On prime twins in arithmetic progressions (Q2367079): Difference between revisions

Put, for a non-zero integer \(k\), \[ \psi(x,q,a,2k)=\sum_{{{0<m,n\leq x,m-n=2k} \atop {n\equiv a\pmod q}}} \Lambda(m)\Lambda(n). \] If \((a+2k,q)=1\), then \[ E(x,q,a,2k)=\psi(x,q,a,2k)-2\prod_{p>2} \left( 1- {1\over {(p-1)^ 2}}\right) \prod_{{{k,p>2} \atop {p| q}}} \left({{p-1} \over {p-2}}\right) \cdot {{x-| 2k|} \over {\varphi(q)}} \] is expected to be an error term. \textit{H. Maier} and \textit{C. Pomerance} [Trans. Am. Math. Soc. 322, 201-237 (1990; Zbl 0706.11052)] proved an analogue of the Bombieri-Vinogradov mean value theorem for the function \(E\). \[ \sum_{q\leq Q} \max_{(a,q)=1} \sum_{0<2k\leq x} | E(x,q,a,2k)|\ll x^ 2(\log x)^{-A} \] if \(Q=x^ \delta\) for some small \(\delta>0\). Using a variant of Linnik's dispersion method, the author gives a fairly simple proof for the Maier-Pomerance theorem with \(Q=x^{1/2}(\ln x)^{-B(A)}\).