Molecular geometry optimization by semiempirical methods using an optimized minimization of electronic and nuclear energies

Abstract

A method for locating minima on semiempirical potential energy surfaces is proposed. The method utilizes standard variable metric minimization techniques of optimization theory. By relaxing the criterion for convergence of the scf energy, the number of iterations may be substantially reduced for each evaluation of the semiempirical energy. Although this leads to an increase in the total number of scf calculations required for the location of the minimum energy geometry, the total number of matrix diagonalizations is decreased. By using the density matrix from one scf calculation as the starting point for the next the quality of the wave function gradually increases along the optimization route. Numerous examples are presented, demonstrating a savings in computation time with no loss in accuracy of the final wave function or geometry.