You are on page 1of 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

## 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 Islams 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

1.1.
1.2.
1.3.
1.4.
1.5.
1.6.

## n-dimensional Space ............................................................................................... 1

Superscript and Subscript ....................................................................................... 1
The Einstein's Summation Convention ...................................................................... 1
Dummy Index ....................................................................................................... 1
Free Index ............................................................................................................. 2
Krnecker Delta ..................................................................................................... 2
Exercises ............................................................................................................... 5

## Chapter2 Tensor Algebra ..................................................................................... 6-30

2.1.
2.2.
2.3.
2.4.
2.5.
2.6.
2.7.
2.8.
2.9.
2.10.
2.11.
2.12.
2.13.
2.14.
2.15.

Introduction .......................................................................................................... 6
Transformation of Coordinates ................................................................................ 6
Covariant and Contravariant Vectors ......................................................................... 7
Contravariant Tensor of Rank Two .......................................................................... 9
Covariant Tensor of Rank Two ................................................................................ 9
Mixed Tensor of Rank Two..................................................................................... 9
Tensor of Higher Order ......................................................................................... 14
Scalar or Invariant ................................................................................................ 15
Addition and Subtraction of Tensors ....................................................................... 15
Multiplication of Tensors (Outer Product of Tensors) ............................................... 16
Contraction of a Tensor ........................................................................................ 18
Inner Product of Two Tensors .............................................................................. 18
Symmetric Tensors .............................................................................................. 20
Skew-symmetric Tensor ....................................................................................... 20
Quotient Law ....................................................................................................... 24

xii

2.16.
2.17.

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

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

3.1.
3.2.
3.3.
3.4.
3.5.
3.6.
3.7.
3.8.
3.9.
3.10.
3.11.
3.12.
3.13.

## The Metric Tensor ............................................................................................... 31

Conjugate Metric Tensor (Contravariant Tensor) ...................................................... 34
Length of a Curve ................................................................................................ 42
Associated Tensor ................................................................................................ 43
Magnitude of Vector ............................................................................................. 43
Scalar Product of Two Vectors .............................................................................. 44
Angle Between Two Vectors .................................................................................. 45
Angle Between Two Coordinate Curves .................................................................. 47
Hypersurface ....................................................................................................... 48
Angle Between Two Coordinate Hyper surface ........................................................ 48
n-Ply Orthogonal System of Hypersurfaces ............................................................. 49
Congruence of Curves .......................................................................................... 49
Orthogonal Ennuple .............................................................................................. 49
Examples ............................................................................................................ 52
Exercises ............................................................................................................. 54

Chapter4
4.1.
4.2.
4.3.
4.4.
4.5.
4.6.
4.7.
4.8.

## Christoffels Symbols and Covariant Differentiation ................................ 55-84

Christoffels Symbol............................................................................................. 55
Transformtion of Christoffels Symbols .................................................................. 64
Covariant Differentiation of a Covariant Vector ........................................................ 67
Covariant Differentiation of a Contravariant Vector ................................................... 68
Covariant Differentiation of Tensors ....................................................................... 69
Riccis Theorem .................................................................................................. 71
Gradient, Divergence and Curl ............................................................................... 75
The Laplacian Operator ......................................................................................... 80
Exercises ............................................................................................................. 83

5.1.
5.2.
5.3.

## Riemann-Christoffel Tensor ................................................................................... 85

Ricci Tensor ........................................................................................................ 88
Covariant Riemann-Christoffel Tensor .................................................................... 89

5.4.
5.5.
5.6.
5.7.

## Properties of Riemann-Christoffel Tensors of First Kind Ri j k l ................................... 91

Bianchi Identity .................................................................................................... 94
Einstein Tensor .................................................................................................... 95
Riemannian Curvature of V n .................................................................................. 96

Contents

5.8.
5.9.
5.10.
5.11.
5.12.
5.13.

xiii

## Formula For Riemannian Curvature in Terms of Covariant

Curvature Tensor of V n ......................................................................................... 98
Schurs Theorem ............................................................................................... 100
Mean Curvature ................................................................................................. 101
Ricci Principal Directions .................................................................................... 102
Einstein Space ................................................................................................... 103
Weyl Tensor or Projective Curvature Tensor ......................................................... 104
Examples .......................................................................................................... 106
Exercises ........................................................................................................... 109

Chapter6 The e-systems and the Generalized Krnecker Deltas ................ 111-115
6.1.
6.2.
6.3.
6.4.

## Completely Symmetric ......................................................................................... 111

Completely Skew-symmetric ................................................................................ 111
e-system ........................................................................................................... 112
Generalized Krnecker Delta ................................................................................ 112

6.5.

## Contraction of i jk ............................................................................................ 114

Exercises ........................................................................................................... 115

7.1.

7.2.

7.3.
7.4.
7.5.
7.6.
7.7.
7.8.
7.9.

## Reciprocal Base System Covariant and Contravariant Vectors .................................. 122

On The Meaning of Covariant Derivatives ............................................................. 127
Intrinsic Differentiation ....................................................................................... 131
Parallel Vector Fields ........................................................................................... 134
Geometry of Space Curves ................................................................................. 134
Serret-Frenet Formulae ....................................................................................... 138
Equations of A Straight Line ................................................................................ 140
Exercises ........................................................................................................... 141

8.1.
8.2.
8.3.
8.4.
8.5.
8.6.
8.7.
8.8.
8.9.

## Introduction ...................................................................................................... 142

Newtonian Laws ................................................................................................ 142
Equations of Motion of Particle ............................................................................ 143
Conservative Force Field ..................................................................................... 144
Lagrangean Equation of Motion ........................................................................... 146
Applications of Lagrangean Equations ................................................................... 152
Hamiltons Principle ............................................................................................ 153
Integral Energy .................................................................................................. 155
Principle of Least Action ..................................................................................... 156

xiv

8.10.
8.11.
8.12.
8.13.
8.14.
8.15.

## Generalized Coordinates ...................................................................................... 157

Lagrangean Equation of Generalized Coordinates ................................................... 158
Divergence Theorem, Greens Theorem, Laplacian Operator and Stokes
Theorem in Tensor Notation ................................................................................ 161
Gausss Theorem ............................................................................................... 164
Poissons Equation ............................................................................................. 166
Solution of Poissons Equation ............................................................................. 167
Exercises ........................................................................................................... 169

9.1.
9.2.
9.3.
9.4.
9.5.
9.6.
9.7.
9.8.

## Curvature of Curve, Principal Normal................................................................... 170

Geodesics ......................................................................................................... 171
Eulers Condition ............................................................................................... 171
Differential Equations of Geodesics ...................................................................... 173
Geodesic Coordinates ......................................................................................... 175
Riemannian Coordinates ...................................................................................... 177
Geodesic Form of a Line Element ........................................................................ 178
Geodesics in Euclidean Space .............................................................................. 181
Examples .......................................................................................................... 182
Exercises ........................................................................................................... 186

10.1.
10.2.
10.3.
10.4.
10.5.

## Parallelism of a Vector of Constant Magnitude (Levi-Civitas Concept) ..................... 188

Parallelism of a Vector of Variable Magnitude ......................................................... 191
Subspace of Riemannian Manifold ........................................................................ 193
Parallelism in a Subspace .................................................................................... 196
Fundamental Theorem of Riemannian Geometry Statement ..................................... 199
Examples .......................................................................................................... 200
Exercises ........................................................................................................... 203

Chapter11
11.1.
11.2.
11.3.
11.4.
11.5.
11.6.
11.7.

## Riccis Coefficient of Rotation ............................................................................. 205

Reason for the Name Coefficients of Rotation .................................................... 206
Curvature of Congruence .................................................................................... 207
Geodesic Congruence ......................................................................................... 208
Normal Congruence ........................................................................................... 209
Curl of Congruence ............................................................................................ 211
Canonical Congruence ........................................................................................ 213
Examples .......................................................................................................... 215
Exercises ........................................................................................................... 217

Contents

xv

12.1.
12.2.
12.3.
12.4.
12.5.
12.6.
12.7.
12.8.
12.9.
12.10.
12.11.
12.12.
12.13.
12.14.
12.15.

## Introduction ...................................................................................................... 218

Generalized Covariant Differentiation .................................................................... 219
Laws of Tensor Differentiation ............................................................................ 220
Gausss Formula ................................................................................................ 222
Curvature of a Curve in a Hypersurface and Normal Curvature, Meuniers Theorem,
Dupins Theorem ............................................................................................... 224
Definitions ......................................................................................................... 227
Eulers Theorem ................................................................................................ 228
Conjugate Directions and Asymptotic Directions in a Hypersurface.......................... 229
Tensor Derivative of Unit Normal......................................................................... 230
The Equation of Gauss and Codazzi ..................................................................... 233
Hypersurfaces with Indeterminate Lines of Curvature ............................................ 234
Central Quadratic Hypersurfaces .......................................................................... 235
Polar Hyperplane ................................................................................................ 236
Evolute of a Hypersurface in an Euclidean Space ................................................... 237
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 ndimensional 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 = a i x i . By using summation convention,

i =1
drop the sigma sign and write convention as
n

ai x i = a i x i

i =1

## This convention is called Einsteins 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

## Tensors and Their Applications

a jx j = a1x1 + a2 x2 + + an x n

and

## 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 KRNECKER DELTA
The symbol ij , called Krnecker Delta (a German mathematician Leopold Krnecker, 1823-91 A.D.)
is defined by
1 if i = j
ij =
0 if i j
Similarly ij and ij are defined as
1 if i = j
ij = 0 if i j
1 if i = j
ij =
0 if i j

and

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
= ij
x j
It is also written as
2.

x i x k
= ij .
x k x j

ii = 11 + 22 + 33 + + nn

ii

## (by summation convention)

= 1 + 1+ 1 + + 1

ii = n
3.

a ij kj = a ik
Since

31 1
32 2
33 3
3n n
a 3 j 2j = a 2 + a 2 + a 2 + + a 2

## (as j is dummy index)

Preliminaries

= a32

(as 12 = 32 = = n2 = 0 and 22 = 1)

In general,
a ij kj = a i 11k + a i 2 2k + a i 3 3k + + aik kk + + a in nk
ik
a ij kj = a (as 1k = 2k = = nk = 0 and kk = 1)

4.

j k = ik
ij kj = 1i 1k + i2 2k + i3 3k + + ii ik + + in nk
= ik (as 1i = i2 = i3 = = in = 0 and ii = 1)

EXAMPLE 1
Write

d
dx1 dx 2
dx n
+ 2
+ + n
= 1
using summation convention.
dt
x dt
x dt
x dt

Solution
d
dx 1 dx 2
dx n
+
+

+
=
dt
x 1 dt x 2 dt
x n dt
d dx i
= i
dt
x dt
EXAMPLE 2
Expand: (i) aij xixj; (ii) glm gmp
Solution
(i)

1 j
2 j
n j
aij x i x j = a1 j x x + a2 j x x + + a nj x x

## = 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

(aij x i x j ) = a ij
( xi x j )
(i)
x k
xk

## Tensors and Their Applications

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

= a ij xi jk + aij x j ik ,

as

x j
x k

= (aij jk ) xi + (aij ik ) x j
= aik xi + akj xj
= aik xi + aki xi

as aij jk = aik
as j is dummy index

(aij x i x j )
= 2aik xi

xk
(aij x i x j )

(ii)

x k

## as given aik = aki

= 2aikxi

Differentiating it w.r.t. xl :
2 (aij xi x j )
xk xl

= 2 aik

xi
xl

= 2 aik il
2 (aij xi x j )
xk xl

as a ik il = alk.

= 2alk

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
Differentiating it w.r.t. xi partially,

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

Since

## since i and j are dummy indices

x l m
x m l
x + alm
x =0
xi
x i
x l
= li
xi

and

x m
= mi
xi

= jk

Preliminaries

alm il x m + alm im x l = 0
aim x m + ali x l = 0
as alm li = aim and alm mi = ali .
Differentiating it w.r.t. xj partially
aim

x m
x l
+ ali
x j
x j = 0
aim mj + ali 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) a ij xj

(ii)

( ga )

(iii) Aik Bi

3. Evaluate:
(i) x j ij

(ii) ij kj kl

(iii) ij ij

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

ANSWERS
1.
2.

(i) xi xj
(ii) ds2 = g ij dxi dxj
1
(i) a i1 x + a i2 x2 + a i3 x3 + . . . + a in xn
(ii)

( ga ) +

( g a ) + +

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

3.

(i) xi

4. Cij xi xj

(ii) il

(iii) n

(iii) a i x i x 3 .

( ga )

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

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

## 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
j
A
i
or
A
=
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
Ai =

## 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
Aj
Ai
or
Aj =
i
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
Ai =

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 )
i
i
i
dx i = x dx 1 + x dx 2 + + x dx n
1
2
n

d xi =

x i
x j

dx j

vector.
EXAMPLE 2
Show that

## is a covariant vector where 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

( x 1 , x 2 ,..., x n ) = ( x1 , x 2 ,..., x n )
=

1 2

x + 2 x + + n x n
1
x
x
x

x1 x 2
x n
=
+
+

+
i
x
x 1 x i x 2 x i
x n x i

x j
=
x i
x j x i

x j
=
x i
x i x j

or

## It is law of transformation of component of covariant vector. So,

vector.

is component of covariant
x i

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
Let

dx1 dx 2
dx n
,
, ,
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

## vector of the point ( x 1 , x 2 ,..., x n ) in Y-coordinate system are

dx i
. Then x 1 , x 2 , ..., x n or x i being a
dt

d xi
=
dt

x i dx1 x i dx 2
x i dx n
+
+

+
dt1 dt
dx 2 dt
dx n dt

d xi
=
dt

x i dx j
dx j dt

## It is law of transformation of component of contravariant vector. So,

dx i
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
A ij =

x i x j kl
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
Akl
x i x j
Then A ij called components of covariant tensor of rank two.
Aij =

## 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
Al
=
x k x
Then Aij are called components of mixed tensor of rank two.
A ji

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 Krnecker 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 ij and component of Krnecker delta in Y-coordinate be ij , then according to the law of
transformation
ij =

x i x i x l x k
=
x j x k x j x l

ij =

x i x l k
l
x k x j

## This shows that Krnecker ij is mixed tensor of rank two.

EXAMPLE 4
If Ai is a covariant tensor, then prove that

Ai
do not form a tensor..
x j

Solution
Let X and Y be two coordinate systems. As given Ai is a covariant tensor. Then
Ai =

x k
Ak
x i

Differentiating it w.r.t. x j
Ai

j =
x
x j
Ai
x k
=
x j
x i

x k

x i Ak

Ak
2 xk
+
A
k
x j
x i x j

It is not any law of transformation of tensor due to presence of second term. So,

(1)
Ai
is not a
x j

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

x i x l k
l
x k x j

Tensor Algebra

11

x i
=
x k

x l k

x j l

x l k x k
l =
x x
x j
x j
x i
x i
i
=

= ij
ij =
,
as
j
x j
x j
=

x i x k
k

, as

So, ij 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:

## 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
A
... (1)
x q
When coordinate x i be transformed to x i , the law of transformation of contravariant vector is
Ap =

x i p
A
x p
x i x p q
A i = x i x q A from (1)
x i q
i =
A
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.
Ai =

## 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
Aq
... (1)
x p
When coordinate x i be transformed to x i , the law of transformation of a covariant vector is
Ap =

Ai =

x p
Ap
x i

12

## Tensors and Their Applications

x p x q
Aq
x i x p
x q
Aq
Ai =
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.
Ai =

## 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
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
Aqp =

A ji

x i x q p
Aq
=
x p x j
=

A ji =

x x x x
x x x x

As from (1)

x i x s r
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

y = m1 x + m2 y
put x = x1 , y = x 2 , x = x 1 , y = x 2

...(2)

Tensor Algebra

13

## Consider the contravariant transformation

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

From (1)

x
2
2
= l , but x = x1 , y = x , x = x 1 , y = x
x 1

Then
x x 1
=
=l .
x x 1 1
Similarly,

2
2
y
x
y
x
= l2 = 1 ;
= m2 = 2
x
y
x
x
x
x 1
= m1 = 2 ;
y
x

...(3)

So, we have
A 1 = l1 A1 + m1 A 2

A 2 = l2 A1 + m 2 A 2
Consider the covariant transformation

..(4)

x j
A j ; j = 1,2
x i
x1
x 2
A
+
A2
Ai =
1
x i
x1
Ai =

for i = 1, 2 .
x1
x 2
A
+
A2
1
x 1
x 1
x 1
x 2
A
+
A2
A2 =
1
x 2
x 2
A1 =

From (3)

A1 = l1 A1 + m1 A2

A2 = l2 A1 + m2 A2

...(5)

14

## 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
A

Then Ai ii 2 ...i r

i1i 2 ...i r

x i1 x i2

x i r

p1 p 2 ...p r
x p1 x p 2 x p r A
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
A j1 j 2 ... j s =

x q1 x q 2
x q s

x j1 x j 2
x j s

Aq1 ,q 2 ...,q 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
A ij11ij22......i rj s =

x i1 x i 2 x ir
x p1 x p2 x p r

x q1 x q 2
x q s

Aqp11qp22......q ps r
j1
j2
js
x x
x

i i ...i
Then A j11 2j 2 ...rj s are called component of mixed tensor of rank r + s .
2 ... j s
A tensor of type Ai1j1i 2j...
is known as tensor of type (r, s ) , In (r,s), the first component r
ir
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
is a mixed tensor of type (3, 2) in which contravariant tensor of rank three and covariant
Alm
tensor of rank two. Then according to the law of transformation

Almijk =

x i x j x k x a x b
x x x x l x m

Aab

## 2.8 SCALAR OR INVARIANT

A function ( 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.

( x1 , x 2 ,..., x n ) = ( 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
p1
p2
pr
j1
j2
js
x x
x x x
x
and
B ij11ij22......i rj s =

x i1 x i 2
x ir x q1 x q 2
x q 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 i r x q1 x q 2 x q s

x p1 x p2
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
,..., pr

Cqp11,,qp22,...,
qs
x p1 x p2
x p r x j1 x j 2
x j s
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
C ij11ij22......i rj 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
Akij =

x i x j x r pq
Ar
x p x q x k

and
Bkij

x i x j x r pq
Br
=
x p x q x k

then
Akij Bkij =

x i x j x r pq
Ar Brpq
p
q
k
x x x

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

## 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 ...
(which is covariant tensor of rank m and contravariant tensor of rank n). Then
ln
according to the law of transformation.

A ij11ij22......i rj s =

x i1 x i 2
x ir x q1 x q 2
x q r

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

Aqp11qp22......q ps r

x k1 x k 2
x km x1 x 2
x n

x 1 x 2
x m x l1 x l 2
x l n

B112 2......n m

and
km
Bl1kl12k...2 ...
=
ln

## Then their product is

A ij11ij22......i rj s Blk1l12k...2 ...l nkm =

x i1
x p1

x i r x q1
x pr x j1

x q s x k1
x km x 1
x n

x j s x 1
x m x l1
x l m
Aqp11qp22......q ps r B112 2......n m

Tensor Algebra

17

If
...km
i i ...i
k k ...k
C ij1ij2 ......i rjk1l kl 2 ...
= A j11 j22 ...rjs Bl1l12 ...2 ln n
1 2
s 1 2 ln

and
...,p r
C qp11qp2 2......qpsr 112...2...nm = Aqp11qp22...,
B112 2......n m
qs

So,
x q s x k1
x i1
x ir x q 1

x j s x 1
x p1
x p r x j1
x k m x 1
x n
l l C qp11qp2 2......q ps r1 12 2......n h
x m x 1
x n

...km
C ji11ij22......irj ksl11kl 22 ...
ln =

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
Ai =

x i k
A
x k

...(1)

Bj =

x l
Bl
x j

...(2)

and

## Multiply (1) and (2), we get

Ai B j =

x i x l k
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
A ji =

x i x q p
Aq
x p x j

and

Bmkl =

x k x l x t rs
Bt
x r x s x m

18

## Multiplying these, we get

A ji Bmkl =

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

Aimijk =

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

Aimijk =

Since

x s
= sp
x p

x j x k x t pqr
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

l
l
Consider the tensors Akij and Bmn
if we first form their outer product Akij B mn
and contract this by
k
putting l = k then the result is Akij B mn
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
A
Bq
=
and
=
Bi
A
x p
x i
i

x i x q p
A Bq
=
AB
x p x i
x q p
x q
A
B
,
= qp
=
since
q
x p
x p
i

= 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
A ji =

x i x q p
x k x l x t rs
kl
A
Bt
q and Bm =
x p x j
x r x s x m

...(1)

Put k = j then
Bmjl

x j x l x t rs
Bt
x r x s x m

...(2)

## Multiplying (1) & (2) we get

A ij B mjl =
=
A ji Bmjl

x i x q x j x l x t p rs
Aq Bt
x p x j x r x s x m
x i x l x t q p rs
r Aq Bt
x p x s x m

x i x l x t p qs
Aq B t
=
x p x s x m

since

x q x j
x q
=
= qr
x j x r
x r

since

qr Btrs = Btqs

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

## 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
ijk
= Almjik
(2) The tensor Alm
is symmetric if Alm

## THEOREM 2.10 A symmetric tensor of rank two has only

1
n(n + 1) different components in n2

dimensional space.
Proof: Let A ij be a symmetric tensor of rank two. So that Aij = A ji .
A11
21
A
The component of A ij are A31

M
A n1

A12
A 22
A32
M
An2

A13
A 23
A33
M
An3

L
L
L
L
L

A1n

A2n
A 3n

M
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.
So, the remaining different components are

1 2
(n n) . Hence the total number of different
2

components
= n+

1 2
1
(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
ijk
(ii) The tensor Alm
is Skew-symmetric if Alm
= Almjik

THEOREM 2.11

## A Skew symmetric tensor of second order has only

1
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
21
A
ij
The components of A are A 31

M
A n1

[Since A

ii

A12
0
A 32
M
An 2

A13
A 23
0
M

L
L
L
L

An3 L

A1n

A2n
A3n

M
0

## = 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
Aij =

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

## Aij = Sij + Tij

where

Sij =

1
1
( Aij + A ji ) and Tij = ( Aij A ji )
2
2

S ji =

1
( A + Aij )
2 ji

Now,

S ji = Sij
So, Sij is symmetric tensor..
and
Tij =

1
( A + A ji )
2 ij

1
( A Aij )
2 ji
1
= ( Aij A ji )
2

T ji =

T ji = Tij
or

Tij = T ji
So, Tij is Skew-symmetric Tensor..

22

## Tensors and Their Applications

EXAMPLE 9
j k
If = a jk A A . Show that we can always write = b jk A j Ak where b jk is symmetric.

