Let $f:X\to Y$ be a representable map of finite type (or is finite dimensional enough?) Artin stacks, whose fibres (which are schemes) have dimension at most $n$. Then is it true that $R^qf_*\mathbf{Q}_\ell=0$ for all $q\gg 0$?

Note: by taking atlases, I think it is sufficient to let $X,Y$ be schemes.

Edit: Will Sawin pointed out that the question as stated was obviously false, I've edited it to remove that false statement.


$Y$ admits a smooth surjective morphism from a scheme $Z$. Because smooth morphisms are locally of finite type, $Z \to Y$ is locally of finite type, and you can choose an open cover that covers $Y$ and then pass to a finite subcover to make $Z$ of finite type.

Because this morphism is smooth, by smooth base change the pullback of $R^q f_* \mathbf Q_\ell$ to $Z$ is the pushforward of $\mathbb Q_\ell$ from $Z \times_Y X$ to $Z$. Because this morphism is smooth, it suffices to prove a bound for this pushforward.

If $Z \to Y$ is a schematic morphism (this might be a little stronger than the fibers being schemes) then $Z \times_Y X$ is a scheme, also of finite type. Boundedness then follows from classical results - mod $\ell$ cohomology is a limit over the cohomology of neighborhoods, and these are finite type of bounded dimension.

  • $\begingroup$ Whoops, I must have not been thinking when I wrote the question as it was. I've edited it now, hopefully it is less trivial (or perhaps even correct). $\endgroup$
    – Meow
    Jun 18 at 23:42
  • 2
    $\begingroup$ @Meow Isn't the disjoint union of $\mathbb A^m$ for all $m$ an Artin stack? You need some kind of finiteness assumption on $X$. $\endgroup$
    – Will Sawin
    Jun 19 at 1:19
  • $\begingroup$ Right, I should have included a finite type assumption, which has been edited in now. $\endgroup$
    – Meow
    Jun 19 at 9:07
  • $\begingroup$ @Meow See edits. $\endgroup$
    – Will Sawin
    Jun 19 at 14:41
  • $\begingroup$ Thanks for being accomodating of my shifting the goalposts! $\endgroup$
    – Meow
    Jun 19 at 23:52

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.