Noncommutative enhancements of contractions (Q1710394)

There is a common idea that various aspects of birational geometry should be enhanced into the noncommutative setting. The examples considered in this article comes from associating noncommutative algebras to certain contractions: Algorithmic ways to relate minimal models based on cluster theory, new invariants for flips and flops linked to Gopakumar-Vafa invariants, the braiding of flop functors, noncommutative versions of curve counting, full conjectural analytic classification of 3-fold flops, and the new examples of 3-fold flops. \par These examples are either local in nature, or apply only to contraction of curves. This article intends to work in a more general setting: Take a contraction $f:X\rightarrow Y$ with mild assumptions, and enhance the non-isomorphism locus $Z\subset Y$ with a certain sheaf of noncommutative rings $\mathcal D$. This sheaf is used to give new results, in particular a class of autoequivalences. \par In this article, the axiomatic framework applies when $f$ is a crepant resolution of a Gorenstein $d$-fold $Y$, with mild restrictions which holds in many settings including those of \textit{M. Haiman} [J. Am. Math. Soc. 14, No. 4, 941--1006 (2001; Zbl 1009.14001)], \textit{R. V. Bezrukavnikov} and \textit{D. B. Kaledin} [in: Algebraic geometry. Methods, relations, and applications. Collected papers. Dedicated to the memory of Andrei Nikolaevich Tyurin. Moscow: Maik Nauka/Interperiodica. 13--33 (2004; Zbl 1137.14301); translation from Tr. Mat. Inst. Steklova 246, 20--42 (2004)], Procesi bundles [\textit{I. Losev}, Math. Ann. 359, No. 3--4, 729--744 (2014; Zbl 1304.14019)], Toda-Uehara [\textit{Y. Toda} and \textit{H. Uehara}, Adv. Math. 223, No. 1, 1--29 (2010; Zbl 1187.18009)], Springer resolutions of determinantal varieties [\textit{R.-O. Buchweitz} et al., Invent. Math. 182, No. 1, 47--115 (2010; Zbl 1204.14003)], and all known 3-fold crepant resolutions including derived McKay correspondence [\textit{T. Bridgeland} et al., J. Am. Math. Soc. 14, No. 3, 535--554 (2001; Zbl 0966.14028)]. If the Craw-Ishii conjecture holds, the conditions also cover all 3-fold projective crepant resolutions. Finally, the authors show that this framework applies to all constructions with fibres of dimension at most one. \par The authors claim that the construction of the sheaf $\mathcal D$ of noncommutative rings is the most subtle part of the article, and consider this for local to global-problems for one specific example, the crepant divisor-to-curve contraction $f:X\rightarrow Y$ where $Y$ is singular along a one-dimensional locus $Z$ and above every point in $Z$ is an irreducible curve. \par By a contraction, the authors mean a projective birational map $f:X\rightarrow Y$ between $d$-dimensional normal varieties over an algebraically closed field $\Bbbk$, satisfying $\mathbf{R}f_\ast\mathcal O_X=\mathcal O_Y$ where $Y$ is quasi-projective. The authors assume there exists a vector bundle $\mathcal P$ on $X$ containing $\mathcal O_X$ as a summand such that (1) the natural map $f_\ast\mathcal End_X(\mathcal P)\rightarrow\mathcal End_Y(f_\ast\mathcal P)$ is an isomorphism, (2) The functor $\mathbf{R}f_\ast\mathbf{R}\mathcal Hom_X(\mathcal P,-):\text{D}^b(\operatorname{coh}X)\rightarrow\text{D}^b(\bmod\mathcal B)$ is an equivalence with $\mathcal B=f_\ast\mathcal End_X(\mathcal P)$ considered as a sheaf of $\mathcal O_Y$-algebras. Write $\mathcal Q=f_\ast\mathcal P$ so that $\mathcal B=\mathcal End_Y(\mathcal Q)$, and define the subsheaf $\mathcal I$ of $\mathcal B$ consisting of local sections that at each stalk factor through a finitely generated projective module. Then $\mathcal I$ is naturally a subsheaf of two-sided ideals of $\mathcal B$ and can be viewed as a subsheaf of $\mathcal B^{\text{op}}$. Now the authors define $\mathcal D=\mathcal B^{\text{op}}/\mathcal I$, and call it the sheafy contraction algebra. In fact, two alternative ways of describing $\mathcal I$ is given, and so the Global Contraction Theorem can be proved. It says that $\text{Supp}_Y\mathcal D=Z$. This says that $\mathcal D$ is a sheaf of algebras on the locus $Z$, and the ringed space $(Z,\mathcal D)$ is a noncommutative enhancement of $Z$. Notice that the stalks of $\mathcal D$ are not in general local rings, but is interpreted as a sheaf of algebras on the base $Y$ of the contraction $f:X\rightarrow Y$. This sheaf is lifted to give an object on $X$. This is possible because, by construction, $\mathcal D$ is a factor of $\mathcal B^{\text{op}}$ so also carries the natural structure of a $\mathcal B$-module. Thus $\mathcal D$ is an object of $\text{D}^b(\bmod\mathcal B^{\text{op}})$. Across a dual version of the condition (2) above, this defines an object $\mathcal E=f^{-1}\mathcal D\otimes^{\mathbf{L}}_{f^{-1}\mathcal B^{\text{op}}}\mathcal P^\ast$ of $\text{D}^b(\operatorname{coh}X)$. In general, $\mathcal E$ need not be a sheaf, but the authors find that $\mathbf{R}f_\ast\mathcal E=0$ so that in particular, $\text{Supp}_X\mathcal E$ lies in the exceptional locus. Now we can state the most foundational main result, and we do it verbatim: \par Theorem 1.4. Suppose that $f:X\rightarrow Y$ is a contraction where the fibres have dimension at most one. Then the following hold. (1) The completion of $\mathcal D$ at a closed point $z\in Z$ is Morita equivalent to the algebra that prorepresents noncommutative deformations of the reduced fibre over $z$. (2) $\mathcal E$ is a sheaf. (3) $\text{Supp}_X\mathcal E$ equals the exceptional locus of $f$. (4) The restriction of $\mathcal E$ to the formal fibre above a closed point $z\in Z$ recovers the universal sheaf $\mathcal E_z$ from noncommutative deformation theory, up to additive equivalence. \par Next, the article applies these results to what could be called part of noncommutative birational geometry: Crepant contractions and autoequivalences. Using the exact sequence $0\rightarrow\mathcal I\rightarrow\mathcal B^{\text{op}}\rightarrow\mathcal D\rightarrow 0$, there is constructed a functor $\mathsf{Twist}_X(a)$ (sitting in a functorial triangle). When $Z$ is a point, $\mathsf{Twist}_X$ reduces to a noncommutative twist over a contraction algeba, and the existence of a global sheafy contraction makes it possible to avoid the most delicate local-to-global gluing problems. \par The article gives the following criterion for when $\mathsf{Twist}_X$ is an equivalence: Under some mild natural assumptions, assume $f$ is crepant, and that $\hat{\mathcal O}_{Y,z}$ are hypersurfaces for all closed points $z\in Z$, then the following are equivalent: (1) $\mathcal D$ is a Cohen-Macaulay sheaf on $Y$, and $\mathcal E$ is a perfect complex on $X$. (2) $\mathcal D$ is relatively spherical for all closed points $z\in Z$. If these conditions hold, then the functor $\mathsf{Twist}_X$ is an autoequivalence of $\text{D}^b(\operatorname{coh}X)$. Notice that the conditions are automatically satisfied when $\dim Z\leq\dim Y-3$. \par The criterion above is sufficiently general to include other interesting examples: $\mathsf{Twist}_X$ is an autoequivalence for the Springer resolution $X\rightarrow Y$ of the variety of singular $n\times n$ matrices. Other examples when the conditions are not automatic includes partial resolutions of Kleinian singularities, and 3-fold crepant divisor-to-curve examples. The authors leverage their control of $\mathcal E$ in this setting, which by the main theorem 1.4 above is a sheaf with support coinciding with the exceptional locus. Reformulation the condition (1) by letting $\mathcal D$ be a Cohen-Macaulay sheaf on $Y$ in terms of a Cohen-Macaulay property for $\mathcal E$ on $X$, close to verbatim: With the general setup, suppose further that the fibres of $f$ have dimension at most one. Then the following are equivalent. (1) $\mathcal D$ is a Cohen-Macaulay sheaf on $Y$. (2) For all $y\in Z$, and all $x\in f^{-1}(y),\;\mathcal E_x\in\text{CM}_{\dim Z_y+1}\mathcal O_{X,x}$. In particular, if these conditions hold, then above every $y\in Z$, the exceptional locus is equidimensional of dimension $\dim Z_y+1$. Combining with the previous criterion, this dimension condition gives an easy-to-check obstruction to $\mathcal D$ being relatively spherical, and this is demonstrated through various examples. \par The article lifts birational geometry to the noncommutative situation, and contains important ideas and general theory. The results are interesting, varied, and proves the strength of noncommutative geometry.