A note on trace of powers of algebraic numbers (Q2212628)

The paper under review deals with the problem of detecting algebraic integers amongst algebraic numbers, by looking at the traces of their powers. These questions were studied by \textit{G. Pólya} [Rend. Circ. Mat. Palermo 40, 1--16 (1915; JFM 45.0655.02)] and \textit{B. de Smit} [J. Number Theory 45, No. 1, 112--116 (1993; Zbl 0782.11027)], who showed that if \(\alpha\) is an algebraic number and \(\mathrm{Tr}_{\mathbb{Q}(\alpha)/\mathbb{Q}}(\alpha^i) \in \mathbb{Z}\) for some finite, explicit sequence of integers \(i\) (depending on the degree of \(\alpha\)), then \(\alpha\) is an algebraic integer. As a corollary, one sees that \(\alpha\) is an algebraic integer as soon as \(\mathrm{Tr}_{\mathbb{Q}(\alpha)/\mathbb{Q}}(\alpha^i) \in \mathbb{Z}\) for all but finitely many \(i \in \mathbb{N}\). This last assertion is generalised by Theorem 2 of the paper under review, which shows that an algebraic number \(\alpha\) is an algebraic integer as soon as, for any fixed algebraic number \(\lambda\), the trace \(\mathrm{Tr}_{\mathbb{Q}(\alpha, \lambda)/\mathbb{Q}}(\lambda \alpha^i)\) is integral and non-zero for all but finitely many \(i \in \mathbb{N}\). It is necessary that these traces are non-zero, as the examples \(\alpha = 1/\sqrt{2}\) or \(\alpha = (1 + \sqrt{5})/\sqrt{2}\) show. The rest of the paper under review (see in particular Theorem 5 and Theorem 12) aims to study pairs of non-zero algebraic numbers \(\alpha, \lambda\) such that \(\mathrm{Tr}_{F/K}(\lambda \alpha^i) = 0\) for infinitely many \(i \in \mathbb{N}\), where \(K \subseteq F\) are two number fields such that \(\mathbb{Q}(\alpha,\lambda) \subseteq F\). We mention in particular that Corollary 6 shows that \(\mathrm{Tr}_{\mathbb{Q}(\alpha)/\mathbb{Q}}(\alpha^i) = 0\) for infinitely many \(i \in \mathbb{N}\) if and only if \(\alpha \not\in \mathbb{Q}(\alpha^{h})\), where \(h\) denotes the order of the torsion part of the Galois group of \(\alpha\) (that is, the Galois group of the splitting field of the minimal polynomial of \(\alpha\)). Moreover, Corollary 13 proves the analogous result with the extension \(\mathbb{Q} \subseteq \mathbb{Q}(\alpha)\) replaced by \(\mathbb{Q}(\zeta) \subseteq \mathbb{Q}(\alpha,\zeta)\), where \(\zeta\) is a primitive root of unity of order \(h\). Perhaps surprisingly, the proofs of the paper under review use some deep results from Diophantine approximation, such as the theorem of Skolem, Mahler and Lech (see the work of \textit{G. Hansel} [Theor. Comput. Sci. 43, 91--98 (1986; Zbl 0605.10007)] for an elementary proof). Such a theorem does not have a literal analogue for function fields of positive characteristic (see however the work of \textit{H. Derksen} [Invent. Math. 168, No. 1, 175--224 (2007; Zbl 1205.11030)] for a possible analogue). The last section of the paper under review shows that also some of the aforementioned results do not have a literal analogue over function fields of positive characteristic.