PBC lecture notes series: Compiled by Dr.

Abhijit Kar Gupta 1

Introduction to Tensor

Let us think of the coordinate transformation from the Rectangular Cartesian system
( x, y, z ) to the Curvilinear system ( u1 , u 2 , u 3 ):
( x, y, z ) → (u1 , u 2 , u 3 )
We may consider the following two sets of basis vectors:
∂r ∂r ∂r ∂r ∂r ∂r
eˆ1 = , eˆ2 = , eˆ3 = and
∂u1 ∂u1 ∂u 2 ∂u 2 ∂u 3 ∂u 3

Eˆ 1 = ∇u1 ∇u1 , Eˆ 2 = ∇u 2 ∇u 2 , Eˆ 3 = ∇u 3 ∇u 3 .
The above two sets of basis vectors are reciprocal to each other.

To prove:
r = r (u1 , u 2 , u 3 )
∂r ∂r ∂r
∴ dr = du1 + du 2 + du 3
∂u1 ∂u 2 ∂u 3
∂r ∂r ∂r
∴ ∇u1 ⋅ d r = du1 = (∇u1 ⋅ )du1 + (∇u1 ⋅ )du 2 + (∇u1 ⋅ )du 3
∂u1 ∂u 2 ∂u 3
We can thus write
∂r ∂r ∂r
∇u1 ⋅ = 1 , ∇u1 ⋅ = 0 , ∇u1 ⋅ = 0.
∂u1 ∂u 2 ∂u 3
∂r
In general, ⋅ ∇u n = δ mn (Kroenecker delta symbol), where δ mn = 0 , m ≠ n
∂u m
= 1, m = n

Home Work:
Take spherical polar coordinate system ( r ,θ , φ ) for example; verify the reciprocity.

Now for any vector, we can write

∂r ∂r ∂r
A = A1 + A2 + A3 = A1α 1 + A 2 α 2 + A3 α 3 , where A1 , A 2 , A 3 are
∂u1 ∂u 2 ∂u 3
contravariant coefficients.
3
In short, A = ∑ A k α k , where we drop the vector symbols for the sake of convenience.
k =1
We can also write,
PBC lecture notes series: Compiled by Dr. Abhijit Kar Gupta 2

A = A1 ∇u1 + A2 ∇u 2 + A3 ∇u 3 = A1 β1 + A2 β 2 + A3 β 3 , where A1 , A2 , A3 are covariant

coefficients.
3
In short , A = ∑ Al β l
l =1
Following Einstein’s summation convention over repeated index,
A = A k α k , k = 1, 2, 3 (Contravarinat components)
Also,
A = Al β l , l = 1, 2, 3 (Covariant components)
We have to remember that k and l are dummy indices and so in the further calculations
they may be replaced by any other when necessary.

Now we can have the coordinate transformations

( x, y, z ) → (u1 , u 2 , u 3 ) or may be to any other curvilinear system, ( x, y, z ) → (u1 , u 2 , u 3 ) .

Let us consider the coordinate transformation:

(u1 , u 2 , u 3 ) → (u1 , u 2 , u 3 )

We would like to see how the contravariant or covariant components of a vector are
related (as the transformation is carried out from one curvilinear system to another).

Contravariant components:

We have,
u m = u m (u1 , u 2 , u 3 ) ; m = 1, 2, 3.
∂u
∴ du m = m du n ; m, n = 1, 2, 3 (1)
∂u n
[Einstein’s summation convention is followed.]
Considering the position vector in the two systems, r = r (u1 , u 2 , u 3 ) and r = r (u1 , u 2 , u 3 ) ,
we can write,
∂r
dr = du i = α i du i and
∂u i
∂r
dr = du j = α j du j .
∂u j
Thus
α i du i = α j du j (2)
Putting (1) into (2),
∂u
α i i du n = α j du j
∂u n
PBC lecture notes series: Compiled by Dr. Abhijit Kar Gupta 3

Now in the LHS of the above, we can replace the index n by j as n is a dummy index
there.

∂u i
Therefore, we can write α j = α i (3)
∂u j
This is how the basis vectors transform.
To see how the contravariant components transform, consider
A = A k α k and also A = A l α l .
∴ Ak α k = A lα l
∂u k
= A lα k
∂u l

∂u k
∴ Ak = A l
∂u l

∂u k
Similarly, A k = A l . This is how the contravariant components transform.
∂u l

Covariant components:

