Phys.

506 Electricity and Magnetism Winter 2004

Prof. G. Raithel
Problem Set 3
Total 40 Points

1. Problem 9.1 10 Points

This problem deals with the implications of the fact that negative frequencies are not allowed in
the formalism of Chapter 9 of Jackson (and other parts of the textbook that deal with harmonic
time-dependence).

a): For a rigid charge distribution with a body frame with coordinates (r, θ, φ̃) rotating with an angular-
velocity vector ẑω in the laboratory frame with coordinates (r, θ, φ) it is

ρ(x, t) = ρ(r, θ, φ̃) = ρ(r, θ, φ − ωt)

The usual time-dependent multipole moments in the laboratory frame are

Z ½Z ¾
l ∗ 3 l ∗
qlm (t) = r Ylm (θ, φ)ρ(r, θ, φ − ωt)d x = r Ylm (θ, φ̃)ρ(r, θ, φ̃)d3 x exp(−imωt) = q̃lm exp(−imωt)

where q̃lm is a fixed multipole moment in the body frame (q̃lm can be thought of as the usual time-dependent
multipole moment evaluated at t = 0, i.e. q̃lm = qlm (t = 0)). Since ρ is real and Yl,−m = (−1)m Yl,m∗
, it is
m ∗
q̃l,−m = (−1) q̃l,m .

The positive and negative values of m correspond to positive and negative frequencies. Negative frequen-
cies are not allowed in the formalism of Chapter 9 of Jackson (and other parts of the textbook
that deal with harmonic time-dependence). We must therefore re-write equations such that only positive
frequencies occur, and deduce suitable multipole moments.

Consider, for instance, the (real-valued) electrostatic potential in the near-field limit at an observation point
(ro , θo , φo ):

l=∞,m=l
X µ ¶
1 4π 1
Φ(xo , t) = Ylm (θo , φo )qlm (t)
4π²0 2l + 1 rol+1
l=0,m=−l

1 X µ 4π ¶ 1
= [Ylm (θo , φo )qlm (t) + Yl,−m (θo , φo )ql,−m (t)]
4π²0 2l + 1 rol+1
l,m≥0

1 X µ 4π ¶ 1
= [Ylm (θo , φo )q̃lm exp(−imωt) + Yl,−m (θo , φo )q̃l,−m exp(+imωt)]
4π²0 2l + 1 rol+1
l,m≥0

1 X µ 4π ¶ 1 £ ¤
∗ ∗
= l+1
Ylm (θo , φo )q̃lm exp(−imωt) + Yl,m (θo , φo )q̃l,m exp(+imωt)
4π²0 2l + 1 ro
l,m≥0
X µ ¶
1 4π 1
= l+1
Re {Ylm (θo , φo ) 2q̃lm exp(−imωt)}
4π²0 2l + 1 ro
l,m≥0
We imply that for m = 0 the factor 2 in front of q̃lm is dropped. In the complex quantity of the last line
only positive-frequency components mω with m ≥ 0 occur, as required. By inspection of the result we see
that the multipole moments Qlm suited for Chapter 9 are


 2q̃lm , m>0
Qlm = q̃l,0 , m=0

0 , m<0

with corresponding frequencies mω. Since static distributions don’t radiate, the case m = 0 is quite irrelevant.

b): At fixed location x, we perform a discrete temporal Fourier transform,

∞
X
ρ(x, t) = cn (x)fn (t)
n=−∞

RT
with basis functions fn (t) = √1T exp(−inωt), where T = 2π
ω . Note the orthonormality t=0
fn∗ (t)fm (t)dt =
P∞
δnm and the closure n=−∞ fn∗ (t0 )fn (t) = δ(t − t0 ).

Then,

Z T Z T
1
cn (x) = fn∗ (t)ρ(x, t)dt = √ exp(inωt)ρ(x, t)dt
t=0 T t=0

Noting c−n (x) = c∗n (x),

∞
X
ρ(x, t) = c0 f0 + [cn (x)fn (t) + c−n (x)f−n (t)]
n=1
X∞
= c0 f0 + [cn (x)fn (t) + c∗n (x)fn∗ (t)]
n=1
X∞
= c0 f0 + Re [2cn (x)fn (t)]
n=1
Z ∞
"( Z ) #
1 T X 2 T
= ρ(x, t)dt + Re ρ(x, t) exp(inωt)dt exp(−inωt) (1)
T 0 n=1
T 0

Note that all frequencies are positive. By inspection we see that the charge densities to be used in Eq. 9.1 ff
are
( RT
2
ρ(x, t) exp(inωt)dt , n > 0
ρ(x) = T 0
1
RT ,
T 0
ρ(x, t)dt , n=0

where the frequencies nω are all positive, as required. Since static distributions don’t radiate, the case n = 0
is quite irrelevant. The multipole moments for frequency component nω with n > 0 are
 Z Z T
2
Qlm = d3 x dtρ(r, θ, φ − ωt)rl Ylm
∗
(θ, φ) exp(inωt)
T 0
Z Z T
2
= d3 x dtρ(r, θ, φ)rl Ylm
∗
(θ, φ + ωt) exp(inωt)
T 0
Z Z T
2 3
= d x dtρ(r, θ, φ)rl Ylm∗
(θ, φ) exp(−imωt) exp(inωt)
T 0
½ Z ¾
= δmn 2 ρ(r, θ, φ)rl Ylm
∗
(θ, φ)d3 x

For n = 0, drop the factor 2. This result is equivalent to that of part a).

c): We have already generally shown that both methods a) and b) lead to identical multipole moments that
are simply related to the multipole moments in the body frame. For a charge q located at (R, θ = π/2, φ = φ0 )
rotating about the z-axis with frequency ω0 , the body-frame charge density in spherical and cylindrical
coordinates is

δ(r − R) δ(r − R)
ρ(x) = q δ(cos θ)δ(φ − φ0 ) = q δ(z)δ(φ − φ0 )
R2 R

and it is

Monopole moment:

Spherical: Q00 = √q . Cartesian: Q = q. Since m = 0, the frequency of the monopole moment is zero. This
4π
is generally the case, and explains why monopole moments do not occur in radiation problems.

Dipole moment:

Spherical:

Z r
∗ 3
Q11 = 2 rY11 ρ(x)d3 x = −2qR exp(−iφ0 )
8π
Q10 = 0
Q1−1 = 0 (2)

The frequency of Q11 is mω0 = ω0 .

Cartesian: Use Eq. 4.5 in Jackson to find

Q11 − Q1,−1
px = q = qR exp(−iφ0 )
3
−2 8π
Q11 + Q1,−1
py = q = iqR exp(−iφ0 )
3
2i 8π
 Q
pz = q 10 = 0
3
4π

Thus, p = qR exp(−iφ0 )(x̂ + iŷ). The frequency is ω0 , and the dipole moment with explicitly displayed time
dependence is

p = qR exp(−iφ0 )(x̂ + iŷ) exp(−iω0 t)

Note. One may choose the time origin such that the global phase term exp(−iφ0 ) becomes 1.

Note. It is still instructive to obtain the cartesian moments by first calculating the harmonic charge densities
and then their moments. With

δ(r − R)
ρ(x, t) = q δ(z)δ(φ − φ0 − ω0 t)
R

Frequency zero:

Z T
1 δ(r − R) q δ(r − R) q δ(r − R)
ρ0 (x) = q δ(z)δ(φ − φ0 − ω0 t)dt = δ(z) = δ(z)
T 0 R T ω0 R 2π R

which has a zero-frequency cartesian monopole moment, Q = q.

n-th harmonic frequency:

Z T
2 δ(r − R) q δ(r − R)
ρ(x) = q δ(z)δ(φ − φ0 − ω0 t) exp(inω0 t)dt = δ(z) exp(in(φ − φ0 ))
T 0 R π R

from which we can see, for instance, that the electric-dipole components are radiating at the fundamental
(as is generally the case),

Z
q δ(r − R)
px = exp(−inφ0 ) r cos φ δ(z) exp(inφ)rdrdzdφ = qR exp(−iφ0 )δn,1
π R
Z
q δ(r − R)
py = exp(−inφ0 ) r sin φ δ(z) exp(inφ)rdrdzdφ = iqR exp(−iφ0 )δn,1
π R
pz = 0 (3)

Higher moments. As explained above, only moments with m > 0 exist. For m > 0, the spherical moments
have oscillation frequencies mω0 and magnitudes

