|From:||Thomas Zhang <tk.zhang@...3421...>|
|Date:||Fri, 15 Dec 2017 22:56:13 +0000|
Thanks for the replies! Sorry, 1 small correction, the timestep I used was 0.25fs. The times should be divided by half.
I have tried adjusting the timestep to half (0.125fs) during the NEMD step for a smaller system. The two attached files are both equilibrated at 0.25fs timestep, but one is ran with NEMD using 0.125fs, and one is using 0.25fs. Interestingly, I get even more drift. It could be also dependent on the specific velocity configuration. I have seen the similar behaviour with a separate series of repeated runs, only changing the inital velocity seed. Some systems had drift, while others had less.
From: Benjamin Jensen <bdjensen@...1754...>
Sent: Thursday, December 14, 2017 7:51:03 AM
Cc: Thomas Zhang
Subject: Re: [lammps-users] Reaxff c/h/o energy conservation and temperature drift in NVE
The Chenoweth CHO parameters have some pretty steep gradients, and usually requires smaller than typical time steps for energy conservation. I wouldn't be surprised if 0.5 fs was still to large. I'd rerun with decreasing time steps and look for convergence.
On Wed, Dec 13, 2017 at 5:08 PM, Ray Shan <rshan@...1795...> wrote: