L-J and electrostatic force

GROMACS version: 2020.2-dev-20200430-5e78835-unknown
GROMACS modification: Yes/No
Here post your question

I would like to calculate the net electrostatic force and L-J force between two groups of atoms. I know about the option of md -rerun, but that seems to report the net force from all atoms on each atom. Here I only want the interaction between two specific groups of atoms. Although I can draft the calculation using the trajectories and L-J parameters in the .itp file, I wonder if there are readily available analysis tools so that I don’t have to reinvent the wheel?

I learnt that by adding energygrps in the mdp file and run gmx energy with the results will report the L-J or coulombic interaction energy between any two of the groups specified. However, in the online manual, it reads:

“group(s) for which to write to write short-ranged non-bonded potential energies to the energy file (not supported on GPUs)”

Does it mean the L-J and coulombic interaction evaluated is less accurate as the PME correction is not included? And since I usually use GPU to treat the PME, will this method sitll work?

Thank you!

Computing the forces between groups of atoms in a larger system is usually useless. In most cases one can not meaningfully decouple a group of atoms from its surroundings. A clear example is water. Water has a dielectric constant of 80. Thus two ions solvated in water will only feel an effective interaction that is 80 times smaller then they would feel in vacuum. Computing the force between only these two ions would give you the vacuum force, which is meaningless for the case of solvated ions.

That is why we do not have any tools to do this.

The energy groups are used for reporting energies, they have no effect on the computed forces.

I’m evaluating the interaction between two membranes. From shearing tests, I know the membrane-membrane interaction have qualitative difference for two different configurations and I am trying to make quantitative comparison by using metric such as force or energy, so if I can obtain either of these two quantities, it will give me some start on the analysis.

There are some water molecules between the membranes but the mean head group to head group distance across water is somewhere from 0.7-1.2 nm. This is why I believe electrostatic or the vdw are relevant. In addition, at this length scale, water cannot be treated straightforwardly as a continuum. In any case, as you said, by taking account of the dielectric constant, we can acknowledge the lower limit of the interaction (force or energy). Therefore I don’t think it’s overall useless.

In this case the direct electrostatic interactions between the membranes will still be about a factor 80 larger than than the effective interactions as screened by the water. The direct LJ interactions will all be attractive, but that doesn’t say much either, as this will not cause attraction between the membranes in solvent.

Anyhow, I think you can get what you want by doing 3 reruns: with membranes1+2, with membrane1 and membrane2. Then you can subtract the intra-membrane forces and get the forces between the membranes. You then need to sum the forces over all atoms.

Thanks for the instruction. Now I see how it should be done.

About LJ in solvent. In surface force balance experiments to measure interaction between mica surfaces in water, by allowing negatively charged mica surfaces approach each other, one will observe a initial repulsion at hundreds of nm separation due to osmotic pressure of the electrical double layers. However, the two surfaces will jump into contact starting at ~ 7A due to vdw forces. This is why we considering vdw forces to be nontrivial at subnanometer separation even in water.