Real & Complex Analysis Qualifying Exam Solution, Spring 2011 Shiu-Tang Li January 1, 2013 A-1 P (a)Let A be a collection of sequences: {(a1 , a2 , · · · )|ai ∈ Q, ∞ i=1 |ai | < 1 ∞)}. For each (b1 , b2 , · · · ) ∈ l , we can find (a1 , a2 , · · · ) ∈ A s.t. |ai − bi | < 2i when arbitrary is given. (b)The techniques in (a) may be applied to this problem, too. A-2 (a)Let g(x) : x 7→ |x|, R → R. Since g is a continuous function, g −1 maps every B ∈ B(R) to B(R). Therefore, |f |−1 = f −1 ◦ g −1 maps every B ∈ B(R) to a measurable set in X. (b)Let A be a set in X which is not measurable. It’s obvious that f (x) = 1A (x) − 1{X\A} (x) is non-measurable while |f | ≡ 1 is a measurable function. 1 A-3 Assume that at θ = 0, every continuous function f satisfies f (1) = lim N →∞ N X cn n=−N Z 2π N X 1 = lim f (eiθ )e−inθ dθ N →∞ 2π 0 n=−N Z 2π −iN θ 1 (1 − ei(2N +1)θ ) iθ e = lim f (e ) dθ N →∞ 2π 0 1 − eiθ Z 2π 1 e−i(N +1/2)θ − ei(N +1/2)θ = lim f (eiθ ) dθ N →∞ 2π 0 e−iθ/2 − eiθ/2 Z 2π sin((N + 1/2)θ) 1 dθ f (eiθ ) = lim N →∞ 2π 0 sin(θ/2) R 2π Define wn := C[eiθ : 0 ≤ θ < 2π] → R by wn (f (eiθ )) = 0 f (eiθ ) sin((n+1/2)θ) dθ. sin(θ/2) iθ For each g ∈ C[e : 0 ≤ θ < 2π], |wn (g)| ≤ Mg for some Mg > 0 (Since wn (g) → 2πg(1)). Therefore, kwn k ≤ M by uniform boundedness principle, for C[eiθ : 0 ≤ θ < 2π] endowed with sup-norm is a Banach space. By defiR 2π nition, 0 | sin((n+1/2)θ) | dθ ≤ kwn k ≤ M , for every n ∈ N. sin(θ/2) However, this is wrong. Note that Z π Z 2π sin((n + 1/2)θ) sin((n + 1/2)θ) | dθ ≥ | | dθ | sin(θ/2) sin(θ/2) 0 0 Z π sin((n + 1/2)θ) ≥ | | dθ θ/2 0 Z (n+1/2)π sin(z) | dz =2 | z 0 n X 1 ≥2 > M when n is large enough, jπ j=1 which is a contradiction. A-4 Assume that lim|x|→∞ f (x) 6= 0. Therefore, there is a sequence {xn } s.t. |xn | → ∞ and |f (xn )| > a > 0 for all n large enough. Since f is uniformly 2 continuous, there exists a δ so that for |x − y| < δ, |f (x) − f (y)| < a/2. As a result, for all n large enough, |f (y)| > a/2 for y ∈ Bδ (xn ). We may pick a further subsequence {xnP } of {xn } so that these balls are mutually disjoint. k R a 2 As a result, |f | dx ≥ ∞ n=1 2 · δ π = ∞, which contradicts the fact that f ∈ L1 (R). A-5 Apply Fatou’s lemma to f + fn − |f − fn |, we have Z Z Z Z f dµ − lim sup |f − fn | dµ ≥ lim inf(f + fn − |f − fn |) dµ = 2 f dµ. 2 X X Therefore, lim sup R X X X |f − fn | dµ ≤ 0. B-6 For each z0 on U, there is some r > 0 so that f (z) = for all |z − z0 | < r. Taking conjugates, we have f (z) = ∞ X an (z − z0 )n P∞ n=0 an (z − z0 )n for all |z − z0 | < r n=0 ⇒g(z) = f (z) = ∞ X an (z − z 0 )n for all |z − z 0 | = |z − z0 | < r. n=0 That is, g(z) has power point representation at z 0 . This shows g(z) is holomorphic on U . B-7 We first claim that f has only finitely many poles on {z : |z| ≤ R}, for otherwise there would be a cluster point of these poles, and it would be a contradiction to the definition of meromorphic functions. (1/z) 1 Consider g(z) = f(1/z) Since the only k on B0 (1/R) := {z : |z| < R }. isolated singularity is 0, and |g(z)| ≤ C on B0 (1/R), 0 is a removable singularity of g. Redefine g and we may write f (1/z) = (1/z)k g(z) on a deleted neighborhood