MP465 – Advanced Electromagnetism

Problem Set 3
Due by 5pm on Thursday, 9 April 2020

1. The 3-dimensional Levi-Čivita symbol ijk , where all indices range from
1 to 3 (or x to z), is defiined by

 0 if any two indices are the same,
ijk = +1 if (ijk) is an even permutation of (123),
−1 if (ijk) is an odd permutation of (123).


It satisfies the identity

3
X
ijm k`m = δik δj` − δi` δjk
m=1

and is used to define the cross-product via

  3
X
~a × ~b = ijk aj bk .
i
j,k=1

Using the above, prove the following two vector calculus identites:
         
(a) ~a × ∇ ~ × ~b + ~b × ∇ ~ × ~a = ∇ ~ ~a · ~b − ~a · ∇~ ~b − ~b · ∇
~ ~a,
         
(b) ~ × ~a × ~b = ~a ∇
∇ ~ · ~b + ~b · ∇
~ ~a − ~b ∇ ~ ~b.
~ · ~a − ~a · ∇

2. Suppose we have a linear medium subjected to an electromagnetic field,

and we measure the magnitude of the fields inside the medium to be
15 kV · m−1 (which is about the strength of the electric field which
would give you a slight zap when touching a metal doorknob) and
5 mT (the strength of the magnetic field of a typical fridge magnet).
This fields will induce tiny electric and magnetic dipole moments in
each constituent particle of the medium; determine the magnitudes of
these moments for the following media: (a) water, (b) wood and (c)

1
 air. (Assume that all media are at standard temperature and pressure,
i.e. 20◦ C and one atmosphere).
Note: in order to do this problem, you must take it upon yourself to look
up the necessary numbers like electric and magnetic susceptibilities,
mass densities and the like, and you must cite your sources (as all good
scientists must). Not doing so will result in marks being lost.
3. Maxwell’s equations in matter are
~
~ ·D
∇ ~ = ρ, ~ = J~ + ∂ D ,
~ ×H
∇
∂t
~
~ ·B
∇ ~ = 0, ~ = − ∂B ,
~ ×E
∇
∂t
where ρ and J~ are whatever free charge and current densities might be
present. From these, show that for linear media, the energy-conservation
equation
∂u ~ ~ ~ ·S
~
+ J · E = −∇ (1)
∂t
is satisfied if the electromagnetic (EM) field’s energy density u and
energy current (usually called the Poynting vector) S ~ are given by,
respectively,
1 ~ ~ ~ ~

~=E ~ ×H ~
u= D·E+H ·B , S
2
Explanation: The first term in the left-hand side of (1) is the rate of
change of the EM field’s internal energy density; the second term is the
rate of work per unit volume done on a free charge/current distribution
by the EM field (note that the magnetic field does not contribute to
this). Thus, the left-hand side is the rate at which the total energy
density, of both the EM fields and the charge/current distribution,
changes.
The right-hand side is an energy density flow rate due to propagation
of the EM field. Thus, when integrated over a volume V, this equation
states that the energy flowing into V must equal the total rate of energy
change due to both a change in the internal EM energy and work being
done on the currents. Hence, energy is conserved for linear media.

2
4. The Maxwell stress tensor T is defined to be the 3 × 3 matrix with
elements Tij given by

1 ~ ~ ~ ~

Tij := D · E + H · B δij − Di Ej − Hi Bj .
2
Show that for a linear medium with permeability µ and permittivity ,
this quantity satisfies the momentum-conservation equation
3
∂ X ∂Tij
(µSi ) + fi = − (2)
∂t j=1
∂xj

where

f~ = ρE
~ + J~ × B
~

is the force per unit volume felt by the free charge/current distribution
(obtained from the Lorentz force law).
Explanation: µS ~ is the momentum density of the EM field, so the
first term on the left-hand side of (2) is its rate of change. The second
term is the force per unit volume felt by the distribution; since force is
the same as the rate of change of momentum, f~ is the rate at which the
charge/current distribution’s momentum density changes. The right-
hand side is thus the rate of change of the total momentum density.
Thus, following the same train of thought as in the previous problem,
the right-hand side involving the stress tensor must be the rate at which
the total momentum density is being carried away from the fields and
charges (effectively, the force that’s being exerted on its surroundings).
Integrating over a volume V thus shows that the momentum flowing
into V equals the momentum lost by V. Ergo, momentum is conserved.

3
 VECTOR CALCULUS FORMULAE

1. Cartesian coordinates (x, y, z) with constant unit direction vectors êx ,

êy , êz

• position vector: ~r = xêx + yêy + zêz

• line element: d~r = dx êx + dy êy + dz êz
surface element: d~σ = dy dz êx + dx dz êy + dx dy êz
volume element: d3~r = dx dy dz
• gradient of a function f (x, y, z):

~ ∂f ∂f ∂f
∇f = êx + êy + êz
∂x ∂y ∂z

~ y, z) = Ax (x, y, z)êx + Ay (x, y, z)êy +

• divergence of a vector A(x,
Az (x, y, z)êz :

~ = ∂Ax + ∂Ay + ∂Az

~ ·A
∇
∂x ∂y ∂z

~ y, z) = Ax (x, y, z)êx +Ay (x, y, z)êy +Az (x, y, z)êz :

• curl of a vector A(x,
     
~ ×A ~ = ∂A z ∂A y ∂A x ∂A z ∂A y ∂A x
∇ − êx + − êy + − êz
∂y ∂z ∂z ∂x ∂x ∂y

• Laplacian of a function f (x, y, z):

∂ 2f ∂ 2f ∂ 2f
∇2 f = + +
∂x2 ∂y 2 ∂z 2

2. Cylindrical coordinates (r, φ, z) with unit direction vectors êr , êφ , êz

• relation to Cartesian coordinates: x = r cos φ, y = r sin φ, z un-

changed
• relation to Cartesian unit vectors:
 
êr = cos φ êx + sin φ êy êx = cos φ êr − sin φ êφ
↔
êφ = − sin φ êx + cos φ êy êy = sin φ êr + cos φ êφ

with êz the same for both systems.

4
• position vector: ~r = rêr + zêz
• line element: d~r = dr êr + rdφ êφ + dz êz
surface element: d~σ = rdφ dz êr + dr dz êφ + rdr dφ êz
volume element: d3~r = rdr dφ dz
• gradient of a function f (r, φ, z):

~ ∂f 1 ∂f ∂f
∇f = êr + êφ + êz
∂r r ∂φ ∂z

~ φ, z) = Ar (r, φ, z)êr + Aφ (r, φ, z)êφ +

• divergence of a vector A(r,
Az (r, φ, z)êz :

~ = 1 ∂ (rAr ) + 1 ∂Aφ + ∂Az

~ ·A
∇
r ∂r r ∂φ ∂z

~ φ, z) = Ar (r, φ, z)êr +Aφ (r, φ, z)êφ +Az (r, φ, z)êz :

• curl of a vector A(r,
     
~ ~ 1 ∂Az ∂Aφ ∂Ar ∂Az 1 ∂ ∂Ar
∇×A = − êr + − êφ + (rAφ ) − êz
r ∂φ ∂z ∂z ∂r r ∂r ∂φ

• Laplacian of a function f (r, φ, z):

1 ∂ 2f ∂ 2f
 
2 1 ∂ ∂f
∇f = r + 2 2+ 2
r ∂r ∂r r ∂φ ∂z

5
3. Spherical coordinates (r, θ, φ) with unit direction vectors êr , êθ , êφ
• relation to Cartesian coordinates: x = r sin θ cos φ, y = r sin θ sin φ,
z = r cos θ
• relation to Cartesian unit vectors:

êr = sin θ cos φ êx + sin θ sin φ êy + cos θ êz 
êθ = cos θ cos φ êx + cos θ sin φ êy − sin θ êz
êφ = − sin φ êx + cos φ êy


 êx = sin θ cos φ êr + cos θ cos φ êθ − sin φ êφ
↔ êy = sin θ sin φ êr + cos θ sin φ êθ + cos φ êφ
êz = cos θ êr − sin θ êθ


• position vector: ~r = rêr

• line element: d~r = dr êr + rdθ êθ + r sin θdφ êφ
surface element: d~σ = r2 sin θdθ dφ êr + r sin θdr dφ êθ + rdr dθ êφ
volume element: d3~r = r2 sin θdr dθ dφ
• gradient of a function f (r, θ, φ):

~ ∂f 1 ∂f 1 ∂f
∇f = êr + êθ + êφ
∂r r ∂θ r sin θ ∂φ
~ θ, φ) = Ar (r, θ, φ)êr + Aθ (r, θ, φ)êθ +
• divergence of a vector A(r,
Aφ (r, θ, φ)êφ :

~ ·A
~ = 1 ∂ 2
1 ∂ 1 ∂Aφ
∇ 2
r Ar + (sin θAθ ) +
r ∂r r sin θ ∂θ r sin θ ∂φ
~ θ, φ) = Ar (r, θ, φ)êr +Aθ (r, θ, φ)êθ +Aφ (r, θ, φ)êφ :
• curl of a vector A(r,
   
