I have a system of star polymers, 8000 star polymers, each with 11 monomers. I want to find the pairwise forces between the molecules. compute pair/local gives me pairwise per atom forces. I have the option of group/group or force/tally, but I will need to define 8000 groups. Is there any way to get the pairwise forces about the center of mass of the molecules? Will it be possible to calculate pairwise forces per molecule using chunk/atom? Any advice will help.

I need this to calculate the force-velocity correlation with time, which will further help me in computing the memory kernel for a non-Markovian DPD simulation.
I am more interested in the fluctuating force(instantaneous-mean), i.e. Fij-<Fij, where i and j denote the ith molecule and jth molecule respectively.

You may be able to implement this as a modification compute pe/mol/tally, but that would require that you have no long-range electrostatics and no many-body potentials. Beyond that, I don’t see a way outside of modifying pair styles for your specific needs in a way that would not be an acceptable modification for inclusion of the LAMMPS distribution. And if you are using long-range electrostatics, you still have a significant problem.

This sounds like something that is much easier done in post-processing anyway. Forces can be calculated from trajectory coordinates and then correlated with separate velocity records.

I was taking post-processing initially by using the pair/local option and then adding the forces to get intermolecular forces. However, the file sizes are becoming way too large using this method. That’s why, I was seeking help. If some inbuilt function within lammps might make my work a bit easier.