A universal Mn/Fe criterion for Fe-IMCs morphology control in Al–Si casting alloys across a wide range of Fe levels by computational thermodynamics and machine learning

Lightweighting via aluminum alloys is essential for achieving carbon neutrality, and the increasing adoption of large-scale aluminum recycling demands a higher tolerance for Fe impurities. In Al-Si casting alloys, Fe forms brittle intermetallic compounds (Fe-IMCs). Mn is commonly added to transform the detrimental needle-shaped β-Al 4.5 FeSi (β-Fe) or δ-Al 3 FeSi 2 (δ-Fe) phases into less harmful script-like α-Al 15 (Fe, Mn) 3 Si 2 (α-Fe) phase, but insufficient Mn leaves these phases largely unmodified, while excessive Mn leads to sludge formation. In this paper, we established a universal Mn/Fe criterion to control Fe-IMCs morphology by combining computational thermodynamics with machine learning. A customized thermodynamic database for the Al-Si-Mg-Fe-Mn quinary system was developed and validated against experimental phase equilibria and solidification behavior. High-throughput solidification simulations over a wide composition range were performed, and the results, combined with the experimental observations from literature and newly designed A356-based alloys, reveal that needle-shaped Fe-IMCs were fully suppressed when the volume fraction ratio of α-Fe to β-Fe phases exceeds 4 or that of α-Fe to δ-Fe phases exceeds 8. A machine learning model that uses simulated volume fractions of Fe-IMCs as input features was then employed to construct formation diagrams that delineate the stability regions of different Fe-IMCs types, from which a quantitative relationship linking optimal Mn content to Fe and Si levels was derived as w (Mn)=1.242× w (Fe)-0.122× w 2 (Fe)+0.001× w (Si)-0.035× w (Si)× w (Fe) (in wt.%). It is anticipated that this approach may provide a practical, composition-based guideline for tailoring Fe-IMCs types and designing compositions in Al-Si casting alloys with a wide range of Fe levels.