Chapter 1

Electrostatics I. Potential due to

Prescribed Charge Distribution,
Dielectric Properties, Electric Energy
and Force
1.1

Introduction

In electrostatics, charges are assumed to be stationary. Electric charges exert force on other charges
through Coulombs law which is the generic law in electrostatics. For a given charge distribution,
the electric eld and scalar potential can be calculated by applying the principle of superposition.
Dielectric properties of matter can be analyzed as a collection of electric dipoles. Atoms having
no permanent dipole moment can be polarized if placed in an electric eld. Most molecules have
permanent dipole moments. In the absence of electric eld, dipole moments are oriented randomly
through thermal agitation. In an electric eld, permanent dipoles tend to align themselves in the
direction of the applied eld and weaken the eld. Some crystals exhibit anisotropic polarization and
the permittivity becomes a tensor. The well known double diraction phenomenon occurs through
deviation of the group velocity from the phase velocity in direction as well as in magnitude.

1.2

Coulombs Law

Normally matter is charge-neutral macroscopically. However, charge neutrality can be broken

relatively easily by such means as mechanical friction, bombardment of cosmic rays, heat (e.g.,
candle ame is weakly ionized), etc. The rst systematic study of electric force among charged
bodies was made by Cavendish and Coulomb in the 18th century. (Cavendishs work preceded
Coulombs. However, Coulombs work was published earlier. Lesson: it is important to publish
your work as early as possible.) The force to act between two charges, q1 at r1 and q2 at r2 ; follows

the well known Coulombs law,

Figure 1-1: Repelling Coulomb force between like charges.

F = const.

q1 q2
q1 q2
= const.
; (N)
2
r
jr1 r2 j2

(1.1)

where r = jr1 r2 j is the relative distance between the charges. In CGS-ESU (cm-gram-second
electrostatic unit) system, the constant is chosen to be unity. Namely, if two equal charges q1 =
q2 = q separated by r = 1 cm exert a force of 1 dyne = 10 5 N on each other, the charge is dened
as 1 ESU ' 13 10 9 C. The electronic charge in ESU is e = 4:8 10 10 ESU. In MKS-Ampere
(or SI) unit system, 1 Coulomb of electric charge is dened from
1 Coulomb = 1 Ampere 1 sec,
where 1 Ampere of electric current is dened as follows. If two innitely long parallel currents of
equal amount separated by 1 meter exert a force per unit length of 2 10 7 N/m on each other
(attracting if the currents are parallel and repelling if antiparallel), the current is dened to be 1
Ampere. Since the force per unit length to act on two parallel currents I1 and I2 separated by a
distance d is given by
F
I1 I2
= B1 I2 = 0
; (N/m)
(1.2)
l
2 d
the magnetic permeability 0 = 4 10 7 H/m in MKS-Ampere unit system is an assigned constant
introduced to dene 1 Ampere current. (The permittivity "0 is a measured constant that should
be determined experimentally.)
In MKS-Ampere unit system, the measured proportional constant in Coulombs law is approximately 9:0 109 N m2 C 2 : It is customary to write the constant in the form
const. =
2

1
;
4 "0

Figure 1-2: In MKS unit system, I = 1 Ampere current is dened if the force per unit length between
10 7
innite parallel currents 1 m apart is 2 10 7 N/m. The magnetic permeability 0 = 4
H/m is an assigned constant to dene 1 Ampere current.

where
"0 = 8:85

10

12

C2
N m2

Farad
m

is the vacuum permittivity. Since

c2 =

1
"0

; m2 s

the permittivity "0 can be deduced from the speed of light in vacuum that can be measured
experimentally as well.

1.3

Coulomb Electric Field and Scalar Potential

The Coulombs law may be interpreted as a force to act on a charge placed in an electric eld
produced by other charges, since the Coulomb force can be rewritten as
1 q1 (r2 r1 )
4 "0 jr1 r2 j3
= q2 Eq1 ; (N)

F =

where
Eq1 =

1 q1 (r2 r1 )
;
4 "0 jr1 r2 j3

q2

N
V
=
C
m

(1.3)

(1.4)

is the electric eld produced by the charge q1 at a distance r1 r2 : The eld is associated with the
charge q1 regardless of the presence or absence of the second charge q2 : (The factor 4 is the solid
3

angle pertinent to spherical coordinates. If it is ignored as in the CGS-ESU system, it pops up in

planar coordinates. Corresponding Maxwells equation is
r E=

"0

; in MKS-Ampere,

or
r E=4

; in CGS-ESU.

Which unit system to choose is a matter of conveneince and it is nonsense to argue one is superior
to other.)
Since two charges q1 and q2 exert the Coulomb force on each other, assembling a two-charge
system requires work. A dierential work needed to move the charge q2 against the Coulomb force
is
1 q1 q2 (r2 r1 )
dW = F dr2 =
dr2 :
4 "0 jr1 r2 j3
Noting
rr2

1
jr1

r2 j

=+

r1
jr1

r2
;
r2 j3

integration from r2 = 1 to r2 can be readily carried out,

W =

1
q1 q2
; (J)
4 "0 jr1 r2 j

(1.5)

This is the potential energy of two-charge system. Note that it can be either positive or negative. If
W is positive, the charge system has stored a potential energy that can be released if the system is
disassembled. Energy released through nuclear ssion process is essentially of electrostatic nature
associated with a system of protons closely packed in a small volume. On the other hand, if W
is negative, an energy jW j must be given to disassemble the system. For example, to ionize a
hydrogen atom at the ground state, an energy of 13:6 eV is required. In a hydrogen atom, the total
system energy is given by
1
W = mv 2
2
where rB = 5:3

10

11

e2
=
4 "0 rB

e2
< 0;
8 "0 rB

m is the Bohr radius and

1
e2
mv 2 =
= 13:6 eV,
2
8 " 0 rB

is the electron kinetic energy. The electric potential energy is

e2
=
4 "0 rB

27:2 eV.

The potential energy of two-charge system may be written as

W =

q1
1
4 "0 jr1 r2 j

q2 ; (J)

which denes the potential associated with a point charge q1 at a distance r;

=

1 q
; (J C
4 "0 r

= V);

(1.6)

and the electric eld is related to the potential through

E=

r ; (V m

):

(1.7)

Both potential and electric eld E are created by a charge q regardless of the presence or absence
of a second charge.

1.4

Maxwells Equations in Electrostatics

In electrostatics, electric charges are stationary being held by forces other than of electric origin
such as molecular binding force. Since charges are stationary, no electric currents and thus no
magnetic elds are present, B = 0:
For a distributed charge density (r), the electric eld can be calculated from
1
E(r) =
4 "0

(r0 )(r r0 ) 0
dV =
jr r0 j3

1
r
4 "0

(r0 )
dV 0 ;
jr r0 j

(1.8)

since the dierential electric eld due to a point charge dq = (r0 )dV 0 located at r0 is
dE =
Note that
r

1
jr

r0 j

1 (r0 )dV 0
(r
4 "0 jr r0 j3

r0

1
jr

r0 j

r0 ):

r
jr

r0
;
r0 j3

where r0 is the gradient with respect to r0 . Eq. (1.8) allows one to calculate the electric eld for a
given charge density distribution (r):
In order to nd a dierential equation to be satised by the electric led, we rst note that the
surface integral of the electric eld
I
E dS;

can be converted to a volume integral of the divergence of the eld (Gaussmathematical theorem,

not to be confused with Gausslaw),

I

E dS =
=

However, the function 1= jr

r0 j satises a singular Poissons equation,

r2

Therefore,

r EdV
Z
Z
(r0 )
1
2
dV r
dV 0 :
0j
4 "0 V
jr
r
0
V

1
E dS =
"0
S

1
r0 j

jr

dV

(r

(r

1
r ) (r )dV =
"0
0

V0

r0 ):

(1.9)

(r)dV:

(1.10)

This is known as Gauss law for the electric eld. Note that Gauss law is a consequence of the
Coulombs inverse square law. From
Z
Z
1
r EdV =
(r)dV;
(1.11)
"0
it also follows that
r E=

"0

(1.12)

This is the dierential form of the Gausslaw for the longitudinal electric eld and constitutes one
of Maxwells equations.
For the electric eld due to a static charge distribution,
1
E(r) =
4 "0

(r0 )(r r0 ) 0
dV =
jr r0 j3

1
r
4 "0

(r0 )
dV 0 ;
jr r0 j

its curl identically vanishes because

r

E=

1
r
4 "0

(Note that for any scalar function F; r

electrostatics is

rF
r

(r0 )
dV 0
jr r0 j

0:

0:) Therefore, the second Maxwells equation in

E = 0:

(1.13)

This is a special case of the more general Maxwells equation

r

E=

@B
;
@t

(1.14)

which determines the transverse component of the electric eld. Evidently, if all charges and elds
are stationary, there can be no magnetic eld. It should be remarked that a vector eld can be
uniquely determined only if both its divergence and curl are specied. (This is known as Helmholtzs
6

theorem.) In electrodynamics, two vector elds, the electric eld E and magnetic eld B, are to be
found for given charge and current distributions (r; t) and J(r; t): Therefore it is not accidental
that four Maxwells equations emerge specifying the four functions r E; r E; r B; and r B
which determine the longitudinal and transverse components of E and B.
Digression: A vector A can be decomposed into the longitudinal and transverse components Al
and At which are, by denition, characterized by
r

Al = 0; r At = 0:

The longitudinal component can be calculated from

1
r
4

Al (r) =
since
r2

1
jr

r0 j

r0 A (r0 ) 0
dV ;
jr r0 j
4

r0 :

The transverse component is given by

At (r) = A (r)
1
r
=
4

Al (r)
Z
A (r0 )
r
dV 0 ;
jr r0 j

where the identity

r

A =rr A

r2 A;

is exploited. It is known that for a vector to be uniquely dened, its divergence and curl have to
be specied. Since r A =r Al ; divergence of a vector spcies the divergence of the longitudinal
component. Likewise, r A =r At specifes the curl of the tarnverse component. As a concrete
example, let us consider the electric eld. Its divergence is
r E=

"0

and the solution to this dierential equation is

1
r
4 "0

El (r) =

(r0 )
dV 0 :
jr r0 j

The tranverse component is specied by the magnetic eld as

r

E=

@B
:
@t

The vanishing curl of the static electric eld allows us to write the electric eld in terms of a

gradient of a scalar potential

,
E=

r ;

(1.15)

for the curl of a gradient of a scalar function identically vanishes,

r

0:

For a given charge distribution, the potential has already been formulated in Eq. (1.8),
1
(r) =
4 "0
In terms of the scalar function

(r0 )
dV 0 :
jr r0 j

(1.16)

(r); Eq. (1.12) can be rewritten as

r2

"0

(1.17)

which is known as Poissons equation. In general, solving the scalar dierential equation for is
easier than solving vector dierential equations for E.
The physical meaning of the scalar potential is the amount of work required to move adiabatically a unit charge from one position to another. Consider a charge q placed in an electric eld E.
The force to act on the charge is F = qE and if the charge moves over a distance dl; the amount
of energy gained (or lost, depending on the sign of qE dl) by the charge is
qE dl =

qr

dl:

Therefore, the work to be done by an external agent against the electric force to move the charge
from position r1 to r2 is
W =

r2

r1

E dl = q

r2

r1

dl = q [ (r2 )

(r1 )] ;

(1.18)

where (r2 )
(r1 ) is the potential di erence between the positions r1 and r2 : The work is
independent of the choice of the path of integration. The potential is a relative scalar quantity and
a constant potential can be added or subtracted without aecting the electric eld.
In a conductor, an electric eld drives a current ow according to the Ohms law,
J = E; (A m

(1.19)

where (S m 1 ) is the electrical conductivity. In static conditions (no time variation and no ow
of charges), the electric eld in a conductor must therefore vanish. A steady current ow and
constant electric eld can exist in a conductor if the conductor is a part of closed electric circuit.
However, such a circuit is not static because of the presence ow of charge. For the same reason,
a charge given to an isolated conductor must entirely reside on the outer surface of the conductor.
Since E (static) = 0 in a conductor, the volume charge density must also vanish according to
8

= "0 r E = 0: A charge given to a conductor can only appear as a surface charge (C m 2 )

on the outer surface. The time scale for a conductor to establish such electrostatic state may be
estimated from the charge conservation equation
@
+ r J = 0:
@t

(1.20)

The current density can be estimated from the equation of motion for electrons
m

@v
=
@t

eE

m e v;

(1.21)

where c (s 1 ) is the electron collision frequency with the lattice ions. Noting J =
the density of conduction electrons, we obtain
@
+
@t

J=

ne2
E:
m

nev with n

(1.22)

Therefore, the equation for the excess charge density in a conductor becomes
@
@t

@
+
@t

+ ! 2p = 0;

! 2p =

ne2
;
"0 m

(1.23)

which describes a damped plasma oscillation 0 e t i!p t with an exponential damping factor =
13
c =2: The typical electron collision frequency in metals is of order c ' 10 = sec and the time
constant to establish electrostatic state is indeed very short. (If the low frequency Ohms law is
used, an unrealistically short transient time emerges,
@
+
@t

= 0;
0

0e

t=

where = 0 = ' 10 19 sec.)

The principle that a charge given to a conductor must reside entirely on its outer surface is also
a consequence of Coulombs 1=r2 law. For a conductor of an arbitrary shape, the surface charge
distribution is so arranged that the electric eld inside the conductor vanishes everywhere. In Fig.
1-3, one conducting spherical shell surrounds a smaller one. The spheres are connected through
a thin wire. A charge originally given to the inner sphere will end up at the outer surface of the
larger sphere and the electric eld inside should vanish if Coulombs law is correct. Experiments
have been conducted to measure residual electric elds inside and it has been established that the
power in the Coulombs law F _ 1=r is indeed very close to 2 within one part in 1016 : is
probably exactly equal to 2.0. If not, there would be a grave consequence that a photon should
have a nite mass. This is because if a photon has a mass mp ; corresponding Compton wavenumber
kC = mp c=~ will modify the wave equation for all potentials, including the Coulomb potential, in
the form
1 @2
2
=
:
(1.24)
r2 kC
c2 @t2
"0
9

Figure 1-3: Charge initially given to the inner conductor will all end up at the outer surface of the
outer conductor. The absence of the electric eld inside the sphere is a consequence of Coulombs
law.

In static case @=@t = 0; this reduces to

r2

2
kC

"0

(1.25)

and for a point charge (r) = q (r); yields a Debye or Yukawa type potential,
(r) =
and electric eld
Er =

1 q
exp ( kC r) ;
4 "0 r

q
4 "0

1
kC
+
2
r
r

kC r

(1.26)

(1.27)

The power in the Coulombs law Er _ q=ra experimentally established is

2 = O(10 16 );
and this corresponds to an upper limit of photon mass of mp < 10 51 kg. The lower limit of
photon Compton wavelength is C > 4 105 km at which distance a signicant deviation from the
Coulombs law, if any, is expected. (Incidentally, in a plasma, the static scalar potential does have
a Debye form,
1 q
(r) =
exp ( kD r) ;
(1.28)
4 "0 r
where
kD =

ne2
;
"0 T

(1.29)

is the inverse Debye shielding distance, n is the plasma density and T (in Joules) is the plasma

10

temperature. Likewise, in a superconductor, static magnetic eld obeys

r2

kL2 B = 0;

where kL = ! p =c is the London skin depth with ! p =

The potential energy of a two-charge system is
W =

1
q1 q2
=2
4 "0 jr1 r2 j

where

(1.30)

ne2 ="0 me the electron plasma frequency.)

q1 q2
1 1
;
2 4 "0 jr1 r2 j

q1 q2
1 1
;
2 4 "0 jr1 r2 j

(1.31)

(1.32)

is the potential energy associated with each charge. For many charge system, this can be generalized
in the form
1 1 X qi qj
1X
;
(1.33)
W =
j qj =
2
2 4 "0
jri rj j
j

i6=j

where
j

1 X
qi
;
4 "0
jri rj j

(1.34)

i6=j

is the potential at the location of charge qj due to all other charges. For distributed charge with
a local charge density (r) (C/m3 ); the potential energy of a dierential charge dq = dV , which
can be regarded as a point charge if dV is su ciently small, is
dW =
Since

= "0 r E;

1
1
dq =
2
2

dV:

1
dW = "0 r EdV:
2

Integrating over the entire volume, we nd

Z
1
W =
"0
r EdV
2
V
Z
1
=
"0
[r ( E) (r )E] dV
2
V
I
Z
1
1
=
"0
E dS +
"0 E 2 dV:
2
2
S
V
where use is made of Gausstheorem,
Z

r FdV =

F dS;

(1.35)

(1.36)

(1.37)

for an arbitrary well dened vector eld F. The surface integral vanishes because at innity, both
potential and electric eld vanish. Therefore, the potential energy associated with a distributed

11

charge system is
W =

1
"0 E 2 dV;
2

(1.38)

which is positive denite. In the expression for the potential energy of discrete charge system,
the potential energy due to a point charge itself is excluded while in the integral form, spatial
distribution of charge is assumed even for point charges and the so-called self energy is included.
The quantity
1
ue = "0 E 2 ; (J/m3 )
(1.39)
2
is the electric energy density associated with an electric eld. This expression for the energy density
holds regardless of the origin of the electric eld, and can be used for elds induced by time varying
magnetic eld as well.
The self-energy of an ideal point charge evidently diverges because the electric eld proportional
to 1=r2 does in the limit of r ! 0. However, if the electron is assumed to have a nite radius re ;
the self-energy turns out to be of the order of
We '

1 e2
:
4 " 0 re

(1.40)

This should not exceed the rest energy of the electron me c2 : Equating these two, the following
estimate for the electron radius emerges,
re '

1 e2
= 2:85
4 "0 mc2

10

15

m.

(1.41)

Although this result should not be taken seriously, the scattering cross section of free electron
placed in an electromagnetic wave (the process known as Thomson scattering) does turn out to be
=

8 2
r ;
3 e

and the concept of electron radius is not totally absurd. For proton whose radius is also of order
10 15 m, the analogy obviously breaks down.
The Poissons equation in Eq. (1.17) is the Eulers equation to make the energy-like integral
stationary
Z
Z
1
1
2
2
U=
E
dV
=
dV:
(1.42)
0
0 (r )
2
2
Indeed, the variation of this integral
U

=
=

( 0r
Z

0r

12

) dV
+

dV;

becomes stationary ( U = 0) when the Poissons equation holds,

r2 +

"0

= 0:

In other words, electrostatic elds are realized in such a way that the total electric energy becomes
minimum. In general, electric force act so as to reduce the energy in a closed (isolated) system.
Macroscopically, electric force tends to increase the capacitance as we will see in Chapter 2.
If the charge density distribution is known, the potential (r) can be readily found as a solution
of the Poissons equation. However, if the charge density is unknown a priori, as in most potential
boundary value problems, the potential and electric eld must be found rst. Then the charge
density is to be found from = "0 r E; or in the case of surface charge on a conductor surface,
= "0 En ; (C m

);

where En is the electric eld normal to the conductor surface.

1.5

Formal Solution to the Poissons Equation

As shown in the preceding section, if the spatial distribution of the charge density (r) is given,
the solution for the Poissons equation
r2 (r) =
can be written down as
(r) =

1
4

(r)

(1.43)

(r0 )
dV 0 :
jr r0 j

(1.44)

This is understandable because the dierential potential due to a point charge dq = dV is

d

1 (r0 )dV 0
;
4 "0 jr r0 j

and the solution in Eq. (1.44) is a result of superposition.

It is noted that the Poissons equation for the scalar potential still holds even for time varying
charge density,
(r; t)
r2 C (r; t) =
;
(1.45)
"0
if one employs the Coulomb gauge characterized by the absence of the longitudinal vector potential
r A = 0:

13

(1.46)

due to a point charge dq = dV 0 .

Figure 1-4: Dierential potential d

In the Coulomb gauge, the transverse vector potential satises the wave equation,
1 @2
c2 @t2

r2

At =

0 Jt ;

(1.47)

where Jt is the transverse current. The solution for the scalar potential in the Coulomb gauge is
non-retarded,
Z
1
(r0 ; t)
dV 0 ;
(1.48)
C (r; t) =
4 0 V jr r0 j
and so is the resultant longitudinal electric eld,
El (r; t) =

1
=
4 "0

r
jr

r0
0
0
3 (r ; t)dV :
0
rj

(1.49)

Such non-retarded (instantaneous) propagation of electromagnetic disturbance is clearly unphysical

and should not exist. In fact, the non-retarded Coulomb electric eld is exactly cancelled by a term
contained in the retarded transverse electric eld
Z
@At
Jt (t
) 0
jr r0 j
0 @
Et (r; t) =
=
dV
;
=
;
@t
4 @t
jr r0 j
c
as will be shown in Chapter 6.
The potential in Eq. (1.44) is in the form of convolution between the function
G(r; r0 ) = G(r

r0 ) =

14

1
;
4 jr r0 j

(1.50)

and the source function (r0 )= 0 : The function G(r; r0 ) is called the Greens function and introduced
as a solution to the following singular Poissons equation
r2 G =

r0 );

(r

(1.51)

subject to the boundary condition that G = 0 at r = 1: (G subject to this boundary condition

is called the Greens function in free space. It is a particular solution to the singular Poissons
equation. Later, we will generalize the Greens function so that it vanishes on a given closed
surface by adding general solutions satisfying r2 G = 0:) It is evident that the Greens function
is essentially the potential due to a point charge, for the charge density of an ideal point charge
(without spatial extension) q located at r0 can be written as
c

r0 ):

= q (r

(1.52)

Here (r r0 ) is an abbreviation for the three dimensional delta function. For example, in the
cartesian coordinates,
(r r0 ) = (x x0 ) (y y 0 ) (z z 0 );
(1.53)
in the spherical coordinates (r; ; ),
(r

r0 ) =

(r r0 ) [r(
(r r0 )
(
rr0 sin
(r r0 )
(cos
rr0

=
=

0
0

)] [r sin (
0

) (

)]

(1.54)
(1.55)

cos 0 ) (

);

(1.56)

and in the cylindrical coordinates ( ; ; z),

(r

r0 ) =

(
(

) [ (

0)

z0)

)] (z

z 0 ):

) (z

(1.57)
(1.58)

The singular Poissons equation

r2 G =

(r

r0 );

(1.59)

can be solved formally as follows. Let the Fourier transform of G(r

g(k) =

Its inverse transform is

G(r

r)=

G(r

1
(2 )3

r0 )e
Z

ik (r r0 )

g(k)eik (r

dr:

r0 ) 3

Noting that the operator r is Fourier transformed as ik and (r

15

r0 ) be g(k); namely,
(1.60)

d k:

(1.61)

r0 ) as unity, we readily nd from

Eq. (1.59),
g(k) =

1
:
k2

(1.62)

Substitution into Eq. (1.61) yields

0

G(r

r) =
=
=
=

where the polar angle

r r0 ; and

1 ik (r r0 ) 3
e
d k
(2
k2
Z 2
Z 1 Z
1
ikjr r0 j cos
d
e
sin
d
dk
(2 )3 0
0
0
Z
1
(jr r0 j cos ) sin d
(2 )2 0
1
;
4 jr r0 j
1

)3

(1.63)

in the k-space is measured from the direction of the relative position vector
d3 k = k 2 dk sin d d ;
Z 1
eikx dk = 2 (x);

(1.64)
(1.65)

(ax) =

1
(x);
jaj

(1.66)

are noted. (The same technique will be used in nding a Greens function for the less trivial wave
equation,
1 @2
r2
G(r; r0 ; t; t0 ) =
(r r0 ) (t t0 );
(1.67)
c2 @t2
in Chapter 6.)
If the charge density distribution is spatially conned within a small volume such that r
the inverse distance function 1=jr r0 j may be Taylor expanded as follows:
1
jr

r0 j

=
=

1
1
1
1
r
r0 + rr
: r0 r0
r
r
2!
r
1 X
1 r r0
+ 3 + 5
(3ri rj r2 ij )ri0 rj0 +
r
r
2r

r0 ;

(1.68)

i;j

and so is the potential

(r) =

1
4

(r0 )
1
dV 0 =
0
jr r j
4 0

1
r

r
(r )dV + 3
r
0

r0 (r0 )dV 0 +

3rr r2 1
:Q+
2r5

(1.69)
where 1 is the unit dyadic. The quantity q =
is the total charge contained in the (small)
volume V and the corresponding lowest order monopole potential is
R

(r0 )dV 0

monopole

16

1 q
:
4 0r

(1.70)

The vector quantity

p=

r (r)dV;

(1.71)

is the dipole moment and the corresponding dipole potential is

dipole

The tensor
Q = Qij =

1 p r
:
4 0 r3

(1.72)

ri rj (r)dV;

(1.73)

denes the quadrupole moment and the corresponding quadrupole potential is

quadrupole

1 3rr r2 1
1 1 X
:Q=
(3ri rj
5
4 0
2r
4 0 2r5

r2

ij )Qij :

(1.74)

i;j

Example 1 Electric Field Lines of a Dipole

The potential due to a dipole moment directed in z direction p = pez at the origin is given by
(r; ) =

1
4

p
cos :
2
0r

The equipotential surfaces are described by

cos
= const.
r2
The electric eld can be found from
E=

1
4

p sin
2p cos
er +
e
3
r
r3

(1.75)

The equation to describe an electric eld line is, by denition,

dr
rd
=
;
Er
E

(1.76)

which gives
dr
= 2 cot d ;
r
and thus
r = const. sin2 :

(1.77)

Evidently, the electric eld lines are normal to the equipotential surfaces. In Fig.1-5, one equipotential surface and three electric eld lines are shown.
Example 2 Linear quadrupole

17

0.5

-1

00

-0.5

0.5

-0.5

-1

Figure 1-5: Equipotential surface and electric eld lines of an electric dipole pz : The z-axis is
vertical.

A linear quadrupole consists of charges q at z = a and 2q at z = 0 as shown in Fig.1-6. The

charge density is
(r) = q[ (z a) 2 (z) + (z + a)] (x) (y):
Therefore, only Qzz is nonvanishing,
Qzz =
and the quadrupole potential at r
(r) =

1
4

1
(3z 2
5
0 2r

z 2 (r)dV = 2a2 q;

a is given by
r2 )Qzz =

1 2qa2 3 cos2
4 0 r3
2

where
P2 (cos ) =

3 cos2
2

1 2qa2
P2 (cos );
4 0 r3

is the Legendre function of order l = 2.

Alternatively, the potential can be found from the direct superposition of the potentials produced by each charge,
(r; ) =

q
4

1
r2

p
For r > a; the functions 1= r2 + a2

a2

2ar cos

2
1
+p
2
2
r
r + a + 2ar cos

2ar cos can be expanded in terms of the Legendre functions

18

Figure 1-6: Linear quadrupole.

Pl (cos ) as
p

1
r2

a2

2ar cos

1X a
r
r

( 1)l Pl (cos ); r > a:

l=0

Retaining terms up to l = 2 (quadrupole), we recover

(r; ) '

1 2qa2
P2 (cos );
4 0 r3

a:

At r
a; the leading order potential is of quadrupole. The total charge is zero, thus no monopole
potential at r
a. Also, the charge system consists of two dipoles of equal magnitude oriented in
opposite directions. Therefore, the dipole potential vanishes at r
a as well. Equipotential prole
is shown in Fig. ??.

1.6

Potential due to a Ring Charge: Several Methods

A given potential problem can be solved in dierent coordinates systems. Of course, the solution
is unique, and answers found by dierent methods should all agree. For the purpose of becoming
familiar with several coordinates systems and some useful mathematical techniques, here we nd
the potential due to a ring charge having radius a and total charge q uniformly distributed using
several independent methods.

Method 1: Direct integration

The potential is symmetric about the z-axis because of the uniform charge distribution. Therefore, we may evaluate the potential at arbitrary azimuthal angle and here we choose = =2,
that is, observing point in the y z plane. The dierential potential due to a point charge

19

1
0.5

-1

-0.5

0.5

-0.5
-1

Figure 1-7: Equipotential surface of a linear quadrupole. The z axis ( = 0) is in the vertical
direction. The thick line is at +1 unit potential, and the thin line is at 1:

dq = qd 0 =2 located at angle
d
where
to

is

1
4

dq
0

jr

r0 j

r2

Integrating over

= cos cos

a2

2ar cos

= =2) and r0 = (a;

is the angle between the vectors r = (r; ;

cos

p
0

+ sin sin

cos(

d 0;
0

) = sin sin

(1.78)

= =2;
0

): cos

reduces
(1.79)

from 0 to 2 ; we nd
1

(r; ) =
Changing the variable from

to

q
02

r2 + a2

through

2 =

(1.80)

4
K(k 2 );
r2 + a2 + 2ar sin

(1.81)

2ar sin sin

the integral can be reduced to

4
p
r2 + a2 + 2ar sin

=2

k 2 sin2

20

d =p

Figure 1-8: Uniform ring charge.

where
k2 =

4ar sin
;
r2 + a2 + 2ar sin

21

(1.82)

is the argument of the complete elliptic integral of the rst kind dened by
Z

K(k )

=2

The nal form of the potential is

(r; ) =

q
2

f (x) =

r2

a2

=2

p
1

k 2 sin2

d :

(1.83)

1
K(k 2 ):
+ 2ar sin
1
x sin2

(1.84)

4.5
4
3.5
3
2.5
2
0

0.2

0.4

0.6

0.8

Figure 1-9: K(x)

p the complete elliptic integral of the rst kind. It diverges logarithmically at x . 1;
K(x) ' ln 4= 1 x :
The function K(k 2 ) is shown in 1-9. It diverges as x = k 2 approaches unity, that is, near the
ring itself, as expected. However, the divergence is only logarithmic,
lim

k2 !1

= ln

4
p

Example 3 Capacitance of a Thin Conductor Ring

Let us assume a thin conducting ring with ring radius a and wire radius ( a) shown in
Fig.1-10. The potential on the ring surface can be found by letting r = a
; = =2: In this limit,

22

the argument k 2 approaches unity,

k2 =

(a

)2

4a(a
)
2
+ a + 2a(a

'1

2a

and the ring potential becomes

ring

'

q
4

0a

8a

ln

(1.85)

Then the self-capacitance of the ring can be found from

C=

ring

ln

0a

8a

(1.86)

This formula is fairly accurate even for a not-so-thin ring because of the mere logarithmic dependence on the aspect ratio, a= : For example, when a= = 5 (which is a fat torus rather than a thin
ring); Eq. (1.86) still gives a capacitance within 2 % of the correct value. An exact formula for the
capacitance of a conducting torus will be worked out later in terms of the toroidal coordinates.

Figure 1-10: A thin conductor ring with major radius a and minor radius ; a

Method 2: Multipole Expansion in the Spherical Coordinates

As the second method, we directly solve the Poissons equation in the spherical coordinates,
r2

(r)
0

q
(r
2
a2
0

a) (cos ):

(1.87)

Except at the ring itself, the charge density vanishes. Therefore, in most part of space, the potential
satises Laplace equation,
r2 = 0; r 6= a; 6= ;
2
and we seek a potential in terms of elementary solutions to the Laplace equation. Since @=@ = 0;

23

the Laplace equation reduces to

2@
@2
1 1 @
+
+ 2
@r2
r @r
r sin @
Assuming that the potential is separable in the form
r2
R

d2 R 2 dR
+
dr2
r dr

sin

@
@

= 0:

(1.88)

(r; ) = R(r)F ( ), we nd

1 1 d
F sin d

sin

dF
d

= 0:

(1.89)

Since the rst term is a function of r only and the second term is a function of only, each term
must be constant cancelling each other. Introducing a separation constant l(l + 1); we obtain two
ordinary dierential equations for R and F;
d2 R 2 dR
+
dr2
r dr
1 d
sin d
which can be written as

d
d

(1.90)

dF
d

+ l(l + 1)F = 0;

(1.91)

dF
d

+ l(l + 1)F = 0;

(1.92)

sin

(1

l(l + 1)
R = 0;
r2

where = cos : Eq. (1.92) is known as the Legendres dierential equation. The solutions for
R(r) are
1
R(r) = rl and l+1 ;
(1.93)
r
and the solutions for F ( ) are the familiar Legendre functions,
F ( ) = Pl (cos ) and Ql (cos );

(1.94)

where Ql (cos ) is the Legendre function of the second kind. Some low order Legendre functions
for real x (jxj 1) are listed below.
P0 (x) = 1; P1 (x) = x; P2 (x) =
1
1+x
Ql (x) = Pl (x) ln
2
1 x

Wl

1 (x);

3x2 1
5x3 3x
; P3 (x) =
;
2
2

1 (x)

3
= 0; W0 (x) = 1; W1 (x) = x;
2

(1.95)
(1.96)

For a general complex variable z; the denition for Ql (z) is modied as follows:
1
z+1
Ql (z) = Pl (z) ln
2
z 1

Wl

1 (z);

1 (z)

3
= 0; W0 (z) = 1; W1 (z) = z;
2

(1.97)

In the present problem, the potential should remain nite everywhere except at the ring. Since
Ql (cos ) diverges at = 0 and ; it should be discarded and we assume the following series solutions

24

for the potential separately for interior (r < a) and exterior (r > a);

(r; ) =

8
P
r
>
>
Al
>
l
<
a

Pl (cos );

r<a
(1.98)

>
>
>
: P Bl a
l
r

l+1

Pl (cos ); r > a

where Al and Bl are constants to be determined. The potential should be continuous at r = a

because there are no double layers to create a potential jump. Therefore, Al = Bl : To determine
Al ; we substitute the assumed potential into the Poissons equation,
@2
2@
1 1 @
+
+ 2
@r2
r @r
r sin @

sin

@
@

q
(r
2
a2
0

a) (cos );

to obtain
X
l

2 d
d2
+
2
dr
r dr

l(l + 1)
r2

Al Rl (r)Pl (cos ) =

where
Rl (r) =

8
>
>
>
<
>
>
>
:

r
a

a
r

l+1

q
(r
2
02 a

a) (cos );

(1.99)

r<a
(1.100)
; r>a

If we multiply Eq. (1.99) by Pl0 (cos ) sin and integrate over from = 0 to ; the summation
over l disappears because of the orthogonality of the Legendre functions,
Z

Pl (cos )Pl0 (cos ) sin d =

1
1

Pl ( )Pl0 ( )d =

2
2l + 1

ll0 :

(1.101)

We thus obtain
Al

d2
2 d
+
2
dr
r dr

l(l + 1)
r2

Rl (r) =

q
(r
2
02 a

a)

2l + 1
Pl (0):
2

(1.102)

The LHS should contain a delta function (r a) to be compatible with that in the RHS. This can
be seen as follows. The radial function Rl (r) has a discontinuity in its derivative at r = a;
dRl
dr

=
r=a+0

l + 1 dRl
;
a
dr

r=a 0

l
=+ :
a

(1.103)

Therefore, the second order derivative d2 Rl =dr2 yields a delta function,

d 2 Rl
=
dr2

2l + 1
(r
a

25

a):

(1.104)

We thus nally nd the expansion coe cient Al ;

Al =

q
4

0a

Pl (0) =

and the potential,

(r; ) =

q
4

q
4

( 1)l=2 l!
; (l = 0; 2; 4; 6;
l
2
0 a 2 [(l=2)!]

X ( 1)l=2 l!
Rl (r)Pl (cos ):
2l [(l=2)!]2
0a

);

(1.105)

(1.106)

l even

Figure 1-11: The derivative of the radial function R(r) is discontinuous at r = r0 and thus its
second order derivative yields a delta function.
The disappearance of the odd harmonics is understandable because of the up-down symmetry
of the problem. At r
a; the leading order term is the monopole potential,
l=0 (r)

q
4

a
;
0a r

followed by the quadrupole potential,

l=2 (r;

)=

q
4

1 a
0a 2 r

P2 (cos );

and so on. The dipole potential (l = 1) vanishes because of the symmetric charge distribution,
( r) = (r);
Z
p = r (r)dV = 0:
For problems with axial symmetry (@=@ = 0) as in this example, knowing the potential along
the axis (z) is su cient to nd the potential at arbitrary point (r; ) in the spherical coordinates.
This is because for all of the Legendre functions, Pl (1) = 1; Pl ( 1) = ( 1)l : In the case of the ring

26

charge, the axial potential can be readily found,

(z) =

q
4

1
:
+ a2

(1.107)

z2

For jzj > a; this can be expanded as

(z) =

q
4

1
0 jzj

1 a
2 z

3 a
8 z

(1.108)

Therefore, at arbitrary point (r > a; ); the potential is

(r; ) =

q 1
4 0r

1 a
2 r

P2 (cos ) +

3 a
8 r

3 r
8 a

P4 (cos )

; r > a:

(1.109)

P4 (cos )

; r < a:

(1.110)

For interior region (r < a); the potential becomes

(r; ) =

q
4

1
0a

1 r
2 a

P2 (cos ) +

They agree with the potential given in Eq. (1.106).

27

Method 3: Cylindrical Coordinates

In the cylindrical coordinates ( ; ; z); the Poissons equation for the potential due to a ring
charge becomes
q (
a)
@2
1 @
@2
( ; z) =
+
+
(z):
(1.111)
2
2
@
@
@z
2 0
a
Since the z-coordinate extends from

1 to 1; we seek a solution in the form of Fourier transform,

1
( ; z) =
2
where
Since

( ; k)eikz dk;

(1.112)

( ; k) is the one-dimensional Fourier transform of the potential having dimensions of V m.

@
( ; z);
@z

is Fourier transformed as
ik ( ; k);
and (z) as unity, the equation for

( ; k) becomes an ordinary dierential equation,

d2
1 d
+
d 2
d

k2

( ; k) =

Elementary solutions of the dierential equation at

d2
1 d
+
2
d
d

k2

q
2

a)
a

(1.113)

6= a;
( ; k) = 0;

(1.114)

are the zero-th order modied Bessel functions,

( ; k) = I0 (k ); K0 (k );
shown in Fig.1-12.
The Fourier potential ( ; k) which is continuous at
may be constructed in the following form,

( ; k) =

8
>
< AI0 (k )K0 (ka);
>
:

AI0 (ka)K0 (k );

(1.115)

= a and remains bounded everywhere

<a
(1.116)
>a

The coe cient A can be determined readily from the discontinuity in the derivative of
= a;
d
d
= AkI0 (ka)K00 (ka);
= AkI00 (ka)K0 (ka);
d =a+0
d =a 0

28

( ; k) at
(1.117)

5
4
3
2
1
0

0.5

1
x

1.5

Figure 1-12: Modied Bessel functions I0 (x) (starting at 1) and K0 (x) (diverging at x = 0):

from which it follows that

d2
d 2

= Ak I0 (ka)K00 (ka)

I00 (ka)K0 (ka)

a) =

=a

A
(
a

a);

(1.118)

where the Wronskian of the modied Bessel functions

I0 (ka)K00 (ka)

1
;
ak

I00 (ka)K0 (ka) =

(1.119)

has been exploited. From

d2
d 2

A
(
a

=
=a

a) =

we nd
A=

q
2

q
2

a)
a

(1.120)

(1.121)

and the nal form of the potential is

( ; z) =

q
4

8
9
>
>
I
(k
)K
(ka)
0
0
<
=

1>
:

I0 (ka)K0 (k )

>
;

<a
ikz

dk;

(1.122)
>a

Noting that the modied Bessel functions are even with respect to the argument, this may be
further rewritten as
8
9
I0 (k )K0 (ka) >
<a
Z 1>
<
=
q
( ; z) = 2
cos(kz)dk;
(1.123)
>
2 0 0 >
:
;
I0 (ka)K0 (k )
>a
29

The convergence of the integral is rather poor since the function I0 K0 decreases with k only
algebraically. An alternative solution for the potential can be found in terms of Laplace transform
rather than Fourier transform,
( ; z) =

kjzj

( ; k)e

dk:

(1.124)

The Laplaces equation thus reduces to

1 d
d2
+
+ k2
2
d
d

( ; k) = 0;

(1.125)

whose solution is the ordinary Bessel function J0 (k ): We thus assume for

( ; k)

( ; k) = A(k)J0 (k );

(1.126)

where A(k) may still be a function of k: The Poissons equation becomes

Z

@2
1 @
@2
+
+
@ 2
@
@z 2

A(k)J0 (k )e

kjzj

dk =

(
0

a)
a

(z):

(1.127)

Noting
d2
e
dz 2
we thus nd

kjzj

= k2 e

kjzj

kA(k)J0 (k )dk =

2k (z)
q

(1.128)
a)

(1.129)

However, the Bessel function forms an orthogonal set according to

Z

kJ0 (ka)J0 (k )dk =

a)
a

(1.130)

which uniquely determines the function A(k);

A(k) =

q
4

J0 (ka):

(1.131)

Therefore, the nal form of the potential is

( ; z) =

q
4

J0 (ka)J0 (k )e

kjzj

dk:

(1.132)

The convergence of this solution is much faster than the solution in Eq. (1.123) and thus more
suitable for numerical evaluation.

30

Method 4: Toroidal Coordinates

The toroidal coordinates ( ; ; ) are related to the cartesian coordinates (x; y; z) through the
following transformation,
8
R sinh cos
>
> x=
;
>
>
cosh
cos
>
>
>
>
>
>
<
R sinh sin
;
y=
(1.133)
>
cosh
cos
>
>
>
>
>
>
>
>
R sin
>
: z=
:
cosh
cos

This coordinate system is one example in which the Laplace equation is not separable, that is, the
potential cannot be written as a product of independent functions, ( ; ; ) 6= F1 ( )F2 ( )F3 ( ):
However, the potential is partially separable and can be sought in the form
( ; ; )=

cosh

cos F1 ( )F2 ( )F3 ( ):

Figure 1-13: Cross-section of the toroidal coordinates ( ; ; ):

radius R:

(1.134)

! 0 corresponds to a thin ring of

In the toroidal coordinates, the = constant surfaces are toroids having a major radius R coth
and minor radius R= sinh : ! 1 degenerates to a thin ring of radius R: In the other limit, ! 0
describes a thin rod on the z-axis. The = constant surfaces are spherical bowls as illustrated in
Fig.1-13.
Assuming the partial separation for the potential in Eq. (1.134) leads to the following ordinary
31

equations for the functions F1 ( ); F2 ( ) and F3 ( );

1 d
sinh d

sinh

dF1
d

1
4

m2
sinh2

l2

F1 = 0;

(1.135)

d2
+ m2 F3 = 0:
d 2

d2
+ l2 F2 = 0;
d 2

Comparing Eq. (1.135) with the standard form of the dierential equation for the associated
Legendre function Plm (cos ); Qm
l (cos );
1 d
sin d

d
d

sin

m2
sin2

+ l(l + 1)

(Plm ; Qm
l ) = 0;

(1.136)

we see that solutions for F1 ( ) are

(cosh ):
F1 ( ) = Plm 1 (cosh ); Qm
l 1
2

(1.137)

The functions F2 and F3 are elementary,

F2 ( ) = eil ; F3 ( ) = eim ;

(1.138)

and the general solution for the potential may be written as

( ; ; )=

p
cosh

cos

Alm Pl

1
2

l;m

(cosh ) + Blm Ql

1
2

(cosh ) eil

+im

(1.139)

Of course, the potential is a real function. The solution above is an abbreviated form of a more
cumbersome expression,
( ; ; ) =

cosh

cos

1
X

Al Pl

l=0

1
X

1
2

(cosh ) + Bl Ql

1
2

(cosh ) (Cl cos l + Dl sin l )

(El cos m + Fl sin m ) :

m=0

We now consider a conducting toroid with a major radius a and minor radius b: Its surface is
described by 0 = const. where
a = R coth

0;

b=

R
sinh

(1.140)

Because of axial symmetry, only the m = 0 term is present. If the toroid is at a potential V; the
potential o the toroid can be written as
( ; )=

cosh

cos

1
X
l=0

32

Al P l

1
2

(cosh ) cos l ;

(1.141)

where the Legendre function of the second kind Ql 1 (cosh ) has been discarded because it diverges
2
at ! 0 which corresponds to the z axis. (The potential along the z axis should be bounded.)
The expansion coe cients Al can be determined from the boundary condition, = V at = 0 ;
V =
or

1
X

cosh

1
X

cos

Al P l

Al P l

1
2

l=0

(cosh

=p

0 ) cos l

Multiplying both sides by cos l0 and integrating over

Ql

1
2

Pl

1
2

= 1;

Al = V

(cosh

1
2

l=0

0 ) cos l

V
cosh

cos

(1.142)

from 0 to ; we obtain

(cosh

0)

(cosh

0)

l;

(1.143)

(x):

(1.144)

where
0

= 2 (l

1);

and the following integral representation has been exploited,

Z

p
cos l
d = 2Ql
x cos

1
2

Then, the nal form of the potential is

( ; )=

2V p

cosh

cos

1 Q
X
l
l=0

(cosh

0)

(cosh

0)

1
2

Pl

1
2

Pl

1
2

(cosh ) cos(l ) l :

(1.145)

The capacitance of a conducting torus can be found from the behavior of the potential at a
large distance from the torus which should be in the form of monopole potential,
(r ! 1) =
At a large distance from the torus r

R; both

r2 = x2 + y 2 + z 2 =
p

cosh

and

0r

(1.146)

approach zero,

R2 (sinh2 + sin2 )
!
(cosh
cos )2

cos '
Pl

1
2

2+
p
2

(cosh ) ' 1:

33

'

p R
2 ;
r

4R2
;
2+ 2

(1.147)

(1.148)
(1.149)

Therefore, the asymptotic potential is

(r

R) =

1
2

l=0

1
2

2V R X Ql
r
Pl

(cosh

0)

(cosh

0)

q
4

0r

(1.150)

from which the capacitance of the torus can be found,

C

X Ql
q
= 8 0R
V
Pl

1
2
1
2

l=0

(cosh
(cosh

0)
l

0)

= 8 0 a tanh

1 Q
X
l
l=0

where a is the major radius of the torus. The function Ql

(1.144) and Pl 1 (cosh 0 ) from

1
2

Pl

(cosh

1
2
1
2

(cosh

0)

(cosh

0)

0)

l;

(1.151)

can be evaluated from Eq.

Pl

1
2

(cosh

0)

d
(cosh

+ sinh

0 cos

)l+ 2

(1.152)

Example 4 Capacitance of a Fat Torus

We wish to nd the capacitance of a conducting torus having a major radius of a = 50 cm and
minor radius of b = 10 cm. The torus is dened by cosh 0 = 5:0 and R = a tanh 0 = 49 cm. For
cosh 0 = 5:0; the Legendre functions numerically evaluated are:
l

0
1
2

Ql

1
2

(5)

Pl

1:00108
0:05063
0:00384

1
2

(5)

0:74575
2:03557
13:32184

Ql

1
2

Pl

1
2

(5)
(5)

1:34234
0:02487
0:00029

The ratio Ql 1 (5)=Pl 1 (5) rapidly converges as l increases and it su ces to truncate the series at
2
2
l = 2: The capacitance is
C = 8"0 R(1:3423 + 2

0:0249 + 2

0:0003 +

) ' 8 0R

The approximate formula worked out earlier gives

C'

4
ln

0a

8a
b

= 8 0a

1:34;

which is in reasonable agreement with the numerical result.

34

1:393 = 8 0 a

1:36:

1.7

Spherical Multipole Expansion of the Scalar Potential

The potential due to a prescribed charge distribution,

Z

1
(r) =
4 "0

(r0 )
dV 0 ;
jr r0 j

can be expanded in terms of the general spherical harmonics as follows. For this purpose, it is
su cient to expand the Greens function,
G(r; r0 ) =

1
1
;
4 jr r0 j

(1.153)

which satises the singular Poissons equation,

r2 G =

r0 ):

(r

(1.154)

The Greens function G satises Laplace equation except at r = r0 ;

r2 G = 0; r 6= r0 ;

(1.155)

and we rst seek elementary solutions for Laplace equation. Assuming that the function G(r; ; )
is separable in the form
G(r; ; ) = R(r)F1 ( )F2 ( );
(1.156)
and substituting this into Laplace equation
1 @
1 1 @
@2
+
+
@r2 r @r r2 sin @

sin

@2
1
2
r2 sin @ 2

l(l + 1)
r2

R(r) = 0;

(1.158)

m2
sin2

(1.159)

@
@

G(r; ; ) = 0;

(1.157)

we obtain three ordinary equations,

d2
1 d
+
2
dr
r dr
1 d
sin d

sin

d
d

+ l(l + 1)

d2
+ m2 F2 ( ) = 0;
d 2

F1 ( ) = 0;

(1.160)

where l(l + 1) and m2 are separation constants. Solutions of the radial function R(r) are as before,
R(r) = rl ;

1
rl+1

In free space, the Greens function must be periodic with respect to the azimuthal angle : Therefore,
F2 ( ) = eim with m being an integer.

35

Equation (1.159) is known as the modied Legendre equation and its solutions are
F1 ( ) = Plm (cos ); Qm
l (cos );
where

(1.161)

dm
Pl (x);
dxm
dm
x2 )m=2 m Ql (x):
dx

Plm (x) = (1

x2 )m=2

Qm
l (x) = (1

(1.162)
(1.163)

That Plm (cos ) given in Eq. (1.162) satises the modied Legendre equation can be seen as follows.
The ordinary Legendre function Pl (cos ) = Pl ( ) satises
d
d

(1

dPl ( )
d

+ l(l + 1)Pl ( ) = 0:

Dierentiating m times yields

(1

dm+2 Pl
d m+2

2 (m + 1)

dm+1 Pl
+ [l(l + 1)
d m+1

m(m + 1)]

d m Pl
= 0:
d m

Let us assume
2 m=2

F1 ( ) = (1

f ( );

and substitute this into Eq. (1.159). f ( ) satises the same equation as
(1

d2 f
d 2

2 (m + 1)

df
+ [l(l + 1)
d

dm Pl
d m ;

m(m + 1)] f = 0:

Pl
Therefore, f ( ) = dd m
:
Since the Legendre function Pl (x) is a polynomial of order l; the azimuthal mode number m
is limited in the range 0
m
l: For later use, we combine the functions F1 ( ) and F2 ( ) and
introduce the spherical harmonic function dened by

Ylm ( ; ) = const. Plm (cos )eim ;

where the constant is chosen from the following normalization,
Z
Ylm ( ; )Ylm ( ; )d = 1:
Noting

[Plm (cos )]2 sin d =

we nd
const. =

2 (l + m)!
;
2l + 1 (l m)!

2l + 1 (l m)!
;
4
(l + m)!
36

(1.164)

(1.165)

(1.166)

Ylm =

2l + 1 (l m)! m
P (cos )eim :
4
(l + m)! l

(1.167)

For historical reasons, it is customary to write Ylm ( ; ) in the form

Ylm ( ; ) = ( 1)m
Yl;

m(

2l + 1 (l m)! m
P (cos )eim ; for 0
4
(l + m)! l

; ) = ( 1)m Ylm ( ; ) ; for

l;

0;

or for arbitrary m; positive or negative,

Ylm ( ; ) = ( 1)(m+jmj)=2

2l + 1 (l jmj)! jmj
P (cos )eim :
4
(l + jmj)! l

(1.168)

Some low order forms of Ylm ( ; ) are:

l=0 m=0

l=1 m=0

m=

l=2 m=0

m=

m=

1
Y00 = p
4
Y10 =

1 Y1;

Y20 =

1 Y2;

2 Y2;

3
cos
4
r

=
r

3
sin e
8

5 3 cos2
4
2

1
=
4

15
sin cos e
8
15
sin2 e
2

2i

Having found the general solutions of the Greens function, we now assume the following expansion for G;
X
G(r; r0 ) =
Alm gl (r; r0 )Ylm ( ; );
lm

where the radial function gl

(r; r0 )

is

gl (r; r0 ) =

8
rl
>
>
>
; r < r0 ;
>
< r0l+1
>
>
0l
>
>
: r ;
rl+1
37

r > r0 :

In this form, the function gl (r; r0 ) remains bounded everywhere. The expansion coe cient Alm can
be determined by multiplying the both sides of
r2 G =

(r r0 )
(cos
rr0

r0 ) =

(r

cos 0 ) (

);

(1.169)

by Yl0 m0 ( ; ) and integrating the result over the entire solid angle,
Alm

(r r0 )
Ylm ( 0 ;
rr0

d2
gl (r; r0 ) =
dr2

);

where orthogonality of Ylm ( ; );

Z

Ylm ( ; )Yl0 m0 ( ; )d

ll0 mm0 ;

(1.170)

is exploited to remove summation over l and m: Since the radial function gl (r; r0 ) has a discontinuity
in its derivative at r = r0 ; it follows that
d2
gl (r; r0 ) =
dr2

(2l + 1)

r0 )

(r
r2

(1.171)

Therefore,
Alm =

1
Y ( 0;
2l + 1 lm

);

(1.172)

and the desired spherical harmonic expansion of the Greens function is given by
G(r; r0 ) =

X 1
1
=
gl (r; r0 )Ylm ( ; )Ylm ( 0 ;
0
4 jr r j
2l + 1

):

(1.173)

lm

For a given charge distribution (r); this allows evaluation of the potential in the form of spherical
harmonics,
(r) =
=
=

Z
1
(r0 )
dV 0
4 "0
jr r0 j
Z
1 X 4
1
Ylm ( ; ) r0l Ylm ( 0 ; 0 ) (r0 ; 0 ; 0 )dV 0
4 "0
2l + 1 rl+1
lm
1 X 4
1
Ylm ( ; )qlm ; in the region r > r0 ;
l+1
4 "0
2l + 1 r

(1.174)

lm

where
qlm

r0l Ylm ( 0 ;

) (r0 ; 0 ;

)dV 0 ;

(1.175)

is the electric multipole moment of order (l; m):

Example 5 Planar Quadrupole
A planar quadrupole consists of charges +q and
38

q alternatively placed at each corner of a

square of side a as shown in Fig.1-14. The charge density may be written as

(r) = q

(r)

p
(r
2a)
(r a)
(cos ) ( ) +
(cos )
2
2
a
2a

The potential in the far eld region r

(r; ; ) =
where
q2;

r Y2;

(r a)
(cos )
a2

a is of quadrupole nature and given by

1 1
(Y2;2 q2;2 + Y2;
"0 5r3
1
2 ( ; ) (r)dV =
2

2 q2; 2 ) ;

15 2
a qe
2

i =2

The potential reduces to

(r; ; ) '

3 a2
sin2 sin(2 ); r
8 "0 r 3

a:

The reader should check that this is consistent with the direct sum of four potentials in the limit
r
a:

Figure 1-14: Planar quadrupole in the x

1.8

y plane.

Collection of Dipoles, Dielectric Properties

Dielectric properties of material media originate from dipole moments carried by atoms and molecules. Some molecules carry permanent dipole moments. For example, the water molecule has a
dipole moment of about 6 10 30 C m due to deviation of the center of electron cloud from the
center of proton charges. When water is placed in an external electric eld, the dipoles tend to be
39

aligned in the direction of the electric eld and a resultant electric eld in water becomes smaller
than the unperturbed external eld by a factor "="0 ; where " is the permittivity of water. Even a
material composed of molecules having no permanent dipole moment exhibits dielectric property.
For example, a dipole moment is induced in a hydrogen atom placed in an electric eld through
perturbation in the electron cloud distribution which is spatially symmetric in the absence of external electric eld. In this section, the potential and electric eld due to a collection of dipole
moments will be analyzed.
The potential due to a single dipole p located at r0 is
1 (r r0 ) p
:
4 0 jr r0 j3

(r) =

(1.176)

Consider a continuous distribution of many dipoles. It is convenient to introduce a dipole moment

density P =np (C m/m3 = C/m2 ) where n is the number density of dipoles. An incremental
potential due to an incremental point dipole dp = P(r0 )dV 0 is
d

Integration of this yields

(r) =
Since

r0 ) P(r0 ) 0
dV :
jr r0 j3

1 (r

1
4

r
jr

(1.177)

r0 ) P(r0 ) 0
dV :
jr r0 j3

(r

r0
= r0
r0 j3

1
r0 j

jr

(1.178)

(1.179)

where r0 means dierentiation with respect to r0 ; the integral in Eq. (1.178) can be rewritten as
Z

(r

r0 ) P(r0 ) 0
dV =
jr r0 j3

P(r0 )
jr r0 j

r0

dV 0

r0 P (r0 ) 0
dV :
jr r0 j

The rst term in the right can be converted into a surface integral through the Gauss theorem,
Z

P(r0 )
jr r0 j

dV =

P(r0 )
dS0 ;
jr r0 j

which vanishes on a closed surface with innite extent on which all sources should be absent.
Therefore, the potential due to distributed dipole moments is
dipole (r)

1
4

r0 P(r0 ) 0
dV :
jr r0 j

Comparing with the standard form of the potential,

1
(r) =
4 "0

40

(r0 )
dV 0 ;
jr r0 j

(1.180)

we see that
e

r P;

can be regarded as an eective charge density. To distinguish it from the free charge density,
e dened above is called bound charge density. In general, e cannot be controlled by external
means. In a dielectric body, the bound charge appears as a surface charge density. Adding the
monopole potential due to a free charge density free (r); we nd the total potential
(r) =

1
4

0
free (r )
dV 0
jr r0 j

1
4

r0 P(r0 ) 0
dV :
jr r0 j

(1.181)

Noting
r2

1
jr

r0 j

r E=

r2

(r

we nd

In this form too, it is evident that

(1.181) can be rearranged as

r0 );

free

r P

(1.182)

r P can be regarded as an eective charge density. Eq.

free :

r ( 0 E + P) =

(1.183)

Introducing a new vector D (displacement vector) by

D=

0E

+ P;

(1.184)

r D=

free ;

(1.185)

we write Eq. (1.183) as

which is equivalent to the original Maxwells equation

r E=

all

(1.186)

provided that the total charge density all consists of free charges and dipole charges and that
higher order charge distributions (quadrupole, octupole, etc.) are ignorable. The vector D was
named the displacement vector by Maxwell. Its fundamental importance in electrodynamics will be
appreciated later in time varying elds because it was with this displacement vector that Maxwell
was able to predict propagation of electromagnetic waves in vacuum. As briey discussed in Section 2, Maxwells equations would not be consistent with the charge conservation principle if the
displacement current,
@D
@("E)
=
;
@t
@t
were absent.
In usual linear insulators, both D and P are proportional to the local electric eld E which

41

allows us to introduce an eective permittivity ";

D = "0 E + P ="E:

(1.187)

In some solid and liquid crystals, the permittivity takes a tensor form,
D=

E or Di =

ij Ej ;

(1.188)

because of anisotropic polarizability. Double refraction phenomenon in optics already known in

the 17th century is due to the tensorial nature of permittivity in some crystals as we will study in
Chapter 5.

Figure 1-15: In an electric eld, the hydrogen atom exhibits a dipole moment due to the displacement of the center of electron cloud from the proton.

The origin of eld induced dipole moment may be seen qualitatively for the case of hydrogen
atom as follows. In an electric eld, the distribution of the electron cloud around the proton
becomes asymmetric because of the electric force acting on the electron. The center of electron
cloud is displaced from the proton by x given by
m! 20 x = eE;
where ! 0 is the frequency of bound harmonic motion of the electron and m is the electron mass.
(~! 0 is of the order of the ionization potential energy.) Then, the dipole moment induced by the
electric eld is
e2
E:
p = ex =
m! 20
If the atomic number density is n; the dipole moment density P is
P = np =

42

ne2
E;
m! 20

(1.189)

and the displacement vector is given by

D = "0 E + P
= "0 1 +

ne2
"0 m! 20

E = "E:

(1.190)

(1.191)

The permittivity can therefore be dened by

" = "0
where
!p =

! 2p
1+ 2
!0
s

ne2
;
"0 m

(1.192)

is an eective plasma frequency. (The plasma frequency pertains to free electrons in a plasma. ! p
introduced here is so called only for dimensional convenience.)
In the incremental contribution to the permittivity normalized by "0 ,
! 2p
e2
=
n
;
! 20
m"0 ! 20
the quantity
e2
;
4 "0 m! 20
is of the order of r03 where r0 is the Bohr radius. This can be seen from the force balance for the
electron,
e2
mr0 ! 20 =
;
4 "0 r02
which indeed gives
e2
= r03 :
4 "0 m! 20
The quantity
= 4 "0 r03 =

e2
;
m! 20

(1.193)

is called the atomic polarizability. A more rigorous quantum mechanical calculation yields for this
quantity
9
=
4 "0 r03 ;
(1.194)
2
indicating a substantial correction (by a factor 4.5) to the classical picture based on the assumption
of rigid shift of the electron cloud. For molecualr hydrogen H2 ; the polarizability at T = 273 K is
approximately given by
' 5:4 4 "0 r03 :

43

Satisfactory quantum mechanical calculation was performed by Kolos and Wolniewicz relatively
recently (1967). This is in a fair agreement with the permittivity experimentally measured in the
standard conidition 20 C, 1 atmospheric pressure,
" ' 1 + 2:7

10

"0 :

The atomic polarizability and the macroscopic relative permittivity "r are related through a
simple relationship known as the Clausius-Mossoti equation,
"r 1
n
=3
;
"0
"r + 2

(1.195)

where n is the number density of atoms. This may qualitatively be seen as follows. A dipole
moment induced by an atom,
e2
p = ex =
E = Eext ;
(1.196)
m! 20
produces an electric eld

p
;
4 "0 r 3

E0 =

(1.197)

at a distance r: If the number density of atoms is n; the average inter-atom distance R can be found
from
4 3
1
R = :
(1.198)
3
n
Then,
E0 =

4 "0 R 3

Eext ;

(1.199)

and the total electric eld is

E =

Eext
4 "0 R 3
n
Eext :
3"0

(1.200)
(1.201)

Therefore, using this total eld in calculation of the polarization eld P; we nd

n

P =
1

n ="0
n E=
n "0 E;
1
3"0
3"0

(1.202)

which denes the relative permittivity "r as

"r

1=

n ="0
n :
1
3"0

(1.203)

Solving for n ="0 ; we obtain Eq.(1.195). Note that the relative permittivity is an easily measurable
quantity and it reects, somewhat surprisingly, microscopic atomic polarizability through a simple

44

relationship.
In an oscillating electric eld, the displacement x is to be found from the equation of motion,
@2
+ ! 20 x =
@t2

eE0 e

i!t

(1.204)

which yields
x(t) =

eE0 e i!t
:
! 2 ! 20

A resultant permittivity is
! 2p

"(!) = "0

!2

! 20

(1.205)
!

(1.206)

This exhibits a resonance at the frequency ! 0 : The resonance plays an important role in calculation
of energy loss of charged particles moving in a dielectric medium as will be shown in Chapter 8.
Most molecules have permanent electric dipole moments. In the absence of electric elds,
dipoles are randomly oriented due to thermal agitation. When an electric eld is present, each
dipole acquires a potential energy,
U=

p E=

pE cos ;

(1.207)

where is the angle between the dipole moment p and electric eld E: Dipoles are randomly
oriented but should obey the Boltzmann distribution in thermal equilibrium,
n(cos ) = n0 exp

U
kB T

= n0 exp

pE cos
kB T

(1.208)

where n(cos )=n0 is the probability of dipole orientation in direction, n0 is the number density
of the dipoles and kB = 1:38 10 23 J/ K is the Boltzmann constant. The average dipole number
density oriented in the direction of the electric eld can be calculated from
R
n(cos ) cos d
n= R
:
(1.209)
exp pEkBcos
d
T

In practice, the potential energy is much smaller than the thermal energy pE
the exponential function can be approximated by
exp

pE cos
kB T

and we obtain
n'

'1+
n0 p
E:
3kB T

45

pE cos
;
kB T

kB T: Therefore,

(1.210)

(1.211)

Resultant dipole moment density is

P = np =

n 0 p2
E:
3kB T

(1.212)

This denes a permittivity in a medium consisting of molecules having a permanent dipole moment
p;
n 0 p2
:
(1.213)
" = "0 1 +
3"0 kB T
Note that the correction term
"=

n 0 p2
;
3kB T

(1.214)

is inversely proportional to the temperature. This property can be exploited in separating contributions from atomic and molecular polarizability in a given medium. In liquids and solids, the
additional permittivity given above can be comparable with "0 primarily because of the large number density n0 : For example, a water molecule has a dipole moment of p = 6 10 30 C m. At
room temperature, the additional permittivity due to molecular polarizability of water is approximately " ' 11"0 : The measured permittivity of water at room temperature is 81"0 : The atomic
polarizability is thus dominant.

1.9

Boundary Conditions for E and D

In electrostatics, the electric eld E obeys the Maxwells equations,

r E=

"0

and
r

E = 0:

In problems involving dielectrics, it is more convenient to introduce the displacement vector D,

r D=
where

free

free ;

(1.215)

is the free charge density that can be controlled by external means.

Let us consider a boundary of two dielectrics having permittivities "1 and "2 ; respectively.
Integration of r D = free over the volume of a pancake dV = dS dn (dn is the thickness of the
pancake in the direction perpendicular to the surface dS) on the boundary yields
Z

r DdV =

D dS =

free dV:

(1.216)

In the limit of innitesimally thin pancake, this reduces to

(Dn1

Dn2 )dS =

free dndS

46

free dS;

(1.217)

Figure 1-16: Gausslaw applied to a thin pancake at a boundary between two dielectrics.

where free = free dn is the free surface charge density residing on the boundary. Note that the
volume charge density in the presence of surface charge can be written as
=

(n

n0 ):

(1.218)

General boundary condition for the normal component of the displacement vector is thus given by
D1n

D2n =

free :

(1.219)

In the absence of free surface charge, the normal component of the displacement vector should be
continuous,
Dn1 = Dn2 ; no free charge.
(1.220)

H
Figure 1-17: E dl = 0 applied to a rectangle at the boundary of two dielectrics. The tangential
component of the electric eld Et is continuous. This holds in general for time varying elds as
well.

47

The equation r E = 0 demands that the tangential component of electric eld be continuous
across a boundary of dielectrics. This can be seen by integrating r E = 0 over a small rectangular
area on the boundary,
Z
I
r

E dS =

E dl = 0;

(1.221)

E2 ) dl = 0; E1t = E2t :

(1.222)

which yields
(E1

In fact, the continuity of the tangential component of the electric eld holds for general time varying
case,
@B
r E=
;
(1.223)
@t
for the area integral

B dS;

vanishes in the limit of innitesimally thin rectangle, S ! 0: In contrast to charge and current
densities, a singular magnetic eld involving a delta function is not physically realizable because
then the magnetic energy simply diverges. Note that the square of a delta function is not integrable,
Z
2
(x)dx = (0) = 1:
In a conductor, there can be no static electric elds and E = 0 must hold inside conductors:
Therefore, at a conductor surface, the tangential component of the electric eld should vanish and
the electric eld lines fall normal to the surface. The normal component of the electric eld and
the surface charge density are related through
En =

"0

; at conductor surface.

(1.224)

The potential of a conducting body is constant. The surface electric eld depends on the amount
of charge carried by the body and also curvature of the surface. A trivial case is the electric eld
at the surface of charged conducting sphere of radius a;
E=

1 q
:
4 "0 a2

In general, the eld increases as the curvature radius decreases. The electric eld at the tip of
needle can be strong enough to allow emission of electrons as in tunneling electron microscopes. In
power engineering, conductor surfaces should be made smooth and round as much as possible to
avoid breakdown.
Example 6 Dielectric Sphere in an Electric Field
Let us consider an uncharged dielectric sphere having a uniform permittivity " placed in an
uniform external electric eld E0 = E0 ez : The sphere perturbs the external eld because a bound
48

surface charge = r P will be induced on the sphere. The interior and exterior potentials may
be expanded in the spherical harmonics,
(r; ) =

r
a

Al

(r; ) =

Bl

a
r

Pl (cos );

l+1

Pl (cos );

r < a (interior);

r > a (exterior);

(1.225)

(1.226)

where
0 (r;

)=

E0 z =

E0 r cos ;

(1.227)

is the potential associated with the external electric eld E0 = E0 ez . Since there is no double layer
to cause potential jump at the sphere surface, the potential is continuous at the surface r = a;
from which it follows Al = Bl : This also follows from the continuity of the tangential component
( component) of the electric eld,
1@
E =
:
(1.228)
r@
The continuity of the normal (radial) component of the displacement vector requires
"

E0 cos +

X
l

!
l
Al Pl (cos ) =
a

"0

E0 cos +

X
l

!
l+1
Al
Pl (cos ) :
a

(1.229)

Since cos = P1 (cos ); only the l = 1 terms are non-vanishing and we readily nd
A1 =

" "0
aE0 :
" + 2"0

Then the solutions for the potentials are

8
" "0
>
>
E0 r cos ;
r<a
0 (r; ) +
>
>
" + 2"0
<
(r; ) =
>
>
" "0
a3
>
>
: 0 (r; ) +
E0 2 cos ; r > a
" + 2"0 r

(1.230)

(1.231)

The electric eld in the dielectric sphere is uniform and given by

Eiz =

3"0
E0 ;
" + 2"0

r < a:

(1.232)

This is smaller than the external eld E0 since " > "0 : The interior displacement vector is
Diz = "Eiz =

3"
" 0 E0 ;
" + 2"0

49

(1.233)

which is larger than D0 = "0 E0 : The perturbed exterior potential,

" "0 a3
E0 cos ;
" + 2"0 r2

(1.234)

is of dipole form with an eective dipole moment

pz = 4 " 0

" "0 3
a E0 ;
" + 2"0

(1.235)

located at the center of the sphere.

In the limit of "
"0 ; the potential reduces to

(r; ) =

8
0;
>
>
<
>
>
:

r<a

a3
0 (r; ) + 2 E0 cos ; r > a
r

(1.236)

which describes the potential when a conducting sphere is placed in an external electric eld. A
sphere of innite permittivity is mathematically identical to a conducting sphere.
The polarization vector P can be found from
Dz = "0 Ez + Pz ;
Pz =

3(" "0 )
" 0 E0 ;
" + 2"0

(1.237)
r < a:

(1.238)

Since P = 0 outside the sphere, the divergence of the polarization vector yields the bound charge
density induced on the sphere surface,
e

=
=

r P
2(" "0 )
"0 E0 cos
" + 2"0

(r

The total dipole moment carried by the sphere is

pz =
=

4 a3
Pz
3
4 (" "0 )a3
" 0 E0 ;
" + 2"0

which agrees with that deduced from the exterior potential.

50

a):

(1.239)

1.10

Electric Force

The energy density associated with an electric eld

1
"0 E 2 ;
2

(J/m3 )

(1.240)

manifests itself as either pressure or tensile stress depending on the direction of the force relative
to the electric eld. A charge density placed in an electric eld E experiences a force per unit
volume,
=

= "0 (r E)E
= r ("0 EE)

"0 E rE:

(1.241)

1
In electrostatics, r E = 0; and thus the curvature E rE and gradient ErE = rE 2 of the
2
electric eld are identical,
r(E E) = 2E

rE 2 = 2E rE:

E + 2E rE
(1.242)

Therefore,
f =r

"0 EE

1
"0 E 2 1 ;
2

(1.243)

1
"0 E 2
2

(1.244)

where 1 is the unit sensor. The tensor

Tij = "0 Ei Ej

ij ;

is called the Maxwells stress tensor associated with the electric eld. For a linear dielectric, "0
may be replaced with its permittivity ": The force vector is
f = r T;

or fi =

@
Tij :
@xj

(1.245)

Let us consider a simple case: an electric eld in the z-direction, E = Ez ez : The tensor Tij is
diagonal with the following components,
0

B
B
T=B
B
@

1
"0 Ez2
2
0
0

1
"0 Ez2
0
2
1
0
+ "0 Ez2
2

51

C
C
C:
C
A

(1.246)

The force in the directions perpendicular to the eld are

fx =

@
@x

1
"0 Ez2 ;
2

(1.247)

fy =

@
@y

1
"0 Ez2 ;
2

(1.248)

which appear as a pressure acting from a higher energy density region to lower energy density
region. The force in the direction of the electric eld is
fz = +

@
@z

1
"0 Ez2 ;
2

(1.249)

which appears as tension acting from a lower energy density region to higher energy density region.
The force to act on a volume V can be found from the integral
Z
I
F =
f dV =
T dS
V
S
I
1
=
"0 E(E n)
"0 E 2 n dS;
(1.250)
2
S
where n is the unit normal vector on the closed surface S; dS = ndS: For a charged conducting
sphere, the electric eld is radially outward everywhere. At the surface, the force per unit area is
1
"0 Er2 ; (N/m2 );
2
acting radially outward. Since E = 0 inside a conducting sphere, the force acts from lower energy
density region to higher energy density region as expected of tensile stress. Of course, no net force
acts on the sphere.
Example 7 Dielectric Hemispheres in an External Electric Field
As a less trivial example, we consider a dielectric sphere consisting of identical hemispheres with
a narrow gap at an equatorial plane. It is placed in an electric eld with the gap plane perpendicular
to the eld. We wish to nd a force to act between the hemisphere. The force, if any, should be
axial in the direction perpendicular to the gap. The z-component of the integral,
F=

is
Fz =

"E(E n)

"Ez En

1 2
"E n dS
2
1 2
"E nz dS;
2

(1.251)

where on the spherical surface of one of the hemispheres,

Ez = E0

3
" cos2 + "0 sin2
" + 2"0
52

(1.252)

Er = E0
E =

3"
cos ;
" + 2"0

E0

(1.253)

3"0
sin ;
" + 2"0

E 2 = Er2 + E 2
9
= E02
"2 cos2 + "20 sin2
(" + 2"0 )2

(1.254)

(1.255)

In the gap, the eld is uniform,

Figure 1-18: A dielectric sphere with a negligible gap at an equator placed in an electric eld normal
to the gap surface. The hemispheres attract each other. The D-led lines (not E eld lines) shown
are relevant to the preceding Example as well.

Ezgap =

"
3"
Eiz =
E0 ;
"0
" + 2"0

where
Eiz =

3"0
E0 ;
" + 2"0

(1.256)

(1.257)

is the internal electric eld in the dielectric sphere given in Eq. (1.232). The integral over the
hemispherical surface is
Fz1 = 2 a2

=2

" 0 Ez Er

1
"0 E 2 cos
2

sin d

Z =2
9"0 E02
"2 cos2 + ""0 sin2
(" + 2"0 )2 0
9"0 E02 1 2
a2
(" + 2""0 "20 ):
(" + 2"0 )2 4

= 2 a2
=

53

1 2
(" cos2 + "0 sin2 ) cos sin d
2
(1.258)

This force is repelling. The contribution from the at surface in the gap is
9"0 E02 "2
;
(" + 2"0 )2 2

(1.259)

9(" "0 )2
"0 E02 :
4(" + 2"0 )2

(1.260)

a2

Fz2 =
and the net force is
a2

Fz =

The minus sign indicates an attractive force between the hemispheres. In the limit "
force becomes
9 2
Fz =
a "0 E02 :
4

"0 ; the

This corresponds to the case of solid or closed conducting hemispheres.

