Optimization of strong and weak coordinates

We present a new scheme for the geometry optimization of equilibrium and transition state structures that can be used for both strong and weak coordinates. We use a screening function that depends on atom-pair distances to differentiate strong coordinates from weak coordinates. This differentiation significantly accelerates the optimization of these coordinates, and thus of the overall geometry. An adapted version of the delocalized coordinates setup is used to generate automatically a set of internal coordinates that is shown to perform well for the geometry optimization of systems with weak and strong coordinates. For the Baker test set of 30 molecules, we need only 173 geometry cycles with PW91/TZ2P calculations, which compares well with the best previous attempts reported in literature. For the localization of transition state structures, we generate the initial Hessian matrix, using appropriate force constants from a database. In this way, one avoids the explicit computation of the Hessian matrix. (C) 2006 Wiley Periodicals, Inc.