Optical bandgap modeling of thermal annealed ZnO:Ga thin films using neural networks

• Published in: Physica status solidi. A, Applications and materials science Vol. 207; no. 7; pp. 1572 - 1576
• Main Authors: Kim, Chang Eun, Moon, Pyung, Yun, Ilgu, Kim, Sungyeon, Myoung, Jae-Min, Jang, Hyeon Woo, Bang, Jungsik
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Berlin WILEY-VCH Verlag 01.07.2010; WILEY‐VCH Verlag; Wiley-VCH
• Subjects: annealing; Condensed matter: electronic structure, electrical, magnetic, and optical properties; doping; Exact sciences and technology; Ii-vi semiconductors; modeling; optical properties; Optical properties and condensed-matter spectroscopy and other interactions of matter with particles and radiation; Optical properties of specific thin films; Physics; ZnO; Thin films; Thickness; Energy gap; Neural networks; Zinc oxide; Modelling; Absorption spectra; Non linear effect; Thermal annealing; Carrier density; Burstein Moss effect
• ISSN: 1862-6300; 1862-6319
• DOI: 10.1002/pssa.200983715

In this paper, the thermal annealing process modeling for the optical bandgap of ZnO:Ga thin films for transparent conductive oxide was presented using neural network (NNets) based on error backpropagation (BPNN) algorithm and multilayer perceptron (MLP). The thermal annealing process of ZnO:Ga thin films were analyzed by general factorial experimental design. The annealing temperature and film thickness were considered as input factors. To model the nonlinear annealing process, 6 experiments were trained by BPNN which has 2‐4‐1 structures and 2 additional samples were experimented to verify the predicted models. The output response model on optical bandgap and carrier concentration of ZnO:Ga thin films trained by BPNN was represented by surface plot of response surface model. Based on the modeling results, NNets can provide sufficient correspondence between the predicted output values and the measured. The optical bandgap variation of ZnO:Ga thin films by annealing is due to increased carrier concentration and explained by Burstein–Moss effect. The thermal annealing process is nonlinear and complex but the output response can be predicted by the NNets model.