We perform first-principles density-functional theory calculations to investigate the atomic and electronic properties and the formation energies of different defects and dopant complexes (involving $V_{\text{N}}$, $V_{\text{In}}$, O, Si, Mg, and C) in wurtzite InN. We find that O substituted on a N site $\left({\text{O}}_{\text{N}}\right)$ and Si substituted on an In site $\left({\text{Si}}_{\text{In}}\right)$ are the most favorable sites for O and Si impurities, which act as single donors, with the ionized state (${\text{O}}_{\text{N}}^{+}$ and ${\text{Si}}_{\text{In}}^{+}$) being most stable in $p$-type InN. Substitutional C on a N site $\left({\text{C}}_{\text{N}}\right)$ and Mg on an In site $\left({\text{Mg}}_{\text{In}}\right)$ are the most favorable sites for the C and Mg impurities, which are predicted to be single acceptors under more $n$-type conditions. Two O, Si, and Mg atoms in the neutral and charged states prefer to be well separated, exhibiting a mutually repulsive interaction, as do two C atoms in the charged state. Two carbon atoms in the neutral state, however, prefer to be located on neighboring anion sites. We also investigate defect complexes involving vacancies and impurities, as well as defect configurations arising from “codoping” and/or “cluster-doping” concepts, namely, ${\text{Mg}}_m{\text{O}}_n$ $\left(m,n\le 4\right)$ and ${\text{Si}}_i{\text{C}}_j$ $\left(i,j\le 4\right)$ complexes. Rather than being isolated defects, the nitrogen vacancy $V_{\text{N}}$ prefers to be bound to ${\text{Mg}}_{\text{In}}$, forming a neutral ${\text{Mg}}_{\text{In}}{V}_{\text{N}}$ complex, while the indium vacancy $V_{\text{In}}$ prefers to bond to oxygen, but the formation energy for the ${\text{O}}_{\text{N}}{V}_{\text{In}}$ complex (in all charge states) is high. The formation energy of the $3{\text{O}}_{\text{N}}{V}_{\text{In}}$ complex is significantly lower, but is still too high to be an important defect in InN in equilibrium. Our results indicate that the ${\text{Mg}}_m{\text{O}}_n$ complexes could be important defects under both In-rich and N-rich conditions: for $p$-type material, donor defects containing more O than Mg have very low formation energies, while for $n$-type material, acceptor defects containing more Mg than O have low formation energies. These complexes could provide a more efficient $p$- (or $n$-) type doping than single Mg (or than single Si and O). Our results suggest complexes of ${\text{Si}}_i{\text{C}}_j$, on the other hand, offer no advantage over single dopants.

• Received 1 May 2008

DOI:https://doi.org/10.1103/PhysRevB.79.035207