Noncommutative resolution, F-blowups and \(D\)-modules (Q2389247): Difference between revisions

The starting point for the article is the isomorphism between the \(G\)-Hilbert scheme and the F-blowup. The \(G\)-Hilbert scheme is associated to a smooth \(G\)-variety \(M\) with \(G\) a finite group, while the \(e\)th F-blowup is associated to the \(e\) times iteration of the Frobenius morphism \(F:X\rightarrow X\) of a singular variety in positive characteristic. Both are the moduli spaces of certain \(0\)-dimensional subschemes. The authors work over an algebraically closed field of characteristic \(p>0.\) Under some conditions, the isomorphism connects the \(G\)-Hilbert scheme of a \(G\)-variety \(M\) and the \(e\)th F-blowup of the quotient variety \(X=M/G:\) \[ \text{Hilb}^G(M)\cong\text{FB}_e(X). \] In this article the authors try to understand the mechanism behind this phenomenon. The authors mention the known correspondence \[ {\mathbf D}({\mathbf{Coh}}(\text{Hilb}^G(M)))\cong{\mathbf D}({\mathbf{Coh}}^G(M)) \] where \({\mathbf{Coh}}(-)\) (\({\mathbf{Coh}}^G(-)\)) denotes the category of coherent (\(G\)-)sheaves and \({\mathbf D}\) denotes the bounded derived category. Then they ask: Does a similar result hold in the situation of \(\text{FB}_e(-)\)? The authors address these questions in terms of the noncommutative resolution due to Van den Bergh. Let \(M=\mathbb A^d_k\) and \(G\subset\text{SL}_d(k).\) Let \(S\) and \(R\) be the coordinate rings of \(M\) and \(X=M/G.\) Then the endomorphism ring \(A:=\text{End}_R(S)\) is a noncommutative crepant resolution which means that \(A\) is regular in the sense that it has finite global dimension and satisfies the condition corresponding to the crepancy. The \(G\)-Hilbert scheme is identified with some moduli space \(W\) of \(A\)-modules and a coherent \(G\)-sheaf on \(M\) is identified with an \(A\)-module. Thus the first question can be reduced to \[ {\mathbf D}({\mathbf{Coh}}(W))\cong{\mathbf D}(A-{\mathbf{mod}}) \] and the Galois group \(G\) disappears from view. For an affine scheme \(X=\text{Spec}R\) over \(k\), for \(q=p^e,\) the \(e\)th Frobenius morphism of \(X\) is identified with the morphism \(X\rightarrow X_e:=\text{Spec}R^q\) defined by the inclusion map \(R^q\hookrightarrow R.\) The authors suppose that this is a finite morphism, and put \(D_{R,e}:=\text{End}_{R^q}(R).\) The answers to the above questions is then claimed to be the following result: Let \(M=\mathbb A^d_k=\text{Spec} S,\) \(G\subset GL_d(k)\) a small finite subgroup and \(X:=M/G=\text{Spec} R.\) Then for sufficiently large \(e\), there is an equivalence of abelian categories \[ \text{End}_R(S)-{\mathbf{mod}}\cong D_{R,e}-{\mathbf{mod}}. \] Hence \(D_{R,e}\) has global dimension \(d\). Moreover, the moduli spaces \(\text{Hilb}^G(M)\) and \(\text{FB}_e(X)\) correspond to each other via this equivalence and are isomorphic to each other. The author also investigates when \(D_{R,e}\) is a crepant noncommutative resolution, and answers this for different classes of singularities. The article is very compact, containing a lot of advanced theory in few pages and not very easy to read. However the philosophy, ideas, and proofs are nice and good and gives great value to the reader.