OPLS potential types for LAMMPS input file from EMC

EMC
1

Hi all-

I’ve recently been trying to model some polymer systems using OPLS-AA in LAMMPS using EMC to build my initial structures. I had been trying to incorporate the new OPLS/2020 files which are reported in J. Phys. Chem. B 2024, 128, 1, 250–262, though that has proven to be a bit beyond my current skill level. (If anyone has had luck with this let me know)

My question is: when EMC spits out a .params file for LAMMPS from its build with OPLS, it chooses the harmonic, harmonic, harmonic, multi/harmonic, and harmonic styles for bonds, angles, dihedrals, and impropers, respectively. The bonds and angles I get, these match the form of OPLS. But for the dihedrals and impropers, the form specified in OPLS is a four-term cosine series,

image
image830×48 6.32 KB

for both types. The multi/harmonic dihedral potential in LAMMPS is given by
image
and the harmonic improper potential by
image
which are both different from the four-term cosine series. Is there a reason why EMC defaults to those potentials, and is it because they have been worked out to be mathematically equivalent to the original form? I feel like the harmonic improper one especially couldn’t really match the periodic cosine expression no matter what, unless there’s some math I’m forgetting.

In LAMMPS, it seems like it would make the most sense to use the dihedral_style/opls potential

image
image821×236 13.5 KB

and the improper_style/cvff potential
image
instead. Is there a way to get EMC to output these by default, or would this have to be changed manually?

Thank you so much,

Sam

2

Hi Sam,

EMC internally converts the OPLS style to a harmonic series, which means the two are equivalent. The choice for the improper style was made based on earlier papers. The more general form you provided transforms into a harmonic when developing a Taylor polynomial with d = -1 and n = 1, neglecting all higher than quadratic terms. Unfortunately, there is no EMC option to write out the forms as you provided.

3

Pieter,
Thank you for the reply - I realize shortly after posting that this was probably a Taylor expansion thing that I was missing. It makes sense. My main reason for asking is because in LAMMPS, the accelerator versions of certain bonded potentials have Intel-architecture optimizations (like the dihedral_style opls) while others do not (like the multi/harmonic). I would wager a guess, though, that in a large atomistic simulation, even with acceleration the dihedrals are not going to be the computational bottleneck that slows things down.
Sam