Professional Documents
Culture Documents
applic PDF
applic PDF
intentionally left
blank
To
My parents
This page
intentionally left
blank
FOREWORD
It gives me great pleasure to write the foreword to Dr. Nazrul Islams book entitled Tensors and Their
Applications. I know the author as a research scholar who has worked with me for several years. This
book is a humble step of efforts made by him to prove him to be a dedicated and striving teacher who
has worked relentlessly in this field.
This book fills the gap as methodology has been explained in a simple manner to enable students
to understand easily. This book will prove to be a complete book for the students in this field.
Ram Nivas
Professor,
Department of Mathematics and Astronomy,
Lucknow University,
Lucknow
This page
intentionally left
blank
PREFACE
Tensors were introduced by Professor Gregorio Ricci of University of Padua (Italy) in 1887
primarily as extension of vectors. A quantity having magnitude only is called Scalar and a quantity with
magnitude and direction both, called Vector. But certain quantities are associated with two or more
directions, such a quantity is called Tensor. The stress at a point of an elastic solid is an example of a
Tensor which depends on two directions one normal to the area and other that of the force on it.
Tensors have their applications to Riemannian Geometry, Mechanics, Elasticity, Theory of Relativity,
Electromagnetic Theory and many other disciplines of Science and Engineering.
This book has been presented in such a clear and easy way that the students will have no difficulty
in understanding it. The definitions, proofs of theorems, notes have been given in details.
The subject is taught at graduate/postgraduate level in almost all universities.
In the end, I wish to thank the publisher and the printer for their full co-operation in bringing out
the book in the present nice form.
Suggestions for further improvement of the book will be gratefully acknowledged.
This page
intentionally left
blank
CONTENTS
Introduction .......................................................................................................... 6
Transformation of Coordinates ................................................................................ 6
Covariant and Contravariant Vectors ......................................................................... 7
Contravariant Tensor of Rank Two .......................................................................... 9
Covariant Tensor of Rank Two ................................................................................ 9
Mixed Tensor of Rank Two..................................................................................... 9
Tensor of Higher Order ......................................................................................... 14
Scalar or Invariant ................................................................................................ 15
Addition and Subtraction of Tensors ....................................................................... 15
Multiplication of Tensors (Outer Product of Tensors) ............................................... 16
Contraction of a Tensor ........................................................................................ 18
Inner Product of Two Tensors .............................................................................. 18
Symmetric Tensors .............................................................................................. 20
Skew-symmetric Tensor ....................................................................................... 20
Quotient Law ....................................................................................................... 24
xii
2.16.
2.17.
Chapter4
4.1.
4.2.
4.3.
4.4.
4.5.
4.6.
4.7.
4.8.
Christoffels Symbol............................................................................................. 55
Transformtion of Christoffels Symbols .................................................................. 64
Covariant Differentiation of a Covariant Vector ........................................................ 67
Covariant Differentiation of a Contravariant Vector ................................................... 68
Covariant Differentiation of Tensors ....................................................................... 69
Riccis Theorem .................................................................................................. 71
Gradient, Divergence and Curl ............................................................................... 75
The Laplacian Operator ......................................................................................... 80
Exercises ............................................................................................................. 83
5.4.
5.5.
5.6.
5.7.
Contents
5.8.
5.9.
5.10.
5.11.
5.12.
5.13.
xiii
Chapter6 The e-systems and the Generalized Krnecker Deltas ................ 111-115
6.1.
6.2.
6.3.
6.4.
6.5.
7.2.
7.3.
7.4.
7.5.
7.6.
7.7.
7.8.
7.9.
xiv
8.10.
8.11.
8.12.
8.13.
8.14.
8.15.
Chapter11
11.1.
11.2.
11.3.
11.4.
11.5.
11.6.
11.7.
Contents
xv
This page
intentionally left
blank
CHAPTER 1
PRELIMINARIES
1.1 n-DIMENSIONAL SPACE
In three dimensional rectangular space, the coordinates of a point are (x, y, z). It is convenient to write
(x1, x2, x3) for (x, y, z). The coordinates of a point in four dimensional space are given by (x1, x2, x3, x4).
In general, the coordinates of a point in n-dimensional space are given by (x1, x2, x3,...., xn) such ndimensional space is denoted by V n.
1.2 SUPERSCRIPT AND SUBSCRIPT
In the symbol Aklij , the indices i, j written in the upper position are called superscripts and k, l written
in the lower position are called subscripts.
1.3 THE EINSTEIN'S SUMMATION CONVENTION
n
ai x i = a i x i
i =1
a jx j = a1x1 + a2 x2 + + an x n
and
and
Properties
1. If x1, x2, ... xn are independent coordinates, then
x i
= 0 if i j
x j
x i
= 1 if i = j
x j
This implies that
x i
= ij
x j
It is also written as
2.
x i x k
= ij .
x k x j
ii = 11 + 22 + 33 + + nn
ii
= 1 + 1+ 1 + + 1
ii = n
3.
a ij kj = a ik
Since
31 1
32 2
33 3
3n n
a 3 j 2j = a 2 + a 2 + a 2 + + a 2
Preliminaries
= a32
(as 12 = 32 = = n2 = 0 and 22 = 1)
In general,
a ij kj = a i 11k + a i 2 2k + a i 3 3k + + aik kk + + a in nk
ik
a ij kj = a (as 1k = 2k = = nk = 0 and kk = 1)
4.
j k = ik
ij kj = 1i 1k + i2 2k + i3 3k + + ii ik + + in nk
= ik (as 1i = i2 = i3 = = in = 0 and ii = 1)
EXAMPLE 1
Write
d
dx1 dx 2
dx n
+ 2
+ + n
= 1
using summation convention.
dt
x dt
x dt
x dt
Solution
d
dx 1 dx 2
dx n
+
+
+
=
dt
x 1 dt x 2 dt
x n dt
d dx i
= i
dt
x dt
EXAMPLE 2
Expand: (i) aij xixj; (ii) glm gmp
Solution
(i)
1 j
2 j
n j
aij x i x j = a1 j x x + a2 j x x + + a nj x x
g lm g mp = g l 1 g 1 p + g l 2 g 2 p + + g ln g np , as m is dummy index.
EXAMPLE 3
If aij are constant and aij = aji, calculate:
(i) x (aij x i x j )
(ii) x x (aij xi x j )
k
k
l
Solution
(aij x i x j ) = a ij
( xi x j )
(i)
x k
xk
j
i
= a ij xi x + a ij x j x
k
k
= a ij xi jk + aij x j ik ,
as
x j
x k
= (aij jk ) xi + (aij ik ) x j
= aik xi + akj xj
= aik xi + aki xi
as aij jk = aik
as j is dummy index
(aij x i x j )
= 2aik xi
xk
(aij x i x j )
(ii)
x k
= 2aikxi
Differentiating it w.r.t. xl :
2 (aij xi x j )
xk xl
= 2 aik
xi
xl
= 2 aik il
2 (aij xi x j )
xk xl
as a ik il = alk.
= 2alk
EXAMPLE 4
If
aij xi xj= 0
where aij are constant then show that
aij + aji = 0
Solution
Given
aijx ix j= 0
almxlxm= 0
Differentiating it w.r.t. xi partially,
(alm x l x m ) = 0
xi
l m
a lm
(x x ) = 0
xi
a lm
Since
x l m
x m l
x + alm
x =0
xi
x i
x l
= li
xi
and
x m
= mi
xi
= jk
Preliminaries
alm il x m + alm im x l = 0
aim x m + ali x l = 0
as alm li = aim and alm mi = ali .
Differentiating it w.r.t. xj partially
aim
x m
x l
+ ali
x j
x j = 0
aim mj + ali lj = 0
aij + aji= 0
Proved.
EXERCISES
1. Write the following using the summation convention.
(i) (x 1 )2 + (x2 )2 + (x3 )2 + . . . + (xn )2
(ii) ds2 = g 11 (dx1 )2 + g 22 (dx2 )2 + . . . + g nn (dxn )2
(iii) a 1 x1 x3 + a 2 x2 x3 + . . . + a n xn x3
2. Expand the following:
(i) a ij xj
(ii)
( ga )
(iii) Aik Bi
3. Evaluate:
(i) x j ij
(ii) ij kj kl
(iii) ij ij
ANSWERS
1.
2.
(i) xi xj
(ii) ds2 = g ij dxi dxj
1
(i) a i1 x + a i2 x2 + a i3 x3 + . . . + a in xn
(ii)
( ga ) +
( g a ) + +
(i) xi
4. Cij xi xj
(ii) il
(iii) n
(iii) a i x i x 3 .
( ga )
CHAPTER 2
TENSOR ALGEBRA
2.1 INTRODUCTION
A scalar (density, pressure, temperature, etc.) is a quantity whose specification (in any coordinate
system) requires just one number. On the other hand, a vector (displacement, acceleration, force, etc.)
is a quantity whose specification requires three numbers, namely its components with respect to some
basis. Scalers and vectors are both special cases of a more general object called a tensor of order
n whose specification in any coordinate system requires 3n numbers, called the components of tensor.
In fact, scalars are tensors of order zero with 3 = 1 component. Vectors are tensors of order one with
31 = 3 components.
2.2 TRANSFORMATION OF COORDINATES
In three dimensional rectangular space, the coordinates of a point are (x, y, z) where x, y, z are real
numbers. It is convenient to write (x1, x2, x3) for ( x, y , z ) or simply xi where i = 1, 2, 3. Similarly in
n- dimensional space, the coordinate of a point are n-independent variables (x1, x2,..., x n) in X-coordinate
system. Let ( x 1 , x 2 ,..., x n ) be coordinate of the same point in Y-coordinate system.
Let x 1 , x 2 , , x n be independent single valued function of x1, x2,...., xn, so that,
x 1 = x 1 (x 1 , x 2 , ...., x n )
x 2 = x 2 ( x1 , x 2 , ...., x n )
x 3 = x 3 (x 1 , x 2 , ...., x n )
M
x n = x n (x1 , x 2 , ..., x n )
or
x i = x i ( x1 , x 2 , ..., x n ) ;
i = 1, 2, , n
(1)
Tensor Algebra
i = 1, 2, ..., n
The equations (1) and (2) are said to be a transformation of the coordinates from one coordinate
system to another
2.3 COVARIANT AND CONTRAVARIANT VECTORS (TENSOR OF RANK ONE)
Let (x1, x2, ..., xn) or x i be coordinates of a point in X-coordinate system and ( x 1 , x 2 ,..., x n ) or x i be
coordinates of the same point in the Y-coordinate system.
Let A i, i = 1, 2, ..., n (or A 1, A 2, ..., A n) be n functions of coordinates x 1, x2, ..., x n
in X-coordinate system. If the quantities A i are transformed to A i in Y-coordinate system then according
to the law of transformation
x i j
x j
j
A
i
or
A
=
x j
x i A
Then A i are called components of contravariant vector.
Let Ai , i = 1, 2,..., n (or A 1, A 2, , A n) be n functions of the coordinates x1, x2, ..., x n
Ai =
one.
Note: A superscript is always used to indicate contravariant component and a subscript is always used to indicate
covariant component.
EXAMPLE 1
If xi be the coordinate of a point in n-dimensional space show that dxi are component of a
contravariant vector.
Solution
1
2
n
Let x1, x2, ..., xn or xi are coordinates in X-coordinate system and x , x ,..., x or x i are
coordinates in Y-coordinate system.
If
x i = x i (x1 , x 2 ,..., x n )
i
i
i
dx i = x dx 1 + x dx 2 + + x dx n
1
2
n
d xi =
x i
x j
dx j
Solution
Let x1, x2, ..., xn or x i are coordinates in X-coordinate system and x 1 , x 2 , ..., x n or x i are
coordinates in Y-coordinate system.
Consider
( x 1 , x 2 ,..., x n ) = ( x1 , x 2 ,..., x n )
=
1 2
x + 2 x + + n x n
1
x
x
x
x1 x 2
x n
=
+
+
+
i
x
x 1 x i x 2 x i
x n x i
x j
=
x i
x j x i
x j
=
x i
x i x j
or
is component of covariant
x i
EXAMPLE 3
Show that the velocity of fluid at any point is a component of contravariant vector
or
Show that the component of tangent vector on the curve in n-dimensional space are component
of contravariant vector.
Solution
Let
dx1 dx 2
dx n
,
, ,
be the component of the tangent vector of the point ( x1 , x 2 ,..., x n ) i.e.,
dt dt
dt
dx i
be the component of the tangent vector in X-coordinate system. Let the component of tangent
dt
Tensor Algebra
dx i
. Then x 1 , x 2 , ..., x n or x i being a
dt
x i dx1 x i dx 2
x i dx n
+
+
+
dt1 dt
dx 2 dt
dx n dt
d xi
=
dt
x i dx j
dx j dt
dx i
is component of
dt
contravariant vector.
i.e. the component of tangent vector on the curve in n-dimensional space are component of
contravariant vector.
2.4 CONTRAVARIANT TENSOR OF RANK TWO
Let A ij (i, j = 1, 2, ..., n) be n2 functions of coordinates x1, x2, ..., xn in X-coordinate system. If the
quantities A ij are transformed to A ij in Y-coordinate system having coordinates x 1 , x 2 , ..., x n . Then
according to the law of transformation
A ij =
x i x j kl
A
x k x l
ij
Then A are called components of Contravariant Tensor of rank two.
10
Note:
(i) The rank of the tensor is defined as the total number of indices per component.
(ii) Instead of saying that A ij are the components of a tensor of rank two we shall often say Aij is a tensor
of rank two.
THEOREM 2.1 To show that the Krnecker delta is a mixed tensor of rank two.
Solution
Let X and Y be two coordinate systems. Let the component of Kronecker delta in X-coordinate
system ij and component of Krnecker delta in Y-coordinate be ij , then according to the law of
transformation
ij =
x i x i x l x k
=
x j x k x j x l
ij =
x i x l k
l
x k x j
Ai
do not form a tensor..
x j
Solution
Let X and Y be two coordinate systems. As given Ai is a covariant tensor. Then
Ai =
x k
Ak
x i
Differentiating it w.r.t. x j
Ai
j =
x
x j
Ai
x k
=
x j
x i
x k
x i Ak
Ak
2 xk
+
A
k
x j
x i x j
It is not any law of transformation of tensor due to presence of second term. So,
(1)
Ai
is not a
x j
tensor.
THEOREM 2.2 To show that ij is an invariant i.e., it has same components in every coordinate
system.
Proof: Since ij is a mixed tensor of rank two, then
ij =
x i x l k
l
x k x j
Tensor Algebra
11
x i
=
x k
x l k
x j l
x l k x k
l =
x x
x j
x j
x i
x i
i
=
= ij
ij =
,
as
j
x j
x j
=
x i x k
k
, as
So, ij is an invariant.
THEOREM 2.3 Prove that the transformation of a contravariant vector is transitive.
or
Prove that the transformation of a contravariant vector form a group.
Proof:
x i p
A
x p
x i x p q
A i = x i x q A from (1)
x i q
i =
A
A
x q
This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of contravariant vectors is transitive or form a group.
Ai =
Ai =
x p
Ap
x i
12
x p x q
Aq
x i x p
x q
Aq
Ai =
x i
This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of covariant vectors is transitive or form a group.
Ai =
A ji
x i x q p
Aq
=
x p x j
=
A ji =
x x x x
x x x x
As from (1)
x i x s r
As
x r x j
This shows that if we make direct transformation from x i to x i , we get same law of transformation.
This property is called that transformation of tensors form a group.
THEOREM 2.6 There is no distinction between contravariant and covariant vectors when we restrict
ourselves to rectangular Cartesian transformation of coordinates.
Proof: Let P(x, y) be a point with respect to the rectangular Cartesian axes X and Y. Let ( x , y ) be the
coordinate of the same point P in another rectangular cartesian axes X and Y , Let (l1, m1) and (l2, m2)
be the direction cosines of the axes X , Y respectively. Then the transformation relations are given by
x = l1 x + m1 y
...(1)
y = l2 x + m2 y
and solving these equations, we have
x = l1 x + l 2 y
y = m1 x + m2 y
put x = x1 , y = x 2 , x = x 1 , y = x 2
...(2)
Tensor Algebra
13
From (1)
x
2
2
= l , but x = x1 , y = x , x = x 1 , y = x
x 1
Then
x x 1
=
=l .
x x 1 1
Similarly,
2
2
y
x
y
x
= l2 = 1 ;
= m2 = 2
x
y
x
x
x
x 1
= m1 = 2 ;
y
x
...(3)
So, we have
A 1 = l1 A1 + m1 A 2
A 2 = l2 A1 + m 2 A 2
Consider the covariant transformation
..(4)
x j
A j ; j = 1,2
x i
x1
x 2
A
+
A2
Ai =
1
x i
x1
Ai =
for i = 1, 2 .
x1
x 2
A
+
A2
1
x 1
x 1
x 1
x 2
A
+
A2
A2 =
1
x 2
x 2
A1 =
From (3)
A1 = l1 A1 + m1 A2
A2 = l2 A1 + m2 A2
...(5)
14
Then Ai ii 2 ...i r
i1i 2 ...i r
x i1 x i2
x i r
p1 p 2 ...p r
x p1 x p 2 x p r A
are called components of contravariant tensor of rank r.
x q1 x q 2
x q s
x j1 x j 2
x j s
Aq1 ,q 2 ...,q s
Aij11i 2j 2......i rj s are transformed to A ij11ij22......i rj s in Y-coordinate system having coordinates x 1 , x 2 , , x n . Then
according to the law of transformation
A ij11ij22......i rj s =
x i1 x i 2 x ir
x p1 x p2 x p r
x q1 x q 2
x q s
Aqp11qp22......q ps r
j1
j2
js
x x
x
i i ...i
Then A j11 2j 2 ...rj s are called component of mixed tensor of rank r + s .
2 ... j s
A tensor of type Ai1j1i 2j...
is known as tensor of type (r, s ) , In (r,s), the first component r
ir
indicates the rank of contravariant tensor and the second component s indicates the rank of covariant
tensor.
ij
Thus the tensors Aij and A are type (0, 2) and (2, 0) respectively while tensor Aij is type (1, 1).
Tensor Algebra
15
EXAMPLE
ijk
is a mixed tensor of type (3, 2) in which contravariant tensor of rank three and covariant
Alm
tensor of rank two. Then according to the law of transformation
Almijk =
x i x j x k x a x b
x x x x l x m
Aab
( x1 , x 2 ,..., x n ) = ( x 1 , x 2 ,..., x n )
Scalar is also called tensor of rank zero.
For example, Ai Bi is scalar..
2.9 ADDITION AND SUBTRACTION OF TENSORS
THEOREM 2.7 The sum (or difference) of two tensors which have same number of covariant and the
same contravariant indices is again a tensor of the same rank and type as the given tensors.
Proof: Consider two tensors Aij11i 2j 2......i rj s and B ij11i 2j 2......i rj s of the same rank and type (i.e., covariant tensor of
rank s and contravariant tensor of rank r.). Then according to the law of transformation
x i1 x i 2
x ir x q1 x q 2
x q s
A ij11ij22......i rj s =
Aqp11qp22......q ps r
p1
p2
pr
j1
j2
js
x x
x x x
x
and
B ij11ij22......i rj s =
x i1 x i 2
x ir x q1 x q 2
x q s
Bqp11qp22......q ps r
x p1 x p2
x p r x j1 x j2
x j s
Then
A ji11 ij22......i rj s B ij11ij22......i rj s =
x i1 x i 2
x i r x q1 x q 2 x q s
x p1 x p2
x p r x j1 x j 2
x j s
(A
p1 p2 ...p r
q1q 2 ...q s
B qp11qp2 2......qps r
If
A ij11ij22......i rj s B ji11ij22......irj s = C ij11ij22......i rj s
and
Aqp11qp2 2......q ps r Bqp11qp2 2......qps r = C qp11qp2 2......qps r
So,
x i1 x i 2
x ir x q1 x q 2
x q s
,..., pr
Cqp11,,qp22,...,
qs
x p1 x p2
x p r x j1 x j 2
x j s
i , i ,..., i
This is law of transformation of a mixed tensor of rank r+s. So, C j11 , j22 ,..., jrs is a mixed tensor of
C ij11ij22......i rj s =
16
EXAMPLE 5
If Akij and Bnlm are tensors then their sum and difference are tensors of the same rank and type.
Solution
As given Akij and Bkij are tensors. Then according to the law of transformation
Akij =
x i x j x r pq
Ar
x p x q x k
and
Bkij
x i x j x r pq
Br
=
x p x q x k
then
Akij Bkij =
x i x j x r pq
Ar Brpq
p
q
k
x x x
If
Akij B kij = C kij and Arpq B rpq = C rpq
So,
x i x j x r pq
Cr
x p x q x k
The shows that C kij is a tensor of same rank and type as Akij and Bkij .
C kij =
A ij11ij22......i rj s =
x i1 x i 2
x ir x q1 x q 2
x q r
x p1 x p2
x p r x j1 x j 2
x j s
Aqp11qp22......q ps r
x k1 x k 2
x km x1 x 2
x n
x 1 x 2
x m x l1 x l 2
x l n
B112 2......n m
and
km
Bl1kl12k...2 ...
=
ln
x i1
x p1
x i r x q1
x pr x j1
x q s x k1
x km x 1
x n
x j s x 1
x m x l1
x l m
Aqp11qp22......q ps r B112 2......n m
Tensor Algebra
17
If
...km
i i ...i
k k ...k
C ij1ij2 ......i rjk1l kl 2 ...
= A j11 j22 ...rjs Bl1l12 ...2 ln n
1 2
s 1 2 ln
and
...,p r
C qp11qp2 2......qpsr 112...2...nm = Aqp11qp22...,
B112 2......n m
qs
So,
x q s x k1
x i1
x ir x q 1
x j s x 1
x p1
x p r x j1
x k m x 1
x n
l l C qp11qp2 2......q ps r1 12 2......n h
x m x 1
x n
...km
C ji11ij22......irj ksl11kl 22 ...
ln =
i i ...i
...k
This is law of transformation of a mixed tensor of rank r + m + s + n. . So, C j11 j22 ...rj ks1lk12l 2 ...lmn is a
mixed tensor of rank r + m + s + n . or of type (r + m, s + n ) . Such product is called outer product or
open proudct of two tensors.
THEOREM 2.9 If A i and B j are the components of a contravariant and covariant tensors of rank one
then prove that A iB j are components of a mixed tensor of rank two.
Proof: As Ai is contravariant tensor of rank one and B j is covariant tensor of rank one. Then
according to the law of transformation
Ai =
x i k
A
x k
...(1)
Bj =
x l
Bl
x j
...(2)
and
x i x l k
A Bl
x k x j
This is law of transformation of tensor of rank two. So, Ai B j are mixed tensor of rank two.
Such product is called outer product of two tensors.
EXAMPLE 6
Show that the product of two tensors Aij and B mkl is a tensor of rank five.
Solution
As Aij and B mkl are tensors. Then by law of transformation
A ji =
x i x q p
Aq
x p x j
and
Bmkl =
x k x l x t rs
Bt
x r x s x m
18
x i x q x k x l x t p rs
A B
x p x j x r x s x m q t
This is law of transformation of tensor of rank five. So, Aij B mkl is a tensor of rank five.
2.11 CONTRACTION OF A TENSOR
The process of getting a tensor of lower order (reduced by 2) by putting a covariant index equal to a
contravariant index and performing the summation indicated is known as Contraction.
In other words, if in a tensor we put one contravariant and one covariant indices equal, the
process is called contraction of a tensor.
ijk
For example, consider a mixed tensor Alm
of order five. Then by law of transformation,
x i x j x k x s x t pqr
Ast
x p x q x r x l x m
Put the covariant index l = contravariant index i, so that
Almijk =
Aimijk =
x i x j x k x s x t pqr
Ast
x p x q x r x i x m
x j x k x s xt pqr
Ast
x q x r x p x m
x j x k x t s pqr
p Ast
x q x r x m
Aimijk =
Since
x s
= sp
x p
x j x k x t pqr
A pt
x q x r x m
ijk
This is law of transformation of tensor of rank 3. So, Aim
is a tensor of rank 3 and type (1, 2)
ijk
while Alm is a tensor of rank 5 and type (2, 3). It means that contraction reduces rank of tensor by
two.
EXAMPLE 7
If A i and B i are the components of a contravariant and covariant tensors of rank are respectively
then prove that A iB i is scalar or invariant.
Tensor Algebra
19
Solution
As A i and B i are the components of a contravariant and covariant tensor of rank one respectively,
then according to the law of the transformation
x i p
x q
A
Bq
=
and
=
Bi
A
x p
x i
i
= qp A p Bq
p
A i Bi = A B p
x i x q p
x k x l x t rs
kl
A
Bt
q and Bm =
x p x j
x r x s x m
...(1)
Put k = j then
Bmjl
x j x l x t rs
Bt
x r x s x m
...(2)
x i x q x j x l x t p rs
Aq Bt
x p x j x r x s x m
x i x l x t q p rs
r Aq Bt
x p x s x m
x i x l x t p qs
Aq B t
=
x p x s x m
since
x q x j
x q
=
= qr
x j x r
x r
since
qr Btrs = Btqs
This is the law of transformation of a mixed tensor of rank three. Hence Aij B mjl is a mixed tensor
of rank three.
20
1
n(n + 1) different components in n2
dimensional space.
Proof: Let A ij be a symmetric tensor of rank two. So that Aij = A ji .
A11
21
A
The component of A ij are A31
M
A n1
A12
A 22
A32
M
An2
A13
A 23
A33
M
An3
L
L
L
L
L
A1n
A2n
A 3n
M
A nn
i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn are
different. Thus remaining components are (n2 n). In which A 12 = A 21, A 23 = A 32 etc. due to symmetry.
So, the remaining different components are
1 2
(n n) . Hence the total number of different
2
components
= n+
1 2
1
(n n) = n(n + 1)
2
2
THEOREM 2.11
1
n(n 1) different non-zeroo
2
components.
Proof: Let A ij be a skew-symmetric tensor of order two. Then Aij = A ji .
Tensor Algebra
21
0
21
A
ij
The components of A are A 31
M
A n1
[Since A
ii
A12
0
A 32
M
An 2
A13
A 23
0
M
L
L
L
L
An3 L
A1n
A2n
A3n
M
0
i.e., A ij will have n2 components. Out of these n2 components, n components A 11, A 22, A 33, ..., A nn
are zero. Omitting there, then the remaining components are n2 n. In which A 12 = A 21, A 13 = A 31
1 2
etc. Ignoring the sign. Their remaining the different components are (n n) .
2
1
Hence the total number of different non-zero components = n(n 1)
2
Note: Skew-symmetric tensor is also called anti-symmetric tensor.
THEOREM 2.12 A covariant or contravariant tensor of rank two say Aij can always be written as the
sum of a symmetric and skew-symmetric tensor.
Proof: Consider a covariant tensor A ij. We can write A ij as
Aij =
1
1
( A + A ji ) + ( Aij A ji )
2 ij
2
Sij =
1
1
( Aij + A ji ) and Tij = ( Aij A ji )
2
2
S ji =
1
( A + Aij )
2 ji
Now,
S ji = Sij
So, Sij is symmetric tensor..
and
Tij =
1
( A + A ji )
2 ij
1
( A Aij )
2 ji
1
= ( Aij A ji )
2
T ji =
T ji = Tij
or
Tij = T ji
So, Tij is Skew-symmetric Tensor..
22
EXAMPLE 9
j k
If = a jk A A . Show that we can always write = b jk A j Ak where b jk is symmetric.
Solution
As given
= a jk A j A k
Interchange the indices i and j
...(1)
= ak j A k A j
...(2)
1
(a jk + akj ) A j A k
2
= b jk A j A k
where b jk =
1
(a + akj )
2 jk
b jk =
bkj
bkj = b jk
So, b jk is Symmetric.
EXAMPLE 10
Ti T j
If Ti be the component of a covariant vector show that j i
x
x
symmetric covariant tensor of rank two.
Solution
As Ti is covariant vector. Then by the law of transformation
Ti =
x k
Tk
x i
Tensor Algebra
23
Ti
j = x j
x
x k
x i Tk
2 x k
x k Tk
T
+
=
k
x j x i
x i x j
T j
x j
2 xk
x k x l Tk
T
+
k
x j x i
x i x j x l
...(1)
Similarly,
T j
2 xk
x k x l Tk
T
+
k
x i
x i x j
x j x i x l
Interchanging the dummy indices k & l
T j
x i
2 x k
x k x l Tl
T
+
k
x i x j
x i x j x k
...(2)
=
x i x j x l x k
x j x i
This is law of transformation of covariant tensor of rank two. So,
Ti T j
are component of
x j x i
To show that
is Skew-symmetric tensor..
x j x i
Let
Ti T j
Aij =
x j x i
T j Ti
A ji =
xi x j
T
j
i
= x j x i
A ji = Aij
or
Aij = A ji
Ti T j
is Skew-symmetric.
x j x i
Ti T j
So,
are component of a Skew-symmetric covariant tensor of rank two.
x j x i
So, Aij =
24
x p x q
x i x j
A( p, q, r )B r
...(1)
x p x q
x i x j
A ( p , q, r )
x p x q x r
x i x j x k
x p x q x r
x r
x k
Bk
A( p, q , r )B k
A( p , q , r )
x i x i x k
As B k is arbitrary..
So, A(i , j , k ) is covariant tensor of rank three.
EXAMPLE 12
If A (i, j, k)A iB jCk is a scalar for arbitrary vectors A i, B j, Ck. Show that A(i, j, k) is a tensor of
type (1, 2).
Solution
Let X and Y be two coordinate systems. As given A(i , j, k ) A i B j C k is scalar. Then
i
A (i, j, k ) A B C k = A( p, q, r ) A B C r
...(1)
Tensor Algebra
25
A (i, j , k ) A B C k = A( p, q, r)
x p i
A
x i
x q j
B
=
x j
x r k
C
=
x k
Ap =
Bq
Cr
x p xq x k
x i x j xr
Ai B jCk
As A i , B j , C k are arbitrary..
