Three kinds of mutation. (Q663578): Difference between revisions

Let \(H\) be a hereditary finite dimensional algebra with \(n\) isomorphism classes of simple objects. The tilting objects for \(H\) inside \(\mathcal D=\mathcal D^b(H)\), the bounded derived category of finitely generated \(H\)-modules, are of considerable importance, and several variations on the concept have also been studied. In the paper under review, the authors are concerned with three of these: exceptional sequences in \(\text{mod\,}H\), silting objects in \(\mathcal D\), and \(m\)-cluster tilting objects in the \(m\)-cluster category, \(\mathcal C_m=\mathcal D/\tau^{-1}[m]\), which is an orbit category of \(\mathcal D\). There are mutation operations on both the set of \(m\)-cluster tilting objects and the set of exceptional sequences. It is also possible to define a mutation operation for silting objects. The main goal of this paper is to relate these different notions of mutation. A basic object \(T\) in \(\mathcal D\) is said to be partial silting if \(\text{Ext}^i(T,T)=0\) for \(i>0\), and silting if in addition \(T\) is maximal with this property. A basic object \(T\) in \(\mathcal C_m\) is called \(m\)-cluster tilting if \(\text{Ext}_{\mathcal C_m}^i(T,T)=0\) for \(0<i\leq m\) and if \(X\) satisfies \(\text{Ext}_{\mathcal C_m}^i(T,X)=0\) for \(0<i\leq m\), then \(X\) is in \(\text{add\,}T\). This condition is known to be equivalent to \(T\) being maximal with respect to \(\text{Ext}_{\mathcal C_m}^i(T,T)=0\) for \(0<i\leq m\) [see \textit{A. Wrålsen}, J. Algebra 321, No. 2, 532-547 (2009; Zbl 1177.18008) and \textit{Y. Zhou} and \textit{B. Zhu}, J. Algebra 321, No. 10, 2898-2915 (2009; Zbl 1196.16013)], and to \(T\) having \(n\) indecomposable summands. If \(T\) satisfies \(\text{Ext}_{\mathcal C_m}^i(T,T)=0\) for \(0<i\leq m\) and it has \(n-1\) summands, it is called `almost \(m\)-cluster tilting'. If we make a (reasonable, but not canonical) choice of a fundamental domain in \(\mathcal D\) for \(\tau^{-1}[m]\), then the lifting of objects from \(\mathcal C_m\) to \(\mathcal D\) takes \(m\)-cluster tilting objects to silting objects. Conversely, any silting object is the lift of an \(m\)-cluster tilting object for \(m\) sufficiently large. This allows us to deduce the number of summands of a silting object, and to lift the notion of mutation from \(m\)-cluster tilting objects to silting objects. Exceptional sequences were investigated first in the context of algebraic geometry [\textit{A. N. Rudakov}, Lond. Math. Soc. Lect. Note Ser. 148, 1-6 (1990; Zbl 0721.14011)], and later for hereditary finite dimensional algebras by \textit{W. Crawley-Boevey}, [CMS Conf. Proc. 14, 117-124 (1993; Zbl 0828.16012)], and \textit{C. M. Ringel}, [Contemp. Math. 171, 339-352 (1994; Zbl 0851.16010)]. An exceptional sequence in \(\text{mod\,}H\) is a sequence of indecomposable \(H\) modules, \((E_1,\dots,E_r)\), with the properties that \(\Hom(E_i,E_j)=0\) for \(j<i\) and \(\text{Ext}^1(E_i,E_j)=0\) for \(j\leq i\). The maximal length of an exceptional sequence is \(n\); an exceptional sequence of maximal length is called `complete', and one of length \(n-1\) is called `almost complete'. Given a sequence of indecomposable \(H\) modules \((M_1,\dots,M_n)\), a `placement' of \((M_1,\dots,M_n)\) is a sequence \((\widehat M_1,\dots,\widehat M_n)\) of objects in \(\mathcal D\) such that \(\widehat M_i\) is isomorphic to \(M_i[t_i]\) for some \(t_i\in\mathbb Z\). The authors are interested in studying placements of an exceptional sequence which form a silting object. It is easy to see that any complete exceptional sequence admits such a placement. It is natural to ask if it is possible to define a placement for each exceptional sequence in \(\Hom\,H\), such that if two exceptional sequences are related by a single mutation, the same would be true for the corresponding silting objects. Simple examples show that this is too much to ask. In fact, for some \(H\), there exists an exceptional sequence \(\mathcal E\) such that it is not possible to place \(\mathcal E\) and the sequences \(\mathcal E_i\) obtained by applying a single mutation to \(\mathcal E\), in such a way that the corresponding silting object \(\widehat{\mathcal E}\) is related by a single mutation to each of the \(\widehat{\mathcal E_i}\). The main result of the paper is as follows: let \(\mathcal F\) be an almost complete exceptional sequence, and \(\mathcal E_1,\dots,\mathcal E_{n+1}\) be the complete exceptional sequences which can be obtained by adding one term to \(\mathcal F\). Essentially by definition, \(\mathcal E_i\) and \(\mathcal E_{i+1}\) are related by a single mutation. Then there exists a placement \(\widehat{\mathcal E_i}\) of each \(\mathcal E_i\), all of which agree as to the placement of \(\mathcal F\), such that \(\widehat{\mathcal E_i}\) is silting for all \(i\), and such that \(\widehat{\mathcal E_i}\) and \(\widehat{\mathcal E_{i+1}}\) are related by a single mutation. In the course of investigations, the authors define a quiver associated to an exceptional sequence, as follows. The vertex set consists of the terms in the exceptional sequence; there is an arrow from \(E_i\) to \(E_j\) if \(\Hom(E_i,E_j)\neq 0\) or \(\text{Ext}^1(E_j,E_i)\neq 0\). They show that this quiver is always acyclic, and that it is connected provided \(H\) is connected and the exceptional sequence is complete. This acyclicity is crucial for the main result mentioned above.