Étale Equivalence Relations
The theory of étale equivalence relations provides the foundational machinery for constructing algebraic spaces. Every algebraic space arises as the quotient of a scheme by an étale equivalence relation, and conversely, every such quotient is an algebraic space. This perspective makes algebraic spaces concrete: they are "schemes modulo étale gluings."
1. Equivalence Relations in Categories
Let be a category with fiber products. An equivalence relation on an object is a monomorphism such that for every object , the image of is an equivalence relation on the set .
Equivalently, (via the two projections and ) must satisfy:
- Reflexivity: There exists a section of both and (the diagonal factors through ).
- Symmetry: There exists an involution swapping and .
- Transitivity: The map given by factors through .
2. Étale Equivalence Relations
An equivalence relation on a scheme (in the category of schemes over ) is an étale equivalence relation if both structure morphisms are étale.
The étaleness condition is crucial: it ensures that the quotient sheaf (in the étale topology) is an algebraic space. Weaker conditions (e.g., flat or smooth) lead to algebraic stacks rather than algebraic spaces.
An equivalence relation is:
- smooth if and are smooth,
- flat if and are flat and locally of finite presentation.
Smooth equivalence relations give rise to algebraic stacks (Artin stacks), while étale equivalence relations specifically give algebraic spaces.
3. Quotient Sheaves
Given an étale equivalence relation , we form the quotient sheaf in the étale topology:
for each scheme , where the coequalizer is taken in the category of sets, and we sheafify with respect to the étale topology.
Let be an étale equivalence relation on a scheme over . Then the quotient sheaf is an algebraic space over , and the natural map is an étale surjection serving as an atlas.
We verify the three conditions for an algebraic space:
Sheaf condition: By construction, is a sheaf for the étale topology.
Representability of the diagonal: For any scheme and morphisms , we need to be a scheme. After étale base change to , the morphisms lift locally. The fiber product is a scheme since is a scheme and the map is a base change.
Existence of atlas: The morphism is étale and surjective by construction. Étaleness follows because for any , the base change is an étale -scheme (since is étale). Surjectivity follows from the fact that étale-locally, every section of lifts to .
4. The Converse: Every Algebraic Space is a Quotient
Every algebraic space over is isomorphic to for some étale equivalence relation on a scheme .
Let be an étale atlas (which exists by definition). Set . Since the diagonal of is representable, is a scheme. The two projections are étale because they are base changes of the étale morphism . The relation is an equivalence relation (reflexivity from the diagonal , symmetry from the swap map, transitivity from composition of correspondences). By construction, .
5. Examples
The trivial equivalence relation on is (the diagonal). The quotient is itself. This shows every scheme arises from an étale equivalence relation.
Let be a finite group acting freely on a scheme over a field . Define with and . Since is étale over (being a finite discrete group scheme in characteristic not dividing ), the projections are étale. Freeness ensures is a monomorphism. The quotient is an algebraic space.
Let and acting by (assuming and contains a primitive -th root of unity is not needed here; we only need in ). Set . The quotient is an algebraic space. When is invertible in , this is isomorphic to via the map .
Let be an étale surjection of schemes. Then is an étale equivalence relation, and . This provides a tautological presentation of any scheme as a quotient.
Let and the involution swapping the two factors. On the open set (complement of the diagonal), acts freely. The quotient is an algebraic space (and in this case a scheme, since is quasi-projective).
Take and let be the equivalence relation that identifies the two copies everywhere except at the origin. The maps are étale (they are open immersions on the gluing part). The quotient is the line with doubled origin, a non-separated algebraic space (which happens to be a scheme).
Let be a finite Galois extension with group . Let . Then defines an étale equivalence relation, and . This is Galois descent in its simplest form.
Let be schemes with an open subscheme and an étale isomorphism . Set and (with appropriate maps using ). The quotient is the algebraic space obtained by gluing and along . If is not an open immersion on both sides, the result may fail to be a scheme.
Let be an abelian variety over and a finite étale subgroup scheme. Then acts freely on by translation: . The quotient is an abelian variety (hence a scheme), and the map is an isogeny. This is a case where the algebraic space quotient turns out to be a scheme.
Let and (the symmetric group) acting by permuting factors. The quotient is the -th symmetric product (a scheme). The étale equivalence relation on the open locus of distinct points gives an étale presentation of the quotient.
Consider and the equivalence relation given by the free -action restricted to . The Zariski quotient does not exist as a scheme with good properties, but in the étale topology, is a well-defined algebraic space.
Let be a smooth projective curve of genus over , and a fixed-point-free involution (which exists when admits a degree-2 map to a curve unramified everywhere). Then with the identity and gives an étale equivalence relation, and is a curve of genus when is odd.
For any scheme , the diagonal is an étale equivalence relation (in fact a closed immersion if is separated). The quotient is itself. This trivial example illustrates that the notion of étale equivalence relation subsumes the notion of identity.
6. Properties of Étale Equivalence Relations
Every étale equivalence relation in the category of schemes is effective in the étale topology: the natural map is an isomorphism.
This is a key difference from the Zariski topology, where equivalence relations need not be effective. The étale topology has better descent properties, which is precisely why algebraic spaces are defined using it.
For smooth (but non-étale) equivalence relations, the quotient sheaf in the étale topology may not be an algebraic space but rather an algebraic stack. Effectivity still holds, but the quotient has a richer structure (with automorphisms).
7. Refinements and Morita Equivalence
Two étale equivalence relations and define the same algebraic space (up to isomorphism) if and only if they are Morita equivalent: there exists a scheme with étale surjections to both and that are compatible with the equivalence relations.
Two étale equivalence relations and are Morita equivalent if as sheaves on the big étale site.
This means an algebraic space is not "an" equivalence relation, but rather an equivalence class of equivalence relations. Different atlases give different presentations of the same algebraic space, much as different atlases present the same manifold.
8. Effectivity and Descent
An equivalence relation is effective (in a given topology) if the natural map is a coequalizer and .
Let be a flat, locally finitely presented equivalence relation on a scheme . Then is effective in the fppf topology: the quotient fppf sheaf satisfies .
For étale equivalence relations, effectivity already holds in the étale topology:
This means that once we form the quotient sheaf, we recover the original equivalence relation by forming the fiber product.
9. Constructing Étale Equivalence Relations in Practice
In practice, étale equivalence relations arise from:
- Group actions: A group scheme acting on gives (when the action is free and is étale).
- Étale covers with descent data: An étale cover gives .
- Gluing data: Schemes with étale isomorphisms satisfying the cocycle condition.
The last construction is the most general and shows that algebraic spaces are obtained by "étale gluing" of schemes.
References
- D. Knutson, Algebraic Spaces, LNM 203, Springer, 1971, Chapter II.
- M. Artin, "Algebraic Spaces," Yale Mathematical Monographs, 1971.
- The Stacks Project, Tag 025Y: Algebraic Spaces.