How Realistic Are Idealized Copper Surfaces? A Machine Learning Study of Rough Copper-Water Interfaces.

Area of Science:

• Computational materials science
• Surface chemistry
• Electrocatalysis

Background:

• Copper is a key catalyst for electrochemical CO2 reduction (CO2RR), producing valuable products like ethylene and ethanol.
• Understanding the atomic structure of the water-copper interface is critical for elucidating CO2RR mechanisms in aqueous solutions.

Purpose of the Study:

• To investigate the atomic structure of nanometer-scale rough copper-water interfaces.
• To identify unique interfacial environments and their potential role in catalytic processes.

Main Methods:

• Development of molecular dynamics protocols to generate rough copper surfaces.
• Machine-learning interatomic potential-driven molecular dynamics simulations with large atomic systems.
• Active learning workflow for identifying uncertain regions and converting them to DFT-feasible cells.
• Unsupervised machine learning for analyzing local interfacial environments.

Main Results:

• Unique interfacial environments were observed on rough copper surfaces, distinct from model systems.
• Identified stacking-fault-induced configurations and undercoordinated corner atoms.
• Corner atoms consistently showed chemisorbed water molecules, suggesting their catalytic significance.

Conclusions:

• Rough copper surfaces exhibit unique atomic configurations important for CO2RR.
• Undercoordinated corner atoms with chemisorbed water are potential active sites for catalysis.
• This study provides atomic-level insights into copper-water interfaces for improved catalyst design.