1.11

Force in Capacitor

For two electrode systems, the potential dierence between the electrodes V and the charges
residing on each electrode are related through the capacitance C;
Q = CV:

(1.261)

An incremental energy required to increase the charge by dQ is

dU = V dQ =

1
QdQ = CV dV:
C

(1.262)

Therefore the amount of energy stored in a capacitor is

U=

1 Q2
1
= CV 2 :
2 C
2

(1.263)

This is applicable for a single electrode as well if the second electrode is at innity (self-capacitance).
For example, a conducting sphere of radius a has a self-capacitance of
C = 4 "0 a:

(1.264)

If it carries a charge Q; the amount of potential energy associated with it is

U=

Q2
:
8 "0 a

(1.265)

Of course, the energy is stored in the space surrounding the sphere, and the energy can alternatively
be calculated from the integral,
U=

Q2
1
"0 Er2 4 r2 dr =
;
2
8 "0 a

54

(1.266)

where
Er =

Q
;
4 "0 r 2

r>a

(1.267)

is the electric eld.

If a capacitance is described as a function of geometrical factor ; such as electrode separation
distance and electrode size, the electric force tends to act in such a way to increase the capacitance,
1 @C( )
F = V2
; if the voltage is held constant,
2
@
F =

1 2@
Q
2 @

1
C( )

; if the charge is held constant.

(1.268)

(1.269)

For example, the capacitance of parallel plates capacitor consisting of circular disks is
C(d; a) = "0

a2
;
d

(1.270)

where a is the disk radius and d is the separation distance. The disks evidently attract each other
so as to reduce the distance d or increase the capacitance with a force
Fd =

a2
1
"0 2 V 2 =
2 d

The radial force is

Fa = " 0

1 Q2
:
2 "0 a2

d
a 2
V =
Q2 ;
d
"0 a3

(1.271)

(1.272)

which acts so as to increase the radius of the disks. Note that when the charge is xed, the force
acts to reduce the energy stored in the capacitor, while when the voltage is xed, the force acts
to increase the energy. A power supply is a large reservoir of energy and a capacitor connected to
a power supply (the case of xed voltage) is not a closed system. The earlier statement based on
the variational principle that an electrostatic equilibrium is the minimum energy state of course
pertains to closed systems.

55

Problems
1.1 Verify that for each of the expressions of the three dimensional delta functions,
(r

r0 ) = (x

(r

r0 ) =

(r

x0 ) (y
0)

(
0

the volume integral is unity,

(r r0 )
(cos
rr0

r0 ) =

y 0 ) (z
) (z

cos 0 ) (

z 0 );
z 0 );
0

);

r0 )dV = 1:

(r

1.2 The electron cloud in a hydrogen atom is described by the charge density distribution
e
exp
a3

(r) =
where a = 5:3
atom is

10

11

2r
a

m is the Bohr radius. Show that the potential energy of a hydrogen

U=

1 e2
:
4 "0 a

1.3 A planar quadrupole in the x y plane consists of four charges, +q and q alternatively
placed at the corners of a square of side a: What is the potential energy of the quadrupole?
1.4 Four charges e; e; e and e are placed at the corners of a tetrahedron having side a:Find
the electric dipole and quadrupole moments.

1.5 Show that for l = 1;

1
X

rY1;m ( ; )q1;m =

m= 1

56

3
r p;
4

and a resultant dipole potential is consistent with the direct expansion of the potential,
dipole

1 r p
:
4 "0 r 3

1.6 An octupole consists of eight charges (four +q and four q) alternatively placed at the corners
of a cube of side a: Determine the far eld potential (r; ; ) at r
a and also the potential
energy of the octupole.
1.7 Equal octants on a spherical surface of radius a are maintained alternatively at potentials V
and V: Determine the lowest order far eld potential (r; ; ) at r
a: Useful expansion
is
I
X a l+1
0 0
0 0
0
(r) =
Ylm ( ; )
s ( ; )Ylm ( ; )d ; r > a;
r
lm

where

s(

; ) is the surface potental and d

1.8 The potential due to a long line charge

= sin 0 d 0 d 0 :

(C/m) is

( )=
where

2 "0

ln + constant,

is the distance from the line charge. Verify this from the basic formula,
(r) =
=

Z
(r0 )
1
dV 0
4 "0
jr r0 j
Z 1
1
p
dz 0 :
2 + z 02
4 "0 1

Then, show that the capacitance per unit length of a parallel-wire transmission line with wire
separation distance d and common wire radii a ( d) is approximately given by
C
'
l

"0
; (F/m).
d
ln
a
57

What is the inductance per unit length of the transmission line? (Hint: The product
CL
;
l l
is constant and equal to "0 0 : For a coaxial cable lled with insulating material having
permittivity "; the capacitance is
2 "
C
=
;
l
ln(a=b)
and the product

CL
l l

is equal to "

0 :)

1.9 An insulating circular disk of radius a carries a total charge q uniformly distributed over
its area a2 : By rst nding the potential on the axis of the disk, (z); generalize it to a
potential at arbitrary position (r; ) in terms of the spherical harmonic functions Pl (cos ):
(The case of charged conducting disk will be analyzed in Chapter 2.) Check whether the
Poissons equation,
;
r2 =
"0
is satised by your solution. (It should be.)
1.10 Solve the preceding problem using the cylindrical coordinates ( ; z).
1.11 The Legendre polynomial Pl (x) can be generated from
Pl (x) =

1 dl 2
(x
2l l! dxl

1)l ; (Rodriguesformula).

Using this repeatedly, verify the orthogonality of Legendre functions,

Z

1
1

Pl (x)Pl0 (x)dx =

2
2l + 1

ll0 :

1.12 Evaluate numerically the capacitance of a conducting torus having a major radius of 10 cm
and minor radius of 3 cm.
1.13 Two concentric circular rings of radii a and b in the same plane carry charges q and q;
respectively. Show that the far eld potential is of quadrupole nature. What is the potential
energy of the system?
1.14 Two coaxial conductor rings of radii a and b and axial separation distance c carry charges q
and q: What is the dominant far eld potential? Find the mutual capacitance.
1.15 The permittivity of a molecular hydrogen (H2 ) gas under the standard condition, 0 C, one
atmospheric pressure, is
" = "0 + " = 1 + 2:7 10 4 "0 :

58

When

" is written in the form

! 2p
ne2
"
=
const:
;
0
! 20
me ! 20

" = const:

where n is the molecule density, me = 9:1 10 31 kg is the electron mass, and ! 0 is the
frequency of bound harmonic motion of electron ~! 0 = 27:2 eV, what should the constant
be? For atomic hydrogen, the constant is 4.5 as predicted by quantum mechanics.
1.16 Two dipole moments p1 and p2 are a distance r apart. Show that the potential energy of the
two dipole system is given by
U=

1 1
4 "0 r 3

p1 p2

(p1 r)(p2 r)
r2

1.17 Two parallel charge sheets of opposite polarity

(C/m2 ) separated by a small distance
form a double layer. (a) Show that the potential jump across the double layer is
=

"0

"

where =
(C m 1 ) is called the moment of double layer. (b) A circular double layer of
radius a and moment is on the x y plane. Find the potential (r; ):
1.18 Concentric spherical capacitor of radii a (inner) and b (outer) is lled with a nonuniform
dielectric material whose permittivity depends on the radius r; "(r): Determine "(r) if the
radial electric eld between the electrodes is to be constant. Assume that the permittivity at
r = b is "0 :
1.19 The problem of ring charge can alternatively be solved by the method of Laplace transform
in the cylindrical coordinates as shown in Chapter 1,
q
( ; z) =
4 "0

J0 (ka)J0 (k )e

kjzj

dk:

Show that the solution above satises the Poissons equation,

1 @
@2
@2
+
+
@ 2
@r @z 2

( ; z) =

q
(
2 "0 a

a) (z):

1.20 A conducting sphere of radius a is placed in a uniform electric eld E0 = E0 ez whose potential
is 0 = E0 z: The dipole moment induced on the sphere is
pz = 4 "0 a3 E0 ;

59

which produces a potential

0

1 pz
cos :
4 "0 r 2

(a) The exterior electric eld is

r 0
pz
= E 0 ez +
(2 cos er + sin e ) ; r > a:
4 "0 r 3

E(r; ) =

Find a change in the total electric energy and interpret your result. Note that E = 0 in
the sphere.
(b) A dipole p in an electric eld E has a potential energy
U=

p E=

Recover this result from the direct integral,

I
1
U=
2

4 "0 a3 E02 :

0 dS;

over the sphere surface. Note that at a conductor surface,

= "0 Er (r = a):

1.21 A charge q1 is placed at a distance z = a from a dielectric plate having a permittivity ":
Another charge q2 is at the mirror position z = a: Find the force on each charge. As you
will nd, the forces are not equal in magnitude. Explain.

Hint: If a charge q is placed at a distance z = a from the surface of a dielectric plate, the
potential in the air region is
air

1
4 "0

q
jr

aj

60

" "0 q
" + "0 jr + aj

; z>0

and that in the dielectric is

dielectric

2"
q
1
; z < 0;
4 " " + "0 jr aj

where a = aez :
1.22 The permittivity of an unmagnetized plasma is
" (!) = "0

! 2pe
!2

p
where ! pe = ne2 =me "0 is the plasma frequency. Show that a spherical plasma exhibits a
dipole (l = 1) ocsillation at a frequency
! pe
!= p :
3
What are the frequencies of higher multipole modes?

61

Chapter 2

Electrostatics II. Potential Boundary

Value Problems
2.1

Introduction

In Chapter 1, a general formulation was developed to nd the scalar potential

electric eld E =

(r) and consequent

for a given static charge distribution (r): In a system involving conductor

electrodes, often the potential

is specied on electrode surfaces and one is asked to nd the po-

tential in the space o the electrodes. Such problems are called potential boundary value problems.
In this case, the surface charge distribution on the electrodes is unknown and can only be found
after the potential and electric eld have been found in the vicinity of the electrode surfaces from
s

= "0 En ;

(C/m2 )

where
@
;
@n

En =

is the electric eld component normal to the conducting electrode surface with n the normal coordinate.
If the potential is specied on a closed surface, the potential o the surface is uniquely determined in terms of the surface potential. This is known as Dirichlets boundary value problem and
most problems we will consider belong to this category. Solving Dirichlets problems is greatly
facilitated by nding a suitable Greens function for a given boundary shape. However, except
for simple geometries (e.g., plane, sphere, cylinder, etc.), nding Greens functions analytically is
not an easy task. For complicated electrode shapes, potential problems often have to be solved
numerically.

Specifying the normal derivative @ =@n on a closed surface also uniquely determines the potential elsewhere. This category of boundary value problems is called Neumann problem. Physically,
specifying the normal derivative of the potential on a closed surface corresponds to specifying the
surface charge distribution on the surface through
=

"0

@
:
@n

Then, the problem is reduced to nding the potential due to a prescribed charge distribution as
worked out in Chapter 1.
Specifying both the potential itself and its normal derivative everywhere on a closed surface is
in general overdetermining. However, in some problems, the potential is known in one part of a
closed surface and its normal derivative in the remaining part. This constitutes the so-called mixed
boundary value problem.
By introducing suitable coordinates transformation, some potential problems can be reduced
to one dimensional, that is, the potential becomes a total function of a single coordinate variable. This happens if the Laplace equation and potential are completely separable,

(u1 ; u2 ; u3 ) =

F1 (u1 )F2 (u2 )F3 (u3 ): There are some 30 known rectilinear coordinate systems developed in the past
for specic purposes. As one example, we will study the oblate spheroidal coordinates because of
its wide variety of applications in electrostatics and magnetostatics.

2.2

Dirichlet Problems and Greens Functions

If a charge is given to a conductor, the potential of the conductor becomes constant everywhere after
a short transient time as shown in Chapter 1. Electrostatic state is thus quickly established. Since
the volume charge density

should vanish in a conductor, all of the charge given to a conductor

must reside entirely on the conductor surface in the form of singular surface charge density
(C/m2 ): The corresponding volume charge density involves a delta function
=

(n ns );

where n is the coordinate normal to the surface and ns indicates the location of the surface.
After static condition is established, the volume charge density and the electric eld in a conductor both vanish. The potential of a conductor thus becomes constant

= const. If a

charge q is given to an isolated conductor, the potential of the conductor relative to zero potential
at innity is uniquely determined and the proportional constant C denes the self-capacitance of
the conductor,
C=

q
c

; (F):

(2.1)

Let us consider a trivial case, a conducting sphere of radius a carrying a charge q: The potential
outside the sphere is given by
q 1
;
4 "0 r

(r) =

a:

(2.2)

The sphere potential is

s

q
;
4 "0 a

(2.3)

which determines the self-capacitance of the sphere,

C = 4 "0 a; (F).
The outer potential

(2.4)

(r) can be written in the form

(r) =

a
r

s;

r > a;

(2.5)

which indicates that the potential is uniquely determined if the sphere potential

is known. In

general, if the potential is specied everywhere on a closed surface, the potential elsewhere o
the surface is uniquely determined in terms of the surface potential

s (rs )

where rs denotes the

coordinates on the closed surface. This is known as Dirichlets theorem and nding a potential for
given boundary potential distribution on a closed surface is called Dirichlets problem.
The same problem can also be solved in terms of the electric eld on the sphere surface,
Er =

1 q
;
4 "0 a2

(2.6)

q
; (C/m2 ):
4 a2

(2.7)

which can be replaced with a surface charge,

= "0 Er =

The potential due to the uniform surface charge is

(r) =
=
=
where

1
4 "0

jr
2

r0 j

dS

p
2 a
4 "0
r 2 + a2
0
q
; (r > a)
4 "0 r

1
2ar cos

sin d
(2.8)

is measured from the direction of r: (This is allowed because of symmetry. For r < a; the

integral yields
(r) =

q
= const.; (r < a; interior)
4 "0 a

which is also an expected result.) Since

@
@
=
;
@r
@n

Er =

(2.9)

where n is the normal coordinate on the surface directed away from the volume of interest, the
potential can be rewritten as
(r) =

1
4

1
r0 j

jr

@
dS 0 :
@n

(2.10)

As this simple example indicates, potential boundary value problems can be solved in terms of
either the surface potential

or its normal derivative, @ =@n: The latter method may be regarded

as a boundary value problem for the electric eld.

Let us revisit the potential due to a prescribed charge distribution,
1
(r) =
4 "0

(r0 )
dV 0 :
jr r0 j

(2.11)

The potential can be understood as a convolution between the charge density distribution (r) and
the function
G(r; r0 ) =

1
1
;
4 jr r0 j

(2.12)

which is the particular solution to the singular Poissons equation

r2 G =

(r

r0 );

(2.13)

subject to the boundary condition that G vanish at innity. The function G is called Greens
function. Physically, the Greens function dened as a solution to the singular Poissons equation
is nothing but the potential due to a point charge placed at r = r0 : In potential boundary value
problems, the charge density (r) is unknown and one has to devise an alternative formulation
in terms of boundary potential

s (r):

It is noted that the Greens function in Eq. (2.12) is the

particular solution to the singular Poissons equation and we still have freedom to add general
solutions satisfying Laplace equation,
G = Gp + Gg ;

(2.14)

where Gp is the particular solution and Gg is a collection of general solutions satisfying

r2 Gg = 0:

(2.15)

This freedom will play an important role in constructing a Greens function suitable for a given
boundary shape as we will see shortly. In doing so, we exploit the following theorem:

Theorem 1 Greens Theorem: For arbitrary scalar functions

Z

dV =

( r

and ; the following identity holds,

r ) dS:

(2.16)

Proof of this theorem goes as follows. Gausstheorem applied to the function r

Z

r ( r )dV =

( r ) dS:

gives
(2.17)

The LHS may be expanded as

Z

r ( r )dV =

Therefore,

and

r + r2

r + r2

dV =

r + r2

dV =

Exchanging

dV:

(2.18)

( r ) dS:

(2.19)

( r ) dS:

(2.20)

Subtracting Eq. (2.20) from Eq. (2.19) yields

Z

dV =

( r

r ) dS;

(2.21)

which is the desired identity.

Figure 2-1: s (r0 ) is the potential specied on a closed surafce S; n is the coordinate normal to the
surface directed away from the volume wherein the potential (r) is to be evaluated.
We now apply the formula to electrostatic potential problems. Let
5

be the Greens function

= G; satisfying
r2 G =
and

r0 );

(r

(2.22)

be the scalar potential satisfying Poissons equation

r2

"0

(2.23)

Then, the terms in the LHS of Eq. (2.21) become

Z

r2r0 GdV 0

r0

(r

r0 )dV 0

(r);

provided the coordinates r resides in the volume V where we wish to nd the potential, and
Z

Gr2 dV 0 =

1
"0

@
@n

G (r0 )dV 0 :

(2.24)

The RHS of Eq. (2.21) reduces to

I

@G
@n

where

dS;

(2.25)

is the potential on the closed surface and n is the coordinate normal to the surface directed

away from the volume of interest as indicated in Fig.2-1. Therefore, the solution for the potential
(r) is given by
1
(r) =
"0

G (r )dV

@G
@n

@ s
@n

dS:

(2.26)

At this stage, the Greens function is still arbitrary except it should satisfy the singular Poissons
equation in Eq. (2.22). The rst term in the RHS allows evaluation of the potential for a given
charge distribution as we saw earlier. The surface integral involves the potential on the closed
surface

and its normal derivative, namely, the normal component of the electric eld at the

surface.
In usual boundary value problems, the potential on a closed surface is specied as a function
of the surface coordinates. In this case, it is convenient to choose the Greens function so that it
vanishes on the surface,
G = 0 on S:
Then the last term in Eq. (2.26) vanishes, and the solution for the potential becomes
1
(r) =
4 "0

(r0 )
dV 0
jr r0 j

@G 0
dS ;
@n

G = 0 on S:

(2.27)

In particular, if there are no charges in the region of concern

= 0; the potential is uniquely

determined in terms of the surface potential alone,

I

(r) =

@G 0
dS ;
@n

= 0 in V; G = 0 on S:

(2.28)

We have a freedom to make such a choice for the Greens function that it vanish on the closed
surface S through adding general solutions to the particular solution of the singular Poissons
equation. Therefore, solving a potential boundary value problems for a given closed surface S boils
down to nding a Greens function satisfying
r2 G =

(r

r0 );

G = 0 on S:

(2.29)

Once such an appropriate Greens function is found for a given surface shape S; the potential at
arbitrary point can be found from Eq. (2.28) for a specied potential distribution

s (r)

on the

surface.
In the following, Greens functions for some simple surface shapes will be found. It is noted
that three dimensional Greens functions have dimensions of 1/length, two dimensional Greens
functions are dimensionless, and one dimensional Greens functions have dimensions of length.

2.3
2.3.1

Examples of Greens Functions

Plane

Suppose that the potential is specied everywhere on an innite (x; y) plane,

s (x; y):

The plane

is closed at innity and the method of Greens function is applicable. The Greens function is to
be found as a solution to the equation
r2 G =

(x

x0 ) (y

y 0 ) (z

z 0 );

(2.30)

with the boundary condition G = 0; z = 0: Mathematically, the Greens function is equivalent to

the potential due to a point charge placed near a grounded conducting plate that can be readily
worked out using the method of image as shown in Fig.2-2,
1
G(r; r0 ) =
4

p
(x

1
x0 )2 + (y

y 0 )2 + (z

z 0 )2

p
(x

1
x0 )2 + (y

y 0 )2 + (z + z 0 )2

; (2.31)

where the second term in the RHS is the contribution from the image charge at the mirror point.
Note that the Greens function is reciprocal and remains unchanged against the coordinates inter-

change,
G(r; r0 ) = G(r0 ; r):
This is expected from the fact that the delta function in the original singular Poissons equation is
even,
r0 ) = (r0 r):

(r
In the upper region z > 0;
@G
@n

=
=

@G
@z 0 z 0 =0
z
1
:
0
2
2 [(x x ) + (y y 0 )2 + z 2 ]3=2
0 0
s (x ; y )

Therefore, for a surface potential

(2.32)

specied as a function of (x0 ; y 0 ); the potential in the

region z > 0 is given by

(r) =

Figure 2-2: Image charge

q vanishes at the plate.

z
2

1
1

dx0

dy 0

s (x

[(x

x0 )2 + (y

0; y0)

y 0 )2 + z 2 ]3=2

(2.33)

q for a large, grounded conducting plate. The potential due to q and

Let us apply this formula to the boundary condition on the (x; y) plane,

s(

where

)=

V;

<a

0;

>a

(2.34)

p
x2 + y 2 is the radial distance on the plane as shown in Fig. 2-3. Physically, the

boundary condition describes a large conducting plate which is grounded except for a circular

region of radius a whose potential is maintained at V: The potential on the z-axis can be found

Figure 2-3: A large conducting plate is grounded except for a circular region which is at a potential
V:
easily,
zV
2

(z) =

= V

0d 0

+ z 2 )3=2
z
p
;
2
z + a2

02

z > 0:

(2.35)

The axial potential in the lower region z < 0 can be found by observing the up-down symmetry
and for both regions,
(z) = V

jzj
p
2
z + a2

(2.36)

Then, the potential at arbitrary point (r; ) is

(r; ) =

8
>
>
V 1
>
>
<
>
>
>
>
: V

Note that at r

1
2

r
jP1 (cos )j +
a

r
a

1
2

jP3 (cos )j

3
8

r
a

jP5 (cos )j +

; r<a
(2.37)

a
r

jP1 (cos )j

3
8

a
r

jP3 (cos )j +

r>a

a; the potential is of dipole type,

(r

a) _

1
jcos j :
r2

(2.38)

This problem should not be confused with the potential due to an isolated charged conducting disk
which will be discussed later. The potential and electric eld in the upper half region are identical
to those realized by an ideally thin circular capacitor whose top plate is at a potential V and the
lower plate at

V: The appearance of the dipole potential is thus an expected result. (For a thin

capacitor with plate separation distance , the electric eld between the plates diverges but the
product E = 2V remains constant. Such a structure is called a double layer.)
9

2.3.2

Sphere

Figure 2-4: The image of charge q with respect to a grounded conducting sphere is q 0 =
located at a2 r0 =r02 where r0 is the location of the charge q:

qa=r0

In nding a Greens function for a given surface shape, the method of images is most conveniently
exploited. In the case of a sphere having a radius a; the Greens function can be found as a solution
for the potential due to a charge q placed at a distance r0 from the center of a grounded conducting
sphere. In the case of a sphere having radius a, an image charge
q0 =

a
q;
r0

(2.39)

r00 =

a2 0
r;
r02

(2.40)

placed at

together with the charge q, makes the surface potential vanish. This is illustrated in Fig.2-4. The
potential due to charge q placed near a grounded conducting sphere is thus equivalent to that due
to two charges, q and its image charge q 0 ; and is given by

(r) =

a
r0

q @ 1
A
4 "0 jr r0 j jr r00 j
0
q @
1
p
2
02
4 "0
r +r
2rr0 cos

10

1
2

(rr0 =a)

a2

2rr0 cos

A;

(2.41)

which readily yields the Greens function for a sphere,

0
1
1
@p
G(r; r0 ) =
4
r2 + r02 2rr0 cos
Here

q
(rr0 =a)2 + a2

2rr0 cos

A:

is the angle between the two position vectors r = (r; ; ) and r0 = (r0 ; 0 ;

(2.42)

). Its cosine

value is
cos

= cos cos

+ sin sin

cos(

):

(2.43)

The Greens function indeed vanishes on the sphere surface r = a or r0 = a: Again, the Greens
function is invariant against coordinates exchange, r $ r0 ; that is, Greens functions are reciprocal.
For exterior (r > a) potential problems, the normal gradient @G=@n is
@G
@n

@G
@r0

r0 =a

1
4 (r2 + a2

r2
a
;
2ar cos )3=2

r>a

(2.44)

r < a:

(2.45)

and for interior (r < a) problems,

@G
@n

= +

@G
@r0

r0 =a

r2
a

1
4 (r2 + a2

a
2ar cos )3=2

Note that the normal coordinate n is directed away from the volume of interest. If the surface
potential is specied as a function of

and

s(

); and there are no charges, the exterior

potential at an arbitrary point r = (r; ; ) can be found from

(r) =

1
4

a(r2
4

@G
dS
@n
a(r2

(r2 + a2
Z
a2 )
0

2ar cos )3=2

Z 2
0 0
sin d
d

Recalling the expansion of the function 1= jr

1
jr

r0 j

a2 )

s(

)d

s(

(r2 + a2

2ar cos )3=2

(2.46)

r0 j in terms of the spherical harmonic functions

X 4
r0l
Ylm ( ; )Ylm ( 0 ;
2l + 1 rl+1
l;m

11

);

r > r0

(2.47)

the exterior potential can be decomposed into multipole potentials,

X a
(r; ; ) =
r

l+1

Ylm ( ; )

l;m

s(

)Ylm ( 0 ;

)d 0 ;

r > a:

(2.48)

The interior potential can be found using the expansion

1
jr

r0 j

X 4
rl
Ylm ( ; )Ylm ( 0 ;
2l + 1 r0l+1

r < r0 ;

);

(2.49)

l;m

X r
(r; ; ) =
a

Ylm ( ; )

l;m

s(

)Ylm ( 0 ;

)d 0 ;

r < a:

(2.50)

Example 2 Charge near a Floating Conducting Sphere

To become familiar with the Greens function method, let us consider a somewhat trivial problem
of nding the potential when a charge q is placed at a distance d from the center of a oating
a
dq

conducting sphere of radius a: The charge q and its image q 0 =

at (a=d)2 d make the sphere

potential 0 as we have just seen. However, since the oating sphere should carry no net charge, a
charge

q 0 = ad q must be placed at the center of the sphere which raises the sphere potential to
s

q0
q
=
; d > a:
4 "0 a
4 "0 d

Therefore, the exterior potential can be found by summing contributions from q; its image q 0 and
the charge

q 0 at the center,
(r) =

1
4 "0

q
jr

jr

dj

qa=d
qa=d
+
(a=d)2 dj
r

In this expression, the function

1
4

1
jr

dj

a
d jr

1
(a=d)2 dj

is the Greens function which vanishes on the sphere surface, r = a: The last term is in the form
s

a
;
r

where
s

q
; (independent of a)
4 "0 d

12

is the surface potential. Indeed,

I

where

@G
dS =
s
@n

a(r2 a2 )
4
a(r2

a
s ;
r

a2 )
2

sin d

s(

(r2

a2

2ar cos

2ar cos )3=2

a2

1
r2

0 3=2

sin 0 d

is measured from the direction of the vector d. (This is allowed because of the symmetry.)

Example 3 Specied Potential on a Sphere Surface

Figure 2-5:

= +V for 0 <

< =2;

V for =2 <

< :

Let us nd the potential outside a spherical shell of radius a whose top half is maintained at
potential +V and lower half at

V;

s(

)=

8
< +V;
:

V;

;
(2.51)

as shown in Fig.2-5. Because of axial symmetry, only m = 0 terms survive the integration over
the azimuthal angle

: Also, because of up-down antisymmetry, only odd l terms survive the

integration over the polar angle

I

: Noting
; 0 )Yl0 ( 0 ; 0 )d 0
r
Z
2l + 1 1
Pl ( )d ;
2V
4
0

s(

= 2

l = 1; 3; 5;

we readily nd the exterior potential,

(r; ) = V

3 a
2 r

P1 (cos )

7 a
8 r
13

P3 (cos ) +

a:

(2.52)

The interior potential is

(r; ) = V

3r
P1 (cos )
2a

7 r
8 a

P3 (cos ) +

a:

(2.53)

The surface charge density on the sphere can be found from the normal component of the
electric eld,
=
=

"0
"0 V
a

@
@r

r=a+0

3P1 (cos )

7
P3 (cos ) +
2

The total surface charge on the upper hemisphere

q = 2 a2

=2

( ) sin d ;

simply diverges (albeit only logarithmically) and it is not possible to dene the capacitance of the
hemispheres. This is because of the assumption of ideally small gap separating the two hemispheres.
If a small gap

2.3.3

a is assumed, a nite capacitance containing a factor ln(a= ) emerges.

Interior of Cylinder of Finite Length

Figure 2-6: Cylinder of a nite length.

14

The Greens function for the interior of a cylinder of radius a and length l shown in Fig.2-6 can
be found as a solution for the following singular Poissons equation
1 @
@2
@2
1 @2
+
+
+
2
@ 2
@
@ 2 @z 2

r2 G =

0)

G=

) (z

z 0 );

(2.54)

with the boundary condition

G = 0;

= a; z = 0 and l:

Since the Greens function should be periodic with respect to

respect to exchange of

and

(2.55)
and should also be invariant with

; the angular dependence can be assumed to be cos[m(

)] where

m is an integer. Assuming the following separation of variables,

G(r; r0 ) =

Rm ( )Zm (z) cos m(

);

(2.56)

we see that the radial function Rm ( ) and the axial function Zm (z) satisfy, respectively,
1 d
d2
+
2
d
d
d2
dz 2

m2
2

+ k 2 Rm ( ) = 0;

(2.57)

k 2 Zm (z) = 0;

(2.58)

where k 2 is a separation constant which can be either positive or negative.

Let us rst consider the case k 2 > 0: Solutions for Rm ( ) which satises the boundary condition
Rm ( = a) = 0 is the m-th order Bessel function,
Rmn ( ; 0 ) = Jm

xmn
a

Jm

xmn
a

(2.59)

where xmn is the n-th root of Jm (x) = 0: (The Bessel function of the second kind Nm (x) is discarded
because it diverges on the axis,

= 0:)

Solutions for the axial function Zm (z) are e

kz

or sinh(kz) and cosh(kz): The boundary con-

dition for Zm (z) is it vanish at z = 0 and l: Therefore, we can construct the axial function as
follows,
Zm (z; z 0 ) =

8
>
>
< sinh(kmn z) sinh[kmn (l
>
>
: sinh[k (l
mn

z 0 )];

z)] sinh(kmn z 0 );

0 < z < z 0 < l;

(2.60)
0 < z 0 < z < l;

where kmn = xmn =a: A more fancy way to write Zm (z; z 0 ) is

Zm (z; z 0 ) = sinh[kmn min(z; z 0 )] sinhfkmn [l
15

max(z; z 0 )]g:

(2.61)

The Greens function may thus be assumed in the form

G(r; r0 ) =

Amn Rmn ( ; 0 )Zmn (z; z 0 ) cos[m(

)]:

(2.62)

m;n

The expansion coe cient Amn can be determined from the discontinuity in the derivative of the
axial function Zmn (z; z 0 ) at z 0 ;
d
Zmn
dz
d
Zmn
dz

kmn cosh[kmn (l

z 0 )] sinh(kmn z 0 );

z=z 0 +0

= +kmn cosh(kmn z 0 ) sinh[(kmn (l

z=z 0

z 0 )]:

Then, a singularity appears in the second order derivative,

d2
Zmn =
dz 2

z 0 );

kmn sinh(kmn l) (z

(2.63)

which is compatible with the delta function in the RHS of the original singular Poissons equation
in Eq. (2.54). Eq. (2.54) now reduces to
X

Amn kmn sinh(kmn l)Jm (kmn )Jm (kmn 0 ) cos m(

)=

0)

):

(2.64)

mn

Multiplying both sides by

0
0 J (k
0
m mn ) cos m

A0n =

a2 k

and integrating over

1

;
2
mn Jm+1 (kmn a) sinh(kmn l)

2
1
;
2
a2 kmn Jm+1
(kmn a) sinh(kmn l)

Amn =

and

; we nd

m = 0;

(2.65)

(2.66)

1;

where use has been made of the following integral,

Z

2
Jm
(kmn )d =

a2 2
(kmn a):
J
2 m+1

(2.67)

The nal form of the desired Greens function is

xmn
xmn 0
Jm
Zmn (z; z 0 )
1 X
1 Jm
X
1
a
a
G(r; r0 ) =
cos[m(
2
a
xmn Jm+1
(xmn ) sinh(kmn l)
m=0 n=1

16

)]"m ;

(2.68)

where
"m =

1; m = 0
2; m

If one does not like the appearance of "m ; the summation over m can be changed to from
1;
1
X

1
a m=

G(r; r0 ) =

1
X

Jm

1 n=1

If it is assumed that k 2 =

xmn 0
xmn
Zmn (z; z 0 )
Jm
a
a
cos[m(
2
(xmn ) sinh(kmn l)
xmn Jm+1

)]:

1 to

(2.69)

< 0; appropriate general solutions to

1 d
d2
+
2
d
d
d2
+
dz 2

m2

Rm ( ) = 0;

(2.70)

Zm (z) = 0;

(2.71)

are

with

Rm

Rm

= [Km (

Km (

m a) Im ( m

m a) Im

Im (
Im (

m a) Km

m a) Km ( m

Im (km ) ;

)] Im km

<
0

<

< a; (2.72)

< a;

(2.73)

= m =l and
Zm (z) = sin (

m z) sin

mz

(2.74)

from which the Greens function can be constructed. Remaining calculation is left for exercise. The
0)

reader should appreciate how a delta function (

appears from the term

d2 Rm
:
d 2

(2.75)

One may wonder about Greens function for the exterior region of a cylinder of nite length.
This problem appears to be a di cult one and analytical expressions are not available to the
authors knowledge. It may be the case the problem can only be solved numerically.

2.3.4

Long Cylinder (3-Dimensional)

Three dimensional Greens function for a long cylinder satises

r2 G =

@2
1 @
1 @2
@2
+
+
+
2
@ 2
@
@ 2 @z 2

17

G=

0)

) (z

z 0 );

(2.76)

which is to be solved for the boundary conditions

G( = a) = 0;

G(z =

1) = 0:

(2.77)

Following the same procedure as in the preceding example, the interior solution for interior ;

<a

may be assumed as
G(r; r0 ) =

Amn Jm (kmn )Jm (kmn 0 ) cos[m(

)] exp[ kmn jz

m;n

z 0 j];

(2.78)

where
xmn
;
a

kmn =

(2.79)

and xmn is the n-th root of Jm (x) = 0: Since

d2
e
dz 2

kmn jz z 0 j
1
X

2
= kmn
e

kmn jz z 0 j

)] = 2

< a)

cos[m(

z 0 );

2kmn (z

(2.80)

);

(2.81)

m= 1

we readily nd the interior Greens function ( ;

1
1
1 X X Jm (kmn )Jm (kmn 0 )
cos[m(
G(r; r ) =
2
kmn Jm+1
(kmn a)
m= 1 n=1

)]e

kmn jz z 0 j

(2.82)

For exterior of a long cylinder, solutions to the equation

@2
1 @
1 @2
@2
+
+
+
2
@ 2
@
@ 2 @z 2

G=

0)

) (z

z 0 );

can be found in terms of Fourier transform with respect to the z-coordinate. Let G(r; r0 ) be
0

G(r; r ) =

im(

z0 )

Rm ( ; 0 ; k)eik(z

dk:

(2.83)

The radial function Rm ( ; 0 ; k) satises

d2
1 d
+
2
d
d

m2
2

k 2 Rm ( ; 0 ; k) =

0)

(
2

(2.84)

Elementary solutions are the modied Bessel functions Im (k ) and Km (k ) and we can construct

18

following solutions which remain bounded in the region a <

Rm ( ; ; k) =

< 1;

8
>
>
< A(k)Im (k ) + B(k)Km (k ); a <
>
>
: C(k)K (k );
m

a<

<
0

<

< 1;

(2.85)

< 1;

The boundary conditions are Rm ( = a) = 0 and Rm ( ) be continuous at

0;

A(k)Im (ka) + B(k)Km (ka) = 0;

(2.86)

A(k)Im (k 0 ) + B(k)Km (k 0 ) = C(k)Km (k 0 ):

(2.87)

Then,

Rm ( ; 0 ; k) =

8
>
>
A(k) Im (k )
>
>
<
>
>
>
>
: A(k)

1
Km (k 0 )

Im (ka)
Km (k ) ;
Km (ka)
Im (k 0 )

a<

Im (ka)
Km (k 0 ) Km (k ); a <
Km (ka)

<

<

< 1;
< 1:
(2.88)

The unknown function A(k) can be found from the discontinuity in the derivative at
d2
Rm ( ; 0 ; k)
d 2

= kA(k)
=

0 (k 0 )I (k 0 )
0 (k 0 )
Km
Km (k 0 )Im
m
(
Km (k 0 )

A(k)
(
0
Km (k )

0)
0

=
0

0;

(2.89)
(2.90)

where again use has been made of the Wronskian of the modied Bessel functions,
0
Im
(x)Km (x)

0
Im (x)Km
(x) =

We thus nd
A(k) =

1
:
x

(2.91)

Km (k 0 )
;
2

(2.92)

and Rm ( ; 0 ; k) reduces to

Rm ( ; 0 ; k) =

8
1
>
>
Km (k 0 ) Im (k )
>
>
< 2
>
>
>
1
>
:
2

Im (k 0 )

Im (ka)
Km (k ) ;
Km (ka)

a<

<

< 1;
(2.93)

Im (ka)
Km (k 0 ) Km (k ); a <
Km (ka)

19

<

< 1:

The exterior Greens function of a long cylinder is given by

1 X im(
G(r; r ) =
e
2 m

2.3.5

Rm ( ; 0 ; k)eik(z

z0 )

dk:

(2.94)

Long Cylinder (2-Dimensional)

Cross-section of a long cylinder.

are the line charge and its image, respectively, that together

make the cylinder surface an equipotential surface,

( = a) = =(2 "0 ) ln(a= 0 ):
For boundary value problems in which z-dependence is suppressed, it is convenient to formulate
a two dimensional Greens function. Two dimensional Greens function for a long cylinder is to be
found from
r2 G(r; r0 ) =
where

2 (r

r0 );

2 (r

(2.95)

r0 ) is the two-dimensional delta function. In the cylindrical geometry, it is given by

0)

r0 ) =

2 (r

);

(2.96)

and the Greens function satises

@2
1 @
1 @2
+
+
2
@ 2
@
@ 2

G=

0)

):

(2.97)

In this case, the method of image can be exploited very conveniently. Let us consider a long line
charge

(C/m) placed at ( 0 ;

negative line charge

) parallel to a long, grounded conducting cylinder of radius a: A

placed at ( 00 ;

) where
00

20

a2
0

(2.98)

makes the cylinder surface an equipotential surface at a potential

s

2 "0

ln

(2.99)

Since we are seeking a potential that vanishes on the cylinder surface

s

can be subtracted from the potential due to two line charges

(r; r0 ) =

2 "0

where

r0 =

r00

q
s

r0

ln r

and

0 cos(

);

002

00 cos(

a2

(2.100)

(2.101)

cos(

a2

r00 + ln

ln r

02

= a; the constant potential

):

(2.102)

The desired Greens function is

0

G(r; r ) =

1
ln
2

p
(

02

0 =a)2

+ a2

0 cos(

)
0

0 cos(

(2.103)

For exterior Dirichlet problems, the normal derivative at the cylinder surface is
2

@G
=
@n

@G
@ 0

=
0 =a+0

1
2

a
2a cos(

+ a2

> a;

(2.104)

and for interior,

2

@G
@G
=
@n
@ 0

0 =a

1
=
2

a
a
2a cos(

+ a2

< a:

If the potential on a long cylindrical surface is specied as a function of the angle ;

(2.105)
s(

); the

potential o the surface can be calculated from

( ; )=

21

s(

@G 0
d :
@n

(2.106)

For the interior ( < a) ; the potential is given by

( ; ) =
=

1
2
1
2

2
s

2
a2
0 d
a2 + 2 2a cos(
)
1
X
m
1+2
cos[m
a

] d 0;

m=1

(2.107)

where use is made of the following expansion,

a2 +

2
a2
2a cos(

=1+2

1
X

m=1

cos[m

]:

Example 4

Figure 2-7:
= V for 0 < < ;
(Example of 2-D Greens function.)

V for

<

< 0 on the surface of a long cylinder.

As an example, let us consider a long conducting cylinder consisting of two equal troughs. The
upper half in the region 0 <
potential

<

is at a potential V and the lower half

V as shown Fig.2-7. The exterior potential

2

( ; )=V

a2
2

+ a2

1
2a cos(

22

<

< 0 is at a

> a is given by
0

0
2

+ a2

1
2a cos(

(2.108)

The rst integral can be eected by changing the variable from

Z
=

to

through

=2 = ;

=2

1
d
+ + 2a sin(
)
=2
2
0
2
2
6
B ( + a ) tan
2
2
1
6tan B
@
2
a2 4
a2
2

a2
2
a2

a2

tan

tan

a2 )(cot

=2

+ 2a C7
C7
A5

tan ) + 2a
a2

2a sin
2
a2

13

=2

+ tan

+ a2 )(cot + tan ) + 2a
2
a2

(2.109)

where use has been made of the identities,

tan

tan

x
= cot x
2

x + tan

tan

x=

tan x;

y = tan

x+y
1 xy

tan

1
x

Similarly, the second integral yields

2
2

a2

tan

2a sin
2
a2

(2.110)

and the potential becomes

( ; )=

2V

tan

2a sin
2
a2

> a:

(2.111)

tan

2a sin
2
a2

< a:

(2.112)

The interior potential is

( ; )=

2.3.6

2V

Wedge

A wedge is formed by two large plates intersecting at an angle

The potential due to a point charge q at (
plates

= 0 and

= , and

0;

; z0)

as illustrated in Fig.2-8.

with the boundary conditions

= 0 at the

= 1; jzj = 1 essentially gives the Greens function. We thus seek

23

Figure 2-8: A wedge formed by two large conducting plates intersecting at an angle :
a solution to the Poissons equation
@2
1 @
1 @2
@2
+
+
+
2
@ 2
@
@ 2 @z 2

0)

G=

) (z

z 0 );

(2.113)

subject to the those boundary conditions. As in the case of 3-dimensional Greens function for a
long cylinder, we Fourier transform the Greens function,
G(r; ; z) =

1
2

g( ; ; k)eik(z

z0 )

dk:

(2.114)

The angular dependence of the Greens function can be assumed to be

m

sin
which indeed vanishes at

= 0 and

g( ; ; k) =

sin

(2.115)

= : We thus assume

Am Rm ( ) sin

sin

(2.116)

to obtain
"
X
d2
1 d
Am
+
2
d
d
m

1
2

Rm ( ) sin

sin

0)

):

(2.117)

24

The radial function can be composed of the modied Bessel functions,

8
>
>
< Im

Rm ( ) =

>
>
: I
m

(k )Km

(k 0 );

0;

<

(2.118)
0 )K

(k

m =

(k );

< :

The derivative of the radial function Rm ( ) has discontinuity at

0;

and the second order

derivative yields
d2 Rm ( )
=
d 2

);

(2.119)

where the Wronskian of the modied Bessel functions,

I (x)K 0 (x)

1
;
x

I 0 (x)K (x) =

has been substituted. The expansion coe cient Am is thus determined as

1

Am =

(2.120)

and the desired Greens function is

G(r; r0 ) =

Z
2 X
m

Rm ( ; 0 ) cos[k(z

z 0 )]dk sin(

) sin(

);

(2.121)

where
=

(2.122)

We will encounter an application of wedge potential in the section of inversion method later in this
Chapter.
Example 5 Line Charge parallel to a Long Dielectric Cylinder
Consider a long line charge with charge density

(C m

1)

placed parallel to a long dielectric

cylinder of radius a: The distance between the line charge and the cylinder axis is b: The potential
outside the cylinder can be sought by assuming an image line charge
another image
>a (

where

; )=

at the position a2 =b and

at the center,

4 "0

ln

+ b2

2 b cos

4 "0

ln

a2
b

a2
2
cos
b

4 "0

ln

is measured relative to the location of line charge : The interior potential is free from

25

00

singularity and may be assumed to be due to an image

<a (

00

; )=

4 "0

ln

+ b2

at the location of the line charge ;

2 b cos

4 "0

ln b2 :

Note that the constant potential (the last term)

0

ln b2 ;

4 "0

is needed to match same outer potential at r = a;

>a (r

+ 0
ln a2 + b2
4 "0

= a) =

2ab cos

4 "0

ln b2 :

The pertinent boundary conditions are E and Dr be continuous at r = a which yield

0

00

and
0

"0

="

00

Then
0

"0 "
; and
"0 + "

00

2"0
:
"0 + "

The attracting force to act on the unit length of the line charge is given by
F
=
l
2 "0

00

(a2 =b)

2 "0

a2
b (b2

"0
a2 ) "

"
; (N m
0+"

):

A magnetically dual problem is the case of long current I parallel to a magnetic cylinder with
a permeability : The image currents are
I0 =

I at

a2
;
b

and
I 00 =

2.4

2
+

I at the axis.
0

Other Useful Rectilinear Coordinates

The familiar three coordinate systems, cartesian, spherical, and cylindrical, are frequently used in
analyzing potential problems. However, there are some 30 known coordinate systems developed for
specic problems. For simple electrode shapes, potential problems can be rendered one dimensional
26

by a suitable choice of coordinates. However, in some coordinates, solutions to Laplace equations

are not always completely separable. We have encountered one such example in Chapter 1, the
toroidal coordinates, in analyzing the potential due to a ring charge. In this section, some coordinate
systems useful for potential problems will be introduced.

2.4.1

Oblate Spheroidal Coordinates ( ; ; )

Figure 2-9: Oblate spheroidal coordinates ( ; ; ):

= cons. describes the surface of a hyperboloid.

! 0 degenerates to a thin disk of radius a:

The oblate spherical coordinates ( ; ; ) are related to the cartesian coordinates through the
following transformation,

A surface of constant

8
>
>
< x = a cosh sin cos

(2.123)

y = a cosh sin sin

>
>
: z = a sinh cos

is the surface of an oblate spheroid described by

x2 + y 2
z2
+
= 1;
(a cosh )2 (a sinh )2

as shown in Fig.2-9. In the limit of

(2.124)

! 0; the surface degenerates to a thin disk of radius a

with negligible thickness, and in the opposite limit

27

1; the surface approaches a sphere with

a radius r = a cosh

' a sinh : This coordinate system is convenient if electrode shapes are an

oblate sphere or disk. A surface of constant

is a hyperboloid described by
z2
= 1:
(a sin )2

x2 + y 2
(a sin )2

(2.125)

The metric coe cients are

h
h
h

@x
@

@x
@

@x
@

@y
@
@y
@

@y
@

@z
@
@z
@

@z
@

q
= a cosh2

sin2 ;

(2.126)

=h ;

(2.127)

= a cosh sin :

(2.128)

The Laplace equation in the oblate spherical coordinates can thus be written down as
1
h h h

@
@

h h @
h @

@
@

h h @
h @

h h @
h @

@
@

= 0;

(2.129)

@2
= 0:
@ 2

(2.130)

which reduces to
1
cosh

sin

@2
@
@2
@
+ tanh
+ 2 + cot
2
@
@
@
@

Assuming a separated solution

1
cosh

sin

( ; ; ) = F1 ( )F2 ( )eim ; (m = integer), we obtain

d2
d
+ tanh
2
d
d

l(l + 1) +

d2
d
+ cot
+ l(l + 1)
d
d 2

m2
cosh2
m2
sin2

F1 ( ) = 0;

F2 ( ) = 0;

(2.131)

(2.132)

where l(l + 1) is a separation constant. Eq. (2.132) is the standard form of the Legendre equation
and solutions for F2 ( ) are
F2 ( ) = Plm (cos ); Qm
l (cos ):

(2.133)

Eq. (2.131) can be rewritten as

d2
sinh d
+
2
d
cosh d

l(l + 1) +

28

m2
1 + sinh2

F1 ( ) = 0;

(2.134)

which is also the Legendre equation with a variable i sinh : Therefore, solutions for F1 ( ) are
F1 ( ) = Plm (i sinh ); Qm
l (i sinh );

(2.135)

and general solution to Laplace equation can be constructed from these elementary solutions.
If a point charge q is placed at

0; 0;

; the potential in terms of the oblate spheroidal

coordinates can be found as

q
1
4 "0 jr r0 j

(r) =

1
l
q X X (l jmj)!
"0 a
(l + jmj)!

l=0 m= l

0
Plm (i sinh ) Qm
l (i sinh )

Plm (i sinh 0 ) Qm
l (i sinh )

Ylm ( ; ) Ylm

)
< 0
(2.136):
> 0

Derivation of this expression is left for exercise. The Wronskian of the Legendre functions,
Plm (x)

d m
Q (x)
dx l

Qm
l (x)

d m
1 (l + jmj)!
Pl (x) = 2
;
dx
x
1 (l jmj)!

(2.137)

should be useful. Furthermore, the Greens function for an oblate spheroidal surface described by
=
for

0
0

can readily be worked out to be:

<

<

G r; r0

0;
1
l
1 X X (l jmj)! m
i sinh
P (i sinh ) Qm
l
a
(l + jmj)! l

Ylm ( ; ) Ylm

l=0 m= l

1
l
1 X X (l jmj)! Plm (i sinh
a
(l + jmj)! Qm
l (i sinh
l=0 m= l

and for
G r; r0

<

0)
0)

m
i sinh
Qm
l (i sinh ) Ql

0 0
;
;
Ylm ( ; ) Ylm (2.138)

< ;

1
l
1 X X (l jmj)! m
P i sinh
a
(l + jmj)! l

Qm
l (i sinh ) Ylm ( ; ) Ylm

l=0 m= l

1
l
1 X X (l jmj)! Plm (i sinh
a
(l + jmj)! Qm
l (i sinh
l=0 m= l

0)
0)

m
Qm
i sinh
l (i sinh ) Ql

0 0
Ylm ( ; ) Ylm (2.139)
;
:

Example 6 Charged Conducting Disk

A thin disk of radius a is described by

= 0 in the oblate spherical coordinates. If a constant

surface is an equipotential surface, the potential o the surface is a function of

only, that is,

the potential problem becomes one dimensional. This is the most advantageous merit of using a
coordinate system most suitable for particular potential problems. The relevant solution which
29

Figure 2-10: A charged conducting disk of radius a. A disk is described by

vanishes at

= 0; 0

= 1 is the lowest order Legendre function of the second kind,

( ) = AQ0 (i sinh ) + B;

(2.140)

where A and B are constants. Since

h
Q0 (i sinh ) = i tan

(sinh )

i cot

(sinh );

(2.141)

and the boundary condition is

( = 0) = V (disk potential),
we readily nd the potential at an arbitrary ;
( )=
Note that cot

1 (0)

2V

cot

(sinh ):

=2: The far eld potential at

asymptotic form of the function cot

cot

(2.142)

1 or r

a can be found from the

1 x;
1

x'

1
x

1
+
3x3

; x

1:

(2.143)

The leading far eld potential is monopole as expected,

(

1) '

2V

1
sinh

30

'

2V a
:
r

(2.144)

Comparing with the standard monopole potential

(r) =

1 q
;
4 "0 r

(2.145)

we readily nd the total charge carried by the disk,

q = 8"0 aV;
and the self-capacitance of the disk,
C = 8"0 a; (F).

(2.146)

This expression was rst found by Cavendish.

The surface charge distribution on the disk is quite nonuniform because like charges repel each
other. Charge is distributed in such a manner that the tangential electric eld on the disk surface
vanishes. The surface charge density can be found from the normal component of the electric eld,
= "0 En = "0 E ;

(2.147)

where
E

1 @
h @

=0

2V
1
:
a j cos j

(2.148)

Note that
d
cot
dx

x=

1
:
1 + x2

(2.149)

The surface charge density diverges at the edge of the disk where

= =2: The charge residing on

the disk surface can be found from the following surface integral,
q = "0

2"0 aV

d
Z

sin d

= 8"0 aV:

h h
Z

=0
2

This is consistent with the charge found earlier using the monopole potential.
If one uses a coordinate system other than the oblate spheroidal system, solutions will be much
more involved. Let us employ the cylindrical coordinates ( ; ; z): Because of axial symmetry,

31

dependence can be suppressed and we seek a solution in the form of Laplace transform,
( ; z) =

( ; k)e

kjzj

dk:

(2.150)

The Laplace equation without

dependence
@2
1 @
@2
+
+
@ 2
@
@z 2

becomes

( ; z) = 0;

d2
1 d
+
+ k2
2
d
d

(2.151)

( ; k) = 0;

(2.152)

which suggests that

( ; k) = A(k)J0 (k ):

(2.153)

The boundary conditions are:

(

a; z =

0) = V (constant):

The following integral has a peculiar property,

Z

8
>
>
< 2;

sin ax
J0 (bx)dx =
>
x
>
:

sin

if a > b;
(2.154)
1

(a=b); if a < b:

Exploiting this property, we can construct the following solution for the potential,
( ; z) =

2V

sin ka
J0 (k )e
k

kjzj

dk:

(2.155)

The potential in the disk plane (z = 0) is

( ; z = 0) =

8
>
V;
>
>
<

if

>
>
2V
>
:
sin

< a;
(2.156)

(a= ); if

> a:

Example 7 Dipole Moment of a Conducting Disk in an External Electric Field

If a thin conductor disk is placed perpendicular to an external eld, the dipole moment is zero
because of negligible thickness of the disk even though charge separation does take place in such
a manner that disk surfaces are oppositely charged. The external electric eld is little disturbed
by the disk in this case. The maximum disturbance occurs when the disk surface is parallel to the
32

Figure 2-11: Conducting disk in an external electric eld parallel to the disk surface.
eld.
We assume a uniform external electric eld in the x direction and a thin conducting disk placed
in the x

y plane with its axis in the z direction as shown in Fig.2-11. The potential associated

with the external uniform electric eld is

0

The radial function cosh

E0 x

E0 a cosh sin cos :

(2.157)

is actually P11 (i sinh ) and the presence of the disk should yield

a perturbation proportional to the Legendre function of the second kind Q11 (i sinh ) since the
perturbed potential should have the same angular dependence as

0(

; ; ) to satisfy the boundary

condition at the disk: Thus we assume

( ; ; )=

E0 a cosh sin cos + AQ11 (i sinh ) sin cos ;

(2.158)

where Q11 (i sinh ) is actually a real function,

Q11 (i sinh ) = cosh

cot

(sinh )

sinh
cosh2

(2.159)

The constant A can be determined from the boundary condition that the disk potential be zero,
that is,

( = 0) = 0: We thus nd
A=

aE0 ;

33

and the potential becomes

( ; ; )=
Far away from the disk at r

a or

lim ( ; ; ) !
1

aE0 cosh

Q11 (i sinh ) sin cos :

(2.160)

1; the potential approaches

aE0 cosh sin cos +

4E0 a3 sin cos

;
3
r2

(2.161)

where the asymptotic form of Q11 (i sinh );

Q11 (i sinh ) '

2 1
3 sinh2

2 a
3 r

(2.162)

has been substituted. Comparing the dipole term in Eq. (2.161) with the standard dipole potential
dipole

1 p r
;
4 "0 r3

(2.163)

we can readily identify the dipole moment induced by the disk,

p = 4 "0

4a3
E0k ;
3

(2.164)

where E0k is the component of the external electric eld tangential to the disk surface. Note that the
dipole moment is proportional to a3 . The moment is equally applicable for low frequency oscillating
electric eld as long as the wavelength associated with the oscillating led is much longer than the
disk radius, ka =

1: A resultant scattering cross-section of a conducting disk (sphere too)

placed in a low frequency electromagnetic wave is proportional to a6 .

Example 8 Leakage of Electric Field through a Small Hole in a Conducting Plate
Consider a parallel plate capacitor whose grounded, lower plate has a small circular hole of
radius a as shown in Fig.2-12. We wish to nd how the hole perturbs the potential. This problem
has important applications in analyzing leakage of microwaves through a small hole in waveguide
walls.
The unperturbed electric eld E0 between the plates is assumed downward with a corresponding
potential
0 (z)

8
>
>
< E0 z; z > 0
>
>
: 0;

34

z<0

(2.165)

Figure 2-12: The lower plate of a parallel plate capacitor has a small hole of radius a: The electric
eld leaks throught the hole.
where z = a sinh cos : We note
sinh =

iP1 (i sinh ):

(2.166)

Therefore, the perturbed potential can be sought in term of the Legendre function of the second
kind Q1 (i sinh ) which is equivalent to
1

Q1 (i sinh ) = sinh cot

(sinh )

1:

(2.167)

We thus assume the following form for the potential in both regions,

( ; )=

8
>
aE0 sinh cos + A sinh cot
>
<
>
>
:

A sinh cot

1 (sinh

1 (sinh

1 cos ;

1 cos ; 0 <

<

<

< ;

which ensures continuity of the potential at the hole ( = 0): Continuity of the normal component
of the electric eld at the hole requires
@
@

=0
z=+0

@
@

=0
z= 0

from which we readily nd the constant A;

A=

aE0

35

In the region below the lower plate (z < 0); the potential is
( ; )=

aE0

sinh cot

(sinh )

1 cos ;

<

< :

(2.168)

Its asymptotic form is of dipole nature,

(r

E0 a3 1
cos > 0;
3 r2

a) !

(2.169)

(note that cos < 0 in the region below the plate) and the eective dipole moment of the hole is
p=

4 "0 a3
E0 ;
3

(2.170)

which is downward. The far-eld potential in the upper region (z > 0) is

'

E0 a3 1
cos ;
3 r2

(2.171)

in which the dipole term is due to an eective dipole moment upward. The potential at the center
of the hole is
( = 0; = 0 or ) =

aE0

(2.172)

The results of this example, together with those of Example 14 in Chapter 3 (leakage of magnetic
eld through a hole in a superconducting plate), will have important implications on diraction of
electromagnetic waves by an aperture in a conducting plate. Since the eective dipoles are opposite
to each other in the two regions z > 0 and z < 0; it follows in general that
Ez ( z) =

Ez (z);

that is, the electric eld normal to the plate is an odd function of z: This means that the surface
charges

= "0 n E (C/m2 ) induced on both sides of the plate at z = +0 and z =

0 are identical,

where n is the unit normal vector at the plate surface. (Note that n changes its sign from one side
to other.) The component tangential to the plate,
Et = n

E;

is an even function of z;
Et ( z) = Et (z):
Of course, on the surface of the conducting plate, Et vanishes but it does not in the hole. For

36

magnetic elds resulting from a hole in an ideally conducting plate, we will see that the normal
component should vanish at the plate surface
Hz = 0; at z =

0;

and o the plate, it is even with respect to z;

Hz ( z) = Hz (z);
while the tangential component Ht = n

H is an odd function of z;
Ht ( z) =

Ht (z):

It follows that the surface currents

Js = n

H; (A/m)

on both surfaces of the plate are identical.

2.5

Method of Inversion

The method of inversion is useful when an electrode has a spherical shape, either complete spheres
(e.g., two spheres touching) or incomplete sphere (e.g., spherical bowl, solid hemisphere, etc.). For
a given sphere of radius a which we call inverting sphere; the inverted position of a point at r is
dened by
ri =

a2
r:
r2

(2.173)

(See Fig.2-13.) A sphere is inverted into another sphere. If the center of the inverting sphere is
chosen on the surface of a sphere to be inverted, the inverted surface becomes a plane as shown
in Fig.2-14. This is where the merit of method of inversion is found because potential problems of
planar electrodes are often simpler than those involving spheres.
Consider a charge q placed at r0 = (r0 ; 0 ;
=
where
cos

): The potential at position r = (r; ; ) is

q
1
p
;
4 "0 r2 + r02 2rr0 cos

= cos cos

+ sin sin

37

cos(

(2.174)

):

Figure 2-13: Point P at (r; ; ) is inverted with respect to the sphere of radius a to Q at (a2 =r; ; );
i.e., at the image position.

Figure 2-14: If an inverting sphere is centered on a surface of a sphere to be inverted, the sphere is
inverted to an innite plane.
In the inverted space with respect to a sphere of radius a; a charge q 0 will appear at
a
r0

a
r

r0 ;

(2.175)

r:

(2.176)

and the position r is inverted to

The potential at the inverted position is

i

q0
r
4 "0 a4
r2

q0

rr0

4 "0

a2

1
+

a4
r02

p
r2 + r02

38

a4
cos
rr0

1
2rr0 cos

(2.177)

In general, if a function

(r; ; ) satises the Laplace equation, the potential function

a2
; ;
r

a
r

(2.178)

also satises the Laplace equation.

It should be noted that an equipotential spherical surface is in general not inverted to an
equipotential sphere. However, a spherical surface at zero potential is inverted to a zero potential
spherical surface. Since the reference potential can be chosen arbitrarily without aecting the
electric eld, one can always choose the potential of an equipotential spherical surface at zero
potential. For example, the potential of a charged conducting sphere of radius a is
s

1 q
;
4 "0 a

relative to zero potential at innity. However, we can subtract

(2.179)
s

from the potential everywhere

and choose the sphere potential at zero and the potential at innity as
1

1 q
:
4 "0 a

The electric eld remains unchanged through uniform shift of the potential. If an inverting sphere
is chosen in such a way that it has a radius 2a centered at the surface of the conducting sphere of
radius a; the conducting sphere is inverted to an innite plane touching the both spheres as shown
in Fig. 2-15. Since the sphere potential is chosen at zero, the potential of the plane is also zero.
The potential at innity is inverted to
1 q
4 "0 a

2a
=
r

1 2q
;
4 "0 r

(2.180)

where r is the radial distance from the center of the inverting sphere with radius 2a: This is a
potential due to a point charge

2q: Therefore, a charge

2q =

8 "0

s a;

(2.181)

appears at the center of the inverting sphere.

Example 9 Capacitance of Touching Spheres
Using the method of inversion, we can nd the capacitance of two conducting sphere touching
each other as shown in Fig.2-15. The potential of the touching spheres is denoted by

s:

If the

inverting sphere has radius 2a and its center at the touching point, the two spheres become two

39

Figure 2-15: Touching spheres are inverted to parallel plates by a sphere of radius 2a centered at
the touching point. Images appear in the inverted space.
parallel planes separated by a distance 4a: A charge
q=

8 "0 a

s;

(2.182)

appears at the midpoint between the plates after inversion which can be analyzed easily using the
method of multiple images. The following image charges appear: q at jzj = 4a;
at jzj = 12a;

q at jzj = 8a; q

: The amount of total charge on the surface of the original spheres can be found by

re-inverting the image charges,

2a
4a
= q ln 2

Q = 2q

= 8 "0 a

2a
2a
+
8a 12a
s ln 2:

(2.183)

Therefore, the self-capacitance of the touching spheres is

C=

= 8 "0 a ln 2:

(2.184)

The potential

i(

; z) in the inverted space shown in Fig.2-16can be found in the form of Fourier

transform,
i ( ; z) =

A(k) sinh[k(2a

jzj)]J0 (k )dk;

(2.185)

where A(k) is a weighting function to be determined. It is noted that the elementary solution to
the Laplace equation is
J0 (k )e

40

kz

(2.186)

Figure 2-16: Geometry in the inverted space.

and the assumed form of the potential certainly satises the Laplace equation as well. The weighting
function A(k) can be determined by noting
d2
sinh[k(2a
dz 2
and

jzj)] =

2k cosh(2ak) (z);

kJ0 (k )dk =

(2.187)

( ):

(2.188)

The charge density of the point charge q at the origin is

q

( ) (z):

(2.189)

Then, A(k) can be determined from the Poissons equation

r2

"0

(2.190)

as
1
q
;
4 "0 cosh(2ak)

A(k) =

(2.191)

and the potential in the inverted space is

i(

; z) =

q
4 "0

sinh[k(2a jzj)]
J0 (k )dk:
cosh(2ak)

The potential in the original conguration can be found by reinverting

tion
z!

2a
r

z;

where
r2 =

41

+ z2;

2a
r

(2.192)
through the transforma-

is the distance from the center. The result is

"
q 2a
4 "0 r

( ; z) =

with r2 =

2 + z2:

2a
r

sinh k 2a

jzj

!#

cosh(2ak)

Recalling that we have subtracted

"

J0 k

2a
r

dk;

(2.193)

= q=4 "0 a (the sphere potential) from the

potential everywhere to make the sphere potential vanish, we nally obtain

2

6
6
( ; z) = s 6
61
4

(2a)2
r

"

2a
r

sinh k 2a

cosh(2ak)

jzj

!#

"

J0 k

2a
r

7
7
dk 7
7:
5

(2.194)

Figure 2-17: Geometry in the original space.

Example 10 Capacitance of Spherical Bowl

As a second example, we consider a hollow spherical bowl of radius a with an angle 2 subtended
at the center shown in Fig.2-18. As inverting sphere, one can choose a sphere having a radius 2a sin
centered at the edge of the bowl. After inversion, the bowl becomes a semi-innite plane as shown
and a charge q =

8 "0 a sin

will appear at the center of the inverting sphere. Potential

problems involving a semi-innite conducting plate can be analyzed as a limiting case of a wedge.
For a charge q placed at ( 0 ;

; z 0 ) near a wedge intersecting at an angle

42

, the potential is given

Figure 2-18: A bowl (radius a; center angle 2 ) is inverted to a semi-innite plane by a sphere of
radius 2a sin centered at the edge of the bowl.
by
8 PR1
0
>
>
0 I (k )K (k ) cos[k(z
>
m
<
2q
( ; ; z) =
"0 >
PR1
>
>
0
:
0 I (k )K (k ) cos[k(z

z 0 )]dk sin(

);

<

(2.195)
z 0 )]dk sin(

where

) sin(

) sin(

);

>

= m = : Noting
Z

I (k )K (k ) cos[k(z

1
z )]dk = p
2 2
0

where

cosh =
and the sum formula

1
X

m=1

pm cos(mx) =

1
2

02

+ (z
2 0

43

z 0 )2

e
cosh

cosh

1 p2
2p cos x + p2

d ;

(2.196)

(2.197)

1 ;

(2.198)

we see that the potential reduces to

1
q
p
4 "0
2

(r) =

cosh(
p
For a plate

sinh
1
cos[ (

= )
1

cosh

cosh

)= ]

cosh(

1
cos[ ( +

= )

)= ]

d :

(2.199)

= 2 ; and this becomes

(r) =

q
4

2"

1
cos
R

0
cos[(
)=2]
cosh( =2)

1
cos
R0

cos[( + 0 )=2]
cosh( =2)

(2.200)

where
R =
R0 =

02

0 cos(

02

0 cos(

);

):

The potential in the vicinity of the charge q can be found by letting

r=2a sin

r;

; =

1;
(r) =

q 1
4 "0 r

q
1
4 "0 4 a sin

1+

sin

(2.201)

where r is the distance from the charge. The correction due to the presence of the conducting plate
is therefore
=

q
1
4 "0 4 a sin

1+

sin

(2.202)

and in the physical space, the far eld potential due to a charged conducting bowl is in the form
(r

a) =
=

q
1
4 "0 2 r

1a
(sin + )
r

1+
s:

sin
(2.203)

Comparing with the standard monopole potential

=

Q
;
4 "0 r

we nally nd the capacitance of the bowl,

C = 4"0 a( + sin ):

44

(2.204)

For a sphere,

= ; we recover C = 4 "0 a: For a disk of radius R;

also recover

' sin ' R=a

1; and we

C = 8"0 R:
The capacitance of a solid (or closed) hemisphere can be found in a similar manner and given
by
1
p
3

C = 8 "0 a 1

(2.205)

This is left for an exercise.

2.6

Numerical Methods

Analytic solutions in potential problems can only be found for a limited number of applications, and
in practice, it is often necessary to resort to numerical analysis. In this section, we will estimate the
capacitance of a square conductor plate of side a: Mathematically speaking, this problem constitutes
an integral equation for the potential
1
=
4 "0

; which is constant at the conductor,

(r0 )
dS 0 =
jr r0 j

1
4

@
r0 j @n0

jr

dS 0 = V = constant,

(2.206)

where
= "0 En =

"0

@
;
@n

is the unknown surface charge density. The capacitance can be found from
C=

dS:

(2.207)

As a very rough estimate, we recall that the capacitance of a circular disk of radius a is given
by
C = 8"0 a;

(2.208)

and approximate the capacitance of the square plate by

C = 8"0 re = 8"0

0:564a;

(2.209)

where re is the radius of a circular disk having the same area as the plate,
2
re
= a2 ; re = 0:564a:

The nite element numerical method given below yields C ' 8"0
45

0:547a.

Figure 2-19: A square conducting plate of side a is divided into 25 sub-areas. Because of symmetry,
the number of unknown potentials is reduced to 6.
The capacitance is the ratio between the total charge Q and the plate potential V , C = Q=V:
We divide the plate into n

n sub-areas of equal size each with side a=n. Each sub-plate is at an

equal potential V but charges on the sub-plates dier. To illustrate the procedure, we choose n = 5
(25 sub-areas) as shown in Fig. 2-19. Because of symmetry, there are 6 unknown charges to be
found. The potential on each sub-plate can be calculated by summing contributions from charges
on all sub-plates including the charge on itself. The self-potential of one unit can be estimated as
follows. Consider a square plate of side

carrying a uniform surface charge density

(C/m2 ): The

potential at the center of the plate can be found from

=
=
=
=
2

where q =

4
4
4
4

=2

=2

1
dy p
2
"0
x + y2
=2
=2
Z =2 h
p
4
ln
x2 + 2 +
"0 0
p
4 ln 1 + 2
"0
p
q
q
4 ln 1 + 2 =
"0
4 "0
dx

i
ln x dx
3:5255;

(2.210)

is the charge carried by the sub-plate. With this preparation, we can write down

the potential of sub-plate A as follows:

4 "0

2
2
2
+p +
qB
3
17 5
2
1
2
1
2
2
1
+ 1+ p
qC + p + p + p
qD + p + p
qE + p qF
20
2
10 3 2
5
13
2 2
= 4:2023qA + 3:5517qB + 1:4472qC + 1:5753qD + 1:4491qE + 0:3536qF :
(2.211)

3:5255 +

2
1
+ p
4 4 2

qA + 2 +

46

Other potentials

F,

which are all equal, can be written down in a similar way and we obtain

6 simultaneous equations for qA

qF which can be solved easily. A resultant total charge is

Q = 1:743

4 "0

and the capacitance is

C ' 0:3486

4 "0 a

= 0:547

8"0 a:

(2.212)

Accuracy will improve if a larger number of sub-areas are used.

The method can be applied to estimate the capacitance of a conducting cube as well. With 150
sub-areas (25 sub-areas on each side), the following capacitance emerges,
C ' 0:65

4 "0 a;

(2.213)

where a is the side of the cube. An estimate based on a sphere having the same surface area gives
C = 4 "0 re = 0:69
where
re =

4 "0 a;

(2.214)

6
a = 0:69a:
4

A well known nite element method of solving the Laplace equation is based on the fact that the
potential at the center of a cube may be approximated by the average of 6 surrounding potentials
on each face of the cube,
6

1X
6

i:

i=1

This follows from the Taylor expansion of the potential,

(x

; y; z) =

(x; y; z)

; z) =

(x; y; z)

) =

(x; y; z)

(x; y
(x; y; z

@
1 @2
+
@x
2 @x2
1 @2
@
+
@y
2 @y 2
@
1 @2
+
@z
2 @z 2

;
;
;

Adding these 6 equations, we nd

(x

; y; z) + (x; y

; z) + (x; y; z
47

) = 6 (x; y; z) + r2 + O( 4 ):

(2.215)

satises the Laplace equation, r2

Therefore, if

= 0;
6

center

1X
'
6

i;

(2.216)

i=1

valid to order

For 2-dimensional problems in which z-dependence is suppressed, we have

4

center

'

1X
4

i:

(2.217)

i=1

The equation can be applied to each sub-unit having a volume

(3-D) or area

(2-D). Resultant

simultaneous equations can be solved numerically.

Example 11 Potential in a Long Cylinder
Consider a conducting cylinder having a cross-section as shown in Fig. ??. The periodic upper
electrode is at a potential V and the at lower electrode is grounded. In order to apply the nite
element method, we divide the cross section into sub-sections and allocate 10 nodes points as shown.
Applying Eq. (2.217) to the potentials

= 88:2;

1
9

= 63:7;

= 55:7;

2
10

=1

10); we obtain

= 100 +

4;

= 100 +

= 100 +

= 300 +

8;

= 200 +

9;

= 2

10 ;

10

= 2

9:

= 61:8;

4
Solutions are:

i (i

= 31:0;

+2

+2

1
3

5;

9;

6;

3;

10 ;

= 30:2;

= 53:5;

= 27:9;

= 97:1;

= 27:9 all in Volts. A larger number of node points will improve

accuracy.

48

Cross-section of long cylinder with periodic anode structure.

49

Problems
2.1 A ring charge of total charge q and radius b is coaxial with a long grounded conducting
cylinder of radius a (< b): Determine the potential everywhere.
2.2 A ring charge of total charge q and radius b is coaxial with a long uncharged dielectric cylinder
of permittivity " and radius a: Determine the potential everywhere.
2.3 A charge q is placed at an axial distance b from a conducting disk of radius a: Determine
the potential everywhere. Consider two cases, (a) the disk is grounded, and (b) the disk is
oating.
2.4 Show that a charge q at a distance d from the center of a oating conducting spherical shell
of radius a raises the sphere potential to
s

q
; if d > a (q outside the sphere);
4 "0 d

or
s

q
; if d < a (q inside):
4 "0 a

2.5 A large grounded conducting plate has a hemispherical bob of radius a: A charge q is placed
at an axial distance d from the center of the bob. Find the force on the charge.
2.6 Show that the capacitance per unit length of a parallel wire transmission line with a common
wire radius a and separation distance d is
C
=
l
ln

"0
p
d + d2
2a

4a2

!:

2.7 A coaxial cable having inner and outer radii a and b is bent to form a thin toroidal capacitor
with a major radius R (

a; b): Find the capacitance.

2.8 Show that the mutual capacitance between conducting spheres of radii a and b separated by
a large distance d

a; b is approximately given by
Cab ' 4 "0

ab
;
d

and that the capacitance of the sphere of radius a is aected by the sphere of radius b as
Caa ' 4 "0 a 1 +
50

ab
d2

2.9 Rigorous analysis of potential problems involving two conducting spheres can be made by
using the bispherical coordinates dened by
a sin cos
;
cosh
cos
a sin sin
;
cosh
cos
a sinh
:
cosh
cos

x =
y =
z =

= constant surface is a sphere described by

2

x + y + (z
and

a
sinh

acotanh ) =

= constant surface is
(

acotan )2 + z 2 =

a
sin

which is spindle-like shape.

