Machine Learning Force Field Parameters for C-O-H Atoms

Hello all,

I would like to use Machine Learning Force Field (ML FF) parameters for my simulation containing a graphene oxide membrane and salt solution (H2O+ions). I am interested in applying “pair_style mliap” or “pair_style snap” commands for my system. According to my search, there are limited ML FF parameters for a few metal atoms only in the LAMMPS package (example directory). It is been a week now that I have been searching the literature to find ML FFs suitable for my system, however no success yet.

I would truly appreciate it if anyone who has prior experience in this could share any relevant references or insights that could help me implement ML FF parameters in my system.

Thank you in advance.


The chances of someone else doing the exact same work on a complex system (graphene oxide membrane and salt solution (H2O+ions)) and who has already created the machine learning potential you need is pretty low to zero. So you probably will need to train a new machine learning potential yourself, using quantum chemistry calculations as the training set. There is much discussion in the literature on the topic of training ML potentials from DFT.