On a ternary Diophantine equation involving fractional powers with prime variables of a special form (Q2155872): Difference between revisions

In the classical Waring-Goldbach problem one studies the representability of integers as sum of powers of primes. The study of the arithmetical properties of \(p+2\) for primes \(p\) is related to the famous twin prime conjecture. In the present paper, the authors study an interesting hybrid variation of both problems. They prove that for \(1<c<2173/1930=1.1259\cdots\) and \(N\) sufficiently large, the Diophantine equation \[ \lfloor p_1^c\rfloor+\lfloor p_2^c\rfloor+\lfloor p_3^c\rfloor=N \] is solvable in prime variables \(p_1\), \(p_2\), \(p_3\) such that each of \(p_1+2\), \(p_2+2\), \(p_3+2\) has at most \[ \lfloor 11387/(4346-3860c)\rfloor \] prime factors, counted with multiplicity. This improves on a result of \textit{Z. Petrov} [God. Sofiĭ. Univ., Fak. Mat. Inform. 104, 171--183 (2017; Zbl 1474.11171)] who had the smaller range \(1<c<17/16=1.0625\). The proof uses Rosser's sieve and exponential sum techniques (exponent pairs). The paper is very well written, contains a nice historical overview on the problem and an extensive bibliography.