Catalyzed Single-Walled Carbon Nanotube Growth by Machine Learning Molecular Dynamics Simulation

Classical molecular dynamics (MD) simulations have been performed to elucidate the mechanism of single-walled carbon nanotube (SWCNT) growth. However, discussing the chirality has been challenging due to topological defects in simulated SWCNTs. Recently, the application of neural network to interatomic potentials has been actively studied and such interatomic potentials are called neural network potentials (NNPs). NNPs have a better ability for approximate functions and can predict complex systems' energy more accurately than conventional interatomic potentials. In this study, we developed an NNP to describe SWCNT growth more accurately. We successfully demonstrated defect-free and chirality-definable growth of SWCNTs on Fe nanoparticles by MD simulation. Furthermore, it was revealed that edge vacancies of SWCNTs caused defect formation and adatom diffusion contributed to vacancy healing, preventing defect formation.