On the exceptional set for Goldbach's problem in short intervals (Q5938008)

The purpose of this paper is to show that for any \(\theta>1/3\), there is a positive number \(\delta=\delta(\theta)\) such that the following holds: For any real number \(X\geq 1\), the number of even integers in the interval \([X,X+X^{\theta}]\) that cannot be written as a sum of two primes is \(O(X^{\theta-\delta})\), where the implied constant depends at most on \(\theta\). However the proof of Lemma 3 of this article (a zero density estimate for Dirichlet \(L\)-functions) stands upon an error, so this theorem is not yet established completely. But the author has recently informed the reviewer that she could succeed in removing the error, and retrieving the theorem. The latter correction will appear in due course (see Zbl 1111.11312). Although the proof contains the above issue at present, the main theorem of this paper substantially improves the previous best in this direction, proved by Q. Yao in 1982, which asserts the above conclusion with \(7/12\) in place of \(1/3\) for the lower limit of \(\theta\). The proof follows the strategy of the famous work of \textit{H. L. Montgomery} and \textit{R. C. Vaughan} [Acta Arith. 27, 353--370 (1975; Zbl 0301.10043)], and a new form of Deuring-Heilbronn phenomenon is presented. The treatment of the minor arc integral is essentially due to \textit{H. Mikawa} [Tsukuba J. Math. 15, 19--29 (1991; Zbl 0735.11041)].