Decomposition of Electromagnetic Fields

into Electromagnetic Plane Waves

Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544
(July 11, 2010; updated March 18, 2016)

Problem

Give a Fourier analysis of electromagnetic elds E(x, t) and B(x, t) in the half spaces z >
<0
(which are uniform, isotropic, nonconducting media with known free currents J(x, t)) in
terms of plane-wave solutions to Maxwells equations. This analysis can be interpreted as
a representation of these elds in terms of classical photons. In particular, consider the
elds of a charge q that moves with uniform velocity v = v z in a medium with index of

refraction n(), including the limit that v = 0 as well as the case that v > c/n (Cerenkov
radiation [1]), where c is the speed of light in vacuum.
For a static electric eld, such as that of a point charge q at the origin, E = q r/r2 (in
Gaussian units), relevant plane waves of the form ei(kxt) have zero frequency. Then, a
Fourier analysis of the scalar potential V = q/r has the Fourier expansion
V (x) =

q 
= Vk eikx d3 k,
r

(1)

where the Fourier coecient V is given by



Vk

1
q
d3 x
=
V (x) e
=
r
dr
eikr cos d cos
3
2
(2)
(2) 0
1

q
q
q
= 2
sin kr dr = 2 cos kr|
,
0 =
2 k 0
2 k
2 2k 2
ikx

(2)

on averaging to zero the rapid oscillations of the term cos kr for large r.
Then,
q
E = 2 r = V =
r

iq
q
eikx d3 k = 2
2
2
2 k
2

iq
k ikx 3
e
d k= 2
2
k
2

e
k

ikx

d3 k. (3)

This result suggests that the static Coulomb elds can be regarded as consisting of longitu eikx , of zero frequency. This description is appealing from a quantum
dinal plane waves, k
eikx with a longitudinal virtual photon (of mass
view, in which we identify the wave k
2
2
hk/c) ). However, this wave does not satisfy the free-space Maxwell
m given by m = (
equation E = 0, and so is somewhat unsatisfactory from a classical perspective.
In this classical problem, consider only plane waves that satisfy Maxwells equations as
the basis for the Fourier synthesis of the electromagnetic elds (and potentials).
To analyze as large a class of time-dependent elds as possible, consider plane waves
for which the wave vector k is complex, such that the plane waves for which Im(k) = 0
are physically signicant only close to their source charge-current distributions. Waves with
such limited spatial extent are often called evanescent.
1

Solution

This problem is a variant on the analysis of Huygens and Fresnel of a scalar eld (with time
dependence eit ) in terms of spherical waves generated by sources on a surface, typically
planar. Here, we seek a description of the vector electromagnetic elds E and B in terms of
plane waves rather than spherical waves.
The solution builds on the spirit of Weyls representation [2] of a scalar spherical wave
in terms of scalar plane waves,
i
eikr
=
r
2

 

i
ei(kx x+ky y+kz |z|)
dkx dky =
kz
2

where
kz =

k2

kx2

ky2

k2 k2 k2
x
y


i k2 + k2 k2
x

and1
k =

 

eik x
dkx dky ,
kz

if kx2 + ky2 k 2 ,
if kx2 + ky2 > k 2,

(kx , ky , kz )

(kx , ky , kz ) if z < 0.

if z 0,

(4)

(5)

(6)

The plane waves are homogeneous when kx2 + ky2 k 2 , but they are inhomogeneous (evanescent, and signicant only close to the plane z = 0) otherwise. The plane-wave decomposition
(4) is not spherically symmetric, which is a reminder that all plane waves (and especially
evanescent plane waves = classical virtual photons) are convenient mathematical ctions,
rather than entities with crisp physical reality.
The decomposition of three-dimensional scalar waves with a velocity c into homogenous
plane waves of the same velocity was perhaps rst considered by Stoney [4], and elaborated
upon by Whittaker [5]. Inclusion of inhomogeneous waves in the decomposition of scalar
waves was rst considered by Weyl [2]. The representation of electromagnetic elds in terms
of plane-wave solutions to Maxwells equations was perhaps rst considered by Clemmow [6]
and [7], which latter work this note follows.2

2.1

Temporal Fourier Analysis

As usual, a (possibly complex) scalar function f(x, t) has a temporal Fourier representation
f(x, t) =
where

1
f (x, ) =
2

f (x, ) eit d,


f(x, t) eit dt.

(7)

(8)

The notation k follows [3].

If ones attention is restricted to the fields far from the source, a somewhat simpler analysis obtains, as
discussed in [8] (which seems unaware of the full solution of [6, 7]).
2

It is customary to consider only positive angular frequencies, so by noting that f (x, ) =

f (x, ) we can also write
f(x, t) = 2Re


0

f (x, ) eit d.

(9)

Throughout the rest of this note we assume that such a temporal Fourier analysis can be
made for all relevant scalar functions.

2.2

Plane Electromagnetic Waves

We consider uniform, isotropic, nonconducting media with (frequency-dependent) relative

permittivity  and relative permeability . Maxwells equations for the Fourier components
E and B of the electric and magnetic elds in such media are
E =

4
,


B = 0,

E = ik0B ,

B =

4
J in2k0 E , (10)
c

where and J are the Fourier components of the free charge and current densities,
k0 =

,
c

(11)

and n =  is the index of refraction of the medium.

