Rich families and elementary submodels (Q2248501): Difference between revisions

Nonseparable Banach spaces are frequently studied through a proper collection of separable subspaces, which can sometimes be organized to form a long sequence of projections or at least a so-called projectional skeleton. If \(\mathcal{F}\) is a collection of separable subspaces of a Banach space \(X\), a statement \(\phi\) is separably determined (using \(\mathcal{F}\)) if it holds in \(X\) as soon as it holds in \(F\) for every \(F\in\mathcal{F}\). In any non-trivial case, this requests of course that the collection \(\mathcal{F}\) is not too small. The present paper investigates the relationship between two concepts, the so-called rich families (which are cofinal families of separable subspaces closed under increasing sequences) and the subtler notion of elementary submodels, which stems from logic and the Löwenheim-Skolem theorem. The main result of this paper (Theorem 3.2) is that all statements which are separably determined by the method of rich families are also separably determined by the method of elementary submodels. The converse is shown to hold in many cases, and the question is left open if it always holds.