High‐Strength/Toughness High‐Entropy Nitrides Composition Design via High‐Throughput Calculations and Machine Learning

ABSTRACT

The machine learning models for rationally designing high‐entropy nitride coatings with high strength/toughness were developed. The Vickers hardness and Poisson's ratio datasets including ternary, quaternary, and quinary TiN‐ and ZrN‐based nitride coatings were built through high‐throughput first‐principles calculations. Random forest, support vector regression and deep neural network models were optimized and used to predict Vickers hardness and Poisson's ratio of coatings. It was found that support vector regression models displayed the highest predictive accuracies of 0.946 for Vickers hardness and 0.939 for Poisson's ratio in the independent datasets, demonstrating the superior generalization ability. In a composition space encompassing 43 673 unreported quinary high‐entropy nitride coatings, the support vector regression models inferred 280 coatings with hardness exceeding TiN and 26 462 coatings possessing toughness feature, whose prediction accuracies were validated by first‐principles calculations. It indicates that the support vector regression is a robust and reliable model to design high‐entropy nitride coatings with high hardness or toughness, which provides the basis for the design and preparation of multilayer coatings with high hardness and toughness.