This Thin Magnetic Rectangle Could Power High-Performance, High-Efficiency Spintronic Computers

Researchers at Japan's Tohoku University have come up with a theoretical model for a spintronic computer that, they say, could represent a major leap towards brain-like high-performance neuromorphic systems for machine learning and artificial intelligence workloads: spin-wave reservoir computing.

"Our study proposed a physical RC [reservoir computing architecture] that harnessed propagating spin waves," explains Natsuhiko Yoshinaga, co-author and associate professor at Tohoku University's Advanced Institute for Materials Research (WPI-AIMR). "The theoretical framework we developed utilized response functions that link input signals to propagating spin dynamics. This theoretical model elucidated the mechanism behind the high performance of spin wave RC, highlighting the scaling relationship between wave speed and system size to optimize the effectiveness of virtual nodes."

While neuromorphic computing, which is based on the human brain, isn't a new concept, Yoshinaga's team argues for a new approach that can break free from the confines of traditional electronic computers. Instead, the team's framework takes the concept of reservoir computing — which allows for the memorization of previous information and its non-linear transformation through a fixed randomly-generated "reservoir" network — and makes it physical, rather than as a software layer running on traditional computing hardware.

The team's proposed physical reservoir computing architecture uses the propagation of spin waves, based on spintronics rather than electronics, to perform operations at a rate that the researchers say should be "comparable with other state-of-the-art systems." The proposed reservoir is a thin rectangular magnetic device with cylindrical "injectors" to which power can be inserted in order to trigger spin-waves that propagate through the system — though, it must be noted, the device is as yet entirely theoretical, with no physical prototype yet developed.

Despite this, the researchers are confident in the potential of the approach. "By employing the unique properties of spintronics technology, we have potentially paved the way for a new era of intelligent computing," Yoshinaga claims, "leading us closer to realizing a physical device that can be put to use in weather forecasts and speech recognition."

The team's work has been published in the journal npj Spintronics under open-access terms.