[gmx-users] PME question

[Erik Lindahl]

Hi,

> 
> 1) what is the relevance of rcoulomb if there really is no cut-off with
> PME? and what must it equal rlist?

It defines the part of the interactions treated in direct space. Since 
PME changes the shape of the coulombic interaction significantly (it 
will be zero when we reach the cutoff) the reciprocal part should be
evaluated every timestep, and then it doesn't make any sense to have
a long-range part of direct-space coulomb interactions updated every
10 steps or so.

You can still have rvdw>rlist.


> 
> 2) must the system be net-zero charge to use PME? if this is not a
> requirement, how does gromacs avoid infinite terms from repeatedly summing
> the excess charge?
> 

We apply the charge correction term, so this shouldn't be a problem. 
Moving charges that cross box boundaries can still cause strange 
effects, but that's inherent in a lattice summation I guess...

Cheers,

Erik