Engineering ZrO2/WS2 nanocomposite for multilevel memory and high-performance neuromorphic computing

Abstract

Neuromorphic computing, developed from the human brain, has the potential of ongoing evolution of computing. With its massive parallelism and potential for lower power consumption, memristors hold the key to significant advancements in computational competences. However, emerging bilayer memristor devices are still facing challenges likes stress issues in fabricating memory devices to enhance their resistive switching characteristics, such as (RON/OFF, energy consumption, and endurance). As a result, we report a novel approach for hybrid nanocomposite in the present study with the synergistic combination of zirconium oxide (ZrO2) and tungsten disulfide (WS2) as the switching layer in memory devices to resolve this challenge. The stoichiometry dependent Ag/Z70W30/Pt nanocomposite devices showed extremely stable resistive switching, having ultralow power consumption of 72.8 aJ, a high RON/OFF of ∼108, and excellent reliability with an endurance that is greater than ∼106 cycles. Furthermore, memristors effectively emulate diverse bio-synaptic functions, which include paired-pulse facilitation (PPF), paired-pulse depression (PPD) and spike-rate-dependent plasticity (SRDP). Neuromorphic simulations using the MNIST handwritten digit dataset demonstrated the memristor device's high performance, achieving 91.61 % recognition accuracy. Hybrid nanocomposite memristors integrating TMDs and oxides offer a promising paradigm for artificial synapses in neuromorphic computing systems.