of 0, where g is holomorphic on B0 (1/R). ThereP n fore, f (1/z) = ∞ b n=−k n z on a deleted neighborhood of 0, and we define 3 h(z) = Pk n=1 b−n z k for z ∈ C. Now we let all the poles ofP f be {z1 , · · · , zm }. For each zj , we have Laun rent series expansion f (z) = ∞ n=−N (j) aj,n (z − zj ) on a deleted neighborP n hood of zj , and we define fj (z) = −1 n=−N (j) aj,n (z −zj ) for every z ∈ C\{0}. P We claim that F (z) = f (z) − m j=1 fj (z) − h(z) is a bounded entire function. To see F is entire, we note P that F is holomorphic on each zj , since both f (z) − fj (z) and fj (z) − m j=1 fj (z) − h(z) are holomorphic on zj , and this is true for every 1 ≤ j ≤ m. To see F is bounded, we only need to show that f (z) − h(z) is bounded for all |z| large, which is equivalent to show that f (1/z) − h(1/z) is P bounded for all |z| small. By the definition of h, we have n f (1/z)−h(1/z) = ∞ n=0 bn z , which is obviously bounded on a neighborhood of 0. As a result, by PLouville’s theorem, F (z) is a constant a on C. This shows that f (z) = a + m j=1 fj (z) + h(z), where each fj is a rational function and h(z) is a polynomial, and therefore f (z) is also a rational function. B-8 (i) ⇒ (ii): By definition, we have f (z) = (z − a)n g(z) on Ba (), where n ≥ k and g(z) is holomorphic on Ba (), for some > 0. Take C = supz∈Ba () |(z − a)n−k g(z)| = supz∈∂Ba () n−k |g(z)|. f (z) (ii) ⇒ (i): Let g(z) = (z−a) k , we have |g(z)| ≤ C on Ba (), and g(z) is holomorphic on Ba () \ {a}. Riemann’s theorem on removable singularities imply that 0 is a removable singularity of g, and we may redefine g to be holomorphic on Ba (). We may write f (z) = g(z)(z − a)k ; that is, 0 is a zero of f of order ≥ k. B-9 It is obvious that f (z) is differentiable on C \ {0}, and f (0) is not defined. Thus 0 is the only singularity. P∞ P∞ Notice that if both n=1 an and m=1 bm converge absolutely, then by P Tonelli’s theorem n,m |an bm | < ∞. Therefore, we may sum the double array {an bm }n,m in any order. 4 3 5 sin(z) = z − z3! + z5! − · · · for all z ∈ C, and sin( z1 ) = z1 − z313! + z515! − · · · for all z ∈ C \ {0}. Since for every z ∈ C \ {0}, both series are absolutely convergent, we have ∞ ∞ 2n−1 X X 1 1 n+1 z sin(z) sin( ) = (−1) (−1)m+1 2m−1 z (2n − 1)! m=1 z (2m − 1)! n=1 = + ∞ X ∞ X k=0 n=1 ∞ X ∞ X (−1)n+k+1 (−1) k=0 n=1 n+1 z 2n+2k−1 1 (−1)n+1 2n−1 (2n − 1)! z (2n − 1)! 1 z 2n−1 (−1)n+k+1 2n+2k−1 (2n − 1)! z (2n + 2k − 1)! and we find that the coefficient of 1/z in the above Laurent series is 0, so the residue of z = 0 is 0. B-10 RR cos(x) dx+ Let CR = Reiθ , θ goes from 0 to π. By the residue theorem, −R x2x+2x+10 R R x sin(x) R (−1+3i)e−3−i zeiz i −R x2 +2x+10 dx + CR z2 +2z+10 dz = 2πiRes(f ; −1 + 3i) = 2πi · = 6i e−3 π (− cos(1) + 3 sin(1) + 3i cos(1) + i sin(1)). Since 3 Z | CR Z π zeiz Reiθ · eiR cos(θ)−R sin(θ) dz| = | iReiθ dθ| 2 ei2θ + 2Reiθ + 10 z 2 + 2z + 10 R 0 Z π R2 ≤ e−R sin(θ) 2 dθ R − 2R − 10 0 Z π/2 R2 =2 e−R sin(θ) 2 dθ R − 2R − 10 0 Z π/2 2 R2 ≤ e−R· π θ 2 dθ R − 2R − 10 0 π R2 = (1 − e−R ) 2 → 0, 2R R − 2R − 10 as R → ∞. Therefore, as R → ∞, we have Z ∞ x sin(x) e−3 π dx = (3 cos(1) + sin(1)). 2 3 −∞ x + 2x + 10 5