Predicting thermodiffusion in simple binary fluid mixtures

The predictive capabilities of some existing theoretical models to quantify thermodiffusion have been investigated in this work. To do so, the tests have been performed on two model fluids, the hard-sphere and the Lennard-Jones (including spheres and dimers) ones, exploring different mixtures and thermodynamic conditions thanks to extensive molecular simulations. It has been confirmed that the thermal diffusion factor should be expressed as the sum of one term related to the isotope effect and one term related to the "chemical" effects and that a kinetic term is required to quantify thermodiffusion from the gas state to the liquid state. In addition, regarding the isotope effects, it has been obtained that none of the available theoretical models are able to yield a reasonable prediction relatively to the molecular simulations results and that the moment of inertia contribution is one order of magnitude smaller than the mass contribution in the liquid state. Finally, concerning the chemical effects, it has been shown the Shukla and Firoozabadi model, complemented with a kinetic term, is probably the most reasonable option to estimate the chemical contribution to the thermal diffusion factor, even if it fails in capturing the effect of the asymmetry in size and in shape between the species. Overall, this works confirms that there is still a lack of a generic model able to predict accurately thermal diffusion factors, or equivalently Soret coefficient, in simple binary mixtures from the gas state to the liquid state.