1. Laplace Operator and the Laplace Equation
Let U be an open subset of Rn. A function f : U → R is said to be twice continuously differentiable if all of the second partial derivatives of f exist and continuous on U. The set of all real valued twice continuously differentiable functions on U is denoted by C2(U ).
Let f ∈ C2(U ). The Laplacian of f is defined to be
∆f =
n
X
i=1
fxixi.
Definition 1.1. A function f ∈ C2(U ) is said to be harmonic if ∆f = 0.
Let g : (0, ∞) → R be a C2-function. We define a function f : R2\ {0} → R by f (x, y) = g(r), r =p
x2+ y2. Let us compute ∆f. By chain rule,
fx= g0(r)x
r, fy = g0(r)y r and
fxx= g00(r)
x r
2
+ g0(r)r2− x2 r3 , fyy= g00(r)
y r
2
+ g0(r)r2− y2 r3 . We obtain that
∆f = fxx+ fyy= g00(r)x2+ y2
r2 + g0(r)2r2− (x2+ y2) r3
= g00(r) +1 rg0(r).
Hence ∆f = 0 if and only if
g00(r) + 1
rg0(r) = 0, r > 0.
Let h(r) = g0(r) for r > 0 We have rh0(r) + h(r) = 0, for r > 0 i.e. (rh(t))0 = 0 for r > 0.
By mean value theorem, rh0(r) = c1 for r > 0. Therefore g0(r) = h(r) = c1
r for r > 0.
By integrating g0(r), we have
g(r) = Z r
r0
c1
tdt = c1ln r + c2
where c2 = −c1ln r0 and r0 > 0. This shows that f (x, y) = c1
2 log(x2+ y2) + c2. Let us try to solve the Laplace equation
∆f = 0 on B(0, 1).
1
2
Proposition 1.1. Let f ∈ C2(B(0, 1)) ∩ C(D(0, 1)). We set
u(r, θ) = f (r cos θ, r sin θ), 0 < r < 1, θ ∈ R.
Then u is a periodic function in θ with period 2π.
∆f = urr+1
rur+ 1 r2uθθ.
Assume that u(r, θ) = R(r)Θ(θ) is a nonzero solution to ∆f = 0. Then R00(r)Θ(θ) +R0(r)
r Θ(θ) +R(r)
r2 Θ00(θ) = 0.
Dividing the above equation by R(r)Θ(θ) and multiplying the above by r2, we obtain r2R00(r) + rR0(r)
R(r) = −Θ00(θ) Θ(θ) .
The left hand side of the equation is a function of r which is independent of θ while the right hand side of the equation a function of θ which is independent of r. Thus both of the right hand side and the left hand side of the equation must be a constant. Set the constant to be λ We obtain two differential equation
r2R00(r) + rR0(r) −λR(r) = 0,
Θ00(θ) + λΘ(θ) = 0.
When λ = 0, Θ(θ) = aθ + b for some a, b. Since Θ is a period function, a = 0. Therefore r2R00(r) + rR0(r) = 0. We have seen that R(r) = c1log r + c2. Since we assume that f is continuously differentiable at 0, it is continuous at 0. This implies that c1 = 0.
When λ 6= 0, for each n ≥ 1, the function cn(θ) = cos nθ and sn(θ) = sin nθ are both periodic of period 2π and satisfy Θ00+ n2Θ = 0. In this case, we obtain that for each n ≥ 1,
r2R00(r) + rR0(r) − n2R(r) = 0.
To solve this equation, we define a new function φ(t) = R(et), i.e. we make a change of variable r = et. Then we obtain that
φ00(t) − n2φ(t) = 0.
By theory of O.D.E., we find that φ(t) is of the form φ(t) = C1ent+ C2e−nt. This implies that
R(r) = C1rn+ C2r−n.
Again, by the continuity of f, we find C2 = 0. Set un(r, θ) = rncos nθ and vn(r, θ) = rnsin nθ. Then {1, un(r, θ), vn(r, θ)} are all solutions to the Laplace equation. Fourier’s idea: combine all these solutions into one. More precisely, we set
(1.1) u(r, θ) = a0
2 +
∞
X
n=1
rn(ancos nθ + bnsin nθ).
How do we determine a0, an, bn? Observe that (1.2) u(1, θ) = f (cos θ, sin θ) = a0
2 +
∞
X
n=1
(ancos nθ + bnsin nθ).
3
Let us define g(θ) = f (cos θ, sin θ) for θ ∈ R. It follows from the definition that g is a periodic function of period 2π, in fact,
g(θ + 2π) = f (cos(θ + 2π), sin(θ + 2π)) = f (cos θ, sin θ) = g(θ)
for any θ ∈ R. We can also use the notation f |S1 to denote g. We call g the restriction of f to S1 = ∂B(0, 1). Hence if we know g, the restriction of f to the boundary of S1, and express g as an infinite series of the form (1.2), then we can find (solve for) f. This leads to the following Dirichlet problem
(1.3) ∆f = 0 on B(0, 1)
f |S1 = g Hence if g has the following series representation
(1.4) g(θ) = a0
2 +
∞
X
n=1
ancos nθ + bnsin nθ, then f is given by (1.1).
Example 1.1. Solve (1.3) with the boundary condition g(θ) = 1 − cos θ + sin θ + 3 sin 3θ.
The harmonic function is given by
f (r cos θ, sin θ) = 1 − r cos θ + r sin θ + 3r3sin 3θ.
We can also express f in terms of x, y as follows. By De Moivre’s law, we know (x + iy)n= rn(cos nθ + i sin nθ).
We set Cn(x, y) = Re(x + iy)n and Sn(x, y) = Im(x + iy)n. The above function can be rewritten as
f (x, y) = 1 − C1(x, y) + S1(x, y) + 3S3(x, y).
Since C1(x, y) = x and S1(x, y) = y and S3(x, y) = 3xy2− y3, f (x, y) = 1 − x + y + 9x2y − 3y3.
In general, if the boundary condition is given by (1.4), then the solution to (1.3) is given by
(1.5) f (x, y) = a0
2 +
∞
X
n=1
(anCn(x, y) + bnSn(x, y)).
Later, we will show that any continuous periodic function of period 2π always possess such a series representation as (1.4) and discuss the convergence of (1.5). We need the notion of uniform convergence.
Example 1.2. Solve (1.3) with the boundary condition
g(θ) = 1 + 2 cos θ − 3 sin θ + 4 sin2θ + cos2θ + 5 sin θ cos θ + cos3θ + 6 sin3θ.