Loading [MathJax]/jax/element/mml/optable/Latin1Supplement.js

The Stacks project

81.3 Colimits of algebraic spaces

We briefly discuss colimits of algebraic spaces. Let S be a scheme. Let \mathcal{I} \to (\mathit{Sch}/S)_{fppf}, i \mapsto X_ i be a diagram (see Categories, Section 4.14). For each i we may consider the small étale site X_{i, {\acute{e}tale}} whose objects are schemes étale over X_ i, see Properties of Spaces, Section 66.18. For each morphism i \to j of \mathcal{I} we have the morphism X_ i \to X_ j and hence a pullback functor X_{j, {\acute{e}tale}} \to X_{i, {\acute{e}tale}}. Hence we obtain a pseudo functor from \mathcal{I}^{opp} into the 2-category of categories. Denote

\mathop{\mathrm{lim}}\nolimits _ i X_{i, {\acute{e}tale}}

the 2-limit (see insert future reference here). What does this mean concretely? An object of this limit is a system of étale morphisms U_ i \to X_ i over \mathcal{I} such that for each i \to j in \mathcal{I} the diagram

\xymatrix{ U_ i \ar[r] \ar[d] & U_ j \ar[d] \\ X_ i \ar[r] & X_ j }

is cartesian. Morphisms between objects are defined in the obvious manner. Suppose that f_ i : X_ i \to T is a family of morphisms such that for each i \to j the composition X_ i \to X_ j \to T is equal to f_ i. Then we get a functor T_{\acute{e}tale}\to \mathop{\mathrm{lim}}\nolimits X_{i, {\acute{e}tale}}. With this notation in hand we can formulate our lemma.

Lemma 81.3.1. Let S be a scheme. Let \mathcal{I} \to (\mathit{Sch}/S)_{fppf}, i \mapsto X_ i be a diagram of schemes over S as above. Assume that

  1. X = \mathop{\mathrm{colim}}\nolimits X_ i exists in the category of schemes,

  2. \coprod X_ i \to X is surjective,

  3. if U \to X is étale and U_ i = X_ i \times _ X U, then U = \mathop{\mathrm{colim}}\nolimits U_ i in the category of schemes, and

  4. every object (U_ i \to X_ i) of \mathop{\mathrm{lim}}\nolimits X_{i, {\acute{e}tale}} with U_ i \to X_ i separated is in the essential image of the functor X_{\acute{e}tale}\to \mathop{\mathrm{lim}}\nolimits X_{i, {\acute{e}tale}}.

Then X = \mathop{\mathrm{colim}}\nolimits X_ i in the category of algebraic spaces over S also.

Proof. Let Z be an algebraic space over S. Suppose that f_ i : X_ i \to Z is a family of morphisms such that for each i \to j the composition X_ i \to X_ j \to Z is equal to f_ i. We have to construct a morphism of algebraic spaces f : X \to Z such that we can recover f_ i as the composition X_ i \to X \to Z. Let W \to Z be a surjective étale morphism of a scheme to Z. We may assume that W is a disjoint union of affines and in particular we may assume that W \to Z is separated. For each i set U_ i = W \times _{Z, f_ i} X_ i and denote h_ i : U_ i \to W the projection. Then U_ i \to X_ i forms an object of \mathop{\mathrm{lim}}\nolimits X_{i, {\acute{e}tale}} with U_ i \to X_ i separated. By assumption (4) we can find an étale morphism U \to X and (functorial) isomorphisms U_ i = X_ i \times _ X U. By assumption (3) there exists a morphism h : U \to W such that the compositions U_ i \to U \to W are h_ i. Let g : U \to Z be the composition of h with the map W \to Z. To finish the proof we have to show that g : U \to Z descends to a morphism X \to Z. To do this, consider the morphism (h, h) : U \times _ X U \to W \times _ S W. Composing with U_ i \times _{X_ i} U_ i \to U \times _ X U we obtain (h_ i, h_ i) which factors through W \times _ Z W. Since U \times _ X U is the colimit of the schemes U_ i \times _{X_ i} U_ i by (3) we see that (h, h) factors through W \times _ Z W. Hence the two compositions U \times _ X U \to U \to W \to Z are equal. Because each U_ i \to X_ i is surjective and assumption (2) we see that U \to X is surjective. As Z is a sheaf for the étale topology, we conclude that g : U \to Z descends to f : X \to Z as desired. \square

We can check that a cocone is a colimit (fpqc) locally on the cocone.

