The Stacks project

82.6 Comparing fppf and étale topologies

This section is the analogue of Étale Cohomology, Section 58.94.

Let $S$ be a scheme. Let $X$ be an algebraic space over $S$. On the category $\textit{Spaces}/X$ we consider the fppf and étale topologies. The identity functor $(\textit{Spaces}/X)_{\acute{e}tale}\to (\textit{Spaces}/X)_{fppf}$ is continuous and defines a morphism of sites

\[ \epsilon _ X : (\textit{Spaces}/X)_{fppf} \longrightarrow (\textit{Spaces}/X)_{\acute{e}tale} \]

by an application of Sites, Proposition 7.14.7. Please note that $\epsilon _{X, *}$ is the identity functor on underlying presheaves and that $\epsilon _ X^{-1}$ associates to an étale sheaf the fppf sheafification. Consider the morphism of sites

\[ \pi _ X : (\textit{Spaces}/X)_{\acute{e}tale}\longrightarrow X_{spaces, {\acute{e}tale}} \]

comparing big and small étale sites, see Section 82.5. The composition determines a morphism of sites

\[ a_ X = \pi _ X \circ \epsilon _ X : (\textit{Spaces}/X)_{fppf} \longrightarrow X_{spaces, {\acute{e}tale}} \]

If $\mathcal{H}$ is an abelian sheaf on $(\textit{Spaces}/X)_{fppf}$, then we will write $H^ n_{fppf}(U, \mathcal{H})$ for the cohomology of $\mathcal{H}$ over an object $U$ of $(\textit{Spaces}/X)_{fppf}$.

Lemma 82.6.1. Let $S$ be a scheme. Let $X$ be an algebraic space over $S$.

  1. For $\mathcal{F} \in \mathop{\mathit{Sh}}\nolimits (X_{\acute{e}tale})$ we have $\epsilon _{X, *}a_ X^{-1}\mathcal{F} = \pi _ X^{-1}\mathcal{F}$ and $a_{X, *}a_ X^{-1}\mathcal{F} = \mathcal{F}$.

  2. For $\mathcal{F} \in \textit{Ab}(X_{\acute{e}tale})$ we have $R^ i\epsilon _{X, *}(a_ X^{-1}\mathcal{F}) = 0$ for $i > 0$.

Proof. We have $a_ X^{-1}\mathcal{F} = \epsilon _ X^{-1} \pi _ X^{-1}\mathcal{F}$. By Lemma 82.5.1 the étale sheaf $\pi _ X^{-1}\mathcal{F}$ is a sheaf for the fppf topology and therefore is equal to $a_ X^{-1}\mathcal{F}$ (as pulling back by $\epsilon _ X$ is given by fppf sheafification). Recall moreover that $\epsilon _{X, *}$ is the identity on underlying presheaves. Now part (1) is immediate from the explicit description of $\pi _ X^{-1}$ in Lemma 82.5.1.

We will prove part (2) by reducing it to the case of schemes – see part (1) of Étale Cohomology, Lemma 58.94.6. This will “clearly work” as every algebraic space is étale locally a scheme. The details are given below but we urge the reader to skip the proof.

For an abelian sheaf $\mathcal{H}$ on $(\textit{Spaces}/X)_{fppf}$ the higher direct image $R^ p\epsilon _{X, *}\mathcal{H}$ is the sheaf associated to the presheaf $U \mapsto H^ p_{fppf}(U, \mathcal{H})$ on $(\textit{Spaces}/X)_{\acute{e}tale}$. See Cohomology on Sites, Lemma 21.7.4. Since every object of $(\textit{Spaces}/X)_{\acute{e}tale}$ has a covering by schemes, it suffices to prove that given $U/X$ a scheme and $\xi \in H^ p_{fppf}(U, a_ X^{-1}\mathcal{F})$ we can find an étale covering $\{ U_ i \to U\} $ such that $\xi $ restricts to zero on $U_ i$. We have

\begin{align*} H^ p_{fppf}(U, a_ X^{-1}\mathcal{F}) & = H^ p((\textit{Spaces}/U)_{fppf}, (a_ X^{-1}\mathcal{F})|_{\textit{Spaces}/U}) \\ & = H^ p((\mathit{Sch}/U)_{fppf}, (a_ X^{-1}\mathcal{F})|_{\mathit{Sch}/U}) \end{align*}

where the second identification is Lemma 82.3.1 and the first is a general fact about restriction (Cohomology on Sites, Lemma 21.7.1). Looking at the first paragraph and the corresponding result in the case of schemes (Étale Cohomology, Lemma 58.94.1) we conclude that the sheaf $(a_ X^{-1}\mathcal{F})|_{\mathit{Sch}/U}$ matches the pullback by the “schemes version of $a_ U$”. Therefore we can find an étale covering $\{ U_ i \to U\} $ such that our class dies in $H^ p((\mathit{Sch}/U_ i)_{fppf}, (a_ X^{-1}\mathcal{F})|_{\mathit{Sch}/U_ i})$ for each $i$, see Étale Cohomology, Lemma 58.94.6 (the precise statement one should use here is that $V_ n$ holds for all $n$ which is the statement of part (2) for the case of schemes). Transporting back (using the same formulas as above but now for $U_ i$) we conclude $\xi $ restricts to zero over $U_ i$ as desired. $\square$

The hard work done in the case of schemes now tells us that étale and fppf cohomology agree for sheaves coming from the small étale site.

Lemma 82.6.2. Let $S$ be a scheme. Let $X$ be an algebraic space over $S$. For $K \in D^+(X_{\acute{e}tale})$ the maps

\[ \pi _ X^{-1}K \longrightarrow R\epsilon _{X, *}a_ X^{-1}K \quad \text{and}\quad K \longrightarrow Ra_{X, *}a_ X^{-1}K \]

are isomorphisms with $a_ X : \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/X)_{fppf}) \to \mathop{\mathit{Sh}}\nolimits (X_{\acute{e}tale})$ as above.

Proof. We only prove the second statement; the first is easier and proved in exactly the same manner. There is an immediate reduction to the case where $K$ is given by a single abelian sheaf. Namely, represent $K$ by a bounded below complex $\mathcal{F}^\bullet $. By the case of a sheaf we see that $\mathcal{F}^ n = a_{X, *} a_ X^{-1} \mathcal{F}^ n$ and that the sheaves $R^ qa_{X, *}a_ X^{-1}\mathcal{F}^ n$ are zero for $q > 0$. By Leray's acyclicity lemma (Derived Categories, Lemma 13.16.7) applied to $a_ X^{-1}\mathcal{F}^\bullet $ and the functor $a_{X, *}$ we conclude. From now on assume $K = \mathcal{F}$.

By Lemma 82.6.1 we have $a_{X, *}a_ X^{-1}\mathcal{F} = \mathcal{F}$. Thus it suffices to show that $R^ qa_{X, *}a_ X^{-1}\mathcal{F} = 0$ for $q > 0$. For this we can use $a_ X = \epsilon _ X \circ \pi _ X$ and the Leray spectral sequence (Cohomology on Sites, Lemma 21.14.7). By Lemma 82.6.1 we have $R^ i\epsilon _{X, *}(a_ X^{-1}\mathcal{F}) = 0$ for $i > 0$. We have $\epsilon _{X, *}a_ X^{-1}\mathcal{F} = \pi _ X^{-1}\mathcal{F}$ and by Lemma 82.5.5 we have $R^ j\pi _{X, *}(\pi _ X^{-1}\mathcal{F}) = 0$ for $j > 0$. This concludes the proof. $\square$

Lemma 82.6.3. Let $S$ be a scheme and let $X$ be an algebraic space over $S$. With $a_ X : \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/X)_{fppf}) \to \mathop{\mathit{Sh}}\nolimits (X_{\acute{e}tale})$ as above:

  1. $H^ q(X_{\acute{e}tale}, \mathcal{F}) = H^ q_{fppf}(X, a_ X^{-1}\mathcal{F})$ for an abelian sheaf $\mathcal{F}$ on $X_{\acute{e}tale}$,

  2. $H^ q(X_{\acute{e}tale}, K) = H^ q_{fppf}(X, a_ X^{-1}K)$ for $K \in D^+(X_{\acute{e}tale})$.

Example: if $A$ is an abelian group, then $H^ q_{\acute{e}tale}(X, \underline{A}) = H^ q_{fppf}(X, \underline{A})$.

Proof. This follows from Lemma 82.6.2 by Cohomology on Sites, Remark 21.14.4. $\square$

Lemma 82.6.4. Let $S$ be a scheme. Let $f : X \to Y$ be a morphism of algebraic spaces over $S$. Then there are commutative diagrams of topoi

\[ \xymatrix{ \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/X)_{fppf}) \ar[rr]_{f_{big, fppf}} \ar[d]_{\epsilon _ X} & & \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/Y)_{fppf}) \ar[d]^{\epsilon _ Y} \\ \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/X)_{\acute{e}tale}) \ar[rr]^{f_{big, {\acute{e}tale}}} & & \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/Y)_{\acute{e}tale}) } \]

and

\[ \xymatrix{ \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/X)_{fppf}) \ar[rr]_{f_{big, fppf}} \ar[d]_{a_ X} & & \mathop{\mathit{Sh}}\nolimits ((\textit{Spaces}/Y)_{fppf}) \ar[d]^{a_ Y} \\ \mathop{\mathit{Sh}}\nolimits (X_{\acute{e}tale}) \ar[rr]^{f_{small}} & & \mathop{\mathit{Sh}}\nolimits (Y_{\acute{e}tale}) } \]

with $a_ X = \pi _ X \circ \epsilon _ X$ and $a_ Y = \pi _ X \circ \epsilon _ X$.

