Dominant dimension and tilting modules (Q2312824)

This paper studies the (classical) tilting modules over Artin algebras which admit both a projective cover which is also injective and an injective envelope which is also projective. It addresses questions concerning the existence and uniqueness of such modules, properties of the modules when they do exist, and finding classes of examples of algebras admitting such a module. This work builds on Crawley-Boevey and Sauter's observation [\textit{W. Crawley-Boevey} and \textit{J. Sauter}, Math. Z. 285, No. 1--2, 367--395 (2017; Zbl 1405.16028)] that every Auslander algebra admits a unique such tilting module, and is parallel to independent work by the reviewer and \textit{J. Sauter} [in: Proceedings of the 50th symposium on ring theory and representation theory, University of Yamanashi, Japan, 2017. Yamanashi: Symposium on Ring Theory and Representation Theory Organizing Committee. 155--164 (2018; Zbl 1410.16022)] concerning similar questions. (By way of comparison, while Pressland and Sauter also consider non-classical tilting modules, with projective dimension more than \(1\), the present paper deals with the case of classical tilting modules in more detail.) It is shown that an Artin algebra \(\Lambda\) admits a tilting module \(T\) of the above kind if and only if it has dominant dimension at least \(2\) (or equivalently, it is isomorphic to the endomorphism algebra of a generating-cogenerating module over a second algebra), in which case the tilting module is unique. The endomorphism ring of \(T\) is shown to have global dimension at most that of \(\Lambda\), and moreover that the inequality is strict (in which case the difference is exactly \(1\)) if and only if the projective dimension of the Auslander-Reiten translation of \(T\) is not maximal. The authors show that certain triangular matrix algebras extending Auslander algebras admit a tilting module of the desired form, and also classify those Nakayama algebras which do. In these cases, the relevant tilting module may be described explicitly. Most strikingly, the authors use these ideas to give a class of algebras \(\Lambda\) of representation dimension \(4\) for which the finitistic dimension conjecture holds; the class is characterised in terms of properties of the tilting module of interest for the endomorphism algebra of a certain \(\Lambda\)-module. Note that the finitistic dimension conjecture is known to hold for all algebras of representation dimension at most \(3\) [\textit{K. Igusa} and \textit{G. Todorov}, Fields Inst. Commun. 45, 201--204 (2005; Zbl 1082.16011)].