The Goldbach-Vinogradov theorem with three primes in a thin subset (Q1273109)

In 1986, \textit{E. Wirsing} [Analysis 6, 285-308 (1986; Zbl 0586.10032)] showed the existence of a thin subset \(S\) of primes such that every sufficiently large odd integer can be written as a sum of 3 primes in \(S\), and that \[ | S\cap[1,x]| \ll (x\log x)^{(1/3)}. \] This result is the best possible apart from the logarithmic factor. But his probabilistic method did not present how such a thin set \(S\) can be constructed explicitly. Following \textit{D. Wolke}'s work [Acta Math. Hung. 61, 241-258 (1993; Zbl 0790.11075)], the author constructed a thin set \(S\), at the cost of the upper bound for \(| S\cap[1,x]| \), as a union of some sets of primes in arithmetic progressions, such that every sufficiently large odd integer can be written as a sum of 3 primes in \(S\), and that \[ | S\cap[1,x]| \ll x^{9/10}(\log x)^c, \] for some positive absolute constant \(c\). The last function supersedes \(x^{15/16}\) due to Wolke. In order to prove such a conclusion via Wolke's idea, one must find a larger integer \(r\) such that when \((b_j,r)=1\) for \(j=1,2\) and \(3\), every sufficiently large odd integer \(N\), congruent to \(b_1+b_2+b_3\) modulo \(r\), can be represented as \(N=p_1+p_2+p_3\) with primes \(p_j\equiv b_j\) (mod \(r\)) for \(j=1,2,3\). Indeed, the author shows that, roughly speaking, the latter statement is valid for almost all prime moduli \(r\leq R\), for \(R=N^{3/20}(\log N)^{-c}\) with some \(c>0\), while Wolke showed it for \(R=N^{1/11}\). This improvement comes from a mean value estimate on exponential sums over prime numbers supplied by \textit{J. Liu} and \textit{T. Zhan} [Acta Arith. 82, 197-227 (1997; Zbl 0889.11035)] and \textit{T. Zhan} and \textit{J. Liu} [Chin. Sci. Bull. 41, 363-366 (1996; Zbl 0851.11048)], and swiftly yields the aforementioned reduction on \(| S\cap[1,x]| \) by following Wolke's way.