A feature of averaging (Q2391797): Difference between revisions

Let \(1<p<\infty\) and let \(ces(p)\) (named after Cesàro) be the Banach sequence space \[ ces(p)=\left\{x=(x_n)\::\|x\|_{ces(p)}:=\sum_{n=1}^\infty\left(\frac{1}{n}\sum_{k=1}^n |x_k|\right)^p <\infty\right\}. \] For any sequence \(x=(x_n)_n\), let \(Cx=(\frac{1}{n}\sum_{k=1}^n x_k)_n\) and \(|x|=(|x_n|)_n\). Note that (by definition) \(x\in ces(p)\) exactly when \(C|x|\in\ell_p\) and that (thus) \(C|x|\in ces(p)\) exactly when \(CC|x|\in\ell_p\). So, if \(x\in ces(p)\), \(C|x|\in \ell_p\). But, by Hardy's classical inequality, then \(C|x|\in ces(p)\). The converse is true as well as is proved in Paragraph 20.31 of \textit{G. Bennett} [Mem. Am. Math. Soc. 576 (1996; Zbl 0857.26009)]. The main aim of the note under review is a short proof of an apparently new inequality, \[ (CC|x|)_n\geq\frac{1}{6}(C(|x|)_{[n/2]}, \tag{\(*\)} \] where \([\cdot]\) denotes the integer part. As a consequence, \[ \sum_{n=1}^\infty (C|x|)_n^p\leq 6^p \sum_{n=1}^\infty(CC|x|)_{2n}^p, \] from which Bennett's result follows directly. For a linear subspace \(\mathbb{X}\) of \(\mathbb{C}^\mathbb{N}\), denote \(ces(\mathbb{X})=\{x\in\mathbb{C}^\mathbb{N}:C|x|\in \mathbb{X}\}\). Inequality \((*)\) now lets us conclude the ``half Bennett result'' that \(x\in ces(\mathbb{X})\) if \(C|x|\in\mathbb{X}\) whenever \(\mathbb{X}\) is solid. Furthermore, in the cases when \(x\mapsto Cx\) maps \(\mathbb{X}\) into itself, we obtain the ``full Bennett result'' that \[ x\in ces(\mathbb{X})\Leftrightarrow C|x|\in ces(\mathbb{X)}. \] The last part of this very cute little note contains a list of a lot of cases where either the half or the full generalized Bennett theorem is obtained.