Work

The Sharp Global Brun–Titchmarsh Constant and Explicit Wheeled Short-Interval Prime Bounds

Year: 2026
Type: preprint
Abstract: Montgomery and Vaughan (1973) proved π(x+y) − π(x) ≤ 2y/log y for all real x > 0, y > 1, and Yamada (2023) established π(x+y) − π(x) < 2y/(log y + 0.8601) (approaching log y + 0.9912 as y → ∞). Splitting (x, x+y] into φ(W₀) progressions modulo W₀ shrinks each sub-interval to y/W₀ and loses −log W₀ in the denominator. Combining single-board modulo-W₀ pre-wheeling in Selberg's Λ² sieve with the L¹ optimal-weight identity Σ_{d ≤ W, (d,W₀)=1} |λ_d| = Υ_{W₀}(W)/J_{W₀}(W), contour-free envelopes |Υ₂₁₀(W) − (756/(325π²))W| ≤ 1.1949√W and |J₂₁₀(W) − (8/35)(log W + η₂₁₀)| ≤ 3.6056/√W, periodic wheel envelopes U₃₀₀₃₀(⌈y⌉), and iterated 2^s Möbius deletion from Q₆ = 30,030, we prove the sharp global bound π(x+y) − π(x) < 2y/(log y + ξ₁*) for all x > 0, y > 1 with ξ₁* = 5/2 − log 5 = 0.8905620875…, attained uniquely as (x, y) → (2⁻, 5⁺) at {2, 3, 5, 7} with stability 0 < 2 − x ≤ y − 5 ≤ (10/3)(S(x, y) − ξ₁*) for S(x, y) < 9/10. On x ≥ 3, y > 1 we obtain the sharp twin-prime constant ξ_twin* = 2 − log 2 = 1.3068528194… with twin-cluster stability, 8/3 − 2 log 2 on x ≥ 2, and +2.315 on x ≥ 5, y > 12; on x > 0, y > 50 we prove π(x+y) − π(x) < 2y/(log y + 2.4225) < 2y/(log y + 2.42) (C_asymp(210) = 2.427905…, < 2y/(log y + 2.43452686) for y ≥ 10⁴⁸; verified in full across all lengths in Lean 4.19 and via a standalone zero-float 10⁻¹²ℤ Python suite).
Author: Weeks, Alben
Language: English
Open Access status: green
OpenAlex ID: W7234366929
Cites:
Cited by:
Related to: 10