Complex geometry and the asymptotics of Harish-Chandra modules for real reductive Lie groups. III: Estimates on n-homology (Q757618): Difference between revisions

[For part II see Invent. Math. 86, 255-286 (1986; Zbl 0657.22017).] In this paper the authors obtain new information, beyond that provided by the theory of asymptotics, on the number of embeddings of an irreducible admissible representation into principal series representations. Actually, they provide a lower bound for the number of embeddings and conjecture that this lower estimate captures all embeddings. They verify the conjecture for real semisimple Lie groups all of whose Cartan subgroups are connected. This lower bound is computed in the spirit of the Kazhdan-Lusztig conjecture. More precisely, they define, combinatorially, polynomials \(h_{X,w}(u)\), where X indexes a Harish- Chandra module and w indexes an element of the Weyl group. The polynomials are defined through relations between the Kazhdan-Lusztig polynomials for Harish-Chandra modules and Kazhdan-Lusztig polynomials for Verma modules. Let \(h_{X,w^+}(u)\) be the portion of \(h_{X,w}\) with positive coefficients. Then they prove that \(\dim_{{\mathbb{C}}}(H_ 0({\mathfrak n}^-,X)_{w\chi -\rho})\geq h_{X,w^+}(1)\). They conjecture that \(h_{X,w^+}=h_{X,w}\) and that \({\mathfrak h}_{X,w}(1)=\dim_{{\mathbb{C}}}(H_ 0({\mathfrak n}^-,X)_{w\chi -\rho}\). Here \({\mathfrak n}^-\) is the unipotent radical of the Iwasawa-Borel subalgebra associated to an Iwasawa decomposition of G. The proof uses the Jacquet functor and translates the problem to a problem in positive characteristic, via the Riemann-Hilbert correspondence and the equivalence \({\mathfrak O}_ F'\simeq {\mathfrak O}'_{F,q}\).