A sharp isoperimetric bound for convex bodies (Q2480559)

For a probability measure \(\mu\) on a convex body in \(\mathbb R^{n}\), one may define the notion of Minkowski surface area \(\mu^{+}\) of subsets of \(K\) and consider the corresponding isoperimetric problem. Extending results of \textit{M. Dyer} and \textit{A. Frieze} [Proc. Symp. Appl. Math. 44, 123--169 (1991; Zbl 0754.68052)] and \textit{L. Lovász} and \textit{M. Simonovits} [Random Struct. Algorithms 4, No.~4, 359--412 (1993; Zbl 0788.60087)], the author shows the following result: Let \(\mu\) be a log-concave probability measure on \(K\) and \(\| \cdot \|\) a norm on \(\mathbb R^{n}\). Then, for all measurable sets \(S\subseteq K\) with \(\mu (S)\leq 1/2\) holds (diam \(K\)) \(\mu^{+} (S)\geq\mu (S) G(1/\mu (S))\). Here diam is with respect to \(\|\cdot\|\), further \(\mu^{+}(S)\) is the Minkowski surface area with respect to \(\mu\) and \(\|\cdot\|\), and \(G(\cdot )\) is a function which is defined by \(G(2)=2\) and \(G(1/\mu (S))=\gamma^{2} e^{\gamma}/(e^{\gamma} (\gamma - 1)+1)\), where \(\gamma >0\) is the unique solution of the equation \(\mu (S)=e^{\gamma} (\gamma -1)/(e^{\gamma}-1)^{2}\) for \(\mu (S)<1/2\). The proof is based on tools developed by Lovász and Simonovits. The method applies also in the case where \(\mu\) is the uniform distribution on \(K\).