On packing spheres into containers (Q863698)

In this clear, precise and well-written paper, the author disproves Johannes Kepler's claim from 1611 [Progr. Linz. 30 S. \(8^\circ\) (1907; JFM 39.0076.09), Collection d'Histoire des Sciences (1975; Zbl 0409.01004)] that ``The (cubic or hexagonal close) packing [of spheres of the same size] will be the tightest possible, so that in no other arrangement more spheres could be packed into the same container.'' In more modern terms, Kepler's assertion is that \(n(C)\), the maximum cardinality of a packing of spheres of diameter~\(1\) contained in a convex body~\(C\subset\mathbb R^3\), is always attained by hexagonal close packings. The author disproves this for arbitrary smooth convex bodies~\(C\subset\mathbb R^3\), by showing that there exist arbitrarily small \(r>0\) such that the maximum cardinality of a packing of spheres of radius~\(r\) in~\(C\) is not attained by a face-centered cubic or hexagonal close packing set. This is deduced as a corollary from the following more general theorem: For \(d\geq2\), \(C\subset\mathbb R^d\) a smooth convex body, and a family \(\mathcal F\) of packing sets of ``limited complexity'', there exists \(n_0\in\mathbb N\), depending on \(\mathcal F\) and \(C\), such that \(\lambda(C,n)\) is not attained by any packing set in \(\mathcal F\) for any \(n\geq n_0\). Here \(\lambda(C,n)\) is the smallest real \(\lambda>0\) such that \(\lambda C\) contains a packing of \(n\)~balls of diameter~\(1\), and packing sets of ``limited complexity'' include all lattice packing sets. The author professes his belief that the theorem may also apply to some non-smooth containers, e.g., certain polytopes, and also conjectures that it fails for polytopes all of whose facets lie in planes containing hexagonal sublattices of the fcc lattice.