Real & Complex Analysis Qualifying Exam Solution, Spring 2009 Shiu-Tang Li December 21, 2012 A-1 Refer to the solution of A-1(b) in Spring 2010 qualifying exam. A-2 R Since XR |f |p dµ ≤ kf kp∞ µ(X), kf kp ≤ kf k∞ . On the other hand, for every > 0, ( X |f |p dµ)1/p ≥ (kf k∞ −)µ(|f | > kf k∞ −)1/p ⇒ lim inf p→∞ kf kp ≥ kf k∞ − . Arbitrariness of and the fact that kf kp ≤ kf k∞ for every p > 0 shows that limp→∞ kf kp exists and equals kf k∞ . A-3 Since T maps {x : kxkX = 1, x ∈ X} to a precompact set in Y , it follows that T is a bounded linear operator (and is hence continuous). The fact that T is onto and the open mapping theorem tells us T maps every open set in X to some open set in Y . Now for each y ∈ Y , y = T (x) for some x ∈ X. It follows that T (B1 (x)) is a open set that contains y, and it is percompact since B1 (x) is a bounded set. As a result, T (B1 (x)) is a compact neighborhood of y (Since it contains T (B1 (x))). The above arguments show that Y is locally compact. Since Y is also a topological vector space, it follows that Y is finite dimensional. It is not an easy job to prove this; alternatively, we may use the fact that Y is a locally 1 compact normed linear space, of which the proof relies only Riesz’s lemma and is easier. A-4 See the solution of A-3 in Fall 2010 qualifying exam. A-5 ˆ Since f ∈ L1 (R), convergence theoR∞ R ∞ f (x) ∈ C(R)R 1by use of dominated rem. Therefore, −∞ |fˆ(x)| dx ≤ −1 |fˆ(x)| dx + −∞ |xfˆ(x)| dx < ∞, and this shows fˆ(x) ∈ L1 (R), R ∞and now Fourier inversion formula is applicable and we have f (x) = √12π −∞ fˆ(y)eixy dy a.e. R. R∞ √1 fˆ(y)eixy dy is 2π −∞ R∞ ihy √1 fˆ(y)eixy e h−1 dy. 2π −∞ The final step is to show g(x) = differentiable any- where. Consider g(x+h)−g(h) = As h → 0, mean h value theorem along with dominated convergence theorem R ∞ with dominating √i = y fˆ(y)eixy dy. function |2y fˆ(y)| tells us that limh→0 g(x+h)−g(h) h 2π −∞ B-6 First we observe that sin(z) z = P n z 2n n=0 (−1) (2n+1)! = 1+z 2 n z 2(n−2) n=1 (−1) (2n+1)! . P = 1, and we have f (z)z → 1 as z → 0, which implies 0 is Thus limz→0 sin(z) z a simple pole for f , and the residue at 0 is exactly limz→0 f (z)z = 1. The residue formula implies that R γ f (z) dz = 2πiRes(f ; 0) = 2πi. B-7 See B-7 of Spring 2011. But this is much easier - we only need to subtract principal part of each pole from f , to form a bounded entire function. B-8 on Ω. Even if the line integral R We MUST assume that f is continuous 1 f (z) dz = 0 for every piecewise-C closed curve, we still need to assume γ 2 continuity to assure that Morera’s theorem works. For each z0 ∈ Ω, there is some > 0 so that Bz0 () ⊂ Ω. Define R F (z) = γ(z) f (w) dw on Bz0 (), where γ(z) is the line segment(s) starting from z0 = a + bi to c + bi, then c + bi to c + di = z. (z) = For h 6= 0 small enough, For every z ∈ Bz0 () we have F (z+h)−F h R 1 f (w) dw, where γ(z, h) is the line segment(s) starting from z = a + bi h γ(z,h) to c + bi, then c + bi to c + di = z + h, by the assumption of this problem. Therefore, Z 1 F (z + h) − F (z) 1 − f (z)| = | | f (w) dw − hf (z)| h h γ(z,h) h Z 1 f (w) − f (z) dw| =| h γ(z,h) Z 1 ≤ |f (w) − f (z)| dw |h| γ(z0 ,h) 1 ≤ |h| sup |f (w) − f (z)| → 0 as h → 0, |h| w∈γ(z,h) since supw∈γ(z,h) |f (w) − f (z)| → 0 due to continuity of f at z. This implies F 0 (z) = f (z) on Bz0 (). We can define F (z) on every small ball B ∈ Ω in this way. Consider F1 defined on B1 , F2 defined on B2 , B1 ∩ B2 6= ∅, but F1 6= F2 on B1 ∩ B2 , then F10 (z) − F20 (z) = f (z) − f (z) = 0, so that F1 (z) − F2 (z) = c on B1 ∩ B2 . We may redefine F2 by adding a constant to make F1 = F2 on B1 ∩ B2 . S Now we write Ω = z∈Ω Bz (rz ), and by Lindelöf’s Theorem, we have S∞ further Ω = n=1 Bn , where each Bn is of form Bz (rz ), and there is some Fn (z) defined on Bn so that Fn0 = f on Bn . We may apply the technique in the previous paragraph to redefine Fn (z) for every n ∈ N, and define F (z) := Fn (z) on Bn . The proof is complete. B-9 It seems that this problem is misstated. 3 B-10 Z 0 2π 1 dθ = a + sin θ Z Z|z|=1 = |z|=1 1 1 · dz iz a + (z − z −1 )/2i 2 dz i2az + z 2 − 1 √ Solving √ i2az + z 2 − 1 = 0, we have z = −i2a ± 4a2 − 1i. Since a > 1, only −i2a + 4a2 − 1i = √4a2−i lies in the unit disk. By the residue theorem, −1+2a √ R 2 2 dz = 2πiRes( ; −i2a + 4a2 − 1i) = 2πi √4a12 −1i = 2 2 i2az+z −1 |z|=1 i2az+z −1 √ 2π . 4a2 −1 4