Real & Complex Analysis Qualifying Exam Solution, Spring 2010 Shiu-Tang Li January 1, 2013 A-1 R1 (a) Given > 0, P (|fn − fm | > ) < 1p 0 |fn (x) − fm (x)|p dx. Therefore, for each j ∈ N, there exists Nj such that P (|fn − fm | > 21j ) < 21j for every n, m ≥ Nj . If we choose n = Nj , m = Nj+1 , and define Aj := {|fNj −fNj+1 | > S∞ 1 }, we find that {fNj (x)} is Cauchy on ( m=n Am )c for every n; thus 2j {fNj (x)} is Cauchy a.e. x, actually. As a result,gj (x) := fNj (x) → f (x) for some finite a.e. f and the proof is complete. (b)Let f1 = χ[0,1/2] , f2 = χ[1/2,1] , f3 = χ[0,1/4] , f4 = χ[1/4,2/4] , f5 = χ[2/4,3/4] , f6 = χ[3/4,1] , f7 = χ[0,1/8] , f8 = χ[1/8,2/8] , · · · , and so on. A-2 (a)Approach 1. We would show that the inclusion map i is a closed operator. If fn → f in Lp (X), and i(fn ) = fn → g in Lq (X), then by the application of Chebeshev’s inequality we have fn → f in measure and fn → g in measure. Thus, f = g a.e. X. By closed graph theorem, The inclusion i is continuous linear operator and hence it is bounded. Approach 2. The idea is basically due to Shiang Tang. Assume that i is not bounded. there exists kfn kp = 1 for all n but kfn kp ≥ n2n . P Thus, Consider f = n=1 21n |fn |. It is obvious that f ∈ Lp (X); however, for every n ∈ N, kf kq ≥ k 21n fn kq ≥ n, so f ∈ / Lq (X), contradicts the fact that p q L (X) ⊂ L (X). (b) Assume that inf E 0 ∈M 0 µ(E 0 ) = 0, so we could pick a sequence of An ∈ M 0 so that µ(An ) → 0. Let fn = µ(An )−1/p χAn , and it is easy to 1 check that kfn kp = 1 for every n ∈ N. However, kfn kq = µ(An )1/q−1/p → ∞ as n → ∞. (a) above is thus violated. Remarks. To prove conversely, consider inf E 0 ∈M 0 µ(E 0 ) = m > 0. For every Rf ∈ Lp , we claim that f is bounded and hence f ∈ Lq . Otherwise, R P P ∞ ∞ ∞ > PX |f |p dµ = n=0 X |f |p χ{n≤|f |<n+1} dµ ≥ n=0 np µ(n ≤ |f | < n + 1) ≥ ∞ of f will result in n=1 µ(n ≤ |f | < n + 1) = ∞, since unboundednessP infinitely many positive terms that are larger than m in ∞ n=1 µ(n ≤ |f | < n + 1), and this is a contradiction. A-3 (a) Decompose H = V ⊕ V ⊥ . For every h = v + w, h0 = v 0 + w0 , where v, v 0 ∈ V and w, w0 ∈ V ⊥ , (i)πV (h) = πV2 (h) = v – idempotent kvk V (h)k = khk ≤ 1 for all h ∈ H. – Norm (ii) Since khk2 = kvk2 + kw2 k, kπkhk less than 1. (iii) (πV (h), h0 ) = (v, v 0 + w0 ) = (v + w, v 0 ) = (h, πV (h0 )). – Self adjoint (b) Let N (P ) be the null space of P . we claim that N (P ) is closed, for if an ∈ N (P ) and an → a ∈ N (P ), P (an ) = 0 → P (a) = 0 by continuity of P . Now we decompose H = N (P ) ⊕ N (P )⊥ . For every a ∈ N (P ), b ∈ N (P )⊥ , 0 = (P (a), b) = (a, P (b)), hence P (b) ∈ N (P )⊥ for all b ∈ N (P )⊥ . We claim that b = P (b) for all b ∈ N (P )⊥ . Consider P (b − P (b)) = P (b) − P 2 (b) = P (b) − P (b) = 0, showing b − P (b) ∈ N (P ). However, b − P (b) also belongs to N (P )⊥ due to the arguments above. The only possibility is that b − P (b) = 0. Therefore, for every u = a + b ∈ H, a ∈ N (P ) and b ∈ N (P )⊥ , P (a + b) = P (b) = b, and the proof is complete. A-4 P (a) Norm-preserving: For every ξ ∈ l∞ , kξkl∞ = 1, Λη (ξ) ≤ i |ηi | = kηkl1 . Thus kηk(l∞ )∗ ≤ kηkl1 . Conversely, if we choose ξ so that ξi = sgn(ηi ), 2 then kηk(l∞ )∗ ≥ Λη (ξ) = P i |ηi | = kηkl1 . P P Injection: if i ηi ξi = i ηi0 ξi for every ξi ∈ l∞ , then we may choose ξ so that ξj = 1 and ξk = 0 for k 6= j. Thus ηi = ηi0 for every i and hence η = η 0 . Not onto: let c ∈ (l∞ )∗ be such that c(ξ) = lim supn ξn and assume that c = Λη for some η ∈ l1 . If we choose ξ such that ξk vanishes in every subscript k but some j, then we have ηj = 0. Arbitrariness of j shows that η = 0 and c is a zero linear functional, which is a contradiction. (b)µ : σ−finite positive measure; 1 ≤ q < ∞; p is chosen s.t. For a proof, see Rudin page 127. 