On Gaussian primes in sparse sets (Q6962535)

One of \textit{E. Landau}'s famous problems [Jahresber. Dtsch. Math.-Ver. 21, 208--228 (1912; JFM 43.0264.01)] asks for a proof of the existence of infinitely many primes \(p\) of the shape \(p = n^2 + 1\) with \(n\) integer. This motivated efforts to find primes of the shape \N\[\N p = n^2 + b^2 \tag{1}\N\]\Nwith \(n, b\) integer and \(b\) restricted to belong to a special set \(B\subset {\mathbb N}\), which we aim to take as sparse as possible, with the hope of exploiting the algebraic properties of the norm form \(n^2 + b^2\) and arithmetic in \({\mathbb Q}(i)\). \textit{E. Fouvry} and \textit{H. Iwaniec}'s early work [Acta Arith. 79, No. 3, 249--287 (1997; Zbl 0881.11070)] was concerned with sets \(B\) not sparser than powers of logarithms (such as the set of primes). Then the first instance of sets of power density was \textit{J. Friedlander} and \textit{H. Iwaniec}'s work [Ann. Math. (2) 148, No. 3, 945--1040 (1998; Zbl 0926.11068)] where they found an asymptotic formula for such primes with \(B\) itself the set of squares (thus \(p = n^2 + c^4\) for some integer \(c\)). We mention also a recent result in this direction (posterior to the paper under review) due to \textit{B. J. Green} and \textit{M. Sawhney} [``Primes of the form $p^2 + nq^2$'', Preprint, \url{arXiv:2410.04189}], which restricts \textit{both} \(n\) and \(b\) to be prime numbers.\N\NHere the author proves that there exist infinitely many primes of the shape (1) whenever \(B\) is \emph{any} set not too sparse in the sense that \N\[\N|B \cap [1, X]| \gg X^{1-\delta} \tag{2}\N\]\N for some positive~\(\delta\). The generality of the statement is striking; no particular algebraic property of \(B\) is assumed. The author also proves an asymptotic formula for the number of such primes, and for weighted count of the same type as Fouvry-Iwaniec's original statement.\N\NThe author's result that condition (2) is sufficient for counting primes of the shape (1) would follow from Fouvry-Iwaniec's arguments if we knew that Hecke \(L\)-functions for \({\mathbb Q}(i)\) satisfied the Riemann hypothesis. Thus the main achievement in the paper is to show that the arguments can be carried out without this hypothesis. This is done by a subtle reworking of the type I / type II estimates from Fouvry-Iwaniec's arguments, isolating the contribution from characters of low conductors (such as those associated to Siegel zeroes), and keeping track of the possible way that \(B\) can be correlated with them.