Zeta measure associated to \({\mathbb{F}}_ q[T]\) (Q915779)

This is an important and well-written paper. The result contained in it is surprisingly simple and leads one to expect a deep and rich theory that should explain it. Before discussing the paper itself, we recall the situation of the Riemann zeta-function \(\zeta\) (s). Basic to the theory of \(\zeta\) (s) are the two integrals relating it to the exponential function via the \(\Gamma\)-function. The first integral expresses \(\zeta\) (s) - we ignore the \(\Gamma\)-factor - as an integral involving the sum \(\sum^{\infty}_{a=1}e^{-ax};\) this sum can be expressed as u/(1-u) where \(u=e^{-x}\). The second integral, which is obtained in a similar way, relates \(\zeta\) (s) with a theta function. The rational function, \(R(u)=u/(1-u)\) of the first integral can also be obtained from the p-adic theory. In fact, the Dirac measure at a is \(u^ a\) in the standard interpretation of measures as power-series in u-1. Thus, \(R(u)=\sum^{\infty}_{a=1}u^ a\) can be recovered as a sum of Dirac measures. Now let \({\mathfrak A}={\mathbb{F}}_ q[[T]]\) where q is a prime power. We can associate to \({\mathfrak A}\) the \(\zeta\)-function \[ \zeta_{{\mathfrak A}}(s)=\sum_{n\in {\mathfrak A},n\quad monic}n^{-s}. \] This function converges for s a positive integer but, in fact, can be defined on a much larger space where it can be analytically continued. At the ``negative'' integers, \(\zeta_{{\mathfrak A}}(s)\) has good special values that can be interpolated at a finite prime v of \({\mathfrak A}\). This interpolation can be expressed as an integral over involving a v-adic measure \(\mu\) which may also be given as a sum of Dirac measures as above. Using the ``Carlitz-polynomials'' one is able to associate to \(\mu\) an a priori complicated sum of hyper-differential operators as classical theory associates to a measure a power series. From the discussion given above concerning the Riemann zeta-function, one can view this sum of operators as a v-adic reflection of some theory at the infinite prime of \({\mathfrak A}.\) Based on hand calculations of the author, and some computer calculations of the reviewer, it became apparent that the sum associated to \(\mu\) actually converges to a simple answer. This answer only involved the ``magic numbers''; i.e., a collection of numbers which originally arose in the Kummer-Vandiver conjecture for \({\mathfrak A}\). It is these calculations that Thakur establishes in complete generality in the paper being reviewed. It now becomes important to search for their meaning. Recently, the reviewer has been able to use these hyper-derivatives to establish such things as Poisson summation and the like for \({\mathfrak A}\). However, at the present time much remains to be accomplished. We finish with pointing out a curious phenomenon associated to some of the formulae contained in the paper. In Theorem X, the author quotes a result of E.-U. Gekeler on the power sums \(S_ i(k)=sum\) of k-th powers of monics of degree i; such sums occur in zeta-values at negative integers. Gekeler shows that if \(k<q^{i+1}-1,\) then \[ S_ i(k)=M\prod_{j\leq i}L_{i-j}^{q^ j(k_ j-(q-1))}g_ k; \] where \(k=\sum k_ jq^ j\), \(M=\pm\) a multinomial coefficient, and \(g_ k\) is Carlitz's factorial. But, let \({\tilde \pi}\) be the period of the Carlitz module (i.e., the \({\mathfrak A}\)-analog of \(2\pi\) i) and let \(\zeta_{{\mathfrak A}}(j)\) be the value of \(\zeta_{{\mathfrak A}}(s)\) at positive j. Then (up to sign) \[ L^{-q^ j}_{i-j}=\zeta_{{\mathfrak A}}(q^ i-q^ j)/{\tilde \pi}^{q^ i-q^ j}. \] In other words: \(S_ i(k)=M\times factorial\times zeta\)-value; where \(M\in {\mathbb{Z}}/(p).\) Is this formula also a reflection of the same underlying theory that describes the author's calculations of measures?