Re: [AMBER] periodic box drifted during NVE run

From: Quynh Vo <>
Date: Thu, 3 Jul 2014 13:19:34 -0700

> It's not clear what kind of fluid you have, or how you know what the "expected

> value" is. 

Thank you again for your help. My molecule is a small surfactant with a phosphate group and some short alkyl tails. The expected value is the reported self diffusion coefficient (SDC) in literature for this molecule.

> I'd suggest doing a simulation water, where there is a large
> literature on the self-diffusion coefficients for various force-field models.
> Make sure you can reproduce those.

This is the first thing I did before starting simulation my own molecule. I was able to get the expected SDC of TIP3 water model which is about 5.2 (higher than experimental value)

> It's actually unusual to see significant box drift: are you sure you don't
> have the "Flying ice cube" problem? 
I was really surprised to see the box drifted like that too. Before using the center command in cpptraj, the MSD profile of the molecules follows a quadratic trendline so no meaningful SDC could be calculated there. When I used the center command on all molecules (* mask), the MSD is linear but gave a value about 7 times less than experimental value (same thing happened when I used the rst file as the reference). When I used center command on the first molecule of the system, the system is shifted from the original periodic box. I guess this is expected because the position of the first molecule wasn't at the center of the box to begin with. However, the MSD profile is very choppy and doesn't show a clear linear trendline (I have 18ns of simulations: 6 NPT and 12 NVE). So I think the center command is clearly affecting the SDC in a way. I just don't know what would be the best approach for this problem.

Additionally, I am not sure that I do not have the "flying ice cube" problem. But I thought "flying ice cube" only happens with bad thermostat. I used NTT=3 for my NPT runs. I compared KE, PE, and Etotal at the end of NPT runs with NVE runs and those values are conserved. Temperature was well equilibrated and average around 297K for NVE runs. I visualized some random molecules in NVE and they seem to have frequency energy (bonds and angles do vary). Is there anything else I can look for to determine if I have the "flying ice cube" effect?

Thank you again for your help.
Quynh Vo
AMBER mailing list
Received on Thu Jul 03 2014 - 13:30:02 PDT
Custom Search