Pith. sign in

REVIEW 4 major objections 7 minor 46 references

MBFormer: A General Transformer-based Learning Paradigm for Many-body Interactions in Real Materials

T0 review · 4 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A symmetry-aware transformer learns the many-body hierarchy directly from mean-field DFT inputs, predicting GW quasiparticle and exciton energies to within about 0.2 eV.

desk verdict A useful architecture with a verifiable GW result, but the exciton-energy prediction is under-specified to the point of being unverifiable as written. read the letter →

arxiv 2507.05480 v1 pith:KQ5LA2VR submitted 2025-07-07 cond-mat.mtrl-sci physics.comp-ph

Bowen Hou , Xian Xu , Jinyuan Wu , Diana Y. Qiu This is my paper
classification cond-mat.mtrl-sciphysics.comp-ph
keywords transformermany-bodyperturbationtheoryGWapproximationBethe-Salpeterequationexcitonquasiparticlemachinelearningtwo-dimensionalmaterials
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

MBFormer is a transformer-based architecture that treats each Kohn-Sham state from a DFT calculation as a token, embeds the full set of mean-field states as context, and decodes them into many-body observables. The paper claims that this single paradigm can predict GW quasiparticle corrections, exciton energies, oscillator strengths, and exciton wavefunction amplitudes across hundreds of 2D materials, with mean absolute errors around 0.16-0.20 eV for energies. The goal is to show that the whole post-mean-field hierarchy can be learned end-to-end directly from ground-state inputs, bypassing the expensive construction of self-energy and interaction kernels. If true, this would turn a cheap DFT run into a fast estimator of excited-state properties, useful for high-throughput screening of optical and electronic materials.

What carries the argument

The central object is a grid-free encoder-decoder transformer. Input mean-field Bloch states (wavefunction squared, energy, band index, k-point) are mapped by an equivariant variational autoencoder into compact latent tokens; a basis-assembly module combines those single-particle tokens into electron-hole pair tokens via concatenation. Self-attention in the encoder captures correlations among all mean-field states, cross-attention lets query states interact with the ground-state context, and the last decoder layer projects the mixed representation to scalars (energies, oscillator strengths) or to the distribution of exciton wavefunction amplitudes.

What would settle it

Train or evaluate MBFormer on the same 721 materials but recompute the reference exciton energies with a denser k-grid and more empty bands, or with a different pseudopotential; if held-out MAE changes by more than roughly the claimed 0.2 eV tolerance, the reported accuracy does not transfer beyond the reference workflow.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a single, symmetry-aware transformer, trained on the mean-field Kohn-Sham states of 721 two-dimensional semiconductors, can reproduce state-level GW-BSE results. For G0W0 self-energy corrections the test-set MAE is 0.16 eV with R2=0.969; for exciton energies the MAE is 0.20 eV, dropping to 0.15 eV for the first 100 bound excitons. The model also predicts exciton dipole oscillator strengths to a normalized MAE of 0.003 and exciton wavefunction amplitudes, and it can infer exciton properties on fine k-grids after coarse-grid training. The authors argue that the attention mechanism is what carries the many-body physics: removing self-attention raises the BSE validation MSE from 0.16 to 0.63 eV2, while simply deepening the network cannot compensate.

Load-bearing premise

The 0.16-0.20 eV errors are only meaningful if the reference GW-BSE calculations across the 721 materials are accurate and mutually consistent; the paper gives no convergence tests of those labels, so the model may be learning workflow noise rather than physics.

