Higher infinitesimal neighbourhoods (Q1270381)

Let \(Z\) be a configuration of points in the projective space \({\mathbb P}^m\) and call \(h_{Z,k}(n)\) the Hilbert function of the \(k\)-th neighbourhood \(Z^k\). When \(k=2\) and \(Z\) is general, the Hilbert function is known according to the Alexander-Hirschowitz theorem: It has the expected behaviour, except for \(4\) classified cases. When \(k\) grows, the situation is much less understood, even if there are conjectures for the behaviour of \(h_{Z,k}(n)\) when the set \(Z\) is in sufficiently general position. For instance, in this setting \textit{M. V. Catalisano} and \textit{A. Gimigliano} [J. Algebra 183, No. 1, 245-265 (1996; Zbl 0863.14028)] conjectured that \(h_{Z,k}\) is bounded below by the Hilbert function of \(Z_0^k\), \(Z_0\) being a general subset of a rational normal curve. The author studies sets of points in linear general position and their \(j\)-secant varieties, for \(j=1,\dots,k\). The main result is a lower bound for \(h_{Z,h}(n)\) at the crucial value \(n=k+1\), which is consistent with the expected behaviour. In particular it turns out that Catalisano and Gimigliano's conjecture holds at the value \(n=k+1\).