In regions with no free charge or current all six scalar components of E and B obey
the Helmholtz equation,
(12)
2 f + k 2 f = 0,
where

n
.
(13)
c
In general,  and are frequency dependent, which implies that they are complex functions.
However, at frequencies not close to the natural frequencies of the medium the imaginary
parts of  and are very small in many (transparent) media of interest, so we approximate
the index n and the wave number k as being purely real throughout the rest of this note.
Plane-wave solutions to the Helmholtz equation (12) have the form
k=

k0 = nk0 =

f (x) = f0 eikx ,

(14)

where the (possibly complex) wave vector k = Rek + iImk obeys

k 2 = k2 = kx2 + ky2 + kz2 .

(15)

The plane waves are homogeneous when Imk is zero, and inhomogeneous when Imk is
nonzero.
In general, any of the components (kx , ky , kz ) of the wave vector might be complex.
We consider the case that two components are real and one, kz say, is complex. In the
approximation that k is real, eq. (15) tells us that kz is real when kx2 + ky2 k 2, and purely
imaginary when kx2 + ky2 > k 2 . When kz is imaginary the plane waves are evanescent, dying
out in the z-coordinate, and propagating in x and y.
3

 = k/k, such that k

2 = 1. Using
We can introduce the (possibly complex) unit vector k
eq. (14) for components of the plane wave E,k , B,k we see that Maxwells equations (10)
(in regions with no free charge or current) imply that
,k = 0,
kE

k0
B,k

kE
B,k =
,
,k =
k
n

,k = 0,
kB

kB
,k = nE,k . (16)

E,k and B,k are mutually orthogonal, and |B | = n |E |.

Thus, k,
The time-average density of electromagnetic energy stored in a plane wave is
 |E,k |2 |B,k |2
 |E,k |2
u,k  =
+
=
.
16
16
8

(17)

The transport of energy by a plane wave is described by the Poynting vector, whose time
average is
c
cn
 E )] = c  |E,k |2 Rek
 = c u,k  .
Re(E,k H,k ) =
Re[E,k (k
,k
8
8
8n
n
(18)
 has no component in the z-direction, and their energy
In the case of evanescent waves, Rek
ows only perpendicular to the direction in which these waves die out.
there are two independent plane-wave soluFor each direction of the unit wave vector k
tions, commonly characterized by their polarization. We can, however, omit further discussion
of polarization and proceed to relate the plane waves to the source charges and currents.3
S,k  =

2.3

Four-Dimensional Fourier Analysis

We can augment the temporal Fourier analysis that determined the elds E , B and J
with 3-dimensional spatial Fourier analyses of the form
  

B (x) =

B,k eikx d3 k,

where

B,k =

1   
B (x) eikx d3 x,
(2)3

(19)

and the wave vector k is purely real. The Fourier components E,k , B,k and J,k are related
to one another by the transforms of Maxwells equations (10),
kE,k = 4i,k ,

kE,k = k0 B,k ,

kB,k = 0,

kB,k =

4i
J,k n2 k0 E,k .
c
(20)

For evanescent waves the condition that k2 be real implies that the vectors Rek and Imk are orthogonal.
These two directions, together with the direction of their cross product, can be taken as the basis for a
definition of the wave polarization. For example, taking Rek along the x-axis and Imk along the z-axis we
= k (cosh a, 0, i sinh a) for any real number a, whence k2 = k 2 .
can write k = k k
One polarization (called E-polarization by Clemmow [7]) has E perpendicular to both Rek and
Imk and B in their plane, E,E = (0, EE , 0) eikx. The other polarization (H-polarization) has
nE and E = k
B /n imply that B,E =
B,H = (0, nEH , 0) eikx. The relations B = k
(inEE sinh a, 0, nEE cosh a) eikx and E,H = (iEH sinh a, 0, EH cosh a) eikx . For both polarizations,
 , which provides a more detailed justification of eq. (18).
= n |E |2 Rek
Re(E B ) = nE 2 cosh a x

Combining these, we have

B,k =

4i k J,k
4i k J,k
4i
k J,k
,
=
=
2
2
2
2
2
2
c n k0 k
c k k
c kz (k 2 kx2 ky2)

(21)

and hence,
4i
B (x) =
c

  


kz

k J,k eikx d3 k
k 2 kx2 ky2

kz +

k 2 kx2 ky2

(22)

The expansion (22) expresses the magnetic eld in terms of mathematical plane waves, but
are these electromagnetic plane waves? As discussed in sec. 2.2, electromagnetic planes waves
must satisfy the conditions k 2 = k2 , eq. (15), and k B,k = 0, eq. (16), in a medium of
index n. The second, but not the rst of these conditions is met by the expansion (21)-(22).
Note that the denominator of the integrand of eq. (22) vanishes when the rst condition
holds. If we consider, say, the integration over kz to be a contour integration with the contour
completed at innity, then Cauchys
 integral theorem has the eect of enforcing the condition
2
2
2
(15). When kx + ky > k , kz = i kx2 + ky2 k 2 is pure imaginary and the plane waves are
evanescent. Physically, these waves should die out, rather than grow, with |z|, which is
insured when z > 0 by completing the contour on a semicircle at innity for positive
Im(kz )

2
and deforming the contour along the Re(kz ) axis to enclose the pole at kz = k kx2 ky2
(when this is real) but not that at the negative of this value (and completing the contour
on a semicircle at innity for negative Im(kz ) if z < 0). Using Cauchys integral theorem we
now obtain4
 
4 2
k J,k ik x
B (x) =
e
dkx dky ,
(23)
c
kz
where kz and k are dened in eqs. (5)-(6) and the sign holds for z >
< 0. Then, using the
last of eq. (16), E,k = k B,k /nk, to determine the electric eld of the plane wave we
nd,
 
4 2
k (k J,k ) ik x
E (x) =
e
dkx dky .
(24)
cnk
kz
The expansions (23)-(24) are the electromagnetic equivalent of the more familiar Fresnel
diraction integral5 for scalar elds, and have the advantage of being exact throughout all
space. Qualitatively, these expansions indicate that the elds in space are a kind of Fourier
transform of the source distribution, which is often restricted to an aperture in the plane
z = 0.

