5
$\begingroup$

My question is a subquestion of this question. And a repost from this MSE question.

The settings is the following. Let $X$ be a scheme. Assume that the adjunction in TAG01BH of the stacks project is actually an adjoint equivalence. My goal is to proof that $X$ is affine.

This is a special case of the general question: can we reconstruct a scheme from its category of quasi-coherent sheafs? I am aware of the available reconstruction theorems, but I feel that this could be done with elementary methods. Feel free to critique this feeling.

For starters, quasi-separatedness of $X$ should be established before you can invoke Gabriel-Rosenberg and quasi-compactness before you can use Serre's vanishing theorem on $X$. Any hints on how to go about this?

What I tried thus far is trying to use Serre's vanishing theorem. The problem: how to prove that there exist 'enough' acyclic resolutions inside $Qcoh(X)$ as is the case for $Qcoh(\mbox{Spec } \Gamma(X,O_X))$, since $\widetilde{I}$ is flasque for injective $\Gamma(X,O_X)$-modules $I$.

$\endgroup$
4
  • 2
    $\begingroup$ Regarding your last comment: There exists a non-noetherian ring $A$ and an injective $A$-module $E$, such that $\widetilde{E}$ is not a flasque $\mathcal{O}_{\mathrm{Spec}(A)}$-module (cf. SGA VI, Expose 2, Appendice 1, 0.1 or in the Stack Project stacks.math.columbia.edu/tag/0273). Nevertheless, the category $\mathrm{QCoh}(X)$ has enough injective objects for any scheme since it is a Grothendieck abelian category - a result due to Gabber resp. Franke. $\endgroup$ Jun 17, 2015 at 7:15
  • 2
    $\begingroup$ Could you please specify your assumptions? Do you mean the following? (a) For every $\Gamma(X,\mathcal{O}_X)$-module $M$ the canonical map $M \to \Gamma(X,\tilde{M})$ is an isomorphism, (b) for every quasi-coherent $\mathcal{O}_X$-module $F$ the canonical map $\widetilde{\Gamma(X,F)} \to F$ is an isomorphism, and (c) for every two quasi-coherent $\mathcal{O}_X$-modules $F,G$ the canonical map $\Gamma(X,F) \otimes \Gamma(X,G) \to \Gamma(X,F \otimes G)$ is an isomorphism? $\endgroup$ Aug 16, 2015 at 13:02
  • $\begingroup$ The assumptions are (a) & (b). So the unit and co-unit of the adjunction between $\Gamma(X,-)$ and $\widetilde{}$ are natural isomorphisms. Removed the part about the tensor structure. $\endgroup$
    – bbnkttp
    Aug 16, 2015 at 17:09
  • $\begingroup$ - = 'tilde'. My apologizes for the bad formatting. $\endgroup$
    – bbnkttp
    Aug 16, 2015 at 17:15

1 Answer 1

1
$\begingroup$

First prove that the map $X \to \operatorname{Spec} \Gamma( X, \mathcal O_X)$ is a bijection on points, then that it is a homeomorphism, then that the structure sheaf is equivalent.

Given two points $p_1, p_2$ in the same fiber, let $k$ be a field which contains both $\mathcal O_X/p_1$ and $\mathcal O_x/p_2$. Then unless $p_1=p_2$ you can give it two different structures of an $\mathcal O_X$-module and hence a coherent sheaf, but the pushforward of the two is the same. So $p_1$, $p_2$ are actually the same.

Given a closed set $Z \subseteq X$, we must show it arises by pullback from a closed subset of $X \to \operatorname{Spec} \Gamma( X, \mathcal O_X)$. Well $\mathcal O_Z$ arises by pullback from some sheaf on $\operatorname{Spec} \Gamma( X, \mathcal O_X)$. Moreover as $\mathcal O_Z$ is a quotient of $\mathcal O_X$, so that sheaf must be a quotient of the structure sheaf by an ideal. A point is contained in $Z$ if and only if the map to the structure sheaf of that point factors through $\mathcal O_Z$ if and only if the point is in the vanishing set of this ideal. So $Z$ is the pullback of the vanishing set of this ideal, which is closed.

Finally, the fact that the pushforward of $\mathcal O_X$ is the structure sheaf of $\operatorname{Spec} \Gamma( X, \mathcal O_X)$ implies that the map is an isomorphism on structure sheaves.

$\endgroup$

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.