Artificial neural network-based model for estimation of EQE of multi-junction solar cells

External quantum efficiency (EQE) of multi-junction (MJ) solar cells is an important parameter used to optimize the design parameters of the solar cells and to predict the end-of-life (EOL) in space solar cells. The EQE undergoes drastic variations when the solar cell is bombarded with charged particles, due to their complex interactions with cell materials. Usually, elaborate and extensive experimental setup is needed to measure EQE of a MJ solar cell when it is under the influence of charged particles of different energy levels and fluences. In this paper we propose an artificial neural network (ANN)-based model to estimate EQE of triple-junction InGaP/GaAs/Ge solar cells, for proton energies from 30 keV to 10 MeV with fluences ranging from 10¹⁰ to 10¹⁴ ion/cm². Using only a small subset of the measured data (taken from Sato et al. [2]) as training set, we have shown that the ANN-based model can estimate the EQE for the complete range with much lower error than the PC1D model reported by Sato et al. [2]. With extensive simulation results we have shown superior performance of the ANN-based models over PC1D in terms of absolute error, mean square error and correlation coefficient between the measured and estimated EQE, under the influence of a wide range of proton energy energies and fluences.