Parameterization of NDDO wavefunctions using genetic algorithms. An evolutionary approach to parameterizing potential energy surfaces and direct dynamics calculations for organic reactions

We used a genetic algorithm to fit a set of energy differences obtained by neglect-of-diatomic-differential-overlap (NDDO) molecular orbital theory to reference ab initio data, yielding a set of specific reaction parameters (SRP) for the reaction Cl + CH₄. Only a small number ab initio points along a distinguished-coordinate path were used as input, but the surface is well fit both on and off the reaction path over a range of energies three times wider than the input range. The resulting NDDO-SRP potential energy surface is almost four orders of magnitude less expensive to evaluate than the reference ab initio surface and is well suited for direct dynamics calculations.