Rethinking glass structure beyond short-range order: A perspective from atomistic modeling and experimental studies

Abstract

Metallic glasses challenge the classical structure–property doctrine because the lack of long-range periodicity makes their structures complex and inherently multiscale. Although short-range order (SRO) within the first-neighbor cage has provided a useful vocabulary (e.g., icosahedral versus liquid-like motifs), growing evidence shows that SRO alone is not a reliable state variable for properties, such as stability and mechanical performance. In particular, the same nominal SRO motif can correspond to distinct energetic states depending on its embedding at more extended length scale, and local propensity indicators do not uniquely predict whether local plastic activity remains isolated or evolves into avalanches and extensive plastic deformation. We argue that rethinking glass structure requires shifting from per-atom SRO coordinates toward collective variables, defined over the range of relevant cooperativity. Within this framework, we clarify two kinds of medium-range order (MRO) that are frequently conflated in the literature at the intermediate scales (around second nearest- neighbor shell, 0.5–1 nm). One is the topological MRO, manifested as SRO motif connectivity/networking in atomistic models or as crystal-like nanodomains, probed by fluctuation electron microscopy and 4D-STEM. Another is defined by the persistent oscillations of pair distribution functions (i.e., the density wave fluctuations). Their coherence length is also about the same scale, and provides an alternative, more general view of the MRO. This motivates an energy landscape-based viewpoint in which density-fluctuation fields do not map to single basins, but condition the statistical accessibility of basins and activation barriers, thereby shaping the rigidity network that steers cascade propagation. We conclude by outlining quantitative opportunities to connect these scale-dependent collective descriptors, response-based diffraction, and energy landscape sampling into predictive, experimentally anchored structure–property relations. Data-driven atomistic analysis of metallic-glass structures using smooth overlap of atomic positions (SOAP) descriptors and transparent machine learning to identify radius of informative structural environments (RISE) Data-driven atomistic analysis of metallic-glass structures using smooth overlap of atomic positions (SOAP) descriptors and transparent machine learning to identify radius of informative structural environments (RISE)