Chapter 4

Elliptic Differential Equations

BY

Prof. D. C. Sanyal

Retired Professor Of Mathematics

University Of Kalyani
West Bengal, India
E-mail : dcs [email protected]

1
Module-3: Solution of Three-Dimensional Laplace’s
Equation by Separation of Variables Method

1. Introduction
In this module we proceed to find the solution of three-dimensional Laplace’s
equation in different system of coordinates and apply the solutions to some specific
problems.

2. Solution of Three-Dimensional Laplace’s Equation

I. Cartesian coordinates (x, y, z)
Laplace’s equation in three-dimensional cartesian coordinates (x, y, z) is given by
∂2 ψ ∂2 ψ ∂2 ψ
∇2 ψ = + + = 0. (1)
∂x2 ∂y 2 ∂z2
To solve this equation by separation of variables technique, we put

ψ(x, y, z) = X(x)Y (y)Z(z). (2)

Substituting this in (1), we have

1 d 2Y 1 d 2Z 1 d 2X
+ = − = λ21 , (say) (3)
Y ∂y 2 Z dz2 X dx2
where λ21 is a separation constant. Thus, we have
d 2X
+ λ21 X = 0
dx2
whose solution is X(x) = c1 cos λ1 x + c2 sin λ1 x.
Again, from (3), it follows that
1 d 2Z 2 1 d 2Y
− λ 1 = − = λ22 , (say)
Z dz2 Y dy 2
d 2Y 2 d 2Z
so that + λ2 Y = 0, − λ23 Z = 0
dy 2 dz2
where λ23 = λ21 + λ22 and λ22 is another separation constant. Solutions of these two
equations are Y (y) = c3 cos λ2 y + c4 sin λ2 y, Z(z) = c5 cosh λ3 z + c6 sinh λ3 z.
Hence the general solution of the equation (1) is

ψ(x, y, z) = (c1 cos λ1 x + c2 sin λ1 x)(c3 cos λ2 y + c4 sin λ2 y)(c5 cosh λ3 z + c6 sinh λ3 z) (4)

2
Chapter 4 Elliptic Differential Equations

Example 1: Find the potential function ψ(x, y, z) in a rectangular box defined by 0 ≤

x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c, if the potential is zero on all sides and the bottom while
ψ = f (x, y) on the top of the box.

Solution: We take the surfaces z = 0 and z = c as the bottom and the top of the box.
Then the boundary conditions of the problem are

ψ(0, y, z) = ψ(a, y, z) = 0, ψ(x, 0, z) = ψ(x, b, z) = 0,

ψ(x, y, 0) = 0, ψ(x, y, c) = f (x, y).

Now, consider the solution (4) of the Laplace;s equation (1). Then the boundary con-
mπ
ditions ψ(0, y, z) = 0 and ψ(a, y, z) give respectively c1 = 0 and λ1 a = mπ, i.e. λ1 = a ,
(m = 1, 2, · · · ), while the conditions ψ(x, 0, z) = 0 and ψ(x, b, z) = 0 give c3 = 0 and
nπ
λ2 = b , (n = 1, 2, ...) respectively. Also the condition ψ(x, y, 0) = 0 = 0 leads to c5 = 0.
Thus, we have

ψ(x, y, z) = c2 sin λ1 x · c4 sin λ2 y · c6 sinh λ3 z,

i.e. ψ(x, y, z) = A sin λ1 x · sin λ2 y · sinh λ3 z, where A = c2 c4 c6

√ √
2 2
Noting that λ3 = λ1 + λ2 = π ma2 + nb2 = λmn , say, the solution is given, by the use of
2 2

superposition principle, as
∑∑ ( ) ( )
mπx nπy
ψ(x, y, z) = Amn sin sin sinh (λmn z) . (1)
m n
a b

Finally, using the boundary condition ψ(x, y, c) = f (x, y), we have

∑∑ ( ) ( )
mπx nπy
f (x, y) = Amn sin sin sinh (λmn c)
m n
a b

which is a double Fourier sine series and, therefore, it follows that

∫ ∫ ( ) ( )
4 a b mπx nπy
Amn sinh (λmn c) = f (x, y) sin sin (2)
ab 0 0 a b

Thus the required solution is given by (1) in which Amn is obtained from (2).

II. Cylindrical coordinates (r, θ, z)

Laplace’s equation in cylindrical coordinates (r, θ, z) is

∂2 ψ 1 ∂ 1 ∂2 ψ ∂2 ψ
∇2 ψ = + + + = 0. (5)
∂r 2 r ∂r r 2 ∂θ 2 ∂z2

3
 Chapter 4 Elliptic Differential Equations

We assume a solution of the form

ψ = (r, θ, z) = R(r)Θ(θ)Z(z). (6)

Substituting (6) into (5), we get

( )
1 d 2 R 1 dR 1 d 2Θ 1 d 2Z
+ + + =0 (7)
R dr 2 r dr r 2 Θ dθ 2 Z dz2

1 d2Z
Putting Z dz2 = m2 , we have, by solving, Z = c5 emz + c6 e−mz .
1 d2Θ d2Θ
Also, taking Θ dθ 2 = −n2 , i.e. dθ 2
+ n2 Θ = 0, we find Θ = c3 cos nθ + c4 sin nθ.
Thus, from (7), it follows that
( )
d 2 R 1 dR 2 n
2
+ + m − 2 R=0
dr 2 r dr r

whose solution is R(r) = c1 Jn (mr) + c2 Yn (mr), where Jn and Yn are Bessel functions of
order n of first and second kind respectively.
Thus the most general solution of the equation (5) is

ψ(r, θ, z) = {c1 Jn (mr) + c2 yn (mr)} (c3 cos nθ + c4 sin nθ)(c5 emz + c6 e−mz ). (8)

Example 2: A homogeneous thermally conducting cylinder occupies the region 0 ≤

r ≤ a, 0 ≤ θ ≤ 2π, 0 ≤ z ≤ h, where (r, θ, z) are cylindrical coordinates. The top z =
h and the lateral surface r = a are held at 0o , while the base z = 0 is held at 100o .
Assuming that there are no sources of heat generation within the cylinder, find the
steady-temperature distribution within cylinder.

Solution: The temperature ψ must be a single-valued continuous function and for

steady-state, ψ satisfies Laplace’s equation inside the cylinder. Thus we are to solve
the following boundary problem;

∇2 ψ = 0, 0 ≤ r ≤ a, 0 ≤ θ ≤ 2π, 0 ≤ z ≤ h

subject to ψ = 0o at z = h and at r = a,

ψ = 100o at z = 0.

The general solution of the Laplace equation in cylindrical coordinates system (r, θ, z)
is given by (8). Now since the face z = 0 is maintained at 100o and since the other
face z = h and the lateral surface r = a are held at 0o , so the temperature ψ inside the

4
Chapter 4 Elliptic Differential Equations

cylinder must be independent of θ. Also, we note that Yn (mr) is undefined at r = 0.

This is possible only if n = 0 in the general solution. Thus

ψ(r, z) = J0 (mr)(Aemz + Be−mz ).

Using the boundary condition ψ = 0 at z = h, we have B = −Ae2mh , so that

[ ]
ψ(r, z) = AJ0 (mr)emh em(z−h) − e−m(z−h)

i.e. ψ(r, z) = BJ0 (mr) sinh{m(z − h)}, where B = 2Aemh .

Also the boundary condition ψ = 0 at r = a implies that J0 (ma) = 0, which has infinitely
many positive roots ξk , say, where k = 1, 2, 3, · · · and so m = ξk /a. Thus using the
superposition principle, the solution is
∑
∞
ψ(r, z) = Bk J0 (ξk r/a) sinh{ξk (z − h)/a}.
k=1

Now the boundary condition ψ = 100o at z = 0 gives

∑
∞
100 = Bk J0 (ξk r/a) sinh{−ξk h/a}.
k=1

which is a Fourier-Bessel series. Multiplying both sides by rJ0 (ξl r/a) and then inte-
grating, we get
∫ a ∑
∞ ∫ a
100 rJ0 (ξl r/a)dr = Bk sinh{−ξk h/a} rJ0 (ξk r/a)J0 (ξl r/a)dr
0 k=1 0

and using the orthogonal property of Bessel function, viz.


