Large finite lattice packings (Q5946245)

Let \(K\) be a centrally symmetric convex body in \(\mathbb{R}^d\). A finite lattice packing of \(K\) is a family \(C+K\), where \(C\) is a finite subset of a packing lattice of \(K\). Due to J. M. Wills, the parametric density, for a parameter \(\rho>0\), of the packing is defined by \[ \delta (C;K,\rho) ={n\cdot V(K)\over V(\text{conv} C+\rho K)} \] with \(n=\text{card} C\). (Note that for \(\rho=1\) this is the density inside the convex hull of the corresponding translates.) Writing \(C^n\) for the optimal packing if \(n\) is large, \(\text{conv} C^n\) is a segment if \(\rho< \rho_S (>0)\), where \(\rho_S\) denotes the sausage radius, and the inradius \(r(\text{conv} C^n)\) tends to infinity with \(n\) if \(\rho> \rho_C\), with \(\rho_C (\geq \rho_S)\) as critical radius. The authors show that if \(\rho>\rho_C\) in \(\mathbb{R}^d\), then the shape of \(\text{conv} C^n\) is not too far from that of a ball, and if \(r(\text{conv} C^n)\) is bounded but the radius of the largest contained \((d-2)\)-ball tends to infinity, then eventually \(C^n\) is contained in some hyperplane or \((d-2)\)-plane and its shape is not too far from that of a \((d-1)\)-ball or \((d-2)\)-ball, respectively. From this they deduce further interesting results on three-dimensional finite lattice packings, such as the following: if \(\rho_S<\rho < \rho_C\), then \(\text{conv} C^n\) is eventually planar with a shape not too far from that of a disc. If \(K\) is in particular a 3-ball, then \(\rho_S= \rho_C\), verifying the strong sausage conjecture.