Quadratic Gorenstein algebras with many surprising properties (Q2203985)

Let \(k\) be a field of characteristic zero. Artinian Gorenstein algebras presented by quadrics (``quadratic Gorenstein algebras'' for short) were studied by \textit{U. Nagel} and the reviewer [Collect. Math. 64, No. 2, 211--233 (2013; Zbl 1291.13019)], from the point of view of possible Hilbert functions and the Lefschetz properties, making several conjectures and posing many questions. Some of these were answered in the negative by \textit{R. Gondim} and \textit{G. Zappalà} [Proc. Am. Math. Soc. 146, No. 3, 993--1003 (2018; Zbl 1409.13002)]. \textit{M. Mastroeni} et al. [``Quadratic Gorenstein rings and the Koszul property. I'', Preprint, \url{arXiv:1903.08265}; ``Quadratic Gorenstein rings and the Koszul property. II'', Preprint, \url{arXiv:1903.08273}] continued the study of these algebras by bringing in the question of when they are Koszul. They showed that when the regularity is 3 and the codimension is at least 9, there are non-Koszul examples; this negatively answered a question of \textit{A. Conca} et al. [Compos. Math. 129, No. 1, 95--121 (2001; Zbl 1030.13005)]. More specifically, Mastroeni et al. [loc. cit.] asked for which \((r,c)\) does there exist a non-Koszul, quadratic Gorenstein algebra with regularity \(r\) and codimension \(c\). They give an almost complete answer, leaving open \((r,c) = (3,6), (3,7)\) and \((3,8)\). In this paper the authors first settle the case \((3,8)\) in the affirmative using Nagata's idealization construction, and they show that the other two cannot be settled in the same way and in fact their example is minimal with respect to the codimension, with regularity 3. Next, the authors show that a certain property called subadditivity fails in a strong way for quadratic Gorenstein algebras (details omitted here). This leads to an infinite family of quadratic Gorenstein algebras that are non-Koszul, have arbitrarily high degree first syzygies, have non-unimodal Hilbert functions, and do not satisfy the strong or weak Lefschetz properties. This gives a counterexample to a conjecture of Nagel and the reviewer [loc. cit.], extending a previous counterexample of Gondim and Zappalà [loc. cit.]. Finally, the authors extend a method of Roos to show that there is no finite test of the Koszul property for quadratic Gorenstein algebras: they show that for any integer \(\alpha \geq 2\), there is an Artinian quadraatic Gorenstein algebra \(B_\alpha\) with \(\hbox{codim}(B_\alpha) = 14\) and \(\hbox{reg}(B_\alpha) = 3\) such that the free resolution of the residue field \(k\) as a \(B_\alpha\) module is linear for precisely \(\alpha\) steps.