Abstract
Electrochemical CO2 reduction (CO2RR) offers a promising pathway to mitigate carbon emissions while producing sustainable fuels and value-added chemicals. Yet, the complexity of competing reaction pathways, the sensitivity of product selectivity to catalyst structure and environment, and the limitations of conventional density functional theory (DFT) screening hinder rational catalyst discovery. In recent years, machine learning (ML) has emerged as a transformative approach, bridging data-driven prediction with mechanistic understanding. This review highlights how ML accelerates CO2RR research by enabling the rapid screening of vast chemical spaces, uncovering physicochemical descriptors, and integrating predictive models into closed-loop experimental workflows. Advances include the use of graph neural networks, adsorption-energy distributions, Bayesian optimization (BO), and interpretable models that expose key features controlling activity and selectivity. Case studies illustrate how ML-guided pipelines have identified novel Cu-based alloys, single-atom catalysts, and dual-site architectures with experimentally validated performance. Despite progress, challenges remain in domain adaptation, multi-objective optimization, and the incorporation of dynamic electrochemical conditions into models. Collectively, the integration of ML with theory and experiment is reshaping CO2RR research from trial-and-error exploration toward mechanism-driven, autonomous catalyst discovery, paving the way for scalable carbon utilization technologies.
| Original language | English |
|---|---|
| Article number | 217850 |
| Journal | Coordination Chemistry Reviews |
| Volume | 559 |
| DOIs | |
| State | Published - 15 Jul 2026 |
Bibliographical note
Publisher Copyright:© 2026 Elsevier B.V.
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 7 Affordable and Clean Energy
-
SDG 13 Climate Action
Keywords
- Bayesian optimization (BO)
- Catalyst discovery
- DFT
- Descriptor engineering
- Electrochemical CO reduction (CORR)
- Machine learning
ASJC Scopus subject areas
- General Chemistry
- Physical and Theoretical Chemistry
- Inorganic Chemistry
- Materials Chemistry
Fingerprint
Dive into the research topics of 'Machine learning-driven approaches in electrochemical CO2 reduction: Bridging the gap between data and sustainable catalysis'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver