All-Metal-Oxide Heterojunction Optoelectronic Synapses with Multilevel Memory for Artificial Visual Perception...

Area of Science:

• Materials Science
• Neuroscience
• Electrical Engineering

Background:

• Metal-oxide semiconductor-based optoelectronic synaptic transistors are crucial for energy-efficient and stable neuromorphic computing.
• Emulating the human visual system requires devices that can mimic synaptic plasticity and memory functions.

Purpose of the Study:

• To propose and investigate a novel WO3/InWZnO heterojunction optoelectronic synaptic transistor.
• To demonstrate its potential for emulating the human visual system and its application in neuromorphic vision systems.

Main Methods:

• Fabrication of a WO3/InWZnO heterojunction synaptic transistor.
• Characterization of optical responsivity, short-term memory (STM) to long-term memory (LTM) transitions, paired-pulse facilitation (PPF), and post-tetanic potentiation (PTP).
• Implementation of multilayer perceptron (MLP) and U-Net models for handwritten digit recognition and image segmentation.

Main Results:

• Achieved high optical responsivity (58.37 A W^-1) at 650 nm.
• Successfully emulated STM-to-LTM transitions and demonstrated long-term multilevel storage via photogating.
• Exhibited optimal PPF (176%) and PTP (890%) indices under 460 nm light.
• MLP model achieved 87.4% accuracy for distorted handwritten digits; U-Net achieved 74.5% accuracy for image segmentation.

Conclusions:

• The WO3/InWZnO heterojunction synaptic transistor shows significant potential for advanced neuromorphic vision systems.
• The device's ability to emulate synaptic functions and its performance in image recognition and segmentation highlight its feasibility.