Master of Science (MS)
Hiu Yung Wong
Nanotechnology; Electrical engineering; Materials Science
Artificial intelligence (AI) has become a key enabler in many applications but requires fast and energy-efficient computation. IBM is exploring an “Analog AI" acceleration approach using phase change memories (PCM). As such, there is a need to accurately model the physics inside each PCM cell, including the interplay between thermal and electrical dynamics, and the impact of PCM polycrystal grains and grain boundaries in materials such as Ge2Sb2Te5. The thesis uses an existing thermo electrodynamic Technology Computer-Aided Design (TCAD), which was created at IBM, to simulate the intermediate resistance states critical to the “Analog AI" application. The thesis discusses the simulations thermal behavior, simulates against industrial tools, and uses the three-dimensional conservation of energy to confirm the requirement of a TCAD tool. As an accurate thermal simulator, the thesis continues to discuss and implement Poole Frenkel to simulate against experimental PCM device data. Thereafter, it discusses the complication of chalcogenide material and the impact of the polycrystalline electrical impact of grain boundaries. The availability of an accurate TCAD simulation is expected to have significant impact on the understanding and design of future Analog AI systems.
Saltin, Johan, "Physics-Based Modeling of Phase-Change Memory Devices and Materials" (2021). Master's Theses. 5187.