## 102.9 Parasitic modules

The following definition is compatible with Descent, Definition 35.12.1.

Definition 102.9.1. Let $\mathcal{X}$ be an algebraic stack. A presheaf of $\mathcal{O}_\mathcal {X}$-modules $\mathcal{F}$ is parasitic if we have $\mathcal{F}(x) = 0$ for any object $x$ of $\mathcal{X}$ which lies over a scheme $U$ such that the corresponding morphism $x : U \to \mathcal{X}$ is flat.

Here is a lemma with some properties of this notion.

Lemma 102.9.2. Let $\mathcal{X}$ be an algebraic stack. Let $\mathcal{F}$ be a presheaf of $\mathcal{O}_\mathcal {X}$-modules.

1. If $\mathcal{F}$ is parasitic and $g : \mathcal{Y} \to \mathcal{X}$ is a flat morphism of algebraic stacks, then $g^*\mathcal{F}$ is parasitic.

2. For $\tau \in \{ Zariski, {\acute{e}tale}, smooth, syntomic, fppf\}$ we have

1. the $\tau$ sheafification of a parasitic presheaf of modules is parasitic, and

2. the full subcategory of $\textit{Mod}(\mathcal{X}_\tau , \mathcal{O}_\mathcal {X})$ consisting of parasitic modules is a Serre subcategory.

3. Suppose $\mathcal{F}$ is a sheaf for the étale topology. Let $f_ i : \mathcal{X}_ i \to \mathcal{X}$ be a family of smooth morphisms of algebraic stacks such that $|\mathcal{X}| = \bigcup _ i |f_ i|(|\mathcal{X}_ i|)$. If each $f_ i^*\mathcal{F}$ is parasitic then so is $\mathcal{F}$.

4. Suppose $\mathcal{F}$ is a sheaf for the fppf topology. Let $f_ i : \mathcal{X}_ i \to \mathcal{X}$ be a family of flat and locally finitely presented morphisms of algebraic stacks such that $|\mathcal{X}| = \bigcup _ i |f_ i|(|\mathcal{X}_ i|)$. If each $f_ i^*\mathcal{F}$ is parasitic then so is $\mathcal{F}$.

Proof. To see part (1) let $y$ be an object of $\mathcal{Y}$ which lies over a scheme $V$ such that the corresponding morphism $y : V \to \mathcal{Y}$ is flat. Then $g(y) : V \to \mathcal{Y} \to \mathcal{X}$ is flat as a composition of flat morphisms (see Morphisms of Stacks, Lemma 100.25.2) hence $\mathcal{F}(g(y))$ is zero by assumption. Since $g^*\mathcal{F} = g^{-1}\mathcal{F}(y) = \mathcal{F}(g(y))$ we conclude $g^*\mathcal{F}$ is parasitic.

To see part (2)(a) note that if $\{ x_ i \to x\}$ is a $\tau$-covering of $\mathcal{X}$, then each of the morphisms $x_ i \to x$ lies over a flat morphism of schemes. Hence if $x$ lies over a scheme $U$ such that $x : U \to \mathcal{X}$ is flat, so do all of the objects $x_ i$. Hence the presheaf $\mathcal{F}^+$ (see Sites, Section 7.10) is parasitic if the presheaf $\mathcal{F}$ is parasitic. This proves (2)(a) as the sheafification of $\mathcal{F}$ is $(\mathcal{F}^+)^+$.

Let $\mathcal{F}$ be a parasitic $\tau$-module. It is immediate from the definitions that any submodule of $\mathcal{F}$ is parasitic. On the other hand, if $\mathcal{F}' \subset \mathcal{F}$ is a submodule, then it is equally clear that the presheaf $x \mapsto \mathcal{F}(x)/\mathcal{F}'(x)$ is parasitic. Hence the quotient $\mathcal{F}/\mathcal{F}'$ is a parasitic module by (2)(a). Finally, we have to show that given a short exact sequence $0 \to \mathcal{F}_1 \to \mathcal{F}_2 \to \mathcal{F}_3 \to 0$ with $\mathcal{F}_1$ and $\mathcal{F}_3$ parasitic, then $\mathcal{F}_2$ is parasitic. This follows immediately on evaluating on $x$ lying over a scheme flat over $\mathcal{X}$. This proves (2)(b), see Homology, Lemma 12.10.2.

Let $f_ i : \mathcal{X}_ i \to \mathcal{X}$ be a jointly surjective family of smooth morphisms of algebraic stacks and assume each $f_ i^*\mathcal{F}$ is parasitic. Let $x$ be an object of $\mathcal{X}$ which lies over a scheme $U$ such that $x : U \to \mathcal{X}$ is flat. Consider a surjective smooth covering $W_ i \to U \times _{x, \mathcal{X}} \mathcal{X}_ i$. Denote $y_ i : W_ i \to \mathcal{X}_ i$ the projection. It follows that $\{ f_ i(y_ i) \to x\}$ is a covering for the smooth topology on $\mathcal{X}$. Since a composition of flat morphisms is flat we see that $f_ i^*\mathcal{F}(y_ i) = 0$. On the other hand, as we saw in the proof of (1), we have $f_ i^*\mathcal{F}(y_ i) = \mathcal{F}(f_ i(y_ i))$. Hence we see that for some smooth covering $\{ x_ i \to x\} _{i \in I}$ in $\mathcal{X}$ we have $\mathcal{F}(x_ i) = 0$. This implies $\mathcal{F}(x) = 0$ because the smooth topology is the same as the étale topology, see More on Morphisms, Lemma 37.38.7. Namely, $\{ x_ i \to x\} _{i \in I}$ lies over a smooth covering $\{ U_ i \to U\} _{i \in I}$ of schemes. By the lemma just referenced there exists an étale covering $\{ V_ j \to U\} _{j \in J}$ which refines $\{ U_ i \to U\} _{i \in I}$. Denote $x'_ j = x|_{V_ j}$. Then $\{ x'_ j \to x\}$ is an étale covering in $\mathcal{X}$ refining $\{ x_ i \to x\} _{i \in I}$. This means the map $\mathcal{F}(x) \to \prod _{j \in J} \mathcal{F}(x'_ j)$, which is injective as $\mathcal{F}$ is a sheaf in the étale topology, factors through $\mathcal{F}(x) \to \prod _{i \in I} \mathcal{F}(x_ i)$ which is zero. Hence $\mathcal{F}(x) = 0$ as desired.

Proof of (4): omitted. Hint: similar, but simpler, than the proof of (3). $\square$

Parasitic modules are preserved under absolutely any pushforward.

Lemma 102.9.3. Let $\tau \in \{ {\acute{e}tale}, fppf\}$. Let $\mathcal{X}$ be an algebraic stack. Let $\mathcal{F}$ be a parasitic object of $\textit{Mod}(\mathcal{X}_\tau , \mathcal{O}_\mathcal {X})$.

1. $H^ i_\tau (\mathcal{X}, \mathcal{F}) = 0$ for all $i$.

2. Let $f : \mathcal{X} \to \mathcal{Y}$ be a morphism of algebraic stacks. Then $R^ if_*\mathcal{F}$ (computed in $\tau$-topology) is a parasitic object of $\textit{Mod}(\mathcal{Y}_\tau , \mathcal{O}_\mathcal {Y})$.

Proof. We first reduce (2) to (1). By Sheaves on Stacks, Lemma 95.21.2 we see that $R^ if_*\mathcal{F}$ is the sheaf associated to the presheaf

$y \longmapsto H^ i_\tau \Big(V \times _{y, \mathcal{Y}} \mathcal{X}, \ \text{pr}^{-1}\mathcal{F}\Big)$

Here $y$ is a typical object of $\mathcal{Y}$ lying over the scheme $V$. By Lemma 102.9.2 it suffices to show that these cohomology groups are zero when $y : V \to \mathcal{Y}$ is flat. Note that $\text{pr} : V \times _{y, \mathcal{Y}} \mathcal{X} \to \mathcal{X}$ is flat as a base change of $y$. Hence by Lemma 102.9.2 we see that $\text{pr}^{-1}\mathcal{F}$ is parasitic. Thus it suffices to prove (1).

To see (1) we can use the spectral sequence of Sheaves on Stacks, Proposition 95.20.1 to reduce this to the case where $\mathcal{X}$ is an algebraic stack representable by an algebraic space. Note that in the spectral sequence each $f_ p^{-1}\mathcal{F} = f_ p^*\mathcal{F}$ is a parasitic module by Lemma 102.9.2 because the morphisms $f_ p : \mathcal{U}_ p = \mathcal{U} \times _\mathcal {X} \ldots \times _\mathcal {X} \mathcal{U} \to \mathcal{X}$ are flat. Reusing this spectral sequence one more time (as in the proof of Lemma 102.5.1) we reduce to the case where the algebraic stack $\mathcal{X}$ is representable by a scheme $X$. Then $H^ i_\tau (\mathcal{X}, \mathcal{F}) = H^ i((\mathit{Sch}/X)_\tau , \mathcal{F})$. In this case the vanishing follows easily from an argument with Čech coverings, see Descent, Lemma 35.12.2. $\square$

The following lemma is one of the major reasons we care about parasitic modules. To understand the statement, recall that the functors $\mathit{QCoh}(\mathcal{O}_\mathcal {X}) \to \textit{Mod}(\mathcal{X}_{\acute{e}tale}, \mathcal{O}_\mathcal {X})$ and $\mathit{QCoh}(\mathcal{O}_\mathcal {X}) \to \textit{Mod}(\mathcal{O}_\mathcal {X})$ aren't exact in general.

