Energy grps splitting to calculate interaction energy of single water

User discussions
1

GROMACS version:2022.5
GROMACS modification: Yes/No

I simulated TIP4P/ice hexagonal ice with 2000 water molecules (Nw). I divided total potential energy by Nw to get potential energy per mole
-123290/2000=-61.65 kJ/mol

Then I divided the 2000 water system into two energy groups: BW1 (1 water) and BW2 (1999 waters), and calculated the interaction energy.
Energy Average Err.Est. RMSD Tot-Drift

Potential -123290 10 425.428 51.3519 (kJ/mol)
Coul-SR:BW1-BW1 -4.50819 0.0016 0.0632246 -0.0024656 (kJ/mol)
LJ-SR:BW1-BW1 0 0 0 0 (kJ/mol)
Coul-SR:BW1-BW2 -143.349 0.48 12.8556 -2.05087 (kJ/mol)
LJ-SR:BW1-BW2 29.5369 0.36 10.8844 1.58858 (kJ/mol)
Coul-SR:BW2-BW2 -151971 14 544.409 86.7442 (kJ/mol)
LJ-SR:BW2-BW2 29481.2 5.5 193.345 -34.6624 (kJ/mol)
As NBW1=1, I can use the self term (BW1-BW1) directly, but I didn’t know how to normalize the cross term (BW1-BW2). Then I tried this formula
[(Coul-SR:BW1-BW1+LJ-SR:BW1-BW1) / NBW1]+[(Coul-SR:BW1-BW2+LJ-SR:BW1-BW2) / NBW1*2] = -61.41 kJ/mol
I tried this formula for splitting BW1=10 and BW2=1990, and it gave an energy of -61.37 kJ/mol.
Similarly, I tried it for the spc water model at 300 K, which gave me an average potential energy of -41.89 kJ/mol without splitting.
Then I tried splitting the system into BW1 and BW2 of varying fractions and still got the matching energies with the unsplitted system.

  1. I do not know why I need to divide self-term (BW1-BW1) by NBW1 and cross terms (BW1-BW2) by NBW1*2?
  2. Can I use the same approach to calculate the energy of the interface water at the protein by considering the interface water as BW1 and all other system atoms as BW2?
2

Why do you want to calculate interaction energies?

You cannot split interaction energies over groups. You don’t seem to write how you handle the cross term, but you could assign half to one group and half to the other. That might explain the factor 2. But that also shows that splitting in meaningless.

And finally, most simulations are rather inaccurate without PME. PME mesh energies are difficult to split.