Structural Evolution of C_(60) Molecular Crystal Predicted by Neural Network Potential

摘要

The preparation of fullerene and a wide range of derivatives has attractedmuch attention in past decades. Understanding the structural evolutionstarting from fullerene is critical to guide the experimental exploration buthas been paid less attention yet. Using ab initio molecular dynamics or globalstructural search algorithm accompanied with density functional theory calculationscan give a glance of the potential energy surface (PES), but suffershigh cost and low efficiency. Herein, by using neural network potentials andstochastic surface walking global structural search method, an acceleratedenergy calculation, and the higher efficient PES mapping starting from C_(60)molecular crystal are reported. The structural evolution is found to follow anorder of molecular crystal → polymers → opened caged ordered structures →3D) curved carbon and graphite under zero pressure. Under high pressures,e.g., 50 GPa, the evolution follows a pathway of fullerene → sp~2 and sp~3hybrid carbon → sp~3 amorphous carbon and crystal diamond. The electronicproperty calculation of the obtained structures in the evolution pathwayshows a band gap depending on the order parameter of the generatedstructures, suggesting that more novel carbons can be potentially prepared inexperiments, starting from C_(60).