EL2 and the electronic structure of the ${\mathrm{As}}_{\mathrm{Ga}}$-${\mathrm{As}}_{\mathrm{i}}$ pair in GaAs: The role of lattice distortion in the properties of the normal state

The proposal that in its normal state EL2 is an ${\mathrm{As}}_{\mathrm{Ga}}$ ${\mathrm{As}}_{\mathrm{i}}$ pair with a [111] axis and two-bond-length separation is tested by performing electronic structure calculations for that defect pair. The ${\mathrm{As}}_{\mathrm{i}}$ is allowed to minimize its energy by moving along the [111] axis. Its equilibrium position depends on its charge. We find that in the charge state ${\mathrm{As}}_{\mathrm{i}}^{+}$ the ${\mathrm{As}}_{\mathrm{i}}$ seeks a position very close to the hexagonal interstitial site (H), 1.5 bond lengths from the ${\mathrm{As}}_{\mathrm{Ga}}$. In this configuration, unlike that at two-bond-length separation, the electronic structure of the pair accounts for many observed but seemingly unrelated properties of EL2. The calculation suggests that in this model the transition to the metastable state would be triggered by an off-axis Jahn-Teller distortion of the ${\mathrm{As}}_{\mathrm{i}}$. This calculation also predicts that in isolation the ${\mathrm{As}}_{\mathrm{i}}$ has no EPR-active equilibrium states but does have a deep two-electron level, (1+,3+), caused by very large lattice relaxation: In the state ${\mathrm{As}}_{\mathrm{i}}^{3+}$ the interstitial is at a tetrahedral site ($T_d$), while in the state ${\mathrm{As}}_{\mathrm{i}}^{+}$ it is closer to the H point. The problems with identifying EL2 as the pair, however, are twofold. First, the EL2 defect is predicted to have a shallow effective-mass-like level in addition to the two deep levels ascribed to its ${\mathrm{As}}_{\mathrm{Ga}}$ component. Secondly, we have not found a plausible mechanism for binding the two components of the pair to each other.