6
$\begingroup$

Given a scheme $X$ and sum of divisors $D$, you can take the line bundle $$\mathcal{O}_X(D)\ =\ \{\text{functions }f\text{ with [conditions on zeroes/poles]}\}\ \subseteq\ j_*\mathcal{O}_\eta\ =\ \mathcal{M}$$ where $\mathcal{M}$ is the sheaf of rational functions on $X$ and $j:\eta\to X$ is the generic point (or the union of generic points of the irreducible components). It satisfies $\mathcal{O}_X(D+D') =\mathcal{O}_X(D)\otimes\mathcal{O}_X(D')$. Moreover, under weak conditions these are the only coherent sheaves with an inverse under $\otimes$.


One categorical level up, you can consider sheaves of categories on $X$, with $\text{QCoh}_X$ being the analogue of $\mathcal{O}_X$.

My question is: what is the analogue "$\text{QCoh}_X(D)$" of $\mathcal{O}_X(D)$?


If at all possible, the definition should mimic the above, defining it as a subsheaf of categories of $j_*\text{QCoh}_\eta$ defined by similar [conditions], and with the same identity for $D+D'$. Ideally, under the same weak conditions as above, these are the only invertible sheaves of categories.

$\endgroup$
7
  • 2
    $\begingroup$ Let me check if I understand a few things right: (1) You want to complete the analogy Picard group : Divisors :: Brauer group : (something), right? (2) $\eta$ denotes the localization of $X$ away from the generic point and $j : \eta \to X$ is the inclusion, right? Some suggestions: (3) Is it clear what should happen in the affine case? (4) I might suggest editing the question title to make the question clearer. $\endgroup$ Commented Jul 27, 2023 at 17:49
  • 3
    $\begingroup$ I'd say the analogue is modules over Azumaya algebras (which are the invertible objects among quasi-coherent sheaves of categories). $\endgroup$ Commented Jul 27, 2023 at 18:34
  • $\begingroup$ @TimCampion (3/4) Understanding what $\text{QCoh}_X(1)$ is in the affine case should be the hard part of the question ($\text{QCoh}_X(-1)$ should presumably be the sheaves vanishing along $D$). (1) I have no idea how to generalise (G_m bundles <-> invertible sheaves) to (BG_m gerbes <-> "invertible sheaves of categories") but if you know how I'd love to hear! $\endgroup$
    – Pulcinella
    Commented Jul 27, 2023 at 20:51
  • 1
    $\begingroup$ I don’t think there’s a canonical gerbe associated to a divisor (which I think is what your question is asking for), rather there are many - the root stacks (gerbe of n-th roots of O(D) for given n, and the corresponding n-torsion invertible sheaf of categories), cf eg papers of Talpo and Vistoli $\endgroup$ Commented Jul 28, 2023 at 5:09
  • 5
    $\begingroup$ I think these results are originally due to Toën and Antieau-Gepner, but a convenient reference is section 11.5 in Spectral Algebraic Geometry. There is a bijection between invertible quasi-coherent sheaves of $\infty$-categories on a qcqs scheme $X$ and (derived) Azumaya algebras modulo Morita equivalence. You may also be interested in Remark 11.5.5.5: if $X$ is normal and noetherian, these are further equivalent to $B\mathbb G_m$-torsors. $\endgroup$ Commented Jul 28, 2023 at 5:12

0

You must log in to answer this question.