Instability of HfO2 RRAM devices: Comparing RTN and cycling variability

In this study, we present an extensive statistical characterization of the cycling variability and Random Telegraph Noise (RTN) in the HfO₂-based Resistive Random Access Memories (RRAM) cells. Devices with different dielectric stacks are tested under a variety of read (sampling times and read voltage magnitudes) and operational (reset voltages) conditions. A Factorial Hidden Markov Model (FHMM) analysis is employed to reveal the properties of the traps causing multi-level RTN in High Resistive State (HRS), while the I-V data are analyzed through the developed compact model to investigate cycling variability. The activation and deactivation of traps assisting the charge transport through a dielectric barrier in HRS is found to be responsible for the observed RTN while the read current variations can be attributed to the stochastic nature of the filament oxidation process during reset, also leading to a variable number of traps formed in the barrier after each switching cycle. The statistical characterization of RTN and cycling variability, which demonstrates the uncorrelated nature of these phenomena, provides guidelines for scaling and optimization of RRAM device operations and reliability.