A purely exponential Diophantine equation in three unknowns (Q2151162)

The authors consider the exponential equation \[(1)\;\;\; a^x+(ab+1)^y=b^z\ \ \text{in } x,y,z\in\mathbb{Z}_{>0},\] where \(a,b\) are integers \(>1\). They give a list of various pairs \((a,b)\) for which (1) is solvable. This list consists of a couple of infinite families and a finite number of isolated cases. With the exception of two of the infinite families, for all pairs \((a,b)\) from the list the complete resolution of (1) can be deduced from the literature. In the paper under review, the authors consider the two families from the list that had not been treated so far, i.e., \par (A): \(a=(b^{2i}-1)/(b+1)\) with \(i\geq 1\), \((b,i)\not=(3,1),(4,1)\); \par (B): \((a,b)=(F_{2k}, F_{2k+1}))\) with \(k\geq 2\),\par where \((F_n)_{n\geq 0}\) denotes the Fibonacci sequence. The authors prove that in case (A), the only solution of (1) is \((x,y,z)=(1,1,2i)\), while in case (B), the only solution is \((2,1,2)\). The proof consists of ingenious elementary arguments, an application of an estimate of \textit{Y. Bugeaud} [Compos. Math. 132, No. 2, 137--158 (2002; Zbl 1171.11318)] for linear forms in two \(p\)-adic logarithms, and computations.