221 Analysis 2, 2008–09 Suggested solutions to exercise sheet 5 In these questions, the words “measurable” and “integrable” should be taken to mean “Lebesgue measurable” and “Lebesgue integrable”. 1. Either give an example, or explain why no such example exists, of: (a) a (Lebesgue) measurable function f : R → R which is not integrable; (b) an integrable function f : R → R such that −f is not integrable; (c) an integrable function f : R → R such that f 2 is not integrable; (d) a convergent sequence of integrable functions fn : R → R such that lim fn is not integrable. n→∞ Solution. (a) The constant function f (x) = 1 is a measurable function R(since it’s equal R to χR , and χA is measurable iff A is a measurable set). But |f (x)| dx = χR dm = m(R) = ∞, so f is not integrable. (b) No such example Rexists. If f is integrable then f is measurable, so −f R is R measurable. Also, |f (x)| dx < ∞ and | − f | = |f |, so |(−f )(x)| dx = |f (x)| dx < ∞. So −f is integrable. (c) Let f (x) = x−1/2 if x ∈ (0, 1), and f (x) = 0 if x 6∈ (0, 1). Then f is measurable [it can be written as a limit of continuous functions] and f χ(1/n,1) ↑ f as n → ∞, so Z Z 1 M CT FTC x−1/2 dx = lim 2 − 2 · ( n1 )1/2 = 2 < ∞. |f (x)| dx = lim n→∞ n→∞ 1/n So f is integrable. However, f 2 is not integrable since f (x) = x−1 for x ∈ (0, 1) and f 2 χ(1/n,1) ↑ f 2 , as n → ∞, so Z 2 M CT |f (x)| dx = Z 1 lim n→∞ 1/n FTC x−1 dx = lim 0 − log( n1 ) = lim log n = ∞. n→∞ n→∞ (d) Let fn = χ[−n,n]R and let f (x) = 1. Then fn → f as n → ∞ and each fn is integrable [with fn dm = 2n < ∞], but f is not integrable. 2. Evaluate the following expressions, carefully justifying your working. Z ∞ Z 1 Z n 1 1 + nx2 sin(n2 x2 ) √ (a) dx (b) lim dx (c) lim dx n→∞ 0 (1 + x)n n→∞ −n n2 |x|3/2 1 + x2 −∞ 1 1 1 −1 Solution. (a) If x > 1 then 1 + x2 < 4x2 , so √1+x . So for any 2 > 2x n ≥ 1, Z n Z ∞ Z 1 1 1 n −1 FTC 1 √ √ dx ≥ dx ≥ x dx = log n. 2 1 2 1 + x2 1 + x2 −∞ 1 R∞ 1 Since 21 log n → ∞ as n → ∞, this gives −∞ √1+x 2 dx = ∞. [Alternatively,R you can do this by substituting u = sinh−1 (x) on [−n, n] and n −1 1 (n) → ∞ as n → ∞.] showing that −n √1+x 2 dx = 2 sinh 2 1+nx (b) Let fn (x) = (1+x) n χ(0,1) (x). We apply the Dominated Convergence Theorem to the sequence (fn )n≥1 : • For any x ∈ (0, 1) we have fn (x) = 1 + nx2 1+n < →0 n (1 + x) (1 + x)n as n → ∞, so fn → 0 as n → ∞. • If x ∈ (0, 1) and n ≥ 1 then x2 < x, so 1 + nx2 < 1 + nx and 1 + nx2 1 + nx2 1 + nx2 = < 1. ≤ (1 + x)n 1 + nx 1 + nx + n2 x2 + . . . So fn = |fn | ≤ χ(0,1) , which is integrable and independent of n. So χ(0,1) will do as our dominating function. By the Dominated Convergence Theorem, Z n Z Z Z 1 + nx2 DCT dm = lim fn dm = lim fn dm = 0 dm = 0. lim n→∞ n→∞ n→∞ −n (1 + x)n 2 2 x ) (c) Let fn (x) = sin(n if x ∈ (−n, n)\{0}, and fn (x) = 0 for all other x ∈ R. n2 |x|3/2 We again apply the Dominated Convergence Theorem: • For any x ∈ (−n, n) with x 6= 0, using the estimate | sin θ| < |θ| we have |fn (x)| ≤ 1 n2 |x|3/2 →0 as n → ∞. So fn → 0 as n → ∞. • We have |sin θ| ≤ |θ| for any θ ∈ R, so if x 6= 0 and n ≥ 1 then sin(n2 x2 ) n 2 x2 2 3/2 ≤ 2 3/2 = |x|1/2 . n |x| n |x| Observe that x 7→ |x|1/2 is integrable near 0, but not near ∞. So we need another estimate to take care of the growth for large x: | sin θ| ≤ 1 for any θ ∈ R, so sin(n2 x2 ) 1 2 3/2 ≤ 2 3/2 ≤ |x|−3/2 . n |x| n |x| Now |fn (x)| ≤ |x|1/2 χ(−1,1) (x) + |x|−3/2 χ(−∞,−1]∪[1,∞) (x), which is integrable and independent of n. 2 By the Dominated Convergence Theorem, Z n Z Z Z sin(n2 x2 ) DCT lim dx = lim fn dm = lim fn dm = 0 dm = 0. n→∞ −n n2 |x|3/2 n→∞ n→∞ 3. (a) Let a, b ∈ R with a < b and let f, g : [a, b] → R be continuously differentiable. Use the Fundamental Theorem of Calculus to show that Z b Z b 0 f 0 (x)g(x) dx. f (x)g (x) dx = f (b)g(b) − f (a)g(a) − a a [Hint: what is Rb a (f g)0 (x) dx?] (b) Show that if f, g : R → R are continuously differentiable functions and f (x)g(x) → L ∈ R as x → ∞, and f g 0 χ[0,∞) and f 0 gχ[0,∞) are both integrable, then Z ∞ Z ∞ 0 f (x)g (x) dx = L − f (0)g(0) − f 0 (x)g(x) dx. 0 0 R∞ (c) Fix a number t ∈ R. Show that 0 e−x cos(xt) dx = (1+t2 )−1 , justifying your working. [Hint: call the value of this integral J. Integrate by parts twice (justify using (b)) to find an equation that J satisfies, and then solve it.] R∞ (d) For t ∈ R, let I(t) = 0 x−1 e−x sin(xt) dx. Show that I 0 (t) = (1 + t2 )−1 , and solve this differential equation to find I(t). Solution. (a) Since f and g are continuously differentiable, so is f g. By the Fundamental Theorem of Calculus, Z b (f g)0 (x) dx = (f g)(b) − (f g)(a). a Moreover, (f g)0 = f g 0 + f 0 g by the product rule, and f g 0 and f 0 g are continuous, so are integrable on [a, b]. So Z b 0 Z (f g) (x) dx = a b Z 0 f (x)g (x) dx + a b f 0 (x)g(x) dx. a Equating these two expressions and rearranging gives the result. (b) If f : R → R and f χ[0,∞) is integrable then by Exercise 4.2, Z ∞ Z ∞ Z n Z f (x) dx = f (x)χ[0,∞) (x) dx = lim f (x)χ[0,∞) (x) dx = lim n→∞ −∞ 0 n→∞ −n For every n ≥ 1, by part (a) we have Z n Z 0 f (x)g (x) dx = f (n)g(n) − f (0)g(0) − 0 0 3 n f 0 (x)g(x) dx. 0 n f (x) dx. Since f g 0 and f 0 g are integrable, taking the limit as n → ∞ gives Z n Z ∞ 0 f 0 (x)g(x) dx f (x)g (x) dx = lim f (n)g(n) − f (0)g(0) − lim n→∞ n→∞ 0 0 Z ∞ f 0 (x)g(x) dx. = L − f (0)g(0) − 0 R∞ (c) Let J = 0 e−x cos(xt) dx. First take f (x) = e−x and g(x) = 1t sin(xt): these are continuously differentiable, and |f g 0 | and |f 0 g| are both dominated by e−x which is integrable on [0, ∞), so f g 0 χ[0,∞) and f 0 gχ[0,∞) are integrable; so L = limx→∞ (e−x · 1t sin(xt)) = 0 = f (0)g(0) and g 0 (x) = cos(xt), so by (b) Z ∞ Z ∞ Z ∞ −x 1 −1 −x e−x sin(xt) dx. (−e )· t sin(xt) dx = t e cos(xt) dx = − J= 0 0 0 Now take g(x) = − 1t cos(xt). Then g 0 (x) = sin(xt) and all of the integrability and differentiability conditions of (b) are satisfied again. This time, L = 0 and f (0)g(0) = − 1t , so by (b) we have Z ∞ Z −x 1 e sin(xt) dx = 0 − (− t ) − (−e−x ) · (− 1t cos(xt)) dx 0 Z ∞ −1 =t e−x cos(xt) dx = t−1 (1 − J). 1− 0 So J = t−2 (1 − J), which gives J = t−2 1+t−2 = 1 . 1+t2 (d) We check that to find I 0 (t), we can interchange the order of differentiation −x and integration. Let f (x, t) = e sin(xt) χ(0,∞) (x). We have: x • For all t ∈ R, f (x, t) is an integrable function of x. To see this, observe that |e−x sin(xt) | ≤ |e−x xt | = |t|e−x , which is integrable. x x • For x ∈ R we have ∂ f (x, t) ∂t −x • The function g(x) = e = e−x x · cos(xt) · x = e−x cos(xt); is integrable and dominates ∂ f (x, t). ∂t So the hypotheses of the theorem justifying the interchange of integration and differentiation are satisfied, and by part (c), Z Z ∞ Z ∞ d ∞ ∂ 1 (c) 0 I (t) = f (x, t) dx = f (x, t) dx = e−x cos(xt) dt = . dt 0 ∂t 1 + t2 0 0 1 So I(t) is an antiderivative of 1+t 2 , so I(t) = arctan(t) + C for some conR∞ stant C. Since I(0) = 0 0 dx = 0, C = 0 so I(t) = arctan(t). 4. Let (an )n≥1 be a sequence in [−∞, ∞]. Prove the following assertions. (a) lim sup(−an ) = − lim inf an n→∞ n→∞ (b) an converges as n → ∞ if and only if lim inf an = lim sup an . n→∞ n→∞ (c) If an → a as n → ∞ then a = lim inf an = lim sup an . n→∞ 4 n→∞ Solution. (a) If S ⊆ [−∞, ∞] and −S = {−x : x ∈ S} then sup(−S) = − inf S. Applying this to the sets S = {ak : k ≥ n} gives lim sup(−an ) = lim sup(−ak ) = lim −inf ak = − lim inf ak = − lim inf an . n→∞ n→∞ n→∞ k≥n n→∞ k≥n k≥n n→∞ (b) and (c) Suppose that an → a as n → ∞. • If a ∈ R and ε > 0 then there is n0 ≥ 1 with |ak − a| ≤ ε for all k ≥ n0 . So for n ≥ n0 we have |a − inf ak | ≤ ε and |a − sup ak | ≤ ε. Hence k≥n k≥n inf ak → a and sup ak → a as n → ∞, so a = lim inf an = lim sup an . k≥n n→∞ k≥n n→∞ • If a = ∞ and R ∈ R then there is n0 ≥ 1 with ak ≥ R for all k ≥ n0 . So for n ≥ n0 we have inf ak ≥ R and sup ak ≥ R. Hence inf ak → ∞ k≥n k≥n k≥n and sup ak → ∞ as n → ∞, so lim inf an = lim sup an = ∞. n→∞ k≥n n→∞ • If a = −∞ then −an → ∞ so lim inf (−an ) = lim sup(−an ) = ∞ which, n→∞ n→∞ by (a), gives lim inf an = lim sup an = −∞. n→∞ n→∞ Conversely, suppose that lim inf an = lim sup an and let a be this extended n→∞ n→∞ real number. • If a ∈ R and ε > 0 then there is n0 ≥ 1 so that |a − inf ak | ≤ ε and k≥n |a − sup ak | ≤ ε for all n ≥ n0 . So for all n ≥ n0 we have k≥n a − ε ≤ inf ak ≤ an ≤ sup ak ≤ a + ε, k≥n0 k≥n0 so |a − an | ≤ ε for all n ≥ n0 . So an → a as n → ∞. • If a = ∞ and R ∈ R then there is n0 ≥ 1 so that inf ak ≥ R for all k≥n n ≥ n0 . So for all n ≥ n0 we have R ≤ inf k≥n0 ak ≤ an , so an → a as n → ∞. • If a = −∞ and R ∈ R then there is n0 ≥ 1 so that sup ak ≤ R for all k≥n n ≥ n0 . So for all n ≥ n0 we have an ≤ supk≥n0 ak ≤ R, so an → −∞ as n → ∞. 5