It is really starting to appear like many problems we used to look at with conventional forcefields actually require some degree of polarizability. I’ve looked at Justin’s paper on the Drude model and also Jorgensen’s 2013 paper (https://pubs.acs.org/doi/abs/10.1021/jz302085c) and I am nearly convinced.
Please correct me if I’m wrong, but it appears that OPLS-AAP described at the very end of the paper above is implemented in Gromacs as type 1 described here: http://manual.gromacs.org/documentation/2019.1/reference-manual/functions/polarization.html – except I am not sure I understand the reference to a harmonic potential (other than the field-induced dipole on a given atom is a linear function of the field).
A few other things are unclear. Let’s say I want to have a polarizable molecule. My understanding is that the [ polarization ] section is all that’s required. However, what is the meaning of index j in the corresponding entries? Say, I want a CNT or a graphene sheet, in which every atom is polarizable according to OPLS-AAP. The likeliest “polarizer” would be a charged ion, which isn’t part of that topology. How does one set up the molecule, then, and why are two indices required for each entry?
Sasha (used to be Alex in the old mailing list)