Partition Rank and Partition Lattices

Abstract

We introduce a universal approach for applying the partition rank method, an extension of Tao's slice rank polynomial method, to tensors that are not diagonal. This is accomplished by generalizing Naslund's distinctness indicator to what we call a partition indicator. The advantages of partition indicators are two-fold: they diagonalize tensors that are constant when specified sets of variables are equal, and even in more general settings they can often substantially reduce the partition rank as compared to when a distinctness indicator is applied. The key to our discoveries is integrating the partition rank method with Möbius inversion on the lattice of partitions of a finite set. Through this we unify disparate applications of the partition rank method in the literature. We then use our theory to address a finite field analogue of a question of Erdős, generalize work of Pach et. al on bounding sizes of sets avoiding right triangles to bounding sizes of sets avoiding right k-configurations, and complement work of Sauermann and Tao to determine lower bounds on sizes of sets in F n p that must have solutions to a homogeneous linear equation where many but not all variables are equal.