pipette
ESEspañol

Van der Waerden's Problem for Cubic Surfaces

Evan O'Dorney

Preprint

In the authors' words

For a smooth cubic surface , the absolute Galois group acts on the lines of , giving a subgroup \[ G_S\subseteq W(E_6). \] We study the distribution of when the twenty coefficients of a defining cubic form are chosen independently from a large interval . We prove that, for every , the number of smooth integral cubic forms of height at most for which \[ G_S\neq W(E_6) \] is \[ O_\varepsilon\!\left(H^{19+1/21+\varepsilon}\right). \] The principal obstruction is the unique subgroup of index . By a theorem of Saito, membership in this subgroup is equivalent to the signed discriminant of the cubic surface being a square. A recent square-value estimate of Bonolis, Pierce, and Woo then gives the exponent . For each of the other five conjugacy classes of maximal subgroups of , recent results on rational points in projective thin sets give the sharper bound . We also prove the lower bound \[ N_{exc}(H)\gg H^{16} \] by counting cubic surfaces containing a fixed rational line, and exhibit an -sized exceptional family with no rational line, arising from rational Eckardt points. These examples lead us to conjecture that \[ N_{exc}(H)=H^{16+o(1)}. \] Finally, we investigate whether Bhargava's proof of van der Waerden's conjecture for can be adapted to the index- -problem. A double-discriminant argument gives the desired bound whenever the product of the strongly square-dividing primes is sufficiently large. The remaining difficulty comes from weak square divisibility.

Main resultLimitation the authors admit

Appeared: Friday, October 30. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 14 pages, written autonomously with ChatGPT