The binary Goldbach problem with one prime of the form \(p=k^2+l^2+1\) (Q2483161): Difference between revisions

Let \(E(N)\) denote the number of integers \(n\equiv 0\) or \(4\pmod 6\) with \(n\leq N\), which cannot be written in the form \(n= p_1+ p_2\) with primes \(p_1\), \(p_2\) for which \(p_1= a^2+ b^2+ 1\). Then it is shown that \(E(N)= O(N(\log N)^{-A})\) for any fixed \(A\). Inspired by the work of \textit{H. Iwaniec} [Acta Arith. 21, 203--204 (1972; Zbl 0215.35603)], the paper detects primes of the form \(p= a^2+ b^2+ 1\) by using a half-dimensional lower bound sieve coupled with a ``reversal of rôles'' and an upper bound estimate. The former of these removes factors up to \(N^{4/9}\), while the latter looks at primes in the sequence \(1+ uq_1q_2\), with primes \(q_1\equiv q_2\equiv 3\pmod 4\) in the range \(q_i\geq N^{4/9}\). In order to handle the error terms which arise from the sieve, a level-of-distribution result is required, which is proved via the circle method. It is the treatment of this latter result which takes up the bulk of the paper.