Ben Sauerwine

Classical Electrodynamics Homework 7

1) Consider the electrostatic potential in the two-dimensional pie-shaped region

shown:

The potential on the edges is zero. Find the Green’s function for this problem.
v v
Mote that the Green’s function G (x , x ') is defined by
v v v v
∇ 2x G (x , x ') = δ 2 (x − x ')
∂2 ∂2 1 ∂ ⎛ ∂ ⎞ 1 ∂2
∇ 2x = + = ⎜r ⎟ +
∂x 2 ∂y 2 r ∂r ⎝ ∂r ⎠ r 2 ∂θ 2
v
First, I need solutions to the homogeneous equation ∇ 2x u ( x ) = 0 . I recall from last
week that solutions to the angular equation look like u (θ ) = e imx , which to satisfy the
boundary condition along the x-axis and at alpha I will write as
⎛ πm ⎞
u (θ ) = C m sin ⎜ θ ⎟ m > 0 in order to satisfy the boundary conditions where I have
⎝α ⎠
neglected normalization for the time being (it is some constant C m that I’ll fill in
later). Since sine forms a complete set over odd functions and the m = 0 case is not
⎛ πm ⎞
2

an option for sine, I may take m > 0 only. Let k m = ⎜ ⎟ . Now I recall the
⎝α ⎠
⎛ ∂ ∂2 2⎞
solutions to the equation ⎜⎜ ρ + ρ2 − k m ⎟⎟v m = 0 from last week: namely,
⎝ ∂ρ ∂ρ 2
⎠
v m = Aρ km + Bρ − km for k m ≠ 1 .
vm = C + A ln ρ for k m = 0 = m .
1+ ρ 2 ρ 2 −1
vm = A +B for k m = 1
2ρ 2ρ
Now applying my boundary conditions to the problem, I have:
v (< ) m = Am ρ km ⎫
⎪
⎬ 0 ≠ km ≠ 1
v ( > ) m = Bm ρ k m
R kn − km
− − kn ρ
R
[
= Bm ρ k m − R 2 k n ρ − k m ]
⎪⎭

v (< ) m = 0 ⎫⎪
LnR − Lnρ ⎬ 0 = m = k m
v (> ) m = Bm
LnR ⎪⎭ .

v (< ) m = Am ρ ⎫
⎪
(> ) ρ 2 − R2 ⎬ 1 = km
v m = Bm
(
ρ 1+ R2 ) ⎪
⎭
Now I will hybridize an angular eigenfunction expansion with a radial jump
discontinuity. First I will normalize the angular portion:
α
⎛α ⎞ 1
⎛α ⎞
∫ sin 2 ⎜ θ ⎟dα = α m ≠ 0
⎝ πm ⎠ 2
u (θ ) = C m sin ⎜ θ ⎟ m > 0 0
⎝ πm ⎠ 2
Cm =
α
Now I need the Wronskian of my radial solutions. I will take each separately:

A[v< ' v> − v> ' v< ] =

1
ρ
v (< ) m = Am ρ k m ⎫
⎪
⎬ 0 ≠ km ≠ 1
(> )
v m = Bm ρ − − k n ρ
km

R
R kn −km
[
= Bm ρ k m − R 2 k n ρ − k m ]
⎪⎭
2 R 2 kn k n 1
v< ' v > − v > ' v< = − =
ρ Aρ
v (< ) m = 0 ⎫⎪
LnR − Lnρ ⎬ 0 = m = k m
v ( > ) m = Bm
LnR ⎪⎭
v< ' v> − v > ' v< = 0 (no valid m = 0 solutions )
v (< ) m = Am ρ ⎫
⎪ 2R 2 1
ρ 2 − R2 ⎬ 1 = km v< ' v > − v > ' v< = − =
(> )
v m = Bm
(
ρ 1+ R2 ) ⎪ (2
)
1 + R ρ Aρ
⎭
Finally, I aggregate these into one function.
Assume k m ≠ 1 . If such a value does arise, it will be necessary to insert the special
solution and Wronskian shown above for that particular element in the eigenfunction
expansion.
⎛ ⎞⎛ πm ⎞⎛ πm 2πm πm
⎞
2 ∞ ⎛ πm ⎞ ⎛ πm ⎞⎜ α ⎟⎜ α ⎟⎜ α −
G (θ , ρ ;θ ' , ρ ') = ∑ sin ⎜ θ ⎟ sin ⎜ θ ' ⎟⎜ − ρ
πm ⎟⎜ < ⎟⎜ >
ρ − R ρ > α ⎟⎟
α
α m =1 ⎝ α ⎠ ⎝ α ⎠⎜ ⎟
⎝ 2πmR α ⎠⎝ ⎠⎝ ⎠
2

