Minors and resolutions of non-commutative schemes (Q1677177)

Let \(B\) be a ring and \(P\) a finitely generated projective left \(B\)-module. Then a minor of \(B\) is a ring \(A=B_P=(\text{End}_BP)^{\text{op}}\). A thick subcategory \(\mathcal C\) of an abelian or triangulated category \(\mathcal A\) is \textit{localizing (colocalizing)} if the canonical functor \(\mathsf{G}:\mathcal A\rightarrow\mathcal A/\mathcal C\) has a right (left) adjoint \(\mathsf{F}\). If \(\mathcal C\) is both localizing and colocalizing it is a \textit{bilocalizing}. In this case \(\mathsf{G}\) is called a \textit{bilocalization functor}. The category of \(B\)-modules and \(A\)-modules are closely related in several ways: 1) The functors \(\mathsf{F}=P\otimes_A-,\;\mathsf{H}=\text{Hom}_A(P^\vee,-):A-\mathsf{Mod}\rightarrow B-\mathsf{Mod}\) are fully faithful. 2) \(\mathsf{G}=\text{Hom}_B(P,-):B-\mathsf{Mod}\rightarrow A-\mathsf{Mod}\) is exact and essentially surjective. The pairs \((F,G)\) and \((G,H)\) are adjoint, giving the concept that \(\mathsf{G}\) is a \textit{bilocalization functor}. Put \(I=I_P=\text{Im}(P\otimes_A P^\vee\rightarrow B).\) Letting \(\overline B=B/I\), the category \(\overline B-\mathsf{Mod}\) is the kernel of \(\mathsf{G}\) and \(A-\mathsf{Mod}\) is equivalent to the Serre quotient of \(B-\mathsf{Mod}\) modulo \(\overline B-\mathsf{Mod}\). 3) If the global dimension of \(A\) and \(\overline B\) are finite, and under specific given conditions, one can prove that the global dimension of \(B\) is finite. The bounded derived relations are even stronger: Let \(\mathsf{DG}\) be the derived functor of \(\mathsf{G}\), let \(\mathsf{LF}\) be the left derived functor of \(\mathsf{F}\) and let \(\mathsf{RH}\) be the right derived functor of \(\mathsf{H}\). 4) There are adjoint pairs \((\mathsf{LF},\mathsf{DG})\) and \((\mathsf{DG},\mathsf{RH})\), \(\mathsf{LF},\;\mathsf{RH}\) are fully faithful and the category \(\mathcal D(A-\mathsf{Mod})\) is equivalent to the Verdier localization of \(\mathcal D(B-\mathsf{Mod})\) modulo its triangulated subcategory \(\mathcal D_{\overline B}(B-\mathsf{Mod})\) which consists of complexes with cohomology in \(\overline B-\mathsf{Mod}.\) 5) There exists a semi-orthogonal decomposition \(\mathcal D(B-\mathsf{Mod})=\langle\mathcal D_{\overline B}(B-\mathsf{Mod}),\mathcal D(A-\mathsf{Mod})\rangle.\) Let \(A\) be a commutative normal Gorenstein domain and let \(F\) be a reflexive \(A\)-module with the property that the ring \(B=B_F=\text{End}_A(A\oplus F)^{\text{op}}=\begin{pmatrix} A&F\\F^\vee&E\end{pmatrix}\) with \(E=(\text{End}_A F)^{\text{op}}\) is maximal Cohen-Macaulay over \(A\) and of finite global dimension. Van den Bergh suggested that this is a \textit{non-commutative crepant resolution} of \(A\) because of the fact that under some additional assumptions, the existence of such a ring \(B\) implies the existence of an ordinary, commutative crepant resolution. Let \(e=\begin{pmatrix} 1&0\\0&1\end{pmatrix}\in B\), and let \(P=Be\). Then \(A=B_P\), explaining the relation between NCCR's and minors. \textit{Y. A. Drozd} and \textit{G.-M. Greuel} [J. Algebra 246, No. 1, 1--54 (2001; Zbl 1065.14041)] observed a close relation between coherent sheaves over the nodal cubic \(C=V(zy^2-x^3-x^2z)\subset\mathbb P^2\) and representations of the finite dimensional quiver algebra given by \(\Lambda\) with \(3\) nodes and arrows \(\alpha_1,\alpha_2\) from node \(1\) to node \(2\) and \(\beta_1,\beta_2\) from node \(2\) to node \(3\) with the relation \(\beta_1\alpha_1=\beta_2\alpha_2=0.\) Let \(\mathcal I\) be the ideal sheaf of the singular point of \(C\), let \(\mathcal A=\mathcal End_C(\mathcal O_C\oplus\mathcal I).\) Then \((C,\mathcal A)\) is a ringed space, and the category of coherent left \(\mathcal A\)-modules \(\mathcal A-\mathsf{mod}\) can be considered. \(\mathcal D^b(\mathcal A-\mathsf{mod})\) has a tilting complex whose opposite endomorphism algebra is isomorphic to \(\Lambda\), and that implies that the categories \(\mathcal D^b(\mathcal A-\mathsf{mod})\) and \(\mathcal D^b(\Lambda-\mathsf{mod})\) are equivalent. Also, the triangulated category \(\mathsf{Perf}(C)\) of \textit{perfect complexes} on \(C\) is equivalent to a full subcategory of \(\mathcal D^b(\mathcal A-\mathsf{mod})\). This leads to a sheaf-theoretic version of the construction of minors where the commutative scheme \((C,\mathcal O_C)\) is a minor of the non-commutative scheme \((C,\mathcal A)\). The main objective in this article is to establish a general framework for the theory of minors of non-commutative schemes. The article starts by reviewing the necessary results on localization of abelian and triangulated categories. This includes a definition of non-commutative schemes giving a proof of the result characterizing the triangulated category \(\mathsf{Perf}(\mathcal A)\) of perfect complexes over a non-commutative scheme \((X,\mathcal A)\) as the category of \textit{compact objects} of the unbounded derived category of quasi-coherent sheaves \(\mathcal D(\mathcal A).\) The authors give the general definition of a minor \((X,\mathcal A)\) of a non-commutative scheme \((X,\mathcal B)\) and study the relations between \((X,\mathcal A)\) and \((X,\mathcal B)\). The notion of \textit{quasi-hereditary} non-commutative schemes is introduced, generalizing quasi-hereditary semi-primary rings and quasi-hereditary orders. \textit{Strongly Gorenstein} non-commutative schemes are defined, and in particular non-commutative curves are studied. In particular, the authors construct a categorical resolution for a reduced non-commutative curve which they call the \textit{könig resolution} as an analogue of the construction given by him. If the curve is rational, a tilting complex can be constructed which proves that this categorical resolution can be given as the derived category of modules over a finite dimensional quasi-hereditary algebra. In particular it gives an estimate of the Rouquier dimension of the perfect derived category of coherent sheaves over a non-commutative curve. The article is sufficiently detailed and explicit, and the sheafification of the results are very interesting and a motivation for doing non-commutative algebraic geometry. The article shows that commutative problems can be assigned non-commutative geometry which gives new insight to the original objects.