If we ever need this combination of Mie potential with binary interaction parameter custom mixing rules, can revisit the framework I started here:

e6f603e#diff-e8dfc92764bdbb7fa37ed231cdaa988bc9fb319fffb733b34bc3985b03e3020aR705

Since the binary interaction (kappa) parameters are per-pair-type, rather than global or per-particle, I tried to use interaction groups, whereby we divide the nonbonded forces by pair types, and set kappa as global parameters for separate CustomNonbondedPotential objects. But something is not quite right (the validation of OpenMM vs. manually calculated nonbonded potential energies is off).