Real & Complex Analysis Qualifying Exam Solution, Spring 2008 Shiu-Tang Li Finished: January 1, 2013 Last updated: November 25, 2013 A-1 R Pick M > 0 so that {f >M } f dµ < /2 by monotone convergence theorem. and everything is fine now. Now let δ = 2M A-2 Z 1 Z 1 Z 1 2 2 1/2 f (y) dy)2 dx)1/2 kKk∞ ( kT f k2 = ( (T f (x)) dx) ≤ ( 0 0 Z 1 Z0 1 kKk2∞ ( f (y)2 dy) dx)1/2 ≤( 0 Z 1 Z0 1 2 1/2 ≤( kKk∞ dx) ( f (y)2 dy)1/2 0 0 = kKk∞ kf k2 . A-3 We first claim that self-adjoint operators are bounded (Hellinger-Toeplitz). To see this we consider un ∈ H for every n ∈ N, and un → u, T un → v. Since (T un , w) → (v, w) and (T un , w) = (un , T w) → (u, T w) = (T u, w) for every w ∈ H, it follows that T u = v, and closed graph theorem shows us that T is bounded. 1 Let B be a bounded set in H. Our goal is to show that T (B) is precompact. Pick a sequence {T (xn )} from T (B), we may write each T (xn ) = an1 y1 + · · · + anm ym , where {y1 , · · · , ym } is an orthonormal set that spans T (H). Since T is bounded, |T (xn )| is bounded in n, and thus supi,j |aij | < ∞. by Bolzano-Weierstrass theorem we can find a subsequence Pm {T (xnj )}j so that anj ,k → bk ∈ R for 1 ≤ k ≤ m. Therefore, T (xnj ) → k=1 bk yk . Therefore T (B) is precompact, or relative compact. The final task is to show V = K ⊥ . For each u ∈ H, v ∈ K, (T u, v) = (u, T v) = (u, 0) = 0. Thus V ⊂ K ⊥ . On the other hand, since V and K are both closed linear spaces, we have H = V ⊕ V ⊥ = K ⊕ K ⊥ . Besides, since every element w in V ⊥ satisfies (w, T y) = 0 for every y ∈ H, we have (T w, y) = 0 for every y ∈ H and thus T w = 0. This shows V ⊥ ⊂ K. Now for every x ∈ / V , we have x = v + v 0 , where v ∈ V , v 0 ∈ V ⊥ , and v 0 6= 0. Since V ⊂ K ⊥ and V ⊥ ⊂ K, it follows that x ∈ / K ⊥ . Therefore, K ⊥ ⊂ V . The proof of V = K ⊥ is complete. A-4 (The proof may also be found in Theorem 5.14 of [Rudin], but I adopt a slightly different approach here.) Given > 0, for every f ∈ L1 [−π, π], we may find M > 0 large so that kf − fM kL1 [−π,π] < , where fM := (f ∧ M ) ∨ (−M ). We let µ be the Lebesgue measure on [−π, π]. By Lusin’s theorem (Theorem 9. above), since µ(x : fM (x) 6= 0) < ∞, there is a function g ∈ Cc [−π, π] such that µ(fM 6= g) < /M , and supx∈[−π,π] |g(x)| ≤ supx∈[−π,π] |fM (x)| ≤ M . It follows that kg − fM kL1 [−π,π] < 2, which shows C[−π, π] is dense in L1 [−π, π]. Since C[−π, π] contains all uniformly continuous functions on [−π, π], for Pm every g ∈ C[−π, π] we may find h(x) = i=1 ai χ[bi ,bi+1 ) , −π = b1 < b2 < · · · < bm+1 = π so that h is close to g in L1 -norm. 2 Consider Z π m X ai χ[bi ,bi+1 ) e inx dx = −π i=1 = m X i=1 m X Z bi+1 ai einx dx bi ai i=1 1 inbi+1 (e − einbi ) in which → 0 as n → ∞. Therefore, for every f ∈ L1 [−π, π] we may pick h of the form above and kf − hkL1 [−π,π] < δ, and thus Z π | inx f (x)e −π Z π dx| ≤ Z π |f (x) − h(x)| dx + | −π h(x)einx dx| −π ≤ 2δ for all n large enough. This completes the proof. A-5 For every x ∈ l∞ , kT xkl1∗ = supkykl1 =1 | kxkl∞ . This shows that kT k ≤ 1. P i xi yi | ≤ supkykl1 =1 kxkl∞ P i |yi | = For x, y ∈ l∞ , x 6= y, ∃i so that xi 6= yi . Define ei so that (ei )k = δik . We have T x(z) = xi 6= yi = T y(z). P To prove that T is onto, for every x ∈ l1∗ , decompose x = i zi (ei )∗ , where (ei )∗ ∈ l1∗ and (ei )∗ (ej ) = δij . We find that supi |zi | < ∞; otherwise, |x(en )| we may pick a subsquence {znj }j so that |znj | → ∞, and ken kj 1 = |znj | j l is unbounded in j, contradicts the fact that x is bounded. Now we have z = (z1 , z2 , · · · ) ∈ l∞ and T z = x. The proof is complete. B-6 Hard. B-7 P n zn Since cos(z) = 12 (eiz +e−iz ) is an entire function, cos(z) = ∞ n=0 (−1) (2n)! P∞ n z n−2 for every z ∈ C. Therefore, when z 6= 0, cos(z) = n=0 (−1) (2n)! = z2 P∞ zn n n=−2 (−1) (2(n+2))! , which is the Laurent series for F . 3 B-8 For every piecewise C 1 closed curve γ 0 ⊂ 4, Z Z Z Z bZ d f (z) dz = φ(w)g(w − z) dw dz = φ(w(s))g(w(s) − z(t))w0 (s)z 0 (t) ds dt γ0 γ0 γ dZ b Z = c Z φ(w(s))g(w(s) − z(t))w0 (s)z 0 (t) dt ds Z φ(w(s))g(w(s) − z))w0 (s) dz ds γ0 c = c a d = Z a d 0 · w0 (s) ds = 0, c since as a function of z, φ(w(s))g(w(s) − z)) is analytic in the region enclosed by γ 0 . As a result, by Morera’s theorem, f is analytic on 4. (Actually it suffices to show the result above for every triangular closed curve.) B-9 Assume that f (z) 6= 0 for every z ∈ C, and lim inf |z|→∞ |f (z)| > 0. Since for each pole zk , limz→zk |f (z)| = ∞, it follows that inf z∈C |f (z)| > 0 ⇒ 1 1 supz∈C |f (z)| < ∞. Since now f (z) is a bounded entire function, f (z) is a constant function by Louville’s theorem, which contradicts the assumption of the problem. B-10 Define g(w) = 1r (f (Rw + z) − f (z)). Note that g maps the unit disk to itself, g(0) = 0, and g is holomorphic. By Schwarz lemma, we have (i)|g(z)| ≤ |z|, and if the equality holds for some z0 6= 0 in the unit disk, then g is a rotation. (ii)|g 0 (0)| ≤ 1, and if the equality holds, then g is a rotation. By (ii), |g 0 (0)| = | 1r f 0 (Rw + z)R|w=0 | ≤ 1 ⇒ |f 0 (0)| ≤ 4 r . R