One potential solution is to handle this by post-processing the trajectory data. In that case, you can collect information for the average box dimensions and then scale the coordinates at each step appropriately or consistently add the same box length (something to that effect)-- but you’d have to decide if this is a more or less acceptable solution than doing what the compute is already doing.
Best,
Mike
Michael A. Webb
Miller Group
Department of Chemical Engineering
California Institute of Technology
mwebb@…1937… | (775) 343-8821