On weighted covering numbers and the Levi-Hadwiger conjecture (Q891149)

In this paper, the authors introduce a new concept of weighted covering and separation numbers and explore various properties of these quantities. Covering and separation numbers are frequently used tools in diverse mathematical areas that range from analysis and combinatorics to convexity and probability. For two compact sets \(T,K\subset{\mathbb R}^n\) with \(T\) having nonempty interior, the (classical) covering number \(N(K,T)\) is the smallest number of translates of \(T\) whose union covers \(K\), while the separation number \(M(K,T)\) is the maximum number of points in \(K\) with the property that translates of \(T\) with these points are pairwise non-overlapping. The authors have already defined weighted generalizations of the covering and separation numbers in their earlier paper [the first author and \textit{O. Raz}, Adv. Math. 227, No. 1, 730--744 (2011; Zbl 1221.52026)]. Here, they further generalize these notions as follows. Let \(\mu\) be a non-negative Borel regular measure in \({\mathbb R}^n\). Then, \(\mu\) is a covering measure of \(K\) by \(T\) if for the indicator functions \({\mathbf 1}_K\) and \({\mathbf 1}_T\) it holds that \(\mu * {\mathbf 1}_T\geq {\mathbf 1}_K\), where \(*\) denotes convolution. The weighted covering number \(N^*(K,T)\) is the infimum of \(\int_{{\mathbb R}^n}d\mu\) taken over all covering measures. The weighted separation number \(M^*(K,T)\) is defined analogously. The main results of the paper include a relationship between weighted separation and covering numbers which the authors call strong duality. They also prove volume bounds for weighted separation and covering numbers and explore the connection with the classical versions of these quantities. It is notable that some of these statements hold under very general conditions. The classical Levi-Hadwiger problem seeks the minimum number of translates of smaller homothetic copies of a convex body \(K\) that can cover \(K\). According to the conjecture, this minimum is always at most \(2^n\), and equality holds only for parallelotopes. The authors prove an upper bound for the Levi-Hadwiger covering number for centrally symmetric convex bodies that is asymptotically the same as the currently best known upper bound due to Rogers. Furthermore, they introduce a weighted version of the Levi-Hadwiger covering problem using weighted covering numbers (for non-negative discrete covering measures), and solve the case when the convex body is centrally symmetric. They also characterize the equality case. Their result resolves the case of equality in the fractional illumination problem [\textit{M. Naszódi}, Contrib. Discrete Math. 4, No. 2, 83--88 (2009; Zbl 1189.52005)] for centrally symmetric convex bodies.