Ultralow Powered 2D MoS₂-Based Memristive Crossbar Array for Synaptic Applications.

Area of Science:

• Materials Science
• Nanotechnology
• Solid-State Electronics

Background:

• Two-dimensional (2D) materials are crucial for beyond-CMOS electronics, particularly for memristive devices used in information storage and neuromorphic computing.
• Challenges like low device yield and variability in device-to-device (D2D) and cycle-to-cycle (C2C) performance hinder the development of high-density memristive devices.
• 2D materials offer potential for low-power electronic applications due to their unique properties.

Purpose of the Study:

• To demonstrate a memristive crossbar array (MCA) utilizing multilayer 2D molybdenum disulfide (MoS₂) as the resistive switching layer.
• To address challenges of device yield, D2D, and C2C variability in memristive devices.
• To showcase the potential of 2D MoS₂ for high-density, energy-efficient, and stable memristive applications.

Main Methods:

• Fabrication of a (10 × 10) memristive crossbar array (MCA) using multilayer 2D MoS₂ as the resistive switching layer.
• Controlled growth process to ensure uniformity of MoS₂ layers across the array.
• Characterization of device performance, including switching voltages, endurance, retention, power consumption, and energy efficiency.

Main Results:

• Achieved a high device yield of 94% with minimal variability in switching voltages (V_SET: 4.16%, V_RESET: 3.60%).
• Demonstrated excellent endurance (∼24,000 cycles) and retention (1.6 × 10⁶ s).
• Devices exhibited low switching voltages and fast switching speeds, consuming only 53 pW power and 53 aJ energy, and achieved 97.79% accuracy in MNIST recognition simulation.

Conclusions:

• The 2D MoS₂-based MCA offers a promising solution for high-density, low-variability, and energy-efficient memristive devices.
• The controlled growth process is key to achieving uniform MoS₂ layers and stable device performance.
• These findings pave the way for practical applications in information storage and neuromorphic computing.