DAC-SIM

Associated dataset and files for the paper entitled"<b>Accelerating CO2 Direct Air Capture Screening for Metal-Organic Frameworks with a Transferable Machine Learning Force Field</b>"<br><br><b>GoldDAC.zip</b><br>- The dataset used to train and evaluate the developed MLFF (MACE-DAC).<br>- General descriptions were provided in Readme.md file in the .zip file.<br><br><b>DACSIM_optimized_coremof_2019_asr.zip</b><br>- <i>7,906 frameworks</i> undergone geometry optimization using MLFF (MACE-DAC-1) were provided as .cif file.<br>- Excluding frameworks with more than 250 atoms, 8,131 frameworks from CoRE MOF 2019 ASR database were initially considered.<br>- A few frameworks (~2.8 %) were failed during geometry optimization stage.<br><br><b>chemically_correct_promising_mofs.zip</b><br>- <i>119 chemically correct frameworks</i> suitable for DAC application were provided as .cif file.<br>- A total of 161 frameworks were originally regarded as promising during screening stage; however, 42 of them contain chemically impractical moieties (e.g. missing hydrogen atom or overlapping of atoms due to symmetry issues).<br><br><b>DACSIM_screening_coremof_2019_asr.csv</b><br>- Screening results for 7,906 geometry optimized frameworks.<br>- Adsorption properties (e.g. averaged interaction energy, Henry's law coefficient, and heat of adsorption) calculated from the Widom insertion Monte Carlo (MC) simulation were collected.<br>- Two independent calculations were conducted and the mean value of those two calculations were set as the final value.<br><br><b>DACSIM-Qst_vs_ODAC23-Eint</b><br>- Minimum interaction energy (referred to adsorption energy in the orginal paper) obtained from ODAC 2023 database were provided with heat of adsorption (Qst) values calculated in this work.<br>- A total of <i>2,794 frameworks</i> existed across both dataset.<br><br><b>DACSIM_FF_screening.csv</b><br>- Qst calculated using a classical force field (a combination of universal force field and DDEC charges) were provided with Qst values calculated in this work.<br>- A total of <i>2,654 frameworks</i> existed across both dataset.<br><br><b>Feature_analysis.csv</b><br>- A file that describes what chemical features the 119 chemcially correct promising MOFs contain.<br>- If the framework contain certain feature, it designate as 1; otherwise, it designate as 0.<br>本数据集为论文《基于可迁移机器学习力场加速金属有机框架的CO₂直接空气捕获筛选》的配套数据与文件。<br><br><b>GoldDAC.zip</b><br>- 该数据集用于训练与评估所开发的机器学习力场（Machine Learning Force Field，MLFF）MACE-DAC。<br>- 详细说明详见该压缩包内的Readme.md文件。<br><br><b>DACSIM_optimized_coremof_2019_asr.zip</b><br>- 提供了7906个经机器学习力场MACE-DAC-1完成几何优化的金属有机框架（Metal-Organic Frameworks，MOF）结构，格式为.cif文件。<br>- 初始共从CoRE MOF 2019 ASR数据库中选取8131个框架，剔除了原子数超过250的框架。<br>- 约2.8%的框架在几何优化阶段未能完成。<br><br><b>chemically_correct_promising_mofs.zip</b><br>- 提供了119个适用于CO₂直接空气捕获（DAC）应用的化学合理金属有机框架结构，格式为.cif文件。<br>- 初始筛选阶段共认定161个具备应用潜力的框架，但其中42个存在化学上不可行的结构缺陷（如氢原子缺失、因对称性问题导致原子重叠等）。<br><br><b>DACSIM_screening_coremof_2019_asr.csv</b><br>- 包含7906个经几何优化的框架的筛选结果。<br>- 收集了通过Widom插入式蒙特卡洛（Monte Carlo，MC）模拟计算得到的吸附性能数据，例如平均相互作用能、亨利定律系数以及吸附热。<br>- 共开展两次独立计算，以两次结果的平均值作为最终数值。<br><br><b>DACSIM-Qst_vs_ODAC23-Eint</b><br>- 提供了从ODAC 2023数据库中获取的最小相互作用能（原文中称为吸附能），以及本研究中计算得到的吸附热（Qst）数值。<br>- 两个数据集共包含2794个金属有机框架。<br><br><b>DACSIM_FF_screening.csv</b><br>- 提供了使用经典力场（通用力场与DDEC电荷组合）计算得到的吸附热（Qst），以及本研究中计算得到的吸附热（Qst）数值。<br>- 两个数据集共包含2654个金属有机框架。<br><br><b>Feature_analysis.csv</b><br>- 该文件描述了119个化学合理的高潜力金属有机框架所包含的化学特征。<br>- 若框架包含某一特征，则标记为1；反之则标记为0。