P-i-n Heterojunction Photonic Synapses with Multispectral Weak Light Modulation for Full-Color Image Recognition

Photonic synaptic transistors, designed to emulate the multispectral light detection of retinal neurons, play a pivotal role in advancing artificial vision systems capable of full-color image recognition. However, their limited sensitivity to weak and multispectral light significantly hampers their application in full-color image capture and recognition. In this work, a p-i-n structured synaptic transistor is introduced that improves weak light sensitivity and enhances synaptic performance by addressing interface issues. By integrating P3HT/LiF/InZnGdO nanofibers as p-i-n structure, the spectral response is successfully broadened from visible to near-UV region, achieving sensitivity to light intensities as low as 0.05 mW cm −2 . The device exhibits remarkable PPF characteristics across the wavelength range of 395–633 nm, with the PPF index showing a maximum increase of 84% compared to the single InZnGdO nanofiber device. Additionally, multiple neural networks are constructed with data from the synaptic transistor array driven by wavelength-tunable synaptic plasticity to validate the device's feasibility in various application scenarios. The proposed device achieved a recognition accuracy of up to 95.26% for high-precision full-color images based on the public CIFAR-10 dataset. This p-i-n structure offers a novel pathway to enhance the synaptic performance of metal oxide-based synaptic transistors, advancing the development of next-generation artificial vision.