~ ~ 1 ∂ ∂Aθ 1 ∂Ar 1 ∂
∇×A = (sin θAφ ) − êr + − (rAφ ) êθ
r sin θ ∂θ ∂φ r sin θ ∂φ r ∂r
 
1 ∂ ∂Ar
+ (rAθ ) − êφ
r ∂r ∂θ

• Laplacian of a function f (r, θ, φ):

∂ 2f
   
2 1 ∂ 2 ∂f 1 ∂ ∂f 1
∇f = 2 r + 2 sin θ + 2 2
r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂φ2

6
 MP465 – Advanced Electromagnetism

Solutions to Problem Set 3

1. The point of this problem is just to get you used to the types of manip-
ulations you need to do when using the Levi-Civita symbol, because it
appears surprisingly frequently throughout maths and phsyics...

(a) The curl of a vector field ~b is, component-wise,

  X
~ × ~b
∇ = ~ k b`
mk` (∇)
m
k`
X ∂b`
= mk`
mn
∂xk

where all sums will be from 1 to 3 (or over x, y and z – same

thing). Thus,
h  i X  
~a × ∇~ × ~b = ijm aj ∇~ × ~b
i m
jm
!
X X ∂b`
= ijm aj mk`
jm k`
∂xk
X ∂b`
= ijm mk` aj
jk`m
∂xk

Now, because the Levi-Civita symbol is totally antisymmetric,

mk` = −km` = k`m and so the above sum may be rewritten as
!
h 
~ × ~b
i X X ∂b`
~a × ∇ = ijm k`m aj
i
jk` m
∂xk
X ∂b`
= (δik δj` − δi` δjk ) aj .
jk`
∂x k

1
 Now, if we do the sums over j, k and ` successively, we get
h  i X ∂b` ∂b`

~
~a × ∇ × b~ = δik a` − δi` ak
i
k`
∂x k ∂xk
 
X ∂b` ~
= a` − δi`~a · ∇b`
`
∂x i

∂~b ~ i.
= ~a · − ~a · ∇b
∂xi
If we swap the two vectors, we obviously get
h 
~b × ∇
i ∂~a ~ ~
~ × ~a = ~b · − b · ∇ai
i ∂xi
so adding the two gives
h   
~ × ~b + ~b × ∇~ × ~a
i ∂~b ~ i + ~b · ∂~a − ~b · ∇a
~ i.
~a × ∇ = ~a · − ~a · ∇b
i ∂xi ∂xi

The second and last terms together are the ith -component of −(~a ·
~ ~b − (~b · ∇)~
∇) ~ a · ~b)]i is the
~ a, and the product rule tells us that [∇(~
first plus the third terms, and so we’ve proved what we wanted:
         
~a × ∇ ~ × ~b + ~b × ∇ ~ × ~a = ∇ ~ ~a · ~b − ~a · ∇~ ~b − ~b · ∇ ~ ~a.

(b) For this problem, we need to be cogniscent of the fact that the
derivative in the curl acts on a product of the components of ~a and
~b and to use the appropriate theorem of calculus. More specifically,
since
  X
~a × ~b = mk` ak b` ,
m
k`

2
 then
h  i X
~ ~
∇ × ~a × b = ~ j (~a × ~b)m
ijm (∇)
i
jm
!
X ∂ X
= ijm mk` ak b`
jm
∂xj k`
 
X ∂
= ijm mk` (ak b` )
jk`m
∂x j
 
X ∂ak ∂b`
= ijm mk` b ` + ak .
jk`m
∂xj ∂xj

Again using mk` = k`m and summing over m gives

 
h 
~ × ~a × ~b
i X ∂a k ∂b `
∇ = (δik δj` − δi` δjk ) b ` + ak .
i
jk`
∂xj ∂xj

As we did in (a), we now do the sums over j, k and `, in that

order:
h  i X   ∂ak ∂b `
 
∂a k ∂b `

~ × ~a × ~b
∇ = δik b ` + ak − δi` b ` + ak
i
k`
∂x ` ∂x ` ∂x k ∂xk

X ∂ai 
∂b` 
~ · ~a)b` + (~a · ∇)b
~ `
= b ` + ai − δi` (∇
`
∂x ` ∂x `

= (~b · ∇)a ~ · ~b) − (∇

~ i + ai (∇ ~ · ~a)bi − (~a · ∇)b
~ i

which we recognise as the ith component of

         
~ × ~a × ~b = ~a ∇
∇ ~ · ~b + ~b · ∇
~ ~a − ~b ∇ ~ ~b
~ · ~a − ~a · ∇

and that’s exactly what we wanted to prove.

