Thermal transport in n-type B-S co-doped and p-type B-doped diamond: A machine-learning potential study

Achieving reliable n-type doping in diamond is a long-standing challenge for high-power electronics. In this work, we systematically investigate the electronic structure and thermal transport properties of B and B–X (X = N, P, O, S, F, Cl) co-doped diamond using first-principles calculations and a high-precision machine-learning potential (MLP) trained via DeepMD-kit. Our results identify B–S co-doping as a superior n-type strategy, exhibiting a reduced formation energy and n-type conductivity. Crucially, by overcoming the accuracy limitations of empirical potentials in describing acoustic phonon velocities, our MLP rigorously quantifies the thermal conductivity of the co-doped system. Although the heavy mass and strain field of sulfur dopants introduce additional phonon scattering, B-S co-doped diamond retains a substantial thermal conductivity. The results reveal the microscopic interplay between dopant-induced electronic activation and phonon scattering, validating B-S co-doping as a viable route for thermally robust n-type diamond electronics. This work provides a reference for the design of n-type diamond with good thermal conductivity based on machine-learning potentials.