Reliable Memristor Crossbar Array Based on 2D Layered Nickel Phosphorus Trisulfide for Energy-Efficient Neuromorphic...

Area of Science:

• Materials Science
• Nanotechnology
• Computer Engineering

Background:

• Designing energy-efficient memristors for neuromorphic computing is crucial for advancing beyond the von Neumann architecture.
• Existing memristor technologies face challenges in reliability and performance for large-scale synaptic arrays.

Purpose of the Study:

• To develop and characterize novel memristors based on layered nickel phosphorus trisulfide (NiPS₃) for artificial synaptic applications.
• To elucidate the resistive switching mechanism and synaptic plasticity emulation capabilities of NiPS₃-based memristors.

Main Methods:

• Fabrication and electrical characterization of Ti/NiPS₃/Au memristive devices.
• Transmission electron microscopy (TEM) and energy dispersive X-ray spectroscopy (EDS) for microstructural and chemical analysis.
• Implementation in crossbar arrays for neuromorphic operations and artificial neural network simulations.

Main Results:

• NiPS₃ memristors demonstrated uniform bipolar nonvolatile switching with low operating voltage (<1 V), fast switching (<20 ns), and high On/Off ratio (>10²).
• Resistive switching is attributed to Ti conductive filament formation/dissolution, influenced by Ti ion penetration and P/S ion drift, creating vacancies.
• Devices successfully emulated long-term synaptic plasticity, enabling multipattern memorization and multiply-and-accumulate (MAC) operations with high pattern recognition accuracy (≈96.4%).

Conclusions:

• Layered NiPS₃ is a promising material for fabricating high-performance memristors for neuromorphic computing.
• The unique switching mechanism involving ion drift and vacancy formation contributes to the device's favorable characteristics.
• NiPS₃ memristors show significant potential for realizing efficient and accurate artificial neural networks and synaptic arrays.