Exact categories of topological vector spaces with linear topology (Q6567969)

The author of this thorough and systematic article starts the introduction to it with the warning that ``topological algebra is treacherous''. He goes on to clarify why this is the case when one studies topological abelian groups and topological vector spaces. A fundamental manifestation of his claim is that quotients of complete, separated abelian groups by closed subgroups need not be complete. To this end, the author, after giving a sufficient condition for such a quotient to be complete (Proposition 1.4), he shows that any separated topological abelian group, respectively \(k\)-vector space, is the quotient (endowed with the quotient topology) of a complete such by a closed subgroup, respectively subspace (Theorems 2.5, 3.1). The result, in greater generality, is attributed by the author to [\textit{V. I. Arnautov} et al., Introduction to the theory of topological rings and modules. Basel: Marcel Dekker (1996; Zbl 0842.16001)].\N\NA certain tension, present in the categorical treatment of other kinds of structure, occurs in the subject of topological abelian groups or vector spaces. Good structures may fail to have nice categorical properties, while less convenient structures may organize in categories with good properties. The author proceeds to show (Theorems 7.1, 7.2) that the categories of (not necessarily complete) abelian groups and separated abelian groups with a linear topology, as well as those of (not necessarily complete) \(k\)-vector spaces and separated abelian groups with a linear topology, form a quasi-abelian category (i.e an additive one with kernels and cokernels, were the former are stable under pushout and the latter under pullback. Put in the terminology of general category theory, a finitely complete and co-complete additive category which is both co-regular and regular.) Kernels and cokernels arise in the desired way: they are, respectively, injective maps with the induced topology and surjective maps with the quotient topology. On the other hand (Proposition 8.3) the categories of complete separated abelian groups or \(k\)-vector spaces with linear topology are left quasi-abelian (kernels are stable under pushout). (Reviewer's remark: We would like to point out here that, in the non-additive world, the categories of topological groups and Hausdorff topological groups are known to be regular, see [\textit{A. Carboni} et al., Appl. Categ. Struct. 1, No. 4, 385--421 (1993; Zbl 0799.18002)])\N\NHe deals further with the embedding of complete separated abelian groups and \(k\)-vector spaces into the category of pro-objects of ablian groups, respectively \(k\)-vector spaces, identifying their full images as the subcategories of limit-epimorphic objects, i.e such that the morphisms from the limit of the inverse directed system, representing the pro-object to the components, are epimorphic (Theorem 9.4). These embeddings though do not have the desired properties, as the respective categories do not inherit an exact structure from the abelian category of pro-objects.\N\NThe last two sections of the paper present an account of properties of three tensor product structures that can be defined on the categories of topological \(k\)-vector spaces and separated such spaces. These are three different linear topology structures on the ordinary tensor product of the vector spaces. They were introduced in [\textit{A. A. Beilinson}, Mosc. Math. J. 8, No. 1, 1--20 (2008; Zbl 1170.14002)]. As bifunctors, on either category, they preserve kernels and cokernels on each argument, and fixing an argument, they become exact endofunctors of the respective categories with the quasi-abelian structure. The completed versions of these tensor products, as endofunctors of the respective categories of complete separated structures do not have these nice properties, at least not for the maximal exact category structure on them.