In the theoretical part of my contribution I will introduce the Always Stable Predictor-Corrector (ASPC) method. It is suitable for integrating the equations of motion with a right-hand side containing an implicit equation of a self-consistent field (SCF) type, like calculation of induced dipoles. The method combines two principles: stability (errors decay while propagating to next steps) and time reversibility of a sufficiently high order (energy conservation). The method is locally of second order. It contains one mixing (damping) parameter which can be chosen so that the method is always stable for a converging SCF or optimized to convergence properties of given system. Higher-order but only partially stable schemes are also discussed. The ASPC method can be written both in a history form and Gear notation.
The ASPC method is tested using polarizable water and compared with the full iteration method with a predictor an the extended Lagrangian (Car-Parrinello-like) method. Accuracy and efficiency of the methods are discussed.
In the application part two systems are discussed:
- Rock saltl/brine equilibrium by a mixed model of polarizable ions and nonpolarizable water
- An attempt to reparameterize TIP4P with polarizability
CECAM - Centre Européen de Calcul Atomique et Moléculaire Station 13, Bat. PPH, 1015 Lausanne, Switzerland
