4
$\begingroup$

A theorem by Kollár asserts that if $X$ and $Y$ are projective varieties with $X$ smooth, and $f : X \to Y$ is a surjective map, then the higher direct images $R^if_*\omega_X$ vanish for $i$ greater than the generic fiber dimension. I'd like to know if one can weaken the assumption that $X$ is smooth, for example, to $X$ having quotient singularities, or something even weaker (whereby the sheaf $\omega_X$ represents an appropriate dualizing sheaf in such a case).

Kollár's original proof uses the fact that $X$ is smooth to prove that the higher direct images $R^if_*\omega_X$ are torsion-free which I suspect may fail for $X$ having singularities.

Improve this question
$\endgroup$
3
  • $\begingroup$ There are definitely situations where the vanishing still holds if $X$ and $Y$ are not smooth. An easy example is when $f$ is flat. But maybe you are interested only in assumptions on $X$ that make this true for every $f \colon X \to Y$ with $Y$ smooth projective (say)? $\endgroup$
    – R. van Dobben de Bruyn
    Feb 7, 2018 at 0:47
  • $\begingroup$ Thanks, Remy. For my purposes I'd really like to take $Y$ to be singular, and let $X$ be a partial resolution of the singularities of $Y$, so that in particular, $f$ is a birational map, but $X$ may still admit singularities itself. $\endgroup$
    – Mitchell Faulk
    Feb 7, 2018 at 1:01
  • $\begingroup$ Good job on correcting the spelling! ;-) $\endgroup$
    – Sándor Kovács
    Feb 9, 2018 at 17:13

1 Answer 1

Reset to default
5
$\begingroup$

If $X$ has rational singularities, then this vanishing (and the torsion-freeness as well) follows almost trivially. Let $g:Z\to X$ be a resolution of singularities. Then $Rg_*\omega_Z\simeq \omega_X$ and hence $R^if_*\omega_X\simeq R^i(f\circ g)_*\omega_Z=0$ for $i>\dim X-\dim Y$ by Kollár's theorem (I suppose that's what you meant, not what you wrote, right?! ;-)

Quotient singularities are rational, so this includes for what you are hoping. Also, since rational singularities are Cohen-Macaulay there is indeed a dualizing sheaf and it plays nice.

p.s.: If I remember correctly, Kollár also mentions this in his paper(s).

Improve this answer
$\endgroup$
2
  • $\begingroup$ Also, sorry for the follow-up question, but your parenthetical remark suggests I may have formulated my original statement incorrectly? I was trying to follow as closely as possible the formulation written by Koll\'{a}r in Theorem 2.1(ii) of ``Higher Direct Images of Dualizing Sheaves I.'' $\endgroup$
    – Mitchell Faulk
    Feb 7, 2018 at 16:03
  • 4
    $\begingroup$ Your formulation was fine. I just meant that Kollar$\neq$Kollár. :) $\endgroup$
    – Sándor Kovács
    Feb 7, 2018 at 20:42

Your Answer

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

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