Polyanion sodium cathode materials dataset

We have created a polyanion sodium cathode materials dataset that includes optimizations of structures to the lowest energy, ab initio molecular dynamics simulations trajectories sampled at 1000K, and structures generated from ML-driven molecular dynamics simulation at 1000K using active learning algorithms. The dataset consists of structures sampled from four sodium ion polyanionic cathode materials NaTMPO₄(olivine) ,NaTMPO₄(maricite), Na₂TMSiO₄ and Na_2.56TM_1.72(SO₄)₃ , along with various structures incorporating doping of transition metal ions (TM). We consider four different transition metal ions (Fe, Mn, Co, Ni).The dataset consist of 113,532 structures with atomic charge and 184,612 structures without atomic charge.For each sampled structure, we record its crystal composition, total energy, atom-wise force vectors, atom-wise magnetic moments derived from Mulliken analysis, and the atomic charges are obtained through Bader analysis when possible. Our polyanion sodium cathode materials dataset serves as a valuable addition to existing datasets, enabling the exploration of phase space while providing insights into the dynamic behavior of the materials.The dataset is presented in XYZ format. The dataset is divided into single transition metal ions structures and multiple transition metal ion structures. This division is provided for each of the four cathode materials: NaTMPO₄(olivine) ,NaTMPO₄(maricite), Na₂TMSiO₄ and Na_2.56TM_1.72(SO₄)₃ . For example, Na_2.56TM_1.72(SO₄)₃ structures are split into single transition metal ion types<i> Na2M2SO4_alluadite_single_total.xyz </i>and multiple transition metal ion types <i>Na2M2SO4_alluadite_multiple_optimization.xyz. </i>The single TM ion structures are further separated into the sampling method used to gather the dataset, structure optimization <i>Na2M2SO4_alluadite_single_optimization.xyz</i>, AIMD simulation <i>Na2M2SO4_alluadite_single_AIMD.xyz</i>, and ML driven MD sampling <i>Na2M2SO4_alluadite_single_MD_sampling.xyz</i>. This allows the user to either use the whole dataset or only structures gathered using a specific sampling method.The atomic charge dataset is based in the <i>polyanion_cathode_dataset </i>folder and the combined dataset , consisting of 113,532 structures, is available in <i>polyanion_cathode_dataset</i>/<i>Combined.xyz.</i><br>The dataset, which do not include atomic charges but retain all other physical properties, is stored in the <i>polyanion_cathode_dataset_with_optimization_steps</i> folder and is divided into single TM ions structures and multiple TM ion structures, categorized separately for each of the four cathode materials. For example, the structures without atomic charge from the structural optimization process of Na_2.56TM_1.72(SO₄)₃single TM ion structures are collected in <i>polyanion_cathode_dataset_with_optimization_steps/Na2M2SO4_alluadite_single_structural_optimization.xyz</i>.The complete dataset, including all structures with atomic charge as well as the structures without atomic charge obtained during structural optimization (298,144 structures), is available as <i>polyanion_cathode_dataset_with_optimization_steps/Combined.xyz</i>.<br>To extract structural compositions and physical properties, the ase.io.read function from ASE version 3.23.0 is used. An example of how to extract data and plot the physical properties is provided in https://github.com/dtu-energy/cathode-generation-workflow/tree/main/extract_data/read_data.py and https://github.com/dtu-energy/cathode-generation-workflow/tree/main/extract_data/utils.py contains two functions, one used to attached Bader charges to an ASE atom object an another to combine multiple XYZ data files.<br>To cite the data please use the doi https://doi.org/10.11583/DTU.27202446<br>For the sampling density functional theory (DFT) calculation were performed using the Vienna Ab initio simulation package (VASP) version 6.4. The Perdew-Burke-Ernzerhof (PBE) functional with Hubbard-U corrections were applied was utilized for all calculations. The U-values are similar to the ones used for materials project (Fe: 5.3eV, Mn: 3.9eV, Co: 3.32eV, Ni: 6.2eV). For all calculations, an energy cutoff of 520eV was applied, with a smearing width of 0.01eV and convergence criteria set to 1e-5eV for energy and 0.03eV/Å for forces. All calculations were performed with spin polarization. The k-points employed for the four materials were fixed, with NaTMPO₄(olivine) and NaTMPO₄(maricite) utilizing [3,4,6] gamma points, Na₂TMSiO₄employing [3,4,4] gamma points and Na_2.56TM_1.72(SO₄)₃ utilizing [2,3,4] gamma points. When constructing supercells, the gamma point in the direction of cell enlargement was halved.All molecular dynamics (MD) simulations are conducted using the Langevin thermostat with a friction constant of 0.003. The temperature is maintained at 1000K to facilitate diffusion events, and a time step of 1fs as employed throughout the simulations. All simulations are executed within the canonical (NVT) ensemble and a sample frequency was set to 1fs.<br>Versions:<br>3: The dataset consist purely on structures with atomic charge4: The dataset is divided into two parts, one with structures with atomic charge and with a subset of the structures without atomic charge<br>5: Finished things up for the publication. Clean the dataset a bit and changed the sign of the atomic charges to be in the unit of e instead of -e. ReadMe file and description has also been updated.

