Physics 505 Fall 2007

Homework Assignment #5 — Solutions

Textbook problems: Ch. 3: 3.13, 3.17, 3.26, 3.27

3.13 Solve for the potential in Problem 3.1, using the appropriate Green function obtained
in the text, and verify that the answer obtained in this way agrees with the direct
solution from the differential equation.
Recall that Problem 3.1 asks for the potential between two concentric spheres
of radii a and b (with b > a), where the upper hemisphere of the inner sphere
and the lower hemisphere of the outer sphere are maintained at potential V , and
where the other hemispheres are at zero potential. Since this problem involves
the potential between two spheres, we use the Dirichlet Green’s function

G(~x, ~x 0 )
l
a2l+1
  
X 4π 1 1 r>
= l
r< − l+1 − Ylm (Ω)Ylm ∗ (Ω0 )
2l + 1 1 − a 2l+1 r< l+1
r> b2l+1


l,m b
(1)
Because there are no charges between the spheres, the Green’s function solution
for the potential only involves the surface integral
Z
1 ∂G
Φ(~x ) = − Φ(~x 0 ) 0 da0
4π S ∂n
Here, we note a subtlety in that the boundary surface is actually disconnected,
and includes both the inner sphere of radius a and the outer sphere of radius b.
This means that the potential may be written as a sum of two contributions
Z Z
1 0 ∂G 2 0 1 ∂G
Φ(~x ) = − Φ(~x ) 0 a dΩ − Φ(~x 0 ) 0 b2 dΩ0 (2)
4π r0 =a ∂n 4π r0 =b ∂n
We now compute the normal derivatives of the Green’s function (1)
∂G  ∂G 
= − =
∂n0 r0 =a ∂r0 r< =r0 =a
 
  
4π X 1 a l+1  a l+1  r l
=− 2 − Ylm (Ω)Ylm ∗ (Ω0 )
a a 2l+1 r b b


1−
l,m b

and
∂G  ∂G 
= 0 =
∂n0 r0 =b ∂r r> =r0 =b

  
4π X 1 r l  a l  a l+1
=− 2
2l+1 − Ylm (Ω)Ylm ∗ (Ω0 )
b 1− a b b r
l,m b
Inserting these expressions into (2) yields

Φ(~x )
X Z 
1
 
a l+1  a l+1  r l 
0 m∗ 0 0
= Va (Ω )Yl (Ω )dΩ − Ylm (Ω)
a 2l+1 r b b


l,m 1− b
X Z 
1
 
r l  a l  a l+1 
+ Vb (Ω0 )Ylm ∗ (Ω0 )dΩ0 − Ylm (Ω)
a 2l+1 b b r


l,m 1− b

This is the general expression for the solution to the boundary value problem
where Va (Ω) is the potential on the inner sphere and Vb (Ω) is the potential on
the outer sphere.
For the upper/lower hemispheres problem, we note that azimuthal symmetry
allows us to restrict the m values to m = 0 only. In this case, the spherical
harmonic expansion reduces to a Legendre polynomial expansion

Φ(~x )
X Z 1    
2l + 1 a l+1  a l+1  r l
= Va (ζ)Pl (ζ)dζ 
2l+1  − Pl (cos θ)
−1 2 1 − ab r b b
l
X Z 
2l + 1
 
r l  a l  a l+1

+ Vb (ζ)Pl (ζ)dζ   − Pl (cos θ)
a 2l+1 b b r


l 2 1 − b

where ζ = cos θ0 . Since Va = V for ζ > 0 and Vb = V for ζ < 0, this above
expression reduces to
  
