On the Craw-Ishii conjecture (Q1703579)

Let \(G\le \operatorname{SL}(n, \mathbb C)\) be a finite group. The McKay correspondence describes the phenomenon that many properties of crepant resolutions of the quotient variety \(\mathbb C^n/G\) are determined by the representation theory of \(G\). One explanation for this is that crepant resolutions are often moduli spaces of \(G\)-equivariant sheaves on \(\mathbb C^n\). Indeed, for \(n=2\), the crepant resolution is unique (it is the minimal resolution), and it agrees with the \(G\)-Hilbert scheme; see [\textit{Y. Ito} and \textit{I. Nakamura}, Lond. Math. Soc. Lect. Note Ser. 264, 151--233 (1999, Zbl 0954.14001)]. The \(G\)-Hilbert scheme is the moduli space of \(G\)-clusters, that is \(G\)-invariant subschemes \(Z\subset \mathbb C^n\) with \(H^0(\mathcal O_Z)\cong \mathbb C\langle G\rangle\), where \(\mathbb C\langle G\rangle\) denotes the regular representation of \(G\). In dimension 3, [\textit{A. Craw} and \textit{A. Ishii}, Duke Math. J. 124, No. 2, 259--307 (2004, Zbl 1082.14009)] proved that every crepant resolution of \(\mathbb C^3/G\), where \(G\le \operatorname{SL}(3, \mathbb C)\) is a finite abelian subgroup, is a moduli space of \(\theta\)-stable \(G\)-constellations for some stability parameter \(\theta\in \operatorname{Hom}_{\mathbb Z}(R(G), \mathbb Q)\) with \(\theta(\mathbb C\langle G\rangle)=0\). Here, a \(G\)-constellation is a \(G\)-equivariant coherent sheaf \(F\) on \(\mathbb C^3\) with \(H^0(F)\cong \mathbb C\langle G\rangle\), and it is called \(\theta\)-stable, if for every \(G\)-equivariant subsheaf \(E\subset F\), we have \(\theta(E) >0\). For a finite subgroup \(G\le \operatorname{GL}(3,\mathbb C)\) which is not contained in \(\operatorname{SL}(3,\mathbb C)\), the appropriate replacement for a crepant resolution of the quotient \(\mathbb C^3/G\) is that of a relative minimal model. The paper under review makes the following conjecture: For every finite subgroup \(G\le \operatorname{GL}(3,\mathbb C)\) and every relative minimal model \(Y\to \mathbb C^3/G\), there exists a stability parameter \(\theta \in \operatorname{Hom}_{\mathbb Z}(R(G), \mathbb Q)\) such that \(Y\) is isomorphic to an irreducible component of the moduli space of \(\theta\)-stable \(G\)-constellations. This conjecture is proven for a quite large class of cyclic subgroups, not contained in \(\operatorname{SL}(3,\mathbb C)\). The proofs heavily use toric geometry. An important ingredient is the notion of \(G\)-bricks. These are certain finite sets of Laurent monomials defining \(\mathbb C\)-bases of torus invariant \(G\)-constellations.