From: Jason Swails (jason.swails_at_gmail.com)
Date: Wed Apr 16 2014 - 08:27:34 CDT
On Mon, Apr 14, 2014 at 5:25 PM, Jonathan Phillips <994phij_at_gmail.com>wrote:
> Hi,
> I'm a bit confused by the manual (looking at 2.9). Was wondering if
> anybody could help clear up my confusion.
>
> The manual says:
> "NAMD uses a non-bonded pair list which contained all pairs of atoms for
> which non-bonded interactions should be calculated."
> This implies that if you have pairlistdist set to anything less than the
> box size, true 'full electrostatics' will never be calculated.
>
No. And I think what you meant was if the cutoff was half the box size
(the cutoff will be the _radius_ of the sphere for which direct
interactions are computed).
> You'll calculate PME (if it's on) for every pair of atoms in the
> pairlist (and their images in other boxes), but nothing for the atoms which
> aren't in the pairlist.
>
No. See below for my comments about PME. What you'll be missing is the
force contributions from the van der Waals interactions outside the cutoff
(assuming NAMD uses a long-range dispersion correction, the energy and
virial will be more or less correct for "infinite" cutoffs subject to the
typical assumptions that particle density is the bulk density).
>
> But the default for pairlistdist is cutoff, and the manual says this about
> cutoff:
> "Next, consider the case where full electrostatics are calculated. In this
> case, the electrostatic interactions are not truncated at any distance. In
> this scheme, the cutoff parameter has a slightly different meaning for the
> electrostatic interactions — it represents the local interaction distance,
> or distance within which electrostatic pairs will be directly calculated
> every timestep."
> This suggests that you can get full electrostatics without setting a
> particularly large pairlistdist (or else they'd have mentioned pairlistdist
> in that section). Which then suggests that pairlistdist applies only to
> short range interactions.
>
Correct.
> Finally, the manual says:
> "PME stands for Particle Mesh Ewald and is an efficient full
> electrostatics method for use with periodic boundary conditions."
> This suggests that when PME is on, full electrostatics is used, and when
> PME is off, it is not used, but that's not explicitly stated.
>
Ewald sums are a way of computing long-range electrostatics. For an
infinitely periodic system, you cannot possibly do a brute-force
sum-over-pairs of all atoms in all periodic cells (there are an infinite
number of atoms). So you MUST make a choice for how you include long-range
electrostatics. Ewald sums work for periodic systems with neutral unit
cells and give you the exact long-range electrostatics for the infinitely
periodic system. (PME is a special case of Ewald that uses an
interpolating charge grid so the discrete FFT can be used instead of a
traditional Fourier transform.) Ewald sums work by splitting the
electrostatic contributions up into a short-range sum in direct space (a
sum over pairs of atoms in the pairlist) and a long-range sum in Fourier
space. Together, they recover the exact, full electrostatics of the
infinitely periodic system.
So are the pair lists referring specifically to short range interactions,
> and is my guess right about PME?
>
There are other ways of "accounting" for long-range electrostatic effects
(like reaction field and isotropic periodic sum), but using a cutoff (even
with switching or shifting functions) is not one of them.
Hope this helped,
Jason
-- Jason M. Swails BioMaPS, Rutgers University Postdoctoral Researcher
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:21 CST