Deep Learning GW Quasiparticle Hamiltonians for Many-Body Excited-State Electronic Structure at Scale

Abstract

Accurate quasiparticle electronic structures are the foundation for understanding excited-state properties of materials and explaining optoelectronic, quantum, and transport phenomena. First-principles GW calculations nevertheless remain computationally intensive for large or configurationally complex systems. Here we introduce DeepH-GW, a deep-learning framework that predicts an effective GW quasiparticle Hamiltonian directly from atomic structure. Building on the local, equivariant message-passing architecture of DeepH, DeepH-GW is trained on high-fidelity plane-wave GW calculations through a real-space Hamiltonian-reconstruction interface. This approach combines the systematic accuracy and broad chemical applicability of plane-wave methods with linear-scaling neural-network inference. Across the systems examined, DeepH-GW reproduces quasiparticle band structures with errors on the order of a few meV. Moreover, we show that, despite the intrinsic nonlocality of many-body interactions, DeepH-GW exhibits strong cross-scale transferability from relatively small training structures to substantially larger supercells. We demonstrate that the framework can accurately capture GW-level electron-phonon band-gap renormalization through predictions in supercells with thermal displacements, illustrating the power of the approach. DeepH-GW therefore provides a practical route toward large-scale many-body simulations and foundation models for excited-state electronic structure.

The Multi-Electron Unified Wave Equation in Quantum Chemistry: Symmetrized Riesz Exchange Kernels, Sobolev Domain Classification, and O(N4) Quasiparticle Electronic Structure across Bond Dissociation
Mehardeep Singh
ChemRxiv22 Oct 2026PreprintCC BY-NC-NDFull textAtomic, molecular & optical physicsChemRxiv
Optical characterization of excited-state spin Hamiltonians and clock transitions at telecommunication wavelength in Tm3+:YAlO3
Akshay Babu Karyath, Luozhen Li, Julien Bertrand +6
arXiv25 Sep 2026PreprintCC BY 4.0Full textAtomic, molecular & optical physics Optics & photonicsarXiv
Dynamical Phonon Anharmonicity Drives Charge-Density-Wave Softening and Electronic Quasiparticle Decoherence in Monolayer 1H-NbSe2
Yi Xia
arXiv2 Nov 2026PreprintarXiv licenceFull textCondensed matterarXiv
Can end-to-end learning from raw electronic structure explain magnetic anisotropy?
Rafał Topolnicki, Jan Navrátil, Piotr Błoński
arXiv8 Oct 2026PreprintCC BY 4.0Full textComputational physics Condensed matterarXiv
Electronic-Structure-Informed Many-Body Encoding for Quantum Machine Learning
Liangnuo Zhang, Tanner Culpitt, Qiwei Yin +2
arXiv29 Oct 2026PreprintarXiv licenceFull textPhysical chemistry Quantum physics & quantum informationarXiv

Cite

BibTeX RIS

BibTeX

@misc{gong2026deep,
  title = {{Deep Learning GW Quasiparticle Hamiltonians for Many-Body Excited-State Electronic Structure at Scale}},
  author = {Xiaoxun Gong and Zechen Tang and Woochang Kim and Yang Li and Wenhui Duan and Yong Xu and Steven G. Louie},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.36962},
  eprint = {2609.36962},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.36962},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Deep Learning GW Quasiparticle Hamiltonians for Many-Body Excited-State Electronic Structure at Scale
AU  - Xiaoxun Gong
AU  - Zechen Tang
AU  - Woochang Kim
AU  - Yang Li
AU  - Wenhui Duan
AU  - Yong Xu
AU  - Steven G. Louie
PY  - 2026
DA  - 2026/09/29
PB  - arXiv
DO  - 10.48550/arxiv.2609.36962
UR  - https://arxiv.org/abs/2609.36962
AB  - Accurate quasiparticle electronic structures are the foundation for understanding excited-state properties of materials and explaining optoelectronic, quantum, and transport phenomena. First-principles GW calculations nevertheless remain computationally intensive for large or configurationally complex systems. Here we introduce DeepH-GW, a deep-learning framework that predicts an effective GW quasiparticle Hamiltonian directly from atomic structure. Building on the local, equivariant message-passing architecture of DeepH, DeepH-GW is trained on high-fidelity plane-wave GW calculations through a real-space Hamiltonian-reconstruction interface. This approach combines the systematic accuracy and broad chemical applicability of plane-wave methods with linear-scaling neural-network inference. Across the systems examined, DeepH-GW reproduces quasiparticle band structures with errors on the order of a few meV. Moreover, we show that, despite the intrinsic nonlocality of many-body interactions, DeepH-GW exhibits strong cross-scale transferability from relatively small training structures to substantially larger supercells. We demonstrate that the framework can accurately capture GW-level electron-phonon band-gap renormalization through predictions in supercells with thermal displacements, illustrating the power of the approach. DeepH-GW therefore provides a practical route toward large-scale many-body simulations and foundation models for excited-state electronic structure.
ER  -

Abstract

Related

Cite