2.4

Surface Currents Only on the Plane z = 0

We can conrm the general result (23) by a dierent argument for the case that the currents
are conned to the plane z = 0. Then, the magnetic eld obeys the following symmetry with
4

This argument follows sec. 3.4 of [7], where the discussion is artificially restricted to currents on the
plane z = 0.
5
See, for example, http://www.fourieroptics.org.uk/fresnelDiff.html

respect to this plane,6

Bx (x, y, z) = Bx (x, y, z),

Bz (x, y, z) = Bz (x, y, z).

(25)
From consideration of small loops with areas normal to the x and y axes and which surround
a surface-current-density element K(x, y, 0, t), the fourth Maxwell equation and eq. (25)
imply that
Bx (x, y, 0+, t) =

By (x, y, z) = By (x, y, z),

2
Ky (x, y, 0, t),
c

By (x, y, 0+ , t) =

and

2
Kx (x, y, 0, t). (26)
c

Of course, the temporal Fourier components obey the same form,

B,x (x, y, 0+) =

2
K,y (x, y, 0),
c

B,y (x, y, 0+ ) =

and

2
K,x (x, y, 0).
c

(27)

The surface current density K can be further analyzed in a two-dimensional Fourier transform,
 
K,k ei(kx x+ky y) dkx dky ,
(28)
K (x, y, 0) =
where
K,k

1
=
(2)2

 

K (x, y, 0)ei(kx x+ky y) dx dy.

(29)

The expansion of B into homogeneous and inhomogeneous electromagnetic plane waves

(which latter die out with increasing |z|) has the general form
 

B (x) =
so that

B (x, y, 0+ ) =

B,k eik

 

dkx dky ,

(30)

B,k ei(kx x+ky y) dkx dky .

(31)

Then, equations (27) and (28) tell us that

B,k+ ,x =

2
K,k,y ,
c

B,k+ ,y =

2
K,k,x ,
c

(32)

and the condition (16) that k+ B,k+ = 0 leads to

B,k+ ,z =

2 kx K,k,y ky K,k,x
.
c
kz

(33)

Equations (32)-(33) can be combined into the form

B,k+ =

2 k+ K,k
2 k+ K,k+
=
c
kz
c
kz

(z > 0),

(34)

where we note that K,k = K,k+ = K,k since the Fourier component K,k does not
depend on kz .
6

See, for example, [9].

For z < 0, a similar argument for B (x, y, 0 ) leads to eqs. (34) with the substitution of
k for k+ . Hence,
2
B (x) =
c

 

k K,k ik x
e
dkx dky ,
kz

(35)

where kz and k are dened in eqs. (5) and (6) and the sign holds for z >
< 0. This is a
special case of the previous result (23), since the volume current density J corresponding to
the surface current density K on the plane z = 0 is
Jx = Kx (z),

Jy = Ky (z),

Jz = 0,

(36)

for which the Fourier transforms are

J,k ,x

1
=
(2)3

  

J,x (z) e

ik x

1
d x=
(2)3
3

 

K,x ei(kx x+ky y) dx dy =

1
K,k ,x,
2
(37)

etc.

2.5

Oscillating Electric Dipole

,
As a rst example, we consider a small oscillating electric dipole of moment p = p0 eit p
. We follow the usual method of Hertz (see, for example, sec. 9.2 of [10])
so that p = p0 p
which shows that, on integration by parts and using the continuity equation J = i ,


J d3 x =

x( J ) d3 x = i

x d3 x = ip .

(38)

For a point dipole at the origin we therefore write

J = ip 3 (x).
Then,
J,k

1
=
(2)3

  

J (x) eik

(39)

d3 x =

ip
.
(2)3

(40)

Taking the dipole to be in vacuum, its magnetic eld is given by eq. (23) as
ik
B (x) =
2
where k = k0 = /c, kz =

 

k p ik x
e
dkx dky = A ,
kz

(41)

k 2 kx2 ky2 and

kp
A (x) =
2

 

eik x
eikr
dkx dky = ikp
,
kz
r

(42)

recalling Weyls expansion (4). Thus, we obtain the standard form for the vector potential
of an oscillating point dipole (or conversely, we can use the standard form of the vector
potential together with eq. (42) to provide a proof of eq. (4)).
7

The corresponding plane-wave expansion of the electric eld follows from eq. (24),
E (x) =

i
2

 

k (k p ) ik x
e
dkx dky .
kz

(43)

The expansions (41)-(43) contain both homogeneous and inhomogeneous plane waves, which
is satisfactory in that the elds of an oscillating dipole contain energy in the near zone
that is exchanged with the source rather than radiated away. We can say that this energy
corresponds to the inhomogeneous plane waves, whose energy ow is the same in magnitude
for kx and for ky such that the average ow of energy in the inhomogeneous waves is
zero. However, the details of the characterization of the inhomogeneous waves is somewhat
unsatisfactory in that these depend on the arbitrary choice of orientation of the plane called
z = 0. To this author, there is very limited physical reality to the inhomogeneous plane
waves identied in the expansions (41)-(43).

2.6

Radiated Power

When considering the power radiated by a time-dependent current distribution, our expansion of the elds in plane waves (in which the plane z = 0 plays a special role) leads us to
evaluate the power crossing a plane with z > 0. Because the orientation of the x-y-z axes
is arbitrary, we obtain a general result. In this way we can deduce the total power radiated
into any half space, as discussed in sec. 3.1.2 of [7].
However, this does not provide a very detailed picture of the ow of radiated power. Far
from all current sources the radiated power ows radially outwards from the centroid of the
source, and the plane-wave expansion of the elds, together with appropriate approximations,
leads to the usual characterization of the power radiated into a specied solid angle, as
discussed in sec. 3.2 of [7]. Of course, in this approximation the spirit of the plane-wave
expansion is abandoned in favor of the more usual approach based on spherical waves.
Close to the sources (Fresnel zone in optics, near zone in antenna theory) the plane-wave
expansion provides an alternative description to the more usual formulation in terms of the
elds E (x) and B (x). However, because plane waves have, by denition, innite transverse
extent, the physical meaning of such waves in small volumes is ambiguous (to this author).
It remains that the standard description of the (time-average) ow of energy through an
electromagnetic eld is via the Poynting vector,
S =

c
Re(E H ),
8

(44)

c
Re(E H ).
8

(45)

and its temporal Fourier components,

S  =

Further decomposition of the elds, by spatial Fourier transforms, into electromagnetic plane
waves does not, in general, provide much additional physical insight as to the nature of the
ow of energy.
Of possible interest will be the plane-wave expansion of the elds of a charge with uniform
velocity.
8

2.7

Charge with Uniform Velocity

The rst calculation of the elds of a charge with uniform velocity may have been made by
Maxwell [11] who understood that the elds are fore-aft symmetric, but whose miscalculation
of asymmetric elds via Lorenz retarded potentials [12, 13] held up acceptance of this
powerful tool for many years. The rst published calculation of the elds of a charge with
v c was made by J.J. Thomson [14], and the rst calculation valid for any v < c was given
by Heaviside [15]. See also [16]. Shortly thereafter, Heaviside also calculated the elds for

motion with v > c [17], anticipating by many years what is now called the Cerenkov
eect
[1].
The plane-wave expansion of the elds of a uniformly moving charge is treated in sec. 7.2
of [7]. A somewhat earlier discussion was given in [18]. See also [19].
The current density of charge q that moves with velocity v along the x-axis can be written
.
J = qv (x vt) (y) (z) x

(46)

Then, its temporal Fourier transform is, recalling eq. (13),

qx
(y) (z)
J =
2

(x vt)eit v dt =

q eix/v x
q eickx/nv x
(y) (z) =
(y) (z).
2
2

(47)

The spatial Fourier transform of this is

J,k

qx
=
(2)4

  

eix/v (y) (z) eikx d3 x =

qx
(kx /v),
(2)3

(48)

noting that (k) = eikx dx/2. Using this in eq. (23), the temporal Fourier components
of the magnetic eld are given by
 

ik x
4 2
k x
q
(k

/v)
e
dkx dky
x
3
c
(2)
kz

ky z i(ky ykz z)
q eix/v kz y
=
e
dky ,
2c
kz

B (x) =

(49)

where
kz =

k 2 (1 c2 /n2 v 2) ky2 =

2 (n2 c2/v 2 )/c2 ky2 = i

 2 2
c /v n2 + ly2 ,
c

(50)

and ly = cky /. Recalling that E,k = k B,k /nk = k cB,k /n2 , the temporal
Fourier expansion of the electric eld follows from eq. (49) as
 
ky z ik x
q
kz y

(k

/v)
k

e
dkx dky
E (x) =
x
2n2
kz

+ ky y
kz z i(ky ykz z)
q eix/v kx (1 n2 v 2/c2 ) x
=
e
dky .
2v
kz

(51)

2.7.1

v < c/n

When the speed of the charge is less than the speed of light c/n in the surrounding medium,
kz is purely imaginary according to eq. (50), and all plane waves in the expansion (49) are
evanescent. No radiation (to innity [20]) is emitted by a charge moving uniformly at
sublight speed.
The eld B is axially symmetric, and azimuthal, with respect to the axis of motion, which
is the x-axis here. The azimuthal eld B (x, t) (about the x-axis) at distance r from the
x-axis can be evaluated as By (x, 0, r , t) using eq. (49),


B,y (x, 0, r ) eit d

0



q
ix/v (r /c) c2 /v2 n2 +l2y it
Re
dl

d
e
e
e
=
y
c2
0

q
dly
=
Re

2

2
r
ic(x vt)/r v c2 /v 2 n2 + l2

B = By (x, 0, r , t) = 2Re

q
2
r

ly2

2

dly




[c(x vt)/rv] + c /v n +
q
d

2i


 
2
r
dly l + i [c(x vt)/r v]2 + c2 /v 2 n2 2 
y

2 2

[c(x vt)/r v]2 + c2/v 2 n2 + ly2

2

[c(x vt)/r v] + c /v n +

where ly = cky /

ly2

ly =i

[c(xvt)/r v]2 +c2 /v2 n2

(1 n v /c )r
qv
,
2 3/2
c [(x vt)2 + (1 n2v 2/c2 )r
]

(52)

where the integral in ly was evaluated by completing the contour at +. This is the usual
result for the magnetic eld of a charge moving at constant, sublight speed.
For completeness, we also calculate the axially symmetric electric eld E(x, 0, r , t) in
the x-z plane. Comparing eqs. (49) and eq. (51), we infer that
Ez (x, 0, r , t) =

c
q
(1 n2 v 2/c2 )r
By (x, 0, r , t) =
.
2 3/2
v
 [(x vt)2 + (1 n2 v 2/c2 )r
]

(53)

