Answers are grounded in this paper and cite their sources.
For an odd prime p, let np be the least quadratic nonresidue and rp the least prime quadratic residue modulo p. We give two explicit Fourier test functions for the extremal constant 𝒞(1) of Carneiro, Milinovich, Quesada-Herrera and Ramos (Mathematics of Computation, doi:10.1090/mcom/4154), whose published upper bound is 𝒞(1)−2 < 0.7615. A single-shift function with 36 rational coefficients certifies 𝒞(1)−2 < 0.760777412, and a two-shift function with 24 rational coefficients certifies 𝒞(1)−2 < 0.76079030. Under the generalized Riemann hypothesis for Dirichlet L-functions, Theorems 1 and 2 of that paper then give limsup np/(log p)2 < 0.760777412 and the same bound for rp. The Fourier inequalities themselves are unconditional. Equivalently 𝒞(1) > 1.1464924. This removes about 38% of the interval for 𝒞(1)−2 left open by the authors' Theorem 5. Each witness is certified by at least two separately written interval-arithmetic verifiers. Single-shift witness: an Arb trapezoid bound with an exact second-derivative estimate, and a Cauchy–Schwarz bound with exact panel means; both use exact rational root isolation for the numerator. Two-shift witness: Arb complex ball quadrature, an outward-rounded mpmath trapezoid bound, and exact root isolation with validated Simpson panels. A Lean 4 file (Mathlib v4.24.0) proves the infinite-tail estimate and the final ratio comparison for the two-shift witness. Floating-point evaluations of the extremal quotient in the original variables agree with both certificates. Files: fourier-bounds-note.pdf is the note. fourier-bounds-replication.zip contains the exact witnesses, all verifiers, the floating-point cross-checks, the optimizer, the Lean source and the TeX source. Context: L. Alpöge's announcement of a Lean formalization covering Dirichlet GRH motivated this search. The bounds on 𝒞(1)−2 do not depend on it. If that result is accepted, it discharges the GRH hypothesis of both corollaries about primes. OpenAI Codex assisted with the numerical searches, verification code and preparation of the note. Anthropic's Claude independently reran all verifiers and the Lean check before deposit.
No takes yet. Share an insight, caveat, or question.
Lennart Lopin (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: