Improved Learning of Molecular Energetics Through an Electron-Wise Joint Charge Density and Energy Objective

Abstract

We present a graph neural network-based learning framework trained to jointly predict electron densities and molecular energies. The model is trained to predict electron densities calculated with a Generalized Gradient Approximation (GGA) functional while simultaneously learning to predict molecular energies obtained with hybrid functional calculations at the B3LYP level of theory. Our results indicate that the rich spatial information in electron density distribution can be used to improve and accelerate the learning of accurate energies. By sharing an equivariant molecular representation across density and energy prediction heads, the model learns complementary molecular quantities within a unified framework. This provides a new promising route for practical use cases of many recent charge density learning frameworks for atomic scale simulations.

The energetics of force errors in machine-learned molecular dynamics
Peng Kang, Da Wan, Shulin Bai +5
arXiv8 Sep 2026PreprintarXiv licenceFull textComputational physics Condensed matterarXiv
Surrogate-assisted multi-objective topology optimization design method for prefabricated large-span Suspen-Dome based on structural-mapping-method and improved-PGSA
Yuchen Dai, Ailin Zhang, Liuhaoxiang Wang +3
PLOS One2 Nov 2026Peer-reviewedCC BYFull textMaterials & polymers Algebra & number theoryPLOSEurope PMCPMC
Improved electron beam stability in laser wakefield acceleration through control of laser angular chirp and wavefront stability
Zhan Jin, Yan-Jun Gu, Shingo Sato +8
Physical Review Research23 Oct 2026Peer-reviewedCC BY 4.0Full textHigh-energy & particle physicsAPSCrossrefDOAJ
Long-Range Machine Learning Interatomic Potentials for Defect Energetics in SrTiO₃
Yee Chit Wong, Marcel F. Langer, Reinhard J. Maurer +1
arXiv1 Oct 2026PreprintarXiv licenceFull textCondensed matterarXiv
Interplay between photon condensation and electron-electron interactions in molecular systems
Matteo Parisi, Elisabetta Paladino, Giuseppe A. Falci +4
arXiv20 Oct 2026PreprintarXiv licenceFull textCondensed matter Quantum physics & quantum informationarXiv

Cite

BibTeX RIS

BibTeX

@misc{ionas2026improved,
  title = {{Improved Learning of Molecular Energetics Through an Electron-Wise Joint Charge Density and Energy Objective}},
  author = {Vadim Ionas and Jonas Elsborg and Felix Ærtebjerg and Arghya Bhowmik},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.32858},
  eprint = {2609.32858},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.32858},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Improved Learning of Molecular Energetics Through an Electron-Wise Joint Charge Density and Energy Objective
AU  - Vadim Ionas
AU  - Jonas Elsborg
AU  - Felix Ærtebjerg
AU  - Arghya Bhowmik
PY  - 2026
DA  - 2026/09/26
PB  - arXiv
DO  - 10.48550/arxiv.2609.32858
UR  - https://arxiv.org/abs/2609.32858
AB  - We present a graph neural network-based learning framework trained to jointly predict electron densities and molecular energies. The model is trained to predict electron densities calculated with a Generalized Gradient Approximation (GGA) functional while simultaneously learning to predict molecular energies obtained with hybrid functional calculations at the B3LYP level of theory. Our results indicate that the rich spatial information in electron density distribution can be used to improve and accelerate the learning of accurate energies. By sharing an equivariant molecular representation across density and energy prediction heads, the model learns complementary molecular quantities within a unified framework. This provides a new promising route for practical use cases of many recent charge density learning frameworks for atomic scale simulations.
ER  -

Abstract

Related

Cite