Study Of Rtn Signals In Resistive Switching Devices Based On Neural Networks

Random Telegraph Noise (RTN) in Resistive Random Access Memories (RRAM) is an important phenomenon both for the investigation of device physics and for reliability issues. The characteristics of these signals depend on the number of active traps, on the interaction between these traps at different times, on the occurrence of anomalous effects, etc. Using the Locally Weighted Time Lag Plot (LWTLP), a fast numerical procedure, data from RTN current-time (I-t) traces can be represented with a pattern that allows a deeper understanding of the device physics. In the context of self-organizing maps, a neural network devoted to clustering, we have analyzed the LWTLPs to classify the RTN traces obtained from a long measurement with more than 3 million data points. This RTN pattern classification, obtained in an unsupervised learning scheme, allows a comprehensive characterization of the signals and the physics underlying the device operation.