In compactly generated categories, compact objects control everything: they generate the category, and maps out of them into filtered colimits factor through a finite stage. In dualizable categories there may be no nonzero compact objects at all, and their role is taken over by compact morphisms, which we introduced in Section 3.1. We develop their factorization properties and use them to characterize dualizability.
The argument proceeds from individual maps to whole categories: the lifting criteria lead to interpolation, compact exhaustions give generators, and (AB6) packages the result as a commutation property of products and filtered colimits. The final section uses these criteria to prove automatic dualizability.
Throughout, for a category \(C\) with filtered colimits we write \(\colim \colon \Ind (C) \to C\) for the colimit functor and \(\varepsilon \colon \hat {y} \to y\) for the natural transformation of Section 3.1, whenever \(C\) is compactly assembled. (In the talk, as in [Krause et al. (2024)], the colimit functor was denoted \(k\).) The talk mostly worked with dualizable categories. Several of the results below hold more generally for compactly assembled categories, and we state them in this generality whenever the proof is the same; see [Krause et al. (2024), Sections 2.2--2.4] for a systematic treatment of the non-stable case.
Sections
Weakly compact morphisms
Elementwise compactness, interpolation, and the stable compactness criterion.
Compactly exhaustible objects
Compact exhaustions and a characterization of dualizable categories.
A universal property of ŷ
Filtered-colimit-preserving approximations and the universal property of the finest presentation.
The axiom (AB6)
Products of ind-objects and the Clausen–Scholze characterization of dualizability.
Automatic dualizability and images
Dualizability via generating strongly continuous functors, and images of functors.