The number of powers of 2 in a representation of large even integers by sums of such powers and of two primes (Q4486469)

\textit{Yu. V. Linnik} [Mat. Sb., Nov. Ser. 32, 3-60 (1953; Zbl 0051.03402)] proved that every large even integer \(N\) is a sum of two primes and of a bounded number of powers of 2. Calling \(r_k(N)\) the number of such representations, \textit{J. Liu, M.-C. Liu} and \textit{T. Wang} [Sci. China, Ser. A 41, 1255-1271 (1998; Zbl 0924.11086)] recently showed that \(r_k(N)\gg N(\log N)^{k-2}\) for all even \(N\geq N_k\), where \(N_k\) depends on \(k\) alone, provided that \(k\geq 54 000\). The author proves the same result, in the smaller range \(k\geq 25 000\), and also that for any \(\varepsilon>0\), any \(k\geq k_0(\varepsilon)\) and any \(N\geq N_k(\varepsilon)\) one has \[ \sum_{2\mid n,\;n\leq N} \Bigl( r_k'(n)-2(\log_2 N)^k(\log N)^{-1} \Bigr)^2 \leq 2\varepsilon N(\log_2 N)^{2k}(\log N)^{-2}, \] where \(r_k'(n)\) denotes the number of representations of \(n\) in the form \(p+2^{\nu_1}+\cdots+2^{\nu_k}\), \(p\) being a prime number. Both proofs depend on an extremely careful and detailed analysis of all the constants involved in zero free regions and zero density estimates for the Dirichlet \(L\)-functions, in order to improve a key estimate in Liu, Liu and Wang's paper referred to above.