Real Analysis – September 2011 Part A 1. Let E and F be Lebesgue measurable subsets of R, with E ⊂ F. Suppose that F is Borel measurable, but suppose that E is not Borel measurable. Prove or disprove: F − E is uncountable. True. Proof: By way of contradiction, assume that F − E is countable, thus F − E is Borel measurable. Now, notice that Ec = ( F − E) ∪ F c , where both F − E and F c are Borel measurable. So, Ec is Borel measurable, and thus, E would also be Borel measurable, which is a contradiction. Thus, F − E must be uncountable. 2. Let f : [0, 1] → [0, 1] be a nondecreasing continuous function satisfying f (0) = 0 and f (1) = 1. Prove or give a counterexample: Z 1 0 f 0 ( x ) dx = 1. False. Counterexample: Cantor function Let C be the Cantor set. n That is, let o −n , where each a = 0 or 2 C = x ∈ [0, 1] : x = ∑∞ a 3 n n =1 n ( ) m − 1 ∞ 3 [−1 [ 3k + 1 3k + 2 = x ∈ [0, 1] \ , m 3 3m m =1 k =0 n o = x ∈ [0, 1] \ “open middle thirds” . Notice that µ(C ) = 0, since C is totally disconnected, where µ is the Lebesgue measure. Define the Cantor function f : [0, 1] → [0, 1] as ∞ an −n 2 , if x ∈ Cas defined above ∑ 2 n = 1 f (x) = p 2−k , for some p, k ∈ Z, if x ∈ [0, 1] \ C so that if x ∈ [0, 1] \ C, then f ( x ) is equal to the value of f at the endpoints of the “open middle third” interval that x is in. Notice: f (0) = 0 and f (1) = 1, (i.e. f goes from 0 to 1 as x goes from 0 to 1). Also, f is continuous everywhere. Now, since µ(C ) = 0 and f 0 ( x ) = 0 for every x 6∈ C, f is differentiable almost everywhere with f 0 ≡ 0. Thus, f is nondecreasing, also. So f satisfies the hypotheses. Z 1 Z 1 However, 0 f ( x ) dx = 0 dx = 0 6= 1 = f (1) − f (0). 0 0 1 3. Prove or disprove: a. Every locally compact metric space is complete. b. Every locally compact inner product space is finite dimensional. a. False. Counterexample: Let E = (0, 1) ⊆ R. Then E is a locally compact metric space since for every x∈ (0, 1) x x+1 there exists some interval [ a, b] ⊂ (0, 1) such that x ∈ [ a, b]. Let [ a, b] = , . 2 2 However, E is not complete: consider the sequence ∞ 1 n n =2 1 1 1 which is Cauchy since − → 0 as m, n → ∞. However, lim = 0 6∈ E. Thus, E is n→∞ n m n not a complete metric place. b. True. Proof: By way of contradiction, assume that X is a locally compact inner product space with infinite dimension. Then, every inner product space has an orthonormal basis, there since ∞ exists a countable set u j j=1 of orthonormal basis elements such that ( 0 if j 6= k u j , uk = 1 if j = k Now, since X is locally compact, it’s locally compact at 0. So, there is a compact K and open U such that 0 ∈ U ⊆ K. Also, since U is open, there is some δ > 0 such that (−δ, δ) ⊆ U. δ δ Now, scale each u j by 2 so that 2 u j − 0 = 2δ u j = 2δ u j = 2δ · 1 = 2δ . Thus, δ 2 uj ∈ (−δ, δ) for all j. But, δ δ 2 u j − 2 uk = = = = = n rD δ 2 (u j − uk ) , 2δ (u j − uk ) E q u j − uk , u j − uk q δ u j , u j − 2 u j , uk + uk , uk 2 √ δ 2 1−0+1 δ 2 √δ 2 6→ 0 as j, k → ∞. o ⊆ K is not Cauchy and cannot have a convergent subsequence. But X is locally compact and K is compact, so every sequence in K should have a convergent subsequence. This has led us to a contradiction. Thus, X must be finite dimensional. Thus, δ 2 uj 2 4. Let f : [0, 1] → R be absolutely continuous, and assume that f 0 ∈ L2 ([0, 1]) and that f (0) = 0. Show that the following limit exists, and compute its value. lim x −1/2 f ( x ) x →0+ Since f is absolutely continuous, f 0 exists a.e. Z x and is integrable on [0, x ] for x ∈ [0, 1] and, by the Fundamental Theorem of Calculus, f 0 (t) dt = f ( x ) − f (0) = f ( x ). Also, note 0 Z 1 21 0 2 0 2 f (t) dt that f ∈ L ([0, 1]) implies that < ∞. 0 Now, 0 ≤ lim x −1/2 f ( x ) x →0+ = = ≤ = ≤ = = = lim x −1/2 · f ( x ) x →0+ Z x −1/2 0 lim x f ( t ) dt x →0+ 0 Z x 0 lim x −1/2 f ( t ) dt x →0+ 0 on [0, x ] lim x −1/2 f 0 (t) + 1 x →0 lim x −1/2 12 · f 0 (t) + x →0 2 lim x −1/2 x →0+ Z x 0 lim x −1/2 x1/2 x →0+ lim x →0+ Z R ! 21 |1|2 dt Z R By Hölder’s Inequality Z x 0 2 f ( t ) dt ! 12 0 0 2 χ[0,x] f (t) dt 2 χ[0,x] f 0 (t) dt ! 12 ! 21 Now, let { xn } be a sequence of positive real numbers in [0, 1] such that xn → 0 as 2 n → ∞. Define gn ( x ) := χ[0,xn ] f 0 ( x ) . Then, for x ∈ [0, 1], gn ( x ) → 0 as n → ∞ 2 and gn ( x ) ≤ f 0 ( x ) , which is integrable, so by the Lebesgue Dominated Convergence Theorem, Z 21 12 Z 12 Z 21 Z 0 2 = lim gn (t) dt = lim gn (t) dt = 0 dt = 0. lim χ[0,x] f (t) dt x →0+ R n→∞ R R n→∞ Thus, lim x −1/2 f ( x ) = 0. R x →0+ 3 Part B 5. Suppose that f : R → C is locally integrable (i.e., f · χK ∈ L1 (R), for every compact K), and suppose that g : R → C is continuous. Prove or disprove: a. If g has compact support, then f ∗ g is uniformly continuous. b. If g is bounded and if f has compact support, then f ∗ g is uniformly continuous. a. False. Counterexample: Let f (x) = ex if x ∈ [−1, 0] x + 1 g( x ) = − x + 1 if x ∈ (0, 1] 0 otherwise. and Then, g is bounded and continuous, and has compact support since g = 0 outside of [−1, 1]. Also, f is locally integrable. However, Z f ( x − y) g(y) dy f ∗ g (x) = R = = Z R e x−y g(y) dy Z −1 −∞ = ex = e x e x −y 0 Z −1 h · 0 dy + −1 Z 1 e−y y dy − − ye Z 0 −y −e −y = e x − 2 − e + e −1 0 i0 −1 e x −y (y + 1) dy + e−y y dy + − h Z 1 − ye −y Z 1 −1 −e 0 e−y dy −y i1 0 e x −y (−y + 1) dy + Z ∞ 1 e x−y · 0 dy + h −e −y i1 −1 which is not uniformly continuous since e x is not uniformly continuous. b. True. Since f has compact support, f = 0 outside some compact interval [− N, N ]. Since f is locally integrable, f is integrable on [− N, N ], so we have that Z Z N f = f = M f < ∞, for some M f . R −N Since compactly supported functions are dense in L1 (R Z ), given ε > 0 there exists some f ( x ) − h( x ) dx < ε. continuous function h with compact support such that R 4 Now, since g is bounded, there is some Mg such that | g| ≤ Mg . Then, f ∗ g ( x1 ) − f ∗ g ( x2 ) = = ≤ ≤ = = ≤ Z Z f ( x1 − y) g(y) dy − f ( x2 − y) g(y) dy R R Z f ( x − y ) − f ( x − y ) g ( y ) dy 2 1 R Z f ( x1 − y) − f ( x2 − y) g(y) dy R Mg Mg Mg Mg Z R f ( x1 − y) − f ( x2 − y) dy Z R f (v) − f (v + u) (−dv), if u = x2 − x1 and v = x1 − y Z R "Z f (v) − h(v) + h(v) − h(v + u) − h(v + u) − f (v + u) dv Z h(v) − h(v + u) dv f (v) − h(v) dv + R R # Z h(v + u) − f (v + u) dv + R Z R h(v) − h(v + u) dv + Mg ε < Mg ε + Mg < 2Mg ε + Mg (2R)ε ε 2Mg (1 + R) . = −R err... are we done? if |u| < δ (since h is continuous) 5 6. Consider a function f : R → R which is periodic with period one and which satisfies f ( x ) = x for | x | < 1/2. a. Compute the Fourier series for f . ∞ b. Use your result from item (a) to compute a. The Fourier series for f is So, c0 = Z 1 2 − 12 ∑ 1 . n2 n =1 ∑ cn e2πinx where cn = n ∈Z Z 1 2 − 12 f ( x )e−2πinx dx. x dx = 0, and for n 6= 0, cn = 1 2 Z − 12 = = = xe−2πinx dx − xe−2πinx 2πin 12 − Z x =− 21 e−2πinx − xe−2πinx − 2πin (2πin)2 1 2 − 12 e−2πinx dx −2πin 12 x =− 21 eπin e−πin eπin −e−πin − − + 4πin 4πin (2πin)2 (2πin)2 2i sin (πn) 2 cos (πn) + −4πin (2πin)2 (−1)n = + 0. −2πin = So, the Fourier series for f is (−1)n 2πinx ∑ −2πin e . n∈Z,n6=0 b. Parseval’s Identity tells us that k f k L2 = k fˆkl 2 . Thus, k f k2L2 = k fˆk2l 2 . Well, Z 1 (−1)n 2πinx 2 2 2 2 2 k f k L2 = x dx and k fˆkl 2 = ∑ −2πin e − 12 n∈Z,n6=0 = 1 12 = 1 4π 2 n2 n∈Z,n6=0 = 2 ∑ ∞ ∞ Thus, 1 ∑ n2 n =1 = π2 . 6 1 4π 2 n2 n =1 ∑ 6 7. Let f : R → R be a bounded Lebesgue measurable function, let K be a compact subset Z of R, and let g(t) = cos(tx ) f ( x ) dx, t ∈ R. K Prove or disprove: a. g has compact support. b. g is continuously differentiable. a. False. Counterexample: π − sin(πt) − sin(tx ) Let f ( x ) = χK , where K = [0, π ]. Then g(t) = = , cos(tx ) dx = t t 0 0 so, g(t) = 0 for t ∈ Z \ {0}. Thus, the support of g is R \ Z \ {0} , which is not compact. Z π b. True. First, let’s show that g is differentiable: 0 g (t) = = R R g(t + h) − g(t) K cos ( t + h ) x f ( x ) dx − K cos( tx ) f ( x ) dx lim = lim h h h →0 h →0 Z cos (t + h) x − cos(tx ) f ( x ) lim dx, since K is compact (bounded). h h →0 K Let {hn } be a sequence such that hn the Mean Value Theorem cos (t + h ) x − cos(tx ) n = hn ≤ → 0 as n → ∞. Now, fix t ∈ R and x ∈ K. Then, by for some c ∈ [t, t + hn ] x sin(cx ) x ≤ sup x = M, since K is compact. x ∈K cos (t + h ) x − cos(tx ) n Now, f ( x ) ≤ M f ( x ), and f ∈ L1 (K ), so, by the Lebesgue hn Dominated Convergence Theorem, we have Z cos (t + h) x − cos(tx ) f ( x ) 0 g ( x ) = lim dx h h →0 K Z cos (t + hn ) x − cos(tx ) f ( x ) = lim dx n→∞ K hn Z cos (t + hn ) x − cos(tx ) = lim f ( x ) dx hn K n→∞ = Z K − x sin(tx ) f ( x ) dx, which exists (and is finite) since K is compact. Thus, g0 ( x ) exists. 7 Now, to show that g0 ( x ) is continuous. Let {tn } be a sequence in R such that tn → t0 as n → ∞. Then, we want to show that 0 0 lim g (tn ) = g (t0 ). Well, − x sin(tn x ) f ( x ) ≤ x f ( x ) and x f ( x ) ∈ L1 (K ). So, by the n→∞ Lebesgue Dominated Convergence Theorem, we have lim n→∞ Z K − x sin(tn x ) f ( x ) dx = = Z lim − x sin(tn x ) f ( x ) dx K n→∞ Z K = Z K − x sin lim tn x f ( x ) dx, since sin(tx ) is continuous n→∞ − x sin(t0 x ) f ( x ) dx = g 0 ( t0 ). Thus, g0 (t) is continuous. So, g(t) is continuously differentiable, as required. 8