A Bioinspired Low-Power Optoelectronic Synaptic Transistor for Artificial Visual Recognition and Multilevel Optical...

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Neuromorphic computing aims to overcome the von Neumann bottleneck using bio-inspired designs.
• Artificial visual systems require wide light response and low power consumption, which are current challenges.

Purpose of the Study:

• To develop an optoelectronic synaptic transistor (OST) emulating avian visual processing.
• To achieve wide light spectrum response and multilevel optical storage with low power consumption.

Main Methods:

• Fabrication of an organic OST using solution-processed organic semiconductors and a biodegradable PVA electret.
• Characterization of synaptic behaviors including postsynaptic current, plasticity, potentiation, and depression.
• Implementation of MNIST image recognition using an artificial neural network (ANN) and simulation of biological memory behaviors.

Main Results:

• The OST successfully emulated avian retina and synapse functions, responding to UV and RGB light.
• Demonstrated multilevel optical storage with 300 conductance states at low light intensity (1 μW/cm²).
• Achieved 90.8% recognition rate in MNIST image recognition and low power consumption (137 pJ/pulse).

Conclusions:

• The developed organic OST shows significant potential for intelligent, efficient machine vision and neuromorphic systems.
• The device offers a simple, low-temperature fabrication process with biodegradable materials.
• Highlights opportunities for future advancements in artificial intelligence hardware.