Physics 506 Winter 2008

Homework Assignment #5 Solutions

Textbook problems: Ch. 9: 9.22, 9.23, 9.24
Ch. 10: 10.1
9.22 A spherical hole of radius a in a conducting medium can serve as an electromagnetic
resonant cavity.
a) Assuming innite conductivity, determine the transcendental equations for the
characteristic frequencies
lm
of the cavity for TE and TM modes.
Because of the spherical symmetry, it is natural to describe the modes of the
spherical cavity in terms of a vector spherical wave expansion. These waves fall
into either TE or TM modes, depending on whether r

E = 0 or r

H = 0,
respectively. The TE (or magnetic multipole) modes are given by

H =
i
k

[j
l
(kr)

X
lm
],

E = Z
0
j
l
(kr)

X
lm
(1)
where we have chosen the spherical Bessel function j
l
(kr) since it is regular at
r = 0. FOr a perfect conductor, we impose the boundary conditions H

= 0 and
E

= 0 at r = a. More precisely, we demand

r

H

r=a
= 0, r

E

r=a
= 0
These are equivalent to the condition j
l
(ka) = 0, and leads to the quantization
k
nlm
= x
ln
/a where x
ln
is the n-th zero of the spherical Bessel function j
l
. The
TE
nlm
frequencies are thus

nlm
=
x
ln
c
a
, j
l
(x
ln
) = 0, l 1, |m| l
Each frequency specied by l and n is (2l + 1)-fold degenerate, with azimuthal
quantum number labeled by m.
The TM (or electric multipole) modes are similar, although the boundary condi-
tions are somewhat more involved. The modes themselves are given by

H = j
l
(kr)

X
lm
,

E = Z
0
i
k

[j
l
(kr)

X
lm
] (2)
This time, the H

= 0 boundary condition is automatic, while the E

= 0
condition gives
r (

[j
l
(kr)

X
lm
])

r=a
= 0
This vector quantity may be simplied using
r(

V ) =

(r

V )

V (r

V =

(r

V )
_
1 + r

r
_

V =

(r

V )

r
r

V
Using

V = j
l
(kr)

X
lm
with r

X
lm
= 0 gives
r ([j
l
(kr)

X
lm
]) =

r
(rj
l
(kr))

X
lm
(3)
Hence the E

= 0 boundary condition leads to the TM

nlm
frequencies

nlm
=
y
ln
c
a
,
d
dx
[xj
l
(x)]

z=y
ln
= 0, l 1, |m| l
The y
ln
correspond to zeros of [xj
l
(x)]

or equivalently j
l
(x) + xj

l
(x).
b) Calculate numerical values for the wavelength
lm
in units of the radius a for the
four lowest modes for TE and TM waves.
The numerical values for the wavelengths are obtained from the zeros x
ln
and
y
ln
. For TE
nlm
modes, the rst four zeros of j
l
(x) are
x
11
= 4.4934, x
21
= 5.7635, x
31
= 6.9879, x
12
= 7.7253
Since k
nlm
= x
ln
/a and
nlm
= 2/k
nlm
, we end up with
nlm
/a = 2/x
ln
or

11m
a
= 1.398,

12m
a
= 1.090,

13m
a
= 0.899,

21m
a
= 0.813
All these modes are (2l + 1)-fold degenerate.
For TM
nlm
modes, the rst four zeros of [xj
l
(x)]

are
y
11
= 2.7437, y
21
= 3.8702, y
31
= 4.9734, y
41
= 6.0619
with corresponding wavelengths

11m
a
= 2.290,

12m
a
= 1.623,

13m
a
= 1.263,

14m
a
= 1.036
Note that the next mode, given by y
12
= 6.1168 is nearly degenerate with y
41
.
c) Calculate explicitly the electric and magnetic elds inside the cavity for the lowest
TE and lowest TM mode.
The lowest TE and TM modes both have l = 1. Thus we begin with an overview
of l = 1 vector spherical harmonics

