Bessels 2020

1 Solutions in cylindrical coordinates: Bessel

functions
1.1 Bessel functions
Laplace’s equation in cylindrical coordinates is:
   
1 ∂ ∂Φ 1 ∂ 1 ∂Φ ∂ 2Φ
ρ + + =0
ρ ∂ρ ∂ρ ρ ∂φ ρ ∂φ ∂z 2
Separate variables: Let Φ = R (ρ) W (φ) Z (z) . Then we find:
 
1 ∂ ∂R 1 ∂ 2W 1 ∂ 2Z
ρ + + =0
Rρ ∂ρ ∂ρ W ρ2 ∂φ2 Z ∂z 2
The last term is a function of z only, while the sum of the first two terms is a
function of ρ and φ only. Thus we take each part to be a constant called k2 .
Then
∂ 2Z
= k2 Z
∂z 2
and the solutions are
Z = e±kz
This is the appropriate solution outside of a charge distribution, say above a
plane, (Φ → 0 as z → ±∞), or inside a cylinder with grounded walls and
non-zero potential on one end.
The remaining equation is:
 
1 ∂ ∂R 1 ∂ 2W
ρ + + k2 = 0
Rρ ∂ρ ∂ρ W ρ2 ∂φ2

Now multiply through by ρ2 :

 
ρ ∂ ∂R 1 ∂ 2W
ρ + k2 ρ2 + =0
R ∂ρ ∂ρ W ∂φ2
Here the last term is a function of φ only and the first two terms are functions
of ρ only. Again we often want a solution that is periodic with period 2π, so
we choose a negative separation constant:
∂2W
= −m2 W ⇒ W = e±imφ
∂φ2
Finally we have the equation for the function of ρ :
 
∂ ∂R
ρ ρ + k2 ρ2 R − m2 R = 0
∂ρ ∂ρ

1
To see that this equation is of Sturm-Liouville form, divide through by ρ :
 
∂ ∂R m2
ρ + k2 ρR − R=0 (1)
∂ρ ∂ρ ρ
Now we have a Sturm-Liouville equation (slreview notes eqn. 1) with f (ρ) = ρ,
g (ρ) = m2 /ρ, eigenvalue λ = k2 and weighting function w (ρ) = ρ. Equation (1)
is Bessel’s equation. The solutions are orthogonal functions. Since f (0) = 0,
we do not need to specify any boundary condition at ρ = 0 if our range is
0 ≤ ρ ≤ a, as is frequently the case. (We do specify that R remain finite.) We
do need a boundary condition at ρ = a.
It is simpler and more elegant to solve Bessel’s equation if we change to the
dimensionless variable x = kρ. Then:
 
∂ ∂R m2
k kρ + k2 ρR − k R = 0
∂kρ ∂kρ kρ
 
d dR m2
x + xR − R = 0
dx dx x
The equation has a singular point at x = 0. So we look for a series solution of
the Frobenius type (cf Lea Chapter 3 §3.3.2):
∞
[
R = xp an xn
n=0
∞
[
R = (n + p) an xn+p−1
n=0
  [∞
d dR
x = (n + p)2 an xn+p−1
dx dx n=0

Then the equation becomes:

∞
[ ∞
[ ∞
[
(n + p)2 an xn+p−1 + an xn+p+1 − m2 an xn+p−1 = 0
n=0 n=0 n=0
p−1
The indicial equation is given by the coefficient of x :
2 2
p − m = 0 ⇒ p = ±m
Thus one of the solutions (with p = m) is analytic at x = 0, and one (with
p = −m) is not. To find the recursion relation, look at the k + p − 1 power of x:
(k + p)2 ak + ak−2 − m2 ak = 0
and so
ak−2 ak−2
ak = − 2 =−
(k + p) − m2 k2 + 2kp + p2 − m2
ak−2 ak−2
= − 2 =−
k + 2kp k (k ± 2m)

2
Let’s look first at the solution with p = +m. We can step down to find each ak .
If we start the series with a0, then k will always be even, k = 2n, and
−1 −1
a2n = a2n−4
2n (2n + 2m) (2n − 2) (2n − 2 + 2m)
(−1)3
= a2n−6
23 n (n − 1) (n − 2) 23 (n + m) (n + m − 1) (n + m − 2)
(−1)n 1
= a0 n
2 n! 2n (n + m) (n + m − 1) · · · (m + 1)

The usual convention is to take

1
a0 = (2)
2m Γ (m + 1)

Then
n
1 (−1) 1
a2n =
2m Γ (m + 1) 2n n! 2n (n + m) (n + m − 1) · · · (m + 1)
(−1)n
= (3)
n!Γ (n + m + 1) 22n+m

and the solution is the Bessel function:

∞
[ (−1)n  x m+2n
Jm (x) = (4)
n=0
n!Γ (n + m + 1) 2

The function Jm (x) has only even powers if m is an even integer and only odd
powers if m is an odd integer. The series converges for all values of x.
Let’s see what the second solution looks like. With p = −m the recursion
relation is:
ak−2
ak = (5)
k (k − 2m)
where again k = 2n. Now if m is an integer we will not be able to determine
a2m because the recursion relation blows up. One solution to this dilemma is
to start the series with a2m . Then we can find the succeeding a2n :

−a2(n−1)+2m −a2(n−2)+2m (−1)n Γ (m + 1)

a2(n+m) = 2
= 4 = a2m
2 (n + m) n 2 (n + m) (n + m − 1) n (n − 1) n!Γ (n + m + 1) 22n

which is the same recursion relation we had before. (Compare the equation
above with equation (3). Thus we do not get a linearly independent solution
this way1 . (This dilemma does not arise if the separation constant is taken to
be −ν 2 with ν non-integer. In that case the second recursion relation provides
1 This happens because the two roots of the indicial equation differ by an integer: 2m.

3
a series J−ν (x) that is linearly independent of the first.) Indeed we find:
∞
[ (−1)n Γ (m + 1)
J−m (x) = a x2(n+m)−m
2n 2m
n=0
n!Γ (n + m + 1) 2
∞
[ n
(−1) Γ (m + 1) 2m  x m+2n
= a2m
n=0
n!Γ (n + m + 1) 2

and if we choose

(−1)m
a2m =
Γ (m + 1) 2m
then
m
J−m (x) = (−1) Jm (x) (6)
With this choice Jν (x) is a continuous function of ν. (Notice that we can also
express the series using equation (3) for the coefficients, with m → −m and
n → k + m,and where a2n ≡ 0 for n < m. )
We still have to determine the second, linearly independent solution of the
Bessel equation. We can find it by taking the limit as ν → m of a linear
combination of Jν and J−ν known as the Neumann function Nν (x) :

Jν (x) cos νπ − J−ν (x)

Nm (x) = lim Nν (x) = lim
ν→m ν→m sin νπ
Jm+ε (x) cos (m + ε) π − J−(m+ε) (x)
= lim
ε→0 sin (m + ε) π
Jm+ε (x) (cos mπ cos επ − sin mπ sin επ) − J−(m+ε) (x)
= lim
ε→0 sin mπ cos επ + cos mπ sin επ
Jm+ε (x) (−1)m cos επ − J−(m+ε) (x)
= lim
ε→0 (−1)m sin επ

Now we expand the functions to first order in ε. We use a Taylor series for the
Bessel functions. Note that ε appears in the index, not the argument, so we
have to differentiate with respect to ν. .
m
Jm+ε (x) (−1) − J−(m+ε) (x)
Nm (x) = lim
ε→0 (−1)m επ
      
(−1)m m dJν  dJ−ν 
= lim (−1) Jm + ε − J−m + ε
ε→0 επ dν ν=m dν ν=m

Using relation (6), we have:

   
1 dJν  m dJ−ν 
Nm (x) = − (−1)
π dν ν=m dν ν=m

4
 The derivative has a logarithmic term:
% &
dJν d [∞
(−1)n  x 2n
= xν
dν dν n=0
n!Γ (n + ν + 1) 2
dxν [
∞
(−1)
n  x 2n d [
∞
(−1)
n  x 2n
= + xν
dν n=0 n!Γ (n + ν + 1) 2 dν n=0 n!Γ (n + ν + 1) 2

and
dxν d ν ln x
= e = ln xeν ln x = xν ln x
dν dν
and so dJν /dx has a term containing Jν ln x. This term diverges as x → 0
provided that Jv (0) is not zero, i.e. for ν = 0. The function Nv (x) also
diverges as x → 0 for ν = 0, because it contains negative powers of x. (The
series for J−ν starts with a term x−ν .) Nν is finite as x → ∞ because Jν goes
to zero sufficiently fast.
Two additional functions called Hankel functions are defined as linear com-
binations of J and N :
(1)
Hm (x) = Jm (x) + iNn (x) (7)

and
(2)
Hm (x) = Jm (x) − iNn (x) (8)
Compare the relation between sine, cosine, and exponential:

e±ix = cos x ± i sin x

1.2 Properties of the functions

The Bessel functions (Js) are well behaved both at the origin and as x → ∞.
They have infinitely many zeroes. All of them, except for J0 , are zero at x = 0.
The first few functions are shown in the figure.

1.0
J
0.8

0.6

0.4

0.2

0.0
2 4 6 8 10 12 14
-0.2 x
-0.4

The first three Bessel functions. J0 , J1 (red) and J2

5
For small values of the argument, we may approximate the function with the
first term in the series:
1  x m
Jm (x) ≈ for x 1 (9)
Γ (m + 1) 2
The Neumann functions are not well behaved at x = 0. N0 has a logarithmic
singularity, and for m > 0, Nm diverges as an inverse power of x :
2
N0 (x) ≈ ln x for x 1
π  m
(m − 1)! 2
Nm (x) ≈ − for x 1, m > 0 (10)
π x
For large values of the argument, both J and N oscillate: they are like damped
cosine or sine functions:
u 
2 mπ π 
Jm (x) ≈ cos x − − for x 1, m (11)
πx 2 4
u 
2 mπ π 
Nm (x) ≈ sin x − − for x 1, m (12)
πx 2 4
and thus the Hankel functions are like complex exponentials:
u k 
(1,2) 2 mπ π l
Hm ≈ exp ±i x − − for x 1, m (13)
πx 2 4
Notice that if m > 1, the large argument expansions apply for x m rather
than the usual x 1.

1.3 Relations between the functions

As we found with the Legendre functions, we can determine a set of recursion
relations that relate successive Jm (x) . For example (Lea §8.4.3)
 
d Jm (x) Jm+1 (x)
m
=− (14)
dx x xm
which is valid for m ≥ 0. In particular, with m = 0 we obtain:
J1 (x) = −J0 (x) (15)
d
(xm Jm (x)) = xm Jm−1 (x) (16)
dx
From (14) and (16) we may obtain
2m
Jm+1 + Jm−1 = Jm (17)
x
and similarly
dJm
Jm+1 − Jm−1 = −2 (18)
dx
The same relations hold for the Ns and the Hs.

6
1.4 Orthogonality of the Jm
Since the Bessel equation is of Sturm-Liouville form, the Bessel functions are
orthogonal if we demand that they satisfy boundary conditions of the form
(slreview notes eqn 2). In particular, suppose the region of interest is ρ = 0
to ρ = a, and the boundary conditions are Jm (ka) = 0. We do not need a
boundary condition at ρ = 0 because the function f (ρ) = ρ is zero there. Then
the eigenvalues are
αmn
kmn =
a
where αmn is the nth zero of Jm . (The zeros are tabulated in standard references
such as Abramowitz and Stegun. Also programs such as Mathematica and
Maple can compute them.) Then
] a
a2 2
ρJm (kmn ρ) Jm (kmn3 ρ) dρ = [Jm (kmn a)] δ nn3 (19)
0 2

1.5 Solving a potential problem.

Example. A cylinder of radius a and height h has its curved surface and its
bottom grounded. The top surface has potential V. What is the potential inside
the cylinder?
The potential has no dependence on φ and so only eigenfunctions with m = 0
contribute. The potential is zero at ρ = a,so the solution we need is J0 (kρ)with
eigenvalues chosen to make J0 (ka) = 0. Thus the eigenvalues are given by
k0n a = α0n , where α0n are the zeros of the function J0 . The remaining function
of z must be zero at z = 0, so we choose the hyperbolic sine. Thus the potential
is: ∞
[
Φ (ρ, z) = an J0 (k0n ρ) sinh (k0n z)
n=1

Now we evaluate this at z = h :

∞
[
V = Φ (ρ, h) = an J0 (k0n ρ) sinh (k0n h)
n=1

Next we make use of the orthogonality of the Bessel functions. Multiply both
sides by ρJ0 (k0r ρ)and integrate from 0 to a. (Note here that the weight function
w (ρ) = ρ. This is the first time we have seen a weight function that is not 1.)
Only one term in the sum, with n = r, survives the integration.
] a ] a ∞
[
V ρJ0 (k0r ρ) dρ = ρJ0 (k0r ρ) an J0 (k0n ρ) sinh (k0n h) dρ
0 0 n=1
] a
2
= ar ρ [J0 (k0r ρ)] dρ sinh (k0r h)
0
a2 2
= ar [J0 (k0r a)] sinh (k0r h)
2

7
To evaluate the left hand side, we use equation (16) with m = 1 :
] a ]
1 a d
ρJ0 (kρ) dρ = (kρJ1 (kρ)) dρ
0 k 0 dkρ
1 a
= ρJ1 (kρ)|a0 = J1 (ka)
k k
So
Va 2
ar = J1 (k0r a)
k0r a [J0 (k0r a)]2 sinh (k0r h)
2

V 2
=
k0r a J1 (k0r a) sinh (k0r h)

where we used the result from equation (15) that J0 = −J1 . Finally our solution
is: ∞
[ J0 (α0n ρ/a) sinh (α0n z/a)
Φ = 2V
α
n=1 0n 1
J (α0n ) sinh (α0n h/a)
The first two zeros of J0 are: α01 = 2.4048, α02 = 5.5201, and thus the first
two terms in the potential are:
#         $
J0 2.4048 ρa sinh 2.4048 az J0 5.5201 aρ sinh 5.5201 az
Φ = 2V  +   + ···
2.4048J1 (2.4048) sinh 2.4048 ha 5.5201J1 (5.5201) sinh 5.5201 ha

1.6 Modified Bessel functions

Suppose we change the potential problem so that the top and bottom of the
cylinder are grounded but the outer wall at ρ = a has potential V (φ, z) . Then
we would need to choose a negative separation constant so that the solutions of
the z-equation are trigonometric functions:

∂ 2Z
= −k2 Z ⇒ Z = a sin kz + b cos kz
∂z 2
At z = 0, Z (z) = 0, so we need the sine, and therefore set b = 0. We also need
Z (h) = 0, so we choose the eigenvalue k = nπ/h.
This change in sign of the separation constant also affects the equation for
the function R (ρ) because the sign of the k2 term changes.
 
∂ ∂R m2
ρ − k2 ρR − R=0
∂ρ ∂ρ ρ

or, changing variables to x = kρ :

 
∂ ∂R m2
x − xR − R=0 (20)
∂x ∂x x

8
which is called the modified Bessel equation. The solutions to this equation
are Jm (ikρ) . It is usual to define the modified Bessel function Im (x) by the
relation:
1
Im (x) = m Jm (ix) (21)
i
so that the function Im is always real (whether or not m is an integer). Using
equation (4) we can write a series expansion for Im :
∞  m+2n
1 [ (−1)n ix
Im (x) =
im n=0 n!Γ (n + m + 1) 2
[∞
1  x m+2n
= (22)
n=0
n!Γ (n + m + 1) 2

As with the Js, if m is an integer, I−m is not independent of Im : in fact:

m m
I−m (x) = im J−m (ix) = im (−1) Jm (ix) = (−1) i2m Im (x) = Im (x) (23)

The second independent solution is usually chosen to be:

π m+1 (1)
Km (x) = i Hm (ix) (24)
2
Then these functions have the limiting forms:
1  x m
Im (x) ≈ for x 1 (25)
Γ (m + 1) 2

and
x
K0 (x) ≈ −0.5772 − ln for x 1 (26)
 m 2
Γ (m) 2
Km (x) ≈ for m > 0, x 1 (27)
2 x

At large x, x 1, m, the asymptotic forms are:

1
Im (x) ≈ √ ex (28)
2πx
and u
π −x
Km (x) ≈ e (29)
2x
(cf Lea Chapter 3 Example 3.9) These functions, like the real exponentials,
do not have multiple zeros and are not orthogonal functions. Note that the Is
are well behaved at the origin but diverge at infinity. For the Ks, the reverse
is true. They diverge at the origin but are well behaved at infinity (See figure
below).

9
 10
I,K
8

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
x
Modified Bessel functions. K solid I dashed Black m = 0,Blue 1, Red 2

The recursion relations satisifed by the modified Bessel functions are similar
to, but not identical to, the relations satisfied by the Js. For the Is, again we
can start with the series:
  ∞  x 2n
d Im 1 [ 1
m
= m
dx x 2 n=0 n!Γ (n + m + 1) 2
1 [
∞
n  x 2n−1
=
2m n=1 n!Γ (n + m + 1) 2

Now let k = n − 1 :
  ∞  x 2k+1
d Im 1 [ 1
=
dx xm 2m k!Γ (k + m + 2) 2
k=0

1 [∞
1  x 2k+m+1
=
xm k!Γ (k + m + 2) 2
k=0
 
d Im Im+1 (x)
= (30)
dx xm xm
and similarly
d
(xm Im ) = xm Im−1 (31)
dx
Expanding out and combining, we get:

2Im = Im+1 + Im−1

2m
Im = Im−1 − Im+1 (32)
x

10
For the Ks, the relations are:
 
d d Km (x) Km+1 (x)
(xm Km ) = −xm Km−1 ; =−
dx dx xm xm

and consequently:
2m
Km−1 − Km+1 = − Km
x
Km−1 + Km+1 = −2Km (33)

1.7 Combining functions

When solving a physics problem, we start with a partial differential equation and
a set of boundary conditions. Separation of variables produces a set of coupled
ordinary differential equations in the various coordinates. The standard solution
method (notes3half §1.2) requires that we choose the separation constants by
fitting the zero boundary conditions first. In a standard 3-dimensional problem,
once we have chosen the two separation constants we have no more freedom and
the third function is determined.
When solving Laplace’s equation in cylindrical coordinates, the functions
couple as follows:

Zero boundary conditions in ρ : The eigenfunctions are of the form:

 ρ  z z  ±imφ
Jm αmn Amn sinh αmn + Bmn cosh αmn e
a a a
 
The set of functions Jm αmn aρ e±imφ form a complete orthogonal set on the
surfaces z = constant that bound the region.

Zero boundary conditions in z : The eigenfunctions are of the form:

  nπρ   nπρ   nπz 
Amn Im + Bmn Km sin e±imφ
h h h
  ±imφ
The set of functions sin nπzh e form a complete orthogonal set on the
boundary surface ρ =constant.
Thus in solutions of Laplace’s equation, Js in ρ always couple with the hy-
perbolic sines and hyperbolic cosines (or real exponentials) in z, while the Is
and Ks in ρ always couple with the sines and cosines (or complex exponentials)
in z.

1.8 Finishing the problem

Example: Suppose the potential on the curved wall of the cylinder is V (a, φ, z) =
V sin φ The top and bottom are grounded

11
 The solution is of the form
+∞ [
[ ∞  nπz  k  nπρ   nπρ l
Φ (ρ, φ, z) = sin eimφ Amn Im + Bmn Km
m=−∞ n=1
h h h

The solution must be finite on axis at ρ = 0, so Bmn = 0. Now we evaluate the

potential at ρ = a
+∞ [
[ ∞  nπz   nπa 
Φ (a, φ, z) = sin eimφ Amn Im = V sin φ
m=−∞ n=1
h h

