnkh = 0; (h, k = 1, 2, …, n – 1, h ≠ k ).
Proof: Let the given congruence en| be regarded as nth of the required ennuple. Let en|i be unit tangent
to given congruence.
X ij = ½ (en| i, j + en| j,i) … (1)
Let us find a quantity ρ and n quantities ei satisfying the n + 1 equations

e n|i ei = 0
 i, j = 1, 2, …, n … (2)
( X ij – ωgij e i + ρE n| j = 0

where ω is a scalar invariant.


Writing equation (2) in expansion form, we have
e n|1e1 + en |2 e 2 + …… + en| n e n = 0 … (3)
and

( X 1 j – ωg1 j ) ei + ( X 2 j – ωg 2 j ) e 2 + … ( X nj – ωg nj ) e n + ρ e n| j = 0
214 Tensors and Their Applications

for j = 1, 2, …, n.
1 2 n
( X 11 – ωg11 ) e + ( X 21 – ωg 21 ) e + … + ( X n1 – ωg n 1 ) e + ρ en |1 = 0
( X 12 – ωg12 ) e1 + ( X 22 – ωg 22 ) e 2 + … + ( X n 2 – ωg n 2 )e n + ρ en |2 = 0
M
( X 1n – ωg1 n ) e1 + ( X 2 n – ωg 2 n ) e 2 + … + ( X nn – ωg nn ) e n + ρ en |n = 0
from (3)
e n|1 e1 + en |2 e 2 + … en |n e n + ρ.0 = 0
eliminating ρ and the quantities e1, e2, …en, we have the equation

(X 11 – ωg11 ) (X 21 – ωg 21 ) … (X n1 – ωg n1 ) en |1
( X12 – ωg12 ) ( X 22 – ωg 22 ) … ( X n2 – ωg n2 ) e n|2
M
( X1n – ωg1n ) (X 2n – ωg 2n ) … (X nn – ωg nn ) en |n
en |1 en |2 … en| n 0

which is of degree n – 1 in ω. Hence there will be n – 1 roots of ω and these roots be ω1, ω2, …,
ωn – 1. All roots are real. Let ωh be one of these roots and let the corresponding values of ρ and ei be
denoted by ρh and e ih| respectively. Then ρh and e ih| will satisfy the equation (2), we have
en |i e ih| = 0 … (4)
and ( X ij – ωg ij ) eih | + ρh en | j = 0 … (5)
Similarly ωk be another root of these roots, we have
en |i e ik| = 0 … (6)
( X ij – ωg ij ) eik | + ρh en | j = 0 … (7)

Multiplying (5) by e kj| and (7) ehj| and using (4) and (6), we get

( X ij – ωh g ij ) ehi |ekj| = 0 … (8)

( X ij – ωk g ij ) eki |e hj| = 0 … (9)

Since X ij and g ij are symmetric tensor in i and j. Now, interchanging i and j in equation (9), we
get
( X ij – ωk g ij ) ehi |e kj| = 0 … (10)
Subtracting (10) and (8), we get
(ωh – ωk ) g ij e ih| ekj | = 0 … (11)
Since ωk – ωk ≠ 0 as ωh ≠ ωh.
⇒ g ij eih| ekj| = 0 … (12)
Ricci's Coefficients of Rotation and Congruence 215

This shows that e ih| and e kj| unit vectors are orthogonal to each other and hence the congruence
e h| and e k| (h ≠ k) are orthogonal to each other. Hence the n – 1 congruence eh| (h = 1, 2, …, n) thus
determined form an orthogonal ennuple with en|.
Using equation (12), equation (10) becomes
X ij e ih| ekj| = 0

Since from (1), X ij =


1
e (
2 n |i , j
)
+ e n| j , i . Then we have,

1
( e
2 n |i , j
)
+ en | j ,i ehi |ekj| = 0

en |i , j ehi | ekj| + en| j ,i eik ehj = 0


γnhk + γnkh = 0; (h, k = 1, 2, …, n – 1 such that h ≠ k ) … (13)
Conversely if equation (13) is true then (n – 1) congruences eh| of the orthogonal ennuple and
canonical with respect to e n| . Hence necessary and sufficient condition that the n – 1 congruences e h|
of an orthogonal ennuple be canonical with respect to en| are
γnhk + γnkh = 0; (h, k = 1, 2, …, n such that h ≠ k )