2) Suppose an electrostatic potential φ ( x ) satisfies the following differential

∂ 2φ
equation: = −α 0δ ( x ) + α 1δ ' (x )
∂x 2
The term − α 0δ ( x ) represents an infinitely extended charge distribution in
the yz-plane while α 1δ ' (x ) represents a sheet of dipole. Find the general
solution to this equation.

∂ 2φ
The general solution to = 0 is, of course φ ( x ) = cx + d . Now writing the
∂x 2
Green’s function for this at zero, I have the To this I need to add a particular
solution. Writing:
A[φ <φ > '−φ.>φ < '] = 1 c = 1 d = 0
.
G (x; x') = x< = min ( x, x') = xθ ( x − x') + x'θ ( x'− x )
Now I use Green’s function to invert the original differential equation:
∂ 2φ
= −α 0δ ( x ) + α 1δ ' (x ) . The solutions will be:
∂x 2
φ0 (x ) = cx + d
∞
φ ( x ) = ∫ G(x, x')[− α 0δ (x') + α1δ ' ( x')]dx' + φ0 (x )
−∞
∞
= ∫ x[− α 0δ (x')θ (x − x') + α1δ ' (x')θ (x − x')] + x' [− α 0θ (x'− x )δ (x') + α1θ ( x'− x )δ ' (x')]dx' + φ0 ( x )
−∞
I’d like to know what these compositions of delta-function and Heaviside
functions mean. I know that
∞ ∞ ∞
−i
δ (x ) = ( ) ( )
1 i 1 −ikx
∫ δ ∫ θ ∫
− ikx −ikx
e dk ' x = ke dk x = e dk
2π −∞ 2π −∞ 2π −∞ k
Now I write
 ∞ ∞ ∞ ∞

∫−∞δ (x')θ (x − x')dx' = −∫∞δ (x') 2π dk = θ ( x )

i 1 −ik ( x− x ') i 1
∫−∞ k e ∫ke
−ikx
dkdx' =
2π −∞
∞ ∞
δ ' (x')θ (x − x') = ∫ δ ' (x')
i 1 −ik ( x − x ' )

−∞
2π ∫ke
−∞
dkdx' =

dv = δ ' ( x')dx' v = δ (x') u = e −ik ( x − x ') du = ike−ik ( x − x ') dx'

i 1 ⎡ −ik ( x − x ') ⎤
∞ ∞
δ (x') −∞ − ∫ δ (x')ike−ik ( x− x ') dx'⎥ dk
∞
=∫ ⎢e
−∞
2π k ⎣ −∞ ⎦
∞ ∞
=∫
i 1 −ikx
2π k
[
e − ike−ikx dk = ∫ ⎢
2π ⎣ k
]
1 ⎡ i −ikx ⎤
e + e −ikx ⎥ dk = θ ( x ) + δ (x )
−∞ −∞ ⎦
∞ ∞
x'θ (x'− x )δ ( x') = ∫ x' δ ( x')
i 1 −ik ( x ' − x )

−∞
2π ∫ke
−∞
dkdx' = 0

∞ ∞
x'θ (x'− x )δ ' ( x') = ∫ x' δ ' (x')
i 1 −ik ( x ' − x )

−∞
2π −∞
∫ke dkdx' =

dv = δ ' ( x')dx' v = δ (x') u = x' e −ik ( x '− x ) ( )

du = − ikx' e −ik ( x '− x ) + e −ik ( x '− x ) dx'
i 1 ⎡ −ik ( x '− x ) ⎤
∞ ∞
= ∫−∞ 2π k ⎢⎣ x ' e δ ( x ')
∞

−∞
− ∫ δ ( (
x') − ikx' e −ik ( x '− x ) + e −ik ( x '− x ) dx'⎥ dk )
−∞ ⎦
sgn( x )
∞
i 1 ikx
=−∫ e dk = −
−∞
2π k 2
Above, sgn represents the signum function. Now that I’ve done the hard part, all
that’s left is for me to make some replacements in my integral:
∞
φ (x ) = ∫ x[− α 0δ (x')θ (x − x') + α 1δ ' (x')θ (x − x')] + x' [− α 0θ (x'− x )δ (x') + α1θ (x'− x )δ ' ( x')]dx' + φ 0 (x )
−∞

= − xα 0θ ( x ) + xα1 (θ ( x ) + δ ( x )) − α1 sgn(x ) + φ 0 ( x )
1
2
= − xα 0θ ( x ) + xα1θ ( x ) − α 1 sgn(x ) + cx + d
1
2
I’d test this function by plotting its second derivative, but I don’t know what the
first derivative of the delta-function means! Thus, I’ll have to take this on faith.

3) (Jackson 2.7) Consider a potential problem in the half-space defined by

z ≥ 0 , with Dirichlet boundary conditions on the plane z = 0 (and infinity).
v v
(a) Write down the appropriate Green function G (x , x ') .
v
At infinity, I will let my potential go to zero. Now writing ∇ 2φ = f ( x ) and thinking
carefully, perhaps I can find a simpler form. Looking at Jackson formula (2.3), I see that
for Dirichlet boundary conditions, he has simply built a solution around the identity
⎛ 1 ⎞ v v
∇ 2 ⎜⎜ v v ⎟⎟ = −4πδ ( x − x ') by simply building a zero at the boundaries and ensuring that
⎝ x − x' ⎠
the function to transform is zero in the region corresponding to the subtracted portion.
v v 1 ⎡ ⎤
Now I see that G (x , x ') = −
1 1
⎢v − v ⎥.
4π ⎢ x − xˆx' x − xˆx' y − xˆx' z x − ˆx' x − xˆx' y + xˆx' z ⎥
x
⎣ ⎦
(b) If the potential on the plane z = 0 is defined to be φ = V inside a circle of
radius a centered at the origin and φ = 0 outside that circle, find an integral
expression for the potential at the point P specified in terms of cylindrical
coordinates.
v v v v v v v v v
Use φ (r ) = ∫ G (r , r ') f ( x )dV + ∫ φ (r ')∇' G (r , r ') ⋅ dS , from the notes. Assuming that there
V D
v v v v v v
is no charge density in the main volume, I have φ (r ) = ∫ φ (r ')∇' G (r , r ') ⋅ dS , and I only
D
need the divergence of the density function at the border.
v v v 1 v ⎡ ⎤
∇ ' = G ( x , x ') = −
1 1
∇' ⎢ v − v ⎥
4π ⎢ x − xˆx' x − xˆx' y − xˆx' z x − ˆx' x − xˆx' y + xˆx' z ⎥
x
⎣ ⎦
⎡ ⎤
1 v ⎢ 1 1 ⎥
=− ∇' −
4π ⎢ (x − x' ) + (x − x' ) + (x − x' )
2 2 2
(x x − x' x ) + (x y − x' y )2 + (x z + x' z )2
2 ⎥
⎣ x x y y z z ⎦
⎡ ⎤
2 ⎛ − 1 ⎞ v ⎢ − (x x − x' x )xˆ − (x y − x' y )yˆ − (x z − x' z )zˆ − (x x − x' x )xˆ − (x y − x' y )yˆ + (x z + x' z )zˆ ⎥
=− ⎜ ⎟∇' −
4π ⎝ 2 ⎠ ⎢ (x − x' )2 + (x − x' )2 + (x − x' )2 3 (x x − x' x )2 + (x y − x' y )2 + (x z + x' z )2 ⎥⎦
3⎥
⎢⎣ x x y y z z

