LECTURE 18: SOLVING THE LAPLACE EQUATIONS III AND POISSON EQUATIONS MINGFENG ZHAO August 05, 2015 Solving the Laplace equation in a wedge with cut-outs Example 1. Consider the following problem: 1 1 urr + ur + 2 uθθ = 0, a < r < b, 0 < θ < α, r r (1) BC : u (r, 0) = u (r, α), a ≤ r ≤ b, θ θ u(b, θ) = 0, u(a, θ) = f (θ), 0 < θ < α. First, let’s look the following problem: 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r (2) BC : uθ (r, 0) = uθ (r, α), a ≤ r ≤ b, u(b, θ) = 0, 0 < θ < α. Let u(r, θ) = R(r)Θ(θ) be a non-zero separated solution to (2), then 1 1 R00 (r)Θ(θ) + R0 (r)Θ(θ) + 2 R(r)Θ00 (θ) = 0. r r That is, we have R00 (r) + 1r R0 (r) Θ00 (θ) =− = λ. R(r) Θ(θ) It’s easy to see that λr = λθ = 0, that is, λ is a constant. Since uθ (r, 0) = uθ (r, α) = 0, then Θ0 (0) = Θ0 (α) = 0. So Θ(θ) satisfies: (3) Θ00 + λΘ = 0, Θ0 (0) = Θ0 (α) = 0. For the eigenvalue problem (3), we know that the eigenvalues are λ = nπ S are Θ(θ) = C cos θ , with n ∈ N {0}. α 1 nπ 2 α , and the corresponding eigenfunctions 2 MINGFENG ZHAO nπ For each n ≥ 0, when λ = α C1 + C2 ln r, 1 , since R00 (r) + R0 (r) − λR(r) = 0, then R(r) = r C1 + C2 ln b = 0, C1 r nπ α if n = 0, . + C2 r− nπ α , if n ≥ 1. if n = 0, C[− ln b + ln r], Since u(b, θ) = 0, then , that is, R(r) = i h 2nπ nπ C b nπ C r nπ − nπ α − e α b r− α , if n ≥ 1. 1 α + C2 b α = 0, if n ≥ 1. In summary, for any n ≥ 0, we can find a non-zero solution to (2): − ln b + ln r, if n = 0, h nπ i nπ un (r, θ) = 2nπ r α − b α cos θ , if n ≥ 1. α if n = 0, Let u(r, θ) = a0 [− ln b + ln r] + ∞ X i nπ h nπ 2nπ nπ θ an r α − b α r− α cos α n=0 be the solution to (1), it’s easy to see that uθ (r, 0) = uθ (r, α) = u(b, θ) = 0. Since u(a, θ) = f (θ), then ∞ X h nπ i nπ 2nπ nπ an a α − b α a− α cos θ . α n=0 f (θ) = a0 [− ln b + ln a] + Then we get Z 1 2 a0 [− ln b + ln a] = · 2 α α f (θ) dθ, h and an a nπ α −b 2nπ α − nπ α a i 0 Z 2 = α α f (θ) cos nπ α 0 dθ, for all n ≥ 1. That is, a0 = 1 α[− ln b + ln a] Z α f (θ) dθ, and an = α a 0 Z 2 h nπ α −b 2nπ α a− i nπ α α f (θ) cos nπ 0 Solving the Laplace equation with Dirichlet problem in a disc Consider the following problem: BC : (4) 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < 2π, r r u(r, 0) = u(r, 2π), uθ (r, 0) = uθ (r, 2π), 0 < r ≤ a, u(a, θ) = f (θ), 0 ≤ θ < 2π, u(r, θ) is bounded as r → 0. First, let’s look the following problem: 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r (5) BC : u(r, 0) = u(r, 2π), uθ (r, 0) = uθ (r, 2π), 0 < r ≤ a, u(r, θ) is bounded as r → 0. α dθ, for all n ≥ 1. LECTURE 18: SOLVING THE LAPLACE EQUATIONS III AND POISSON EQUATIONS 3 Let u(r, θ) = R(r)Θ(θ) be