MSci 4261 ELECTROMAGNETISM

LECTURE NOTES IV
In this part of the notes, we will discuss how Maxwell’s equations can be reduced to second-order
equations for the vector and scalar potentials. We will exhibit the solutions to these equations by defining
the Green functions.
4.1 Scalar and Vector Potentials
Recall the Maxwell equations in a vacuum with sources -

∇ · B = 0,
∂B (1)
∇×E=− ,
∂t

and
1
∇·E= ρ,
0
(2)
∂E
∇ × B = µ 0 0 + µ0 J.
∂t
We can solve the first two equations by introducing the vector potential A and the scalar potential Φ, and
writing
B = ∇ × A,
∂A (3)
E = −∇Φ − .
∂t
However, note that there is a redundancy in this solution - B and E are unchanged if we make the transfor-
mations
A → A + ∇Λ,
∂Λ (4)
Φ→Φ− ,
∂t
for any function Λ. This redundancy is called a gauge freedom, and the transformations (4) are called gauge
transformations or a gauge symmetry. Whilst this may seem a little recondite, gauge symmetries turn out to
be central to our understanding of modern fundamental interactions, in particular in the Standard Model.
They are also crucial to our understanding of the quantum electromagnetic field.
The other two Maxwell equations (2) can be written, using (3), as

∂ 1
∇2 Φ +
∇ · A) = − ρ
(∇
∂t 0
(5)
1 ∂2A 1 ∂Φ
∇2 A − 2 2 − ∇(∇
∇·A+ 2 ) = −µ0 J.
c ∂t c ∂t

One way to proceed is to fix the gauge freedom (4) by imposing what is called a gauge fixing condition. This
effectively fixes Λ. An example is the Lorentz gauge

1 ∂Φ
∇·A+ = 0. (6)
c2 ∂t

The general idea here is to solve (5) for the potentials A, Φ (in terms of the sources ρ, J), and then (3) gives
the electromagnetic fields E, B. In the Lorentz gauge, the equations (5) become

1 ∂2Φ 1
uΦ := ∇ 2 Φ −
t =− ρ
c2 ∂t2 0
(7)
1 ∂2A
uA := ∇2 A − 2 2 = −µ0 J
t
c ∂t
1
(u
t is the d’Alembertian).

Now note the following:

(1) Notice that the left-hand sides of the two equations in (7) are very similar - in fact the potentials are
part of a four-vector Aµ as we will see in a later lecture;
(2) The operator tu is a wave operator, and the equations (7) are wave equations;
(3) There are other gauge fixing conditions - for example the Coulomb gauge, which is ∇ · A = 0. In this
gauge we obtain
1
∇2 Φ = − ρ
0
so here the potential Φ describes the instantaneous Coulomb potential due to the charge density ρ, and

t
uA = −µJt , (∗)

where Jt is transverse (∇·Jt = 0). Physically then, in the Coulomb gauge the scalar potential determines
the ”near” Coulomb fields in terms of Φ and the vector potential A determines the (tranverse) radiation
fields in terms of Jt .

A proof of (*) follows: Firstly note (see lecture notes 2)

1
∇2 = −4πδ 3 (x0 − x)
|x0 − x|

(where the derivatives in ∇ are with respect to x), and the identity

∇2 V = ∇∇ · V − ∇ × ∇ × V

for any vector field V. Thus

−1 J(x0 )
Z Z
J(x) = δ 3 (x0 − x)J(x0 ) d3 x0 = ∇2
4π |x0 − x|
1 J(x0 ) 3 0 1 J(x0 ) 3 0
Z Z
= − ∇∇ · d x + ∇×∇× d x
4π |x0 − x| 4π |x0 − x|
=: Jl + Jt

with ∇ × Jl = 0 and ∇ · Jt = 0, so that these fields are longitudinal and transverse respectively. Now
∇2 Φ = − 10 ρ in the Coulomb gauge, so that

1 ρ(x0 )
Z
Φ= d 3 x0
4π0 |x0 − x|

whence
1 1 ρ̇(x0 ) µ0 ∇0 · J(x0 ) 3 0
Z Z
∇ Φ̇ = ∇ d 3 x0 = − ∇ d x = µ 0 Jl
c2 4π0 c2 |x0 − x| 4π |x0 − x|
(as ∇ · J + ρ = 0). Thus (see (5))
1
t
uA = −µ0 J + ∇Φ̇ = −µ0 Jt .
c2

