无机晶体声子数据库

该数据集是由橡树岭国家实验室创建的，包含近5000种无机晶体，每个晶胞包含最多12个原子。数据集旨在为评估通用机器学习原子间势（uMLIPs）的性能提供一个基准。与现有声子数据库相比，该数据库更全面地覆盖了具有相对简单晶体结构、较大布里渊区和丰富声子色散特征的无机晶体，以反映微妙的原子间相互作用。数据集创建过程从较小的晶胞开始，逐步构建，最终包含86种元素。该数据集的应用领域在于解决实时分析和解释非弹性中子散射数据的问题。

This dataset was developed by Oak Ridge National Laboratory, containing nearly 5000 inorganic crystals, each with up to 12 atoms per unit cell. Its core purpose is to serve as a benchmark for evaluating the performance of universal machine learning interatomic potentials (uMLIPs). Compared with existing phonon databases, this dataset more comprehensively covers inorganic crystals featuring relatively simple crystal structures, large Brillouin zones, and abundant phonon dispersion characteristics that reflect subtle interatomic interactions. The dataset was constructed through a step-by-step process starting from smaller unit cells, ultimately encompassing 86 distinct chemical elements. This dataset is designed to address the challenges of real-time analysis and interpretation of inelastic neutron scattering data.