On the desingularization of the unipotent variety
On the desingularization of the unipotent variety
复制标题
论单能品种的去奇异化
DOI:
10.1007/bf01390010
复制
发表时间:
1976
影响因子:
3.1
通讯作者:
R. Steinberg
中科院分区:
文献类型:
--
作者:
R. Steinberg
Theorem 1.1. Let G be a (connected) semisimple algebraic group with separable universal covering, V the variety of unipotent elements of G, B a Borel subgroup of G, and W the subset of all (u, g B) in V x G/B such that gI vg B. Then~: W-~ V is a desingularization of V.Here the nonsingular elements of V are just the regular elements, all conjugate, studied in some detail in [16]. Next, in case G is simple, come the subregular elements, dense among the singular elements of V and forming a single class of codimension 2 along which V has its generic singularity. Thig turns out to be a Kleinian singularity, an isolated rational double point on an algebraic surface, the one with the same label as that of G: for example, the surface singularity xy= z r+ l, commonly labeled A,, occurs as the generic singularity for V in case G is of type A r. Furthermore, with some restrictions on char k, the characteristic of the base field, the universal deformation of the surface singularity can be naturally achieved within the corresponding group G. In view of these remarkable facts (and others), conjectured by Grothendieck and proved by Brieskorn and Tits (mostly in unpublished notes; see, however,[5] and the last part of [17]) attention is focused on the deeper singularities of V, hence on all of the fibres of n. Observe that the fibre above u is (G/B)., the variety of" flags" fixed by u, or, equivalently, the variety of Borel subgroups containing u. In this article we obtain some results about the dimensions of these fibres and the numbers of irreducible components, relating the latter to elements of the Weyl group W. The basic idea, introduced in [17], is that given two components Y, Z of (G/B), there exists a unique w in the Weyl group such that gl B and g2 B are in the attitude w, that is, g~ lg2~ BwB, for a dense open set of (gxB, g2 B) in YxZ. This enables us to get a sort of classification of pairs of components in terms of elements of W and of components in terms of involutions and to prove that (.) dim (G/B).< 1/2 (dim G,-r). Here and elsewhere G. is the centralizer of u and r