Cellular objects and Shelah's singular compactness theorem (Q899557)

Shelah's Singular Compactness Theorem states that if \(A\) is an abelian group of size \(\mu\), where \(\mu\) is a singular cardinal, such that all subgroups of size less than \(\mu\) are free, then \(A\) itself is free. This theorem has various generalizations which all look informally like the sentence ``if \(\mu\) is a singular cardinal and \(S\) a structure all of whose substructures of cardinality less than \(\mu\) are free, then \(S\) is free itself''. The goal of the paper is to make the above sentence precise. Two formulations are proposed: the functorial form and the cellular form. The functorial form starts from a functor \(F:\mathcal{A} \to \mathcal{B}\) preserving filtered colimits from an accessible category to a finitely accessible category: think of \(\mathcal{A}\) as the category of sets, \(\mathcal{B}\) as the category of groups and \(F\) as the free group functor. The functorial form states under a technical condition for \(F\) that if \(X\in \mathcal{B}\) is of size a singular cardinal \(\mu\), if all subobjects of \(X\) of size less than \(\mu\) are in the image of \(F\), then \(X\) itself is in the image of \(F\). The functorial form is actually a little bit more general since it only assumes that ``enough'' objects of \(X\) lie in the image of \(F\). The cellular form starts from a set of morphisms \(I\) of \(\mathcal{B}\) which is supposed now to be locally finitely presentable (the cocompleteness of \(\mathcal{B}\) is required indeed): think of \(I\) as the singleton \(\{\varnothing \subset \mathbb{Z}\}\). The cellular form states that if \(X\in \mathcal{B}\) is of size a singular cardinal \(\mu\), if all subobjects of \(X\) of size less than \(\mu\) are \(I\)-cellular (i.e. the unique map from the initial object is a transfinite composition of pushouts of maps of \(I\)), then \(X\) itself is \(I\)-cellular.