Machine-learning enhanced thermal stability investigation of single Shockley stacking faults in 4H-SiC

Haonan Chen,Wenyu Kang,Wei Lin,Junyong Kang
DOI: https://doi.org/10.1016/j.commatsci.2024.113077
IF: 3.572
2024-05-15
Computational Materials Science
Abstract:A comprehensive understanding of single Shockley stacking faults (1SSFs) would be crucial for defect control in silicon carbide (SiC). Nevertheless, the thermal stability of 4H-SiC 1SSFs remains unclear due to limitations in conventional technologies. In this work, a machine learning force field (MLFF) for the Si-C system was developed, and the outcomes were compared with other simulations and experiments in three aspects. The MLFF was then utilized to investigate the thermal stability of ten established 1SSF models (3072 atoms for each) at seven temperatures. The results revealed that antiphase boundaries (APBs) positively related to thermal stability, and APB-lacking 1SSFs susceptible to annihilation. Among the stably existing 1SSFs, a bidirectional transition process of 30° partial dislocations (PDs) between symmetric reconstruction (SR) and asymmetric reconstruction (AR) was visualized. The distorted bilayers caused by 90° PDs were also observed. Notably, two novel structures were identified with robust thermal stability that could naturally exist in 4H-SiC. Additionally, the specific vibration modes of 1SSFs in phonon spectra were characterized, which offers potential applications in non-destructive defect detection in SiC materials. This study provides advanced insights into the evolution and thermal stability of 1SSFs at the atomic level.
materials science, multidisciplinary
What problem does this paper attempt to address?