I was wondering if someone could clear up the difference between domain decomposition in GROMACS and LAMMPS. As I understand it LAMMPS creates ghost atom copies of atoms in surrounding domains and these are used to calculate the force on central domains. This allows LAMMPS to avoid using the minimum image convention and one can use a cutoff distance greater than half the box length. What is different about how GROMACS implements domain decomposition that means that it has to use the minimum image convention?