Simultaneous rational specialization for open quasi-finite affine families
ProvedAlgebra.exists_open_simultaneous_rational_specializationLet be an algebraically closed field, let with finite, and let be a finite index set. For each , let be a finite-type -algebra, with the compatible -algebra structure. Assume that
is open and quasi-finite.
Fix and -algebra homomorphisms whose restrictions to are evaluation at . A polynomial in finitely many labelled elements of the , with coefficients in , can be evaluated at any tuple of such homomorphisms. Suppose
Then there is a Zariski-open neighborhood of the evaluation ideal in such that, whenever has , there are -algebra homomorphisms satisfying
Thus finitely many rational points can be specialized together while retaining any prescribed polynomial nonvanishing condition. This includes empty index sets and empty coefficient-variable sets; nilpotents in the are allowed.
Formalization Note. The variables of are the disjoint union of the underlying sets of the ; each polynomial still has finite support. Finite type is stated separately because Mathlib's Algebra.QuasiFinite records finite-dimensional residue-field fibers without including finite type. This is an auxiliary algebraic synthesis of universal openness and rational-point existence, not a verbatim numbered source theorem. Its proof remains Open.
Verified reduction (8 October 2026). The finite tensor-product algebra and the rational-point specialization are now proved. The proof constructs the universal evaluation, proves finite generation of the tensor product, and uses its Jacobson property to find maximal ideals avoiding the nonvanishing element. Algebraic closedness produces the required compatible rational homomorphisms. The sole Open input is openness of finite tensor products over a normal Noetherian domain. That child contains no rational-point or nonvanishing-polynomial conclusion. The original formal statement is unchanged.
import Mathlib set_option autoImplicit false open scoped BigOperators Topology
/-- Simultaneous rational specialization with one nonvanishing constraint. -/
theorem Algebra.exists_open_simultaneous_rational_specialization
(K : Type*) [Field K] [IsAlgClosed K]
(σ ι : Type*) [Finite σ] [Finite ι]
(R : ι → Type*) [∀ i, CommRing (R i)] [∀ i, Algebra K (R i)]
[∀ i, Algebra (MvPolynomial σ K) (R i)]
[∀ i, IsScalarTower K (MvPolynomial σ K) (R i)]
[∀ i, Algebra.FiniteType (MvPolynomial σ K) (R i)]
(hopen : ∀ i, IsOpenMap
(PrimeSpectrum.comap (algebraMap (MvPolynomial σ K) (R i))))
(hquasi : ∀ i, Algebra.QuasiFinite (MvPolynomial σ K) (R i))
(c₀ : σ → K) (e₀ : ∀ i, R i →ₐ[K] K)
(h₀ : ∀ i (Q : MvPolynomial σ K),
e₀ i (algebraMap (MvPolynomial σ K) (R i) Q) = MvPolynomial.eval c₀ Q)
(P : MvPolynomial (Σ i, R i) K)
(hP : MvPolynomial.eval (fun t => e₀ t.1 t.2) P ≠ 0) :
∃ U : Set (PrimeSpectrum (MvPolynomial σ K)), IsOpen U ∧
MvPolynomial.pointToPoint (k := K) c₀ ∈ U ∧
∀ c : σ → K, MvPolynomial.pointToPoint (k := K) c ∈ U →
∃ e : ∀ i, R i →ₐ[K] K,
(∀ i (Q : MvPolynomial σ K),
e i (algebraMap (MvPolynomial σ K) (R i) Q) = MvPolynomial.eval c Q) ∧
MvPolynomial.eval (fun t => e t.1 t.2) P ≠ 0 := by sorry