Assessing Microstructure Phase Maps (AMASE)

Phase diagrams are “treasure maps” in materials innovation. However, they traditionally assume defect-free materials, whereas real-world microstructures are often dominated by defects––such as grain boundaries, phase boundaries, dislocations, and stacking faults––that have their own phase behaviours and distinct rules for evolving, interacting, and co-existing. This discrepancy significantly ties materials innovation to slow, trial-and-error approaches. Project AMASE is envisioned to deliver “roadmaps”, introducing two novel concepts of Multi-Defect Phase Diagrams and Microstructure Phase Maps for accurate microstructure predictions.AMASE will combine atomistic simulations, machine learning, thermodynamics, and multi-phase-field simulations via a novel CALPHAD-integrated density-based concept. These will be realised through three pillars: first, bridging atomistic simulations, coarse-graining, and machine learning analyses to develop Representative Field Variable(s) that unify descriptions of various defects; second, developing CALPHAD-integrated free energy functionals, iterated with a machine learning framework, and used to generate Multi-Defect Phase Diagrams; and third, spatiotemporal mapping of various microstructures by coupling the results of the first two pillars with a multi-phase-field approach to obtain Static and Dynamic Microstructure Phase Maps. These aims are closely entangled with three critical engineering challenges: (i) mitigating liquid metal embrittlement in steels, (ii) reducing hydrogen embrittlement in Al-alloys, and (iii) improving the formability of Mg-alloys.Built on the PI’s pioneering contributions in defect thermodynamics and scale-bridging methods, AMASE will deliver a scalable predictive toolkit compatible with widely used platforms such as Thermo-Calc, pyCALPHAD, and OpenPhase, promising to significantly improve development cycles and setting a new paradigm that offers transformative solutions for high-impact industrial challenges.