本团队构建了一款聚阴离子型钠离子正极材料数据集，涵盖经结构优化至最低能量的构型、1000K下采样的从头算分子动力学（ab initio molecular dynamics, AIMD）模拟轨迹，以及基于主动学习算法在1000K下通过机器学习驱动分子动力学生成的构型。
该数据集的构型采样自四种钠离子聚阴离子正极材料：NaTMPO₄（橄榄石型，olivine）、NaTMPO₄（磷钠锰铁矿型，maricite）、Na₂TMSiO₄以及Na₂.₅₆TM₁.₇₂(SO₄)₃，同时包含各类过渡金属离子（transition metal, TM）掺杂的构型。
本数据集选取了四种不同的过渡金属离子：Fe、Mn、Co与Ni。
该数据集包含113532条带原子电荷的构型，以及184612条不带原子电荷的构型。
针对每条采样构型，我们记录了其晶体组分、总能量、原子受力矢量、通过马利肯布居分析（Mulliken analysis）得到的原子磁矩；若条件允许，还通过贝德分析（Bader analysis）获取原子电荷。
本聚阴离子钠离子正极材料数据集可作为现有数据集的重要补充，既支持相空间探索，又能为材料的动态行为研究提供参考。
数据集采用XYZ格式存储。
数据集按过渡金属离子种类分为单过渡金属离子构型与多过渡金属离子构型两类，且该分类覆盖上述四种正极材料。
例如，Na₂.₅₆TM₁.₇₂(SO₄)₃的构型被分为单过渡金属离子类型文件`Na2M2SO4_alluadite_single_total.xyz`与多过渡金属离子类型文件`Na2M2SO4_alluadite_multiple_optimization.xyz`。
单过渡金属离子构型还可按采样方法进一步分类：结构优化构型文件`Na2M2SO4_alluadite_single_optimization.xyz`、从头算分子动力学模拟构型文件`Na2M2SO4_alluadite_single_AIMD.xyz`，以及机器学习驱动分子动力学采样构型文件`Na2M2SO4_alluadite_single_MD_sampling.xyz`。用户可选择使用完整数据集，或仅调取特定采样方法得到的构型。
带原子电荷的数据集存储于`polyanion_cathode_dataset`文件夹中，该文件夹内的合并数据集（包含113532条构型）可通过`polyanion_cathode_dataset/Combined.xyz`获取。不含原子电荷但保留其余所有物理属性的数据集存储于`polyanion_cathode_dataset_with_optimization_steps`文件夹，同样按单过渡金属离子构型与多过渡金属离子构型分类，且针对四种正极材料分别归档。例如，Na₂.₅₆TM₁.₇₂(SO₄)₃的单过渡金属离子构型中，来自结构优化过程的不带原子电荷的构型被收录于`polyanion_cathode_dataset_with_optimization_steps/Na2M2SO4_alluadite_single_structural_optimization.xyz`。完整数据集（包含所有带原子电荷的构型，以及结构优化过程中得到的不带原子电荷的构型，总计298144条）可通过`polyanion_cathode_dataset_with_optimization_steps/Combined.xyz`获取。
可使用原子模拟环境（Atomic Simulation Environment, ASE）3.23.0版本中的`ase.io.read`函数提取构型组分与物理属性。数据提取及物理属性可视化的示例代码可参考https://github.com/dtu-energy/cathode-generation-workflow/tree/main/extract_data/read_data.py；https://github.com/dtu-energy/cathode-generation-workflow/tree/main/extract_data/utils.py包含两个工具函数，其一用于为ASE原子对象添加贝德电荷，其二用于合并多条XYZ格式数据文件。
引用该数据集请使用DOI：https://doi.org/10.11583/DTU.27202446
本数据集的采样基于密度泛函理论（density functional theory, DFT）计算，所有计算均采用维也纳从头算模拟包（Vienna Ab initio Simulation Package, VASP）6.4版本完成。计算中使用了引入Hubbard-U修正的Perdew-Burke-Ernzerhof（PBE）泛函，U值设置与材料项目（Materials Project）一致：Fe为5.3eV、Mn为3.9eV、Co为3.32eV、Ni为6.2eV。所有计算均设置520eV的截断能、0.01eV的展宽宽度，能量收敛标准为1×10⁻⁵eV，受力收敛标准为0.03eV/Å，且均开启自旋极化。四种材料的k点设置固定：NaTMPO₄（橄榄石型）与NaTMPO₄（磷钠锰铁矿型）采用[3,4,6]γ点，Na₂TMSiO₄采用[3,4,4]γ点，Na₂.₅₆TM₁.₇₂(SO₄)₃采用[2,3,4]γ点。构建超胞时，沿晶胞扩大方向的γ点数量减半。
所有分子动力学（molecular dynamics, MD）模拟均采用朗之万恒温器（Langevin thermostat）完成，摩擦常数设置为0.003。模拟温度维持在1000K以促进扩散过程，时间步长设为1fs，所有模拟均在正则系综（canonical ensemble, NVT）下进行，采样频率设为1fs。
版本说明：
版本3：仅包含带原子电荷的构型
版本4：数据集分为两部分，其一为带原子电荷的构型，其二为部分不带原子电荷的构型
版本5：完成出版准备工作，对数据集进行了精简调整，将原子电荷的单位改为e（原单位为-e），同时更新了README文件与数据集说明文档。