The y-component of Faradays law tells us that

Ex(x, 0, r , t)
Ez (x, 0, r , t) 1 By (x, 0, r , t)
3q
(1 n2 v 2/c2 )2 (x vt)r
=
,

=
2 3/2
r
x
c
t
 [(x vt)2 + (1 n2 v 2/c2 )r
]
(54)
which integrates to
Ex (x, 0, r , t) =

(1 n2 v 2/c2 )(x vt)

q
.
2 3/2
 [(x vt)2 + (1 n2 v 2/c2 )r
]

(55)

Thus, the electric eld is radial with respect to the position of the charge, (x vt, 0, 0), and
is related to the magnetic eld by
nv
B.
(56)
E=
c
10

2.7.2

The Plane-Wave Spectrum of the Coulomb Field (v = 0)

In the limit that the speed v of the charge goes to zero the magnetic eld (52) vanishes,
while the electric eld becomes the Coulomb eld E(v = 0) = q r/r2 , where r is the distance
from the charge to the observer.
To deduce the electric eld from its plane-wave expansion (51) we write
E(x, t) = 2Re


0

E (x) eit d




ky z ik x it
kz y
q
d
e
e
Re dkx dky
(kx /v) k


kz
0
 
+ ky y
kz z ik x ikx vt
q
kx (1 n2 v 2/c2 ) x
=
e
e
dkx dky ,
(57)
Re

kz

where now
k=

n
kx nv
=
,
c
c

and

kz =

k 2 kx2 ky2 = i kx2 (1 n2 v 2/c2 ) + ky2 .

(58)

In the limit that v = 0, kz = i kx2 + ky2, k 2 = 0 = (k )2 and we obtain the expansion

 
+ ky y
kz z ik x
q
kx x
e
dkx dky
Re

kz

 
ik eik x
q

= Re
dkx dky .

kx2 + ky2

E(x, v = 0) =

(59)

The Fourier components Ek of this expansion obey the relation k Ek = 0 as v and

k go to zero. So we can say that the expansion (59) expresses the Coulomb eld in terms
of zero-frequency electromagnetic plane waves, all of which are evanescent since kz is pure
imaginary. This expansion is therefore conceptually superior to that of eq. (3) (although
eq. (59) suers from the arbitrariness of the choice of the plane z = 0.)
An expansion of the Coulomb eld in evanescent plane waves was perhaps rst given in
[21].
Presuambly the classical gravitational eld can be decomposed into gravitational plane
waves. Then, a static gravitational eld consists of zero-frequency plane waves (virtual
gravitons), which are not aected by the gravitational eld since they have zero frequency
(zero energy in the quantum view). Hence, the exterior gravitational eld of a black hole
can exist without being sucked into it. Similarly, a black hole can have an exterior static
electric eld consisting of zero-frequency (zero energy) virtual photons.
2.7.3

Cerenkov
Radiation: c/n < v < c

When a charge q moves with speed v greater than that of light, c/n, in a medium (but
with v < c, of course), the plane-wave expansion (49) contains both homogeneous and
inhomogeneous waves, and radiation is therefore emitted. This is sometimes considered to
be paradoxical in that the charge is not obviously accelerating. However, the radiation exists
only when the charge moves through a medium with index of refraction greater than 1, in
11

which case the charges in the medium are accelerated by the passing charge q, and we can say
that is the medium, rather than the charge itself, which emits the radiation. Of course, the
radiated energy must come from the charge itself, so there must be a (small) back reaction
of the medium on the passing charge, which decelerates the latter.
The temporal expansion of the magnetic eld is, from eq. (49),
q eix/v
B (x) =
2c
where
kx =

ck

=
v
nv

ky z i(ky ykz z)
kz y
e
dky ,
kz

and

kz =

k 2(1 c2/n2 v 2 ) ky2 .

(60)

(61)


For plane waves in the x-y plane, kz = 0 and ky = k 1 c2/n2 v 2 = kx (nv/c) 1 c2 /n2 v 2 =


kx n2 v 2/c2 1, which is real, so these waves are homogeneous, and carry energy away from


the charge q. Similarly, for plane waves in the x-z plane, ky = 0 and kz = kx n2 v 2/c2 1.
The wave vector k for the homogeneous waves (radiation eld) does not have a continuous
angular distribution, but always makes angle C to the y-z plane, where
tan C =

kx
kx
1
,
=
=
ky (kz = 0)
kz (ky = 0)
n2v 2/c2 1

so that
cos C = 

1
1 + tan2 C

c
.
nv

(62)

(63)

angle.
The angle C is the famous Cerenkov

Since k E,k = 0, the electric eld points only in a single direction, namely at the

Cerenkov
angle C to the negative x-axis (and the magnetic eld circles about the x-axis).
This eld conguration was rst depicted by Heaviside [17].
The temporal Fourier expansion of the electric eld follows from eq. (51) as
+ ky y
kz z i(ky ykz z)
q eix/v  kx (n2 v 2/c2 1) x
e
dky .
(64)
2v
kz
The electric eld in, say, the x-z plane for z > 0 consists of plane waves with ky = 0, so we
have that
E (x) =

