Cartesian Tensors

Reference: H. Jeffreys Cartesian Tensors

Cartesian tensors 1/35

1 Coordinates and Vectors

z = x3
e3
y = x2
e2

e1

x = x1

Cartesian tensors 2/35

Coordinates
xi i = 1, 2, 3

x1 = x x2 = y x3 = z (1)

Unit vectors
ei i = 1, 2, 3

e1 = e x = i e2 = e y = j e3 = ez = k (2)

General vector (formal definition to follow) denoted by compo-

nents e.g. u = u i

Cartesian tensors 3/35

Notation
The boldface notation for vectors is referred to as dyadic nota-
tion
The subscript notation is tensor notation.
Summation convention
Einstein: repeated index means summation:
3
ui vi = ∑ ui vi
i=1
(3)
3
u ii = ∑ u ii
i=1

Cartesian tensors 4/35

2 Orthogonal Transformations of Coordinates

The behaviour of quantities under orthogonal transformations of

the coordinate system is the basis of Cartesian tensors.
We want to formulate equations in such a way that they are in-
dependent of the specific coordinate system.

Cartesian tensors 5/35

 x3
x 3′ x 2′
x2
x 1′

x1

General linear transformation

x i′ = a ij x j
a ij = Transformation Matrix

Cartesian tensors 6/35

Position vector
Consider the position vector expressed in terms of coordinates
and unit vectors in two related coordinate systems
r = x i e i = x i ′e i ′ (4)

In view of the transformation from the unprimed to the primed

system:
r = a ij x j e i ′ = x j ( a ij e i ′ ) (5)

Therefore we can write:

e j = a ij e i ′ (6)

Cartesian tensors 7/35

so that we have the two companion transformations:
x i ′ = a ij x j e i = a ji e j ′ (7)

Kronecker Delta
δ ij = 1 if i = j
= 0 otherwise
In matrix form

1 0 0
δ ij = 0 1 0 (8)
0 0 1

Cartesian tensors 8/35

Substitution property
δ ij T jk... = T ik... (9)

In the summation over j the only term of the sum that makes any
contribution is that for which j = i .

2.1 Orthogonal transformation

So far, what we have described is valid for any linear transfor-
mation.
Now impose the condition that both the original and the primed
reference frames are orthonormal
e i ⋅ e j = δ ij and e i′ ⋅ e ′j = δ ij (10)

Cartesian tensors 9/35

Use transformation of the unit vectors:
e i ⋅ e j = a ki e k′ ⋅ a lj e l′

= a ki a lj e k′ ⋅ e l′
(11)
= a ki a lj δ kl
= a ki a kj
NB the last operation is an example of the substitution property
of the Kronecker Delta.
Since e i ⋅ e j = δ ij , then the orthonormal condition on a ij is

a ki a kj = δ ij (12)

Cartesian tensors 10/35

In matrix notation:
aT a = I (13)
We also have as a consequence of the properties of matrices, that
aa T = I (14)
In tensor notation:
( aa T ) ij = a ik a jk = δ ij (15)

Any of equations (12), (13), (14) or (15) defines an orthogonal

transformation.

Cartesian tensors 11/35

2.2 Reverse transformations
x i′ = a ij x j ⇒ a ik x i′ = a ik a ij x j = δ kj x j = x k

∴x k = a ik x i′ ⇒ x i = a ji x ′j
i.e. the reverse transformation is simply determined by the trans-
pose.
Similarly, following from
e j = a ij e i ′ (16)

we have
e i′ = a ij e j (17)

Cartesian tensors 12/35

Transformations for both coordinates and basis vec-
tors
The complementary set of transformations is then
x i ′ = a ij x j e i ′ = a ij e j (18)

2.3 Interpretation of the matrix a ij

Since
e i′ = a ij e j

then the a ij are the components of e i′ wrt the unit vectors in the
original system.e.g.
e 1 ′ = a 11 e 1 + a 12 e 2 + a 13 e 3 (19)

Cartesian tensors 13/35

2.4 Example: 2D rotation
It is easiest here to determine the
y relationship between the unit ba-
y′ e ′ sis vectors:
y ey
x′
ex′ e x ′ = cos θ e x + sin θe y
(20)
e y ′ = – sin θ e x + cos θe y
θ
x In matrix form:
ex
ex′ ex
cos θ sin θ 0
e y ′ = – sin θ cos θ 0 e y (21)
ez ′ 0 0 1 e
z

Cartesian tensors 14/35

Then the transformation equation for the coordinates is:

x′ cos θ sin θ 0 x
y′ = – sin θ cos θ 0 y (22)
z′ 0 0 1 z

3 Scalars, Vectors & Tensors

We define these objects by the way in which they transform with

respect to orthogonal coordinate transformations.