Editorial extensions

If this is right

  • A single trained model can replace GW and BSE calculations for a range of materials, reducing the cost of excited-state screening from many-body perturbation theory to a forward pass through DFT inputs.
  • Because the model is grid-free and accepts variable numbers of source and query states, the same architecture can be retrained for other many-body targets, such as self-energy operators or higher-order Green's functions.
  • Fine-k-grid inference means the costly BSE convergence step can be skipped at prediction time: a model trained on coarse grids can resolve exciton dispersions, oscillator strengths, and wavefunction distributions on dense grids.
  • The explicit attention-matrix output provides a direct view of which mean-field states each exciton or quasiparticle draws from, offering interpretable access to the many-body correlation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the learned functional is as universal as claimed, a natural extension is to 3D materials, heterostructures, or defect states, but the current reference-label quality and the single database impose an unknown transfer penalty across DFT setups.
  • The reported errors are lower bounds in the sense that they are measured against labels from one consistent workflow; recomputed labels with different pseudopotentials or k-grids would likely show larger deviations.
  • For degenerate or nearly degenerate exciton states, the random rotation over the degenerate subspace acts as label noise; this is visible in the larger JSD for higher-energy excitons and could be treated by equivariant losses rather than raw state matching.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes MBFormer, an encoder-decoder transformer that takes Kohn-Sham DFT states as input tokens and predicts many-body properties derived from GW and GW-BSE calculations. The method is demonstrated on 721 two-dimensional materials from C2DB, with reported MAEs of 0.16 eV for G0W0 quasiparticle corrections and 0.20 eV for exciton energies, together with predictions of oscillator strengths and exciton wavefunction amplitudes. A separate experiment on monolayer hBN claims coarse-to-fine k-grid inference for excitonic properties. The authors argue that the attention mechanism is essential for capturing many-body correlations, based on an ablation that removes self-attention.

Significance. If the claims are substantiated, MBFormer would represent a meaningful step toward machine-learning models that map ground-state DFT data to excited-state many-body observables, including two-particle quantities that are difficult to obtain at DFT cost. The held-out test-set protocol, with the model trained on external GW-BSE labels rather than fitted to the target properties, gives the work a plausible generalization claim, and the inclusion of exciton wavefunction and oscillator-strength prediction goes beyond single-property ML models. However, the manuscript as submitted does not specify how the model converts per-electron-hole-pair decoder outputs into exciton eigenvalues, and it lacks baseline comparisons, error bars, and reference-data convergence analysis. The significance is therefore conditional on the missing BSE decoding specification and on the availability of the supplemental details.

major comments (4)
  1. [Sec. II.D, Eq. (5), Fig. 4(a)] The manuscript does not specify how the per-target-state scalar output of BSE-MBFormer is converted into exciton energies. In Sec. II.B the decoder returns a scalar vector in R^{N' x 1} with one entry per input target state; for the BSE task the target states are electron-hole pairs |cvk>. Exciton energies in Eq. (5) are eigenvalues of the N' x N' BSE Hamiltonian, and no mapping from per-pair scalars to the spectrum is given: there is no sorted-eigenvalue loss, no diagonalization of a predicted kernel, and no assignment rule. The attention row representing |Phi_alpha| is a wavefunction, not an energy. Because the reported exciton MAE of 0.20 eV, the R2 = 0.966 parity plot, and the fine-k-grid R2 = 0.999 result all rest on this unspecified step, the central two-particle claim is not verifiable from the submitted manuscript; the relevant SI Notes are not available to the reader.
  2. [Sec. II.E, first paragraph] The description of the fine-k-grid experiment is ambiguous. The text says the model is trained 'on 6 coarse k-grids with 84 data augmentations (10% of the dataset)' from a total of 900 BSE Hamiltonians; 10% of 900 is 90, not 84, and 'data augmentations' is not defined. It is also unclear whether the training set consists of 6 grids, 84 Hamiltonians, or 84 augmentations of 6 grids, and whether region 1 in Fig. 5(a) is entirely excluded from training. These details are essential for evaluating the extrapolation claim that the model can infer excitonic properties on fine k-grids from coarse-grid training data.
  3. [Sec. II.D, Fig. 4(b)] The ablation that supports the claim that attention is crucial for BSE is not documented as a controlled experiment. The figure compares a 5-layer BSE-MBFormer without and with self-attention, but the text does not state that the two models are matched in parameter count, embedding dimension, training budget, or optimizer settings. Without this information, the observed 75% reduction in validation MSE cannot be attributed specifically to the self-attention module. In addition, the assertion that self-attention within the electron-hole pair basis is sufficient is justified only by a physical argument and by an unavailable SI Note 1; no cross-attention ablation is shown for the BSE model.
  4. [General, Sec. III] The claimed state-of-the-art performance is not supported by baseline comparisons or reference-data uncertainty analysis. The paper does not compare against existing GW-ML models such as the per-state representation model of Ref. [22], nor against simple baselines like linear regression on KS eigenvalues or a GNN-based predictor. The reported results come from a single random train/test split without error bars or multiple seeds, and the GW-BSE labels are assumed consistent and converged across the 721 C2DB materials, but no convergence tests or label-error analysis are presented. Consequently, the reported MAEs may not reflect true generalization error across different DFT setups, pseudopotentials, or material classes.
minor comments (7)
  1. [Abstract] The phrase 'determinant coefficient' should be 'coefficient of determination', and R2 is described as a 'determinant coefficient' in Fig. 4(a).
  2. [Figure captions 3 and 4] The captions contain the typo 'applicaiton' instead of 'application'.
  3. [Sec. II.E] The text says 'scaler value prediction' when it means 'scalar value prediction'.
  4. [Sec. II.B and II.D] Essential technical details are repeatedly deferred to SI Notes 1, 2, 4, and 5, including the E2-VAE wavefunction embedder, the positional encoding of k-points and band indices, and the BSE energy decoding. At least the definitions needed to reproduce the core claims should appear in the main text or in a supplement available to the reviewers.
  5. [General] No data or code availability statement is provided; given the complexity of the pipeline and the reliance on an unpublished supplement, this hinders reproducibility.
  6. [Fig. 5(a) and Sec. II.E] The region labeled 'area 1' is used in the text and figure caption but is only implicitly defined; a precise description of which k-grids belong to area 1 and how it is selected should be stated.
  7. [Eq. (2)] The arrow notation in Eq. (2) represents a mapping whose input and output dimensionalities depend on the material, so the notation should be clarified to avoid implying a fixed-dimensional function.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: predictions are trained on external GW-BSE labels and tested on held-out materials; self-citations are not load-bearing.

full rationale

The central derivation chain is a supervised learning pipeline: DFT mean-field states are embedded by MBFormer, and the model is trained against G0W0 self-energy corrections and BSE exciton energies, oscillator strengths, and wavefunction amplitudes from external C2DB GW-BSE calculations (Secs. II.C-II.E, Figs. 3-5). Test-set MAEs are reported for held-out materials and held-out k-grids, so the claimed predictions are not fitted to the target quantities by construction. The only self-citations are architectural lineage: the E2-VAE 'extending our earlier work that employed a standard VAE' [39] and a prior low-rank electron-hole kernel paper [25] mentioned as related effort; neither supplies the GW-BSE labels nor defines the loss function, and the attention-importance conclusion is supported by an in-paper ablation (Fig. 4b) rather than by citation. The reviewer-identified gap that the BSE decoder's scalar-per-electron-hole-pair output is not explicitly mapped to sorted exciton eigenvalues is a reproducibility and specification omission, not a circular reduction: no equation in the paper defines the predicted exciton energy as the same fitting target used to train the model. Accordingly, no circular step meets the evidence threshold.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The model's learned weights are the ML model itself, not hidden physical constants; the hand-chosen preprocessing choices are listed above.

free parameters (5)
  • GW source-state count (conduction bands included as source tokens) = 60 in convergence study; actual not stated
    Model accuracy depends on how many empty bands are provided; the paper shows validation loss improves from 0.068 to 0.039 eV^2 as this count grows.
  • Number of exciton states used for the energy MAE = First 100 low-energy excitons per material
    The reported MAE improves from 0.20 eV to 0.15 eV when restricting to these states, so the headline error depends on this selection.
  • Oscillator strength max-renormalization = Maximum rescaled to 1
    Degenerate exciton states are normalized to max 1 before training; the reported MAE of 0.003 is in this normalized unit, not absolute oscillator strength.
  • k-grid training set size = 6 coarse k-grids, 84 data augmentations (10% of 900)
    The coarse-to-fine extrapolation result is measured on a model trained on a deliberately small subset; the exact grid shifts and augmentation details are in SI.
  • Architecture hyperparameters (depth, embedding dimension, attention heads) = Not specified in main text
    These affect all reported results and are not given in the arXiv v1.
assumptions (4)
  • domain assumption GW-BSE reference data from C2DB are accurate and converged
    The supervised labels are treated as ground truth; the paper does not provide convergence tests.
  • domain assumption Many-body Hamiltonian can be expressed as a functional of the set of DFT KS orbitals (Eq. 1-2)
    The entire tokenization strategy depends on this representability assumption.
  • ad hoc to paper Self-attention within the BSE basis is sufficient to capture electron-hole interaction; cross-attention to the source environment is unnecessary
    Stated in Section II D with physical explanation but no systematic comparison.
  • ad hoc to paper The ablation without self-attention is a fair control; hyperparameters are matched
    The claim that self-attention causes the accuracy gain assumes the no-attention baseline is equivalent in other respects.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MBFormer: A General Transformer-based Learning Paradigm for Many-body Interactions in Real Materials." pith.science (2026). https://pith.science/paper/KQ5LA2VR

@misc{pith2026250705480,
  author       = {Pith},
  title        = {Pith review of: MBFormer: A General Transformer-based Learning Paradigm for Many-body Interactions in Real Materials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KQ5LA2VR}},
  note         = {Machine review of arXiv:2507.05480}
}
read the original abstract

Recently, radical progress in machine learning (ML) has revolutionized computational materials science, enabling unprecedentedly rapid materials discovery and property prediction, but the quantum many-body problem -- which is the key to understanding excited-state properties, ranging from transport to optics -- remains challenging due to the complexity of the nonlocal and energy-dependent interactions. Here, we propose a symmetry-aware, grid-free, transformer-based model, MBFormer, that is designed to learn the entire many-body hierarchy directly from mean-field inputs, exploiting the attention mechanism to accurately capture many-body correlations between mean-field states. As proof of principle, we demonstrate the capability of MBFormer in predicting results based on the GW plus Bethe Salpeter equation (GW-BSE) formalism, including quasiparticle energies, exciton energies, exciton oscillator strengths, and exciton wavefunction distribution. Our model is trained on a dataset of 721 two-dimensional materials from the C2DB database, achieving state-of-the-art performance with a low prediction mean absolute error (MAE) on the order of 0.1-0.2 eV for state-level quasiparticle and exciton energies across different materials. Moreover, we show explicitly that the attention mechanism plays a crucial role in capturing many-body correlations. Our framework provides an end-to-end platform from ground states to general many-body prediction in real materials, which could serve as a foundation model for computational materials science.

Figures

Figures reproduced from arXiv: 2507.05480 by the authors.

Figure 1
Figure 1. Attention mechanism and many-body correlation learning paradigm. a. Schematic of the formation of excited quasi-particles (N ± 1 problem) and excitons (N ± 2 problem), which can be described by GW and BSE formalisms, respectively. b. MBFormer: an encoder-decoder transformer architecture to learn many-body correlations and excited states. Starting from a DFT calculation, the encoder embeds the mean field information … view at source ↗
Figure 2
Figure 2. Architecture of MBFormer. a. Overview of the MBFormer architecture, including N￾stacked encoder and decoder blocks, which are connected by the cross-attention mechanism. The raw input includes mean-field states information src ={(ϕnk(r) 2 , ϵnk, n, k)} and target basis state information tgt ={(ϕ Q nk (r) 2 , ϵ Q nk , n Q, k Q)}, where ϕnk(r) is the wavefunction, ϵnk is the energy, n is the band index, and k is the c… view at source ↗
Figure 3
Figure 3. MBFormer applicaiton: extrapolation of G0W0 energies for general 2D materials. a. Parity plot comparing the exact G0W0 correction to the machine learning (ML) predicted ones. The histogram reflects the distribution of calculated and ML-predicted GW corrections. The orange(blue) points represent the training(test) dataset. 80%(20%) of total 721 2D materials are used for training(test). To better visualize the trainin… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: MBFormer applicaiton: extrapolation of excitonic properties for general 2D ma￾terials. a. Parity plot comparing the exact calculated exciton energies to the machine learning (ML) predicted exciton energies. The orange(blue) points represent the training(test) dataset. …
Figure 5
Figure 5. Figure 5: Inference of excitonic properties from coarse to fine k-grid sampling. a. An overview of calculated augmented dataset (900 datapoints) of monolayer hBN used to train the MBFormer model. Each individual point stands for a BSE Hamiltonian constructed from a specific k-gr…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 44 canonical work pages

