Predicting atomic structure and mechanical properties in quinary L1₂-Strengthened cobalt-based superalloys using machine learning-driven first-principles calculations

L1₂-strengthened Cobalt (Co)-based superalloys are promising high-temperature materials for aero-engine applications. To make first-generation Co-Al-W-based superalloys industrially viable, it's crucial to enhance the mechanical properties and solvus temperature of the metastable L1₂ phase. Introducing additional transition metal (TM) elements into the FCC matrix is a promising strategy. Although first-principles calculations are invaluable for materials design, their high computational cost and low-efficiency for the multi-component systems, particularly those doped with TM elements, limit their practical use. In this study, we combine machine learning with first-principles calculations to accelerate the predictions of atomic structure and mechanical property. Using datasets from first-principles calculations, our ML models predict the trend in element occupancy, doping position, and mechanical attributes of the L1₂ phase. The ML models, further refined with first-principles data, efficiently predict properties for Nb-doped systems, outperforming traditional counterparts. This methodology expedites calculations and promises advancements in designing various advanced materials, including multiple-principal-element alloys.