proetale

4 The pro-étale topology

4.1 Weakly étale morphisms of schemes

Definition 4.1 Weakly étale morphisms, [ BS15 , Definition 4.1.1 ]
#

A map \(f \colon Y \to X\) is weakly étale if \(f\) is flat and \(\Delta _f\colon Y \to Y \times _{X} Y\) is flat.

Lemma 4.2 Local property of weakly étale morphisms

\(f \colon Y \to X\) is weakly étale if and only if for every affine open \(U\) of \(Y\) and every affine open \(V\) of \(X\) with \(f(U) \subset V\), the induced ring homomorphism \(\Gamma (X, V) \to \Gamma (Y, U)\) is weakly étale.

Lemma 4.3

Etale morphisms are weakly étale.

Proof

Any étale morphism is flat and its diagonal is an open immersion, hence flat.

Weakly étale is stable under composition and base change.

Proof

This follows because flat is stable under composition and base change.

4.2 The site

Definition 4.5 Quasi-compact cover
#

A jointly surjective family \((f_i \colon Y_i \to X)_{i \in I}\) of morphisms of schemes is quasi-compact if for every affine open \(U\) of \(X\) there exist quasi-compact opens \(V_i\) in \(Y_i\) such that \(U = \bigcup _{i \in I} f_i(V_i)\).

Lemma 4.6

Let \((f_i \colon Y_i \to X)_{i \in I}\) be a jointly-surjective family of morphisms of schemes. If \(f_i\) is flat and locally of finite presentation for every \(i \in I\), the family is quasi-compact.

Proof

In this case, every \(f_i\) is an open map, from which the claim easily follows.

Definition 4.7
#

Let \(\mathcal{P}\) be a morphism property of schemes and \(X\) be a scheme. The small \(\mathcal{P}\)-site of \(X\) is the category of \(X\)-schemes with structure morphism satisfying \(\mathcal{P}\) and with covers given by quasi-compact, jointly surjective families of morphisms satisfying \(\mathcal{P}\).

Definition 4.8 The fpqc site

Let \(X\) be a scheme. The fqpc site of \(X\), denoted by \(X_{\mathrm{fpqc}}\), is the \(\mathcal{P}\)-site of \(X\) with \(\mathcal{P} = \text{flat}\).

Proposition 4.9

Let \(F\) be a presheaf on the category of schemes. Then \(F\) is a sheaf in the fpqc topology if and only if it is a sheaf in the Zariski topology and satisfies the sheaf property for \(\{ V \to U\} \) with \(V\), \(U\) affine and \(V \to U\) faithfully flat.

Theorem 4.10

The fpqc site is subcanonical, i.e., every representable presheaf is a sheaf.

Proof

TBA.

Definition 4.11 The pro-étale site, [ BS15 , Definition 4.1.1 ]

Let \(X\) be a scheme. The pro-étale site of \(X\), denoted by \(X_{\mathrm{proét}}\), is the \(\mathcal{P}\)-site of \(X\) with \(\mathcal{P} = \text{weakly étale}\).

Remark 4.12
#

Despite the name proétale, Definition 4.11 does not mention pro-étale morphisms. The name is justified (see [ BS15 , Remark 4.1.3 ] ), but we don’t formalise the justification (yet).

Proposition 4.13 [ BS15 , Lemma 4.2.4 ]

The site \(X_{\mathrm{proét}}\) is subcanonical.

Proof

Follows from 4.10, since every pro-étale sheaf is an fpqc-sheaf.

4.2.1 Affine covers

Definition 4.14 [ BS15 , Definition 4.2.1 ]
#

An object \(U\) of \(X_{\mathrm{proét}}\) is called a pro-étale affine if it can be written as \(U = \lim U_i\) over a small cofiltered diagram \(i \mapsto U_i\) of affine schemes étale over \(X\). The subcategory of pro-étale affines in \(X_{\mathrm{proét}}\) is denoted by \(X^{\mathrm{aff}}_{\mathrm{proét}}\).

