The increasing atmospheric CO2 levels from fossil fuel use have led to global warming. The IPCC report warns that global warming will exceed 1.5°C in the 21st century if greenhouse gas emissions, mainly CO2, are not drastically reduced. To combat this, Europe aims to achieve net-zero emissions by 2050 through the European Green Deal. A critical step is converting CO2 and renewable electricity into fuels and chemicals. One major challenge of this process is the lack of an electrocatalyst that can reduce CO2 to high-energy, high-value long-chain hydrocarbons, like propanol. This project aims to use Machine Learning and simulations to help developing the effective CO2 electroreduction catalysts.
In this reaction, oxide-derived copper catalysts have attracted widespread attention due to their excellent ability to promote carbon-carbon coupling. The outstanding performance of oxide derived Cu is typically attributed to their unique surface structures. However, the intense dynamic behaviour of such catalysts under reaction conditions leads to surface restructuring. Due to the catalyst's active nature, it is highly susceptible to oxidation during characterization processes using ex-situ experimental techniques, making it difficult to reflect its true structure during the reaction. Consequently, there remains significant controversy regarding the active sites, particularly the existence forms and distributions of oxygen atoms in the material under reaction conditions. Furthermore, traditional simulation methods such as first-principles calculations and classical molecular dynamics simulations struggle to balance accuracy and speed, thus failing to capture this complex dynamic process. Limited understanding of this fundamental process hinders further optimization of catalysts and reaction conditions. To address this challenge, the objective of the project is to provide atomic-level insights with the help of advanced machine learning techniques, thus providing design principles for better catalyst development.
