0% found this document useful (0 votes)
2 views

bessel-examples

The document discusses the Dirichlet problem for a cylinder, focusing on solving the Laplace equation in cylindrical coordinates with specific boundary conditions. It details the separation of variables method, leading to Bessel functions as solutions, and outlines the process for finding eigenvalues and eigenfunctions for the Laplace equation in a cylinder. Additionally, it presents an exercise related to the radial eigenvalue problem for the Laplace equation in a ball.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
2 views

bessel-examples

The document discusses the Dirichlet problem for a cylinder, focusing on solving the Laplace equation in cylindrical coordinates with specific boundary conditions. It details the separation of variables method, leading to Bessel functions as solutions, and outlines the process for finding eigenvalues and eigenfunctions for the Laplace equation in a cylinder. Additionally, it presents an exercise related to the radial eigenvalue problem for the Laplace equation in a ball.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

More examples with Bessel functions

A. Eremenko
March 21, 2021

1. Dirichlet problem for a cylinder

This problem describes time-independent solutions of the wave and heat


equations in a cylinder.
A cylinder is described in cylindrical coordinates by inequalities

0 ≤ r ≤ L, 0 ≤ z ≤ H,

where L is the radius and H is the height. We want to solve the Laplace
equation in cylindrical coordinates
1 1
urr + ur + 2 uθθ + uzz = 0, (1)
r r
under the boundary conditions

u(r, θ, 0) = f (r, θ), (2)

u(r, θ, H) = g(r, θ), (3)


u(L, θ, z) = h(θ, z). (4)
where f, g, h are some given functions. Besides, we have the periodicity
conditions

u(r, −π, z) = u(r, π, z), uθ (r, −π, z) = uθ (r, π, z), (5)

which come from the nature of cylindrical coordinates.


As always, the problem is split into 3 problems:
a) f = g = 0,

1
b) g = h = 0,
c) f = h = 0,
and the complete solution is the sum of three solutions of a),b),c).
The first step is common for all three problems: separation of the vari-
ables. We look for solutions of the form u(r, θ, z) = R(r)Θ(θ)Z(z). Plugging
this form, we obtain
1 1
R00 ΘZ + R0 ΘZ + 2 RΘ00 Z + RΘZ 00 = 0.
r r
The general rule is to separate those variables first for which the boundary
conditions are homogeneous. So for all problems a), b), c) we separate θ part
first, since the conditions (5) are homogeneous. So we rewrite our equation
as
R00 R0 Z 00 Θ00
r2 + r + r2 =− .
R R Z Θ
Both sides must be constant, say equal to µ, and for θ we obtain

Θ00 + µ2 Θ = 0

with periodic boundary conditions. As we know, this implies that µ2 = m2


where m is a non-negative integer, and to each positive m correspond two
eigenfunctions, which can we written in the real form

Θm (θ) = am cos(mθ) + bm sin(mθ), m = 0, 1, 2, . . . ,

or in the complex form

Θm = cm eimθ , −∞ < m < +∞.

(To m = 0 corresponds only one eigenfunction, a constant).


The remaining equation is

R00 R0 Z 00
r2 + r + r2 − m2 = 0,
R R Z
and we separate r from z:

R00 1 R0 m2 Z 00
+ − 2 =− . (6)
R rR r Z

2
In problem a),
the boundary conditions for z are homogeneous, (they come from (2), (3)
with f = g = 0), so we consider the z part first. We have

Z 00 + λZ = 0, Z(0) = Z(H) = 0.

This is a familiar problem: the eigenvalues ares

λn = π 2 n2 /H 2 , (7)

and eigenfunctions are


πnz
 
Zn (z) = sin , n = 1, 2, 3, . . . .
H
The remaining equation in r then becomes

R00 1 R0 m2
+ − 2 − λn = 0,
R rR r
or
R00 R0  
r2 + r − λn r2 + m2 R = 0.
R R
This looks like the equation reducible to the Bessel equation, except the
wrong sign of λn (compare with eigenvalue problem for a disk). So it is
reduced to Bessel by the change of the variable
q 
πnr
  √
R(r) = y −λn r = y i , i= −1, (8)
H
where y satisfies the Bessel equation

x2 y 00 + xy 0 + (x2 − m2 )y = 0. (9)

The general solution of this last equation is

y(x) = aJm (x) + bYm (x),

where Jm and Ym are Bessel functions of the first and second kind, and since
our function R is supposed to be bounded, while Ym (x) → −∞ as x → 0,
we must take b = 0.

