Attribution Non-Commercial (BY-NC)

Просмотров: 2.2K

Attribution Non-Commercial (BY-NC)

- Goldstein H., Poole C., Safko J. - Classical Mechanics
- tensor analysis
- Tensor Calculus
- A Student’s Guide to Lagrangians and Hamiltonians
- Group Theory - The Application to Quantum Mechanics [Meijer-Bauer]
- Introduction to Tensor Calculus & Continuum Mechanics
- Vector and Tensor Analysis 1950
- 1948-Vector and Tensor Analysis
- - Differential Geometry With Applications To Mechanics And Physics
- Classical Dynamics
- 1107034264_Particle
- Tensor Analysis for Students of Physics and Engineering. Read 02.24.2008
- Quantum Field Theory 2013
- Tensor Analysis Intro
- Functional and Structural Tensor Analysis for Engineers
- Introduction to Vectors and Tensors
- Vector and Tensor Analysis
- Relativity an Introduction to Special and General Relativity
- Tensor Analysis
- Tensor

Вы находитесь на странице: 1из 262

intentionally left

blank

Copyright © 2006, New Age International (P) Ltd., Publishers

Published by New Age International (P) Ltd., Publishers

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 Islam’s book entitled “Tensors and Their

Applications. I know the author as a research scholar who has worked with me for several years. This

book is a humble step of efforts made by him to prove him to be a dedicated and striving teacher who

has worked relentlessly in this field.

This book fills the gap as methodology has been explained in a simple manner to enable students

to understand easily. This book will prove to be a complete book for the students in this field.

Ram Nivas

Professor,

Department of Mathematics and Astronomy,

Lucknow University,

Lucknow

This page

intentionally left

blank

PREFACE

‘Tensors’ were introduced by Professor Gregorio Ricci of University of Padua (Italy) in 1887

primarily as extension of vectors. A quantity having magnitude only is called Scalar and a quantity with

magnitude and direction both, called Vector. But certain quantities are associated with two or more

directions, such a quantity is called Tensor. The stress at a point of an elastic solid is an example of a

Tensor which depends on two directions one normal to the area and other that of the force on it.

Tensors have their applications to Riemannian Geometry, Mechanics, Elasticity, Theory of Relativity,

Electromagnetic Theory and many other disciplines of Science and Engineering.

This book has been presented in such a clear and easy way that the students will have no difficulty

in understanding it. The definitions, proofs of theorems, notes have been given in details.

The subject is taught at graduate/postgraduate level in almost all universities.

In the end, I wish to thank the publisher and the printer for their full co-operation in bringing out

the book in the present nice form.

Suggestions for further improvement of the book will be gratefully acknowledged.

This page

intentionally left

blank

CONTENTS

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

1.1. n-dimensional Space ............................................................................................... 1

1.2. Superscript and Subscript ....................................................................................... 1

1.3. The Einstein's Summation Convention ...................................................................... 1

1.4. Dummy Index ....................................................................................................... 1

1.5. Free Index ............................................................................................................. 2

1.6. Krönecker Delta ..................................................................................................... 2

Exercises ............................................................................................................... 5

2.1. Introduction .......................................................................................................... 6

2.2. Transformation of Coordinates ................................................................................ 6

2.3. Covariant and Contravariant Vectors ......................................................................... 7

2.4. Contravariant Tensor of Rank Two .......................................................................... 9

2.5. Covariant Tensor of Rank Two ................................................................................ 9

2.6. Mixed Tensor of Rank Two..................................................................................... 9

2.7. Tensor of Higher Order ......................................................................................... 14

2.8. Scalar or Invariant ................................................................................................ 15

2.9. Addition and Subtraction of Tensors ....................................................................... 15

2.10. Multiplication of Tensors (Outer Product of Tensors) ............................................... 16

2.11. Contraction of a Tensor ........................................................................................ 18

2.12. Inner Product of Two Tensors .............................................................................. 18

2.13. Symmetric Tensors .............................................................................................. 20

2.14. Skew-symmetric Tensor ....................................................................................... 20

2.15. Quotient Law ....................................................................................................... 24

xii Tensors and Their Applications

2.17. Relative Tensor .................................................................................................... 26

Examples ............................................................................................................ 26

Exercises ............................................................................................................. 29

3.1. The Metric Tensor ............................................................................................... 31

3.2. Conjugate Metric Tensor (Contravariant Tensor) ...................................................... 34

3.3. Length of a Curve ................................................................................................ 42

3.4. Associated Tensor ................................................................................................ 43

3.5. Magnitude of Vector ............................................................................................. 43

3.6. Scalar Product of Two Vectors .............................................................................. 44

3.7. Angle Between Two Vectors .................................................................................. 45

3.8. Angle Between Two Coordinate Curves .................................................................. 47

3.9. Hypersurface ....................................................................................................... 48

3.10. Angle Between Two Coordinate Hyper surface ........................................................ 48

3.11. n-Ply Orthogonal System of Hypersurfaces ............................................................. 49

3.12. Congruence of Curves .......................................................................................... 49

3.13. Orthogonal Ennuple .............................................................................................. 49

Examples ............................................................................................................ 52

Exercises ............................................................................................................. 54

4.1. Christoffel’s Symbol............................................................................................. 55

4.2. Transformtion of Christoffel’s Symbols .................................................................. 64

4.3. Covariant Differentiation of a Covariant Vector ........................................................ 67

4.4. Covariant Differentiation of a Contravariant Vector ................................................... 68

4.5. Covariant Differentiation of Tensors ....................................................................... 69

4.6. Ricci’s Theorem .................................................................................................. 71

4.7. Gradient, Divergence and Curl ............................................................................... 75

4.8. The Laplacian Operator ......................................................................................... 80

Exercises ............................................................................................................. 83

5.1. Riemann-Christoffel Tensor ................................................................................... 85

5.2. Ricci Tensor ........................................................................................................ 88

5.3. Covariant Riemann-Christoffel Tensor .................................................................... 89

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

5.5. Bianchi Identity .................................................................................................... 94

5.6. Einstein Tensor .................................................................................................... 95

5.7. Riemannian Curvature of V n .................................................................................. 96

Contents xiii

Curvature Tensor of V n ......................................................................................... 98

5.9. Schur’s Theorem ............................................................................................... 100

5.10. Mean Curvature ................................................................................................. 101

5.11. Ricci Principal Directions .................................................................................... 102

5.12. Einstein Space ................................................................................................... 103

5.13. Weyl Tensor or Projective Curvature Tensor ......................................................... 104

Examples .......................................................................................................... 106

Exercises ........................................................................................................... 109

Chapter–6 The e-systems and the Generalized Krönecker Deltas ................ 111-115

6.1. Completely Symmetric ......................................................................................... 111

6.2. Completely Skew-symmetric ................................................................................ 111

6.3. e-system ........................................................................................................... 112

6.4. Generalized Krönecker Delta ................................................................................ 112

6.5. Contraction of ä iá jkâ ã ............................................................................................ 114

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

7.1. Length of Arc .................................................................................................... 116

7.2. Curvilinear Coordinates in E 3 .............................................................................. 120

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

7.4. On The Meaning of Covariant Derivatives ............................................................. 127

7.5. Intrinsic Differentiation ....................................................................................... 131

7.6. Parallel Vector Fields ........................................................................................... 134

7.7. Geometry of Space Curves ................................................................................. 134

7.8. Serret-Frenet Formulae ....................................................................................... 138

7.9. Equations of A Straight Line ................................................................................ 140

Exercises ........................................................................................................... 141

8.1. Introduction ...................................................................................................... 142

8.2. Newtonian Laws ................................................................................................ 142

8.3. Equations of Motion of Particle ............................................................................ 143

8.4. Conservative Force Field ..................................................................................... 144

8.5. Lagrangean Equation of Motion ........................................................................... 146

8.6. Applications of Lagrangean Equations ................................................................... 152

8.7. Hamilton’s Principle ............................................................................................ 153

8.8. Integral Energy .................................................................................................. 155

8.9. Principle of Least Action ..................................................................................... 156

xiv Tensors and Their Applications

8.11. Lagrangean Equation of Generalized Coordinates ................................................... 158

8.12. Divergence Theorem, Green’s Theorem, Laplacian Operator and Stoke’s

Theorem in Tensor Notation ................................................................................ 161

8.13. Gauss’s Theorem ............................................................................................... 164

8.14. Poisson’s Equation ............................................................................................. 166

8.15. Solution of Poisson’s Equation ............................................................................. 167

Exercises ........................................................................................................... 169

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

9.2. Geodesics ......................................................................................................... 171

9.3. Euler’s Condition ............................................................................................... 171

9.4. Differential Equations of Geodesics ...................................................................... 173

9.5. Geodesic Coordinates ......................................................................................... 175

9.6. Riemannian Coordinates ...................................................................................... 177

9.7. Geodesic Form of a Line Element ........................................................................ 178

9.8. Geodesics in Euclidean Space .............................................................................. 181

Examples .......................................................................................................... 182

Exercises ........................................................................................................... 186

10.1. Parallelism of a Vector of Constant Magnitude (Levi-Civita’s Concept) ..................... 188

10.2. Parallelism of a Vector of Variable Magnitude ......................................................... 191

10.3. Subspace of Riemannian Manifold ........................................................................ 193

10.4. Parallelism in a Subspace .................................................................................... 196

10.5. Fundamental Theorem of Riemannian Geometry Statement ..................................... 199

Examples .......................................................................................................... 200

Exercises ........................................................................................................... 203

11.1. Ricci’s Coefficient of Rotation ............................................................................. 205

11.2. Reason for the Name “Coefficients of Rotation” .................................................... 206

11.3. Curvature of Congruence .................................................................................... 207

11.4. Geodesic Congruence ......................................................................................... 208

11.5. Normal Congruence ........................................................................................... 209

11.6. Curl of Congruence ............................................................................................ 211

11.7. Canonical Congruence ........................................................................................ 213

Examples .......................................................................................................... 215

Exercises ........................................................................................................... 217

Contents xv

12.1. Introduction ...................................................................................................... 218

12.2. Generalized Covariant Differentiation .................................................................... 219

12.3. Laws of Tensor Differentiation ............................................................................ 220

12.4. Gauss’s Formula ................................................................................................ 222

12.5. Curvature of a Curve in a Hypersurface and Normal Curvature, Meunier’s Theorem,

Dupin’s Theorem ............................................................................................... 224

12.6. Definitions ......................................................................................................... 227

12.7. Euler’s Theorem ................................................................................................ 228

12.8. Conjugate Directions and Asymptotic Directions in a Hypersurface.......................... 229

12.9. Tensor Derivative of Unit Normal......................................................................... 230

12.10. The Equation of Gauss and Codazzi ..................................................................... 233

12.11. Hypersurfaces with Indeterminate Lines of Curvature ............................................ 234

12.12. Central Quadratic Hypersurfaces .......................................................................... 235

12.13. Polar Hyperplane ................................................................................................ 236

12.14. Evolute of a Hypersurface in an Euclidean Space ................................................... 237

12.15. Hypersphere ......................................................................................................238

Exercises ........................................................................................................... 241

Index .................................................................................................................... 243-245

This page

intentionally left

blank

CHAPTER – 1

PRELIMINARIES

In three dimensional rectangular space, the coordinates of a point are (x, y, z). It is convenient to write

(x1, x2, x3) for (x, y, z). The coordinates of a point in four dimensional space are given by (x1, x2, x3, x4).

In general, the coordinates of a point in n-dimensional space are given by (x1, x2, x3,...., xn) such n-

dimensional space is denoted by V n.

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.

n

Consider the sum of the series S = a1 x1 + a2 x 2 + ... + a n x n = Σ a i x i . By using summation convention,

i =1

drop the sigma sign and write convention as

n

Σ ai x i = a i x i

i =1

“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.

Any index which is repeated in a given term is called a dummy index or dummy suffix. This is also called

Umbral or Dextral Index.

e.g. Consider the expression ai xi where i is dummy index; then

ai xi = a1 x1 + a 2 x 2 + ⋅ ⋅ ⋅ + an x n

2 Tensors and Their Applications

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

These two equations prove that

a ix i = a j x j

So, any dummy index can be replaced by any other index ranging the same numbers.

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.

The symbol δij , called Krönecker Delta (a German mathematician Leopold Krönecker, 1823-91 A.D.)

is defined by

1 if i = j

δij =

0 if i ≠ j

Similarly δij and δij are defined as

1 if i = j

δ ij = 0 if i ≠ j

1 if i = j

and δ ij =

0 if i ≠ j

Properties

1. If x1, x2, ... xn are independent coordinates, then

∂x i

= 0 if i ≠ j

∂x j

∂x i

= 1 if i = j

∂x j

This implies that

∂x i

= δij

∂x j

∂x i ∂x k

It is also written as = δ ij .

∂x k ∂x j

2. δ ii = δ11 + δ 22 + δ 33 + ⋅ ⋅ ⋅ + δ nn (by summation convention)

δ ii = 1 + 1+ 1 + ⋅ ⋅ ⋅ + 1

δ ii = n

3. a ij δkj = a ik

a 3 j δ 2j = a δ2 + a δ2 + a δ2 + ⋅ ⋅ ⋅ + a δ 2

31 1 32 2 33 3 3n n

Since (as j is dummy index)

Preliminaries 3

In general,

a ijδ kj = a i 1δ1k + a i 2 δ2k + a i 3 δ 3k + ⋅ ⋅ ⋅ + aik δkk + ⋅ ⋅ ⋅ + a in δ nk

ik

a ij δkj = a (as δ1k = δ2k = ⋅ ⋅ ⋅ = δnk = 0 and δ kk = 1)

i j

4. δ j δk = δik

EXAMPLE 1

dφ ∂φ dx1 ∂φ dx 2 ∂φ dx n

Write = 1 + 2 + ⋅⋅⋅ + n using summation convention.

dt ∂x dt ∂x dt ∂x dt

Solution

dφ ∂φ dx 1 ∂φ dx 2 ∂φ dx n

= + + ⋅ ⋅ ⋅ +

dt ∂x 1 dt ∂x 2 dt ∂x n dt

dφ ∂φ dx i

= i

dt ∂x dt

EXAMPLE 2

Expand: (i) aij xixj; (ii) glm gmp

Solution

aij x i x j = a1 j x x + a2 j x x + ⋅ ⋅ ⋅ + a nj x x

1 j 2 j n j

(i)

= a11 x1 x1 + a22 x 2 x 2 + ⋅ ⋅ ⋅ + ann x n x n

aij x i x j = a11 ( x1 )2 + a22 ( x 2 ) 2 + ⋅ ⋅ ⋅ + ann (x n )2

(as i and j are dummy indices)

EXAMPLE 3

If aij are constant and aij = aji, calculate:

∂ ∂

(i) ∂x (aij x i x j ) (ii) ∂x ∂x (aij xi x j )

k k l

Solution

∂ ∂

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

∂x k ∂xk

4 Tensors and Their Applications

∂x j ∂x

= a ij xi ∂ x + a ij x j ∂ x

i

k k

∂x j

= a ij xi δ jk + aij x j δ ik , as = δ jk

∂x k

= (aij δ jk ) xi + (aij δ ik ) x j

= aik xi + akj xj as aij δjk = aik

= aik xi + aki xi as j is dummy index

∂(aij x i x j )

∂ xk = 2aik xi as given aik = aki

∂(aij x i x j )

(ii) = 2aikxi

∂x k

Differentiating it w.r.t. xl :

∂ 2 (aij xi x j ) ∂xi

= 2 aik

∂xk ∂xl ∂xl

= 2 aik δ il

∂ 2 (aij xi x j )

= 2alk as a ik δil = alk.

∂xk ∂xl

EXAMPLE 4

If aij xi xj= 0

where aij are constant then show that

aij + aji = 0

Solution

Given

aijx ix j= 0

⇒ almxlxm= 0 since i and j are dummy indices

Differentiating it w.r.t. xi partially,

∂

(alm x l x m ) = 0

∂xi

∂ l m

a lm (x x ) = 0

∂xi

∂x l m ∂x m l

a lm x + alm x =0

∂xi ∂x i

∂x l ∂x m

Since = δ li and = δmi

∂xi ∂xi

Preliminaries 5

almδ il x m + alm δ im x l = 0

aim x m + ali x l = 0

as alm δli = aim and alm δmi = ali .

Differentiating it w.r.t. xj partially

∂x m ∂x l

aim + ali

∂x j ∂x j = 0

aim δmj + ali δ lj = 0

aij + aji= 0 Proved.

EXERCISES

(i) (x 1 )2 + (x2 )2 + (x3 )2 + . . . + (xn )2

(ii) ds2 = g 11 (dx1 )2 + g 22 (dx2 )2 + . . . + g nn (dxn )2

(iii) a 1 x1 x3 + a 2 x2 x3 + . . . + a n xn x3

2. Expand the following:

∂ i

(i) a ij xj (ii) i

( ga ) (iii) Aik Bi

∂x

3. Evaluate:

(i) x j δ ij (ii) δ ij δkj δkl (iii) δ ij δij

ANSWERS

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

1

∂ 1 ∂ 2 ∂ n

(ii) 1

( ga ) + 2

( g a ) + ⋅⋅⋅ + n

( ga )

∂x ∂x ∂x

4. Cij xi xj

CHAPTER – 2

TENSOR ALGEBRA

2.1 INTRODUCTION

A scalar (density, pressure, temperature, etc.) is a quantity whose specification (in any coordinate

system) requires just one number. On the other hand, a vector (displacement, acceleration, force, etc.)

is a quantity whose specification requires three numbers, namely its components with respect to some

basis. Scalers and vectors are both special cases of a more general object called a tensor of order

n whose specification in any coordinate system requires 3n numbers, called the components of tensor.

In fact, scalars are tensors of order zero with 3° = 1 component. Vectors are tensors of order one with

31 = 3 components.

In three dimensional rectangular space, the coordinates of a point are (x, y, z) where x, y, z are real

numbers. It is convenient to write (x1, x2, x3) for ( x, y , z ) or simply xi where i = 1, 2, 3. Similarly in

n- dimensional space, the coordinate of a point are n-independent variables (x1, x2,..., x n) in X-coordinate

system. Let ( x 1 , x 2 ,..., x n ) be coordinate of the same point in Y-coordinate system.

Let x 1 , x 2 , …, x n be independent single valued function of x1, x2,...., xn, so that,

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

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

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

M M

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

or

x i = x i ( x1 , x 2 , ..., x n ) ; i = 1, 2, …, n …(1)

Tensor Algebra 7

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

Let (x1, x2, ..., xn) or x i be coordinates of a point in X-coordinate system and ( x 1 , x 2 ,..., x n ) or x i be

coordinates of the same point in the Y-coordinate system.

Let A i, i = 1, 2, ..., n (or A 1, A 2, ..., A n) be n functions of coordinates x 1, x2, ..., x n

in X-coordinate system. If the quantities A i are transformed to A i in Y-coordinate system then according

to the law of transformation

∂x i j ∂x j

Ai = A or A j

= i

∂x j ∂x i A

Then A i are called components of contravariant vector.

Let Ai , i = 1, 2,..., n (or A 1, A 2, …, A n) be n functions of the coordinates x1, x2, ..., x n

in X-coordinate system. If the quantities Ai are transformed to Ai in Y-coordinate system then

according to the law of transformation

∂x j ∂x i

Ai =

i

Aj or Aj = Ai

∂x ∂x j

Then A i are called components of covariant vector.

The contravariant (or covariant) vector is also called a contravariant (or covariant) tensor of rank

one.

Note: A superscript is always used to indicate contravariant component and a subscript is always used to indicate

covariant component.

EXAMPLE 1

If xi be the coordinate of a point in n-dimensional space show that dxi are component of a

contravariant vector.

Solution

1 2 n

Let x1, x2, ..., xn or xi are coordinates in X-coordinate system and x , x ,..., x or x i are

coordinates in Y-coordinate system.

If

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

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

i i i

1

∂x 2

∂x n

∂x

8 Tensors and Their Applications

∂x i

d xi = dx j

∂x j

It is law of transformation of contravariant vector. So, dx i are components of a contravariant

vector.

EXAMPLE 2

∂φ

Show that is a covariant vector where φ is a scalar function.

∂x i

Solution

Let x1, x2, ..., xn or x i are coordinates in X-coordinate system and x 1 , x 2 , ..., x n or x i are

coordinates in Y-coordinate system.

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

∂φ 1 ∂φ 2 ∂φ

∂φ = ∂x + 2 ∂x + ⋅ ⋅ ⋅ + n ∂x n

∂x 1

∂x ∂x

∂φ ∂φ ∂x1 ∂φ ∂x 2 ∂φ ∂x n

= + + ⋅ ⋅ ⋅ +

∂x i

∂x 1 ∂x i ∂x 2 ∂x i ∂x n ∂x i

∂φ ∂φ ∂x j

=

∂x i ∂x j ∂x i

∂φ ∂x j ∂φ

or =

∂x i ∂x i ∂x j

∂φ

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

∂x i

vector.

EXAMPLE 3

Show that the velocity of fluid at any point is a component of contravariant vector

or

Show that the component of tangent vector on the curve in n-dimensional space are component

of contravariant vector.

Solution

dx1 dx 2 dx n

Let , , …, be the component of the tangent vector of the point ( x1 , x 2 ,..., x n ) i.e.,

dt dt dt

dx i

be the component of the tangent vector in X-coordinate system. Let the component of tangent

dt

Tensor Algebra 9

dx i

vector of the point ( x 1 , x 2 ,..., x n ) in Y-coordinate system are . Then x 1 , x 2 , ..., x n or x i being a

dt

function of x1 , x 2 , ..., x n which is a function of t. So,

d xi ∂x i dx1 ∂x i dx 2 ∂x i dx n

= + + ⋅ ⋅ ⋅ +

dt dt1 dt dx 2 dt dx n dt

d xi ∂x i dx j

=

dt dx j dt

dx i

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

dt

contravariant vector.

i.e. the component of tangent vector on the curve in n-dimensional space are component of

contravariant vector.

Let A ij (i, j = 1, 2, ..., n) be n2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the

quantities A ij are transformed to A ij in Y-coordinate system having coordinates x 1 , x 2 , ..., x n . Then

according to the law of transformation

∂x i ∂x j kl

A ij = A

∂x k ∂x l

ij

Then A are called components of Contravariant Tensor of rank two.

Let A ij (i, j = 1, 2, ..., n) be n 2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the

quantities Aij are transformed to Aij in Y-coordinate system having coordinates x 1 , x 2 ,...,x n , then

according to the law of transformation,

∂x k ∂x l

Aij = Akl

∂x i ∂x j

Then A ij called components of covariant tensor of rank two.

Let Aij (i, j = 1, 2, ..., n) be n2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the

quantities Aij are transformed to A ji in Y-coordinate system having coordinates x 1 , x 2 ,..., x n , then

according to the law of transformation

∂x i ∂x l k

= A ji Al

∂x k ∂x

Then Aij are called components of mixed tensor of rank two.

10 Tensors and Their Applications

Note: (i) The rank of the tensor is defined as the total number of indices per component.

(ii) Instead of saying that “ A ij are the components of a tensor of rank two” we shall often say “ Aij is a tensor

of rank two.”

THEOREM 2.1 To show that the Krönecker delta is a mixed tensor of rank two.

Solution

Let X and Y be two coordinate systems. Let the component of Kronecker delta in X-coordinate

system δij and component of Krönecker delta in Y-coordinate be δij , then according to the law of

transformation

∂x i ∂x i ∂x l ∂x k

δij = =

∂x j ∂x k ∂x j ∂x l

∂x i ∂x l k

δij = δl

∂x k ∂x j

This shows that Krönecker δij is mixed tensor of rank two.

EXAMPLE 4

∂Ai

If Ai is a covariant tensor, then prove that do not form a tensor..

∂x j

Solution

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

∂x k

Ai = Ak

∂x i

Differentiating it w.r.t. x j

∂Ai ∂ ∂x k

∂x j =

∂x j ∂x i Ak

∂Ai ∂x k ∂Ak ∂ 2 xk

= + A …(1)

∂x j ∂x i ∂x j

k

∂x i ∂x j

∂Ai

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

∂x j

tensor.

THEOREM 2.2 To show that δij is an invariant i.e., it has same components in every coordinate

system.

Proof: Since δij is a mixed tensor of rank two, then

∂x i ∂x l k

δij = δl

∂x k ∂x j

Tensor Algebra 11

∂x i ∂x l k

=

∂x k ∂x j δ l

∂ x i ∂x k ∂x l k ∂x k

= , as δl =

∂x ∂ x

k j

∂x j ∂x j

∂x i ∂x i

δij = = δ i

, as = δij

∂x j

j

∂x j

So, δij is an invariant.

or

Prove that the transformation of a contravariant vector form a group.

Proof: Let Ai be a contravariant vector in a coordinate system x i (i = 1, 2,...,n) . Let the coordinates

When coordinate x i be transformed to x i , the law of transformation of a contravariant vector is

∂x p q

Ap = A ... (1)

∂x q

When coordinate x i be transformed to x i , the law of transformation of contravariant vector is

∂x i p

Ai = A

∂x p

∂x i ∂x p q

A i = ∂x i ∂x q A from (1)

∂x i q

A i = A

∂x q

This shows that if we make direct transformation from x i to x i , we get same law of transformation.

This property is called that transformation of contravariant vectors is transitive or form a group.

or

Prove that the transformation of a covariant vector form a group.

Proof: Let Ai be a covariant vector in a coordinate system x i (i = 1, 2, ..., n) . Let the coordinates x i be

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

When coordinate x i be transformed to x i , the law of transformation of a covariant vector is

∂x q

Ap = Aq ... (1)

∂x p

When coordinate x i be transformed to x i , the law of transformation of a covariant vector is

∂x p

Ai = Ap

∂x i

12 Tensors and Their Applications

∂x p ∂x q

Ai = Aq

∂x i ∂x p

∂x q

Ai = Aq

∂x i

This shows that if we make direct transformation from x i to x i , we get same law of transformation.

This property is called that transformation of covariant vectors is transitive or form a group.

or

Prove that the equations of transformation a tensor (Mixed tensor) posses the group property.

Proof: Let Aij be a mixed tensor of rank two in a coordinate system x i (i = 1, 2,...,n) . Let the coordinates

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

When coordinate x i be transformed to x i , the transformation of a mixed tensor of rank two is

∂x p ∂x s r

Aqp = As ... (1)

∂x r ∂x q

When coordinate x i be transformed to x i , the law of transformation of a mixed tensor of rank

two is

∂x i ∂x q p

A ji = Aq

∂x p ∂x j

∂x ∂x ∂x ∂x

i q p s

r

= p j r q

As from (1)

∂x ∂x ∂x ∂x

∂x i ∂x s r

A ji = As

∂x r ∂x j

This shows that if we make direct transformation from x i to x i , we get same law of transformation.

This property is called that transformation of tensors form a group.

THEOREM 2.6 There is no distinction between contravariant and covariant vectors when we restrict

ourselves to rectangular Cartesian transformation of coordinates.

Proof: Let P(x, y) be a point with respect to the rectangular Cartesian axes X and Y. Let ( x , y ) be the

coordinate of the same point P in another rectangular cartesian axes X and Y , Let (l1, m1) and (l2, m2)

be the direction cosines of the axes X , Y respectively. Then the transformation relations are given by

x = l1 x + m1 y

...(1)

y = l2 x + m2 y

and solving these equations, we have

x = l1 x + l 2 y

...(2)

y = m1 x + m2 y

put x = x1 , y = x 2 , x = x 1 , y = x 2

Tensor Algebra 13

∂x i j

A i = A ; j = 1,2

∂x j

∂x i 1 ∂x i 2

A i = A + 2A

∂x1 ∂x

for i = 1, 2 .

∂x 1 1 ∂x 1 2

A 1 = A + 2A

∂x1 ∂x

∂x 1 ∂x 2 2

2

A 2 = A + 2 A

∂x1 ∂x

∂x

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

2 2

From (1)

∂x 1

Then

∂x ∂x 1

= =l .

∂x ∂x 1 1

Similarly,

∂x ∂x 1

= m1 = 2 ;

∂y ∂x

∂y ∂x 2

∂y ∂x

= m2 = 2

2 ...(3)

= l2 = 1 ;

∂x ∂x ∂y ∂x

So, we have

A 1 = l1 A1 + m1 A 2

..(4)

A 2 = l2 A1 + m 2 A 2

Consider the covariant transformation

∂x j

Ai = A j ; j = 1,2

∂x i

∂x1 ∂x 2

Ai = A + A2

∂x i

1

∂x1

for i = 1, 2 .

∂x1 ∂x 2

A1 = A + A2

∂x 1

1

∂x 1

∂x 1 ∂x 2

A2 = A + A2

∂x 2

1

∂x 2

From (3)

A1 = l1 A1 + m1 A2

...(5)

A2 = l2 A1 + m2 A2

14 Tensors and Their Applications

A 1 = A1 and A 2 = A2

Hence the theorem is proved.

(a) Contravariant tensor of rank r

i i ...i

Let Ai1i 2 ...ir be nr function of coordinates x1, x2, ..., xn in X-coordinates system. If the quantities A i 2 r

are transformed to A i1i 2 ...i r in Y-coordinate system having coordinates x 1 , x 2 , ..., x n . Then according

to the law of transformation

∂x i1 ∂x i2 ∂x i r

A i1i 2 ...i r

= K

p1 p 2 ...p r

∂x p1 ∂x p 2 ∂x p r A

Then Ai ii 2 ...i r are called components of contravariant tensor of rank r.

(b) Covariant tensor of rank s

Let A j1 j 2 ... j s be n s functions of coordinates x1, x2, ..., xn in X-coordinate system. If the quantities

A j1 j 2 ... j s are transformed to A j1 j 2 ... j s in Y- coordinate system having coordinates x 1 , x 2 ,...,x n . Then

according to the law of transformation

∂x q1 ∂x q 2 ∂x q s

A j1 j 2 ... j s = ⋅ ⋅ ⋅ Aq1 ,q 2 ...,q s

∂x j1 ∂x j 2 ∂x j s

Then A j1 j 2 ... j s are called the components of covariant tensor of rank s.

i i ...i

Let A j11 2j 2 ...rj s be nr+s functions of coordinates x1, x2, ..., xn in X-coordinate system. If the quantities

Aij11i 2j 2......i rj s are transformed to A ij11ij22......i rj s in Y-coordinate system having coordinates x 1 , x 2 , … , x n . Then

according to the law of transformation

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

A ij11ij22......i rj s = ⋅ ⋅ ⋅ Aqp11qp22......q ps r

∂x p1 ∂x p2 ∂x p r ∂x ∂x

j1 j2

∂x js

i i ...i

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

2 ... j s

ir is known as tensor of type (r, s ) , In (r,s), the first component r

indicates the rank of contravariant tensor and the second component s indicates the rank of covariant

tensor.

ij

Thus the tensors Aij and A are type (0, 2) and (2, 0) respectively while tensor Aij is type (1, 1).

Tensor Algebra 15

EXAMPLE

ijk

Alm is a mixed tensor of type (3, 2) in which contravariant tensor of rank three and covariant

tensor of rank two. Then according to the law of transformation

∂x i ∂x j ∂x k ∂x a ∂x b αβγ

Almijk = Aab

∂x α ∂x β ∂xγ ∂x l ∂x m

A function φ ( x1 , x 2 , ..., x n ) is called Scalar or an invariant if its original value does not change upon

transform ation of coordinates from x1, x2, ..., xn to x 1 , x 2 , ..., x n . i.e.

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

Scalar is also called tensor of rank zero.

For example, Ai Bi is scalar..

THEOREM 2.7 The sum (or difference) of two tensors which have same number of covariant and the

same contravariant indices is again a tensor of the same rank and type as the given tensors.

Proof: Consider two tensors Aij11i 2j 2......i rj s and B ij11i 2j 2......i rj s of the same rank and type (i.e., covariant tensor of

rank s and contravariant tensor of rank r.). Then according to the law of transformation

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

A ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Aqp11qp22......q ps r

∂x ∂x

p1 p2

∂x ∂x ∂x

pr j1 j2

∂x js

and

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

B ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Bqp11qp22......q ps r

∂x p1 ∂x p2 ∂x p r ∂x j1 ∂x j2 ∂x j s

Then

∂x i1 ∂x i 2

∂x p1 ∂x p2

⋅ ⋅ ⋅

∂x i r ∂x q1 ∂x q 2 … ∂x q s

∂x p r ∂x j1 ∂x j 2 ∂x j s

(A p1 p2 ...p r

q1q 2 ...q s

± B qp11qp2 2......qps r )

If

A ij11ij22......i rj s ± B ji11ij22......irj s = C ij11ij22......i rj s

and

Aqp11qp2 2......q ps r ± Bqp11qp2 2......qps r = C qp11qp2 2......qps r

So,

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

⋅ ⋅ ⋅

C ij11ij22......i rj s = ⋅ ⋅ ⋅ Cqp11,,qp22,...,

,..., pr

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

i , i ,..., i

This is law of transformation of a mixed tensor of rank r+s. So, C j11 , j22 ,..., jrs is a mixed tensor of

rank r+s or of type (r, s).

16 Tensors and Their Applications

EXAMPLE 5

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

Solution

As given Akij and Bkij are tensors. Then according to the law of transformation

∂x i ∂x j ∂x r pq

Akij = Ar

∂x p ∂x q ∂x k

and

∂x i ∂x j ∂x r pq

Bkij = Br

∂x p ∂x q ∂x k

then

Akij ± Bkij =

∂x i ∂x j ∂x r pq

∂x ∂x ∂x

p q k

(

Ar ± Brpq )

If

Akij ± B kij = C kij and Arpq ± B rpq = C rpq

So,

∂x i ∂x j ∂x r pq

Cr

C kij =

∂x p ∂x q ∂x k

The shows that C kij is a tensor of same rank and type as Akij and Bkij .

THEOREM 2.8 The multiplication of two tensors is a tensor whose rank is the sum of the ranks of

two tensors.

i i ...i

Proof: Consider two tensors A j11 2j 2 ...rj s (which is covariant tensor of rank s and contravariant tensor of

k k2 ...km

rank r) and Bl11l 2 ...

ln (which is covariant tensor of rank m and contravariant tensor of rank n). Then

according to the law of transformation.

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

A ij11ij22......i rj s = ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Aqp11qp22......q ps r

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

and

∂x k1 ∂x k 2 ∂x km ∂xβ1 ∂x β2 ∂x βn

Bl1kl12k...2 ...km

= ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ Bβα11βα2 2......βαn m

∂x α1 ∂x α 2 ∂x α m ∂x l1 ∂x l 2 ∂x l n

ln

∂x i1 ∂x i r ∂x q1 ∂ x q s ∂ x k1 ∂ x km ∂ x β 1 ∂x βn

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

∂x p1 ∂x pr ∂x j1 ∂x j s ∂x α1 ∂x α m ∂x l1 ∂x l m

Tensor Algebra 17

If

C ij1ij2 ......i rjk1l kl 2 ...

...km i i ...i k k ...k

1 2 s 1 2 ln

= A j11 j22 ...rjs Bl1l12 ...2 ln n

and

C qp11qp2 2......qpsβr α1β12α...2β...nαm = Aqp11qp22...,

...,p r

qs

Bβα11βα2 2......βαn m

So,

∂x i1 ∂x ir ∂x q 1 ∂x q s ∂x k1

C ji11ij22......irj ksl11kl 22 ...

...km

ln =

⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅

∂x p1 ∂x p r ∂x j1 ∂x j s ∂x α1

∂x k m ∂x β1 ∂ x βn

⋅ ⋅ ⋅ α ⋅ l ⋅ ⋅ ⋅ l C qp11qp2 2......q ps βrα1 β1α2 2......βαn h

∂x m ∂x 1 ∂x n

i i ...i ...k

This is law of transformation of a mixed tensor of rank r + m + s + n. . So, C j11 j22 ...rj ks1lk12l 2 ...lmn is a

open proudct of two tensors.

THEOREM 2.9 If A i and B j are the components of a contravariant and covariant tensors of rank one

then prove that A iB j are components of a mixed tensor of rank two.

Proof: As Ai is contravariant tensor of rank one and B j is covariant tensor of rank one. Then

according to the law of transformation

∂x i k

Ai = A ...(1)

∂x k

and

∂x l

Bj = Bl ...(2)

∂x j

Multiply (1) and (2), we get

∂x i ∂x l k

Ai B j = A Bl

∂x k ∂x j

This is law of transformation of tensor of rank two. So, Ai B j are mixed tensor of rank two.

Such product is called outer product of two tensors.

EXAMPLE 6

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

Solution

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

∂x i ∂x q p ∂x k ∂x l ∂x t rs

A ji = Aq and Bmkl = Bt

∂x p ∂x j ∂x r ∂x s ∂x m

18 Tensors and Their Applications

∂x i ∂x q ∂x k ∂x l ∂x t p rs

A ji Bmkl = A B

∂x p ∂x j ∂x r ∂x s ∂x m q t

This is law of transformation of tensor of rank five. So, Aij B mkl is a tensor of rank five.

The process of getting a tensor of lower order (reduced by 2) by putting a covariant index equal to a

contravariant index and performing the summation indicated is known as Contraction.

In other words, if in a tensor we put one contravariant and one covariant indices equal, the

process is called contraction of a tensor.

ijk

For example, consider a mixed tensor Alm of order five. Then by law of transformation,

∂x i ∂x j ∂x k ∂x s ∂x t pqr

Almijk = Ast

∂x p ∂x q ∂x r ∂x l ∂x m

Put the covariant index l = contravariant index i, so that

∂x i ∂x j ∂x k ∂x s ∂x t pqr

Aimijk = Ast

∂x p ∂x q ∂x r ∂x i ∂x m

∂x j ∂x k ∂x s ∂xt pqr

= Ast

∂x q ∂x r ∂x p ∂x m

∂x j ∂x k ∂x t s pqr ∂x s

= δ p Ast Since = δ sp

∂x q ∂x r ∂x m ∂x p

∂x j ∂x k ∂x t pqr

Aimijk = A pt

∂x q ∂x r ∂x m

ijk

This is law of transformation of tensor of rank 3. So, Aim is a tensor of rank 3 and type (1, 2)

ijk

while Alm is a tensor of rank 5 and type (2, 3). It means that contraction reduces rank of tensor by

two.

Consider the tensors Akij and Bmn

l

if we first form their outer product Akij B mn

l

and contract this by

putting l = k then the result is Akij B mn

k

which is also a tensor, called the inner product of the given

tensors.

Hence the inner product of two tensors is obtained by first taking outer product and then

contracting it.

EXAMPLE 7

If A i and B i are the components of a contravariant and covariant tensors of rank are respectively

then prove that A iB i is scalar or invariant.

Tensor Algebra 19

Solution

As A i and B i are the components of a contravariant and covariant tensor of rank one respectively,

then according to the law of the transformation

∂x i p ∂x q

Ai= A and Bi = Bq

∂x p ∂x i

Multiplying these, we get

∂x i ∂x q p

ABi= i A Bq

∂x p ∂x i

∂x q p ∂x q

= A B , since = δ qp

∂x p

q

∂x p

= δ qp A p Bq

p

A i Bi = A B p

This shows that A iB i is scalar or Invariant.

EXAMPLE 8

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

tensor of rank 3.

Solution

As Aij is mixed tensor of rank 2 and B mkl is mixed tensor of rank 3. Then by law of transformation

∂x i ∂x q p ∂x k ∂x l ∂x t rs

A ji = A kl

q and Bm = Bt ...(1)

∂x p ∂x j ∂x r ∂x s ∂x m

Put k = j then

∂x j ∂x l ∂x t rs

Bmjl = Bt ...(2)

∂x r ∂x s ∂x m

Multiplying (1) & (2) we get

∂x i ∂x q ∂x j ∂x l ∂x t p rs

A ij B mjl = Aq Bt

∂x p ∂x j ∂x r ∂x s ∂x m

∂x i ∂x l ∂x t q p rs ∂x q ∂x j ∂x q

= δr Aq Bt since = = δ qr

∂x p ∂x s ∂x m ∂x j ∂x r ∂x r

∂x i ∂x l ∂x t p qs

A ji Bmjl = Aq B t since δ qr Btrs = Btqs

∂x p ∂x s ∂x m

This is the law of transformation of a mixed tensor of rank three. Hence Aij B mjl is a mixed tensor

of rank three.

20 Tensors and Their Applications

A tensor is said to be symmetric with respect to two contravariant (or two covariant) indices if its

components remain unchanged on an interchange of the two indices.

