The Elliott-Halberstam conjecture implies the Vinogradov least quadratic nonresidue conjecture (Q2346705)

For any odd prime \(p\), let \(n(p)\) be the least positive integer which is a quadratic non-residue of \(p\). Vinogradov conjectured that \(n(p)\ll_{\varepsilon} p^{\varepsilon}\) for any fixed \(\varepsilon>0\). This is known to follow from the Generalized Riemann Hypothesis, but the present article shows that it is also a consequence of the Elliot--Halberstam conjecture. Indeed it would suffice to know that \[ \sum_{q\leq X^{1-\varepsilon}, p\nmid q}\left|\psi(x;q,p)-\frac{x}{\phi(q)}\right| \ll_{\varepsilon}\frac{x}{\log x} \] uniformly for all primes \(p\leq x\), for every fixed \(\varepsilon>0\). A similar result is established for other characters \(\chi\), where one considers the least positive integer \(n\) with \(\chi(n)\not=1\). The idea behind the proof is quite simple. If \(\chi(r)=1\) for every prime \(r\leq p^{\kappa}\) say, for some fixed \(\kappa>0\), then one can show that there is an \(X\in[q^2,q^A]\) with \(A=A(\kappa)\), such that \[ \left|\sum_{X<n\leq 2X}\chi(n)\Lambda(n)\right|\gg_{\kappa}X. \] This follows immediately from ``classical'' results about zeros of Dirichlet \(L\)-functions, or can be deduced by (complicated) elementary means, as is done in the paper. The first key idea is then to use the periodicity of \(\chi\) to deduce that \[ \left|\sum_{X-p<n\leq 2X-p}\chi(n)(\Lambda(n+p)-1)\right|\gg_{\kappa}X. \] One now writes \(\chi=f*\mu\), and sees that \(f(r^e)=0\) for any prime \(r\leq p^{\kappa}\), and \(|f(r^e)|\leq 2\) in general. It follows that \(f(n)\ll_{\kappa}1\) for \(n\leq 2X\). The use of \(f\), with these special properties, forms the second key idea in the proof. One now replaces \(\chi(n)\) by \(\sum_{rs=n}f(r)\) and finds that the contribution for \(r\leq X^{1-\varepsilon}\) can be handled by the Elliot-Halberstam conjecture, producing a bound \(O_{\kappa,\varepsilon}(X(\log X)^{-1})\). On the other hand, since \(r\) only contains prime factors \(>p^{\kappa}\) a sieve upper bounds shows that the remaining \(r\) contribute \(O_{\kappa}(\varepsilon X)\). This leads to a contradiction if one takes \(\varepsilon\) small enough compared with \(\kappa\). The paper also shows how to handle primitive roots, modulo primes \(p\) for which \(p-1\) has a bounded number of prime divisors. There is also a discussion showing how a general Type II estimate extending the Bombieri-Vinogradov range, of the type developed by \textit{Y. Zhang} [Ann. Math. (2) 179, No. 3, 1121--1174 (2014; Zbl 1290.11128)], implies a non-trivial bound for a character sum \(\sum_{n\leq N}\chi(n)\) with \(N\) somewhat below \(p^{1/2}\).