Complete solution of the polynomial version of a problem of Diophantus (Q1434329)

The problem of Diophantus referred to in the title consists in finding sets of integers \(\{a_1,\dots,a_m\}\) such that \(a_ia_j+1\) is a perfect square for all \(1\leq i<j\leq m\). Let us call such a set a Diophantine \(m\)-tuple. There has been a lot of activity towards understanding Diophantine \(m\)-tuples in recent years, most notably by the first author who showed that there exist at most finitely many (effectively computable) Diophantine quintuples and no Diophantine sextuple [see \textit{A. Dujella}, J. Reine Angew. Math. 566, 183--214 (2004; Zbl 1037.11019)]. In the paper under review, the authors look at a polynomial version of Diophantus's problem. The above-mentioned result immediately implies that there do not exist Diophantine quintuples of non-zero polynomials with integer coefficients, not all constant. The main result of this paper shows that all the Diophantine quadruples \(\{a,b,c,d\}\) of such polynomials are regular, in the sense that \((a+b-c-d)^2=4(ab+1)(cd+1)\). The proof of this result is long and technical. Assuming \(\{a,b,c\}\) are given, one needs to show that \(d\) is uniquely determined. First, the authors reduce this problem to solving equations of the form \(u_n=v_m\), where \((u_n)_{n\geq 0}\) and \((v_m)_{m\geq 0}\) are binary recurrent sequences of polynomials whose initial terms and characteristic equations depend on \(a,~b\) and \(c\). Then they use a very precise determination of the initial terms of these sequences together with several gap principles and congruence relations to reach the desired conclusion.