Periods of \(t\)-motives and transcendence (Q1362091)

This important paper uses the theory of solitons in characteristic \(p\) to study the transcendency properties of special values of gamma functions associated to the Carlitz module. Thus this paper is an outgrowth of the techniques originated by \textit{G. Anderson} [in The arithmetic of function fields, de Gruyter, 51-73 (1992; Zbl 0797.11056); and Duke Math. J. 73, 491-542 (1994; Zbl 0807.11032)]. To begin the function field theory, we need to describe the set-up of the ``two \(T\)'s:'' in the arithmetic of \(\mathbb{F}_q[T]\), the element \(T\) really plays two distinct roles: in extensions of \(k:= \mathbb{F}_q(T)\) it is a scalar whereas via a Drinfeld module, or \(T\)-module, it is an operator. In function field theory, it is possible to actually separate these two distinct roles of \(T\) by using rings of the form \(\mathbb{F}_q [T] \otimes_{\mathbb{F}_q} \mathbb{F}_q[T]\), etc. Here the first copy of \(\mathbb{F}_q[T]\) contains the scalars while the second contains the operators. In order to have a simplicity of notation, the author lets \(T:=T \otimes 1= \text{``}T\) as scalar'' and \(t:=1 \otimes T= \text{``}T\) as operator.'' Next we briefly recall the particular \(\Gamma\)-function that is being studied in this paper, for more we refer the reader to Section 9.9 of [\textit{D. Goss}, Basic structures of function field arithmetic, Springer (1996; Zbl 0874.11004)]. The \(\Gamma\)-function will live in the universe of scalars. It is defined as \[ \Gamma (z):= {1\over z} \prod_{n\in \mathbb{F}_q [T] \text{ monic}} \left(1+ {z\over n} \right)^{-1}; \] it clearly converges for all \(z\neq n\), where \(n\) is a monic polynomial or 0. It is called the geometric gamma function and is easily related to the exponential of the Carlitz module. It has been known for over two decades that it is possible to describe Drinfeld modules in terms of locally free sheaves in a manner following Krichever's work on KdV [see, e.g., \textit{D. Mumford}, in: Proc. Int. Symp. Algebraic Geometry Kyoto 1977, 115-153 (1978; Zbl 0423.14007)]. Let \(f\) be a monic polynomial in \(t\) and let \(\zeta_f\) be a generator of the \(\mathbb{F}_q[t]\) module of \(f\)-division points under the Carlitz module. Let \(X_f\) be a smooth projective model of the field \(k_f:= \mathbb{F}_q (T,\zeta_f)\). Following the analogy with KdV, Anderson was able to construct certain functions defined on an open subset of \(X_f\times_{\text{Spec} (\mathbb{F}_q)} X_f\) which he called solitons; i.e., Anderson shows that there is a meromorphic function \(\varphi\) on \(X_f \times X_f\) with the property that the values of certain twists of \(\varphi\) at generic points are essentially the partial products used to define the geometric gamma function. The calculation of the divisor of the soliton by Anderson and Sinha established that it satisfies an equation similar to that of the ``shtuka function'' associated to sign normalized rank one Drinfeld modules (Basic structures\dots subsection 7.11). This then allows the author to construct certain geometric \(t\)-modules which play a role similar to the Jacobians of Fermat curves. These \(t\)-modules naturally come equipped with complex multiplication making them behave like rank 1 modules. On the other hand, these \(t\)-modules are not 1-dimensional and so it is not automatic that they are uniformizable (i.e., that their exponential functions are surjective). The key to the author's uniformizability results is the long diagram in subsection 1.3.3. This diagram, due to \textit{G. Anderson} and \textit{D. Thakur} [Ann. Math., II. Ser. 132, 159-191 (1990; Zbl 0713.11082)], reinterprets the canonical long exact sequence of the exponential of the Carlitz module in terms of homological algebra. The author also writes down a similar sequence for his geometric \(t\)-modules. The uniformizability is then established through the use of a ``Green's function'' which plays a role very similar to the one it plays in classical arithmetic [see \textit{B. Gross}, Arithmetic geometry, 327-339 (1986; Zbl 0605.14027)]. As the special values of the solitons are related to the gamma function, it is natural to suspect that the periods of the geometric \(t\)-modules of the author can be related to gamma values, and indeed such results are established by the author. Transcendency results are then deduced via results of Gelfond-Schneider-Lang type due to \textit{Jing Yu} [e.g., Duke Math. J. 58, 559-575 (1989; Zbl 0687.12008)]. The reader may well wonder about an analog of the ``two \(T\)'s'' for classical arithmetic. However, as \(\mathbb{Z} \otimes_\mathbb{Z}\) \(\mathbb{Z}\) is \(\mathbb{Z}\), such a separation is impossible (and one needs to check which role \(\mathbb{Z}\) is playing in any given situation). By abuse of language and science, one can view this as a sort of ``wave-particle duality'' in number theory\dots . It is my pleasure to thank Dinesh Thakur for kindly sharing his very valuable insights into the paper being reviewed.