1 Problem 3.

17

1.1

Equation 3.138 in Jackson:

r G(~x; ~x0) = 4 (
2
0 ) (' '0 ) (z z0)

Substituting in the denition of the laplacian in cylindrical coordinates:

 
1 @  @G(~x; ~x0) + 1 @ G(~x; ~x0) + @ G(~x; ~x0) = 4 ( 0)('
2 2
'0 ) (z z0) (1)
 @ @ 2 @'2 @z 2 

Now, we want to show that the Green function can be expressed in the following form:
1 1  
4X X
im(' '0 )
 nz  nz 0  n   n 
G(~x; ~x0 ) = e sin sin Im  < Km >
L n=1 m= 1 L L L L

First, we're going to conrm that this solution satises the above equation (that is, it's a
valid Green function). Then, we're going to show that it is the Green function for the given
setup. Plugging in the above Green function into equation (1):
1 1    
4X X
im(' '0 )
 nz  nz 0 1 @ @  n   n 
L
e sin L
sin L  @
 Im
@ L
 < Km
L
> (2)
n=1 m= 1
  
1m +n2
2 2
Im
 n 
 < Km
 n
> = 4 ( 0 ) (' '0 ) (z z0)
2
L 2
L L

We will need the following identities of the Dirac delta function:

X1
0 0
 (' ')= eim(' ' ) (3)
m= 1
1  
0 2 X  nz  nz 0
 (z z)= sin sin (4)
L n=1
L L

Hence, the right hand side of equation (2) becomes:

1 X 1  
8  ( 0 X
im(' '0 )
 nz  nz 0
) e sin L sin L
L n=1 m= 1

1
Thus, we have two summations equal to one another. Setting their coecients equal to one
another and canceling the exponential and sin terns from both sides of the equation:
     
4 1 @  @ I  n   K  n   1 m + n  I  n   K  n  2
2 2

m < m > m < m >

L  @ @ L L 2 L2 L L
8 (
= L 0 )
  n   n      n   n 
@ @ n2  2 2
  Im  < Km > m 2
+ Im < Km > = 2( 0 )
@ @ L L L2 L L

Letting x = n L
, this becomes the Bessel equation for x 6= x0. Clearly, Im (x<) Km (x>)
is a solution to this equation. Hence, we have shown that the proposed Green function is
indeed a valid Green function.
To show that this is the Dirichlet Green function for the system at hand, we must show that
is zero for ~x0 is on the surfaces:
 The exponential function has no zeroes for nite '0.


 sin nzL 0 is zero at z = 0; L, as expected. Additionally, it is zero when z0 is an integer
multiple of L, which is allowed since using the method of images would eectively gives
rise to an innite number of surfaces at these places.
 I has no zeroes. Im (m 6= 0) is zero only at 0 = 0.
0

 Km has no zeroes for nite 0.

Finally, this Green function must not blow up anywhere. Im and Km are the only functions
which blow up { they blow up at innity and zero (respectively). However, since the argument
of Im is the lesser of ; 0 and the argument of Km is the greater of ; 0 this Green function
does not blow up at all.
Hence, we have shown that this is the Green function for the system at hand.

1.2

We want to show that the following Green function also solves the same setup:
1 Z 1
0
G(~x; ~x ) = 2
X 0
dkeim(' ' ) Jm (k)Jm (k0 )
sinh(kz<) sinh(k(L z> ))
m= 1 1 sinh(kL)

2
Again, we're going to show that this Green function satises equation (1). Plugging it in
yields:
1 Z 1   
2
X
dke im(' '0 ) sinh(kz<) sinh(k(L z> )) 1 @ @
 Jm (k)Jm (k0 )
m= 1 1 sinh(kL)  @ @
  
+ 1 m +k 2 2
Jm (k)Jm (k0 ) = 4 ( 0 ) (' '0 ) (z z0)
 2

By substituting equation (3) into the right side of the above expression, we have two sum-
mations equal to one another. Equating the terms of each summation and cancelling the
exponentials yields:
Z 1   
2 dk
sinh(kz< ) sinh(k (L z> )) 1 @ @ 0
 Jm (k)Jm (k )
1  sinh(kL)  @

@

+ 1 m + k Jm(k)Jm(k0) = 4 ( 0)(z z 0)

2
2 2

Now, substituting the following identity into the right side of the above expression:
Z 1
 ( 0 ) =  kJm (k)Jm (k0 )dk
0

we obtain two integrals equal to each other. Equating the integrands yields:
     
2 sinh(kz< ) sinh(k (L z> )) 1 @ @ 0
 Jm (k)Jm (k ) +
1 0
m + k Jm (k)Jm (k ) 2 2
sinh(kL)  @ @  2

= 4k(z z 0)
   
@ @

  Jm (k)Jm (k0 ) + m +  k Jm (k)Jm (k0 )
2 2 2
@ @
sinh(kL)
= 2 k(z 2
z0)
sinh(kz<) sinh(k(L z>))
Letting x = k, this becomes the Bessel equation for x 6= x0, which is clearly solved by
Jm (k)Jm (k0 ), indicating that this function satises equation 1 and is a valid Green function.
Again, to show that this Green function is zero on the surfaces, we note that sinh (kz 0) is
zero at z = 0 while sinh (k(L z 0)) is zero at z = L. All of the functions are nite, so the
Green function does not blow up anywhere.

3
2 Problem 3.26

We want to show that the following is the Green function for the volume between two
concentric spheres with radii a and b (where a < b):
1 l
X r<
G(~x; ~x0 ) = l+1
Pl (cos )
l=0
r>
1   
X 1 l + 1 0 l l (ab)2l+1 rl r0l
+ b2l+1 a2l+1 l
(rr ) l + 1 (rr0)l+1 +a 2l+1
r0l+1
+ rl+1
Pl (cos )
l=0

First, we note that the rst summation is equal to j~x ~x0j by equation 3.38 in Jackson. Hence,1

the second summation must satisfy Laplace's equation with respect to r0 (equations 1.40,
1.41 in Jackson). We know that the general solution to Laplace's equation with azimuthal
symmetry is: 1
X

F = Al r0l + Bl r0 l Pl (cos ) ( +1)

l=0

So, we need to solve for Al and Bl and show that F is equal to the second summation in the
Green function which we're trying to validate.
Because this is a Neumann Green function, it must satisfy the Neumann boundary conditions:
@GN
= 4 = 1
@n 0 S a +b 2 2

Note that S = 4(a + b ), which is the total surface area of all surfaces. Since the normal
2 2

vector points away from the region of interest (and the region of interest is the area between
the spheres), n0 = r0 for r0 = a and n0 = r0 for r0 = b. Also, noting that the normal
derivative of the rst summation of the Green function is zero for r 6= r0 yields the following
boundary conditions:

@F 
= 1
@r r0 a a + b
0  2 2
 =

@F 
= 1
@r0  a +b r 0 =b
2 2

The left hand sides of the above equations are summations. We'll exploit the orthogonality
of the Legendre polynomials to simplify the computation. Because P (cos ) = 1, these 0
equations become:

@F 
= 1 P (cos )
@r r0 a a + b
0  2 2 0

@F 
= 1 P (cos )
0  a +b
0
@r r 0 =b
2 2

4
Because Legendre polynomials are orthogonal, the l = 0 term of the summation has to satisfy
the above relations by itself, meaning that the remaining terms have to be zero. That is:
 
@ r0l 
@r0 rl+1
+ Al r0l + Bl r0 (l+1) 
0 =0
 l r =a
@ r 
0 0
@r r l+1
+ Al r0l + Bl r0 (l+1) 
0 =0
r =b

For l > 0. Evaluating the derivatives, these equations become:

al 1
l
rl+1
+ Al lal 1
+ Bl ( l 1)a l 2
=0
(l + 1) rbl+2 Al lbl
l 1
1
+ Bl ( l 1)b l 2
=0
   " #
al 1
lal 1
( l 1)a l 2
Al
= l r l+1
( l 1)b rl 1
lbl 1 l 2
Bl (l + 1) bl+2

Solving this linear equation yields:

 
A =
1 a2l+1 l + 1 l 1
+ l r
l
b2l+1 a2l+1 rl+1
 
a2l+1 l b2l+1
Bl = (b2l+1 a2l+1) l + 1 rl+1 + r
l 1

Plugging these terms into:

r<
l
gl (r; r0 ) = l+1
+ Al r0l + Bl r0 (l+1)
(5)
r>

yields the radial Green function which we're validating (for l > 0).

2.1

As we noted above, the l = 0 term of the summation satises the Neumann boundary
conditions by itself. That is:
 0l 
@ r
+ A r 0 l
+ B r 0 l

 ( +1)
= 1
0  a +b
l l
@r r l +1
l ;r 0 a =0 =
2 2

 l 
@ r
+ A r 0l + B r0 l 
 ( +1)
= 1
0 0  a +b
l l
@r r l +1
l ;r 0 b =0 =
2 2

5
For l > 0. Evaluating the derivatives, these equations become:
Ba =
1 2
a +b
0 2 2

1 Bb = 1 2
a +b
0
b 2 2 2

The A term vanishes because it is multiplied by l (and l = 0). Solving for B is trivial:
0 0

a2
B0 = a2 + b2
Plugging this into equation (5) for l = 0 yields:
 
g (r; r0 ) =
1 + A (r ) a2
r0 1
a2 + b2
0 0
r>
In order for A (r), it must play no role in nding (~x). That is, the following must evaluate
0
to zero:
1 Z (~r0)A (r)d r0 + 1 I @  A (r)da0 3
4" 0 V 4 @n0 0
S
0

Because A (r) does not depend on r0, we can pull it out of both integrals:
0
Z I  I 
A (r)
1 0 0 1 @ 0 1 Q 0
4" | V ({z
~r )d r + A (r) da = A (r) En0 da
3
0
0
}
4  @n0
S |{z}
0
4  " S
0
0

Q En0

The term in parentheses is zero by Gauss' Law. Hence, A (r) may be any arbitrary function 0
of r, as long as it does not depend on r0.

3 Problem 3.27

Because there are no charges in this system, the rst integral in equation 1.46 in Jackson is
zero. Hence:
I
1
(~x) = 4 @n @
GN da0
0
S |{z}
Er
!
I 1
= 41
X
(E cos )
0 gl (r; r0 )Pl (cos ) da0
S l=0

Using the following identity:

1
Pl (cos ) =
4  X  0 0
2l + 1 m 1 Yl;m( ; ' )Yl;m(; ')
=

6
and the fact that Y ; (; ') = cos , (~x) becomes:
10

1 1 I
E0 X X 4  0  0 0 02
(~x) = 4 l=0 m= 1 2l + 1 S Y1;0(; ')gl (r; r )Yl;m( ; ' )Yl;m(; ')r sin drdd'
Because Er is zero on the inner sphere, we only need to integrate over the outer sphere,
Hence, r0 = b. So, we're able to pull b , gl (r; b), and Yl;m(; ') out of the integral:
2

(~x) = 
1 1
E0 X X 4  
gl (r; b)Yl;m (; ')b 2
I
 (0 ; '0 )da0
Y1;0 (; ')Yl;m
4 l=0 m= 1 2l + 1 S
| {z }
l=1;m=0

= E3 g (r; b) Y| ; {z
0
1 (; '}) b
10
2

cos 
   
E0
+ b a 2(rb) + 21 ((ab
r 1 ) +a r + b 3
= 3 b2 3
rb) 3
b r
b cos  2
3
2 2
2

   
E 1 b 3 ab ar
= 3 r + 1 (a=b) b 2rb + 2 r + b cos 
0
2 3 3

3 3 2 2
  a  
E 1
= 3 1 (a=b) r 1 b + 2r + 2 r + b cos 
0 3 a ar 3 3 3

3 2 3
 
E 1 3 a
= 3 1 (a=b) 3r + 2 r cos 
0
3

3 2
 
cos 
= E 1 (a=b) r + 2r a 3

0 3 2

~ = r(~x)
E
We obtain the components of E using the denition of r' in spherical coordinates:
@
E = r
@r  
= E 1 cos  a 3

(a=b) 1 r
0 3 3

E =
1 @
r @  
sin 
= E 1 (a=b) r + 2r a 3

0 3 2

3.1

Converting (~x) to cylindrical coordinates:

7
  p 
cos cos (z=  + z ) p
1 2 2
a
 3
(~x) = E 0
1 "(a=b) 3
 +z + 2 2
2 ( + z ) 2 2
#
=E 1 z+
za 3

0
1 (a=b) 3
2 ( + z ) =
2 2 3 2

We obtain the components of E using the denition of r' in cylindrical coordinates:

@
E =

@
= E0 3za3
1 (a=b)3 2 (2 + z 2)5=2
@
Ez = @z " p #
= 1 2 ( 2 + z 2 ) a 3 3z 2 a 3  2 + z 2
3=2

E0
1 (a=b)3 1 + 2 (2 + z 2)3