Elekes-Szabó for collinearity on cubic surfaces
Martin Bays, Jan Dobrowolski, Tingxiang Zou
TL;DR
This work extends the Elekes-Szabó paradigm to the 3D orchard problem on complex cubic surfaces, achieving a full classification of when quadratic growth in collinear triples can occur without plane concentration. It combines model-theoretic coarse pseudofinite dimension methods with geometric and incidence-geometry techniques: a coherent-action/nilpotence argument handles reducible cubics, while a Szemerédi–Trotter–type energy analysis paired with fixed-point geometry of Geiser involutions handles irreducible cubics. The main finitary takeaway is that, for large finite point sets on a cubic surface not equal to three planes through a common line, substantial triple-line configurations must concentrate on a single plane, thereby mirroring the planar orchard phenomenon in a higher-dimensional setting. These results illuminate the higher-dimensional Elekes-Szabó landscape by identifying nilpotent-group actions as the natural mechanism behind quadratic incidences in the collinearity setting on cubic surfaces. The methods open avenues for broader applications to Ind-definable group actions and to other ternary incidence relations arising from algebraic surfaces.
Abstract
We study the orchard problem on cubic surfaces. We classify possibly reducible cubic surfaces $X\subseteq \mathbb{P}^3(\C)$ with smooth components on which there exist families of finite sets (of unbounded size) with quadratically many 3-rich lines which do not concentrate (in a natural sense) on any projective plane. Namely, we prove that such a family exists precisely when $X$ is a union of three planes sharing a common line. Along the way, we obtain a general result about nilpotency of groups admitting an algebraic action satisfying an Elekes-Szabó condition, and we prove the following purely algebrogeometric statement: if the composition of four Geiser involutions through sufficiently generic points $a,b,c,d$ on a smooth irreducible cubic surface has infinitely many fixed points, then a single plane contains $a,b,c,d$ and all but finitely many of the fixed points.
