MA244 Analysis III Solutions. Sheet 3. Questions for credit: 6 (4 points), 8 (5 points), 12 (8 points) and 15 (8 points) 0.1 Further exercises in integration R∞ R∞ R∞ R1 2 2 2 2 2 e−x dx = 0 e−x dx + 1 e−x dx. Therefore, 0 e−x dx and 1 e−x dx are either both convergent or both divergent. The exponential function is 2 2 increasing x ≥ 1. Therefore, e−x ≤ e−x for x ≥ 1. R ∞ −xand −x ≤ −x R R for But 1 e := limR→∞ 1 e−x dx = limR→∞ (e−1 − e−R ) = e−1 , convergent. Therefore, the probability integral converges by the comparison principle. R 1/2 R1 2. B(p, q) = 0 xp−1 (1 − x)q−1 dx + 1/2 xp−1 (1 − x)q−1 dx. The Euler integral converges iff each of the integrals in the right hand side converges. For R 1/2 x ∈ (0, 1/2], xp−1 (1 − x)q−1 ≤ xp−1 max(1, (1/2)q−1 ). The integral 0 xp−1 dx R 1/2 converges for p > 0 (lecture material). Therefore, 0 xp−1 (1 − x)q−1 dx converges by the comparison principle. Similarly, for x ∈ [1/2, 1), xp−1 (1−x)q−1 ≤ R1 R 1/2 (1 − x)q−1 max(1, (1/2)p−1 ). The integral 1/2 (1 − x)q−1 dx = 0 xq−1 dx conR1 verges for q > 0 (lecture material). Therefore, 1/2 xp−1 (1 − x)q−1 dx converges by the comparison principle. We conclude that the Euler integral of the first kind converges for p > 0, q > 0. R∞ R1 R∞ 3. 0 xp−1 e−x dx = 0 xp−1 e−x dx + 1 xp−1 e−x dx. The Euler integral converges iff each integral in Rthe right hand side converges. For x ∈ (0, 1], xp−1 e−x ≤ 1 xp−1 . The integral 0 xp−1 dx converges for p > 1 (lecture material). Therefore, R 1 p−1 −x x e dx converges by the comparison principle. For x ∈ [1, ∞), xp−1 e−x ≤ 0 x|p−1| e−x = (x|p−1| e−x/2 )e−x/2 ≤ supx∈[0,∞) (x|p−1| e−x/2 )e−x/2 ≤ Cp e−x/2 , where Cp > 0 is a p-dependent constant. (The global maximum of g(x) = x|p−1| e−x/2 is achieved at xc = 2|p − 1|. The valueR at the global maximum is Cp = ∞ −x/2 −|p−1| (2|p − 1|)|p−1| dx converges (question 1). R ∞e p−1 −x> 0.) The integral 1 e Therefore, 1 x e dx converges by the comparison principle. We conclude that the Euler integral of the second kind converges for p > 0. 1. R∞ 0 4. Let {p0 = a < p1 < . . . < pN = b} be a finite set of points such that F is continuously differentiable on each of [pk−1 , pk ], k = 1, 2, . . . N . Such a set exists the definition of piece-wise continuous differentiability. By FTC2, R pk by 0 F = F (pk ) − F (pk−1 ). Therefore, pk−1 Z b 0 F = a N Z X k=1 pk pk−1 0 F = N X (F (pk ) − F (pk−1 )) = F (b) − F (a). k=1 Notice the crucial role the continuity of F plays in the proof. 5. Let k ∈ Z. sign(sin(x)) = 1 for x ∈ (2kπ, (2k + 1)π) and sign(sin(x)) = −1 for x ∈ ((2k + 1)π, (2k + 2)π). Let F : F (x) = x − 2kπ for x ∈ [2kπ, (2k + 1)π], F (x) = π−(x−(2k+1)π) for x ∈ [(2k+1)π, (2k+2)π]. (A ’saw tooth’ function.) Notice that F is continuous, piece-wise continuously differentiable and F 0 (x) = R πn sign(sin(x)) for x 6= πk. By the theorem of Question 4, 0 sign(sin(x))dx = R πn 0 F = F (πn) − F (0) = F (πn), which is zero for even n and π for odd n. 0 Ra R0 Ra R0 Ra t=−x 6. In each case −a f (x)dx = −a f (x)dx+ 0 f (x)dx = − a f (−t)dt+ 0 f (x)dx = Ra (f (x) + f (−x))dx. Therefore: 0 Ra Ra (a) If f (x) = f (−x), −a f = 2 0 f . Ra Ra (b) If f (x) = −f (−x), −a f = 0 0 · dx = 0. Both statements are true. R π/2 π/2 R π/2 7. In = 0 sin0 (x) cosn−1 (x)dx = sin(x) cosn−1 (x)|x=0 − 0 sin(x)(cosn−1 )0 (x)dx = R π/2 R π/2 (n − 1) 0 sin2 (x) cosn−2 (x)dx = (n − 1) 0 (1 − cos2 (x)) cosn−2 (x)dx = (n − 1)(In−2 − In ). So, n−1 In−2 . In = n By explicit calculation, I0 = π/2, I1 = 1. If n is even, n−1 (n − 1)(n − 3) (n − 1)(n − 3) . . . 3 · 1 In−2 = In−4 = . . . = I0 n n(n − 2) n(n − 2) . . . 4 · 2 (n − 1)(n − 3) . . . 3 · 1 π . = n(n − 2) . . . 4 · 2 2 In = If n is odd, n−1 (n − 1)(n − 3) (n − 1)(n − 3) . . . 2 · 4 In−2 = In−4 = . . . = I1 n n(n − 2) n(n − 2) . . . 3 · 1 (n − 1)(n − 3) . . . 4 · 2 = n(n − 2) . . . 3 · 1 In = So, 9·7·5·3 π 63π 63π · = 9 = , 10 · 8 · 6 · 4 · 2 2 2 512 8·6·4·2 27 128 I9 = = = . 9·7·5·3 315 315 I10 = 8. The theoretical underpinning for this question is Theorem 22 of the course. Let xk = k/n, k = 1, . . . , n − 1. (a) n−1 n−1 X 1X k lim = lim xk n→∞ n2 n→∞ n k=1 k=1 Z T hm.22 = 0 1 1 xdx = . 2 (b) lim n→∞ n X k=1 n 1 1X 1 = lim n + k n→∞ n k=1 1 + xk T hm.22 Z 1 = 0 1 dx = log(1+x)|10 = log(2). 1+x (c) n n X kp 1X p xk lim = lim p+1 n→∞ n→∞ n n k=1 k=1 T hm.22 Z = 0 1 xp dx = 1 . p+1 9. The integral is positive: f is continuous on [0, 2π], so the integral exists in the usual sense. Z 2π Z π Z 2π Z π f= f+ f= (f (x) + f (x + π))dx, 0 0 π 0 f (x) + f (x + π) = sin(x)/x + sin(x + π)/(xR + π) = sin(x)(1/x − 1/(x + π)) = π > 0 for x ∈ (0, π). Therefore, f > 0 by integral bounds. sin(x) x(x+π) R1 √ R1√ R √ 2 dx − 1 1 · dx = 2 − ( 1 + x 1 + x 1 + 0)dx > 0, as the inte0 0 0 grand is continuous, non-negative, and positive R1√ R 1 at at least one point of 2 the integration interval. So, 0 1 + x dx > 0 1 · dx R1 R1 R1 (b) 0 x2 sin2 (x)dx − 0 x sin2 (x)dx = 0 x(x − 1) sin2 (x)dx < 0, as the integrand is continuous, non-positive, negative R 1 2 and R 1 at at2 least one point of 2 the integration interval. So, 0 x sin (x)dx < 0 x sin (x)dx R2 R2 R2 2 2 (c) 1 ex dx− 1 ex dx = 1 ex (ex −x −1)dx > 0, as the integrand is continuous, 2 non-negative (x2 − x ≥ 0 for x ≥ 1, so ex −x at at least R 2 ≥x21), andR 2positive x one point of the integration interval. So, 1 e dx > 1 e dx 10. (a) 0.2 Uniform convergence. 11. The integrand is continuous (hence uniformly continuous) on D = [0, 10] × [0, 10]. More over, the partial derivative of the integrand with respect to x is also continuous on D. Therefore, the integral is differentiable with Rrespect to 2 b the limits of integration (FTC1) and the c-derivative of the integral a e−cy dy is equal to the integral of the c-derivative of the integrand. Therefore, applying the chain rule formula from the hint we find Z 10 2 0 −x3 f (x) = −e − y 2 e−xy dy x R ∞ −αx −βx R ∞ −αx 0 12. Let F (α) = 0 e −e dx. Then F (α) = −e = − α1 . Therefore, x 0 F (α) = − log(α) + C(β), where C(β is a constant of integration. It can be found by noticing that F (β) = 0, which gives C(β) = log(β). Therefore, F (α) = log(β/α). R∞ 13. (a) F (p) = 0 e−pt dt = p1 R∞ 1 (b) F (p) = 0 e−pt e−αt dt = p+α R∞ R∞ R∞ 1 (c) F (p) = 0 e−pt cos(βt)dt = 2i1 ( 0 e−pt eiβt dt + 0 e−pt e−iβt dt) = 21 ( p−iβ + p 1 ) = p2 +β 2 p+iβ P P∞ xn 1 14. (a) For any x ∈ [−1, 1], |xn /n2 | ≤ 1/n2 . As ∞ n=1 n2 < ∞, n=1 n2 converges uniformly on [−1, 1] by the M-test. P P∞ sin(nx) 1 (b) For any x ∈ R, sin(nx)/2n ≤ 1/2n . But ∞ n=1 2n < ∞. So, n=1 2n converges on R uniformly by the M-test. 15. In each of the three cases below, both the power series S(x) and the series for S 0 (x) converge uniformly on [a, b]. Therefore, the term-wise differentiation is justified. Integrating S 0 to obtain S is done using F T C2. P∞ k−1 P 1 xk 0 = 1−x . Therefore, S(x) = (a) Let S(x) = ∞ k=1 x k=1 k . Then S (x) = −log(1 − x) + C, where C is an integration constant. As S(0) = 0, C = 0. P xk Therefore, ∞ k=1 k = −log(1 − x). P P 1 k−1 xk k−1 (b) Let S(x) = ∞ . Then S 0 (x) = ∞ = 1+x . Therek=1 (−1) k=1 (−x) k fore, S(x) = log(1 + x) + C, where C is an integration constant. As P xk S(0) = 0, C = 0. Therefore, ∞ k=1 k = log(1 + x). P P∞ 2k−2 x2k−1 1 0 (c) Let S(x) = ∞ = 1−x 2 . Therefore, k=1 2k−1 . Then S (x) = k=1 x 1+x 1 S(x) = 2 log 1−x + C, where C is an integration constant. As S(0) = 0, P xk 1 1+x C = 0. Therefore, ∞ k=1 k = − 2 log 1−x . November the 3rd, 2014 Daniel Ueltschi and Oleg Zaboronski.