Effect of the PMMA Buffer Layer on the Electrical Stability of CuI Memristor

Abstract

Electrochemical metallization-based memristors have attracted considerable attention as artificial synaptic devices owing to their ion-mediated conductance modulation. Although halide-based resistive switching (RS) layers are promising for neuromorphic applications, the use of active metals such as Ag or Cu as the top electrode (TE) can induce unstable switching behavior through unwanted interfacial reactions and metal migration into the halide layer. In this study, a poly(methyl methacrylate) (PMMA) buffer layer was inserted between the Ag TE and the CuI RS layer to improve the electrical and structural stability of CuI-based memristors. The PMMA buffer layer suppressed Ag diffusion into the CuI layer, thereby preserving the Ag TE for up to six months. This suppression of electrode degradation consequently enabled stable resistive switching and long-term electrical stability without significant degradation. X-ray diffraction and time-dependent scanning electron microscopy – energy dispersive X-ray spectroscopy analyses were performed to investigate the presence of Ag and its diffusion behavior over time. In addition, ABAQUS-based finite element analysis was conducted to simulate the time-dependent Ag concentration distribution, revealing that the maximum Ag concentration in the PMMA/CuI device was reduced by approximately 88% compared with that in the device without PMMA. The synaptic characteristics of the devices were further evaluated through long-term potentiation and depression measurements and implemented in artificial neural network simulation. As a result, the recognition accuracy increased from 66.0% for the CuI device to 79.4% for the PMMA/CuI device, corresponding to an approximately 20% improvement. Furthermore, spike-dependent conductance modulation was demonstrated by varying the spike amplitude, number, and duration, confirming the capability of the PMMA/CuI device to emulate synaptic weight modulation. These results demonstrate that PMMA acts as an effective diffusion-modulating buffer layer for improving the reliability of CuI-based artificial synaptic devices.