On a conjecture of Erdős, Pólya and Turán on consecutive gaps between primes (Q722382)

A conjecture of Erdős, Pólya and Turán, described by \textit{P. Erdős} [Recent progress in analytic number theory, Vol. 1, 1--3 (Durham, 1979), Academic Press London-New York (1981; Zbl 0459.10002)], states that if \(\alpha_1,\ldots,\alpha_m\) are non-zero integer constants, not all of the same sign, then the sum \[ \sum_{i=1}^m\alpha_i d_{n+i}\;\;\;(d_n:=p_{n+1}-p_n) \] changes sign infinitely often as \(n\rightarrow\infty\). In his paper Erdős remarks that he could not even prove that \(d_{n}>d_{n+1}+d_{n+2}\) infinitely often. The present paper proves the conjecture, even establishing the stronger result that for every \(m\) there is a positive exponent \(c=c(m)\) such that \[ \frac{d_n}{(\log n)^c\max(d_{n-m},\ldots,d_{n-1},d_{n+1},\ldots,d_{n+m})} \] is unbounded as \(n\to\infty\). (The author comments that Konyagin has shown that one can take \(c(m)=\tfrac14\) for all \(m\)). The proof uses a result of \textit{W. D. Banks} et al. [Proc. Lond. Math. Soc. (3) 113, No. 4, 515--539 (2016; Zbl 1410.11114)], which is itself established via the Maynard-Tao method.