Minimally dominant elements of finite Coxeter groups

TL;DR

This paper studies special elements in finite Coxeter groups called minimally dominant elements, revealing their geometric properties, conjugacy relations, and their role in constructing transverse slices and partitions in algebraic groups.

Contribution

It introduces the concept of minimally dominant elements, characterizes their properties, and connects them to Lusztig's partitions and transverse slices in reductive groups.

Findings

Minimally dominant elements are characterized by maximal length involutions.

These elements are conjugate via simple and cyclic shifts, not pseudo-Anosov braids.

They provide an alternative approach to Lusztig's inverse Kazhdan-Lusztig map.

Abstract

Recently, Lusztig constructed for each reductive group a partition by unions of sheets of conjugacy classes, which is indexed by a subset of the set of conjugacy classes in the associated Weyl group. Sevostyanov subsequently used certain elements in each of these Weyl group conjugacy classes to construct strictly transverse slices to the conjugacy classes in these strata, generalising the classical Steinberg slice, and similar cross sections were built out of different Weyl group elements by He-Lusztig. In this paper we observe that He-Lusztig's and Sevostyanov's Weyl group elements share a certain geometric property, which we call minimally dominant; for example, we show that this property characterises involutions of maximal length. Generalising He-Nie's work on twisted conjugacy classes in finite Coxeter groups, we explain for various geometrically defined subsets that their elements are conjugate by simple shifts, cyclic shifts and strong conjugations. We furthermore derive for a large class of elements that the principal Deligne-Garside factors of their powers in the braid monoid are maximal in some sense. This includes those that are used in He-Lusztig's and Sevostyanov's cross sections, and explains their appearance there; in particular, all minimally dominant elements in the aforementioned conjugacy classes yield strictly transverse slices. These elements are conjugate by cyclic shifts, their Artin-Tits braids are never pseudo-Anosov in the conjectural Nielsen-Thurston classification and their Bruhat cells should furnish an alternative construction of Lusztig's inverse to the Kazhdan-Lusztig map and of his partition of reductive groups.