Finite sample sets for elementary affine loci
DefinitionWeierstrassEllipticZeta_ElementaryLocusSamplesalgebraic-geometrynumber-theory
For an elementary affine locus with anchor r in covering coordinates (t,z,u), define a finite set of sample points for a polynomial of bidegree (m,n):
- Point: just r.
- Line of slope α: r+(j,0,αj), for integers 0≤j≤m+n.
- Whole fibre: r+(i,0,j), for integers 0≤i≤m and 0≤j≤n.
The integers are mapped into ℂ. The definition contains only these points. Its sample cardinalities, containment in the locus, and sufficiency for testing polynomial vanishing are proved in the companion theorem. It does not choose the complex anchor or slope, and does not bound the canonical chart cost.
Definition code
import Definitions.Def_WeierstrassEllipticZeta_ElementaryLoci
import Mathlib.Data.Finset.Prod
import Mathlib.Data.Finset.Range
noncomputable section
open scoped Classical
namespace WeierstrassEllipticZeta
/-- Integer parameter samples sufficient to test a bihomogeneous polynomial on an
elementary affine locus. The theorem establishing sufficiency is separate. -/
def elementaryLocusSamples (shape : ElementaryLocusShape) (r : Fin 3 → ℂ)
(m n : ℕ) : Finset (Fin 3 → ℂ) :=
match shape with
| .point => {r}
| .line α => (Finset.range (m + n + 1)).image
(fun j : ℕ => r + ![(j : ℂ), 0, α * (j : ℂ)])
| .fibre => ((Finset.range (m + 1)).product (Finset.range (n + 1))).image
(fun ij : ℕ × ℕ => r + ![(ij.1 : ℂ), 0, (ij.2 : ℂ)])
end WeierstrassEllipticZeta
Source
Derived finite interpolation criterion for the mission's elementary locus-vanishing condition. At a fixed elliptic coordinate, substituting univariate polynomials T,U gives degree at most m*deg(T)+n*deg(U). The line restriction has degree at most m+n and is tested at m+n+1 integer parameters. A whole fibre is tested on an (m+1) by (n+1) integer grid, using the separate degree bounds m and n. The proof uses Mathlib's polynomial root-count theorem, https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/Algebra/Polynomial/Roots.lean. This is supporting coordinate algebra for the locus-selection framework, not a claim that Philippon (1986), section 5, is fully formalized: https://www.numdam.org/item/10.24033/bsmf.2060.pdf. No platform theorem dependencies are used by the complete proof. The forward sketch and full converse preserve all witnesses, the local cost, and C. Selecting a successful complex anchor/slope and proving the global cost bound remain open; the finite sample count is not an integer-search bound.