Data-driven optimization of hardness and toughness of high-entropy nitride coatings

High coating hardness and toughness are mutually contradicting properties and are challenging to be achieved simultaneously. Combining the vast component space of high entropy systems and the powerful high -dimensional data processing tools is expected to be the best solution to this problem. In this paper, high -entropy nitride coatings data for quinary and hexagonal systems were collected and machine learning predic-tion models were trained. Using a new material system combined with multi-objective optimization, high -entropy nitride coatings with the optimal hardness and elastic modulus combination were successfully ob-tained and verified by experiments. In addition, the partial dependence heatmaps were used to visualize how elemental content affects mechanical properties prediction in this system. This approach helped to better interpret the optimization results and discover the unknown mapping relationships between elemental content and the mechanical properties of high-entropy nitrides in machine learning models.