The driving force and the kinetics of the chemically induced liquid film migration (CILM) have been investigated. The experimental results of the rate of CILM in W-Ni investigated by Yoon and Huppmann, has been examined. The volume diffusion in W grain is so rapid that the coherency strain energy in the thin diffusion layer on the surface of retreating grain can provide the driving force for CILM. The kinetics can be described in terms of solution and reprecipitation, and therefore the mobility of the process can be unambiguously determined by diffusion across the liquid film. The calculated migration rate, 1.3±1.0㎛/min, agrees at least in order of magnitude with the observed rate of 1.1㎛/min in W-Ni system. The result of this analysis indicates that a large part of the driving force for forming an equilibrium W-Ni alloy is dissipated by Ni diffusion into W solid to form a coherent layer of a nearly equilibrium composition, and a small fraction due to the strain energy is left to drive the migration. The result obviously has important implications for understanding diffusion induced grain boundary migration (DIGM) phenomenon. Two groups of experiments, Group 1 and Group 2, have been performed to observe the dependence of CILM on the driving force in W-Ni-Fe. In Group 1 experiment, spherical W particles of 70-100㎛ have been annealed in the presence of small amount of Ni-Fe melt at $1,550^\circ{C}$. Many liquid films at particle contacts migrate into neighbouring particles as reported earlier by Yoon and Huppmann in W-Ni. When the relative amount of Fe to Ni increases while keeping the total at.% of Ni-Fe constant, the migration rate increased. In Group 2 experiment, the migration of liquid films between binary W-Ni grains obtained by liquid phase sintering is induced by adding Fe to form a ternary W-Ni-Fe alloy. In group 2 experiment, fine powder mixtures of 92W-8Ni by weight are liquid phase sintered and annealed for 5 hours at $1,550^\circ{C}$ until W grains grow considerably by Ostwald ripening. Each specimen is reannealed for various times up to 4 hours after adding Fe powder. After adding Fe, liquid films between grains were observed to migrate into neighbouring grains. The migration rate increases with the amount of Fe added. The experimentally observed migration rates at various compositions in both experiments agree qualitatively, with the prediction based on the coherency strain energy as the driving force. Because of lack of the available data of misfit parameter, an accurate quantitative analysis is not possible. At very high Fe content (50Fe-50Ni), the migration rate decreases again in both experiments. This result can be attributed to the decrease of diffusivity of W in liquid with the relative increase of Fe to Ni, though the exact reason is not clear.
After considerable migration, some of the liquid films reverse the direction of migration leaving the unstable phase grain of decrescent moon shape. Such phenomena occur, even when the radius of curvature of solid-liquid interface, which normally impedes the migration, apparently remains unchanged. The phenomena thus do not seem to occur soley by curvature effect. The change of the driving force with time can be the cause for the phenomena, since the mobility of CILM is expected to be constant with time. The time dependence of coherency strain energy at the surface of retreating grain has been analyzed by taking advantage of the fact that coherency stress is analogous to thermal stress. The analysis shows that the functional nature of coherency stress is the cause for the decrease of driving force with time. When the total solute content increases in the retreating grain with time, or when the size of an equlibrium phase coherently adherent to the retreating grain increases with time, the average lattice parameter of the grain will approach that of equilibrium phase. The magnitude of strain at the surface of the grain, and hence the driving force, is thus expected to decrease with time. A mathematical model for the rate of migration, in which the coherency strain energy is assumed to be the driving force for the liquid film migration, is developed. The agreement between the result of numerical calculation based on the model and the experiment is quite good. Also, the result of calculation shows that the interface between the unstable phase and the stable one would rather be incoherent than coherent.