Different one-dimensional simulation techniques for studying the Kirkendall effect are developed and analyzed. Comparisons with experimentally observed marker migration show good agreement for small shifts and comparisons with observed Kirkendall porosity show reasonable agreement under the assumption that a certain supersaturation is needed before the vacancies coalesce into pores.

A convenient approach in simulations of kinetics is to use thermodynamic software, e.g. Thermo-Calc, to calculate thermodynamic quantities, e.g. chemical potentials, required in the simulation. The main drawback with such an approach is that it will generate a large amount of additional computational work. To overcome this problem a method that decreases the amount of computational work has been developed. The new method is based on artificial neural networks (ANN). By training the ANN to estimate thermodynamic quantities a significant increase in computational speed was obtained.

By calculating the dissipation of available driving force due to diffusion inside migrating interfaces an approach for including the effect of solute drag in computer simulations of grain growth and phase transformations has been developed. The new method is based on an effective interfacial mobility and simulations of grain growth have been performed in binary and ternary systems using experimentally assessed model parameters.

Keywords: Diffusion, Kirkendall effect, Phase transformations, Random walk, Solute drag, Interfacial mobility