Monte Carlo simulations of real fluids

HSU

15. February 2023

Univ.-Prof. Dr.Ing. Karsten Meier (Chair of Thermodynamics, HSU)

Thermodynamic properties of gases and liquids over wide ranges of temperature and pressure can be determined by molecular-dynamics or Monte Carlo simulations. In such simulations, the interactions between molecules are almost exclusively modeled by simple effective pair potentials, such as the Lennard-Jones potential. With such an approach, accurate predictions are only possible in small regions of temperature and pressure.  Our research aims at the very accurate prediction of thermodynamic properties and phase equilibria of noble gases, nitrogen, and carbon dioxide using sophisticated ab initio potentials, which are derived from quantum-chemical calculations of the interaction energies, by semi-classical Monte Carlo simulations. We include nonadditive three-body interactions and Feynman-Hibbs corrections for quantum effects, which yield significant contributions to the macroscopic properties.

The calculation of the interaction energy between the molecules in Monte Carlo simulations requires large computational resources, especially for mathematically complex ab initio potentials. The evaluation of the pairwise interactions between molecules scales with N2, the evaluation of the nonadditive three-body interactions with N3, where N is the number of particles (typically N = 100…1500).

Our in-house command-line code is written in Fortran 90 and parallelized with MPI. It enables the simulation of many single-phase thermodynamic properties by Metropolis Monte Carlo in seven different statistical ensembles and of vapor-liquid equilibria by the NpT+test particle method for several potential models of noble gases and linear molecules. The parallelization is mainly applied to the calculation of the interaction energies between molecules. A typical simulation run with 500 particles for the noble gas argon requires about 40 days on a 28-core compute node. Our aim of this cooperation is to reduce the run time of our software at least by a factor of two.