Overcoming the minimum image constraint using the closest point search.

Finding the set of nearest images of a point in a simulation cell with periodic (torus) boundary conditions is of central importance for molecular dynamics algorithms. To compute all pairwise distances closer than a given cutoff in linear time requires region-based neighbor-listing algorithms. Available algorithms encounter increasing difficulties when the cutoff distance exceeds half the shortest cell length. This work provides details on two ways to directly and efficiently generate region-region interaction lists in n-dimensional space, free from the minimum image restriction. The solution is based on a refined version of existing algorithms solving the closest vector problem. A self-contained discussion of lattice reduction methods for efficient higher-dimensional searches is also provided. In the MD setting, these reduction criteria provide useful guidelines for lattice compaction.