Hi, Axel,
Thanks for your quick reply, there are only small difference in the intensity. But when i use the different random number (seed) to run the exact same simulation with the same number of cores, the intensities are also not converged, is that also because of the “statistic uncertainty”? or something wrong with my specification in the input script?
Best Regards
Jiasen Guo
Hi, Axel,
Thanks for your quick reply, there are only small difference in the
intensity. But when i use the different random number (seed) to run the
exact same simulation with the same number of cores, the intensities are
also not converged, is that also because of the "statistic uncertainty"? or
something wrong with my specification in the input script?
have you spent some time reading up on the issues that i mentioned?
if you properly understand what is going on, you should be able to
answer this question yourself.
remember that in your paper/thesis you cannot write: "this is
acceptable because some random dude on the internet told me so".
axel.
BTW: there are plenty of publications and text book material on
estimating/quantifying statistical errors in properties derived from
molecular dynamics simulations.
nudge, nudge. wink, wink. know what i mean? 
Axel’s point is that if you for a long time on different # of cores,
or with a different random seed, the trajectories of the 2 simulations
will diverge. For equilibrium simulations it’s a statistical difference,
but it can affect all kinds of things you measure from MD snapshots.
See the beginning of Section 12 in the LAMMPS manual for a brief
overview, or better look in the literature as Axel suggested.
Steve