Methodology for stability evaluation on the Multi-level storages of oxide-based conductive bridge RAM (CBRAM)

Multi-level per cell (MLC) is a prime approach to realize high-density storage for resistive random access memory (RRAM). In this work, we investigated the data retention and random telegraph noise (RTN) characteristics on intermediate resistance states (IRS) which were programmed in Set and Reset process for the first time. The main observations are as follows: 1. Although the IRS-SET have higher decay factor `τ´, the IRS-Reset has better data retention than IRS-Set; 2. The MLC-Set states exhibit lower mean P2p values and better resistance distribution compared with the IRS-Reset states. A comprehensive physical model was proposed to account for the results. Moreover, implicative approaches to increase the robustness of MLC storage are suggested.