An isoperimetric inequality in the plane with a log-convex density (Q1623067): Difference between revisions

The classical isoperimetric problem is to find the sets with minimal boundary measure under a volume constraint or, equivalently, the sets with maximal volume and given boundary measure, that are called isoperimetric sets. The typical approach to this problem is to show the existence of minimizers in the class of sets with finite perimeter. If \(f\) is a positive lower semi-continuous density on \(\mathbb{R}^2\), the weighted volume \(V_f=f\mathcal {L}^2\) is defined on the \(\mathcal {L}^2\)-measurable sets in \(\mathbb {R}^2\), and \(P_f(E)=\int_{\mathbb R^2}f d|D\chi_E|\in[0,\infty]\) is the \(f\)-weighted perimeter of a set \(E\) of finite perimeter in \(\mathbb {R}^2\), then the weighted isoperimetric problem is \(I_f(v)=\inf \{P_f(E)\}\), where \(V_f(E)=v>0\) and \(|Du|\) is the total variation of the distributional derivative \(|Du|\) of an integrable function \(u\) with bounded variation on \(\mathbb R^2\). If \(f\) takes the form \(f:\mathbb {R}^2\to (0,+\infty );\;x\mapsto e^{h(|x|)}\), where \(h:[0,+\infty )\to\mathbb{R}\) is a non-decreasing convex function, then \(h'_+(x,v)\) is the directional derivative of \(h\) in direction \(v\) for \(x\geq 0\) and \(v\geq 0\), and \(h'_-(x,v)\) is the directional derivative of \(h\) in direction \(v\) for \(x>0\) and \(v\leq 0\). Hence \(\rho_+=h'_+(\cdot,+1)\), \(\rho_-=h'_+(\cdot,-1)\), \(\rho=1/2(\rho_++\rho_-)\) on \((0,\infty)\), and the function \(h\) is locally of bounded variation and is differentiable with \(h'=\rho\). In this paper, the author studies minimizers for \(I_f\), finds the minimizer and shows its uniqueness. The author shows that the minimizer of \(I_f\) is a centered ball \(B\) in \(\mathbb{R}^2\) with \(V_f(B)=v\), and if \(R=\inf\{\rho>0\}\in[0,\infty)\), \(v_0=V(B(0,R)\), \(v>0\), and \(E\) is a minimizer for \(I_f\), then the following conditions are satisfied: (i)\, if \(v\leq v_0\), then \(E\) is equivalent to a ball \(B\) in \(\overline B(0,R)\) with \(V(B)=V(E)\), (ii)\, if \(v>v_0\), then \(E\) is equivalent to a centred ball \(B\) with \(V(B)=V(E)\).