Lemma 4.15 [ BS15 , Remark 4.2.3 ]

The category \(X^{\mathrm{aff}}_{\mathrm{proét}}\) admits limits indexed by a connected diagram and the forgetful functor to \(\mathrm{Sch} / X\) preserves them. If \(X\) is affine, \(X^{\mathrm{aff}}_{\mathrm{proét}}\) has all small limits.

Definition 4.16

Let \(X\) be affine. The affine pro-étale site of \(X\) is the category \(X^{\mathrm{aff}}_{\mathrm{proét}}\) with the covering families are the quasi-compact covers.

Lemma 4.17

The topos \(\mathrm{Shv}(X_{\mathrm{et}})\) is generated by the affine étale site of \(X\), i.e., the full subcategory of \(X_{\mathrm{et}}\) consisting of affine, étale schemes over \(X\).

Proof

TBA.

Lemma 4.18 [ BS15 , Lemma 4.2.4 ]

The topos \(\mathrm{Shv}(X_{\mathrm{proét}})\) is generated by \(X^{\mathrm{aff}}_{\mathrm{proét}}\).

Proof

TBA.

4.2.2 Repleteness

An important property of the pro-étale site is the following:

Proposition 4.19 [ BS15 , Proposition 4.2.8 ]

The category of sheaves on \(X_{\mathrm{proét}}\) has enough weakly contractible objects.

Proof

TBA.

Corollary 4.20

The category of sheaves on \(X_{\mathrm{proét}}\) is replete.

4.3 Comparison with étale cohomology

Let \(X\) be a scheme. Since an étale map is also weakly étale by 4.3 we may define:

Definition 4.21

We denote by \(\nu = \nu _X\) the morphism of sites \(X_{\mathrm{proét}} \to X_{\mathrm{et}}\) induced by the functor from \(X_{\mathrm{et}}\) to \(X_{\mathrm{proét}}\) given by \((U \to X) \mapsto (U \to X)\).

Lemma 4.22

The morphism of sites \(\nu \) is continuous.

Proof

This follows from the definitions, because every étale covering is a quasi-compact covering by 4.6.

We denote by \(\nu _{p}\colon \mathrm{PreShv}(X_{\mathrm{proét}}) \to \mathrm{PreShv}(X_{\mathrm{et}})\) and \(\nu ^{p}\colon \mathrm{PreShv}(X_{\mathrm{et}}) \to \mathrm{PreShv}(X_{\mathrm{proét}})\) the pushforward and pullback functors on the level of presheafs.

Lemma 4.23

Let \(F\) be a presheaf on \(X_{\mathrm{et}}\). Then \((\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}\) preserves filtered colimits.

Proof

Let \(U = \lim _{\lambda } U_{\lambda }\) be a cofiltered limit. By definition of \(X^{\mathrm{aff}}_{\mathrm{proét}}\) and because pro-pro-étale is pro-étale, we may assume that the \(U_{\lambda }\) are étale over \(X\). Then

\[ (\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}(\lim U_{\lambda }) = \operatorname {colim}_{\mu } F(V_{\mu }) , \]

where the \(V_{\mu }\) run over factorisations \(U = \lim U_{\lambda } \to V_{\mu } \to X\) where \(V_{\mu } \to X\) is étale. Since the \(U_{\lambda }\) are locally of finite presentation over \(X\), they are finitely presented in the categorical sense. Hence, for every such factorisation there exists a \(\lambda \) such that \(U = \lim U_{\lambda } \to V_{\mu }\) factors via \(U_{\lambda }\). This shows that the \(U_{\lambda }\) lie cofinal in all \(V_{\mu }\), so we obtain

\[ \operatorname {colim}_{\mu } F(V_{\mu }) = \operatorname {colim}_{\lambda } F(U_{\lambda }) = \operatorname {colim}_{\lambda } (\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}(U_{\lambda }) . \]
Lemma 4.24

The topology on \(X^{\mathrm{aff}}_{\mathrm{proét}}\) is generated by relatively pro-representable one-hypercovers with respect to the inclusion \(X^{\mathrm{aff}}_{\mathrm{ét}} \to X^{\mathrm{aff}}_{\mathrm{proét}}\).