THEOREM 11.8 Necessary and sufficient conditions that n –1 mutually orthogonal congruences e h|
orthogonal to a normal congruence e n| , be canconical with respect to the later are γnhk = 0 where k, h
= 1, 2,…, n – 1 such that h ≠ k.
Proof: By theorem (11.7), Necessary and sufficient conditions that (n – 1) congruences e h| of an
orthogonal ennuple be canonical with respect to en| are
γnhk + γnkh = 0 (h, k = 1, 2, …, n – 1 such that h ≠ k).
If the congruence en| is normal. Then
γnhk = γnkh
The given condition γ nhk + γ nkh = 0 becomes
γnhk + γnhk = 0
2γnhk = 0
γnhk = 0 Proved.

EXAMPLE 1
1
If en| are the congruences canonical with respect to en| prove that (i) ωh = γnhk (ii) ρh = γhnn,
2
(iii) If en| is a geodesic congruence, the congruences canonical with respect to it are given by
( X ij − ωg ij )ei = 0 = 0
Solution
Suppose (n – 1) congruences eh| of an orthogonal ennable in a V n are canonical with respect to
the cougruence eh| then
en |iehi | = 0 … (1)
216 Tensors and Their Applications

and ( X ij – ωh g ij ) ehi | + ρ h en | j = 0 … (2)


where
1
X ij = (e + en| j,i) … (3)
2 n| i,j
Since e in| unit tangents then
g ij ehi |ehj| = 0 … (4)

(i) Multiplying equation (2) by e ih| , we get

( X ij – ωh g ij ) ehi | ehj| + ρh e n| j e hj| = 0

X ij eih| ehj| – ω h g ij eih |ehj| = 0 since en| j ehj| = 0 (from (1))

X ijeih| ehj| – ωh = 0, since g ijehi |ehj| = 1


1
(e n|i , j + en| j , i ) e ih|e hj| – ωh = 0; from (3)
2
1
(e e i e j + en | j ,i eih| ehj | ) – ωh = 0
2 n|i , j h | h|
1 (γ + γ ) – ω = 0
nhh nhh h
2
ωh = γnhh
(ii) Multiplying (2) by enj| , we get

( X ij – ωh g ij ) ehi | enj| + ρh e n| j e nj| = 0

X ij eih| enj| – ω h g ij ehi |enj| + ρ h × 1 = 0 from (1)


1
(e n|i , j + en| j , i ) e ih|e hj| + ρh = 0, since e ih| e hj| = 0
2
1
(e ei e j + en | j ,i ehi | ehj | ) + ρh = 0
2 n|i , j h | n|
1
(γ + γnnh) + ρh = 0
2 nhn
1
ρh = – γ , since γnnh = 0
2 nhn
1
or ρh = γ
2 hnn
(iii) If en| is a geodesic congruence, then γhnn = 0 from the result (ii), we have
1
ρh = × 0
2
ρh = 0
Ricci's Coefficients of Rotation and Congruence 217

from equation (2), we get


( X ij – ω h gij ) ehi | + 0en| j = 0 or ( X ij – ωh g ij ) e ih| = 0
this gives ( X ij – ωg ij ) e i = 0 .

EXAMPLE 2
Prove that when a manifold admits an orthogonal systyem of n normal congruences then any of
these in canonical with respect to each other congruence of the system.
solution
Let e ih| (h = 1, 2, …, n) be unit tangents to n normal congruences of an orthogonal ennuple in a
V n. So that
γlhk = 0, where l, h, k = 1, 2,...n such that l, h, k being unequal.
It is required to show that a congruence eh| is canonical with respect to the congruence ek| (h, k = 1, 2,
…, h ≠ k). We know that the n – 1 congruence en| of an orthogonal ennuple be canonical to en| iff
γnhk + γnkh = 0
This condition is satisfied by virtue of equation (1). Hence (n – 1) congruences eh| of an orthogonal
ennuple are canonical to en|.
Similarly we can show that any n – 1 congruences are canonical to the remaining congruence.
It follows from the above results that any one congruence is canonical with respect to each other
congruence of the system.

EXERCISE

1. If φ is a scalar invariant
n
∑ φ,ij eih|ehi | = ∆ φ2
h =1

2. The coefficient of ρ n – 1 in the expansion of the determinant |φ,ij – ρg ij | is equal to ∇ 2 φ.


3. If eh| are the unit tangents to n mutually orthogonal normal congruences and e11 + be21 is also a
normal congruence then ae11 – be21 is a normal congruence.
4. Show that
∂Q
∂ ∂φ

∂ ∂Q
∂sh ∂sk ∂sk ∂sh =
l

( γ lhk – γ lkh )
∂sl
where S h denotes the arc length of a curve through a point P of an ennuple and Q is a scalar invariant.
5. If the congruence eh| (h = 1, 2, … n – 1) of an orthogonal ennuple are normal, prove that they are
canonical with respect to other congruence eh| .
CHAPTER – 12

HYPERSURFACES

12.1 INTRODUCTION
We have already studied (Art 10.3 chapter 10) that if m > n then we call V n to be a subspace of V m and
consequently V m is called enveloping space of V n. We also know that there are m – n linearly independent
normals Nα to V n. (Art 10.3 Theorem 10.5, chapter 10). If we take m = n + 1 then V n is said to be
hypersurface of the enveloping space V n + 1
Let V n be Riemannian space of n dimensions referred to cordinates xi and having the metric ds2 =
gij dx dxj. Let V n be Riemannian space of m dimensions referred to coordinates yα and having the metric
i

ds2 = aαβ dyα dyβ. Where m > n. Let Greek letters α, β, γ … take the values 1, 2, …, m and latin indices
i, j, k, … take the values 1, 2, …n. Then we have, the relation between aαβ and gij
α β
gij = aαβ ∂y ∂y … (1)
∂xi ∂x j
Since the function yα are invariants for transformations of the coordinates xi in V n, their first
covariant derivatives with respect to the metric of V n are the same as their ordinary derivatives with
respect to the variables xi.
∂yα
i.e., yα, i =
∂xi
Then equation (1) can be written as

gij = aαβ y,αi yβ, j … (2)

The vector of V n + i whose contravariant components are y,αi is tangential to the curve of
parameter xi in V n. Consequently if N α are the contravariant components of the unit vector normal to
V n. Then we have
a αβ N β y,αi = 0, (i = 1, 2, …, n) … (3)
and a αβ N αN β = 1 … (4)
Hypersurface 219

12.2GENERALISED COVARIANT DIFFERENTIATION


Let C be any curve in V n and s its arc length. The along this curve the x's and the y's may be expressed
as function of s only. Let uα and ν β be the components in the y's of two unit vector fields which are
parallel along C with respect to V m. Similarly w i the components in x's of a unit vector field which is
parallel along C with respect to V n. Now, uα is parallel along C relative to V m then we have,
dyβ
uα,β =0
ds
 ∂u α  γ  dy β
 β – uγ   =0
 ∂y α β ds

∂u α dy β  γ  dy β
γ 
– u =0
∂y β ds α β ds

duα  γ  dy β
– uγ   =0
ds α β ds

duα  α  dy β

= u γ β γ ds … (1)
ds  
similarly v β is parallel along C relative to V m then
α
dvβ γ  γ  dy
= – v α β ds … (2)
ds  
and wi is parallel along C relative to V n then

j  i  dx
k
dwi
= – w  j k  ds … (3)
ds  
The Christoffel symbol with Greek indices being formed with respect to the aαβ and the y's and
christoffel symbol with Latin indices with respect to the gij and the x's.
Let Aβαi be a tensor field, defined along C, which is mixed tenser of the second order in the y's and
α
a covariant vector in the x's. Then the product uα v β wi Aβi is scalar invariant and it is a function of s
along c. Its derivative with respect to s is also a scalar invariant.
α
Differentiating uα v β wi Aβi with respect to s, we have

d dAα dv β dw i
(uα v β wi Aβαi ) = u v β w i βi + du α v β w i A α + u α
wi α
Aβi
+ uα v β Aβαi
dt α βi ds ds
ds ds

dAβαi duδ β i δ dv δ dw j
u αvβ Aβαj
β
= uα v w i
+ v w Aβi + uα w i Aδαi +
ds ds ds ds
220 Tensors and Their Applications

dAβαi γ
β i  α  dy γ α β i  δ  dy
= u αv w + Aβδi .u αv βwi   – Aδi .u αv w βγ  ds
ds δγ  ds  

i  j  dx
k
α β
– Aβj uα v w ik  ds , by equation (1), (2) & (3).
 