Then
x p xq x k
A( p, q, r )
A ( i, j , k ) =
xi x j xr
So, A(i, j, k) is tensor of type (1, 2).
2.16
Consider a covariant symmetric tensor Aij of rank two. Let d denote the determinant Aij with the
elements Aij i.e., d = Aij and d 0 .
Now, define A ij by
A ij =
d
A is a contravariant symmetric tensor of rank two which is called conjugate (or Reciprocal) tensor
of Aij .
ij
THEOREM 2.13 If Bij is the cofactor of Aij in the determinant d = |Aij| 0 and Aij defined as
A ij =
Bij
d
(i)
ki
Aij Bij = d
Aij
Bij
d
=1
Aij Aij = 1,
given
Aij =
Bij
d
26
(ii)
Aij B kj = 0
Aij
B kj
d
= 0,
Aij Akj = 0
d 0
if i k
Aij Akj = ik
i.e.,
2.17 RELATIVE TENSOR
i i ...i
If the components of a tensor A j11 2j 2 ...rj s transform according to the equation
2 ...k r
Alk11l 2k...
ls
x
=
x
Aij11i 2j 2......i rj s
x k1 x k2
x k r x j1 x j2
x js
x i1 x i 2
xi r x l r x l 2
x l s
x
is the Jacobian of transformation. If
x
= 1, the relative tensor is called a tensor density. If w = 0 then tensor is said to be absolute.
i i ...i
Hence A j11 2j 2 ...rj r is called a relative tensor of weight , where
MISCELLANEOUS EXAMPLES
1. Show that there is no distinction between contravariant and covariant vectors when we
restrict ourselves to transformation of the type
i m
i
x i = am x + b ;
where a's and b's are constants such that
a ir ami = rm
Solution
Given that
or
i m
i
x i = am x + b
a im x m = x i bi
...(1)
...(2)
or
i i
i i
xs = as x b as
..(3)
Tensor Algebra
27
x i
a is
s =
x
The contravariant transformation is
...(4)
x i s
A = a is A s
=
A
x s
The covariant transformation is
...(5)
x s
As = a is As
...(6)
x i
Thus from (5) and (6), it shows that there is no distinction between contravariant and
covariant tensor law of transformation
Ai =
2. If the tensors aij and g ij are symmetric and u i , v i are components of contravariant vectors
satisfying the equations
(aij kgij )u i = 0, i, j = 1, 2,...,n
( aij k'gij )v i = 0, k k .
Prove that gij u i v j = 0, a ij u i v j = 0.
Solution
The equations are
(aij kgij )u i = 0
...(1)
(aij k gij )v i = 0
j
...(2)
j
(k k ) g ij u v = 0
g ij u i v j = 0 since k k k k 0
Multiplying (1) by v j , we get
a ij v j u i kgij u i v j = 0
a ij u i v j = 0 as g ij u i v j = 0.
Proved.
28
i k
( j l + l j ) Aik = 0
Solution
Given Aij is a Skew-symmetric tensor then Aij = A ji .
Now,
i
i k
i k
( j l + l j ) Aik = 0
as A jl = Alj
4. If aij is symmetric tensor and bi is a vector and a ij bk + a jk bi + a kib j = 0 then prove that
a ij = 0 or bk = 0 .
Solution
The equation is
aij bk + a jk bi + akib j = 0
aij bk + a jk bi + akib j = 0
x p x q
x r
x p x q x r
x p x q
x r
b
+
a
b
+
a
b
=0
r
pq
r
pq
r
x i x j x k
x j x k
x i
x k x i
x j
x p x q x r x p x q x r x p x q x r
+ j
+
a pq br i
j
k
x x k x i x k x i x j
x x x
= 0
a pq br = 0 a pq = 0 or br = 0
a ij = 0 or bk = 0
5. If a mn x x = bmn x x for arbitrary values of x r , show that a(mn) = b(mn) i.e.,
a mn + a nm = bmn + bnm
m n
m n
If a mn and bmn are symmetric tensors then further show the a mn = bmn .
Solution
Given
a mn x m x n = bmn x m x n
(a mn bmn )x m x n = 0
Tensor Algebra
29
EXERCISES
1. Write down the law of transformation for the tensors
(i) Aij
(ii) Bkij
ijk
(iii) Clm
2. If Arpq and Bts are tensors then prove that Arpq Bts is also a tensor..
3. If Aij is a contravariant tensor and Bi is covariant vector then prove that Aij Bk is a tensor of rank
three and Aij Bj is a tensor of rank one.
4. If Ai is an arbitrary contravariant vector and Cij Ai Aj is an invariant show that Cij + Cji is a covariant
tensor of the second order.
5. Show that every tensor can be expressed in the terms of symmetric and skew-symmetric tensor.
6. Prove that in n-dimensional space, symmetric and skew-symmetric tensor have
n
n
(n + 1) and (n 1)
2
2
ji
= 0 holds
w.r.t to a basis then prove that either f = g and U ij is skew-symmetric or f = g and U ij is symmetric.
i k
i k
8. If Aij is skew-symmetric then ( B j Bl + B l B j ) Aik = 0 .
ij
30
10. If Arpq is a tensor of rank three. Show that Arpr is a contravariant tensor of rank one.
ij
k
k
11. If a k i j is a scalar or invariant, i , j , are vectors then akij is a mixed tensor of type (2, 1).
h i h k
12. Show that if a hijk = 0 where i and i are components of two arbitrary vectors then
13. Prove that Aij Bi C j is invariant if Bi and C j are vector and Aij is tensor of rank two.
14. If A(r, s, t) be a function of the coordinates in n-dimensional space such that for an arbitrary vector
Br of the type indicated by the index a A(r, s, t)Br is equal to the component Cst of a contravariant
tensor of order two. Prove that A(r, s, t) are the components of a tensor of the form Arst .
15. If Aij and Aij are components of symmetric relative tensors of weight w. show that
A
ij
ij
= A
x
x
w 2
and
Aij = Aij
x
x
w+ 2
16. Prove that the scalar product of a relative covariant vector of weight w and a relative contravariant
vector of weight w is a relative scalar of weight w + w .
CHAPTER 3
. . . . . . . . .. . . . . . . . . . . . . . . . .
+ g n1 dx n dx1 + g n 2 dx n dx 2 + + g nn (dx n )2
ds 2 = g ij dx i dx j (i, j = 1,2,...n)
...(1)
...(1)
32
So,
If
dxi =
x i 1 x i 2
x i n
dx
+
dx
+
+
dx
x i
x 2
x n
dxi =
x i k
dx
x k
x i k
x j l
j
dx
x
and
d
x
=
x k
x l
dxk
x j l
dx
x l
x j k l
x dx
x l
x j k l
dx dx
x l
g kl g ij x x
x k x l
or
or
g kl gij
k l
dx dx = 0
x i x j
= 0 as dx k and dx l are arbitrary..
x k x l
g kl = g ij
x j x j
x k x l
g ij = g kl
x k x l
x i x j
...(2)
33
1
1
(g + g ji ) + ( gij g ji )
2 ij
2
g ij = Aij + Bij
1 g +g
( ij
ji ) = symmetric
2
1
Bij = (g ij g ji ) = Skew-symmetric
2
Aij =
where
Now,
(4)
Bij dxi dx j = 0
So, g ij is symmetric since Aij is symmetric. Hence g ij is a covariant symmetric tensor of rank
two. This is called fundamental Covariant Tensor.
THEOREM 3.2 To show that g ij dx i dx j is an invariant.
Proof: Let x i be coordinates of a point in X-coordinate system and x i be coordinates of a same point
in Y-coordinate system.
Since g ij is a Covariant tensor of rank two.
Then,
g ij = g kl
x k x 1
x i x j
34
x k x l
=0
x i x j
g ij g kl
k
l
gij g kl x x
x i x j
i j
dx dx = 0
( g ij dx dx ) = g kl
x k x l i j
dx dx
x i x j
= g kl
x k i x l
dx
dx j
x i
x j
g ij dx i dx j = g kl dx k dx l
So, g ij dx i dx j is an ivariant.
3.2 CONJUGATE METRIC TENSOR: (CONTRAVARIANT TENSOR)
The conjugate Metric Tensor to g ij , which is written as g ij , is defined by
Bij
g ij = g (by Art.2.16, Chapter 2)
where Bij is the cofactor of g ij in the determinant g = g ij 0 .
By theorem on page 26
Aij Akj = ik
g ij g kj = ik
So,
Note
(i)
(ii)
EXAMPLE 1
Find the Metric and component of first and second fundamental tensor is cylindrical coordinates.
Solution
Let (x1, x2, x3) be the Cartesian coordinates and (x1, x2 , x3 ) be the cylindrical coordinates of a
point. The cylindrical coordinates are given by
x = r cos , y = r sin , z = z
So that
x1 = x, x 2 = y, x 3 = z and x 1 = r , x 2 = , x 3 = z
...(1)
Let g ij and g ij be the metric tensors in Cartesian coordinates and cylindrical coordinates
respectively.
35
12
2 2
3 2
ds 2 = (dx ) + (dx ) +(dx )
...(2)
i
j
ds 2 = g ij dx dx
But
...(3)
x i x j
, since g ij is Covariant Tensor of rank two. (i, j = 1, 2, 3)
x i x j
for i = j = 1.
2
g11
x1
x 2
x 3
= g11 1 + g 22 1 + g 33 1
x
x
x
x
y
z
g11 = g11 + g 22 + g 33
r
r
r
and g11 = g 22 = g 33
Put i = j = 2.
x 1
g 22 = g11 2
x
2
+ g 22 x
x 2
g 22
3
+ g 33 x
x 2
x
y
z
+ g 22
+ g 33
= g11
36
Since g11 = g 22 = g 33 = 1
z
x
y
=0
= r sin ,
= r cos,
g 22 = ( r sin )2 + (r cos )2 + 0
= r 2 sin 2 + r 2 cos 2
g 22 = r 2
Put i = j = 3.
2
g 33
x1
x 2
x 3
g
+
g
+
g
= 11 3
22
33
3
3
x
x
x
2
x
y
z
= g11 + g 22 + g 33
z
z
z
Since
x
y
z
= 0,
= 0,
= 1 . So, g 33 = 1 .
z
z
z
So,
g11 = 1, g 22 = r 2 , g 33 = 1
and
( )
1
ds 2 = g11 dx
( )
+ g 22 dx 2
( )
+ g 33 dx 3
g ij
since
2
= g 21 g 22 g 23 = 0 r 0
g 31 g32 g 33 0 0 1
1
g = g ij = 0
0
2
g= r
0
r2
0
0
0
1
37
and
g 11 =
cofactor of g11 in g
g
g 11 =
B11 r 2
= 2 =1
g
r
g 22 =
B12 1
= 2
g
r
g 33 =
B33 r 2
= 2 =1
g
r
Bij
g
g 12 = g13 = g 21 = g 23 = g 31 = g 32 = 0
0 0
1
0 .
r2
0 1
EXAMPLE 2
Find the matrix and component of first and second fundamental tensors in spherical coordinates.
Solution
Let ( x1 , x 2 , x 3 ) be the cartesian coordinates and ( x 1 , x 2 , x 3 ) be the spherical coordinates of a
point. The spherical coordinates are given by
x = r sin cos , y = r sin sin , z = r cos
So that x1 = x, x 2 = y , x 3 = z and x 1 = r , x 2 = , x 3 =
Let g ij and g ij be the metric tensors in cartesian and spherical coordinates respectively..
The metric in cartesian coordinates is given by
2
2
2
ds 2 = dx + dy + dz
( ) + (dx ) + (dx )
ds 2 = dx1
But
2 2
3 2
i
j
ds 2 = g ij dx dx ; (i, j = 1, 2, 3)
38
x i x j
x i x j
g11
x 1
x 2
x 3
g
+
g
+
g
= 11 1
22
33
1
1
x
x
x
g11
x
y
z
= g11 + g 22 + g 33
r
r
r
x 1
= g11 2
x
2
+ g 22 x
x 2
3
+ g 33 x
x 2
x
y
z
g 22 = g11 + g 22 + g33
since g11 = g 22 = g 33 = 1
x
y
z
= r cos cos ,
= r cos sin ,
= r sin
39
Put i = j = 3
2
g 33
x1
x 2
x 3
g
+
g
+
g
= 11 3
22
33
3
3
x
x
x
g 33
x
y
z
= g11 + g 22 + g33
since g11 = g 22 = g 33 = 1
x
y
z
= r sin sin ,
= r sin cos ,
=0
and
and
( )
1
ds 2 = g11 dx
( )
+ g 22 dx 2
( )
+ g 33 dx 3
ds 2 = dr 2 + r 2 d 2 + r 2 sin 2 d 2
(ii) The Metric tensor or first fundamental tensor is
g ij
and
g g g = 0 r 2
0
= 21 22 23
g 31 g32 g 33 0 0 r 2 sin 2
g = g ij
1
=0
0
0
r2
0
0
0
= r 4 sin 2
r 2 sin 2
(iii) The cofactor of g are given by B11 = 1, B22 = r 2 , B33 = r 2 sin 2 and B12 = B21 =
B31 = B13 = B23 = B32 = 0
The second fundamental tensor or conjugate tensor is g ij =
Bij
g
40
g 11 =
r 4 sin 2
r 4 sin 2
=1
=
g 11
g
22
B 22 r 2 sin 2
= 4 2
=
g
r sin
g 22 =
g
33
1
r2
B33
r2
=
=
g
r 4 sin 2
g 33 =
1
r sin 2
2
12
13
21
31
32
and g = g = g = g = g = 0
Hence the fundamental tensor in matrix form is
g
21
= g
g 31
11
ij
12
g 22
g 32
0
g
0
23
g =
g 33
0
0
1
r2
13
r 2 sin 2
EXAMPLE 3
If the metric is given by
( ) ( )
2
ds 2 = 5 dx 1 + 3 dx 2
( )
+ 4 dx 3
(i, j = 1, 2, 3)
( )
1
ds 2 = g11 dx
So,
41
1 2
2 2
3 2
1
2
ds 2 = g11 (dx ) + g 22 (dx ) + g33 (dx ) + 2 g12 dx dx
+ 2 g 23 dx 2 dx 3 + 2 g13 dx1 dx 3
...(1)
g = g ij = g 21 g 22 g 23 = 3 3 2 = 4
g 31 g 32 g33
0 2 4
(ii) Let Bij be the cofactor of g ij in g.
Then
B11 = Cofactor of g11 =
3 2
=8
2 4
B22 = Cofactor of g 22 =
5 0
= 20
04
B33 = Cofactor of g 33 =
5 3
=6
3 3
3 2
= 12 = B 21
0 4
B 23
Since
3 3
= 6 = B31
0 2
= Cofactor of g 23 =
g ij =
5 3
0
= 10 = B 32
Bij
g
We have
B
8
g 11 = 11 = = 2; g 22 = 5, g 33 = 3 , g 12 = g 21 = 3, g 13 = g 31 = 3 , g 23 = g 32 = 5
g
4
2
2
2
...(2)
42
Hence,
g ij
3
3
2
5
3
5
2
3 5
3
2 2
2
2
If s be arc length of the curve between the points P1 and P2 on the curve which correspond to
the two values t1 and t 2 of the parameter t.
s =
P2
P1
ds =
t2
t1
12
i
j
gij dx dx dt
dt dt
NULL CURVE
dx i dx j
= 0 along a curve. Then s = 0. Then the points P1 and P2 are at zero distance, despite
dt dt
of the fact that they are not coincident. Such a curve is called minimal curve or null curve.
If g ij
EXAMPLE 4
A curve is in spherical coordinate xi is given by
2
1 1
3
2
x1 = t, x = sin and x = 2 t 1
t
Solution
In spherical coordinate, the metric is given by
1 2
1 2
2 2
1
2 2
3 2
ds 2 = (dx ) + ( x ) (dx ) + ( x sin x ) (dx )
43
2
1 1
, x3 = 2 t 2 1
x1 = t, x = sin
t
given
dx 2 =
dt
t t 1
2
, dx 3 =
2t
t 2 1
dt
2t
dt
1 1
+t
+ t sin sin
dt
t t 2 1
t t 2 1
ds = (dt )
1 2
1
1
2t dt
2 dt , dx 3 = 2 t 2 1
2
1 t
1
t
dx 1 = dt , dx 2 =
dt 2
4t 2
2
dt
+
+
(dt )2
=
ds
2
2
t 1 t 1
2
5t 2
2
ds 2 = t 2 1 dt
ds =
t 1
2
dt
t2
t1
3.4
ds =
5 t2 1
=
dt =
2 12
t2 1
1
t
15 units
ASSOCIATED TENSOR
i i ... i
A tensor obtained by the process of inner product of any tensor A j11 2j 2 ... rj s with either of the fundamental
44
A=
or
g ij Ai A j
i j
A 2 = g ij A A
i
Also, A2 = A j A j as g ij A = A j
i.e., square of the magnitude is equal to scalar product of the vector and its associate.
g ij Ai A j
A 2 = g ij Ai A j
or
A vector of magnitude one is called Unit vector. A vector of magnitude zero is called zero vector
or Null vector.
3.6 rSCALAR
r r
r PRODUCT OF TWO VECTORS
Let A and B be two vectors. Their scalar product is written as A B and defined by
r r
A B = Ai Bi
r r
i
i j
j
Also,
A B = A B i = g ij A B since Bi = g ij B
r r
i
ij
i
ij
A B = Ai B = g Ai B j since B = g B j
Thus
i.e.,
r r
i
i j
2
A A = A Ai = g ij A A = A
r
A = A = g ij Ai A j
Angle
two vectors
r between
r
Let A and B be two vectors. Then
r r
r r
A B = A B cos
r r
g ij A i B j
AB
cos = r r =
AB
g ij A i A j g ij B i B j
since
r
A =
r
g ij Ai A j ; B =
g ij B i B j
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
45
r
r
THEOREM 3.3 The necessary and sufficient condition that the two vectors A and B at 0 be orthogonal
r r
if g ( A, B ) = 0
r
r
Proof: Let be angle between the vectors A and B then
r r
r r
A B = A B cos
r r
or
A B = AB cos
g ij A i B j = AB cos
cos =
g ij Ai B j
...(1)
AB
r
r
If A and B are orthogonal then = cos = 0 then from (1)
2
i j
g ij A B = 0
r r
r r
i j
g A, B = 0 since g A, B = g ij A B
( )
( )
i j
Conversely if g ij A B = 0 then from (1)
cos = 0 = .
2
r
r
So, two vectors A & B are orthogonal.
r
r
r
r
Note: (i) If A and B be unit vectors. Then A = B = 1. Then
Proved.
r r
cos = A B = gij A i B j
r
r
(ii) Two vectors A and B will be orthogonal if angle between them is i.e., = then
2
2
cos = cos = = 0
2
46
= l 2 + 2lm cos + m 2
Since
i j
2
g ij A i A j = A 2 = 1; g ij B B = B = 1.
and
r
r
r
g ij A i B j = cos; as A & B are unit vector i.e., A = 1 A 2 = 1 .
(l + m cos )2 + m 2 m 2 cos 2
(l + m cos) 2 + m 2 (1 cos2 ) 0
This inequality holds for the real values of l & m.
if
1 cos2 0
cos 2 1
cos 1
Proved.
...(1)
ij
B 2 = g Ai A j
...(2)
and
From equation (1)
A 2 = (gij Ai ) A j
A2 = Aj A j
...(3)
...(4)
47
since g ij Ai = A j
from (3) and (4)
2
A2 = B
A=B
i
So, magnitude of Ai and A are equal.
...(2)
B i = dx i = (0,...0, x q , 0...0)
...(3)
cos =
g ij A i B j
g ij A i A j
g ij B i B j
g pq A p B q
g pp A p A p g qq B q B q
g pq A p B q
g pp g qq A p B q
g pq
g pp g qq
g ij
g ii g jj
...(4)
48
=0
2
g ij = 0
and
(x i ) = constant
...(1)
( x i ) = constant
...(2)
...(3)
is orthogonal to dx i . Hence
is normal to = constant, since dx i is
i
x
x i
tangential to hypersurface (1).
This shows that
is normal to the hypersurface (2). If is the angle between the hypersurface (1)
x i
and (2) then is also defined as the angle between their respective normals. Hence required angle
is given by
Similarly
x i x j
g ij i
g ij i
j
x x
x x j
g ij
cos =
...(4)
49
If we take
and
= x p = constant
...(5)
= x q = constant
The angle between (5) and (6) is given by
...(6)
x p x q
x i x j
q
q
x p x p
ij x x
g ij i
g
x x j
x i x j
g ij
cos =
g ij ip qj
g ij ip qj g ij qi qj
g pq
cos =
g pp g qq
...(7)
cos ij =
g ii g jj
...(8)
cos ij = 0
50
THEOREM 3.6 To find the fundamental tensors g ij and g ij in terms of the components of the unit
tangent e h ( h = 1, 2,... n ) to an orthogonal ennuple.
Proof: Consider n unit tangents e ih (h = 1, 2,...n) to conguence e h ( h = 1, 2,... n ) of an orthogonal ennuple
in a Riemannian Vn . The subscript h followed by an upright bar simply distinguishes one congruence
from other. It does not denote tensor suffix.
The contravariant and covariant components of eh | are denoted by eh | and eh |i respectively..
Suppose any two congruences of orthogonal ennuple are eh | and ek | so that
g ij ehi |ekj| = hk
...(1)
ehi |ek |i = hk
from (1),
g ij ehi | ekj| = 0
g ij ehi | ehj| = 1
and
We define
ehi | =
cofactorof eh |i in determinant eh |i
e h|i
e e
i
h| h | j
h =1
= ij
...(2)
Multiplying by e jk
n
e e
i
h| h| j
h =1
g j k = ij g jk
e e
i k
h | h|
or
h =1
= g ik
...(3)
e e
i
h | h | j g ik
h =1
or
i
= j g ik
g jk =
h |k eh | j
...(4)
g ij =
h |i
h =1
e h| j
...(5)
51
g ij =
i
h|
e hj|
...(6)
h =1
u =
C e
i
h h|
...(7)
h =1
Then
n
u i ek |i =
C e e
i
h h | k |i
h =1
or
C
h
h
k
= Ck
h =1
C k = u i ek |i
...(8)
C k = projection of u i on ek |i
i.e.,
Using (8), equation (7) becomes
n
ui =
u e
j
i
h| j ek |
h =1
Now,
u 2 = u iu i =
C e C e
i
h h|
k k |i
from (7)
C C e e
i
k h | k| i
h ,k
C C
h
h
k
h ,k
C C
h
(C
)2
h
n
u2 =
h =1
52
Miscellaneous Examples
1. If p and q are orthogonal unit vectors, show that
( g hj g ik g hk gij ) p h q i p j q k = 1
Solution
Since p and q are orthogonal unit vectors. Then
g ij pi q j = 0, p 2 = q 2 = 1 .
Now,
h
j i k
h k i j
( g hj g ik g hk gij ) p h q i p j q k = g hj g ik p p q q g hk g ij p q q p
= ( g hi p h p j ) ( gik q i q k ) (g hk p j q k ) ( gij q i p j )
= p2.q2 0.0
=1.1
= 1 (since g hi p h p j = 1 & g hk p h q k = 0 )
2. If is the inclination of two vectors A and B show that
sin =
2
(g hi g ik g hk g ij ) A h A j B i B k
g hj gik A h A j B j B k
Solution
If be the angle between the vectors A and B then
g ij A j B i
cos =
gij A i A j g ik B i B k
sin 2 = 1
( gij B i A j ) (g hk A h B k )
( g hj A h A j ) ( g ik B i B k )
(g hj g ik g hk g ij ) A h A j B i B k
g hj g ik A h A j B i B k
3. If X ij are components of a symmetric covariant tensor and u, v are unit orthogonal to w and
satisfying the relations
( X ij g ij )u i + w j = 0
( X ij g ij ) v i + w j = 0
where prove that u and v are orthogonal and that
53
X ij u i v j = 0
Solution
Suppose X ij is a symmetric tensor. Since u i , v j are orthogonal to wi then
u i wi = 0
...(1)
v i wi = 0
...(2)
( X ij gij )u i + w j = 0
...(3)
( X ij gij )vi + w j = 0
where .
...(4)
given
Multiply (3) & (4) by v j , u j respectively and using (1) and (2), we have
( X ij g ij )u iv j = 0
...(5)
( X ij g ij )v iu j = 0
...(6)
Interchanging the suffixes i & j in the equation (6) and since g ij , X ij are symmetric, we get
( X ij g ij )u iv j = 0
Subtract (6) & (7) we get
... (7)
( )g iju i v j = 0
Since and 0.
Hence,
g ij u i v j = 0
So, u and v are orthogonal.
Using (8) in equation (5) & (6), we get
...(8)
X ij u i v i = 0
Proved.
Solution
Since g ij is a symmetric tensor of rank two. Then
g ij =
x k x l
g kl
x i x j
x k
=
x i
x l
x j
g kl
x
= J (Jacobian)
x
g kl = g
&
g ij = g
54
So,
g = gJ 2
or
J=
g
g
g 1 2
d x d x ... d x n
g
g d x1d x 2... d x n
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
1
1
1
, g 33 =
, g 22 =
g11
g 22
g 33
8. If is the angle between the two vectors Ai and B i at a point, prove that
2
sin =
(g hi gik g hk gij ) A h Ai B j B k
ghi g jk Ah Ai B j B k
9. Show that the angle between two contravariant vectors is real when the Riemannian metric is positive
definite.
CHAPTER 4
[ij, k ] =
1 gi k g j k g i j
+
k , (i, j, k = 1, 2,...n )
2 x j
x i
x
called Christoffel 3-index symbols of the first kind.
...(1)
k
= g k l [ij, l ]
i j
...(2)
and
called Christoffel 3-index symbols of second kind, where g i j are the components of the metric Tensor
or fundamental Tensor.
There are n distinct Christoffel symbols of each kind for each independent g i j . Since g i j is
1
n (n + 1) independent components. So, the number of
2
1
1
independent components of Christoffels symbols are n n(n + 1) = n 2 (n + 1) .
2
2
symmetric tensor of rank two and has
k
THEOREM 4.1 The Christoffel's symbols [ij, k ] and are symmetric with respect to the indices i
i j
and j.
Proof: By Christoffels symbols of first kind
[ij, k ] =
g ij
1 g ik g jk
j +
k
i
2 x
x
x
[ ji, k ] =
1 g jk gik g ji
+
k
2 x i
x j
x
56
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.
[ij, m] =
(ii)
[ik , j ] + [ jk , i] =
k
g km
i j
g ij
x k
g ij
i
j
= g jl g im
l
k
x
m k
Proof: (i) By Christoffels symbol of second kind
(iii)
k
= g k l [ij, l ]
i j
Multiplying this equation by g k m , we get
k
g k m = g k m g k l [ij, l ]
i j
= lm [ij , l ] as g k m g k l = lm
k
g km = [ij, m]
i j
(ii) By Christoffels symbol of first kind
and
[ik , j ] =
1 g k j g i j g i k
2 x i x k x j
...(1)
[ jk ,i ] =
1 g ki g ji g jk
2 x j x k
x i
...(2)
[ik , j ] + [ jk , i ] =
1 g i j g ji
+
g = g ji
2 x k x k since ij
[ik , j ] + [ jk , i] =
57
1 g ij g ij
2
=
2 x k x k
(iii) Since g ij g lj = li .
Differentiating it w.r.t. to x k , we get
g ij
glj
x k
+ glj
g ij
=0
x k
g lj
x k
+ g lm g lj
g lm glj
mj
g ij
=0
x k
g
g ij
g ij g lm ljk
k =
x
x
g lj
g ij
ij lm
g
g
{
[
lk
,
j
]
+
[
jk
,
l
]
}
= [lk , j ] + [ jk , l ] .
=
since
x k
x k
g im
lm
ij
ij
lm
= g g [lk , j ] g g [ jk , l ]
x k
g im
i
m
= g lm g ij
k
x
l k
j k
Interchanging m and j, we get
i
j
g ij
lj
im
g
=
x k
l k
m k
or
g ij
i
j
g ij g im
as g lj = g jl
k =
x
l k
m k
Proved.
( )
i log g
THEOREM 4.3 To show that i j =
x j
Proof: The matrix form of g ik is
g ik
21 g 22 ... G2 n
=
M
g n1 g n 2 ... g nn
58
and
g = gik
g11
g 21
=
M
g n1
g n 2 ... gnn
But
g ik g = lk
Take
l = k.
il
g ik g ik = kk = 1
g ik = [g ik ]1 =
Gik
(Theorem 2.13 , Pg 25)
g
g = g ik Gik
g ik x j
x
= Gik
gik
x j
But Gik = gg ik
g
ik gik
= gg
j
x j
x
g
1 g
g ik ikj
j =
g x
x
1 g
g ik
= g ik { [ jk , i] + [ij, k ] } as
= [ jk , i] + [ij, k ]
g x j
x j
ik
ik
= g [ jk , i] + g [ij , k ]
k i
= +
j k i j
...(1)
59
i i
1 g
+ as k is dummy indices
j =
g x
j i j i
i
1 g
= 2
j
g x
j i
i
1 g
=
j
2 g x
j i
log( g )
x
j i
Proved.
EXAMPLE 1
If g ij 0 show that
x j
= x j [ik , ] i k
i k
( [j, ] + [j , ] )
Solution
By Christoffels symbol of second kind
= g [ik , ]
i
k
Multiplying it by g , we get
g = g g [ik , ]
i
k
g = [ik , ] as g g = 1
i k
[ik , ]
g
=
j
x j
x
i k
g
since
g
+
x j i k
ik,
j =
i k x
x j
g
x j
= [j, ] + [j, ]
+ ([j, ] + [j , ]) =
[ik , ]
j
i
k
i
k
x
x j
60
x j
= x j [ik , ] i k ( [i, ] + [j, ] )
i k
Solved.
