(257q) Global Optimization of Continuous Force Fields for Organic Compounds
AIChE Annual Meeting
Monday, November 9, 2015 - 6:00pm to 8:00pm
Continuous force fields have traditionally been optimized in an arduous manner that requires extensive simulation and resimulation as the force field parameters are refined and tested. Taking cues from experience with step potentials, we have developed an optimization protocol facilitated by thermodynamic perturbation theory (TPT) combined with molecular dynamics of continuous potential models (CMD) using the LAMMPS simulator. Reference fluid simulations are supplemented with limited simulations of the full potential with “draft” candidates for the optimal potential model, permitting the characterization of complete equations of state that permit variable values of the potential parameters. The methodology follows the publications of Ghobadi and Elliott (J. Chem. Phys., 141:094708, 141:024708, 140:234104). The potential parameters can then be optimized without further simulation. Finally, the optimized potentials are tested with full potential simulations.
The potential model of interest in the present work is the 14-12-8-6 model, cut and shifted at 1.2nm. This permits accentuation of the steepness of the potential, as in the m-6 Mie model, while retaining the efficiency and parallelization of the Lennard-Jones (LJ) model. The cut and shifted potential is better adapted for applications to inhomogeneous fluids like interfaces or most biological applications. For the most part, results are similar to those with the m-6 model. Vapor pressure deviations average near 9% for n-alkanes modeled with transferable united atom models, and 3% for noble gases. Saturated liquid density deviations average near 1% below a reduced temperature of 0.9. Compressed fluid densities deviations average near 0.5% for the better potential models. Critical temperatures are matched to within 1K. Critical pressures are matched to within 0.1MPa and critical densities are matched to within 0.01 g/cm3. These results establish a firm foundation for optimal force field development going forward to encompass branches, rings, alcohols, amines, etc.
To demonstrate the efficiency of this protocol, vapor pressure, critical properties, saturated liquid density, and compressed fluid density are used to characterize potential models for n-alkanes, branched alkanes, alcohols, ethers, aldehydes, ketones, olefins, and aromatic compounds.