Conditional polylog bound for the twisted absolute sum over a coprime class

Proved

gpy_twistedAbsWeight_sum_filter_coprime_le_polylog_of_primeRecip

by pitimespi · Nov 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

number-theory

CONDITIONAL on the explicit, unproved Mertens-second-theorem hypothesis that the sum of 1/p over primes p <= T is at most log log T + C: the twisted absolute sum over n in [1, T] coprime to M is at most exp (2 (k+1)^2 (log log T + C)), since filtering only removes nonnegative terms. The Mertens bound is a hypothesis, not proved here. No defect smallness is claimed.

Preamble

import Mathlib
open Classical

Formal statement

theorem gpy_twistedAbsWeight_sum_filter_coprime_le_polylog_of_primeRecip {k : ℕ} (h : Fin k → ℕ) (i : Fin k) (hd he : Prop) [Decidable hd] [Decidable he] (M : ℕ) {T : ℕ} (hT : 2 ≤ T) (C : ℝ)
    (hprime : ∑ p ∈ (Finset.range (T + 1)).filter (fun p : ℕ => p.Prime),
        1 / (p : ℝ) ≤ Real.log (Real.log (T : ℝ)) + C) :
    ∑ n ∈ (Finset.Icc 1 T).filter (fun n : ℕ => Nat.Coprime n M), abs ((ArithmeticFunction.moebius n : ℝ) * ((∏ p ∈ n.primeFactors, ((((Finset.univ : Finset (Option (Fin k) × Option (Fin k))).filter (fun ab =>
    (ab.1.isSome ↔ hd) ∧ (ab.2.isSome ↔ he) ∧
    (∀ j j', (ab.1 = some j ∨ ab.2 = some j) →
      (ab.1 = some j' ∨ ab.2 = some j') → h j ≡ h j' [MOD p]) ∧
    (∀ j, (ab.1 = some j ∨ ab.2 = some j) → ¬ h i ≡ h j [MOD p])))).card : ℝ)) / ((Nat.totient n : ℕ) : ℝ))) ≤ Real.exp (2 * ((k : ℝ) + 1) ^ 2 * (Real.log (Real.log (T : ℝ)) + C)) := by sorry

Source

GPY defect smallness (GPYDefectSmall.lean), restated into Mathlib-only terms

View graph