LECTURE 17: SOLVING THE LAPLACE EQUATIONS II MINGFENG ZHAO August 04, 2015 Solving the Laplace equation with Neumann boundary condition in a rectangle Recall that Divergence Theorem: If ∆u = uxx + uyy = 0 in a domain D, then Z Z Z ∂u dS(x, y). 0= ∆u dxdy = div ∇u = ∂n D D ∂D Therefore, in order to have a solution to the problem: ∆u = uxx + uyy = 0, (x, y) ∈ D, ∂u (x, y) = f (x, y), (x, y) ∈ ∂D. ∂n Then we must have an additional condition that: Z f (x, y) dS(x, y) = 0. BC : ∂D Theorem 1. Let D be a “nice” bounded domain in R2 and f (x, y) be a “nice” function on ∂D, then the problem ∆u = uxx + uyy = 0, (x, y) ∈ D, BC : ∂u (x, y) = f (x, y), ∂n (x, y) ∈ ∂D. has a solution if and only if Z f (x, y) dS(x, y) = 0. ∂D Moreover, any two solutions differ by a constant. For the Laplacian BC : (1) AC : equation in a rectangle D = [0, a] × [0, b], since ∂D consists of four line segments, then ∆u = uxx + uyy = 0, uy (x, 0) = f1 (x), (x, y) ∈ D, uy (x, b) = f2 (x), 0 ≤ x ≤ a, ux (0, y) = g1 (y), ux (a, y) = g2 (y), 0 ≤ y ≤ b, Z Z a Z a Z b Z b ∂u dS(x, y) = − f1 (x) dx + f2 (x) dx − g1 (y) dy + g2 (y) dy. 0= 0 0 0 0 ∂D ∂n 1 2 MINGFENG ZHAO To solve (1), we need to solve the following five problems: (2) BC : uy (x, 0) = f1 (x) − α1 =: f˜1 (x), AC : ux (0, y) = 0, ux (a, y) = 0, Z a [f1 (x) − α1 ] dx = 0. ∆u = uxx + uyy = 0, (x, y) ∈ D, uy (x, b) = 0, 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 (3) BC : uy (x, 0) = 0, AC : ux (0, y) = 0, ux (a, y) = 0, Z a [f2 (x) − α2 ] dx = 0. ∆u = uxx + uyy = 0, (x, y) ∈ D, uy (x, b) = f2 (x) − α2 =: f˜2 (x), 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 (4) (5) BC : uy (x, 0) = 0, AC : ux (0, y) = g1 (y) − β1 =: g̃1 (y), Z b [g1 (y) − β1 ] dy = 0. (x, y) ∈ D, uy (x, b) = 0, 0 ≤ x ≤ a, ux (a, y) = 0, 0 ≤ y ≤ b, 0 BC : uy (x, 0) = 0, AC : ux (0, y) = 0, ux (a, y) = g2 (y) − β2 =: g̃2 (y), Z b [g2 (y) − β2 ] dy = 0. BC : AC : (6) ∆u = uxx + uyy = 0, ∆u = uxx + uyy = 0, (x, y) ∈ D, uy (x, b) = 0, 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ∆u = uxx + uyy = 0, (x, y) ∈ D, uy (x, 0) = α1 , uy (x, b) = α2 , 0 ≤ x ≤ a, ux (0, y) = β1 , ux (a, y) = β2 , 0 ≤ y ≤ b, 0 = −aα1 + aα2 − bβ1 + bβ2 . All α1 , α2 , β1 and β2 are constant. It’s easy to see that the solution to (1) is just the sum of solutions to (2), (3), (4), (5) and (6). • For (6), we can solve it directly: To solve (6), we are looking for a solution of the form u(x, y) = F (x) + G(y) to (6), then ux (x, 0) = F 0 (x), uy (x, y) = G0 (y). LECTURE 17: SOLVING THE LAPLACE EQUATIONS II 3 By the boundary condition in (6), we need F 0 (0) = β1 , F 0 (a) = β2 , G0 (0) = α1 , and G0 (b) = α2 . Then let’s try quadratic polynomials for F and G, that is, F (x) = β2 − β1 2 x + β1 x, 2a and G(y) = α2 − α1 2 y + α1 y. 2b In this case, we have ∆u = F 00 (x) + G00 (y) = = = α2 − α1 β2 − β1 + a b aα2 − aα1 + bβ2 − bβ1 ab 0, By the additional condition in (6). Therefore, the solution to (6) is given by: u(x, y) = β2 − β1 2 α2 − α1 2 x + β1 x + y + α1 y . 2a 2b • We are going to solve (2), (3), (4), and (5) by using the method of separation of variables: In the following, we (7) just solve (2). First, let’s look the problem: ∆u = uxx + uyy = 0, (x, y) ∈ D, BC : uy (x, b) = 0, 0 ≤ x ≤ a, ux (0, y) = 0, ux (a, y) = 0, 0 ≤ y ≤ b. Let u(x, y) = X(x)Y (y) be a non-zero separated solution to (7), then X 00 (x)Y (y) + X(x)Y 00 (y) = 0, that is, − X 00 (x) Y 00 (y) = = λ. X(x) Y (y) It’s easy to see that λx = λy = 0, that is, λ is a constant. Since ux (0, y) = ux (a, y) = 0, then X 0 (0) = X 0 (a) = 0. Then X(x) satisfies: (8) X 00 + λX = 0, X 0 (0) = X 0 (a) = 0. For the eigenvalue problem (7), we know that the eigenvalues are λ = nπ S functions are X(x) = C cos x , with n ∈ N {0}. a nπ 2 a , and the corresponding eigen- 4 MINGFENG ZHAO For λ = nπ 2 a , since uy (x, b) = 0, then Y 0 (b) = 0. Then Y satisfies 2 Y 00 − λY = Y 00 − nπ Y = 0, a Y 0 (b) = 0. (9) Then the solution to (9) is given by: h nπ i 2π nπ Y (y) = C e a y + e a b e− a y . Hence, for any n ≥ 0, we can find a non-zero solution to (7): i h nπ 2π nπ un (x, y) = C e a y + e a b e− a y . Let u(x, y) = ∞ X h nπ i nπ 2π nπ an e a y + e a b e− a y cos x a n=0 be the solution to (2). It’s easy to see that uy (x, b) = 0 for 0 ≤ x ≤ a, and ux (0, y) = ux (a, y) = 0 for 0 ≤ y ≤ b. Since uy (x, 0) = f˜1 (x), then f˜1 (x) = ∞ X an · n=0 = ∞ X an · n=1 h nπ a −e 2π a b · nπ nπ i x cos a a i nπ 2nπ nπ h 1 − e a b cos . a a By the additional condition in (2), we know that a Z f˜1 (x) dx = 0. 0 Then for all n ≥ 1, we have an · i 2Z a nπ 2nπ nπ h 1−e a b = f˜1 (x) cos x dx. a a 0 a That is, we have, an = Z 2 h nπ 1 − e 2nπ a b i 0 a nπ f˜1 (x) cos x dx, a for all n ≥ 1. Therefore, the solution to (2) is given by: u(x, y) = ∞ X h nπ i nπ 2π nπ x , an e a y + e a b e− a y cos a n=0 LECTURE 17: SOLVING THE LAPLACE EQUATIONS II 5 where a0 is any constant and an = Z 2 h nπ 1 − e 2nπ a b i a 0 nπ f˜1 (x) cos x dx, a for all n ≥ 1. Solving the Laplace equation with Robin boundary condition in a rectangle Let D = [0, a] × [0, b], consider the following Robin boundary condition for the Laplace equation: ∆u = uxx + uyy = 0, (x, y) ∈ D, (10) BC : u(x, 0) = 0, u(x, b) = f (x), 0 ≤ x ≤ a, ux (0, y) = 0, ux (a, y) = 0, 0 ≤ y ≤ b. First, let’s look the problem: ∆u = uxx + uyy = 0, (x, y) ∈ D, (11) BC : u(x, 0) = 0, 0 ≤ x ≤ a, ux (0, y) = 0, ux (a, y) = 0, 0 ≤ y ≤ b. Let u(x, y) = X(x)Y (y) be a non-zero separated solution to (11), then X 00 (x)Y (y) + X(x)Y 00 (y) = 0, that is, − Y 00 (y) X 00 (x) = = λ. X(x) Y (y) It’s easy to see that λx = λy = 0, that is, λ is a constant. Since ux (0, y) = ux (a, y) = 0, then X 0 (0) = X 0 (a) = 0. Then X(x) satisfies: X 00 + λX = 0, X 0 (0) = X 0 (a) = 0. (12) nπ 2 For the eigenvalue problem (11), we know that the eigenvalues are λ = , and the corresponding eigenfunctions a nπ S are X(x) = C cos x , with n ∈ N {0}. nπ 2 a For λ = , since u(x, 0) = 0, then Y (0) = 0. Then Y satisfies a 2 Y 00 − λY = Y 00 − nπ Y = 0, a (13) Y (0) = 0. Then the solution to (13) is given by: Y (y) = Cy, if n = 0, nπ C e nπ a y − e− a y , if n 6= 1. 6 MINGFENG ZHAO Hence, for any n ≥ 0, we can find a non-zero solution to (11): un (x, y) = Cy, if n = 0, nπ nπ C e nπ a y − e− a y cos x , if n = 6 1. a Let u(x, y) = a0 y + ∞ X nπ nπ nπ x an e a y − e− a y cos a n=1 be the solution to (10). It’s easy to see that u(x, 0) = 0 for 0 ≤ x ≤ a, and ux (0, y) = ux (a, y) = 0 for 0 ≤ y ≤ b. Since u(x, b) = f (x), then f (x) = a0 b + ∞ X nπ nπ nπ an e a b − e− a b cos x . a n=1 Then we have a0 b = 1 2 · 2 a a Z f (x) dx, , 0 nπ nπ 2 and an · e a b − e− a b = a Z a f (x) cos 0 nπ x dx, a for all n ≥ 1. which implies that 1 a0 = ab Z a f (x) dx, 0 2 and an = nπ b nπ a e − e− a b Z a 0 nπ x dx, f˜1 (x) cos a for all n ≥ 1. Therefore, the solution to (10) is given by: u(x, y) = a0 y + ∞ X nπ nπ nπ x an e a y − e− a y cos a n=1 where a0 = 1 ab Z a f (x) dx, and an = 0 e nπ a b 2 nπ − e− a b Z 0 a nπ f˜1 (x) cos x dx, a Solving the Laplace equation in a wedge Example 1. Consider the following problem: (14) BC : 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r u(r, 0) = u(r, α) = 0, 0 < r ≤ a, u(a, θ) = f (θ), 0 < θ < α, u(r, θ) is bounded as r → 0. for all n ≥ 1. LECTURE 17: SOLVING THE LAPLACE EQUATIONS II 7 First, let’s look the following problem: 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r (15) BC : u(r, 0) = u(r, α) = 0, 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 (15), 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, α) = 0, then Θ(0) = Θ(α) = 0. So Θ(θ) satisfies: Θ00 + λΘ = 0, (16) Θ(0) = Θ(α) = 0. nπ 2 For the eigenvalue problem (16), we know that the eigenvalues are λ = , and the corresponding eigenfunctions α nπ are Θ(θ) = C sin θ , with n ∈ N. α nπ nπ nπ 1 For each n ≥ 1, when λ = , since R00 (r) + R0 (r) − λR(r) = 0, then R(r) = C1 r α + C2 r− α . Since u(r, θ) is α r nπ bounded as r → 0, then C2 = 0, that is, R(r) = C1 r α . In summary, for any n ≥ 1, we can find a non-zero solution to (15): un (r, θ) = Cr nπ α sin nπ θ . α Let u(r, θ) = ∞ X an r nπ α sin n=1 nπ θ α be the solution to (14), it’s easy to see that u(r, 0) = u(r, α) = 0. Since u(a, θ) = f (θ), then f (θ) = ∞ X an a nπ α sin n=1 nπ θ . α Then we get an a nπ α 2 = α α Z f (θ) sin nπ 0 α dθ, for all n ≥ 1. dθ, for all n ≥ 1. That is, nπ 2a− α an = α Z α f (θ) sin 0 nπ α 8 MINGFENG ZHAO Therefore, the solution to (14) is: u(r, θ) = ∞ X an r nπ α sin n=1 nπ θ , α where nπ 2a− α an = α Z α f (θ) sin nπ 0 α dθ, for all n ≥ 1. Example 2. Let u0 and u1 be two constants, consider the following problem: BC : (17) Let w(r, θ) = 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r u(r, 0) = u0 , u(r, α) = u1 , 0 < r ≤ a, u(a, θ) = f (θ), 0 < θ < α, u(r, θ) is bounded as r → 0. u1 − u0 θ + u0 , then wr = wrr = wθθ = 0, w(r, 0) = u0 and w(r, α) = u1 . Let v(r, θ) = u(r, θ) − w(r, θ), α then v satisfies 1 1 vrr + vr + 2 vθθ = 0, 0 < r < a, 0 < θ < α, r r u1 − u0 θ − u0 , BC : v(r, 0) = u(r, α) = 0, v(a, θ) = f (θ) − α v(r, θ) is bounded as r → 0. 0 < θ < α, By Example 1, then v(r, θ) = ∞ X n=1 an r nπ α sin nπ θ , α where nπ 2a− α an = α Z α 0 nπ u1 − u0 dθ, f (θ) − θ − u0 sin α α for all n ≥ 1. Therefore, the solution to (17) is given by ∞ nπ X nπ u1 − u0 u(r, θ) = w(r, θ) + v(r, θ) = θ + u0 + an r α sin θ , α α n=1 where nπ 2a− α an = α Z 0 α nπ u1 − u0 f (θ) − θ − u0 sin dθ, α α for all n ≥ 1. LECTURE 17: SOLVING THE LAPLACE EQUATIONS II Example 3. Consider the following BC : (18) 9 problem: 1 1 urr + ur + 2 uθθ = 0, 0 < r < a, 0 < θ < α, r r uθ (r, 0) = uθ (r, α) = 0, 0 < r ≤ a, u(a, θ) = f (θ), 0 < θ < α, 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 (19) BC : u (r, 0) = u (r, α) = 0, θ θ u(r, θ) is bounded as r → 0. Let u(r, θ) = R(r)Θ(θ) be a non-zero separated solution to (19), 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: Θ00 + λΘ = 0, (20) Θ0 (0) = Θ0 (α) = 0. For the eigenvalue problem (20), we know that the eigenvalues are λ = nπ S θ , with n ∈ N {0}. are Θ(θ) = C cos α For each n ≥ 0, when λ = nπ α nπ 2 α , and the corresponding eigenfunctions C1 + C2 ln r, 1 , since R00 (r) + R0 (r) − λR(r) = 0, then R(r) = r Since u(r, θ) is bounded as r → 0, then C2 = 0, that is, R(r) = C1 r nπ α . In summary, for any n ≥ 0, we can find a non-zero solution to (19): un (r, θ) = Cr nπ α cos nπ θ . α Let u(r, θ) = ∞ X n=0 an r nπ α cos nπ θ α C1 r nπ α + C2 r− if n = 0, . nπ α , if n ≥ 1. 10 MINGFENG ZHAO be the solution to (18), it’s easy to see that uθ (r, 0) = uθ (r, α) = 0. Since u(a, θ) = f (θ), then f (θ) = ∞ X an a nπ α cos n=1 nπ θ . α Then we get a0 = 1 2 · 2 α Z Z α α f (θ) dθ, and an a nπ α = 0 2 α Z α f (θ) cos nπ 0 α dθ, for all n ≥ 1. That is, 1 a0 = α nπ f (θ) dθ, 0 2a− α and an = α Z α f (θ) cos nπ α 0 dθ, for all n ≥ 1. dθ, for all n ≥ 1. Therefore, the solution to (18) is: u(r, θ) = ∞ X an r nπ α cos n=0 nπ θ , α where a0 = 1 α Z nπ α f (θ) dθ, 0 and an = 2a− α α Z α f (θ) cos 0 nπ α 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