Large tilting sheaves over weighted noncommutative regular projective curves (Q507693)

Due to Beilinson, a form of tilting theory can be used to establish a connection between algebraic geometry and representation theory of finite dimensional algebras. The starting example is \(\mathbb X=\mathbb P_1(k)\) over a field \(k\), where the the derived Hom-functor \(\mathbf{R}\text{Hom}(T,-)\) defines an equivalence between the derived categories of \(\text{Qcoh}\mathbb X\) and \(\text{Mod}\Lambda\), \(\Lambda\) the the path algebra of the quiver with two nodes and double arrows, that is, the Kronecker algebra. This follows by considering the tiling sheaf \(T=\mathcal O\oplus\mathcal O(1)\) in \(\text{coh}\mathbb X\) with endomorphism ring \(\Lambda\). More examples exist where a noetherian tilting object \(T\) in a triangulated category \(\mathcal D\) provides an equivalence between \(\mathcal D\) and the derived category of \(\text{End}(T)\), e.g. for Calabi-Yau and cluster categories. To be able to work through the introduction, the reader will need to know the definition of a noncommutative curve (or noncommutative scheme) in general. Section 2 in the article contains one of the best reviews of these concepts, and the reader will gain by going back to the introduction after reading Section 2: A noncommutative curve \(\mathbb X\) is a category \(\mathcal H\) of coherent sheaves over \(\mathbb X\). \(\mathcal H\) is small, connected, abelian, and every object in \(\mathcal H\) is noetherian. \(\mathcal H\) is a \(k\)-category with finite-dimensional \(\text{Ext}\) -spaces, and there is an autoequivalence \(\tau\) on \(\mathcal H\) (Auslander-Reiten translation) such that Serre duality holds. Finally, \(\mathcal H\) contains an object of infinite length. Now, an indecomposable coherent sheaf \(E\) which is not of finite length, is torsion free and is called a vector bundle. Then one can write \(\mathcal H=\mathcal H_+\vee\mathcal H_0\) with \(\mathcal H_+\) the class of vector bundles, and so \(\mathcal H_0=\underset{x\in\mathbb X}\coprod\mathcal U_x\) where \(\mathbb X\) is an index set and every \textit{tube} \(\mathcal U_x\) is a connected uniserial length category. For this definitions to make sense, the authors demands that the set \(\mathcal X\) consists of infinitely many points, and then \(\mathbb X\) is called a \textit{weighted noncommutative regular projective curve} over \(k\). It follows (in the domestic case) that for all points \(x\in\mathbb X\) there are exactly \(p(x)<\infty\) simple objects in \(\mathcal U_x\), and that for almost all, \(p(x)=1\). The numbers \(p(x)\) with \(p(x)>1\) are called the weights. Weighted noncommutative regular curves are \textit{noncommutative smooth proper curves} (in the sense of Stafford and van den Bergh), and if \(\mathcal H\) in addition to the other properties also admits a \textit{canonical tilting object}, \(\mathbb X\) is a \textit{noncommutative curve of genus 0}. The weighted projective lines introduced by \textit{W. Geigle} and \textit{H. Lenzing} [Lect. Notes Math. 1273, 265--297 (1987; Zbl 0651.14006)] provide the basic framework for this article. Their main advantage is that they contains a tilting bundle in the category \(\text{coh}(\mathbb X)\). Here, the corresponding finite-dimensional algebras are the canonical algebras which is a much studied object in representation theory. The tilting objects are \textit{small} if they are noetherian objects, and their endomorphism rings are finite-dimensional algebras. In general, an object \(T\) in a Grothendieck category \(\mathfrak H\) is called \textit{tilting} if \(T\) generates the objects in \(T^{\perp_1}=\{X\in\mathfrak H|\text{Ext}^1(T,X)=0\}\). These large tilting objects, occurring frequently, are not necessarily in derived correspondence to finite dimensional algebras, but they are connected to \textit{recollements of triangulated categories}, proving strong relationship between the derived categories involved. The large tilting models are interesting because of their connection with localization of categories: Let \(R\) be a Dedekind domain. Then the tilting modules over \(R\) are parametrized by the subsets \(V\subseteq\text{Max-Spec} R\), obtained by localizations at sets \(V\) of simple modules: The universal localization \(R\hookrightarrow R_V\) yields the tilting module \(T_V=R_V\oplus R_V/R\), and the set \(V=\emptyset\) corresponds to the only finitely generated tilting module, the regular module \(R\). For arbitrary tame hereditary algebras, the classification of tilting modules is complicated due to the possible presence of finite dimensional direct summands from non-homogeneous tubes. Infinte dimensional tilting modules are parametrized by pairs \((B,V)\) where \(B\) is called a branch module, and \(V\subseteq\mathbb X\). The tilting module corresponding to \((B,V)\) has finite dimensional part \(B\) and an infinite dimensional part of the form \(T_V\) inside a suitable subcategory. The main goal of this article, is the classification of large tilting objects in hereditary Grothendieck categories (categories acting like the category of vector bundles). Of particular interest is the category \(\text{Qcoh}\mathbb X\) of quasicoherent sheaves over a weighted noncommutaive regular projective curve \(\mathbb X\) over a field \(k\). Also, it is proved how the known results for tame hereditary algebras extend to the general setting. Note in the following that the authors uses the different perpendicular classes, e.g. \(T^{\perp_1}=\{\mathcal F|\text{Ext}^1(\mathcal F,T)\}=0\). For a locally coherent Grothendieck category \(\mathfrak H,\) a class \(\mathcal S\subseteq\in\mathcal H\), \(\mathcal H\) the class if finitely presented objects in \(\mathfrak H\), is called \textit{resolving} if it generates \(\mathfrak H\) and has a specified closure property. The authors prove that if \(\text{pd}(S)\leq 1\) for all \(S\in\mathcal S\), there is a tilting object \(T\in\mathfrak H\) with \(T^{\perp_1}=\mathcal S^{\perp_1}\). This is a first existence theorem, leading up to the main results of the article: For \(\mathbb X\) a weighted noncommutative regular projective curve and \(\mathfrak H=\text{Qcoh}\mathbb X\), the assignment \(\mathcal S\mapsto\mathcal S^{\perp_1}\) is a bijection between resolving classes \(\mathcal S\in\mathcal H\) and tilting classes \(T^{\perp_1}\) of finite type. The authors develop a lot of tools to do the announced classification, and these results are interesting by themselves. Also, many techniques and inductive arguments are pinpointed, and of great value in other circumstances. They are use to prove the main content of the article which is the classification of large tilting sheaves: More or less verbatim: Let \(\mathbb X\) be of tubular type. Then every large tilting sheaf in \(\text{Qcoh}\mathbb X\) has a well defined slope \(w\). If \(w\) is irrational, then there is up to equivalence precisely one tilting sheaf of slope \(w\). If \(w\) is rational or \(\infty\), then the large tilting sheaves of slope \(w\) are classified like in the domestic case. From the above theorem, the authors consider the various special cases, and ends out with many general, interesting results. The article is very clear, and it gives an excellent introduction to noncommutative geometry and some of its applications. The article proves the necessity of noncommutative geometry by generalising results from commutative geometry, and most important; it gives the connection between representation theory and algebraic geometry.