Proof. This follows immediately from working out the definitions of the morphisms involved, see Topologies on Spaces, Section 71.7 and Section 82.5. $\square$

Lemma 82.6.5. In Lemma 82.6.4 if $f$ is proper, then we have

  1. $a_ Y^{-1} \circ f_{small, *} = f_{big, fppf, *} \circ a_ X^{-1}$, and

  2. $a_ Y^{-1}(Rf_{small, *}K) = Rf_{big, fppf, *}(a_ X^{-1}K)$ for $K$ in $D^+(X_{\acute{e}tale})$ with torsion cohomology sheaves.

Proof. Proof of (1). You can prove this by repeating the proof of Lemma 82.5.6 part (1); we will instead deduce the result from this. As $\epsilon _{Y, *}$ is the identity functor on underlying presheaves, it reflects isomorphisms. Lemma 82.6.1 shows that $\epsilon _{Y, *} \circ a_ Y^{-1} = \pi _ Y^{-1}$ and similarly for $X$. To show that the canonical map $a_ Y^{-1}f_{small, *}\mathcal{F} \to f_{big, fppf, *}a_ X^{-1}\mathcal{F}$ is an isomorphism, it suffices to show that

\begin{align*} \pi _ Y^{-1}f_{small, *}\mathcal{F} & = \epsilon _{Y, *}a_ Y^{-1}f_{small, *}\mathcal{F} \\ & \to \epsilon _{Y, *}f_{big, fppf, *}a_ X^{-1}\mathcal{F} \\ & = f_{big, {\acute{e}tale}, *} \epsilon _{X, *}a_ X^{-1}\mathcal{F} \\ & = f_{big, {\acute{e}tale}, *}\pi _ X^{-1}\mathcal{F} \end{align*}

is an isomorphism. This is part (1) of Lemma 82.5.6.

To see (2) we use that

\begin{align*} R\epsilon _{Y, *}Rf_{big, fppf, *}a_ X^{-1}K & = Rf_{big, {\acute{e}tale}, *}R\epsilon _{X, *}a_ X^{-1}K \\ & = Rf_{big, {\acute{e}tale}, *}\pi _ X^{-1}K \\ & = \pi _ Y^{-1}Rf_{small, *}K \\ & = R\epsilon _{Y, *} a_ Y^{-1}Rf_{small, *}K \end{align*}

The first equality by the commutative diagram in Lemma 82.6.4 and Cohomology on Sites, Lemma 21.19.2. Then second equality is Lemma 82.6.2. The third is Lemma 82.5.6 part (2). The fourth is Lemma 82.6.2 again. Thus the base change map $a_ Y^{-1}(Rf_{small, *}K) \to Rf_{big, fppf, *}(a_ X^{-1}K)$ induces an isomorphism

\[ R\epsilon _{Y, *}a_ Y^{-1}Rf_{small, *}K \to R\epsilon _{Y, *}Rf_{big, fppf, *}a_ X^{-1}K \]

The proof is finished by the following remark: a map $\alpha : a_ Y^{-1}L \to M$ with $L$ in $D^+(Y_{\acute{e}tale})$ and $M$ in $D^+((\textit{Spaces}/Y)_{fppf})$ such that $R\epsilon _{Y, *}\alpha $ is an isomorphism, is an isomorphism. Namely, we show by induction on $i$ that $H^ i(\alpha )$ is an isomorphism. This is true for all sufficiently small $i$. If it holds for $i \leq i_0$, then we see that $R^ j\epsilon _{Y, *}H^ i(M) = 0$ for $j > 0$ and $i \leq i_0$ by Lemma 82.6.1 because $H^ i(M) = a_ Y^{-1}H^ i(L)$ in this range. Hence $\epsilon _{Y, *}H^{i_0 + 1}(M) = H^{i_0 + 1}(R\epsilon _{Y, *}M)$ by a spectral sequence argument. Thus $\epsilon _{Y, *}H^{i_0 + 1}(M) = \pi _ Y^{-1}H^{i_0 + 1}(L) = \epsilon _{Y, *}a_ Y^{-1}H^{i_0 + 1}(L)$. This implies $H^{i_0 + 1}(\alpha )$ is an isomorphism (because $\epsilon _{Y, *}$ reflects isomorphisms as it is the identity on underlying presheaves) as desired. $\square$

Lemma 82.6.6. In Lemma 82.6.4 if $f$ is finite, then $a_ Y^{-1}(Rf_{small, *}K) = Rf_{big, fppf, *}(a_ X^{-1}K)$ for $K$ in $D^+(X_{\acute{e}tale})$.

