Interpretation of Nonzero 14-LJ and 14-Coul Terms for Nonbonded Groups

GROMACS version: GROMACS/2023.3-foss-2022b-CUDA-12.1.1-PLUMED-2.9.2
GROMACS modification: No

My dynamics looks normal but there are nonzero interactions for 14-LJ and 14-Coul terms. It is my understanding that these terms should be exactly 0 for nonbonded molecules. My force field is OPLS with custom parameters for some ligands and cofactors. The dynamics looked normal, but these 14-LJ/Coul terms are very concerning to me. I’m using the pull code if that matters.

EDIT: I think its a bug with the energygrps reporting in GROMACS 2023 version, because I ran the same dynamics and analysis with the same topology on GROMACS 2021 and I get the expected 0s for all the intermolecular 14-LJ and 14-Coul energies.

Any help or further clarification would be greatly appreciated!

I don’t understand what you mean. 1-4 interactions are always intramolecular, not intermolecular. Nothing has changed in this regards between 2023 and 2025. Any molecules with at least three bonds in a row will have 1-4 interactions.

I think this may be related to this post I found: Report on non-zero intermolecular Coulomb-14/LJ-14 from Gromacs/2023-gpuvolta and gromacs/2023

In short, I think my version (or perhaps just the my particular installation) has a bug in reporting intermolecular 14 interactions, where it reports nonzero intermolecular interaction energies for 14-LJ and 14-Coul. On my same HPC, I have access to a 2021 version of GROMACS and I can repeat my workflow on the same topology/gro/mdp files without this issue in the 14 energy components, so I don’t think its a problem with the simulation setup.

I don’t understand how you know that this 1-4 term is intermolecular. There is no division in intra and intermolecular 1-4 energies.

Do you mean that the 1-4 energy term (a single term only) should be zero and is non-zero?

I mean that I use energygrps to calculate the interaction energy between nonbonded groups from my trajectory. I mean these 14 energy terms should be nonzero.

Now I’m even more confused. These 1-4 energy terms should be non-zero, but they are zero? I suppose you mean the other way around.

Anyhow, I found the issue causing this and it was fixed in release 2024.3: incorrect gmx energy decomposition for separate molecules (#5109) · Issues · GROMACS / GROMACS · GitLab
All your results are correct, except for the reporting of the 1-4 energies.

oh yes I mean they should be zero, sorry for the confusion. Good to know this energygrps reporting issue has been fixed in the 2024 version. I will notify my HPC staff so they can install the 2024 version - thank you!

Better install the lastest version: 2025.1.