PHYS 20141 ELECTROMAGNETISM : 2021/22

Section 1 : Mathematical Preliminaries

P. Campbell
(for use in conjunction with the teaching material on the PHYS20141 Blackboard area )

Department of Physics & Astronomy

University of Manchester

September 28, 2021

 1
Mathematical preliminaries

WEEK 1 SUPPORTING TEXTS:

Grant & Phillips, Chapter 1 and Appendix B
Griffiths, Chapter 1
Any previous Vector Algebra notes

In this lecture series we bring together ideas introduced in PHYS10342 “Electricity and
Magnetism” and PHYS10372 “Mathematics 2” (and in your quantum mechanics and
relativity courses) and develop the formalism with which you can approach your future
problems in modern physics. Studying electromagnetism can be a mathematically in-
tensive pursuit. You already have foundations in some of the mathematics we will use
(vectors and polar coordinates), but other material (the -function and Laplace/Poisson
equation) are presented here initially and covered in more detail in the course on “Math-
ematics of waves and fields”, PHYS20171 (and MATH20401).

1.1 Scalars and vectors

A scalar field (r) is a quantity with magnitude (positive or negative) at a position

r = (x, y, z). Thus, a scalar field is a quantity which has di↵erent values at di↵erent
locations. Examples include energy, charge density and the electric potential.
A vector field v(r) has a magnitude and direction at r. Examples include the electric
and magnetic fields, force, velocity acceleration and current density. We write vectors in
terms of a basis:

0 1
v1
v = v1 î + v2 ĵ + v3 k̂ ⌘ @ v2 A ⌘ (v1 , v2 , v3 ) . (1.1)
v3

2
 1.1 Scalars and vectors 3

The vector has been written in terms of an orthonormal basis,

0 1
1
î = x̂ = êx = ê1 = @ 0 A, (1.2)
0
0 1
0
ĵ = ŷ = êy = ê2 = @ 1 A , (1.3)
0
0 1
0
k̂ = ẑ = êz = ê3 = @ 0 A. (1.4)
1
To construct a unit vector, one divides a vector by its length. Thus a unit vector v̂ is
v
v̂ = , (1.5)
|v|
where the magnitude of v is
q
v = |v| = v12 + v22 + v32 (1.6)
The scalar product between two vectors (in three dimensions) is
3
X
a · b = a 1 b1 + a 2 b2 + a 3 b3 = ai bi = ab cos ✓, (1.7)
i=1
where ✓ is the angle between the vectors a and b.
The vector (or cross) product is
î ĵ k̂
a⇥b = a1 a2 a3 (1.8a)
b 1 b 2 b3
= abn̂ sin ✓ (1.8b)
= b ⇥ a. (1.8c)
The triple product is
a1 a2 a3
a · (b ⇥ c) = b1 b2 b3 (1.9a)
c1 c2 c3
= c · (a ⇥ b) (1.9b)
= b · (c ⇥ a). (1.9c)
Two useful vector identities are:
a ⇥ (b ⇥ c) = (a · c)b (a · b)c, (1.10)
 1.2 Vector calculus 4

(a ⇥ b) · (c ⇥ d) = (a · c)(b · d) (a · d)(b · c). (1.11)

The use of vectors is common place in the study of electromagnetism. For example,
Coulomb’s law between two point charges is
1 q1 q2 q1 q2 r 1 r2
Force = ) F12 = . (1.12)
4⇡"0 r2 4⇡"0 |r1 r2 | 3

1.2 Vector calculus

Until we state otherwise, we will work in the Cartesian basis. Let = (x, y, z) be a
scalar field and
3
X
v = v(x, y, z) = vx (x, y, z)î + vy (x, y, z)ĵ + vz (x, y, z)k̂ = vi êi (1.13)
i=1

be a vector field.

1.2.1 The grad operator

In Cartesian coordinates, the gradient (grad) is an operator that acts upon a scalar field
to give a vector field,
X@ 3
@ @ @
r = î + ĵ + k̂ = êi . (1.14)
@x @y @z @xi
i=1

1.2.2 The div operator

The divergence (div) operator takes a vector field and returns a scalar field, via

X @vi 3
@vx @vy @vz
r·v = + + = . (1.15)
@x @y @z @xi
i=1

Consider the vector (which is represented in figure 1.1),

v = xî + y ĵ, (1.16)

then

r · v = 2 > 0. (1.17)

If r · v > 0, one has the interpretation of a “source”, and if r · v < 0 then of a “sink”.
Vector fields with these properties are illustrated in figure 1.1.
 1.2 Vector calculus 5

Fig. 1.1. On the left is a vector field with r · v > 0, for example v = xî + y ĵ, and on the left is
one with r · v < 0.

Fig. 1.2. The vector v = y î + xĵ which has a curl which is in the positive z-direction. which
would be out of the page.

1.2.3 The curl operator

Another important vector operation is the curl,

î ĵ k̂
@ @ @
r⇥v = @x @y @z . (1.18)
vx vy vz

Consider v = y î + xĵ which is illustrated in figure 1.2, then r ⇥ v = 2(0, 0, 1) = 2k̂.

1.2.4 The Laplacian operator and vector identities

Finally, the Laplacian, which is a scalar operator that can act on both vectors and scalars,
@ @ @
r · r = r2 = 2
+ 2 + 2, (1.19)
@x @y @z
in Cartesian coordinated. Hence,

@2 @2 @2
r2 = + + , (1.20)
@x2 @y 2 @z 2
r2 v = r2 vx î + r2 vy ĵ + r2 vz k̂. (1.21)
(1.22)
 1.2 Vector calculus 6

There are 3 important vector identities which we will use during the course
r⇥r = 0, (1.23)
r · (r ⇥ v) = 0, (1.24)
2
r ⇥ (r ⇥ v) = r(r · v) r v, (1.25)
which hold for all scalar fields and vector fields v. These are all easy to show in
component notation, for example
î ĵ k̂
@ @ @
r⇥r = @x @y @z
@ @ @
@x @y @z
✓ 2 ◆ ✓ ◆ ✓ ◆
@ @2 @2 @2 @2 @2
= î + ĵ + k̂
@y@z @z@y @z@x @x@z @x@y @y@x
= 0, (1.26)
if partial derivatives commute.

1.2.5 Integral theorems

The fundamental theorem of calculus states that
Z b
df
dx = f (b) f (a), (1.27)
a dx
or, the integral of a derivative is determined by the behavior of the function on the
boundary. There are two analogues of this in vector calculus.
The divergence theorem links the divergence of a vector field inside a volume and the
flux of the field through a surface enclosing the volume,
Z Z
r · vdV = v · dS, (1.28)
V S

where dS = dS n̂ and n̂ is the outward facing normal vector of the surface S enclosing
the volume V as illustrated in the left-hand portion of figure 1.3.
Stokes’ theorem links the curl of a vector field through a surface and the integral of
the field around a loop enclosing the surface as illustrated in the right hand portion of
figure 1.3 ,
Z I
r ⇥ v · dS = v · d`. (1.29)
S L

The direction of the normal vector to the surface n̂, such that dS = n̂dS, is given by the
right-hand screw rule relative to the sense of the line integral. In this case, this means
that n̂ is into the page.
 1.2 Vector calculus 7

V S

n̂

Fig. 1.3. On the left is an illustration of the situation used in the divergence theorem: volume V ,
closed surface S enclosing V and the outward facing normal n̂. On the right is that for Stokes’
theorem: a closed surface, S, and bounding loop, L. For this direction of the loop, the right-hand
screw rule implies that the normal, n̂ is into the page.

1.2.6 Curvilinear coordinates

In the Cartesian coordinate system, vectors are written in terms of the basis î, ĵ, k̂, which
are constant in space (and time). Plane polar coordinates in 2D are an example of a
curvilinear coordinate system, whereby the basis vectors vary with position
r̂ = cos ✓î + sin ✓ĵ, ✓ˆ = sin ✓î + cos ✓ĵ. (1.30)
In (r, ✓) coordinates the integration form changes,
dxdy ! rdrd✓, (1.31)
and the position vector becomes
✓ ◆
r cos ✓
r = rr̂ = . (1.32)
r sin ✓
We will also use 3D curvilinear coordinate systems: cylindrical polar coordinates
(r, ✓, z) and spherical polar coordinates (r, ✓, ).
Cylindrical polar coordinates: In the case of cylindrical polar coordinates:
0 1
⇣ ⌘ ⇣ ⌘ r cos ✓
ˆ ẑ , dxdydz ! rdrd✓dz,
î, ĵ, k̂ ! r̂, ✓, r = rr̂r + zẑ = @ r sin ✓ A . (1.33)
z
A very important point to note is that the gradient, divergence, curl and Laplacian
operators change in curvilinear coordinates; the reason being that the surface and volume
elements of the bases are position dependent.
Writing a vector in cylindrical polar coordinates as
v = vr r̂ + v✓ ✓ˆ + vz ẑ, (1.34)
 1.2 Vector calculus 8

then
@ 1@ ˆ @
r =r̂ + ✓+ ẑ, (1.35)
@r r @✓ @z
1 @ 1 @v✓ @vz
r·v = (rvr ) + + , (1.36)
r @r
✓ r◆@✓ ✓ @z ◆ ✓ ◆
1 @vz @v✓ @vr @vz ˆ 1 @rv✓ @vr
r⇥v = r̂ + ✓+ ẑ,
r @✓ @z @z @r r @r @✓
✓ ◆ 2 2
1 @ @ 1 @ @
r2 = r + 2 2 + . (1.37)
r @r @r r @✓ @z 2
The volume element becomes
dV = rdrd✓dz. (1.38)

Spherical polar coordinates : Writing a vector in spherical polar coordinates as

v = vr r̂ + v✓ ✓ˆ + v ˆ, (1.39)
then
@ 1@ ˆ 1 @ ˆ
r = r̂ + ✓+ , (1.40)
@r r @✓ r sin ✓ @
1 @ 1 @ 1 @v
r·v = 2
r 2 vr + (sin ✓v✓ ) + , (1.41)
r @r r sin ✓ @✓ r sin ✓ @
✓ ◆ ✓ ◆
1 @ @v✓ 1 1 @vr @
r⇥v = (sin ✓v ) r̂ + (rv ) ✓ˆ
r sin ✓ @✓ @ r sin ✓ @ @r
✓ ◆
1 @ @vr ˆ
+ (rv✓ ) , (1.42)
r @r @✓
✓ ◆ ✓ ◆
2 1 @ 2@ 1 @ @ 1 @2
r = r + sin ✓ + . (1.43)
r2 @r @r r2 sin ✓ @✓ @✓ r2 sin2 ✓ @ 2
The volume element becomes
dV = r2 sin ✓drd✓d . (1.44)
 1.3 Dirac -function 9

1.3 Dirac -function

The Dirac -function is a function defined by

(x a) = 0 unless x = a. (1.45)

and
Z 1
(x a)f (x) dx = f (a), (1.46)
1

which implies that

Z 1
(x a) dx = 1. (1.47)
1

Formally, the value of (x a) at x = a is infinite such that the above integral is finite.
In 3 dimensions, the -function is
(3)
(r a) = (x a1 ) (y a2 ) (z a3 ). (1.48)

An analogy to a -function is the normalized Gaussian function,

1 (x a)2
f (x) = p e 2 2 , (1.49)
2⇡ 2
R1
in the limit that the width ! 0. This functional form has the property that 1 f (x)dx =
1 independent of . The function f (x) is plotted for a=0 and = 1.0, 0.3 and 0.1 in
figure 1.4.
We use -functions to model point distributions. For example, the charge density due
to a distribution of point charges is modeled via
n
X
(3)
⇢(r) = qi (r ri ) , (1.50)
i=1

so that the total charge is

Z X
Q= ⇢dV = qi , 8ri 2 V. (1.51)
V

1.4 Laplace’s & Poisson’s Equations

Poisson’s equation is

r2 f = 4⇡g, (1.52)

and Laplace’s equation is that with g = 0. They are relevant to problems involving
potentials for example in electromagnetism and gravity.
 1.4 Laplace’s & Poisson’s Equations 10

4
!=1
3.5
! = 0.3
3 ! = 0.1
2.5
f(x)

2
1.5
1
0.5
0
-4 -2 0 2 4
x

Fig. 1.4. The Gaussian function for = 1.0, 0.3 and 0.1. Notice that the area under the curve
remains the same, but the function becomes more and more peaked at x = 0 as ! 0.

If the scalar field f only depends on r, then we say that f is spherically symmetric.
Thus, Laplace’s equation for a spherically symmetric scalar field f = f (r) is
✓ ◆
1 d 2 df
r = 0, (1.53)
r2 dr dr
which is integrated once to give
df
r2 = A, (1.54)
dr
and twice to give
A
f =B , (1.55)
r
where A, B are constants. If one has the condition that f (r = 1) = 0 then B = 0, so
that
A
f (r) = . (1.56)
r
A potential of this form gives rise to a 1/r2 force-law.
 1.4 Laplace’s & Poisson’s Equations 11

1.4.1 Uniqueness of solutions to Poisson’s equation*

Suppose f1 and f2 are solutions to Poissons equation in some region V , bounded by S,
with the same boundary conditions. Now define h = f1 f2 , so that r2 h = 0 in V and
h = 0 on S.
To prove the uniqueness theorem (i.e. that if two solutions exist to Poisson’s equation
with the same boundary conditions, then they are the same solution), we consider
r · (hrh) = hr2 h + |rh|2 , (1.57)
which can then be integrated over V ,
Z Z
r · (hrh) dV = hr2 h + |rh|2 dV, (1.58)
V V

but since r2 h = 0 within V ,

Z Z Z
hrh · dS = r · (hrh) dV = |rh|2 dV, (1.59)
S V V

where the first equality comes from the divergence theorem. Now, h = 0 on S, and
hence,
Z
|rh|2 dV = 0. (1.60)
V

As |rh|2 is positive and its integral vanishes it must be zero, rh = 0. Thus h is a

constant in V , which must be zero due to the boundary condition, and f1 = f2 .

1.4.2 General solution to Poisson’s equation*

One can show that
✓ ◆ ✓ ◆
1 r r0 1
r = , r 2
= 4⇡ (3)
r r0 . (1.61)
|r r0 | |r r0 |3 |r r0 |
The general solution of r2 f = 4⇡g is
Z
g(r0 )
f (r) = dV 0 . (1.62)
V |r r0 |
We can confirm this by acting with the Laplacian upon the proposed solution,
Z ✓ ◆
2 0 0 2 1
r f (r) = dV g(r )r (1.63)
V |r r0 |
Z
= 4⇡ dV 0 g(r0 ) (3) r r0 (1.64)
V
= 4⇡g(r). (1.65)
 1.4 Laplace’s & Poisson’s Equations 12

By the uniqueness theorem, because it is a solution it is the only solution.

If we define ✓0 to be the angle between the vectors r and r0 then
|r r0 |2 = r2 + (r0 )2 2r · r0
= r2 + (r0 )2 2rr0 cos ✓0 . (1.66)
Hence if the integral in (1.62) is over the whole of space and the function g(r) = g(r),
that is, it is only a function of the radial coordinate, then
Z 1 Z ⇡ Z 2⇡
g(r0 )
f (r) = f (r) = dr0 (r0 )2 d✓0 sin ✓0 d 0 2 . (1.67)
0 0 0 (r + (r0 )2 2rr0 cos ✓0 )1/2
 PHYS 20141 ELECTROMAGNETISM : 2021/22

Section 2 : Maxwell’s equations in vacuum

P. Campbell
(for use in conjunction with the teaching material on the PHYS20141 Blackboard area)

Department of Physics & Astronomy

University of Manchester

September 28, 2021

 2
Maxwell’s equations in vacuum

WEEK 2 SUPPORTING TEXTS:

Grant & Phillips; Chapter 10 and take !, µ = 1 for vacuum
Griffiths; Gauss 2.2.1, No Monopoles 5.3.2
STATIC Faraday-Lens 2.2.4, Ampère’s law 5.3.3
DYNAMIC Faraday-Lens 7.2.1, Ampère-Maxwell law 7.3
CONTINUITY page 222

Maxwell’s equations are the governing set of equations for electromagnetism. They de-
scribe how electric and magnetic fields are linked, and how they are generated. Histor-
ically, they were deduced empirically. In this section we will derive the equations and
discuss some of the applications to vacuum situations.

2.1 Currents the continuity equation

We start by considering the sources of electromagnetism: the charge density ρ(r, t) and
the current density, j(r, t). The charge density is determined by the amount of charge
in a small piece of volume (just as mass density is mass divided by volume). So, if dq is
the amount of charge in a volume dV , then
! !
Q = dq = ρ(r, t)dV. (2.1)

The current is a flux through a surface

! !
I = dI = j(r, t) · dS. (2.2)

We note that the charge density is a scalar field and the current density is a vector
field and both can be defined at any point in space and time. The current, I, is a flux
through a surface and is, as such, not a vector. It can only be postive or negative (or
zero) but can however be thought of as having a direction – this direction is specified by
the orientation of the lower dimensional surface, eg. a wire, through which the current
flows.

2
 2.1 Currents the continuity equation 3

dA = n̂dA

Fig. 2.1. Current j through a surface, with surface element dS = n̂dA.

