Automating Molecular Forcefield Development and Improving Parameterization Methods

At the heart of all Molecular Dynamics simulations lies an energy potential that seeks to capture the underlying quantum mechanical interactions between atoms and molecules. However, describing these interactions is difficult, and developing parameters and functional forms for molecular forcefields is a major roadblock for researchers looking to study new systems computationally. We have simplified forcefield parameterization through an iterative optimization approach that automatically generates new training sets, calculates their energies using Density Functional Theory, and fits the results to the desired potential. We also implement a method of joining distinctive forcefields that allows for flexible choices of short- and long-range potentials that efficiently model reactive environments. This method is illustrated by modeling lead-sulfide quantum dots and their passivating ligands. Finally, we propose methods of weighting training set data based on geometry and energy considerations to increase model accuracy during simulations