Automated Endurance Characterization of Phase Change Memory

Phase change memory cells are outstanding candidates for processing-in-memory and neuromorphic computing. The high endurance, low cycle-to-cycle variability, and low read noise especially suit many applications, whereas the high cell-to-cell variability poses a challenge, since each cell serves slightly different results. Therefore, automated characterization is necessary to create a sufficient statistical database to thoroughly study the switching behavior. This paper introduces a sophisticated algorithm for the endurance measurement of phase change memory cells using the aixMATRIX setup by aixACCT Systems. The algorithm is able to perform endurance measurements on devices over an entire sample, adjusting the biasing parameters according to the switching behavior of each cell and sorting out nonfunctional cells. The stepping between the cells is performed automatically. We highlight the benefits of our algorithm by providing an in-depth analysis of all devices on one phase change memory in a bridge geometry sample.