Higher-Order Reverse Isoperimetric Inequalities for Log-concave Functions
Dylan Langharst, Francisco Marín Sola, Jacopo Ulivelli
TL;DR
This work develops the $m$-th order extension of key affine reverse inequalities from convex geometry to the realm of integrable log-concave functions. Central to the approach is the asymmetric LYZ body $igra f ig floor$, which enables a functional formulation of Zhang’s projection inequality in the $m$-order setting and yields a concise proof via covariogram–LYZ connections. The authors introduce and analyze functional radial mean bodies $R_p^m f$ and Ball bodies $K_p(g)$, including new regimes for $p o(-1,0]$ and $p oty$, and study their asymptotics, linking them to $ ext{vol}_{nm}(oldsymbol{ ext{Pi}}^{oldsymbol{ullet},m}igra f ig floor)$ and to $ ext{supp}(g_{f,m})=D^m( ext{supp}(f))$. They establish a functional, $m$-th order Rogers–Shephard inequality based on sup-convolution and the functional covariogram, providing sharp equality conditions tied to simplex-type structures in level sets. Overall, the paper broadens the affine-geometry toolkit for function spaces, connects Ball-type and radial-mean body concepts to the log-concave setting, and offers new structural insights with potential applications in high-dimensional convex-analytic problems and related functional inequalities.
Abstract
The Rogers-Shephard and Zhang's projection inequalities are two reverse, affine isoperimetric-type inequalities for convex bodies. Following a classical work by Schneider, both inequalities have been extended to the so-called $m$th-order setting. In this work, we establish the $m$th-order analogues for these inequalities in the setting of log-concave functions. Our proof of the functional Zhang's projection inequality employs properties of the asymmetric LYZ body, significantly streamlining the argument and producing a novel approach for the case $m=1$. Furthermore, we introduce and analyze the radial mean bodies of a log-concave function, thereby providing a functional generalization of Gardner and Zhang's radial mean bodies. These are new even in the case $m=1$. Our development leverages an extension of Ball bodies, which may be of independent interest.
