Tensor Analysis - Copy To PDF
Tensor Analysis - Copy To PDF
Tensor Analysis - Copy To PDF
Introduction
Scalars are specified by magnitude only and vectors have magnitude as well as
direction. Tensors are a natural and logical generalization of vectors. There are
associated with magnitude and two or more directions.
A tensor can be specified only by its components which depend upon the system of
reference. The components of the same tensor will be different for two different
sets of axes with different orientations.
Consider the relation connecting electrical current density J in a medium and the
applied electric field 𝐸⃗ .The relation is 𝐽 = 𝜎𝐸⃗
It should be realized that this law is a special case and apply strictly only to
isotropic media or to a media possessing high symmetry. In reality, many media
are anisotropic and the current flows in a direction different from that of the
electric field .In such a situation the above equation can be written in a generalized
form as
𝐽𝑥 = 𝜎𝑥𝑥 𝐸𝑥 + 𝜎𝑥𝑦 𝐸𝑦 + 𝜎𝑥𝑧 𝐸𝑧
Where Jx, Jy, Jz and Ex, Ey, Ez are respectively, the Cartesian components of 𝐽 and
𝐸⃗ and σij (i, j= x, y, z) are said to components of the conductivity tensor of
the medium. The conductivity tensor has components. In the Cartesian co-ordinate
system xyz, J has three components, E has three components and σ has 9
components.
Physical law must be independent of any particular co-ordinate system. The above
equations which is written in the xyz co-ordinate system
𝐽𝑥 𝜎𝑥𝑥 𝜎𝑥𝑦 𝜎𝑥𝑧 𝐸𝑥
[𝐽𝑦 ] = [𝜎𝑦𝑥 𝜎𝑦𝑦 𝜎𝑦𝑧 ] [𝐸𝑦 ]
𝐽𝑧 𝜎𝑧𝑥 𝜎𝑧𝑦 𝜎𝑧𝑧 𝐸𝑧
The elements of the new conductivity tensor are related to the elements of old
conductivity tensor in accordance with a certain mathematical rule. A study of the
consequences of the requirements of co-ordinate transformation leads is to tensor
analysis
The development of tensor analysis was due to Ricci(1889) and his pupil Levi-
civita (1901).The term tensor was introduced by Einstein (1916).It is of great use
in rigid body dynamics , elasticity ,hydrodynamics ,electromagnetic theory ,the
general theory of relativity and numerous other field of science and Engineering
Co-ordinate Transformations
Let XOY and X′OY′ be two co-ordinate systems in two dimensional spaces. Let
the co-ordinates of the point P be (x, y) in XOY and (𝑥̅ , 𝑦̅) in X′OY′
Then
𝑥̅ = 𝑥 𝑐𝑜𝑠𝜃 + 𝑦 𝑠𝑖𝑛𝜃
𝑦̅ = −𝑥 𝑠𝑖𝑛𝜃 + 𝑦 𝑐𝑜𝑠𝜃
𝑥̅ cos 𝜃 sin 𝜃 𝑥
[ ]=[ ][ ]
𝑦 − sin 𝜃 cos 𝜃 𝑦
If x and y are replaced by x′ and 𝑥̅ and 𝑦̅ are replaced by𝑥̅ 1 𝑎𝑛𝑑 𝑥̅ 2
𝑥̅ 1 = cos 𝜃𝑥 1 + sin 𝜃𝑥 2
𝑥̅ 2 = (−sin 𝜃)𝑥 1 + cos 𝜃𝑥 2
𝑥̅ 1 = 𝑓(𝑥 1 , 𝑥 2 ) 𝑎𝑛𝑑 𝑥̅ 2 = 𝑔(𝑥 1 , 𝑥 2 )
𝑥̅ 1 = 𝑥̅ 1 (𝑥 1 , 𝑥 2 ) 𝑎𝑛𝑑 𝑥̅ 2 = 𝑥̅ 2 (𝑥 1 , 𝑥 2 )
Let (x1 ,x2…xN) be a point in one system .Let the co-ordinates of the same point in
another system be (𝑥̅ 1 , 𝑥̅ 2 , … . 𝑥̅ 𝑁 )
Then 𝑥̅ 1 = 𝑥̅ 1 (𝑥 1 , 𝑥 2 … . 𝑥 𝑁 )
𝑥̅ 2 = 𝑥̅ 2 (𝑥 1 , 𝑥 2 … . 𝑥 𝑁 )
𝑥̅ 𝑁 = 𝑥̅ 𝑁 (𝑥 1 , 𝑥 2 … . 𝑥 𝑁 )
It is obvious that a point can have only one set of co-ordinates in a particular
system i.e. there should be one to one correspondence between 𝑥̅ 𝑖 𝑎𝑛𝑑 𝑥̅ 𝑖 .This
requirement is satisfied if the functions involved above are single valued
continuous and have continuous derivatives. Then conversely, to each set of co-
ordinates (𝑥̅ 1 , 𝑥̅ 2 … . 𝑥̅ 𝑁 ) there correspond a unique set
𝑥 𝑖 = 𝑥 𝑖 (𝑥̅ 1 , 𝑥̅ 2 … . . 𝑥̅ 𝑁 ) ⟶ (2)
The simplified notation used by Einstein is aixi. Here we have omitted Σ sign. Here
we adopt the convention that whenever an index (superscript or subscript)is
repeated in a given term , we have to sum over that index from 1 to N , under
otherwise specified .This is called the summation convention.
Instead of using the index ‘i’, we could have used another letter, say ‘p’ and the
sum could be written as apxp. Any index that is repeated in a term to indicate the
summation is called a dummy index. An index occurring only once in a given term
is called a free index.
𝑥 𝑖 = 𝑥 𝑖 (𝑥̅ 1 , 𝑥̅ 2 , … 𝑥̅ 𝑁 ), 1 ≤ 𝑖 ≤ 𝑁 ⟶ (1)
Differentiating
𝜕𝑥 𝑖
𝑑𝑥 𝑖 = ∑𝑁
𝛼=1 𝑑𝑥̅ 𝛼 1 ≤ 𝑖 ≤ 𝑁 ⟶ (2)
𝜕𝑥̅ 𝛼
𝑥̅ 𝛼 = 𝑥̅ 𝛼 (𝑥 1 , 𝑥 2 , … . 𝑥 𝑁 ), 1 ≤ 𝑖 ≤ 𝑁 ⟶ (3)
𝜕𝑥 −𝛼
𝑑𝑥 −𝛼 = ∑𝑁
𝑖=1 𝑑𝑥 𝑖 1 ≤ 𝛼 ≤ 𝑁 ⟶ (4)
𝜕𝑥 𝑖
𝜕𝑥̅ 𝛼
𝑑𝑥 −𝛼 = 𝑑𝑥 𝑖 1 ≤ 𝑖 ≤ 𝑁 ⟶ (6)
𝜕𝑥 𝑖
The index ‘i’ appear only once in eqn (2) and (5), hence ‘i’ is a free index .In eqns
(4) and (6),’α’ is a free index.
1 𝑖𝑓 𝑖=𝑗
𝛿𝑗 𝑖 = {
0 𝑖𝑓 𝑖≠𝑗
𝑑𝑥 𝑖
Therefore we can write = 𝛿 𝑖𝑗
𝑑𝑥 𝑗
𝑑𝑥̅ 𝛼
Similarly the co-ordinates 𝑥̅ 𝛼 are independent of each other .Hence = 𝛿ᵦ𝛼 .
𝑑𝑥̅ 𝛽
𝑑𝑥 𝑖 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛼
=
𝑑𝑥 𝑗 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛼
Therefore we have . = 𝛿 𝑖𝑗
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑘
Similarly = 𝛿 𝛼𝛽
𝜕𝑥 𝑘 𝜕𝑥̅ 𝛽
Rank of a tensor
The rank of a tensor is the number of indices in the symbol represents a tensor. It is
the number of ways by which a physical quantity changes or it is the number of
components of physical quantity.
Tensors are specified in terms of their rank and components. Scalar or invariants
are tensors of rank zero. Vector (having N1 components in N-dimensional space)
are tensors of rank 1.Displacement, velocity, electric field , momentum, vector
potential or examples. Stress, conductivity, moment of inertia etc in anisotropic
media requires N2 components for its complete description in an N-dimensional
space and is examples of tensor of rank 2.
Let the co-ordinates of P be (x1,x2) and that of Q be (x1+δx1, x2+δx2).Let the co-
ordinates be changed from
(𝑥 1 , 𝑥 2 )𝑡𝑜(𝑥̅ 1 , 𝑥̅ 2 )
𝑥̅ 1 = 𝑥̅ 1 (𝑥 1 , 𝑥 2 )
𝜕𝑥̅ 1 𝜕𝑥̅ 1
𝛿𝑥̅ 1 = 1
𝛿𝑥 1 + 𝛿𝑥 2
𝜕𝑥 𝜕𝑥 2
In N dimension
𝜕𝑥̅ 1 𝜕𝑥̅ 1 𝜕𝑥̅ 𝑁
𝛿𝑥̅ 1 = 1
𝛿𝑥 1 + 2
𝛿𝑥 2 + ⋯ + 𝛿𝑥 𝑁
𝜕𝑥 𝜕𝑥 𝜕𝑥 𝑁
𝜕𝑥̅ 1
= ∑𝑁
𝑖=1 𝛿𝑥 𝑖
𝜕𝑥 𝑖
𝜕𝑥̅ 1
= 𝛿𝑥 𝑖
𝜕𝑥 1
In general
𝜕𝑥̅ 𝛼
𝛿𝑥̅ 𝛼 = 𝛿𝑥 𝑖
𝜕𝑥 𝑖
𝜕𝑥̅ 𝛼
𝐴̅𝛼 = 𝑖 𝐴𝑖
𝜕𝑥
𝐴1̅ , 𝐴̅2 , 𝐴̅3 … 𝐴̅𝑁 in another co-ordinate system (𝑥̅ 1 , 𝑥̅ 2 , … ) by the transformation
𝜕𝑥̅ 𝛼
equations 𝐴𝛼̅ = 𝑖 𝐴𝑖 ,they are the components of a contra variant vector or the
𝜕𝑥
contra variant tensor of first rank.
𝜕∅ 𝜕∅ 𝜕∅
, ,… ⟶ (1)
𝜕𝑥 1 𝜕𝑥 2 𝜕𝑥 𝑁
But 𝑥̅ 1 = 𝑥̅ 1 (𝑥 1 , 𝑥 2 … … 𝑥 𝑁 )
𝜕∅ 𝜕∅ 𝜕𝑥 1 𝜕∅ 𝜕𝑥 2 𝜕∅ 𝜕𝑥 𝑁
= . + . + ⋯+ .
𝜕𝑥̅ 1 𝜕𝑥 1 𝜕𝑥̅ 1 𝜕𝑥 2 𝜕𝑥̅ 1 𝜕𝑥 𝑁 𝜕𝑥̅ 1
In general
𝜕∅ 𝜕∅ 𝜕𝑥 1 𝜕∅ 𝜕𝑥 2 𝜕∅ 𝜕𝑥 𝑁
= . + + ⋯+ .
𝜕𝑥̅ 𝛼 𝜕𝑥 1 𝜕𝑥̅ 𝛼 𝜕𝑥 2 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑁 𝜕𝑥̅ 𝛼
𝜕∅ 𝜕∅ 𝜕𝑥 𝑖
i.e. = .
𝜕𝑥̅ 𝛼 𝜕𝑥 1 𝜕𝑥̅ 𝛼
N quantities A1, A2…AN in one co-ordinate system are said to be the component of
covariant tensor of rank 1 if they transform to a new co-ordinate system
𝑥̅ 𝛼 (𝑥̅ 1 , 𝑥̅ 2 , … . 𝑥̅ 𝑁 ) according to the relation
𝜕𝑥 𝑖
𝐴̅𝛼 = 𝐴𝑖
𝜕𝑥̅ 𝛼
1.A set of N2 function Aij are said to be the components of a contra variant tensor
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽
of rank 2,if they transform according to the relation 𝐴̅𝛼𝛽 = 𝑖 𝐴𝑖𝑗 ,under co-
𝜕𝑥 𝜕𝑥 𝑗
ordinate transformation ; where 𝐴̅𝛼𝛽 are the components of the tensor in barred
co-ordinate system
2. A set of N2 quantities Aij are said be the components of a co-variant tensor rank
two if they transforms according to the relation
𝜕𝑥 𝑖 𝜕𝑥 𝑗
𝐴̅𝛼𝛽 = 𝐴𝑖𝑗 under co-ordinate transformations where 𝐴̅𝛼𝛽 are the
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽
components of the tensor in the barred co-ordinate system.
3.A set of N2 functions Aij are said to be the components of a tensor of contra
variant with rank one and co-variant rank one (simply a mixed tensor of rank 2 if
they transforms according to
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐴̅𝛼𝛽 = 𝑖 𝐴𝑖𝑗 under co-ordinate transformations
𝜕𝑥 𝜕𝑥̅ 𝛽
General definitions
𝑖 ,𝑖 ….𝑖
A set of Np+q functions𝐴𝑗1,𝑗2 ….𝑗𝑝 , are said to be the components of a mixed tensor of
1 2 𝑞
contra variant rank P and co-variant rank q (total rank p+q) if they transform under
co-ordinate transformations to
𝛼 ,𝛼 ,…𝛼 𝜕𝑥̅ 𝛼1 𝜕𝑥̅ 𝛼2 𝜕𝑥̅ 𝛼𝑝 𝜕𝑥 𝑗1 𝜕𝑥 𝑗2 𝜕𝑥 𝑗𝑞 𝑖 ,𝑖 ,….𝑖
𝐴𝛽̅ 1,𝛽 2,….𝛽𝑝 = . …… . . 𝛽2 … … 𝐴𝑗1,𝑗2 ,….𝑗𝑝
1 2 𝑞 𝜕𝑥 𝑖1 𝜕𝑥 𝑖2 𝜕𝑥 𝑖𝑝 𝜕𝑥̅ 𝛽1 𝜕𝑥̅ 𝜕𝑥̅ 𝛽𝑞 1 2 𝑞
A scalar is a tensor of rank zero and this has N0=1 component. A vector is a tensor
of rank 1 and has N1=N components in N-dimensional space.
𝐴1 𝐴1
2 𝐴
𝐴𝑖 = [ 𝐴 ] 𝑜𝑟𝐴𝑖 = [ 2 ]
⋮ ⋮
𝑁 𝐴𝑁
𝐴
A second order tensor can be represented by an N×N square matrix
Properties
𝜕𝑥 𝑖
𝟏. = 𝛿𝑗𝑖
𝜕𝑥 𝑗
𝜕𝑥̅ 𝛼
= 𝛿𝛽𝛼
𝜕𝑥̅ 𝛽
𝟐. 𝛿𝑗𝑖 𝐴𝑖 = 𝐴𝑖
= 𝐴𝑖
3. 𝛿𝑖𝑖 = 𝑁
=N
𝑗
4. 𝛿𝑗𝑖 𝛿𝑘 = 𝛿𝑘𝑖
𝑗
𝛿𝑗𝑖 𝛿𝑘 = 𝛿1𝑖 𝛿𝑘1 + 𝛿2𝑖 𝛿𝑘2 + ⋯ +𝛿𝑖𝑖 𝛿𝑘𝑖 + ⋯ 𝛿𝑁𝑖 𝛿𝑘𝑁
= 𝛿𝑘𝑖
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑖
5. 𝛼
. = = 𝛿𝑘𝑖
𝜕𝑥̅ 𝜕𝑥 𝑘 𝜕𝑥 𝑘
Proof:
𝜕𝑥 𝑖 𝜕𝑥 𝑖 𝜕𝑥̅ 1 𝜕𝑥 𝑖 𝜕𝑥̅ 2 𝜕𝑥 𝑖 𝜕𝑥̅ 𝑁
𝛿𝑘′ = = + + ⋯+
𝜕𝑥 𝑘 𝜕𝑥̅ 1 𝜕𝑥 𝑗 𝜕𝑥̅ 2 𝜕𝑥 𝑗 𝜕𝑥̅ 𝑁 𝜕𝑥 𝑗
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛼
= .