a non-zero separated solution to (5), then 1 1 R00 (r)Θ(θ) + R0 (r)Θ(θ) + 2 R(r)Θ00 (θ) = 0. r r That is, we have R00 (r) + 1r R0 (r) Θ00 (θ) =− = λ. R(r) Θ(θ) It’s easy to see that λr = λθ = 0, that is, λ is a constant. Since u(r, 0) = u(r, 2π) and uθ (r, 0) = uθ (r, 2π), then Θ(0) = Θ(2π) and Θ0 (0) = Θ0 (2π). So Θ(θ) satisfies: Θ00 + λΘ = 0, (6) Θ(0) = Θ(2π), Θ0 (0) = Θ0 (2π). For the eigenvalue problem (6), we know that the eigenvalues are λ = n2 , and the corresponding eigenfunctions are S Θ(θ) = C1 cos(nx) + C2 sin(nx), with n ∈ N {0}. C1 + C2 ln r, if n = 0, 1 . For each n ≥ 0, when λ = n2 , since R00 (r) + R0 (r) − λR(r) = 0, then R(r) = r nπ C r nπ α + C r− α , if n ≥ 1. 1 2 C, if n = 0, Since u(r, θ) is bounded as r → 0, then C2 = 0, that is, R(r) = Crn , if n ≥ 1. In summary, for any n ≥ 0, we can find a non-zero solution to (5): 1, if n = 0, un (r, θ) = rn [C cos(nθ) + C sin(nθ)], if n ≥ 1. 1 2 Let u(r, θ) = a0 + ∞ X [an rn cos(nθ) + bn rn sin(nθ)] n=1 be the solution to (4), it’s easy to see that u(r, 0) = u(r, 2π) and uθ (r, 0) = uθ (r, 2π). Since u(a, θ) = f (θ), then ∞ a0 X f (θ) = + [an an cos(nθ) + bn an sin(nθ)]. 2 n=1 Then we get Z 1 2π f (θ) dθ, a0 = π 0 1 an a = π Z a−n an = π Z n 2π f (θ) cos(nθ) dθ, 0 2π 1 and bn a = π Z a−n and bn = π Z n f (θ) sin(nθ) dθ, for all n ≥ 1. f (θ) sin(nθ) dθ, for all n ≥ 1. 0 That is, 1 a0 = π Z 2π f (θ) dθ, 0 2π f (θ) cos(nθ) dθ, 0 0 2π 4 MINGFENG ZHAO The Poisson’s integral formula Recall that the solution to the following problem: 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < 2π, r r BC : u(r, 0) = u(r, 2π), u (r, 0) = u (r, 2π), 0 < r ≤ a, θ θ (7) u(a, θ) = f (θ), 0 ≤ θ < 2π, u(r, θ) is bounded as r → 0. is given by: ∞ a0 X + [an rn cos(nθ) + bn rn sin(nθ)], 2 n=1 u(r, θ) = where 1 a0 = π Z 2π f (θ) dθ, 0 a−n an = π 2π Z f (θ) cos(nθ) dθ, 0 a−n and bn = π Z 2π f (θ) sin(nθ) dθ, for all n ≥ 1. 0 Then we have u(r, θ) = = = = Let z = 1 2π Z 1 2π Z 2π f (τ ) dτ + 0 ∞ −n Z X a n=1 2π f (τ ) dτ + 0 π ∞ X a−n rn π n=1 2π f (τ ) cos(nτ ) dτ rn cos(nθ) + 0 a−n π 2π Z f (τ ) sin(nτ ) dτ 0 2π Z f (τ )[cos(nτ ) cos(nθ) + sin(nτ ) sin(nθ)] dτ 0 Z ∞ X a−n rn 2π f (τ ) dτ + f (τ ) cos[n(τ − θ)] dτ π 0 0 n=1 # " Z ∞ 1 2π 1 X r n cos[n(θ − τ )] dτ. + f (τ ) π 0 2 n=1 a 1 2π Z 2π r n r i(θ−τ ) e , then cos[n(θ − τ )] = Re z n for all n 6= 1, which implies that a a ∞ n X r n=1 a cos[n(θ − τ )] ∞ X = n=1 Re z n = Re ∞ X zn n=1 1 If |z| < 1, that is, r < a 1−z ar cos(θ − τ ) . a2 − 2ar cos(θ − τ ) + r2 = Re = Therefore, we get u(r, θ) = 1 π Z 0 2π 1 ar cos(θ − τ ) f (τ ) + 2 a2 − 2ar cos(θ − τ ) + r2 dτ rn sin(nθ) LECTURE 18: SOLVING THE LAPLACE EQUATIONS III AND POISSON EQUATIONS = a2 − r2 2π Z 0 2π 5 f (τ ) dτ, a2 − 2ar cos(θ − τ ) + r2 which is called the Poisson integral formula. Solving the Laplace equation with Neumann problem in a disc Consider the following problem: (8) BC : 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < 2π, r r u(r, 0) = u(r, 2π), uθ (r, 0) = uθ (r, 2π), 0 < r ≤ a, AC : ur (a, θ) = f (θ), 0 ≤ θ < 2π, u(r, θ) is bounded as r → 0 Z 2π f (θ) dθ = 0. 0 First, let’s look the following problem: (9) 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r BC : u(r, 0) = u(r, 2π), uθ (r, 0) = uθ (r, 2π), 0 < r ≤ a, u(r, θ) is bounded as r → 0. Let u(r, θ) = R(r)Θ(θ) be a non-zero separated solution to (9), then 1 1 R00 (r)Θ(θ) + R0 (r)Θ(θ) + 2 R(r)Θ00 (θ) = 0. r r That is, we have R00 (r) + 1r R0 (r) Θ00 (θ) =− = λ. R(r) Θ(θ) It’s easy to see that λr = λθ = 0, that is, λ is a constant. Since u(r, 0) = u(r, 2π) and uθ (r, 0) = uθ (r, 2π), then Θ(0) = Θ(2π) and Θ0 (0) = Θ0 (2π). So Θ(θ) satisfies: (10) Θ00 + λΘ = 0, Θ(0) = Θ(2π), Θ0 (0) = Θ0 (2π). For the eigenvalue problem (10), we know that the eigenvalues are λ = n2 , and the corresponding eigenfunctions are S Θ(θ) = C1 cos(nx) + C2 sin(nx), with n ∈ N {0}. 6 MINGFENG ZHAO C1 + C2 ln r, if n = 0, 1 0 . R (r) − λR(r) = 0, then R(r) = r C r nπ − nπ 1 α + C2 r α , if n ≥ 1. C, if n = 0, Since u(r, θ) is bounded as r → 0, then C2 = 0, that is, R(r) = Crn , if n ≥ 1. In summary, for any n ≥ 0, we can find a non-zero solution to (9): 1, if n = 0, un (r, θ) = rn [C cos(nθ) + C sin(nθ)], if n ≥ 1. For each n ≥ 0, when λ = n2 , since R00 (r) + 1 2 Let u(r, θ) = a0 + ∞ X [an rn cos(nθ) + bn rn sin(nθ)] n=1 be the solution to (8), it’s easy to see that u(r, 0) = u(r, 2π) and uθ (r, 0) = uθ (r, 2π). Since ur (a, θ) = f (θ), then f (θ) = ∞ X [nan an−1 cos(nθ) + nbn an−1 sin(nθ)]. n=1 Then we have Z 2π f (θ) dθ = 0, nan a n−1 0 1 = π Z 2π f (θ) cos(nθ) dθ, n−1 and nbn a 0 1 = π Z 2π f (θ) sin(nθ) dθ, for all n ≥ 1. 0 That is, a1−n an = nπ Z 2π f (θ) sin(nθ) dθ, 0 a1−n and bn = nπ Z 2π f (θ) sin(nθ) dθ, for all n ≥ 1. 0 Solving the Poisson equation with Dirichlet boundary condition in a rectangle by using eigenfunction expansions For the Poisson equation in a rectangle D = [0, a] × [0, b], since ∂D consists of four line segments, then uxx + uyy = f (x, y), (x, y) ∈ D, (11) BC : u(x, 0) = f1 (x), u(x, b) = f2 (x), 0 ≤ x ≤ a, u(0, y) = g1 (y), u(a, y) = g2 (y), 0 ≤ y ≤ b. To solve the problem (11): Step I: Choose any “nice” function w(x, y) such that w(0, y) = g1 (y) and w(a, y) = g2 (y) for 0 ≤ y ≤ b. For example, we can just choose