EXAMPLE 2
k
Show that if g ij = 0 for i j then (i) = 0 whenever i, j and k are distinct.
i j
i 1 log gii
i 1 log g ii
i
1 g jj
=
=
(ii) =
(iii)
(iv)
i
j
2 gii x i
i i 2 x
i j 2 x
j j
Solution
The Christoffels symbols of first kind
[ij, k ] =
(a) If i = j = k
The equation (1) becomes
[ii, i] =
(b) If i = j k
The equation (1) becomes
[ii, k ] =
1 g jk g ik g ij
2 x i
x j x k
1 gii
2 x i
1 g i k gi k g i i
+ i k
2 x i
x
x
Since g ik = 0 as i k (given)
1 g ii
1 g jj
or [ jj, i] =
k
2 x
2 xi
[ii, k ] =
[ij,i ] =
1 g ji g ii g ij
+ j i
2 x i
x
x
(c) i = k j
1 g ii
, as g ij = 0, i j
2 x j
(d) i j k
[ij, k ] = 0
as g ij = 0 , g j k = 0 , i j k
...(1)
61
k
=0
i j
i
= g ii [ii, i ]
i i
(ii)
= g ii
=
1 gii
from (a)
2 x i
1
1 g ii
g ii =
i as
gii
2 gii x
i
1 log g ii
=
2 x i
i i
(iii)
i=k j
i
ii
= g [ij, i]
i j
1
1
ii
= g [ij, i] as g =
ii
gii
i
1 log gii
=
2 x j
i j
(iv)
j=ki
i
ii
= g [ jj, i]
j
j
1 1 g j j
=
from (b)
g ii 2 x i
1 g j j
i
= 2g
i
i i x
j j
EXAMPLE 3
If ds 2 = dr 2 + r 2 d 2 + r 2 sin 2 d 2 , find the values of
1
3
[22, 1] and [13, 3], (ii)
and
2 2
1 3
Solution
(i)
Solved.
62
Also,
22
g 11 = 1, g =
1
1 g 33 =
,
2
r sin 2
r2
(See Ex. 2, Pg. 39, and g ij = 0, for i j. )
[ij, k ] =
1 g jk g ik g ij
+
, i, j, k = 1, 2,3
2 x i
x j x k
[22, 1] =
=
1 g 21 g 21 g 22
+ 2 1 Since g21 = 0.
2 x 2
x
x
1 0
0 r 2
2 + 2 1
2 x
x
x
1 r 2
= r
2 r
Taking i = 1, j = k = 3 in (1)
[13,3] =
=
[13, 3] =
r sin 2
(ii) Christoffel symbols of the second kind are given by
k
= g kl [ij, l] = g k 1 [ij ,1] + g k 2 [ ij ,2] + g k 3 [ ij,3]
i j
Taking k = 1, i = j = 2.
1
= r
2 2
...(1)
63
and
3
31
32
33
= g [13, 1] + g [13, 2] + g [13, 3]
1
3
=
1
[13, 3] Since g 31 = g 32 = 0
2
r sin
2
3
1
1
2
= 2 2 r sin =
1
3
r
r sin
1 gik g ik g ij
+
2 x i
x j x k
...(1)
x p x 2
g pq
x i x j
...(2)
x k
x p x q
x i x j g pq
x k
x p x q
x i x j
p
q g
pq
g pq + x x
i
j
x x x k
g i j
2 x p x q x p 2 x q
=
x k x i x j + x i x k x j
xk
p
q g
r
pq x
g pq + x x
x i x j x r x k
...(3)
2 x p x q x p 2 x q
+ k
= i k
j
x x i x j
x x x
r
q g
p
rq x
g pq + x x
x k x j x p x i
...(4)
64
and interchanging j, k and also interchange q, r in the last term of equation (3)
2 x p x q x p 2 x q
g ik
+ i
= j i
i
x j
x x k x j
x x x
p
r
q g
pr
g pq + x x x
i
k
j
q
...(5)
Substituting the values of equations (3), (4) and (5) in equation (1), we get
[ij, k ] =
1 2 x p x q
x p x q x r
2 i j k g pq + i
2 x x x
x x j x k
[ij, k ] =
2 x p x q
x p x q x r 1 g rp g qr g pq
g
+
+ p
pq
x i x j x k
x i x i x k 2 x q
x
x r
g rp g qr g pq
q +
x p
x r
x
2 x p x q
x p x q x r
g
+
[ pq, r]
...(6)
pq
x i x j x k
x i x i x k
It is law of transformation of Christoffel's symbol of the first kind. But it is not the law of
transformation of any tensor due to presence of the first term of equation (6).
So, Christoffel's symbol of first kind is not a tensor.
[ij, k ] =
k
in another coordinate system
i j
2 x p x q
x p x q x r
g pq + i
[ pq, r ] from (6)
i
j
l
x x x
x x j x l
is contravariant tensor of rank two.
[ij, l ] =
kl
As g
g kl =
x k x l st
g
x s x t
Now
g kl [ij, l ] =
x k x l st 2 x p x q
x k x l st x p x q x r
g
g
+
g
[pq, r]
pq
x s x t
x i x j x l
x s x t
x i x j x l
x l x q st 2 x p
x k
g
g
+
pq
x t x l x i x j
x s
x l x r x p x q st
x t x l x i x j g [ pq, r ]
x k
x s
x k 2 st 2 x p
x k r x p x q st
x l x q
g
g
+
g
[
pq
,
r
]
= qt
as
t
pq
t
s
i j
s
i
j
t
l
x
x x
x
x x
x x
x k 2 x p sq
x k x p x q sr
g
g
+
g [ pq, r ] as rt g st = g sr
pq
x s x i x j
x s x i x j
65
x p 2 x p s x k x p x 2 s
p + s
x s x i x j
x x i x j p q
s
sr
= sp and g [ pq, r ] = p q
g kl [ij, l ] =
Since g sq g pq
k
=
i j
x k 2 x s
x k x p x q
+
x s x i x j x s x i x j
p q
....(7)
It is law of transformation of Christoffels symbol of the second kind. But it is not the law of
transformation of any tensor. So, Christoffels symbol of the second kind is not a tensor.
Also, multiply (7) by
x s
, we get
x k
x s
dx k
Since
k
x s x k 2 x s
x s x k x p x q
+
=
x k x s x i x j x k x s x i x j
i j
p q
x s x k
= ss = 1
x k x s
x s
x k
k
2 xs
x p x q
+
=
x i x j x i x j
i j
x s
2 xs
=
x k
x i x j
p q
k x p x q s
i j i
j
x x p q
...(8)
x s x i x j
x s x i x j p q
i j
When coordinate x i be transformed to x i . Then
k x i x j x r
r
2 x k x r
+
=
u
v
k
x u x v x k
i j x x x
u v
...(1)
66
s x k x p x q x i x j x r
2 x s x k x i x j x r
+
= p q
x s x i x j x u x v x k x i x j x s x u x v x k
+
2 x k x r
x u x v x k
2 k
r
r s x p x q x r
x x
=
+
+
u
v
s
u
v
k
u v p q x x x
x x x
2 x s x r x i x j
x i x j x s x u x v
as
p
i
x x
...(2)
...(3)
x v
x s
x i
x i x s x i
x u + x i x v x u
2 s
= x
x u x v
2 x s x j x i
x s 2 x i
2 xs
+
=
x i x j x v x u
x i x u x v
x u x v
Mutiply (5) by
...(4)
x r
.
x s
2 x s x j x i x r
2 x i x s x r
2 x s x r
+
=
x i x j x v x u x s x u x v x s x s
x u x v x s
Replace dummy index i by k in second term on L.H.S.
2 x s x j x i x r
2 x k x r
2 x s x r
+
=
x i x j x v x u x s x u x v x k
x u x v x s
Using (5) in equation (2), we get
r
s x p x q x r
2 x s x r
+
=
u v
s
x u x v x s
u v
p q x x x
...(5)
...(6)
67
The equation (6) is same as the equation (1). This shows that if we make direct transformation
from x i to x i we get same law of transformation. This property is called that transformation of
Christoffel's symbols form a group.
4.3 COVARIANT DIFFERENTIATION OF A COVARIANT VECTOR
Let Ai and Ai be the components of a covariant vector in coordinate systems x i and x i respectively..
Then
x p
Ap
Ai =
x i
...(1)
k x p x q s
i
j
i j x x p q
2 xs
in equation (3), we have
x i x j
x s
Ai
= k
x j
x
k x p x q s
x p A p x q
A
+
s
x i x q x j
i j x i x j p q
k x s
x p x q
= k As i
x x j
i j x
k
Ai
x p x q
A
+
=
k
x j
x i x j
i j
s
x p x q A p
As + i
x x j x q
p q
A p
s
q As
x
p q
...(2)
...(3)
68
k
x p x q
Ai
A
=
k
x i x j
x j
i j
A p
s
x q As p q
...(4)
Ai
k
Ak
j
x
i j
...(5)
x p x q
Ap, q
x i x j
..(6)
or
...(1)
2 xs
x s k i x p x q s
=
x k i j
x i x j p q
x j x i
substituting the value of
2 xs
in the equation (1), we get
x j x i
A s
x s
=
A j
x k
A s A q
x s
=
A q x j
x k
k i x p x q s i x s A i
A i
A + i
x x j p q
x x j
i j
k i x p i x q
A i A
x
x j
i j
s x s A i
+ i
j
p q x x
69
Interchanging the dummy indices i and k in the first term on R.H.S. and put
A s x q
x s
=
x q x j
x i
x q
x j
A s
x s
p s
x q + A p q = x i
i
A i
+ Ak
j
x
k
x p i
A = A p we get
x i
i k x q p s x s A i
A j A
+ i
j
x
k j
p q x x
i
x i x q
=
j
x s x j
k A i
A + j
j
x
A s
p s
+
A
q
x
p q
...(2)
j
x
k j
Using (3), the equation (2) can be expressed as
A,i j =
...(3)
x p x q
Ap, q
...(4)
x i x j
It is law of transformation of a mixed tensor of rank two. Thus, Ai , j is a mixed tensor of rank
two. Ai , j is called covariant derivative of Ai with respect to x j .
A,ij =
x p x q
A pq
x i x j
...(1)
A pq
x k
x p x q
x i x j
A pq
x k
Aij
x p x q Apq x r
2 x p x q
x p 2 x q
+
A
+
A pq ...(2)
=
pq
x i x j x r x k x k x i x j
x i x k x j
x k
as
x p x q A pq
+ k
i
j
k
x x x
x
A pq x r
(since A pq components in x i coordinate)
x r x k
70
A pq
2 x p x q
2 x l x q
A
=
lq
x i x k x j
x i x k x j
A pq
2 x p x q
x q
A
=
lq
x i x k x j
x j
h x l l x p x r
h
i
k
i k x
p r x x
h
x l x 2 l
x p x q x r
2 x p x q
A
lq
lq
= i k
x h x j p r
x i x j x k
x i x k x j
A pq
h
2 x p x q
l
Ahj
i
k
j =
x x x
i k
p
as Ahj = Alq
x p x q x r
Alq i
r
x x j x k
...(3)
x l x q
by equation (1)
x h x j
and
Apq
x p 2 x q
x p 2 x l
A
=
pl
x i x j x k
x i x j x k
= Apl
x p
x i
h x l l x q x r
h
j
k
j k x
q r x x
h
x p x l l x q x r x p
=
A
Apl
pl
x i x h q r x j x k x i
j k
Apq
h
x p 2 x q
l x p x q x r
A
A pl
=
i
ih
j
k
x i x j x k
j k
q r x x x
...(4)
Substituting the value of equations (3) and (4) in equation (2) we get,
A pq
h
h
l
l x p x q x r
r Alq p r A pl q r
+ Ahj +
=
Aih
i
j
k
k
x x x
x
i k
j k
x
Aij
h
h
A pq
l
l x p x q x r
ih
hj
lq
pl
r
i
j
k
x k j k
i k
p r
q r x x x
x
Aij
Aij , k =
71
Aij
h
h
Aih Ahj , then
x
j k
i k
k
Aij , k = Apq , r
x p x q x r
x i x j x k
It is law of transformation of a covariant tensor of rank three. Thus, A ij,k is a covariant tensor
of rank three.
So, Aij , k is called covariant derivative of Aij w.r.t. to x k .
Similarly we define the covariant derivation xk of a tensors A ij and Aij by the formula
A ij
i
j
+ A lj + A il
A ,k =
k
x
l k
l k
A ij
i
l
+ A lj Ali
Aij ,k =
k
x
l k
j k
ij
and
ij ... l
In general, we define the covariant deriavative xk of a mixed tensor Aab
by the formula
... c
ij ... l
Aab
... c ,k
Note:
ij ...l
Aab
pj ... l i
ip... l j
ij... p l
...c
+ Aab
=
+ Aab... c
+ + Aab... c
...c
k
x
p k
p k
p k
p
ij... l p
ij ... l p
A ijpb......l c
Aap... c
Aab ... p
a k
b k
c k
Ai , k is also written as Ai , k = k Ai .
ij
i
l
+ lj il
x
l k
j k
k
i i
= 0+
j k j k
ij,k = 0 as
ij
i i
= 0; lj =
x
l k j k
k
Also, consider first the tensor gij and the covariant derivative of gij is
g ij , k =
g ij
m
m
g mj gim
x
i k
j k
k
72
g ij , k =
g ij
But
x k
g ij
m
[ik , j ] [ jk , i] as g mj = [ik , j ]
x
i k
k
= [ik , j ]+ [ jk, i]
g ij , k = g ij g ij
x k x k
g ij , k = 0
So,
But
im
gmj,k = 0 and ij , k = 0. So, g , k = 0 as g mj 0
EXAMPLE 4
Prove that if A ij is a symmetric tensor then
Ai j, j =
g jk
1
Ai j g A jk
j
2
x i
g x
Solution
Given that A ij be a symmetric tensor. Then
A ij = A ji
...(1)
We know that
Ai j, k
Ai j
j
l
+ Ail Al j
=
k
x
l k
i k
Ai j, j
Ai j
j
l
+ Ail Al j
=
j
x
l j
i j
Put k = j, we get
log g
Ai j
+ Ail
Al j g hl [ ij, h]
j
x
x l
Ai j
Ai j g
+
A jh [ij, h] since A ij is symmetric.
x j
g x j
Ai j, j =
...(2)
1 Ai j g
A jk [ij, k ]
g
x j
...(3)
73
But
A jk [ij , k ] =
1 jk g jk g ki g ij
A i +
2
x j x k
x
A jk [ij , k ] =
g ij
1 jk g jk
jk g ki
jk
A
+
A
A
2
x i
x j
x k
g
g ki
A kj kji
j =
x
x
On Interchanging the dummy indices j & k.
A jk
A jk
...(4)
g ji
g ki
jk
A
=
since Aij = A ji
x k
x j
g
g ki
A jk ijk = 0 as g ij = g ji
j
x
x
Using (5), equation (4) becomes
A jk
A jk [ij , k ] =
...(5)
1 jk g jk
A
2
x i
g jk
1 Ai j g
1
A jk
j
2
g
x
x i
Proved.
EXAMPLE 5
k k
Prove that are components of a tensor of rank three where
i j a i j b
are the Christoffel symbols formed from the symmetric tensors aij and bij .
Solution
Since we know that from equation (8), page 65.
2 xs
x s
=
x i x j
x k
x s
x k
or
k x p x q s
i
j
i j x x p q
k
2 xs
x p x q s
+
=
x i x j x i x j p q
i j
k
2 xs
x p x q
+
= i j
x i x j
i j
x x
s x k
s
p q x
k
k
and
i j a
i j b
74
and
2 xs
k
x p x q
+
=
i
j
x i x j
x x
i j a
s x k
s
p q a x
k
2 xs
x p x q
= i j+ i
i j b
x x j
x x
s x k
s
p q b x
Subtracting, we obtain
s
k k
s x p x q x k
i
=
j
s
p q a p q b x x x
i j a i j b
Put
s
s
s
= A pq
p q a p q b
Then above equation can written as
x p x q x k
x i x j x s
It is law of transformation of tensor of rank three.
s
Aijk = Apq
k k
So, are components of a tensor of rank three.
i j a i j b
EXAMPLE 6
If a specified point, the derivatives of gij w.r.t. to xk are all zero. Prove that the components of
covariant derivatives at that point are the same as ordinary derivatives.
Solution
Given that
g ii
= 0, i , j, k at P 0
...(1)
x k
Let Aij be tensor..
Now, we have to prove that Aij , k =
Aij ,k =
Aij
x k
A ij
at P0 .
i
i
+ A j
A
x
k
j k
k
...(2)
75
g ij
k
Since and [ij,.k ] both contain terms of the type
and using equation (1) we get
x k
i j
k
= 0 = [ij, k] at P0 .
i j
So, equation (2) becomes
A ij
Aij ,k =
at P0
x k
4.7 GRADIENT, DIVERGENCE AND CURL
(a) Gradient
If be a scalar function of the coordinates, then the gradient of is denoted by
grad =
x i
A i
i
+ Ak
i
x
k i
1 g Ak
g
x k
Solution:
If Ai be components of contravariant vector then
i
div A i = A, i =
A i
i
+ Ak
i
x
k i
Since
1 g
log g =
=
k
x
g x k
k i
So,
div A i =
A i
1 g k
+
A
i
x
g x k
76
div A i =
1 g Ak
g
x k
div A i =
...(1)
Proved
(c) Curl
Let Ai be a covariant vector then
Ai
k
Ak
j
x
i j
Ai , j =
A j
k
Ak
x
j i
Aj, i =
and
Ai A j
Thus
curl Ai = Ai , j A j ,i
Note: curl A i is a skew-symmetric tensor.
Since
A j, i A i,j = (A i,j A j,i)
EXAMPLE 8
If Aij be a skew-symmetric tensor of rank two. Show that
Aij , k + A jk ,i + Aki, j =
Aij
x
A jk
x
Aki
x j
Solution
Since we know that
Aij , k =
A jk ,i =
Aki, j =
Aij
l
l
Alj Ail
i
k
x
j k
k
A jk
h
l
Alk A jl
x
j i
j k
i
Aki
l
Ali
j
x
k
l
Akl
j
i j
77
Aij
x k
A jk
x i
l
Aki l
A
+
A
lj
jl
x j i k
k i
l
l l
l
Ali + Ail
Akl + Alk
j k i j
j i
k j
l
Since is symmetric i.e.,
i k
l l
= etc.
i k k i
=
Aij
x
k
Since Aij is skew-symmetric. Then Alj
A jk
x
Aki l
( Alj + A jl )
x j i k
l
( Ali + Ail ) ( Akl + Alk )
j
i j
= A jl Alj + A jl = 0 . Similarly,,
Aij
x
A jk
x
Aki
x j
THEOREM 4.5 A necessary and sufficient condition that the curl of a vector field vanishes is that
the vector field be gradient.
Proof: Suppose that the curl of a vector Ai vanish so that
curl Ai = Ai , j A j , i = 0
...(1)
Ai A j
=0
x j xi
A j
Ai
=
j
x i
x
A j j
Ai
j
dx
dx
=
xi
x j
A j dx j
dAi =
x i
Integrating it we get
Ai =
A j dx j
i
x
78
A j dx j
i
x
Ai =
, where =
x i
or
Ai = .
Conversely suppose that a vector Ai is such that
=
A j dx j
Ai = , is scalar..
To prove curl Ai = 0
Now,
Ai = =
and
So,
x i
2
Ai
=
x j xi
x j
A j
2
=
x i
x i x j
Ai A j
=0
x j xi
So,
curl Ai = Ai , j A j ,i =
So,
curl Ai = 0
Ai A j
=0
x j x i
Proved.
THEOREM 4.6 Let and be scalar functions of coordinates xi. Let A be an arbitrary vector then
(i)
(ii)
() = +
(iii)
2 () = 2 + 2 + 2
(iv)
div () = 2 +
1 g Ai
g
x i
...(1)
replace A i by A i , we get
div( A ) =
i
1 g Ai
g
x i
1
g
79
( g Ai )
+
g
A
i
x i
x
i
A +
1 ( g A )
g
= A i + div A i
Thus
div( A) = A + div A
...(2)
( )
x i
+ i
i
x
x
() = +
Thus
(iii) Taking divergence of both sides in equation (3), we get
div () = div [ + ]
2 () = div () + div ()
= + div () + + div ()
2 () = div () + div () + 2
Thus,
2 () = 2 + 2 + 2
(iv) Replace A by in equation (3), we get
div () = + div ()
div () = + 2
THEOREM 4.7 Let Ai be a covariant vector and a scalar function. Then
(i)
(ii)
curl () =
( )
...(3)
80
= , jAi + Ai , j , i A j A j , i
= ( Ai , j A j , i) + ( Ai , j A j ,i )
= Ai + curl Ai
So,
curl (A) = A + curlA
...(1)
curl () = curl () + .
curl () = .
Proved.
g g kr r
k
x
g x
Proof: Since
2 = div grad
...(1)
and
grad =
,
x r
(Covariant vector) is
x r
x r
1 g Ak
, (Sec Ex. 7, Pg 75)
g
x k
81
2 = div g kr r =
x
g g kr r
k
x
Proved.
EXAMPLE 9
Show that, in the cylindrical coordinates,
2V =
1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2
by Tensor method
Solution
The cylindrical coordinates are (r , , z ). If V is a scalar function of (r , , z ).
Now,
2V = V
Since
V = i
V
V
V
+j
+k
r
Let
V
V
V
, A2 =
, A3 =
...(1)
r
z
Then V = iA1i + jA2 j + kA3 k , since V is covariant tensor. The metric in cylindrical coordinates
A1 =
is
ds 2 = dr 2 + r 2 d2 + dz 2
here, x1 = r , x 2 = , x 3 = z .
Since ds 2 = g ij dx i dx j
g11 = 1, g 22 = r 2 , g 33 = 1
and others are zero.
g11 g12 g13
1 0 0
g = g ij = g 21 g 22 g 23 = 0 r 2 0 = r 2
g 31 g 32 g33 0 0 1
g =r
Now,
div (V ) = div Ai =
k
1 gA
x k
g
82
1 ( g A )
=
div (V ) =
1 ( g A )
g
( g A )
x
1 ( rA1 ) ( rA 2 ) ( rA3 )
+
+
r r
We can write
A k = g kq Aq (Associated tensor)
A k = g k1 A1 + g k 2 A2 + g k 3 A3
Put k = 1
11
12
13
A1 = g A1 + g A2 + g A3
A1 = g 11 A1 as g 12 = g 13 = 0
Similarly,
A 2 = g 22 A2
A3 = g 33 A3
and
g 11 =
Cofactor of g11 in g r 2
= 2 =1
g
r
g 22 =
Cofactor of g 22 in g
1
= 2
g
r
g 33 =
Cofactor of g 33 in g r 2
= 2 = 1 (See. Pg. 34, Ex.1)
g
r
So,
11
A1 = g A1 = A1
1
22
A 2 = g A2 = 2 A2
r
33
A3 = g A3 = A3 .
or A1 = A1, A 2 = 1 A2 , A3 = A3 .
r2
from (1), we get
V
1 V
V
2
3
A1 = r , A = r 2 , A = z
from (2),
...(2)
83
So,
1 V 1 V V
i
2V = div A = r r r r + r r 2 + z r z
div (V ) =
1 V
r
r r r
1 V
+
r
V
+ r
z z
1 V 1 2 V
2V
+
r
+
r
r r r r 2
z 2
div (V ) =
1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2
2V =
1 V 1 2V 2V
r
+
+
r r r r 2 2 z 2
EXERCISES
1. Prove that the expressions are tensors
(a) Aij , l =
(b)
Air jk ,l
Aj Ai
jl
x i l
Aij
l
Air jk
x
r A
r
A Ar Ar
l jk j l i k k l ij + l ijk
2. Prove that
Ai j, j
1 ( Ai
g
g)
j
j k
Ak
i j
( g Ai jk ) is a tensor..
g xk
4. Prove that the necessary and sufficient condition that all the Christoffel symbols vanish at a point is
that g ij are constant.
5. Evaluate the Christoffel symbols in cylindrical coordinates.
6. Define covariant differentiation of a tensor w.r. to the fundamental tensor gij . Show that the covariant
differentiation of sums and products of tensors obey the same result as ordinary differentiation.
7. Let contravariant and covariant components of the same vector A be Ai and Ai respectively then
prove that
i
div A = div Ai
84
by tensor method.
1 2 V
1
V
1
2V
r
+ 2
sin
+ 2 2
2
r sin 2
r r r r sin
CHAPTER 5
RIEMANN-CHRISTOFFEL TENSOR
5.1 RIEMANN-CHRISTOFFEL TENSOR
If A i is a covariant tensor then the covariant x j derivative of A i is given by
A i,j =
Ai
A
x j i j
...(1)
A A
i A
A
j
j
x
i j i k
j
A
i A
i k x
i k
x k
Ai
2
Ai , jk =
x k x
+
i k
Interchanging j and k in equation (2), we get
Ai
A
+
A
j
j k x
j k i
i k
Ai
x j x k
x j
2
Ai , kj =
A A
i j
k
k
j
x
i k x
i k x
A
A
A j k
i k x
i j x
...(2)
86
+
i j
Subtract equation (3) from (2), we get
`
Ai , jk Ai , kj
Ai
A +
k
k i x
k j
=
i k
A
i
i j
A
A
A +
k
x
j
i j k
...(3)
i k A
x j
Interchanging of and in the first and third term of above equation, we get
Ai , jk Ai , kj
i k
=
x j
i j
x k
+
i k
j i j
Ai , jk Ai , kj = A Rijk
...(4)
...(5)
where
Rijk
i k
i j
+
=
x j
xk
i k
j i j
...(6)
Since A i is an arbitrary covariant tensor of rank one and difference of two tensors A i, jk Ai, kj is
a covariant tensor of rank three. Hence it follows from quotient law that Rijk is a mixed tensor of rank
four. The tensor Rijk is called Riemann Christoffel tensor or Curvature tensor for the metric g ij dx i dx j .
The symbol Rijk is called Riemanns symbol of second kind.
Now, if the left hand side of equation (4) is to vanish i.e., if the order of covariant differentiation
is to be immaterial then
Rijk = 0
Since A is arbitrary. In general Rijk 0, so that the order of covariant differentiation is not
immaterial, It is clear from the equation (4) that a necessary and sufficient condition for the validity of
inversion of the order of covariant differentiation is that the tensor Rijk vanishes identically..
Remark
The tensor
i
Rikl
x k
= i
j k
l
x
k
+
i
j
l
j k
l
j l
...(7)
Riemann-Christoffel Tensor
87
THEOREM 5.1 Curvature tensor Rijk is anti symmetric w.r.t. indices j and k.
Proof: We know that from (7) curvature tensor
x j
j k
xk
Rijk = +
Rijk
i j
i k
i k
=
x j
i k
i j
+
x k
j
i k k
i j
R ikj
i j i k +
=
k i j j i k
x k
x j
i k
=
j
x
Rik
i j
x k
j i
k k i j
= Rijk
x k
Rijk
=
i j
Rijk
i k
= i k
x j
j k
i j
i j
+
x k
j
i k
i k i j
...(1)
88
Similarly
R j ki
j
i
j
k
= k i +
x
x
k j i j k i
...(2)
and
Rki
k j
=
xi
k i
+
x j
i
k j k i
...(3)
x j
Rjk
=
x k
k
=
x j
=
2 log g
x j x k
j
+
x k
j
k k
2 log g
xk x j
log g
Since k =
and and are free indices Rj k = 0 .
x k
Riemann-Christoffel Tensor
89
Rij = Rij
x
x
j
=
+
i j i i j
Rij = j
x
Rij =
2 log g
x j x i
i
j +
j i i j
x
i
j
+
x
j i i j
...(1)
R ji = i
x
R ji =
2 log g
x i x j
j
i +
i j j i
x
i
j
x
i j i j
...(2)
...(1)
is known as the covariant Riemann-Christoffel tensor or the Riemann-Christoffel tensor of the first
kind.
