Ultra-Low-Power and Reconfigurable Optoelectronic Memtransistor Based on Vertical Nb-WSe₂/Te van der Waals...

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• The Von Neumann architecture faces energy efficiency limitations.
• Neuromorphic computing hardware offers a solution.
• Synaptic memtransistors are key for energy-efficient brain-like systems.

Purpose of the Study:

• To fabricate an ultra-low-power, reconfigurable optoelectronic memtransistor.
• To emulate synaptic plasticity and logic functions.
• To demonstrate its potential in neuromorphic computing applications.

Main Methods:

• Fabrication of a vertical Nb-doped WSe2/Te van der Waals heterostructure.
• Optical stimulation to emulate synaptic behaviors (facilitation, plasticity).
• Implementation in a convolutional neural network (CNN) for image recognition.

Main Results:

• The device exhibits ultra-low energy consumption (<1 aJ), significantly lower than biological synapses.
• Programmable AND/OR logic emulation via photoelectronic control.
• Achieved 92.32% accuracy on CIFAR-10 benchmark with CNN implementation, robust to noise.

Conclusions:

• Nb-WSe2/Te vdW memtransistors are highly energy-efficient optoelectronic synapses.
• The device enables multimodal synaptic conditioning and robust CNN performance.
• These memtransistors show promise for ultra-low-power, reconfigurable neuromorphic computing hardware.