Professional Documents
Culture Documents
intentionally left
blank
Copyright © 2006, New Age International (P) Ltd., Publishers
Published by New Age International (P) Ltd., Publishers
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.
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
PRELIMINARIES
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.
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
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
ANSWERS
∂ 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.
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
∂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.
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.
∂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.
∂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
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
∂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
Then A j11 2j 2 ...rj s are called component of mixed tensor of rank r + s .
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 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 .
∂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
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
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.
∂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.
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
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 .
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
EXAMPLE
[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
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
∂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
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
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 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).
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
(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
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
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
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
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
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
EXERCISES
(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.
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
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
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
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.
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
+ 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
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
1 0 0
since g = g ij = 0 r2 0
0 0 1
2
g= r
Metric Tensor and Riemannian Metric 37
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
So that x1 = x, x 2 = y , x 3 = z and x 1 = r , x 2 = θ , x 3 = φ
Let g ij and g ij be the metric tensors in cartesian and spherical coordinates respectively..
The metric in cartesian coordinates is given by
ds 2 = dx + dy + dz
2 2 2
ds 2 = g ij dx dx ; (i, j = 1, 2, 3)
i j
But
38 Tensors and Their Applications
∂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
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
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
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
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
∫
t2
ds = 5 dt = =
2 12 15 units
t1 1
t2 −1 1
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.
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
= g ij l A A + g ij lmA B + m lg ij B A + m g ij B B
2 i j i j i j 2 i j
= l 2 + 2lm cos θ + m 2
Since
g ij A i A j = A 2 = 1; g ij B B = B = 1.
i j 2
and
r r r
g ij A i B j = cosθ; as A & B are unit vector i.e., A = 1 ⇒ A 2 = 1 .
Since square of magnitude of any vector ≥ 0.
So, the square of the magnitude of lAi + mB i ≥ 0.
or l 2 + 2lm cos θ + m 2 ≥ 0
(l + m cos θ )2 + m 2 − m 2 cos 2 θ ≥0
(l + m cosθ) 2 + m 2 (1 − cos2 θ) ≥ 0
This inequality holds for the real values of l & m.
if 1 − cos2 θ ≥ 0
⇒ cos 2 θ ≤ 1
cos θ ≤ 1
Proved.
THEOREM 3.5 The magnitude of two associated vectors are equal.
Proof: Let A and B be magnitudes of associate vectors A i and A i respectively. Then
i j
A 2 = g ij A A ...(1)
and
ij
B 2 = g Ai A j ...(2)
From equation (1)
A 2 = (gij Ai ) A j
A2 = Aj A j ...(3)
since g ij Ai = A j
from (3) and (4)
2
A2 = B
⇒ A=B
i
So, magnitude of Ai and A are equal.
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
g ij
cos wij =
g ii g jj
48 Tensors and Their Applications
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.
∂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
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
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)
∑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)
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
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
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
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
Take l = k.
g ik g ik = δ kk = 1
g ik = [g ik ]−1 =
Gik
⇒ (Theorem 2.13 , Pg 25)
g
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
1 g 33 = 1
g 11 = 1, g =
22
Also, ,
r2 r sin 2 θ
2
1 ∂g jk ∂g ik ∂g ij
[ij, k ] =
2 ∂x i
+ − , i, j, k = 1, 2,3 ...(1)
∂x j ∂x k
Taking i = j = 2 and k = 1 in (1)
1 ∂g 21 ∂g 21 ∂g 22
[22, 1] = + 2 − 1 Since g21 = 0.
2 ∂x 2 ∂x ∂x
1 ∂0 ∂0 ∂r 2
= 2 + 2 − 1
2 ∂x ∂x ∂x
1 ∂r 2
= − = −r
2 ∂r
Taking i = 1, j = k = 3 in (1)
1 ∂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
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
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.
∂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 .
∂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 .
∂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 .
∂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
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
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
EXAMPLE 7
∂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 ]
∇ 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
= 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.
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
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.
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
∂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
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
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
∂ ∂ 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
∂ 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
∂
∂ 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
∂ 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
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
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.
∂ ∂ 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
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)
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
∂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
∂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 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
∂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 .
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
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 .
∑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.
⇒ ( 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)
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
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.
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 .
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
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
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
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
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
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 )
r 2 sin 2 θ 1
K= = = constant.
r 4 sin 2 θ r 2
EXERCISES
1. Show that
∂[ jl, i] ∂[ jk , i] α α
Rijkl = − + [il , α ] − [ik , α]
∂xk ∂xl j k j l
2. Show that
1 ∂ g il
2 2
2
∂ g jk ∂ 2 g ik ∂ g jl
+ g ( [ jk , β][il , α] − [ jl , β][ik , α ] ) .
Rijkl = + − − αβ
2 ∂x j ∂xk ∂xi ∂xl ∂ x j ∂xl ∂xi ∂ xk
3. Using the formula of the problem 2. Show that
Rijkl = − R jikl = − Rijlk = Rklij and Rijkl + Riklj + R iljk = 0
4. Show that the curvature tensor of a four dimensional Riemannian space has at the most 20 distinct
non-vanishing components.
h
5. (a) If prove that the process of contraction applied to the tensor Rijk generates only one new tensor
Rij which is symmetric in i and j.
1
(b) If ds2 = g11 ( dx1 ) 2 + g 22 (dx 2 )2 + g 33 ( dx3 )2 . Prove that Rij =
Rihhj , (h, i, j being unequal).
g hh
6. Show that when in aV3 the coordinates can be chosen so that the components of a tensor g ij are zero
when i, j, k are unequal then
1
(i) Rhj = Rhiij
g ii
110 Tensors and Their Applications
1 1
(ii) Rhh = Rhiih + Rhjjh
g ii g jj
7. Prove that if
α 1 ∂R
Ri = g Rij then Ri , α = 2 i
α αj
∂x
and hence deduce that when n > 2 then scalar curvature of an Einstein space is constant.
8. If the Riemannian curvature K of Vn at every point of a neighbourhood U of Vn is independent of the
direction chosen, show that K is constant throughout the neighbourhood U. Provided n > 2.
9. Show that a space of constant curvature K0 is an Einstein space and that R = K 0 n (1 –n ).
10. Show that the necessary and sufficient condition that Vn be locally flat in the neighbourhood of 0 is
that Riemannian Christoffel tensor is zero.
11. Show that every V2 is an Einstein space.
12. For two dimensional manifold prove that
R
K= −
2
13. Show that if Riemann-Christoffel curvature tensor vanishes then order of covariant differentiation is
commutative.
CHAPTER – 6
The concept of symmetry and skew-symmetry with respect to pairs of indices can be extended to
cover to pairs of indices can be extended to cover the sets of quantities that are symmetric or skew-
symmetric with respect to more than two indices. Now, consider the sets of quantities A i ... i k or
Ai l ... i k depending on k indices written as subscripts or superscripts, although the quantities A may not
represent tensor.
6.3 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
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 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
and
(2) ei1i2 ... in = δi1i 2 ... in
1 2 ... n
δ ijk
αβγ = δ αβ1 + δ αβ2 + δαβ 3 = δαβ
ij1 ij 2 ij 3 ij
Hence
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
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
5. Prove that εijk g is a covariant tensor of rank three where where ε ijk is the usual permutation
symbol.
CHAPTER – 7
GEOMETRY
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).
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
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
= 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
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).
∂ 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
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.
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)
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 ∂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
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 .
∂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
∂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
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
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.
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
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
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
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
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
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
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
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
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
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.
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.
=
1
2
(
m g jk δij x& k + δki x& j )
1 1
= m g jk δ ij x& k + m g jk δ ki x& j
2 2
=
1
2
(
m gik x& k + g ji x& j )
1
= m (gij x& + gij x& ) gij = g ji
j j
as
2
Analytical Mechanics 147
∂T
= m g ij x& j
∂x& i
∂T
or = m g ik x& k …(2)
∂x& i
Differentiating equation (2) with respect to t, we get
d ∂T
= m
dt ∂x& i
d
g x& k
dt ik
( )
d k
= m g ik x& + g ik &x&
k
dt
∂g ik dx j k
= ∂x j dt x& + g ik &x&
k
m
d ∂T ∂g
= m ikj x& x& + m g ik &x&
j k k
… (3)
dt ∂x& i ∂x
1
Since T = m g jk x& j x&k
2
Differentiating it with respect to xi, we get
∂T 1 ∂g jk j k
i =
m x& x& … (4)
∂x 2 ∂x i
Now,
d ∂T ∂T ∂g 1 ∂g
− i = m ikj x& j x&k + m g ik&x&k − m jki x& j x& k
dt ∂x& i ∂x ∂x 2 ∂x
1 gik j k 1 1 ∂g
= m g ik &x& + m j x& x& + m g ik x& j x& k − m jki x& j x& k
k
2 ∂x 2 2 ∂x
1 g ik j k 1 ∂g ij k j 1 ∂g jk j k
= m g ik &x& + x& x& + m k x& x& − m i x& x&
k
m
2 ∂x j 2 ∂x 2 ∂x
1 ∂gik ∂g ij ∂g jk j k
= m g ik &x& + 2 m j + k − x& x&
k
∂x ∂x ∂x i
[
= m g 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
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
O
Y2
φ
R
mg cos φ
Y1 θ P
mg sin φ
Y3 mg
Fig. 8.1.