2. For this problem, we need to determine the polarisability and magneti-

sation induced by the given fields in the materials presented. Recall
that the definitions of the electric and magnetic susceptibilities are

P~ = 0 χe E,
~ ~ = χm H.
M ~

3
 The first of these will give the polarisability directly from the electric
~ = B/µ,
field, but we have to manipulate the second a bit: recall that H ~
where µ = (1 + χm )µ0 is the material’s permeability. This means that
we get the magnetisation from the magnetic field via M ~ = χm B/µ.
~
But this isn’t all: we want the electric and magnetic dipole of each
constituent atom/molecule in our material, which means we need a
number density ρN for each material, i.e. the number of particles per
unit volume. Now, if the total volume is V , then it has a total mass
ρM V , where ρM is the mass density of the material. If the mass of each
constituent is A (we’re using m, M and µ for other quantities, so best
to use a nonstandard variable for mass here), then N = ρM V /A, so
ρN = N/V = ρM /A is the number density. Thus, if p~ and m ~ are the
~ ~
constituent dipoles, then P = ρN p~ and M = ρN m.
~ Thus,
0 χe A ~ χm A ~
p~ = E, m
~ = B.
ρM µρM

So now we’re in a position to put in some numbers. Me, whenever

I’m confronted with the need for physical quantities, I go straight to
www.engineeringtoolbox.com, and I get these:

material mass density relative permittivity relative permeability

water 1000 80 0.999992
wood 500-800 2-6 1.00000043
air 1.204 1.000536 1.00000037

(All densities are in kilogrammes per cubic metre, and for air and water,
I’ve taken the values at 20◦ C.) Note that it’s the relative permittivity
/0 and relative permeability µ/µ0 listed above, and we can see that
µ ≈ µ0 holds for all three materials (as we mentioned before). However,
µ = (1+χm )µ0 gives χm = µ/µ0 −1, so we can’t use that approximation
to get that, because they’d all be zero. But they’re easily computed
from the exact values above: for water, wood and air, we get magnetic
susceptibilities of −8.0×10−6 , 4.3×10−7 and 3.7×10−7 respectively. (So
water is diamagnetic and wood and air are paramagnetic.) Similarly,
χe = /0 − 1 and so we get electric susceptibilities of 79, 1-5 and
5.36 × 10−4 .

4
 Now we need constituent masses. Water is easy: it’s very close to 18
times the proton mass 1.67 × 10−27 kg, so we’ll use that. Similarly, if
we take air to be 20% O2 and 80% N2 , we get 28.8 times the proton
mass. Thus, we have everything we need for (a) and (c), and putting
in the numbers gives the electric dipoles to be 3.15 × 10−34 A · m · s
for water and 2.84 × 10−36 A · m · s for air, both in the same direction
as the electric field. The magnetic dipoles are 9.57 × 10−31 A · m2 for
water and 5.88 × 10−29 A · m2 for air, with the former in the opposite
~ and the latter in the same direction.
direction to B
Now, wood is the tricky one, because obviously not all wood is the
same, which is why there’s a range of values above. So let’s take the
midpoints, i.e. a mass density of 650 kg · m−3 and an electric suscepti-
bility of 3. Also, what we take as a constituent mass is not obvious, so
we’ll make a very crude approximation that it’s made entirely of cellu-
lose, which Wikipedia tells me has a molar mass of 162.1406 grammes
per mole, so we’ll take 162 proton masses as its constituent mass. With
these assumptions, we get an electric dipole of 1.66 × 10−36 A · m · s and
7.12 × 10−31 A · m2 for the magnetic dipole, both in the same direction
as the fields inducing them.
3. We’ll do this and the next problem by computing both sides of the
equations given and showing that they’re equal rather than deriving
them from scratch.
First off, for a linear medium, the energy density and Poynting vector
are
~ ~ 1 ~ ~ ~ = 1E~ × B.
~
u= E ·E+ B · B, S
2 2µ µ
So the time derivative of the energy density is

∂u ~ ~
~ · ∂E + 1 B
= E ~ · ∂B
∂t ∂t µ ∂t
and since Maxwell’s equations tell us that
~
∂E
µ ~ ×B
= ∇ ~ − µJ,
~
∂t
~
∂B
= −∇~ × E,
~
∂t
5
 we get
 
