A higher-dimensional analogue of Carlitz-Drinfeld theory (Q1898244)

Let \({\mathbf A}= \mathbb{F}_r [T]\) be the polynomial ring over a finite field. As is well-known, this ring has much in common with the integers \(\mathbb{Z}\). It has a division algorithm, it has unique factorization with a finite unit group, it is discrete in the topology given by the ``prime at infinity'' and is co-compact in the completion \({\mathbf K}\) (of \(\mathbb{F}_r (T)\)) given by this infinite prime. In the classical case of \(\mathbb{Z}\), we have the usual theory of \(L\)-functions based on the exponentiation \(a^s= e^{(s \log a)}\) for positive \(a\). In the case of \(\mathbb{F}_r [T]\) we can define exponentiation in the following manner. Let \(a\in {\mathbf A}\) be monic of degree \(d\). Note that ``monic'' is the correct generalization of ``positive''; moreover the notion of monicity can be defined locally at \(\infty\) via the Laurent expansion of \(a\) in terms of \(1/T\), and so can be extended to all of \(\mathbb{K}^*\). Let \(\langle a\rangle:= a/ T^d\), so \(\langle a\rangle\) is a 1-unit in \({\mathbf K}\) and, as such, may be raised to the power \(y\) for any \(y\in \mathbb{Z}_p\) (simply use the binomial theorem -- the sum will trivially converge). Put \(S_\infty:= \mathbb{C}^*_\infty \times \mathbb{Z}_p\) where \(\mathbb{C}_\infty\) is the completion of an algebraic closure of \({\mathbf K}\); then if \(s= (x,y)\in S_\infty\) we set \(a^s:= x^d\cdot \langle a\rangle^y\). One checks that this definition has all the usual properties of exponentiation and leads to a theory of \(L\)-series for Drinfeld \(\mathbb{F}_r [T]\)-modules, etc. For more general affine rings of curves there is also a theory of \(L\)-series. The subtlety lies in the fact that \({\mathbf A}\) will in general have a non-trivial class group. Let \(I\) thus be any ideal of \({\mathbf A}\). As explained above, the notion of monicity (for any \({\mathbf A})\) can be defined in general via expansions at \(\infty\); one sees that there is always a number \(e\) so that \(I^e\) is principal and generated by a monic element \(\alpha\). One defines \(\langle \alpha\rangle^y\) exactly as above. Moreover, one sees easily that the 1-units in \(\mathbb{C}_\infty\) are a \(\mathbb{Q}_p\)-module. As such one sets \(\langle I\rangle^y:= \langle \alpha \rangle^{y/ e}\) and, for \(s\in S_\infty\), \(I^s:= x^{\deg I}\cdot \langle I\rangle^y\). In this fashion one constructs \(L\)-series in the context of Drinfeld modules for any \({\mathbf A}\). The most basic example of these \(L\)- functions is the zeta function of \({\mathbf A}:= \sum_I I^{-s}\) where \(I\) runs over all the non-trivial ideals of \({\mathbf A}\). The basic connection between this function and Drinfeld modules lies in an analog of Euler's famous result relating special zeta values and periods of rank one Drinfeld modules. Recently, in ``A higher-dimensional generalization of the Goss zeta function'' [J. Number Theory 50, 363-375 (1995; Zbl 0819.11055)]\ \textit{M. Kapranov} presented a remarkable generalization of the above theory. Kapranov works with \(\mathbb{F}_r [T_1, \dots, T_n ]\) for arbitrary \(n\). Replacing \(\infty\) with a maximal flag of subspaces \(X= \mathbb{P}^n/ \mathbb{F}_r \supset X_{n-1} \supset \dots \supset X_0\), he defines an \(n\)-dimensional complete local field \({\mathbf K}\simeq \mathbb{F}_{r^m} ((t_1)) \dots ((t_n))\) in the sense of Parshin. Using this he is able to define an exponentiation generalizing the above for ``monic'' \(f(T_1, \dots, T_n)\) where \(S_\infty\) is now \((\mathbb{C}^*_\infty)^n \times \mathbb{Z}_p\). Kapranov then defined the zeta function of \(\mathbb{F}_r [T_1, \dots, T_n ]\) which has much in common with the basic curve case \(n=1\). However,and very importantly, as yet there is no analog of Euler's result. In the note being reviewed, the author shows how one can carry over formally some of the theory from \(n=1\) to arbitrary \(n\). There are many questions left open by these results. However, there is one main question that needs to be answered; a positive answer to this question should lead to solutions of all others. Recall that the theory for \(n=1\) corresponds to the theory of Drinfeld modules and exponentials, and that a Drinfeld module is an algebraic object with an analytic uniformization. \textit{Question}: Do there exist analogs of Drinfeld modules for \(n>1\)? Any progress along these lines would be very valuable.