7
$\begingroup$

Given a morphism of algebraic stacks $f: X \to Y$ (possibly non-representable) what sufficient conditions will guarantee that the derived push-forward of the structure sheaf on $X$ is isomorphic to the structure sheaf on $Y$? I am looking for a stacky version of a result of Buch-Mihalcea (Thm 3.1 of Quantum K-theory of Grassmannians https://arxiv.org/abs/0810.0981, which itself is a variation of a result of Kollar)

Thm 3.1 of BM: Let $f : X \rightarrow Y$ be a surjective map of projective varieties with rational singularities. Assume that the general fiber of $f$ is rational, i.e. $f^{−1}(y)$ is an irreducible rational variety for all closed points in a dense open subset of $Y$. Then $f_∗[\mathcal{O}_X]=[\mathcal{O}_Y]$ in the K-theory of Y.

I am happy to assume, in any proposed generalization, that the stacks are proper DM stacks (even smooth) and that the general fiber is an irreducible rational variety.

$\endgroup$
  • 2
    $\begingroup$ The stack structure on $Y$ is irrelevant: if $Y$ is a smooth DM stack, then there exists a smooth surjective morphism from a smooth scheme to $Y$. You can base change your original morphism by this smooth cover. So assume that $Y$ is a scheme. Then the morphism $f$ factors through the coarse moduli space of $X$, $p:X\to |X|$. In characteristic $0$ (or assuming tame), $Rp_*\mathcal{O}_X$ is the structure sheaf of $|X|$. Then, if $|f| : |X|\to Y$ satisfies the hypotheses of Buch-Mihalcea, then $[Rf_*\mathcal{O}_X]$ equals $[R|f|_*\mathcal{O}_{|X|}]$, and this equals $[\mathcal{O}_Y]$. $\endgroup$ – Jason Starr Jan 15 '17 at 10:48
  • $\begingroup$ The example I wrote was wrong! I will try to find a correct example, e.g., a quartic surface with a worse singularity. $\endgroup$ – Jason Starr Jan 15 '17 at 19:33
  • $\begingroup$ Oops, there are no counterexamples . . . posting answer below. $\endgroup$ – Jason Starr Jan 15 '17 at 20:00
3
$\begingroup$

I am amplifying the above comments. The theorem of Buch-Mihalcea also holds for stacks.

Let $k$ be a field of characteristic $0$. Let $Y$ be a Deligne-Mumford stack that is finite type over $k$ and that is normal: there exists a surjective, étale morphism $h:\widetilde{Y}\to Y$ with $\widetilde{Y}$ a normal scheme. Let $f:X\to Y$ be a proper, surjective morphism of Deligne-Mumford stacks. Assume that $X$ is integral, and assume that $X$ has rational singularities: there exists a surjective, étale morphism $g:\widetilde{X}\to X$ such that $\widetilde{X}$ is a $k$-scheme that has rational singularities. Finally, assume that the geometric generic fiber of $f$ is isomorphic (as a stack) to a smooth proper variety that is rationally connected, or even just $\mathcal{O}$-acyclic.

Proposition. The natural map $\mathcal{O}_Y\to f_*\mathcal{O}_X$ is an isomorphism, and $R^q f_*\mathcal{O}_X$ equals $0$ for every $q>0$.

Because $h$ is flat, $h^*R^qf_*\mathcal{O}_X$ equals $R^q \widetilde{f} \mathcal{O}_{X\times_Y \widetilde{Y}}$, where $\widetilde{f}:X\times_Y \widetilde{Y} \to \widetilde{Y}$ is the projection. Thus, without loss of generality, assume that $Y$ is a normal scheme. As in the comment, there exists a coarse moduli space $p:X \to |X|$. The morphism $f$ factors as the composition of $p$ and a surjective, proper $k$-morphism, $$|f|:|X|\to Y.$$ Because the characteristic is $0$, $X$ is a tame stack. Thus, $\mathcal{O}_{|X|} \to p_*\mathcal{O}_X$ is an isomorphism and $R^qp_*\mathcal{O}_X$ equals $0$ for all $q>0$. Therefore, it suffices to prove that $\mathcal{O}_Y\to |f|_*\mathcal{O}_{|X|}$ is an isomorphism and $R^q|f|_*\mathcal{O}_{|X|}$ equals $0$ for all $q>0$. Via Buch-Mihalcea, it suffices to prove that $|X|$ has rational singularities. This follows from Proposition 5.13, p. 157 of "Birational Geometry of Algebraic Varieties" by Kollár and Mori.

$\endgroup$
  • $\begingroup$ Jason Starr, That's excellent, thanks! Did you drop the projectivity hypotheses in BM on purpose, or was that an oversight? $\endgroup$ – Chris Woodward Jan 15 '17 at 22:09
  • $\begingroup$ Because of Chow's lemma and existence of a projective resolution of singularities (in characteristic $0$), the version with $|f|$ proper follows from the version with $|f|$ projective. $\endgroup$ – Jason Starr Jan 15 '17 at 22:35
  • $\begingroup$ To be clear, I think Jason also meant to require that Y has rational singularities (as in the original question). $\endgroup$ – Nick Addington Jul 23 '19 at 23:30

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.