A modular approach to the generalized Ramanujan-Nagell equation (Q2166103)

\textit{N. Terai} [Bull. Aust. Math. Soc. 90, No. 1, 20--27 (2014; Zbl 1334.11020)] discussed the solution of \(x^2+d^y=k^z,\, x,y,z\in \mathbb{N}\) in the case \(d = 2k -1\) and conjectured that for any integer \(k>1\), the equation has only one positive integer solution \((x , y , z) = (k - 1, 1, 2)\). His conjecture has been verified in some special cases most of which focus on the case \(k\not\equiv 0\pmod 4\). In the same paper, N.\ Terai used some classical methods to the equation for \(k\equiv 0\pmod 4\) with \(k \leq 30\). However, his method does not apply for \(k \in \{12, 24\}\). Until in 2017, \textit{M. A. Bennett} and \textit{N. Billerey} [Math. Comput. 86, No. 305, 1375--1401 (2017; Zbl 1407.11049)] used the modular approach to solve the case \(k \in \{12, 24\}\). It follows that the case \(k\equiv 0\pmod 4\) and \(2k - 1\) is an odd prime power is a very difficult case to the conjecture. In the current paper, using the modular approach, the authors prove that if \(k\equiv 0\pmod 4\), \(30 < k < 724\) and \(2k - 1\) is an odd prime power, then under the GRH, the conjecture is true.