Standard DFT underestimates the band gap of InSb and gives poor effective masses. This paper compares G0W0, the hybrid functionals HSE and HSEsol, and DFT+U, with the hybrid mixing parameter and the Hubbard U calibrated by Bayesian optimization (BMach). The optimized DFT+U approach reproduces the experimental band gap, effective masses and Luttinger parameters at a much lower computational cost than the high-level methods.
High-fidelity electronic structure and properties of InSb: G0W0 and Bayesian-optimized hybrid functionals and DFT + U approaches
Ritwik Das, Anne-Sophie Grimault-Jacquin, Frédéric Aniel. Physical Review B 112, 075136 (2025).
PhysicsAI/MLInSbSemiconductorsInfrared materialsGWHybrid functionalsDFT+UBayesian optimizationMachine learning