Linnik's approximation to Goldbach's conjecture, and other problems (Q2344336): Difference between revisions

A classical result by Linnik states that every sufficiently large even integer \(n\) can be expressed as the sum of two primes and \(K\) powers of \(2\), where \(K\) is a fixed positive integer. This result is viewed as an approximation to the binary Goldbach conjecture, since the sequence of \(K\)-fold sums of powers of \(2\) is rather thin. Linnik did not specify the value of \(K\) in his result, but through the years a number of researchers have shown that progressively smaller values of \(K\) are admissible. Presently, we know that \(K = 12\) is admissible unconditionally, and that \(K = 7\) is admissible on the assumption of the Generalized Riemann Hypothesis for Dirichlet \(L\)-functions. An unconditional bound for \(K\) was obtained independently by \textit{Z. Liu} and \textit{G. Lü} [Int. J. Number Theory 7, No. 5, 1317--1329 (2011; Zbl 1237.11042)] and by \textit{C. Elsholtz} (unpublished), while a conditional bound was established a decade earlier by \textit{D. R. Heath-Brown} and \textit{J. C. Puchta} [Asian J. Math. 6, No. 3, 535--566 (2002; Zbl 1097.11050)] and by \textit{J. Pintz} and \textit{I. Z. Ruzsa} [Acta Arith. 109, No. 2, 169--194 (2003; Zbl 1031.11060)]. The proofs of these results involve a delicate analysis that fuses together an application of the circle method, sieve methods, results on the distribution of primes in arithmetic progressions, and harmonic analysis of lacunary trigonometric polynomials. At the end, one is left to select \(K\) as the least integer satisfying the inequality: \[ \lambda^{K-2} < C \tag{*}, \] where \(\lambda\) and \(C\) are explicit constants that occur in the proof. The paper under review starts with a brief overview of the inequality (*). In particular, the authors explain the structure of the constant \(C\), which is a rather formidable expression (see equation (2) in the paper) built up of \(5\) other constants, several of them of independent interest and the subjects of independent research. The authors describe those constants and provide estimates and/or approximations for several of them. Then they focus on one of those five constants and on the value of \(\lambda\). The estimation of both values involves a healthy amount of numerical calculations that had not been ``pushed to the limit'' in prior work on the problem. The main thrust of the paper under review is the re-examination of that numeric work in order to push it reasonably close to its limit and to examine what one can then say about the value of \(K\) in Linnik's theorem. The authors come tantalizingly close to improving both known records: They show that one needs \(K \geq 11.0953\) unconditionally and \(K \geq 6.09353\) on GRH. The paper concludes with a brief section that surveys several theorems on sums of powers of primes and powers of \(2\) inspired by the above work on Linnik's problem (e.g., sums of four squares of primes and \(K\) powers of \(2\)). The authors give references to the state of art results in each problem and report possible improvements on those results that can be achieved with the help of the numeric work they carried in regard to the value of \(\lambda\) in Linnik's theorem.