Detection of Open Metal Sites in Metal–Organic Frameworks Using Machine Learning

Curbing the rise in carbon footprints is one of the major endeavors of scientists nowadays. One way of achieving this feat is by trapping carbon dioxide. Researchers have spent most of the twenty-first century studying ways to capture CO2 from higher CO2 concentrations in the air. Nonetheless, the constant ascent in the carbon dioxide levels and lack of effective means which could keep up with this ever-increasing number is compelling scientists to foray into alternative methods. One such method involves employing various adsorbents, such as M OFs containing several open metal sites for CO2 adsorption, to absorb CO2 from lower CO2 concentrations in the air. The authors of this study use machine learning methods to predict open metal sites in computation-ready, experimental metal–organic frameworks (CoRE MOFs). Two models—a deep neural network and a k-nearest neighbors model—were used and compared to verify our hypothesis. This research paper will form a pillar in the field of carbon engineering and a pioneer in the study of carbon capture using MOFs by segmenting them based on a higher probability of adsorption.