Proof. Let $V \to Y$ be a surjective étale morphism where $V$ is a scheme. It suffices to prove the base change map is an isomorphism after restricting to $V$. Hence we may assume that $Y$ is a scheme. As the morphism is finite, hence representable, we conclude that we may assume both $X$ and $Y$ are schemes. In this case the result follows from the case of schemes (Étale Cohomology, Lemma 58.94.6 part (2)) using the comparison of topoi discussed in Section 82.3 and in particular given in Lemma 82.3.1. Some details omitted. $\square$

Lemma 82.6.7. In Lemma 82.6.4 assume $f$ is flat, locally of finite presentation, and surjective. Then the functor

\[ \mathop{\mathit{Sh}}\nolimits (Y_{\acute{e}tale}) \longrightarrow \left\{ (\mathcal{G}, \mathcal{H}, \alpha ) \middle | \begin{matrix} \mathcal{G} \in \mathop{\mathit{Sh}}\nolimits (X_{\acute{e}tale}),\ \mathcal{H} \in \mathop{\mathit{Sh}}\nolimits ((\mathit{Sch}/Y)_{fppf}), \\ \alpha : a_ X^{-1}\mathcal{G} \to f_{big, fppf}^{-1}\mathcal{H} \text{ an isomorphism} \end{matrix} \right\} \]

sending $\mathcal{F}$ to $(f_{small}^{-1}\mathcal{F}, a_ Y^{-1}\mathcal{F}, can)$ is an equivalence.

Proof. The functor $a_ X^{-1}$ is fully faithful (as $a_{X, *}a_ X^{-1} = \text{id}$ by Lemma 82.6.1). Hence the forgetful functor $(\mathcal{G}, \mathcal{H}, \alpha ) \mapsto \mathcal{H}$ identifies the category of triples with a full subcategory of $\mathop{\mathit{Sh}}\nolimits ((\mathit{Sch}/Y)_{fppf})$. Moreover, the functor $a_ Y^{-1}$ is fully faithful, hence the functor in the lemma is fully faithful as well.

Suppose that we have an étale covering $\{ Y_ i \to Y\} $. Let $f_ i : X_ i \to Y_ i$ be the base change of $f$. Denote $f_{ij} = f_ i \times f_ j : X_ i \times _ X X_ j \to Y_ i \times _ Y Y_ j$. Claim: if the lemma is true for $f_ i$ and $f_{ij}$ for all $i, j$, then the lemma is true for $f$. To see this, note that the given étale covering determines an étale covering of the final object in each of the four sites $Y_{\acute{e}tale}, X_{\acute{e}tale}, (\mathit{Sch}/Y)_{fppf}, (\mathit{Sch}/X)_{fppf}$. Thus the category of sheaves is equivalent to the category of glueing data for this covering (Sites, Lemma 7.26.5) in each of the four cases. A huge commutative diagram of categories then finishes the proof of the claim. We omit the details. The claim shows that we may work étale locally on $Y$. In particular, we may assume $Y$ is a scheme.

Assume $Y$ is a scheme. Choose a scheme $X'$ and a surjective étale morphism $s : X' \to X$. Set $f' = f \circ s : X' \to Y$ and observe that $f'$ is surjective, locally of finite presentation, and flat. Claim: if the lemma is true for $f'$, then it is true for $f$. Namely, given a triple $(\mathcal{G}, \mathcal{H}, \alpha )$ for $f$, we can pullback by $s$ to get a triple $(s_{small}^{-1}\mathcal{G}, \mathcal{H}, s_{big, fppf}^{-1}\alpha )$ for $f'$. A solution for this triple gives a sheaf $\mathcal{F}$ on $Y_{\acute{e}tale}$ with $a_ Y^{-1}\mathcal{F} = \mathcal{H}$. By the first paragraph of the proof this means the triple is in the essential image. This reduces us to the case where both $X$ and $Y$ are schemes. This case follows from Étale Cohomology, Lemma 58.94.4 via the discussion in Section 82.3 and in particular Lemma 82.3.1. $\square$


Comments (0)


Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 0DGE. Beware of the difference between the letter 'O' and the digit '0'.