Exploring the phase change and structure of carbon using a deep learning interatomic potential

Small-scale systems based on periodic boundary conditions often cannot accurately describe real-world situations, especially when conducting molecular dynamics simulations to study phase transitions, where it is very necessary to use large-scale systems. However, studying phase transitions in large-scale systems is an important and difficult task. Though ab initio molecular dynamics (AIMD), based on density functional theory (DFT), provides advantages in terms of accuracy, it is very difficult to study phase transitions in large-scale systems due to the considerable computational time required. In addition, although traditional empirical potentials are faster, their lower calculation accuracy makes it difficult to use them for phase transition studies. It is crucial to devise a method that has enabled a promising fusion of computational efficiency and precision to effectively investigate phase transitions in large-scale systems. In this work, the obtained machine learning potential function of carbon through deep neural networks not only demonstrates strong scalability but also effectively enables the study of the formation mechanisms of amorphous diamond and polycrystalline diamond using C 60 crystals and graphene as precursors under high-pressure high-temperature conditions (HPHT). Furthermore, the structure search software (AIRSS) was used to generate numerous initial structures which were optimized using the machine learning potential, a process which led to finding new structural clusters of carbon. Interestingly, the predictive capabilities of the machine learning potential for symmetric and asymmetric carbon clusters aligned well with the Gaussian approximation potential (GAP), yet the former demonstrated higher computational efficiency, making it more suitable for carbon material research. The results of this work signify significant progress in the field of carbon transition study, opening up new possibilities for exploring and understanding carbon materials with improved computational efficacy. A machine learning potential for carbon, developed using deep neural networks, allows efficient phase transition studies in large-scale systems and is transferable for searching carbon cluster structures.