Writing the vector A in terms of covariant components,

A = Al β l and also A = Am β m
∴ Al β l = Am β m
⇒ A1∇u1 + A2 ∇u 2 + A3 ∇u 3 = A1∇u1 + A2 ∇u 2 + A3 ∇u 3
Now,
∂u ∂u ∂u
A1∇u1 + A2 ∇u 2 + A3 ∇u 3 = ( A1 1 + A2 2 + A3 3 )iˆ +
∂x ∂x ∂x
∂u1 ∂u 2 ∂u 3 ˆ ∂u ∂u ∂u
( A1 + A2 + A3 ) j + ( A1 1 + A2 2 + A3 3 )kˆ
∂y ∂y ∂y ∂z ∂z ∂z
∂u ∂u ∂u
= Ak k iˆ + Al l ˆj + Am m kˆ , k , l , m =1, 2, 3 (4)
∂x ∂y ∂x
Again,
∂u ∂u ∂u
A1∇u1 + A2 ∇u 2 + A3 ∇u 3 = ( A1 1 + A2 2 + A3 3 )iˆ +
∂x ∂x ∂x
∂u ∂u ∂u ∂u ∂u ∂u
( A1 1 + A2 2 + A3 3 ) ˆj + ( A1 1 + A2 2 + A3 3 )kˆ
∂y ∂y ∂y ∂z ∂z ∂z
PBC lecture notes series: Compiled by Dr. Abhijit Kar Gupta 4

∂u l ˆ ∂u ∂u
= Al i + Am m ˆj + An n kˆ , l , m , n =1, 2, 3 (5)
∂x ∂y ∂x
Comparing (4) and (5),
∂u k ∂u
Ak = Al l
∂x ∂x
∂u k ∂u l
Ak = Al (I)
∂y ∂y
∂u ∂u
Ak k = Al l
∂z ∂z
Let us now consider,
∂u m
u m = u m (u1 , u 2 , u 3 ) ⇒ du m = du n
∂u n
∂u m ∂u m ∂u n
∴ =
∂x ∂u n ∂x
∂u m ∂u m ∂u n
= ( II )
∂y ∂u n ∂y
∂u m ∂u m ∂u n
=
∂z ∂u n ∂z
From (I) and (II) we can write
∂u ∂u
Ak k = Al l
∂x ∂x
∂u l ∂u m ∂u ∂u k
= Al = Al l
∂u m ∂x ∂u k ∂x
[Q m , k are dummy indices]

∴ ∂u l
Ak = Al
∂u k

∂u l
Similarly, Ak = Al . This is how the covariant components transform.
∂u k

The set of components, A k ’s or Al ’s which transform according to the rules

∂u k ∂u
Ak = A l or Ak = Al l , comprise a tensor of rank 1.
∂u l ∂u k
• Tensor of rank 1 = Vector.

Note:
PBC lecture notes series: Compiled by Dr. Abhijit Kar Gupta 5

For the contravariant components of a vector A and the covariant components of

another vector B , we have
A k Bk = A k Bk
The above is invariant under coordinate transformation and it is a scalar.
∂u ∂u ∂u ∂u l k
A k Bk = ( A k k )( Bk l ) = k A B k = A k Bk .
∂u l ∂u k ∂u l ∂u k

Home Work:
∂u k ∂u
From A k Bk = A k Bk , show that Bk = Bl when A k = A l k .
∂u l ∂u l

Tensor of Rank 2:

Consider the line element in Cartesian system,

ds 2 = dx 2 + dy 2 + dz 2
We can write the above as
ds 2 = g jk dx j dx k , where j , k =1, 2, 3
Here we assumed, x1 ≡ x , x 2 ≡ y and x 3 ≡ z .
Thus g jk ’s form a matrix and it is called metric tensor.

 g11 g 12 g 13   1 0 0 
   
g =  g 21 g 22 g 23  =  0 1 0 
g g 32 g 33   0 0 1 
 31

The metric tensor in spherical polar coordinate system ( r , θ , φ ):

1 0 0 
 
g = 0 r 2
0  [
Q ds 2 = dr 2 + r 2 dθ 2 + r 2 sin 2 θ .dφ 2 ]
 0 0 r 2 sin 2 θ 
 
In cylindrical coordinate system ( ρ , φ , z ):
1 0 0
 
g = 0 ρ 2 0 [
Q ds 2 = dρ 2 + ρ 2 dφ 2 + dz 2 ]
0 0 1
 

We know ds 2 , being a scalar, is invariant under coordinate transformation and the

metric tensor g is a symmetric matrix.
∴ We can write,
PBC lecture notes series: Compiled by Dr. Abhijit Kar Gupta 6

∂x m i ∂x n
g kl dx k dx l = g mn dx m dx n = g mn dx dx j
∂x i
∂x j

∂x m ∂x n i j
⇒ g kl dx k dx l = g mn ⋅ dx dx
∂x i ∂x j
If we replace i , j by k , l in the above (as they are dummy indices),
∂x m ∂x n
g kl = g mn ⋅
∂x i ∂x j
∴ g kl ’s are elements of a symmetric covariant tensor of rank 2.

• If the elements are transformed by two derivatives, it is a tensor of rank 2 and so on.
• If the elements are transformed by one derivative, it is a tensor of rank 1 (vector).
• Tensor or rank 0 is a scalar.