Computing limit linear series with infinitesimal methods (Q2466306)

The paper concerns with the classical interpolation problem in Algebraic Geometry to determine the dimension \(d\) of the system of hypersurfaces of given degree and having multiplicity \(m_i\) at given points \(p_i\). The previous problem is usually attacked using specialization methods. The points \(p_i\) are moved to a special position then one computes the dimension \(d_0\) in this special position and check if \(d_0\) is equal to the expected dimension \(d_e\) of the starting system. Since, by semi-continuity one has \(d_e \leq d \leq d_0\) we obtain \(d=d_e\). The most common specialization methods are the Horace method [\textit{J. Alexander} and \textit{A. Hirschowitz}, Invent. Math. 140, No. 2, 303--325 (2000; Zbl 0973.14026)], collision of fat points [\textit{L. Evain}, J. Algebra. Geom. 8, 787--796 (1999; Zbl 0953.14027)] and degeneration of the projective space [\textit{C. Ciliberto} and \textit{R. Miranda}, J. Reine Angew. Math. 501, 191--220 (1998; Zbl 0943.14002)]. Unluckily all these methods are hardly usable when there are few points with high multiplicities. Thus, the author introduces a method of specialization, based on a study of deformations of a space of sections, which tackles the numerical difficulties appearing in these difficult cases. Successively the author applies this method to extend results by Nagata relative to the Hilbert functions of fat points in the plane. In fact, a consequence of Alexander-Hirschowitz is that the Hilbert function of a generic union of \(k\) points in the plane of multiplicity \(m_1,\dots, m_k\) is \[ H_Z(d)=\min\left( \frac{(d+1)(d+2)}{2}, \sum_{i=1}^k \frac{m_i(m_i+1)}{2}\right) \] provided \(k>>\max(m_i)\). The case with a fixed number of points (\(\geq 10\)) and big multiplicities have been considered by Nagata. In the article the author proves a result which compute the Hilbert function for every degree.