Exact algorithms for \(p\)-adic fields and epsilon constant conjectures (Q733352)

There are several algorithms for computations in number fields. This is not so for \(p\)-adic fields. The problem is how to represent \(p\)-adic objects. One way is to use approximations and analyze the errors given by the approximations. However sometimes it is possible to give precise representatives of \(p\)-adic objects and in that way there are no errors due to approximations. The latter is the approach taken by the authors. The main goal of this paper is to describe algorithms for some computational problems in Galois extensions of \(p\)--adic fields. The strategy is to represent a local Galois extension as the completion of a global Galois extension. In this way it is possible to use the existing algorithms of global number fields and avoid to consider approximations. The main result of the paper is: There exists an algorithm which for a given prime number \(p\) and positive integer \(n\) proves or disproves the local epsilon constant conjecture of the second author [J. Lond. Math. Soc. (2) 70, No. 2, 289--306 (2004; Zbl 1068.11075)] for every Galois extension \(L/{\mathbb Q} _ p\) of degree \(n\). As a corollary, the authors obtain that there exists an algorithm which for a given positive integer \(n\) either proves the global epsilon conjecture of [Proc. Lond. Math. Soc. (3) 87, No. 3, 545--590 (2003; Zbl 1056.11071)] for every Galois extension \(L/{\mathbb Q}\) of degree \(n\) or finds a counterexample to the local epsilon constant conjecture. In general, this approach cannot be used to computationally disprove the global conjectures. For instance, suppose that an algorithm finds a counterexample \(M/{\mathbb Q}_ p\), with \([M:{\mathbb Q}_ p]\) dividing \(n\), to the local epsilon constant conjecture. If \(p\) is odd, there exists a Galois extension \(L\) of \({\mathbb Q}\) that is a counterexample to the global epsilon constant conjecture, though not necessarily of degree \(n\). However, if the counterexample is for \(p=2\), we cannot conclude that this disproves the global epsilon constant conjecture. Theorem 1.1 and its corollaries are considered as theoretical results and therefore the efficiency of the algorithms are not considered in the paper.