Deformations of rational surface singularities and reflexive modules with an application to flops (Q1621481): Difference between revisions

This article considers the relation between deformations of a rational surface singularity with a reflexive module, and deformations of a partial resolution of the singularity with the locally free strict transform of the module. The results indicate how a family of small resolutions of a 3-dimensional index one terminal singularity and its flop are obtained by blowing up in a maximal Cohen-Macaulay module and its syzygy. The article is motivated by the work on the geometrical McKay correspondence which can be said to give a one-to-one correspondence between the isomorphism classes of indecomposable reflexive modules \(\{M_i\}\) and the prime components \(\{E_j\}\) of the exceptional divisor in the minimal resolution \(\tilde X\rightarrow X\) of a rational double point (RDP), that is \(A_n\), \(D_n\), \(E_{6 - 8}.\) For a natural class of special reflexive modules named Wunram modules after its inventor, the correspondence holds for any rational surface singularity. \textit{M. Van den Bergh} [Duke Math. J. 122, No. 3, 423--455 (2004; Zbl 1074.14013)] used the endomorphism ring of a higher dimensional Wunram module to prove derived equivalences for flops, and this again led to attention to the \(2\)-dimensional case with interesting results by \textit{O. Iyama} and \textit{M. Wemyss} [Math. Z. 265, No. 1, 41--83 (2010; Zbl 1192.13012); Ill. J. Math. 55, No. 1, 325--341 (2011; Zbl 1258.13015)] and \textit{M. Wemyss} [Math. Ann. 350, No. 3, 631--659 (2011; Zbl 1233.14012)]. In this article, the authors prove that blowing up a rational surface singularity \(X\) in a reflexive module \(M\) gives a partial resolution \(f:Y\rightarrow X\) where \(Y\) is normal, dominated by the minimal resolution , and where the strict transform \(\mathcal M=f^\Delta(M)\) is locally free. This partial resolution is determined by the first Chern class \(c_1(\mathcal F)\) of the strict transform \(\mathcal F\) of \(M\) to \(\tilde X.\) Thus, in particular, any partial resolution dominated by the minimal resolution is given by blowing up in a Wunram module, and the authors mention the RDP-resolution obtained by contracting the \((-2)\)-curves in the minimal resolution in particular; this is given by blowing up in the canonical module \(\omega_X.\) Consider the category of deformations \(\text{Def}_{Y,\mathcal M}\) of the pair \((Y,\mathcal M)\) blowing down to \((X,M)\). The main result in the present article says that the blowing down map \(\alpha:\text{Def}_{Y,\mathcal M}\rightarrow\text{Def}_{X,M}\) is injective and that it commutes with the forgetful map \(\beta:\text{Def}_{Y,\mathcal M}\rightarrow\text{Def}_{Y}\) and the blowing down map \(\delta:\text{Def}_Y\rightarrow\text{Def}_X.\) Furthermore, the forgetful map \(\beta\) is smooth and an isomorphism in many situations. The blowing down map \(\delta\) is a Galois covering onto the Artin component \(A\) on spaces, which for RDPs equals \(\text{Def}_X\). In general, it not injective, making the injectivity of \(\alpha\) surprising. The authors prove that \(\beta\) is an isomorphism if \(M\) is Wunram, implying that \(\delta\) factors through a closed embedding \(\alpha\beta^{-1}:\text{Def}_Y\subseteq\text{Def}_{(X,M)}\) realizing deformations of the pair as conjectured by \textit{C. Curto} and \textit{D. R. Morrison} [J. Algebr. Geom. 22, No. 4, 599--627 (2013; Zbl 1360.14053)] in the RDP case. A deformation of the pair \((X,M)\) in the geometric image of \(\text{Def}_{(Y,\mathcal M)}\) lifts to a deformation of \((Y,\mathcal M)\) without any base change. In general, \(\text{Def}_{X,M}\) is not dominated by \(\text{Def}_{(Y,\mathcal M)}\) even for RDPs. A main ingredient in Wahl's proof that the covering \(\text{Def}_{\tilde X}\rightarrow A\) has Galois action by a product of Weyl groups is the injectivity of \(\delta\) in the case that \(Y\) is the RDP