Neural Network Prediction of Interatomic Interaction in Multielement Substances and High-Entropy Alloys: A Review

One of the most exciting tools that have entered the arsenal of modern science and technology in recent years is machine learning, which can efficiently solve problems of approximation of multidimensional functions. There is a rapid growth in the development and application of machine learning in physics and chemistry. This review is devoted to the possibilities of predicting interatomic interactions in multielement substances and high-entropy alloys using artificial intelligence based on neural networks and their active machine learning, which provides a comprehensive overview and analysis of recent research on this topic. The relevance of this direction is due to that the prediction of the structure and properties of materials by means of atomistic quantum mechanical modeling based on density functional theory (DFT) is difficult in many cases because of the rapid increase in computational costs with increasing size in accordance with the size of the object. Machine learning methods make it possible to reproduce real interparticle interaction potentials of the system using the available DFT calculations, and then, on their basis, to model the required properties by the molecular dynamics method on a multiply increased spatiotemporal scale. As a starting point, we introduce machine learning principles, algorithms, descriptors, and databases in materials science. The design of the potential energy surface and interatomic interaction potentials in solid solutions, high-entropy alloys, high-entropy metal compounds with carbon, nitrogen, and oxygen, as well as in bulk amorphous materials, is described.