Empowering the Sustainable Development of High-End Alloys via Interpretive Machine Learning.

The extensive use of scarce, expensive, and toxic elements in high-performance metal alloys restricts their sustainable development. Here we propose a novel alternative alloying-element design strategy that combines physicochemical-factor screening, a "black-box" interpretative method based on SHapley Additive exPlanation analysis, and sensitivity analyses of elemental influence. A "white-box" model of alloy compositions and properties is therefore established that enables the rational selection of abundant elements and the efficient designs of alloys with substitution for scarce alloying elements. The success of this design strategy is demonstrated by reducing the Co content in the C70350 alloy series (e.g., Cu-1.3Ni-1.4Co-0.56Si-0.03Mg (wt.%)). Indeed, Cu-1.95Ni-0.5Co-0.6Si-0.2Mg-0.1Cr (wt.%) is obtained as an ultra-low-Co-containing alloy by substituting only a trace amount of Cr for Co in the Cu-Ni-Co-Si-Mg system, followed by compositional optimization. Although the Co content is reduced by 64% (i.e., from 1.4 to 0.5 wt.%), the properties of the alloy (ultimate tensile strength and electrical conductivity of 850 MPa and 47.2%IACS, respectively) are comparable to those of the C70350 alloy. This study contributes to the sustainable and green development of metallic materials by providing a new avenue for substituting abundant elements for scarce and undesired elements in metal alloys.