Statistical, simulation and modeling analysis of variability in memristors with single and bilayer dielectrics of HfO2 and Al2O3, a comparison

Variability is a key aspect of memristors that hinders their usage in massive commercial applications. This study investigates the cycle-to-cycle variability of resistive switching devices fabricated using three different dielectric configurations: two monolayer insulators (Al 2 O 3 or HfO 2 ) or a HfO 2 /Al 2 O 3 bilayer. Thousands of resistive switching (RS) current-voltage (I-V) curves were measured under ramped voltage stress and the RS parameters were extracted using different numerical methodologies. The variability of the obtained RS parameters was analyzed using the conventional 1D coefficient of variation (σ/μ, where σ stands for the standard deviation and μ for the mean), as well as a recently introduced 2D coefficient of variation, providing a deeper insight into the joined variability of the switching voltages and currents. An enhanced version of the Stanford compact model was employed to model the experimental I-V curves, in addition to an analysis based on a 3D circuit breaker (CB) simulator, which incorporates thermal effects for controlling the switching of the CB. These two modeling approaches successfully captured the observed RS behavior and variability trends along the thousands of measured cycles and for the three considered technologies. Furthermore, the simulation study sheds light on how the different thermal properties of the studied technologies impact on the conductive filament evolution. • Variability is studied in memristors with HfO2, Al2O3 oxides. • An advanced statistical analysis is performed to study variability. • Compact modeling is employed to characterize the measurements. • A 3D simulator based in circuit breakers is used for the study. • Memristors with a bilayer of HfO 2 and Al 2 O 3 have less variability.