 α
β i dAβi δ  α  dy
γ
α  δ  dy
γ
α  j  dx
k 
u
= α v w  + Aβi 
.  – Aδi   – Aβj  
. 
 ds δ γ  ds β γ  ds i k  ds 
= Scalar, along C
Since the outer product uαv βwi is a tensor and hence from quotient law that the expression within
α α
the bracket is tenser of the type Aβi and this tensor is called intrinsic derivative of Aβi with respect to
s.
The expression within the bracket can also be expressed as

dxk  ∂Aβαi α δ   j 


 + Aβδi   y,γk – Aδαi   y,γk – Aβαj   
ds  ∂xk
 δγ  βγ  ik 

dxk
Since is arbitrary..
ds
So, by Quotient law the expression within the bracket is a tensor and is called tensor derivative of
α
Aβi with respect to xk. It is denoted by Aβαi;k . Then we have

∂Aβαi α δ   j
Aβαi;k = + Aβδi   y,γk – Aδαi   y,γk – Aβαj  .
dx k δγ
  βγ
  ik 

Aβαi;k is also defined as generalised covariant derivative Aβi


α
with respect to xk .
Note:– Semi-colon(;) is used to denote tensor differentiation.

12.3 LAWS OF TENSOR DIFFERENTIATION

THEOREM 12.1 Tensor differentiation of sums and products obeys the ordinary rules of differentiation.
Proof: Suppose Aβα , Bβα and B γ are tensors in V m.
(i) To prove that

( Aβα + Bβα ); k = Aβα;k + Bβα; k

Let the sum Aβα + Bβα be denoted by the tensor Cβα .


Now,

∂Cβα α α 
Cβα;k = + Cβα   y, ck – Caα   y, ck
∂x βc
k
a c
Hypersurface 221

∂( Aβα + Bβα ) α a 


∴ ( Aβα + Bβα ); k = + ( Aβa + Aβα )   y,ck – ( Aaα + Baα )  y ,ck
β c 
k
dx a c 

 ∂Aβα 
 aα c αa  c 
=  ∂x k + Aβ   y, k – Aa   y , k +
 ac β c 

 ∂Bβα α   a 
 + Bβa   y,ck – Baα   y,ck 
 ∂xk
  ac βc 

( Aβα + Bβα ); k = Aβa ;α + Bβa ;k Hence the result (i)


(ii) Prove that

( Aβα + Bγ ); k = Aαβ; k Bγ + Aαβ Bγ ; k

Let Aβα Bγ = Dβγ


α α
Then Dβγ is a tensor
we have
α
∂Dβγ α  a a 
D αβγ ; k = + D;ak   yc – Dαaγ   y,ck – Dαβa   y,ck
∂x k
ac   βc  γc 

∂( Aβα B γ )  α
Bγ   y ,ck – Aaα Bγ   y,ck – Aβα B a   y ,ck
a a
∴ ( A αβ Bγ ) ; k = + Aβγ
a

βc  βc 
k
dx ac 
 ∂Aβα 
α α c αa  c   ∂B 
Bγ + Aβα  γ – Ba   y,ck 
 a
=  dxk + Aβ   y, k – Aa   y , k
 ac β c   dx
k
γc  

( Aβα B γ ); k = Aβα; k Bγ + Aaα Bγ ; k


Hence the result (ii).
Note:– α, β, γ, a, c take values from 1 to m while k take values from 1 to n.

THEOREM 12.2 To show that aαβ; j = 0

or
To prove that the metric tensor of the enveloping space is generalised covariant constant with respect
to the Christoffel symbol of the subspace.
Proof: We have (see pg. 220)
∂aαβ γ γ 
aαβ;i = – aγβ   y,δi – aαγ   y,δi
dxi αβ
  βδ 
∂aαβ
= – ( [αδ, β] + [βδ, α ] ) y,δi
dx i
222 Tensors and Their Applications

∂aαβ ∂aαβ δ
or aαβ;i = – y,i
dx i
∂ yδ

∂aαβ ∂aαβ ∂yδ ∂ aαβ ∂ aαβ


or aαβ;i = i

δ
⋅ i = – = 0 Proved.
dx ∂y dx ∂x i ∂x i

12.4 GAUSS'S FORMULA


At a point of a hypersurface V n of a Riemannian space V n + 1, the formula of Gauss are given by

y;αij = Ωij N α
Proof: Since yα is an invariant for transformation of the x's and its tensor derivative is the same as its
covariant derivative with respect to the x’s, so that

dyα
yα; i = yα, i = … (1)
dxi
Again tenser derivative of equation (1) with respect to x’s is

y;αij = ( y;i ); j = ( y,αi )


α

∂ l  α 
= ( y,αi ) – y,αl   + yβ,i y,γj  
βγ 
j
dx ij 

∂  ∂y α  α  l  β γ  α 
= – y,l   + y,i y, j  
∂x j  ∂xi  ij βγ 

∂ 2 yα l  α 
y;αij = – y,αl   + yβ,i y,γj   … (2)
βγ 
i j
∂x ∂x ij
Interchanging j and i in (2), we have

∂ 2yα l  α 
y;αji = – y ,αl   + y β, j y ,γi  
∂x ∂x βγ 
j i
 ji 

∂ 2yα l  α 
y;αji = – y ,αl   + y β,l y ,γj   … (3)
∂x ∂x βγ
i j
ij 

α  α 
[On interchanging β and γ in third term of R.H.S. of (3) and using βγ  = γβ  ].
   
On comparing equation (2) and (3), we have

y;αij = y;αji

So, y;αij is symmetrical with respect to indices i & j.


Hypersurface 223

Let g ij dx idx j and aαβdyα dyβ be fundamental forms corresponding to V n and V n + 1 respectively..
Then

∂yα ∂yβ
gij = aαβ
∂xi ∂x j

or gij = aαβ y;αi y;βj


Taking tensor derivative of both sides with respect to xk

gij;k = aαβ;k y;αi y;βj + aαβ y;αi y;βjk + aαβ y;αik y;βj

But gij;k = 0 and aαβ;k = 0.


we have,

0 = a αβ y ;αi y;βjk + aαβ y;αik y;βj

or aαβ y;αik y,βj + aαβ y,αi y;βjk = 0, using (1), Art. 12.4 … (4)
By cyclic permutation on i, j, k in (4), we have

aαβ yα; ji y,βk + aαβ y α, j y;βki = 0 … (5)

and a αβ y;αkj y,βi + aαβ y α,k y β;ij = 0 … (6)


subtracting equation (4) from the sum of (5) and (6), we get.

2 aαβ y;αij y ,βk = 0

or a αβ y ;αij y β,k = 0 … (7)

This shows that y;αij is normal (orthogonal) to y,βk . Since y,βk is tangential to V n and hence y;αij is

normal to V n. Then we can write

y;αij = N αΩij (8)


where Nα is unit vector normal to V n and Ω ij is a symmetric covariant tensor of rank two. Since
y;αij is a function of x's the tenser Ω ij is also a function of x's.
The equation (8) are called Gauss’s formula.
From equation (8)

y;αij = Ωij N α

y;αij aαβ N β = Ω ij aαβ N β N α

= Ω ij N 2
224 Tensors and Their Applications

y;αij aαβ N β = Ωij , N=1

or Ω ij = y;αija αβN β.
The quadratic differential form

Ωij dxi dx j
is called the second fundamental form for the hypersurface V n of V n + 1. The components of tenser Ω ij
are said to be coefficient of second fundamental form.
Note: The quadratic differential form g ij dx idx j is called first fundamental form.

12.5 CURVATURE OF A CURVE IN A HYPERSURFACE AND NORMAL CURVATURE


If U α and u i be the contravariant components of the vector u relative to V n and V n + 1 respectively then
we have (from chapter 10, Theorem 10.4)

∂y α i
U α = u = y,αi ui … (1)
∂ xi
Let the derived vector to vector u along C with respect to metric of V n and Vn + 1 are denoted by
p and q respectively. Then

pi dxj
= u,i j
ds
α d yβ
and qα = U,β
ds
dx j
= U ;αj , (from equation 1, Art 12.4) … (2)
ds
Taking the tensor derivative of each side of equation (1) with respect to x's, we have
U ;αj = y; ij u + y, i u , j
α i α i

By Gauss's formula, we have y;αij = Ωij N α

U ;αj = Ω ij N u + y ,i u , j
α i α i
Then

Putting the value of u;αj in equation (2), we have

α i α i dx j
qα = (Ωij N u + y, i u, j )
ds
 j 
i dx  α
 Ω α i
or qα =  ij u  N + y,i p … (3)
 ds 

i dx j
where pi = u, j
ds
Hypersurface 225

Now suppose that vector u is a the unit tangent t to the curve C. Then the derived vectors q and
p are the curvature vectors of a C relatively to V n + 1 and V n respectively. Then equation (3) becomes.

 dx i dx j  α
 N + y ,αi p i
qα =  Ωij ds ds  … (4)

dx i dx j
Taking Kn = Ω ij , we get
ds ds
α
qα = Kn Nα + y,i pi ...(5)
Kn is called Normal curvature of V n at any point P of the curve C and Kn Nα is called normal
curvature vector of V n + 1 in the direction of C.
Meaunier's Theorem
If Ka and Kn are the first curvature of C relative to V n + 1 and normal curvature of V n respectively
r r
and w is the angle between N and C (C being the unit vector of V n + 1 then the relation between Ka, Kn
and w is given by
Kn = Ka cos ω
Proof: We know that
qα = Kn N α + p i y,αi ...(1)

dxi dx j
where Kn = Ωij
ds ds
Let Ka and Kg be the first curvatures of C with respect V n + 1 and V n respectively then

Ka = a αβ q α q β , Kg = gij pi p j
r r
Let ω be the angle between N and C .
r r
Then N ⋅ C = N ⋅ C cos ω

= cos ω as |N| = |C| = 1


r r
N .C = cos ω ...(2)
r r
If b is a unit vector of V n + 1 in the direction principal normal of C with respect to V n then
equation (1) becomes
r r r
Ka C = K g b + K n N … (3)
r
Taking scalar product of equation (3) with N , we have
r r r r r r
K a N .C = K g N. b + K n N. N

Ka cos ω = Kg ⋅ 0 + K n ⋅1; from (2)


Kn = Ka cos ω … (4)
Proved.
226 Tensors and Their Applications

EXAMPLE 1
Show that the normal curvature is the difference of squares of geodesic curvatures.
Solution
We know that (from Meurier’s Theorem, equation 3)
r r r
KaC = Kg b + Kn N
Taking modulus of both sides, we get
2 2
Ka2 = Kg + Kn

2 2
or Kn2 = Ka – K g .

Theorem 12.3 To show that the first curvature in Vn + 1 of a geodesic of the hypersurface Vn is the
normal curvature of the hypersurface in the direction of the geodesic.
Proof: From, example 1, we have
2 2
Ka2 = Kg + Kn

If C is a geodesic of V n then p i = 0 ⇒ Kg = 0
Then we have
Ka2 = Kn2
⇒ Ka = K n. Proved.

Dupin's Theorem
The sum of normal curvatures of a hypersurface V n for n mutually orthogonal directions is an
invariant and equal to Ω ij g ij .
Proof: Let ehi | (h = 1, 2, …, n) be unit tangents to n congruences of an orthogonal ennuple in a V n. Let
Knh be normal curvature of the hypersurface V n in the direction of the congruence eh | . Then
Knh = Ωij eih | ehj|
The sum of normal curvatures for n mutually orthogonal directions of an orthogonal ennuple is a
V n is
n
∑ Knh = ∑ Ω
n

ij eih| ehj|
h =1 h =1

= Ωij ∑e
h =1
i
h| ehj|

= Ω ij g ij
= Scalar invariant Proved.
Hypersurface 227

12.6 DEFINITIONS
(a) First curvature (or mean curvature) of the hypersurface Vn at point P.
It is defined as the sum of normal curvatures of a hypersurface V n form mutually orthogonal directions
at P and is denoted by M. Then
M = Ω ij g ij
(b) Minimal Hypersurface
The hypersurface V n is said to be minimal if M = 0
i.e., Ω ij g ij = 0
(c) Principle normal curvatures
The maximum and minimum values of Kn are said to be the principle normal curvatures of V n at P.
Since these maximum and minimum values of Kn correspond to the principal directions of the symmetric
tensor