Parameterizing DFT+ <i>U</i> corrections for III–V semiconductor alloys

Abstract

(InxGa1−x)(As1−ySby) alloys are key materials for short-wave infrared (SWIR) sensors and related optoelectronic devices, where accurate band structure parameters are required for predictive TCAD simulations. Building on a previously proposed Bayesian Optimization approach for calibrating Hubbard parameters for crystals [Yu et al., njp Comput. Mater. 6, 180 (2020)], we develop a DFT+U framework in which Hubbard-U parameters on the III–V p states are calibrated to experimental alloy bandgaps and hybrid-functional reference band structures using Bayesian optimization. These parameters are then transferred to ternary alloys via a composition-dependent interpolation scheme applied to special quasi-random supercells. Using this calibrated approach, we obtain bandgap bowing parameters and effective masses for the four ternary subsystems of the (InxGa1−x)(As1−ySby) alloy in good agreement with experimental data and empirical bowing models. The same calculations provide the composition dependence of the EL and EX conduction band valleys, yielding a consistent ab initio set of band parameters across the relevant alloy space. These results provide device-ready input for TCAD simulations of SWIR detectors based on (InxGa1−x)(As1−ySby) and more generally illustrate how Bayesian calibration of DFT+U can be used to parameterize band structures in complex III–V alloys.