Lecture 3 of Tensors: Tensor of rank 2

Tensor of rank r

• A general tensor T of rank 2 has 9 components Tpq

• We can generalise these ideas to generate a tensor of any rank we choose
• Under a transformation to new coordinates (x′1 , x′2 , x′3 ) defined by the rotation matrix L = [ℓij ]
• For example: a tensor T of rank 4 has 34 = 81 entries Tpqrs
ℓij = n′i · nj
• Under a change of coordinates (x1 , x2 , x3 ) to (x′1 , x′2 , x′3 ), with corresponding change in
the rank 2 tensor transforms according to the rule basis (n1 , n2 , n3 ) to (n′1 , n′2 , n′3 ), defined by the rotation matrix L = [ℓij ] where
3 X
X 3 ℓij = n′i · nj , the tensor transforms according to the rule
′
Tpq = ℓpi ℓqj Tij
3 X
3 X 3 X
3
i=1 j=1 ′
X
Tpqrs = ℓpi ℓqj ℓrk ℓsl Tijkl
• In matrix form, we can alternatively write i=1 j=1 k=1 l=1

T ′ = LT LT • etc.

18 19

The Kronecker delta cont. . .

Basic tensors: the Kronecker delta

• The Kronecker delta has many special properties:

• The identity matrix is the simplest rotation ◦ it is symmetric: δij = δji for all i, j ∈ {1, 2, 3}
3 3
 
1 0 0 X X
  ◦ δij Ajk = Aik , δij Akj = Aki for any tensor A of rank 2
I=
0 1 0 j=1 j=1

0 0 1 ◦ the only tensor of rank 2 which is isotropic (same values in all coordinate systems) is

• In tensor notation, I has components δij where pδij where p is a constant scalar

′
 ◦ the Kronecker delta is hence itself an isotropic tensor: δij = δij because
1 if i = j
δij = δ ′ = LδLT = LLT = I .
0 if i 6= j
◦ orthogonality can be expressed in terms of the Kronecker delta

• This defines a tensor of rank 2! You can show this by showing that the gradient of a vector is a ni · nj = δij
∂u ∂x
tensor in general (Uij = ∂x i ) and then writing δ as δ = ∂x i as a gradient.
j j
◦ vector calculus rules too
• δij is called the Kronecker delta ∂xi
= δij
∂xj

20 21
 Basic tensors: the alternating tensor The alternating tensor cont. . .

• An permutation of a list of numbers (a1 a2 . . . aN ) is just a reordering of the N numbers • Properties of the alternating tensor
• A permutation is even if we can exchange pairs of numbers an even number of times to get the ◦ ǫijk is a tensor of rank 3, with 33 = 27 components
permutation (similarly odd permutations) ◦ ǫijk is antisymmetric: swapping a pair of indices changes the sign
• So (2 3 1) is an even permutation of (1 2 3), while (3 2 1) is an odd permutation of (1 2 3) ǫijk = −ǫjik = −ǫikj = −ǫkji
(1 2 3) 7→ (2 1 3) 7→ (2 3 1) (1 2 3) 7→ (3 2 1) ǫijk = ǫjki = ǫkij

• The alternating tensor (also called the Lévi-Civita tensor ) is defined to be ◦ ǫijk is isotropic
 ǫ′ijk = ǫijk
0 if 2 (or 3) of i, j and k are equal

◦ the only isotropic tensor of order 3 is Kǫijk , for a constant scalar K

ǫijk = 1 if (i j k) is an even permutation of (1 2 3)
• Note that the only isotropic tensor of order 4 is, for constant scalars α, β , γ


−1 if (i j k) is an odd permutation of (1 2 3)


µijkl = αδij δkl + βδik δjl + γδil δjk

• For example
ǫ112 = 0, ǫ123 = 1, ǫ231 = 1, ǫ321 = −1, etc. (see later for significance. . . )

22 23

The Einstein summation convention More on the summation convention

• Tensor notation rapidly gets out of hand; the summation convention makes life easier. • The summation convention makes life much easier

• Two rules • Can express some familiar concepts neatly and simply (and bring in δij and ǫijk naturally)
1. Instead of writing e.g. aij = bij for all i, j ∈ {1, 2, 3}, if a suffix occurs once (and only • For example
once) on each side of an equation assume that it ranges over all possible choices (namely ai = δij aj
1, 2, 3). Called a free suffix. e.g. just write aij = bij
P • Vector products are amenable to the summation convention: if a and b have components
2. Instead of writing the sign, if a suffix occurs twice (and only twice) in an expression then
(a1 , a2 , a3 ) and (b1 , b2 , b3 ) respectively, then
sum over that suffix. Called a dummy suffix (doesn’t matter what we call it).
P3 ◦ Scalar product: a · b
For example, the rule for transforming tensors of rank 1: vi′ = j=1 ℓij vj , is written, using
the summation convention, as a · b = ai bi = δij aj bi = δij δil aj bl (sum over i, i, j , and i, j, l respectively)

vi′ = ℓij vj = ℓip vp = ℓim vm (j , p, m (etc.) are the dummy suffices) ◦ Vector product: the i-component of a × b is

• Expressions like api bqi cqi are meaningless with the summation convention [a × b]i = ǫijk aj bk (sum over j and k )

24 25