A note on the dynamical zeta function of general toral endomorphisms (Q5961995)

The authors consider \(d\)-dimensional toral endomorphism which can be represented by an integer matrix \(M \in \text{Mat}(d,\mathbb Z)\), with action mod \(1\) on the \(d\)-torus \(\mathbb T^d \simeq\mathbb R^d /\mathbb Z^d\). Given such an integer matrix \(M\) the Artin-Mazur zeta function \(\zeta_M\) is defined by \[ \zeta_M(z) := \exp \left( \sum_{m=1}^\infty \frac{a_m}{m} z^m \right) = \prod_{m=1}^\infty (1-z^m)^{-c_m}. \] The coefficients \(a_m\) are given by \[ a_m = | \det ( \mathbf{1}-M^m)|. \] If \(M\) is a quasihyperbolic matrix, i.e.\ \(M\) has no roots of unity in its spectrum, then \(a_m\) and \(c_m\) are related by \[ a_m = \sum_{l\mid m} l \, c_l \qquad \text{and} \qquad c_m = \frac{1}{m} \sum_{l \mid m} \mu \left( \frac{m}{l} \right) a_l. \] \textit{S.\ Smale} [Bull.\ Am.\ Math.\ Soc.\ 73, 747--817 (1967; Zbl 0202.55202)] related already the Artin-Mazur zeta function and the Lefschetz zeta function for hyperbolic toral automorphisms (i.e.\ the associated matrix \(M\) has no eigenvalues in the unit circle). The authors give a different proof. From their introduction: ``In this note, we explain a different approach via elementary geometry and linear algebra, which bypasses more advanced topological methods as well as the need to calculate eigenvalues. A key observation is that all arguments apply to general endomorphisms without additional effort. [\dots] Also, we make several steps explicit to facilitate their computational use.''