ProofComplete

Algebraic Spaces - Key Proof

We sketch the proof of Keel-Mori's theorem on the existence of coarse moduli spaces for Deligne-Mumford stacks, which is fundamental for relating stacks to algebraic spaces.

ProofKeel-Mori Theorem (Sketch)

Let X\mathcal{X} be a separated Deligne-Mumford stack of finite type over a Noetherian base SS with finite stabilizers. We construct a coarse moduli space XX which is a separated algebraic space.

Step 1: Construction of the Coarse Space

Define XX as a functor X:SchSopSetX: \textbf{Sch}_S^{\text{op}} \to \textbf{Set} by: X(T)={morphisms f:TX with trivial stabilizers}/isomorphismX(T) = \{\text{morphisms } f: T \to \mathcal{X} \text{ with trivial stabilizers}\}/\text{isomorphism}

We must show that XX is representable by an algebraic space. The key is to prove:

  • XX is a sheaf in the étale topology
  • The diagonal XX×XX \to X \times X is representable
  • There exists a scheme with a surjective étale morphism to XX

Step 2: Sheaf Condition

Let {TiT}\{T_i \to T\} be an étale cover. If fi:TiXf_i: T_i \to \mathcal{X} are morphisms with trivial stabilizers that agree on overlaps Ti×TTjT_i \times_T T_j, we need to show they glue to f:TXf: T \to \mathcal{X}.

Since X\mathcal{X} is a stack, the fif_i glue to a morphism f:TXf: T \to \mathcal{X}. We must verify that ff has trivial stabilizers. This follows from the étale local nature of stabilizers: if stabilizers are trivial after pulling back to an étale cover, they are trivial on the base.

Step 3: Representability of the Diagonal

Given two morphisms f1,f2:TXf_1, f_2: T \rightrightarrows X (with trivial stabilizers), we must show the equalizer is representable by a scheme. This equalizer consists of points where f1f_1 and f2f_2 agree as morphisms to X\mathcal{X}.

Since X\mathcal{X} has representable diagonal, the equalizer of f1,f2f_1, f_2 viewed as morphisms to X\mathcal{X} is a closed subscheme of TT. The triviality of stabilizers ensures this equalizer is the same whether we work in X\mathcal{X} or its coarse space.

Step 4: Étale Cover by a Scheme

Choose an étale atlas UXU \to \mathcal{X} where UU is a scheme. We must show that UXU \to X is étale and surjective.

Surjectivity: For any field kk and xX(k)x \in X(k), the point xx corresponds to a morphism Spec(k)X\text{Spec}(k) \to \mathcal{X} with trivial stabilizer. Since UXU \to \mathcal{X} is surjective, this lifts to Spec(k)U\text{Spec}(k') \to U for some field extension k/kk'/k. Thus UXU \to X is surjective on geometric points.

Étaleness: To show UXU \to X is étale, we must verify it is locally of finite presentation, flat, and unramified. These properties follow from the corresponding properties of UXU \to \mathcal{X} and the fact that XX\mathcal{X} \to X is universal among morphisms to algebraic spaces.

Step 5: Universal Property

Let ϕ:XY\phi: \mathcal{X} \to Y be a morphism to an algebraic space YY. We must show ϕ\phi factors uniquely through XX.

For any scheme TT and morphism f:TXf: T \to \mathcal{X} with trivial stabilizer, compose to get TYT \to Y. This is compatible with the equivalence relation defining XX, hence descends to a morphism XYX \to Y. Uniqueness follows from the density of points with trivial stabilizers.

Step 6: Properness

The morphism π:XX\pi: \mathcal{X} \to X is proper. To see this, use the valuative criterion: given a commutative diagram with Spec(K)\text{Spec}(K) and Spec(R)\text{Spec}(R) (a DVR), a morphism Spec(R)X\text{Spec}(R) \to X and Spec(K)X\text{Spec}(K) \to \mathcal{X} lifting the generic point, we must find a unique lift Spec(R)X\text{Spec}(R) \to \mathcal{X}.

Since X\mathcal{X} is separated, such lifts form a closed subspace of the Hom-space, which is proper. The existence and uniqueness follow from the properness of X\mathcal{X} over SS and the representability of the diagonal.

Step 7: Pushforward of Structure Sheaf

Finally, πOX=OX\pi_*\mathcal{O}_\mathcal{X} = \mathcal{O}_X follows from the fact that sections of OX\mathcal{O}_\mathcal{X} that are invariant under all automorphisms descend to XX. Since stabilizers are finite and we work over a base where finite group averaging is possible, invariant functions are precisely those that descend.

This completes the sketch of the Keel-Mori theorem.

This proof demonstrates how stack-theoretic properties translate into properties of algebraic spaces, bridging the gap between the stacky world and classical geometry.