Coupled ferroelectric-anisotropic optoelectronic synapse for polarization-sensitive neuromorphic vision

Polarization-sensitive photodetection and non-volatile memory are both vital for neuromorphic vision hardware but are rarely integrated within a single device. This challenge arises from interfacial instabilities and depolarization fields at the 2D/ferroelectric junctions that degrade remanent polarization and long-term retention. Here, we demonstrate a polarization-resolved optoelectronic synapse based on a 2D ReS2 channel and a ferroelectric Hf0.5Zr0.5O2 (HZO) gate dielectric in a metal-ferroelectric-metal-insulator-semiconductor (MFMIS) ferroelectric field-effect transistor (FeFET). Co-modulation of ferroelectric polarization and photoexcited carrier trapping enables high responsivity, strong detectivity, and long-term optoelectronic retention. Coupling between the polarization anisotropy of ReS2 and ferroelectric memristive states enables gate-tunable polarization ratios and polarization-resolved learning. Furthermore, the optoelectronic synapse exhibits linear and energy-efficient optical–electrical modulation with 2.0 fJ per event. An ANN built from these synapses achieves 97.33% accuracy in iris recognition under unpolarized light, while a 3×3 FeFET-based CNN performs butterfly classification under polarized illumination through polarization-resolved feature extraction. This work establishes a unified ferroelectric-anisotropic platform for energy-efficient, polarization-resolved neuromorphic vision. Neuromorphic vision hardware calls for all-in-one integration of photodetection and computing. Huo et al. report an optoelectronic synapse based on a ReS2 channel integrated to a metal–ferroelectric–metal–insulator–semiconductor architecture, showing ultra-low dark current and high photoresponsivity.