[14p-K507-1~17] CS.5 Code-sharing Session of 6.5 & 7.5

3:15 PM - 3:30 PM JST(6:15 AM - 6:30 AM UTC)

[14p-K507-9] Molecular dynamics simulation of diffusion process of adsorbed excess hydrogen on H-terminated Si(100) 2x1 surface by using self-made machine learning potential

〇Kohji Inagaki¹ (1.Osaka Univ.)

Keywords:

machine learning potential,reaction process,surface diffusion

It is difficult to analyse reaction processes beyond activation barriers with machine learning potentials that are constructed using normal stable structures as training data. We have developed a method to efficiently collect atomic structures that act as reaction barriers and create machine learning potentials that include these structures in the learning data. The diffusion process of excess hydrogen adsorbed on a silicon (100)2x1 hydrogen-terminated surface was also analysed by simulation using this method. These results are presented.