AI hardware for neuromorphic computing applications – memory device fabrication and characteristics

• Main Author: Liu, Jixuan
• Other Authors: Zhang Dao Hua
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/166459
• Institution: Nanyang Technological University

In this project, an overview of the field of both conventional and emerging memory technologies is provided. Then, a novel Resistive Random Access Memory (RRAM) is proposed to design and fabricate. Besides, RRAM devices under different parameters are tested and the performance test results are analysed. The presented RRAM is one kind of three-terminal electronic synapse device. The main innovative part of the device is the use of chalcogenide material, Ge2Sb2Te5 (GST), instead of traditional transition metal oxide (TMO). Under the applied electric field, GST acts as an adjustable conductive path by allowing small metal atoms like Ag to diffuse in. By changing the structure and thickness of the electrodes and GST layer, devices with analog resistance switching characteristics and different electrical properties are investigated. Eventually, a high on/off ratio (~20) of the device with good linearity for conductance updates is achieved. Due to the superiority of good linearity of conductance update and low power consumption, this new type of RRAM is supposed to have a promising future with a wide range of application prospects, such as In-Memory Computation, Neuromorphic Computing, Security Applications, and Non-volatile SRAM. Keywords: Resistive Random Access Memory (RRAM), three-terminal electronic synapse devices, Ge2Sb2Te5 (GST), Neuromorphic Computing