Thermodiffusion models

Three approaches to study thermodiffusion in binary and multicomponent mixtures are explored in this chapter, viz., the nonequilibrium thermodynamics, algebraic correlations, and artificial neural network. The first method employs the principles of nonequilibrium thermodynamics to explain thermodiffusive separation, by considering the heat and mass fluxes in the mixture as linear functions of forces such as temperature gradient and chemical potential. The second method is based on the observation of relations between the thermodiffusion parameters and parameters such as the mixture composition and pure component/mixture properties. Finally, in artificial neural networks, a data mining of a reasonably large set of experimental data is undertaken and a model is developed that predicts the thermodiffusion data based on the principles of associative thinking. To this end, mathematical functions are integrated in the model to quantify the decision-making process. Expressions corresponding to all three methods are discussed in this chapter.