Ultra robust negative differential resistance memristor for hardware neuron circuit implementation.

Area of Science:

• Materials Science
• Computer Engineering
• Neuroscience

Background:

• Neuromorphic computing aims for efficient computation using brain-inspired architectures.
• Memristor-based artificial neurons offer energy efficiency and scalability but face reliability challenges.
• Existing memristors struggle with stability, hindering practical neuromorphic applications.

Purpose of the Study:

• To develop an ultra-robust and efficient artificial neuron using a negative differential resistance (NDR) memristor.
• To demonstrate the memristor's stability, high temperature resistance, and endurance for neuromorphic applications.
• To integrate the NDR memristor into advanced neural network circuits for enhanced functionality and reliability.

Main Methods:

• Fabrication of an NDR memristor utilizing an AlAs/In0.8Ga0.2As/AlAs quantum well (QW) structure.
• Characterization of device performance, including variation, temperature resistance, and endurance (>10^11 cycles).
• Implementation of the NDR memristor in a reduced Fitz Hugh Nagumo (FN) neuron circuit and a multimodal impulse neural network.

Main Results:

• The proposed NDR memristor exhibits exceptional stability, low variation (0.264%), and high temperature resistance (400°C).
• Devices demonstrated over 10^11 switching cycles at room temperature and 10^9 cycles at 400°C, ensuring long-term operation (>310 years at 10 Hz).
• The integrated FN neuron circuit showed diverse dynamics and functions, while the neural network achieved 91.74% accuracy in temperature-labeled image classification.

Conclusions:

• The AlAs/In0.8Ga0.2As/AlAs QW NDR memristor provides a highly reliable and efficient solution for neuromorphic computing.
• This work presents a competitive method for constructing robust artificial neurons and neural network hardware.
• The developed technology enables diverse neuromorphic functionalities and paves the way for advanced, high-temperature-resistant computing systems.