Solutions
Transcription
Solutions
Math 140B - Winter 2015 - Final Exam Problem 1. Consider the sequence fn : [0, 1] → R given by ( nx for 0 ≤ x ≤ n1 fn (x) = 1 for n1 < x ≤ 1. (i) Find the pointwise limit f of the sequence fn . (ii) Calculate the distance between fn and f in the supremum norm d(fn , f ) = sup |fn (x) − f (x)|. x∈[0,1] Deduce that the sequence fn does not converge uniformly to f . (iii) Show that the sequence fn converges in L2 to the function f . Solution: (i) We keep x fixed and make n → ∞. When x = 0, we have fn (0) = 0, hence the limit is f (0) = 0. For x 6= 0, we have n1 < x for n sufficiently large. Thus fn (x) = 1 for n large, hence the limit when n → ∞ is f (x) = 1. Thus, the pointwise limit is ( 0 if x = 0 f (x) = . 1 if 0 < x ≤ 1 (ii) We have 0 fn (x) − f (x) = nx − 1 0 if x = 0 if 0 < x ≤ n1 . if n1 < x ≤ 1 We have d(fn , f ) = sup |fn (x) − f (x)| = sup |nx − 1| = sup 1 − nx = 1. 1 x∈(0, n ] x∈[0,1] 1 x∈(0, n ] Since d(fn , f ) 6→ 0, it follows that fn does not converge uniformly to f . (iii) We show that the (square of the) L2 -distance between fn and f converges to zero: Z 1 |fn (x) − f (x)|2 dx → 0. 0 We compute Z 1 2 Z |fn (x) − f (x)| dx = 0 0 1 n 2 Z |nx − 1| dx = 0 1 n n2 x2 − 2nx + 1 dx = n2 x3 1 x= 1 − nx2 + x|x=0n = → 0. 3 3n Problem 2. Let 0 < a < b < 1 be fixed. Show that polynomials of the form P (x) = c2015 x2015 + c2016 x2016 + . . . + cn xn , n ≥ 2015 (not fixed), are uniformly dense in the space of real valued continuous functions over [a, b], but not dense in the set of continuous functions over [0, 1]. Solution:Let A denote the set polynomials of the form P (x) = c2015 x2015 + c2016 x2016 + . . . + cn xn , n ≥ 2015. Clearly, sums, products and scalar multiples of polynomials of this form are also of this form, so A is an algebra. Furthermore, A separates points and vanishes nowehere on [a, b]. Indeed, the polynomial Q(x) = ∈ A can be used to separate points over [a, b]: x2015 x1 6= x2 =⇒ x2015 6= x2015 . 1 2 By the same reasoning, the polynomial Q(x) = x2015 can be used to show that A vanishes nowehere: for x1 ∈ [a, b] =⇒ x2015 6= 0 (since x1 6= 0 as a > 0). 1 By Stone-Weierstraβ, A is uniformly dense in C[a, b]. On the other hand, all polynomials in A vanish at 0: P (0) = 0. If the algebra A were dense in C[0, 1] then for each function f ∈ C[0, 1], we could find a sequence of polynomials Pn ⇒ f , Pn ∈ A. However Pn (0) = 0, which would imply f (0) = 0. Nonetheless, there are continuous functions f which do not vanish at 0, for instance f (x) = x + 1. Contradiction! Problem 3. Consider the function f : [−π, π) → R given by f (x) = ex . n π −π e −e (i) Show that cn = (−1) 1−in · 2π (ii) Find the value of the sum . ∞ X n=1 n2 1 . +1 Solution: (i) We compute the Fourier coefficients Z π Z π 1 1 e(1−in)x x=π 1 e(1−in)π − e−(1−in)π 1 −inx f (x)e dx = e(1−in)x dx = |x=−π = . cn = 2π −π 2π −π 2π 1 − in 2π 1 − in Using that eiπ = −1, so that einπ = e−inπ = (−1)n , the above formula simplifies to (−1)n eπ − e−π . 2π 1 − in There is no need to do the calculation separately for n = 0 since the denominator of the above fraction never vanishes. (ii) We compute 1 eπ − e−π √ |cn | = . 