Ultralow Powered 2D MoS<sub>2</sub>-Based Memristive Crossbar Array for Synaptic Applications
Saurabh Yadav, Chandrabhan Patel, Manoj Kumar Rajbhar, Mayank Dubey, Dhananjay D. Kumbhar, Tukaram D. Dongale, Vishal Khandelwal, Saravanan Yuvaraja, Xiaohang Li, Shaibal Mukherjee
Abstract
Two-dimensional materials are increasingly integral to beyond-CMOS electronics, facilitating the development of emerging memristive device technology for information storage and neuromorphic computing. Despite their emergence, some critical challenges including low device yield, substantial device-to-device (D2D), and cycle-to-cycle (C2C) variability factors hinder the development of high-density memristive devices for future low-power electronic applications. Here, we demonstrate a memristive crossbar array (MCA) in which multilayer 2D MoS 2 acts as a resistive switching layer that offers lower switching voltages with a few microseconds pulse width. Additionally, the use of 2D MoS 2 further excels in integration density and energy efficiency, which significantly helps to achieve a device yield of 94%. Moreover, the 2D MoS 2 controlled growth process ensures the uniformity of MoS 2 layers across a (10 × 10) crossbar array that enhances the stability of fabricated MCA’s having minimal variability in device switching voltages ( V SET: 4.16% and V RESET: 3.60%). The fabricated devices show excellent endurance (∼24,000 cycles) and retention (1.6 × 10 6 s). Furthermore, due to lower switching voltages and fast switching speed, the fabricated devices consume 53 pW power and 53 aJ energy, making them more energy-efficient and achieving an impressive 97.79% accuracy in MNIST digit recognition through synaptic behavior simulation.