[lammps-users] periodicity and differences between timesteps

  in order to check a calculation of the heat flux J = d M / dt where M = \sum_a x_a e_a and e_a is the total energy at atom a and x_a the position at atom a I've run into something I can't figure out. The heat flux J and the time derivative of the moment M using a finite difference from one timestep to the next (for a very small timestep) is consistent for a non-periodic system. However, with a periodic system there is no agreement.

I've tried a number of things:
* translate the box: the results are invariant
* check that atom indices/labels at one timestep match those at the next : true for the system I am testing with
* change the system size : some times the errors are better than others
* "unwrapped" the coordinates using the function at the end of the email : no correspondence
* split the moment calculation into J = \sum v_a e_a + x_a (d e_a / dt) : the problem is in the energy rate term

none of these turned up what is going amiss. Any insight would be appreciated.
btw I am using "atom_style dpd"

I don't see any problem w/ what you're doing.
If you expect x*e to be invariant with PBC then
you'd need to unwrap the coords. If you track
it timestep to timestep is it OK on some steps?
PBC are only enforce on timesteps when reneighboring