Large-Scale Computational Screening of Metal–Organic Frameworks for D2/H2 Separation

Fei Wang,Zhiyuan Bi,Lifeng Ding,Qingyuan Yang

DOI: https://doi.org/10.1016/j.cjche.2022.04.003

IF: 3.8

2022-04-01

Chinese Journal of Chemical Engineering

Abstract:Deuterium (D2) is one of the important fuel sources that power nuclear fusion reactors. The existing D2/H2 separation technologies that obtain high-purity D2 are cost-intensive. Recent research has shown that metal–organic frameworks (MOFs) are of good potential for D2/H2 separation application. In this work, a high-throughput computational screening of 12020 computation-ready experimental MOFs is carried out to determine the best MOFs for hydrogen isotope separation application. Meanwhile, the detailed structure-performance correlation is systematically investigated with the aid of machine learning. The results indicate that the ideal D2/H2 adsorption selectivity calculated based on Henry coefficient is strongly correlated with the 1/ΔAD feature descriptor; that is, inverse of the adsorbility difference of the two adsorbates. Meanwhile, the machine learning (ML) results show that the prediction accuracy of all the four ML methods is significantly improved after the addition of this feature descriptor. In addition, the ML results based on extreme gradient boosting model also revealed that the 1/ΔAD descriptor has the highest relative importance compared to other commonly-used descriptors. To further explore the effect of hydrogen isotope separation in binary mixture, 1548 MOFs with ideal adsorption selectivity greater than 1.5 are simulated at equimolar conditions. The structure-performance relationship shows that high adsorption selectivity MOFs generally have smaller pore size (0.3–0.5 nm) and lower surface area. Among the top 200 performers, the materials mainly have the sql, pcu, cds, hxl, and ins topologies. Finally, three MOFs with high D2/H2 selectivity and good D2 uptake are identified as the best candidates, of all which had one-dimensional channel pore. The findings obtained in this work may be helpful for the identification of potentially promising candidates for hydrogen isotope separation.

engineering, chemical