[gmx-users] COM removal

[Erik Lindahl]

Hi,

> When I do not apply COM removal to the system, the system gains a net 
> momentum due to truncation errors and the box starts drifting. To 
> prevent this from happening, COM removal can be applied to the entire 
> system. This does some sort of rescaling to prevent the box from 
> moving.

Correct. It's not really a scaling, though: you just remove the net 
velocity of the system.


> However, when I start applying COM removal to individual groups, 
> doesn't this affect the true dynamics  of the system? If I need to 
> calculate diffusion coefficients of  a particular group (Say group 
> AAA)  then I can still do this calculation if Comm-grps = system was 
> chosen, but not when comm-grps = AAA .
>
> Am I correct ?

Yes, partly. The whole idea with COM-removal is to compensate for 
non-physical artefacts in simulations.

The most obviuous problem is that in contrast to the real world, 
nothing stops your box from moving in space.

>
> I need to calculate diffusion coefficients of some peptides at 
> lipid/water interfaces, but have used COMM removal on the peptides, 
> lipids and SOL separately rather than the whole system. I think I need 
> to rerun these simulations. Don't I?
>

For lipid bilayers it's a good idea to remove COM separately for the 
water, upper layer and lower layer, since the lipid layers can show 
non-physical diffusion against each other in the XY-plane.

If you apply COM to the peptide, it won't be able to move relative to 
the water or lipids.

Cheers,

Erik