Solution
As given
= a jk A j A k
Interchange the indices i and j

...(1)

= ak j A k A j

...(2)

## Adding (1) and (2),

2 = (a jk + akj ) A j A k
=

1
(a jk + akj ) A j A k
2

= b jk A j A k
where b jk =

1
(a + akj )
2 jk

## To show that b jk is symmetric.

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

b jk =
bkj

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
x
x
symmetric covariant tensor of rank two.
Solution
As Ti is covariant vector. Then by the law of transformation
Ti =

x k
Tk
x i

Tensor Algebra

23

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

Ti
j = x j
x

x k

x i Tk

2 x k
x k Tk
T
+
=
k
x j x i
x i x j
T j
x j

2 xk
x k x l Tk
T
+
k
x j x i
x i x j x l

...(1)

Similarly,
T j

2 xk
x k x l Tk
T
+
k
x i
x i x j
x j x i x l
Interchanging the dummy indices k & l
T j
x i

2 x k
x k x l Tl
T
+
k
x i x j
x i x j x k

...(2)

## Substituting (1) and (2), we get

x k x l Tk Tl
Ti T j

=
x i x j x l x k
x j x i
This is law of transformation of covariant tensor of rank two. So,

Ti T j

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

j
i
= x j x i

A ji = Aij
or

Aij = A ji

Ti T j

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
So, Aij =

24

## 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
A ( i, j, k ) B k =

x p x q
x i x j

A( p, q, r )B r

...(1)

## Since B k is contravariant vector. Then

x k
x r
Bk = r Br
Br = k B k
or
x
x
So, from (1)
A ( i, j, k ) B k =
A ( i, j, k ) B k =
A ( i, j , k ) =

x p x q
x i x j

A ( p , q, r )

x p x q x r
x i x j x k
x p x q x r

x r
x k

Bk

A( p, q , r )B 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

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
A
Ai =
or
x p
x j q
B or
Bj =
x q
x k
C k = r Cr
or
x
So, from (1)
i

A (i, j , k ) A B C k = A( p, q, r)

x p i
A
x i
x q j
B
=
x j
x r k
C
=
x k

Ap =
Bq
Cr

x p xq x k
x i x j xr

Ai B jCk

As A i , B j , C k are arbitrary..
Then

x p xq x k
A( p, q, r )
A ( i, j , k ) =
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
A ij =

## Cofactor of Aij is the determinan t Aij

d
A is a contravariant symmetric tensor of rank two which is called conjugate (or Reciprocal) tensor
of Aij .
ij

THEOREM 2.13 If Bij is the cofactor of Aij in the determinant d = |Aij| 0 and Aij defined as
A ij =

Bij
d

## Then prove that Aij A = .

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

(i)

ki

Aij Bij = d
Aij

Bij
d

=1

Aij Aij = 1,

given

Aij =

Bij
d

26

(ii)

Aij B kj = 0
Aij

B kj
d

= 0,

Aij Akj = 0

d 0
if i k

## from (i) & (ii)

1 if i = k
Aij Akj = 0 if i k

Aij Akj = ik

i.e.,
2.17 RELATIVE TENSOR

i i ...i
If the components of a tensor A j11 2j 2 ...rj s transform according to the equation
2 ...k r
Alk11l 2k...
ls

x
=
x

Aij11i 2j 2......i rj s

x k1 x k2
x k r x j1 x j2
x js

x i1 x i 2
xi r x l r x l 2
x l s

x
is the Jacobian of transformation. If
x
= 1, the relative tensor is called a tensor density. If w = 0 then tensor is said to be absolute.

i i ...i
Hence A j11 2j 2 ...rj r is called a relative tensor of weight , where

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

a ir ami = rm
Solution
Given that
or

i m
i
x i = am x + b
a im x m = x i bi

...(1)
...(2)

## Multiplying both sides (2) by a ir , we get

a ir ami x m = a ir x i bi a ir
rm x m = a ir x i bi a ir as given a ir ami = rm
i i
i i
x r = a r x b a r as rm x m = x r

or

i i
i i
xs = as x b as

x s
= a is
x i

..(3)

Tensor Algebra

27

## Now, from (1)

i s
i
x i = as x + b

x i
a is
s =
x
The contravariant transformation is

...(4)

x i s
A = a is A s
=
A
x s
The covariant transformation is

...(5)

x s
As = a is As
...(6)
x i
Thus from (5) and (6), it shows that there is no distinction between contravariant and
covariant tensor law of transformation
Ai =

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
j

...(2)
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
i

(k k ) g ij u v = 0
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

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

i

i k

( j l + l j ) Aik = 0
Solution
Given Aij is a Skew-symmetric tensor then Aij = A ji .
Now,
i

i k

## ( j l + l j ) Aik = ij kl Aik + il kj Aik

= ij Ail + il Aij
= A jl + Alj
i

i k

( j l + l 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

a pq

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

x p x q x r x p x q x r x p x q x r
+ j
+
a pq br i

j
k
x x k x i x k x i x j
x x x

= 0

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.,
a mn + a nm = bmn + bnm
m n

m n

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

## 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.
6. Prove that in n-dimensional space, symmetric and skew-symmetric tensor have

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

30

## Tensors and Their Applications

10. If Arpq is a tensor of rank three. Show that Arpr is a contravariant tensor of rank one.
ij
k
k
11. If a k i j is a scalar or invariant, i , j , are vectors then akij is a mixed tensor of type (2, 1).
h i h k
12. Show that if a hijk = 0 where i and i are components of two arbitrary vectors then

## 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
A

ij

ij

= A

x
x

w 2

and

Aij = Aij

x
x

w+ 2

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 ndimensional 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

## 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.
i
j
i
j
ds 2 = g ij dx dx = gij dx dx

So,

## The theorem will be proved in three steps.

(i) To show that dxi is a contravariant vector.
x i = x i (x1, x2 ,...x n )

If

dxi =

x i 1 x i 2
x i n
dx
+
dx
+

+
dx
x i
x 2
x n

dxi =

x i k
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
dxi =

x i k
x j l
j
dx
x
and
d
x
=
x k
x l

## from equation (2)

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

dxk

x j l
dx
x l

x j k l
x dx
x l
x j k l
dx dx
x l

## Since g ij dx i dx j = g kl dx k dx l (i, j are dummy indices).

i
j

g kl g ij x x

x k x l

or

or

g kl gij

k l
dx dx = 0

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

x j x j
x k x l

g ij = g kl

x k x l
x i x j

...(2)

33

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

g ij =

1
1
(g + g ji ) + ( gij g ji )
2 ij
2

g ij = Aij + Bij
1 g +g
( ij
ji ) = symmetric
2
1
Bij = (g ij g ji ) = Skew-symmetric
2
Aij =

where

Now,

(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.
Then,

g ij = g kl

x k x 1
x i x j

34

## Tensors and Their Applications

x k x l
=0
x i x j

g ij g kl

k
l

gij g kl x x

x i x j

i j
dx dx = 0

( g ij dx dx ) = g kl

x k x l i j
dx dx
x i x j

= g kl

x k i x l
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 = ik
g ij g kj = ik

So,
Note

(i)
(ii)

## Tensors g ij and g ij are Metric Tensor or Fundamental Tensors.

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 cos , y = r sin , z = z
So that
x1 = x, x 2 = y, x 3 = z and x 1 = r , x 2 = , x 3 = z

...(1)

Let g ij and g ij be the metric tensors in Cartesian coordinates and cylindrical coordinates
respectively.

35

## The metric in Cartesian coordinate is given by

2
2
2
ds 2 = dx + dy + dz

12
2 2
3 2
ds 2 = (dx ) + (dx ) +(dx )

...(2)

i
j
ds 2 = g ij dx dx

But

## = 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
g ij = g ij

x i x j
, since g ij is Covariant Tensor of rank two. (i, j = 1, 2, 3)
x i x j

for i = j = 1.
2

g11

x1
x 2
x 3
= g11 1 + g 22 1 + g 33 1
x
x
x

## since g12 = g13 = = g32 = 0 .

2

x
y
z
g11 = g11 + g 22 + g 33
r
r
r

and g11 = g 22 = g 33

## Since x = r cos , y = r sin , z = z

x
y
z
= cos,
= sin ,
=0
r
r
r
= 1.
g11 = cos 2 + sin 2 + 0
g11 = 1

Put i = j = 2.
x 1
g 22 = g11 2
x

2
+ g 22 x

x 2

g 22

3
+ g 33 x

x 2

x
y
z
+ g 22
+ g 33

= g11

36

## Tensors and Their Applications

Since g11 = g 22 = g 33 = 1
z
x
y
=0
= r sin ,
= r cos,

g 22 = ( r sin )2 + (r cos )2 + 0
= r 2 sin 2 + r 2 cos 2
g 22 = r 2
Put i = j = 3.
2

g 33

x1
x 2
x 3

g
+
g
+
g
= 11 3
22
33
3
3
x
x
x
2

x
y
z
= g11 + g 22 + g 33
z
z
z
Since

x
y
z
= 0,
= 0,
= 1 . So, g 33 = 1 .
z
z
z

So,

g11 = 1, g 22 = r 2 , g 33 = 1

and

## (i) The metric in cylindrical coordinates

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

( )

1
ds 2 = g11 dx

( )

+ g 22 dx 2

( )

+ g 33 dx 3

## since g12 = g13 = = g32 = 0

2
2
2
2
ds 2 = dr + r (d ) + d

g ij

since

## g11 g12 g13 1 0 0

2
= g 21 g 22 g 23 = 0 r 0
g 31 g32 g 33 0 0 1

1
g = g ij = 0
0
2
g= r

0
r2
0

0
0
1

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
ij
The second fundamental tensor or conjugate tensor is g =

and

g 11 =

cofactor of g11 in g
g

g 11 =

B11 r 2
= 2 =1
g
r

g 22 =

B12 1
= 2
g
r

g 33 =

B33 r 2
= 2 =1
g
r

Bij
g

g 12 = g13 = g 21 = g 23 = g 31 = g 32 = 0

## Hence the second fundamental tensor in matrix form is 0

0 0
1
0 .
r2
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 cos , y = r sin sin , z = r cos
So that x1 = x, x 2 = y , x 3 = z and x 1 = r , x 2 = , x 3 =
Let g ij and g ij be the metric tensors in cartesian and spherical coordinates respectively..
The metric in cartesian coordinates is given by
2
2
2
ds 2 = dx + dy + dz

( ) + (dx ) + (dx )

ds 2 = dx1
But

2 2

3 2

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

38

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

On transformation
g ij = g ij

x i x j
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
+
g
22
33
x i x j
x i x j
x i x 1
since i, j are dummy indices.
Put i = j = 1
g ij = g11

g11

x 1
x 2
x 3

g
+
g
+
g
= 11 1
22
33
1
1
x
x
x

g11

x
y
z
= g11 + g 22 + g 33
r
r
r

## Since x = r sin cos , y = r sin sin , z = r cos

x
z
y
= cos
= sin sin ,
= sin cos ,
r
r
r
and g11 = g 22 = g 33 = 1 .
So,
g11 = (sin cos )2 + (sin sin )2 + cos2
g11 = 1
put i = j = 2
g 22

x 1
= g11 2
x

2
+ g 22 x

x 2

3
+ g 33 x

x 2

x
y
z
g 22 = g11 + g 22 + g33

since g11 = g 22 = g 33 = 1
x
y
z
= r cos cos ,
= r cos sin ,
= r sin

g 22 = r 2

## Metric Tensor and Riemannian Metric

39

Put i = j = 3
2

g 33

x1
x 2
x 3

g
+
g
+
g
= 11 3
22
33
3
3
x
x
x

g 33

x
y
z
= g11 + g 22 + g33

since g11 = g 22 = g 33 = 1
x
y
z
= r sin sin ,
= r sin cos ,
=0

and

## g 33 = ( r sin sin )2 + (r sin cos )2 + 0

g 33 = r 2 sin 2
So, we have
g11 = 1, g 22 = r 2 , g 33 = r 2 sin 2
g12 = g13 = g 21 = g 23 = g31 = g 32 = 0

and

## (i) The Metric in spherical coordinates is

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

( )

1
ds 2 = g11 dx

( )

+ g 22 dx 2

( )

+ g 33 dx 3

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

g ij
and

## g11 g12 g13 1 0

g g g = 0 r 2
0
= 21 22 23

g 31 g32 g 33 0 0 r 2 sin 2

g = g ij

1
=0
0

0
r2
0

0
0
= r 4 sin 2
r 2 sin 2

(iii) The cofactor of g are given by B11 = 1, B22 = r 2 , B33 = r 2 sin 2 and B12 = B21 =
B31 = B13 = B23 = B32 = 0
The second fundamental tensor or conjugate tensor is g ij =

Bij
g

40

g 11 =

## cofactor of g11 in g B11

=
g
g

r 4 sin 2
r 4 sin 2
=1
=

g 11
g

22

B 22 r 2 sin 2
= 4 2
=
g
r sin

g 22 =
g

33

1
r2

B33
r2
=
=
g
r 4 sin 2

g 33 =

1
r sin 2
2

12
13
21
31
32
and g = g = g = g = g = 0
Hence the fundamental tensor in matrix form is

g
21
= g
g 31

11

ij

12

g 22
g 32

0
g

0
23
g =

g 33
0

0
1
r2

13

r 2 sin 2

EXAMPLE 3
If the metric is given by

( ) ( )
2

ds 2 = 5 dx 1 + 3 dx 2

( )

+ 4 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)

( )

1
ds 2 = g11 dx

## + 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.,

## Metric Tensor and Riemannian Metric

So,

41

1 2
2 2
3 2
1
2
ds 2 = g11 (dx ) + g 22 (dx ) + g33 (dx ) + 2 g12 dx dx

+ 2 g 23 dx 2 dx 3 + 2 g13 dx1 dx 3

...(1)

## Now, the given metric is

1 2
2 2
3 2
1
2
2
3
ds 2 = 5(dx ) + 3(dx ) + 4(dx ) 6dx dx + 4dx dx
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
5 3 0

## g11 g12 g13

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
B11 = Cofactor of g11 =

3 2
=8
2 4

B22 = Cofactor of g 22 =

5 0
= 20
04

B33 = Cofactor of g 33 =

5 3
=6
3 3
3 2
= 12 = B 21
0 4

## B12 = Cofactor of g12 =

B13 = Cofactor of g13 =

B 23

Since

3 3
= 6 = B31
0 2

= Cofactor of g 23 =

g ij =

5 3
0

= 10 = B 32

Bij
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

...(2)

42

Hence,

g ij

3
3
2
5
3
5
2

3 5
3

2 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.
s =

P2
P1

ds =

t2

t1

12

i
j
gij dx dx dt

dt dt

NULL CURVE
dx i dx j
= 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.
If g ij

EXAMPLE 4
A curve is in spherical coordinate xi is given by

## Find length of arc 1 t 2.

2
1 1
3
2
x1 = t, x = sin and x = 2 t 1
t

Solution
In spherical coordinate, the metric is given by
1 2
1 2
2 2
1
2 2
3 2
ds 2 = (dx ) + ( x ) (dx ) + ( x sin x ) (dx )

## Metric Tensor and Riemannian Metric

43

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

given

dx 2 =

dt
t t 1
2

, dx 3 =

2t
t 2 1

dt

2t

dt
1 1
+t
+ t sin sin
dt

t t 2 1
t t 2 1

ds = (dt )

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

dx 1 = dt , dx 2 =

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

5t 2
2
ds 2 = t 2 1 dt
ds =

t 1
2

dt

## Now, the length of arc, 1 t 2, is

2

t2

t1

3.4

ds =

5 t2 1

=
dt =
2 12
t2 1

1
t

15 units

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 i Aijk = Ajk ; g j Aijk = Aik ; g k Aijk = Aij
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

## 3.5 MAGNITUDE OF VECTOR

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

A=
or

g ij Ai A j

i j
A 2 = g ij A A

i
Also, A2 = A j A j as g ij A = A j
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

A 2 = g ij Ai A j
or
A vector of magnitude one is called Unit vector. A vector of magnitude zero is called zero vector
or Null vector.
3.6 rSCALAR
r r
r PRODUCT OF TWO VECTORS
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
i
i j
j
Also,
A B = A B i = g ij A B since Bi = g ij B
r r
i
ij
i
ij
A B = Ai B = g Ai B j since B = g B j
Thus

i.e.,

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

Angle
two vectors
r between
r
Let A and B be two vectors. Then
r r
r r
A B = A B cos
r r
g ij A i B j
AB
cos = r r =

AB
g ij A i A j g ij B i B j
since

r
A =

r
g ij Ai A j ; B =

g ij B i B j

## This is required formula for cos .

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 be angle between the vectors A and B then
r r
r r
A B = A B cos
r r
or
A B = AB cos
g ij A i B j = AB cos

cos =

g ij Ai B j

...(1)

AB

r
r
If A and B are orthogonal then = cos = 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

( )

( )

i j
Conversely if g ij A B = 0 then from (1)

cos = 0 = .
2
r
r
So, two vectors A & B are orthogonal.

r
r
r
r
Note: (i) If A and B be unit vectors. Then A = B = 1. Then

Proved.

r r
cos = A B = gij A i B j

r
r

(ii) Two vectors A and B will be orthogonal if angle between them is i.e., = then
2
2

cos = cos = = 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 2 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 be the angle between unit vectors A and B then
i j
j
ij
ij
i
cos = g ij A B = A j B = A j B Bi = g A j Bi = A Bi

46

## To show that is real i.e., |cos| 1.

Consider the vector lAi + mB i when l and m are scalars. The square of the magnitude of
i
i
j
j
lAi + mB i = g ij (lA + mB ) (lA + mB )
2 i j
i j
i j
2
i j
= g ij l A A + g ij lmA B + m lg ij B A + m g ij B B

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

and

r
r
r
g ij A i B j = cos; 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 + m 2 0

(l + m cos )2 + m 2 m 2 cos 2

(l + m cos) 2 + m 2 (1 cos2 ) 0
This inequality holds for the real values of l & m.
if
1 cos2 0

cos 2 1
cos 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)

ij
B 2 = g Ai A j

...(2)

and
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,
C
s
where
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 is required angle then

cos =

cos =

g ij A i B j
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
g pp g qq

## which is required formula for .

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

g ij
g ii g jj

...(4)

48

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, ... un1) represent n 1 dimensional subspace V n1 of V n. If we
eliminating the parameters u1, u2, ...un1, 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.
i

## Let be a scalar function of coordinates x i . Then (x ) = constant determines a family of

hypersurface of V n.
3.10 ANGLE BETWEEN TWO COORDINATE HYPERSURFACE
Let

and

(x i ) = constant

...(1)

( x i ) = constant

...(2)

## represents two families of hypersurfaces.

Differentiating equation (1), we get
i
dx = 0
x i

...(3)

is orthogonal to dx i . Hence
is normal to = constant, since dx i is
i
x
x i
tangential to hypersurface (1).
This shows that

is normal to the hypersurface (2). If is the angle between the hypersurface (1)
x i
and (2) then is also defined as the angle between their respective normals. Hence required angle
is given by
Similarly

x i x j

g ij i
g ij i
j
x x
x x j
g ij

cos =

...(4)

## Metric Tensor and Riemannian Metric

49

If we take

and

= x p = constant

...(5)

= x q = constant
The angle between (5) and (6) is given by

...(6)

x p x q
x i x j
q
q
x p x p
ij x x
g ij i
g
x x j
x i x j
g ij

cos =

g ij ip qj

g ij ip qj g ij qi qj
g pq

cos =

g pp g qq

...(7)

g ij

cos ij =

g ii g jj

...(8)

ij =

cos ij = 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| = hk

...(1)

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

and
We define

ehi | =

cofactorof eh |i in determinant eh |i
e h|i

## Also, from the determinant property, we get

n

e e

i
h| h | j

h =1

= ij

...(2)

Multiplying by e jk
n

e e

i
h| h| j

h =1

g j k = ij g jk

e e

i k
h | h|

or

h =1

= g ik

...(3)

n

e e

i
h | h | j g ik

h =1

or

i
= j g ik

g jk =

h |k eh | j

...(4)

n

g ij =

h |i

h =1

e h| j

...(5)

51

g ij =

i
h|

e hj|

...(6)

h =1

## 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

u =

C e

i
h h|

...(7)

h =1

Then
n

u i ek |i =

C e e

i
h h | k |i

h =1

or

C
h

h
k

= Ck

h =1

C k = u i ek |i

...(8)

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

ui =

u e
j

i
h| j ek |

h =1

Now,

u 2 = u iu i =

C e C e
i
h h|

k k |i

from (7)

C C e e

i
k h | k| i

h ,k

C C
h

h
k

h ,k

C C
h

(C

)2

h
n

u2 =

h =1

## 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,
h
j i k
h k i j
( 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

= ( 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 is the inclination of two vectors A and B show that
sin =
2

(g hi g ik g hk g ij ) A h A j B i B k
g hj gik A h A j B j B k

Solution
If be the angle between the vectors A and B then
g ij A j B i

cos =

gij A i A j g ik B i B k

## But sin 2 = 1 cos 2

sin 2 = 1

( gij B i A j ) (g hk A h B k )
( 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 g ij )u i + w j = 0
( X ij g ij ) v i + w j = 0
where 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)

( X ij gij )u i + w j = 0

...(3)

( X ij gij )vi + w j = 0
where .

...(4)

given

Multiply (3) & (4) by v j , u j respectively and using (1) and (2), we have
( X ij g ij )u iv j = 0

...(5)

( X ij g 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 g ij )u iv j = 0
Subtract (6) & (7) we get

... (7)

( )g iju i v j = 0
Since and 0.
Hence,
g ij u i v j = 0
So, u and v are orthogonal.
Using (8) in equation (5) & (6), we get

...(8)

X ij u i v i = 0

Proved.

## g dx1 dx 2 ...dx n for the element volume.

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

x k x l
g kl
x i x j

g ij
Since

x k
=
x i

x l
x j

g kl

x
= J (Jacobian)
x
g kl = g

&

g ij = g

54

So,
g = gJ 2
or
J=

g
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
dx 1dx 2 ... dx n =

g 1 2
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

## adu + bdv + cdw + 2 fdvdw + 2 gdwdu + 2 hdudv

3. Show that for an orthogonal coordinate system
g 11 =

1
1
1
, g 33 =
, g 22 =
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
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.
5. Prove that the number of independent components of the metric g ij cannot exceed

8. If is the angle between the two vectors Ai and B i at a point, prove that
2

sin =

(g hi gik g hk gij ) A h Ai B j B k
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

[ij, k ] =

1 gi k g j k g i j

+
k , (i, j, k = 1, 2,...n )
2 x j
x i
x
called Christoffel 3-index symbols of the first kind.

...(1)

k
= g k l [ij, l ]
i j

...(2)

and

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
n (n + 1) independent components. So, the number of
2
1
1
independent components of Christoffels symbols are n n(n + 1) = n 2 (n + 1) .
2
2
symmetric tensor of rank two and has

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 Christoffels symbols of first kind

[ij, k ] =

g ij
1 g ik g jk
j +
k
i
2 x
x
x

[ ji, k ] =

1 g jk gik g ji

+
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

(i)

[ij, m] =

(ii)

[ik , j ] + [ jk , i] =

k
g km
i j

g ij
x k

g ij

i
j
= g jl g im

l
k
x

m k
Proof: (i) By Christoffels symbol of second kind
(iii)

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

= lm [ij , l ] as g k m g k l = lm

k
g km = [ij, m]
i j
(ii) By Christoffels symbol of first kind

and

[ik , j ] =

1 g k j g i j g i k

2 x i x k x j

...(1)

[ jk ,i ] =

1 g ki g ji g jk

2 x j x k
x i

...(2)

## adding (1) and (2),

[ik , j ] + [ jk , i ] =

1 g i j g ji

+
g = g ji
2 x k x k since ij

## Christoffel's Symbols and Covariant Differentiation

[ik , j ] + [ jk , i] =

57

1 g ij g ij
2
=
2 x k x k

(iii) Since g ij g lj = li .
Differentiating it w.r.t. to x k , we get
g ij

glj
x k

+ glj

g ij
=0
x k

## Multiplying this equation by g lm , we get

g ij g lm

g lj
x k

+ g lm g lj
g lm glj
mj

g ij
=0
x k

g
g ij
g ij g lm ljk
k =
x
x
g lj
g ij
ij lm

g
g
{
[
lk
,
j
]
+
[
jk
,
l
]
}
= [lk , j ] + [ jk , l ] .
=
since
x k
x k

g im
lm
ij
ij
lm
= g g [lk , j ] g g [ jk , l ]
x k

g im
i
m
= g lm g ij

k
x
l k
j k
Interchanging m and j, we get
i
j
g ij
lj
im

g
=

x k
l k
m k

or

g ij
i
j
g ij g im
as g lj = g jl
k =
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

g ik

## g11 g12 ... g1n

g

21 g 22 ... G2 n

=
M

g n1 g n 2 ... g nn

58

and

g = gik

g11
g 21
=
M

## g12 ... g1n

g 22 ... g 2 n

g n1

g n 2 ... gnn

But

g ik g = lk

Take

l = k.

il

g ik g ik = kk = 1

g ik = [g ik ]1 =

Gik
(Theorem 2.13 , Pg 25)
g

g = g ik Gik

## Differentiating w.r.t. g ik partially

g
g ik
=1
= Gik since
g ik
g ik
Now,
g
g g ik
j =

g ik x j
x
= Gik

gik
x j

But Gik = gg ik
g
ik gik
= gg
j
x j
x
g
1 g
g ik ikj
j =
g x
x
1 g
g ik
= g ik { [ jk , i] + [ij, k ] } as
= [ jk , i] + [ij, k ]
g x j
x j
ik
ik
= g [ jk , i] + g [ij , k ]

k i
= +
j k i j

...(1)

## Christoffel's Symbols and Covariant Differentiation

59

i i
1 g
+ as k is dummy indices
j =
g x
j i j i
i
1 g
= 2
j
g x
j i
i
1 g
=
j
2 g x
j i
log( g )
x

j i

Proved.

EXAMPLE 1
If g ij 0 show that

x j

= x j [ik , ] i k

i k

( [j, ] + [j , ] )

Solution
By Christoffels symbol of second kind

= g [ik , ]
i
k

Multiplying it by g , we get

g = g g [ik , ]
i
k

g = [ik , ] as g g = 1
i k

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

[ik , ]
g
=
j
x j
x
i k
g

since
g

+
x j i k

ik,

j =
i k x
x j

g
x j

= [j, ] + [j, ]

+ ([j, ] + [j , ]) =
[ik , ]
j
i
k
i
k
x
x j

60

## Tensors and Their Applications

x j

= x j [ik , ] i k ( [i, ] + [j, ] )

i k

Solved.

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
j
2 gii x i
i i 2 x
i j 2 x
j j
Solution
The Christoffels symbols of first kind

[ij, k ] =
(a) If i = j = k
The equation (1) becomes

[ii, i] =
(b) If i = j k
The equation (1) becomes

[ii, k ] =

1 g jk g ik g ij

2 x i
x j x k

1 gii
2 x i

1 g i k gi k g i i
+ i k
2 x i
x
x

Since g ik = 0 as i k (given)
1 g ii
1 g jj
or [ jj, i] =
k
2 x
2 xi

[ii, k ] =

[ij,i ] =

1 g ji g ii g ij
+ j i

2 x i
x
x

(c) i = k j

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

...(1)

## Christoffel's Symbols and Covariant Differentiation

61

k
=0
i j
i
= g ii [ii, i ]
i i

(ii)

= g ii
=

1 gii
from (a)
2 x i

1
1 g ii
g ii =
i as
gii
2 gii x

i
1 log g ii
=
2 x i
i i
(iii)

i=k j
i
ii
= g [ij, i]
i j
1
1
ii
= g [ij, i] as g =
ii
gii
i
1 log gii
=
2 x j
i j

(iv)

j=ki
i
ii

= g [ jj, i]
j
j

1 1 g j j
=
from (b)
g ii 2 x i
1 g j j
i
= 2g
i
i i x
j j

EXAMPLE 3
If ds 2 = dr 2 + r 2 d 2 + r 2 sin 2 d 2 , find the values of
1
3
[22, 1] and [13, 3], (ii)
and
2 2
1 3
Solution
(i)

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

Clearly,
g11 = 1, g 22 = r 2 , g 33 = r 2 sin 2 and gi j = 0 for i j

Solved.

