Researchers turn wasted spin into power for memory chips

A surprising twist in physics could reshape low-power computing. A team led by the Korea Institute of Science and Technology (KIST), with collaborators at DGIST and Yonsei University, has shown that “spin loss,” long treated as wasted energy in spintronic devices, can instead be harnessed to switch magnetization, the fundamental operation behind data storage and computation. Their study, published in Nature Communications, reports a threefold improvement in energy efficiency using this new principle, potentially accelerating the development of ultra-low-power memory and AI semiconductors.

Spintronics and the problem of spin loss

Spintronics relies on the spin property of electrons, rather than just their charge, to encode digital information. Magnetic materials can store data in the form of spin orientation: up equals 1, down equals 0. Traditionally, flipping these states requires forcing electron spins into the magnet with a strong current. But much of that spin leaks away, a loss that researchers have struggled to minimize because it wastes power and reduces efficiency.

Turning loss into a resource

The new work reframes this waste as an advantage. By directly channeling current into the magnetic layer, the team found that spin loss itself produces a reaction force strong enough to flip magnetization. Counterintuitively, the greater the spin loss, the easier it becomes to switch the state. This inversion of conventional wisdom suggests that dissipation can be engineered to make devices more efficient, rather than less.

“Until now, the field of spintronics has focused only on reducing spin losses, but we have presented a new direction by using the losses as energy to induce magnetization switching,” said Dr. Dong-Soo Han, senior researcher at KIST.

Experimental proof and scalability

In systematic tests, the researchers built simple device structures pairing ferromagnetic metals with antiferromagnetic insulators. These combinations allowed magnons—quantized spin waves—to dissipate energy in a way that generated internal torque sufficient to reverse magnetization. Devices fabricated with this approach consumed less current and achieved deterministic switching, all without the need for exotic materials or complicated architectures.

The design is also compatible with semiconductor fabrication methods already in use, making industrial application more feasible. According to the team, the principle could support:

• AI semiconductors and neuromorphic chips
• Ultra-low-power memory and logic circuits
• Probabilistic and stochastic computing hardware
• Edge devices requiring efficient, small-scale integration

Key Findings

• Sample: Devices combining ferromagnetic metals with antiferromagnetic insulators
• Method: Harnessing magnonic spin dissipation instead of minimizing it
• Result: Up to 3x higher energy efficiency in magnetization switching
• Location: Korea Institute of Science and Technology (KIST), DGIST, Yonsei University
• Publication: Nature Communications, July 2025
• Safety/Scalability: Works with existing semiconductor processes, suitable for miniaturization

Implications for the AI era

Harnessing spin loss opens a new design space for spintronic materials. Instead of battling dissipation, engineers may deliberately enhance it to lower energy costs. This is particularly important for artificial intelligence, where both data storage and computation are power-intensive. By enabling magnetization control with less current, the discovery supports efforts to reduce energy demand while scaling down device size.

Takeaway

Researchers in Korea have demonstrated that spin loss, once considered a waste, can serve as a power source for switching magnetization in spintronic devices. The approach triples efficiency and aligns with standard semiconductor manufacturing, positioning it as a promising foundation for ultra-low-power memory and AI computing technologies.

Journal: Nature Communications
DOI: 10.1038/s41467-025-61073-w