Linear independence of the values of \(q\)-hypergeometric series and related functions (Q2490442)

The main aim of this paper is to extend linear independence theorems of \textit{J.-P. Bézivin} [Manuscr. Math. 61, 103--129 (1988; Zbl 0644.10025)] under the setting of \textit{Y. André} [Ann. Math. (2) 151, 741--756 (2000; Zbl 1037.11050)]. To give an impression of the results, let \(K\) be an algebraic number field, and let \(\mathcal{M}\) denote the set of all places of \(K\), the absolute value \(|\;|_w\) of \(K\) being canonically normalized for any \(w\in \mathcal{M}\). Assuming distinct \(\theta_1,\dots , \theta_h\in K^\times\), let \(\widetilde{\mathcal{M}}:=\{w\in\mathcal{M}\mid\max(|\theta_1|_w,\dots ,|\theta_h|_w)>1\}\), and suppose that there exists \(v\in \widetilde{\mathcal{M}}\) with \(|\theta_1|_v>|\theta_2|_v\geq\dots \geq |\theta_h|_v>1\). With \(\lambda_1,\dots ,\lambda_h\in K^\times\), suppose \(A(n):=\sum_{i=1}^h \lambda_i\theta_i^n \neq 0\) for all \(n\in \mathbb{N}_0\), and assume that, for every \(w\in \widetilde{\mathcal{M}}\), there exist \(n_w\in \mathbb{N}\) and a real \(c_w>1\) with \(|A(n+1)|_w\geq c_w|A(n)|_w\) for every \(n\geq n_w\). Then, for every \(w\in \widetilde{\mathcal{M}}\), the series \(\Phi(z):= \sum_{n=0}^\infty z^n/\prod_{\nu=0}^n A(\nu)\) defines an entire function on the algebraic closure \(\mathbb{C}_w\) of the completion of \(K\) with respect to \(w\). On these conditions, let \(\alpha_1,\dots ,\alpha_m\in K^\times\) such that no quotient \(\alpha_i/\alpha_j\) with \(i\neq j\) belongs to the multiplicative subgroup of \(K^\times\) generated by \(\theta_1,\dots ,\theta_h\). Then, for any \(\ell\in \mathbb{N}_0\), the equation \(a_0+\sum_{i=0}^\ell \sum_{j=1}^m a_{ij}\Phi^{(i)}(\alpha_j)=0, \underline{a}:=(a_0,a_{01},\dots ,a_{0m},\dots ,a_{\ell 1},\dots ,a_{\ell m})\in K^{1+(\ell+1)m}\) in \(K_w\) for all \(w\in \widetilde{\mathcal{M}}\) implies \(\underline{a}=\underline{0}\). As in Bézivin's paper quoted above, the rationality criterion of Borel-Dwork is the main ingredient of the proof. Notice that the case \(|\theta_h|_v\leq 1\) is considered as well, but then additional hypotheses are needed. A second aim of this paper is to give a qualitative refinement of the second author's [Math. Ann. 325, 123--136 (2003; Zbl 1025.11023)] main result under the setting of André.