Lemma 102.9.4. Let $\mathcal{X}$ be an algebraic stack. Let $\alpha : \mathcal{F} \to \mathcal{G}$ and $\beta : \mathcal{G} \to \mathcal{H}$ be maps in $\mathit{QCoh}(\mathcal{O}_\mathcal {X})$ with $\beta \circ \alpha = 0$. The following are equivalent:

1. in the abelian category $\mathit{QCoh}(\mathcal{O}_\mathcal {X})$ the complex $\mathcal{F} \to \mathcal{G} \to \mathcal{H}$ is exact at $\mathcal{G}$,

2. $\mathop{\mathrm{Ker}}(\beta )/\mathop{\mathrm{Im}}(\alpha )$ computed in either $\textit{Mod}(\mathcal{X}_{\acute{e}tale}, \mathcal{O}_\mathcal {X})$ or $\textit{Mod}(\mathcal{X}_{fppf}, \mathcal{O}_\mathcal {X})$ is parasitic.

Proof. We have $\mathit{QCoh}(\mathcal{O}_\mathcal {X}) \subset \textit{LQCoh}^{fbc}(\mathcal{O}_\mathcal {X})$, see Section 102.8. Hence $\mathop{\mathrm{Ker}}(\beta )/\mathop{\mathrm{Im}}(\alpha )$ computed in $\textit{Mod}(\mathcal{X}_{\acute{e}tale}, \mathcal{O}_\mathcal {X})$ or $\textit{Mod}(\mathcal{X}_{fppf}, \mathcal{O}_\mathcal {X})$ agree, see Proposition 102.8.1. From now on we will use the étale topology on $\mathcal{X}$.

Let $\mathcal{E}$ be the cohomology of $\mathcal{F} \to \mathcal{G} \to \mathcal{H}$ computed in the abelian category $\mathit{QCoh}(\mathcal{O}_\mathcal {X})$. Let $x : U \to \mathcal{X}$ be a flat morphism where $U$ is a scheme. As we are using the étale topology, the restriction functor $\textit{Mod}(\mathcal{X}_{\acute{e}tale}, \mathcal{O}_\mathcal {X}) \to \textit{Mod}(U_{\acute{e}tale}, \mathcal{O}_ U)$ is exact. On the other hand, by Lemma 102.4.1 and Sheaves on Stacks, Lemma 95.14.2 the restriction functor

$\mathit{QCoh}(\mathcal{O}_\mathcal {X}) \xrightarrow {x^*} \mathit{QCoh}((\mathit{Sch}/U)_{\acute{e}tale}, \mathcal{O}) \xrightarrow {{-}|_{U_{\acute{e}tale}}} \mathit{QCoh}(U_{\acute{e}tale}, \mathcal{O}_ U)$

is exact too. We conclude that $\mathcal{E}|_{U_{\acute{e}tale}} = (\mathop{\mathrm{Ker}}(\beta )/\mathop{\mathrm{Im}}(\alpha ))|_{U_{\acute{e}tale}}$.

If (1) holds, then $\mathcal{E} = 0$ hence $\mathop{\mathrm{Ker}}(\beta )/\mathop{\mathrm{Im}}(\alpha )$ restricts to zero on $U_{\acute{e}tale}$ for all $U$ flat over $\mathcal{X}$ and this is the definition of a parasitic module. If (2) holds, then $\mathop{\mathrm{Ker}}(\beta )/\mathop{\mathrm{Im}}(\alpha )$ restricts to zero on $U_{\acute{e}tale}$ for all $U$ flat over $\mathcal{X}$ hence $\mathcal{E}$ restricts to zero on $U_{\acute{e}tale}$ for all $U$ flat over $\mathcal{X}$. This certainly implies that the quasi-coherent module $\mathcal{E}$ is zero, for example apply Lemma 102.4.2 to the map $0 \to \mathcal{E}$. $\square$

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).