  1. [22]

    that do not incorporate generative modeling. D. Application to two-particles BSE prediction: universal materials inference Next, we move to excitonic property prediction, which can be more complicated due to the much larger Hilbert space. Exciton eigenstates and eigenvalues are obtained by diagonalizing an effective two-particle Hamiltonian including both...

    work page

  2. [1]

    Machine learning interatomic potentials as emerging tools for materials science,

    Volker L Deringer, Miguel A Caro, and Gábor Csányi, “Machine learning interatomic potentials as emerging tools for materials science,” Advanced Materials31, 1902765 (2019)

    work page 2019

  3. [2]

    Uma: A family of universal models for atoms,

    Brandon M Wood, Misko Dzamba, Xiang Fu, Meng Gao, Muhammed Shuaibi, Luis Barroso-Luque, Kareem Abdelmaqsoud, Vahe Gharakhanyan, John R Kitchin, Daniel S Levine,et al. , “Uma: A family of universal models for atoms,” arXiv preprint arXiv:2506.23971 (2025)

    arXiv 2025

  4. [3]

    Deep-learning density functional perturbation theory,

    He Li, Zechen Tang, Jingheng Fu, Wen-Han Dong, Nianlong Zou, Xiaoxun Gong, Wenhui Duan, and Yong Xu, “Deep-learning density functional perturbation theory,” Physical Review Letters132, 17 096401 (2024)

    work page 2024

  5. [4]

    Graph trans- former networks for accurate band structure prediction: An end-to-end approach,

    Weiyi Gong, Tao Sun, Hexin Bai, Jeng-Yuan Tsai, Haibin Ling, and Qimin Yan, “Graph trans- former networks for accurate band structure prediction: An end-to-end approach,” arXiv preprint arXiv:2411.16483 (2024)

    work page arXiv 2024

  6. [5]

    Transferable equivariant graph neural networks for the hamiltonians of molecules and solids,

    Yang Zhong, Hongyu Yu, Mao Su, Xingao Gong, and Hongjun Xiang, “Transferable equivariant graph neural networks for the hamiltonians of molecules and solids,” npj Computational Materials 9, 182 (2023)

    work page 2023

  7. [6]

    Time-reversal equivariant neural network potential and hamiltonian for magnetic materials,

    Hongyu Yu, Yang Zhong, Junyi Ji, Xingao Gong, and Hongjun Xiang, “Time-reversal equivariant neural network potential and hamiltonian for magnetic materials,” (2022)

    work page 2022

  8. [7]

    Transferable E(3) equivariant parameterization for Hamiltonian of molecules and solids

    YangZhong, HongyuYu, MaoSu, XingaoGong, andHongjunXiang,“Transferablee(3)equivariant parameterization for hamiltonian of molecules and solids,” arXiv preprint arXiv:2210.16190 (2022)

    work page Pith review arXiv 2022

Show all 46 references
  1. [8]

    A generative model for inor- ganic materials design,

    Claudio Zeni, Robert Pinsler, Daniel Zügner, Andrew Fowler, Matthew Horton, Xiang Fu, Zilong Wang, Aliaksandra Shysheya, Jonathan Crabbé, Shoko Ueda,et al., “A generative model for inor- ganic materials design,” Nature , 1–3 (2025)

    2025

  2. [9]

    Fine-tuned language models generate stable inorganic materials as text,

    Nate Gruver, Anuroop Sriram, Andrea Madotto, Andrew Gordon Wilson, C Lawrence Zitnick, and Zachary Ulissi, “Fine-tuned language models generate stable inorganic materials as text,” arXiv preprint arXiv:2402.04379 (2024)

    2024 arXiv

  3. [10]

    General frame- work for e (3)-equivariant neural network representation of density functional theory hamiltonian,

    Xiaoxun Gong, He Li, Nianlong Zou, Runzhang Xu, Wenhui Duan, and Yong Xu, “General frame- work for e (3)-equivariant neural network representation of density functional theory hamiltonian,” Nature Communications14, 2848 (2023)

    2023

  4. [11]

    Universal materials model of deep-learning density functional theory hamiltonian,