Now evaluating this at the border where x' z = 0 and taking the surface normal ẑ , I have
⎡ ⎤
1 ⎢ − ( x x − x' x )xˆ − (x y − x' y )yˆ − ( x z )zˆ − ( x x − x' x )xˆ − (x y − x' y )yˆ + ( x z )zˆ ⎥ ⋅ zˆ
−
4π ⎢ 3 ⎥
(x x − x' x )2 + (x y − x' y ) + ( x z − x' z ) (x x − x' x ) + (x y − x' y ) + ( x z + x' z )
3
2 2 2 2 2
⎢⎣ ⎥⎦
2 xz
=
4π
(x x − x' x )2 + (x y − x' y ) + ( x z )
3
2 2

So now integrating over the surface, I have

v v v v v v 1 ∞ 2π v xz
φ (r ) = ∫ φ (r ')∇' G (r , r ') ⋅ dS = ∫ ∫ φ (r ') dS
2π 0 0 (x x − x' x ) + (x y − x' y ) + (x z )
3
2 2 2
D

a
.
xz
=V∫ dS
( x x − x' x ) + (x ) + (x )
3
− x' y
2 2 2
0
y z

(c) Show that, along the axis of the circle, the potential is given by
v ⎛ ⎞
φ (r ) = V ⎜⎜1 −
z
⎟.
2 ⎟
⎝ a +z ⎠
2

Using my integral from part b,

 a
xz ρ ⎛ −x ⎞
V∫ dρ = V ⎜ z
+ 1⎟
ρ 2 + x z2 ⎜ 2 ⎟
⎝ a + xz
2
0
⎠
(d) Show that at large distances, e.g. ρ 2 + z 2 >> a 2 , the potential can be
expanded in a power series in ρ 2 + z 2 ( ) −1
, and that the leading terms are

φ=
Va 2 z ⎡
⎢1 −
3a 2
+
(
5 3ρ 2 a 2 + a 4
+
⎤
⎥
)
2
+z (ρ 2
3
2 2
) (
⎣⎢ 4 ρ + z
2 2
)
8 ρ 2 + z2
2
(
...
⎥⎦ )
Recalling my integral from part b,
2π a 2π a
v xz ρ ' xz ρ '
φ (r ) =
V V
∫ ∫ dρ 'dθ ' = ∫∫ dρ ' dθ '
2π 0 0
(
(x x − x' x )2 + (x y − x' y )2 + x z 2 2
2π
) (ρ )
3 3
+ ρ ' −2 ρρ ' cosθ '+ x z
0 0 2 2 2 2

as part of the problem statement, I have ρ 2 + z 2 >> a 2 , ρ ' < a , I may expand in ρ ' = a
2π a 2π a
v xz ρ ' xz ρ '
φ (r ) ≈
V V 1
∫∫ dρ ' dθ ' = ∫∫ dρ ' dθ '
2π
(ρ ) 2π
(ρ )
3 3 3
+ ρ ' −2 ρρ ' cosθ '+ x z + xz ⎛ θ ' ⎞⎟
⎜1 + ρ ' −2 ρρ ' cos
0 0 2 2 2 2 2 2 2 0 0 2 2

⎜ ρ 2 + xz 2 ⎟
⎝ ⎠
Expanding this in ρ 2 + z 2 >> a 2 , I have

One way to see this is:

