A group of over 90 AI agents powered by Claude Opus 5.5 identified two material candidates for next-generation spintronic memory within three days. Both materials are classified as Luttinger-compensated antiferromagnets, which allow for spin sorting without generating parasitic magnetic fields.
The first candidate, a manganese and iron oxide, was designed from scratch by the AI. While its electronic properties exceed thermal noise at room temperature, the required synthesis temperature of 900–1300 degrees Celsius makes achieving the necessary atomic order difficult.
The second candidate, KV[Cr(CN)6], was first synthesized in 1999. Although its spin-sorting properties were previously overlooked, the material maintains magnetic order up to 376 K. Researchers have now identified its potential as a semiconductor for high-speed electronics.
Calculations for both materials were performed using density functional theory and published on GitHub. Experimental data from a 1999 powder sample showed a residual magnetic moment of 0.125 Bohr magnetons per unit, though further synthesis is required for verification.
