The eigenfunctions of the Hilbert matrix (Q1935029): Difference between revisions

The authors consider the Hilbert matrix \(H_{\lambda}=(\frac{1}{n+m+\lambda})_{n, m\geq 0}\), where \(\lambda \in {\mathbb C}\), \(\lambda \notin {\mathbb Z}\). Let \(X\) be one of the Banach spaces \(H^p, A^{-\tau }, A^{-\tau }_0\) of analytical functions on the unit disc of \(\mathbb C\). The norms in these spaces are given by \[ \|f\|^p=\sup _{0<r<1}\int \limits _0^{2\pi }|f(r\exp (i\theta))|^p\,\frac{d\theta}{2\pi}, \] \[ \|f\|_\tau=\sup (1-|z|)^{\tau}|f(z)|, \] and the space \(A_0^{-\tau }\) is distinguished be the condition \((1-|z|)^{\tau}|f(z)|\to 0\) as \(|z|\to 1^-\). The matrix \(H_{\lambda }\) determines a linear operator (also denoted as \(H_{\lambda }\)) in \(X\) with respect to the basis \(z^n\), \(n=0, 1, \dots \). This operator is expressed using a path integral. Using this, the authors prove that \(H_{\lambda }\) is bounded operator in \(H^p\), \(1<p<\infty\), and \(A^{-\tau }\), \( A^{-\tau }_0\), \(0<\tau <1\). (They state, in fact, a more general result.) Then some differential operators are introduced which ``almost commute'' with \(H_{\lambda }\) (this means that the commutator \([H_{\lambda }, D]\) is one-dimensional). The most interesting operator acts as \(Df(z)=z(z-1)^2f''(z)+(z-1)[(\lambda +2)z-\lambda]f'(z)+\lambda f(z)\). The eigenfunctions of \(D\) are calculated; they are expressed by hypergeometrical functions \(_2F_1\). This gives eigenfunctions of \(H_{\lambda }\). By using these results, all the eigenvalues and the eigenfunctions of \(H_{\lambda }\) in the Banach space \(X\) are described. For example, in the case \(H^p\), \(p\geq 2\), and \(A^{-\tau}\), \(0<\tau <1/2\), any eigenvalue has the form \(\pm \csc\pi \lambda \), where \((1-\lambda)^{\lambda }\in X\). Moreover, the multiplicities of these eigenvalues are \([N/2]\) and \([N-1/2]\), where \(N\) is the largest integer for which \((1-z)^{\lambda} \in X\). The eigenfunctions are, respectively, \(f_{2k}\) and \(f_{2k+1}\), where \(f_n \) is given by \((1-z)^{-n-\lambda }(1+z)^n\).