A general Pietsch domination theorem (Q616721)

Let \(X\), \(Y\) and \(E\) be (arbitrary) nonvoid sets, \(\mathcal{H}\) be a family of mappings from \(X\) to \(Y\), \(G\) be a Banach space and \(K\) be a compact Hausdorff topological space. Let \[ R : K\times E \times G\longrightarrow[0,\infty)~\mathrm{and~} S :{\mathcal{H}}\times E \times G\longrightarrow[0,\infty) \] be arbitrary mappings and \(0<p<\infty\). A mapping \(f\in{\mathcal{H}}\) is said to be \textit{\(R\)-\(S\)-abstract \(p\)-summing} is there is a constant \(C_{1}>0\) so that \[ \left(\sum_{j=1}^{m}S(f,x_{j},b_{j})^{p}\right)^{\frac1p}\leq C_{1}\sup_{\varphi\in K}\left(\sum_{j=1} ^{m}R\left( \varphi,x_{j},b_{j}\right) ^{p}\right)^{\frac1p} \] for all \(x_{1},\dots,x_{m}\in E,\) \(b_{1},\dots,b_{m}\in G\) and \(m\in\mathbb{N}\). Consider the following conditions: {\parindent6mm \begin{itemize}\item[(1)] For each \(f \in \mathcal{H}\), there is \(x_{0} \in E\) such that \(R(\varphi, x_{0}, b) = S(f, x_{0}, b) = 0\) for every \(\varphi\in K\) and \(b \in G\). \item[(2)] The mapping \(R_{x,b} : K \longrightarrow[0, \infty)~,~R_{x,b}(\varphi) = R(\varphi,x,b)\) is continuous for every \(x \in E\) and \(b \in G\). \item[(3)] It holds that \(R\left( \varphi,x,\eta b\right) \leq\eta R\left( \varphi,x,b\right) ~\mathrm{and~} \eta S(f,x,b) \leq S(f,x,\eta b)\) for every \(\varphi\in K\), \(x \in E\), \(0 \leq\eta\leq1\), \(b\in G\) and \(f \in{\mathcal{H}}\). \end{itemize}} In [``A unified Pietsch domination theorem,'' J.~Math.\ Anal.\ Appl.\ 365, No.\,1, 269--276 (2010; Zbl 1193.46026)], \textit{G.\,Botelho, D.\,Pellegrino} and \textit{P.\,Rueda} gave an abstract version of Pietsch's domination theorem which unifies a number of known domination theorems for classes of linear and nonlinear mappings and showed that the Pietsch-type domination does not depend really on any algebraic condition. More precisely, it was proved that under the conditions (1), (2) and (3), a map \(f\in{\mathcal{H}}\) is \(R\)-\(S\)-abstract \(p\)-summing if and only if there is a constant \(C>0\) and a regular Borel probability measure \(\mu\) on \(K\) such that \[ S(f,x,b)\leq C\left( \int_{K}R\left( \varphi,x,b\right) ^{p}d\mu\left( \varphi\right) \right) ^{\frac{1}{p}} \] for all \(x\in E\) and \(b\in G\). In the paper under review, it is proved that the same characterization holds true without the hypotheses (1) and (3) on \(R\) and \(S\). First, the use of Ky Fan Lemma as in Pietsch's original proof permits the authors to remove (1). Then, an argument due to M.\,Mendel and G.\,Schechtman that appears in [\textit{J.\,D.\thinspace Farmer} and \textit{W.\,B.\thinspace Johnson}, ``Lipschitz \(p\)-summing operators'', Proc.\ Am.\ Math.\ Soc.\ 137, No.\,9, 2989--2995 (2009; Zbl 1183.46020)] allows them to get the result with no hypothesis on \(S\) and just supposing that \(R\) satisfies (2).