3
So we obtain:
πnr
 
Rm,n (r) = Jm i .
H
Combining Rm,n , Θn , Zm , we obtain a series for problem a):
πnr imθ πnz
X  
u(r, θ, z) = cm,n Jm i e sin .
m,n H H

To satisfy the boundary condition (4) we plug r = L, and obtain a double


Fourier series

cm,n Jm (iπnL/H)eimθ sin(πnz/H).


X
h(θ, z) =
m,n

and Fourier formulas give us


Z π Z H
1
cm,n = h(θ, z)e−imθ sin(πnz/H)dzdθ,
πHJm (iπnL/H) −π 0
where we used the normalization factors
Z π Z H
imθ 2 H
|e | dθ = 2π, and sin2 (πnz/H)dz = .
−π 0 2
This solves problem a).
Remark. Functions i−ν Jν (ix) are sometimes called modified Bessel func-
tions and the standard notation for them is Iν . The power of i multiple is
added to make them real on the positive ray.

Problems b), and c).

We return to (6). This time the problem in r is homogeneous, so we


consider the r-part first. Denoting the common value of the RHS and LHS
of (6) by −λ2m we obtain the r-part:

r2 R00 + rR0 + (λ2m r2 − m2 )R = 0, (10)

with the boundary condition

R(L) = 0, (11)

which comes from (4) with h = 0. This is reduced to Bessel equation by


setting R = y(λm r), so that y will satisfy (9), and taking into account that

4
R(0) must be finite, we obtain R = Jm (λm r). Now the boundary condition
(11) implies
λm,k = xm,k /L,
where xm,k is the k-th zero of Jm . Notice that λm,k are all real since Bessel
functions of order > −1 have only real roots, and without loss of generality
we may consider only positive zeros since only λ2m enters our equation (10).
Once λm,k is found, it remains to solve the z part in (6). We have

Z 00 − λ2m,k Z = 0,

whose general solution can be written as

Z(z) = a cosh(λm,k z) + b sinh(λm,k z).

So the general solution satisfying homogeneous boundary conditions for the


cases b), c) has the form

Jm (λm,k r)eimθ (am,k cosh(λm,k z) + bm,k sinh(λm,k z)) .


X
um,k (r, θ, z) =
m,k

To satisfy the boundary conditions (3), (4), we plug z = 0 or z = H and use


Fourier-Bessel formulas.
For example, for problem c), the boundary condition at z = 0 is zero, so
we set am,k = 0, and obtain

bm,k Jm (xm,k r/L)eimθ sinh (xm,k H/L) ,


X
g(r, θ) =
m,k

and Fourier formulas give


Z π Z L
1
bm,k = g(r, θ)Jm (xm,k r/L)e−imθ rdrdθ.
πL2 Jm+1 (xm,k )2 sinh(xm,k H/L) −π 0

Notice that we integrate rdr because Bessel functions are orthogonal with
weight r, and we used the formula for the square norm
Z L
2 L2
Jm (xm,k r/L)rdr = Jm+1 (xm,k ),
0 2
which is formula (29) in the handout “Bessel functions”.

2. Eigenvalue problem for Laplace equation in a cylinder

5
To solve heat of wave equation in a cylinder, we need the following eigen-
value problem
∆u + λ2 u = 0,
with some boundary conditions. (I denoted the eigenvalue by λ2 for conve-
nience of some further formulas). Let us take for example,
u(x) = 0 on the boundary of the cylinder.
In cylindrical coordinates this becomes
1 1
urr + ur + 2 uθθ + uzz + λ2 u = 0,
r r
u(L, θ, z) = 0, −π ≤ θ ≤ π, 0 ≤ z ≤ H,
u(r, θ, 0) = u(r, θ, H) = 0, 0 ≤ r ≤ L, −π ≤ θ ≤ π.
plus the periodicity condition (5) which comes from the nature of cylindrical
coordinates.
As always, the first step is separation of variables, and plugging u = RΘZ
we obtain
1 1
R00 ΘZ + R0 ΘZ + 2 RΘ00 Z + RΘZ 00 + λ2 RΘZ = 0.
r r
Separate θ first:
00
2R R0 2Z
00
2 2 Θ00
r +r +r +λ r =− . (12)
R R Z Θ
We conclude that the separation constant must be of the form m2 , where m
is a non-negative integer, and eigenfunctions are
Θ±m (θ) = e±imθ , m = 0, 1, 2, . . . .
Replacing Θ00 /Θ in (12) by −m2 and dividing on r2 we obtain
R00 1 R0 m2 2 Z 00
+ − 2 +λ =− ,
R rR r Z
so the common value of the RHS and LHS is a constant, and we obtain a
familiar problem for z variable, with boundary conditions Z(0) = Z(H) = 0,
from which we conclude that this separation constant must be π 2 n2 /H 2 and
πnz
Zn (z) = sin , n = 1, 2, 3, . . . .
H
6
Now for the r part we obtain

π 2 n2 2
! !
2 00 0 2
r R + rR + λ − r − m2 R = 0, (13)
H2

and this is reduced to the Bessel equation by the change of the variable
q
R(r) = f ( λ2 − π 2 n2 /H 2 r).

Then as always, we conclude that the solution must be a scaled Bessel func-
tion Jm since the second linear independent solution of the Bessel equation is
not bounded near 0. Moreover, we know that all roots of the Bessel function
are real, so we must have
λ2 ≥ π 2 n2 /H 2 . (14)
If equality holds in (14), then (13) is an Euler equation whose linearly inde-
pendent solutions are R(r) = rm and R(r) = r−m , when m = 0 the second
solution is log r. In any case, the second solution is infinite at 0, so it must
be rejected. But then the solution R(r) = rm cannot satisfy the boundary
condition R(L) = 0. Thus we must have a strict inequality in (14).
Thus q
R(r) = Jm ( λ2 − π 2 n2 /H 2 r),
and the boundary condition R(L) = 0 implies

λ2m,n,k = (xm,k /L)2 + (πn/H)2 . (15)

So we found eigenvalues, and eigenfunctions are


πiz
um,n,k (r, θ, z) = eimθ sin Jm (λm,n,k r).
H
Here m is any integer (positive or negative), while n and k are positive
integers.

Exercise. Radial eigenvalue problem for Laplace equation in a ball.

Consider the eigenvalue problem

∆u + λ2 u = 0

7
in a ball |x| ≤ L in 3-space, with zero boundary condition. Here one has to
use spherical coordinate’s (r, φ, θ) which will be fully discussed later. But let
us restrict the problem to eigenfunctions which depend on the radius r = |x|
only. Find all such eigenfunctions and corresponding eigenvalues.
For radial functions in spherical coordinates, the expression of the Lapla-
cian in the following
2
∆r u = urr + ur .
r
Here subscript r in ∆r means that we drop all terms which involve differen-
tiation with respect to φ, θ.
So we want to solve the equation
2
y 00 + y 0 + λ2 y = 0, (16)
r
with boundary conditions that y(0) is finite, and
y(L) = 0.
I suggest two ways of solving this problem.
1. Try to find a simple change of the variable which reduces this equation to
a Bessel equation.
2. If you cannot do this, try to solve it using power series. The associated
Euler equation is
r2 y 00 + 2ry 0 = 0,
where we dropped the term λ2 r2 . The corresponding characteristic equation
is
ρ2 + ρ = 0,
so solutions of Euler’s equation are 1 and 1/r. This suggests that a solution
of (16) which is bounded at 0 must be of the form

an r n .
X
y(r) =
n=0

Plug this form, determine the coefficients an explicitly, and recognize the
resulting function y, it turns out to be elementary. This permits you to find
eigenvalues and eigenfunctions.
Once you have an explicit solution of (16), you can return to step 1, and
find a simple change of the variable which reduces it to Bessel equation.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy