MATH 321: Real Variables II Lecture #30 University of British Columbia Lecture #30: Instructor: Scribe: March 26, 2008 Dr. Joel Feldman Peter Wong Theorem. (Pointwise Convergence of Fourier Series) Assume: f : R → C be 2π-periodic and Riemann integrable on [−π, π]; x ∈ R |f (x + t) − f (x+ ) ≤ M |t| for all 0 < x < δ |f (x + t) − f (x− ) ≤ M |t| for all − δ < x < 0 f (x+ ) f (x− ) = limt→0 t<0 for some δ > 0 and M < ∞, then limn→∞ Sn = 21 [f (x+ ) + f (x− )] Proof. Last class we defined f (x−t)−f (x+ ) sin(t/2) , if t ∈ [−π, 0), (x− ) g(t) = f (x−t)−f , if t ∈ (0, π], sin(t/2) 0, if t = 0. and showed that Sn (x) − 12 [f (x+ ) − f (x− )] Z π i 1 = 4πi g(t)e 2 t eint dt − = −π 1 (nth Fourier 2i 1 4πi π Z i g(t)e− 2 t e−int dt −π coefficient of g(t)e(i/2)t ) − 1 2i (−nth Fourier coefficient of g(t)e−(i/2)t ) so the Theorem follows from Lemma. g(t) is Riemann integrable on [−π, π]. Proof. Let ε > 0, ε < δ. Note that 1 sin(t/2) (1) On [−π, −ε] ∪ [ε, π], g(t) is Riemann-integrable since f (x − t), f (x± ), are integrable. (2) There is a constant M 0 > 0 such that |g(t)| ≤ M 0 for all t ∈ [−ε, ε] because M(t/2) |g(t)| ≤ 2 sin(t/2) ≤ M0 since limt→0 Z t sin(t/2) π −π g(t) dt − Z = 2. Thus, π g(t) dt = −π = Z Z −ε g(t) dt + −π ε g(t) dt − ε −ε Z ε −ε π g(t) dt − Z −ε −π g(t) dt − Z ε g(t) dt − −ε Z π g(t) dt ε ε g(t) dt ε −ε R g(t) dt + g(t) dt < M 0 (2ε), Z for all 0 < ε < δ. This means Z ε −ε −ε R R ε Since −ε g(t) dt < M 0 (2ε) and Z g(t) dt − Z R = . Hence, g ∈ R. ε g(t) dt ≤ 4M 0 ε −ε 2 MATH 321: Lecture #30 Theorem. (Uniform Convergence of Fourier Series) If f : R → C is 2π-periodic and is piecewise C 1 , then Sn (f, x) converges uniformly to f on any closed interval that contains no point of discontinuity of f , Definition. f is piecewise C 1 on [−π, π] if there exist n ∈ N and −π = x0 < x1 < x2 < · · · < xn = π such that (1) for each 1 ≤ j ≤ n, f has at least one continuous derivative on (xj−1 , xj ) (2) limt→0± f (xj + t) and limt→0± f 0 (xj + t) exist. That is, f (x+ j ), f (x− j ), f 0 (x+ j ), f 0 (x− j ) −π = x0 x2 x3 x4 = π m=−∞ cm e imx , then f 0 has x1 all exist. Proof of Theorem: Step 1: f is continuous and piecewise C 1 . P∞ imx m=−∞ im cm e If f has Fourier series expansion P∞ MATH 321: Real Variables II Lecture #31 University of British Columbia Lecture #31: Instructor: Scribe: March 28, 2008 Dr. Joel Feldman Peter Wong Theorem. (Uniform Convergence of Fourier Series) If f : R → C is 2π-periodic and is piecewise C 1 , then the Fourier series of f converges uniformly on any closed interval that does not contain a point of discontinuity of f . [ ][ ][ ] Proof. Step 1: If f is continuous and piecewise C 1 , then the Fourier series of f converges uniformly to f . Step (1a): If f has Fourier series P∞ n=−∞ an einx , then f 0 has Fourier series P∞ n=−∞ in an einx . Proof. The n-th Fourier coefficient of f 0 is Z π 1 f 0 (x)e−inx dx 2π −π n Z 1 X xk 0 f (x)e−inx dx = 2π xk−1 k=1 (where f 0 (x) is continuous on [−π, π] except at −π = x0 < x1 < x2 < · · · < xn = π) Z xk −ε n 1 X = lim f 0 (x)e−inx dx ε→0 x 2π +ε k−1 k=1 ) ( xk −ε Z xk −ε n X 1 −inx −inx + in f (x)e dx = lim f (x)e ε→0 2π xk−1 xk−1 +ε k=1 x− n n Z X k =xk 1 X xk 1 −inx +(in) = f (x)e−inx dx f (x)e + 2π 2π x x =x k−1 k−1 k=1 k=1 k−1 | {z } {z } | the xk with k = 1, . . . , n − 1 contributions cancel; and the x0 = −π, xn = π terms cancel by periodicity 1 Rπ f (x)e−inx dx an = 2π −π = 0 + in an = in an . Step (1b): If f is piecewise C 1 and continuous, the Fourier series converges uniformly. We know ∞ X an einx n=−∞ converges pointwise to f (x) since |f (x + t) − f (x)| ≤ sup |f 0 ||t| ≤ M |t| and we know ∞ X n=−∞ in an einx 2 MATH 321: Lecture #31 converges in the mean to f 0 (x). So for each x, ∞ N N X X X X inx inx inx inx an e an e = an e − an e = f (x) − n=−∞ n=−N n=−N |n|>N s s X X X X 1 1 n an ≤ |n an |2 |an | = ≤ n n2 But P∞ 1 n=−∞ n2 converges and lim N →∞ Step 2: Let F (x) = ( π−x 2π , 0, P∞ n=−∞ s X |n|>N |n|>N |n|>N 1 n2 |n an |2 converges to =0 and |n|>N lim N →∞ 1 2π Rπ s X −π |n|>N |f 0 (x)|2 dx < ∞. So, |n an |2 |n|>N 0 < x < 2π x = 0, 2 y 1 2 b −3π b b −π −2π π 2π 3π x and F (x + 2π) = F (x) for all x ∈ R. By Problem Set 9, #3, F (x) has Foureir series ∞ X 1 2πi 1 inx ne n=−∞ n6=0 which converges uniformly on [ε, 2π − ε] for any ε > 0. Step 3: (Proof of theorem) Suppose f has jump discontinuities on [−π, π) at p1 , . . . , pm . Let, for 1 ≤ j ≤ m, − let δj = f (p+ j ) − f (pj ). Write f (x) = n X δj F (x − pj ) + f (x) − δj F (x − pj ) j=1 j=1 | n X {z } FS conv. uniformly except at pj | {z } continuous FS converges uniformly