## 93.15 Algebraic stacks, overhauled

Some basic results on algebraic stacks.

Lemma 93.15.1. Let $S$ be a scheme contained in $\mathit{Sch}_{fppf}$. Let $f : \mathcal{X} \to \mathcal{Y}$ be a $1$-morphism of algebraic stacks over $S$. Let $V \in \mathop{\mathrm{Ob}}\nolimits ((\mathit{Sch}/S)_{fppf})$. Let $y : (\mathit{Sch}/V)_{fppf} \to \mathcal{Y}$ be surjective and smooth. Then there exists an object $U \in \mathop{\mathrm{Ob}}\nolimits ((\mathit{Sch}/S)_{fppf})$ and a $2$-commutative diagram

$\xymatrix{ (\mathit{Sch}/U)_{fppf} \ar[r]_ a \ar[d]_ x & (\mathit{Sch}/V)_{fppf} \ar[d]^ y \\ \mathcal{X} \ar[r]^ f & \mathcal{Y} }$

with $x$ surjective and smooth.

Proof. First choose $W \in \mathop{\mathrm{Ob}}\nolimits ((\mathit{Sch}/S)_{fppf})$ and a surjective smooth $1$-morphism $z : (\mathit{Sch}/W)_{fppf} \to \mathcal{X}$. As $\mathcal{Y}$ is an algebraic stack we may choose an equivalence

$j : \mathcal{S}_ F \longrightarrow (\mathit{Sch}/W)_{fppf} \times _{f \circ z, \mathcal{Y}, y} (\mathit{Sch}/V)_{fppf}$

where $F$ is an algebraic space. By Lemma 93.10.6 the morphism $\mathcal{S}_ F \to (\mathit{Sch}/W)_{fppf}$ is surjective and smooth as a base change of $y$. Hence by Lemma 93.10.5 we see that $\mathcal{S}_ F \to \mathcal{X}$ is surjective and smooth. Choose an object $U \in \mathop{\mathrm{Ob}}\nolimits ((\mathit{Sch}/S)_{fppf})$ and a surjective étale morphism $U \to F$. Then applying Lemma 93.10.5 once more we obtain the desired properties. $\square$

This lemma is a generalization of Proposition 93.13.3.

Lemma 93.15.2. Let $S$ be a scheme contained in $\mathit{Sch}_{fppf}$. Let $f : \mathcal{X} \to \mathcal{Y}$ be a $1$-morphism of algebraic stacks over $S$. The following are equivalent:

1. for $U \in \mathop{\mathrm{Ob}}\nolimits ((\mathit{Sch}/S)_{fppf})$ the functor $f : \mathcal{X}_ U \to \mathcal{Y}_ U$ is faithful,

2. the functor $f$ is faithful, and

3. $f$ is representable by algebraic spaces.

Proof. Parts (1) and (2) are equivalent by general properties of $1$-morphisms of categories fibred in groupoids, see Categories, Lemma 4.35.8. We see that (3) implies (2) by Lemma 93.9.2. Finally, assume (2). Let $U$ be a scheme. Let $y \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{Y}_ U)$. We have to prove that

$\mathcal{W} = (\mathit{Sch}/U)_{fppf} \times _{y, \mathcal{Y}} \mathcal{X}$

is representable by an algebraic space over $U$. Since $(\mathit{Sch}/U)_{fppf}$ is an algebraic stack we see from Lemma 93.14.3 that $\mathcal{W}$ is an algebraic stack. On the other hand the explicit description of objects of $\mathcal{W}$ as triples $(V, x, \alpha : y(V) \to f(x))$ and the fact that $f$ is faithful, shows that the fibre categories of $\mathcal{W}$ are setoids. Hence Proposition 93.13.3 guarantees that $\mathcal{W}$ is representable by an algebraic space. $\square$

Lemma 93.15.3. Let $S$ be a scheme contained in $\mathit{Sch}_{fppf}$. Let $u : \mathcal{U} \to \mathcal{X}$ be a $1$-morphism of stacks in groupoids over $(\mathit{Sch}/S)_{fppf}$. If

1. $\mathcal{U}$ is representable by an algebraic space, and

2. $u$ is representable by algebraic spaces, surjective and smooth,

then $\mathcal X$ is an algebraic stack over $S$.

Proof. We have to show that $\Delta : \mathcal{X} \to \mathcal{X} \times \mathcal{X}$ is representable by algebraic spaces, see Definition 93.12.1. Given two schemes $T_1$, $T_2$ over $S$ denote $\mathcal{T}_ i = (\mathit{Sch}/T_ i)_{fppf}$ the associated representable fibre categories. Suppose given $1$-morphisms $f_ i : \mathcal{T}_ i \to \mathcal{X}$. According to Lemma 93.10.11 it suffices to prove that the $2$-fibered product $\mathcal{T}_1 \times _\mathcal {X} \mathcal{T}_2$ is representable by an algebraic space. By Stacks, Lemma 8.6.8 this is in any case a stack in setoids. Thus $\mathcal{T}_1 \times _\mathcal {X} \mathcal{T}_2$ corresponds to some sheaf $F$ on $(\mathit{Sch}/S)_{fppf}$, see Stacks, Lemma 8.6.3. Let $U$ be the algebraic space which represents $\mathcal{U}$. By assumption

