[gmx-users] problem with free energy calculations

Berk Hess gmx3 at hotmail.com
Thu Feb 8 10:58:37 CET 2007




>From: chris.neale at utoronto.ca
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] problem with free energy calculations
>Date: Thu,  8 Feb 2007 02:51:34 -0500
>
>Hi David Mobley,
>
>could you please clarify your "Nose-Hoover defect" comment or ref it 
>please?
>Also, somehow I missed it.. what others do you agree with? Are you  
>referring to comment of berk Hess? If so, then Berk: are you  suggesting 
>that multiple tc_grps here is inappropriate? Why so?

I assume David was refering to my comment.

With any global thermostat you have to worry about ergodicity.
When all degrees of freedom are coupled "strong enough",
this in not an issue. When there are nearly decoupled or completely
decoupled degrees of freedom one global thermostat will
not give you the correct ensemble.
Using two thermostats, one for the decoupled molecule and one
for the rest, would not solve the problem, as this leaves the degrees
of freeom between the molecule and the rest uncoupled.
Gromacs currently sets the velocities of these degrees of freedom
to zero. This is appropriate for instance a system consisting of
a water and an octane slab, but not for free-energy decoupling
a molecule.

A solution might be to add a separate thermostat for the mentioned
degrees of freedom, but also there one might run into problems
with ergodicity.

The simplest solution is using a local thermostat that works locally
on each degree of freedom. In Gromacs this can be done with
the integrators sd and bd. But one should not use bd as it slows
down the sampling.

Berk.

_________________________________________________________________
Live Search, for accurate results! http://www.live.nl




More information about the gromacs.org_gmx-users mailing list