3
$\begingroup$

Let $X$ be a complete nonsingular curve and $S$ a scheme over $k$ algebraically closed, and $\cal{F}$ a coherent sheaf on $X \times S$, generated by finitely many global sections and flat over $S$ (via the projection). So my question is the following: is the locus of points $s\in S$ such that $\cal{F}_s$ is locally free on $X$ open? And if so, is there an easy way to see this?

$\endgroup$
4
  • 3
    $\begingroup$ Welcome new contributor. That is false without a hypothesis that the curve is proper. When the curve is proper, you can prove that with “Fitting ideals”. $\endgroup$ Mar 13, 2022 at 1:38
  • $\begingroup$ Thank you! Yes, the curves I'm working with are in fact proper (I edited the question). Thank you so much! $\endgroup$
    – nolatos
    Mar 13, 2022 at 1:42
  • $\begingroup$ @JasonStarr Is it important that $X$ is a curve? And maybe it could be generalized to a relatively proper map $X\to S$ instead of a trivial family? $\endgroup$
    – Z. M
    Mar 13, 2022 at 11:42
  • $\begingroup$ @Z.M You are correct. The same result holds for a proper, flat morphism. $\endgroup$ Mar 13, 2022 at 12:24

1 Answer 1

7
$\begingroup$

To elaborate on Jason's comment: consider a morphism of schemes $f:\mathrm{X}\rightarrow\mathrm{S}$, and an $f$-flat coherent sheaf $\mathscr{F}$ on $\mathrm{X}$. Then

(1) the singular locus $\mathrm{Sing}(\mathscr{F})$ of $\mathscr{F}$ (given by the points $x\in\mathrm{X}$ such that $\mathscr{F}$ is not locally free at $x$) is closed - if $r$ is the rank of $\mathscr{F}$, it is cut out by the $r$-th Fitting ideal sheaf of $\mathscr{F}$;

(2) $\mathscr{F}$ is locally free at $x\in\mathrm{X}$ if and only if $\mathscr{F}_{f(x)}$ is locally free at $x$, where $\mathscr{F}_{f(x)}$ denotes the restriction of $\mathscr{F}$ to $f^{-1}(f(x))$ - this follows from flatness of $\mathscr{F}$ and the Nakayama lemma.

By (1) and (2), $f(\mathrm{Sing}(\mathscr{F}))$ consists of all $s\in\mathrm{S}$ such that $\mathscr{F}_{s}$ is not locally free. If $f$ is proper, then of course the latter image is closed in $\mathrm{S}$.

Lemma 2.1.8 in The Geometry of Moduli Spaces of Sheaves by Huybrechts & Lehn is a good reference.

$\endgroup$
1
  • $\begingroup$ Thank you so much, that was very helpful! $\endgroup$
    – nolatos
    Mar 14, 2022 at 21:19

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.