3.1 Scalar (f):

f ( x ′i ) = f ( xi ) (23)

Cartesian tensors 15/35

Example of a scalar is f = r 2 = x i x i . Examples from fluid dy-
namics are the density and temperature.

3.2 Vector (u):

Prototype vector: x i
General transformation law:
x i′ = a ij x j ⇒ u i′ = a ij u j (24)

as the transformation law for a generic vector.

Cartesian tensors 16/35

3.3 Gradient operator
Suppose that f is a scalar. The gradient of f is defined by
∂f
( grad f ) i = ( ∇f ) i = (25)
∂ xi
Need to show this is a vector by its transformation properties.

∂f ∂ f ∂x j
= (26)
∂ x i′ ∂ x j∂ x ′
i
Since,
x j = a kj x k′ (27)

Cartesian tensors 17/35

then
∂x j
= a kj δ ki = a ij
∂ x i′
(28)
∂f ∂f
and = a ij
∂ x i′ ∂x j

Hence the gradient operator satisfies our definition of a vector.

Scalar Product
u ⋅ v = ui vi = u1 v1 + u2 v2 + u3 v3 (29)

is the scalar product of the vectors u i and v i .

Cartesian tensors 18/35

Exercise:
Show that u ⋅ v is a scalar.

3.4 Tensor
Prototype second rank tensor x i x j
General definition by transformation of components:
′ = a a T
T ij (30)
ik jl kl
Exercise:
Show that u i v j is a second rank tensor if u i and v j are vectors.

Cartesian tensors 19/35

Exercise:
∂u i
u i, j =
∂x j
is a second rank tensor. (Introduces the comma notation for par-
tial derivatives.) In dyadic form this is written as grad u or ∇u .

3.5 Divergence
Exercise:
Show that the quantity
∂v i
∇ ⋅ v = div v = (31)
∂ xi

Cartesian tensors 20/35

is a scalar.

4 Products and Contractions of Tensors

It is easy to form higher order tensors by multiplication of lower

rank tensors, e.g. T ijk = T ij u k is a third rank tensor if T ij is a
second rank tensor and u k is a vector (first rank tensor). It is
straightforward to show that T ijk has the relevant transformation
properties.
Similarly, if T ijk is a third rank tensor, then T ijj is a vector.
Again the relevant transformation properties are easy to prove.

Cartesian tensors 21/35

5 Differentiation following the motion
This involves a common operator occurring in fluid dynamics.
Suppose the coordinates of an element of fluid are given as a
function of time by
xi = xi ( t ) (32)

vi

Cartesian tensors 22/35

The velocities of elements of fluid at all spatial locations within
a given region constitute a vector field, i.e. v i = v i ( x j, t )
If we follow the trajectory of an element of fluid, then on a par-
ticular trajectory x i = x i ( t ) . The acceleration of an element is
then given by:
dv i
d ∂v i ∂v i dx j ∂v i ∂v i
fi = = v i ( x j ( t ), t ) = + = + vj (33)
dt dt ∂t ∂ x jdt ∂t ∂x j

Exercise: Show that f i is a vector.

Cartesian tensors 23/35

6 The permutation tensor ε ijk

ε ijk = 0 if any of i, j, k are equal

= 1 if i, j, k unequal and in cyclic order (34)
= – 1 if i, j, k unequal and not in cyclic order
e.g.
ε 112 = 0 ε 123 = 1 ε 321 = – 1 (35)

Is ε ijk a tensor?

In order to show this we have to demonstrate that ε ijk , when de-

fined the same way in each coordinate system has the correct
transformation properties.
Cartesian tensors 24/35
Define
′ = ε
ε ijk lmn a il a jm a kn
= ε 123 a i1 a j2 a k3 + ε 312 a i3 a j1 a k2 + ε 231 a i2 a j3 a k1
+ ε 213 a i2 a j1 a k3 + ε 321 a i3 a j2 a k1 + ε 132 a i1 a j3 a k2
= a i1 ( a j2 a k3 – a j3 a k2 ) – a i2 ( a j1 a k3 – a j3 a k1 )
+ a i3 ( a j1 a k2 – a j2 a k1 )

a i1 a i2 a i3
= a j1 a j2 a j3
a k1 a k2 a k3

Cartesian tensors 25/35

In view of the interpretation of the a ij , the rows of this determi-
nant represent the components of the primed unit vectors in the
unprimed system. Hence:
′ = e′ ⋅ (e′ × e′ )
ε ijk i j k
This is zero if any 2 of i, j, k are equal, is +1 for a cyclic permu-
tation of unequal indices and -1 for a non-cyclic permutation of
′ . Thus ε
unequal indices. This is just the definition of ε ijk ijk
transforms as a tensor.

