DFT dataset for high entropy materials

V2023.04.06 DFT dataset:

This DFT dataset contains about 84k structures, including ordered structures with no more than 8 atoms and SQSs with 27, 64, or 125 atoms. See our paper (publisher version; arXiv version) and Github for more details. 

The csv file hea.2023-04-06.csv contains both the initial unrelaxed and the final relaxed structures, the energies of relaxed structures, and various attributes (atomic magnetic moment and charge, volume, space group number etc.). SROs_structure_ini.csv contains the short-range order parameters of the initial structures. See explore_data.py for data loading and an overview of the data distribution.

The csv files structure_featurized.dat_all.csv and structure_ini_featurized.dat_all.csv contain the Matminer features ready for training descriptor-based machine learning models. See demo_ML_training.py for an example. 

Note: The trajectory data (energies and forces for structures during the DFT relaxations) is not published with this paper; it will be released later with our future work on machine learning force fields for HEMs.

版本V2023.04.06的密度泛函理论（DFT）数据集：
本数据集共包含约8.4万个结构，涵盖原子数不超过8的有序结构，以及原子数分别为27、64或125的特殊准随机结构（SQSs）。更多细节可参阅我们的论文（正式出版版；arXiv预印本版）及GitHub仓库。
文件hea.2023-04-06.csv同时收录初始未弛豫结构与最终弛豫结构、弛豫后结构的能量，以及各类属性参数（原子磁矩与电荷、晶胞体积、空间群编号等）。SROs_structure_ini.csv包含初始结构的短程序参数。可参考explore_data.py文件了解数据加载方式与数据分布概览。
文件structure_featurized.dat_all.csv与structure_ini_featurized.dat_all.csv包含可直接用于训练基于描述符的机器学习模型的Matminer特征。示例代码可参阅demo_ML_training.py。
注：本数据集未附带轨迹数据（即DFT弛豫过程中结构的能量与原子受力数据）；该数据将随我们后续针对高熵合金（HEMs）的机器学习力场相关研究一同发布。