Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedPipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

The stable locus has a geometric quotient

Statement

Assume the Axiom of Choice inherited from the named suppliers. Let G be a complex reductive affine algebraic group acting algebraically on an affine algebraic set X, with categorical quotient π:X→X/ ⁣/G (Finite generation of invariants and the affine categorical quotient) and stable locus Xs (Stable points of an affine action). Then π(Xs) is open in X/ ⁣/G, one has Xs=π−1(π(Xs)) (so that Xs is an open G-stable subset of X), and the restriction πs:Xs→π(Xs) is a geometric quotient (Categorical and geometric quotients of classical varieties). In particular the fibres of πs are exactly the G-orbits, each orbit in Xs is closed in X, and Oπ(Xs)=(π∗sOXs)G.

Facts & Assumptions

Given: AC; a complex reductive affine algebraic group G acting on an affine algebraic set X; the categorical quotient π:X→X/ ⁣/G; the stable locus Xs; and Y={y∈X:dim⁡Gy≥1}, the locus of positive-dimensional stabilizers.

[F1]

Stable points. A point x is stable exactly when its orbit Gx is closed in X and its stabilizer is finite, equivalently dim⁡Gx=0 (Stable points of an affine action).

[F2]

The affine quotient theorem. The morphism π is a surjective categorical quotient, every fibre of π contains exactly one closed G-orbit, and for closed G-stable Y,Y′⊆X one has π(Y∩Y′)=π(Y)∩π(Y′), so the image of a closed G-stable subset is closed in X/ ⁣/G (Finite generation of invariants and the affine categorical quotient, clauses (iv) and (v)).

[F3]

Semicontinuity. For every n the locus {x:dim⁡Gx≥n} is closed in X; in particular Y is closed, and it is G-stable because Ggx=gGxg−1 has the same dimension as Gx (Semicontinuity of stabilizer and orbit dimension).

[F4]

Separation by an invariant. If Z⊆X is closed and G-stable and x satisfies π(x)∉π(Z), there is f∈C[X]G with f(x)≠0 and f∣Z=0 (Invariants separate a stable point from a disjoint closed invariant subset).

[F5]

Orbit dimensions and closures. For every orbit one has dim⁡G=dim⁡Gx+dim⁡Gx; every irreducible component of the closure Gy‾ has dimension dim⁡Gy and the boundary consists of orbits of strictly smaller dimension; every orbit closure contains a closed orbit, and orbits of minimal dimension are closed (Orbit dimension and closed orbits for complex group actions).

[F6]

Geometric quotients. A G-invariant morphism is a geometric quotient when it is surjective with fibres exactly the orbits, when a subset of the target is open exactly when its preimage is open, and when on open subsets the pullback of regular functions is an isomorphism onto the invariant regular functions (Categorical and geometric quotients of classical varieties).

[F7]

Naturality of the Reynolds operator. For a morphism f:V→W of rational G-modules one has RW∘f=fG∘RV (Complete reducibility and the Reynolds operator for a complex reductive group, clause (ii)); applied to the localisation A→Ag of the coordinate ring this gives (Ag)G=(AG)g for g∈AG: an invariant fraction equals RA(a)/gm, and the localized inclusion (AG)g→Ag is injective: a numerator a∈AG killed by a power of g in A is killed by that same power in the subring AG. Here Ag is rational, since a/gm lies in the image of the finite-dimensional rational span of a divided by the invariant denominator.

[F8]

Principal-open functions. On an affine algebraic set T with coordinate ring B, OT(DT(f))=Bf (Regular functions on a principal open are the principal localization). Such opens form a basis, as a point outside a closed polynomial zero locus has a defining polynomial nonzero there.

Proof

technique · direct
1.1F3

The subset Y={y:dim⁡Gy≥1} is closed in X and G-stable by [F3].

2.1F2F5step 1.1

Let x∈Xs. Then π(x)∉π(Y): if π(y)=π(x) for some y∈Y, then the fibre F=π−1(π(x)) is closed, G-stable and contains both the closed orbit Gx and Gy; by the unique closed orbit property [F2] the orbit Gx is the unique closed orbit in F, so the closed orbit contained in Gy‾ by [F5] must be Gx, giving Gx⊆Gy‾. Since dim⁡Gy=dim⁡G−dim⁡Gy≤dim⁡G−1<dim⁡G=dim⁡Gx, the point x is not in Gy, so Gx lies in the boundary of Gy‾; but every orbit in that boundary has dimension strictly smaller than dim⁡Gy, by [F5], contradicting dim⁡Gx=dim⁡G>dim⁡Gy. Hence π(x)∉π(Y).

