Chapter 5

Infinite Matrices

[6/100%]
Show
LaTeX
Problem 44

Every operator has a column-finite matrix. More precisely, if AA is an operator on a Hilbert space H\mathbf{H}, then there exists an orthonormal basis {ej}\left\{ e_j\right\} for H\mathbf{H} such that, for each jj, the matrix entry ⟨Aej,ei⟩\left\langle Ae_j, e_i \right\rangle vanishes for all but finitely many ii's.

?
Problem 45

If αij≥0\alpha_{ij} \geq 0, if pi>0p_i > 0 and qj>0q_j > 0 (i,j=0,1,2,…)(i, j = 0, 1, 2, \ldots ), and if β\beta and γ\gamma are positive numbers such that

∑iαijpi≤βqj(j=0,1,2,…),∑jαijqj≤γpi(i=0,1,2,…), \begin{align} \sum _i \alpha _{ij} p_i & \leq \beta q_j & (j = 0, 1, 2, \ldots ), \\ \sum _j \alpha _{ij} q_j & \leq \gamma p_i & (i = 0, 1, 2, \ldots ), \end{align}

then there exists an operator AA (on a separable infinite-dimensional Hilbert space, of course) with ∥A∥2≤βγ\left\| A\right\|^2 \leq \beta \gamma and with matrix ⟨αij⟩\left\langle \alpha_{ij} \right\rangle (with respect to a suitable orthonormal basis).

?
Problem 46

There exists an operator AA (on a separable infinite-dimensional Hilbert space) with ∥A∥≤π\left\| A\right\| \leq \pi and with matrix ⟨1/(i+j+1)⟩\left\langle 1/(i+j+1) \right\rangle (i,j=0,1,2,…)(i, j = 0, 1, 2, \ldots ).

?
Remark.
?

This matrix is named after Hilbert and is accordingly called the Hilbert matrix; the norm of the matrix is in fact equal to π\pi.

Problem 47

The matrix ⟨2−(i+j+1)⟩\left\langle 2^{-(i+j+1)} \right\rangle is bounded. What is its norm?

?
Remark.
?

A matrix whose ⟨i,j⟩\left\langle i, j \right\rangle entry is a function of i+ji+j only is called a Hankel matrix. The Hilbert matrix (Problem 46) is the Hankel matrix that corresponds to the function φ\varphi defined by φ(x)=1/(x+1)\varphi (x) = 1/(x+1) (i.e., αij=φ(i+j)\alpha_{ij} = \varphi (i+j), i,j=0,1,2,…i, j = 0, 1, 2, \ldots); the matrix ⟨2−(i+j+1)⟩\left\langle 2^{-(i+j+1)} \right\rangle is the Hankel matrix corresponding to φ(x)=2−(x+1)\varphi (x) = 2^{-(x+1)}.

Problem 48

Is the Hilbert matrix positive?

?
Remark.
?

The Gramian of a finite or infinite sequence {fn}\left\{ f_n\right\} of vectors is the matrix whose ⟨i,j⟩\left\langle i, j \right\rangle entry is the inner product ⟨fi,fj⟩\left\langle f_i, f_j \right\rangle.

Problem 49

Under what conditions on a sequence {fn}\left\{ f_n\right\} of vectors in a Hilbert space H\mathbf{H} does the series ∑nαnfn\sum_n \alpha_n f_n converge in H\mathbf{H} for every sequence α\alpha in l2l^2?

?