62

## Tensors and Their Applications

Also,

22
g 11 = 1, g =

1
1 g 33 =
,
2
r sin 2
r2
(See Ex. 2, Pg. 39, and g ij = 0, for i j. )

## (i) Christoffel Symbols of first kind are given by

[ij, k ] =

1 g jk g ik g ij
+

, i, j, k = 1, 2,3
2 x i
x j x k

## Taking i = j = 2 and k = 1 in (1)

[22, 1] =
=

1 g 21 g 21 g 22
+ 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)

[13,3] =
=

## 1 g33 g13 g13

+ 3 3
2 x1
x
x
1 r 2 sin 2
since g13 = 0
2
r

[13, 3] =

r sin 2
(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

2
2

2
2

= r
2 2

...(1)

## Christoffel's Symbols and Covariant Differentiation

63

and
3
31
32
33
= g [13, 1] + g [13, 2] + g [13, 3]
1
3

=

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

3
1
1
2

= 2 2 r sin =
1
3
r
r sin

## 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
[ ij , k ] =

1 gik g ik g ij
+

2 x i
x j x k

...(1)

g ij =

x p x 2
g pq
x i x j

...(2)

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

g ij
x k

x k

x p x q

x i x j g pq

x k

x p x q

x i x j

p
q g

pq
g pq + x x

i
j
x x x k

g i j

2 x p x q x p 2 x q

=
x k x i x j + x i x k x j
xk

p
q g
r

pq x
g pq + x x

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
x i

2 x p x q x p 2 x q
+ k
= i k
j
x x i x j
x x x

r
q g
p

rq x
g pq + x x

x k x j x p x i

...(4)

64

## Tensors and Their Applications

and interchanging j, k and also interchange q, r in the last term of equation (3)
2 x p x q x p 2 x q
g ik
+ i
= j i
i
x j
x x k x j
x x x

p
r
q g

pr
g pq + x x x
i
k
j
q

...(5)

Substituting the values of equations (3), (4) and (5) in equation (1), we get
[ij, k ] =

1 2 x p x q
x p x q x r
2 i j k g pq + i
2 x x x
x x j x k

[ij, k ] =

2 x p x q
x p x q x r 1 g rp g qr g pq

g
+
+ p
pq
x i x j x k
x i x i x k 2 x q
x
x r

g rp g qr g pq
q +

x p
x r
x

2 x p x q
x p x q x r
g
+
[ pq, r]
...(6)
pq
x i x j x k
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.
[ij, k ] =

## (ii) Law of Transformation of Christoffel's Symbol of the Second Kind

k
kl
kl
g
[
ij
,
l
]
=
is function of coordinates x i and g [ij, l ] =
Let
i j
i
x . Then

k
in another coordinate system
i j

2 x p x q
x p x q x r
g pq + i
[ pq, r ] from (6)
i
j
l
x x x
x x j x l
is contravariant tensor of rank two.
[ij, l ] =

kl
As g

g kl =

x k x l st
g
x s x t

Now
g kl [ij, l ] =

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

g
g
+
pq
x t x l x i x j
x s

x l x r x p x q st

x t x l x i x j g [ pq, r ]

x k
x s

x k 2 st 2 x p
x k r x p x q st
x l x q

g
g
+

g
[
pq
,
r
]
= qt
as
t
pq
t
s
i j
s
i
j
t
l
x
x x
x
x x
x x

x k 2 x p sq
x k x p x q sr
g
g
+
g [ pq, r ] as rt g st = g sr
pq
x s x i x j
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
p + s

x s x i x j
x x i x j p q
s
sr
= sp and g [ pq, r ] = p q

g kl [ij, l ] =
Since g sq g pq

k
=
i j

x k 2 x s
x k x p x q
+
x s x i x j x s x i x j

p q

....(7)

It is law of transformation of Christoffels symbol of the second kind. But it is not the law of
transformation of any tensor. So, Christoffels symbol of the second kind is not a tensor.
Also, multiply (7) by

x s
, we get
x k
x s
dx k

Since

k
x s x k 2 x s
x s x k x p x q
+
=
x k x s x i x j x k x s x i x j
i j

p q

x s x k
= ss = 1
x k x s

x s
x k

k
2 xs
x p x q
+
=
x i x j x i x j
i j

x s
2 xs
=
x k
x i x j

p q

k x p x q s
i j i

j
x x p q

...(8)

## It is second derivative of x s with respect to x s in the terms of Christoffels symbol of second

kind and first derivatives.
THEOREM 4.4 Prove that the transformation of Christoffels 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 Christoffels symbols of
second kind (equation (7)) is
k
x k 2 x s
x k x p x q s
+
=

x s x i x j
x s x i x j p q
i j
When coordinate x i be transformed to x i . Then
k x i x j x r
r
2 x k x r
+
=

u
v
k
x u x v x k
i j x x x
u v

...(1)

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

2 k
r
r s x p x q x r
x x

=
+
+

u
v
s
u
v
k
u v p q x x x
x x x

2 x s x r x i x j
x i x j x s x u x v
as

p
i

x x

...(2)

x s x i
x s
=
x i x u
x u

...(3)

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

x v

x s

x i

x i x s x i

x u + x i x v x u

2 s

= x

x u x v

2 x s x j x i
x s 2 x i
2 xs
+
=
x i x j x v x u
x i x u x v
x u x v
Mutiply (5) by

...(4)

x r
.
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
+
=
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
s
x u x v x s
u v
p q x x x

...(5)

...(6)

## 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
Ap
Ai =
x i

...(1)

## Differentiating (1) partially w.r.t. to x j ,

Ai
2 x p
x p A p
A
+
=
p
x j
x j x i
x i x j
Ai
2 x p
x p A p x q
A
+
=
p
x j
x j x i
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
As + i
j =
j
i
x
x x
x x q x j
Since we know that from equation (8), page 65,
2 xs
x s
=
x i x j
x k
Substituting the value of

k x p x q s
i

j
i j x x p q

2 xs
in equation (3), we have
x i x j
x s
Ai
= k
x j
x

k x p x q s
x p A p x q

A
+

s
x i x q x j
i j x i x j p q

k x s
x p x q
= k As i
x x j
i j x
k
Ai
x p x q
A
+
=

k
x j
x i x j
i j

s
x p x q A p

As + i
x x j x q
p q

A p
s
q As

x
p q

...(2)

...(3)

68

k
x p x q
Ai

A
=

k
x i x j
x j
i j

A p
s

x q As p q

...(4)

Ai , j =

Ai
k
Ak
j
x
i j

...(5)

Ai , j =

x p x q
Ap, q
x i x j

..(6)

## 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
A
x s
x s i
A
As =
x i
Ai =

or

## Differentiating it partially w.r.t. to x j , we get

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

...(1)

2 xs
x s k i x p x q s
=

x k i j
x i x j p q
x j x i
substituting the value of

2 xs
in the equation (1), we get
x j x i
A s
x s
=
A j
x k

A s A q
x s
=
A q x j
x k

k i x p x q s i x s A i
A i

A + i
x x j p q
x x j
i j
k i x p i x q
A i A
x
x j
i j

s x s A i

+ i
j
p q x x

## Christoffel's Symbols and Covariant Differentiation

69

Interchanging the dummy indices i and k in the first term on R.H.S. and put
A s x q
x s
=
x q x j
x i
x q
x j

A s

x s
p s

x q + A p q = x i

i
A i
+ Ak
j
x
k

x p i
A = A p we get
x i

i k x q p s x s A i
A j A
+ i
j
x
k j
p q x x
i

x i x q
=
j
x s x j

k A i
A + j
j
x

A s

p s

+
A

q
x
p q

...(2)

## Now, we introduce the comma notation

Ai
i
+ Ak

j
x
k j
Using (3), the equation (2) can be expressed as
A,i j =

...(3)

x p x q
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 .
A,ij =

## 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
Aij =

x p x q
A pq
x i x j

...(1)

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

Aij

A pq
x k

x p x q

x i x j

A pq

x k
Aij

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

x k
as

x p x q A pq

+ k
i
j
k
x x x
x

A pq x r
(since A pq components in x i coordinate)
x r x k

70

A pq

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

A pq

2 x p x q
x q
A
=
lq
x i x k x j
x j

h x l l x p x r
h

i
k
i k x
p r x x

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

h x l l x p x r
2 xl
h i
i
k =
k
x x
i k x
p r x x
A pq

h
x l x 2 l
x p x q x r
2 x p x q
A

lq
lq
= i k
x h x j p r
x i x j x k

x i x k x j

A pq

h
2 x p x q
l
Ahj

i
k
j =
x x x
i k
p
as Ahj = Alq

x p x q x r
Alq i
r
x x j x k

...(3)

x l x q
by equation (1)
x h x j

and
Apq

x p 2 x q
x p 2 x l
A
=
pl
x i x j x k
x i x j x k
= Apl

x p
x i

h x l l x q x r
h

j
k
j k x
q r x x

h
x p x l l x q x r x p
=
A

Apl
pl

x i x h q r x j x k x i
j k
Apq

h
x p 2 x q
l x p x q x r
A

A pl
=

i
ih
j
k
x i x j x k
j k
q r x x x

...(4)

Substituting the value of equations (3) and (4) in equation (2) we get,
A pq
h
h
l
l x p x q x r
r Alq p r A pl q r
+ Ahj +
=
Aih
i
j
k
k

x x x
x
i k
j k
x
Aij

h
h
A pq
l
l x p x q x r

ih
hj
lq
pl
r
i
j
k
x k j k
i k
p r
q r x x x
x
Aij

## Christoffel's Symbols and Covariant Differentiation

Aij , k =

71

Aij

h
h
Aih Ahj , then
x
j k
i k
k

Aij , k = Apq , r

x p x q x 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
+ A lj + A il
A ,k =
k
x
l k
l k
A ij
i
l
+ A lj Ali
Aij ,k =

k
x
l k
j k
ij

and

ij ... l
In general, we define the covariant deriavative xk of a mixed tensor Aab
by the formula
... c
ij ... l
Aab
... c ,k

Note:

ij ...l
Aab
pj ... l i
ip... l j
ij... p l
...c
+ Aab
=
+ Aab... c
+ + Aab... c

...c
k
x
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

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
ij,k =

ij

i
l
+ lj il

x
l k
j k
k

i i
= 0+

j k j k
ij,k = 0 as

ij

i i
= 0; lj =

x
l k j k
k

Also, consider first the tensor gij and the covariant derivative of gij is
g ij , k =

g ij

m
m
g mj gim

x
i k
j k
k

72

## Tensors and Their Applications

g ij , k =
g ij

But

x k

g ij

m
[ik , j ] [ jk , i] as g mj = [ik , j ]
x
i k
k

= [ik , j ]+ [ jk, i]

g ij , k = g ij g ij
x k x k
g ij , k = 0

So,

## Since we know that g im g mj = ij . Similarly taking covariant derivative, we get

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

But

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

EXAMPLE 4
Prove that if A ij is a symmetric tensor then
Ai j, j =

g jk

1
Ai j g A jk
j
2
x i
g x

Solution
Given that A ij be a symmetric tensor. Then
A ij = A ji

...(1)

We know that
Ai j, k

Ai j
j
l
+ Ail Al j
=
k
x
l k
i k

Ai j, j

Ai j
j
l
+ Ail Al j
=
j
x
l j
i j

Put k = j, we get

log g
Ai j
+ Ail
Al j g hl [ ij, h]
j
x
x l

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

Ai j, j =

...(2)

1 Ai j g
A jk [ij, k ]
g
x j

...(3)

## Christoffel's Symbols and Covariant Differentiation

73

But
A jk [ij , k ] =

1 jk g jk g ki g ij
A i +

2
x j x k
x

A jk [ij , k ] =

g ij
1 jk g jk
jk g ki
jk
A
+
A

A
2
x i
x j
x k

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

A jk

...(4)

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

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

A jk

A jk [ij , k ] =

...(5)

1 jk g jk
A
2
x i

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

Ai j, j =

g jk
1 Ai j g
1
A jk
j
2
g
x
x i

Proved.

EXAMPLE 5
k k
Prove that are components of a tensor of rank three where
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
=
x i x j
x k
x s
x k
or

k x p x q s
i

j
i j x x p q

k
2 xs
x p x q s
+

=

x i x j x i x j p q
i j
k
2 xs
x p x q
+

= i j
x i x j
i j
x x

s x k

s
p q x

k
k
and
i j a
i j b

74

## Using this equation, we can write

and

2 xs
k
x p x q
+

=
i
j

x i x j
x x
i j a

s x k

s
p q a x

k
2 xs
x p x q
= i j+ i
i j b
x x j
x x

s x k

s
p q b x

Subtracting, we obtain
s
k k
s x p x q x k

i
=
j
s
p q a p q b x x x
i j a i j b
Put
s
s
s

= A pq
p q a p q b
Then above equation can written as
x p x q x k
x i x j x s
It is law of transformation of tensor of rank three.
s
Aijk = Apq

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..
Now, we have to prove that Aij , k =
Aij ,k =

Aij
x k

A ij

at P0 .

i
i
+ A j
A

x
k
j k
k

...(2)

## Christoffel's Symbols and Covariant Differentiation

75

g ij
k
Since and [ij,.k ] both contain terms of the type
and using equation (1) we get
x k
i j
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 be a scalar function of the coordinates, then the gradient of is denoted by
grad =

x i

## which is a covariant vector.

(b) Divergence
The divergence of the contravariant vector Ai is defined by
div A i =

A i
i
+ Ak
i
x
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
i
Prove that div A =

1 g Ak
g
x k

Solution:
If Ai be components of contravariant vector then
i
div A i = A, i =

A i
i
+ Ak
i
x
k i

Since

1 g
log g =
=
k
x
g x k
k i
So,
div A i =

A i
1 g k
+
A
i
x
g x k

76

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

A k
1 g k
+
A
k
x
g x k

div A i =

1 g Ak
g
x k

div A i =

...(1)
Proved

(c) Curl
Let Ai be a covariant vector then
Ai
k
Ak
j
x
i j

Ai , j =

A j

k
Ak
x
j i

Aj, i =

and

## are covariant tensor.

So, Ai , j A j , i =

Ai A j

## 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 , k + A jk ,i + Aki, j =

Aij
x

A jk
x

Aki
x j

Solution
Since we know that
Aij , k =

A jk ,i =

Aki, j =

Aij

l
l
Alj Ail
i
k
x

j k
k

A jk

h
l
Alk A jl

x
j i
j k
i

Aki
l
Ali
j
x
k

l
Akl
j
i j

77

## Adding these, we get

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

Aij
x k

A jk

x i

l
Aki l

A
+
A

lj
jl
x j i k
k i

l
l l
l
Ali + Ail
Akl + Alk
j k i j
j i
k j
l
Since is symmetric i.e.,
i k

l l
= etc.
i k k i
=

Aij
x

k
Since Aij is skew-symmetric. Then Alj

A jk
x

Aki l
( Alj + A jl )
x j i k

l
( Ali + Ail ) ( Akl + Alk )
j
i j
= A jl Alj + A jl = 0 . Similarly,,

## Ali + Ail = 0 and Akl + Alk = 0

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

Aij
x

A jk
x

Aki
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 = , is scalar..

Since from (1),
Ai , j A j ,i = 0

Ai A j

=0
x j xi
A j
Ai
=
j
x i
x
A j j
Ai
j
dx
dx
=
xi
x j

A j dx j
dAi =
x i

Integrating it we get
Ai =

A j dx j
i
x

78

## Tensors and Their Applications

A j dx j
i
x

Ai =
, where =
x i
or
Ai = .
Conversely suppose that a vector Ai is such that
=

A j dx j

Ai = , is scalar..
To prove curl Ai = 0
Now,
Ai = =

and
So,

x i

2
Ai
=
x j xi
x j
A j
2
=
x i
x i x j
Ai A j

=0
x j xi

So,

curl Ai = Ai , j A j ,i =

So,

curl Ai = 0

Ai A j

=0
x j x i
Proved.

THEOREM 4.6 Let and be scalar functions of coordinates xi. Let A be an arbitrary vector then
(i)

(ii)

() = +

(iii)

2 () = 2 + 2 + 2

(iv)

div () = 2 +

## Proof: (i) Since we know that

div A i =

1 g Ai
g
x i

...(1)

replace A i by A i , we get
div( A ) =
i

1 g Ai
g
x i

## Christoffel's Symbols and Covariant Differentiation

1
g

79

( g Ai )

+
g
A

i
x i
x
i

A +

1 ( g A )
g

= A i + div A i
Thus
div( A) = A + div A

...(2)

## (ii) By definition of gradient,

() =

( )
x i

+ i
i
x
x

() = +
Thus
(iii) Taking divergence of both sides in equation (3), we get
div () = div [ + ]
2 () = div () + div ()
= + div () + + div ()
2 () = div () + div () + 2
Thus,
2 () = 2 + 2 + 2
(iv) Replace A by in equation (3), we get
div () = + div ()
div () = + 2
THEOREM 4.7 Let Ai be a covariant vector and a scalar function. Then
(i)

(ii)

curl () =

## Proof: (i) Let Ai be a covariant vector then

curl A = curl Ai = Ai , j A j ,i
Replacing Ai by Ai , we get

( )

...(3)

80

## Tensors and Their Applications

= , jAi + Ai , j , i A j A j , i
= ( Ai , j A j , i) + ( Ai , j A j ,i )
= Ai + curl Ai
So,
curl (A) = A + curlA

...(1)

## (ii) Replacing A by in equation (1), we get

curl () = curl () +
Interchange of and , we get
Since curl ( ) = 0.
So,

curl () = curl () + .

curl () = .

Proved.

## 4.8 THE LAPLACIAN OPERATOR

The operator 2 is called Laplacian operator read as "del square".
THEOREM 4.8 If is a scalar function of coordinates x i then
2 =

g g kr r
k
x
g x

Proof: Since
2 = div grad

...(1)

and
grad =

,
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)
Now, the contravariant vector A k associated with
A k = g kr
Since
div A i =

(Covariant vector) is
x r

x r

1 g Ak
, (Sec Ex. 7, Pg 75)
g
x k

81

## So, from (1)

1

2 = div g kr r =
x

g g kr r

k
x

Proved.

EXAMPLE 9
Show that, in the cylindrical coordinates,
2V =

1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2

by Tensor method
Solution
The cylindrical coordinates are (r , , z ). If V is a scalar function of (r , , z ).
Now,
2V = V
Since
V = i

V
V
V
+j
+k
r

Let
V
V
V
, A2 =
, A3 =
...(1)
r

z
Then V = iA1i + jA2 j + kA3 k , since V is covariant tensor. The metric in cylindrical coordinates
A1 =

is
ds 2 = dr 2 + r 2 d2 + dz 2
here, x1 = r , x 2 = , 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 =

k
1 gA
x k
g

82

## Tensors and Their Applications

1 ( g A )
=

div (V ) =

1 ( g A )
g

( g A )
x

1 ( rA1 ) ( rA 2 ) ( rA3 )
+
+

r r

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
11
12
13
A1 = g A1 + g A2 + g A3

A1 = g 11 A1 as g 12 = g 13 = 0
Similarly,
A 2 = g 22 A2
A3 = g 33 A3
and
g 11 =

Cofactor of g11 in g r 2
= 2 =1
g
r

g 22 =

Cofactor of g 22 in g
1
= 2
g
r

g 33 =

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

So,
11
A1 = g A1 = A1
1
22
A 2 = g A2 = 2 A2
r
33
A3 = g A3 = A3 .

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

...(2)

## Christoffel's Symbols and Covariant Differentiation

83

So,
1 V 1 V V
i
2V = div A = r r r r + r r 2 + z r z

div (V ) =

1 V
r
r r r

1 V
+
r

V
+ r

z z

1 V 1 2 V
2V
+
r
+
r

r r r r 2
z 2

div (V ) =

1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2

2V =

1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2

EXERCISES
1. Prove that the expressions are tensors
(a) Aij , l =

(b)

Air jk ,l

Aj Ai
jl
x i l

Aij
l

Air jk
x

r A
r

A Ar Ar

l jk j l i k k l ij + l ijk

2. Prove that

Ai j, j

1 ( Ai
g

g)
j

## 3. If A ijk is a skew-symmetric tensor show that

j k
Ak
i j

( 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

## 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
2
V =

by tensor method.

1 2 V
1

V
1
2V
r
+ 2
sin
+ 2 2
2
r sin 2
r r r r sin

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
A i,j =

Ai
A
x j i j

...(1)

## Differentiating covariantly the equation (1) w.r. to x k , we get

Ai , j
A, j
Ai ,
Ai , jk =
k
x
i k
j k
=

A A

i A
A

j
j
x

i j i k
j

A
i A
i k x
i k

x k

Ai
2

Ai , jk =

x k x

+
i k
Interchanging j and k in equation (2), we get

Ai
A
+

A
j
j k x
j k i

i k

Ai

x j x k
x j
2

Ai , kj =

A A
i j

k
k
j
x
i k x
i k x

A
A
A j k
i k x
i j x

...(2)

86

## Tensors and Their Applications

+
i j
Subtract equation (3) from (2), we get
`

Ai , jk Ai , kj

Ai

A +

k
k i x
k j

=
i k

A
i

i j
A

A
A +
k
x
j
i j k

...(3)

i k A

x j

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

Ai , jk Ai , kj

i k
=
x j

i j
x k

+
i k

j i j

Ai , jk Ai , kj = A Rijk

...(4)

...(5)

where

Rijk

i k
i j

+
=

x j
xk
i k

j i j

...(6)

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 Rijk is a mixed tensor of rank
four. The tensor Rijk is called Riemann Christoffel tensor or Curvature tensor for the metric g ij dx i dx j .
The symbol Rijk is called Riemanns 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
Rijk = 0
Since A is arbitrary. In general Rijk 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 Rijk vanishes identically..
Remark
The tensor

i
Rikl

x k
= i

j k

l
x
k
+
i

j
l

j k

l

j l

...(7)

Riemann-Christoffel Tensor

87

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

x j

j k

xk

Rijk = +

Rijk

i j

i k

i k
=

x j

i k

i j
+

x k
j

i k k

i j

R ikj

i j i k +
=
k i j j i k
x k
x j

i k

=
j
x

Rik

i j
x k

j i

k k i j

= Rijk

## So, Rijk antisymmetric w.r.t. indices j and k.

Theorem 5.2 To prove that
Rijk + R jk i + R ki j = 0
Proof: Since we know that

x k

Rijk
=

i j

Rijk

i k

= i k
x j

j k

i j

i j
+

x k
j

i k

i k i j

...(1)

88

Similarly
R j ki

j
i
j
k

= k i +

x
x
k j i j k i

...(2)

and

Rki

k j

=
xi

k i
+
x j
i

k j k i

...(3)

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

Rij k + R j k i + Rki j = 0
This is called cyclic property.
5.2 RICCI TENSOR
The curvature tensor Rijk can be contracted in three ways with respect to the index and any one of
its lower indices
Rj k , Ri k , Rij
Now, from equation (7), art. 5.1,

x j

Rjk
=

x k

k
=

x j
=

2 log g
x j x k

j
+

x k
j

k k

2 log g
xk x j

log g
Since k =
and and are free indices Rj k = 0 .
x k

## Also for Rij .

Write R i j for R i j

Riemann-Christoffel Tensor

89

Rij = Rij

x
x
j
=
+

i j i i j

Rij = j
x

Rij =

2 log g
x j x i

i
j +
j i i j
x

i
j

+

x
j i i j

...(1)

R ji = i
x

R ji =

2 log g
x i x j

j
i +
i j j i
x

i
j

x
i j i j

...(2)

## (Since and 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 Rik :
Rik = Rik = R ik
5.3 COVARIANT RIEMANN-CHRISTOFFEL TENSOR
The associated tensor
Ri j k l = g i R jkl

...(1)

is known as the covariant Riemann-Christoffel tensor or the Riemann-Christoffel tensor of the first
kind.
Expression for Rijkl
Rijkl = gi Rjkl

90

## Tensors and Their Applications

= g i k j l l j k + j l k j k l

x x

(2)

Now,
g i

=
x k j l
x k
=

g i
gi j l j l x k

[ jl , i ] g i
k
x k
j l x

...(3)

Similarly,
g i

[ jk , i ] gi

=
l
l
x j k
x l
j k x

...(4)
udv duv vdu

By the formula dx = dx dx

## Using (3) and (4) in equation (2), we get

Ri j k l =

Ri j k l =

[ jl , i] gi [ jk , i ]
k
+
x k
xl
j l x

g i

l + gi
j k x
j l

g i

k
j k

[ jl , i ] [ jk , i ] g i g i

+
l j l k + g i j l k g i j k l
k
l
j
k
x
x

x
x

[ jl , i ] [ jk , i]

+
( [il, ] + [l , i ] ) ( [ik , ] + [k , i] )
k
l
x
x
j k
j l

+ [k , i]
[l , i ]
j l
j k

Ri j k l =

[ jl , i ] [ jk , i]

+
[il, ] [ik , ]
k
l
x
x
j k
j l

...(5)

It is also written as

Ri j k l

k
k
+ j k
x
= x
[ jk , i ] [ jl, i] [ik , ]

j l
[il, ]

...(6)

## But we know that

[ jl, i] =

1 gli g ji g jl

+
i
2 x j x l
x

...(7)

Riemann-Christoffel Tensor

91

[ jk ,i ] =

and

1 g ki g ji g ik

+
i
2 x j x k
x

...(8)

## Using (7) and (8), equation (5) becomes

1
Rijkl =
2 x k

g li g ji g jl
j +

xl
xi
x

1

l
2 x

g ki g ji g jk
j + k
x
x i
x

+
[il, ] [ik , ]
j k
j l
2
2
g jk
g jl
1 g il
g ik
+

= 2 j k
i
l
j
l
i
k
x x
x x x x
x x
2

Rijkl

[il , ] [ik , ]
j k
j l

...(9)

Since

= g [ jk , ] and j l = g [ jl , ]
j
k

Ri j k l =

2 g jk
2 g jl
1 2 g il
2 g ik
+

2 x j x k x i x l x j x l x i x l

+ g [ jk , ][il, ] g [ jl , ][ik , ]

...(10)

## This is expression for Ri j k l .

The equation (9) can also be written as
Ri j k l

2 g jk
2 g jl
1 2 gil
2 gik
+

= 2 x j x k x i x l x j x l x i x k

+ g

j k

g
i l
j l i k

(i)

R jikl = Rijkl

(ii)

Rijlk = Rijkl

(iii)

Rklij = Rijkl

(iv)

92

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

Rijkl =

2 g jk
2 g jl
1 2 gil
2 g ik
+

2 x j x k x i x l 2 x j x l x i x k

+
[il, ] [ik , ]
j k
j l

...(1)

## (i) Interchanging of i and j in (1), we get

R jikl

2
2 g jk
2 gik
2 gil
1 g jl
=

= i k
2 x x
x j x l x i x l x j x k

+
[ jl, ] il [ jk , ]
i k

2 g jk
2 g jl
1 2 gil
2 gik
+

2 x j x k x i x l x i x k x j x l

+ [ jl , ]
i k

[ jk , ]
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
2 g ji
2 g jl
1 2 g kl
2 g ki
+

R kjil =
2 x j x i x k x l x j x l x k x i
Now interchange j and l, we get
R klij

2
2 g lj
1 g kj
2 g li
2 g ki
= l i+ k j l j k i
2 x x x x
x x
x x

For

[kj, ] = g

l i
k j
l i

=
g
k j
l i

=
[li, ]
k j

[li, ] = k j [li, ]

l i

+ [kl, ] [ki, ]
j i
j l

+ [kj, ] [ki, ]
l i
l j

Riemann-Christoffel Tensor

93

So,
R klij =

2
2 g lj
1 g kj
2 gli
2 g ki
+

2 x l xi x k x j x l x j x k x i

+
[li, ] [ki, ]
k j
l j

## R klij = Rijkl , from (1)

from equation (1)
Rijkl

2
2 g jk
2 g jk
2 g jl
1 g il
= 2 j k + i l j l j k
x x
x x x x
x x

+
[il , ] j l [ik , ]
j k

Riklj

2
2 g kj
1 g ij
2 g kl
2 g il
= k l + i j k j i l
2 x x x x
x x
x x

+ [ij, ] [il , ]
k l
k j

2
2 g lj
2 gij
2 g lk
1 g ik
+

2 x l x j x i x k x l x k x i x j

+
[ik , ] lk [ij, ]
l j

Riljk =

## 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
n (n 1) .
does not exceed
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 non1
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
2

components of Rijkl is

n (n 1)
.
12

94

## Tensors and Their Applications

Hence the number of distinct non-vanishing components of the curvature tensor Rijkl does not
1 2 2
n (n 1) .
exceed
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

R ijkl

R ijkl

R ijkl, m

i

k
l
x x
mk
+
=
ii
m

jk
jl

jk
i

jl
=
x k

i

ml
m

jl

i

jk + i m m i

x l
mk jl jk ml

i
i
2 2

j l j k
=
x m x k
x m x l

...(1)

i m
Since j k , j l etc. are constant at pole.

