colabfit/Open_Catalyst_2025_OC25_Val
收藏资源简介：
Open Catalyst 2025 OC25 Val数据集是Open Catalyst 2025（OC25）数据集的验证分割，专门用于固液界面研究。该数据集包含催化剂/溶剂/离子/吸附物结构的单点DFT计算，覆盖88种元素、8种溶剂（水、甲醇、四氯化碳、二甲亚砜、苯、己烷、四氢呋喃、乙醚）、9种离子物种（铯离子、氢氧根离子、锂离子、硫酸根离子、钙离子、四甲基铵离子、碳酸氢根离子、氢离子、氟离子）以及来自OC20集合的吸附物和反应中间体。表面来源于39,821个Materials Project体结构，米勒指数≤3。结构高度非平衡态，采样自短ab initio分子动力学模拟（10-50步，1000K，NVT系综）或短DFT弛豫（5个离子步骤）。验证分割包含203,630个结构，代表分布外体-溶剂组合（约26万种独特配对中保留2.5%）。与训练集相比，验证计算使用更严格的DFT收敛标准（EDIFF=1e-6 eV），以提供更高质量的力标签。所有DFT计算均使用VASP 6.3.2软件，采用非自旋极化RPBE泛函，辅以D3色散校正（零阻尼），平面波截断能400 eV，k点倒易密度40，并在z方向施加偶极校正。
The Open Catalyst 2025 (OC25) Val Dataset is the validation split of the Open Catalyst 2025 (OC25) dataset, specifically designed for solid-liquid interface research. This dataset contains single-point DFT calculations for catalyst/solvent/ion/adsorbate structures, covering 88 elements, 8 solvents (water, methanol, carbon tetrachloride, dimethyl sulfoxide (DMSO), benzene, hexane, tetrahydrofuran (THF), diethyl ether), 9 ionic species (cesium ion, hydroxide ion, lithium ion, sulfate ion, calcium ion, tetramethylammonium ion, bicarbonate ion, hydrogen ion, fluoride ion), as well as adsorbates and reaction intermediates from the OC20 collection. The surfaces are derived from 39,821 bulk structures in the Materials Project, with Miller indices ≤ 3. The structures are highly non-equilibrium states, sampled from short ab initio molecular dynamics simulations (10–50 steps, 1000 K, NVT ensemble) or short DFT relaxations (5 ionic steps). The validation split contains 203,630 structures, representing out-of-distribution (OOD) bulk-solvent pairs, with 2.5% retained from approximately 260,000 unique pairs. Compared to the training set, the validation calculations use more stringent DFT convergence criteria (EDIFF = 1e-6 eV) to provide higher-quality force labels. All DFT calculations were performed using VASP 6.3.2 software, employing the non-spin-polarized RPBE functional with D3 dispersion correction (zero damping), a plane-wave cutoff energy of 400 eV, a reciprocal k-point density of 40, and dipole correction applied along the z-direction.
