Theoretical and computational study of voltage-controlled terahertz synaptic devices for neuromorphic systems

Area of Science:

• Terahertz technology
• Plasmonics
• Neuromorphic engineering

Background:

• Neuromorphic computing aims to mimic the brain's structure and function.
• Terahertz (THz) technology offers unique properties for high-speed, low-latency applications.
• Artificial synapses are key components for building neuromorphic systems.

Purpose of the Study:

• To demonstrate electrically tunable terahertz artificial synapses.
• To explore their application in neuromorphic computing and single-pixel imaging.
• To investigate the modulation capabilities of split-ring plasmon resonator (SRPR) arrays coupled with Schottky diodes.

Main Methods:

• Simulated electrically tunable terahertz artificial synapses using SRPR arrays and Schottky diodes.
• Modulated synaptic weight by applying gate voltage bias to Schottky diodes.
• Implemented a ResNet-18 neural network using an array of these synapses.
• Utilized the synaptic array as a spatial light modulator in a single-pixel THz imaging system.

Main Results:

• Achieved tunable transmission modulation of SRPR-based synapses from 36% to 85%.
• Demonstrated precise control over synaptic weights.
• Attained 93.4% accuracy for image classification on the CIFAR-10 dataset using a ResNet-18 network.
• Showcased reduced system complexity in single-pixel terahertz imaging.

Conclusions:

• Electrically tunable terahertz artificial synapses offer precise synaptic weight control.
• These synapses are promising for high-performance neuromorphic computing.
• The developed system enables efficient single-pixel terahertz imaging.
• This work opens new avenues for brain-inspired computing and intelligent imaging systems.