Maximizing Riesz capacity ratios: conjectures and theorems

TL;DR

This paper develops a shape optimization framework for Riesz capacity ratios, conjecturing maximizers as geometric shapes like balls or simplices in different parameter regions, and proves several cases especially in low dimensions.

Contribution

It introduces a new optimization approach for Riesz capacities, proposing conjectures for maximizers across parameter regions and proving results in low-dimensional cases.

Findings

Maximizers are conjectured to be balls, simplices, or intervals depending on parameters.

Proved maximality for certain shapes in dimensions 1 and 2.

Identified symmetry-breaking transition regions in the parameter space.

Abstract

A shape optimization program is developed for the ratio of Riesz capacities $\text{Cap}_q(K)/\text{Cap}_p(K)$, where $K$ ranges over compact sets in $\mathbb{R}^n$. In different regions of the $pq$-parameter plane, maximality is conjectured for the ball, the vertices of a regular simplex, or the endpoints of an interval. These cases are separated by a symmetry-breaking transition region where the shape of maximizers remains unclear. On the boundary of $pq$-parameter space one encounters existing theorems and conjectures, including: Watanabe's theorem minimizing Riesz capacity for given volume, the classical isodiametric theorem that maximizes volume for given diameter, Szeg\H{o}'s isodiametric theorem maximizing Newtonian capacity for given diameter, and the still-open isodiametric conjecture for Riesz capacity. The first quadrant of parameter space contains P\'{o}lya and Szeg\H{o}'s conjecture on maximizing Newtonian over logarithmic capacity for planar sets. The maximal shape for each of these scenarios is known or conjectured to be the ball. In the third quadrant, where both $p$ and $q$ are negative, the maximizers are quite different: when one of the parameters is $-\infty$ and the other is suitably negative, maximality is proved for the vertices of a regular simplex or endpoints of an interval. Much more is proved in dimensions $1$ and $2$, where for large regions of the third quadrant, maximizers are shown to consist of the vertices of intervals or equilateral triangles.