Visible actions on flag varieties of exceptional groups and a generalization of the Cartan decomposition (Q397840)

Let \( G \) be a compact semisimple group and \( \sigma \) be a Chevalley-Weyl involution (this terminology seems not so popular; it is an involution which acts as \( (-1) \) on a Cartan subalgebra of \( \mathfrak{g} = \mathrm{Lie}(G) \)). The author classifies a pair \( (L, H) \) of Levi subgroups for which \( G = L G^{\sigma} H \) holds when \( G \) is of exceptional type. The author has already classified such pairs for classical groups of types B, C and D [\textit{Y. Tanaka}, Bull. Aust. Math. Soc. 88, No. 1, 81--97 (2013; Zbl 1347.22012)]; [J. Math. Soc. Japan 65, No. 3, 931--965 (2013; Zbl 1296.22015)]; [Tohoku Math. J. (2) 65, No. 2, 281--295 (2013; Zbl 1277.22014)]. For type A, Kobayashi classified such pairs [\textit{T. Kobayashi}, Transform. Groups 12, No. 4, 671--694 (2007; Zbl 1147.53041)], and this completes the classification. The author, as well as Kobayashi, is interested in such pairs because of the investigation of visible actions. If \( L, H \) are both symmetric subgroups (i.e., its connected component is fixed by a certain involution), such classification goes down to a result of Matsuki, which generalizes the classical Cartan decomposition [\textit{T. Matsuki}, J. Algebra 197, No. 1, 49--91, Art. No. JA977123 (1997; Zbl 0887.22009)]. In the exceptional case, there are a few cases which do not fall into Matsuki's setting. According to the classification given in Theorem 1.1, there are no such pairs except for the types \( E_6 \) and \( E_7 \) among the simple exceptional groups. When \( G \) is of type \( E_6 \) or \( E_7 \), one of the Levi subgroups \( L = G^{\tau} \) is always a symmetric subgroup (up to a switch of the position) and \( H \) is a subgroup of the symmetric subgroup \( G^{\tau'} \), where \( \tau \) and \( \tau' \) are involutions. Notably, the semisimple part of the subgroup \( H \) seems always to be of maximal rank in the semisimple part of \( G^{\tau'} \). The proof uses Matsuki's result first for the pair \( (G^{\tau}, G^{\tau'}) \), and then turns to decompose \( G^{\tau'} \) further. To prove the exhaustion of the classification, the author uses representation theory; namely the classification of multiplicity free decompositions of tensor product representations by \textit{P. Littelmann} [J. Algebra 166, No. 1, 142--157 (1994; Zbl 0823.20040)] and \textit{J. R. Stembridge} [Represent. Theory 7, 404--439 (2003; Zbl 1060.17001)] is a key.