[gmx-users] NVE problems

[istvan at kolossvary.hu]

Hi, 

I posted a message right before Christmas (gmx-users digest, Vol 1 #1166) 
about problems with NVE. I have a feeling that the nsgrid error that I 
mentioned is not specific to the problem, it just shows a general numerical 
instability due to the NVE simulation. In other cases Gromacs blows up with 
other numerical errors. In any case, I guess the simplest way to present the 
problem is to ask why should Gromacs blow up on alanine dipeptide in a 3x3x3 
nm box using NVE instead of NPT? I can hardly think of anything simpler. The 
system was minimized, PR-equilibrated (FF was OPLS-AA), but then a fully 
unconstrained run with PME that would normally run without problems in NPT, 
blows up after only 1 ns using a 1 fs time step when I turn off temperature 
and pressure control. Shouldn't a 1 fs time step be adequate for NVE without 
LINCS/SHAKE? Can anyone experienced with NVE in Gromacs help me? 

Many thanks, 

  Istvan Kolossvary