Instability in harmonic restraints when equilibrating large systems

From: Tristan Croll (tristan.croll_at_qut.edu.au)
Date: Fri Jul 24 2015 - 16:52:05 CDT

I have a tendency to use the constraints function in NAMD rather than fixed atoms when initially soaking the water into a new system, to give the protein a little bit of freedom to "breathe". Like many, I also often run into the message, "FATAL ERROR: Periodic cell has become too small for original patch grid!" when starting up very large systems. Through trial and error I'd found that temporarily increasing langevinPistonPeriod to 500 or so usually fixed the issue, but this morning I decided to watch more closely what was going on using IMD Connect.

In this case, at least, the culprit was clearly a runaway oscillation in the harmonic restraints. In the first 20 or so steps I saw the protein begin to make larger and larger wholesale jumps in position, in alternate directions with each step. In the last step before the simulation crashed, the portion I was watching jumped almost entirely off screen. Simply adding:

constraintScaling 0.5

to the config file allowed the simulation to run smoothly with langevinPistonPeriod at 250 (but it still crashed at 200).

Upshot: it might be a good idea to increase the damping.

Cheers,

Tristan

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:15 CST