Accelerated discovery through multi-property screening strategy: Achieving strength-ductility synergy in TiZrHfNbMoTa RHEAs

Abstract

Refractory high-entropy alloys (RHEAs) exhibit excellent potential for high-temperature applications. However, their vast compositional space and the challenge of optimizing multiple properties impede further development. To address these challenges, this study integrates high-throughput computation and machine learning to identify compositions that meet multiple criteria, including density, strength, ductility, and resistance to high-temperature softening. A Bayesian multi-property optimization framework is employed to efficiently enhance both compressive yield strength and ductility. After three iterative cycles, the optimized alloys exhibit favorable combined yield strength and ductility compared to the RHEAs used for model training. Among these, the Ti₂₈Zr₁₅Hf₂₀Nb₁₄Mo₁₆Ta₇ alloy demonstrates notable overall properties, including a compressive yield strength of 1327 ± 24 MPa, ductility of 37.9 ± 1 %, and specific yield strength at room temperature of 155.9 MPa·g⁻¹·cm³. Additionally, it exhibits excellent resistance to high-temperature softening and low density.