Discovery of High-Performing Metal–Organic Frameworks for Efficient SF₆/N₂ Separation: A Combined Computational Screening, Machine Learning, and Experimental Study

Abstract

Effective capture and recovery of sulfur hexafluoride (SF 6 ) from SF 6 /N 2 mixture is an urgent challenge. Considering the existence of a large number of metal− organic frameworks (MOFs), the computational screening of MOFs is strongly desired before experimental efforts. In this work, the top-performance MOF adsorbents were identified from the most recent computation-ready, experimental metal−organic frameworks (CoRE MOFs) based on various metrics. The degree of unsaturation (unsat) and the number of hydrogen per unit cell (H) revealed with the optimal machine learning (ML) model are important factors for effective SF 6 /N 2 separation. One of the screened MOF candidates, FIRNAX01(TKL-107), was synthesized and the separation performance exceeded all the reported MOFs. Our computational screening not only offers effective prediction but also paves the way for accelerating the development of novel MOFs.

Non-CO2 greenhouse gas separation using advanced porous materials

“…Experimental evaluation of TKL107 further verified its high performance with the highest TSN among all of the reported MOFs. 329 In addition to SF 6 , the high throughput computation of other fluorinated gases was also explored. 330,331…”

Efficient Removal of Greenhouse Gases: Machine Learning-Assisted Exploration of Metal–Organic Framework Space

“…138,141−143 to combine computational screening and ML to separate SF 6 / N 2 . 148 They obtained performance metric values (such as work capacity, selectivity, and reproducibility) from 2,877 MOFs using GCMC, and the best-performing MOFs were selected (Figure 11A). Seven descriptors of 11 geometric descriptors (e.g., φ, PLD, Vfree, gASA, vASA, ρ, and AV) and 76 chemical descriptors (unsaturation, oxygen-to-metal ratio, nitrogen-tooxygen ratio, etc.)…”

“…A similar behavior was observed for Ni­(NDC)­(TED) 0.5 and Ni­(adc)­(dabco) 0.5 . He et al attempted to combine computational screening and ML to separate SF 6 /N 2 . They obtained performance metric values (such as work capacity, selectivity, and reproducibility) from 2,877 MOFs using GCMC, and the best-performing MOFs were selected (Figure A).…”

“…(E) Separation performance (TSN) of 0FIRNAX01 (TKL-107; labeled with a red star symbol) with various benchmark MOFs. Reproduced with permission from ref . Copyright 2023 American Chemical Society.…”