Machine Learning Interatomic Potential-Enabled Discovery of Chlorofullerenes

Chlorination is an effective method for stabilizing fullerenes with pentagon adjacencies, transforming reactive sp² carbon sites into sp³ hybridization to release local strain and reorganize orbital and spin distributions. Yet, the immense variety of fullerene isomers and their diverse chlorination sites pose significant challenges for accurate structural prediction. This study introduces a density functional theory (DFT)-accurate machine learning potential (MLP) for the C-Cl system, facilitating high-throughput screening of chlorinated fullerenes. The MLP precisely replicates a range of experimentally identified chlorofullerenes and establishes correlations between parent fullerenes and their chlorinated derivatives in terms of structural features and stability, offering theoretical insights into the formation principles of chlorofullerenes.