[gmx-users] Re: oscillations in constrain forces

[Berk Hess]

>From: fangel at usc.es
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] Re: oscillations in constrain forces
>Date: Tue, 22 Jun 2004 09:22:28 +0200
>
>Hi Berk,
>what means long time averages? between 0 and 5 ns the oscillations doesn't
>decrease and the average value seems to be zero, in fact those oscillations
>look like noise.. could be that the force I want to measure is too small?

The oscillations should not decrease and they should look like noise.
In MD molecules continuously collide which you would see as noise
in the constraint force.
What is long depends on the size of your molecules. 5 ns is usually
about the minimum that you would need.

You should make sure the normal constraints do not pose
problems. I guess the default accuracies might not be enough,
to be certain you should do a run with:
shake_tol = 1e-6
when you use shake, or
lincs_order=8
lincs_iter=2
when you use lincs.

I have recently converted the pull code to always use double
precision internally. But just to be sure you can do a simulation
with a double precision compiled mdrun.

Berk.

_________________________________________________________________
MSN Search, for accurate results! http://search.msn.nl