Analytic properties of zeta functions and subgroup growth (Q5931224)

The issue of counting the subgroups of index \(n\) (resp. \(\leq n\)) in a given group has received growing attention during the last decades. The state of the art in the early nineties has been documented in \textit{A. Lubotzky}'s paper in [Groups '93 Galway/St. Andrews'. Lond. Math. Soc. Lect. Note Ser. 212, 368-404 (1995; Zbl 0862.20020)]. If \(G\) is a finitely generated nilpotent group and \(a_n\) the number of subgroups of index \(n\) in \(G\), then the zeta function \(\zeta_G(s) := \sum_n a_n n^{-s}\) converges in some right half-plane and has an Euler product decomposition, the Euler factor at the prime \(p\) counting the subgroups of index a power of \(p\). Early examples suggested that the zeta functions arising in this way are closely linked to Dedekind zeta-functions. In particular it was thought that the Euler factors (known to be rational functions in \(p^{-s}\) for torsion free \(G\)) fall into finitely many classes of rational functions. One of the main results of the paper under review is that \(\zeta_G(s)\) has rational abscissa of convergence \(\alpha(G)\), that it can be continued meromorphically to a slightly larger half-plane and that the only pole with real part \(\alpha(G)\) is at \(\alpha(G)\). Using Tauberian theorems this implies that the number of subgroups of \(G\) with index at most \(N\) is asymptotically \(c N^{\alpha(G)}(\log N)^b\) for some real number \(c\) and some nonnegative integer \(b\). Similar results are achieved for the zeta functions counting normal subgroups in \(G\) or counting ideals (or subrings) of a ring \(R\) whose additive group is finitely generated torsion free. In addition to its concrete results, an important aspect of the current paper is to open up a new way to look at these zeta functions. The authors associate to \(G\) a reduced scheme \(D\) defined over the rationals such that for almost all primes \(p\) the \(p\)th Euler factor of \(\zeta_G\) can be calculated in terms of the number of points modulo \(p\) on the irreducible components of \(D\). Therefore, \(\zeta_G\) often will not behave like Dedekind zeta-functions and the Euler factors will not only consist of finitely many classes of rational functions. The main tool used in the paper is a thorough study of certain \(p\)-adic integrals, called cone integrals, following ideas of Denef, who applied such integrals in his work on Igusa zeta functions.