In a collaborative effort led by researchers from Skolkovo Institute of Science and Technology in Moscow, a team of scientists has developed an artificial intelligence (AI) driven approach to identify new compounds for capturing carbon dioxide (CO2) emissions from industrial sources. The team, comprising Davide Garbelotto, Alexander Lobo, Urvi Awasthi, Oleg Medvedev, Srayanta Mukherjee, Anton Aristov, Konstantin Polunin, Alex De Mur, Leonid Zhukov, Azad Huseynov, and Murad Abdullayev, focused on ionic liquids (ILs) as a promising alternative to traditional amine-based solvents for CO2 capture.
The researchers employed a five-stage pipeline to discover ILs with optimal properties for CO2 capture. First, they generated IL candidates by pairing available cation and anion molecules. Next, they used a graph neural network (GNN)-based molecular property prediction model to estimate the temperature- and pressure-dependent CO2 solubility and viscosity of these candidates. The team then converted solubility data into working capacity and regeneration energy using Van’t Hoff modeling. Following this, they identified the best set of candidates through Pareto optimization, which balances multiple objectives. Finally, they filtered the candidates based on feasible synthesis routes.
The study, published in the journal Nature Communications, identified 36 feasible IL candidates that could potentially reduce operational expenditures (OPEX) by 5-10% and capital expenditures (CAPEX) by up to 10%. These savings are attributed to lower regeneration energy requirements and reduced corrosivity of the new ILs compared to traditional amine-based solvents. The AI-driven approach offers a novel strategy for carbon capture in refineries and other industrial facilities, contributing to more efficient and cost-effective emission reduction efforts.
The practical applications of this research are significant for the energy industry, particularly in sectors with high CO2 emissions such as refineries, power plants, and chemical manufacturing facilities. By adopting these newly discovered ILs, companies could enhance their carbon capture capabilities, reduce operational costs, and contribute to global efforts to mitigate climate change. The AI-driven discovery process also sets a precedent for future research, demonstrating the potential of machine learning techniques in accelerating the development of advanced materials for energy and environmental applications.
This article is based on research available at arXiv.