Enhancement of the predictive power of molecular dynamics simulations for the determination of self-diffusion coefficient and viscosity demonstrated for propane

摘要

For the development and optimization of classical effective force fields (FFs) commonly used in molecular dynamics (MD) simulations, often experimental data are required. An alternative fully-predictive approach is the use of simplified pair-specific, ab initio-based FFs (AI-FFs) derived from quantum calculations in the limit of zero density. This approach, which could successfully be employed in our recent work for methane (CH4) and carbon dioxide (CO2), has not been investigated for relatively large systems including propane (C3H8). In the present study, C3H8 was selected as model system to evaluate the predictive power of MD simulations with respect to the determination of the self-diffusion coefficient and viscosity. For this aim, a semi-rigid version of our new simplified all-atom AI-FF was applied over a broad density range from the superheated vapor to the gas state and supercritical region up to the compressed liquid state. To elucidate the effects of the AI-FF on the calculated dynamical properties of C3H8, corresponding MD simulations representing the first investigations for the studied thermodynamic states were also performed using commonly employed, effective literature FFs in the form of TraPPE and OPLS. With the exception of the compressed liquid state, it was found that our ab initio-based approach improves the prediction results for the self-diffusion coefficient and viscosity at the studied thermodynamic states corresponding to pressures between (0.1 and 5) MPa and temperatures between (293.15 and 403.15) K compared to the use of effective FFs. This could be demonstrated by comparison with the ab initio-based calculations at the limit of zero density and the scarce experimental data. It could be figured out that the combination of an all-atom and a rigid or semi-rigid representation in our ab initio-based FFs for C3H8 as well as CH4 and CO2 suits well for a reliable simulation of dynamical properties over a broad density range. (C) 2019 Elsevier B.