A nonlocal free boundary problem with Wasserstein distance (Q6050109)

In this interesting paper, the author uses Fourier transform techniques to study the probability measures \(\rho\in\mathcal{M}(\mathbb{R}^2)\) minimizing the functional \[ J[\rho]=\int\int\log\left(\frac{1}{|x-y|}\right)d\rho(x)d\rho(y)+d^2(\rho,\rho_0), \] where \(\rho_0\) is a probability measure, and \(d(\rho,\rho_0)\) is the Wasserstein distance. The first main result of the paper proves that if \(\rho_0\) has compact support, then there exists a probability measure with compact support \(\rho\) minimizing \(J\). This is done by showing the existence of a minimizing sequence with uniformly bounded supports. The key ingredients are a Carleson's estimate, along with a localization argument for the Fourier transforms of the measures. Furthermore, minimizers are shown to be absolutely continuous with respect to the Lesbegue measure. Additionally, the author shows that the potential \(U^{\rho}=-\log|x|\ast\rho\) solves a degenerate obstacle problem, where the obstacle \(\psi\) is the transport potential, that is, \begin{align*} \Delta U^{\rho}&=\Delta \psi\chi_{\{\rho>0\}},\\ \rho(U^{\rho}-\psi)&=0. \end{align*} The singular set of the free boundary of the corresponding obstacle problem is shown to have Hausdorff dimension less than \(n-1\), and to be contained in a rectifiable set. Finally, it is shown that \(U^{\rho}\) solves a nonlocal Monge-Ampère equation, whose linearization leads to \(\rho_t=\operatorname{div}(\rho\nabla U^{\rho})\).