Quiver varieties of type \(A\). (Q1772465)

Let \(d=(d_1,d_2,\dots,d_n)\) and \(v=(v_1,v_2,\dots,v_n)\) be two vectors of non-negative integers. Associated to the pair \((d,v)\) (or equivalently to a collection of vector spaces having dimensions \(d_i\) and \(v_i\)), \textit{H. Nakajima} [Duke Math. J. 76, No. 2, 365-416 (1994; Zbl 0826.17026)] introduced certain quiver varieties \(M(d,v)\) and \(M_0(d,v)\) with a map \(\pi\colon M(d,v)\to M_0(d,v)\) (that is equivariant under the action of the product of a certain collection of general linear groups). Let \(M_1(d,v)=\pi(M(d,v))\). These varieties were used by Nakajima to study representations of Kac-Moody algebras. Nakajima conjectured (and proved in a special case) that the varieties \(M(d,v)\) and \(M_1(d,v)\) could be identified with certain varieties defined geometrically using nilpotent matrices, Slodowy slices, and flag varieties (in such a way as to preserve the aforementioned group action). Somewhat more precisely, let \(N=\sum_{i=i}^nid_i\) and consider the Lie algebra of matrices \(M_N(\mathbb{C})\). Using \(v\), a partition of \(N\) is defined, and then these geometric varieties are determined by the nilpotent orbit corresponding to that partition and an associated flag variety. The main result of this paper is a proof of Nakajima's conjecture. The proof is accomplished by explicitly constructing maps between the varieties and then showing that the maps are isomorphisms. The proof in part makes use of a modified path algebra to give a description of the coordinate ring of \(M_0(d,v)\).