Predicting the pathways of string-like motions in metallic glasses via path featurizing graph neural networks

String-like motions (SLMs) cooperative, “snake”-like movements of
particles—are crucial for dynamics in diverse glass formers.
 Despite their ubiquity, questions persist: do SLMs prefer
specific paths? If so, can we predict these paths? Here, in Al-Sm glasses,
our iso-configurational ensemble simulations reveal that SLMs indeed
follow certain paths. By designing a graph neural network (GNN) to
featurize the environment around directional paths, we achieve a
high-fidelity prediction of likely SLM pathways solely based on the static
structure. GNN gauges a structural measure to assess each path’s
propensity to engage in SLMs, akin to a “softness” metric, but for paths
rather than for atoms. Our GNN interpretation reveals the critical role of
the bottleneck zone along paths in steering SLMs. By monitoring
“path-softness”, we elucidate SLM-favored paths transit from fragmented to
interconnected upon glass transition. Our findings reveal that, beyond
atoms or clusters, glasses have another dimension of structural
heterogeneity: “paths”.

类弦运动（String-like motions, SLMs）是粒子协同完成的"蛇形"运动，是各类玻璃形成体中动力学行为的关键基础。尽管这类运动普遍存在，但仍有核心问题待解：类弦运动是否偏好特定路径？若存在偏好路径，能否对其进行预测？本研究针对铝钐（Al-Sm）玻璃体系，通过等构型系综（iso-configurational ensemble）模拟证实，类弦运动确实会沿特定路径行进。我们设计了一种图神经网络（Graph Neural Network, GNN）对定向路径周围的环境进行特征提取，仅通过静态结构信息即可实现类弦运动潜在路径的高精度预测。该图神经网络通过量化一种结构指标，来评估每条路径参与类弦运动的倾向性，这类似于针对路径而非原子的"软度"度量指标。对该图神经网络的可解释性分析表明，路径沿线的瓶颈区域在调控类弦运动中发挥着关键作用。通过监测"路径软度"，我们阐明了在玻璃化转变过程中，类弦运动偏好的路径会从碎片化状态转变为互联连通状态。本研究结果表明，除原子与团簇之外，玻璃体系还存在另一维度的结构非均匀性：即"路径"维度。