Expression for Rijkl
Rijkl = gi Rjkl
90
= g i k j l l j k + j l k j k l
x x
(2)
Now,
g i
=
x k j l
x k
=
g i
gi j l j l x k
[ jl , i ] g i
k
x k
j l x
...(3)
Similarly,
g i
[ jk , i ] gi
=
l
l
x j k
x l
j k x
...(4)
udv duv vdu
By the formula dx = dx dx
Ri j k l =
[ jl , i] gi [ jk , i ]
k
+
x k
xl
j l x
g i
l + gi
j k x
j l
g i
k
j k
[ jl , i ] [ jk , i ] g i g i
+
l j l k + g i j l k g i j k l
k
l
j
k
x
x
x
x
[ jl , i ] [ jk , i]
+
( [il, ] + [l , i ] ) ( [ik , ] + [k , i] )
k
l
x
x
j k
j l
+ [k , i]
[l , i ]
j l
j k
Ri j k l =
[ jl , i ] [ jk , i]
+
[il, ] [ik , ]
k
l
x
x
j k
j l
...(5)
It is also written as
Ri j k l
k
k
+ j k
x
= x
[ jk , i ] [ jl, i] [ik , ]
j l
[il, ]
...(6)
[ jl, i] =
1 gli g ji g jl
+
i
2 x j x l
x
...(7)
Riemann-Christoffel Tensor
91
[ jk ,i ] =
and
1 g ki g ji g ik
+
i
2 x j x k
x
...(8)
g li g ji g jl
j +
xl
xi
x
1
l
2 x
g ki g ji g jk
j + k
x
x i
x
+
[il, ] [ik , ]
j k
j l
2
2
g jk
g jl
1 g il
g ik
+
= 2 j k
i
l
j
l
i
k
x x
x x x x
x x
2
Rijkl
[il , ] [ik , ]
j k
j l
...(9)
Since
= g [ jk , ] and j l = g [ jl , ]
j
k
Ri j k l =
2 g jk
2 g jl
1 2 g il
2 g ik
+
2 x j x k x i x l x j x l x i x l
+ g [ jk , ][il, ] g [ jl , ][ik , ]
...(10)
2 g jk
2 g jl
1 2 gil
2 gik
+
= 2 x j x k x i x l x j x l x i x k
+ g
j k
g
i l
j l i k
R jikl = Rijkl
(ii)
Rijlk = Rijkl
(iii)
Rklij = Rijkl
(iv)
92
2 g jk
2 g jl
1 2 gil
2 g ik
+
2 x j x k x i x l 2 x j x l x i x k
+
[il, ] [ik , ]
j k
j l
...(1)
2
2 g jk
2 gik
2 gil
1 g jl
=
= i k
2 x x
x j x l x i x l x j x k
+
[ jl, ] il [ jk , ]
i k
2 g jk
2 g jl
1 2 gil
2 gik
+
2 x j x k x i x l x i x k x j x l
+ [ jl , ]
i k
[ jk , ]
i l
R jikl = Rijkl
or
Rijkl = R jikl
R kjil =
2 x j x i x k x l x j x l x k x i
Now interchange j and l, we get
R klij
2
2 g lj
1 g kj
2 g li
2 g ki
= l i+ k j l j k i
2 x x x x
x x
x x
For
[kj, ] = g
l i
k j
l i
=
g
k j
l i
=
[li, ]
k j
[li, ] = k j [li, ]
l i
+ [kl, ] [ki, ]
j i
j l
+ [kj, ] [ki, ]
l i
l j
Riemann-Christoffel Tensor
93
So,
R klij =
2
2 g lj
1 g kj
2 gli
2 g ki
+
2 x l xi x k x j x l x j x k x i
+
[li, ] [ki, ]
k j
l j
2
2 g jk
2 g jk
2 g jl
1 g il
= 2 j k + i l j l j k
x x
x x x x
x x
+
[il , ] j l [ik , ]
j k
Riklj
2
2 g kj
1 g ij
2 g kl
2 g il
= k l + i j k j i l
2 x x x x
x x
x x
+ [ij, ] [il , ]
k l
k j
2
2 g lj
2 gij
2 g lk
1 g ik
+
2 x l x j x i x k x l x k x i x j
+
[ik , ] lk [ij, ]
l j
Riljk =
components of Rijkl is
n (n 1)
.
12
94
Hence the number of distinct non-vanishing components of the curvature tensor Rijkl does not
1 2 2
n (n 1) .
exceed
12
Remark
When Rijkl is of the form Riiii i.e., all indices are same
In this case, Riiii has no components.
5.5 BIANCHI IDENTITY
It states that
R ijkl, m + R ijlm, k + R ijmk ,l = 0
and
Proof: Introducing geodesic coordinate1 in which Christoffel symbols are constant with the pole at P0 .
Since we know that
R ijkl
R ijkl
R ijkl, m
i
k
l
x x
mk
+
=
ii
m
jk
jl
jk
i
jl
=
x k
i
ml
m
jl
i
jk + i m m i
x l
mk jl jk ml
i
i
2 2
j l j k
=
x m x k
x m x l
...(1)
i m
Since j k , j l etc. are constant at pole.
So, their derivatives are zero.
x l
R ijlm = i
j l
x m
i
j m
j l
j m
in curve . Geodesic.
Riemann-Christoffel Tensor
95
i
i
i
= j m j l +
l
x
xm
l
R ijlm, k
j m m
j l
i
2 i
2
j m j l
=
x k x l
x k x m
...(2)
and
R ijmk
R ijmk
R ijmk
m
k
x
x
m
+
=
i i
j m j k j m
j k
i
i
j k j m + i i
=
x m
x k
m j k k j m
i
2 i
2
j k
j m
x m x l
x k x l
...(3)
...(4)
96
1
R, m = 0
2
1
Rmk , k km R , k = 0
2
since R , m = km R, k
k 1 k
Rm m R = 0
2
, k
1
Rmk km R is divergence free.
2
1 i
1
i
The tensor Rmk km R = Gmk or R j j R is known as Einstein Tensor..
2
2
5.7 RIEMANN CURVATURE OF A V n
Consider two unit vectors pi and qi at a point P 0 of V n. These vectors at P 0 determine a pencil of
directions deferred by ti = pi + qi. and being parameters. One and only one geodesic will pass
through P 0 in the direction of p i . Similarly one and only one geodesic will pass through in the direction
qi. These two geodesics through P 0 determined by the orientation of the unit vectors pi and qi. Let this
surface is denoted by S.
The Gaussian curvature of S at P 0 is defined to be the Riemannian Curvature of Vn at P0 for the
orientation determined by pi and qi.
Let the coordinates y i of Vn are Riemannian coordinates with origin at P 0. The equation of
surfaces S in given by
i.e.,
y i = ( p i + q i)s
...(1)
y i = p iu 1 + q i u 2
...(2)
Riemann-Christoffel Tensor
97
y i y j
(, = 1, 2)
u u
k
Let and
be the Christoffel symbols corresponding to the coordinates y i and u .
i
j
R
Let and R hijk be curvature tensor corresponding to the metrices b du du and g ij dy i dy j .
Since the Greek letters , , , take values 1,2 and so that the number of independent non1
1 2 2
vanishing components of R is n2 (n 2 1) for n = 2, i.e., they are
2 2 1 = 1. Let us
12
12
transform the coordinate system u to u and suppose that the corresponding value of R1212 are
'
.
R1212
Then
'
= R
R1212
u u u u
u 1 u 2 u 1 u 2
= R1
u1 u u u
u 2 u u u
+
R
2
u 1 u 2 u 1 u 2
u 1 u 2 u 1 u 2
= R12
u 1 u 2 u u
u 2 u 1 u u
+
R
21
u1 u 2 u1 u 2
u1 u 2 u1 u 2
= R1212
u1 u 2 u1 u 2
u1 u 2 u 2 u1
+
R
1221
u 1 u 2 u 1 u 2
u 1 u 2 u 1 u 2
+ R 2112
u 2 u1 u1 u 2
u 2 u1 u 2 u 1
+
R
2121
u11 u 2 u 2 u 2
u 1 u 2 u 1 u 2
2
u 1 u 2
u1 u 2 u 2 u u 2 u1
2
+
= 2112 u1 u 2
u 1 u 2 u 2 u 2 u 1 u 2
= R1212 J 2 where J =
so,
= R1212 J 2
R1212
u1
u 1
u1
u 2
u 2
u 1
u 2
u 2
u
u
...(3)
98
Again
b = b
b = b
or
from (3) and (4), we get
u u
u u
u
u
u
u
b = bJ 2
...(4)
R1212
R
= 1212 = K (say)
b
b
...(5)
This shows that the quantity K is an invariant for transformation of coordinates. The invariant K
is defined to be the Guasian curvature of S. Hence K is the Riemannian Curvature of S at P0 .
Since Riemannian Coordinates yi with the origin at P0 . We have as geodesic coordinates with the
pole at P0 .
Therefore
Then
k
,
= 0 at P0
i j g b
and
at P0 .
[, ] b [, ] b
+
at P0 .
R =
u
u
R1212 =
[21,1] b
u
[22,1] b
u1
at P0
...(6)
K = b
u 2
u 1
...(7)
Riemann-Christoffel Tensor
99
[, ]b
y i y j y k
[ij, k ]g
u u u
[21, 1]b
y i y j y k
[ij, k ]g
u 2 u1 u1
...(8)
[ij, k ]g
qi p j p k
u 2
i j k
= q p p
[ij, k ]g y h
y h u 2
i j k h
= q p p q
[ij, k ]g
y h
i k j h
= qq p p
[ij, h ]g
...(9)
y k
Similarly,
[22, 1] b
u
i k j h
= qq p p
[ij, h ]g
...(10)
y k
h i j k
R1212 = p q p q Rhijk at P0
Since
b = g ij
y i y j
u u
b11 = g ij
y i y j
= g ij p i p j
1
1
u u
b11 = g hj p h p j
at P0
...(11)
100
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
...(12)
...(13)
This is formula for Riemannian Curvature of Vn at P0 determined by the orientation of Unit vectors p i
and q i at P0 .
5.9 SCHURS THEOREM
If at each point, the Riemannian curvature of a space is independent of the orientation choosen then it
is constant throughout the space.
Proof: If K is the Riemannian curvature of Vn at P for the orientation determined by unit vectors p i
and q i then it is given by
p h q i p j q k R hijk
K=
(g ik g hj g ij g hk ) p h qi p j q k
Let K be independent of the orientation choosen. Then equation (1) becomes
K=
...(1)
R hijk
g ik g hj gij g hk
Rhijk = K (g ik g hj g ij g hk )
...(2)
...(3)
Taking the sum of (3) and two similar equations obtained by cyclic permutation of the suffices,
j, k and l.
Riemann-Christoffel Tensor
101
...(4)
Here n > 2 therefore three or more distinct values to indices j, k, m can be given .
Multiplying (4) by g hj and using g hj g hl = lj , we get
(ng ik ih g hk )k , l +(g il kj gik l j ) K , j +( ih g h l ngil )K , k = 0
or,
(n 1) g ik K , l + 0 + (1 n) gil K , k = 0 for ij = 0, i j
g ik K ,l g il K , k = 0
Multiplying by g ik and using
g ik ik = n, g il g ik = kl ,
we get
nK,l kl K ,k = 0 or (n 1)K ,l = 0
or
K, k = 0 as (n 1)K ,l = 0
K
= 0.
x l
integrating it, we get K = constant. This proves that the partial derivatives of K w.r.t. to xs are all zero.
or
ehp e qk rk sk R pqrs
ehp e qk e rh eks ( g pr q qs g ps g qr )
e hp1 e qk1 e rh1 e ks1 R pqrs
= (e p e r g ) (e q e s g ) (e p e s g ) (e q e r g )
pr
ps
h h
k k qs
h k
k h qr
...(1)
102
e h eh g ps = 0 etc.
and
Using these in equation (1), then equation (1) becomes
K hk =
K hk =
n
e
k =1
11 0 0
e hp e kq ehr
e ks R pqrs
...(2)
hk
k =1
Put
e hp ekq e rh e ks R pqrs
hk
k =1
p q r s
e e e
h k h k
R pqrs
= M h . Then
n
p r
M h = eh eh
e
k =1
q s
e
k k
R pqrs
p r qs
= e h e h g R pqrs
p r qs
= e h e h g R qprs
p r
M h = eh e h R pr
...(3)
h =1
p r
= eh |e h| R pr
= g pr R pr
= R
n
or
M
h =1
= g pr R pr == R
This proves that the sum of mean curvatures for n mutual orthogonal directions is independent of the
directions chosen to complete an orthogonal ennuple and has the value R.
5.11 RICCIS PRINCIPAL DIRECTIONS
Let e ih1 is not a unit vector and the mean curvature M h is given by
Mh =
Rij e hi e hj
g ij e hi ehj
Riemann-Christoffel Tensor
103
( Rij + M h g ij ) e ih ehj = 0
...(1)
...(1)
1
= nR
from (1)
Rij =
R
g
n ij
R
g at every point of the space.
n ij
104
Rhijk
K = (g g g g )
ik hj
ij hk
Rhijk = K (g ik g hj g ij g hk )
Multiplying by g hk
Kg hk ( g ik g hj g ij g hk ) = g hk Rhijk
K (hi g hj ng ij ) = Rij
Since g hk g ik = hi ; g hk g hk = n and g hk Rhi jk = Rij
K (g ij ng ij ) = Rij
K (1 n) g ij = Rij
...(1)
ij
Multiplying by g , we get
Kn(1 n) = R as g ij Rij = R
...(2)
R
1
= Rgij
n(1 n ) n
R
g
n ij
This is necessary and sufficient condition for the space Vn to be Einstein space.
Rij =
1
( g R g kh g ij )
1 n ki hj
Theorem 5.6 A necessary and sufficient condition for a Riemannian V n (n > 3) to be of constant
curvature to that the Weyl tensor vanishes identically throughout Vn .
Proof: Necessary Condition:
Let K be Riemannian Curvature of Vn . Let K = constant.
We have to prove that Whijk = 0
Since we know that
K=
p h q i p j q k R hijk
(g hj g ik g ij g hk ) p h qi p j q k
= constant
Riemann-Christoffel Tensor
105
R hijk
g hj gik gij g hk
...(1)
Multiplying by g hk , we get
g hk Rhijk = g hk K (g hj g ik g ij g hk )
Rij = K (kj gik ng ij )
Rij = K (1 n) g ij
...(2)
...(3)
R
g
n ij
...(4)
1
[ g R g hk g ij ]
1 n ik hj
Whijk = Rhijk +
1 R
R
g ik g hj g hk g ij
1n
n
n
R
= Rhijk + n(1 n) [ gik g hj g hk gij ]
= Rhijk +
Rhijk
R
Whijk = Rhijk + K
Rhijk
K
, by equation (3)
Whijk = 2 R hijk
Since K is constant. The equation (5) shows that Whijk = 0 .
This proves necessary condition.
106
Sufficient Condition
Let Whijk = 0. Then we have to prove that K is constant.
Now, Whijk = 0.
Rhijk +
1
[ g R g hk Rij ] = 0
1 n ik hj
Multiplying by g hk , we get
Rij +
1
[g hk g ik Rhj g hk g hk Rij ]
1 n
1
Rij +
[h R nRij ]
1 n i hj
R
Rij + ij (1 n)
1 n
2 R ij
=0
=0
=0
=0
Rij = 0
K=
R hijk p h qi p j q k
(g hj g ik 0 g ij ) p h qi p j q k
R hijk p h q i p j q k
( p h p j g hj ) (q i q k g ik )
R hijk p h q i p j q k
p2 q 2
h i j k
K = Rhijk p q p q since p 2 = 1, q 2 = 1
K = constant as Rij = 0
This proves sufficient condition.
EXAMPLE 1
For a V2 referred to an orthogonal system of parametric curves (g 12 = 0) show that
R12 = 0, R11 g 22 = g 22 g11 = R1221
ij
R = g Rij =
Consequently
Rij =
1
Rg .
2 ij
2R1221
g11 g 22
Riemann-Christoffel Tensor
107
Solution
Given that g12 = 0 so that g 12 = 0.
Also,
g ij =
1
1
1
22
g 11 =
& g =g .
g ij
g11
22
g = g ij =
g11 g12
g
0
= 11
= g11 g 22
g 21 g 22
0 g 22
R12
= g 21 R2121 as R1121 = 0
=0
R 2112
g 22
So,
...(1)
R22 =
R1221
g11
R1221
g11
...(2)
2R1221
g11 g 22
ij
i1
i2
R = g Rij = g Ri 1 + g Ri 2
(3)
108
R 11 R 22
+
g 11 g 22
R1221
R1221
[Q by eqn. (3)]
= g g +g g
22 11
11 22
2R1221
R=
g11 g 22
(iv) To prove Rij =
1
Rg
2 ij
R=
2R1221
g11 g 22
R=
2R1221
as g = g11 g 22
g
R1221 =
1
Rg
2
1
Rg = R 22 g11 .
2
it becomes
R11 =
Rg
Rg g
Rg
= 11 22 = 11
2 g 22
2g 22
2
Rg
Rg11 g 22 Rg22
=
=
R22 =
2 g11
2g11
2
So,
1
1
Rg11 & R22 = g 22
2
2
1
= g12 as R12 = 0 = g12
2
R11 =
R12
This prove that Rij =
1
R .
2 ij
EXAMPLE 2
The metric of the V 2 formed by the surface of a sphere of radius r is ds 2 = r 2 d 2 + r 2 sin 2 d 2
1
in spherical polar coordinates. Show that the surface of a sphere is a surface of constant curvature 2 .
r
Solution
Given that
ds 2 = r 2 d 2 + r 2 sin 2 d 2
Since r is radius of curvature then r is constant.
Riemann-Christoffel Tensor
g11 = r 2 ,
109
g 22 = r 2 sin 2 , g12 = 0,
g = g11 g 22 = r 4 sin 2 .
We can prove
R 1221 = r2 sin 2
Now, the Riemannian curvature K of V n is given by
p hq i p jq k
.
K= h i j k
p q p q ( g hj gik g hk gij )
At any point of V2 there exists only two independent vectors.
Consider two vectors whose components are (1, 0) and (0, 1) respectively in V2 . Then
K=
R 1212
R
= 1212
g11 g 22
g
K=
r 2 sin 2 1
=
= constant.
r 4 sin 2 r 2
EXERCISES
1. Show that
Rijkl =
[ jl, i] [ jk , i]
+
[
il
,
[ik , ]
xk
xl
j k
j l
2. Show that
2
g jk 2 g ik
g jl
1 g il
+
2 x j xk xi xl x j xl xi xk
Rijkl =
+ g ( [ jk , ][il , ] [ jl , ][ik , ] ) .
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
(i) Rhj =
1
Rhiij
g ii
110
(ii) Rhh =
1
1
Rhiih +
Rhjjh
g ii
g jj
7. Prove that if
1 R
Ri = g Rij then Ri , = 2 i
x
and hence deduce that when n > 2 then scalar curvature of an Einstein space is constant.
8. If the Riemannian curvature K of Vn at every point of a neighbourhood U of Vn is independent of the
direction chosen, show that K is constant throughout the neighbourhood U. Provided n > 2.
9. Show that a space of constant curvature K0 is an Einstein space and that R = K 0 n (1 n ).
10. Show that the necessary and sufficient condition that Vn be locally flat in the neighbourhood of 0 is
that Riemannian Christoffel tensor is zero.
11. Show that every V2 is an Einstein space.
12. For two dimensional manifold prove that
K=
R
2
13. Show that if Riemann-Christoffel curvature tensor vanishes then order of covariant differentiation is
commutative.
CHAPTER 6
112
6.3 e-SYSTEM
Consider a skew-symmetric system of quantities ei1 ... i n ( or ei1 ,..., i n ) in which the indices i1 ... in assume
values 1,2,...n. The system ei1 ... i n ( or ei1... i n ) is said to be the e-system if
= 1; when i1 , i2 ,..., in
EXAMPLE 1
Find the components of system eij when i, j takes the value 1,2.
Solution
The components of system eij are
e11 , e12 , e21 , e 22 .
By definition of e-system, we have
e11 = 0,
e12 = 1,
e 21 = e12 = 1
e22 = 0,
EXAMPLE 2
Find the components of the system ei jk .
Solution
By the definition of e-system,
e123 = e 231 = e321 = 1
e213 = e132 = e321 = 1
eijk = 0 if any two indices are same.
6.4 GENERALISED KRNECKER DELTA
A symbol i1j1...... ikjk depending on k superscripts and k subscripts each of which take values from 1 to n,
is called a generalised Krnecker delta provided that
(a) it is completely skew-symmetric in superscripts and subscripts
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
i1 ... i k
j ... j = 1; where odd number of transpositions arrange the superscripts
1
k
i.e.,
i.e.,
i.e.,
i i ... i
Theorem 6.1 To prove that the direct product e 1 2 n e j1 j2 ... jn of two systems e i1 ... in and e j1 j 2 ... jn is the
ej
1 j2 ... jn
i i ... i
(ii) ei1i 2 ... i n = 1j12j 2 ...n jn
jn
i i ... i
= 1j1 2j 2 ... nj n
114
and
1 2 ... n
(2) ei1i2 ... in = i1i 2 ... in
12
+ 1; if is an even permutatio n of 12
1; if is an odd permutatio n of 12
=
0; if is not permutatio n of 12
Similarly results hold for all values of and selected from the set of numbers 1, 2, 3.
Hence
+ 1; if i j is an even permutatio n of
1; if i j is an odd permutatio n of
ij
= 0; if two of the subscripts or superscripts are equal or when the
subscripts and superscripts are not formed from the same numbers.
1
If we contract ij . To contract ij first contract it and the multiply the result by . We obtain
2
a system depending on two indices
1 ij 1 i 1
i2
i3
i = 2 j = 2 ( 1 + 2 + 3 )
1 12
1
13
It i = 1 in i then we get = 2 + 3
2
115
...(1)
...(2)
and
i1j1i2j...2 ...i r jr = n (n 1) (n 2) (n r + 1) =
n!
n r!
...(3)
or
e i1 i 2 ... i n e i1 i 2 ... i n = n!
...(4)
...(5)
EXERCISE
1. Expand for n = 3
i
(a) j
12 i j
(c) ij x y
ij 2 1
ij i j
ij
(c) e a a = e a .
(b) ij x x
ij
(d) ij
2. Expand for n = 2
ij 1 2
(a) e ai a j
(b) e ai a j
ijk
3. Show that ijk = 3 ! if i, j, k = 1, 2, 3.
5. Prove that ijk g is a covariant tensor of rank three where where ijk is the usual permutation
symbol.
CHAPTER 7
GEOMETRY
7.1 LENGTH OF ARC
Consider the n-dimensional space R be covered by a coordinate system X and a curve C so that
i
...(1)
C : x i = x (t ), (i = 1,2..., n)
which is one-dimensional subspace of R. Where t is a real parameter varying continuously in the
interval t1 t t 2 . The one dimensional manifold C is called arc of a curve.
1
2
1 2
dx n
n dx dx
Let F x , x ,..., x , dt dt ,..., dt be a continuous function in the interval t1 t t 2 . Wee
i
dx
assume that F x, > 0, unless every dx = 0 and that for every positive number k
dt
dt
1
2
1
2
n
1 2
dx
dx
dx n
x , x ,..., x n , dx , dx ,... dx .
F x1 , x 2 ,..., x n , k
,k
,..., k
kF
=
dt
dt
dt
dt dt
dt
The integral
dx
F x, dt
t1
dt
is called the length of C and the space R is said to be metrized by equation (2).
s=
t2
...(2)
dx
Different choices of functions F x, lead to different metric geometrices.
dt
If one chooses to define the length of arc by the formula
dx p dx q
dt, (p,q = 1, 2, ..., n)
...(3)
t1
dt dt
dx p
dx p dxq
, then the resulting
where g pq ( x)
is a positive definite quadratic form in the variable
dt dt
dt
geometry is the Riemannian geometry and space R metrized in this way is the Riemannian n-dimensional
space R n.
s=
t2
g pq ( x)
Geometry
117
Consider the coordinate transformation T : x i = x i ( x1 ,..., x n ) such that the square of the element
of arc ds,
p
q
ds 2 = g pq dx dx
...(4)
ds 2 = dx i dx i
...(5)
dx
THEOREM 7.1 A function F x, satisfying the condition F x, k = kF x , for every
dt
dt
dt
k > 0. This condition is both necessary and sufficient to ensure independence of the value of the
t2
dx
integral s = t F x, dt of a particular mode of parametrization of C. Thus if t in C : x i = xi (t)
1
dt
is replaced by some function t = (s ) and we denote x i [(s )] by i () . so that x i (t ) i (s ) we have
equality
t2
t1
dx
F x , dt =
dt
s2
s1
F (, ) ds
ds i
and t1 = (s1 ) and t2 = (s 2 ).
ds
Proof: Suppose that k is an arbitary positive number and put t = ks so that t1=ks 1 and t2=ks 2. Then
i
C : x i = x (t ) becomes
i
where =
C : x i (ks) = i (t )
and
or
dx i (ks )
dx i (ks)
=k
ds
dt
t2
dx
s=
F x , dt we get
t1
dt
i (s ) =
s=
s=
s2
s1
s2
s1
dx (ks )
F x (ks),
kds
dt
dx
dx
We must have the relation F (, ) = F x, k = kF x, . Conversely, if this relation is true
dt
dt
118
for every line element of C and each k > 0 then the equality of integrals is assumed for every choice of
parameter t = ( s), 1 (s ) > 0, s1 s s 2 with and t2 = (s 2 ).
i
Note: (i)
(ii)
dx
Here take those curves for which x i (t ) and
are continuous functions in t1 t t 2.
dt
A function F x,
dx
dx
dx
satisfying the condition F x, k = kF x, for every k > 0 is called
dt
dt
dt
i
dx
.
dt
EXAMPLE 1
What is meant, consider a sphere S of radius a, immersed in a three-dimensional Euclidean
manifold E 3 , with centre at the origin (0, 0, 0) of the set of orthogonal cartesian axes O X 1 X 2 X 3 .
Solution
Let T be a plane tangent to S at (0, 0,a) and the points of this plane be referred to a set of orthogonal
cartesian axes O Y 1Y 2 as shown in figure. If we draw from O (0, 0, 0) a radial line OP , interesting
the sphere S at P ( x1 , x 2 , x 3 ) and plane T at Q( x 1 , x 2 , a) then the points P on the lower half of the
sphere S are in one-to-one correspondence with points ( x 1 , x 2 ) of the tangent plane T..
3
X
O
P1
1
P2
C
Q2
K
Q1
Fig. 7.1
If P ( x1 , x 2 , x 3 ) is any point on the radial line OP,, then symmetric equations of this line is
x1 0
x2 0 x3 0
=
=
= 2
x1 0
x 0 a 0
or
x1 = x 1 , x 2 = x 2 , x 3 = a
Since the images Q of points P lying on S, the variables x i satisfy the equation of S,
( x1 )2 + (x 2 )2 + (x 3 ) 2 = a 2
( ) ( )
2 x 1 + x 2
or
+ a2 = a2
...(6)
Geometry
119
x =
1
and
( x 1 ) 2 + ( x 2 )2 + a 2
ax 2
, x2 =
( x 1 )2 + (x 2 )2 + (x 3 ) 2
a2
x =
3
...(7)
( x 1) 2 + ( x 2 ) 2 + ( x 3 ) 2
These are the equations giving the analytical one-to-one correspondence of the points Q on T and
points P on the portion of S under consideration.
Let P1 ( x1 , x 2 , x 3 ) and P2 ( x1 + dx1 , x 2 + dx 2 , x 3 + dx 3 ) be two close points on some curve C lying
on S. The Euclidean distance P1 P2 along C, is given by the formula
ds 2 = dx i dx i , (i = 1, 2, 3)
...(8)
Since
x i
p
dx = x p d x , (P =1, 2)
i
xi xi
d x pd x q
p
q
x x
= g pq ( x )d x pd x q , (p,q = 1, 2)
xi x i .
x p x q
If the image K of C on T is given by the equations
where g pq (x ) are functions of x i and g pq =
x 1 = x 1 ( t )
K:
x 2 = x 2 (t ), t1 t t2
t2
t1
g pq
dx p dxq
dt
dt dt
1 2
2 2
1
+
(
x
)
+
(
x
)
a 2
(d x 1)2 + (d x 2) 2 +
ds2 =
and
...(9)
120
2
1
d x1 d x 2
+
+ 12 x1 d x x 2 d x
t 2 dt
dt
dt
a
dt
dt
s= t
1
1 2
2 2
1
1 + 2 (x ) + ( x )
a
So, the resulting formulas refer to a two-dimensional manifold determined by the variables ( x 1 , x 2 ) in
the cartesian plane T and that the geometry of the surface of the sphere imbeded in a three-dimensional
Euclidean manifold can be visualized on a two-dimensional manifold R 2 with metric given by equation (9).
1
If the radius of S is very large then in equation (9) the terms involving 2 can be neglected. Then
a
equation (9) becomes
1 2
2 2
...(10)
ds 2 = (d x ) + (d x ) .
Thus for large values of a, metric properties of the sphere S are indistinguishable from those of
the Euclidean plane.
The chief point of this example is to indicate that the geometry of sphere imbedded in a Euclidean
3-space, with the element of arc in the form equation (8), is indistinguishable from the Riemannian
geometry of a two-dimensional manifold R 2 with metric (9). the latter manifold, although referred to
a cartesian coordinate system Y, is not Euclidean since equation (9) cannot be reduced by an admissible
transformation to equation (10).
Geometry
121
2
3
x1 = c1 , x = c2 , x = c3
3
Y
(2)
3
2
X
1
X
2
Y
1
Y
Fig. 7.2
intersect in one and only one point. The surfaces defined by equation (2) the coordinate surfaces
and intersection of coordinate surface pair-by pair are the coordinate lines. Thus the line of intersection
of x1 = c1 and x 2 = c2 is the x 3 coordinate line because along this the line the variable x 3 is the only
one that is changing.
EXAMPLE 2
Consider a coordinate system defined by the transformation
3
X2
O
X
Y1
Fig. 7.3
x 1 = x1 sin x 2 cos x 3
...(3)
x 2 = x1 sin x 2 sin x 3
...(4)
x 3 = x1 cos x 2
...(5)
122
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)
(x 1 ) 2 + ( x 2 ) 2
...(7)
or
(x1 )2 + (x 2 )2
x3
1 2
2 2
1 ( x ) + ( x )
tan
x =
x3
...(8)
2
1 x
x 3 = tan 1
x
...(9)
So, the inverse transformation is given by the equations (6), (8) and (9).
If x1 > 0, 0 < x 2 < , 0 x 3 < 2. This is the familiar spherical coordinate system.
7.3 RECIPROCAL BASE SYSTEMS
Covariant and Contravariant Vectors
r r r
Let a cartesian coordinate system be determined by a set of orthogonal base vectors b1 , b2 ,b3 then the
position vector rr of any point P (x 1 , x 2 , x 3 ) can be expressed as
r
r
...(1)
r = bi x i (i = 1, 2, 3)
Geometry
123
r
Since the base vectors bi are independent of the position of the point P (x 1 , x 2 , x 3 ) . Then from (1),
r
r
dr = bi dx i
...(2)
If P (x 1 , x 2 , x 3 ) and Q (x 1 + dx 1 , x 2 + d x 2, x 3 + d x 3 ) be two closed point. The square of the
element of arc ds between two points is
r r
ds 2 = dr dr
from equation (2),
r i r
j
ds 2 = bi dx b j dx
r r
= bi b j d x id x j
r r
i
j
ds 2 = ij d x d x ; since bi b j = ij
3
Y
X
a3
3
2
X
a2
a1
1
X
P
b3
r
O
b2
b1
2
Y
1
Y
Fig. 7.4.
r r r
r r
r r
(b1 , b2 , b3 areorthogonal base vector i.e.,b1 b1 = 1 & b1 b2 = 0) .
ds 2 = d x id x i;
as ij = 1, i = j
= 0, i j
a familiar expression for the square of element of arc in orthogonal cartesian coordinates.
Consider the coordinate transformation
i
1
2
3
x i = x (x , x , x ), (i = 1, 2, 3)
r
define a curvilinear coordinate system X. The position vector r is a function of coordinates x i .
i.e.,
r
r
r = r (x i ), (i = 1, 2, 3)
Then
r
r
r i
dr =
dx
...(3)
x i
124
and
ds 2 = d rr d rr
r
r
r r
d x id x j
=
xi x j
ds 2 = g ij dx i dx j
where
g ij
r
r
r r
=
xi x j
r
r
The vector
is a base vector directed tangentially to X i - coordinate curve.
x i
Put
r
r
r
= ai
i
x
Then from (3) and (4)
r r
r
r
dr = ai d x i and g ij = a i a j
...(4)
...(5)
...(6)
r xi
r
, as d x j are arbitrary
a j = bi
x j
r
So, the base vectors a j transform according to the law for transformation of components of
covariant vectors.
r
The components of base vectors a i , when referred to X-coordinate system, are
r
r
r
a1 : (a1 , 0,0), a 2 : (0, a2 , 0), a3 : (0,0, a3 ).
...(7)
and they are not necessarily unit vectors.
In general,
r r
r r
g11 = ar1 ar1 1, g 22 = a 2 a2 1, g 33 = a3 a3 1.
If the curvilinear coordinate system X is orthogonal. Then
r r
r r
g ij = a i a j = ai a j cos ij = 0, if i j.
r
r
Any vector A are can be written in the form A = k d rr where k is a scalar..
...(8)
...(9)
Geometry
125
r
r r i
d
r
=
d x we have
Since
xi
r
r
r
i
A = x i (kdx )
r
r i
A = ai A
r
where Ai = kdxi . The numbers Ai are the contravariant components of the vector A
Consider three non-coplanar vectors
r r
r r
r r
a 2 a3 ar 2 = a3 a1 , ar 3 = a1 a2
r1
...(10)
a = rr r ,
[ar1ar2 ar3 ]
[ar1ar2ar3 ]
[a1a 2a3 ]
r r
r
r
rr r
where a 2 a3 , etc. denote the vector product of a 2 and a3 and [a1 a2 a3 ] is the triple scalar
r r r
prduct a1 a2 a3 .
Now,
r r r
rr r
a2 a3 a1 [a1 a2 a3 ]
r1 r
=
= 1.
a a1 = [ar ar ar ]
[ar1ar2ar3 ]
1 2 3
r r r
r r r
a2 a3 a2 [a2 a2 a3 ]
r1 r
=
= 0.
a a 2 = [ar ar ar ]
[ar1ar2 ar3 ]
1 2 3
r r r
Since [a2 a2 a3 ] = 0.
Similarly,
r r
r r
a 1 a3 = a 2 a1 = = 0
r r
r r
a 2 a2 = 1, a 3 a 3 = 1
r r
a i a j = ij
EXAMPLE 3
rr r
To show that [a1 a2 a3 ] =
Solution
...(11)
126
and
g= g ij
a22
a32
g = a1 a2 a3
from eqn. (11) and (12), we have
[ar1ar2ar3 ] = g
...(12)
r r r
1
Since the triple products [a 1 a 2 a 3 ] =
. Moreover,,
g
r r
r r
r r
a2 a3
r
a 3 a1 r
a1 a 2
r
,
a
=
,
a
=
a1 = r1 r 2 r 3
r r r
r r r
2
3
[a a a ]
[a1 a 2 a 3 ]
[a1 a 2 a 3 ]
The system of vectors ar1 , ar 2 , ar3 is called the reciprocal base system.
r r r
Hence if the vectors a 1 , a 2 , a 3 are unit vectors associated with an orthogonal cartesian coordinates
then the reciprocal system of vector defines the same system of coordinates.
Solved.
r
r ri
The differential of a vector r in the reciprocal base system is d r = a d xi .
where dxi are the components of drr. Then
r r
ds 2 = dr dr
r
r
= (a i dxi ) (a j dx j )
r r
= a i a j dxi dx j
where
ij
ds 2 = g dxi dx j
r r
g ij = a i a j = g ji
...(13)
Geometry
127
The system of base vectors determined by equation (10) can be used to represent an arbitrary
r r
r
vector A in the form A = a i Ai , where Ai are the covariant components of A.
r
r
Taking scalar product of vector Ai a i with the base vector a j , we get
r r
r r
Ai a i a j = Ai ij = A j as a i a j = ij .
7.4 ON THE MEANING OF COVARIANT DERIVATIVES
r
r
A
r
THEOREM 7.2 If A is a vector along the curve in E 3 . Prove that
= A,j a
j
x
i
r
A
Also, prove that A,i j = j . Where Ai are component of A.
x
r
r
Proof: A vector A can be expressed in the terms of base vectors ai as
r
r
A = Ai ai
r
r
r
r
where a i = i and Ai are components of A.
x
r
The partial derivative of A with respect to x j is
r
r
A
Ai r
i ai
a
+
A
=
i
x j
x j
x j
r r
Since g ij = a i a j .
Differentiating partially it w.r.t. x k , we have
r
r
g ij
a j
a i r
a j + k
=
x k
x k
x
Similarly,
r
r
g jk
a j r
ak
a
+
=
k
x i
xri
xri
g ik
ai r
a
ak + kj
and
j =
j
x
x
x
r
Since A can be written as
r
r i ri
A = a i A = a Ai
r
Taking scalar product with a j , we have
r r
r r
a i a j A i = a i a j Ai
...(1)
r
ai
r
aj
r
ai
g ij A i = ij Ai = A j
r r
r r
As ai a j = g ij , a i a j = ij and ij Ai = A j .
We see that the vector obtained by lowering the index in Ai is precisely the covariant vector Ai .
128
r
The two sets of quantities Ai and Ai are represent the same vector A referred to two different
base systems.
EXAMPLE 4
Show that g i g j = ij .
Solution
Since we know that
Then
g i = ari ar and g j = ar j ar
r r r r
g i g j = (ai a ) (a j a )
r r
r r
= (ai a j ) (a a )
r r
= ij as ai a j = ij
g i g j = ij as = 1.
But
r
r
r
ai =
x i
r
r
r
r
a j
ai
2r
2r
=
=
=
x j
x j dx i x i x j xi
So,
r
r
a j
ai
=
x j
x i
Now,
[ij, k ] =
Substituting the value of
or
Hence
1 g ik g jk g ij
+
k
2 x j
x i
x
g
g ik g jk
, i and ijk , we get
j
x
x
x
r
1 a i r
[ij,k] = 2 j a k
2 x
r
ai r
ak
[ij,k] =
x j
r
ai
rk
1 rk
j = [ ij, k ]a , as r = a
x
ak
r
a i r
r r
a = [ij, k ]a k a
x j
, Christoffels symbol
Geometry
129
r r
= [ij, k ]g k , Since g k = a k a
r
ai r
[ i j , k ] g k =
,
.
a
=
as
i j
x j
i j
r
r
ai
= a
i j
x j
r
ai
Substituting the values of
in equation (1), we get
x j
...(2)
r
Ai r i r
A
ai + A a
=
x j
x j
i j
A r
r
a + A i a
=
j
x
i j
r
A i r
A
=
j + A a
x j
i j
x
r
A
A i
r
=
A
,
a
since
A
=
+ A
j
,j
x j
x j i j
Thus, the covariant derivative A,j of the vector A is a vector whose components are the
r
A
r
components of
j referred to the base system ai .
x
r
ai
=
0
If the Christoffel symbols vanish identically i.e.,
the
= 0, from (2).
x j
i j
Substituting this value in equation (1), we get
r
A
Ai r
ai
=
x j
x j
But
Ai i
+
A
x j j
i
Ai
= 0.
A,i j =
as
x j
j
A,i j =
Proved.
r
r
THEOREM 7.3 If A is a vector along the curve in E 3 . Prove that A j , k a j wheree A j are components
r
of A. .
r
Proof: If A can be expressed in the form
r
r
A = Ai a i
130
r
where Ai are components of A.
r
The partial derivative of A with respect to x k is
r
r
A
Ai ri
a i
a + Ai k
=
x k
x k
x
r r
Since a i a j = ij , we have,
...(1)
a
+
a
=0
j
x k
x k
r
r
r i a j
a i r
a j = a k
x
x k
r r
But a i a = i . Then
r r
= ai a
,
j k
r
a i
x k
r
a i
x k
Since
r
a j
r
= a
x
j k
k
r
a j = i
j k
r
i
a j =
j k
r
i r
a i
a j ,
=
x k
j k
1 rj
as ar = a
j
r
a i
substituting the value of
in equation (1), we get
x k
r
Ai ri
i r j
A
a Ai
a
k = x k
x
j k
=
Ai r j
i rj
a Ai
a
k
x
j k
r
Ai
i r
A
Ai a j
k =
k
x
j k
x
r
A j
i
A
rj
A
=
Ai
=
A
a
,
Since
j,
k
k
k
j, k
x
x
j k
Proved.
Geometry
131
=
dt
x j dt
r
j
dA
r dx
= A, j a
dt
dt
Since we know
r
A r
r
A
A
,
a
=
j + a
j
=
x
i j
x j
So,
r
A i r dx j
dA
= j + i j A a dt
dt
x
r
dA i dx j r
dA
+ A
=
a
dt
dt
i j
dt
dA i dx j
+ A
is called the absolute or Intrinsic derivative of A with respect
dt
i
j
dt
A
.
to parameter t and denoted by
t
The formula
A dA i dx j
A A
=
.
So, t = dt + i j A dt is contravariant vector. If A is a scalar then, obviously,
t
t
Some Results
(i) If Ai be covariant vector
dAi dx
Ai
= dt i A dt
t
(ii)
(iii)
A ij Aij i j dx j i dx
=
+
+
A
A
t
dt
dt
dt
A ij
t
A ij
i dx i dx
+
A
A
dt j dt i dt
132
(iv)
A ijk
t
A ijk
i dx i dx i dx
+
A jk
A
A
dt
dt j k dt k j dt
EXAMPLE 5
If g ij be components of metric tensor, show that
g ij
t
= 0.
Solution
The intrinsic derivative of g ij is
g ij
t
dx
dx
g j
g i
dt i
dt j
dt
dg ij
g ij dx
dx
dx
j
i
dt j
dt
x dt i
g ij
dx
g i
j
= x
i
j dt
g ij
t
as
g ij
dx
= [i, j ] [j, i]
x
dt
gj = [ i , j ] and
g i = [ j , i ] .
j
i
g ij
But
x
g ij
So,
= [i, j ] + [ j, i].
= 0.
EXAMPLE 6
Prove that
d ( g ij A i A j )
dt
= 2 g ij A i
A i
t
Solution
Since g ij Ai a j is scalar..
Then
d ( g ij A i A j )
dt
(g ij A i A j )
t
Geometry
133
( A i A j )
, since g ij is independent of t.
t
A i j
A j
A + Ai
= g ij
t
t
Interchange i and j in first term, we get
= g ij
d ( g ij A i A j )
dt
d ( g ij A i A j )
dt
i A j
A j
+ Ai
= g ij A
as g ij is symmetric.
t
t
= 2 g ij Ai
A j
t
Proved.
EXAMPLE 7
Prove that if A is the magnitude of Ai then
A, j =
Ai , j Ai
A
Solution
Given that A is magnitude of Ai . Then
Since
g ik A i A k = Ai A i
g ik A i A k = A 2
Taking covariant derivative w.r. to x j , we get
g ik A,i j A k + g ik A i A,kj = 2 AA, j
Interchange the dummy index in first term, we get
g ki A,kj A i + g ik A i A,kj = 2 AA, j
2 g ik Ai A,kj = 2 AA, j
g ik A i A,kj = AA, j
Ai g ik A,kj = AA, j
Ai Ai , j = AA, j since g ik A,kj = Ai , j
A, j =
Ai , j Ai
A
Proved.
134
3
X
3
Y
2
X
C
1
X
2
Y
1
Y
r
r
if A is a parallel field along C then the vector A do not change along the curve and we can write
r
r
dA
= 0. It follows that the components Ai of A satisfy a set of simultaneous differential equations
dt
A i
= 0 or
t
dAi i dx
+
=0
A
dt
dt
This is required condition for the vector field Ai is parallel.
7.7 GEOMETRY OF SPACE CURVES
Let the parametric equations of the curve C in E 3 be
C : x i = x i (t ), t1 t t 2
(i = 1, 2, 3).
s=
t2
t1
gij
dxi dx j
dt
dt dt
...(1)
...(2)
dx i dx j
=1
ds ds
...(3)
Geometry
Put
135
dx i
= i . Then equation (3) becomes
ds
g ij i j = 1
...(4)
r
r
The vector , with components i , is a unit vector. Moreover,, is tangent to C, since its
dx i
components i , when the curve C is referred to a rectangular Cartesian coordinate Y,, becomes i =
.
ds
These are precisely the direction cosines
tangent vector to the curve C.
r of the
r
Consider
a
pair
of
unit
vectors
and
(with
components i and i respectively) at any point
r
P of C. Let is tangent to C at P Fig. (7.6).
+ d
P(x)
(Q(x + dx)
C
r + dr
O
Fig. 7.6
r
r
The cosine of the angle between and is given by the formula
i j
cos = g ij
r
r
and if and are orthogonal, then equation (5) becomes
g ij i j = 0
...(6)
i j
j i
+ gij
=0
s
s
(7)
r
Any vector satisfying equation (6) is said to be normal to C at P..
Now, differentiating intrinsically, with respect to the are parameter s, equation (4), we get
g ij
...(5)
i j
i j
+ gij
=0
s
s
j
=0
s
136
g ij i
j
=0
s
j
either vanishes or is normal to C and if does not vanish
s
j
we denote the unit vector co-directional with
by j and write
s
we see that the vector
j =
j
1 j K =
,
s
K s
...(8)
...(9)
j j
j
+ g ij i
=0
s
s
or
g ij i
j
i j
= gij
s
s
= Kg ij i j Since
g ij i
g ij i
i
= K i
s
j
= K , since g ij i j = 1.
s
j
+K =0
s
j
+ g ij i j K = 0 as g ij i j = 1
s
j
g ij i
+ K j = 0
s
g ij i
j
+ K j is orthogonal to i .
s
Geometry
137
r
Now, we define a unit vector v , with components v j , by the formula
1 j
j
v i = s + K
...(10)
r
r
i
r
+ Ki the vector v will be orthogonal to both and .
s
To choose the sign of in such a way that
where =
g eijk i j v k = 1
r r
so that the triad of unit vectors , and r forms a right handed system of axes.
Since eijk
xi
g
=
is a relative tensor of weight 1 and
xj
...(11)
obsolute tensor and hence left hand side of equation (11) is an invariant v k in equation (11) is determined
by the formula
ijk
v k = i j
...(12)
1 ijk
2 = ds
s
ds
138
where
r
r
c is a vector perpendicular to .
With each point P of C we can associate a constant K , such that c K = is a unit vector..
Since
r
c
r
=
K
1 r
r
a , from (15)
=
K s
r
r
= a , since = 1
K s
1 i
i
i
or
= K , K > 0 whereK =
K s
s
s
i
1
i
i
i
i
+
K
i
i
i
(ii) = s
= K where = s + K
or
s
(iii)
= k
s
First two formulas have already been derived in article (7.7), equation (8) and (10).
Proof of (iii)
From equation (12), article (7.7), we have
i
(i)
i =
ijk i j = k
where i , i , k are mutually orthogonal.
Taking intrinsic derivative with respect to s, we get
j
i
k
j + ijk i
=
s
s
s
From formulas (i) and (ii), we get
ijk
ijk K i j + ijk i ( j K j ) =
ijk i ( j K j ) =
ijk i j K ijk j i =
k
s
k
,
s
k
s
Since ijk i j = 0
Geometry
139
ijk i j = 0
Since
k
s
ijk i j =
k
s
k
k
=
s
or
i
= i
s
This is the proof of third Serret-Frenet Formula
Expanded form of Serret-Frenet Formula.
(i)
(ii)
(iii)
di i j dx k
i
+
ds j k ds = K
or
d 2 xi i dx j dx k
+
= K i
ds 2 j k ds ds
di i j dx k
+
i
i
ds jk
ds = K
dv i i j dx k
i
+
ds jk
ds =
EXAMPLE 8
Consider a curve defined in cylindrical coordinates by equation
x1 = a
2
x = ( s )
x3 = 0
so that g11 = 1,
g22 = (x1 )2 ,
g33 = 1,
gij = 0,
i j
It is easy to verify that the non-vanishing Christoffel symbols are (see Example 3, Page 61)
1
2
2
1
= x1 , = = 1 .
22
12
21
x
140
d
dx i
The components of the tangent vector to the circle C are i =
so that 1 = 0, 2 =
,
ds
ds
3 = 0.
Since is a unit vector, gijij= 1 at all points of C and this requires that
(x1 )2 dds
2 d
= a =1
ds
d
1
So, = 2 and by Serret-Frenet first formula (expanded form), we get
ds
a
1 2 dx 2
d1 1 j dx k
1
+
K = ds j k ds = 2 2 ds =
a
2 2 dx1
d2 2 j dx k
K2 = ds + j k ds = 2 1 ds = 0
d3 3 j dx k
K = ds + j k ds = 0
1
, 1 = 1, 2 = 0, 3 = 0
a
s1 s s 2,
(i = 1, 2, 3).
dAi i dx
+
A
ds
ds = 0
(1)
We shall make use of equation (1) to obtain the equations of a straight line in general curvilinear
r
coordinates. The characteristic property of straight lines is the tangent vector to a straight line is
r
directed along the straight line. So that the totality of the tangent vectors forms a parallel vector field.
Geometry
141
i
Thus the field of tangent vector =
dxi
must satisfy equation (1), we have
ds
d 2 xi i dx dx
i
+
=
= 0
s
ds 2 ds ds
The equation
d 2 xi i d x d x
+
EXERCISE
1. Show that
d (gij AiB j )
dt
2. Show that Ai , j A j , i =
= g ij
A i j
B i
B + g ij A i
t
t
Ai A j
x j xi
d ( gij Ai B j )
dxk
i
i
= Ai,k B + Bi,k A
5. Show that
2 i
dK i
K ( i Ki )
=
ds
ds 2
2i
s2
2i
d i
dK i
( K 2 + 2 )i
ds
ds
d
= ( K v )
ds
s2
i
6. Find the curvature and tension at any point of the circular helix C whose equations in cylindrical
coordinates are
C : x1 = a, x2 = , x 3 =
Show that the tangent vector at every point of C makes a constant angle with the direction of X 3 axis. Consider C also in the form y1 = a cos, y2 = a sin, y3 = . Where the coordinates yi are
rectangular Cartesion.
CHAPTER 8
ANALYTICAL MECHANICS
8.1 INTRODUCTION
Analytical mechanics is concerned with a mathematical description of motion of material bodies subjected
to the action of forces. A material body is assumed to consist of a large number of minute bits of matter
connected in some way with one another. The attention is first focused on a single particle, which is
assumed to be free of constraints and its behaviour is analyzed when it is subjected to the action of
external forces. The resulting body of knowledge constitutes the mechanics of a particle. To pass from
mechanics of a single particle to mechanics of aggregates of particles composing a material body, one
introduces the principle of superposition of effects and makes specific assumptions concerning the
nature of constraining forces, depending on whether the body under consideration is rigid, elastic,
plastic, fluid and so on.
8.2 NEWTONIAN LAWS
1. Every body continues in its state of rest or of uniform motion in a straight line, except in so
far as it is compelled by impressed forces to change that state.
2. The change of motion is proportional to the impressed motive force and takes place in the
direction of the straight line in which that force is impressed.
3. To every action there is always an equal and contrary reaction; or the mutual actions of two
bodies are always and oppositely directed along the same straight line.
The first law depends for its meaning upon the dynamical concept of force and on the kinematical
idea of uniform rectilinear motion.
The second law of motion intorduces the kinematical concept of motion and the dynamical idea
of force. To understand its meaning it should be noted that Newton uses the term motion in the sense
of momentum, i.e., the product of mass by velocity, this, "change of motion" means the time of change
of momentum.
In vector notation, the second law can be stated as
r
d (mv )
r
(1)
F = dt
Analytical Mechanics
143
(1)
and the curve C the trajectory of the particle. Let at time t, particle is at P {x i (t )}.
If v i be the component of velocity of moving particle then
dx i
dt
and if ai be the component of acceleration of moving particle then
vi =
ai
v i dv i i j d x k
= t = dt + j k v d t
ai
d 2 x i i dx j dx k
+
=
dt 2 j k dt dt
(2)
(3)
v i
i
is the intrinsic derivative and the are the Christoffel symbols calculated from the
t
jk
metric tensor gij
If m be the mass of particle. Then by Newtons second law of motion
where
Fi = m
v i
= ma i
t
(4)
r
We define the element of work done by the force F in producing a displacement drr by invariant
r
dW = F .drr .
r
Since the components of F and drr are F i and dxi respectively..
144
Then
dW = gij F i dxj
(5)
j
i
= F j dx where F j = gij F
The work done in displacing a particle along the trajectory C, joining a pair of points P 1 and P 2, is
line integral
W=
P2
P1
Fi dx i
(6)
W=
m g ij
v i
dx j
t
m g ij
v i dx j
dt
t dt
P2
P1
P2
P1
P2
P1
v i j
v dt
t
m g ij
(7)
d
( g v iv j )
dt ij
t
d
vi j
( g ij v i v j ) = 2 g ij
v
or
dt
t
vi j 1 d
(g ij v i v j )
g ij
v =
2 dt
t
using this result in equation (7), we get
W=
P2 m
P1
d
( g v i v j ) dt
2 dt ij
m
[g v i v j ]PP12
2 ij
Let T2 and T1 is kinetic energy at P 2 and P 1 respectively.
W = T2 T1
W=
m
m v2
gij v i v j =
is kinetic energy of particle.
2
2
We have the result that the work done by force F i in displacing the particle from the point P 1 to
1
the point P 2 is equal to the difference of the values of the quantity T =
mv 2 at the end and the
2
beginning of the displacement.
where T =
P2
P1
(8)
Analytical Mechanics
145
of the work function W. The negative of the work function W is called the force potential or
potential energy.
We conclude from equation (8) that
V
Fi = i
(9)
x
i
where potential energy V is a function of coordinates x . Hence, the fields of force are called conservative
V
.
x i
THEOREM 8.2 A necessary and sufficient condition that a force field Fi, defined in a simply
connected region, be conservative is that Fi,j = Fj,i.
if F i =
V
x i
Now,
Fi k
Fk
x j i j
V
i
x k F
F i,j =
k
x j
i j
F i,j =
k
2V
Fk
j
i
x x j i
(1)
and
F j,i =
F j
k
Fk
x j i
i
Similarly,
F j,i =
2V
k
Fk
i
j
x x
ji
F j k
Fi
as due to symmetry..
j =
x i i j
x
(2)
146
Take F i =
V
x i
Then
Fi
2V
j i
j =
x
x x
2V
= i j
x x
V
=
x i x j
F j
Fi
j =
x
x i
Fi =
V
.
x i
Hence, F i is conservative.
8.5 LAGRANGEAN EQUATIONS OF MOTION
Consider a particle moving on the curve
C : x i = x i (t )
At time t, let particle is at point P (xi).
The kinetic energy T =
Since x& i = v i.
1
mv 2 can be written as
2
1
i j
T = m g ij x& x&
2
1
m g jk x& j x& k
2
Differentiating it with respect to x& i , we get
or
T=
(1)
x& j k
&k
1
T
j x
&
&
m
g
x
+
x
jk
= 2
x& i
i
&
x
x& i
1
m g jk ij x& k + ki x& j
2
1
1
m g jk ij x& k + m g jk ki x& j
2
2
1
m gik x& k + g ji x& j
2
1
j
j
= m (gij x& + gij x& )
2
)
as
gij = g ji
Analytical Mechanics
147
T
= m g ij x& j
x& i
T
= m g ik x& k
x& i
or
(2)
d T
d
g x& k
= m
dt x& i
dt ik
d
k
k
= m g ik x& + g ik &x&
dt
g ik dx j k
k
m
= x j dt x& + g ik &x&
d T
g
j k
k
= m ikj x& x& + m g ik &x&
dt x& i
x
1
m g jk x& j x&k
2
Differentiating it with respect to xi, we get
T
1 g jk j k
m
x& x&
i =
x
2 x i
Now,
(3)
Since T =
d T
dt x& i
(4)
T
g
1 g
i = m ikj x& j x&k + m g ik&x&k m jki x& j x& k
x
2 x
x
k
= m g ik &x& +
1 gik j k 1
1 g
m j x& x& + m g ik x& j x& k m jki x& j x& k
2 x
2
2 x
k
= m g ik &x& +
1 g ik j k 1 g ij k j 1 g jk j k
m
x& x& + m k x& x& m i x& x&
2 x j
2 x
2 x
1 gik g ij g jk j k
k
x& x&
= m g ik &x& + 2 m j + k
x
x i
x
= m g ik &x&k + m [ jk , i ] x& j x& k
= m g ik &x&k + m g il gil [ jk , i ]x& j x& k
= m g il &x&l + m g il gil [ jk , i]x& j x& k
148
l
= m g il &x&l + x& j x& k ,
jk
d T T
= m g il a l ,
dt x& i x i
where al is component of acceleration
or
l
as g il [ jk , i] =
j k
l j k
l
l
Since a = x&& + j k x& x&
d T T
= m ai
dt x& i x i
d T T
= Fi
(5)
dt x& i x i
where F i = m ai, component of force field. The equation (5) is Lagrangean equation of Motion.
For a conservative system, F i =
d T
dt x& i
V
. Then equation (5) becomes
x i
V
T
i = i
x
x
d T (T V )
=0
(6)
dt x& i
x i
Since the potential V is a function of the coordinates xi alone. If we introduce the Lagrangean
function
L=TV
Then equation (6) becomes
or
d L L
=0
dt x& i x i
(7)
EXAMPLE 1
Show that the covariant components of the acceleration vector in a spherical coordinate system
with
ds 2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2 are
a1 = &x&1 x1 (x& 2 )2 x1 (x& 3 sin x 2 ) 2
and
a2 =
d
( x1 )2 x& 2 (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2
dt
a3 =
d
(x1 sin x 2 ) 2 x& 3
dt
Analytical Mechanics
149
Solution
In spherical coordinate system, the metric is given by
ds2 = (dx1 ) 2 + ( x1 dx 2 )2 + (x1 ) 2 sin 2 x 2 (dx 3 ) 2
If v is velocity of the paiticle then
2
2
2
3
ds dx1
1 2 dx
1
2 2 dx
2
=
+
(
x
)
+
(
x
sin
x
)
v =
dt
dt
dt dt
= F i and m ai = F i
dt x& i x i
where F i and ai are covariant component of force field and acceleration vector respectively.
So,
T=
d T T
= m ai
dt x& i x i
(2)
Take i = 1,
m a1 =
d T T
dt x&1 x 1
m a1 =
1 d
m
m (2x&1 ) 2x 1 ( x& 2 )2 + 2 x1 (sin x 2 ) 2 ( x& 3 )2
2 dt
2
a1 =
dx&1
x1 ( x& 2 )2 + x1 (sin x 2 ) 2 ( x& 3 )2
dt
d T
dt x& 2
m a2 =
1 d 1 2 2 1
m ( x ) 2 x& m 2( x1 )2 sin x 2 cos x 2 (x&3 ) 2
2 dt
2
a2 =
(1)
T
2
x
d
( x1 )2 x& 2 (x1 ) 2 sin x 2 cos x 2 ( x& 3 )2
dt
150
Take i = 3
m a3 =
d T T
dt x&3 x 3
m a3 =
1 d
m 2( x& 3 ) ( x1 sin x 2 )2 0
2 dt
a3 =
d 3 1 2
x& ( x ) (sin x 2 ) 2
dt
EXAMPLE 2
Use Lagrangean equations to show that, if a particle is not subjected to the action of forces then
its trajectory is given by yi = ait + bi where ai and bi are constants and the yi are orthogonal cartesian
coordinates.
Solution
If v is the velocity of particle. Then we know that,
v 2 = g ij y& i y& j
where yi are orthogonal cartesian coordinates.
Since
gij = 0, i j
gij = 1, i = j
So,
v 2 = ( y& i ) 2
But,
1 2
mv , T is kinetic energy..
2
1
i 2
T = m( y& )
2
The Lagrangean equation of motion is
T=
d T T
= Fi
dt y& i yi
Since particle is not subjected to the action of forces.
So, F i = 0
Then
d 1
m 2 y& i 0 = 0
dt 2
dy& i
=0
dt
Analytical Mechanics
151
dy& i
=0
dt
or
y& i = ai
yi = ait + bi
i
i
where a and b are constant.
EXAMPLE 3
Prove that if a particle moves so that its velocity is constant in magnitude then its acceleration
vector is either orthogonal to the velocity or it is zero.
Solution
If v i be the component of velocity of moving particle then
vi
dx i
=
dt
or
v i = x& i
(g v i v j ) = 0
t ij
v i j
v j
g ij
v + vi
t
t
=0
g ij
v i j
v j
v + g ij v i
=0
t
t
g ij
v i j
v i
v + g ji v j
= 0, (Interchange dummy index i and j in second term)
t
t
2 g ij
g ij
v i j
v = 0 as gij = gji
t
vi j
v =0
t
v i
v i
= 0.
is either orthogonal to v i or zero i.e.,
t
t
152
=0
dt x& i x i
(1)
1
m y& i y& i .
2
Hence, the equation (1) becomes m &y&i = 0. Integrating it we get yi = ait + bi, which represents
a straight line.
