A Sylvester-Gallai theorem for cubic curves (Q2136190): Difference between revisions

The theorem of Sylvester-Gallai, a classic result of combinatorial geometry, states that for any finite set \(A\) of points in \(\mathbb{R}^2\), not contained in a line, there exists a straight line that intersects \(A\) in exactly two points. It is conjectured that for any finite point set \(A \subset \mathbb{R}^2\) that is not contained in an algebraic curve of degree \(d\) there exist a -- not necessarily irreducible -- algebraic curve of degree \(d\) that intersects \(\mathbb{P}\) in precisely \(d(d+3)/2\) points (the number of points that typically determine a curve of degree \(d\)). This conjecture was shown to be true for \(d = 2\) by \textit{J. A. Wiseman} and \textit{P. R. Wilson} in [Discrete Comput. Geom. 3, No. 4, 295--305 (1988; Zbl 0643.51009)]. Since then, several other proofs for the case \(d = 2\) have appeared. This article adds a new proof that allows for a generalization to cubic curves and sufficiently large point sets. While details of their proof are a bit involved and complicated, the basic idea is straightforward and elegant: In the projective space of cubic curves a plane is determined by requiring the cubics to pass through seven suitably chosen points. The dual version of the original Sylvester-Gallai theorem then guarantees existence of a point in that space that leads to a solution cubic. The authors write quite openly about the caveats of their approach and also explain reasons for some of its weak points: \begin{itemize} \item The point set \(A\) is required to have a cardinality of at least \(250\). This number is certainly not optimal and could be reduced. However, by using only ideas from this article, it seems impossible to get entirely rid of a restriction on the cardinality of \(A\). \item Unlike the line and conic versions of the Sylvester-Gallai theorem, the authors cannot guarantee that the nine points determine a \emph{unique} cubic. \item Generalization to higher degrees seem difficult. The authors outline a possible approach based on Sylvester-Gallai theorems for points and hyperplanes in higher dimensions. However, nothing seems to be known in that regard and it is entirely possible that these theorems fail to be true. \end{itemize}