2
$\begingroup$

Most examples of ample line bundles that are not globally generated have less number of global sections than the dimension of the variety. Assuming ampleness, is the existence of "enough" global sections sufficient to guarantee globally generated-ness? More precisely:

Let $X$ be a smooth, projective $\mathbb{C}$-variety of dimension $n \ge 2$ and $L$ an ample invertible sheaf on $X$ with $h^0(L) \ge n+1$. Is $L$ globally generated?

$\endgroup$
2
  • $\begingroup$ I guess you are assuming the variety is connected. $\endgroup$
    – user127776
    Commented Apr 1, 2022 at 12:17
  • $\begingroup$ @user127776 Yes, the variety is connected. Moreover as it is also assumed to be smooth, it is irreducible as well. $\endgroup$
    – user43198
    Commented Apr 1, 2022 at 12:19

1 Answer 1

6
$\begingroup$

The answer is no for curves, and then a product construction show that the answer is no in every dimension.

Take a hyperelliptic curve $X$ of genus $g \geq 3$ and let $P$ be a 2-torsion line bundle on $X$, such that $P = \mathcal{O}_C(p-q)$, where $p$, $q$ are distinct Weierstrass points. Now set $L = \omega_C \otimes P$.

Since $L$ has strictly positive degree on a curve, it is ample. Moreover, by Serre duality, $h^1(C, \, L)=h^0(C, \, P)=0$, hence $$h^0(C, \, L)=g-1 >2$$ by Riemann-Roch. On the other hand, we also have $$h^1(C, \, L (-p))=h^1(C, \, \omega_C (-q)) = h^0(C, \, \mathcal{O}_C(q))=1,$$ hence $$h^0(C, \, L(-p))=1+(1-g)+ (2g-3)=g-1.$$ But then $|L|=|L-p|$, in other words, $p$ is in the base locus of $L$ and so $L$ is not globally generated.

Finally, take two copies $(C_1, \, L_1)$, $(C_2, \, L_2)$ of the polarized pair $(C, \, L)$ and consider the line bundle on $C_1 \times C_2$ given by $$\mathscr{L}=L_1 \boxtimes L_2=p_1^* L_1 \otimes p_2^*L_2.$$ This is ample by Nakai-Moishezon, since it has positive self-intersection and intersects positively every effective curve in $C \times C$. By the Künneth formula, we have $$H(C_1 \times C_2, \, \mathscr{L})=H^0(C_1, \, L_1) \otimes H^0(C_2, \, L_2)$$ and so a basis of sections for $\mathscr{L}$ is given by elements of the form $\sigma_1 \boxtimes \sigma_2$, with $\sigma_i \in H^0(C_i, \, L_i)$. But all these elements vanish at the point $(p, \, p) \in C_1\times C_2$, hence $\mathscr{L}$ is not globally generated.

This provides a counterexample to your question in dimension $n=2$, and similar constructions provide counterexamples in every dimension. Note that, by increasing the genus of $C$, we can make the integer $h^0(\mathscr{L})$ arbitrary large.

$\endgroup$
6
  • $\begingroup$ Are there examples in higher dimension as well (I had the hypothesis in my question that dimension of the variety is at least 2). $\endgroup$
    – user43198
    Commented Apr 1, 2022 at 12:22
  • $\begingroup$ Ah, ok. Take a product with another curve and take the pull-back of L. It should work. $\endgroup$ Commented Apr 1, 2022 at 12:23
  • $\begingroup$ I am guessing you want to take the tensor product of the pullback of the ample line bundles. I am a little worried that this new line bundle might not be ample. Do you have an argument why it should be ample? $\endgroup$
    – user43198
    Commented Apr 1, 2022 at 12:26
  • $\begingroup$ The box product should work. $\endgroup$ Commented Apr 1, 2022 at 12:48
  • $\begingroup$ Thanks, this is very helpful. $\endgroup$
    – user43198
    Commented Apr 1, 2022 at 12:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .