Reflection functors for quiver varieties and Weyl group actions. (Q1434232)

To a finite oriented graph \(\Gamma\) without edge loops (a quiver), two sets \(\mathbf v\) and \(\mathbf w\) of positive integers indexing the vertices of the quiver and to an element \(\zeta\in\mathbb{R}^3\otimes\mathbb{R}^{{\text{Vertices}}(\Gamma)}\) is associated an hyper-Kähler manifold \({\mathfrak M}_\zeta(\mathbf{v,w})\), called a quiver variety [\textit{H.~Nakajima}, Duke Math. J. 76, No. 2, 365--416 (1994; Zbl 0826.17026)]. The construction of quiver varieties can be seen as an abstract and generalized version of the ADHM description of the moduli spaces of anti-self-dual connections on ALE spaces [\textit{P.~B.~Kronheimer} and \textit{H.~Nakajima}, Math. Ann. 288, No. 2, 263--307 (1990; Zbl 0694.53025)]. So it is natural to wonder whether geometrical phenomena typical of moduli spaces of instantons have a counterpart in the general setting of quiver varieties. In particular, there is a Weyl group action on ALE spaces, which, by pull-back on connections, induces a natural action of the Weyl group on the moduli spaces of anti-self-dual connections. It turns out that the maps describing the action of elements in the Weyl group have a purely quiver theoretic description, and hence make sense for arbitrary quiver varieties. In the general setting of quiver varieties, the maps corresponding to elements in the Weyl group are called reflection functors, since they are close relatives of the Bernstein-Gel'fand-Ponomarev reflection functors [\textit{I.~N.~Bernstein}, \textit{I.~M.~Gel'fand} and \textit{V.~A.~Ponomarev}, Usp. Mat. Nauk 28, No. 2(170), 19--33 (1973; Zbl 0269.08001)]. In this paper the author provides a detailed study of the reflection functors for an arbitrary quiver manifold, from the point of view of hyper-Kähler geometry. Most of the results of the paper have been mentioned in [\textit{H.~Nakajima}, loc. cit.], but a proof of the Weyl group relation has not been given there; moreover a definition of reflection functors for simple reflections has been given in the not so widely known [\textit{H.~Nakajima}, in Proc. Symp. Representation Theory, Yamagata (1992)]. So, during the last ten years various authors have defined Weyl group actions on quiver varieties and reflection functors, at least for simple reflections, [\textit{W.~Crawley-Boevey} and \textit{M.~P.~Holland}, Duke Math. J. 92, No. 3, 605--635 (1998; Zbl 0974.16007), \textit{G.~Lusztig}, Ann. Inst. Fourier 50, No. 2, 461--489 (2000; Zbl 0958.20036) and \textit{A.~Maffei}, Ann. Sc. Norm. Super. Pisa, Cl. Sci. (5) 1, No. 3, 649--686 (2002; Zbl 1143.14309)]. It should be remarked that the aforementioned authors have studied quiver varieties and Weyl group actions on them from a complex (or algebraic) geometric perspective, whereas Nakajima's point of view in this paper is that of hyper-Kähler geometry. The author proves that the reflection functors are hyper-Kähler isometries, that his definition is identical with the definitions given by Crawley-Boevey-Holland, Lusztig and Maffei and with the definition previously given by the author himself [in loc. cit.], and that the reflection functor for the longest element of the Weyl group is identified with Lusztig's new symmetry for quiver varieties [\textit{G.~Lusztig}, Duke Math. J. 105, No. 2, 239--265 (2000; Zbl 1017.20040)] when the quiver is a graph of type ADE. Note that Lusztig's new symmetry is defined only on Lagrangian submanifolds of the quiver variety, while the corresponding Nakajima symmetry is defined on the whole variety and has a very simple description in terms of the ADHM construction: It maps an anti-self dual connection to its dual connection. The paper is extremely well written, with introductory sections on hyper-Kähler structures and quiver varieties which make it accessible to a wide mathematical audience. In section 3, Nakajima defines reflection functors for simple reflections and identifies them with the reflection functors defined by Lusztig and Maffei [loc. cit.] which are known to satisfy the Weyl group relations; so an action of the Weyl group is defined. In section 4, reflection functors for general elements of the Weyl group are defined. It is natural to wonder whether the composition of these general reflection functors corresponds to the multiplication in the Weyl group (Conjecture 4.15(3) in the paper). This would provide a direct proof that reflection functors define a Weyl group representation, whereas proofs currently available rely on the checking of the Weyl group relations. This conjecture is verified in section 7 for quivers which are graphs of Dynkin type.