Let $X$ be a scheme and $\mathcal S$ a site which is a full subcategory of the category $Aff/X$ of affine schemes with a map to $X$. If I understand correctly, the category $QCoh^\mathcal S(X)$ of $\mathcal S$-quasicoherent sheaves is the global sections of the $\mathcal S$-stackification of the functor $Mod: \mathcal S^{op} \to Cat$, $(Spec A \to X) \mapsto Mod_A$, where $Mod_A$ is the category of $A$-modules.

The interesting phenomenon (cf. the stacks project, which is probably using slightly different definitions) is that for reasonable $\mathcal S$ (the above link gives precise conditions), $QCoh^\mathcal S(X)$ is actually independent of $\mathcal S$. I'm looking for a high-concept explanation of this fact.

The best I can figure is the following. For reasonable topologies, the functor $Mod: \mathcal S \to Cat$ is already a stack, so its global sections can be computed using an $\mathcal S$-cover of the terminal object (which I've technically left out of the category $\mathcal S$, but that's okay). Since $X$ is a scheme, it comes with a Zariski cover, which is also a $\mathcal S$-cover for reasonable topologies. Thus $QCoh^\mathcal S(X)$ is simply computed in the same way for reasonable topologies $\mathcal S$, so of course it agrees.

This suggests that a statement of this meta-principle (an alternative to the one found in the stacks project) would say that

Claim: Let $X$ be a scheme $\mathcal S$ be a full subcategory of $Aff/X$, equipped with a Grothendieck topology. Then $QCoh^\mathcal S(X) = QCoh^{Zariski}(X)$ if

  1. $Mod : \mathcal S^{op} \to Cat$ is a stack, and

  2. Every Zariski cover is an $\mathcal S$-cover.


  1. Is the above claim correct?

  2. If not (or if so!) is there some other high-concept way to see that $QCoh^\mathcal S(X)$ is independent of $\mathcal S$ for reasonable $\mathcal S$?

  • $\begingroup$ Isn't this fpqc descent for quasicoherent sheaves: stacks.math.columbia.edu/tag/023R ? $\endgroup$ Commented Dec 12, 2018 at 20:06
  • $\begingroup$ @HarryGindi yes $\endgroup$
    – Tim Campion
    Commented Dec 12, 2018 at 20:07
  • $\begingroup$ Condition 2 does not really make sense as $\mathcal S$ is defined as a topology on $Aff/X$, which presumably means the topology defines coverings within $Aff/X$. In this generality one can define global sections as choices of an object in $\mathcal F(U)$ for each open set $U$, compatible with maps, and then 1 is obviously sufficient because the maps are the same in each topology. $\endgroup$
    – Will Sawin
    Commented Dec 12, 2018 at 20:12

1 Answer 1


The basic reason is that a sheaf of modules is quasi-coherent if and only if it is a Cartesian presheaf. Therefore a certain covering suffices for the condition to hold. The condition on coarse covers determine the behavior on fine covers, therefore being quasi-coherent in Zariski topology (or in étale topology on an algebraic space, DM-stack) already makes the sheaf quasi-coherent in any finer topology. The theory is specially simple when you have a base of "affine" schemes that cover the base space.

The Cartesian condition is just a condition on presheaves, so, in a sense, topology is not important. Risking to commit self-promotion, this is treated in the context of geometric stacks in the paper "A functorial formalism for quasi-coherent sheaves on a geometric stack", https://arxiv.org/abs/1304.2520 (see also Expo. Math., 33 (2015) pp. 452-501). The case of affine schemes is explained in section 2 and it is generalized to stacks with affine diagonal in section 3. The case of a semi-separated scheme follows adapting (and somehow simplifying) the arguments there. In particular, one does not need to use the small flat site in this case

  • 2
    $\begingroup$ Sorry, what is a Cartesian sheaf? Is there a site called the "cartesian site"? $\endgroup$
    – Tim Campion
    Commented Dec 13, 2018 at 5:40
  • $\begingroup$ @TimCampion I've made an edit. Hope this is more explicit now. $\endgroup$
    – Leo Alonso
    Commented Dec 13, 2018 at 9:25
  • $\begingroup$ @Tim Campiom: see definition 3.1 on page 44 of Vistoli's notes (homepage.sns.it/vistoli/descent.pdf). I think a presheaf $\mathcal{F}$ on a category $\mathcal{C}$ is cartesian when it is cartesian when you think of it as a fibered (in sets) category $\mathcal{F}\to\mathcal{C}$. Compare with Définition (12.3) in Champ algébriques by Laumon--Moret-Bailly $\endgroup$
    – Qfwfq
    Commented Dec 14, 2018 at 16:47

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.