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
Qn) Show that Kronecker delta is a mixed tensor
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑖 𝜕𝑥 𝑗
= .
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
= 𝛿𝑗𝑖
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛽
Algebra of tensors
1. Equality
𝑖 ,𝑖 ……𝑖 𝑖 ,𝑖 …𝑖
Two tensors 𝐴𝑗1,𝑗2 …..𝑗 𝑝 𝑎𝑛𝑑𝐴𝑗1,𝑗2 ….𝑗𝑞 are said to be equal if and only if they have the
1 2 𝑞 1 2 𝑞
same contra variant rank and the same co-variant rank and every component of one
is equal to the corresponding component of the other
𝑖 ,𝑖 ……..𝑖 𝑖 ,𝑖 ……𝑖
𝐴𝑗1,𝑗2 ……..𝑗𝑝 = 𝐵𝑗 1,𝑗 2…….𝑗𝑝
1 2 𝑞 1 2 𝑞
If two tensors have the same contra variant rank and same co-variant rank they are
said to be of the same type
2. Null tensor
If all the Nr components of a tensor of rank ‘r’ vanish, it is said to be a null tensor.
Two tensors of same type and rank can be added or subtracted from one another.
The result will have the same type and rank.
𝑖𝑗 𝑖𝑗 𝑖𝑗 𝑖𝑗
If 𝐴𝑘 and 𝐵𝑘 are two tensors, then their sum 𝐶𝑘 and difference 𝐷𝑘 are of the
same type and rank as that of A and B
𝜕𝑥̅ 𝜕𝑥̅ 𝜕𝑥 𝛼 𝛽 𝑘
𝛼𝛽 𝑖𝑗
𝐵̅𝛾 = 𝑖 . 𝑗 . 𝛾 𝐵𝑘
𝜕𝑥 𝜕𝑥 𝜕𝑥̅
𝛼𝛽 𝛼𝛽 𝛼𝛽
𝐶𝛾̅ = 𝐴̅𝛾 + 𝐵̅𝛾
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝑖𝑗 𝑖𝑗
= . . (𝐴𝑘 + 𝐵𝑘 )
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝛼𝛽 𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝑖𝑗
𝐶𝛾̅ = . . 𝐶𝑘
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
̅𝛾𝛼𝛽 = 𝐴𝛼𝛽
Difference 𝐷 ̅𝛾 − 𝐵̅𝛾𝛼𝛽 e
𝛼 𝛽 𝑘
̅𝛾𝛼𝛽 = 𝜕𝑥̅ 𝑖 . 𝜕𝑥̅ 𝑗 . 𝜕𝑥 𝛾 (𝐴𝑖𝑗
𝐷
𝑖𝑗
𝑘 − 𝐵𝑘 )
𝜕𝑥 𝜕𝑥 𝜕𝑥̅
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝑖𝑗
= . . 𝐷𝑘
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
̅𝛾𝛼𝛽 =
𝐷 . .
𝑖𝑗
𝐷𝑘
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝜕𝑥̅ 𝜕𝑥 𝛼 𝛽 𝑘
𝛼𝛽 𝑖𝑗
𝐴̅𝛾 = 𝑖 . 𝑘 . 𝛾 𝐴𝑘
𝜕𝑥 𝜕𝑥 𝜕𝑥̅
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝑖𝑗
= . . 𝐴𝑘
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑘 𝑖𝑗
= . . 𝐴𝑘
𝜕𝑥 𝑗 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛾
𝛼𝛽 𝛽𝛼
i.e. 𝐴̅𝛾 = 𝐴̅𝛾
𝜕𝑥̅ 𝜕𝑥̅ 𝜕𝑥 𝛼 𝛽 𝑘
𝛼𝛽 𝑖𝑗
𝐴̅𝛾 = 𝑖 . 𝑗 . 𝛾 𝐴𝑘
𝜕𝑥 𝜕𝑥 𝜕𝑥̅
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝑖𝑗
= . . (−𝐴𝑘 )
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑘 𝑖𝑗
= . . 𝐴𝑘
𝜕𝑥 𝑗 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛾
𝛽𝛼
= −𝐴̅𝛾
Any tensor can be represented as the sum of two tensor, one of which is symmetric
and the other is skew symmetric in a pair of contra variant and co-variant indices.
𝑖𝑗
Let 𝐴𝑘 be a tensor
𝑖𝑗 1 𝑖𝑗 𝑗𝑖 1 𝑖𝑗 𝑗𝑖
𝐴𝑘 = [𝐴𝑘 + 𝐴𝑘 ] + [𝐴𝑘 − 𝐴𝑘 ]
2 2
= B+C
Illustration
𝑖𝑗 𝑙
Let 𝐴𝑘 𝑎𝑛𝑑 𝐵𝑚 are two tensors. The transformation equations are
𝜕𝑥̅ 𝜕𝑥̅ 𝜕𝑥 𝛼 𝛽 𝑘
𝛼𝛽 𝑖𝑗
𝐴̅𝛾 = 𝑖 . 𝑗 . 𝛾 𝐴𝑘
𝜕𝑥 𝜕𝑥 𝜕𝑥̅
𝜕𝑥̅ 𝜕𝑥 𝜌 𝑚
𝜌
𝐵̅𝛿 = 𝑙 . 𝛿 𝐵𝑚
𝑙
𝜕𝑥 𝜕𝑥̅
𝜕𝑥̅ 𝛾
𝐵̅𝛾 = 𝑘 𝐵𝑘
𝜕𝑥
𝜕𝑥 𝑚
𝐶𝛿̅ = 𝐶𝑚
𝜕𝑥̅ 𝛿
𝛼 𝛾 𝜕𝑥 𝑗 𝜕𝑥 𝑚
̅ 𝛼𝛾 = 𝜕𝑥̅ 𝑖 . 𝜕𝑥̅ 𝑘
𝐷 𝐷𝑗𝑚
𝑗𝑘
𝛽𝛿 𝜕𝑥 𝜕𝑥 𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛿
Contraction of a tensor
𝑖𝑗𝑘
Consider a tensor 𝐴𝑙𝑚 of contra variant rank 3 and covariant rank 2 which has N5
compound .Let any one of the contra variant indices be equated to any one of co-
variant indices and summed over from 1 to N.
𝑖𝑗𝑘
Consider a set of the form 𝐴𝑙𝑚 . Here ‘i’ is a dummy index while j, k and m are
free indices. By convention we have
𝑖𝑗𝑘 1𝑗𝑘 2𝑗𝑘 3𝑗𝑘 𝑁𝑗𝑘
𝐴𝑖 = 𝐴1𝑚 + 𝐴2𝑚 + 𝐴3𝑚 + ⋯ + 𝐴𝑁𝑚
𝑖𝑗𝑘
The entity 𝐴𝑙𝑚 has N3 components i.e. it is a tensor of rank 3.
𝑖𝑗𝑘
Consider the transformation of the tensor 𝐴𝑙𝑚
𝑖𝑗𝑘
This shows that 𝐴𝑖𝑚 is a tensor of contravariant rank 2 and covariant rank 1.This
process is known as the contraction of a tensor. When a tensor is contracted by
equaling one of its contra variant indices to one of its co-variant indices, the
resulting entity is a tensor whose contra variant and covariant ranks is reduced by
one each, thus reduced the total rank by 2.
