Free energy minimizers with radial densities: classification and quantitative stability
Shrey Aryan, Lauro Silini
TL;DR
The paper analyzes the weighted isoperimetric problem in $\mathbb{R}^n$ with radial density $f=e^\psi$ and radial potential $g$, introducing the energy $\mathcal{E}(F)=\mathcal{P}_f(F)+\mathcal{G}_f(F)$ under a fixed weighted volume. It provides a complete 1D classification, shows large-volume optimality by centered balls under $\kappa$-uniform admissibility via calibration, and reveals counterexamples demonstrating that $\psi''+g'>0$ is not universally sufficient for global optimality in higher dimensions. The authors develop a detailed geometric analysis of optimal sets via spherical symmetrization, deriving a constant weighted mean curvature condition and a two-curve (upper/lower) ODE structure for the generating profile, leading to a full uniqueness result for minimizers and an isoperimetric-type inequality. Finally, they establish sharp quantitative stability: a quadratic bound $\mathcal{E}(E)-\mathcal{E}(B_R) \ge c \lvert E\triangle B_R\rvert_f^2$ for volume-constrained near-minimizers, leveraging a Fuglede-type near-spherical estimate and De Giorgi regularity theory to extend to the general setting. These results advance the understanding of weighted isoperimetry in Euclidean/Riemannian contexts and provide precise stability controls for radial densities and exterior potentials.
Abstract
We study the isoperimetric problem with a potential energy $g$ in $\mathbb{R}^n$ weighted by a radial density $f$ and analyze the geometric properties of minimizers. Notably, we construct two counterexamples demonstrating that, in contrast to the classical isoperimetric case $g = 0$, the condition $\ln(f)'' + g' \geq 0$ does not generally guarantee the global optimality of centered spheres. However, we demonstrate that centered spheres are globally optimal when both $f$ and $g$ are monotone. Additionally, we strengthen this result by deriving a sharp quantitative stability inequality.
