On a Diophantine equation of Erdős and Graham (Q2197527)

In the book of \textit{ P. Erdős} and \textit{R. L. Graham} [Old and new problems and results in combinatorial number theory. Genéve: L'Enseignement Mathématique, Université de Genève (1980; Zbl 0434.10001)] problem 63 is about the equation \[ \frac{n}{2^n}=\sum_{i=1}^k\frac{a_i}{2^{a_i}},\tag{1} \] with \(k>1\). One may ask if for any given \(n\) there is a solution in \(k,a_1,\ldots,a_k\), or if for any given \(k\) there is a solution in \(n,a_1,\ldots,a_k\). It is also an interesting problem if any \(\alpha\in (0,2)\) can be represented as \[ \alpha=\sum_{i=1}^{\infty}\frac{a_i}{2^{a_i}}.\tag{2} \] Formerly \textit{P. B. Borwein} and \textit{T. A. Loring} [Math. Comput. 54, No. 189, 377--394 (1990; Zbl 0699.10019)] investigated these questions. The authors formulate basic theoretical results on the solutions of (1). The solutions for \(k\le 8\) are enumerated.They prove that \[ a_k\le 2^{k+2}+2k(\log_2k-1)-4, \] implying that \(a_k\) can be bounded in terms of \(k\). It is shown that for given \(k\) the equation has infinitely many effectively computable solutions \(n,a_1,\ldots,a_k\). On the other hand, an infinite set \(K\) is constructed, such that for any \(k\in K\) (1) has at least five solutions in \(n,a_1,\ldots,a_k\). It is shown that for \(n\le 10^4\) (1) has a solution in \(k,a_1,\ldots,a_k\). They construct an infinite set \(R\) of rational numbers, such that for each \(x\in R\) the number \(x\) has at least nine representations in the form (2). In the proof the authors solve a discrete logarithm problem and apply among others greedy algorithm.