7
$\begingroup$

Let $X$ be a projective, smooth variety over $\mathbb{C}$, and let $D$ be a irreducible, big Cartier divisor(notice, I do not assume nefness). Then is it true that $${\rm{H}}^1(X, K_X + D) = 0\quad?$$

On the one hand, I doubt it because the result seems to be written nowhere; on the other hand, I feel the conditions here are much nicer than many vanishing theorem, and I was wondering if some modification would reduce it to one of the vanishing theorem.

Besides, any references/results/comments on vanishing theorem for big divisors are great welcome!

$\endgroup$

1 Answer 1

6
$\begingroup$

Let $X\to \mathbb P ^1$ be a family of smooth quadric surfaces degenerating to $X_0=\mathbb F _2=\mathbb P (\mathcal O _{\mathbb P ^1}\oplus \mathcal O _{\mathbb P ^1} (2))$ where $0\in \mathbb P ^1$. We denote by $E$ the $-2$ curve in $X_0$, $F$ the fiber of $X_0\to \mathbb P ^1$ and by $f:X\to Z$ the corresponding small contraction (following the notation of arXiv:0901.0389 remark 4.4 and remark 4.9). For $b>>0$ consider $L$ a line bundle whose restriction to the general fiber is $\mathcal O _{\mathbb P ^1} (b+1)\boxtimes \mathcal O _{\mathbb P ^1} (b-1)$ and whose restriction to the special fiber is $\mathcal O _{X_0}(b(E+2F)+E)$ and $D\in |L|$ a general member.

We claim that $R^1f_*\omega _X(D)\ne 0$. To see this consider the short exact sequence $0\to \omega _X(D-X_0)\to \omega _X(D)\to \omega _{X}(D)|_{X_0}\to 0$. Since $R^if_*\omega _X(D-X_0)\cong R^if_*\omega _X(D)\otimes \mathcal O_{Z}(-Z_0)$ (by the projection formula), then it follows that $R^1f_* \omega _X(D)\to R^1f_*\omega _{X}(D)|_{X_0}\cong R^1f_* \omega _{X_0}(D|_{X_0})$ is surjective. Consider now the short exact sequence $0\to \omega _{X_0}\to \omega _{X_0}(D|_{X_0})\to \omega _{D|_{X_0}}\to 0$. Since $R^1f_*\omega _{X_0}=0$ (by Grauert-Riemanschneider), it follows that $R^1f_* \omega _{X_0}(D|_{X_0})\to R^1f_*\omega _{D|_{X_0}}$ is surjective. Since $|L|_{X_0}|=|b(E+2F)|+E$ where $E+2F$ is the pull back of an ample divisor $H$ on $X_0$, we have $R^1f_*\omega _{D|_{X_0}}\cong R^1f_*\omega _E\otimes \mathcal O _{Z_0}(bH)$. Since $H^1(\omega _E)\cong H^0(\mathcal O _E)^\vee\ne 0$, we have $R^1f_*\omega _{D|_{X_0}}\ne 0$ and the claim follows.

It should be easy to see that $D$ is big, it is not nef because $D\cdot E =E^2=-2$, moreover we can arrange that $H^1(f_*\omega _X(D))=0$ (if necessary replace $D$ by $D$ plus the pull-back of a sufficiently ample divisor on $Z$ and apply Serre vanishing). Then, by the Leray spectral sequence $H^1(\omega _X (D))\cong H^0(R^1f_*\omega _X (D))\ne 0$.

Maybe there are easier examples, but this is the first one that came to mind. I do not think there can be surface examples, because if $D$ is irreducible and not nef, then $D^2<0$, but as $D$ is big, $D\sim _{\mathbb Q}D'\neq D$ subtracting common components of $D$ we obtain $\lambda D\sim _{\mathbb Q}D''$ with $0<\lambda$ and the support of $D''$ does not contain $D$. But then $D^2=\frac 1 {\lambda ^2} D\cdot D''\geq 0$ a contradiction.

$\endgroup$
1
  • $\begingroup$ Thank you for your terrific answer!! I should look the above mentioned paper in more details! $\endgroup$
    – Li Yutong
    Feb 4, 2015 at 1:18

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.