Cartesian tensors 26/35

6.1 Uses of the permutation tensor
Cross product
Define
c i = ε ijk a j b k (36)

then
c 1 = ε 123 a 2 b 3 + ε 132 a 3 b 2 = a 2 b 3 – a 3 b 2
c 2 = ε 231 a 3 b 1 + ε 213 a 1 b 3 = a 3 b 1 – a 1 b 3 (37)
c 3 = ε 312 a 1 b 2 + ε 321 a 2 b 1 = a 1 b 2 – a 2 b 1

These are the components of c = a × b .

Cartesian tensors 27/35

6.2 Triple Product
In dyadic notation the triple product of three vectors is:
t = u ⋅ (v × w) (38)
In tensor notation this is
t = u i ε ijk v j w k = ε ijk u i v j w k (39)

6.3 Curl
∂u k
( curl u ) i = ε ijk (40)
∂x j

Cartesian tensors 28/35

e.g.
∂u 3 ∂u 2 ∂u 3 ∂u 2
( curl u ) 1 = ε 123 + ε 132 = – (41)
∂ x2 ∂ x3 ∂ x2 ∂ x3
etc.

6.4 The tensor ε iks ε mps

Define
T ikmp = ε iks ε mps (42)

Properties:
• If i = k or m = p then T ikmp = 0 .
Cartesian tensors 29/35
• If i = m we only get a contribution from the terms s ≠ i and
k ≠ i, s . Consequently k = p . Thus ε iks = ± 1 and
ε mps = ε iks = ± 1 and the product ε iks ε iks = ( ± 1 ) 2 = 1 .
• If i = p , similar argument tells us that we must have s ≠ i and
k = m ≠ i . Hence, ε iks = ± 1 , ε mps = −+ 1 ⇒ ε iks ε mps = – 1 .
So,
i = m, k = p ⇒ 1 unless i = k ⇒ 0
i = p, k = m ⇒ – 1 unless i = k ⇒ 0
These are the components of the tensor δ im δ kp – δ ip δ km .

∴ε iks ε mps = δ im δ kp – δ ip δ km (43)

Cartesian tensors 30/35

6.5 Application of ε iks ε mps
∂
[ curl ( u × v ) ] i = ε ijk ( ε klm u l v m )
∂x j
∂
= ε ijk ε klm ( ul vm )
∂x j (44)

 ∂u l ∂v m
= ( δ il δ jm – δ im δ jl )  vm + ul 
∂x j ∂x j 

Cartesian tensors 31/35

We then use the substitution property of δ ij to show that:

∂u i ∂u j ∂v m ∂v i
[ curl ( u × v ) ] i = vm – vi + ui – uj
∂ xm ∂x j ∂ xm ∂x j

∂u i ∂v i ∂v j ∂u j (45)
= vj – uj + ui – vi
∂x j ∂x j ∂x j ∂x j
= ( v ⋅ ∇u – u ⋅ ∇v + u∇ ⋅ v – v∇ ⋅ u ) i
The Laplacean
2 2 2
∂ φ ∂ φ ∂ φ ∂ 2φ
∇2φ = + + = ---------------- (46)
∂ x 12 ∂ x 22 ∂ x 32 ∂x i ∂x i

Cartesian tensors 32/35

7 Tensor Integrals
7.1 Green’s Theorem
v In dyadic form:
ni
∫ ( ∇ ⋅ v ) dV = ∫ ( v ⋅ n ) dS (47)
V V S
S In tensor form:
∂v i
∫ ∂ xi dV = ∫ v i n i dS = S
(48)
V S
= Flux of v through S

Cartesian tensors 33/35

Extend this to tensors:
∂T ij
∫ ∂ x j dV = ∫ T ij n j dS (49)
V S
= Flux of T ij through S

Cartesian tensors 34/35

7.2 Stoke’s Theorem
ti C In dyadic form:

S ∫ ( curl u ) ⋅ n dS = ∫ u ⋅ t ds (50)
S C
In tensor form:
∂u k
∫ ε ijk
∂x j
n i dS = ∫ u i t i ds (51)
S C

Cartesian tensors 35/35