𝑖𝑗𝑘 𝑖𝑗𝑘 𝑖𝑗𝑘 𝑖𝑗𝑘
𝐴𝑙𝑖 , 𝐴𝑗𝑚 , 𝐴𝑙𝑘 are various contracted forms of tensor 𝐴𝑙𝑚
𝑖𝑗𝑘
A tensor can be repeatedly contracted. A tensor 𝐴𝑖𝑚 of total rank 5, on contraction
𝑖𝑗𝑘
gives the tensor 𝐴𝑖𝑚 of total rank 3 which can be further contracted to give the
𝑖𝑗𝑘 𝑖𝑗𝑘
tensor 𝐴𝑖𝑗 𝑜𝑟 𝐴𝑖𝑘 contravariant rank 1.
Qn) Contract the tensor 𝑨𝒊𝒋 and establish that scalars are tensors of rank zero.
𝜕𝑥̅ 𝜕𝑥 𝛼 𝑗
𝐴𝛽𝛼̅ = 𝑖 . 𝛽 . 𝐴𝑗𝑖
𝜕𝑥 𝜕𝑥̅
When α=β
𝜕𝑥̅ 𝜕𝑥 𝛼 𝑗
𝐴𝛽𝛼̅ = 𝑖 . 𝛼 . 𝐴𝑗𝑖
𝜕𝑥 𝜕𝑥̅
𝜕𝑥 𝑗 𝑗
= 𝐴𝑗𝑖 = 𝛿𝑖 𝐴𝑗𝑖
𝜕𝑥 𝑖
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝜕𝑥 𝑙
𝛼𝛽 𝑖𝑗
𝐴̅𝛶𝛿 = 𝑖 𝐴𝑘𝑙
𝜕𝑥 𝜕𝑥 𝑘 𝜕𝑥̅ 𝛶 𝜕𝑥̅ 𝛿
𝜕𝑥̅ 𝛽 𝜕𝑥 𝑙 𝑖𝑗
= 𝛿𝑖𝑘 𝐴𝑘𝑙
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛿
𝛼𝛽 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑙 𝑖𝑗
𝐴̅𝛼𝛿 = 𝐴𝑖𝑙
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛿
𝑗 𝑗
𝛿𝑗𝑙 𝐴𝑙 = 𝐴𝑗
𝛽 𝑗
𝐴𝛽̅ = 𝐴𝑗
𝑖𝑗
Thus a double contraction on the tensor 𝐴𝑘𝑙 is an invariant
𝑖𝑗
During the first contraction, the rank of 𝐴𝑘𝑙 becomes 2 which was initially 4.After
𝑗
the second contraction the rank of it becomes zero which was 2. Thus 𝐴𝑗 becomes
scalar which is an invariant or a tensor of rank zero.
Inner product
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑞 𝑖𝑗 𝑝
= 𝛿𝑝𝑘 𝐴𝑘𝑙 𝐵𝑞
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝜎
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑞 𝑖𝑗
= 𝐴𝑘 𝐵𝑞𝑘
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝜎
𝑖𝑗
The N3 entities 𝐴𝑘 𝐵𝑞𝑘 transforms like the components of a tensor of contravariant
rank 2 and covariant rank 1.Let the components of new tensor in barred and the
unbarred system will be
𝛼𝛽 𝛼𝛽 𝛾 𝑖𝑗 𝑖𝑗
Let 𝐶𝜎̅ = 𝐴̅𝛾 𝐵𝜎 𝐶𝑞 = 𝐴𝑘 𝐵𝑞𝑘
𝑖𝑗 𝑖𝑗 𝑝
𝐶𝑞 is said to be inner product of two tensors 𝐴𝑘 𝑎𝑛𝑑𝐵𝑞
𝑖𝑗 𝑝 𝑖𝑗 𝑝 𝑖𝑗 𝑝
𝐴𝑘 𝐵𝑖 , 𝐴𝑘 𝐵𝑗 are examples of inner product of 𝐴𝑘 𝑎𝑛𝑑𝐵𝑞 . In taking the inner
product of two tensors, one contra variant index of one tensor should be equated to
one covariant index of the other.
Transformation equation is
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐴𝛼̅ 𝐵̅𝛽 = 𝑖 𝐴𝑖 𝐵𝑗
𝜕𝑥 𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐶𝛽̅ 𝛼 = 𝐶𝑖
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛽 𝑗
Where 𝐶𝑗𝑖 = 𝐴𝑖 𝐵𝑗
𝑗
= 𝛿𝑖 . 𝐶𝑗𝑖
𝐶𝛼̅ 𝛼 = 𝐶𝑖𝑖 i.e. the transformation not connected by a partial derivative. It is an invariant.
Form two inner products of rank 2 and obtain a rank zero by suitable operations on
the tensors 𝐴𝑖𝑗 𝑎𝑛𝑑 𝐵𝑘𝑙
Transformation equation is
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
𝐴̅𝛼𝛽 𝐵̅𝛾𝛿 = 𝐴𝑖𝑗 𝐵𝑘𝑙
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝜕𝑥 𝑙
𝛾𝛿 𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
𝐶𝛶̅ = 𝐶𝑖𝑗𝑘𝑙
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝜕𝑥 𝑙
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛿
= 𝐶𝑖𝑗𝑘𝑙
𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛿
= 𝐶𝑖𝑗𝑖𝑙
𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
̅ 𝛼𝛿 can be written as 𝐷
From the above relation 𝐶𝛼𝛽 ̅𝛽𝛿 which is a tensor of rank two.
Since there are product of two partial derivatives in the transformation equation,the
above relation can be written as
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛿
̅𝛽𝛿 =
𝐷 𝐷𝑗𝑙
𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
Contracting again δ=β
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛽
̅𝛽 =
𝐷 𝐷𝑗𝑙
𝛽 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
𝑗 𝑗
= 𝛿𝑘 𝐷𝑗𝑙 = 𝐷𝑗
𝛽
̅ is a tensor of rank 0, which is invariant under
From the above relation 𝐷𝛽
transformation
i=j
=N
Quotient law
Quotient law states that if the inner product of an entity with an arbitrary tensor is a
tensor, the entity is a tensor
If the inner product of the function X (i,j)and Ak is find to be a tensor then X(i,j) is
also a tensor.
Illustration
A quantity A (i,j) is such that its ‘inner product’ with an arbitrary tensor Bk is a
contra variant tensor Ci show that A(i,j) a tensor. Deduce its nature and transform
equation
A (i, j) Bk = Ci
A (i , j) Bk = D ⟶(1)
𝜕𝑥 𝑗
Multiplying the R.HS and L.H.S of eqn (5) with
𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛽 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐴̅(𝛼,𝛽) 𝑗 𝛽
= 𝐴(𝑖,𝑗)
𝜕𝑥 𝜕𝑥̅ 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐴̅(𝛼,𝛽) = 𝑖 𝐴(𝑖,𝑗)
𝜕𝑥 𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝐴𝛽̅𝛼 = 𝑖 𝐴(𝑖,𝑗)
𝜕𝑥 𝜕𝑥̅ 𝛽
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
= 𝐴𝑗𝑖 ⟶ (𝟔)
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛽
The relation is tensorial, and the nature of the tensor is given by the above relation.
X (i, j, k) Al = C ⟶(1)
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥 𝑘
𝑋̅(𝛼,𝛽,𝛾) 𝐴 =
𝛼 𝑖
𝐵𝑗𝑘
𝜕𝑥̅ 𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛾
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥 𝑘
𝑋̅(𝛼,𝛽,𝛾) 𝐴 =
𝛼 𝑖
𝑋(𝑖,𝑗,𝑘)
𝜕𝑥̅ 𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝛼
Multiplying by we get
𝜕𝑥 𝑖
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗 𝜕𝑥 𝑘
𝑋̅(𝛼,𝛽,𝛾) 𝑖 𝑋(𝑖,𝑗,𝑘)
𝜕𝑥 𝜕𝑥̅ 𝛽 𝜕𝑥̅ 𝛾
Qn) If 𝑨𝒊𝒌 𝒂𝒏𝒅 𝑩𝒊𝒍 are two tensors, show that their outer product is not a
tensor.
𝑖𝑗
Transformation equation for𝐴𝑘 𝑎𝑛𝑑 𝐵𝑙𝑖
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
𝛼𝛽 𝑖𝑗
𝐴̅𝛾 = 𝑖 𝐴𝑘
𝜕𝑥 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑙
𝐵̅𝛿𝛼 = 𝑖 𝐵𝑙𝑖
𝜕𝑥 𝜕𝑥̅ 𝛿
𝑖𝑗
Outer product is 𝐴𝑘 𝑎𝑛𝑑 𝐵𝑙𝑖 is given by
𝜕𝑥̅ 𝛼 2 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝜕𝑥 𝑙 𝑖𝑗
=( ) 𝐴𝑘 𝐵𝑙𝑖
𝜕𝑥 𝑖 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
𝛼 2 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘 𝜕𝑥 𝑙
̅ 𝛼𝛼𝛽 = (𝜕𝑥̅ 𝑖 )
𝐶𝛾𝛿
𝑖𝑖𝑗
𝐶𝑘𝑙
𝜕𝑥 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
𝑖𝑖𝑗
From the above relation it is clear that 𝐶𝑘𝑙 is not a tensor because its
transformation equation contains partial derivative of degree two i.e. degree greater
𝑖𝑖𝑗 𝑖𝑗
than one. So 𝐶𝑘𝑙 , which is the outer product of 𝐴𝑘 𝑎𝑛𝑑 𝐵𝑙𝑖 is not a tensor
𝒊𝒋 𝒊𝒋
Qn) If 𝑨𝒌 is a tensor show that 𝑨𝒌𝒍 is not a tensor.
𝑖𝑗
Given 𝐴𝑘 is a tensor .So its transformation equation is
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽 𝜕𝑥 𝑘
𝛼𝛽 𝑖𝑗
𝐴̅𝛾 = 𝑖 𝐴𝑘
𝜕𝑥 𝜕𝑥 𝑗 𝜕𝑥̅ 𝛾
𝛼 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑘 𝜕𝑥 𝑙
̅𝛾𝛿 = 𝜕𝑥̅ 𝑖
𝐴𝛼𝛼 𝐴𝑖𝑖𝑘𝑙
𝜕𝑥 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
𝜕𝑥̅ 𝛼 2 𝜕𝑥 𝑘 𝜕𝑥 𝑙
=( ) 𝐴𝑖𝑖𝑘𝑙
𝜕𝑥 𝑖 𝜕𝑥̅ 𝛾 𝜕𝑥̅ 𝛿
From the above relation it is clear that 𝐴𝑖𝑖𝑘𝑙 is not a tensor. Became the
transformation equation of 𝐴𝑖𝑖𝑘𝑙 has terms which are partial derivatives of degree
2(greater than 1). A tensor is a quantity whose transformation equation will contain
terms which are partial derivatives of degree 0 or 1
=dxi dxi
[𝑑𝑠 2 = 𝑑𝑟 2 + 𝑟 2 𝑑𝜃 2 + 𝑟 2 𝑠𝑖𝑛²𝜃𝑑∅2 ]
Where gij are functions of xi , such a co-ordinate system is called Riemannian and
the corresponding space is called Riemannian space (Riemannian geometry is the
study of curved surface).If all the coefficients gij are independent of xi , the space
becomes a Euclidean space
Since the distance between any two points is independent of the co-ordinate
system, ds2 must be a scalar invariant
i.e. gij dxi dxj must be a tensor of rank 0.But dxi and dxj are contra variant vectors
and their outer product must be contra variant tensor of rank 2.Hence by quotient
law gij must be a co-variant tensor of rank 2.It is called the metric tensor or the first
fundamental tensor in Riemannian space. The components of gij can be represented
as a matrix is the form
𝑔11 𝑔12 … 𝑔1𝑁
𝑔21 𝑔22 …
𝑔𝑖𝑗 = [ ⋮ 𝑔2𝑁 ]
⋮ ⋮
𝑔𝑁1 𝑔𝑁2 … 𝑔𝑁𝑁
|gij|=1
2. In spherical polar co-ordinates
𝑑𝑠 2 = 𝑑𝑟 2 + 𝑟 2 𝑑𝜃 2 + 𝑟 2 𝑠𝑖𝑛2 𝜃𝑑∅2
1 0 0
𝑔𝑖𝑗 = [0 𝑟 2 0 ]
0 0 𝑟 𝑠𝑖𝑛2 𝜃
2
|𝑔𝑖𝑗 | = 𝑟 4 𝑠𝑖𝑛2 𝜃
3. In cylindrical co-ordinates
2
𝑑𝑠 2 = 𝑑𝜌2 + 𝑝2 𝑑∅2 + 𝑑 2
1 0 0
𝑔𝑖𝑗 = [0 𝜌2 0]
0 0 1
|𝑔𝑖𝑗 | = 𝜌2
Qn) Find the metric for a two dimensional surface of a sphere of constant
radius ‘a’.
r = a, const
Generally ds2=dr2+r2dθ2+r2sin2θdϕ2
dr = 0
ds2=a2dθ2+a2sin2θdϕ2
|𝑔𝑖 | = 𝑎4 𝑠𝑖𝑛2 𝜃
Qn) Consider a three dimensional co-ordinate system (u, v, w) which are
related to the Cartesian co-ordinate system by x = vw, y = uw and z = uw.
Obtain the Riemannian metric of the co-ordinate system in terms of u, v, w
and Show that |gij|= 4 u2 v2 w2
ds2= dx2+dy2+dz2
dx2= (v dw + w dv )2
= v2dw2+w2dv2+2vwdwdv
= u2dw2+w2du2+2uwdwdu
dz2 = (udv+vdu)2
= u2dv2+v2du2+2uvdudv
ds2 = dx2+dy2+dz2
= v2dw2+w2dv2+2vwdwdv+u2dw2+w2du2+2uwdwdu+u2dv2+v2du2+2uvdudv
𝑣̅ 2 + 𝑤 2 𝑢𝑣̅ 𝑢𝑤
𝑔𝑖𝑗 = [ 𝑢𝑣̅ 𝑢 + 𝑤2
2
𝑣̅𝑤 ]
𝑢𝑤 𝑣̅𝑤 𝑢 + 𝑣̅ 2
2
𝑣̅ 2 + 𝑤 2 𝑢𝑣̅ 𝑢𝑤
𝑔𝑖𝑗 = [ 𝑢𝑣̅ 𝑢 + 𝑤2
2
𝑣̅𝑤 ]
𝑢𝑤 𝑣̅𝑤 𝑢 + 𝑣̅ 2
2
= [v2+w2][(u2+w2)(u2+v2)-v2w2] – uv [uv(u2+v2)-uvw2]+uw[uwv2+uw(u2+w2)]
= (v2+w2)[u4+u2v2+w2u2+w2v2-v2w2] -uv[uvu2+uv3-
uvw2]+uw[uwv2+uwu2+uw3]
= 4u2v2w2
Qn) Find the fundamental metric tensor and its determinant in which the line
𝟐 𝟐
element is given by 𝒅𝒔𝟐 = 𝟑𝒅𝒙𝟏 + 𝟑𝒅𝒙𝟑 + 𝟒𝒅𝒙𝟏 𝒅𝒙𝟐 + 𝟖𝒅𝒙𝟏 𝒅𝒙𝟑 +
𝟖𝒅𝒙𝟐 𝒅𝒙𝟑 𝒂𝒏𝒅 𝑺𝒉𝒐𝒘 𝒕𝒉𝒂𝒕 |𝒈𝒊𝒋 | = 𝟒
3 2 4
𝑔𝑖𝑗 = [2 0 4]
4 4 3
|gij|=3(-16)-2(6-16) +4(8)
= - 48+20+32
=4
ds2 is invariant
But 𝑥 𝑖 = 𝑥 𝑖 (𝑥̅ 𝛼 )
𝑥 𝑗 = 𝑥 𝑗 (𝑥̅ 𝛽 )
𝜕𝑥 𝑖 𝜕𝑥 𝑗
𝑑𝑠 2 = 𝑔𝑖𝑗 𝛼
𝑑𝑥̅ 𝛼 𝑑𝑥̅ 𝛽
𝜕𝑥̅ 𝜕𝑥̅ 𝛽
𝜕𝑥 𝑖 𝜕𝑥 𝑗
= 𝑔𝑖𝑗 𝑑𝑥̅ 𝛼 𝑑𝑥̅ 𝛽 ⟶ (3)
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽
𝜕𝑥 𝑖 𝜕𝑥 𝑗
𝑔̅𝛼𝛽 = 𝑔𝑖𝑗 ⟶ (4)
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽
i.e. gij is a covariant tensor of rank two . α and β are the indices which represent the
coordinates (𝑥̅ 1 , 𝑥̅ 2 , … 𝑥̅ 𝑁 ).Similarly i and j are indices representing the co-
ordinates (x1,x2……xN).There is no change in the relation on interchange of indices
i and j .So gij is symmetric
gij = gji
Let gij represents a symmetric co-variant tensor such that gij together with its
partial derivatives are continuous.
Let g = |gij| ≠ 0
𝑔11 𝑔12 … 𝑔1𝑁
𝑔 𝑔₂₂ …
𝑔𝑖𝑗 = [ 21 ⋮ 𝑔2𝑁 ]
⋮ ⋮
𝑔𝑁1 𝑔𝑁2 … 𝑔𝑁𝑁
|gij | = g11G11+g12G12+…..g1NG1N
=g
𝐺 𝑖𝑗
Where Gij is the co-factor of the term gij. Let us define 𝑔𝑖𝑗 =
𝑔
gij is called the reciprocal metric tensor or the tensor conjugate to gij . It is a
symmetric contra variant tensor of rank 2.
For N=3
𝐺 𝑖𝑗 𝐺 𝑘𝑗
i.e. 𝑔𝑖𝑗 = 1 𝑎𝑛𝑑 𝑔𝑖𝑗 =0
𝑔 𝑔
|gij|= r7sin2θ
𝑟 4 𝑠𝑖𝑛2 𝜃
0 0
𝑟 4 𝑠𝑖𝑛2 𝜃
𝑟 2 𝑠𝑖𝑛2 𝜃
𝑔𝑖𝑗 = 0 0
𝑟 4 𝑠𝑖𝑛2 𝜃
𝑟2
[ 0 0
𝑟 4 𝑠𝑖𝑛2 𝜃]
1 0 0
1
= [0 𝑟2
0 ]
1
0 0
𝑟 2 𝑠𝑖𝑛2 𝜃
The fundamental relation for the line element is ds2=gij dxi dxj.It shows gij is a co-
variant tensor of rank two. Calculate the inner product of gij with another tensor.
By quotient law gij gj must be invariant. But gij is a co-variant tensor rank of two.
Again by quotient law gij is a contra variant tensor of rank two.
Associated tensor
A tensor obtained from an arbitrary tensor by forming an inner product with the
given tensor and any one of the first metric or reciprocal metric tensor is called an
associated tensor of the given tensor.
Ai gij ⟶Aj
Ai gij ⟶Aj
𝐴𝑗𝑖 𝑔 𝑗𝑘 ⟶ 𝐴𝑖𝑘
Taking the inner product of a tensor with first metric tensor lowers one of its
indices converts one of its contra variant indexes to co-variant. Similarly taking the
inner product of a tensor with reciprocal metric tensor raises its index i.e. converts
one of its co-variant index to contra variant
𝑗
𝐴𝑘𝑗 𝑔𝑘𝑖 ⟶ 𝐴𝑖
𝑖𝑗 𝑗
𝐴𝑘 𝑔𝑖𝑙 ⟶ 𝐴𝑘𝑙
𝑖𝑗 𝑖𝑗𝑚
𝐴𝑘𝑙 𝑔𝑙𝑚 ⟶ 𝐴𝑘
Qn) Show that in a Cartesian co-ordinate system, the contra variant and the
co-variant components of a vector are identical
Ai and Ai be the contra variant and co-variant components of a vector with respect
to the Cartesian co-ordinate
Ai=Aj gij
For Cartesian co-ordinate system gij (no summation over i) and gij=0 if i≠j
Ai= Ai gij = Ai , i.e. there is no distinction between co-variant 2 contra variant
components.
Tensor calculus
𝑖 𝜕𝑥 𝑖
𝑥,𝛼 =
𝜕𝑥̅ 𝛼
𝜕𝑥̅ 𝛼
𝑥̅,𝑖𝛼 ⟶
𝜕𝑥 𝑖
𝑖 𝛼 𝜕𝑥 𝑖 𝜕𝑥̅ 𝛼
𝑥,𝛼 𝑥̅,𝑗 = . = 𝛿𝑗𝑖
𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝑖 𝜕²𝑥 𝑖
𝑥,𝛼𝛽 =
𝜕𝑥̅ 𝛼 𝜕𝑥̅ 𝛽
𝑗 𝜕𝑥̅ 𝛼 𝜕𝑥 𝑗
𝛼
𝑥̅,𝑖𝑗 𝑥,𝛽 = . = 𝛿𝛽𝛼
𝜕𝑥 𝑗 𝜕𝑥̅ 𝛽
𝜕𝑔𝑖𝑗
𝑔𝑖𝑗,𝑘 =
𝜕𝑥 𝑘
𝑖𝑗 𝜕𝑔𝑖𝑗
𝑔,𝑘 =
𝜕𝑥 𝑘
Christoffel symbols
= [k, ij]
1 𝜕𝑔𝑖𝑘 𝜕𝑔𝑗𝑘 𝜕𝑔𝑖𝑗
[𝑖𝑗, 𝑘] = [ + − ]
2 𝜕𝑥 𝑗 𝜕𝑥 𝑖 𝜕𝑥 𝑘
Properties
= 𝛿𝑘𝑚 [𝑖𝑗, 𝑚]
Expanding
Multiplication of the first symbol of the reciprocal tensor converts it into the
symbol of second kind. Similarly multiplication of second kind symbol by first
metric tensor converts it into a symbol of first kind.