EXAMPLE

(1) The tensor A ij is symmetric if A ij = A ji

ijk

(2) The tensor Alm ijk

is symmetric if Alm = Almjik

1

THEOREM 2.10 A symmetric tensor of rank two has only n(n + 1) different components in n-

2

dimensional space.

Proof: Let A ij be a symmetric tensor of rank two. So that Aij = A ji .

21

A A 22 A 23 L A2n

The component of A ij are A31 A32 A33 L A 3n

M M M L M

A n1

An2 An3 L A nn

i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn are

different. Thus remaining components are (n2– n). In which A 12 = A 21, A 23 = A 32 etc. due to symmetry.

1 2

So, the remaining different components are (n − n) . Hence the total number of different

2

components

1 2 1

= n+ (n − n) = n(n + 1)

2 2

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

ijk

(ii) The tensor Alm ijk

is Skew-symmetric if Alm = − Almjik

1

THEOREM 2.11 A Skew symmetric tensor of second order has only n(n − 1) different non-zeroo

2

components.

Proof: Let A ij be a skew-symmetric tensor of order two. Then Aij = − A ji .

Tensor Algebra 21

21

A 0 A 23 L A2n

The components of A are A 31

ij

A 32 0 L A3n

M M M L M

A n1

An 2 An3 L 0

[Since A ii

= − Aii ⇒ 2 A ii = 0 ⇒ Aii = 0 ⇒ A11 = A 22 = ⋅ ⋅ ⋅ A nn = 0]

i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn

are zero. Omitting there, then the remaining components are n2 –n. In which A 12 = – A 21, A 13 = – A 31

1 2

etc. Ignoring the sign. Their remaining the different components are (n − n) .

2

1

Hence the total number of different non-zero components = n(n − 1)

2

Note: Skew-symmetric tensor is also called anti-symmetric tensor.

THEOREM 2.12 A covariant or contravariant tensor of rank two say Aij can always be written as the

sum of a symmetric and skew-symmetric tensor.

Proof: Consider a covariant tensor A ij. We can write A ij as

1 1

Aij = ( A + A ji ) + ( Aij − A ji )

2 ij 2

Aij = Sij + Tij

1 1

where Sij = ( Aij + A ji ) and Tij = ( Aij − A ji )

2 2

Now,

1

S ji = ( A + Aij )

2 ji

S ji = Sij

So, Sij is symmetric tensor..

and

1

Tij = ( A + A ji )

2 ij

1

T ji = ( A − Aij )

2 ji

1

= − ( Aij − A ji )

2

T ji = −Tij

or Tij = −T ji

So, Tij is Skew-symmetric Tensor..

22 Tensors and Their Applications

EXAMPLE 9

j k

Solution

As given

φ = a jk A j A k ...(1)

Interchange the indices i and j

φ = ak j A k A j ...(2)

2φ = (a jk + akj ) A j A k

1

φ = (a jk + akj ) A j A k

2

φ = b jk A j A k

1

where b jk = (a + akj )

2 jk

To show that b jk is symmetric.

Since

1

b jk = (a + akj )

2 jk

1

bkj = (a kj + a jk )

2

1

= (a jk + akj )

2

bkj = b jk

So, b jk is Symmetric.

EXAMPLE 10

∂Ti ∂T j

If Ti be the component of a covariant vector show that j − i are component of a Skew-

∂x ∂x

symmetric covariant tensor of rank two.

Solution

As Ti is covariant vector. Then by the law of transformation

∂x k

Ti = Tk

∂x i

Tensor Algebra 23

∂Ti ∂ ∂x k

j = ∂x j ∂x i Tk

∂x

∂2 x k ∂x k ∂Tk

= T +

∂x j ∂x i

k

∂x i ∂x j

∂T j ∂ 2 xk ∂x k ∂x l ∂Tk

= T + ...(1)

∂x j ∂x j ∂x i

k

∂x i ∂x j ∂x l

Similarly,

∂T j

∂ 2 xk ∂x k ∂x l ∂Tk

= T +

∂x i ∂x i ∂x j

k

∂x j ∂x i ∂x l

Interchanging the dummy indices k & l

∂T j ∂2 x k ∂x k ∂x l ∂Tl

= T + ...(2)

∂x i ∂x i ∂x j

k

∂x i ∂x j ∂x k

Substituting (1) and (2), we get

− = −

∂x j ∂x i ∂x i ∂x j ∂x l ∂x k

∂Ti ∂T j

This is law of transformation of covariant tensor of rank two. So, − are component of

∂x j ∂x i

a covariant tensor of rank two.

∂Ti ∂T j

To show that − is Skew-symmetric tensor..

∂x j ∂x i

Let

∂Ti ∂T j

Aij = −

∂x j ∂x i

∂T j ∂Ti

A ji = −

∂xi ∂x j

∂T ∂T

= − ∂ x j − ∂x i

i j

A ji = − Aij

or Aij = − A ji

∂Ti ∂T j

So, Aij = − is Skew-symmetric.

∂x j ∂x i

∂Ti ∂T j

So, − are component of a Skew-symmetric covariant tensor of rank two.

∂x j ∂x i

24 Tensors and Their Applications

By this law, we can test a given quantity is a tensor or not. Suppose given quantity be A and we do not

know that A is a tensor or not. To test A, we take inner product of A with an arbitrary tensor, if this

inner product is a tensor then A is also a tensor.

Statement

If the inner product of a set of functions with an atbitrary tensor is a tensor then these set of

functions are the components of a tensor.

The proof of this law is given by the following examples.

EXAMPLE 11

Show that the expression A(i,j,k) is a covariant tensor of rank three if A(i,j,k)B k is covariant

tensor of rank two and B k is contravariant vector

Solution

Let X and Y be two coordinate systems.

As given A (i, j, k)B k is covariant tensor of rank two then

∂x p ∂x q

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

∂x i ∂x j

∂x k ∂x r

Bk = r Br or Br = k B k

∂x ∂x

So, from (1)

∂x p ∂x q ∂x r

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

∂x i ∂x j ∂x k

∂x p ∂x q ∂x r

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

∂x i ∂x j ∂x k

∂x p ∂x q ∂ x r

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

∂ x i ∂x i ∂x k

As B k is arbitrary..

So, A(i , j , k ) is covariant tensor of rank three.

EXAMPLE 12

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

type (1, 2).

Solution

Let X and Y be two coordinate systems. As given A(i , j, k ) A i B j C k is scalar. Then

i j p q

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

Tensor Algebra 25

∂x i p ∂x p i

Ai = A or Ap = A

∂x p ∂x i

∂x j q ∂x q j

Bj = B or Bq = B

∂x q ∂x j

∂x k ∂x r k

C k = r Cr or Cr = C

∂x ∂x k

So, from (1)

i j ∂x p ∂xq ∂x k

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

∂x i ∂x j ∂xr

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

Then

∂x p ∂xq ∂x k

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

∂xi ∂x j ∂xr

So, A(i, j, k) is tensor of type (1, 2).

Consider a covariant symmetric tensor Aij of rank two. Let d denote the determinant Aij with the

elements Aij i.e., d = Aij and d ≠ 0 .

Now, define A ij by

Cofactor of Aij is the determinan t Aij

A ij =

d

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

ij

of Aij .

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

Bij

A ij =

d

Then prove that Aij A = . δ ki

kj

(i) Aij Bij = d

Bij

⇒ Aij =1

d

Bij

Aij Aij = 1, given Aij =

d

26 Tensors and Their Applications

(ii) Aij B kj = 0

B kj

Aij = 0, d ≠0

d

Aij Akj = 0 if i ≠ k

from (i) & (ii)

1 if i = k

Aij Akj = 0 if i ≠ k

i.e., Aij Akj = δik

i i ...i

If the components of a tensor A j11 2j 2 ...rj s transform according to the equation

ω

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

Alk11l 2k...

2 ...k r

= Aij11i 2j 2......i rj s ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅

∂x

ls

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

∂x

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

i i ...i

is the Jacobian of transformation. If

∂x

ω = 1, the relative tensor is called a tensor density. If w = 0 then tensor is said to be absolute.

MISCELLANEOUS EXAMPLES

1. Show that there is no distinction between contravariant and covariant vectors when we

restrict ourselves to transformation of the type

x i = am x + b ;

i m i

a ir ami = δ rm

Solution

Given that

x i = am x + b

i m i

...(1)

or a im x m = x i − bi ...(2)

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

a ir ami x m = a ir x i − bi a ir

δ rm x m = a ir x i − bi a ir as given a ir ami = δ rm

x r = a r x − b a r as δ rm x m = x r

i i i i

xs = as x − b as

i i i i

or

Differentiating Partially it w.r.t. to x i

∂x s

= a is ..(3)

∂x i

Tensor Algebra 27

x i = as x + b

i s i

∂x i

a is

s = ...(4)

∂x

The contravariant transformation is

∂x i s

A = A = a is A s

i ...(5)

∂x s

The covariant transformation is

∂x s

As = a is As

Ai = ...(6)

∂x i

Thus from (5) and (6), it shows that there is no distinction between contravariant and

covariant tensor law of transformation

2. If the tensors aij and g ij are symmetric and u i , v i are components of contravariant vectors

satisfying the equations

(aij − kgij )u i = 0, i, j = 1, 2,...,n

( aij − k'gij )v i = 0, k ≠ k ′ .

Prove that gij u i v j = 0, a ij u i v j = 0.

Solution

The equations are

j j

Multiplying (1) and (2) by u and v respectively and subtracting, we get

aij u i v j − aij v i u j − kgij u i v j + k ′gij u j v i = 0

Interchanging i and j in the second and fourth terms,

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

− kgij v j u i + k ′gij u i v j = 0

(k ′ − k ) g ij u v = 0

i j

g ij u i v j = 0 since k ≠ k ′ ⇒ k − k ′ ≠ 0

Multiplying (1) by v j , we get

a ij v j u i − kgij u i v j = 0

a ij u i v j = 0 as g ij u i v j = 0. Proved.

28 Tensors and Their Applications

i k i k

(δ j δl + δl δ j ) Aik = 0

Solution

Given Aij is a Skew-symmetric tensor then Aij = − A ji .

Now,

i k i k

(δ j δl + δl δ j ) Aik = δ ij δ kl Aik + δil δ kj Aik

= A jl + Alj

i k i k

(δ j δl + δl δ j ) Aik = 0 as A jl = − Alj

4. If aij is symmetric tensor and bi is a vector and a ij bk + a jk bi + a kib j = 0 then prove that

a ij = 0 or bk = 0 .

Solution

The equation is

aij bk + a jk bi + akib j = 0

⇒ aij bk + a jk bi + akib j = 0

By tensor law of transformation, we have

∂x p ∂x q ∂x r ∂x p ∂x q ∂x r ∂x p ∂x q ∂x r

a pq b + a b + a b =0

∂x i ∂x j ∂x k

r pq

∂x j ∂x k

r

∂x i

pq

∂x k ∂x i

r

∂x j

∂x p ∂x q ∂x r ∂x p ∂x q ∂x r ∂x p ∂x q ∂x r

a pq br i + j + = 0

∂x ∂x ∂x ∂x ∂x k ∂x i ∂x k ∂x i ∂x j

j k

⇒ a pq br = 0 ⇒ a pq = 0 or br = 0

⇒ a ij = 0 or bk = 0

5. If a mn x x = bmn x x for arbitrary values of x r , show that a(mn) = b(mn) i.e.,

m n m n

a mn + a nm = bmn + bnm

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

Solution

Given

a mn x m x n = bmn x m x n

(a mn − bmn )x m x n = 0

Tensor Algebra 29

Differentiating again w.r.t. x j partially

(aij − bij ) + (a ji − b ji ) = 0

a ij + a ji = bij + b ji

So,

2 amn = 2bmn

a mn = bmn

EXERCISES

(i) Aij

(ii) Bkij

ijk

(iii) Clm

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

3. If Aij is a contravariant tensor and Bi is covariant vector then prove that Aij Bk is a tensor of rank

three and Aij Bj is a tensor of rank one.

4. If Ai is an arbitrary contravariant vector and Cij Ai Aj is an invariant show that Cij + Cji is a covariant

tensor of the second order.

5. Show that every tensor can be expressed in the terms of symmetric and skew-symmetric tensor.

n n

6. Prove that in n-dimensional space, symmetric and skew-symmetric tensor have (n + 1) and (n − 1)

2 2

independent components respectively.

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

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

i k i k

ij i

9. Explain the process of contraction of tensors. Show that aij a = δ j .

30 Tensors and Their Applications

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

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

ij k k

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

h i h k

13. Prove that Aij Bi C j is invariant if Bi and C j are vector and Aij is tensor of rank two.

14. If A(r, s, t) be a function of the coordinates in n-dimensional space such that for an arbitrary vector

Br of the type indicated by the index a A(r, s, t)Br is equal to the component Cst of a contravariant

tensor of order two. Prove that A(r, s, t) are the components of a tensor of the form Arst .

15. If Aij and Aij are components of symmetric relative tensors of weight w. show that

w− 2 w+ 2

ij ij ∂x ∂x

A = A and Aij = Aij

∂x ∂x

16. Prove that the scalar product of a relative covariant vector of weight w and a relative contravariant

vector of weight w′ is a relative scalar of weight w + w ′ .

CHAPTER – 3

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

. . . . . . . . .. . . . . . . . . . . . . . . . .

+ +

+ g n1 dx n dx1 + g n 2 dx n dx 2 + ⋅ ⋅ ⋅ ⋅ + g nn (dx n )2

ds 2 = g ij dx i dx j (i, j = 1,2,...n) ...(1)

using summation convention.

Where g ij are the functions of the coordinates x i such that

g = g ij ≠ 0

The quadratic differential form (1) is called the Riemannian Metric or Metric or line element for n-

dimensional space and such n-dimensional space is called Riemannian space and denoted by Vn and

g ij is called Metric Tensor or Fundamental tensor..

The geometry based on Riemannian Metric is called the Riemannian Geometry.

THEOREM 3.1 The Metric tensor g ij is a covariant symmetry tensor of rank two.

Proof: The metric is given by

i j

ds 2 = g ij dx dx ...(1)

32 Tensors and Their Applications

system. Then metric ds2 = gij dxidxj transforms to ds 2 = g ij d x i dx j .

Since distance being scalar quantity.

ds 2 = g ij dx dx = gij dx dx

i j i j

So, ...(2)

The theorem will be proved in three steps.

(i) To show that dxi is a contravariant vector.

If x i = x i (x1, x2 ,...x n )

∂x i 1 ∂x i 2 ∂x i n

dxi = dx + dx + ⋅ ⋅ ⋅ + dx

∂x i ∂x 2 ∂x n

∂x i k

dxi = dx

∂x k

It is law of transformation of contravariant vector. So, dx i is contravariant vector..

(ii) To show that g ij is a covariant tensor of rank two. Since

∂x i k ∂x j l

dxi = dx and d x j

= ∂x

∂x k ∂x l

from equation (2)

∂x i ∂x j l

g ij dx i dx j = g ij dxk dx

∂x k ∂x l

∂x i ∂x j k l

g ij dx i dx j = g ij k ∂x dx

∂x ∂x l

∂x i ∂x j k l

k l

g kl dx dx = ij kg dx dx

∂x ∂x l

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

k l

g kl − g ij ∂x ∂x

i j

dx dx = 0

∂x k ∂x l

∂x i ∂x j

or g kl − gij = 0 as dx k and dx l are arbitrary..

∂x k ∂x l

∂x j ∂x j

g kl = g ij

∂x k ∂x l

∂x k ∂x l

or g ij = g kl

∂x i ∂x j

Metric Tensor and Riemannian Metric 33

1 1

g ij = (g + g ji ) + ( gij − g ji )

2 ij 2

g ij = Aij + Bij

1 g +g

where Aij = ( ij ji ) = symmetric

2

1

Bij = (g ij − g ji ) = Skew-symmetric

2

Now, g ij dx i dx j = ( Aij + Bij )dxi dx j from (3)

Bij dxi dx j = B ji dx i dx j

Bij dxi dx j = − B ij dx i dx j

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),

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.

Proof: Let x i be coordinates of a point in X-coordinate system and x i be coordinates of a same point

in Y-coordinate system.

Since g ij is a Covariant tensor of rank two.

∂x k ∂x 1

Then, g ij = g kl

∂x i ∂x j

34 Tensors and Their Applications

∂x k ∂x l

⇒ g ij − g kl =0

∂x i ∂x j

i j

gij − g kl ∂x ∂x

k l

dx dx = 0

∂x i ∂x j

i j ∂x k ∂x l i j

( g ij dx dx ) = g kl dx dx

∂x i ∂x j

∂x k i ∂x l

= g kl dx dx j

∂x i ∂x j

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

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

Bij

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

By theorem on page 26

So, g ij g kj = δik

Note (i) Tensors g ij and g ij are Metric Tensor or Fundamental Tensors.

(ii) g ij is called first fundamental Tensor and g ij second fundamental Tensors.

EXAMPLE 1

Find the Metric and component of first and second fundamental tensor is cylindrical coordinates.

Solution

Let (x1, x2, x3) be the Cartesian coordinates and (x1, x2 , x3 ) be the cylindrical coordinates of a

point. The cylindrical coordinates are given by

x = r cosθ , y = r sin θ, z = z

So that

x1 = x, x 2 = y, x 3 = z and x 1 = r , x 2 = θ, x 3 = z ...(1)

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

respectively.

Metric Tensor and Riemannian Metric 35

ds 2 = dx + dy + dz

2 2 2

12 2 2 3 2 ...(2)

i j

But ds 2 = g ij dx dx

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

+ g 32 dx 3 dx 2 + g33(dx3)3 ...(3)

Comparing (2) and (3), we have

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

On transformation

∂x i ∂x j

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

∂x i ∂x j

for i = j = 1.

2 2

∂x1 ∂x 2 ∂x 3

g11 = g11 1 + g 22 1 + g 33 1

∂x ∂x ∂x

since g12 = g13 = ⋅ ⋅ ⋅ = g32 = 0 .

2 2 2

∂x ∂y ∂z

g11 = g11 + g 22 + g 33

∂r ∂r ∂r

Since x = r cos θ, y = r sin θ, z = z

∂x ∂y ∂z

= cosθ, = sin θ, =0

∂r ∂r ∂r

and g11 = g 22 = g 33 = 1.

g11 = cos 2 θ + sin 2 θ + 0

g11 = 1

Put i = j = 2.

2 2 2

∂x 1 2 3

g 22 = g11 2 + g 22 ∂x + g 33 ∂x

∂x 2 ∂x 2

∂x

∂x

2

∂y

2

∂z

2

g 22 = g11 + g 22 + g 33

∂θ ∂θ ∂θ

36 Tensors and Their Applications

Since g11 = g 22 = g 33 = 1

∂x ∂y ∂z

= −r sin θ, = r cosθ, =0

∂θ ∂θ ∂θ

g 22 = (− r sin θ )2 + (r cos θ)2 + 0

= r 2 sin 2 θ + r 2 cos 2 θ

g 22 = r 2

Put i = j = 3.

2

∂x1 ∂x 2 ∂x 3

g

= 11 3 + g

22

+ g

g 33 3 33 3

∂x ∂x ∂x

∂x

2

∂y ∂z

= g11 + g 22 + g 33

∂z ∂z ∂z

∂x ∂y ∂z

Since = 0, = 0, = 1 . So, g 33 = 1 .

∂z ∂z ∂z

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

(i) The metric in cylindrical coordinates

i j

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

ds 2 = g11 dx ( )

1 2

( )

+ g 22 dx 2

2

( )

+ g 33 dx 3

2

ds 2 = dr + r (dθ ) + dφ

2 2 2 2

2

g ij = g 21 g 22 g 23 = 0 r 0

g 31 g32 g 33 0 0 1

1 0 0

since g = g ij = 0 r2 0

0 0 1

2

g= r

Metric Tensor and Riemannian Metric 37

B11 = r 2 , B22 = 1, B33 = r 2

and B12 = B21 = B13 = B 23 = B31 = B32 = 0

Bij

The second fundamental tensor or conjugate tensor is g =

ij

.

g

cofactor of g11 in g

g 11 =

g

B11 r 2

g 11 = = 2 =1

g r

B12 1

g 22 = = 2

g r

B33 r 2

g 33 = = 2 =1

g r

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

1 0 0

1

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

r2

0 0 1

EXAMPLE 2

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

Solution

point. The spherical coordinates are given by

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

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

Let g ij and g ij be the metric tensors in cartesian and spherical coordinates respectively..

The metric in cartesian coordinates is given by

ds 2 = dx + dy + dz

2 2 2

2 2 2 3 2

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

i j

But

38 Tensors and Their Applications

On transformation

∂x i ∂x j

g ij = g ij

∂x i ∂x j

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

∂x1 ∂x 1 ∂x 2 ∂x 2 ∂x 3 ∂x 3

g ij = g11 + g + g

∂x i ∂x j

22

∂x i ∂x j

33

∂x i ∂x 1

since i, j are dummy indices.

Put i = j = 1

2 2 2

∂x 1 ∂x 2 ∂x 3

g

= 11 1 + g

22

+ g

g11 1 33

∂x ∂x ∂x

1

∂x

2

∂y

2

∂z

2

g11 = g11 + g 22 + g 33

∂r ∂r ∂r

Since x = r sin θ cos φ, y = r sin θ sin φ , z = r cos θ

∂x ∂y ∂z

= sin θ cos φ, = sin θ sin φ, = cos θ

∂r ∂r ∂r

and g11 = g 22 = g 33 = 1 .

So,

g11 = (sin θ cos φ )2 + (sin θ sin φ )2 + cos2 θ

g11 = 1

put i = j = 2

2 2

∂x 1 2 3

= g11 2 + g 22 ∂x + g 33 ∂x

g 22 ∂x 2 ∂x 2

∂x

∂x

2

∂y

2

∂z

2

g 22 = g11 + g 22 + g33

∂θ ∂θ ∂θ

since g11 = g 22 = g 33 = 1

∂x ∂y ∂z

= r cos θ cos φ, = r cos θ sin φ, = −r sin θ

∂θ ∂θ ∂θ

g 22 = r 2

Metric Tensor and Riemannian Metric 39

Put i = j = 3

2 2 2

∂x1 ∂x 2 ∂x 3

g

= 11 3 + g

22

+ g

g 33 3 33

∂x ∂x ∂x

3

2 2 2

∂x ∂y ∂z

g 33 = g11 + g 22 + g33

∂φ ∂φ ∂φ

since g11 = g 22 = g 33 = 1

∂x ∂y ∂z

and = −r sin θ sin φ, = r sin θ cos φ, =0

∂φ ∂φ ∂φ

g 33 = (− r sin θ sin φ)2 + (r sin θ cos φ )2 + 0

g 33 = r 2 sin 2 θ

So, we have

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

(i) The Metric in spherical coordinates is

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

i j

ds 2 = g11 dx ( )

1 2

( )

+ g 22 dx 2

2

( )

+ g 33 dx 3

2

ds 2 = dr 2 + r 2 dθ 2 + r 2 sin 2 θdφ 2

(ii) The Metric tensor or first fundamental tensor is

g g g = 0 r 2

g ij = 21 22 23 0

g 31 g32 g 33 0 0 r 2 sin 2 θ

and

1 0 0

g = g ij =0 r2 0 = r 4 sin 2 θ

0 0 r 2 sin 2 θ

(iii) The cofactor of g are given by B11 = 1, B22 = r 2 , B33 = r 2 sin 2 θ and B12 = B21 =

B31 = B13 = B23 = B32 = 0

Bij

The second fundamental tensor or conjugate tensor is g ij = .

g

40 Tensors and Their Applications

g 11 = =

g g

r 4 sin 2 θ

=

r 4 sin 2 θ

g 11 =1

B 22 r 2 sin 2 θ

g 22

= = 4 2

g r sin θ

1

g 22 =

r2

B33 r2

g 33

= =

g r 4 sin 2 θ

1

g 33 =

r sin 2 θ

2

and g = g = g = g = g = 0

12 13 21 31 32

0 0 0

g 11

g 12

g

13

1

21 23 0 0

g ij

= g g 22 g = r2

g 31

g 32 g 33 0 0

1

r 2 sin 2 θ

EXAMPLE 3

If the metric is given by

( ) ( )

ds 2 = 5 dx 1 + 3 dx 2

2 2

( )

+ 4 dx 3

2

− 6dx 1dx 2 + 4dx 2 dx 3

Solution

i j

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

ds 2 = g11 dx

1

( ) 2

+ g12 dx 1dx 2 + g13 dx1 dx 3 + g 21dx 2 dx1

Since g ij is symmetric ⇒ g ij = g ji

Metric Tensor and Riemannian Metric 41

1 2 2 2 3 2 1 2

So,

Now, the given metric is

1 2 2 2 3 2 1 2 2 3

...(2)

Comparing (1) and (2) we have

g11 = 5, g 22 = 3, g 33 = 4, 2g12 = −6 ⇒ g12 = −3 = g 21

2 g 23 = 4 ⇒ g 23 = 2 = g 32 , g13 = 0 = g 31

g = g ij = g 21 g 22 g 23 = − 3 3 2 = 4

g 31 g 32 g33 0 2 4

Then

3 2

B11 = Cofactor of g11 = =8

2 4

5 0

B22 = Cofactor of g 22 = = 20

04

5 −3

B33 = Cofactor of g 33 = =6

−3 3

−3 2

B12 = Cofactor of g12 = − = 12 = B 21

0 4

−3 3

B13 = Cofactor of g13 = = −6 = B31

0 2

5 −3

B 23 = Cofactor of g 23 = – = −10 = B 32

0 2

Bij

Since g ij =

g

We have

B 8

g 11 = 11 = = 2; g 22 = 5, g 33 = 3 , g 12 = g 21 = 3, g 13 = g 31 = − 3 , g 23 = g 32 = − 5

g 4 2 2 2

42 Tensors and Their Applications

Hence,

3

2 3 −

2

= 5 −

5

g ij 3

2

− 3 5 3

−

2 2 2

3.3 LENGTH OF A CURVE

Consider a continuous curve in a Riemannian Vn i.e., a curve such that the coordinate x i of any

current point on it are expressible as functions of some parameter, say t.

The equation of such curve can be expressed as

x i = xi (t)

The length ds of the arc between the points whose coordinate s are x i and x i + dx i given by

i j

ds 2 = g ij dx dx

If s be arc length of the curve between the points P1 and P2 on the curve which correspond to

the two values t1 and t 2 of the parameter t.

12

t2 i j

gij dx dx dt

∫ ∫

P2

ds = dt dt

s =

P1 t1

NULL CURVE

dx i dx j

If g ij = 0 along a curve. Then s = 0. Then the points P1 and P2 are at zero distance, despite

dt dt

of the fact that they are not coincident. Such a curve is called minimal curve or null curve.

EXAMPLE 4

A curve is in spherical coordinate xi is given by

−1 1

x1 = t, x = sin and x = 2 t − 1

2 3 2

t

Find length of arc 1 ≤ t ≤ 2.

Solution

In spherical coordinate, the metric is given by

ds 2 = (dx ) + ( x ) (dx ) + ( x sin x ) (dx )

1 2 1 2 2 2 1 2 2 3 2

Metric Tensor and Riemannian Metric 43

−1 1

x1 = t, x = sin , x3 = 2 t 2 − 1

2

given

t

dx 1 = dt , dx 2 =

1 1 1

(

− 2 dt , dx 3 = 2 ⋅ t 2 − 1

−1 2

)2t dt

1 t

2 2

1−

t

dt 2t

dx 2 = − , dx 3 = dt

t t −1 2

t 2 −1

2 2

dt 2t

2

−1 1

ds = (dt )

2

+t − + t sin sin dt

2

t t 2 − 1

2

t t 2 − 1

dt 2 4t 2

ds =2dt 2

+ + (dt )2

t −1 t −1

2 2

5t 2 2

ds 2 = t 2 − 1 dt

t

ds = 5 dt

t −12

2

t 5 t2 −1

∫

2

∫

t2

ds = 5 dt = =

2 12 15 units

t1 1

t2 −1 1

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

Associated Vector

Consider a covariant vector Ai . Then g ik Ai = A k is called associated vector of Ai . Consider a

contravariant vector A j . Then g jk A j = Ak is called associated vector of A j .

44 Tensors and Their Applications

i

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

A= g ij Ai A j

i j

or A 2 = g ij A A

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

i

i.e., square of the magnitude is equal to scalar product of the vector and its associate.

The magnitude or length A of covariant vector Ai . Then A is defined by

A= g ij Ai A j

or A 2 = g ij Ai A j

A vector of magnitude one is called Unit vector. A vector of magnitude zero is called zero vector

or Null vector.

Let A and B be two vectors. Their scalar product is written as A ⋅ B and defined by

r r

A ⋅ B = Ai Bi

r r

A ⋅ B = A B i = g ij A B since Bi = g ij B

i i j j

Also,

r r

A ⋅ B = Ai B = g Ai B j since B = g B j

i ij i ij

Thus

r r

A ⋅ A = A Ai = g ij A A = A

i i j 2

r

i.e., A = A = g ij Ai A j

Angle

r between two vectors

r

Let A and B be two vectors. Then

r r r r

A ⋅ B = A B cos θ

r r

A⋅B g ij A i B j

⇒ cos θ = r r =

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

r r

since A = g ij Ai A j ; B = g ij B i B j

This is required formula for cos θ .

Definition

r r

The inner product of two contravariant vectors A (or A i ) and B (or Bi ) associated with a symmetric

tensor g ij is defined as g ij A i B j . It is denoted by

r r i j

g ( A , B ) = g ij A B

Metric Tensor and Riemannian Metric 45

r r

THEOREM 3.3 The necessary and sufficient condition that the two vectors A and B at 0 be orthogonal

r r

if g ( A, B ) = 0

r r

Proof: Let θ be angle between the vectors A and B then

r r r r

A ⋅ B = A B cos θ

r r

or A ⋅ B = AB cos θ

g ij A i B j = AB cos θ

g ij Ai B j

⇒ cos θ = ...(1)

AB

r r

If A and B are orthogonal then θ = π ⇒ cos θ = 0 then from (1)

2

i j

g ij A B = 0

r r r r

⇒ ( ) ( ) i j

g A, B = 0 since g A, B = g ij A B

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

i j

cos θ = 0 ⇒ θ = π .

r r 2

So, two vectors A & B are orthogonal. Proved.

r r r r

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

r r

cos θ = A ⋅ B = gij A i B j

r r π π

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

2 2

π

cos θ = cos θ = = 0

2

THEOREM 3.4 To show that the definition of the angle between two vectors is consistent with the

requirement cos 2θ ≤ 1.

OR

To justify the definition of the angle between two vectors.

OR

To show that the angle between the contravariant vectors is real when the Riemannian Metric is

positive definition.

r r

Proof: Let θ be the angle between unit vectors A and B then

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

i j j ij ij i

46 Tensors and Their Applications

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

i i j j

= g ij l A A + g ij lmA B + m lg ij B A + m g ij B B

2 i j i j i j 2 i j

= l 2 + 2lm cos θ + m 2

Since

g ij A i A j = A 2 = 1; g ij B B = B = 1.

i j 2

and

r r r

g ij A i B j = cosθ; as A & B are unit vector i.e., A = 1 ⇒ A 2 = 1 .

Since square of magnitude of any vector ≥ 0.

So, the square of the magnitude of lAi + mB i ≥ 0.

or l 2 + 2lm cos θ + m 2 ≥ 0

(l + m cos θ )2 + m 2 − m 2 cos 2 θ ≥0

(l + m cosθ) 2 + m 2 (1 − cos2 θ) ≥ 0

This inequality holds for the real values of l & m.

if 1 − cos2 θ ≥ 0

⇒ cos 2 θ ≤ 1

cos θ ≤ 1

Proved.

THEOREM 3.5 The magnitude of two associated vectors are equal.

Proof: Let A and B be magnitudes of associate vectors A i and A i respectively. Then

i j

A 2 = g ij A A ...(1)

and

ij

B 2 = g Ai A j ...(2)

From equation (1)

A 2 = (gij Ai ) A j

A2 = Aj A j ...(3)

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.

Let a Vn referred to coordinate x i , (i = 1, 2, ... n) . For a coordinate curve of parameter x l, the coordinate

x l alone varies. Thus the coordinate curve of parameter x l is defined as

x i = c i , ∀i except i = l ...(1)

i,

where C s are constants.

Differentiating it, we get

dx i = 0, ∀i , except i = l and dx l ≠ 0

Let Ai and B i be the tangent vectors to a coordinate curve of parameters x p and x q respectively..

Then

Ai = dx i = (0,...0, x p , 0...0) ...(2)

If θ is required angle then

g ij A i B j

cos θ =

g ij A i A j g ij B i B j

g pq A p B q

=

g pp A p A p g qq B q B q

g pq A p B q

=

g pp g qq A p B q

g pq

cos θ = ...(4)

g pp g qq

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

g ij

cos wij =

g ii g jj

48 Tensors and Their Applications

π

cos wij = cos =0

2

⇒ g ij = 0

3.9 HYPERSURFACE

The n equations xi = xi (u1) represent a subspace of Vn . If we eliminate the parameter u1, we get

(n –1) equations in x j, s which represent one dimensional curve.

Similarly the n equations xi = xi (u1,u2) represent two dimensional subspace of V n. If we eliminating

the parameters u1, u2, we get n –2 equations in xi,s which represent two dimensional curve V n. This

two dimensional curve define a subspace, denoted by V 2 of V n.

Then n equations xi = xi (u1, u2, ... un–1) represent n – 1 dimensional subspace V n–1 of V n. If we

eliminating the parameters u1, u2, ...un–1, we get only one equation in x i, s which represent n –1

dimensional curve in V n. This particular curve is called hypersurface of V n.

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

i

hypersurface of V n.

Let

φ (x i ) = constant ...(1)

represents two families of hypersurfaces.

Differentiating equation (1), we get

∂φ i

dx = 0 ...(3)

∂x i

∂φ ∂φ

This shows that is orthogonal to dx i . Hence is normal to φ = constant, since dx i is

∂x i

∂x i

tangential to hypersurface (1).

∂ψ

Similarly is normal to the hypersurface (2). If ω is the angle between the hypersurface (1)

∂x i

and (2) then ω is also defined as the angle between their respective normals. Hence required angle ω

is given by

∂φ ∂ψ

g ij

∂x i ∂x j

cos ω = ...(4)

∂φ ∂φ ∂ψ ∂ψ

g ij i g ij i

∂x ∂x ∂x ∂x j

j

Metric Tensor and Riemannian Metric 49

If we take

φ = x p = constant ...(5)

and ψ = x q = constant ...(6)

The angle ω between (5) and (6) is given by

∂x p ∂x q

g ij

∂x i ∂x j

cos ω =

∂x p ∂x p ij ∂x ∂x

q q

g ij i g

∂x ∂x j ∂x i ∂x j

g ij δip δ qj

=

g ij δip δ qj g ij δqi δ qj

g pq

cos ω = ...(7)

g pp g qq

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

g ij

cos ωij = ...(8)

g ii g jj

π

ωij =

2

⇒ cos ωij = 0

from (8), we have g = 0 .

ij

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.

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

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

ehi |ek |i = δ hk

from (1),

g ij ehi | ekj| = 0

We define

cofactorof eh |i in determinant eh |i

ehi | =

e h|i

Also, from the determinant property, we get

n

∑e e

h =1

i

h| h | j = δij ...(2)

Multiplying by e jk

n

∑e e

h =1

i

h| h| j g j k = δ ij g jk

or ∑e e

h =1

i k

h | h| = g ik ...(3)

n

∑e e

h =1

i

h | h | j g ik = δ j g ik

i

or g jk = ∑e h |k eh | j ...(4)

from (3) and (4)

n

g ij = ∑e

h =1

h |i e h| j ...(5)

Metric Tensor and Riemannian Metric 51

n

g ij = ∑e

h =1

i

h| e hj| ...(6)

Corollary: To find the magnitude of any vector u is zero if the projections of u on eh | are all zero.

Proof: Let

n

i

u = ∑C e

h =1

i

h h| ...(7)

Then

n n

u i ek |i = ∑C e e

h =1

i

h h | k |i = ∑C δ

h =1

h

h

k = Ck

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

i.e., C k = projection of u i on ek |i

Using (8), equation (7) becomes

n

ui = ∑u e

h =1

j i

h| j ek |

Now,

u 2 = u iu i =

∑ C e ∑ C e

i

h h| k k |i

from (7)

h k

= ∑C C e e

h ,k

h

i

k h | k| i

= ∑C C δ

h ,k

h k

h

k

= ∑C C

h

h h

u2 = ∑ (C

h =1

h )2

Hence the magnitude of a vector u is zero iff all the projections of u (i.e. of ui) on n mutually

orthogonal directions ehi | are zero.

52 Tensors and Their Applications

Miscellaneous Examples

1. If p and q are orthogonal unit vectors, show that

( g hj g ik − g hk gij ) p h q i p j q k = 1

Solution

Since p and q are orthogonal unit vectors. Then

g ij pi q j = 0, p 2 = q 2 = 1 .

Now,

( g hj g ik − g hk gij ) p h q i p j q k = g hj g ik p p q q − g hk g ij p q q p

h j i k h k i j

= ( g hi p h p j ) ( gik q i q k ) − (g hk p j q k ) ( gij q i p j )

= p2.q2 – 0.0

=1.1

= 1 (since g hi p h p j = 1 & g hk p h q k = 0 )

2. If θ is the inclination of two vectors A and B show that

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

sin θ =

2

g hj gik A h A j B j B k

Solution

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

g ij A j B i

cos θ =

gij A i A j g ik B i B k

But sin 2 θ = 1 − cos 2 θ

( gij B i A j ) (g hk A h B k )

sin 2 θ = 1 −

( g hj A h A j ) ( g ik B i B k )

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

=

g hj g ik A h A j B i B k

3. If X ij are components of a symmetric covariant tensor and u, v are unit orthogonal to w and

satisfying the relations

( X ij − αg ij )u i + γw j = 0

( X ij − βg ij ) v i + δw j = 0

where α ≠ β prove that u and v are orthogonal and that

Metric Tensor and Riemannian Metric 53

X ij u i v j = 0

Solution

u i wi = 0 ...(1)

v i wi = 0 ...(2)

where α ≠ β.

Multiply (3) & (4) by v j , u j respectively and using (1) and (2), we have

( X ij − αg ij )u iv j = 0 ...(5)

( X ij − βg ij )v iu j = 0 ...(6)

Interchanging the suffixes i & j in the equation (6) and since g ij , X ij are symmetric, we get

( X ij − αg ij )u iv j = 0 ... (7)

Subtract (6) & (7) we get

(β − α)g iju i v j = 0

Since β ≠ α and β − α ≠ 0.

Hence,

g ij u i v j = 0 ...(8)

So, u and v are orthogonal.

Using (8) in equation (5) & (6), we get

X ij u i v i = 0 Proved.

4. Prove the invariance of the expression g dx1 dx 2 ...dx n for the element volume.

Solution

∂x k ∂x l

g ij = g kl

∂x i ∂x j

Taking determinant of both sides

∂x k ∂x l

g ij = g kl

∂x i ∂x j

∂x

Since = J (Jacobian)

∂x

g kl = g & g ij = g

54 Tensors and Their Applications

So,

g = gJ 2

or

g

J= g

Now, the transformation of coordinates from x l to x i , we get

∂x 1 2 n

dx 1 dx 2 ... dx n = ∂x d x d x ... dx

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

g 1 2

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

g

EXERCISES

1. For the Metric tensor g ij defined g kl and prove that it is a contravariant tensor.

2. Calculate the quantities g i j for a V3 whose fundamental form in coordinates u, v, w, is

2 2 2

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

3. Show that for an orthogonal coordinate system

1 1 1

g 11 = , g 22 = , g 33 =

g11 g 22 g 33

4. For a V2 in which g11 = E , g12 = F , g 21 = G prove that

g = EG − F 2 , g 11 = G g , g 12 = − F g , g 22 = E g

1

5. Prove that the number of independent components of the metric g ij cannot exceed n( n + 1) .

2

6. If vectors u i , vi are defined by u i = g ij u j , v i = g ij v j show that u i = g ij u j , u i vi = u i v i and u i g ij u j = u i g ij u j

7. Define magnitude of a unit vector. prove that the relation of a vector and its associate vector is

reciprocal.

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

(g hi gik − g hk gij ) A h Ai B j B k

2

sin θ = ghi g jk Ah Ai B j B k

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

definite.

CHAPTER – 4

DIFFERENTIATION

The German Mathematician Elwin Bruno Christoffel defined symbols

1 ∂gi k ∂g j k ∂g i j

[ij, k ] =

2 ∂x j

+

∂x i

− k , (i, j, k = 1, 2,...n )

∂x

...(1)

called Christoffel 3-index symbols of the first kind.

k

and = g k l [ij, l ] ...(2)

i j

called Christoffel 3-index symbols of second kind, where g i j are the components of the metric Tensor

or fundamental Tensor.

There are n distinct Christoffel symbols of each kind for each independent g i j . Since g i j is

1

symmetric tensor of rank two and has n (n + 1) independent components. So, the number of

2

independent components of Christoffel’s symbols are n ⋅ n(n + 1) = n 2 (n + 1) .

1 1

2 2

k

THEOREM 4.1 The Christoffel's symbols [ij, k ] and are symmetric with respect to the indices i

i j

and j.

Proof: By Christoffel’s symbols of first kind

1 ∂g ik ∂g jk ∂g ij

[ij, k ] = j +

2 ∂x

∂ − k

∂x i

∂x

Interchanging i and j, we get

1 ∂g jk ∂gik ∂g ji

[ ji, k ] = + − k

2 ∂x i ∂x j ∂x

56 Tensors and Their Applications

1 ∂g i k ∂g j k ∂g i j

+ − k since g ij = g ji

2 ∂x j

=

∂x i ∂x

[ ji, k ] = [ ij , k ]

Also, by Christoffel symbol of second kind

k

= g k l [ij, l ]

i j

= g k l [ ji, l ] since [ij , l ] = [ ji, l ]

k k

= j i

i j

Proved.

THEOREM 4.2 To prove that

k

(i) [ij, m] = g km

i j

∂g ij

(ii) [ik , j ] + [ jk , i] =

∂x k

∂g ij i j

(iii) = − g jl − g im

∂x k

l k m k

Proof: (i) By Christoffel’s symbol of second kind

k

= g k l [ij, l ]

i j

Multiplying this equation by g k m , we get

k

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

i j

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

k

g km = [ij, m]

i j

(ii) By Christoffel’s symbol of first kind

1 ∂g k j ∂g i j ∂g i k

[ik , j ] = + −

2 ∂x i ∂x k ∂x j

...(1)

1 ∂g ki ∂g ji ∂g jk

and [ jk ,i ] = + − ...(2)

2 ∂x j ∂x k ∂x i

adding (1) and (2),

1 ∂ g i j ∂g ji

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

2 ∂ x k ∂x k since ij

Christoffel's Symbols and Covariant Differentiation 57

1 ∂g ij ∂g ij

[ik , j ] + [ jk , i] = ⋅2 =

2 ∂x k ∂x k

(iii) Since g ij g lj = δ li .

∂glj ∂g ij

g ij + glj =0

∂x k ∂x k

Multiplying this equation by g lm , we get

∂g lj ∂g ij

g ij g lm + g lm g lj =0

∂x k ∂x k

∂g ij ∂g

g lm glj k =

− g ij g lm ljk

∂x ∂x

∂g ij ∂g lj

δ mj = − g ij lm

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

∂x k ∂x k

∂g im { } {

= − g g [lk , j ] − g g [ jk , l ]

lm ij ij lm

}

∂x k

∂g im i m

= − g lm − g ij

∂x k

l k j k

Interchanging m and j, we get

∂g ij i

lj

j

im

= − g − g

∂x k l k m k

∂g ij i j

or k =

− g ij − g im as g lj = g jl

∂x l k m k

Proved.

i ∂ log g

THEOREM 4.3 To show that i j =

( )

∂x j

Proof: The matrix form of g ik is

g

21 g 22 ... G2 n

g ik =

M

g n1 g n 2 ... g nn

58 Tensors and Their Applications

g 21 g 22 ... g 2 n

and g = gik =

M

g n1 g n 2 ... gnn

But g ik g = δlk

il

Take l = k.

g ik g ik = δ kk = 1

g ik = [g ik ]−1 =

Gik

⇒ (Theorem 2.13 , Pg 25)

g

⇒ g = g ik Gik ...(1)

∂g ∂g ik

= Gik since =1

∂g ik ∂g ik

Now,

∂g ∂g ∂g ik

j =

∂x ∂g ik ∂x j

∂gik

= Gik

∂x j

But Gik = gg ik

∂g ik ∂gik

= gg

∂x j

∂x j

1 ∂g ∂g

j =

g ik ikj

g ∂x ∂x

1 ∂g ∂g ik

= g ik { [ jk , i] + [ij, k ] } as = [ jk , i] + [ij, k ]

g ∂x j ∂x j

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

ik ik

k i

= +

j k i j

Christoffel's Symbols and Covariant Differentiation 59

1 ∂g i i

j = + as k is dummy indices

g ∂x j i j i

1 ∂g i

= 2

g ∂x j

j i

1 ∂g i

=

2 g ∂x j

j i

∂ log( g ) i

= Proved.

∂x j

j i

EXAMPLE 1

If g ij ≠ 0 show that

β ∂ β

= ∂x j [ik , α] − i k

g αβ

∂ ( [βj, α] + [αj , β] )

∂x j i k

Solution

By Christoffel’s symbol of second kind

β

= g [ik , α]

βα

i k

Multiplying it by g αβ , we get

β

g αβ = g αβ g [ik , α ]

βα

i k

β

g αβ = [ik , α ] as g αβ g = 1

βα

i k

∂ β ∂

j αβ

g = [ik , α]

∂x i k ∂x j

∂ β β ∂g α β ∂

g αβ + j =

[

ik,α ]

∂ x j i k i k ∂ x ∂x j

∂g αβ

since = [αj, β] + [βj, α]

∂x j

∂ β β ∂

g αβ j + ([αj, β] + [βj , α]) = [ik , α]

∂x

i k i k ∂x j

60 Tensors and Their Applications

β ∂ β

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

∂

g αβ Solved.

∂x j i k

EXAMPLE 2

k

Show that if g ij = 0 for i ≠ j then (i) = 0 whenever i, j and k are distinct.

i j

i 1 ∂ log gii i 1 ∂ log g ii i 1 ∂g jj

(ii) = (iii) = (iv) = −

i i 2 ∂x i j 2 ∂x 2 gii ∂x i

i j

j j

Solution

The Christoffel’s symbols of first kind

1 ∂g jk ∂g ik ∂g ij

[ij, k ] = + − ...(1)

2 ∂x i ∂x j ∂x k

(a) If i = j = k

The equation (1) becomes

1 ∂gii

[ii, i] =

2 ∂x i

(b) If i = j ≠ k

The equation (1) becomes

1 ∂g i k ∂gi k ∂g i i

[ii, k ] = 2 ∂x i

+ i – k

∂x ∂x

Since g ik = 0 as i ≠ k (given)

1 ∂g ii 1 ∂g jj

[ii, k ] = − or [ jj, i] = −

2 ∂x k

2 ∂xi

(c) i = k ≠ j

1 ∂g ji ∂g ii ∂g ij

[ij,i ] = + j − i

2 ∂x i ∂x ∂x

1 ∂g ii

= , as g ij = 0, i ≠ j

2 ∂x j

(d) i ≠ j ≠ k

[ij, k ] = 0 as g ij = 0 , g j k = 0 , i ≠ j ≠ k

k

= g kl [ij, l ] since g k l = 0, k ≠ l

i j

Christoffel's Symbols and Covariant Differentiation 61

k

=0

i j

i

(ii) = g ii [ii, i ]

i i

1 ∂gii

= g ii ⋅ from (a)

2 ∂x i

1 ∂g ii g ii =

1

= i as

2 gii ∂x gii

i 1 ∂ log g ii

=

i i 2 ∂x i

(iii) i=k≠ j

i

= g [ij, i]

ii

i j

= g [ij, i] as g =

1 ii 1

ii gii

i 1 ∂ log gii

=

i j 2 ∂x j

(iv) j=k≠i

i

= g [ jj, i]

ii

j j

1 1 ∂g j j

= − from (b)

g ii 2 ∂x i

i − 1 ∂g j j

= 2g Solved.

i i ∂x

i

j j

EXAMPLE 3

1 3

(i) [22, 1] and [13, 3], (ii) and

2 2 1 3

Solution

Clearly,

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

62 Tensors and Their Applications

1 g 33 = 1

g 11 = 1, g =

22

Also, ,

r2 r sin 2 θ

2

(i) Christoffel Symbols of first kind are given by

1 ∂g jk ∂g ik ∂g ij

[ij, k ] =

2 ∂x i

+ − , i, j, k = 1, 2,3 ...(1)

∂x j ∂x k

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

1 ∂g 21 ∂g 21 ∂g 22

[22, 1] = + 2 − 1 Since g21 = 0.

2 ∂x 2 ∂x ∂x

1 ∂0 ∂0 ∂r 2

= 2 + 2 − 1

2 ∂x ∂x ∂x

1 ∂r 2

= − = −r

2 ∂r

Taking i = 1, j = k = 3 in (1)

[13,3] = + 3 − 3

2 ∂x1 ∂x ∂x

1 ∂r 2 sin 2 θ

= since g13 = 0

2 ∂r

[13, 3] =

r sin 2 θ

(ii) Christoffel symbols of the second kind are given by

k

= g kl [ij, l] = g k 1 [ij ,1] + g k 2 [ ij ,2] + g k 3 [ ij,3]

i j

Taking k = 1, i = j = 2.

1

= g 11 [22, 1] + g 12 [22, 2 ] + g13 [22, 3]

2 2

1

= 1[22, 1] + 0[22, 2 ]+ 0[22, 3] Since g 12 = g 13 = 0

2 2

1

= −r

2 2

Christoffel's Symbols and Covariant Differentiation 63

and

3

= g [13, 1] + g [13, 2] + g [13, 3]

31 32 33

1 3

=

1

[13, 3] Since g 31 = g 32 = 0

r sin θ

2 2

3 1 1

= 2 2 ⋅ r sin θ =

2

1 3 r sin θ r

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

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

1 ∂gik ∂g ik ∂g ij

[ ij , k ] = + − ...(1)

2 ∂x i ∂x j ∂x k

∂x p ∂x 2

g ij = g pq ...(2)

∂x i ∂x j

∂g ij ∂ ∂x p ∂x q

∂x k

=

∂x k ∂x i ∂x j g pq

∂ ∂x p ∂x q q ∂g

g pq + ∂x ∂x

p

pq

=

∂x k ∂x i ∂x j ∂x ∂x ∂x k

i j

∂g i j ∂ 2 x p ∂x q ∂x p ∂ 2 x q q ∂g

g pq + ∂x ∂x pq ∂x

p r

∂xk

= ∂x k ∂x i ∂x j + ∂x i ∂x k ∂x j ∂x i ∂x j ∂x r ∂x k

...(3)

Interchanging i, k and also interchanging p, r in the last term in equation (3)

∂g kj ∂ 2 x p ∂x q ∂x p ∂ 2 x q q ∂g

g pq + ∂x ∂x rq ∂x

r p

= i k + k ...(4)

∂x i ∂x ∂x ∂x ∂x ∂x i ∂x j ∂x k ∂x j ∂x p ∂x i

j

64 Tensors and Their Applications

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

∂ g ik ∂ 2 x p ∂x q ∂x p ∂ 2 x q q ∂g

g pq + ∂x ∂x ∂x

p r

= j i + i

pr

...(5)

∂x j ∂x ∂x ∂x

i

∂x ∂x k ∂x j ∂ x i

∂x k

∂ x j

∂ x q

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

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

[ij, k ] = 2 i j k g pq + i q + −

2 ∂x ∂x ∂x ∂x ∂x j ∂x k ∂x ∂x p ∂x r

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

[ij, k ] = g + + p −

∂x i ∂x i ∂x k 2 ∂x q

pq

∂x i ∂x j ∂x k ∂x ∂x r

∂ 2 x p ∂x q ∂x p ∂x q ∂x r

[ij, k ] = g + [ pq, r] ...(6)

∂x i ∂x j ∂x k

pq

∂x i ∂x i ∂x k

It is law of transformation of Christoffel's symbol of the first kind. But it is not the law of

transformation of any tensor due to presence of the first term of equation (6).

So, Christoffel's symbol of first kind is not a tensor.

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

k k

Let g kl

[ij , l ] = is function of coordinates x i and g [ij, l ] =

kl

in another coordinate system

i j i j

i

x . Then

∂ 2 x p ∂x q ∂x p ∂x q ∂x r

[ij, l ] = g pq + i [ pq, r ] from (6)

∂x ∂x ∂x

i j l

∂x ∂x j ∂x l

kl

As g is contravariant tensor of rank two.

∂x k ∂x l st

g kl = g

∂x s ∂x t

Now

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

g kl [ij, l ] = g g + g [pq, r]

∂x s ∂x t ∂x i x j ∂x l

pq

∂x s ∂x t ∂x i ∂x j ∂x l

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

∂x t ∂x l ∂x i ∂x j g [ pq, r ]

g g +

=

∂x s ∂x t ∂x l ∂x i x j pq

∂x s

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

= δ g g + δ g [ pq , r ] as = δqt

∂x

t

∂x x

pq

∂x

t

s i j s

∂x ∂x

i j

∂x ∂x

t l

∂x k ∂ 2 x p sq ∂x k ∂x p ∂x q sr

= g g + g [ pq, r ] as δ rt g st = g sr

∂x s ∂x i ∂x j

pq

∂x s ∂x i ∂x j

Christoffel's Symbols and Covariant Differentiation 65

∂x p ∂ 2 x p s ∂x k ∂x p ∂x 2 s

g kl [ij, l ] = δp + s

∂x s ∂x i ∂x j ∂x ∂x i ∂x j p q

s

= δ sp and g [ pq, r ] = p q

sr

Since g sq g pq

k ∂x k ∂ 2 x s ∂x k ∂x p ∂x q s

= + ....(7)

i j ∂x s ∂x i ∂x j ∂x s ∂x i ∂x j p q

It is law of transformation of Christoffel’s symbol of the second kind. But it is not the law of

transformation of any tensor. So, Christoffel’s symbol of the second kind is not a tensor.

∂x s

Also, multiply (7) by , we get

∂x k

∂x s k ∂x s ∂x k ∂ 2 x s ∂x s ∂x k ∂x p ∂x q s

= +

dx k i j ∂x k ∂x s ∂x i ∂x j ∂x k ∂x s ∂x i ∂x j p q

∂x s ∂x k

Since = δ ss = 1

∂x k ∂x s

∂x s k ∂ 2 xs ∂x p ∂x q s

= +

∂x k i j ∂x i ∂x j ∂x i ∂x j p q

∂ 2 xs ∂x s k ∂x p ∂x q s

= i j − i j ...(8)

∂x i ∂x j ∂x k ∂x ∂x p q

kind and first derivatives.

THEOREM 4.4 Prove that the transformation of Christoffel’s Symbols form a group i.e., possess the

transitive property.

Proof: Let the coordinates x i be transformed to the coordinate system x i and x j be transformed to x i .

When coordinate x i be transformed to x i , the law of transformation of Christoffel’s symbols of

second kind (equation (7)) is

k ∂x k ∂ 2 x s ∂x k ∂ x p ∂ x q s

= + ...(1)

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

r k ∂x i ∂x j ∂x r ∂ 2 x k ∂x r

= +

i j ∂x ∂x ∂x ∂x u ∂x v ∂x k

u v k

u v

66 Tensors and Their Applications

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

= p q +

∂x s ∂x i ∂x j ∂x u ∂x v ∂x k ∂x i ∂x j ∂x s ∂x u ∂x v ∂x k

∂ 2 x k ∂x r

+

∂x u ∂x v ∂x k

r s ∂x p ∂x q ∂x r 2 k

∂ x ∂x

r

= + +

u v p q ∂x ∂x ∂x

u v s u v k

∂x ∂x ∂x

∂ 2 x s ∂x r ∂x i ∂x j

...(2)

∂x i ∂x j ∂x s ∂x u ∂x v

∂x

p

∂x

i

∂x

p

as i u = u

∂x ∂x ∂x

Since we know that

∂x s ∂x i ∂x s

= ...(3)

∂x i ∂x u ∂x u

∂ ∂x s ∂x i ∂x s ∂ ∂x i

= ∂ x

2 s

∂x v ∂x i ∂x u + ∂x i ∂x v ∂x u

∂x u ∂x v

∂ 2 x s ∂x j ∂x i ∂x s ∂ 2 x i ∂ 2 xs

+ = ...(4)

∂x i ∂x j ∂x v ∂x u ∂x i ∂x u ∂x v ∂x u ∂x v

∂x r

Mutiply (5) by .

∂x s

∂ 2 x s ∂x j ∂x i ∂x r ∂ 2 x i ∂x s ∂x r ∂ 2 x s ∂x r

+ =

∂x i ∂x j ∂x v ∂x u ∂x s ∂x u ∂x v ∂x s ∂x s ∂x u ∂x v ∂x s

Replace dummy index i by k in second term on L.H.S.

∂ 2 x s ∂x j ∂x i ∂x r ∂ 2 x k ∂x r ∂ 2 x s ∂x r

+ = ...(5)

∂x i ∂x j ∂x v ∂x u ∂x s ∂x u ∂x v ∂x k ∂x u ∂x v ∂x s

Using (5) in equation (2), we get

r s ∂x p ∂x q ∂x r ∂ 2 x s ∂x r

= u v + ...(6)

p q ∂x ∂x ∂x ∂x u ∂x v ∂x s

s

u v

Christoffel's Symbols and Covariant Differentiation 67

The equation (6) is same as the equation (1). This shows that if we make direct transformation

from x i to x i we get same law of transformation. This property is called that transformation of

Christoffel's symbols form a group.

Let Ai and Ai be the components of a covariant vector in coordinate systems x i and x i respectively..

Then

∂x p

Ai = Ap ...(1)

∂x i

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

∂Ai ∂2 x p ∂x p ∂A p

= A +

∂x j ∂x j ∂x i

p

∂x i ∂x j

∂Ai ∂2 x p ∂x p ∂A p ∂x q

= A + ...(2)

∂x j ∂x j ∂x i

p

∂x i ∂x q ∂x j

It is not a tensor due to presence of the first term on the R.H.S. of equation (2).

Now, replace dum m y index p by s in the first term on R.H.S. of (2), we have

∂Ai ∂ 2 xs ∂x p ∂A p ∂x q

j =

As + i ...(3)

∂x ∂x ∂x

j i

∂x ∂x q ∂x j

Since we know that from equation (8), page 65,

∂ 2 xs ∂x s k ∂x p ∂x q s

= − i j

∂x i ∂x j ∂x k i j ∂x ∂x p q

∂ 2 xs

Substituting the value of in equation (3), we have

∂x i ∂x j

∂Ai ∂x s k ∂x p ∂x q s ∂x p ∂A p ∂x q

= k − s A +

∂x j ∂x i j ∂x i ∂x j p q ∂x i ∂x q ∂x j

k ∂x s ∂x p ∂x q s ∂x p ∂x q ∂A p

= k As − i As + i

i j ∂x ∂x ∂x j p q ∂x ∂x j ∂x q

∂Ai k ∂x p ∂x q ∂A p s

A + q − As

∂x j

= k

∂x i ∂x j ∂x

i j p q

68 Tensors and Their Applications

∂Ai k ∂x p ∂x q ∂A p s

− Ak =

∂x i ∂x j ∂x q − As p q ...(4)

∂x j i j

Now, we introduce the comma notation

∂Ai k

Ai , j = − Ak ...(5)

∂x j

i j

Using (5), the equation (4) can be expressed as

∂x p ∂x q

Ai , j = Ap, q ..(6)

∂x i ∂x j

It is law of transformation of a covariant tensor of rank two. Thus, Ai , j is a covariant tensor of

rank two.

So, Ai , j is called covariant derivative of Ai with respect to x j .

Let Ai and A i be the component of contravariant vector in coordinate systems x i and x i respectively..

Then

∂x i s

Ai = A

∂x s

∂x s i

or As = A

∂x i

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

∂A s ∂ 2 xs ∂x s ∂A i

= A i

+ ...(1)

∂x j ∂x j ∂x i ∂x i ∂x j

Since from equation (8) on page 65,

∂ 2 xs ∂ x s k i ∂ x p ∂x q s

= A −

∂x j ∂x i ∂ x k i j ∂ x i ∂x j p q

∂ 2 xs

substituting the value of in the equation (1), we get

∂x j ∂x i

∂A s ∂x s k i ∂x p ∂x q s i ∂x s ∂A i

= A − i A + i

∂A j ∂x k i j ∂x ∂x j p q ∂x ∂x j

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

= A − i A + i

∂A q ∂x j ∂x k i j ∂x ∂x j p q ∂x ∂x

j

Christoffel's Symbols and Covariant Differentiation 69

∂x p i

Interchanging the dummy indices i and k in the first term on R.H.S. and put A = A p we get

∂x i

∂A s ∂x q ∂x s i k ∂x q p s ∂x s ∂A i

= A − j A + i

∂x q ∂x j ∂x i k j ∂x p q ∂x ∂x

j

∂x q ∂A s p s

∂x s i k ∂A i

∂x j ∂x q + A p q = ∂x i k A + j

j ∂x

i ∂A s

∂A i ∂x i ∂x q p s

+ Ak = + A ...(2)

∂x

∂x ∂x s ∂x j p q

j

k j q

Now, we introduce the comma notation

∂Ai i

+ Ak

A,i j = ...(3)

∂x j

k j

Using (3), the equation (2) can be expressed as

∂x p ∂x q

A,ij = Ap, q ...(4)

∂x i ∂x j

It is law of transformation of a mixed tensor of rank two. Thus, Ai , j is a mixed tensor of rank

two. Ai , j is called covariant derivative of Ai with respect to x j .

Covariant derivative of a covariant tensor of rank two.

Let Ai j and Ai j be the components of a covariant tensor of rank two in coordinate system x i and x i

respectively then

∂x p ∂x q

Aij = A pq ...(1)

∂x i ∂x j

∂Aij ∂x p ∂x q ∂A pq ∂ ∂x p ∂x q

+ k A pq

∂x k

=

∂x ∂x ∂x

i j k

∂x ∂x i ∂x j

∂Aij ∂x p ∂x q ∂Apq ∂x r ∂ 2 x p ∂x q ∂x p ∂ 2 x q

= + A + A pq ...(2)

∂x k ∂x i ∂x j ∂x r ∂x k ∂x k ∂x i ∂x j

pq

∂x i ∂x k ∂x j

∂A pq ∂A pq ∂x r

as = (since A pq components in x i coordinate)

∂x k ∂x r ∂x k

70 Tensors and Their Applications

∂ 2 x p ∂x q ∂ 2 x l ∂x q

A pq = A

∂x i ∂x k ∂x j lq

∂x i ∂x k ∂x j

∂ 2 x p ∂x q ∂x q h ∂x l l ∂x p ∂x r

A pq = A h − i

∂x i ∂x k ∂x j lq

∂x j i k ∂x p r ∂x ∂x

k

∂ 2 xl h ∂x l l ∂x p ∂x r

k = h − i

∂x ∂x

i

i k ∂x p r ∂x ∂x

k

h ∂x l ∂x 2 l ∂ x p ∂ x q ∂x r

∂ 2 x p ∂x q −

A pq

= i k Alq Alq

∂x i ∂x k ∂x j ∂x h ∂ x j p r ∂x i ∂x j ∂x k

∂ 2 x p ∂x q h l ∂x p ∂x q ∂x r

A pq j = Ahj − Alq i ...(3)

∂x ∂x ∂x

i k

i k p r ∂x ∂x j ∂x k

∂x l ∂x q

as Ahj = Alq by equation (1)

∂x h ∂x j

and

∂x p ∂ 2 x q ∂x p ∂ 2 x l

Apq = A

∂x i ∂x j ∂x k pl

∂x i ∂x j ∂x k

∂x p h ∂x l l ∂x q ∂x r

= Apl h − j

∂x i j k ∂x q r ∂x ∂x

k

h ∂x p ∂x l l ∂x q ∂x r ∂x p

= pl

A − Apl

j k ∂x i ∂x h q r ∂x j ∂x k ∂x i

∂x p ∂ 2 x q h l ∂x p ∂x q ∂x r

Apq = A − i A pl ...(4)

∂x i ∂x j ∂x k

ih

j k q r ∂x ∂x ∂x

j k

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

∂Aij ∂A pq l l ∂x p ∂x q ∂x r h h

= r − Alq p r − A pl q r + Ahj + Aih

∂x k ∂x ∂x ∂x ∂x

i j k

i k j k

∂Aij h h ∂A pq l l ∂x p ∂x q ∂x r

− Aih − Ahj = − Alq − A pl

∂x k j k i k ∂x

r

p r q r ∂x ∂x ∂x

i j k

Christoffel's Symbols and Covariant Differentiation 71

∂Aij h h

Aij , k = − Aih − Ahj , then

∂x k

j k i k

∂x p ∂x q ∂x r

Aij , k = Apq , r

∂x i ∂x j ∂x k

It is law of transformation of a covariant tensor of rank three. Thus, A ij,k is a covariant tensor

of rank three.

So, Aij , k is called covariant derivative of Aij w.r.t. to x k .

Similarly we define the covariant derivation xk of a tensors A ij and Aij by the formula

∂A ij i j

ij

A ,k = + A lj + A il

∂x k

l k l k

∂A ij i l

and Aij ,k = + A lj − Ali

∂x k

l k j k

In general, we define the covariant deriavative xk of a mixed tensor Aab

ij ... l

... c

by the formula

∂Aab

ij ...l

pj ... l i ip... l j ij... p l

ij ... l

Aab = ...c

+ Aab ...c + Aab... c + ⋅ ⋅ ⋅ + Aab... c

∂x

... c ,k k

p k p k p k

p ij... l p ij ... l p

− A ijpb......l c − Aap... c − ⋅ ⋅ ⋅ − Aab ... p

a k b k c k

Note: Ai , k is also written as Ai , k = ∇ k Ai .

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

Proof: The covariant derivative xp of Kronecker delta is

∂δ ij i l

δ ij,k = + δ lj − δ il

∂x k

l k j k

i i

= 0+ −

j k j k

∂δ ij i i

δ ij,k = 0 as = 0; δlj =

∂x k

l k j k

Also, consider first the tensor gij and the covariant derivative of gij is

∂g ij m m

g ij , k = − g mj − gim

∂x k

i k j k

72 Tensors and Their Applications

∂g ij m

g ij , k = − [ik , j ]− [ jk , i] as g mj = [ik , j ]

∂x k

i k

∂g ij

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

∂x k

So, g ij , k = ∂g ij − ∂g ij

∂x k ∂x k

g ij , k = 0

W e can perform a sim ilar calculation for the tensor gij.

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

k g mj + g g mj , k = δ j , k

g ,im im i

im

But

EXAMPLE 4

Prove that if A ij is a symmetric tensor then

Ai j, j =

1 ∂

( 1

Ai j g − A jk )

∂g jk

g ∂x ∂x i

j

2

Solution

A ij = A ji ...(1)

We know that

∂Ai j j l

Ai j, k = + Ail − Al j

∂x k

l k i k

Put k = j, we get

∂Ai j j l

Ai j, j = + Ail − Al j ...(2)

∂x j

l j i j

=

∂Ai j

+ Ail

(

∂ log g )

− Al j g hl [ ij, h]

∂x j

∂x l

∂Ai j Ai j ∂ g

= + − A jh [ij, h] since A ij is symmetric.

∂x j g ∂x j

Ai j, j =

(

1 ∂ Ai j g )

− A jk [ij, k ] ...(3)

g ∂x j

Christoffel's Symbols and Covariant Differentiation 73

But

1 jk ∂g jk ∂g ki ∂g ij

A jk [ij , k ] = A i + −

2 ∂x ∂x j ∂x k

1 jk ∂g jk jk ∂g ki

∂g ij

A jk [ij , k ] = A + −

jk

A A

2 ∂x i ∂x j ∂x k

∂g ki ∂g

A jk

j =

A kj kji ...(4)

∂x ∂x

On Interchanging the dummy indices j & k.

∂g ki ∂g ji

A jk jk

= A since Aij = A ji

∂x j ∂x k

∂g ki ∂g

⇒ A jk − A jk ijk = 0 as g ij = g ji ...(5)

∂x j

∂x

Using (5), equation (4) becomes

1 jk ∂g jk

A jk [ij , k ] = A

2 ∂x i

Ai j, j =

(

1 ∂ Ai j g 1)

− A jk

∂g jk

Proved.

g ∂x j

2 ∂x i

EXAMPLE 5

k k k k

Prove that − are components of a tensor of rank three where and

i j a i j b i j a i j b

are the Christoffel symbols formed from the symmetric tensors aij and bij .

Solution

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

∂ 2 xs ∂x s k ∂x p ∂x q s

= − i j

∂x i ∂x j ∂x k i j ∂x ∂x p q

∂x s k ∂ 2 xs ∂x p ∂x q s

= +

∂x k i j ∂x i ∂x j ∂x i ∂x j p q

k ∂ 2 xs ∂x p ∂x q s ∂x k

or = i j + s

i j ∂x ∂x ∂x i ∂x j p q ∂x

74 Tensors and Their Applications

k ∂ 2 xs ∂x p ∂x q s ∂x k

= + s

∂x ∂x ∂x i ∂x j p q a ∂x

i j

i j a

k ∂ 2 xs ∂x p ∂x q s ∂x k

and = i j+ i s

i j b ∂x ∂x ∂x ∂x j p q b ∂x

Subtracting, we obtain

k k s s ∂x p ∂x q ∂x k

=− − i

p q a p q b ∂x ∂x ∂x

j s

i j a i j b

Put

s s

− = A pq

s

p q a p q b

Then above equation can written as

s ∂x p ∂x q ∂x k

Aijk = Apq

∂x i ∂x j ∂x s

It is law of transformation of tensor of rank three.

k k

So, − are components of a tensor of rank three.

i j a i j b

EXAMPLE 6

If a specified point, the derivatives of gij w.r.t. to xk are all zero. Prove that the components of

covariant derivatives at that point are the same as ordinary derivatives.

Solution

Given that

∂g ii

= 0, ∀ i , j, k at P 0 ...(1)

∂x k

Let Aij be tensor..

∂Aij

Now, we have to prove that Aij , k = at P0 .

∂x k

∂A ij i i α

Aij ,k = + A αj − Aα ...(2)

∂x α k

k

j k

Christoffel's Symbols and Covariant Differentiation 75

k ∂g ij

Since and [ij,.k ] both contain terms of the type and using equation (1) we get

i j ∂x k

k

= 0 = [ij, k] at P0 .

i j

So, equation (2) becomes

∂A ij

Aij ,k = at P0

∂x k

(a) Gradient

If φ be a scalar function of the coordinates, then the gradient of φ is denoted by

∂φ

grad φ =

∂x i

which is a covariant vector.

(b) Divergence

The divergence of the contravariant vector Ai is defined by

∂A i i

div A i = + Ak

∂x i

k i

It is also written as A,ii

The divergence of the covariant vector Ai is defined by

div Ai = g ik Aik

EXAMPLE 7

i (

1 ∂ g Ak )

g ∂x k

Solution:

∂A i i

div A i = A, i = + Ak

i

∂x i

k i

Since

i

=

∂

∂x

log g = (

1 ∂ g

g ∂x k

)

k i

k

So,

∂A i 1 ∂ g k

div A i = + A

∂x i

g ∂x k

76 Tensors and Their Applications

∂A k 1 ∂ g k

div A i = + A

∂x k

g ∂x k

div A i =

1 ∂ g Ak ( ) ...(1)

g ∂x k

Proved

(c) Curl

Let Ai be a covariant vector then

∂Ai k

Ai , j = − Ak

∂x j

i j

∂A j k

and Aj, i = − Ak

∂x i

j i

are covariant tensor.

∂Ai ∂A j

So, Ai , j − A j , i = − is covariant tensor of second order, which is called curl of Ai .

∂x j ∂x i

Thus

curl Ai = Ai , j − A j ,i

Note: curl A i is a skew-symmetric tensor.

Since

A j, i –A i,j = – (A i,j –A j,i)

EXAMPLE 8

If Aij be a skew-symmetric tensor of rank two. Show that

∂Aij ∂A jk ∂Aki

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

∂x k

∂x i

∂x j

Solution

Since we know that

∂Aij l l

Aij , k = − Alj − Ail

∂x k

i k j k

∂A jk h l

A jk ,i = − Alk − A jl

∂x i

j i j k

∂Aki l l

Aki, j = − Ali − Akl

∂x j

k j i j

Christoffel's Symbols and Covariant Differentiation 77

∂Aij ∂A jk ∂Aki l l

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

∂x j i k

lj jl

∂x k ∂x i k i

l l l l

− Ali + Ail − Akl + Alk

k j j k i j j i

l l l

Since is symmetric i.e., = etc.

i k i k k i

∂Aij ∂A jk ∂Aki l

= + + − ( Alj + A jl )

∂x k

∂x i

∂x j i k

l l

− ( Ali + Ail ) − ( Akl + Alk )

k j i j

Since Aij is skew-symmetric. Then Alj = − A jl ⇒ Alj + A jl = 0 . Similarly,,

Ali + Ail = 0 and Akl + Alk = 0

So,

∂Aij ∂A jk ∂Aki

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

∂x k

∂x i

∂x j

THEOREM 4.5 A necessary and sufficient condition that the curl of a vector field vanishes is that

the vector field be gradient.

Proof: Suppose that the curl of a vector Ai vanish so that

curl Ai = Ai , j − A j , i = 0 ...(1)

To prove that Ai = ∇φ, φ is scalar..

Since from (1),

Ai , j − A j ,i = 0

∂Ai ∂A j

⇒ − =0

∂x j ∂xi

∂Ai ∂A j

⇒ =

∂x j

∂x i

∂Ai ∂A j j

⇒

j

dx = dx

∂x j ∂xi

⇒ dAi =

∂

∂x i

A j dx j ( )

Integrating it we get

Ai = ∫

∂

∂x i

(

A j dx j )

78 Tensors and Their Applications

∂

∂x

=i ∫

A j dx j

∂φ

Ai =

∂x i

, where φ = ∫ A j dx j

or Ai = ∇φ.

Conversely suppose that a vector Ai is such that

Ai = ∇φ, φ is scalar..

To prove curl Ai = 0

Now,

∂φ

Ai = ∇φ =

∂x i

∂Ai ∂ 2φ

=

∂x j ∂x j ∂xi

∂A j ∂ 2φ

and =

∂x i ∂x i ∂x j

∂Ai ∂A j

So, − =0

∂x j ∂xi

∂Ai ∂A j

So, curl Ai = Ai , j − A j ,i = − =0

∂x j ∂x i

THEOREM 4.6 Let φ and ψ be scalar functions of coordinates xi. Let A be an arbitrary vector then

(i) div (φA) = φ div A + A ⋅ ∇φ

(ii) ∇(φψ) = φ∇ψ + ψ∇φ

div A i =

(

1 ∂ g Ai ) ...(1)

g ∂x i

replace A i by φA i , we get

div( φA ) =

i

(

1 ∂ g φAi )

g ∂x i

Christoffel's Symbols and Covariant Differentiation 79

1 ∂ ( g Ai ) i ∂φ

= φ + g A ⋅

∂x i ∂x

i

g

1 ∂( g A )

i

∂φ

⋅A +φ

i

=

∂x

i

∂x

i

g

= ∇φ ⋅ A i + φ div A i

Thus

div( φA) = ∇φ ⋅ A + φ div A ...(2)

(ii) By definition of gradient,

∂(φψ )

∇(φψ) =

∂x i

∂ψ ∂φ

= φ +ψ i

∂x i

∂x

Thus ∇(φψ) = φ∇ψ + ψ∇φ ...(3)

(iii) Taking divergence of both sides in equation (3), we get

div (∇φψ) = div [φ∇ψ + ψ∇φ ]

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

Thus,

∇ 2 (φψ) = φ∇ 2 ψ + ψ∇ 2 φ + 2∇φ ⋅ ∇ψ

(iv) Replace A by ∇ψ in equation (3), we get

div (φ∇ψ) = ∇φ ⋅ ∇ψ + φ div (∇ψ)

div (φ∇ψ) = ∇φ ⋅ ∇ψ + φ∇ 2 ψ

(i) curl (φA) = A × ∇φ + φ curlA

(ii) curl (ψ∇φ) = ∇φ × ∇ψ

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

curl A = curl Ai = Ai , j − A j ,i

Replacing Ai by φAi , we get

( )

curl (φAi ) = (φAi ), j − φA j , i

80 Tensors and Their Applications

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

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

= Ai × ∇φ + φcurl Ai

So,

curl (φA) = A × ∇φ + φcurlA ...(1)

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

curl (φ∇ψ) = φcurl (∇ψ) + ∇ψ × ∇φ

Interchange of φ and ψ , we get

curl (ψ∇φ) = ψcurl (∇φ) + ∇φ × ∇ψ.

Since curl (∇φ ) = 0.

So,

curl (ψ∇φ) = ∇φ × ∇ψ. Proved.

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

1 ∂ ∂φ

∇ 2φ = g g kr r

g ∂x ∂x

k

Proof: Since

∇ 2 φ = div grad φ ...(1)

and

∂φ

grad φ = ,

∂x r

which is covariant vector.

But we know that any contravariant vector A k associated with Ar (covariant vector) is

A k = g kr Ar (Sec Art. 3.4, Pg 43)

∂φ

Now, the contravariant vector A k associated with (Covariant vector) is

∂x r

∂φ

A k = g kr

∂x r

Since

div A i =

1 ∂ g Ak( )

, (Sec Ex. 7, Pg 75)

g ∂x k

Christoffel's Symbols and Covariant Differentiation 81

∂φ

∂ g g kr r

∂φ 1 ∂ x

∇ 2 φ = div g kr r = Proved.

∂x g ∂x k

EXAMPLE 9

Show that, in the cylindrical coordinates,

1 ∂ ∂V 1 ∂ 2V ∂ 2V

∇ 2V = r + +

r ∂r ∂r r 2 ∂θ 2 ∂z 2

by Tensor method

Solution

The cylindrical coordinates are (r , θ, z ). If V is a scalar function of (r , θ, z ).

Now,

∇ 2V = ∇ ⋅ ∇V

Since

∂V ∂V ∂V

∇V = i +j +k

∂r ∂θ ∂z

Let

∂V ∂V ∂V

A1 =, A2 = , A3 = ...(1)

∂r ∂θ ∂z

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

is

ds 2 = dr 2 + r 2 dθ2 + dz 2

here, x1 = r , x 2 = θ, x 3 = z .

Since ds 2 = g ij dx i dx j

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

and others are zero.

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

g 31 g 32 g33 0 0 1

Now,

(

1 ∂ gA

k

)

g ∂x k

82 Tensors and Their Applications

1 ∂( g A )

1

1 ∂( g A )

2

∂( g A )

3

= + +

∂x

1

∂x

2

∂x

3

g g

1 ∂ ( rA1 ) ∂ ( rA 2 ) ∂ ( rA3 )

div (∇V ) = + + ...(2)

r ∂r ∂θ ∂z

We can write

A k = g kq Aq (Associated tensor)

A k = g k1 A1 + g k 2 A2 + g k 3 A3

Put k = 1

A1 = g A1 + g A2 + g A3

11 12 13

A1 = g 11 A1 as g 12 = g 13 = 0

Similarly,

A 2 = g 22 A2

A3 = g 33 A3

and

Cofactor of g11 in g r 2

g 11 = = 2 =1

g r

Cofactor of g 22 in g 1

g 22 = = 2

g r

Cofactor of g 33 in g r 2

g 33 = = 2 = 1 (See. Pg. 34, Ex.1)

g r

So,

A1 = g A1 = A1

11

1

A 2 = g A2 = 2 A2

22

r

A3 = g A3 = A3 .

33

or A1 = A1, A 2 = 1 A2 , A3 = A3 .

r2

from (1), we get

∂V 1 ∂V ∂V

A1 = ∂r , A = r 2 ∂θ , A = ∂z

2 3

from (2),

Christoffel's Symbols and Covariant Differentiation 83

So,

1 ∂ ∂V ∂ 1 ∂V ∂ ∂V

∇ 2V = div A = r ∂r r ∂r + ∂θ r r 2 ∂θ + ∂ z r ∂z

i

1 ∂ ∂V ∂ 1 ∂V ∂ ∂V

div (∇V ) = r + + r

r ∂r ∂r ∂θ r ∂θ ∂z ∂z

1 ∂ ∂V 1 ∂ 2 V ∂ 2V

= r + + r

r ∂r ∂r r ∂θ 2 ∂z 2

1 ∂ ∂V 1 ∂ 2V ∂ 2V

div (∇V ) = r + +

r ∂r ∂r r 2 ∂θ 2 ∂z 2

1 ∂ ∂V 1 ∂ 2V ∂ 2V

∇ 2V = r + +

r ∂r ∂r r 2 ∂θ 2 ∂z 2

EXERCISES

∂Aij α α

(a) Aij , l = − Aαj − Ai α

∂x i l jl

l

∂Air jk α r α α r Aα

Air jk ,l = − A − Ar − Ar

l αjk j l i αk k l ijα + α l ijk

(b)

∂x l

i

2. Prove that

j

1 ∂ ( Ai g) j k

Ai j, j = j

− Ak

g ∂x i j

1 ∂

3. If A ijk is a skew-symmetric tensor show that ( g Ai jk ) is a tensor..

g ∂xk

4. Prove that the necessary and sufficient condition that all the Christoffel symbols vanish at a point is

that g ij are constant.

5. Evaluate the Christoffel symbols in cylindrical coordinates.

6. Define covariant differentiation of a tensor w.r. to the fundamental tensor gij . Show that the covariant

differentiation of sums and products of tensors obey the same result as ordinary differentiation.

7. Let contravariant and covariant components of the same vector A be Ai and Ai respectively then

prove that

i

div A = div Ai

84 Tensors and Their Applications

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

10. A necessary and sufficient condition that the covariant derivative vector be symmetric is that the

vector must be gradient.

11. Show that, in spherical coordinates

1 ∂ 2 ∂V 1 ∂ ∂V 1 ∂ 2V

2

∇ V = r + 2 sin θ + 2 2

r ∂r ∂ r r sin θ ∂ θ

2

∂ θ r sin θ ∂ φ2

by tensor method.

CHAPTER – 5

RIEMANN-CHRISTOFFEL TENSOR

If A i is a covariant tensor then the covariant x j derivative of A i is given by

∂Ai α

A i,j = − Aα ...(1)

∂x j i j

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

∂Ai , j α α

Ai , jk = − Aα, j − Ai , α

∂x k

i k j k

∂ ∂A α α ∂A β

i − A − α −

= ∂x α

Aβ

∂x k j

i j i k ∂ x j

α j

α ∂A γ

− αi − Aγ

i k ∂x i k

α

∂

∂ Ai

2

i j α ∂ Aα α ∂ Aα

Ai , jk = − Aα − −

∂x k ∂x ∂x i k ∂x i k ∂ x

j k k j

α β α ∂Ai α γ

+ Aβ − α + Aγ ...(2)

i k α j j k ∂x j k i α

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

α

∂

∂ Ai

2

i k α ∂A α ∂A

Ai , kj = − Aα − αj − αk

∂x j ∂x k ∂x j i k ∂x i j ∂x

86 Tensors and Their Applications

α β α ∂Ai α γ

` + Aβ − α + Aγ ...(3)

i j α k k i ∂x k j i α

Subtract equation (3) from (2), we get

α

∂ α

∂

α β i j α β

Ai , jk − Ai , kj = Aβ − Aα − Aβ + i k A

α

i k α j ∂x k

i j α k ∂x j

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

α α

∂ ∂

i k − i j β α β α Aα

Ai , jk − Ai , kj = + −

β k

...(4)

∂x j ∂x k i k β j i j

Ai , jk − Ai , kj = Aα Riαjk ...(5)

where

α α

∂ ∂

i k i j β α β α

Riαjk = − + − ...(6)

∂x j ∂xk i k β j i j β k

Since A i is an arbitrary covariant tensor of rank one and difference of two tensors A i, jk – Ai, kj is

a covariant tensor of rank three. Hence it follows from quotient law that Riαjk is a mixed tensor of rank

four. The tensor Riαjk is called Riemann Christoffel tensor or Curvature tensor for the metric g ij dx i dx j .

The symbol Riαjk is called Riemann’s symbol of second kind.

Now, if the left hand side of equation (4) is to vanish i.e., if the order of covariant differentiation

is to be immaterial then

Riαjk = 0

Since A α is arbitrary. In general Riαjk ≠ 0, so that the order of covariant differentiation is not

immaterial, It is clear from the equation (4) that “a necessary and sufficient condition for the validity of

inversion of the order of covariant differentiation is that the tensor Riαjk vanishes identically..

Remark

The tensor

∂ ∂ i i

∂x k ∂x α k α l

l

i

= i + ...(7)

Rikl i α α

j k

j l j k j l

Riemann-Christoffel Tensor 87

THEOREM 5.1 Curvature tensor Riαjk is anti symmetric w.r.t. indices j and k.

Proof: We know that from (7) curvature tensor

∂ ∂ α α

∂x j ∂ xk β j β k

Riαjk = α α +

β β

i j i k i j i k

α α

∂ ∂

i k i j α β α β

Riαjk = − + −

∂x j ∂x k β j i k β k i j

Interchanging j and k, we get

α α

∂ ∂

i j − i k + α β − α β

R iαkj = β k i j β j i k

∂x k ∂x j

α α

∂ ∂

i k i j α β α β

= − − + −

∂x ∂x k β j i k β k i j

j

Rikα j = – Riαjk

Riαjk + R αjk i + R kiα j = 0

Proof: Since we know that

∂ ∂ α α

∂x j

∂x k β j β k

+

= α α

α

Rijk β β

i j i k i j i k

α α

∂ ∂

i j α β β α

Rijαk = i k − + − ...(1)

∂x j ∂x k β j i k i j β k

88 Tensors and Their Applications

Similarly

α α

∂ ∂

α β β α

= k − i +

j i j k

R αj ki − ...(2)

∂x ∂x β k j i j k β i

and

α α

∂ ∂

k j k i α β β α

Rkiα = − + − ...(3)

∂ xi ∂x j

j

β i k j k i β j

Riαj k + R αj k i + Rkαi j = 0

This is called cyclic property.

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

its lower indices

∂ ∂ α α

∂x j ∂x k β j β k

+

= α α

α

Rαjk β β

α j α k α j α k

α α

∂ ∂

α k α j α β α β

= − + −

∂x j ∂x k β j α k β k α j

∂ 2 log g ∂ 2 log g

= –

∂x j ∂ x k ∂ xk∂ x j

α ∂ log g

Since α k = and α and β are free indices Rααj k = 0 .

∂x k

Also for Riαj α .

Write R i j for R αi j α

Riemann-Christoffel Tensor 89

∂ ∂ α α

β j β α

α

∂x j

∂x

Rij = Rijαα = +

α α β β

i j i α i j i α

α α

∂ ∂

α j + α β − α β

Rij = j −

i i

β j i α β α i j

∂x ∂x α

α

∂

∂ 2 log g α β α β

− α +

i j

Rij = − ...(1)

∂x j ∂x i ∂x j i α β αi j

β

α α

∂ ∂

α i + α β − α β

R ji = i −

j j

β i j α β α j i

∂x ∂x α

α

∂

∂ 2 log g β α α β

− α +

i j

R ji = − ...(2)

∂x i ∂x j ∂x i α β j β α i j

(Since α and β are dummy indices in third term).

Comparing (1) and (2), we get

Rij = R ji

Thus Rij is a symmetric Tensor and is called Ricci Tensor..

For Riααk :

Riααk = − Rikαα = −R ik

The associated tensor

Ri j k l = g iα R αjkl ...(1)

is known as the covariant Riemann-Christoffel tensor or the Riemann-Christoffel tensor of the first

kind.

Expression for Rijkl

Rijkl = gi α Rαjkl

90 Tensors and Their Applications

∂ α ∂ α β α β α

= g iα k j l − l j k + j l β k − j k β l …(2)

∂x ∂x

Now,

∂ α ∂ α α ∂g i α

g iα =

∂x k j l gi α j l − j l ∂x k

∂x k

∂[ jl , i ] α ∂g i α

= − k ...(3)

∂x k j l ∂x

Similarly,

∂ α ∂[ jk , i ] α ∂gi α

g iα l = − l ...(4)

∂x j k ∂x l j k ∂x

udv duv vdu

By the formula dx = dx − dx

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

∂[ jl , i] α ∂gi α ∂[ jk , i ] α ∂g i α β α β α

Ri j k l = − k − + l + gi α − g iα

∂x k j l ∂ x ∂ xl j k ∂x j l β k j k β l

∂[ jl , i ] ∂[ jk , i ] α ∂g iα α ∂g iα βα β α

= − + l − j l k + g i α j l β k − g iα j k β l

∂x k

∂x l

j k ∂x ∂x

∂[ jl , i ] ∂ [ jk , i] α α

Ri j k l = − + ( [il, α ] + [αl , i ] ) − ( [ik , α] + [αk , i] )

∂x k

∂x l

j k j l

β β

+ [βk , i] − [βl , i ]

j l j k

∂[ jl , i ] ∂ [ jk , i] α α

Ri j k l = − + [il, α] − [ik , α ] ...(5)

∂x k

∂x l

j k j l

It is also written as

∂ ∂ α α

Ri j k l = ∂x k

∂x k + j k j l ...(6)

[ jk , i ] [ jl, i] [ik ,α ] [il, α ]

But we know that

1 ∂gli ∂g ji ∂g jl

[ jl, i] = + − i ...(7)

2 ∂x j ∂x l ∂x

Riemann-Christoffel Tensor 91

1 ∂g ki ∂g ji ∂g ik

and [ jk ,i ] = + − i ...(8)

2 ∂x j ∂x k ∂x

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

1 ∂ ∂g li ∂g ji ∂g jl 1 ∂ ∂g ki ∂g ji ∂g jk

Rijkl = j + − − j + k −

2 ∂x k ∂x ∂ xl ∂ xi 2 ∂x

l

∂x ∂x ∂x i

α α

+ [il, α ] − [ik , α ]

j k j l

1 ∂ g il

2 2

∂ g jk ∂ g jl

2 2

∂ g ik +

= 2 j k + − −

Rijkl

∂x ∂x ∂x ∂x

i l

∂x ∂x ∂x ∂x

j l i k

α α

[il , α] − [ik , α ] ...(9)

j k j l

Since

α α

= g αβ [ jk , β] and j l = g [ jl , β]

αβ

j k

1 ∂ 2 g il ∂ 2 g jk ∂ 2 g ik ∂ 2 g jl

Ri j k l = + − −

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

The equation (9) can also be written as

1 ∂ 2 gil ∂ 2 g jk ∂ 2 gik ∂ 2 g jl

+ − −

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

Ri j k l

α β α β

αβ

+ g αβ − g

j k i l j l i k

(iii) Rklij = Rijkl

(iv) Rijkl + Riklj + Riljk = 0

92 Tensors and Their Applications

1 ∂ 2 gil ∂ 2 g jk ∂ 2 g ik ∂ 2 g jl

Rijkl = + − −

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

α α

+ [il, α ] − [ik , α ] ...(1)

j k j l

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

1 ∂ g jl ∂ 2 g jk α

2

∂ 2 gik ∂ 2 gil

R jikl = i k = − − +

i k

[ jl, α]− iαl [ jk , α]

2 ∂x ∂x ∂x j ∂x l ∂x i ∂x l ∂x j ∂x k

1 ∂ 2 gil ∂ 2 g jk ∂ 2 g jl ∂ 2 gik α

+ [ jl , α ] − α

= − + − − [ jk , α ]

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

R jikl = − Rijkl

or Rijkl = − R jikl

(ii) Interchange l and k in equation (1) and Proceed as in (i)

(iii) Interchange i and k in (1), we get

1 ∂ 2 g kl ∂ 2 g ji ∂ 2 g ki ∂ 2 g jl α

+ [kl, α] − α [ki, α]

R kjil = + − −

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

Now interchange j and l, we get

1 ∂ g kj ∂ 2 g lj α

+ [kj, α] − α [ki, α ]

2

∂ 2 g li ∂ 2 g ki

R klij = l i+ k j − l j − k i l i

2 ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x l j

For

α α αβ β

[kj, α ] = g

l i l i k j

β αβ α

= g

k j l i

β

= [li, β]

k j

α α

[li, β] = k j [li, α]

l i

Riemann-Christoffel Tensor 93

So,

1 ∂ g kj ∂ 2 g lj α α

2

∂ 2 gli ∂ 2 g ki

R klij = + − − + [li, α] − [ki, α]

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

from equation (1)

1 ∂ g il ∂ 2 g jk ∂ 2 g jk ∂ 2 g jl α

2

Rijkl = 2 j k + i l − j l − j k

+

j k

[il , α] − jα l [ik , α ]

∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x

1 ∂ g ij ∂ 2 g kj α

+ [ij, α] − α [il , α]

2

∂ 2 g kl ∂ 2 g il

Riklj = k l + i j − k j − i l k l

2 ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x k j

1 ∂ g ik ∂ 2 g lj ∂ 2 gij ∂ 2 g lk α

2

Riljk = + − − +

l j

[ik , α] − lαk [ij, α ]

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

On adding these equations, we get

Rijkl + Riklj + R iljk = 0

Theorem 5.3 Show that the number of not necessarily independent components of curvature tensor

1 2 2

does not exceed n (n − 1) .

12

Or

Show that number of distinct non-vanishing components of curvature tensor does not exceed

1 2 2

n (n − 1) .

12

Proof: The distinct non-vanishing components of Rijkl of three types.

(i) Symbols with two distinct indices i.e., Rijij . In this case total number of distinct non-vanishing

1

components of Rijkl are n(n − 1) .

2

(ii) Symbols with three distinct indices i.e., Rijik . In this case, total number of distinct non-

1

vanishing components Rijkl are n(n − 1) (n − 2) .

2

(iii) Symbols Rijkl with four distinct indices. In this case, total number of distinct non-vanishing

n (n − 1)

2 2

components of Rijkl is .

12

94 Tensors and Their Applications

Hence the number of distinct non-vanishing components of the curvature tensor Rijkl does not

1 2 2

exceed n (n − 1) .

12

Remark

When Rijkl is of the form Riiii i.e., all indices are same

In this case, Riiii has no components.

It states that

R ijkl, m + R ijlm, k + R ijmk ,l = 0

and Rhjkl, m + R hjlm, k + Rhjmk ,l = 0

Proof: Introducing geodesic coordinate1 in which Christoffel symbols are constant with the pole at P0 .

Since we know that

∂ ∂ i i

∂x ∂x

k l

mk ml

R ijkl = +

ii m m

jk jl jk jl

i i

∂ ∂

jl − jk + i m − m i

R ijkl =

∂x k ∂x l mk jl jk ml

i i

∂2 ∂2

j l − j k

R ijkl, m = ...(1)

∂x m ∂x k ∂x m ∂x l

i m

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

So, their derivatives are zero.

∂ ∂ i i

∂x l ∂x m α l α m

+

R ijlm = i i α α

j l j m j l j m

Riemann-Christoffel Tensor 95

i i

∂ ∂

i α i α

= j m − j l + −

∂x ∂ xm

l

α l j m α m j l

i 2 i

∂2 ∂

j m − j l

R ijlm, k = ...(2)

∂x k ∂x l ∂x k ∂x m

and

∂ ∂ i i

∂x ∂x β m β k

m k

R ijmk = +

i i β β

j m j k j m j k

i i

∂ ∂

j k − j m + i β − i α

R ijmk =

∂x m ∂x k β m j k β k j m

i 2 i

∂2 ∂

j k j m

R ijmk = − ...(3)

∂x m ∂x l ∂x k ∂x l

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

R ijkl, m + R ijlm, k + R ijmk ,l = 0 ...(4)

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..

1 i

Theorem 5.4 To prove the tensor R j − δ j R is divergence free.

i

2

Proof: We know that from equation (5)

96 Tensors and Their Applications

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 jk R jm , k − g hl R hm, l = 0

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

R , m − R mk , k − Rmk , k = 0

R , m − 2 Rmk , k = 0

1

Rmk , k − R, m = 0

2

1

Rmk , k − δ km R , k = 0 since R , m = δkm R, k

2

k 1 k

Rm − δ m R = 0

2 , k

1

Rmk − δ km R is divergence free.

2

1 1 i

The tensor Rmk − δ km R = Gmk or R j − δ j R is known as Einstein Tensor..

i

2 2

Consider two unit vectors pi and qi at a point P 0 of V n. These vectors at P 0 determine a pencil of

directions deferred by ti = αpi + βqi. α and β being parameters. One and only one geodesic will pass

through P 0 in the direction of p i . Similarly one and only one geodesic will pass through in the direction

qi. These two geodesics through P 0 determined by the orientation of the unit vectors pi and qi. Let this

surface is denoted by S.

The Gaussian curvature of S at P 0 is defined to be the Riemannian Curvature of Vn at P0 for the

orientation determined by pi and qi.

Let the coordinates y i of Vn are Riemannian coordinates with origin at P 0. The equation of

surfaces S in given by

y i = ( p i α + q iβ)s ...(1)

i.e., y i = p iu 1 + q i u 2 ...(2)

Riemann-Christoffel Tensor 97

1

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

Let ds 2 = bαβ du α du β be the metric for the surface S. where

∂y i ∂ y j

bαβ = g ij (α, β = 1, 2)

∂uα ∂uβ

k γ

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

g

i j α β

Let αβγδ and R hijk be curvature tensor corresponding to the metrices b αβ du αdu β and g ij dy i dy j .

R

Since the Greek letters α, β, γ, δ take values 1,2 and so that the number of independent non-

1

vanishing components of Rαβγδ is n2 (n 2 − 1) for n = 2, i.e., they are

12

1 2 2

12

(

⋅ 2 2 − 1 = 1. Let us )

transform the coordinate system u α to u ′ α and suppose that the corresponding value of R1212 are

'

R1212 .

Then

∂u α ∂u β ∂u γ ∂u δ

'

R1212 = Rαβγδ

∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2

∂u1 ∂u β ∂u γ ∂u δ ∂u 2 ∂u β ∂u γ ∂u δ

= R1βγδ + R βγδ

∂u ′1 ∂u′ 2 ∂u ′1 ∂u′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2

2

∂u 1 ∂u 2 ∂u γ ∂u δ ∂u 2 ∂u 1 ∂u γ ∂u δ

= R12 γδ + R 21γδ

∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2 ∂u′1 ∂u′ 2

= R1212 + R

∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2 ∂u ′1 ∂u ′ 2

1221

+ R 2112 + R

∂u11 ∂u ′ 2 ∂u′ 2 ∂u ′ 2 ∂u ′1 ∂u′ 2 ∂u ′1 ∂u′ 2

2121

∂u 1 ∂u 2 ∂u1 ∂u 2 ∂u 2 ∂u ′ ∂u 2 ∂u1

2

R − 2 +

= 2112 ∂u′1 ∂u′ 2 ∂u ′1 ∂u ′ 2 ∂u′ 2 ∂u′ 2 ∂u ′1 ∂u ′ 2

∂u1 ∂u1

∂u ′1 ∂u ′ 2 ∂u

= R1212 J 2 where J = =

∂u 2 ∂u 2 ∂u′

so, ∂u ′1 ∂u ′ 2

′ = R1212 J 2

R1212 ...(3)

98 Tensors and Their Applications

Again

∂u γ ∂u δ

b′αβ = bαβ

∂u ′ α ∂u ′β

∂u γ ∂u δ

⇒ b′αβ = bαβ

∂u′ α ∂u′ β

or b′ = bJ 2 ...(4)

from (3) and (4), we get

′

= 1212 = K (say)

R1212 R

...(5)

b′ b

This shows that the quantity K is an invariant for transformation of coordinates. The invariant K

is defined to be the Guasian curvature of S. Hence K is the Riemannian Curvature of S at P0 .

Since Riemannian Coordinates yi with the origin at P0 . We have as geodesic coordinates with the

pole at P0 .

Therefore

k γ

, = 0 at P0

i j g α βb

Then

∂[ij , h]g ∂[ik , h] g

Rhijk = − + at P0 .

∂y k ∂y j

and

∂[βγ, α] b ∂[βδ, α] b

Rαβγδ = − + at P0 .

∂u δ ∂u γ

∂[21,1] b ∂[22,1] b

R1212 = − + at P0 ...(6)

∂u 2

∂u1

from (5) we get

K = b − +

∂u 1

...(7)

∂u 2

CURVATURE TENSOR OF V n

k γ

Let and be the Christoffel symbols of second kind relative to the metrices bαβdu αdu β

i j g α β

Riemann-Christoffel Tensor 99

∂y i ∂y j ∂y k

[αβ, γ]b = [ij, k ]g ...(8)

∂u α ∂u β ∂u γ

∂y i ∂y j ∂y k

[21, 1]b = [ij, k ]g

∂u 2 ∂u1 ∂u1

= q i p j p k [ij , k ] g using (2)

Now,

∂[21, 1] b ∂[ij, k ]g

= qi p j p k

∂u 2

∂u 2

i j k

∂[ij, k ]g ∂y h

= q p p

∂y h ∂u 2

i j k h

∂[ij, k ]g

= q p p q

∂y h

Interchanging h and k, we get

∂[21, 1] b i k j h

∂[ij, h ]g

= qq p p ...(9)

∂u 2 ∂y k

Similarly,

∂[22, 1] b i k j h

∂[ij, h ]g

= qq p p ...(10)

∂u 1

∂y k

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

R1212 = p q p q − ∂y k + ∂y j

h i j k

at P0

h i j k

R1212 = p q p q Rhijk at P0 ...(11)

Since

∂y i ∂y j

bαβ = g ij

∂u α ∂u β

∂y i ∂y j

b11 = g ij = g ij p i p j

∂u ∂u

1 1

b11 = g hj p h p j

100 Tensors and Their Applications

Similarly

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

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

b11 b12

b = b b = b11b22 − b12 b21

21 22

[

b = p h qi p j qk g hig ik − gij ghk ] ...(12)

Dividing (11) and (12), we get

R1212 p h q i q k Rhijk

K= b = ...(13)

p h qi p j q k ( gik g hj − g ij g hk )

This is formula for Riemannian Curvature of Vn at P0 determined by the orientation of Unit vectors p i

and q i at P0 .

If at each point, the Riemannian curvature of a space is independent of the orientation choosen then it

is constant throughout the space.

Proof: If K is the Riemannian curvature of Vn at P for the orientation determined by unit vectors p i

and q i then it is given by

p h q i p j q k R hijk

K= ...(1)

(g ik g hj − g ij g hk ) p h qi p j q k

Let K be independent of the orientation choosen. Then equation (1) becomes

R hijk

K=

g ik g hj − gij g hk

Rhijk = K (g ik g hj − g ij g hk ) ...(2)

We have to prove that K is constant throughout the space Vn .

If N = 2, the orientation is the same at every point. So, consider the case of Vn when n > 2.

Since g ij are constants with respect to covariant differentation, therefore covariant differentation

of (2) gives

Rhijk ,l = ( g ik g hj − g ij g hk ) K ,l ...(3)

where K, l is the partial derivative of K.

Taking the sum of (3) and two similar equations obtained by cyclic permutation of the suffices,

j, k and l.

Riemann-Christoffel Tensor 101

Here n > 2 therefore three or more distinct values to indices j, k, m can be given .

Multiplying (4) by g hj and using g hj g hl = δ lj , we get

or,

(n − 1) g ik K , l + 0 + (1 − n) gil K , k = 0 for δ ij = 0, i ≠ j

g ik K ,l − g il K , k = 0

Multiplying by g ik and using

g ik δik = n, g il g ik = δ kl ,

we get

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

or K, k = 0 as (n − 1)K ,l = 0

∂K

or = 0.

∂x l

integrating it, we get K = constant. This proves that the partial derivatives of K w.r.t. to x’s are all zero.

Consequently K is constant at P. But P is an arbitrary point of Vn . Hence K is constant throughout Vn .

The sum of mean curvatures of a Vn for a mutually orthogonal directions at a point, is independent of

the ennuple choosen. Obtain the value of this sum.

Or

Prove that the mean curvature (or Riccian Curvature) in the direction ei at a point of a Vn is the

sum of n – 1 Riemmanian curvatures along the direction pairs consisting of the direction and n – 1

other directions forming with this directions an orthogonal frame.

i

Proof: Let e hi be the components of unit vector in a given direction at a point P of a Vn . Let e k be the

components of unit vector forming an orthogonal ennuple.

Let the Riemannian curvature at P of Vn for the orientation determined by l hi and e ik (h ≠ k ) be

denoted by K hk and given by

ehp e qk λ rk λ sk R pqrs

K hk =

ehp e qk e rh eks ( g pr q qs − g ps g qr )

= (e p e r g ) (e q e s g ) − (e p e s g ) (e q e r g ) ...(1)

h h pr k k qs h k ps k h qr

102 Tensors and Their Applications

e hp e hr g pr = 1

p s

and e h eh g ps = 0 etc.

Using these in equation (1), then equation (1) becomes

e hp ekq e rh e ks R pqrs

K hk =

1×1 − 0 × 0

K hk = e hp e kq ehr e ks R pqrs ...(2)

n n

∑K

k =1

hk = ∑e p q r s

e e e

h k h k

R pqrs

k =1

n

Put ∑K

k =1

hk = M h . Then

n

p r

M h = eh eh ∑e

k =1

q s

e

k k

R pqrs

p r qs

= e h e h g R pqrs

= − e h e h g R qprs

p r qs

M h = − eh e h R pr

p r

...(3)

This shows that M h is independent of (n – 1) orthogonal direction choosen to complete an orthogonal

ennuple. Here M h is defined as mean curvature or Riccian curvature of Vn for the direction e hp1 .

Summing the equation (1) from h = 1 to h = n, we get

n

∑M

h =1

h = − eh |e h| R pr

p r

= − g pr R pr

= –R

n

or ∑M

h =1

h = − g pr R pr == −R

This proves that the sum of mean curvatures for n mutual orthogonal directions is independent of the

directions chosen to complete an orthogonal ennuple and has the value − R.

Let e ih1 is not a unit vector and the mean curvature M h is given by

Rij e hi e hj

Mh =

g ij e hi ehj

Riemann-Christoffel Tensor 103

⇒ ( Rij + M h g ij ) e ih ehj = 0

∂M h

g jk ehj e hk + 2 ( Rij + Mg ij ) e hj

∂ ehi = 0 ...(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 .

A space, which is homogeneous relative to the Ricci tensor Rij is called Einstein space.

If space is homogeneous then we have

Rij = λg ij ...(1)

1

⇒ λ = nR

from (1)

R

Rij = g

n ij

R

Hence a space is an Einstein space if Rij = g at every point of the space.

n ij

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

104 Tensors and Their Applications

Rhijk

K = (g g − g g )

ik hj ij hk

Rhijk = K (g ik g hj − g ij g hk )

Multiplying by g hk

Kg hk ( g ik g hj − g ij g hk ) = g hk Rhijk

K (δhi g hj − ng ij ) = Rij

K (g ij − ng ij ) = Rij

K (1 − n) g ij = Rij ...(1)

ij

Multiplying by g , we get

Kn(1 − n) = R as g ij Rij = R ...(2)

from (1) & (2)

R 1

Rij = (1 − n) gij ⋅ = Rgij

n(1 − n ) n

R

⇒ g

Rij =

n ij

This is necessary and sufficient condition for the space Vn to be Einstein space.

Weyl Tensor denoted as Whijk and defined by

1

Whijk = Rhijk + ( g R − g kh g ij )

1 − n ki hj

Theorem 5.6 A necessary and sufficient condition for a Riemannian V n (n > 3) to be of constant

curvature to that the Weyl tensor vanishes identically throughout Vn .

Let K be Riemannian Curvature of Vn . Let K = constant.

We have to prove that Whijk = 0

Since we know that

p h q i p j q k R hijk

K= = constant

(g hj g ik − g ij g hk ) p h qi p j q k

Riemann-Christoffel Tensor 105

Then

R hijk

K= ...(1)

g hj gik − gij g hk

Multiplying by g hk , we get

g hk Rhijk = g hk K (g hj g ik − g ij g hk )

Rij = K (1 − n) g ij ...(2)

g ij Rij = K (1 − n) g ij g ij

R = K (1 − n )n ...(3)

Putting the value of K from (3) in (2), we get

R

Rij = g ...(4)

n ij

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

Now, the W tensor is given by

1

Whijk = Rhijk + [ g R − g hk g ij ]

1 − n ik hj

from (5), we get

1 R R

Whijk = Rhijk + g ik g hj − g hk g ij

1−n n n

R

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

R Rhijk

= Rhijk + ⋅ , by. eqn. (1)

n(1 − n) K

Rhijk

Whijk = Rhijk + K , by equation (3)

K

Whijk = 2 R hijk

This proves necessary condition.

106 Tensors and Their Applications

Sufficient Condition

Let Whijk = 0. Then we have to prove that K is constant.

Now, Whijk = 0.

1

⇒ Rhijk + [ g R − g hk Rij ] = 0

1 − n ik hj

Multiplying by g hk , we get

1

Rij + [g hk g ik Rhj − g hk g hk Rij ] =0

1− n

1

Rij + [δh R − nRij ] =0

1 − n i hj

R

Rij + ij (1 − n) =0

1− n

⇒ 2 R ij =0

⇒ Rij = 0

⇒ g = 0 or Rhijk = 0

hk

If g hk = 0 then

R hijk p h qi p j q k

K=

(g hj g ik − 0 g ij ) p h qi p j q k

R hijk p h q i p j q k R hijk p h q i p j q k

K= =

( p h p j g hj ) (q i q k g ik ) p2 ⋅ q 2

h i j k

K = Rhijk p q p q since p 2 = 1, q 2 = 1

K = constant as Rij = 0

This proves sufficient condition.

EXAMPLE 1

For a V2 referred to an orthogonal system of parametric curves (g 12 = 0) show that

R12 = 0, R11 g 22 = g 22 g11 = R1221

2R1221

R = g Rij =

ij

g11 g 22

Consequently

1

Rij = Rg .

2 ij

Riemann-Christoffel Tensor 107

Solution

Also,

1 1 1

& g =g .

22

g ij = ⇒ g 11 =

g ij g11 22

i j

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

1 2 2 2

g11 g12 g 0

and g = g ij = = 11 = g11 g 22

g 21 g 22 0 g 22

(i) To prove R12 = 0

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

hk h1

= g 21 R2121 as R1121 = 0

R12 =0

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

R11 = g hk Rh 11k = g 2 k R 211k

R 2112

R11 = g R 2112 = g

22

22

So,

R 2112

R11 = ...(1)

g 22

and R22 = g hk Rh 22k = g 1k R122k

R1221

= g R1221 =

11

g11

R1221

So, R22 = ...(2)

g11

from (1) and (2)

R 11g22 = R 1221=R 22g11 … (3)

(iii) To prove

2R1221

R=

g11 g 22

R = g Rij = g Ri 1 + g Ri 2

ij i1 i2

108 Tensors and Their Applications

R 11 R 22

= +

g 11 g 22

R1221

= g g +g g

R1221

[Q by eqn. (3)]

22 11 11 22

2R1221

R=

g11 g 22

1

(iv) To prove Rij = Rg

2 ij

2R1221

R=

g11 g 22

2R1221

R= as g = g11 g 22

g

1

R1221 = Rg

2

The eqn (3) expressed as

1

R11 g 22 = Rg = R 22 g11 .

2

it becomes

Rg Rg g Rg

R11 = = 11 22 = 11

2 g 22 2g 22 2

Rg Rg11 g 22 Rg22

R22 = = =

2 g11 2g11 2

So,

1 1

R11 = Rg11 & R22 = g 22

2 2

1

R12 = g12 as R12 = 0 = g12

2

1

This prove that Rij = R .

2 ij

EXAMPLE 2

The metric of the V 2 formed by the surface of a sphere of radius r is ds 2 = r 2 dθ 2 + r 2 sin 2 θdφ 2

1

in spherical polar coordinates. Show that the surface of a sphere is a surface of constant curvature 2 .

r

Solution

Given that

ds 2 = r 2 d θ 2 + r 2 sin 2 θd φ 2

Since r is radius of curvature then r is constant.

Riemann-Christoffel Tensor 109

We can prove

R 1221 = r2 sin 2 θ

Now, the Riemannian curvature K of V n is given by

p hq i p jq k

K= h i j k .

p q p q ( g hj gik − g hk gij )

Consider two vectors whose components are (1, 0) and (0, 1) respectively in V2 . Then

R 1212 R

K= = 1212

g11 g 22 g

r 2 sin 2 θ 1

K= = = constant.

r 4 sin 2 θ r 2

EXERCISES

1. Show that

∂[ jl, i] ∂[ jk , i] α α

Rijkl = − + [il , α ] − [ik , α]

∂xk ∂xl j k j l

2. Show that

1 ∂ g il

2 2

2

∂ g jk ∂ 2 g ik ∂ g jl

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

Rijkl = + − − αβ

2 ∂x j ∂xk ∂xi ∂xl ∂ x j ∂xl ∂xi ∂ xk

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

Rijkl = − R jikl = − Rijlk = Rklij and Rijkl + Riklj + R iljk = 0

4. Show that the curvature tensor of a four dimensional Riemannian space has at the most 20 distinct

non-vanishing components.

h

5. (a) If prove that the process of contraction applied to the tensor Rijk generates only one new tensor

Rij which is symmetric in i and j.

1

(b) If ds2 = g11 ( dx1 ) 2 + g 22 (dx 2 )2 + g 33 ( dx3 )2 . Prove that Rij =

Rihhj , (h, i, j being unequal).

g hh

6. Show that when in aV3 the coordinates can be chosen so that the components of a tensor g ij are zero

when i, j, k are unequal then

1

(i) Rhj = Rhiij

g ii

110 Tensors and Their Applications

1 1

(ii) Rhh = Rhiih + Rhjjh

g ii g jj

7. Prove that if

α 1 ∂R

Ri = g Rij then Ri , α = 2 i

α αj

∂x

and hence deduce that when n > 2 then scalar curvature of an Einstein space is constant.

8. If the Riemannian curvature K of Vn at every point of a neighbourhood U of Vn is independent of the

direction chosen, show that K is constant throughout the neighbourhood U. Provided n > 2.

9. Show that a space of constant curvature K0 is an Einstein space and that R = K 0 n (1 –n ).

10. Show that the necessary and sufficient condition that Vn be locally flat in the neighbourhood of 0 is

that Riemannian Christoffel tensor is zero.

11. Show that every V2 is an Einstein space.

12. For two dimensional manifold prove that

R

K= −

2

13. Show that if Riemann-Christoffel curvature tensor vanishes then order of covariant differentiation is

commutative.

CHAPTER – 6

KRÖNECKER DELTAS

The concept of symmetry and skew-symmetry with respect to pairs of indices can be extended to

cover to pairs of indices can be extended to cover the sets of quantities that are symmetric or skew-

symmetric with respect to more than two indices. Now, consider the sets of quantities A i ... i k or

Ai l ... i k depending on k indices written as subscripts or superscripts, although the quantities A may not

represent tensor.

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.

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.

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

ei1 ... in (or ei1 ... in ) = −1; when i1 , i2 ,..., in

an odd permutation of number1, 2, ..., n

= 0 in all other cases

EXAMPLE 1

Find the components of system eij when i, j takes the value 1,2.

Solution

The components of system eij are

e11 , e12 , e21 , e 22 .

By definition of e-system, we have

e11 = 0, indices are same

e12 = 1, since i j has even permutation of 12

e 21 = − e12 = −1 since i j has odd permutation of 12

e22 = 0, indices are same

EXAMPLE 2

Solution

By the definition of e-system,

e123 = e 231 = e321 = 1

e213 = e132 = e321 = −1

eijk = 0 if any two indices are same.

A symbol δi1j1...... ikjk depending on k superscripts and k subscripts each of which take values from 1 to n,

is called a generalised Krönecker delta provided that

(a) it is completely skew-symmetric in superscripts and subscripts

The e-Systems and the Generalized Krönecker Deltas 113

(b) if the superscripts are distinct from each other and the subscripts are the same set of

numbers as the superscripts.

The value of symbol

= 1; an even number of transposition is required to arrange the

superscripts in the same order as subscripts.

δ j ... j = −1; where odd number of transpositions arrange the superscripts

i1 ... i k

1 k

in the same order as subscripts

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

EXAMPLE 3

Find the values of δ ijkl .

Solution

By definition of generalised Kronecker Delta, δ ijkl = 0 if i = j or k = l or if the set. ij is not the set kl.

δ11

pq = δ pq = δ13 = ⋅ ⋅ ⋅ = 0

22 23

i.e.,

δ ijkl = 1 if kl is an even permutation of ij

i.e., δ12

12

= δ 21

21

= δ13

13

= δ31

31

= δ 23

23

= ⋅⋅⋅ =1

i.e., δ12

21

= δ13

31

= δ13

31 = δ12 = ⋅ ⋅ ⋅ = − 1

21

i i ... i

Theorem 6.1 To prove that the direct product e 1 2 n e j1 j2 ... jn of two systems e i1 ... in and e j1 j 2 ... jn is the

generalized Krönecker delta.

i i ... i n

Proof: By definition of generalized Krönecker delta, the product e 1 2 ej

1 j2 ... jn

has the following values.

(i) Zero if two or more subscripts or superscripts are same.

(ii) +1, if the difference in the number of transpositions of i1 , i2 ,...,in and j1 , j2 ,..., jn from

1,2,...n is an even number.

(iii) –1, if the difference in the number of transpositions of i1, i2,...,in and j1, j2, ...jn from 1, 2,..n

an odd number.

Thus we can write

= δ 1j1 2j 2 ... nj n

i i ... i

e i 1i 2 ... i n e j1 j 2 ... jn

i i ... i

i i ... i

114 Tensors and Their Applications

(ii) -1; if i1 , i2 ,...,in is an odd permutation of numbers 1,2,...n

(iii) 0; in all other cases

Hence by Definition of generalized krönecker delta, we can write

(1) e 1 2 n = δ11 22... nn

i i ... i i i ... i

and

(2) ei1i2 ... in = δi1i 2 ... in

1 2 ... n

jk

δ ijk

αβγ = δ αβ1 + δ αβ2 + δαβ 3 = δαβ

ij1 ij 2 ij 3 ij

α β 3.

Hence

− 1; if αβ is an odd permutatio n of 12

δ12

αβ =

0; if αβ is not permutatio n of 12

Similarly results hold for all values of α and β selected from the set of numbers 1, 2, 3.

Hence

+ 1; if i j is an even permutatio n of αβ

− 1; if i j is an odd permutatio n of αβ

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

ij

subscripts and superscripts are not formed from the same numbers.

1

If we contract δ ijαβ . To contract δ ijαβ first contract it and the multiply the result by . We obtain

2

a system depending on two indices

1 ij 1 i 1

δ iα = 2 δαj = 2 (δ α1 + δ α2 + δα 3 )

i2 i3

The e-Systems and the Generalized Krönecker Deltas 115

1 12

It i = 1 in δiα then we get δ α = δα 2 + δ α3

1

2

13

( )

This vanishes unless α = 1 and if α = 1 then δ11 = 1.

(i) 0 if i ≠ α, (α, i = 1, 2, 3, )

(ii) 1 if i = α .

By counting the number of terms appearing in the sums. In general we have

1 ij

δ iα = n −1 δ αj and δ ij = n(n − 1)

ij

...(1)

We can also deduce that

(n − k )! i1i2 ... ir i r −1... i k

δi1j1i2j...2 ...i r jr = (n − r )! δ j1 j2 ... j r j r −1... j k ...(2)

and

n!

δi1j1i2j...2 ...i r jr = n (n −1) (n − 2) ⋅ ⋅ ⋅ (n − r + 1) = ...(3)

n − r!

or

and from (2) we deduce the relation

EXERCISE

1. Expand for n = 3

i α 12 i j αβ i j ij

(a) δ αδ j (b) δ ij x x (c) δ ij x y (d) δij

2. Expand for n = 2

(c) e aα aβ = e a .

ij 1 2 ij 2 1 ij i j ij

(a) e ai a j (b) e ai a j

ijk

i ... i

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

k

5. Prove that εijk g is a covariant tensor of rank three where where ε ijk is the usual permutation

symbol.

CHAPTER – 7

GEOMETRY

Consider the n-dimensional space R be covered by a coordinate system X and a curve C so that

C : x i = x (t ), (i = 1,2..., n)

i

...(1)

which is one-dimensional subspace of R. Where t is a real parameter varying continuously in the

interval t1 ≤ t ≤ t 2 . The one dimensional manifold C is called arc of a curve.

1 2 n dx dx

1 2

dx n

Let F x , x ,..., x , dt dt ,..., dt be a continuous function in the interval t1 ≤ t ≤ t 2 . Wee

dx i

assume that F x, > 0, unless every dx = 0 and that for every positive number k

dt dt

dx 1

dx 2

dx n 1 2 1 2 n

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

dt dt

=

dt dt dt dt

The integral

dx

∫

t2

s= F x, dt ...(2)

t1 dt

is called the length of C and the space R is said to be metrized by equation (2).

dx

dt

If one chooses to define the length of arc by the formula

dx p dx q

∫

t2

s= g pq ( x)dt, (p,q = 1, 2, ..., n) ...(3)

t1 dt dt

dx p dxq dx p

where g pq ( x) is a positive definite quadratic form in the variable , then the resulting

dt dt dt

geometry is the Riemannian geometry and space R metrized in this way is the Riemannian n-dimensional

space R n.

Geometry 117

Consider the coordinate transformation T : x i = x i ( x1 ,..., x n ) such that the square of the element

of arc ds,

p q

ds 2 = g pq dx dx ...(4)

can be reduced to the form

ds 2 = dx i dx i ...(5)

Then the Riemannian manifold R n is said to reduce to an n-dimensional Euclidean manifold E n.

The Y-coordinate system in which the element of arc of C in E n is given by the equation (5) is

called an orthogonal cartesian coordinate system. Obviously, E n is a generalization of the Euclidean

plane determined by the totality of pairs of real values ( x 1 , x 2 ). If these values ( x 1 , x 2 ) are associated

with the points of the plane referred to a pair of orthogonal Cartesian axes then the square of the

element of arc ds assumes the familiar form

ds 2 = (d x1)2 + (d x 2 )2 .

dx dx dx

THEOREM 7.1 A function F x, satisfying the condition F x, k = kF x , for every

dt dt dt

k > 0. This condition is both necessary and sufficient to ensure independence of the value of the

dx

∫

t2

integral s = t F x, dt of a particular mode of parametrization of C. Thus if t in C : x i = xi (t)

1 dt

is replaced by some function t = φ(s ) and we denote x i [φ(s )] by ξ i (λ) . so that x i (t ) ξ i (s ) we have

equality

dx

∫

t2

∫

s2

F x , dt = F (ξ, ξ′) ds

t1 dt s1

ds i

where ξ′ =

i

and t1 = φ (s1 ) and t2 = φ (s 2 ).

ds

Proof: Suppose that k is an arbitary positive number and put t = ks so that t1=ks 1 and t2=ks 2. Then

i

C : x i = x (t ) becomes

C : x i (ks) = ξ i (t )

dx i (ks ) dx i (ks)

and ξ′ i (s ) = =k

ds dt

dx

∫

t2

Substituting these values in s= F x , dt we get

t1 dt

dx (ks )

∫ F x (ks),

s2

kds

dt

s= s1

∫

s2

or s= s1

dx dx

We must have the relation F (ξ, ξ′ ) = F x, k = kF x, . Conversely, if this relation is true

dt dt

118 Tensors and Their Applications

for every line element of C and each k > 0 then the equality of integrals is assumed for every choice of

parameter t = φ ( s), φ1 (s ) > 0, s1 ≤ s ≤ s 2 with and t2 = φ (s 2 ).

i

dx

Note: (i) Here take those curves for which x i (t ) and are continuous functions in t1 ≤ t ≤ t 2.

dt

dx dx dx

(ii) A function F x, satisfying the condition F x, k = kF x, for every k > 0 is called

dt dt dt

i

dx

positively homogeneous of degree 1 in the .

dt

EXAMPLE 1

What is meant, consider a sphere S of radius a, immersed in a three-dimensional Euclidean

manifold E 3 , with centre at the origin (0, 0, 0) of the set of orthogonal cartesian axes O − X 1 X 2 X 3 .

Solution

Let T be a plane tangent to S at (0, 0,–a) and the points of this plane be referred to a set of orthogonal

cartesian axes O′ –Y 1Y 2 as shown in figure. If we draw from O (0, 0, 0) a radial line OP , interesting

the sphere S at P ( x1 , x 2 , x 3 ) and plane T at Q( x 1 , x 2 , −a) then the points P on the lower half of the

sphere S are in one-to-one correspondence with points ( x 1 , x 2 ) of the tangent plane T..

3

X

2

X

O

P2

P1

C 2

O´ Y

1

X

Q2

1 K

Y Q1

Fig. 7.1

If P ( x1 , x 2 , x 3 ) is any point on the radial line OP,, then symmetric equations of this line is

x1 − 0 x2 − 0 x3 − 0

= 2 = =λ

x1 − 0 x − 0 −a− 0

or

x1 = λx 1 , x 2 = λx 2 , x 3 = −λa ...(6)

Since the images Q of points P lying on S, the variables x i satisfy the equation of S,

( x1 )2 + (x 2 )2 + (x 3 ) 2 = a 2

or ( ) ( )

λ2 x 1 + x 2

2 2

+ a2 = a2

Geometry 119

ax1 ax 2

x =

1 , x2 =

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

− a2

and x =

3 ...(7)

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

These are the equations giving the analytical one-to-one correspondence of the points Q on T and

points P on the portion of S under consideration.

Let P1 ( x1 , x 2 , x 3 ) and P2 ( x1 + dx1 , x 2 + dx 2 , x 3 + dx 3 ) be two close points on some curve C lying

on S. The Euclidean distance P1 P2 along C, is given by the formula

ds 2 = dx i dx i , (i = 1, 2, 3) ...(8)

Since

∂x i p

dx = ∂x p d x , (P =1, 2)

i

∂ xi ∂ xi

ds 2 = d x pd x q

∂x ∂ x

p q

= g pq ( x )d x pd x q , (p,q = 1, 2)

∂ xi ∂ x i .

where g pq (x ) are functions of x i and g pq =

∂ x p ∂x q

If the image K of C on T is given by the equations

x 1 = x 1 ( t )

K:

x 2 = x 2 (t ), t1 ≤ t ≤ t2

dx p dxq

∫

t2

s= g pq dt

t1 dt dt

1 1 2

(d x 1)2 + (d x 2) 2 + (x d x − x 2 d x 1)2

a2

{ }

2

ds2 = 1 2 2 ...(9)

1 +

a 2 ( x 1 2

) + ( x )

and

120 Tensors and Their Applications

2 2 2

d x1 d x 2 2 1

+ + 12 x1 d x − x 2 d x

t 2 dt dt

dt a dt

s= t ∫

1 1

{

1 + 2 (x ) + ( x )

1 2 2 2

}

dt

a

So, the resulting formulas refer to a two-dimensional manifold determined by the variables ( x 1 , x 2 ) in

the cartesian plane T and that the geometry of the surface of the sphere imbeded in a three-dimensional

Euclidean manifold can be visualized on a two-dimensional manifold R 2 with metric given by equation (9).

1

If the radius of S is very large then in equation (9) the terms involving 2 can be neglected. Then

a

equation (9) becomes

ds 2 = (d x ) + (d x ) .

1 2 2 2

...(10)

Thus for large values of a, metric properties of the sphere S are indistinguishable from those of

the Euclidean plane.

The chief point of this example is to indicate that the geometry of sphere imbedded in a Euclidean

3-space, with the element of arc in the form equation (8), is indistinguishable from the Riemannian

geometry of a two-dimensional manifold R 2 with metric (9). the latter manifold, although referred to

a cartesian coordinate system Y, is not Euclidean since equation (9) cannot be reduced by an admissible

transformation to equation (10).

Let P (x ) be the point, in an Euclidean 3-space E 3 , referred to a set of orthogonal Cartesian coordinates Y..

Consider a coordinate transformation

T : x i = x i (x 1 , x 2 , x 3 ), (i = 1, 2, 3)

∂ xi

Such that J = ∂ x j ≠ 0 in some region R of E 3 . The inverse coordinate transformation

T −1 : x i = x ( x , x , x ), (i = 1, 2, 3)

i 1 2 3

will be single values and the transformations T and T −1 establish one-to-one correspondence

between the sets of values ( x1 , x 2 , x 3 ) and ( x 1 , x 2 , x 3 ).

The triplets of numbers ( x1 , x 2 , x 3 ) is called curvilinear coordinates of the points P in R.

1 2 3

If one of the coordinates x , x , x is held fixed and the other two allowed to vary then the point

P traces out a surface, called coordinate surface.

If we set x1 = constant in T then

x1 ( x 1 , x 2 , x 3 ) = constant ...(1)

defines a surface. If constant is allowed to assume different values, we get a one-parameter family of

surfaces. Similarly, x 2 (x 1 , x 2 , x 3 ) = constant and x 3 ( x 1 , x 2 , x 3 ) = constant define two families of

surfaces.

The surfaces

Geometry 121

x1 = c1 , x = c2 , x = c3

2 3

… (2)

3 3

Y X

2

X

P 1

X

2

Y

1

Y

Fig. 7.2

intersect in one and only one point. The surfaces defined by equation (2) the coordinate surfaces

and intersection of coordinate surface pair-by pair are the coordinate lines. Thus the line of intersection

of x1 = c1 and x 2 = c2 is the x 3 − coordinate line because along this the line the variable x 3 is the only

one that is changing.

EXAMPLE 2

Consider a coordinate system defined by the transformation

3

Y

1

X

X2

2

O Y

3

X

Y1

Fig. 7.3

x 3 = x1 cos x 2 ...(5)

122 Tensors and Their Applications

The surfaces x1 = constant are spheres, x 2 = constant are circluar cones and x3 = constant are

planes passing through the Y 3-axis (Fig. 7.3).

The squaring and adding equations (3), (4) and (5) we get,

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

On solving

( x 1 )2 + (x 2 )2 + (x 3 ) 2 = (x1 )2

x1 = (x 1 ) 2 + ( x 2 ) 2 + ( x 3 )2 ...(6)

Now, squaring and adding equations (3) and (4), we get

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

( x 1 )2 + (x 2 )2 = ( x1 )2 (sin x 2 ) 2

x1 sin x 2 = (x 1 ) 2 + ( x 2 ) 2 ...(7)

Divide (7) and (5), we get

(x1 )2 + (x 2 )2

tan x 2 =

x3

2 2

−1 ( x ) + ( x )

1 2

2 tan

or x = x3 ...(8)

Divide (3) and (4), we get

x2

1 =

tan x 3

x

−1 x

2

⇒ x 3 = tan 1

...(9)

x

So, the inverse transformation is given by the equations (6), (8) and (9).

If x1 > 0, 0 < x 2 < π, 0 ≤ x 3 < 2π. This is the familiar spherical coordinate system.

Covariant and Contravariant Vectors

r r r

Let a cartesian coordinate system be determined by a set of orthogonal base vectors b1 , b2 ,b3 then the

position vector rr of any point P (x 1 , x 2 , x 3 ) can be expressed as

r r

r = bi x i (i = 1, 2, 3) ...(1)

Geometry 123

r

Since the base vectors bi are independent of the position of the point P (x 1 , x 2 , x 3 ) . Then from (1),

r r

dr = bi dx i ...(2)

element of arc ds between two points is

r r

ds 2 = dr ⋅ dr

from equation (2),

r i r

ds 2 = bi dx ⋅ b j dx

j

r r

= bi ⋅ b j d x id x j

r r

ds 2 = δij d x d x ; since bi ⋅ b j = δ ij

i j

3 3

Y X

2

a3 a2 X

a1 1

X

P

r

b3

O b2 2

Y

b1

1

Y

Fig. 7.4.

r r r r r r r

(b1 , b2 , b3 areorthogonal base vector i.e.,b1 ⋅ b1 = 1 & b1 ⋅ b2 = 0) .

as äij = 1, i = j

ds 2 = d x id x i;

= 0, i ≠ j

a familiar expression for the square of element of arc in orthogonal cartesian coordinates.

Consider the coordinate transformation

i 1 2 3

x i = x (x , x , x ), (i = 1, 2, 3)

r

define a curvilinear coordinate system X. The position vector r is a function of coordinates x i .

i.e.,

r r

r = r (x i ), (i = 1, 2, 3)

Then

r

r ∂r i

dr = dx ...(3)

∂x i

124 Tensors and Their Applications

and

ds 2 = d rr ⋅ d rr

r r

∂r ∂r

= ⋅ d x id x j

∂ xi ∂ x j

ds 2 = g ij dx i dx j

where

r r

∂r ∂r

g ij = ⋅ ...(4)

∂ xi ∂ x j

r

∂r

The vector is a base vector directed tangentially to X i - coordinate curve.

∂x i

Put

r

∂r r

= ai ...(5)

∂x i

r r r r

dr = ai d x i and g ij = a i ⋅ a j ...(6)

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

r r

a j dx j = bi d x i

r r ∂x i j

a j dx j = bi dx

∂x j

r r ∂ xi

⇒ a j = bi , as d x j are arbitrary

∂x j

r

So, the base vectors a j transform according to the law for transformation of components of

covariant vectors.

r

The components of base vectors a i , when referred to X-coordinate system, are

r r r

a1 : (a1 , 0,0), a 2 : (0, a2 , 0), a3 : (0,0, a3 ). ...(7)

and they are not necessarily unit vectors.

In general,

r r r r

g11 = ar1 ⋅ ar1 ≠ 1, g 22 = a 2 ⋅ a2 ≠ 1, g 33 = a3 ⋅ a3 ≠ 1. ...(8)

If the curvilinear coordinate system X is orthogonal. Then

r r r r

g ij = a i ⋅ a j = ai a j cos θij = 0, if i ≠ j. ...(9)

r r

Any vector A are can be written in the form A = k d rr where k is a scalar..

Geometry 125

r

r ∂r i

Since d r = d x we have

∂ xi

r

r ∂r i

A = ∂x i (kdx )

r r i

A = ai A

r

where Ai = kdxi . The numbers Ai are the contravariant components of the vector A

Consider three non-coplanar vectors

r r r r r r

r1 a 2 × a3 ar 2 = a3 × a1 , ar 3 = a1 × a2

a = rr r ,

[a1a 2a3 ] [ar1ar2 ar3 ] [ar1ar2ar3 ] ...(10)

r r r r rr r

where a 2 × a3 , etc. denote the vector product of a 2 and a3 and [a1 a2 a3 ] is the triple scalar

r r r

prduct a1 ⋅ a2 × a3 .

Now,

r r r rr r

a2 × a3 ⋅ a1 [a1 a2 a3 ]

r1 r = = 1.

a ⋅ a1 = [ar ar ar ] [ar1ar2ar3 ]

1 2 3

r r r r r r

r1 r a2 × a3 ⋅ a2 [a2 a2 a3 ]

a ⋅ a 2 = [ar ar ar ] = = 0.

1 2 3 [ar1ar2 ar3 ]

r r r

Since [a2 a2 a3 ] = 0.

Similarly,

r r r r

a 1 ⋅ a3 = a 2 ⋅ a1 = ⋅ ⋅ ⋅ = 0

r r r r

a 2 ⋅ a2 = 1, a 3 ⋅ a 3 = 1

Then we can write

r r

a i ⋅ a j = δij

EXAMPLE 3

rr r

To show that [a1 a2 a3 ] = [ ]

g and ar1ar 2 ar 3 = 1 where g = g ij .

g

Solution

The components of base vectors ai are

r r r

a1 : (a1 , 0,0), a 2 : (0, a2 , 0) and a3 : (0, 0, a3 )

Then

a1 0 0

rr r

[a1a2a3 ] = 0 a2 0 = a1a2a3 ...(11)

0 0 a3

126 Tensors and Their Applications

and

g= g ij = g 21 g 22 g 23

g 31 g 32 g33

g11 = ar1 ⋅ ar1 ⇒ a12 = g11

Similarly

a22 = g 22 , a 32 = g 33

and

g12 = ar1 ⋅ ar2 = 0, g13 = ar1 ⋅ ar3 = 0 etc.

So,

a12 0 0

g= 0 a22 0 = a12 a22 a32

0 0 a32

g = a1 a2 a3 ...(12)

from eqn. (11) and (12), we have

[ar1ar2ar3 ] = g

r r r 1

Since the triple products [a 1 a 2 a 3 ] = . Moreover,,

g

r r r r r r

r a2 × a3 r a 3 × a1 r a1 × a 2

a1 = r1 r 2 r 3 , a 2 = r r r , a 3 = r r r

[a a a ] [a1 a 2 a 3 ] [a1 a 2 a 3 ]

The system of vectors ar1 , ar 2 , ar3 is called the reciprocal base system.

r r r

Hence if the vectors a 1 , a 2 , a 3 are unit vectors associated with an orthogonal cartesian coordinates

then the reciprocal system of vector defines the same system of coordinates. Solved.

r r ri

The differential of a vector r in the reciprocal base system is d r = a d xi .

where dxi are the components of drr. Then

r r

ds 2 = dr ⋅ dr

r r

= (a i dxi ) ⋅ (a j dx j )

r r

= a i ⋅ a j dxi dx j

ij

ds 2 = g dxi dx j

r r

where g ij = a i ⋅ a j = g ji ...(13)

Geometry 127

The system of base vectors determined by equation (10) can be used to represent an arbitrary

r r r

vector A in the form A = a i Ai , where Ai are the covariant components of A.

r r

Taking scalar product of vector Ai a i with the base vector a j , we get

r r r r

Ai a i ⋅ a j = Ai δij = A j as a i ⋅ a j = δ ij .

r

r ∂A r

THEOREM 7.2 If A is a vector along the curve in E 3 . Prove that = A,αj aα

∂x j

∂A i r

Also, prove that A,i j = j . Where Ai are component of A.

∂x

r r

Proof: A vector A can be expressed in the terms of base vectors ai as

r r

A = Ai ai

r

r ∂r r

where a i = i and Ai are components of A.

∂x

r

The partial derivative of A with respect to x j is

r r

∂A ∂Ai r i ∂ai

= a + A ...(1)

∂x j ∂x j

i

∂x j

r r

Since g ij = a i ⋅ a j .

Differentiating partially it w.r.t. x k , we have

r r

∂g ij ∂a i r ∂a j r

= ⋅a j + k ⋅ ai

∂x k ∂x k ∂x

Similarly,

r r

∂g jk ∂a j r ∂ak r

= ⋅ a + ⋅ aj

∂x i k

∂xri ∂xri

∂g ik ∂ai r ∂a r

and j =

⋅ ak + kj ⋅ ai

∂x ∂x j

∂x

r

Since A can be written as

r r i ri

A = a i A = a Ai

r

Taking scalar product with a j , we have

r r r r

a i ⋅ a j A i = a i ⋅ a j Ai

⇒ g ij A i = δ ij Ai = A j

r r r r

As ai ⋅ a j = g ij , a i ⋅ a j = δ ij and δ ij Ai = A j .

We see that the vector obtained by lowering the index in Ai is precisely the covariant vector Ai .

128 Tensors and Their Applications

r

The two sets of quantities Ai and Ai are represent the same vector A referred to two different

base systems.

EXAMPLE 4

Show that g iα g jα = δ ij .

Solution

Since we know that

g i α = ari ⋅ arα and g jα = ar j ⋅ ar α

Then

r r r r

g i α g jα = (ai ⋅ aα ) (a j ⋅ a α )

r r r r

= (ai ⋅ a j ) (aα ⋅ a α )

r r

= δ ij δ αα as ai ⋅ a j = δ ij

g i α g jα = δij as δ αα = 1.

But

r

r ∂r

ai =

r ∂x i r

r r

∂ai ∂ 2r ∂ 2r ∂a j

= = =

∂x j ∂x j dx i ∂x i ∂x j ∂xi

So,

r r

∂ai ∂a j

=

∂x j ∂x i

Now,

1 ∂g ik ∂g jk ∂g ij

+ − k , Christoffel’s symbol

2 ∂x j

[ij, k ] =

∂x i ∂x

∂g ik ∂g jk ∂g

Substituting the value of , i and ijk , we get

∂x j

∂x ∂x

r

1 ∂a i r

[ij,k] = ⋅ 2 j ⋅ a k

2 ∂x

r

∂ai r

[ij,k] = ⋅ ak

∂x j

r

∂ai rk 1 rk

or j = [ ij, k ]a , as r = a

∂x ak

Hence

r

∂a i r α r r

⋅ a = [ij, k ]a k ⋅ a α

∂x j

Geometry 129

r r

= [ij, k ]g kα , Since g kα = a k ⋅ a α

r α

∂ai r α α [ i j , k ] g kα =

. a = , as

∂x j i j i j

r α r

∂ai

= aα ...(2)

∂x j i j

r

∂ai

Substituting the values of in equation (1), we get

∂x j

r

∂A ∂Ai r α i r

= ai + A a α

∂x j ∂x j i j

∂A r α

α r

= a + A i aα

j α

∂x i j

r

∂A ∂A α α i r

= j + A aα

∂x j ∂x i j

r

∂A α r α ∂A α α i

= A , a α since A = + A

∂x j j ,j

∂x j i j

Thus, the covariant derivative A,αj of the vector A α is a vector whose components are the

r

∂A r

components of j referred to the base system ai .

∂x

r

α ∂ai

If the Christoffel symbols vanish identically i.e., = 0 the = 0, from (2).

i j ∂x j

Substituting this value in equation (1), we get

r

∂A ∂Ai r

= ai

∂x j ∂x j

∂Ai i β

But A,i j = + A

∂x j β j

∂Ai i

A,i j = as = 0.

∂x j β j

Proved.

r r

THEOREM 7.3 If A is a vector along the curve in E 3 . Prove that A j , k a j wheree A j are components

r

of A. .

r

Proof: If A can be expressed in the form

r r

A = Ai a i

130 Tensors and Their Applications

r

where Ai are components of A.

r

The partial derivative of A with respect to x k is

r r

∂A ∂Ai ri ∂a i

= a + Ai k ...(1)

∂x k ∂x k ∂x

r r

Since a i ⋅ a j = δij , we have,

r r

∂a i r r i ∂a j

⋅ a + a ⋅ =0

∂x k

j

∂x k

r r

∂a i r r i ∂a j

⋅a j = − a ⋅ k

∂x k ∂x

r

r r α ∂a j α r

= − ai ⋅ aα , Since = a α

j k ∂x k

j k

r r

But a i ⋅ aα = δiα . Then

r

∂a i r α

⋅ a j = − δiα

∂x k j k

r

∂a i r i

⋅a j = −

∂x k j k

r 1 rj

∂a i i r

= − a j , as ar = a

∂x k j k j

r

∂a i

substituting the value of in equation (1), we get

∂x k

r

∂A ∂Ai ri i r j

a − Ai a

k = ∂x k

∂x j k

∂Ai r j i rj

= a − Ai a

∂x k

j k

r

∂A ∂Ai i r

k =

− Ai a j

∂x ∂x

k

j k

r

∂A rj ∂A j i

= A a , Since A j, k = − Ai

∂x k j, k

∂x k

j k

Proved.

Geometry 131

r

Let a vector field A(x ) and

C : x i = x (t ), t1 ≤ t ≤ t 2

i

r

be a curve in some region of E 3 . The vector A(x ) depend on the parameter t and if A(x ) is a

differentiable vector then

r r

dA ∂A dx j

= ⋅

dt ∂x j dt

r

α r dx

j

dA

= A, j aα

dt dt

Since we know

r ∂A α α r

∂A α r

= A, j a α = j + aα (See Pg. 127, Theo. 7.2)

∂x j ∂x i j

So,

r

dA ∂A α α i r dx j

= j + i j A aα dt

dt ∂x

r

dA dAα α i dx j r

= + A aα

dt dt i j dt

dAα α i dx j

The formula + A is called the absolute or Intrinsic derivative of A α with respect

dt

i j dt

δA α

to parameter t and denoted by .

δt

δA α dAα α i dx j δA δA

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

δt δt

Some Results

(i) If Ai be covariant vector

δAi dAi α dx β

= dt − i β Aα dt

δt

δA ij δAij i αj dx β j iα dx β

(ii) = + A + A

δt dt α β dt α β dt

δA ij δA ij

i α dx β α i dx β

(iii) = + A − A

δt dt α β j dt i β α dt

132 Tensors and Their Applications

δA ijk δA ijk

i α dx β α i dx β α i dx β

(iv) = + A jk − A − A

δt dt α β dt j β αk dt k β j α dt

EXAMPLE 5

δg ij

If g ij be components of metric tensor, show that = 0.

δt

Solution

δg ij dg ij α dx β α dx β

= − g αj − g iα

δt dt i β dt β j dt

∂g ij dx β α dx β α dx β

= − g α − g α

∂x β dt i β dt β j

j i

dt

∂g ij α α dx β

−

= ∂x β g α – g iα

β j dt

j

i β

δg ij ∂g ij dx β

= β − [iβ, j ] − [βj, i]

δt ∂x dt

α α

as gαj = [ i β, j ] and g i α = [β j , i ] .

i β β j

∂g ij

But = [iβ, j ] + [ jβ, i].

∂x β

δg ij

So, = 0.

δt

EXAMPLE 6

Prove that

d ( g ij A i A j ) δA i

= 2 g ij A i

dt δt

Solution

Since g ij Ai a j is scalar..

Then

d ( g ij A i A j ) δ(g ij A i A j )

=

dt δt

Geometry 133

δ( A i A j )

= g ij , since g ij is independent of t.

δt

δA i j δA j

= g ij A + Ai

δt δt

Interchange i and j in first term, we get

d ( g ij A i A j ) i δA j δA j

= g ij A + Ai as g ij is symmetric.

dt δt δt

d ( g ij A i A j ) δA j

= 2 g ij Ai

dt δt

Proved.

EXAMPLE 7

Prove that if A is the magnitude of Ai then

Ai , j Ai

A, j =

A

Solution

Given that A is magnitude of Ai . Then

Since

g ik A i A k = Ai A i

g ik A i A k = A 2

Taking covariant derivative w.r. to x j , we get

g ik A,i j A k + g ik A i A,kj = 2 AA, j

Interchange the dummy index in first term, we get

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 , j Ai

A, j =

A

Proved.

134 Tensors and Their Applications

Consider a curve

C : x i = x (t ), t1 ≤ t ≤ t 2 , (i = 1, 2, 3)

i

r

in some region of E 3 and a vector A localized at point P of C. If we construct at every point of C a

vector equal to A in magnitude and parallel to it in direction, we obtain a parallel field of vector along the

curve C.

3

3 X

Y 2

X

C

1

X

P

O 2

Y

1

Y

r r

if A is a parallel field along C then the vector A do not change along the curve and we can write

r

dA r

= 0. It follows that the components Ai of A satisfy a set of simultaneous differential equations

dt

∂A i

= 0 or

∂t

dAi i α dxβ

+ A =0

dt α β dt

This is required condition for the vector field Ai is parallel.

Let the parametric equations of the curve C in E 3 be

C : x i = x i (t ), t1 ≤ t ≤ t 2 (i = 1, 2, 3).

The square of the length of an element of C is given by

i j

ds 2 = g ij dx dx ...(1)

and the length of arc s of C is defined by the integral

dxi dx j

∫

t2

s= gij dt ...(2)

t1 dt dt

from (1), we have

dx i dx j

g ij =1 ...(3)

ds ds

Geometry 135

dx i

Put = λi . Then equation (3) becomes

ds

g ij λi λj = 1 ...(4)

r r

The vector λ , with components λi , is a unit vector. Moreover,, λ is tangent to C, since its

dx i

components λi , when the curve C is referred to a rectangular Cartesian coordinate Y,, becomes λi = .

ds

These are precisely the direction cosines

r of the tangent vector to the curve C.

r

r

Consider a pair of unit vectors λ and µ (with components λi and µ i respectively) at any point

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

λ

λ + dλ

µ

C

r + dr

r

O

Fig. 7.6

r r

The cosine of the angle θ between λ and µ is given by the formula

cos θ = g ij λ λ

i j

...(5)

r r

and if λ and µ are orthogonal, then equation (5) becomes

g ij λi µ j = 0 ...(6)

r

Any vector µ satisfying equation (6) is said to be normal to C at P..

Now, differentiating intrinsically, with respect to the are parameter s, equation (4), we get

δλi j δλj i

g ij λ + gij λ =0 …(7)

δs δs

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

δλi j δλi j

g ij λ + gij λ =0

δs δs

Since g ij is symmetric. Then

δλj

2 g ij λi =0

δs

136 Tensors and Their Applications

δλ j

⇒ g ij λi =0

δs

δλj

we see that the vector either vanishes or is normal to C and if does not vanish

δs

δλj

we denote the unit vector co-directional with by µ j and write

δs

1 δλ j K = δλ

j

µj = ,

δs ...(8)

K δs

The vector µ j is called the Principal normal vector to the curve C at the point P and K is the

curvature of C.

r r

The plane determined by the tangent vector λ and the principal normal vector µ is called the

osculating plane to the curve C at P.

r

Since µ is unit vector

g ij µi µ j = 1 ...(9)

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

δλj j δλ j

g ij µ + g ij λi =0

δs δs

or

δµ j δλi j

g ij λi = − gij µ

δs δs

δλi

= − Kg ij µ i µ j Since = K µi

δs

δµ j

g ij λi = − K , since g ij µ i µ j = 1.

δs

δµ j

g ij λi +K =0

δs

δµ j

g ij λi + g ij λi λ j K = 0 as g ij λi λj = 1

δs

δµ j

g ij λi + Kλ j = 0

δs

δµ j

This shows that the vector + Kλ j is orthogonal to λi .

δs

Geometry 137

r

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

1 δµ j

v i = τ δs + Kλ

j

...(10)

δµi r r r

where τ = + Kλi the vector v will be orthogonal to both λ and µ .

δs

To choose the sign of τ in such a way that

g eijk λi µ j v k = 1 ...(11)

r r

so that the triad of unit vectors λ , µ and νr forms a right handed system of axes.

2

∂ xi

Since eijk is a relative tensor of weight −1 and g = it follows that ε ijk = g eijk is an

∂xj

obsolute tensor and hence left hand side of equation (11) is an invariant v k in equation (11) is determined

by the formula

v k = ε λi µ j

ijk

...(12)

α 1 ijk

where λ i and µ j are the associated vectors g iα λα and g iα µ and ε ijk = e is an absolute tensor..

g

The number τ appearing in equation (10) is called the torsion of C at P and the vector vr is the

binormal.

We have already proved that in Theorem 7.2, Pg. 127.

r

∂A r

= A,αi a α

∂x i

r

if the vector field A is defined along C, we can write

r

∂A ∂x i α ∂x r

i

= A a ...(13)

∂x i ∂s ∂s α

, i

δA α α dx

i

= A,i

δs ds

Then equation (13) becomes

r r r

dA δA α r ∂A dx i dA

= a ⋅ ⋅ ⋅ ; as = ...(14)

ds δs α ∂x i ds ds

r r

Let r be the position vector of the point P on C then the tangent vector λ is determined by

r

dr r r

= λi ai = λ

ds

from equation (14), we get

r r

d 2r dλ δλα r r

⇒ 2 = ds

= aα = c ...(15)

ds δs

138 Tensors and Their Applications

r r

where c is a vector perpendicular to λ.

r r

With each point P of C we can associate a constant K , such that c K = µ is a unit vector..

Since

r

c r

= µ

K

r 1 δλα r

µ = a , from (15)

K δs α

from equation (8), we get

r r α

µ = µα aα , since µ α = 1 δλ

K δs

The serret-frenet formulas are given by

1 δλi δλi δλ i

= Kµ , K > 0 whereK =

i

(i) µi = or

K δs δs δs

1 δµ i

i δµi δµi

(ii) ν = τ δs + K λ = τν – Kλ where τ = δs + K λ

i

i or i i

δs

k

δν

(iii) = – τµ k

δs

First two formulas have already been derived in article (7.7), equation (8) and (10).

Proof of (iii)

From equation (12), article (7.7), we have

ε ijk λ i µ j = ν k

Taking intrinsic derivative with respect to s, we get

δλi δµ j δνk

ε ijkµ j + ε ijk λ i =

δs δs δs

From formulas (i) and (ii), we get

δνk

ε ijk Kµ iµ j + εijk λ i (τν j – Kλ j ) =

δs

δνk

ε ijk λ i (τν j – Kλ j ) = , Since ε ijk µ i µ j = 0

δs

δνk

ε ijk λ iν j τ – Kε ijk λ j λi =

δs

Geometry 139

Since ε ijk λ i λ j = 0

δνk

ε ijk λi ν j τ =

δs

Since ε ijk λ i ν j = µ k , but ε ijk are skew-symmetric.

Then

ε ijk λ i ν j = – µ k

So,

δνk

– µkτ =

δs

δνk

= – τµ

k

or

δs

δν i

⇒ = – τµi

δs

This is the proof of third Serret-Frenet Formula

Expanded form of Serret-Frenet Formula.

dλi i j dx k d 2 xi i dx j dx k

(i) + λ i + = Kµ i

ds j k ds = Kµ or

ds 2 j k ds ds

dµi i j dx k

+ µ

ds = τν – Kλ

(ii) i i

ds jk

dv i i j dx k

+ ν

ds = – τµ

(iii) i

ds jk

EXAMPLE 8

Consider a curve defined in cylindrical coordinates by equation

x1 = a

2

x = θ( s )

x3 = 0

This curve is a circle of radius a.

The square of the element of arc in cylindrical coordinates is

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

1 2 1 2 2 2 3 2

It is easy to verify that the non-vanishing Christoffel symbols are (see Example 3, Page 61)

1 2 2 1

= x1 , = = 1 .

22

12

21 x

140 Tensors and Their Applications

dx i dθ

The components of the tangent vector λ to the circle C are λi = so that λ1 = 0, λ2 = ,

ds ds

λ3 = 0.

Since λ is a unit vector, gijλiλj= 1 at all points of C and this requires that

2 2

2 dθ

= a =1

ds

2

dθ

So, = 2 and by Serret-Frenet first formula (expanded form), we get

1

ds a

dλ1 1 j dx k 1 2 dx 2 1

Kµ = ds j k ds = 2 2λ ds = –

+1 λ

a

dλ2 2 j dx k 2 2 dx1

Kµ2 = ds + j k λ ds = 2 1λ ds = 0

dλ3 3 j dx k

Kµ = ds + j k λ ds = 0

3

1

Since µ is unit vector, g ij µi µ j = 1 and it follows that K = , 1 = –1, µ 2 = 0, µ 3 = 0

a µ

Let Ai be a vector field defined along a curve C in E 3 such that

i

C : x i = x (s ) . s1 ≤ s ≤ s 2, (i = 1, 2, 3).

s being the arc parameter.

If the vector field Ai is parallel then from article 7.6 we have

δA i

=0

δs

dAi i α dx β

or + A

ds = 0 (1)

ds α β

We shall make use of equation (1) to obtain the equations of a straight line in general curvilinear

r

coordinates. The characteristic property of straight lines is the tangent vector λ to a straight line is

r

directed along the straight line. So that the totality of the tangent vectors λ forms a parallel vector field.

Geometry 141

dxi

Thus the field of tangent vector λ =

i

must satisfy equation (1), we have

ds

δλi d 2 xi i dx α dx β

= + = 0

δs ds 2 α β ds ds

d 2 xi i d x α d x β

The equation + = 0 is the differential equation of the straight line.

ds 2 α β ds d s

EXERCISE

d (gij AiB j ) δA i j δB i

1. Show that = g ij B + g ij A i

dt δt δt

∂ Ai ∂A j

2. Show that Ai , j – A j , i = –

∂x j ∂ xi

d ( gij Ai B j )

= Ai,k B + Bi,k A

i i

4. Show that

dxk

5. Show that

δ2 λi dK i

= µ – K ( τνi – Kλi )

ds 2 ds

δ 2µi dτ i dK i

= ν – ( K 2 + τ2 )µi – λ

δs2 ds ds

δ 2νi dτ

= τ( Kλ – τv ) – µ

i i i

δs2 ds

6. Find the curvature and tension at any point of the circular helix C whose equations in cylindrical

coordinates are

C : x1 = a, x2 = θ, x 3 = θ

Show that the tangent vector λ at every point of C makes a constant angle with the direction of X 3 -

axis. Consider C also in the form y1 = a cosθ, y2 = a sinθ, y3 = θ. Where the coordinates yi are

rectangular Cartesion.

CHAPTER – 8

ANALYTICAL MECHANICS

8.1 INTRODUCTION

Analytical mechanics is concerned with a mathematical description of motion of material bodies subjected

to the action of forces. A material body is assumed to consist of a large number of minute bits of matter

connected in some way with one another. The attention is first focused on a single particle, which is

assumed to be free of constraints and its behaviour is analyzed when it is subjected to the action of

external forces. The resulting body of knowledge constitutes the mechanics of a particle. To pass from

mechanics of a single particle to mechanics of aggregates of particles composing a material body, one

introduces the principle of superposition of effects and makes specific assumptions concerning the

nature of constraining forces, depending on whether the body under consideration is rigid, elastic,

plastic, fluid and so on.

1. Every body continues in its state of rest or of uniform motion in a straight line, except in so

far as it is compelled by impressed forces to change that state.

2. The change of motion is proportional to the impressed motive force and takes place in the

direction of the straight line in which that force is impressed.

3. To every action there is always an equal and contrary reaction; or the mutual actions of two

bodies are always and oppositely directed along the same straight line.

The first law depends for its meaning upon the dynamical concept of force and on the kinematical

idea of uniform rectilinear motion.

The second law of motion intorduces the kinematical concept of motion and the dynamical idea

of force. To understand its meaning it should be noted that Newton uses the term motion in the sense

of momentum, i.e., the product of mass by velocity, this, "change of motion" means the time of change

of momentum.

In vector notation, the second law can be stated as

r

r d (mv )

F = dt … (1)

Analytical Mechanics 143

r r

F = ma … (2)

r

r d (mv )

from (1) if F = 0 then dt = 0.

So that

r

m v = constant.

hence vr is constant vector..

Thus the first law is a consequence of the second.

The third law of motion states that accelerations always occur in pairs. In term of force we may

say that if a force acts on a given body, the body itself exerts an equal and oppositely directed force on

some other body. Newton called the two aspects of the force of action and reaction.

THEOREM 8.1 The work done in displacing a particle along its trajectory is equal to the change in

the kinetic energy of particle.

Proof: Let the equation of path C of the particle in E 3 be

i

C : x i = x (t ) … (1)

and the curve C the trajectory of the particle. Let at time t, particle is at P {x i (t )}.

If v i be the component of velocity of moving particle then

dx i

vi = … (2)

dt

and if ai be the component of acceleration of moving particle then

δv i dv i i j d x k

ai = δt = dt + j k v d t

d 2 x i i dx j dx k

ai = + … (3)

dt 2 j k dt dt

δv i i

where is the intrinsic derivative and the are the Christoffel symbols calculated from the

δt jk

metric tensor gij

If m be the mass of particle. Then by Newton’s second law of motion

δv i

Fi = m = ma i … (4)

δt

r

We define the element of work done by the force F in producing a displacement drr by invariant

r

dW = F .drr .

r

Since the components of F and drr are F i and dxi respectively..

144 Tensors and Their Applications

Then

dW = gij F i dxj … (5)

j i

= F j dx where F j = gij F

The work done in displacing a particle along the trajectory C, joining a pair of points P 1 and P 2, is

line integral

∫

P2

W= Fi dx i … (6)

P1

using equation (4) then equation (6) becomes

δv i

∫

P2

W= m g ij dx j

P1 δt

δv i dx j

∫

P2

= m g ij dt

P1 δt dt

δv i j

∫

P2

W= m g ij v dt … (7)

P1 δt

Since g ij v i v j is an invariant then

δ (g ij vi v j ) d

= ( g v iv j )

δt dt ij

d δvi j

or ( g ij v i v j ) = 2 g ij v

dt δt

δvi j 1 d

⇒ g ij v = (g ij v i v j )

δt 2 dt

using this result in equation (7), we get

P2 m d

W= ∫P1

( g v i v j ) dt

2 dt ij

m

[g v i v j ]PP12

W=

2 ij

Let T2 and T1 is kinetic energy at P 2 and P 1 respectively.

W = T2 – T1

m m v2

where T = gij v i v j = is kinetic energy of particle.

2 2

We have the result that the work done by force F i in displacing the particle from the point P 1 to

1

the point P 2 is equal to the difference of the values of the quantity T = mv 2 at the end and the

2

beginning of the displacement.

∫

P2

The force field F i is such that the integral W = Fi dx i is independent of the path.