(a) Finding the metric coe cients h ; h

and h ; show that the Laplace equation in the

bispherical coordinates is
@
@

1
cosh

@
cos @

1 @
sin @

sin
cosh

@
cos @

sin2

1
(cosh

@2
= 0:
cos ) @ 2

(b) Show that the general solution to the Laplace equation is in the form
( ; ; )=

p
cosh

cos

X
1
(Alm e(l+ 2 ) + Blm e

(l+ 12 )

)Plm (cosh )eim :

l;m

As in the oblate spheroidal coordinates, in this coordinate system too, the Laplace
equation is not separable.
2.10 Using the inversion method, show that the capacitance of a solid or closed conducting hemisphere of radius a is given by
C = 8 "0 a 1

1
p
3

2.11 Find a 2D Greens function for the interior of long cylinder having a semicircular cross-section
of radius a:

51

Hint: Assume
G(r; r0 ) =

( P

m Am (

m [Bm (

where r = ( ; ); r0 = ( 0 ;

= 0 )m sin(m ) sin(m 0 );

<
0

= 0 )m + Cm ( 0 = )m ] sin(m ) sin(m );

<

< a;
< a;

):

2.12 A conducting disk of radius a is placed parallel to an external electric eld E0 : The dominant
perturbation to the potential is dipole as shown in Example 6. What is the leading higher
order correction?
2.13 Cylindrical capacitors have cross-sections as shown. Estimate graphically the capacitance per
unit length for each. For (a), analytic expression for the capacitance is
C
=
l

where

= 4a;

cosh

2 "0
2+
1
2

2
2

d2

1 2

= 2a; and d = a: For (b), one has to resort to numerical analysis for an

exact value.

2.14 Find numerically the capacitance of a conducting cube of side a: What do you estimate for
the lower and upper bounds of the capacitance?
2.15 Derive Eq. (??), the expression for the potential due to a point charge in the oblate spheroidal
coordinates ( ; ; ) : The charge is at

0; 0;

2.16 Derive Eqs. (2.138) and (2.139), the Greens function for an oblate spheroid described by
=

(const.)

2.17 The prolate spheroidal coordinates ( ; ; ) is convenient to solve potential problems involving
prolate spheroids (sphere elongated along the z axis). The coordinate transformation is
52

dened by
x = a sinh sin cos ;
y = a sinh sin sin ;
z = a cosh cos :
In the limit of
limit

! 0;

= const. surface describes a thin rod having a length 2a; and in the

! 1; it approaches a sphere with radius a cosh

coe cients are:

' a sinh . Show that the metric

q
sinh2 + cos2 = h ;

= a

= a sinh cos :

Then, show that general solution of Laplaces equation r2

= 0 in the prolate spheroidal

coordinates is in the form

( ; ; )=

im
m
m
:
[Alm Plm (cosh ) + Blm Qm
l (cosh )] [Clm Pl (cos ) + Dlm Ql (cos )] e

l;m

In the lowest order l = 0; possible one dimensional solutions are

( ) = Q0 (cosh ) = ln coth

( ) = Q0 (cos ) = ln cot

53

Chapter 3

Magnetostatics
3.1

Introduction

Magnetostatics is a branch of electromagnetic studies involving magnetic elds produced by steady

non-time varying currents. Evidently currents are produced by moving charges undergoing translational motion. An eective current (called magnetization current) is also produced if magnetic
dipoles are nonuniformly distributed. To realize steady currents, a large number of charges must
be involved so that collection of charges can be regarded as a continuous uid. In non-time varying
cases (@=@t = 0), the charge conservation principle requires that
@
+ r J = r J =0; in steady state.
@t
This implies that a steady current is transverse. A static magnetic eld can also be produced by a
nonuniform magnetization (magnetic dipole density in the case of a collection of magnetic dipoles)
M (A/m) which produces a magnetization current
JM = r

M; (A m

):

The magnetization current is transverse (divergence-free), since r JM = r (r M) 0:

The magnetic eld B exerts a force perpendicular to the velocity of charged particles. The force
to act on a charge q is
F = q (E + v B) :
The magnetic eld does not do any work on charged particles since the rate of work v F is
independent of B;
v F = qv E:
(Note that v (v

B) =0:) In contrast to the electric eld which is a true (or polar) vector, the

magnetic eld is a pseudo (or axial) vector, or more precisely, constitutes a pseudotensor,
2

0
6
ij
B = 4 Bz
By

Bz
0
Bx

3
By
7
Bx 5 :
0

The magnetic eld due to a prescribed current distribution can be calculated using the BiotSavarts law. This law is generic because all other laws in magnetostatics, such as the Amperes law
and vanishing divergence of the magnetic eld r B = 0; follow from it. Calculation of the magnetic
eld is facilitated by introducing a magnetic vector potential A; which is related to the magnetic
eld through B = r A: This expression is consistent with r B = 0; since r (r A)
0
identically. It also follows that r A = 0 in static cases.
One important objective of magnetostatics is to derive formulae for the self-inductance and
mutual inductance for given current congurations. Use of the vector potential greatly facilitates
calculation of the magnetic ux,
Z
Z
I
= B dS = r A dS = A dl;
and inductance,
L=

3.2

Biot-Savarts Law and Vector Potential

Figure 3-1: Geometry to illustrate Biot-Savarts law.

For a given, non-time-varying current density distribution J(r); the magnetic eld B(r) can be
calculated from the Biot-Savarts law,
B(r) =

J(r0 )
jr

(r r0 ) 0
dV ; (T)
r0 j3

(3.1)

where
0

= 4

10

(T m A

= 4

10

(H m

);

is the vacuum magnetic permeability. As explained in Chapter 1,

introduced to dene 1 Ampere current. Noting
r
jr

r0
=
r0 j3

jr

r0 j

(3.2)
0

is an assigned constant

(3.3)

we may rewrite the law in the form

B(r) =

J(r0 )
dV:
jr r0 j

(3.4)

Then, one of the Maxwells equations

r B = 0;

(3.5)

immediately follows because of the identity for an arbitrary vector eld A;

r (r

A)

0:

Physical meaning of r B =0 is that there be no magnetic charges in contrast to the electric eld
r E = ="0 ; and that the magnetic eld is transverse without longitudinal component. Eq. (3.4)
implies that the magnetic eld is a curl of a vector eld dened by
A(r) =

J(r0 )
dV 0 :
jr r0 j

(3.6)

This vector quantity is called the magnetic vector potential. In terms of the vector potential, the
magnetic eld is thus given by
B(r) =r A:
(3.7)
The magnetic eld given in Eq. (3.1) is a pseudo vector because it is invariant against coordinates inversion, r ! r: In contrast, the Coulomb electric eld,
1
E (r) =
4 "0

(r

r0 ) (r0 ) 0
dV ;
jr r0 j3

is a true (or polar) vector, because it changes sign on coordinate inversion.

As will be shown in Chapter 4, for time varying current density, the vector potential is to be
modied by taking into account the retardation due to nite propagation time of electromagnetic
disturbances,
Z
J(r0 ; t
) 0
A(r;t) = 0
dV ;
4
jr r0 j
where
=

jr

r0 j

B(r; t) =r A and r B = 0 continue to hold in general.

The curl of the magnetic eld is
r

B=r

A = r (r A)

r2 A:

(3.8)

The divergence of the vector potential vanishes for static elds, since
Z

1
r
J(r0 )dV 0
4
jr r0 j
Z
1
0
=
r0
J(r0 )dV 0
4
jr r0 j
Z
Z
J(r0 )
1
0
0
0
=
r0
dV
+
r0 J(r0 )dV 0
0
4
jr r j
4
jr r0 j
= 0;
0

r A =

where it is noted that

r0

J(r0 )
jr r0 j

dV 0 =

J(r0 )
dS0 = 0;
jr r0 j

and that for steady current ow, r J = 0: Therefore, for static magnetic elds,
r

B=

r2 A:

r2
Z
4

(3.9)

However,
r A =
=
=

J(r0 )dV 0

jr

r0 j

(r

r0 )J(r0 )dV 0
(3.10)

0 J(r):

This means that the vector potential for a prescribed current distribution J(r)
A(r) =

J(r0 )
dV 0 ;
jr r0 j

is the particular solution for the vector Poissons equation

r2 A =

0 J(r);

(3.11)

0 J:

(3.12)

which is identical to the familiar Amperes law

r

B=

In summary, the Biot-Savarts law is generic in the sense other well known laws in magnetostatics
follow from it. In particular, we have derived the following vectorial relationships from the BiotSavarts law:
r B = 0;
r B = 0 J;
B = r A;
r A = 0;
r2 A =
0J

absence of magnetic monopoles (always)

Amperes law for static elds
always
for static elds; always in Coulomb gauge (by denition)
static Amperes law in terms of the vector potential A

The Amperes law in Eq. (3.12) is valid only for static elds. For more general time-varying elds,
it is generalized as
@E
;
(3.13)
r B = 0 J + "0
@t
where
D = "0

@E
; (A/m2 )
@t

(3.14)

is the displacement current density originally conceived by Maxwell. The divergence of Eq. (3.13)
yields (note that "0 r E = )
r (r

B) = 0 =

r J+

@
@t

that is, the Maxwells equation in Eq. (3.13) is consistent with the charge conservation law,
@
+ r J = 0:
@t

(3.15)

The divergence of the vector potential A can be assigned an arbitrary scalar function without
aecting the electric and magnetic elds. However, for static elds, the divergence of the vector
potential identically vanishes. For time varying elds, the following choice is often made,
r A=
where

1 @
; Lorentz gauge
c2 @t

(3.16)

is the scalar potential. As we will see, this choice for r A allows complete decoupling

between

and A in the sense that the potentials satisfy similar inhomogeneous wave equations,
r2

1 @2
c2 @t2

1
(r; t);
"0

(3.17)

r2

1 @2
c2 @t2

A=

0 J(r; t):

(3.18)

If the Coulomb gauge, characterized by r A = 0 (absence of longitudinal vector potential), is

used instead, such decoupling cannot be achieved. In Coulomb gauge, the scalar potential C and
transverse vector potential A? satisfy the following,
r2

1 @2
c2 @t2

r2

1
(r; t);
"0

where J? is the transverse current. Solution of

(r; t) =

0 J? (r; t);

A? =
C

1
4 "0

(3.19)

(3.20)

(r; t) is non-retarded,
(r0 ; t)
dV 0 ;
jr r0 j

(3.21)

and so is the longitudinal electric eld,

EC (r; t) =

1
4 "0

(r

r0 ) (r0 ; t) 0
dV :
jr r0 j3

(3.22)

As we will see in Chapter 4, this non-retarded electric eld is cancelled and observable electric eld
is in fact always retarded. Observable electromagnetic elds, E and B; should be independent of
the choice of gauge.
In the vector Poissons equation for the vector potential,
r2 A =

0 J;

(3.23)

if the current density J is unidirectional, so is the vector potential and A k J holds. For example,
if only Jz exists, the resultant vector potential is also in the z-direction, Az : If the current density
is azimuthal J ; so is the vector potential, and only A exists. It is noted that in noncartesian
coordinates,
r2 A i 6= r2 Ai ;
(3.24)

in general. For example, in the spherical coordinates,

r2 A

r(r A)

r (r A)
2
2 @A
2 cot
2 @A
er
=
r 2 Ar
Ar
A
2
2
2
2
r
r @
r
r sin @
1
2 @Ar
2 cos @A
+ r2 A
e
2 A + r2 @
2
r sin
r2 sin2 @
2 cos @A
2 @Ar
1
+
e :
A
+
+ r2 A
2
2
r sin @
r2 sin
r2 sin2 @

(3.25)

If the current is purely azimuthal, J = J e ; the azimuthal component of the vector potential
satises
1
r2 A
A =
(3.26)
0J :
r2 sin2
In axially symmetric cases with @=@ = 0; elementary solutions for
1
A = 0;
r2 sin2

r2 A
are

A (r; ) = rl Pl1 (cos ) ;

1
rl+1

Pl1 (cos ) :

(3.27)

(3.28)

Then the lowest order harmonic is l = 1 (dipole) which is consistent with the absence of magnetic
monopoles.

3.3

Boundary Conditions

The vanishing divergence of the magnetic eld r B = 0 requires that the normal component of
the eld B be continuous at any boundary,
B1n = B2n :

(3.29)

The static Amperes law,

r

B=

(3.30)

0 J;

indicates that the tangential component can be discontinuous in the presence of a surface current,
n

(B1

B2 ) =

0 Js ;

(3.31)

where n is a unit vector normal to the surface and Js (A/m) is the surface current density. The
surface current density Js is the total current density regardless of its origin, including translational
motion of charge (conduction current), collection of magnetic dipoles (magnetization current), etc.
As will be shown later, for practical applications, it is convenient to separate the conduction current
Jcond = v ( (r) the charge density and v = dr=dt the velocity of charge density) and rewrite the

Amperes law in terms of a vector H,

r
where
H=

H = Jcond ;

M; (A m

(3.32)

(3.33)

with M the magnetization vector or the magnetic dipole moment density. The boundary condition
for the tangential component of the vector H is
n

(H1

H2 ) = Jcs ;

where Jcs is the conduction surface current density having dimensions of A m

3.4

(3.34)
1.

Some Examples

In this section, vector potentials will be worked out for some simple current congurations.
Example 1 Ring Current

Figure 3-2: Ring current. Because of axial symmetry, nding the vector potential on the y
plane will su ce.

Consider a thin circular ring current I with radius a: The current is unidirectional in the
direction and the current density can be written as
J =I

(r

a)
a
8

(cos ) :

(3.35)

Therefore, the vector potential is also unidirectional and only the azimuthal component A is
nonvanishing. When Ar = A = 0; in Eq. (3.25), the component of the vector Poissons equation
r2 A =

(3.36)

0 J(r);

in the spherical coordinates is

1

r2

A =

r2 sin2

where
r2 =

0J

0I

@
@2
2 @
1
+
+
@r2 r @r r2 sin @

(r

a)
a

sin

@
@

(cos ) ;

(3.37)

(3.38)

is the scalar Laplacian. Note that @=@ = 0 because of axial symmetry of the problem. Since
bounded general solutions to the dierential equation
@2
2 @
1
@
+
+
@r2 r @r r2 sin @

@
@

sin

are

rl Pl1 (cos );

rl+1

1
r2 sin2

A (r; ) = 0;

Pl1 (cos );

(3.39)

(3.40)

the solution for A (r; ) may be assumed in the form

A (r; ) =

8 X
r
>
>
Al
>
>
a
>
< l

Pl1 (cos );

r<a
(3.41)

>
X
>
a
>
>
>
Al
:
r

l+1

Pl1 (cos ); r > a

where the continuity of A at r = a is required from the continuity in the radial component of the
magnetic eld,
1 @
(sin A ):
(3.42)
Br (r; ) =
r sin @
The appearance of Pl1 (cos ); even though the problem is axisymmetric, is due to the nature of the
vector Laplace equation. Eq. (3.39) is formally identical to scalar Laplace equation with m = 1:
The coe cient Al can be determined by multiplying Eq. (3.39) by Pl10 (cos ) sin and integrating
the result over = 0 to ;
X
l

Al

d2
2 d
+
2
dr
r dr

l(l + 1)
r2

gl (r)

1
1

Pl1 ( )Pl10 d =

0I

(r

a)
a

Pl10 (0);

(3.43)

where

= cos and

8
>
>
>
<

gl (r) =

Noting

1
1

Pl1 ( )Pl10 ( )d =

Pl1 (0) =

and the singularity at r = a;

>
>
>
:

r
a

a
r

l+1

r<a
(3.44)
; r>a

2 (l + 1)!
2l + 1 (l 1)!

8
l 1
>
>
>
< ( 1) 2

2
l(l + 1)
2l + 1

ll0 ;

l!!
; l odd
(l 1)!!

>
>
>
: 0;

d2
gl (r) =
dr2

ll0

(3.45)

(3.46)

l even

(2l + 1)

(r

a)
a

(3.47)

we nally obtain
Al =

0I

Pl1 (0)
=
2l(l + 1)

0 I(

and
A (r; ) =

0I

1
X

1)

l 1
2

gl (r)

l=1

(l 2)!!
;
2(l + 1)!!

l = 1; 3; 5;

Pl1 (0) 1
P (cos ):
2l(l + 1) l

(3.48)

(3.49)

Note the absence of the monopole potential, l = 0:

Far away from the ring r
a; the vector potential is dominated by the dipole term (l = 1) as
expected,
mz
A (r; ) ' 0 2 sin ;
(3.50)
4 r
or in vector form,

m r
;
4
r3
0

(3.51)

m = a2 I ez ;

(3.52)

A =
where

is the magnetic dipole moment of the ring current,

Z
1
m =
r JdV
2
Z
1
=
( e )rJ r2 sin drd d
2
Z
(r a) (cos ) 3
I
=
(sin ez cos e )
r sin drd d
2
a
=
a2 I ez :

10

The vector potential of the ring current can alternatively be found from the following direct
integration,
Z
J(r0 )
0
A(r) =
dV 0 ;
(3.53)
4
jr r0 j
where for a lamentary ring current, the volume integral can be replaced with line integral,
J(r0 )dV 0 = I dl = Ia d 0 e 0 :
Because of the axial symmetry, the vector potential can be evaluated at an arbitrary location and
we choose = =2; a point on the y z plane. Then only the x component of Iad 0 e 0 contributes
to the integral,
Z 2
sin 0
p
d 0;
A (r; ) = 0 Ia
2
2
4
r +a
2ar cos
0
where

cos

= cos cos

+ sin sin

2
= sin sin 0 :

By changing the variable from

to

cos

through
2 =

the integral can be manipulated into the following form,

A (r; ) =

0 Ia

r2

a2

1
1
(2
+ 2ar sin k 2

where
k2 =

r2

k 2 )K(k 2 )

4ar sin
;
+ a2 + 2ar sin

2E(k 2 ) ;

(3.54)

(3.55)

and K(k 2 ) and E(k 2 ) are the complete elliptic integrals of the rst and second kind, respectively,
dened by
Z =2
Z =2 p
1
2
2
p
K(k ) =
d ; E(k ) =
1 k 2 sin2 d :
(3.56)
0
0
1 k 2 sin2
Using this result, the inductance of a thin conductor ring of loop radius a and wire radius b
with a
b can be estimated. The magnetic ux enclosed by the ring current itself is
Z
I
=
B dS =
A dl:
(3.57)
S

11

Figure
H 3-3: A thin conductor ring. The magnetic ux enclosed by the ring can be found from
= A dl where A = A is the vector potential on the surface of the ring.
On the inner surface of the wire, r = a b; and = =2: The argument of the elliptic functions k 2
approaches unity and we may approximate the elliptic functions by
lim K(k 2 ) = ln

k2 !1

4
1

= ln

k2

8a
b

lim E(k 2 ) = 1;

k2 !1

where
1

k2 = 1

4a(a b)
b)2 + a2 + 2a(a

(a

b)

'

b2
:
4a2

(3.58)

(3.59)

Then, the vector potential on the wire surface is

A '

0I

ln

8a
b

2 ;

(3.60)

and the magnetic ux enclosed by the ring is

= 2 (a

b)A
8a
0 aI ln
b

'

2 :

(3.61)

The external self inductance of the ring is therefore given by

Lext =

0a

ln

8a
b

2 :

(3.62)

External inductance pertains to the magnetic ux outside the conductor. Magnetic eld also
exists in the conductor. For a straight conducting wire, the internal inductance per unit length is
Li
= 0 ; (H/m).
l
8

12

(3.63)

The internal inductance Li is dened by

1
Li I 2 = magnetic energy in the conductor.
2

(3.64)

For a straight wire of radius a carrying a current I uniformly distributed across the cross-section,
the magnetic eld in the conductor can be found by using Amperes law,
2

B ( )=

B ( )=

0 2 I;
a

0I

(3.65)

):

(3.66)

2 a2

As will be shown, the magnetic energy density is

1
2

B 2 ; (J m

Then, the magnetic energy stored per unit length of the wire is
Z

1
2

B 2 ( )2

d =

1 0 2
I ; (J m
28

),

(3.67)

which is independent of the wire radius a:This denes the internal inductance per unit length,
Li
= 0 ; (H m
l
8

).

(3.68)

If the current density is not uniform as in strong skin eect, the internal inductance must be
evaluated according to Eq. (3.64).
In the cylindrical coordinates ( ; ; z); the vector potential A ( ; z) satises
@2
1 @
@2
+
+
@ 2
@
@z 2

1
2

A ( ; z) =

0I

a) (z):

(3.69)

The solution for this equation can readily be found in terms of Fourier transform involving the
modied Bessel functions I1 (k ) and K1 (k ) as

A ( ; z) =

8 R1
>
I1 (k )K1 (ka) cos(kz)dk;
Ia < 0
>
: R1
0

I1 (ka)K1 (k ) cos(kz)dk;

< a;
(3.70)
> a:

This expression or its alternative based on Laplace transform,

A ( ; z) =

0 Ia

J1 (ka)J1 (k )e

kjzj

dk;

is useful in calculating the self-inductance of a uniformly wound solenoid.

13

(3.71)

Example 2 Rotating Charged Conducting Sphere

If a conducting sphere of radius a carrying a surface charge density rotates about its axis at
an angular frequency !, the current density becomes singular and is given by
J = !a (r

a) sin :

(3.72)

Solutions for the vector potential satisfying

Figure 3-4: Rotating charged conductor sphere.

2 @
1
@
@2
+
+ 2
2
@r
r @r r sin @

sin

@
@

A (r; ) =

r2 sin2

!a (r

a) sin ;

(3.73)

can be sought in the form

A (r; ) =

8
P
r
>
>
>
l Al
<
a
>
>
>
: P Al a
l
r

Pl1 (cos );

r<a
(3.74)

l+1

Pl1 (cos ); r > a

where Al is a constant. The radial function

gl (r) =

8
>
>
>
<
>
>
>
:

r
a

a
r

l+1

r>a
(3.75)

14

; r<a

has discontinuity in its derivative at the surface r = a and thus

d2
gl (r) =
dr2
The sin
present,

(2l + 1)

(r

a)
a

(3.76)

dependence of the singular current density requires that only the l = 1 harmonic be
P11 (cos ) = sin :

Therefore, the solution for the vector potential is

8 r
>
sin ;
r<a
>
>
< a
2
!a
A (r; ) = 0
>
3
>
a 2
>
:
sin ; r > a
r
The outer eld is of pure dipole with an eective dipole moment
!a4
!qa2
=
;
3
3

mz =

q = 4 a2 :

The exterior magnetic eld is given by

B = r

A
m
z
0
(2 cos er + sin e ) :
4 r3

=
The interior magnetic eld is uniform,
B =
=

!a

3
2

!a

ez
(cos er

sin e ); r < a:

(3.77)

The polar component of the magnetic eld B is discontinuous at the surface with a jump
B (r = a + 0)

B (r = a

0) =

!a sin =

where
Js = !a sin ;
is the surface current density.
Example 3 Rotating Charged Sphere

15

(A/m)

0 Js ;

(3.78)

If a uniformly charged sphere of radius a and uniform charge density (C m

its axis at an angular frequency ! (rad s 1 ); the current density J (r; ) is

J (r; ) =

The vector potential satises

@2

2 @
@
1
+
+ 2
2
@r
r @r r sin @

@
sin
@

3)

rotates about

8
>
< ! r sin ; r < a;
>
:

0;

1
2
r sin2

(3.79)

r > a:

A (r; ) =

8
>
<
>
:

0!

r sin ; r < a;
(3.80)

0;

r > a:

In the region outside the sphere r > a where there is no current, the solution may be assumed to
be
X
a l+1 1
(3.81)
Pl (cos ); r > a:
A (r; ) =
Al
r
l

Interior solution may be sought in the form of a sum of general and particular solutions. The
particular solution can be readily found by assuming
A (r; ) = Br3 sin :

(3.82)

Substituting this into

@2
2 @
1
@
+
+ 2
2
@r
r @r r sin @

sin

@
@

A (r; ) =

r2 sin2

0!

r sin ;

(3.83)

we nd the particular solution,

1
10

AP (r; ) =

r3 sin ;

0!

(3.84)

and the interior solution may be assumed to be

A (r; ) =

Bl

r
a

Pl1 (cos )

1
10

0!

r3 sin ;

r < a;

(3.85)

where the rst term in the RHS is the general solutions. Continuity of the magnetic eld at the
sphere surface requires that both A and @A =@r be continuous. Then, only the l = 1 component
is nonvanishing and we easily nd
A1 =

1
15

0!

a3 ; B1 =

16

1
6

0!

a3 :

(3.86)

As in the preceding example, the eld outside the sphere is pure dipole. This is due to the particular
angular dependence of the current density J _ sin : Comparing the outer vector potential
A (r; ) =

1
15

a5

0!

sin
;
r2

(3.87)

with the standard dipole eld, we nd the dipole moment of the rotating charged sphere
4
! a5 :
15

mz =

(3.88)

The reader should check this result agrees with the dipole moment directly calculated from
Z
1
m=
r J dV:
(3.89)
2 V
Example 4 Long Straight Currents
Let us assume a straight current I directed along the z axis. The cartesian component of the
vector potential Az satises scalar Poissons equation
r 2 Az =

0 Jz ;

(3.90)

Jz = I (x) (y):

(3.91)

where

The Poissons equation is mathematically identical to that for the scalar potential due to a long
line charge,
r2

"0

(x) (y);

which has a solution

( )=
where

2 "0

ln ;

x2 + y 2 is the radial distance in the cylindrical coordinates. By analogy, we nd

0I

Az ( ) =

ln :

(3.92)

A resultant magnetic eld is the familiar one,

B=r

Az =

17

0I

e :

(3.93)

For an innite array of identical z directed currents I located in the plane x = 0 at y = 0;

2a;
, the potential is given by
Az (x; y) =
=
=

0I

4
0I

4
0I

ln (x2 + y 2 )[x2 + (y
ln

x2
y2
2 x
a

1+

ln cosh

a;

a)2 ][x2 + (y + a)2 ]

x2
x2
1
+
(y a)2
(y + a)2
2 y
cos
+ constant,
a

y 2 (y

1+

a)2 (y + a)2
(3.94)

where the following closed forms of the innite products are used,
x2
1+ 2
y

x2
1+
(y a)2

y 2 (y

a)2 (y + a)2

= 1

cos

2 x
a

cosh

x2
1+
(y + a)2

=
1
2 y
a

cos

cos
2 y
a

constant.

2 y
a

(3.95)

(3.96)

Calculation of the magnetic eld B(x; y) is left for exercise.

3.5

Inductance

The magnetic ux enclosed by a closed current circuit is in general proportional to the current
I: The external inductance is dened by the ratio,
Le =

(3.97)

In linear systems without ferromagnetic materials, the inductance is a constant determined solely
by geometrical factors. The magnetic ux can be calculated in terms of either the magnetic eld
B or the vector potential A since
Z
Z
I
B dS = (r A) dS =
A dl:
(3.98)
=
S

As an example, consider a parallel wire transmission line with a common wire radius a and
center to center separation distance d: In the absence of skin eect (uniform current density), the
vector potential outside the wires is given by
Az =

0I

ln

1;

> a:

(3.99)

Then, the magnetic ux enclosed by the unit length of the transmission line is
I
1
=2
Az ( 1 = a; 2 )d ;
l
2
18

(3.100)

Figure 3-5: Parallel wire transmission line. The current in the conductors is assumed to be uniform
(no skin eect).

where

is the azimuthal angle about the center of the current I; and

2

The integral needed is

1
2

ln

"

d2 + a2

d
a

+1

2ad cos :

d
2 cos
a

(3.101)

d = 2 ln

d
a

(3.102)

Then the ux per unit length is

0I

ln

d
a

; (Wb m

d
a

; (H m

(3.103)

and the external inductance per unit length is

Le
=
l

ln

).

(3.104)

This holds for an arbitrary ratio of d=a (> 1) as long as the current ows uniformly in both wires
(no skin eect). Each wire has an internal inductance of
Li
= 0 ; (H m
l
8

).

(3.105)

Therefore, the total inductance per unit length of the transmission line is
L
=
l

ln

d
a

1
; (H m
4
19

), (uniform current).

(3.106)

If the current is concentrated at the wire surface as in the case of strong skin eect, the internal
inductance vanishes and the external inductance is modied as
!
p
Le
d + d2 a2
0
; (H m 1 ), (strong skin eect),
(3.107)
=
ln
l
2a
which is dual of the capacitance of parallel wire transmission line,
C
=
l

"0
p
d + d2
2a

ln

a2

! ; (F m

).

(3.108)

The similarity between the capacitance and inductance is due to the fact that the scalar potential
and the cartesian component of the vector potential Az both satisfy the scalar Laplace equation,
r2

r2 Az = 0;

= 0;

(3.109)

in the case of parallel wire transmission line.

Example 5 Inductance of a Solenoid
The inductance of closely wound solenoid of circular cross-section as shown in Fig.3-6 can be
found using the vector potential of a single current loop of radius a,

A ( ; z) =

8 Z
>
>
>
>
Ia <

Z
>
>
>
>
:

I1 (k )K1 (ka) cos(kz)dk;

< a;

(3.110)
1

I1 (ka)K1 (k ) cos(kz)dk;

> a:

Consider a solenoid of radius a and length l with a total number of windings of N: The vector
potential at arbitrary axial location z on the solenoid surface = a is
0 N Ia

A (a; z) =

dz

I1 (ka)K1 (ka) cos[k(z

z 0 )]dk:

(3.111)

A dierential segment of length dz has a number of windings N dz=l and the magnetic ux linked
to the segment is
d

= 2 aA
=

0I

N
dz
l
Z l
Z 1
N 2 2
dz 0
I1 (ka)K1 (ka) cos[k(z
2a dz
l
0
0

z 0 )]dk:

(3.112)

z 0 )]dk;

(3.113)

The total ux linked to the solenoid is therefore

=

0I

N
l

2a2

dz

dz 0

20

I1 (ka)K1 (ka) cos[k(z

and the inductance is given by

L =
=

(3.114)

I
0

N
l

2a2

dz

dz 0

I1 (ka)K1 (ka) cos[k(z

z 0 )]dk; (H).

(3.115)

This is exact as long as windings are dense enough so that only the azimuthal component of the
vector potential is present.

Figure 3-6: Solenoid with uniform winding density.

For a long solenoid l

a; we note
1
lim I1 (x)K1 (x) = ;
x!0
2

and

cos[k(z

z 0 )]dk =

(z

(3.116)

z 0 ):

(3.117)

Then, the inductance reduces to the familiar form of the inductance of a long solenoid,
L'

N2

a2
=
l

a2 2
N ; (H) ; (long solenoid l
l

a).

(3.118)

An expression valid to order (a=l)2 can be worked out using the approximation that the magnetic
eld is axial everywhere inside the solenoid. The magnetic eld on the axis of a circular loop current
is
I
a2
Bz (z) = 0
:
(3.119)
2 (z 2 + a2 )3=2
21

For a solenoid of winding number density n, length l and radius a; the axial magnetic eld is
2
0 nIa

Bz (z) =

dz 0
z)2 + a2 ]3=2

[(z 0

0 nI
2

z
+p
2
z + a2
z)2 + a2

(l

(3.120)

For a long solenoid, the magnetic eld inside the solenoid can be approximated by the axial eld.
Therefore, the ux linkage is
'

2
0n I
a2
2

0n

a I

N
l

a2

N
l

a2

2 2

l
(l

z
+p
2
2
2
z + a2
z) + a

l2 + a2

dz

a ;

(3.121)

and the self-inductance is given by

L '
'

p
l2 + a2
l

a+

a ;

1 a2
2 l

a;

(3.122)

(3.123)

The exact inductance in Eq. (3.114) may be written in the form

L=

(aN ) 2
f (a=l);
l

where f (a=l) is the correction factor given by

a
f
l

=
=

Z
Z l
Z 1
2 l
0
dz
dz
I1 (ka)K1 (ka) cos[k(z
l 0
0
0
Z
a
a
8 1 sin2 (x=2)
I1
x K1
x dk:
2
x
l
l
0

z 0 )]dk

This had been tabulated by Nagaoka. The function f (a=l) is shown in Fig.3-7.

3.6

Neumanns Formula for Mutual Inductance

The mutual inductance among electric circuits can be conveniently calculated using Neumanns
formula. Consider two closed current loops C1 and C2 : To nd the mutual inductance of the two
loops, we let the loop 1 carry a current I1 and calculate the magnetic ux linked to the loop 2. The
vector potential created by the current I1 at a point on the loop 2 is
A=

0 I1

22

dl1
;
r12

(3.124)

1
0.8
0.6
0.4
0.2
00

0.5

1
x

1.5

Figure 3-7: Nagaoka factor f (a=l) for a solenoid of radius a and length l: In the gure, x = a=l:

where r12 is the distance between the segment dl1 and the position on loop 2 as illustrated in
Fig.3-8. Then, the magnetic ux enclosed by loop 2 is given by
I
A dl2
12 =
I I
dl1 dl2
0 I1
:
(3.125)
=
4
r12
The mutual inductance between the two loops is dened by

Figure 3-8: Geometry for two circuits mutual inductance (Neumanns formula).

23

12

M12 =

I1

I I

dl1 dl2
:
r12

(3.126)

It is evident that the mutual inductance is reciprocal,

M12 = M21 ;

(3.127)

and depend only on the geometrical shapes and their mutual orientation.
Example 6 Coaxial Parallel Square Loops

Figure 3-9: Coaxial parallel square loops. The mutual inductance can be found from the sum of
contributions from all parallel and anti-parallel pairs.

The mutual inductance between two identical square loops, which are coaxial and whose sides
are parallel, can be calculated as a sum of mutual inductances of parallel wires. If two parallel
wires of length a are a distance d apart, the mutual inductance is
M

=
=
=

dl1

ln

"

a ln

dl2 p
p

1
(l1

l2 )2 + d2
!

x2 + d2
dx
d
!
#
p
p
a + a2 + d2
a2 + d2 + d ; (H).
d
x+

(3.128)

Applying this results to all parallel pairs of the square loops, we nd

M12 =

"

!
p
2 + d2
a
0
a ln
a2 + d2 + d
d
!
#
p
p
p
a + 2a2 + d2
p
a ln
+ 2a2 + d2
a2 + d2 ; (H).
a2 + d2
a+

Example 7 Mutual Inductance of Coaxial Circular Loops

24

(3.129)

Figure 3-10: Coaxial circular loops.

If two circular loops of radii a and b are coaxial and their planes are a distance c apart, the
mutual inductance can readily be found from the vector potential at the loop B due to a current I
in the loop A,
1
1
Ia
(2 k 2 )K(k 2 ) 2E(k 2 ) ;
(3.130)
A = 0 p
2
2
2
k
(a + b) + c
where

k2 =

4ab
:
(a + b)2 + c2

(3.131)

The magnetic ux enclosed by the loop B is thus

= 2 bA
1
2 0 abI
(2
= p
(a + b)2 + c2 k 2

k 2 )K(k 2 )

2E(k 2 ) ;

(3.132)

2E(k 2 ) ; (H).

(3.133)

and the mutual inductance is

Mab =

1
2

k 2 )K(k 2 )

(a + b)2 + c2 (2

An alternative expression for the mutual inductance can be found in terms of spherical harmonic
p
expansion. The vector potential at the loop B is, if a < b2 + c2 ;
A (r =

p
b2 + c2 ;

0)

0I

l odd

where
cos

=p

25

a
2
b + c2

b
:
b2 + c2

l+1

Pl1 (0) 1
P (cos
2l(l + 1) l

0 );

(3.134)

(3.135)

If a >

b2 + c2 ;

A (r =

b2 + c2 ;

0)

0I

l odd

b2 + c2
a

!l

Pl1 (0) 1
P (cos
2l(l + 1) l

0 ):

(3.136)

The ux linked to the loop B is therefore given by

= 2 bA (r =

b2 + c2 ;

0 );

(3.137)

and the mutual inductance is

Mab = b

l odd

and
Mab = b

a
2
b + c2

l odd

b2 + c2
a

l+1

!l

Pl1 (0) 1
P (cos
l(l + 1) l

Pl1 (0) 1
P (cos
l(l + 1) l

0 );

if a <

0 ); if a >

p
b2 + c2 ;

p
b2 + c2 :

(3.138)

(3.139)

Figure 3-11: Geometry for calculating the mutual inductance of two circular loops whose axes
intersect at an angle :
If two circular loops are not coaxial but their axes intersect as shown in Fig.3-11, the mutual
inductance can be calculated as follows. First, we note the vector potential due to a circular current
of radius a sin residing on a sphere of radius a is given by (see Eq. (3.49))
A (r; ) =

0I

X a
r
l

l+1

sin
P 1 (cos )Pl1 (cos ):
2l(l + 1) l

26

(3.140)

The radial magnetic eld is

1 @
(sin A )
r sin @
X a l+1
0 I sin
Pl1 (cos )Pl (cos );
2r
r

Br =
=

(3.141)

where use is made of the identity,

1 d
[sin Pl1 (cos )] = l(l + 1)Pl (cos ):
sin d

(3.142)

The magnetic ux linked to the loop B can be found by integrating the radial magnetic eld over
the incomplete spherical surface at radius b: Denoting the angular location on the sphere by ( 0 ; 0 )
where 0 is measured from the axis of the loop B and 0 is the angle about it, we note that the
argument of the Legendre function Pl (cos ) becomes
cos = cos

cos + sin

sin cos

Then the Legendre function can be expanded as,

Pl (cos

cos + sin

sin cos ) = Pl (cos 0 )Pl (cos ) + 2

1
X

Plm (cos 0 )Plm (cos ) cos m ; (3.143)

m=1

and the magnetic ux through the loop B is

= b

0 Ib sin sin

sin 0 Br (r = b)

X
l

where the integration over

1
a
l(l + 1) b

l+1

Pl1 (cos )Pl1 (cos )Pl (cos );

(3.144)

reduces to
Z

sin 0 Pl (cos 0 )d

sin
P 1 (cos ):
l(l + 1) l

(3.145)

The desired mutual inductance is

Mab =

0 b sin

sin

X
l

3.7

1
a
l(l + 1) b

l+1

Pl1 (cos )Pl1 (cos )Pl (cos ); (H).

(3.146)

Multipole Expansion of Static Vector Potential

The vectorial nature of the vector potential prohibits us from expanding the vector potential directly
in terms of the spherical harmonics in contrast to the case of scalar potential. However, the
vector potential can be indirectly expanded in spherical harmonic functions. In magnetostatics,

27

the Coulomb gauge r A = 0 allows us to write A in terms of another vector function F as

A=r

(3.147)

F;

since r (r F) 0 holds identically. The vector F introduced may be regarded as a generating

function for the vector potential and it may not have any physical meanings. The vector eld F
can be decomposed into radialor longitudinalcomponent, and transversecomponent in the
form
F= r +r r ;
(3.148)
where and are scalar functions. If the vector potential satises vector Laplace equation r2 A =
0; then and both satisfy scalar Laplace equation,
r2

r2 = 0;

= 0;

(3.149)

and thus can be expanded in spherical harmonics,

;

Alm rl +

l;m

Blm
rl+1

Ylm ( ; ):

(3.150)

The transverse function does not contribute to the magnetic eld B: (Why? Note the identity
r (r r ) = 0.) Then, can de discarded and
A=r

r :

(3.151)

The dierential operator

r

r=

@
1 @
+e
;
sin @
@

(3.152)

is the well known angular momentum operator and operates only on the harmonic function Ylm ( ; ):
Evidently, Eq. (3.151) generates components A and A only, but the remaining radial component
Ar can be calculated from r A = 0: Note that the radial component Ar ; which exists in
r

(r

r );

does not contribute to the magnetic eld.

Let us revisit the vector potential due to a ring current. Because of the axial symmetry, the
scalar function may be expanded in the form

(r; ) =

8 P
r
>
>
al
>
>
a
< l

>
>
P
a
>
>
:
al
r
l

Pl (cos );

r<a
(3.153)

l+1

28

Pl (cos ); r > a

and the vector potential can be calculated as

A = r
=

@
r =e
@
X
e
al gl (r)Pl1 (cos );

(3.154)

where
gl (r) =

and use is made of

8
>
>
>
<

r
a

a
r

l+1

r<a
(3.155)

>
>
>
:

d
Pl (cos ) =
d

; r>a

Pl1 (cos ):

(3.156)

The vector potential is therefore consistent with that worked out in Example 1.
If the current density J(r) is conned in a limited spatial region, the vector potential at a
su ciently large distance can be Taylor expanded as
A(r) =
=
where r

1
4 r
0

J(r0 )
dV 0
jr r0 j

1
J(r )dV +
4 r3
0

(r r0 )J(r0 )dV 0 +

(3.157)

r0 is assumed. The rst term in RHS vanishes for static current, since
Z
Z
0
Ji dV =
rri0 JdV 0
Z
Z
0
0
=
r (ri J)dV
ri0 r JdV 0
= 0;

Note that for static currents, r J = 0 holds. The absence of monopole vector potential is consistent
with the absence of magnetic charges (magnetic monopoles), r B = 0: Therefore, the lowest order
far eld vector potential is the dipole eld,
Z
1
A(r) ' 0 3 (r r0 )J(r0 )dV 0 :
(3.158)
4 r
Using
r

(r0 J) = r0 (r J)

29

(r r0 )J;

and
Z

ri0 (r

J)dV

ri0 rj Jj dV 0

= rj
= rj
=
the integral can be reduced to
Z

= rj

ri0 Jj dV 0

ri0 (rrj0 J)dV 0

(ri0 rj0 J)dV 0

r
Z
rj rj0 Ji d0 =

1
2

(r r0 )J(r0 )dV 0 = r

Z
Z

rj

rj0 (rri0 J)dV 0

(r r0 )JdV 0 ;

J(r0 )

r0 dV 0

(3.159)

and the dipole vector potential can be written in the form

A(r) =

m r
;
r3

(3.160)

where m is the magnetic dipole moment,

1
m=
2

JdV; (A m2 ):

(3.161)

For a lamentary loop current, the volume integral reduces to line integral,
JdV = Idl;
and the dipole moment can be found from the contour integral,
I
1
m= I
r dl = IS;
2 C
where

1
S=
2

dl;

(3.162)

(3.163)

is the area enclosed by the loop.

3.8

Magnetic Materials

If magnetic dipoles are continuously distributed, it is convenient to introduce a magnetic dipole

moment density,
M(r) = n(r)m(r); (A m 1 );
(3.164)
where n(r) is the number density of dipoles and m(r) is the dipole moment. Although individual
dipoles do not carry macroscopic current, a collection of many dipoles can eectively produce
30

a macroscopic current. However, such an eective current is divergence-free and thus does not
contribute to charge transfer. The vector potential in the presence of distributed magnetic dipoles
can be approximated by
A(r) =

Jc (r0 )
dV 0 + 0
0
jr r j
4

M(r0 )
jr

(r r0 ) 0
dV ;
r0 j3

where Jc (r) is the conduction current due to translational motion of charges. The second term in
RHS is the contribution from the distributed magnetic dipoles. Singling out the conduction current
Jc in magnetostatics is similar to separating the electric charges into free and bound charges in
dielectrics,
Z
Z
0
1
P(r0 ) (r r0 ) 0
1
f (r )
0
dV :
dV
+
(r) =
4 "0
jr r0 j
4 "0
jr r0 j3
Noting

r0
1
0
;
3 = r jr
0
r0 j
rj

r
jr
Z

M(r0 )
jr

(r r0 ) 0
dV =
r0 j3

M(r0 )
Z

r0

1
jr

r0 j

dV 0

M(r0 )
r0 M(r0 )
dV 0
jr r0 j
jr r0 j
I
Z
M(r0 )
r0 M(r0 ) 0
=
dS+
dV
r0 j
jr r0 j
S jr
Z
r0 M(r0 ) 0
dV ;
=
jr r0 j

r0

(3.165)

we nd that the curl of the magnetic eld becomes

r

B=

r2 A =

r2

0 Jc

0r

M:

(3.166)

where use is made of the identity

jr

r0 j

(r

r0 ):

Rearranging Eq. (3.166) yields

r

= Jc :

(3.167)

It is customary to introduce a vector H dened by

H=

B
0

31

M;

(3.168)

and an alternative form of Amperes law is

r

H = Jc :

(3.169)

In vacuum and nonmagnetic media, M = 0; and

B=
holds. The permeability

0 H;

of linear magnetic materials is dened by

B = H:

Permanent magnets are highly nonlinear. In permanent magnets, there are no conduction current,
Jc = 0; and thus
r H = 0;
(3.170)
must holds. This means that the vector H can be assigned a scalar potential,
H=

m:

(3.171)

In permanent magnets, the two vectors B and H are in general oriented in opposite directions as
shown in the following simple example.
Example 8 Magnetized Sphere
If an iron sphere of radius a is uniformly magnetized with an axial uniform magnetization
M = M0 ez ;the elds B and H can be found as follows. The scalar potential m satises the
Laplace equation,
r2 m = 0:
In the sphere r < a the uniform magnetization produces uniform magnetic elds B and H;
Bz =

0 (Hz

+ M0 ); r < a;

(3.172)

where Bz and Hz are constants. Then the solution for the scalar potential may be assumed as

m (r;

)=

8
>
>
<
>
>
:

Hz r cos ;

r<a

a3
Hz 2 cos ; r > a
r

(3.173)

where continuity of m ; that is, continuity of the tangential component of H; H ; at the surface
r = a is taken into account. The continuity of the radial component of the magnetic eld Br yields
0 (M0

+ Hz ) =
32

0 Hz ;

(3.174)

or

1
M0 :
3

Hz =

(3.175)

This is the intensity of the vector H inside the sphere. The interior magnetic eld is

Figure 3-12: Spherical permanent magnet. The B prole is shown qualitatively in the left gure
and H prole in the right gure. Note the discontinuity in the H eld lines at the surface.

Bz =

2
3

0 M0 :

(3.176)

The eld outside the sphere is of dipole type,

B=

0H =

In the absence of conduction current, r

3
0 M0 a
3r3

(2 cos er + sin e ) :

H = 0 requires
I
H dl = 0;

(3.177)

(3.178)

along an arbitrary closed path. If the path intersects with the magnetized sphere, the vector H
H
must change its sign at the sphere surface in order to satisfy H dl = 0: The magnetic eld B
found in this example is mathematically identical to that in Example 2 worked out for a rotating
charged conducting sphere. This is not surprising because an eective current density associated
with a uniform magnetization indeed yields
Jef f

= r

= M0 (r

a) sin e ;

which is identical to that due to a rotating surface charge on a conducting sphere.

33

(3.179)

In the absence of conduction current, Jc = 0; the following integral over the entire volume
should vanish,
Z
B HdV = 0:
(3.180)
all space

This is because the LHS can be written in terms of the scalar potential m as
Z
Z
Z
B r m dV =
r (B m )dV +
m r BdV = 0:

(3.181)

The magnetic energy associated with a permanent magnet can thus be calculated from either
Z
1
2
Em =
(3.182)
0 H dV;
2
all space
or
Em =

magnet

1
2

0M

(3.183)

HdV:

In the case of spherical permanent magnet, the total magnetic energy is

Z
1
4 3 1
2
Em =
H
Hr2 + H 2 dV
a
+
2 0 z
3
2 0 r>a
2
4
=
M 2 a3 +
M 2 a3
27 0 0
27 0 0
2
=
M 2 a3 ;
9 0 0

(3.184)

where the elds in the exterior region r > a

Hr =

2M0 a3
M0 a3
cos
;
H
=
sin ;
3r3
3r3

have been substituted. The energy agrees with that from

Z
1
1
4 3 2
2
a =
0 M HdV =
0 M0
2
6
3
9
magnet r<a

2 3
0 M0 a :

Example 9 Image Currents

If a current I is placed parallel to a at surface of highly permeable bloc, an image current
in the same direction as the current itself appears. This is because of the boundary condition
that the tangential component of the magnetic eld should vanish at the surface. The image of a
current segment perpendicular to the surface is opposite to the current also because of the boundary
condition.

34

Image currents I 0 in iron (left) and superconductor (right).

If the permeability of the bloc is nite, the vector potential in air region can be calculated by
summing contributions from the current I itself and an image current in the iron,
0

I0 =

I;

and the vector potential in the bloc by assuming a current

2
+

I 00 =

I;
0

at the location of the current.

For a superconducting surface, the normal component of the magnetic eld should vanish.
Therefore, the image of a current parallel to superconducting surface is opposite, while for a vertical
current, the image is in the same direction. Fig. ?? illustrates the two cases.

3.9

Magnetic Force and Stress Tensor

The volume force to act on a current density J is

f =J
Substituting J = r

B=

0;

B; (N m

):

(3.185)

we may rewrite f as
f

(r

B)

1
0

(B r)B

1
rB 2 :
2

(3.186)

However,
r (BB) = (r B)B + (B r)B
= (B r)B:

35

(3.187)

Therefore, the force density can be written in the form

1

f=

BB

B2
1 ;
2

(3.188)

where 1 is the unit tensor. The magnetic stress tensor is dened by

Tm =

BB

B2
1 :
2

(3.189)

This is reminiscent to the electric stress tensor we have encountered in Chapter 1,

Te = "0 EE

E2
1 :
2

(3.190)

The magnetic energy density is

um =

B2
; (J/m3 = N/m2 )
2 0

(3.191)

and the total magnetic energy is

Um =
=
=
=

B2
dV
2 0
Z
1
B (r A)dV
2 0
Z
Z
1
1
r (A B)dV +
A (r
2 0
2 0
Z
1
(A J)dV; (J).
2

B)dV
(3.192)

For a closed current loop, this may be written as

1
Um = I
2

A dl;

where the vector potential can be divided into that part due to the current I itself and another
part due to currents in other circuits. Recalling the denition of self and mutual inductances, the
magnetic energy associated with a single current loop can be written as
1
1 X
Mi Ii ;
Um1 = LI 2 + I
2
2

(3.193)

where Mi is the mutual inductance between the circuit under consideration and other circuits. For
two loops, the total magnet energy stored is given by
1
1
Um = L1 I12 + L2 I22 + M12 I1 I2 ;
2
2

36

(3.194)

because of reciprocity M12 = M21 : The mutual interaction energy M12 I1 I2 can be either negative
or positive depending on the directions of current ows I1 and I2 : However, the total magnetic
energy Um is evidently positive denite.
As electric forces act so as to increase self and mutual capacitances, magnetic forces tend to
increase inductances. The force in the direction of geometrical metric factor i contained in the
inductances can be calculated from
Fi=

1
L( i )I 2 ;
2

@
@ i

(3.195)

for the case of self-inductance, and mutual interaction force from

Fi=

@
(M12 ( i )I1 I2 ) :
@ i

(3.196)

For example, the self-inductance of a circular current loop of loop radius a and wire radius
L=

0a

8a

ln

2+

li
;
2

is

(3.197)

where li is the factor of internal inductance due to the magnetic energy stored in the wire. For
uniform current distribution (no skin eect), li = 1=2: If the loop carries a current I; it tends to
expand so as to increase a with a force
1 2 @
I
L(a)
2 @a
1
8a
I 2 ln
2 0

Fa =
=

1+

li
:
2

(3.198)

Example 10 Toroidal Magnet

Consider a toroidal magnet with a major radius R; minor radius a; and number of windings N:
By the Amperes law, the magnetic eld in the magnet can be readily found,
B =

NI
;
2

a<

< R + a:

The magnetic ux is

0N I

R+a

R a

0N I R

q
a2

(a

R2

a2 :

)2

d
(3.199)

Therefore, the inductance of the magnet is

L=

2
R
0N

37

p
R2

a2 :

(3.200)

The magnet tends to shrink in the major radius direction with a force
FR =
=

1
2
1
2

2
0 (N I)

@
R
@R

2
0 (N I)

R2

R2

a2

a2
< 0:

(3.201)

In the minor radius direction, the magnet tends to expand with a force
Fa =
=

1
2
1
2

p
@
R2 a2
R
@a
a
2
> 0:
0 (N I) p 2
R
a2
2
0 (N I)

(3.202)

In large fusion devices, mechanical support structures must be able to handle these magnetic forces
which are enormous.

3.10

Boundary Value Problems

As discussed briey, the Maxwells equation for the magnetic eld B;

r B = 0;

(3.203)

requires that the normal components of B be continuous across a boundary of two magnetic media,
B1n = B2n :

(3.204)

For the eld H;

r

H = Jc ;

(3.205)

the tangential components may be a provided there exists a surface conduction current Js (A/m)
owing on the boundary surface,
n (H1 H2 ) = Js ;
(3.206)
In the absence of conduction surface current, the tangential components of H should be continuous.
On the surface of highly permeable medium such as iron, the magnetic eld lines falls almost
normal to the surface provided iron body and current circuits are not linked topologically. Linked
and unlinked examples are shown in Fig. 3-13. In unlinked cases, the eld H in iron should vanish
if iron
0 : Since
B1 cos

= B2 cos

2;

or

1 H1 cos 1

2 H2 cos 2 ;

(3.207)

and
H1 sin

= H2 sin

38

2;

(3.208)

Figure 3-13: Iron rings topologically unlinked (left) and linked (right) to a current loop I: For
unlinked ring, H = 0; while for linked ring, H 6= 0 in iron even if
0 : The magnetic eld lines
fall normal to iron surface if a current is unlinked.

we nd
1

tan

=
1

tan

(3.209)

Therefore, if 2
1 ; the angle 1 must approach 0. The eld H2 in iron should vanish also.
In linked cases, the eld H remains nite even in iron and should satisfy
I
H dl = I;
(3.210)
along a path in iron. The magnetic eld lines do not necessarily fall normal to the iron surface.
For example, if a current is placed at the center of hollow iron cylinder, the magnetic eld line at
the iron surface is tangential everywhere.
Example 11 Ring Current around a Long Iron Core
We consider a circular ring current I of radius b which is coaxial with a long, highly permeable
iron cylinder of radius a as shown in Fig.3-14. If the iron cylinder has no return legs, it is not
topologically linked to the current ring. Therefore, the boundary condition for the magnetic eld
at the iron surface is
Bz ( = a; z) = 0:
(3.211)
Because of axial symmetry, the vector potential A ( ; z) can be found from the following equation,
1 @
@2
@2
+
+
@ 2
@
@z 2

A ( ; z) =

0I

b) (z):

(3.212)

Fourier transformation through

A ( ; z) =

1
2

1
1

39

A ( ; k)eikz dk;

(3.213)

Figure 3-14: A ring current coaxial with an iron cylinder.

reduces the problem to one dimensional,

d2
1 d
+
d 2
d

k2

A ( ; k) =

0I

b):

(3.214)

Bounded solutions in each region may be assumed as

A ( ; k) =

8
>
< AI1 (k ) + BK1 (k ); a <
>
:

CK1 (k );

< b;
(3.215)

> b;

where the coe cients A; B; and C may be functions of the Fourier variable k: Continuity of A ( ; k)
at = b yields
AI1 (kb) + BK1 (kb) = CK1 (kb):
(3.216)
The axial magnetic eld is
Bz =
Since

1 @
( A ):
@

1 d
1 d
[ I1 (k )] = kI0 (k );
[ K1 (k )] =
d
d

(3.217)

kK0 (k );

(3.218)

the boundary condition Bz = 0 at the cylinder surface gives

B=

I0 (ka)
A;
K0 (ka)
40

(3.219)

and
C=

I0 (ka)
I1 (kb)
+
K1 (kb) K0 (ka)

A:

(3.220)

The coe cient A can be found from the discontinuity of the radial derivatives of the vector potential,
d2
A ( ; k)
d 2

I1 (kb) 0
K (kb)
K1 (kb) 1

= Ak
=b

A
(
bK1 (kb)

I10 (kb)

b)

b);

(3.221)

where use is made of the Wronskian of the modied Bessel functions,

0
Im
(x)Km (x)

0
Im (x)Km
(x) =

1
:
x

Comparing with the RHS of the original equation in Eq. (3.214), we nd

0 IbK1 (kb);

A=

(3.222)

and the solution of the physical vector potential is

A ( ; z) =

8 Z
>
>
>
>
Ib <

Z
>
>
>
>
:

K1 (kb) I1 (k ) +

I0 (ka)
K1 (k ) cos(kz)dk; a <
K0 (ka)

< b;
(3.223)

1
0

I0 (ka)
K1 (k ) I1 (kb) +
K1 (kb) cos(kz)dk;
K0 (ka)

> b:

The rst terms indicate the vector potential due to the current ring alone while the second terms are
corrections due to the presence of the iron cylinder. The iron cylinder increases the ring inductance
by an amount
Z1
I0 (ka) 2
L = 2 0 b2
K (kb)dk:
(3.224)
K0 (ka) 1
0

The magnetic eld can be calculated from

B=r
The radial component is, for a <
B ( ; z) =
=

A :

< b;

@A
@z Z
0 Ib

K1 (kb) I1 (k ) +

41

I0 (ka)
K1 (k ) k sin(kz)dk;
K0 (ka)

(3.225)

and for

> b;
B ( ; z) =

0 Ib

K1 (k ) I1 (kb) +

The axial component in the region a <

Bz ( ; z) =
=

(3.226)

< b is

1 @
( A )
@
Z 1
0 Ib
K1 (kb) I0 (k )
0

and for

I0 (ka)
K1 (kb) k sin(kz)dk:
K0 (ka)

I0 (ka)
K0 (k ) k cos(kz)dk;
K0 (ka)

(3.227)

I0 (ka)
K1 (kb) k cos(kz)dk:
K0 (ka)

(3.228)

> b;
Bz ( ; z) =

0 Ib

K0 (k ) I1 (kb) +

If the permeability of the iron cylinder is not innite but nite , the vector potential in the
region a < < b is modied as

A ( ; z) =

0 Ib

Z1

(K1 (kb)I1 (k ) + f (k)K1 (kb)K1 (k )) cos(kz)dk;

(3.229)

where
f (k) =

(
(

0 )kaI0 (ka)I1 (ka)

0 )kaI1 (ka)K0 (ka)

:
0

Derivation of this modication is left for an exercise.

Example 12 Magnetic Shielding

Figure 3-15: Cross-section of a cylindrical iron shell placed in an external magnetic eld. The eld
inside the shell is greatly reduced.

42

An iron cylinder having inner and outer tic eld radii a; b and permeability
is placed in
an external magnetic eld B0 with its axis perpendicular to the eld. The magnetic eld inside
the cylinder is greatly reduced if
0 even if the thickness of the cylinder is small. Since no
conduction currents are present, the magnetic eld H can be generated from a scalar potential m
which satises the Laplace equation,
r2 m = 0;
(3.230)
or

@2
1 @
1 @2
+
+
2
@ 2
@
@ 2

m(

; ) = 0:

(3.231)

The external magnetic eld can be generated from

0

H0 cos :

(3.232)

Since the magnetic eld due to the presence of the iron cylinder should have the same angular
dependence, we may assume the following solutions in each region,
8
>
c1 cos ;
0 < < a;
>
>
>
<
c3
cos ;
a < < b;
c2 +
(3.233)
m( ; ) =
>
>
c
>
4
>
cos
H0 cos ; b < :
:
The boundary conditions are both

and @

m =@

be continuous at

c1 a = c2 a +
c2 b +
0 c1

c2

c3
b2

c3
;
a

c3
c4
=
b
b
=
=

(3.234)

H0 b;

c2
0

= a and b: Then,

(3.235)

c3
;
a2
c4
+ H0 :
b2

(3.236)
(3.237)

These simultaneous equations can be readily solved. For the interior magnetic eld, only c1 is
needed. Its solution is
c1 =

4H0 b2
0 2
b

2:

a2

1+
0

If

0;

we recover c1 =

H0 : If

0;

c1 =

(3.238)

1
0

we nd
4H0

43

b2

b2

a2

(3.239)

and thus the interior magnetic eld

Hi = 4H0

b2

b2

a2

H0 :

(3.240)

Example 13 Superconducting Disk in an External Magnetic Field

In this example, we analyze how a superconducting disk disturbs a uniform external magnetic
eld. The boundary condition for the magnetic eld at a surface of superconducting body is that
the normal component vanish because magnetic eld cannot penetrate into superconductors. The
same boundary condition should also apply for ordinary conductors in oscillating magnetic eld if
the skin depth given by
1
; (m)
(3.241)
=p
f 0

is small compared with conductor dimensions. In both cases, electric currents ow at the conductor
surface in a manner to cancel the external magnetic eld.

Figure 3-16: Superconducting disk in an external magnetic eld with its face normal to the eld.

If the disk is ideally thin, the greatest disturbance to the magnetic eld occurs when the disk
axis is parallel to the magnetic eld because in this case, the magnetic dipole moment induced
on the disk surface becomes maximum. The potential of the external uniform magnetic eld is as
before,
(3.242)
m0 = H0 z = H0 r cos :
It is convenient to use the oblate spheroidal coordinates ( ; ; ): A thin disk is described by = 0:
Because of axial symmetry, @=@ = 0; general solutions to the Laplaces equation may be assumed
as
X
[Al Pl (i sinh ) + Bl Ql (i sinh )]Pl (cos ):
(3.243)
m( ; ) =
l

Since the external eld has cos = P1 (cos ) dependence, only the l = 1 harmonic is relevant, and

44

the potential reduces to

m(

; ) = [AP1 (i sinh ) + BQ1 (i sinh )] cos :

(3.244)

Noting z = a sinh cos in the oblate spheroidal coordinates, and P1 (i sinh ) = i sinh ; we nd
A = iaH0 :

(3.245)

The coe cient B can be found from the boundary condition B = 0 at the disk surface, namely,
@
@

= 0 at

= 0:

(3.246)

Since
Q1 (i sinh ) = sinh arccot(sinh )

1;

(3.247)

we readily nd
B=

aH0 ;

where arccot(0) = =2 is noted. The solution for

m(

; ) = aH0

The behavior of the potential at r

sinh +

m(

(3.248)
; ) is thus given by

(sinh arccot(sinh )

(3.249)

a can be found using the asymptotic form of Q1 (i sinh );

1
;
3 sinh2

Q1 (i sinh ) !
with result
m(

1) cos :

1; ) '

H0 r cos

1;

(3.250)

2a3 H0
cos :
3 r2

(3.251)

The dipole moment of the disk placed perpendicular to an external magnetic eld is thus given by
m=

2a3
H0 :
3

(3.252)

Let us check if this is consistent with the dipole moment expected from
Z
Z
1
1
m=
r JdV =
r Js dS;
2
2
where
Js = n
=

H
2

H0 p
a2
45

e ;

(3.253)

which exists on both sides of the disk. Then,

2

mz =

H0

8
H0 a3 ;
3

a2

which agrees with that identied from the far eld potential. Note that
Z

x3
2
dx = :
2
3
1 x

Example 14 Leakage through a Hole in a Superconducting Plate

A uniform magnetic eld parallel to a superconducting plate exists in one of the regions separated by the plate. The plate has a hole of radius a:We wish to nd the magnetic eld in the other
region. This problem is similar to Example 7 in Chapter 2. The leakage led is expected to be of
dipole nature. Since the unperturbed eld can be described by the potential
0

H0 x =

H0 a cosh sin cos ;

in the oblate spheroidal coordinates ( ; ; ); we assume

m(

; ; )=

(3.254)

H0 a cosh sin cos + AQ11 (i sinh ) sin cos ;

where
Q11 (i sinh ) = cosh

sinh
cosh2

arccot(sinh )

(3.255)

(3.256)

It is convenient to dene the function arccot x without discontinuity in its derivative. The asymptotic forms are:
1
1
arccot x ' +
+ ; x ! 1;
(3.257)
x 3x3
1
1
arccot x '
+
; x ! +1:
(3.258)
x 3x3
? 0 corresponds to z ? 0; respectively. In the region z !
and the coe cient A can thus be readily found,
A=

aH0

1; the uniform eld should vanish

(3.259)

In the upper region, the far eld potential becomes

m( ; ; ) =

H0 x +

2H0 a3 r x
;
3
r3

46

a; z > 0:

(3.260)

while in the lower region

m (r) =

2H0 a3 r x
;
3
r3

a; z < 0:

(3.261)

The eective magnetic dipole moments for the region below the hole is
m=

8a3
H0 ; z < 0:
3

(3.262)

In the region z > 0; the eective magnetic dipole moment is

m=

8a3
H0 ; z > 0:
3

(3.263)

The formulae derived here may be conveniently used in analyzing leakage radiation from a small
hole drilled in the wall of waveguides.

Figure 3-17: Leakage of magnetic eld through a hole in a superconducting plate.

47

Problems
3.1 Find the magnetic eld at the center of (a) an equilateral triangular loop and (b) square loop
both having side a and carrying current I:
3.2 Find the magnetic eld at point P:

3.3 A current I ows along a at spiral coil described by

( )=

R
;
2 N

where N is the number of turns and R is the outermost radius of the coil. Show that the
magnetic eld along the axis is given by
Bz (z) =

0N I

3.4 An elliptic loop of axes a and b (a

the center of the loop is given by

"

R+

ln

R2 + z 2
z

#
R
p
:
R2 + z 2

b) carries a current I: Show that the magnetic eld at

Bz =

0I

where
l = 4aE 1

Discuss the limiting cases of a = b and a

b2
a2

ab is its area. E k 2 is the complete elliptic

is the circumference of the loop and S =

integral of the second kind dened by
E k2 =

l
;
4S

=2 p

k 2 sin2 d :

b:

3.5 Find the mutual interaction force between a circular loop current I1 of radius a and a coplanar
straight long current I2 when the center of the loop is at a distance b from the line current.

48

3.6 Show that an alternative expression for the vector potential due to a circular loop current
(radius a; current I) is
0 Ia

A ( ; z) =

J1 (ka)J1 (k )e

kjzj

dk:

3.7 The loop in the preceding problem is placed with its plane parallel to the surface of a large
magnetic bloc of permeability at a distance d: Using the method of image, show that the
inductance of the loop increases by
L=

0a

2
k

where
k=p

2
E(k 2 ) ;
k

k K(k 2 )

a
:
a2 + d2

3.8 A circular current loop of radius a is concentric with an iron sphere of innite permeability
of radius b. Show that the presence of the iron sphere increases the loop inductance by
L=

0b

1
X
l=0

(2l 1)!!
(2l)!!

b
a

4l+2

3.9 Show that the mutual inductance between two circular loops of radii a and b with parallel
49

axes a distance c apart and loop planes a distance d apart is given by

M=

0 ab

J1 (ka)J1 (kb)J0 (kc)e

kd

dk:

3.10 Evaluate the Nagaoka factor numerically for a=l = 0:1; 0:5 and 1.
3.11 A current ows along a circular tunnel in a highly permeable body. Do the magnetic eld lines
fall normal to the tunnel surface? What about the case in which a return current coexists?
Draw qualitatively the magnetic eld lines for both cases.

3.12 A conductor having a rectangular cross-section a

J0 = J0 ez : Find the magnetic eld:

50

b carries a uniform axial current density

3.13 A current sheet I with a width 2a is placed on one of iron walls as shown. Show that the
vector potential in the gap is given by
Az =

0I

4 a

a
a

h
ln cosh

(x

x0 )

cos

dx0 :

3.14 A current I is placed on the midplane of air gap formed by iron as shown. Find the vector
potential and magnetic eld in the gap.

3.15 A thin superconducting ring of major radius a and minor (wire) radius b ( a) is placed in
an external magnetic eld with its plane perpendicular to the eld. Determine the vector
potential and the magnetic dipole moment induced by the ring.
3.16 If the magnetic eld and current density are parallel to each other, the Lorentz force J
vanishes and Amperes law reduces to
r

B = B;

where (m 1 ) is a constant. Such arrangement is pertinent to designing superconducting

wires which are fragile and also to low pressure plasma equilibria such as Reversed Field
Pinch (RFP) and Spheromak. Solve the equation in (a) the cylindrical coordinates assuming
B = 0; @=@z = @=@ = 0; and (b) in spherical coordinates with the boundary condition Br
= 0 at r = a:

51

Chapter 4

Time Varying Fields, Simple Waves

4.1

Introduction

In this Chapter, the charge density

The charge conservation law

and current density J are generalized to be varying with time.

@
+ r J = 0;
@t

(4.1)

imposes a constraint between the two quantities (charge density in C m 3 ) and J (the current
density in A m 2 ) and various electrodynamic laws must be consistent with this basic law. A time
varying magnetic ux induces an electric eld through Faradays law,
I

E dl =

d
dt

B dS:

(4.2)

Likewise, a time varying electric eld induces a magnetic eld through the displacement current,
I

d
B dl = 0 "0
dt
C

E dS;

(4.3)

even in the absence of the current J = 0. As is well known, the displacement current played a
crucial role in Maxwells prediction that electromagnetic waves can propagate through vacuum.
The generalized set of Maxwells equations,
r E=
r

E=

(4.4)

@B
;
@t

(4.5)

"0

r B = 0;
r

B=

J + "0

(4.6)
@E
@t

(4.7)

is consistent with the charge conservation law, since the divergence of the LHS of the last equation
indeed vanishes (recall that r (r B) =0 identically) which requires that
r J + "0

4.2

@
@
r E=r J+
= 0:
@t
@t

(4.8)

Faradays Law

In 1831, Faraday discovered that an electric current was induced along a conductor loop when a
magnetic ux enclosed by the loop changed with time. This important discovery gave an answer
to the old question prior to Faradays time whether a magnetic eld could induce an electric eld
because it had been known that an electric eld, via an electric current, could induce magnetic
eld. What Faraday found was that an electric eld (or electromotive force, emf) was induced by
a time varying magnetic ux. The integral form of the Faradays law,
emf =

E dl =

d
dt

B dS;

(4.9)

was later put into a dierential form by Maxwell,

r

E=

@B
:
@t

(4.10)

The negative sign is due to Lenz and indicates that the emf is so induced as to oppose the change in
the magnetic ux. It should be noted that many of the experiments originally done by Faraday were
actually due to motional emf in which motion of conductors across a magnetic eld was responsible
for generation of emf without apparent time variation of the magnetic eld itself. An object moving
in a magnetic eld experiences an eective electric eld given by
Eef f = v

B:

(4.11)

This may be seen by noting that change in the magnetic ux enclosed by a loop consists of two
parts, one due to time variation of the magnetic eld,
d

@B
dSdt;
@t

and the other due to the change in the shape of the loop,
d

=
=
=

@S
dt
@t

B (v dl)dt
I
(v B) dldt:

Figure 4-1: Induction of emf in time varying magnetic eld (upper gures) and motional emf in
static magnetic eld (lower gure).

Then the total ux change is

d
=
dt

@B
dS
@t

(v

B) dl;

and the emf induced along the loop is given by

emf =

E dl =

@B
dS +
@t

(v

B) dl:

(4.12)

An emf can be generated by letting a conductor move across a stationary magnetic eld as done in
most electric generators.

4.3

Displacement Current and Wave Equation

As shown in Introduction, the displacement current density

"0

@E
;
@t

plays a crucial role for the Maxwells equations to be consistent with the charge conservation law,
@
+ r J = 0:
@t

The magnetic eld induced by the displacement current may be best visualized in a capacitor being
slowly charged as shown in Fig.4-2. The current ows on the surface of electrodes. The radially

Figure 4-2: The surface currents on the capacitor electrode are consistent with the magnetic elds
both outside and inside.

outward surface current on the outer surface of the upper plate,

I

Js1 =

(4.13)

is consistent with the boundary condition for the magnetic eld,

n1 H = Js1 ;
where
H=

I
2

e ;

(4.14)

(4.15)

is the magnetic eld expected from the Amperes law. In the space between the electrodes, no
conduction current exists but there exists an azimuthal magnetic eld,
B ( )=

1
2

0 "0

@Ez
;
@t

(4.16)

as if there were a uniform conduction current equal to

Jz = "0

@Ez
:
@t

(4.17)

The magnetic eld is required to exist to satisfy the boundary condition at the inner electrode
surface,
n2 H = Js2 ;
(4.18)

where

I
;
2 a2

Js2 =

(4.19)

is the radially inward surface conduction current on the inner surface of the upper plate. Since
I=
and
Ez =

dq
;
dt

(4.20)

q
;
"0 a2

(4.21)

@Ez 1
:
@t 2

(4.22)

we nd
H =

"0

This is consistent with the Maxwells equation,

1 @
@Ez
( H ) = "0
;
@
@t

(4.23)

@Ez
:
@t

(4.24)

or its integral form,

2

H =

"0

As is well known, the displacement current was instrumental for Maxwell to predict that electromagnetic elds obey a wave equation. To see what wave equations the electromagnetic elds
should satisfy, let us take a curl of the Faradays law,
r
r

@B
;
@t
@
r
@t

E =

(r

E) =

(4.25)

B:

The LHS can be expanded as

r

(r

E) = r(r E)

r2 E:

Since
r E=

"0

; r

B=

J + "0

@E
@t

Eq. (4.25) reduces to the following inhomogeneous wave equation,

r2

1 @2
c2 @t2

E=

1
r +
"0

@J
:
@t

(4.26)

where c is the speed of light in vacuum,

c2 =

1
"0

:
0

(4.27)

Likewise, the magnetic eld obeys

1 @2
c2 @t2

r2

B=

0r

(4.28)

J:

If the displacement current were absent (which, incidentally, is equivalent to the assumption that
c ! 1), both elds would merely satisfy vector Poissons equation,
r2 E =

1
r +
"0

r2 B =

0r

@J
;
@t
J;

which do not exhibit any propagation nature with a nite speed.

