Neat idempotents and tiled orders having large global dimension (Q1858215): Difference between revisions

Neat idempotents have been introduced by \textit{I. Ágoston, V. Dlab}, and \textit{T. Wakamatsu} [Commun. Algebra 19, No. 2, 433-442 (1991; Zbl 0743.16015)] in connection with quasi-hereditary algebras. For a finite dimensional algebra \(A\) over a field \(K\), a primitive idempotent \(e\in A\) is said to be neat if \(\text{Ext}^i_A(S,S)=0\) holds for the simple modules \(S_A:=eA/\text{Rad }eA\) and all \(i>0\). It was proved there that the global dimension of \(A\) can be estimated in terms of \(C:=(1-e)A(1-e)\) when \(e\) is neat. The author applies this idea to tiled orders, which leads to a new version of Jategaonkar's conjecture (in its original form disproved by Kirkman and Kuzmanovich, and the author). So the question arises whether each tiled order of finite global dimension admits a neat idempotent. If this were true, an exact upper bound for the exponent sum \(\sum\lambda_{ij}\) of a tiled order \(\Lambda=(\Pi^{\lambda_{ij}})\) with \(\text{gld }\Lambda<\infty\) would be known. As a main application of neat idempotents, the author gives a short (and new) construction of \textit{W. S. Jansen} and \textit{C. J. Odenthal}'s series of tiled orders \(\Gamma_n\) with \(\text{gld }\Gamma_n=2n-8\) [J. Algebra 192, No. 2, 572-591 (1997; Zbl 0877.16008)] for all (not necessarily even) \(n\geq 6\). Moreover, he gives an example of a tiled order \(\Lambda\) in \(\text{M}_8(K)\) with \(\text{gld }\Lambda=10\).