Fingerprinting the vibrational signatures of dopants and defects in a fully random alloy: An ab initio case study of Si, Se, and vacancies in In0.5Ga0.5As

Abstract

Correct identification of local configurations of dopants and point defects in random alloys poses a challenge to both computational modeling and experimental characterization methods. In this paper, we propose and implement a computationally efficient approach to address this problem. Combining special quasirandom structures, virtual crystal approximation, and real-space lattice static Green’s functions, we are able to calculate, at moderate computational cost, the local phonon density of states (LPDOSs) of impurities in a random alloy crystal for system sizes, surpassing the capabilities of a conventional, cubic-scaling, density functional theory. We validate this method by showing that our LPDOS predictions of substitutional silicon in GaAs and InAs are in excellent agreement with the experimental data. For the case study, we investigate a variety of local configurations of Si and Se substitutional dopants and cation vacancies in quasirandom In0.5Ga0.5As alloys. In all cases, the impurity LPDOS in a random alloy exhibits qualitatively different signatures from those in the pure binary compounds GaAs and InAs. Specifically, they are characterized by a wide continuous band (rather than narrow discrete peaks) of vibrational modes at frequencies typically higher than the bulk modes, a sign of coupling between localized vibrations of the impurity and those of its random neighboring host atoms. The accuracy and computational cost of this approach open a way to the simulation of impurities in random structures on a large scale and the prediction of vibrational signatures of alloys with defects.