Explicit prime-sieve quotient bound with a positive denominator
ProvedHelfgott.prime_sieve_quotient_explicit_error_boundexplicit-boundsgoldbachlarge-sieveminor-arcs
For a positive integer , define
where the infinite sum is over all primes and is Euler's constant. Put
For every real and , the quantity
is strictly positive, and
This converts the proved all-modulus remainder estimate and logarithmic lower bound into an unconditional bound for the actual sieve quotient. The maximum retains whichever lower bound is stronger. The theorem is valid even when the asymptotic lower bound alone is negative; no unproved numerical denominator condition is required. The constant 8.217 is the certified platform error constant, not the sharper 7.284 quoted in Helfgott's paper. The final uniform sieve constant and minor-arc integral estimate are separate obligations.
Preamble
import Definitions.Def_Helfgott_PrimeSieveCoprimeAsymptoticData import Mathlib.NumberTheory.Harmonic.EulerMascheroni open Helfgott Finset open scoped BigOperators Classical
Formal statement
theorem Helfgott.prime_sieve_quotient_explicit_error_bound (q : ℕ) (hq : 0 < q) (R S : ℝ) (hR : 1 ≤ R) (hS : 1 < S) :
let C := Real.eulerMascheroniConstant +
(∑' p : Nat.Primes, Real.log (p : ℕ) /
(((p : ℕ) : ℝ) * (((p : ℕ) : ℝ) - 1))) +
∑ p ∈ q.primeFactors, Real.log p / (p : ℝ)
let E := fun T : ℝ => (q : ℝ) / (q.totient : ℝ) * (8217 / 1000 : ℝ) *
(∏ p ∈ q.primeFactors, (1 + (p : ℝ) ^ (-(2 / 3 : ℝ))) /
(1 + ((p : ℝ) ^ (1 / 3 : ℝ) + (p : ℝ) ^ (2 / 3 : ℝ)) /
((p : ℝ) * ((p : ℝ) - 1)))) * T ^ (-(1 / 3 : ℝ))
0 < max (Real.log S) (Real.log S + C - E S) ∧
primeSieveG q R / primeSieveG q S ≤
(Real.log R + C + E R) / max (Real.log S) (Real.log S + C - E S) := by sorrySource
H. A. Helfgott, The ternary Goldbach conjecture is true, arXiv:1312.7748v2, Section 5.2, equations (5.10)-(5.14) and the quotient in (5.7). https://arxiv.org/html/1312.7748v2 . Uses the proved platform remainder Helfgott.actual_coprime_prime_sieve_error_8217 and Helfgott.prime_sieve_logarithmic_lower_bound; retains their certified error constant 8.217.