X
1m
=
1

LY
1m
It is natural to write the angular momentum operator

L in terms of raising and
lowering components
L
+
= L
x
+ iL
y
, L

= L
x
iL
y
, L
z
Using
L
+
Y
lm
=
_
l(l + 1) m(m + 1)Y
l,m+1
L

Y
lm
=
_
l(l + 1) m(m1)Y
l,m1
L
z
Y
lm
= mY
l
m
for l = 1 gives
X
+
11
= 0, X
z
11
=
1

2
Y
11
, X

11
= Y
10
X
+
10
= Y
11
, X
z
10
= 0, X

10
= Y
1,1
X
+
1,1
= Y
10
, X
z
1,1
=
1

2
Y
1,1
, X

1,1
= 0
(4)
A vector with components (V
+
, V

, V
z
) can be converted to spherical coordinates
(V
r
, V

, V

) according to
V
r
=
1
2
(V
+
e
i
+ V

e
i
) sin + V
z
cos
V

=
1
2
(V
+
e
i
+ V

e
i
) cos V
z
sin
V

=
i
2
(V
+
e
i
V

e
i
)
Using the explicit form of the spherical harmonics then gives
X
r
11
= 0, X

11
=
_
3
16
e
i
, X

11
= i
_
3
16
cos e
i
X
r
10
= 0, X

10
= 0, X

10
= i
_
3
8
sin
X
r
1,1
= 0, X

1,1
=
_
3
16
e
i
, X

1,1
= i
_
3
16
cos e
i
We are now ready to examine the explicit electric and magnetic elds. From the
expression (1) for TE
nlm
modes, we have

E
11m
= Z
0
j
1
(kr)

X
1m
,

H
11m
=
i
Z
0
k

E
11m
The m = 0 mode is the most straightforward to write down

E
110
= iZ
0
_
3
8
j
1
(kr) sin

H
110
=
1
kr
_
3
8
_
2j
1
(kr) cos r [krj
0
(kr) j
1
(kr)] sin

_
(5)
Note that we have used the spherical Bessel function identity
j

l
() = j
l1
()
l + 1

j
l
()
Even more explicitly, we have
j
1
() =
sin

2

cos

[j
1
()]

= j
0
() j
1
() =
_
1

2
1
_
sin +
cos

The m = 1 mode is given by

E
111
= Z
0
_
3
16
j
1
(kr)e
i
(

+ i cos

H
111
=
1
kr
_
3
16
e
i
_
2j
1
(kr) sin r [krj
0
(kr) j
1
(kr)](cos

+ i

)
_
(6)
while the m = 1 mode is given by

E
11,1
= Z
0
_
3
16
j
1
(kr)e
i
(

i cos

H
11,1
=
1
kr
_
3
16
e
i
_
2j
1
(kr) sin r + [krj
0
(kr) j
1
(kr)](cos

)
_
(7)
We now turn to the lowest TM mode, which is the TM
11m
mode with elds given
by (2)

H
11m
= j
1
(kr)

X
1m
,

E
11m
=
iZ
0
k

H
11m
It ought to be clear the the roles of

E and

H are interchanged between the TE
and TM modes. In particular, the TM
11m
elds may be obtained from the TE
11m
elds of (5), (6) and (7) through the substitution

E Z
0

H, Z
0

H

E
(This is essentially the action of electric-magnetic duality.) Explicitly, the TM
11m
modes correspond to

H
110
= i
_
3
8
j
1
(kr) sin

E
110
=
Z
0
kr
_
3
8
_
2j
1
(kr) cos r [krj
0
(kr) j
1
(kr)] sin

H
111
=
_
3
16
j
1
(kr)e
i
(

+ i cos

E
111
=
Z
0
kr
_
3
16
e
i
_
2j
1
(kr) sin r [krj
0
(kr) j
1
(kr)](cos

+ i

)
_

H
11,1
=
_
3
16
j
1
(kr)e
i
(

i cos

E
11,1
=
Z
0
kr
_
3
16
e
i
_
2j
1
(kr) sin r + [krj
0
(kr) j
1
(kr)](cos

)
_
Note, however, that the wavenumbers k
nlm
are quantized dierently for the TE
versus the TM modes.
9.23 The spherical resonant cavity of Problem 9.22 has nonpermeable walls of large, but
nite, conductivity. In the approximation that the skin depth is small compared to
the cavity radius a, show that the Q of the cavity, dened by equation (8.86), is given
by
Q =
a

for all TE modes

Q =
a

_
1
l(l + 1)
x
2
lm
_
for TM modes
where x
lm
= (a/c)
lm
for TM modes.
In order to calculate the Q factor, we need to obtain both the stored energy and
the power loss at the walls. We start with the simpler case of TE modes, given
by (1). The energy density for harmonic elds is
u =

0
4
|

E|
2
+

0
4
|

H|
2
However, the energy is equally distributed between

E and

H. Thus for TE modes
we may immediately write down
u =

0
2
|

E|
2
=

0
2
j
l
(kr)
2
|

X
lm
|
2
The stored energy is given by integrating this over the volume of the sphere
U =

0
2
_
j
l
(kr)
2
|

X
lm
|
2
r
2
drd =

0
2
_
a
0
j
l
(kr)
2
r
2
dr
Note that we have used orthonormality of the vector spherical harmonics to sim-
plify the integral. We now use the normalization integral for spherical Bessel
functions
_
a
0
j
l
(x
lm
/a)j
l
(x
ln
/a)
2
d =
1
2
a
3
[j

l
(x
ln
)]
2

mn
to obtain
U
lmn
=

0
a
3
4
j

l
(x
ln
)
2
(8)
The power loss is given in terms of the tangential magnetic eld at the conducting
surface
P =
1
2
_
| r

H|
2
da
Using

H = (i/k)

j
l
(kr)

X
lm
from (1) as well as the vector identity (3) gives
P
lmn
=
1
2
_
r=a
_
1
kr
d
dr
rj
l
(kr)
_
2
|

X
lm
|
2
r
2
d
=
1
2k
2
([rj
l
(kr)]

)
2

r=a
=
1
2k
2
(j
l
(ka) + kaj

l
(ka))
2
=
a
2
2
j

l
(x
ln
)
2
(9)
where in the last line we made use of the fact that ka = x
ln
and that j
l
(x
ln
) = 0.
Combining (8) and (9) then gives the Q factor for TE modes
Q
lmn
=
U
lmn
P
lmn
=

0
a
2
=
a

where we made use of the denition of the skin depth =

_
2/
0
.
The calculation for TM modes is similar. However, the appropriate spherical
Bessel function normalization integral needs to be modied for integrating to zeros
of [xj
l
(x)]

. Here we simply state that the appropriate normalization integral may

be written as
_
a
0
j
l
(
m
/a)j
l
(
n
/a)
2
d =
1
2
a
3
_
1 +
p(p 1) l(l + 1)

2
n
_
[j
l
(
n
)]
2

mn
where
n
is the n-th positive zero of
[x
p
j
l
(x)]

= 0
The elds for the TM modes are given in (2), while the characteristic frequencies
are given in terms of zeros of [xj
l
(x)]

. We thus set p = 1 in the above normal-

ization integral and use the notation y
ln
to denote the n-th zero of [xj
l
(x)]

= 0.
The expression for the TM stored energy then becomes
U
lmn
=

0
2
_
a
0
j
l
(kr)
2
r
2
dr =

0
a
3
4
_
1
l(l + 1)
y
2
ln
_
j
l
(y
ln
)
2
The power loss is
P
lmn
=
1
2
_
| r

H|
2
da =
1
2
_
r=a
j
l
(kr)
2
| r

X
2
lm
|r
2
d =
a
2
2
j
l
(y
mn
)
2
As a result, the Q factor for a TM
lmn
mode is
Q
lmn
=
U
lmn
P
lmn
=

0
a
2
_
1
l(l + 1)
y
2
ln
_
=
a

_
1
l(l + 1)
y
2
ln
_
9.24 Discuss the normal modes of oscillation of a perfectly conducting solid sphere of radius
a in free space.
a) Determine the characteristic equations for the eigenfrequencies for TE and TM
modes of oscilation. Show that the roots for always have a negative imaginary
part, assuming a time dependence of e
it
.
Setting up this perfectly conducting sphere problem is similar to what we did
for the spherical hole problem. However, an important feature of the sphere in
free space is that the volume of the resonant cavity is unbounded (ie it is all of
space outside of the radius a). An important physical consequence of this is that
oscillating electromagnetic elds will radiate out to innity. Since power is lost
to innity, these so-called normal modes are actually unstable in the sense that
they decay away after a while. Such modes are generally denoted quasi-normal
modes, and are described by a complex frequency . For
=
0

i
2
(10)
the electric and magnetic elds behave as

E e
it
= e
t/2
e
i
0
t
Hence the imaginary part of the quasi-normal mode frequency governs the decay
of the elds. Since energy is proportional to the square of the elds, the energy
decays as e
t
. Note that 0 is essential for this to make sense. If were
negative, then the mode would grow exponentially with time. Clearly this would
violate energy considerations. In fact, so long as radiation is emitted and reaches
innity, the mode must necessarily decay. In this case, we may argue that is
strictly positive. In terms of the frequency in (10), energy conservation then
demands that always has a negative imaginary part.
In order to actually work out the quasi-normal mode frequencies, we note that
TE modes are given by the analog of (1) for the exterior problem with outgoing
radiation

H =
i
k

h
(1)
l
(kr)

X
lm
,

E = Z
0
h
(1)
l
(kr)

X
lm
Here we have used physical outgoing radiation boundary conditions to select the
rst spherical Hankel function h
(1)
l
. The TE boundary conditions are identical to
what we found above, namely
r

H

r=a
= 0, r

E

r=a
= 0
This corresponds to the equation
h
(1)
l
(ka) = 0 (TE modes)
Unlike in the case of the spherical Bessel functions j
l
() and n
l
(), the spherical
Hankel functions do not admit any real zeros. One way to see this is to note that
h
(1)
l
() is dened as the complex combination
h
(1)
l
() = j
l
() + in
l
()
If were real, then the only way for h
(1)
l
() to vanish is if both real and imaginary
parts [ie j
l
() and n
l
()] were to simultaneously vanish for the same . However,
it is easy to see that the zeros of j
l
and n
l
never coincide. Therefore, the zeros of
h
(1)
l
are always complex. In fact, h
(1)
l
has precisely l zeros in the complex plane.
To show this, we note that h
(1)
l
() may be written as a complex polynomial in
1/ times the outgoing spherical wave factor e
i
/. In particular
h
(1)
l
() = (i)
l+1
e
i

s=0
(l + s)!
s!(l s)!
_
i
2
_
s
Ignoring the irregular point at innity, the zeros of h
(1)
l
then correspond to the
zeros of the polynomial
P
l
() =
l

s=0
(2l s)!
(l s)!s!
(2i)
s
Since this is a polynomial of degree l, it admits precisely l complex zeros. In
fact, it can be shown that all these zeros have negative imaginary part, and
approximately lie along an arc in the lower half complex plane. The zeros of
h
(1)
l
() are plotted for small values of l as
-6 -4 -2 2 4 6
-4
-2
2
4
l=1 l=2 l=3 l=4 l=5

The TE
nlm
frequencies are thus

nlm
=
x
ln
c
a
, h
(1)
l
(x
ln
) = 0, l 1, |m| l, n = 1, 2, . . . , l
where x
ln
denotes the n-th zero of the spherical Hankel function h
(1)
l
.
The TM modes may also be worked out in a similar fashion. In particular, the
analog of (2) for the exterior problem is

H = h
(1)
l
(kr)

