Atomistic simulation of boron implantation and defect-mediated diffusion in silicon via DFT-MD and deep potential molecular dynamics.

Ion implantation followed by thermal annealing is a key process in semiconductor doping, where the resulting dopant distribution critically determines device performance. However, quantitatively predicting post-annealing impurity profiles remains challenging due to the complex coupling between implantation-induced defect generation and defect-mediated diffusion. Here, we develop a first-principles-based atomistic framework that combines density functional theory molecular dynamics (DFT-MD) for ion implantation with deep potential molecular dynamics (DeePMD) for large-scale annealing simulations. In this approach, all-electron DFT is used to construct accurate interatomic interactions governing implantation, while the deep potential (DP) model enables efficient and accurate simulations of defect-mediated diffusion at extended time and length scales. The DFT-MD + DP framework significantly improves the prediction of boron dopant profiles in silicon, reducing the normalized root mean square error by 50-69% compared with conventional MC/BCA + KMC methods, with excellent agreement with experimental secondary ion mass spectrometry (SIMS) measurements. The improved accuracy originates from a physically consistent description of both implantation-induced defect structures and interstitial-mediated diffusion mechanisms. In particular, the framework captures the role of excess self-interstitials and the kick-out mechanism governing boron diffusion, establishing a direct link between atomistic processes and macroscopic dopant profiles. These results demonstrate that a first-principles-based atomistic approach enables quantitatively accurate predictions of dopant transport in silicon and provides new insights into defect-mediated diffusion processes.