(2l + 1)V Nl
X a l+1  a l+1  r l
Φ(~x ) =   − Pl (cos θ)
a 2l+1 r b b


l 2 1− b

X (2l + 1)(−1)l V Nl  r l  a l  a l+1 

+   − Pl (cos θ)
a 2l+1 b b r


l 2 1 − b
X (2l + 1)V Nl  a l+1  a l+1  r l 
=   −
a 2l+1 r b b


l 2 1− b
  
l r l  a l  a l+1
+ (−1) − Pl (cos θ)
b b r

where
1 l=0
(
Z 1
Nl = Pl (ζ)dζ = Γ(j − 1 )
0
(−1)j+1 √ 2 l = 2j − 1 odd
2 πj!
 If desired, the potential may be rearranged to read
X (2l + 1)V Nl   a l   a l+1
l+1
Φ(~x ) =   1 + (−1)
a 2l+1 b r


l 2 1− b
  l+1   r l 
l l+1 a
+ (−1) 1 + (−1) Pl (cos θ)
b b
∞
"
V X (−1)j+1 (4j − 1)Γ(j − 12 )  a 2j−1   a 2j
= +V √   1+
2 a 4j−1 b r


j=1 4 πj! 1 − b
 #
 a 2j   r 2j−1
− 1+ P2j−1 (cos θ)
b b

which agrees with the solution to Problem 3.1 that we have found earlier.

3.17 The Dirichlet Green function for the unbounded space between the planes at z = 0
and z = L allows discussion of a point charge or a distribution of charge between
parallel conducting planes held at zero potential.
a) Using cylindrical coordinates show that one form of the Green function is

G(~x, ~x 0 )
∞ ∞  nπz   nπz 0   nπ 
4 X X im(φ−φ0 )  nπ 
= e sin sin Im ρ< Km ρ>
L n=1 m=−∞ L L L L

In cylindrical coordinates, the polar direction φ is periodic with period 2π. This
suggests that the Green’s function could be expanded as a Fourier series in eimφ .
Similarly, the boundary conditions G = 0 at z = 0 and z = L motivates the use of
a Fourier sine series sin(nπz/L) in the z coordinate. More precisely, a complete
Fourier expansion in φ and z would give
X 0 0
 nπz   n0 πz 
G(~x, ~x 0 ) = g(ρ, ρ0 )eimφ eim φ sin sin
L L
m,n,m0 ,n0

However, it turns out that m and m0 (and n and n0 ) do not need to be chosen
to be independent. This can be seen from the Green’s function equation (given
here as a differential equation in ~x )

∇x G(~x, ~x 0 ) = −4πδ 3 (~x − ~x 0 )

In cylindrical coordinates, this reads

1 ∂2 ∂2
 
1 ∂ ∂ 4π
ρ + 2 2 + 2 G(ρ, φ, z; ρ0 , φ0 , z 0 ) = − δ(ρ − ρ0 )δ(φ − φ0 )δ(z − z 0 )
ρ ∂ρ ∂ρ ρ ∂φ ∂z ρ
(3)
Using the completeness relations
∞
X 0
eim(φ−φ ) = 2πδ(φ − φ0 ) (4)
m=−∞

and
∞
X  nπz   nπz 0  L
sin sin = δ(z − z 0 )
n=1
L L 2

suggests that we take

∞ ∞  nπz   nπz 0 
X X 0
0
G(~x, ~x ) = g(ρ, ρ0 )eim(φ−φ ) sin sin (5)
n=1 m=−∞
L L

Substituting this decomposition into (3) gives

m2  nπ 2
 
1 d d 4
ρ − 2 − g(ρ, ρ0 ) = − δ(ρ, ρ0 )
ρ dρ dρ ρ L Lρ

Making the substitution

nπρ
x=
L
converts (the homogeneous part of) this to a modified Bessel equation

d2 m2
  
1 d 4
2
+ − 1+ 2 g(x, x0 ) = − δ(x, x0 )
dx x dx x Lx

At this stage, the solution becomes standard. Noting that the modified Bessel
function Im (x) blows up as x → ∞ and the function Km (x) blows up as x → 0,
we are left with
AIm (x) x < x0

0
g(x, x ) =
BKm (x) x > x0
where the coefficients A and B are determined by the matching conditions

d d 4
g< = g> , g< = g> +
dx dx Lx0

at x = x0 . This system may be solved to yield

4 Km (x0 )
A=
Lx0 Im
0 (x0 )K (x0 ) − I (x0 )K 0 (x0 )
m m m
0
4 Im (x )
B=
Lx0 Im
0 (x0 )K (x0 ) − I (x0 )K 0 (x0 )
m m m
 Noting that the modified Bessel functions satisfy the Wronskian formula

1
Iν (x)Kν0 (x) − Iν0 (x)Kν (x) = −
x
finally gives
4 Im (x)Km (x0 ) x < x0

0
g(x, x ) =
L Im (x0 )Km (x) x > x0
4
= Im (x< )Km (x> )
L
where
x< = min(x, x0 ), x> = max(x, x0 )
Converting x back to ρ and substituting into (5) then gives the desired Dirichlet
Green’s function
∞ ∞  nπz   nπz 0   nπρ 
0 4 X X im(φ−φ0 ) <
 nπρ 
>
G(~x, ~x ) = e sin sin Im Km
L n=1 m=−∞ L L L L

b) Show that an alternative form of the Green function is

∞ ∞
sinh(kz< ) sinh[k(L − z> )]
Z
X 0
0
G(~x, ~x ) = 2 dk eim(φ−φ ) Jm (kρ)Jm (kρ0 )
m=−∞ 0 sinh(kL)

This alternative form of the Green’s function is derived by expanding in φ and ρ

instead of φ and z. For the ρ expansion, we use the integral relation
Z ∞
1
kJν (kρ)Jν (kρ0 )dk = δ(ρ − ρ0 )
0 ρ

along with the completeness relation (4) to motivate the decomposition

∞ Z ∞
X 0
0
G(~x, ~x ) = kdk gk (z, z 0 )eim(φ−φ ) Jm (kρ)Jm (kρ0 ) (6)
m=−∞ 0

Since the Bessel function Jm (kρ) satisfies the Bessel equation

d2 m2
  
1 d 2
+ + k − 2 Jm (kρ) = 0
dρ2 ρ dρ ρ

the substitution of (6) into the Greens’ function equation (3) gives

d2
 
− k gk (z, z 0 ) = −2δ(z − z 0 )
2
dz 2
 Since gk (z, z 0 ) vanishes at z = 0 and z = L, this is a standard one-dimensional
Green’s function problem. Writing

z < z0

0 A sinh(kz)
gk (z, z ) =
B sinh[k(L − z)] z > z 0

we find that the matching and jump conditions become

2
A sinh(kz 0 ) = B sinh[k(L − z 0 )], A cosh(kz 0 ) = −B cosh[k(L − z 0 )] +
k
This may be solved to give
2 sinh[k(L − z 0 )] 2 sinh(kz 0 )
A= , B=
k sinh(kL) k sinh(kL)
so that
2
gk (z, z 0 ) = sinh(kz< ) sinh[k(L − z> )]
k sinh(kL)
Substituting this into (6) then yields
∞ k
sinh(kz< ) sinh[k(L − z> )]
Z
X 0
G(~x, ~x ) = 2 dk eim(φ−φ ) Jm (kρ)Jm (kρ0 )
m=−∞ 0 sinh(kL)

3.26 Consider the Green function appropriate for Neumann boundary conditions for the
volume V between the concentric spherical surfaces defined by r = a and r = b, a < b.
To be able to use (1.46) for the potential, impose the simple constraint (1.45). Use
an expansion in spherical harmonics of the form
∞
X
0
G(~x, ~x ) = gl (r, r0 )Pl (cos γ)
l=0

l+1
where gl (r, r0 ) = r<
l
/r> + fl (r, r0 ).
a) Show that for l > 0, the radial Green function has the symmetric form
l
r<
gl (r, r0 ) = l+1
+
r>
l (ab)2l+1 r0l
  l 
1 l+1 0 l 2l+1 r
(rr ) + +a + l+1
(b2l+1 − a2l+1 ) l l + 1 (rr0 )l+1 r0l+1 r

