Enhancing strength-ductility synergy in BCC refractory high-entropy alloys via machine learning and bimodal heterostructure design

The brittleness of most refractory high-entropy alloys (RHEAs) severely limits their structural applications. Given the vast compositional space, we employ machine learning combined with the Pareto frontier optimization to screen for promising Ti-V-Nb-Ta compositions, followed by building up a bimodal heterostructure to enhance both yield strength and ductility. Guided by the Pareto frontier algorithm, Ti₂₄V₃₂Nb₁₆Ta₂₈ and Ti₂₈V₃₄Nb₁₆Ta₂₂ alloys were developed, exhibiting high yield strengths (745 MPa and 747 MPa) and notable tensile elongations (21.1% and 24.9%), respectively. During tensile deformation, the as-cast alloys activate multiple deformation mechanisms, including dislocation glide, twinning, and kinking. The synergistic operation of these mechanisms effectively relieves local stress concentrations and contributes to the improved ductility. Following cold rolling and annealing, the optimized heterostructured Ti₂₄V₃₂Nb₁₆Ta₂₈ alloy achieves a yield strength of 931 MPa and an elongation of 31.1%, corresponding to increases of 25% and 47% over the as-cast state. The improvement stems from the coordinated deformation between coarse and fine grains. Specifically, the fine-grained regions contribute to strength by hindering dislocation slip, while the coarse grain regions act as a plastic buffer, activating multiple slip systems and {112}<111> mechanical twins to coordinate local deformation, relieve stress concentrations, and prevent premature cracking. This synergistic interaction suppresses strain localization and delays crack propagation, ultimately enabling the alloy to achieve both high yield strength and high ductility. Our design strategy provides an effective pathway to achieve good strength-ductility synergy in RHEAs.