The Stacks project

Proof. If $U \subset V \subset X$ are affine open, then the ring map $\mathcal{O}(V) \to \mathcal{O}(U)$ is flat. Hence the equivalence between (4) and (5) is a special case of Cohomology on Sites, Lemma 21.43.11 (the proof also clarifies the statement).

The ringed site $(X_{Zar}, \mathcal{O})$ and the ringed space $(X, \mathcal{O}_ X)$ have the same categories of modules by Descent, Remark 35.8.3. Via this equivalence the quasi-coherent modules correspond by Descent, Proposition 35.8.9. Hence we get a canonical exact equivalence between the triangulated categories in (1) and (2).

The discussion preceding the lemma shows that we have an equivalence of ringed topoi $(\mathop{\mathit{Sh}}\nolimits (X_{affine, Zar}), \mathcal{O}) \to (\mathop{\mathit{Sh}}\nolimits (X_{Zar}), \mathcal{O})$ and hence an equivalence between abelian categories of modules. Since the notion of quasi-coherent modules is intrinsic (Modules on Sites, Lemma 18.23.2) we see that this equivalence preserves the subcategories of quasi-coherent modules. Thus we get a canonical exact equivalence between the triangulated categories in (2) and (3).

To get an exact equivalence between the triangulated categories in (3) and (4) we will apply Cohomology on Sites, Lemma 21.29.1 to the morphism $\epsilon : (X_{affine, Zar}, \mathcal{O}) \to (X_{affine}, \mathcal{O})$ above. We take $\mathcal{B} = \mathop{\mathrm{Ob}}\nolimits (X_{affine})$ and we take $\mathcal{A} \subset \textit{PMod}(X_{affine}, \mathcal{O})$ to be the full subcategory of those presheaves $\mathcal{F}$ such that $\mathcal{F}(V) \otimes _{\mathcal{O}(V)} \mathcal{O}(U) \to \mathcal{F}(U)$ is an isomorphism. Observe that by Descent, Lemma 35.11.2 objects of $\mathcal{A}$ are exactly those sheaves in the Zariski topology which are quasi-coherent modules on $(X_{affine, Zar}, \mathcal{O})$. On the other hand, by Modules on Sites, Lemma 18.24.2, the objects of $\mathcal{A}$ are exactly the quasi-coherent modules on $(X_{affine}, \mathcal{O})$, i.e., in the chaotic topology. Thus if we show that Cohomology on Sites, Lemma 21.29.1 applies, then we do indeed get the canonical equivalence between the categories of (3) and (4) using $\epsilon ^*$ and $R\epsilon _*$.

We have to verify 4 conditions:

1. Every object of $\mathcal{A}$ is a sheaf for the Zariski topology. This we have seen above.

2. $\mathcal{A}$ is a weak Serre subcategory of $\textit{Mod}(X_{affine, Zar}, \mathcal{O})$. Above we have seen that $\mathcal{A} = \mathit{QCoh}(X_{affine, Zar}, \mathcal{O})$ and we have seen above that these, via the equivalence $\textit{Mod}(X_{affine, Zar}, \mathcal{O}) = \textit{Mod}(X, \mathcal{O}_ X)$, correspond to the quasi-coherent modules on $X$. Thus the result by the discussion in Schemes, Section 26.24.

3. Every object of $X_{affine}$ has a covering in the chaotic topology whose members are elements of $\mathcal{B}$. This holds because $\mathcal{B}$ contains all objects.

4. For every object $U$ of $X_{affine}$ and $\mathcal{F}$ in $\mathcal{A}$ we have $H^ p_{Zar}(U, \mathcal{F}) = 0$ for $p > 0$. This holds by the vanishing of cohomology of quasi-coherent modules on affines, see Cohomology of Schemes, Lemma 30.2.2.

This finishes the proof. $\square$