Perovskite-Based Memristor with 50-Fold Switchable Photosensitivity for In-Sensor Computing Neural Network.

Area of Science:

• Materials Science
• Optoelectronics
• Computer Vision

Background:

• In-sensor computing offers efficient machine vision by processing visual data directly on the sensor.
• Controlling sensor photosensitivity is crucial but challenging for in-sensor computing applications.

Purpose of the Study:

• To develop a perovskite material with tunable photosensitivity for advanced in-sensor computing.
• To demonstrate a proof-of-concept visually enhanced neural network system utilizing this novel sensor.

Main Methods:

• Fabrication of a ternary cationic halide perovskite Cs$_{0.05}$FA$_{0.81}$MA$_{0.14}$Pb(I$_{0.85}$Br$_{0.15}$)$_3$ (CsFAMA).
• Characterization of the perovskite's optoelectronic properties, including external quantum efficiency (EQE) and responsivity.
• Demonstration of switchable photoresponsivity via ion migration and readout voltage adjustment.
• Integration of the perovskite sensor array into a visually enhanced neural network system.

Main Results:

• The CsFAMA perovskite exhibits high EQE (>80%) across the visible spectrum (400-750 nm) with a peak responsivity of 0.45 A/W at 750 nm.
• A 50-fold enhancement in photoresponsivity was achieved by tuning ion migration and readout voltage.
• The perovskite sensor array demonstrated simultaneous optimization of imaging and recognition, improving object recognition accuracy by 17% in low-light environments.

Conclusions:

• The developed CsFAMA perovskite enables controllable photosensitivity, a key advancement for in-sensor computing.
• The switchable photosensitivity facilitates enhanced performance in machine vision tasks, particularly under challenging lighting conditions.
• This work paves the way for more efficient and accurate visual sensing and processing systems.