Machine learning assisted photocurrent wavelength classification in a self-powered broadband photoelectrochemical photodetector based on Bi2Te0.6Se2.4/MoSe2 heterojunctions

Self-powered operation and broadband spectral adaptability are critical for the next generation of photodetectors, enabling sustainable performance and multispectral functionality. The broadband wavelength response capability can expand the application scenarios of a photodetector. However, achieving both broadband photoresponse and accurate wavelength recognition remains a significant challenge. Herein, we develop a self-powered photoelectrochemical photodetector based on Bi 2 Te 0.6 Se 2.4 /MoSe 2 heterojunctions, capable of identifying 256, 365, 546, and 650 nm and simulated sunlight with the assistance of machine learning. To achieve precise energy band alignment with MoSe 2 and enhance the heterojunction interface, the selenium content in Bi 2 Te 3-x Se x and the mass ratio of the composite components were systematically optimized. These structural optimizations improve charge carrier separation and transport via band structure engineering and interfacial modulation, resulting in superior photoelectrochemical performance. By integrating material design with machine-learning-assisted wavelength discrimination, this study demonstrates a multidisciplinary approach to broadband detection and highlights the potential of such devices in optical communication, environmental monitoring, and multispectral imaging applications. • Integration of Bi 2 Te 0.6 Se 2.4 and MoSe 2 for PEC photodetection, combining complementary light absorption and efficient charge separation. • Simultaneous realization of zero-bias operation and broadband photoresponse from ultraviolet to visible regions. • Enabling accurate classification of incident light wavelengths based on photocurrent features. • Demonstrating a synergistic strategy that integrates material design, device engineering, and data analytics for multifunctional PEC photodetectors.