Laplace Equation

we shall discuss the methods to solve the Laplace equation,

∆u = 0, in Ω,

with various cases of domains Ω and the boundary conditions of the Dirichlet, Neumann, or
Robin type boundary conditions on Γ. Here ∆ = ∇2 = ∇ · ∇ is the Laplace operator and
the domain Ω, considered in R2 or R3 , is enclosed by a piecewise smooth boundary Γ. We
first consider,

∆u(x, y) = uxx (x, y) + uyy (x, y) = 0, (x, y) ∈ R = {(x, y) : 0 < x < a, 0 < y < b},

u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, u(x, 0) = 0, u(x, b) = f (x), 0 ≤ x ≤ a.

By putting u(x, y) = X(x)Y (y) in the Laplace equation and separating the variables, we get

X ′′ Y ′′
=− = λ.
X Y
For X we get the SLP, X ′′ − λX = 0, 0 < x < a, with the BCs X(0) = 0 and X(a) = 0.
This SLP will have a non-zero solution for λ < 0. Hence we take λ = −µ2 , µ > 0. We get
the eigenvalues λn = −n2 π 2 /a2 and the eigenfunctions Xn (x) = sin(nπx/a). For Y we get
the SLP Y ′′ − (n2 π 2 /a2 )Y = 0 which has the solution

Yn (y) = Cn e(nπ/a)y + Dn e−(nπ/a)y .

The Bc Y (0) = 0 implies Dn = −Cn . Hence we may take The index often starts from n=1
because n=0 yields trivial or
nπy constant solutions that may not
Yn (x) = Bn sinh .
a contribute in meaningful ways to
the physical solution.
Hence ∞
X nπy nπx The summation form allows for constructing
u(x, y) = Bn sinh sin . solutions as linear combinations of
a a
n=1 orthogonal eigenfunctions.

1
The index often starts from n=1 because =0
yields trivial or constant solutions that may not
contribute in meaningful ways to the physical
solution.
Finally, to determine Bn , we use the BC u(x, b) = f (x), 0 ≤ x ≤ a, which gives,
nπb nπx
f (x) = Bn sinh sin .
a a
Hence Z a
nπb 2 nπx
Bn sinh = f (x) sin dx
a a 0 a
which gives Z a
2 nπx
Bn = f (x) sin dx.
a sinh(nπb/a) 0 a
Laplace Equation in Symmetric Domains Now we consider the equation ∆u = 0 in a
circular domain Ω = {(r, θ) : 0 < r < a, 0 ≤ θ ≤ 2π}. We have U (r, θ) = u(x(r, θ), y(r, θ)).
Using the chain rule and x2 + y 2 = r2 , tan−1 (y/x) = θ, we have

ux = Ur rx + Uθ θx , uy = Ur ry + Uθ θy ,

uxx = Urr rx2 + Ur rxx + 2Urθ rx θx + Uθ θxx + Uθθ θx2 ,

uyy = Urr ry2 + Ur ryy + 2Urθ ry θy + Uθ θyy + Uθθ θy2 .
x 1 xrx y2 y 2xy
rx = , rxx = − 2 = 3 , θx = − 2 2
, θxx = 2 ,
r r r r x +y (x + y 2 )2
y 1 yry x2 x 2xy
ry = , ryy = − 2 = 3 , θy = 2 , θ yy = − .
r r r r x + y2 (x2 + y 2 )2
Hence

∆u = Urr (rx2 + ry2 ) + Ur (rxx + ryy ) + Uθ (θxx + θyy ) + 2Urθ (rx θx + ry θy ) + Uθθ (θx2 + θy2 )
1 1
= Urr + Ur + 2 uθθ .
r r
Now, we consider the BVP,

∆U (r, θ) = 0, (r, θ) ∈ {(r, θ) : 0 < r < a, 0 ≤ θ ≤ 2π}, U (a, θ) = f (θ).

