On the existence of symplectic resolutions of symplectic reductions (Q971912)

The author studies the existence of symplectic resolutions of the symplectic reduction of a coregular representation of \(\mathrm{Sl}_{2}\). If \(G\) is any semisimple complex linear algebraic group and \(V\) is a finite dimensional complex vector space on which \(G\) acts, then \(V\oplus V^{*}\) has a natural symplectic form and \(G\) acts symplectically on it. The moment map \(\mu:(V\oplus V^{*})^{\oplus 2}\longrightarrow\mathfrak{g}^{*}\), where \(\mathfrak{g}=\mathrm{Lie}(G)\), is a \(G-\)equivariant map, hence \(\mu^{-1}(0)\subseteq (V\oplus V^{*})^{\oplus 2}\) is a \(G-\)invariant subset. The quotient \((V\oplus V^{*})///G:=\mu^{-1}(0)//G\) is the symplectic reduction of \(V\oplus V^{*}\). If \(\mathcal{I}_{\mu}\) is the ideal corresponding to \(\mu^{-1}(0)\), then \((V\oplus V^{*})///G\) is isomorphic as schemes to \(\mathrm{Spec}(\mathbb{C}[V\oplus V^{*}]/\mathcal{I}_{\mu})^{G}\). It is not necessarily a symplectic variety, but it is if \(\mu^{-1}(0)\) is normal, the stabilizer of every regular point of \(\mu^{-1}(0)\) is trivial, and \((\mu^{-1}(0)_{\mathrm{sing}}//G)\subseteq ((V\oplus V^{*})///G)_{\mathrm{sing}}\) (see Section 2). Suppose that \(G=\mathrm{Sl}_{2}\) and that the representation \(V\) is coregular, i. e. the quotient \(V//G\) is smooth. Coregular representations for simple, connected and simply connected groups are classified by \textit{G. W. Schwarz} [Invent. Math. 50, 1--12 (1978; Zbl 0391.20033)], and for \(\mathrm{Sl}_{2}\) there are only the following cases: (1) \(\mathbb{C}^{2}\), \(\mathbb{C}^{2}\oplus\mathbb{C}^{2}\) and \(\mathbb{C}^{2}\oplus\mathbb{C}^{2}\oplus\mathbb{C}^{2}\); (2) \(S^{3}\mathbb{C}^{2}\) and \(S^{4}\mathbb{C}^{2}\); (3) \(\mathfrak{sl}_{2}\) and \(\mathfrak{sl}^{\oplus 2}_{2}\); (4) \(\mathfrak{sl}_{2}\oplus\mathbb{C}^{2}\). The case (1) is treated in section 3: the author shows that \(\mathbb{C}^{2}\oplus\mathbb{C}^{2}///\mathrm{Sl}_{2}\) is non-reduced, hence it is not a symplectic variety, while \((\mathbb{C}^{2}\oplus\mathbb{C}^{2})^{\oplus 2}///\mathrm{Sl}_{2}\) is the union of two normal components, each admitting a symplectic resolution. Finally, \((\mathbb{C}^{2}\oplus\mathbb{C}^{2})^{\oplus 3}///\mathrm{Sl}_{2}\) is a symplectic variety which is isomorphic to the subscheme of \(\mathfrak{so}_{6}\) of matrices \(A\) such that \(A^{2}=0\) and \(\mathrm{Pf}(QA)=0\), where \(\mathrm{Pf}(QA)\) is the Pfaffian of the 15 skew-symmetric \(4\times 4-\)minors of \(A\). This corresponds to the closure of the nilpotent orbit \(\mathcal{O}_{[2^{2},1^{2}]}\), which is singular at the origin, and which admits two symplectic resolution: one by the cotangent bundle, and one by its dual. The method is to study the action of \(\mathrm{Sp}_{2n}\) on \((\mathbb{C}^{2n})^{\oplus 2m}\): the author gives a criterion for reducedness and normality of \(\mu^{-1}(0)\), and describes the quotient \(\mu^{-1}(0)//\mathrm{Sl}_{2}\) as the scheme \(Z\) of matrices \(A\in\mathfrak{so}_{2m}\) such