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
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.
∫
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
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
∫
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.”
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
∂x r j ∂x r
= q& + ,
∂q j ∂t ( j = 1, 2, ..., n)
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
d ∂T ∂T ∂x r
−
dt ∂q& i
∂qi
= ∑
m a r
∂q i
… (8)
∂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)
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
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
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
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
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
∫ 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
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'
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
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 ∂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)
1 ∇ 2ψ
ψ(P ) = − ∫
4π ∞ r
dV … (6)
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
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.
∫
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
∫ 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
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
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.
∫
B
s= F dt ...(3)
A
174 Tensors and Their Applications
Since curve C is geodesic, then the integral (3) should be stationary, we have from Euler's
condition
∂F d ∂F
− =0 ...(4)
∂x i dt ∂ x& i
Differentiating equation (2) with respect to x k and x& k we get,
∂F 1 ∂g ij i j
= x& x&
∂x k
2 s& ∂ x k
∂F 1 1
and k =
2 gik x& i = g ik x&i
∂x& 2s& s&
d ∂F 1 1 ∂ g ik j i 1
= − 2 &s& g ik x& + x& x& + g ik &x& i
i
dt ∂ x& k s& s& ∂x j s&
Putting these values in equation (4), we get
1 ∂g ij i j 1 1 ∂ g ik j i 1
x& x& − − 2 &s&g ik x& i + x& x& + g ik &x&i = 0
2 s& ∂ x k s &
s ∂ x j &
s
&s& ∂g 1 ∂ g ij i j
g ik &x&i − g ik x& i + ikj − x& x& = 0
s& ∂x 2 ∂x k
&s&
g ik &x&i − g x& i + [k , ij ] x&i x& j = 0
&s ik
multiplying it by g km , we get
&s& km
g km gik &x&i − g g ik x&i + g km [k , ij ] x& i x& j = 0
s&
m
g km gik = δim and g [k , ij] = i j
km
But
&s& m m i j
x&&m − x& + x& x& = 0
s& i j
d 2xm 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.
∂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
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
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.
⇒ 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
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.
g ij = a ij = δij = 1, if i = j
0, if i ≠ j
∂g ij ∂a ij
= =0
∂x k
∂x k
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
∑ (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
or a ij ( y i2 − y1i ) ( y i3 − y1i ) = 0
or ∑ (y
i =1
i
2 − y1i ) ( y 3i − y1i ) = 0 ...(1)
(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)
(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
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
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
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.
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
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.
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
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
eh|i eli| ,j = 0
∑γh
lhk eh |m = el |m , j ekj |
Replacing m by i, we get
∑ γ lhkei|m = el |i , j e kj|
h
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 … (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 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
= ∑∑ 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
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.
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
∂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
THEOREM 11.6 If a congruence of curves satisfy two of the following conditions it will also satisfy
the third
(a) that it be a satisfy the third
(b) that it be a geodesic congruence
(c) that it be irrotational.
Proof: Consider an orthogonal ennuple and e ih| (h = 1, 2, …, n) be n unit tangent to n congruences of
this orthogonal ennnuple.
