In multi-component alloys, the composition dependence of bulk and interface properties forms high-dimensional datasets that grow exponentially with the number of elements, making true multi-component simulations computationally prohibitive. To overcome this challenge, we have successfully applied tensor completion techniques (in collaboration with Prof. L. Delathauwer) and neural networks to efficiently obtain, represent, and utilize Gibbs energies, diffusion mobilities, thermodynamic factors, and atomic mobilities as functions of composition (as for example modeled within the CALPHAD approach) in phase-field simulations. This approach enables us to fully capture composition-dependent properties, allowing for the study of cross-element interactions, complex diffusion behavior, and intricate phase transformation pathways. We also explored the use of neural network models to determine and/or represent other material properties in a phase-field model.

Further reading

• Combining thermodynamics with tensor completion techniques to enable multicomponent microstructure predictions (2020)

• Phase-field approach to simulate BCC-B2 phase separation in the AlnCrFe2Ni2 medium-entropy alloy (2022)

• Integration of machine learning with phase field method to model the electromigration induced Cu6Sn5 IMC growth at anode side Cu/Sn interface (2020)

• Combining multi-phase field simulation with neural network analysis to unravel thermomigration accelerated growth behavior of Cu6Sn5 IMC at cold side Cu-Sn interface (2020)

• Features and classification of solid solution behavior of ternary Mg alloys (2024)

• Interaction of elements in dilute Mg alloys: a DFT and machine learning study (2022)