(ii) Simple Pendulum
Let a pendulum bob of mass m be supported by an extensible string. In spherical coordinates, the
metric is given by
If xi be rectangular coordinate system, then T =
ds2 = dr 2 + r 2 d 2 + r 2 sin 2 d 2
If T be the kinetic energy, then
T=
1 2 1
mv = m (r&2 + r 2 & 2 + r 2 sin 2 & 2 )
2
2
Y2
r
R
mg cos
Y1
mg sin
Y3
P
mg
Fig. 8.1.
= Fi
dt x& i x i
x1 = r , x 2 = , x 3 = .
So, take
x1 = r
i = 1, 2, 3
(1)
Analytical Mechanics
153
d T T
= mg cos R
dt r& r
from (1), we have
R
&r& r&2 r sin 2 &2 = g cos
m
(2)
Take x2 = , we have
r && + 2r&& r sin cos & 2 = g sin
(3)
(4)
If the motion is in one plane, we obtain from equations (2), (3), and (4), by taking & = 0.
R
r&& r & 2 = g cos m
r && + 2r&& = g sin
g
If r& = 0, we get, && = sin which is equation of simple pendulum supported by an
r
inextensible string. For small angles of oscillation the vibration is simple harmonic. For large vibration
the solution is given in the term of elliptic functions.
8.7 HAMILTONS PRINCIPLE
If a particle is at the point P 1 at the time t1 and at the point P 2 at the time t2, then the motion of the
particle takes place in such a away that
t2
t1
(T + Fi xi ) dt = 0
where xi = xi (t) are the coordinates of the particle along the trajectory and xi + x i are the coordinates
along a varied path beginning at P 1 at time t1 and ending at P 2 at time t2.
Proof: Consider a particle moving on the curve
t1 t t 2
C : x i = x (t ),
At time t, let particle is at P(xi). If T is kinetic energy. Then
i
T=
1
m g ij x& i x& j
2
or T = T ( x i , x& i ) i.e, T is a function of xi and x& i . Let C ' be another curve, joining t1 and t2 close
to be C is
C' : x i (, t ) = x i (t ) + x i (t)
154
At t1 and t2
xi = x i = x i + x i
x i (t1 ) = 0 and x i (t 2 ) = 0
But T = T ( x i , x& i ).
If T be small variation in T.. Then
T i T i
T = x&i x& + x i x
Now,
{(T + F )x }dt =
t2
t1
t1
T i
x dt +
x i
t2
t1
t2
t1
T i
x& dt +
x&i
t2
t1
T i
x dt +
x i
T i 2
x& i x
t1
t2
t1
T i 2
then & i x = 0.
x
t1
So,
(T + F x )dt =
(T + F x )dt =
t1
t1
+
t2
t2
t1
t2
t1
x dt
t2
1
d
dt
T
i
x&
i
x dt
Fi x i dt
t2 T
t1
i
d T
i i + Fi x dt
x dt x&
= Fi
dt x& i x i
or
t1
Fi x i dt
d T i
x dt +
dt x&i
Since x i (t1 ) = 0, x i (t 2 ) = 0.
t2
t2
T
as 1st term
x& i
T
T
i x i + i x& i + Fi x i dt
x&
x
t2
T d T
= Fi
x i dt x& i
t2
t1
Fi xi dt
Analytical Mechanics
155
So,
t2
t1
t2
t1
(T + Fi x i ) dt =
t2
t1
[Fi + Fi ] xi dt
(T + Fi x i ) dt = 0
Proved.
or
T=
1
m g ij x& i x& j
2
T=
1
m g ij v i v j
2
As T is invariant. Then
Taking intrinsic derivative with respect to t, we get
dT
T
=
dt
t
=
m gij v i v j
t 2
v i j
1
v j
m g ij
v + vi
2
t
t
1 v i j
v j i
m gij
v + g ij
v
2
t
t
1 vi j
vi j
= 2 m gij t v + g ji t v , Since i and j are dummy indices.
1
v i j
m
2
g
v
=
ij
2
t
dT
v i j
v
= m gij
dt
t
or
dT
v j i
m
g
v
=
ij
dt
t
as g ij = g ji
156
j i
= m g ij a v as
dT
= m ai v i,
dt
v j
= aj
t
since gij aj = ai
dT
= F i v i,
dt
Since F i = m ai is a covariant component of force field.
But given F i is conservative, then
V
, where V is potential energy..
x i
Fi =
So,
dT
V i
v
=
dt
x i
V dx i
x i dt
dT
dV
=
dt
dt
dT dV
d
+
= 0 (T + V ) = 0
dt
dt
dt
T + V = h, where h is constant.
Proved.
P2
mv.ds
(1)
p1
A=
P2
p1
m g ij
t ( P2 )
t ( P1 )
dx i
dx j
dt
m gij
dx i dx j
dt
dt dt
Analytical Mechanics
Since T =
157
1
dx i dx j
m g ij
, we have
2
dt dt
A=
t ( P2 )
2T dt
t ( P1 )
This integral has a physical meaning only when evaluated over the trajectory C, but its value can
be computed along any varied path joining the points P 1 and P 2.
Let us consider a particular set of admissible paths C' along which the function T + V, for each
value of parameter t, has the same constant value h. The integral A is called the action integral.
The principle of least action stated as of all curves C' passing through P 1 and P 2 in the
neighbourhood of the trajectory C, which are traversed at a rate such that, for each C', for every value
of t, T + V = h, that one for which the action integral A is stationary is the trajectory of the particle.
8.10 GENERALIZED COORDINATES
In the solution of most of the mechanical problems it is more convenient to use some other set of
coordinates instead of cartesian coordinates. For example, in the case of a particle moving on the
surface of a sphere, the correct coordinates are spherical coordinates r, , where and are only
two variable quantities.
Let there be a particle or system of n particles moving under possible constraints. For example, a
point mass of the simple pendulum or a rigid body moving along an inclined plane. Then there will be
a minimum number of independent coordinates required to specify the motion of particle or system of
particles. The set of independent coordinates sufficient in number to specify unambiguously the system
configuration is called generalized coordinates and are denoted by q 1 , q 2 , ... q n where n is the total
number of generalized coordinates or degree of freedom.
Let there be N particles composing a system and let x(i ) , (i = 1, 2,3), ( = 1, 2,...N ) be the positional
coordinates of these particles referred to some convenient reference frame in E 3. The system of N free
particles is described by 3N parameters. If the particles are constrained in some way, there will be
certain relations among the coordinates x(i ) and suppose that there are r such independent relations,
f i ( x1(1) , x(21) , x 3(1 ) ; x1( 2 ) , x(22 ) , x(32 ) ;...x1( N ) x(2N ) x(3N ) ) = 0, (i = 1, 2, ..., r)
(1)
By using these r equations of constraints (1), we can solve for some r coordinates in terms of the
remaining 3N r coordinates and regard the latter as the independent generalized coordinates qi. It is
more convenient to assume that each of the 3N coordinates is expressed in terms of 3N r = n
independent variables qi and write 3N equations.
x(i ) = x(i ) (q1 , q 2 ,...,q n , t )
(2)
where we introduced the time parameter t which may enter in the problem explicitly if one deals with
moving constraints. If t does not enter explicitly in equation (2), the dynamical system is called a
natural system.
The velocity of the particles are given by differentiating equations (2) with respect to time. Thus
x&(i ) =
x i( )
q j
q& j +
x i( )
t
(3)
158
(4)
Differentiating it we get
f j i f j
q& +
(5)
t = 0
q i
It is clear that they are integrable, so that one can deduce from them equations (4) and use them
to eliminate the superfluous coordinates .
In some problems, functional relations of the type
(6)
F j (q1 , q 2 ,...,q k ; q&1 ,...,q& k , t ) = 0, ( j = 1, 2, 3, ..., m)
arise which are non-integrable. If non-integrable relations (6) occurs in the problems we shall say that
the given system has k m degrees of freedom, where m is the number or independent non-integrable
relations (6) and k is the number of independent coordinates. The dynamical systems involving nonintegrable relations (6) are called non-holonomic to distinguish them from holonomic systems in which
the number of degrees of freedom is equal to the number of independent generalized coordinates.
In other words, a holonomic system is one in which there are no non-integrable relations involving
the generalized velocities.
8.11 LAGRANGEAN EQUATIONS IN GENERALIZED COORDINATES
Let there be a system of particle which requires n independent generalized coordinates or degree of
freedom to specify the states of its particle.
The position vectors xr are expressed as the function of generalized coordinates q i , (i = 1, 2,...,n)
and the time t i.e.,
1
2
n
(r = 1, 2, 3, )
xr = xr (q , q ,...,q , t );
The velocity x& r of any point of the body is given by
x r dq j x r
x& = q j dt + t
r
x r j x r
q& +
,
t ( j = 1, 2, ..., n)
q j
(1)
involve n independent parameters qi. The velocities x& r in this case are given by
x r j
q& ,
=
x&
q j
r
(r = 1, 2, 3; j = 1, 2, ..., n)
(2)
Analytical Mechanics
159
(k = 1, 2, ..., n)
(3)
1
m g rs x& r x& s ,
2
(r,s = 1,2,3,)
(4)
where m is the mass of the particle located at the point xr. The grs are the components of the metric
tensor.
Substituting the value of x& r from equation (2), then equation (4) becomes
1
x r x s i j
q& q&
T = 2 m g rs i
q q j
T=
1
a q&i q& j
2 ij
aij = m g rs
where
x r r s
,
q i q j
(5)
(r, s = 1, 2, 3), (i, j = 1, ..., n)
1
a q&i q& j is an invariant and the quantities aij are symmetric, we conclude that the aij
2 ij
are components of a covariant tensor of rank two with respect to the transformations (3) of generalized
coordinates.
Since T =
Since the kinetic energy T is a positive definite form in the velocities q& i , | aij |> 0. Then we
construct the reciprocal tensor aij .
Now, from art. 8.5, Pg. 146, by using the expression for the kinetic energy in the form (5), we
obtain the formula,
d T
dt q& i
T
l l j k
(6)
l
where the Christoffel symbol are constructed from the tensor akl.
jk
Put
l
q&&l + q& j q& k = Q l
jk
so, the equation (6), becomes
d T
dt q& i
T
l
qi = ail Q
= Qi (i = 1, 2, ..., n)
(7)
160
x& r x r x& r
2 xr j
x&r
d x r
&
=
,
=
q
=
Now, from the realtions & j
and qi dt q i and using equations
q
q j qi x i q j
(2) and (4).
Then by straightforward calculation, left hand member of equation (7) becomes
d T
dt q& i
T
x r
m
a
=
r
qi
q i
(8)
(9)
Fr's
where
are the components of force F acting on the particle located at the point P..
From the equation (9), we have
m ar
x r
=
q i
Fr
x r
q i
Fr
x r
q i
qi =
(10)
Fr
x r
q i
qi = Qi
(11)
qi = 0
(12)
Analytical Mechanics
161
Since L (q, q& ) is a function of both the generalized coordinates and velocities.
L i L i
dL
= & i q&& + i q&
dt
q
q
from (12), we have
L
d L
i =
q
dt q& i
(13)
i
q&
d L i
q&
dt q& i
(14)
T=
1
a q& i q& j .
2 ij
(2)
162
r
If the components of F relative to an arbitrary curvilinear coordinate system X are denoted by
F i then the covariant derivative of F i is
F ,ij =
F i i k
+ F
x j k j
r
The invariant F, ij in cartesian coordinates represents the divergence of the vector field F .
Also,
r
i j
i
j
F n = g ij F n = F ni since g ij n = ni
Hence we can rewrite equation (1) in the form
F dV
i
,i
F n dS
i
(3)
and
xi
Put F i = i and from the divergence of we get
= i =
= i =
x i
F,i j = g ij Fi , j = g ij (i , j + i j )
Substituting this in equation (3), we get
ij
(i , j + i j )dV
n dS
i
(4)
Since = i , then
g ij i , j = 2
(5)
(g
ij
i , j + g ij i . J ) dV
n dS
S
( + ) dV = n dS
2
dV = n dV
2
V
i
where n = i n =
.
n
(6)
Analytical Mechanics
163
dV
2
n dV
S
(7)
( ) dV
2
n n dS
(8)
2
k
ij
2 = g i j k
i j x
x x
i
and the divergence of the vector F is
F,ii =
F i i
+ F
x i j i
(9)
(10)
log g
=
j
i
xj
The equation (10) becomes
But we know that
F,ii
or
ij
If putting F i = g
F,ii
F i
+ j log g F j
=
i
x
x
1 ( g F i )
x i
g
g g ij j
1
x
g ij j , i =
i
x
g
(11)
(12)
164
g g ij j
1
x
2 = g ij j ,i =
i
g
x
It is expansion form of Laplacian operator.
(iv) Stoke's Theorem
r
Let a portion of regular surface S be bounded by a closed regular curve C and let F be any vector point
function defined on S and on C. The theorem of Stokes states that
r
r
n. curl F ds = F . ds
(13)
r
where is the unit tangent vector to C and curl F is the vector whose components in orthogonal
cartesian coordinates are determined from
e1
r
curl F = x1
F1
e2
x 2
F2
e3
r
x 3 = F
F3
(14)
ijk F j , k ni ds
S
The integral
F dx
i
C
i
Fi
dx i
ds
ds
(16)
r
is called the circulation of F along the contour C.
Analytical Mechanics
165
r
F .n dS =
m cos r 2 dw
r 2 cos
r dw
and dw is the solid angle subtended by dS.
cos
Thus, we have,
r
F .n dS
m dw = 4 m
=
where dS =
(1)
dS
r
d
P
m
S
Fig. 8.2.
i =1
mi cos i
ri 2
r
F .n dS = 4
(2)
i =1
The result (2) can be easily generalized to continuous distributions of matter whenever such
distribution no where melt the surface S.
The contribution to the flux integration from the mass element dV contained within V, is
cos dV
r
dS
F .n dS =
r2
S
r
F .n dS =
cos dV
r2
V
dS
(3)
166
where
denotes the volume integral over all bodies interior to S. Since all masses are assumed to be
interior to S,r never vanishes. So that the integrand in equation (3) is continuous and one can interchange
to order of integration to obtain
r
F .n dS =
cos dS
dV
s
r2
(4)
cos dS
= 4. Since it represents the flux due to a unit mass contained within S.
r2
S
Hence
But
F.n dS
4 dV = 4 m
(5)
S"
r
where the subscript i on F n refers to the flux due to the masses located inside S and m is the total
mass within S. On the other hand, the net flux over S' due to the masses outside S, by Gauss's theorem
is
r
( F n)o ds = 0
S'
r
where the subscript o on F n refers to the flux due to the masses located outside S.
Now if we S' and S" approach S, we obtain the same formula (5) because the contribution to the
r
total flux from the integral (F n)o dS is zero.
S'
r
F .n ds =
r
div F dV
F.n ds = 4 dV
from these, we have
Analytical Mechanics
167
r
(div F 4 ) dV = 0
Since this relation is true for an arbitrary V and the integrand is piecewise continuous, then
r
div F = 4
By the definition of potential function V, we have
r
F = V
and
div V = 2V
So,
r
div F = 4
div ( V ) = 4
2V = 4
Q (y )
r
P(x)
n
Fig. 8.3.
168
1
has a discontinuity at x i = y i , delete the point P(x) from region of integration by
r
surrounding it with a sphere of radius and volume V'. Apply Greens symmetrical formula to the
1
region V V' within which
and V possess the desired properties of continuity..
r
2
2 1
= 0.
In region V V', =
r
Then equation (1) becomes
Since
1 2
dV =
r
V V'
1
1r
dS +
r n
n
S'
1
1r
ds
r n
n
(2)
where n is the unit outward normal to the surface S + S' bounding V V' . S' being the surface of the
= .
sphere of radius and
n
r
Now
S'
1
1r
dS =
r n
S'
1
1r
r r
r
dS
2
r dw
2
r r r
S'
r r + dw
S'
S'
1
( 1r )
dw 4
r n
n
r r =
S'
(3)
where is the mean value of V over the sphere S' and w denote the solid angle.
1
2
3
Let ( x , x , x ) = (P ) as r 0 then as 0 from (3), we have
S'
1
1r
r n
n = 4 (P )
Since 0 then
1 2
dV =
r
r dV
V'
= 0.
1
1r
dS 4 (P )
r n
n
Analytical Mechanics
169
1r
1 1 2
1 1
1
dV
+
dS
dS
(P ) = 4 r
4 S r n
4 S n
V
(4)
m
m
and
r r 2
r
(5)
where m is constant.
If integration in equation (4) is extended over all space, so that r . Then, using equation (5),
equation (4) becomes
(P ) =
1 2
dV
4 r
(6)
dV
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
n dS
i
dV
2
where i =
.
x i
CHAPTER 9
...(1)
The intrinsic derivative (or desired vector) of t i along the curve C is called the first curvaturee
r
r
vector of curve C relative to Vn and is denoted by p . The magnitude of curvature vector p is called
first curvature of C relative Vn and is denoted by K:
So,
K=
g ij p i p j
i
P = t , j ds
t i
i
i dx
+
t
= j
j ds
x
j
t i dx j
dx i
+
t
ds
x j ds
dx dx j i
dt i
= ds + ds ds j
d 2 xi
ds
dx j dx k i
ds ds k
171
d 2 xi
dx j dx k i
pi = ds 2 + ds ds j k as j k = k j
t1
to
f (x i , x&i ) dt
to be staionary are
f
x
d f
=0
dt x& i
dxi
x& = dt i = 1, 2, 3,...
Proof: Let C be a curve in a Vn and A, B two fixed points on it. The coordinates x i of the current point
P on C are functions of a single parameter t. Let t0 and t1 be the values of the parameter for the points
A and B respectively.
To find the condition for the integral
where
t1
t0
f (x i , x&i ) dt
...(1)
to be stationary.
Let the curve suffer an infinitesimal deformation to C , the points A and B remaining fixed while
the current points P(xi) is displaced to P' (xi + i) such that i = 0 at A and B both.
172
P'
Fig. 9.1
F (x
t1
t0
& i dt
+ i , x& i +
By Taylor's theorem
f
f
+ k +
y
x
F ( x + h , y + k ) = f ( x, y ) + h
Then
I =
I =
t1
F i F i
i
i
F ( x , x& ) + i + i & + dt
x
x
t0
t1
t0
F ( x i , x& i ) dt +
t1 F
t 0 x i
i +
F i
& dt
x& i
F
f
i i + i & i dt
t0 x
x&
t1
I = I I =
t1
t0 x i
i +
F i
& dt
x&i
...(2)
zi & j
& i =
x
x j
where
Now,
F i 1
&
dt
= x&i
t0 x
&i
t0
t1 F
t1
t0
t1
t0
d F i
dt
dt x&i
d F i
dt
dt x&i
...(3)
173
t
F
1
since i i (t ) = 0, i (t1 ) = i (t0 ) = 0
x&
t0
ti F
t0 x i
t1 F
t0 x i
d F i
dt
dt x&i
...(4)
d F i
dt = 0
dt x&i
Since i are arbitrary and hence the integrand of the last integral vanishes, so that
F d F
= 0, (i = 1, 2,..., n)
xi dt x&i
Hence the necessary and sufficient condition for the integral (1) to be stationary are
(5)
F d F
= 0,
(i = 1, 2, , n)
x i dt dx i
These are called Euler's conditions for the integral I to be stationary.
9.4 DIFFERENTIAL EQUATIONS OF GEODESICS
To obtain the differential equations of a geodesic in a Vn , using the property that it is a path of minimum
(or maximum) length joining two points A and B on it.
Proof: Consider a curve C in Vn joining two fixed points A and B on it and x i (t ) be the coordinates of
point P on it.
The length of curve C is
s=
ds
=
dt
gij
gij
dxi d x j
dt
dt dt
...(1)
dxi d x j
dt dt
Put
ds
=
dt
or
gij
d xi d x j
= F (say)
dt dt
...(2)
g ij x& i x& j = F
s& =
B
A
F dt
...(3)
174
Since curve C is geodesic, then the integral (3) should be stationary, we have from Euler's
condition
F
x
d F
=0
dt x& i
...(4)
and
1 g ij i j
x& x&
2 s& x k
F
1
1
2 gik x& i = g ik x&i
k =
x&
2s&
s&
d F
1
1 g ik j i 1
i
x& x& + g ik &x& i
= 2 &s& g ik x& +
dt x& k
s& x j
s&
s&
g
&s&
1 g ij i j
x& x& = 0
g ik x& i + ikj
s&
2 x k
x
g ik &x&i
g ik &x&i
&s&
g x& i + [k , ij ] x&i x& j = 0
&s ik
multiplying it by g km , we get
g km gik &x&i
&s& km
g g ik x&i + g km [k , ij ] x& i x& j = 0
s&
m
km
g km gik = im and g [k , ij] = i j
But
x&&m
d 2xm
dt
&s& m m i j
x& + x& x& = 0
s&
i j
&s& dx m m dx j dxk
Replacing dummy
+
= 0 index i by k
s& dt j k dt dt
m dx j dx k
+
= 0 ...(6)
j k ds ds
...(5)
175
dx m
ds
=0
,k
Then the intrinsic derivative (or derived vector) of the unit tangent to a geodesic in the direction of the
curve is everywhere zero. In otherwords, a geodesic of Vn is a line whose first curvature relative to
Vn is identically zero.
THEOREM 9.2 To prove that one and only one geodesic passes through two specified points lying in
a neighbourhood of a point O of a Vn .
OR
To prove that one and only one geodesic passes through a specified point O of Vn in a prescribed
direction.
Proof: The differential equations of a geodesic curve in a Vn are
d 2xm
ds 2
m dx j dx k
+
=0
j k ds ds
These equations are n differential equations of the second order. Their complete integral involves 2n
arbitrary constants. These may be determined by the n coordinates of a point P on the curve and the n
components of the unit vector in the direction of the curve at P. Thus, in general, one and only one
geodesic passes through a given point in a given direction.
9.5 GEODESIC COORDINATES
A cartesian coordinate system is one relative to which the coefficients of the fundamental form are
constants. Coordinates of this nature do not exists for an arbitrary Riemannian V n.. It is, however,
possible to choose a coordinate system relative to which the quantities gij are locally constant in the
neighbourhood of an arbitrary point P 0 of V n. Such a cartesian coordinate system is known as geodesic
coordinate system with the pole at P 0.
The quantities g ij are said to be locally constants in the neighbourhood of a point P0 if
g ij
= 0 at P0
xk
and
g ij
xk
0 elsewhere
176
The covariant derivative of Aij at P0 with respect to x k reduces to the corresponding ordinary
derivatives. Hence
Aij , k =
Aij
x k
at P0
THEOREM 9.3 The necessary and sufficient condition that a system of coordintes be geodesic with
pole at P0 are that their second covariant derivatives with respect to the metric of the space all vanish
at P0.
Proof: We know that (equation 8, Pg. 65)
2xs
x j x j
x s
x j x i
or
x s k x p x q s
x k i j x i x j p q
x s k x p x q s
=
x k i j x i x j p q
...(1)
x j x i
x s
=
x k
x p x q s
=
i
j
x x p q
x j
=
k x p x q s
i
j
i j x x p q
x s
x i
x s
x k
(x ) x
s
,i
= ( x ,is ), j
k
x s
s
since k = x, k at P0
i
j
x
k
since i j = 0 at P0
s
,k
Thus,
x,sij
x p x q
x i x j
p q
Necessary Condition
Let x s be a geodesic coordinate system with the pole at P0 so that
s
= 0 at P0
p q
x,sij = 0 at P0
Sufficient Condition
Conversely suppose that x,sij = 0 at P0 .
...(2)
177
s
x p
x q
0
0 at P0
= 0 at P0 , as
and
x i
x j
p q
= ds
...(1)
the subscript zero indicating as usual that the function is to be evaluated at P0 . The quantities i
represents that only one geodesic will pass through P 0 in the direction of i in V n. Let yi be the
coordinates of a point P on the geodesic C such that
yi = si
...(2)
where s is the arc length of the curve from P0 to P . The coordinates y i are called Riemannian
coordinates.
The differential equation of geodesic C in terms of coordinates y i relative to Vn is given by
d 2 y i i dy i dy j
+
=0
ds 2 j k ds ds
...(3)
i
is a christoffel symbol relative to the coordinates y i .
j k
where
i i j
= 0
j k
...(4)
=
0
as
s
j k s s
or
i i j
y y = 0
j k
...(5)
178
= 0 at P0
j k
Hence the Riemannian coordinates are geodesic coordinate with the pole at P0 .
THEOREM 9.4 The necessary and sufficient condition that the coordinates y i be Riemannian coordinates
i i j
y y = 0 hold throughout the Riemannian Vn .
j k
is that
i i j
y y = 0 (from equation 5) throughout
Proof: If y are Riemannian coordinates then the condition
j k
i
the Riemannian Vn .
d 2 y i dy i dy j
i i j
i.e.,
(dx )
1 2
1 1
= g11dx dx
g 11 = 1
...(1)
2
3
n
v i = ( 0, dx , dx ,..., dx )
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)
179
Now,
ti =
and
dx1
dx1
d 2xi
ds 2
d xi d x i
= 1
ds
dx
= 1 and
=
d 2xi
dxi
dx1
= 0 for i 1
= 0 for i = 1, 2,..., n
dx12
=0
ds 2 j k ds ds
using above results, we have
i dx 1 dx1
=0
11 ds ds
i
=0
1 1
g ij [11, j ] = 0
[11, j ] = 0
1 2 g 1 j g 11
2 x i
x j
ij
as g 0
= 0, since g 11 = i g 11 = 0
x j
So,
g 1 j
x1
= 0 for j 1
...(3)
g1 j
x1
= 0, ( j = 2, 3, , n)
ds 2 = gij dx dx
1 1
j
k
ds 2 = g 11dx dx + g jk dx dx
1 2
j
k
ds 2 = ( dx ) + g jk dx dx ; ( j = 2, 3, , n, k = 2, 3, , n)
...(4)
The line element (4) is called geodesic form of the line element.
We note that the coordinate curves with parameter x 1 are orthogonal to the coordinate curve
x i = c i (i = 1, 2, ..., n) at all points and hence to the hypersurfaces x 1 = c at each point.
Note 2: The existence of geodesic form of the line element proves that the hypersurfaces = x1 = constant form a
system of parallels i.e., the hypersurfaces = x 1 = constant are geodesically parallel hypersurfaces.
Note 1:
180
THEOREM 9.5 The necessary and sufficient condition that the hypersurfaces = constant form a
system of parallel is that ()2 = 1.
Proof: Necessary Condition
Suppose that hypersurface = constant form a system of parallels then prove that ( 2) = 1.
Let us take the hypersurface = 0 as the coordinate hypersurface x1 = 0. Let the geodesics
cutting this hypersurface orthogonally, be taken as coordinate lines of parameter x1. Then the parameters
x1 measures are length along these geodesics from the hypersurface x1 = 0. This implies the existence
of geodesic form of the line element namely
1 2
i
j
ds 2 = ( dx ) + g ij dx dx
...(1)
where i, j = 2, 3,..., n.
From (1), we have
g 11 = 1, g 1i = 0 for i 1.
ij
= g
x i dx j
x1 x1
= g ij 1i 1j
i
j
x x
()2 = g11 = 1
so
()2 = 1
Sufficient Condition
Suppose that ()2 = 1 then prove that the hypersurface = constant from a system of parallels.
Let us taken = x1 and orthogonal trajectories of the hypersurfaces = x1 constant as the
coordinate lines of parameter x1. Then the hypersurfaces
x1 = constant
xi = constant (i 1) are orthogonal to each other. The condition for this g1i = 0 for i 1 .
Now, given that ()2 = 1
g ij
=1
x i x j
g ij
x1 x1
=1
x i x j
g ij 1i 1j = 1
g11 = 1
181
Thus
g11 = 1 and g1 i = 0 for i 1.
Consequently
g 11 = 1, g = 0 , for i 1.
1i
is given by
ds 2 = (dx1 ) 2 + g ik dx i dx k ; (i, k = 2, 3, , n)
which is geodesic form of the line element. It means that the hypersurfaces = x1 = constant
form a system of parallels.
9.8 GEODESICS IN EUCLIDEAN SPACE
Consider an Euclidean space S n for n-dimensions. Let y i be the Euclidean coordinates. The differential
equation of geodesics in Euclidean space is given by
d 2 y i i dy j dy k
+
=0
ds 2 j k ds ds
...(1)
a ij
x k
=0
k
This implies that = 0, [ij, k ] = 0 relative to S n .
i
j
Then equation (1) becomes
d 2 yi
ds 2
=0
...(2)
182
THEOREM 9.6 Prove that the distance l between two points P (y i) and Q (y' i ) in Sn is given by
(y y )
n
i 2
l=
i =1
Proof: We know that geodesics in S n are straight line. Then equation of straight line in S n may be
taken as
yi = ais + bi
...(1)
Let P ( y i ) and Q ( y' i) lie on equation (1). Then
yi = ais + bi, y i = a i s + bi
y i y i = a i (s s )
Then equation (2) becomes
y i y i = a i l
n
l2
i =1
(ai ) 2 =
( y
y i )2
i =1
(a )
i 2
=1
i =1
So,
n
( y
y i )2
i =1
l=
( y
y i )2
i =1
EXAMPLE 1
Prove that Pythagoras theorem holds in S n .
Solution
Consider a triangle ABC right angled at A i.e., BAC = 90 o .
C (y i3)
A ( y1i )
B ( y2i )
Fig. 9.2
...(2)
183
or
or
(y
i =1
i
2
y1i ) ( y 3i y1i ) = 0
...(1)
(AB) =
(y
i
2
y1i ) 2
...(2)
i
3
y1i )2
...(3)
i
3
y i2 )2
...(4)
i =1
n
(AC) =
(y
i =1
n
(BC) =
(y
i =1
(BC) =
[( y
i
3
i =1
n
[( y
i
3
i =1
n
( y i3 y1i )2 +
i =1
n
(y
i
1
y i2 ) 2 + 2 0, [from (1)]
i =1
( y i3 y1i )2 +
i =1
( y
i
1
y2i )2
i =1
...(1)
184
or
d
=
d
Now,
f ( )
d
f ()
, Then, d =
1
f ( )
=
i =
x i
x
()2 =
f ( )
1
f ( )
1
1
( ) 2 =
f (); from (1)
f ( )
f ( )
()2 = 1
This proves that the hypersurfaces = constant form a system of parallels and therefore the
hypersurfaces = constant.
