LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES MINGFENG ZHAO July 21, 2015 Two-point boundary value problems A two-point boundary value problem for the second order linear differential equation has the form: P (x)y 00 + Q(x)y 0 + R(x)y = f (x), ∀a < x < b, α1 y(a) + β1 y 0 (a) = γ1 α2 y(b) + β2 y 0 (b) = γ2 , where P (x), Q(x) and R(x) are functions, αi , βi and γi are constants for i = 1, 2. Remark 1. For a two-point boundary value problem, there might be a solution, or no solution or many solutions. Example 1. Let a be a constant, solve the boundary value problem y 00 + y = 0, y(0) = 1 and y(π) = a. Try y = erx to be a solution to y 00 + y = 0, then y 0 = rerx , y 00 = r2 erx and 0 = y 00 + y = r2 erx + erx = erx (r2 + 1). Then r2 + 1 = 0, that is, r = ±i. So the general solution to y 00 + y = 0 is y(x) = C1 cos(x) + C2 sin(x). Since y(0) = 1 and y(π) = a, then C1 = 1, − C1 = a. and So if a 6= −1, then there is no solution to y 00 + y = 0, y(0) = 1 and y(π) = a; if a = −1, then for any constant C2 , y(x) = cos(x) + C2 sin(x) is a solution to y 00 + y = 0, y(0) = 1 and y(π) = a = −1. 1 2 MINGFENG ZHAO Example 2. Solve the boundary value problem y 00 + y = 0, y(0) = y(π) = 0. By the computations in Example 1, we know that the general solution to y 00 + y = 0 is: y(x) = C1 cos(x) + C2 sin(x). Since y(0) = y(π) = 0, then C1 = 0, and − C1 = 0. Therefore, the solution to y 00 + y = 0, y(0) = y(π) = 0 is: y(x) = C2 sin(x). Eigenvalue problems Definition 1. Consider the two-point boundary value problem: P (x)y 00 + Q(x)y 0 + R(x)y = λr(x)y, (1) α1 y(a) + β1 y 0 (a) = 0 α2 y(b) + β2 y 0 (b) = 0, ∀a < x < b, where P (x), Q(x), R(x) and r(x) are functions, αi and βi are constants for i = 1, 2, and λ is a constant. We say that λ is an eigenvalue of (1) if there is a non-zero solution φ(x) to (1). In this case, a non-zero solution φ(x) to (1) is called an eigenfunction corresponding to the eigenvalue λ. Example 3. For the following eigenvalue problem: X 00 (x) + λX(x) = 0, 0 < x < L, (2) X(0) = 0, X(L) = 0. By the computations in the part of Solving the heat equation without source with homogeneous Dirichlet nπ 2 boundary value in Lecture 8, the eigenvalues for (2) are λn = for n ∈ N, and the corresponding eigenfunctions L nπ are Xn = sin x . L Example 4. For the following eigenvalue problem: X 00 (x) + λX(x) = 0, 0 < x < L, (3) X 0 (0) = 0, X 0 (L) = 0. LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES 3 By the computations in the part of Solving the heat equation without source with homogeneous Neumann nπ 2 S boundary condition in Lecture 8, the eigenvalues for (3) are λn = for n ∈ N {0}, and the corresponding L nπ eigenfunctions are Xn = cos x . L Example 5. For the following eigenvalue problem: X 00 (x) + λX(x) = 0, −L < x < L, (4) X(−L) = X(L), X 0 (−L) = X 0 (L). By the computations in the part of Solving the heat equation on a circular ring (with a periodic boundary nπ 2 S condition) in Lecture 8, the eigenvalues for (4) are λn = for n ∈ N {0}, and the corresponding eigenfunctions L nπ nπ x and X̃n = sin x . are Xn = cos L L Fourier sine series Question 1. Assume that f (x) = ∞ X bn sin n=1 nπ x , L x ∈ [0, L], how can we compute bn by using f (x)? Lemma 1. Let L > 0 and m ∈ Z, then L Z mπ x dx cos L 0 L, if m = 0, . 0, if m 6= 0. = Proof. If m = 0, then Z L cos 0 mπ x dx L Z L = 1 dx = L. 0 If m 6= 0, then Z L cos 0 mπ x dx L = 1 mπ L · sin mπ x L L 0 = L · [sin(mπ) − sin(0)] mπ = 0. 4 MINGFENG ZHAO Lemma 2. Let L > 0 and m, n ∈ N, then L Z 0 0, if m 6= n = . L , if m = n. 2 mπ nπ sin x sin x dx L L Proof. Recall the identity: sin(α) sin(β) = cos(α − β) − cos(α + β) . 2 Then we have Z L sin 0 mπ nπ x sin x dx L L L (m − n)π (m + n)π cos x − cos x dx L L 0 Z Z (m − n)π (m + n)π 1 L 1 L x dx − x dx = cos cos 2 0 L 2 0 L Z 1 L (m − n)π = x dx Since m, n ∈ N and by Lemma 1 cos 2 0 L 0, if m 6= n = , By Lemma 1. L , if m = n. 2 1 2 = Z ∞ X nπ mπ x for all x ∈ L, for any m ∈ N, both sides multiply sin and then L L n=1 integrate over [0, L], “informally” we have Now assume that f (x) = Z 0 L bn sin mπ x dx f (x) sin L Z L = 0 Z = = ∞ X # nπ mπ bn sin x sin x dx L L n=1 ∞ LX 0 = " ∞ X bn sin n=1 L Z bn sin mπ nπ x sin x dx L L mπ nπ x sin x dx L L n=1 0 L bm , 2 By Lemma 2. Then we have bm 2 = L Z L f (x) sin 0 mπ x dx, L for all m ≥ 1 . LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES 5 Example 6. Solve the problem: ut = uxx , 0 < x < 1, BC: u(0, t) = 0, u(1, t) = 0, IC: u(x, 0) = x, 0 ≤ x ≤ 1. t > 0, t > 0, By the result in the part of Solving the heat equation without source with homogeneous Dirichlet boundary condition in Lecture 8, we know that for any n ∈ N, we can find a non-zero solution to ut = uxx , 0 < x < 1, t > 0, : BC: u(0, t) = 0, u(1, t) = 0, t > 0. un (x, t) = e−n 2 π2 t sin(nπx). Now let’s write x as an “infinitely” linear combination of {sin(nπx)}∞ n=1 : x= ∞ X bn sin(nπx), ∀x ∈ [0, 1]. n=1 Then for all n ≥ 1, we have bn = 2 1 Z x sin(nπx) dx Z = 1 0 1 2 x sin(nπx) dx. 0 Z For x sin(nπx) dx, use the integration by parts, we have Z 1 x d cos(nπx) · − nπ Z 1 − x d cos(nπx) nπ Z 1 − x cos(nπx) − cos(nπx) dx nπ 1 1 − x cos(nπx) − sin(nπx) + C nπ nπ Z x sin(nπx) dx = = = = = − x cos(nπx) sin(nπx) + . nπ n2 π 2 So we get bn = x cos(nπx) sin(nπx) 2 − + nπ n2 π 2 1 0 6 MINGFENG ZHAO cos(nπ) sin(nπ) 2 − + 2 2 −2·0 nπ n π = 2 cos(nπ) nπ 2(−1)n+1 . nπ = − = Hence “informally” we get x= ∞ X 2(−1)n+1 · sin(nπx), nπ n=1 ∀x ∈ [0, 1]. So the solution is: u(x, t) = ∞ X 2(−1)n+1 −n2 π2 t ·e sin(nπx). nπ n=1 ∞ X 2(−1)n+1 · sin(nπx) in [0, 1], then Remark 2. Since we know that x = nπ n=1 ∞ nπ X 2(−1)n+1 · sin nπ 2 n=1 ∞ X 2(−1)2k+1+1 (2k + 1)π = · sin (2k + 1)π 2 1 2 = k=0 = ∞ X k=0 = 2 π · sin kπ + (2k + 1)π 2 ∞ 2 X (−1)k . π 2k + 1 k=0 Hence we have ∞ π X (−1)k 1 1 1 = = 1 − + − + ··· . 4 2k + 1 3 5 7 k=0 Example 7. Solve the problem: ut = uxx , 0 < x < 1, BC: u(0, t) = 0, IC: u(x, 0) = f (x), t > 0, u(1, t) = 0, t > 0, 0 ≤ x ≤ 1, where f (x) = 2x, 2 − 2x, 1 if 0 < x ≤ , 2 1 if < x < 1, 2 LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES 7 By the computations in the part of Solving the heat equation and separation of variables in Lecture 8, we ut = c2 uxx , 0 < x < 1, t > 0, know that for any n ∈ N, we can find a non-zero solution to : BC: u(0, t) = 0, u(L, t) = 0, t > 0. un (x, t) = e−n 2 π2 t sin(nπx). Now let’s write f (x) as an “infinitely” linear combination of {sin(nπx)}∞ n=1 : f (x) = ∞ X bn sin(nπx), ∀x ∈ [0, 1]. n=1 Then for all n ≥ 1, we have bn = 2 1 f (x) sin(nπx) dx 0 Z = 1 Z 1 2 2 (2 − 2x) sin(nπx) dx 2x sin(nπx) dx + 2 1 2 0 Z = 1 Z 1 2 4 1 Z Z 1 2 0 1 sin(nπx) dx − 4 x sin(nπx) dx + 4 x sin(nπx) dx. 1 2 By the computation in Example 6, we have Z x cos(nπx) sin(nπx) x sin(nπx) dx = − + . nπ n2 π 2 Then we have " 1 2 Z 4 x sin(nπx) dx 0 Z # cos nπ sin nπ 2 2 = 4 − + −4·0 nπ n2 π 2 2 cos nπ 4 sin nπ 2 2 = − + nπ n2 π 2 1 4 sin(nπx) dx = 1 2 = Z 1 −4 x sin(nπx) dx = 1 2 = − 1 2 4 cos(nπx) nπ 1 1 2 4 cos(nπ) 4 cos nπ 2 − + nπ nπ " # 1 nπ sin nπ cos(nπ) sin(nπ) 2 cos 2 2 −4 − + 2 2 +4 − + nπ n π nπ n2 π 2 4 sin nπ 4 cos(nπ) 2 cos nπ 2 2 − + . nπ nπ n2 π 2 So we have bn 2 cos nπ 4 sin nπ 4 cos(nπ) 4 cos nπ 2 2 2 = − + − + nπ n2 π 2 nπ nπ 8 MINGFENG ZHAO 4 sin nπ 4 cos(nπ) 2 cos nπ 2 2 + − + nπ nπ n2 π 2 8 sin nπ 2 = . n2 π 2 Then b2k = 8 sin(kπ) (2k)2 π 2 = 0 8 sin b2k+1 = = = (2k+1)π 2 (2k + 1)2 π 2 8 sin kπ + π2 (2k + 1)2 π 2 8(−1)k , (2k + 1)2 π 2 ∀k ≥ 0. So “informally” we get f (x) = ∞ X k=0 8(−1)k sin[(2k + 1)πx], (2k + 1)2 π 2 ∀x ∈ [0, 1]. So the solution is: u(x, t) = ∞ X k=0 2 2 8(−1)k e−(2k+1) π t sin[(2k + 1)πx]. 2 2 (2k + 1) π Remark 3. Notice that 1 = f 2 1 = ∞ X k=0 = ∞ X k=0 = ∞ X k=0 = ∞ X k=0 8(−1)k sin (2k + 1)2 π 2 (2k + 1)π 2 π 8(−1)k sin kπ + 2 2 (2k + 1) π 2 8(−1)k · (−1)k (2k + 1)2 π 2 8 . (2k + 1)2 π 2 Hence we get ∞ (5) X π2 1 1 1 1 = = 1 + 2 + 2 + 2 + ··· . 8 (2k + 1)2 3 5 7 k=0 LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES Remark 4. One can use (5) to prove the following identity: 1 1 1 π2 1 = 1 + 2 + 2 + 2 + 2 + ··· . 6 2 3 4 5 In fact, for any n ≥ 1, there exist unique k ≥ 0 and unique odd integer m such that n = 2k m, then 1+ 1 1 1 1 + 2 + 2 + 2 + ··· 22 3 4 5 = = ∞ X 1 2 n n=1 ∞ X ∞ X k=0 i=0 = 1 [2k (2i + 1)]2 ∞ ∞ X 1 1 X k 4 i=0 (2i + 1)2 k=0 = ∞ X 1 π2 · 4k 8 By (5) k=0 = ∞ π2 X 1 8 4k = 1 π2 · 8 1− k=0 = = 1 4 π2 4 · 8 3 π2 . 6 Fourier cosine series Question 2. Assume that ∞ f (x) = nπ a0 X an cos + x , 2 L n=1 x ∈ [0, L], how can we compute an by using f (x)? S Lemma 3. Let L > 0 and m, n ∈ N {0}, then Z L cos 0 mπ nπ x cos x dx L L 0, if m 6= n = L, if m = n = 0 . L , if m = n 6= 0. 2 9 10 MINGFENG ZHAO Proof. Recall the identity: cos(α) cos(β) = cos(α + β) + cos(α − β) . 2 Then we have Z L cos 0 nπ mπ x cos x dx L L L (m + n)π (m − n)π cos x + cos x dx L L 0 Z Z (m + n)π (m − n)π 1 L 1 L cos cos x dx + x dx. 2 0 L 2 0 L 1 2 = = Z Then • If m 6= n, since m, n ∈ N S {0}, by Lemma 1, then L Z cos 0 nπ mπ x cos x dx = 0. L L • If m = n = 0, then L Z cos 0 Z L mπ nπ x cos x dx = 1 dx = L. L L 0 • If m = n 6= 0, then m = n ∈ N, by Lemma 1, then Z L cos 0 mπ nπ L x cos x dx = . L L 2 ∞ Now assume f (x) = nπ a0 X an cos + x in [0, L], then 2 L n=1 • Both sides integrate over [0, L], then “informally” we have Z L # ∞ nπ a0 X + an cos x dx 2 L 0 n=1 Z L ∞ nπ a0 L X + an cos x dx 2 L 0 n=1 Z f (x) dx = 0 = = L " a0 L , 2 By Lemma 1. Then we get 2 a0 = L Z L f (x) dx. 0 LECTURE 9: TWO-POINT BOUNDARY VALUE PROBLEMS AND FOURIER SERIES • For any m ∈ N, both sides multiply cos 11 mπ and then integrate over [0, L], “informally” we have L # Z L Z L" ∞ mπ mπ nπ a0 X f (x) cos x dx = + x cos x dx an cos L 2 L L 0 0 n=1 # Z L" ∞ mπ mπ X nπ a0 cos cos x + dx = an cos 2 L L 2 0 n=1 Z Z L ∞ mπ mπ nπ X a0 L = cos cos dx x dx + an cos 2 0 L L 2 0 n=1 am L , 2 = By Lemma 3. Then we get 2 = L am Z L f (x) cos 0 mπ x , L ∀m ∈ N. In summary, we have am = 2 L Z L f (x) cos 0 mπ x dx, L for all m ≥ 0 . Example 8. Solve the problem: ut = uxx , 0 < x < 1, t > 0, BC: ux (0, t) = 0, ux (1, t) = 0, IC: u(x, 0) = x, 0 ≤ x ≤ 1. t > 0, By the result in the part of Solving the heat equation without source with homogeneous Neumann boundary condition in Lecture 8, we know that for any n ∈ N, we can find a non-zero solution to ut = uxx , 0 < x < 1, t > 0, : BC: u (0, t) = 0, u (1, t) = 0, t > 0. x x un (x, t) = e−n 2 π2 t cos(nπx). Now let’s write x as an “infinitely” linear combination of {cos(nπx)}∞ n=0 : ∞ x= a0 X + an cos(nπx), 2 n=1 ∀x ∈ [0, 1]. Then for all n ≥ 0, we have an = Then 2 1 Z 1 Z x cos(nπx) dx = 2 0 1 x cos(nπx) dx. 0 12 MINGFENG ZHAO • When n = 0, then 1 Z x dx = 2 · an = 2 0 x2 2 1 = 1. 0 • When n ≥ 1, use the integration by parts, we have Z Z 1 x cos(nπx) dx = x d sin(nπx) · nπ Z 1 x d sin(nπx) = nπ Z 1 = x sin(nπx) − sin(nπx) dx nπ 1 1 = x sin(nπx) + cos(nπx) + C nπ nπ = x sin(nπx) cos(nπx) + + C. nπ n2 π 2 So we get 1 x sin(nπx) cos(nπx) = 2 + nπ n2 π 2 0 sin(nπ) cos(nπ) 1 = 2 + −2· 2 2 nπ n2 π 2 n π an = = = 2 cos(nπ) − 2 n2 π 2 2[(−1)n − 1] n2 π 2 0, if n is even, −4 , if n = 2k + 1. (2k + 1)2 π 2 Then x= ∞ 1 4 X 1 − 2 cos[(2k + 1)πx], 2 π (2k + 1)2 ∀x ∈ [0, 1]. k=0 So the solution is: u(x, t) = ∞ 1 4 X −(2k+1)2 π2 t cos[(2k + 1)πx] − 2 e · . 2 π (2k + 1)2 k=0 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