Energy-saving computers with magnetic vortices

Researchers at Johannes Gutenberg University Mainz (JGU) have succeeded in improving the framework for Brownian reservoir computation by recording and transmitting hand gestures to a system that then uses skyrmions to detect these individual gestures. “We were impressed to see that our hardware approach and concept worked as well as – or even better than – power-hungry software solutions using neural networks,” says Grischa Beneke, a member of Professor Mathias Kläui’s research group at the JGU Institute of Physics. Working with other experimental and theoretical physicists, Beneke was able to demonstrate that simple hand gestures can be recognized using Brownian reservoir computation with a relatively high degree of precision.

Tank calculations do not require any training and reduce energy consumption

Tank computing systems are similar to artificial neural networks. Their advantage is that they do not require extensive training, which reduces their overall energy consumption. “All we have to do is train a simple output mechanism to map the output,” Beneke explained. The exact computational processes remain unclear and are not important in detail. The system can be compared to a pond into which rocks have been thrown, creating a complex pattern of ripples on the surface. In the same way that the ripples suggest the number and position of rocks thrown in, the output mechanism of the system provides information about the original input.

In their latest article recently published in Nature communicationresearchers describe how they recorded simple hand gestures, such as a left or right swipe, with Range-Doppler radar, using two Infineon Technologies radar sensors. The radar data is then converted into appropriate voltages, which are fed to a tank, which in this case consists of a multilayer stack of thin films made of different materials, formed into a triangle with contacts at each corner. Two of the contacts supply a voltage that causes the skyrmion to move in the triangle. “In response to the signals provided, we detect complex movements,” described Grischa Beneke. “These movements of the skyrmion allow us to infer the movements that the radar system has recorded.” Skyrmions are chiral magnetic vortices that are believed to have great potential for use in unconventional computing devices and as information carriers in innovative data storage devices. “Skyrmions are truly amazing. Initially, we considered them only as candidates for data storage, but they also have great potential for applications in computers in combination with sensor systems,” emphasized Professor Mathias Kläui, supervisor of this research field at JGU.

Comparison of the results obtained using Brownian reservoir calculations with those recorded using a software-based approach shows that the accuracy of gesture recognition is similar or even better with Brownian reservoir calculations. The advantage of combining reservoir calculations with the Brownian computation concept is that the skyrmions are free to perform random movements, since local differences in magnetic properties have less influence on the way they respond. This means that the skyrmions, unlike how they usually respond, can be moved using only very low currents – a significant improvement in energy efficiency compared to the software approach. Since the data collected by the Doppler radar and the internal dynamics of the reservoir operate on similar time scales, the sensor data can be fed directly into the reservoir. The time scales of the system can be adapted to solve many other problems.

“We found that radar data of different hand gestures are detected on our hardware resource with an accuracy at least as good as that of the state-of-the-art software-based neural network approach,” the researchers concluded in their paper. Nature communicationAccording to Beneke, further improvements should be possible in the reading process, which is currently used by the magneto-optical Kerr microscope (MOKE). Using a magnetic tunnel junction instead could help reduce the size of the entire system. Signals delivered by the magnetic tunnel junction are already being emulated to demonstrate the capacity of the tank.