The conservation of charge implies that the rate of loss of charge through a surface is
given by the current, that is, Q̇ = −I. Hence, we find that
! ! !
ρ̇ dV = Q̇ = − j · dS = − ∇ · j dV, (2.3)
V S V
where the final equality follows from the divergence theorem. Therefore
!
{ρ̇ + ∇ · j} dV = 0, (2.4)
V
and, as this must hold over any volume V , the integrand itself must be zero,
ρ̇ + ∇ · j = 0. (2.5)
This is the continuity equation.
Now consider a current through a surface, as illustrated in Figure 2.1. One can define
a velocity field v such that j = ρv which corresponds to the local velocity of the positive
charge carriers. Hence,
! ! ! !
I = dI = j · dS = ρv · n̂dS = −nevdrift dS, (2.6)

where vdrift = −v · n̂ is the electron drift velocity through the surface [ms−1 ], n" is the
number density [m−3 ], e is the charge on the electron (e = 1.6 × 10−19 C) and dS is
the area of the surface [m2 ]. The minus sign arises as the current is defined to be in
the direction of the +ve charge carriers and the electron drift velocity is in the opposite
direction.
 2.2 Integral forms of Maxwell’s equations 4

2.2 Integral forms of Maxwell’s equations

2.2.1 Various definitions
In this section we present a few definitions (that are hopefully familiar).
• Electric field, or electric flux density, units [Vm−1 ], E(r, t).
• Magnetic field, or magnetic flux density, units [T], B(r, t).
• Electric flux through a surface, units [Vm],
!
ΦE = E · dS. (2.7)
S

• Magnetic flux through a surface, units [Wb] = [Tm2 ] = [Vs],

!
ΦB = B · dS. (2.8)
S

The force per unit charge carries units [NC−1 ], and is computed via the Lorentz force
law,
f = E + v × B. (2.9)
The electromagnetic energy density has units [Jm−3 ] and in vacuum is
1 1
u = ε0 |E|2 + |B|2 , (2.10)
2 2µ0
and thus the total electromagnetic energy in a volume V is
!
Utot = u dV, (2.11)
V

whose unit is [J]. The electromotive force is the integrated component of the electric field
along a line,
!
E = E · d!, (2.12)

whose unit is [V].

 2.2 Integral forms of Maxwell’s equations 5

2.2.2 Gauss’ law

A range of experimental results, many taken in collaboration with Weber, led Gauss
to assert that: “the flux of electric field through a closed surface S is proportional to
the amount of charge in a volume V that the surface encloses”. The mathematical
formulation is
1
! !
Q
E · dS = ΦE = = ρ dV. (2.13)
S ε0 ε0 V
The constant of proportionality, ε0 , which is known as the permittivity of free space, can
be measured experimentally and quantifies the strength of electrostatics (in the sense
that it quantifies the absolute strength of the electric field, and hence electrostatic force,
due to a particular charge).
We can use the divergence theorem on the LHS to turn the area integral into a volume
integral,
1
! ! !
E · dS = ∇ · E dV = ρ dV, (2.14)
S V ε0 V
so that
! # $
1
∇·E− ρ dV = 0. (2.15)
V ε0

As with the case of the continuity equation, this integral must hold over any integration
volume V , and therefore its integrand must be zero, meaning that
1
∇·E= ρ, (2.16)
ε0
which is the differential form of Gauss’ law.

2.2.3 No magnetic monopoles

No magnetic monopoles have ever been detected. Hence, we assert that:“the flux of
magnetic field through any closed surface is zero”, which can be expressed as
!
B · dS = ΦB = 0. (2.17)
S

Using identical arguments to those used in the case of Gauss’ law one derives that
∇·B = 0 which must be true for every magnetic field, both static and time-varying. A
consequence of this relationship is that magnetic field lines never start (or end) anywhere:
there are no “sources” or “sinks” of magnetic field; they must either be closed, or infinite.
 2.2 Integral forms of Maxwell’s equations 6

2.2.4 Faraday-Lenz law

Again relying on (considerable) experimentation, Faraday deduced that: “the electromo-
tive force (EMF) induced in a closed loop is equal to the rate of change of magnetic flux
linked in the loop” and Lenz observed that: “the EMF is generated in a direction such
that the current flow will annul the change in flux ”. These two empirical results estab-
lish the first connection between the two apparently unrelated phenomena of electricity
and magnetism (electromagnetic induction) and can be written in a single mathematical
equation,
d
E =− ΦB . (2.18)
dt
The minus sign is the consequence of Lenz’s observation. Using the definition of the
EMF and Stoke’s theorem,
! ! !
E · d! = ∇ × E · dS = − Ḃ · dS. (2.19)
L S S

Using similar arguments as for Gauss’s law, but this time asserting that this must be
true for all surfaces, one can deduce the differential form of the Faraday-Lenz law

∇ × E + Ḃ = 0 . (2.20)

2.2.5 Ampère’s law and the displacement current

Ampère, experimenting with magnets, deduced that: “the circulation of the magnetic
field around a closed loop L is proportional to the current passing through the surface
that the loop encloses”,
% & !
B · d! = µ0 Ii = µ0 j · dS . (2.21)
L S

The constant of proportionality, µ0 , known as the permeability of free space, defines the
strength of magnetic field for a given current, in the same way that the permittivity of
free space defines the strength of electric field for a given charge. Using Stokes’ theorem
and the same argument used in the previous section, one can deduce the differential form
of Ampère’s law

∇ × B = µ0 j . (2.22)

Ampère’s law has been experimentally tested in many situations. However, it has an
important limitation. If one take the divergence of each side the differential equation,
one finds that ∇ · j = 0 since ∇ · ∇ × B = 0, therefore, for consistency with the continuity
equation we require ρ̇ = 0, that is, Ampère’s law only applies in situations where the
charge density is time independent.
 2.2 Integral forms of Maxwell’s equations 7

Maxwell realised that one must introduce an extra term into Ampère’s law to ame-
liorate this problem. He thought of Ampère’s law as holding for some effective current
which is
jeff = j + jdisp (2.23)
= j + ε0 Ė, (2.24)
and the extra current, jdisp whose form can be deduced simply from Gauss’ law, was
termed the displacement current. Therefore, the integral form of the Ampère-Maxwell
law is given by
% ! ' (
B · d! = µ0 j + ε0 Ė · dS. (2.25)
L S
from which we can see that magnetic fields can be created by both a current, or a
time-varying electric field. The differential form
∇ × B − µ0 ε0 Ė = µ0 j, (2.26)
can be deduce by now familiar arguments.

2.2.6 Summary
The integral forms of Maxwells equations are
1
! !
E · dS = ρ dV, (2.27a)
S ε0 V
!
B · dS = 0, (2.27b)
S

d
E = − ΦB , (2.27c)
dt
% ! ' (
B · d! = µ0 j + ε0 Ė · dS. (2.27d)
L S
 2.2 Integral forms of Maxwell’s equations 8

From them we can deduce the differential forms of Maxwell’s equations

1
∇·E = ρ, (2.28a)
ε0

∇ · B = 0, (2.28b)

∇ × E + Ḃ = 0, (2.28c)

∇ × B − µ0 ε0 Ė = µ0 j. (2.28d)
Maxwell’s equations govern how electric and magnetic fields are sourced by charge and
current density and how they interact with each other. The two fields are infact the same
phenomenon, but for the moment it useful to note a few points:
• There are two sourceless equations ∇·B = 0 and ∇×E+ Ḃ = 0 and 2 sourced equations
∇ · E = ε10 ρ and ∇ × B − µ0 ε0 Ė = µ0 j which define how electric and magnetic fields
are created by charge and current densities.
• There are two equations which can be thought of as imposing spatial constraints on the
fields, ∇ · E = ε10 ρ and ∇ · B = 0, and two equations which govern the time evolution
of the fields Ḃ = −∇ × E and Ė = µ01"0 (∇ × B − µ0 j).
• One can derive the continuity equation from the Ampère-Maxwell law by taking the di-
vergence of both sides and substituting in Gauss’ law. This implies that the continuity
equation does not bring extra dynamical constraints on the theory.
• The measured values of ε0 = 8.854 × 10−12 F m−1 and µ0 = 4π × 10−7 H m−1 , give
µ0 ε0 = c−2 to high precision. Since one is computed from the magnitude of the
electric field created by a given charge and the other is the magnetic field created by a
particular current, there is little chance this is coincidental and instead points to some
deep connections between Maxwell’s law and relativity (which we will explore later).
 2.2 Integral forms of Maxwell’s equations 9

Fig. 2.2. Schematic of the electric field in the regimes where (a) ∇ · E > 0 which could correspond
to the electric field around a positive charge, (b) ∇ · E = 0 an example of which could be the
field through a capacitor and (c) ∇ · E < 0 which could correspond to the field around a negative
charge.

WEEK 3 SUPPORTING TEXTS:

Grant & Phillips; Sections 3.1, 3.2 & 4.2 → 4.4
Griffiths; Chapters 2 & 5

2.2.7 Time independent form of Maxwell’s equations

If all fields in Maxwell’s equations are independent of time, that is, all time derivatives
are zero, then Maxwell’s equations become
ρ
∇·E= , ∇ × E = 0, (2.29a)
ε0

∇·B=0 ∇ × B = µ0 j. (2.29b)
Note that E and B have decoupled and have no influence on each other. The first two
equations are those of electrostatics, and the second two are those of magnetostatics. The
fact that ∇ × E = 0 is often termed as E being irrotational. In addition, the continuity
equation implies that ∇ · j = 0 which implies that the current flow is incompressible.
Electric fields satisfy
Q
! !
E · dS = , E · d! = 0, (2.30)
S ε0 L

or alternatively the differential forms ∇ · E = ρ/ε0 and ∇ × E = 0. From these facts and
the patterns we expect for different values of the divergence of a vector field (section 1),
we can deduce that electric field lines end on point charges. Various situations of interest
are shown in fig. 2.2.
Magnetic fields satisfy
! " !
B · d! = µ0 Ii B · dS = 0, (2.31)
L S
 2.2 Integral forms of Maxwell’s equations 10

Fig. 2.3. The magnetic field due to a current carrying wire.

or alternatively the differential forms ∇ · B = 0 and ∇ × B = µ0 j. The mathematical

statement that the divergence of the magnetic field is zero implies that magnetic field
lines never end, that is, there are no magnetic poles. Intuitively, this means that the
number of magnetic field lines going into a region is identical to the number coming out
of a region. A typical magnetic field configuration is illustrated in Figure 2.3
From section 1 ∇ × ∇φ = 0 for all φ and ∇ · (∇ × v) = 0 for all v. Therefore, if we set
B = ∇ × A, E = −∇φ, (2.32)
then the sourceless Maxwell’s equations, ∇ · B = 0 and ∇ × E = 0, are automatically
satisified. φ is the scalar electrostatic potential and A is the vector magnetic potential.
We note that the definition of A is not unique in the sense that A! = A + ∇ψ gives the
same magnetic field for any choice of ψ. This extra, so called gauge degree of freedom
can be removed by the choice of the Coulomb gauge,
∇ · A = 0, (2.33)
which we will discuss further in a later section. Substituting these relations into the
sourced Maxwell equations, one finds
ρ
∇2 φ = − , ∇2 A = −µ0 j. (2.34)
ε0
These are both Poisson equations with 4πg = ρ/ε0 and 4πg = µ0 j respectively.
The scalar and vector potentials have physical interpretations. We define the potential
difference,
dV = φ (r + d!) − φ(r) = d! · ∇φ = −d! · E, (2.35)
and thus the potential in going from A → B is found via
# B # B
∆VA→B = dV = − E · d!. (2.36)
A A
 2.3 Electostatics and magnetostatics 11

One can also substitute the expression for B into that for the magnetic flux, φB and
deduce that
# # #
φB = B · dS = ∇ × A · dS = A · d!. (2.37)
S S L

This is invariant under the transformation A! = A + ∇ψ since the line integral of ∇ψ

around a closed curve is zero.

2.3 Electostatics and magnetostatics

2.3.1 Electrostatics
We have seen that in the time independent, or static, limit that the scalar potential φ
satisfies
ρ
∇2 φ = − , (2.38)
ε0
which can be solved
1 ρ(r! )
#
φ(r) = dV ! . (2.39)
4πε0 V |r − r! |
The electric field can be computed by taking the gradient,

1 ρ(r! ) (r − r! )
#
E(r) = −∇φ = dV ! . (2.40)
4πε0 V |r − r! |3
The integral for the potential sums up the contributions of the infinitesimal charge dq =
ρdV ! at position r! in dV ! and weights it by the inverse distance between the point at
which one is computing the field, r, and the position r! . This is illustrated in Figure 2.4.
The integral for E performs a similar calculation but the weighting is the inverse square
distance and there is a vectorial sum of the field from all the points in the volume.
We can take the charge distribution to be a set of point charges. This is modelled by
a sum of delta-functions,
"
ρ(r) = qi δ (3) (r − ri ) , (2.41)
i

where ri is the location of charge qi . The property of the δ-function that

#
dV ! δ (3) (r! − r)F (r! ) = F (r), (2.42)
V

makes the electrostatic potential and electric field integrals straight-forward to evaluate.
 2.3 Electostatics and magnetostatics 12

(r )

r-r
r

r
O

Fig. 2.4. Geometry for calculating the electric field at a point r due to a charge distribution which
is only non-zero inside some (shaded) volume. The integral (2.40) sweeps over the shaded region
picking up the value of the charge density at every location weighted by the inverse distance from
the point to r.

They are
1 " qi
φ(r) = , (2.43)
4πε0 |r − ri |
i
1 " qi
E(r) = ŝi (2.44)
4πε0
i
|r − ri |2
1 " qi (r − ri )
= 3 , (2.45)
4πε0 |r − r i |
i

where ŝi (si = r − ri ) is a unit vector from the location where the electric field is to be
evaluated, r, to the location ri of the ith charge (which has charge qi ).
Consider a system with two point charges. The electric field at r due to a point charge
q2 is given by
1 q2 (r − r2 )
E2 (r) = . (2.46)
4πε0 |r − r2 |3
This can be used to compute the force on charge 1 due to charge 2
q1 q2 r 1 − r 2
F12 = q1 E2 (r1 ) = = −F21 . (2.47)
4πε0 |r1 − r2 |3
Hence, we have shown that Coulomb’s law is a consequence of Maxwell’s equations in
the time independent limit.

2.3.2 Magnetostatics
In the time independent limit, the magnetic vector potential satisfies
∇2 A = −µ0 j, (2.48)
 2.3 Electostatics and magnetostatics 13

which is solved, in a similar way to the electrostatic potential, by

µ0 j(r! )
#
A(r) = dV ! . (2.49)
4π V |r − r! |
Thus, in analogy with the electostatic case, when we specify the current density j we can
integrate to find the magnetic vector potential, and thus the magnetic field,
B(r) = ∇ × A(r) (2.50)
$ %
µ0 j(r! )
#
= dV ! ∇ × (2.51)
4π V |r − r! |
$ %
µ0 1
#
!
= dV ∇ × j(r! ) (2.52)
4π V |r − r! |
µ0 j(r! ) × (r − r! )
#
= dV ! . (2.53)
4π V |r − r! |3
Note, the gradient operator ∇ only acts on the unprimed r coordinates and hence j(r! )
is constant and we have used the fact that ∇ × (ψv) = ∇ψ × v + ψ∇ × v.
For a length of wire, d! carrying current I, then
jdV = Id!, (2.54)
so that the expression for magnetic field becomes
µ0 I d! × (r − r! )
#
B(r) = , (2.55)
4π |r − r! |3
which is the Biot-Savart law.
There is an analgoue of Coulomb’s law in the case of current carrying loops. Its
derivation in the general case is quite involved, so before attempting it we will consider
the simpler case of two parallel, infinite wires a distance d apart, carrying currrents I1
and I2 respectively as illustrated in Figure 2.5. The force acting on an infinitesimal
vectorial line element d! = dlẑ is given by
F12 = I1 d! × B2 (r1 ), (2.56)
where B2 (r1 ) is the magnetic field created by wire 2 evaluated at a point on wire 1. Since
the wires are infinite, only line segments which are diametrically opposite have any effect
with those either side cancelling each other out. Using Ampère’s law one finds that the
magnetic field due to wire 2 is given by
µ0 I2
B2 (r) = θ̂, (2.57)
2π|r − r2 |
where r2 is a point on wire 2 and hence
µ0 I1 I2
dF12 = d' ẑ × θ̂. (2.58)
2πd
 2.3 Electostatics and magnetostatics 14

d!1 = d!ẑ

I1 I2

Fig. 2.5. Two inifinite wire with current I1 and I2 flowing in the z-direction. If I1 and I2 are
both positive. there is an attractive force per unit length acting between the wires.

r1 r2

I2
d!2
I1
d!1

Fig. 2.6. Two loops with current I1 and I2 flowing, respectively.

