"Extreme events" in STT-MRAM speed retention and reliability (Conference Presentation)

Abstract

Fast operation speed, high retention and high reliability are the most attractive features of the spin transfer torque magnetic random access memory (STT-MRAM) based upon perpendicular magnetic tunneling junction (pMTJ). For state-of-the-art pMTJ STT-MRAM, its device performance is fundamentally determined by material “extreme events” physics. For example, nanosecond write bit error rate is determined by extremely high probability (>(1-10^(-7))) stochastic magnetization switching events, retention is determined by magnetization configurations with extremely low switching probability, reliability is determined by extremely low probability (<10^(-15)) tunneling junction break-down events. Despite their critical importance, accurately modeling, testing and prediction of have been a long-standing challenging physics and engineering issue due to their low occurrence rates. In this presentation, we will discuss our unique modeling and testing approaches to understand and predict in STT-MRAM write, read, retention and reliability. Specifically, we will present our model that accurately calculates extremely low write BER for various magnetization configurations. We will review our study of thermal magnetization switching through the dynamic optimal reversal path approach, capable of characterizing extreme thermal magnetization switching events under both low frequency (e.g. static retention) and high frequency (e.g. fast read) excitations. We will also discuss a new MTJ breakdown reliability model that quantifies extreme events uniformly at different failure mode regions.