BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Oxidation-aware machine learning interatomic potential for magneti
 te Fe3O4
DTSTART:20260506T130000Z
DTEND:20260506T140000Z
DTSTAMP:20260512T174700Z
UID:indico-event-18836@indico.psi.ch
CONTACT:matthias.krack@psi.ch
DESCRIPTION:Speakers: Libor Vojacek (PSI)\n\nMagnetite Fe3O4 is the oldest
  known magnetic material\, yet the physics of its Verwey transition is sti
 ll not fully solved due to its coupled structural\, charge\, and orbital i
 nteractions. The difficulty hinges on the multivalent nature of iron (Fe2+
 /Fe3+)\, which is a central motif in redox chemistry and functional oxides
 .Machine learning interatomic potentials (MLIP) trained on DFT+U+V data ha
 ve recently proven to efficiently tackle the oxidation state of multivalen
 t metallic ions with near ab initio accuracy [1]\, adding to their intrins
 ic efficiency enabling the simulations of large and/or complex systems. Th
 is is especially relevant for magnetite where the low-temperature monoclin
 ic phase requires a large (224 atom) supercell with a combinatorial number
  of 64 choose 32 ~ 1018 possible charge orderings\, making ab initio calcu
 lations expensive. In this work\, we train an oxidation-aware MLIP for mag
 netite and demonstrate its capabilities side-by-side with a conventional (
 unaware) fine-tuned foundation model [2]. While the oxidation-aware potent
 ial provides the oxidation state directly as a control variable\, the conv
 entional potential can capture it intrinsically through the distinct volum
 es of the octahedral oxygen cages [3].\n\nC. Malica and N. Marzari\, Npj C
 omput. Mater. 11 (2025) 212\nS. Batzner\, A. Musaelian\, L. Sun\, M. Geig
 er\, J. P. Mailoa\, M. Kornbluth\, N. Molinari\, T. E. Smidt\, and B. Kozi
 nsky\, Nat Commun 13 (2022) 2453\nL. Vojáček and N. Marzari\, in prepara
 tion (2026)\n\n\nhttps://indico.psi.ch/event/18836/
LOCATION:OVGA/200
URL:https://indico.psi.ch/event/18836/
END:VEVENT
END:VCALENDAR