resolution. This follows directly from the authors result because \(\text{Def}_{X,\omega_X}\cong\text{Def}_{X}\). The results indicate that there are interesting relations to \(\text{Def}_{X}\), for instance the component structure. The main application of the result above is a generalization of three conjectures of Curto and Morrison [loc. cit.] concerning the nature of small partial resolutions of \(3\)-dimensional index one terminal singularities and their flops. When \(g:W\rightarrow Z\) is a small partial resolution and \(X\subseteq Z\) is a sufficiently generic hyperplane section with strict transform \(f:Y\rightarrow X\), a result of Reid states that \(f\) is a partial resolution of an RDP. Thus \(g\) is a \(1\)-parameter deformation of \(f\) and so an element in \(\text{Def}_Y\). The authors prove that \(Y\) then is the blowing up of \(X\) in a reflexive module \(M\). As a consequence, \(\alpha\beta^{-1}\) takes this \(g\) to a \(1\)-parameter deformation of \((Z,N)\) of the pair \((X,M).\) Verbatim: Corollary. There is a maximal Cohen-Macaulay \(\mathcal O_Z\)-module \(N\) such that (i) The small partial resoltuion \(W\rightarrow Z\) is given by blowing up \(Z\) in \(N.\) (ii) Blowing up \(Z\) in the syzygy module \(N^+\) of \(N\) gives the unique flop \(W^+\rightarrow Z\). (iii) The length of the flop equals the rank of \(N\) if the flop is simple. A version of this statement is given for flat families of small partial resolutions and flops. It is proved that there is a family of pairs \((\mathbf{X},\mathbf{M})\) in \(\text{Def}_{(X,M)}\) such that the blowing up of \(\mathbf{X}\) in \(\mathbf{M}\) and in the syzygy \(\mathbf{X}^+\) give two simultaneous partial resolutions \(\mathbf{Y}\rightarrow\mathbf{X}\leftarrow\mathbf{Y}^+\) inducing any local family of flops \(g\) by pullback, for any \(g\) with hyperplane section \(f\). In fact, \(\mathbf{X}\) equals \(A_1, D_4, E_6, E_7, E_8\) for \(l=1,2,3,4,5,6\) respectively, so that the result above proves that there is in each case a unique reflexive module \(M\) of rank \(l\) such that any simple flop of length \(l\) is obtained by pullback from the \(\mathbf{Y}\rightarrow\mathbf{X}\leftarrow\mathbf{Y}^+.\) This gives the universal simple flop of length \(l\) realized as blowing-ups in families of reflexive modules (as suggested by Curto and Morrison [loc. cit.]). The RDPs are hypersurfaces, and any maximal Cohen-Macaulay module is given by a matrix factorization. The conjectures of Curto and Morrison [loc. cit.] are stated by matrix factorizations, and they are verified for \(A_n\) and \(D_n\) by brute force computations. The present argument is conceptual, coordinate-free, and makes the conjectures transparent. The singularities considered in this work will all be henselisations of finite algebras and the results will therefore have finite type representations locally in the étale topology. Work by Donovan, Wemyss et.al. links properties of various noncommutative algebras to flops. This involves quiver algebras, mutations, tilting theory and GIT-constructions with endomorphism algebras as input. This work offers a direct proof of the original Curto-Morrison conjectures using deformation theory where the blowing up ideal for the small, partial resolution is obtained directly from the \(2\)-dimensional Wunram module. Any flop with fixed RDP hyperplane section and Dynkin diagram is a pullback from a pair of universal blowing ups. The article is impressing. The deformation functors are studied conceptually, more as fibred categories than as functors, and natural transformations are transformed into maps between the categories of resolutions of singularities. The article gives a lot of techniques that that can be used in the study of more general contractions, and shows a brilliant use of deformation theory in general. Also, the article is self contained with respect to the deformation theory, and contains all preliminaries needed for understanding the importance of the results.