Share Email Print

Proceedings Paper

Mutually synchronized spin Hall nano-oscillators for neuromorphic computing (Conference Presentation)
Author(s): Mykola Dvornik; Ahmad A. Awad; Philipp Dürrenfeld; Afshin Houshang; Ezio Iacocca; Randy K. Dumas; Johan Åkerman

Paper Abstract

Deep Machine Learning is the emerging brain-inspired computing approach that employs artificial neural networks to solve such important problems as image and voice recognition, market behavior prediction, etc. It however still relies on digital CMOS technologies that approach their fundamental limits. As a consequence, there is now significant research activity aimed at finding hardware platforms that would allow for the native implementation of the artificial neural networks. There are already models available that describe human brain operation via synchronization phenomena in complex networks of nonlinear oscillators. This research topic remains mostly theoretical, or numerical, since large-scale oscillator networks are needed, but not easily implemented. However, it was recently demonstrated that so-called spin torque and spin Hall nano-oscillators can act as artificial neurons [1], and their propensity for mutual synchronization on the nano-scale can open up for very large non-linear oscillator networks with different degrees of mutual interactions. To this end, we here present the first experimental demonstration of mutual synchronization of nano-constriction spin Hall nano-oscillators (SHNOs) [2]. The mutual synchronization is observed both as a strong increase in the power and coherence of the electrically measured microwave signal. The mutual synchronization is also optically probed using scanning micro-focused Brillouin light scattering microscopy (µ-BLS), providing the first direct imaging of synchronized nano-magnetic oscillators. By tailoring the connection region between the nano-constrictions, we have been able to synchronize SHNOs separated by up to 4 micrometers. In addition, we have demonstrated mutual synchronization of as many as nine SHNOs. Our results opens up a direct route for the design of very large SHNO based oscillator networks and pave the way for the development of a spintronic brain-inspired computing technology. [1] J. Grollier, D. Querlioz, M.D. Stiles, PIEEE 104, 2024 (2016) [2] A. A. Awad, P. Dürrenfeld, A. Houshang, M. Dvornik, E. Iacocca, R. K. Dumas and J. Åkerman, Nature Physics 13, 292–299 (2017).

Paper Details

Date Published: 19 September 2017
Proc. SPIE 10357, Spintronics X, 103572J (19 September 2017); doi: 10.1117/12.2278026
Show Author Affiliations
Mykola Dvornik, Göteborgs Univ. (Sweden)
Ahmad A. Awad, Göteborgs Univ. (Sweden)
Philipp Dürrenfeld, Göteborgs Univ. (Sweden)
Afshin Houshang, Göteborgs Univ. (Sweden)
Ezio Iacocca, Göteborgs Univ. (Sweden)
Randy K. Dumas, Göteborgs Univ. (Sweden)
Johan Åkerman, Göteborgs Univ. (Sweden)

Published in SPIE Proceedings Vol. 10357:
Spintronics X
Henri-Jean Drouhin; Jean-Eric Wegrowe; Manijeh Razeghi; Henri Jaffrès, Editor(s)

© SPIE. Terms of Use
Back to Top