The COF space: materials features, gas adsorption, and separation performances assessed by machine learning

Abstract

Covalent organic frameworks (COFs) are promising materials for gas adsorption; however, only a small number of COFs has been studied for a few types of gas separations to date. To unlock the full potential of the COF space, composed of 69 784 different types of materials, we studied the adsorption of five important gas molecules, CO2, CH4, H2, N2, and O2 in COFs at various pressures combining high-throughput molecular simulations and machine learning. Adsorbent performances of COFs were then explored for industrially critical separations, such as CO2/CH4, CO2/H2, CO2/N2, CH4/H2, CH4/N2, and O2/N2. The key structural and chemical properties of the most promising adsorbents were revealed. Our work offers the most extensive dataset produced for COFs in the literature composed of similar to 4.3 million data points for all synthesized and hypothetical COFs' structural, chemical, and energetic features; gas adsorption properties; and selectivities to facilitate the materials discovery.