Variability assessment and mitigation by predictive programming in Pr0.7Ca0.3MnO3 based RRAM

Filamentary resistive switching based RRAM (e.g. HfO₂) depends on soft breakdown mechanism which is fundamentally stochastic and hence prone to variability. In comparison, manganites based non-filamentary resistive switching is explored from a variability perspective. We demonstrate excellent within device (WID) variability reduction in resistance levels (11x for HRS and >10x for LRS) compared to HfO₂ RRAM. DC sweep based device to device (DTD) variability in a 6 level programming process is measured to show large but partially systematic variations. We propose a predictive programming scheme to take advantage of the partially systematic nature of PCMO variability to demonstrate resistance level distributions with >1.5x improvement in variability compared to conventional programming. The PCMO DTD variability for HRS is also 2-4x lower than the WID HfO₂ variability.