摘要：Metal–organic framework(MOF)membranes hold great promise in energy-efficient chemical *** outstanding challenges of the microstructural design stem from(1)thinning of membranes to immensely reduce the mass-transfer resistance(for high permeances);(2)tuning of orientation to optimize the selective transport of gas molecules,and(3)reinforcement of intercrystalline structure to subside leakage through defective gaps(for high selectivity).Here,we propose the ZIF-L membrane that is completely confined into the voids of the alumina support through an interfacial assembly process,producing an appealing membrane-interlocked-support(MIS)composite architecture that meets the requirements of the microstructural design of MOF ***,the membranes show average H2 permeances of above 4000 GPU and H_(2)/CO_(2) separation factor(SF)of above 200,representing record-high separation performances of ZIF-L membranes and falling into the industrial target zone(H_(2) permeance>1000 GPU and H_(2)/CO_(2) SF>60).Furthermore,the ZIF-L membrane possessing the MIS composite architecture that is established with alumina particles as scaffolds shows mechanical stability,scraped repeatedly by a piece of silicon rubber causing no selectivity loss.

摘要：合金材料因其丰富可调的几何结构和电子性质,在催化和材料科学领域得到了广泛的应用.其中合金形成能作为一个重要的物理量,对合金材料的形成和催化活性有重要影响.近年来,随着人工智能和数据库的发展,利用机器学习的方法研究和设计新的材料成为新的研究焦点.基于此,通过人工智能的多任务压缩感知算法,结合AB2合金形成能数据库展开了合金形成能描述符和预测研究.首先建立了相应合金形成能的通用描述符,并展开了特征敏感性分析,揭示出合金材料组分的电子性质和几何性质的影响及其相互依赖关系.研究结果显示,该描述符的预测误差低于8.10 kJ·mol^(-1),具有清晰的物理可阐述性,并预测了大量未知合金材料的形成能.