Localized Redox Filament Formation in Biopolymer-Based RRAM Devices for In-Memory Volatile Molecule Sensing

Abstract

Memristors and resistive random access memory (RRAM) utilize a range of active ions, material compositions, and structural designs, each influencing switching performance in different ways. Gaining insight into the core mechanisms that drive these dynamic behaviors and how they are affected by material and structural choices is essential for engineering devices with tailored electrical characteristics. The performance of memristor devices largely depends on the controlled formation and rupture of conductive filaments (CFs) within the switching layer (SL), which serves as the core functional component. This study focuses on the characterization and investigation of the filament formation process, exploring how switching layer thickness, electrode geometry, and lateral gap width influence filament dynamics in biopolymer-based RRAM devices. Different spatial CFs were compared both qualitatively and quantatively. Using polyvinylpyrrolidone (PVP) as the switching layer, devices with varying thicknesses and lateral gaps were fabricated via photolithography and spin-coating techniques. Surface and cross-sectional analyses revealed that thicker PVP SL (250 nm) facilitated multiple filament nucleation due to increased defect density, while thinner SL (30 nm) promoted localized filament growth but required higher forming voltages. Electrode geometry was found to play a crucial role, with cone-shaped electrodes demonstrating enhanced filament localization and lower forming voltages (1.77±0.25 V) with the narrowest dispersion compared to rectangular (2.00±0.64 V) and arch-shaped (1.91±0.36 V) designs, owing to improved electric field distribution. Increasing the lateral gap width (W) from 3 µm to 15 µm resulted in higher forming voltages and an increased number of nucleation sites, while filament localization was preserved due to the optimized layer thickness and geometry. The electrochemical oxidation of the positive bias electrode and the reduction of Ag⁺ ions at the negative electrode were confirmed through FESEM-EDX analysis. The planar structure was further studied for in memory volatile organic molecules sensing during the HRS. The Characterization and visualization of the filament formation process under varying conditions deepen our understanding of electrochemically driven mechanisms and offer valuable design principles for optimizing the switching layer and electrode architecture in redox-based RRAM devices.