Reflection Prediction of Black Silicon Texture Under the Guidance of Image Recognition Technology

Black silicon texturing has been considered to be a promising wafer texturing technology and can effectively improve the short-circuit current and, thus, power conversion efficiency, which has been widely used in the multicrystalline silicon photovoltaic industry. In this article, we propose a new method to “recognize” the black silicon morphology from the scanning electron microscope (SEM) images and to predict the resultant reflectance. ImageJ software was applied to recognize and analyze the SEM images for the first time, to obtain the acid-etched pit parameters. Then, we applied the recognized parameters into the sigmoid-Boltzmann equation to obtain an accurate distribution of the pit size. Finally, a modified model has been established to describe the exponential relations between the reflectance and the ratio of pit depth d over wavelength λ ( d / λ ), pit size, and pit coverage proportion. For the typical pit size 500–800 nm in mass production, its depth is about half of the wavelength λ . Thus, we can predict the reflectance of the black silicon surface precisely, by monitoring their SEM images. The average reflectance deviation between the measured and the predicted reflectance is ∼3.7% relatively.