On the first infinitesimal neighborhood of a linear configuration of points in \(\mathbb P^2\) (Q2496194)

This paper studies the relation between a reduced 0-dimensional subscheme of \({\mathbb P}^2\) and its first infinitesimal neighborhood; in particular, given a reduced \({\mathbf X}\subset {\mathbb P}^2\) with assigned Hilbert function, what can we say about the Hilbert function and the graded Betti numbers of its first infinitesimal neighborhood \({\mathbf Z}\)? Examples are well known of different \({\mathbf X}\)'s having the same Hilbert function but for whom the schemes \({\mathbf Z}\)'s do not. Recall that if \({\mathcal I}_{\mathbf X}\) is the ideal sheaf of \({\mathbf X}\), then \({\mathbf Z}\) is defined by the ideal sheaf \({\mathcal I}_{\mathbf Z}= {\mathcal I}_{\mathbf X}^2\). If \(S_2\) is the set of all possible Hilbert functions for reduced 0-dimensional subschemes of \({\mathbb P}^2\), then there is a 1-1 correspondence between \(S_2\) and the set of all type vectors, i.e. all \(r\)-ples, \(\forall r\), \((d_1,\dots,d_r)\) of integers with \(0<d_1<\cdots<d_r\). Hence here type vectors are another way (more handy for the problem under analysis) to encode the information carried out by the Hilbert function. Let \(T=(d_1,\dots,d_r)\) be a type vector, then a set \({\mathbf X}\) of distinct points in the plane will be said to be a linear configuration of type \(T\) if there are \(r\) distinct lines \(L_1,\dots,L_r\) in the plane such that \({\mathbf X}=\bigcup _{i=1}^r {\mathbf X}_i\), and each \({\mathbf X}_i\) is made of \(d_i\) distinct points on \(L_i\), none of them on \(L_j\), for \(i\neq j\). For any such a scheme \({\mathbf X}\) we have that its Hilbert function and graded Betti numbers are completely determined by \(T\) (and its Hilbert function is the one associated to \(T\)). The question which is answered in this paper is the following: ``For which \(T\)'s we have that all linear configurations \({\mathbf X}\) of type \(T\) are such that their first infinitesimal neighborhoods \({\mathbf Z}\) have all the same Hilbert function (and graded Betti numbers)?'' It turns out that what must happen for \(T=(d_1,\dots,d_r)\) to completely determine such data is the following: Consider the sequence \((d_1,2d_1,d_2,2d_2,\dots,d_r,2d_r)\); re-order it in increasing order so to get, say, \(T'=(m_1,\dots,m_{2r})\) and let \(\Delta T'\) be its difference sequence; then what you must have is that between any two zero entries of \(\Delta T'\) there is at least one entry \(>1\). If this happens, the Hilbert function of \({\mathbf Z}\) for any linear configuration \({\mathbf X}\) of type \(T\) is determined by \(T\), and easily computed from \(T'\); moreover \({\mathbf Z}\) can be obtained by liaison (via a series of double links). Vice versa, if the above condition does not hold, it is possible to exhibit two linear configurations \({\mathbf X_1},{\mathbf X_2}\) of type \(T\) such that their first infinitesimal neighborhoods \({\mathbf Z_1}, {\mathbf Z_2}\) have different Hilbert functions. A slightly stronger condition on \(\Delta T'\) gives the same kind of situation also for the graded Betti numbers of \({\mathbf Z}\). The problem of determining, for a reduced \({\mathbf X}\) with given Hilbert function, if there are a minimum and maximum Hilbert functions for \({\mathbf Z}\) is discussed too. The existence of such a maximum is proved (although no algorithm is given to recover it from the Hilbert function of \({\mathbf X}\)).