On an analogue of Hardy's inequality (Q1325293)

Let \(1<p <\infty\) be fixed. Then Hardy's inequality \[ \sum_{n=0}^ \infty \Biggl( (n+1)^{-1} \sum_{k=0}^ n x_ k \Biggr)^ p\leq \biggl( {p\over {p-1}} \biggr)^ p \sum_{n=0}^ \infty x^ p_ n \] holds for each \((x_ k)\) with \(x_ k\geq 0\) \((k=0,1,\dots)\) which is equivalent to \(\ell_ p\subset \text{ces}_ p\) and the statement that the operator norm of the identity injection of \(\ell_ p\) into \(\text{ces}_ p\) is \({p\over {p-1}}\). Quite similar, for a given bounded sequence \((p_ n)\) with \(1<p_ n\) \((n=0,1,\dots)\) the inclusion \((*)\) \(\ell(p_ n) \subset \text{ces} (p_ n)\) implies and is implied by an analogue of Hardy's inequality. The present paper deals with the question: when does \((*)\) hold? Theorem 1 contains a characterization of the case \(\text{ces} (p_ n)= \{0\}\); obviously, in that case \((*)\) fails. Theorem 2 gives a sufficient condition for \((*)\) to fail, in the case when \((p_ n)\) is not a convergent sequence, whereas Theorems 3 and 4 are addressed to the case when \((p_ n)\) is a convergent sequence. The paper is closed by some remarks, problems and conjectures.