The Pulse
Vals AI Agents Identify Two Magnetic Semiconductor Candidates
Vals AI says Claude Opus 5.5 agents helped design YBaMnFeO₅ and identify KV[Cr(CN)₆] as candidates for spin-selective semiconductors. The predictions come from simulations; neither material’s spin sorting has been measured, and the new comp
AI.info Team · Written with AI ·
A 376 K result—and an important distinction
A magnetic powder first made in 1999 stayed magnetically ordered up to 376 kelvin, above room temperature. That result is not evidence that its spin-selective electronic behavior has been measured: Vals AI says the band gap and spin sorting remain untested. In a post dated October 4, researcher Geby Jaff describes calculations by a team of Claude Opus 5.5 agents that identify the old material, KV[Cr(CN)₆], alongside a newly designed compound, YBaMnFeO₅.
The pair is proposed as a route to spintronic memory: materials that could sort electrons by spin while producing little or no net magnetism. Vals AI presents both as candidates, not working memory materials. Its calculations predict the desired properties in ideal crystal structures, and the company shares its inputs, outputs and analysis code in a public computational ledger.
Why the calculations target spin windows
In ordinary antiferromagnets, neighboring magnetic moments cancel, but electrons at a given energy are not necessarily sorted by spin. Luttinger-compensated materials seek to combine that near-zero net moment with different local environments for opposite spins, allowing electrons to sort by spin at the edges of a semiconductor’s band gap.
Vals AI calls the energy range with a single spin orientation a “spin window.” Room-temperature thermal fluctuations are about 26 millielectronvolts, the post says; larger windows could help preserve spin sorting against that thermal activity. The team used density functional theory, a quantum-mechanical simulation method, with PBE+U and HSE06 approximations. The reported band gaps and spin windows come from HSE06.
YBaMnFeO₅ has a predicted ordering problem
The newly designed compound contains yttrium, barium, manganese, iron and oxygen. Vals AI says it could not find evidence that YBaMnFeO₅ had previously been made or proposed as this kind of magnet. Its calculations predict a 2.35 eV band gap, with spin windows of 1.0 eV for holes and 1.4 eV for electrons.
The same simulations estimate magnetic order up to about 420 K, or about 490 K after calibrating against a known magnet. But the prediction depends on manganese and iron occupying a precise checkerboard arrangement. Vals AI’s simulations indicate that arrangement becomes disordered at around 950 K; the post says making this oxide typically requires temperatures of 900 to 1,300 degrees Celsius, which could scramble the structure and erase its spin sorting.
KV[Cr(CN)₆] still needs a direct test
The second candidate is a Prussian-blue-family material, KV[Cr(CN)₆], that was first made in 1999. Vals AI calculates a band gap of about 2.1 eV and spin windows of 2.6 eV for holes and 1.6 eV for electrons. The original hydrated powder showed a small residual magnetic moment—0.125 Bohr magnetons per formula unit—and remained ordered up to 376 K, according to the post.
Those measured magnetic properties belong to the old, water-containing powder, not a newly verified sample with the predicted semiconductor behavior. The calculations for KV[Cr(CN)₆] instead describe a perfect, dry crystal. Vals AI also reports that its two simulation methods disagree about water’s effect: HSE06 predicts that spin sorting survives, while PBE+U predicts that the hole window shrinks by more than half.
Jaff writes: “The next step is to make KV[Cr(CN)₆] again and measure its spin sorting directly.” Geby Jaff, Vals AI