Re: AMBER: Confusion about Langevin dymamics

From: David Mobley <dmobley.gmail.com>
Date: Fri, 31 Aug 2007 09:05:13 -0700

Somewhat tangentially, I actually like using Langevin as a thermostat
(with weak friction) because it samples the correct distribution of
temperatures, and doesn't have some of the artifacts of thermostats
like Berendsen.

But as Dr. Case pointed out, any thermostat will muck with your
dynamics. This doesn't matter if what you're computing are
thermodynamic properties. But if you're trying to do kinetics this is
probably important.

David


On 8/31/07, David A. Case <case.scripps.edu> wrote:
> On Fri, Aug 31, 2007, Hayden Eastwood wrote:
> >
> > How can you be certain of molecular time scale events if the bulk water
> > being modelled over-estimates the viscosity of the surrounding medium?
> >
>
> You can't. If time scales are of interest to you (as they often are to me)
> you should not use Langevin or any other thermostat, since all temperature
> regulation interrupts Newtonian dynamics.
>
> A typical friction term for Langevin simulations (say of gamma_ln = 5) is only
> 10% of the true solvent friction. So, if you do use a Langevin thermostat
> with explicit solvent, the dynamics won't be all that wrong because of double
> counting. Izaguirre, Skeel, and co-workers haverecommended using Langevin
> coupling with even smaller friction constants, and you might want to study
> some of those papers, e.g.:
>
> %A J.A. Izaguirre
> %A D.P. Catarello
> %A J.M. Wozniak
> %A R.D. Skeel
> %T Langevin stabilization of molecular dynamics
> %J J. Chem. Phys.
> %V 114
> %P 2090-2098
> %D 2001
>
> [Aside: if you are interested in molecular time scales, you should also
> carefully consider which water model you use, since popular ones are often
> rather poor in modeling the viscosity of actual water.]
>
> ....dac
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Sep 02 2007 - 06:07:34 PDT
Custom Search