Z
δ(r − R)
Qlm = 2q rl ∗
δ(cos θ)δ(φ − φ0 )Ylm (θ, φ)r2 drd cos θdφ
R2
 = 2qRl Ylm
∗
(π/2, φ0 )
s
2l + 1 (l − m)! m
= 2qRl exp(−imφ0 ) P (0)
4π (l + m)! l

l+m (l+m)!
Thereby, for even l − m it is Plm (0) = (−1) 2
2l ( l−m
, while for odd l − m it is Plm (0) = 0. Thus,
2 )!( 2 )!
l+m

radiation occurs at all harmonic frequencies mω0 . The lowest-order non-zero multipole at frequency mω0 is
Ql=m,m . Non-zero higher-order multipoles at frequency mω0 are Ql=m+2,m , Ql=m+4,m etc.
2. Problem 9.2 10 Points

According to Problem 9.1, it is


 2q̃lm , m>0
Qlm = q̃l,0 , m=0

0 , m<0

where the frequencies are mω and

Z
q̃lm = rl Ylm
∗
(θ, φ)ρ(r, θ, φ(t = 0))d3 x

Here,

q
ρ(x, t = 0) = δ(r − R)δ(cos θ) [δ(φ) + δ(φ + π) − δ(φ + π/2) − δ(φ + 3π/2)]
R2

and

Z
q
Qlm = 2 rl+2 ∗
δ(r − R)δ(cos θ) [δ(φ) + δ(φ + π) − δ(φ + π/2) − δ(φ + 3π/2)] Ylm (θ, φ)drd cos θdφ
R2
= 2qRl [Ylm
∗ ∗
(0, 0) + Ylm ∗
(0, π) − Ylm (0, π/2) − Ylm∗
(0, 3π/2)]
s
2l + 1 (l − m)! m
= 2qRl P (0) [1 + exp(imπ) − exp(imπ/2) − exp(im3π/2)]
4π (l + m)! l
s " #
l 2l + 1 (l − m)! l+m (l + m)!
= 2qR (−1) 2 l ¡ l−m ¢ ¡ l+m ¢ × 4 (4)
4π (l + m)! 2 2 ! 2 !

where for the result to be different from zero it must be both l − m even and m = 2 + 4p with integer
p = 1, 2, 3... Thus, the lowest non-zero moments are Q22 , Q42 , Q62 , ... and Q66 , Q86 , ... Also, there is no
magnetic dipole moment, because the net circular current is zero. Thus, in the long wavelength limit the
leading radiation term comes from Q22 , which radiates at a frequency 2ω and has a value, following the
above formula, of

r
2 30
Q22 = qR
π

√
Since the sidelength a = R 2, it also is

r
2 15
Q22 = qa
2π

There are no other non-zero spherical quadrupole moments.

The quadrupole radiation field in the far field is given by Eqs. 9.169 and 9.149 (applied in that order),

r r r
−i3 ck 4 3 2 15 qa2 ck 3 1
H = exp(ikr − 2iωt) qa X22 = exp(ikr − 2iωt) X22
ikr3 × 5 2 2π r 20π
E = Z0 H × n

The radiation pattern is, following Eqn. 9.151,

dP Z0 2 2
= 2 |aE (2, 2)| |X22 |
dΩ 2k

q q
ck4 3 2 15
where aE (2, 2) = i 15 2 qa 2π . Using further that k = 2ω/c and the table 9.1 one finds

dP Z0 ω 6 q 2 a4
= (1 − cos4 θ)
dΩ 2π 2

This result can also be obtained directly from the fields, because

dP r2
= Re [n̂ · (E × H∗ )]
dΩ 2

This may be integrated, or one may use Eq. 9.154, to find

Z
dP Z0 2 8Z0 ω 6 q 2 a4
P = dΩ = 2 |aE (2, 2)| =
4π dΩ 2k 5πc4

Note. From Q22 and Eqns. 4.6 one may also derive the cartesian quadrupole tensor

 
1 i 0
Q = 3qa2  i −1 0  ,
0 0 0

and then use Eqs. 9.45 and 9.49 to arrive at the same result. This method is less elegant, however.
3. Problem 9.3 10 Points

We show that there is a non-zero electric-dipole moment. From that it follows that the leading radiation
term in the long-wavelength approximation is the electric-dipole radiation.

