Multi-principal element alloys, Machine learning, Lattice distortion, Compositional design, Yield strength, Mining engineering. Metallurgy, TN1-997

Screening alloys with excellent strength and ductility in the huge compositional space of multi-principal element alloys (MPEAs) is challenging. This study combines thermodynamic calculations, first principles calculations, and machine learning (ML) techniques to design CoNiCrMo MPEAs with high lattice distortion and good face-centered cubic (FCC) phase stability in order to achieve excellent mechanical properties. Mean square atomic displacement (MSAD) serves as a key parameter for obtaining high lattice distortion, and solvus temperature (ST) is used as a key indicator to evaluate the stability of FCC phases, which are the two main indicators we focus on in alloy screening using ML methods. The results indicate electronegativity difference is the descriptor of the highest importance in the lattice distortion prediction of the ML model. Meanwhile, the MSAD value is basically the same as the electronegativity difference value with compositional variation, which provides an important reference for alloy design. The lattice friction stresses and the Hall-Petch coefficients of the designed Ni60Cr35Mo5 alloy are 189.9 MPa and 749.0 MPa μm−1/2, respectively. The synergistic effect of Cr and Mo elements can better improve the MSAD value and solid solution strengthening effect, for larger average bond distances and standard deviations. The method proposed in this study not only provides a fast and efficient design pathway for the development of MPEAs, but also has wide applicability to other metallic materials.