| Abstract Scope |
This paper presents an iterative computational/experimental approach for the rapid discovery of novel alloys in the Ti-V-Nb-Mo-Hf-Ta-W refractory space. We employ a Bayesian material design cycle to simultaneously maximize high specific hardness and high specific elastic modulus. High-throughput computational thermodynamics and intelligent filtering reduce the vast alloy space, followed by iterative synthesis, processing, and characterization of 24-alloy batches. Bayesian optimization guided the selection of subsequent batches, yielding a 34% increase in the Pareto front hypervolume after just two iterations, with 7 alloys in the second batch outperforming all from the first. This work demonstrates the power of batch Bayesian optimization (BBO) for materials design without relying on prior physical models. |