Physics Colloquium: Using Magnetic Domain Walls for Cognitive Computing
Prof. Jean Anne C. Incorvia, The University of Texas at Austin
Abstract: Over the past decades, tremendous effort has gone into designing and building computers to accomplish tasks outside the reach of humans. But, despite the revolution in computing that has occurred, there are tasks where humans easily outperform computers. The efficiency and adaptiveness of the mammalian brain enables many immersive tasks, such as language understanding, coherent processing of multiple senses to make a decision, object perception, and consciousness.
To achieve true brain-like computing, we need to understand 1) the advanced cognitive features of the brain to accomplish tasks and 2) how to translate those features to nanoengineered materials and devices that are 3) compatible with circuits and systems to perform actual tasks. While the basic neuromorphic computing building blocks are neurons that integrate and fire, as well as synapses that store a weight, higher-order behaviors of these components are crucial to these tasks.
There are rich dynamical behaviors in magnetic materials that can be used to this end. We will present our recent results on understanding and leveraging the materials properties of magnetic domain walls (DWs) and magnetic tunnel junctions (MTJs) for cognitive computing. We will show that by engineering the lithographic shape of patterned spintronic synapse devices, we can tune their behavior for inference tasks, where linear weight changes and symmetry are important, or for online learning, where avoiding forgetting of previously-learned data is important.
We will simulate how similar devices can also act as neurons with brain-inspired “edgy-relaxed” behavior that is useful for processing repeated tasks. We will also show how interactions between the magnetic devices can emulate the interactions between neurons that is important for energy-efficient learning. The results show that spintronics can be a powerful monolithic platform for advanced computing.
Bio: Jean Anne C. Incorvia is an Assistant Professor and holds the Fellow of Advanced Micro Devices (AMD) Chair in Computer Engineering in the Department of Electrical and Computer Engineering at The University of Texas at Austin, where she directs the Integrated Nano Computing (INC) Lab. Prof. Incorvia develops nanodevices for the future of computing using emerging physics and materials. Dr. Incorvia received her bachelor’s in physics from UC Berkeley in 2008 and her Ph.D. in physics from Harvard University in 2015, cross-registered at MIT. She completed a postdoc at Stanford University before starting at UT Austin in 2017. She has received the 2020 IEEE Magnetics Society Early Career Award, a 2020 US National Science Foundation CAREER award, and is a 2021 Intel Rising Star.