3.1F2F4F5step 2.1

By [F4] applied to the closed G-stable set Y and the point x∈Xs of step 2.1, there is f∈C[X]G with f(x)≠0 and f∣Y=0. Put Xf={f≠0}. Since f is invariant, Xf=π−1(D(f)), so it is open, saturated and G-stable. Every y∈Xf satisfies y∉Y, hence dim⁡Gy=0 and dim⁡Gy=dim⁡G. If Gy were not closed, its boundary would contain an orbit of dimension strictly smaller than dim⁡Gy by [F5]. But Gy‾⊆π−1(π(y))⊆Xf, since the quotient fibre is closed and contains Gy; every point of this closure lies outside Y, so each orbit in the boundary has dimension dim⁡G, a contradiction. Thus Xf⊆Xs. Applying the same construction to every stable point gives Xs=⋃fXf, where f ranges over invariants vanishing on Y.

4.1F2step 3.1

Consequences for openness and saturation: Xf=π−1(D(f)), so π(Xf)=D(f) by surjectivity of π, so it is open in X/ ⁣/G and π(Xs)=⋃fD(f) is open. If π(x′)=π(x) with x∈Xs, choose the invariant f constructed at x in step 3.1; then π(x)∈D(f), so x′∈π−1(D(f))=Xf⊆Xs. Thus Xs=π−1(π(Xs)) is saturated, and it is open and G-stable in X.

4.2F6F7F8step 3.1

Structure sheaf: let f∈C[X]G with Xf⊆Xs. By [F8] the invariant regular functions on π−1(D(f))=Xf form (Af)G with A=C[X], and the Reynolds localisation identity [F7] identifies this with (AG)f=Oπ(Xs)(D(f)), so the pullback along πs is an isomorphism onto the invariant functions on each principal piece; the identity is compatible with restriction and glues over the basis of the D(f). This is condition (iii) of [F6], and it also gives Oπ(Xs)=(π∗sOXs)G.

5.1F1F2F6step 4.1

The fibres of πs are exactly the orbits: if π(x′)=π(x) with x,x′∈Xs, then Gx and Gx′ are both closed orbits in the same fibre, so they coincide by uniqueness of the closed orbit in that fibre [F2]. Together with surjectivity of πs onto π(Xs) this gives condition (i) of the geometric quotient [F6].

6.1F2F6step 5.1

Quotient topology: let W⊆Xs be open. Then W is open in X because Xs is open, and G⋅W is open, G-stable and satisfies π−1(π(G⋅W))∩Xs=G⋅W∩Xs by step 5.1. The complement X∖G⋅W is closed and G-stable, so its image is closed in X/ ⁣/G by [F2]; intersecting with π(Xs) gives π(G⋅W)∩π(Xs)=π(Xs)∖π(X∖G⋅W), which is open in π(Xs). Since π(W)=π(G⋅W), condition (ii) of [F6] holds: a subset of π(Xs) is open exactly when its preimage in Xs is open.

7.1F1F2F6F7step 4.1step 5.1step 6.1step 4.2∎

Conclusion: by steps 4.1 and 5.1 the map πs:Xs→π(Xs) is surjective with fibres exactly the orbits, and by steps 6.1 and 4.2 it satisfies the quotient topology and invariant-function conditions, so it is a geometric quotient by [F6]; the orbits in Xs are closed in X by the definition of stability [F1], and the sheaf identity of step 4.2 completes the statement. All Axiom of Choice content is inherited from the quotient, semicontinuity, separation and orbit suppliers used above.

Remarks

  • This is Brion's proof of Proposition 1.26 (printed pp. 9-10): the stable locus is a union of saturated invariant principal opens obtained from separating functions, and on it the quotient fibres are exactly the orbits. The topological condition is checked after saturating the open set, so no claim is made that images of arbitrary invariant opens outside Xs are open.
  • The class of the principal bundle, that is the local triviality of πs as a Gm-bundle in the example, is not asserted by this theorem; it is verified directly in ex-gm-quotient-of-affine-plane.

Depends on

Used by

Dependency tree · two levels

64 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources