Exploration of Dual-Atom Catalysts for CO2 Reduction Reaction Under Varying Potentials Using First-Principles and Machine Learning Approaches

Abstract

Carbon dioxide (CO₂) is a major greenhouse gas, contributing to global warming and climate change by raising surface temperatures and causing extreme weather events. This necessitates effective CO₂ removal and conversion into chemical products, aligning with the demands of energy materials science. Recent advancements in catalytic methods, particularly electrocatalysis, have focused on converting CO₂ into valuable fuels and industrial products using various catalysts. Among these, carbon-supported double atom catalysts (DACs) have demonstrated promise due to their close catalytic site proximity, enhancing electron transfer efficiency and stability. This study employs density functional theory (DFT) to explore the geometric configurations, electronic structures, and adsorption properties of 380 carbon-supported DACs for CO₂ reduction reactions (CO₂RR). By analyzing reaction pathways involving HCOOH, CO, CH₄/CH₃OH, and hydrogen evolution reactions (HER) at 722 distinct active sites, we assess the performance and selectivity of these DACs across varying potentials. The integration of machine learning (ML) algorithms into the computational analyses allows for accurate predictions of intermediate adsorption energies and site selectivity, achieving a coefficient of determination (R²) of approximately 0.90 and a mean absolute error (MAE) of around 0.2. Ultimately, this comprehensive DFT-ML approach identifies several promising candidates for CO₂RR and elucidates key descriptors that impact their performance.

Graphical Abstract