Noncommutative resolutions and rational singularities (Q841560)

In this article, it is important that \(k\) is a fixed algebraically closed field of characteristic \(0\). The authors prove that the center of a homologically homogeneous, finitely generated \(k\)-algebra has rational singularities. Assume \(X=\text{Spec} R\) for an affine, Gorenstein \(k\)-algebra \(R\). In this article, a commutative resolution of singularities is a crepant homomorphism \(f:Y\rightarrow X,\) i.e. \(f^\ast\omega_Y=\omega_X.\) Bondal and Orlov conjectured that two such resolutions are derived equivalent, and this was later proved by Bridgeland. The authors generalize this to a \textit{third} noncommutative crepant resolution explaining Bridgeland's proof. This observation leads to different approaches to the Bondal-Orlov conjecture and related topics. The question now is how the existence of a noncommutative crepant resolution affects the original commutative singularity. It is known that if a Gorenstein singularity has a crepant resolution then it has rational singularities. The authors asks wether this is true for a noncommutative crepant resolution. The article answers this affirmatively. Let \(\Delta\) be a prime affine \(k\)-algebra that is finitely generated as a module over its center \(Z(\Delta).\) \(\Delta\) is called homologically homogeneous of dimension \(d\) if all simple \(\Delta\)-modules have the same projective dimension \(d.\) The properties of homologically homogeneous rings are close to commutative regular rings, and the idea is to use such a ring \(\Delta\) as a noncommutative analogue of a crepant resolution. Formally, a noncommutative crepant resolution of \(R\) is any homologically homogeneous ring of the form \(\Delta=\text{End}_R(M)\) where \(M\) is a reflexive and finitely generated \(R\)-module. The main result of the article is the following: Theorem. Let \(\Delta\) be a homologically homogeneous \(k\)-algebra. Then the center \(Z(\Delta)\) has rational singularities. In particular, if a normal affine \(k\)-domain \(R\) has a noncommutative crepant resolution, then it has rational singularities. Also, examples are given proving that this theorem may fail in positive characteristic. The article starts with the properties of homologically homogeneous rings, based on tame orders: If \(\Delta\) is a prime ring with simple Artinian ring of fractions \(A\) (i.e. \(\Delta\) is a prime order in \(A\)), \(\Delta\) is called a \textit{tame \(R\)-order} if it is a finitely generated and reflexive \(R\)-module such that \(\Delta_{\mathfrak p}\) is hereditary for all prime ideals \(\mathfrak p\) in \(R\) of height \(1\). A homologically homogeneous ring \(\Delta\) of dimension \(d\) is Cohen Macaulay (CM) over its center \(Z(\Delta)\), both GK\(\dim\Delta\) and the global homological dimension gl\(\dim\Delta\) of \(\Delta\) equal \(d\), the center \(Z=Z(\Delta)\) is an affine CM normal domain, and finally, \(\Delta\) is a tame \(Z\)-order. The rest of the article is then used to prove the main theorem. This involves reduction to the Calabi-Yau case for proving that \(Z\) has rational singularities by a generalization of the commutative method where one constructs a Gorenstein cover of a \(\mathbb Q\)-Gorenstein singularity. The article ends with examples proving, among other things, that the main theorem may fail in the case where \(k\) has positive characteristic. The article is precise, and illustrates noncommutative algebraic geometry in a concrete way. It also give useful criterions and ideas to be followed in other settings in noncommutative geometry.