R&D: Nonvolatile Bipolar Optical Fiber Memory Based on Phase Change Materials
Bipolar memory can implement synaptic weight update mechanism within neural network system’s synapses, thereby showcasing excellent storage, programming, and neural network computing potential.
This is a Press Release edited by StorageNewsletter.com on January 15, 2024 at 2:36 pmACS Photonics has published an article written by Jiapeng Sun, Hanyang Li, Yifan Qin, Siying Cheng, Jinhua Mou, Jiaxing Gao, Yu Zhang, and Zhihai Liu, College of Physics and Optoelectronic Engineering, Harbin Engineering University, Harbin 150001, China
Abstract: “Developing an optical fiber storage platform that integrates storage and computation poses a challenge. This article presents a bipolar memory that incorporates phase change materials by combining two tapered fibers with a microsphere. The state of Ge2Sb2Te5 is altered by modulating an external laser, resulting in repeatable or randomly accessible five-level data storage. Multistage writing is achieved using a 532 nm pump laser and laser energy levels ranging from 0.423 to 1.206 mJ, while a 793 nm continuous wave laser with an average power of 4 to 11 mW completes the multistage reset. The bipolar memory demonstrates nonvolatility, high signal-to-noise ratio, excellent repeatability, 75 ns write response time, 180 ns reset response time, and 18 dB contrast. Furthermore, we develop a continuous programming platform using quaternary ASCII coding as an alternative to English letters. The bipolar memory can implement the synaptic weight update mechanism within the neural network system’s synapses, thereby showcasing excellent data storage, programming, and neural network computing potential.“