The Stacks project

\begin{equation*} \DeclareMathOperator\Coim{Coim} \DeclareMathOperator\Coker{Coker} \DeclareMathOperator\Ext{Ext} \DeclareMathOperator\Hom{Hom} \DeclareMathOperator\Im{Im} \DeclareMathOperator\Ker{Ker} \DeclareMathOperator\Mor{Mor} \DeclareMathOperator\Ob{Ob} \DeclareMathOperator\Sh{Sh} \DeclareMathOperator\SheafExt{\mathcal{E}\mathit{xt}} \DeclareMathOperator\SheafHom{\mathcal{H}\mathit{om}} \DeclareMathOperator\Spec{Spec} \newcommand\colim{\mathop{\mathrm{colim}}\nolimits} \newcommand\lim{\mathop{\mathrm{lim}}\nolimits} \newcommand\Qcoh{\mathit{Qcoh}} \newcommand\Sch{\mathit{Sch}} \newcommand\QCohstack{\mathcal{QC}\!\mathit{oh}} \newcommand\Cohstack{\mathcal{C}\!\mathit{oh}} \newcommand\Spacesstack{\mathcal{S}\!\mathit{paces}} \newcommand\Quotfunctor{\mathrm{Quot}} \newcommand\Hilbfunctor{\mathrm{Hilb}} \newcommand\Curvesstack{\mathcal{C}\!\mathit{urves}} \newcommand\Polarizedstack{\mathcal{P}\!\mathit{olarized}} \newcommand\Complexesstack{\mathcal{C}\!\mathit{omplexes}} \newcommand\Pic{\mathop{\mathrm{Pic}}\nolimits} \newcommand\Picardstack{\mathcal{P}\!\mathit{ic}} \newcommand\Picardfunctor{\mathrm{Pic}} \newcommand\Deformationcategory{\mathcal{D}\!\mathit{ef}} \end{equation*}

Lemma 54.89.7. Let $K$ be a field. Let $n \geq 1$ be invertible in $K$. Consider a commutative diagram

\[ \xymatrix{ X \ar[d] & X' \ar[l]^ p \ar[d]_{f'} & Y \ar[l]^ h \ar[d]^ e \\ \mathop{\mathrm{Spec}}(K) & S' \ar[l] & T \ar[l]_ g } \]

of schemes with $X' = X \times _{\mathop{\mathrm{Spec}}(K)} S'$ and $Y = X' \times _{S'} T$ and $g$ quasi-compact and quasi-separated. The canonical map

\[ p^{-1}E \otimes _{\mathbf{Z}/n\mathbf{Z}}^\mathbf {L} (f')^{-1}Rg_*F \longrightarrow Rh_*(h^{-1}p^{-1}E \otimes _{\mathbf{Z}/n\mathbf{Z}}^\mathbf {L} e^{-1}F) \]

is an isomorphism if $E$ in $D^+(X_{\acute{e}tale}, \mathbf{Z}/n\mathbf{Z})$ has tor amplitude in $[a, \infty ]$ for some $a \in \mathbf{Z}$ and $F$ in $D^+(T_{\acute{e}tale}, \mathbf{Z}/n\mathbf{Z})$.

Proof. This lemma is a generalization of Lemma 54.89.6 to objects of the derived category; the assertion of our lemma is true because in Lemma 54.89.6 the scheme $X$ over $K$ is arbitrary. We strongly urge the reader to skip the laborious proof (alternative: read only the last paragraph).

We may represent $E$ by a bounded below K-flat complex $\mathcal{E}^\bullet $ consisting of flat $\mathbf{Z}/n\mathbf{Z}$-modules. See Cohomology on Sites, Lemma 21.44.4. Choose an integer $b$ such that $H^ i(F) = 0$ for $i < b$. Choose a large integer $N$ and consider the short exact sequence

\[ 0 \to \sigma _{\geq N + 1}\mathcal{E}^\bullet \to \mathcal{E}^\bullet \to \sigma _{\leq N}\mathcal{E}^\bullet \to 0 \]

of stupid truncations. This produces a distinguished triangle $E'' \to E \to E' \to E''[1]$ in $D(X_{\acute{e}tale}, \mathbf{Z}/n\mathbf{Z})$. For fixed $F$ both sides of the arrow in the statement of the lemma are exact functors in $E$. Observe that

\[ p^{-1}E'' \otimes _{\mathbf{Z}/n\mathbf{Z}}^\mathbf {L} (f')^{-1}Rg_*F \quad \text{and}\quad Rh_*(h^{-1}p^{-1}E'' \otimes _{\mathbf{Z}/n\mathbf{Z}}^\mathbf {L} e^{-1}F) \]

are sitting in degrees $\geq N + b$. Hence, if we can prove the lemma for the object $E'$, then we see that the lemma holds in degrees $\leq N + b$ and we will conclude. Some details omitted. Thus we may assume $E$ is represented by a bounded complex of flat $\mathbf{Z}/n\mathbf{Z}$-modules. Doing another argument of the same nature, we may assume $E$ is given by a single flat $\mathbf{Z}/n\mathbf{Z}$-module $\mathcal{E}$.

