Learning Electron Density with Virtual Atoms for Transferable Atomistic Representations

Abstract

Electron density is a fundamental quantity of electronic structure, yet most machine learning density models treat it solely as a reconstruction target and are developed separately from machine learning interatomic potentials. Here, we introduce a virtual-atom formulation within the DPA3 framework, in which density query points are represented as virtual nodes and processed by the same parameterized encoder used for real atoms. An asymmetric neighbor construction allows virtual atoms to receive information from nearby real atoms without altering the real-atom interaction graph. The resulting framework supports density, energy, and force prediction through task-specific readout heads. The proposed model achieves state-of-the-art density reconstruction accuracy on the Materials Project benchmark, matches the best reported performance on nickel-manganese-cobalt battery cathode materials, and is competitive with dedicated density models on QM9. When density labels are used jointly with energy and force supervision, the model improves in-domain accuracy, makes local environment information more accessible under linear probing, and reduces zero-shot force errors on larger QM40 molecules. Density-only checkpoints also provide useful initializations for downstream energy-force learning under the fixed fine-tuning protocol. Together, these results support electron density as a source of electronic-structure supervision for interatomic-potential representations.

Electronic configuration vectors provide transferable elemental representations for machine learning across the periodic table
Raghavendra Rajagopalan, Ananth Govind Rajan
ChemRxiv3 Sep 2026PreprintCC BY-NC-NDFull textMaterials & polymersChemRxiv
Response-state Learning for Transferable Vibrational Spectroscopic Characterization with Electron Prior
Zetong Li, Zhuosong Xie, Hengyu Fan +6
arXiv20 Oct 2026PreprintarXiv licenceFull textPhysical chemistryarXiv
ROSE: Deep Learning exchange-correlation functional from reduced density-matrix representations
Kirill Kulaev, Alexander Ryabov, Bogdan Protsenko +4
arXiv1 Nov 2026PreprintCC BY 4.0Full textPhysical chemistry Computational physicsarXiv
Hybridizing expert descriptors and learned embeddings yields transferable Mat2Vec-* element representations
Xiaotong Liu, Qiong Feng, Haotian Chen +2
Communications Chemistry30 Oct 2026Peer-reviewedCC BY-NC-ND 4.0Full textGeneral chemistryNatureCrossrefEurope PMCPMC
Transformer-inspired architectures make machine-learned electron densities accurate and transferable enough to bypass self-consistency
Jeheon Woo
arXiv29 Oct 2026PreprintarXiv licenceFull textPhysical chemistry Computational physicsarXiv

Cite

BibTeX RIS

BibTeX

@misc{liu2026learning,
  title = {{Learning Electron Density with Virtual Atoms for Transferable Atomistic Representations}},
  author = {Yuzhi Liu and Xinzijian Liu and Chengqian Zhang and Anyang Peng and Duo Zhang and Han Wang},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.14521},
  eprint = {2610.14521},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.14521},
}

RIS

TY  - UNPB
TI  - Learning Electron Density with Virtual Atoms for Transferable Atomistic Representations
AU  - Yuzhi Liu
AU  - Xinzijian Liu
AU  - Chengqian Zhang
AU  - Anyang Peng
AU  - Duo Zhang
AU  - Han Wang
PY  - 2026
DA  - 2026/10/10
PB  - arXiv
DO  - 10.48550/arxiv.2610.14521
UR  - https://arxiv.org/abs/2610.14521
AB  - Electron density is a fundamental quantity of electronic structure, yet most machine learning density models treat it solely as a reconstruction target and are developed separately from machine learning interatomic potentials. Here, we introduce a virtual-atom formulation within the DPA3 framework, in which density query points are represented as virtual nodes and processed by the same parameterized encoder used for real atoms. An asymmetric neighbor construction allows virtual atoms to receive information from nearby real atoms without altering the real-atom interaction graph. The resulting framework supports density, energy, and force prediction through task-specific readout heads. The proposed model achieves state-of-the-art density reconstruction accuracy on the Materials Project benchmark, matches the best reported performance on nickel-manganese-cobalt battery cathode materials, and is competitive with dedicated density models on QM9. When density labels are used jointly with energy and force supervision, the model improves in-domain accuracy, makes local environment information more accessible under linear probing, and reduces zero-shot force errors on larger QM40 molecules. Density-only checkpoints also provide useful initializations for downstream energy-force learning under the fixed fine-tuning protocol. Together, these results support electron density as a source of electronic-structure supervision for interatomic-potential representations.
ER  -

Abstract

Related

Cite