Wednesday, 09 October, 2013
SPEAKER: Prof. Lukas Vlcek, ORNL
ABSTRACT: As computer simulations are increasingly used to complement or replace experiments, highly accurate descriptions of microscopic interactions are required for realistic predictions. One way to achieve this goal, while maintaining computational efficiency, is to use simple models with highly optimized interaction parameters. Here, we present a novel method for efficient global optimization of empirical molecular models. The speed of the technique, based on the free energy perturbation principles, makes it possible to explore many-dimensional parameter spaces and identify optimal models within minutes upon collecting data from a single simulation. The approach has been used to develop molecular models of aqueous electrolytes and optimize H2O/CO2 interactions for fluid-phase equilibria. In the process, inconsistencies have been noticed between molecular simulations and the currently accepted values of single-ion hydration thermodynamic properties. Our results suggest significant errors in the reference data, which are routinely used in ion hydration studies. The implications of this finding for the development of electrolyte models will be discussed. The presentation will conclude with an example of mesoscale simulations that use a simple lattice gas model to achieve a realistic representation of near-critical CO2 confined in a porous material.