Documentation

TauCeti.NumberTheory.LSeries.WienerIkehara.BoundaryGrowth

A growth bound for nonnegative coefficients, and the summability it supplies #

The smoothed Wiener--Ikehara asymptotic TauCeti.LSeries.tendsto_tsum_term_mul_fourier_atTop still carries a summability hypothesis: the Fourier-weighted series βˆ‘ a n 𝓕 psi (log (n / x) / 2Ο€) / n has to converge on the boundary line Re s = 1, where the coefficients are no longer damped by n ^ (-(sigma - 1)). This file removes that hypothesis for nonnegative coefficients, which is the only case Wiener--Ikehara is about.

The input is coefficient nonnegativity together with the boundary remainder data on the real segment sigma ∈ (1, 2]; the growth bound for the partial sums is derived from them, not assumed. Nonnegativity turns the boundary data into the one-sided estimate βˆ‘ β€–a nβ€– / n ^ sigma ≀ B / (sigma - 1) on (1, 2], and inserting sigma = 1 + 1 / log t into it bounds βˆ‘_{n ≀ t} β€–a nβ€– by a multiple of t log t. That is weaker than the Chebyshev bound O(t) which Wiener--Ikehara ultimately proves, but it is available before any Tauberian argument, and one logarithm to spare is all the summability needs: the Fourier transform of a smooth compactly supported function decays faster than |v| ^ (-3), so the factor attached to a n is O((log n) ^ (-3)), and the Abel-summation bound TauCeti.LSeries.LSeriesSummable_mul_of_norm_le converts O(t log t) partial sums into a convergent series.

Only the values of the boundary remainder on the real segment sigma ∈ (1, 2] enter the growth bound, so the results below are stated with the boundary data restricted to that segment; the final asymptotic specializes the half-plane hypotheses it inherits from TauCeti.LSeries.tendsto_tsum_term_mul_fourier_atTop_of_contDiff.

Main results #

References #

The one-sided bound coming from the boundary data #

theorem TauCeti.LSeries.tsum_norm_term_le_of_boundary {a : β„• β†’ β„‚} {A : β„‚} {G : β„‚ β†’ β„‚} (ha : 0 ≀ a) (hG : ContinuousOn (fun (sigma : ℝ) => G ↑sigma) (Set.Icc 1 2)) (hG' : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ G ↑sigma = LSeries a ↑sigma - A / (↑sigma - 1)) (hsum : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ LSeriesSummable a ↑sigma) :
βˆƒ (B : ℝ), 0 ≀ B ∧ βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ (Summable fun (n : β„•) => β€–LSeries.term a (↑sigma) nβ€–) ∧ βˆ‘' (n : β„•), β€–LSeries.term a (↑sigma) nβ€– ≀ B / (sigma - 1)

For nonnegative coefficients, a boundary remainder G continuous on the real segment [1, 2] bounds the Dirichlet series on (1, 2] by B / (sigma - 1): the remainder is bounded on the compact segment, and the pole term contributes β€–Aβ€– / (sigma - 1).

No analytic continuation is used, only the values of G on that segment and the identity G = LSeries a - A / (s - 1) on its interior.

The partial-sum bound #

theorem TauCeti.LSeries.isBigO_sum_Icc_norm_of_boundary {a : β„• β†’ β„‚} {A : β„‚} {G : β„‚ β†’ β„‚} (ha : 0 ≀ a) (hG : ContinuousOn (fun (sigma : ℝ) => G ↑sigma) (Set.Icc 1 2)) (hG' : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ G ↑sigma = LSeries a ↑sigma - A / (↑sigma - 1)) (hsum : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ LSeriesSummable a ↑sigma) :
(fun (t : ℝ) => βˆ‘ k ∈ Finset.Icc 1 ⌊tβŒ‹β‚Š, β€–a kβ€–) =O[Filter.atTop] fun (t : ℝ) => t * Real.log t

A crude growth bound for the partial sums. Nonnegative coefficients whose Dirichlet series has a boundary remainder continuous on the segment [1, 2] satisfy βˆ‘_{1 ≀ n ≀ t} β€–a nβ€– = O(t log t).

The proof inserts sigma = 1 + 1 / log t into tsum_norm_term_le_of_boundary: the truncation n ≀ t costs a factor t ^ sigma = e t, and the bound B / (sigma - 1) is B log t. This is one logarithm short of the Chebyshev bound O(t) that Wiener--Ikehara eventually delivers, but it needs no Tauberian input.

The Fourier weight #

theorem TauCeti.LSeries.LSeriesSummable_mul_fourier_of_nonneg {a : β„• β†’ β„‚} {A : β„‚} {G : β„‚ β†’ β„‚} {psi : ℝ β†’ β„‚} {x : ℝ} (ha : 0 ≀ a) (hG : ContinuousOn (fun (sigma : ℝ) => G ↑sigma) (Set.Icc 1 2)) (hG' : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ G ↑sigma = LSeries a ↑sigma - A / (↑sigma - 1)) (hsum : βˆ€ (sigma : ℝ), 1 < sigma β†’ sigma ≀ 2 β†’ LSeriesSummable a ↑sigma) (hpsi : ContDiff ℝ (β†‘βŠ€) psi) (hsupp : HasCompactSupport psi) (hx : 0 < x) :
LSeriesSummable (fun (n : β„•) => a n * FourierTransform.fourier psi (1 / (2 * Real.pi) * Real.log (↑n / x))) 1

The Fourier-weighted series converges on the boundary line. For nonnegative coefficients with a boundary remainder continuous on the segment [1, 2] and a smooth compactly supported test function, the series tested at scale x > 0 is summable at s = 1. This discharges the standing summability hypothesis of TauCeti.LSeries.tendsto_tsum_term_mul_fourier_atTop_of_contDiff.

theorem TauCeti.LSeries.tendsto_tsum_term_mul_fourier_atTop_of_nonneg {a : β„• β†’ β„‚} {A : β„‚} {G : β„‚ β†’ β„‚} {psi : ℝ β†’ β„‚} (ha : 0 ≀ a) (hG : ContinuousOn G {z : β„‚ | 1 ≀ z.re}) (hG' : βˆ€ (z : β„‚), 1 < z.re β†’ G z = LSeries a z - A / (z - 1)) (hsum : βˆ€ (sigma : ℝ), 1 < sigma β†’ LSeriesSummable a ↑sigma) (hpsi : ContDiff ℝ (β†‘βŠ€) psi) (hsupp : HasCompactSupport psi) :
Filter.Tendsto (fun (x : ℝ) => βˆ‘' (n : β„•), LSeries.term a 1 n * FourierTransform.fourier psi (1 / (2 * Real.pi) * Real.log (↑n / x))) Filter.atTop (nhds (2 * ↑Real.pi * A * psi 0))

The smoothed Wiener--Ikehara asymptotic for nonnegative coefficients. Testing the Dirichlet series of a nonnegative coefficient system against a smooth compactly supported function on the line Re s = 1 gives the limit 2Ο€ A psi 0, where A is the residue subtracted off by the continuous boundary remainder G.

This is TauCeti.LSeries.tendsto_tsum_term_mul_fourier_atTop_of_contDiff with its summability hypothesis discharged: for nonnegative coefficients the boundary data itself forces the Fourier-weighted series to converge at every large scale. The hypotheses are now exactly the analytic input of Wiener--Ikehara.