A concept of optically triggered and electrically controlled ultra-fast neuromorphic computing processor based on an antiferromagnetic/heavy metal (AFM/HM) heterostructure is proposed. The AFM/HM-based artificial neurons are excited with short THz-range pulses, triggering precession in AFM. Bias electric current in the HM layer can be used to modify the resonance frequency of precession. The conversion of the precession into the electric current in the HM-layer occurs via the inverse spin Hall effect. A model of a neuromorphic processor is, thus, proposed, consisting of excitatory AFM-based artificial neurons—oscillators, and processing neurons—detectors. We show that the use of optical excitation can significantly increase the processing speed of neuromorphic computing at low power consumption. Examples of the implementation of the simplest logical operations (OR, AND) are demonstrated.
Optically initialized and current-controlled logical element based on antiferromagnetic-heavy metal heterostructures for neuromorphic computing
Note: This paper is part of the APL Special Collection on Ultrafast and Terahertz Spintronics.
A. Mitrofanova, A. Safin, O. Kravchenko, S. Nikitov, A. Kirilyuk; Optically initialized and current-controlled logical element based on antiferromagnetic-heavy metal heterostructures for neuromorphic computing. Appl. Phys. Lett. 14 February 2022; 120 (7): 072402. https://doi.org/10.1063/5.0079532
Download citation file: