LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES MINGFENG ZHAO July 23, 2015 Fourier series Example 1. Assume that f (x) = x, 0 < x < 2 represents one full period of the function. Find the Fourier series. By the assumption, we know that the period of f is 2, then the Fourier series of f is: ∞ f (x) = a0 X + [an cos(nπx) + bn sin(nπx)] , 2 n=1 where an = 2 Z 1 1 Z f (x) cos(nπx) dx ∀n ≥ 0, 0 2 = x cos(nπx) dx 0 = bn = = 2, if n = 0, 0, if n ≥ 1 Z 1 2 f (x) sin(nπx) dx 1 0 Z 2 x sin(nπx) 0 = − 2 , nπ ∀n ≥ 1. So we get x=1− ∞ 2 X sin(nπx) , π n=1 n ∀x ∈ (0, 2). Even and odd functions Definition 1. Let f be a function on the domain I which is symmetric with respect to 0 (that is, x ∈ I if and only if −x ∈ I), then 1 2 MINGFENG ZHAO I. we say that f is an even function if f (−x) = f (x) for all x ∈ I. II. we say that f is an odd function if f (−x) = −f (x) for all x ∈ I. Remark 1. It’s easy to see the following properties of even and odd functions: • The sum (difference) and product (quotient) of two even functions are even. • The sum (difference) of two odd functions is odd; the product (quotient) of two odd functions is even. • The sum (difference) of an odd function and an even function is neither even nor odd; the product (quotient) of two such functions is odd. • Any function can be expressed as a sum of an even function and an odd function. Lemma 1. For any L > 0, then 0, Z f (x) dx = −L 2 Z if f is odd in [−L, L], L L f (x) dx, if f is even in [−L, L]. 0 Proof. In fact, we have Z L 0 Z f (x) dx L Z = f (x) dx + −L −L f (x) dx 0 0 Z Z f (−x) · −dx + = f (x) dx L 0 L Z L = Z f (−x) dx + 0 L f (x) dx 0 L Z = [f (−x) + f (x)] dx 0 = 0, Z 2 if f is odd in [−L, L], L f (x) dx, if f is even in [−L, L]. 0 Remark 2. Let f be a function on [−L, L], then the Fourier series of f (x) is given by: ∞ f (x) = nπ nπ i a0 X h + an cos x + bn sin x , 2 L L n=1 where an = 1 L Z L f (x) cos −L nπ x dx L ∀n ≥ 0, LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES bn = 1 L L Z f (x) sin −L nπ x dx L 3 ∀n ≥ 1. Then nπ nπ x is odd, and f (x) sin x is even. By L L • If f is an odd function on [−L, L], then for any n ≥ 0, f (x) cos Lemma 1, then an bn 0 ∀n ≥ 0, Z nπ 2 L f (x) sin x dx, = L 0 L = ∀n ≥ 1. Then we get f (x) = ∞ X bn sin n=1 nπ x , L and bn = 2 L L Z f (x) sin 0 mπ x , L ∀n ≥ 1. nπ nπ x is even, and f (x) sin x is odd. By L L • If f is an even function on [−L, L], then for any n ≥ 0, f (x) cos Lemma 1, then an bn = = 2 L L Z 0, f (x) cos 0 nπ x dx L ∀n ≥ 0, ∀n ≥ 1. Then we get ∞ f (x) = nπ a0 X + an cos x , 2 L n=1 and an = 2 L L Z f (x) cos 0 mπ x , L ∀n ≥ 0. Half-range expansions If we are given a function f on [0, L], and we want to represent f by a Fourier series with period 2L, then we have two choices: f (x), 0<x<L • Fourier sine series: We get an odd extension of f on [−L, L], that is, fo (x) = , −f (−x), −L < x < 0 and fo (x + 2L) = fo (x) for all x ∈ R, then ∞ X nπ fo (x) = bn sin x , L n=1 and 2 bn = L Z L f (x) sin 0 mπ x , L ∀n ≥ 1. 4 MINGFENG ZHAO f (x), 0<x<L • Fourier cosine series: We get an even extension of f on [−L, L], that is, fe (x) = , f (−x), −L < x < 0 and fe (x + 2L) = fe (x) for all x ∈ R, then ∞ nπ a0 X fe (x) = + x , an cos 2 L n=1 2 an = L and Z L f (x) cos 0 mπ x , L ∀n ≥ 0. Example 2. Expand f (x) = x, 0 ≤ x ≤ 2 in a half-range: (a) Fourier sine series; (b) Fourier cosine series. x, 0 < x < 2 (a) Fourier sine series: The odd extension of f (x) is: fo (x) = , and fo (x + 4) = fo (x) for all x ∈ R, x, −2 < x < 0 then x= ∞ X bn sin n=1 nπ x , 2 0 < x < 2, where bn Z 2 nπ x dx 2 0 2 nπ x sin x dx 2 0 2 2 Z = = x sin 4(−1)n+1 . nπ = Then ∞ nπ 4 X (−1)n+1 x= sin x , x ∈ (−2, 2). π n=1 n 2 x, 0<x<2 (a) Fourier cosine series: The even extension of f (x) is: fe (x) = , and fe (x + 4) = fe (x) for all −x, −2 < x < 0 x ∈ R, then ∞ nπ a0 X + an cos x , 0 < x < 2, x= 2 2 n=1 where an 2 2 Z Z 2 nπ x dx 2 0 2 nπ = x cos x dx 2 0 if n = 0, 2, = n 4[(−1) − 1] , if n ≥ 1. 2 n π2 = x cos LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES Then ∞ ∞ nπ X 4 X (−1)n − 1 x = 1+ cos x π 2 n=1 n2 2 n=1 ∞ (2k + 1)π 8 X 1 cos = 1− 2 x , π (2k + 1)2 2 ∀x ∈ [−2, 2]. k=0 Complex form of Fourier series Recall Euler’s identity eiθ = cos(θ) + i sin(θ), then we have cos(θ) = eiθ + e−iθ , 2 and sin(θ) = eiθ − e−iθ eiθ − e−iθ = −i · . 2i 2 Let ∞ f (x) = nπ i nπ a0 X h + x + bn sin x . an cos 2 L L n=1 Then we have f (x) = = = Let c0 = a0 1 = 2 2L Z ∞ nπ nπ i a0 X h + an cos x + bn sin x 2 L L n=1 nπ nπ nπ nπ ∞ a0 X ei L x − e−i L x ei L x + e−i L x + − ibn · an · 2 2 2 n=1 ∞ ∞ a0 X an − ibn i nπ x X an + ibn −i nπ x + e L + e L . 2 2 2 n=1 n=1 L f (x) dx, for any n ≥ 1, let −L cn = an − ibn , 2 an + ibn = cn . 2 and c−n = Then we get f (x) = ∞ X cn ei nπ L x . n=−∞ Since 1 an = L Z L nπ f (x) cos x dx, L −L 1 and bn = L Z L f (x) sin −L nπ x dx, L Then for all n ≥ 1, we have cn = = an − ibn 2 Z L h nπ nπ i 1 f (x) cos x − i sin x dx 2L −L L L ∀n ≥ 1. 5 6 MINGFENG ZHAO = c−n = = = 1 2L L Z f (x)e−i nπ L x dx −L an + ibn 2 Z L h nπ nπ i 1 f (x) cos x + i sin x dx 2L −L L L Z L nπ 1 f (x)ei L x dx. 2L −L In summary, we have cm = Z 1 2L L f (x)e−i mπ L x dx, ∀m ∈ Z. −L −1, if −π ≤ x < 0, Example 3. Let f (x) = 1, if 0 < x < π. Then cn π f (x)e−inx dx −π Z 0 Z π 1 −inx −inx − e dx + e dx . 2π −π 0 = Z Z 1 2π = 0 For e−inx dx, then −π Z 0 e −inx Z dx 0 einy · −dy = −π π Z π = Let y = −x, then dx = −dy einy dy 0 Z = π einx dx Let y = x. 0 So we get cn = = = = Z π Z π 1 − einx dx + e−inx dx 2π 0 0 Z π 1 [−einx + e−inx ] dx 2π 0 Z −i π sin(nx) dx π 0 0, if n is even −2i , if n is odd. nπ LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES 7 So we get ∞ X f (x) = − k=−∞ 2i ei(2k+1)x . (2k + 1)π Convergence of Fourier series Theorem 1. Suppose that f and f 0 are piecewise continuous on the interval [−L, L). Suppose that f is defined outside the interval [−L, L) so that it is periodic with period 2L. Then f has a Fourier series: ∞ nπ i nπ a0 X h + x + bn sin x , an cos 2 L L n=1 f (x) = where an = bn = 1 L L Z nπ x dx ∀n ≥ 0, L −L Z nπ 1 L x dx ∀n ≥ 1. f (x) sin L −L L f (x) cos Moreover, the Fourier series converges to f (x) at all points where f is continuous, and it converges to at all points where f is discontinuous. Proof. For any N ≥ 1, let N SN (x) = nπ nπ i a0 X h + an cos x + bn sin x . 2 L L n=1 For any x ∈ R and fix it, for any n ∈ Z, then Z x−L Z L nπ y −i nπ L dy = f (x − z)e−i L (x−z) · −dz f (y)e −L Let y = x − z x+L Z x+L f (x − z)ei = nπ L z · e−i nπ L x dz x−L = e−i nπ L x Z x+L f (x − z)ei nπ L z dz x−L = e−i nπ L x Z L f (x − z)ei nπ L z dz Since f has period 2L. −L By the computation in Complex form of Fourier series in Lecture 11, then SN (x) = N X cn e i nπ L x n=−N = N X n=−N e i nπ L x 1 · 2L Z L −L f (y)e−i nπ L y dy f (x+) + f (x−) 2 8 MINGFENG ZHAO 1 2L = For N X ei nπ L y , if y = 0, then n=−N N X ei nπ L y Z " L f (x − y) · −L # N X e = 2N + 1; if x 6= y, let θ = ei nπ L z dy. n=−N n=−N N X i nπ L y N X = n=−N π y, then L einθ n=−N = e−iN θ − ei(N +1)θ 1 − eiθ = e−iN θ − ei(N +1)θ e−i 2 · θ 1 − eiθ e−i 2 θ e−i = 2N +1 θ 2 − ei θ 2N +1 θ 2 θ e−i 2 − ei 2 2N +1 2 θ −2i sin θ2 −2i sin = 2N +1 2 θ sin θ2 sin = sin = (2N +1)π y 2L π sin 2L z . Let 2N + 1, if z = 0, DN (z) = (2N +1)π sin z 2L , if z 6= 0. π sin 2L z Notice that Z L Z Dn (z) dz L = −L N X ei nπ L z dz −L n=−N Z L = dz + −L = X 2L + 1≤|n|≤N X 1≤|n|≤N = 2L, Z Z L ei nπ L z dz −L L h nπ nπ i cos z + i sin z dz L L −L By Lemma 3 and Lemma 4 in Lecture 10. LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES 9 So we get |f (x) − SN (x)| = = = ≤ Let F (z) = Z L 1 f (x − z)DN (z) dz f (x) − 2L −L Z L 1 Z L 1 f (x)DN (z) dz − f (x − z)DN (z) dz 2L −L 2L −L 1 Z L [f (x) − f (x − z)]DN (z) dz 2L −L Z L 1 |f (x) − f (x − z)||DN (z)| dz. 2L −L f (x) − f (x − z) for z 6=, then z 1 |f (x) − SN (x)| ≤ 2L Z L |F (z)| · |zDN (z)| dz. −L If f is differentiable at x, one can use the Riemann’s Lemma to prove that |f (x) − SN (x)| → 0, as N → ∞. Remark 3. Riemann’s Lemma states that: for any integrable function f on [−L, L], then Z L f (x)einx dx = 0. lim n→∞ −L 0, if −L < x < 0, Example 4. Let L > 0 and f (x) = , and let f be defined outside (−L, L) so that f (x + 2L) = L, if 0 < x < L f (x) for all x. Find the Fourier series for f and determine where it converges. It’s easy to see that f satisfies the conditions in Theorem 1, then f has a Fourier series: ∞ nπ nπ i a0 X h f (x) = + an cos x + bn sin x , 2 L L n=1 where a0 = = an 1 L Z 1 L Z = 1 = 1 L L f (x) dx −L L L dx 0 Z L f (x) cos −L nπ x dx L 10 MINGFENG ZHAO Z nπ 1 L L cos x dx L 0 L Z L nπ = cos x dx L 0 nπ L L = sin x nπ L = 0 = = bn = = = = L sin(nπ) nπ 0 ∀n ≥ 1, Z nπ 1 L f (x) sin x dx L −L L Z nπ 1 L L sin x dx L 0 L Z L nπ x dx sin L 0 nπ L L − cos x nπ L 0 L [cos(nπ) − 1] nπ 1 − (−1)n , ∀n ≥ 1. nπ = − = So we get ∞ f (x) = = nπ L X 1 − (−1)n + sin x 2 n=1 nπ L ∞ L 2X 1 (2k + 1)π + sin x . 