A Hasse-type principle for exponential Diophantine equations over number fields and its applications (Q1795159)

In the paper under review, the authors study the \({\mathcal S}\)-unit equation \[ \alpha_1 \beta_{11}^{u_{11}}\cdots \beta_{1\ell}^{u_{1\ell}}+\cdots+\alpha_k\beta_{k1}^{u_{k1}}\cdots \beta_{k\ell}^{u_{k\ell}}=\gamma. \] Here, \(\boldsymbol{\alpha}=(\alpha_1,\ldots,\alpha_k)\), \(\boldsymbol{\beta}=(\beta_{ij})_{\substack{1\le i\le k\\ 1\le j\le \ell}}\) and \(\gamma\) are algebraic numbers and the \(\mathbf{u}=(u_{ij})_{\substack{1\le i\le k\\ 1\le j\le \ell}}\) are non-negative integers. Let \({\mathbb L}\) be a field containing the components of \(\boldsymbol{\alpha}\) and \(\boldsymbol{\beta}\). The authors propose a conjecture which essentially says that if there is no non-negative integer solution \(u\) to the above equation then there must exist an ideal \({\mathcal M}\) of \({\mathcal O}_{\mathbb L}\) which signals the non-existence of such solutions. That is, the local equation obtained by asking that the difference between the left and right--hand sides above being a member of \({\mathcal M}\) does not have a non-negative integer solution in the \(\mathbf{u}\). The advantage of such a conjecture, if true, and if furthermore one could compute \({\mathcal M}\), then the non-existence of a solution modulo \({\mathcal M}\) amounts in a finite computation. This is a variant of a conjecture of Skolem. The authors offer a theorem in support of this conjecture. The theorem says that if one writes \(H\) for the set of all right--hand sides as \(\mathbf{u}\) runs over matrices with non-negative entries, then for every ideal \({\mathcal U}\) of \({\mathcal O}_{\mathbb L}\) and any \(\varepsilon>0\), there is an ideal \({\mathcal M}\) which is a multiple of \({\mathcal U}\) such that the cardinality of \(H\mod {\mathcal M}\) is smaller than \(N({\mathcal M})^{\varepsilon}\), where \(N({\mathcal M})\) is the norm of \({\mathcal M}\). The authors also provide a method to solve particular instances of this equation and as an illustration they find all triples of Balancing numbers \((B_u,B_v,B_w)\) (which are members of the Lucas sequence \(\{B_n\}_{n\ge 0}\) of characteristic equation \(x^2-6x+1\)) which sum up to a power of \(2,3,5\) or \(7\). While this last numerical result can also be found using linear forms in logarithms of algebraic numbers, the authors apply their approach and find, for each equation and each potential \textit{maximal} solution which they have found by an exhaustive search into the small box \(\max\{u,v,w\}\le 100\) a suitable modulus, that indeed indicates that the solution that they have cannot be extended to a larger one. For example, for \(b=2\) the modulus is \(2^9\cdot 3\cdot 17\cdot 257\cdot 7681\).