There are several approaches to this problem. However, we first consider the
Neumann boundary condition (1.45)

∂G(~x ~x 0 ) 

4π
= −
∂n0 bndy S
For this problem with two boundaries, the surface area S must be the area of both
boundaries (ie it is the total area surrounding the volume). Hence S = 4π(a2 +b2 ),
and in particular this is uniform (constant) in the angles. As a result, this will
only contribute to the l = 0 term in the expansion of the Green’s function. More
precisely, we could write

∂G(~x ~x 0 )  X ∂gl (r, r0 )
  1
= P (cos γ) = − P0 (cos γ)

l
∂n0 ∂n0 a2 + b2
 
bndy

l bndy

Since the Legendre polynomials are orthogonal, this implies that

∂gl (r, r0 ) 

1
= − δl,0
∂n0 bndy a2 + b2

Noting that the outward normal is either in the −r̂0 or the r̂0 direction for the
sphere at a or b, respectively, we end up with two boundary condition equations
∂gl (r, r0 )  ∂gl (r, r0 ) 
 
1 1
0
= 2 2
δl,0 0
=− 2 δl,0 (7)
∂r 
a a +b ∂r 
b a + b2

Now that we have written down the boundary conditions for gl (r, r0 ), we proceed
to obtain its explicit form. The suggestion of the problem is to write
l
r<
gl (r, r0 ) = l+1
+ fl (r, r0 )
r>
Since
1 X rl
<
0
= l+1
Pl (cos γ)
|~x − ~x | r> l

we see that the first term in gl (r, r0 ) is designed to give the singular source delta
function. The remaining term
X
F (~x, ~x 0 ) = fl (r, r0 )Pl (cos γ)
l

then solves the homogeneous equation ∇2x0 F (~x, ~x 0 ) = 0. But we know how to
solve Laplace’s equation in spherical coordinates, and the result is that the radial
function must be of the form
1
fl (r, r0 ) = Al r0l + Bl
r0l+1
Note that we are taking the Green’s function equation to act on the ~x 0 variable,
where ~x may be thought of as a parameter (constant) giving the location of the
delta function source. We thus have
l
r< 1
gl (r, r0 ) = l+1
+ Al r0l + Bl 0l+1 (8)
r> r
All that remains is to use the boundary conditions (7) to solve for Al and Bl . For
the inside sphere (at a), we have

al−1 1 δl,0
l l+1
+ lAl al−1 − (l + 1)Bl l+2 = 2 (9)
r a a + b2
while for the outside sphere we have

rl 1 δl,0
−(l + 1) + lAl bl−1 − (l + 1)Bl =− (10)
bl+2 bl+2 a2 + b2
For l 6= 0 we rewrite these equations as
 2l+1    
la −(l + 1) Al −la2l+1 /rl+1
=
lb2l+1 −(l + 1) Bl (l + 1)rl

which may be solved to give

    
Al 1 −(l + 1) (l + 1) −la2l+1 /rl+1
=
Bl l(l + 1)(b2l+1 − a2l+1 ) −lb2l+1 la2l+1 (l + 1)rl
rl
 
(a/r)2l+1 + (l + 1)/l
= 2l+1
b − a2l+1 a2l+1 + l/(l + 1)(ab/r)2l+1

Inserting this into (8) yields

l
rl
  
0 r< a 2l+1 l + 1
gl (r, r ) = l+1 + 2l+1 + r0l
r> b − a2l+1 r l
  
2l+1 l  ab 2l+1 1
+ a +
l+1 r r0l+1
l
r<
= l+1
r>
 0l
l (ab)2l+1 rl
 
1 l+1 0 l 2l+1 r
+ 2l+1 (rr ) + +a + 0l+1
b − a2l+1 l l + 1 (rr0 )l+1 rl+1 r
2l+1

