Machine learning and molecular fingerprint screening of high-performance 2D/3D MOF membranes for Kr/Xe separation

Separation of Xe and Kr is extremely important in several applications, such as spent nuclear fuel reprocessing. In this work, high-throughput computational screening (HTCS) was used to simulate the dynamic behavior of Kr/Xe separation for 6013 computation-ready, experimental metal–organic framework membranes (CoRE-MOFMs). First, the structure–performance relationships of the metal–organic framework membranes (MOFMs) for Kr/Xe separation were analyzed by univariate analysis. Then, five machine learning (ML) algorithms (random forest (RF), decision tree (DT), support vector machine (SVM), k-nearest neighbors (KNN) and extreme gradient boosting (XGB)) were employed for classification and regression of permeability (P) and permselectivity (S). Besides, the excellent bits of linkers were determined by molecular fingerprints (MFs), and the excellent nodes and separation mechanisms were also discussed. Finally, three design strategies were proposed to boost the Kr/Xe separation performance of MOF membranes. Combining HTCS, ML and MF, we provide a new direction for designing high-performance MOF membranes for Kr/Xe separation.