This is a question from Ravi Vakil's notes I've been stuck on, namely 18.2.I.

Let $X$ be a scheme over a field $k$, and let $K/k$ be any field extension. Let $\mathcal{L}$ be an invertible sheaf on $X$. Then I would like to show $\mathcal{L}$ is base-point free if and only if the pullback of $\mathcal{L}$ to $X_K$, say $\mathcal{L}\otimes K$, is base-point free.

I know that $H^0(X, \mathcal{L})\otimes K\cong H^0(X_K, \mathcal{L}\otimes K)$. In particular, a basis of $H^0(X,\mathcal{L})$ over $k$ yields a basis of $H^0(X_K, \mathcal{L}\otimes K)$ after tensoring with $K$.

If $\mathcal{L}$ is base-point free, then so too must the pullback be base-point free. Otherwise there exists a base point $p$ in $X_K$, whose image in $X$ is $q$. As $q$ is not a base-point of $\mathcal{L}$, there exists some global section $s$ such that $s$ doesn't vanish at $q$, and thus the pullback of $s$ doesn't vanish at $p$.

I'm not sure how to proceed showing the converse, namely that $\mathcal{L}\otimes K$ being base-point free implies $\mathcal{L}$ is. If I try to proceed by contradiction I run into the issue where a point of #X# may not have a point in $X_K$ lying over it (as $K/k$ is not necessarily algebraic).

Could someone point me in the right direction?

  • 1
    $\begingroup$ Try to compute the vanishing of the pullback of one of these global sections in the basis of $H^0(X,L)$. There is a formula in the big list of properties of the pullback in chapter 16 in Ravi. Then study how the intersections of the vanishing behave after pullback also. I think that works. Also, I think there will always be a point in $X \times_k K$ lying above any given point $p$ of $X$ - this just because $k(p) \otimes_k K$ is never the zero ring, so when you compute the scheme theoretic preimage it is not empty. $\endgroup$ – Lorenzo Najt Apr 7 '16 at 21:24

To show that any point of $X$ has a point in $X_K$ lying over it, just recall that surjectivity is stable under base change (and $\operatorname{Spec}(K)\rightarrow \operatorname{Spec}(k)$ is surjective).

To see why this finishes the argument (for future askers of this question), notice that we have a map $\Gamma(X,\mathcal{L})\rightarrow \Gamma(X_K,\mathcal{L}\otimes K)$ and for any $p \in X_K$ mapping to $q \in X$, we have a diagram:

$$\require{AMScd} \begin{CD} \Gamma(X,\mathcal{L}) @>>> \Gamma(X_K,\mathcal{L}\otimes K)\\ @VVV @VVV \\ \mathcal{L}_q @>>> (\mathcal{L}\otimes K)_p \end{CD}$$

Then $q$ is a basepoint of $\mathcal{L}$ iff the left hand map is the $0$ map (because $\mathcal{L}$ is a line bundle and so any non-zero element of the stalk will generate it) and since neither of the horizontal maps are $0$, this would imply that $p$ is a basepoint of $\mathcal{L} \otimes K$, a contradiction of our hypothesis, so no points in the image of $X_K$ are basepoints and we're done by surjectivity.


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.