E(x, 0, z > 0, t) = 2Re


0

E (x, 0, z > 0) eit d

q
+ z] ei[(x+z/ tan C )/vt] d
Re
[ tan C (n2 v 2/c2 1) x
=
v
0


2q
x
+
z/
tan

C
z]
t .
= [tan C (n2 v 2/c2 1) x
(65)
v
v

12

At time t = 0 the electric eld in the x-z plane for z > 0 is nonzero only along the line
z = x tan C , as shown in the gure. By a similar argument the magnetic eld in the z-z
plane is nonzero only along this line. The electric and magnetic elds are is azimuthally

symmetric, so the elds are nonzero only on the Cerenkov

cone. The present argument
predicts innite elds on this cone, whereas in reality the index n exceeds unity for only a
nite range of frequency, and the elds extend slightly outside the cone, and are nite.
To deduce the frequency spectrum of the radiated power, we rst note that the total
energy d2 U that crosses an area element dArea, integrated over all time, is


c
dArea
S dArea dt =
E B dt
dU =
4

 
c
dArea
E B eit d dt
=
4



c
c
dArea
E B d = dArea Re
E B d,
=
2

0
2

(66)

since E () = E () and B () = B (). Equal amounts of energy cross any plane at

z > 0 or at z < 0, so the total energy radiated is twice that which crosses a plane at z > 0,
2c
U=

 

dx dy z Re


0

(E B )z>0 d.

(67)

The energy radiated per unit frequency interval and per unit path length of the charges
motion along the x-axis is independent of x. Since Bx = 0, we have


2c
d2 U

=
dy Re(E,x B,y
)z>0
d dx

  
2c q q
kx (n2 v 2/c2 1) i(ky y+kz z) i(ky y+kz  z)
Re
=
e
e
dy dky dky
2v 2c
kz
 
i(kz kz  )z)
2c q q 2 2 2
 e
(n v /c 1) Re
=
2(ky ky )
dky dky
2v 2 2c
kz
 2Im(kz )z
q 2 2 2 2
e
=
(n
v
/c

1)
Re
dky
2
v
kz

 (n/c) 1c2 /n2 v2

q 2n2
dky
2
2 2
=
(1 c /n v )
2 2 2
2
2
2
2
c
(n/c) 1c /n v
( n /c )(1 c2 /n2 v 2) ky2


q 2
c2
=
1

,
c2
n2 v 2

(68)

where we note that in the fourth line the integrand is real only when kz is real. Equation

(68) is the standard result for the energy spectrum of Cerenkov

radiation [22], which has
the surprising feature (of little practical import) that a magnetic medium of index n emits

times as much Cerenkov

radiation as does a dielectric medium of the same index. As usual,
we note that the index n can be greater than unity for only a nite range of frequencies, so
that the total power radiated over all frequencies is nite.

13

The x-component of the electric eld at the charge is, using eq. (64),
Ex (vt, 0, 0, t) = 2Re
=


0

E,x (vt, 0, 0) eit d

q 2 2 2
d
(n v /c 1) Re
v 2

(69)


1
dky =
kz


0

q
c2
d 2 1 2 2 .
c
nv

This is a peculiar result in that we might have expected the electric eld to diverge at the
charge.7 The eld (69) acts opposite to the direction of the charges velocity and decelerates
it. The work done by the electron per unit path length is qEx, whose Fourier component
at frequency equals the energy radiated per unit path length. That is, the work done by

the electron on the Cerenkov

eld is transformed into the Cerenkov
radiation.
For additional discussion of the relation of radiation by moving charges to the plane-wave
decomposition of their elds, see [23].

2.8

TEM Waves in a Coaxial Transmission Line

We consider a (vacuum) coaxial cable centered on the x-axis (so that the currents are close
to the plane z = 0) with perfect conductors of radii a and b
> a. This cable supports TEM
waves with currents in a cylindrical coordinate system (r = y 2 + z 2 , , x),
I(r = a, x) = I(r = b, x) = I0 ei(kx 0 t),

(70)

where 0 = kc is the angular frequency of the waves, and elds

E(a < r < b) =

2I0 i(kx0 t)
r,
e
cr

B(a < r < b) =

2I0 i(kx0 t)
e
.
cr

(71)

The TEM elds are, of course, possible static elds multiplied by the waveform ei(kx0 t) .
Because the elds are zero for r < a and r > b they are not simply plane waves of the
form ei(kx0 t) .
To display the plane-wave expansion (23)-(24) of the TEM elds, we rst note that the
nonzero Fourier components of the current density are
  

J0 (x) eik x d3 x
3
(2)


1 2I0    ikx (r a) (r b)
ik x 3
e
e
dx

=
(2)3 c
2a
2b


  
(r

b)
dx
(r

a)
1 2I0

ei(kkx )x
ei(ky r cos kz r sin ) r dr d
=
(2)3 c
2
a
b
 2 

1 2I0
i(ky a cos kz a sin )
i(ky b cos kz b sin )
)
e

e
d.
(72)
(k

k
=
x
(2)3 c
0

J0 ,k =

Thus, kx = k. Then according to eq. (5),

kz =
7

k 2 kx2 ky2 = iky ,

For v < c/n, kz is pure imaginary and Ex(vt, 0, 0, t) = 0 at the charge according to eq. (69).

14

(73)

so that the electromagnetic plane-wave decomposition of the TEM wave in a coaxial cable
consists only of inhomogeneous, plane electromagnetic waves. This is consistent with the
usual view that an (innite) coaxial cable transmits waves along its axial direction but
does not radiate. Also, the waves in the transmission line are closely associated with the
conductors of the line, and are not free in the sense of homogeneous, plane electromagnetic
waves.
The analysis for TEM waves on an arbitrary two-conductor transmission line parallel to
the x-axis diers from the above only in the form of the double integral in eq. (72) over
the transverse (y-z) plane; eq. (73) holds in all cases, and the plane-wave expansion involves
only inhomogeneous (evanescent) electromagnetic plane waves.8 Similarly, the plane-wave
expansion for waves in/on cylindrical waveguides (including optical bers as wall as hollow
metallic guides) only involves inhogeneous plane electromagnetic waves.

2.9
2.9.1

DC Magnets
Solenoid

We consider a solenoid magnet of length l and radius a that carries uniform azimuthal current
per unit axial length I = B0 /0 , such that the interior, axial magnetic eld is b0 in the limit
of large l. Taking the z-axis to be that of the solenoid, which extends over 0 < z < l, the
current density is
J(x, t) =

B0

(r a)
0

(0 < z < l),

(74)

in a cylindrical coordinate system (r, , z). The temporal Fourier transform of this dc current
density is
J (x) =

B0
= B0 (r a)()( sin x
+ cos y
)
(r a)()
0
0

(0 < z < l).

(75)

That is, J has only a zero-frequency component. When computing the 4-dimensional
Fourier transform of the azimuthally symmetric function (75), we can dene the (half)plane
+ kz z. Thus,
= 0 to contain the wave vector k, which can then be written as k = kx x
  

J,k =
=
=
=

J (x) eik x d3 x
3
(2)
 
 l
B0 ()
+ cos y
) eikx r cos eikz z
r dr d dz (r a)( sin x
(2)30
0
2aB0 () eikz l/2 sin(kz l/2)  2
+ cos y
) eikx a cos
d ( sin x
(2)3 0 kz
0
iaB0()J1 (kx a)eikz l/2 sin(kz l/2)
.

y
(76)
2 20kz

In the quantum view, these inhomogeneous waves correspond to virtual photons, so the plane-wave
photons associated with TEM waves on a transmission line are not real, even when the wave velocity (of
the total wave) is c as for a vacuum transmission line.

15

The spectrum of plane waves of the (dc) solenoid magnet has only = 0, so k = 0 and
all waves are inhomogenous (virtual photons). The wave vector (kx , 0, kz ) has kx 1/2a
from the Bessel function J1(kx a), while |kz | ranges between 0 and 2/l according to the
factor sin(kz l/2)/kz .
2.9.2

Helical Wiggler

A so-called helical wiggler [24] made from a double helix winding on a cylinder of radius a
with period 2/k0 has a purely transverse magnetic eld along its axis given by9
sin k0 z),
B(0, 0, z) = B0 (
x cos k0 z + y

(77)

where the current I in each helical winding is

I=

B0
.
+ k0 K1 (k0 a)

k02 aK0(k0 a)

(78)

With some eort, the Fourier transform J,k could be computed exactly, but it is clear
that the virtual photons of the static magnetic eld have zero frequency/energy and have
kz = k0 in the limit of a long wiggler.

2.10

DC Current

As a nal example (which could well have been the rst, given its simplicity), we consider a
DC current density of any form. This current density may or may not be electrically neutral.
In any case, the temporal Fourier transform of the DC current has only a zero-frequency
component, so the plane-wave decomposition contains only inhomogeneous waves (virtual
photons). That is, the DC current does not radiate, although in a model of the current due
to moving charges, these charges are accelerated as they traverse the current loop.10

References

[1] P.A. Cerenkov,

Visible Radiation Produced by Electrons Moving in a Medium with
Velocities Exceeding That of Light, Phys. Rev. 52, 378 (1937),
http://physics.princeton.edu/~mcdonald/examples/EM/cerenkov_pr_52_378_37.pdf

[2] H. Weyl, Ausbreitung elektromagnetischer Wellen u

ber einem ebenen Leiter, Ann. Phys.
365, 481 (1919), http://physics.princeton.edu/~mcdonald/examples/EM/weyl_ap_365_481_19.pdf
[3] G.C. Sherman, A.J. Devaney and L. Mandel, Plane-Wave Expansions of the Optical
Field, Opt. Comm. 6, 115 (1972),
http://physics.princeton.edu/~mcdonald/examples/EM/sherman_oc_6_115_72.pdf
9

See, for example, sec. 2.3 of [25].

This result is implicit in the first-ever computation of electromagnetic radiation, by FitzGerald (1883)
[26], and was verified for circular currents by J.J. Thomson in 1903 [27] as a possible model for the ground
state of an atom.
10

16

[4] G.J. Stoney, Discussion of a New Theorem in Wave Propagation, Phil. Mag. 43, 273
(1897), http://physics.princeton.edu/~mcdonald/examples/EM/stoney_pm_43_273_97.pdf
How to apply the Resolution of Light into Uniform Undulations of Flat Wavelets to the
Investigation of Optical Phenomena, Phil. Mag. 5, 264 (1903),
http://physics.princeton.edu/~mcdonald/examples/EM/stoney_pm_5_264_03.pdf

Flat-Wavelet Resolution, Phil. Mag. 9, 577 (1905),

http://physics.princeton.edu/~mcdonald/examples/EM/stoney_pm_9_577_05.pdf

[5] E.T. Whittaker, On the partial dierential equations of mathematical physics, Math.
Ann. 57, 333 (1903), http://physics.princeton.edu/~mcdonald/examples/EM/whittaker_ma_57_333_02.pdf
[6] H.G. Booker and P.C. Clemmow, The Concept of an Angular Spectrum of Plane Waves,
and Its Relation to that of Polar Diagram and Aperture Distribution, Proc. IEE 97, 11
(1950), http://physics.princeton.edu/~mcdonald/examples/EM/booker_piee_97_11_50.pdf
[7] P.C. Clemmow, The Plane Wave Spectrum Representation of Electromagnetic Fields
(Pergamon, Oxford, 1966).
[8] H.A. Haus, On the radiation from point charges, Am. J. Phys. 547, 1126 (1986),
http://physics.princeton.edu/~mcdonald/examples/EM/haus_ajp_54_1126_86.pdf

