Optimal configurations of finite sets in Riemannian 2-manifolds (Q5933885)

One observes in different instances that two-dimensional discrete point patterns, as a result of some optimality condition, tend to become regular hexagonal. In this paper, the author develops a tool for proving corresponding results in a number of cases. First, he extends L. Fejes Tóth's theorem on sums of moments, well known in discrete geometry, to a Riemannian 2-manifold \(M\), in the form of the following asymptotic result. Let \(f : [0,\infty)\to [0,\infty)\) satisfy a certain growth condition (holding, for example, for all positive powers), let \(J\) be a Jordan measurable set in \(M\) with \(\omega_M(J) > 0\) (\(\omega_M\) is the Riemannian area measure). For a finite point set \(S\subset M\), let \[ I(J,S):=\int_J \min_{p\in S}\{f(\rho_M(p,x))\}d\omega_M(x) \] (\(\rho_M\) is the Riemannian metric). Then \[ \inf \{I(J,S):S\subset M,\# S=n\}\sim\int_{H_n}f(\|x\|)dx\quad \text{as}\quad n\to\infty, \] where \(H_n\) is a regular hexagon in the Euclidean plane of area \(\omega_M(J)/n\) and centre 0. As in the classical case, this extension shows that, for problems of different types, regular hexagonal arrangements are optimal or almost optimal. Next, the author succeeds in proving the following converse. If \((S_n)\) is a sequence of sets in \(M\) with \(\# S_n = n\) and such that \[ I(J,S_n)\sim \inf\{I(J, S) : S\subset M, \#S = n\}\quad\text{as}\quad n\to\infty, \] then \((S_n)\) is uniformly distributed in \(J\) and (in a precise sense) asymptotically a regular hexagonal pattern of edge length \((2\omega_M(J)/\sqrt{3}n)^{1/2}\). The proof is complicated. Applications are made to the volume approximation of three-dimensional smooth convex bodies by polytopes with the number of facets tending to infinity, to the isoperimetric problem for polytopes in three-dimensional Minkowski spaces, and to optimal numerical integration formulae for Hölder classes of functions of two variables.