The exotic Robinson-Schensted correspondence (Q1950620): Difference between revisions

Let \( G \) be a complex reductive group and \( \theta \in \mathrm{Aut}(G) \) an involutive automorphism. Then the pair \( (G, K) \), where \( K = G^{\theta} \) is a subgroup of fixed points of \( \theta \), is called a symmetric pair and \( G/K \) is a symmetric space. In a series of works [Invent. Math. 75, 205--272 (1984; Zbl 0547.20032); Adv. Math. 56, 193--237 (1985; Zbl 0586.20018); ibid. 57, 226--265 (1985; Zbl 0586.20019); ibid. 266--315 (1985; Zbl 0594.20031); Adv. Math. 59, 1--63 (1986; Zbl 0602.20035); Adv. Math. 61, 103--155 (1986; Zbl 0602.20036)], \textit{G. Lusztig} began to study the theory of character sheaves on \( G \), but not over the complex numbers, but over finite fields. In a very rough and naive sense, character sheaves are perverse sheaves on \( G \) (or on some geometric object related to \( G \)) which is parametrized by conjugacy classes of \( G \), and their characteristic polynomials at \( x \in G \) gives the values of irreducible characters (this is first observed by Green for type A group). Most interesting part of this theory is about the character sheaves on the unipotent variety, in which case, the characteristic polynomials mentioned above are a variant of Kostka polynomials, and there exists a rich combinatorial structure. \textit{V. Ginzburg} [Astérisque No. 173--174, 9--10, 199--255 (1989; Zbl 0706.22005)], then \textit{I. Grojnowski} [MIT thesis, 1992] extended the theory of character sheaves on \( G \) to symmetric spaces \( G/K \). Let \( V \) be a symplectic vector space of dimension \( 2 n \), and put \( G = \mathrm{GL}(V) \) and \( K =\mathrm{Sp}(V) \) which is a symmetric subgroup of \( G \). In this paper, the authors try to study theory of ``exotic'' character sheaves on symmetric spaces, i.e., those on a product \( V \times G/K \) of a vector space \( V \) and \( G/K \) on which \( K \) acts diagonally. According to Grojnowski's formulation, this is much related to structures of geometric cells for the diagonal action of \( K \) on \( V \times G/B \). By \textit{T. Matsuki} [J. Algebra 375, 148--187 (2013; Zbl 1332.22017)], it is known that there are finitely many \( K \)-orbits on \( V \times G/B \), and some reformulation of Matsuki's parametrization tells us that in fact \( K \)-orbits are parametrized by decorated fixed point free involutions in \( S_{2n} \), the Weyl group of \( G \). Here ``decorated'' involution means a pair \( (w, \alpha) \) of an involution \( w \) and a subset \( \alpha \subset \{ 1, 2, \dots, 2 n \} \) satisfying a certain condition which is essentially due to \textit{R. Travkin} [Sel. Math., New Ser. 14, No. 3--4, 727--758 (2009; Zbl 1230.20047)] in the case of mirabolic triple flag varieties. Let us denote by \( R_{2n}' \) the parameters of \( K \)-orbits on \( V \times G/B \) thus obtained. Let \( Y \subset T^{\ast} (V \times G/B) \) be a conormal variety, which is defined to be a union of conormal bundles over \( K \) orbits. Let \( \pi : Y \to V \times V^{\ast} \times \mathcal{N} \) be a product of projection \( T^{\ast} (V \times G/B) \to T^{\ast}(V) = V \times V^{\ast} \) to the first factor and a moment map \( T^{\ast} (G/B) \to \mathcal{N} \), where \( \mathcal{N} \) denotes the nilpotent cone in \( {\mathfrak{1}}{g} = \mathrm{Lie}(G) \). The image \( Z := \pi(Y) \subset V \times V^{\ast} \times \mathcal{N} \) together with the following two maps plays a central role in this paper. One of the maps is the projection \( p_1: Z \to V \times \mathcal{N} \) to the enhanced nilpotent cone \textit{P. N. Achar} and \textit{A. Henderson} [Adv. Math. 219, No. 1, 27--62 (2008; Zbl 1205.14061)] defined by \( p_1(v, u, x) = (v, x) \). There is another map \( p_2 : Z \to V \times \mathcal{N}_{{\mathfrak{1}}{g}^{-\theta}} \) defined by \( p_2(v, u, x) = (v, x + v \otimes u) \), and this is a well-defined map to the exotic nilpotent cone \textit{S. Kato} [Duke Math. J. 148, No. 2, 305--371 (2009; Zbl 1183.20002)]. Note that nilpotent \( G \)-orbits on the enhanced nilpotent cone \( V \times \mathcal{N} \) are parametrized by bipartitions of size \( 2 n \); thus we denote by \( \mathcal{O}_{\mu;\nu} \) the nilpotent orbit which corresponds to a bipartition \( (\mu; \nu) \). Similarly, nilpotent \( K \)-orbits on the exotic nilpotent cone \( V \times \mathcal{N}_{{\mathfrak{1}}{g}^{-\theta}} \) are parametrized by bipartitions of size \( n \); let us denote by \( \mathbb O_{\widetilde{\mu};\widetilde{\nu}} \) the nilpotent orbit corresponding to a bipartition \( (\widetilde{\mu}; \widetilde{\nu}) \) but with size \( n \). Interesting feature here is that \( p_2(p_1^{-1}(\mathcal{O}_{\mu; \nu})) \) is a single nilpotent orbit \( \mathbb O_{\widetilde{\mu};\widetilde{\nu}} \). The authors make many interesting observations on these nilpotent orbits (and their relations), such as closure relations, the structure of fibers of the maps \( p_1 \) and \( p_2 \), dimensions, and explicit combinatorial description of the correspondence \( (\mu; \nu) \to (\widetilde{\mu}; \widetilde{\nu}) \), for example. It is well known that the Springer fiber \( (G/B)_x \) of the moment map \( T^{\ast}(G/B) \to \mathcal{N} \) is naturally parametrized by standard Young tableaux \( T \) of certain shape determined by \( x \in \mathcal{N} \). Steinberg used this fact to get Robinson-Schensted correspondence in terms of the geometry of the double flag variety \( G/B \times G/B \) [\textit{R. Steinberg}, J. Algebra 113, No. 2, 523--528 (1988; Zbl 0653.20039)]. Now we can state a main theorem in the paper under review. Let us denote by \( Q_{2n}' \) the set of bipartitions of size \( 2 n \) for which \( \mathcal O_{\mu; \nu} \) is in \( p_1(Z) \). Then the pair \( ((\mu; \nu), T) \), where \( (\mu; \nu) \in Q_{2n}' \) and \( T \in \mathrm{SYT}(\mu + \nu) \) is a standard Young tableau, parametrizes irreducible components of the conormal variety \( Y \), which in tern parametrize \( K \)-orbits on \( V \times G/B \). Put \( T_{2n}' = \{ ((\mu; \nu), T) \mid (\mu; \nu) \in Q_{2n}', T \in \mathrm{SYT}(\mu + \nu) \} \). Thus there exists a natural bijection \( R_{2n}' \simeq T_{2n}' \), which is a generalization of Robinson-Schensted correspondence. The bijection is achieved in purely geometric language, and its combinatorial description is still open. The above ``main'' theorem is in fact interesting enough, but the content of this paper is much more rich. It will play a role as a fundamental and concrete example in the theory of exotic character sheaves which is developing rapidly now.