[gmx-users] Freeze Velocity?

[Erik Lindahl]

Hi Jon,

The main problem is that constant velocity corresponds to zero 
acceleration, i.e. no forces on the atoms, which is equivalent to 
excluding all interactions between it and the system.

If you just change the velocity after the update step you'll get 
inconsistent integration, but depending on your application you might 
be able to live with that.

It might be possible to imagine a case where the average force on the 
group is zero, but it is allowed to rotate - still, it's a bit strange 
conceptually.

In contrast, freezing just means we don't perform the update step, and 
acceleration groups means there are external forces.

If you still want to implement it, the best place is in update.c where 
we do the acceleration group stuff. Calculate the average force on a 
group and then subtract it.

This won't work if there are bonds between this group and other atoms, 
since the bond constraining step will adjust forces. I don't think 
there is any theoretical way you could fulfill both zero average force 
and constrained bonds with other atoms...

Cheers,

Erik


On Mar 21, 2005, at 5:43 PM, pappaj2 at rpi.edu wrote:

>
> I am trying to set the velocity for a group so that is does not vary 
> for
> the course of a simulation. This is analogous to the freeze command 
> with
> respect to atom position. I know of no way to do this in the existing
> framework of the .mdp file options. I do not have much experience 
> coding
> but I think it should be pretty easy to write this option into the 
> grompp
> file, though perhaps I am being naive. I am asking if you have a simple
> list of files  and locations where I need to change code, should I be
> altering the grompp file or is there some way to do it using user 
> defined
> options? I figure I can set the force on the group equal to zero at 
> each
> time step, or set the velocity equal to the specified velocity after 
> each
> time step. This latter way may avoid screwing up pressure calculations 
> and
> such.
>
> 				       Thanks,
> 					  Jon P.
>
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.