On twin primes associated with the Hawkins random sieve (Q2503371)

The paper under review is concerned with asymptotic estimates for the number of twin primes associated to the Hawkins random sieve. This sieve is a probabilistic analogue to the sieve of Eratosthenes. It starts with the set \(S_1\) of natural numbers greater than 1. Every element of \(S_1\setminus \{2\}\) is then sieved out, independently of the others, with probability \(1/2\). Let \(S_2\) be the resulting set and define \(P_2:=\min S_2\). Now every element of \(S_2\setminus \{P_2\}\) is sieved out with probability \(1/P_2\). In general, at the \(n\)-th step, the set \(S_{n+1}\) is obtained from the set \(S_n\setminus \{P_n\}\) by sieving out every element with probability \(1/P_n\), where \(P_n:=\min S_n\). The sequence of the numbers \(\{P_1,P_2,\dots\}\), known as Hawkins primes, mimics the sequence of primes in a statistical sense. For instance, it is known that the analogue of the prime number theorem for Hawkins primes holds with probability 1. \textit{M. C. Wunderlich} [Acta Arith. 26, 59--81 (1974; Zbl 0257.10033)] proved that also the analogue of the twin primes conjecture for Hawkins primes with difference \(k=1\) holds with probability 1. Following the lines of Wunderlich, the auhors extend this result to \(k=2\) which requires rather more work but is straightforward. The authors observe that for larger \(k\), the calculations in Wunderlich's method become extremely complicated. They develop a different approach which allows them to extend Wunderlich's result to arbitrary \(k\) without running into such complicated calculations. Using their approach, they further extend this result to \(l\)-tuples of Hawkins primes.