A proof of the refined PRV conjecture via the cyclic convolution variety (Q2230604)

This paper gives a new proof of the PRV (Parthasarathy-Ranga Rao-Varadarajan) conjecture and its refined version using the technique of geometric Satake correspondence. The PRV conjecture asserts that the representation \(V(\nu)\) of a reductive group \(G\) must appear in the tensor product \(V(\lambda)\otimes V(\mu)\) of represetnations of \(G\), where \(\nu\) is the dominant translate of \(\lambda + u\mu\) for any element \(u\) in the Weyl group \(W\). The refined version gives a lower bound of the multiplicity. The PRV conjecture was originally proved by \textit{S. Kumar} [Invent. Math. 93, No. 1, 117--130 (1988; Zbl 0668.17008)] and \textit{O. Mathieu} [Compos. Math. 69, No. 1, 37--60 (1989; Zbl 0678.17012)] independently, and the refined version was proved by \textit{S. Kumar} [Invent. Math. 97, No. 2, 305--311 (1989; Zbl 0691.17004)]. Later on, several new proofs were discovered, including the method of Littelmann path model. The geometric Satake correspondence relates representation theory of \(G\) to the geometry of affine Grassmanians of the Langlands dual group \(G^\vee\) of \(G\). Very often, the representation theoretic questions can be translated to geometric questions. A new proof of the PRV conjecture (not including the refined version) has been obtained earlier by \textit{T. Richarz} [Doc. Math. 19, 209--246 (2014; Zbl 1326.14121)] using geometric Satake. Nevertheless, the approach presented in this paper is different from Richarz's method. The novelty is that the author explictly constructs certain compontents of the cyclic convolution variety, whose cardinality exactly gives the lowe bound of the refined PRV. These components are realized as the closures of \(G(\mathcal{O})\)-orbits, and one just needs to argue that the orbits have the expected dimension.