Since ẑ × θ̂ = −r̂, we deduce that, if I1 , I2 > 0, then there is a attractive force per unit
length of

dF12 µ0 I1 I2
= . (2.59)
d' 2πd
Now let us consider two general loops, as illustrated in Figure 2.6. The force on loop
1 due to the magnetic field from loop 2 is given by
!
F12 = I1 d!1 × B2 (r1 ), (2.60)
 2.3 Electostatics and magnetostatics 15

where, by the Biot-Savart law,

µ0 I2 d!2 × (r1 − r2 )
!
B2 (r1 ) = . (2.61)
4π |r1 − r2 |3
Hence, the force is
$ %
µ0 I1 I2 d!2 × (r1 − r2 )
! !
F12 = d!1 × . (2.62)
4π |r1 − r2 |3
Now
d!1 × (d!2 × (r1 − r2 )) (r1 − r2 ) d!2 (d!1 · (r1 − r2 ))
3
= −d!1 · d!2 3
+ , (2.63)
|r1 − r2 | |r1 − r2 | |r1 − r2 |3
and the second term is a perfect differential around a closed loop which can be set to
zero. Hence, we can deduce that
µ0 I1 I2 r1 − r2
! !
F12 = − d!1 · d!2 . (2.64)
4π |r1 − r2 |3
 2.4 Calculating E and B fields 16

WEEK 4 SUPPORTING TEXTS:

Grant & Phillips; Gauss 1.4, Ampère 4.4
Electric dipole 1.5.3, Magnetic dipole 4.3
Griffiths; 2.1.4 (example 2.8), 5.2 (see examples)
Electric dipole 3.4.4

2.4 Calculating E and B fields

In this section we will derive expressions for the electric and magnetic fields in various
static situations. We will use the general integral solutions derived in the previous section
on electostatics and magnetostatics. In the first two cases there is also a symmetry which
facilitates the use of Gauss’ law, or Ampère’s law. This is much quicker than the direct
integration, but only works in specialized circumstances (whereas the direct integration
can be used in any situation). Of course, it is not always possible to compute the integrals
in closed form.

2.4.1 Electric field due to spherically symmetric charge distribution

Consider a spherically symmetric distribution of charge. This means that the charge
density ρ depends only on r, and not any of the other polar coordinates:

ρ(r) −→ ρ(r). (2.65)

We will compute the electric field by two methods. First via the potentials (i.e. calculate
φ and then find E = −∇φ), and then via Gauss’ law.
One can define the spherical polar coordinate system (r! , θ! , φ! ) so that the angles θ!
and φ! are measured relative to the vector r, that is, the north pole is chosen to be in
the direction of r. We can then compute the potential by
! ∞ ! π ! 2π
1 !2 ! ! ! ρ(r! )
φ(r) = r dr sin θ dθ dφ!
4πε0 0 0 0 (r2 + r!2 − 2rr! cos θ! )1/2
! ∞ ! π
1 ρ(r! )
= 2π r!2 dr! sin θ! dθ!
4πε0 0 0 (r2 + r!2 − 2rr! cos θ! )1/2
"# $1/2 %π
r2 + r!2 − 2rr! cos θ!
! ∞
1 !2 ! !
= r dr ρ(r )
2ε0 0 rr!
0
! ∞
1 ! ! !
# ! !
$
= dr r ρ(r ) r + r − |r − r |
2ε0 r 0
& ! r ! ∞ '
1 1 ! !2 ! ! ! !
= dr r ρ(r ) + dr r ρ(r ) , (2.66)
ε0 r 0 r
 2.4 Calculating E and B fields 17
ẑ

r
(x, y, z)

θ̂

Fig. 2.7. The set-up used in the calculation of the magentic field due to an infinite current carrying
wire.

and the electric field is given by

( r )
∂φ 1
!
E = −∇φ = − r̂ = dr! r!2 ρ(r! ) r̂. (2.67)
∂r ε0 r 2 0

This is quite a lengthy and involved calculation. If we use Gauss’ law then we must
think a little about the symmetry of the situation, and choose a sensible’ Gaussian
surface over which to integrate. By symmetry we can deduce that E = Er r̂ and then
choose a spherical surface with radius r,
Q(r) 4π r ! !2 !
! !
2
E · dS = 4πr Er = = dr r ρ(r ), (2.68)
ε0 ε0 0
where Q(r) is the charge inside a sphere of radius, r. Therefore,
! r
1
Er = dr! r!2 ρ(r! ), (2.69)
ε0 r 2 0
which was arrived at through much simpler algebra.

2.4.2 Magnetic field due to current-carrying wire

Let us compute the magnetic field generated by a wire (in the ẑ-direction) carrying
current I as illustrated in Figure 2.7. The wire must be infinitely long, otherwise edge
effects would complicate the calculation. We will first use the Biot-Savart law and find
that it is a complicated task and eventually use Ampère’s law to do the same calculation
in a fraction of time. Nonetheless, it serves as an illustration of how to apply the Biot-
Savart law in more general circumstances.
First, we define r = (x, y, z) which is the location at which we want to find the value
 2.4 Calculating E and B fields 18

of the magnetic field. We have a position vector inside the current distribution,
r! = (0, 0, z ! ), d! = (0, 0, 1)dz ! , (2.70)
where −∞ < z ! < +∞. The Biot-Savart law for this geometry becomes
µ0 I (0, 0, 1)dz ! × (x, y, z − z ! )
!
B(r) = (2.71)
4π (x2 + y 2 + (z − z ! )2 )3/2
! ∞
µ0 I dz !
= (−y, x, 0) 3/2
. (2.72)
4π −∞ (x2 + y 2 + (z − z ! )2 )

In order to perform the integral, we make the substitution

z − z! −dz !
* = tan α ⇔ * = sec2 α dα. (2.73)
x2 + y 2 x2 + y2
Hence,
π/2
µ0 I (−y, x, 0)
!
B(r) = cos α dα, (2.74)
4π x2 + y 2 −π/2

which is now easily integrated to give

µ0 I (−y, x, 0)
B(r) = . (2.75)
2π x2 + y 2
We now notice the presence of the azimuthal unit vector θ̂ = (−y/r, x/r, 0) or =
(− sin θ, cos θ, 0), and arrive at our final answer. The magnetic field due to a current
carrying wire is
µ0 I
B(r) = θ̂. (2.76)
2πr
To use Ampère’s law to do the same calculation one uses the azimuthal symmetry of
the wire, which implies B = Bθ θ̂, to guide the choice of curve. We choose a circle around
the enclosing the wire of radius r and then the magnetic field is just
!
B · d! = Bθ 2πr = µ0 I. (2.77)

Therefore, via Ampère’s law, the magnetic field around a current carrying wire is in
obvious agreement with the magnetic field calculated via the Biot-Savart law.

2.4.3 Charged circular loop

Now let us consider a system where one cannot use Gauss’ law due to the absence of
an appropriate symmetry: the case of a charge circular loop of radius R and charge per
unit length λ which is illustrated in Figure 2.8. We will place the loop in the x − y plane
 2.4 Calculating E and B fields 19
ẑ
(x, y, z)

Fig. 2.8. The set-up used in the calculation of the electric field due to a circular loop of charge.

and centre it at the origin. Adopting a cylindircal polar coordinate system, the charge
density, ρ(r! ) is zero except when r! = R and z ! = 0 and therefore it is given by

ρ(r! , θ! , z ! ) = λδ(r! − R)δ(z ! ). (2.78)

We will use a Cartesian system for r = (x, y, z) and a cylindrical coordinate system for
r! = (r! cos θ! , r! sin θ! , z ! ) and hence
2
|r − r! |2 = x2 + y 2 + r! − 2xr! cos θ! − 2yr! sin θ! + (z − z ! )2 . (2.79)

Therefore, the electric field is given by

 
∞ ∞ 2π x − r! cos θ!
1 λδ(r! − R)δ(z ! )
! ! !
E(x, y, z) = dz ! r! dr! dθ!  y − r! sin θ!  .(2.80)
4πε0 −∞ 0 0 [F (x, y, z, r! , θ! , z ! )]3
z − z!

where F (x, y, z, r! , θ! , z ! ) = (x2 + y 2 + r! 2 − 2xr! cos θ! − 2yr! sin θ! + (z − z ! )2 )1/2 . The r!

and z ! integrations can be performed readily using the properties of the δ-function and
hence we deduce that
 
! 2π x − R cos θ!
λR 1
E(x, y, z) = dθ!  y − R sin θ!  . (2.81)
4πε0 0 [F (x, y, z, R, θ! , 0)]3
z

It is difficult to compute a closed form for the θ! -integral, but it can be done for points
on the axis of the loop, that is when x = y = 0 for which one obtains
 
0
λRz
E(x, y, z) = 0 . (2.82)
2ε0 (R2 + z 2 )3/2
1
 2.5 Electric & magnetic dipoles 20

r! r

Fig. 2.9. The set-up used in calculating the electric field due to a localized distribution of charge
at some point r which is a long way distant from the charge.

2.5 Electric & magnetic dipoles

We have seen in the previous section that one can calculate the electric and magnetic
fields in specific geometries. However, you will have noticed that it can be quite diffi-
cult. Often one will make approximations at large distance, which can yield important
information, and this is the topic for the end of this section.
First, consider the electrostatic potential and electric field at some point r which is
a long way away from some localized distribution of charge as illustrated in Figure 2.9.
This implies that |r − r! | ≈ r, where r! is a coordinate inside the charge distribution and
hence
1 !
! ρ(r ) 1 Q
! !
φ(r) = dV ≈ dV ! ρ(r! ) = , (2.83)
4πε0 |r − r |! 4πε0 r 4πε0 r
1 ! !
! (r − r ) ρ(r ) 1 r Q
! !
E(r) = dV 3 ≈ 3
dV ! ρ(r! ) = 2
r̂. (2.84)
4πε0 |r − r |! 4πε 0 r 4πε 0r

Therefore, we see that in the far-field approximation, the potential/field due to an ex-
tended, but localized charge distribution looks like the potential/field of a point particle
whose charge is the integrated charge density,
!
Q = dV ρ(r). (2.85)

We will define the electric dipole moment about r to be

!
# $
p(r) = dV ! r! − r ρ(r! ). (2.86)

It is a vector quantity with units [C m].

Let us consider two point charges, q1 = +q and q2 = −q, at positions r1 = +a and
r2 = −a respectively. The charge density for this configuration can be represented by
two delta-functions
/ $0
ρ(r! ) = q δ (3) r! − a − δ (3) r! + a ,
# $ #
(2.87)
 2.5 Electric & magnetic dipoles 21

and using this one can compute the electric dipole moment,
!
$ / $0
dV ! r! − r q δ (3) r! − a − δ (3) r! + a
# # $ #
p(r) = (2.88)
= q [(a − r) − (−a − r)] = 2qa. (2.89)
Therefore, denoting d = 2|a| – the separation between the charges – we have
|p| = p = qd. (2.90)
For a typical molecular separation of 0.1 nm the magnitude of the electric dipole moment
is 1.6 × 10−29 C m.
The electrostatic potential due to these point charges, and the electric field are given
by
( )
q 1 1
φ(r) = − , (2.91)
4πε0 |r − a| |r + a|
( )
q r−a r+a
E(r) = 3
− . (2.92)
4πε0 |r − a| |r + a|3
Now,
*
|r − a| = r2 + a2 − 2r · a (2.93)
1
a 2 2r · a
= r 1+ 2 − 2 , (2.94)
r r
and, expanding as a Taylor series in a/r, which corresponds to finding the fields at a
distance from the dipole which is much greater than the separation between the charges
of the dipole, one can deduce that
1 1/ r·a 0 1 1 / r·a 0
= 1 + 2 + ··· , = 1 + 3 + · · · . (2.95)
|r − a| r r |r − a|3 r3 r2
Therefore, inserting these into the expressions for the electrostatic potential,
q 2 r·a / r·a 03
φ(r) = 1 + 2 + ··· − 1 − 2 + ··· (2.96)
4πε0 r r r
q p · r̂
≈ 2r · a = . (2.97)
4πε0 r3 4πε0 r2
Similarly for the electric field,
1
E(r) ≈ [3 (r̂ · p) r̂ − p] . (2.98)
4πε0 r3
As a simple example let us take a dipole aligned along the z-axis where the charges
are separated by d as illustrated in Figure 2.10, that is
d
a= (0, 0, 1), p = qd(0, 0, 1), (2.99)
2
 2.5 Electric & magnetic dipoles 22

Fig. 2.10. Two charges (separated by a distance d) in the z-direction as used in the example of
an electric dipole.

and use
r̂ = (sin θ, 0, cos θ). (2.100)
This means that we are considering the plane y = 0.
One then can easily calculate that in the plane y = 0 we have
 
3 sin θ cos θ
qd cos θ qd 
φ= 2
, E= 0 . (2.101)
4πε0 r 4πε0 r3 2
3 cos θ − 1

Note that for the dipole φ ∝ 1/r2 and E ∝ 1/r3 whereas for a point charge, φ ∝ 1/r
and E ∝ 1/r2 . This is because the total charge of the dipole is zero,
!
Q = dV ρ = 0, (2.102)

and hence the terms which are ∝ 1/r in φ and ∝ 1/r2 in E are zero.
One can perform similar calculations in the case of magnetic fields but it a little more
involved and hence we will just state the main results. The magnetic dipole moment is
I
4
# ! $
m(r) = r − r × d!, (2.103)
2
which can be used to show that
µ0 m × r̂ µ0 1
A(r) = , B(r) = [3 (r̂ · m) r̂ − m] . (2.104)
4π r2 4π r3
 2.5 Electric & magnetic dipoles 23

Fig. 2.11. On the left is representation of the electric field lines due to two point charges and on
the right is the same for the magentic field lines created by a cricular loop carrying a current.
In both cases the field lines have the same dipolar pattern at long distances, but they are very
different near the sources.

For a simple closed loop, the amplitude of the magnetic dipole moment is the
current multiplied by the area of the loop and its direction is perpendicular to the
loop and the sense is given by the right-hand screw rule for the current.
 2.6 Structure of Maxwell’s equations 24

2.6 Structure of Maxwell’s equations

WEEK 5 Part i) SUPPORTING TEXTS:
Grant & Phillips; Chapter 14
Griffiths; Chapter 12 (focus on 12.3 up to 12.3.2)

2.6.1 Potential formulation

Rather than work with the vector fields E, B which (collectively) have 6 components,
we can write Maxwell’s equation in terms of a scalar φ and single vector A field (which
only has four degrees of freedom). This provides us with a much simpler set of equations
and also hints at the true degrees of freedom in electromagnetism. We call φ the electric
potential and A the magnetic vector potential (or just the magnetic potential).
To show how this is possible, recall the following identities from vector calculus
∇ × ∇f = 0, (2.105)
∇ · ∇ × v = 0, (2.106)
which hold for all scalar fields f and vector fields v. Now consider the source-free Maxwell
equations,
∇ · B = 0, (2.107)
∇ × E + Ḃ = 0. (2.108)
If we write
B = ∇ × A, (2.109)
then (2.107) is satisfied automatically due to the vector identity (2.106). Using (2.109)
in (2.108) we find that
! "
∇ × E + Ȧ = 0, (2.110)

which is satisfied by choosing E + Ȧ = −∇φ, that is,

E = −Ȧ − ∇φ, (2.111)
where φ is a scalar field, which we call the electric potential (the minus sign is chosen by
convention).
As in the static case, the choice of φ and A is not unique. To show this, if we transform
φ, A according to
φ −→ φ! = φ − Ψ̇, (2.112a)
A −→ A! = A + ∇Ψ, (2.112b)
 2.6 Structure of Maxwell’s equations 25

where Ψ is an arbitrary scalar field, then the electric and magnetic fields are unchanged
implying that a family of solutions for φ and A exist. We can convince ourselves that
this is true for the electric field,
! " ! "
E! = −Ȧ! − ∇φ! = − Ȧ + ∇Ψ̇ − ∇ φ − Ψ̇ = −Ȧ − ∇φ = E, (2.113)

and one can perform a similar calculation for the magnetic field.
This is an example of gauge freedom, and it is “controlled” by imposing a condition
on φ and A that links them so that unique electric and magnetic fields are found. This
is called fixing the gauge and a popular choice is the Lorenz gauge†
1
φ̇ + ∇ · A = 0. (2.114)
c2
When φ is independent of time, this gauge condition becomes ∇ · A = 0, which is called
the Coulomb gauge.
Let us presume that the Lorenz gauge is not satisfied for some φ and A, then if we
perform a gauge transformation (2.112), we obtain
# $
1 ! ! 1 1 2
φ̇ + ∇ · A = 2 φ̇ + ∇ · A − Ψ̈ − ∇ Ψ . (2.115)
c2 c c2
If we now choose Ψ to be a solution of
1 1
2
Ψ̈ − ∇2 Ψ = 2 φ̇ + ∇ · A, (2.116)
c c
which is (in principle) a solvable wave equation since the RHS of the equation is just
a known forcing term, then it is always possible to find a Ψ so that the Lorenz gauge
condition is satisfied.
We define the electric and magnetic fields in terms of a scalar and vector field, according
to
B = ∇ × A, E = −Ȧ − ∇φ. (2.117)
We can now insert these into the sourced Maxwell equations, that is, Gauss’ law and the
Ampère-Maxwell law. To begin with, let us take Gauss’ law and replace the electric field
E with the scalar and magnetic potentials using (2.117)
! " ∂
∇ · E = −∇ · ∇φ + Ȧ = −∇2 φ − ∇ · Ȧ = −∇2 φ − ∇ · A. (2.118)
∂t
We can use the Lorenz gauge condition (2.114) to provide an expression for ∇ · A in
terms of φ̇, so that
1 ∂2φ 1
∇ · E = −∇2 φ + = ρ. (2.119)
c2 ∂t2 ε0
† This is not a spelling mistake: the Lorenz gauge and Lorentz transformation were invented by different people!
 2.6 Structure of Maxwell’s equations 26

Similarly with the Ampère-Maxwell law,

! "
∇ × B − µ0 ε0 Ė = ∇ × ∇ × A + µ0 ε0 ∇φ̇ + Ä
! "
= ∇(∇ · A) − ∇2 A + µ0 ε0 −c2 ∇(∇ · A) + Ä
1
= −∇2 A + Ä
c2
= µ0 j. (2.120)

If we define the differential “wave” operator,

1 ∂2
!= − ∇2 , (2.121)
c2 ∂t2

then (2.119) and (2.120) become

ρ
!φ = , !A = µ0 j, (2.122)
ε0

which are two sourced wave equations. Therefore, we have shown that by introducing the
potentials φ and A, Maxwell’s equations become two sourced wave equations. The scalar
potential φ is sourced by the charge density ρ, and the vector potential A is sourced by
the current density j. Notice that the speed of these waves is c (the speed of light).

2.6.2 Lorentz transformations*

A Lorentz transformation in the z-direction, between two frames S and S ! , can be im-
plemented by performing the coordinate transformation
! vz "
t! = t! (t, z) = γ t − 2 , (2.123a)
c
z ! = z ! (t, z) = γ (z − vt) , (2.123b)
!
# $
vz
t = t(t! , z ! ) = γ t! + 2 , (2.123c)
c
z = z(t! , z ! ) = γ z ! + vt!
% &
(2.123d)

with the x and y coordinates remaining unchanged, and where the Lorentz factor γ is
defined as usual by
$−1/2
v2
#
γ= 1− 2 . (2.124)
c
 2.6 Structure of Maxwell’s equations 27

By the chain rule we can compute the partial derivative operators in the “primed” frame
in terms of quantities in the “unprimed” frame,
# $
∂ ∂t ∂ ∂z ∂ ∂ ∂
= + = γ + v , (2.125a)
∂t! ∂t! ∂t ∂t! ∂z ∂t ∂z
# $
∂ ∂ v ∂
= γ + . (2.125b)
∂z ! ∂z c2 ∂t
We can combine these expressions to find, for example, how the wave operator (2.121)
transforms under a Lorentz transformation. The wave operator in the “primed” frame is

1 ∂2 ∂2
!! = −
c2 ∂t!2 ∂z !2
∂ 2 v ∂ 2
# $ # $
1 2 ∂ 2 ∂
= γ +v −γ +
c2 ∂t ∂z ∂z c2 ∂t
γ2 ∂2 ∂2 2
# 2
v ∂2 v2 ∂ 2
# $ $
2 ∂ 2 ∂
= + 2v +v −γ +2 2 +
c2 ∂t2 ∂t∂z ∂z 2 ∂z 2 c ∂z∂t c4 ∂t2
v2 ∂ 2 v2 ∂ 2
' # $ # $ (
2 1
= γ 1− 2 − 1− 2
c2 c ∂t2 c ∂z 2
1 ∂2 ∂2
= 2 2
− 2, (2.126)
c ∂t ∂z
that is,

!! = !, (2.127)

which means that the wave operator ! is invariant under Lorentz transformations.
Let us consider what information we can gain by performing a Lorentz transformation
on the conservation equation: if charge is conserved in one frame it is conserved in all
frames. This means that the conservation equation in the “primed” frame is still zero,
∂ρ!
+ ∇! · j! = 0, (2.128)
∂t!
and using (2.125) to express the derivative operators in terms of quantities in the un-
primed frame, this becomes

∂jx! ∂jy!
# $ # $
∂ ∂ ! ∂ v ∂
γ +v ρ + + +γ + j ! = 0. (2.129)
∂t ∂z ∂x ∂y ∂z c2 ∂t z
We now rearrange this,

∂ ! ! v " ∂ % ! & ∂j ! ∂jy!

γ ρ + 2 jz! + γ vρ + jz! + x + = 0, (2.130)
∂t c ∂z ∂x ∂y
 2.6 Structure of Maxwell’s equations 28

and compare it with the conservation equation in the unprimed frame,

∂ρ ∂jx ∂jy ∂jz
+ + + = 0, (2.131)
∂t ∂x ∂y ∂z
we can read off
! v "
ρ = γ ρ! + 2 jz! , (2.132a)
c
!
jx = jx , (2.132b)
jy = jy! , (2.132c)
jz! !
% &
jz = γ + vρ . (2.132d)
These are the “inverse Lorentz transformations”, and so one can readily construct
! v "
ρ! = γ ρ − 2 j z , (2.133a)
c
jx! = jx , (2.133b)
jy! = jy , (2.133c)
jz! = γ (jz − vρ) . (2.133d)
Therefore, if an observer O at rest sees a charge and current density (ρ, j), an observer O!
moving a speed v along the ẑ-direction relative to the first observer will see a different set
of charge and current densities, (ρ! , j! ). This means that to say whether or not a system
has current or charge is an observer dependent statement: in the same way that distance
and time intervals are dependent on the frame in which they are measured, charge and
current densities are dependent on reference frame. There is however a quantity which
can be constructed from the charge and current densities which is frame independent
(i.e. all observers will agree on its value). This quantity is
J 2 = c2 ρ2 − |j|2 = c2 ρ!2 − |j! |2 . (2.134)
By using (2.125) on the Lorenz gauge condition (2.114), one can show that the scalar
and vector potentials transform according to
φ! = γ (φ − vAz ) , (2.135a)
A!x = Ax , (2.135b)
A!y
= Ay , (2.135c)
! v "
A!z = γ Az − 2 φ . (2.135d)
c
We can now use these derived transformations to see how the wave equations (2.122)
transform under a Lorentz transformation,
ρ!
# $
! ! ρ
! φ = !γ (φ − vAz ) = γ (!φ − v!Az ) = γ − vµ0 jz = , (2.136)
ε0 ε0
 2.6 Structure of Maxwell’s equations 29

and similarly,

!! A!z = µ0 jz! . (2.137)

This means that Maxwell’s equations are Lorentz invariant and hence one could say that
Maxwell “anticipated” Einstein’s discovery of Special Relativity.

2.6.3 Lorentz transformation of electric and magnetic fields*

So far we have seen that charge and current densities change under a Lorentz transfor-
mation. Since charge and current source electric and magnetic fields, we should expect
that the E and B fields will also change under a Lorentz transformation. We will use
(2.135) to perform a Lorentz transformation on φ and A and use these expressions to
calculate the electric and magnetic fields from (2.117).
The electric field in S ! is given by
∂A!
# $
E ! = − ∇! φ ! + . (2.138)
∂t!
We now use (2.125) and (2.135) to obtain
 ∂   
∂x
# $ Ax
 γ(φ − vAz ) + γ ∂ + v ∂ 
 
∂
E! = −  ∂y A y

∂t ∂z
∂
γ( ∂z + cv2 ∂t
∂
) γ(Az − cv2 φ)
 
 # $ 
∂φ ∂Ax ∂Ax ∂Az
 ∂x + ∂t + v ∂z − ∂x 
 # $ 
 ∂φ ∂Ay ∂Ay ∂Az

= −γ   ∂y + ∂t + v ∂z − ∂y
.
 (2.139)
 # $ # $ 
2 v 2 ∂Az
γ 1 − vc2 ∂φ
 
∂z + γ 1 − c2 ∂t

In component form, this is just

' # $(
∂φ ∂Ax ∂Ax ∂Az
Ex! = −γ + +v − , (2.140a)
∂x ∂t ∂z ∂x

' # $(
∂φ ∂Ay ∂Ay ∂Az
Ey! = −γ + +v − , (2.140b)
∂y ∂t ∂z ∂y

' (
∂φ ∂Az
Ez! = − + . (2.140c)
∂z ∂t
To be able to identify the terms on the right-hand-side of these expressions with the
 2.6 Structure of Maxwell’s equations 30

electric and magnetic fields in the unprimed frame, we should remember that
∂Ay
 
∂Az
∂y − ∂z
B = ∇ × A =  ∂A ∂Az 
∂z − ∂x  . (2.141)
 x

∂Ay ∂Ax
∂x − ∂y

This enables us to realise, for example, that the term multiplying v in (2.140a) is By and
that the first two terms are Ex . Hence, we can deduce that

Ex! = γ(Ex − vBy ), (2.142a)

Ey! = γ(Ey + vBx ), (2.142b)
Ez! = Ez . (2.142c)

In a similar fashion, we use B = ∇ × A to find the components of the magnetic field

in the primed frame in terms of the components in the unprimed frame:
5 5
5 î ĵ k̂ 5
5 5
5 ∂ ∂ ∂ v ∂
B! = ∇! × A! = 5 ∂x γ( + )
5
∂y ∂z c2 ∂t 55
5 Ax Ay γ(Az − cv2 φ) 5
5

∂Ay v ∂φ v ∂Ay
 
γ( ∂A
∂y
z
− ∂z − c ∂y
2 − c ∂t
2 )
=  γ( ∂A ∂Az v ∂Ax v ∂φ  . (2.143)
− + + )
 x
∂z ∂x c ∂t
2 c2 ∂x 
∂Ay ∂Ax
∂x − ∂y

We can then read off the components of B! in terms of the components of E, B:

v
Bx! = γ(Bx + E ),
c2 y
(2.144a)
v
By! = γ(By − E ),
c2 x
(2.144b)
Bz! = Bz . (2.144c)

Equations (2.142) and (2.144) reveal that only the components of E and B in the plane
perpendicular to the Lorentz transformation are affected: Ez , Bz are unaffected.

2.6.4 Electro/magnetostatics and special relativity*

We can make connection with the physical picture in a particular case. Let us first
remember that:

(i) for a wire carrying a static line charge density λ, by using Gauss’ law the electric
field is
λ
E= r̂; (2.145)
2πε0 r
 2.6 Structure of Maxwell’s equations 31

(ii) and for a line current, by using Ampère’s law, the magnetic field is
µ0 I
B=
θ̂. (2.146)
2πr
Now recall the Lorentz transformations of the electric (2.142) and magnetic (2.144) fields,
for a Lorentz boost along the ẑ-direction:
v
Ex! = γ(Ex − vBy ), Bx! = γ(Bx + E )
c2 y
(2.147a)
v
Ey! = γ(Ey + vBx ), By! = γ(By − E ),
c2 x
(2.147b)
Ez! = Ez , Bz! = Bz (2.147c)
Let us consider a physical system where we observe a charge-carrying-wire in the rest
frame of the wire, whose electric field (in Cartesian components) is
 
cos θ
λ 
E= sin θ  , B = 0. (2.148)
2πε0 r
0
Now we transform to a frame moving at speed v in the ẑ-direction, by using the Lorentz
transformations (2.147). For the moving observer, the fields around the wire become
 
cos θ
γλ  λ!
E! = sin θ  ≡ r̂, (2.149)
2πε0 r 2πε0 r
0
 
sin θ !
γλv   ≡ µ0 I θ̂,
B! = − cos θ (2.150)
2πc2 ε0 r2 2πr
0
which implies that the charge and current densities in S !
λ! = λγ, I ! = −γvλ = −λ! v. (2.151)
This impies that a static observer sees a line charge density λ and no magnetic field,
while a moving observer sees both an electric field and a magnetic field. The wire is
Lorentz contracted, that is, lengths in the z-direction are reduced by L = L0 /γ. Since
the total charge remains constant λ = Q/L0 and hence λ! = Q/L. The magnetic field is
generated by a current I ! = −λ! v. These “new” charge and current densities also follow
from (2.133), where one can set jz = 0 and find ρ! , jz! .
Importantly, even at low velocities, where γ → 1, the “relativistic” effects are observed:
E! & E + v × B, (2.152)
! 2
B & B − (1/c ) v × E, (2.153)
!
j & j − ρv, (2.154)
ρ! & ρ − (1/c2 ) j · v. (2.155)
 PHYS 20141 ELECTROMAGNETISM : 2021/22

Section 3 : Electromagnetic effects in simple materials

P. Campbell
(for use in conjunction with the teaching material on the PHYS20141 Blackboard area)

Department of Physics & Astronomy

University of Manchester

October 20, 2021

 3
Electromagentic effects in simple materials

WEEK 5 Part ii) SUPPORTING TEXTS:

Grant & Phillips; Section 3.3 (images), 4.1 (conductors)
Griffiths; Section 2.5 (conductors), 3.2 (images)

So far we have studied Maxwells equations in a vacuum and we have solved for the
electric and magnetic fields in various simple configurations of charge and/or current. We
have seen that these calculations can become very difficult even for apparently simple
configurations – for example, the cases of the off-axis fields due to circular loops of
charge and current. One mole of a material contains 6 × 1023 atoms or molecules and
each contains positive (protons) and negative (electrons) charges in non-uniform charge
distributions which are often moving, creating currents. In this section of the course, we
will see that these charges can lead to complicated collective phenomena which, while
small at the atomic scale, can produce significant macroscopic effects. We will show how
these phenomena can be modelled within Maxwell’s equations.

3.1 Conductors
In a conductor a significant fraction of its electrons are “free to move” - these are the so
called conduction electrons. This means that the resistance, R, to the flow of a current is
very low and the corrreponding conductivity, σ, is very high. Ohm deduced an empirical
relationship linking the current I, voltage V and resistance, R,
1 1
!
I= V = E · d!. (3.1)
R R
We define the conductivity, σ, by
1
σdS = d!, (3.2)
R
where σ has units [Ω−1 m−1 ]. Substituting this relation into Ohm’s law and using the
definition of the current, this enables us to deduce a relationship between the electric
field, conductivity and current

j = σE, (3.3)

2
 3.2 Method of images 3
charge + q

Fig. 3.1. Schematic setup for the method of images. A charge q sits a distance a above an earthed
conducting plane

which is the form of Ohm’s law which we can use in the context of Maxwell’s equations.
We will now use this expression for the current in the conservation equation, ρ̇+ ∇ ·j =
0, and Gauss’ law, ∇ · E = ε10 ρ, to show that
σ
ρ̇ + ∇ · j = ρ̇ + σ∇ · E = ρ̇ + ρ = 0. (3.4)
ε0
This equation can be integrated to give

ρ(r, t) = ρ(r, 0)e−t/tR , (3.5)

where we defined tR = ε0 /σ which can be interpreted as a relaxation time scale for

the movement of the conduction electrons to smooth out non-uniformities in the charge
density. Notice that for good conductors, σ is very high so that tR is very small. Typical
values are
for a metal with σ ≈ 107 Ω−1 m−1 ,
"
8 × 10−19 s
tR ≈ (3.6)
8 × 103 s for an insulator with σ ≈ 10−15 Ω−1 m−1 .

3.2 Method of images

The method of images exploits the uniqueness theorem. The solution to a set of differen-
tial equations under a given set of boundary conditions is the solution, regardless of the
physical motivation for the boundary conditions. Specifically in the method of images
we use the fact that at a perfect conductor the electric potential is zero, which is also
the case at a location half-way between a charge and anti-charge.
Consider a point charge a distance a above an infinite perfectly conducting plane in
the x, y-plane at z = 0. The plane is earthed so that φ(z = 0) = 0. See Figure 3.1.
We want to solve Poisson’s equation for φ, subject to the condition φ(x, y, 0) = 0 –
note that E = 0 inside the conductor. The method of images provides a way of solving
this problem. One imagines that the plane contains an “image charge” at z = −a with
 3.2 Method of images 4

charge −q, and one ignores the plane. Then, the potential due to this charge– anti-charge
pair is thus
# $
q 1 1
φ(r) = − (3.7)
4πε0 |r − a| |r + a|
% '
q 1 1
= & −& . (3.8)
4πε0 x2 + y 2 + (z − a)2 x2 + y 2 + (z + a)2
If we now set z = 0 in this expression,
φ(x, y, 0) = 0, (3.9)
which is exactly the boundary condition of the charge above the conducting plane. Hence,
the solution satisfies the boundary condition, and by the uniqueness theorem it is the
only solution.
The physical picture, for a positive charge +q, the conduction electrons in the slab are
attracted to the surface of the conductor and their overall effect is equivalent to a charge
−q at z = −a. (Infact there is a surface charge density induced which can be computed
using Gauss’ law.)
There is a force on the charge due to the “image charge”:
F = qEimage (r = a), (3.10)
where
q r+a
Eimage = − (3.11)
4πε0 |r + a|3
 3.3 Capacitance, relative permittivity & dielectrics 5

+Q

d E Gaussian surface Voltage, V

−Q

Fig. 3.2. Parallel plate capacitor with the plates separated by a distance d which have been
charged by a power supply with potential difference ∆V . The electric field is in the z-direection
and will be directed from the positively charged plate to the negative.