Proof

The topology on \(X^{\mathrm{aff}}_{\mathrm{proét}}\) is generated by surjections \(V \to U\) in \(X^{\mathrm{aff}}_{\mathrm{proét}}\) and finite standard Zariski coverings. Both of these can be relatively pro-represented.

Further denote by \(\nu _{*} \colon \mathrm{Shv}(X_{\mathrm{proét}}) \to \mathrm{Shv}(X_{\mathrm{et}})\) and \(\nu ^* \colon \mathrm{Shv}(X_{\mathrm{et}}) \to \mathrm{Shv}(X_{\mathrm{proét}})\) the corresponding pushforward and pullback functors. Since \(\nu ^*\) is exact, we have \(R \nu ^* = \nu ^*\) by abuse of notation and we simply write \(\nu ^*\) everywhere. In contrast to [ BS15 ] , we write \(R \nu _{*}\) for the derived functor of \(\nu _{*}\).

Proposition 4.25

Let \(F\) be a sheaf on \(X_{\mathrm{et}}\). Then \((\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}\) is a sheaf. In particular, \((\nu ^*F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}} \cong (\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}\).

Proof

By 4.24 and 4.23, this follows from 3.17. The in particular follows because we have a commutative diagram

\[ \begin{tikzcd} \affet{X} \arrow{r} \arrow{d} & \affproet{X} \arrow{d} \\ \et{X} \arrow{r} & \proet{X}. \end{tikzcd} \]
Corollary 4.26 [ BS15 , Lemma 5.1.1 ]

Let \(F\) be a sheaf on \(X_{\mathrm{et}}\) and \(U = \lim _i U_i\) in \(X^{\mathrm{aff}}_{\mathrm{proét}}\). Then \(\nu ^*F(U) = \operatorname {colim}_i F(U_i)\).

Proof

By 4.25 we have \((\nu ^*F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}} \cong (\nu ^p F)|_{X^{\mathrm{aff}}_{\mathrm{proét}}}\). The claim now follows from 4.23.

Corollary 4.27

For any \(F \in \mathrm{Shv}(X_{\mathrm{et}})\), the unit \(F \to \nu _{*} \nu ^{*} F\) is an isomorphism.

Proof

By 4.17, we may check this on any affine \(U\) étale over \(X\). Since \(U = \lim U\), the formula of 4.26 shows the desired isomorphism.

Corollary 4.28 [ BS15 , Lemma 5.1.2, fully faithful part ]

\(\nu ^*\) is fully faithful.

Proof

Follows formally from 4.27.

Definition 4.29

A sheaf \(F\) in \(\mathrm{Shv}(X_{\mathrm{proét}})\) is called classical if it lies in the essential image of \(\nu ^{*}\).

We can recognize the classical sheafs among all pro-étale sheafs by checking if they preserve affine filtered colimits:

Corollary 4.30 [ BS15 , Lemma 5.1.2, essential image part ]

A sheaf \(F\) in \(\mathrm{Shv}(X_{\mathrm{proét}})\) is classical if and only if for any \(U \in X^{\mathrm{aff}}_{\mathrm{proét}}\) with presentation \(U = \lim U_i\), it holds that \(F(U) = \operatorname {colim}F(U_i)\).

Proof

The only if part follows directly from Corollary 4.26. Conversely, let \(F\) be in \(\mathrm{Shv}(X_{\mathrm{proét}})\) and suppose \(F\) satisfies the formula in the statement. Then \(\nu ^* \nu _{*} F \to F\) is an isomorphism. Indeed, by Lemma 4.18, we may check this on sections in \(X^{\mathrm{aff}}_{\mathrm{proét}}\). Let \(U = \lim U_i\) be in \(X^{\mathrm{aff}}_{\mathrm{proét}}\). Then again by Corollary 4.26

\[ \nu ^* \nu _{*} F (U) = \operatorname {colim}\nu _{*} F(U_i) = \operatorname {colim}F(U_i) = F(U) \qedhere . \]
Remark 4.31
#

Note that the proof of relies on the highly nontrivial Lemma 4.18.

Lemma 4.32 [ BS15 , Lemma 5.1.6 ]

Let \(K\) be in \(D^+(X_{\mathrm{et}})\) and \(U = \lim _i U_i\) in \(X^{\mathrm{aff}}_{\mathrm{proét}}\). Then \(R \Gamma (U, \nu ^*K) = \operatorname {colim}_i R\Gamma (U_i, K)\).

