An exponential polynomial is controlled by its first derivatives at the origin
ProvedFourExp.expPoly_value_le_derivscomplex-analysisexponential-polynomialstranscendence
Values of an exponential polynomial are bounded by its low-order derivatives at .
Let be distinct complex numbers with for some . For each let be a polynomial that is zero or has degree less than , and put
If for every , then for every with
Proof idea (van der Poorten's method, as in Waldschmidt 1971).
- The numbers , , determine the coefficients of . The system is a confluent Vandermonde determinant , so Cramer's rule gives .
- The are the coefficients of the Hermite interpolation polynomial of at the nodes , each with multiplicity .
- Newton's form with divided differences, written as contour integrals over , gives .
Why polynomial coefficients. Rescaling keeps a function in this form (compose each with an affine map), and FourExp.expPoly_zero_count_scaled needs exactly that.
What it is for. It is the second estimate behind FourExp.expPoly_zero_count_scaled.
Preamble
import Mathlib open Finset
Formal statement
namespace FourExp
theorem expPoly_value_le_derivs
{l : ℕ} (q : Fin l → ℕ) (w : Fin l → ℂ) (hw : Function.Injective w)
(P : Fin l → Polynomial ℂ) (hP : ∀ j, P j = 0 ∨ (P j).natDegree < q j)
(W R D : ℝ) (hW0 : 0 ≤ W) (hW : ∀ j, ‖w j‖ ≤ W) (hR : 0 ≤ R)
(hD : ∀ s : ℕ, s < ∑ j, q j →
‖iteratedDeriv s (fun z : ℂ => ∑ j, (P j).eval z * Complex.exp (w j * z)) 0‖ ≤ D)
(u : ℂ) (hu : ‖u‖ ≤ R) :
‖∑ j, (P j).eval u * Complex.exp (w j * u)‖
≤ ((∑ j, q j : ℕ) : ℝ) * (W + 1) ^ ((∑ j, q j) + 1) * Real.exp (R * (W + 1)) * D := by
sorry
end FourExpSource
M. Waldschmidt, Indépendance algébrique des valeurs de la fonction exponentielle, Bull. Soc. Math. France 99 (1971), 285–304, §4, (4.5)–(4.13).