Using Magnetic Spin Textures for Cognitive Computing

Jean Anne Incorvia¹, Sam Liu¹, Thomas Leonard¹, Can Cui¹, Mahshid Alamdar¹, Suyogya Karki¹, Vivian Rogers¹, Otitoaleke Akinola¹, Priyamvada Jadaun², Joseph S. Friedman³, Christopher H. Bennett⁴, T. Patrick Xiao⁴, Matthew J. Marinella⁴, and David Paydarfar¹

¹The University of Texas at Austin, USA
²IMEC, Belgium
³The University of Texas at Dallas, USA
⁴Sandia National Laboratories, Albuquerque, New Mexico, USA

There are rich dynamical behaviors in magnetic materials that are bio-mimetic and highly applicable to cognitive computing. Here, we will present our recent results on understanding and leveraging the materials properties of magnetic domain walls (DWs), skyrmions, and magnetic tunnel junctions (MTJs) for energy efficient cognitive computing. We will show that DW motion in a magnetic track can be engineered to have highly linear position vs. time behavior, acting as an artificial synapse [1-2]. We will show a brain-inspired “edgy-relaxed” behavior seen in biological neurons can be implemented inherently in the DW-MTJ device as an artificial neuron. [3]. We will show results on designing a coupled skyrmions neuron that can be dynamically modulated based on context and environment [4].

References:

[1] S. Liu, T. P. Xiao, C. Cui, et al., “A domain wall-magnetic tunnel junction artificial synapse with notched geometry for accurate and efficient training of deep neural networks.” Applied Physics Letters 118, 202405 (2021). DOI 10.1063/5.0046032.

[2] T. Leonard, S. Liu, M. Alamdar, et al., “Shape-Dependent Multi-Weight Magnetic Artificial Synapses for Neuromorphic Computing.” Under review. ArXiv: 2111.11516. https://arxiv.org/abs/2111.11516.

[3] S. Liu, C. H. Bennett, J. S. Friedman, et al., “Controllable reset behavior in domain wall-magnetic tunnel junction artificial neurons for task-adaptable computation.” IEEE Magnetics Letters 12, 4500805 (2021). DOI 10.1109/LMAG.2021.3069666.

[4] P. Jadaun, C. Cui, S. Liu, et al., “Adaptive and advanced cognition implemented with a context-aware and flexible neuron for next-generation artificial intelligence.” Under review. ArXiv: 2010.15748. https://arxiv.org/abs/2010.15748.