We take U (r, θ) = R(r)Θ(θ) and obtain

r2 R′′ + rR′ Θ′′
=− = λ.
R Θ
For R we get
r2 R′′ + rR′ − λR = 0.
This is a Cuachy-Eluer equation and we take R(r) = rn . The n(n−1)+n−λ = 0 which gives
λ = n2 . Hence we may take λ = µ2 , and µ = ±n, n = 0, 1, 2, . . . . Thus Rn (r) = Cn rn +Dn r−n .
For Θ we get Θn (θ) = An cos nθ+Bn sin nθ. For the interior problem, we discard the solutions
r−n , i.e., we take Dn = 0 for all n, and hence
∞
X
U (r, θ) = A0 + [An cos nθ + Bn sin nθ] rn .
n=1

2
Now, U (a, θ) = f (θ) gives
∞
X
f (θ) = A0 + an [An cos nθ + Bn sin nθ].
n=1

Hence Z π
1
A0 = f (θ) dθ,
2π −π
Z π Z π
n 1 n 1
a An = f (θ) cos nθ dθ, a Bn = f (θ) sin nθ dθ.
π −π π −π

We may also write

∞
X  r n
U (r, θ) = A0 + [A∗n cos nθ + Bn∗ sin nθ] , (1)
n=1
a
Z π
1
A0 = f (θ) dθ, (2)
2π −π
π
1 π
Z Z
1
A∗n = f (θ) cos nθ dθ, ∗
Bn = f (θ) sin nθ dθ, A∗n = an An , ; Bn∗ = an Bn . (3)
π −π π −π
If we have the Neumann condition (∂U/∂n) = (∂U/∂r)(a, θ) = g(θ), instead of the
Dirichlet condition, then
∞
X
g(θ) = [nan−1 An cos nθ + nan−1 Bn sin nθ],
n=1
Z π Z π
1 1
An = g(θ) cos nθ dθ, Bn = g(θ) sin nθ dθ.
πnan−1 −π πnan−1 −π

Since, by the divergence theorem,

ZZ Z
2
∇ U (r, θ) rdrdθ = 0 = ∇U · n̂ ds.
Ω Γ

Now,  
xUr yUθ yUr xUθ x y 
∇U = − 2 , + 2 , n̂ = , , ds = adθ.
r r r r r r
Hence Z π
g(θ) dθ = 0.
−π

This is the compatibility condition which g(θ) must satisfy.

Poisson’s Formula By 2π-periodicity of f (θ), We also have
Z 2π
1
A0 = f (θ) dθ, (4)
2π 0

3
 Z 2π Z 2π
1 1
A∗n = f (θ) cos nθ dθ, Bn∗ = f (θ) sin nθ dθ, (5).
π 0 π 0

Using (4) and (5) in (1), we get

" ∞
#
Z 2π
1 1 X  r n
U (r, θ) = f (ϕ) + cos n(θ − ϕ) dϕ.
π 0 2 n=1 a

We have the following identity for z = reiθ :

∞
1 X n 1 z 1 − r2 + 2ir sin θ
+ z = + = .
2 n=1 2 1−z 2(1 − 2r cos θ + r2

Hence ∞
a2 − r 2
 
1 X  r n 1
+ cos n(ϕ − θ) = . (6)
2 n=1 a 2 a2 − 2ar cos(θ − ϕ) + r2
Using (6) we get the Poisson integral formula:
Z 2π
1 a2 − r 2
U (r, θ) = f (ϕ) dϕ. (7)
2π 0 a2 − 2ar cos(ϕ − θ) + r2
Remark: For a ≤ b ≤ c, we have
Z c Z b Z c
f (x) dx = f (x) dx + f (x)dx.
a a b

Choose a unique k such that

s − π ≤ 2πk ≤ s + π.
Then Z s+π Z 2πk Z s+π
F (x) dx = f (x) dx + f (x) dx
s−π s−π 2πk
Z 2πk+2π Z s+π
= f (x) dx + f (x) dx
s+π 2πk
Z 2πk+2π Z 2π
= f (x) dx = f (x) dx.
2πk 0
If we consider

