Arithmetic equivalence for function fields, the Goss zeta function and a generalisation (Q963001)

Let \(A={\mathbb F}_q[T]\) be the polynomial ring over the finite field \({\mathbb F}_q\) of characteristic \(p\), and let \(F={\mathbb F}_q(T)\) be its quotient field. Let \(K\) and \(L\) be two geometric extensions of \(F\) inside a finite Galois extension \(N\) of \(F\). The authors first show that the following three properties are equivalent: (1) [Arithmetic equivalence] For every irreducible polynomial \({\mathfrak p}\) in \(A\) outside a set of Dirichlet density zero, the splitting type of \({\mathfrak p}\) in \(K\) and \(L\) is the same. (2) [Split equivalence] For every irreducible polynomial \({\mathfrak p}\) in \(A\) outside a set of Dirichlet density zero, the number of prime ideals in \(K\) and \(L\) above \({\mathfrak p}\) is the same. (3) [Gaßmann equivalence] Every conjugacy class in \(\text{Gal}(N/F)\) meets \(\text{Gal}(N/K)\) and \(\text{Gal}(N/L)\) in the same number of elements. This is analogous to the classical situation for number fields (as extensions of \(F={\mathbb Q}\)), where all three properties are equivalent to \(K\) and \(L\) having the same Dedekind zeta function. See the encyclopedic source [\textit{N.~Klingen}, Arithmetical similarities. Prime decomposition and finite group theory. Oxford: Clarendon Press (1998; Zbl 0896.11042)]. Now for function fields we have the notion of the Weil zeta function, which takes values in \({\mathbb C}\) and depends only on the function field itself, and the characteristic \(p\) valued Goss zeta function, which depends also on \(F\) and \(A\). But although Gaßmann equivalent extensions have the same Weil and Goss zeta functions, the converse is not true. Extensions \(K\) and \(L\) with the same Goss zeta function and \([K:F]<p\) and \([L:F]<p\) are shown to be Gaßmann equivalent, while there are counterexamples that are Galois of degree \(p\). Therefore the authors introduce a characteristic \(0\) valued lift \(\zeta_K^{[0]}(s)\) of the Goss zeta function of \(K\), which combines the ability to distinguish the places of \(F\) with being suitable for counting their extensions in \(K\) (not just modulo \(p\)). Then they show that \(\zeta_K^{[0]}(s)=\zeta_L^{[0]}(s)\) is indeed equivalent to any of the above three equivalences. In particular, \(\zeta_K^{[0]}(s)=\zeta_L^{[0]}(s)\) implies that \(K\) and \(L\) have the same Weil zeta function and the same Goss zeta function. But they need not be isomorphic as extensions of \(F\); as in the classical situation, the minimal degree of a counterexample is \(7\).