Keywords: Modeling & Simulation
Abstract: In contrast to conventional electronics that deal with the charge of an electron, spintronics utilizes the spin of an electron to manipulate, transmit, store and detect information. Spintronic devices can be fabricated using back-end-of-the-line CMOS processes and, therefore, realized in modern fabrication facilities without much re-tooling. Spintronics offers great opportunities for realizing unconventional and domain-specific computing paradigms that have advantages in terms of energy and area efficiency compared to von Neumann computing. Examples of unconventional information processing approaches are neuromorphic computing, probabilistic computing, Ising machines, Bayesian inference, and processing-in-memory, etc. In this talk, I will discuss the state-of-the-art in diverse spintronic devices including non-volatile digital memory, memristors, p-bits, high-frequency oscillators, and signal modulators, that can enhance the overall circuit functionality of CMOS chips. I will present my group’s recent work in developing multi-physics and multi-scale models that can facilitate the heterogeneous integration of spintronics with silicon. I will use these models to explain recent experimental findings and bridge the gap between physics and applications to unconventional computing and signal processing. I will conclude my talk by summarizing the limits, challenges, and opportunities of spintronics and guiding the development and deployment of most promising “More-than-Moore” and “Beyond Moore” technologies.