[gmx-users] where is Coul-LR?

[Mark Abraham]

On 10/11/2011 6:03 AM, Yun Shi wrote:
> So would it be reasonable to set rcoulomb = 2 or even 3 nm when 
> rerunning a trajectory? I am looking at a ligand-antibody system, and 
> I guess the long-range electrostatic interactions will not be small.

You can set rcoulomb to anything you like, but there's no reason to 
suppose the value of the computed Coulomb energy has any connection to a 
relevant observable. The force field was not designed to work at the 
cutoff values you suggest, and a longer cutoff is not necessarily more 
accurate. Neither was the force field parametrized to be able to be 
decomposed this way.

> A trick proposed by Nicolas in the mailing list during 2007 is to set 
> charges to 0.00 for everything except the atoms we are interested in. 
> This would make Coul-LR: = Coul.-recip., which is calculated by PME 
> (more accurate than just setting rcoulomb to 3nm?) and we can read 
> from .edr file. So I wonder if this method is theoretically sound?

There is no sense in which Coul-LR relates to Coul-recip. They are 
computed in fundamentally different ways, and affect Coul-SR differently 
too.

Mark