Let $C$ be a smooth projective curve and $E$ a vector bundle of rank $r$ on $C$. We say that $E$ is nef/ample if $\mathcal{O}_{\mathbb{P}(E)}(1)$ is so. Equivalently (see Hartshorne's papers on 'Ample vector bundles' and 'Ample vector bundles on curves'), $E$ is ample if and only if for any coherent $F$, $S^m(E)\otimes F$ is globally generated for all $m\geq n_0$.

The statement I'm slightly stuck on is the following comment in a paper of Fujita 'On Kahler fibre spaces over curves': If $C$ has genus $g\geq2$ and $H^1(C, E)=0$ for some vector bundle $E$, then $E$ is ample.

This follows easily in the rank $1$ case from Riemann-Roch. I suspect the general case will also be easy but I have been through Lazarsfeld's book and the standard references with no luck so far. Any help appreciated!

  • $\begingroup$ Riemann-Roch tells you that $\mu(E) \geq g - 1$. Does it help? $\endgroup$ – Piotr Achinger Dec 16 '10 at 13:30
  • 2
    $\begingroup$ Have you forgotten to mention the condition on $H^1(C,E)$? $\endgroup$ – Rex Dec 16 '10 at 13:46

Here's a partial answer. Suppose we're in characteristic 0 (Fujita would be assuming this), and that $rank(E)=2$. By cor 7.6 of Hartshorne's ample vector bundles paper, it suffices to check that $deg(E)>0$ and $deg(L)>0$ for ay quotient line bundle. From Riemann-Roch as in Piotr's comment, we get $$deg(E) + rank(E)(1-g) = h^0(E)\ge 0$$ which implies positivity of $deg(E)$. On a curve $H^1(E)=0$ implies the vanishing for any quotient bundle, and so in particular for $L$. Combing this with the above argument, gives $deg(L)>0$.

I think this can be pushed, but I'd better back to the less fun things that I'm supposed to be doing now.

  • $\begingroup$ Thank you! Your comment about $H^1$ of a quotient bundle also being trivial is spot on, as from Theorem 2.4 from Hartshorne's 'Ample vector bundles on curves', (over $\mathbb{C}$) $E$ is ample if and only if every quotient bundle has positive degree, and by Riemann-Roch (or at least the vector bundle equivalent $\chi(E) = \deg(E) + rk(E)\chi(\mathcal{O}_X)$) it follows that if $E\to E'\to0$, then $\deg(E')>0$. If you'd like to incorporate this in your answer for completeness I'd be glad to accept it. $\endgroup$ – Frank Dec 16 '10 at 16:23
  • $\begingroup$ I didn't see this comment before I posted my answer, but it is essentially the same argument. $\endgroup$ – J.C. Ottem Dec 16 '10 at 16:45
  • $\begingroup$ Frank & J.C.: Good work! $\endgroup$ – Donu Arapura Dec 17 '10 at 14:10

This builds on the idea of Donu Arapura: We use the following criterion of Hartshorne (Thm. 2.4 in this article): $E$ is ample iff $Q$ has positive degree on $C$ for any quotient bundle $Q$ of $E$. Let $r=rank(E)$ and $s=rank(Q)$. Here

$$deg(Q)=s(g-1)+\chi(Q)> \chi(Q)$$

so it suffices to show that $\chi(Q)\ge 0$. But this is elementary, since $H^1(E)$ implies $H^1(Q)=0$ by the exact sequence $$ 0\to L \to E\to Q \to 0. $$

  • $\begingroup$ Thank you, just posted this as a comment to Donu's answer. $\endgroup$ – Frank Dec 16 '10 at 16:44

It follows from Gieseker's Criterion- we have to show that E can't have t any trivial quotient. Use the Cohomology seq for 0 -> K -> E -> O_C -> 0


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.