Math 118 :: Winter 2009 :: Midterm solutions 1. • Split e−ikx = cos(kx) − i sin(kx). Since the function is odd, the cosine terms give a zero contribution. We have Z Z π cos(kx) π 1 π cos(kx)dx − x sin(kx)xdx = . k −π k −π −π The integral vanishes. Notice that cos(kx) = (−1)k . What is left is −2π (−1)k , k Hence the Fourier series expansion is x = fˆk = −i Z k 6= 0. 1 2π P∞ ikx fˆ , k k=−∞ e π sin(kx)x dx = 2iπ −π (−1)k , k with fˆ0 = 0 and k 6= 0. • By comparing areas, we see that N X k −1 N +1 Z ≥ x−1 dx. 1 1 The latter integral diverges, because log(N + 1) → ∞ as N → ∞. • The FS coeffcients decay like a constant times k −1 in modulus, so they are not absolutely summable. • Consider g(x) = πx for x ∈ [−1, 1]. The trick for Chebyshev expansions is to view g as h(θ) = g(cos θ) = π cos θ, θ ∈ [0, 2π], and to perform a FS of h with cosine terms only (by evenness in θ.) Now use the formulas Z ĥk = 2π cos (kθ)h(θ) dθ, 0 h(θ) = ∞ 1 1X cos (kθ)ĥk , ĥ0 + 2π π k ∈ Z+ . k=1 For h(θ) = π cos θ the expansion has only one term – h itself – corresponding to k = 1. Another way to see this is to compute the inner product of cos θ times cos(kθ), and to notice that it is nonzero only for k = 1. We identify ĥ1 = π 2 , and all the other ĥk are zero. Back to g(x) = πx, we have the Chebyshev expansion 1 πx = ĥ1 T1 (x) = πT1 (x), π with T1 (x) = x. • The Chebyshev expansion has only one term, vs. a FS that decays like 1/k, so the Chebyshev expansion converges a lot faster. Because of the x = cos θ trick, the fast decay of the Cheb. expansion is inherited from the fast decay of the FS of a smooth periodic function. On the other hand x is not periodic in any interval, so the truncated FS suffers from the Gibbs effect. 2. We saw in class that Z sin k = k ∞ e−ikx −∞ 1 χ (x) dx. 2 [−1,1] So by Plancherel, Z 1 2π ∞ −∞ Hence Z ∞ sin k k −∞ 3. ĝk = h 2 1 1 dx. 4 −1 sin k k N X Z dk = 2 dk = π. e−ikjh (−1)j fj . j=1 Thanks to the relation π/h = N/2, and −1 = eiπ , we can combine two factors as e−ikjh (−1)j = e−ij(k−N/2)h . So we have ĝk = fˆk−N/2 . Or ĝk = fˆk+N/2 is also correct. 4. • Z ix |e − 1| = |(−i) • So | f (x) − f (y) |≤ x−y Z |e−iky | | iy Z e dy| ≤ 0 f (x) − f (y) = x−y x Z x 1dy = x. 0 e−ikx − e−iky ˆ f (k) dk. x−y e−ik(x−y) − 1 ˆ | |f (k)| dk ≤ x−y Z |k| |fˆ(k)| dk.