Native point defects in ZnO

We have performed a comprehensive first-principles investigation of native point defects in ZnO based on density functional theory within the local density approximation (LDA) as well as the $\mathrm{LDA}+U$ approach for overcoming the band-gap problem. Oxygen deficiency, manifested in the form of oxygen vacancies and zinc interstitials, has long been invoked as the source of the commonly observed unintentional $n$-type conductivity in ZnO. However, contrary to the conventional wisdom, we find that native point defects are very unlikely to be the cause of unintentional $n$-type conductivity. Oxygen vacancies, which have most often been cited as the cause of unintentional doping, are deep rather than shallow donors and have high formation energies in $n$-type ZnO (and are therefore unlikely to form). Zinc interstitials are shallow donors, but they also have high formation energies in $n$-type ZnO and are fast diffusers with migration barriers as low as $0.57\phantom{\rule{0.3em}{0ex}}\mathrm{eV}$; they are therefore unlikely to be stable. Zinc antisites are also shallow donors but their high formation energies (even in Zn-rich conditions) render them unlikely to be stable under equilibrium conditions. We have, however, identified a different low-energy atomic configuration for zinc antisites that may play a role under nonequilibrium conditions such as irradiation. Zinc vacancies are deep acceptors and probably related to the frequently observed green luminescence; they act as compensating centers in $n$-type ZnO. Oxygen interstitials have high formation energies; they can occur as electrically neutral split interstitials in semi-insulating and $p$-type materials or as deep acceptors at octahedral interstitial sites in $n$-type ZnO. Oxygen antisites have very high formation energies and are unlikely to exist in measurable concentrations under equilibrium conditions. Based on our results for migration energy barriers, we calculate activation energies for self-diffusion and estimate defect-annealing temperatures. Our results provide a guide to more refined experimental studies of point defects in ZnO and their influence on the control of $p$-type doping.