Lemma 81.3.2. Let S be a scheme. Let B be an algebraic space over S. Let \mathcal{I} \to (\mathit{Sch}/S)_{fppf}, i \mapsto X_ i be a diagram of algebraic spaces over B. Let (X, X_ i \to X) be a cocone for the diagram in the category of algebraic spaces over B (Categories, Remark 4.14.5). If there exists a fpqc covering \{ U_ a \to X\} _{a \in A} such that

  1. for all a \in A we have U_ a = \mathop{\mathrm{colim}}\nolimits X_ i \times _ X U_ a in the category of algebraic spaces over B, and

  2. for all a, b \in A we have U_ a \times _ X U_ b = \mathop{\mathrm{colim}}\nolimits X_ i \times _ X U_ a \times _ X U_ b in the category of algebraic spaces over B,

then X = \mathop{\mathrm{colim}}\nolimits X_ i in the category of algebraic spaces over B.

Proof. Namely, for an algebraic space Y over B a morphism X \to Y over B is the same thing as a collection of morphism U_ a \to Y which agree on the overlaps U_ a \times _ X U_ b for all a, b \in A, see Descent on Spaces, Lemma 74.7.2. \square

We are going to find a common partial generalization of Lemmas 81.3.1 and 81.3.2 which can in particular be used to reduce a colimit construction to a subcategory of the category of all algebraic spaces.

Let S be a scheme and let B be an algebraic space over S. Let \mathcal{I} be an index category and let i \mapsto X_ i be a diagram in the category of algebraic spaces over B, see Categories, Section 4.14. For each i we may consider the small étale site X_{i, spaces, {\acute{e}tale}} whose objects are algebraic spaces étale over X_ i, see Properties of Spaces, Section 66.18. For each morphism i \to j of \mathcal{I} we have the morphism X_ i \to X_ j and hence a pullback functor X_{j, spaces, {\acute{e}tale}} \to X_{i, spaces, {\acute{e}tale}}. Hence we obtain a pseudo functor from \mathcal{I}^{opp} into the 2-category of categories. Denote

\mathop{\mathrm{lim}}\nolimits _ i X_{i, spaces, {\acute{e}tale}}

the 2-limit (see insert future reference here). What does this mean concretely? An object of this limit is a diagram i \mapsto (U_ i \to X_ i) in the category of arrows of algebraic spaces over B such that for each i \to j in \mathcal{I} the diagram

\xymatrix{ U_ i \ar[r] \ar[d] & U_ j \ar[d] \\ X_ i \ar[r] & X_ j }

is cartesian. Morphisms between objects are defined in the obvious manner. Suppose that f_ i : X_ i \to Z is a family of morphisms of algebraic spaces over B such that for each i \to j the composition X_ i \to X_ j \to Z is equal to f_ i. Then we get a functor Z_{spaces, {\acute{e}tale}} \to \mathop{\mathrm{lim}}\nolimits X_{i, spaces, {\acute{e}tale}}. With this notation in hand we can formulate our next lemma.

Lemma 81.3.3. Let S be a scheme. Let B be an algebraic space over S. Let \mathcal{I} \to (\mathit{Sch}/S)_{fppf}, i \mapsto X_ i be a diagram of algebraic spaces over B. Let (X, X_ i \to X) be a cocone for the diagram in the category of algebraic spaces over B (Categories, Remark 4.14.5). Assume that

  1. the base change functor X_{spaces, étale} \to \mathop{\mathrm{lim}}\nolimits X_{i, spaces, {\acute{e}tale}}, sending U to U_ i = X_ i \times _ X U is an equivalence,

  2. given

    1. B' affine and étale over B,

    2. Z an affine scheme over B',

    3. U \to X \times _ B B' an étale morphism of algebraic spaces with U affine,

    4. f_ i : U_ i \to Z a cocone over B' of the diagram i \mapsto U_ i = U \times _ X X_ i,

    there exists a unique morphism f : U \to Z over B' such that f_ i equals the composition U_ i \to U \to Z.