w(x, y) = g2 (y) − g1 (y) x + g1 (y). a LECTURE 18: SOLVING THE LAPLACE EQUATIONS III AND POISSON EQUATIONS 7 Step II: Let v(x, y) = u(x, y) − w(x, y), then vxx + vyy = uxx + uyy − wxx − wyy = f (x, y) − wxx − wyy =: f˜(x, y) v(x, 0) v(x, b) v(0, y) v(a, y) (12) = u(x, 0) − w(x, 0) = f1 (x) − w(x, 0) = f˜1 (x) = v(x, b) − w(x, b) = f2 (x) − w(x, b) = f˜2 (x) = u(0, y) − w(0, y) = g1 (y) − g1 (y) = 0 = u(a, y) − w(a, y) = g2 (y) − g2 (y) = 0. Then v satisfies the following problem: vxx + vyy = f˜(x, y), (x, y) ∈ D, BC : v(x, 0) = f˜1 (x), v(x, b) = f˜2 (x), v(0, y) = v(a, y) = 0, 0 ≤ y ≤ b. 0 ≤ x ≤ a, To solve (12), we are going to use the method of eigenfunction expansions, which is a “generalization” of “variation of parameters”: Step II-1: First let’s look at the problem: vxx + vyy = 0, (x, y) ∈ D, (13) BC: v(0, y) = v(a, y) = 0, 0 ≤ y ≤ b. 8 MINGFENG ZHAO By using the separation of variables, we know that the “ general ” solution to (13) is: v(x, y) = ∞ nπ X nπ nπ an e a y + bn e− a y sin x . a n=1 Step II-2: For each y > 0, we can think f˜(·, y) is a function of x, then we can write it as an “infinite” linear combination n nπ o∞ x of sin , that is, a n=1 ∞ nπ X f˜(x, y) = x , f˜n (t) sin a n=1 where Z a nπ f˜(x, t) sin x dx, ∀n ≥ 1. a 0 n nπ o∞ Step II-3: Write f˜1 (x) and f˜2 (x) as “infinite” linear combinations of sin x , that is, a n=1 ∞ ∞ nπ nπ X X f˜1 (x) = x , and f˜2 (x) = x , f˜1,n sin f˜2,n sin a a n=1 n=1 2 f˜n (t) = a where 2 f˜1,n = a Z 0 a nπ x dx, f˜1 (x) sin a 2 and f˜2,n = a Z 0 a nπ x dx, f˜2 (x) sin a ∀n ≥ 1. Step II-4: Let v(x, y) = ∞ X vn (y) sin n=1 nπ x , a be the solution to (12). It’s easy to see that v(0, y) = v(a, y) = 0 for all 0 ≤ y ≤ b. Moreover, “informally” we have ∞ X n=1 vn00 (y) sin ∞ nπ X nπ nπ 2 x − x vn (y) sin a a a n=1 = vxx + vyy = f˜(x, y) = ∞ X vn (0) sin n=1 nπ x a n=1 vn (b) sin nπ x a nπ x f˜n (y) sin a n=1 = v(x, 0) = f˜1 (x) = ∞ X ∞ X ∞ X nπ f˜1,n sin x a n=1 = v(x, b) = f˜2 (x) = ∞ X nπ f˜2,n sin x . a n=1 LECTURE 18: SOLVING THE LAPLACE EQUATIONS III AND POISSON EQUATIONS (14) 9 Then for any n ≥ 1, vn (y) satisfies the following ODE with the initial value problem: n2 π 2 vn00 (y) − 2 vn (y) − f˜n (y) = 0, a vn (0) = f˜1,n , vn (b) = f˜2,n . So if for any n ≥ 1, vn (y) is the solution to (14), then v(x, y) = ∞ X vn (y) sin n=1 nπ x is the solution to (12). a Step III: The solution to (11) is given by: u(x, y) = = w(x, y) + v(x, y) w(x, y) + ∞ X n=1 vn (y) sin nπ x . a Department of Mathematics, The University of British Columbia, Room 121, 1984 Mathematics Road, Vancouver, B.C. Canada V6T 1Z2 E-mail address: mingfeng@math.ubc.ca