∫a 
 a2 2

 2 Jn+1 (αi ) for i = j
xJn (αi x)xJn (αi x)dx = 

0 
 0 for i , j,

where αi , αj are the roots of the equation Jn (x) = 0, it follows that

∫ a
a2 2
100 rJ0 (ξk r/a)dr = Bk sinh{−ξk h/a} · J (ξ )
0 2 1 k
∫
Also using the relation xJ0 (x)dx = J1 (x), we have

a2 a2 2
100 · J (ξ k ) = B k sinh(−ξ k h/a) · J (ξ )
ξk2
1
2 1 k
200
so that Bk = .
ξk sinh(−ξk h/a)J1 (ξk )

5
 Chapter 4 Elliptic Differential Equations

Hence the required temperature distribution is

∑
∞
J0 (ξk r/a) sinh{ξk (z − h)/a}
ψ(r, z) = 200 ,
ξk J1 (ξk ) sinh(−ξk h/a)
k=1

where ξk , (k = 1, 2, · · · ), are the roots of the equation J0 (ξ) = 0.

III. Spherical polar coordinates (r, θ, ϕ)

Laplace’s equation in spherical polar coordinates (r, θ, ϕ) is
( ) ( )
2 ∂ 2 ∂ψ 1 ∂ ∂ψ 1 ∂2 ψ
∇ ψ= r + sin θ + = 0. (9)
∂r ∂r sin θ ∂θ ∂θ sin2 θ ∂ϕ2
To solve this equation, we put

ψ(r, θ, ϕ) = R(r)F(θ, ϕ). (10)

Substituting this in (9), we get

( ) { ( ) }
1 d 2 dR 1 ∂ ∂F 1 ∂2 F
r =− sin θ + = −λ, say, (11)
R dr dr F sin θ ∂θ ∂θ sin θ ∂ϕ2
where λ is separation constant. Then
( )
d 2 dR
r + λR = 0 (12)
dr dr
{ ( ) }
1 ∂ ∂F 1 ∂2 F
and sin θ + − λF = 0. (13)
sin θ ∂θ ∂θ sin θ ∂ϕ2
Using the transform R = r m and λ = −n(n + 1) in (12), the auxiliary equation is

m(m − 1) + 2m − n(n + 1) = 0 ⇒ m = n, −(n + 1).

Thus R(r)c1 r n + c2 /r n+1 (14)

Also, the equation (13) with λ = −n(n + 1) is

{ ( ) }
1 ∂ ∂F 1 ∂2 F
sin θ + + n(n + 1)F = 0 (15)
sin θ ∂θ ∂θ sin θ ∂ϕ2
For solution of this equation, we put

F(θ, ϕ) = Θ(θ)Φ(ϕ) (16)

so that from (15)

{ ( ) }
1 d 2 Φ sin θ d dΘ
− = sin θ + n(n + 1)Θ sin θ = m2 , say,
2
(17)
Φ dϕ2 Θ dθ dθ

6
Chapter 4 Elliptic Differential Equations

where m2 is another separation constant. Then

d 2Φ
+ m2 Φ = 0 =⇒ Φ(ϕ) = c3 cos mϕ + c4 sin mϕ, (18)
dθ 2
provided m , 0. If m = 0, we have the axis-symmetric case. Again, from (17), we get
{ }
d dΘ
sin θ (sin θ) + n(n + 1)Θ · sin θ − m2 Θ = 0
2
dθ dθ

which, on putting cos θ = µ, reduces to

{ }
2 d 2Θ dΘ m2
(1 − µ ) 2 − 2µ + n(n + 1) − Θ=0
dµ dθ 1 − µ2
This is the well-known associated Legendre equation and has the general solution

Θ(µ) = c5 Pnm (µ) + c6 Qnm (µ), −1 ≤ µ < 1

i.e. Θ(θ) = c5 Pnm (cos θ) + c6 Qnm (cos θ), −1 ≤ θ < π, (19)

where Pnm and Qnm are associated Legendre functions of the first and second kind re-
spectively. Since the function Qnm (cos θ) has a singularity at θ = 0, so we choose c6 = 0.
Thus the general solution of Laplace’s equation (9) is obtained from (10), (14), (16),
(18) and (19) in the form
∑ ∞ (
∞ ∑ )
n c2
ψ(r, θ, ϕ) = c1 r + (c3 cos mϕ + c4 sin mϕ)Pnm (cos θ) (20)
r n+1
m=0 n=0

In particular, for the axisymmetric case, the general solution is

∑ ∞ (
∞ ∑ )
c2
ψ(r, θ) = n
c1 r + n+1 Pnm (cos θ) (21)
r
m=0 n=0

Problem 1 (Interior Dirichlet problem for a sphere):

The interior Dirichlet problem for a sphere is defined as follows:
To find the value of ψ at any point in the interior of the sphere r = a such that

∇2 ψ = 0, 0 ≤ r ≤ a, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π

subject to ψ(a, θ, ϕ) = f (θ, ϕ) on r = a.

Since r = 0 is a point within the sphere, we must put c2 = 0 in the general solution
(20). Thus, we have
∑
∞ ∑
∞
ψ(r, θ, ϕ) = r n [Amn cos mϕ + Bmn sin mϕ] Pnm (cos θ), 0 ≤ r ≤ a. (22)
m=0 n=0

7
 Chapter 4 Elliptic Differential Equations

where Amn and Bmn are new constants after adjustment. The use of the boundary
condition ψ(a, θ, ϕ) = f (θ, ϕ), we get
∑
∞ ∑
∞
f (θ, ϕ) = an [Amn cos mϕ + Bmn sin mϕ] Pnm (µ), (23)
m=0 n=0

where µ = cos θ and it is assumed that f (θ, ϕ) is expansible is series of associated

Legendre function.
Now, multiplying both sides of (23) by Pnm (µ) cos mϕ and performing double inte-
gration with respect to µ, (−1 ≤ µ ≤ 1), and ϕ, (0 ≤ ϕ ≤ 2π), we get
∫ 1 ∫ 2π ∫ 2π [∫ 1 ]
f (θ, ϕ)Pnm (µ) cos mϕ dµ dϕ = a Amn n 2
cos mϕ {Pnm (µ)}2 dµ dϕ
−1 0 0 −1
2πan Amn · (m + n)!
= .
(2n + 1) · (n − m)!

where, we have used the orthogonality relation of associate Legendre function. Thus,
we have ∫ 2π
(2n + 1) · (n − m)!
Amn = f (θ, ϕ)Pnm (µ) cos mϕ dµ dϕ.
2πan · (n + m)! 0
Similarly, multiplying both sides of (23) by Pnm (µ) sin mϕ and integrating as above, we
obtain ∫ 2π
(2n + 1) · (n − m)!
Bmn = f (θ, ϕ)Pnm (µ) sin mϕ dµ dϕ.
2πan · (n + m)! 0
Substituting the above values of Amn and Bmn in (22), we get
( ) ∫ ∫
1 ∑ ∑ (2n + 1) · (n − m)! r n 1 2π
∞ ∞
ψ(r, θ, ϕ) = f (η, χ)Pnm (µ)Pnm (µ′ ) ×
2π ·(n + m)! a −1 0
n=0 m=0
[cos mχ cos mη + sin mχ sin mη] dηdχ
∑
∞ ∑
∞ ( )n ∫ 1 ∫ 2π
1 (2n + 1) · (n − m)! r
i.e ψ(r, θ, ϕ) = f (η, χ)Pnm (cos χ)Pnm (cos η) ×
2π ·(n + m)! a −1 0
n=0 m=0
cos m(χ − η)dηdχ

which is the desired solution of interior Dirichlet problem for a sphere.