2 π 2k + 1 L k=0 Since f is continuous in (−L, 0) and (0, L), then ∞ L 2X 1 (2k + 1)π f (x) = + sin x , 2 π 2k + 1 L ∀x 6= ZL. k=0 Notice that f can not be continuous at x = ZL, but we have f (0−) + f (0+) f (L−) + f (L+) f (mL−) + f (mL+) L = = = , 2 2 2 2 On the other hand, we know that ∞ L 2X 1 (2k + 1)π L + sin · (mL) = , 2 π 2k + 1 L 2 k=0 ∀m ∈ Z. ∀m ∈ Z. LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES 11 Gibbs phenomenon Theorem 2. Let f be a piecewise continuously differentiable function which is periodic with period 2L > 0. Suppose that f is not continuous at x0 but f (x0 +) and f (x0 −) exists such that f (x0 +) − f (x0 −) = α 6= 0. For the Fourier series of f : ∞ f (x) = nπ i nπ a0 X h + an cos x + bn sin x , 2 L L n=1 where Let SN (x) = an = bn = 1 L Z L nπ x dx ∀n ≥ 0, L −L Z nπ 1 L x dx ∀n ≥ 1. f (x) sin L −L L f (x) cos N −1 h nπ i nπ X a0 + x + bn sin x , then there exists a positive constant C > 0 such that an cos 2 L L n=1 L lim SN x0 + = f (x0 +) + C · α N →∞ N L = f (x0 −) − C · α lim SN x0 − N →∞ N −1, −π ≤ x < 0, Example 5. Let f (x) = , then f is odd, and the Fourier series of f is: 1, 0≤x<π ∞ f (x) = a0 X [an cos(nx) + bn sin(nπ)] . + 2 n=1 Then we have an = = bn = Z 1 1 f (x) cos(nx) dx −1 for all n ≥ 0, 0, 1 1 Z 1 f (x) sin(nx) dx −1 Z = 1 2 1 sin(nx) dx 0 12 MINGFENG ZHAO = 2[1 − (−1)n ] , nπ for all n ≥ 1. So we have f (x) = ∞ 2 X 1 − (−1)n sin(nx) π n=1 n = ∞ 4X 1 sin[(2k + 1)x], π 2k + 1 ∀x ∈ (−π, 0) [ (0, π). k=0 For any N ≥ 1, let SN (x) = N −1 4 X sin[(2k + 1)x] · . π 2k + 1 k=0 Then SN 2π 2N = N −1 2π 4 X sin (2k + 1) 2N · π 2k + 1 k=0 = = = N −1 4 2π X sin 2k+1 2N · 2π · · 2k+1 π 2N 2N · 2π k=0 N −1 2 X sin 2k+1 2N · 2π · 2k+1 N 2N · 2π k=0 N −1 1 2π X sin 2k+1 2N · 2π · · . 2k+1 π N 2N · 2π k=0 It’s easy to see that SN 2π 2N is exactly the midpoint Riemann sum of sin(x) with N intervals over [0, 2π], which x implies that lim SN N →∞ Z 2π For π 2π 2N 1 = π Z 2π sin(x) 1 dx = x π 0 Z 0 π sin(x) dx + x Z 2π π sin(x) dx . x sin(x) dx, we have x Z 2π π π Z sin(x) dx x sin(y + π) dy y+π = 0 π sin(y) dy y+π π 0 sin(x) dx x+π Z = − 0 Z = − Let x = y + π Let y = x. Then lim SN N →∞ 2π 2N 1 = π Z 0 π 1 1 sin(x) · − x x+π Z dx = 0 π sin(x) dx > 0. x(x + π) LECTURE 11: HALF RANGE FOURIER SERIES AND CONVERGENCE OF FOURIER SERIES 13 Since SN (x) is odd, then lim SN N →∞ 2π − 2N Z =− 0 π sin(x) dx < 0. x(x + π) Notice that f (0+) − f (0−) = 2. 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