[9] K.T. McDonald, Symmetries of Electromagnetic Fields Associated with a Plane Conducting Screen (Jan. 14, 2012),
http://physics.princeton.edu/~mcdonald/examples/emsymmetry.pdf

[10] J.D. Jackson, Classical Electrodynamics, 3rd ed. (Wiley, New York, 1999).
[11] J.C. Maxwell, On a Method of making a Direct Comparison of Electrostatic with Electromagnetic Force; with a Note on the Electromagnetic Theory of Light, Phil. Trans.
Roy. Soc. London 158, 643 (1868),
http://physics.princeton.edu/~mcdonald/examples/EM/maxwell_ptrsl_158_643_68.pdf

[12] L. Lorenz, Ueber die Identitat der Schwingungen des Lichts mit den elektrischen
Stromen, Ann. Phys. 207, 243 (1867),
http://physics.princeton.edu/~mcdonald/examples/EM/lorenz_ap_207_243_67.pdf

On the Identity of the Vibration of Light with Electrical Currents, Phil. Mag. 34, 287
(1867), http://physics.princeton.edu/~mcdonald/examples/EM/lorenz_pm_34_287_67.pdf
[13] K.T. McDonald, Maxwells Objection to Lorenz Retarded Potentials (Oct. 26, 2009),
http://physics.princeton.edu/~mcdonald/examples/maxwell.pdf

[14] J.J. Thomson, On the Electric and Magnetic Eects produced by the Motion of Electried Bodies, Phil. Mag. 11, 229 (1881),
http://physics.princeton.edu/~mcdonald/examples/EM/thomson_pm_11_229_81.pdf

[15] O. Heaviside, On the Electromagnetic Eects due to the Motion of Electrication

through a Dielectric, Phil. Mag. 27, 324 (1889),
http://physics.princeton.edu/~mcdonald/examples/EM/heaviside_pm_27_324_89.pdf

17

[16] J.J. Thomson, On the Magnetic Eects produced by Motion in the Electric Field, Phil.
Mag. 28, 1 (1889), http://physics.princeton.edu/~mcdonald/examples/EM/thomson_pm_28_1_89.pdf
[17] O. Heaviside, Electrical Papers, Vol. 2 (Macmillan, New York, 1894), p. 496 (reprinted
from The Electrician, Sept. 6, 1889, p. 458),
http://physics.princeton.edu/~mcdonald/examples/EM/heaviside_electrical_papers_2.pdf

[18] G. Toraldo di Francia, On the Theory of Some Cerenkovian

Eects, Nuovo Cim. 16, 61
(1960), http://physics.princeton.edu/~mcdonald/examples/EM/toraldodifrancia_nc_16_61_60.pdf
[19] R. Asby and E. Wolf, Evanescent Waves and the Electromagnetic Field of a Moving
Charged Particle, J. Opt. Soc. Am. 61, 52 (1971),
http://physics.princeton.edu/~mcdonald/examples/EM/asby_josa_61_52_71.pdf

[20] K.T. McDonald, On the Denition of Radiation by a Systems of Charges (Sept. 6, 2010),
http://physics.princeton.edu/~mcdonald/examples/radiation.pdf

[21] J.L. Agudn and A.M. Platzeck, On the Limit for 0 of the Toraldo di Francia
Expansion, Lett. Nuovo Cim. 22, 523 (1978),
http://physics.princeton.edu/~mcdonald/examples/EM/agudin_lnc_22_523_78.pdf

[22] I. Tamm, Radiation Emitted by Uniformly Moving Electrons, J. Phys. (USSR) 1, 439
(1939), http://physics.princeton.edu/~mcdonald/examples/EM/tamm_jpussr_1_439_39.pdf
[23] J.D. Lawson, Cherenkov Radiation, Physical and Unphysical, and its Relation to
Radiation from an Accelerated Electron, Am. J. Phys. 33, 1002 (1965),
http://physics.princeton.edu/~mcdonald/examples/EM/lawson_ajp_33_1002_65.pdf

Guided Surface Waves and Radiation from Charges Moving in Curved Paths, Am. J.
Phys. 34, 601 (1966),
http://physics.princeton.edu/~mcdonald/examples/EM/lawson_ajp_34_601_66.pdf

Some Attributes of Real and Virtual Photons, Contemp. Phys. 11, 575 (1970),
http://physics.princeton.edu/~mcdonald/examples/QED/lawson_cp_11_575_70.pdf

A synthetic approach to some electromagnetic wave phenomena, Eur. J. Phys. 5, 104

(1984), http://physics.princeton.edu/~mcdonald/examples/EM/lawson_ejp_5_104_84.pdf
[24] B.M. Kincaid, A short-period helical wiggler as an improved source of synchrotron
radiation, J. Appl. Phys. 48, 2684-2691 (1977),
http://physics.princeton.edu/~mcdonald/examples/accel/kincaid_jap_48_2684_77.pdf

[25] K.T. McDonald and H. Mitter, The Helical Wiggler (Oct. 12, 1986),
http://physics.princeton.edu/~mcdonald/examples/helical.pdf

[26] K.T. McDonald, FitzGeralds Calculation of the Radiation of an Oscillating Magnetic

Dipole (June 20, 2010), http://physics.princeton.edu/~mcdonald/examples/fitzgerald.pdf
[27] K.T. McDonald, Why Doesnt a Steady Current Loop Radiate? (Dec. 1, 2001),
http://physics.princeton.edu/~mcdonald/examples/steadycurrent.pdf

18