1 l+1 l (ab)
= 2l+1 2l+1
(r< r> )l +
b −a l l + 1 (r< r> )l+1
l l

2l+1 r< 2l+1 r>
+b l+1
+a l+1
r> r<
l a2l+1 l b2l+1
  
l+1 l l
= r + r> +
l(b2l+1 − a2l+1 ) < l + 1 r< l+1 l + 1 r>l+1

(11)
which is valid for l 6= 0. Note that in the last few lines we have been able to
rewrite the Green’s function in terms of a product of u(r< ) and v(r> ) where u
and v satisfies Neumann boundary conditions at r = a and r = b, respectively.
 This is related to another possible method of solving this problem. Using the
Legendre identity
∞
X 2l + 1
Pl (cos γ) = δ(φ − φ0 )δ(cos θ − cos θ0 )
4π
l=0

the Green’s function equation may be reduced to the one-dimensional problem

 
d 02 d
r − l(l + 1) gl (r, r0 ) = −(2l + 1)δ(r − r0 )
dr0 dr0

Using the general method for the Sturm-Liouville problem, the Green’s function
is given by
2l + 1
gl (r, r0 ) = − ul (r< )vl (r> ) (12)
Al
where u(r0 ) and v(r0 ) solve the homogeneous equation and the constant Al is
fixed by the Wronskian, W (u, v) = Al /r02 . For l 6= 0 the boundary conditions (7)
are homogeneous
u0 (r0 )r0 =a = 0 v 0 (r0 )r0 =b = 0
 

It is easy to see that these are satisfied by

l a2l+1 l b2l+1
u(r0 ) = r0l + v(r0 ) = r0l +
l + 1 r0l+1 l + 1 r0l+1
Computing the Wronskian gives

 u v  l(2l + 1)(a2l+1 − b2l+1 )

 
 u v0  =
 0 
(l + 1)r02

which allows us to identify

l
Al = −(2l + 1) (b2l+1 − a2l+1 )
l+1

This gives the result of the last line of (11).

b) Show that for l = 0

a2
 
0 1 1
g0 (r, r ) = − + f (r)
r> a2 + b2 r0

where f (r) is arbitrary. Show explicitly in (1.46) that answers for the potential
Φ(~x ) are independent of f (r).
The l = 0 case involves a non-homogeneous boundary condition. Hence the result
of (12) will not work. Of course, we can still work out the one-dimensional delta
 function problem with matching and jump conditions at r0 = r. However it is
more direct to return to (9) and (10) and to simply solve those conditions for
l = 0. Both (9) and (10) result in

a2
B0 = −
a2 + b2
while leaving A0 completely undetermined. Finally, since r is thought of as a
parameter, this indicates that A0 = f (r) can be an arbitrary function of r. The
l = 0 Green’s function is given by (8)

1 a2 1
g0 (r, r0 ) = − 2 + f (r)
r> a + b2 r 0
Incidentally, we note that without the inhomogeneous Neumann boundary con-
dition term −4π/S there will be no solution to the system (9) and (10) for l = 0
(unless b is taken to ∞). This demonstrates the inconsistency of simply setting
∂G/∂n0 = 0 for the Neumann Green’s function.
Note that, by setting f (r) = −a2 /[(a2 + b2 )r] we obtain a symmetrical Green’s
function
a2
 
0 1 1 1
g0 (r, r ) = − 2 +
r> a + b2 r 0 r
On the other hand, the choice of f (r) is unphysical. This arises because, for the
Neumann Green’s function, the f (r) contribution to the potential is given by
∂Φ(~x 0 )
Z I
1 0 3 0 1
Φ(~x ) = ρ(~x )f (r) d x + 0
f (r) da0
4π0 V 4π S ∂n
Z I 
f (r) 0 3 0 ~ 0 0
= ρ(~x ) d x − 0 E(~x ) · dâ
4π0 V S
 I 
f (r) ~ x ) · dâ = 0
0 0
= qenc − 0 E(~
4π0 S

by Gauss’ law. It is important not to mix up r and r0 in this derivation.

