Device Modeling and Circuit Design for Beyond-CMOS and Artificial Intelligence Hardware

Abstract

Emerging technologies provide potential solutions to overcome the limitations of modern CMOS technologies. Specifically, as power density limitations impede further CMOS scaling, emerging technologies including spintronics, memristors, ambipolar transistors, and other beyond-CMOS devices are promising replacements for conventional CMOS transistors due to features such as non-volatility, low energy consumption, high operation speed, or high logical expressiveness. However, the inadequacies of current modeling techniques inhibit the design of large-scale systems that exploit these novel devices. This dissertation describes a variety of proposals for beyond-CMOS computing systems using emerging technologies. Spintronic technologies such as domain wall-magnetic tunnel junctions (DW-MTJs) and magnetic skyrmions are particularly exciting for highly-efficient non-volatile information processing. Additionally, unconventional electronic switching devices including ambipolar transistors and memristors are also investigated as replacements to CMOS and for hybrid emerging technology-CMOS computing systems. These beyond-CMOS devices and circuits exhibit the versatility and scalability of emerging technologies. By exploiting low energy switching, reconfigurability, non-volatility and unique operating mechanisms, these emerging technologies provide the potential to revolutionize information processing for the next generation of computing system.