A glass-box foundation model for density functional theory

Abstract

Machine-learned exchange–correlation functionals promise accuracy beyond traditional forms, yet so far they trade interpretability for flexibility. Here we introduce DL-xDH26, a glass-box foundation model for density functional development on the fifth rung of Jacob’s ladder, with separate exchange and correlation expert towers, interpretable enhancement-factor and shift parameters relative to a physical base, and exact degree-one homogeneity. Trained exclusively on relative energies, it recovers atomic exchange and correlation energies and their quantum shell structure, and produces smooth, size-consistent dissociation curves without dispersion corrections, while reaching errors below 1 kcal/mol on the GMTKN55 and MGCDB84 benchmark suites. Lightweight fine-tuning then specializes the same backbone to cyclization reactions, distorted organic geometries, and organometallic reaction energetics while retaining its broad accuracy.

Octupole deformation in even-even Ra isotopes from covariant density functional theory with localized exchange terms in a three-dimensional lattice space
Z.Y. Dong, Z.X. Ren, Q. Zhao +1
Physics Letters B30 Oct 2026Peer-reviewedCC BY 4.0Full textNuclear physicsCrossref
The inverse problem of time-dependent density functional theory on the torus
Asbjørn Bækgaard Lauritsen, Mathieu Lewin, Jakob Oldenburg
HAL8 Sep 2026PreprintarXiv licenceFull textPhysical chemistry Mathematical physicsHALarXiv
Bulk Boundary Condition for Semi-Infinite Surfaces in Density Functional Theory
Sayan Bhowmik, Andrew J. Medford, Phanish Suryanarayana
Journal of Chemical Theory and Computation16 Oct 2026Peer-reviewedCC BY 4.0Full textAtomic, molecular & optical physicsPMC
Time-Dependent Density Functional Theory with Coulomb Interactions
Asbjørn Bækgaard Lauritsen, Mathieu Lewin, Jakob Oldenburg
HAL8 Sep 2026PreprintarXiv licenceFull textPhysical chemistry Mathematical physicsHALarXiv
Current density functional theory in the age of generalized Kohn-Sham theories
Giovanni Vignale
arXiv24 Sep 2026PreprintarXiv licenceFull textCondensed matterarXiv

Cite

BibTeX RIS

BibTeX

@misc{bi2026glassbox,
  title = {{A glass-box foundation model for density functional theory}},
  author = {Sheng Bi and Shiqi Liu and Igor Ying Zhang and Xin Xu},
  howpublished = {ChemRxiv},
  year = {2026},
  month = sep,
  doi = {10.26434/chemrxiv.15008506/v1},
  url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008506/v1},
  note = {CC BY},
}

RIS

TY  - UNPB
TI  - A glass-box foundation model for density functional theory
AU  - Sheng Bi
AU  - Shiqi Liu
AU  - Igor Ying Zhang
AU  - Xin Xu
PY  - 2026
DA  - 2026/09/08
PB  - ChemRxiv
DO  - 10.26434/chemrxiv.15008506/v1
UR  - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008506/v1
AB  - Machine-learned exchange–correlation functionals promise accuracy beyond traditional forms, yet so far they trade interpretability for flexibility. Here we introduce DL-xDH26, a glass-box foundation model for density functional development on the fifth rung of Jacob’s ladder, with separate exchange and correlation expert towers, interpretable enhancement-factor and shift parameters relative to a physical base, and exact degree-one homogeneity. Trained exclusively on relative energies, it recovers atomic exchange and correlation energies and their quantum shell structure, and produces smooth, size-consistent dissociation curves without dispersion corrections, while reaching errors below 1 kcal/mol on the GMTKN55 and MGCDB84 benchmark suites. Lightweight fine-tuning then specializes the same backbone to cyclization reactions, distorted organic geometries, and organometallic reaction energetics while retaining its broad accuracy.
ER  -

Abstract

Related

Cite