Abstract:

Metal-organic frameworks (MOFs) have attracted much attention in the field of gas adsorption and separation due to their high porosity and ultra-high specific surface area, and the database of MOFs has been enriched as a result. The use of high-throughput computational screening methods can provide rich structural properties and performance data, which is beneficial to screening materials with high performance from a large number of metal-organic framework materials. In order to fully explore the information within the data, machine learning is used as an auxiliary tool that can reveal the implicit metal-organic framework structure and property relationships. To gain a greater understanding of the performance trends of metal-organic framework materials in different applications, especially in gas storage and separation, machine learning methods are also widely used. The latest research progress in machine learning prediction and design of metal-organic framework materials applied to the adsorption and separation of combustible gases is reviewed in terms of the descriptors of metal-organic frameworks suitable for machine learning work, and the screening and prediction of material properties by using machine learning methods, which accelerates the pace of the design and development of metal-organic frameworks, and guides the direction and rules of material synthesis, reducing the cost of manpower and material resources.

Key words: MOFs, adsorption, separation, computer simulation, machine learning

摘要：

金属有机框架（MOFs）由于其高孔隙率和超高的比表面积在气体吸附和分离领域受到广泛关注，金属有机框架数据库也因此丰富。使用高通量计算筛选方法可以提供丰富的结构性质和性能数据，有利于从大量的金属有机框架材料中筛选具有高性能的材料。为了充分挖掘数据内的信息，将机器学习用作辅助工具，可以揭示隐含的金属有机框架结构和性能关系；能够对金属有机框架材料在不同应用中的性能趋势有更多的理解。特别是在气体储存和分离方面，机器学习方法也被广泛应用。从适用于机器学习工作的金属有机框架的描述符，利用机器学习方法筛选及预测材料性质等方面综述了机器学习预测和设计应用于可燃气体吸附分离的金属有机框架材料的最新研究进展，加快金属有机框架的设计和开发步伐，指引材料的合成方向和规律，降低了人力物力成本。

关键词: MOFs, 吸附, 分离, 计算机模拟, 机器学习