Machine learning boosts molecular design of metal-organic framework for efficient CF4 capture

CF4 serves as a prominent plasma etching gas in the microelectronics industry. The efficient separation of CF4 from N2 not only holds the potential to notably mitigate greenhouse gas emissions but also offers economic benefits for the semiconductor industry. Herein, we performed high-throughput computational screening of computation-ready, experimental metal–organic frameworks and identified 7 metal–organic frameworks demonstrating superior separation performance. Among these, MURBEI emerged as one of the most promising candidates. Further enhancement in separation performance was achieved, with an increase of 9.1 %, 41.8 %, and 18.2 % calculated for its working capacity, selectivity, and a trade-off between selectivity and working capacity guided by the categorical boosting model (R2 = 0.859). Dispersion-corrected density functional theory calculations revealed that the introduced alkyl groups could enhance the van der Waals interactions between framework and CF4 molecule, thereby facilitating more efficient capture of CF4. These findings provide theoretical guidance for the design of novel metal–organic framework-based adsorbents for CF4 capture.