X
lm
,

E = Z
0
i
k

[h
(1)
l
(kr)

X
lm
]
This time, the conducting sphere boundary conditions lead to
d
dx
[xh
(1)
l
(x)]

x=ka
= 0 (TM modes)
This time, there are l + 1 zeros, which also approximately lie along arcs in the
lower half complex plane
-6 -4 -2 2 4 6
-4
-2
2
4
l=1 l=2 l=3 l=4 l=5

Hence the TM
nlm
frequencies are

nlm
=
y
ln
c
a
,
d
dx
[xh
(1)
l
(x)]

x=y
ln
= 0, l 1, |m| l, n = 1, 2, . . . , l+1
b) Calculate the eigenfrequencies for the l = 1 and l = 2 TE and TM modes.
Tabulate the wavelength (dened in terms of the real part of the frequency) in
units of the radius a and the decay time (dened as the time taken for the energy
to fall to e
1
of its initial value) in units of the transit time (a/c) for each of the
modes.
For l = 1 and l = 2, the spherical Hankel functions are explicitly
h
(1)
1
() =
e
i

_
1 +
i

_
, h
(1)
2
() =
ie
i

_
1 +
3i

2
_
The zeros of h
(1)
l
(l = 1, 2) are then
x
11
= i
x
21
=

3
2

3i
2
, x
22
=

3
2

3i
2
while the zeros of [h
(1)
l
()]

(l = 1, 2) are
y
11
=

3
2

i
2
, y
12
=

3
2

i
2
y
21
1.596i, y
22
1.807 0.702i, y
23
1.807 0.702i
Since the complex frequencies are given by these zeros multiplied by c/a, we end
up with
Mode
nlm
/a /(a/c)
TE
11m
1/2
TE
12m
4/

3 1/3
TM
11m
4/

3 1
TM
12m
0.313
TM
22m
3.476 0.712
where the wavelength and the energy decay time is given by
=
2c

i
2
10.1 a) Show that for arbitrary initial polarization, the scattering cross section of a per-
fectly conducting sphere of radius a, summed over outgoing polarizations, is given
in the long-wavelength limit by
d
d
(
0
, n
0
, n) = k
4
a
6
_
5
4
|
0
n|
2

1
4
| n ( n
0

0
)|
2
n
0
n
_
where n
0
and n are the directions of the incident and scattered radiations, respec-
tively, while
0
is the (perhaps complex) unit polarization vector of the incident
radiation (
0

0
= 1; n
0

0
= 0).
If all polarizations are specied, the conducting sphere scattering cross section is
given by
d
d
( n, ; n
0
,
0
) = k
4
a
6
|

1
2
( n

) ( n
0

0
)|
2
(11)
What we would like to do is to sum this over both orthogonal outgoing polariza-
tions. One way to do this is to introduce a linear polarization basis transverse
to the outgoing direction n. To do so, we rst assume the scattering is not in
the forward direction. Then the incoming direction n
0
may be used to dene
orthogonal polarizations

1
=
n n
0
sin
,
2
= n
1
=
n( n n
0
) n
0
sin
where is the angle between n and n
0
. In particular, we may write sin
2
=
1 ( n n
0
)
2
. In this case, the cross section summed over outgoing polarizations
becomes
d
d
( n; n
0
,
0
) =
k
4
a
6
1 ( n n
0
)
2
_
|( n n
0
)
0

