Accelerated discovery of CO₂ electrocatalysts using active machine learning

Area of Science:

• Electrochemistry
• Materials Science
• Catalysis

Background:

• Growing global energy demand necessitates renewable energy solutions.
• Electrochemical reduction of carbon dioxide (CO2) offers a pathway for storing intermittent solar and wind energy.
• Copper-based catalysts are key for producing valuable multi-carbon products from CO2, but current efficiency and productivity are limiting.

Purpose of the Study:

• To develop novel electrocatalysts for efficient CO2 reduction to ethylene.
• To overcome the limitations of existing copper electrocatalysts in terms of energy efficiency and productivity.
• To leverage computational and machine learning approaches for catalyst discovery.

Main Methods:

• Density functional theory calculations combined with active machine learning to identify promising electrocatalyst compositions.
• Electrochemical reduction experiments to evaluate catalyst performance.
• In situ X-ray absorption spectroscopy to investigate catalyst structure and mechanism.

Main Results:

• Cu-Al electrocatalysts demonstrated the highest reported Faradaic efficiency for CO2 to ethylene conversion (>80%).
• Achieved high current density (400 mA/cm²) at 1.5 V vs. RHE with a 55% ethylene power conversion efficiency at 150 mA/cm².
• Computational studies indicated that Cu-Al alloys provide optimal CO binding sites and surface orientations for efficient and selective CO2 reduction.

Conclusions:

• Cu-Al electrocatalysts represent a significant advancement over pure copper for CO2 electroreduction.
• The synergistic effects in Cu-Al alloys, including favorable Cu coordination, enhance C-C bond formation for ethylene production.
• This work highlights the power of integrating computation and machine learning in designing advanced multi-metallic electrocatalysts.