2
3 ⎡ ρ ' 2 −2 ρρ ' cos θ ' ⎤ 15 ⎡ ρ ' 2 −2 ρρ ' cos θ ' ⎤
2π a
v xz
φ (r ) ≈
V
∫0 ∫0 2 ⎢⎢ ρ 2 + x 2 ⎥⎥ + 8 ⎢⎢ ρ 2 + x 2 ⎥⎥ − ... p' dρ ' dθ '
1 −
2π
(ρ )
3
2
+ xz
2 2 ⎣ z ⎦ ⎣ z ⎦
2π

∫ cos θ dθ = π
2

3 ⎡ ρ ' 2 ⎤ 15 ρ ' 4 +2 ρ 2 ρ ' 2

a
V xz
= 2π ∫ 1 − ⎢ ⎥+ − ... p ' dρ '
2π
(ρ 2
+ xz )
3
2 2 0 (
2 ⎣⎢ ρ 2 + x z 2 ⎦⎥ 8 ρ 2 + x 2 2
z )
3 ⎡ a2 ⎤ 5 a 4 + 3ρ 2 a 2
a
Va 2 xz
= ∫1 − ⎢ 2 ⎥+
(
− ... p ' dρ '
)
2ε 0
(ρ ) 4 ⎣⎢ ρ + x z 2 ⎦⎥ 8 ρ 2 + x 2 2
3
+ xz
2 2 2 0
z

or in Mathematica:
 Va 2 z
This gives my desired result once I factor out , and some rearrangement of
( )
3
2ρ +z 2 2 2

terms. Notice that I could have just as well written:

To verify that this is indeed my desired result, I allow the radial component of this to go
to zero:

and simply compare this to the expansion of my earlier result:

And I see that they agree.

4) (Jackson 2.19) Show that the two-dimensional Green function for Dirichlet
boundary conditions for the annular region, b ≤ ρ ≤ c (concentric cylinders)
has the expansion
⎛ ρ 2 ⎞ ⎛ c2 ⎞
ln⎜⎜ <2 ⎟⎟ ln⎜⎜ 2 ⎟⎟
b ⎠ ⎝ ρ> ⎠ ∞
cos[m(φ − φ ')] ⎛⎜ m b 2 m ⎞⎟⎛⎜ 1 ρ > m ⎞⎟
G= ⎝ + 2 ∑ ρ − −
⎡ ⎛ b ⎞ 2 m ⎤ ⎜⎝ ρ < m ⎟⎠⎜⎝ ρ > m c 2 m ⎟⎠
<
⎛ c2 ⎞ m =1
ln⎜⎜ 2 ⎟⎟ m ⎢1 − ⎜ ⎟ ⎥
⎝b ⎠ ⎢⎣ ⎝ c ⎠ ⎥⎦
I see from the form of the Green’s function that there is no dependence on the z
component included here, and indeed if one were to be included it should be added using
the jump discontinuity method with values at positive and negative infinity zero so that
⎡ 1 ∂ ⎛ ∂ ⎞ 1 ∂2 ∂2 ⎤
⎢ ⎜⎜ ρ ⎟⎟ + 2 + ⎥φ = 0
⎣ ρ ∂ρ ⎝ ∂ρ ⎠ ρ ∂θ ∂z 2 ⎦
2

∞
φ = A(ρ )B(θ )C (z ) B(θ ) = ∑C m e imθ C ( z ) = cz + d no dependence
m = −∞

⎡ 1 ∂ ⎛ ∂ ⎞ m2 ⎤
⎢ ⎜⎜ ρ ⎟⎟ − 2 ⎥φ = 0
⎣ ρ ∂ρ ⎝ ∂ρ ⎠ ρ ⎦
Stealing my results from last week, I have:
Am (ρ ) = Aρ m + Bρ − m , and I must kill either boundary. To do this I take:
A< ( ρ ) = ρ m − b 2 m ρ − m A> ( ρ ) = c −2 m ρ m − ρ − m , and computing the Wronskian gives
 AW = A[ A< A> '− A< ' A> ]
1
AW =
ρ
[ ]
AW = − A (mρ m −1 + mb 2 m ρ − m −1 )(c − 2 m ρ m − ρ − m ) − (mc − 2 m ρ m −1 + mρ − m −1 )(ρ m − b 2 m ρ − m )

= 2mA
(1 − b 2m
c −2m )
A=−
1
ρ 2m(b c − 2 m − 1)
2m

except in the case where m = 0, where I have instead:

Am ( ρ ) = c + d ln ρ , and to kill the boundaries I take
A< ( ρ ) = ln b − ln ρ A> ( ρ ) = ln c − ln ρ . (Note here that I could choose any integer
coefficient I like on the logarithms: this yields the freedom to choose any power in the
final solution. I don’t like squares that Jackson uses, so I have done it linearly.) This
time computing the Wronskian gives
AW = A[ A< A> '− A< ' A> ]
1
AW =
ρ
⎡⎛ 1 ⎞ ⎛ 1⎞ ⎤ ⎛ 1 ⎞⎛ b ⎞
AW = − A⎢⎜⎜ − ⎟⎟(ln c − ln p ) − ⎜⎜ − ⎟⎟(ln b − ln p )⎥ = − A⎜⎜ ⎟⎟⎜ ln ⎟
⎣⎝ ρ ⎠ ⎝ ρ⎠ ⎦ ⎝ ρ ⎠⎝ c ⎠
1
A=−
⎛b⎞
ln⎜ ⎟
⎝c⎠
Now finally taking the normalization of my angular solutions, I get
∞
B(θ ) = ∑C m e imθ , but I notice that in this normalization, this linear combination of
m = −∞
solution gives a zero normalization factor, indicating that I must split this up into cosine
and sine portions. I will take only the cosine portion in order to leave the m = 0 case
∞ ∞
intact. Doing so gives me B(θ ) = ∑ C m cos[mθ ] + ∑ S n sin[nθ ] , and normalization gives
m =0 n =1
1
me a factor of C 0 = 1 C m = S m = . Combining these, then, gives
π
 ⎛ρ ⎞ ⎛ρ ⎞
ln⎜ < ⎟ ln⎜ > ⎟
G (θ , ρ ;θ ' , ρ ') = − ⎝ ⎠ ⎝ ⎠
b c
⎛b⎞
2π ln⎜ ⎟
⎝c⎠
∞
−1 ⎛ m b 2 m ⎞⎛ ρ > m ⎞
+ ∑ [cos(mθ ) cos(mθ ') + sin (mθ )sin (mθ ')]
1 ⎜ ρ < − − m ⎟⎜ 2 m − 1− m ⎟
π
m =1 ⎛ ⎛ b ⎞ 2m ⎞ ⎜⎝ ρ < ⎟⎠⎜⎝ c ρ > ⎟⎠
2m⎜ ⎜ ⎟ −1⎟
⎜⎝ c ⎠ ⎟
⎝ ⎠
⎛ρ ⎞ ⎛ c ⎞
ln⎜ < ⎟ ln⎜⎜ ⎟⎟
⎝ b ⎠ ⎝ ρ > ⎠ ∞ cos(m(θ − θ ')) ⎛⎜ m b 2 m ⎞⎟⎛⎜ ρ > 1 ⎞⎟
m

=− −∑ ρ − −
⎛ ⎛ b ⎞ 2 m ⎞ ⎜⎝ ρ < ⎟⎠⎜⎝ c ρ > −m ⎟⎠
< −
⎛c⎞ m 2 m
2π ln⎜ ⎟ m =1
2mπ ⎜1 − ⎜ ⎟ ⎟
⎝b⎠ ⎜ ⎝c⎠ ⎟
⎝ ⎠

Note the places where I disagree with Jackson here: one is the overall sign, one is in the
term with the logarithms. It turns out that it really makes no difference had I chosen
[ ]
Log ρ any power as my linearly independent logarithmic term, so I could raise each
element to any integer power I like there. The other point is where I have an additional 2
pi in the denominator along with m. My normalization factors definitely brought
1
a C0 = 1 Cm = S m = for that integral, and my Wronskian certainly brought a two
π
factor in the denominator.