Uniform weighted approximation by finite damped Mobius remainder sums
OpenNymanBeurling.damped_moebius_mean_square_approximationWrite and let be the Möbius function. For every , the proposed approximation assertion asks for and such that, simultaneously for every integer cutoff ,
The coefficients of the approximant to the constant sequence are therefore . The choice of and precedes the universal quantifier over ; a bound for just one cutoff would not suffice. The terms are nonnegative, so the condition controls the full weighted mean-square error.
This is an open unconditional analytic assertion. Bagchi's construction proves the relevant approximation under the Riemann hypothesis; RH is not an assumption in this formal target, and the paper is not being cited as an unconditional proof. Together with an elementary tail-budget choice and rational rounding, this assertion implies arbitrarily small finite rational remainder certificates. It replaces an unrestricted coefficient search by a family with the two parameters and , while retaining the central unresolved analytic difficulty.
import Mathlib set_option autoImplicit false open scoped BigOperators
theorem NymanBeurling.damped_moebius_mean_square_approximation
(δ : ℝ) (hδ : 0 < δ) :
∃ (N : ℕ) (ε : ℝ), 0 < ε ∧ ε < 1/2 ∧ ∀ H : ℕ,
(∑ n ∈ Finset.range H,
(1 + ∑ k : Fin N,
((ArithmeticFunction.moebius (k.val+2) : ℝ) / (k.val+2 : ℝ)^ε) *
(((n+1) % (k.val+2) : ℕ) : ℝ) / (k.val+2 : ℝ))^2 /
((n+1 : ℝ)*(n+2 : ℝ))) < δ := by sorry