3.27 Apply the Neumann Green function of Problem 3.26 to the situation in which the
normal electric field is Er = −E0 cos θ at the outer surface (r = b) and is Er = 0 on
the inner surface (r = a).
a) Show that the electrostatic potential inside the volume V is

a3
 
r cos θ
Φ(~x ) = E0 1+ 3
1 − p3 2r

where p = a/b. Find the components of the electric field

a3 a3
   
cos θ sin θ
Er (r, θ) = −E0 1− 3 , Eθ (r, θ) = E0 1+ 3
1 − p3 r 1 − p3 2r
 Since there is no charge between the spheres, the solution to be boundary value
problem is given by
∂Φ(~x 0 )
I
1
Φ(~x ) = G(~x, ~x 0 ) da0
4π S ∂n0
Z
1
=− Er (Ω0 )G(~x, ~x 0 )b2 dΩ0
4π r0 =b
E 0 b2
Z
= G(~x, ~x 0 ) cos θ0 dΩ0
4π r0 =b
∞ Z
E 0 b2 X
= gl (r, r0 )Pl (cos γ) cos θ0 dΩ0
4π 0
r =b
l=0

By writing
4π X m
Pl (cos γ) = Y (Ω)Ylm ∗ (Ω0 )
2l + 1 m l
p
and noting that cos θ = 4π/3Y10 (Ω), we end up with the expansion
r
4π X gl (r, b)Ylm (Ω)
Z
Φ(~x ) = E0 b 2
Ylm ∗ (Ω0 )Y10 (Ω0 )dΩ0
3 2l + 1
l,m
r
4π g1 (r, b)Y10 (Ω)
= E 0 b2
3 3
2
E0 b cos θ
= g1 (r, b)
3
where we have used orthogonality of the spherical harmonics. Inserting l = 1 into
(11) then gives

E0 b2 cos θ a3 3b a3
   
2 E0 r cos θ
Φ(~x ) = r+ 2 = 1+ 3 (13)
3 b3 − a3 2r 2 1 − (a/b)3 2r
This is the potential for a constant electric field combined with an electric dipole.
Defining p = a/b, the components of the electric field are

a3 a3
   
∂Φ E0 cos θ 1 ∂Φ E0 sin θ
Er = − =− 1− 3 , Eθ = − = 1+ 3
∂r 1 − p3 r r ∂θ 1 − p3 2r
 
Note that the boundary conditions Er r=a = 0 and Er r=b = −E0 cos θ are
obviously satisfied. On the other hand, the parallel component of the field, Eθ ,
is non-vanishing on both surfaces (except at the poles). Physically, this indicates
that these surfaces are not conductors.
b) Calculate the Cartesian or cylindrical components of the field, Ez ad Eρ , and
make a sketch or computer plot of the lines of electric force for a typical case of
p = 0.5.
Rewriting (13) as

a3 z a3 z
   
E0 E0
Φ(~x ) = z+ 3 = z+
1 − p3 2r 1 − p3 2(ρ2 + z 2 )3/2

we obtain
a3 (1 − 3ẑ 2 )
 
∂Φ E0
Ez = − =− 1+
∂z 1 − p3 2r3
3a3 ẑ ρ̂
 
∂Φ E0
Eρ = − =− −
∂ρ 1 − p3 2r3
p
where ρ̂ = ρ/r, ẑ = z/r, and r = ρ2 + z 2 . As indicated above, this corresponds
to a constant electric field combined with an electric dipole. For E0 > 0, a sketch
of the electric field lines looks like
2

-2 -1 1 2

-1

-2

This sketch indicates that the radial component of the electric field vanishes at
the surface of the inner sphere.