So, their derivatives are zero.

x l
R ijlm = i

j l

x m
i

j m

## 1 Details of geodesic coordinate given in chapter curvature

j l

j m

in curve . Geodesic.

Riemann-Christoffel Tensor

95

i
i

i
= j m j l +
l
x
xm
l

R ijlm, k

j m m

j l

i
2 i
2

j m j l
=
x k x l
x k x m

...(2)

and

R ijmk

R ijmk

R ijmk

m
k
x
x
m
+
=
i i

j m j k j m

j k

i
i

j k j m + i i
=

x m
x k
m j k k j m
i
2 i
2

j k
j m

x m x l
x k x l

...(3)

## 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
i
Theorem 5.4 To prove the tensor R j j R is divergence free.
2
Proof: We know that from equation (5)

96

## 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
Rmk , k

1
R, m = 0
2

1
Rmk , k km R , k = 0
2

since R , m = km R, k

k 1 k
Rm m R = 0
2

, k
1
Rmk km R is divergence free.
2
1 i
1
i
The tensor Rmk km R = Gmk or R j j R is known as Einstein Tensor..
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 = pi + qi. and 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

i.e.,

y i = ( p i + q i)s

...(1)

y i = p iu 1 + q i u 2

...(2)

Riemann-Christoffel Tensor

97

## where s = u1 and s = u 2 , three parameters namely , , s can be reduced to two parameters

u and u 2 . Here u1 and u 2 are coordinates of any current point on S.
1

## Let ds 2 = b du du be the metric for the surface S. where

b = g ij

y i y j
(, = 1, 2)
u u

k

Let and
be the Christoffel symbols corresponding to the coordinates y i and u .
i
j

R
Let and R hijk be curvature tensor corresponding to the metrices b du du and g ij dy i dy j .
Since the Greek letters , , , take values 1,2 and so that the number of independent non1
1 2 2
vanishing components of R is n2 (n 2 1) for n = 2, i.e., they are
2 2 1 = 1. Let us
12
12
transform the coordinate system u to u and suppose that the corresponding value of R1212 are

'
.
R1212
Then

'
= R
R1212

u u u u
u 1 u 2 u 1 u 2

= R1

u1 u u u
u 2 u u u
+
R
2

u 1 u 2 u 1 u 2
u 1 u 2 u 1 u 2

= R12

u 1 u 2 u u
u 2 u 1 u u
+
R
21
u1 u 2 u1 u 2
u1 u 2 u1 u 2

= R1212

u1 u 2 u1 u 2
u1 u 2 u 2 u1
+
R
1221
u 1 u 2 u 1 u 2
u 1 u 2 u 1 u 2

+ R 2112

u 2 u1 u1 u 2
u 2 u1 u 2 u 1
+
R
2121
u11 u 2 u 2 u 2
u 1 u 2 u 1 u 2

2
u 1 u 2
u1 u 2 u 2 u u 2 u1

2
+
= 2112 u1 u 2
u 1 u 2 u 2 u 2 u 1 u 2

= R1212 J 2 where J =
so,
= R1212 J 2
R1212

u1
u 1

u1
u 2

u 2
u 1

u 2
u 2

u
u

...(3)

98

## Tensors and Their Applications

Again
b = b
b = b

or
from (3) and (4), we get

u u
u u
u
u

u
u

b = bJ 2

...(4)

R1212

R
= 1212 = K (say)
b
b

...(5)

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

Then

k

,
= 0 at P0
i j g b

and

Rhijk =
+
y k
y j

at P0 .

[, ] b [, ] b
+
at P0 .
R =
u
u

R1212 =

[21,1] b
u

[22,1] b
u1

at P0

...(6)

## from (5) we get

1 [21, 1] b [22, 1]b
+

K = b
u 2
u 1

...(7)

## 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

Let and be the Christoffel symbols of second kind relative to the metrices bdu du
i j g

Riemann-Christoffel Tensor

99

[, ]b

y i y j y k
[ij, k ]g
u u u

[21, 1]b

y i y j y k
[ij, k ]g
u 2 u1 u1

...(8)

Now,
[21, 1] b
u

[ij, k ]g

qi p j p k

u 2

i j k
= q p p

[ij, k ]g y h
y h u 2

i j k h
= q p p q

[ij, k ]g
y h

[21, 1] b
u 2

i k j h
= qq p p

[ij, h ]g

...(9)

y k

Similarly,
[22, 1] b
u

i k j h
= qq p p

[ij, h ]g

...(10)

y k

## Using (9) and (10), equation (6) becomes

[ij, h ]g [ik , h]g
h i j k
R1212 = p q p q y k + y j

h i j k
R1212 = p q p q Rhijk at P0

Since
b = g ij

y i y j
u u

b11 = g ij

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

b11 = g hj p h p j

at P0

...(11)

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

21
22

...(12)

## Dividing (11) and (12), we get

p h q i q k Rhijk
R1212
=
K= b
p h qi p j q k ( gik g hj g ij g hk )

...(13)

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 SCHURS 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=

(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
K=

...(1)

R hijk
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 = lj , we get
(ng ik ih g hk )k , l +(g il kj gik l j ) K , j +( ih g h l ngil )K , k = 0
or,
(n 1) g ik K , l + 0 + (1 n) gil K , k = 0 for ij = 0, i j
g ik K ,l g il K , k = 0
Multiplying by g ik and using
g ik ik = n, g il g ik = kl ,
we get

nK,l kl K ,k = 0 or (n 1)K ,l = 0
or

K, k = 0 as (n 1)K ,l = 0

K
= 0.
x l
integrating it, we get K = constant. This proves that the partial derivatives of K w.r.t. to xs are all zero.
or

## 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
K hk =

ehp e qk rk sk R pqrs
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 )
pr
ps
h h
k k qs
h k
k h qr

...(1)

102

## Since unit vectors e hi , ik are orthogonal. Therefore

e hp e hr g pr = 1
p s

e h eh g ps = 0 etc.
and
Using these in equation (1), then equation (1) becomes

K hk =
K hk =
n

e
k =1

11 0 0
e hp e kq ehr

e ks R pqrs

...(2)

hk

k =1

Put

e hp ekq e rh e ks R pqrs

hk

k =1

p q r s
e e e
h k h k

R pqrs

= 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
p r qs
= e h e h g R qprs
p r
M h = eh e h R pr

...(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

h =1

p r
= eh |e h| R pr

= g pr R pr
= R
n

or

M
h =1

= 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 RICCIS PRINCIPAL DIRECTIONS
Let e ih1 is not a unit vector and the mean curvature M h is given by
Mh =

Rij e hi e hj
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
= 0
ehi

...(1)

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 = g ij

...(1)

## Inner multiplication by g ij , we get

R = n since Rij g ij = R and g ij gij = n

1
= nR

from (1)
Rij =

R
g
n ij

## Hence a space is an Einstein space if Rij =

R
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 (hi g hj ng ij ) = Rij
Since g hk g ik = 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)

Rij = (1 n) gij

R
1
= Rgij
n(1 n ) n

R
g
n ij
This is necessary and sufficient condition for the space Vn to be Einstein space.

Rij =

## 5.13 WEYL TENSOR OR PROJECTIVE CURVATURE TENSOR

Weyl Tensor denoted as Whijk and defined by
Whijk = Rhijk +

1
( 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
K=

p h q i p j q k R hijk
(g hj g ik g ij g hk ) p h qi p j q k

= constant

Riemann-Christoffel Tensor

105

## Since K be independent of the orientation determined by the vector p i and q i .

Then
K=

R hijk
g hj gik gij g hk

...(1)

Multiplying by g hk , we get
g hk Rhijk = g hk K (g hj g ik g ij g hk )
Rij = K (kj 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)

Rij =

R
g
n ij

...(4)

## The equation (3) shows that R is constant since K is constant.

Now, the W tensor is given by
Whijk = Rhijk +

1
[ g R g hk g ij ]
1 n ik hj

Whijk = Rhijk +

1 R
R
g ik g hj g hk g ij

1n
n
n

## from (5), we get

R
= Rhijk + n(1 n) [ gik g hj g hk gij ]
= Rhijk +

Rhijk
R

## , by. eqn. (1)

n(1 n) K

Whijk = Rhijk + K

Rhijk
K

, by equation (3)

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.
Rhijk +

1
[ g R g hk Rij ] = 0
1 n ik hj

Multiplying by g hk , we get
Rij +

1
[g hk g ik Rhj g hk g hk Rij ]
1 n
1
Rij +
[h R nRij ]
1 n i hj
R
Rij + ij (1 n)
1 n
2 R ij

=0
=0
=0
=0

Rij = 0

## Since Rij = 0 g hk Rhijk = 0.

hk
g = 0 or Rhijk = 0

## If Rhijk = 0, then clearly K = 0. So, K is constant.

If g hk = 0 then
K=

K=

R hijk p h qi p j q 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
( p h p j g hj ) (q i q k g ik )

R hijk p h q i p j q k
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
ij
R = g Rij =

Consequently
Rij =

1
Rg .
2 ij

2R1221
g11 g 22

Riemann-Christoffel Tensor

107

Solution
Given that g12 = 0 so that g 12 = 0.
Also,
g ij =

1
1
1
22
g 11 =
& g =g .
g ij
g11
22

## The metric of V2 is given by

i
j
ds 2 = g ij dx dx ; (i, j = 1, 2)
1 2
2 2
ds 2 = g11 (dx ) + g 22 (dx ) Since g12 = 0.

and

g = g ij =

g11 g12
g
0
= 11
= g11 g 22
g 21 g 22
0 g 22

## (i) To prove R12 = 0

hk
h1
R12 = g Rh12 k = g Rh 2 as Rh 122 = 0

R12

= g 21 R2121 as R1121 = 0
=0

## (ii) To prove R11 g 22 = R22 g11 = R1221

R11 = g hk Rh 11k = g 2 k R 211k
R 2112
22
R11 = g R 2112 = g
22
So,
R11 =
and

R 2112
g 22

11
= g R1221 =

So,

...(1)

R22 =

R1221
g11

R1221
g11

...(2)

## from (1) and (2)

R 11g22 = R 1221=R 22g11
(iii) To prove
R=

2R1221
g11 g 22

ij
i1
i2
R = g Rij = g Ri 1 + g Ri 2

(3)

108

## = g 11 R11 + g 22 R22 for g 12 = 0

=

R 11 R 22
+
g 11 g 22

R1221
R1221
[Q by eqn. (3)]
= g g +g g
22 11
11 22
2R1221
R=
g11 g 22
(iv) To prove Rij =

1
Rg
2 ij
R=

2R1221
g11 g 22

R=

2R1221
as g = g11 g 22
g

R1221 =

1
Rg
2

## The eqn (3) expressed as

R11 g 22 =

1
Rg = R 22 g11 .
2

it becomes
R11 =

Rg
Rg g
Rg
= 11 22 = 11
2 g 22
2g 22
2

Rg
Rg11 g 22 Rg22
=
=
R22 =
2 g11
2g11
2
So,
1
1
Rg11 & R22 = g 22
2
2
1
= g12 as R12 = 0 = g12
2

R11 =
R12
This prove that Rij =

1
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 d 2 + r 2 sin 2 d 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 2 + r 2 sin 2 d 2
Since r is radius of curvature then r is constant.

Riemann-Christoffel Tensor

g11 = r 2 ,

109

g 22 = r 2 sin 2 , g12 = 0,

g = g11 g 22 = r 4 sin 2 .

We can prove
R 1221 = r2 sin 2
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
K=

R 1212
R
= 1212
g11 g 22
g

K=

r 2 sin 2 1
=
= constant.
r 4 sin 2 r 2

EXERCISES
1. Show that
Rijkl =

[ jl, i] [ jk , i]

+
[
il
,

[ik , ]
xk
xl
j k
j l

2. Show that
2
g jk 2 g ik
g jl
1 g il
+

2 x j xk xi xl x j xl xi xk

## 3. Using the formula of the problem 2. Show that

2

Rijkl =

+ g ( [ jk , ][il , ] [ jl , ][ik , ] ) .

## 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
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
(b)

(i) Rhj =

1
Rhiij
g ii

110

## Tensors and Their Applications

(ii) Rhh =

1
1
Rhiih +
Rhjjh
g ii
g jj

7. Prove that if
1 R

Ri = g Rij then Ri , = 2 i
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
K=

R
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

KRNECKER 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 skewsymmetric 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

## an even permutation of number1, 2, ..., n

= 1; when i1 , i2 ,..., in

## = 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,
e12 = 1,

## indices are same

since i j has even permutation of 12

e 21 = e12 = 1

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 KRNECKER DELTA
A symbol i1j1...... ikjk depending on k superscripts and k subscripts each of which take values from 1 to n,
is called a generalised Krnecker delta provided that
(a) it is completely skew-symmetric in superscripts and subscripts

## The e-Systems and the Generalized Krnecker 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.

i1 ... i k
j ... j = 1; where odd number of transpositions arrange the superscripts
1
k

## = 0, in all other cases the value of the symbolis zero

EXAMPLE 3
Find the values of ijkl .
Solution
By definition of generalised Kronecker Delta, ijkl = 0 if i = j or k = l or if the set. ij is not the set kl.
22
23
11
pq = pq = 13 = = 0

i.e.,

= 21
= 13
= 31
= 23
= =1
12
12
21
13
31
23

i.e.,

## and ijkl = 1 if kl is an odd permutation of ij.

31
21
= 13
12
= 13
21
31 = 12 = = 1

i.e.,

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

i i ... i n

ej

1 j2 ... jn

## (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
e i 1i 2 ... i n e j1 j 2 ...
THEOREM 6.2 To prove that
(i) e i 1 i 2 ... i n = 1j1 2j 2 ... nj n
i i ... i

i i ... i
(ii) ei1i 2 ... i n = 1j12j 2 ...n jn

jn

i i ... i
= 1j1 2j 2 ... nj n

114

## 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 krnecker delta, we can write
i i ... i
i i ... i
(1) e 1 2 n = 11 22... nn

and
1 2 ... n
(2) ei1i2 ... in = i1i 2 ... in

## 6.5 CONTRACTION OF ijk

jk
Let us contract i
on k and . For n = 3, the result is
ij1
ij 2
ij 3
ij
ijk
= 1 + 2 + 3 =

## This expression vanishes if i and j are equal or if and are equal.

If i = 1, and j = 2, we get 123
3.
Hence

12

+ 1; if is an even permutatio n of 12

1; if is an odd permutatio n of 12
=
0; if is not permutatio n of 12

Similarly results hold for all values of and selected from the set of numbers 1, 2, 3.
Hence
+ 1; if i j is an even permutatio n of

1; if i j is an odd permutatio n of

ij
= 0; if two of the subscripts or superscripts are equal or when the

subscripts and superscripts are not formed from the same numbers.

1
If we contract ij . To contract ij first contract it and the multiply the result by . We obtain
2
a system depending on two indices
1 ij 1 i 1
i2
i3
i = 2 j = 2 ( 1 + 2 + 3 )

## The e-Systems and the Generalized Krnecker Deltas

1 12
1
13
It i = 1 in i then we get = 2 + 3
2

115

## This vanishes unless = 1 and if = 1 then 11 = 1.

Similar result can be obtained by setting i = 2 or i = 3. Thus i has the values.
(i) 0 if i , (, i = 1, 2, 3, )
(ii) 1 if i = .
By counting the number of terms appearing in the sums. In general we have
1 ij
ij
i = n 1 j and ij = n(n 1)
We can also deduce that
(n k )! i1i2 ... ir i r 1... i k
i1j1i2j...2 ...i r jr = (n r )! j1 j2 ... j r j r 1... j k

...(1)

...(2)

and
i1j1i2j...2 ...i r jr = n (n 1) (n 2) (n r + 1) =

n!
n r!

...(3)

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

(a) j

12 i j

(c) ij x y

ij 2 1

ij i j
ij
(c) e a a = e a .

(b) ij x x

ij

(d) ij

2. Expand for n = 2
ij 1 2

(a) e ai a j

(b) e ai a j

ijk
3. Show that ijk = 3 ! if i, j, k = 1, 2, 3.

i ... i

## 1j1... jk Ai1 ... ik = k! A j1 ... j k

k

5. Prove that ijk g is a covariant tensor of rank three where where 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
i
...(1)
C : x i = x (t ), (i = 1,2..., n)
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
1 2
dx n
n dx dx
Let F x , x ,..., x , dt dt ,..., dt be a continuous function in the interval t1 t t 2 . Wee

i
dx
assume that F x, > 0, unless every dx = 0 and that for every positive number k
dt
dt
1
2
1
2
n

1 2
dx
dx
dx n
x , x ,..., x n , dx , dx ,... dx .
F x1 , x 2 ,..., x n , k
,k
,..., k
kF
=

dt
dt
dt
dt dt
dt

The integral
dx
F x, dt
t1
dt
is called the length of C and the space R is said to be metrized by equation (2).
s=

t2

...(2)

dx
Different choices of functions F x, lead to different metric geometrices.
dt
If one chooses to define the length of arc by the formula
dx p dx q
dt, (p,q = 1, 2, ..., n)
...(3)
t1
dt dt
dx p
dx p dxq
, then the resulting
where g pq ( x)
is a positive definite quadratic form in the variable
dt dt
dt
geometry is the Riemannian geometry and space R metrized in this way is the Riemannian n-dimensional
space R n.
s=

t2

g pq ( x)

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)

ds 2 = dx i dx i

...(5)

## can be reduced to the form

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
t2
dx
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 = (s ) and we denote x i [(s )] by i () . so that x i (t ) i (s ) we have

equality

t2
t1

dx
F x , dt =
dt

s2
s1

F (, ) ds

ds i
and t1 = (s1 ) and t2 = (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
i
where =

C : x i (ks) = i (t )
and

or

dx i (ks )
dx i (ks)
=k
ds
dt
t2
dx
s=
F x , dt we get
t1
dt

i (s ) =

s=

s=

s2
s1
s2
s1

dx (ks )

F x (ks),
kds
dt

## F [(s ),( s )]ds

dx

dx
We must have the relation F (, ) = 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 = ( s), 1 (s ) > 0, s1 s s 2 with and t2 = (s 2 ).
i

Note: (i)
(ii)

dx
Here take those curves for which x i (t ) and
are continuous functions in t1 t t 2.
dt

A function F x,

dx
dx

dx
satisfying the condition F x, k = kF x, for every k > 0 is called
dt
dt

dt
i

## positively homogeneous of degree 1 in the

dx
.
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
O
P1
1

P2
C

Q2
K

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
x1 0
x 0 a 0
or
x1 = x 1 , x 2 = x 2 , x 3 = a
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

( ) ( )

2 x 1 + x 2

or

+ a2 = a2

...(6)

Geometry

119

## Solving for and substituting in equation (6), we get

ax1

x =
1

and

( x 1 ) 2 + ( x 2 )2 + a 2

ax 2

, x2 =

( x 1 )2 + (x 2 )2 + (x 3 ) 2

a2

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

ds 2 =

xi xi
d x pd x q
p
q
x x

= g pq ( x )d x pd x q , (p,q = 1, 2)
xi x i .
x p x q
If the image K of C on T is given by the equations
where g pq (x ) are functions of x i and g pq =

x 1 = x 1 ( t )
K:
x 2 = x 2 (t ), t1 t t2

s=

t2
t1

g pq

dx p dxq
dt
dt dt

## A straight forward calculation gives

1 1 2
(x d x x 2 d x 1)2
a2
2
1

1 2
2 2
1
+
(
x
)
+
(
x
)
a 2

(d x 1)2 + (d x 2) 2 +
ds2 =
and

...(9)

120

## Tensors and Their Applications

2

2
1
d x1 d x 2

+
+ 12 x1 d x x 2 d x

t 2 dt
dt
dt
a
dt
dt
s= t
1
1 2
2 2
1
1 + 2 (x ) + ( x )
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

1 2
2 2
...(10)
ds 2 = (d x ) + (d x ) .
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
i 1
2
3
T 1 : x i = x ( x , x , x ), (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
...(1)
x1 ( x 1 , x 2 , x 3 ) = constant
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
2
3
x1 = c1 , x = c2 , x = c3

3
Y

(2)

3
2
X

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

X2

O
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)

tan x 2 =

or

(x1 )2 + (x 2 )2
x3

1 2
2 2

1 ( x ) + ( x )
tan
x =

x3

...(8)

## Divide (3) and (4), we get

x2
tan x 3
1 =
x

2
1 x
x 3 = tan 1
x

...(9)

So, the inverse transformation is given by the equations (6), (8) and (9).
If x1 > 0, 0 < x 2 < , 0 x 3 < 2. 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
...(1)
r = bi x i (i = 1, 2, 3)

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
j
ds 2 = bi dx b j dx
r r
= bi b j d x id x j
r r
i
j
ds 2 = ij d x d x ; since bi b j = ij

3
Y

X
a3

3
2
X

a2

a1

1
X

P
b3

r
O

b2

b1

2
Y

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) .
ds 2 = d x id x i;

as ij = 1, i = j

= 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
g ij

r
r
r r

=
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
i
x
Then from (3) and (4)
r r
r
r
dr = ai d x i and g ij = a i a j

...(4)

...(5)

...(6)

## Now, from equations (2) and (6), we get

r
r
a j dx j = bi d x i
r x i j
r
dx
a j dx j = bi
x j

r xi
r
, as d x j are arbitrary
a j = bi
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.
If the curvilinear coordinate system X is orthogonal. Then
r r
r r
g ij = a i a j = ai a j cos ij = 0, if i j.
r
r
Any vector A are can be written in the form A = k d rr where k is a scalar..

