Machine Learning Guided Exploration of High Strength Thin-Film Alloys

Machine learning based prediction models have been spotlighted as a new approach to discover alloy compositions with outstanding properties. It is however challenging to acquire a large amount of strength data for vast composition ranges for the models. Combinatorial synthesis and high-throughput experiments can provide a practical pathway to obtain various composition dependent properties of thin films libraries. In this study, magnetron sputter deposition was used to produce ternary thin film alloys with composition spreads. X-ray diffraction, scanning electron microscopy, and nanoindentation were performed to characterize composition dependent microstructural properties and hardness. Then a machine learning model that can predict hardness of the alloy thin films was constructed. We show that the model has a high accuracy and can be applied to develop novel thin films with high strength.