Halfspaces minimise nonlocal perimeter: a proof via calibrations (Q778064): Difference between revisions

The paper deals with a nonlocal version of Plateau's problem. The author considers a nonlocal extension of the total variation functional, featuring a positive kernel which may be not summable at the origin. This includes fractional kernels and extends previous literature to the case where the kernel is not monotone or not radial. The main result of the paper is that half-spaces are the unique minimizers of the nonlocal perimeter in a ball. The minimality is proven by showing that halfspaces admit calibrations, where the classical notion of calibration is adapted to the nonlocal setting. In this context, a calibration \(\zeta\) of a given set \(E\) is a scalar function, defined on pairs of points \((x,y)\) in the space, with values in \([-1,1]\), and satisfying two technical properties: The first property is a nonlocal extension of the vanishing divergence condition. The second property implies that \(\zeta(x,y)=-1\) if \(x\in E\) and \(y\in E^{c}\), while \(\zeta(x,y)=1\) if \(x\in E^{c}\) and \(y\in E\); thus (under regularity assumptions) \(\zeta(x,y)\) gives the sign of the scalar product between the vector \(y-x\) and the inner normal to \(E\) at a point close to the segment between \(x\) and \(y\). It is proven that a set with a calibration is a local minimizer; moreover, any local minimizer is associated to a calibration. The final section concerns an application to a \(\Gamma\)-convergence result for nonlocal functionals with a certain scaling of the kernel. It is shown that the functionals converge to (an extension of) De Giorgi's perimeter, featuring an anisotropic norm. The main tool to prove this result is a characterization of the anisotropic norm, obtained by using the minimality of half-spaces shown in the first part of the paper.