Prime in quadratic progressions on average (Q546372)

\textit{G. H. Hardy} and \textit{J. E. Littlewood} [Acta Math. 44, 1--70 (1923; JFM 48.0143.04)] conjectured that \[ \sum\limits_{n\leq x} \Lambda\left(n^2+k\right) \sim \mathfrak{S}(k)x \quad \text{as } x\rightarrow \infty, \] where \(\Lambda(n)\) is the von Mangoldt function and \(\mathfrak{S}(k)\) is a constant that depends only on \(k\). In [\textit{S. Baier} and \textit{L. Zhao}, Math. Ann. 338, 963--982 (2007; Zbl 1136.11059)], it was shown that this holds in an average sense. More precisely, the following result was proved: Given \(A,B,C>0\), we have, for \(x^2L^{-A}\leq y\leq x^2\), that \[ \sum\limits_{n\leq x} \Lambda\left(n^2+k\right) = \mathfrak{S}(k)x+O\left(xL^{-B}\right) \] holds for all square-free \(k\) not exceeding \(y\) with at most \(O\left(yL^{-C}\right)\) exceptions. Here, as in the following, \(L=\log x\). In the paper under review, the \(y\)-range is widened if the above asymptotics is replaced by a suitable lower bound. The authors show that there exist effective positive constants \(\delta'\) and \(\delta''\) such that for \(x^{2-\delta'}\leq y\leq x\) \[ \sum\limits_{x/2<n\leq x} \Lambda\left(n^2+k\right) \gg GxL^{-2} \] holds true for all integers \(1\leq k\leq y\) with at most \(O\left(y^{1-\delta''}\right)\) exceptions, where \(G\) depends (weakly) on \(x\) and possible Siegel zeros. The basic approach is to use the circle method to count the number of primes \(p\) and positive integers \(n\) in the relevant ranges such that \[ k=p-n^2. \] This is a problem similar to that of counting the solutions \(p\) and \(n\) of \[ k=p+n^2, \] which was studied in [\textit{R. Brünner}, \textit{A. Perelli} and \textit{J. Pintz}, Acta Math. Hung. 53, 347--365 (1989; Zbl 0683.10039)]. The idea in the last-mentioned paper was to calculate the effect of possible Siegel zeros and to use Bombieri's density theorem to estimate the contribution of the ``main'' terms attached to them. The authors carry this idea over to the paper under review. It should be noted that in [\textit{S. Baier} and \textit{L. Zhao}, ``On primes in quadratic progressions'', Int. J. Number Theory 5, 1017--1035 (2009; Zbl 1221.11197)], the \(y\)-range for which the expected \textit{asymptotics} holds has been widened to \(x^{1+\varepsilon}\leq y\leq x^2\) by a completely different method that uses a dispersion approach.