Definable categories (Q1689577)

As the authors say: ``The paper belongs to the realm of categorical logic where one studies the interplay between syntax and semantics using the language of category theory'' and: ``The most important new concept in this paper is that of a \textit{definable category}''. One may notice that the notion of definable category is introduced via equivalence with some subcategory of another category and thus is not ``absolute'': \par Definition 2.1.1. A full subcategory $\mathcal{L}$ of a locally finitely presentable category $\mathcal{K}$ is a \textit{definable subcategory} if it is closed in $\mathcal{K}$ under products, directed colimits and pure subobjects. Furthermore, a category is \textit{definable} if it is equivalent to a definable subcategory of a locally finitely presentable category. \par A functor between two definable categories is an \textit{interpretation functor} if it preserves products and directed colimits. \par The question of existence of an ``intrinsic'' characterization of definable categories remains open: ``in general it is hard to check whether a given category is definable''. \par Main inspiration for this notion of definability comes from the world of additive categories where similar approach was developed earlier [\textit{M. Prest}, Mem. Am. Math. Soc. 987, v, 109 p. (2011; Zbl 1229.03034)]. The paper proceeds by study of its relations with other, better established notions. \par The general outline of main results about definable categories is given in author's abstract: \par ``Definable subcategories are precisely the finite-injectivity classes. We prove a 2-duality between the 2-category of small exact categories and the 2-category of definable categories, and provide a new proof of its additive version. We further introduce a third vertex of the 2-category of regular toposes and show that the diagram of 2-(anti)-equivalences between three 2-categories commutes; the corresponding additive triangle is well-known.'' \par As a whole, the paper a very detailed presentation of the results mentioned above including proofs (if not proved elsewhere). \par The relationship with regular logic (i.e., the $(\top,\wedge, \exists)$-fragment of first-order logic) is important. The following 2-equivalence between 2-categories is established (theorem 3.2.5): \[Reg(-, Set): \mathbb{EX}\leftrightarrows \mathbb{DEF}: (-, Set)^{\Pi\rightarrow}\] Here the catedory $\mathbb{DEF}$ has as its objects definable categories, as 1-morphisms interpretation functors and as 2-morphisms natural transformations, and $\mathbb{EX}$ is the category of small (Barr-)exact categories, regular functors and natural transformations. According to the authors it is one of the main contributions of the paper. \par In additive case (Section 4), the new proofs of similar duality results are obtained via more general technique developed using the new notion of definability. \par In Section 5 (the last technical section of the paper) the relationship with regular toposes is studied. ``We... complete the picture of 2-(anti)-equivalences between three 2-categories.'' The connection with regular logic and regular theories is discussed in some detail. \par Future directions are outlined in Section 6. For example (the following is the most detailed proposal): \par ``It is natural to ask whether the syntax-semantics duality for regular logic is entirely `logical' in nature in the sense that it does not depend on the base category. More specifically, it would be useful to replace Set or Ab by a sufficiently nice monoidal category $\mathcal{V}$, and a functor by a $\mathcal{V}$-enriched functor... we may ask for $\mathcal{V}$ to be locally finitely presentable, exact, and symmetric monoidal closed. Such class of monoidal categories clearly includes Set, Ab, presheaf toposes and module categories.''