Re: [AMBER] force distribution analysis and Ewald

From: David Cerutti <>
Date: Tue, 3 Oct 2017 22:09:42 -0400

Ewald electrostatics gives you kqq/r for an infinite lattice of your
system. As such you must consider the short-ranged and long-ranged
components of the forces together--the short-ranged forces computed in
"direct space" are a correction that brings the reciprocal space forces
back to kqq/r for point charges (as Jason said, the reciprocal space forces
and energies are valid for Gaussian charges of the same magnitudes and
positions as the original point charges).

The reciprocal space sum IS pairwise decomposable--just not in a way that's
fast. If you want to compute the force that the primary image plus lattice
images of two particular particles in your system exert on one another,
you'll need to write a Matlab script to do the particle <==> mesh mapping
for each particle, take that FFT on the grid, calculate the direct space
erfc(r)/r interaction, and then add it all up. Or, I can give you my
routines for doing that.


On Tue, Oct 3, 2017 at 3:37 PM, Jason Swails <> wrote:

> > On Oct 3, 2017, at 2:57 PM, Thomas Pochapsky <>
> wrote:
> >
> > Hi, We are trying to do a force distribution analysis of our solvated
> > dynamics tracks (looking for pairwise forces that might help trace out
> > mechanical coupling pathways in the protein). We used PME for long
> > range interactions, is this likely to introduce errors when looking at
> > short range interactions such as adjacent pairwise interactions?
> Yes. The short-ranged interactions are damped artificially by a
> neutralizing Gaussian charge distribution. So it will impact short-range
> interactions.
> The effect of this neutralizing charge distributions is canceled in
> reciprocal space which is not pairwise decomposable. Ergo there's no
> straightforward way to pull the contribution of a particular pair
> interaction from the reciprocal space sum.
> > Philosophical suggestions as well as practical ones welcome.
> The amount of an interaction that is pushed into reciprocal space depends
> on Ewald parameters -- e.g. the charge distribution width alpha. The
> narrower the distribution, the faster the short range interactions are
> damped and more of it is pushed into reciprocal space.
> HTH,
> Jason
> --
> Jason M. Swails
> _______________________________________________
> AMBER mailing list
AMBER mailing list
Received on Tue Oct 03 2017 - 19:30:02 PDT
Custom Search