1 p + 1 q = 1. A-5 (Poincare’s recurrence theorem) Given E ∈ M , let the collection of sets satisfying this property be A = AE , and note that A is measurable. We claim that {T −n (A)}n are pairwise disjoint. If, x ∈ T −n (A) ∩ T −m (A), m > n, then T n (x) ∈ A and T m−n T n (x) ∈ A, a contradiction to the definition of A. P P S −n −n (A)) = ∞ (A)) = ∞ Therefore, ∞ > µ(X) ≥ µ( ∞ n=1 µ(A), n=1 µ(T n=1 T the only possibility is that µ(A) = 0. B-6 (a) Res(f ; 1) = limz→1 (z − 1)f (z) = 6+2 = R1; Res(f ; −1) = limz→−1 (z + 2·4 −6+2 1)f (z) = −2·2 = 1; by the residue theorem, γ f (z) dz = 2πi(Res(f ; 1) + Res(f ; −1)) = 4πi. 3 (b) 1 1 z · (1 − 2 )−1 · (1 + )−1 2 3z z 3 ∞ ∞ m X X 1 1 mz = (6z + 2) · 2 · (−1) m 3z n=0 z 2n m=0 3 f (z) = (6z + 2) · = = ∞ X 2 · (−1)m n,m=0 ∞ X k=−∞ m−2 1 z m−1 2 m 1 z (−1) + z 2n 3m 3 z 2n 3m X 2(−1)m k 2(−1)m z . + m m+1 3 3 m,n≥0:m−2−2n=k m,n≥0:m−1−2n=k X Where the interchange of summation is due to absolute and uniform converP∞ 2(−1)2n P∞ 2(−1)2n+1 −1 gence. The coefficient of z is + n=0 32n+2 = 2. Therefore, n=0 32n R R 2 f (z) dz = γ z dz = 4πi. γ B-7 Z 2π 2n Z 1 1 ( (z + z −1 ))2n dz iz |z|=1 2 Z (z 2 + 1)2n 1 dz = n i4 |z|=1 z 2n+1 cos θ dθ = 0 1 (z 2 + 1)2n 2π 2n = 2πi · n Res( ; 0) = n · . 2n+1 i4 z 4 n B-8 (a) Let f 0 (z0 ) = 0. If f 0 (z) ≡ 0 on U , then f is a constant, and of course it is not injective. So by uniqueness theorem we may assume that there is some neighborhood Bz0 (R) of z0 such that f 0 (z) 6= 0 for z ∈ Bz0 (R) \ {0}. Since f 0 (z0 ) = 0, we have f (z) = f (z0 ) + (z − z0 )k g(z) in a neighborhood B := Bz0 (r) of z0 , where k ≥ 2, and g is holomorphic and non-vanishing on B. And here we choose r < R. Let m := 12 supξ∈∂B |(z − z0 )k g(z)| > 0, we have that (z − z0 )k g(z) + m and (z − z0 )k g(z) have both k zeros on B by 4 Rouche’s theorem. Therefore, f (z) − f (0) + m has k zeros on B. Since (f (z)−f (0)+m)0 = f 0 (z) 6= 0 on B \{0}, These k zeros {z1 , · · · , zk } of f (z) − f (0) + m on B are all distinct, and obviously none of them equals 0. Therefore, f (z1 ) = · · · = f (zk ), thus f is not injective. (b) False. Let f (z) = z 2 defined on an open connected set which contains {1, −1} but does not contain {0}. B-9 2+z Consider a(z) = 2−z , which is analytic on HD2 := {z : |z| < 2, Re z > 0}, and it maps HD2 to Q := {z = a + bi : a > 0, b > 0}. To see this, notice that Arg(2+z) = Arg(z−(−2)), and if we pick −2,2, and z ∈ HD2 as vertices of a triangle, it is an obtuse triangle. Therefore, 0 < Arg(2 + z) − Arg(2 − z) < π2 , 2+w and this shows 2+z ∈ Q. Besides, for any z ∈ Q, z = 2−w ⇒ w = 2z−2 ∈ 2−z z+1 2z−2 −1 −1 HD2 , we may define a (z) : Q → HD2 by a (z) = z+1 , which is the inverse map of a. Therefore, a is a conformal map (analytic and bijective) from HD2 to Q. Let b(z) = z 2 : Q → H := {z : Im z > 0}. It is easy to check that b is also a conformal map. i−z Finally, we define c(z) = i+z , which is analytic on H, and for z = a + bi, b > 0, |i − z| = | − a + (1 − b)i| < |a + (1 + b)i| = |i + z|, which shows that c : H → D := {z : |z| < 1}. To see c(z) is onto and 1 − 1, for any 1−r2 −2i sin(θ) 1−reiθ i−z w = reiθ ∈ D, r < 1, w = i+z ⇒ z = i 1+re iθ = i 1+r 2 +2r cos(θ) ∈ H. Thus, we may define the inverse map of c by c−1 (z) = i 1−z : D → H. These facts 1+z show that c is a conformal map from H to D. Therefore, F (z) := (c ◦ b ◦ a)(z) = 2+z 2 i−( 2−z ) 2+z 2 i+( 2−z ) is a conformal map from HD2 to D. If φ(z) is another conformal map from HD2 to D, then φ ◦ F −1 is a a−z conformal map from D to D, which takes form eiθ 1−az for some a ∈ D. As a−F (z) a result, φ(z) = eiθ 1−aF (z) . 5 B-10 (w) Define g(ζ) = f (ζ)−f , g(w) = f 0 (w). Since g is continuous on the unit ζ−w disk D, and is holomorphic on D \ {0}, thus g is holomorphic on D by Morera’s theorem. 2M As a result, when z 6= w, we have |g(w)| ≤ supζ∈∂B0 ( 1+r ) |g(ζ)| ≤ (1+r)/2 . 2 4M That is, |f (z) − f (w)| ≤ 1+r |z − w|, and the constant is independent of our choices of z, w. 6