Mathematics 442C Suggested solutions to exercise sheet 2 1. (a) Observe that a(1 − ba) = (1 − ab)a. Thus (1+b(1−ab)−1 a)(1−ba) = 1−ba+b(1−ab)−1 (1−ab)a = 1−ba+ba = 1. Similarly, (1 − ba)b = b(1 − ab), so (1 − ba)(1 − b(1 − ab)−1 a) = 1 − ba + b(1 − ab)(1 − ab)−1 a = 1. (b) Let λ ∈ σ(ab) \ {0}. Then λ − ab = λ(1 − λ−1 ab) 6∈ Inv A, so 1 − λ−1 ab 6∈ Inv A. By (a), 1 − λ−1 ba 6∈ Inv A, so λ − ba = λ(1 − λ−1 ba) 6∈ Inv A. So λ ∈ σ(ba) \ {0}. Hence σ(ab) \ {0} ⊆ σ(ba) \ {0}. The reverse inclusion follows by symmetry. 2. Since a and a−1 are invertible, 0 is not in σ(a) or σ(a−1 ). If λ ∈ C with λ 6= 0 then λ−1 − a−1 = λ−1 (a − λ)a−1 by (⋆). Since λ−1 (a − λ) and a−1 commute, by Remark 1.2.5(iii) we have λ ∈ σ(a) ⇐⇒ a − λ 6∈ Inv A ⇐⇒ λ−1 − a−1 6∈ Inv A ⇐⇒ λ−1 ∈ σ(a−1 ). 3. For |λ| < 1/r(a) we have r(λa) = |λ|r(a) < 1 by Theorem 1.3.8, so 1 − λa is invertible and S(λ) is well-defined. For λ, µ in the domain of S, we have S(µ) − S(λ) 1 = (1 − µa)−1 − (1 − λa)−1 µ−λ µ−λ 1 (1 − µa)−1 (1 − λa) − (1 − µa) (1 − λa)−1 = µ−λ = (1 − µa)−1 a(1 − λa)−1 . by (⋆) The map µ 7→ 1−µa is continuous, inversion is continuous Inv A → Inv A and multiplication is continuous on A. Hence S(µ)−S(λ) → (1 − λa)−1 a(1 − λa)−1 µ−λ as µ → λ. So S is holomorphic. 4. If a ∈ S ′ then ab = ba for all b ∈ S, so ab = ba for all b ∈ T , so a ∈ T ′ . Hence S ′ ⊆ T ′. If a ∈ S then ab = ba for all b ∈ S ′ . Hence a ∈ S ′′ . So S ⊆ S ′′ . Since S ⊆ S ′′ we have S ′ ⊇ (S ′′ )′ = S ′′′ . Moreover, S ′ ⊆ (S ′ )′′ = S ′′′ . So we have equality. 5. We know that A/I is a Banach algebra. Since (a + I)(1 + I) = a + I = (1 + I)(a + I), the element 1 + I is an identity element for A/I, and k1 + Ik = inf k1 − bk ≤ k1 − 0k = 1. b∈I If k1 + Ik < 1 then there is b ∈ I with k1 − bk < 1. By Theorem 1.2.7, b = 1 − (1 − b) ∈ Inv A. So b ∈ I ∩ Inv A; since A is a proper ideal, this contradicts Lemma 1.5.5. So k1 + Ik = 1, and 1 + I is the unit for A/I. Here’s an even more elementary way to prove that k1 + Ik ≥ 1: since I is a proper ideal, 1 6∈ I, so 1 + I 6= 0, hence k1 + Ik 6= 0. Moreover, k1 + Ik = k(1 + I)2 k ≤ k1 + Ik2 , and cancelling k1 + Ik gives k1 + Ik ≥ 1. 6. If θ : M2 (C) → C is a homomorphism then θ(ab − ba) = θ(ab) − θ(ba) = θ(a)θ(b) − θ(b)θ(a) = 0 for all a, b ∈ M2 (C). Let for i, j = 1, 2, let eij ∈ M2 (C) be the matrix with a 1 in the (i, j) position and 0s elsewhere. Then eij ekℓ = δjk eiℓ , so θ(e12 ) = θ(e12 e22 − e22 e12 ) = 0 Hence θ(e11 ) = θ(e12 e21 ) = θ(e12 )θ(e21 ) = 0 and θ(e22 ) = θ(e21 e12 ) = θ(e21 )θ(e12 ) = 0, and θ(e21 ) = θ(e21 e11 ) = θ(e21 )θ(e11 ) = 0. Since M2 (C) = span{e11 , e12 , e21 , e22 } and θ sends each of these generators to 0, we have θ = 0. 7. (a) If fn ∈ K and fn → f ∈ C(D) then fn converges uniformly to f , so fn converges pointwise to f . If z ∈ T then fn (z) = 0 for all n ≥ 1. Hence f (z) = limn→∞ fn (z) = 0, so f ∈ K. Hence K is closed in C(D). If h ∈ C(T) and f, g ∈ K and λ ∈ C then for z ∈ T we have (f + λg)(z) = f (z) + λg(z) = 0 and (hf )(z) = (f h)(z) = f (z)h(z) = 0 · h(z) = 0, so f + λg and hf = f h are in K. Hence K is an ideal in C(T). (b) The mapping θ is well-defined, since if f +K = g +K then f −g ∈ K, so (f −g)|T = 0, so f |T = g|T . It is easy to check that θ is linear. Moreover, θ(f g) = (f g)|T = (f |T ) · (g|T ) = θ(f )θ(g), so θ is a homomorphism. Let f ∈ C(D). If g ∈ K then kf |T k = sup |f (z) − g(z)| ≤ sup |f (z) − g(z)| = kf − gk. z∈T z∈D Hence kf |T k ≤ inf g∈K kf − gk = kf + Kk. Let ε > 0. Since f is a continuous function on a compact set, it is uniformly continuous. Hence there is δ > 0 such that |f (z) − f (w)| ≤ ε whenever z, w ∈ D with |z − w| ≤ δ. Let s be a continuous function [0, 1] → [0, 1] with s(1) = 0 and s|[0,1−δ] = 1, and let g(z) = s(|z|)f (z) for z ∈ D. Observe that g is continuous, and if 1 − δ ≤ |z| ≤ 1 then z z | ≤ δ, so |f (z) − f ( |z| )| ≤ ε. Hence |z − |z| kf + Kk ≤ kf − gk = sup |f (z) − g(z)| = = sup (1 − s(|z|))|f (z)| ≤ 1−δ≤|z|≤1 ≤ sup 1−δ≤|z|≤1 |f (z) − g(z)| sup 1−δ≤|z|≤1 z∈D sup |f (z)| 1−δ≤|z|≤1 z z )| + |f (z) − f ( |z| )| ≤ kf |T k + ε. |f ( |z| Since ε > 0 was arbitrary, this shows that kf + Kk ≤ kf |T k = kθ(f )k. The reverse inequality has already been established, so we have equality. Hence θ is isometric. In particular, θ is injective. If g ∈ C(T), consider ( z ) if z 6= 0, |z|g( |z| f : D → C, z 7→ 0 if z = 0. Then f ∈ C(D) and θ(f + K) = f |T = g. Hence θ is surjective. Hence θ is an isometric isomorphism of Banach algebras. 2 8. (a) We have S ∗ S = I, so σ(S ∗ S) = {1}. Also, SS ∗ e1 = 0, so SS ∗ is not invertible. Hence 0 ∈ σ(SS ∗ ). By Exercise 1, σ(SS ∗ ) \ {0} = σ(S ∗ S) \ {0} = {1}, so σ(SS ∗ ) = {0, 1}. (b) We have T (λI − S ∗ ) = I, so T = λ−1 (I + T S ∗ ). Thus T e1 = λ−1 e1 , and for n ≥ 2 we have T en = λ−1 (en + T en−1 ). Solving this recurrence relation (by induction, if you like) gives T en = λ−1 en + λ−2 en−1 + · · · + λ−n e1 . (c) If x = P n≥1 xn en ∈ H then X X kS ∗ xk2 = k xn en−1 k2 = |xn |2 ≤ kxk2 . n≥2 n≥2 So kS ∗ xk ≤ kxk; hence kS ∗ k ≤ 1, and so σ(S ∗ ) ⊆ D. If λ ∈ D and λI − S ∗ is invertible, then T (λI − S ∗ ) = I for some T ∈ B(H). By (b), we have kT en k2 = |λ|−2 + |λ|−4 + · · · + |λ|−2n ≥ n so kT en k → ∞ as n → ∞. So T is not bounded, which is a contradiction. So λI − S ∗ is not invertible, so λ ∈ σ(S ∗ ). Hence σ(S ∗ ) = D. (d) If T ∈ Inv B(H) then T R = RT = I for some R ∈ B(H). Hence I = I ∗ = (T R)∗ = (RT )∗ , so I = R∗ T ∗ = T ∗ R∗ . So T ∗ ∈ Inv B(H). Now if T ∗ ∈ Inv B(H) then T = (T ∗ )∗ ∈ Inv B(H). Hence λ ∈ σ(S) ⇐⇒ λI − S 6∈ Inv B(H) ⇐⇒ (λI − S)∗ 6∈ Inv B(H) ⇐⇒ λI − S ∗ 6∈ Inv B(H) ⇐⇒ λ ∈ D ⇐⇒ λ ∈ D. So σ(S) = D. P xk 9. (a) We have ex = ∞ n=0 k! for every real number x, and pn (x) is the nth partial sum of this series. (b) We have X (x + y)k X 1 k X xℓ y m pn (x + y) = = . xr y k−r = k! k! r ℓ! m! 0≤k≤n 0≤r≤k≤n 0≤ℓ,m≤n, ℓ+m≤n So pn (x)pn (y) − pn (x + y) = X xj y k X xℓ y m X xj y k − = . j! k! ℓ! m! j!k! 0≤j,k≤n 0≤ℓ,m≤n 1≤j,k≤n, ℓ+m≤n Hence the coefficient of xj y k is either 0 or 1 j! k! j+k>n > 0. (c) We have pn (x) → ex , pn (y) → ey and pn (x + y) → ex+y . Hence qn (x, y) → ex ey − ex+y = 0 as n → ∞. P ak 1 1 ak k k (d) The series ∞ k=0 k! converges absolutely, since k k! k = k! ka k ≤ k! kak , P∞ kakk P∞ ak = ekak < ∞. Hence pn (a) converges to so k k! k ≤ k=0 k! P∞ k=0 ak k=0 k! as n → ∞. 3 P (e) Since ab = ba and qn (x, y) = 0≤j,k≤n tjk xj y k for some tjk ≥ 0 by (b), we have X tjk aj bk kpn (a)pn (b) − pn (a, b)k = 0≤j,k≤n ≤ X tjk kaj bk k by the triangle inequality 0≤j,k≤n ≤ X kakj kbkk 0≤j,k≤n = qn (kak, kbk). (f) We have kea eb − ea+b k = lim kpn (a)pn (b) − pn (a + b)k ≤ lim qn (kak, kbk) = 0 n→∞ n→∞ by (e) and (c). Hence ea eb = ea+b . (g) If a ∈ A then a commutes with −a. Hence e−a ea = ea e−a = ea−a = e0 = 1, so e−a = (ea )−1 and ea is invertible. Since a product of invertible elements is invertible, the result follows. 4