A reduction of Keller's conjecture (Q1819376)

The famous unsolved conjecture of O. H. Keller says that there exist two cubes in any n-dimensional cube tiling which share a complete (n-1)- dimensional face. \textit{G. Hajós}, Ĉasopis Pêst. Mat. Fys. 74, 157- 162 (1949), formulated this conjecture by means of factorisations of finite abelian groups as follows: If G is a finite abelian group (written additively) and \[ G= K + [g_ 1]_{q_ 1}+...+ [g_ n]_{q_ n} \] is one of its factorisations into ''simplices'' \[ [g_ j]_{q_ j}=\{0,g,2g,...,(q_ j-1)g\} \] and a subset K, then there is an index i such that \(1\leq i\leq n\) and \(q_ ig_ i=k-k'\) where k, k'\(\in K.\) In this paper the author considerably reduces the Keller's conjecture in his Theorem. Keller's conjecture is equivalent to the following statement: If G is an internal direct sum of cyclic groups generated by the elements \(g_ 1,g_ 2,...,g_ m\), respectively, each of them is of order 4, and if \(G= K+ [g_ 1]_ 2+...+ [g_ m]_ 2\) is a factorization, then always there exists a generator \(g_ i\) (1\(\leq i\leq m)\) such that \(2g_ i=k-k'\) where k, k'\(\in K.\) The proof improves the way of Hajós. It is based on geometric methods which enable to transform a cube tiling, giving counterexample for Keller's conjecture, onto such another one with periodic structure, leading to group in the theorem. In the reviewer's opinion this new formulation open a new direction in attacking the Keller's conjecture by combinatorial methods.