Boundary of hard and soft zones of a partially recrystallized alloy determines the strength‑ductility trade‑off: a machine learning study

In physical metallurgy, partially recrystallized microstructures show great potential for performance improvement through tailored thermo-mechanical processing. It is widely accepted that increasing the recrystallized faction generally reduces strength while enhancing ductility. However, deviations from this trend have long been recognized, stemming from competing effects of texture, dislocations, grain boundaries, and precipitates. These anomalies also raise an important question: can hidden factors beyond recrystallized fraction and aforementioned factors break the strength-ductility trade-off? Using machine learning, we identify the boundary between soft recrystallized and hard non-recrystallized regions as the determinant to the mechanical properties of partially recrystallized alloys. Through micro-digital image correlation (μDIC), we verify that a pronounced strain gradient is generated near this boundary, which contributes to additional strain hardening capacity and ductility. The increased boundary density thus overcomes the strength-ductility trade-off in partially recrystallized alloys.