New approach to the rapid semiempirical calculation of molecular electrostatic potentials based on the am1 wave function: Comparison with ab initio hf/6-31g* results

A new approach to the computation of molecular electrostatic potentials based on the AM1 wave function is described. In contrast to the prevailing philosophy, but consistent with the underlying NDDO approximation, no deorthogonalization of the wave function is carried out. The integrals required for the computation of the electronic contributions to the molecular electrostatic potential are evaluated in a manner similar to that of the AM1 core-electron attraction integrals, while the nuclear contributions are computed using a new semiempirical function—Z_A(S^AS^A, S^pS^p)1 + exp – ω_A(R_Ai – δ_A)]]—where the atomic parameters ω_A and δ_A are obtained by calibration against the results of ab initio HF/6-31G* calculations. Isopotential contour maps for guanine and cytosine obtained with the new method are qualitatively almost indistinguishable from their HF/6-31G* counterparts, while quantitative comparisons for the minima for a wide range of molecules are reproduced with an rms error of 5.2 kcal mol^?1. The locations of the “lone-pair” minima for a wide range of heterosubstituted organic molecules generally fall within 0.02 Å of the corresponding HF/6-31G* minima while those in the π-regions of unsaturated molecules are generally within 0.2 Å. Because of the rapid integral evaluation, the fully semiempirical method described here is extremely economical. For example, for the guanine–cytosine base pair it is >500 times faster than calculations in which the complete integral matrix is computed analytically from the deorthogonalized AM1 wave function. © John Wiley & Sons, Inc.