Machine-Learning-Assisted First-Principles Calculations of Strained InAs 1 − x Sb x Alloys for Curved Focal-Plane Arrays

Curved focal-plane arrays offer significant advantages compared with their flat counterparts. However, the curving of the material induces strains which alter its optoelectronic properties. A comprehensive framework is presented for the computational investigation of the band gaps of ${\mathrm{InAs}}_{1\ensuremath{-}x}{\mathrm{Sb}}_{x}$ alloys under various strain conditions that are relevant for manufacturing curved focal-plane arrays. The framework consists of both standard and hybrid functional density functional theory (DFT) calculations, finite-element analysis (FEA) calculations, and Gaussian process (GP) regression. The DFT calculations are used for investigating the change of the band gap of the material under various strain conditions. This dataset is then used for training a GP model which is utilized to assess the effects of strain across the device, based on the FEA simulations. The results show excellent predictive capabilities for the machine-learning model at a significantly reduced computational cost and are directly transferable to other device designs.