Systematic first principles parameterization of force fields for metal-organic frameworks using a genetic algorithm approach.

A systematic strategy is proposed to derive the necessary force field parameters directly from first principles calculations of nonperiodic model systems to reproduce both the structure and curvature of the reference potential energy surface. The parameters are determined using a genetic algorithm combined with a novel fitness criterion based on a representation of structure and curvature in a set of redundant internal coordinates. Due to the efficiency of this approach it is possible to abandon the need for transferability of the parameters. The method is targeted for the application on metal-organic frameworks (MOFs), where parameters for molecular mechanics force fields are often not available, because of the wide range of possible inorganic fragments involved. The scheme is illustrated for Zn4O-based IRMOF materials on the example of MOF-5. In a "building block" approach parameters are derived for the two model systems basic zinc formate (Zn4O(O2CH)6), and dilithium terephthalate with reference data obtained from density functional theory. The resulting potential gives excellent agreement with the structure, vibrational frequencies, thermal behavior and elastic constants of the periodic MOF-5.