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 d..., This dataset contains the string-like motion (SLM) probability data of Al90Sm10 metallic glasses that are established by the iso-configurational ensemble simulations. There are a total of 61 independent Al90Sm10 samples that are quenched to and relaxed at 400 K. Additional details, including the train/val/test splits used in the paper, can be found in \"Readme.txt\"., , # Predicting the Pathways of String-like Motions in Metallic Glasses via Path-Featurizing Graph Neural Networks [https://doi.org/10.5061/dryad.2z34tmptt](https://doi.org/10.5061/dryad.2z34tmptt) This dataset contains the string-like motion (SLM) probability data of Al90Sm10 metallic glasses that are established by the iso-configurational ensemble simulations. ## Description of the data and file structure There are a total of 61 independent Al90Sm10 samples that are quenched to and relaxed at 400 K. **Files of each sample directory:** **[data.dump]** The glass configuration in the format of LAMMPS dump. Atom type 1 is for Al, 2 is for Sm. **[string_probability.csv]** The string-like motion (SLM) probability data of each glass configuration. The columns are [\"source_id\", \"final_id\", \"atom_types\", \"string_probability\"] \"source_id\": the source id of the string. \"final_id\": the final id of the string. For example, if source_id is 103 and final_id is 107, this indicates a string s...类串运动（String-like Motions, SLMs），即粒子协同的“蛇形”运动，是多种玻璃形成体系动力学过程的核心机制。尽管这类运动普遍存在，但仍有核心问题待解：类串运动是否偏好特定路径？若存在偏好，能否对这些路径进行预测？ 本研究针对铝钐（Al-Sm）玻璃体系，通过等构型系综模拟发现，类串运动确实遵循特定路径。我们设计了图神经网络（Graph Neural Network, GNN）对定向路径周围的环境进行特征化，仅基于静态结构即可实现高精度的类串运动潜在路径预测。该图神经网络可量化一种结构指标，以评估每条路径参与类串运动的倾向，这类似于针对路径而非原子的“柔软度”度量标准。通过对图神经网络的解释分析，我们揭示了路径沿线的瓶颈区域在调控类串运动中的关键作用。通过监测“路径柔软度”，我们阐明了玻璃转变过程中，类串运动偏好的路径从碎片化状态转变为互联状态的过程。本研究发现，除原子与团簇外，玻璃体系还存在另一类[原文截断]。 本数据集包含通过等构型系综模拟得到的Al₉₀Sm₁₀金属玻璃的类串运动概率数据。 数据集共包含61个独立的Al₉₀Sm₁₀样品，所有样品均经淬火处理并在400 K下完成弛豫。 更多细节（包括论文中使用的训练/验证/测试集划分方案）可查阅"Readme.txt"文件。 # 基于路径特征化图神经网络预测金属玻璃的类串运动路径 [https://doi.org/10.5061/dryad.2z34tmptt] 本数据集包含通过等构型系综模拟得到的Al₉₀Sm₁₀金属玻璃的类串运动概率数据。 ## 数据与文件结构说明 数据集共包含61个独立的Al₉₀Sm₁₀样品，所有样品均经淬火处理并在400 K下完成弛豫。 **每个样品目录下的文件：** **[data.dump]** 该文件为LAMMPS dump格式的玻璃体系构型文件，其中原子类型1代表铝（Al）原子，类型2代表钐（Sm）原子。 **[string_probability.csv]** 该文件存储了每个玻璃构型的类串运动概率数据。 文件列字段为["source_id", "final_id", "atom_types", "string_probability"] - "source_id": 串的起始ID - "final_id": 串的终止ID。例如，若source_id为103且final_id为107，则表示一条串的[原文截断]