Gauss sums for function fields (Q752761)

Let C be a complete, smooth curve over the field \(F_ q\). Let \(\infty\) be a fixed prime of C and A the ring of functions regular away from \(\infty\). Let \(\phi\) be a rank one Drinfeld A-module defined over a finite extension of the fraction field k of A. Let \({\mathcal P}\subset A\) be a prime and let \(\phi\) [\({\mathcal P}]\) be the \({\mathcal P}\) torsion points of \(\phi\). As is well-known \(\phi\) [\({\mathcal P}]\) is isomorphic to A/\({\mathcal P}\). Choose one such isomorphism \(\psi\) :A/\({\mathcal P}\to \phi [{\mathcal P}]\) (for example one can use the exponential function of \(\phi\)). Now let L be an extension of k which contains both \(\phi\) [\({\mathcal P}]\) and a constant field extension isomorphic to A/\({\mathcal P}\). Choose an \(F_ q\)- linear injection \(\chi\) : A/\({\mathcal P}\to L\); all other such injections are \(q^ j\)-th powers of \(\chi\). To \(\psi\),\(\chi\) the author associates the ``basic Gauss sum'': \[ \sum_{z\in A/{\mathcal P}^*}\chi (z^{- 1})\psi (z). \] For characters not coming in this fashion, the author defines the Gauss sum by multiplying the basic Gauss sums in the fashion of L. Carlitz using q-adic expansions. Such sums were originally studied by the author in the case \(A=F_ q[T]\) [Invent. Math. 94, 105-112 (1988; Zbl 0629.12014)] and, quite amazingly, all the results one would expect from classical theory hold. In the paper being reviewed, the author examines the Gauss sums in more general contexts. In particular he handles non-polynomial class {\#}1 rings A. One already sees in this situation that the results are not quite so simple as in the polynomial case. The author also discusses the case of composite moduli. Here, unlike the classical theory, one obtains an addition rule where things can also be quite unruly. - Perhaps the neatest result in the paper appears in the appendix. In it, the author in a few short lines establishes (in the prime case described above) that the Gauss sums are never zero!