Let $\mathscr{A}$ be an abelian category, and $S\subset\mathscr{A}$ some set of objects in $\mathscr{A}$ (i.e. $S$ is not a proper class), so we can fairly easily define the category $\mathscr{C}_S$ to be the smallest full subcategory of $\mathscr{A}$ containing $S$ such that $\mathscr{C}_S$ is closed under subobjects, quotients, and extensions, making $\mathscr{C}_S$ the minimal Serre subcategory of $\mathscr{A}$ containing $S$.

But will $\mathscr{C}_S$ be a small category?

The reason I'm asking this is that it is often convenient, when working with Abelian categories, to consider some set $S\subset\mathscr{A}$, and then construct some small Abelian full exact subcategory $\mathscr{C}$ of $\mathscr{A}$ containing $S$ and applying the Freyd-Mitchell embedding theorem to introduce elements (radically simplifying diagram chases).


Not in general, even if by "small" you mean "essentially small" (i.e., there is a small set of isomorphism classes of objects). First of all, it is possible for a single object to have a proper class of subobjects (if this does not happen, the category $\mathscr{A}$ is called well-powered). I don't know an explicit example off the top of my head, but you should be able to build one by taking a category with a proper class of subobjets and then "freely" making it abelian.

It is also not enough just to assume your category is well-powered. It is possible in a well-powered abelian category to have two objects $A$ and $B$ such that there is a proper class of non-isomorphic extensions of $B$ by $A$ (see my answer https://math.stackexchange.com/a/1766337/86856 for an explicit example).

On the other hand, if you know that your category is well-powered and that there is only a set of extensions of any two objects (up to isomorphism), then it is easy to see that $\mathscr{C}_S$ is essentially small. Indeed, define $S_0=S$ and let $S_{n+1}$ be the collection of all subobjects, quotients, and extensions of objects of $S_n$. Since $S_n$ is essentially small and there is (up to isomorphism) only a set of subobjects/quotients of each object of $S_n$ and only a set of extensions of each pair of objects of $S_n$, $S_{n+1}$ is also essentially small. The union $\bigcup_nS_n$ is then essentially small and is closed under subobjects, quotients, and extensions, and so is equal to $\mathscr{C}_S$.

(If you want a category that is literally small and not just essentially small, you can take $S_{n+1}$ to just have one representative of each isomorphism class of subobject/quotient/extension, and then each $S_n$ will be a set and $\bigcup_n S_n$ will be a set of objects which contains a representative of every isomorphism class in $\mathscr{C}_S$.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.