A modified electrostatic model was applied to impurity diffusion, and a Z0-value of 0.4 was assumed for the effective charge on the Ni ion. A comparison of calculated and experimental diffusion parameters revealed that the sign of the difference between the impurity diffusion and self-diffusion energies, and the sign of the difference between the impurity diffusion and self-diffusion coefficients, were always correctly predicted. On the other hand, these comparisons revealed systematic deviations in the case of 5p impurities. These could not be explained by current impurity diffusion models.

G.Neumann, V.Toelle: Applied Physics A, 1996, 63[4], 377-80