Disproving a weaker form of Hooley's conjecture

TL;DR

This paper demonstrates that a weakened form of Hooley's conjecture regarding the variance of primes in arithmetic progressions, even with a damping weight, does not hold in certain ranges, challenging previous assumptions.

Contribution

The paper proves that the estimate $G_\eta(x;q) \ll x\log q$ fails in the range $q \asymp \log \log x$, disproof of a weaker Hooley's conjecture with damping factors.

Findings

The estimate $G_\eta(x;q) \ll x\log q$ is false for $q \asymp \log \log x$.

Damping factors do not restore the validity of Hooley's conjecture in the studied range.

Abstract

Hooley conjectured that $G(x;q) \ll x\log q$, as soon as $q\to +\infty$, where $G(x;q)$ represents the variance of primes $p \leq x$ in arithmetic progressions modulo $q$, weighted by $\log p$. In this paper, we study $G_\eta(x;q)$, a function similar to $G(x;q)$, but including the weighting factor $\eta\left(\frac{p}{x}\right)$, which has a dampening effect on the values of $G_\eta$. Our study is motivated by the disproof of Hooley's conjecture by Fiorilli and Martin in the range $q \asymp \log \log x$. Even though this weighting factor dampens the values, we still prove that an estimation of the form $G_\eta(x;q) \ll x\log q$ is false in the same range.