Machine-learning-assisted exploration of anion-pillared metal organic frameworks for gas separation

An efficient paradigm in materials design is developed by combining abandoned experimental data, computational structure descriptors, and random forest algorithm. As a validating case study, we achieve the precise prediction of adsorption properties of C2H2, C2H4, and CO2 in anion-pillared MOFs. Several MOFs with top performance for the separation of CO2/C2H2 and C2H2/C2H4 are successfully explored and synthesized. The structure-properties relationship is also described quantitatively to offer more precise and intuitive guidance to the

Automated summary

An efficient paradigm for materials design is developed by combining abandoned experimental data, computational structure descriptors, and random forest algorithm. As a case study, the adsorption properties of C2H2, C2H4, and CO2 in anion-pillared MOFs are precisely predicted, and several MOFs with top performance for CO2/C2H2 and C2H2/C2H4 separation are successfully explored and synthesized. A quantitative structure-properties relationship is also provided to offer more accurate and intuitive guidance.