From theorem (11.2), we have
∑γ
h =1
nhk eh |i
m
= e n |i , m e k |
∑γ
h, k =1
nhk eh|i e k | j = e n|i, j … (2)
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)
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
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
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
e n|i ei = 0
i, j = 1, 2, …, n … (2)
( X ij – ωgij e i + ρE n| j = 0
( 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
)
+ en | j ,i ehi |ekj| = 0
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
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
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
∂u α dy β γ dy β
γ
– u =0
∂y β ds α β ds
duα γ dy β
– uγ =0
ds α β ds
duα α dy β
= u γ β γ ds … (1)
ds
similarly v β is parallel along C relative to V m then
α
dvβ γ γ dy
= – v α β ds … (2)
ds
and wi is parallel along C relative to V n then
j i dx
k
dwi
= – w j k ds … (3)
ds
The Christoffel symbol with Greek indices being formed with respect to the aαβ and the y's and
christoffel symbol with Latin indices with respect to the gij and the x's.
Let Aβαi be a tensor field, defined along C, which is mixed tenser of the second order in the y's and
α
a covariant vector in the x's. Then the product uα v β wi Aβi is scalar invariant and it is a function of s
along c. Its derivative with respect to s is also a scalar invariant.
α
Differentiating uα v β wi Aβi with respect to s, we have
d dAα dv β dw i
(uα v β wi Aβαi ) = u v β w i βi + du α v β w i A α + u α
wi α
Aβi
+ uα v β Aβαi
dt α βi ds ds
ds ds
dAβαi duδ β i δ dv δ dw j
u αvβ Aβαj
β
= uα v w i
+ v w Aβi + uα w i Aδαi +
ds ds ds ds
220 Tensors and Their Applications
dAβαi γ
β i α dy γ α β i δ dy
= u αv w + Aβδi .u αv βwi – Aδi .u αv w βγ ds
ds δγ ds
i j dx
k
α β
– Aβj uα v w ik ds , by equation (1), (2) & (3).
α
β i dAβi δ α dy
γ
α δ dy
γ
α j dx
k
u
= α v w + Aβi
. – Aδi – Aβj
.
ds δ γ ds β γ ds i k ds
= Scalar, along C
Since the outer product uαv βwi is a tensor and hence from quotient law that the expression within
α α
the bracket is tenser of the type Aβi and this tensor is called intrinsic derivative of Aβi with respect to
s.
The expression within the bracket can also be expressed as
dxk
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
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
∂Cβα α α
Cβα;k = + Cβα y, ck – Caα y, ck
∂x βc
k
a c
Hypersurface 221
∂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
∂( 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
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
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
∂ 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
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
gij;k = aαβ;k y;αi y;βj + aαβ y;αi y;βjk + aαβ y;αik y;βj
or aαβ y;αik y,βj + aαβ y,αi y;βjk = 0, using (1), Art. 12.4 … (4)
By cyclic permutation on i, j, k in (4), we have
This shows that y;αij is normal (orthogonal) to y,βk . Since y,βk is tangential to V n and hence y;αij is
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.
∂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
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 Ωij .
(d) Principal directions of the hypersurface at a point P.
The principal directions determined by the symmetric tensor Ωij at P are said to be principal directions
of the hypersurface at P.
(e) Line of curvature in V n
A line of curvature in a hypersurface V n is a curve such that its direction at any point is a principal
direction.
Hence we have n congruences of lines of curvature of a V n.
THEOREM 12.4 To show that the mean curvature of a hypersurface is equal to the negative of the
divergence of the unit normal.
or
To show that the first curvature of a hypersurface is equal to the negative of the divergence of the unit
normal.
or
To show that the normal curvature of a hypersurface for any direction is the negative of the tendency
of the unit normal in that direction.
Proof: Let N be the unit normal vector to the hypersurface V n in y's and let Nα be its covariant
components. Let t be the unit tangent vector to N congruences eh| (h = 1, 2, …, n) of an orthogonal
ennuple in V n and let Thα| be the contravariant components in V n + 1 of t. Since t is orthogonal to N,
therefore
Th|α N α = 0 … (1)
Now Thα|,βThβ| is the first curvature of the curve eh| and N α,β . Thβ| . Thα| is the tendency of Nα is the
direction of eh| . Hence equation (3) implies that the normal component of the first curvature of the eh|
relative to V n + 1 or the normal curvature of V n in the direction of the curve eh|
= – tendency of N in the direction of the curve eh| … (4)
Taking summation of both of (3) and (4) for h = 1, 2, …, n
we have
i.e., mean curvature or first curvature of a hypersurface = – divergence of the unit normal.
Corollary: To prove that
M = – divn + 1N
Proof: Since N is a vector of unit magnitude i.e., constant magnitude, its tendency is zero. also by
definition, the div n + 1 N and div n N differ only by the tendency of the vector N in its direction. But the
tendency of N is zero hence it follows that
divn N = – div n + 1 N
Hence M = – div n N = – div n + 1 N.
