[gmx-users] 4xRE: COM motion removal (further clarification)

[Berk Hess]

> > The center of mass of the total system is a free degree of freedom.
> > Since no forces are working on it, its momentum should not change.
> > But do to integration errors it could. To avoid this annoying, though
> > unharmful, behavior one should use comm_mode=linear, ntscomm=1.
>
>To see this change in the z-coordinate of the COM I have to use 
>comm_mode=none
>or com_mode=linear with nstcomm>>1.
>
> > You can not remove the separate com motion of the surface and
> > the water as this would influence the process you want to study.
>
>I would imagine this is a situation in bulk simulation but what is the 
>situation
>in my case (frozen surface+droplet)? I have tried to remove COM motion of 
>the
>droplet and the whole system and the final results are pretty equal (also 
>equal
>to calculation with comm_mode=none but without lateral motion on surface).
>However when removing COM motion of water droplet I have to use much larger
>nstcomm (>2000) to see as quick spreading as in case of comm_mode=none or
>comm_mode=linear with nstcomm=100.

Again, never use nstcomm>1, it never makes sense (except if the comm removal
would really take a significant amount of cpu time.

When you freeze the surface you have an overall reference
and your surface would never drift, therefore you do not need
comm removal for the system.
You should not remove the comm for the droplet, as it should move
with respect to the surface.

Berk.