Artificial Synapse with High Weight-Updating Performance Based on Charge-Trapping Mechanism

Area of Science:

• Neuromorphic Engineering
• Materials Science
• Nanotechnology

Background:

• Artificial synapses are crucial for neuromorphic hardware, aiming to mimic biological synaptic dynamics for efficient neural networks.
• Current two-dimensional (2D) material heterojunctions face limitations in dynamic weight updating due to high nonlinearity and low symmetricity.

Purpose of the Study:

• To engineer novel 2D heterojunction artificial synapses with improved weight-updating characteristics.
• To investigate the role of charge-trapping layers in modulating synaptic performance.
• To simulate adaptive behaviors, such as the human eye's response, using optoelectronic modulation.

Main Methods:

• Fabrication of 2D ReS₂/CTL/h-BN heterojunctions by treating h-BN with oxygen plasma to create a charge-trapping layer (CTL).
• Characterization of synaptic performance, including memory window and weight-updating characteristics (long-term potentiation/depression - LTP/D).
• Analysis of the mechanism underlying the influence of trap states on device performance and optimization of the device structure.

Main Results:

• The fabricated device demonstrated a large memory window and excellent synaptic performance.
• The device successfully simulated adaptive behavior, mimicking the human eye, via optoelectronic double-pulse modulation.
• Optimized LTP/D weight-updating showed reduced nonlinearity (0.63) and improved symmetricity (41.25), surpassing previously reported devices.

Conclusions:

• The developed 2D ReS₂/CTL/h-BN heterojunction synapses offer superior weight-updating performance compared to existing technologies.
• The study provides critical insights into optimizing synaptic devices by understanding and controlling trap states in charge-trapping layers.
• This research paves the way for more efficient and high-quality artificial neural networks in neuromorphic computing.