Machine Learning Discovery in High Entropy Materials—Derivatives of Cantor Alloys and Complex Glasses

High entropy materials, including alloys (HEAs) and metallic glasses (HEMGs), are at the forefront of materials science due to their exceptional properties derived from compositional complexity. This work employs machine learning (ML) to uncover and refine the properties of these materials, demonstrating significant advancements in HEAs and initiating the exploration of HEMGs. For HEAs, we applied Bayesian optimization with compositional constraints, enhancing the mechanical properties of the Cantor alloy and discovering compositions like Fe21Cr20Mn5Co20Ni34 and Fe6Cr22Mn5Co32Ni35, which show in silico marked improvements in "single crystal" yield stress [1]. This success showcases the potential of ML in optimizing the complex compositional landscape of HEAs and paves the way for the design of several material properties simultaneously. The brittleness and ductility of metallic glasses are of high interest. Extending this methodology to HEMGs, we will show results on the optimization of the mechanical properties for the system ZrCuAl. The data driven approach allows to map yield properties and elastic modulus as non linear function of the composition and cooling rate. Our ML-driven approach enhances HEAs and gives the basis for innovative advancements in HEMGs, showcasing the potential of ML and Bayesian methods in the discovery and optimization of high entropy materials for diverse applications.<br/>[1] Torsti, V., Mäkinen, T., Bonfanti, S., Koivisto, J., & Alava, M. J. (2024). Improving the mechanical properties of Cantor-like alloys with Bayesian optimization. APL Machine Learning, 2(1)