Unravelling conducting substrate dependent resistive switching in ZnO@ β -SiC memristors for neuro-inspired computing

Abstract The rapid progress of artificial intelligence and machine learning hardware heavily depends on conventional von Neumann computing systems, where the physical separation between memory and processing units limits high processing by consuming high power. Inspired by the human brain, resistive switching (RS) devices have emerged as promising candidates for in-memory computing due to their potential for low-power operation and long-term stability. Although oxide and perovskite-based memristors have been extensively investigated, achieving stable and reliable device performance remains challenging. In this study, we propose a hybrid oxide- ceramic composite, ZnO@ β -SiC, to enhance device stability, reliability, and overall performance. These devices exhibit robust bipolar RS along with synaptic functionalities. Furthermore, conducting substrate-dependent device characteristics have also been investigated. Devices fabricated on Pt coated Si substrates demonstrate ultra-low switching voltages ( V set ∼ 100 mV) with abrupt resistance change, while those on ITO require comparatively higher switching voltage (∼0.8 V). Moreover, a detailed underlying mechanism based on morphological and microstructural information has been proposed. Importantly, the ITO-based devices, owing to their gradual resistance modulation, achieved ∼97% image recognition accuracy on the MNIST dataset using artificial neural network simulations. These results highlight the potential of ZnO@ β -SiC composite devices for applications in both resistive random-access memory and neuromorphic computing.