[EXTERNAL] issues with calculating diffusion coefficient when jumps

The only way to be confident is to convince yourself that your results do not depend on the details of the simulation setup. This will be true if you measure MSD only after the inserted hydrogen atoms have equilibrated. What this requires will depend strongly on the particular material structure that the hydrogen atoms are encountering. It is easy to construct materials where this is computational intractable, which might be the case here. That’s why it is a good idea to first reproduce simulations from previous publications, which by definition are tractable. But this only works if the publications are trustworthy. There are lots of tricks that can be used to help speed up equilibration. The best place to find them is in the detailed descriptions provided in high quality publications. This really goes beyond the scope of this mailing list.

Thanks for your reply and advice. I have been doing MD for a little while now but it can certainly become stressful when you have to ask the question of whether the publication you are using is the reference, or if there are issues in published results. If I may follow up with two short questions, directly related to MSD, that would be great. I won’t ask anything further as I know I am getting a little outside the scope.

  1. You say to equilibrate H after adding, which I agree with. Would a fix/nve limit or a minimize make sense? Or do you mean runnning some kind of NVT/NPT prior to doing the MSD. If I run the NVT/NPT it would still have some kind of large jump, I just wouldn’t have it included in my MSD calc. I assume the desire would be no big jump at all correct?

  2. This question is about MSD. If MSD is being done in several small intervals for example 50ps-200ps, 51ps-201ps, 52ps-202ps etc (essentially changing the initial reference point). to get an average from a large number of data points (this is the method used when MSD is infrequent jumps), then does it really matter if the atoms jump some large arbitrary distance in the beginning? Assuming I don’t count those data points, and it doesn’t explode out again like that, then why does that initial large ‘explosion’ to somewhere else matter from an MSD point of view?

Thanks again and like I said I won’t ask any more questions beyond these two.


See comments below: