Poisson polyhedra in high dimensions (Q2354363)

Random constructions and basic probabilistic reasoning often provide the existence of an object with desirable properties which is not accessible via a purely deterministic approach. A notoriously difficult problem, to which this fundamental observation may well apply, is the hyperplane conjecture. In one of several equivalent formulations, it asserts the existence of a universal constant \(c>0\) such that for any space dimension \(n\) and for any convex body \(K\subset {\mathbb R}^n\) of volume one, there is a hyperplane \(L\) in \({\mathbb R}^n\) such that the intersection \(K\cap L\) has \((n-1)\)-dimensional volume at least \(c\). The best lower bound for \(c\) known up to date is decreasing with the space dimension \(n\) and is of the order \(n^{-1/4}\), see [\textit{B. Klartag}, Geom. Funct. Anal. 16, No. 6, 1274--1290 (2006; Zbl 1113.52014)] and [\textit{B. Klartag} and \textit{G. Kozma}, Isr. J. Math. 170, 253--268 (2009; Zbl 1221.52010)]. The hyperplane conjecture is known to be true for special classes of convex bodies, see the survey articles [the second author, Lect. Notes Math. 2068, 205--238 (2013; Zbl 1275.60017)] and [the third author, in: New perspectives in stochastic geometry. Oxford: Oxford University Press. 45--76 (2010; Zbl 1202.60025)]. However, despite considerable effort over a period of now nearly 30 years, a general proof is still missing. Instead, Klartag's bound may even be a natural threshold. These insights have recently led to contributions which investigate possible counterexamples, with a special focus on randomly generated polytopes (see [Klartag and Kozma, loc. cit.]). The major object of investigation in the present paper is a parametric class of random polyhedra. For an arbitrary space dimension \(n\geq 2\) the authors define an isotropic Poisson hyperplane process in \({\mathbb R}^n\) which depends on a distance exponent \(r>0\) and an intensity \(\gamma\). This hyperplane process gives rise to a random hyperplane tessellation and thus to a system of random polyhedra, which are the cells of the tessellation. The cell containing the origin is called the zero cell of the tessellation, and is denoted by \(Z_0\). If for example \(r=n\), then the zero cell is equal in distribution to the typical cell of a classical Poisson-Voronoi tessellation, see [\textit{R. Schneider} and \textit{W. Weil}, Stochastic and integral geometry. Berlin: Springer (2008; Zbl 1175.60003)]. The authors consider the normalized zero cell, which is a re-scaled version of \(Z_0\) of unit volume. A special case of one of the main result of the paper is the following assertion. \textbf{Theorem 1.1.} Assume that \(r=b\, n^\alpha\) for some \(b>0\) and \(\alpha > 1/2\). Then, as the space dimension \(n\) tends to infinity, the probability that the hyperplane conjecture holds for the normalized zero cell tends to one. Theorem 1.1 is related to general investigations dealing with the combinatorial structure and the geometry of the zero cells \(Z_0\) obtained within the class of random tessellations considered in the present paper. The starting point is a set of identities connecting the number of \(l\)-dimensional faces with certain dual intrinsic volumes of \(Z_0\). In the special case \(r=1\), these identities reduce to a result of [\textit{R. Schneider}, Adv. Appl. Probab. 41, No. 3, 682--694 (2009; Zbl 1182.60008)] involving the well-known intrinsic volumes. The authors also investigate the expected measure of the \(l\)-skeleton. Evaluating these bounds as \(n\to \infty\) is the basis for the asymptotic results of the present paper. Another aspect of the paper deals with the question of how close the random polyhedra \(Z_0\) are to a Euclidean ball. In a last step, the authors determine the asymptotic behavior of \(Ef_l(Z_0)\), which is the mean number of \(l\)-dimensional faces of the zero cell \(Z_0\), for some fixed \(l\), as the space dimension \(n\) tends to infinity. The paper is organized as follows: After setting up the framework together with some background material in Section 2, basic identities are presented in Section 3.1. Special formulae for \(r=1\) are contained in Section 3.2, whereas Section 3.3 focuses on sections with subspaces and the transfer principle. The hyperplane conjecture for \(Z_0\) is discussed in Section 3.4, while Section 3.5 deals with the asymptotic behavior of the isoperimetric ratio and with the combinatorial structure of the zero cell in high dimensions. The detailed proofs of the main results are provided in the final Section 4.