A new proof of the Alexander-Hirschowitz interpolation theorem (Q666280)

Fix \(p_1,\dots, p_h \in \mathbb{P}^r\) distinct points and fix \(m_1, \dots ,m_h\) positive integers. Let \({\mathcal{L}}_{r,d}\) be the linear system of hypersurfaces of \(\mathbb{P}^r\) of degree \(d\) and consider \[ {\mathcal{L}}:={\mathcal{L}}_{r,d}(m_1, \dots, m_d) \] the subsystem of those divisors of \({\mathcal{L}}_{r,d}\) having multiplicity at least \(m_i\) at \(p_i\), \(i=1, \dots, n\). Its virtual dimension is defined to be \[ \nu({\mathcal{L}}):={r+d \choose r}-1-\sum{i=1}^n {r+m_i-1 \choose r} \] i.e., the virtual dimension of \({\mathcal{L}}_{r,d}\) minus the number of conditions imposed by all multiple points \(p_i\). This number cannot be less than \(-1\), hence we define the expected dimension to be \[ \epsilon({\mathcal{L}}):=\text{max}\{\nu({\mathcal{L}}),-1\}. \] If the conditions imposed by the assigned points are not linearly independent, the effective dimension of \({\mathcal{L}}\) is greater than the expected one: in that case we say that \({\mathcal{L}}\) is special. Otherwise, if the effective and the expected dimension coincide, we say that \({\mathcal{L}}\) is non-special. What is known, as yet, is essentially concentrated in the Alexander-Hirschowitz Theorem which says that a general collection of double points in \(\mathbb{P}^r\) gives independent conditions on the linear system \({\mathcal{L}}\) of the hypersurfaces of degree \(d\), with a well known list of exceptions. In this paper the author presents a new proof of this theorem which consists in performing degenerations of \(\mathbb{P}^r\) and analyzing how \({\mathcal{L}}\) degenerates. The degenerations used here were introduced by \textit{C. Ciliberto} and \textit{R. Miranda} [J. Reine Angew. Math. 501, 191--220 (1998; Zbl 0943.14002)] and, originally proposed by Z. Ran, to study higher multiplicity interpolation problem. The original approach consists in degenerating the plane to a reducible surface, with two components intersecting along a line, and simultaneously degenerating the linear system \({\mathcal{L}}\) to a linear system \({\mathcal{L}}_0\) obtained as fibered product of linear systems on the two components over the restricted system on their intersection. The limit linear system \({\mathcal{L}}_0\) is somewhat easier than the original one, in particular this degeneration argument allows to use induction either on the degree or on the number of imposed multiple points. This contruction provides a recursive formula for the dimension of \({\mathcal{L}}_0\) involving the dimensions of the systems on the two components. In this paper the author generalizes this approach to the case with \(r \geq 3\) and completes the proof of Alexander-Hirschowitz Theorem with this method, exploiting induction on both \(d\) and \(r\). A tricky point of this approach is the study of the transversality of the restrictions of the systems on the intersection of the two components. In the planar case, Ciliberto and Miranda proved it using the finiteness of the set of inflection points of linear systems on \(\mathbb{P}^1\). In higher dimension transversality is more complicated. In Section 2.2 and in Section 3.1, the author presents a new approach to this problem: if at least one of the two restricted systems is a complete linear system, then the dimension of the intersection is easily computed. Anyhow, this is not sufficient to finish the proof of Alexander-Hirschowitz Theorem. For instance, it does not work in the cubic case. The solution to this obstacle is to blow up a codimension three subspace \(L\) of \(\mathbb{P}^r\), instead of a point. This approach to the cubic case is not so different from the one proposed by \textit{M. C. Brambilla} and \textit{G. Ottaviani} [J. Pure Appl. Algebra 212, No. 5, 1229--1251 (2008; Zbl 1139.14007)] where they give another alternative, and short, proof of Alexander-Hirschowitz Theorem, in the case \(d\geq 4\), and propose a new and simpler degeneration argument in the cubic case. Also the quartic case must be analysed separately. Indeed, twisting by a negative multiple of the exceptional component of the central fiber, one reduces to quadrics that are special. The proof, in this case, involves a geometric argument that exploits the property of cubics of containing all lines through two distinct double points. The constructions in this paper, besides its intrinsic intent, gives hope for further extensions to greater multiplicities.