Machine-learning assisted filterless color imaging with donor–acceptor ratio engineered self-driven organic photodetectors

Miniaturized optical wavelength-sensing devices based on solution-processed organic materials hold great promise for integration into portable and wearable technologies. Yet, the realization of self-powered compact wavelength sensors remains elusive. Here, we report a self-powered wavelength sensor built from broadband photodetectors featuring a meticulously engineered PM6: L8-BO active layer. By systematically varying the donor-acceptor stoichiometries and implementing these blends in nano-scale active layers (50 nm and 100 nm) that modulate the internal optical field distribution, we tailor the spectral responsivity of individual sensor units, yielding distinct wavelength-dependent optoelectronic signatures. An array of these wavelength-discriminating units enables quantitative discrimination and identification of incident light wavelengths. The device accurately resolves wavelengths from 380 nm to 850 nm with a resolution of better than ～1 nm, determined through the photocurrent ratio mapping of the four photodetector elements. As a proof of concept, we demonstrate the device’s capability in wavelength recognition and full-color imaging, underscoring its potential for compact, self-powered, and versatile optical sensing platforms.