Elementary kinematical model of thermal diffusion in liquids and gases.

An elementary hydrodynamic and Brownian motion model of the thermal diffusivity D(T) of a restricted class of binary liquid mixtures, previously proposed by the author, is given a more transparent derivation than originally, exposing thereby the strictly kinematic-hydrodynamic nature of an important class of thermodiffusion separation phenomena. Moreover, it is argued that the solvent's thermometric diffusivity alpha appearing in that theory as one of the two fundamental parameters governing D(T) should be replaced by the solvent's (isothermal) self-diffusivity D(S). In addition, a corrective multiplier of O(1) is inserted to reflect the general physicochemical noninertness of the solute relative to the solvent, thus enhancing the applicability of the resulting formula D(T)=lambdaD(S)beta to "nonideal" solutions. Here, beta is the solvent's thermal expansivity and lambda is a term of O(1), insensitive to the physicochemical nature of the solute (thus rendering D(T) primarily dependent upon only the properties of the solvent). This formula is, on the basis of its derivation, presumably valid only under certain idealized, albeit well-defined, circumstances. This occurs when the solute molecules are: (i) large compared with those of the solvent; and (ii) present only in small proportions relative to those of the solvent. When the solute is physicochemically inert, it is expected that lambda=1. When these conditions are met, the resulting thermal diffusivity of the mixture is, in theory, independent of any and all properties of the solute. Moreover, because beta is algebraically signed, the thermal diffusivity can either by positive or negative, according as the solvent expands or contracts upon being heated. This formula for D(T) is compared with available experimental data for selected binary liquid mixtures. Reasonable agreement is found in almost all circumstances with lambda near unity, the more so the higher the temperature, especially when the solute-solvent mixture properties closely approximate those where agreement would be expected and conversely. Finally, it is pointed out that for the restricted circumstances described, the formula D(T)=lambdaD(S)beta is equally credible for gases. Here, based on gas-kinetic theory, it is possible to furnish the theoretical value of lambda. Overall, while spanning a range of about five orders of magnitude, the D(T) values given by this elementary formula are shown to apply with reasonable accuracy to: (i) liquids (including circumstances for which D(T) is negative) as well as gases; (ii) all combinations of solvents and solutes tested (the latter including, for example, polymer molecules and metallic colloidal particles); and (iii) all sizes of solute molecules, from angstroms to submicron.