EXAMPLE 3
Show that it is always possible to choose a geodesic coordinates system for any Vn with an
arbitrary pole P 0.
Solution
Let P0 be an arbitrary pole in a Vn . Let us consider general coordinate system x i . suppose the
value of x i and P0 are denoted by x0i . Now consider a new coordinate system x j defined by the
equation.
x
1
2
h i
l
m
m
( x x0 ) ( x x0 )
l
m
j
m
m
j
= a m ( x x0 ) + a h
...(1)
The coefficients amj being constants and as such that their determinant do not vanish.
Now we shall prove that this new system of coordinated x j defined by equation (1) is a geodesic
coordinate system with pole at P0 i.e., second covariant derivative of x j vanishes at P..
Differentiating equation (1) with respect to x m , we get
x j
x
x j
x m
1
2
= a j at P
m
0
h
i
l
2 ( x x0 )
l
m
j
j
= a m + ah
...(2)
...(3)
185
x j
m = a j 0
x
m
0
and therefore the transformation given by equation (1) is permissible in the neighbourhood of P0 .
Differentiating equation (2) with respect to x j , we get
2 x j
j m = a j h
x x
h
0
l m 0
...(4)
(x )
j
,lm 0
2x
= l
x x 0 l m 0
j
x j
x h
h h j
a h , (from (3) and (4))
l m 0 lm
j
= ah
(x , )
j
lm 0
=0
= , ij... l
d xi d x j
dxl
ds ds
ds
...(1)
Solution
Since the coordinates x i lie on a geodesic. Then
d 2 xi
ds 2
i dx j dx k
+
=0
j k ds ds
...(2)
or
d
dx i
= , i
ds
ds
, i dx j
d 2xi
+
,i
x j ds
ds 2
,i dx i dx j
= x j ds ds ,i
i d x j d x k
, from (2)
j k ds ds
...(3)
186
,i dx i dx j
= x i ds ds , m
=
m dx j dx k
j k ds ds
, i dx i dx j
m dxi dx j
, i
m dx i dx j
,m
= j
i j ds ds
x
Equations (3) and (4) imply that the equation (1) holds for p = 1 and p = 2.
Suppose that the equation (1) holds for p indices r1 , r2 ,...,r p so that
d p
dx r1 dx r1
dx p
ds p
ds ds
ds
Differentiating the equation (5) with respect to s, we get
d p +1
ds p +1
= , r1 , r2 , ...,
rp
...(5)
r
r
r
, r1 r2 ... r p dx r1
dx p dx p +1
d 2 x r1 dx r2
dx p
+
r
r1 r2 ... rp
ds
ds
ds
ds
x p +1
ds 2 ds
+ , r1 r2 ... rp
substituting value of
...(4)
dx r1
d 2x p
ds
ds 2
...(6)
d 2 x r1
etc. from (2) in (6) and adjusting dummy indices, we have
ds 2
, r r ... r
m
m
1 2
p
d p +1
,
m
r
...
r
,
r
r
...
m
r
2
p
1 2
p +1 = x p +1
r
r
r
r
ds
1 p +1
p p +1
dx r1 dx r2
dx p +1
ds ds
ds
r
d x r1 d x r2
d x p +1
ds ds
ds
This shows that the equation (1) holds for next values of p. But equation (1) holds for p = 1, 2,
... Hence equation (1) holds for all values of p.
= ,r1 r2 .. r p rp +1
EXERCISES
1. Prove that at the pole of a geodesic coordinate system, the components of first covariant derivatives
are ordinary derivatives.
2. If x i are geodesic coordinates in the neighbourhood of a point if they are subjected to the
transformation
i
xi = x +
1 i
c x j xk xl
6 jkl
187
where Cs are constants then show that xi are geodesic coordinates in the neighbourhood of O.
3. Show that the principal normal vector vanishes identically when the given curve is geodesic.
4. Show that the coordinate system x i defined by
1 i j k
i
x i = x + 2 j k x x
d2x 1 d
dx
1 d
dt
d2y
d2z
d 2t d (log f ) dx dt
Ans : 2
(log f ) +
(log
f
)
=
0
;
=
0
;
=
0
;
=
0
2 dx
ds
dx
ds ds
ds
ds 2 f 2 dx
ds2
ds2
ds2
7. Find the differential equations for the geodesics in a cylindrical and spherical coordinates.
8. Find the rate of divergence of a given curve C from the geodesic which touches it at a given point.
CHAPTER 10
PARALLELISM OF VECTORS
m i dx
j + u m j
=0
ds
x
j
ui d x j
m i dx
+
u
=0
x j ds
m j ds
du i
i
+ um
ds
m
dx j
j ds = 0
(2)
Parallelism of Vectors
189
This concept of parallelism is due to Levi-Civita. The vector ui is satisfying the equation (1) is
said to a parallel displacement along the curve
Now, multiplying equation (1) by gil, we get
dx j
g il u,i j
ds = 0
or
( gil u,i j )
dx j
=0
ds
( g il u i ) , j
dx j
=0
ds
ul , j
dx j
ds = 0
u i, j
dx j
=0
ds
ui
m dx j
u
j
m
=0
i j ds
x
dui
m dx j
um
=0
ds
i j ds
The equation (2) and (3) can be also written as
m i
j
dui = u m j d x
(3)
(4)
m
j
dui = um i j d x
(5)
The equation (4) and (5) give the increment in the components ui and ui respectively due to
displacement dxj along C.
and
THEOREM 10.1 If two vectors of constant magnitudes undergo parallel displacements along a given
curve then they are inclined at a constant angle.
Proof: Let the vectors ui and v i be of constant magnitudes and undergo parallel displacement along a
curve C, we have (from equation (1), Pg. 188.)
dx j
= 0
ds
j
d
x
vi, j
= 0
ds
u,i j
at each point of C.
Multiplying (1) by gil, we get
dx j
( g il u,i j )
=0
ds
(1)
190
or
ul , j
dx j
=0
ds
u i, j
dx j
=0
ds
(2)
vi , j
dx j
=0
ds
(3)
Similarly,
uv sin
dx j
ds
j
d
dx j
i dx
vi + u i vi , j
= u, j
ds
ds
ds
ds
d
Either sin = 0 or
=0
ds
Either = 0
or
= constant.
=0
d d x m m d x j d x k
+
ds ds j k ds ds = 0
x j
dx m
ds
d x j m d x j d xk
ds + j k ds ds = 0
(4)
Parallelism of Vectors
j
x
dx m
ds
191
m dx k dx j
+
=0
j k ds ds
d xm dx j
ds ds = 0
, j
or
t,mj
dx j
=0
ds
dx m
suffer a parallel displacement along a geodesic curve.This
ds
confirms that geodesic is an auto-parallel curve.
Proved.
This shows that the unit tangent vector
dx j
=0
ds
(2)
The equation (1) shows that v i is of variable constant and parallel with respect to Riemannian V n
so that
v ,ij
dx j
dxj
i
= ( u ), j
ds
ds
i
i
= (, j u + u, j )
dx j
ds
= , j
dx j i
dxj
u + u,i j
ds
ds
= , j
dx j i
u Since
ds
d x j i
u
=
x j ds
v,i j
d i
dx j
u
=
ds
ds
d vi
= ds
from (1)
u,i j
dx j
=0
ds
192
i
= v
d (log )
ds
dx j
d (log )
= v i f (s) where f (s) =
(3)
ds
ds
Hence a vector v i of variable magnitude will be parallel with respect to V n if equation (3) is satisfied.
Conversely suppose that a vector v i of variable magnitude such that
dx j
= v i f (s )
v,ij =
ds
to show that v i is parallel, with respect to V n .
Take
v,i j
ui = v i (s)
(5)
Then
dx j
dx j
i
= (v ), j
ds
ds
j
dx j i
i dx
+ , j
v
= v, j
ds
ds
dx j i
i
v
f
(
s
)
+
v
=
x j ds
d
dx j
i
u,i j
= v f (s ) +
(6)
ds
ds
d
= 0.
Select such that f ( s ) +
ds
Then equation (6) becomes
dx j
u,i j
=0
ds
This equation shows that the vector u i is of constant magnitude and suffers a parallel displacement
along curve C. The equation (5) shows that v i is parallel along C.
Hence necessary and sufficient condition that a vector v i of variable magnitude suffers a parallel
displacement along a curve C is that
u ,ij
v,i j
dx j
= v i f (s ).
ds
EXAMPLE 1
Show that the vector v i of variable magnitude suffers a parallel displacement along a curve C if
and only if
dx k
(v i v i,k v i v l,k )
= 0,
i = 1, 2, ..., n.
ds
Solution
From equation (4), we have
dx j
v ,ij
= v i f (s )
(1)
ds
Parallelism of Vectors
193
Multiplying by v l, we get
dx j
= v l v i f (s )
ds
Interchange the indices l and i, we get
v l v,i j
dx j
= v i v l f (s)
ds
Subtract (1) and (2), we get
v i v,l j
(v l u,ik v i v l,k )
(2)
dx k
= 0 by interchanging dummy indices j and k.
ds
y i j
dx dx
x j
y
x j
(1)
g ij dx i x j
g ij dx i dx j g ij x i x j
a dy y
a dy dy a y y
(1)
194
y i y
dx
dx j
x i
x j
y i y j
y i y j
d
x
d
x
a
x
x .
x i
x j
xi
x j
a
=
a
y y i j
dx dx
x i x j
y y i j
y y i j
dx
dx
a
x x .
xi x j
x i x j
a
=
a
cos =
Since g ij = a
gij dx i dx j
g ij dx i dx j g ij x i x j
(2)
y y
(from equation (1), art. 10.3)
x i x j
Proved.
THEOREM 10.4 If U and ui denote the components of the same vector in Riemannian Vm and Vn
respectively then to show that
i y
= u
U
x i
Proof: Let the given vector be unit vector at any point P of V n. Let the component of the same vector
x's and y's be a i and A respectively. Let C be curve passing through s in the direction of the given
vector then
y y dx i y i
= i
= i a
=
(1)
A
s
x ds
x
But the components of a vector of magnitude a are a times the corresponding components of the
Unit vector in the same direction. Then
i
i
(2)
U = aA , u = aa
Multiplying (1) by a, we get
y
(aa i )
=
aA
x i
y i
=
u , using (2)
U
x i
Or
Proved.
THEOREM 10.5 To show that there are m-n linearly vector fields normal to a surface V n immersed in
a Riemannian V m.
Proof: Since V n is immersed in V m, the coordinates y of points in V m are expressible as functions of
coordinates xi in V n.
Parallelism of Vectors
195
Now,
dy
y dx i
=
ds
x i ds
i
for the curve x = s, we have
...(1)
dy
y
=
ds
x i
...(2)
dy
y
is a vector tangential to the curve in V m. Then from equation (2) it follows that
ds
xi
i
y
N =0
xi
y
(a N ) i = 0
x
y
N i = 0,
x
a
a N y ,i = 0
...(3)
i = 1, 2, , n & = 1, 2, , m
y
(m > n). The coefficient matrix i
x
is
y y y
2 y x
+
a
x i x j x k
xi x j x k
where [, ] and [i j, k ] are the Christoffels symbols of first kind relative to metrics a dy a dy and
g ij dx i dx j .
Solution
Since relation between a and g ij is given by
y y
x i x j
Differentiating it with respect to xk, we get
g ij = a
g ij
x k
a y y y
2 y y
y 2 y
+
a
+
a
=
x x i x j x k
x i x k x j
x i x j x k
(1)
196
Similarly
g jk
x i
a y y y
2 y y
y 2 y
+
a
+
a
=
y x i x j x k
x j x i x k
x j x k x i
(2)
and
a y y y
g ki
2 y y
y 2 y
+
a
+
a
=
x j
x x i x j x k
x k x j x i
x k x i x j
But we know that
[ij, k ] =
and,
1 g jk g ki g ij
2 x i
x j x k
1 g g g
[, ] = 2 +
y
y
y
(3)
(4)
(5)
Substituting the value of (1), (2), (3) in equation (4) and using (5) we get,
[ij, k ] = [, ]
y y y
2 y y
+
a
.
x i x j x k
xi x j x k
dT
ds
y
x i
2
dt i y
dx j
i y
+
t
=
ds x i
x i x j ds
i
= t
(1)
(2)
Now,
dx j
ds
dy
= T,
ds
T
dy
+
T
=
ds
x
T dy
dy
+
T
=
x ds
ds
i
p i = t, j
dT
dy
+
T
= ds
ds
(3)
(4)
Parallelism of Vectors
197
dT
and T from (2) and (1) in equation (4), we get
ds
2
dt i y
dx j
i y
i y dy
+
t
+
t
q =
ds x i
x i x j ds
dx i ds
t i y dx j
dx j
i y
i y dy
+
t
+
t
x j x i ds
x i x j ds
x i ds
j
t i y dx j
dx j
i y
i y y dx
+
t
+
t
x j x i ds
x i x j ds
x i x. j ds
t i y d x j i d x j 2 y
y y
+
t
+
+
ds x i x j x j x i
x j x i ds
(5)
y y y
2 y y
+
a
x i x j x k
x i x j x k
[ ji, k ] = [, ]
y y y
2 y y
+
a
x j x i x k
x i x j x k
y y y
2 y y
a
+
a
= j
x x i x k
x i x j x k
y
a
= x k
Multiplying (5) by a
(6)
y
, we get, using (6),
x k
Or
2
y y
j i + iy j
x x
x x
j
y
t i dx j y
y
i dx
q
a
+
t
[ij, k ]
=
ds
x k
x j ds x i
x k
dx j t i
p
y
g + t i g pk ,
j ik
k =
ds x
x
i j
dx j
= ds
t i
i
j gik + t a g ik
x
a j
= g ik
dx j i
t,
ds j
since g ij = a
y y
x i x j
198
i
= ( gik t ), j
= tk, j
dx j
ds
dx j
ds
y
dx j
t
=
k, j
ds
x k
y
q k = pk
x
q
or
(7)
Proved.
Properties of Vm
(i) If a curve C lies in a subspace V n of V m and a vector field in V n is parallel along the curve C
with regard to V m, then to show that it is also a parallel with regard to V n.
Proof: If a vector field t is a parallel along C with respect to V m. then its derived vector q
vanishes i.e.,
q = 0, = 1, 2,... m.
Hence equation (7) becomes pk = 0, k = 1, 2, ..., n. This shows that the vector field t is also
parallel along C with respect to V n.
(ii) To show that if a curve C is a geodesic in a V n it is a geodesic in any subspace V n of V m.
y
Proff : Science, q
= pi
x i
Let t be unit tangent vector to the curve C them T is unit tangent vector to C relative to V m
and ti is unit tangent vector to C relative to V n.
Now,
pi = 0
curve C is a geodesic in V n
q =0
curve C is a geodesic in V m
Also,
q = 0, pi = 0, i
This shows that if a curve C in V n is a geodesic relative to V m then the same curve is also a geodesic
relative to V n.
(iii) A necessary and sufficient condition that a vector of constant magnitude be parallel with
respect to V n along C in that subspace, is that its derived vector relative to V m for the
direction of the curve be normal to V n.
Proof: Let t be a vector of constant magnitude. This vector t is parallel along C relative to Vn iff pi =
0, i
iff q
But q
For
y
=0
x i
y
= 0 implies that q is normal to V n.
x i
y
= y,i lying in V m is tangential to a coordinate curve of parameter xi in V n.
xi
Parallelism of Vectors
199
These statements prove that a necessary and sufficient condition that a vector of constant
magnitude be parallel along C relative to V n is that its derived vector i.e., q along C relative
to V m be normal to V n.
(iv) A necessary and sufficient condition that a curve be a geodesic in V n is that its principal
subnormal relative to V m the enveloping space be normal to V n at all points of the curve.
Proof: In particular let the vector t be unit tangent vector to the curve C. In this case q is
called principal normal the curve C. Also pi = 0, i implies that the curve C is a geodesic
relative to V n.
Using result (iii), we get at once the result (iv).
(v) To prove that the tendency of a vector is the same whether it is calculated with respect to
V m.
Proof: Since, we have
y
q i = pi
x
q
y dx i
dxi
p
=
i
ds
x i ds
dy
dxi
= pi
ds
ds
dy dy
dx j dx i
T,
= ti, j
ds ds
ds ds
i.e., tendency of T along C = tendency of ti along C.
i.e., tendency of t along C relative to V m = tendency of t along C relative to V n.
q
( gij p i q j ),k
dx k
=0
ds
200
dx k j
dx k i j
i j dx
g ij p,ik
q
+
g
p
q
+
g
ij
,k ds ij ,k ds P q = 0
ds
(3)
gij , k dx p i q j = 0
ds
gij, k = 0
dx k
0)
(Since and are unit vectors and
ds
g ij
m
m
g mj gim = 0
k
x
i k
j k
g ij
[ik , j ] [ jk , i ] = 0
x k
g ij
= [ik , j] + [ jk , i ]
x k
Now, using equation (4), we have
g jk g ki g ij
+
= 2[ij, k ]
x i
x j x k
pi
qi
[ij,k] =
k
lk
But we know that i j = g [ij, l ]
from (5), we have
1 g jk g ki g ij
2 x i
x j x k
k
1 lk g jl g li gij
= g i + j l
2
x
x
i j
x
(4)
(5)
Proved.
EXAMPLE 3
If t i and T are contravariant components in x's and y's respectively, of n vector field in V n
immersed in V m. Show that
T, j = y,ij t i + y ,i t,i j
Solution
Since we know that
T
i
= t
y
x i
Parallelism of Vectors
201
i
T = t y, j .
Taking covariant differentiation of both sides, we get
T, j = (t i y ,i ), j
= t,i j y,i + t i ( y ,i ), j
T, j = t i y ,ij + t,i j y ,i
EXAMPLE 4
Show that g ij
dx i dx j
remains constant along a geodesic.
ds ds
Solution
i
Let t =
dx i
. Then
ds
dx i dx j
= g ij t i t j = t 2
ds ds
Since we know that geodesics are autoparallel curves. Then
gij
t,i j
dx j
=0
ds
t,i j t j = 0
or
(1)
Now,
dt 2
d
d
( g ij t i t j ) = (ti t i )
=
ds
ds
ds
dx j
i
= (t i t i ), j t j
= ( ti t ) , j
ds
dt 2
= (ti , j t i )t i + (t ,i j t i )ti = 0,
ds
from (1)
Integrating it we get
t2 = constant.
So, g ij
dx i dx j
remains constant along a geodesic.
ds ds
EXAMPLE 5
If ti are the contravariant components of the unit tangent vector to a congruence of geodesics.
Show that
t i (ti , j + t j ,i ) = 0
and also show that | ti , j + t j ,i |= 0.
202
Solution
Let t i denote unit tangent vector to a congruence of geodesic so that
t,i j t j = 0
Since geodesics are auto-parallel curves. Then to prove that
(i)
(1)
t i (ti , j + t j ,i ) = 0
(ii) | ti , j + t j ,i |= 0
Since
t i ti = t 2 = 1
t i ti = 1
Taking covariant derivative of both sides, we get
(t i t i ), j = 0
or
t,i j t i + t i ti , j = 0
ti , j t i = 0
(2)
from (1)
g ik t,i j t j = 0
or
tk, j t j = 0
t k ,i t i = 0
Thus
t i t j, i = 0
t j ,i t i = 0
t i t j, i = 0
(3)
| ti , j + t j ,i | = 0
Solved.
EXAMPLE 6
When the coordinates of a V 2 are chosen so that the fundamental forms is
(dx&' ) + 2 g12dx1 dx 2 + (dx 2 )2 , prove that the tangent to either family of coordinate curves suffers a
parallel displacement along a curve of the other family.
Parallelism of Vectors
203
Solution
The metric is given by
i
j
ds 2 = g ij dx dx
i,j = 1,2.
Comparing it with
1 2
1
2
2 2
ds 2 = (dx ) + 2 g12 dx dx + (dx )
(1)
We have,
g11 = 1, g22 = 1.
In this case, we have coordinate curves of parameters x1 and x2 respectively. The coordinate x1
curve is defined by
xi = ci, i, except i = 1.
(2)
2
and the coordinate x curve is defined by
x i = d i, i, except i = 2
(3)
i
i
where c and d are constants.
Let p i and q i be the components of tangents vectors to the curves (2) and (3) respectively. Then
we have
pi = dxi = 0, i, except i = 1
and
qi = dxi = 0, i, except i = 2
So,
pi = (dx i, 0) and qi = (0,dx i)
Let t be the unit tangent vector to the curve (2).
Hence
pi
dx i
i
t
=
= (1,0)
=
p
ds
where
p = dxi
So, we have
dx j
=0
ds
Hence the tangent to the family of coordinates curves (2) suffers a parallel displacement along a
curve of the family of curves (3).
p,i j
EXERCISES
1. Explain Levi-Civita's concept of parallelism of vectors and prove that any vector which undergoes a
parallel displacement along a geodesic is inclined at a constant angle to the curve.
2. Show that the geodesics is a Riemannian space are given by
d 2 x m m dx i dx k
+
=0
ds 2
i k ds ds
Hence prove that geodesics are auto-parallel curves.
204
CHAPTER 11
(1)
(2)
The indices l, h and k are not tensor indices. But these indices in lhk are arranged in proper way,
the first index l indicates the congruence whose unit tangent is considered, the second h indicates the
direction of projection and the third k is used for differentiation.
THEOREM 11.1 To prove thast the Riccie's coefficients of rotation are skew-symmetric in the first
two indices i.e.,
lhk = lhk
Proof: If e ih| (h = 1, 2, , n) be n unit tangents to n congruences e h| of an orthgonal ennuple in a Vn
then
eh|i eli| = 0
convariant differentiation with respect to xj, we get
eh|i eli| ,j = 0
eh |i , j eli| + eli|, j eh|i = 0
206
or
(1)
or
or
lhk eh|i
h
= el|i , j ekj|
j
= el |i , j e k|
(e e )
i
h|
h| m
= el |i , j ekj|im since
ehi | eh|m =
h
i
m
= (e l |i , j m ) e kj |
i
lhk eh |m
= el |m , j ekj |
Replacing m by i, we get
lhkei|m
h
= el |i , j e kj|
u,ij ehj| = 0
(1)
(2)
207
d
i
j
ds m = (u e h|i ), j em |
= (u i e h|i , j + u i, j eh |i ) emj |
sin
at the point P 0, =
d
u i e h |i , j e mj | + u ,i j e mj |
ds m =
(3)
, we have
2
d
u i eh|i , j emj |
ds m =
= u i eh|i , j emj |
= eli|eh|i, j emj | ;
from (1)
d
ds m = hlm
d
ds m = hlm
(4)
d
In Eucliden space of three dimensions, ds is the arc-rate of rotation of the vector e ih| about the
m
curve Cm. Hence the quantities hlmare called coefficients of rotation of the ennuple. Since it was
discovered by Ricci and hence it is called Ricci's coefficient of rotation.
11.3 CURVATURE OF CONGRUENCE
The first curvature vector pl| of curve of the congruence is the derived vector of e ih| in its own
direction. Where e ih| (h = 1, 2, , n) be unit tangents to n congruence eh| of an orthogonal ennuple in
a Vn .
If phi | be the contravariant component of first curvature vector pl|. Then by definition, we have
phi | = ehi , j ehj|
from theorem (11.2), we have
phi | =
hlh eil|
l
(1)
208
The magnitude of p ih| is called curvature of the curve of congruence e h| and denoted by K h | .
Now,
K h2| = g ij phi | p hj|
= g ij
i
hlh e l |
i
hmh em |
hmh hmh
l
m hlm
hmh , Since g ei e i = l
m
ij l| m|
K h2| =
hmh )
hlh eil|
l
Since
eli|
0 , h and hence
phi |
congruence.
Hence C is a geodesic congruence iff hlh = 0, h .
But
hlh = lhh
So, C is a geodesic congruence iff lhh = 0, h .
or C is a geodesic congruence iff lhh = 0.
Hence lhh = 0 are the necessary and sufficient conditions that congruence with unit tangents ehi |
be geodesic congruence. Again lhh is the tendency of the vector eli| in the direction vector e ih| .
209
Thus a congruence C of an orthogonal ennuple is a geodesic congruence iff the tendency of all
other congruences in the direction of C vanish identically.
11.5 NORMAL CONGRUENCE
A normal congruence is one which intersects orthogonally a family of hypersurfaces.
THEOREM 11.4 Necessary and sufficient conditions that the congruence eh | of an orthogonal ennuple
be normal is that
nqp = npq
Proof: Consider a congruence C of curves in a V n. Let ( x1 , x 2 , , x n ) = constant be a family of
hypersurfaces. To determines a normal congruence whose tangent vector is grad or .
Let ti be the convariant components of unit tangent vector to C. The congruence C is a normal
congruence to a family of hypersurface ( x1 , x 2 , , x n ) = constant if
,
,1
, 2
= n = y (say)
=
(1)
t1
t2
tn
In order that (n 1) differential equations given by equation (1) admit a solution which is not
constant, these must constitute a complete system.
From (1)
,i
= y or , i = yti
ti
yti =
x i
ti + y
t i
x j
x i x j
Interchanging indices i and j, we get
t j
y
2
t
+
y
=
j
x i
x i
x j x i
Subtracting (2) and (3) we get
t y
t
y
j ti + y i j i t j + y i j = 0
x x
dx
dx
t j y
t
y
y ij i + t i j t j i = 0
x x
x
x
(2)
(3)
210
Multiplying by tk,
ti t j
y
y
ik ti i = 0
ytk j i + ti tk
j
x
x
dx
dx
By cyclic permutation of i, j, k in (4), we get
t j t
yti k kj
x
x
(4)
y
y
+ t j t i k i k ti
=0
x
x j
(5)
and
t
t
y
y
yti ki ik + t k t j i ti t j k = 0
x x
x
x
(6)
+ t j tk ti
x i x k
t k (ti , j t j ,i ) + ti (t j , k t k , j ) + t j (t k , i ti ,k ) = 0
or
= 0
(7)
as y 0 , where i, j, k = 1, 2, , n.
These are the necessary and sufficient conditions that the given congruence be a normal congruence.
Now suppose that the congruence is one of an orthogonal ennuple in V n. Let en|i be the unit
tangents of given congruence C so that ti = en|i
Then equation (7) becomes
en|k (en| i,j en| j,i) + en| i (en| j,k en| k,j) + en| j (en|k,i en|i,k) = 0
(8)
Now, multiplying equation (8) by e ip| eqR| , we get
e n|k (en |i , j en | j ,i ) e ip| eqk| + en |i (en | j , k en| k , j ) e ip| eqk + en | j (en |k ,i en |i , k ) e ip |eqk|
where p and q are two new indices chosen from 1, 2, , n 1. i.e., p,q and n are unequal.
e n |k eqk| = q = 0, q n
h
But
en |i e kp| = hp = 0, p n
so, we have
e n| j (en |k ,i en | j , k ) eip | eqk| = 0
e n| j (en |k , i e iq |e kp | en | j , k eip |e kq| ) = 0
en|j (nqp npq) = 0
Since
211
npq = nqp,
(9)
Conversely if equation (9) in true then we get equation (8). Which implies that equation (7) are
satisfied bty en|i Hence en| is a normal congruence.
Thus necessary and sufficient conditions that the congruence en| of an orthogonal ennuple be a
normal congruence are that
nqp = npq (p, q = 1, 2, , n 1 such that p q)
THEOREM 11.5 Necessary and sufficient conditions that all the congruences of an orthogonal ennuple
be normal.
Proof: If all the congruences of a orthogonal ennuple are normal. Then
nqp = npq (p, q = 1, 2, , n 1 such that p q)
If the indices h, k, l and unequal then
hkl = hlk
(1)
But due to skew-symmetric property i.e., hlk = lhk.
So,
hkl = hlk = lhk = lkh, from (1)
= klh (skew-symmetric property).
= khl, from (1)
hkl = hkl , (skew-symmetric property).
hkl + hkl = 0
2 hkl = 0
hkl = 0
212
nhk eh |i
h =1
m
= e n |i , m e k |
h,k =1
= en |i ,m mj
h, k =1
nhk eh|i e k | j
= e n|i, j
(2)
h,k =1
n ,k =1
curl en| i =
from (2)
nhk eh |i ek | j
h ,k =1
h , k =1
nkh ek | j e h|i
curl en|i =
h ,k =1
(3)
curl en| i =
h ,k =1
n 1
( nhk nkh ) eh |i ek | j +
h =1
k =1
curl en|i
h , k =1
n1
curl en|i
( nhk nkh ) e h |i e k | j +
h ,k
n 1
nhn e h|i en | j
h =1
n1
nkn en |i
ek| j
k =1
n1
h =1
nhn ( eh |i en | j
en |i ek | j )
(4)
213
if
i.e., if hnn = 0
i.e., if the congruence e n| is a geodesic congruence. Further, if first and second term on right
hand side of equation (4) both vanishes then
curl en|i = 0
Hence we have proved that if the congruence e n| satisfies any two of the following conditions
then it will also satisfy the third.
(a) e n| is a normal congruence
(b)
e n| is irrotational
(c)
e n| is a geodesic congruence
(1)
(2)
(3)
214
for j = 1, 2, , n.
