Researchers from the Universities of Münster and Heidelberg, led by Prof. Rudolf Bratschitsch, have developed a new method that allows spin waves to travel significantly longer distances through waveguides, paving the way for more energy-efficient AI hardware. As the demand for AI grows, so does the strain on global energy systems—driving the need for innovative computing technologies. Spin waves, which can carry and process information, offer a promising alternative. Using their new technique, the team built the largest spin waveguide network to date and demonstrated precise control over wave properties such as wavelength and boundary reflections. The breakthrough was published in Nature Materials.

Figure 1. Visualization of a Spin Waveguide Network Created with an Ion Beam (Bottom: Antenna and Network; Top Right: Ion Beam; Top Left: Spin Wave).

Exploring Spin Waves and Their Magnetic Properties

Electron spin, a quantum property representing intrinsic angular momentum, plays a key role in determining a material's magnetic characteristics through the collective alignment of spins. When an alternating current is applied to a magnetic material using an antenna, it produces a varying magnetic field that can excite spin waves within the material. Figure 1 shows Visualization of a Spin Waveguide Network Created with an Ion Beam (Bottom: Antenna and Network; Top Right: Ion Beam; Top Left: Spin Wave).

Spin waves have previously been harnessed to build individual components like logic gates—which convert binary inputs into outputs—and multiplexers that choose among multiple input signals. However, these components had not yet been integrated into a larger, connected circuit—until now.

According to Rudolf Bratschitsch, the challenge in building larger spin-wave networks—similar to those in electronics—has been the significant attenuation of spin waves within nanoscale waveguides, especially those narrower than a micrometer. To overcome this, the research team used yttrium iron garnet (YIG), a material known for having the lowest spin-wave attenuation [1]. They etched individual waveguides into a 110-nanometer-thick YIG film using a silicon ion beam and successfully created a large network consisting of 198 nodes. This new method enables the flexible and reproducible fabrication of complex, high-quality spin-wave structures.

References:

1. https://scitechdaily.com/ai-hardware-revolution-scientists-create-largest-spin-waveguide-network/

Cite this article:

Janani R (2025), AI Hardware Breakthrough: Scientists Build Largest-Ever Spin Waveguide Network, AnaTechMaz, pp. 242