Effect of Cooling Rate and Composition in the Mechanical Properties and Avalanche Dynamics of CuZrAl Metallic Glasses—Insights from Hybrid Simulations and Machine Learning Potentials

This study investigates the role of cooling rate and composition in the mechanical properties and avalanche dynamics of CuZrAl metallic glasses using advanced computational techniques.<br/>We employ a hybrid simulation approach combining Monte Carlo and Molecular Dynamics (MD) methods to generate equilibrated samples at temperatures below the conventional glass transition. This enables us to achieve a stable glassy regime, facilitating a detailed examination of the kinetics, thermodynamics, and rheology of CuZrAl glasses.<br/><br/>Our findings reveal an enhanced understanding of the stability and behavior of the CuZrAl system, and we observe abrupt stress drops corresponding to shear band precursors and dynamics. We further utilize machine learning interatomic potentials specifically developed for the Zr-Cu-Al system, trained on a comprehensive Density Functional Theory database. This model allows for large-scale MD simulations with near ab initio accuracy, validating the MLIP model's accuracy through comparison with other computational techniques and providing a comparative analysis showcasing the prediction of elastic properties.<br/><br/>Additionally, we explore avalanche dynamics and local structural arrangements in Zr-Cu-Al metallic glasses under shear. By systematically varying the ratio of the elements, we examine the relation between composition and local configurations in shear transformation zones. Our findings on avalanche dynamics reveal that avalanches are universal in the steady state, and unlike Lennard-Jones systems, CuZrAl shows avalanches approaching the peak stress. The gap distribution is flat in the steady state with a power-law exponent in the small strain regime.<br/><br/>Overall, we propose an integrated approach to advance our understanding of complex metallic glasses and provides comparisons with experimental data.