The role of machine learning in carbon neutrality: Catalyst property prediction, design, and synthesis for carbon dioxide reduction

Abstract

Achieving carbon neutrality is an essential part of responding to climate change caused by the deforestation and over-exploitation of natural resources that have accompanied the development of human society. The carbon dioxide reduction reaction (CO₂RR) is a promising strategy to capture and convert carbon dioxide (CO₂) into value-added chemical products. However, the traditional trial-and-error method makes it expensive and time-consuming to understand the deeper mechanism behind the reaction, discover novel catalysts with superior performance and lower cost, and determine optimal support structures and electrolytes for the CO₂RR. Emerging machine learning (ML) techniques provide an opportunity to integrate material science and artificial intelligence, which would enable chemists to extract the implicit knowledge behind data, be guided by the insights thereby gained, and be freed from performing repetitive experiments. In this perspective article, we focus on recent advancements in ML-participated CO₂RR applications. After a brief introduction to ML techniques and the CO₂RR, we first focus on ML-accelerated property prediction for potential CO₂RR catalysts. Then we explore ML-aided prediction of catalytic activity and selectivity. This is followed by a discussion about ML-guided catalyst and electrode design. Next, the potential application of ML-assisted experimental synthesis for the CO₂RR is discussed. Finally, we present specific challenges and opportunities, with the aim of better understanding research and advancements in using ML for the CO₂RR.