Next, we use the same arguments for the variable $F$ to reduce to the case where $F$ is given by a single sheaf of $\mathbf{Z}/n\mathbf{Z}$-modules $\mathcal{F}$. Say $\mathcal{F}$ is annihilated by an integer $m | n$. If $\ell $ is a prime number dividing $m$ and $m > \ell $, then we can look at the short exact sequence $0 \to \mathcal{F}[\ell ] \to \mathcal{F} \to \mathcal{F}/\mathcal{F}[\ell ] \to 0$ and reduce to smaller $m$. This finally reduces us to the case where $\mathcal{F}$ is annihilated by a prime number $\ell $ dividing $n$. In this case observe that

\[ p^{-1}\mathcal{E} \otimes _{\mathbf{Z}/n\mathbf{Z}}^\mathbf {L} (f')^{-1}Rg_*\mathcal{F} = p^{-1}(\mathcal{E}/\ell \mathcal{E}) \otimes _{\mathbf{F}_\ell }^\mathbf {L} (f')^{-1}Rg_*\mathcal{F} \]

by the flatness of $\mathcal{E}$. Similarly for the other term. This reduces us to the case where we are working with sheaves of $\mathbf{F}_\ell $-vector spaces which is discussed

Assume $\ell $ is a prime number invertible in $K$. Assume $\mathcal{E}$, $\mathcal{F}$ are sheaves of $\mathbf{F}_\ell $-vector spaces on $X_{\acute{e}tale}$ and $T_{\acute{e}tale}$. We want to show that

\[ p^{-1}\mathcal{E} \otimes _{\mathbf{F}_\ell } (f')^{-1}R^ qg_*\mathcal{F} \longrightarrow R^ qh_*(h^{-1}p^{-1}\mathcal{E} \otimes _{\mathbf{F}_\ell } e^{-1}\mathcal{F}) \]

is an isomorphism for every $q \geq 0$. This question is local on $X$ hence we may assume $X$ is affine. We can write $\mathcal{E}$ as a filtered colimit of constructible sheaves of $\mathbf{F}_\ell $-vector spaces on $X_{\acute{e}tale}$, see Lemma 54.72.2. Since tensor products commute with filtered colimits and since higher direct images do too (Lemma 54.51.7) we may assume $\mathcal{E}$ is a constructible sheaf of $\mathbf{F}_\ell $-vector spaces on $X_{\acute{e}tale}$. Then we can choose an integer $m$ and finite and finitely presented morphisms $\pi _ i : X_ i \to X$, $i = 1, \ldots , m$ such that there is an injective map

\[ \mathcal{E} \to \bigoplus \nolimits _{i = 1, \ldots , m} \pi _{i, *}\mathbf{F}_\ell \]

See Lemma 54.73.4. Observe that the direct sum is a constructible sheaf as well (Lemma 54.72.9). Thus the cokernel is constructible too (Lemma 54.70.6). By dimension shifting, i.e., induction on $q$, on the category of constructible sheaves of $\mathbf{F}_\ell $-vector spaces on $X_{\acute{e}tale}$, it suffices to prove the result for the sheaves $\pi _{i, *}\mathbf{F}_\ell $ (details omitted; hint: start with proving injectivity for $q = 0$ for all constructible $\mathcal{E}$). To prove this case we extend the diagram of the lemma to

\[ \xymatrix{ X_ i \ar[d]^{\pi _ i} & X'_ i \ar[l]^{p_ i} \ar[d]^{\pi '_ i} & Y_ i \ar[l]^{h_ i} \ar[d]^{\rho _ i} \\ X \ar[d] & X' \ar[l]^ p \ar[d]_{f'} & Y \ar[l]^ h \ar[d]^ e \\ \mathop{\mathrm{Spec}}(K) & S' \ar[l] & T \ar[l]_ g } \]

with all squares cartesian. In the equations below we are going to use that $R\pi _{i, *} = \pi _{i, *}$ and similarly for $\pi '_ i$, $\rho _ i$, we are going to use the Leray spectral sequence, we are going to use Lemma 54.54.3, and we are going to use Lemma 54.88.6 (although this lemma is almost trivial for finite morphisms) for $\pi _ i$, $\pi '_ i$, $\rho _ i$. Doing so we see that

\begin{align*} p^{-1}\pi _{i, *}\mathbf{F}_\ell \otimes _{\mathbf{F}_\ell } (f')^{-1}R^ qg_*\mathcal{F} & = \pi '_{i, *}\mathbf{F}_\ell \otimes _{\mathbf{F}_\ell } (f')^{-1}R^ qg_*\mathcal{F} \\ & = \pi '_{i, *}((\pi '_ i)^{-1} (f')^{-1}R^ qg_*\mathcal{F}) \end{align*}

Similarly, we have

\begin{align*} R^ qh_*(h^{-1}p^{-1} \pi _{i, *}\mathbf{F}_\ell \otimes _{\mathbf{F}_\ell } e^{-1}\mathcal{F}) & = R^ qh_*(\rho _{i, *}\mathbf{F}_\ell \otimes _{\mathbf{F}_\ell } e^{-1}\mathcal{F}) \\ & = R^ qh_*(\rho _ i^{-1}e^{-1}\mathcal{F}) \\ & = \pi '_{i, *}R^ qh_{i, *} \rho _ i^{-1}e^{-1}\mathcal{F}) \end{align*}

Simce $R^ qh_{i, *} \rho _ i^{-1}e^{-1}\mathcal{F} = (\pi '_ i)^{-1} (f')^{-1}R^ qg_*\mathcal{F}$ by Lemma 54.89.6 we conclude. $\square$


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 0F1J. Beware of the difference between the letter 'O' and the digit '0'.