Thermodiffusion in multicomponent mixtures. Thermodynamic, algebraic, and neuro-computing models (Q2276432)

This short monograph refers to an important transport phenomenon in multicomponent mixtures, which stems from a competition between the Soret effect (thermodiffusion in the case of a nonuniform temperature distribution) and the Fickian diffusion process due to concentration gradients. Its manifestations can be observed in many processes occurring in nature, including biological applications and neural network computing analogies. The authors present three basic theoretical approaches aimed at quantifying thermodiffusion, one derives from linear response theory in nonequilibrium thermodynamics of multicomponent mixtures, the second refers to a development of various algebraic simplifications that quantify a coupling of extensive thermodynamic data to an appropriate equation of state, and, in the third, neural networks are employed to predict the thermodiffusion data based on the principles of associative thinking. The resultant estimates of enthalpy or fugacity give a quite accurate quantitative description of thermodiffusion processes in specific liquids. Layered neural network computing algorithms are advanced as reliable tools for the description of thermodiffusion in multilayered mixtures. The book is written from the physicist's point of view and its addressees are professionals in thermal engineering and material science, plus postdoctoral researchers and graduate students of physics and applied science faculties.