Topical categories of domains (Q4719352)

This is an important paper, which deserves to be carefully studied both by topos-theorists and by theoretical computer scientists. The author's main theme is a new variation on the old song that toposes are generalized spaces: he considers toposes as \textit{topology-free} spaces, by which he means not that they have no topology, but rather that the topology is so thoroughly `integrated' into the definition of a topos that one does not have to consider it --- all the constructions one can conceivably talk about are automatically continuous. (Of course this is an oversimplification, and the author is well aware that the truth is not quite so simple. But it makes a good slogan.) To illustrate the theme, recall the familiar idea that the classifying topos \([T]\) of a geometric theory \(T\) may be considered as `the space of all \(T\)-models' in a natural way. If one has two geometric theories \(T\) and \(T'\), and a `geometric' way of constructing a \(T'\)-model from a \(T\)-model, this will give rise to a geometric morphism \([T]\to[T']\); in other words, the construction is automatically `continuous' with respect to the `topologies' imposed on the two generalized spaces. Turning to the mathematical content of the present paper: by a `topical category', the author means an internal category in the \(2\)-category \textbf{Top} of (Grothendieck) toposes and geometric morphisms. Because \textbf{Top} has (pullbacks and) cotensors with \textbf{2}, any topos gives rise to an `intrinsic' topical category; but these intrinsic categories are often less interesting than one might expect. For example, the intrinsic topical category of sets (corresponding to the object classifier) is not cartesian closed, because exponentiation is not a geometric construction (in particular, it is not continuous). However, not all topical categories are constructed this way. The author shows how one can construct a topical category of strongly algebraic domains (that is, one whose topos of objects is the classifying topos for strongly algebraic domains), and shows that it is cartesian closed as a topical category. This means that any constructions one can perform on strongly algebraic domains, making use of function-spaces as well as the more obviously `geometric' constructions, have the automatic continuity property referred to above. Although the paper is a long one, its length derives from the author's insistence on clear exposition of what are in actuality quite sophisticated ideas. It is a paper which will repay reading by beginners, who will find it an accessible introduction to the \(2\)-categorical aspects of topos theory and geometric logic, as well as by experts who will enjoy a fascinating new view of their subject.