Methane Adsorption Database of MOFs for Machine Learning V2.0

Metal-organic frameworks (MOFs) possess high surface areas and customizable properties, making them among the most promising materials for gas adsorption. The structures of MOFs vary widely due to differences in metal nodes, organic linkers, and their combinations, with hundreds of thousands of distinct structures identified to date. Efficiently screening and designing MOFs with high storage capacities for gas adsorption is a key challenge in advancing MOFs and is critical for the development of carbon capture and energy storage technologies. This database includes 261,612 MOFs that can be directly utilized for training machine learning models. All these MOF structures are derived from public databases. The dataset features 14 geometric descriptors for each MOF, along with methane adsorption capacities obtained from grand canonical Monte Carlo simulations at 5.8 bar and 65 bar, which are typical pressures for methane storage. V2.0: The latest MOSAEC algorithm, developed by Woo et al., was employed to identify structures with impossible or unlikely metal oxidation states. These chemically invalid samples were subsequently removed from the dataset.金属有机框架（Metal-organic frameworks, MOFs）拥有极高的比表面积与可定制化的结构性质，是目前最具应用前景的气体吸附材料之一。由于金属节点、有机配体及其组合方式存在差异，MOFs的结构种类繁多，截至目前已鉴定出数十万种不同的结构。高效筛选并设计具备高气体吸附存储容量的MOFs，是推动该材料发展的核心挑战，同时对于碳捕获与储能技术的研发至关重要。本数据集共包含261612种可直接用于机器学习模型训练的MOFs，所有MOF结构均源自公开数据库。该数据集为每种MOF提供14项几何描述符，并附带了在5.8巴与65巴（甲烷存储的典型压力）下，通过巨正则系综蒙特卡洛（grand canonical Monte Carlo）模拟得到的甲烷吸附容量。V2.0版本：本数据集采用Woo等人开发的最新MOSAEC算法，识别出存在不合理或不可能的金属氧化态的结构，并将这些化学无效样本从数据集中移除。