Brownian Reservoir Computing Innovative power-saving calculators utilize magnetic vortices and ambient heat

Related Vendors

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

Researchers at the Johannes Gutenberg University Mainz have developed a first prototype of an innovative computing platform that utilizes, among other things, ambient thermal energy and magnetic skyrmions. The combination of Brownian computing and reservoir computing is designed to be particularly energy-efficient.

A large portion of the energy used today is consumed in the form of electrical power for processing and storing data, as well as for operating the necessary devices. And this share will continue to rise – that's what many forecasts unanimously predict. Novel concepts, such as neuromorphic computing, use energy-saving approaches to solve this problem. In a joint project by experimental and theoretical physicists at the Johannes Gutenberg University Mainz (JGU), such an approach has now been implemented as part of an ERC Synergy Grant: Brownian Reservoir Computing. An article on this has been published in the journal "Nature".

Brownian Reservoir Computing is a combination of two unconventional computing methods: Brownian computing and reservoir computing. Brownian computing takes advantage of the fact that the processes required for computations usually take place at room temperature and, thus, the thermal energy of the environment can be used. This saves electricity. Thermal energy often manifests as the random movement of particles. This "Brownian motion" is, therefore, the namesake of the project.

Reservoir computing for particularly efficient data processing at low cost

In reservoir computing, on the other hand, the complex reaction of a physical system to external stimuli is exploited to process data in a particularly resource-efficient manner. The majority of the "computing work" is performed by the physical system itself, which does not require additional energy. According to the researchers, it is particularly impressive that such a reservoir computer can be very easily adapted to various tasks, as the physical system does not need to be adjusted.

Scientists led by Prof. Dr. Mathias Kläui from the Institute of Physics at the JGU, with the help of Prof. Dr. Johan Mentink from the Dutch Radboud University Nijmegen, have now succeeded in developing a prototype that combines these two computing methods. The prototype is capable of performing Boolean logic operations, which can be used as standard tests for reservoir computing.

For the physical system, magnetic skyrmions occurring in metal thin films were chosen. These magnetic vortices behave like particles and can be moved by electric currents. The behavior of the skyrmions is influenced not only by the applied current but also by their Brownian motion. The latter can lead to significant energy savings, as the Brownian motion of the skyrmions automatically resets the computer after each operation and prepares it for the next calculation.

First prototype developed in Mainz

Although there have been many theoretical concepts for skyrmion-based reservoir computing in recent years, it was the combination with the Brownian computing concept that enabled the Mainz scientists to develop the first functioning prototype. "The prototype is lithographically easy to manufacture and can hypothetically be reduced to nanometer size," says experimental physicist Klaus Raab.

"This success is due to the excellent collaboration between experimental and theoretical physics here in Mainz," emphasizes theoretical physicist Maarten Brems. Project leader Mathias Kläui adds: "I am thrilled that support from a Synergy Grant of the European Research Council has made it possible to collaborate with excellent colleagues from theoretical physics in Nijmegen, resulting in our joint work. I see great potential in unconventional computing, which is also excellently supported here in Mainz by the Carl Zeiss Foundation in the project Emergent Algorithmic Intelligence." (me)