Combination of Surface Plasmon Resonance and Interface Engineering Enables Drastically Enhanced Polarization‐Sensitive In‐Sensor Computing Neuromorphic Vision

ABSTRACT Polarization‐sensitive neuromorphic visual sensors based on 2D materials can sense and process vector information of images, which is increasingly important for accurate image recognition. However, this is normally precluded by their crystallography‐governed intrinsically weak anisotropy and unsatisfactory interface engineering with plasmonic materials. Here, we present a high‐performance polarization‐sensitive neuromorphic vision sensor (NVS) that leverages surface plasmon resonance (SPR) from Au─Ag plasmonic nanostructures and engineered interface barriers to enhance the polarization performance of a ReSe 2 in‐sensor computing structure. Periodically arrayed Au─Ag elliptical cylinders on ReSe 2 , aligned with the b‐axis, yield strong SPR‐enhanced polarization in the NVS. Direct Ag/ReSe 2 contact forms anti‐barrier layers that effectively modulate photocarriers for neuromorphic vision. This device configuration leads to a polarized photocurrent response ratio as high as ∼18 under the illumination of light with a wavelength of 825 nm and enables volatile memristor behaviors simultaneously. With this sensor, polarization imaging with a high degree of linear polarization (DoLP) was demonstrated. Meanwhile, a mixed input reservoir model was proposed for effective and efficient high‐precision image recognition. This work paves a new way for designing high‐performance intelligent vision systems based on multi‐dimensional polarization information.