Atomistic simulation study of diamond doping based on machine learning potential

Abstract

Diamond is regarded as a highly promising semiconductor material due to its outstanding physical properties. However, achieving efficient p-type and n-type doping remains a significant challenge. In this work, we systematically investigate the B–O and B–S co-doping behaviors in diamond using molecular dynamics and Monte Carlo hybrid simulations based on machine learning potentials. The results indicate that the dopant ratio determines the formation type of impurity complexes, thereby governing the overall electronic properties of the material. Our simulations reveal that a B : O ratio of 4 : 1 leads to the exclusive formation of B₄O complexes, which exhibit the lowest formation energy (− 1.92 eV) and a low ionization energy (0.11 eV), making them a promising co-doping strategy for high-quality p-type diamond. For B–S co-doping, a B : S ratio of 1 : 1 results in BS complexes with significantly reduced formation energy compared to S single-doping and a low ionization energy (0.56 eV), meeting the requirements for n-type diamond, while a B : S ratio of 4 : 1 is more suitable for p-type diamond. These findings emphasize the critical role of precise dopant ratio control in obtaining high-performance p/n-type diamond semiconductors and provide atomic-scale mechanistic insights for the controllable synthesis of co-doping diamond.