Effectiveness of in situ NH3 annealing treatments for the removal of oxygen from GaN surfaces

An in situ NH3 annealing procedure for the cleaning of GaN(0 0 0 1) is studied in detail using density functional theory (DFT), microkinetic modeling and X-ray photoelectron spectroscopy (XPS). The microkinetic model was calibrated and tested against published H2 and NH3 temperature programmed desorption (TPD) experiments on GaN(0 0 0 1). We find that an NH3 treatment is efficient for the removal of carbon contaminants, but a complete removal of oxygen contaminants cannot be achieved. The remaining oxygen coverage after the treatment was estimated from XPS measurements to be 0.92 ML. In contrast, our microkinetic model based on DFT derived parameters predicts complete removal of OH species and a final oxygen coverage of 0.19 ML. We assign the difference between model and experiments to the formation of a surface oxide phase, which is not included in the model. DFT results also indicate strong adsorbate-adsorbate interactions for H, N, NH, NH2, O, and OH on the GaN(0 0 0 1) surface which were incorporated into the microkinetic model to a first approximation. XPS experiments and microkinetic modeling demonstrate that the final surface composition shows little dependence on process parameters such as temperature or the time the sample is kept at an elevated temperature. Furthermore, the microkinetic model suggests that complete removal of OH from the surface can also be achieved using a NH3/H2 mixture, or even pure H2 as a hydrogen source. The amount of H2 present in the feed changes the coverage of NHx species, but a certain amount of adsorbed oxygen is always left on the surface.