I wonder if in latest Gromacs the [pairs] section of the topology file is still used or if Gromacs figures out close by atoms automatically. Gromacs seems to consider non bonded pair interactions between atoms of one molecule even when these are not given in the [pars] section of the toplogy file, so I wonder what is special about those pairs given in the [pairs] section. Is the [pairs] section historic like the charge groups (see section “Charge groups” in Molecular Dynamics — GROMACS 2021.2 documentation)? Or does this section exist to have finer control for 1-4 interactions than for more remote interactions? Does this depend on parameters?
Gromacs knows connectivity through [bonds], does not need [pairs] to figure close by atoms. But [pairs] is required for defining 1-4 atoms. 1-4 is force-field specific, controls intral-molecular LJ/electrostatic interactions of dihedral’s terminal atoms.
So is the point that 1-4 pairs are treated specially because historically some force fields included all 1-4 interactions in the dihedral terms while others didn’t so that one had to specify this somhow?
I still wonder if current Gromacs:
- treats 1-4 pair interactions automatically and the [pairs] section is just a historsic artifact
- never treats 1-4 pair interactions automatically, so that one have to have a pairs section
- treats 1-4 pairs automatically if no pairs section is there and excludes 1-4 pairs from automatic treatment if there is a pairs section
- treats whatever pairs are given in the pairs section specially - say with better precision or more control - no matter if they are 1-4 or not - and all other pairs automatically
- does something else
I would highly appreciate some documentation pointers on this topic - I couldn’t find anything beyond the one I mentioned above.
Each force field treats 1-4 interactions differently. Some are scaled, some have different LJ and Coulombic scaling factors, some (like CHARMM) do not scale 1-4 interactions but sometimes include special 1-4 LJ terms (intramolecular NBFIX). So a
[pairs] directive is always required and is always used. Without it, 1-4 interactions would otherwise be entirely excluded using the standard nrexcl = 3. So essentially, GROMACS excludes nonbonded interactions between 1-4 pairs, then re-incorporates them as bonded interactions (via
[pairs]) and may treat them with special scaling terms. The software has always done this.
Thanks, this fully answers my question!