1
2
( n ( n n
0
)) ( n
0

0
)|
2
+|( n( n n
0
) n
0
)
0

1
2
( n ( n( n n
0
) n
0
)) ( n
0

0
)|
2

=
k
4
a
6
1 ( n n
0
)
2
_
|( n n
0
)
0

1
2
( n( n n
0
) n
0
) ( n
0

0
)|
2
+|( n n
0
)( n
0
)
1
2
( n
0
n) ( n
0

0
)|
2

=
k
4
a
6
1 ( n n
0
)
2
_
| n ( n
0

0
)
1
2
( n n
0
) n ( n
0

0
)|
2
+|( n n
0
)( n
0
)
1
2
( n
0
)|
2

=
k
4
a
6
1 ( n n
0
)
2
_
| n ( n
0

0
)|
2
(1
1
2
( n n
0
))
2
+| n
0
|
2
(
1
2
( n n
0
))
2

Note that we have used transversality of the initial polarization, n

0

0
= 0. To
proceed, we expand the squares and rewrite the above as
d
d
( n; n
0
,
0
) =
k
4
a
6
1 ( n n
0
)
2
_
(
5
4
( n n
0
))(| n ( n
0

0
)|
2
+| n
0
|
2
)
(1 ( n n
0
)
2
)(
1
4
| n ( n
0

0
)|
2
+| n
0
|
2

(12)
The second line cancels the denominator. However the rst line needs a bit of
work. We now use the fact that
0
is a unit polarization vector orthogonal to n
0
.
As a result, the three vectors
n
0
,
0
, n
0

0
(13)
form a normalized right-handed coordinate basis spanning the three-dimensional
space. (There is a slight subtlety if
0
is complex, although the end result is okay,
provided we are careful with magnitude squares.) The components of n expanded
in this basis are
n n
0
, n
0
, n ( n
0

0
)
and since n is a unit vector, the sum of the squares of these components must be
one. In other words
( n n
0
)
2
+| n
0
|
2
+| n ( n
0

0
)|
2
= 1
where we have been careful about complex quantities. Using this result, we see
that the denominator in (12) can be completely eliminated, resulting in
d
d
( n; n
0
,
0
) = k
4
a
6
[
5
4
( n n
0
)
1
4
| n ( n
0

0
)|
2
| n
0
|
2
] (14)
b) If the incident radiation is linearly polarized, show that the cross section is
d
d
(
0
, n
0
, n) = k
4
a
6
_
5
8
(1 + cos
2
) cos
3
8
sin
2
cos 2
_
where n n
0
= cos and the azimuthal angle is measured from the direction of
the linear polarization.
As stated, the scattering angle is given by n n
0
= cos . The azimuthal angle
is the one between n and
0
, measured in the plan perpendicular to n
0
. What
this means is that, using the basis vectors (13) with
0
real, the components of n
can be written as
n = n
0
cos +
0
sin cos + ( n
0

0
) sin sin
or alternatively
n n
0
= cos , n
0
= sin cos , n ( n
0

0
) = sin sin
Substituting this into (14) gives
d
d
(, ) = k
4
a
6
[
5
4
cos
1
4
sin
2
sin
2
sin
2
cos
2
]
= k
4
a
6
[
5
4
cos
1
8
sin
2
(1 cos 2)
1
2
sin
2
(1 + cos 2)]
= k
4
a
6
[
5
8
(1 + cos
2
) cos
3
8
sin
2
cos 2]
c) What is the ratio of scattered intensities at = /2, = 0 and = /2, = /2?
Explain physically in terms of the induced multipoles and their radiation patterns.
At = /2, we have
d
d
(/2, ) = k
4
a
6
[
5
8

3
8
cos 2]
Hence
d
d
(/2, 0) =
1
4
k
4
a
6
,
d
d
(/2, /2) = k
4
a
6
Scattering at 90

is fairly easy to understand physically. For = 0, the scattered

wave is lined up with the incident polarization
0
. Since the polarization is given
by the electric eld vector, this indicates that the induced electric dipole of the
sphere is lined up with the direction of the scattered wave. Since the radiation
must be transverse, no dipole radiation can be emitted on axis, and in this case
the scattering must be purely magnetic dipole in nature. On the other hand, for
= /2, the scattered wave is lined up with the incident magnetic eld, and
hence the scattering must be purely electric dipole in nature. This demonstrates
that the maximum strength of magnetic dipole scattering is a quarter that of
electric dipole scattering. This is in fact evident by the factor of 1/2 in the
magnetic dipole term in the cross section expression (11).