On sums of three unit fractions with polynomial denominators (Q5933542)

The author conjectured that for every positive integer \(m\) there is a solution of \(\frac{m}{n}=\frac{1}{x_1}+ \frac{1}{x_2}+\frac{1}{x_3}\) in positive integers \(x_1,x_2, x_3\), for all sufficiently large integers \(n > n_0(m)\) [see \textit{W. Sierpiński} [Mathesis 65, 16-32 (1956; Zbl 0075.03302)]. This extended conjectures of Erdős and Straus, and Sierpiński. Partial answers were found by means of polynomial identities \(\frac{m}{ax+b}=\frac{1}{F_1(x)}+ \frac{1}{F_2(x)}+\frac{1}{F_3(x)}\) which solve the problem for certain residue classes. In this paper the author proves that identities of the following type cannot exist. (1) \(m \equiv 0 \bmod 4\), \(a>0\), \((a,b)=1\), and \(b\) is a quadratic residue modulo \(a\). (2) \(a>0, m>3b>0\). In particular, this means that no finite set of polynomial identities can solve the original conjecture. It was known since long (see for example Section D11 [\textit{R. K. Guy}, Unsolved problems in number theory. 2nd ed., Springer-Verlag (1994; Zbl 0805.11001)]) that Schinzel proved such results. Fortunately, they are now published.