The index of representations associated with stabilisers (Q852675): Difference between revisions

The authors study the index of finite dimensional representations. If \(Q\) is an algebraic group and \(\mathfrak{q}\) is its Lie algebra then the index \(\text{ind}(\mathfrak{q},V)\) of a finite dimensional \(Q\)-module \(V\) is by definition the minimum of the codimensions of the \(Q\)-orbit in the dual of \(V\). This notion is due to Rais. By a result of Vinberg, there is an upper bound on this index, namely ind\((\mathfrak{q},V^*)\) is at most \(\text{ind}(\mathfrak{q}_v,(V/\mathfrak{q}\cdot v)^*)\) (for all \(v\in V\)), where \(\mathfrak{q}_v\) is the stationary subalgebra of \(v\). If \(Q\) is semi-simple and \(V\) the adjoint representation it suffices to check nilpotent \(v\in V\). Here, equality was conjectured by \textit{A. G. Èlashvili} [Funct. Anal. Appl. 6, 44--53 (1972); translation from Funkts. Anal. Prilozh. 6, No. 1, 51--62 (1972; Zbl 0252.22015)]. This was proved recently by work of \textit{J.-Y. Charbonnel} [Bull. Soc. Math. Fr. 132, No. 4, 477--508 (2004; Zbl 1092.14055)] and by work of \textit{O. S. Yakimova} (classical cases) [The centralisers of nilpotent elements in classical Lie algebras, Funct. Anal. Appl., in press, preprint math.RT/0407065]. The article studies conditions that imply equality. To do that, the authors pursue two goals. First the authors find properties of \(v\) such that equality (*): \(\text{ind}(\mathfrak{q},V^*)= \text{ind}(\mathfrak{q}_v,(V/\mathfrak{q}\cdot v)^*)\) holds and second they find representations \(V\) such that equality (*) holds for all \(v\in V\). The authors say that the \(Q\)-module \(V\) (short: \((Q:V)\)) has GIB (good index behaviour) if (*) holds for all \(v\in V\). And \((Q:V)\) has GNIB (good nilpotency index behaviour) if (*) holds for all nilpotent \(v\in V\). Let now \(G\) be a reductive group and \(V\) a \(G\)-module. Then the authors show that \((G:V)\) has GIB if and only if GNIB holds for any slice representation of \((G:V)\). Furthermore, if \((G:V)\) has only finitely many orbits, then GNIB implies GIB (this applies in particular in the case where \(V\) is the adjoint representation of a simple Lie algebra). The authors consider in particular isotropy representations of symmetric pairs. These have finitely many orbits, so we are in the situation where it is enough to check GNIB. Let \((G,G_0)\) be a symmetric pair with induced \(\mathbb{Z}_2\)-grading \(\mathfrak{g}=\mathfrak{g}_0\oplus\mathfrak{g}_1\), and let \((G_0,\mathfrak{g}_1)\) the isotropy representation. The authors then study these isotropy representations. They show in sections 3--5 that for the symmetric pairs \((\text{SL}_n, \text{SO}_n)\), \((\text{SL}_{2n},\text{Sp}_{2n})\), \((\text{Sp}_{2n}, \text{GL}_n)\) and \((\text{SO}_{2n}, \text{GL}_n)\), the isotropy representations all have GNIB. Furthermore, they show that most of the remaining isotropy representations do not have GNIB. As a result, if the Lie algebra of \(G\) is simple, classical, they are able to completely describe which isotropy representations have GNIB (and thus GIB). The result for \(\mathfrak{sl}_n\) is given in Theorem 0.1 at the end of the introduction.