Statement
The normal curvature Kn of V n for any direction of ar in V n is given by
n
Kn = ∑ Kh cos2 α h
h =1
where Kh are the principal curvature and α h are the angles between direction of ar and the congruence
eh|
Proof: The principal directions in V n determined by the symmetric covariant tensor teser Ω ij are
given by
(Ωij – K h g ij ) p ih| = 0 … (1)
where Kh are the roots of the equation
Ωij – Kgij = 0 … (2)
Ωij p i p j
Kn = g p i p i … (3)
ij
Hypersurface 229
or ai = ∑e
h=1
i
h|
cosα h … (6)
r r
where cos α h| = a ⋅ e h | = ai eh|i
r
ai being the contravariant components of a and α h being the inclination of vector a to eh|. The
r
normal curvature of V n for the direction of a is given by
Kn = Ωij a ia j
from (6), we get
Kn = Ω ij ∑e i
h| cos α h
∑e j
k| cos α h
h h
n
= ∑
h ,k =1
(Ωij e ih| ekj | ) cosα h cosαk
= ∑
h =1
(Ωij ehi |ehj| ) cos 2 α h
n
Kn = ∑ Kh cos α h
2
… (7)
h =1
This is a generalisation of Euler’s Theorem.
A direction in V n which is self-conjugate is said to be asymptotic and the curves whose direction
are along asymptotic directions are called asymptotic lines.
Therefore the direction the vector ar at a point of V n be asymptotic if
Ωij a i a j = 0 … (2)
The asymptotic lines at a point of a hypersurface satisfies the differential equation
Ω ij dxi dx j = 0 … (3)
THEOREM 12.5 If a curve C in a hypersurface V n has any two of the following properties it has the
third
(i) it is a geodesic in the hypersurface Vn
(ii) it is a geodesic in the enveloping space Vn + 1
(iii) it is an asymptotic line in the hypersurface V n.
Proof: Let C be a curve in the hypersurface V n. The normal curvature Kn of the hypersurface V n in the
direction of C is given by
2 2
Ka2 = Kg + Kn … (1)
where Ka and Kg are the first curvatures of C relative to enveloping space V n + 1 and hypersurface V n
respectively.
Suppose C is a geodesic in the hypersurface V n [i.e., (i) holds] then Kg = 0.
If C is also a geodesic in the enveloping space V n + 1 [i.e., (ii) holds] then Ka = 0.
Now, using these values in equation (1), we have
Kn2 = 0 ⇒ Kn = 0
Implies that C is an asymptotic line is the hypersurface V n i.e., (iii) holds.
Hence we have proved that (i) and (ii) ⇒ (iii)
Similarly we have proved that
(ii) and (iii) ⇒ (i)
and (i) and (iii) ⇒ (ii)
α ∂ yα
i.e., y;αi = y,i = … (1)
∂ xi
The unit Normal N α be the contravariant vector in the y’s whose tensor derivative with respect
to the x’s is
∂N α α β δ
N;αi = + N y, i … (2)
∂x i β δ
Hypersurface 231
2 aαβ N β N ;αi = 0
a αβ N β N ;αi = 0 … (4)
which shows that N;αi is orthogonal to the normal and therefore tangential to the hypersurface.
aαβN α y,βi = 0
or g ki A kj + Ωij = 0
α α ∂ y α ∂ yβ
since aαβ y, k y, i = a αβ = g ki
∂x ∂x
k i
Multiplying this equation by gim, we get
g ki g im A kj + Ωij g im = 0
232 Tensors and Their Applications
δm k
k A j + Ωij g
im
=0
Aim + Ωij g im = 0
im
Aim = – Ωij g
Substituting this value in equation (5), we get
N;αi = Ω ij g y, k = – Ω ji g
α
y,αk
ik jk
N;αi = – Ωij g jk
y ,αk … (6)
Theorem 12.6 The derived vector of the unit normal with respect to the enveloping space, along a
curve provided it be a line of curvature of the hypersurface.
Proof: Since the tensor derivative of N α is
Ωijδilei = λ ei g il
(Ω il – λg il ) ei = 0 (l = 1, 2, …, n)
This equation implies that the direction of e i is a principal direction for the symmetric tensor Ω il
i.e., e i is a principal direction for the hypersurface V n. Hence by definition the curve C is a line of
curvature V n.
Hypersurface 233
where A i is a covariant tenser of rank one and difference of two tensers Ai, jk – Ai,kj is a covariant
tenser of rank three.
It y,αi are components of a covariant tensor of rank one in x 's. Then replacing A i by y,αi in
equation (1), we get
Multiplying equation (5) by a αβ yβ,l and summed with respect to α. Using the relations
a αβ y ,αi y,βl = 0
and g αβ = N β y,αl = 0,
we get
Rlijk = (Ωlj Ωik – Ωlk Ωij ) + R βγδε y β,l y,γi y ,δj y,εk … (6)
Multiplying (6) by a αβ N β and summing with respect to α. Using relations
a αβ N β y α,l = 0
and aαβ N β B β = 0
M = Ωij g ij = ω gij g ij = ωn
M
⇒ ω= .
n
So that the conditions for indeterminate lines of curvature are expressible as
M
Ω ij = g , from (i) … (9)
n ij
If all the geodesics of a hypersurface V n are also geodesics of an enveloping V n + 1. They
hypersurface V n is called a totally geodesic hypersurface of the hypersurface V n + 1.
THEOREM 12.7 A totally geodesic hypersurface is a minimal hypersurface and its-lines of curvature
are indeterminate.
Proof: We know that
2 2
Ka2 = Kn + K g … (1)
and a hypersurface is said to be minimal if
M=0 … (2)
and the lines of curvature are indeterminate if
M
Ω ij = g ij … (3)
n
If a hypersurface V n is totally geodesic then geodesics of V n are also geodesics of V n + 1.
i.e., Ka = 0 = k g
Now, from (1), we have
Kn = 0
But normal curvature Kn is zero for an asymptotic direction. Hence a hypersurface V n is totally
geodesic hypersurface iff the normal curvature Kn zero for all directions in V n and hence
Ω ij = 0
M = Ω ij g ij = 0
i.e., equation (2) is satisfied.
Hence, the totally geodesic hypersurface is minimal hypersurface.
In this case equation (3) are satisfied hence the lines of curvature are indeterminate.
Hypersurface 235
Let aij be the components in the x ′s of a symmetric tensor of the rank two and evaluated at the
pole O. Then the equation
y i aij y j
=1 … (2)
represents a central quadratic hypersurface
Substituting the value of equation (1) in equation (2), we get
sξi aij sξi = 1
1
ξiaijξi = … (3)
s2
The equation (3) showing that the two values of s are equal in magnitude but opposite in sign.
The positive value of s given by equation (3) is the length of the radius of the quadric (2) for the
direction ξ i .
THEOREM 12.8 The sum of the inverse squares of the radii of the quadric for n mutually orthogonal
directions at O is an invariant equal to a ij g ij .
Proof: If e ih| , (h = 1, 2, …n) are the contravariant components of the unit tangents at O to the curves
of an orthogonal ennuple in Sn. The radius Sh relative to the direction e ih| is given by
1
aij ehi | ehj | =
s 2h
n h
= a ij ∑
h =1
eih | e hj|
n
∑ (s h )–2 = a ij g ij Proved.
h =1
236 Tensors and Their Applications
THEOREM 12.9 The equation of hyperplane of contact of the tangent hypercone with vertex at the
point Q (y ) .
Proof: Given (from equation 2, pg. 235)
aij y i y j = 1
Differentiating it
aij dy i y j + aij y i dy j = 0
or aij dy i y j + aij y j dy i = 0
2 aij dy i y j = 0
aij dy i y j = 0
This shows that dy i is tangential to the quadric. Hence yj is normal to the quadric.
The tangent hyperplane at the point P ( y j) is given by
(Y i – y i ) aij y j = 0
aij Y i y j = aij y i y j
Consequently for all positions of the hyperplane through R, Q lies on the hyperplane
yi a ~yj =1
ij … (6)
This is required equations of the polar hyperplane of R and R is the pole of this hyperplane.
Hypersurface 237
Let P ( y α ) be a point on the unit normal N α such that distance of P from P is ρ in the direction
of N α such that
y α = y α + ρN α … (4)
The vector ( y,αi + ρN ,αi ) dx i is tangential to V n whereas N αdρ is a normal vector. Therefore if the
Multiplying it by y,αi and summing with respect to α, we get, using equation (3), as
( gil – ρ Ω ij δlj ) d xi = 0
( g il – ρΩil ) dx i = 0
12.15 HYPERSPHERE
The locus of a point in Sn which moves in such way that it is always at a fixed distance R from a fixed
point C (b α ) is called a hypersphere of radius R and centre C. Therefore the equation of such a
hypersphere is given by
n
∑ (y
α =1
α
– b α )2 = R 2 … (1)
THEOREM 12.10 The Riemannian curvatrure of a hypersphere of radius R is constant and equal to
1
.
R2
Proof: Let the hypersurface be a V n of Sn + 1 and let its centre be taken as origin of Euclidean
coordinates in Sn + 1. Then the hypersphere is given by
∑ (y
α
α 2
) = R 2, (α = 1, 2, …, n + 1) … (2)
For the point in V n the y's are functions of the coordinates xi on the hypersphere.
Differentiating equation (2) with respect to x i, we get
∑ y α y,αi =0 … (3)
α
and again differentiating it with respect to x j, we get.
∑y
α
α
,j y,αi + ∑y
α
α
y,αij = 0 … (4)
By Gauss formula,
y,αij = Ω ij N α … (5)
Using (5), equation (4) becomes
g ij + ∑ y α Ωij N α =0 … (6)
α
From equation (3) it follows that y,αi is perpendicular to y α . But y,αi is tangential to V m. Hence
y α is normal to V n.. The equation (2) implies that the components of the unit vector N α are given by
yα
N α = ⇒ y α = RN α … (7)
R
Using (7), equation (6) becomes
g ij + ∑RNα
α
N α Ωij = 0
or g ij + RΩ ij ∑ (N
α
α 2
) =0
or g ij + R Ωij = 0 since ( N α )2 = 1
240 Tensors and Their Applications
2. To prove that for a space Vn with positive constant Riemannian curvature K these exists sets of n + 1 real
coordinate y α satisfying the condition
n +1
∑( y α
) = 1 1
where R 2 =
α =1
K K
∑ (y α ) 2 =
1
Proved.
α K
EXAMPLE 2
Show that the directions of two lines of curvature at a point of a hypersurface are conjugate.
Solution
EXAMPLE 3
Show that the normal curvature of hyper surface V n in an asymptotic direction vanishes.
Solution
Let us consider a curve C in a V n. If C is an asymptotic line then it satisfies the differential
equation
Ω ij dxi dx j = 0
dxi dx j
i.e., Ωij =0 … (1)
ds ds
Now the normal curvature Kn of the hypersurface V n in an asymptotic direction in a V n is given
by
dxi dx j
Kn = Ωij
ds ds
i.e., Kn = 0 from (1)
EXAMPLE 4
To prove that if the polar hyperplane of the point R passes through a point P then that of P passes
through R.
Solution
11. Prove that the necessary and sufficient condition that system of hypersurface with unit normal N be
isothermic is that
ρ ( N div N – N .∇N ) = 0.
12. Show that the normal curvature of a subspace in an asymptotic direction is zero.
13. If straight line through a point P in S n meets a hyperquadric in A and B and the polar hyperplane of
P in Q prove that P, Q are harmonic conjugates to A, B.
14. What are evolutes of a hypersurface in an Euclidean space. Show that the varieties ρ1 = constant in
evolute are parallel, having the curves of parameter ρ1 as orthogonal geodesics of V n .
INDEX
Bianchi identity 94 D
Binormal 137
Degree of freedom 157
C Dextral Index 1
Divergence 75
Canonical congruence 213
Divergence Theorem 161
Christoffel’s symbols 55
Dummy index 1
Completely skew-symmetric 111
Dupin’s Theorem 226
Completely symmetric 111
Concept 188 E
Congruence of curves 49
Einstein space 103
Conjugate (or Reciprocal) Symmetric Tensor 25
Einstein tensor 95
Conjugate directions 229
Einstein’s Summation Convention 1
Conjugate metric tensor 34
Euler’s condition 171
Conservative force field 144
Euler’s Theorem 228
Contraction 18
Evolute 237
Contravariant tensor of rank r 14
Contravariant Tensor of rank two 9 F
Contravariant vector 7
Covariant tensor of rank s 14 First curvature 227
First curvature vector of curve 170
244 Tensors and Their Applications