Investigation on the RESET switching mechanism of bipolar Cu/HfO₂/Pt RRAM devices with a statistical methodology

The RESET switching of bipolar Cu/HfO 2 /Pt resistance random access memory (RRAM) is investigated. With a statistical methodology, we systematically analyze the RESET voltage ( V RESET ) and RESET current ( I RESET ). V RESET shows a U-shape distribution as a function of R ON according to the scatter plot of the raw experimental data. After data correction by a series resistance ( R S ), V RESET is nearly constant, while I RESET decreases linearly with R CF . These behaviours are consistent with the thermal dissolution model of RESET. Moreover, the I RESET and V RESET distributions are strongly affected by the R ON distribution. Using a 'resistance screening' method, the I RESET and V RESET distributions are found to be compatible with the Weibull distribution model. The Weibull slopes of the V RESET and I RESET distributions are independent of R CF , indicating that the RESET point corresponds to the initial phase of conductive filament (CF) dissolution, according to our cell-based model for the unipolar RESET of RRAM devices. The scale factor of the V RESET distributions is roughly constant, while that of the I RESET distributions scale with 1/ R CF . Accordingly, the RESET switching of the HfO 2 -based solid electrolyte memory is compatible with the thermal dissolution mechanism, improving our understanding on the physics of resistive switching of RRAM devices.