Hi everyone,

I have calculated the interaction energy for a system with two single atoms and was quite surprised how much the Coulomb interaction differs when using periodic boundary conditions (PBC), Particle Mesh Ewald (PME) or no periodic boundary conditions.

I calculated the interaction energy for a system with 2 particles (0.4nm distance, q1=-1e, q2= 0.0842e) for the following three situations:

1. PBC with PME

2. PBC withou PME

3. no PBC (not possible with the latest version of gromacs, so one need to travel back in versions)

The short range Coulomb interactions (Coulomb (SR) ) and the Coul-SR:a_1-a_2 differed for both situations with PBC. Only for a system with no PBC, the analytical (-29.2468 kJ/mol) and simulated result were the same and Coulomb (SR) = Coul-SR:a_1-a_2.

Is this expected? And if yes, does anyone have a good intuition to explain this?

I used the following mdp parameters:

• PBC with PME (Coulomb (SR) = -209.594 and Coul-SR:a_1-a_2 = -4.12159)

◦ pbc = xyz

◦ energygrps = a_1 a_2

◦ cutoff-scheme = Verlet

◦ ns_type = grid

◦ nstlist = 20

◦ rlist = 1.2

◦ vdwtype = cutoff

◦ vdw-modifier = force-switch

◦ rvdw-switch = 1.0

◦ rvdw = 1.2

◦ coulombtype = PME

◦ rcoulomb = 1.2

◦ pme_order = 4

◦ fourierspacing = 0.16

• PBC without PME (Coulomb (SR) = -77.7984 and Coul-SR:a_1-a_2 =-19.4982)

◦ pbc = xyz

◦ energygrps = a_1 a_2

◦ cutoff-scheme = Verlet

◦ ns_type = grid

◦ nstlist = 20

◦ rlist = 1.2

◦ vdwtype = cutoff

◦ vdw-modifier = force-switch

◦ rvdw-switch = 1.0

◦ rvdw = 1.2

◦ coulombtype = Cut-off;

◦ rcoulomb = 1.2

• no PBC Coulomb (SR)=-29.2468 and Coul-SR:a_1-a_2 = -29.2468

◦ energygrps = a_1 a_2

◦ pbc = no

◦ cutoff-scheme = group

◦ nstlist = 0

◦ rvdw = 0

◦ rcoulomb =0

◦ rlist = 0

For the systems with PBC, I have used gromacs/2020.3-AVX2-GPU.

For the system without PBC, I used gromacs/5.1/64.