...(8)

...(9)

Geometry

125

r
r r i
d
r
=
d x we have
Since
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
a 2 a3 ar 2 = a3 a1 , ar 3 = a1 a2
r1
...(10)
a = rr r ,
[ar1ar2 ar3 ]
[ar1ar2ar3 ]
[a1a 2a3 ]
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
a2 a3 a2 [a2 a2 a3 ]
r1 r
=
= 0.
a a 2 = [ar ar ar ]
[ar1ar2 ar3 ]
1 2 3

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 = ij

EXAMPLE 3
rr r
To show that [a1 a2 a3 ] =
Solution

g

## 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
0 0 a3

...(11)

126

and

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
g= 0
0

a22

## 0 = a12 a22 a32

a32

g = a1 a2 a3
from eqn. (11) and (12), we have
[ar1ar2ar3 ] = g

...(12)

r r r
1
Since the triple products [a 1 a 2 a 3 ] =
. Moreover,,
g
r r
r r
r r
a2 a3
r
a 3 a1 r
a1 a 2
r
,
a
=
,
a
=
a1 = r1 r 2 r 3
r r r
r r r
2
3
[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

where

ij
ds 2 = g dxi dx j
r r
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 ij = A j as a i a j = 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,j a
j
x
i
r
A
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
=
i
x j
x j
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 j
a i r
a j + k
=
x k
x k
x
Similarly,
r
r
g jk
a j r
ak

a
+
=
k
x i
xri
xri
g ik
ai r
a
ak + kj
and
j =
j
x
x
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

...(1)

r
ai
r
aj
r
ai

g ij A i = ij Ai = A j

r r
r r
As ai a j = g ij , a i a j = ij and 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 i g j = ij .
Solution
Since we know that
Then

g i = ari ar and g j = ar j ar
r r r r
g i g j = (ai a ) (a j a )

r r
r r
= (ai a j ) (a a )

r r
= ij as ai a j = ij
g i g j = ij as = 1.
But

r
r
r
ai =
x i
r
r
r
r
a j
ai
2r
2r
=
=
=
x j
x j dx i x i x j xi

So,

r
r
a j
ai
=
x j
x i

Now,
[ij, k ] =
Substituting the value of

or
Hence

1 g ik g jk g ij

+
k
2 x j
x i
x

g
g ik g jk
, i and ijk , we get
j
x
x
x
r
1 a i r
[ij,k] = 2 j a k
2 x
r
ai r
ak
[ij,k] =
x j
r
ai
rk
1 rk
j = [ ij, k ]a , as r = a
x
ak

r
a i r
r r
a = [ij, k ]a k a
x j

, Christoffels symbol

Geometry

129

r r
= [ij, k ]g k , Since g k = a k a

r

ai r
[ i j , k ] g k =
,
.
a
=
as

i j
x j
i j
r
r
ai
= a
i j
x j
r
ai
Substituting the values of
in equation (1), we get
x j

...(2)

r
Ai r i r
A
ai + A a
=
x j
x j
i j

A r
r
a + A i a
=
j
x
i j
r
A i r
A
=
j + A a
x j
i j
x
r
A
A i
r

=
A
,
a
since
A
=
+ A
j

,j
x j
x j i j
Thus, the covariant derivative A,j of the vector A is a vector whose components are the
r
A
r
components of
j referred to the base system ai .
x
r

ai
=
0
If the Christoffel symbols vanish identically i.e.,
the
= 0, from (2).
x j
i j
Substituting this value in equation (1), we get
r
A
Ai r
ai
=
x j
x j
But

Ai i
+
A
x j j
i
Ai

= 0.
A,i j =
as
x j
j
A,i 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
=
x k
x k
x
r r
Since a i a j = ij , we have,

...(1)

## Differentiating it partially w.r. to x k , we get

r
r
a i r
r i a j

a
+
a

=0
j
x k
x k
r
r
r i a j
a i r
a j = a k
x
x k

r r
But a i a = i . Then

r r
= ai a
,
j k
r
a i
x k
r
a i
x k

Since

r
a j

r
= a
x
j k
k

r

a j = i

j k
r
i
a j =
j k
r
i r
a i
a j ,
=
x k
j k

1 rj
as ar = a
j

r
a i
substituting the value of
in equation (1), we get
x k
r
Ai ri
i r j
A
a Ai
a
k = x k
x
j k
=

Ai r j
i rj
a Ai
a
k
x
j k

r
Ai
i r
A
Ai a j
k =
k
x
j k
x
r
A j
i
A
rj
A
=
Ai

=
A
a
,
Since
j,
k
k
k
j, k
x
x
j k
Proved.

Geometry

131

## 7.5 INTRINSIC DIFFERENTIATION

r
Let a vector field A(x ) and
i
C : x i = x (t ), t1 t t 2
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
j
dA
r dx
= A, j a
dt
dt
Since we know

r
A r
r
A
A
,
a
=
j + a
j

=
x
i j
x j

## (See Pg. 127, Theo. 7.2)

So,
r
A i r dx j
dA
= j + i j A a dt
dt

x
r
dA i dx j r
dA
+ A
=
a
dt
dt
i j
dt
dA i dx j
+ A
is called the absolute or Intrinsic derivative of A with respect
dt
i
j
dt

A
.
to parameter t and denoted by
t
The formula

A dA i dx j
A A
=
.
So, t = dt + i j A dt is contravariant vector. If A is a scalar then, obviously,
t
t

Some Results
(i) If Ai be covariant vector
dAi dx
Ai
= dt i A dt
t

(ii)
(iii)

A ij Aij i j dx j i dx
=
+
+
A
A
t
dt
dt
dt
A ij
t

A ij

i dx i dx
+
A
A
dt j dt i dt

132

## Tensors and Their Applications

(iv)

A ijk
t

A ijk

i dx i dx i dx
+

A jk
A
A
dt
dt j k dt k j dt

EXAMPLE 5
If g ij be components of metric tensor, show that

g ij
t

= 0.

Solution
The intrinsic derivative of g ij is
g ij
t

dx
dx
g j

g i
dt i
dt j
dt

dg ij

g ij dx
dx
dx

j
i

dt j
dt
x dt i

g ij
dx

g i

j
= x
i

j dt

g ij
t
as

g ij
dx
= [i, j ] [j, i]
x
dt

gj = [ i , j ] and
g i = [ j , i ] .
j
i
g ij

But

x
g ij

So,

= [i, j ] + [ j, i].
= 0.

EXAMPLE 6
Prove that
d ( g ij A i A j )
dt

= 2 g ij A i

A i
t

Solution
Since g ij Ai a j is scalar..
Then
d ( g ij A i A j )
dt

(g ij A i A j )
t

Geometry

133

( A i A j )
, since g ij is independent of t.
t
A i j
A j
A + Ai
= g ij

t
t
Interchange i and j in first term, we get
= g ij

d ( g ij A i A j )
dt
d ( g ij A i A j )
dt

i A j
A j
+ Ai
= g ij A
as g ij is symmetric.
t
t

= 2 g ij Ai

A j
t
Proved.

EXAMPLE 7
Prove that if A is the magnitude of Ai then
A, j =

Ai , j Ai
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
A, j =

Ai , j Ai
A

Proved.

134

## 7.6 PARALLEL VECTOR FIELDS

Consider a curve
i
C : x i = x (t ), t1 t t 2 , (i = 1, 2, 3)
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
X

3
Y

2
X
C
1
X

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
r
dA
= 0. It follows that the components Ai of A satisfy a set of simultaneous differential equations
dt
A i
= 0 or
t
dAi i dx
+
=0
A
dt
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
and the length of arc s of C is defined by the integral

s=

t2
t1

gij

dxi dx j
dt
dt dt

...(1)

...(2)

## from (1), we have

g ij

dx i dx j
=1
ds ds

...(3)

Geometry

Put

135

dx i
= i . Then equation (3) becomes
ds
g ij i j = 1

...(4)
r
r
The vector , with components i , is a unit vector. Moreover,, is tangent to C, since its
dx i
components i , when the curve C is referred to a rectangular Cartesian coordinate Y,, becomes i =
.
ds
These are precisely the direction cosines
tangent vector to the curve C.
r of the
r
Consider
a
pair
of
unit
vectors
and

(with
components i and i respectively) at any point

r
P of C. Let is tangent to C at P Fig. (7.6).

+ d
P(x)

(Q(x + dx)

C
r + dr

O
Fig. 7.6

r
r
The cosine of the angle between and is given by the formula
i j
cos = g ij

r
r
and if and are orthogonal, then equation (5) becomes
g ij i j = 0

...(6)

i j
j i
+ gij
=0
s
s

(7)

r
Any vector 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
g ij

## as g ij is constant with respect to s.

Interchange indices i and j in second term of equation (7) we get
g ij

...(5)

i j
i j
+ gij
=0
s
s

2 g ij i

j
=0
s

136

## Tensors and Their Applications

g ij i

j
=0
s

j
either vanishes or is normal to C and if does not vanish
s
j
we denote the unit vector co-directional with
by j and write
s
we see that the vector

j =

j
1 j K =
,
s
K s

...(8)

## where K > 0 is so chosen as to make j a unit vector..

The vector 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 and the principal normal vector is called the
osculating plane to the curve C at P.
r
Since is unit vector
g ij i j = 1

...(9)

## Also, differentiating intrinsically with respect to s to equation (6), we get

g ij

j j
j
+ g ij i
=0
s
s

or
g ij i

j
i j

= gij
s
s
= Kg ij i j Since

g ij i
g ij i

i
= K i
s

j
= K , since g ij i j = 1.
s

j
+K =0
s

j
+ g ij i j K = 0 as g ij i j = 1
s
j

g ij i
+ K j = 0
s

g ij i

## This shows that the vector

j
+ K j is orthogonal to i .
s

Geometry

137

r
Now, we define a unit vector v , with components v j , by the formula

1 j
j
v i = s + K

...(10)

r
r
i
r
+ Ki the vector v will be orthogonal to both and .
s
To choose the sign of in such a way that
where =

g eijk i j v k = 1
r r
so that the triad of unit vectors , and r forms a right handed system of axes.
Since eijk

xi
g
=
is a relative tensor of weight 1 and
xj

...(11)

## it follows that ijk = g eijk is an

obsolute tensor and hence left hand side of equation (11) is an invariant v k in equation (11) is determined
by the formula
ijk
v k = i j

...(12)

1 ijk

## where i and j are the associated vectors g i and g i and ijk =

e is an absolute tensor..
g
The number 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,i a
x i
r
if the vector field A is defined along C, we can write
r
i
A x i
x r
A
a
=
...(13)
,
i
s
x i s
Using definition of intrinsic derivative,
i
A
dx
= A,i
s
ds
Then equation (13) becomes
r
r
r
dA
A r
A dx i dA
a ; as
=
=
...(14)
ds
s
ds
x i ds
r
r
Let r be the position vector of the point P on C then the tangent vector is determined by
r
dr
r r
= i ai =
ds
from equation (14), we get
r
r
d r
r
d 2r
=
a = c
...(15)

2 = ds
s
ds

138

## Tensors and Their Applications

where

r
r
c is a vector perpendicular to .

With each point P of C we can associate a constant K , such that c K = is a unit vector..
Since

r
c
r
=
K
1 r
r
a , from (15)
=
K s

r
r
= a , since = 1
K s

## 7.8 SERRET-FRENET FORMULA

The serret-frenet formulas are given by

1 i
i
i
or
= K , K > 0 whereK =
K s
s
s
i

1
i
i
i
i
+
K

i
i
i
(ii) = s
= K where = s + K
or
s

(iii)
= k
s
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
i

(i)

i =

ijk i j = k
where i , i , k are mutually orthogonal.
Taking intrinsic derivative with respect to s, we get
j
i
k
j + ijk i
=
s
s
s
From formulas (i) and (ii), we get
ijk

ijk K i j + ijk i ( j K j ) =
ijk i ( j K j ) =
ijk i j K ijk j i =

k
s
k
,
s
k
s

Since ijk i j = 0

Geometry

139

ijk i j = 0

Since

k
s

ijk i j =

## Since ijk i j = k , but ijk are skew-symmetric.

Then
ijk i j = k
So,
k =

k
s

k
k
=
s

or

i
= i
s
This is the proof of third Serret-Frenet Formula
Expanded form of Serret-Frenet Formula.

(i)

(ii)

(iii)

di i j dx k
i
+

ds j k ds = K

or

d 2 xi i dx j dx k
+

= K i
ds 2 j k ds ds

di i j dx k
+
i
i
ds jk
ds = K
dv i i j dx k
i
+
ds jk
ds =

EXAMPLE 8
Consider a curve defined in cylindrical coordinates by equation
x1 = a
2
x = ( s )
x3 = 0

## This curve is a circle of radius a.

The square of the element of arc in cylindrical coordinates is
1 2
1 2
2 2
3 2
ds 2 = (dx ) + ( x ) (dx ) + (dx )

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

d
dx i
The components of the tangent vector to the circle C are i =
so that 1 = 0, 2 =
,
ds
ds
3 = 0.
Since is a unit vector, gijij= 1 at all points of C and this requires that

(x1 )2 dds

2 d
= a =1
ds

d
1
So, = 2 and by Serret-Frenet first formula (expanded form), we get
ds
a
1 2 dx 2
d1 1 j dx k
1
+

K = ds j k ds = 2 2 ds =
a

2 2 dx1
d2 2 j dx k
K2 = ds + j k ds = 2 1 ds = 0

d3 3 j dx k
K = ds + j k ds = 0

## Since is unit vector, g ij i j = 1 and it follows that K =

1
, 1 = 1, 2 = 0, 3 = 0
a

## Similarly we can shows that = 0 and 1 = 0, 2 = 0, 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
A i
=0
s
or

dAi i dx
+
A
ds
ds = 0

(1)

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 to a straight line is
r
directed along the straight line. So that the totality of the tangent vectors forms a parallel vector field.

Geometry

141

i
Thus the field of tangent vector =

dxi
must satisfy equation (1), we have
ds

d 2 xi i dx dx
i
+
=
= 0

s
ds 2 ds ds
The equation

d 2 xi i d x d x
+

## = 0 is the differential equation of the straight line.

ds 2 ds d s

EXERCISE
1. Show that

d (gij AiB j )
dt

2. Show that Ai , j A j , i =

= g ij

A i j
B i
B + g ij A i
t
t

Ai A j

x j xi

## 3. If Ai = g ij A j show that Ai , k = gi A,k

4. Show that

d ( gij Ai B j )
dxk

i
i
= Ai,k B + Bi,k A

5. Show that
2 i
dK i
K ( i Ki )
=
ds
ds 2
2i
s2
2i

d i
dK i
( K 2 + 2 )i

ds
ds
d

= ( K v )
ds
s2
i

6. Find the curvature and tension at any point of the circular helix C whose equations in cylindrical
coordinates are
C : x1 = a, x2 = , x 3 =
Show that the tangent vector at every point of C makes a constant angle with the direction of X 3 axis. Consider C also in the form y1 = a cos, y2 = a sin, y3 = . 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
d (mv )
r
(1)
F = dt

Analytical Mechanics

143

## If we postulate the invariance of mass then equation (1) can be written as

r
r
(2)
F = ma
r
d (mv )
r
from (1) if
= 0.
F = 0 then dt
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
dt
and if ai be the component of acceleration of moving particle then
vi =

ai

v i dv i i j d x k
= t = dt + j k v d t

ai

d 2 x i i dx j dx k
+
=
dt 2 j k dt dt

(2)

(3)

v i
i
is the intrinsic derivative and the are the Christoffel symbols calculated from the
t
jk
metric tensor gij
If m be the mass of particle. Then by Newtons second law of motion
where

Fi = m

v i
= ma i
t

(4)

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

W=

P2

P1

Fi dx i

(6)

W=

W=

m g ij

v i
dx j
t

m g ij

v i dx j
dt
t dt

P2

P1

P2

P1

P2

P1

v i j
v dt
t

m g ij

(7)

## Since g ij v i v j is an invariant then

(g ij vi v j )

d
( g v iv j )
dt ij

t
d
vi j
( g ij v i v j ) = 2 g ij
v
or
dt
t
vi j 1 d
(g ij v i v j )
g ij
v =

2 dt
t
using this result in equation (7), we get
W=

P2 m

P1

d
( g v i v j ) dt
2 dt ij

m
[g v i v j ]PP12
2 ij
Let T2 and T1 is kinetic energy at P 2 and P 1 respectively.
W = T2 T1
W=

m
m v2
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.
where T =

## 8.4 CONSERVATIVE FORCE FIELD

The force field F i is such that the integral W =

P2

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
.
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.
if F i =

V
x i

Now,
Fi k
Fk
x j i j
V
i
x k F
F i,j =
k
x j
i j
F i,j =

k
2V
Fk
j
i
x x j i

(1)

and
F j,i =

F j

k
Fk
x j i
i

Similarly,
F j,i =

2V
k
Fk
i
j
x x
ji

## From equation (1) and (2), we get

F i,j = F j,i
conversely, suppose that F i,j = F j,i
Then
F j k
Fi k
Fk =
Fk
j
x
x i j i
i j

F j k
Fi
as due to symmetry..
j =
x i i j
x

(2)

146

Take F i =

V
x i

Then
Fi
2V
j i
j =
x
x x
2V
= i j
x x
V

=
x i x j
F j
Fi
j =
x
x i
Fi =

## So, we can take

V
.
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).
The kinetic energy T =

Since x& i = v i.

1
mv 2 can be written as
2
1
i j
T = m g ij x& x&
2

1
m g jk x& j x& k
2
Differentiating it with respect to x& i , we get
or

T=

(1)

x& j k
&k
1
T
j x

&
&
m
g
x
+
x
jk
= 2
x& i
i
&

x
x& i

1
m g jk ij x& k + ki x& j
2

1
1
m g jk ij x& k + m g jk ki x& j
2
2

1
m gik x& k + g ji x& j
2
1
j
j
= m (gij x& + gij x& )
2

)
as

gij = g ji

Analytical Mechanics

147

T
= m g ij x& j
x& i
T
= m g ik x& k
x& i

or

(2)

## Differentiating equation (2) with respect to t, we get

d T
d
g x& k

= m
dt x& i
dt ik

d
k
k
= m g ik x& + g ik &x&
dt

g ik dx j k

k
m
= x j dt x& + g ik &x&

d T
g
j k
k
= m ikj x& x& + m g ik &x&
dt x& i
x
1
m g jk x& j x&k
2
Differentiating it with respect to xi, we get
T
1 g jk j k
m
x& x&
i =
x
2 x i
Now,

(3)

Since T =

d T

dt x& i

(4)

T
g
1 g
i = m ikj x& j x&k + m g ik&x&k m jki x& j x& k
x
2 x
x
k
= m g ik &x& +

1 gik j k 1
1 g
m j x& x& + m g ik x& j x& k m jki x& j x& k
2 x
2
2 x

k
= m g ik &x& +

1 g ik j k 1 g ij k j 1 g jk j k
m
x& x& + m k x& x& m i x& x&
2 x j
2 x
2 x

1 gik g ij g jk j k
k
x& x&
= m g ik &x& + 2 m j + k
x
x i
x
= 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

## = m g il x&&l + g il [ jk , i ]x& j x& k

l
= m g il &x&l + x& j x& k ,
jk

d T T

= m g il a l ,
dt x& i x i
where al is component of acceleration
or

l
as g il [ jk , i] =
j k

l j k
l
l
Since a = x&& + j k x& x&

d T T

= 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.
For a conservative system, F i =
d T

dt x& i

V
. Then equation (5) becomes
x i

V
T
i = i
x
x

d T (T V )

=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=TV
Then equation (6) becomes
or

d L L

=0
dt x& i x i

(7)

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

and

a2 =

d
( x1 )2 x& 2 (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2
dt

a3 =

d
(x1 sin x 2 ) 2 x& 3
dt

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
3
ds dx1
1 2 dx
1
2 2 dx
2
=
+
(
x
)
+
(
x
sin
x
)

v =
dt
dt
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
mv
2
1
1 2
1 2
2 2
1
2 2
3 2
T = m (x& ) + ( x ) (x& ) + ( x sin x ) (x& )
2
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,
T=

d T T

= m ai
dt x& i x i

(2)

Take i = 1,
m a1 =

d T T

dt x&1 x 1

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

## from (1), we get

a1 =

dx&1
x1 ( x& 2 )2 + x1 (sin x 2 ) 2 ( x& 3 )2
dt

## a1 = &x&1 x1 (x& 2 )2 x1 (x& 3 sin x 2 ) 2

Take i = 2,
m a2 =

d T

dt x& 2

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 =

(1)

T
2
x

d
( x1 )2 x& 2 (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2
dt

150

## Tensors and Their Applications

Take i = 3
m a3 =

d T T

dt x&3 x 3

m a3 =

1 d
m 2( x& 3 ) ( x1 sin x 2 )2 0
2 dt

a3 =

d 3 1 2
x& ( x ) (sin x 2 ) 2
dt

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
mv , T is kinetic energy..
2
1
i 2
T = m( y& )
2
The Lagrangean equation of motion is
T=

d T T

= Fi
dt y& i yi
Since particle is not subjected to the action of forces.
So, F i = 0
Then

d 1

m 2 y& i 0 = 0
dt 2

dy& i
=0
dt

Analytical Mechanics

151

dy& i
=0
dt

or

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
vi

dx i
=
dt

or

v i = x& i

## given |v| = constant.

Since
g ij vi v j = | v |2 = constant
Taking intrinsic derivative with respect to t, we get

(g v i v j ) = 0
t ij
v i j
v j
g ij
v + vi
t
t

=0

g ij

v i j
v j
v + g ij v i
=0
t
t

g ij

v i j
v i
v + g ji v j
= 0, (Interchange dummy index i and j in second term)
t
t
2 g ij
g ij

v i j
v = 0 as gij = gji
t
vi j
v =0
t

## This shows that acceleration vector

v i
v i
= 0.
is either orthogonal to v i or zero i.e.,
t
t

152

## 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
dt x& i x i

(1)

1
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
If xi be rectangular coordinate system, then T =

ds2 = dr 2 + r 2 d 2 + r 2 sin 2 d 2
If T be the kinetic energy, then
T=

1 2 1
mv = m (r&2 + r 2 & 2 + r 2 sin 2 & 2 )
2
2

Y2
r

R
mg cos

Y1

mg sin
Y3

P
mg

Fig. 8.1.

## from Lagrangean equation of motion

d T T

= Fi
dt x& i x i
x1 = r , x 2 = , x 3 = .
So, take

x1 = r

i = 1, 2, 3

(1)

Analytical Mechanics

153

d T T

= mg cos R
dt r& r
from (1), we have
R
&r& r&2 r sin 2 &2 = g cos
m

(2)

Take x2 = , we have
r && + 2r&& r sin cos & 2 = g sin

(3)

## and take x3 = , we have

d 2& 2
(r sin ) = 0
dt

(4)

If the motion is in one plane, we obtain from equations (2), (3), and (4), by taking & = 0.
R
r&& r & 2 = g cos m
r && + 2r&& = g sin
g
If r& = 0, we get, && = sin which is equation of simple pendulum supported by an
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 HAMILTONS 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
t1

(T + Fi xi ) dt = 0

where xi = xi (t) are the coordinates of the particle along the trajectory and xi + x 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
t1 t t 2
C : x i = x (t ),
At time t, let particle is at P(xi). If T is kinetic energy. Then
i

T=

1
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 (, t ) = x i (t ) + x i (t)

154

## Tensors and Their Applications

At t1 and t2
xi = x i = x i + x i

x i (t1 ) = 0 and x i (t 2 ) = 0

But T = T ( x i , x& i ).
If T be small variation in T.. Then
T i T i
T = x&i x& + x i x
Now,

{(T + F )x }dt =
t2

t1

t1

T i
x dt +
x i

t2
t1

t2
t1

T i
x& dt +
x&i

t2
t1

T i
x dt +
x i

T i 2
x& i x

t1

t2
t1

T i 2
then & i x = 0.
x
t1
So,

(T + F x )dt =

(T + F x )dt =
t1

t1

+
t2

t2

t1

t2
t1

x dt

t2
1

d
dt

T
i
x&

i
x dt

Fi x i dt

t2 T

t1

i
d T
i i + Fi x dt
x dt x&

## since particle satisfies the Lagrangean equation of motion. Then

d T T

= Fi
dt x& i x i
or

t1

Fi x i dt

d T i

x dt +
dt x&i

Since x i (t1 ) = 0, x i (t 2 ) = 0.

t2

t2

T
as 1st term
x& i

## Integrating second term by taking

T
T

i x i + i x& i + Fi x i dt
x&
x

t2

T d T

= Fi
x i dt x& i

t2
t1

Fi xi dt

Analytical Mechanics

155

So,

t2
t1
t2

t1

(T + Fi x i ) dt =

t2
t1

[Fi + Fi ] xi dt

(T + Fi x i ) dt = 0

Proved.

## 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

or

T=

1
m g ij x& i x& j
2

T=

1
m g ij v i v j
2

As T is invariant. Then
Taking intrinsic derivative with respect to t, we get
dT
T
=
dt
t
=

m gij v i v j
t 2

v i j
1
v j
m g ij
v + vi
2
t
t

1 v i j
v j i
m gij
v + g ij
v
2
t
t

1 vi j
vi j
= 2 m gij t v + g ji t v , Since i and j are dummy indices.

1
v i j
m
2
g
v
=
ij
2
t
dT
v i j
v
= m gij
dt
t
or

dT
v j i
m
g
v
=
ij
dt
t

as g ij = g ji

156

## Tensors and Their Applications

j i
= m g ij a v as

dT
= m ai v i,
dt

v j
= aj
t

since gij aj = ai

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
, where V is potential energy..
x i

Fi =
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
A=

P2

mv.ds

(1)

p1

## evaluated over the path

i
C : x i = x (t ), t1 t t 2
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

A=

P2
p1

m g ij

t ( P2 )
t ( P1 )

dx i
dx j
dt

m gij

dx i dx j
dt
dt dt

Analytical Mechanics

Since T =

157

1
dx i dx j
m g ij
, we have
2
dt dt
A=

t ( P2 )

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, , where and 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 ) , (i = 1, 2,3), ( = 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 ) 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 ) = x(i ) (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 ) =

x i( )
q j

q& j +

x i( )
t

(3)

158

## 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& +
(5)
t = 0
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
(6)
F j (q1 , q 2 ,...,q k ; q&1 ,...,q& k , t ) = 0, ( j = 1, 2, 3, ..., m)
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 nonintegrable 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
(r = 1, 2, 3, )
xr = xr (q , q ,...,q , t );
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& +
,
t ( j = 1, 2, ..., n)
q j

## 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
q& ,
=
x&
q j
r

(r = 1, 2, 3; j = 1, 2, ..., n)

(2)

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
T=

1
m g rs x& r x& s ,
2

(r,s = 1,2,3,)

(4)

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
q& q&
T = 2 m g rs i
q q j
T=

1
a q&i q& j
2 ij

aij = m g rs

where

x r r s
,
q i q j

(5)
(r, s = 1, 2, 3), (i, j = 1, ..., n)

1
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 T =

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
dt q& i

T
l l j k

## qi = a il &q& + j k q& q&

(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
dt q& i

T
l

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
&
=
,
=
q
=
Now, from the realtions & j
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
dt q& i

T
x r

m
a
=
r
qi
q i

## in which a j = g ij a i is acceleration of the point P.

Also, Newton's second law gives
m ar = F r

(8)

(9)

Fr's

where
are the components of force F acting on the particle located at the point P..
From the equation (9), we have

m ar

x r
=
q i

Fr

x r
q i

Fr

x r
q i

d T
dt q& i

qi =

(10)

Qi =

Fr

x r
q i

The equations
d 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
dt q& i

qi = 0

## where L = T V is the kinetic potential.

(12)

Analytical Mechanics

161

Since L (q, q& ) is a function of both the generalized coordinates and velocities.
L i L i
dL
= & i q&& + i q&
dt
q
q
from (12), we have

L
d L
i =
q
dt q& i

(13)

L i d L
dL
q&& + i
=
dt
dt q&
q& i
=

i
q&

d L i
q&
dt q& i

(14)

L i
T i
q& =
q& = 2T
i
&
q
q& i
since

T=

1
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 n ds
(1)

## where n 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
F 1 F 2 F 3
r
+
+
div F =
x 1 x 2 x 3

(2)

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 ,ij =

F i i k
+ F
x j k j

r
The invariant F, ij in cartesian coordinates represents the divergence of the vector field F .
Also,
r
i j
i
j
F n = g ij F n = F ni since g ij n = ni
Hence we can rewrite equation (1) in the form

F dV
i
,i

F n dS
i

(3)

## (ii) Symmetrical form of Green's Theorem

Let ( x1 , x 2 , x3 ) and (x1 , x 2 , x3 ) be two scalar function in V.. Let i and i be the gradients of
and respectively, so that

and
xi
Put F i = i and from the divergence of we get
= i =

= i =

x i

F,i j = g ij Fi , j = g ij (i , j + i j )
Substituting this in equation (3), we get

ij

(i , j + i j )dV

n dS
i

(4)

Since = i , then
g ij i , j = 2

(5)

## Also, the inner product g ij i j can be written as

g ij i j = .
where denote the gradient and 2 denote the Laplacian operator..
Hence the formula (4) can be written in the form

(g

ij

i , j + g ij i . J ) dV

n dS
S

( + ) dV = n dS
2

dV = n dV
2

V
i
where n = i n =

.
n

(6)

Analytical Mechanics

163

dV
2

n dV
S

(7)

( ) dV
2

n n dS

(8)

## This result is called a symmetric form of Green's theorem.

(iii ) Expansion form of the Laplacian Operator
The Laplacian of is given by
ij
2 = g i , j from (5)
when written in the terms of the christoffel symbols associated with the curvilinear coordinates
i
x covering E 3,

2
k
ij
2 = g i j k
i j x
x x
i
and the divergence of the vector F is
F,ii =

F i i
+ F
x i j i

(9)

(10)

log g
=
j
i

xj

The equation (10) becomes
But we know that

F,ii

or
ij
If putting F i = g

F,ii

F i

+ j log g F j
=
i
x
x

1 ( g F i )
x i
g

x j

g g ij j
1
x
g ij j , i =
i
x
g

ij
2 = g j, i

(11)

(12)

164

## Hence equation (12) becomes

g g ij j
1
x
2 = g ij j ,i =
i
g
x
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
n. curl F ds = F . ds
(13)

r
where is the unit tangent vector to C and curl F is the vector whose components in orthogonal
cartesian coordinates are determined from
e1

r
curl F = x1
F1

e2

x 2
F2

e3

r
x 3 = F
F3

(14)

## 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 = ijk F j , k
(15)
r
we define the vector G to be the curl of F .
dx i
r
i
ijk
Since n . curl F = ni G = F j ,k ni and the components of the unit tangent vector and
.
ds
Then equation (13) may be written as

ijk F j , k ni ds
S

The integral

F dx
i

C
i

Fi

dx i
ds
ds

(16)

r
is called the circulation of F along the contour 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 m 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 be the angle between the
unit outward normal to n 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
F .n dS =

m cos r 2 dw
r 2 cos

r dw
and dw is the solid angle subtended by dS.
cos
Thus, we have,
r
F .n dS
m dw = 4 m
=

where dS =

(1)

dS

r
d
P
m
S

Fig. 8.2.

r
F .n =

i =1

mi cos i
ri 2

## and total flux is

r
F .n dS = 4

(2)

i =1

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 dV contained within V, is
cos dV
r
dS
F .n dS =
r2
S

r
F .n dS =

cos dV

r2
V

dS

(3)

166

## Tensors and Their Applications

where

denotes the volume integral over all bodies interior to S. Since all masses are assumed to be

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
F .n dS =

cos dS
dV
s
r2

(4)

cos dS
= 4. Since it represents the flux due to a unit mass contained within S.
r2
S
Hence
But

F.n dS

4 dV = 4 m

(5)

## 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 .n )i dS = 4 m

S"

r
where the subscript i on F n 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 n)o ds = 0

S'
r
where the subscript o on F n 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 n)o dS is zero.

S'

## 8.14 POISSON'S EQUATION

By divergence theorem, we have

r
F .n ds =

## and by Gauss's Theorem,

r
div F dV

F.n ds = 4 dV
from these, we have

Analytical Mechanics

167

r
(div F 4 ) dV = 0

Since this relation is true for an arbitrary V and the integrand is piecewise continuous, then
r
div F = 4
By the definition of potential function V, we have
r
F = V
and

div V = 2V

So,

r
div F = 4
div ( V ) = 4
2V = 4

## which is equation of poisson.

If the point P is not occupied by the mass, then = 0. Hence at all points of space free of matter
the potential function V satisfies Laplace's equation
2V = 0
8.15 SOLUTION OF POISSONS EQUATION
We find the solution of Poissons Equation by using Greens symmetrical formula. We know that
Green's symmetrical formula

( 2 2 ) dV = n n dS
(1)
S

## where V is volume enclosed by S and and are scalar point functions.

1
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.
Put =

Q (y )

r
P(x)
n

Fig. 8.3.

168

## Tensors and Their Applications

1
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 and volume V'. Apply Greens symmetrical formula to the
1
region V V' within which
and V possess the desired properties of continuity..
r
2
2 1
= 0.
In region V V', =
r
Then equation (1) becomes
Since

1 2
dV =
r
V V'

1
1r

dS +

r n
n

S'

1
1r

ds

r n
n

(2)

where n is the unit outward normal to the surface S + S' bounding V V' . S' being the surface of the

= .
sphere of radius and
n
r
Now

S'

1
1r

dS =

r n

S'

1
1r

r r
r

dS

2
r dw
2

r r r
S'

r r + dw
S'

S'

1
( 1r )

dw 4
r n

n
r r =

S'

(3)

where is the mean value of V over the sphere S' and w denote the solid angle.
1
2
3
Let ( x , x , x ) = (P ) as r 0 then as 0 from (3), we have

S'

1
1r

r n
n = 4 (P )

## Then equation (2) becomes

Since 0 then

1 2
dV =
r

r dV

V'

= 0.

1
1r

dS 4 (P )

r n
n

Analytical Mechanics

169

1r
1 1 2
1 1
1

dV
+
dS

dS
(P ) = 4 r
4 S r n
4 S n
V

(4)

## This gives the solution of Poisson's equation at the origin.

If is regular at infinity, i.e., for sufficiently large value of r, is such that
()

m
m

and
r r 2
r

(5)

where m is constant.
If integration in equation (4) is extended over all space, so that r . Then, using equation (5),
equation (4) becomes
(P ) =

1 2
dV
4 r

(6)

## But is a potential function satisfying the Poisson's equation i.e. 2 = 4.

Hence, from (6), we get
(P ) =

dV
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

n dS
i

dV
2

where i =

.
x i

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

dx j

i
P = t , j ds

t i
i
i dx

+
t

= j

j ds
x

j
t i dx j
dx i
+
t

ds
x j ds

dx dx j i

dt i

= ds + ds ds j

d 2 xi
ds

dx j dx k i

ds ds k

j

## Curvature of Curve, Geodesic

171

d 2 xi

dx j dx k i

pi = ds 2 + ds ds j k as j k = k j

## If n is a unit vector in the direction of p , then we have

r
p = k n
The vector n 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

to

f (x i , x&i ) dt

to be staionary are
f
x

d f

=0
dt x& i
dxi

x& = dt i = 1, 2, 3,...
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

where

t1

t0

f (x i , x&i ) dt

...(1)

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 + i) such that i = 0 at A and B both.

172

P'

Fig. 9.1

## In this case the value of integral (1) becomes I

So,
I =

F (x
t1

t0

& i dt
+ i , x& i +

By Taylor's theorem
f
f
+ k +
y
x

F ( x + h , y + k ) = f ( x, y ) + h

Then
I =

I =

t1

F i F i
i
i
F ( x , x& ) + i + i & + dt
x
x

t0
t1
t0

F ( x i , x& i ) dt +

t1 F

t 0 x i

i +

F i
& dt
x& i

I = I +

F
f

i i + i & i dt
t0 x
x&
t1

I = I I =

t1

t0 x i

i +

F i
& dt
x&i

...(2)

zi & j
& i =
x
x j

where
Now,

F i 1
&

dt
= x&i
t0 x
&i
t0
t1 F

t1
t0

t1
t0

d F i

dt
dt x&i

d F i

dt
dt x&i

...(3)

## Curvature of Curve, Geodesic

173
t

F
1
since i i (t ) = 0, i (t1 ) = i (t0 ) = 0

x&
t0

I =

ti F

t0 x i

## The integral I is stationary if I = 0 .

i.e., if

t1 F

t0 x i

d F i

dt
dt x&i

...(4)

d F i

dt = 0
dt x&i

Since i are arbitrary and hence the integrand of the last integral vanishes, so that
F d F

= 0, (i = 1, 2,..., n)
xi dt x&i
Hence the necessary and sufficient condition for the integral (1) to be stationary are

(5)

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
s=

ds
=
dt

gij

gij

dxi d x j
dt
dt dt

...(1)

dxi d x j
dt dt

Put
ds
=
dt

or

gij

d xi d x j
= F (say)
dt dt

...(2)

g ij x& i x& j = F

s& =

s=

B
A

F dt

...(3)

174

## Tensors and Their Applications

Since curve C is geodesic, then the integral (3) should be stationary, we have from Euler's
condition
F
x

d F

=0
dt x& i

...(4)

## Differentiating equation (2) with respect to x k and x& k we get,

F
x

and

1 g ij i j
x& x&
2 s& x k

F
1
1
2 gik x& i = g ik x&i
k =
x&
2s&
s&
d F
1
1 g ik j i 1
i
x& x& + g ik &x& i

= 2 &s& g ik x& +
dt x& k
s& x j
s&
s&

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
j
&
&
2 s& x k
s
s
s

g
&s&
1 g ij i j
x& x& = 0
g ik x& i + ikj
s&
2 x k
x

g ik &x&i

g ik &x&i

&s&
g x& i + [k , ij ] x&i x& j = 0
&s ik

multiplying it by g km , we get
g km gik &x&i

&s& km
g g ik x&i + g km [k , ij ] x& i x& j = 0
s&
m

km
g km gik = im and g [k , ij] = i j

But

x&&m

d 2xm
dt

&s& m m i j
x& + x& x& = 0
s&
i j

&s& dx m m dx j dxk
Replacing dummy
+

= 0 index i by k
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
ds 2

m dx j dx k
+

= 0 ...(6)
j k ds ds

...(5)

175

## which may also written as

dx k
ds

dx m

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
ds 2

m dx j dx k
+

=0
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

and

g ij
xk

0 elsewhere

## 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
h
Aij h
Aih A h j , see pg. 71
Aij , k = x k
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 , k =

Aij
x k

at P0

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 j x j
x s

x j x i

or

x s k x p x q s

x k i j x i x j p q

x s k x p x q s
=

x k i j x i x j p q

...(1)

## Interchanging the coordinate system x i and x i in equation (1), we get

x s

x j x i

x s
=

x k

x p x q s

=
i
j
x x p q
x j
=

k x p x q s
i

j
i j x x p q
x s

x i

x s

x k

(x ) x
s
,i

= ( x ,is ), j

k
x s
s
since k = x, k at P0
i
j

x

k
since i j = 0 at P0

s
,k

Thus,
x,sij

x p x q
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 .

...(2)

177

s x p x q
i
j = 0
p q x x

s
x p
x q

0
0 at P0

= 0 at P0 , as
and
x i
x j
p q

## 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 i the
quantities defined by
dxi

= ds

...(1)

the subscript zero indicating as usual that the function is to be evaluated at P0 . The quantities i
represents that only one geodesic will pass through P 0 in the direction of i in V n. Let yi be the
coordinates of a point P on the geodesic C such that
yi = si
...(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
ds 2 j k ds ds

...(3)

i
is a christoffel symbol relative to the coordinates y i .
j k

where

i i j
i
0+
= 0 since dy = i
j k
ds
or

i i j

= 0
j k

...(4)

i yi y j
yi
= i

=
0
as
s
j k s s
or

i i j

y y = 0
j k

...(5)

178

## 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
y y = 0 hold throughout the Riemannian Vn .
j k

is that

i i j
y y = 0 (from equation 5) throughout
Proof: If y are Riemannian coordinates then the condition
j k
i

the Riemannian Vn .
d 2 y i dy i dy j
i i j

## Conversely if j k y y = 0 hold then ds 2 +

= 0 are saitsfied by y i = s i .
j k ds ds

## Hence yi are Riemannian coordinates.

9.7 GEODESIC FORM OF A LINE ELEMENT
Let be a scalar invariant whose gradiant is not zero. Let the hypersurface = 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 i is given by
i
j
u 2 = g ij u u

i.e.,

(dx )

1 2

1 1
= g11dx dx

g 11 = 1

## Now, if v i is the tangent vector to the hypersurface x1 = 0 then we have

...(1)

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.

## 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.

...(2)

## Curvature of Curve, Geodesic

179

Now,
ti =

and

dx1
dx1
d 2xi
ds 2

d xi d x i
= 1
ds
dx

= 1 and
=

d 2xi

dxi
dx1

= 0 for i 1

= 0 for i = 1, 2,..., n

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
1 2 g 1 j g 11

2 x i
x j

ij
as g 0

= 0, since g 11 = i g 11 = 0

x j

So,
g 1 j
x1

= 0 for j 1

...(3)

## from equations (1), (2), and (3), we have

g 11 = 1, g1j = 0; ( j = 2, 3, , n),

g1 j
x1

= 0, ( j = 2, 3, , n)

## The line element is given by

i

ds 2 = gij dx dx

1 1
j
k
ds 2 = g 11dx dx + g jk dx dx
1 2
j
k
ds 2 = ( dx ) + g jk dx dx ; ( j = 2, 3, , n, k = 2, 3, , n)

...(4)

The line element (4) is called geodesic form of the line element.
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 = x1 = constant form a
system of parallels i.e., the hypersurfaces = x 1 = constant are geodesically parallel hypersurfaces.

Note 1:

180

## Tensors and Their Applications

THEOREM 9.5 The necessary and sufficient condition that the hypersurfaces = constant form a
system of parallel is that ()2 = 1.
Proof: Necessary Condition
Suppose that hypersurface = constant form a system of parallels then prove that ( 2) = 1.
Let us take the hypersurface = 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
1 2
i
j
ds 2 = ( dx ) + g ij dx dx

...(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,
ij
()2 = = g

ij
= g

x i dx j

x1 x1
= g ij 1i 1j
i
j
x x

()2 = g11 = 1

so

()2 = 1
Sufficient Condition
Suppose that ()2 = 1 then prove that the hypersurface = constant from a system of parallels.
Let us taken = x1 and orthogonal trajectories of the hypersurfaces = 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 ()2 = 1

g ij

=1
x i x j

g ij

x1 x1
=1
x i x j

g ij 1i 1j = 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 = 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
ds 2 j k ds ds

...(1)

## In case of Euclidean coordinates the fundamental tensor g ij is denoted by aij and

g ij = a ij = ij = 1, if i = j
0, if i j
g ij
x

a ij
x k

=0

k
This implies that = 0, [ij, k ] = 0 relative to S n .
i
j
Then equation (1) becomes
d 2 yi
ds 2

=0

## 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 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 )
Then equation (2) becomes
y i y i = a i l
n

l2

i =1

(ai ) 2 =

( y

y i )2

i =1

## But a i is the unit tangent vector to the geodesics. Then

n

(a )

i 2

=1

i =1

So,
n

( y

y i )2

i =1

l=

( y

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

...(2)

183

## Then the lines AB and AC are orthogonal to each other. So,

AB AC = 0
a ij ( y i2 y1i ) ( y i3 y1i ) = 0

or

or

(y
i =1

i
2

y1i ) ( y 3i y1i ) = 0

...(1)

n

(AB) =

(y

i
2

y1i ) 2

...(2)

i
3

y1i )2

...(3)

i
3

y i2 )2

...(4)

i =1
n

(AC) =

(y
i =1
n

(BC) =

(y
i =1

n

(BC) =

[( y

i
3

i =1
n

[( y

i
3

i =1
n

## y1i ) + ( y1i y i2 )]2

( y i3 y1i )2 +

i =1
n

(y

i
1

y i2 ) 2 + 2 0, [from (1)]

i =1

( y i3 y1i )2 +

i =1

( y

i
1

y2i )2

i =1

## (BC)2 = ( AC)2 + ( AB)2

Hence Pythagoras theorem holds in S n .
EXAMPLE 2
Prove that if is any solution of the differential equation ()2 = f () then the hypersurfaces =
constant constitute a system of parallels.
Solution
Given that
()2 = f ()

...(1)

184

## Then prove that the hypersurfaces = constant form a system of parallel.

Suppose
=

or

d
=
d

Now,

f ( )

d
f ()

, Then, d =

1
f ( )

=
i =
x i
x

()2 =

f ( )

1
f ( )

1
1
( ) 2 =
f (); from (1)
f ( )
f ( )

()2 = 1
This proves that the hypersurfaces = constant form a system of parallels and therefore the
hypersurfaces = 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.
x

1
2

h i
l
m
m
( x x0 ) ( x x0 )
l
m

j
m
m
j
= a m ( x x0 ) + a h

...(1)

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
x
x j

x m

1
2

= a j at P
m
0

h
i
l
2 ( x x0 )
l
m

j
j
= a m + ah

...(2)

...(3)

## 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

0
l m 0

...(4)

## But we know that

(x )

j
,lm 0

2x
= l

x x 0 l m 0
j

x j

x h

h h j
a h , (from (3) and (4))
l m 0 lm

j
= ah

(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 is any scalar
function of the x's show that
d p
ds p

= , ij... l

d xi d x j
dxl

ds ds
ds

...(1)

Solution
Since the coordinates x i lie on a geodesic. Then
d 2 xi
ds 2

i dx j dx k
+
=0

j k ds ds

...(2)

## Here the number of suffices i j...l is p.

We shall prove the theorem by mathematical induction method.
Since xs are functions of s and is a scalar function of x s , we have
d
dxi
=
ds
x i ds
d 2
ds2

or

d
dx i
= , i
ds
ds

, i dx j
d 2xi
+

,i
x j ds
ds 2

,i dx i dx j
= x j ds ds ,i

i d x j d x k
, from (2)

j k ds ds

...(3)

186

,i dx i dx j
= x i ds ds , m
=

m dx j dx k

j k ds ds

, i dx i dx j
m dxi dx j

## (adjusting the dummy index.)

,
m
x j ds ds
i j ds ds

, i
m dx i dx j

,m
= j
i j ds ds
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
d p

dx r1 dx r1
dx p

ds p
ds ds
ds
Differentiating the equation (5) with respect to s, we get
d p +1
ds p +1

= , r1 , r2 , ...,

rp

...(5)

r
r
r
, r1 r2 ... r p dx r1
dx p dx p +1
d 2 x r1 dx r2
dx p

+
r
r1 r2 ... rp
ds
ds
ds
ds
x p +1
ds 2 ds

+ , r1 r2 ... rp
substituting value of

...(4)

dx r1
d 2x p

ds
ds 2

...(6)

d 2 x r1
etc. from (2) in (6) and adjusting dummy indices, we have
ds 2
, r r ... r
m
m

1 2
p

d p +1

,
m
r
...
r
,
r
r
...
m
r
2
p
1 2
p +1 = x p +1
r
r
r
r

ds
1 p +1
p p +1

dx r1 dx r2
dx p +1

ds ds
ds
r

d x r1 d x r2
d x p +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.
= ,r1 r2 .. r p rp +1

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
i
xi = x +

1 i
c x j xk xl
6 jkl

## Curvature of Curve, Geodesic

187

where Cs 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
i
x i = x + 2 j k x x

## is geodesic coordinate system with the pole at the origin.

5. Obtain the equations of geodesics for the metric
ds2 = e2 kt ( dx2 + dy2 + dz2 ) + dt 2
6. Obtain the differential equations of geodesics for the metric
1
2
2
2
2
ds2 = f ( x) dx + dy + dz + f ( x) dt
2

d2x 1 d
dx
1 d
dt
d2y
d2z
d 2t d (log f ) dx dt
Ans : 2

(log f ) +
(log
f
)
=
0
;
=
0
;
=
0
;

=
0
2 dx
ds
dx
ds ds

ds
ds 2 f 2 dx
ds2
ds2
ds2

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-CIVITAS

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
j
ui

m i dx
j + u m j
=0
ds
x
j
ui d x j
m i dx
+
u

=0
x j ds
m j ds

du i
i
+ um
ds
m

dx j

j ds = 0

(2)

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

or

( gil u,i j )

dx j
=0
ds

( g il u i ) , j

dx j
=0
ds

ul , j

dx j
ds = 0

u i, j

dx j
=0
ds

ui
m dx j

u
j
m
=0
i j ds
x
dui
m dx j
um
=0
ds
i j ds
The equation (2) and (3) can be also written as
m i
j
dui = u m j d x

(3)

(4)

m
j
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.
and

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
= 0

ds

j
d
x
vi, j
= 0
ds

u,i j

at each point of C.
Multiplying (1) by gil, we get
dx j
( g il u,i j )
=0
ds

(1)

190

or

ul , j

dx j
=0
ds

u i, j

dx j
=0
ds

(2)

vi , j

dx j
=0
ds

(3)

Similarly,

## Let be the angle between ui and v i then

ui.v i = uv cos
Differentiating it with respect to arc length s, we get
d
d (u i vi )
(uv cos ) =
ds
ds
i
= (u v i ), j

uv sin

dx j
ds

j
d
dx j
i dx
vi + u i vi , j
= u, j
ds
ds
ds

## Using equation (1) and (3), then equation (4) becomes

d
uv sin
=0
ds
d
sin
= 0, as u 0, v 0

ds
d
Either sin = 0 or
=0

ds
Either = 0

or

= constant.

## 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
+
ds 2
j k ds ds

=0

d d x m m d x j d x k
+
ds ds j k ds ds = 0

x j

dx m

ds

d x j m d x j d xk

ds + j k ds ds = 0

(4)

Parallelism of Vectors

j
x

dx m

ds

191

m dx k dx j
+

=0

j k ds ds

d xm dx j

ds ds = 0

, j

or

t,mj

dx j
=0
ds

dx m
suffer a parallel displacement along a geodesic curve.This
ds
confirms that geodesic is an auto-parallel curve.
Proved.
This shows that the unit tangent vector

## 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 = ui
(1)
where is a function of arc length s.
If ui is parallel with respect to Riemannian V n along the curve C. Then,
u ,ij

dx j
=0
ds

(2)

The equation (1) shows that v i is of variable constant and parallel with respect to Riemannian V n
so that
v ,ij

dx j
dxj
i
= ( u ), j
ds
ds
i
i
= (, j u + u, j )

dx j
ds

= , j

dx j i
dxj
u + u,i j
ds
ds

= , j

dx j i
u Since
ds

d x j i
u
=
x j ds
v,i j

d i
dx j
u
=
ds
ds
d vi
= ds

from (1)

u,i j

dx j
=0
ds

192

## Tensors and Their Applications

i
= v

d (log )
ds

dx j
d (log )
= 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 i f (s )
v,ij =
ds
to show that v i is parallel, with respect to V n .
Take
v,i j

ui = v i (s)

(5)

Then
dx j
dx j
i
= (v ), j
ds
ds
j
dx j i
i dx
+ , j
v
= v, j
ds
ds
dx j i
i
v
f
(
s
)

+
v
=
x j ds
d
dx j
i
u,i j
= v f (s ) +
(6)
ds
ds

d
= 0.
Select such that f ( s ) +
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
u ,ij

v,i j

dx 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 f (s )
ds
Interchange the indices l and i, we get
v l v,i j

dx j
= v i v l f (s)
ds
Subtract (1) and (2), we get
v i v,l j

(v l u,ik v i v l,k )

(2)

dx 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 and having
the metric ds2 = ax dy dy, where m > n. Let Greek letters , , 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 ) 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 dy dy
y
x i
y
a
gij = i
x

y i j
dx dx
x j
y
x j

(1)

## As dxi and dxj are arbitrary.

This gives relation between gij and a .
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 x j defined at any point of V n and suppose that the same vectors
in V m are represented by dy and y respectively. If is the angle between dxi and x i then
cos =

g ij dx i x j
g ij dx i dx j g ij x i x j

cos =

a dy y
a dy dy a y y

(1)

194

## Tensors and Their Applications

y i y
dx
dx j
x i
x j
y i y j
y i y j
d
x
d
x
a
x
x .

x i
x j
xi
x j
a

=
a

y y i j
dx dx
x i x j
y y i j
y y i j
dx
dx
a
x x .

xi x j
x i x j
a

=
a

cos =

Since g ij = a

gij dx i dx j
g ij dx i dx j g ij x i x j

(2)

y y
(from equation (1), art. 10.3)
x i x j

## from (1) & (2)

cos = cos = .

Proved.

THEOREM 10.4 If U and ui denote the components of the same vector in Riemannian Vm and Vn
respectively then to show that

i y
= 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 respectively. Let C be curve passing through s in the direction of the given
vector then
y y dx i y i
= i
= i a
=
(1)
A
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

i
i
(2)
U = aA , u = aa
Multiplying (1) by a, we get

y
(aa i )
=
aA
x i
y i
=
u , using (2)
U
x i

Or

Proved.

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 of points in V m are expressible as functions of
coordinates xi in V n.

Parallelism of Vectors

195

Now,
dy
y dx i
=
ds
x i ds
i
for the curve x = s, we have

...(1)

dy
y
=
ds
x i

...(2)

dy
y
is a vector tangential to the curve in V m. Then from equation (2) it follows that
ds
xi
i

## 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
Since

y
N =0
xi
y
(a N ) i = 0
x
y
N i = 0,
x
a

a N y ,i = 0

...(3)

i = 1, 2, , n & = 1, 2, , m

## The equation (3) are n equations in m unknowns N

y
(m > n). The coefficient matrix i
x

is

## 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 . This shows that there
are m n linearly independent normals to V n to V m.
EXAMPLE 2
Show that
[ij, k] = [, ]

y y y
2 y x
+
a

x i x j x k
xi x j x k

where [, ] and [i j, k ] are the Christoffels symbols of first kind relative to metrics a dy a dy and
g ij dx i dx j .
Solution
Since relation between a and g ij is given by
y y
x i x j
Differentiating it with respect to xk, we get
g ij = a

g ij
x k

a y y y
2 y y
y 2 y
+
a
+
a
=

x x i x j x k
x i x k x j
x i x j x k

(1)

196

## Tensors and Their Applications

Similarly
g jk
x i

a y y y
2 y y
y 2 y
+
a
+
a
=

y x i x j x k
x j x i x k
x j x k x i

(2)

and
a y y y
g ki
2 y y
y 2 y
+
a
+
a
=

x j
x x i x j x k
x k x j x i
x k x i x j
But we know that
[ij, k ] =

and,

1 g jk g ki g ij

2 x i
x j x k

1 g g g

[, ] = 2 +
y
y
y

(3)

(4)

(5)

Substituting the value of (1), (2), (3) in equation (4) and using (5) we get,
[ij, k ] = [, ]

y y y
2 y y
+
a
.

x i x j x k
xi x j x k

## 10.4 PARALLELISM IN A SUBSPACE

THEOREM 10.6 Let T 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 q are derived vectors of t
along C relative to Vn and Vm respectively. Then
y
= pi
xi
Proof: Since from equation (1), theorem, (10.4), Pg. 194 we have
q

dT
ds

y
x i
2
dt i y
dx j
i y
+
t
=
ds x i
x i x j ds
i
= t

(1)
(2)

Now,
dx j
ds

dy
= T,
ds
T

dy

+
T

=
ds
x

T dy
dy
+
T

=
x ds
ds

i
p i = t, j

dT
dy
+
T

= ds
ds

(3)

(4)

Parallelism of Vectors

## Putting the value of

197

dT
and T from (2) and (1) in equation (4), we get
ds
2

dt i y
dx j
i y
i y dy
+
t
+
t
q =

ds x i
x i x j ds
dx i ds

t i y dx j
dx j
i y
i y dy
+
t
+
t

x j x i ds
x i x j ds
x i ds

j
t i y dx j
dx j
i y
i y y dx
+
t
+
t

x j x i ds
x i x j ds
x i x. j ds

t i y d x j i d x j 2 y
y y
+
t
+
+

ds x i x j x j x i
x j x i ds

(5)

## But we know that

[ij, k ] = [, ]

y y y
2 y y
+
a

x i x j x k
x i x j x k

[ ji, k ] = [, ]

y y y
2 y y
+
a

x j x i x k
x i x j x k

y y y
2 y y
a
+
a
= j

x x i x k
x i x j x k

y
a
= x k
Multiplying (5) by a

(6)

y
, we get, using (6),
x k

Or

2
y y
j i + iy j
x x
x x

j
y
t i dx j y
y
i dx
q
a
+
t
[ij, k ]
=

ds
x k
x j ds x i
x k

dx j t i
p
y
g + t i g pk ,
j ik
k =

ds x
x
i j

dx j
= ds

t i

i
j gik + t a g ik
x

a j

= g ik

dx j i
t,
ds j

since g ij = a

y y
x i x j

198

## Tensors and Their Applications

i
= ( gik t ), j

= tk, j

dx j
ds

dx j
ds

y
dx j
t
=
k, j
ds
x k

y
q k = pk
x
q

or

(7)
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 q
vanishes i.e.,
q = 0, = 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.
y
Proff : Science, q
= pi
x i
Let t be unit tangent vector to the curve C them T 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

q =0
curve C is a geodesic in V m
Also,
q = 0, 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
iff q
But q
For

y
=0
x i

y
= 0 implies that q is normal to V n.
x i

y
= y,i 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., q 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 q 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
q i = pi
x
q

y dx i
dxi
p
=
i
ds
x i ds

dy
dxi
= pi
ds
ds

dy dy
dx j dx i
T,
= ti, j
ds ds
ds ds
i.e., tendency of T 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.
q

## 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
=0
(1)
ds
dx j
q ,ij
and
=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,
p,i j

( gij p i q j ),k

dx k
=0
ds

200

## Tensors and Their Applications

k

dx k j
dx k i j
i j dx

g ij p,ik
q
+
g
p
q
+
g
ij
,k ds ij ,k ds P q = 0
ds

(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
0)
(Since and are unit vectors and
ds
g ij
m
m
g mj gim = 0
k
x
i k
j k
g ij
[ik , j ] [ jk , i ] = 0
x k
g ij
= [ik , j] + [ jk , i ]
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
pi

qi

[ij,k] =
k
lk
But we know that i j = g [ij, l ]

from (5), we have

1 g jk g ki g ij

2 x i
x j x k

k
1 lk g jl g li gij
= g i + j l
2
x
x
i j
x

(4)

(5)

Proved.

EXAMPLE 3
If t i and T are contravariant components in x's and y's respectively, of n vector field in V n
immersed in V m. Show that
T, j = y,ij t i + y ,i t,i j
Solution
Since we know that
T

i
= t

y
x i

Parallelism of Vectors

201

i
T = t y, j .
Taking covariant differentiation of both sides, we get

T, j = (t i y ,i ), j
= t,i j y,i + t i ( y ,i ), j
T, j = t i y ,ij + t,i j y ,i
EXAMPLE 4
Show that g ij

dx i dx j
remains constant along a geodesic.
ds ds

Solution
i
Let t =

dx i
. Then
ds

dx i dx j
= g ij t i t j = t 2
ds ds
Since we know that geodesics are autoparallel curves. Then
gij

t,i j

dx j
=0
ds
t,i j t j = 0

or

(1)

Now,
dt 2
d
d
( g ij t i t j ) = (ti t i )
=
ds
ds
ds
dx j
i
= (t i t i ), j t j
= ( ti t ) , j
ds
dt 2
= (ti , j t i )t i + (t ,i j t i )ti = 0,
ds

from (1)

Integrating it we get
t2 = constant.
So, g ij

dx i dx j
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
Since geodesics are auto-parallel curves. Then to prove that
(i)

(1)

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

Thus

t i t j, i = 0

t j ,i t i = 0

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
1 2
1
2
2 2
ds 2 = (dx ) + 2 g12 dx dx + (dx )

(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
pi
dx i
i
t
=
= (1,0)
=
p
ds

where

p = dxi

So, we have
dx 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).
p,i j

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

## 3. Establish the equivalence of the following definitions of a geodesic.

(i) It is an autoparallel 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 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 lhk, so that

lhk = el|i, j eih|ekj|

(1)

## The invariants lhk are Ricci's Coefficients of Rotation.

Since i being a dummy index, has freedom of movement. Then equation (1) may be written as
lhk = el |i , j eh|i ekj|

(2)

The indices l, h and k are not tensor indices. But these indices in lhk 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.,
lhk = lhk
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

## multiplying by ekj| and summing for j, we get

eh |i , j eli|eki | + eii|, j eh|i e kj| = 0
hlk + lhk = 0
hlk = lhk

or

(1)

llk = llk
2llk = 0
llk = 0

or
or

lhk eh|i
h

= el|i , j ekj|

## Proof: since we know that

lhk = el|i, j eih|ekj|
Multiplying by eh|m and summing for h.

## lhk eh|m = el|i , j eih|ekj|eh|m

h
h

j
= el |i , j e k|

(e e )
i

h|

h| m

= el |i , j ekj|im since

ehi | eh|m =
h

i
m

= (e l |i , j m ) e kj |
i

lhk eh |m

= el |m , j ekj |

Replacing m by i, we get

lhkei|m
h

= el |i , j e kj|

## 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
and

u,ij ehj| = 0

(1)
(2)

207

## If is the angle between the vectors ui and e ih| , we have

cos = ui eh|i
Differentiating it with respect to arc length sm along Cm, we get
sin

d
i
j
ds m = (u e h|i ), j em |
= (u i e h|i , j + u i, j eh |i ) emj |

sin
at the point P 0, =

d
u i e h |i , j e mj | + u ,i j e mj |
ds m =

(3)

, we have
2

d
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 |

d
ds m = hlm
d
ds m = hlm

(4)

d
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 hlmare 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 | =

hlh eil|
l

(1)

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

i
hlh e l |

i
hmh em |

## g ij eli| emi | hlh hmh

hmh hmh

l
m hlm

hmh , Since g ei e i = l
m
ij l| m|

K h2| =

hmh )

## 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 | =

hlh eil|
l

Since

eli|

0 , h and hence

phi |

## = 0 iff hlh = 0, h . But pli| = 0 iff the congruence C is geodesic

congruence.
Hence C is a geodesic congruence iff hlh = 0, h .
But
hlh = lhh
So, C is a geodesic congruence iff lhh = 0, h .
or C is a geodesic congruence iff lhh = 0.
Hence lhh = 0 are the necessary and sufficient conditions that congruence with unit tangents ehi |
be geodesic congruence. Again lhh 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
nqp = npq
Proof: Consider a congruence C of curves in a V n. Let ( x1 , x 2 , , x n ) = constant be a family of
hypersurfaces. To determines a normal congruence whose tangent vector is grad or .
Let ti be the convariant components of unit tangent vector to C. The congruence C is a normal
congruence to a family of hypersurface ( x1 , x 2 , , x n ) = constant if
,
,1
, 2
= 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)
,i
= y or , i = yti
ti

yti =
x i

## Differentiating it with respect to x j , we get

y
x j

ti + y

t i
x j

x i x j
Interchanging indices i and j, we get
t j
y
2
t
+
y
=
j
x i
x i
x j x i
Subtracting (2) and (3) we get
t y
t
y
j ti + y i j i t j + y i j = 0
x x
dx
dx
t j y
t
y
y ij i + t i j t j i = 0
x x
x
x

(2)

(3)

210

## Tensors and Their Applications

Multiplying by tk,
ti t j
y
y
ik ti i = 0
ytk j i + ti tk
j
x
x
dx
dx
By cyclic permutation of i, j, k in (4), we get
t j t
yti k kj
x
x

(4)

y
y
+ t j t i k i k ti
=0
x
x j

(5)

and
t
t
y
y
yti ki ik + t k t j i ti t j k = 0
x x
x
x

(6)

## On adding (4) , (5) and (6) we get

t
t j
t j t
y t k ij i + ti k kj
x
x
x
x

+ t j tk ti

x i x k

t k (ti , j t j ,i ) + ti (t j , k t k , j ) + t j (t k , i ti ,k ) = 0

or

= 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.
e n |k eqk| = q = 0, q n
h

But

en |i e kp| = 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 (nqp npq) = 0
Since

211

## nqp npq = 0, en| j 0

npq = nqp,
(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
nqp = npq (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
nqp = npq (p, q = 1, 2, , n 1 such that p q)
If the indices h, k, l and unequal then
hkl = hlk
(1)
But due to skew-symmetric property i.e., hlk = lhk.
So,
hkl = hlk = lhk = lkh, from (1)
= klh (skew-symmetric property).
= khl, from (1)
hkl = hkl , (skew-symmetric property).
hkl + hkl = 0
2 hkl = 0

hkl = 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

h

212

## Putting l = n and j = m, we have

n

nhk eh |i

h =1

m
= e n |i , m e k |

n

h,k =1

= en |i ,m mj

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
=

h,k =1

n ,k =1

n

curl en| i =

from (2)

nhk eh |i ek | j

h ,k =1

h , k =1

nkh ek | j e h|i

curl en|i =

h ,k =1

(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

curl en| i =

h ,k =1

n 1

( nhk nkh ) eh |i ek | j +

h =1

k =1

## Since we know that nnk = nnh = nnn= 0.

So,
n 1

curl en|i

h , k =1
n1

curl en|i

( nhk nkh ) e h |i e k | j +

h ,k

n 1

nhn e h|i en | j

h =1

## ( nhk nkh ) eh|i ek | j +

n1

nkn en |i

ek| j

k =1

n1

h =1

nhn ( eh |i en | j

en |i ek | j )

(4)

213

## The first term on R.H.S. of equation (4) vanishes

if
nhk nkh = 0
i.e.,
if
nhk = nkh.
i.e., if the congruence e n| is normal.
Again the second term of R.H.S of equation (4) vanishes
nhn = 0

if

i.e., if 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
nhk + nkh = 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 = (en| i, j + en| j,i)

(1)

## Let us find a quantity and n quantities ei satisfying the n + 1 equations

e n|i ei = 0
i, j = 1, 2, , n
( X ij gij e i + E n| j = 0

(2)

## where is a scalar invariant.

Writing equation (2) in expansion form, we have
e n|1e1 + en |2 e 2 + + en| n e n = 0
and
( X 1 j g1 j ) ei + ( X 2 j g 2 j ) e 2 + ( X nj g nj ) e n + e n| j = 0

(3)

214

## Tensors and Their Applications

for j = 1, 2, , n.
1

( X 11 g11 ) e + ( X 21 g 21 ) e + + ( X n1 g n 1 ) e + en |1 = 0
( X 12 g12 ) e1 + ( X 22 g 22 ) e 2 + + ( X n 2 g n 2 )e n + en |2 = 0
M
( X 1n g1 n ) e1 + ( X 2 n g 2 n ) e 2 + + ( X nn g nn ) e n + en |n = 0
from (3)
e n|1 e1 + en |2 e 2 + en |n e n + .0 = 0
eliminating and the quantities e1, e2, en, we have the equation

(X 11 g11 ) (X 21 g 21 )
( X12 g12 ) ( X 22 g 22 )

(X n1 g n1 )
( X n2 g n2 )

(X 2n g 2n )

(X nn g nn )

M
( X1n g1n )
en |1

en |2

en| n

en |1
e n|2
en |n
0

## which is of degree n 1 in . Hence there will be n 1 roots of and these roots be 1, 2, ,

n 1. All roots are real. Let h be one of these roots and let the corresponding values of and ei be
denoted by h and e ih| respectively. Then h and e ih| will satisfy the equation (2), we have
en |i e ih| = 0
and
( X ij g ij ) eih | + h en | j = 0
Similarly k be another root of these roots, we have

(4)
(5)

en |i e ik| = 0

(6)

( X ij g ij ) eik | + h en | j = 0

(7)

Multiplying (5) by e kj| and (7) ehj| and using (4) and (6), we get
( X ij h g ij ) ehi |ekj| = 0

(8)

( X ij k 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 k g ij ) ehi |e kj| = 0
Subtracting (10) and (8), we get

(10)

(h k ) g ij e ih| ekj | = 0

(11)

Since k k 0 as h h.

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
+ e n| j , i . Then we have,
2 n |i , j

1
e
+ en | j ,i ehi |ekj| = 0
2 n |i , j
en |i , j ehi | ekj| + en| j ,i eik ehj = 0
nhk + nkh = 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
nhk + nkh = 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 nhk = 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
nhk + nkh = 0 (h, k = 1, 2, , n 1 such that h k).
If the congruence en| is normal. Then
nhk = nkh
The given condition nhk + nkh = 0 becomes
nhk + nhk = 0
2nhk = 0
nhk = 0

Proved.

EXAMPLE 1
1
If en| are the congruences canonical with respect to en| prove that (i) h = nhk (ii) h = hnn,
2
(iii) If en| is a geodesic congruence, the congruences canonical with respect to it are given by
( X ij g 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
where

( X ij h g ij ) ehi | + h en | j = 0

X ij =

(2)

1
(e
+ en| j,i)
2 n| i,j

(3)

## Since e in| unit tangents then

g ij ehi |ehj| = 0

(4)

## (i) Multiplying equation (2) by e ih| , we get

( X ij h g ij ) ehi | ehj| + h e n| j e hj| = 0
X ij eih| ehj| h g ij eih |ehj| = 0

(from (1))

## X ijeih| ehj| h = 0, since g ijehi |ehj| = 1

1
(e n|i , j + en| j , i ) e ih|e hj| h = 0;
2

from (3)

1
(e e i e j + en | j ,i eih| ehj | ) h = 0
2 n|i , j h | h|
1 ( + ) = 0
nhh
nhh
h
2
h = nhh
(ii) Multiplying (2) by enj| , we get
( X ij h g ij ) ehi | enj| + h e n| j e nj| = 0
X ij eih| enj| h g ij ehi |enj| + h 1 = 0

1
(e n|i , j + en| j , i ) e ih|e hj| + h = 0,
2

from (1)
since e ih| e hj| = 0

1
(e ei e j + en | j ,i ehi | ehj | ) + h = 0
2 n|i , j h | n|
1
( + nnh) + h = 0
2 nhn
h =
or

h =

1
, since nnh = 0
2 nhn

2 hnn

(iii) If en| is a geodesic congruence, then hnn = 0 from the result (ii), we have
1
h = 0
2

h = 0

217

## ( X ij h gij ) ehi | + 0en| j = 0 or

( X ij h g ij ) e ih| = 0

this gives ( X ij g 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
lhk = 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
nhk + nkh = 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 is a scalar invariant
n

,ij eih|ehi | =
2

h =1

## 2. The coefficient of n 1 in the expansion of the determinant |,ij g ij | is equal to 2 .

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

Q
( lhk lkh )

=
sl
sh sk sk sh
l

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 N 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 =
i
gij dx dxj. Let V n be Riemannian space of m dimensions referred to coordinates y and having the metric
ds2 = a dy dy. Where m > n. Let Greek letters , , take the values 1, 2, , m and latin indices
i, j, k, take the values 1, 2, n. Then we have, the relation between a and gij

gij = a y y

xi x j

(1)

Since the function y 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.
y, i =

i.e.,

y
xi

## Then equation (1) can be written as

gij = a y,i y, j
The vector of V n

+ i

(2)

## whose contravariant components are y,i is tangential to the curve of

parameter xi in V n. Consequently if N are the contravariant components of the unit vector normal to
V n. Then we have
a N y,i = 0,
and

N N =

(i = 1, 2, , n)

(3)
(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 u and 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, u is parallel along C relative to V m then we have,
u,

dy
=0
ds

u
dy
u
=0
ds
y
u dy
dy

=0

y ds
ds
du
dy
u
=0
ds
ds
dy
du
= u ds
ds

(1)

dy
dv
= v ds
ds

(2)

## and wi is parallel along C relative to V n then

j i dx
dwi
= w j k ds
ds

(3)

The Christoffel symbol with Greek indices being formed with respect to the a and the y's and
christoffel symbol with Latin indices with respect to the gij and the x's.
Let Ai be a tensor field, defined along C, which is mixed tenser of the second order in the y's and

a covariant vector in the x's. Then the product u v wi Ai is scalar invariant and it is a function of s
along c. Its derivative with respect to s is also a scalar invariant.

## Differentiating u v wi Ai with respect to s, we have

dA
d
dv
dw i
i
(u v wi Ai ) = u v w i i + du v w i A +
u
w
A
+
u v Ai

i
dt
ds
ds
ds
ds

= u v w

dAi
ds

du i dv
dw j
v w Ai +
u w i Ai +
u v Aj
ds
ds
ds

220

## Tensors and Their Applications

i
= u v w

dy

i dy
+ Ai .u v wi
Ai .u v w ds
ds

ds

dAi

i j dx
Aj u v w ik ds , by equation (1), (2) & (3).

k

i dAi
dy
dy
j dx
u
v
w

+
A
.

A
.

i
i
j
=
ds
ds
i k ds
ds

= Scalar, along C
Since the outer product uv wi is a tensor and hence from quotient law that the expression within

## the bracket is tenser of the type Ai

and this tensor is called intrinsic derivative of Ai
with respect to
s.
The expression within the bracket can also be expressed as

dxk
ds

Since

Ai

Ai

+ Ai y,k Ai y,k Aj
xk

ik

dxk
is arbitrary..
ds

So, by Quotient law the expression within the bracket is a tensor and is called tensor derivative of
with respect to xk. It is denoted by Ai;k . Then we have
Ai;k =

Ai

j
+ Ai y,k Ai y,k Aj .

dx

ik
k

## Ai;k is also defined as generalised covariant derivative Ai

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 A , B and B are tensors in V m.
(i) To prove that
( A + B ); k = A;k + B; k
Let the sum A + B be denoted by the tensor C .
Now,
C;k

+ C y, ck Ca y, ck
x
a c
c
k

Hypersurface

221

( A

B ); k

( A + B )
dx

+ ( Aa + A ) y,ck ( Aa + Ba ) y ,ck
a c
c

a c
a c

+
A
y

A
y

,
k
a
,
k
= x k
+
ac
c

B

a

+ Ba y,ck Ba y,ck
xk
ac
c

## ( A + B ); k = Aa ; + Ba ;k Hence the result (i)

(ii) Prove that
( A + B ); k = A; k B + A B ; k

A B = D

Let

Then D
is a tensor

we have
D ; k =

( A

B ) ; k =

a
a
+ D;ak yc Da y,ck Da y,ck
ac
c
c

( A B )
dx

a
a
a
+ A
B y ,ck Aa B y,ck A B a y ,ck
ac
c
c

a
c
a c

+
A
y

A
y
B + A Ba y,ck

,
k
a
,
k
= dxk
k
ac
c
c
dx

( A B ); k = A; k B + Aa B ; k
Hence the result (ii).
Note: , , , a, c take values from 1 to m while k take values from 1 to n.
THEOREM 12.2 To show that a; 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)
a;i =

a
dxi

a
dx i

a y,i a y,i

( [, ] + [, ] ) y,i

222

or

a;i =

or

a;i =

a
dx

a
dx

a
y,i
y

a y
a a
i =

y
dx
x i
x i

= 0

Proved.

## 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;ij = ij N

Proof: Since y is an invariant for transformation of the x's and its tensor derivative is the same as its
covariant derivative with respect to the xs, so that
y; i = y, i =

dy

(1)

dxi

y;ij

j
dx
ij

y l
y,l + y,i y, j
x j xi
ij

2 y

l

y,l + y,i y,j
x x
ij

i

(2)

## Interchanging j and i in (2), we have

y;ji

2y
l

y ,l + y , j y ,i
=
j
i
x x
ji

y;ji

2y
l
y ,l + y ,l y ,j
=
i
j
x x
ij

(3)

## [On interchanging and in third term of R.H.S. of (3) and using = ].

On comparing equation (2) and (3), we have
y;ij = y;ji

Hypersurface

223

Then
gij = a

y y
xi x j

or

## Taking tensor derivative of both sides with respect to xk

gij;k = a;k y;i y;j + a y;i y;jk + a y;ik y;j
But

gij;k = 0

and

a;k = 0.

we have,
0 = a y ;i y;jk + a y;ik y;j
or

(4)

## By cyclic permutation on i, j, k in (4), we have

and

a y; ji y,k + a y , j y;ki = 0

(5)

## a y;kj y,i + a y ,k y ;ij = 0

(6)

subtracting equation (4) from the sum of (5) and (6), we get.
2 a y;ij y ,k = 0
a y ;ij y ,k = 0

or

(7)

This shows that y;ij is normal (orthogonal) to y,k . Since y,k is tangential to V n and hence y;ij is
normal to V n. Then we can write
y;ij = N ij
where

(8)

is unit vector normal to V n and ij is a symmetric covariant tensor of rank two. Since

## The equation (8) are called Gausss formula.

From equation (8)
y;ij = ij N

y;ij a N = ij a N N
= ij N 2

224

y;ij a N = ij ,
or

ij

N=1

y;ija N .

## The quadratic differential form

ij dxi dx j

is called the second fundamental form for the hypersurface V n of V n + 1. The components of tenser 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 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 i
u = y,i ui
(1)
U
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

and

pi

u,i j

dxj
ds

= U,

d y
ds

= U ;j

dx j
, (from equation 1, Art 12.4)
ds

(2)

Taking the tensor derivative of each side of equation (1) with respect to x's, we have
U ;j = y; ij u + y, i u , j

## By Gauss's formula, we have y;ij = ij N

Then

U ;j = ij N u + y ,i u , j

## Putting the value of u;j in equation (2), we have

dx j
ds

i
i
= (ij N u + y, i u, j )

or

i
i dx

u
= ij
N + y,i p
ds

where

i
pi = u, j

dx j
ds

(3)

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
N + y ,i p i
ds

q = ij ds

Kn = ij

Taking

dx i dx j
,
ds ds

(4)

we get

q = Kn N + y,i pi

...(5)

Kn is called Normal curvature of V n at any point P of the curve C and Kn N 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

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
Proof: We know that
q = Kn N + p i y,i

...(1)

dxi dx j
ds ds
Let Ka and Kg be the first curvatures of C with respect V n + 1 and V n respectively then
where

Kn = ij

Ka =
r

a q q , Kg =

gij pi p j

## Let be the angle between N and C .

Then

r r
N C = N C cos

= cos
r r
N .C = cos

If b is a unit vector of V n +
equation (1) becomes

as |N| = |C| = 1
...(2)
r

## in the direction principal normal of C with respect to V n then

r
r
r
Ka C = K g b + K n N

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 = Kg 0 + K n 1;

Kn = Ka cos

(3)

from (2)
(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 Meuriers 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
Kn2 = Ka K g .

or

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 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 = ij eih | ehj|
The sum of normal curvatures for n mutually orthogonal directions of an orthogonal ennuple is a
V n is
n

Knh =
h =1

ij

eih| ehj|

h =1

= ij

i
h|

ehj|

h =1

= 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 = ij g ij
(b) Minimal Hypersurface
The hypersurface V n is said to be minimal if M = 0
i.e.,
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 ij .
(d) Principal directions of the hypersurface at a point P.
The principal directions determined by the symmetric tensor ij at P are said to be principal directions
of the hypersurface at P.
(e) Line of curvature in V n
A line of curvature in a hypersurface V n is a curve such that its direction at any point is a principal
direction.
Hence we have n congruences of lines of curvature of a V n.
THEOREM 12.4 To show that the mean curvature of a hypersurface is equal to the negative of the
divergence of the unit normal.
or
To show that the first curvature of a hypersurface is equal to the negative of the divergence of the unit
normal.
or
To show that the normal curvature of a hypersurface for any direction is the negative of the tendency
of the unit normal in that direction.
Proof: Let N be the unit normal vector to the hypersurface V n in y's and let N be its covariant
components. Let t be the unit tangent vector to N congruences eh| (h = 1, 2, , n) of an orthogonal
ennuple in V n and let Th| be the contravariant components in V n + 1 of t. Since t is orthogonal to N,
therefore
Th| N = 0

(1)

## Taking covariant derivative of equation (1) with regard to y provides

Th| , N + Th| N , = 0
Multiplying equation (2) by Th , we get
Th| , Th| N + Th| Th| N , = 0

(2)

228

## (Th| ,Th| ) N = N , Th| Th|

(3)

Now Th|,Th| is the first curvature of the curve eh| and N , . Th| . Th| is the tendency of N is the
direction of eh| . Hence equation (3) implies that the normal component of the first curvature of the eh|
relative to V n + 1 or the normal curvature of V n in the direction of the curve eh|
= tendency of N in the direction of the curve eh|

(4)

## Taking summation of both of (3) and (4) for h = 1, 2, , n

we have
i.e., mean curvature or first curvature of a hypersurface = divergence of the unit normal.
Corollary: To prove that
M = divn + 1N
Proof: Since N is a vector of unit magnitude i.e., constant magnitude, its tendency is zero. also by
definition, the div n + 1 N and div n N differ only by the tendency of the vector N in its direction. But the
tendency of N is zero hence it follows that
divn N = div n + 1 N
Hence
M = div n N = div n + 1 N.
12.7 EULER'S THEOREM
Statement
The normal curvature Kn of V n for any direction of ar in V n is given by
n

Kn =

Kh cos2 h
h =1

where Kh are the principal curvature and h are the angles between direction of ar and the congruence
eh|
Proof: The principal directions in V n determined by the symmetric covariant tensor teser ij are
given by
(ij K h g ij ) p ih| = 0

(1)

ij Kgij = 0

(2)

## and phi | are the unit tangents to n congruences of lines of curvature.

The roots of the equation (2) are the maximum and minimum values of the quality Kn defined by
ij p i p j

Kn = g p i p i
ij

(3)

Hypersurface

229

## Multiplying equation (1) by p ki | , (K h), we get

(ij K n gij ) pih| p ik| = 0
The principal directions satisfy the equation

(h k )

ij pih | pkj| = 0

(4)

Let eh| be the unit tangents to the n congruences of lines of curvature. Then the principal curvatures
are given by
Kh = ij ehi | ehi |

(h = 1, 2, , n)

## Any other unit vector ar in V n is expressible in the form

r
a =

eh| cos h
h
n

or
where

ai

(5)

h=1

h|

cos h

(6)

r r
cos h| = a e h | = ai eh|i

ai being the contravariant components of a and h being the inclination of vector a to eh|. The
r
normal curvature of V n for the direction of a is given by
Kn = ij a ia j
from (6), we get

Kn = ij

h ,k =1
n

i
h|

cos h

j
k|

cos h

## (ij e ih| ekj | ) cos h cosk

(ij ehi |ehj| ) cos 2 h

h =1

2
Kn = Kh cos h
h =1

(7)

## This is a generalisation of Eulers Theorem.

12.8 CONJUGATE DIRECTIONS AND ASYMPTOTIC DIRECTIONS IN A HYPERSURFACE
r
The directions of two vector, ar and b , at a point in V n are said to be conjugate if
ij a ib j = 0

(1)

and two congruences of curves in the hypersurface are said to be conjugate if the directions of the two
curves through any point are conjugate.

230

## Tensors and Their Applications

A direction in V n which is self-conjugate is said to be asymptotic and the curves whose direction
are along asymptotic directions are called asymptotic lines.
Therefore the direction the vector ar at a point of V n be asymptotic if
ij a i a j = 0

(2)

## The asymptotic lines at a point of a hypersurface satisfies the differential equation

ij dxi dx j = 0

(3)

THEOREM 12.5 If a curve C in a hypersurface V n has any two of the following properties it has the
third
(i) it is a geodesic in the hypersurface Vn
(ii) it is a geodesic in the enveloping space Vn + 1
(iii) it is an asymptotic line in the hypersurface V n.
Proof: Let C be a curve in the hypersurface V n. The normal curvature Kn of the hypersurface V n in the
direction of C is given by
2
2
Ka2 = Kg + Kn

(1)

where Ka and Kg are the first curvatures of C relative to enveloping space V n + 1 and hypersurface V n
respectively.
Suppose C is a geodesic in the hypersurface V n [i.e., (i) holds] then Kg = 0.
If C is also a geodesic in the enveloping space V n + 1 [i.e., (ii) holds] then Ka = 0.
Now, using these values in equation (1), we have
Kn2 = 0

Kn = 0

## Implies that C is an asymptotic line is the hypersurface V n i.e., (iii) holds.

Hence we have proved that (i) and (ii) (iii)
Similarly we have proved that
(ii) and (iii) (i)
and (i) and (iii) (ii)
12.9 TENSOR DERIVATIVE OF THE UNIT NORMAL
The function y are invariants for transformations of the coordinates x i 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 x i.
i.e.,

y;i = y,i =

y
xi

(1)

The unit Normal N be the contravariant vector in the ys whose tensor derivative with respect
to the xs is
N;i =

N
+ N y, i
x i

(2)

Hypersurface

231

aN N = 1

Since

(3)

## Tensor derivative of this equation with respect to

xi

gives

aN N;i + a N;i N = 0

## Interchanging and in Ist term, we get

or

a N N ;i + a N ;i N = 0
a N N ;i + a N ;i N = 0
2 a N N ;i = 0
a N N ;i = 0

(4)

which shows that N;i is orthogonal to the normal and therefore tangential to the hypersurface.
Thus N;i can be expressed in terms to tangential vectors y,k to V n so that
N;i = Ai y,k

## where Aik is a mixed tensor of second order in V n to be determined.

Since unit normal N is orthogonal to tangential vector y,i in V n. Then
aN y,i = 0

## Taking tensor derivative with respect to x j, we get

a N ;j y,i + a N y,ij = 0
or

a N ;j y ,i + a N ij N = 0

or

a N ;j y ,i + ij (a N N ) = 0

## from equation (3),

since y,ij = ij N

a N N = 1

or

a N ;j y,i + ij = 0

or

g ki A kj + ij = 0

y y

= g ki
since a y, k y, i = a
k
i
x x

## Multiplying this equation by gim, we get

g ki g im A kj + ij g im = 0

(5)

232

## Tensors and Their Applications

k
im
m
=0
k A j + ij g

Aim + ij g im = 0
im
Aim = ij g

## Substituting this value in equation (5), we get

N;i = ij g y, k = ji g

ik

N;i = ij g

jk

jk

y,k

y ,k

(6)

## This is the required expression for the tensor derivative of N .

Theorem 12.6 The derived vector of the unit normal with respect to the enveloping space, along a
curve provided it be a line of curvature of the hypersurface.
Proof: Since the tensor derivative of N is
N;i = ij g

jk

y ,k

(1)

N;i ei = ij g

Then

jk

y,k ei

(2)

## The direction of N;i ei is identical with that of e i

N;i ei = y,i ei ,

Then

( is scalar constant)

## from (2), we have

ij g jk y,k ei = y,i e i
Multiplying both sides of this equation by a y,l
ij g jk e i (a y ,k y,l ) = e i (a y,i y ,l )
ij g jkei gkl = ei gil

since g il = a y ,i y ,l

ijilei = ei g il
il ei ei gil = 0

( il g il ) ei = 0

(l = 1, 2, , n)

This equation implies that the direction of e i is a principal direction for the symmetric tensor il
i.e., e i is a principal direction for the hypersurface V n. Hence by definition the curve C is a line of
curvature V n.

Hypersurface

233

## 12.10 THE EQUATION OF GAUSS AND CODAZZI

since we know that (pg. 86, equation 5)
p
Ai, jk Ai, kj = A p Rijk

(1)

where A i is a covariant tenser of rank one and difference of two tensers Ai, jk Ai,kj is a covariant
tenser of rank three.
It y,i are components of a covariant tensor of rank one in x 's. Then replacing A i by y,i in
equation (1), we get
p
y,ijk y,ikj = y,p Rijk
= y,p g ph Rhijk ,

p
Since Rijk
= g phRhijk

(2)

## where Rhijk are Riemann symbols for the tensor g ij

We know that

and

y,ij = ij N

(3)

N ,i = ij g ik y,k

(4)

## are Riemann symbols for the tensor a

Let R
and evaluated at points of the hypersurface
using equation (3) and (4), equation (2) becomes

## y,p g ph [ Rhijk (hj ik hk ij )] N (ij , k ik , j ) R

y,i y , j y,k

(5)

Multiplying equation (5) by a y,l and summed with respect to . Using the relations
a y ,i y,l = 0
and

g = N y,l = 0,

we get
Rlijk = (lj ik lk ij ) + R y ,l y,i y ,j y,k

(6)

## Multiplying (6) by a N and summing with respect to . Using relations

a N y ,l = 0
and

a N B = 0

we get ij ,k ik , j + R N y ,i y,j y ,k = 0

(7)

Hence, The equation (6) are generalisation of the Gauss Characteristic equation and equation
(7) of the Mainardi-Codazzi equations.

234

## 12.11 HYPERSURFACES WITH INDETERMINATE LINES OF CURVATURE

A point of a hypersurface at which the lines of curvature are indeterminate is called an Umbilical Point.
The lines of curvature may be indeterminate at every point of the hypersurface iff
ij = g ij

(8)

where is an invariant
The mean curvature M of such a hypersurface is given by
M = ij g ij = gij g ij = n

M
.
n

ij =

M
g ,
n ij

from (i)

## If all the geodesics of a hypersurface V n are also geodesics of an enveloping V n

hypersurface V n is called a totally geodesic hypersurface of the hypersurface V n + 1.

(9)
+ 1.

They

THEOREM 12.7 A totally geodesic hypersurface is a minimal hypersurface and its-lines of curvature
are indeterminate.
Proof: We know that
2
2
Ka2 = Kn + K g

## and a hypersurface is said to be minimal if

M=0
and the lines of curvature are indeterminate if
ij =

M
g ij
n

(1)
(2)

(3)

## If a hypersurface V n is totally geodesic then geodesics of V n are also geodesics of V n + 1.

i.e.,
Ka = 0 = k g
Now, from (1), we have
Kn = 0
But normal curvature Kn is zero for an asymptotic direction. Hence a hypersurface V n is totally
geodesic hypersurface iff the normal curvature Kn zero for all directions in V n and hence
ij = 0

M = ij g ij = 0
i.e., equation (2) is satisfied.
Hence, the totally geodesic hypersurface is minimal hypersurface.
In this case equation (3) are satisfied hence the lines of curvature are indeterminate.

Hypersurface

235

## 12.12 CENTRAL QUADRATIC HYPERSURFACE

Let xi be the cartesian in Euclidean space S n, so that the components gij of the fundamental tensor are
constants. Let yi be the Riemmannian coordinates. If a fixed point O is taken as a pole and s the distance
of any point P then Riemannian coordinates y i of P with pole O are given by
y i = s i
(1)
i
is unit tangent in the direction of OP.
Let aij be the components in the x s of a symmetric tensor of the rank two and evaluated at the
pole O. Then the equation
y i aij y

=1

(2)

## represents a central quadratic hypersurface

Substituting the value of equation (1) in equation (2), we get
si aij si = 1
iaiji =

(3)

s2

The equation (3) showing that the two values of s are equal in magnitude but opposite in sign.
The positive value of s given by equation (3) is the length of the radius of the quadric (2) for the
direction i .
THEOREM 12.8 The sum of the inverse squares of the radii of the quadric for n mutually orthogonal
directions at O is an invariant equal to a ij g ij .
Proof: If e ih| , (h = 1, 2, n) are the contravariant components of the unit tangents at O to the curves
of an orthogonal ennuple in Sn. The radius Sh relative to the direction e ih| is given by
aij ehi | ehj | =
n

or

h =1

(s h )2

1
s 2h
h

h =1

= a ij
n

h =1

(s h )2

eih | e hj|

h =1

= a ij g ij

Proved.

236

## Tensors and Their Applications

THEOREM 12.9 The equation of hyperplane of contact of the tangent hypercone with vertex at the
point Q (y ) .
Proof: Given (from equation 2, pg. 235)
aij y i y j = 1
Differentiating it
aij dy i y j + aij y i dy j = 0
or

aij dy i y j + aij y j dy i = 0
2 aij dy i y j = 0
aij dy i y j = 0

This shows that dy i is tangential to the quadric. Hence yj is normal to the quadric.
The tangent hyperplane at the point P ( y j) is given by
(Y i y i ) aij y j = 0
aij Y i y j = aij y i y j
aij Y i y j = 1

since aij y i y j = 1

(4)

## This equation represents the equation of tangent hyperplane at P ( y i) .

If the tangent hyperplane P ( y j) passes through the point Q (yi). Then we have
y i aij y j = 1

(5)

Thus all points of the hyperquadric, the tangent hyperplanes at which pass through Q lie on the
hyperplane (5) on which yi is the current point. This is the hyperplane of contact of the tangent
hypercone whose vertex is Q ( y i ) .
12.13 POLAR HYPERPLANE
The polar hyperplane of the point R (~y i ) with respect to quadric (2) is the locus of the vertices of the
hypercones which touch the hyperquadric along its intersections with hyperplanes through R. If Q ( y i )
is the vertex of such tangent hypercone, then R lies on the hyperplane of contact of Q so that
yi a ~
yj =1
ij

Consequently for all positions of the hyperplane through R, Q lies on the hyperplane
yi a ~
yj =1
ij

(6)

This is required equations of the polar hyperplane of R and R is the pole of this hyperplane.

Hypersurface

237

## 12.14 EVOLUTE OF A HYPERSURFACE IN EUCLIDEAN SPACE

Consider a hypersurface V n of Euclidean space Sn + 1 and let x i (i = 1, 2, n) be coordinates of an
arbitrary point P of V n whose components relative to Sn + 1 are

y ( = 1, 2,L n + 1)
Let N be a unit normal vector at P relative to Sn + 1 so that tensor derivate N becomes covariant
derivative.
So,

N ;i = N ,i = ij g jk y,k

(1)

y;ij = y,ij = ij N

(2)

n +1

and

gij =

=1

y,i y,j

(3)

Let P ( y ) be a point on the unit normal N such that distance of P from P is in the direction
of N such that
y = y + N
(4)
Suppose P undergoes a displacement dx i in V n then the corresponding displacement d y of P is
given by
d y = ( y,i + N ,i ) dx i + N d

(5)

## The vector ( y,i + N ,i ) dx i is tangential to V n whereas N d is a normal vector. Therefore if the

displacement of P ( y ) be along the normal to the hypersurface then we have
( y,i N ,i 0) dx i = 0

(6)

## Using equation (1) in equation (6), we get

( y,i g jk y,k ) dx i = 0
Multiplying it by y,i and summing with respect to , we get, using equation (3), as
( gil ij g jk glk ) d xi = 0
( gil ij lj ) d xi = 0
( g il il ) dx i = 0
( gij ij ) d xi = 0

(7)

This shows that the directions dxi given by equation (7) are principal directions of the hypersurface
where the roots of the equation |gij ij| = 0 are called principal radii of normal curvature. The
locus of P ( y ) satisfying the condition (4) is called evolute of the hypersurface V n of Sn + 1 where
is a root of (7). The evolute is also a hypersurface of Sn + 1.

238

## Tensors and Their Applications

12.15 HYPERSPHERE
The locus of a point in Sn which moves in such way that it is always at a fixed distance R from a fixed
point C (b ) is called a hypersphere of radius R and centre C. Therefore the equation of such a
hypersphere is given by
n

(y

=1

b )2 = R 2

(1)

THEOREM 12.10 The Riemannian curvatrure of a hypersphere of radius R is constant and equal to
1
.
R2
Proof: Let the hypersurface be a V n of Sn + 1 and let its centre be taken as origin of Euclidean
coordinates in Sn + 1. Then the hypersphere is given by

(y

= R 2,

( = 1, 2, , n + 1)

(2)

For the point in V n the y's are functions of the coordinates xi on the hypersphere.
Differentiating equation (2) with respect to x i, we get

y y,i

=0

## and again differentiating it with respect to

,j

y,i +

(3)
x j,

we get.

y,ij = 0

(4)

By Gauss formula,
y,ij = ij N
Using (5), equation (4) becomes
g ij +

y ij N

=0

(5)

(6)

From equation (3) it follows that y,i is perpendicular to y . But y,i is tangential to V m. Hence
y is normal to V n.. The equation (2) implies that the components of the unit vector N are given by
N

y
y = RN
R

g ij +

RN

or
or

g ij + R ij

N ij = 0

(N

) =0

g ij + R ij = 0

since ( N )2 = 1

(7)

240

## Tensors and Their Applications

2.

To prove that for a space Vn with positive constant Riemannian curvature K these exists sets of n + 1 real
coordinate y satisfying the condition
n +1

( y
=1

) = 1
K

where R 2 =

1
K

## Proof: Using (9) in equation (1), we get

(y ) 2

1
K

Proved.

EXAMPLE 2
Show that the directions of two lines of curvature at a point of a hypersurface are conjugate.
Solution
The principal direction e ih| are given by
ij ehi | ehj| = 0

(1)

## The shows that principal directions at a point of a hypersurface are conjugate.

Thus we say that two congruences of lines of curvature are conjugate.
EXAMPLE 3
Show that the normal curvature of hyper surface V n in an asymptotic direction vanishes.
Solution
Let us consider a curve C in a V n. If C is an asymptotic line then it satisfies the differential
equation
ij dxi dx j = 0
dxi dx j
=0
(1)
ds ds
Now the normal curvature Kn of the hypersurface V n in an asymptotic direction in a V n is given
i.e.,

ij

by
Kn = ij
i.e.,

Kn = 0

dxi dx j
ds ds
from (1)

EXAMPLE 4
To prove that if the polar hyperplane of the point R passes through a point P then that of P passes
through R.
Solution
Let P( y i ) and R ( y i ) be two points. The polar hyperplane of R ( y i ) is

242

## Tensors and Their Applications

11. Prove that the necessary and sufficient condition that system of hypersurface with unit normal N be
isothermic is that

( N div N N .N ) = 0.
12. Show that the normal curvature of a subspace in an asymptotic direction is zero.
13. If straight line through a point P in S n meets a hyperquadric in A and B and the polar hyperplane of
P in Q prove that P, Q are harmonic conjugates to A, B.
14. What are evolutes of a hypersurface in an Euclidean space. Show that the varieties 1 = constant in
evolute are parallel, having the curves of parameter 1 as orthogonal geodesics of V n .

INDEX
A
Absolute 131
Addition and Subtraction of Tensors 15
Associated tensor 43
Asymptotic directions 229

B
Bianchi identity 94
Binormal 137

C
Canonical congruence 213
Christoffels symbols 55
Completely skew-symmetric 111
Completely symmetric 111
Concept 188
Congruence of curves 49
Conjugate (or Reciprocal) Symmetric Tensor 25
Conjugate directions 229
Conjugate metric tensor 34
Conservative force field 144
Contraction 18
Contravariant tensor of rank r 14
Contravariant Tensor of rank two 9
Contravariant vector 7
Covariant tensor of rank s 14

## Covariant tensor of rank two 9

Covariant vector 7
Curl 76
Curl of congruence 211
Curvature 136
Curvature of Congruence 207
Curvature tensor 86

D
Degree of freedom 157
Dextral Index 1
Divergence 75
Divergence Theorem 161
Dummy index 1
Dupins Theorem 226

E
Einstein space 103
Einstein tensor 95
Einsteins Summation Convention 1
Eulers condition 171
Eulers Theorem 228
Evolute 237

F
First curvature 227
First curvature vector of curve 170

244

## First fundamental tensor 34

Free Index 2
Fundamental tensor 31
Fundamental theorem 199

G
Gauss Characteristic equation 233
Gausss formula 222
Gausss theorem 164
Generalised Krnecker delta 112
Generalized coordinates 157
Geodesic congruence 208
Geodesic coordinate system 175
Geodesics 171
Gradient 75
Greens Theorem 162

H
Hamiltons principle 153
Hypersphere 238
Hypersurface 48, 218

I
Inner product of two tensors 18
Integral of energy 155
Intrinsic derivative 131

K
Kinetic energy 144
Krnecker Delta 2

L
Lagrangean equation 148
Lagrangean function 148
Laplaces equation 167
Laplacian operator 80, 163
Length of a curve 42
Levi-civitas concept 188
Line element 31
Line of curvature 227

## Tensors and Their Applications

M
Magnitude of vector 44
Mainardi-Codazzi equations 233
Mean curvature 101
Meauniers Theorem 225
Metric Tensor 31
Minimal curve 42
Minimal Hypersurface 227, 234
mixed tensor of rank r + s 14
mixed tensor of rank two 9

N
Newtonian Laws 142
Normal congruence 209
n-ply orthogonal system of hypersurfaces 49
Null curve 42

O
Orthogonal Cartesian coordinates 120
Orthogonal ennuple 49
Osculating plane 136
Outer Product of Tensor 16

P
Parallel vector fields 134
Parallelism 188
Poissons equation 166
Polar hyperplane 236
Potential energy 145
Principal directions of the hypersurface 227
Principle normal vector 136
Principle normal curvatures 227
Principle of least action 156
Projective curvature tensor 104

Q
Quotient Law 24

R
Reciprocal base systems 122

Index

Relative Tensor 26
Ricci tensor 88
Riccis principal directions 102
Riccis coefficients of rotation 205
Riccis Theorem 71
Riemann curvature 96
Riemanns symbol 86
Riemann-Christoffel Tensor 85
Riemannian coordinates 177
Riemannian Geometry 31, 116
Riemannian Metric 31
Riemannian space 31

245

## Stokes Theorem 164

Straight line 140
Subscripts 1
Superscripts 1
Symmetric Tensors 20

T
Tensor density 26
Torsion 137
Totally geodesic hypersurface 234
Transformation of Coordinates 6

U
S
Scalar product of two vectors 44
Schurs theorem 100
Second fundamental tensors 34
Serret-Frenet formula 138
Simple Pendulum 152
Skew-Symmetric Tensor 20

Umbral 1
Unit principal normal 171

W
Weyl tensor 104
Work function W 145