Machine Learning Enhanced Multiscale Material Mechanics Modeling for Complex Alloys

Complex alloys, including chemically disordered high-entropy alloys (HEAs) and structurally disordered metallic glasses (MGs), have drawn significant attention due to their broad tunability and exceptional properties, with applications ranging from structural materials to damage-resistant systems. However, predictive modeling of these alloys remains challenging because of their complex atomic environments, non-equilibrium states, and dynamic evolution under load. This talk will first focus on chemically disordered HEAs, which are formed by mixing multiple elements in near-equal proportions. We investigate how chemical environments and local ordering states govern deformation mechanisms and dislocation behavior. Machine learning (ML) interatomic potentials reproduce defect energetics and dislocation core structures with near-quantum accuracy, while graph neural network–accelerated Monte Carlo simulations efficiently probe chemical ordering states. High-throughput simulations further reveal how tuning diffuse antiphase boundary energies and lattice distortion modulates screw-to-edge slip resistance, offering pathways to balance strength and ductility in NbMoTaW HEAs. For structurally disordered MGs, which lack long-range order and deform through localized atomic rearrangements, we employ multiscale mechanics models that embed atomic-level disorder into coarse-grained frameworks. ML models parameterize flow defects and feed into the shear transformation zone dynamics model, enabling us to capture collective deformation and shear band formation. Particular emphasis is placed on how nanoscale heterogeneity, arising from short- and medium-range order, influences the nucleation and propagation of shear bands. Together, these case studies illustrate how ML-enhanced multiscale modeling bridges atomic disorder to macroscopic performance, accelerating the design and discovery of high-performance structural materials across vast configurational and compositional spaces.