On the notion of thermophoretic velocity

The phenomenology of thermal diffusion of a particle in a solvent involves the Soret coefficient of the particle/solvent mixture. It is usually considered that the migration velocity of the particle in the solvent is proportional to the Soret coefficient. I show here that this view is wrong because ordinary diffusion contributes to migration if the solvent is not homogeneous. I examine several examples (NaCl in water, polystyrene in ethylbenzene, maghemite nanoparticles) to show that this contribution can be sufficiently strong to change the sign of the migration velocity and vitiate the interpretation of experimental data, measurement of the Soret coefficient, the comparison of experiment with theory and inter-experimental comparisons.