[gmx-developers] position restraint algorithms

[Berk Hess]

David van der Spoel wrote:
> Hi,
>
> in the latest development code the position restraint routine is 
> treated differently from other bonded-force routines, because of 
> possible communication requirements. However, this is only necessary 
> when restraining the center of mass (COM) of a group of atoms, isn't 
> it? An alternative implementation could use the COM vsite and restrain 
> that. In that case the complexity is moved somewhere else, but posres 
> gets cleaner.
This is not related to restraining the COM of a group of atoms, that is 
impossible with position restraints.
It is related with pressure scaling.
>
> I'm asking because I would like to implement a position-dependent 
> force on the COM of a molecule. The force should in the end be 
> tabulated rather than analytical, and hence I can re-use most of the 
> mechanisms for tabulated bonds (except I need to add a one-particle 
> table type). Any suggestions?
I would say that this related more to the pull code, that is the only 
code that uses COMs.
Without a reference group you have absolute COM restraining.
Currently the options are umbrella (harmonic) and constant_force (which 
was incorrect
in mdp_opt.html, I just committed a fix). You could add a table option 
there.

Berk.