AMath 567, Autumn 2011 Sample Solutions for Assignment 8. Reading: Sec. 7.1-7.4, Sec. 8.1-8.2. 1. Apply Parseval’s relation to the function f (x) = x on (−π, π) to evaluate The Fourier coefficients of f (x) = x are: fˆ(0) = n 6= 0, √1 2π Rπ −π P∞ 1 n=1 n2 . x dx = 0, and for 1 Z π −inx fˆ(n) = √ xe dx 2π −π " # Z π i −inx π i −inx 1 x e e dx (using integration by parts) = √ − n −π n 2π −π 1 i = √ π (e−inπ + einπ ) + (e−inπ − einπ ) 2π n √ i √ i = 2π cos(nπ) = ± 2π . n n By Parseval’s identity, the sum of squares of absolute values of the Fourier coefficients is equal to the square of the L2 [−π, π]-norm of x, namely, Z π x2 dx = −π 2π 3 . 3 Since fˆ(0) = 0 and fˆ(−n) = −fˆ(n), the sum of squares of the Fourier coefficients is twice the sum of squares of those corresponding to positive n; hence ∞ ∞ X X 2π 3 1 π2 1 = , or, = . 4π 2 2 3 6 n=1 n n=1 n 2. Exercise 7.3 on p. 183. First note that the functions 1 1 1 √ , √ cos(kx), √ sin(kx), k = 1, 2, . . . , π π 2π form an orthonormal basis for L2 [−π, π]. Since the functions cos(kx) , k = 0, 1, 2, . . . are even functions, we have Rπ Rπ −π cos(kx) cos(jx) dx = 2 0 cos(kx) cos(jx) dx, and so the set √ 1 2 √ , √ cos(kx), k = 1, 2, . . . , π π 1 is an orthonormal set in L2 [0, π]. RSimilarly, since the functions sin(kx), k = Rπ π 1, 2, . . . are odd functions, we have −π sin(kx) sin(jx) dx = 2 0 sin(kx) sin(jx) dx, and so the set √ 2 √ sin(kx), k = 1, 2, . . . , π is an orthonormal set in L2 [0, π]. Let f ∈ L2 [0, π]. Then f can be extended to [−π, π] in such a way that the extended function is in L2 [−π, π] and is either even or odd: ( fe (x) = f (−x) −π ≤ x < 0 , fo (x) = f (x) 0≤x≤π ( −f (−x) −π ≤ x < 0 . f (x) 0≤x≤π In either case, let g denote the extended function. Then g can be expanded in a Fourier series consisting of sin’s and cos’s: g(x) = where ∞ a0 X (ak cos(kx) + bk sin(kx)), + 2 k=1 1Zπ g(x) cos(kx) dx, k = 0, 1, 2, . . . , ak = π −π 1Zπ bk = g(x) sin(kx) dx, k = 1, 2, . . . . π −π Since fe (x) is an even function and sin(kx) is odd and cos(kx) is even, we have Z π −π fe (x) sin(kx) dx = 0, Z π −π fe (x) cos(kx) dx = 2 Z π 0 fe (x) cos(kx) dx. Thus f (x) can be expanded on [0, π] as f (x) = ∞ α0 X + αk cos(kx), 2 k=1 where 2Zπ f (x) cos(kx) dx. π 0 P Equivalently, f (x) = ∞ α̂ f (x), where the fk ’s are the functions defined k=0 Rπ k k in the problem and α̂k = 0 f (x)fk (x) dx. Since this series converges to fe in L2 [−π, π], it converges to f in L2 [0, π], and so the functions fk , k = 0, 1, 2, . . ., form a complete orthonormal set, i.e., an orthonormal basis for L2 [0, π]. Similarly, since fo (x) is an odd function, we have αk = Z π −π fo (x) cos(kx) dx = 0, Z π −π fo (x) sin(kx) dx = 2 2 Z π 0 fo (x) sin(kx) dx. Thus f (x) can be expanded on [0, π] as ∞ X f (x) = βk sin(kx), k=1 where 2Zπ βk = f (x) sin(kx) dx. π 0 P β̂ e (x), where the ek ’s are the functions defined Equivalently, f (x) = ∞ k=1 Rπ k k in the problem and β̂k = 0 f (x)ek (x) dx. Since this series converges to fo in L2 [−π, π], it converges to f in L2 [0, π], and so the functions ek , k = 1, 2, . . ., form a complete orthonormal set, i.e., an orthonormal basis for L2 [0, π]. 3. Exercise 7.7 on p. 184. Let y = (π/L)x and define v(y, t) ≡ u(x, t) = u((L/π)y, t). Then vt (y, t) = ut (x, t) and vyy (y, t) = (L/π)2 uxx (x, t). Thus v(y, t) satisfies the initialboundary value problem: vt = (π/L)2 vyy , v(0, t) = v(π, t) = 0, t > 0, v(y, 0) = f ((L/π)y), 0 ≤ y ≤ π. q Since the set of sin functions { 2/π sin(ny) : n = 1, 2, . . .} form an orthonormal basis for L2 [0, π], f ((L/π)y) can be expanded in terms of these functions: ∞ X 2Zπ fn sin(ny), fn = f ((L/π)s) sin(ns) ds. f ((L/π)y) = π 0 n=1 Look for a solution v(y, t) of the form v(y, t) = tiating term-by-term, we find vt (y, t) = ∞ X P∞ vn0 (t) sin(ny), vyy (y, t) = − n=1 ∞ X vn (t) sin(ny). Differenvn (t)n2 sin(ny). n=1 n=1 2 Equating Fourier coefficients of vt and (π/L) vyy gives vn0 (t) = −(π/L)2 n2 vn (t), n = 1, 2, . . . . Solving for vn and applying the initial condition: 2 n2 t vn (t) = e−(π/L) vn (0) = e−(π/L) 2 n2 t fn . Substituting this into the expression for v(y, t) gives ∞ X v(y, t) = 2 n2 t e−(π/L) fn sin(ny). n=1 Finally, since v(y, t) = u(x, t), u(x, t) = ∞ X 2 n2 t e−(π/L) n=1 3 fn sin(n(π/L)x). 4. Exercise 7.10 on p. 184. If u(x, t) = f (x + ct) + g(x − ct), then, differentiating using the chain rule, we find ut (x, t) = cf 0 (x+ct)−cg 0 (x−ct), so that utt (x, t) = c2 f 00 (x+ct)+c2 g 00 (x− ct); ux (x, t) = f 0 (x+ct)+g 0 (x−ct), so that uxx (x, t) = f 00 (x+ct)+g 00 (x+ct). Thus, utt = c2 uxx . 5. Exercise 8.8 on p. 213. Let ϕx (z) = hx, zi and ϕy (z) = hy, zi be two linear functionals in H ∗ and define an inner product on H ∗ by hϕx , ϕy i = hy, xi. First check that this is an inner product: (a) hϕx , ϕx i = hx, xi ≥ 0 with equality if and only if x = 0, which is the case if and only if ϕx = 0; (b) hϕy , ϕx i = hx, yi = hy, xi = hϕx , ϕy i; (c) and for any scalars α and β, hϕx , αϕy + βϕw i = hϕx , ϕᾱy+β̄w i = hᾱy + β̄w, xi = αhy, xi + βhw, xi = αhϕx , ϕy i + βhϕx , ϕw i. Now show that H ∗ is complete with the norm defined by this inner product. Let (ϕxn ) be a Cauchy sequence in H ∗ . Then given > 0 there exists N such that n, m > N implies kϕxn − ϕxm k = hϕxn − ϕxm , ϕxn − ϕxm i1/2 < . But since hϕxn − ϕxm , ϕxn − ϕxm i = hxn − xm , xn − xm i, this means that (xn ) is a Cauchy sequence in H, hence has a limit x ∈ H. Claim: ϕxn converges to ϕx in H ∗ . To see this, note that kϕxn − ϕx k = hϕxn − ϕx , ϕxn − ϕx i1/2 = hxn − x, xn − xi1/2 = kxn − xk → 0. Thus any Cauchy sequence (ϕxn ) in H ∗ has a limit ϕx in H ∗ , where x = limn→∞ xn , so H ∗ is a Hilbert space with this inner product. 4