Accurate first-principle bandgap predictions in strain-engineered ternary III-V semiconductors

Abstract

Tuning the bandgap in ternary III-V semiconductors via modification of the composition or the strain in the material is a major approach for the design of optoelectronic materials. Experimental approaches screening a large range of possible target structures are hampered by the tremendous effort to optimize the material synthesis for every target structure. We present an approach based on density functional theory efficiently capable of providing the bandgap as a function of composition and strain. Using a specific density functional designed for accurate bandgap computation (TB09) together with a band unfolding procedure and special quasirandom structures, we develop a computational protocol efficiently able to predict bandgaps. The approach's accuracy is validated by comparison to selected experimental data. We thus map the phase space of composition and strain (we call this the 'bandgap phase diagram') for several important III-V compound semiconductors: GaAsP, GaAsN, GaPSb, GaAsSb, GaPBi, and GaAsBi. We show the application of these diagrams for identifying the most promising materials for device design. Furthermore, our computational protocol can easily be generalized to explore the vast chemical space of III-V materials with all other possible combinations of III-and V-elements.

Systematic strain-induced bandgap tuning in binary III–V semiconductors from density functional theory

“…The DFT computational setup for the quaternary calcula-tions follows the approach used in the binary and ternary datasets calculations in our previous studies. 57,58 The computations are performed using the projector-augmented wave (PAW) method 105,106 as implemented in the Vienna ab-initio simulation package (VASP). [107][108][109][110] The generalized gradient approximation (GGA) based exchange-correlation functional by Perdew, Burke, and Ernzerhof (PBE) 47 with a cut-off energy of 550 eV for the planewave basis set is chosen in all calculations.…”

“…The bandgap natures are determined using the Bloch-spectral-weight-based protocol as described in Ref. 58. The 'fold2Bloch' 117 code is used to determine the Bloch spectral weights.…”

“…where a eqm is the equilibrium lattice parameter of the specific structure that will be strained. 57,58 Following the convention from Refs. 57 and 58, in the following, we indicate tensile strain with positive sign and compressive strain with negative sign.…”

“…56 Our previous study on mapping bandgaps for binary and ternary III-V semiconductors using the TB09 functional over a large range of composition and strain values showed excellent agreement with experiment at moderate computational costs. 57,58 However, due to the large composition space of quaternary III-V semiconductor materials leading to an estimated amount of 10,000,000s 59 of calculations per combination elements, the computational costs of these methods are still too large to enable high-throughput studies.…”

“…The excellent performance of the approach is demonstrated. Subsequently, we construct the composition-strain-bandgap relationship, the "bandgap phase diagram", 58,88 for GaAsPSb using our ML model. Although many of the binary and ternary subsystems of GaAsPSb, namely GaAs, GaP, GaSb, GaAsP, GaAsSb, and GaPSb, have been successfully synthesized and found special applications in different research fields, 4,7,49,[89][90][91][92][93][94][95][96][97][98][99][100][101][102][103][104] this particular quaternary compound has not been studied yet.…”