Physical Simulation of the ReRAM reliability and retention through Kinetic Monte Carlo modeling

In the era of advanced technology , non-volatile (NV) Resistive Random Access Memory (ReRAM or RRAM) represents a revolutionary paradigm, able to withstand the Big Data requirement. Thanks to its intrinsic low power consumption, low cost, fast access times and high memory density, RRAM is forecasted to be a promising heir of flash technology. Based on a simple two terminal metal-insulator-metal (in this work, said to be TiN/HfO2/Ti) structure, its conductivity can be controlled by the application of a proper bias voltage during the programming phase, where the device moves from SET (HRS -> LRS) to RESET (LRS -> HRS) state. This study revolves around the assessment of Oxide-based RRAM (OxRAM) retention capability at high (up to +150°C) operating temperatures. This is accomplished through Kinetic Monte Carlo (KMC) modeling, which takes into account the stochastic nature of resistive switching (RS). Conductive filament (CF) morphology variations and spread LRS resistance distributions are analysed and the long-term device performance is examined simulating accelerated life tests (ALTs). By focusing previously on the Ti/HfO2 interface effects and, then, to CF dissolution due to Vo spreading, several annealing simulations are carried out for distinct baking times, up to 10 hours, and temperatures, up to 230°C. Their effect, in conjunction with a right choice of activation energy parameters, is studied in terms of resistance drift distributions and QQ plots. Ultimately, the OxRAM lifetime is extracted through Weibull and Time-to-Failure graphs, exhibiting a slope of Ea=1.56eV= Ea,ion|TiOx and a maximum 10 years temperature of 106°C, which makes it a reliable device for typical commercial NVM applications.