On additive complements. III (Q402622)

Two sets \(A,B\) of non-negative integers are called ``additive complements'' if every sufficiently large positive integer is sum of two numbers, one of them belonging to \(A\) and the other one to \(B.\) Of course \(A\) and \(B\) may have elements in common or even be equal. It is known (see [\textit{J. Fang} and \textit{Y. Chen}, Proc. Am. Math. Soc. 138, No. 6, 1923--1927 (2010; Zbl 1202.11010)] for a reference) that the counting functions \(A(x)=|A\cap[1,x]|\) and \(B(x)\) of two additive complements \(A\) and \(B\) verify \(A(x)B(x)>x-c\) for a constant \(c\) and all \(x\geq 1\). Consequently \[ \liminf_{x\to+\infty}x^{-1}A(x)B(x)\geq 1. \] \textit{A. Sárközy} and \textit{E. Szemerédi} [Acta Math. Hung. 64, No. 3, 237--245 (1994; Zbl 0816.11013)] proved in 1994 that if \noindent \(\limsup_{x\rightarrow+\infty}x^{-1}A(x)B(x)\leq1\) (that is, if \(\lim_{x\rightarrow+\infty}x^{-1}A(x)B(x)=1),\) then \[ \lim_{x\to+\infty}A(x)B(x)-x=+\infty.\tag{1} \] The authors of the paper under review proved in [Proc. Am. Math. Soc. 138, No. 6, 1923--1927 (2010; Zbl 1202.11010)] that (1) still holds when \[ \limsup_{x\rightarrow+\infty}x^{-1}A(x)B(x)>2 ~~\text{or}~~\limsup_{x\rightarrow+\infty}x^{-1}A(x)B(x)<{5\over 4}. \] In the above result, the number 2 cannot be replaced by a smaller one: this was pointed out by the authors in [ibid. 139, No. 3, 881--883 (2011; Zbl 1202.11010)]. It was an open problem to replace \({5\over 4}\) by a bigger number. This is done in this paper: the authors managed to replace \({5\over 4}\) by \(3-\sqrt 3.\) The paper is well written, contains a clear introduction to the subject and informs the reader about results on additive complements in the set of \textit{all} integers. \texttt{Open problems:} Is it possible to replace \(3-\sqrt 3\) by a bigger number? Possibly, find the biggest such number. Estimate the growth of \(A(x)B(x)-x\) when (1) holds.