3
$\begingroup$

Reflexive sheaves on a regular quasi-projective variety can be characterized by the following property that they are the kernel of a surjection from a vector bundle to a torsion-free sheaf. I wonder what the class of reflexive sheaves that are kernel of a surjection from a vector bundle to a reflexive sheaf consists of? They are clearly reflexive but can any reflexive sheaf written in that form? Or under what conditions every reflexive sheaf is a kernel of a surjection from a vector bundle to another reflexive sheaf? Is there any other way to characterize these specific reflexive sheaves?

$\endgroup$

1 Answer 1

5
$\begingroup$

If the ambient variety is smooth, the locus of points where a reflexive sheaf is not locally free has codimension at least 3. And for a sheaf which is a kernel of a surjection from locally free to reflexive, this sheaf has codimension at least 4. So, the classes are different.

In general, one can consider so-called locally $m$-syzygy sheaves --- these are sheaves $F$ for which there is a resolution $$ 0 \to F \to E_1 \to E_2 \to \dots \to E_m $$ (not necessarily exact in the rightmost term) with $E_i$ locally free. For $m = 1$ these are torsion-free sheaves, for $m = 2$ reflexive, and for $m = 3$ your class.

$\endgroup$
6
  • $\begingroup$ Is there a reference that proves this? Is this still true of the base field is not algebraically closed? $\endgroup$
    – user127776
    Commented Dec 23, 2020 at 4:42
  • 2
    $\begingroup$ The case $m=1$ (torsion-free) is Tag 0AUU, and then $m=2$ follows from that and Tag 0AV2, and $m=3$ is your class by definition. $\endgroup$ Commented Dec 23, 2020 at 4:49
  • 2
    $\begingroup$ Ah, can't you just prove that the same way as Tag 03BN? I think the point is that when you localise to a point of codimension $\leq 3$, you get a maximal Cohen–Macaulay module (see Tag 00NF), which is free by Tag 00NT. $\endgroup$ Commented Dec 23, 2020 at 5:00
  • 2
    $\begingroup$ I learnt this stuff from the book of Okonek-Schneider-Spindler. $\endgroup$
    – Sasha
    Commented Dec 23, 2020 at 5:16
  • 1
    $\begingroup$ As an answer to the question I asked above. The converse is not true i.e. if a reflexive sheaf is a vector bundle on the complement of a codimension 4 subvariety, it doesn't need to be locally 3-syzygy sheaf. I figured this out through an unexpected contradiction, there might be an easier way of showing it. $\endgroup$
    – user127776
    Commented Dec 23, 2020 at 23:50

You must log in to answer this question.

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