A Near-Black Box Parameter Optimizer for NDDO-Descendant Semiempirical Methods: Reparameterizations for MNDO and AM1

We provide a detailed description of an enhanced version of our previous geometry-corrected parameter optimization algorithm capable of accounting for reference geometries, including all pertinent equations necessary for its implementation. As a demonstration of the utility of our novel algorithm, reparameterizations for MNDO and AM1 using 1,187 CHNO molecules in the PM7 training set are reported and compared to analogous results obtained from the PARAM program used in development of the PMx models. Our AM1 reparameterization yielded unexpectedly large improvements, and our findings indicate that derived geometrical reference functions are ill-suited for parameter optimization and may also inadvertently incentivize smaller force constants for chemical bonds. Together, these results suggest that judicious parameter refinement could substantially enhance the performance of NDDO-descendant semiempirical models.