Output trajectory for dynamical calculation

Hi all,

I want to calculate MSD, ISF etc dynamical properties.

But I am confused about how to dump trajectory:

If I dump trajectory in a log frequency, using variable “logfreq” or a file containg dump timesteps, in this case, I can only select one snap as time origin. I am not sure if one time origin is sufficient?

And it would be better if anyone could tell me other ways which could choose many snaps as time origins, except for the way with equal interval between snaps.

Two comments:

  1. a number of analysis computations can be done by LAMMPS itself on-the-fly and in parallel while doing the simulation. That would save you the need to adjust your trajectory storage for that purpose.

  2. I have done a bit of post-processing analysis myself and written some tools/software for that purpose and never considered anything else but writing out evenly spaced trajectory data. Even for the case where you want to process overlapping chunks of data (which rarely makes sense these days considering how much faster and more parallel computers are these days compared to the times when trying to squeeze out as much information from a data set as possible was a necessity, e.g. when I was a grad student 20+ years ago) this gives you the most flexibility and options. If you want to process frames in non-linear fashion in your analysis, you can still do that.