Calculation of Transport Parameters Using ab initio and AMOEBA Polarizable Force Field Methods

J Phys Chem A. 2021 Jun 10;125(22):4918-4927. doi: 10.1021/acs.jpca.1c03028. Epub 2021 May 26.

Abstract

The transport properties of chemical species such as coefficients of diffusion, thermal conductivity, and viscosity have been widely used in combustion modeling. Lennard-Jones parameters fitted from the accurate intermolecular potential energy surfaces are crucial to obtain such information. Hence, a fast and accurate energy function is always desired for this purpose. In this study, the quality of a widely used polarizable force field AMOEBA was examined for the interaction between noble gases and n-alkanes. First, the intermolecular energy was compared between AMOEBA, MP2/CBS, MP2/aug'-cc-pVDZ, and QCISD(T)/CBS. The root mean squared error of the original AMOEBA was 10.31 cm-1 against QCISD(T)/CBS for all conformations. This was comparable with the errors of 10.84 and 7.75 cm-1 for MP2/aug'-cc-pVDZ and MP2/CBS, respectively. Further optimizing the van der Waals parameters of noble gases, the error of the force field against QCISD(T)/CBS was reduced to 6.24 cm-1, even better than the MP2/CBS results. Based on the optimized force field parameters, the intermolecular Lennard-Jones parameters were derived using the spherically averaged method and one-dimensional minimization method for a set of (n-alkanes, noble gases) pairs. The discrepancy of the one-dimensional minimization predicted Lennard-Jones collision rates from the tabulated values was typically within 10%, while it could be as large as 20-30% for the spherically averaged method. Additionally, the binary diffusion coefficients were calculated using the present Lennard-Jones parameters. In this case, the parameters derived from the spherically averaged method perform better. The mean unsigned error of the diffusion coefficients is usually within 5%, which is in good agreement with the experimental results. The results demonstrate that the AMOEBA force field can be used to generate the transport parameters systematically.