Tensors: an informal definition

Properties of the transformation matrix

Recall: we showed that under a change of coordinates, a vector v can be written (in the new • A tensor is an object that represents a physical or mathematical quantity and transforms, under
coordinates) v ′ = Lv where ℓij = n′i · nj . changes of coordinates, according to certain transformation laws.

Question: • A tensor of rank r in 3-dimensional space has the properties that:

(a) Show using graphical arguments that

◦ It needs 3r components to be described fully (in m dimensions, it needs mr components).
◦ The components are divided into 3 groups (each with 3r−1 members), each of which is
ℓ211 + ℓ212 + ℓ213 = 1. associated with one of the three Cartesian axes

(b) Using the definition of ℓij , show that if the basis n′1 , n′3 , n′3 is orthogonal, then ◦ The components transform according to well-defined rules so that the concepts represented
by the tensors are invariant when you change coordinates.
ℓ11 ℓ21 + ℓ12 ℓ22 + ℓ13 ℓ23 = 0.
• We showed that v ′ = Lv for vectors, where L is the transformation matrix (direction cosines).
(c) Is there something special about i = 1 in (a) or (b)?
• DEFINITION: Any quantity u with the property u′ = Lu under a change of coordinates is a
(d) Hence show that L is orthogonal. (Recall an orthogonal matrix is a matrix whose transpose is it’s tensor of the first order or a tensor of rank 1 (ie a vector).
inverse).
• Note: yes, this is an obscure definition ... more precise later.

10 11

Example: tensor of rank 0

Recap: a vector (rank 1)

• Any scalar can be a tensor of rank zero

• Any vector is a tensor of rank 1. It has 31 = 3 components
• Example: the boiling point of water (at a pressure of 1 atmosphere)
• Components come in 3 groups of 31−0 = 1, each associated with a Cartesian coordinate axis
◦ 100◦ C
(the component in that direction)
◦ 212◦ F
• The vector transforms according to the transformation rule:
◦ 373.15◦ K
◦ to transform the rank one tensor v from (x1 , x2 , x3 ) coordinates with basis (n1 , n2 , n3 ), to
◦ 80◦ Réaumur (!) (in which 0 is the freezing point)
v ′ in (x′1 , x′2 , x′3 ) coordinates with basis (n′1 , n′2 , n′3 )
• Note the difference between the boiling point (the tensor) and its different values in different ◦ v ′ = Lv , where L = [ℓij ] is the matrix of direction cosines ℓij = n′i · nj
coordinate systems (the temperature scale)
• e.g. the vector v = (1, 2, 3)T is a tensor of rank 1, in the standard Cartesian coordinates
• Obviously which scale we choose cannot affect the actual boiling point of water: invariance is
(x, y, z) with basis vectors (i, j, k)
important.

12 13
 Example: tensor of rank 2
Example: tensor of rank 2 cont. . .

• Example: consider a rank one tensor v = (v1 , v2 , v3 ) in (x1 , x2 , x3 ) coordinates

• Consider the rank one tensor v = (1, 2, 3), and transformation given by rotation of 90◦ about
• This defines a tensor T with 32 = 9 components Tpq , where p = 1, 2, 3, q = 1, 2, 3 with
the x3 -axis (ie the example we had before).
Tpq = vp vq
• Then the tensors T = [vp vq ] and T ′ = [vp′ vq′ ] are given by
• In new coordinates, the tensor becomes    
′
1 2 3 4 −2 6
Tpq = vp′ vq′    
T =
 2 4 6
 T′ = 
−2 1 −3

• We know that the rank 1 tensor transforms according to 3 6 9 6 −3 9
3
X 3
X
vp′ = vq′ =
P3 P3
ℓpi vi , ℓqj vj ′
• Alternatively, transform according to the rule Tpq = i=1 j=1 ℓpi ℓqj Tij
i=1 j=1
′ T
• In practice, we can show that T = LT L
• So
3
! 3 
3
3 X 3
3 X    T  
′
Tpq =
X
ℓpi vi 
X
ℓqj vj  =
X
ℓpi ℓqj vi vj =
X
ℓpi ℓqj Tij 0 1 0
1 2 3 0 1 0 4 −2 6
     
i=1 j=1 i=1 j=1 i=1 j=1 T′ = 
−1 0 0  2 4
  −1 0 0 = −2 1 −3
6    

• DEFINITION: any set of quantities Tpq that transforms according to 0 0 1 3 6 9 0 0 1 6 −3 9

′
P3 P3
Tpq = i=1 j=1 ℓpi ℓqj Tij under a change of coordinates is a tensor of rank 2.

14 15

Example: another rank 2 tensor cont. . .

Example: another rank 2 tensor

• The stress tensor σ is defined to be

• The most important physical tensor of rank 2 is the stress tensor  
σ11 σ12 σ13
• When an elastic body (or fluid) is deformed, internal forces are set up. At any particular point P  
in such a body, consider forces acting on a plane A containing P σ=σ
 21 σ22 σ23 

σ31 σ32 σ33
• Stress is force per unit area: let area of A be 1
• Look at forces acting on A • Using tensor calculus, we can find a rotation L to transform the stress tensor to new coordinates
◦ force is a vector: has three components with respect to axes n1 , n2 , n3 x′1 , x′2 , x′3 , so that
 
′
◦ assume n1 is perpendicular to A σ11 0 0
′
 
◦ label the three components σ11 , σ12 , σ13 ; σ =
 0
′
σ22 0 
′
0 0 σ33
• Similarly σpq is the stress acting in the nq direction, on a plane ⊥ to np
′ ′ ′
• Rotation gives the planes in which failure may occur, and the principal stresses σ11 , σ22 , σ33

16 17