[gmx-users] Question about integrator md-vv and associated documentation

James james at ryley.com
Tue Sep 18 23:27:20 CEST 2018


There is a long-standing typo in the description of md-vv. It says:

"*The* the kinetic energy, which is determined from the whole step
velocities and is therefore slightly too high."

Should that be "*There *the kinetic energy..."?

The docs then go on to say:

"The advantage of this integrator is more accurate, reversible Nose-Hoover
and Parrinello-Rahman coupling integration based on Trotter expansion, as
well as (*slightly too small*) full step velocity output."

Can anyone quantify "slightly too small"? I am using double precision
gromacs with Verlet cutoffs and the md-vv integrator, on 5.0.4, no
thermostat. I am trying to watch potential energy be converted into
friction/temperature. I am seeing temperature go down slightly on runs
where I would expect it to go up slightly, and wondering if the "slightly
too small" velocity output is the reason. (I cannot use md-vv-avek -- I
assume it is not implemented for Verlet cutoffs, at least in 5.0.4).

I know most people don't need this level of accuracy, but for my models,
I've compared single and double precision, and md versus md-vv, and double
precision with md-vv substantially reduces total energy drift. I'm just
trying to figure out if I can improve it further because it is still a
little off, and the drift is always negative (again, going back to whether
this is because of the "slightly too small" issue).

James Ryley

More information about the gromacs.org_gmx-users mailing list