# Set theoretic issue of localization of abelian categories

For a small abelian category in which every object is also a set, consider its localization with respect to a Serre subcategory (thus a quotient category), is it true that under this localization functor $T$, the image of an object is still a set? I wonder if it is always possible to turn this functor $T$ into an endofunctor $F$ such that its image is equivalent to the localized category, as it is considered in triangulated categories where a localization functor is defined as an endofunctor together with a natural transformation satisfying certain properties. In this sense, the first question seems to be resolved naturally.

• It sounds to me like the assumption that "every object is also a set" is misguided. What would be an example of a small abelian category in which not every object is a set? The very notion of "small" presupposes set theory, and so the objects of a small category are automatically sets. If you are considering a set theory with urelements then we can get rid of them by moving to an isomorphic category. – Andrej Bauer Oct 14 '15 at 20:22
• Is it possible to have an (essentially) small category in which each object is not a set? – user81489 Oct 14 '15 at 21:05
• Let us suppose we are working in Zermelo-Fraenkel set theory without urelements. Then everything is a set. I do not even understand why you worry about the set-theoretic nature of the objects. Why do you even care whether the objects are sets, bananas or coconuts? – Andrej Bauer Oct 14 '15 at 21:33
• My problem focuses on the localized category, and I wonder when applied the localization functor $T$ the image $T(M)$ of an object $M$ (which is a set) remains a set or not. In the category of finitely generated abelian groups, the localization with respect to the Serre subcategory generated by $\mathbb{Z}/(2)$ say corresponds to the functor given by tensoring with $\mathbb{Z}_{(2)}$. In particular, the image of every abelian group is still a set, the elements are given by equivalence classes of fractions. So I wonder if it is true in general that $T(M)$ is a set for arbitrary localization. – user81489 Oct 14 '15 at 21:46
• Do you mean rather that every object of the category has an underlying set? – David Roberts Oct 15 '15 at 1:40

Perhaps @user81489 means the following. Suppose given a concretizable category (note: every small category $\mathcal C$ is concretizable, by the functor $\prod_{C\in \mathcal C} \hom(C,-)$; see also this post on SBS.). Is its localization with respect to a Serre subcategory also concretizable?
• Yes. A morphism in $\mathcal C[\mathcal W^{-1}]$ is an equivalence class of zig-zag paths in $\mathcal C$ whose "zag" components are in $\mathcal W$. If $\mathcal C$ is small, there are only a set of such zig-zags, and hence only a set of equivalence classes. – Theo Johnson-Freyd Oct 15 '15 at 17:04