WEEK 7 SUPPORTING TEXTS:

Grant & Phillips; Sections 2.1, 2.2 (2.2.2) and start of 2.3
Griffiths; Section 4.1, focus on 4.1.2 and 4.1.3 then 4.2 and start of 4.3

3.3 Capacitance, relative permittivity & dielectrics

Consider a parallel plate capacitor of two plates separated by a distance d. The plates
are connected to a power supply with potential difference ∆V , until the plates carry
charges +Q and −Q respectively. The power supply is disconnected leaving the poten-
tial difference between the plates. If we draw a Gaussian surface around the plates as
illustrated in Figure 3.2, then we can determine the electric field under the assumption
that there is a vacuum between the plates. Gauss’ law gives
Q
!
E · dS = Ez A = , (3.12)
ε0
where A is the surface area of the plates, and we have aligned the system along the z-axis
(that is the plates lie in the x − y-plane). Hence we deduce a constant electric field
Q
Ez = . (3.13)
Aε0
The potential difference between the top and bottom plates is
Qd
∆V = Ez d = . (3.14)
Aε0
We define capacitance to be the ratio between the charge and potential difference,
Q ε0 A
C= = , (3.15)
∆V d
whose unit is the Farad, [F]. This is a property of the system which is only dependent
on the geometry of the system, in this case, only on A and d. We can also compute the
 3.4 Polarization 6

Substance εr
Air (1atm) 1.00059
Air (100atm) 1.0548
Teflon 2.1
Polyethylene 2.25
Mylar 3.1
Glass 5 – 10
Water 80.4
Strontium Titanate 310

Table 3.1. Typical values of the relative permittivity (dielectric constant) for some
materials. Note that these values are dependent on enviromental properties such as
density, pressure and temperature.

energy of a parallel plate capacitor,

∆V 2
" #
1 1 1
!
2
U = ε0 |E| dV = ε0 Ad = C(∆V )2 . (3.16)
2 2 d 2
If we now add a material (for example, glass) between the plates, the measured voltage
is observed to drop, meaning that the capacitance increases. We define the relative
permittivity to be
C
εr = , (3.17)
Cvacuum
which is also referred to as the dielectric constant. The quantity C is the measured
value of the capacitance with some material separating the plates and Cvacuum is the
capacitance for the same geometry, that is the same physical separation and shape of
the parallel plates, but separated by a vacuum. The values of εr are listed in Table 3.1
for some typical materials
The class of materials which exhibit this kind of behaviour are known as dielectrics.
They are electrical insulators, meaning that the materials have low conductivity, and
the electrons are bound into atoms and molecules. Hence, there are few free electrons.
When an electric field is applied to a dielectric the intrinsic dipoles within the material
can become aligned, while atoms/molecules with no intrinsic dipole moment become
polarized creating dipole. We will discuss the possible mechanisms for these phenomena
in the following sections.

3.4 Polarization
Polarization of a material object occurs when the constituents of the substance align in
some preferred direction associated with an electric field. To understand how the atoms
 3.4 Polarization 7
+q F
Eext

F −q

Fig. 3.3. Dipole in an external electric field.

of a material object respond to an applied electromagnetic field we will consider the

action of a single electric dipole in an external electric field Eext as shown in Figure 3.3.
Recall that the dipole moment is p = qd.
The work done by the dipole is the energy change due to the dipole

Uext = −qEext .d cos θ, (3.18)

where θ is the angle between the dipole and the external field. The torque on the dipole
is

τext = F.d sin θ = dqEext sin θ. (3.19)

If θ = 0 the energy due to the external field is minimized and the torque is zero. There-
fore, dipoles will attempt to rotate to align with the applied field to minimize the energy
and eliminate torque.
For a general dipole, the energy due to and torque on a dipole in an applied electric
field Eext are calculated via

Uext = −p · Eext , τext = p × Eext . (3.20)

The lowest energy state is acheived by setting p ∝ Eext which corresponds to a state
with zero torque.
Therefore, if an electric field is applied to a material, two things happen:

(i) intrinsic dipoles will align to minimize energy and eliminate torques;
(ii) atoms and molecules can be polarized, inducing a dipole moment.
We define a macroscopic vector field, the polarization, P, whose units are [Cm−2 ] to
represent the combined effect. The polarization vector can be computed from

P = np, (3.21)

where n is the number density of atoms or molecules and p is the average dipole moment
 3.4 Polarization 8

due to the alignment of intrinsic dipoles and induced dipoles in the material. The polar-
ization of a material will in general be a function of the applied electric field, P ≡ P(E)
which is called a constitutive relation.
For a linear isotropic material we will take the constitutive relation to be
P = χE ε0 E. (3.22)
In index notation this reads
Pi = χ E ε 0 E i . (3.23)
The coefficient χE is the electric susceptibility, and determines the strength of the re-
sponse of the polarization properties of the material due to an applied electric field E.
This will be a property of the material and will be a function of environmental properties
of the material. We will see later that it is related to the dielectic constant, εr .
We can generalize (3.23) and introduce (a) anisotropic direction dependancies and (b)
non-linear responses. These complicated anisotropic non-linear responses can be encoded
in an expression of the form
3 3
(1) (2)
$ $
Pi = ε 0 χij Ej + ε0 χijk Ej Ek . (3.24)
j=1 j,k=1

(2) (1)
Taking the simple case where χijk = 0, if the matrix χij is diagonal and all entries are
equal then the material is isotropic then (3.23) is recovered. The off-diagonal terms
(1) (2)
in χij reflect anisotropic responses. The final term, χijk , is a tensor and contains
information about quadratic responses of the polarization vector to the applied electric
field. Including the anisotropic and non-linear responses will describe systems closer to
reality (for instance, if a substance is constructed from a regular lattice of atoms there
will be a natural set of preferred directions), but they will make the problem incredibly
complicated. For this reason we will focus on the linear isotropic response (3.23).

3.4.1 Mechanisms for polarization

We shall now return to linear isotropic materials, for which the polarization in terms of
the applied electric field is given by
P = np = χE ε0 E, (3.25)
and we will consider mechanisms for the generation of the internal polarization of the
material. There many different mechanisms but we will focus on two: one due to the
alignment of intrinsic dipoles and one where dipoles are created within a dielectric ma-
terial.
As we have already mentioned, a realistic dielectric material will have an enormously
 3.4 Polarization 9

large number of atoms (∼ 6 × 1023 atoms per mole). These atoms will be arranged in
highly complicated structures within the substance. It may therefore seem a fruitless task
to attempt to construct a model for how a dielectric material will respond to an applied
electric field. However, we can use our intuition for the behaviour of a single dipole
to build simple theoretical pictures that will enable us to understand the macroscopic
behavior of a dielectric material in the presence of an applied electric field.

3.4.1.1 Alignment
When the atoms or molecules of a dielectric material have an intrinsic dipole moment,
pint , then in the absence of an externally applied electric field there is no energy penalty
in having the dipoles have random alignments. This is shown in Figure 3.4(a). However,
when an external electric field, Eext , is applied to the dielectric material there are large
energies associated with dipoles which have large angles between the dipole moment
and the electric field, which is what we calculated in (3.18). To reduce this energy the
dipoles collectively align with the applied electric field; this is shown in Figure 3.4(b).
The polarization that this induces can be calculated†, and is given by
np2int
Palign = Eext . (3.26)
3kB T
This formula is of the form (3.23), where the electric susceptibility χE is inversely pro-
portional to the temperature.
Note:

• The largest possible field in air is ≈ 106 Vm−1 , beyond which the molecules break
down. This threshold can be a little larger in dielectrics.
• The energy at this threshold electric field is ∆Uext = 2pEext ≈ 2×10−4 eV' Uthermal =
3 −2
2 kB T ≈ 4 × 10 eV, at room temperature. This means that complete alignment will
not take place even at the very largest electric fields possible. However, the fact that
there are so many atoms means that there will be an observable polarization even at
relatively low values of the electric field.

3.4.1.2 Induced
We will now consider a material that does not have any intrinsic dipoles. When an
electric field is applied to this material, the electron “cloud” around an atom will distort
relative to the “cloud” without an applied field. This will create a dipole from a given
atom because the negative electrons are shifted relative to the positive nucleus as shown
in Figure 3.5. These induced dipoles will then align with the applied electric field, in a
manner similar to that discussed in §3.4.1.1. This mechanism is sometimes called atomic
† Grant & Phillips p 62-64
 3.4 Polarization 10

(a) No applied electric field (b) Applied electric field

Fig. 3.4. Schematic of the distribution of intrinsic dipoles within a substance. In (a) there is
no applied external electric field and the dipoles have a “random” alignment. In (b) there is an
external electric field, which has the effect of aligning the intrinsic dipoles to reduce the energy
of the dipoles.

polarization. In Figure 3.6 we give a schematic showing how the induced polarization
generates a surface charge density due to an applied electric field.
We can calculate the induced dipole moment around a given atom due to some exter-
nally applied field Eext . If the external field causes an offset d for a distribution of radius
R0 , then

q2f
qEext = (3.27)
4πε0 d2

is the force due to the offset, with f ≈ d3 /R03 is the fractional charge offset. Hence,

qd ≈ 4πε0 R03 Eext (3.28)

is the induced dipole moment due to the offset, and therefore the polarization,

Pinduced = nαε0 Eext , (3.29)

where

α ≡ 4πR03 (3.30)

is the atomic polarization.

3.4.1.3 Overall
The polarization response of a realistic dielectric material will be a combination of the
induced and alignment phenomena discussed above; which process dominates will be
different for different materials. The overall polarization response of a dielectric material
 3.4 Polarization 11

Fig. 3.5. Schematic of how an electric dipole moment can be induced by applying an external
electric field.

(a) No applied electric field

Net negative surface charge density

Net positive surface charge density

(b) Applied electric field

Fig. 3.6. Schematic of the induced polarization mechanism. In (a) we show the distribution of
positive charges inside a medium, without applied electric field. In (b) an electric field has been
applied, which moves the charge distributions, and creates a net positive/negative surface charge
density. This is the collective behavior of the single case in Figure 3.5.

is given by
P = Palign + Pinduced
np2intrinsic
= Eext + nαε0 Eext
3kB T
p2int
" #
= ε0 n α + Eext , (3.31)
3kB T ε0
 3.5 Electrostatics in dielectrics 12

which is of the form P = χE ε0 E if we identify the electric susceptibility with

p2int
" #
χE = n α + . (3.32)
3kB T ε0
One can measure the strength of the two effects exploiting the different T dependencies
but we should note that the actual dipole moments and offsets are very small.

3.5 Electrostatics in dielectrics

We will now study electric fields inside dielectric materials, and explore how the polariza-
tion induced by the external electric field affects the electric field inside the material. For
simplicity we will only consider static situations. Let us consider a polarization vector
P = Px (x)î, Px > 0. (3.33)
Electrons will move leftwards out of the dielectric material at one end and into the
material at the other end. This creates a shift in the charge, given by
∂Px
∆Q = − (P (x + δx) − Px (x)) δyδz ≈ − δxδyδz, (3.34)
∂x
but, the shift in the charge is just given by the elemental volume multiplied by the value
of the bound charge density
∆Q = ρbound δxδyδz, (3.35)
so that
∂Px
ρbound = − . (3.36)
∂x
Repeating the calculation for the other spatial directions, one can see that
ρbound = −∇ · P. (3.37)
This bound charge is neutral. The total bound charge is found by integrating the bound
charge density within the material, and adding to it the total surface charge density;
! !
Qbound = ρbound dV + σdS = 0, (3.38)
V S
where σ is the surface charge density. Thus,
! ! !
∇ · P dV = P · n̂ dS = σdS, (3.39)
V S S
where the first equality used the divergence theorem and n̂ is a unit normal of the
enclosing surface. Hence, the surface charge density due to the polarization is given by
σP = P · n̂. (3.40)
 3.5 Electrostatics in dielectrics 13

Recall the equations of electrostatics, ∇ · E = ε10 ρ, ∇ × E = 0. We shall split the charge

density ρ into two contributions: one from the bound charge ρbound and another from
the free charge ρfree . Thus, Gauss’ law reads
" # " #
1 1 1
∇·E= ρ= ρbound + ρfree = − ∇ · P + ρfree . (3.41)
ε0 ε0 ε0
A trivial rearrangement of this formula reveals that
∇ · (ε0 E + P) = ρfree . (3.42)
This can be rewritten as a modified Gauss’ law
∇ · D = ρfree , (3.43)
where we define
D ≡ ε0 E + P, (3.44)
which is called the electric displacement vector [Cm−2 ]. The integral form of Gauss’ law
becomes
!
D · dS = Qfree . (3.45)
S
Note that for a linear isotropic material, P = χE ε0 E, so that
D = (1 + χE ) ε0 E. (3.46)
 3.5 Electrostatics in dielectrics 14

+Q

d D Voltage, V

−Q

Fig. 3.7. Parallel plate capacitor with dielectric.

WEEK 8 SUPPORTING TEXTS:

Grant & Phillips; sections 2.3 then 4.3, 4.6, 4.7
Griffiths; sections 4.3, 4.4 then 6.1
We will now use these formulae to study a parallel plate capacitor with a dielectric
between the plates, as in Figure 3.7. The surface charge density of the plates is
Q
σplate = . (3.47)
A
If we now apply Gauss’ law,
!
D · dS = Dz A = Q, (3.48)

with Dz = (1 + χE ) ε0 Ez . We can then obtain the components of the electric and

polarization vectors,
Q χE Q
Ez = , Pz = . (3.49)
(1 + χE ) ε0 A (1 + χE ) A
The potential difference is
Qd
V = Ez d = , (3.50)
(1 + χE ) ε0 A
and hence the capacitance,
Q ε0 A
C= = (1 + χE ) = (1 + χE ) C0 , (3.51)
V d
where C0 is the capacitance without the dielectric. Therefore, at constant Q (i.e. with-
out the battery connected), we have calculated that the electric field is lowered by
a factor of (1 + χE ), relative to the electric field between two capacitor plates separated
by vacuum, but the capacitance increases relative to that of a vacuum separated capac-
itor. Also notice that when the dielectric is not present, χE = 0 so that the polarization
vanishes.
 3.5 Electrostatics in dielectrics 15

(a) No applied field. (b) With an applied field.

Fig. 3.8. Schematic of the behaviour of a dielectric between two parallel plates. In (a) there is no
applied field, and the intrinsic dipoles within the dielectric substance are randomly aligned. In
(b) we switch on the external field, by connecting the plates across a voltage source. This aligns
the dipoles.

We now define the relative permittivity,

εr ≡ 1 + χ E , (3.52)
and thus the electric displacement vector becomes
D = εr ε0 E. (3.53)
Note,
εr − 1
σtop = P · n̂top = − σplate , (3.54a)
εr

εr − 1
σbottom = P · n̂bottom = + σplate . (3.54b)
εr
The energy of the capacitor is
1 1 Aεr ε0 1 1
!
U = CV 2 = (Ez d)2 = AdDz Ez = D · E dV, (3.55)
2 2 d 2 2
and we note that the expression for the energy of an electric field is modified in the
presence of a dielectric.
Figure 3.8 shows a schematic of what is actually happening inside the dielectric ma-
terial. Intrinsic dipoles have a random orientation before any applied field is switched
on. When we connect the plates to a battery (i.e. to a voltage supply) the electric field
between the plates causes the dipoles to align. At non-zero temperature this alignment
may only be partial with perhaps only a slight preponderance of dipoles aligned with the
field.
 3.6 Interfaces between dielectrics 16

Region 1 Region 2 Region 1 Region 2

L
S

δ"
n̂ n̂

area δS

d d
(a) Surface S enclosing an interface (b) Loop L enclosing an interface
between dielectrics. between dielectrics.

Fig. 3.9. Interface between two different dielectrics (a) a volume with unit normals to the surface
S and (b) a loop of length δ% parallel to the boundary.

3.6 Interfaces between dielectrics

(1)
Consider the boundary between two regions 1 and 2, with relative permitivitties εr and
(2)
εr respectively, and with no free charge on the boundary.
Let S be a surface of thickness d and area δS encompassing the boundary, as in Figure
3.9(a). As d → 0, because there is no charge on the boundary,
!
(1) (2)
0 = D · dS = −D⊥ δS + D⊥ δS, (3.56)

and therefore D⊥ is continuous.

Now let L be a loop of length δ% and with d which encompasses the boundary, as in
Figure 3.9(b). As d → 0,
"
(1) (2)
0 = E · d! = −E// δ% + E// δ%, (3.57)

which implies that E// is continuous across the boundary.

We can apply these continuity arguments to the boundary shown in Figure 3.10. We
may write the electric fields in the two regions as
# $ # $
(1) sin θ1 (2) sin θ2
E = E1 , E = E2 . (3.58)
cos θ1 cos θ2
If E// is continuous then
E1 sin θ1 = E2 sin θ2 , (3.59)
 3.6 Interfaces between dielectrics 17

E(1)

Region 1 θ1

Region 2 θ2

E(2)

Fig. 3.10. An electric field across two regions of differing relative permitivitties.

and if D⊥ is continuous then

D1 cos θ1 = D2 cos θ2 . (3.60)
Hence,
D1 D2
cot θ1 = cot θ2 , (3.61)
E1 E2
and therefore,
ε(1) (2)
r cot θ1 = εr cot θ2 . (3.62)
 3.7 Inductance and relative permeability 18

3.7 Inductance and relative permeability

Consider a solenoid, of length %, carrying current I, and have N turns per unit length.
We choose an Ampereian surface at a distance r from the centre of the solenoid.
Applying Ampere’s law,
!
B% = B · d! = N %µ0 I, (3.63)

and hence the magnetic field due to the current is

B = µ0 N I. (3.64)

The magnetic flux for the solenoid is

!
ΦB = B · dS = πr2 N %B = µ0 N 2 πr2 %I. (3.65)

Considering Faraday’s law,

∂ΦB ∂I
E =− = −µ0 N 2 πr2 % . (3.66)
∂t ∂t
and we define inductance
E
L≡− = µ0 N 2 πr2 %, (3.67)
I˙
whose unit is the Henry [H].
If we add a magnetic material within the solenoid the voltage will change indicating
that the inductance has changed. We define the relative permeability,
L
µr ≡ , (3.68)
Lvac
where Lvac is the inductance in the vacuum (i.e. without the magnetic material).

3.8 Magnetization and magnetic susceptibility

Magnetization is a phenomenon akin to polarization. Just as for the polarization in
dielectric materials, we will define the magnetization,

M = nm, (3.69)

where m = IAn̂ is the average magnetic dipole moment, and A is the area of the dipole
loop. In analogy with the dielectric case, if the magnetic dipole is in an external magnetic
field Bext , the energy and torque of the dipole is

Uext = −m · Bext , τext = m × Bext , (3.70)

 3.9 Diamagnetism and paramagnetism 19

and once again the (magnetic) dipoles will align to the applied (magnetic) field to reduce
the energy and eliminate torques.
We define the magnetic susceptibility χB , and for a linear isotropic medium the mag-
netization is given in terms of the magnetic field by
χB
M= B. (3.71)
µ0
• χB < 0 corresponds to diamagnetism – equivalent to induced polarization,
• χB > 0 corresponds to paramagnetism – equivalent to alignment polarization.

Note, we here follow the Grant & Phillips definition of magnetic susceptibility (page
184). We will also consider the possibility of a nonlinear response, M = M(B), of a
ferromagnet.

3.9 Diamagnetism and paramagnetism

There are two “linear” mechanisms for producing magnetization from a magnetic field.
These mechanisms are analogous to the intrinsic and induced mechanisms we studied
for electric polarization. The two mechanisms are called diagmagnetism and paramag-
netism. There is an important non-linear mechanism, called ferromagnetism, which we
will discuss in a subsequent section.
Consider an electron orbiting a nucleus at some radius r; we will approximate the
motion to be circular. The “circular” motion of the (charged) electron means that there
is a current,
ev
I= , (3.72)
2πr
so that the magnetic dipole moment is
ev e
m= πr2 = L, (3.73)
2πr 2me
where the angular momentum is

L = me rv. (3.74)

The current flows in the opposite direction to the electrons, giving

e
m=− L. (3.75)
2me
This is the intrinsic dipole moment of an atom, due to an electron moving in a circular
orbit around a nucleus. We recall
% from quantum mechanics that the orbital angular
momentum is given by |L| = %(% + 1) !, so that the magnetic dipole moment will be
proportional to the Bohr magneton, |m| ∝ µB , where µB = e!/2me .
 3.9 Diamagnetism and paramagnetism 20

For more than one electron the moments superpose linearly,

& e &
mtot = mi = − Li . (3.76)
2me
i i
'
If i Li = 0 there is no intrinsic dipole moment.
In the absence of an external magnetic field, the electrostatic force between the nucleus
and the electron is
Ze2
F = mrω 2 = qE ⇒ me ω02 r = , (3.77)
4πε0 r2
which rearranges to give the rotational frequency,
(
Ze2
ω0 = . (3.78)
4me πε0 r3
If we now switch on an external magnetic field, remembering for circular motion that
v = rω, the Lorentz force law, F = e(E + v × B), yields
Ze2
me ω 2 r = + eωr∆Bext , (3.79)
4πε0 r2
where ∆Bext is the instantaneous change in the magnetic field. In comparison to
the zero field system, there is an instantaneous change in the angular momentum,
|∆L| = 12 er02 ∆Bext . Although a correct treatment of this problem demands a quan-
tum mechanical description of the perturbation to the atomic system, we can use this
change in angular momentum to estimate a change in the “classical” dipole moment.
The change, and thus the induced, magnetic dipole is given by
e e2 r02
mind ≡ − ∆L= − Bext . (3.80)
2me 4me
For Z electrons, the full result (averaging over the different orientations of the electron
orbits) is
e2
m=− Z'r2 (Bext , (3.81)
6me
where 'r2 ( is the mean square radius. The magnetization is then given by
ne2 Z'r2 (
M=− Bext , (3.82)
6me
and this is diamagnetism. Note that it is in the opposite direction to the externally
applied magnetic field.
'
If i Li )= 0 there is an intrinsic dipole moment, mint . As with polar molecules in
 3.9 Diamagnetism and paramagnetism 21

a dielectric the imposition of an external magnetic field Bext will create an alignment.
One finds that
nm2int
M= Bext . (3.83)
3kB T
This is paramagnetism.
Thus, the overall magnetic susceptibility due to diamagnetism and paramagnetism is
given by
# 2
Ze2 'r2 (
$
mint
χB = µ0 n − . (3.84)
3kB T 6me
The first term is the contribution due to alignment (i.e. paramagnetism) and the second
term is due to the induced field (i.e. diamagnetism).
 3.10 Magnetostatics in a magnet 22

WEEK 9 SUPPORTING TEXTS:

Grant & Phillips; Section 4.3, 5
Griffiths; Section 6.1 to 6.4

3.10 Magnetostatics in a magnet

In dielectrics there is a bound charge, ρbound = −∇ · P and a surface charge density
σsurface = P · n̂. In a magnet there is an equivalent bound current and surface current,
jbound = ∇ × M, jsurface = M × n̂, (3.85)
where n̂ is a (unit) vector normal to the surface. The correspondance between the
dielectric and magnetic case is rather simple: the polarization goes over to magnetization,
and the “dot” products become “cross” products - but note the signs.
The equations of magnetostatics are ∇ · B = 0, ∇ × B = µ0 j. As with dielectrics we
will split the current density j into contributions due to bound and free currents,
j = jbound + jfree , (3.86)
and Ampère’s law now becomes
∇ × B = µ0 (jbound + jfree ) = µ0 (∇ × M + jfree ) . (3.87)
We now define the magnetic intensity vector,
1
H≡ B − M, (3.88)
µ0
so that Ampère’s law reads
∇ × H = jfree . (3.89)
The units of H are [Am−1 ]. The integral form of Ampère’s law becomes,
!
H · d! = Ifree . (3.90)
!
The free current, Ifree is the parameter we can control from outside the material, just as
ρfree was in the case of dielectrics.
In diamagnetic and paramagnetic substances the magnetization is linked linearly to
the magnetic field via
χB
M= B. (3.91)
µ0
Using this in (3.88) the magnetic intensity vector reads
1 − χB
H= B. (3.92)
µ0
 3.11 Interfaces between magnets 23

The magnetic energy becomes

"
1
U= B · H dV. (3.93)
2
Let us reconsider the solenoid of length $. The modified Ampere’s law now reads
!
H · d! = H$ = N $I. (3.94)

Hence,
H = N I. (3.95)
1−χB
Since from (3.88) we have H = µ0 B, then

µ0 N 2 πr2 $ dI
E =− , (3.96)
1 − χB dt
and the inductance,
E Lvac
L=− = , (3.97)
I˙ 1 − χB
which motivates us to define the relative permeability,
1
µr ≡ . (3.98)
1 − χB
The magnetic field and magnetic intensity vectors are thus linked via
B = µr µ0 H. (3.99)

3.11 Interfaces between magnets

As with dielectrics, consider the boundary between two regions having relative perme-
(1) (2)
abilities µr , µr , with no free currents on the boundary. Hence, one can find that
!
(Ampere) H · d! = 0 ⇒ H// continuous, (3.100)
"
(no monopoles) B · dA = 0 ⇒ B⊥ continuous. (3.101)

3.12 Ferromagnetism
In the derivation of paramagnetism in Section 3.9, we implicitly assumed that the intrinsic
dipole moments did not interact with each other, which enabled us to write down
nm2intrinsic
M= Bext . (3.102)
3kB T
 3.12 Ferromagnetism 24

In certain magnetic materials (such as iron, nickel and cobalt) interactions do occur and
can lead to the formation of “domains” of size 0.1-1mm, where in a given domain all
magnetic moments have the same alignment. This is ferromagnetism. For ferromagnets,
the linear expressions linking the magnetization, magnetic field and magnetic intensity
vectors (3.91, 3.99) are not applicable. In general however, the magnetization is some
function of the magnetic field, M = M(B) and the relative permeability is given by

1 ∂B
µr = . (3.103)
µ0 ∂H

A full treatment of these interacting dipoles is complicated, and requires quantum

mechanics and field theory (and usually numerical techniques to solve the equations).
The aim here is to develop a picture of what is going on “inside” a ferromagnet.
Consider two magnetic dipole moments, m1 , m2 . The energy associated with these
moments having different alignments is given by ∆U12 ∝ −m1 · m2 . So, to reduce this
energy the dipole moments should align with each other, m1 ∝ m2 .
The formation of the domains themselves is rather interesting, and can be thought of as
being related to the behaviour of a substance through a phase transition. We are familiar
with the idea of something fundamental changing when a substance undergoes a phase
transition. For example, when water cools down below 0◦ C, some degrees of freedom
of the motion of the molecules are removed. The phase transition temperatures are
markers of when this fundamental shift takes place, which affects the collective behaviour
of the constituent water molecules. For phase transitions we thus speak of some “critical
temperature” Tc at which the transition occurs. In ferromagnetism we call the critical
temperature the Curie temperature, TC .
We can describe the formation of domains as follows – let us imagine we have a block of
“ferromagnetic material”, and we start the system at high temperatures. For “high tem-
peratures”, T > TC , the interactions between the magnetic dipoles are relatively weak
(this can be thought about as the thermal fluctuations being dominant over the mag-
netic interaction energy associated with dipole moments). In this phase the substance
is purely paramagnetic, in which case the magnetization and magnetic field vectors are
proportional M ∝ B. Once the temperature of the substance is reduced to just below the
Curie temperature, T = TC − ', interactions between nearby dipoles become important
and “bubbles” of magnetic domains begin to form – this process is called bubble nucle-
ation. As the temperature of the system continues to be reduced the domains increase in
size until the entire substance becomes partitioned into domains. The domains will then
begin to meet at “domain walls” – this produces another complicated type of interaction
which we will not discuss. Once all these magnetic domains have been created they can
be collectively aligned by applying an external magnetic field.
In Figure 3.11 we give a schematic picture of how domains nucleate within a substance
 3.13 Ideal ferromagnets 25

T > TC T = TC − ! T < TC T " TC

Fig. 3.11. Formation of ferromagnetic domains via a phase transition. For high temperatures
relative to the Curie temperature, TC , the substance does not have any domains. As the temper-
ature is decreased domains nucleate and grow. Eventually, magnetic domains completely cover
the substance, where neighbouring domains meet at “domain walls”.

as the temperature is decreased below the Curie temperature. Typically, ferromagnetic

domains are of the order 0.1mm – 1mm in size and contain 1017 – 1020 dipoles.

3.13 Ideal ferromagnets

We will consider first the case of an ideal ferromagnet, which we define as a substance
for whom the magnetization M is a constant within the material (at zero temperature).
This would mean that there is a single magnetic domain within a substance. To get some
feel for the numbers and scales, for example, within iron, we first consider the number
density of magnetic dipoles,
NA ρ 6 × 1023 mol−1 × 7800kg m−3
n= = ≈ 8.4 × 1028 m−3 . (3.104)
[Atomic Weight Fe] 56g mol−1
Multiplying this number by the Bohr magneton gives the rough order of magnitude of
the magnetization:

nµB ≈ 8.4 × 1028 m−3 × 9.3 × 10−24 A m2 ≈ 8 × 105 A m−1 (3.105)

sush that

|M| ≈ few × nµB = few × 8 × 105 A m−1 . (3.106)

The corresponding magnetic field is

|B| ≈ µ0 |M| ≈ few T. (3.107)

When the magnetization is constant there are no bound currents (∇ × M = 0), only
 3.13 Ideal ferromagnets 26

Fig. 3.12. Sketch of the field lines around a ferromagnet. Notice that the surface of the ferro-
magnet looks as though there are a set of magnetic charges sourcing / sinking the magnetic field
lines.

jsurface = M × n̂. The equations of magnetostatics with no free current are given by the
now familiar formulae
∇ × H = 0, ∇ · H = −∇ · M = ρm . (3.108)
These should be compared with the equations of electrostatics, ∇ × E = 0, ∇ · E = ρ/ε0 ,
where we call ρm an effective magnetic charge. Notice that since the magnetization is
constant within an ideal ferromagnet, ∇ · M = 0; however, this does not mean that
ρm = 0 everywhere. On the surface of an ideal ferromagnet, ρm = −∇ · M (= 0. This
is in analogy with the induced surface charge density in conductors. In Figure 3.12 we
illustrate the H field lines pointing towards from the edges (n.b. the field could equally
well point away from the edges) of the region where M is a constant (i.e. from within
the ideal ferromagnet), inducing a surface magnetic charge density where ∇ · M (= 0.
Let us write H = −∇ψ, where ψ is a magnetic scalar potential. From (3.108) it im-
mediately follows that the magnetic scalar potential satisfies a Poisson equation, sourced
by the magnetic charge density,
∇2 ψ = −ρm = ∇ · M. (3.109)
The solution to this equation is
"
1 ∇r! · M(r$ )
ψ(r) = − dV $ . (3.110)
4π |r − r$ |
 3.13 Ideal ferromagnets 27

WEEK 10a SUPPORTING TEXTS:

Grant & Phillips; Section 5.4
Griffiths; Section 6.4.2
This integral can be simplified by using boundary conditions. We first rewrite the
integrand as
" # $ # $%
1 M(r! ) 1
!
! !
ψ(r) = − dV ∇r! · − M(r ) · ∇r! .
4π |r − r! | |r − r! |
Now, the first term is a boundary-term, and is thus zero by the divergence theorem. The
second term can be rewritten via
# $ # $
1 1
∇ r! = −∇ , (3.111)
|r − r! | |r − r! |
so that
# $
1 1
!
! !
ψ(r) = − dV M(r ) · ∇ (3.112)
4π |r − r! |
1 M(r! )
!
= − ∇ · dV ! . (3.113)
4π |r − r! |
Hence, we have derived a formula for calculating the scalar magnetic potential if the
magnetization is known,
1 M(r! )
!
ψ(r) = − ∇ · dV ! . (3.114)
4π |r − r! |
We will now perform this integration for the case of a uniformly magnetized sphere,
whose magnetization vector is modeled via
"
M0 ẑ r < a,
M= (3.115)
0 r > a.
Hence, using this in (3.114), we obtain
! 2π ! a ! π
M(r! ) 1
!
dV ! = M 0 ẑ dφ !
dr ! !2
r dθ! sin θ!
|r − r! | 0 0 0 (r + r − 2rr! cos θ! )1/2
2 !2
! a
2πM0 ẑ & '
= dr! r! r + r! − |r − r! | , (3.116)
r 0

which integrates to yield

2πM0 ẑ r(3a2 − r2 )
!
"
! M(r )
!
r < a,
dV = (3.117)
|r − r! | 3r 2a3 r > a.
Importantly note that everything is independent of the angles θ, φ.
 3.14 Electromagnets and hysteresis 28

(a) B-field (b) H-field

Fig. 3.13. Magnetic fields around a uniformly magnetized sphere.

Therefore, the magnetic scalar potential can be computed,

"
1 z r < a,
ψ = M0 1 3 (3.118)
3 r2
a cos θ r > a.
The contribution outside, r > a, is just that for a magnetic dipole so that B = µ0 H.
Inside the uniformly magnetized sphere, r < a, we have
1
H = −∇ψ = − M0 ẑ, (3.119a)
3

2
B = µ0 (H + M) = µ0 M0 ẑ. (3.119b)
3
Hence, inside the magnetic sphere, the magnetization M increases, H drops and B
increases. Also note that B and H are anti-parallel, which is a generic feature of
ferromagnets. Recall that our calculation was for T = 0. For T > 0, a combination of
shape and anisotropy effects come into play making the problem much more complicated
and resulting in a non-linear response.
See Figure 3.13 for a sketch of the fields inside and outside the spherical ferromagnet.
Notice that the H-field is lower inside than the B-field, and points in the opposite
direction.

3.14 Electromagnets and hysteresis

We will now qualitatively discuss electromagnets, and how an applied current can be
used to generate a permanent magnetization.
 3.14 Electromagnets and hysteresis 29
B/T
A
BS
saturation

spins flip

O
H/Am−1

Fig. 3.14. Sketch of the behaviour of the magnetic intensity and field vectors as the current
surrounding a ferromagnet is increased. As the current is varied between O and A the energy
supplied by the current flips the domains; above A the magnetization has saturated with (ap-
proximately) all domains in the substance being aligned. The increase of the magnetic field is
now very slow.

We will need to imagine a ferromagnetic substance with a current carrying wire wound
around the substance. The wire will carry a current I, whose value we can control. Since
H = H(I), by controlling the current we are infact controlling the magnetic intensity
vector which will in turn affect the magnetic field.
When there is zero applied current, I = 0, the domains inside the ferromagnet are
randomly aligned so that there is no net magnetization, |M| ≈ 0. As the current is
increased the domains begin to flip due to the energy input from the current carrying
wire. The flip of the domains causes H to increase, causing an increase in the magnetic
field B = µr µ0 H and magnetization M ∝ B; typically, µr ≈ 1000. When the current is
increased, the magnetization will reach a maximum M = Mmax (i.e. saturate) when all
domains in the substance are aligned; the only way of increasing the magnetic field is by
a slow linear increase in H. In Figure 3.14 we show a sketch of the behaviour of B(H).
In Figure 3.15 we show a continuation of the sketch in Figure 3.14 where we now show
what happens if we reduce the current. In going from O → A the current is increased until
the saturation point. If we now reduce the current we travel along the curve A → B some
domains will flip back, but not alll, which will leave a remnant magnetization so that M
will remain significant. At B there is zero applied current, H = 0, and the magnetization
is permanent. If we now applied an opposite current, we travel along B → C, at which
point enough of the domains will have flipped to cancel out the magnetic field. This value
of H = HC is called the coercive field. Between C → D the domains all begin to flip in
 3.14 Electromagnets and hysteresis 30
B/T saturation ↑

A
B

C
O F H/Am−1

saturation ↓
D

Fig. 3.15. Hysteresis. The remnant magnetic field is the value of B at B and the coercive field
is the value of H at C.

the opposite direction, until magnetization saturation occurs at D (the magnetization at

D is opposite to that at A).
The work done in going along the round-trip
A→B→C→D→E→F →A (3.120)
is given by
(
U = −V H · dB, (3.121)

which is just the area under the hysteresis curve (which is non-zero).
 PHYS 20141 ELECTROMAGNETISM : 2021/22

Section 4 : Electromagnetic waves

P. Campbell
(for use in conjunction with the teaching material on the PHYS20141 Blackboard area)

Department of Physics & Astronomy

University of Manchester

November 24, 2021

 4
Electromagnetic waves

WEEK 10b SUPPORTING TEXTS:

Grant & Phillips; Sections 11.1, 11.2
Griffiths; Sections 9.1 and start of 9.2
So far we have considered static electromagnetic fields: we have used Maxwells equa-
tions to determine the value of the electric E and magnetic B field vectors space due to
some distribution of charge and/or current. In this chapter we will allow the E and B
fields to have their time dependence, and we will show that Maxwell’s equations can be
rewritten as sourced wave equations. In general these wave equations are complicated,
there are however certain simple situations in which the equations can be solved exactly.
The first situation we study is one where space is completely free of charge and current
(i.e. free space). We will find that plane waves are solutions to Maxwell’s equations
here. We will then use these plane wave solutions to study reflection, polarization and
the energy flux of EM waves. We then allow space to contain currents, which is relevant
in, for example, conductors and plasma.
To solve Maxwell’s equations we will use an ansatz, which means that we make an edu-
cated “guess” for the algebraic form of the solution; importantly, the ansatz will contain
parameters and functions, whose values are unknown and unconstrained. We then insert
this ansatz into the relevant equations and deduce constraints on the parameters within
the ansatz. The dispersion relation, ω = ck, is a popular example of such a constraint.

4.1 Solutions to the wave equation in 1D

We will begin by reviewing some useful results about one dimensional wave equations.
Suppose that φ(x, t) is some time and location dependent field, where the dependence of
φ on space and time is determined by the 1D wave equation,

1 ∂2φ ∂2φ
− = 0. (4.1)
v 2 ∂t2 ∂x2
The solutions to this wave equation are of the form

φ(x, t) = f (x − vt) + g(x + vt). (4.2)

2
 4.1 Solutions to the wave equation in 1D 3

The first term is a “right moving” solution, and the second a “left moving” solution. For
example, if we take a “plane wave” ansatz
φ(x, t) = Aei(kx−ωt) + Bei(kx+ωt) , (4.3)
and substitute this into the wave equation we find that the ansatz should be subject to
the dispersion relation ω = vk. The phase of these waves is ϕ = kx − ωt, and hence parts
of the field with the same phase are those for whom ϕ = 2πn, where n ∈ Z. The phase
velocity is
ω
vp = , (4.4)
k
and the group velocity
∂ω
vg = . (4.5)
∂k
If φ is a real scalar field, the physical part of the solution is
Re(φ) = |A| cos (kx − ωt + arg A) . (4.6)
For example, if arg A = 0,
Re(φ) = A cos (kx − ωt) , (4.7)
or if arg A = ± π2 , then
! π"
Re(φ) = A cos kx − ωt ± = ∓A sin(kx − ωt). (4.8)
2
We have done this for the right-moving part, but the equivalent is also true for the
left-moving term (but in the opposite direction).
For notational clarity, in this section the Cartesian basis vectors are denoted x̂, ŷ, ẑ.
We use k̂ to denote the direction wavevector.
The wave equation in 3D is given by
1 ∂2φ
− ∇2 φ = 0, (4.9)
v 2 ∂t2
whose solutions are given by
φ = Aei(k·r−ωt) + Bei(k·r+ωt) . (4.10)
Notice that the wave-number k from the 1D case has been promoted to the wave-vector
k. If we set B = 0 for simplicity, then the solution reads
φ = Aei(k·r−ωt) = Aei(kx x+ky y+kz z−ωt) . (4.11)
We can calculate the spatial derivative of this solution:
∂φ
= ikx Aei(k·r−ωt) = ikx φ. (4.12)
∂x
 4.2 Maxwell’s equations in free space & the electric wave equation 4

If we calculate the other spatial derivatives we find

∇φ = ikφ, (4.13)
and if we calculated the Laplacian of the solution we find
∇2 φ = −|k|2 φ. (4.14)
Furthermore, if we calculated the second time derivative of the solution we would obtain
∂2φ
= −ω 2 φ. (4.15)
∂t2
We can now substitute (4.14) and (4.15) into the 3D wave equation (4.9) to obtain the
dispersion relation:
ω2
− + |k|2 = 0, (4.16)
v2
which readily rearranges to
ω = v|k|. (4.17)
In 1D, the expression
kx − ωt = 2πn (4.18)
gives the points of constant phase which move at speed v. In 3D the equivalent expression
yields
k · r − ωt = 2πn, (4.19)
which, at a given time, describes a plane of constant phase. We thus refer to (4.10) as
plane waves. Planes of constant phase travel in the direction of k with speed v.

4.2 Maxwell’s equations in free space & the electric wave equation
In free space there are no charges or currents (i.e. ρ = 0, j = 0). Hence, Maxwell’s
equations in free space read
∇ · E = 0, ∇ · B = 0, (4.20a)
2
Ḃ = −∇ × E, Ė = c ∇ × B. (4.20b)
If we now take the time derivative of the last of these equations,
Ë = c2 ∇ × Ḃ = −c2 ∇ × (∇ × E) = −c2 ∇(∇ · E) − ∇2 E ,
# $
(4.21)
and by noting that ∇ · E = 0 in free space, this becomes
1
Ë − ∇2 E = 0, (4.22)
c2
 4.2 Maxwell’s equations in free space & the electric wave equation 5

Fig. 4.1. Sketch of a plane wave.

In a similar fashion one can deduce that

1
B̈ − ∇2 B = 0, (4.23)
c2
although it is important to note that the equations for E and B cannot be solved
independently. Equations (4.22) and (4.23) are wave equations, for the electric and
magnetic fields. Notice that both waves move at speed c and do not have a source (the
RHS of these equations is identically zero; later on we will see the appearance of sources).
A solution to these wave equations is the single-wavelength (i.e. monochromatic) plane
waves. For the electric field, possible example solutions are
E = E0 ei(k·r−ωt) or E0 cos (k · r − ωt) or E0 sin (k · r − ωt) . (4.24)
In Figure 4.1 we sketch a plane wave.
Note that we only take the real part of the complex exponential. The “amplitude” of
the electric field, E0 , is the polarization vector and is a constant. The time average of
the electric field is
% 2& 1
|E| = |E0 |2 . (4.25)
2
By differentiating the ansatz (4.24) one can obtain
Ë = −ω 2 E, ∇2 E = −k 2 E, (4.26)
so that when used in the wave equation (4.22) we find
ω = ±ck (4.27)
still holds. Note that the wavenumber k = |k| is
2π
k= . (4.28)
λ
What this means is that Maxwell’s equations predict the existence of plane wave solutions
traveling at speed v = c with dispersion relation ω = c|k|.
 4.2 Maxwell’s equations in free space & the electric wave equation 6

We can use the plane wave ansatz

E = E0 ei(k·r−ωt) , B = B0 ei(k·r−ωt) (4.29)
and Maxwell’s equations to obtain important relationships between the electric and mag-
netic field vectors.
Before we do this in generality, let us consider a “simple” electric field vector given by
E = E0 x̂ cos(kz − ωt). (4.30)
We can then obtain
∇2 E = −k 2 E, Ë = −ω 2 E. (4.31)
' '
' x̂ ŷ ẑ ''
∂ ∂ ∂ '
'
∇ × E = '' ∂x ∂y ∂z ' = −kE0 ŷ sin(kz − ωt) = −Ḃ. (4.32)
' E cos(kz − ωt) 0 0 '
0

The last equality follows from the Faraday-Lenz law. After integrating it, we find
kE0
B= ŷ cos(kz − ωt), (4.33)
ω
ie. B0 = (kE0 /ω)ŷ. Hence, for the explicit example, E ∝ x̂ we find that B ∝ ŷ and
k ∝ ẑ.

WEEK 11 SUPPORTING TEXTS:

Grant & Phillips; Sections 11.4, 11.5, 11.6
Griffiths; Sections 9.2.1, 9.3 and start of 9.4
We will now perform this analysis for a general monochromatic plane wave
E = E0 ei(k·r−ωt) . (4.34)
It is useful to first deduce that
∇ · E = ik · E, ∇ × E = ik × E. (4.35)
From Gauss’ law in free space, ∇ · E = 0, we find
k · E0 = 0, (4.36)
which means that the wavevector k and electric field E are orthogonal. Also,
−Ḃ = ∇ × E = ik × E0 ei(k·r−ωt) , (4.37)
which can be integrated to give
k × E0 i(k·r−ωt)
B= e . (4.38)
ω
 4.2 Maxwell’s equations in free space & the electric wave equation 6

We can use the plane wave ansatz

E = E0 ei(k·r−ωt) , B = B0 ei(k·r−ωt) (4.29)
and Maxwell’s equations to obtain important relationships between the electric and mag-
netic field vectors.
Before we do this in generality, let us consider a “simple” electric field vector given by
E = E0 x̂ cos(kz − ωt). (4.30)
We can then obtain
∇2 E = −k 2 E, Ë = −ω 2 E. (4.31)
! !
! x̂ ŷ ẑ !!
∂ ∂ ∂ !
!
∇ × E = !! ∂x ∂y ∂z ! = −kE0 ŷ sin(kz − ωt) = −Ḃ. (4.32)
! E cos(kz − ωt) 0 0 !
0

The last equality follows from the Faraday-Lenz law. After integrating it, we find
kE0
B= ŷ cos(kz − ωt), (4.33)
ω
ie. B0 = (kE0 /ω)ŷ. Hence, for the explicit example, E ∝ x̂ we find that B ∝ ŷ and
k ∝ ẑ.

WEEK 11 SUPPORTING TEXTS:

Grant & Phillips; Sections 11.4, 11.5, 11.6
Griffiths; Sections 9.2.3, 9.4.1 and 9.4.2
We will now perform this analysis for a general monochromatic plane wave
E = E0 ei(k·r−ωt) . (4.34)
It is useful to first deduce that
∇ · E = ik · E, ∇ × E = ik × E. (4.35)
From Gauss’ law in free space, ∇ · E = 0, we find
k · E0 = 0, (4.36)
which means that the wavevector k and electric field E are orthogonal. Also,
−Ḃ = ∇ × E = ik × E0 ei(k·r−ωt) , (4.37)
which can be integrated to give
k × E0 i(k·r−ωt)
B= e . (4.38)
ω
 4.3 Poynting vector, intensity and radiation pressure 7

This tells us that the magnitude of the magnetic field is

k × E0
B = B0 ei(k·r−ωt) , B0 = . (4.39)
ω
We can now see that the following identities hold:
E0 · B0 = k · E0 = k · B0 = 0, (4.40)
these mean that the electric, magnetic and wavevectors are all mutually orthogonal. In
addition we find that
kE0 1
B0 = = E0 . (4.41)
ω c
If we form the vector product of E0 with B0 , we further obtain
E0 × (k × E0 )
E 0 × B0 =
ω
1" 2 #
= E0 k − (k · E0 )E0
ω
E02
= k, (4.42)
ω
and therefore,
k̂ = Ê0 × B̂0 . (4.43)
Another important property of the EM waves is that E and B are in phase ie. the terms
in the exponential are the same in both.

4.3 Poynting vector, intensity and radiation pressure

The energy density for an electromagnetic field is given by
1 1
u = ε0 |E|2 + |B|2 , (4.44)
2 2µ0
and the energy, U , is calculated by integrating the energy density over all space,
$
U = dV u. (4.45)

We can use the electric and magnetic fields which describe plane waves (4.29) to compute
the energy density. One finds
1 % & '(2 1 % & '(2
u = ε0 E02 Re ei(k·r−ωt+θ) + E 2
Re e i(k·r−ωt+θ)
2 2µ0 c2 0
% & '(2
= ε0 E02 Re ei(k·r−ωt+θ) . (4.46)
 4.3 Poynting vector, intensity and radiation pressure 8

Notice that the contributions to u from the electric and magnetic fields are identical. We
can read off that the maximum energy density is
umax = ε0 E02 , (4.47)
and we can calculate that the average energy density is
1
%u& = ε0 E02 , (4.48)
2
as cos2 (k · r − ωt + θ) = 12 .
) *

We now consider the volume covered by a wave in time-interval ∆t through surface

area ∆A. Then the average energy density enclosed by this volume is given by
1
%enclosed energy density& = %u& c∆t∆A = ε0 E02 c∆t∆A = ∆Utot . (4.49)
2
Hence, the average flux through a surface is given by
∆Utot E2
F= = 0 , (4.50)
∆t∆A 2Z0
where we have defined the impedance of free space
+
µ0
Z0 ≡ = 377 Ω. (4.51)
ε0
We will now calculate the rate of change of the energy density,
1 ∂ 1 ∂ 1
u̇ = ε0 (E · E) + (B · B) = ε0 E · Ė + B · Ḃ. (4.52)
2 ∂t 2µ0 ∂t µ0
Substituting with Maxwell equations in free space, Ḃ = −∇×E, Ė = c2 ∇×B, we obtain
1
u̇ = (E · ∇ × B − B · ∇ × E) = −∇ · N, (4.53)
µ0
where we have defined
1
N= E×B (4.54)
µ0
which is the Poynting vector. It is an energy flux; the units of the Poynting vector are
V m−1 T
[N] = = A V m−2 = W m−2 , (4.55)
Wb A−1 m−1
i.e. Watts per square metre. The rate of change of energy is
$ $
U̇ = − dV ∇ · N = − N · dS, (4.56)

where the second equality arises from the divergence theorem.

 4.3 Poynting vector, intensity and radiation pressure 9

In summary, we have calculated the rate of change of the total energy U̇ in the same
volume, and we have found that it is equal to the total flux of the vector N out of a
surface (in vacuum). Therefore we identify N as the energy flux through a surface.
In materials it is often useful to define the Poynting vector as,

N = E × H, (4.57)

which differs to the earlier definition if µr (= 1. The difference arises due to the presence
of currents (and energy loss to driving these currents). In the former definition the energy
expended driving all currents is considered. In the latter, E × H, the energy expended
on free currents alone is balanced by the flux.
Using the 4.54 definition we can insert a plane wave solution (4.29) into the Poynting
vector,
1 E0 × (k × E0 ) % & i(k·r−ωt+θ) '(2
N = E×B= Re e
µ0 ωµ0
E02
= k̂ cos2 (k · r − ωt + θ) , (4.58)
Z0
where Z0 is the impedance of free space (4.51). We note that the direction of wave
motion (i.e. k̂) is the same as that of N, so that the waves energy flux goes in the same
direction as the wavevector (which is not the case for the electric or magnetic fields). We
can read off the maximum value that the Poynting vector takes
E02
Nmax = , (4.59)
Z0
and again note that the units of Nmax are W m−2 . The irradiance (or intensity) is defined
to be the average of the Poynting vector,
E02
%N & = . (4.60)
2Z0
The momentum of a photon is p = U/c, and hence the momentum flux is %N & /c
[Nm−2 ], and this is the radiation pressure, Pr . At a boundary the radiation creates a
force, Fr , and the magnitude of this force depends on whether or not the radiation is
absorbed or reflected by the boundary:

• If all radiation is absorbed then

A %N &
Fr = . (4.61)
c
• If all radiation is reflected then
2A %N &
Fr = . (4.62)
c
 4.4 EM waves in the presence of a current 10

4.4 EM waves in the presence of a current

We will now study how electromagnetic waves propagate in the presence of currents
(but still in the absence of charge). In general the Maxwell equations become rather
complicated and difficult to solve, but if we can impose generic constraints on the current,
by specifying how it is related to the electric field, and the theory becomes soluble.
In the presence of current and absence of charge (i.e. set ρ = 0, j (= 0), Maxwell’s
equations become

∇ · E = 0, ∇ · B = 0, (4.63a)

Ḃ = −∇ × E, Ė = c2 (∇ × B − µ0 j) . (4.63b)

We can find the time derivative of the third Maxwell equation and substitute the fourth
for Ė to obtain

B̈ = −∇ × Ė = −c2 ∇ × (∇ × B − µ0 j) = c2 ∇2 B + µ0 ∇ × j ,
" #
(4.64)

that is,
1
B̈ − ∇2 B = µ0 ∇ × j. (4.65)
c2
In a similar fashion we can obtain
1 ∂j
2
Ë − ∇2 E = −µ0 . (4.66)
c ∂t
Equations (4.65) and (4.66) are two sourced wave equations (these should be compared
with (4.22, 4.23) which were the unsourced wave equations). We will concentrate on
computing the electric field, as the magnetic field can be calculated from the electric
field via Ḃ = −∇ × E.
To continue we will consider two physical cases. The cases are “modeled” by choosing
a particular dependence of the current upon the electric field.

4.4.1 Conductor
In a conductor obeying Ohm’s law, the current is linearly related to the electric field,

j = σE, (4.67)

where the constant of proportionality, σ, is called the conductivity, and is a parameter

which is specific to particular materials. We can insert this into the electric fields wave
equation (4.66) to find
1
Ë + µ0 σ Ė = ∇2 E. (4.68)
c2
 4.4 EM waves in the presence of a current 11

This is a damped wave equation, and can be solved with an ansatz,

& '
E = E0 Re ei(k·r−ωt) . (4.69)

When we insert this ansatz into the wave equation we find a modified dispersion
relation,
ω2
− − iµ0 σω = −k 2 . (4.70)
c2
Let us define
, -−1
µ0 σω σ & σ ' f
R= 2 2 = = 1.8 × 109 . (4.71)
ω /c 2πε0 f 108 Ω−1 m−1 1GHz
For most cases of interest R ) 1, which implies that
ω2
* µ0 σω, (4.72)
c2
so that
k 2 = iµ0 σω = eiπ/2 µ0 σω, (4.73)
and therefore,
√ 1 √
k = ±eiπ/4 µ0 σω = ± √ (1 + i) µ0 σω . (4.74)
2
Hence, substituting this back into the ansatz for a z-directed wave we obtain
!√ "
µ0 σω
i z−ωt
E = E0 e−z/δ e 2
, (4.75)
where we have defined the skin depth,
+
2
δ= . (4.76)
µ0 σω
This is the length scale over which the amplitude of an EM wave decays inside a con-
ductor. We can imagine a scenario where an EM waves comes from a vacuum, into a
conducting slab, and back out into a vacuum. The wave would look like an oscillatory
solution in the first a vacuum, a damped oscillatory solution in the conductor (with the
amplitude being reduced exponentially) and an oscillatory solution in the second vacuum
– the amplitude here is the final amplitude in the conductor; for a sketch of this situ-
ation see Figure 4.2. Notice that a good conductor will have a very high conductivity,
and therefore a very small skin depth: this means that electromagnetic waves will not
propagate very far into a good conductor.
The behaviour gives rise to the skin effect phenomenon apparent in conductors carrying
alternating current (AC), especially at high frequency. In all conductors the AC current
 4.4 EM waves in the presence of a current 12

Fig. 4.2. Sketch of the behaviour of an electromagnetic wave inside a conductor. The EM wave
comes in from the left through free space, enters the conductor at the solid line. The conduc-
tivity within the medium causes the wave to become damped (we have sketched on the damping
envelope) so that by the time the wave leaves the conductor (at the dashed line) the amplitude
has decreased.

transport is non-uniformly distributed with the majority of the conduction only occurring
near the surface of the conductor. The thin conducting layer (referred to as the “skin”)
has a depth characterised by the skin depth, δ. In copper at 50 MHz this layer is only
∼ 10−5 m thick.

4.4.2 Plasma
A plasma is a state of matter which is partially ionised and electrons move under the
influence of the electric field, ignoring the much weaker magnetic field. Thus, the only
force acting on the electron is due to the electric field:

Fe = me r̈ = −eE, (4.77)

where the second equality follows from the Lorentz force law. Also, by the definition of
current density,

j = −ene ṙ, (4.78)

one can readily deduce that

∂j e 2 ne
= E, (4.79)
∂t me
 4.5 Reflection of EM waves from a perfect conductor 13

where ne is the electron number density in the plasma and me is the mass of the electron.
We can now use this expression in the wave equation for the electric field (4.66),
1 µ 0 e 2 ne
Ë + E = ∇2 E. (4.80)
c2 me
Substituting the ansatz (4.69) into the wave equation (4.80), we obtain the dispersion
relation
ω 2 µ 0 e 2 ne
− + = −k 2 . (4.81)
c2 me
This can be rearranged to obtain
.
ck = ± ω 2 − ωp2 , (4.82)

where we have defined

c 2 µ 0 e 2 ne e 2 ne
ωp2 = = (4.83)
me ε 0 me
which is called the plasma frequency.
Notice that from (4.82) we can see that if ω < ωp , then k is purely imaginary and EM
waves do not propagate. This means that the EM radiation inside the plasma must
have small wavelengths (high frequencies) to propagate: as the density of the plasma
increases the allowed wavelengths will become shorter.

4.5 Reflection of EM waves from a perfect conductor

We will now consider some physical situations involving electromagnetic radiation. Firstly
we study a case where a beam of electromagnetic waves is incident upon a perfect con-
ductor. We can use our knowledge of the electric field inside a perfect conductor to
deduce boundary conditions for the EM waves, and use this to find the amplitude of any
reflected wave in terms of the amplitude of the incident.
The boundary and coordinate system for this problem are as shown in Figure 4.3. An
EM wave E(I) is incident upon a perfect conductor (remembering that within a perfect
conductor, E = 0) at an angle θI to the normal. A wave E(R) is reflected at an angle θR .
We may write generic forms of the incident and reflected waves as
    
− cos θI 0 & '
(I) (I)  (I) 
E =  E1 sin θI  + E2 0  Re ei(k·r−ωt) , (4.84a)
0 1
    
cos θR 0 & ! '
(R) (R)
E(R) = E1  sin θR  + E2  0  Re ei(k ·r−ωt) , (4.84b)
0 1
 4.5 Reflection of EM waves from a perfect conductor 14
conductor E = 0

y=0

θR
θI
y E(R)

x E(I)

Fig. 4.3. Setup, definitions and coordinate system for reflection of an EM wave from a perfect
conductor.

where the wavevectors of the incident and reflected waves are given by

k = k (sin θI , cos θI , 0) , (4.85)

"
k = k (sin θR , − cos θR , 0) . (4.86)

Thus, the waves have the same frequency, but change direction. The task is now to link
(R) (I)
θI and θR and the amplitudes Ei , Ei .
We must now impose boundary conditions. First, the phase must be continuous at
y = 0, which implies that k · r = k" · r at y = 0. Therefore,

kx sin θI = kx sin θR ⇒ θI = θR , (4.87)

which is the well known result that the angle of reflection is the same as the angle of
incidence. Secondly, E// = 0 at y = 0, because E// is continuous and there are no
electric fields inside a conductor. Thus, in the x-direction,
(I) (R) (I) (R)
−E1 cos θI + E1 cos θR = 0 ⇒ E1 = E1 , (4.88)

and in the z-direction,

(I) (R) (I) (R)
E2 + E2 =0 ⇒ E2 = −E2 . (4.89)

The interpretation of these results is that no energy is lost during reflection. Thus, the
total electric field is
(I)
 
−E1 cos θI [cos (k · r − ωt) − cos (k" · r − ωt)]
Etot = E(I) + E(R) =  E1(I) sin θI [cos (k · r − ωt) + cos (k" · r − ωt)]  ,
 
(I)
E2 [cos (k · r − ωt) − cos (k" · r − ωt)]
(4.90)
 4.5 Reflection of EM waves from a perfect conductor 15

or, after some manipulation,

(I)
 
−2E1 cos θI sin (ωt − kx sin θI ) sin (ky cos θI )
Etot =  2E1(I) sin θI cos (ωt − kx sin θI ) cos (ky cos θI )  . (4.91)
 
(I)
2E2 sin (ωt − kx sin θI ) sin (ky cos θI )
We notice that these are standing waves.
There is a surface charge density “induced” by the EM wave being reflected on the
conductor; the surface charge density σ is calculated by finding the component of the
total electric field along the surface of the conductor:
σ (I)
= Etot · −ŷ|y=0 = −2E1 sin θI cos (ωt − kx sin θI ) . (4.92)
ε0
 4.6 Polarization of EM waves & Stokes’ parameters 16

WEEK 12 SUPPORTING TEXTS:

Grant & Phillips; Sections 11.2, 11.6.2
Griffiths; Section 9.1.4, 9.3.1 and 9.3.3

4.6 Polarization of EM waves & Stokes’ parameters

To generally discuss the polarization of electromagnetic radiation we will consider a z-
directed monochromatic EM wave (being z-directed means that the wavevector is given
by k = k(0, 0, 1)). The electric field vector is given by
! "
E = Re E0 ei(kz−ωt) , (4.93)

where E0 is a “complex amplitude”, which we can write as

# $
1 iθ i(θ+α)
E0 = √ E0,x e x̂ + E0,y e ŷ , (4.94)
2
where E0,x and E0,y are real. The parameter θ is present in both terms and represents
an overall phase that dictates whether the function is a cosine, sine or a combination
of the two. The parameter α represents a relative phase. This amplitude term has two
interesting special cases: α = 0 and α = π/2.
The first case, α = 0 means that the two components are in phase so that the electric
field vector becomes (for θ = 0)
% &
1
E= √ E0,x x̂ + E0,y ŷ cos(kz − ωt). (4.95)
2
This is linear polarization along the axis √12 E0,x x̂ + E0,y ŷ and is illustrated in Figure
' (

4.4(a).
The second interesting possibility is characterized by α = ±π/2 and E0,x = E0,y in
which case the two components are out of phase by ±π/2. The electric field vector is
given by
# $
E0,x
E = √ x̂ cos(kz − ωt) ∓ ŷ sin(kz − ωt) , (4.96)
2
at θ = 0. For the “+” solution the electric field has right circular polarization, and for
the “−” solution the electric field has left circular polarization. For example, if we take
α = −π/2 then the components of the electric field vector are

Ex = ER cos(kz − ωt), Ey = ER sin(kz − ωt), (4.97)

where ER = √1 E0,x . The electric field will then trace out a circle of radius ER going
2
 4.6 Polarization of EM waves & Stokes’ parameters 17

Ey Ey
√1 E0,y ER
2

χ
Ex Ex
ER
− √12 E0,x √1 E0,x
2
−ER

− √12 E0,y
−ER

(a) Linear polarization (b) Right circular polarization

(from point of view of the receiver)

Fig. 4.4. Behaviour of the components of the electric field vector in the plane perpendicular to
the wave motion for linear and right circular polarization. The angle χ ≡ kz − ωt and the arrow
denotes the directed the electric field rotates around the Ex − Ey plane.

anti-clockwise in the Ex − Ey plane as shown in Figure 4.4(b). If we had taken α = π/2

then it would have moved clockwise.
In general the electric field traces out an ellipse in the Ex − Ey plane, and it is useful
to characterize the polarization of a wave via the Stokes parameters I, Q, U, V . We will
always operate in the complex representation, such that

cos(k · r − ωt) = Re ei(k·r−ωt) , sin(k · r − ωt) = Re ei(k·r−ωt− 2 ) .

) * ) π *
(4.98)

Defining two new bases (L̂, R̂) and (â, b̂)

1 1
L̂ = √ (x̂ + iŷ) , R̂ = √ (x̂ − iŷ) . (4.99a)
2 2

1 1
â = √ (x̂ + ŷ) , b̂ = √ (x̂ − ŷ) . (4.99b)
2 2
Note that (â, b̂) are at 45◦ to x̂, ŷ. We define the Stokes parameters via

I = |Ex |2 + |Ey |2 , Q = |Ex |2 − |Ey |2 , (4.100a)

U = |Ea |2 − |Eb |2 , V = |E$ |2 − |Er |2 , (4.100b)

where the components of the electric field vector in the various bases are given by

Ea = â · E, Eb = b̂ · E, E$ = L̂∗ · E, Er = R̂∗ · E. (4.101)

The parameters (Q, U ) describe linear polarization and V circular polarization. The
parameter I is the average intensity. See Figure 4.5 for a diagrammatic representation.
 =0
=0
Q<

Q>

4.6 Polarization of EM waves & Stokes’ parameters 18

Ey Ey
Q=0
Q=0
Q<0
U <0 U >0

Ex Ex
Q>0

U >0 U <0
V >

(a) Linear polarization

Ey Ey

U< U>

Ex Ex

U> U <V > 0 V <0

(b) Circular polarization

Fig. 4.5. The properties of the Stokes parameters for (a) linear and (b) circular polarization, in
the (Ex , Ey )-plane.

For a monochromatic plane wave one can calculate that

V <
I 2 = Q2 + U 2 + V 2 . (4.102)

The fraction of a wave which is linearly polarized is

+
Q2 + U 2
fL = , (4.103)
I
and the fraction with circular polarization,
V
fC = , (4.104)
I
with

fL2 + fC2 = 1. (4.105)

Also note that

U = Ey Ex∗ + Ex Ey∗ = 2Re(Ey Ex∗ ), (4.106a)

 4.7 EM waves in dielectric and magnetic media 19
' (
V = i Ey∗ Ex − Ey Ex∗ = 2Im(Ey Ex∗ ), (4.106b)
which means that all parameters may be determined in the x̂, ŷ basis.

4.7 EM waves in dielectric and magnetic media

Recall that for electrostatics in a dielectric medium we have
ρbound = −∇ · P, D = ε0 E + P, (4.107)
and for magnetostatics in a magnetic medium,
1
jbound = ∇ × M, H= B − M. (4.108)
µ0
In non-static situations there is an extra current due to the time dependence of the
polarization, P:
jbound = ∇ × M + Ṗ. (4.109)
The last term is the displacement current created by the polarization. Maxwell’s equa-
tions become
∇ · D = ρfree , ∇ · B = 0, (4.110a)

Ḃ = −∇ × E, Ḋ = ∇ × H − jfree , (4.110b)
where ρfree and jfree are the free charge and currents, which are the quantities that can
be externally controlled.
For simplicity let us set the free charge and current to zero, ρfree = 0, jfree = 0, and
consider linear isotropic materials (i.e. not ferromagnets) so that we can take
1
D = εr ε0 E, H= B. (4.111)
µr µ0
Thus, the Ampere-Maxwell equation becomes
c2
Ė = ∇ × B = v 2 ∇ × B, (4.112)
εr µ r
where
c c
v= =√ , (4.113)
n εr µ r
√
and n = εr µr is the refractive index. Therefore, we see that EM waves move a speed
v < c in dielectric/magnetic media. The dispersion relation becomes
ck
ω = ±vk = ± . (4.114)
n
 4.7 EM waves in dielectric and magnetic media 20
E(T)
Refractive index nT

θT
y=0

θR
θI
y E(R)

x E(I)
Refractive index nI

Fig. 4.6. Setup, definitions and coordinate system for refraction of an EM wave at a boundary
between two regions of different refractive indices.

It is important to note that the refractive index n may be dependent upon frequency.
Consider the situation depicted in Figure 4.6: an EM wave in a medium with refractive
index nI hits a boundary with another medium with refractive index nT . The EM wave
then splits into reflected and transmitted components.
We write the field vectors as
(I)
E(I) = E0 ei(k·r−ωt) , (4.115a)
(R) (R) !
E = E0 ei(k ·r−ωt) , (4.115b)
(T) !!
E(T) = E0 ei(k ·r−ωt) , (4.115c)
where the wavevectors are
k = k(sin θI , cos θI , 0), (4.116a)
k% = k(sin θR , − cos θR , 0), (4.116b)
%% %%
k = k (sin θT , cos θT , 0), (4.116c)
where
nI ω nT ω
k= , k %% = . (4.117)
c c
As for reflection from a conductor, we demand from the boundary that the phase is
continuous
k · r = k% · r = k%% · r at y = 0. (4.118)
Hence,
kx sin θI = kx sin θR ⇒ θI = θR (4.119)
 4.7 EM waves in dielectric and magnetic media 21

as before, and
kx sin θI = k %% x sin θT , (4.120)
which, if we insert the definitions of k and k %% in terms of the refractive indices, reads
nT sin θT = nI sin θI , (4.121)
which is Snell’s law.
One can go further and compute the electric fields E(R) and E(T) in terms of E(I) , which
gives what are known as Fresnel’s equations, which give rise to a number of phenomenon
such as the Brewster angle.
If we consider an EM wave going from air into glass (i.e. nI = 1, nT = 1.5), so that
sin θT = 0.66 sin θI , (4.122)
or, in the opposite direction (from glass into air),
sin θT = 1.5 sin θI . (4.123)
In this second case, sin θI > 1/1.5, sin θT is greater than unity, and there is no transmitted
wave, and everything is reflected back into the glass. This occurs at the critical angle
(when nI > nT ),
θcrit = sin−1 (nT /nI ), (4.124)
such that θI > θcrit only leads to a reflected wave: this is called total internal reflection.