Density functions for prime and relatively prime numbers (Q1237089)

This is a sequel to papers by \textit{P.Erdős} and \textit{J.L.Selfridge} [Proc. Manitoba Conf. numer. Math. 1971, 1-14 (1971; Zbl 0267.10054)] and \textit{D.Hensley} and \textit{I.Richards} [Acta Arith. 25, 375-391 (1974; Zbl 0285.10004)]. Let \(\rho^*(x)\) be the maximum number of primes in any interval beyond \(x\) of length \(x\). Let \(r^*(x)\) be the maximum number of pairwise coprime integers in any interval of length \(x\). A finite set \(S\) of integers is ''\(\rho^*-admissible\)'' if for each prime \(p\) some residue class \((mod p)\) excludes all elements of \(S\). \(S\) is ''\(r^*-admissible\)'' if for each prime \(p\) some residue class \((mod p)\) excludes all but at most one element of \(S\). The prime \(k\)-tuples hypothesis asserts that if \(\{b_1< b_2< \dots< b_k\}\) is \(\rho^*-admissible\) then there are infinitely many positive integers \(n\) for which all of \(n+b_1, n+b_2, \dots, n+b_k\) are prime. Under the prime \(k\)-tuples hypothesis it is proved that \(\rho^*(x)\) is the number of elements in a maximal \(\rho^*-admissible\) set in any interval of length \(x\) (proposition 4). With no hypothesis (proposition 5) \(r^*(x)\) is the maximum number of elements in any \(r^*\)-admissible set in any interval of length \(x\). Sieve methods are used to get upper an lower bounds on \(r^*(x)-\rho^*(x)\). Namely, theorem 1: There is an effectively computable \(c> 0\) for which \(r^*(x)-\rho^*(x)> x^c\) for all sufficiently large \(x\). Theorem 2: Under the prime \(k\)-tuples hypothesis, \[ r^*(x)-\rho^*(x)=o(x/\log^2x)\quad\text{as}\quad x\to\infty. \] The previously known lower bound was \(log x\). Since Hensley and Richards have proved under the prime \(k\)-tuples hypothesis that \(\rho^*(x)< \pi(x)+Kx/\log^2x\), then it appears that \(r^*(x)\sim\rho^*(x)\). This is not surprising, however, under the prime \(k\)-tuples hypothesis we have the even stronger fact that \(r^*(X)-\rho^*(X)=o(x/\log^2x)\) where as \(\rho^*(x)-\pi(x)> Kx/\log^2x\). Thus it seems that \(\rho^*(x)\) is much closer to \(r^*(x)\) than to \(\pi(x)\). Of course, the prime \(k\)-tuples hypothesis is a rather strong assumption which has as yet not been verified even for \(k = 2\).