Skip to content- Sources: primary, ledger, discussion
- Summary: A write-up reports that a team of Claude Opus 5.5 agents ran density functional theory calculations at two levels of approximation, PBE+U and HSE06, alongside literature search, and put forward two Luttinger-compensated magnetic semiconductor candidates. Neither the band gap nor the spin sorting has been measured for either candidate, so the result is simulation and literature search rather than a discovery. The published ledger records five corrections to the agents' own records: a YBaMnFeO5 hull distance of +13.7 meV/atom rather than +2.6, computed before competing phases finished and no longer at the edge of stability, an HSE06 run the run index marked converged whose exact-exchange loop had stopped early, two partly relaxed geometries, one truncated output, and a 2008 hybrid-functional paper by Middlemiss, Lawton and Wilson whose spin-resolved plot already shows the same-spin band edges, which narrows the novelty claim.
- Why it matters: The published inputs, raw outputs and ledger made five errors in the agents' own records findable, which is the part of an agent-science claim a reader can check.
- Follow-up: Watch for a measured band gap or spin sorting result for either candidate.