ABSTRACT The mechanical properties, namely the elastic constants (C11, C12, and C44), bulk B, shear Cs, and Young’s modulus Y0, of the In1-xGaxAsyP1-y lattice-matched GaAs and InP substrates, were estimated using an artificial neural network (ANN method). This study aimed to create ANN networks that facilitate the estimation of the mechanical properties of alloys subjected to different temperatures and hydrostatic pressures. We aimed to predict the mechanical properties of InGaAsP alloys matched to InP and GaAs substrates as a function of temperature and hydrostatic pressure. ANN learning was performed based on the results validated by the empirical pseudopotential method data within the virtual crystal approximation, including the actual disorder potential. An ANN is a nonlinear process modeling technique frequently encountered in materials science. The predictive curves show high reliability, and underline the importance of using this approach to replace the time-consuming and costly experimental test, as well as to predict the mechanical response of the material where conventional modeling approaches have failed. Our method can consolidate the theoretical references found in the literature, it can also serve the field of engineering the mechanical properties of semiconductors, facilitate the successful design of optoelectronic devices, and serve future experimental works.
Read full abstract