𝜕𝑔𝑖𝑗
4. = 𝑔𝑖𝑗,𝑘 = [𝑖𝑘, 𝑗] + [𝑖𝑘, 𝑖]
𝜕𝑥𝑘
R.H.S = [i k, j] + [j k, i]
1 1
= [𝑔𝑖𝑗,𝑘 + 𝑔𝑘𝑗,𝑖 − 𝑔𝑖𝑘,𝑗 ] + [𝑔𝑗𝑖,𝑘 + 𝑔𝑘𝑖,𝑗 − 𝑔𝑗𝑘,𝑖 ]
2 2
𝜕𝑔𝑖𝑗
𝑔𝑖𝑗,𝑘 =
𝜕𝑥 𝑘
𝑖𝑚 𝜕𝑔𝑖𝑚 𝑚 𝑖
5. 𝑔𝑙𝑚 = = −𝑔𝑖𝑗 [{𝑗 𝑙 }] − 𝑔𝑚𝑘 [{ }]
𝜕𝑥 𝑙 𝑘𝑙
We have 𝑔𝑖𝑗 𝑔𝑗𝑘 = 𝛿𝑘𝑖 ⟶ (1)
Differentiating w: r: to xl
𝜕𝑔𝑖𝑗 𝜕𝑔𝑗𝑘
𝑔𝑗𝑘 + 𝑔𝑖𝑗 =0
𝜕𝑥 𝑙 𝜕𝑥 𝑙
𝜕𝑔𝑖𝑗 𝜕𝑔𝑗𝑘
𝑔𝑗𝑘 = −𝑔𝑖𝑗
𝜕𝑥 𝑙 𝜕𝑥 𝑙
Multiplying gkm
𝜕𝑔𝑖𝑗 𝜕𝑔𝑗𝑘
𝑔𝑘𝑚 𝑔𝑗𝑘 = −𝑔𝑘𝑚 𝑔𝑖𝑗
𝜕𝑥 𝑙 𝜕𝑥 𝑙
𝜕𝑔𝑖𝑗
𝛿𝑗𝑚 = −𝑔𝑘𝑚 𝑔𝑖𝑗 [𝑔𝑗𝑘,𝑙 ]
𝜕𝑥 𝑙
On expanding
𝜕𝑔𝑖𝑚
= −𝑔𝑘𝑚 𝑔𝑖𝑗 [[𝑗𝑙, 𝑘] + [𝑘𝑙, 𝑗]]
𝜕𝑥 𝑙
The symbols are [11,1] , [11,2] , [12,1] , [12,2] , [21,1] , [21,2] , [22,1] , [22,2]
x1 = θ, x2 = ϕ
1 𝜕𝑔11 𝜕𝑔11 𝜕𝑔11
[11,1] = [ + − ]
2 𝜕𝜃 𝜕𝜃 𝜕𝜃
1 𝜕𝑎2
= =0
2 𝜕𝜃
1 𝑎2
= =0
2 𝜕∅
1 −𝜕𝑎2
= − =0
2 𝜕∅
1 𝜕
= (𝑎2 𝑠𝑖𝑛2 𝜃) = 𝑎2 𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃
2 𝜕𝜃
[21, 2] = a2sinθcosθ
[21, 1] = 0
1 𝜕𝑔21 𝜕𝑔21 𝜕𝑔22
[22,1] = [ + − ]
2 𝜕𝜙 𝜕𝜙 𝜕𝜃
1 𝜕
= ×− 𝑎2 𝑠𝑖𝑛2 𝜃
2 𝜕𝜃
= - a2sinθcosθ
1 𝜕𝑔22 𝜕𝑔22 𝜕𝑔22
[22,2] = [ + − ]
2 𝜕𝜙 𝜕𝜙 𝜕∅
1 𝜕𝑔22
=
2 𝜕∅
1 𝜕
= 𝑎2 𝑠𝑖𝑛2 𝜃 = 0
2 𝜕∅
Non- zero Christoffel symbols are [12, 2], [21, 2] and [22, 1]
=0
1
Г12 = 𝑔1𝑖 [12, 𝑖]
=0
2
Г11 = 𝑔21 [11, 𝑖]
=0+0=0
2
Г12 = 𝑔2𝑖 [12, 𝑖]
1 ⟶ r, 2 ⟶ θ, 3 ⟶ ϕ
=0
Г123 = 𝑔1𝑖 [2 3, 𝑖]
1
= . 𝑟 2 2𝑠𝑖𝑛𝜃𝑐𝑜𝑠𝜃 ⟶ 𝑔13 = 0
2
Г123 = 0
Qn) Show that Christoffel symbols are not tensors .How does Christoffel
symbols of the first and the second kinds transform under co-ordinate
transformation?
Consider the first metric tensor gij. It is the second order co-variant tensor.
𝑖 𝑖 𝑗 𝑖 𝑗 𝑗
= [𝑥,2𝛾 𝑥,𝛽 + 𝑥,𝛼 𝑥,𝛽𝛾 ] 𝑔𝑖𝑗 + 𝑥,𝛼 𝑥,𝛽 𝑥,𝛾𝑘 𝑔𝑖𝑗,𝑘 ⟶ (2)
𝜕 𝜕𝑥 𝑘
[Since 𝑔𝑖𝑗,𝛾 = 𝑔𝑖𝑗 = 𝑥,𝛾𝑘 𝑔𝑖𝑗,𝑘 ]
𝜕𝑥 𝑘 𝜕𝑥̅ 𝑘
𝑖 𝑗 𝑗 𝑗
𝑔̅𝛾𝛼,𝛽 = [𝑥,𝛾𝛽 𝑥,𝛼 + 𝑥,𝛾𝑖 𝑥,𝛼𝛽 ] 𝑔𝑖𝑗 + 𝑥,𝛾𝑖 𝑥,𝛼 𝑥𝛽𝑘 𝑔𝑖𝑗,𝛾 𝑘 ⟶ (4)
1
̅̅̅̅̅̅̅
[𝛼𝛽, 𝛾 ] = [𝑔̅𝛼𝛾,𝛽 + 𝑔̅𝛽𝛾,𝛼 − 𝑔̅𝛼𝛽,𝛾 ]
2
1
= [(4) + (3) − (2)]
2
Since gij is a symmetric tensor .the indices i and j could be interchanged in the
coefficient of gij on the RHS of eqn (2).In the last term on RHS we perform the
index changes i ⟶ j⟶ k⟶ i.
1
[ ̅̅̅̅̅̅̅
𝛼𝛽, 𝛾 ] = [𝑔̅𝛽𝛾,𝛼 + 𝑔̅𝛼𝛾,𝛽 − 𝑔̅𝛼𝛽,𝛾 ]
2
𝑖 𝑗 𝑖 1 𝑗
= 𝑥,𝛼𝛽 𝑥,𝛾 𝑔𝑖𝑗 + 𝑥,𝛼 𝑥𝛽 𝑥𝛾𝑘 [𝑔𝑘𝑖,𝑗 + 𝑔𝑗𝑘,𝑖 − 𝑔𝑖𝑗,𝑘 ]
2
𝑖 𝑗 𝑖 𝑗
= 𝑥,𝛼𝛽 𝑥,𝛾 𝑔𝑖𝑗 + 𝑥,𝛼 𝑥𝛽 𝑥𝛾𝑘 [𝑖𝑗, 𝑘] ⟶ (5)
This is the law of transformation of the christoffel symbols of the first kind from
𝑖
one co-ordinate system to another. The 𝑥,𝛼𝛽 term makes the expression non
tensorial. Hence christoffel symbol is not a tensor.
̅𝜌𝛼𝛽 = 𝑔̅ 𝛾𝜌 [𝛼𝛽,
Г ̅̅̅̅̅̅̅
𝛾]
𝑖 𝑗 𝑗
= 𝑥,𝛼𝛽 𝑥,𝛾 𝑔̅ 𝛾𝜌 𝑔𝑖𝑗 + 𝑥,𝛼
𝑖
𝑥𝛽 𝑥𝛾𝑘 𝑔̅ 𝛾𝜌
𝑖 𝑗 𝛾 𝜌 𝑗 𝛾
= 𝑥,𝛼𝛽 𝑥,𝛾 𝑥̅,𝑘 𝑥̅𝑙 𝑔𝑘𝑙 𝑔𝑖𝑗 + 𝑥,𝛼
𝑖
𝑥𝛽 𝑥𝛾𝑘 𝑥̅,𝑘 𝑔𝑘𝑙 [𝑖𝑗, 𝑘]
𝑗 𝜌 𝑗 𝜌
𝑖
= 𝑥,𝛼𝛽 𝛿𝑘 ̅𝑥,𝑙 𝑔𝑘𝑙 𝑔𝑖𝑗 + 𝑥,𝛼
𝑖
𝑥,𝛽 𝑥̅𝑙 Г𝑙𝑖𝑗
𝑖 𝜌 𝑗 𝜌
= 𝑥,𝛼𝛽 𝑥̅,𝑙 𝛿𝑖𝑙 + 𝑥,𝛼
𝑖
𝑥,𝑙 Г𝑖𝑗
𝑥,𝛽 ̅̅̅
𝜌 𝑗 𝜌
𝑖
= 𝑥,𝛼𝛽 𝑖
𝑥̅,𝑖 + 𝑥,𝛼 𝑥𝛽 ̅𝑥𝑙 Г𝑙𝑖𝑗
This gives the law of transformation of the christoffel symbols of the second kind.
