Unveiling the invisible: Polarization-sensitive ferroelectric photomemristors for enhanced image recognition

Photoresponsive memristors (i.e., photomemristors) have been recently highly regarded to tackle data latency and energy consumption challenges in conventional Von Neumann architecture-based image recognition systems. However, their efficacy in recognizing low-contrast images is quite limited, and while preprocessing algorithms are usually employed to enhance these images, which naturally introduce delays that hinder real-time recognition in complex conditions. To address this challenge, here we present a self-driven polarization-sensitive ferroelectric photomemristor inspired by advanced biological systems. The proposed prototype device is engineered to extract image polarization information, enabling real-time and in-situ enhanced image recognition and classification capabilities. By combining the anisotropic optical feature of the two-dimensional material (ReSe₂) and ferroelectric polarization of single-crystalline Diisopropylammonium Bromide (DIPAB) thin film, tunable and self-driven polarized responsiveness with intelligence was achieved. With remarkable optoelectronic synaptic characteristics of the fabricated device, a significant enhancement was demonstrated in recognition probability—averaging an impressive 85.9% for low-contrast scenarios, in contrast to the mere 47.5% exhibited by traditional photomemristors. This holds substantial implications for the detection and recognition of subtle information in diverse scenes such as autonomous driving, medical imaging, and astronomical observation.