Equation (4.26) can be solved symbolically as
E=
r

1
r +
"0

1
1 @2
c2 @t2

@J
@t

(4.29)

where

1
;
1 @2
2
r
c2 @t2
is the propagator integral operator which yields a retarded solution for the electric eld,
E(r; t) =
=

Z
1
r0
4 "0
Z
1
r
4 "0

(r0 ; t
) 0
dV
0
jr r j
(r0 ; t
) 0
dV
0
jr r j

where
=

jr

@
4 @t
0

r0 j

1
jr
Z

@
r0 j @(t

J(r0 ; t

)dV 0

J(r0 ; t
) 0
dV ;
0
jr r j

In Eq. (4.31),
1
4 "0

(r0 ; t
) 0
dV ; A (r; t) = 0
0
jr r j
4
6

(4.31)

(4.32)

is the time required for electromagnetic disturbances to propagate over a distance jr

the source at r0 and the observer at r: Note that
Z
r0 (r0 ; t
) 0
dV
jr r0 j
Z
(r0 ; t
)
r r0
=
r0
(r0 ; t
) dV 0
jr r0 j
jr r0 j3
Z
(r0 ; t
) 0
= r
dV :
0
jr r j

(r; t) =

(4.30)

J(r0 ; t
) 0
dV ;
0
jr r j

r0 j between

are the retarded scalar and vector potential, respectively. For a single charged particle moving at
a velocity vp (t) ; the scalar potential becomes
(r; t) =
=

Z
e
4 "0
e
4 "0 (1

rp t 1c jr r0 j
dV 0
jr r0 j
1
;
n ) jr rp j ret
r0

(4.33)

where
jret means every time dependent quantity is to be evaluated at the retarded time t0 to be
determined implicitly from
1
t t0
r rp t0 = 0:
c
Likewise, the vector potential due to a moving charge is
A (r; t) =

e
4

(1

v
) jr

rp j

(4.34)

ret

These potentials were rst formulated by Lienard and Wiechert. We will use the potentials in
Chapter 8 in formulating the radiation electromagnetic elds due to moving charges.
Equation (4.26) can also be written as
E =
r2
=
r2

1
1 @2
c2 @t2
1
1 @2
c2 @t2

1
r +
"0

@J
@t

1
r +
"0

@Jl
+
@t

@Jt
@t

(4.35)

where Jl is the longitudinal component of the current density satisfying

@
+ r Jl = 0;
@t

(4.36)

and Jt = J Jl is the transverse component which does not aect the charge density because of
the identity (or denition)
r Jt = 0:
(4.37)
Since r

Jl = 0; it follows
r(r Jl ) = r2 Jl ;

and a symbolic solution for the longitudinal current density is

Jl =

1 @
r :
r2 @t

(4.38)

Substituting into Eq. (4.35), we obtain

E=

1
r +
"0 r2

1
1 @2
c2 @t2

@Jt
:
@t

(4.39)

This formulation is consistent with the choice of Coulomb gauge for the scalar and vector potentials
as we will see in more detail in Chapter 6. Solution for the electric eld is
E(r; t) =

1
r
4 "0

(r0 ; t)
dV 0
jr r0 j

@
4 @t
0

Jt (r0 ; t
) 0
dV :
0
jr r j

(4.40)

Note that in this formulation, the Coulomb electric eld (the rst term in the RHS) is non-retarded
and instantaneous. This unphysical result is a consequence of the Coulomb gauge which singles
out the transverse current Jt . In fact, the non-retarded longitudinal electric led is cancelled by
a term contained in the last term and all physically observable electromagnetic elds are retarded
because of the nite propagation speed c:

4.4

Fields and Potentials

By now, it is clear that there exists two kinds of electric eld, one dened in terms of the scalar
potential,
E1 = r ;
(4.41)
and another originating from time varying magnetic eld,
r

@B
=
@t

E2 =

@
r
@t

A:

(4.42)

The eld E1 is longitudinal because r E1 = 0: In Eq. (4.42), the longitudinal components of

E and A vanish. In fact Maxwells equations do not impose any conditions on the longitudinal
component of A; and choice of r A is arbitrary. Therefore, we are allowed to assume
E =

@
A
@t
@
Al
@t

@
At ;
@t

where Al and At are the longitudinal and transverse component of the vector potential. Substitution in r E = ="0 yields
@
r E = r2
r Al = :
(4.43)
@t
"0
In Coulomb gauge, Al is so chosen as to satisfy
Coulomb gauge: r Al = 0:

(4.44)

(Note that Al may not be zero identically.) Eq. (4.43) becomes

r2

(r; t)
;
"0

(r; t) =

(4.45)

where C is the scalar potential in the Coulomb gauge. It should be emphasized that both C
and are time dependent and in Coulomb gauge the scalar potential responds to a change in the
charge density instantaneously. In Lorenz gauge, r Al is assigned as
Lorenz gauge: r Al +

1 @ L
= 0:
c2 @t

(4.46)

In this case, the scalar potential obeys a wave equation,

1 @2
c2 @t2

r2

"0

(4.47)

This yields a retarded solution,

1
L (r; t) =
4 "0

1
r
c

r0 ; t

r0 j

jr

r0

dV 0 :

(4.48)

Proof is straightforward if the following are noted:

r2
r2

r0 ; t

1
jr

1
r
c

r0 j
r0

r0

1 @2
c2 @t2

;
1
r
c

r0 ; t

r0

For the vector potential, we rewrite the Ampere-Maxwells law

r

B=

J + "0

@E
@t

in terms of the potentials as

r2 A

1 @2A
c2 @t2

r (r A) +

1 @
c2 @t

0 J:

(4.49)

In Coulomb gauge with r A =0; this reduces to

r2 A
r2 A

1 @2A
c2 @t2
1 @2A
c2 @t2

0J

0 Jt ;

1 @
r
c2 @t

since the longitudinal component of the current Jl vanishes through the continuity equation
@
+ r Jl = 0:
@t
9

(4.50)

Note that

@
@
= "0 r E =
@t
@t

and thus
"0

Jl

@
r
@t

"0

@ 2
r
@t

C;

= 0:

In Coulomb gauge, the vector potential is transverse.

In Lorenz gauge, Eq. (4.49) becomes a wave equation for A ;
1 @2A
=
c2 @t2

r2 A

(4.51)

0 J;

where J = Jl + Jt ; A = Al + At : In Lorenz gauge, the potentials are symmetric in the sense that
both and A satisfy the same wave equation,
r2

1 @2
c2 @t2

=c
A

c
J

(4.52)

and appropriately form a covariant four vector ( =c; A) : In contrast, the formulation of electromagnetic elds in terms of Coulomb gauge is not invariant under the Lorentz transformation. All
potentials and elds are retarded in Lorenz gauge while in the Coulomb gauge, the scalar potential
is non-retarded. The appearance of non-retarded scalar potential is due to the choice r A = 0; or
the assumption that the vector potential is purely transverse.

4.5

Poynting Vector: Energy and Momentum Conservation

The complete set of Maxwells equations is

r E=
r

E=

"0

@B
;
@t

r B = 0;
r

B=

J + "0

@E
@t

It is noted that the charge density contains all kinds of charges, free charges, bound charges, etc.
Likewise, the current density J contains all kinds of current, conduction currents, magnetization
currents, etc. If free charges are singled out, the rst equation can be written in terms of the
displacement vector D;
r D = free ;
(4.53)
where
D = "0 E + P = "E;
10

(4.54)

with P being the polarization vector, or the electric dipole moment density. The current density
is likewise decomposed into the part due to the motion of free charges and the other due to time
variation of the polarization vector,
@
r P + r Jp = 0;
@t
from which
Jp =

@P
:
@t

(4.55)

Therefore, the magnetic induction equation can be rewritten as

r

B=

J + "0

@E @P
+
@t
@t

J+

@D
@t

(4.56)

where the current density J consists of conduction and magnetization currents,

J = Jc +Jm = Jc +r

M:

If the conduction current is singled out, Eq. (4.56) can be rewritten in terms of the vector H;
r

H = Jc +

where
H=

@D
;
@t

(4.57)

M:

In macroscopic applications, free charges and conduction currents are the quantities that can be
controlled by external means and Eqs. (4.53) and (4.57) are often more convenient than the original
forms.
The Maxwells equations are also consistent with energy conservation law. To show this, we
dene the Poynting vector by
E H; (W m 2 ):
(4.58)
As its dimensions imply, the Poynting vector indicates the ow of electromagnetic energy per unit
area per unit time, that is, power density. The divergence of the Poynting vector is
r (E

H) = (r E) H (r H) E
@D
@B
=
H
Jc +
E
@t
@t
1
@ 1
=
H 2 + "E 2
E J;
@t 2
2

provided that the permittivity " and permeability

11

are independent of the frequency !: If not, the

energy densities should be modied as

1 @ [!" (!)] 2 1 @ [! (!)] 2
E ;
H ;
2
@!
2
@!

(4.59)

respectively. We will return to this problem shortly in the following Section. Integrating over a
volume, we nd
Z

r (E

H) dV =

(E

d
(Um + Ue )
dt

H) dS =

where
Ue =

1
2

"E 2 dV;

Um =

1
2

E JdV;

H 2 dV;

(4.60)

(4.61)

are the total electric and magnetic energies stored in the volume and
Z
E JdV; (J s 1 = W)

(4.62)

is the rate of electromagnetic energy conversion into other forms of energy, e.g., creation of heat
through Joule dissipation and acceleration of charged particles. Therefore,
I
(E H) dS;
(4.63)
S

can be interpreted as the electromagnetic power ow into the volume.

For a system consisting of charged particles, the momentum conservation can be shown in a
similar way. The mechanical momentum Pm follows the equation of motion,
dPm
=
dt

( E+J

B) dV; (N).

(4.64)

Substituting
= "0 r E
and
J=

1
0

"0

@E
;
@t

we nd
dPm
dt

=
=

where

Z
Z

"0 (r E)E +

!
r T

!
T = Tij = "0 Ei Ej

1 @
E
c2 @t

1
"0 E 2
2

ij

"0

@E
@t

B dV

H dV;

1
0

12

Bi Bj

(4.65)

1
2

B2
0

ij ;

(N m

(4.66)

is the Maxwells stress tensor and use is made of the following vector identities,
rE 2 = r(E E) = 2E (r

E) + 2 (E r) E;

r (EE) = (r E)E + (E r) E
rB 2 = r(B B) = 2B

(r

B) + 2 (B r) B

r (BB) = (r B)B + (B r) B = (B r) B:
Eq. (4.65) suggests that the vector
1
E
c2

H; (N s m

(4.67)

can be regarded as the electromagnetic momentum density and

1
E
c

H; (N m

(4.68)

as the electromagnetic momentum ux density. For a system of charged particles, momentum

conservation thus requires inclusion of the electromagnetic momentum as well as mechanical momentum.
From the momentum density in Eq. (4.67), the angular momentum density associated with
electromagnetic elds is naturally dened by
1
r
c2

(E

H); (J s m

(4.69)

and the total angular momentum associated with electromagnetic elds by

Z
1
L = 2 r (E H)dV ; (J s).
c

(4.70)

Its ux density is
1
R= r
c

(E

H); (J m

):

(4.71)

Both the momentum and angular momentum densities are proportional to the Poynting vector,
namely, the energy ow. If for example a system is losing energy through radiation of electromagnetic energy, the system is necessarily losing momentum and angular momentum as well. Consider
a charged particle undergoing circular motion, e.g., electron in a magnetic eld. Since the electron is continuously accelerated by the centripetal force, it radiates electromagnetic energy. At
the same time it loses its angular momentum to radiation. Therefore, it is natural to expect that
electromagnetic elds (with proper polarization) carry an angular momentum with them.

13

4.6

Plane Electromagnetic Waves

One special mode of electromagnetic waves in free space is a plane wave in which the amplitude
of electric and magnetic eld remains constant. Without loss of generality, we may assume wave
propagation in the z direction and an electric eld in the x direction,
E(z; t) = E0 ei(kz

!t)

ex :

In free space, the electric eld satises the wave equation,

@2
@z 2

1 @2
c2 @t2

provided

E(z; t) = 0;

(4.72)

!
= c:
k

From
r

E=

(4.73)
@B
;
@t

we nd
E = !B;

(4.74)

where k = kez : The magnetic eld is thus in the y direction,

B=

k
E0 ei(kz
!

!t)

ey ;

(4.75)

and its amplitude is

B0 =

E0
E0
E0
; or H0 =
;
=
c
c 0
Z

where
Z=

"0

(4.76)

= 376:8; ( )

(4.77)

is the impedance of free space.

The electric and magnetic energy densities associated with a plane wave are the same, for
1
1
1
"0 E 2 = "0 Z 2 H 2 =
2
2
2

0H

(4.78)

This equipartition of wave energy is similar to that in mechanical waves in which kinetic and
potential energies are equal. The total wave energy density is therefore given by
u=2

1
"0 E 2 = "0 E 2 ; (J m
2

14

(4.79)

and the Poynting ux may be written in terms of either the electric or magnetic eld as
Sz = Ex Hy = c"0 Ex2 =

Ex2
; (W m
Z

(4.80)

or
Sz = c

2
0 Hy

= ZHy2 :

(4.81)

For a harmonic wave with an amplitude E0 , the average (rms) wave energy density is given by
1
uave = "0 E02 ; (W m
2

(4.82)

and corresponding rms Poynting ux is

1
1 E02
Szave = c"0 E02 =
; (W m
2
2 Z

):

(4.83)

Electromagnetic waves radiated by a localized source approach plane waves at a su ciently

large distance but they can never be pure plane waves. Plane waves can be constructed from two
circularly polarized waves with opposite helicities, one rotating with positive helicity and another
with negative helicity. Helicity of electromagnetic waves is related to the angular momentum
associated with the waves. Evidently, a plane polarized wave carry zero angular momentum. A
more general theory of electromagnetic radiation will be developed in Chapter 5.
In a dielectric medium, the Maxwells equations are modied as
r
r

@B
;
@t

E=
B=

@("E)
;
@t

(4.84)

where " is the permittivity which in general depends on the wave frequency and spatial position
and also the electric eld. The origin of the permittivity is in the current induced by the electric
eld in a material medium. In the magnetic induction equation
r

B=

J + "0

@E
@t

if the current density is proportional to the electric eld through a conductivity ; J = E; we have
r

B=

E + "0

@E
@t

0"

@E
;
@t

(4.85)

where the permittivity is given by

" = "0 1 + i

!"0

(4.86)

It should be noted that the current density is due to deviation of electron orbit from bound harmonic

15

motion in molecules. The equation of motion for an electron placed in an oscillating electric eld is
m

d2
+ ! 20 x =
dt2

eE0 e

i!t

(4.87)

where m is the electron mass and ! 0 is the frequency of bound harmonic motion. The current
density is
dx
i! eE
J = ne
= 2
;
dt
!
! 20 m
and the conductivity becomes
=

i!
!2

ne2
;
! 20 m

(4.88)

where n is the number density of electrons. Then the permittivity is given by

"(!) = "0

! 2p

! 20

!2

(4.89)

where ! p is the plasma frequency dened by

! 2p =

ne2
:
"0 m

(4.90)

The phase velocity of electromagnetic waves in a dielectric is

and the group velocity is

!
1
=p
k
"(!)
d!
=
dk

(4.91)

!
1
:
1 ! d" k
1+
2 "(!) d!

(4.92)

(At frequencies remotely separated from the resonance frequency ! 0 ; the group velocity coincides
with the energy propagation velocity. Near the resonance, however, the group velocity exceeds c and
it loses the meaning of energy propagation velocity. The concept of signal velocity was introduced
by Brillouin.) The impedance is accordingly modied as
E
Z(!) =
=
H

"(!)

By denition, the group velocity is equal to

d!
Poynting ux
=
;
dk
Energy density

16

(4.93)

where the Poynting ux is

E2
S=
=
Z

"(!)

E 2 ; (W m

):

Therefore the wave energy density in a dielectric medium is

u =
=
=

1+

1 ! d"
2 "(!) d!

"(!)E 2

1 d
1
[!"(!)]E 2 +
H2
2 d!
2 0
1
1
1 ! 2p (! 2 + ! 20 )
"0 E 2 +
"0 E 2 +
2
2
2
2 (!
!0)
2

0H

; (J m

where the relationship

1
1
"(!)E 2 =
2
2

0H

(4.94)

is used. This result is valid only if the group velocity can be regarded as energy propagation velocity
which may not hold near the resonance ! ' ! 0 if the dielectric is dissipative.
The origin of the additional factor
1 d" 2
! E ;
2 d!
in the electric energy density is due to electron kinetic and potential energies in an oscillating
electric eld. From the equation of motion of a bound electron,
@2
+ ! 20 x =
@t2

e
E;
m

we readily nd the electron kinetic energy density,

1
1 ne2
!2
1 ! 2 ! 2p
2
nmv 2 =
E
=
"0 E 2 ;
2
2 m (! 2 ! 0 )2
2 (! 2 ! 0 )2

(4.95)

and potential energy density

1 ! 20 ! 2p
1
nm! 20 x2 =
"0 E 2 :
2
2 (! 2 ! 0 )2

(4.96)

Therefore, the total energy density associated with the electric eld is
1
1 ! 2 ! 2p
1 ! 20 ! 2p
2
"0 E 2 +
"
E
+
"0 E 2 ; (J m
0
2
2 (! 2 ! 0 )2
2 (! 2 ! 0 )2

(4.97)

which is consistent with that conveniently calculated from

1 d
[!"(!)]E 2 :
2 d!

(4.98)

Since the wave under consideration is strongly dispersive, there is no simple energy equipartition
as in the case of nondispersive waves. Note that the electron kinetic and potential energies are the

17

result of forced oscillations.

The fact that
E

@D
@ ("E)
=E
;
@t
@t

(4.99)

is not always equal to

1 @E 2
"
;
2 @t

(4.100)

can be seen if we realize that the permittivity " is in general frequency dependent. Waves become
dispersive and wave amplitude is bound to decrease although slowly. In this case,
@ ("E)
@t

i!" (!) E0 e

i!t

' ( i! 0 + ) " (! 0 ) + i
=
where

i! 0 " (! 0 ) E0 e

i!t

@"
E0 e
@! 0

+ " (! 0 ) + ! 0

i!t

@" dE0
e
@! 0 dt

i!t

(4.101)

(< 0) is the damping rate of the amplitude,

dE0
= E0 :
dt

(4.102)

Therefore, time variation of electric energy density should be calculated from

E

@ ("E)
1
@" dE 2
=
" (! 0 ) + ! 0
;
@t
2
@! 0 dt

(4.103)

and the electric energy density becomes

ue =

1
@"
" (! 0 ) + ! 0
E2:
2
@! 0

(4.104)

Likewise, the magnetic energy density should be generalized as

um =

1
2

(! 0 ) + ! 0

@
H 2:
@! 0

(4.105)

These expressions were originally formulated by von Laue.

4.7

Wave Reection and Transmission - Normal Incidence

Reection and transmission of plane electromagnetic waves at a boundary of two dielectrics can be
conveniently analyzed in terms of impedance mismatch. Let a plane wave of amplitude Ei in air be
incident normally on a at dielectric surface of impedance Z2 as shown in Fig.4-3: The impedance
of air is very close to the vacuum impedance, Z1 = 377 : The incident Poynting ux is split into

18

those of reected (Er ) and transmitted (Et ) waves,

E2 E2
Ei2
= r + t ; (W m
Z1
Z1
Z2

):

(4.106)

At the boundary, continuity of the electric elds, which are all tangential to the surface, yields

Figure 4-3: A plane wave incident normal to a dielectric boundary.

Ei + Er = Et :

(4.107)

Solving these equations for Er and Et ; we nd

Er =

Z2 Z1
Ei;
Z2 + Z1

(4.108)

Et =

2Z2
Ei :
Z2 + Z1

(4.109)

In terms of the magnetic elds,

Z1 Hi2 = Z1 Hr2 + Z2 Ht2 ;

(4.110)

Hi + Hr = Ht ;

(4.111)

yield
Hr =

Z1 Z2
Hi;
Z1 + Z2

Ht =

2Z1
Hi :
Z1 + Z2

(4.112)

Note that the polarity of either electric or magnetic eld of the reected wave is reversed. This
is understandable because the incident and reected elds must satisfy the following vectorial
relationships,
k Ei = ! 0 Hi ;

19

Er = !

0 Hr :

Either Er or Hr must change sign on reection for the reected Poynting ux to be in opposite
direction relative to the incident ux.
It should be pointed out that the analysis presented above is solely based on energy conservation and no consideration was given to momentum conservation. In fact, if only the momenta
associated with the three waves (incident, reected, and transmitted) are considered, momentum
is not conserved. Since the momentum ux densities associated with each wave are
1 Ei2
= "1 Ei2 ;
c1 Z1
Reected wave : "1 Er2 ;
Incident wave :

(4.113)

Transmitted : "2 Et2 ;

the following momentum unbalance emerges,
"1 (Ei2 + Er2 )

"2 Et2 =

2(Z12 Z22 )
2(Z1 Z2 )
"1 Ei2 =
"1 Ei2 :
2
(Z1 + Z2 )
Z1 + Z2

(4.114)

This is actually taken up by the dielectric body as mechanical momentum. Recall that an innitely
massive body can absorb momentum without absorbing energy. If Z2 > Z1 ; the dielectric is pushed
to the left. In fact, at the boundary, the electric energy densities are discontinuous, and the
dierence is
1
"2 Et2
2

1
2Z22
"1 Et2 = ("2 "1 )
E2
2
(Z1 + Z2 )2 i
2(Z12 Z22 )
=
"1 Ei2
(Z1 + Z2 )2
2(Z1 Z2 )
=
"1 Ei2 :
Z1 + Z2

This appears as a force per unit area on the boundary surface acting from the higher energy density
side to the lower because electric force in the direction perpendicular to the eld appears as pressure.
Reection at a conductor surface can be analyzed in a similar manner by modifying the impedance appropriately. For a medium having a conductivity ; the impedance is given by
Z=

i! 0
:
i!"0

This can be seen from the Maxwells equation in the presence of conduction current,
r

@E
@t
i!"0 E;

H = J + "0
=

20

(4.115)

and the eective permittivity in a conductor is dened by

"e = "0

i!

Therefore, the impedance in a conductor is

Z=

"e

i! 0
:
i!"0

For ordinary conductors, the conduction current far dominates over the displacement current even
in microwave frequency range. Then,
Z'

i!

= (1

r
!
i)
2

(4.116)

A complex impedance indicates strong dissipation of electromagnetic energy. The magnitude of

the impedance
r
! 0
jZj =
;
(4.117)

p
and electromagnetic waves
is much smaller than the free space impedance Z0 =
0 ="0 ' 377
incident on a conductor surface suers strong reection. However, reection can never be complete.
Bath room mirrors coated with aluminum has power reection coe cient of about 90% at optical
frequency ! ' 1014 rad/sec.
Example 1 Impedance Matching
A dielectric lm a quarter wavelength thick coated on a surface of another dielectric (say, optical
glass) can eliminate reection of electromagnetic waves normally incident if the impedance of the
lm is chosen to be a geometric mean,
Zf =

p
Z0 Zg :

(4.118)

This condition follows from cancellation between two reected waves, one at the air-lm boundary
and another at the lm-glass boundary,
Zf Z0
Zg Zf
+ ei
= 0;
Zf + Z0
Zg + Zf
f

where the phase factor ei = 1 is due to additional propagation distance

2=
of the wave
4
2
reected at the lm-glass boundary. Here, f is the wavelength in the dielectric lm. Solving
Zf Z0
Zg Zf
=
;
Zf + Z0
Zg + Zf

21

for Zf ; we nd

p
Z0 Zg :

Zf =

Likewise, reection from a conductor plate can be avoided by placing a thin conducting plate
of thickness d at a quarter wavelength in front of the conductor surface if the conductance of the
plate is chosen to be
1
=
:
(4.119)
Z0 d

4.8

Reection and Transmission at Arbitrary Incident Angle

For an arbitrary incident angle 1 ; reection and transmission at a dielectric boundary can be
analyzed by exploiting the boundary conditions for the electric and magnetic elds as follows. An
incident wave continues to be assumed plane polarized. The refracted angle 2 is related to the
incident angle 1 through the well known Snells law,
sin
sin

n2
;
n1

(4.120)

p
p
where n1 = "1 ="0 and n2 = "2 ="0 are the indices of refraction of respective media. The Snells
law follows from the conservation of the wavenumber parallel to the boundary,
k1 sin

= k2 sin

2;

(4.121)

and the change in the wave propagation velocity,

!
1
=p
k1
"1
or

!
1
=p
k2
"2

;
0

(4.122)

k2
k1
=
:
n1
n2

(4.123)

The normal wavenumber in the medium 2 is

kz = k2 cos

= k2

n1
n2

sin2

1:

This becomes pure imaginary when

sin

>

n2
; n2 < n1 ;
n1

This is the condition for total reection which occurs when a wave is incident on a medium with a
smaller index of refraction (e.g., from glass to air), n2 < n1 : An imaginary wavenumber indicates
exponential damping in the region n2 from the surface,
E0 e

kz z

22

where kz is the damping factor in the axial (z) direction,

kz = k2

n1
n2

sin2

1:

The wavenumber component along the surface is

kk = k2

n1
sin :
n2

If the second medium is vacuum (or air) n2 = 1; k2 = !=c = k0 ; and

kk = k0 n1 sin :
In total reection, electromagnetic elds in the region of smaller index of refraction exponentially
decay from the surface. Such waves are called evanescent. Important applications of evanescent
waves are being found in high resolution microscopy even with visible light.
In analysis of wave reection and transmission at a boundary between dierent media, use of
the continuity of tangential electric and magnetic elds, Et ; Ht ; is su cient because continuity of
normal components Dn ; Bn is redundant. From Maxwells equation,
r

H=

@D
; or k
@t

H=

!D;

it is evident that continuity of Dn demands continuity of n (k H) = k (n H). However, the

tangential component of H; n H; is continuous and the wavenumber parallel to the boundary is
also continuous. (Note that in k (n H) = (k n) H; only the parallel component of k appears.)
The latter (continuity of k n) follows from the conservation of wave momentum parallel to the
boundary which manifests itself in the form of well known Snells law. Likewise, from
r

E=

@B
; or k
@t

E = !B;

we see that continuity of Bn automatically follows because of continuity of tangential component

of the electric eld.

4.8.1

H in the Incident Plane (E tangential to the boundary)

We rst consider the case in which the magnetic eld of the incident wave is in the incident plane,
or the electric eld is parallel to the boundary surface. The reected and refracted electric elds
are also parallel to the surface and the continuity of the tangential component of the electric eld
gives
Ei + Er = Et :
(4.124)

23

Figure 4-4: A plane wave with incident angle 1 when the magnetic eld is in the incidence plane
and the electric eld is parallel to the boundary surface.

Since for each wave, the following vectorial relationship holds,

E=

0 !H;

(4.125)

the continuity of tangential component of the H eld yields

k1 cos

1 (Ei

Er ) = k2 cos

2 Et :

(4.126)

Recalling the Snells law, we thus nd

Er =

sin(
sin(

Et =

2 cos
sin(

In the limit of normal incidence (small angles

Er =

sin(
sin(

1)

2
1

2)

Ei '

sin
sin

1)

1+

2)

1 sin 2
1

2)

1 ; 2 ),

sin
1 + sin
2

where
Zi =

Ei ;

Ei :

(4.128)

we recover

Ei =

"i

(4.127)

n 1 n2
Z2 Z1
Ei =
Ei ;
n2 + n1
Z2 + Z1

(4.129)

(4.130)

is the impedance of respective media.

When the conditions for total reection are met, the impedance dened by the ratio between

24

the tangential component of the electric and magnetic eld,

Z2TE =

E2
H2 cos

r
i

=
2

"2

n1
n2

sin2

(4.131)

is also pure imaginary which indeed ensures total reection of electromagnetic waves,
Z1
i Z2TE
= 1:
TE
Z1 + i Z2
In total reection, the phase dierence between the incident and reected waves is
TE

4.8.2

= 2 tan

Z1
Z2TE

= 2 tan

sin2

(n2 =n1 )2

cos

(4.132)

E in the Incident Plane (H tangential to the boundary)

Figure 4-5: The electric eld is in the incidence plane and the magnetioc eld is paralllel to the
boundary.

In this case, the magnetic eld is parallel to the boundary plane and the continuity of tangential
component of the magnetic eld is simply
Hi + Hr = Ht :

(4.133)

For each wave, the electric eld is related to the magnetic eld through
E=

1
k
!"
25

H:

(4.134)

Therefore, continuity of the tangential component of the electric eld yields

1
cos
n1

1 (Hi

1
cos
n2

Hr ) =

2 Ht

1
cos
n2

2 (Hi

+ Hr ):

(4.135)

Solving for Hr ; we nd
Hr =
=

sin 1 cos
sin 1 cos
tan( 1
tan ( 1 +

sin
+
sin
1
2)
Hi :
2)

2 cos 2
2 cos 2

Hi

(4.136)
(4.137)

The transmitted magnetic eld is

Ht =

1+

tan(
tan (

2)

1+

2)

The reected electric eld is

Er =

Hi =

tan(
tan (

1)

2
1

2 sin 1 cos
1 + 2 ) cos(

sin(

2)

1
1

2)

Hi :

Ei ;

(4.138)

and transmitted electric eld is

Et =

2 sin 2 cos
sin( 1 + 2 ) cos(

1
2)

Ei :

(4.139)

If
1

(4.140)

the reected wave vanishes completely. Under this condition, Snells law becomes
n2
sin
=
n1
sin

sin
cos

= tan

1;

(4.141)

and this particular angle

B

= tan

n2
n1

(4.142)

is called the Brewsters angle. If incident wave is plane polarized with the magnetic eld oriented
parallel to, say, a surface of glass, reection can be avoided at the Brewsters angle. This principle
is often exploited in designing reecting mirrors in lasers so that output laser beam has a high
degree of planar polarization.
Because the reection coe cient depends on wave polarization, unpolarized wave with random
polarization becomes partially polarized on reection and transmission at a dielectric boundary.
The magnetic reection coe cient derived in Eq. (4.136) will play an important role in analyzing
the transition radiation discussed in Chapter 8.
When the conditions for total reection are met, the impedance dened by the ratio between

26

the tangential component of the electric and magnetic eld is

Z2TM

E2 cos
=
H2

r
i

"2

n1
n2

sin2

1:

(4.143)

The phase dierence between the incident and reected waves is

TM

4.9

= 2 tan

Z1
Z2TM

= 2 tan

!
p
sin2 1 (n2 =n1 )2
:
(n2 =n1 )2 cos 1

(4.144)

Circularly and Elliptically Polarized Plane Waves

Planar polarization discussed in the preceding section is a highly idealized mode of propagation of
electromagnetic waves. A plane polarized wave can be decomposed into two circularly polarized
plane waves of opposite helicity, one with positive helicity and another with negative helicity.
Helicity of an electromagnetic wave is closely related with the angular momentum carried by the
wave.
Circularly polarized waves propagating in the z-direction can be described by the electric eld
vectors,
E (z; t) = E0 (ex iey )ei(kz !t) ;
(4.145)
where the positive sign is for positive helicity and minus sign is for negative helicity. The sum of
these two waves of opposite helicity trivially yields a plane polarized wave with the electric eld in
the x-direction. Corresponding magnetic elds are
B

=
=

1
k E
!
k
E0 (ey
!

iex )ei(kz

!t)

(4.146)

and Poynting ux is
S=E

H =2

E02
ez ;
Z

(4.147)

where the factor 2 accounts for the two independent modes of equal amplitude.
A general form of mixed helicity may be written as
E(z; t) = (E1 ex + E2 iey )ei(kz

!t)

(4.148)

where E1 and E2 are complex amplitude. Corresponding magnetic eld,

B(z; t) =

k
(E1 ey
!

E2 iex )ei(kz

!t)

(4.149)

is of course normal to the electric eld,

E B = 0:

27

(4.150)

However, the scalar product E B in general does not vanish.

Figure 4-6: Trace of the head of the rotating electric eld vector associated with a circularly
polarized wave with positive helicity. In the case of negative helicity, the direction of rotation
relative to the wavevector is reversed.

Example 2 Reection and Transmission of Circularly Polarized Wave

Consider a circularly polarized wave incident at an angle i to a at surface of a dielectric. The
electric eld of the incident wave may be decomposed into two components, one in the incident
plane and another perpendicular to the incident plane,
Ei = Ek0 + iE?0 :
The reected wave of the parallel component is
Ekr =

tan(
tan(

1)

2
1

Ek0 ;

2)

while the reected perpendicular component is

sin(
sin(

E?r =

1)

2
2

1)

E?0 :

The transmitted (refracted) components are

Ekt =

2 sin 1 cos
sin( 1 + 2 ) cos(

E?t =

2 cos
sin(

2
1

1 sin 2
1

2)

2)

Ek0 ;

E?0 :

Both reected and transmitted waves are elliptically polarized. In particular, if the incident angle is
at the Brewsters angle, 1 = =2 2 ; Ekr vanishes and the reected wave becomes plane polarized.
28

On total reection of a circularly polarized electromagnetic wave, the reected wave becomes
elliptically polarized because the phases of TE and TM components dier as evident from Eqs.
(4.132) and (4.144).

4.10

StokesParameters

Consider an electromagnetic wave propagating in the z direction with electric eld components
E(r; t) = (Ex ex + Ey ey: )ei(kz

!t)

(4.151)

The amplitudes Ex and Ey may be complex allowing for nite phase dierence,
Ex = jEx j ei x ; Ey = jEy j ei y :

(4.152)

is nite,
If = x
y = 0; the led is a simple superposition of two linearly polarized waves. If
the wave is in general elliptically polarized. In optics, direct measurement of the phase dierence
is not easy. What is normally measured is the intensity or the quadratic quantities of the electric
eld,
jEx j2 ; jEy j2 ; jEx j jEy j cos( x
(4.153)
y ); and jEx j jEy j sin( x
y ):
Stokesparameters are dened, with I = jEx j2 + jEy j2 the total intensity, by
s0 =
s1 =
s2 =
s3 =

1
I
1
I
2
I
2
I

jEx j2 + jEy j2 = 1;
jEx j2

jEy j2 ;

jEx j jEy j cos(

jEx j jEy j sin(

x
x

(4.154)
y );
y );

and satisfy
s21 + s22 + s23 = 1;

(4.155)

if the waves are purely coherent. If not, s2 and s3 are to be modied as

s2 =

2
2
Re Ex Ey ; s3 = Im Ex Ey ;
I
I

(4.156)

where the bar indicates time average. For incoherent waves,

s21 + s22 + s23 < 1:

(4.157)

Natural light is characterized by a collection of many waves with random phases and complete
depolarization, s1 = s2 = s3 = 0 even though each wave may be highly monochromatic. For a
plane wave polarized in the x direction,
s1 = 1; s2 = s3 = 0:
29

For a plane wave polarized in the direction

= =4 (along the plane x = y),

s1 = s3 = 0; s2 = 1:
For a circularly polarized wave with positive (negative) helicity,
s1 = s2 = 0; s3 =

1:

In experiments, Stokes parameters can be determined by rotating a polarizer plate. jEx j2

and jEy j2 can be readily found by aligning the polarization direction along x and y direction,
respectively. If the polarizer is at angle from the x axis, the intensity measured at that angle is
I( ) = (jEx jei

cos + jEy jei

sin ) jEx je

= jEx j2 cos2 + jEy j2 sin2 + 2 jEx j jEy j cos sin cos(

By choosing

cos + jEy je

sin
y ):

= =4; for example, we have

I

1
jEx j2 + jEy j2 + 2jEx jjEy j cos(
2

y)

(4.158)

Finally, exploiting a uniaxial (or biaxial) crystal, a so-called quarter wavelength plate can be fabricated to induce =2 relative phase delay between Ex and Ey due to the dierent propagation
velocities of ordinary and extraordinary modes. The intensity in this case is
I 0 ( ) = (jEx jei

cos + ijEy jei

sin ) jEx je

= jEx j2 cos2 + jEy j2 sin2

If

cos

ijEy je

2 jEx j jEy j cos sin sin(

sin

y)

= =4; this reduces to

I0

Therefore, jEx jei x and jEy jei

suring the four intensities,

1
jEx j2 + jEy j2
2

2jEx jjEy j sin(

y)

(4.159)

and corresponding Stokesparameters can be determined by mea-

I ( = 0) ; I

; I

and I 0

The tensor dened by

Iij = Ei Ej ; (i; j = x; y)

(4.160)

is called polarization tensor. Unless the eld is purely coherent, the time averaged intensity Ei Ej
may still vary slowly with time. Iik is Hermitian and thus can be diagonalized through the eigen-

30

values

and

which are the roots of

det (Iij

ij )

= 0:

(4.161)

Corresponding two polarization eigenvectors n(1) and n(2) can be determined from
(1)

Iij nj

(1)
1 ni ;

(2)

Iij nj

(2)
1 ni :

(4.162)

In terms of the eigenvectors, the polarization tensor can be written in the form
!
I =

(1) (1)
1n n

(2) (2)
2n n :

(4.163)

As an example, let us consider superposition of two plane polarized waves, one in the x direction
and another in the direction (cos ; sin ) in the x y plane. The relative phase between the two
waves is assumed to be random and the intensities of the waves are I1 and I2 : The electric eld is
E = (E1 + E2 ei cos ; E2 ei sin );
where I1 = E12 ; I2 = E22 ; and

is random phase. Then

I1 + I2 cos2
I2 cos sin

Iij =
Eigenvalues are
1;2

1
=
2

min =

I1 + I2

max

I2 cos sin
I2 sin2

q
(I1 + I2 )2

The ratio
=

(4.164)

I1 + I2
I1 + I2 +

p
p

(4.165)

4I1 I2 sin2

(I1 + I2 )2

4I1 I2 sin2

(I1 + I2 )2

4I1 I2 sin2

(4.166)

(4.167)

may be called the degree of depolarization.

= 1 corresponds to completely unpolarized state,
while = 0 corresponds to a plane polarized wave. When I1 = I2 = I; the polarization tensor
reduces to
!
1 + cos2
cos sin
Iij = I
:
(4.168)
cos sin
sin2
Eigenvalues are
1;2

=1

cos ;

(4.169)

and the degree of depolarization is

=

1 cos
:
1 + cos

(4.170)

The eigenvectors are

n(1) =

cos ; sin
2
2

; n(2) =

31

sin ; cos
2
2

(4.171)

Derivation of the eigenvectors is left for exercise.

The Stokesparameters si and the polarization tensor are related through
1
Iij = I
2

ij

3
X

m=1

where
(1)
ij

(m)
sm ij

1
0

0
1

(2)
ij

1 + s1 s2
s2 + is3 1

1
= I
2

0 1
1 0

(3)
ij

0
i

is3
s1

i
0

(4.172)

(4.173)

are Paulis spin matrices. When s1 = 1; s2 = s3 = 0; Iij reduces to

Iij = I

1 0
0 0

which describes a plane polarized wave in the x direction. When s2 = 1; s1 = s3 = 0;

1 1
1 1

1
Iij = I
2

This describes the case of plane polarization in the direction

1
Iij = I
2

1
i

i
1

= =4: When s3 = 1; s1 = s2 = 0;

which describes circular polarization with positive helicity. Finally, s3 =

corresponding
!
1 i
1
Iij = I
;
2
i 1

1; s1 = s2 = 0 and

describes circular polarization with negative helicity.

4.11

Propagation along a Conductor Rod

In this section, we analyze propagation of electromagnetic waves along a conductor rod of radius
a and conductivity : A simple transverse Magnetic (TM) mode is considered with the following
eld components, E ; Ez ; and H : The axial electric eld Ez outside the conductor rod satises the
scalar wave equation,
@2
1 @
@2
+
+
@ 2
@
@z 2

Ez ( ; z; t) =

32

1 @2
Ez ( ; z; t):
c2 @t2

(4.174)

Figure 4-7: Field proles of electromagentic wave propagating along a conductor rod.

For a harmonic wave with time dependence e

i!t ;

this reduces to

@2
1 @
@2
!2
+
+
+
@ 2
@
@z 2
c2

Ez ( ; z) = 0:

(4.175)

Furthermore, since the wave is propagating along the rod, we may single out the z dependence in
the form eikz and reduce the wave equation to
d2
1 d
+
+
2
d
d
where
2

!
c

Ez ( ) = 0;

k2 :

(4.177)

Elementary solutions to Eq. (4.176) are the Bessel functions J0 (

binations,
(1)
H0 ( ) = J0 ( ) + iN0 ( );
(2)

H0 (

) = J0 (

(4.176)

iN0 (

);

); N0 (

) and their linear com(4.178)

(4.179)

known as Hankel functions of the rst and second kind, respectively. Their asymptotic behavior at
large argument is
r
h
i
2
(1)
H0 (x) !
exp i x
;
x
4
r
h
i
2
(2)
H0 (x) !
exp i x
;
x
4

indicating radially outward and inward propagation, respectively. In the present case, both k
and are complex because of dissipation in the rod. It is reasonable to assume radially outward

33

propagation (radiation from the rod) and we choose

(1)

Ez ( ) = E0 H0 (

):

(4.180)

From r E = 0; we nd the radial component of the electric eld,

ik

E ( )=
and from r

E = i!

0 H;

);

(4.181)

the azimuthal component of the magnetic eld,

H ( )=

where use is made of

(1)

E0 H1 (

i!"0

E0

(1)

H1 (

);

(4.182)

d
(1)
(1)
xH1 (x) = xH0 (x);
dx
d (1)
(1)
H (x) = H1 (x):
dx 0

The boundary condition at the rod surface is

Ez ( = a)
=
H ( = a)
or

Z=

i!

(surface impedance);

(4.183)

(1)

H0 ( a)
(1)

H1 ( a)

= i!"0 Z:

(4.184)

For a given wave frequency !; this equation determines and thus k; the axial wavenumber and
in this respect, it is a dispersion relation.
For a small impedance Z; approaches 0, and the axial wavenumber k becomes
k=

!
;
c

as expected. Noting
2
(1)
lim H0 (x) ! 1 + i (ln(x=2) +

x!0

E) ;

(1)

lim H1 (x) !

x!0

21
;
x

(4.185)

we see that the axial electric eld becomes negligible and the transverse elds approach those of
TEM mode,
E ( )
a
E ( ) = E0 ; H ( ) =
;
(4.186)
Z0
p
where Z0 =
0 ="0 is the impedance of free space. The elds in coaxial cables used for transmission
of electromagnetic waves can be approximated by those given above,
E =

V
1
;
ln(b=a)

H =

34

E
I
=
;
Z
2

(4.187)

where V is the potential dierence between the inner and outer conductor with radii a and
p
b;respectively, I is the current and Z =
0 =" with " the permittivity of the insulating material lling the cable. The characteristic impedance of the cable is
Zcable

V
=
=
I

0 ="

b
a

ln

(4.188)

Similarly, the impedance of a parallel wire transmission line with conductor radius a and separation
distance D is approximately given by
Z'

4.12

0 ="0

D
a

ln

a:

(4.189)

Skin Eects in Conductors

In a conductor, the conduction current dominates over the displacement current and a simple Ohms
law,
J = E;
(4.190)
may be assumed where
Maxwells equations

(S/m) is the conductivity. Eliminating the electric eld between the

r

@B
; r
@t

E=

B'

E;

yields the following diusion equation for the magnetic eld,

r2 B =

@B
:
@t

If the eld is oscillating at a frequency !; we obtain the following ordinary dierential equation,
r2 B + i!

B = 0:

Penetration of the magnetic eld into a conductor slab can be described by

d2 By
+ i!
dz 2

By = 0:

(4.191)

This has a bounded solution

B(z) = B0 eikz ;
where
k=

i!

1 + ip
= p
!
2

(4.192)

The magnetic eld decays exponentially from the conductor surface and the quantity
=

2
!

35

; (m)

(4.193)

is called the skin depth. Damping of the electromagnetic elds is evidently due to Ohmic dissipation
in the conductor.
For a cylindrical conductor rod with radius a( c=!) in which an axial current Jz ( )e i!t is
excited, Jz obeys
1 d
d2
+
+ i! 0
Jz ( ) = 0;
(4.194)
2
d
d
This has a bounded solution
Jz ( ) = E0

J0 (k )
;
J0 (ka)

(4.195)

where E0 is the electric eld at the rod surface. The impedance per unit length of the rod can thus
be dened by
Z
E0
k J0 (ka)
=
=
; (Ohms/m):
(4.196)
l
I
2 a J1 (ka)
where use is made of
I=2

a
J1 (ka)
E0
:
k
J0 (ka)

Jz ( ) d = 2

In the low frequency limit jkj a

(4.197)

1; series expansion of the Bessel functions can be exploited,

J0 (x) ' 1

x2
x
; J1 (x) '
4
2

x2
8

The impedance reduces to

Z
1
=
l
a2
where

i!

(4.198)

Li
= 0 (H/m)
l
8

(4.199)

is the well known internal inductance of a cylindrical rod carrying a uniform current (no skin eect).
In the high frequency limit (strong skin eect), the ratio J0 (ka)=J1 (ka) approaches i; and the
impedance is
Z
1
=
(1 i):
(4.200)
l
2 a
This is also reasonable since with strong skin eect, the current ow is limited in a thin layer with
an area 2 a : Note that the inductive reactance is identical to the resistance in this limit.
If a conductor cylinder is placed in an oscillating axial magnetic eld Hz0 e i!t as in inductive
rf heating, the current ows in the azimuthal direction. J ( ) satises
1 d
d2
+
2
d
d

1
2

+ i!

J ( ) = 0;

< a:

An appropriate solution is
J ( )=k

J1 (k )
Hz0 ;
J0 (ka)

36

< a;

(4.201)

where k is still given by

k=

4.13

i!

Skin Eect in a Plasma

A charge neutral plasma contains equal amount of positive and negative charge density. When
placed in an electric eld oscillating at a high frequency, a plasma current due to electron motion
is induced. The conductivity can be found by letting ! 0 = 0 in Eq. (4.88) (because in a plasma,
electrons are free),
i ne2
=
;
(4.202)
! m
and corresponding permittivity
" = "0

! 2p
!2

In a collisional plasma with electron collision frequency

" = "0

:
c;

! 2p
! (! + i c )

(4.203)
this is modied as
!

(4.204)

as can be easily worked out by introducing a nite collision frequency in the equation of motion.
For a frequency much smaller than the plasma frequency, the permittivity becomes negative,
"'

"0

! 2p
;
!2

(4.205)

and wave propagation is forbidden. The wavenumber is complex in this case

k=i

!p
;
c

(4.206)

and the wave amplitude decays in a manner

E0 ei(kz

!t)

= E0 exp

!p
z e
c

i!t

(4.207)

where z is the distance in the plasma from the vacuum-plasma boundary. This means that an
electromagnetic wave incident on a plasma cannot penetrate into plasma except for a distance of
the order of the skin depth dened by
c
=
:
(4.208)
!p
The wave is completely reected if the plasma is collisionless. Reection of low frequency radio
waves by the ionoshperic plasma and is a well known example.

37

In a collisional plasma with !

!p;

c;

the skin depth is modied as

c
'
!p

!
:
2 c

(4.209)

Derivation of this formula is left for exercise.

Strictly speaking, the collisionless skin depth = c=! p is valid for a cold plasma with negligible
electron temperature. To implement the eects of nite electron temperature, it is necessary to
employ a kinetic theory to nd a conductivity : The electron velocity distribution function f
(r; v; t) obeys the kinetic equation,
@f
+ v rf
@t

e
(E + v
m

B)

@f
= 0:
@v

(4.210)

As it is, it is a nonlinear equation because the electromagnetic elds E and B associated with a
wave aect the distribution function f: Let us assume a wave propagating along the z direction
with electric eld polarized in the x direction Ex (z; t) = E0 ei(kz !t) : The distribution function may
be linearized as f = f1 + FM ; where f1 is the perturbation and FM is unperturbed Maxwellian
distribution. Noting
mv
@FM (v 2 )
=
FM (v 2 );
(4.211)
@v
Te
and thus
(v

B)

we nd
f1 =

@FM (v 2 )
= 0;
@v

e
Ex
Te
vx FM :
i (kvz !)

(4.212)

The current density can be found from the rst order moment,
Z
Je = ne vf1 dv:

(4.213)

Only Jx is nonvanishing, and given by

Jx =
=
=
where

= !=kvT e with vT e =

ne2
Ex
i
Te

vx2
kvz

FM dv

Z 1
ne2
1
e
i
Ex p
mkvT e
1 t
2
ne
i
Ex Z ( ) ;
mkvT e

t2

dt
(4.214)

2Te =m being the thermal velocity of electrons and Z ( ) is known

38

as the plasma dispersion function. The permittivity is therefore given by

" = "0

!
! 2p
Z( ) :
1+
!kvT e

If the electron temperature is negligible

(4.215)

1; the function Z ( ) approaches

Z( )'

and we recover the case of cold plasma,

" = "0
In the opposite limit

1;

! 2p
!2

p
Z( )'i ;

and

! 2p
p
" ' i "0
:
!kvT e

The damping factor (inverse skin depth) in this limit is given by

Im k =

1=6

'

!p
c

2=3

!
vT e

1=3

(4.216)

This is often called anomalous skin eect. (This is probably misnomer because there is nothing
anomalous in the derivation.)

4.14

Waves in Anisotropic Dielectrics

Some crystals exhibit anisotropy in polarizability. The permittivity in such media becomes a tensor,
and the displacement vector D and electric eld E are related through a dielectric tensor ";
D = " E; or Di ="ij Ej ;

(4.217)

where " is a diagonal tensor consisting of three permittivities in each axial direction, x; y; and z;
1
"x 0 0
B
C
" = @ 0 "y 0 A :
0 0 "z
0

(4.218)

In uniaxial crystals, polarization occurs preferentially along one axis, say, "x = "y 6= "z : In
anisotropic media, the phase and group velocities are in general oriented in dierent directions
and the well known double refraction phenomenon (already known in the 17th century) can be
39

explained in terms of the dielectric anisotropy.

The Maxwells equations to describe wave propagation in an anisotropic medium in which there
are no sources ( free = 0; J = 0) are
r

@B
; r
@t

E=

B=

@D
=
@t

Then
r

E=

@2E
;
@t2

0"

r (r E) r2 E+

0"

0"

@2E
= 0:
@t2

@E
:
@t

(4.219)

(4.220)
(4.221)

Note that in an anisotropic medium, the divergence of the electric eld, r E; does not necessarily
vanish, although
r D = free = 0;
(4.222)
must hold. After Fourier decomposition, Eq. (4.221) reduces to
k(k E)

k2 E + !2

0"

E = 0;

(4.223)

or
(k 2

!2

ki kj

ij

0 "ij )Ej

= 0:

(4.224)

The dispersion relation of electromagnetic waves in a uniaxial crystal is therefore given by the
determinant,
det(k 2 ij ki kj ! 2 0 "ij ) = 0:
(4.225)
Introducing ^"ij = "ij ="0 and the index of refraction,
n=

c
k;
!

(4.226)

we rewrite the dispersion relation as

det(n2

ij

ni nj

^"ij ) = 0:

(4.227)

In uniaxial crystals, polarizability along one axis (say, z-axis) diers from those along other
axes. We assume
"x = "y "? ; " z "k :
(4.228)
Since the crystal is symmetric about the z-axis, the wave vector k can be assumed to be in the
x z plane without loss of generality,
kx = k? ;

40

kz = kk ;

(4.229)

or
n? =

ck?
;
!

nk =

ckk
:
!

(4.230)

The dispersion relation then becomes

n2k

^"? 0

n? nk

n? nk

n2
0

^"? 0
n2?

= 0:

(4.231)

^"k

Expanding the determinant, we nd

n2 = ^"? ;

(4.232)

and
^"k n2k + ^"? n2?

^"k^"? = 0:

(4.233)

The rst mode of propagation is independent of the propagation angle as if in an isotropic medium.
It is called the ordinary mode because it maintains the properties of electromagnetic waves in
isotropic media. The electric eld is in the y-direction and the dispersion relation in Eq. (4.232) is
equivalent to the following dierential equation,
@2
@2
+
@x2 @z 2

0 "? !

Ey (x; y) = 0:

(4.234)

In this mode, the phase and group velocities are in the same direction (in the direction of k)
although in magnitude they dier because of the frequency dependence of the permittivity "? (!):
The second mode has a peculiar property and for this reason is called extraordinary mode. Let
the angle between k and the z-axis be : The dispersion relation in Eq. (4.233) becomes
ck
!

^"k cos2 + ^"? sin2

The phase velocity given by

s
!
cos2
=c
k
^"?

^"k^"? = 0:

sin2
ek ;
^"k

(4.235)

(4.236)

is directed along k: The group velocity is

d!
dk

=
=

@!
1 @!
ek +
e
@k
k@
!
c2
1
ek +
k
!=k ^"k

1
^"?

sin cos e :

(4.237)

This is evidently not parallel to the phase velocity unless ^"k = ^"? (isotropic dielectric). The
magnitude of the group velocity is always larger than that of the phase velocity.
Electromagnetic waves in a plasma conned by a magnetic eld is another example of anisotropic
medium which accommodates variety of waves. A simple case of cold plasma will be considered in
41

Problem 4.11.

Figure 4-8: Double refraction of randomly polarized incident wave by a uniaxial crystal. Propagation of the O-mode (ordinary mode) is unaected but that of the X-mode (extraordinary mode) is.
The group velocity of the X-mode in the crystal deviates from the incident direction.

An important consequence of the presence of the ordinary and extraordinary modes is the well
known double refraction caused by some crystals. Consider a light beam with random polarizations
incident normal to a surface of a uniaxial crystal as shown in Fig. 4-8. Unless the optical axis (zaxis in the geometry assumed) coincides with the direction of the incident beam, the beam is split
into two beams at the surface. In the crystal, one beam propagates along the direction of the
incident beam and another at an angle. The phase velocities of the ordinary and extraordinary
modes are in the same direction as the incident beam but the group velocity of the extraordinary
mode deviates from the incident direction. Wave energy propagates at the group velocity and thus
beam splitting occurs.
Some isotropic dielectrics can become uniaxial media if placed in an electric eld. This is
because the permittivity is in general nonlinear and the component in the direction of the eld
becomes eld-dependent,
" = "(0) + E n ;
(4.238)
where is a constant, n = 1 is for Pockels eect and n = 2 is for Kerr eect. Pockels eect in
some liquids is widely used for laser switching and optical modulation.

42

Problems
4.1 A sphere of radius a carries a charge q which is decreasing due to emission of charge in
every radial direction. Show that there is no magnetic eld even though there exists a radial
conduction current density.
4.2 A cylindrical permanent magnet with radius a and a uniform axial magnetization Mz is
rotating at an angular frequency ! about its axis. What is the emf induced in the closed
circuit shown?

Problem 4.2. Unipolar generator.

4.3 The permittivity of an unmagnetized plasma is given by
"(!) = "0

! 2pe
!2

where ! pe is the electron plasma frequency,

! pe =

ne2
:
" 0 me

Show that the energy density of a plane electromagnetic wave in a plasma is

1
1 ! 2pe
1
u = "0 E 2 + "0 2 E 2 +
2
2 !
2
and interpret the term
1 ! 2pe 2
"0
E :
2 !2

43

0H

= "0 E 2 ;

4.4 An isotropic dielectric has a permittivity in the form

"(!) = "0

! 2p
!2

! 20

Show that the group velocity does not exceed c: What if a nite dissipation is allowed,
! 2p
! 2 + 2i !

"(!) = "0

! 20

where is a damping constant? For simplicity, assume ! p = ! 0 and plot Re (d!=dk) as a

function of ! for various =! 0 :
4.5 Show that if a dielectric medium is loss free, its dielectric tensor should be Hermitian,
"ij = "ji :
Note: In the absence of external magnetic eld, the tensor is symmetric "ij = "ji : In this
case, the loss-less condition is Im ("ij ) = 0:
4.6 Find the permittivity and thickness of a dielectric lm to be coated on a glass surface to
eliminate reection of light wave of wavelength = 550 nm. Assume that the glass has an
index of refraction of n = 1:5:
4.7 A light beam linearly polarized is incident on a glass (nglass = 1:5) surface at an angle i = 45 :
Find the amplitude and polarization of the reected and refracted waves. Consider both cases
of polarization, electric eld in the incident plane and magnetic eld in the incident plane.
4.8 A light beam of circular polarization in glass is incident on a at glass-air boundary at an
angle i = 60 : Find the polarization of totally reected beam.
4.9 A uniaxial crystal has "xx = "yy = 1:3"0 and "zz = 1:5"0 : The optical axis is at angle 70 to
a at surface. A light beam of random polarization is incident normal to the surface. Find
the propagation direction of the extraordinary mode in the crystal.
4.10 A uniaxial crystal has the following permittivity tensor,
0

1
3"0 0
0
B
C
!
" = @ 0 3"0 0 A :
0
0 2"0
A light beam in air is incident on a at surface(x y plane)of the crystal at an angle i
from the normal which coincides with the optical axis z: Show that the refraction law for the

44

extraordinary mode is
tan

p
^"? sin

=q
^"k ^"k

sin2

;
i

where ^"? = 3; ^"k = 2: What is the refraction law for the ordinary mode?
4.11 Find the impedance per unit length of a copper wire of radius 3 mm at f = 60, 103 and 106
Hz. Copper conductivity is = 5:8 107 S/m at room temperature.
4.12 Using the equation of motion for electrons in a cold magnetized plasma,
me

@v
=
@t

e (E + v

B0 ) ;

where B0 (external magnetic eld) is in z direction, show that the dielectric tensor is given
by
3
2
! 2p
! 2p
i
0
7
6 1 !2
2
2)
!(! 2
7
6
2
2
7
6
!
!
!
p
ps
7
" = "0 6
1
0
7
6 i
2)
2
!2
7
6 !(! 2
4
! 2p 5
0
0
1
!2
p
where ! p = ne2 =me "0 is the plasma frequency and e = eB0 =me is the cyclotron frequency.
The wavevector k may be assumed to be k = k? ex +kk ez without loss of generality because of
axial symmetry. Note that the tensor is Hermitian. (In this analysis, the electron temperature
is ignored.)
4.13 A laser beam passes through a glass window of refractive index 1.50 into water with refractive
index 1.33. The beam is E-polarized (electric eld in the incident plane). Design the glass
window to avoid reection at both surfaces.

45

Chapter 5

Radiation from Macroscopic Sources

(Antennas and Apertures)
5.1

Introduction

Strictly speaking, radiation of electromagnetic waves from antennas and apertures should be analyzed as a boundary value problem incorporating the boundary conditions for the electric and
magnetic elds. This is because the current distribution in antennas and electromagnetic elds in
apertures are likely to be modied by the elds created by antennas and apertures, and incorporating the boundary conditions should be able to solve radiation problems self-consistently. However,
such a procedure is usually very much involved even in simple geometry and often requires solving
integral equations. In this Chapter, we will assume that feedback from the elds on the prescribed
current and eld distribution is negligibly small. This simplication makes numerous radiation
problems analytically tractable.

5.2

Radiation Vector Potential

In Chapter 4, we have seen that the vector potential obeys a wave equation
r2

1 @2
c2 @t2

A=

0 J (r; t) ;

(5.1)

provided the Lorenz gauge is chosen,

r A+

1 @
= 0:
c2 @t

(5.2)

Eq. (1) has the following retarded solution,

A(r; t) =

J(r0 ; t
) 0
dV ;
0
jr r j
1

(5.3)

where
=

1
r
c

r0 ;

(5.4)

is the time required of electromagnetic disturbance to propagate from the source position r0 to the
observing position r: It is noted that J and A consist of longitudinal and transverse components.

Figure 5-1: The vector potential dA at a distance jr r0 j from the current segment JdV 0 located
at r0 is created by the current = jr r0 j =c seconds earlier because of nite propagation speed of
electromagnetic disturbance.

If the current is oscillating at a frequency !; and if the current density is separable in the form
J(r; t) = J(r)e

i!t

; (A m

(5.5)

the vector potential becomes

A(r; t) =

J(r0 )
exp
jr r0 j

i! t

jr

r0 j

For a localized, small radiation source, the observation point is at r

may be approximated by
ei(kr
A(r; t) '
4
r
0

!t)

J(r0 )e

ik r0

dV 0 ;

(5.6)

r0 ; and the vector potential

dV 0 ;

(5.7)

where k = !=c; k = ker ; and the following approximation is used,

k r

r0 ' kr

r 0
r = kr
r

k r0 for r

r0 :

(5.8)

The magnetic eld associated with the vector potential can be calculated from
B=r

(5.9)

A;

and the transverse radiation electric eld from

1 @E
=r
c2 @t

(5.10)

B:

If a localized source is radiating electromagnetic energy at a constant power P; the surface

integral of the Poynting ux on an arbitrary sphere with radius r should yield the power,
I
I
1
P =
jEj2 r2 d = Z0 jHj2 r2 d = constant (in W);
(5.11)
Z0
where
d

= sin d d ;

(5.12)

is the dierential solid angle. Therefore, as far as radiation of energy is concerned, retaining both
electric and magnetic elds proportional to 1=r will su ce: Since the vector potential already has
1=r dependence, the dierential operator r in calculating the electric and magnetic elds can be
approximated by
r

1 i(kr
e
r

!t)

r i(kr
e
r3

1
+ ik ei(kr
r

!t)

!t)

1
' ik ei(kr
r

!t)

provided kr
1, that is, in the region at least several wavelengths away from the source region.
The region kr
1 is called radiation or far-eld zone. This indicates that as far as radiation power
is concerned, the magnetic eld may be approximated by
B = r

' ik

A; kr

1:

(5.13)

In the far eld region kr

1 eects of radiation source become ignorable and electric and magnetic elds are essentially detached from the source. The radiation power calculated through the
approximation is pure real indicating a net power radiated by the source.
In order to nd a reactive power due to electromagnetic energy stored in the vicinity of a
source, one has to integrate the Poynting ux on a closed surface near the source. The radiation
zone approximation fails completely in this case. Also, even in radiation zone, higher order terms
are often required in evaluating radiation of angular momentum L,
1
dL
=
dt
c

(E

H)r2 d ; (J).

Since
r

(E

H) = E(r H)
3

H(r E);

(5.14)

it can be seen that radial components, either Er or Hr ; proportional to 1=r2 ; yield a constant rate
of radiation of angular momentum,
1
r (E H) _ 2 :
r
Example 1 Center-Fed Half Wavelength Dipole Antenna

Figure 5-2: Half wavelength center-fed antenna.

Radiation by a center-fed half wavelength dipole antenna is of practical importance because

of its common use in communication. It is formed by bending inner and outer conductors at a
distance =4 from the open end of a coaxial cable. The standing current wave along the antenna
may be approximated by
2
;
(5.15)
I(z; t) = I0 cos(kz)e i!t ; k =
if feedback from the radiated eld is ignorable. Then, the vector potential at kr
0 I0 i(kr !t)

Az (r; t) =

4 r
0 I0 i(kr
e
4 r

!t)

=4

cos(kz 0 )e

ikz 0 cos

1 is

dz

=4

2
cos
cos
2
k sin2

(5.16)

e :

(5.17)

and the radiation magnetic eld in the region is

B ' ik
=

A
I0
i 0 ei(kr
4 r

!t)

2
cos
cos
sin
2

Corresponding Poynting ux is
Sr = Z jH j2
=

ZI02
1
cos2
cos
4 2 r2 sin2
2

(5.18)

(5.19)

and angular distribution of the radiation power is given by

dP
d

= r 2 Sr
ZI02 1
cos2
cos
4 2 sin2
2

Integration over the entire solid angle yields the total radiation power,
P

=
=

dP
d
d
sin d
2 Z

ZI0
2

ZI02 1
cos2
cos
4 2 sin2
2

1
cos2
cos
sin
2

d :

(5.20)

The integral can be done numerically and amounts to 1.219. Then

P = 73:1 ( )I02 = Rrad I 2 ; (W)
where
Rrad =

0 ="0

(5.21)

1:219 = 73:1 ( )

(5.22)

is the well known radiation resistance of the center-fed half wavelength dipole antenna. The angular
dependence of the Poynting ux
f( ) =

1
cos2
cos
2
sin2

(5.23)

is shown below.
The calculation presented above entirely ignores the reactive power which may exist due to storage of electric and/or magnetic energy in the vicinity of the antenna. To account for such reactive
power, we must deviate from the far-eld analysis and integrate the Poynting ux directly on a
surface close to the antenna surface. Instead of calculating the Poynting ux on the antenna surface
of nite radius a, we calculate the Poynting ux on a cylindrical surface of radius a surrounding an
ideally thin antenna of length =2 as shown in Fig. 5-4. In this approximation, the magnetic eld
on the antenna surface may be replaced by the static form without retardation,
H (z) =

I0
I(z)
=
cos(kz);
2 a
2 a

<z<

(5.24)

0.3

0.2
0.1
-1.0

-0.8

-0.6

-0.4

-0.2

0.2

-0.1

0.4

0.6

0.8

1.0

-0.2
-0.3

Figure 5-3: Angular ( ) dependence of the Poynting ux radiated by a half wavelength dipole
antenna.

The electric eld on the antenna surface is zero within our assumption of ideally conducting antenna
except at the gap at the midpoint. However, the electric eld due to the current lament assumed
at the axis is nite at the surface a distance a away. It can be calculated from
1 @
E=r
c2 @t

B = r(r A)

r2 A;

(5.25)

where A is the vector potential on the surface. It can be found from the integral,
0

Az (z) =

=4
=4

where
R(z; z 0 ) =

I0 cos kz 0 ikR(z;z 0 ) 0
e
dz ;
R(z; z 0 )

p
(z

(5.26)

z 0 )2 + a2 ;

(5.27)

is the distance between a point on the surface ( = a; z) and a current segment I(z 0 )dz 0 at z 0 : With
this approximation, Eq. (5.25) reduces to
i

@ 2 Az
!
E
(z)
=
+ k 2 Az ;
z
c2
@z 2

(5.28)

since in current-free region the vector potential satises the Helmholtz equation
r2 + k 2 Az = 0:
The integral in
Ez (z) = i

c2 0 I0
4 !

=4

cos(kz 0 )

=4

@2
+ k2
@z 2

(5.29)

eikR(z;z ) 0
dz ;
R(z; z 0 )

(5.30)

Figure 5-4: The Poynting ux on the surface of =2 antenna of nite radius a is approximated by
that on the cylindrical surface of radius a surrounding a thin antenna.

can be performed by noting

@ eikR(z;z )
=
@z R(z; z 0 )

@ eikR(z;z )
;
@z 0 R(z; z 0 )

(5.31)

and by integrating by parts twice with the result

Ez (z) = i
where
R1 =

Z0 I0
4

eikR1
eikR2
+
R1
R2

+ a2 ;

R2 =

(5.32)

z+

+ a2 :

(5.33)

The radial outward Poynting ux on the antenna surface is therefore given by

S

Ez H

Z0 I02
8 2a

eikR1
eikR2
+
R1
R2

cos(kz);

(5.34)

and the power leaving through the antenna surface is

P

= 2 a
=

=4

Z0 I02
i
4

S dz
=4

=4
=4

eikR1
eikR2
+
R1
R2

cos(kz)dz:

(5.35)

2
1.5
1
0.5
00

0.02

0.04

0.06

0.08

0.1

Figure 5-5: Real (lower curve) and imaginary (upper curve) parts of the integral.

The integral has to be performed numerically. Introducing

x=

4z

A=

4a

we rewrite the integral in the form

f (A) =

i p
i p
(1 x)2 + A2
exp
(1 + x)2 + A2
1 B exp
2
2
B
p
p
+
@
(1 x)2 + A2
(1 + x)2 + A2
1

C
C cos
x dx;
A
2

(5.36)

which is shown in Fig. 5-5 as a function of the normalized antenna radius A = 4a= : For A < 0:01;
the radiation impedance is constant and approximately equal to
Zrad ' 73:1

i42:0

= 73:1 + j42:0

( )
( ) (in engineering notation)

The real part comfortably agrees with the radiation resistance calculated earlier in the far eld zone.
The reactive part of the impedance is inductive due to dominant magnetic energy compared with
the electric capacitive energy. However, the reactance is a very sensitive function of the antenna
length. It vanishes if the antenna length is chosen at l ' 0:49 and further decrease in the length
makes the reactance capacitive. Radiation from a center-fed antenna can be analyzed by assuming
a standing wave form,
I(z) = I0 sin[k(l jzj)];
and is left for exercise.The axial electric eld Ez (z) in Eq. (5.32) can be alternatively (perhaps
more conveniently) found from
@
@
Az ;
(5.37)
Ez =
@z
@t
8

where

is the retarded scalar potential given by

Z ijr r0 j
e
1
r0 dV 0
4 "0
jr r0 j
Z iR
1
e
z 0 dz 0 ;
4 "0
R l

(z) =
=
with

R=
and

q
(z

(5.38)

z 0 )2 + a2 ;

the linear charge density that can be found as

@ l @I (z)
+
= 0;
@t
@z
l

i
(z) = I0 sin kz; (C m
c

):

(5.39)

Then,
Ez =
=
=

@
@
Az
@z
@t
Z =4
iI0
@
0
4 "0 c
@z
=4
iI0
4 "0 c

eikR1
R1

eiR
R
eikR2
R2

i! 0 I0
sin kz dz +
4
0

iZ0 I0
4

eikR1
R1

eikR2
R2

=4
=4

cos kz 0 ikR 0
e dz
R

(5.40)

which agrees with Eq. (5.32). In radiation zone, the scalar potential is irrelevant but in near eld
region, it should be considered together with the vector potential in a self consistent manner.

5.3

Radiation from Sources Smaller than Wavelength: Multipole

Radiation

If the size of a radiating source is small compared with the wavelength kr0
potential at r
r0
Z
0
0 i(kr !t)
A(r; t) '
e
J(r0 )e i k r dV 0 ;
4 r
can be expanded in a series of k r0 ;
A(r; t) '

4 r

i(kr !t)

J(r0 ) 1

In magnetostatics, the rst integral vanishes,

Z
J(r0 )dV 0 = 0;

ik r0 +

dV 0 :

1, the far eld vector

as shown in Chapter 3. However, for time varying current density, it does not, since
Z
Z
Z
Ji dV =
rxi JdV =
xi r JdV
Z
@
d
=
xi dV = pi ;
@t
dt
where
pi =

xi dV;

(5.41)

is the i-th component of the electric dipole moment,

Z
p = r dV; (C m).
In general, for time varying sources,
Z

J(r;t)dV =

d
p:
dt

(5.42)

The lowest radiation eld from a small source is therefore of electric dipole nature.
The integral in the next order,
Z
(k r)JdV;
is a combination of electric quadrupole and magnetic dipole as seen from the following manipulation,
Z
Z
kj rj Ji dV = kj rj rri JdV
Z
Z
Z
= kj r (ri rj J)dV kj ri rrj JdV kj ri rj r JdV
Z
Z
@
=
kj ri Jj dV + kj ri rj dV;
(5.43)
@t
or going back to the vector form
Z

(k r)JdV =

d
r(k J)dV +
dt

(k r)r dV:

(5.44)

However,
(r

J) = r(k J)

(5.45)

J(k r):

Therefore,
Z

(k r)JdV

1
k
2

1
d
(r J)dV + k
2
dt
1
d
m+ k
Q;
2
dt

10

rr dV
(5.46)

where
m=

Q=

1
2

JdV; (A m2 = J T

(5.47)

is the magnetic dipole moment and

rr (r) dV; (C m2 )

is the electric quadrupole moment tensor. Then, the radiation vector potential to order k r0 is
A(r) '

4 r

p+ik
_

1
_ eikr ; kr0
ik Q
2

1 (small source)

(5.48)

where the dot indicates time derivative. The radiation magnetic eld can be found from
0H

H(r) '

1
p
+ n (n
c

4 c

= B ' ik
m)

(5.49)

A;
1
n
2c

...
n Q

eikr
;
r

(5.50)

where n = r=r is the radial unit vector and the vector k =kn has been replaced by a time derivative
operator wherever appropriate,
!
n@
k= n=i
:
(5.51)
c
c @t
The radiation power can be calculated from
Z
P = Z0 jHj2 r2 d ; (W).

(5.52)

For example, the power radiated by an electric dipole oscillating at a frequency ! is given by
P

=
=

Z
2
Z0 ! 4
p
n
d
(4 c)2
Z
2
1
1
p n d ;
3
4 "0 4 c

where
p=

d2
p:
dt2

The solid angle integration can be performed by exploiting

Z
4
ni nj d =
ik ; i; j = x; y; z:
3
The result is
electric dipole: Pp =

(5.53)

1 2! 4 p2
1 2
p2
=
; (W)
4 "0 3c3
4 "0 3c3

11

(5.54)

(5.55)

where p = jpj is the magnitude of the dipole moment, e.g.,

p2 = p2x + p2y + p2z :

(5.56)

Similarly, the power radiated by a magnetic dipole is

magnetic dipole: Pm =

2
1 2! 4 m2
1 2m
=
; (W)
4 "0 3c5
4 "0 3c5

(5.57)

which is of higher order by a factor (v=c)2 ' (ka)2 compared with the radiation power due to
electric dipoles where v is the typical velocity of charges associated with the current J = nev and
a is the typical size of the radiation source.
The radiation power due to an electric quadrupole is given by
P =

1 1
"0 16 2 4c4
0

...
n Q

d :

(5.58)

The x component of the vector

q=n

(n Q) ;

(5.59)

is
qx = ny (nx Qxz + ny Qyz + nz Qzz )
Using

we nd

Similarly,

and

ni nj nk nl d

4
(
15

nz (nx Qxy + ny Qyy + nz Qzy ):

ij kl

ik jl

il jk );

(5.60)

(5.61)

qx2 d

4
Q2xy + Q2xz + 4Q2yz + (Qyy
15

Qzz )2 :

(5.62)

qy2 d

4
Q2xy + Q2yz + 4Q2xz + (Qxx
15

Qzz )2 ;

(5.63)

qz2 d

4
Q2xz + Q2yz + 4Q2xy + (Qxx
15

Qyy )2 ;

(5.64)

where the symmetry of the quadrupole moment Qij = Qji is noted. Adding the three components
yields
Z
4
6(Q2xy + Q2xz + Q2yz ) + (Qxx Qyy )2 + (Qyy Qzz )2 + (Qzz Qxx )2
q2d
=
15
2
3
X
4 4 X 2
=
3
Qij (
Qii )2 5 ;
(5.65)
15
ij

12

and the radiation power is

electric quadrupole: PQ =

1
1 4
3
4 "0 60c5

Qij

ij

!2 3
Qii 5 ; (W).

(5.66)

Example 2 Charge Undergoing Nonrelativistic Circular Motion

Figure 5-6: Charge e undergoing nonrelativistic circular motion. The dominant radiation process
is of electric dipole nature.