Due to the presence of the first term in R.H.S, this is not a tensor.
𝝏
Qn) Show that г𝒊𝒊𝒋 = (𝐥𝐧 √𝒈), 𝒋 = (𝒍𝒏√𝒈)
𝝏𝒙𝒋
g = | gij |
1 1 𝜕𝑔
=
√𝑔 2√𝑔 𝜕𝑥 𝑗
1 𝜕𝑔
=
2𝑔 𝜕𝑥 𝑗
= gik Gik
𝜕𝑔
= 𝐺 𝑖𝑘
𝜕𝑔𝑖𝑘
𝜕𝑔 𝜕𝑔 𝜕𝑔𝑖𝑘
=
𝜕𝑥 𝑗 𝜕𝑔𝑖𝑘 𝜕𝑥 𝑗
𝜕𝑔𝑖𝑘
= 𝐺 𝑖𝑘
𝜕𝑥 𝑗
𝜕𝑔
= 𝑔𝑔𝑖𝑘 [𝑔𝑖𝑘 , 𝑗]
𝜕𝑥 𝑗
= 𝑔[Г𝑖𝑖𝑗 + Г𝑘𝑘𝑗 ]
1 𝜕𝑔
Г𝑖𝑖𝑗 = = (ln √𝑔)
2𝑔 𝜕𝑥 𝑗 ,𝑗
Covariant derivative
𝐴̅𝛼,𝛽 = 𝑥𝛼𝛽
𝑖 𝑖
𝐴𝑖 + 𝑥,𝛼 𝐴𝑖,𝛽 ⟶ (2)
𝑗
𝐴̅𝛼,𝛽 = 𝑥𝛼𝛽
𝑖 𝑖
𝐴𝑖 + 𝑥,𝛼 𝑥,𝛽 𝐴𝑖,𝑗 ⟶ (3)
𝑖
To evaluate 𝑥,𝛼𝛽 consider the transformation of Christoffel symbol of second kind
𝜌 𝜌 𝑗 𝜌
̅ 𝑖
Г𝛼𝛽 = 𝑥,𝛼𝛽 𝑖
𝑥̅𝑖 + 𝑥,𝛼 𝑥,𝛽 𝑥̅,𝑙 Г𝑙𝑖𝑗
𝑘
Taking inner product with 𝑥,𝜌
𝑘̅ 𝜌 𝑘 𝑖 𝜌 𝑘 𝑖 𝑗
𝑥,𝜌 Г𝛼𝛽 = 𝑥,𝜌 𝑥,𝛼𝛽 𝑥̅𝑖 + 𝑥,𝜌 𝑥,𝛼 𝑥,𝛽
𝑗
𝑘
= 𝑥,𝛼𝛽 𝑖
𝑥,𝛼 𝑥𝛽 𝛿,𝑙𝑘 Г𝑙,𝑖𝑗
𝑗
𝑘
= 𝑥,𝛼𝛽 𝑖
𝑥,𝛼 𝑥𝛽 Г𝑘,𝑖𝑗
𝑘 ̅𝜌 𝑗
𝑘
𝑥,𝛼𝛽 = 𝑥,𝜌 𝑖
Г𝛼𝛽 − 𝑥,𝛼 𝑥,𝛽 Г𝑘,𝑖𝑗 ⟶ (4)
̅𝜌𝛼𝛽 𝐴𝑖 − 𝑥,𝛼
𝐴̅𝛼,𝛽 = 𝑥𝜌𝑖 Г 𝑖 𝑗 𝑖 𝑗
𝑥,𝛽 𝐴𝑘 Г𝑘𝑖𝑗 + 𝑥,𝛼 𝑥,𝛽 𝐴𝑖,𝑗
𝜌 𝑗
𝐴̅𝛼,𝛽 − 𝐴𝜌̅ Г = 𝑥,𝛼
𝑖
𝑥,𝛽 [𝐴𝑖,𝑗 − 𝐴𝑘 Г𝑘𝑖𝑗 ]
𝛼𝛽
𝑖
𝐴𝑖; 𝑗 = 𝐴𝑖,𝑗 + 𝐴𝑘 Г𝑗𝑘
𝑖𝑗 𝑖𝑗 𝑗
𝐴 ;𝑘 = 𝐴,𝑘 + 𝐴𝑙𝑗 Г𝑖𝑙𝑘 + 𝐴𝑖𝑙 Г𝑙𝑘
𝑖
𝐴𝑗,𝑘 𝑖
= 𝐴𝑗,𝑘 + 𝐴𝑗𝑙 Г𝑖𝑙𝑘 − 𝐴𝑖𝑙 Г𝑗𝑘
𝑙
Qn) Show that the covariant derivative of the
a) metric tensor
b) Kronecker delta are identically zero
𝑙
1) 𝑔𝑖𝑗,𝑘 = 𝑔𝑖𝑗,𝑘 − 𝑔𝑖𝑙 Г𝑗𝑘 − 𝑔𝑙𝑗 Г𝑙𝑖𝑘
Hence RHS = 0
Ai: jk = (Ai;j);k
𝜕
=
𝜕𝑥 𝑘
[𝐴𝑖,𝑗 − 𝐴𝑙 Г𝑙𝑖𝑗 ] − Гℎ𝑖𝑘 [𝐴ℎ,𝑗 − 𝐴𝑙 Г𝑙ℎ𝑗 ] − Г𝑗𝑘
ℎ
[𝐴𝑖,ℎ − 𝐴𝑙 Г𝑙𝑖ℎ ]
𝐴𝑖 ;𝑗𝑘 ≠ 𝐴𝑖,𝑘𝑗
𝐴𝑖 ;𝑘𝑗 = 𝐴𝑖,𝑘𝑗 − 𝐴𝑙 Г𝑙𝑖𝑘,𝑗 − 𝐴𝑙,𝑗 Г𝑙𝑖𝑘 − Гℎ𝑖𝑗 𝐴ℎ,𝑘 + Гℎ𝑖𝑗 𝐴𝑙 Г𝑙ℎ − Гℎ𝑘𝑗 𝐴𝑖,ℎ − Гℎ𝑘𝑗 𝐴𝑙 Г𝑙𝑖ℎ
⟶ (3)
1. Let l = i
𝑙 𝑖
𝑅𝑖.𝑗𝑘 ⟶ 𝑅𝑖.𝑗𝑘
𝑙
𝑅𝑖.𝑗𝑘 = Г𝑙𝑖𝑗,𝑘 − Г𝑙𝑖𝑘,𝑗 + Гℎ𝑖𝑗 Г𝑙ℎ𝑘 − Гℎ𝑖𝑘
𝑖
𝑅𝑖.𝑗𝑘 = Г𝑖𝑖𝑗,𝑘 − Г𝑖𝑖𝑘,𝑗 + Гℎ𝑖𝑗 Г𝑖ℎ𝑘 − Гℎ𝑖𝑘 Г𝑖ℎ𝑗
=0
2. Let l = j
𝑙 𝑗
𝑅𝑖.𝑗𝑘 = 𝑅𝑖.𝑗𝑘
𝑗 𝑗 𝑗 𝑗
= Г𝑖𝑗,𝑘 − Г𝑖𝑘,𝑗 + Гℎ𝑖𝑗 Гℎ𝑘 − Гℎ𝑖𝑘 Гℎ𝑗
𝑗
𝑅𝑖.𝑗𝑘 = 𝑅𝑖𝑘 is called Ricci tensor