Non-commutative crepant resolution of minimal nilpotent orbit closures of type A and Mukai flops (Q2401693)

A non-commutative crepant resolution (NCCR) of a Gorenstein algebra \(R\) is an endomorphism ring \(\text{End}_R(M)\) of a maximal Cohen-Macaulay \(R\)-module \(M\) such that \(\text{End}_R(M)\) is a maximal Cohen-Macaulay \(R\)-module of finite global dimension. The three main problems are the following {\parindent=6mm \begin{itemize}\item[(1)] Construct a NCCR of \(R\) and characterize the tilting object \(M\). \item[(2)] Construct a derived equivalence between the NCCR and a commutative crepant resolution. \item[(3)] Construct a (commutative) crepant resolution as a moduli space of modules over the NCCR. \end{itemize}} Let \(V=\mathbb C^N\) be a complex vector space of dimension \(N\geq 2\). Consider the subset \(B(1)\) of \(\text{End}_{\mathbb C}(V)\) given by \(B(1)=\{X\in\text{End}_{\mathbb C}(V)|X^2=0,\;\dim\text{Ker}X=N-1\}\) which is a nilpotent orbit aof type \(A\). Then \(\overline{B(1)}\) is normal, and has only symplectic singularities. Thus its coordinate ring \(R\) is normal and Gorenstein. Also, since \(\text{codim}_{\overline{B(1)}}(\partial B(1))\geq 2\), \(R\simeq H^0(B(1),\mathcal O_{B(1)})\) as \(\mathbb C\)-algebras. The main objective in this article is to study NCCRs of a minimal nilpotent orbit closure \(\overline{B(1)}\) of type \(A\). Let \(H\) be a subgroup of \(\text{SL}_N\) such that \(\text{SL}_N/H\simeq B(1)\). Then \(H\) is isomorphic to a particular subgroup of \(\text{SL}_N\) on which it can be defined a character \(H\ni A\mapsto c^{-a}\). Let \(\mathcal M_a\) be the homogeneous line bundle on \(B(1)\) corresponding to this character, and set \(M_a=H^0(B(1),\mathcal M_a).\) It is proved in this article that a direct sum of \(R\)-modules \((M_a)_a\) gives a NCCR of \(R\). More precisely, each \(M_a\) is a Cohen-Macaulay \(R\)-module for \(N+1\leq a\leq N-1\), and for \(0\leq k\leq N-1\), the \(R\)-module \(\bigoplus_{a=-N+k+1}^kM_a\) gives a NCCR \(\text{End}_R(\bigoplus_{a=-N+k+1}^kM_a)\) of \(R\). This result is proved by \textit{tilting bundles} on the known (commutative) crepant resolutions \(Y\) and \(Y^+\) of \(\overline{B(1)}\) that are the total spaces of the cotangent bundles on \(\mathbb P(V)\) and \(\mathbb P(V^\ast)\) respectively. Consider the projections \(\pi: Y\rightarrow\mathbb P(V)\) and \(\pi^\prime: Y\rightarrow\mathbb P(V^\ast)\). It is proved that the bundles \(\mathcal T_k=\bigoplus_{a=_n+k+1}^k\pi^\ast\mathcal O_{\mathbb P(V)}(a)\) and \(\mathcal T_k^+=\bigoplus_{a=_n+k+1}^k{\pi^\prime}^\ast\mathcal O_{\mathbb P(V^\ast)}(a)\) are tilting bundles on \(Y,\;T^+\). Also there is a canonical isomorphism of \(R\)-algebras \(\Lambda_k=\text{End}_Y(\mathcal T_k)\simeq\text{End}_{Y^+}(\mathcal T^+_{n-k-1})\simeq \text{End}_R(\bigoplus_{a=-N+k+1}^kM_a).\) By the theory of tilting bundles, there is an equivalence of categories \(D^b(Y)\simeq D^b(\mathsf{mod}(\Lambda_k))\). The author provides another NCCR \(\Lambda^\prime\) of \(R\) that is not isomorphic to \(\Lambda_k\) but which is derived equivalent to \(\Lambda_k\). This is interesting in itself (by proving that reconstruction of algebras are not possible). The article describes a NCCR \(\Lambda_k\) as the path algebra of the \textit{double Beilinso quiver} with explicitly given relations. Similar results for non-commutative resolutions of determinantal varieties are obtained by Buchweitz, Leuschke, and Van den Bergh, and also by Weyman and Zhao. Let \(S=\text{Sym}^\bullet(V\otimes_{\mathbb C} V^\ast)\), let \(v_1,\dots,v_N\) be the standard basis of \(V=\mathbb C^N\), and let \(f_1,\dots,f_N\) be the dual basis of \(V^\ast\). Regarding \(x_{ij}=v_j\otimes f_i\in S\) as the variables of the affine coordinate ring of the affine variety \(\text{End}_{\mathbb C}(V)\simeq V^\ast\otimes_{\mathbb C}V\), it is clear that because \(\overline{B(1)}\) is a closed subvariety of \(\text{End}_{\mathbb C}(V)\), \(R\) is a quotient of \(S\). The author proves that as \(S\)-algebra, the non-commutative algebra \(\Lambda_k\) is isomorphic to the path algebra \(S\widetilde{\Gamma}\) of the double Beilinson quiver \(\widetilde{\Gamma}\) with \(N\) vertices, and with relations \(v_iv_j=v_jv_i,\;f_if_j=f_jf_i,\;v_jf_i=f_iv_j=x_{ij}\) for all \(1\leq i,j\leq N\), \(\sum_{i=1}^Nf_iv_i=0=\sum_{i=1}^Nv_if_i.\) By using this result, the crepant resolutions \(Y\) and \(Y^+\) are recovered from the quiver \(\widetilde{\Gamma}\) as moduli spaces of representations. The article contains a characterization of the simple representations of the quiver. It is proved that a simple representation corresponds to a point of a crepant resolution that lies over a non-singular point of \(\overline{B(1)}\). The relations between a crepant resolution \(Y (Y^+)\) and a NCCR \(\Lambda_k\) is seen as a generalization of the \textit{McKay correspondence}. Classically, this states that for a finite group \(G\subset\text{SL}_2\) there are relations between the geometry of a quotient variety \(\mathbb C^2/G\) and the representations of the group \(G\). The generalization is understood as general relationships between the crepant resolution \(\widetilde{\mathbb C^2/G}\) of \(\mathbb C^2/G\) and a quotient stack \([\mathbb C^2/G].\) \(\widetilde{\mathbb C^2/G}\) is called a geometric resolution of \(\mathbb C^2/G\), and because a coherent sheaf on a quotient stack \([\mathbb C^2/G]\) is canonically identified with a module over the skew group algebra \(\mathbb C[x,y]\sharp G\), a smooth stack \([\mathbb C^2/G]\) is said to be an algebraic resolution of \(\mathbb C^2/G\). This leads to interpret the McKay correspondence as a correspondence between geometric and algebraic resolutions: A geometric resolution of \(\overline{B(1)}\) is \(Y\) or \(Y^+\), and an algebraic resolution is the NCCR \(\Lambda_k\). The diagram of two crepant resolutions \(Y\mathop{\longrightarrow}\limits^{\phi}\overline{B(1)}\mathop{\longleftarrow}\limits^{\phi^+} Y^+\) is a local model of a class of flops called \textit{Mukai flops}. By blowing up, Kawamata and Namikawa defines functors \(\text{KN}_k:D^b(Y)\rightarrow D^b(Y^+)\) and \(\text{KN}_k^\prime:D^b(Y^+)\rightarrow D^b(Y)\) which by composing with \(\psi_k\) and \(\psi_{N-k-1}\) gives an equivalence \(D^b(Y)\rightarrow D^b(Y^+)\). It is a main result that this composition coincides with the Kawamata-Namikawa functors. Finally, the author constructs \textit{multi-mutation} functors as an analogue to the Iyama-Wemyss's mutation functor. It is proved that a composition of two multi-mutation functors correspond to a P-twist \(P_k\in\text{Auteq}(D^b(\text{mod}(\Lambda_{N+k})))\). In the case of three-dimentional flops, Donovan and Wemyss proved that a compostion of two IW functors corresponds to a spherical twist. The results in this article say that in the case of Mukai flops, a composition of many IW mutations corresponds to a P-twist. The article is sufficiently detailed and gives a very good introduction to non-commutative algebraic geometry. Furthermore, the results are very nice and generalizes known results in the derived setting.