Electrical Characterization of TiO₂-Based RRAM for Neuromorphic Computing Applications

Abstract

Resistive Random Access Memory (RRAM) is a promising memory technology because of its simple structure, high storage density, low power consumption, CMOS-compatibility, and non-volatile characteristics. TiO₂-based RRAM exhibits bipolar resistive switching through the formation and rupture of oxygen vacancy based conductive filaments. Beyond providing potential alternatives for conventional flash memory, RRAM is also attractive for neuromorphic computing, where programmable conductance states can emulate biological synaptic weights and enable energy-efficient in-memory computation. In this study, we investigate behaviors of 10-µm Al/TiO₂/Au RRAM devices consisting of an Al active electrode, a TiO₂ insulating switching layer, and an Au passive electrode. Devices were fabricated using photolithography, thin-film deposition, and lift-off, with Au deposited by thermal evaporation, TiO₂ deposited by RF sputtering, and Al deposited by DC sputtering. Measurements were performed with a semiconductor parameter analyzer attached to a probe station, by sweeping voltage and recording current. A thorough electrical characterization of electroforming behavior, endurance, ON/OFF ratio, and cycle-to-cycle variability was conducted. Intermediate resistance states were also obtained by tuning SET compliance current or RESET voltage to control conductive-filament formation and partial rupture. Along with a memory window of roughly 10x, the device demonstrated programming capability for over 8 conductance states. This result features the potential of Al/TiO₂/Au RRAM as an analog synaptic device for neuromorphic computing applications.