$\mathcal{T}_ i' = \mathcal{U} \times _{u, \mathcal{X}, f_ i} \mathcal{T}_ i$

is representable by an algebraic space $T'_ i$ over $S$. Hence $\mathcal{T}_1' \times _\mathcal {U} \mathcal{T}_2'$ is representable by the algebraic space $T'_1 \times _ U T'_2$. Consider the commutative diagram

$\xymatrix{ & \mathcal{T}_1 \times _{\mathcal X} \mathcal{T}_2 \ar[rr]\ar '[d][dd] & & \mathcal{T}_1 \ar[dd] \\ \mathcal{T}_1' \times _\mathcal {U} \mathcal{T}_2' \ar[ur]\ar[rr]\ar[dd] & & \mathcal{T}_1' \ar[ur]\ar[dd] \\ & \mathcal{T}_2 \ar '[r][rr] & & \mathcal X \\ \mathcal{T}_2' \ar[rr]\ar[ur] & & \mathcal{U} \ar[ur] }$

In this diagram the bottom square, the right square, the back square, and the front square are $2$-fibre products. A formal argument then shows that $\mathcal{T}_1' \times _\mathcal {U} \mathcal{T}_2' \to \mathcal{T}_1 \times _{\mathcal X} \mathcal{T}_2$ is the “base change” of $\mathcal{U} \to \mathcal{X}$, more precisely the diagram

$\xymatrix{ \mathcal{T}_1' \times _\mathcal {U} \mathcal{T}_2' \ar[d] \ar[r] & \mathcal{U} \ar[d] \\ \mathcal{T}_1 \times _{\mathcal X} \mathcal{T}_2 \ar[r] & \mathcal{X} }$

is a $2$-fibre square. Hence $T'_1 \times _ U T'_2 \to F$ is representable by algebraic spaces, smooth, and surjective, see Lemmas 93.9.6, 93.9.7, 93.10.4, and 93.10.6. Therefore $F$ is an algebraic space by Bootstrap, Theorem 79.10.1 and we win. $\square$

An application of Lemma 93.15.3 is that something which is an algebraic space over an algebraic stack is an algebraic stack. This is the analogue of Bootstrap, Lemma 79.3.6. Actually, it suffices to assume the morphism $\mathcal{X} \to \mathcal{Y}$ is “algebraic”, as we will see in Criteria for Representability, Lemma 96.8.2.

Lemma 93.15.4. Let $S$ be a scheme contained in $\mathit{Sch}_{fppf}$. Let $\mathcal{X} \to \mathcal{Y}$ be a morphism of stacks in groupoids over $(\mathit{Sch}/S)_{fppf}$. Assume that

1. $\mathcal{X} \to \mathcal{Y}$ is representable by algebraic spaces, and

2. $\mathcal{Y}$ is an algebraic stack over $S$.

Then $\mathcal{X}$ is an algebraic stack over $S$.

Proof. Let $\mathcal{V} \to \mathcal{Y}$ be a surjective smooth $1$-morphism from a representable stack in groupoids to $\mathcal{Y}$. This exists by Definition 93.12.1. Then the $2$-fibre product $\mathcal{U} = \mathcal{V} \times _{\mathcal Y} \mathcal X$ is representable by an algebraic space by Lemma 93.9.8. The $1$-morphism $\mathcal{U} \to \mathcal X$ is representable by algebraic spaces, smooth, and surjective, see Lemmas 93.9.7 and 93.10.6. By Lemma 93.15.3 we conclude that $\mathcal{X}$ is an algebraic stack. $\square$

Lemma 93.15.5. Let $S$ be a scheme contained in $\mathit{Sch}_{fppf}$. Let $j : \mathcal X \to \mathcal Y$ be a $1$-morphism of categories fibred in groupoids over $(\mathit{Sch}/S)_{fppf}$. Assume $j$ is representable by algebraic spaces. Then, if $\mathcal{Y}$ is a stack in groupoids (resp. an algebraic stack), so is $\mathcal{X}$.

Proof. The statement on algebraic stacks will follow from the statement on stacks in groupoids by Lemma 93.15.4. If $j$ is representable by algebraic spaces, then $j$ is faithful on fibre categories and for each $U$ and each $y \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{Y}_ U)$ the presheaf

$(h : V \to U) \longmapsto \{ (x, \phi ) \mid x \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{X}_ V), \phi : h^*y \to f(x)\} /\cong$

is an algebraic space over $U$. See Lemma 93.9.2. In particular this presheaf is a sheaf and the conclusion follows from Stacks, Lemma 8.6.11. $\square$

Comment #2121 by Docteur Cottard on

I believe that many results in this section (and others) are also true when "algebraic stack" is replaced by "Deligne-Mumford stack". I'm thinking, for instance, in Lemma 15.4. It would be nice to at least simply say it in the statements (the proof is usually the same).

Comment #2142 by on

First I want to point out that all of these results have meaning for Deligne-Mumford stacks as Deligne-Mumford stacks are just special cases of algebraic stacks. But I understand that you mean something else, and yes, many of these types of results have to be added. I think we will add them later and some of them are already present (implicitly perhaps). For example, the correct version of 93.15.4 for Deligne-Mumford stacks is that if $f : \mathcal{X} \to \mathcal{Y}$ is a DM morphism and $\mathcal{Y}$ is Deligne-Mumford, then $\mathcal{X}$ is Deligne-Mumford. Moreover a morphism which is representable by algebraic spaces is DM. All of this is in the chapter on morphisms of algebraic stacks: see Lemmas 100.4.3 and 100.4.11 and Theorem 100.21.6.

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).