$\newcommand{\oh}{\mathcal{O}} \newcommand{\QCoh}{\mathsf{QCoh}} \newcommand{\ra}{\rightarrow} \newcommand{\F}{\mathcal{F}} \newcommand{\Mod}{\text{-}\mathsf{Mod}}$I'm working through the proof of Proposition 10 of Murfet's notes. The proposition is as follows:

Let $f : X \ra Y$ be a morphism of schemes, where $X$ is noetherian and $Y = \text{Spec} A$ is affine. Then for any $ \F \in \QCoh(X)$ and $i \geq 0$ there exists a canonical isomorphism of sheaves of modules on $Y$ natural in $\F$ $$ \beta : R^if_* \F \longrightarrow \widetilde{H^i(X, \F)}.$$

The proof goes as follows (I've denoted my questions with "Q:"):

$R^if_* \F$ has a canonical $\oh_Y$-module structure and $H^i(X, \F)$ has a canonical $\Gamma(X, \F)$-module structure, and since $\Gamma(X, \F)$ has the structure of an $A$-module, $H^i(X, \F)$ has an $A$-module structure. Also, $X$ noetherian implies that $f_* \F \in \QCoh(Y)$. Therefore, we have a canonical isomorphism of sheaves $$ f_* \F \cong \widetilde{\Gamma(X, \F)}. $$ Q: I'm not sure why we have this. I know that if in general $\F$ is a quasi-coherent sheaf on $X$, then for $U_i$ opens of a particular cover of $X$, we have $\F|_{U_i} \cong \widetilde{F(U_i)}$. But I'm not sure as to how we derive the above? I've tried $$ f_* \F|_{U_i} = \F(f^{-1}(-))|_{U_i} $$ but this doesn't really get me anywhere.

Proceeding with the proof, we have for $i=0$ a canonical isomorphism natural in $\F$ $$ \mu^0 : R^0f_* \F \cong f_* \F \cong \widetilde{\Gamma(X, \F)} = \widetilde{H^0(X, \F)} \quad \checkmark $$ Now, since the tilde functor $\widetilde{-}: A \Mod \ra \oh_Y \Mod$ is exact, we have two cohomological $\delta$-functors $\{ R^i f_*(-) \}_{i \geq 0}$ and $\{ \widetilde{H^i(X, -)} \}_{i \geq 0}$ between $\QCoh(X)$ and $\oh_Y \Mod$.

Q: Why does this follow from exactness of the tilde functor? Apologies; this may be obvious (I'm not too brushed up on my $\delta$-functor knowledge).

Quasi-coherent sheaves may be embedded into flasque quasi-coherent sheaves. Hence, both functors are effaceable for $i>0$.

Q: Why does the $\delta$-functors being effaceable follow from this? Effaceable means (in this case) that for any object $\F \in \QCoh(X)$ there exists a monomorphism $u : \F \ra \mathcal{G}$ such that $\{ R^i f_*(u) \}_{i \geq 0} = 0$ and $\{\widetilde{H^i(X, u)} \}_{i \geq 0}=0$, some $\mathcal{G}$. I think that this is because we can say $u$ is the embedding into a flasque quasi-coherent sheaf, and since sheaf cohomology vanishes for flasque sheaves and higher direct images, we get the result. Is this correct modulo details?

Then by a theorem of Grothendieck, both of the $\delta$-functors are universal (i.e. a universal $\delta$-functor is characterised by the property that giving any morphism from it to any other $\delta$-functor is equivalent to giving just the $0$th degree). Therefore, $\mu^0$ gives rise to the canonical natural equivalence that we require.

Apologies for the lengthy post and thank you for any answers!


1 Answer 1


Q1: On an affine scheme $Z$, we know that for any quasicoherent sheaf $\mathcal{A}$ we have that $\mathcal{A}\cong \widetilde{\mathcal{A}(Z)}$. Applying that to the case at hand, we know that $f_*\mathcal{F}$ is a quasicoherent sheaf on $Y$, so $f_*\mathcal{F}\cong \widetilde{f_*\mathcal{F}(Y)}$. But by the definition of the pushforward, we have that $f_*\mathcal{F}(Y) = \mathcal{F}(f^{-1}(Y)) = \mathcal{F}(X)$, and so we may conclude that $f_*\mathcal{F} \cong \widetilde{\mathcal{F}(X)}$.

Q2: One functor described here is just $R^\bullet f_*(-)$, which is known to be a cohomological $\delta$-functor basically by definition (it's a right derived functor of a left exact functor). The other functor here is the functor $\widetilde{H^\bullet(X,-)}$, which can be written as the composite of the two functors $\widetilde{-}$ and $H^\bullet(X,-)$. The first functor, $H^\bullet(X,-)$ is already a $\delta$-functor, and $\widetilde{-}$ being exact implies that it preserves the exactness of all the sequences and diagrams required to verify $H^\bullet(X,-)$ as a $\delta$-functor, so the composite is a $\delta$-functor.

Q3: Yes, this is essentially correct.

  • $\begingroup$ Great - everything makes sense now. Thank you for your answer! $\endgroup$
    – mathphys
    May 31, 2019 at 12:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.