2
4.2 The Delta-Function

Our discussion of Green functions will rely on use of Dirac delta functions. The delta-function is a
continuum generalisation of the finite-dimensional Kronecker delta. Recall that this is defined by

1 n = n0

δnn0 =
0 n 6= n0

(n is an integer) so that X
δnn0 fn = fn0 .
n

The continuous analogue of this (in one dimension) is denoted δ(x − x0 ). This has the properties

δ(x − x0 ) = 0 x 6= x0 ;

δ(x − x0 ) = δ(x0 − x);

and Z
dx δ(x − x0 ) f (x) = f (x0 )

for “all” functions f (there are some weak restrictions which will not concern us here). Some consequences
of these properties are Z
δ(x − x0 ) dx0 = 1

if the integration range includes the value x0 = x (the integral is zero if this is not the case),

1
δ(cx) = δ(x),
c
(c is a constant), and
1  
δ(x2 − a2 ) = δ(x − a) + δ(x + a)
2|a|
(a is a constant) which we will need in the following. The proof of this relation follows by noting that
Z ∞ Z ∞ Z ∞
2 2 2
f (a ) = δ(y − a )f (y)dy = δ(y − a )f (y)dy = δ(x2 − a2 )f (x2 )2x dx.
−∞ 0 0

1
One can then check that this is the same as 2|a| (δ(x − a) + δ(x + a)) (inserted in the same integral) in each
of the cases a > 0, a < 0.
An explicit representation of the delta function, which we will use often in the following, is
Z ∞
1
δ(λ) = dωe−iλω .
2π −∞

(Using the Fourier transform formulae above, one can check that this representation satisfies
Z ∞
δ(λ − λ0 )f (λ)dλ = f (λ0 )
−∞

for any function f .)

The one-dimensional delta function defined above can be immediately generalised to higher dimensions,
in Cartesian coordinates. For example, in three dimensions, define

δ 3 (x − x0 ) = δ(x − x0 )δ(y − y 0 )δ(z − z 0 )

3
where x = (x, y, z), x0 = (x0 , y 0 , z 0 ). In other coordinates a Jacobian arises - for example, in spherical polars

1
δ 3 (r − r0 ) = δ(r − r0 )δ(θ − θ0 )δ(φ − φ0 )
r2 sin θ
(recall that dxdydz = r 2 sin θdrdθdφ). The Jacobian term is needed so that the properties of the delta
function continue to be satisfied.
4.3 Green functions
In this section we will see how to formally solve wave equations using inverse differential operators, or
Green functions. First consider the equation

1 ∂2A
uA = ∇2 A −
t = −µ0 J.
c2 ∂t2
R∞
Integrate this with −∞ dteiωt to get

∇2 + k 2 )A(x, ω) = −µ0 J(x, ω)

(∇ (8)

in terms of the Fourier transform fields. We define k 2 = ω 2 /c2 .

We wish to solve this equation for the electromagnetic potentials A, given the source fields J.
Suppose that there is a function Gk (x, x0 ) satisfying

∇2 + k 2 ) Gk (x, x0 ) = −4πδ 3 (x − x0 ).
(∇ (9)

Such a Gk is called a Green function for the operator ∇ 2 + k 2 . One can think of δ 3 (x − x0 ) as the functional
unit 1 (just like the Kronecker delta is the (infinite-dimensional but discrete) unit matrix), so that G k is
proportional to the inverse of the operator ∇ 2 + k 2 .
Then we have the result that
µ0
Z
A(x, ω) = Gk (x, x0 )J(x0 , ω)d3 x0 , (10)
4π

is the solution of (8). (This may be viewed as solving (8) functionally by writing A as the inverse of the
differential operator, ie the Green function, times the right-hand side of (8).) The above result is easy to see:

µ
Z  
∇2 + k 2 )A(x, ω) = 0
(∇ ∇2 + k 2 )x Gk (x, x0 ) J(x0 , ω)d3 x0
(∇
4π
Z
= −µ0 δ 3 (x − x0 )J(x0 , ω)d3 x0 = −µ0 J(x, ω)

Thus, to solve (8) for A, one need only solve the simpler equation (9) for Gk , then use (10) to find A.

