Theoretical study of preferred dopants for n-type transparent conducting oxides

Traditionally, it is believed that the conduction band edges of $d^{0}$ or $d^{10}$ oxides are derived mostly from cation states, thus substitutional doping on anion sites is expected to cause less perturbation and produce shallow donor levels in these materials. Using first-principles calculations, we show that although this paradigm is applicable for more covalent oxides such as SnO$_{\mathrm{2}}$ where F$_{\mathrm{O}}$ is a better n-type dopant than Sb$_{\mathrm{Sn}}$, for more ionic oxides such as ZnO, the conduction band edge actually contains a considerable amount of O $s$ orbitals, thus F$_{\mathrm{O}}$ in ZnO causes larger perturbation and consequently produces deeper donor levels than cation site doping such as Al$_{\mathrm{Zn}}$. This observation can be explained by coupling of cation state with high lying oxygen orbitals. The origin of the preferred n-type dopability of oxides, the potential of oxygen vacancy as n-type dopant, and the selection of chemical potential for n-type doping will also be discussed.