Multilevel Resistive Switching and Synaptic Functionality in an Energy Efficient Ferroelectricity Assisted Memristor Guided by Ionic Dielectric Barrier

Abstract

A solution‐based method is used to fabricate a multilevel memristive device, in which the ferroelectric polarization of the LiNbO 3 layer is controlled by the ionic‐polarization of the Li 5 AlO 4 barrier layer. The high resistance state (HRS) to low resistance state (LRS) ratio in this LiNbO 3 /Li 5 AlO 4 /LiNbO 3 stacked configuration is approximately 10 4 . The device exhibits exceptional reliability, retaining steady performance over 10 3 switching cycles, sustaining ∼ 10 6 pulse operations, and preserving data retention beyond 10 5  s with an ON/OFF ratio close to 10 3 . Excellent device uniformity (±3%) and ReRAM‐like behavior demonstrated across 50 devices, confirming reproducibility. The device has the potential to create compact and multifunctional neuromorphic systems due to its reliable resistive switching behavior and support for basic logic operations. Oxygen‐vacancy‐assisted filament formation produces polarization‐modulated conduction pathways that govern the resistive states and allow for tunable multilevel switching under different compliance currents. Apart from its memory capabilities, the device successfully simulates synaptic processes like potentiation and depression, which reflect biological learning mechanisms. These neuromorphic features are further applied in the training of a convolutional neural network, achieving a training accuracy of about 98%. The study shows that polarization‐assisted transport and vacancy‐mediated switching work well together, providing a promising path toward high‐performance and scalable neuromorphic computing architectures.