MAXWELL STRESS TENSOR

Link to: physicspages home page.

To leave a comment or report an error, please use the auxiliary blog.
References: Griffiths, David J. (2007), Introduction to Electrodynamics,
3rd Edition; Pearson Education - Problem 8.3.
Post date: 9 Jun 2014.
Here we’ll look at a purely classical, non-relativistic form of the tensor
in electromagnetism. In doing so, we’ll look only at the spatial components
of the tensor, so it becomes a 3 × 3 matrix.
The derivation starts with a calculation of the total force due to electro-
magnetic fields on the charges and currents within some volume V. From
the Lorentz force law, we have
ˆ
F = ρ (E + v × B) d3 r (1)
ˆV
= (ρE + J × B) d3 r (2)
V
We can think of the integrand as a force density, or force per unit volume
f:

f ≡ ρE + J × B (3)
We can express this entirely in terms of fields by using Maxwell’s equa-
tions:

ρ = 0 ∇ · E (4)
1 ∂E
J = ∇ × B − 0 (5)
µ0 ∂t
So we get
 
1 ∂E
f = (0 ∇ · E) E + ∇ × B − 0 ×B (6)
µ0 ∂t
We now need to do a bit of vector calculus gymnastics. From the product
rule

∂ ∂E ∂B
(E × B) = ×B+E× (7)
∂t ∂t ∂t
1
 MAXWELL STRESS TENSOR 2

and from Faraday’s law

∂B
= −∇ × E (8)
∂t

Combining these two we get

∂E ∂ ∂B
×B = (E × B) − E × (9)
∂t ∂t ∂t
∂
= (E × B) + E × (∇ × E) (10)
∂t
We can insert this into 6 and while we’re at it, we can add on a term
1
µ0 (∇ · B) B. This is always zero because ∇ · B = 0, but it gives the equation
a symmetry that will be useful in a minute. We get for the force density:

1 1 ∂E
f = 0 (∇ · E) E + (∇ · B) B + (∇ × B) × B − 0 ×B (11)
µ0 µ0 ∂t
1 1
= 0 (∇ · E) E + (∇ · B) B + (∇ × B) × B− (12)
µ0 µ0
∂
0 (E × B) − 0 E × (∇ × E) (13)
∂t
Now another identity from vector calculus says

∇ (A · B) = A × (∇ × B) + B × (∇ × A) + (A · ∇) B + (B · ∇) A (14)

If A = B = E, we get

∇ E 2 = 2E × (∇ × E) + 2 (E · ∇) E


(15)

so

1
∇ E 2 − (E · ∇) E


E × (∇ × E) = (16)
2
1
∇ B 2 − (B · ∇) B


B × (∇ × B) = (17)
2
Putting this into 12 we get
 MAXWELL STRESS TENSOR 3

1 ∂
f = 0 (∇ · E) E + (∇ · B) B − 0 (E × B) − (18)
µ0 ∂t
1
B × (∇ × B) − 0 E × (∇ × E)
µ0
1 ∂
= 0 (∇ · E) E + (∇ · B) B − 0 (E × B) − (19)
µ0 ∂t
 
1 1 1
∇ 0 E 2 + B 2 + 0 (E · ∇) E + (B · ∇) B
2 µ0 µ0
1
= 0 [(∇ · E) E + (E · ∇) E] + [(∇ · B) B + (B · ∇) B] − (20)
µ0
 
1 1 ∂
∇ 0 E 2 + B 2 − 0 (E × B)
2 µ0 ∂t

It might not seem that we’re making any progress, since the equations
just get longer with each alteration. However, we can now introduce the
←→
Maxwell stress tensor T which is a 3 × 3 matrix with components defined
by

   
1 2 1 1 2
Tij ≡ 0 Ei Ej − δij E + Bi Bj − δij B (21)
2 µ0 2

Note that the tensor is symmetric: Tij = Tji . If we define the scalar
product of the tensor with an ordinary vector to be another vector:

h ← →i
a · T = ∑ ai Tij (22)
j i

where the subscript j indicates the jth component of the resulting vector,
then the divergence is
 MAXWELL STRESS TENSOR 4

h ← →i
∇· T = ∑ ∂iTij (23)
j i
 
1 2
= 0 ∑ (∂i Ei ) Ej + Ei (∂i Ej ) − δij ∂i E + (24)
i 2
 
1 1 2
(∂i Bi ) Bj + Bi (∂i Bj ) − δij ∂i B
µ0 ∑ i 2
 
1 2
= 0 (∇ · E) Ej + (E · ∇) Ej − ∂j E + (25)
2
 
1 1 2
(∇ · B) Bj + (B · ∇) Bj − ∂j B
µ0 2
←
→
Comparing this with 20, we see that we can write f in terms of T and
the Poynting vector as

←→ ∂S
f = ∇ · T − 0 µ0 (26)
∂t
The total force on the volume is then
ˆ
F = fd3 r (27)
V
ˆ  
←
→ ∂S 3
= ∇ · T − 0 µ0 d r (28)
V ∂t
From the formula 23 for the divergence, we can see that the vector re-
sulting from the divergence has as its components the divergences of each
←→
column of T . Therefore we can apply the divergence theorem to the first
term in the integrand to get
ˆ ˆ
←
→ ∂
F= T · da − 0 µ0 Sd3 r (29)
S ∂t V

where S is any surface that encloses only the charges and currents within
V.
Example. We can revisit the problem of finding the magnetic force between
the two halves of a spherical shell of surface charge density σ rotating with
angular velocity ω = ωẑ. In our earlier solution we used the Biot-Savart
law and integrated over each differential ring in the rotating sphere. Using
the stress tensor, we can integrate over any volume that encloses the upper
half of the sphere, so we can choose the half space consisting of all space
 MAXWELL STRESS TENSOR 5

above the xy plane (we’re assuming that the centre of the sphere is at the
origin, so the xy plane contains the sphere’s equator). Since the distribution
of charges and currents is finite, all fields will go to zero at infinity, so we
need to integrate only over the xy plane.
We saw earlier that the magnetic field inside and outside the sphere is
  
 2µ0 Rωσ cos θr̂ − sin θθ̂θ = 2µ0 Rωσ ẑ r < R
3  3
B = µ R4 ωσ  (30)
 0 1 θ
3 r 3 2 cos θr̂ + sin θ θ̂ r > R
In the xy plane, θ = π/2 so the field is
(
2µ0 Rωσ
3 ẑ r<R
B= µ0 R4 ωσ 1 (31)
− 3 r3 ẑ r>R
Since we’re interested only in the magnetic field, we can ignore E here,
although there is a repulsive force between the two hemispheres due to the
electric field as well. Also, as the currents are steady, ∂S/∂t = 0.
From the symmetry ofh the problem, the force is in the z direction, so we
←→ i
need to work out only T · da . We get Txz = Tyz = 0 because Bx =
z
By = 0 on the xy plane, so we’re left with just Tzz :
(
2
1 2 9 µ0 σ 2 ω 2 R2 r < R
Tzz = B = 1 2 2 R8
(32)
2µ0 z 18 µ0 σ ω r6 r>R
The total force is then (the minus sign is because Tzz > 0 and da points
towards −z):
ˆ
←
→
F= T · da (33)
S
" ˆ R ˆ ∞ #
2 2π r dr
= −ẑ µ0 σ 2 ω 2 R2 2π r dr + µ0 σ 2 ω 2 R8 6
(34)
9 0 18 R r
 
2π π
= −ẑ µ0 σ 2 ω 2 R4 + µ0 σ 2 ω 2 R4 (35)
9 36
π
= − µ0 σ 2 ω 2 R4 ẑ (36)
4
This agrees with the result we got earlier using the Biot-Savart law.

P INGBACKS
Pingback: Maxwell stress tensor: force between two charges
Pingback: Momentum in electromagnetic fields
 MAXWELL STRESS TENSOR 6

Pingback: Maxwell stress tensor and electromagnetic waves