∇2 U (r, θ) = 0, 1 < r < 3, U (1, θ) = sin θ, U (3, θ) = 0,

then we take ∞  
X Dn n
U (r, θ) = A0 + [An cos nθ + Bn sin nθ] Cn r + n .
n=1
r
∞
X
sin θ = A0 + [An cos nθ + Bn sin nθ] [Cn + Dn ]
n=1

4
 ∞  
X
n Dn
0 = A0 + [An cos nθ + Bn sin nθ] Cn 3 + n .
n=1
3
Z π
1
A0 = sin θ dθ = 0.
2π −π
We have An = 0 for n = 0, 1, 2, . . . and Cn = Dn = 0 for n = 2, 3, . . . and
D1
B1 (C1 + D1 ) = 1, 3C1 + = 0.
3
Hence  
9
U (r, θ) = −B1 C1 sin θ r − .
r
For r = 1 we get
sin θ = 8B1 C1 sin θ.
Hence B1 C1 = 1/8. Thus,
9 − r2
 
U (r, θ) = sin θ. (8)
8r
Verification: Clearly, U given by (8) satisfies the BCs. To check that ∇2 U (r, θ) = 0 in
1 < r < 3, 0 ≤ θ ≤ 2π, it is enough to check that V = (sin θ)/r (or v = y/(x2 + y 2 )) and
W = r sin θ (or w = y) satisfy

Urr + (Ur /r) + (Uθθ /r2 ) = 0 ( or uxx + uyy = 0). (9)

We have
sin θ 2 sin θ sin θ
Vr = − 2
, Vrr = 3
, Vθθ = − .
r r r
Wr = sin θ, Wrr = 0, Wθθ = −rsin θ.
We see that V and W satisfy (9). Since U = (9V + W )/8, it follows by linearity of (9) that
U also satisfies (9).
If we take the Neumann conditions Ur (1, θ) = sin θ, and Ur (3, θ) = 0, then we get all
An = 0, except A1 and all Bn = 0. Hence
D1
B1 (C1 − D1 ) = 1, C1 − = 0.
9
Hence D1 = 9C1 and B1 C1 = −1/8. Thus,
 
1 9
U (r, θ) = − r+ sin θ.
8 r

5
Laplace Operator in Spherical Coordinates
Referring to the figure,

we have
x = r cos θ sin ϕ, y = r sin θ sin ϕ, z = r cos ϕ.
Also, p
r 2 = x2 + y 2 + z 2 , θ = tan−1 (y/x), ϕ = tan−1 ( x2 + y 2 /z),
where square root of non-negative number is taken non-negative. Let

w(r, θ, ϕ) = u((x(r, θ, ϕ), y(r, θ, ϕ), z(r, θ, ϕ)).

Then

uxx + uyy + uzz = wrr (rx2 + ry2 + rz2 ) + wθθ (θx2 + θy2 + θz2 ) + wϕϕ (ϕ2x + ϕ2y + ϕ2z )

+2wrθ (rx θx + ry θy + rz θz ) + 2wrϕ (rx ϕx + ry ϕy + rz ϕz ) + 2wθϕ (θx ϕx + θy ϕy + θz ϕz )

wr (rxx + ryy + rzz ) + wθ (θxx + θyy + θzz ) + wϕ (ϕxx ϕyy + ϕzz ). (1)
Now,

rx = x/r, ry = y/r, rz = z/r, θx = −y/(x2 + y 2 ), θy = x/(x2 + y 2 ), θz = 0,

p
xz yz x2 + y 2
ϕx = p , ϕy = p , ϕz = − 2 ,
(x2 + y 2 + z 2 ) x2 + y 2 (x2 + y 2 + z 2 ) x2 + y 2 x + y2 + z2
1 x2 1 y2 1 z2
rxx = − 3 , ryy = − 3 , rzz = − 3 ,
r r r r r r
2xy 2xy
θxx = 2 2 2
, θyy = − 2 , θzz = 0,
(x + y ) (x + y 2 )2
2x2 x2
 
z
ϕxx = p 1− 2 − ,
(x2 + y 2 + z 2 ) x2 + y 2 x + y 2 + z 2 x2 + y 2

6
 2y 2 y2
 
z
ϕyy = p 1− − ,
(x2 + y 2 + z 2 ) x2 + y 2 x2 + y 2 + z 2 x2 + y 2
p
2z x2 + y 2
ϕzz = 2 .
(x + y 2 + z 2 )2
Thus,
1 1 1 1
rx2 + ry2 + rz1 = 1, θx2 + θy2 + θz2 = = 2 2 , ϕ2x + ϕ2y + ϕ2z = 2 = 2,
x2 +y 2 r sin ϕ 2
x +y +z 2 r

rx θx + ry θy + rz θz = rx ϕx + ry ϕy + rz ϕz = θx ϕx + θy ϕy + θz ϕz = 0,
2 z cot ϕ
rxx + ryy + rzz = , θxx + θyy + θzz = 0, ϕxx + ϕyy + ϕzz = p = 2 .
r 2 2 2 2
(x + y + z ) x + y 2 r
Putting these expressions in (1), we get

2 1 1 cot ϕ
uxx + uyy + uzz = wrr + wr + 2 2 wθθ + 2 wϕϕ + 2 wϕ
r r sin ϕ r r
 
1 2 1 1
= 2 (r wr )r + (sin ϕ wϕ )ϕ + wθθ .
r sin ϕ sin2 ϕ
Laplace Equation in Spherical Domain Now we consider the interior BVP,
2 1 1 cot ϕ
= wrr + wr + 2 2 wθθ + 2 wϕϕ + 2 wϕ = 0, 0 < r < a,
r r sin ϕ r r

w(a, θ, ϕ) = f (θ, ϕ),

where w is assumed to be bounded. We consider the case of azimuthal symmetry (symmetry
in the variable θ) when f (θ, ϕ) = f (ϕ) and w(r, θ, ϕ) = w(r, ϕ). Taking w(r, ϕ) = R(r)Φ(ϕ)
we get    
1 d 2 dR 1 d dΦ
r =− sin ϕ = k.
R dr dr sin ϕ dϕ dϕ
We first consider the equation in |phi. We have
 
d dΦ
sin ϕ = −k sin ϕ Φ. (∗)
dϕ dϕ

Let ψ = cos ϕ (or ϕ = cos−1 ψ) and Ψ(ψ) = Φ(cos−1 ψ) = Φ(ϕ). Then

dΦ(ϕ) dΨ(ψ) dψ dΨ(ψ)

= · = − sin ϕ .
dϕ dψ dϕ dψ

Hence (∗) becomes  

d 2 dΨ(ψ)
(1 − ψ ) + kψ(ψ) = 0. (∗∗)
dψ dψ

7
The equation (∗∗) is of the from of Legendre’s equation if we take k = n(n + 1) with the
solution Ψn (ψ) = Pn (ψ), n = 0, 1, 2, . . . . Thus (∗) has solutions Φn (ϕ) = Ψn (ψ) = Pn (ψ) =
Pn (cos ϕ). Now for R, we get

r2 R′′ + 2rR − n(n + 1)R = 0.

This is a Cauchy-Euler equation with the solution

Bn
Rn (r) = An rn + , n = 0, 1, 2, . . . .
rn+1
Thus,
∞ ∞
X
n
X Bn
w(r, ϕ) = An r Pn (cos ϕ) + P (cos ϕ).
n+1 n
n=0 n=0
r
For the interior problem, Bn = 0 and for r = a, we get
∞
X
w(a, ϕ) = f (ϕ) = An an Pn (cos ϕ).
n=0

Putting ϕ = cos−1 ψ, we get

∞
X
−1 −1
w(a, cos ψ) = f (cos ψ) = An an Pn (ψ).
n=0

Let g(ψ) = f (cos−1 ψ) = f (ϕ). Then

2n + 1 1 2n + 1 π
Z Z
An = g(ψ)Pn (ψ)dψ = f (ϕ) sin ϕPn (cos ϕ)dϕ.
2an −1 2an 0

Similarly, for the exterior problem, An = 0 and

(2n + 1)an π
Z
Bn = f (ϕ) sin ϕPn (cos ϕ) dϕ.
2 0

Example: Let a = 1 and f (ϕ) = 1 + cos ϕ. The for interior problem, we have
∞
X
1 + cos ϕ = An Pn (cos ϕ).
n=0

Since, P0 (x) = 1, P1 (x) = x, clearly, A0 = 1 = A1 and An = 0 for n = 2, 3, . . . . For

exterior problem and a = 1 we have Bn are the same as An for the interior problem. Thus,
B0 = B1 = 1 and Bn = 0 for n = 2, 3, . . . .
Example: Let
110, 0 ≤ ϕ < ϕ/2,
f (ϕ) =
0, π/2 < ϕ ≤ π.

8
For the interior problem with a = 1 and the f, given above, we get
Z π/2 Z 1
An = 55(2n + 1) Pn (cos ϕ) sin ϕ dϕ = 55(2n + 1) Pn (x)dx.
0 0

Using
M
X (2n − 2m)!
Pn (x) = (−1)m xn−2m ,
m=0
2n m!(n − m)!(n − 2m)!
we get
M 1
(2n − 2m)!
X Z
m
An = 55(2n + 1) (−1) n xn−2m dx
m=0
2 m!(n − m)!(n − 2m)! 0

M
X (2n − 2m)!
= 55(2n + 1) (−1)m
m=0
2n m!(n − m)!(n − 2m + 1)!
Example: For the interior problem, let a = 1 and f (ϕ) = cos 2ϕ. Since
 
2 4 2 1 4 3 2 1 1
w = cos ϕ =⇒ f (ϕ) = cos 2ϕ = 2 cos ϕ − 1 = w − 1 = P2 (w) − = w − − .
3 3 3 2 2 3
Hence
4 1 4 1 2 1
U (r, ϕ) = r2 P2 (w(ϕ)) − = r2 P2 (cos ϕ) − = r2 (3 cos2 ϕ − 1) − .
3 3 3 3 3 3
Maximum Principle for the Laplace Equation

Let D ⊂ R2 be a bounded domain with it boundary denoted as Γ. Consider a continuous

function u(x, y) on D = D ∪ Γ satisfying the Laplace equation

uxx + uyy = 0, in D.

9
 Let
MD = max u(x, y), MΓ = max u(x, y). the minimum (maximum) of u in D + T only
(x,t)∈D (x,t)∈Γ occur on T unless u is a constant in D+ T
Then
MΓ = MD .
Proof: Clearly, MΓ ≤ MD . Suppose that MΓ < MD . Then there exists (x0 , y0 ) ∈ D such
that u(x0 , y0 ) = MD . Consider the auxiliary function

MD − MΓ
v(x, y) = u(x, y) + [(x − x0 )2 + (y − y0 )2 ],
4R2
where R > 0 is large enough so that the disc (x − x0 )2 + (y − y0 )2 ≤ R2 includes D̄. Then
v(x0 , y0 ) = MD and for any (x, y) ∈ Γ we have v(x, y) < MD . Therefore maximum of v(x, y)
attained in D. But in D we have
MD − MΓ M − MΓ
vxx + vyy = uxx + uyy + 2
= D 2 > 0 (∗),
2R 2R
since uxx + uyy = 0 in D. If there is (x1 , y1 ) ∈ D, such that

v(x1 , y1 ) = max v(x, y),

D

then
vxx (x1 , y1 ) ≤ 0, vyy (x1 , y1 ) ≤ 0.
Hence
vxx (x1 , y1 ) + vyy (x1 , y1 ) ≤ 0,
which is a contradiction to (∗). Hence MD = MΓ .

10