∂u ~· 1~ ~ ~ 1~ ~ ~
= E ∇×B−J − B ·∇×E
∂t µ µ
so
∂u ~ ~ 1 ~ ~ ~ ~ ~ ~

+J ·E = − B·∇×E−E·∇×B .
∂t µ

Now let’s compute the divergence of the Poynting vector using the
Levi-Civita symbol:

∇ ~ = 1∇
~ ·S ~ · (E~ × B)~
µ
1X ∂ ~ ~ i
= (E × B)
µ i ∂xi
1X ∂
= (ijk Ej Bk )
µ ijk ∂xi
 
1X ∂Ej ∂Bk
= ijk Bk + Ej
µ ijk ∂xi ∂xi
!
1 X ∂Ej X ∂Bk
= Bk kij − Ej jik
µ kij
∂xi jik
∂xi

since ijk = kij = −jik . In the first term, doing the sums over i and j
P
first gives k Bk (∇ ~ × E)
~ k=B ~ ·∇~ × E,
~ and in the second, doing the
P
(ik) sums gives j Ej (∇ ~ × B)~ j=E ~ ·∇~ × B,
~ so

~ = 1 B
 
~ ·S
∇ ~ ·∇
~ ×E
~ −E
~ ·∇
~B~
µ
and therefore, as desired,
∂u ~ ~ ~ · S.
~
+ J · E = −∇
∂t

4. The first step for this problem is quite similar to what we did in Problem

6
3, namely,
∂ X ∂
(µSi ) =  ijk (Ej Bk )
∂t jk
∂t
 ! ! 
X ~
∂E ~
∂B 
=  ijk  Bk + Ej
jk
∂t ∂t
j k
 
X 1~ ~ − J~ Bk − 
X 
~ ×E~

= ijk ∇×B ijk Ej ∇
jk
µ j jk
k
 
1 ~ ~ ×B ~ − J~ × B~ − E
~ × (∇ ~ × E) ~
= (∇ × B) .
µ i

Thus, using the definition of f~,

 
∂ 1 ~ ~ ~ ~ ~ ~ ~
(µSi ) + fi = (∇ × B) × B − E × (∇ × E) + ρE
∂t µ i
 
1~ ~ × B)
~ − E
~ × (∇
~ × E)
~ + E(
~ ∇ ~ · E)
~
= − B × (∇
µ i

~ · E.
since Gauss’ law for electric fields gives ρ = ∇ ~ Now, if we invoke
Problem 1(a) above, we see that
h 
~
i ∂~a ~ i
~a × ∇ × ~a = ~a · − ~a · ∇a
i ∂xi
1 ∂ ~ i.
= (~a · ~a) − ~a · ∇a
2 ∂xi
If we use this, we see that
 
∂ ∂ ~ ~ 1 ~ ~ 1 ~ ~
(µSi ) + fi = − E·E+ B · B + (B · ∇)Bi
∂t ∂xi 2 2µ µ
+(E~ · ∇)E
~ i + (∇
~ · E)E
~ i.

Now to compute the right-hand side of the equation we want to verify:

for a linear medium, the Maxwell stress tensor will be
 
~ ~ 1 ~ ~ 1
Tij = E·E+ B · B δij − Ei Ej − Bi Bj .
2 2µ µ

7
Therefore,
X ∂Tij X ∂   1 ~ ~

1

= ~ ~
E·E+ B · B δij − Ei Ej − Bi Bj
j
∂xj j
∂x j 2 2µ µ
  
X ∂ ~ ~ 1 ~ ~
= δij E·E+ B·B
j
∂xj 2 2µ
 
∂Ei ∂Ej 1 ∂Bi 1 ∂Bj
−  Ej + Ei + Bj + Bi .
∂xj ∂xj µ ∂xj µ ∂xj

If we now do the sum, we get

X ∂Tij  
∂ ~ ~ 1 ~ ~
= E·E+ B·B
j
∂x j ∂x i 2 2µ
 
~ ~ ~ ~ 1 ~ ~ 1 ~ ~
− (E · ∇)Ei + Ei (∇ · E) + (B · ∇)Bi + Bi (∇ · B) .
µ µ

~ ·B
But ∇ ~ = 0, so this is

X ∂Tij  
∂ ~ ~ 1 ~ ~ 1 ~ ~
= E·E+ B·B − (B · ∇)Bi
j
∂xj ∂xi 2 2µ µ
~ · ∇)E
−(E ~ i − (∇
~ · E)E
~ i

and this proves that

∂ X ∂Tij
(µSi ) + fi = −
∂t j
∂xj

as we’d hoped.