Atomic Charge Schemes Comparison for Fe Single Atom in Graphitic carbon: Insights from Quantum Simulations and Machine Learning

Zhiyu Wang, Jirui Jin, Mingie Liu

University of Florida

Single-atom catalyst is a new class of catalyst with high activity and selectivity which is widely utilized in electroreduction reaction such as oxygen reduction reaction, CO₂ reduction reaction, etc. The charge state of single transition metal atom is highly related to the catalytic performance as the adsorption strength is associated with the number of d-electrons for molecular system. Therefore, the dependency on the charge state of single metal center with geometric and electronic properties is important for rational catalyst design but still lacking. We systematically studied different charge schemes such as Bader, DDEC6, Mulliken, Hirshfeld and Charge Model 5 (CM5) which are commonly used charge analysis methods in catalysis field. Various sizes and edges of graphene nanoflakes supported Fe single atom with various positions, coordination environments are selected as the model system. We found that among the designed variables, the charge state shows a strong dependency with the coordinated atoms. DDEC6 charge shows a strong correlation with Hirsfeld and Mulliken charges. Subsequently, the 20 designed features including 12 electronic features and 8 geometric features are selected for machine learning models. The average bond length between metal center and coordination atoms are chosen as top importance features for 4 charge schemes. CM5 charge is the only one which depends on the electronic properties. This study offers an in-depth understanding of charge distribution of single atom catalysts and bridges the structural-performance relationship by the charge state descriptor in future studies.