We make use of orthogonality by mutltiplying by sin pπz/h and integrating from

0 to h.
+∞ [
[ ∞ ] h  nπz   nπa  ] h
pπz imφ pπz
sin sin dze Amn Im = V sin φ sin dz
m=−∞ n=1 0
h h h 0 h
+∞
[  pπa   h
h imφ h pπz 
e Amp Im = V sin φ − cos
m=−∞
2 h pπ h 0
h
= V [1 − (−1)p ] sin φ
pπ
The result is non-zero only for odd p.
Next we make use of the orthogonality of the eimφ
+∞
[ ]  pπa  ] 2π
h 2π imφ −im3 φ h 3
e e dφAmp Im = 2V sin φe−im φ dφ p odd
m=−∞
2 0 h pπ 0
= 0 p even

To do the integral on the right hand side, express the sine in terms of exponen-
tials.
] 2π ] 2π iφ
3 e − e−iφ −im3 φ
sin φe−im φ dφ = e dφ
0 0 2i
2π
= (δ m3 1 − δ m3 ,−1 )
2i
Only one term with m = m survives the integration on the left hand side. Thus
h  pπa  h 2π
2πAm3 p Im3 = 2V (δ m3 1 − δ m3 ,−1 )
2 h pπ 2i
4V 1 (δ m3 1 − δ m3 ,−1 )
Am3 p =  
pπ 2i Im3 pπa h
%  nπρ   nπρ  &
4V [
∞  nπz  1
iφ I1  h  −iφ I−1  h 
Φ (ρ, φ, z) = sin e −e
π
n=1, odd
h 2in I1 nπah I−1 nπa
h

12
But I1 = I−1, (eqn 23), so

4V
∞
[  nπz  sin φ I  nπρ 
1
Φ (ρ, φ, z) = sin  h 
π h n I1 nπa h
n=1, odd

The dimensions are correct, and the 1/n gives us reasonable convergence. The
ratio  
I1 nπρ
h
  ≤ 1 for ρ ≤ a.
I1 nπa
h
The potential at ρ = 0 is zero, as expected. The m = 1 "source" (the potential
on the surface) generates an m = 1 response. Again we can look at the first few
terms (here 10) . Let’s take h/a = 2 and φ = π/4 (red) and Φ = π/2 (black)

phi/V 1.0
0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
rho/a

1.9 Continuous set of eigenvalues: the Fourier Bessel Trans-

form
In Lea Chapter 7 we approached the Fourier transform by letting the length
of the domain in a Fourier series problem become infinite. The orthogonality
relation for the exponential functions:
] L  nπx   mπx 
1
exp i exp −i dx = δ mn
2L −L L L
becomes ] ∞
1 3
eikx e−ik x dx = δ (k − k )
2π −∞

13
That is, the Kronecker delta becomes a delta function, and the countable set
of eigenvalues nπ/L becomes a continuous set of values −∞ < k < ∞.
The same thing happens with Bessel functions. With a finite domain in ρ,
say 0 ≤ ρ ≤ a, we can determine a countable set of eigenvalues from the set of
zeros of the Bessel functions Jm . If our domain in ρ becomes infinite, then we
cannot determine the eigenvalues, and instead we have a continuous set. The
orthogonality relation:
] a  ρ  ρ a2 2
Jm αmn Jm αmk ρdρ = [J (αmn )] δ nk
0 a a 2 m
becomes: ] ∞
δ (k − k )
ρJm (kρ) Jm (k ρ) dρ = (34)
0 k
(The proof of this relation is in Lea Appendix 7.) Then the solution to the
physics problem is determined as an integral over k. For example, a solution of
Laplace’s equation may be written as:
[ ] ∞
Φ (ρ, φ, z) = eimφ A (k) f (kz) Jm (kρ) dk
m 0

where f (kz) depends on the boundary conditions in z. It will be a combination

of the exponentials e−kz and e+kz .  
Example: Suppose the potential is V (ρ) = V0 aρ sin aρ on a plane at z = 0,
and we want to find the potential for z > 0. Then the appropriate function of z
is e−kz , chosen so that Φ → 0 as z → ∞ ( a long way from the plane). We also
need the Jm , which remain finite at ρ = 0. Then the solution is of the form:
[ ] ∞
imφ
Φ (ρ, φ, z) = e Am (k) e−kz Jm (kρ) dk
m 0

