R&D: Programming and Read Performances Optimization of PCM Via Multi-Objective Genetic Algorithm and Improved Finite Element Analysis

SSRN has published an article written by Pu Tang, Ming Tao, and Jing Xiao, Hunan University, 2 Lushan South Rd, Changsha, CA 410082, China.

Abstract: “Phase Change Memory (PCM) has attracted widespread attention for its high-speed reading and writing capabilities and high-density non-volatile storage. However, reducing the reset current of PCM can result in an increase in its set resistance, which affects the programming or readout performances. Therefore, further optimization of PCM under this trade-off remains a significant challenge. In this paper, multi-objective genetic algorithm(MOGA) is utilized to address this problem through the optimization of PCM’s geometry. An enhanced PCM finite element model, considering the Seebeck, Peltier, and Thomson effects, is employed as the fitness function of MOGA to improve the accuracy of PCM’s reset arguments and obtain precise solution. The PCM geometry solution set with optimal performance in different technology nodes are presented in this paper. Specifically, for the 22-nm node, the optimization result shows that the reset current can be reduced by up to 10.3% when the PCM is employed for low-power applications, such as neuromorphic computing, and the read latency time can be reduced by up to 27.6% when the PCM is utilized for fast reading purposes, such as in PCM-DRAM main memory. This proposed optimization scheme has great potential in improving the performance of PCM.“