The Stacks project

60.26 Frobenius action on crystalline cohomology

In this section we prove that Frobenius pullback induces a quasi-isomorphism on crystalline cohomology after inverting the prime $p$. But in order to even formulate this we need to work in a special situation.

Situation 60.26.1. In Situation 60.7.5 assume the following

  1. $S = \mathop{\mathrm{Spec}}(A)$ for some divided power ring $(A, I, \gamma )$ with $p \in I$,

  2. there is given a homomorphism of divided power rings $\sigma : A \to A$ such that $\sigma (x) = x^ p \bmod pA$ for all $x \in A$.

In Situation 60.26.1 the morphism $\mathop{\mathrm{Spec}}(\sigma ) : S \to S$ is a lift of the absolute Frobenius $F_{S_0} : S_0 \to S_0$ and since the diagram

\[ \xymatrix{ X \ar[d] \ar[r]_{F_ X} & X \ar[d] \\ S_0 \ar[r]^{F_{S_0}} & S_0 } \]

is commutative where $F_ X : X \to X$ is the absolute Frobenius morphism of $X$. Thus we obtain a morphism of crystalline topoi

\[ (F_ X)_{\text{cris}} : (X/S)_{\text{cris}} \longrightarrow (X/S)_{\text{cris}} \]

see Remark 60.9.3. Here is the terminology concerning $F$-crystals following the notation of Saavedra, see [Saavedra].

Definition 60.26.2. In Situation 60.26.1 an $F$-crystal on $X/S$ (relative to $\sigma $) is a pair $(\mathcal{E}, F_\mathcal {E})$ given by a crystal in finite locally free $\mathcal{O}_{X/S}$-modules $\mathcal{E}$ together with a map

\[ F_\mathcal {E} : (F_ X)_{\text{cris}}^*\mathcal{E} \longrightarrow \mathcal{E} \]

An $F$-crystal is called nondegenerate if there exists an integer $i \geq 0$ a map $V : \mathcal{E} \to (F_ X)_{\text{cris}}^*\mathcal{E}$ such that $V \circ F_{\mathcal{E}} = p^ i \text{id}$.

Remark 60.26.3. Let $(\mathcal{E}, F)$ be an $F$-crystal as in Definition 60.26.2. In the literature the nondegeneracy condition is often part of the definition of an $F$-crystal. Moreover, often it is also assumed that $F \circ V = p^ n\text{id}$. What is needed for the result below is that there exists an integer $j \geq 0$ such that $\mathop{\mathrm{Ker}}(F)$ and $\mathop{\mathrm{Coker}}(F)$ are killed by $p^ j$. If the rank of $\mathcal{E}$ is bounded (for example if $X$ is quasi-compact), then both of these conditions follow from the nondegeneracy condition as formulated in the definition. Namely, suppose $R$ is a ring, $r \geq 1$ is an integer and $K, L \in \text{Mat}(r \times r, R)$ are matrices with $K L = p^ i 1_{r \times r}$. Then $\det (K)\det (L) = p^{ri}$. Let $L'$ be the adjugate matrix of $L$, i.e., $L' L = L L' = \det (L)$. Set $K' = p^{ri} K$ and $j = ri + i$. Then we have $K' L = p^ j 1_{r \times r}$ as $K L = p^ i$ and

\[ L K' = L K \det (L) \det (M) = L K L L' \det (M) = L p^ i L' \det (M) = p^ j 1_{r \times r} \]

It follows that if $V$ is as in Definition 60.26.2 then setting $V' = p^ N V$ where $N > i \cdot \text{rank}(\mathcal{E})$ we get $V' \circ F = p^{N + i}$ and $F \circ V' = p^{N + i}$.

Theorem 60.26.4. In Situation 60.26.1 let $(\mathcal{E}, F_\mathcal {E})$ be a nondegenerate $F$-crystal. Assume $A$ is a $p$-adically complete Noetherian ring and that $X \to S_0$ is proper smooth. Then the canonical map

\[ F_\mathcal {E} \circ (F_ X)_{\text{cris}}^* : R\Gamma (\text{Cris}(X/S), \mathcal{E}) \otimes ^\mathbf {L}_{A, \sigma } A \longrightarrow R\Gamma (\text{Cris}(X/S), \mathcal{E}) \]

becomes an isomorphism after inverting $p$.

Proof. We first write the arrow as a composition of three arrows. Namely, set

\[ X^{(1)} = X \times _{S_0, F_{S_0}} S_0 \]

and denote $F_{X/S_0} : X \to X^{(1)}$ the relative Frobenius morphism. Denote $\mathcal{E}^{(1)}$ the base change of $\mathcal{E}$ by $\mathop{\mathrm{Spec}}(\sigma )$, in other words the pullback of $\mathcal{E}$ to $\text{Cris}(X^{(1)}/S)$ by the morphism of crystalline topoi associated to the commutative diagram

\[ \xymatrix{ X^{(1)} \ar[r] \ar[d] & X \ar[d] \\ S \ar[r]^{\mathop{\mathrm{Spec}}(\sigma )} & S } \]

Then we have the base change map

60.26.4.1
\begin{equation} \label{crystalline-equation-base-change-sigma} R\Gamma (\text{Cris}(X/S), \mathcal{E}) \otimes ^\mathbf {L}_{A, \sigma } A \longrightarrow R\Gamma (\text{Cris}(X^{(1)}/S), \mathcal{E}^{(1)}) \end{equation}

see Remark 60.24.8. Note that the composition of $F_{X/S_0} : X \to X^{(1)}$ with the projection $X^{(1)} \to X$ is the absolute Frobenius morphism $F_ X$. Hence we see that $F_{X/S_0}^*\mathcal{E}^{(1)} = (F_ X)_{\text{cris}}^*\mathcal{E}$. Thus pullback by $F_{X/S_0}$ is a map

60.26.4.2
\begin{equation} \label{crystalline-equation-to-prove} F_{X/S_0}^* : R\Gamma (\text{Cris}(X^{(1)}/S), \mathcal{E}^{(1)}) \longrightarrow R\Gamma (\text{Cris}(X/S), (F_ X)^*_{\text{cris}}\mathcal{E}) \end{equation}

Finally we can use $F_\mathcal {E}$ to get a map

60.26.4.3
\begin{equation} \label{crystalline-equation-F-E} R\Gamma (\text{Cris}(X/S), (F_ X)^*_{\text{cris}}\mathcal{E}) \longrightarrow R\Gamma (\text{Cris}(X/S), \mathcal{E}) \end{equation}

The map of the theorem is the composition of the three maps (60.26.4.1), (60.26.4.2), and (60.26.4.3) above. The first is a quasi-isomorphism modulo all powers of $p$ by Remark 60.24.9. Hence it is a quasi-isomorphism since the complexes involved are perfect in $D(A)$ see Remark 60.24.13. The third map is a quasi-isomorphism after inverting $p$ simply because $F_\mathcal {E}$ has an inverse up to a power of $p$, see Remark 60.26.3. Finally, the second is an isomorphism after inverting $p$ by Lemma 60.25.6. $\square$


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 07N0. Beware of the difference between the letter 'O' and the digit '0'.