Cube-tilings of \(\mathbb{R}^ n\) and nonlinear codes (Q1329187)

Let \(K_ n\) denote the largest integer such that every tiling of \(\mathbb{R}^ n\) by unit cubes contains two cubes that have a common face of dimension \(K_ n\). This number is connected with a famous conjecture (Keller's conjecture) that \(K_ n = n - 1\) for all dimensions \(n\), which has previously been disproved by the authors [Bull. Am. Math. Soc., New Ser. 27, No. 2, 279-283 (1992; Zbl 0759.52013)] for \(n \geq 10\). (It is true for \(n \leq 6\), and unsettled for \(n = 7,8,9\).) In the paper under review they refine this result and prove \(K_ n \leq n - \sqrt{n/3}\) for all \(n\). They use coding theory terminology to describe their result in order to emphasize the close relation with nonlinear codes over \(GF(2)\), but they do not use results from the theory. Two more results conclude the paper: If, for some dimension, there is a tiling with certain special properties, then a linear bound on \(K_ n\) would follow. However, they also formulate a conjecture which, if true, would imply that such a special tiling cannot exist. And finally, an explicit example proves \(K_{10} \leq 7\).