Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Assume I have a line bundle $\mathcal{L}$ on a projective, nonsingular variety $X$ over, say, an algebraically closed field (in fact, you may assume $\mathbb{C}$). Given global sections $s_0,\ldots,s_k\in\mathcal{L}(X)$, how to prove that they do not generate $\mathcal{L}$ globally?

I'd think that this is best done by contradiction, but I can't think of any theorems that strongly rely on the assumption that a line bundle is globally generated by certain particular sections. Note that $\mathcal{L}$ may very well be globally generated (in fact, in my case, it most certainly is) - but I want to prove that certain sections do not generate it globally. So, what I am basically asking for, are results that include a base-point free linear system in their assumptions, so I can use that to produce a contradiction. My first impule was to consider the induced morphism $\phi:X\to\mathbb{P}^k$, but that's already where I get stuck.

If it helps, the case $X=\mathbb{P}^n$ and $\mathcal{L}=\mathcal{O}_X(1)$ is already interesting to me, but it'd (of course) be better if the strategy works even in the more general setting.

share|improve this question

1 Answer 1

up vote 5 down vote accepted

1) Let's analyze the situation for $X=\mathbb P(V)=\mathbb P(k^{n+1})$ and $\mathcal L=\mathcal O_{X}(1)$ .
The sections of the vector space $\Gamma(X,\mathcal L)=V^*$ generate the line bundle $\mathcal L$ .
However I claim that the sections of any strict subspace $H\subsetneq \Gamma(X,\mathcal L)=V^*$ do not generate $\mathcal L$.
Indeed , we may assume that $H\subsetneq V^*$ is a hyperplane.
It consists of the linear forms $a_0x_0+...+a_nx_n \in V^*$ satisfying $p_0a_0+...+p_na_n=0$ for some $(p_0,\cdots,p_n)\neq 0\in V$ .
Then it is tautological that all sections $s\in H$ vanish at $(p_0:\cdots:p_n)\in X$.

r) Let's analyze the situation for $X=\mathbb P(V)=\mathbb P(k^{n+1})$ and $\mathcal L^r=\mathcal O_{X}(r)\quad (r\geq 2)$
I claim that there exists a hyperplane $H\subsetneq \Gamma(X,\mathcal L^r)=S^r(V^*)=(k[x_0,\cdots,x_n])_r$ whose sections still generate $\mathcal L^r$.
A hyperplane $H_p\subsetneq (k[x_0,\cdots,x_n])_r$ consists of those homogeneous polynomials $\Sigma_{|I|=r}a_Ix^I\in (k[x_0,\cdots,x_n])_r$ of degree $r$ satisfying the relation $\Sigma_{|I|=r}p_Ia_I=0$ for some fixed family $p=(p_I)_{|I|=r}\in \mathbb P^N=\mathbb P(S^rV)$.

Given an arbitrary point $b=(b_0:\cdots :b_n)\in \mathbb P^n$, we must show that there exists a family $(a_I)_{|I|=r}$ such that $\Sigma_{|I|=r}p_Ia_I=0$ but $\Sigma_{|I|=r}a_Ib^I\neq 0$: indeed that family will define a section $s=\Sigma a_Ix^I\in H_p$ not vanishing at $b$.

For that it suffices to take $p\in \mathbb P^N \setminus v(\mathbb P^n)$, i.e. outside of the image of the Veronese embedding $$v=v_r:\mathbb P^n=\mathbb P(V)\to \mathbb P^N=\mathbb P(S^rV):(c_0:\cdots:c_n)\mapsto (\cdots:c^I:\cdots)\quad (|I|=r)$$

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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