On the residues of products of prime numbers (Q579309)

This paper contains a modest attack to the problem proposed by \textit{P.Erdős} that for any sufficiently large prime \(q\) and any residue class \(a\not\equiv 0\) modulo \(q\) the congruence \(p_1p_2\equiv a (\bmod q)\) can be solved in primes \(p_1\leq q\) and \(p_2\leq q\). All considerations are subject to the quasi-Riemann hypothesis \(H(\theta_q,x)\), i.e., it is supposed that for all characters \(\chi\) modulo \(q\) the \(L(s,\chi)\) do not vanish in the domain \(\text{Re }s>\theta_q\), \(|\text{Im }s| < x^{1-\theta_q}.\) The generalized Riemann hypothesis is \(H(1/2,\infty)\) but this is not enough to imply the above conjecture. There are three possible ways to weaken it, which can be satisfied (i) with almost all residue classes mod \(q\), (ii) with the product of three primes instead of two, and (iii) with a little bit larger primes \(p_1\) and \(p_2.\) It is proved that (i) if \(H(\theta_q,q)\) is true then \(p_1p_2\equiv a (\bmod q)\), \(p_1\leq q\), \(p_2\leq q\) can be solved for all but \(cq^{2\theta_q-1}\log^5q\) residue classes \(a\not\equiv 0\) modulo \(q\); (ii) if \(H(\theta_q,q)\) is true with \(\theta_q < 1-(3+\epsilon)\frac{\log\log q}{\log q}\) then \(p_1p_2p_3\equiv a (\bmod q)\), \(p_1 \leq q\), \(p_2 \leq q\), \(p_3 \leq q\) can be solved; (iii) if the generalized Riemann hypothesis is true then \(p_1p_2\equiv a\) (mod q), \(p_1 \leq cq \log^4q\), \(p_2 \leq cq \log^4q\) can be solved.