Lower bounds for matrices (Q1080920)

Let A be a nonnegative matrix mapping \(\ell^ p\) into \(\ell^ q(0<p<\infty,0<q<\infty)\) and let x be an element of \(\ell^ p\). This paper deals with the problem of finding best possible constants \(\lambda\) independent of x such that the inequality (*) \(\| Ax\|_ q\geq \lambda \| x\|_ p\) holds for every x \((\| \cdot \|_ p\) denotes the usual norm on \(\ell^ p)\). The validity of (*) is shown for \(p\geq 1\), \(0<q\leq p\), and an explicit formula for \(\lambda\) as well as a detailed discussion of the case of equality in (*) are given. Moreover, the author considers the special cases of Hilbert and Cesàro matrices H and C, respectively, the latter one extending a result by \textit{R. Lyons} [Proc. Am. Math. Soc. 86, 694 (1982; Zbl 0503.47006)] to the space \(\ell^ p\). Finally, the values of \(\| Hx\|_ p\) and \(\| Cx\|_ p\) are compared and the integral analogue of (*) is studied. As an important tool of proof, the author uses summation by parts in \(\ell^ p\).