Core algebra revisited (Q879364)

Motivated by \textit{John Reynolds}' work in parametric polymorphism, the author first defines the notion of core object in any category with finite products, as an object \(C\) together with a natural transformation \((-) \times C \to (-)\) which is universal among such, in the sense that any other natural transformation \((-) \times A \to (-)\) is induced by a unique morphism \(A \to C\). Examples in the category of \(M\)-sets, \(M\) being various types of monoids, are given in details. From this definition, the author then extracts an equational description of what he calls core algebras, which, in his own words, can be seen as ``something of a generalization of commutative monoids''. He shows that his equations are complete, in the sense that any equation that holds for all categorical cores is a consequence of them. This follows from the fact that any core algebra is a subalgebra of the core of some category. The completeness of the chosen operators is proved in the author's recent paper [\textit{P. Freyd}, Theory Appl. Categ. 18, 303--320 (2007; Zbl 1139.18001)]. It is also shown that every Grothendieck topos has a core, but not every elementary topos. Using an appropriate example of a functor that does not preserve core objects, the author concludes that ``the core [\dots] requires more than the internal logic of a topos in its definition. What is required is polymorphism''.