Finite coefficient rank criterion for polynomial interpolation
ProvedTranscendenceTheory.bounded_polynomial_family_rank_obstructionelliptic-functionslinear-algebrapolynomials
Let be a field and let be finite index sets. Fix and polynomials , for , , and , all of degree at most (the zero polynomial is allowed).
Form matrices over , with rows indexed by , by
There is a right-hand side that cannot be represented by a constant linear combination of the columns of if and only if adjoining all right-hand-side columns increases rank:
The coefficient cutoff is exactly coefficients per polynomial equation. No nonemptiness assumptions, nonzero polynomial assumptions, or independence hypotheses on the columns are required. This criterion replaces a family of polynomial interpolation failures by a single finite matrix rank comparison.
Preamble
import Mathlib.LinearAlgebra.Matrix.Rank import Mathlib.Data.Matrix.ColumnRowPartitioned import Mathlib.LinearAlgebra.FiniteDimensional.Lemmas import Mathlib.LinearAlgebra.Finsupp.LinearCombination import Mathlib.Algebra.Polynomial.BigOperators import Mathlib.Algebra.Polynomial.Degree.Operations open scoped BigOperators
Formal statement
theorem TranscendenceTheory.bounded_polynomial_family_rank_obstruction
(K α ι κ : Type*) [Field K] [Fintype α] [Fintype ι] [Fintype κ]
(d : ℕ) (A : α → ι → Polynomial K) (B : α → κ → Polynomial K)
(hA : ∀ a i, (A a i).natDegree ≤ d)
(hB : ∀ a k, (B a k).natDegree ≤ d) :
(∃ k, ¬ ∃ u : ι → K, ∀ a, ∑ i, Polynomial.C (u i) * A a i = B a k) ↔
let Ac : Matrix (α × Fin (d + 1)) ι K := fun e i => (A e.1 i).coeff e.2.val
let Bc : Matrix (α × Fin (d + 1)) κ K := fun e k => (B e.1 k).coeff e.2.val
Ac.rank < (Matrix.fromCols Ac Bc).rank := by sorry
Source
Derived finite coefficient rank criterion for the interpolation frontier https://prove2.me/theorems/4af1c4a0-3334-49b6-b559-e122c13d1774. Primary sources: Mathlib revision 0df444a360eaa60ab8c11dca51a86af692955474, LinearAlgebra/Matrix/Rank.lean, Matrix.rank_eq_finrank_span_cols, line 398: https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/LinearAlgebra/Matrix/Rank.lean#L398; LinearAlgebra/FiniteDimensional/Lemmas.lean, Submodule.finrank_lt_finrank_of_lt, line 235: https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/LinearAlgebra/FiniteDimensional/Lemmas.lean#L235; Algebra/Polynomial/Degree/Operations.lean, Polynomial.coeff_eq_zero_of_natDegree_lt, line 85: https://github.com/leanprover-community/mathlib4/blob/0df444a360eaa60ab8c11dca51a86af692955474/Mathlib/Algebra/Polynomial/Degree/Operations.lean#L85. The complete theorem is derived here from finite-dimensional column spans and coefficient vanishing above degree; it is not a verbatim theorem from the mission paper. Mission context: Senthil Kumar K, Algebraic independence of values of Weierstrass elliptic and zeta functions (2026), Appendix A.2, Theorem A.3 and Proposition A.1, https://doi.org/10.1017/S001309152610145X. The connecting reduction preserves all hypotheses, witnesses, root conditions and numerical bounds, while bounding the exact branch coefficient indices by 2N|Z| and adjoining every right-hand-side column simultaneously. Both directions are checked in Lean. The uniform geometric estimate remains open.