Identification of atomic rearrangements in amorphous alloys based on machine learning

Abstract

The disordered structure inherent in amorphous alloys precludes the existence of well-defined structural defects analogous to crystals. In this investigation, a novel machine-learning parameter termed Structural Atomic Rearrangement (SAR) is formulated. SAR integrates structural parameters, thermodynamic vibrational entropy, and kinetic activation energy. This departure from the conventional reliance solely on microstructure for macroscopic property characterization enables SAR to comprehensively and accurately identify atomic rearrangements. In the context of deformation, SAR proves effective in pinpointing local regions undergoing plastic rearrangement, offering a distinctive signal preceding the formation of shear bands. This innovative approach challenges the conventional notion and underscores the capability of SAR in capturing rearranged atoms in a more nuanced manner than conventional structural parameters alone.