Re: [AMBER-Developers] [AMBER] Any issues if I make PMEMD understand bonds span imaging?

From: Jason Swails <>
Date: Wed, 6 Apr 2016 16:56:41 -0400

Moving this to the dev list where it's more appropriate

On Wed, Apr 6, 2016 at 4:49 PM, David Cerutti <> wrote:

> It's been asked before, but I've been contemplating my next career moves
> and a number of problems in materials science would be a lot easier to do
> with Amber if it could simulate systems in which bonded structures could
> extend infinitely. Having implemented the functionality in mdgx, I would
> say that, in principle, the only thing that needs to change in PMEMD /
> sander to have a bond traverse the box imaging would be to import the box
> dimensions, make appropriate copies of the coordinates, then transform and
> image the copies to compute forces and energies.
> Any objections if I take this on? Might be something nice to do early in a
> cycle, so there can be lots of testing before the next release. Can anyone
> suggest things I might break if I try this?

​collective variable restraints. It may also complicate
exclusion/exception rules.

​Are you going to change the whole bonded routines (bonds, angles, and
torsions) to use MIC displacements​ instead of raw displacements? Are you
going to introduce new valence lists that list *just* the terms that will
span the PBC?

If there's ever a case where a bond distance becomes large compared to the
box size (not common with bonds, but *may* be common with collective
variable restraints), then you start breaking the definition of a "chemical
bond" (i.e., two specific atoms bonded with each other), since the bond
definition will jump between an atom and one of its images.

I've thought about this before... GROMACS also does this, IIRC -- you may
look at how they handle it for inspiration.

All the best,
Jason M. Swails
AMBER-Developers mailing list
Received on Wed Apr 06 2016 - 14:00:05 PDT
Custom Search