Machine learning optical band gaps of doped-ZnO films

Abstract

Tailoring the optical bandgap, Eg, of ZnO nanostructured thin films is of great interest to meet increasing demands for diverse practical applications, such as optoelectronics, spintronics, and photonics. Processing parameters, dopant selection, and various combinations of the two have significant effects on the crystal structure and microstructure of fabricated films, and thus optical performance. Empirical results and previous models through the first principle show that changes in lattice parameters and the grain size correlate with those in Eg. But correlations are complicated and the models rarely apply to complex situations such as co-doped films. In this work, the Gaussian process regression (GPR) model is developed to elucidate the statistical relationship among the lattice parameters, grain size, and energy bandgap for doped-ZnO films. A total of 65 ZnO films with Eg ranging from 2.100 eV to 3.760 eV are explored. The modeling approach demonstrates a high degree of accuracy and stability, contributing to efficient and low-cost estimations of Eg.