The multiplicity one case of Lusztig's conjecture. (Q984456)

This is about Lusztig's conjecture on the irreducible rational characters of a reductive algebraic group \(G\) over a field of positive characteristic \(p\). The author uses the formulation of the conjecture in terms of the Jordan-Hölder multiplicity \([\Delta(x\cdot_p0):L(w\cdot_p0)]\), where \(\Delta(x\cdot_p0)\) is a baby Verma module for the restricted Lie algebra of \(G\), and \(L(w\cdot_p0)\) is one of the irreducibles whose character one tries to compute recursively. The character of the standard module \(\Delta(x\cdot_p0)\) is well known. Assume that \(p\) is bigger than the Coxeter number. The conjecture predicts that the above Jordan-Hölder multiplicity equals \(p_{x,w}(1)\), where the `periodic polynomial' \(p_{x,w}\) is a close cousin of the Kazhdan-Lusztig polynomial \(P_{x,w}\). The main result of the paper is that indeed the multiplicity equals one if and only if \(p_{x,w}(1)\) equals one. The method of proof is not the one originally envisaged by Lusztig. Instead of perverse sheaves the author employs intersection sheaves on moment graphs. For an excellent introduction to this approach and its background see the author's paper [Bull. Lond. Math. Soc. 42, No. 6, 957-972 (2010; Zbl 1234.20054)]. Suffice it to say here that the moment graphs of Braden and MacPherson provide a rather combinatorial tool to study certain intersection cohomology complexes in characteristic zero, or hypercohomology of some parity sheaves in positive characteristic. In the hands of the author they allow a more precise grip than the perverse sheaves used in earlier attacks on the Lusztig conjecture.