1
( X 11 g11 ) e + ( X 21 g 21 ) e + + ( X n1 g n 1 ) e + en |1 = 0
( X 12 g12 ) e1 + ( X 22 g 22 ) e 2 + + ( X n 2 g n 2 )e n + en |2 = 0
M
( X 1n g1 n ) e1 + ( X 2 n g 2 n ) e 2 + + ( X nn g nn ) e n + en |n = 0
from (3)
e n|1 e1 + en |2 e 2 + en |n e n + .0 = 0
eliminating and the quantities e1, e2, en, we have the equation
(X 11 g11 ) (X 21 g 21 )
( X12 g12 ) ( X 22 g 22 )
(X n1 g n1 )
( X n2 g n2 )
(X 2n g 2n )
(X nn g nn )
M
( X1n g1n )
en |1
en |2
en| n
en |1
e n|2
en |n
0
(4)
(5)
en |i e ik| = 0
(6)
( X ij g ij ) eik | + h en | j = 0
(7)
Multiplying (5) by e kj| and (7) ehj| and using (4) and (6), we get
( X ij h g ij ) ehi |ekj| = 0
(8)
( X ij k g ij ) eki |e hj| = 0
(9)
Since X ij and g ij are symmetric tensor in i and j. Now, interchanging i and j in equation (9), we
get
( X ij k g ij ) ehi |e kj| = 0
Subtracting (10) and (8), we get
(10)
(h k ) g ij e ih| ekj | = 0
(11)
Since k k 0 as h h.
g ij eih| ekj| = 0
(12)
215
This shows that e ih| and e kj| unit vectors are orthogonal to each other and hence the congruence
e h| and e k| (h k) are orthogonal to each other. Hence the n 1 congruence eh| (h = 1, 2, , n) thus
determined form an orthogonal ennuple with en|.
Using equation (12), equation (10) becomes
X ij e ih| ekj| = 0
Since from (1), X ij =
1
e
+ e n| j , i . Then we have,
2 n |i , j
1
e
+ en | j ,i ehi |ekj| = 0
2 n |i , j
en |i , j ehi | ekj| + en| j ,i eik ehj = 0
nhk + nkh = 0; (h, k = 1, 2, , n 1 such that h k )
(13)
Conversely if equation (13) is true then (n 1) congruences eh| of the orthogonal ennuple and
canonical with respect to e n| . Hence necessary and sufficient condition that the n 1 congruences e h|
of an orthogonal ennuple be canonical with respect to en| are
nhk + nkh = 0; (h, k = 1, 2, , n such that h k )
THEOREM 11.8 Necessary and sufficient conditions that n 1 mutually orthogonal congruences e h|
orthogonal to a normal congruence e n| , be canconical with respect to the later are nhk = 0 where k, h
= 1, 2,, n 1 such that h k.
Proof: By theorem (11.7), Necessary and sufficient conditions that (n 1) congruences e h| of an
orthogonal ennuple be canonical with respect to en| are
nhk + nkh = 0 (h, k = 1, 2, , n 1 such that h k).
If the congruence en| is normal. Then
nhk = nkh
The given condition nhk + nkh = 0 becomes
nhk + nhk = 0
2nhk = 0
nhk = 0
Proved.
EXAMPLE 1
1
If en| are the congruences canonical with respect to en| prove that (i) h = nhk (ii) h = hnn,
2
(iii) If en| is a geodesic congruence, the congruences canonical with respect to it are given by
( X ij g ij )ei = 0 = 0
Solution
Suppose (n 1) congruences eh| of an orthogonal ennable in a V n are canonical with respect to
the cougruence eh| then
en |iehi | = 0
(1)
216
and
where
( X ij h g ij ) ehi | + h en | j = 0
X ij =
(2)
1
(e
+ en| j,i)
2 n| i,j
(3)
(4)
(from (1))
from (3)
1
(e e i e j + en | j ,i eih| ehj | ) h = 0
2 n|i , j h | h|
1 ( + ) = 0
nhh
nhh
h
2
h = nhh
(ii) Multiplying (2) by enj| , we get
( X ij h g ij ) ehi | enj| + h e n| j e nj| = 0
X ij eih| enj| h g ij ehi |enj| + h 1 = 0
1
(e n|i , j + en| j , i ) e ih|e hj| + h = 0,
2
from (1)
since e ih| e hj| = 0
1
(e ei e j + en | j ,i ehi | ehj | ) + h = 0
2 n|i , j h | n|
1
( + nnh) + h = 0
2 nhn
h =
or
h =
1
, since nnh = 0
2 nhn
2 hnn
(iii) If en| is a geodesic congruence, then hnn = 0 from the result (ii), we have
1
h = 0
2
h = 0
217
( X ij h g ij ) e ih| = 0
this gives ( X ij g ij ) e i = 0 .
EXAMPLE 2
Prove that when a manifold admits an orthogonal systyem of n normal congruences then any of
these in canonical with respect to each other congruence of the system.
solution
Let e ih| (h = 1, 2, , n) be unit tangents to n normal congruences of an orthogonal ennuple in a
V n. So that
lhk = 0, where l, h, k = 1, 2,...n such that l, h, k being unequal.
It is required to show that a congruence eh| is canonical with respect to the congruence ek| (h, k = 1, 2,
, h k). We know that the n 1 congruence en| of an orthogonal ennuple be canonical to en| iff
nhk + nkh = 0
This condition is satisfied by virtue of equation (1). Hence (n 1) congruences eh| of an orthogonal
ennuple are canonical to en|.
Similarly we can show that any n 1 congruences are canonical to the remaining congruence.
It follows from the above results that any one congruence is canonical with respect to each other
congruence of the system.
EXERCISE
1. If is a scalar invariant
n
,ij eih|ehi | =
2
h =1
=
sl
sh sk sk sh
l
where S h denotes the arc length of a curve through a point P of an ennuple and Q is a scalar invariant.
5. If the congruence eh| (h = 1, 2, n 1) of an orthogonal ennuple are normal, prove that they are
canonical with respect to other congruence eh| .
CHAPTER 12
HYPERSURFACES
12.1 INTRODUCTION
We have already studied (Art 10.3 chapter 10) that if m > n then we call V n to be a subspace of V m and
consequently V m is called enveloping space of V n. We also know that there are m n linearly independent
normals N to V n. (Art 10.3 Theorem 10.5, chapter 10). If we take m = n + 1 then V n is said to be
hypersurface of the enveloping space V n + 1
Let V n be Riemannian space of n dimensions referred to cordinates xi and having the metric ds2 =
i
gij dx dxj. Let V n be Riemannian space of m dimensions referred to coordinates y and having the metric
ds2 = a dy dy. Where m > n. Let Greek letters , , take the values 1, 2, , m and latin indices
i, j, k, take the values 1, 2, n. Then we have, the relation between a and gij
gij = a y y
xi x j
(1)
Since the function y are invariants for transformations of the coordinates xi in V n, their first
covariant derivatives with respect to the metric of V n are the same as their ordinary derivatives with
respect to the variables xi.
y, i =
i.e.,
y
xi
+ i
(2)
parameter xi in V n. Consequently if N are the contravariant components of the unit vector normal to
V n. Then we have
a N y,i = 0,
and
N N =
(i = 1, 2, , n)
(3)
(4)
Hypersurface
219
dy
=0
ds
u
dy
u
=0
ds
y
u dy
dy
=0
y ds
ds
du
dy
u
=0
ds
ds
dy
du
= u ds
ds
(1)
dy
dv
= v ds
ds
(2)
(3)
The Christoffel symbol with Greek indices being formed with respect to the a and the y's and
christoffel symbol with Latin indices with respect to the gij and the x's.
Let Ai be a tensor field, defined along C, which is mixed tenser of the second order in the y's and
a covariant vector in the x's. Then the product u v wi Ai is scalar invariant and it is a function of s
along c. Its derivative with respect to s is also a scalar invariant.
dA
d
dv
dw i
i
(u v wi Ai ) = u v w i i + du v w i A +
u
w
A
+
u v Ai
i
dt
ds
ds
ds
ds
= u v w
dAi
ds
du i dv
dw j
v w Ai +
u w i Ai +
u v Aj
ds
ds
ds
220
i
= u v w
dy
i dy
+ Ai .u v wi
Ai .u v w ds
ds
ds
dAi
i j dx
Aj u v w ik ds , by equation (1), (2) & (3).
k
i dAi
dy
dy
j dx
u
v
w
+
A
.
A
.
i
i
j
=
ds
ds
i k ds
ds
= Scalar, along C
Since the outer product uv wi is a tensor and hence from quotient law that the expression within
dxk
ds
Since
Ai
Ai
+ Ai y,k Ai y,k Aj
xk
ik
dxk
is arbitrary..
ds
So, by Quotient law the expression within the bracket is a tensor and is called tensor derivative of
with respect to xk. It is denoted by Ai;k . Then we have
Ai;k =
Ai
j
+ Ai y,k Ai y,k Aj .
dx
ik
k
+ C y, ck Ca y, ck
x
a c
c
k
Hypersurface
221
( A
B ); k
( A + B )
dx
+ ( Aa + A ) y,ck ( Aa + Ba ) y ,ck
a c
c
a c
a c
+
A
y
A
y
,
k
a
,
k
= x k
+
ac
c
B
a
+ Ba y,ck Ba y,ck
xk
ac
c
A B = D
Let
Then D
is a tensor
we have
D ; k =
( A
B ) ; k =
a
a
+ D;ak yc Da y,ck Da y,ck
ac
c
c
( A B )
dx
a
a
a
+ A
B y ,ck Aa B y,ck A B a y ,ck
ac
c
c
a
c
a c
+
A
y
A
y
B + A Ba y,ck
,
k
a
,
k
= dxk
k
ac
c
c
dx
( A B ); k = A; k B + Aa B ; k
Hence the result (ii).
Note: , , , a, c take values from 1 to m while k take values from 1 to n.
THEOREM 12.2 To show that a; j = 0
or
To prove that the metric tensor of the enveloping space is generalised covariant constant with respect
to the Christoffel symbol of the subspace.
Proof: We have (see pg. 220)
a;i =
a
dxi
a
dx i
a y,i a y,i
( [, ] + [, ] ) y,i
222
or
a;i =
or
a;i =
a
dx
a
dx
a
y,i
y
a y
a a
i =
y
dx
x i
x i
= 0
Proved.
Proof: Since y is an invariant for transformation of the x's and its tensor derivative is the same as its
covariant derivative with respect to the xs, so that
y; i = y, i =
dy
(1)
dxi
y;ij
y l
y,l + y,i y, j
x j xi
ij
2 y
l
y,l + y,i y,j
x x
ij
i
(2)
2y
l
y ,l + y , j y ,i
=
j
i
x x
ji
y;ji
2y
l
y ,l + y ,l y ,j
=
i
j
x x
ij
(3)
Hypersurface
223
y y
xi x j
or
gij;k = 0
and
a;k = 0.
we have,
0 = a y ;i y;jk + a y;ik y;j
or
(4)
and
a y; ji y,k + a y , j y;ki = 0
(5)
(6)
subtracting equation (4) from the sum of (5) and (6), we get.
2 a y;ij y ,k = 0
a y ;ij y ,k = 0
or
(7)
This shows that y;ij is normal (orthogonal) to y,k . Since y,k is tangential to V n and hence y;ij is
normal to V n. Then we can write
y;ij = N ij
where
(8)
is unit vector normal to V n and ij is a symmetric covariant tensor of rank two. Since
y;ij a N = ij a N N
= ij N 2
224
y;ij a N = ij ,
or
ij
N=1
y;ija N .
is called the second fundamental form for the hypersurface V n of V n + 1. The components of tenser ij
are said to be coefficient of second fundamental form.
Note: The quadratic differential form g ij dx idx j is called first fundamental form.
12.5 CURVATURE OF A CURVE IN A HYPERSURFACE AND NORMAL CURVATURE
If U and u i be the contravariant components of the vector u relative to V n and V n + 1 respectively then
we have (from chapter 10, Theorem 10.4)
y i
u = y,i ui
(1)
U
xi
Let the derived vector to vector u along C with respect to metric of V n and Vn + 1 are denoted by
p and q respectively. Then
and
pi
u,i j
dxj
ds
= U,
d y
ds
= U ;j
dx j
, (from equation 1, Art 12.4)
ds
(2)
Taking the tensor derivative of each side of equation (1) with respect to x's, we have
U ;j = y; ij u + y, i u , j
U ;j = ij N u + y ,i u , j
i
i
= (ij N u + y, i u, j )
or
i
i dx
u
= ij
N + y,i p
ds
where
i
pi = u, j
dx j
ds
(3)
Hypersurface
225
Now suppose that vector u is a the unit tangent t to the curve C. Then the derived vectors q and
p are the curvature vectors of a C relatively to V n + 1 and V n respectively. Then equation (3) becomes.
dx i dx j
N + y ,i p i
ds
q = ij ds
Kn = ij
Taking
dx i dx j
,
ds ds
(4)
we get
q = Kn N + y,i pi
...(5)
Kn is called Normal curvature of V n at any point P of the curve C and Kn N is called normal
curvature vector of V n + 1 in the direction of C.
Meaunier's Theorem
If Ka and Kn are the first curvature of C relative to V n + 1 and normal curvature of V n respectively
r
and w is the angle between N and C (C being the unit vector of V n + 1 then the relation between Ka, Kn
and w is given by
Kn = Ka cos
Proof: We know that
q = Kn N + p i y,i
...(1)
dxi dx j
ds ds
Let Ka and Kg be the first curvatures of C with respect V n + 1 and V n respectively then
where
Kn = ij
Ka =
r
a q q , Kg =
gij pi p j
r r
N C = N C cos
= cos
r r
N .C = cos
If b is a unit vector of V n +
equation (1) becomes
as |N| = |C| = 1
...(2)
r
r
r
r
Ka C = K g b + K n N
r
Taking scalar product of equation (3) with N , we have
r r
r r
r r
K a N .C = K g N. b + K n N. N
Ka cos = Kg 0 + K n 1;
Kn = Ka cos
(3)
from (2)
(4)
Proved.
226
EXAMPLE 1
Show that the normal curvature is the difference of squares of geodesic curvatures.
Solution
We know that (from Meuriers Theorem, equation 3)
r
r
r
KaC = Kg b + Kn N
or
Theorem 12.3 To show that the first curvature in Vn + 1 of a geodesic of the hypersurface Vn is the
normal curvature of the hypersurface in the direction of the geodesic.
Proof: From, example 1, we have
2
2
Ka2 = Kg + Kn
If C is a geodesic of V n then p i = 0 Kg = 0
Then we have
Ka2 = Kn2
Ka = K n.
Proved.
Dupin's Theorem
The sum of normal curvatures of a hypersurface V n for n mutually orthogonal directions is an
invariant and equal to ij g ij .
Proof: Let ehi | (h = 1, 2, , n) be unit tangents to n congruences of an orthogonal ennuple in a V n. Let
Knh be normal curvature of the hypersurface V n in the direction of the congruence eh | . Then
Knh = ij eih | ehj|
The sum of normal curvatures for n mutually orthogonal directions of an orthogonal ennuple is a
V n is
n
Knh =
h =1
ij
eih| ehj|
h =1
= ij
i
h|
ehj|
h =1
= ij g ij
= Scalar invariant
Proved.
Hypersurface
227
12.6 DEFINITIONS
(a) First curvature (or mean curvature) of the hypersurface Vn at point P.
It is defined as the sum of normal curvatures of a hypersurface V n form mutually orthogonal directions
at P and is denoted by M. Then
M = ij g ij
(b) Minimal Hypersurface
The hypersurface V n is said to be minimal if M = 0
i.e.,
ij g ij = 0
(c) Principle normal curvatures
The maximum and minimum values of Kn are said to be the principle normal curvatures of V n at P.
Since these maximum and minimum values of Kn correspond to the principal directions of the symmetric
tensor ij .
(d) Principal directions of the hypersurface at a point P.
The principal directions determined by the symmetric tensor ij at P are said to be principal directions
of the hypersurface at P.
(e) Line of curvature in V n
A line of curvature in a hypersurface V n is a curve such that its direction at any point is a principal
direction.
Hence we have n congruences of lines of curvature of a V n.
THEOREM 12.4 To show that the mean curvature of a hypersurface is equal to the negative of the
divergence of the unit normal.
or
To show that the first curvature of a hypersurface is equal to the negative of the divergence of the unit
normal.
or
To show that the normal curvature of a hypersurface for any direction is the negative of the tendency
of the unit normal in that direction.
Proof: Let N be the unit normal vector to the hypersurface V n in y's and let N be its covariant
components. Let t be the unit tangent vector to N congruences eh| (h = 1, 2, , n) of an orthogonal
ennuple in V n and let Th| be the contravariant components in V n + 1 of t. Since t is orthogonal to N,
therefore
Th| N = 0
(1)
(2)
228
(3)
Now Th|,Th| is the first curvature of the curve eh| and N , . Th| . Th| is the tendency of N is the
direction of eh| . Hence equation (3) implies that the normal component of the first curvature of the eh|
relative to V n + 1 or the normal curvature of V n in the direction of the curve eh|
= tendency of N in the direction of the curve eh|
(4)
Kn =
Kh cos2 h
h =1
where Kh are the principal curvature and h are the angles between direction of ar and the congruence
eh|
Proof: The principal directions in V n determined by the symmetric covariant tensor teser ij are
given by
(ij K h g ij ) p ih| = 0
(1)
(2)
Kn = g p i p i
ij
(3)
Hypersurface
229
(h k )
ij pih | pkj| = 0
(4)
Let eh| be the unit tangents to the n congruences of lines of curvature. Then the principal curvatures
are given by
Kh = ij ehi | ehi |
(h = 1, 2, , n)
eh| cos h
h
n
or
where
ai
(5)
h=1
h|
cos h
(6)
r r
cos h| = a e h | = ai eh|i
ai being the contravariant components of a and h being the inclination of vector a to eh|. The
r
normal curvature of V n for the direction of a is given by
Kn = ij a ia j
from (6), we get
Kn = ij
h ,k =1
n
i
h|
cos h
j
k|
cos h
h =1
2
Kn = Kh cos h
h =1
(7)
(1)
and two congruences of curves in the hypersurface are said to be conjugate if the directions of the two
curves through any point are conjugate.
230
A direction in V n which is self-conjugate is said to be asymptotic and the curves whose direction
are along asymptotic directions are called asymptotic lines.
Therefore the direction the vector ar at a point of V n be asymptotic if
ij a i a j = 0
(2)
(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
y;i = y,i =
y
xi
(1)
The unit Normal N be the contravariant vector in the ys whose tensor derivative with respect
to the xs is
N;i =
N
+ N y, i
x i
(2)
Hypersurface
231
aN N = 1
Since
(3)
xi
gives
aN N;i + a N;i N = 0
a N N ;i + a N ;i N = 0
a N N ;i + a N ;i N = 0
2 a N N ;i = 0
a N N ;i = 0
(4)
which shows that N;i is orthogonal to the normal and therefore tangential to the hypersurface.
Thus N;i can be expressed in terms to tangential vectors y,k to V n so that
N;i = Ai y,k
a N ;j y ,i + a N ij N = 0
or
a N ;j y ,i + ij (a N N ) = 0
since y,ij = ij N
a N N = 1
or
a N ;j y,i + ij = 0
or
g ki A kj + ij = 0
y y
= g ki
since a y, k y, i = a
k
i
x x
(5)
232
Aim + ij g im = 0
im
Aim = ij g
ik
N;i = ij g
jk
jk
y,k
y ,k
(6)
jk
y ,k
(1)
Then
jk
y,k ei
(2)
Then
( is scalar constant)
since g il = a y ,i y ,l
ijilei = ei g il
il ei ei gil = 0
( il g il ) ei = 0
(l = 1, 2, , n)
This equation implies that the direction of e i is a principal direction for the symmetric tensor il
i.e., e i is a principal direction for the hypersurface V n. Hence by definition the curve C is a line of
curvature V n.
Hypersurface
233
(1)
where A i is a covariant tenser of rank one and difference of two tensers Ai, jk Ai,kj is a covariant
tenser of rank three.
It y,i are components of a covariant tensor of rank one in x 's. Then replacing A i by y,i in
equation (1), we get
p
y,ijk y,ikj = y,p Rijk
= y,p g ph Rhijk ,
p
Since Rijk
= g phRhijk
(2)
and
y,ij = ij N
(3)
N ,i = ij g ik y,k
(4)
(5)
Multiplying equation (5) by a y,l and summed with respect to . Using the relations
a y ,i y,l = 0
and
g = N y,l = 0,
we get
Rlijk = (lj ik lk ij ) + R y ,l y,i y ,j y,k
(6)
a N B = 0
we get ij ,k ik , j + R N y ,i y,j y ,k = 0
(7)
Hence, The equation (6) are generalisation of the Gauss Characteristic equation and equation
(7) of the Mainardi-Codazzi equations.
234
(8)
where is an invariant
The mean curvature M of such a hypersurface is given by
M = ij g ij = gij g ij = n
M
.
n
M
g ,
n ij
from (i)
(9)
+ 1.
They
THEOREM 12.7 A totally geodesic hypersurface is a minimal hypersurface and its-lines of curvature
are indeterminate.
Proof: We know that
2
2
Ka2 = Kn + K g
M
g ij
n
(1)
(2)
(3)
M = ij g ij = 0
i.e., equation (2) is satisfied.
Hence, the totally geodesic hypersurface is minimal hypersurface.
In this case equation (3) are satisfied hence the lines of curvature are indeterminate.
Hypersurface
235
=1
(2)
(3)
s2
The equation (3) showing that the two values of s are equal in magnitude but opposite in sign.
The positive value of s given by equation (3) is the length of the radius of the quadric (2) for the
direction i .
THEOREM 12.8 The sum of the inverse squares of the radii of the quadric for n mutually orthogonal
directions at O is an invariant equal to a ij g ij .
Proof: If e ih| , (h = 1, 2, n) are the contravariant components of the unit tangents at O to the curves
of an orthogonal ennuple in Sn. The radius Sh relative to the direction e ih| is given by
aij ehi | ehj | =
n
or
h =1
(s h )2
1
s 2h
h
h =1
= a ij
n
h =1
(s h )2
eih | e hj|
h =1
= a ij g ij
Proved.
236
THEOREM 12.9 The equation of hyperplane of contact of the tangent hypercone with vertex at the
point Q (y ) .
Proof: Given (from equation 2, pg. 235)
aij y i y j = 1
Differentiating it
aij dy i y j + aij y i dy j = 0
or
aij dy i y j + aij y j dy i = 0
2 aij dy i y j = 0
aij dy i y j = 0
This shows that dy i is tangential to the quadric. Hence yj is normal to the quadric.
The tangent hyperplane at the point P ( y j) is given by
(Y i y i ) aij y j = 0
aij Y i y j = aij y i y j
aij Y i y j = 1
since aij y i y j = 1
(4)
(5)
Thus all points of the hyperquadric, the tangent hyperplanes at which pass through Q lie on the
hyperplane (5) on which yi is the current point. This is the hyperplane of contact of the tangent
hypercone whose vertex is Q ( y i ) .
12.13 POLAR HYPERPLANE
The polar hyperplane of the point R (~y i ) with respect to quadric (2) is the locus of the vertices of the
hypercones which touch the hyperquadric along its intersections with hyperplanes through R. If Q ( y i )
is the vertex of such tangent hypercone, then R lies on the hyperplane of contact of Q so that
yi a ~
yj =1
ij
Consequently for all positions of the hyperplane through R, Q lies on the hyperplane
yi a ~
yj =1
ij
(6)
This is required equations of the polar hyperplane of R and R is the pole of this hyperplane.
Hypersurface
237
y ( = 1, 2,L n + 1)
Let N be a unit normal vector at P relative to Sn + 1 so that tensor derivate N becomes covariant
derivative.
So,
N ;i = N ,i = ij g jk y,k
(1)
y;ij = y,ij = ij N
(2)
n +1
and
gij =
=1
y,i y,j
(3)
Let P ( y ) be a point on the unit normal N such that distance of P from P is in the direction
of N such that
y = y + N
(4)
Suppose P undergoes a displacement dx i in V n then the corresponding displacement d y of P is
given by
d y = ( y,i + N ,i ) dx i + N d
(5)
(6)
(7)
This shows that the directions dxi given by equation (7) are principal directions of the hypersurface
where the roots of the equation |gij ij| = 0 are called principal radii of normal curvature. The
locus of P ( y ) satisfying the condition (4) is called evolute of the hypersurface V n of Sn + 1 where
is a root of (7). The evolute is also a hypersurface of Sn + 1.
238
12.15 HYPERSPHERE
The locus of a point in Sn which moves in such way that it is always at a fixed distance R from a fixed
point C (b ) is called a hypersphere of radius R and centre C. Therefore the equation of such a
hypersphere is given by
n
(y
=1
b )2 = R 2
(1)
THEOREM 12.10 The Riemannian curvatrure of a hypersphere of radius R is constant and equal to
1
.
R2
Proof: Let the hypersurface be a V n of Sn + 1 and let its centre be taken as origin of Euclidean
coordinates in Sn + 1. Then the hypersphere is given by
(y
= R 2,
( = 1, 2, , n + 1)
(2)
For the point in V n the y's are functions of the coordinates xi on the hypersphere.
Differentiating equation (2) with respect to x i, we get
y y,i
=0
,j
y,i +
(3)
x j,
we get.
y,ij = 0
(4)
By Gauss formula,
y,ij = ij N
Using (5), equation (4) becomes
g ij +
y ij N
=0
(5)
(6)
From equation (3) it follows that y,i is perpendicular to y . But y,i is tangential to V m. Hence
y is normal to V n.. The equation (2) implies that the components of the unit vector N are given by
N
y
y = RN
R
RN
or
or
g ij + R ij
N ij = 0
(N
) =0
g ij + R ij = 0
since ( N )2 = 1
(7)
240
To prove that for a space Vn with positive constant Riemannian curvature K these exists sets of n + 1 real
coordinate y satisfying the condition
n +1
( y
=1
) = 1
K
where R 2 =
1
K
(y ) 2
1
K
Proved.
EXAMPLE 2
Show that the directions of two lines of curvature at a point of a hypersurface are conjugate.
Solution
The principal direction e ih| are given by
ij ehi | ehj| = 0
(1)
ij
by
Kn = ij
i.e.,
Kn = 0
dxi dx j
ds ds
from (1)
EXAMPLE 4
To prove that if the polar hyperplane of the point R passes through a point P then that of P passes
through R.
Solution
Let P( y i ) and R ( y i ) be two points. The polar hyperplane of R ( y i ) is
242
11. Prove that the necessary and sufficient condition that system of hypersurface with unit normal N be
isothermic is that
( N div N N .N ) = 0.
12. Show that the normal curvature of a subspace in an asymptotic direction is zero.
13. If straight line through a point P in S n meets a hyperquadric in A and B and the polar hyperplane of
P in Q prove that P, Q are harmonic conjugates to A, B.
14. What are evolutes of a hypersurface in an Euclidean space. Show that the varieties 1 = constant in
evolute are parallel, having the curves of parameter 1 as orthogonal geodesics of V n .
INDEX
A
Absolute 131
Addition and Subtraction of Tensors 15
Associated tensor 43
Asymptotic directions 229
B
Bianchi identity 94
Binormal 137
C
Canonical congruence 213
Christoffels symbols 55
Completely skew-symmetric 111
Completely symmetric 111
Concept 188
Congruence of curves 49
Conjugate (or Reciprocal) Symmetric Tensor 25
Conjugate directions 229
Conjugate metric tensor 34
Conservative force field 144
Contraction 18
Contravariant tensor of rank r 14
Contravariant Tensor of rank two 9
Contravariant vector 7
Covariant tensor of rank s 14
D
Degree of freedom 157
Dextral Index 1
Divergence 75
Divergence Theorem 161
Dummy index 1
Dupins Theorem 226
E
Einstein space 103
Einstein tensor 95
Einsteins Summation Convention 1
Eulers condition 171
Eulers Theorem 228
Evolute 237
F
First curvature 227
First curvature vector of curve 170
244
G
Gauss Characteristic equation 233
Gausss formula 222
Gausss theorem 164
Generalised Krnecker delta 112
Generalized coordinates 157
Geodesic congruence 208
Geodesic coordinate system 175
Geodesics 171
Gradient 75
Greens Theorem 162
H
Hamiltons principle 153
Hypersphere 238
Hypersurface 48, 218
I
Inner product of two tensors 18
Integral of energy 155
Intrinsic derivative 131
K
Kinetic energy 144
Krnecker Delta 2
L
Lagrangean equation 148
Lagrangean function 148
Laplaces equation 167
Laplacian operator 80, 163
Length of a curve 42
Levi-civitas concept 188
Line element 31
Line of curvature 227
M
Magnitude of vector 44
Mainardi-Codazzi equations 233
Mean curvature 101
Meauniers Theorem 225
Metric Tensor 31
Minimal curve 42
Minimal Hypersurface 227, 234
mixed tensor of rank r + s 14
mixed tensor of rank two 9
N
Newtonian Laws 142
Normal congruence 209
n-ply orthogonal system of hypersurfaces 49
Null curve 42
O
Orthogonal Cartesian coordinates 120
Orthogonal ennuple 49
Osculating plane 136
Outer Product of Tensor 16
P
Parallel vector fields 134
Parallelism 188
Poissons equation 166
Polar hyperplane 236
Potential energy 145
Principal directions of the hypersurface 227
Principle normal vector 136
Principle normal curvatures 227
Principle of least action 156
Projective curvature tensor 104
Q
Quotient Law 24
R
Reciprocal base systems 122
Index
Relative Tensor 26
Ricci tensor 88
Riccis principal directions 102
Riccis coefficients of rotation 205
Riccis Theorem 71
Riemann curvature 96
Riemanns symbol 86
Riemann-Christoffel Tensor 85
Riemannian coordinates 177
Riemannian Geometry 31, 116
Riemannian Metric 31
Riemannian space 31
245
T
Tensor density 26
Torsion 137
Totally geodesic hypersurface 234
Transformation of Coordinates 6
U
S
Scalar product of two vectors 44
Schurs theorem 100
Second fundamental tensors 34
Serret-Frenet formula 138
Simple Pendulum 152
Skew-Symmetric Tensor 20
Umbral 1
Unit principal normal 171
W
Weyl tensor 104
Work function W 145