A big data approach to calibrate a single transferable polarizable force field parameter set for alcohols

In this work, parameters are developed for a polarizable force field for linear alcohol models, using our previously described QM/MM approach to attribute condensed-phase polarizabilities. Starting from the non-polarizable GROMOS 53A5/53A6 parameter set, van der Waals and Coulomb interaction parameters are optimized to reproduce pure-liquid (thermodynamic, dielectric and transport) properties, as well as hydration free energies. Here we show that transferability of force field parameters can be improved by a systematic (analytical) approach to the calibration e↵ort. Designing an analytical model that is able to accurately predict the e↵ect of parameter changes on the physicochemical properties reproduced in simulation, a large set of 17489 models (from over 35 million parameter combinations) can be identified that all reproduce pureliquid properties of the series methanol-butanol within experimental accuracy. Moreover, 34 models could be selected without special repulsive van der Waals parameters to distinguish between hydrogen-bonding and non-hydrogen-bonding atom pairs, making the force field simple and transparent. Further simulations were performed using the parameters of the best-ranked model, demonstrating that hydration, dielectric and transport properties were also well reproduced. In addition, results from simulations of butane suggested that the obtained aliphatic parameters are transferable to alkanes.