This example of radiation from a nonrelativistic single charge is discussed here because radiation
elds are of multipole nature. In the lowest order, dipole radiation is expected followed by electric
quadrupole radiation and higher. The magnetic dipole radiation is absent because the magnetic
dipole associated with the circulating charge is constant. The current due to a moving single charge
is
J = ev (t) (r rp ) ;
where e is the charge, rp (t) is the instantaneous location of the charge, and v (t) = drp =dt is the
velocity. If a charge e is undergoing circular motion with nonrelativistic velocity, v = !a
c; the
dipole moment is
p (t) = erp (t) = ea(cos !tex + sin !tey )
= ea [sin cos(

!t)er + cos cos(

!t)e

sin(

where the conversions

ex = rx = r(r sin cos )
= sin cos er + cos cos e
13

sin e ;

!t)e ] ;

(5.67)

ey = ry = r(r sin sin )

= sin sin er + cos sin e + cos e ;
are substituted. It is convenient to introduce complex notation,
1
p = p ea (sin er + cos e + ie ) ei(
2

!t)

p
The factor 1= 2 is required for the amplitude jpj to remain ea: The vector potential is
A(r; t) =

4 r

pe
_ ikr =

4 r

( i!)pei(kr

!t+ )

(5.68)

and the radiation magnetic eld is

H(r; t) '
=

ik

ea!k 1
p (e + i cos e ) :
4 r 2

(5.69)

Note that the eld rotates with the charge as explicitly indicated by the factor ei(kr !t+ ) : The
magnetic eld is transverse (TM mode). In general, electric multipoles radiate TM modes. The
radiation eld is elliptically polarized,
H
= i cos :
(5.70)
H
In particular, at

= 0; the eld is circularly polarized with positive helicity,

H
= i;
H

At = =2; the component vanishes and the led is plane polarized. At

circularly polarized but with negative helicity,
H
=
H

i:

(5.71)
= ; the eld is again

(5.72)

The radiation power is constant because the magnitude of the dipole moment is constant, and is
given by
1 2! 4 (ea)2
P =
:
(5.73)
4 "0 3c3
However, a! 2 = jvj
_ is the centripetal acceleration experienced by the charge and the power can be
rewritten as
1 2(ev)
_ 2
P =
:
(5.74)
4 "0 3c3
This is the well known Larmors formula for radiation power emitted by a nonrelativistic charge
under acceleration v_ regardless of its direction relative to the velocity v.

14

Let us consider higher order modes. The magnetic dipole moment of the circulating charge is
constant,
Z
1
r JdV
m =
2
Z
1
=
e!r0 r e [r rp (t)]dV
2
1
=
e!r02 ez (constant).
2
Therefore, there is no magnetic dipole radiation. The non-vanishing components of the quadrupole
moment tensor of the circulating charge are
Qxx = ea2 cos2 !t;

Qyy = ea2 sin2 !t;

Qxy = Qyx = ea2 cos !t sin !t:

(5.75)

Corresponding radiation power is

1 4 X ...2
1
Qij
3
4 "0 60c5
1 8
4 "0 5

ij

(ea2 )2 ! 6
c5

X ...
Qii
i

!2 3

(5.76)

It should be noted that the quadrupole radiation elds all oscillate at 2! (twice the circulation
frequency) and corresponding radial wavenumber is 2k = 2!=c: The quadrupole radiation power is
smaller than the dipole radiation power by an order (ka)2 as expected, for ka is the small expansion
parameter employed in the course of derivation. The appearance of the quadrupole radiation elds
even for pure circular motion of the charge is strictly due to retardation of radiation elds which
causes frequency modulation in the electromagnetic elds. In relativistic cases, this eect is much
more pronounced and radiation intensity of higher harmonics becomes signicant. Radiation from
relativistic charges will be analyzed in Chapter 8.
Example 3 Radiation of Angular Momentum in the Preceding Example
The charge loses its kinetic energy to radiation elds at the rate
d
dt

1
M v2
2

=P =

1 2! 4 (ea)2
:
4 "0 3c3

(5.77)

Since the angular momentum of the charge is

Lz = M vr0 ;

(5.78)

radiation elds carry away angular momentum at the rate

dLz
=
dt

M r0
15

dv
P
= :
dt
!

(5.79)

In general, if radiation elds have eim dependence, the ratio becomes

1 dLz
m
= :
P dt
!

(5.80)

Here, we wish to verify Eq. (5.79) through direct calculation of the rate of angular momentum
radiation,
Z
1 2
dL
= r
r (E H )d :
(5.81)
dt
c
Since an electric dipole radiates TM mode, it follows that
(E

H )=

H (E r);

(5.82)

and the radial component of the electric eld Er ; that has been ignored in calculation of energy
radiation, must be retained. For the ux of angular momentum r (E H )=c to be proportional
to r2 ; the radial electric eld must be proportional to 1=r2 ; since the dominant radiation magnetic
eld is proportional to 1=r:
Since the radial electric eld is of higher order than the radiation eld, we may continue to use
the radiation magnetic eld,
1

H =

'

4 r

eikr k

1 ikr
e n
4 cr

p_ =

p
:

(5.83)

rr A + k 2 A :

(5.84)

The electric eld is to be found from

"0

@
E=r
@t

H=

1
0

A=

Noting
r A=

we nd
Er =

eikr
p_
r

ik
r

1
r2

eikr p_r ;

eikr
eikr
p
_
=
(n p):
_
r
2 "0 cr2
2 "0 cr2

Therefore, the radiation ux of the angular momentum is

=
=

1
r (E H )
c
1
(r E)H
c
1
1
(n
4 "0 2 c3 r2

16

p
)(n p);
_

(5.85)

and the rate of angular momentum radiation is given by

dL
1
1
=
dt
4 "0 2 c3

(n

p
)(n p)d
_
:

(5.86)

In the case of circulating charge,

1
p(t) = p p0 (sin er + cos e + ie )ei(
2
we have
p_ =

; p0 = ea;

i
p !p0 (sin er + cos e + ie )ei(
2

and
n

!t)

p
=

1
p ! 2 p0 (cos e + ie )e
2

i(

!t)

!t)

(5.87)

(5.88)

(5.89)

Then, the rate of angular momentum loss and radiation power are related through
dL
dt

=
=
=

Z
1 ! 3 p20
sin2 d ez
4 "0 4 c3
1 2! 3 (ea)2
ez
4 "0 3c3
P
ez ;
!

(5.90)

as expected.
Example 4 Circular Loop Antenna

Figure 5-7: Small circular loop antenna with ka

1:

If a circular loop antenna of radius a carries a slowly oscillating current I0 e i!t where !a
c
or ka
1; the lowest order radiation eld is of magnetic dipole type. The magnetic dipole moment
17

is
mz = a2 I0 e

i!t

(5.91)

and radiation magnetic eld can be found from

H(r) =
=

1
::
eikr n (n m)
2
4 c r
eikr 2
! m0 sin e ;
4 c2 r

(5.92)

where
m0 = a2 I0 :
The peak radiation power is
P =

1 2! 4 m20
;
4 "0 3c5

(5.93)

1 ! 4 m20
:
4 "0 3c5

(5.94)

and the time averaged power is

Pave =

As the frequency increases, the dipole approximation breaks down. We will return to this problem
in Chapter 7.

5.4

Radiation from Apertures

Radiation from an aperture such as microwave horn antennas should be analyzed as a vector
boundary value problem. A rigorous formulation along this line will be made in Chapter 7. Here, a
nearly equivalent (but less rigorous) formulation based on a ctitious magnetic currentdeveloped
by Schelkuno in 1936 will be given. The absence of magnetic charge (magnetic monopole) and its
ow (magnetic current) is well established. However, as the tangential component of the magnetic
eld on a given boundary surface can be replaced by a surface electric current, the tangential
component of electric eld can be replaced by a ctitious magnetic current.

Figure 5-8: The tangential magnetic eld on an aperture can be replaced by a surface electric
current, while an electric eld by a ctitious magnetic current.

18

Let us consider an area S on which tangential components of electric and magnetic elds Et
and Ht are specied as illustrated in Fig. 5-8. In the method developed by Schelkuno, the normal
components of the electromagnetic eld are ignored, and the formula does not fully agree with that
based on the Greens dyadic in Chapter 7. The magnetic eld can be replaced by a surface electric
current
Js = n Ht ;
owing on a superconducting surface. Here n is the unit normal on the surface. The radiation
magnetic vector potential due to the surface current is given by
Z
Js
0
0
A(r) =
eikjr r j dS
4
jr r0 j
Z
n Hs ikjr r0 j
0
=
e
dS:
(5.95)
4
jr r0 j
For the tangential surface electric eld, Es ; Schelkuno introduced an electric vector potential,
Z
1
Es n ikjr r0 j
F(r) =
e
dS;
(5.96)
4
jr r0 j
and called the quantity Es n as a surface magnetic current. The electric eld in terms of both
vector potentials is given by
E(r) =
=

ic2
r r A r F
!
Z
i
n Hs ikjr
r r
e
4 "0 !
jr r0 j

r0 j

dS +

r0 ; kr
1; this reduces to
Z
i ikr
E(r) '
e
fk (n Es ) Zk
4 r

1
r
4

n
jr

Es ikjr
e
r0 j

r0 j

dS:

(5.97)

In the radiation zone r

where
Z=

0 ="0

[er (n

= 377

Hs )]g e

ik r0

dS;

(5.98)

(5.99)

Note that the electric eld derived above is explicitly transverse to the radial direction. In Chapter
7, a similar formula based on the Greens dyadic for the wave equation will be derived,
Z
0
i ikr
e
[k (n Es ) + !n Bs k(n Es )] e ik r dS:
(5.100)
E(r) =
4 r
This is not explicitly transverse but is in fact identical to Eq. (5.98) as will be shown in Chapter 7.
For a small hole in a conducting plate, the electric and magnetic dipole moments derived earlier
can be conveniently used to analyze leakage of electromagnetic energy through the hole. The electric

19

dipole moment of a small hole in a conducting plate is

p = 4 "0

a3
E? ;
3

(5.101)

where E? is the normal component of the unperturbed electric eld behind the plate. The magnetic
dipole moment of the hole is
8 a3
2a3
Hk =
Hk ;
(5.102)
m= 4
3
3
where Hk is the component of the magnetic eld parallel to the plate. If the elds are oscillating,
so are the dipole moments and they radiate electromagnetic waves. The oscillation frequency !
must be low enough for the dipole approximation to hold, !a
c; or ka
1; that is, the hole size
be much smaller than the wavelength :
Example 5 Radiation from an Open End of Coaxial Cable

Figure 5-9: Radiation from an open end of coaxial cable. (The lowest order TEM mode in the
cable is assumed.)

The lowest order electromagnetic eld propagating in an coaxial cable is azimuthally symmetric
and the eld components are given by
E ( ; t) =

V0 ei(kz
ln(b=a)

!t)

(5.103)

E ( ; t)
;
(5.104)
Z0
p
where a and b are the inner and outer radii and Z0 = "= 0 is the impedance of the insulating
material lling the cable. If the cable is terminated at an open end, voltage and current standing
waves are formed. The magnetic eld vanishes at an open end and the electric eld is doubled. The
electric eld radiated by the open end can be calculated from Eq. (??) with ignoring the magnetic
H ( ; t) =

20

eld,
E(r) '
=
'
where kb

ikr

(n

Es )e

ik r0

dS
4 r
Z b Z 2
i ikr kV0
d 0e
d
e
er
ez
4 r
ln(b=a)
0
a
eikr k 2 V0
(b2 a2 ) sin e ;
8r ln(b=a)
k

lk sin cos(

(5.105)

1 has been assumed. The radiation power is given by

P

= r
=
=

jEj2
d
Z0

1 1
64 Z0

k 2 V0
(b2
ln(b=a)

a )

1
24 Z0

k2
(b2
ln(b=a)

a2 )

sin3 d

V02 :

(5.106)

This denes the radiation resistance

Rrad =

24

ln(b=a)
2
k (b2 a2 )

Z0 :

(5.107)

Rrad is large because of the assumption kb

1 and indicates that radiation from an open end
of a coaxial cable is extremely ine cient. Note that in the integration over the solid angle, the
entire space is covered, 0 < < ; 0 < < 2 : If a coaxial cable is connected to a hole in a large
conducting plate, the integration is limited to 0 < < =2: However, the radiation electric eld
should be doubled because of image eect. The Poynting ux is quadrupled. However, since
integration covers the half space, the radiation power is only doubled.
Example 6 Leakage of Microwave through a Small Hole
Branching microwave power to another waveguide is often done with a device called directional
coupler shown in Fig.5-11. An even number of small holes are drilled a quarter wavelength apart
through a wall separating two waveguides. It is evident that microwave propagates toward the
region A because of constructive interference in that direction, while in region B, waves suer
destructive interference. In this example, we analyze radiation from a single small hole drilled at
the center of a wider wall of a rectangular waveguide.
If the hole radius is much smaller than the wavelength, the hole acts as oscillating dipoles
with electric and magnetic dipole moments,
3

p = 4 "0

E? ; m =

21

Hk :

(5.108)

Figure 5-10: For radiation from an open end of coaxial cable in a large conducting plate, the
radiation electric eld should be doubled compared with the case of free end. This is due to an
image current formed by the conducting plate.

The fundamental mode in a rectangular waveguide is the TE10 mode whose elds are
Ey (x) = E0 sin

E0
sin
x ; Hz _ cos
x ;
Z10
a
a

x ; Hx =

where
Z10 = p
1

Z0
(! c =!)2

(5.109)

(5.110)

is the impedance of the TE10 mode with ! c = c=a the cuto frequency. Since Hz vanishes at the
hole, the dipoles of the hole become
3

p = 4 "0

E 0 ey ; m = 8

E0
ex ;
3 Z10

(5.111)

Substituting this into the radiation magnetic eld due to dipoles in Eq. (??), we obtain
H(r) '
=

where Z0 =

0 ="0 :

1 ::
eikr
::
p n + (m n) n
4 c
c
r
3
2
ikr
p
k E0 e
(1 2 1 (! c =!)2 ) cos e
3 Z0 r
p
+(2 1 (! c =!)2 cos ) sin e ;
1

Calculation of radiation power, which is proportional to

Example 7 Resonant Slot Antenna

22

(5.112)
6

; is left for exercise.

Figure 5-11: Top: Directional coupler. g is the axial wavelength in the waveguide. Propagation
toward region B is prohibited because of destructive interference. Bottom: A small hole at the
midpoint of the wider wall radiates as electric and magnetic dipoles.

Slot antennas are widely used in microwave radiation. In this example, a simple resonant half
wavelength slot antenna is considered. As explained in Example 5, radiation electric eld from an
aperture in a large conducting plate is twice the eld due to an open aperture because of image
eect. In Fig. 5.4, the electric eld across the slot is assumed to be
Ex = E0 cos(kz):
Then the radiated electric eld is found from
Z
0
i ikr
E(r) '
e k
(n Es )e ik r dS
4 r
Z =4
i2aE0 ikr
=
e k ez
cos kz 0 e
4 r
=4

(5.113)

ikz 0 cos

dz 0 ;

(5.114)

where the antenna is assumed to be in the z-direction. The integral is identical to that encountered
in the case of half wavelength antenna, and the radiation electric eld reduces to

E (r; ) =

ieikr aE0 cos 2 cos

r
sin

23

(5.115)

Figure 5.12:

=2 resonant slot antenna.

24

Problems
5.1 A quarter wavelength antenna is erected vertically from the ground which may be regarded
as a perfect conductor in the lowest order approximation. Show that the radiation resistance
is 73=2 = 36:5 :
5.2 Assuming the following current distribution in a center-fed dipole antenna of length 2l and
radius a,
I(z) = I0 sin[k(l jzj)];
l z l;
nd, through numerical analysis, the length at which the antenna impedance becomes pure
real. Consider the cases of a= = 0:001 and 0.01.
5.3 Show that if the permittivity and permeability tensors are symmetric, "ij = "ji ;
r (Ea

Hb

Eb

ij

ji ;

Ha ) = 0;

where Ea(b) is the electric eld due to antenna a(b) and Ha(b) is the magnetic eld due to
antenna a(b): This is known as the reciprocity theorem.
5.4 An antenna array is formed by six dipole antennas half wavelength apart. The phase of
feeding currents can be controlled Im = I0 ei(m 1) ; where m = 1 6: Determine the angular
and phase dependence of the far-eld Poynting ux f ( ; ) and observe that the radiation
pattern can be rotated by changing the phase :

Six-antenna array (top view).

5.5 An antenna of length l carries a travelling wave I(z; t) = I0 ei(kz
power is given by
P =

Z0 2
I
4 0

!t) :

Show that the radiation

sin2 [(kl=2)(1 cos )] sin3

d ; (W),
(1 cos )2
25

where Z0 =

0 ="0 :

5.6 Two half wavelength antennas are a distance d apart as shown. One is driven by a power
supply and the other is passively driven by the rst antenna through a mutual impedance
approximately given by
Z12 =
where
r1 =

Z0
i
4

p
d2 + (z2

=4
=4

eikr1
eikr2
+
r1
r2

=4)2 ; r2 =

cos(kz2 )dz2 ;

d2 + (z2 + =4)2 :

Determine the angular dependence of the radiation pattern. (This problem shows the principle
of Yagi-Uda antenna. The driven antenna is called a parasitic antenna, or reector/director
depending on its eect on the radiation pattern. In practice, several parasitic antennas are
installed to increase directivity as in TV antennas.)
Hint: To nd I2 in terms of I1 ; use
V2 = 0 = Z12 I1 + Z22 I2 ;
where Z22 = Z11 = 73:1

i42:0

Yagi-Uda antenna. One is active and the other is passive (driven by the rst antenna).
5.7 A rectangular aperture of area a b in a conducting plate is illuminated uniformly by a short
wavelength incident wave. Find the diracted electric eld.
5.8 Calculate the power radiated by a hole in the wall of a rectangular waveguide using the eld
given in Eq. (5.112).
5.9 Two identical charges undergo circular motion at an orbit radius a and rotation frequency ! 0
staying on a diameter. Find (a) the radiation power and (b) the rate of angular momentum
radiation. Assume a! 0
c (nonrelativistic).
5.10 Generalize the preceding problem for the case of N identical charges distributed along a circle
at equal spacing. In the limit N ! 1; radiation is expected to disappear. Does it?
26

5.11 An electric dipole moment is suddenly created at r = 0; t = 0; p (t) = pz U (t) where U (t) is
the Heaviside step function. (Consider two charges q and q originally overlapped that are
suddenly pulled apart at t = 0.) Find exact expressions for the electric and magnetic elds.
5.12 A magnetic dipole m precesses at a frequency ! and angle 0 from precession axis. Find the
radiation power. (Radiation from pulsers is believed to be due to this mechanism.)

27

Chapter 6

Harmonic Expansion of
Electromagnetic Fields
6.1

Introduction

For a given current source J(r; t); the vector potential can in principle be found by solving the
inhomogeneous vector wave equation,
r2

1 @2
c2 @t2

A (r; t) =

0 J (r; t) ;

provided the Lorenz gauge is chosen. In source free region, the electromagnetic elds E and H can
then be deduced from the vector potential through
B=r
"0

@E
=r
@t

A;

B; (in source-free region J = 0):

In this Chapter, we develop spherical harmonic expansion of the electromagnetic elds. In experiments, the electromagnetic elds, not the potentials, are measured. Radiation electromagnetic
elds can be decomposed into two basic vector components, Transverse Magnetic (TM) and Transverse Electric (TE), where transverse is with respect to the direction of wave propagation r.
These fundamental modes are normal to each other and provide convenient base vectors in analysis
of radiation elds.

6.2

Wave Equations and Greens Function

The set of Maxwells equations,

r E=

"0

(6.1)

E=

@B
;
@t

(6.2)

r B = 0;
r

B=

(6.3)

J + "0

@E
@t

(6.4)

can be reduced to two inhomogeneous wave equations for the two potentials
Substitution of
@A
;
E= r
@t

and A as follows.
(6.5)

into Eq. (6.1) yields

r2 +
Also, substitution of B = r
r2 A

@
r A=
@t

"0

(6.6)

A and Eq. (6.5) into Eq. (6.4) yields

1 @2A
c2 @t2

r r A+

1 @
c2 @t

0 J:

(6.7)

If the Lorenz gauge

r A+

1 @
= 0;
c2 @t

(6.8)

is chosen for r A (i.e., the longitudinal component of the vector potential), then Eqs. (6.6) and
(6.7) are completely decoupled, and reduce, respectively, to
r2
r2

1 @2
c2 @t2
1 @2
c2 @t2

A=

"0

(6.9)

(6.10)

0 J:

If the Coulomb gauge

r A = 0;
is chosen instead, that is, if the longitudinal component of the vector potential is chosen to be zero,
such decoupling cannot be achieved,
;
(6.11)
r2 C =
"0
r2 A

1 @2A
c2 @t2

1 @
r
c2 @t

0 J:

(6.12)

Since the longitudinal current Jl satises the charge conservation law,

@
+ r Jl = 0;
@t
it follows that

1 @ 2
r
c2 @t

0r

(6.13)

Jl ;

(6.14)

and the wave equation for the vector potential in Coulomb gauge involves only the transverse
component of the vector potential,
1 @ 2 At
=
c2 @t2

r2 At

0 Jt :

(6.15)

As briey pointed out in Chapter 3, the scalar potential in Eq. (6.11) is not subject to retardation
due to nite propagation speed of electromagnetic disturbance. The part of the electric eld
generated from the gradient of the scalar potential
E=

@A
;
@t

is also nonretarded because of instantaneous propagation. Physical (retarded) electric eld is contained in the vector potential. In the Lorenz gauge, both potentials are appropriately retarded and
we will therefore employ Lorenz gauge. (As we will see in the following section, the instantaneous
Coulomb eld is in fact cancelled by a term in the retarded transverse vector potential.)
The scalar potential and three cartesian components of the vector potential Ai all satisfy the
wave equation in the form
1 @2
Ai = fi (r; t);
(6.16)
r2
c2 @t2
where fi (r; t) is a source function, either the charge density or the current density. The Greens
function for the scalar wave equation G(r; t) can be found as a solution for the following singular
wave equation,
1 @2
r2
G(r r0 ; t t0 ) =
(r r0 ) (t t0 );
(6.17)
c2 @t2
with the boundary condition that
G = 0 at r = 1 and t =

1:

Once the Greens function is found, the solution to the original wave equation can be written down
in the form of convolution,
Z
Ai (r; t) = G(r r0 ; t t0 )fi (r0 ; t)dV 0 dt0 :
(6.18)
We seek a solution for the Greens function by the method of Fourier transform. Let the Fourier
transform of G(r; t) be g(k; !);
g(k; !) =

G(R; ) =

dV

1
(2 )4

d e
3

d k

i(k R ! )

G(R; );

d!g(k; !)ei(k R

! )

(6.19)

(6.20)

where R = r

r0 ; = t

t0 : Then, Eq. (6.17) in the Fourier-Laplace space (k; !) becomes

!2
c2

k2 +

g(k; !) =

or
g(k; !) =

1;

1
:
(!=c)2

k2

(6.21)

Substitution of g(k; !) into Eq. (6.20) yields

1

G(R; ) =

(2 )4

dk

d!

ei(k R ! )
:
k 2 (!=c)2

Integration over ! can be done easily,

Z

1
1

k2

(Note that with a new variable s =

ic2

i1
i1

i!

(!=c)2

d! = 2

c
sin(kc ):
k

(6.22)

i!; the integral reduces to

Z

es
2 1
ds =
2
2
s + (ck)
ck 2 i

i1
i1

ckes
2
ds =
sin(kc );
2
2
s + (ck)
ck

(6.23)

which is the standard inverse Laplace transform. Causality is appropriately handled in inverse
Laplace transform.) Then,
G(R; ) =

c
(2 )3

dk

d k sin(ck )eikR cos sin ;

where is the polar angle in the k space measured from the direction of R = r
The integrations can be carried out as follows:
G(R; ) =
=
=

(6.24)

r0 ; and R = jr

r0 j :

kdk sin(ck )
eikR cos sin d
(2 )2 0
0
Z 1
c
2 sin(ck ) sin(kR)dk
4 2R 0
c
[ (c
R)
(c + R)] ;
4 R

(6.25)

where in the nal step, use is made of

Z

cos(ak)dk =

(a):

The function

c
(c
R

R) =

1
R

R
c

(6.26)

describes an impulse propagating radially outward at a speed c, and the function

c
1
(c + t) =
R
R

R
c

(6.27)

describes an impulse propagating radially inward. For radiation elds due to a source spatially
limited, only the outgoing solution is physically meaningful and we adopt it as the Greens function,
1
4 R
1
4 jr r0 j

G(R; ) =
=

R
c
jr

t0

r0 j

; (m

):

(6.28)

It is straightforward to check that G(R; t) satises the inhomogeneous wave equation in Eq. (6.17)
if the following identity is recalled,
r2

r0 j

jr

r0 :

Having found the Greens function for the wave equation, the general solution to the inhomogeneous wave equation,
1 @2
r2
Ai = fi (r; t);
(6.29)
c2 @t2
can be written down as
1
Ai (r; t) =
4

dV

dt0

fi (r0 ; t0 )
jr r0 j

t0

jr

r0 j

(6.30)

If the source function f (r; t) is separable in the form,

fi (r; t) = fi (r)e
Eq. (6.30) reduces to
1
Ai (r; t) =
e
4

i!t

i!t

(6.31)

eikjr r j
fi (r0 )dV 0 ;
jr r0 j

(6.32)

where k = !=c: The spatial part of this solution,

Ai (r) =

1
4

eikjr r j
fi (r0 )dV 0 ;
jr r0 j

(6.33)

satises the Helmholtz equation,

r 2 + k 2 Ai =

fi (r):

(6.34)

Corresponding Greens function is

G(r

eikjr r j
;
r)=
4 jr r0 j
0

(6.35)

which satises the Helmholtz equation,

r2 + k 2 G r

r0 =

r0 :

(6.36)

The formulation developed here agrees with that in the preceding Chapter which has been derived
through a qualitative argument based on the retarded nature of electromagnetic disturbance. In
static cases ! = 0; k = 0; we trivially recover the static potentials,
1
(r) =
4 "0
A(r) =

(r0 )
dV 0 ;
jr r0 j
J(r0 )
dV 0 :
jr r0 j

In the Greens function

G(r

r0 ; t

t0 ) =

1
4 jr r0 j

t0

jr

r0 j

(6.37)

the retarded nature of electromagnetic disturbance can be clearly seen. For a source located at
r0 and observer at r; electromagnetic phenomena observed at time t is due to source disturbance
created at the time jr r0 j =c seconds earlier than t:

6.3

Coulomb and Lorenz Gauges

The divergence of the vector potential r A can be assigned an arbitrary scalar function without
aecting the electromagnetic elds E and B: The Maxwells equations do not specify r A: In fact,
potential transformation involving a scalar function ;
A0 = A + r ;

@
;
@t

@A
= E;
@t

(6.38)

does not aect the electromagnetic elds,

E0 =

B0 = r

@A0
=
@t
A0 = r

A = B:

In Lorenz gauge characterized by

r A+
the function

1 @
= 0;
c2 @t

(6.39)

must satisfy the homogeneous wave equation,

r2

1 @2
c2 @t2

= 0;

(6.40)

and in the Coulomb gauge choice r A = 0;

must satisfy Laplace equation

r2

= 0:

(6.41)

In classical electrodynamics, the scalar function can be chosen to be zero because a reasonable
boundary condition L;C ! 0 at r ! 1 for both equations
r2

1 @2
c2 @t2

= 0; r2

= 0;

allows only L;C = 0:

In macroscopic electromagnetic analysis, Lorenz gauge dened by
r A+

1 @
= 0;
c2 @t

is more convenient because all potentials and electromagnetic elds are explicitly retarded. In
contrast, the scalar potential in the Coulomb gauge is not retarded and consequent Coulomb electric
eld is also nonretarded. Such instantaneous elds are clearly unphysical, and should be cancelled
somehow by a counterpart. In this section, it is shown that the electromagnetic elds formulated in
Lorenz gauge, which are all retarded appropriately, are indeed consistent with those in the Coulomb
gauge. The instantaneous Coulomb electric eld emerging in the Coulomb gauge is cancelled.
In Coulomb gauge, the scalar potential satises time dependent Poisson equation,
r2 (r; t) =
Its solution is nonretarded,
(r; t) =

1
4 "0

(r; t)
:
"0
(r0 ; t)
dV 0 ;
jr r0 j

(6.42)

(6.43)

and resultant Coulomb electric eld is also nonretarded,

E(r; t) =
=

r (r; t)
Z
(r r0 ) (r0 ; t) 0
1
dV :
4 "0
jr r0 j3

(6.44)

As we have seen, the vector potential in Coulomb gauge (r A = 0) is transverse and obeys the
wave equation,
1 @2
At =
Jl );
(6.45)
r2
0 Jt =
0 (J
c2 @t2
where Jt is the transverse component of the current density. Note that the longitudinal current Jl
and the charge density are related through the charge conservation law,
@
+ r Jl = 0:
@t

(6.46)

Solution for the transverse vector potential is retarded,

At (r; t) =

1
r0 j

jr

Jt r0 ; t

jr

r0 j

dV 0 :

(6.47)

The electric eld in the Coulomb gauge is thus give by

EC (r; t) =
=

@At
Z @t 0
(r r ) (r0 ; t) 0
1
dV
4 "0
jr r0 j3
r

@
4 @t
0

1
jr

r0 j

J t r0 ; t

jr

r0 j

dV 0 : (6.48)

The instantaneous Coulomb eld (the rst term in the RHS) must be somehow cancelled by a term
stemming from the time derivative of the transverse vector potential for the electric eld to be
consistent with that from Lorenz gauge which guarantees that all elds are retarded.
It is convenient to work with spatial Fourier transforms of the potentials and elds. Since the
inverse Laplace transform of the Fourier-Laplace Greens function
g(k; !) =
is

c2
2

1
1

e
!2

k2

i !

(ck)2

1
;
(!=c)2

d! = c2

(6.49)

sin(ck )
;
ck

(6.50)

the particular solution for an inhomogeneous wave equation

1 d2
+ k 2 A(k; t) = f (k; t);
c2 dt2
can be given in the form of convolution,
A(k; t) =

c
k

sin(ck )f (k; t

)d :

(6.51)

Applying this to the transverse vector potential, we nd

At (k; t) =

c
0
k

sin(ck ) [J(k; t

Jl (k; t

)] d :

(6.52)

Since in the Coulomb gauge, the scalar potential is given by

(k; t) =

(k; t)
;
"0 k 2

(6.53)

the electric eld becomes

ik
(k; t)
"0 k 2
ik
(k; t)
"0 k 2

EC (k; t) =
=

@
At (k; t)
@t
Z 1
c @
sin(ck ) [J(k; t
0
k @t 0

Jl (k; t

)] d :

(6.54)

Noting
Z 1
@
sin(ck )Jl (k; t
)d
@t 0
Z 1
@
=
sin(ck ) Jl (k; t
)d
@
Z0 1
cos(ck )Jl (k; t
)d ;
= ck

(6.55)

and
Jl (k; t) =

ik @
(k; t);
k 2 @t

(6.56)

we nd
Z 1
ik
c @
EC (k; t) =
(k; t)
sin(ck )J(k; t
)d
0
"0 k 2
k @t 0
Z
ik
ik c 1
+
(k;
t)
sin(ck ) (k; t
)d
"0 k 2
"0 k 0
Z
Z 1
ik c 1
c @
=
sin(ck ) (k; t
)d
sin(ck )J(k; t
0
"0 k 0
k @t 0

)d : (6.57)

Note that the instantaneous elds (the rst and third terms in the RHS) exactly cancel each other
and the electric eld is appropriately retarded and consistent with the result in the Lorenz gauge,
EL (r; t) =

1
r
4 "0

(r0 ; ) 0
dV
jr r0 j

where

jr

=t

r0 j

@
4 @t
0

J(r0 ; ) 0
dV ;
jr r0 j

(6.58)

Fourier transform of Eq. (6.58) indeed recovers

EL (k; t) =

ik c
"0 k

sin(ck ) (k; t

c @
0
k @t

)d

sin(ck )J(k; t

)d ;

(6.59)

which is identical to the eld formulated in the Coulomb gauge.

The magnetic eld is generated by the transverse vector potential,
B=r
(note that r

A=r

At ;

(6.60)

Al = 0 by denition) and is thus explicitly retarded irrespective of the choice of the

9

gauge.

6.4

Elementary Spherical Waves

For a current source oscillating at a frequency !; the vector potential can be calculated from Eq.
(6.33),
Z ikjr r0 j
e
0
i!t
e
J(r0 )dV 0 :
A(r;t) =
4
jr r0 j
Here, we wish to expand the spatial Greens function,
0

eikjr r j
;
r)=
4 jr r0 j
0

G(r

(6.61)

in terms of spherical harmonics as done in static cases,

Static

1
4 jr r0 j
X 1 (r0 )l
=
Ylm ( ; )Ylm ( 0 ;
2l + 1 rl+1
:

r0 ) =

G(r

); r > r0 :

(6.62)

l;m

Evidently, the objective of the harmonic expansion is to identify electric and magnetic multipole
components in the radiation electromagnetic elds. The Greens function
r0 ) =

G(r

eikjr r j
;
4 jr r0 j

satises the singular Helmholtz equation,

r2 + k 2 G =

(r

r0 ):

(6.63)

We wish to construct a series expansion of the Greens function in terms of solutions to the homogeneous Helmholtz equation,
r2 + k 2 f (r) = 0:
(6.64)
Let the function f (r) be separated in the form
f (r) = R(r)F ( ; ):
Substitution into Eq. (6.64) yields
2 d
d2
+
+ k2
2
dr
r dr
1 @
sin @

sin

@
@

l(l + 1)
r2

R(r) = 0;

1 @2
+ l(l + 1) F ( ; ) = 0;
sin2 @ 2
10

(6.65)

(6.66)

where l(l + 1) is a separation constant. The angular function F ( ; ) is unchanged from the static
case,
F ( ; ) = Ylm ( ; ):
(6.67)
To nd solutions for the radial function R(r); let us assume
u(r)
R(r) = p :
r

(6.68)

Substitution into Eq. (6.65) yields the following equation for u(r);
1 d
d2
+
+ k2
2
dr
r dr

(l + 21 )2
u(r) = 0;
r2

(6.69)

whose solutions are Bessel functions of half-integer order,

u(r) = J

1 (kr);
l+ 2

1 (kr):
l+ 2

(6.70)

It is customary to introduce spherical Bessel functions dened by

jl (kr) =
where the numerical factor

r J

1 (kr)
l+ 2

kr

nl (kr) =

r N

1 (kr)
l+ 2

p
kr

(6.71)

=2 is to let both functions approach the form of spherical wave,

jjl (kr)j ; jnl (kr)j !

1
for kr
kr

1:

(6.72)

Noting that the asymptotic forms of the ordinary Bessel functions are
Jn (x) !

2
cos x
x

2n + 1
4

(6.73)

Nn (x) !

2
sin x
x

2n + 1
4

(6.74)

we observe that jl (x) and nl (x) asymptotically approach

jl (x) !

1
cos x
x

l+1
2

; x

(6.75)

nl (x) !

1
sin x
x

l+1
2

; x

1:

(6.76)

In radiation analyses, it is convenient to introduce spherical Hankel functions dened by

(1)

First kind: hl (x) = jl (x) + inl (x);

11

(6.77)

(2)

Second kind: hl (x) = jl (x)

inl (x):

(6.78)

1;

(6.79)

The asymptotic forms of these functions are:

(1)

hl (x) ! ( i)l+1
(2)

hl (x) ! il+1

eix
; x
x
ix

; x

1;

(6.80)

which describe outgoing and incoming spherical waves, respectively.

Having found elementary spherical harmonic solutions for the homogeneous Helmholtz equation,
we are now ready to construct the Greens function in terms of those elementary solutions. Since
(1)
(2)
any linear combinations of the four radial functions jl (kr); nl (kr); hl (kr) and hl (kr) satisfy the
Helmholtz equation, and the Greens function should be bounded everywhere, we may assume
0

G(r; r ) =

1 X
l
X

Alm jl (kr0 )hl (kr)Ylm ( ; )Ylm ( 0 ;

1 X
l
X

Alm jl (kr)hl (kr0 )Ylm ( ; )Ylm ( 0 ;

(1)

); for r > r0 ;

(6.81)

); for r < r0 ;

(6.82)

l=0 m= l

and
G(r; r0 ) =

(1)

l=0 m= l

where Alm is an expansion coe cient of the (l; m) harmonic and continuity on the surface r = r0
is imposed. Note that for a small argument, only jl (x) remains bounded,
jl (x) '

xl
;
(2l + 1)!!

1:

(6.83)

The spherical Bessel function of the second kind nl (x) diverges at small x as
nl (x) '

(2l 1)!!
;
xl+1

1:

(6.84)

(2)

The radially converging (incoming) solution hl (kr) is discarded because we are interested in
radiation of electromagnetic eld from localized sources.
The coe cient Alm can be determined through the familiar procedure exploiting the discontinuity in the radial derivative. Substituting Eqs. (6.81) and (6.82) into the original equation
(r2 + k 2 )G =

(r

(r r0 )
(
rr0 sin

r0 ) =

) (

);

multiplying both sides by Yl0 m0 ( ; ); and integrating the result over the entire solid angle by noting
the orthogonality of the harmonic functions,
Z
Ylm ( ; )Yl0 m0 ( ; )d = ll0 mm0 ;
(6.85)
12

we obtain

d2
2 d
+
+ k2
dr2 r dr

Alm
where

l(l + 1)
r2

gl (r; r0 ) =

(r r0 )
;
rr0

(6.86)

8
0 (1)
0
>
< jl (kr )hl (kr); r > r ;

gl (r; r0 ) =

>
:

(1)
jl (kr)hl (kr0 );

(6.87)

r < r0

The derivative of the radial function gl (r; r0 ) is discontinuous at r = r0 and thus second order
derivative yields the singularity compatible with the RHS,
1
d2
gl (r; r0 ) = ik
(r
dr2
(kr)2

r0 );

(6.88)

where use is made of the Wronskian of the spherical Bessel functions,

(1)

jl (x)[hl (x)]0

(1)

jl0 (x)hl (x) = i jl (x)n0l (x)

jl0 (x)nl (x) =

i
:
x2

(6.89)

We thus nd a rather simple result,

Alm = ik;

(6.90)

and the desired spherical harmonic expansion of the Greens function is

8
1 P
l
P
>
(1)
>
ik
jl (kr)hl (kr0 )Ylm ( ; )Ylm ( 0 ;
>
>
>
l=0 m= l
<

r0 j

eikjr
=
4 jr r0 j >
>
1 P
l
>
P
(1)
>
>
jl (kr0 ) hl (kr)Ylm ( ; )Ylm ( 0 ;
: ik

r < r0 ;

);

(6.91)
0

); r > r0 :

l=0 m= l

In the limit of static elds k ! 0 ( that is, ! ! 0); we indeed recover

8
1 P
l
P
1 (r0 )l
>
>
Y ( ; )Ylm ( 0 ;
>
>
l+1 lm
>
2l
+
1
r
< l=0 m= l

1
=
4 jr r0 j >
>
1 P
l
>
P
>
>
:
l=0 m=

rl

1
Y ( ; )Ylm ( 0 ;
0 )l+1 lm
2l
+
1
(r
l

r > r0

);

(6.92)
0

); r < r0 :

The expansion in Eq. (6.91) allows us to write the vector potential in the form
A(r) =

= ik

Z
0

eikjr r j
J(r0 )dV 0
0
jr r j

1 X
l
X

(1)

hl (kr)Ylm ( ; )

l=0 m= l

13

J(r0 )jl (kr0 )Ylm ( 0 ;

)dV 0 :

(6.93)

For cartesian components Ai (r); (i = x; y; z); Eq. (6.93) gives

Ai (r) = ik

1 X
l
X

h(kr)Ylm ( ; )

l=0 m= l

Ji (r0 )jl (kr0 )Ylm ( 0 ;

)dV 0 :

However, for non-cartesian components (e.g., components in the spherical coordinates), it is necessary to single out the component of the current density J(r) in the same direction as the vector
potential A at the observing position
A(r) = ik

1 X
l
X

(1)
hl (kr)Ylm (

; )eA

l=0 m= l

eA J(r0 )jl (kr0 )Ylm ( 0 ;

)dV 0 ;

(6.94)

where eA is the unit vector in the direction of A; eA = A=A: The problem with this representation
is that the direction of the vector potential eA is not known a priori. It is more convenient if we can
nd basic electromagnetic eld vectors which are normal to each other everywhere. The Transverse
Electric (TE) and Transverse Magnetic (TM) eigenvectors will serve exactly for this purpose and
form a convenient dyadic for radiation electromagnetic elds.

6.5

TE (Transverse Electric) and TM (Transverse Magnetic) Base

Vectors

In electromagnetic wave problems, boundary conditions are usually imposed on the electric and
magnetic elds E and B; rather than the potentials and A: (Note that the electromagnetic elds
are gauge invariant, while the potentials are not.) Also, in experiments, what is usually measured
is the electric eld. For these reasons, it is desirable to formulate spherical harmonic expansion of
the electric eld.
In Chapter 3 on magnetostatics, we saw that r A = 0 allows us to write the vector potential
in the form
A = r F;
where F is a vector eld. In the Lorenz gauge, the divergence of A does not vanish,
r A=

1 @
;
c2 @t

which suggests
A = rf + r

F;

where f is a scalar eld. However, the eld f does not contribute to the magnetic eld B = r A
since r rf
0 and we may continue to assume A = r F even for time varying elds. (The
scalar eld f does aect the formulation of the electric eld.) The vector eld F may further be

14

decomposed into radial and transverse components as follows,

F=r +r
where
and

and
,

r ;

(6.95)

are scalar functions. If the vector potential satises the Helmholtz equation, so do
(r2 + k 2 ) = 0;

(r2 + k 2 ) = 0;

(6.96)

and thus can be expanded in spherical harmonics. (Note that Laplacian r2 and operator r
commute.) The magnetic eld in terms of the scalar functions is given by
B = r
= r

A=r

(r

r )

r ) + k2 r

(r

rr2

r ;

(6.97)

where use is made of

r (r

r ) = 0:

(6.98)

The second term in the RHS is evidently transverse to r, that is, the scalar eld is a generating
function for transverse magnetic (TM) modes.
The scalar eld generates transverse electric (TE) modes, since in the current free region, the
electric eld is given by
i!
E = r B
c2
=
r rr2
= k 2 [r

r +r

(r
(r

rr2 )
r )] :

(6.99)

The TE and TM modes are the desired base vectors because vector functions r r and r (r r )
are normal to each other. To prove this, let us assume
(r) =
Expanding
r

[gl (r)r

Alm gl (r)Ylm ( ; );

rYlm ] = rgl (r)

(r) =

Blm gl (r)Ylm ( ; ):

(6.100)

[r

rYlm ( ; )] + gl (r)r (r rYlm )

l(l + 1)
= rgl (r) [r rYlm ( ; )] gl (r)
Ylm er gl (r)rYlm
r
l(l + 1)
d
=
gl (r)
Ylm er
[rgl (r)] rYlm ;
(6.101)
r
dr

we see that the two vectors r rYlm and r [gl (r) r rYlm ] are indeed normal to each other,
since (r rYlm ) r = 0 and (r rYlm ) rYlm = 0: (The product
r (r

rYlm ) rYlm =
15

2im

d
jYlm j2 ;
d

evidently does not vanish. Interpret this.)

The operator
r rYlm ( ; ) = iLYlm ( ; );

L=

ir

r;

(6.102)

is the angular momentum operator and its explicit form is

r

rYlm ( ; ) =

1 @Ylm
@Ylm
+e
:
sin @
@

(6.103)

Noting
@e
=
@

sin er

cos e ;

we can readily show that

(r

r) (r

1 @
sin @

r)2 Ylm =

rYlm ) = (r

sin

@Ylm
@

1 @ 2 Ylm
=
sin2 @ 2

l(l + 1)Ylm ; (6.104)

or
L2 Ylm = l(l + 1)Ylm :

(6.105)

Other useful properties are:

r = er

(r

@
@r

i
r
r2

rYl0 m0 ) d

rYlm ) (r

r (r

(6.106)

L;
= l(l + 1)

i y

@
@z

@
@y

; Ly =
L

i z

Lx

@
@x

(6.107)

rYlm ) = 0;

(6.108)

L2 LYlm = LL2 Ylm ; r2 LYlm = Lr2 Ylm ; L

Lx =

ll0 mm0 ;

iLy = e

@
@z
i

p
l (l + 1)
p
L Ylm ( ; ) = l (l + 1)
L+ Ylm ( ; ) =

; Lz =

LYlm = iLYlm ;

i x

@
@y

@
@
+ i cot
@
@

@
@x

(6.109)
i

@
;
@

(6.110)
(6.111)

m (m + 1)Yl;m+1 ( ; ) ;

(6.112)

m (m

(6.113)

1)Yl;m

1(

; );

Lz Ylm ( ; ) = mYlm ( ; ) :

(6.114)

We are now ready to expand the raw form of the vector potential
A(r) =
=

0 ik

J(r0 )eikjr r j 0
dV
jr r0 j

1 X
l
X

(1)
hl (kr)Ylm (

l=0 m= l

16

; )

jl (kr0 )Ylm ( 0 ;

)J(r0 )dV 0 ;

(6.115)

in terms of the TE and TM base vectors. For TE modes, the vector potential should be in the form
ATE = r

r :

(6.116)

Therefore, it is necessary to identify TE component of the source current J(r0 ); which can be
eected by
TE
Jlm
(r) = jl (kr)(r rYlm ) J(r):
(6.117)
The TE component of the vector potential can thus be assumed in the form
ATE (r) =

0 ik

1 X
l
X

(1)

alm hl (kr)r

l=1 m= l

rYlm ( ; )

jl (kr0 )[r0

r0 Ylm ( 0 ;

)] J(r0 )dV 0 ; (6.118)

where alm is expansion coe cient. Note that in the summation over l; the monopole term l = 0
vanishes because rY00 = 0: It is convenient to choose alm so that
Z
alm (r rYlm ) (r rYlm )d = 1;
(6.119)
or
alm =

1
:
l(l + 1)

(6.120)

Corresponding TE base vector is

1
uTE
r
lm (r) = p
l(l + 1)

which is normalized as

uTE
lm (r)

uTE
lm (r)

rYlm ( ; );

(6.121)

(6.122)

= 1:

The TE component of the vector potential is therefore given by

A

TE

(r) =

0 ik

1 X
l
X

l=1 m= l

0 ik

1
(1)
h (kr)r
l(l + 1) l

1 X
l
X

l=1 m= l

1
(1)
h (kr)r
l(l + 1) l

rYlm ( ; )

rYlm ( ; )

jl (kr0 )[r0
Z

jl (kr0 ) r0

r0 Ylm ( 0 ;
J(r0 )

)] J(r0 )dV 0

0
r0 Ylm ( 0 ; (6.123)
)dV 0 ;

The current density J(r0 ) may be separated into the conduction current Jc and magnetization
current Jm = r M (r),
J(r) = Jc (r) + r M (r) ;
(6.124)
where M (r) is the magnetic dipole moment density.
For the TM modes, it is convenient to let the TM base vector have the same dimensions as TE

17

base vector. For this purpose, we rewrite the decomposition of the vector potential in the form
A=r

1
r + r
k

(r

r ):

(6.125)

Following the same procedure developed for the TE modes, we nd

ATM (r) = i

Z
1
l
1X X
1
(1)
r
[h
(kr)r
rY
(
;
)]
r0 [jl (kr0 )r0 r0 Ylm ( 0 ;
lm
0
l
k
l(l + 1)

)] J(r0 )dV 0 :

l=1 m= l

(6.126)

Corresponding TM base vector in the radiation zone is

which is also normalized as

1
er
uTM
lm (r) = p
l(l + 1)
I

uTM
lm (r)

[r

uTM
lm (r)

rYlm ( ; )];

(6.127)

= 1:

(6.128)

)] J(r0 )

(6.129)

Note that in Eq. (6.126),

r0

[jl (kr0 )r0

r0 Ylm ( 0 ;

is the TM component of the current density.

For a small source kr0
1; the spherical Bessel function jl (kr0 ) can be approximated by
jl (kr0 ) '

(kr0 )l
:
(2l + 1)!!

(6.130)

Then, the moment for the TE modes becomes (the primes are omitted for brevity)
Z
=
=

[jl (kr)r

rYlm ( ; )] Jc (r)dV
Z
kl
rl (r Jc ) rYlm dV
(2l + 1)!!
Z
kl
r(rl Ylm ) (r Jc )dV:
(2l + 1)!!

(6.131)

This contains the current circulation r Jc which produces magnetic multipole moments. The
magnetic multipole moment is in general dened by
Z
1
r(rl Ylm ) (r Jc )dV:
(6.132)
mlm =
l+1
In terms of the magnetic moment mlm ; the integral may be rewritten as
Z

[jl (kr)r

rYlm ( ; )] Jc (r)dV =

18

(l + 1)k l
mlm :
(2l + 1)!!

(6.133)

The contribution from the magnetization current r M can be calculated in a similar manner,
Z
0
mlm =
rl Ylm r MdV;
and the (l; m) component of the TE vector potential is given by
ATE
lm (r) =

0i

k l+1
(1)
(mlm + m0lm )hl (kr)r
l(2l + 1)!!

rYlm ( ; ):

(6.134)

The radiation power associated with the vector potential can readily be calculated as follows:
Z
2
TE
2
Z HTE
d ;
(6.135)
Plm = r
lm
where the magnetic eld HTE
lm (r) in the radiation region kr
1

HTE
lm (r) '

ATE
lm

ik

k l+1 ( i)l (mlm + m0lm ) 1 i(kr

e
l(2l + 1)!!
r

=
Then
HTE
lm

1 is approximately given by

Noting

!t)

k 2(l+1)
2
mlm + m0lm jr
2
[l(2l + 1)!!]
Z

rYlm ( ; )j2 d

jr

[r

rYlm ( ; )]:

rYlm ( ; )j2 :

= l(l + 1);

(6.136)

(6.137)

(6.138)

we nd the radiation power in the form

TE
Plm

l + 1 k 2(l+1)
jmlm + m0lm j2
"0 l [(2l + 1)!!]2
0

1 4 l + 1 k 2(l+1)
jmlm + m0lm j2
4 "0 c l [(2l + 1)!!]2

(6.139)

1
l + 1 k 2(l+1)
2
4 c
jmG + m0G
lm j ;
4 "0
l [(2l + 1)!!]2 lm

(6.140)

where the latter expression facilitates comparison with the formulation in the Gaussian unit system
in which the magnetic multipole moment is dened by
Z
1 1
mG
=
r(rl Ylm ) (r J)dV;
(6.141)
lm
cl+1
Electric multipoles radiate TM modes. This may be seen from the TM component of the current

19

density,
Z
=

[jl (kr)r

rYlm ( ; )] J(r)dV

jl (kr)[r rYlm ( ; )] r JdV

Z
=
r [jl (kr)rYlm ] r JdV
Z
=
jl (kr)Ylm r [r (r J)]dV
Z
=
jl (kr)Ylm r r(r J) r2 J dV
Z
Z
d
=
Ylm [rjl (kr)]r JdV k 2 jl (kr)Ylm r JdV;
dr

(6.142)

where it is noted that the function jl (kr)Ylm ( ; ) satises the Helmholtz equation,
r2 + k 2 jl (kr)Ylm ( ; ) = 0:
Again, if the current is separated into the condution current and magnetization currents, J =
Jc + r M; Eq. (6.142) can be rewritten as
Z

d
Ylm [rjl (kr)]r JdV
dr
Z
d
=
Ylm [rjl (kr)]r Jc dV
dr
For a small source kr0
dominant. Recalling

jl (kr)Ylm r JdV

jl (kr)Ylm r Jc dV + k

jl (kr)Ylm r (r

M) dV:
(6.143)

1, the rst term in the RHS containing electric multipole moment is

@
+ r Jc = 0;
@t

and
jl (kr) '

(kr)l
(2l + 1)!!

for kr

1;

we may approximate Eq. (6.143) by

Z

d
Ylm [rjl (kr)]r Jc dV
dr
(l + 1)k l+1
0
' ic
qlm + qlm
;
(2l + 1)!!
where the second term is of order ka

jl (kr)Ylm r Jc dV + k

jl (kr)Ylm r (r

M) dV
(6.144)

1 (a being the source size) and thus ignorable, qlm is the

20

electric multipole moment dened earlier,

qlm =

rl Ylm ( ; ) (r)dV;

(6.145)

0 is an eective electric quadrupole moment created by the magnetic dipole moment density,
and qlm
0
qlm

ik
(l + 1) c

rl Ylm ( ; )r (r

M) dV:

(6.146)

The radiation vector potential of TM mode is thus given by

ATM
lm (r) =

0c

kl
qlm r
l(2l + 1)!!

(1)

hl (kr)r

i
rYlm ( ; ) ;

(6.147)

and corresponding radiation power is

TM
Plm

3
0c

1
k 2(l+1) l + 1
0 2
qlm + qlm
:
4 c
2
4 "0
[(2l + 1)!!]
l

6.6

k 2(l+1) l + 1
0 2
jqlm + qlm
j
[(2l + 1)!!]2 l
(6.148)

Radiation of Angular Momentum

In chapter 5, radiation of angular momentum from a circulating charge was discussed. In general,
if a charge undergoes circular motion, it radiates (loses) angular momentum as well as energy, for
the two quantities are intimately related. In this section, a general formulation for radiation of
angular momentum is given. Since the momentum ux density associated with electromagnetic
elds is given by
1
E H ;
(6.149)
c
The angular momentum ux density associated with electromagnetic elds is
1
r
c

(E

H ):

(6.150)

In the radiation zone, this must be proportional to 1=r2 if radiation of angular momentum accompanies radiation of energy. Therefore, terms proportional to 1=r3 in the Poynting ux, which have
been ignored in the calculation of radiation power, should be retained.
As a concrete example, let us consider TM modes due to electric multipoles. The vector potential
of TM mode is
ATM
lm (r) =

0c

kl
qlm r
l(2l + 1)!!

21

(1)

hl (kr)r

i
rYlm ( ; ) :

(6.151)

Corresponding magnetic and electric elds are

1

HTM
lm (r) =

ATM
lm

k l+2
(1)
qlm hl (kr)r
l(2l + 1)!!

rYlm ( ; );

(6.152)

and
i
r HTM
lm
!"0
ic
k l+2
qlm r
!"0 l(2l + 1)!!

ETM
lm (r) =
=

(1)

[hl (kr)r

rYlm ( ; )];

(6.153)

where use is made of the identity

h
(1)
hl (kr)r

i
rYlm ( ; ) = 0:

The expressions for the electromagnetic elds are exact. The radiation ux of angular momentum
associated with an (l; m) mode is

=
=
=

1
r (Elm Hlm )TM
c
ic
k 2(l+2)
jqlm j2 r
!"0 [l(2l + 1)!!]2

(1)

rYlm ( ; ))]

(1)

rYlm ( ; )] hl

[r

(hl (kr)r

k 2(l+2)
ic
jqlm j2 r r
!"0 [l(2l + 1)!!]2

[hl (kr)r

2
ic
k 2(l+2)
(1)
jqlm j2 l(l + 1) hl (kr) Ylm ( ; )r
2
!"0 [l(2l + 1)!!]

In radiation zone kr

(1)

[hl
(1)

(kr)r

(kr)r

i
rYlm ( ; )]
rYlm ( ; )

rYlm ( ; ):

(6.154)

1; this reduces to
ic
k 2(l+1) l + 1
jqlm j2 Ylm ( ; )r
!"0 [(2l + 1)!!]2 l

rYlm ( ; )

1
;
r2

(6.155)

and the integration over the solid angle yields the rate of angular momentum radiation,
dLlm
1
k 2(l+1) l + 1
m
=
4 c
jqlm j2 ez ;
2
dt
4 "0
[(2l + 1)!!]
l
!

(6.156)

where use is made of

Z
=
=

Ylm ( ; )r rYlm ( ; )d
Z
im Ylm ( ; )Ylm ( ; )d ez

imez ;

22

(6.157)

Comparing with the radiation power given in Eq. (6.148), we see that they are related through a
simple ratio,
L_ z
m
= :
(6.158)
P
!
Evidently, this is not consistent with the familiar quantum mechanical result,
p
p
~ l(l + 1)
l(l + 1)
L_ z
=
=
:
P
~!
!

(6.159)

The discrepancy is not surprising because the quantum mechanical formula is for a single photon while the result obtained in classical electrodynamics pertains to (innitely) many photons.
Quantum mechanical formula for N photons is
L_ z
=
P

p
N 2 m2 + N l(l + 1)
N!

and the classical result can be recovered in the limit of N

6.7

m2

(6.160)

1:

Some Examples

Example 1 Spherical Dipole Antenna

Figure 6-1: Spherical antenna.

Two ideally conducting hemispheres of radius a insulated from each other at the equator with
a small gap are connected to an oscillating voltage source V0 e i!t : In this case, the surface current
ows in the direction and creates an azimuthal magnetic elds B : The resultant radiation elds

23

are thus TM (Transverse Magnetic) and we may assume the magnetic eld in the form
B=r

r ;

(6.161)

where
is a scalar function. Since the system is axially symmetric without dependence on the
azimuthal angle ; the scalar function can be expanded in spherical harmonics as
(r; ) =

(1)

al hl (kr)Pl (cos );

r > a;

(6.162)

l 1

which yields the magnetic eld

B = r
=

r
X

(1)

al hl (kr)Pl1 (cos )e ;

r > a:

(6.163)

Pl1 (cos ):

(6.164)

Note that

d
Pl (cos ) =
d

d
Pl (cos ) =
d(cos )

sin

The electric eld associated with the magnetic eld can be found from the Maxwells equation,
"0

@E
=r
@t

(6.165)

B;

which yields
E(r) =

1 @
ic2
(sin B )er
! r sin @

1 @
(rB )e
r @r

(6.166)

Substituting B (r) in Eq.(6.163) yields the transverse component of the electric eld,
E (r; ) =

ic2 X
l (1)
al hl (kr)
!
r

(1)
1 (kr)

khl

Pl1 (cos );

(6.167)

where the following recurrence formula for the spherical Bessel functions has been used,
d (1)
h (x) =
dx l

l (1)
h (x)
x l
(1)
1 (x)

= hl

(1)

hl+1 (x)
l + 1 (1)
h (x):
x l

(6.168)

Since the sphere is ideally conducting, the component of the electric eld should vanish except
at the gap. Therefore, at the sphere surface r = a; the electric eld may be approximated by
E (r = a; ) =

V0
a

24

(6.169)

which satises the obvious requirement that

a

E (a; )d = V0 :

(6.170)

Then, for arbitrary angle ; the following relation must hold

V0
a

ic2 X
l (1)
al hl (ka)
!
r

(1)
1 (ka)

khl

Pl1 (cos ):

(6.171)

Multiplying both sides by Pl1 (cos ) sin and integrating over ; we readily nd the expansion
coe cient al ;
Pl1 (0)
! 2l + 1
V0 ;
(6.172)
al = 2
ic 2l(l + 1) lh(1) (ka) kah(1) (ka)
l

where

l 1

8
l 1
l!!
>
>
2
(
1)
; l odd
>
<
(l
1)!!
1
Pl (0) =
>
>
>
: 0;
l even

(6.173)

and use has been made of the integral

Z

1
1

Pl1 (x)Pl10 (x)dx =

2
l(l + 1)
2l + 1

ll0 :

(6.174)

The disappearance of even harmonics is expected because of the up-down anti-symmetry of the
problem. (Note that Pl1 (x) is an odd function of x if l is odd.)
The surface current density Js on the sphere surface can be found from the boundary condition
for the magnetic eld,
Js = n
=
=

H (r = a; )
H (a; )e
1 X
(1)
al hl (ka)Pl1 (cos )e :

(6.175)

0 l 1

The current at the gap

= =2 where the voltage V0 appears is

I = 2 aJs ( = =2)
2 aX
(1)
al hl (ka)Pl1 (0)
=
0 l 1

(1)
hl (ka)
2 a! X 2l + 1
= i
P 1 (0)
2
2l(l + 1) lh(1) (ka) kah(1) (ka) l
0c
l 1

l 1

25

V0 :

(6.176)

This denes the radiation admittance of the spherical dipole antenna,

Y

I
V0
r X
(1)
kahl (ka)
"0
2l + 1
= i
Pl1 (0)
(1)
(1)
l(l
+
1)
0
lh (ka) kah (ka)
=

l 1

(6.177)

l 1

A spherical antenna has limited practical applications. However, the procedure developed in
this example is applicable to more practical problems such as half wavelength dipole antennas. A
thin rod antenna can be analyzed rigorously using the prolate spheroidal coordinates.
Example 2 Circular Loop Antenna

Figure 6-2: Circular loop antenna. ka is arbitrary.

Let us consider a circular oscillating current I0 e
a: The current density may be written as
J = I0 e

i!t

(r

i!t

carried by a thin conductor ring of radius

a)
a

e :

!
As shown in Chapter 5, if the ring radius is much smaller than the wavelength ka = a
c
problem reduces to radiation by a magnetic dipole with a dipole moment
mz (t) = a2 I0 e

i!t

(6.178)
1; the

For an arbitrary value of !a=c; higher order multipole elds must be retained. Since the loop
current radiates TE modes, the TE vector potential in Eq. (6.118) can be directly applied with a
result
X
1
(1)
h (kr)r rYl0 ( ; )Ml ;
(6.179)
ATE (r) = 0 ik
l(l + 1) l
l

26

where
Ml =

rYl0 ( 0 )] J(r0 )dV 0 ;

jl (kr0 )[r0

(6.180)

is the TE component of the current density. Note that because of the axial symmetry, only m = 0
components are nonvanishing. Since
r0

dYl0
e 0
dr0
2l + 1 1
Pl (cos 0 )e 0 ;
4

rYl0 ( 0 ) =
=

we nd
Ml =

2 aI0

2l + 1
jl (ka)Pl1 (0):
4

(6.181)

The vector potential is therefore given by

ATE (r) =

0 I0 ka

X 2l + 1
(1)
jl (ka)hl (kr)Pl1 (0)Pl1 (cos )e ;
2l(l + 1)

(6.182)

and the magnetic eld in the radiation zone kr

H(r) '
=

ik

1 is

ATE

akI0

eikr X
2l + 1
( i)l+1
jl (ka)Pl1 (0)Pl1 (cos )e :
r
2l(l + 1)

(6.183)

l 1

The radiation power associated with the l-th harmonic mode is

Z
l
2
P = r Z0 jHl j2 d
Z
2l + 1 2
2
[Pl1 (cos )]2 sin d
= Z0 (akI0 )2
jl (ka)Pl1 (0) 2
2l(l + 1)
0
2l
+
1
2
=
Z0 (akI0 )2 jl (ka)Pl1 (0)
; l = 1; 3; 5;
l(l + 1)
In the long wavelength limit ka

(6.184)

1; the dipole term (l = 1) is dominant, and we recover

P

(ka)2
6

Z0 (akI0 )2

1 2! 4 m2
;
4 "0 3c5

where m = a2 I0 is the magnetic dipole moment of the ring current.

27

(6.185)

6.8

Spherical Harmonic Expansion of a Plane Wave

A plane wave of planar polarization is characterized by constant amplitudes of electromagnetic

elds and unidirectional propagation assumed here in the z-direction,
E0 eikz ex ;

H0 eikz ey :

(6.186)

The purpose of this section is to decompose the plane wave into spherical harmonics. Once achieved,
such a representation will be very useful in analyzing scattering of electromagnetic wave by an object
placed in the plane wave. Scattered waves are evidently no longer plane waves but they consist
of many (often innitely many) spherical harmonic waves. A key observation to be made is that
there is one-to-one correspondence between spherical harmonic component in the incident plane
wave and the one in the scattered wave. This is because a scattered harmonic mode characterized
by mode numbers (l; m) can only be produced by a mode in the incident wave having exactly the
same angular dependence. For example, TM (l; m) harmonic component in the scattered wave is
generated by the same TM harmonic component contained in the incident wave.

Figure 6-3: Scattered electromagnetic waves consist of spherical waves. The incident plane wave
can be decomposed into spherical waves to facilitate analysis.
The rst step is to expand the propagator function eikz = eikr cos in terms of spherical harmonics. This can be eected by considering the limiting case of the scalar Greens function,
1

XX
eikjr r j
(1)
G(r; r ) =
=
ik
jl (kr)hl (kr0 )Ylm ( ; )Ylm ( 0 ;
0
4 jr r j
0

);

r0 > r:

(6.187)

l=0 m=l

(1)

In the limit of r0 ! 1; that is, when the source is far away, hl (kr0 ) approaches
(1)
hl (kr0 )

! ( i)

28

l+1 e

ikr0

kr0

(6.188)

Noting also
k r0 r ! kr0
we nd
ik r

=4 i

1 X
l
X

k r;

r0

r;

( i)l+1 jl (kr)Ylm ( ; )Ylm ( 0 ;

);

(6.189)

l=0 m= l

where the angles ( 0 ; 0 ) are those of the wavevector k = (k; 0 ;

wave propagating in the z-direction, it follows that 0 = 0 and
Plm (1) =

0
0

): Since we have assumed a plane

becomes irrelevant. Noting

m0 ;

and taking the complex conjugate of Eq. (6.189) yields the following identity,
eikz = eikr cos =

1
X

il (2l + 1)jl (kr)Pl (cos ):

(6.190)

l=0

The electric eld assumed to be in the x-direction can be converted into components in the
spherical coordinates as
E0 ex = E0 r(r sin cos )
= E0 (sin cos er + cos cos e

sin e ):

(6.191)

This eld can be represented as a sum of TE and TM modes. To eect such a representation, we
assume the following expansion,
E0 eikr cos ex =

alm jl (kr)r

lm

rYlm ( ; ) +

1X
blm r
k

[jl (kr)r

lm

rYlm ( ; )] ;

or substituting the expansion in Eq. (6.190) for eikr cos in the LHS,
E0

1
X

il (2l + 1)jl (kr)Pl (cos )(sin cos er + cos cos e

l=0
1
l
XX

alm jl (kr)r

l=1 m=l

rYlm ( ; ) +

1X
blm r
k

[jl (kr)r

lm

sin e )

rYlm ( ; )] ;

(6.192)

where alm and blm are expansion coe cients for TE and TM modes, respectively. The TE coe cient
alm can be determined by multiplying both sides by r rYlm ( ; ) and integrating the result over
the solid angle. Exploiting the following orthogonality relationships,
Z
[r rYlm ( ; )] [r rYl0 m0 ( ; )]d = l(l + 1) ll0 mm0 ;
[r

rYlm ( ; )] r

[jl (kr)r

29

rYlm ( ; ) = 0;

we nd
alm

il (2l + 1)
= E0
l(l + 1)

The presence of cos and sin

ishing. For m = +1; we nd
al;1

Pl (cos )(cos cos e

sin e ) (r

functions in the integral makes only m =

il (2l + 1)
= E0
l(l + 1)

Pl (cos )(cos cos e

sin e ) (r

rYlm )d :

(6.193)

1 components nonvan-

rYl;1 )d ;

(6.194)

where
Yl;1 ( ; ) =
=

Yl;
s

1(

; )

2l + 1
P 1e
4 l(l + 1) l

(6.195)

Since
Z
=

Pl (cos )(cos cos e

sin e ) (r rYl;1 )d
s
Z
dP 1 (cos )
2l + 1
cos
i
Pl (cos )
Pl1 (cos ) + l
4 l(l + 1) 0
sin
d

sin d ;

(6.196)

and
Z

dPl1 (cos )
Pl (cos ) sin d
d
0
Z
d
Pl1 (cos ) [Pl (cos ) sin ]d
d
Z
Z0
[Pl1 (cos )]2 sin d ;
Pl1 (cos )Pl (cos ) cos d +

=
=

(6.197)

the integral reduces to

Z
=

Z0

Pl (cos )

dP 1 (cos )
cos
Pl1 (cos ) + l
sin
d

sin d

[Pl1 (cos )]2 sin d

2
l(l + 1):
2l + 1

(6.198)

Then, nally,
al;1 =

il+1

30

(2l + 1)
E0 :
l(l + 1)

(6.199)

The coe cient al;

can be found in a similar manner,

al;

il+1

= al;1 =

(2l + 1)
E0 :
l(l + 1)

(6.200)

The appearance of m = 1 components in the spherical harmonic expansion of a plane wave is

understandable, for a plane wave can be decomposed into two circularly polarized waves of opposite
helicity.
The coe cient blm of the TM mode can be found in a similar manner. It is more convenient to
work with the magnetic eld,
1
r E
i!
1
1 X
al; 1 r
i!

B =
=

[jl (kr)r

l=1

1
k X
bl;
1] +
i!

rYl;

1 jl (kr)r

l=1

rYl;

1;

(6.201)

where use is made of the identity,

r

[jl (kr)r

rYl;

1]

= rr [jl (kr)r

= 0 + k 2 jl (kr)r

rYl;

1]

rYl;

1:

r2 [jl (kr)r

rYl;

1]

(6.202)

The magnetic eld associated with the plane wave is

B0 eikr cos ey =

E0 X l
i (2l + 1)jl (kr)Pl (cos ) (sin sin er + cos sin e + cos e ) :
c
l

Then,
1
E0 X l
i (2l + 1)jl (kr)Pl (cos ) (sin sin er + cos sin e + cos e )
c
l=1

1
1 X
al;
i!

1r

l=1

Multiplying both sides by r

bl;

[jl (kr)r

rYl;

rYl;

1]

Z
il+1 (2l + 1)
Pl (cos )
l(l + 1)
s
(2l + 1)
il+1
E0
l(l + 1)
al;

k X
bl;
i!

1 jl (kr)r

l; 1

rYl;

1:

(6.203)

and integrating the result over the solid angle, we nd

@Yl;
cos
sin
sin
@

= E0
=

1:

+ cos

@Yl;
@

(6.204)

(Calculation steps are left for exercise.) Then the desired spherical harmonic expansion of a plane

31

wave is
X il
E0 eikz ex = E0
2i
l; 1

4 (2l + 1)
jl (kr)r
l(l + 1)

rYl;

1(

1
r
k

; )

[jl (kr)r

rYl;

1(

; )] :
(6.205)

The accompanying magnetic eld is expanded as

ikz

B0 e

ey =

E0 X l
i
c
l; 1

(2l + 1) 1
r
l(l + 1) k

[jl (kr)r

rYl;

1(

; )]

jl (kr)r

rYl;

1(

; ) :

(6.206)
These expansions of the plane wave greatly facilitate analysis of scattering of plane electromagnetic
waves by an object. An important observation to be made is that the electromagnetic waves
re-radiated (scattered) by an object also consist of TE and TM modes in the form
(1)

TE mode: hl (kr)r
TM mode:

1
r
k

rYl;

(1)

[hl (kr)r

1(

rYl;

; );

(6.207)

1(

(6.208)

; )];

since the boundary conditions for electric and magnetic elds at the object require that scattered
wave should have exactly the same angular dependence as the incident plane wave. In examples to
follow, we analyze scattering of a plane wave by ideally conducting sphere and by dielectric sphere.

6.9

Scattering by an Ideally Conducting Sphere

Let us consider a highly conducting sphere of radius a placed in a plane electromagnetic wave as
shown in Fig.6-4. In two extreme cases, ka
1 (low frequency limit) and ka
1 (high frequency
limit), the problem can readily be solved without detailed analysis based on the spherical harmonic
expansion. If ka
1; the dipole approximation can be used. As we have seen, the electric dipole
moment of a conducting sphere placed in a static electric eld E0 is
p = 4 "0 a3 E0 ;

(6.209)

and the magnetic dipole moment of a superconducting sphere placed in a static magnetic eld H0
is
m = 2 a3 H0 :
(6.210)
These expressions remain valid in oscillating elds as long as the oscillation frequency is su ciently
small so that ka = !a=c
1: Each dipole radiates linearly independent modes. The radiation
powers of the modes are thus additive. The power radiated by the electric dipole is
PE =

2! 4 a6 2
1 2! 4 p2
=
4
"
E ;
0
4 "0 3c3
3c3 0

32

(6.211)

and that due to the magnetic dipole is

PM =

1
1 2! 4 m2
2! 4 a6 2
=
E ;
4
"
0
4 "0 3c5
4
3c3 0

(6.212)

p
where H0 = E0 =Z; Z =
0 ="0 is substituted in the magnetic dipole moment. The total reradiated (scattered) power is
P = 4 "0

2! 4 a6 2
E
3c3 0

1+

1
4

(6.213)

and corresponding scattering cross-section is given by

=

P
P
=
=
Si
c"0 E02

2
8
+
3
3

a2 (ak)4 :

(6.214)

The dependence _ k 4 ; which is common to all dipole radiation by objects much smaller than the
wavelength of incident wave, is known as Rayleighs law.
The reader may wonder why the magnetic dipole radiates a power comparable with that by
electric dipole in this case. As we have seen, the radiation elds due to a magnetic dipole is of
higher order by a factor (ka)2 compared with those due to electric dipole. The reason is as follows.
The low order vector potential
Z
0 ikr
A(r) '
e
J(r0 )(1 ik r0 )dV 0 ;
(6.215)
4 r
must be applied separately for electric and magnetic dipoles because the surface currents on the
conducting sphere induced by the incident electric and magnetic elds are entirely dierent. The
current due to the electric eld ows between the poles while the current induced by the magnetic
eld is azimuthal. The surface current in the polar direction is of order
Js ' a!"0 E0 = akH0 ;

(6.216)

while the azimuthal current is of order

Js ' H0 :

(6.217)

The polar current is smaller than the azimuthal current by a factor ak: (This is not surprising
because current ows in the azimuthal direction without any hindrance while in the polar direction,
current ow results in charge accumulation at the poles.) Therefore, the radiation elds from the
electric and magnetic dipoles are comparable.
In the opposite limit of ka
1 (high frequency or short wavelength limit), geometric and
physical optics approximation can be applied. The problem reduces to reection by the illuminated
surface and diraction by the other half surface in the shadow. Each surface contributes equally
to the scattering cross-section,
= 2 a2 ;
(6.218)
33

although the angular distribution of scattered Poynting uxes are entirely dierent. We will revisit
this problem in chapter 7.

Figure 6-4: Geometry assumed in analysis of scattering by a conducting sphere.

For arbitrary value of ka; the spherical harmonic expansion of the plane wave can be exploited as
follows. The scattered wave can be decomposed into TE and TM modes having the same angular
dependence as those contained in the incident plane wave. We therefore assume the following
expansion for the electric and magnetic elds of the scattered wave,
Esc (r) =

E0

il+1

l; 1

Bsc (r) =

B0

il

l; 1

where Ml;

and Nl;

(2l + 1)
(Al Ml;
l(l + 1)
(2l + 1)
(Al Nl;
l(l + 1)

Bl Nl;

Bl Ml;

1) ;

1) ;

r>a

(6.219)

r>a

are the TE and TM base vectors,

Ml;
Nl;

1 (r)

1 (r)

(1)

= hl (kr)r

1
r
k

(1)

rYl;

[hl (kr)r

1;

rYl;

(6.220)
1 ];

(6.221)

and Al and Bl are expansion coe cients to be determined. If the sphere is ideally conducting, the
boundary conditions at the sphere surface are that the tangential component of the electric eld
and normal component of the magnetic eld both vanish. Since the total eld at the surface is the
sum of incident and scattered waves, the explicit forms of the boundary conditions are:
(Einc +Esc )t = 0;

at r = a;

(6.222)

(Binc +Bsc )n = 0;

at r = a:

(6.223)

and

34

In fact, the coe cients Al and Bl can be determined from (Einc +Esc )t = 0 alone,
Al =

jl (ka)
(1)

d
[rjl (kr)]
dr
r=a
=
d
(1)
[rh (kr)]
dr l
r=a

Bl =

where
[xjl (x)]0 =

(6.224)

hl (ka)
[kajl (ka)]0
(1)

[kahl (ka)]0

(6.225)

d
d
(1)
(1)
[xjl (x)]; [xhl (x)]0 =
[xhl (x)]:
dx
dx

(6.226)

(1)

Note that for gl (kr) = jl (kr) or hl (kr);

r
=

[gl (kr)r rYl; 1 ]

l(l + 1)
Yl; 1 er
gl (kr)
r

d
[rgl (kr)]rYl;
dr

1;

(6.227)

and the coe cient Bl has been determined from the tangential component of the electric eld of
the TM mode. The boundary condition for the magnetic eld is automatically satised. (Verify
this statement.)

1
0.8
0.6
0.4
0.2
00

10

Figure 6-5: Normalized scattering cross-section of an ideally conducting sphere.

function of ka:

=(2 a2 ) as a

The radiation power of the l-th harmonic mode can be readily found as
Pl

Z
Z
(2l + 1) 2
2
2
2
=
r jAl j
jMl1 j d + jBl j
jNl1 j2 d
l(l + 1)
2 (2l + 1)
jAl j2 + jBl j2 ;
= c"0 E02
k2
2c"0 E02

35

(6.228)

where the factor 2 accounts for an equal contribution from the (l; m =
scattering cross-section is given by
(ka) =

1) modes. Then the

Pl
2 X
=
(2l + 1) jAl j2 + jBl j2
k2
c"0 E02
l
l=1
1
0
2
2
1
0
[kaj
(ka)]
2 X
j
(ka)
l
l
A ; (m2 ):
+
(2l + 1) @ (1)
(1)
0
k2
hl (ka)
[kahl (ka)]
l=1

(6.229)

(ka) normalized by 2 a2 (this is the scattering cross-section in the geometrical optics limit ka
1)
is plotted in Fig.?? in the range 0 < x(= ka) < 10: In the long wavelength limit, ka
1; the
spherical Bessel functions approach
jl (x) !
(1)

hl (x) !

xl
l+1
; [xjl (x)]0 !
xl for x
(2l + 1)!!
(2l + 1)!!
i

1;

l(2l 1)!!
(2l 1)!!
(1)
; [xhl (x)]0 ! i
for x
xl+1
xl+1

(6.230)

1:

(6.231)

The lowest order terms of l = 1 remain nite in this limit,

2
1
i (ka)3 ; B1 = i (ka)3 ;
3
3

A1 =

(6.232)

which yields a scattering cross-section

'

10 4 6
k a ; ka
3

1:

(6.233)

In short wavelength limit ka

1; the gure indicates that indeed approaches 2 a2 .
In radar engineering, the Poynting ux scattered back toward a radar is of main interest. In
the geometry assumed in Fig. 6-4, the direction of back scattering corresponds to = : Since
Yl;1 ( ; ) + Yl;

1(

; )=

2i

2l + 1
P 1 (cos ) sin ;
4 l(l + 1) l

dPl1
dPl1
1
l(l + 1)
+
Pl1 = 0; and
= ( 1)l
at
d
sin
d
2
the scattered eld at

= ;

(6.234)

(6.235)

becomes

Esc (r) =

E0

il+1

l; 1

iE0

(2l + 1)
(Al Ml;
l(l + 1)

eikr X 2l + 1
( 1)l (Al
kr
2

Bl Nl;

Bl )ex :

1)

(6.236)

At

= ; the scattered eld is plane polarized in the same direction as the incident electric eld.

36

The Poynting ux at

is

c"0 E 2 X
( 1)l (2l + 1)(Al
Sr ( = ) = 2 20
4r k

Bl ) :

The radar scattering cross section is dened by

radar

X
4
2
(2l + 1)( 1)l (Al
=
r
S
(
=
)
=
r
k2
c"0 E02

Bl ) ;

(6.237)

where 4 is the total solid angle.

Example 3 Scattering by a Dielectric Sphere
Scattering by a sphere of dielectric and magnetic properties can be analyzed in a similar manner.
We assume a sphere of radius a having relative permittivity "r and relative permeability r placed
in a plane wave. In general, "r and r are both complex to account for dissipation (absorption) of
electromagnetic energy. The incident wave is described by the elds
E0 eikz ex =

E0

il+1

l; 1

(2l + 1)
jl (kr)r
l(l + 1)

rYl;

1(

1
r
k

; )

[jl (kr)r

rYl;

1(

; )] ;
(6.238)

and
H0 eikz ey =

E0 X l
i
Z0
l; 1

where Z0 =

0 ="0 :

(2l + 1) 1
r
l(l + 1) k

[jl (kr)r

rYl;

1(

; )]

jl (kr)r

rYl;

1(

; ) :
(6.239)

The scattered wave continues to be assumed in the form

Ee (r) = E0

l 1

l; 1

He (r) =

(2l + 1)
(Al Ml;
l(l + 1)

Bl Nl;

1) ;

r>a

(6.240)

(2l + 1)
(Al Nl;
l(l + 1)

Bl Ml;

1) ;

r>a

(6.241)

E0 X l
i
Z0
l; 1

where Z0 =
0 ="0 is the impedance in the free space. The elds inside the sphere should be
bounded at r = 0 and thus may be assumed as
Ei (r) = E0

il

l; 1

Hi (r) =

(2l + 1)
l(l + 1)

Cl jl (k 0 r)r

(2l + 1)
l(l + 1)

Cl

E0 X l
i
Zs
l; 1

1
r
k0

rYl;

[jl (k 0 r)r

37

Dl

rYl;

1
r
k0

[jl (k 0 r)r

1]

Dl jl (k 0 r)r

rYl;

rYl;

1]

(6.242)

(6.243)

where
Zs =

"r

Z0 ; k 0 =

p
"r

r k;

(6.244)

to account for the impedance and index of refraction of the sphere. Continuity of the tangential
components of the electric eld E and magnetic eld H yields the following conditions:
(1)

jl (ka) + Al hl (ka) = Cl jl (k 0 a);

(1)

[kajl (ka)]0 + Bl [kahl (ka)]0 = p

(1)

jl (ka) + Bl hl (ka) =

(1)

[xjl (x)]0 =

"r

Dl [k 0 ajl (k 0 a)]0 ;

(6.246)

Dl jl (k 0 a);

(6.247)

Cl [k 0 ajl (k 0 a)]0 :

(6.248)

d
d
(1)
(1)
[xjl (x)]; [xhl (x)]0 =
[xhl (x)]:
dx
dx

(6.249)

[kajl (ka)]0 + Al [kahl (ka)]0 =

where

1
"r

(6.245)

Solving for the coe cients Al and Bl , we nd

Al =

Bl =

0
0
r jl (k a)[kajl (ka)]
(1)
hl (ka)[k 0 ajl (k 0 a)]0

"r jl (k 0 a)[kajl (ka)]0

(1)

jl (ka)[k 0 ajl (k 0 a)]0

r jl (k

0 a)[kah(1) (ka)]0
l

jl (ka)[k 0 ajl (k 0 a)]0

(1)

hl (ka)[k 0 ajl (k 0 a)]0

"r jl (k 0 a)[kahl (ka)]0

(6.250)

(6.251)

The case of ideally conducting sphere can be recovered in the limit "r ! 1 and
scattering cross-section can readily be found,

! 0: The

1
2 X
(2l + 1) jAl j2 + jBl j2 :
k2

(6.252)

l=1

In the long wavelength limit, ka; k 0 a

A1 = i
For a dielectric sphere with
given by

2
3

1
2 "r 1
(ka)3 ; B1 = i
(ka)3 :
3 "r + 2
r +2
r

(6.253)

= 1; the scattering cross-section in the low frequency regime is thus

'

In the limit of "r

sphere,

1; the leading order terms are the dipoles, l = 1;

8
3

"r 1
"r + 2

k 4 a6 ; ka

1:

(6.254)

1; we recover the electric dipole portion of the cross-section of a conducting

=

8 4 6
k a :
3

Note that the case of conducting sphere can be recovered, mathematically, in the limit
38

(6.255)
r

= 0 for

A1 and "r = 1 for B1 : The case of conducting sphere should be analyzed by incorporating the
impedance
r
i!
;
Z=
i!" + c
where
letting

6.10

is the conductivity. Ideal conductor is characterized by Z = 0 which can be realized by

! 1; or " ! 1:
c ! 1; or

Scattering by a Cylinder

In this section, we consider scattering of electromagnetic waves by a cylindrical object having a

length l su ciently longer than the wavelength, kl
1: If the incident wave propagates perpendicular to the cylinder axis, the problem becomes two dimensional. For general incident angle, the
incident wave can be decomposed into normal and axial components. The axial component suers
little scattering and it is su cient to consider only the case of normal incidence.

Figure 6-6: Two possible polarizations of the incident eld relative to the cylinder axis, Ei parallel
to the axis (top) and perpendicular (bottom).
Scattering by a conducting cylinder can be analyzed in a manner similar to the case of conducting
sphere. If the incident wavelength is much shorter than the radius of the cylinder, ka
1; the
geometrical optics approximation applies. Regardless of the polarization of the incident wave
relative to the cylinder axis, the scattering cross-section per unit length of the cylinder is given by
l

= 2a + 2a = 4a; ka

1;

(6.256)

where 2a is the contribution form the illuminated surface and another 2a is the contribution from
the shadow surface due to forward diraction. Analysis for this case, which is quite parallel to the
39

case of a conducting sphere, is left for an exercise..

In long wavelength limit ka
1; the polarization direction of the incident wave becomes
important. This is understandable because the current induced on the cylinder surface sensitively
depends on the orientation of the incident electric eld. As intuitively expected, the surface current
ows much more easily along the axis than in azimuthal direction. Therefore, the scattering crosssection when Ei k ez should be much larger than the case Ei ? ez :
The axial components of the electric and magnetic elds satisfy the following 2-dimensional
scalar Helmholtz equation,
1 @
@2
1 @2
+ k2
+
+
2
@ 2
@
@ 2

Ez
Hz

= 0:

(6.257)

General solutions can be found by assuming a form R( )eim ; where the radial function satises
the Bessel equation
d2
1 d
m2
+
+ k 2 R( ) = 0;
(6.258)
2
2
d
d
R( ) = Jm (k ); Nm (k ):
For wave analysis, it is convenient to dene the Hankel functions of the rst and second kinds by
(1)
Hm
(k ) = Jm (k ) + iNm (k );

(6.259)

(2)
(k ) = Jm (k )
Hm

(6.260)

iNm (k ):

The asymptotic form of the rst kind is

(1)
Hm
(k

)!

2
exp ik
k

2m + 1
4

(6.261)

p
(1)
which has an amplitude dependence 1=
appropriate for cylindrical waves. Hm (k ) describes an
(2)
outgoing wave and Hm (k ) an incoming wave.
We rst analyze the case of incident electric eld along the cylinder axis Ei k ez : The incident
wave is assumed to be propagating in the negative x-direction,
Ei (r) = E0 e

ik cos

ez :

(6.262)

The scattered electric eld is also in the z-direction and we assume

Ezsc =

(1)
a m Hm
(k )eim :

(6.263)

If the cylinder is ideally conducting, the boundary condition is that the tangential component of

40

the electric eld vanish at the cylinder surface,

E0 e

ika cos

(1)
am Hm
(ka)eim = 0:

(6.264)

Multiplying by e

im0

and integrating over ; we nd the expansion coe cient am ;

E0

am =

(1)
Hm (ka)

( i)m

i(ka cos +m )

Jm (ka)

E0 ;

(6.265)

X
Jm (ka) (1)
( i)m (1)
Hm (k )eim :
H
(ka)
m
m

(6.266)

(1)

Hm (ka)
and the scattered electric eld is
Ezsc =

E0

The Poynting ux in the radiation zone k

1 is

E 2 2 X Jm (ka)
S = 0
(1)
Z0 k m Hm
(ka)

(6.267)

(6.268)

and the total scattered power is given by

P

S d

E02 4 X Jm (ka)
(1)
Z0 k m Hm
(ka)

The scattering cross-section is

4 X Jm (ka)
=
(1)
l
k m Hm
(ka)

The function
1 X Jm (x)
f (x) =
(1)
x m Hm
(x)

(6.269)

is plotted below. In the short wavelength regime it approaches unity and the scattering cross section
is =l = 4a as expected from the geometrical optics approximation. In the long wavelength regime
ka
1; m = 0 mode is dominant and the cross-section diverges in a manner
2

'

1
ln

ka
2

41

2;

where

= 0:5772

is the Eulers constant. Note that in the limit x ! 0; J0 (x) ! 1 and

(1)

H0 (x) ! i

x
2h
ln
+
2

; x ! 0:

Physically, the symmetric mode (m = 0) corresponds to radiation by a long cylindrical antenna

having a length much larger than the wavelength. Fig.6-7 shows =l normalized by 4a as a function
of x = ka:
10
1 X
x

m= 10

Jm (x)
Jm (x) + jYm (x)

0
0

10

ka

Figure 6-7: Scattering cross-section per unit length of a long conducting cylinder of radius a.
as a fucntion of x = ka: The electric eld is polarized along the cylinder.

=4al

If the incident electric eld is perpendicular to the cylinder axis, it is more convenient to use
the magnetic eld which is axial. The incident magnet eld is
B0 e

ik cos

ez ;

(6.270)

and we assume the scattered magnetic eld in the form

Bzsc (r) =

(1)
bm Hm
(k )eim :

(6.271)

Corresponding electric eld can be found from the Maxwells equation,

1 @Esc
=r
c2 @t

42

Bsc :

(6.272)

The

component of the scattered electric eld is thus

E sc ( ; ) =

while the

ic

X bm @
(1)
[Hm
(k )]eim ;
k
@
m

(6.273)

component of the incident electric eld is

E i ( ; ) = cB0 cos e

ik cos

(6.274)

The boundary condition of vanishing tangential component of the electric eld at the cylinder
surface yields
X bm @
(1)
B0 cos e ika cos
i
[Hm
(ka)]eim = 0:
(6.275)
k
@a
m
0

Multiplying by eim

and integrating over ; we obtain

bm =

B0 ( i)m

[Jm (ka)]0
(1)

[Hm (ka)]0

(6.276)

where the prime means dierentiation with respect to the argument ka and the following transformation is used,
Z 2
cos e i(ka cos +m ) d = i( i)m 2 [Jm (ka)]0 :
0

A resultant scattering cross-section is

4 X [Jm (ka)]0
=
(1)
l
k m [Hm
(ka)]0

(6.277)

Fig.6-8 shows =l normalized by 4a: In contrast to the preceding case Ei = E0 ez ; the cross-section
in the long wavelength regime ka
1 is bounded and given

l
since if ka

'

3
4

k 3 a4 ;

(6.278)

1; the dominant harmonics are

b0 =

In short wavelength regime ka

ization.

i B0
(ka)2 ; b
4

B0
(ka)2 :
4

(6.279)

1; the cross-section approaches 4a as in the case of axial polar-

43

0.8

0.6

0.4

0.2

10

Figure 6-8: ( =l)=4a of a long conducting cylinder when the electric eld is normal to the cylinder.
Problems
6.1 A conducting sphere is placed in a low frequency plane electromagnetic wave such that ka
1
where a is the sphere radius. Finding the eective electric and magnetic dipole moments, show
that the scattering cross section of the sphere is given by
=

(8 + 2) 4 6
k a ;
3

where 8 parts is due to electric polarization and 2 parts due to magnetic dipole. According
to the radiation magnetic eld due to a small source in Eq. (5.51),
H(r) '

1 eikr
4 c r

1
n+ n
c

(n

m)

1
n
2c

...
(n Q) ;

the radiation power due to magnetic dipole is supposed to be of higher order than that due to
electric dipole by a factor (ka)2 : In scattering by a conducting sphere, they are comparable.
Resolve this apparent paradox.
p
6.2 A conducting sphere of radius a has a complex index of refraction "="0 = nr +ini : Determine
the scattering cross-section as a function of ka:
6.3 Show that the transverse component of the current density J is given by
Jt =

1
r
4

44

J(r0 )
dV 0 ;
jr r0 j

and consequently the longitudinal component by

Jl =

1
rr
4

J(r0 )
dV 0 =
jr r0 j

Hint:
r2

1
r0 j

jr

1
r
4

r0 J(r0 ) 0
dV :
jr r0 j

r0 ):

(r

6.4 The scattering cross section of small objects (ka

1 with a being the size) obeys the
4
Rayleighs law _ k which is often used to explain blueness of sky and redness of sunset.
Scattering of sunlight by air molecules requires uctuation in the air density. Explain why.
You may treat the atmosphere as a uid.
6.5 A conducting sphere of radius a is coated with a material having a relative permittivity "r and
permeability r (both may be complex). The thickness of coating is : Analyze low frequency
scattering in the limit ka
1 and show that the radiation elds are characterized by the
dipole terms
2 r (1
)
(kb)3 2 +
;
A1 = i
3 2 + + r (1
)
B1 =
where b = a + ;
section.

(kb)3 2(1
3 2(1

) 2"r (1 + 2 )
;
) + "r (1 + 2 )

= a3 =(a + )3 : Discuss possible eects of coating on the scattering cross

6.6 For a tangential electric eld on the surface of a sphere of radius a,

Es ( ; ) = E ( ; )e + E ( ; )e ;
nd a general expression for the radiation electric eld.
6.7 A conducting sphere of radius a has a narrow planar gap cut at polar angle
the radiation admittance of the sphere is
i
Y =
2 ka sin2
Z0
where Z0 =

1
X
2l + 1
[P 1 (cos
0
2l(l + 1) l

(2)

hl (ka)

0 )]

l=1

(2)
d
da [ahl (ka)]

0 ="0 :

6.8 In the Lorenz gauge, the scalar and vector potentials are
1
(r) =
4 "0
A (r) =

Z
45

eikjr r j
jr r0 j
0

0:

r0 dV 0 ;

eikjr r j
J r0 dV 0 ;
jr r0 j

; (Siemens)

Show that

where the charge density and current density are related through
i! + r J = 0:
Show that the electric eld is given by
E (r) =
=

@A
@t

r
i!
4

1
1 + 2 rr
k

eikjr r j
J r0 dV 0 ;
jr r0 j

where 1 is the unit tensor. Furthermore, show that the concrete form of the tensor operator
in the spherical coordinates is

1
1 + 2 rr
k

where R = jr

2i
kR

2
6 (kR)2
6
eikR
6
0
=6
4 R 6
4
0

i
1+
kR

1
(kR)2
0

0
1+

i
kR

1
(kR)2

7
7 eikR
7
;
7
74 R
5

r0 j : Note that the radial (longitudinal) component is proportional to

1
R

Hint:

2
(kR)2

2i
kR

@er
@er
=e ;
= sin e :
@
@

6.9 The Lorenz gauge is characterized by

r A+

1 @
= 0:
c2 @t

(a) Show that

@
@t

r A+

1 @
c2 @t

= 0;

yields the Coulombs law,

r E=

"0

(b) What does

r r A+

1 @
c2 @t

= 0;

yield?
(c) Show that
r A+
46

1 @
= 0;
c2 @t

is consistent with the charge conservation law,

@
+ r J = 0:
@t
(d) Repeat a and b for the Coulomb gauge, namely, interpret
@
r A =0; rr A = 0:
@t

47

Chapter 7

Diraction: Boundary Value

Problems of Electromagnetic Waves
7.1

Introduction

In electrostatics, a prescribed potential distribution on a closed surface uniquely determines the potential
elsewhere according to the Dirichlets formulation,
(r) =

@G
@n

0
0
s (r )dS ;

G = 0 on S;

provided G(r; r0 ), the Greens function, is so chosen that it vanishes on the closed surface S: The normal
derivative of the Greens function @G=@n can be regarded as a projection or mapping operator to convert a
given distribution of the surface potential s (r0 ) into the potential at an arbitrary observing position r: For
scalar potentials, the projection operator is a scalar function, or more precisely, 1 1 matrix function.
For any vector elds, too, prescribing a vector on a closed surface uniquely determines the vector led
elsewhere. However, a vector eld changes its direction as well as its magnitude over a closed surface. A
vector eld prescribed on a closed surface is to be converted into a vector eld with denitive direction and
magnitude at the observing position. A projection operator that transforms one vector eld on a closed
surface into another elsewhere must be a tensor (or dyadic) composed of appropriate eigenvectors for the
vector eld of concern. For electromagnetic elds, the TE and TM eigenvectors identied in Chapter 5 can
be conveniently used for this purpose.
In this Chapter, a basic formulation will be developed for vector boundary value problems of electromagnetic elds, E and B: It will then be applied to some diraction problems in which boundary electromagnetic
elds are known. However, in most practical problems, it is di cult to know precisely boundary elds because of the feedback from the diracted elds themselves. A rigorous analysis of diraction problems usually
requires solving integral equations based on the boundary conditions for electromagnetic waves.

7.2

Vector Greens Theorem

For arbitrary vector elds E and F, by exploiting the expansion

r (E

F) = r

E r
1

E r

F;

(7.1)

r (F

E) = r

the following identity can easily be proven,

Z
(E r r F F r r

F r

E)dV =

F r

(F

(7.2)

E;

F) dS:

(7.3)

Since
E r

r2 F]

F = E [r(r F)
= r [E(r F)]

(7.4)

(r E)(r F)

E r F;

(7.5)

Eq. (7.3) can be modied as

Z

(F r2 E

E r2 F)dV =

[(Fr E

Er F) n

(n

E) F

E r

F] dS;

(7.6)

where n is the unit normal vector on the closed surface directed away from the volume V; dS = ndS: (See
Fig. 7.1.) This is a mathematical identity and holds for arbitrary vector elds E and F.

Figure 7-1: For Es ; the electric eld specied on the closed surface S; the electric eld o the surface E is
uniquely determined. n is the normal unit vector directed away from the region of interest.

We now let the vector E be the electric eld and F be one component of Greens dyadic. In source free
region, the electric eld E satises the Helmholtz equation,
(r2 + k 2 )E = 0;

(7.7)

and the Greens dyadic satises the singular inhomogeneous Helmholtz equation
(r2 + k 2 )G =

(r

r0 )1;

(7.8)

where 1 is the unit dyadic dened in terms of appropriate eigenvectors for the electromagnetic elds. Then,
2

Eq. (7.4) yields an expression for the electric eld E in terms of the electric eld specied on the closed
surface S;
I
E(r) =
[(n E)r G + (n r E) G + (n E) (r G)] dS;
(7.9)
S

where r E = 0 (no charges on S) has been assumed. Note that r G is a vector and r
For a diagonal Greens dyadic
G = G1;

G is a dyadic.
(7.10)

it follows that
r G = rG;
(n

(n

E) G = (n

E) (r

(7.11)
r

G) = (n

(7.12)

E)G;
rG:

E)

(7.13)

Therefore, Eq. (7.9) reduces to

E(r) =

[(n E)rG + (n

E)G + (n

rG] dS:

E)

(7.14)

The scalar Greens function G satises

(r2 + k 2 )G =

r0 );

(r

(7.15)

and its explicit expression in terms of spherical harmonic expansion is

G(r; r0 )

0
1
eikjr r j
0
4 jr r j
(
)
1 X
l
(1)
X
jl (kr0 ) hl (kr)
= ik
Ylm ( ; ) Ylm
(1)
jl (kr) hl (kr0 )
l=0 m= l

r > r0
r < r0

(7.16)

As shown in Chapter 6, the solution for the vector potential that satises the Helmholtz equation
r2 + k 2 A =

(7.17)

0 J;

can be decomposed into TE and TM modes. In the outer region r > r0 ; we have the following decomposition,
A(r)

0 ik

1 X
l
X

l=1 m= l

0i

1 X
l
X

1
(1)
h (kr)r
l(l + 1) l

l=1 m= l

1
r
l(l + 1)

rYlm ( ; )

(1)
[hl (kr)r

r0 Ylm ( 0 ;

jl (kr0 )r0

rYlm ( ; )]

r0

[jl (kr0 )r0

) J(r )dV 0
r0 Ylm ( 0 ;

)] J(r (7.18)
)dV 0 :

This can be written in a more compact form using the Greens dyadic G;
A(r) =

eikjr r j
0
J(r )dV 0 =
jr r0 j

G J(r )dV 0 ;

(7.19)

where the Greens dyadic is

0

eikjr r j
1
G(r; r ) =
4 jr r0 j
X
1
(1)
= ik
h (kr)jl (kr0 )r
l(l + 1) l
lm
iX
1
(1)
+
r [hl (kr)r
k
l(l + 1)
0

r0 Ylm ( 0 ;

rYlm ( ; )r0
rYlm ( ; )]r0

lm

[jl (kr0 )r0

(7.20)
r0 Ylm ( 0 ;

)]:

(7.21)

Note that the TE eigenvector

(1)

hl (kr)r

rYlm ( ; );

(7.22)

and TM eigenvector
(1)

[hl (kr)r

rYlm ( ; )];

(7.23)

are orthogonal to each other for given mode numbers l and m and the Greens dyadic is evidently diagonal.
For diraction by an aperture in a large, at conducting screen, Eq. (7.14) simplies to
E(r) =

(n

rGdS:

E)

ap erture

(7.24)

This formula was originally found by Smythe based on an equivalent layer of magnetic dipole (equivalent to
a sheet of magnetic current),
2
n EdS;
(7.25)
dm =
i! 0
which produces a vector potential
2
A (r) =
i!

rG

(n

E)dS:

(7.26)

(Recall that an oscillating magnetic dipole creates a vector potential,

A (r) =
provided kr

0 rG

m'

ikr
0e

4 r

ik

m;

1:) Corresponding electric eld is

E(r)

=
=

Z
@
A =2
rG (n E)dS
@t
ap erture
Z
2
(n E) rGdS:
ap erture

(Explain why we can use

E(r) =

@
A (r) ;
@t

here instead of going through the usual procedure

1 @E
=r
c2 @t

B=r

(r

A);

in terms of longitudinal and transverse components of A:) The tangential component of the diracted
magnetic eld in the aperture vanishes (that is, in the aperture, the tangential magnetic eld is equal to that

of the incident wave which is unaected by the aperture) and only the tangential component of the electric
eld contributes to diraction. Likewise, the normal component of the diracted electric eld vanishes in
the aperture. These follow from the symmetry property of the diracted electric and magnetic eld in front
of and behind the conducting plate. The diracted electric eld normal to the plate is odd with respect to
the normal coordinate z,
Ez (z) = Ez ( z);
(7.27)
and so is the tangential magnetic eld,
B (z) =

ez

ez

B ( z) ;

(7.28)

but the tangential electric eld and normal magnetic eld are even,
ez

E (z) = ez

E ( z) ; Bz (z) = Bz ( z):

(7.29)

Therefore in the aperture, the tangential component of the diracted magnetic eld should vanish. On the
surface of the conducting plate away from the aperture, tangential component of diracted magnetic eld
can exist with a corresponding surface current
n

H = Js ; (A m

):

Since the normal unit vector n changes sign from one side (z < 0) to other (z > 0) ; Js = n H is even
with respect to z; Js (z = 0) = Js (z = +0) : The surface charges induced on both sides of the conducting
plate are also equal since
= "0 n E is even with respect to z: We have encountered such symmetry
properties in Chapters 2 and 3 when we analyzed leakage of electric and magnetic elds through a hole in a
superconducting plate. In both cases, eective dipoles for elds in regions z < 0 and z > 0 are opposite to
each other and perturbed electric and magnetic elds due to the hole have such symmetry properties which
hold in general even for radiation elds as long as the plate can be regarded as ideally conducting.
The Smythes formula, Eq. (7.24), may be derived purely mathematically as follows. For a at boundary,
the normal unit vector n is constant, say n = ez . First, we note
I
I
I
I
(n E) rGdS =
r (Gn E) dS + Gr (n E) dS = Gr (n E) dS;
(7.30)
since for a closed surface,

[G (n

E)]dS = 0:

(7.31)

Recalling r E =0 and exploiting the fact that for a unidirectional vector n;

r

(n

E)

=
=

r (n E)

(E r) n

(n r) E + n (r E)

E (r n)

(n r) E;

(7.32)

(E r) n+ (n r) E + n (r

(n r) E + n (r

E) ;

E) + E

(r

n)
(7.33)

we nd for a constant vector n;

n (r

E) = r

(n

E) + r (n E) :

(7.34)

Then
I

Gn

(r E) dS

=
=

Gr
(n

(n
E)

E) dS +
rGdS

Gr(n E)dS
(n E)rG dS;

and Eq. (7.14) reduces to

E(r)

=
=

I
2

[(n E)rG + (n

E)G + (n

S
Z

(n

E)

rG dS;

[(n E)rG + (n

whole at surface

or
E(r) =

whole at surface

E)

rG] dS

E)G] dS:

In the case of diraction by an aperture in a conducting plate, n E =0 except in the aperture. Therefore,
Z
E(r) = 2
(n E) rG dS:
(7.35)
ap erture

This should be equal to

E(r) =

[(n E)rG + i!(n

B)G] dS;

(7.36)

conducting plate

since in the aperture where no eld discontinuity exists, both n E and n B vanish. (They are odd functions
of z:)
Rigorous analysis of diraction by an aperture in a conducting screen is extremely di cult mainly because
in Eqs. (7.14) and (7.24), the elds on the surface and in aperture are the diracted elds which are to be
found! However, in some limiting cases, sensible analytic solutions can be found. In short wavelength limit
ka
1 where a is the size of aperture, geometric optics approximation may be used. In this case, the eld
in the aperture may be approximated by the incident eld. In long wavelength limit ka
1; the eld in the
aperture is totally unknown. However, diracted power can be calculated accurately using eective dipoles.
Then the eld intensity in the aperture can be estimated to be kaEi
Ei where Ei is the incident electric
eld.

7.3
7.3.1

Some Examples of Aperture Diraction

Long Slit

Let a plane wave be incident on a at opaque screen having a long slit with an opening width a: The problem
is essentially two dimensional and the diracted wavevector k may be assumed to lie in the x y plane.
Since the slit extends from z 0 = 1 to 1; we are not allowed to assume r
r0 and the diracted electric

Figure 7-2: Diraction by a long slit of width a in a conducting screen. In short wavelength limit ka
1;
the electric eld in the slit opening may be approximated by that of the incident wave. Smythes formula in
Eq. (7.24) is applicable.

eld has to be calculated from

E(r) =

(n

r0 Gdy 0 ;

E)

(7.37)

where G is the two-dimensional Greens function satisfying

(r2 + k 2 )G2 =
Solution for G2 is
G2 =
(1)

r0 ) =

2 (r

x0 ) (y

(x

i (1) h p 2
H
k x + (y
4 0

y 0 ):

(7.38)

i
y 0 )2 ;

(7.39)

(7.40)

where H0 (x) is the Hankel function of the rst kind. Its asymptotic form is
(1)
H0 (x)

Using n0 =

'

2 ikx
e
kx

i4

1:

n (unit normal directed into the diracted region) and noting

r 0 G2 =

we rewrite Eq. (7.37) as

E(r) =

2r

rG2 ;
G(n0 E)dy 0 :

(7.41)

Substituting
(1)
H0

h p
k x2 + (y

y 0 )2

2 ik
e
k
r
2 ik
e
=
k
'

i4

ik y 0

i4

iky 0 sin

(7.42)

where
=
and

(7.43)

is the angle between k and x-axis, we obtain

E(r) '

with

p
x2 + y 2 ;

a
k
2

(n

r
2 ik
E0 )
e
k

i4

sin

(7.44)

dened by

ka
sin :
(7.45)
2
The function sin = is the familiar Fraunhofer diraction form factor to describe the angular ( ) dependence
of the eld intensity. The radiation (diracted) power per unit length of the slit is
=

P
=
l
In short wavelength limit ka
obtain an expected result

=2
=2

c"0 jEj d :

1; the integration limits may be approximated by from

(7.46)
1 to 1; and we

P
2
= c"0 jE0 j a; ka
1:
(7.47)
l
In the opposite limit ka
1; the electric eld in the slit opening cannot be replaced by the incident eld
because the current induced in the conducting plate greatly disturbs the incident eld. The problem is dual
of scattering of electromagnetic waves by a thin, long conducting plate which will be analyzed as a separate
problem.

7.3.2

Circular Aperture

Figure 7-3: Geometry of circular aperture of radius a in an opaque screen.

Radiation from a circular aperture on a large conducting plate is a classical problem and of practical importance. In short wavelength limit ka
1 (a the hole radius), the eld in the aperture may be approximated

by that of the incident wave E0 : The diracted electric eld is given by

Z
0
i
0
E(r) =
k (n E0 )eikr e ik r dS;
2 r

(7.48)

where in the geometry shown,

k r0 = kr0 sin sin

The integral can be performed as follows:

Z
=

a
0

e ikr sin sin d

r dr
0
0
Z a
J0 (kr0 sin )r0 dr0
2

2 a
J1 (ka sin ):
k sin

(7.49)

The diracted electric eld thus reduces to

E(r) =

ia2
k
r

E0 )eikr

(n

J1 (ka sin )
:
ka sin

(7.50)

p
Since the amplitude of the Bessel function J1 (x) decreases with x in a manner J1 (x) _ 1= x; the electric
eld is appreciable only at small : In order of magnitude, diracted Poynting ux has an angular spread
about the z axis,
1
sin ' '
1:
ka
The rst intensity minimum occurs at the rst root of J1 (x) = 0; ka sin ' 3:83; or
'

3:83
= 1:22 ;
ka
D

D = 2a (diameter of the hole).

(7.51)

This is the familiar diraction limited resolving power of circular apertures such as mirrors in telescopes and
parabolic microwave antennas. The radiation power associated with the diracted eld is
Z
2
2
P = c"0 r
jEj d

Since ka
Noting

=2

J12 (ka sin )

sin d :
sin2

(7.52)

1; the upper limit of the integral may be extended to 1 and sin

may be approximated by :

= c"0 jE0 j 2 a2

J12 (ka )

d =

1
; (independent of ka);
2

we nd
2

P = c"0 jE0 j

a2 :

(7.53)

This is an expected result in the short wavelength limit (ka

1) and simply corresponds to the incident
power going through the circular aperture.
In long wavelength limit ka
1; the above analysis breaks down completely, for the electromagnetic elds
in the aperture are entirely dierent from those associated with the incident wave. For normal incidence on a
conducting plate, the electric eld vanishes at the plate but the magnetic eld is doubled because of complete

reection. This does not mean there exists a tangential magnetic eld of H =2H0 in the aperture because
the aperture should not aect the incident magnetic eld and the tangential component of the magnetic
eld in the aperture is H = H0 :What we can do is to nd an eective magnetic dipole moment of the hole
using 2H0 as the unperturbed eld as we did in analyzing the leakage of magnetic eld through a hole in a
superconducting plate. The dipole moment of the hole is
8a3
3

m=

16a3
H0 ;
3

2H0 =

(7.54)

which radiates at a power

P

1
1 2! 4 m2
2 4 "0 3c5
64
2
(ka)4 a2 Z0 jH0 j ;
27

=
=

where the factor

1
2

(7.55)

is due to radiation into half space (behind the screen). The transmission cross-section is
t

'

P
2

Z0 jH0 j

64
(ka)4 a2 _ k 4 a6 :
27

(7.56)

The dependence _ k 4 (or ! 4 ) is the common feature of Rayleigh scattering of electromagnetic waves by
small objects (in this case an aperture). _ a6 indicates extremely sensitive dependence of the cross section
on the hole radius. To order k 6 ; the cross section is
'

64
22
(ka)4 a2 1 + (ka)2 +
27
25

; ka

1:

(7.57)

Since the radiation power is proportional to

Z

(n

Ea )e

ik r0

' Ea2 a4 ;

dS

where Ea is the electric eld in the aperture, it is evident that the eld is of the order of
Ea ' kaE0

E0 ;

being much smaller than the incident eld E0 : This is expected since the incident wave is essentially shortcircuited at the plate and only a small eld can leak through the aperture. The electric eld in the aperture
has been worked out by Bouwkamp,
Ex (x; y) = i

Ey (x; y) = i

4k
xy
p
3
a2 x2

4k x2 + 2y 2 2a2
4k
p
E0 = i
2
2
2
3
3
a
x
y

y2
2

E0 ;
2
sin2
p
a2

(7.58)
2a2
2

E0 :

(7.59)

Ex is an odd function of x and y and does not contribute (integrates to 0). The eld diverges at the rim
= a but is integrable. Integration over the aperture yields
Z

d Ey (x; y) =

10

8ka3
E0 :
3

(7.60)

The diracted electric eld can then be found using the Smythes formula which is equivalent to the eld
due to an eective magnetic dipole moment in Eq. (7.54).
Diraction in the regime ka ' O(1) is di cult to analyze and requires numerical analysis. The transmission cross-section peaks at ka ' 1:56 and its value is max ' 1:8
a2 . This behavior is similar to the
case of scattering by a conducting sphere.

7.3.3

Radiation from an Open End of a Rectangular Waveguide

Figure 7-4: Radiation from an open end of a rectangular waveguide. The electric eld in the opening is
Ey (x) = E0 sin( x=a) and the tangential component of the magnetic eld is Hx (x) = E(x)=Z10 where Z10
is the impedance of the TE10 mode.
The electric eld of the TE10 mode in a rectangular waveguide is given by
E(x; z; t) = E0 sin

x ei(kz z

!t)

ey ;

(7.61)

where a is the width of the waveguide and

kz =
is the axial wavenumber with

!p
1
c

(! c =!)2 ;

(7.62)

c
;
(7.63)
a
mode. The accompanying magnetic eld transverse to the axis is
!c =

being the cuto frequency of the TE10

H(x; z; t) =
where
Z10 =

E0
sin
x ei(kz z
Z10
a
r

"0

p
1

(! c =!)2 ;

!t)

ex ;

(7.64)

(7.65)

is the impedance of the TE10 mode. (The axial magnetic eld Hz associated with the TE10 mode does not
11

contribute to radiation from the aperture.) If a waveguide is truncated, the open radiates electromagnetic
waves. The Smythes formula is not applicable because it only pertains to radiation from apertures in a
large conducting plate. In the general diraction formula
I
E(r) =
(n E)r0 G + G(n r0 E) + (n E) r0 G dS;
(7.66)
S

where
G(r; r0 ) =
we assume r

r0 . Then,

0
eikjr r j
;
4 jr r0 j

1 ikr
e
4 r

G(r; r0 ) '
and
r0 G =

rG '

ik

ik r0

1 ikr
e
4 r

;
ik r0

Also,
r
Then,
ieikr
E(r) '
4 r

E=

(n

@B
= i!B:
@t

E) + !n0 B

k(n0 E) e

ik r0

dS;

(7.67)

where n0 = n is the normal unit directed toward the region where the electric eld is to be evaluated. It is
not obvious that the contribution from the second and third terms in the integrand,
I
0
0
!n0 B k(n E) e ik r dS;
(7.68)
S

is explicitly transverse (perpendicular to k) which should be the case since we are calculating radiation eld.
To prove this, we rst show that
I
r

(n

B) G (r; r0 ) dS;

can be expressed in terms of the normal component of the electric eld as follows.
I
I
r
G (n B) dS =
r0 G (n B) dS
S
S
I
=
r0 G (B n) dS
S
I
=
r0 G B ndS
IS
I
=
r0 (GB) ndS
Gr0 B ndS
S
S
I
i!
= 0+ 2
G (E n) dS;
c S
where use is made of the identity

r0

A ndS = 0:

12

Therefore,
r

G (n

B) dS

I
B) dS + k 2
G (n B) dS
IS
I S
i!
r G (E n) dS + k 2
G (n B) dS
c2
S
IS
I
i!
(E n) r0 GdS + k 2
G (n B) dS;
c2 S
S

= rr

=
=

G (n

and the diracted eelctric eld can alternatively be calculated from

I
E(r) =
(n E)r0 G + G(n r0 E) + (n E) r0 G dS
S
I
i
r r (Gn H) dS:
=
rG (n E) +
!"
0
S
Then Eq. (7.67) may be rewritten as
ieikr
E(r) =
k
4 r

n0

1
k
!"0

(n

H) e

ik r0

dS;

(7.69)

which is explicitly transverse to k: As mentioned in Chapter 5, this formula was rst derived by Schelkuno in
term of a ctitious magnetic current to replace the tangential component of the electric eld on a boundary.
The price to be paid for introducing a magnetic current is that the Maxwells equation, r B = 0; has to be
violated. In the derivation presented here, no magnetic current is assumed.
Denoting the surface integral over the open end of the waveguide by
I( ; )

dx

dy sin

i
k sin sin

=a)2

x e

ik(x sin cos +y sin sin )

=a
(1 + e
sin2 cos2

ika sin cos

k2

)(1

ikb sin sin

);

(7.70)

we nally nd the electric eld radiated from the open end of a rectangular waveguide,
E(r) =

ikeikr
I( ; )
4 r

"

cos
1

(! c =!)2

+ 1 sin e +

Calculation of the radiation power is left for an exercise.

7.3.4

p
1

1
(! c =!)2

+ cos

cos e

(7.71)

Scattering by a Conducting Sphere Revisited

This problem has been solved rigorously in the preceding Chapter in terms of spherical harmonic expansion.
However, the result is not very illuminating physically particularly in the short wavelength limit ka
1
wherein summation of a large number of functions involving spherical Bessel functions and their derivatives
must be performed. In this limit, geometrical optics approximation should be able to yield the scattered
electric eld provided the boundary conditions for the electromagnetic elds are appropriately incorporated.
The incident plane wave seesthe cross-section of the sphere a2 which corresponds to the scattering crosssection due to scattering by the illuminated hemispherical surface facing the incident wave. Scattering by
the illuminated surface is nothing but reection by a spherical convex mirror. The Poynting ux associated
with reection is uniform (isotropic) and independent of the angles and . The total correct cross-section
13

Figure 7-5: Scattering by a conducting sphere in short wavelength limit ka

1:

as revealed from the rigorous analysis was 2 a2 : The additional a2 is due to shadow scattering from the
hemispherical surface in the shadow of the incident wave where the eld vanish. This vanishing eld can be
interpreted as cancellation between the eld of the incident wave and that of the scattered wave. Shadow
scattering is dual of diraction by a circular aperture in a conducting screen which radiates a power
2

a2 ;

Psh = c"0 jE0 j

ka

1:

(7.72)

Although the scattered power through shadow scattering is identical to that by the illuminated surface, the
Poynting ux is sharply peaked behind the sphere as in the case of diraction with an angular spread of
order
' =a
1:
If the sphere is ideally conducting, the boundary conditions for the electric and magnetic elds are
Et = 0;

Bn = 0:

(7.73)

Let the electric eld of the incident wave be Ei and that of scattered eld be Es : The vanishing tangential
component of the total eld requires that
n

(Ei +Es ) = 0:

(7.74)

The vanishing normal component of the magnetic eld requires that

n (Bi +Bs ) = 0:
In the spirit of geometrical optics valid in the short wavelength regime,
Ei =

c
ki Bi ;
k

Es =

14

c
ks Bs ;
k

(7.75)

where ki is the wavevector of the incident wave and ks is that of the scattered wave. From
(Ei +Es ) = n (ki Bi +ks Bs ) = 0;

(7.76)

and
n

ki = n

ks ;

n ki =

n ks ;

(7.77)

it follows that
(Bi Bs ) = 0:

(7.78)

Likewise, from
n (ki Ei ks Es ) = 0;
we nd
n (Ei Es ) = 0:
With these preparations, we now apply Eq. (7.67) to the illuminated and shadow surfaces separately.
The contribution from the illuminated surface is
Z
0
ieikr
0
0
(k k0 ) (n Ei ) + (k k0 )(n Ei ) e ik r dS;
Eill (r) =
4 r
where the incident magnetic eld has been eliminated through
!Bi = k0 Ei :
In the geometry shown in Fig. 7-5, the incident wave propagates in the negative z-direction,
Ei (r) = E0 e

ikz 0

ikr 0 cos

= E0 e

(7.79)

Therefore, the electric eld diracted by the illuminated surface is

Eill (r) =

ieikr
4 r

(k

k0 )

(n

E0 ) + (k

ikr 0 cos

ik r0

ikr 0 cos

ik r0

k0 )(n E0 ) e

dS:

(7.80)

dS:

(7.81)

Similarly, the contribution from the shadow surface is

Esh (r) =

ieikr
4 r

(k + k0 )

(n

E0 )

(k

k0 )(n E0 ) e

The total diracted eld is given by the sum of Eill and Esh : However, the angular dependence of the eld
intensities is entirely dierent as explained in the introduction, namely, jEill j is insensitive to and ; while
jEsh j sharply peaked in the direction = (forward scattering). Therefore, the scattered power can be
calculated separately as if the elds were incoherent.
Noting r0 = a on the sphere surface, we nd the phase function reduces to
kr0 cos + k r0

= ka (1 + cos ) cos
0

= ka f ( ; ; ;

15

):

+ sin sin

cos(

)
(7.82)

By assumption ka

1: Therefore, the exponential function

ikaf

e
rapidly oscillates as

and

(7.83)

are varied. The slow angular dependence in the amplitude function

(k

eikr
(k
i
4 r

k0 )

k0 )

can be ignored in the integration over

Eill (r) '

(n

(n
0

and

E0 ) + (k

k0 )(n E0 )

and can be taken out of the integral,

0

Ei ) + (k

k0 )(n Ei )

0; 0
0
0

ikaf ( 0 ;

sin 0 d 0 d 0 ;

where 00 and 00 indicate the angular location on the sphere surface which makes the dominant contribution
to the phase integral
Z
0
0
(7.84)
e ikaf ( ; ) sin 0 d 0 d 0 :
Since ka
1; major contribution to the integral comes from the angular location where the function f ( 0 ; 0 )
becomes stationary,
@f
@f
= 0:
(7.85)
0 =
@
@ 0
This determines
0
;
= :
2
This is a trivial result well expected from optical reection. At an observing angular location ( ; ); only
the wave reected at ( 0 = =2; 0 = ) can be detected.
Let us Taylor expand the phase function f ( 0 ; 0 ) about the stationary phase point 0 = 0 = =2 and
0
= 0= ;
!
0

f ( ; 0; ;

Integration over

) ' cos

sin2

)2 +

(7.86)

yields
Z

=2

sin

exp ika cos

' sin

"

2
2

"

ika cos( =2)

Note that the integration limits can be extended to

sin 0 d
2

exp ika cos

1 because only the region

16

(7.87)
0

' =2 contributes to the

integral. Similarly,
Z

exp ika cos

'
'
Therefore,

sin2

exp ika cos

1
p

sin2

ika cos( =2) sin( =2)

Z
= a2

ikaf ( 0 ;

)2 d

)2 d

(7.88)

(7.89)

dS

i
exp ( 2ika cos( =2)) :
ka cos( =2)

At the particular angular location ( =2; ); the magnitude of the vector jk

2k cos
and the normal vector n0 is in the direction of k

(7.90)
k0 j is
(7.91)

k0 ;

n0 =

1; ;
2

(7.92)

Then,
(k

k0 )

2k cos

2k cos

(n0 E0 ) + (k

k0 )(n0 E0 )

[n0 (n0 E0 ) + n0 (n0 E0 )]

[2n0 (n0 E0 )

E0 ] :

(7.93)

The vector
2n0 (n0 E0 )

E0 ;

(7.94)

has a magnitude of E0 ; that is,

j2n0 (n0 E0 )

E0 j = E0 ;

and is perpendicular to the scattered wavevector k. (Prove this statement.) Therefore, the electric eld
scattered (or reected) by the illuminated surface is
Eill (r) =

a ikr
e
2r

2ika cos( =2)

The amplitude jEill (r)j is independent of the angles and

power can readily be found as
Pill

= c"0 E02 r2
=

c"0 E02

17

[2n0 (n0 E0 )

E0 ] :

(7.95)

as expected from isotropic reection and scattered

a
2r
a2 :

4
(7.96)

The scattering cross-section due to scattering (essentially reection) by the illuminated surface is therefore
ill

= a2 :

(7.97)

Scattering by the shadow surface is identical to diraction by a circular aperture on a conducting screen
analyzed earlier, for the boundary conditions
Ei + Es = 0;

Bi + Bs = 0;

(7.98)

indicate that the elds of the incident wave can be used on the surface of the shadow hemisphere. (The
sign inversion does not aect the diracted power.) The shape of the shadow surface is irrelevant as long as
its area projected normal to the incident wave is circular with an area a2 : To see this, let us consider the
electric eld scattered by the shadow surface in Eq. (7.81),
Esh (r) =

ieikr
4 r

(k + k0 )

(n

E0 )

(k

k0 )(n E0 ) ei(k0

k) r0

dS 0 :

(7.99)

Since there is no stationary phase point in the shadow (it only occurs in the illuminated surface) and the
0
function ei(k0 k) r rapidly oscillates because of the assumption ka
1; the integral will be nonvanishing
only if k0 k ' 0 which means predominantly forward scattering. The change in the wavevector k = k k0
is thus perpendicular to k and k0 ;
k = k sin e
= k sin (cos ex + sin ey );

(7.100)

where is polar angle of scattered wavevector k now measured form the negative z axis, e is the radial
unit vector in the cylindrical coordinates and is the azimuthal angle about the z axis. The phase function
becomes
0
(k0 k) r = ak sin sin 0 cos 0
;
(7.101)
and the eld amplitude may be approximated by
0

(k + k0 )
which is independent of

and
Z
'
=

(n

E0 )

(k + k0 ) (n E0 )
Z
Z
2kE0 a2
sin 0 d 0
2kE0

2k

(ez

E0 )

(k
d 0e

k) r0

k0 )(n E0 ) ei(k0

(7.102)

iak sin sin

cos(

dS 0

cos

Z 0a
0

where

2k cos 0 E0 ;

k0 )(n E0 ) '

. Then

(k

d e ik sin cos(
Z
d
d 0 e ik sin

cos(

= a sin 0 ; n0 is the unit vector normal to the circular at surface, and k

18

(7.103)
E0 ' k0

E0 = 0 is

noted. The electric eld scattered by the spherical surface in the shadow is thus given by
Esh (r) =

ieikr
k
2 r

(ez

E0 )

ik sin cos(

d :

(7.104)

This is the negative of the eld diracted by a circular aperture given in Eq. (7.48). Integrations over
and 0 yield
Esh (r) =

ia
k
r

(ez

E0 )

J1 (ka sin )
'
k sin

ia
k
r

(ez

E0 )

J1 (ka )
;
k

1:

(7.105)

The radiation power associated with shadow diraction is identical to that in the case of circular aperture,
Psh = c"0 E02

a2 :

(7.106)

Therefore, the total power reradiated by the sphere is

P = Pill + Psh = c"0 E02

2 a2 ;

(7.107)

and the total scattering cross-section in the short wavelength limit is

= 2 a2 ; ka

1:

(7.108)

This is consistent with the asymptotic (ka ! 1) value of the general cross-section derived earlier,
(ka) =

1
h
i
2 X
2
2
(2l
+
1)
jA
j
+
jB
j
;
l
l
k2

(7.109)

d
[ajl (ka)]
da
:
d
(1)
[ahl (ka)]
da

(7.110)

l=1

where
Al =

jl (ka)
(1)

hl (ka)

; Bl =

The summation over l in Eq. (7.109) requires up to l ' O(ka) for su cient accuracy and thus becomes
cumbersome in optical regime where ka can be huge. Also, it should be noted that the assumption of ideally
conducting sphere ("e = 1; e = 0) entirely breaks down in high frequency (short wavelength) regime.
The eective relative permittivity of ordinary metals in optical frequency regime is in general complex and
remains of the order of unity. (The relative magnetic permeability in the regime may be assumed to be
unity.)

7.4

Scattering by a Knife Edge

In scattering by a knife edge, there is no geometrical scale size to speak of except for k where is the radial
distance from the edge tip. In short wavelength optical regime, assuming an incident electric eld along
the plane above the edge may provide the lowest order approximation. In rigorous approach, the boundary
conditions for the electromagnetic elds must be incorporated. We assume an observing point at ( ; ): The

19

Figure 7-6: Geometry for analyzing scattering by a conducting knife edge. Top: special case of normal
incidence. Bottom: general incidence angle.

distance between a point on the y axis and the observing point is

0

p
( sin

y 0 )2 +

cos2

Therefore, for small ; the path dierence between

and

y 0 sin +

'
0

1 y 02
:
2

(7.111)

is

1 y 02
;
2
and corresponding phase dierence is
1 ky 02
:
2

(7.112)

The electric eld at (x0 ; y) is thus given by

E0
E(y) = p
2

F (x) =

The function, shown in Fig. (??),

exp

1
x

exp
1

i 2
t dt:
2

i 2
t dt;
2

is known by Fresnels integral and its special values are:

F ( 1) = 0; F (0) =

1
(1 + i); F (1) = 1 + i:
2
20

(7.113)

(7.114)

At large positive y (well above the edge); the incident eld E0 is recovered as expected. Well below the edge
in the shadow region, y ! 1; the eld vanishes also as expected. At y = 0; the eld is 12 E0 and intensity
is 14 of the incident intensity. Fig. (7-8) shows the intensity as a function of the normalized vertical distance
q
2
y!y
x0 :
F (x) =

exp
1

i 2
t dt
2

1.2

0.8
0.6
0.4
0.2
-0.2 0
-0.2

0.2

Figure 7-7: Fresnel integral F (x) =

0.4

Rx

0.6

0.8

1.2

exp i 2 t2 dt: F ( 1) = 0; F (1) = 1 + i:

1.2
1
0.8
0.6
0.4
0.2

-4

-2

Figure 7-8: Intensity I=I0 as a function of normalized vertical distance y

1:37I0 :

2
x0 .

I(y = 0) =

1
4 I0 ;

Imax '

The analysis above is rather primitive because no consideration was given to the boundary conditions
of the electric and magnetic elds. For incident electric eld parallel to the conducting plate, E0 = E0 ez
(E-mode), the total eld (incident + scattered) must vanish at the surface of the conductor plate. If the

21

incident wave approaches the plate at an angle

Ez ( ; )

i
2
1

E0 e
i

0;

the total wave eld is given by

ik cos(

ik cos( +

E0 e

0)

Z 2 pk

0)

= cos

k = sin

0
2

ei 2 t dt
+ 0
2

ei 2 t dt:

(7.115)

If the incident wave is so polarized that its magnetic eld is parallel to the plate, H0 = H0 ez (H-mode), the
boundary condition is
@Hz
3
E _
= 0 at =
;
:
@
2 2
This yields
Hz ( ; )

i
2

H0 e
i

ik cos(

ik cos( +

H0 e

0)

0)

Z 2 pk
Z

= cos

k = sin

0
2

ei 2 t dt
+ 0
2

ei 2 t dt:

(7.116)

The reader should verify that the solutions (rst formulated by Sommerfeld) satisfy the Helmholtz equation
and the respective boundary conditions.

22

Problems
7.1 An electric eld E( ; ) = E ( ; )e + E ( ; )e is specied on the surface of a sphere of radius a:
Determine the radiation electric eld.
7.2 A magnetic eld B( ; ) = B ( ; )e + B ( ; )e is specied on the surface of a sphere of radius a:
Determine the radiation magnetic eld.
7.3 A plane wave is incident normal to a long conducting cylinder having a square cross-section of side
a:The electric eld is axial (Ez only) and the incident wave falls normal to one of the rectangular
surfaces. Derive an integral equation for the surface current density Jz (r): (The integral equation
can be solved numerically following the procedure developed for nding the capacitance of a square
conducting plate.)
7.4 Repeat the preceding problem for H mode (Hz only).
7.5 A point light source is placed at a distance z0 on the axis of an opaque disk of radius a (
the light intensity along the axis behind the disk as a function of distance z:

): Determine

7.6 A point light source is placed at a distance z0 on the axis of a circular opening of radius a in an opaque
screen. Determine the light intensity behind the disk as a function of axial distance z:
7.7 Show that in Fresnel diraction of incident wave normal to the plate, the maximum intensity occurs
p
x0 =2 and is given by Imax ' 1:37I0 where I0 is the intensity of the incident wave.
at y ' 1:2
7.8 Show that near a knife edge, an E-mode axial electric eld
p
Ez ( ; ) = A
sin

and corresponding magnetic eld

H=

A
2i!

1
p

cos

ex + sin

ey ;

satisfy the boundary conditions. The x component of the magnetic eld is discontinuous at = 0
and = 2 : Interpret this peculiarity. Show that the intensity of a wave scattered by a knife edge is
insensitive to the observing angle : (A knife edge appears shiny regardless of observing angle.)
7.9 The transmission cross section of a small (ka
T

1) circular aperture in a conducting plate is

=

64 4 6
k a ;
27

where a is the hole radius. Show that the magnitude of the electric eld in the aperture should be of
the order of
Ea ' kaE0 ;
where E0 is the incident eld.
Note: For incident eld polarized in the x direction, Bouwkamp found the electric eld component in
the aperture responsible for diraction,
Ey (x; y) =

8k x2 + 2y 2 2a2
p
E0 :
3
a2 x2 y 2
23

(Ex (x; y) is an odd function of both x and y:) Integration over the aperture yields
Z

Ey (x; y) dS =

Ey (r; ) rdrd =

16 3
ka E0 ;
3

which is consistent with the eective magnetic dipole,

m=

8 3
a H0 :
3

p
;
At the edge of a thin conductor, the electric eld parallel to the edge vanishes in a manner Ek =
p
while the normal component diverges as E? = 1=
where is the distance from the edge. For
example, the electric eld at the rim of a charged conducting disk behaves as
1

E? _ q

a2

(a

=p

1
:
2a

Such properties can be exploited to nd electric eld when the boundary involves sharp conducting
edge.)
7.10 In low frequency limit, scatterimg by a conducting sphere may be analyzed using dipole approximation.
Relevant dipole moments are
p = 4 "0 a3 E0 ; m = 2 a3 H0 ;
where E0 and H0 are the elds of the incident plane wave. Show that the dierential scattering cross
section is
"
#
2
2
1
1
d
2
4 6
2
=k a
1
cos
sin + cos
cos
;
d
2
2
the total scattering cross section is
=

10 4 6
k a ;
3

and that the force exerted on the sphere is

Fz =

14 4 6
k a "0 E02 ; (N).
3

24

Chapter 8

Radiation by Moving Charges

8.1

Introduction

The problem of radiation of electromagnetic waves by a single charged particle moving at an

arbitrary velocity had correctly been formulated independently by Lienard and Wiechert before
the advent of the special relativity theory. This is because once emitted from a charged particle,
electromagnetic waves propagate at the speed c irrespective of the velocity of the charged particle,
just as sound waves propagate at a speed independent of the source velocity. (The major nding
made by Einstein was that electromagnetic waves still propagate at the speed c regardless of the
observers velocity in contrast to the case of sound waves.)
The scalar and vector potentials due to a moving charge can be found rigorously using the
Greens function for the wave equation. Then radiation electromagnetic elds can readily be calculated. In nonrelativistic regime, the radiation power only depends on the acceleration of charged
particles. As the velocity approaches c; however, signicant increase in the radiation power occurs.
Furthermore, in highly relativistic limit, radiation occurs primarily along the direction of velocity
within an angular spread of order ' 1= about the velocity irrespective of the direction of acp
2
is the relativity factor with = v=c: Hence radiation frequency
celeration. Here = 1= 1
is subject to strong Doppler shift. For example, in synchrotron radiation due to highly relativistic
electron beam bent or undulated by a magnetic eld, radiation even in hard x-ray regime can be
created.
In material medium, radiation processes without acceleration on charged particles are possible.
If the velocity of a charged particle exceeds the velocity of electromagnetic waves in the medium
v>p

1
"

;
0

where " is the permittivity, Cherenkov radiation occurs. Furthermore, if a charged particle crosses a
boundary of two dielectric media, the transition radiation occurs even if the condition for Cherenkov
radiation is not met. Transition radiation is due to sudden change in the normalized velocity from
p
p
1 = v "1 0 to 2 = v "2 0 which may be regarded as an eective acceleration even though the
1

particle velicty v remains constant.

8.2

Lienard-Wiechert Potentials

The charge and current densities of a moving point charge are singular and described by
(r; t) = e [r

rp (t)];

J(r; t) = ev(t) [r

(8.1)

rp (t)];

(8.2)

where e is the charge, rp (t) is the instantaneous location of the charge and v(t) = drp (t)=dt is the
instantaneous velocity of the charge which may be changing with time. Exploiting the Greens
function for the wave equation,
G(r

r0 ; t

t0 ) =

1
4 jr r0 j

jr

t0

r0 j

(8.3)

we can write down solutions for the inhomogeneous wave equations,

1 @2
c2 @t2

r2

in the form

e
(r; t) =
4 "0
A(r; t) =

0e

r2

1 @2
c2 @t2

dV

dV 0

A=

dt0
dt0

1
jr

(8.4)

0 J;

(8.5)

[r0 rp (t0 )] [f (t0 )];

t+

jr

r0 j

The volume integrations can be carried out immediately with the results
Z
e
1
(r; t) =
[f (t0 )]dt0 ;
4 "0
jr rp (t0 )j
A(r; t) =

0e

v(t0 )
[f (t0 )]dt0 ;
jr rp (t0 )j

where f (t0 ) is now

f (t0 ) = t0

(8.6)

v(t0 )
[r0 rp (t0 )] [f (t0 )];
jr r0 j

where
f (t0 ) = t0

r0 j

"0

t+

jr

rp (t0 )j
:
c

(8.7)

(8.8)

(8.9)

(8.10)

The integral involving the delta function can be simplied as

Z
Z
0
0
0
g(t ) [f (t )]dt =
g(t0 ) [f (t0 )]
g(t0 )
jdf =dt0 j

1
df
df (t0 )
dt0

(8.11)

f =0

where t0 is now understood as a solution for t0 satisfying f (t0 ) = 0 or

t0

t+

rp (t0 )j
= 0:
c

jr

(8.12)

This is in general an implicit equation for t0 : The time derivative of f (t0 ) is

n(t0 ) vp (t0 )
=1
c

df
=1
dt0

n(t0 )

where n(t0 ) is the unit vector along the relative distance r

time t and time t0 are related through
n(t0 )

[1
or

dt
=1
dt0

(t0 );
rp (t0 ) and

(8.13)
= v=c: The observing

(t0 )]dt0 = dt;

n(t0 )

(t0 ):

After performing time integration, we nally obtain

(r; t) =

e
4 "0 1

A(r; t) =

0e

1
n(t0 )
1
n(t0 )

1
e
1
=
;
rp (t0 )j
4 "0 (t0 ) jr rp (t0 )j

(t0 ) jr

v(t0 )
v(t0 )
0e
=
;
(t0 ) jr rp (t0 )j
4
(t0 ) jr rp (t0 )j

(8.14)

(8.15)

where
(t0 ) = 1

n(t0 )

(t0 ):

(8.16)

These retarded potentials, called Lienard-Wiechert potentials, had been formulated in 1898. They
are applicable to arbitrary velocity of the charged particle. Retarded nature of the potentials clearly
appears in the condition that all time varying quantities, rp (t0 ); v(t0 ); n(t0 ); must be evaluated at
t0 ; not at the observing time t because of nite propagation speed of electromagnetic disturbance.
Having found the retarded potentials, we are now ready to calculate the electromagnetic elds
due to a moving point charge. The electric eld is to be found from
E(r; t) =

@A
;
@t

(8.17)

where spatial and time derivatives pertain to r (the coordinates of the observing location) and
3

t (observing time). Since the potentials

and A are implicit functions of r and t; it is more
convenient to use the original integral representations, Eqs. (8.8) and (8.9), respectively, for proper
dierentiation with respect to r and t: For example, the spatial derivative of the scalar potential
can be performed as follows. Letting R(t0 ) = jr rp (t0 )j ; and introducing a unit vector in the
direction r rp (t0 );
r rp (t0 )
n(t0 ) =
;
(8.18)
jr rp (t0 )j
we nd
r

=
=
=
=

Z
1
e
r
[f (t0 )] dt0
4 "0
jr rp (t0 )j
Z
e
1
@
n
[f (t0 )] dt0
4 "0
@R R
Z
e
n
n d
[f (t0 )]
[f (t0 )] dt0
2
4 "0
R
cR df
e
n
1 d
n
+
;
2
0
4 "0
R
c dt
R

(8.19)

where use is made of the integration by parts,

Z

d
[f (t0 )]dt0
df
Z
1 d
=
g(t0 )
[f (t0 )]dt0
df dt0
dt0
g (t0 )
d
1
=
df =dt0 dt0 df =dt0
g(t0 )

g (t0 )
(t0 )

1 d
(t0 ) dt0

(8.20)

(8.21)
f (t0 )=0

(8.22)

f (t0 )=0

Similarly,
e 1 d
@A
=
@t
4 "0 c dt0

(8.23)

and the electric eld becomes

E(r; t) =

e
4 "0

n
1 d
+
2
R
c dt0

n
R

(8.24)

f (t0 )=0

To proceed further, we need concrete expressions for the derivatives,

dn
dt0

d
dt0

and

The unit vector n along the distance vector R = r

1
R

rp (t0 ) changes its direction only through the

velocity component perpendicular to R;

v? 0
dt ;
R

dn =

(8.25)

as can be seen in Fig. 8-1. This yields

Figure 8-1: The change in the unit vector n is caused by the perpendicular velocity v? :
dn
=
dt0

v?
:
R

(8.26)

Also,
d
dt0

1
R

=
=
=

1
d
[(1 n )R]
2
( R) dt0
1 h
v?
(n _ )R c(1 n
( R)2
c
1
2
n
R(n _ ) :
2
( R)
c

)n

i
(8.27)

Substitution of Eqs. (8.26) and (8.27) to Eq. (8.24) gives

E(r; t) =
=

e
4 "0

"

n
( R)2

n
( R)2

2
e
1
(n
3 R2
4 "0

)+

1
c

3R

1
R(n _ )
c

n
n

[(n

_]

2R

c
:

#
(8.28)

f (t0 )=0

The rst term in the RHS,

ECoulomb =

2
e
1
(n
3 R2
4 "0

;
f (t0 )=0

(8.29)

is the Coulomb eld corrected for relativistic eects. It is proportional to 1=R2 and thus does not
contribute to radiation of energy. The second term,
Erad =

e
4 "0 c

1
3R

[(n

_]

(8.30)

f (t0 )=0

contains acceleration _ and is proportional to 1=R: This is the desired radiation electric eld due
to a moving charged particle.
The magnetic eld can be calculated in a similar manner from
B = r A
1
[n E]f (t0 )=0 :
=
c

(8.31)

Derivation of this result is left for an exercise.

8.3

Radiation from a Charge under Linear Acceleration

If the acceleration is parallel (or anti-parallel) to the velocity, _

reduces to
"
#
e
n (n _ )
Erad =
4 "0
c 3R
0

= 0; the radiation electric eld

:

(8.32)

f (t )=0

The angular distribution of radiation power at the observing time t is

dP (t)
d

= c"0 jErad j2 R2
"
1 e2 n (n
=
4 "0 4 c
c 3

Denoting the angle between _ and n by ; we have [n

(n

_)

#2

(8.33)

f (t0 )=0

_ )]2 = _ 2 sin2 ; and thus

"
#
_ 2 sin2
dP (t)
1 e2
=
d
4 "0 4 c (1
cos )6

(8.34)

f (t0 )=0

However, the power P (t) in the above formulation is the rate of energy radiation at t; the observing
time, which is not necessarily equal to the energy loss rate of the charge at the retarded time t0
determined from f (t0 ) = 0: To nd the radiation power at the retarded time P (t0 ) ; let us consider
the amount of dierential energy dE=d radiated during the time interval between t0 and t0 +dt0 : By
denition dE = P (t0 )dt0 : The radiation energy dE is sandwitched between two eccentric spherical
surfaces with a volume
dV = R2 cdt0 (1
cos ):

Therefore, the dierential radiation energy is

dE
1
= "0 jErad j2 R2 cdt0 (1
d
2

cos );

and
dP (t0 )
d

"
_ 2 sin2
1 e2
(1
4 "0 4 c (1
cos )6
"
#
_ 2 sin2
1 e2
4 "0 4 c (1
cos )5
0

cos )

f (t0 )=0

(8.35)

f (t )=0

In nonrelativistic limit j j
1; the radiation occurs predominantly in the direction perpendicular
to the acceleration = =2. The total radiation power in this case is
P

'
=

2 Z
1 e2 _
sin2 d
4 "0 4 c
2
1 2e2 _
;
1:
4 "0 3c

(8.36)

This is the well known Larmors formula for radiation power due to a nonrelativistic charge. Since
in nonrelativistic limit,
n [(n
) _ ] ' n (n _ );
(8.37)
Larmors formula is applicable for acceleration in arbitrary direction relative to the velocity.
For arbitrary magnitude of the velocity ; the radiation power can be found from
2

P (t ) =
=
=

1 e2 _
4 "0 4 c
2
1 e2 _

sin2
d
(1
cos )5
Z
sin3
2
d
4 "0 4 c
cos )5
0 (1
2
1 2e2 _ 6
;
4 "0 3c

where

=p

(8.38)

(8.39)

is the relativity factor and the following integral is used,

Z

sin3
d =
(1
cos )5

1
1

1
(1

x2
4
1
dx =
2 3:
5
x)
3 (1
)

(8.40)

The angular dependence of the radiation intensity,

sin2
;
(1
cos )6
peaks at angle

where
cos

In highly relativistic limit

1; the angle

p
0

1 + 24
4

(8.41)

(8.42)

becomes of order

'

1;

(8.43)

which indicates a very sharp pencil or beam of radiation along the direction of the velocity : (This
is also the case for acceleration perpendicular to the velocity as shown in the following section.)
Angular distribution of radiation intensity I ( )) for = 0; 0:2; 0:9 and 0:999 is shown below for a
p
2
:
common acceleration. Note that the radiation intensity rapidly increases with = 1= 1

1.0

0.8
0.6

0.5

0.4
0.2
-1 -0.8

-0.4

0
-0.2

0.2 0.4 0.6 0.8 1

-1.0

-0.5

-0.4

0.5

-0.5

-0.6
-0.8

-1.0

-1

= 0 ( = 1:0):

= 0:2 ( = 1:02):

y
2000

0
2000 4000 6000 8000 10000 12000 14000

-2000

= 0:9 ( = 2:3):

9e+7
-9e+7

2e+8

4e+8

6e+8

8e+8

= 0:99 ( = 7:09):

1e+9 1.2e+9

Polar plot of I( ) for several

factors but with common parallel acceleration.

For linear acceleration, the momentum change of the charged particle is

d
dp
=
dt
dt

mv

=m

v:
_

(8.44)

Therefore, the radiation power can be rewritten as

dp
dt0

P (t ) =

1
2e2
4 "0 3c3 m2

dE
dx0

1
2e2
4 "0 3c3 m2

1
2e4
E2
4 "0 3c3 m2 ac

(8.45)

where Eac is an external acceleration electric eld and

dE
;
dx0
is the energy gradient of a linear accelerator which is at most of the order of 100 MeV/m in practice.
The radiation loss in linear accelerators is negligibly small compared with energy gain. This is one
of the advantages of high energy linear accelerators. Note that the radiation power due to linear
acceleration is independent of the particle energy or the relativity factor :

8.4

Radiation from a Charge in Circular Motion

In circular motion, the acceleration is perpendicular to the velocity. Here we consider highly
relativistic motion of a charged particle with
' 1, for nonrelativistic case has already been
discussed in Chapter 4. A geometry convenient for analysis to follow is shown in Fig.8-2. A
particle undergoes circular motion with an orbit radius in the x z plane and it passes the
origin at t0 = 0: At that instant, the acceleration is in the x direction while the velocity is in the z
direction,
_ = _ ex ;
= ez :
The radiation electric eld can then be written down in terms of cartesian components,
E(r; t) =
=

1
e
n
4 "0 c 3 R
_
e
[(
4 "0 c 3 R

[(n

_]

cos ) cos e + (1

cos ) sin e ] ;

(8.46)

Figure 8-2: Particle undergoing circular motion in the x

At t = 0; the particle passes the origin.

z plane with radius

and frequency ! 0 :

and the angular distribution of radiation power is given by

2

2 _
dP (t0 )
1 e
=
d
4 "0 4 c (1

1
(1
cos )5

cos )2

sin2 cos2

(8.47)

The total radiation power is

2

P (t0 ) =

2 _
Z
1 e
4 "0 4 c
(1

1
(1
cos )5

cos )2

1
2

sin2 cos2

cos )2

=
=

2 _
Z 2
Z
1 e
sin d
d
4 "0 4 c 0
(1
0

1 2e2 jvj
_ 2
4 "0 3c3

1
(1
cos )5

Relevant integrals are:

sin2 cos2
(8.48)

Z
1
1

1
1

dx
(1

1
(1

x)3

2
2 2
)

(1

=2

x2
4
1
4
dx =
=
2
5
3
x)
3 (1
3
)

;
6

In highly relativistic case, the acceleration may be approximated by

jvj
_ =

v2

10

'

c2

(8.49)

Then, the radiation power in terms of the orbit radius

P (t0 ) '

is

1 2e2 c 4
:
4 "0 3 2

(8.50)

To maintain the radiation loss in a circular accelerator at a tolerable level, the orbit radius must
be increased as the particle energy mc2 increases. Note that the radiation power is a sensitive
function of the particle energy in contrast to the case of linear acceleration.
As an example, let us consider the Betatron, the well known inductive electron accelerator
invented by Kerst. In the Betatron, the electron cyclotron orbit is maintained constant,
mc2

= ecB (t) = ecB0 sin !t ' ecB0 !t;

where only the initial phase of the sinusoidal magnetic eld is useful for acceleration, !t
the rate of electron energy gain is
d
dt

1. Then

mc2 = ec B0 ! = const.

Equating this to the radiation energy loss,

P t0 =

1 2
4 "0 3c3

we nd
max

where
re =

'

3!eB0 3
2mc2 re

e2
= 2:8
4 "0 mc2

c2

1=4

10

15

(8.51)

m;

(8.52)

is the classical radius of electron. If = 50 cm, ! = 2

60 rad/s, and B0 = 0:5 T (5 kG), the
upper limit of is about 400 and the maximum electron energy attainable is approximately 200
MeV.

