Intramolecular polarisable multipolar electrostatics from the machine learning method Kriging

We describe an intramolecularly polarisable multipolar electrostatic potential model for ethanol, which acts as a pilot molecule for this proof-of-concept study. We define atoms via the partitioning prescribed by quantum chemical topology (QCT). A machine learning method called Kriging is employed to capture the way atomic multipole moments vary upon conformational change. The multipole moments predicted by the Kriging models are used in the calculation of atom-atom electrostatic interaction energies. Charge transfer is treated in the same way as dipolar polarisation and the polarisation of higher rank multipole moments. This method enables the development of a new and more accurate force field.