Hybrid WS₂-Based Memristor With Tunable Conductance Modulation for Neuromorphic and Nociceptive Learning

Area of Science:

• Materials Science
• Nanotechnology
• Electrical Engineering

Background:

• Memristive devices are crucial for next-generation computing and artificial intelligence.
• Developing high-performance, energy-efficient memristors with tunable properties is an ongoing challenge.

Purpose of the Study:

• To report a novel high-performance memristive device integrating WS2, TiO2, and BaTiO3.
• To investigate the device's potential for neuromorphic computing and adaptive sensory applications.

Main Methods:

• Fabrication of a hybrid memristor device with a WS2 switching layer, TiO2 interface, and BaTiO3 dielectric.
• Characterization of the device's electrical properties, including switching characteristics, retention, endurance, and energy efficiency.
• Evaluation of synaptic plasticity and nociceptor-like behaviors.
• Integration into an artificial neural network (ANN) for performance assessment on the MNIST dataset.

Main Results:

• The device exhibits low-voltage analog multilevel switching (±0.5 V) with a wide memory window (>10) and excellent stability (retention >10⁴ s, endurance >10⁵ cycles).
• Ultralow switching energy (≈54.7 pJ) and high uniformity (cycle-to-cycle variation <3.6%) were achieved.
• Demonstrated 5-bit resistance states, successful reproduction of various synaptic plasticity functions, and precise conductance modulation using the ISPVA algorithm.
• Mimicked nociceptor-like behaviors and achieved 97.4% recognition accuracy on the MNIST dataset when integrated into an ANN.

Conclusions:

• The WS2-based hybrid memristor offers a promising platform for energy-efficient neuromorphic and adaptive sensory applications.
• The device's tunable defect properties and hybrid structure enable high performance and versatile functionalities.