Artin's conjecture for abelian varieties (Q2374203)

The following theorem is proved: let \(A\) be an abelian variety of dimension~\(r\) over \({\mathbb Q}\) and let \((a_1,\dots,a_g)\in A({\mathbb Q})\) be \({\mathbb Q}\)-linearly independent points. (In particular, the \({\mathbb Q}\)-rank of \(A\) is at least~\(g\).) Suppose also that at least one of the \(a_i\) is not divisible by \(2\) in \(A({\mathbb Q})\) and, for simplicity, that all \(2\)-torsion in \(A\) is rational. Consider the function \(f_{A,a,{\mathbb Q}}(x)\) that counts primes \(p<x\) of good reduction at which the finite group \(\bar A({\mathbb F}_p)/\langle \bar a_1,\dots\bar a_g\rangle\) has ``at most \((2r-1)\) cyclic components'' (that is, it needs that many generators: this is what is called the length of the abelian group in lattice theory). Then \(f_{A,a,{\mathbb Q}}\gg x/(\log x)^2\). The author asserts that this is ``the right generalization of the classical Artin's primitive root conjecture in the context of abelian varieties'', and in particular denies that title to the the conjecture of [\textit{S. Lang} and \textit{H. Trotter}, Bull. Am. Math. Soc. 83, 289--292 (1977; Zbl 0345.12008)]. That is a matter of opinion, but the method of proof here is elementary (some mention is made of Tate modules but they are not used): by contrast, the Lang-Trotter conjecture and subsequent developments of it raise subtle questions about density of primes and, consequently, the behaviour of zeta functions.