Grain-boundary diffusion rates in the gold-silver system were measured at relatively low temperatures by the surface-accumulation method which was analyzed in Paper I. The specimen was a polycrystalline gold film possessing columnar grains on which a silver layer was initially deposited epitaxially on one surface. During subsequent low-temperature annealing lattice diffusion was frozen out, and diffusion then occurred along the grain boundary and free-surface short circuits. The silver, therefore, diffused into the film from the silver layer along the boundaries, eventually reaching the opposite surface where it accumulated and was measured by Auger spectroscopy. The silver layer acted as an effective constant silver source, and grain-boundary diffusivities were calculated from the accumulation data. However, the exact location of the effective constant source in the silver layer could not be determined and this led to an uncertainty in the values of the grain-boundary diffusivities of a factor of 10. Lower- and upper-bound values were therefore described by D b(lower bound) =7.8×10-6 exp(-0.62eV/kT) and D b(upper bound) =7.8×10-5 exp(-0.62eV/kT) cm 2/s in the temperature range 30-269 °C. An examination of available grain-boundary diffusion data (including the present) suggests a tendency for the observed activation energy to decrease with decreasing temperature, and this was ascribed to a spectrum of activated jumps in the grain boundary and/or a spectrum of grain-boundary types in the specimen employed. The constant source behavior was tentatively ascribed, at least in part, to a grain-boundary Kirkendall effect" resulting from the faster diffusion of silver than gold. The work indicates a need for increased understanding of the details of grain-boundary diffusion in alloys."
