[gmx-users] Reducing fast motion with Langevin Dynamics

[Berk Hess]

>From: Bob Johnson <robertjo at physics.upenn.edu>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: GROMACS Mailing List <gmx-users at gromacs.org>
>Subject: [gmx-users] Reducing fast motion with Langevin Dynamics
>Date: Tue, 31 Jan 2006 15:18:26 -0500
>
>Hello everyone,
>I am simulating the adsorption of single stranded DNA on a carbon nanotube 
>under
>aqueous conditions. I would like to reduce the fast motions in my system so 
>that
>I can increase my timestep. Is there a way to do this with Langevin 
>Dynamics? I
>would like to keep explicit solvent so I would have no use for the random 
>force
>that is included in Langevin Dynamics. All I would like is to have a 
>friction
>term that reduces fast motion in my system.
>Thanks,
>Bob Johnson

Langevin Dynamics is a too general term.
Do you mean stochastic dynamics or Brownian dynamics?

SD is identical to plain MD for not too large friction, so no improvement 
there.

BD or SD with large friction might allow for larger timesteps,
but will make all dynamics slower, so also that is not an option.

In general integrators can never reduce fast motions, as they only 
integrate.
What will help is removing the fastest motions in the system,
which are angular vibrations of hydrogens, by constraining angles
and/or turning hydrogens into virtual sites.
This can be done with the -vsite option of pdb2gmx,
but I don't know if you can use pdb2gmx to generate your DNA topology.

Berk.