Comparison of CHARMM and OPLS-aa forcefield predictions for components in one model asphalt mixture

Abstract

Molecular dynamics simulations have been widely applied to understand the effect of chemical composition on the major physical and mechanical properties and micro-structure of asphalt. However, the choice of forcefields has been mostly empirical. In this project, two commonly used forcefields, the Chemistry at Harvard Macromolecular Mechanics (CHARMM) forcefield and the Optimized Potentials for Liquid Simulations in all atom version (OPLS-aa) forcefield were evaluated, side by side for simulation of the three main components of one model asphalt mixture to compare the effectiveness of the forcefields. The CHARMM forcefield parameters for two smaller components —including n-docosane and 1,7-dimethylnaphthalene molecules—were obtained using the CGENFF program; the forcefield parameters for an average asphaltene molecule were obtained with the VMD forcefield toolkit and Gaussian quantum calculations. Comparing molecule structures of asphaltene, 1,7-dimethylnaphthalene and n-docosane (n-C22) predicted using the CHARMM forcefield to those from traditional OPLS-aa forcefield, the energy-minimized structures are very similar. At a high temperature such as 300 K and above, slight bending on the aromatic ring regions and the alkane chains was observed for the asphaltene molecule; kinking of the n-docosane molecule chain structure was also observed using both forcefields; and the structure of 1,7-dimethylnaphthalene is consistent at both low and high temperatures. Comparing the properties predicted based on the long-term simulation trajectories, the density, diffusion coefficient and g(r) result of asphaltene, 1,7-dimethylnaphthalene and n-docosane from the CHARMM and Nanoscale Molecular Dynamics (NAMD) program are very close to those from the OPLS-aa forcefield and LAMMPS program. However, the CHARMM forcefield correctly predicted the crystallization of n-docosane at 300 K while the OPLS-aa forcefield did not.