This strange magnetism could enable tomorrow's AI
Ultra-thin films of ruthenium dioxide (RuO2) have been shown to exhibit altermagnetism-a magnetic state now recognized as a third fundamental category. That single result checks three boxes that matter for memory: speed, density, and reliability. For AI and high-performance compute, that combination is the bottleneck.
Why a third kind of magnetism matters
Ferromagnets are easy to write but sensitive to stray fields, which limits how tightly you can pack bits. Antiferromagnets are stable but hard to read electrically because their spins cancel out. Altermagnets thread the needle: zero net magnetization with spin-split electronic bands, enabling electrical readout while staying resistant to external magnetic noise.
RuO2 has long been a candidate, but results across labs were inconsistent. The sticking point was materials quality and crystallographic orientation in thin films.
What the team did
An international group from NIMS, The University of Tokyo, Kyoto Institute of Technology, and Tohoku University grew RuO2 thin films with a single crystallographic orientation on sapphire. Controlling the substrate choice and growth conditions removed the variability that had dogged prior attempts.
They used X-ray magnetic linear dichroism to map spin order and confirmed that the net magnetization cancels. Crucially, they detected spin-split magnetoresistance-direct electrical evidence of a spin-split band structure. First-principles calculations of magneto-crystalline anisotropy matched the measurements, closing the loop and confirming altermagnetism in RuO2.
Why it matters for memory and AI workloads
Altermagnetic RuO2 thin films point to memory that writes fast, packs densely, and shrugs off stray fields. That reduces error rates at scale and supports tighter integration near compute. Less overhead in error correction and shielding translates to better performance per watt.
- Higher areal density without magnetic cross-talk from neighbors.
- Electrical readout via spin-split transport, despite zero net magnetization.
- Orientation control as a practical knob to tune device performance.
Methods that speed up discovery
The synchrotron-based magnetic analysis pipeline used here is more than a one-off. It provides a template to screen and characterize other altermagnets quickly, helping the spintronics community converge on materials that scale in real devices and accelerating related Research.
What's next
The team plans to develop memory devices built on RuO2 thin films. Key steps include integrating orientation-controlled films with standard stacks, validating write/read speeds, retention, and endurance, and confirming that the altermagnetic signatures persist at device dimensions and operating temperatures.
Research team and funding
- Team: Zhenchao Wen, Cong He, Hiroaki Sukegawa, Seiji Mitani, Tadakatsu Ohkubo (NIMS); Jun Okabayashi (The University of Tokyo); Yoshio Miura (Kyoto Institute of Technology); Takeshi Seki (Tohoku University).
- Funding: JSPS Grants-in-Aid for Scientific Research (22H04966, 24H00408); MEXT X-NICS (JPJ011438); GIMRT Program (IMR, Tohoku University); Cooperative Research Projects (RIEC, Tohoku University).
The study was published in Nature Communications on September 24, 2025. See the journal for details: Nature Communications.
Your membership also unlocks:
