Machine Learning Assisted Wavelength Recognition in Cu2O/Si Self-Powered Photodetector Arrays for Advanced Image Sensing Applications

The ability of a photodetector array (PDA) to detect multiple wavelengths significantly expands its range of potential applications. However, effectively detecting and distinguishing between different wavelength bands remains a challenge for these arrays. This study introduces an approach for wavelength recognition in PDAs by integrating machine learning techniques with solution-processed Cu2O/Si heterojunction photodetectors. We propose a simple solution-processing method to fabricate a PDA consisting of a 4 × 4 array of p-Cu2O/n-Si photodiodes. This method involves low-power UV irradiation of a molecular precursor film containing Cu (II) complexes to produce a p-type Cu2O thin film on a Si substrate. A UV-shielding glass plate is used as a patterning mask, and water is used to wash away the UV-shielded areas. Using machine learning techniques, we effectively classify various wavelengths of light, including UV, visible, and near-infrared, and accurately predict their corresponding photocurrents in the Cu2O/Si heterojunction. Notably, the PDA enables clear identification of images across different light wavelengths. This PDA paves the way for advanced applications in multispectral imaging and sensing technologies.