On regularity of small primes in function fields (Q909711)

The paper concerns calculations with analogs of Bernoulli numbers in the arithmetic of global fields over \({\mathbb{F}}_ q\), \(q=p^ n\). These rational numbers are used to study the p-class numbers of cyclotomic abelian extensions of the rational field. As an example of the results obtained we have the following neat result on certain elliptic curves (viewed, of course, as coverings of the rational field): Let q be odd with \(q\equiv 1\) (mod 3). Set \(m=q-1\), and let c be a non-square in \({\mathbb{F}}_ q\). Then \[ Y^ 2=X^ 3+c\text{ has p-division points over }{\mathbb{F}}_ q \Leftrightarrow 1=(4c)^{m/6}\begin{pmatrix} 2m/3\\ m/3\end{pmatrix} \text{ in }{\mathbb{F}}_ q. \] The paper also contains a potentially very important conjecture which we now describe. This conjecture involves the Bernoulli-analog which arises from the function field version of Euler's computation of zeta-values at even positive integers. These elements are called ``Bernoulli-Carlitz numbers''. Let A be the polynomial ring over \({\mathbb{F}}_ q\) and let k be the rational field (i.e., quotient field of A). Let f be a prime polynomial in A. Associated to f we have the cyclotomic function field k(f) and its ``real'' (i.e., totally-split at \(\infty)\) subfield \(k(f)^+\). Let \(\tilde C(f)\) be the p-primary class group of the ring of A-integers of \(k(f)^+\). (It is important to note that this group differs from the class group of the field \(k(f)^+\) by the regulator group which may be non-trivial.) Upon tensoring \(\tilde C(f)\) with the Witt ring W of \(A/(f)\) one can decompose into isotypic components under the Teichmüller character \(\omega: A/(f)^*\to W^*\). From the fact that \(k(f)^+\) splits completely at \(\infty\) one sees that only ``even'' (i.e., divisible by \(r-1\)) powers arise. One knew (by using the appropriate ``Kummer'' morphism constructed by S. Okada) that if an isotypic component was non-zero then an associated Bernoulli-Carlitz number was non-zero. However, the naive approach to formulating a converse did not work. In this paper the author formulates a very reasonable converse. His idea is to take into account the action of Frobenius. More precisely, one knows that the Frobenius will permute the various isotypic components, and these components live or die together. Thus Gekeler defines the following equivalence relation on positive integers: \(i\equiv_ fj\) \(\Leftrightarrow\) there exists an integer t with \(i\equiv p^ tj\) (mod \(q^{\deg (f)}-1)\). Then the author gives some evidence for the following Conjecture: Suppose that for all \(0<i,j<q^{\deg (f)}-1\), i,j even, and \(i\equiv_ fj\), we have f divides (the numerator of) the j-th Bernoulli- Carlitz number. Then \(\tilde C(f)(\omega^ i)\) should be non-trivial.