that \(A^{2}=0\) and \(\mathrm{rk}(A)\leq \min\{2n,m\}\), with the reduced structure. Using the natural action of \(SO_{2m}\) on \(Z\), then \(Z\) is given by union of closures of nilpotent orbits. By \textit{B. Fu} [Invent. Math. 151, No. 1, 167--186 (2003; Zbl 1072.14058)], the author guarantees the existence of symplectic resolutions. The cases (2) and (3) are classical: \((S^{3}\mathbb{C}^{2}\oplus S^{2}\mathbb{C}^{2})///\mathrm{Sl}_{2}\simeq\mathbb{C}^{2}/(\mathbb{Z}/4)\) is a symplectic variety having an \(A_{3}-\)singularity at \(\{0\}\), which can be resolved symplectically by two successive blow-ups, while \((S^{4}\mathbb{C}^{2}\oplus S^{4}\mathbb{C}^{2})///\mathrm{Sl}_{2}\) is a symplectic variety isomorphic to a subset of \((\mathbb{C}^{2})^{\oplus 3}/S_{3}\), which can be resolved symplectically by the Hilbert scheme \(\mathrm{Hilb}^{3}(\mathbb{C}^{2})\). The case \(\mathfrak{sl}_{2}^{\oplus 2}///Sl_{2}\) gives a symplectic variety with an \(A_{1}-\)singularity at the origin, which can be resolved by a single blow-up. The case \(\mathfrak{sl}_{2}^{\oplus 4}///\mathrm{Sl}_{2}\) is a symplectic variety isomorphic to the subscheme of \(\mathfrak{sp}_{4}\) of matrices \(A\in\mathfrak{sp}_{4}\) such that \(A^{2}=0\). By \textit{M. Lehn} and \textit{C. Sorger} [J. Algebr. Geom. 15, No. 4, 753--770 (2006; Zbl 1156.14030)], the cotangent bundle gives a symplectic resolution. The case (5) is the last treated. First, the author computes the invariant ring \(\mathbb{C}[\mathfrak{sl}_{2}\oplus\mathbb{C}^{2}]^{\mathrm{Sl}_{2}}\), the generating invariants of the action of \(\mathrm{Sl}_{2}\) on \((\mathfrak{sl}_{2}\oplus\mathbb{C}^{2})^{\oplus 2}\) (Propositions 5.1 and 5.2) and the ideal \(\mathcal{I}_{\mu}\). Using \texttt{Singular}, in Proposition 5.3 it is shown that \(\mathbb{C}[\mu^{-1}(0)]^{\mathrm{Sl}_{2}}\simeq \mathbb{C}[z_{1},\dots,z_{8}]/(h_{1},\dots,h_{9})\), and a precise expression of the polynomials \(h_{1},\dots,h_{9}\) is given. Hence \(\mu^{-1}(0)//\mathrm{Sl}_{2}\) is a subvariety \(Z\) of \(\mathbb{C}^{8}\) whose ideal is \(V(h_{1},\dots,h_{9})\). Then the author shows that \(Z\) is a \(4-\)dimensional symplectic variety with a stratification \(\{0\}\subseteq Z_{\mathrm{sing}}\subseteq Z\), where \(Z_{\mathrm{sing}}\) is the singular locus of \(Z\) and \(\{0\}\) is the singular point of \(Z_{\mathrm{sing}}\). The blow-up \(\widetilde{Z}\) of \(Z\) along \(Z_{\mathrm{sing}}\) is a symplectic resolution of \(Z\). The interesting phenomenon pointed out by the author is that \(\pi^{-1}(0)\) is the union \(E_{1}\cup E_{2}\) of two projective planes intersecting at one point. One can then perform a Mukai flop in \(E_{1}\) or in \(E_{2}\): the author shows using \texttt{Singular} that the two obtained Mukai flops \(W_{1}\) and \(W_{2}\) are two symplectic resolutions of \(Z\) which are not isomorphic (and each one is not isomorphic to \(\widetilde{Z}\)), but which are equivalent in the sense of \textit{B. Fu} and \textit{Y. Namikawa} [Ann. Inst. Fourier 54, No. 1, 1--19 (2004; Zbl 1063.14018)].