Densest packings of translates of strings and layers of balls

Abstract

Let $L \subset {\Bbb R}^3$ be the union of unit balls, whose centres lie on the $z$-axis, and are equidistant with distance $2d \in [2, 2\sqrt{2}]$. Then a packing of unit balls in ${\Bbb R}^3$ consisting of translates of $L$ has a density at most $\pi /(3d\sqrt{3-d^2})$, with equality for a certain lattice packing of unit balls. Let $L \subset {\Bbb R}^4$ be the union of unit balls, whose centres lie on the $x_3x_4$ coordinate plane, and form either a square lattice or a regular triangular lattice, of edge length $2$. Then a packing of unit balls in ${\Bbb R}^4$ consisting of translates of $L$ has a density at most $\pi ^2/16$, with equality for the densest lattice packing of unit balls in ${\Bbb R}^4$. This is the first class of non-lattice packings of unit balls in ${\Bbb R}^4$, for which this conjectured upper bound for the packing density of balls is proved. Our main tool for the proof is a theorem on $(r,R)$-systems in ${\Bbb R}^2$. If $R/r \le 2 \sqrt{2}$, then the Delone triangulation associated to this $(r,R)$-system has the following property. The average area of a Delone triangle is at least $\min \{ V_0, 2r^2 \} $, where $V_0$ is the infimum of the areas of the non-obtuse Delone triangles. This general theorem has applications also in other problems about packings: namely for $2r^2 \ge V_0$ it is sufficient to deal only with the non-obtuse Delone triangles, which is in general a much easier task. Still we give a proof of an unpublished theorem of L. Fejes T\'oth and E (=J.) Sz\'ekely: for the $2$-dimensional analogue of our question about equidistant strings of unit balls, we determine the densest packing of translates of an equidistant string of unit circles with distance $2d$, for the first non-trivial interval $2d \in (2{\sqrt{3}},4)$.