Artificial neural network-based forecasting of strained InGaN/GaN quantum well-related optical absorption

In this paper, we propose a Multi-Layer Perceptron (MLP) approach constituting a specific Artificial Neuron Network (ANN) architecture transforming the field of physics by providing robust alternatives to labor- and time-intensive empirical research to predict the far-field optical absorption spectra of strained (In,Ga)N/GaN quantum well, an important challenge in the photovoltaic area. Our model incorporates the effects of indium surface segregation and built-in electric field, offering a comprehensive analysis of the low-lying electronic states. The accuracy and generalization of the proposed model were evaluated by comparing the predicted results with actual data from calculations using the mean squared error and the correlation coefficient. The results show that the ANN-MLP architecture achieves high predictive accuracy, particularly for QWs with large barriers and low Indium content. The best mean square error and correlation coefficient for the MLP network are respectively $2.3{10}^{-3}$ and $98.3\%$ which verify the high efficiency and accuracy of the proposed neural network model. These findings reveal that ANN-based predictive models streamline the study of optical properties in low-dimensional materials and have the potential to replace traditional methods, accelerating advancements in next-generation optoelectronic device design.