Introduction: There are certain categories of group schemes or group varieties in which one can define exact sequences. For instance, the categories of commutative group schemes of finite type over a field $k$ and of commutative affine group schemes over $k$ are abelian (see for instance here). So let's say that we have a short exact sequence $$ 0 \to A \to B \to C \to 0$$ in such a suitable category (bonus question: examples of abelian categories of group varieties?). There is, a priori, no reason why the sequence should be split. Yet people call such a $B$ (together with the maps) fitting in the above exact sequence an extension of $C$ by $A$. Note that this is true also for other categories and does not have to do with algebraic geometry specifically, but I am most interested in this setting.

My question: What kind of information can one deduce, in general, about $B$, knowing only $A$ and $C$? Does one know something about the underlying topological space of $B$ or of $B(k)$ up to homotopy, or even up to homeomorphism? Is it any better if we consider maps up to isogeny? Feel free to restrict to some specific category where there are better results.

I am asking this because the terminology seems to suggest, and it seems to me that people often assume, that $B$ is some kind of direct sum between $A$ and $C$, even though there is a priori no reason for this to be the case. The isomorphism theorem does imply that $B/A \cong C$, but this does not imply that we know how to obtain $B$ knowing $A$ and $C$, since there might be many possibilities for doing so.

Approximate paraphrase 1: A paraphrase of my question is: what do all possible $B$'s that fit in the above sequence have in common, topologically?

Approximate paraphrase 2: Is there some exact functor $F$ from some "geometric" category $\mathcal{I}$ to some "topological" category $\mathcal{J}$ such that in $\mathcal{J}$ exact sequences are always split or $F(B)$ is at least somehow isomorphic to $F(A) \oplus F(C)$ or something similar?

Note: I am aware of the fact that, at least for modules, the set of the $B$'s fitting in the above exact sequence, up to some equivalence and with some sum operation, is isomorphic to $Ext^1(C,A)$, but I think that the term extension gave rise to the notation $Ext$ rather than the converse, so this fact does not provide any useful insight.


Your Answer

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

Browse other questions tagged or ask your own question.