Gale duality and free resolutions of ideals of points (Q1298083): Difference between revisions

How is the resolution of the ideal of a set of distinct generic points in \({\mathbb{P}}^n\) like? A conjecture about the graded Betti numbers of such resolutions (known as the ``minimal resolution conjecture'', MRC) was given by \textit{A. Lorenzini} [J. Algebra 156, No. 1, 5-35 (1993; Zbl 0811.13008)], and it has been proved in many cases and even asymptotically (when the number of points is much bigger then \(n\), by a result of Hirschowitz and Simpson). Examples found computationally (Schreyer 1993) suggested nevertheless that the MRC could be false in general, even if no ``geometrical reason'' for those (three) counterexamples was known. In the beautiful paper under review, such reason is found, enclosed in the theory of ``Gale transforms'' (they could be viewed as duality of linear series on a finite Gorenstein scheme), a way to associate to a set \(\Gamma \) of \(\gamma \geq r+3\) distinct points in \({\mathbb{P}}^r\) another set \(\Gamma ^\prime \) of \(\gamma \) points in \({\mathbb{P}}^s\), with \(s={\gamma -r -2}\). The idea, expressed in classical language is this: With an appropriate choice of the coordinates, we can suppose that the coordinates of the points of \(\Gamma \) are the rows of a \((r+1)\times \gamma\) matrix \((I_{r+1}|B)\); then the coordinates of the points in \(\Gamma ^\prime \) are the rows of the matrix \((B^T|I_{s+1})\). In the paper the relation among graded Betti numbers of \(\Gamma \) and of \(\Gamma ^\prime\) are found, and the ``mystery'' of the counterexamples to MRC is solved, moreover an infinite family of counterexample is determined (in any \({\mathbb{P}}^r\) with \(r\geq 6\), \(r\neq 9\)).