Re: [AMBER] Born Mayer Potential

From: Jason Swails <>
Date: Tue, 12 Jul 2016 22:28:25 -0400

On Tue, Jul 12, 2016 at 9:37 PM, David A Case <>

> On Tue, Jul 12, 2016, Vivien Ramothe wrote:
> > The parameters that have been calculated for dispersion potential are
> made
> > to be
> > used with the Born Mayer potential: V(r)=f6*(C6/r6)+f8/(C8/r8),
> > the f6 and f8 function are Tang Toennis damping functions.
> > I'm reading the user manual since quite a few days now,
> > even if if didn't see anything about this kind of potential in it.
> That is correct: this sort of function is not implemented.
> > It seems i found a way to implement it myself by using the minimize
> option,
> > of parmed command. but i'm not sure about it since i never used it one
> my
> > own.
> You would not do this with parmed, but would need to modify the source
> code in sander or pmemd. There *is* support for for 12-6-4 potential, so
> you
> could probably use that as a template to construct a 6-8 potential.

​ParmEd has no knowledge of this 6-8 potential, so the "minimize" option
won't use the force field you want.

> > Could you tell me please if it is possible to implement my own kind of
> > potential?
> It's certainly *possible*, but you would have to be willing to do
> programming
> yourself. OpenMM might be a better option here, since you can define new
> types of potentials at a fairly high (Python) level.

To expand very slightly, what OpenMM would allow you to do here is to
specify just the energy expression for your nonbonded potential, and would
then generate optimized CPU and GPU kernels with analytical derivatives,
eliminating the need to debug those particular calculations (of course you
still need to validate that you implemented the potential correctly, but
you can rely on OpenMM to get all of the details of computing the atomic
interactions with forces based on your energy expression correctly).

OpenMM also has better support for virtual sites than Amber does (Amber
only supports a small number of common virtual site arrangements).​

> > How are the long range correction calculated on a potential that AMBER
> > doesn't konw?
> Amber has long-range corrections for dispersion (an r^-6 potential), and
> you
> can use that as a template to implement something similar for a 6-8
> function.
> Again, programming is required. (Not sure what OpenMM can do in this
> regard.)

OpenMM integrates the energy expression you provide analytically to derive
the long-range correction, provided you turn the long-range correction
*on*. OpenMM will complain if you feed it a divergent or conditionally
convergent energy expression and ask for the long-range correction, though


Jason M. Swails
AMBER mailing list
Received on Tue Jul 12 2016 - 19:30:02 PDT
Custom Search