Unlocking CO2 Reduction Potentials: The Synergy of Double-Atom Catalysts and Machine Learning

CO₂ is the primary greenhouse gas responsible for global warming and climate change. By reducing CO₂ emissions, we can help stabilize global temperatures and mitigate the adverse effects of climate change, such as extreme weather events, rising sea levels, and loss of biodiversity. The electrocatalytic reduction of CO₂ involves converting the carbon dioxide into useful chemicals and fuels using electrochemical processes. This technology is considered a promising strategy for carbon recycling and sustainable energy production. The process utilizes catalysts to facilitate the reaction at an electrode surface, where CO₂ is reduced to products like carbon monoxide (CO), methane (CH₄), ethylene (C₂H₄), and even alcohols, using green electrical energy. The efficiency of the CO₂ reduction process heavily relies on the catalysts used. Various materials, including metals, metal oxides, and organic compounds, have been explored as potential catalysts. There are several technical challenges to overcome, including the development of more active, selective, and stable catalysts, the need for high energy efficiency, and the integration of such technology into existing industrial processes.

In a new study published in ACS Catalysis by Professor Zhongfang Chen from University of Puerto Rico together with Linke Yu and Professor Fengyu Li from Inner Mongolia University and Dr. Jingsong Huang and Dr. Bobby Sumpter from Oak Ridge National Laboratory and Dr. William Mustain from University of South Carolina, the researchers conducted comprehensive studies to understand the efficacy of double-atom catalysts (DACs) featuring an inverse sandwich structure, particularly for the electrocatalytic reduction of CO₂ into valuable C₁ products. Their approach combined the precision of first-principles density functional theory (DFT) calculations with the predictive power of machine learning (ML) to investigate into the catalytic potential of these novel structures.

The team utilized spin-polarized DFT to investigate the structural, electronic, and catalytic properties of DACs anchored on defective graphene. They focused on both homonuclear (same metal atoms) and heteronuclear (different metal atoms) DACs, exploring a wide range of metal combinations to determine their stability and catalytic activity. They integrated ML models with DFT findings and identified key atomic and structural features that contribute to catalytic activity. This novel approach allowed them to predict the performance of a broader set of DACs beyond those directly studied through DFT. Moreover, they conducted stability assessments, including thermodynamic and electrochemical stability evaluations using parameters such as binding energy, formation energy, and dissolution potential. Furthermore, the team analyzed the reaction pathways and intermediates involved in CO₂ reduction over these DACs, focusing on the energetics of various steps to identify potential-determining steps and the ultimate products of the reaction.

The researchers found that among the homonuclear DACs, Rh₂⊥gra (Rhodium dimers perpendicular to graphene) demonstrated exceptional catalytic activity, outperforming its homonuclear counterparts and Rh-based DACs with non-inverse sandwich structures. Additionally, 14 heteronuclear DACs showed significant stability and catalytic prowess. Notably, RhIr⊥gra and RhPt⊥gra emerged as standout candidates, featuring remarkably low limiting potentials indicative of high catalytic efficiency. The inverse sandwich configuration, where metal dimers are perpendicularly bonded to the graphene plane, was found to be crucial. This structure facilitated moderate CO₂ adsorption strength, aligning with the optimal catalyst behavior as per Sabatier’s principle.

The authors’ detailed analysis of reaction mechanisms revealed that the DACs predominantly favored the reduction of CO₂ to CH₃OH and CH₄, with the competition between CO₂RR and HER being effectively managed through selective CO₂ adsorption. The study’s findings shed light on the electron transfer mechanisms, highlighting the role of graphene as an electron reservoir and the importance of the metal dimers in mediating electron transfer to the adsorbed CO₂. According to the authors, the integration of ML not only provided insights into the relationship between the DACs’ atomic properties and their catalytic activity but also enabled the prediction of the performance of an extensive array of DACs, identifying 154 potential candidates with promising catalytic activities.

In conclusion, the new study by Professor Zhongfang Chen and colleagues showcased a new class of DACs with high catalytic activity for CO₂ reduction, supported by a novel design strategy that leverages the unique inverse sandwich structure. The combination of DFT and ML provided a comprehensive understanding and predictive capability, paving the way for the development of efficient catalysts for CO₂ utilization and sustainable energy applications.