Proof

TBA.

Proposition 4.33 [ BS15 , Corollary 5.1.6 ]

For any \(K\) in \(D^{+}(X_{\mathrm{et}})\) the unit \(K \to R \nu _{*} \nu ^* K\) is an isomorphism.

Proof

TBA.

Corollary 4.34

Let \(K\) be in \(D^{+}(X_{\mathrm{et}})\). Then

\[ R \Gamma (X_{\mathrm{et}}, K) \cong R \Gamma (X_{\mathrm{proét}}, \nu ^{*} K) . \]
Proof

By 4.33, we have \(R \Gamma (X_{\mathrm{et}}, K) \cong R \Gamma (X_{\mathrm{et}}, R \nu _{*}\nu ^*K)\). Since \(\nu _*\) maps injective sheaves to acyclic ones by general theory, we have \(R \Gamma (X_{\mathrm{et}}, -) \circ \nu _{*} = R (\Gamma (X_{\mathrm{et}}, -) \circ \nu _*) = R \Gamma (X_{\mathrm{proét}}, -)\). So the claim follows.

Definition 4.35 [ Jan88 , § 3 ]
#

Denote by \(\Gamma _{\mathrm{cont}}(X, -)\colon \mathrm{Ab}(X_{\mathrm{et}})^{\mathbb {N}} \to \mathrm{Ab}\) the functor

\[ \left( F_n \right)_{n \in \mathbb {N}} \mapsto \Gamma \left(X, \lim F_n\right). \]

For an inverse system \((F_n)_{n \in \mathbb {N}}\), we call the cohomology groups \(H^i \mathrm{R} \Gamma _{\mathrm{cont}}\left(X, -\right)\left( \left( F_{n} \right)_{n \in \mathbb {N}} \right) \), denoted by \(H^i_{\mathrm{cont}}(X_{\mathrm{et}}, (F_{n})_{n \in \mathbb {N}})\), the continuous étale cohomology of \(X\) with coefficients in \(\left( F_n \right)_{n \in \mathbb {N}}\).

Let \((F_n)_{n \in \mathbb {N}}\) be an inverse system of abelian sheaves on \(X_{\mathrm{et}}\) with epimorphic transition maps. Then there exists a canonical identification

\[ R \Gamma _{cont}(X, (F_n)_n) \cong R \Gamma (X_{\mathrm{proét}}, \lim \nu ^* F_n) . \]

In particular, for every \(i \ge 0\) we have

\[ H^i_{\mathrm{cont}}(X_{\mathrm{et}}, (F_n)_n) \cong H^i(X_{\mathrm{proét}}, \lim \nu ^* F_n) . \]
Proof

We repeatedly use the fact that \(\Gamma \) commutes with inverse limits both on \(X_{\mathrm{et}}\) and \(X_{\mathrm{proét}}\). Then we have

\begin{align*} R \Gamma _{\mathrm{cont}}(X, (F_n)_n) & \cong R \left( \Gamma (X_{\mathrm{et}}, -) \circ \lim \right) ( (F_n)_n ) \\ & \cong R \left( \lim \circ \Gamma (X_{\mathrm{et}}, -) \right) ( (F_n)_n ) \\ & \cong R \lim ( R \Gamma (X_{\mathrm{et}}, F_n)) \\ & \cong R \lim ( R \Gamma (X_{\mathrm{proét}}, \nu ^{*} F_n)) \\ & \cong R \left( \lim \circ \Gamma (X_{\mathrm{proét}}, -) \right) ( (\nu ^* F_n)_n ) \\ & \cong R \left( \Gamma (X_{\mathrm{proét}}, -) \circ \lim \right) ( (\nu ^* F_n)_n ) \\ & \cong R \Gamma (X_{\mathrm{proét}}, R \lim \nu ^* F_n) \\ & \cong R \Gamma (X_{\mathrm{proét}}, \lim \nu ^* F_n) \\ \end{align*}

The fourth isomorphism is 4.34 and the last is 2.7.