Presentation Information
[8a-N403-7]Molecular dynamics study of a-(SiO2)x(Al2O3)1-x /GaN interface using machine learning force field
〇Koki Sato1, Mutsunori Uenuma2, Ryousuke Jinnouchi1, Ryoji Asahi1 (1.Nagoya Univ., 2.AIST)
Keywords:
machine learning,gallium nitride,first-principles calculation
Forming a high-quality gate insulator is essential for the practical implementation of GaN-based devices, yet oxidations and defects induced between the insulator and GaN substrate often lead to degradation of the device performance. In this study, we assessed the structure and defects of the a-AlSiO/GaOx/GaN interface through MD simulations employing a machine-learning interatomic potential and subsequent DFT analysis. Pure a-Al2O3 resulted in significant increase of the defect density due to recrystallization during annealing, whereas SiO2 incorporation promoted Ga–O passivation and effectively suppressed in-gap defect states. Furthermore, inserting and precisely controlling the GaOx layer was shown to effectively improve the Ga–O passivation.