Exactness of sufficiently long rational congruences
ProvedTranscendenceTheory.bounded_rational_congruence_rigiditycommutative-algebraelliptic-functionspolynomials
Let R be any commutative ring. Let Q,A∈R[X], let α,λ∈R, and let d,s be nonnegative integers such that
Then
Both statements are in R[X]. Thus a sufficiently long congruence between these bounded-degree expressions is already an exact polynomial identity.
In the strict bound on A, the zero polynomial has degree −∞, so A=0 is permitted even when d=0. Lean expresses the bound on Q using natural degree, which assigns zero to the zero polynomial and is equivalent to the stated upper bound because d≥0. There is no field, integral-domain, inverse or nonvanishing assumption.
Preamble
import Mathlib.Algebra.Polynomial.Div open Polynomial
Formal statement
theorem TranscendenceTheory.bounded_rational_congruence_rigidity
(R : Type*) [CommRing R] (Q A : Polynomial R) (α z : R) (d s : ℕ)
(hQ : Q.natDegree ≤ d) (hA : A.degree < (d : WithBot ℕ)) (hds : d ≤ s) :
X ^ (s + 1) ∣ (1 - C z * X) * A - C α * Q ↔
(1 - C z * X) * A = C α * Q := by sorry
Source
Derived interpolation-rigidity step for https://prove2.me/theorems/c334622d-2651-4fed-ae97-42d3ef348d2a. Mission context: Senthil Kumar K (2026), Algebraic independence of values of Weierstrass elliptic and zeta functions, Appendix A.2, Theorem A.3 and Proposition A.1; https://doi.org/10.1017/S001309152610145X. This is an algebraic tool for the interpolation frontier; the geometric zero estimate remains open. Primary Lean sources: Mathlib Algebra/Polynomial/Div.lean (X_pow_dvd_iff), Degree/Defs.lean and Degree/Operations.lean (degree inequalities and coefficient vanishing), and Eval/Degree.lean (degree bounds under coefficient maps), revision 0df444a360eaa60ab8c11dca51a86af692955474. See https://leanprover-community.github.io/mathlib4_docs/Mathlib/Algebra/Polynomial/Degree/Operations.html and https://leanprover-community.github.io/mathlib4_docs/Mathlib/Algebra/Polynomial/Div.html. The polynomial (1-z*X)*A-alpha*Q has degree at most d when deg(A)<d and deg(Q)<=d. Divisibility by X^(s+1) for s>=d therefore makes it zero. The mission connection proves the degree bound for weighted numerators after scalar extension, and uses an exact case split at d=2*N*card(Z), preserving all witnesses and bounds.