2π n2 + 1 In particular |cn | = |c−n |, so the above gives ∞ ∞ ∞ X X (eπ − e−π )2 (eπ − e−π )2 X 1 + . |cn |2 = |c0 |2 + 2 |cn |2 = 2 2 2+1 4π 2π n n=−∞ cn = n=1 n=1 Next, we compute Z π Z π 1 1 1 e2π − e−2π f (x)2 dx = e2x dx = . 2π −π 2π −π 2π 2 Using Parseval’s theorem, we know Z π ∞ X 1 2 f (x) dx = |cn |2 . 2π −π n=−∞ Substituting, we find ∞ X 1 1 e2π − e−π π 1 eπ + e−π π = − + · = − + · . n2 + 1 2 (eπ − e−π )2 2 2 eπ − e−π 2 n=1 Problem 4. Let f : R → R be an integrable 2π-periodic function with zero integral. (i) Using Parseval’s theorem, show that the Fourier coefficients cn → 0. (ii) If f is differentiable with continuous derivative, show that ncn → 0. (iii) Show that Z Z π π f (x)2 dx ≤ f 0 (x)2 dx. −π −π Solution: (i) We have Z 1 2π ∞ X π 2 f (x) = −π |cn |2 . n=−∞ Since the series on the right converges, we conclude cn → 0 as n → ∞. (ii) We have seen in the homework that the Fourier coefficients of f 0 equal incn . Indeed, the coefficients of f 0 can be computed integrating by parts and using periodicity Z π Z π 1 1 0 0 −inx −inx π −inx cn = f (x)e dx = f (x)e |x=−π − f (x)(−in)e dx 2π −π 2π −π Z in π = f (x)e−inx dx = incn . 2π −π Since f 0 is continuous, hence integrable, it follows from part (i) applied to the function f 0 that ncn → 0. (iii) By Parseval, we have 1 2π Z ∞ X π f (x)2 dx = −π |cn |2 . n=−∞ On the other hand, also by Parseval, we have Z π ∞ ∞ X X 1 |incn |2 = n2 |cn |2 . f 0 (x)2 dx = 2π −π n=−∞ n=−∞ Furthermore, Z π 1 c0 = f (x) dx = 0 2π −π so c0 does not appear in either Fourier series. For n 6= 0, we have ∞ ∞ X X 2 2 2 2 |cn | ≤ n |cn | =⇒ |cn | ≤ n2 |cn |2 . n=−∞ Hence the above identities show that Z π Z 2 f (x) dx ≤ −π n=−∞ π −π f 0 (x)2 dx. Problem 5. Consider φ : R → R a nonnegative continuous function with φ(0) = 0. Let F be the family of functions f : [0, 1] → R such that f (0) = 0 and for all x, y ∈ [0, 1] we have |f (x) − f (y)| ≤ φ(x − y). (i) Show that F is uniformly bounded. (ii) Show that F is equicontinuous. (iii) Show that any sequence of functions in F has a subsequence converging to a continuous function. Solution: (i) We set y = 0. This yields that for all f ∈ F we have |f (x)| ≤ φ(x). Since φ is continuous, it is bounded by some constant M over the interval [0, 1]. Thus |f (x)| ≤ φ(x) ≤ M shows that the family F is uniformly bounded. (ii) Fix > 0. Since φ is continuous and φ(0) = 0, it follows that there exists δ > 0 such that if |t| < δ we have φ(t) < . For t = x − y, we conclude that if |x − y| < δ we have |f (x) − f (y)| ≤ φ(x − y) < , for all f ∈ F. This is exactly the definition of equicontinuity. (iii) By Arzel` a-Ascoli, any sequence in F has a uniformly convergent subsequence. The limit function is continuous since it is uniform limit of continuous functions (members of F are continuous; in fact, even stronger, F is equicontinuous). Problem 6. Assume that f : (0, 1) → R is a function with bounded derivative |f 0 (x)| ≤ A for all x ∈ (0, 1). Show that f is bounded. (i) Prove that 1 1 + f . 2 2 (ii) Conclude from (i) that f is bounded over (0, 1). (iii) Take f (x) = sin x1 . Show f is bounded, but f 0 is unbounded over the interval (0, 1). Therefore, the converse of (ii) does not hold. |f (x)| ≤ A x − Solution: (i) Using the mean value theorem, we conclude that for all x, there exists c between 12 and x such that 1 1 1 1 0 1 0 f (x) − f = x− f (c) =⇒ f (x) − f = x − |f (c)| ≤ A x − . 2 2 2 2 2 Using the triangle inequality, we find 1 1 1 + f ≤ A x − + f 1 . |f (x)| ≤ f (x) − f 2 2 2 2 (ii) We have x − 12 ≤ 12 for all x ∈ (0, 1) so that above becomes A 1 |f (x)| ≤ + f , 2 2 showing that f is bounded by M = A2 + f 12 . (iii) We have |f (x)| ≤ 1, so f is bounded. On the other hand, by the chain rule 1 1 f 0 (x) = − 2 cos . x x If xn = 1 2nπ ∈ (0, 1) for n ∈ Z>0 , then f 0 (xn ) = −(2nπ)2 cos(2nπ) = −(2nπ)2 → −∞ showing that f 0 is unbounded. Problem 7. Assume that f : [0, 1] → R is a bounded function such that f is integrable over any interval [c, 1] for c > 0. Using Riemann’s criterion, show that f is integrable over [0, 1]. Solution:Since f is bounded, there exists a constant M > 0 such that |f (x)| ≤ M . We fix > 0. We exhibit a partition P of [0, 1] such that U (P, f ) − L(P, f ) < . Let c = 4M . Since f is integrable over [c, 1], by Riemann’s criterion, there exists a partition Q = {c = x0 < x1 < x2 < . . . < xn = 1} of [c, 1] such that U (Q, f ) − L(Q, f ) < . 2 Let P be the partition P = {0} ∪ Q = {0 < c = x0 < x1 < . . . < xn = 1}. Then U (P, f ) − L(P, f ) = (sup f − inf f ) · c + (U (Q, f ) − L(Q, f )) ≤ 2M c + (U (Q, f ) − L(Q, f )) [0,c] [0,c] + = . 4M 2 By Riemann’s criterion, it follows that f is integrable. < 2M · Problem 8. Consider the series f (x) = ∞ X n=1 x . 1 + n2 x2 (i) Show that the series converges uniformly over (−∞, −a) ∪ (a, ∞) for any a > 0. (ii) Conclude that the series converges pointwise on R, and the function f is continuous over R \ {0}. (iii) Show that f is not continuous at x = 0, and conclude that the series does not converge uniformly on R. Solution: (i) We use the Weierstraβ M -test. For x ∈ (−∞, −a) ∪ (a, ∞), we have 1 1 |x| x 1 + n2 x2 ≤ n2 x2 = n2 |x| < n2 a . P∞ 1 P∞ 1 x 1 P∞ Since 1 n2 a = a n=1 n2 converges, it follows that n=1 1+n2 x2 converges uniformly over (−∞, −a) ∪ (a, ∞). (ii) Clearly, f (0) = 0, so we do get pointwise convergence at x = 0. For x 6= 0, there exists a > 0, for instance a = |x| 2 , such that |x| > a. Thus x ∈ (−∞, −a) ∪ (a, ∞), and over this interval we proved uniform, hence pointwise convergence. In fact, the limit function f is continuous over (−∞, −a) ∪ (a, ∞) for any a > 0, being uniform limit of continuous functions (the partial sums are clearly continuous). Since any x ∈ R \ 0 belongs to an interval (−∞, −a) ∪ (a, ∞) for a > 0 (take a = |x| 2 ), this shows that f is in fact continuous at all points x ∈ R \ {0}. (iii) Assume that f is continuous at 0. Now, f (0) = 0 by direct substitution. Then 1 f → f (0) = 0, k → ∞. k We obtain a contradiction once we show that 1 1 f ≥ for all k. k 2 To see the above inequality, we estimate X k k k k 1 X X X 1 k k 1 k 1 k f ≥ ≥ = = = . 2 = 2 2 2 2 n k k +n k +k 2k 2k 2 1+ 2 n=1 k n=1 n=1 n=1 Thus f is discontinuous at 0. Since uniform limit of continuous functions is continuous and as f is discontinuous at 0, it follows that the series does not converge uniformly on R.