2
$\begingroup$

The problem is the following:

Let $k$ be a field and let $N,G,Q: (\operatorname{Sch}/k)_{\operatorname{fppf}} \longrightarrow \operatorname{Grps}$ be fppf-sheaves from the big fppf-site of $\operatorname{Sch}/k$ to the category of groups. Assume that we have an exact sequence of fppf-sheaves $$ e\longrightarrow N \longrightarrow G \longrightarrow Q \longrightarrow e $$ and that $N$ and $Q$ are both algebraic group schemes. Then $G$ is representable by an algebraic group scheme.

I know that after some fppf base change $Q'\longrightarrow Q$ we have an equality $$ G\times_Q Q' \cong N\times_k Q' $$ and so it becomes representable. Moreover, if $N$ is affine I can use fppf-descent obtaining the desired scheme. Nevertheless I am not sure how to deal with the general case.

This is an exercise of Milne's book on Algebraic Groups (Ex 5-10).

$\endgroup$

1 Answer 1

3
$\begingroup$

Here is a possible road to a solution (I am fairly sure that Milne had something more elementary in mind). Algebraic spaces satisfy fppf descent; hence $G$ is a group object is in the category of algebraic spaces of finite type over $k$. But it is well known that a group object is in the category of finitely generated algebraic spaces over a field is in fact a group scheme. This seems to be "well known", although I have never seen a complete proof. Over an algebraically closed field this follows immediately from the result of Grothendieck that every algebraic space contains a dense open subscheme. In the general case this requires a little more care.

$\endgroup$
1
  • 3
    $\begingroup$ You can complete the argument using my answer to this question: mathoverflow.net/q/194326 plus the fact that algebraic group schemes over a field are quasiprojective. $\endgroup$ Sep 6, 2019 at 18:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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