Bourbaki

Exercise 3, § 4

Topological Vector Spaces, V, p. 75

Bourbaki marks this one as harder than the rest.

Let (aij)(i,j)∈I×I(a_{ij})_{(i,j) \in I \times I} be a countably infinite double family of complex numbers. Assume that there exist two numbers β>0\beta > 0, γ>0\gamma > 0, and a family (pi)i∈I(p_i)_{i \in I} of numbers >0> 0 satisfying the relations ()

∑ipi∣aij∣≤βpj,∑j∣aij∣pj≤γpi \sum_i p_i |a_{ij}| \leq \beta p_j, \quad \sum_j |a_{ij}| p_j \leq \gamma p_i
for all i,ji, j in II. a) Show that there exists a continuous endomorphism uu from ℓc2(I)\ell^2_c(I), with norm ≤(βγ)1/2\leq (\beta \gamma)^{1/2}, such that for all x=(xi)i∈Ix = (x_i)_{i \in I} in ℓc2(I)\ell^2_c(I), we have u(x)=yu(x) = y, where y=(yi)y = (y_i) is given by yi=∑jaijxjy_i = \sum_j a_{ij} x_j (for x=(xi)x = (x_i) and y=(yi)y = (y_i) in CI\mathbf{C}^I), put vij=∣xi∣(pj∣aij∣/pi)1/2v_{ij} = |x_i| (p_j |a_{ij}| / p_i)^{1/2}, wij=∣yj∣(pi∣aij∣/pj)1/2w_{ij} = |y_j| (p_i |a_{ij}| / p_j)^{1/2}, and find a bound for ∑i,jvijwij\sum_{i,j} v_{ij} w_{ij}. b) For II, take the set of all integers ≥1\geq 1, and let aij=(i+j)−1a_{ij} = (i + j)^{-1}. Show that the conditions () are satisfied with pi=i−1/2p_i = i^{-1/2} and β=γ=π\beta = \gamma = \pi (these being the best possible constants) (compare the series in () with an integral). In an analogous manner, treat the case where I=NI = \mathbf{N} and aij=(i+j+1)−1a_{ij} = (i + j + 1)^{-1} ("Hilbert's matrix"). \ c) Let H=H2(D)\mathscr{H} = \mathbf{H}^2(D) be the Hardy space, consisting of all functions f(z)=∑n=0∞anznf(z) = \sum_{n=0}^\infty a_n z^n, analytic in the open disc D:∣z∣<1D : |z| < 1 and such that ∥f∥2=∑n∣an∣2<+∞\|f\|^2 = \sum_n |a_n|^2 < +\infty; then ∥f∥\|f\| is a norm on H\mathscr{H}, for which H\mathscr{H} is isomorphic to ℓc2(N)\ell^2_c(\mathbf{N}). Given two functions f,gf, g in H\mathscr{H}, show that the function t↦f(t)g(t)t \mapsto f(t) g(t) of the real variable tt is integrable on [0,1][0, 1] with respect to the Lebesgue measure and that the formula B(f,g)=∫01f(t)g(t) dtB(f, g) = \int_0^1 f(t) g(t) \, dt defines a continuous bilinear form on H×H\mathscr{H} \times \mathscr{H} (consider B(f,f)=∫01f(t)2 dtB(f, f) = \int_0^1 f(t)^2 \, dt for a function f∈Hf \in \mathscr{H} of the form f(z)=∑n=0Nanznf(z) = \sum_{n=0}^N a_n z^n with an≥0a_n \geq 0 for 0≤n≤N0 \leq n \leq N; use Cauchy's theorem to establish the relation
∫−11f(t)2 dt=−i∫0πf(eiθ)2eiθdθ \int_{-1}^1 f(t)^2 \, dt = -i \int_0^\pi f(e^{i\theta})^2 e^{i\theta} d\theta
from which we get B(f,f)≤12∫−ππ∣f(eiθ)∣2dθ=π∥f∥2B(f, f) \leq \frac{1}{2} \int_{-\pi}^\pi |f(e^{i\theta})|^2 d\theta = \pi \|f\|^2. In this way, get the result of b), according to which the Hilbert's matrix defines an endomorphism of norm ≤π\leq \pi of the hilbertian space ℓc2(N)\ell^2_c(\mathbf{N}).)

Solution

No solution has been written yet.

Transcribed from the printed volume. Not translated.