Function spaces arising from kernel operators (Q618825)

This paper is devoted to the characterization of rearrangement invariant function spaces to which particular classical inequalities that hold for well-known Banach function spaces can be extended. Sometimes, this problem is equivalent to determining the optimal domain of a kernel operator. In particular, using the technique proposed by the authors, some classical inequalities for relevant operators in analysis -- as Sobolev type inequalities or the Riesz potentials -- can be extended to larger rearrangement invariant (r.i.) Banach function spaces. In fact, the largest r.i. spaces satisfying these inequalities can be characterized. The problem can be reduced often to the following one. Consider an operator \(T_K:L^\infty[0,1] \to X\) defined by a positive kernel \(K\). The space \([T_K,X](\Omega):= \{u:\Omega \to \mathbb R: T_K u^* \in X \}\) is the natural candidate for representing optimal domains preserving r.i. spaces. Some properties of this space are analysed in terms of the increasing properties of the kernel. Theorem 3.10 and Theorem 3.13 provides reasonable conditions for \([T_K,X](\Omega)\) to be a quasi-normed linear space for r.i. spaces \(X\). In particular, the second one depends on some growth properties in one or both variables that are given in terms of inequalities; for instance, a kernel \(K\) has bounded growth in the second variable if there is a constant \(C>0\) such that \(K(x,2y) \leq C K(x,y)\), \((x,y) \in [0,1] \times [0,1/2]\) (property \(bg_2\)). Propositions 3.20 and 3.21 provides a lot of information on the space \([T_K,X](\Omega)\), concretely on normability and completeness. The paper finishes with some comments on the case of spaces with the Fatou property (Section 4). The paper contains a lot of information: examples, counterexamples and applications are given in each section, and the discussions on the results are useful and clarifying.