How nice are free completions of categories? (Q2310781)

The authors study completions under small coproducts and small colimits of categories that need not be necessarily small. As a completion of a category \(\mathcal{K}\) under small coproducts (respectively, small colimits) is meant a category \(\Sigma \mathcal{K}\) (resp. \(P\mathcal{K}\)) with small coproducts (resp. small colimits) and a functor \(\eta\) from \(\mathcal{K}\) to the latter such that, for any functor \(F\) from \(\mathcal{K}\) to a category \(\mathcal{E}\) with small coproducts (resp. small colimits), there exists a unique up to natural isomorphism functor \(F^*\) from \(\Sigma \mathcal{K}\) (resp. \(P\mathcal{K}\)) to \(\mathcal{E},\) which preserves small coproducts (resp. small colimits) and such that \(F^* \eta \) is naturally isomorphic to \(F.\) These properties describe the respective completions uniquely up to equivalence of categories. The completions can be described as the small sums (respectively, small colimits) of representable functors within the functor category \([\mathcal{K}^{op}, \mathrm{Set}].\) These completions are well understood in the case \(\mathcal{K}\) is small and many of their categorical properties are known. For example the completion under small colimits of a small category is equivalent the category of presheaves on it, hence it is a topos, in particular it is (locally) cartesian closed. The situation is not as well understood in the case \(\mathcal{K}\) is not small. The authors make a number of significant contributions as to whether, and under what assumptions, several categorical properties are enjoyed by such completions in this case. In what concerns the completion under small sums, the authors characterize when \(\Sigma \mathcal{K}\) is extensive and, under the assumption that \(\mathcal{K}\) has finite products, they characterize when is \(\Sigma \mathcal{K}\) cartesian closed and show as a corollary that \(\Sigma \mathcal{K}\) is cartesian closed for a complete additive \(\mathcal{K}.\) Under the assumption that \(\mathcal{K}\) is complete, they characterize when is \(\Sigma \mathcal{K}\) locally cartesian closed. In what concerns the completion under small colimits, the authors characterize when it is complete -their condition though is just a restatement of the characterization given in [\textit{B. J. Day} and \textit{S. Lack}, J. Pure Appl. Algebra 210, No. 3, 651--663 (2007; Zbl 1120.18001)], Theorem 3.8, -- while they provide characterizations for when is it Cartesian closed (provided \(\mathcal{K}\) has finite products) and show that this is equivalent to being locally cartesian closed, provided \(\mathcal{K}\) is complete. They show that, under mild assumptions about cones of finite diagrams in \(\mathcal{K},\) \(P\mathcal{K}\) is a pretopos but (again under some mild assumptions on \(\mathcal{K}\)) \(P\mathcal{K}\) is an elementary topos iff \(\mathcal{K}\) is essentially small. Then they move to the question whether \(P\mathcal{K}\) is well-powered or co-well-powered. They show that the latter property implies the former and, again under mild assumptions on \(\mathcal{K},\) vice-versa. They offer some sufficient conditions for \(P\mathcal{K}\) being co-well-powered as well as some negative result and several examples in that direction.