Geometry optimisation at the semiempirical self-consistent-reaction-field level using the AMPAC and MOPAC programs
Abstract
The self-consistent-reaction-field (SCRF) solvation model as recently implemented in the AMPAC and MOPAC semiempirical SCF-MO programs can result in significant errors in the calculated energies of some polar molecules. This is due to geometry optimisation using approximate gradients evaluated assuming first-order invariance of the density matrix with respect to geometry. For highly polar molecules in particular, we suggest that SCRF geometry optimisations are carried out with full re-evaluation of the density matrix for each gradient, a procedure that results in an increase in the computing time, but increased accuracy.