Physics-based Simulation of RRAM for Reliable Memory Technology

Abstract

Resistive Random Access Memory (RRAM) is an emerging non-volatile memory candidate that offers fast switching speed, low power consumption, and high scalability. RRAM functions by applying a positive voltage to an electrode to form a conductive filament through an insulating switching layer, resulting in high current flow (SET), and applying a negative voltage to rupture the filament and reduce current flow (RESET). However, its relatively large device-to-device and cycle-to-cycle variability remains a critical challenge, limiting RRAM’s reliability. While possible sources of variability have been identified, a comprehensive understanding of the impact of specific device parameters on RRAM’s variability is still missing. This work creates a novel, physics-based model of RRAM that simulates its switching behavior by tracking the evolution of both the conductive filament and the gap distance between the filament and top electrode. By updating these lengths at discrete time intervals while considering various parameters including temperature, electrode resistance, and activation energy, the model accurately calculates current as a ratio of the filament length to the total switching layer thickness and produces a current-voltage (I-V) characteristic of the simulated RRAM device. These I-V curves match experimental RRAM devices with the same parameters, tested using a probe station, validating the model. The study found that a local temperature rise inside a device may cause cycle-to-cycle variability. Activation energy and switching layer thickness were found to be important physical parameters in device-to-device variability, with activation energy decreasing and switching layer thickness increasing SET/RESET voltages and memory window. These findings provide quantitative guidelines for designing and experimentally developing RRAM devices with increased reliability in emerging computing hardware applications.