We also show that the magnetic-dipole moment is zero. This step is not really necessary, because electric-
dipole radiation dominates magnetic-dipole radiation of the same order by a factor of order (kd)−2 .

From Eq. 3.38 in Jackson we see that in the near field the scalar potential produced by the hemispheres is

(2j − 1/2)Γ(j − 1/2) ³ a ´2j

∞
V (t) X
Φ(r, θ, t) = √ (−1)j+1 P2j−1 (cos θ)
π j=1 j! r

The electric-dipole term of that corresponds to j = 1, i.e.

µ ¶2 ½ ¾ ½ ¾
V R 3V R2 p
ΦE1 (r, θ, t) = √ (3/2)Γ(1/2) (cos θ) cos(ωt) = Re cos θ exp(−iωt) = Re cos θ exp(−iωt)
π r 2r2 4π²0 r2

We thus see by comparison that the complex dipole moment vector at frequency ω is

p = 6V R2 π²0 ẑ

In the long-wavelength = small-source approximation, a non-vanishing electric-dipole moment produces the

dominant radiation. Eqns. 9.19, 9.23 and 9.24 then yield, in the radiation zone,

3V exp(ikr)
H = − (kR)2 sin θφ̂
2Z0 r
3V exp(ikr)
E = − (kR)2 sin θθ̂
2 r
dP 9V 2
= (kR)4 sin2 θ
dΩ 8Z0
V2
P = 3π(kR)4
Z0

ω
where k = c.

To show that the magnetic-dipole moment is zero, we note that for symmetry reasons the surface current
has even spatial parity, i.e. K(θ, φ) = K(π − θ, φ + π). Thus,

Z 2π Z 1
1
m = x × K(θ, φ)R2 dφd cos θ
2 φ=0 cos θ=−1
Z 2π Z 1
1
= [x × K(θ, φ) + (−x) × K(π − θ, φ + π)] R2 dφd cos θ
2 φ=0 cos θ=0
 Z 2π Z 1
1
= [x × K(θ, φ) − x × K(θ, φ)] R2 dφd cos θ
2 φ=0 cos θ=0
= 0
4. Problem 9.5 10 Points

a): For A(x), copy Eqns. 9.13-9.16 of Jackson. For Φ(x), write the analogue of Eq. 9.30 for Φ(x),

µ ¶Z µ ¶
1 exp(ikr) 1 1 exp(ikr) 1
Φ(x) = − ik ρ(x0 )n̂ · x0 d3 x0 = − ik n̂ · p
4π²0 r r 4π²0 r r

b):

µ ¶
iµ0 ω exp(ikr)
B = ∇×A=− ∇× p
4π r
·µ ¶ ¸
iµ0 ω exp(ikr) exp(ikr)
= − ∇ ×p+ (∇ × p) =
4π r r
µ · ¸¶
iµ0 ω exp(ikr) 1
= − n̂ ik − 2 ×p
4π r r
· ¸
ck 2 µ0 exp(ikr) 1
= 1− (n̂ × p) (5)
4π r ikr

One way to obtain E is

∂
E = − A − ∇Φ
∂t ( " õ !#
¶2 · µ ¶ µ ¶¸ · µ ¶ µ ¶¸)
1 exp(ikr) 2 1 1 1 1 1 1
= k p + r̂ r̂ · p − ik + 2 − θ̂ θ̂ · p − ik − φ̂ φ̂ · p − ik
4π²0 r r r r r r r

where we have used ∂θ (r̂ · p) = p · θ̂ and ∂φ (r̂ · p) = (p · φ̂) sin θ. The result simplifies to

µ ¶n o
1 exp(ikr) 2 1 exp(ikr) 1
E = k {p − pr r̂} − − ik θ̂pθ + φ̂pφ − 2r̂pr
4π²0 r 4π²0 r2 r

Noting that r̂ = n̂, the first curly bracket equals (n̂ × p) × n̂ and the second p − 3n̂(n̂ · p) we find the final
result,

½ µ ¶ ¾
1 2 exp(ikr) 1 ik
E = k (n̂ × p) + [3n̂(n̂ · p) − p] − 2 exp(ikr) (6)
4π²0 r r3 r