[gmx-users] pme in gmx

[Erik Lindahl]

Hi.

> Hi, All:
> 	I have some basic questions about realization of pme in
> Gromacs3.0 especially the electrostatic part. I searched the mailing 
> list
> but I could not find specific answers so I raise them here. Any hints 
> will
> be sincerely welcome!
>
> 1) the way of pme realization in gmx: I assume that the do_fnbf()
> did the calculation of direct space inside the rlist/rcoul with the
> formula: 1/r - erf/r which was used in table function of the inner loop
> if you set electrostatic type to be PME; the do_pme() did the
> calculation of k space; the do_ewaldlrcorrection() exclude the bound
> interaction for direct space inside the rlist/rcoul.

That's correct. The correction routine also applies things like 
charge/dipole corrections, though.

> 2) I think there is no switch or shift function used in PME around the
> rlist/rcoul, is it?

  No, but PME is already switched of by design; the direct-space 
potential has decayed to pme_rtol relative to the non-switched value at 
the cutoff. If you need better conservation of energies, you can 
decrease this if you decrease the reciprocal grid spacing too.


>
> 3) Are there any difference in PME realization between different 
> versions
> of gmx, eg. 3.0.5, 3.1.4...?
>

No, not really. We improved the dipole corrections somewhere along the 
versions thanks to David Bostick, but that's a very small effect.

Cheers,

Erik