To solve (9), note that ∇ 2 + k 2 is x-translation invariant, and so the solution of (9) can only depend upon
the difference
r = x − x0 .
∇2 is a scalar under rotations) so in fact the solution can
The operator ∇2 + k 2 is also rotation invariant (∇
only depend upon
r = |r| = |x − x0 |.
Now use spherical polar coordinates (r, θ, φ) (for x − x0 ). Then by the above arguments, Gk is a function of
r only. In spherical polars,

1 ∂2   ∂ ∂ 
∇2 Gk (r) = rG k (r) + , terms
r ∂r2 ∂θ ∂φ

4
so that (9) becomes
1 d2
rGk + k 2 Gk = −4πδ (3) (r).


2
(11)
r dr
To solve this, consider the case when r 6= 0. Then the equation above can be written

d2
rGk + k 2 (rGk ) = 0,


dr2
which has solution
rGk = Aeikr + Be−ikr ,
with A, B constants.
1
When r → 0, the r term on the left-hand side of (11) dominates so that the equation becomes

1 d2
rGk = −4πδ (3) (r)


2
(12).
r dr
This is just what we would get from Poisson’s equation
ρ
∇2 Φ = − ,
0

if we took Φ = Gk and ρ = 4π0 δ (3) (r). The latter corresponds to a point charge of magnitude q = 4π0 ,
located at the origin. The solution to this Poisson equation is then

q 1 1
Φ= = ,
4π0 r r

so that the solution to (12) above is

1
Gk = .
r

Combining the information above we conclude that the solution of (9) is

eikr e−ikr
Gk = A +B , with A + B = 1,
r r

the condition on A, B arising from the requirement that Gk → 1r as r → 0. Physical considerations will
dictate the choice of constant A (or B).
Let
e±ikr
G(±) = .
r
Since Z ∞
1
A(x, t) = dω e−iωt A(x, ω),
2π −∞
the time-dependent A(x, t) will have e−iωt factors. Thus we consider the functions

(±) −iωt e−i(ωt∓kr)

Gk e = .
r

The function G+ 0
k is an outgoing (from the interaction point r = x − x = 0) wave, which will be the normal
−
physical choice, so that we would take B = 0. Gk is correspondingly an incoming wave.

Having found
eikr
Gk (r) = G+
k (r) =
r
5
then from (3) above the potential is given by
0
µ0 eik|x−x | 3 0
Z
A(x, ω) = J(x0 , ω) d x.
4π |x − x0 |

Thus
∞ ∞ 0
1 µ0 eik|x−x |
Z Z Z
0
A(x, t) = dω e−iωt d 3 x0 dt0 J(x0 , t0 )eiωt .
2π −∞ 4π −∞ |x − x0 |
The ω integration involves
∞ ikr ∞
1 −iω(t−t0 ) e 1 1
Z Z
dω e = dω eiω(r/c−τ ) = δ(τ − r/c),
2π −∞ r 2πr −∞ r

where
τ := t − t0
(and k = ω/c), using the representation of the delta function given earlier. Thus
∞
µ0 1 0
Z Z 
A(x, t) = d 3 x0 dt0 J(x0 , t0 ) δ t − (t − r/c)
4π −∞ r

or
µ0 1
Z
A(x, t) = d3 x0 J(x0 , t − r/c) . (13)
4π r
These are called the retarded potentials. Physically, what is happening is that values of the currents J at the
point x0 at time t − r/c affect or disturb the electromagnetic potentials A at the point x at time t, where
the disturbance travels at the speed of light, so that ct = |x − x0 | = r.
Similarly, choosing the solution G−k in the above arguments, one gets the advanced potentials, equation
(5) with t + r/c replacing t − r/c in J.

To summarise, we can say that if we define

δ(τ − r/c) 1  1 
G+ (x, t; x0 , t0 ) = = δ t − (t 0
+ |x − x 0
|)
r |x − x0 | c

then
µ0
Z Z
3 0
A(x, t) = d x dt0 G+ (x, t; x0 , t0 ) J(x0 , t0 ) (14)
4π
solves
1 ∂2
 
2
t
uA = ∇ − 2 2 A = −µ0 J.
c ∂t
In equation (14) above the “response”of A at (x, t), to the source J at (x0 , t0 ) is “propagated” by the Green
function G+ (x, t; x0 , t0 ).
A very similar analysis applies to solving

1
t
uΦ = − ρ
0

for the scalar potential Φ in the Lorentz gauge. Thus we have solved Maxwell’s equations, in the presence
of sources, using the vector and scalar potentials, and Green functions. In the following lectures we will see
how to perform an expansion of this solution in order to identify the different physical contributions.