Accelerated Discovery of High Entropy Alloys with Breakthrough Hardness via Inverse Design Strategy

High entropy alloys (HEAs) have gained substantial attention owing to their excellent properties. Nevertheless, identifying HEAs with high hardness from the extensive compositional space remains a challenging task. In this work, we proposed a machine learning based inverse design strategy combining with a self-developed proactive searching progress method to accelerate the discovery of HEAs with enhanced hardness. Three recommended candidates with predicted high hardness values were synthesized by experiments. The results validated that two of the three designed HEAs Cr_17.7Fe_20.9Ni_20.2Ti_22.2V_19.0 and Al_31.1Co_29.8Cr_2.4Cu_0.1Fe_10.8Ti_17.0V_8.8 exhibited hardness values exceeding 1000 HV. Notably, Cr_17.7Fe_20.9Ni_20.2Ti_22.2V_19.0 demonstrated a hardness of 1177 HV, surpassing the maximum hardness in the original dataset. The SHAP analysis reveals that the d-valence electron concentration (${\bar{e}}^d$) is one of the significant factors influencing hardness, and it has a positive impact on hardness when ${\bar{e}}^d$ is below 5.4. This work proved the feasibility of our strategy in developing new HEAs with breakthrough hardness, which might be instructive to other material fields.