[lammps-users] Box bounds


I have a question concerning the box bounds that are printed both to a dump file when using the atom style, and the bounds printed though thermodynamic output when using the lx, ly, or lz attributes.

The simulations I'm running should have the box constantly changing shape. I'm applying a ramped velocity profile to a solid to simulate stretching. The velocity is such that every 1000 timesteps the strain should increment by 0.0001 (the original lengths are around 150 A, so this corresponds to a change in length of 0.015 A every 1000 timesteps.) However, the box bound values look piecewise rather than continuous. For instance: in one simulation from timestep 57000 to 85000 the box length stayed flat at 155.59288 A, then jumped to 156.03231 A. In this case, this corresponds to a jump in strain from 0.0037 to 0.0065. How are the values for lx, ly, and lz calculated? Am I expecting precision that is impossible?


Joe Heidenreich

If you are using free surfaces and setting the velocities of atoms at
the surface, then
the atoms will move continuously. If you are using shrink-wrap
boundaries, then the
box is re-fit around the atoms on steps when reneighboring takes
place. The values
in the dump file will reflect whatever the box size is on the step the
snapshot is
written, but the atoms may be slightly inside or outside the box. So I wouldn't
worry about the exact box size, assuming you have non-periodic boundaries, since
the atoms are in the right position.