Computing resolutions of quotient singularities (Q729754)

The authors present algorithms to compute the Cox rings of certain quotient singularities and of some closely related objects. Precisely, consider a finite subgroup \(G \subset {\text{GL}}(n):={\text{GL}}(n,\mathbb C)\) not containing any pseudo-reflection and the quotient variety \(X_0={\mathbb C}^n/G\). Firstly, they give an algorithm presenting the Cox ring \({\mathcal R}(X_0)\) as a quotient \({\mathbb C}[T_1, \ldots, T_s]/I_0\), where a set of generators of the ideal \(I_0\) is described. Now, \({{\mathcal R} (X_0})\) is a multi-graded ring. They also give a matrix that determines this multi-grading. Next, they present an algorithm to construct a resolution \(X \to X_0\) of the singularities of \(X_0\) and to describe the multi-graded Cox ring \({\mathcal R} (X)\). More precisely, the idea is to construct an embedding of \(X_0\) into a toric variety \(Z_0\) and a modification \(Z \to Z_0\) which is a resolution of the singularities of \(Z_0\). Letting \(X\) denote the strict transform of \(X_0\) to \(Z\), they give a description of \({\mathcal R} (X)\) as a quotient of a polynomial ring by an ideal whose generators are specified, they also obtain a matrix defining the multi-grading. This variety \(X\) is a ``candidate'' to be a resolution of the singularities of \(X\). Indeed, if \(X\) is smooth, then it is a resolution of \(X_0\). The variety \(X\) might be singular, but they give an algorithm to decide whether \(X\) is smooth or not. In the construction, among other methods they use techniques from toric and tropical geometries. Although the algorithms are rather complicated, by using the Computer Algebra System \textbf{Singular} the authors are able to implement it in a number of concrete examples. For instance, they do it for subgroups \(G \subset {\text{GL}}(3)\) of order \(\leq 12\), and they study some interesting properties of the resulting resolutions (e.g., to decide when they are crepant). In the final section they implement the algorithm in two examples involving \( {\text{GL}}(4)\), although it seems that in this case (\(n=4\)) a more general implementation requires too many computations for the presently available machines.