Joseph S. Friedman


Magnetic Domain Wall Neuron

The magnetic domain wall neuron emulates the behavior of neurobiological neurons through manipulation of a magnetic domain wall. This is the first artificial neuron that intrinsically provides the leaking, integrating, firing, and lateral inhibition capabilities without any additional devices or circuitry. This structure is used to perform handwritten digit recognition with 94% accuracy.

Leaky Integrate-and-Fire Neuron

Artificial neurons and synapses are integrated into neural networks to emulate the behavior of neurobiological systems. While artificial synapses need only maintain a resistance state, artificial neurons must provide more complex behavior as exemplified by the leaky integrate-and-fire neuron model. Though neuron behavior can be implemented in software or with inefficient CMOS/electronic hardware, the ideal artificial neuron would provide the leaking, integrating, and firing capabilities within a single device.

The three-terminal magnetic domain wall neuron is the first artificial neuron that provides these three capabilities without requiring any additional devices or circuitry. As illustrated in the micromagnetic simulation video, these three functionalities are achieved through manipulation of the position of the domain wall along the ferromagnetic nanowire track.
  • Leaking: in the absence of any external excitation, the magnetic field from the fixed ferromagnet pushes the domain wall to the left.
  • Integrating: the application of current through the ferromagnetic track pushes the domain wall to the right through spin-transfer torque.
  • Firing: when the domain wall crosses beneath the tunnel barrier, the resistance across the tunnel barrier switches.

Intrinsic Lateral Inhibition

Neurobiological neurons have been shown to exhibit lateral inhibition in addition to leaking, integrating, and firing, whereby integration and firing in one neuron impedes integration and firing in a neighboring neuron. This magnetic domain wall neuron intrinscially provides this lateral inhibition behavior, as the magnetic field created by one neuron can inhibit integration by a neighboring neuron. This is the first artificial neuron proposal that provides lateral inhibition without any additional devices or circuitry.

Handwritten Digit Recognition

These laterally-inhibited domain wall neurons can be integrated into a neural network to perform neuromorphic tasks such as handwritten digit recognition. The presentation of each handwritten digit causes current to flow through synapses to the ten-neuron output layer, where each domain wall neuron represents the recognition of a particular digit. As shown in the video, the neurons integrate the current by moving the domain walls to the right, eventually firing and resetting the neuron output layer. This neuron output layer achieved a 94% recognition rate of the handwritten digits.

Related Publications

  1. W. H. Brigner, N. Hassan, L. Jiang-Wei, X. Hu, D. Saha, C. H. Bennett, M. J. Marinella, J. A. C. Incorvia, F. Garcia-Sanchez, J. S. Friedman, Shape-based Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron, IEEE Transacations on Electron Devices (accepted).
  2. O. Akinola, X. Hu, C. H. Bennett, M. Marinella, J. S. Friedman, J. A. C. Incorvia, Three-Terminal Magnetic Tunnel Junction Synapse Circuits Showing Spike-Timing-Dependent Plasticity, Journal of Physics D: Applied Physics 52:49, 49LT01 (2019).
  3. W. H. Brigner, N. Hassan, X. Hu, L. Jiang-Wei, D. Saha, C. H. Bennett, M. J. Marinella, F. Garcia-Sanchez, J. A. C. Incorvia, J. S. Friedman, Magnetic Domain Wall Neurons with Intrinsic Leaking, Conference on Magnetism and Magnetic Materials, Nov. 2019.
  4. C. Cui, N. Hassan, C. H. Bennett, M. J. Marinella, J. S. Friedman, J. A. C. Incorvia, Optimized Lateral Inhibition in Magnetic Domain Wall Tracks for Neuromorphic Computing, Conference on Magnetism and Magnetic Materials, Nov. 2019.
  5. N. Hassan, X. Hu, L. Jiang-Wei, W. H. Brigner, O. G. Akinola, F. Garcia-Sanchez, M. Pasquale, C. H. Bennett, J. A. C. Incorvia, J. S. Friedman, Magnetic Domain Wall Neuron with Intrinsic Leaking and Lateral Inhibition Capability, Proc. SPIE Spintronics XII, Aug. 2019 (invited).
  6. O. G. Akinola, M. Alamdar, N. Hassan, X. Hu, T. Leonard, J. S. Friedman, J. A. C. Incorvia, Three-Terminal Magnetic Tunnel Junctions for In-Memory and Neuromorphic Computing, Proc. SPIE Spintronics XII, Aug. 2019 (invited).
  7. C. H. Bennett, J. A. C. Incorvia, X. Hu, N. Hassan, J. S. Friedman, M. M. Marinella, Semi-Supervised Learning and Inference in Domain-Wall Magnetic Tunnel Junction (DW-MTJ) neural Networks, Proc. SPIE Spintronics XII, Aug. 2019 (invited).
  8. W. H. Brigner, X. Hu, N. Hassan, C. H. Bennett, J. A. C. Incorvia, F. Garcia-Sanchez, J. S. Friedman, Graded-Anisotropy-Induced Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron, IEEE Journal on Exploratory Solid-State Computational Devices and Circuits 5:1, 19-24 (2019).
  9. N. Hassan, X. Hu, L. Jiang-Wei, W. H. Brigner, O. G. Akinola, F. Garcia-Sanchez, M. Pasquale, C. H. Bennett, J. A. C. Incorvia, J. S. Friedman, Neuromorphic Computing with Domain Wall-Based Three-Terminal Magnetic Tunnel Junctions: Neurons, Joint IEEE International Magnetics Conference & Conference on Magnetism and Magnetic Materials, Jan. 2019.
  10. O. Akinola, E. J. Kim, N. Hassan, J. S. Friedman, J. A. C. Incorvia, Neuromorphic Computing with Domain Wall-Based Three-Terminal Magnetic Tunnel Junctions: Synapse, Joint IEEE International Magnetics Conference & Conference on Magnetism and Magnetic Materials, Jan. 2019.
  11. N. Hassan*, X. Hu*, L. Jiang-Wei, W. H. Brigner, O. G. Akinola, F. Garcia-Sanchez, M. Pasquale, C. H. Bennett, J. A. C. Incorvia, J. S. Friedman, Magnetic Domain Wall Neuron with Lateral Inhibition, Journal of Applied Physics 124:15, 152127 (2018).

This research is sponsored in part by the National Science Foundation under CCF award #1910800.