Answers are grounded in this paper and cite their sources.
In a companion paper [1] we measured, for the primes of a fixed residue class a mod q,the distribution of the gaps between consecutive such primes modulo 2q2, and showedthat a sieved Hardy–Littlewood model Sz predicts it with no free parameter. Here weestablish how that distribution is connected to the zeros of ζ(s) and of the Dirichlet L-functions modulo q, i.e. to the objects of the (generalised) Riemann Hypothesis. Theconnection is a chain with two links. The first link is classical: by the explicit formulathe number N of class-a primes in a short interval fluctuates as a sum over the zeros, andunder RH and the pair correlation of the zeros its variance is given, for a fixed class, byVar N/EN = [log(qX/h) − ∑p|qlog pp−1 − γ − log 2π + 1]/ log X, a corollary of Montgomery–Soundararajan and of a lemma of Lemke Oliver–Soundararajan and Kuperberg; we verifyit on windows of width 109 at 1011, . . . , 1015 for q = 1, 3, 5, 7 (χ2 = 78 on 60 comparisons,Poisson excluded by 40–340 standard deviations). The second link is new. We definethe density elasticity εs = ∂ log ps/∂ log ρ of the gap-class frequencies, prove that in Szconditioning on the local count is an exponential tilt of the step weights, so that the gapclasses respond to a local excess of primes exactly as to a change of height (“local = global”),and show that the zero-driven fluctuation δN enters each class count as δNs = (1 + εs) ps δNplus multinomial noise, with transmission coefficients 1 + εs computed exactly from thesieve. The elasticities of the primes, measured by regression over 104 sub-windows at eachheight and for every residue class a, agree with S29 (χ2 = 10.5/10, 20.8/20, 31.7/30 forq = 3, 5, 7), and a blind test at q = 7, with the predictions deposited with SHA-256 digestsbefore any gap modulo 98 was read, gives χ2 = 24.5 on 24 degrees of freedom; the Cramérmodel restricted to the class is rejected (χ2 in the thousands) and orders the classes wrongly.A model-free check confirms the chain: the global derivative of the frequencies along theheight, measured across 1011–1015 with precision 0.003, equals the local response to thecount. The zeros thus reach the gap classes through one scalar, the local density, withcomputable weights: the long-gap class 0 mod 98 carries them at about two thirds of itsshare, the class 42 mod 98 at about 1.2 times. The same chain shows that the gap-classstatistic cannot locate a zero: a statistic that does not distinguish the zeros from a changeof height carries no information about their position beyond the count itself
No takes yet. Share an insight, caveat, or question.
Cristoni et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: