pipette
ENEnglish

Challenging Atomistic Foundation Models on Disordered Carbons

Joseph Hart, Samuel G. H. Brookes, Fabian L. Thiemann, Christoph Schran

Preprint

En palabras de los autores

Foundation machine-learned interatomic potentials (MLIPs) promise broad transferability, yet their out-of-domain reliability is hard to establish. Here, we use pure and O/H-containing disordered carbon spanning multiple phases, competing hybridisations and porous frameworks to stress-test six popular foundation MLIPs. We first screen all six on density functional theory (DFT) energy and force errors and binding curves, then subject four to melt-quench-anneal molecular dynamics and evaluate the resulting structural observables. Binding curves and O/H-containing trajectories reveal unphysical extrapolation and dynamical failures that low DFT energy and force errors fail to reveal. Of the models tested, MACE-MH-1 is the most consistently robust, with low DFT errors and stable MD. Despite differing architectures, OMAT-based models share structural biases, highlighting the importance of training coverage. All four models generate structures of nanoporous carbon that reproduce the principal experimental and specialised carbon potential signatures. Incorporating O/H leaves short-range order largely unchanged but shifts pore-size distributions towards larger pores. These results emphasise the need for application-specific validation and extend molecular dynamics modelling of nanoporous carbon towards experimentally realistic structures.

Resultado principalEl resumen no menciona limitaciones.

Apareció: martes, 20 de octubre. arXiv. Preprint, todavía sin revisión por pares.

Comentario de los autores: 40th Conference on Neural Information Processing Systems (NeurIPS 2026). Workshop: Sim2Science: ML with Imperfect Scientific Models