A uniformly bounded subsequence of damped Möbius mean squares
OpenNymanBeurling.damped_moebius_bounded_mean_squareDoes there exist one finite B for which arbitrarily small positive ε admit arbitrarily large N such that every finite weighted square sum of the residual R(N,ε,m) is at most B? Here R(N,ε,m)=1+∑_{1≤d≤N} μ(d)d^{−ε}(m mod d)/d, with the natural weight 1/[m(m+1)]. The same B must work for all the chosen exponents, denominator cutoffs, and coordinate cutoffs.
This is an open unconditional assertion. Only one finite upper bound is required; the individual mean squares may stay separated from zero. Since the damped residuals already converge on each fixed coordinate prefix in the iterated limit, this bound suffices for weak Hilbert-space convergence along a diagonal. Orthogonal separation then puts the constant sequence in the norm-closed span of the remainders. Bagchi’s conditional norm convergence under RH implies this boundedness.
import Mathlib open Filter Topology Finset
theorem NymanBeurling.damped_moebius_bounded_mean_square :
∃ B : ℝ, ∃ᶠ ε : ℝ in 𝓝[>] 0, ∃ᶠ N : ℕ in atTop, ∀ H : ℕ,
∑ n ∈ Finset.range H, (1 + ∑ d ∈ Finset.Ioc 0 N,
((ArithmeticFunction.moebius d : ℝ)/(d:ℝ)^ε)*(((n+1)%d:ℕ):ℝ)/(d:ℝ))^2 /
((n+1:ℝ)*(n+2:ℝ)) ≤ B := by sorry