Then X = \mathop{\mathrm{colim}}\nolimits X_ i in the category of all algebraic spaces over B.

Proof. In this paragraph we reduce to the case where B is an affine scheme. Let B' \to B be an étale morphism of algebraic spaces. Observe that conditions (1) and (2) are preserved if we replace B, X_ i, X by B', X_ i \times _ B B', X \times _ B B'. Let \{ B_ a \to B\} _{a \in A} be an étale covering with B_ a affine, see Properties of Spaces, Lemma 66.6.1. For a \in A denote X_ a, X_{a, i} the base changes of X and the diagram to B_ a. For a, b \in A denote X_{a, b} and X_{a, b, i} the base changes of X and the diagram to B_ a \times _ B B_ b. By Lemma 81.3.2 it suffices to prove that X_ a = \mathop{\mathrm{colim}}\nolimits X_{a, i} and X_{a, b} = \mathop{\mathrm{colim}}\nolimits X_{a, b, i}. This reduces us to the case where B = B_ a (an affine scheme) or B = B_ a \times _ B B_ b (a separated scheme). Repeating the argument once more, we conclude that we may assume B is an affine scheme (this uses that the intersection of affine opens in a separated scheme is affine).

Assume B is an affine scheme. Let Z be an algebraic space over B. We have to show

\mathop{\mathrm{Mor}}\nolimits _ B(X, Z) \longrightarrow \mathop{\mathrm{lim}}\nolimits \mathop{\mathrm{Mor}}\nolimits _ B(X_ i, Z)

is a bijection.

Proof of injectivity. Let f, g : X \to Z be morphisms such that the compositions f_ i, g_ i : X_ i \to Z are the same for all i. Choose an affine scheme Z' and an étale morphism Z' \to Z. By Properties of Spaces, Lemma 66.6.1 we know we can cover Z by such affines. Set U = X \times _{f, Z} Z' and U' = X \times _{g, Z} Z' and denote p : U \to X and p' : U' \to X the projections. Since f_ i = g_ i for all i, we see that

U_ i = X_ i \times _{f_ i, Z} Z' = X_ i \times _{g_ i, Z} Z' = U'_ i

compatible with transition morphisms. By (1) there is a unique isomorphism \epsilon : U \to U' as algebraic spaces over X, i.e., with p = p' \circ \epsilon which is compatible with the displayed identifications. Choose an étale covering \{ h_ a : U_ a \to U\} with U_ a affine. By (2) we see that f \circ p \circ h_ a = g \circ p' \circ \epsilon \circ h_ a = g \circ p \circ h_ a. Since \{ h_ a : U_ a \to U\} is an étale covering we conclude f \circ p = g \circ p. Since the collection of morphisms p : U \to X we obtain in this manner is an étale covering, we conclude that f = g.

Proof of surjectivity. Let f_ i : X_ i \to Z be an element of the right hand side of the displayed arrow in the first paragraph of the proof. It suffices to find an étale covering \{ U_ c \to X\} _{c \in C} such that the families f_{c, i} \in \mathop{\mathrm{lim}}\nolimits _ i \mathop{\mathrm{Mor}}\nolimits _ B(X_ i \times _ X U_ c, Z) come from morphisms f_ c : U_ c \to Z. Namely, by the uniqueness proved above the morphisms f_ c will agree on U_ c \times _ X U_ b and hence will descend to give the desired morphism f : X \to Z. To find our covering, we first choose an étale covering \{ g_ a : Z_ a \to Z\} _{a \in A} where each Z_ a is affine. Then we let U_{a, i} = X_ i \times _{f_ i, Z} Z_ a. By (1) we find U_{a, i} = X_ i \times _ X U_ a for some algebraic spaces U_ a étale over X. Then we choose étale coverings \{ U_{a, b} \to U_ a\} _{b \in B_ a} with U_{a, b} affine and we consider the morphisms

U_{a, b, i} = X_ i \times _ X U_{a, b} \to X_ i \times _ X U_ a = X_ i \times _{f_ i, Z} Z_ a \to Z_ a

By (2) we obtain morphisms f_{a, b} : U_{a, b} \to Z_ a compatible with these morphisms. Setting C = \coprod _{a \in A} B_ a and for c \in C corresponding to b \in B_ a setting U_ c = U_{a, b} and f_ c = g_ a \circ f_{a, b} : U_ c \to Z we conclude. \square

Here is an application of these ideas to reduce the general case to the case of separated algebraic spaces.

Lemma 81.3.4. Let S be a scheme. Let B be an algebraic space over S. Let \mathcal{I} \to (\mathit{Sch}/S)_{fppf}, i \mapsto X_ i be a diagram of algebraic spaces over B. Assume that

  1. each X_ i is separated over B,

  2. X = \mathop{\mathrm{colim}}\nolimits X_ i exists in the category of algebraic spaces separated over B,

  3. \coprod X_ i \to X is surjective,

  4. if U \to X is an étale separated morphism of algebraic spaces and U_ i = X_ i \times _ X U, then U = \mathop{\mathrm{colim}}\nolimits U_ i in the category of algebraic spaces separated over B, and

  5. every object (U_ i \to X_ i) of \mathop{\mathrm{lim}}\nolimits X_{i, spaces, {\acute{e}tale}} with U_ i \to X_ i separated is of the form U_ i = X_ i \times _ X U for some étale separated morphism of algebraic spaces U \to X.

Then X = \mathop{\mathrm{colim}}\nolimits X_ i in the category of all algebraic spaces over B.

Proof. We encourage the reader to look instead at Lemma 81.3.3 and its proof.

Let Z be an algebraic space over B. Suppose that f_ i : X_ i \to Z is a family of morphisms such that for each i \to j the composition X_ i \to X_ j \to Z is equal to f_ i. We have to construct a morphism of algebraic spaces f : X \to Z over B such that we can recover f_ i as the composition X_ i \to X \to Z. Let W \to Z be a surjective étale morphism of a scheme to Z. We may assume that W is a disjoint union of affines and in particular we may assume that W \to Z is separated and that W is separated over B. For each i set U_ i = W \times _{Z, f_ i} X_ i and denote h_ i : U_ i \to W the projection. Then U_ i \to X_ i forms an object of \mathop{\mathrm{lim}}\nolimits X_{i, spaces, {\acute{e}tale}} with U_ i \to X_ i separated. By assumption (5) we can find a separated étale morphism U \to X of algebraic spaces and (functorial) isomorphisms U_ i = X_ i \times _ X U. By assumption (4) there exists a morphism h : U \to W over B such that the compositions U_ i \to U \to W are h_ i. Let g : U \to Z be the composition of h with the map W \to Z. To finish the proof we have to show that g : U \to Z descends to a morphism X \to Z. To do this, consider the morphism (h, h) : U \times _ X U \to W \times _ S W. Composing with U_ i \times _{X_ i} U_ i \to U \times _ X U we obtain (h_ i, h_ i) which factors through W \times _ Z W. Since U \times _ X U is the colimit of the algebraic spaces U_ i \times _{X_ i} U_ i in the category of algebraic spaces separated over B by (4) we see that (h, h) factors through W \times _ Z W. Hence the two compositions U \times _ X U \to U \to W \to Z are equal. Because each U_ i \to X_ i is surjective and assumption (2) we see that U \to X is surjective. As Z is a sheaf for the étale topology, we conclude that g : U \to Z descends to f : X \to Z as desired. \square


Comments (2)

Comment #6067 by Dario Weißmann on

there are two instances introducing the notation of the 2-limit of the etale sites in this section, one in the beginning and one directly after 0GFQ.

Comment #6196 by on

Yes that is right, but they are different sites. So I think this is OK.


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.