Existence theorems for dualizing complexes over non-commutative graded and filtered rings (Q1372672)

We aim to complete some of the results in [\textit{P. Jørgensen}, Proc. Am. Math. Soc. 125, No. 3, 709-716 (1997; Zbl 0860.14002); [9]=\textit{P. Jørgensen}, Commun. Algebra 25, No. 2, 575-591 (1997; Zbl 0871.16021); [14]=\textit{A. Yekutieli} and \textit{J. J. Zhang}, Proc. Am. Math. Soc. 125, No. 3, 697-707 (1997; Zbl 0860.14001); [13]=\textit{A. Yekutieli}, J. Algebra 153, No. 1, 41-84 (1992; Zbl 0790.18005)]. In these papers the authors are concerned with local cohomology, Serre duality, and dualizing complexes over graded rings. Recall that a dualizing complex over a non-commutative ring \(A\) is roughly speaking a bounded complex of \(A\)-bimodules finitely generated on both sides such that \(\text{RHom}_A(-,R)\) defines a duality between suitable subcategories of \(D(A)\) and \(D(A^0)\). In the commutative case, dualizing complexes are fundamental in Grothendieck's duality formalism for coherent sheaves. In the non-commutative case, it is pointed out in [14]\ that the existence of a (``balanced'') dualizing complex implies Serre duality, but the converse is less clear. Below, we will give a necessary and sufficient criterion for the existence of (balanced) dualizing complexes over connected graded rings (Theorem 6.3). This resembles [9, Theorem 3.3]\ which (roughly) states that the Matlis dual of local cohomology defines a duality between \(D^b_{fg}(A)\) and \(D^b_{fg}(A^0)\). However, Jørgensen proves his result only for quotients of AS-Gorenstein rings and even in this somewhat more restricted situation he doesn't quite prove that one obtains a balanced dualizing complex in the sense of [13]. Our criterion is based upon a general local duality formula (Theorem 5.1), a slightly restricted version of which is also proved in [9]\ (Jørgensen's methods would also yield our version). In Section 8 of this paper, we develop the rudiments of a theory of dualizing complexes over ungraded rings. This section really poses more problems than it answers, but we are nevertheless able to show the existence of dualizing complexes over some filtered rings, thereby answering a question by Yekutieli at the end of the introduction to [13]. The following corollary (a special case of Corollary 8.7) is a typical example of what can be obtained. Corollary 1.1. Let \(k\) be a field and assume that \(A\) is a \(k\)-algebra carrying a filtration \(k=F_0A\subset F_1A\subset\cdots\) such that \(\text{gr }A\) is commutative and finitely generated. Then \(A\) has a dualizing complex and in particular \(D^b_{fg}(A)\) is dual to \(D^b_{fg}(A^0)\). As a byproduct of our methods we obtain a general formula for the dualizing complex over a Gorenstein ring (Proposition 8.4). It is tempting to conjecture that such a formula might be used to prove existence of dualizing complexes over more general classes of non-graded rings, but we have not yet been able to do so. Finally in Section 9, we compute the dualizing complex of a Koszul AS-regular algebra [[2]=\textit{M. Artin}, \textit{W. F. Schelter}, Adv. Math. 66, 171-216 (1987; Zbl 0633.16001)]. Specializing to dimension three yields a connection between the matrix \(Q\), prominent in [2]\ and the automorphism defined by the canonical normalizing element in degree three, which was introduced in [\textit{M. Artin}, \textit{J. Tate} and \textit{M. Van den Bergh}, Prog. Math. 86, 33-85 (1990; Zbl 0744.14024), Invent. Math. 106, No. 2, 335-388 (1991; Zbl 0763.14001)].