Evaluating Φ on the plane at z = 0, we get:

  [ ] ∞
a ρ
V0 sin = Φ (ρ, φ, 0) = eimφ Am (k) Jm (kρ) dk
ρ a m 0

Now we can make use of the orthogonality of the eimφ . Multiply both sides by
3
e−im φ and integrate over the range 0 to 2π. On the LHS, only the term with
m = 0 survives the integration, and on the RHS only the term with m = m
survives.
] 2π   ] 2π [ ] ∞
a ρ −im3 φ imφ −im3 φ
V0 sin e dφ = e e dφ Am3 (k) Jm3 (kρ) dk
0 ρ a 0 m 0
  ] ∞
a ρ
2πV0 sin δ m 0 = 2π
3 Am3 (k) Jm3 (kρ) dk
ρ a 0

14
—a Fourier Bessel transform. Next2 multiply both sides by ρJm (k ρ) , integrate
from 0 to ∞ in ρ, and use equation (34) to get:
] ∞  ] ∞] ∞
a ρ
V0 sin Jm (k ρ) ρdρδ m0 = Am (k) Jm (kρ) Jm (k ρ) ρdρdk
0 ρ a 0 0
] ∞
δ (k − k ) Am (k )
= Am (k) dk =
0 k k
which determines the coefficient Am (k ) in terms of the known potential on the
plane. Only A0 is non-zero, as expected from the azimuthal symmetry.
On the RHS, let ρ/a = x. We get (dropping the primes on k)
] ∞
A0 (k)
V0 a2 sin xJ0 (xka) dx =
0 k
This integral is GR 6.671#7. So
+
0 if 0 < 1 < ka
A (k) = V0 a2 k √ 1
if 0 < ka < 1
1−(ka)2

So finally we have
] 1/a
kJ (kρ) −kz
Φ (ρ, z) = V0 a t 0
2
e dk
0 1 − (ka)2
] 1  
uJ0 u aρ −uz/a
= V0 √ e du
0 1 − u2
with u = ka. For z = 0 we have
] 1  
uJ0 u ρa 1 ρ
Φ (ρ, 0) = V0 √ du = V0 sin GR 6.554#2
0 1−u 2 ρ/a a
which gives us back the potential on the plane.
The charge density on the plane is given by
 σ

E · ẑ  =
z=0 ε0
So we need
 ]  
∂Φ  V0 u2 J0 u ρa
1
= − √ du
∂z z=0 a 0 1 − u2
] 1
(−1)n  u ρ 2n
∞
[
V0 u2
= − √ du
a 0 1 − u2 n=0 [n!]2 2 a
]
(−1)n  ρ 2n 1 u2(n+1)
[∞
V0
= − 2
√ du
a n=0 [n!]
2a 0 1 − u2
2 We also drop the primes on the m for convenience.

15
Let x = sin θ, dx = cos θdθ Then the integral is
] π/2 ] π/2
sin2(n+1) θ (2n + 1)!! π
In = cos θdθ = sin2(n+1) θ dθ = GR 3.621#3, Lea Ch2 P29d
0 cos θ 0 (2n + 2)!! 2

Thus


V0 π [ (−1)n  ρ 2n (2n + 1)!!
∞
∂Φ 
= −
∂z z=0 a 2 n=0 [n!]2 2a 2n+1 (n + 1)!
V0 π [ (−1)n  ρ 2n (2n + 1)!!
∞
= −
a 4 n=0 (n!)2 2a 2n (n + 1)!

and so the charge density on the plane is

∞  n
V0 π [ (−1)n ρ2 (2n + 1)!!
σ = ε0 Ez = ε0
a 4 n=0 (n!)2 8a2 (n + 1)!

The plot shows the series up to n = 40.

1.0
sigma a/epsV

0.8

0.6

0.4

0.2

0.0
1 2 3 4 5 6 7 8 9 10
rho/a
-0.2

Convince yourself that this agrees with the field line diagram. (The dashed line
is the potential.)

16