Re: [AMBER] limit exceeded and high temperature

From: Jason Swails <jason.swails.gmail.com>
Date: Fri, 11 Sep 2015 11:26:17 -0400

On Fri, Sep 11, 2015 at 10:52 AM, Ruth Helena Tichauer <rhtichau.laas.fr>
wrote:

> Hi Amber community,
>
> I’m running a molecular dynamics simulation of a protein with its ligand
> and an Mg +2 ion in implicit solvent although there are tree water
> molecules in the active region as they are necessary for the reaction to
> proceed.
>
> I’m modeling quantum mechanically (QMMM) a few residues with the PME
> method, as well as the Mg ion, the tree water molecules and the ligand
> (which contains phosphates).
>
> The structure has been successfully quantum minimised prior to the MD/QMMM
> run and in the later (which, for now, nstlim=1000 only) the initial
> temperature is 0°K and the final one is 100°K (tempi=0.0, temp0=100.0)
>
> I started with dt=0.002 and no shake options. I got an error message which
> said: “vlimit exceeded for step..”
> and also: "This is usually a a symptom of some deeper problem with the
> energetics of the system”
>

​A timestep of 2 fs is about the upper limit of what you can achieve using
SHAKE and *no* QM and still have stable integration. It is much too large
with SHAKE and QM.



>
> Searching on the amber archives, I found that I should decrease dt and add
> some shake options (ntc=2 and nff=2) in the input file. Yet, these changes
> (dt=0.001 and even dt=0.0005) didn’t help. Well, by diminishing dt without
> adding the shake options, the simulation reaches the final step but the
> temperature and the total energy are incredibly (unrealistically) high.
>

​This is caused by very high forces, most likely. Perhaps the structure
was not minimized well enough (the minimization output prints out the max
gradient -- you should check that this is decreasing steadily during the
minimization and reaches a reasonably small value at the end).

Also, try a simulation where you print out every step and see if the
trajectory or output file gives any clues as to what is going wrong.

HTH,
Jason

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 11 2015 - 08:30:04 PDT
Custom Search