8.5

Fourier Spectrum of Radiation Fields

The formulae such as Eqs. (8.45) and (8.50) only tell us the total radiation power integrated
over the frequency. Radiation eld emitted by highly relativistic particle is hardly monochromatic
but cnsists of broad frequency spectrum. Knowing such frequency spectrum of radiation is of
practical importance for identifying radiation source. A typical example is synchrotron radiation
due to highly relativistic electrons bent by, or trapped in, a magnetic eld. In nonrelativistic limit,
the radiation elds all have a single frequency component corresponding to the classical electron
cyclotron frequency ! c = eB=m: However, as the relativity factor increases, the radiation elds

11

consist of harmonics of the fundamental frequency ! c = eB= m. In highly relativistic case

1;
3
2
the frequency spectrum becomes almost continuous peaking at the frequency ! ' ! c = eB=m:
Frequency spectrum of the radiation electric eld can be formulated by directly applying Fourier
transformation on the eld in Eq. (8.24),
E(r; !) =
=

E(r; t)ei!t dt
1
"
Z
)
1 e 1 n [(n
3
4 "0 c 1
R

Changing the variable from t to t0 and assuming r

E(r; !) =

1 e ei!r=c
4 "0 c r

[(n

ei!t dt:

(8.53)

f (t0 )=0

rp ; we obtain
)

_]

_]

The unit vector n(t0 ) may be regarded constant since r

Then,
n [(n
) _]
d n
' 0
2
dt

exp i! t0

1
cn

rp (t0 )

dt0 :

(8.54)

rp and thus approximated by n ' r=r:

(n

(8.55)

and Eq. (8.53) can be integrated by parts,

E(r; !) =

i! e ei!r=c
4 "0 c r

(n

) exp i! t0

1
cn

rp (t0 )

dt0 :

(8.56)

Any time derivatives contained in the physical electric eld are merely multiplied by i! in the
Fourier space and the disappearance of the acceleration _ is not surprising. Amazing fact about
Eq. (8.56) is that it is applicable to radiation elds which do not require particle acceleration
such as Cherenkov and transition radiation provided a proper velocity of electromagnetic waves in
dielectrics is substituted for c:
The radiation energy (not power) associated with the electric eld is
c"0 r

Z Z

jE(r; t)j2 dtd ; (J).

(8.57)

However, since the electric eld E(r; t) and its Fourier transform E(r; !) are related through
Parsevals theorem:

jE(r; t)j2 dt =

1
2

jE(r; !)j2 d!;

(8.58)

the radiation energy can be written in terms of the Fourier transform E(r; !) as
1
c"0 r2
2

Z Z

jE(r; !)j2 d d!:

12

(8.59)

The quantity
dI(!)
d

1
c"0 r2 jE(r; !)j2 ;
2

(8.60)

can therefore be identied as the radiation energy per unit solid angle per unit frequency. Substituting Eq. (8.56), we nd
dI(!)
e2 2
1
=
!
d
4 "0 8 2 c

(n

) exp i! t

1
cn

rp (t )

dt

1 < ! < 1;

(8.61)

or
dI(!)
e2 2
1
=
!
d
4 "0 4 2 c

) exp i! t0

(n

1
cn

rp (t0 )

dt0 ; 0 < ! < 1:

(8.62)

Let us work on a few examples.

8.6

Synchrotron Radiation I

Synchrotron radiation is due to highly relativistic electrons trapped in a magnetic eld. The
radiation beam rotates together with an electron and is directed along the direction of the velocity
with an angular spread of order
' 1=
1: If the orbiting frequency is ! 0 = eB= me ; the
radiation beam shines a detector for a duration
t0 '

!0

1
;
!0

(8.63)

as seen by the electron at the retarded time t0 . Since

t
=1
t0

=1

'

1
2

(8.64)

which is entirely due to Doppler eect, the pulse width detected is of order
t'

t0
2

1
:
3!
0

(8.65)

Therefore, synchrotron radiation is dominated by frequency components in the range

!'

!0 =

2 eB

me

(8.66)

In ultrarelativistic case, the frequency spectrum of synchrotron radiation can extend to very high
frequencies even for a modest magnetic eld.
We calculate the amount of energy radiated in one period of cyclotron motion T = 2 =! 0 : Since
the radiation power is constant, the radiated energy is
1
E=
T

!0
I (!) d! =
2
13

I (!) d!:

(8.67)

The time integration in the energy spectrum,

Z

e2 2
1
dI (!)
=
!
d
4 "0 4 2 c

T =2

(n

) exp i! t

1
cn

T =2

rp (t )

dt

; 0 < ! < 1;

(8.68)

can be extended from 1 to 1 since the characteristic frequency of the radiation eld is much
higher than the fundamental frequency !
!0;
e2 2
dI (!)
1
=
!
d
4 "0 4 2 c

) exp i! t0

(n

1
cn

rp (t0 )

dt0 ; 0 < ! < 1:

(8.69)

To perform the integration, we assume the trajectory shown in Fig. (8-2) in which an electron
passes the origin at t = 0: The vector n is assumed to be in the y z plane since the radiation
prole is essentially symmetric about the z axis. The trajectory is described by
rp t0 =
and the velocity is
=

cos ! 0 t0 ex + sin ! 0 tez ;

!0
(sin ! 0 tex + cos ! 0 tez ) :
c

(8.70)

(8.71)

Then,
1
n rp (t0 ) = sin (! 0 t) cos :
c
c

(8.72)

)=

(8.73)

Since
n

(n

sin ! 0 tex + cos ! 0 t sin e? ;

where
e? = n

ex ;

(8.74)

is a unit vector perpendicular to both n and x axis, and the radiation lasts for a very short time
and is limited within a small angle ; Eq. (8.73) reduces to
'

! 0 tex + e? :

Within the same order of accuracy, the phase function !(t

! t

n rp
c

where v has been approximated by c but 1

1

!
2

1
2

by
'

14

n rp =c) can be approximated by

sin ! 0 t cos
c

= ! t
'

(8.75)

1
2

t+

c2 3
t ;
3 2

(8.76)

Then,
Z

h
exp
i! t
?

n rp i
dt
c
1
Z 1
1
c2 3
i!
2
(! 0 tex
e? ) exp
+
t
+
t
=
2
2
3 2
1
Z 1
!
1
c2 3
2
t sin
= 2i! 0
t
+
dt ex
+
t
2
2
3 2
0
Z 1
!
1
c2
2
cos
+ 2 t + 2 t3 dt e? :
2
2
3
0
1

dt

(8.77)

The integrals reduce to the modied Bessel functions of fractional orders (or the Airys functions),
Z

3
x
2

cos

3
x
2

sin

where

Then
2i! 0

=p
t sin

cos

!
2

1+

2 2

1
2

1
3

!0t

!
2

1
3

(8.78)

1
d = p K2=3 (x) ;
3

!
1+
3 3!0

; x=

t+

1
d = p K1=3 (x) ;
3

c2 3
t
3 2

c2
t + 2 t3
3

2 2 3=2

dt = 2i

dt = 2

:
2

!0
1
2

(8.79)

!0

(8.80)

1
p K2=3 (x) ;
3
1
p K2=3 (x) ;
3

and for I (!) =d ; we obtain

dI (!)
d

where

1
e2 ! 2
4 "0 3 2 c ! 20

1 3e2 2 2
!
^ 1+
4 "0 2 c

1
2

1
2

1+

2
K2=3
(x) +

+
2

2
K2=3
[^
! 1+

i
2
K1=3
(x)

2 3=2

]+

2
K1=3
[^
! 1+

2 3=2

o
] ;

and !
^ is the normalized frequency,
!
^=

!
:
3 3!0

(8.81)

The modied Bessel functions K2=3 (x) and K1=3 (x) both diverge at x ! 0: However, xK2=3 (x)
2 (x) and x2 K 2 (x)
and xK1=3 (x) are well behaving and vanish at small x: Fig. (8-3) shows 2x2 K2=3
1=3
which represent radiation intensities associated with electric eld polarization along ex (that is, in
the particle orbit plane) and e? ; respectively.
The energy spectrum I (!) emitted during one revolution (T = 2 =! 0 ) can be found by inte-

15

y 0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

2 (x) (solid line) and x2 K 2 (x) (dotted line). The factor of 2 in 2x2 K 2 (x)
Figure 8-3: 2x2 K2=3
1=3
2=3
assumes ' 1= :

grating dI (!) =d

=
'
=

over the solid angle,

I (!)
Z 2
Z
1 3e2 2 2
0 0
d 0 1+ 2
sin d
!
^
4 "0 2 c
0
0
n
o
3=2
2
2
2 3=2
2
1+
K2=3 [^
! 1+
] + 2 K1=3
[^
! 1+ 2
]
Z =2
n
1 6e2 2 2
2
cos d 1 + 2
1 + 2 K2=3
[^
! 1+
!
^
4 "0
c
=2
Z
n
1 6e2 2 1
3=2
2
!
^
1+ 2
1 + 2 K2=3
[^
! 1+ 2
]+
4 "0 c
1
1 6e2
f (^
!) ;
4 "0 c

2 3=2

]+

2
K1=3
[^
! 1+

1+

1+

2
K2=3
[^
! 1+

2 3=2

]+

2
K1=3
[^
! 1+

o
] d

o
]

(8.82)

where 0 = 2
is the polar angle from the axis of electron revolution (y axis),
angle about the y axis, and the function f (^
! ) is dened by
f (^
! ) = 2^
!2

2 3=2

2
K1=3
[^
! 1+

2 3=2

is the azimuthal

2 3=2

o
] d ;

(8.83)

and shown in Fig. (8-4) (linear scale) and Fig. (8-5) (log-log scale). In Fig. (8-5), the straight line
3!
1=3 : f (^
in the low frequency regime !
! ) peaks
0 has a slope of 1=3; and indicates I (!) _ !
3! ;
at !
^ ' 0:14; or ! ' 0:42 3 ! 0 ; and its peak value is about 0.83. In high frequency regime !
0
the spectrum decays exponentially. The energy radiated per revolution can be calculated as
16

0.8

0.6

0.4

0.2

Figure 8-4: The function f (x) = f !=3

E=

I (!) d! '

1 6e2
3
4 "0 c

3!

!0

plotted in linear scale.

Z

f (^
! ) d^
!;

(8.84)

where the integral numerically evaluated is approximately

Z

Then
E=

f (^
! ) d^
! ' 0:713:

I (!) d! =

1 4:08e2
4 "0
c

4!

; (J)

and the radiation power is

P =E

!0
1 0:65e2 c
=
2
2
4 "0

(8.85)

which agrees reasonably well with Eq. (8.50),

P =

1 2e2 c
4 "0 3 2

'

1 0:667e2 c
2
4 "0

; (W).

The discrepancy may be attributed to the various approximations made in the analysis.

8.7

Synchrotron Radiation II

An alternative approach to nding the frequency spectrum of synchrotron radiation is to apply

discrete Fourier analysis in terms of harmonics of the fundamental frequency ! 0 directly to radiation

17

.1

.1e-1

.1e-2

.1e-3

.1e-2

.1e-1

.1

1.

Figure 8-5: Log-log plot of the function f (x) = f !=3

3!

elds. If a charge e is in circular motion with a constant angular frequency ! 0 , the radiation eld
contains higher harmonics of ! 0 and can be Fourier decomposed as follows. Let us recall the
Lienard-Wiechert vector potential,
A(r; t) =

ev(t0 )
n )R(t0 )

(1

(8.86)

f (t0 )=0

where R = jr rp (t0 )j and the subscript f (t0 ) = 0 indicates that all time dependent quantities
should be evaluated at the retarded time t0 determined from the implicit equation for t0 ,
f (t0 ) = t0

jr

t+

rp (t0 )j
= 0:
c

(8.87)

il! 0 t

(8.88)

The vector potential can be Fourier decomposed as

X

A(r; t) =

Al (r)e

where
Al (r) =

0e

1
T

ev(t0 )
eil!0 t dt;
(1 n )R

(8.89)

with T = 2 =! 0 being the period of the circular motion. Changing the integration variable from t
to t0 by noting
dt
=1 n ;
(8.90)
dt0

18

leads to

eikl r
Al (r) '
4 r T
0e

v(t0 )ei(l!0 t

kl n rp )

dt0 ;

(8.91)

where kl = l! 0 =c. Note that the period T remains unchanged through the transformation. Let the
particle trajectory be
rp (t0 ) = (cos ! 0 t0 ex + sin ! 0 t0 ey );
(8.92)
v(t0 ) = ! 0 ( sin ! 0 t0 ex + cos ! 0 t0 ey ):

(8.93)

Since all radiation elds rotate with the charge, the observing point can be chosen at arbitrary
azimuthal angle and we choose = =2; so that n = (1; ; = =2): Then
n rp (t0 ) = sin sin ! 0 t0 :

(8.94)

The velocity in the spherical coordinates is

v(t0 ) = ! 0 (sin cos ! 0 t0 er + cos cos ! 0 t0 e + sin ! 0 t0 e ):
Thus, the

(8.95)

component of Al (r) is given by

Al =

0 e ikl r

4 r

!0
cos
T

Letting x = ! 0 t0 ; and noting

sin sin ! 0 t0 =c)

cos ! 0 t0 eil!0 (t

dt0 :

(8.96)

= ! 0 =c; we can rewrite this as

ikl r

Al = e

0e

! 0 cos
4 r 2

cos xeil(x

sin sin x)

dx:

(8.97)

The integral reduces to

Z

cos xeil(x

sin sin x)

dx

=
=

[Jl+1 (l sin ) + Jl
2
Jl (l sin );
sin

and thus nally,

Al =

0 ec ikl r

Similarly,
Al = i

sin )]

cot Jl (l sin ):

(8.98)

! 0 ikl r 0
e Jl (l sin );
4 r

(8.99)

4 r

1 (l

0e

where use has been made of the recurrence formula of the Bessel functions,
Jl

1 (x)

Jl+1 (x) = 2Jl0 (x):

19

(8.100)

The far-eld radiation magnetic eld can be found from

i

Hl '

Al ;

kl

(8.101)

which yields
e ! 0 kl ikl r 0
e Jl (l sin );
4 r
eckl ikl r
e
cot Jl (l sin ):
=i
4 r

Hl =
Hl

(8.102)
(8.103)

The radiation power associated with the l-th harmonic is

Z
2
jHl j2 d
Pl = c 0 r
Z
1 (el! 0 )2
2 02
=
Jl (l sin ) + cot2 Jl2 (l sin ) sin d :
8 0
c
0
Since Pl = P l , the total power is
P =

1
X

Pl ;

(8.104)

(8.105)

l=1

where Pl is now
1 (el! 0 )2
Pl =
4 0
c

2 02
Jl (l

In nonrelativistic limit

sin ) + cot2 Jl2 (l sin ) sin d ;

1; the l = 1 term is dominant. For x

1:

(8.106)

1;

1
1
J1 (x) ' x; J10 (x) ' :
2
2

(8.107)

Then the lowest order radiation power agrees with the Larmors formula,
P1 '
=

1 (e! 0 )2 2
4 0 c
4
2
2
1 2e a
;
4 0 3 c3

(1 + cos2 ) sin d

(8.108)

where a = v 2 = is the acceleration.

The integral in Eq. (8.105) cannot be reduced to elementary functions. However, the total
power given in Eq. (8.105) should reduce to Eq. (8.50),
P =

1 2 e2 a2
4 0 3 c3

1 2 e2 a2
1
2 2:
3
4 0 3 c (1
)

(8.109)

To show this, we modify the integral by noting

Jl2 (x)

J0 (2x sin ) cos(2l)d ;

20

(8.110)

Jl02 (x)

l2
x2

J0 (2x sin ) cos 2

cos(2l)d :

(8.111)

(8.112)

Then
2 02
Jl (

l sin ) + cot

Jl2 (

Furthermore,

l sin ) =

J0 (2 l sin sin x)(

cos 2x

1) cos(2lx)dx:

J0 (2 l sin x sin ) sin d =

sin(2 l sin x)
;
l sin x

(8.113)

cos 2x

(8.114)

and the power Pl reduces to

1 (e! 0 l)2
Pl =
4 0
c

sin(2 l sin x)
(
l sin x

1) cos(2lx)dx:

Noting
Z

sin x sin(2 l sin x) cos(2lx)dx

=
=
=

1
2
2

[sin(2l + 1)x + sin(1

2l)x] sin(2 l sin x)dx

[J2l+1 (2 lx)

J2l

1 (2

l)]

0
(2 l)
J2l

(8.115)

and
Z

sin(2 l sin x) cos(2lx)

dx
sin x
0
Z Z
= 2l
cos(2l sin x) cos(2lx)dxd
0
0
Z
= 2 l
J2l (2l )d ;

(8.116)

the power Pl can be rewritten as

1 2(e! 0 l)2
Pl =
4 0 c l

2 0
J2l (2l

and the total power is

P =

l(1

J2l (2l )d

(8.117)

1
X

Pl :

l=1

Relevant sum formula of the Bessel functions is

1
X

J2l (2lx) =

l=1

21

x2
2(1

x2 )

(8.118)

Dierentiating by x;

0
2lJ2l
(2lx) =

Also,

x
:
(1 x2 )2

(8.119)

J2l (2lx)dx =

6(1

2 3:
)

(8.120)

This is one of Kapteyn series formulae. (See for example, Mathematical Formulae (in Japanese),
(Iwanami, Tokyo, 1960), vol. 3, p. 212.) Then, nally, the total radiation power becomes
P =

1 2e2 a2
4 0 3c3

a=

v2

(8.121)

which is consistent with the known radiation power from a charge undergoing circular motion.
Analytic expression for the radiation power Pl can be found by exploiting following approximation,
!
l2=3
1
;
(8.122)
J2l (2 l) ' p 1=3 Ai
2
l
where Ai(x) is the Airy function dened by

For large x

1
Ai(x) = p

1
f (x) = p

cos

1 3
t + xt dt:
3

cos

1 3
t + xt dt
3

0
20

(8.123)

1; the function takes the form

Ai(x) '

Also, Ai0 (0) =

1
exp
2x1=4

2 3=2
x
; x
3

1:

(8.124)

0:4587: Using these approximations, we nd the following approximate formulae,

Pl '

1
4 "0

0:5175

1 e2 ! 20
p
Pl '
4 "0 2 c

e2 ! 20 1=3
l ; 1
c
exp

2 l
3 3

; l

(8.125)

(8.126)

Note that the radiation power increases with l in the manner Pl _ l1=3 up to l ' 3 beyond which
Pl decays exponentially. This is consistent with the analysis in the preceding section.

8.8

Free Electron Laser

In a synchrotron radiation source, many beamlines can be installed by bending an electron beam.
In straight sections of race track, no radiation occurs. However, by inserting a device called wiggler,
22

Figure 8-6: In a wiggler, an electron beam is modulated by a periodic magnetic eld. Electrons
acquire spatially oscillating perpendicular displacement x(z) and velocity vx (z) which together with
the radiation magnetic eld BRy produces a ponderomotive force vx (z) BRy (z) directed in the z
direction. The force acts to cause electron bunching required for amplication of coherent radiation.

high intensity radiation can be extracted. A wiggler consists of periodically alternating magnets
and gives an electron beam periodic kick perpendicular to both the beam velocity and magnetic
eld. Electrons receive kicks at an interval
t0 =
where

(8.127)

is the wavelengthof the periodic wiggler structure. Because of Doppler shift, this time
1
interval is shortened by a factor
' 2 2 for a stationary detector in front of the beam,
1
w

t'

w
:
2c

(8.128)

Therefore, the wavelength of resultant radiation is approximately given by

'

w
2

w;

(8.129)

(8.130)

and the frequency by

!'

2c
w

The intensity of free electron laser can be orders of magnitude higher than that of synchrotron
radiation because of coherent amplication through the periodic structure. Electrons tend to be
bunched in the wiggler as the electron beam travels through the periodic structure. In contrast, no
collective interaction between electrons and electromagnetic waves exists in synchrotron radiation.
In electron bunching, the magnetic ponderomotive force plays a major role. Let us assume a

23

periodic wiggler magnetic eld in y direction,

By = B0 cos

= B0 cos kw z; kw =

(8.131)

The Lorentz force is

F = ev

B;

or
Fx =

evz B0 cos kw z:

Then electron acquires a velocity vx in x direction,

vx =

eB0
sin kw z;
m kw

(8.132)

evx

(8.133)

and a resultant ponderomotive force is

BR ;

where BR is the radiation magnetic eld propagating in the form ei(kz !t) along the beam. Since
the acceleration due to the wiggler magnetic eld is in x direction, the radiation electric eld is
predominantly in x direction and radiation magnetic eld BR is in y direction. Then the ponderomotive force directed in z direction is proportional to ei[(k+kw )z !t] and propagates at a velocity
!
:
k + kw
When this propagation velocity matches the electron beam velocity c, strong interaction between
the radiation eld and electron motion takes place and electrons tend to be bunched. This results
in positive feedback for wave amplication. From the condition
!
k
' c; or
= ;
k + kw
k + kw
we readily recover
k=

8.9

Radiation Accompanying

kw ' 2

kw :

Decay

Equation (8.61) for the angular distribution of radiation energy can be applied to cases in which
particle acceleration is not involved explicitly. In decay, an energetic electron (or positron) is
suddenly released from a nucleus together with neutrino. The situation is equivalent to sudden
acceleration of an electron. The duration of acceleration t is limited by the uncertainty principle

24

t mc2 & ~: Therefore, the upper limit of the frequency spectrum should be of the order of
mc2
:
~

! max '

(8.134)

In

decay, the maximum value of is of order of 30.

Let us assume that an electron suddenly acquires a velocity
= v=c and then travels at a
constant velocity. The integration in Eq. (8.61) is limited from t = 0 to 1;
dI(!)
d

=
=

where
yields

e2 2
1
!
4 "0 4 2 c
1
4 "0 4

1 e2 1
ln
4 "0 c

sin

ei!(1

cos )t

dt

e2
2 c (1

is the angle between the velocity

I(!) =

sin
;
cos )2

(8.135)

and the unit vector n: Integration over the solid angle

1+
1

2 = const.;

! . ! max =

mc2
:
~

The frequency spectrum is at up to ! max : Therefore, the total energy radiated through
is approximately given by
E'

1 e2 1
ln
4 "0 c

1+
1

mc2
1
'
ln(4
~

mc2 ;

(8.136)
decay

(8.137)

where the dimensionless quantity ,

=

1 e2
1
'
;
4 "0 c~
137

is the ne structure constant. The energy emitted as radiation through

of the electron energy.

8.10

(8.138)
decay is a small fraction

Cherenkov Radiation

Cherenkov radiation occurs when a charged particle travels faster than electromagnetic waves in
a material medium. It does not require acceleration of charges and the basic mechanism is very
similar to that of sound shock waves in gases. As in the case of
decay, we assume a charge
travelling along a straight line at a velocity = v=c(!); where
1

c(!) = p

"(!)

25

;
0

is the velocity of electromagnetic waves in a dielectric having a permittivity "(!): Eq. (8.61) should
be modied as follows after taking into account the proper denition of c(!);
Z

e2
dI(!)
= 0 2
! 2 v 2 sin2
d
4 4 c(!)
Note that the integration limits are from
Z

ei!(1

ei!(1

cos )t

dt :

(8.139)

1 to 1: The time integral is singular,

cos )t

dt = 2

[!(1

cos )];

and the condition for radiation is

cos =
or
(!) =

< 1;

v
> 1:
c(!)

(8.140)

Then,
dI(!)
e2
= 0
! 2 v 2 sin2
d
4 (2 )2 c(!)

[!(1

cos )]:

(8.141)

Square of a delta function is not integrable and the radiation energy simply diverges. This is merely
due to the assumption that the charge is radiating forever from t = 1 to 1 which is of course
unphysical. It is more appropriate to consider a radiation power rather than energy. For this
purpose, we consider a thin slab of the dielectric of thickness dz. The transit time over the distance
dz is T = dz=v and we calculate energy radiated during that time,
Z

e2
dI(!)
= 0 2
! 2 v 2 sin2
d
4 4 c(!)

T =2

i!(1

cos )t

dt :

(8.142)

T =2

The integral can be carried out easily,

Z
Thus

where

T =2
T =2

ei!(1

cos )t

dt =

2
!(1

cos )

dI(!)
e2 ! 2
= 0 2
sin2
d
4 4 c(!)
1
= (1
2

sin (1

sin

cos )! T:

26

cos )

T!
:
2

(dz)2 ;

(8.143)

(8.144)

In high frequency regime ! T

1; the function (sin = )2 may be approximated by a delta
function ( ): Integration over the solid angle yields
dI(!)
dz

1
4 "0

e
c0

1
4 "0

e
c0

! 1
! 1

1
"(!) 0 v 2

(8.145)

p
where c0 = 1= "0 0 is the speed of light in vacuum. The rate of energy loss due to Cherenkov
emission is given by
dE
1
=
dz
4 "0

e
c0

2Z 1

1
"(!) 0 v 2

! 1

d!;

0 < ! < 1:

(8.146)

The result obtained is meaningful only if

> 1;

1
or v > c(!) = p
"(!)

(8.147)

which is the condition for Cherenkov radiation. The instantaneous radiation power can be estimated
from
Z
d
P =
I(!)d!
dt
Z
1
e 2
1
=
v ! 1
d!:
(8.148)
4 "0 c0
"(!) 0 v 2
In order to nd the eld proles emitted through Cherenkov radiation, we start from the wave
equations for the potentials in a material medium,
@2
@t2

~" (r; t) =

free ;

(8.149)

@2
@t2

A(r; t) =

0 J;

(8.150)

r2

"
0~

r2

"
0~

where ~" is the dielectric operator containing time derivative,

~" = ~"

@
@t

(8.151)

For a charged particle e travelling at a constant velocity v; the charge density and current density
are described by
= e (r vt);
(8.152)
J = ev (r

27

vt):

(8.153)

Then, after Fourier-Laplace transformation, the Fourier potentials can readily be found,
2 e

(k; !) =
"(!)

(!

!2
c2 (!)

k2
2 e

A(k; !) =

0v
(!
!2
c2 (!)

k2

(8.154)

k v);

(8.155)

k v);

where, as before,
1
"(!)

c2 (!) =

(8.156)

and the transformation

Z Z

(r

vt)ei(!t

k r)

dV dt = 2

(!

k v);

(8.157)

is substituted. Since the physical electric eld is

E(r; t) =

r (r; t)

@
A(r; t);
@t

(8.158)

the Fourier component of the electric eld is given by

E(k; !) =
=

ik (k; !) + i!A(k; !)
!
v
k
2
c (!)
2 ie
!2
"(!) k 2
c2 (!)

(!

k v):

(8.159)

Similarly, the Fourier-Laplace component of the magnetic eld is

B(k; !) = ik

A(k; !)
k v
= 2 ie 0
(!
2
!
k2
c2 (!)

k v):

The physical electromagnetic elds can then be found through inverse transformations,
Z
Z
1
3
d k d!E(k; !)ei(k r !t) ;
E(r; t) =
(2 )4
B(r; t) =

1
(2 )4

d k

d!B(k; !)ei(k r

!t)

(8.160)

(8.161)

(8.162)

To proceed further, we assume that the charged particle is travelling along the z axis at a
constant velocity v: The system is symmetric about the axis and we may assume an observing
point in the x z plane without loss of generality. We denote the cylindrical coordinates of the
28

Figure 8-7: Geometry for Fourier inverse transform.

observing point by ( ;

= 0; z) and spherical Fourier coordinates by k =(k; ; ): Then,

k r = k(z cos + sin cos ):

Because of the cylindrical symmetry, radiation of energy is expected in the radial direction ; and
the relevant Poynting vector is
S

= (E

H )

Ez (r; t)H (r; t):

(8.163)

The energy radiated per unit length along the particle trajectory (z axis) can be calculated from
dE
dz

=
=

2
Z

Ez (r; t)H (r; t)dt

Ez (r; !)H (r; !)d!;

(8.164)

where Ez (r; !) is the Laplace transform of the electric eld,

Ez (r; !) =

1
(2 )3

d3 kEz (k; !)eik r ;

(8.165)

and H (r; !) is the Laplace transform of the magnetic eld,

H (r; !) =

1
(2 )3

29

d3 kH (k; !)eik r :

(8.166)

The Laplace transform of the axial electric eld Ez (r; !) can be calculated as follows:
Ez (r; !) =
=

1
(2 )3

ie
(2 )2

i 0 e!
(2 )2
i

0 e!

d3 kEz (k; !)eik r

Z

k 2 dk

sin d

1Z

1
2

1=

1
Z

i!z
c(!)

exp

!v
c2 (!)

k cos

d
k2

"(!)
(
i!
z
exp
+
c(!)
0

"

1=

where
=

! q
c(!)

J0

kc(!)
;
!

!2
c2 (!)

kv cos )eik(z cos

(!
1
(

)2

cos

)#

(8.167)

v
:
c(!)

+ sin cos )

(8.168)

Letting
2

2;

(8.169)

we nally obtain

Ez (r; !) =
=

0 e!

2
ie
2

exp

! exp

i!z
c(!)

i!z
c(!)

where
=

!
c (!)

1
2

1
2

K0 (

J0
2

!
c(!)

1+

);

1;

(8.170)

(8.171)

and use is made of the integral representation of the modied Bessel function K0 (ax);
Z

K0 (ax) =

In the asymptotic regime j j

1; K0 (

tJ0 (at)
dt:
t2 + x2

(8.172)

) approaches
r

(8.173)

For this to be propagating radially outward in the form eik ; we must choose
s

!
c (!)

!
i
c (!)

1
2

30

1
1

2;

> 1:

(8.174)

With this choice for ; the Laplace transform of the axial electric eld becomes proportional to
"

i!
exp
c(!)

1
2

!#

(8.175)

and the radial and axial wavenumbers can be identied as

s
!
1
!
1
k =
kz = ;
2;
c(!)
v

(8.176)

respectively. Cherenkov radiation is conned in a cone characterized by an angle ;

sin =

c(!)
;
v

(8.177)

as shown in Fig. 8-8.

Figure 8-8: Cherenkov cone. Radiation elds are conned in the cone.

The Laplace transform of the azimuthal magnetic eld is given by

Z
1
H (r; !) =
d3 kH (k; !)eik r
(2 )3
Z
Z
Z 2
ie
kv sin
2
=
k
dk
sin
d
d
2
!2
(2 )
k2
0
0
2

(!

kv cos )

c (!)

ik(z cos + sin cos )

e
e
K1 (
2

) exp

i!
z ;
c(!)

31

(8.178)

where the following integral

x2 J1 (bx)
dx = aK1 (ab);
x2 + a2

(8.179)

is noted. (Calculation steps are left for an exercise.) Substituting Ez (r; !) and H (r; !) into Eq.
(8.164), we nd
Z 1
i 0 e2
1
dE
K0 ( )K1 ( )d!:
(8.180)
=
! 1
2
2
dz
(2 )
1
In the asymptotic region j j

1; this reduces to
dE
1 e2
=
dz
4 "0 c20

! 1

1
2

d!;

(8.181)

in agreement with the earlier result, Eq. (8.146). Eq. (8.180) can be used even when Cherenkov
condition is not satised. In this case, energy loss is through near eld Coulomb interaction between
a charged particle and ions and electrons in molecules in the dielectric media. We will return to
this problem in Section 8.12.

8.11

Transition Radiation

Transition radiation occurs when a charge crosses a boundary of two dielectric media. No acceleration is required, nor is it necessary for charge to move faster than the speed of light as in Cherenkov
radiation. In this respect, transition radiation is a least demanding radiation mechanism. Radiation emitted from a charge approaching a conductor is an extreme case of transition radiation with
an innite permittivity, and may be regarded as the inverse process of radiation accompanying
decay. Disappearance, rather than creation, of charge is responsible for transition radiation.
We rst consider a simple case: a charge e approaching normally a conducting plate at a velocity
v (> 0). On impact, the charge is assumed to come to rest. A conducting plate is mathematically
equivalent to an innitely permissive dielectric plate. An image charge e moving in the opposite
direction in the conducting plate can be introduced so that the current density is
Jz (r; t) =

ev [ (z + vt) + (z

vt)] (x) (y);

1 < t < 0;

(8.182)

where at t = 0 (or z = 0) the particle is brought to rest. Its Laplace transform is

Jz (r; !) =
=

Jz (r;t)ei!t dt

e exp

!
jzj
v

(x) (y):

(8.183)

Let the observing point be at P located at (r; ): (The system is symmetric about the axis and thus

32

Figure 8-9: Radiation from a charge impinging on a metal surface. Sudden deceleration at the
metal surface is the inverse process of radiation accompanying beta decay.

is ignorable.) The vector potential at P can be calculated in the usual manner,

Z
Jz (r0 ; !) ikjr r0 j 0
0
0 ikr
Az (r; !) =
e
dV '
e
Jz (r0 ;!)e ik r dV 0
0
4
jr r j
4 r
Z 1h
i
e
!
!
0
0
0
=
exp
i z 0 e ikz cos + exp
i z 0 eikz cos dz 0
eikr
4 r
v
v
0
e
1
1
1
= i 0 eikr
+
; z > 0;
(8.184)
4 r
k 1 + cos
1
cos
0

and the magnetic eld from

H (r; !) '
=

ikAz sin = 0
e ikr
1
e
+
4 r
1 + cos
1

33

1
cos

sin :

(8.185)

The angular distribution of radiation energy is thus given by

dI(!)
d

=
=
=

1 2
r c 0 jH (r; !)j2
2
2
e2 2 c 0
1
1
+
sin2
32 3
1 + cos
1
cos
1 e2 v 2
sin2
;
1 < ! < 1:
2
4 "0 2 2 c3 (1
cos2 )2

Integration over the solid angle in the region z > 0 (0 <

I(!) =

1 e2 1
4 "0 2 c

1+

ln

1+
1

(8.186)

< =2) yields

2

; ! > 0:

R1
Evidently, the radiation energy 0 I(!)d! diverges. This is due to the assumption of a perfect
conductor. In practice, metals cannot be regarded as perfect conductor. The condition that the
surface impedance of metal
r
i! 0
;
Z=
i!"0 +
p
="0 imposes an upper limit
be su ciently small compared with the free space impedance Z0 =
R !m a x 0
I(!)d!:
of the frequency, !"0
and a cuto emerges in the integral 0

Figure 8-10: Transition radiation emitted by a charge passing through a dielectric boundary.

We now analyze the case of a dielectric slab having a relative permittivity "r = "="0 : In this

34

case, the particle continues to travel after passing the boundary and the current density is now
Jz (r; t) =

ev (x) (y) (z + vt);

1 < t < 1:

Its Laplace transform is

Jz (r; !) =

ev (x) (y)e

!
ivt

(8.187)

(8.188)

The contribution to the radiation elds from the region z > 0 consists of two parts, one directly
from the charge (as in free space) and the other via reection at the dielectric boundary. Denoting
the magnetic reection coe cient by in the Fresnels formulae,
p
"
"r cos
p r
=
"r cos + "r

sin 2
;
sin 2

(8.189)

and following the same procedure as in the case of conductor plate, we nd

H 1 (r; !) '

ev ikr
e
4 cr

1+

cos

sin ;

cos

z > 0:

(8.190)

The contribution from the region z < 0 involves refraction at the boundary, and thus additional
retardation because of the longer path length. In Fig.8-10, for z < 0, we observe
l = r + z 0 tan 00 sin ; z 0 < 0
l0 = z 0 = cos 00 ;
p
sin 0
";
00 =
sin
and thus
kl +

kl0 = kr

kz

sin 2 :

(8.191)

Then the contribution from the region z < 0 to the integral becomes
H 2 (r; !) =
Note that the factor 1
total magnetic eld is H

ev
(1
4 cr

1+

1
"r

sin 2

eikr sin

(8.192)

here indicates the eld amplitude transmitted into the air region. The
1 + H 2 ; and the angular distribution of the radiation energy is given by
dI(!; )
r2 c
=
d
2

where
A=

1
1+

cos

In the case of ideal conductor "r ! 1;

jH (r;!)j2 =

cos

2 2
0e v
A2 sin 2
32 3 c

1+

1
p

sin 2

"r

= 1; we recover

dI(!; )
1 e2 v 2
=
d
4 0 2 2 c3
35

sin
2
cos2

(8.193)

(8.194)

When "r = 1; = 0; radiation evidently disappears.

The factor A in Eq. (8.194) does not fully agree with that in the original work by Frank and
Ginzburg,
1+
1
p
+
;
(8.195)
A0 =
1 + cos
1
cos
"r (1 +
"r sin 2 )
although this too vanishes when "r ! 1 and reduces to the case of conducting medium when
"r ! 1:

8.12

Energy Loss of Charged Particles Moving in Dielectrics

The formula derived in Eq. (8.180) yields a physically meaning energy loss rate even when the
Cherenkov condition is not satised,
< 1: A charged particle moving in a dielectric medium
collides with atoms and lose its energy through Coulomb interaction with electrons in atoms.
Electrons in an atom are bounded. However, they do respond to electromagnetic disturbance and
absorb energy through the resonance "(!) = 0, where
"(!) = "0

! 2p

!2

! 20

(8.196)

Resonance of the type

1
x

x0

(8.197)

can be handled mathematically by introducing an imaginary part,

1
x

x0

=P

1
x

x0

(x

x0 );

(8.198)

where P stands for the principal part. This is justiable because the imaginary part of the function

1
x x0
=
x0 + i"
(x x0 )2 + "2

remains nite even in the limit " ! 0;

lim

"!0

(x

(x

"
;
x0 )2 + "2

"
dx = :
x0 )2 + "2

(8.199)

(8.200)

Physically, the resonance leads to absorption of wave energy by charged particles in a material
medium dielectrics, plasmas, etc.
The characteristic scale length of interaction between a charged particle and atoms in a dielectric is evidently of the order of atomic size which indicates that the interaction is of near-eld,
nonradiating nature dominated by longitudinal (electrostatic) elds. As we will see, the major
contribution to the energy loss occurs through the pole of the dielectric function, "(!) = 0:
In the near-eld region
1; the modied Bessel functions K0 ( ); K1 ( ) may be approx-

36

imated by
K0 (

where

= 0:5772

)'

ln

K1 (

)'

E;

2
1

(8.201)

(8.202)

is the Eulers constant. The real part of Eq. (8.180) becomes

dE
e2
' 0 2 Re
dz
(2 )

1
1

with
!
=
c(!)

1
"(!) 0 v 2

i! 1
s

1;

ln

d!;

< 1:

(8.203)

(8.204)

The dielectric function "(!) is in the form

"(!) = "0

! 2p

!2

! 20

Therefore, the integration can be carried out by evaluating the pole contribution at "(!) = 0; which
occurs at
q
(8.205)
! 2p + ! 20 ;
!=
and exploiting Plemeljs formula,
lim

!0

1
1
=P
a+i
x a

(x

a);

where P indicates the principal part of the singular function 1=(x

dE
1
'
dz
4 "0

e! p
v

a): The result is

1
1:12v
A:
ln @ q
! 2p + ! 20
0

(8.206)

As the minimum distance ; the intermolecular distance may be substituted because the shell
electrons eectively shield the electric eld of the charge well inside the atom.
In a plasma, ! 0 is evidently zero (because electrons in a plasma are free). Then,
dE
1
'
dz
4 "0

e! p
v

where
min

'

ln

3
4 n

min ! p

(8.207)

1=3

is the average distance between ions. For a Maxwellian electron distribution with a temperature

37

Te ; the average energy loss rate may be estimated from

1 (e! p )2
dE
ln
'
dt
4 "0 3vT e
where
D

p
8 2
n
3

3
D

vT e
;
!p

(8.208)

(8.209)

is the Debye shielding length.

8.13

Bremsstrahlung

Whenever a charged particle collides with another charged particle, electromagnetic radiation occurs
due to acceleration by Coulomb force. Collisions between like particles (e.g., electron-electron)
emit quadrupole radiation while collisions between unlike particles (e.g., electron-ion) emit dipole
radiation. Bremsstrahlung is due to collisions between electrons and ions and provides a basic
mechanism for x-ray production.
Let an electron approach an ion having a charge Ze with a velocity v (
c) and impact
parameter b: The acceleration due to Coulomb force is of the order of
a'

1 Ze2
;
4 "0 mb2

(8.210)

and consequent radiation power can be estimated from the Larmors formula,
P '

1 2e2 a2
:
4 "0 3c3

(8.211)

The total energy radiated can in principle be found by integrating the power over time along the
electron trajectory. However, in experiments, one is seldom interested in measuring radiation power
or energy associated with a single electron. What is more relevant is the radiation associated with
a beam of electrons impinging on an ion. In this case, some electrons have impact parameters
vanishingly small. However, the impact parameter has a lower bound imposed by the uncertainty
principle,
~
(8.212)
bmin ' ;
p
where p is the electron momentum. For an impact parameter b; the time duration in which the
acceleration is signicant is
b
t' :
(8.213)
v

38

Therefore, energy radiated by a single electron is

E

= P t
1
2
'
3
(4 "0 ) 3c3

Ze3
m

1
b3 v

(8.214)

For an electron beam having a density n; the radiation power can thus be estimated from
P

= ne v

bmin

E2 bdb

4
1
(4 "0 )3 3

Ze3
m

ne mv
:
c3 ~

(8.215)

If the ion density is ni ; the quantity P ni denes the power density,

1
4
3
(4 "0 ) 3

Ze3
m

ne ni mv
;
c3 ~

(W/m3 ):

(8.216)

The frequency spectrum of radiation energy may qualitatively be found as follows. Since the
characteristic time of acceleration
b
(8.217)
= ;
v
is short, radiation occurs as an impulse and the spectrum is at in the region 0 < ! < v=b; and
vanishes for ! > v=b: Since the minimum impact parameter is
bmin '

~
;
mv

(8.218)

the upper limit of the frequency spectrum extends to

! max '

mv 2
; or ~! . mv 2 :
~

(8.219)

This is essentially a statement of energy conservation, that is, the maximum photon energy emitted during bremsstrahlung is limited by the incident electron kinetic energy, which is reasonable.
Therefore, the frequency spectrum of bremsstrahlung is

I(!; b) =

8
1
2 (Ze3 )2
v
>
>
>
< (4 "0 )3 3 mc2 mv 2 b2 ; 0 < ! < b ;
>
>
>
: 0;

(8.220)

v
!> :
b

I(!; b) has dimensions of J/frequency. It is convenient to introduce a radiation cross-section (!)

39

dened by
Z

(!) =

bmax

I(!; b)2 bdb

bmin

4
1
(4 "0 )3 3
4
1
3
(4 "0 ) 3

=
=
The integral over the frequency,

(Ze3 )2
ln
mc2 mv 2
(Ze3 )2
ln
mc2 mv 2

bmax
bmin
mv 2
:
~!

(8.221)

(!)d!;

(8.222)

evidently diverges at the lower end ! ! 0. To remedy this di culty, Bethe and Heitler recognized
that if the velocity v is understood as the mean value of the initial and nal velocities, i.e., before
and after emission of a photon,
1
(vinitial + vnal )
2
p
p
1
p
E + E ~! ;
2m

v =
=

(8.223)

the integral remains nite,

Z

(!)d! =

=
=

(Ze3 )2

1
4
3
(4 "0 ) 3 mc2 mv 2
1
4
3
(4 "0 ) 3
4
1
3
(4 "0 ) 3

! max

Z
(Ze3 )2 1
ln
mc3 ~ 0
(Ze3 )2
:
mc3 ~

0 p
p
E + E ~!
B
ln @
~!

1+

p
1
p
x

dx; x =

C
A d!

~!
;
E

(8.224)

The integral in the intermediate step is unity. Multiplying by the ion density ni ; we thus obtain
the bremsstrahlung rate per unit length,
dE
dz

= ni
=

(!)d!

1
4 (Ze3 )2 ni
;
(4 "0 )3 3 m2 c3 ~

(8.225)

and radiation power density,

P=V =

1
4 (Ze3 )2 ne ni v
; (W m
3
(4 "0 ) 3
mc3 ~

):

(8.226)

This agrees with the earlier qualitative estimate in Eq. (8.216).

Relativistic correction to the classical bremsstrahlung formulae can be readily found if we move
40

to the electron frame wherein the electron velocity is nonrelativistic. Since the energy and frequency
are Lorentz transformed in the same manner, and the transverse dimensions are Lorentz invariant,
it follows that the radiation cross-section (!) is Lorentz invariant,
lab (! lab )

(! 0 );

where the primed quantities are those in the electron frame. The frequencies ! lab and ! 0 are related
through the relativistic Doppler shift,
! 0 = ! lab (1

cos ); ! lab = ! 0 (1 +

cos 0 );

(8.227)

where and 0 are the angles with respect to the electron velocity in each frame. Since in the
electron frame, the radiation is conned in a small angle about 0 = =2; we have
!0 '

! lab

(8.228)

The collision time is shortened by the factor since the transverse eld is intensied by the same
factor through Lorentz transformation. Therefore, the maximum impact parameter is modied as
bmax =

2v
v
=
;
0
!
! lab

(8.229)

and the radiation cross-section in the laboratory frame becomes

(!) =

(! 0 ) =

1
4 e4 (Ze)2
ln
(4 "0 )3 3 m2 c3 v 2

2 mv 2

~!

(8.230)

It is noted that the minimum impact parameter remains unchanged through the transformation
because it is essentially the Compton length based on the uncertainty principle.

8.14

Radiation due to Electron-Electron Collision

In this case the dipole radiation is absent because in the center of mass frame, two electrons stay
at opposite positions, r1 = r2 ; and the dipole moment identically vanishes.. The lowest order
radiation process is that due to electric quadrupole. (The magnetic dipole moment also vanishes.)
The quadrupole moment tensor is
1
Qij = exi xj ;
(8.231)
2
where xi is the i-th component of the relative distance 2r: In Chapter 5, a general formula for the
quadrupole radiation power has been derived. Noting
d3
:::
:: :
: ::
:::
(xi xj ) = x i xj + 3xi xj + 3xi xj + xi x j ;
dt3

41

(8.232)

Figure 8-11: Colliding electrons in the center of mass frame. The impact parameter is b:

::

xi =
:::

xi =

we nd

1 2e2
xi ;
4 "0 mr3

1 2e2
4 "0 m

d3
1 2e2
(xi xj ) =
3
dt
4 "0 mr3
:::

Q=

1 e3
4 "0 mr3

xi
r3

(8.233)

3xi vr
r4

4(vi xj + vj xi )

(8.234)
6vr xi xj
r

6vr xi xj
r

4(vi xj + vj xi )

(8.235)
(8.236)

Substituting this into the quadrupole radiation power,

2

1
1 4 X ::: 2
P =
3
Qij
4 "0 60c5

X :::
Qii

ij

!2 3

5;

(8.237)

we obtain, after somewhat lengthy calculations,

P =

1
2
4 "0 15c5

e3
m

v 2 + 11v 2
r4

(8.238)

where v is the component of the velocity related to the initial angular momentum bv0 = r(t)v (t)
with v0 the velocity at r ! 1: Energy conservation reads
1
1
1 e2
mv02 = mv 2 +
:
2
2
4 "0 2r

42

(8.239)

Then

4e2
(bv0 )2
+ 11 2 ;
4 "0 mr
r

v 2 + 11v 2 = v02

(8.240)

and the total radiation energy can be found by integrating the radiation power over time along the
trajectory,
2Z 1
2
e3
1
4e2
(bv0 )2
1
2
v
dt:
(8.241)
+
11
E=
0
4
4 "0 15c5 m
4 "0 mr
r2
1 r
This can be converted to an integral over the distance r by noting
dt =

E=

e3

2
1
4 "0 15c5

dr
=s
vr

dr
v02

(bv0 )2
r2
v2
1 0
s
r4

rmin

(8.242)

4e2
4 "0 mr
4e2
(bv0 )2
+ 11 2
4 "0 mr
r
(bv0 )2
r2

v02

dr; (J)

(8.243)

4e2
4 "0 mr

where rmin is the distance of the closest approach,

rmin =

4e2

1 @
+
m
2v02

4e2
m

+ 4(bv0 )2 A :

(8.244)

However, the radiation energy by a single electron pair is of no practical interest. What is more
relevant is the radiation power emitted by an electron beam impinging on a single electron which
can be evaluated from
Z 1
E(b)bdb
P = 2 nv0
0

1
4
4 "0 15c5

e3

2 nv0

bdb

rmin

v2
1 0
s
r4

4e2
(bv0 )2
+ 11 2
4 "0 mr
r

v02

(bv0 )2
r2

dr; (W). (8.245)

4e2
4 "0 mr

The double integral reduces to

Z

4e2
(bv0 )2
+ 11 2
4 "0 mr
r

v2
1 0
bdb
dr 4 s
r
rmin

25
v0
3

rmin

v02

1
r5

4re2
mv02

43

(bv0 )2
r2
3=2

dr =

4e2
4 "0 mr
5 mv03
:
6 e2

(8.246)

Therefore,
P =

1 4 ne4 v04
; (W).
4 "0 9 mc5

(8.247)

This denes a radiation cross-section,

=

P
1
nmv03
2

8 v 2
r ; (m2 )
9 c e

where 21 nmv03 is the energy ux density of the beam and

re =

1 e2
= 2:85
4 "0 mc2

is the classical radius of electron.

44

10

15

m;

(8.248)

Problems
8.1 A charged particle with charge e and mass m undergoes motion in external electric and
magnetic elds E and B: Show that the radiation power is given by
1 2e2 6 2
a
(
a)2
4 "0 3c3
2e2 2
1
(E + v B)2
4 "0 3m2 c3

P (t0 ) =
=

where a is the acceleration. Since a2

( E)2 .

E)2 ;

a)2 is positive denite, so is (E + v

Note: In an instantaneous rest frame of the particle, v = = 0 and

power in that frame is thus with respect to the proper time,

B)2

= 1: The radiation

dE 0
1
2e2
= P( ) =
E02 ;
d
4 "0 3m2 c3
where E0 is the electric eld also in that frame. Since the electric eld is Lorentz transformed
as
E0k = Ek ; E0? = (E? + v B);
we readily see that
P ( ) = P (t0 );
that is, the proper radiation power is Lorentz invariant. This is not surprising because both
time and energy are Lorentz transformed in a similar manner.
8.2 Carry out the integration in the radiation power due to acceleration perpendicular to the
velocity
2 _
?
1 e
0
P (t ) =
4 "0 4 c

(1

1
(1
cos )5

to obtain
2 _
1 2e
0
P (t ) =
4 "0 3c

cos )2

1
2

sin2 cos2

2
4

Plot the angular distribution of the radiation intensity

f( ) =
for

(1

1
(1
cos )6

cos )2

1
2

sin2 cos2

= 0 and =2 in a highly relativistic case.

8.3 A nonrelativistic electron with velocity v collides head-on with a heavy negative ion having
a charge Ze:Find the total energy radiated. What is the dominant frequency component?
45

8.4 A highly relativistic electron passes by a proton at an impact parameter b:Assuming that
the electron trajectory is unaected by the proton and radiation, show that the total energy
radiated is
2
4
e6
1
e6 2
1
:
'
E=
2
4 "0 12m2e c3 b3 v 1
4 "0 4m2e c4 b3
What is the dominant frequency component contained in the radiation eld?
8.5 An electron having an initial energy E(t = 0) = 0 mc2 undergoes cyclotron motion in a
magnetic eld B: Show that the rate of energy loss to radiation is given by
dE(t)
1 2e4 B 2 2
E
=
dt
4 "0 m4 c5

m2 c4 :

Integrate this and nd an expression for E(t):

8.6 An electron is placed in a plane wave E0 ei(kx
Show that the radiation power is

!t) :

P = I0

If ~!

mc2 ; electron recoil may be ignored.

8 2
r ;
3 e

where I0 = c"0 E02 (W/m2 ) is the intensity of the incident wave and
re =

e2
= 2:8
4 "0 mc2

is the classical radius of electron.

=

10

15

(m),

8 2
r ;
3 e

is known as Thomson scattering cross section.

Note: For ~!
mc2 ; Klein and Nishina, solving the Dirac equation to take into account
electron recoil, obtained
mc2 1
2~!
' re2
+ ln
:
~! 2
mc2
8.7 The light from Crab nebula is believed to be synchrotron radiation emitted by highly energetic
electrons trapped in a weak intergalactic magnetic eld. If the electron energy is 2 1011 eV
and B = 10 7 T, what is the dominant wavelength component?
8.8 Perform numerically the integration in the radiation power of l-th harmonic of synchrotron
radiation,
Z
sin(2 l sin x) 2
1 (e! 0 )2 l
Pl =
( cos 2x 1) cos(2lx)dx;
4 0
c
sin x
0
for the case

= 0:99 ( ' 7) and plot the result as a function of l:

8.9 Estimate the bremsstrahlung cooling rate (in W/m3 ) of a hydrogen plasma for an electron
temperature of 10 keV and plasma density of n = 1020 m 3 : If the plasma is conned by a
magnetic eld of 5 T, what is the cooling rate due to cyclotron radiation if reabsorption by
46

the plasma is ignored? (Cyclotron radiation can be reabsorbed by a plasma if its density is
high enough and its contribution to energy loss may be ignored.)
8.10 Assuming a permittivity in the form
" (!) =

! 2p

!2

! 20

"0 ;

carry out the integration in the rate of Cherenkov radiation,

dE
1 e2
=
dz
4 "0 c20

! 1

1
2

d!:

8.11 Two identical charges are moving along a circular orbit of radius a at an angular velocity !:
Assume nonrelativistic motion, !a
c: Angular separation of the charges is 0 : If 0 = 0;
the problem reduces to a charge 2e rotating at ! which radiates as a dipole. If 0 = ; the
dipole moment vanishes and higher order quadrupole radiation should be considered. At what
angular separation 0 does the dipole radiation power become equal to quadrupole radiation
power? Is it necessary to consider magnetic dipole radiation?
8.12 Four identical charges are placed along a circle of radius a with equal angular spacing =2:
If the system rotates about the axis at angular velocity !; what is the lowest order radiation
power? What would happen if the number of charges N increases?
8.13 Generalize the quadrupole radiation in electron-electron collision to relativistic velocity.

47

Chapter 9

Conned and Guided Waves

9.1

Introduction

Electromagnetic waves can be conned in a volume surrounded by a conductor wall as in microwave

cavities and also guided to propagate in conducting tubes and dielectric bers. Parallel wire
transmission lines and coaxial cables are also used in transmitting low frequency waves. In this
chapter, electromagnetic waves in cavities and waveguides will be discussed. In waveguides, axial
electric and magnetic elds Ez and Hz satisfy scalar wave equation and can be readily solved by
incorporating the boundary conditions. Transverse elds E? and H? can then be found in the
form of linear combination of the axial elds. In conductor waveguides, electromagnetic waves can
be classied into modes without axial magnetic eld (TE mode) and modes without axial magnetic
eld (TM mode). The axial phase velocity of electromagnetic waves in conductor waveguides with
a smooth wall is always larger than c: In slow waveguides with a periodically structured wall, the
phase velocity can be adjusted close to c as needed in linear accelerators. In dielectric bers, there
can be no pure TE or TM modes because evanescent elds outside the ber do aect mode structure
through the boundary condition at the ber surface. Optical bers used in communication all have
gradual radial ( ) change in the index of refraction, n( ); and poses a problem of wave propagation
in a nonuniform medium.

9.2

Spherical Cavity

Electromagnetic waves in a spherical cavity surrounded by a conducting wall may be analyzed in

terms of the TE and TM spherical eigenvectors formulated in Chapter 6,
E = i![r
B=r
where

and

r +r
(r

(r

r ) + k2 r

r )];

(9.1)

r ;

(9.2)

both satisfy the Helmholtz equation

@2
@r2

2 @
1 1 @
+ 2
r @r r sin @

Here k = !=c: Solutions for

and

sin

@
@

1
r2 sin2

@2
@

(r)

+ k2 @

(r)

A = 0:

(9.3)

are in the form of standing (non-propagating) wave,

8
< sin m
(r; ; ; t); (r; ; ; t) = jl (kr)Plm (cos )
: cos m

9
=
;

i!t

(9.4)

Note that nl (kr) is discarded because the elds should remain nite at r = 0: As in mechanical
standing waves wherein the sum of potential energy and kinetic energy is constant, in electromagnetic standing waves the sum of electric energy and magnetic energy is constant. This is guaranteed
because the magnetic eld is out of phase with the electric eld by =2 as seen from Eqs. (9.1) and
(9.2). For TE modes without radial component, the boundary condition that E and E vanish at
the inner wall r = a yields
jl (ka) = 0:

(9.5)

cxls
;
a

(9.6)

Then the resonance frequency is given by

! ls =

where xls is the s-th root of jl (x) = 0. Some roots are listed below.
l=0
s=1

4:493

5:763

6:988

9:095

10:417

7:725

10:904 12:323 13:698

14:066 15:515 16:924

For TM modes, the radial component of the magnetic eld should vanish at r = a: This yields
d
[rjl (kr)] = 0 at r = a;
dr

(9.7)

and the resonance frequency of TM mode is

! ls =
where x0ls is the s-th root of

d
dx [xjl (x)]

cx0ls
;
a

(9.8)

= 0:

l=0

=2

2:744

3:870

4:973

3 =2

6:177

7:443

8:722

5 =2

9:317

10:713 12:064

7 =2 12:486 13:921 15:413

s=1

The smallest eigenvalue is that of l = 1 TM mode,

c
! T M;l=1 ' 2:744 ; ka = 2:744:
a
The Q value (cavity quality factor) is dened by
time averaged energy stored in a cavity
time averaged rate of energy absorption by the wall
R 0 2
R
2
2
2 H dV
V
V H dV
R
= !q
=
;
! 0 R 1
H)2 dS
S (n
(n H)2 dS

Q = !

S 2

(9.9)

where the magnetic eld is assumed to be in the form of standing wave H(r)e

2
!

1
f

=p

i!t ;

and

(9.10)

is the skin depth of the wall. Note that in the case of standing wave, the sum of electric and
magnetic energies is constant and equal to either
0

2
or
"0
2

H 2 dV;

E 2 dV:

Time averaged dissipation rate at the wall is

1
2

!
2

(n

H)2 ; (W/m2 );

where the factor 1=2 is for rms (time average) value.

For the lowest order TM mode with l = 1; the azimuthal magnetic eld is in the form
H = Cj1 (kr) sin ;

(9.11)

where C is a constant. Then, for ka = 2:744;

QT M;l=1 =

j0 (ka) j2 (ka)
j12 (ka)

a
' 0:734 ;

(9.12)

where use is made of the integral,

Z

x2 jl2 (x) dx =

x3 2
j (x)
2 l

jl

1 (x) jl+1 (x)

(9.13)

Alternative expression is
QT M;l=1 = 0:734

= 0:32 :

(9.14)

The atmospheric region between the ionosphere and ground earth can be regarded as a thin
4

cavity for electromagnetic waves. The simplest standing wave is the one which is azimuthally
symmetric (@=@ = 0) : The

component of the wave equation

r2 + k 2 B = 0; k = !=c;

(9.15)

yields
1

r2

+ k 2 B = 0;

r2 sin2

(9.16)

provided Br = B = 0: Solution for B (r; ) is

ul (r) 1
Pl (cos ) ;
r

(9.17)

l (l + 1)
r2

ul (r) = 0:

(9.18)

ic2
ul (r)
l (l + 1)
Pl (cos ) ;
!r
r

(9.19)

ic2 dul (r) 1

Pl (cos ) :
!r dr

(9.20)

B (r; ) =
where ul (r) satises

d2 ul
+ k2
dr2

Corresponding electric eld is

Er =

E =

In the narrow region between the ionosphere and ground, we may approximate
l (l + 1)
l (l + 1)
'
;
r2
a2

(9.21)

where a is the earth radius. Then

ul (r) = cos

l (l + 1)
(r
a2

k2

a) ;

(9.22)

and the condition E = 0 at r = a yields a resonance frequency,

!=

c
l (l + 1) :
a

(9.23)

The fundamental resonance (known as Schumann resonance) frequency is approximately 10 Hz.

(Observed frequency is approximately 8 Hz.) The Schumann mode can be excited by lightning
5

discharges and is being continuously monitored as one of the most important electromagnetic
phenomenon of global scale. Note that the resonance frequency is essentially the inverse of the
transit time around the globe,

9.3

= 2 a=c:

Perturbative Change in Cavity Resonance Frequency

For a given eigenmode, the eigenvalue k 2 of a cavity mode can be found from
k02

E)2 dV

V (r

2
R
V

S (n
E2 dV

E) (r

subject to the condition that r E = 0 in the volume and n

E)dS

(9.24)

E = 0 on the wall surface. Note

that the normal vector n is directed away from the cavity volume. The surface integral evidently
vanishes for exact eigenmodes but is retained here for the purpose of variational calculation because
trial functions may not exactly satisfy the boundary conditions. If the cavity surface is only slightly
deformed, the unperturbed electric eld may still be used in the majority of region except at the
deformation. Let us assume a small dent of volume
to order
k

V and
=
=
'
=

V and surface area

S: The new eigenvalue

S is
V (r

E)2 dV

E) (r E)dS
S 0 + S (n
:
2
V
v E dV
R
R
H
E)2 dV
E)2 dV 2 S 0 + S (n E) (r E)dS
V (r
V (r
R
R
2
2
V E dV
V E dV
R
H
R
R
2
E)2 dV
(n E) (r
E)2 dV
V (r
V E dV
V R(r
R
R
R
2 S
1+
2
2
2
2
E
dV
E
dV
E
dV
V
V
V
V E dV
R
R
H
E2 dV
E)2 dV
(n E) (r E)dS
V R(r
R
k02 1 + R V 2
2 S
;
2
2
V E dV
V E dV
V E dV
V

E)dS

(9.25)

where S 0 is the unperturbed surface on which n

E = 0 and

S is the surface of

V facing the

cavity interior. The surface integral can be converted into volume integral as
I

(n

E) (r

E)dS =

=
=
where n0 =

(n0 E) (r

E)dS

n0 [E

(r

E)]dS

r [E

(r

E)]dV
Z
2
2
E) dV + k0
E2 dV

(r
V

n is the normal vector directed away from

2

k =

V [(r

(9.26)

V: Then, the change in the eigenvalue is

E)2 k02 E2 ]dV

:
2
V E dV

(9.27)

Since the electric and magnetic energy are identical in the cavity, we nally nd
2

k '

k02

R
R

V
V

H2 dV
H2 dV

R
R

V
V

E2 dV
E2 dV

(9.28)

where the rst term indicates the fraction of magnetic energy removed and second term the fraction
of electric energy removed from the cavity. Corresponding change in the resonance frequency is
!
!0

1
'
2

R
R

V
V

H2 dV
H2 dV

R
R

V
V

E2 dV
E2 dV

(9.29)

This result indicates that if a dent is made at a position on the inner wall where the magnetic energy
is concentrated, the eigenvalue (and thus the frequency) increases through an eective decrease in
the inductance. If a dent is made where the electric energy is dominant, the resonance frequency
decreases through an eective increase in the capacitance. This is analogous to a perturbation in
inductance and capacitance. If a conductor is inserted into an inductor, the inductance decreases
and if a conductor is inserted in a capacitor, the capacitance increases.
If a small object is placed in the volume of a cavity, it will perturb the electromagnetic elds.
A resultant frequency shift may be analyzed in terms of dipole approximation for a body whose
dimension is su ciently small compared with the wavelength. For example, a conductor sphere of

Figure 9-1: A dent on a cavity wall with volume

which n E0 = 0 with E0 the unperturbed eld.

V and area

S: S 0 is the unperurbed area on

radius a induces an electric dipole moment

p = 4 "0 a3 E(rs );

(9.30)

2 a3 H(rs );

(9.31)

and magnetic dipole moment

m=

where E(rs ) and H(rs ) are the elds at the location of the sphere. The perturbed electric energy
is
Ee =

p E(rs ) =

4 "0 a3 E2 (rs );

(9.32)

and perturbed magnetic energy is

Em =

m B (rs ) = +2

3 2
0 a H (rs ):

(9.33)

Then the magnetic energy removed is

2

3 2
0 a H (rs ):

and the frequency change is given by

!
!0

2"0 E2 (rs ) + 0 H2 (rs ) 3

R
a :
"0 V E2 dV
8

(9.34)

9.4

TE and TM Modes in Conductor Waveguides

In an air-lled waveguide, the electromagnetic elds satisfy the vector wave equations,
r2

1 @2
c2 @t2

E = 0;

(9.35)

r2

1 @2
c2 @t2

H = 0:

(9.36)

Since we are interested in waves propagating along the waveguide extended in the z-direction, the
z and time dependence can be singled out in the form
E(r; t) = E(r? )ei(kz z

!t)

; H(r; t) = H(r? )ei(kz z

!t)

(9.37)

where r? indicates the coordinates transverse to the z-axis. The axial components Ez and Hz both
satisfy scalar wave equation,
r2

1 @2
c2 @t2

Ez = 0;

(9.38)

r2

1 @2
c2 @t2

Hz = 0;

(9.39)

Finding solutions to the axial elds Ez and Hz is su cient to determine electromagnetic elds in
a waveguide, for the Maxwells equations
r

E = i!

0 H;

H=

i!"0 E;

(9.40)

yield transverse components entirely in terms of the axial components as follows,

E? =

i
(!=c)2

kz2

(kz r? Ez + !

H? =

i
(!=c)2

kz2

(kz r? Hz

0 r ? Hz

ez ) ;

(9.41)

!"0 r? Ez

ez ) :

(9.42)

Modes having no axial electric elds, Ez = 0; are called Transverse Electric (TE) modes and modes
having no axial magnetic elds, Hz = 0; are called Transverse Magnetic (TM) modes. As for
radiation elds, these two modes form basic eigenfunctions.

9.5

Rectangular Waveguides

Figure 9-2: Rectangular waveguide.

We assume a conductor tube having a cross-section a

b: The conductor walls are assumed to

be ideally conducting, that is, the skin-depth

==

=p

1
f

(9.43)

is su ciently small and Ohmic dissipation is negligible. Later, we will relax this assumption in
evaluating the damping factor of electromagnetic waves conned in a waveguide. However, for
analyzing mode structure, the assumption of ideal conductor does not introduce signicant errors.
For TE modes, it is su cient to nd the axial magnetic eld Hz (r; t) = Hz (x; y)ei(kz z

!t)

which

obeys the following Helmholtz equation,

@2
!
@2
+
+
@x2 @y 2
c

kz2 Hz (x; y) = 0:

(9.44)

The boundary condition for Hz (x; y) can be found from the vanishing tangential components of the
electric eld at the wall of the waveguide,
r? Hz (x; y) = 0 at x = 0; a and y = 0; b:

10

(9.45)

This condition is satised if we choose

n
m
x cos
y ;
a
b

Hz (x; y) = H0 cos

(9.46)

where m and n are integers. Either m or n can be zero but not simultaneously. Substitution into
Eq. (9.44) yields a dispersion relation,
! 2 = c2

m
a

n
b

+ kz2 ;

(9.47)

which determines the axial wavenumber kz for a given wave frequency ! and given dimensions of
the waveguide. The minimum frequency allowed for wave propagation (kz > 0) in a waveguide
occurs at kz = 0 and is given by
r

! cmn = c

m
a

n
b

(9.48)

This is called the cuto frequency of the TEmn mode. If a > b; the smallest cuto frequency is
that of the TE10 mode (m = 1; n = 0),
! c10 =

c
c
; or fc10 =
(Hz):
a
2a

(9.49)

The TE10 mode is most commonly used in microwave communication.

The phase and group velocities along the wave guide are given by
!
=p
kz
1

c
(! cmn =!)2

p
d!
=c 1
dkz

> c;

(! cmn =!)2 < c:

(9.50)

(9.51)

In the waveguide, waves propagate along a zig-zag path being reected at the walls. The velocity
along the zig-zag path is simply c; for the total wavenumber is
k=

m
a

n
b

+ kz2 :

(9.52)

The phase velocity along the z-axis larger than c is due to a kz which is smaller than k: The group
velocity along the z-axis corresponds to the axial component of c.
11

For TM modes, the axial electric eld Ez (x; y) satisfying the Helmholtz equation
@2
@2
!
+
+
@x2 @y 2
c

kz2 Ez (x; y) = 0;

(9.53)

and the boundary conditions

Ez = 0 at x = 0; a and y = 0; b

(9.54)

is given by
m
n
x sin
y ;
a
b

Ez (x; y) = E0 sin

(9.55)

where m and n are nonzero integers. The cuto frequency of the TM mode is
r

! cmn = c

9.6

m
a

n
b

; m

1; n

1:

(9.56)

TE10 Mode

In this section, some properties of the TE10 mode will be studied. We assume an electric eld in
the form
Ey (x; z; t) = E0 sin

x ei(kz z

!t)

(9.57)

Corresponding magnetic eld can be found from

H =
=

1
r E
i! 0
kz
E0 sin
x ei(kz z
! 0
a

!t)

ex

=a
!

cos

x ei(kz z

!t)

ez :

(9.58)

The magnetic eld prole seen from top is shown below for the case kz = =a: Also shown is the
prole of the surface current on the lower wall,
Js = n

12

H:

(9.59)

1.5

z1

0.5

00 0.2 0.4 0.6 0.8 1

x

Figure 9-3: Magnetic eld prole of the TE10 mode in a rectangular waveguide.

The power transmitted in the waveguide can be found by integrating the Poynting vector in the
axial direction given by
Sz =

Ey Hx =

dx

kz 2 2
E sin
x ;
! 0 0
a
b

dy Sz (x)

(9.61)

Z
Z b
kz 2 a 2
sin
E
x
dy
! 0 0 0
a
0
kz 2 ab
E
; (peak value).
! 0 0 2

(9.60)

(9.62)
(9.63)

The rms power is

Prms =
=
=

kz 2 ab
E
! 0 0 4
r
ab "0 p
1
4
0

ab E02
; (W)
TE
4 Z10

13

(9.64)
(! c10 =!)2 E02

(9.65)
(9.66)

1.5

z1

0.5

00 0.2 0.4 0.6 0.8 1

x

Figure 9-4: Surface current density pattern on the upper inner surface.
where
TE
Z10

=p

is the impedance of the TE10 mode.

0 ="0

(! c10 =!)2

(9.67)

In practice, the nite (not innite) conductivity of the wall causes dissipation of electromagnetic
energy and the power decays as the wave propagates along the waveguide. To evaluate the damping
factor, we recall that the Poynting ux normal to a conductor wall is
r

S? = Z jHt j =
where
Zs =

i!

i!

jHt j2 ;

(9.68)

(9.69)

is the surface impedance of the conductor and Hs is the tangential component of the magnetic eld
at the wall surface. The real part of S? indicates net dissipation,
Re

i!

jHt j =

14

!
2

jHt j2 :

(9.70)

For the TE10 mode, the tangential component of the magnetic eld at the side walls at x = 0 and
a is
Hz (x = 0; a) =

=a
!

E0 ei(kz z

!t)

(9.71)

Therefore, the time averaged loss rate per unit length of the side walls is
r
!
dP1
=2
dz
2

=a

E02
b:
2

(9.72)

Similarly, on the top and bottom walls,

dP2
dz

"
=a
! 0
= 2
2
! 0
r
1 "0 ! 0 2
=
E0 a;
2 0
2
r

and the total loss rate is

"0
dP
=
dz
0

!
2

kz
! 0

E02
a
4
(9.73)

!c
!

a
E02 :
2

(9.74)

The power damping factor is given by

r
1 dP
!"0 a + 2b(! c =!)2
p
=2
P dz
2 ab 1 (! c =!)2

and the eld intensity decays as E0 e

copper waveguide ( ' 6

2 ;

(9.75)

along the waveguide. For example, the damping factor in a

107 S/m) having a cross-section 2

1 cm2 excited at a frequency f = 9

GHz is approximately 2 = 0:04/m. The microwave power e-folds in a distance of 1=2 = 25 m.

Example 1 Rectangular Cavities
Electromagnetic elds conned in a rectangular wave guide closed at z = 0 and L form complete
standing waves in all directions, x:y; z: We assume L > a > b without loss of generality. It is
su cient to consider TM modes because after suitable coordinates changes, TM modes can be
recovered from TE modes. The axial electric eld Ez (r; t) must vanish at x = 0; a and at y = 0; b
and thus may be assumed to be
Ezlmn (r; y) = E0 sin

n
m
x sin
y cos
a
b

15

l
z ;
L

(9.76)

where cos

l
Lz

is chosen so that x and y components of the electric eld vanish at z = 0 and L:

The resonance frequency of (l; m; n) is thus given by

s

! lmn = c

m
a

where m; n are nonzero integers and l = 0; 1; 2;

l
L

(9.77)

The lowest resonance frequency is

! 011 = c

9.7

n
b

(9.78)

Excitation of Rectangular TE Modes

Electromagnetic waves in waveguides can be excited by a current source placed in the guide. To
illustrate the general methodology, we consider a thin vertical current placed at x = a=2 in a
rectangular waveguide. The y component of the wave equation,
r2

1 @2
c2 @t2

E=

1
r +
"0

@J
:
@t

(9.79)

is
r2
where the charge density

1 @2
c2 @t2

Ey =

@Jy
;
@t

(9.80)

has been ignored. Since the transverse electric eld of TE modes can

be generated by the axial magnetic eld Hz as

E? =

0
2
k?

@
r ? Hz
@t

ez ;

(9.81)

we obtain
Ey =

0 @ @Hz
2 @t @x :
k?

(9.82)

Substitution into Eq. (9.82) yields

r2

1 @2
c2 @t2

1 @Hz
2 @x =
k?

Jy =

16

I(t)

a
2

(z):

(9.83)

Let us assume TEm0 modes are excited by the current. Then the axial magnetic eld may be
expanded as
Hz (x; z; t) =

1
X

An (z; t) cos

m=1

m
x ;
a

(9.84)

where Am (z; t) is an expansion function and the eigenfunction for TEmn mode is recalled,
Hzmn (x; y) = cos

n
m
x cos
y :
a
b

(9.85)

Multiplying Eq. (9.83) by

sin
and integrating over the cross section area a
1 @2
c2 @t2

r2z

m0
x ;
a
b; we nd

2
km0
Am0 (z; t) =

2 m
m
I(t) (z) sin
2
a
2

(9.86)

2 = (m =a)2 :
with km0

If the current is an impulse, I(t) = q0 (t) where q0 is the amount of charge transferred, the
Laplace transform of Am0 (z; t) obeys
d2
dz 2

s2
c2

2
km0
Am0 (z; s) =

2 m
m
q0 (z) sin
2
a
2

(9.87)

which can be solved as

Am0 (z; s) =

mq0
1
p
exp
2
2
a
(s=c)2 + km0

q
2 jzj sin m
(s=c)2 + km0
2

(9.88)

Inverse Laplace transformation yields

8
p
m
< cq0 J0 m
(ct)2 z 2 sin
a
a
2
Am0 (z; t) =
:
0;

; ct > jzj

(9.89)

ct < jzj

Note that the pulse front propagates at the speed c: Of course, the pulse shape is severely distorted
as it propagates over a large distance because of the dispersion of the electromagnetic modes
conned in a waveguide.
For a steady oscillating current I(t) = I0 e

i!t ;

17

we may assume Am0 (z; t) = A(z)e

i!t ;

where

Am0 (z) satises

!2
d2
+
dz 2
c2

2
km0
Am0 (z) =

2 m
m
I0 (z) sin
2
a
2

(9.90)

Noting
d2 ikjzj
e
= 2ik (z)
dz 2

k 2 eikjzj ;

we nd the following steady state solution,

Am0 (z) =

9.8

mI0
p
2
ia (!=c)2

sin

2
km0

m
2

p
2
ei (!=c)

2 jzj
km0

(9.91)

Circular Waveguide

Figure 9-5: Circular waveguide. The lower gure shows qualitatively the electric and magnetic eld
proles.

Let us now consider electromagnetic waves conned in a conductor tube of radius a: The
Helmholtz equation for the axial electric and magnetic elds is
@2
@

@2

1 @
1
+ 2
@
@

!
c

18

kz2 @

Ez
Hz

A = 0:

(9.92)

General solutions bounded at the axis

= 0 are

Ez ( ; ); Hz ( ; ) = Jm (k )eim ;

(9.93)

where m is an integer and

!
c

k2 =

kz2 > 0:

(9.94)

(The case k 2 < 0 will be considered later for slow wave circular waveguides.) For TE modes, the
boundary condition H ( = a) = 0 requires
d
Jm (k ) = 0 at
d

= a;

(9.95)

0 (x) = 0 relevant
and for TM modes, Hz ( = a) = 0 requires Jm (ka) = 0: In the table, roots of Jm

to TE modes and roots of Jm (x) = 0 relevant to TM modes are shown. [Note the special case
J00 (x) =

J1 (x):]
Roots of

0 (x) = 0
Jm

m=

1st

3.832

1.841

2nd

7.016

3rd

10.173

Roots of

Jm (x) = 0

m=

3.053

1st

2.405

3.832

5.136

5.331

6.706

2nd

5.520

7.016

8.417

8.536

9.969

3rd

8.654

10.173

11.620

The dispersion relation of the TE modes is

r

ka = a
or

! 2 = c2

"

!
c

x0mn
a

kz2 = x0mn ;

+ kz2 ;

(9.96)

(9.97)

and the cuto frequency is

!c =

c 0
x :
a mn

(9.98)

The smallest root is the rst root of J10 (x) = 0; x11 ' 1:841: This is the fundamental mode in a
circular waveguide and corresponds to the TE10 mode in a rectangular waveguide. Through gradual
tapering, the rectangular TE10 mode can be converted to circular TE11 mode.
19

Example 2 Cylindrical Cavity

Consider TE modes in a circular waveguide closed at z = 0; L:The axial magnetic eld may be
assumed to be
l
z ;
L

Hz (r) = H0 Jm (k )eim sin

where
k=

!
c

kz2 :

From
i

E? =

(!=c)2

i!
(!=c)2

kz2

(kz r? Ez + !

kz2

r ? Hz

0 r ? Hz

ez )

ez ;

we nd the transverse electric eld,

E? =

i!
(!=c)2

im

kz2

@Hz
e
@

Hz e

From the condition that E ( = a) = 0; it follows that

kmnl =

! mnl
c

l
L

x0mn
;
a

0 (x) = 0: The resonance frequency is

where x0mn is the n-th root of Jm

! mnl = c

x0mn
a

l
L

The transverse magnetic eld is

H? =
=

ikz
r ? Hz
(!=c)2 kz2
ikz
@Hz
im
e +
Hz e
(!=c)2 kz2
@

The TE0n1 mode is commonly used in practical applications because the surface current on the end
walls is purely azimuthal (J only as can be seen from Js = n

20

H) and thus a thin circular gap

does not signicantly perturb the elds (and resonance frequency). The resonance frequency can
be controlled by varying the length L and a small but nite gap cannot be avoided in moveable
plungers. The TE0n1 is evidently independent of

and thus degeneracy problems between cos m

and sin m modes can also be avoided.

9.9

TM Mode in Circular Slow Waveguide

Waveguides used in linear electron accelerators must accommodate TM modes having a phase
velocity close to c. Waveguides with smooth inner walls can only accommodate modes having
phase velocities larger than c and thus cannot be used for this purpose. Modes must be TM
because TE modes have no electric eld in the axial direction needed to accelerate electrons.

Figure 9-6: Circular slow waveguide of radius a with diaphragms of radius b. The spacing between
diaphragms is much smaller than the axial wavelength z :

Slow waveguides have conductor diaphragms placed periodically along the axis as shown in
Fig.9-7. The purpose of the diaphragms is to increase the capacitance per unit length of the
waveguide which contributes to slowing down the phase velocity of electromagnetic waves. We
consider modes symmetric about the axis, m = 0: The waveguide has a radius a and diaphragms
have holes of radius b: In the region

< b; the Helmholtz equation for the axial electric eld is

1 @
!
@2
+
+
@ 2
@
c

kz2 Ez ( ) = 0;

< b:

(9.99)

Since we are interested in modes having an axial phase velocity slightly smaller than c; that is,
!
. c;
kz

21

(9.100)

the quantity (!=c)2

kz2 must be negative, and thus solution for Ez ( ) may be assumed to be

Ez ( ) = AI0 (k );

<b

(9.101)

where
!
c

k 2 = kz2
In the diaphragm region b <

> 0:

(9.102)

< a; the electric elds lines are essentially straight provided the

axial period of the diaphragms is su ciently smaller than the axial wavelength. We assume that
this condition is met. Then, the wave equation in the region b <
@2
1 @
!
+
+
@ 2
@
c

Ez ( ) = 0; b <

< a may be approximated by

< a:

(9.103)

< a:

(9.104)

General solutions are

Ez ( ) = BJ0

!
c

+ CN0

!
c

; b<

The boundary conditions are:

Ez ( = a) = 0;

(9.105)

and
Ez and H be continuous at

= b:

(9.106)

These boundary conditions yield

BJ0

!
!
a + CN0
a = 0;
c
c

AI0 (kb) = BJ0

!
!
b + CN0
b ;
c
c

A
ch
!
! i
I1 (kb) =
BJ1
b + CN1
b ;
k
!
c
c
where J00 (x) =

J1 (x); N00 (x) =

(9.107)
(9.108)
(9.109)

N1 (x); I00 (x) = I1 (x) are noted. Eqs. (9.107) through (9.109)

give the following dispersion relation

!
!
J0
a N0
b
ck I0 (kb)
c
c
=
!
!
! I1 (kb)
J1
b N0
a
c
c

22

J0
J0

!
!
b N0
a
c
c :
!
!
a N1
b
c
c

(9.110)

For the purpose of accelerating highly relativistic electrons, the axial phase velocity !=kz must be
close to c or k ' 0: Then I0 (kb) ' 1; I1 (kb) ' kb=2; and the LHS of Eq. (9.110) reduces to 2c=!b:
For a given rf frequency ! and the size of the waveguide a; the dispersion relation
!
!
J0
a N0
b
2c
c
c
'
!
!
!b
J1
b N0
a
c
c

J0
J0

!
!
b N0
a
c
c ;
!
!
a N1
b
c
c

(9.111)

can be solved numerically to determine the aspect ratio a=b of a slow wave circular waveguide.
Fig.9-7 shows the function

2a
f (x) =
xb

J0 (x) N0
J1

b
x
a

b
x N0 (x)
!a
a
; x=
;
b
c
J0 (x) N1
x
a
J0

b
x N0 (x)
a

(9.112)

when a=b = 2:5: The rst root occurs at x ' 3:89 and for a given rf frequency !; the outer radius
a can thus be determined.

1
0.5

0 3.5

3.6

3.7

3.8

3.9

-0.5
-1

Figure 9-7: Root of f (x) = 0 when a=b = 2:5:

9.10

Dielectric Waveguides

An optical ber can conne light waves because of total reection at the surface. In contrast to
conductor waveguides, light waves in optical waveguides cannot be pure TE or TM modes. This
is because electromagnetic elds outside, as well as inside, the optical ber must be considered
23

simultaneously. Although the outer elds are evanescent (otherwise waves cannot be conned), the
elds near the surface do aect those inside.
We rst consider a simple case of step change in the index of refraction,
8
< n;
n( ) =
: 1;

< a;
> a:

Such an optical ber is of no practical interest, for bers used in optical communication all have
graded index of refraction with a gradual change with the radius : The axial electric eld Ez (r; t)
satises the wave equations in both regions,
@2
1 @
1 @2
@2
+
+
+
+
2
@ 2
@
@ 2 @z 2

0 "!

@2
1 @
1 @2
@2
+
+
+
+
2
@ 2
@
@ 2 @z 2

0 "0 !

Ez< (r) = 0;

< a;

(9.113)

Ez> (r) = 0;

> a:

(9.114)

The azimuthal dependence may be assumed to be eim and the axial dependence eikz z ;
E(r) = E( )eim

+ikz z

Then,
d2
1 d
+
2
d
d

m2

d2
1 d
+
d 2
d

m2

kz2 Ez< ( ) = 0;

< a;

(9.115)

kz2 Ez> ( ) = 0;

> a;

(9.116)

0 "!

0 "0 !

which admit the following bounded solutions,

Here
k1 =

Ez< ( ) = AJm (k1 );

< a;

(9.117)

Ez> ( ) = BKm (k2 );

> a:

(9.118)

0 "!

kz2 ; k2 =

kz2

0 "0 !

2:

Note that the outer eld should be evanescent for the waveguide to conne light waves.

24

(9.119)

Similarly, the axial magnetic eld Hz (r) may be assumed to be

Hz< ( ) = CJm (k1 );

< a;

(9.120)

Hz> ( ) = DKm (k2 );

> a:

(9.121)

The transverse elds E? and H? can then be calculated by referring to Eqs. (9.41) and (9.42).
The azimuthal components of the elds are
E<( ) =

i
k12

kz

E>( ) =

i
k22

kz

H <( ) =

i
k12

im

im

kz

AJm (k1 )

BJm (k2 )

im

0
0 k1 CJm (k1

) ;

(9.122)

0
0 k2 DKm (k2

) ;

(9.123)

0
CJm (k1 ) + !"k1 AJm
(k1 ) ;

(9.124)

0
DKm (k2 ) + !"0 k2 BKm
(k1 ) :

(9.125)

and
H >( ) =

i
k12

kz

im

The continuity of Ez ; E ; Hz and H yields

1
k12
1
k12

kz
kz

im
AJm (k1 a)
a

AJm (k1 a) = BKm (k2 a);

(9.126)

CJm (k1 a) = DKm (k2 a);

(9.127)

0
0 k1 CJm (k1 a)

im
0
CJm (k1 a) + !"k1 AJm
(k1 a)
a

1
k22

kz

1
k22

kz

im
BKm (k2 a)
a

0
0 k2 DKm (k2 a)

; (9.128)

im
0
DKm (k2 a) + !"0 k2 BKm
(k1 a) : (9.129)
a

Then the determinantal dispersion relation is

1
1
+ 2
2
k1
k2

mkz
a

!
k1 c1

=
+

!2
k1 k2

0 (k a)
Jm
1
Jm (k1 a)

1
1
+ 2
2
c1 c2

!
k2 c2

0 (k a)
Km
2
Km (k2 a)

0 (k a)K 0 (k a)
Jm
1
m 2
:
Jm (k1 a)Km (k2 a)

(9.130)

Figure 9-8 shows the dispersion relation, namely, axial wavenumber kz normalized by k0 = !=c
as a function of the normalized frequency k0 a = !a=c when m = 1 and n = 1:1: It can be seen that
25

a cuto occurs at !a=c ' 1:386: Fig.9-9 shows the case when m = 2: The cuto frequency increases
to !a=c ' 5:43:

1.1
1.08
1.06
1.04
1.02
10

10

15
x

20

25

30

Figure 9-8: ckz =! vs. !a=c when n = 1:1; m = 1: The cuto frequency is ! c a=c ' 1:386:

1.1
1.08
1.06
1.04
1.02
10

10

15
x

20

25

30

Figure 9-9: ckz =! vs. !a=c when n = 1:1; m = 2: The cuto frequency is ! c a=c ' 5:43:

26

9.11

Graded Index Fibers

In optical bers used in practical communication, the index of refraction is designed to have gradual,
rather than step, variation with the radius. Quadratic variation is commonly employed,
2 2

n( ) = n0 1

= constant

(9.131)

because the electromagnetic elds are then well conned with a Gaussian prole e

a2

. The

corresponding permittivity is
"( ) = "0 n20 1

2 2 2

(9.132)

Since
r ["(r)E] = "(r)r E + E r" = 0;

(9.133)

E = !2

(9.134)

the wave equation

r

0 "(r)E;

reduces to
r2 E + ! 2

0 "(r)E

+r

E r"
"

= 0:

(9.135)

If the change in the permittivity is small, the last term can be ignored in the lowest order approximation, and we obtain a simple wave equation with an inhomogeneous permittivity,
r2 E + ! 2

0 "(r)E

' 0:

(9.136)

In the cylindrical geometry, a cartesian component of the transverse electric eld satises
@2
1 @
1 @2
@2
+
+
+
+ !2
2
2
2
@ 2
@
@z
@

0 "(

) Ei = 0:

For weak variation of "( );

"( ) = "0 n20 (1

2 2 2

) ' "0 n20 1

27

2 2

(9.137)

and axially symmetric mode @=@ = 0; Eq. (9.137) reduces to

d2
1 d
+
d 2
d

kz2 + k02 1

2 2

k02 = ! 2 "0

2
0 n0 :

E( ) = E0 e

a2

E( ) = 0;

(9.138)

where
(9.139)

Assuming
2

we nd
1
a2 = p k0 ;
2
4
p k0 :
2

kz2 = k02

(9.140)
(9.141)

The electric eld is conned with a Gaussian prole in the radial direction. The e-folding radial
distance is

p
4
1
2
;
w= =p
a
k0

(9.142)

which is called beam radius. A constant beam radius is maintained only for appropriate injection
of light wave at the input end. If not, the beam radius varies with the axial distance accompanied
by periodic focusing and defocusing as intuitively expected from the picture of repeated total
reections.

28

Problems
1. A waveguide has a semicircular cross section of radius a: Determine the lowest cuto frequency.
2. Prove Eqs. (??) and (??).
3. Each corner of a rectangular waveguide having a cross section a
with

b is smoothed as shown

a; b: Find the change in the cuto frequency of the TE10 mode.

4. The cuto frequency of a ridge waveguide shown may be estimated from equivalent capacitance and inductance,
C
d
1
' "0 (F/m); L l =
ab (H m)
l
g
2 0
r
2g
1
:
=c
!c = p
abd
LC
Explain.
5. Thin diaphragms along the wide walls of a rectangular waveguide act as an eective capacitance, while diaphragms along the narrow walls act as an eective inductance. Give
qualitative explanations.
6. In the magic T shown, TE10 mode entering port 3 (or 4) is equally divided between ports
1 and 2 but does not come out of port 4 (or 3). Explain in terms of cuto of higher order
modes.
7. Two waveguides are coupled through two small holes drilled in the wider walls separated by
a quarter wavelength in each direction as shown. TE10 mode entering port 1 comes out of
ports 2 and 4, but not port 3. Explain.
8. Show that the dispersion relation of TM modes in a coaxial cable having inner and outer radii
a and b is given by
Jm (ka)Nn (kb)

Jm (kb)Nm (ka) = 0;

where
k2 =

!
c

kz2 :

When a = 1 mm, b = 4 mm, what are cuto frequencies? Consider m = 0 and 1:

29

9. Repeat the preceding problem for TE modes to show that the dispersion relation in this case
is
0
(ka)Nn0 (kb)
Jm

0
0
Jm
(kb)Nm
(ka) = 0:

10. Wave propagation along a conducting helix with a pitch angle

is expected to be characterized

by the axial propagation speed,

!
' c sin ;
kz
because electromagnetic waves are guided along the spiral path of the helix. Show that an
approximate dispersion relation is given by
!
c

cot2

= k2

I0 (ka)K0 (ka)
;
I1 (ka)K1 (ka)

where
k 2 = kz2

30

!
c

Chapter 10

Electromagnetism and Relativity

10.1

Introduction

If I am moving at a large velocity along a light wave, what propagation velocity should I measure?
This was a question young Einstein asked himself and in 1905, he published a monumental paper
on special relativity which formulated how to transform coordinates, velocity and electromagnetic
elds between two inertial frames. Einstein postulated that:
1. all physical laws remain intact in any inertial frames, and
2. the light velocity c is invariant against inertial coordinate transformation.
Postulate 1 means that, for example, the Maxwells equations in a moving frame remain formally
identical to those in the laboratory frame, provided the spatial coordinates, time, and electromagnetic elds are appropriately transformed. Electromagnetic elds appear and disappear as we
change observing frame. For example, if a charge is moving in the laboratory frame, we observe
both electric and magnetic elds. In the frame of the moving charge, the current and magnetic eld
evidently disappear. However, the electromagnetic elds (primed quantities) in the charge frame
are subject to the transformation
E0k = Ek ;

E0? =

(E? + V

B? ) ;

(10.1)

B0k = Bk ;

B0? =

(B?

E? ) ;

(10.2)

where k and ? are relative to the direction of the velocity V: Since in this example, Bk = 0
and B? = V E? in the laboratory frame, the magmatic eld in the frame of the moving charge
indeed vanishes consistent with our intuition. The pertinent static Maxwells equations are satised
in both frames
(10.3)
laboratory frame: r E = ; r B = 0 J
"0
in the frame of moving charge: r0 E0 =
1

"0

; r0

B0 = 0; (J0 = 0):

(10.4)

Here the primed operators and quantities are those in the moving frame which are subject to the
Lorentz transformation.
As shown in Chapter 8, electromagnetic elds due to a charged particle moving at an arbitrary
velocity can be correctly formulated by the Lienard-Wiechert potentials which had been discovered
prior to the theory of relativity. Electromagnetic disturbances propagate at the velocity c regardless
of the velocity of the charge, just as sound waves emitted by a moving source propagate at a
sound velocity independent of the source velocity. A major dierence between sound waves and
electromagnetic waves occurs for stationary source and moving observer. For sound waves, if an
observer is approaching a source at a velocity VO ; the apparent sound velocity becomes cs + VO
because both cs and VO are well dened with respect to the medium of sound waves, namely, air. In
electromagnetic waves that can propagate in vacuum, there is no preferred inertial frame to dene
velocities and a moving observer will measure the same propagation velocity c regardless of the
relative velocity between two inertial frames. Of course, the frequency and wavelength are Doppler
shifted but the product
= 0 0 = c or the ratio !=k = ! 0 =k 0 = c remains invariant.

10.2

CGS-ESU System

In this Chapter, the CGS-ESU (Electro-Static Unit) unit system is used so that electromagnetic
elds E (statvolt/cm ' 300 volt/cm = 3 104 volt/m) and B (gauss = 10 4 T) have the same
dimensions. In CGS-ESU, the Coulombs law is adopted to connect the mechanical world and
electromagnetic world. (Recall that in SI, the magnetic force is adopted to dene 1 ampere current
which in CGS-USU is 3 109 stat-ampere.) If two equal charges separated by 1 cm exert a force
of 1 dyne (= erg/cm = 10 7 J/10 2 m = 10 5 N) on each other, the charge is dened to be 1 ESU
' 3 109 C. The Coulombs law in CGS-ESU system is
Coulombs law: F =

q1 q2
(dyne).
r2

(10.5)

The electronic charge is e = 4:8 10 10 ESU (= 1:6 10 19 C). The potentials

and A also
have common dimensions (statvolt) in CGS-ESU. This is particularly convenient in theoretical
electrodynamics because the elds E (polar vector) and B (axial or pseudo vector) are in fact
components of a unied 4 4 eld tensor and the potentials , A form a four vector ( ; A):The
Maxwells equations in this unit system are:
r E=4
r B = 0; r

; r

E=

B=

1 @B
;
c @t

4
1 @E
J+
;
c
c @t

(10.6)
(10.7)

and the relationships between the elds and potentials are

E=

1 @A
; B=r
c @t
2

A:

(10.8)

The wave equations for the potentials are modied as

1 @2
=
c2 @t2
1 @2
A =
c2 @t2

r2
r2

(10.9)

4
J;
c

(10.10)

subject to the Lorentz gauge,

r A+

1@
= 0:
c @t

(10.11)

Electromagnetic force in the CGS-ESU system is

1
F (dyne) = e E + v
c
the Poynting ux is
S=

1
B ; f (dyne/cm3 ) = E + J
c

c
E
4

B; (erg cm

1
E
4

B; (dyne cm

B; (dyne sec cm

the ux of momentum is

sec

);

);

B;

(10.12)

(10.13)

(10.14)

and the momentum density is

1
4 c

):

(10.15)

):

(10.16)

Electromagnetic energy density is

1
8

E 2 + B 2 ; (erg cm

The vacuum impedance for a plane wave is unity (dimensionless),

B=

c
k
!

E; jBj = jEj :

(10.17)

In CGS-ESU system, macroscopic proportional constants inevitably have unfamiliar units. For
example, the capacitance has dimensions of length (cm) as seen from its denition,
[C] =

[q]
= length.
[ ]

(10.18)

The conductivity relates the electric eld and current density, J = E; and has dimensions of
frequency, sec 1 ; since
[J]
[q] cm 2 sec 1
[ ]=
=
= sec 1 :
[E]
[q] cm 2

10.3

Lorentz Transformation

The null result of Michelson-Morleys extensive interference experiments to detect the ether velocity
was explained by Lorentz who assumed that a moving object contracts in the direction of its velocity
by a factor ;
q
1
0
2
L = L0 = 1
L0 :
(10.19)

This was followed by the nding by Lorentz and Poincar that if the spatial coordinates, time and
electromagnetic elds are all transformed according to what is known as Lorentz transformation,
the Maxwells equations remain intact. If a relative velocity V is assumed in the x direction,
the laboratory coordinates (ct; x; y; z) and coordinates in the moving frame (ct0 ; x0 ; y 0 ; z 0 ) may be
assumed to be related through a linear transformation,
t0 =
0

x
y

0 (t

aV x)

1 (x

V t)

= y; z = z;

provided that the two coordinate systems coincide at t = 0: The invariance of the coordinates
perpendicular to the relative velocity, y = y 0 and z = z 0 ; follows from isotropy of space which is
implicitly assumed. For light pulse emitted at t = t0 = 0 and x = x0 = 0 in the positive x direction
(same direction as V ),
x0 = ct0 ; x = ct;
which yield
c=

1
01

V
;
acV

(10.20)

while for light pulse emitted in the negative x direction

x0 =

ct; x =

or

ct;

c+V
0 1 + acV
1

c=

(10.21)

From Eqs. (10.20) and (10.21), we nd

0

and a =

1
:
c2

(10.22)

To determine ; consider a light pulse emitted along the y 0 axis (x0 = 0; that is, x = V t) in the
moving frame,
1
V2
V
x
=
c
1
t:
y 0 = ct0 = c t
c2
c2

In the laboratory frame, light propagation is tilted due to the relative motion between the two
coordinates, (ct)2 = y 2 + (V t)2 or
p
y = c2 V 2 t:
Since y 0 = y; we nd

=r

1
V2
c2

(10.23)

Desired transformation between (x; t) and (x0 ; t0 ) is

x0 =

(x

V t);
V
t
x ;
c2

t0 =

and the four dimensional coordinates in the laboratory frame (ct; x; y; z) and those in the moving
frame (ct0 ; x0 ; y 0 ; z 0 ) are related through
2
6
6
6
6
4

ct0
x0
y0
z0

7 6
7 6
7=6
7 6
5 4

V
c

0
0
1
0

V
c

0
0

0
0

0
0
0
1

32
76
76
76
76
54

Its inverse transformation can be found by replacing V with

2
6
6
6
6
4

ct
x
y
z

7 6 V
7 6 c
7=6
7 6 0
5 4
0

V
c

0
0

0
0
1
0

0
0
0
1

ct
x
y
z

7
7
7
7
5

(10.24)

V;

32
76
76
76
76
54

ct0
x0
y0
z0

3
7
7
7
7
5

(10.25)

Graphically, Lorentz transformation may be visualized as contraction in the (ct; x) plane with a
quasi angle dened by
cosh = ; sinh = ; tanh = ;
as illustrated in Fig. (10-1). Note that the coordinates (ct0 ; x0 ) are not orthonormal if those in the
laboratory frame are so chosen. Since
tanh(

2)

tanh 1 + tanh
1 + tanh 1 tanh

2
2

1+

1 2

sum of two velocities cannot exceed c:

Time dilation and length contraction can be visualized as follows. A clock stationary at x = 0
in the laboratory frame moves along the vertical ct axis (x = 0) as time elapses. 1 second in the
laboratory frame appears as second in the moving frame. If seen from the moving frame, the
clock is moving and a moving clock ticks slower. If a clock is stationanry at x0 = 0 in the moving

Figure 10-1: Graphical representation of Lorentz transformation for the case = 0:5: Hyperbolic
curves show (ct)2 x2 = 1 (interval of time-like events) and (ct)2 x2 = 1 (length of space-like
object).

frame, it travelsalong the t0 axis (x0 = 0). In the laboratory frame, 1 second in the moving frame
appears as second. In both cases, a moving clock ticks slower.
Likewise, a stick one meter long in the moving frame is contarcted by a factor if seen from
the laboratory frame. Note that length measurements should be done for a common time in both
frames.
It is convenient to introduce a metric tensor dened by
2

6
6
gij = g = 6
6
4
ij

1
0
0
0

0
1
0
0

0
0
1
0

0
0
0
1

3
7
7
7
7
5

(10.26)

A contravariant vector (vector in ordinary sense) xi = (x0 ; x1 ; x2 ; x3 ) = (ct; x; y; z) can be converted

to a covariant vector through
xi = gij xj = (x0 ; x1 ; x2 ; x3 )

(10.27)

s2 = xi xi :

(10.28)

so that

Note that we follow Einsteins convention, that is, repeated subscripts and superscripts mean
summation is to be taken,
3
X
gij xj =
gij xj ;
(10.29)
j=0

xi xi = x0 x0 + x1 x1 + x2 x3 + x4 x4 :

(10.30)

s2 can be either positive or negative. The Lorentz transformation for the coordinates
2
6
6
6
6
4

ct0
x0
y0
z0

7 6
7 6
7=6
7 6
5 4

0
0

0
0
1
0

0
0

0
0
0
1

32
76
76
76
76
54

ct
x
y
z

3
7
7
7
7
5

can be written in the form

x0i = Lij xj ;

(10.31)

where Lij is the Lorentz transformation mixed tensor,

Lij

6
6
=6
6
4

0
0

Its inverse tensor is

Lij

0
0

6
6
=6
6 0
4
0

0
0

0
0
1
0

0
0
0
1

0
0
1
0

0
0
0
1

(10.32)

(10.33)

7
7
7
7
5
7
7
7
7
5

A four dimensional vector that transforms according to the Lorentz transformation law is called
a four-vector. The position vector xi evidently forms a four vector. For an object moving at a
velocity v in the laboratory frame, the velocity four vector is
vi =

dxi
=
d

where
d =

dt =

(c; vx vy ; vz ) ;

(10.34)

(10.35)

dt;

is the proper time as measured in the moving frame of the object. This is the celebrated time
dilation eect. The magnitude of the four velocity is constant,
v i vi = c2 :

(10.36)

The momentum four vector is dened by

E
;p
c

pi =

E
; px ; py ; pz ;
c

(10.37)

where
E = mc2 ;

(10.38)

is the energy and p is the momentum of mass of a particle having a mass m;

p = mv:

(10.39)

The magnitude of this four vector is also invariant,

i

p pi =

E
c

p2 =

(1

)(mc)2 = (mc)2 :

(10.40)

Example 1 How are the velocity and acceleration transformed?

For a relative velocity V between two inertial frames, spatial coordinates are transformed as
r0k = (rk

Vt); r0? = r? ;

(10.41)

where k and ? indicate components parallel and perpendicular to the velocity V: Since the time is
transformed as
v V
dt;
(10.42)
dt0 =
1
c2
the velocity in the direction of the relative velocity is transformed as
vk0
and the normal component as
0
v?
=

dr0k
dt0

dr0?
=
dt0

vk V

vV
c2

v?

vV
c2

(10.43)

(10.44)

For example, if two particles are approaching each with a velocity v relative to the laboratory frame,
the relative velocity in the frame of either particle is
2v
;
1 + (v=c)2
which cannot exceed c:Note that 2v itself can exceed c:However, a relative velocity between two
objects is meaningful only if it is measured in rest frame of either object. 2v pertains to an
observer in the laboratory frame in which both objects are moving. Therefore, 2v itself is not a
very meaningful velocity. As an example, consider head on collision of two protons. We assume
each proton has a velocity v = 0:9c relative to the laboratory frame. In the rest frame of either
8

proton, the two protons approach with a relative velocity

V =

2v
= 0:9945c;
1 + (v=c)2

and the kinetic energy available for nuclear interaction is

1)mc2 = 8 GeV.

To nd transformation of acceleration, we note the acceleration parallel to the relative velocity

is transformed as
dvk0
ak
0
ak = 0 =
;
(10.45)
vV 3
3 1
dt
2
c

and the normal component as

a0? =

0
dv?
=
dt0

a?

vV 2
c2

v?

vV 3
c2

a V
:
c2

(10.46)

Note that transformation of the normal acceleration involves parallel acceleration as well.
The inverse transformation for the acceleration is
a0k

dvk
ak =
=
dt
a? =

a0?

1+

v0 V 2
c2

1+

(10.47)

v0 V 3
c2

0
v?

1+

a0 V
3 c2 :

v0 V
c2

(10.48)

In an instantaneously rest frame of a particle, v0 = V; and v? = 0: Then

a0k =

a0? =
The current density J and charge density

3 ak

V2
c2
2

ak ;

a? :

(10.49)

(10.50)

form the following four vector

J i = (c ; J) = (c ; Jx ; Jy ; Jz );
where

n0

(v) = en(v) = e q

J = en(v)v;

v 2
c

(10.51)

(10.52)
(10.53)

n0

n(v) = q
1

v 2
c

(10.54)

is the density of charged particles corrected for length contraction in the direction of the velocity v
and n0 is the charge density in the rest frame, v = 0: The magnitude of the current four vector is
J i Ji = (c 0 )2 = const.

(10.55)

where 0 is the proper charge density in the rest frame of the charge.
In Lorentz gauge, the potentials satisfy the decoupled wave equations,
r2

1 @2
c2 @t2

r2

1 @2
c2 @t2

A=

(10.56)

4
J:
c

(10.57)

Therefore, a resultant four vector potential is

Ai = ( ; A) = ( ; Ax ; Ay ; Az );
which satises the wave equation
1 @2
c2 @t2

r2 Ai =

4
Ji :
c

Noting
1 @2
c2 @t2

r2 =

@ @
= g ii @i @i = @ i @i ;
@xi @xi

the wave equation can readily be Lorentz transformed as

@ 0j @j0 A0i =

4 0
J
c i

since @ i @i is Lorentz invariant. The electromagnetic wave equation is thus Lorentz invariant which
guarantees the constancy of the wave propagation velocity c.

10.4

Transformation of Electromagnetic Fields

The electric and magnetic elds, E and B; do not form four vectors. This is due to dierent
vectorial nature of the respective elds. The electric eld is a polar vector (or true vector) because
it changes the sign if coordinates are reversed, r ! r: In contrast, the magnetic eld
1
B(r) =
c

(r

r0 ) J(r0 ) 0
dV ;
jr r0 j3

10

remains unchanged against coordinate inversion since both r r0 and J(r) change sign. The magnetic eld is an axial vector (or pseudo vector). Rather they are components of an antisymmetric
pseudo tensor,
2
3
0
Bz By
6
7
B ij = 4 Bz
0
Bx 5
By Bx
0

The conventional magnetic Lorentz force

1
fm = J
c
now takes a form

1
i
fm
= Jj B ij ; (i; j = 1; 2; 3)
c

where
Jj = ( Jx ; Jy ; Jz )
is the covariant current density. Combining the electric eld and magnetic eld into a single eld
tensor
2
3
0
Ex
Ey
Ez
6
7
6 Ex
0
Bz By 7
ij
6
7;
F =6
0
Bx 7
4 Ey Bz
5
Ez
By Bx
0

the conventional electromagnetic force

1
f = E+ J
c
can be rewritten as

1
f i = F ij Jj ; i = 1; 2; 3; j = 0; 1; 2; 3:
c

The component f 0
1
f0 = E J
c
indicates the work done by the electromagnetic eld. A resultant force four vector is
fj =

1
E J; f
c

Using the eld tensor in Eq. (), we now reformulate Maxwells equations as follows. Since
E=

1 @A
=
c @t

rA0

1 @A
; B=r
c @t

the eld tensor can be written as

F ij = @ i Aj

11

@ j Ai

A;

where @ i is the contravariant derivative

@i =

1@
;
c @t

@
=
@xi

@
;
@x

@
;
@y

@
@z

For example,
F ii = @ i Ai

@ i Ai = 0; (no summation here)

F 01 = @ 0 A1
F 12 = @ 1 A2
and so on. Dierentiating F ij = @ i Aj

1 @Ax @
+
= Ex ;
c @t
@x
@Ay
@Ax
+
= Bz ;
@ 2 A1 =
@x
@y
@ 1 A0 =

@ j Ai with respect to xi covariantly, we obtain

@i F ij = @i @ i Aj

@i @ j Ai = @i @ i Aj

@ j @i Ai :

The rst term in the RHS is

@i @ i Aj =

1 @2
c2 @t2

r 2 Aj =

4 j
J ;
c

while the second term vanishes because of our choice of Lorentz gauge,
@i Ai =

1@
+ r A = 0:
c @t

Thus
@i F ij =

4 j
J :
c

For j = 0; noting J 0 = c ; we recover Gausslaw,

@i F ij = r E = 4

For j = 1; 2; 3; we also recover j-th component of generalized Amperes law,

r

B=

4
1 @E
J+
:
c
c @t

Another identity satised by F ij is

@ i F jk + @ j F ki + @ k F ij = 0;
as can be readily checked by substituting F ij = @ i Aj @ j Ai : When i = 0; j = 1; k = 2; Eq. ()
yields
@ 0 F 12 + @ 1 F 20 + @ 2 F 01 = 0;

12

or

1@
( Bz )
c @t

@
Ey
@x

@
( Ex ) = 0;
@y

E)z =

1 @Bz
:
c @t

that is, we recover Faradays law

(r

Furthermore, for i = 1; j = 2; k = 3; we recover

r B = 0:
In order to nd how the eld tensor F ij is Lorentz transformed, let us consider an arbitrary
contravariant vector B j dened by
B j = F ij Ai :
After Lorentz transformation, this becomes
B 0j = F 0ij A0i ;
where
B 0j = Ljk B k ;

A0i = Lik Ak :

Then,
Ljk B k = F 0ij Lim Am
Ljk F lk Al = F 0ij Lil Al
Since Al is arbitrary, we obtain
F 0ij Lim = Ljn F mn :
Multiplying both sides by Lrm and noting
Lim Lrm =

r
i;

we nd
F 0ij = Lim Ljn F mn = Lim F mn Lnj :
For example,
F 001 = L0m L1n F mn
= L00 L11 F 01 + L01 L10 F 10
=
=

Ex +

2 2

Ex = F 01 ;

13

Ex

that is, the electric eld parallel to the relative velocity V is invariant. For F 002 ; we nd
F 002 = L00 L22 F 02 + L01 L22 F 12
=
and so on. The overall result is
2
0
6
6
Ex
F 0ij = 6
6 (E
Bz )
y
4
(Ez + By )

(Ey

Ex
0

Bz );

(Ey
(Bz

(Bz
Ey )
(By + Ez )

Bz )
Ey )

0
Bx

(Ez + By )
(By + Ez )
Bx
0

3
7
7
7
7
5

For a relative velocity in an arbitrary direction, the electromagnetic elds are transformed according
to
E0k = Ek ;

B0k

E0? = (E? +
B0? =

= Bk ;

(B?

B? );
E? );

where k and ? indicate components parallel and perpendicular to the relative velocity V:
The invariance of Bx ; the magnetic eld parallel to V; can be seen from the following observation.
Bx appears only in F 23 = F 32 : Since the component F ij transforms similar to the coordinates
xi and xj ; Bx trarnsforms as y and z which are invariant. Therefore, Bx does not change through
Lorentz transformation. Likewise, F 02 = Ey transforms as x0 = ct and x2 = y;
Ey0 = (Ey
F 03 =

Bz );

Ez as
Ez0 = (Ez + By );

and so on. F 01 = Ex = F 01 is invariant since the Lorentz transformation corresponds to

rotation in the (ct; x) plane and F 00 = F 11 = 0; F 01 = F 10 form an antisymmetric tensor.
Example 2 A charge e is moving at a velocity V along the x axis. Find the electric eld and
compare it with the eld expected from the Lienard-Wirchert potentials.
In the frame of the moving charge, the scalar potential and electric eld are
0

where
r0 =

e
er0
0
;
E
=
;
r0
r03

p
x02 + y 02 + z 02 :

The vector potential in the moving frame is zero, A0 = 0: The x component of the electric eld is
14

invariant,
Ex = Ex0
ex0
=
r03
= e

(x
2 (x

= e(1

V t)

V t)2 + y 2 + z 2 ]3=2
x Vt
2

V t)2 + (1

(x
Similarly,
Ey = Ey0 = e(1

Ez = Ez0 = e(1

)(y 2 + z 2 )

3=2

)
V t)2 + (1

(x

)(y 2 + z 2 )

3=2

)(y 2 + z 2 )

3=2

)
V t)2 + (1

(x

;
:

Therefore, the electric eld in the laboratory frame is

2

E = e(1

(x

V t)ex + y + z
2

V t)2 + (1

(x

)(y 2 + z 2 )

3=2

Equivalence of this expression to the Coulomb eld emerging from the Lienard-Wiechert potentials,
2

E = e(1

3 R2

;
ret

can be readily proven by noting

n
1

(x

=
(x

ret

V t)ex + y + z
2

V t)2 + (1

)(y 2 + z 2 )

1=2

where ret means the retarded time. Denoting the angle between the x axis and the position
vector R = (x V t)ex + y + z by ; we nd
E=

eR
R3 (1

1
2

sin2 )3=2

The Coulomb eld is radial with respect to the present location of the charge. At
Ek =
and at

e
(1
R2

= =2;
E? =

)=

= 0;

e 1
;
R2 2

e
:
R2

In highly relativistic case, the eld is dominated by components perpendicular to the velocity.

15

Angular dependence of the electric eld for the case

2

1
2

(1

= 0:9 is shown below in polar plot.

sin2 )3=2

-2

-1

-1

-2

-3

Example 3 Show that the quantities E 2

B 2 and E B are invariant in Lorentz transformation.

Since
2
;
E 2 = Ek2 + E?

and
E0k = Ek ;

E0? = (E? +

B? );

we nd
E 02 = Ek2 +

(E? +

B? )2 :

B 02 = Bk2 +

(B?

E? )2 :

Similarly,

Then
E 02

B 02 = Ek2
= Ek2

= E2

Bk2 +
Bk2 +
B2:

2
2

(E? +
(1

B? )2

2
)(E?

2
B?
)+2

16

E? )2

(B?

[E? (

B ? ) + B? (

E? )]

For E B;
E0 B0 = Ek0 Bk0 + E0? B0?
= Ek Bk +
= Ek Bk +

(E? +
(1

B? ) (B?

E? )

)E? B?

= E B:
Note that
(A

B) (C

D) = (A C)(B D)

(A D)(B C):

The invariance of E B means that if E and B are normal to each other in one reference frame,
they remain so in any other frames. If E or B is zero in one reference frame, in other frames they
are normal to each other. Furthermore, if the elds in the laboratory frame are E and B; there
exists a frame moving at a velocity

E B
V
=c 2
;
(V =c)2
E + B2

wherein electric and magnetic elds are parallel to each other.

10.5

Energy and Momentum Tensor

As shown in Chapter 1 and 3, the electromagnetic force

1
f = E+ J
c

B;

can be expressed as the divergence of the Maxwells stress tensor T ij ;

fi =

@j T ij =

1
@j
4

1 2
(E + B 2 ) ij
2

Ei Ej

Bi Bj ; (i; j = 1; 2; 3):

The time component of the force four vector was

1
f0 = J E:
c
Since
J E=

1 @
(E 2 + B 2 )
8 @t

where S is the Poynting vector

S=

c
E
4

r S;

B;

f 0 can be written as
f0 =

1 @
(E 2 + B 2 )
8 c @t

17

1
r (E
4

B) ;

and a resultant four dimensional Maxwells stress tensor is

2
1
(E 2 + B 2 ) 41 (E B)x 41 (E B)y
6 81
6 4 (E B)x
T 11
T 12
T =6
6 1 (E B)y T 21 (= T 12 )
T 22
4 4
1
B)z T 31 (= T 13 ) T 32 = (T 23 )
4 (E

1
4

(E B)z
T 13
T 23
T 33

3
7
7
7
7
5

The quaintly
1
1
S=
E
c
4

B;

is the momentum ux density and thus

1
4 c

B;

is the momentum density of electromagnetic elds. The angular momentum density is accordingly
given by
1
r (E B);
4 c
and the total electromagnetic angular momentum is
Z
1
r (E B)dV:
4 c
In four dimensional form, the angular momentum tensor can thus be dened by
Z
1
ij
K =
(xi T ik xj T ik )d k ;
c
where d k is the area element in the four dimensional (hyper) space having the dimensions of
volume (cm3 ):

10.6

Relativistic Mechanics

In terms of the velocity four vector

v i = (c; v) = (c; vx ; vy ; vz );
and the eld tensor
F

ij

6
6
=6
6
4

0
Ex
Ey
Ez

Ex
0
Bz
By

18

Ey
Bz
0
Bx

Ez
By
Bx
0

3
7
7
7
7
5

the force four vector to act on a charged particle having a charge e can be formulated as
e
F i = F ij vj ;
c
where
vj = (c; v) = (c; vx ; vy ; vz );
is the covariant four velocity. F i reduces to
F i = (e

E; F);

where
F = e(E +

B);

is the space component of the electromagnetic force. Note that

F=e

E;

reecting the fact that the magnetic eld does not do any work on charged particles.
The magnitude of the velocity four vector is constant,
v i vi =

(1

)c2 = c2 :

The corresponding momentum four vector

E
;p ;
c

pi =
also has a constant magnitude,
E
c

pi p i =

p2 = (mc)2 ;

where
E 2 = (cp)2 + (mc2 )2 ;
is the energy of the particle.
The Lagrangian in nonrelativistic mechanics,
1
L = mv 2 + e(
2

);

can be readily generalized as

L =
=

mc

mc2

+ e(

+ v (P

19

)
mv)

e ;

where
P=

e
e
@L
= mv + A = p+ A;
@v
c
c

is the canonical momentum. The equation of motion can be derived from Lagrange equation
@L
@v

d
dt

= rL:

Noting
r(v A) = v rA + A rv + v
= v rA + v

A+A

A;

since the velocity should be xed in carrying out spatial dierentiation, we nd

d
e
e
(p + A) = (v rA + v
dt
c
c

A) er :

Noting
dA
@A
=
+ v rA; r
dt
@t

A = B;

E=

1 @A
;
c @t

we recover the familiar equation of motion for a charged particle,

dp
d
1
= ( mv) = e E + v
dt
dt
c

B :

Since the momentum p and energy E are related through

p=

v
E;
c2

the acceleration a = dv=dt can be readily found,

a=

e
[E +
m

E)] :

(10.58)

Example 4 Analyze the motion of an electron in the Coulomb eld of a heavy ion carrying a
charge Ze:
Since the electric eld is static, the energy of electron is conserved,
p
E = c p2 + (mc)2

Ze2
= E0 (const.)
r

(10.59)

By suitable choice of coordinates, the problem can be made two dimensional so that electron motion
is conned in the plane (r; ): The momentum can be decomposed into
p2 = p2r +

20

L
r

(10.60)

where L = rp ; the angular momentum, is also conserved. Then,

s

c p2r +
In nonrelativistic limit v

L
r

+ (mc)2

Ze2
= E0 .
r

(10.61)

c, we have
1
2m

L
r

p2r +

Ze2
= const.
r

(10.62)

In the limit r ! 0; the LHS diverges. (Note that in nonrelativistic limit, the momentum is bounded.)
This means that in nonrelativistic limit, the electron cannot approach the ion indenitely. In
relativistic case, however, the LHS of Eq. () remains nite when r ! 0 provided pr ! 1:
The electron trajectory can be found by dierentiating Eq. (10.61) with respect to time,
L2 1
Ze2
+
= 0:
m r3
r2

d
(m r)
_
dt

(10.63)

L = const. means
m r2 _ = L = const.
Then time derivative can be converted into angular derivative through
L
d
d
=
2
dt
m r d

(10.64)

and Eq. () reduces to

d2 1
+ 1
d 2r

where
=
When

Ze2
1
= 2 2 E0 ;
r
L c

Ze2
:
cL

(10.65)

(10.66)

< 1; the solution for r( ) is quasi-oscillatory,

r( ) =

b
p
1 + a cos( 1

(10.67)

where a is an integration constant determined by the initial condition and

b=

(cL )2 (Ze2 )2
:
Ze2 E0

(10.68)

p
2 is in general an irrational number, the oscillation is quasi-periodic without closed
Since 1
p
2 ) in
orbits. The electron cannot approach the ion indenitely. If > 1; or Ze2 > L ; cos( 1
p
2
Eq. (10.67) becomes cosh(
1 ) and in this case the electron can collapse on the ion.
21

Example 5 Find the power radiated by a charge explicitly in terms of the external electric and
magnetic elds.
In nonrelativistic limit, the radiation power is given by the Larmors formula,
P =

2 (ea)2
:
3 c3

(10.69)

This is still applicable in an instantaneous rest frame of the charged particle,

P =

2 (ea0 )2
;
3 c3

(10.70)

where a0 is the acceleration in that frame. According to Example 1, the acceleration in the laboratory frame a and that in the instantaneous rest frame of the charge a0 are related through
a0 =

ak +

a? ;

(10.71)

as can be found by choosing v = V: Since

ak =

e
e
E ; a? =
(E? +
m 3 k
m

B);

(10.72)

we nd for the radiation power

P

=
=
=

2 (ea0 )2
3 c3
2 e4 h 2
Ek + 2 (E? +
3 m2 c3
2 e4 2
(E +
B)2
3 m2 c3

B)2
(

E)2 :

(10.73)

If the electric eld is parallel to the velocity and there is no magnetic eld as in linear accelerators,
the power becomes independent of the particle energy,
P =

2 e4
E2:
3 m2 c3 k

In the absence of the electric eld, we obtain

P =

2 e4
3 m2 c3

22

B)2 :

10.7

Radiation Damping

Consider an electron subject to an acceleration a for a duration T: In nonrelativistic limit, the

electron acquires a kinetic energy of order
1
Ekin = m(aT )2 ;
2
while the amount of energy radiated is
Erad =

2 (ea)2
T:
3 c3

It is reasonable to conjecture that the radiation energy should not exceed the kinetic energy, Erad <
Ekin for otherwise, one must wonder where the radiation energy comes from. This imposes a limit
on T;
e2
T &
= ' 10 23 sec,
mc3
where

is approximately the transit time of light over the classical electron radius,
re =

e2
= 2:85
mc2

10

13

cm.

(To realize such a time scale in cyclotron motion in a magnetic eld, we need an unrealistically
large magnetic eld, B ' 1012 T.) In this time scale, radiation reaction on dynamics of charged
particle is expected to become signicant. However, the uncertainty principle imposes a more strict
constraint,
E t & ~:
If we choose the self energy of the electron for E = mc2 ; we nd
t&
where

~c e2
~
=
= 137 ;
mc2
e2 mc3
e2
1
=
;
~c
137

is the ne structure constant. Therefore, in nonrelativistic cases, quantum mechanical eects

become important well before recoil force due to radiation need to be considered.
However, in radiation from a charge undergoing harmonic motion, radiation reaction is well
dened and has been observed experimentally. Recoil force exerted by radiation may be estimated
from energy balance,
Z

F vdt =

Z
2 e2 T 2
a dt
3 c3 0
Z
2 e2 T d2 v
v 2 dt:
3 c3 0
dt
23

Then
F =

2 e2 d2 v
;
3 c3 dt2

and the equation of motion of harmonic oscillator is modied as

x

...
x + ! 20 x =

eE
e
m

i!t

Solution for x(t) is

x(t) =

e i!t
eE
m !2 + i !3

! 20

which remains bounded at the resonance ! ' ! 0 due to radiation damping. The shift in the
resonance frequency is given by
5 3 2
!'
! :
8 0

24

Mathematical Formulae

1. Vector Formulae
Bold characters are vector functions and f is a scalar function.
A (B

C) = C (A
(B

A
(A

B) (C

B) = B (C

C) = B(A C)

A)

C(A B)

D) = (A C)(B D)

(A D)(B C)

r (f A) = rf A + f r A
r
r(A B) = A

(r

r (A
r

(A

(f A) = rf
B) + B

A + fr
(r

B) = B (r

B) = Ar B

A (r

A)

rf

r (r
(r

A) + (B r)A + (A r)B
B)

Br A + (B r)A

r
r

(A r)B

A)

A) = r(r A)

r2 A

r r = 3; r = position vector
r2

1
jr

r=0

r0 j

(r

r0 )

df
@f
=
+ (v r)f; v = velocity
dt
@t
dA
@A
Substantive derivative:
=
+ (v r)A
dt
@t
@v
@v 1 2
dv
Substantive derivative:
=
+ (v r)v =
+ rv + (r v)
dt
@t
@t
2
Z
I
Gausstheorem:
r AdV =
A dS
Substantive derivative:

Stokestheorem

(r

A dS =

A dl

A) dS = 0 for closed surface

AdV =

dS

A=

dS

2. Delta Function
Z

1
(at) =
(t);
jaj

=
=
=

g(t) [f (t)]dt =

g(t0 )
where f (t0 ) = 0
df
dt t=t0

r0 ) (3-D delta function)

(r
=

t0 )dt = f (t0 )

f (t) (t

(x x0 ) (y y 0 ) (z z 0 ) (Cartesian)
0
(r r0 ) (
)
0
(
) (spherical)
0
rr
sin
(r r0 )
0
(cos
cos 0 ) (
) (spherical)
rr0
0)
(
0
(
) (z z 0 ) (cylindrical)

3. Curvilinear Coordinates
Let ui (x; y; z) (i = 1; 2; 3) be a system of curvilinear coordinates. The metric coe cients are
s
@x 2
@y 2
@z 2
hi =
+
+
;
@ui
@ui
@ui
and the length segments in each direction are hi dui ei (ei unit vector).
Area elements
dSi = hj hk duj duk ej ek
Volume element
dV = h1 h2 h3 du1 du2 du3
Gradient
rf =
Divergence
r A=

1
h1 h 2 h3

3
X
1 @f
hi @ui
i=1

@
@
@
(h2 h3 A1 ) +
(h3 h1 A2 ) +
(h1 h2 A3 )
@u1
@u2
@u3

Curl

A=

e1
h 2 h3

e2
h 3 h1

e3
h1 h 2

@
@u1

@
@u2

@
@u3

h1 A1

h2 A2

h3 A3

Scalar Laplacian
r2 f = r rf =

1
h1 h2 h 3

@
@u1

h2 h3 @f
h1 @u1

+
2

@
@u2

h3 h1 @f
h2 @u2

@
@u3

h1 h2 @f
h3 @u3

Vector Laplacian (denition)

r2 A

r(r A)

(r

A)

Spherical Coordinates (r; ; )

Transformation
x = r sin cos
y = r sin sin
z = r cos
Metric coe cients
hr = 1; h = r; h = r sin :
Transformation of the unit vectors
2
3 2
er
sin cos
4 e 5 = 4 cos cos
sin
e
2
3 2
ex
sin cos
4 ey 5 = 4 sin sin
ez
cos

sin sin
cos sin
cos

cos
sin
0

cos cos
cos sin
sin

sin
cos
0

32

3
ex
5 4 ey 5
ez
32
3
er
54 e 5
e

For example,

ex = rx = r(r sin cos )

= sin cos er + cos cos e

sin e

Derivative of the unit vectors

@er
@e
@er
=e ;
= sin e ;
=
@
@
@
Gradient
rf =

er ;

@e
@e
= cos e ;
=
@
@

sin er

1 @f
1 @f
@f
er +
e +
e
@r
r@
r sin @

Divergence
r A=

1 @ 2
1 @
1 @A
(r Ar ) +
(sin A ) +
2
r @r
r sin @
r sin @

Curl

e
r sin

e
r

@
@r

@
@

@
@

Ar

rA

r sin A

er
2
r sin
r

A=

Scalar Laplacian
r2 f =

@2f
2 @f
1
@
+
+ 2
@r2
r @r
r sin @
3

sin

@f
@

@2f
1
2
r2 sin @ 2

cos e :

Vector Laplacian
r2 A

r(r A)

(r A)
2
2 @A
2 cot
2 @A
=
r 2 Ar
er
Ar
A
2
2
2
2
r
r @
r
r sin @
1
2 @Ar
2 cos @A
+ r2 A
e
2 A + r2 @
2
r sin
r2 sin2 @
1
2 @Ar
2 cos @A
+ r2 A
e
+ 2 2
2 A + r 2 sin
2
@
@
r sin
r sin

Note that in non-cartesian coordinates,

(r2 A)i 6= r2 Ai

Cylindrical Coordinates ( ; ; z)
Transformation

x =

cos

y =

sin

z = z
Metric coe cients
h = 1; h = ; hz = 1
Derivatives of the unit vectors
@e
@e
=e ;
=
@
@
Gradient
rf =

1 @f
@f
@f
e +
e +
ez
@
@
@z

Divergence
r A=

1 @
1 @A
@Az
( A )+
+
@
@
@z

Curl

A=

ez

@
@

@
@

@
@z

Az

Scalar Laplacian
r2 f =

@2f
1 @f
1 @2f
@2f
+
+
+
2
@ 2
@
@z 2
@ 2

Vector Laplacian
r2 A

r(r A)

r (r A)
2 @A
1
=
r2 A
A e
2 @
2
2 @A
1
+ r2 A + 2
A e + r2 Az ez
2
@

4. Special Functions
Bessel Functions Zm (x) = Jm (x); Nm (x) [or Ym (x) in some books]
J0 (x); J1 (x)
y

1
0.75
0.5
0.25
0
0

10

15

20
x

-0.25

J0 (x) (solid line) and J1 (x) (dashed line).

Y0 (x); Y1 (x)
y

0.5

0
5

10

15

20
x

-0.5

-1

Y0 (x) (solid line) and Y1 (x) (dashed line).

Dierential equation
d2
1 d
+
+ k2
d 2
d

m2
2

Zm (k ) = 0

Wronskian
0
Jm (x)Nm
(x)

0
Jm
(x)Nm (x) =

2
x

Series representation of Jm (x)

x
2
x
2

Jm (x) =
Jm (x) '

n=0
m

( 1)n
x
(m + n)!n! 2

for small x

1
2x

Generating functions
x
exp
2

1
t

eix sin =

2n

; J

m (x)

= ( 1)m Jm (x)

1 2
1 4
4 x + 64 x for small
1 3
16 x for small x

J0 (x) ' 1

J1 (x) '

1
X

1
1
X

Jm (x)tm

m= 1
1
X

Jm (x)eim

m= 1

n-th root of J0 (x) = 0; x0n ; and value of J1 (x0n )

x0n
2:40483
J1 (x0n ) 0:51915

5:52008 8:65373
0:34026 0:27145

11:79153 14:93092
0:23246 0:20655

18:07106
0:18773

n-th root of J1 (x) = 0; x1n ; and value of J0 (x1n )

x1n
J0 (x1n )
For small x (
Jm (x) '

3:83171 7:01559
0:40276 0:30012

10:17347 13:32369
0:24970 0:21836

16:47063 19:61586
0:19647 0:18006

1);
x
2

Nm (x) '

ln

x
+
2

= 0:5772

Hankel functions of the rst and second kinds

(1;2)
Hm
(x) = Jm (x)

Asymptotic forms at large x

iNm (x)

Jm (x) '

Nm (x) '
(1;2)
Hm
(x) '

2
cos x
x

2m + 1
4

2
sin x
x

2m + 1
4

2
exp
x

i x

Z00 (x) =

Z1 (x)

2m + 1
4

Recurrence formulae
1
[Zm 1 (x) Zm+1 (x)]
2
x
Zm (x) =
[Zm+1 (x) + Zm 1 (x)]
2m
0
Zm
(x) =

(Eulers constant).

d m
[x Zm (x)] = xm Zm 1 (x)
dx
Integral representations (There are many. A few are listed.)
Z 2
Z
1
2 1 cos(xt)
p
Jm (x) =
cos(m
x sin )d ; J0 (x) =
dt
2 0
1 t2
0
Z
2 1 cos(xt)
p
dt
N0 (x) =
1 t2
1
Z 1
cos xt
2m+1 x m
Nm (x) = p
dt
1
2
1)m+(1=2)
( 2 m) 1 (t
Integrals
Z

Jm (ax)dx =

1
m
tan
a
2

Nm (ax)dx =

2
K(a=b); K : complete elliptic integral of the rst kind
b
8
Z 1
< 1=b; b > a > 0
1=2b; a = b > 0 (step function)
J0 (ax)J1 (bx)dx =
:
0
0;
a>b>0

J0 (ax)J0 (bx)dx =

1
;
a

(a

xJ1 (ax)J1 (bx)dx =

b)

; (derivative of the above with respect to a)

In fact for any integer m;

Z

xJm (ax)Jm (bx)dx =

+x (ax)J

x (ax)dx

=J

(a

b)
a

(2a);

>1

J (ax)dx =

2
a

+
2

1
2

+1

: gamma function

J (ax)J (bx)
i
dx = J (by)H (1) (ay)
2
2
x
y
2
0
Z 1
1
e ax J0 (bx)dx = p
2
a + b2
0
p
Z 1
( a2 + b2 a)
p
e ax J (bx)dx =
b a2 + b2
0

ax

1
a2 + b2 + c2
J (bx)J (cx)dx = p Q 1
; Q : Legendre function of the 2nd kind
2
2bc
bc
p
Z 1
b2
b2
2 2
exp
e a x J2 (px)dx =
I
2a
8a2
8a2
0
Z 1
1
b2
2 2
e a x x2 J0 (bx)dx = 2 exp
2a
4a2
0
Z 1
1
p2 + q 2
pq
2 2
e a x J (px)J (qx)dx = 2 exp
I
2
2
2a
4a
2a
0
7

sin(ax)J (bx)dx =

p 1
sin sin 1 (a=b)
b2 a2
b p
p
cos 2
a2 b2 (a+ a2 b2 )

sin(ax)J0 (bx)dx =

Sum
J0 (x) + 2

8
<

: p

1
X

0;
1
a2

b>a
; a<b

b>a
b2

; a>b

J2n (x) = 1

n=1
1
X

Jn2 (x) = 1

n= 1
1
X

n2 Jn2 (x) =

n=0

J0 (x) + 2

1
X

x2
4

( 1)n J2n (x) = cos x

n=1
1
X

( 1)n J2n+1 (x) =

n=0
1
X

1
sin x
2

(2n + 1)J2n+1 (x) =

n=0

x
2

1
X
1
x2
J
(2nx)
=
2n
n2
2

n=1
1
X

J2n (2nx) =

n=1
1
X

x2
2(1

n2 J2n (2nx) =

n=1
1
X

n=1

x2 )

x2 (1 + x2 )
2(1 x2 )4

J2n (2nt)dt =

x3
6(1 x2 )3

Spherical Bessel Functions zl (x) = jl (x); nl (x)

Spherical Bessel functions are elementary functions. Some low order forms are:
j0 (x) =
n0 (x) =

sin x
;
x

cos x
;
x

j1 (x) =
n1 (x) =

Denition
jl (x)

sin x

x cos x
x2

j2 (x) =

cos x + x sin x
;
x2

2x

Jl+ 1 (x);

nl (x)

(3

n2 (x) =
r

2x

x2 ) sin x
x3
(3

x2 ) cos x + 3x sin x
x3

Nl+ 1 (x)
2

3x cos x

(1;2)

hl

(x) = jl (x)

inl (x)

Dierential equation
d2
2 d
+
+ k2
2
dr
r dr

l(l + 1)
r2

zl (kr) = 0

Wronskian
jl (x)n0l (x)
For x

1
x2

1;
jl (x) '

For x

jl0 (x)nl (x) =

xl
;
(2l + 1)!!

nl (x) '

(2l 1)!!
xl+1

1;
jl (x) '

1
cos x
x

l+1
2

(1)

hl (x) ' ( i)l+1

nl (x) '

eix
;
x

(2)

1
sin x
x

hl (x) ' il+1

l+1
2
ix

Recurrence formulae
(2l + 1)zl (x) = x[zl

1 (x)

zl+1 (x)]

dzl (x)
= lzl 1 (x) (l + 1)zl+1 (x)
dx
d l+1
[x zl (x)] = xl+1 zl 1 (x)
dx
d
[x l zl (x)] = x l zl+1 (x)
dx

(2l + 1)

Modied Bessel Functions Im (x); Km (x)

I0 (x); I1 (x)
y
25
20
15
10
5
0
0

1.25

2.5

3.75

5
x

I0 (x) (solid line) and I1 (x) (dashed line).

K0 (x); K1 (x)

10

7.5

2.5

0
0

0.5

1.5

2.5

3
x

K0 (x) (solid line) and K1 (x) (dashed line).

Denition
Im (x) = e

im =2

x
2

Jm (ix)
1
X
0

(x=2)2n
n!(m + n)!

" k
#
1
k+m
X 1
x
( 1)m X (x=2)m+2k X 1
+ ln
+
+
2
2
k!(m + k)!
n
n

Km (x) = ( 1)m+1 Im (x)

n=1

k 0

1
2

m
X1

( 1)r

(m

r
r!

r=0

1)! x
2

n=1

2r m

Dierential equation
d2
1 d
m2
2
+
+
k
+
2
d 2
d

Im (k )
Km (k )

Wronskian
0
Im
(x)Km (x)

0
Im (x)Km
(x) =

=0

1
x

Series representation of Im (x)

x
2

Im (x) =
For x

1
X

n=0

1
x
m!(m + n)! 2

2n

1
I0 (x) ' 1 + x2 +
4
x
1
I1 (x) ' + x3 +
2 16

Recurrence formulae
Im
Km

1 (x)

1 (x)

2m
Im (x);
x

Im

2m
Km (x);
x

Km

Im+1 (x) =

Km+1 (x) =

10

1 (x)

1 (x)

0
Im+1 (x) = 2Im
(x)

+ Km+1 (x) =

0
2Km
(x)

I00 (x) = I1 (x);

K00 (x) =

K1 (x)

Integral representation
Z

Z 1
tJ0 (xt)
cos xt
p
dt
=
dt
2+1
t
t2 + 1
0
0
Z 1 2
t J1 (xt)
0
K1 (x) = K0 (x) =
dt
1 + t2
0
!
Z 1
3
2x3=2
=p
cos(t3 + xt)dt; (Airys integral)
x 0
33=2
K0 (x) =

K1=3

Legendre Functions Plm (x); Qm

l (x)
Dierential equation
(1

x2 )

d2
dx2

2x

d
+ l(l + 1)
dx

m2
1 x2

Plm (x)
Qm
l (x)

=0

1 dl 2
(x
1)l (Rodriguesformula)
2l l! dxl
1
1+x
Ql (x) = Pl (x) ln
Wl 1 (x); x real and jxj 1
2
1 x
1
z+1
Ql (z) = Pl (z) ln
Wl 1 (z) for general complex z
2
z 1
3
5
2
W 1 (x) = 0; W0 (x) = 1; W1 (x) = x; W2 (x) = x2
;
2
2
3
Special values (m = 0)
Pl (1) = 1; Pl ( 1) = ( 1)l
Pl (x) =

Pl (0) = 0 for odd l; Pl (0) = ( 1)l=2

(l

1)!!
for even l
l!!

l(l + 1)
; Pl0 (0) = (l + 1)Pl+1 (0)
2
Denition of Plm (x); Qm
1)
l (x) in terms of Pl (x); Ql (x) (x real, jxj
Pl0 (1) =

Plm (x) = 1

x2

m=2

dm
Pl (x);
dxm

Qm
l (x) = 1

x2

dm
Pl (z);
dz m

2
Qm
l (z) = z

m=2

dm
Ql (x)
dxm

For general complex argument z

Plm (z) = z 2

m=2

m=2

Orthogonality of Plm (x)

Z

1
1
1
1

Plm (x)Plm (x)

1 (l + m)!
dx =
2
1 x
m (l m)!
Plm (x)Plm
0 (x)dx =
11

mm0

2 (l + m)!
2l + 1 (l m)!

ll0

dm
Ql (z)
dz m

Wronskian
m = 0; Pl (x)Q0l (x)

Plm (x)

dQm
l (x)
dx

Pl0 (x)Ql (x) =

dPlm (x) m
Ql (x) =
dx

1
1

x2

22m [(l + m + 1)=2] [(l + m)=2 + 1]

1 x2 [(l m + 1)=2] [(l m)=2 + 1]
1 (l + m)!
; 0 m l
1 x2 (l m)!

=
Generating function

(cos + i sin cos )l = Pl (cos ) + 2

l
X

im

m=1

l!
P m (cos ) cos(m )
(l + m)! l

Low order forms of Plm (x); (x = cos )

m=0
P0 (x) = 1; P1 (x) = x; P2 (x) =
l = 1; m = 1

P11 (x) =
l = 2; m = 1; 2
P21 (x) = 3

p
1

P22 (x) = 3(1

5x3 3x
3x2 1
; P3 (x) =
;
2
2

x2 = sin

x2 x = 3 sin cos =
3
x2 ) = 3 sin2 = (1
2

3
sin 2
2
cos 2 )

l = 3; m = 1; 2; 3
P31 (x) =

3p
1
2

x2 (5x2

3
1) = (sin + 5 sin 3 )
8
15
(cos
4

P32 (x) = 15(1

x2 )x =

P33 (x) = 15(1

x2 )3=2 = 15 sin3 =

cos 3 )
15
(2 sin 2 + 7 sin 4 )
4

Qm
l (x) (associated Legendre function of the second kind)
Special case z = i sinh
Q0 (i sinh ) =

i cot

(sinh ); Q1 (i sinh ) = sinh cot

Q11 (i sinh ) = cosh cot

(sinh )

(sinh )

tanh

Spherical harmonic function Ylm ( ; )

m

Positive m : Ylm ( ; ) = ( 1)

2l + 1 (l m)! m
P (cos )eim for 0
4
(l + m)! l
12

Negative m : Yl;

m(

; ) = ( 1)m Ylm ( ; ) for

General form
Ylm ( ; ) = (

m+jmj
1) 2

2l + 1 (l jmj)! jmj
P (cos )eim ; for
4
(l + jmj)! l

1
Y00 ( ; ) = p
4
r

3
cos
4
r
3
sin e
8

Y10 ( ; ) =
Y1;

1(

; ) =

Y20 ( ; ) =
Y2;

1(

; ) =

Y2;

2(

; ) =

5 1
(3 cos2
1)
4 2
r
15
sin cos e i
8
r
1 15
sin2 e i2
4 2

Orthogonality of Ylm ( ; )
I
Ylm ( ; )Yl0 m0 ( ; )d
Toroidal functions Pl

1
2

(cosh ); Ql

d
d2
+ coth
2
d
d

1
2

ll0 mm0 ;

= sin d d

(cosh ) satisfy
l2 +

1
4

m2 cosech2

F( ) = 0

Integral representations
Plm 1 (cosh ) =
2

(cosh ) =
Qm
l 1
2

( 1)m (2l 1)!!

2m+1 (2l 2m 1)!!
( 1)m (2l 1)!!
2m+1 (2l 2m 1)!!

cos m
1

(cosh + cos sinh )l+ 2

cosh mt
1

(cosh + cosh t sinh )l+ 2

dt

Gamma Function
Denition
(z) =

e t tz

dt

Properties
(z + 1) = z (z) ;

(z) (1

z) =

sin ( z)

If z is a positive integer, z = n;
(n + 1) = n!
13

z+

1
2

1
2

z =

cos ( z)

Special values
1
2

(1) = 1;

n+

1
2

(2n
2n

1)!! p

(x)
y

2.5

0
-5

-2.5

2.5

5
x

-2.5

-5

Gamma function

(x) :

Elliptic Integrals K (k 2 ) and E (k 2 )

Complete Elliptic Integrals of the First Kind K(k 2 ) and Second Kind E k 2
Denitions
Z =2 p
Z =2
1
2
2
p
d ; E k =
1 k 2 sin2 d ; 0
K k =
2
2
0
0
1 k sin
Special values

K(0) = E (0) =

; lim K (1

") = ln

"!0

K (x) =

=2

E (x) =

x sin2

x sin2 d

=2 p

14

4
p
"

; E (1) = 1

k2

3
2.5
2
1.5
1
0.5
0
0

0.25

0.5

0.75

1
x

K k 2 (thick line) and E k 2 (thin line).

Relationship between K and E
E(k 2 ) = 1

k2

d
kK k 2
dk

Integrals that can be reduced to the elliptic integrals

Z =2
sin2
1
p
E k2
d = 2 K k2
2
2
k
0
1 k sin
Z =2
cos2
1
p
(1 k 2 )K k 2
d = 2 E k2
2
2
k
0
1 k sin
Z =2
1
1
d
d =
E k2 =
kK k 2
2
2
3=2
2
1 k
dk
(1 k sin )
0
Z =2
sin2
1
d = 2
E k2
(1 k 2 )K k 2
2
2
3=2
2
k
(1
k
)
(1 k sin )
0
Z =2
1
cos2
d = 2 K k2
(1 k 2 )E k 2
2 sin2 )3=2
k
(1
k
0

Series Expansion of Elementary Functions

1
=1
1+x

x + x2

x3 +

; jxj < 1

1
1 2
x + x3 +
2!
3!
1 2
1
cos x = 1
x + x4
2!
4!
1 3
1
sin x = x
x + x5
3!
5!
1 3
2 5
17 7
tan x = x + x + x +
x +
3
15
315
ex = 1 + x +

15

1 2
1
x + x4 +
2!
4!
1 3
1 5
sinh x = x + x + x +
3!
5!
1 3
2
17 7
tanh x = x
x + x5
x +
3
15
315
1 2 1 3
x + x
; jxj < 1
ln(1 + x) = x
2
3
cosh x = 1 +

Innite Products
1
Y

1+

x2
n2

sinh x
x

x2
n2

sin x
x

n=1
1
Y

n=1
1
Y

1+

n=1
1
Y

n=1
1
Y

n= 1

(2n

1)2

= cosh

x
2

= cos

x
2

x2
(2n

1)2

1+

x2
(a 2n )2

cosh x cos a
1 cos a

x2
(a 2n )2

cos x cos a
1 cos a

n= 1
1
Y

x2

16

+iuihi?Uit
41 M1 G1 Mdfnvrq/

Fodvvlfdo Hohfwurg|qdplfv/

wklug hglwlrq +4<<<,1 Vwdqgdug judgxdwh wh{werrn

lq Qruwk Dphulfd1
51 O1 G1 Odqgdx dqg H1 P1 Olivklwv/

Hohfwurg|qdplfv lq Frqwlqxrxv Phgld1

Uljrurxv wuhdwphqw

ri hohfwurg|qdplfv lq pdwwhu1
61 O1 G1 Odqgdx dqg H1 P1 Olivklwv/

Fodvvlfdo Wkhru| ri Ilhogv1

Udgldwlrq ri hohfwurpdjqhwlf

zdyhv dqg judylw| zdyhv1 +Fodvvlfdo hohfwurg|qdplfv dqg jhqhudo uhodwlylw|1,

71 Z1 U1 Vp|wkh/

Vwdwlf dqg G|qdplf Hohfwulflw|1

Qxphurxv h{dpsohv ri vwdwlf sureohpv1

81 F1 Frkhq0Wdqqrxgml/ M1 Gxsrqw0Urf/ dqg J1 Ju|qehuj/

THG1

Skrwrqv dqg Dwrpv= Lqwurgxfwlrq wr

Uljrurxv irxqgdwlrq ri vhfrqg txdqwl}dwlrq1

91 Y1 Y1 Edw|jlq dqg L1 Q1 Wrsw|jlq/

Sureohpv lq Hohfwurg|qdplfv1

Froohfwlrq ri vrph :33

sureohpv zlwk vroxwlrqv1

:1 V1 Udpr/ M1 U1 Zklqqhu|/ dqg W1 Ydq Gx}hu/

Ilhogv dqg Zdyhv lq Frppxqlfdwlrq1

Fodvvlf

hqjlqhhulqj wh{werrn1 Jrrg fkdswhuv rq dqwhqqd udgldwlrq dqg rswlfdo frppxqlfdwlrq1

;1 S1 Prrq dqg G1 H1 Vshqfhu/

Ilhog Wkhru| iru Hqjlqhhuv1

Jrrg uhihuhqfh iru ydulrxv xvhixo

frruglqdwhv1
<1

Kdqgerrn ri Pdwkhpdwlfdo Ixqfwlrqv/

hg1 P1 Deudprzlw} dqg L1 H1 Vwhjxq1 Vwdqgdug pdwk

kdqgerrn1
431 S1 Pruvh dqg K1 Ihvkedfk/

Phwkrgv ri Wkhruhwlfdo Sk|vlfv1

Fodvvlf wh{werrn ri wkhruhwlfdo

sk|vlfv1
441 L1 V1 Judgvkwh|q dqg L1 P1 U|}kln/

Wdeoh ri Lqwhjudov/ Vhulhv/ dqg Surgxfwv1

Jrrg uhihuhqfh

ri lqwhjudov ri vshfldo ixqfwlrqv1

451 S1 M1 Qdklq/

Rolyhu Khdylvlgh= Vdjh lq Vrolwxgh1

Iru ixq uhdglqj1 +Khdylvlgh lv rqh ri p|

khurv1,
461 I1 N1 Ulfkwp|hu/ H1 K1 Nhqqdug/ dqg M1 Q1 Frrshu/
klvwru| ri ghyhorsphqw ri prghuq sk|vlfv1

Lqwurgxfwlrq wr Prghuq Sk|vlfv1

Jrrg