In GROMACS nrexcl sets the exclusions for non bonded interaction. I have nrexcl = 2 which means that neighbours 2 bonds away are excluded i.e., 1-2 and 1-3 are excluded. gen-pair = yes and fudgeLJ = 0.5 (which is what I use) means that 1-4 interactions are considered and are scaled by 0.5. If I read the documentation for special_bonds, this translates to: special_bonds lj 0.0 0.0 0.5 (which is what I use).
Before even going to a 3-mer, I just checked a 2-mer so its just a dumbell. The forces are all comparabale, and in the first frame are equal.
The following are forces that GROMACS considers (converted to LAMMPS units)
0.61407 2.09926 -0.0856839
-0.61407 -2.09926 0.0856839
0.628615 2.14897 -0.0877142
-0.628615 -2.14897 0.0877142
0.60336 2.06263 -0.0841899
-0.60336 -2.06263 0.0841899
The following are forces that LAMMPS consideres:
1 0.614067 2.09925 -0.0856838
2 -0.614067 -2.09925 0.0856838
1 0.594722 2.03312 -0.0829845
2 -0.594722 -2.03312 0.0829845
1 0.537906 1.83889 -0.0750566
2 -0.537906 -1.83889 0.0750566
In this case, I expected the radius of gyration of this dumbell to be almost overlapping. But they are not.