P1

Therefore the integrand F i dxi is an exact differential

dW = F i dxi … (8)

Analytical Mechanics 145

of the work function W. The negative of the work function W is called the force potential or

potential energy.

We conclude from equation (8) that

∂V

Fi = – i … (9)

∂x

i

where potential energy V is a function of coordinates x . Hence, the fields of force are called conservative

∂V

if F i = – .

∂x i

THEOREM 8.2 A necessary and sufficient condition that a force field Fi, defined in a simply

connected region, be conservative is that Fi,j = Fj,i.

∂V

Proof: Suppose that F i conservative. Then F i = –

∂x i

Now,

∂Fi k

F i,j = − Fk

∂x j i j

∂V

∂ − i

∂x − k F

F i,j = k

∂x j i j

∂ 2V k

=− − Fk … (1)

∂x ∂x j i

j i

and

∂F j k

F j,i = − Fk

∂x j i

i

Similarly,

∂ 2V k

F j,i = – − Fk … (2)

∂x ∂x

i j

ji

From equation (1) and (2), we get

F i,j = F j,i

conversely, suppose that F i,j = F j,i

Then

∂Fi k ∂F j k

− Fk = − Fk

∂x j

i j ∂x i j i

⇒

∂Fi ∂F j k

as due to symmetry..

j =

∂x ∂x i i j

146 Tensors and Their Applications

∂V

Take F i = −

∂x i

Then

∂Fi ∂ 2V

j =

− j i

∂x ∂x ∂x

∂ 2V

= − i j

∂x ∂x

∂ ∂V

= −

∂x i ∂x j

∂Fi ∂F j

j =

∂x ∂x i

∂V

So, we can take Fi = − .

∂x i

Hence, F i is conservative.

Consider a particle moving on the curve

C : x i = x i (t )

At time t, let particle is at point P (xi).

1

The kinetic energy T = mv 2 can be written as

2

1

T = m g ij x& x&

i j

2

Since x& i = v i.

1

or T= m g jk x& j x& k … (1)

2

Differentiating it with respect to x& i , we get

∂T 1 ∂x& j k &k

j ∂x

m g &

x + &

x

= 2 ∂x& i & i

∂x& i

jk

∂x

=

1

2

(

m g jk δij x& k + δki x& j )

1 1

= m g jk δ ij x& k + m g jk δ ki x& j

2 2

=

1

2

(

m gik x& k + g ji x& j )

1

= m (gij x& + gij x& ) gij = g ji

j j

as

2

Analytical Mechanics 147

∂T

= m g ij x& j

∂x& i

∂T

or = m g ik x& k …(2)

∂x& i

Differentiating equation (2) with respect to t, we get

d ∂T

= m

dt ∂x& i

d

g x& k

dt ik

( )

d k

= m g ik x& + g ik &x&

k

dt

∂g ik dx j k

= ∂x j dt x& + g ik &x&

k

m

d ∂T ∂g

= m ikj x& x& + m g ik &x&

j k k

… (3)

dt ∂x& i ∂x

1

Since T = m g jk x& j x&k

2

Differentiating it with respect to xi, we get

∂T 1 ∂g jk j k

i =

m x& x& … (4)

∂x 2 ∂x i

Now,

d ∂T ∂T ∂g 1 ∂g

− i = m ikj x& j x&k + m g ik&x&k − m jki x& j x& k

dt ∂x& i ∂x ∂x 2 ∂x

1 gik j k 1 1 ∂g

= m g ik &x& + m j x& x& + m g ik x& j x& k − m jki x& j x& k

k

2 ∂x 2 2 ∂x

1 g ik j k 1 ∂g ij k j 1 ∂g jk j k

= m g ik &x& + x& x& + m k x& x& − m i x& x&

k

m

2 ∂x j 2 ∂x 2 ∂x

1 ∂gik ∂g ij ∂g jk j k

= m g ik &x& + 2 m j + k − x& x&

k

∂x ∂x ∂x i

148 Tensors and Their Applications

[

= m g il x&&l + g il [ jk , i ]x& j x& k ]

l l

= m g il &x&l + x& j x& k , as g il [ jk , i] =

jk j k

d ∂T ∂T l j k

Since a = x&& + j k x& x&

l l

− = m g il a l ,

dt ∂x& i ∂x i

where al is component of acceleration

d ∂T ∂T

or − = m ai

dt ∂x& i ∂x i

d ∂T ∂T

− = Fi … (5)

dt ∂x& i ∂x i

where F i = m ai, component of force field. The equation (5) is Lagrangean equation of Motion.

∂V

For a conservative system, F i = – . Then equation (5) becomes

∂x i

d ∂T ∂T ∂V

− i = − i

dt ∂x& i ∂x ∂x

d ∂T ∂(T − V )

or − =0 … (6)

dt ∂x& i ∂x i

Since the potential V is a function of the coordinates xi alone. If we introduce the Lagrangean

function

L=T–V

Then equation (6) becomes

d ∂L ∂L

− =0 … (7)

dt ∂x& i ∂x i

EXAMPLE 1

Show that the covariant components of the acceleration vector in a spherical coordinate system

with

ds 2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2 are

a2 =

d

dt

[ ]

( x1 )2 x& 2 − (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2

and a3 =

d

dt

[

(x1 sin x 2 ) 2 x& 3 ]

Analytical Mechanics 149

Solution

In spherical coordinate system, the metric is given by

ds2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2

If v is velocity of the paiticle then

2 2 2

ds dx1 2 3

2

1 2 dx 2 2 dx

2 = + ( x ) + ( x1

sin x )

dt dt

v = dt dt

v 2 = ( x&1 )2 + (x1 ) 2 ( x& 2 ) 2 + ( x1 sin x 2 )2 (x& 3 ) 2

If T be kinetic energy then

1 2

T= mv

2

1

2

[

T = m (x& ) + ( x ) (x& ) + ( x sin x ) (x& )

1 2 1 2 2 2 1 2 2 3 2

] … (1)

By Lagrangean equation of motion

d ∂T ∂T

− = F i and m ai = F i

dt ∂x& i ∂x i

where F i and ai are covariant component of force field and acceleration vector respectively.

So,

d ∂T ∂T

− = m ai … (2)

dt ∂x& i ∂x i

Take i = 1,

d ∂T ∂T

m a1 = −

dt ∂x&1 ∂x 1

from (1), we get

m a1 =

1 d m

[

m (2x&1 ) − 2x 1 ( x& 2 )2 + 2 x1 (sin x 2 ) 2 ( x& 3 )2

2 dt 2

]

a1 =

dx&1

dt

[

− x1 ( x& 2 )2 + x1 (sin x 2 ) 2 ( x& 3 )2 ]

a1 = &x&1 − x1 (x& 2 )2 − x1 (x& 3 sin x 2 ) 2

Take i = 2,

d ∂T ∂T

m a2 = − 2

dt ∂x& 2 ∂x

m a2 =

1 d 1 2 2 1

[ ]

m ( x ) 2 x& − m 2( x1 )2 sin x 2 cos x 2 (x&3 ) 2

2 dt 2

a2 =

d

dt

[ ]

( x1 )2 x& 2 − (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2

150 Tensors and Their Applications

Take i = 3

d ∂T ∂T

m a3 = −

dt ∂x&3 ∂ x 3

m a3 =

1 d

[ ]

m 2( x& 3 ) ( x1 sin x 2 )2 − 0

2 dt

a3 =

dt

[

d 3 1 2

x& ( x ) (sin x 2 ) 2 ]

EXAMPLE 2

Use Lagrangean equations to show that, if a particle is not subjected to the action of forces then

its trajectory is given by yi = ait + bi where ai and bi are constants and the yi are orthogonal cartesian

coordinates.

Solution

If v is the velocity of particle. Then we know that,

v 2 = g ij y& i y& j

where yi are orthogonal cartesian coordinates.

Since

gij = 0, i ≠ j

gij = 1, i = j

So,

v 2 = ( y& i ) 2

But,

1 2

T=mv , T is kinetic energy..

2

1

T = m( y& )

i 2

2

The Lagrangean equation of motion is

d ∂T ∂T

−

dt ∂y& i ∂yi

= Fi

So, F i = 0

d 1

Then m 2 y& i – 0 = 0

dt 2

dy& i

m =0

dt

Analytical Mechanics 151

dy& i

or =0

dt

⇒ y& i = ai

⇒ yi = ait + bi

i i

where a and b are constant.

EXAMPLE 3

Prove that if a particle moves so that its velocity is constant in magnitude then its acceleration

vector is either orthogonal to the velocity or it is zero.

Solution

If v i be the component of velocity of moving particle then

dx i

vi = or v i = x& i

dt

given |v| = constant.

Since

g ij vi v j = | v |2 = constant

Taking intrinsic derivative with respect to t, we get

δ

(g v i v j ) = 0

δt ij

δv i j δv j

g ij v + vi =0

δt δt

δv i j δv j

g ij v + g ij v i =0

δt δt

δv i j δv i

g ij v + g ji v j = 0, (Interchange dummy index i and j in second term)

δt δt

δv i j

2 g ij v = 0 as gij = gji

δt

δvi j

g ij v =0

δt

δv i δv i

This shows that acceleration vector is either orthogonal to v i or zero i.e., = 0.

δt δt

152 Tensors and Their Applications

(i) Free-Moving Particle

If a particle is not subjected to the action of forces, the right hand side of equation (5), 148, vanishes.

Then we have

d ∂T ∂T

− =0 … (1)

dt ∂x& i ∂x i

1

If xi be rectangular coordinate system, then T = m y& i y& i .

2

Hence, the equation (1) becomes m &y&i = 0. Integrating it we get yi = ait + bi, which represents

a straight line.

(ii) Simple Pendulum

Let a pendulum bob of mass m be supported by an extensible string. In spherical coordinates, the

metric is given by

ds2 = dr 2 + r 2 d φ2 + r 2 sin 2 φ d θ 2

If T be the kinetic energy, then

1 2 1

T= … (1)

2 2

O

Y2

φ

R

mg cos φ

Y1 θ P

mg sin φ

Y3 mg

Fig. 8.1.

d ∂T ∂T

− = Fi i = 1, 2, 3

dt ∂x& i ∂x i

x1 = r , x 2 = φ, x 3 = θ.

So, take x1 = r

Analytical Mechanics 153

d ∂T ∂T

− = mg cos φ − R

dt ∂r& ∂r

from (1), we have

R

&r& − rφ&2 − r sin 2 φ θ&2 = g cos φ − … (2)

m

Take x2 = φ, we have

and take x3 = θ, we have

d 2& 2

(r θ sin φ) = 0 … (4)

dt

If the motion is in one plane, we obtain from equations (2), (3), and (4), by taking θ& = 0.

R

r&& − r φ& 2 = g cos φ − m

If r& = 0, we get, &φ& = − sinφ which is equation of simple pendulum supported by an

g

r

inextensible string. For small angles of oscillation the vibration is simple harmonic. For large vibration

the solution is given in the term of elliptic functions.

If a particle is at the point P 1 at the time t1 and at the point P 2 at the time t2, then the motion of the

particle takes place in such a away that

∫

t2

(δT + Fi δxi ) dt = 0

t1

where xi = xi (t) are the coordinates of the particle along the trajectory and xi + δx i are the coordinates

along a varied path beginning at P 1 at time t1 and ending at P 2 at time t2.

Proof: Consider a particle moving on the curve

C : x i = x (t ),

i

t1 ≤ t ≤ t 2

At time t, let particle is at P(xi). If T is kinetic energy. Then

1

T= m g ij x& i x& j

2

or T = T ( x i , x& i ) i.e, T is a function of xi and x& i . Let C ' be another curve, joining t1 and t2 close

to be C is

C' : x i (ε, t ) = x i (t ) + δx i (t)

154 Tensors and Their Applications

At t1 and t2

xi = x i = x i + ε δ x i

⇒ δx i (t1 ) = 0 and δx i (t 2 ) = 0

But T = T ( x i , x& i ).

If δT be small variation in T.. Then

∂T i ∂T i

δT = ∂x&i δx& + ∂x i δx

Now,

∂T ∂T

∫ {(δT + F )δx }dt = ∫

t2

i δx i + i δx& i + Fi δx i dt

t2

i

t1 i t1 ∂x ∂x&

∂T i ∂T i

∫ ∫ ∫

t2 t2 t2

= δx dt + δx& dt + Fi δx i dt

t1 ∂x i t1 ∂x&i t1

∂T

Integrating second term by taking as 1st term

∂x& i

∂T i ∂T i 2

t

d ∂T i

∫ ∫ ∫

t2 t2 t2

= δx dt + ∂x& i δx − δx dt + Fi δxi dt

t1 ∂x i t1 t1 dt ∂x&i t1

Since δx i (t1 ) = 0, δx i (t 2 ) = 0.

∂T i 2

t

then & i δx = 0.

∂x t1

So,

∫ (δT + F δx )dt = ∫

∂T d ∂T i

∫t

t2 t2 i t2

i

δ x dt − i δ x dt

∂ x&

i i

t1 t1

∂x 1 dt

∫

t2

+ Fi δx i dt

t1

t2 ∂T i

∫ (δT + F δx )dt = ∫

t2 d ∂T

i − i + Fi δx dt

i

∂x dt ∂x&

i

t1 t1

since particle satisfies the Lagrangean equation of motion. Then

d ∂T ∂T

− = Fi

dt ∂x& i ∂x i

∂T d ∂T

or − = – Fi

∂x i dt ∂x& i

Analytical Mechanics 155

So,

∫ ∫

t2 t2

(δT + Fi δx i ) dt = [−Fi + Fi ] δxi dt

t1 t1

∫

t2

(δT + Fi δx i ) dt = 0 Proved.

t1

THEOREM 8.3 The motion of a particle in a conservative field of force is such that the sum of its

kinetic and potential energies is a constant.

Proof: Consider a particle moving on the curve

C : x i = x i (t ), t1 ≤ t ≤ t 2

i

At time t, let particle is at P (x ). If T is kinetic energy. Then

1

T= m g ij x& i x& j

2

1

or T= m g ij v i v j

2

As T is invariant. Then

Taking intrinsic derivative with respect to t, we get

dT δT

=

dt δt

δ 1

= m gij v i v j

δt 2

1 δv i j δv j

= m g ij v + vi

2 δt δt

1 δv i j δv j i

= m gij v + g ij v

2 δt δt

1 δvi j δvi j

= 2 m gij δt v + g ji δt v , Since i and j are dummy indices.

1 δv i j

= m 2 g v as g ij = g ji

δt

ij

2

dT δv i j

= m gij v

dt δt

dT δv j i

or = m g v

δt

ij

dt

156 Tensors and Their Applications

δv j

j i

= m g ij a v as = aj

δt

dT

= m ai v i, since gij aj = ai

dt

dT

= F i v i,

dt

But given F i is conservative, then

∂V

Fi = – , where V is potential energy..

∂x i

So,

dT ∂V i

=– v

dt ∂x i

∂V dx i

= −

∂x i dt

dT dV

= −

dt dt

dT dV d

+ = 0 ⇒ (T + V ) = 0

dt dt dt

⇒ T + V = h, where h is constant.

Proved.

Let us consider the integral

∫

P2

A= mv.ds (1)

p1

C : x i = x (t ), t1 ≤ t ≤ t 2

i

where C is the trajectory of the particle of mass m moving in a conservative field of force.

In the three dimensional space with curvilinear coordinates, the integral (1) can be written as

dx i

∫

P2

A= m g ij dx j

p1 dt

dx i dx j

∫

t ( P2 )

= m gij dt

t ( P1 ) dt dt

Analytical Mechanics 157

1 dx i dx j

Since T = m g ij , we have

2 dt dt

∫

t ( P2 )

A= 2T dt

t ( P1 )

This integral has a physical meaning only when evaluated over the trajectory C, but its value can

be computed along any varied path joining the points P 1 and P 2.

Let us consider a particular set of admissible paths C' along which the function T + V, for each

value of parameter t, has the same constant value h. The integral A is called the action integral.

The principle of least action stated as “of all curves C' passing through P 1 and P 2 in the

neighbourhood of the trajectory C, which are traversed at a rate such that, for each C', for every value

of t, T + V = h, that one for which the action integral A is stationary is the trajectory of the particle.”

In the solution of most of the mechanical problems it is more convenient to use some other set of

coordinates instead of cartesian coordinates. For example, in the case of a particle moving on the

surface of a sphere, the correct coordinates are spherical coordinates r, θ, φ where θ and φ are only

two variable quantities.

Let there be a particle or system of n particles moving under possible constraints. For example, a

point mass of the simple pendulum or a rigid body moving along an inclined plane. Then there will be

a minimum number of independent coordinates required to specify the motion of particle or system of

particles. The set of independent coordinates sufficient in number to specify unambiguously the system

configuration is called generalized coordinates and are denoted by q 1 , q 2 , ... q n where n is the total

number of generalized coordinates or degree of freedom.

Let there be N particles composing a system and let x(i α) , (i = 1, 2,3), (α = 1, 2,...N ) be the positional

coordinates of these particles referred to some convenient reference frame in E 3. The system of N free

particles is described by 3N parameters. If the particles are constrained in some way, there will be

certain relations among the coordinates x(i α) and suppose that there are r such independent relations,

f i ( x1(1) , x(21) , x 3(1 ) ; x1( 2 ) , x(22 ) , x(32 ) ;...x1( N ) x(2N ) x(3N ) ) = 0, (i = 1, 2, ..., r) … (1)

By using these r equations of constraints (1), we can solve for some r coordinates in terms of the

remaining 3N – r coordinates and regard the latter as the independent generalized coordinates qi. It is

more convenient to assume that each of the 3N coordinates is expressed in terms of 3N – r = n

independent variables qi and write 3N equations.

x(i α) = x(i α) (q1 , q 2 ,...,q n , t ) … (2)

where we introduced the time parameter t which may enter in the problem explicitly if one deals with

moving constraints. If t does not enter explicitly in equation (2), the dynamical system is called a

natural system.

The velocity of the particles are given by differentiating equations (2) with respect to time. Thus

∂x i(α ) ∂x i(α )

x&(i α) = q& j + … (3)

∂q j ∂t

158 Tensors and Their Applications

For symmetry reasons, it is desirable to introduced a number of superfluous coordinates qi and

describe the system with the aid of k > n coordinates q1, q2,..., qk. In this event there will exist certain

relations of the form

f j (q1 ....,q k , t ) = 0 … (4)

Differentiating it we get

∂f j i ∂f j

q& +

∂t = 0 … (5)

∂q i

It is clear that they are integrable, so that one can deduce from them equations (4) and use them

to eliminate the superfluous coordinates .

In some problems, functional relations of the type

F j (q1 , q 2 ,...,q k ; q&1 ,...,q& k , t ) = 0, ( j = 1, 2, 3, ..., m) … (6)

arise which are non-integrable. If non-integrable relations (6) occurs in the problems we shall say that

the given system has k – m degrees of freedom, where m is the number or independent non-integrable

relations (6) and k is the number of independent coordinates. The dynamical systems involving non-

integrable relations (6) are called non-holonomic to distinguish them from holonomic systems in which

the number of degrees of freedom is equal to the number of independent generalized coordinates.

In other words, a holonomic system is one in which there are no non-integrable relations involving

the generalized velocities.

Let there be a system of particle which requires n independent generalized coordinates or degree of

freedom to specify the states of its particle.

The position vectors xr are expressed as the function of generalized coordinates q i , (i = 1, 2,...,n)

and the time t i.e.,

1 2 n

xr = xr (q , q ,...,q , t ); (r = 1, 2, 3, )

∂x r dq j ∂x r

x& = ∂q j dt + ∂t

r

∂x r j ∂x r

= q& + ,

∂q j ∂t ( j = 1, 2, ..., n)

Consider the relation, with n degree of freedom,

1 2 n

xr = xr (q , q ,...,q ) … (1)

involve n independent parameters qi. The velocities x& r in this case are given by

∂x r j

x&r= q& , (r = 1, 2, 3; j = 1, 2, ..., n) …(2)

∂q j

Analytical Mechanics 159

q k = q k (q1 ,...,q n ), (k = 1, 2, ..., n) … (3)

in accordance with the contravariant law.

The kinetic energy of the system is given by the expression of the form

1

T= Σ m g rs x& r x& s , (r,s = 1,2,3,) … (4)

2

where m is the mass of the particle located at the point xr. The grs are the components of the metric

tensor.

Substituting the value of x& r from equation (2), then equation (4) becomes

1 ∂x r ∂x s i j

T = 2 Σ m g rs i q& q&

∂q ∂q j

1

T= a q&i q& j … (5)

2 ij

∂x r ∂r s

where aij = Σ m g rs , (r, s = 1, 2, 3), (i, j = 1, ..., n)

∂q i ∂q j

1

Since T = a q&i q& j is an invariant and the quantities aij are symmetric, we conclude that the aij

2 ij

are components of a covariant tensor of rank two with respect to the transformations (3) of generalized

coordinates.

Since the kinetic energy T is a positive definite form in the velocities q& i , | aij |> 0. Then we

construct the reciprocal tensor aij .

Now, from art. 8.5, Pg. 146, by using the expression for the kinetic energy in the form (5), we

obtain the formula,

d ∂T ∂T l l j k

∂qi = a il &q& + j k q& q&

−

dt ∂q& i (6)

l

where the Christoffel symbol are constructed from the tensor akl.

jk

Put

l

q&&l + q& j q& k = Q l

jk

so, the equation (6), becomes

d ∂T ∂T

− l

dt ∂q& i ∂qi = ail Q

= Qi (i = 1, 2, ..., n) … (7)

160 Tensors and Their Applications

Now, from the realtions & j = , = &

q =

and qi dt q i and using equations

∂q ∂q j ∂qi ∂x i ∂q j ∂ ∂

(2) and (4).

Then by straightforward calculation, left hand member of equation (7) becomes

d ∂T ∂T ∂x r

−

dt ∂q& i

∂qi

= ∑

m a r

∂q i

… (8)

Also, Newton's second law gives

m ar = F r … (9)

where Fr's

are the components of force F acting on the particle located at the point P..

From the equation (9), we have

∂x r ∂x r

∑ m ar

∂q i

= ∑ Fr

∂q i

and equation (8) can be written as

d ∂T ∂T ∂x r

dt ∂q& i

−

∂qi = ∑ Fr

∂q i

… (10)

comparing (7) with (8), we conclude that

∂x r

Qi = ∑ Fr

∂q i

where vector Qi is called generalized force.

The equations

d ∂T ∂T

−

dt ∂q& i ∂qi = Qi … (11)

are known as Lagrangean equations in generalized coordinates.

They give a system of n second order ordinary differential equations for the generalized coordinates qi.

The solutions of these equations in the form

C : q i = qi (t)

Represent the dynamical trajectory of the system.

If there exists a functions V (q1 , q 2 ,...,q" ) such that the system is said to be conservative and for

such systems, equation (11) assume the form

d ∂L

∂L

−

∂qi = 0

dt ∂q& i

… (12)

where L = T – V is the kinetic potential.

Analytical Mechanics 161

Since L (q, q& ) is a function of both the generalized coordinates and velocities.

∂L i ∂L i

= & i q&& + i q&

dL

… (13)

dt ∂q ∂q

∂L d ∂L

.

dt ∂q& i

from (12), we have i =

∂q

Then equation (13), becomes

dL ∂L i d ∂L i

= q&& + i q&

dt ∂q& i dt ∂q&

d ∂L i

q&

dt ∂q& i

= … (14)

∂L i ∂T i

q& = q& = 2T

&

∂q i

∂q& i

1

since T= a q& i q& j .

2 ij

Thus, the equation (14) can be written in the form

d ( L − 2T ) d (T + V )

= =0

dt dt

which implies that T + V = h (constant).

Thus, along the dynamical trajectory, the sum of the kinetic and potential energies is a constant.

STOKE'S THEOREM IN TENSOR NOTATION

r

Let F be a vector point function in a closed region V bounded by the regular surface S. Then

r r

∫ ∫

div F = F ⋅ nˆ ds … (1)

V S

Briefly the theorem states that the integral with subscript V is evaluated over the volume V while

r

the integral in the right hand side of (1) measures the flux of the vector quantity F over the surface S.

r

In orthogonal cartesian coordinates, the divergence of F is given by the formula

r ∂F 1 ∂F 2 ∂F 3

div F = + + … (2)

∂x 1 ∂x 2 ∂x 3

162 Tensors and Their Applications

r

If the components of F relative to an arbitrary curvilinear coordinate system X are denoted by

F i then the covariant derivative of F i is

∂F i i k

F ,ij = + F

∂x j k j

r

The invariant F, ij in cartesian coordinates represents the divergence of the vector field F .

Also,

r

F ⋅ nˆ = g ij F n = F ni since g ij n = ni

i j i j

∫ F dV ∫ F n dS

i i

,i i

= … (3)

V S

Let φ ( x1 , x 2 , x3 ) and ψ (x1 , x 2 , x3 ) be two scalar function in V.. Let φ i and ψ i be the gradients of φ

and ψ respectively, so that

∂φ ∂ψ

and ∇φ = φ i = ∇ψ = ψ i =

∂xi ∂x i

Put F i = φψi and from the divergence of we get

F,i j = g ij Fi , j = g ij (φψi , j + ψi φ j )

Substituting this in equation (3), we get

∫g (φψi , j + ψ i φ j )dV ∫ φψ n dS

ij i

= i … (4)

V S

Since ∇ψ = ψi , then

g ij ψi , j = ∇ 2 ψ … (5)

g ij ψi φ j = ∇φ.∇ψ

where ∇ denote the gradient and ∇ 2 denote the Laplacian operator..

Hence the formula (4) can be written in the form

∫(g φψi , j + g ij ψi .φ J ) dV ∫ φ n̂ ⋅ ∇ψ dS

ij

=

V S

∫ (φ∇ ψ + ∇φ ⋅ ∇ψ) dV = ∫ φ n̂ ⋅ ∇ψ dS

2

V S

∫ φ ∇ ψ dV = ∫ φ n̂ ⋅ ∇ ψ − ∫ ∇φ ⋅ ∇ψ dV

2

… (6)

V S V

∂ψ .

where nˆ ⋅ ∇ψ = ψi n =

i

∂n

Analytical Mechanics 163

∫ ψ∇ φ dV ∫ ψn̂ ⋅ ∇φ − ∫ ∇ψ ⋅ ∇φ dV

2

= … (7)

V S V

Subtracting equation (5) from equation (6), we get

∂ψ ∂φ

∫ (φ∇ ψ − ψ∇ φ) dV ∫ φ ∂n − ψ ∂n dS

2 2

= … (8)

V S

(iii ) Expansion form of the Laplacian Operator

The Laplacian of ψ is given by

∇ 2 ψ = g ψi , j from (5)

ij

when written in the terms of the christoffel symbols associated with the curvilinear coordinates

i

x covering E 3,

∂2 ψ k ∂ψ

∇ 2 ψ = g i j − k

ij

… (9)

∂x ∂x i j ∂x

i

and the divergence of the vector F is

∂F i i

F,ii = + F j

…(10)

∂x i j i

i ∂

But we know that = log g

j i ∂ xj

The equation (10) becomes

∂F i ∂

F,ii = + j log g F j

∂x ∂x

i

1 ∂( g F i )

or F,ii = … (11)

g ∂x i

∂ψ

If putting F i = g

ij

= g ij ψ j in equation (11), we get

∂x j

∂ψ

∂ g g ij j

1 ∂x

g ij ψ j , i = … (12)

g ∂x i

∇ 2 ψ = g ψ j, i

ij

164 Tensors and Their Applications

∂ψ

∂ g g ij j

1 ∂x

∇ 2 ψ = g ij ψ j ,i =

∂x i

g

It is expansion form of Laplacian operator.

(iv) Stoke's Theorem

r

Let a portion of regular surface S be bounded by a closed regular curve C and let F be any vector point

function defined on S and on C. The theorem of Stokes states that

r r

∫

nˆ. curl F ds = F .λ̂ ds ∫ … (13)

S C

r

where λ is the unit tangent vector to C and curl F is the vector whose components in orthogonal

cartesian coordinates are determined from

e1 e2 e3

∂ ∂ ∂ r

r

curl F = ∂ x1 ∂x 2 ∂x 3 = ∇ × F … (14)

F1 F2 F3

where ei being the unit base vectors in a cartesian frame.

We consider the covariant derivative F i,j of the vector F i and form a contravariant vector

Gi = – ε ijk F j , k … (15)

r

we define the vector G to be the curl of F .

r dx i

Since n̂ . curl F = ni G = −ε F j ,k ni and the components of the unit tangent vector λ and

i ijk

.

ds

Then equation (13) may be written as

∫

− εijk F j , k ni ds

= ∫ Fi

dx i

ds

ds … (16)

S C

r

∫ F dx

i

The integral i is called the circulation of F along the contour C.

c

The integral of the normal component of the gravitational flux computed over a regular surface S

containing gravitating masses within it is equal to 4 πm where m is the total mass enclosed by S.

Proof: According to Newton's Law of gravitation, a particle P of mass m exerts on a particle Q of unit

m

mass located at a distancer r from P. Then a force of magnitude F = 2 .

r

Consider a closed regular surface S drawn around the point P and let θ be the angle between the

unit outward normal to n̂ to S and the axis of a cone with its vertex at P.. This cone subtends an

element of surface dS.

Analytical Mechanics 165

r m cos θ r 2 dw

∫ F .nˆ dS = ∫ r 2 cos θ

S S

2

r dw

where dS = and dw is the solid angle subtended by dS.

cos θ

Thus, we have,

r

∫

F .nˆ dS

= ∫

m dw = 4 πm

… (1)

S S

n

F

θ

dS

dω

P

m

Fig. 8.2.

r

n

mi cos θ i

F .nˆ = ∑

i =1 ri 2

and total flux is

r n

∫ F .nˆ dS = 4 π ∑m

i =1

i … (2)

S

The result (2) can be easily generalized to continuous distributions of matter whenever such

distribution no where melt the surface S.

The contribution to the flux integration from the mass element ρ dV contained within V, is

r cos θρ dV

∫ F .nˆ dS = ∫ r2

dS

S S

cos θ ρ dV

r dS

∫ F .nˆ dS = ∫ ∫ r2 … (3)

S V

S

166 Tensors and Their Applications

where ∫ denotes the volume integral over all bodies interior to S. Since all masses are assumed to be

V

interior to S,r never vanishes. So that the integrand in equation (3) is continuous and one can interchange

to order of integration to obtain

r cos θ dS

∫ F .nˆ dS = ∫ ρ ∫

s r2

dV … (4)

V

S

cos θ dS

But ∫r2

= 4π. Since it represents the flux due to a unit mass contained within S.

S

Hence

r

∫ F.nˆ dS = ∫

4 π ρ dV = 4 πm

… (5)

V

S

where m denotes the total mass contained within S. Proved.

Gauss's theorem may be extended to cases where the regular surface S cuts the masses, provided

that the density S is piecewise continuous.

Let S cut some masses. Let S' and S" be two nearby surfaces, the first of which lies wholly within

S and the other envelopes S. Now apply Gauss's theorem to calculate the total flux over S" produced by

the distribution of masses enclosed by S since S" does not intersect them.

We have

r

∫( F .nˆ )i dS = 4 πm

S"

r

where the subscript i on F ⋅ nˆ refers to the flux due to the masses located inside S and m is the total

mass within S. On the other hand, the net flux over S' due to the masses outside S, by Gauss's theorem

is

r

∫

( F ⋅ nˆ)o ds = 0

S'

r

where the subscript o on F ⋅ nˆ refers to the flux due to the masses located outside S.

Now if we S' and S" approach S, we obtain the same formula (5) because the contribution to the

r

∫

total flux from the integral (F ⋅ nˆ)o dS is zero.

S'

By divergence theorem, we have

r r

∫ F .nˆ ds = ∫

V

div F dV

S

and by Gauss's Theorem,

r

∫ F.nˆ ds = 4 π∫ ρ dVV

S

from these, we have

Analytical Mechanics 167

r

∫ (div F − 4πρ ) dV = 0

v

Since this relation is true for an arbitrary V and the integrand is piecewise continuous, then

r

div F = 4 πρ

By the definition of potential function V, we have

r

F = −∇V

and div ∇V = ∇ 2V

So,

r

div F = 4 πρ

div ( −∇V ) = 4 πρ

∇ 2V = – 4 πρ

which is equation of poisson.

If the point P is not occupied by the mass, then ρ = 0. Hence at all points of space free of matter

the potential function V satisfies Laplace's equation

∇ 2V = 0

We find the solution of Poisson’s Equation by using Green’s symmetrical formula. We know that

Green's symmetrical formula

∂ψ ∂φ

∫ ∫

(φ∇ 2 ψ − ψ∇ 2 φ) dV = φ ∂n − ψ ∂n dS … (1)

S

V

where V is volume enclosed by S and φ and ψ are scalar point functions.

1

Put φ = where r is the distance between the points P ( x1 , x 2 , x3 ) and Q ( y1 , y 2 , y 3 ) and V is

r

the gravitational potential.

n

Q (y )

r

S

P(x)

Fig. 8.3.

168 Tensors and Their Applications

1

Since has a discontinuity at x i = y i , delete the point P(x) from region of integration by

r

surrounding it with a sphere of radius ε and volume V'. Apply Green’s symmetrical formula to the

1

region V – V' within which and V possess the desired properties of continuity..

r

2 1

In region V − V', ∇ φ = ∇ = 0.

2

r

Then equation (1) becomes

1 2 1 ∂ψ ∂ 1r 1 ∂ψ ∂ 1r

∫ r

∇ ψ dV = ∫

r ∂n

− ψ

∂n

dS +

∫

r ∂n

− ψ

∂n

ds … (2)

V −V' S S'

where n̂ is the unit outward normal to the surface S + S' bounding V – V' . S' being the surface of the

∂ ∂

sphere of radius ε and =− .

∂n ∂r

Now

1 ∂ψ ∂ 1r 1 ∂ψ ∂ 1r

∫

r ∂n

− ψ

∂n

dS =

∫ −

r ∂r

− ψ

∂r

dS

S' S'

1 ∂ψ ψ 2

= ∫ − r ∂r − r 2

r dw

S'

∂ψ

= − ∫ r ∂r + ψ dw

S'

1 ∂ψ ∂( 1r ) ∂ψ

= − ε

∫

r ∂n

− ψ

∂n ∫

dw − 4πψ

∂r r = ε … (3)

S' S'

where ψ is the mean value of V over the sphere S' and w denote the solid angle.

1 2 3

1 ∂ψ ∂ 1r

∫

r ∂n

− ψ

∂n = −4πψ (P )

S'

1 2 1 ∂ψ ∂ 1r

∫ r

∇ ψ dV = ∫

r ∂n

− ψ

∂n

dS − 4πψ (P )

V S

1

∫ r ∇ ψ dV

2

Since ε → 0 then = 0.

V'

Analytical Mechanics 169

1 1 2 1 1 ∂ψ 1 ∂ 1r

−

ψ(P ) = 4 π r

V

∇ ψ dV∫+

4π S r ∂n

dS − ∫ψ

4π S ∂n

dS ∫ … (4)

If ψ is regular at infinity, i.e., for sufficiently large value of r, ψ is such that

m ∂ψ m

(ψ) ≤ and ≤ … (5)

r ∂r r 2

where m is constant.

If integration in equation (4) is extended over all space, so that r → ∞ . Then, using equation (5),

equation (4) becomes

1 ∇ 2ψ

ψ(P ) = − ∫

4π ∞ r

dV … (6)

Hence, from (6), we get

dV

ψ(P ) = ∫ρ r

∞

EXERCISES

1. Find, with aid of Lagrangian equations, the trajectory of a particle moving in a uniform gravitational

field.

2. A particle is constrained to move under gravity along the line yi = ci s (i = 1, 2, 3). Discuss the motion.

3. Deduce from Newtonian equations the equation of energy T + V = h, where h is constant.

4. Prove that

∫ ψ n dS ∫ ∇ ψ dV

i 2

i =

S V

∂ψ

where ψ i = .

∂x i

5. Prove that the curl of a gradient vector vanishes identically.

CHAPTER – 9

Let C be a curve in a given Vn and let the coordinates x i of the current point on the curve expressed

as functions of the arc length s. Then the unit tangent t to the curve the contravariant components

dxi

t i = ds ...(1)

The intrinsic derivative (or desired vector) of t i along the curve C is called the first curvaturee

r r

vector of curve C relative to Vn and is denoted by p . The magnitude of curvature vector p is called

first curvature of C relative Vn and is denoted by K:

So,

K= g ij p i p j

r

where P i are contravariant components of p so that

i dx j

P = t , j ds

i

∂t i i

α i dx

= j + t

∂x α j ds

∂t i dx j α dx i

j

= + t

∂x j ds ds α j

dt i dxα dx j i

= ds + ds ds α j

d 2 xi dx j dx k i

= 2

+ , Replacing dummy index α by k

ds ds ds k j

Curvature of Curve, Geodesic 171

d 2 xi dx j dx k i i i

pi = ds 2 + ds ds j k as j k = k j

r

If n̂ is a unit vector in the direction of p , then we have

r

p = k nˆ

The vector n̂ is called the Unit principal normal.

9.2 GEODESICS

Geodesics on a surface in Euclidean three dimensional space may be defined as the curve along which

lies the shortest distance measured along the surface between any two points in its plane.

But when the problem of find the shortest distance between any two given points on a surface is

treated properly, it becomes very complicated and therefore we define the geodesics in V3 as follows:

(i) Geodesic in a surface is defined as the curve of stationary length on a surface between any

two points in its plane.

(ii) In V3 geodesic is also defined as the curve whose curvature relative to the surface is

everywhere zero.

By generalising these definitions we can define geodesic in Riemannian Vn as

(i) Geodesic in a Riemannian Vn is defined as the curve of minimum (or maximum) length

joining two points on it.

(ii) Geodesic is the curve whose first curvature relative to Vn is zero at all points.

THEOREM 9.1 The Euler condition for the integral

∫

t1

f (x i , x&i ) dt

to

to be staionary are

∂f d ∂f

− =0

∂x i dt ∂ x& i

dxi

where x& = dt i = 1, 2, 3,...

i

Proof: Let C be a curve in a Vn and A, B two fixed points on it. The coordinates x i of the current point

P on C are functions of a single parameter t. Let t0 and t1 be the values of the parameter for the points

A and B respectively.

To find the condition for the integral

∫

t1

f (x i , x&i ) dt ...(1)

t0

to be stationary.

Let the curve suffer an infinitesimal deformation to C ′, the points A and B remaining fixed while

the current points P(xi) is displaced to P' (xi + ηi) such that ηi = 0 at A and B both.

172 Tensors and Their Applications

P' C’

A B

P C

Fig. 9.1

So,

∫ F (x )

t1

I′ =

i

& i dt

+ ηi , x& i + η

t0

By Taylor's theorem

∂f ∂f

F ( x + h , y + k ) = f ( x, y ) + h + k + ⋅ ⋅ ⋅ ⋅

∂x ∂y

Then

t1 ∂F i ∂F i

∫ F ( x , x& ) + i η + i η& + ⋅ ⋅ ⋅ dt

i i

I′ = t0 ∂x ∂x

t1 ∂F ∂F i

∫ ∫

t1

I′ = F ( x i , x& i ) dt + ηi + η& dt

t0 t 0 ∂x i ∂x& i

t1 ∂F ∂f

I′ = I + ∫ i η i + i η& i dt

t0 ∂ x ∂ x&

t1 ∂F ∂F i

δI = I ′ − I = ∫

t0 ∂ x i

ηi + η& dt

∂ x&i

...(2)

∂ zi & j

where η& i = x

∂x j

Now,

∂F i 1

t

t1 ∂ F d ∂F i

∫

t1

∫ &

η = ∂ x&i η – η dt

i

dt

&i

t0 ∂ x t0

t0 dt ∂ x&i

d ∂F i

∫

t1

= − η dt ...(3)

t0 dt ∂ x&i

Curvature of Curve, Geodesic 173

∂F

t

1

since i ηi (t ) = 0, η i (t1 ) = ηi (t0 ) = 0

∂x& t0

ti ∂F d ∂F i

δI = ∫

t0 ∂ x i

− η dt

dt ∂ x&i ...(4)

The integral I is stationary if δI = 0 .

t1 ∂F d ∂F i

i.e., if ∫

t0 ∂ x i

− η dt = 0

dt ∂ x&i

Since ηi are arbitrary and hence the integrand of the last integral vanishes, so that

∂F d ∂F

− = 0, (i = 1, 2,..., n) … (5)

∂ xi dt ∂ x&i

Hence the necessary and sufficient condition for the integral (1) to be stationary are

∂F d ∂F

– = 0, (i = 1, 2, …, n)

∂x i dt dx i

These are called Euler's conditions for the integral I to be stationary.

To obtain the differential equations of a geodesic in a Vn , using the property that it is a path of minimum

(or maximum) length joining two points A and B on it.

Proof: Consider a curve C in Vn joining two fixed points A and B on it and x i (t ) be the coordinates of

point P on it.

The length of curve C is

dxi d x j

∫

B

s= gij dt ...(1)

A dt dt

ds dxi d x j

= gij

dt dt dt

Put

ds d xi d x j

= gij = F (say) ...(2)

dt dt dt

or s& = g ij x& i x& j = F

Then equation (1) becomes

∫

B

s= F dt ...(3)

A

174 Tensors and Their Applications

Since curve C is geodesic, then the integral (3) should be stationary, we have from Euler's

condition

∂F d ∂F

− =0 ...(4)

∂x i dt ∂ x& i

Differentiating equation (2) with respect to x k and x& k we get,

∂F 1 ∂g ij i j

= x& x&

∂x k

2 s& ∂ x k

∂F 1 1

and k =

2 gik x& i = g ik x&i

∂x& 2s& s&

d ∂F 1 1 ∂ g ik j i 1

= − 2 &s& g ik x& + x& x& + g ik &x& i

i

dt ∂ x& k s& s& ∂x j s&

Putting these values in equation (4), we get

1 ∂g ij i j 1 1 ∂ g ik j i 1

x& x& − − 2 &s&g ik x& i + x& x& + g ik &x&i = 0

2 s& ∂ x k s &

s ∂ x j &

s

&s& ∂g 1 ∂ g ij i j

g ik &x&i − g ik x& i + ikj − x& x& = 0

s& ∂x 2 ∂x k

&s&

g ik &x&i − g x& i + [k , ij ] x&i x& j = 0

&s ik

multiplying it by g km , we get

&s& km

g km gik &x&i − g g ik x&i + g km [k , ij ] x& i x& j = 0

s&

m

g km gik = δim and g [k , ij] = i j

km

But

&s& m m i j

x&&m − x& + x& x& = 0

s& i j

− + = 0 index i by k ...(5)

dt 2

s& dt j k dt dt

This is the differential equation for the geodesic in parameter t.

Taking s = t , s& = 1, &s& = 0. Then equation (5) becomes

d 2xm m dx j dx k

+ = 0 ...(6)

ds 2 j k ds ds

Curvature of Curve, Geodesic 175

dx k dx m

ds ds =0

,k

Then the intrinsic derivative (or derived vector) of the unit tangent to a geodesic in the direction of the

curve is everywhere zero. In otherwords, a geodesic of Vn is a line whose first curvature relative to

Vn is identically zero.

THEOREM 9.2 To prove that one and only one geodesic passes through two specified points lying in

a neighbourhood of a point O of a Vn .

OR

To prove that one and only one geodesic passes through a specified point O of Vn in a prescribed

direction.

Proof: The differential equations of a geodesic curve in a Vn are

d 2xm m dx j dx k

+ =0

ds 2 j k ds ds

These equations are n differential equations of the second order. Their complete integral involves 2n

arbitrary constants. These may be determined by the n coordinates of a point P on the curve and the n

components of the unit vector in the direction of the curve at P. Thus, in general, one and only one

geodesic passes through a given point in a given direction.

A cartesian coordinate system is one relative to which the coefficients of the fundamental form are

constants. Coordinates of this nature do not exists for an arbitrary Riemannian V n.. It is, however,

possible to choose a coordinate system relative to which the quantities gij are locally constant in the

neighbourhood of an arbitrary point P 0 of V n. Such a cartesian coordinate system is known as geodesic

coordinate system with the pole at P 0.

The quantities g ij are said to be locally constants in the neighbourhood of a point P0 if

∂g ij

= 0 at P0

∂ xk

∂g ij

and ≠ 0 elsewhere

∂ xk

k

This shows that [ij, k ] = 0 , i j = 0 at P0 .

Since the covariant derivative of A ij with respect to xk is written as

∂Aij h h

Aij , k = ∂x k − Aih − A h j , see pg. 71

j k i k

176 Tensors and Their Applications

The covariant derivative of Aij at P0 with respect to x k reduces to the corresponding ordinary

derivatives. Hence

∂Aij

Aij , k = at P0

∂x k

THEOREM 9.3 The necessary and sufficient condition that a system of coordintes be geodesic with

pole at P0 are that their second covariant derivatives with respect to the metric of the space all vanish

at P0.

Proof: We know that (equation 8, Pg. 65)

∂2xs ∂x s k ∂x p ∂x q s

= −

∂x j ∂ x j ∂x k i j ∂x i ∂x j p q

∂ ∂x s ∂x s k ∂x p ∂x q s

or = − ...(1)

∂x j ∂x i

∂x k i j ∂x i ∂x j p q

∂ ∂x s ∂x s k ∂x p ∂x q s

= − i j

∂x j ∂x i

∂x k i j ∂x ∂ x p q

∂x p ∂x q s ∂ ∂x s ∂x s k

− j = −

∂x ∂x p q

i

∂x j ∂x i ∂x k

i j

=

∂x

∂

j

(x )− x

s

,i

s

,k

k ∂x s

since k = x, k at P0

∂

s

i j x

k

= ( x ,is ), j since i j = 0 at P0

Thus,

∂x p ∂x q s

x,sij = − ...(2)

∂x i ∂x j p q

Necessary Condition

Let x s be a geodesic coordinate system with the pole at P0 so that

s

= 0 at P0

p q

Hence from (2), we have

x,sij = 0 at P0

Sufficient Condition

Conversely suppose that x,sij = 0 at P0 .

Curvature of Curve, Geodesic 177

s ∂x p ∂x q

i j = 0

p q ∂x ∂x

s ∂x p ∂x q

⇒ = 0 at P0 , as ≠ 0 and ≠ 0 at P0

p q ∂x i ∂x j

So, x s is a geodesic coordinate system with the pole at P 0.

A particular type of geodesic coordinates introduced by Riemann and known as Riemannian coordinates.

Let C be any geodesic through a given point P0 , s the length of the curve measured from P0 and ξ i the

quantities defined by

dxi

ξ = ds

i

...(1)

o

the subscript zero indicating as usual that the function is to be evaluated at P0 . The quantities ξi

represents that only one geodesic will pass through P 0 in the direction of ξi in V n. Let yi be the

coordinates of a point P on the geodesic C such that

yi = sξi ...(2)

where s is the arc length of the curve from P0 to P . The coordinates y i are called Riemannian

coordinates.

The differential equation of geodesic C in terms of coordinates y i relative to Vn is given by

d 2 y i i dy i dy j

+ =0 ...(3)

ds 2 j k ds ds

i

where is a christoffel symbol relative to the coordinates y i .

j k

The differential equation (3) will be satisfied by (2), we have,

i i j i

0+ ξ ξ = 0 since dy = ξ i

j k ds

i i j

or ξ ξ = 0 ...(4)

j k

using equation (2), equation (4) becomes

i yi y j yi

= 0 as = ξi

j k s s s

i i j

or y y = 0 ...(5)

j k

The equation (5) hold throughout the Riemannian Vn .

178 Tensors and Their Applications

i

= 0 at P0

j k

Hence the Riemannian coordinates are geodesic coordinate with the pole at P0 .

THEOREM 9.4 The necessary and sufficient condition that the coordinates y i be Riemannian coordinates

i i j

is that y y = 0 hold throughout the Riemannian Vn .

j k

i i j

Proof: If y are Riemannian coordinates then the condition

i y y = 0 (from equation 5) throughout

j k

the Riemannian Vn .

i i j d 2 y i dy i dy j

Conversely if j k y y = 0 hold then ds 2 + = 0 are saitsfied by y i = s ξi .

j k ds ds

Hence yi are Riemannian coordinates.

Let φ be a scalar invariant whose gradiant is not zero. Let the hypersurface φ = 0 be taken as coordinates

hypersurface x1 = 0 and the geodesics which cut this hypersurface orthogonally as the coordinate

lines of parameter x1 , this parameter measuring the length of arc along a geodesic from the hypersurface

x1 = 0 .

Since dx1 is the length of the vector µi is given by

i j

u 2 = g ij u u

i.e., (dx )

1 2 1 1

= g11dx dx

⇒ g 11 = 1 ...(1)

Now, if v i is the tangent vector to the hypersurface x1 = 0 then we have

2 3 n

v i = ( 0, dx , dx ,..., dx )

since the vectors ui and v i are orthogonal vectors.

Then,

g ij u i v j = 0

⇒ g 1 j u 1 v j = 0, [u i = 0, i = 2, 3, ..., n ]

⇒ g 1 j v j = 0, as u 1 ≠ 0.

⇒ g1 j = 0, for j = 2, 3, …, n. ...(2)

Again the coordinate curves of parameter x1 are geodesics. Then s = x1.

If ti is unit tangent vector to a geodesic at any point then

t1 = 1 and t i = 0, for i = 2, 3, …, n.

Curvature of Curve, Geodesic 179

Now,

d xi d x i

ti = = 1

ds dx

dx1 dxi

⇒ = 1 and = 0 for i ≠ 1

dx1 dx1

d 2xi d 2xi

and = = 0 for i = 1, 2,..., n

ds 2 dx12

Also, the differential equation of geodesic is

d 2 x i i dx j dx k

+ =0

ds 2 j k ds ds

using above results, we have

i dx 1 dx1

=0

11 ds ds

i

⇒ =0

1 1

⇒ g ij [11, j ] = 0

⇒ [11, j ] = 0 as g ≠ 0

ij

1 2 ∂g 1 j ∂ g 11

− = 0, since g 11 = i ⇒ ∂g 11 = 0

2 ∂ x i ∂x j

∂x j

So,

∂g 1 j

= 0 for j ≠ 1 ...(3)

∂x1

∂ g1 j

g 11 = 1, g1j = 0; ( j = 2, 3, …, n), = 0, ( j = 2, 3, …, n)

∂ x1

The line element is given by

i j

ds 2 = gij dx dx

ds 2 = g 11dx dx + g jk dx dx

1 1 j k

ds 2 = ( dx ) + g jk dx dx ; ( j = 2, 3, …, n, k = 2, 3, …, n)

1 2 j k

...(4)

The line element (4) is called geodesic form of the line element.

Note 1: We note that the coordinate curves with parameter x 1 are orthogonal to the coordinate curve

x i = c i (i = 1, 2, ..., n) at all points and hence to the hypersurfaces x 1 = c at each point.

Note 2: The existence of geodesic form of the line element proves that the hypersurfaces φ= x1 = constant form a

system of parallels i.e., the hypersurfaces φ = x 1 = constant are geodesically parallel hypersurfaces.

180 Tensors and Their Applications

THEOREM 9.5 The necessary and sufficient condition that the hypersurfaces φ = constant form a

system of parallel is that (∇φ)2 = 1.

Proof: Necessary Condition

Suppose that hypersurface φ = constant form a system of parallels then prove that (∇φ 2) = 1.

Let us take the hypersurface φ = 0 as the coordinate hypersurface x1 = 0. Let the geodesics

cutting this hypersurface orthogonally, be taken as coordinate lines of parameter x1. Then the parameters

x1 measures are length along these geodesics from the hypersurface x1 = 0. This implies the existence

of geodesic form of the line element namely

ds 2 = ( dx ) + g ij dx dx

1 2 i j

...(1)

where i, j = 2, 3,..., n.

From (1), we have

g 11 = 1, g 1i = 0 for i ≠ 1.

from these values, it follows that

g11 = 1, g1 i = 0, for i ≠ 1

Now,

∂φ ∂φ

(∇φ)2 = ∇φ ⋅ ∇φ = g

ij

∂x i dx j

∂ x1 ∂x1

= g

ij

= g ij δ1i δ1j

∂ x ∂x

i j

so (∇φ)2 = g11 = 1

(∇φ)2 = 1

Sufficient Condition

Suppose that (∇φ)2 = 1 then prove that the hypersurface φ = constant from a system of parallels.

Let us taken φ = x1 and orthogonal trajectories of the hypersurfaces φ = x1 constant as the

coordinate lines of parameter x1. Then the hypersurfaces

x1 = constant

xi = constant (i ≠ 1) are orthogonal to each other. The condition for this g1i = 0 for i ≠ 1 .

Now, given that (∇φ)2 = 1

∂φ ∂φ

⇒ g ij =1

∂x i ∂x j

∂x1 ∂x1

⇒ g ij =1

∂x i ∂x j

⇒ g ij δ1i δ1j = 1

⇒ g11 = 1

Curvature of Curve, Geodesic 181

Thus

g11 = 1 and g1 i = 0 for i ≠ 1.

Consequently

g 11 = 1, g = 0 , for i ≠ 1.

1i

Therefore, the line element

i j

ds 2 = g ij dx dx

is given by

ds 2 = (dx1 ) 2 + g ik dx i dx k ; (i, k = 2, 3, …, n)

which is geodesic form of the line element. It means that the hypersurfaces φ = x1 = constant

form a system of parallels.

Consider an Euclidean space S n for n-dimensions. Let y i be the Euclidean coordinates. The differential

equation of geodesics in Euclidean space is given by

d 2 y i i dy j dy k

+ =0 ...(1)

ds 2 j k ds ds

g ij = a ij = δij = 1, if i = j

0, if i ≠ j

∂g ij ∂a ij

= =0

∂x k

∂x k

k

j

i

Then equation (1) becomes

d 2 yi

=0

ds 2

Integrating it, we get

dyi

= ai, where ai is constant of integration.

ds

Again Integrating, we get

y i = ais + bi, where bi is constant of integration ...(2)

The equation (2) is of the form y = mx + c.

Hence equation (2) represents a straight line. Since equation (2) is a solution of equation (1) and

therefore the geodesic relative to S n are given by equation (2). Hence geodesic curves in Euclidean

space S n are straight lines.

182 Tensors and Their Applications

THEOREM 9.6 Prove that the distance l between two points P (y i) and Q (y' i ) in Sn is given by

∑ (y ′ − y )

n

i i 2

l=

i =1

Proof: We know that geodesics in S n are straight line. Then equation of straight line in S n may be

taken as

yi = ais + bi ...(1)

Let P ( y i ) and Q ( y' i) lie on equation (1). Then

yi = ais + bi, y′ i = a i s′ + bi

y′ i − y i = a i (s ′ − s ) ...(2)

Then equation (2) becomes

y′ i − y i = a i l

n n

l2 ∑

i =1

(ai ) 2 = ∑ ( y′

i =1

i

− y i )2

n

∑ (a )

i =1

i 2

=1

So,

n

l 2 = ∑ ( y′

i =1

i

− y i )2

l= ∑( y′ i

− y i )2

i =1

EXAMPLE 1

Prove that Pythagoras theorem holds in S n .

Solution

Consider a triangle ABC right angled at A i.e., ∠BAC = 90 o .

C (y i3)

A ( y1i ) B ( y2i )

Fig. 9.2

Curvature of Curve, Geodesic 183

AB ⋅ AC = 0

or a ij ( y i2 − y1i ) ( y i3 − y1i ) = 0

or ∑ (y

i =1

i

2 − y1i ) ( y 3i − y1i ) = 0 ...(1)

n

(AB) = 2

∑ (y

i =1

i

2 − y1i ) 2 ...(2)

(AC) = 2

∑ (y

i =1

i

3 − y1i )2 ...(3)

(BC) = 2

∑ (y

i =1

i

3 − y i2 )2 ...(4)

n

(BC) = 2

∑ [( y

i =1

i

3 − y1i ) + ( y1i − y i2 )]2

= ∑ [( y

i =1

i

3 − y1i )2 + ( y1i − y i2 ) 2 + 2( y3i − y1i ) ( y1i − y i2 ) ]

n n

= ∑

i =1

( y i3 − y1i )2 + ∑ (y

i =1

i

1 − y i2 ) 2 + 2 × 0, [from (1)]

n n

= ∑

i =1

( y i3 − y1i )2 + ∑( y

i =1

i

1 − y2i )2

Hence Pythagoras theorem holds in S n .

EXAMPLE 2

Prove that if θ is any solution of the differential equation (∇θ)2 = f (θ) then the hypersurfaces θ =

constant constitute a system of parallels.

Solution

Given that

(∇θ)2 = f (θ) ...(1)

184 Tensors and Their Applications

Suppose

dθ dθ

φ= ∫ f ( θ)

, Then, dφ =

f (θ)

dφ 1

or =

dθ f ( θ)

∂φ ∂φ ∂θ 1

Now, ∇φ = i = = ∇θ

∂x ∂θ ∂x i f ( θ)

2

1

(∇φ)2 = ∇θ

f ( θ)

1 1

= (∇ θ) 2 = f (θ); from (1)

f ( θ) f ( θ)

(∇φ)2 = 1

This proves that the hypersurfaces φ = constant form a system of parallels and therefore the

hypersurfaces θ = constant.

EXAMPLE 3

Show that it is always possible to choose a geodesic coordinates system for any Vn with an

arbitrary pole P 0.

Solution

Let P0 be an arbitrary pole in a Vn . Let us consider general coordinate system x i . suppose the

value of x i and P0 are denoted by x0i . Now consider a new coordinate system x j defined by the

equation.

1 h i

= a m ( x − x0 ) + a h ( x − x0 ) ( x − x0 )

j j m m j l m m

x ...(1)

2 l m

The coefficients amj being constants and as such that their determinant do not vanish.

Now we shall prove that this new system of coordinated x j defined by equation (1) is a geodesic

coordinate system with pole at P0 i.e., second covariant derivative of x j vanishes at P..

Differentiating equation (1) with respect to x m , we get

∂x j 1 h

= a m + ah 2 ( x − x0 )

j j i l

...(2)

∂x m 2 l m

∂x j

= a j at P ...(3)

∂x m m 0

0

Now, the Jacobian determinant

Curvature of Curve, Geodesic 185

∂x j

m = a j ≠ 0

∂x m

0

and therefore the transformation given by equation (1) is permissible in the neighbourhood of P0 .

Differentiating equation (2) with respect to x j , we get

∂2 x j

j m = a j h

∂x ∂x h ...(4)

0 l m 0

But we know that

∂2x h ∂x j

(x )

j

j

= l −

∂x h

∂x ∂x 0 l m 0

,lm 0 m

0

h h j

− a h , (from (3) and (4))

j

= ah

l m 0 lm

(x , )

j

lm 0 =0

Hence equation (1) is a geodesic coordinate system with pole at P0 .

EXAMPLE 4

If the coordinates xi of points on a geodesic are functions of arc lengths s and φ is any scalar

function of the x's show that

d pφ d xi d x j dxl

= φ, ij... l ⋅⋅ ⋅ ...(1)

ds p ds ds ds

Solution

Since the coordinates x i lie on a geodesic. Then

d 2 xi i dx j dx k

+ =0 ...(2)

ds 2 j k ds ds

Here the number of suffices i j...l is p.

We shall prove the theorem by mathematical induction method.

Since x′s are functions of s and φ is a scalar function of x′ s , we have

dφ ∂φ dxi dφ dx i

= or = φ, i ...(3)

ds ∂x i ds ds ds

d 2φ ∂φ, i dx j d 2xi

= + φ ,i

ds2 ∂ x j ds ds 2

∂φ ,i dx i dx j i d x j d x k

= ∂x j ds ds − φ ,i , from (2)

j k ds ds

186 Tensors and Their Applications

∂φ ,i dx i dx j m dx j dx k

= ∂x i ds ds − φ, m

j k ds ds

∂φ, i dx i dx j m dxi dx j

= − φ , m (adjusting the dummy index.)

∂ x j ds ds i j ds ds

∂φ, i m dx i dx j

= j − φ,m

i j ds ds

...(4)

∂x

Equations (3) and (4) imply that the equation (1) holds for p = 1 and p = 2.

Suppose that the equation (1) holds for p indices r1 , r2 ,...,r p so that

r

d pφ dx r1 dx r1 dx p

= φ, r1 , r2 , ...,

⋅ ⋅ ⋅ rp ...(5)

ds p ds ds ds

Differentiating the equation (5) with respect to s, we get

d p +1 φ ∂φ, r1 r2 ... r p dx r1 r r

dx p dx p +1 d 2 x r1 dx r2

r

dx p

= ⋅ ⋅ ⋅ + φ, ⋅ ⋅ ⋅ + ⋅ ⋅⋅

ds p +1

r r1 r2 ... rp

∂ x p +1 ds ds ds ds 2 ds ds

r

dx r1 d 2x p

+ ⋅ ⋅ ⋅φ, r1 r2 ... rp ⋅ ⋅⋅ ...(6)

ds ds 2

d 2 x r1

substituting value of etc. from (2) in (6) and adjusting dummy indices, we have

ds 2

∂ φ, r r ... r m m

d p +1 φ − φ ⋅ ⋅ ⋅ φ

1 2 p

p +1 = ∂x p +1

r , m r ... r ,r r ... m

2 p 1 2

r r

1 p +1 r r

p p +1

ds

r

dx r1 dx r2 dx p +1

⋅⋅⋅

ds ds ds

r

d x r1 d x r2 d x p +1

= φ ,r1 r2 .. r p rp +1

⋅⋅ ⋅

ds ds ds

This shows that the equation (1) holds for next values of p. But equation (1) holds for p = 1, 2,

... Hence equation (1) holds for all values of p.

EXERCISES

1. Prove that at the pole of a geodesic coordinate system, the components of first covariant derivatives

are ordinary derivatives.

2. If x i are geodesic coordinates in the neighbourhood of a point if they are subjected to the

transformation

1 i

xi = x +

i

c x j xk xl

6 jkl

Curvature of Curve, Geodesic 187

where C′s are constants then show that xi are geodesic coordinates in the neighbourhood of O.

3. Show that the principal normal vector vanishes identically when the given curve is geodesic.

4. Show that the coordinate system x i defined by

1 i j k

x i = x + 2 j k x x

i

is geodesic coordinate system with the pole at the origin.

5. Obtain the equations of geodesics for the metric

ds2 = e−2 kt ( dx2 + dy2 + dz2 ) + dt 2

6. Obtain the differential equations of geodesics for the metric

1

ds2 = f ( x) dx + dy + dz + f ( x) dt

2 2 2 2

d2x 1 d dx

2

1 d dt d2y d2z d 2t d (log f ) dx dt

Ans : 2 − (log f ) + (log f ) = 0; = 0; = 0; − = 0

ds 2 dx ds 2 f 2 dx ds ds2 ds2 ds2 dx ds ds

7. Find the differential equations for the geodesics in a cylindrical and spherical coordinates.

8. Find the rate of divergence of a given curve C from the geodesic which touches it at a given point.

CHAPTER – 10

PARALLELISM OF VECTORS

CONCEPT)

Consider a vector field whose direction at any point is that of the Unit Vector ti. In ordinary space, the

field is said to be parallel if the derivative of ti vanishes for all directions ui (say) and at every point of

the field i.e.,

∂t i j

u =0

∂x j

Similarly in a Riemannian V n the field is said to be parallel if the derived vector of ti vanishes at

each point for every direction ui at each point of V n. i.e.,

t,i j = u j = 0

It can be shown that it is not possible for an arbitrary V n. Consequently we define parallelism of

vectors with respect to a given curve C in a V n.

A vector ui of constant magnitude is parallel with respect to V n along the curve C if its derived

vector in the direction of the curve is zero at all points of C i.e.,

dx j i

u =0 … (1)

ds , j

where s is arc-length of curve C.

The equation (1) can be written in expansion form as

∂ui

m i dx

j

j + u m j =0

∂ x ds

∂ui d x j m i dx

j

+ u =0

∂ x j ds m j ds

du i i dx j

+ um

j ds = 0 … (2)

ds m

Parallelism of Vectors 189

This concept of parallelism is due to Levi-Civita. The vector ui is satisfying the equation (1) is

said to a parallel displacement along the curve

Now, multiplying equation (1) by gil, we get

dx j

g il u,i j

ds = 0

dx j

( gil u,i j ) =0

ds

dx j

( g il u i ) , j =0

ds

dx j

ul , j

ds = 0

dx j

or u i, j =0

ds

∂ui m dx j

j − u m =0

∂x i j ds

dui m dx j

− um =0 … (3)

ds i j ds

The equation (2) and (3) can be also written as

m i

dui = − u m j d x

j

… (4)

m j

and dui = um i j d x … (5)

The equation (4) and (5) give the increment in the components ui and ui respectively due to

displacement dxj along C.

THEOREM 10.1 If two vectors of constant magnitudes undergo parallel displacements along a given

curve then they are inclined at a constant angle.

Proof: Let the vectors ui and v i be of constant magnitudes and undergo parallel displacement along a

curve C, we have (from equation (1), Pg. 188.)

dx j

u,i j = 0

ds

j

d x … (1)

vi, j = 0

ds

at each point of C.

Multiplying (1) by gil, we get

dx j

( g il u,i j ) =0

ds

190 Tensors and Their Applications

dx j

ul , j =0

ds

dx j

or u i, j =0 … (2)

ds

Similarly,

dx j

vi , j =0 … (3)

ds

Let φ be the angle between ui and v i then

ui.v i = uv cos θ

Differentiating it with respect to arc length s, we get

d d (u i vi )

(uv cos θ) =

ds ds

i dx j

= (u v i ), j

ds

dθ i dx

j

dx j

− uv sin θ = u, j vi + u i vi , j … (4)

ds ds ds

Using equation (1) and (3), then equation (4) becomes

dθ

− uv sin θ =0

ds

dθ

⇒ sin θ = 0, as u ≠ 0, v ≠ 0

ds

dθ

⇒ Either sinθ = 0 or =0

ds

⇒ Either θ = 0 or θ = constant.

⇒ θ is constant. Since 0 is also a constant.

THEOREM 10.2 A geodesic is an auto-parallel curve.

Proof: The differential equation of the geodesic is given by (See Pg. 174, eqn. 6)

d 2 x m m dx j dx k

+ =0

ds 2 j k ds ds

d d x m m d x j d x k

+

ds ds j k ds ds = 0

∂ dx m d x j m d x j d xk

∂x j ds ds + j k ds ds = 0

Parallelism of Vectors 191

∂ dx m m dx k dx j

j + =0

ds

∂x j k ds ds

d xm dx j dx j

=0

ds ds = 0 or t,mj

, j ds

dx m

This shows that the unit tangent vector suffer a parallel displacement along a geodesic curve.This

ds

confirms that geodesic is an auto-parallel curve. Proved.

Two vectors at a point are said to be parallel or to have the same direction if their corresponding

components are proportional. Consequently the vector v i will be parallel to ui at each point of curve C

provided

v i = φui … (1)

where φ is a function of arc length s.

If ui is parallel with respect to Riemannian V n along the curve C. Then,

dx j

u ,ij =0 … (2)

ds

The equation (1) shows that v i is of variable constant and parallel with respect to Riemannian V n

so that

dx j dxj

v ,ij i

= (φ u ), j

ds ds

dx j

= (φ, j u + φu, j )

i i

ds

dx j i dxj

= φ, j u + φ u,i j

ds ds

dx j i dx j

= φ, j u Since u,i j =0

ds ds

∂φ d x j i

= u

∂x j ds

dx j dφ i

v,i j = u

ds ds

d φ vi

= ds from (1)

φ

192 Tensors and Their Applications

i d (log φ)

= v

ds

dx j d (log φ )

v,i j = v i f (s) where f (s) = … (3)

ds ds

Hence a vector v i of variable magnitude will be parallel with respect to V n if equation (3) is satisfied.

Conversely suppose that a vector v i of variable magnitude such that

dx j

v,ij = = v i f (s )

ds

to show that v i is parallel, with respect to V n .

Take

ui = v i ψ (s) … (5)

Then

dx j dx j

u ,ij = (v ψ), j

i

ds ds

j

i dx dx j i

= v, j ψ + ψ, j v

ds ds

∂ψ dx j i

= v i

f ( s ) ψ + v

∂x j ds

dx j i dψ

u,i j = v ψf (s ) +

ds

… (6)

ds

dψ

Select ψ such that ψf ( s ) + = 0.

ds

Then equation (6) becomes

dx j

u,i j =0

ds

This equation shows that the vector u i is of constant magnitude and suffers a parallel displacement

along curve C. The equation (5) shows that v i is parallel along C.

Hence necessary and sufficient condition that a vector v i of variable magnitude suffers a parallel

displacement along a curve C is that

dx j

v,i j = v i f (s ).

ds

EXAMPLE 1

Show that the vector v i of variable magnitude suffers a parallel displacement along a curve C if

and only if

dx k

(v i v i,k − v i v l,k ) = 0, i = 1, 2, ..., n.

ds

Solution

From equation (4), we have

dx j

v ,ij = v i f (s ) … (1)

ds

Parallelism of Vectors 193

Multiplying by v l, we get

dx j

v l v,i j

= v l v i f (s )

ds

Interchange the indices l and i, we get

dx j

v i v,l j

= v i v l f (s) … (2)

ds

Subtract (1) and (2), we get

dx k

(v l u,ik − v i v l,k ) = 0 by interchanging dummy indices j and k.

ds

Let V n be Riemannian space of n dimensions referred to coordinates xi and having the metric

ds2 = gij dxi dxj. Let V m be Riemannian space of m dimensions referred to coordinates y α and having

the metric ds2 = axβ dyα dyβ, where m > n. Let Greek letters α , β, γ take the values 1, 2, ..., m and Latin

indices i, j, k ... take the values 1, 2, … n.

If the n independent variables xi are such that the coordinates ( y α ) of points in V m are expressed

as a function of xi then V n is immersed in V m i.e. V n is a subspace of V m. Also V m is called enveloping

space of V n.

Since the length ds of the element of arc connecting the two points is the same with respect to V n

or V m. it follows that

gij dxi dxj = a αβdy α dy β

∂y α ∂y β i j

gij dxi dxj = a αβ dx dx

∂x i ∂x j

∂y α ∂y β

⇒ a

gij = αβ i … (1)

∂x ∂x j

As dxi and dxj are arbitrary.

This gives relation between gij and aαβ .

THEOREM 10.3 To show that the angle between any two vectors is the same whether it is calculated

with respect to Vm or Vn.

Proof: Consider two vectors dx i and δx j defined at any point of V n and suppose that the same vectors

in V m are represented by dy α and δy α respectively. If θ is the angle between dxi and δx i then

g ij dx i δx j

cos θ = … (1)

g ij dx i dx j g ij δx i δx j

If φ is the angle between the vectors dyα and δyα then

aαβ dy αδy β

cos φ =

aαβ dy α dy β aαβ δy αδy β

194 Tensors and Their Applications

∂ y α i ∂ yβ

aαβ dx dx j

∂x i ∂x j

=

∂ yα i ∂ yβ j ∂ yα i ∂ yβ j

aαβ d x d x a αβ δx δx .

∂x i ∂x j ∂ xi ∂x j

∂y α ∂y β i j

aαβ dx dx

∂x i ∂x j

=

∂y α ∂y β i j ∂y α ∂y β i j

aαβ dx dx a αβ δx δx .

∂xi ∂x j ∂x i ∂x j

gij dx i dx j

cos φ = … (2)

g ij dx i dx j g ij δx i δx j

∂y α ∂y β

Since g ij = aαβ (from equation (1), art. 10.3)

∂x i ∂x j

from (1) & (2)

cos θ = cos φ ⇒ θ = φ. Proved.

THEOREM 10.4 If Uα and ui denote the components of the same vector in Riemannian Vm and Vn

respectively then to show that

α

i ∂y

α = u

U ∂x i

Proof: Let the given vector be unit vector at any point P of V n. Let the component of the same vector

x's and y's be a i and A α respectively. Let C be curve passing through s in the direction of the given

vector then

∂y α ∂y α dx i ∂y α i

A =α = i = i a … (1)

∂s ∂x ds ∂x

But the components of a vector of magnitude a are a times the corresponding components of the

Unit vector in the same direction. Then

α

U α = aA , u = aa

i i … (2)

Multiplying (1) by a, we get

∂y α

aA α = (aa i )

∂x i

∂y α i

Or α = u , using (2) Proved.

U ∂x i

THEOREM 10.5 To show that there are m-n linearly vector fields normal to a surface V n immersed in

a Riemannian V m.

Proof: Since V n is immersed in V m, the coordinates y α of points in V m are expressible as functions of

coordinates xi in V n.

Parallelism of Vectors 195

Now,

dy α ∂y α dx i

= ...(1)

ds ∂x i ds

i

for the curve x = s, we have

dy α ∂y α

= ...(2)

ds ∂x i

dy α ∂ yα

Since is a vector tangential to the curve in V m. Then from equation (2) it follows that

ds ∂ xi

i α

in V m is tangential to the coordinate curve of transmeter x in V n. Let the Unit vectors N in V m be

normal to each of the above vector fields of V n then

∂y α β

a αβ N =0

∂xi

∂y α

⇒ (aαβ N β ) i = 0 ⇒ a αβ N β y ,αi = 0 ...(3)

∂x

∂y α

⇒ N α i = 0, i = 1, 2, …, n & α = 1, 2, …, m

∂x

∂y α

The equation (3) are n equations in m unknowns N α (m > n). The coefficient matrix i is

∂x

of order m × n and the rank of this matrix is n.

It means that there will be only m – n linearly independent solution of N α . This shows that there

are m – n linearly independent normals to V n to V m.

EXAMPLE 2

Show that

∂y α ∂y β ∂y γ ∂ 2 y α ∂x β

[ij, k] = [αβ, γ ] + a αβ

∂x i ∂x j ∂x k ∂xi ∂x j ∂x k

where [αβ, γ] and [i j, k ] are the Christoffel’s symbols of first kind relative to metrics a αβ dy a dy β and

g ij dx i dx j .

Solution

∂y α ∂y β

g ij = a αβ

∂x i ∂x j

Differentiating it with respect to xk, we get

∂g ij ∂aαβ ∂y α ∂y β ∂y γ ∂ 2 y α ∂y γ ∂y α ∂ 2 y β

= + a αβ

+ a αβ … (1)

∂x k ∂x γ ∂x i ∂x j ∂x k ∂x i ∂x k ∂x j ∂x i ∂x j ∂x k

196 Tensors and Their Applications

Similarly

∂g jk ∂aβγ ∂y α ∂y β ∂y γ ∂ 2 y α ∂y β ∂y α ∂ 2 yβ

= + a αβ + a αβ … (2)

∂x i ∂y α ∂x i ∂x j ∂x k ∂x j ∂x i ∂x k ∂x j ∂x k ∂x i

and

∂g ki ∂a γα ∂y α ∂y β ∂y γ ∂ 2 y α ∂y β ∂y α ∂ 2 y β

= + a αβ

+ a αβ … (3)

∂x j ∂x β ∂x i ∂x j ∂x k ∂x k ∂x j ∂x i ∂x k ∂x i ∂x j

But we know that

1 ∂g jk ∂g ki ∂g ij

[ij, k ] = + − … (4)

2 ∂x i ∂x j ∂x k

1 ∂gβγ ∂g γα ∂g αβ

[αβ, γ] = 2 α + −

∂y γ

and, … (5)

∂y ∂y β

Substituting the value of (1), (2), (3) in equation (4) and using (5) we get,

∂y α ∂y β ∂ y γ ∂ 2 yα ∂ yβ

[ij, k ] = [αβ, γ ] + a αβ .

∂x i ∂x j ∂ x k ∂ xi ∂ x j ∂ x k

THEOREM 10.6 Let T α and t i be the components of the same vector t relative to Vm and Vn

respectively. Let the vector t be defined along a curve C in V n. If p i and qα are derived vectors of t

along C relative to Vn and Vm respectively. Then

∂y α

qα = pi

∂xi

Proof: Since from equation (1), theorem, (10.4), Pg. 194 we have

∂y α

i

T α = t … (1)

∂x i

dT α dt i ∂y α 2 α

i ∂ y dx j

= + t … (2)

ds ds ∂x i ∂x i ∂x j ds

Now,

i dx j

p i = t, j … (3)

ds

β

α dy

qα = T,β

ds

∂T α β

γ α dy

= β + T

∂x γβ ds

∂T α dy β γ α dy

β

= + T

∂xβ ds γβ ds

dT α γ α dy

β

q α

= ds + T … (4)

γβ ds

Parallelism of Vectors 197

dT α

Putting the value of and T γ from (2) and (1) in equation (4), we get

ds

dt i ∂y α i ∂y dy α

2 α γ β

i ∂ y dx j

qα = + t + t

ds ∂x i ∂x i ∂x j ds dx i ds γβ

∂t i ∂y α dx j i ∂y dy α

2 α γ β

i ∂ y dx j

= + t + t

∂x j ∂x i ds ∂x i ∂x j ds ∂x i ds γβ

∂t i ∂y α dx j i ∂y ∂y dx α

2 α γ β

i ∂ y dx j j

= + t + t

∂x j ∂x i ds ∂x i ∂x j ds ∂x i ∂ x. j ds γβ

∂t i ∂ y α d x j i d x j ∂ 2 y α ∂y β ∂y γ α

α + t + +

ds ∂ x i ∂x j ∂x j ∂x i γβ

q = … (5)

∂x j ∂x i ds

But we know that

∂y α ∂y β ∂y γ ∂ 2 y α ∂y γ

[ij, k ] = [αβ, γ] + a αγ

∂x i ∂x j ∂x k ∂x i ∂x j ∂x k

∂y β ∂y γ ∂y δ ∂ 2 y α ∂y δ

[ ji, k ] = [βγ, δ] + a αδ

∂x j ∂x i ∂x k ∂x i ∂x j ∂x k

α ∂y β ∂y γ ∂y δ ∂ 2 y α ∂y δ

= βγ αδ j

a + a αδ

∂x ∂x i ∂x k ∂x i ∂x j ∂x k

∂y δ α ∂y β ∂y γ 2 α

a j i + ∂ iy j

= αδ ∂x k βγ ∂x ∂x ∂x ∂x … (6)

∂y δ

Multiplying (5) by a αδ , we get, using (6),

∂x k

∂y δ α ∂t i dx j ∂y α ∂y δ i dx

j

aαδ q = a αδ + t [ij, k ]

∂x k ∂x j ds ∂x i ∂x k ds

∂y δ dx j ∂t i p ∂y α ∂y β

Or qδ k = g + t i g pk ,

j ik since g ij = aαβ

∂x ds ∂x i j ∂x i ∂x j

∂t i

j gik + t a g ik

dx j i

= ds ∂x

a j

dx j i

= g ik t,

ds j

198 Tensors and Their Applications

i dx j

= ( gik t ), j

ds

dx j

= tk, j

ds

∂y α dx j

qα = t

∂x k

k, j

ds

α

∂y

or q α k = pk …(7)

∂x

Proved.

Properties of Vm

(i) If a curve C lies in a subspace V n of V m and a vector field in V n is parallel along the curve C

with regard to V m, then to show that it is also a parallel with regard to V n.

Proof: If a vector field t is a parallel along C with respect to V m. then its derived vector qα

vanishes i.e.,

qα = 0, α = 1, 2,... m.

Hence equation (7) becomes pk = 0, k = 1, 2, ..., n. This shows that the vector field t is also

parallel along C with respect to V n.

(ii) To show that if a curve C is a geodesic in a V n it is a geodesic in any subspace V n of V m.

α ∂y α

Proff : Science, q = pi

∂x i

Let t be unit tangent vector to the curve C them T α is unit tangent vector to C relative to V m

and ti is unit tangent vector to C relative to V n.

Now, pi = 0 ⇒ curve C is a geodesic in V n

α

q =0 ⇒ curve C is a geodesic in V m

Also, qα = 0, ∀α ⇒ pi = 0, ∀i

This shows that if a curve C in V n is a geodesic relative to V m then the same curve is also a geodesic

relative to V n.

(iii) A necessary and sufficient condition that a vector of constant magnitude be parallel with

respect to V n along C in that subspace, is that its derived vector relative to V m for the

direction of the curve be normal to V n.

Proof: Let t be a vector of constant magnitude. This vector t is parallel along C relative to Vn iff pi =

0, ∀i

∂y α

iff q α =0

∂x i

∂ yα

But qα = 0 implies that qα is normal to V n.

∂x i

∂ yα

For = y,αi lying in V m is tangential to a coordinate curve of parameter xi in V n.

∂ xi

Parallelism of Vectors 199

These statements prove that a necessary and sufficient condition that a vector of constant

magnitude be parallel along C relative to V n is that its derived vector i.e., qα along C relative

to V m be normal to V n.

(iv) A necessary and sufficient condition that a curve be a geodesic in V n is that its principal

subnormal relative to V m the enveloping space be normal to V n at all points of the curve.

Proof: In particular let the vector t be unit tangent vector to the curve C. In this case qα is

called principal normal the curve C. Also pi = 0, ∀i implies that the curve C is a geodesic

relative to V n.

Using result (iii), we get at once the result (iv).

(v) To prove that the tendency of a vector is the same whether it is calculated with respect to

V m.

Proof: Since, we have

∂y α

q α i = pi

∂x

∂y α dx i dxi

qα = p

∂x i ds

i

ds

dy α dxi

qα = pi

ds ds

β α

dy dy dx j dx i

Tα, β = ti, j

ds ds ds ds

i.e., tendency of Tα along C = tendency of ti along C.

i.e., tendency of t along C relative to V m = tendency of t along C relative to V n.

STATEMENT

With a given Riemannian metric (or fundamental tensor) of a Riemannian manifold there is associated

a symmetric affine connection with the property that parallel displacement (or transport) preserves

scalar product.

Proof: Let C be a curve in V n. Let pi and qi be two unit vectors defined along C. Suppose that the unit

vectors pi and qi suffer parallel displacement along the curve C in V n, then we have

dx j

p,i j =0 … (1)

ds

dx j

and q ,ij =0 … (2)

ds

Let gij be the given fundamental tensor of a Riemannian manifold. Hence, the scalar product of

vectors p i and q i is g ij p i q j.

Now,

dx k

( gij p i q j ),k =0

ds

200 Tensors and Their Applications

dx k j k

i j dx

dx k i j

g ij p,ik q + g p +

,k ds ij ,k ds P q = 0

q g

ds

ij … (3)

Using equation (1) and (2), the equation (3) becomes,

k

gij , k dx p i q j = 0

ds

gij, k = 0

dx k

pi qi

(Since and are unit vectors and ≠ 0)

ds

∂g ij m m

− g mj − gim = 0

∂x k

i k j k

∂g ij

− [ik , j ] − [ jk , i ] = 0

∂x k

∂g ij

= [ik , j] + [ jk , i ] … (4)

∂x k

Now, using equation (4), we have

∂g jk ∂g ki ∂g ij

+ − = [ ji, k ] + [ki, j ] + [kj, i ] + [ij, k ] – ([ik , i] + [ jk , i ])

∂x i ∂x j ∂x k

since [ij, k ] = [ ji, k ].

So,

∂g jk ∂g ki ∂g ij

+ − = 2[ij, k ]

∂x i ∂x j ∂x k

1 ∂g jk ∂g ki ∂g ij

[ij,k] = + − … (5)

2 ∂x i ∂x j ∂x k

k

But we know that i j = g [ij, l ]

lk

from (5), we have

k 1 lk ∂g jl ∂g li ∂gij

= g i + j − l Proved.

i j 2 ∂x ∂x ∂x

EXAMPLE 3

If t i and T α are contravariant components in x's and y's respectively, of n vector field in V n

immersed in V m. Show that

T, αj = y,αij t i + y ,αi t,i j

Solution

Since we know that

i ∂y α

T α = t

∂x i

Parallelism of Vectors 201

i α

T α = t y, j .

Taking covariant differentiation of both sides, we get

T, αj = (t i y α,i ), j

EXAMPLE 4

dx i dx j

Show that g ij remains constant along a geodesic.

ds ds

Solution

dx i

Let t =

i

. Then

ds

dx i dx j

gij = g ij t i t j = t 2

ds ds

Since we know that geodesics are autoparallel curves. Then

dx j

t,i j =0

ds

or t,i j t j = 0 … (1)

Now,

dt 2 d d

= ( g ij t i t j ) = (ti t i )

ds ds ds

dx j

i

= ( ti t ) , j = (t i t i ), j t j

ds

dt 2

= (ti , j t i )t i + (t ,i j t i )ti = 0, from (1)

ds

Integrating it we get

t2 = constant.

dx i dx j

So, g ij remains constant along a geodesic.

ds ds

EXAMPLE 5

If ti are the contravariant components of the unit tangent vector to a congruence of geodesics.

Show that

t i (ti , j + t j ,i ) = 0

and also show that | ti , j + t j ,i |= 0.

202 Tensors and Their Applications

Solution

Let t i denote unit tangent vector to a congruence of geodesic so that

t,i j t j = 0 … (1)

Since geodesics are auto-parallel curves. Then to prove that

(i) t i (ti , j + t j ,i ) = 0

(ii) | ti , j + t j ,i |= 0

Since

t i ti = t 2 = 1

t i ti = 1

Taking covariant derivative of both sides, we get

(t i t i ), j = 0

or t,i j t i + t i ti , j = 0

Since t is a free index. Then we have

t,i j t i + ti , j t i = 0

2t i, j t i = 0

⇒ ti , j t i = 0 … (2)

from (1)

g ik t,i j t j = 0

or tk, j t j = 0

⇒ t k ,i t i = 0 ⇒ t j ,i t i = 0 ⇒ t i t j, i = 0

Thus t i t j, i = 0 … (3)

Adding (2) and (3), we get

t i (ti , j + t j ,i ) = 0

Also, since ti ≠ 0, ∀i.

Taking determinants of both sides we get

| t i (ti , j + t j , i ) | = 0

⇒ | ti , j + t j ,i | = 0

Solved.

EXAMPLE 6

When the coordinates of a V 2 are chosen so that the fundamental forms is

(dx&' ) + 2 g12dx1 dx 2 + (dx 2 )2 , prove that the tangent to either family of coordinate curves suffers a

parallel displacement along a curve of the other family.

Parallelism of Vectors 203

Solution

The metric is given by

i j

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

Comparing it with

ds 2 = (dx ) + 2 g12 dx dx + (dx )

1 2 1 2 2 2

(1)

We have,

g11 = 1, g22 = 1.

In this case, we have coordinate curves of parameters x1 and x2 respectively. The coordinate x1

curve is defined by

xi = ci, ∀i, except i = 1. (2)

2

and the coordinate x curve is defined by

x i = d i, ∀i, except i = 2 (3)

i i

where c and d are constants.

Let p i and q i be the components of tangents vectors to the curves (2) and (3) respectively. Then

we have

pi = dxi = 0, ∀i, except i = 1

and qi = dxi = 0, ∀i, except i = 2

So,

pi = (dx i, 0) and qi = (0,dx i)

Let t be the unit tangent vector to the curve (2).

Hence

dx i pi

= = (1,0)

i

= t where p = dxi

ds p

So, we have

dx j

p,i j =0

ds

Hence the tangent to the family of coordinates curves (2) suffers a parallel displacement along a

curve of the family of curves (3).

EXERCISES

1. Explain Levi-Civita's concept of parallelism of vectors and prove that any vector which undergoes a

parallel displacement along a geodesic is inclined at a constant angle to the curve.

2. Show that the geodesics is a Riemannian space are given by

d 2 x m m dx i dx k

+ =0

ds 2 i k ds ds

Hence prove that geodesics are auto-parallel curves.

204 Tensors and Their Applications

(i) It is an auto–parallel curve.

(ii) It is a line whose first curvature relative to Vn identically zero.

(iii) It is the path extremum length between two points on it.

4. If u and v are orthogonal vector fields in a V n , prove that the projection on υ of the desired vector of

u in its own direction is equal to minus the tendency of v in the direction of u.

5. If the derived vector of a vector u i is zero then to show that vector u i has a constant magnitude along

curve.

6. Prove that any vector which undergoes a parallel displacement along a geodesic is inclined at a

constant angle to the curve.

7. Prove that there are m – n linearly independent vector fields normal to a surface Vn immersed in a

Riemannian Vm and they may be chosen in a multiply infinite number of ways. But there is only one

vector field normal to the hyperface.

8. Show that the principal normal vector vanishes identically when the given curve in geodesic.

9. Show that if a curve is a geodesic of a space it is a geodesic of any subspace in which it lies.

10. Define parallelism in a subspace of Riemannian manifold. If a curve C lies in a subspace Vn of Vm and

a vector field in Vn is parallel along C with respect to V m . Then show that it is also parallel with

respect to V n .

CHAPTER – 11

CONGRUENCE

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

The invariants γlhk are Ricci's Coefficients of Rotation.

Since i being a dummy index, has freedom of movement. Then equation (1) may be written as

The indices l, h and k are not tensor indices. But these indices in γlhk are arranged in proper way,

the first index l indicates the congruence whose unit tangent is considered, the second h indicates the

direction of projection and the third k is used for differentiation.

THEOREM 11.1 To prove thast the Riccie's coefficients of rotation are skew-symmetric in the first

two indices i.e.,

γlhk = –γlhk

then

eh|i eli| = 0

convariant differentiation with respect to xj, we get

eh|i eli| ,j = 0

206 Tensors and Their Applications

γhlk + γlhk = 0

or γhlk = –γlhk …(1)

Note: Put l = h in equation (1), we get

γllk = –γllk

or 2γllk = 0

or γllk = 0

h

Proof: since we know that

Multiplying by eh|m and summing for h.

h h

j

= el |i , j e k| ∑ (e e )

h

i

h| h| m

m

h

= (e l |i , j δ m ) e kj |

i

∑γh

lhk eh |m = el |m , j ekj |

Replacing m by i, we get

∑ γ lhkei|m = el |i , j e kj|

h

Let Cm be a definite curve of the congruence whose unit tangent is emi | and P 0 a fidxed point on it. Let

u1 be a unit vector which coincides with the vector eli| at P 0 and undergoes a parallel displayment along

the curve Cm.

Thus

u i = eli| at P 0 … (1)

Ricci's Coefficients of Rotation and Congruence 207

cos θ = ui eh|i

Differentiating it with respect to arc length sm along Cm, we get

dθ

– sin θ i j

ds m = (u e h|i ), j em |

= (u i e h|i , j + u i, j eh |i ) emj |

dθ

– sin θ u i e h |i , j e mj | + u ,i j e mj |

ds m = …(3)

π

at the point P 0,θ = , we have

2

dθ

– u i eh|i , j emj |

ds m =

= u i eh|i , j emj |

dθ

ds m = γhlm

–

dθ

⇒ ds m = – γhlm … (4)

dθ

In Eucliden space of three dimensions, ds is the arc-rate of rotation of the vector e ih| about the

m

curve Cm. Hence the quantities γhlmare called coefficients of rotation of the ennuple. Since it was

discovered by Ricci and hence it is called Ricci's coefficient of rotation.

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

from theorem (11.2), we have

l

208 Tensors and Their Applications

The magnitude of p ih| is called curvature of the curve of congruence e h| and denoted by K h | .

Now,

K h2| = g ij phi | p hj|

= g ij ∑γ i

hlh e l |

∑γ i

hmh em |

l m

l m

= ∑∑ δ

l m

m γ hlm

l

γ hmh , Since g ei e i = l

ij l| m| δm

= ∑ γ hmhγ hmh

m

K h2| = ∑ (γ

m

hmh )

2

If all the curves of a congruence are geodesics then the congruence is called a geodesic congruence.

THEOREM 11.3 A necessary and sufficient condition that congruence C of an orthogonal ennuple

be a geodesic congruence is that the tendencies of all the other congruences of the ennuple in the

direction of C vanish indentically.

Or

To obtain necessary and sufficient conditions that a congruence be a geodesic congruence.

Proof: From equation (1), Pg. 207, we have

phi | = ∑ γ hlh eil|

l

Since eli| ≠ 0 , ∀h and hence phi | = 0 iff γhlh = 0, ∀h . But pli| = 0 iff the congruence C is geodesic

congruence.

Hence C is a geodesic congruence iff γhlh = 0, ∀h .

But

γhlh = – γlhh

So, C is a geodesic congruence iff – γlhh = 0, –∀h .

or C is a geodesic congruence iff γlhh = 0.

Hence γlhh = 0 are the necessary and sufficient conditions that congruence with unit tangents ehi |

be geodesic congruence. Again γlhh is the tendency of the vector eli| in the direction vector e ih| .

Ricci's Coefficients of Rotation and Congruence 209

Thus a congruence C of an orthogonal ennuple is a geodesic congruence iff the tendency of all

other congruences in the direction of C vanish identically.

A normal congruence is one which intersects orthogonally a family of hypersurfaces.

THEOREM 11.4 Necessary and sufficient conditions that the congruence eh | of an orthogonal ennuple

be normal is that

γnqp = γnpq

Proof: Consider a congruence C of curves in a V n. Let φ( x1 , x 2 , …, x n ) = constant be a family of

hypersurfaces. To determines a normal congruence whose tangent vector is grad φ or ∇ φ.

Let ti be the convariant components of unit tangent vector to C. The congruence C is a normal

congruence to a family of hypersurface φ( x1 , x 2 , …, x n ) = constant if

φ,1 φ, 2 φ,

= = … n = y (say) … (1)

t1 t2 tn

In order that (n – 1) differential equations given by equation (1) admit a solution which is not

constant, these must constitute a complete system.

From (1)

φ,i

= y or φ, i = yti

ti

∂φ

⇒ yti =

∂x i

Differentiating it with respect to x j , we get

∂y ∂t i ∂ 2φ

ti + y = … (2)

∂x j ∂x j

∂x i ∂x j

Interchanging indices i and j, we get

∂y ∂t j ∂ 2φ

t + y = … (3)

∂x i

j

∂x i ∂x j ∂x i

∂y ∂t ∂y ∂t

j ti + y i j – i t j + y i j = 0

dx ∂x ∂x dx

∂t ∂t j ∂y ∂y

y ij – i + t i j – t j i = 0

∂x ∂x ∂x ∂x

210 Tensors and Their Applications

Multiplying by tk,

∂ti ∂t j ∂y ∂y

ytk j – i + ti tk – ik ti i = 0 …(4)

dx ∂x ∂x j

dx

By cyclic permutation of i, j, k in (4), we get

∂t j ∂t ∂y ∂y

yti k – kj + t j t i k – i k ti =0 … (5)

∂x ∂x ∂x ∂x j

and

∂t ∂t ∂y ∂y

yti ki – ik + t k t j i – ti t j k = 0 … (6)

∂x ∂x ∂x ∂x

On adding (4) , (5) and (6) we get

∂t ∂t j ∂t j ∂t

y t k ij – i + ti k – kj + t j ∂tk – ∂ti = 0

∂x ∂x ∂x ∂x ∂x i ∂x k

or t k (ti , j – t j ,i ) + ti (t j , k – t k , j ) + t j (t k , i – ti ,k ) = 0 … (7)

as y ≠ 0 , where i, j, k = 1, 2, …, n.

These are the necessary and sufficient conditions that the given congruence be a normal congruence.

Now suppose that the congruence is one of an orthogonal ennuple in V n. Let en|i be the unit

tangents of given congruence C so that ti = en|i

Then equation (7) becomes

en|k (en| i,j – en| j,i) + en| i (en| j,k – en| k,j) + en| j (en|k,i – en|i,k) = 0 … (8)

e n|k (en |i , j – en | j ,i ) e ip| eqk| + en |i (en | j , k − en| k , j ) e ip| eqk + en | j (en |k ,i – en |i , k ) e ip |eqk|

where p and q are two new indices chosen from 1, 2, …, n – 1. i.e., p,q and n are unequal.

But e n |k eqk| = δ q = 0, q ≠ n

h

en |i e kp| = δ hp = 0, p ≠ n

so, we have

en|j (γnqp – γnpq) = 0

Ricci's Coefficients of Rotation and Congruence 211

⇒ γnpq = γnqp, … (9)

Conversely if equation (9) in true then we get equation (8). Which implies that equation (7) are

satisfied bty en|i Hence en| is a normal congruence.

Thus necessary and sufficient conditions that the congruence en| of an orthogonal ennuple be a

normal congruence are that

γnqp = γnpq (p, q = 1, 2, …, n – 1 such that p ≠ q)

THEOREM 11.5 Necessary and sufficient conditions that all the congruences of an orthogonal ennuple

be normal.

Proof: If all the congruences of a orthogonal ennuple are normal. Then

γnqp = γnpq (p, q = 1, 2, …, n – 1 such that p ≠ q)

If the indices h, k, l and unequal then

γhkl = γhlk … (1)

But due to skew-symmetric property i.e., γhlk = – γlhk.

So,

γhkl = γhlk = – γlhk = – γlkh, from (1)

= γklh (skew-symmetric property).

= γkhl, from (1)

γhkl = – γhkl , (skew-symmetric property).

γhkl + γhkl = 0

⇒ 2 γhkl = 0

⇒ γhkl = 0

where (l, h, k = 1, 2, …, n such that h, k, kl are unequal).

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 Tensors and Their Applications

n

∑γ

h =1

nhk eh |i

m

= e n |i , m e k |

n

∑ γ nhk eh|i ek | j = en|i ,me mk| ek| j

h,k =1

= en |i ,mδ mj Since ekm| ek| j = δmj

n

∑γ

h, k =1

nhk eh|i e k | j = e n|i, j … (2)

curl en|i = en| i, j – en| j,i

n n

h,k =1 n ,k =1

n n

curl en| i = ∑

h ,k =1

γ nhk eh |i ek | j – ∑γ

h , k =1

nkh ek | j e h|i

curl en|i = ∑

h ,k =1

(γnhk – γnkh) eh| iek| j … (3)

(i) Let h and k take the values 1, 2, …, n – 1.

(ii) Either h = n or k = n or h = k = n.

Now, the equation (3) becomes

n −1 n –1

curl en| i = ∑

h ,k =1

(γ nhk – γ nkh ) eh |i ek | j + ∑

h =1

(γ nhn – γ nnh ) e h|i en | j

+ ∑ k =1

( γ nnk – γ nkn ) en |i e n| j + (γ nnn – γ nnn ) en |i en | j

So,

n –1 n –1 n–1

curl en|i = ∑

h , k =1

(γ nhk – γ nkh ) e h |i e k | j + ∑

h =1

γ nhn e h|i en | j – ∑γ

k =1

nkn en |i ek| j

n–1 n–1

curl en|i = ∑h ,k

(γ nhk – γ nkh ) eh|i ek | j + ∑γ

h =1

nhn ( eh |i en | j – en |i ek | j ) … (4)

Ricci's Coefficients of Rotation and Congruence 213

if γnhk – γnkh = 0

i.e.,

if γnhk = γnkh.

i.e., if the congruence e n| is normal.

Again the second term of R.H.S of equation (4) vanishes

if γ nhn = 0 i.e., if γ hnn = 0

i.e., if the congruence e n| is a geodesic congruence. Further, if first and second term on right

hand side of equation (4) both vanishes then

curl en|i = 0

Hence we have proved that if the congruence e n| satisfies any two of the following conditions

then it will also satisfy the third.

(a) e n| is a normal congruence

(b) e n| is irrotational

(c) e n| is a geodesic congruence

It has been shown that given a congruence of curves, it is possible to choose, in a multiply infinite

number of ways, n – 1 other congruences forming with the given congruence an orthogonal ennuple.

Consider the system of n – 1 congruence discovered by Ricci, and known as the system canonical with

respect to the given congruence.

THEOREM 11.7 Necessary and sufficient conditions that the n – 1 congruences eh| of an orthogonal

ennuple be canonical with respect to en| are

γnhk + γnkh = 0; (h, k = 1, 2, …, n – 1, h ≠ k ).

Proof: Let the given congruence en| be regarded as nth of the required ennuple. Let en|i be unit tangent

to given congruence.

X ij = ½ (en| i, j + en| j,i) … (1)

Let us find a quantity ρ and n quantities ei satisfying the n + 1 equations

e n|i ei = 0

i, j = 1, 2, …, n … (2)

( X ij – ωgij e i + ρE n| j = 0

Writing equation (2) in expansion form, we have

e n|1e1 + en |2 e 2 + …… + en| n e n = 0 … (3)

and

( X 1 j – ωg1 j ) ei + ( X 2 j – ωg 2 j ) e 2 + … ( X nj – ωg nj ) e n + ρ e n| j = 0

214 Tensors and Their Applications

for j = 1, 2, …, n.

1 2 n

( X 11 – ωg11 ) e + ( X 21 – ωg 21 ) e + … + ( X n1 – ωg n 1 ) e + ρ en |1 = 0

( X 12 – ωg12 ) e1 + ( X 22 – ωg 22 ) e 2 + … + ( X n 2 – ωg n 2 )e n + ρ en |2 = 0

M

( X 1n – ωg1 n ) e1 + ( X 2 n – ωg 2 n ) e 2 + … + ( X nn – ωg nn ) e n + ρ en |n = 0

from (3)

e n|1 e1 + en |2 e 2 + … en |n e n + ρ.0 = 0

eliminating ρ and the quantities e1, e2, …en, we have the equation

(X 11 – ωg11 ) (X 21 – ωg 21 ) … (X n1 – ωg n1 ) en |1

( X12 – ωg12 ) ( X 22 – ωg 22 ) … ( X n2 – ωg n2 ) e n|2

M

( X1n – ωg1n ) (X 2n – ωg 2n ) … (X nn – ωg nn ) en |n

en |1 en |2 … en| n 0

which is of degree n – 1 in ω. Hence there will be n – 1 roots of ω and these roots be ω1, ω2, …,

ωn – 1. All roots are real. Let ωh be one of these roots and let the corresponding values of ρ and ei be

denoted by ρh and e ih| respectively. Then ρh and e ih| will satisfy the equation (2), we have

en |i e ih| = 0 … (4)

and ( X ij – ωg ij ) eih | + ρh en | j = 0 … (5)

Similarly ωk be another root of these roots, we have

en |i e ik| = 0 … (6)

( X ij – ωg ij ) eik | + ρh en | j = 0 … (7)

Multiplying (5) by e kj| and (7) ehj| and using (4) and (6), we get

Since X ij and g ij are symmetric tensor in i and j. Now, interchanging i and j in equation (9), we

get

( X ij – ωk g ij ) ehi |e kj| = 0 … (10)

Subtracting (10) and (8), we get

(ωh – ωk ) g ij e ih| ekj | = 0 … (11)

Since ωk – ωk ≠ 0 as ωh ≠ ωh.

⇒ g ij eih| ekj| = 0 … (12)

Ricci's Coefficients of Rotation and Congruence 215

This shows that e ih| and e kj| unit vectors are orthogonal to each other and hence the congruence

e h| and e k| (h ≠ k) are orthogonal to each other. Hence the n – 1 congruence eh| (h = 1, 2, …, n) thus

determined form an orthogonal ennuple with en|.

Using equation (12), equation (10) becomes

X ij e ih| ekj| = 0

1

e (

2 n |i , j

)

+ e n| j , i . Then we have,

1

( e

2 n |i , j

)

+ en | j ,i ehi |ekj| = 0

γnhk + γnkh = 0; (h, k = 1, 2, …, n – 1 such that h ≠ k ) … (13)

Conversely if equation (13) is true then (n – 1) congruences eh| of the orthogonal ennuple and

canonical with respect to e n| . Hence necessary and sufficient condition that the n – 1 congruences e h|

of an orthogonal ennuple be canonical with respect to en| are

γnhk + γnkh = 0; (h, k = 1, 2, …, n such that h ≠ k )

THEOREM 11.8 Necessary and sufficient conditions that n –1 mutually orthogonal congruences e h|

orthogonal to a normal congruence e n| , be canconical with respect to the later are γnhk = 0 where k, h

= 1, 2,…, n – 1 such that h ≠ k.

Proof: By theorem (11.7), Necessary and sufficient conditions that (n – 1) congruences e h| of an

orthogonal ennuple be canonical with respect to en| are

γnhk + γnkh = 0 (h, k = 1, 2, …, n – 1 such that h ≠ k).

If the congruence en| is normal. Then

γnhk = γnkh

The given condition γ nhk + γ nkh = 0 becomes

γnhk + γnhk = 0

2γnhk = 0

γnhk = 0 Proved.

EXAMPLE 1

1

If en| are the congruences canonical with respect to en| prove that (i) ωh = γnhk (ii) ρh = γhnn,

2

(iii) If en| is a geodesic congruence, the congruences canonical with respect to it are given by

( X ij − ωg ij )ei = 0 = 0

Solution

Suppose (n – 1) congruences eh| of an orthogonal ennable in a V n are canonical with respect to

the cougruence eh| then

en |iehi | = 0 … (1)

216 Tensors and Their Applications

where

1

X ij = (e + en| j,i) … (3)

2 n| i,j

Since e in| unit tangents then

g ij ehi |ehj| = 0 … (4)

1

(e n|i , j + en| j , i ) e ih|e hj| – ωh = 0; from (3)

2

1

(e e i e j + en | j ,i eih| ehj | ) – ωh = 0

2 n|i , j h | h|

1 (γ + γ ) – ω = 0

nhh nhh h

2

ωh = γnhh

(ii) Multiplying (2) by enj| , we get

1

(e n|i , j + en| j , i ) e ih|e hj| + ρh = 0, since e ih| e hj| = 0

2

1

(e ei e j + en | j ,i ehi | ehj | ) + ρh = 0

2 n|i , j h | n|

1

(γ + γnnh) + ρh = 0

2 nhn

1

ρh = – γ , since γnnh = 0

2 nhn

1

or ρh = γ

2 hnn

(iii) If en| is a geodesic congruence, then γhnn = 0 from the result (ii), we have

1

ρh = × 0

2

ρh = 0

Ricci's Coefficients of Rotation and Congruence 217

( X ij – ω h gij ) ehi | + 0en| j = 0 or ( X ij – ωh g ij ) e ih| = 0

this gives ( X ij – ωg ij ) e i = 0 .

EXAMPLE 2

Prove that when a manifold admits an orthogonal systyem of n normal congruences then any of

these in canonical with respect to each other congruence of the system.

solution

Let e ih| (h = 1, 2, …, n) be unit tangents to n normal congruences of an orthogonal ennuple in a

V n. So that

γlhk = 0, where l, h, k = 1, 2,...n such that l, h, k being unequal.

It is required to show that a congruence eh| is canonical with respect to the congruence ek| (h, k = 1, 2,

…, h ≠ k). We know that the n – 1 congruence en| of an orthogonal ennuple be canonical to en| iff

γnhk + γnkh = 0

This condition is satisfied by virtue of equation (1). Hence (n – 1) congruences eh| of an orthogonal

ennuple are canonical to en|.

Similarly we can show that any n – 1 congruences are canonical to the remaining congruence.

It follows from the above results that any one congruence is canonical with respect to each other

congruence of the system.

EXERCISE

1. If φ is a scalar invariant

n

∑ φ,ij eih|ehi | = ∆ φ2

h =1

3. If eh| are the unit tangents to n mutually orthogonal normal congruences and e11 + be21 is also a

normal congruence then ae11 – be21 is a normal congruence.

4. Show that

∂Q

∂ ∂φ

–

∂ ∂Q

∂sh ∂sk ∂sk ∂sh =

l

∑

( γ lhk – γ lkh )

∂sl

where S h denotes the arc length of a curve through a point P of an ennuple and Q is a scalar invariant.

5. If the congruence eh| (h = 1, 2, … n – 1) of an orthogonal ennuple are normal, prove that they are

canonical with respect to other congruence eh| .

CHAPTER – 12

HYPERSURFACES

12.1 INTRODUCTION

We have already studied (Art 10.3 chapter 10) that if m > n then we call V n to be a subspace of V m and

consequently V m is called enveloping space of V n. We also know that there are m – n linearly independent

normals Nα to V n. (Art 10.3 Theorem 10.5, chapter 10). If we take m = n + 1 then V n is said to be

hypersurface of the enveloping space V n + 1

Let V n be Riemannian space of n dimensions referred to cordinates xi and having the metric ds2 =

gij dx dxj. Let V n be Riemannian space of m dimensions referred to coordinates yα and having the metric

i

ds2 = aαβ dyα dyβ. Where m > n. Let Greek letters α, β, γ … take the values 1, 2, …, m and latin indices

i, j, k, … take the values 1, 2, …n. Then we have, the relation between aαβ and gij

α β

gij = aαβ ∂y ∂y … (1)

∂xi ∂x j

Since the function yα are invariants for transformations of the coordinates xi in V n, their first

covariant derivatives with respect to the metric of V n are the same as their ordinary derivatives with

respect to the variables xi.

∂yα

i.e., yα, i =

∂xi

Then equation (1) can be written as

The vector of V n + i whose contravariant components are y,αi is tangential to the curve of

parameter xi in V n. Consequently if N α are the contravariant components of the unit vector normal to

V n. Then we have

a αβ N β y,αi = 0, (i = 1, 2, …, n) … (3)

and a αβ N αN β = 1 … (4)

Hypersurface 219

Let C be any curve in V n and s its arc length. The along this curve the x's and the y's may be expressed

as function of s only. Let uα and ν β be the components in the y's of two unit vector fields which are

parallel along C with respect to V m. Similarly w i the components in x's of a unit vector field which is

parallel along C with respect to V n. Now, uα is parallel along C relative to V m then we have,

dyβ

uα,β =0

ds

∂u α γ dy β

β – uγ =0

∂y α β ds

∂u α dy β γ dy β

γ

– u =0

∂y β ds α β ds

duα γ dy β

– uγ =0

ds α β ds

duα α dy β

= u γ β γ ds … (1)

ds

similarly v β is parallel along C relative to V m then

α

dvβ γ γ dy

= – v α β ds … (2)

ds

and wi is parallel along C relative to V n then

j i dx

k

dwi

= – w j k ds … (3)

ds

The Christoffel symbol with Greek indices being formed with respect to the aαβ and the y's and

christoffel symbol with Latin indices with respect to the gij and the x's.

Let Aβαi be a tensor field, defined along C, which is mixed tenser of the second order in the y's and

α

a covariant vector in the x's. Then the product uα v β wi Aβi is scalar invariant and it is a function of s

along c. Its derivative with respect to s is also a scalar invariant.

α

Differentiating uα v β wi Aβi with respect to s, we have

d dAα dv β dw i

(uα v β wi Aβαi ) = u v β w i βi + du α v β w i A α + u α

wi α

Aβi

+ uα v β Aβαi

dt α βi ds ds

ds ds

dAβαi duδ β i δ dv δ dw j

u αvβ Aβαj

β

= uα v w i

+ v w Aβi + uα w i Aδαi +

ds ds ds ds

220 Tensors and Their Applications

dAβαi γ

β i α dy γ α β i δ dy

= u αv w + Aβδi .u αv βwi – Aδi .u αv w βγ ds

ds δγ ds

i j dx

k

α β

– Aβj uα v w ik ds , by equation (1), (2) & (3).

α

β i dAβi δ α dy

γ

α δ dy

γ

α j dx

k

u

= α v w + Aβi

. – Aδi – Aβj

.

ds δ γ ds β γ ds i k ds

= Scalar, along C

Since the outer product uαv βwi is a tensor and hence from quotient law that the expression within

α α

the bracket is tenser of the type Aβi and this tensor is called intrinsic derivative of Aβi with respect to

s.

The expression within the bracket can also be expressed as

+ Aβδi y,γk – Aδαi y,γk – Aβαj

ds ∂xk

δγ βγ ik

dxk

Since is arbitrary..

ds

So, by Quotient law the expression within the bracket is a tensor and is called tensor derivative of

α

Aβi with respect to xk. It is denoted by Aβαi;k . Then we have

∂Aβαi α δ j

Aβαi;k = + Aβδi y,γk – Aδαi y,γk – Aβαj .

dx k δγ

βγ

ik

α

with respect to xk .

Note:– Semi-colon(;) is used to denote tensor differentiation.

THEOREM 12.1 Tensor differentiation of sums and products obeys the ordinary rules of differentiation.

Proof: Suppose Aβα , Bβα and B γ are tensors in V m.

(i) To prove that

Now,

∂Cβα α α

Cβα;k = + Cβα y, ck – Caα y, ck

∂x βc

k

a c

Hypersurface 221

∴ ( Aβα + Bβα ); k = + ( Aβa + Aβα ) y,ck – ( Aaα + Baα ) y ,ck

β c

k

dx a c

∂Aβα

aα c αa c

= ∂x k + Aβ y, k – Aa y , k +

ac β c

∂Bβα α a

+ Bβa y,ck – Baα y,ck

∂xk

ac βc

(ii) Prove that

α α

Then Dβγ is a tensor

we have

α

∂Dβγ α a a

D αβγ ; k = + D;ak yc – Dαaγ y,ck – Dαβa y,ck

∂x k

ac βc γc

∂( Aβα B γ ) α

Bγ y ,ck – Aaα Bγ y,ck – Aβα B a y ,ck

a a

∴ ( A αβ Bγ ) ; k = + Aβγ

a

βc βc

k

dx ac

∂Aβα

α α c αa c ∂B

Bγ + Aβα γ – Ba y,ck

a

= dxk + Aβ y, k – Aa y , k

ac β c dx

k

γc

Hence the result (ii).

Note:– α, β, γ, a, c take values from 1 to m while k take values from 1 to n.

or

To prove that the metric tensor of the enveloping space is generalised covariant constant with respect

to the Christoffel symbol of the subspace.

Proof: We have (see pg. 220)

∂aαβ γ γ

aαβ;i = – aγβ y,δi – aαγ y,δi

dxi αβ

βδ

∂aαβ

= – ( [αδ, β] + [βδ, α ] ) y,δi

dx i

222 Tensors and Their Applications

∂aαβ ∂aαβ δ

or aαβ;i = – y,i

dx i

∂ yδ

or aαβ;i = i

–

δ

⋅ i = – = 0 Proved.

dx ∂y dx ∂x i ∂x i

At a point of a hypersurface V n of a Riemannian space V n + 1, the formula of Gauss are given by

y;αij = Ωij N α

Proof: Since yα is an invariant for transformation of the x's and its tensor derivative is the same as its

covariant derivative with respect to the x’s, so that

dyα

yα; i = yα, i = … (1)

dxi

Again tenser derivative of equation (1) with respect to x’s is

α

∂ l α

= ( y,αi ) – y,αl + yβ,i y,γj

βγ

j

dx ij

∂ ∂y α α l β γ α

= – y,l + y,i y, j

∂x j ∂xi ij βγ

∂ 2 yα l α

y;αij = – y,αl + yβ,i y,γj … (2)

βγ

i j

∂x ∂x ij

Interchanging j and i in (2), we have

∂ 2yα l α

y;αji = – y ,αl + y β, j y ,γi

∂x ∂x βγ

j i

ji

∂ 2yα l α

y;αji = – y ,αl + y β,l y ,γj … (3)

∂x ∂x βγ

i j

ij

α α

[On interchanging β and γ in third term of R.H.S. of (3) and using βγ = γβ ].

On comparing equation (2) and (3), we have

y;αij = y;αji

Hypersurface 223

Let g ij dx idx j and aαβdyα dyβ be fundamental forms corresponding to V n and V n + 1 respectively..

Then

∂yα ∂yβ

gij = aαβ

∂xi ∂x j

Taking tensor derivative of both sides with respect to xk

gij;k = aαβ;k y;αi y;βj + aαβ y;αi y;βjk + aαβ y;αik y;βj

we have,

or aαβ y;αik y,βj + aαβ y,αi y;βjk = 0, using (1), Art. 12.4 … (4)

By cyclic permutation on i, j, k in (4), we have

subtracting equation (4) from the sum of (5) and (6), we get.

This shows that y;αij is normal (orthogonal) to y,βk . Since y,βk is tangential to V n and hence y;αij is

where Nα is unit vector normal to V n and Ω ij is a symmetric covariant tensor of rank two. Since

y;αij is a function of x's the tenser Ω ij is also a function of x's.

The equation (8) are called Gauss’s formula.

From equation (8)

y;αij = Ωij N α

= Ω ij N 2

224 Tensors and Their Applications

or Ω ij = y;αija αβN β.

The quadratic differential form

Ωij dxi dx j

is called the second fundamental form for the hypersurface V n of V n + 1. The components of tenser Ω ij

are said to be coefficient of second fundamental form.

Note: The quadratic differential form g ij dx idx j is called first fundamental form.

If U α and u i be the contravariant components of the vector u relative to V n and V n + 1 respectively then

we have (from chapter 10, Theorem 10.4)

∂y α i

U α = u = y,αi ui … (1)

∂ xi

Let the derived vector to vector u along C with respect to metric of V n and Vn + 1 are denoted by

p and q respectively. Then

pi dxj

= u,i j

ds

α d yβ

and qα = U,β

ds

dx j

= U ;αj , (from equation 1, Art 12.4) … (2)

ds

Taking the tensor derivative of each side of equation (1) with respect to x's, we have

U ;αj = y; ij u + y, i u , j

α i α i

U ;αj = Ω ij N u + y ,i u , j

α i α i

Then

α i α i dx j

qα = (Ωij N u + y, i u, j )

ds

j

i dx α

Ω α i

or qα = ij u N + y,i p … (3)

ds

i dx j

where pi = u, j

ds

Hypersurface 225

Now suppose that vector u is a the unit tangent t to the curve C. Then the derived vectors q and

p are the curvature vectors of a C relatively to V n + 1 and V n respectively. Then equation (3) becomes.

dx i dx j α

N + y ,αi p i

qα = Ωij ds ds … (4)

dx i dx j

Taking Kn = Ω ij , we get

ds ds

α

qα = Kn Nα + y,i pi ...(5)

Kn is called Normal curvature of V n at any point P of the curve C and Kn Nα is called normal

curvature vector of V n + 1 in the direction of C.

Meaunier's Theorem

If Ka and Kn are the first curvature of C relative to V n + 1 and normal curvature of V n respectively

r r

and w is the angle between N and C (C being the unit vector of V n + 1 then the relation between Ka, Kn

and w is given by

Kn = Ka cos ω

Proof: We know that

qα = Kn N α + p i y,αi ...(1)

dxi dx j

where Kn = Ωij

ds ds

Let Ka and Kg be the first curvatures of C with respect V n + 1 and V n respectively then

Ka = a αβ q α q β , Kg = gij pi p j

r r

Let ω be the angle between N and C .

r r

Then N ⋅ C = N ⋅ C cos ω

r r

N .C = cos ω ...(2)

r r

If b is a unit vector of V n + 1 in the direction principal normal of C with respect to V n then

equation (1) becomes

r r r

Ka C = K g b + K n N … (3)

r

Taking scalar product of equation (3) with N , we have

r r r r r r

K a N .C = K g N. b + K n N. N

Kn = Ka cos ω … (4)

Proved.

226 Tensors and Their Applications

EXAMPLE 1

Show that the normal curvature is the difference of squares of geodesic curvatures.

Solution

We know that (from Meurier’s Theorem, equation 3)

r r r

KaC = Kg b + Kn N

Taking modulus of both sides, we get

2 2

Ka2 = Kg + Kn

2 2

or Kn2 = Ka – K g .

Theorem 12.3 To show that the first curvature in Vn + 1 of a geodesic of the hypersurface Vn is the

normal curvature of the hypersurface in the direction of the geodesic.

Proof: From, example 1, we have

2 2

Ka2 = Kg + Kn

If C is a geodesic of V n then p i = 0 ⇒ Kg = 0

Then we have

Ka2 = Kn2

⇒ Ka = K n. Proved.

Dupin's Theorem

The sum of normal curvatures of a hypersurface V n for n mutually orthogonal directions is an

invariant and equal to Ω ij g ij .

Proof: Let ehi | (h = 1, 2, …, n) be unit tangents to n congruences of an orthogonal ennuple in a V n. Let

Knh be normal curvature of the hypersurface V n in the direction of the congruence eh | . Then

Knh = Ωij eih | ehj|

The sum of normal curvatures for n mutually orthogonal directions of an orthogonal ennuple is a

V n is

n

∑ Knh = ∑ Ω

n

ij eih| ehj|

h =1 h =1

= Ωij ∑e

h =1

i

h| ehj|

= Ω ij g ij

= Scalar invariant Proved.

Hypersurface 227

12.6 DEFINITIONS

(a) First curvature (or mean curvature) of the hypersurface Vn at point P.

It is defined as the sum of normal curvatures of a hypersurface V n form mutually orthogonal directions

at P and is denoted by M. Then

M = Ω ij g ij

(b) Minimal Hypersurface

The hypersurface V n is said to be minimal if M = 0

i.e., Ω ij g ij = 0

(c) Principle normal curvatures

The maximum and minimum values of Kn are said to be the principle normal curvatures of V n at P.

Since these maximum and minimum values of Kn correspond to the principal directions of the symmetric

tensor