Intrinsic volumes of polyhedral cones: a combinatorial perspective (Q2408212): Difference between revisions

In this nicely written survey, the authors present a self-contained introduction to intrinsic volumes of convex cones and their applications in different branches of mathematics, mostly in combinatorics and the theory of hyperplane arrangements. Since the review should be rather concise, let me present subjectively selected parts of the survey. In preliminaries, the authors recall basic facts about convex cones and they define the \(k\)-th intrinsic volumes. For a polyhedral cone \(\mathcal{C} \subset \mathbb{R}^{d}\), we denote by \(\Pi_{C}\) the Euclidean projection, \[ \Pi_{\mathcal{C}}(x) = \text{argmin} \{|| x-y ||^{2}: y \in C\}. \] Now if \(F,G\) are two faces of \(\mathcal{C}\), we define \[ v_{F}(G) = \mathbb{P}\{ \Pi_{G}(g) \in \text{relint} \, F\}, \] where \(g \sim \mathcal{N}(\mathbb{R}^{d})\) is the standard Gaussian vector in \(\mathbb{R}^{d}\) and \(\text{relint} F\) denotes the relative interior of \(F\). Define the \(k\)-th intrinsic volumes of \(C\) with \(0 \leq k \leq d\) to be \[ v_{k}(\mathcal{C}) = \sum_{F \in \mathcal{F}_{k}(\mathcal{C})} v_{F}(\mathcal{C}), \] where the sum goes through all \(k\)-dimensional faces of \(\mathcal{C}\), denoted by \(\mathcal{F}_{k}(\mathcal{C})\). After recalling some natural properties of intrinsic volumes, the authors present the conic Steiner formulae which tells us that the \(d\)-dimensional measure of the \(\epsilon\)-neighbourhood of a convex body \(K \subset \mathbb{R}^{d}\) can be expressed as a polynomial in \(\epsilon\) of degree \(d\) with the intrinsic volumes as coefficients, namely \[ \text{vol}(K + \epsilon \mathbf{B}^{d}) = \sum_{i=0}^{d}V_{i}(K) \omega_{d-i}\epsilon^{d-i}, \] where \(\mathbf{B}^{d}\) denotes the unit ball, \(\omega_{d-i} = \text{vol}(\mathbf{B}^{d-i})\), and the \(V_{i}(K)\) are the Euclidean intrinsic volumes, and they provide a nice generalization of this result in the context of convex cones. Next, the authors present Sommerville's result which tells us that for any polyhedral cone \(\mathcal{C} \subseteq \mathbb{R}^{d}\) one has \[ v_{0}(\mathcal{C}) = \sum_{F \in \mathcal{F}(\mathcal{C})} (-1)^{\dim F} v_{0}(F). \] As a corollary, the authors show the Gauss-Bonnet formulae which tells us that for a polyhedral cone \(\mathcal{C}\) one has \[ \sum_{i=0}^{d} (-1)^{i}v_{i}(\mathcal{C}) = \sum_{i=0}^{d} (-1)^{i}f_{i}(\mathcal{C}) = \kappa(\mathcal{C}), \] where \(f_{i}(\mathcal{C})\) denotes the number of \(i\)-dimensional faces of \(\mathcal{C}\), and \(\kappa(\mathcal{C})\) is equal to either \((-1)^{\dim \mathcal{C}}\) if \(\mathcal{C}\) is a linear space or \(0\) otherwise. In Chapter 5, the authors discuss the Kinematic Formula in the setting of two polyhedral cones, and in Chapter 6 the authors present the Klivans-Swartz relation for hyperplane arrangements. If \(\mathcal{A}\) is a hyperplane arrangement in \(\mathbb{R}^{d}\), we denote by \(\chi_{\mathcal{A},j}(t) = \sum_{k=0}^{j}a_{jk}t^{k}\) the \(j\)th-level charateristic polynomial of the arrangement. Moreover, let us define by \(\mathcal{R}_{j}(\mathcal{A})\) the set of \(j\)-dimensional regions determined by \(\mathcal{A}\) and by \(r_{j}(\mathcal{A})\) we denote the cardinalty of \(\mathcal{R}_{j}(\mathcal{A})\). Finally, let us define the intrinsic volume generating polynomial by \[ P_{\mathcal{C}}(t) = \sum_{k=0}^{d} v_{k}(\mathcal{C}) t^{k}. \] Now we are ready to formulate the most interesting result of the survey (from my very subjective point of view). Theorem. Let \(\mathcal{A}\) be a hyperplane arrangement in \(\mathbb{R}^{d}\). Then for \(0 \leq j \leq d\) one has \[ \sum_{F \in \mathcal{R}_{j}(\mathcal{A})} P_{F}(t) = (-1)^{j}\chi_{\mathcal{A},j}(-t). \] In terms of the intrinsic volumes, for \(0 \leq k \leq j\) one has \[ \sum_{F \in \mathcal{R}_{j}(\mathcal{A})} v_{k}(F) = (-1)^{j-k}a_{jk}. \] In the section devoted to applications, the authors recover, for instance, the intrinsic volume generating polynomial in the case of generic hyperplane arrangements in \(\mathbb{R}^{d}\).