    Yuxiang Wang, Yang Li, Zechen Tang, He Li, Zilong Yuan, Honggeng Tao, Nianlong Zou, Ting Bao, Xinghao Liang, Zezhou Chen,et al., “Universal materials model of deep-learning density functional theory hamiltonian,” Science Bulletin69, 2514–2521 (2024)

    2024

  5. [12]

    Accelerating the calcu- lation of electron-phonon coupling by machine learning methods,

    Yang Zhong, Zhiguo Tao, Weibin Chu, Xingao Gong, and Hongjun Xiang, “Accelerating the calcu- lation of electron-phonon coupling by machine learning methods,” arXiv preprint arXiv:2302.00439 (2023)

    2023 arXiv

  6. [13]

    Electronic excitations: density-functional versus many-body green’s-function approaches,

    Giovanni Onida, Lucia Reining, and Angel Rubio, “Electronic excitations: density-functional versus many-body green’s-function approaches,” Reviews of modern physics74, 601 (2002)

    2002

  7. [14]

    Electron correlation in semiconductors and insulators: Band gaps and quasiparticle energies,

    Mark S Hybertsen and Steven G Louie, “Electron correlation in semiconductors and insulators: Band gaps and quasiparticle energies,” Physical Review B34, 5390 (1986). 18

    1986

  8. [16]

    The computational complexity of density functional theory,

    James Daniel Whitfield, Norbert Schuch, and Frank Verstraete, “The computational complexity of density functional theory,” Many-Electron Approaches in Physics, Chemistry and Mathematics: A Multidisciplinary View , 245–260 (2014)

    2014

  9. [17]

    Computational complexity of interacting electrons and fun- damental limitations of density functional theory,

    Norbert Schuch and Frank Verstraete, “Computational complexity of interacting electrons and fun- damental limitations of density functional theory,” Nature physics5, 732–735 (2009)

    2009

  10. [18]

    Electron-hole excitations and optical spectra from first principles,

    Michael Rohlfing and Steven G. Louie, “Electron-hole excitations and optical spectra from first principles,” Phys. Rev. B62, 4927–4944 (2000)

    2000

  11. [19]

    Coupled-cluster theory in quantum chemistry,

    Rodney J Bartlett and Monika Musiał, “Coupled-cluster theory in quantum chemistry,” Reviews of Modern Physics79, 291–352 (2007)

    2007

  12. [20]

    Nearsightedness of electronic matter,

    Emil Prodan and Walter Kohn, “Nearsightedness of electronic matter,” Proceedings of the National Academy of Sciences102, 11635–11638 (2005)

    2005

  13. [21]

    Deep-learning density functional theory hamiltonian for efficient ab initio electronic-structure calculation,

    He Li, Zun Wang, Nianlong Zou, Meng Ye, Runzhang Xu, Xiaoxun Gong, Wenhui Duan, and Yong Xu, “Deep-learning density functional theory hamiltonian for efficient ab initio electronic-structure calculation,” Nature Computational Science2, 367–377 (2022)

    2022

  14. [23]

    Representing individual electronic states for machine learning gw band structures of 2d materials,

    Nikolaj Rørbæk Knøsgaard and Kristian Sommer Thygesen, “Representing individual electronic states for machine learning gw band structures of 2d materials,” Nature Communications13, 468 (2022)

    2022

  15. [24]

    Spectral operator representations,

    Austin Zadoks, Antimo Marrazzo, and Nicola Marzari, “Spectral operator representations,” arXiv preprint arXiv:2403.01514 (2024)

    2024 arXiv

  16. [25]

    Unified deep learning framework for many-body quantum chemistry via green’s func- tions,

    Christian Venturella, Jiachen Li, Christopher Hillenbrand, Ximena Leyva Peralta, Jessica Liu, and Tianyu Zhu, “Unified deep learning framework for many-body quantum chemistry via green’s func- tions,” arXiv preprint arXiv:2407.20384 (2024)

    2024 arXiv

  17. [26]

    Data- driven low-rank approximation for electron-hole kernel and acceleration of time-dependent gw cal- culations,

    Bowen Hou, Jinyuan Wu, Victor Chang Lee, Jiaxuan Guo, Luna Y Liu, and Diana Y Qiu, “Data- driven low-rank approximation for electron-hole kernel and acceleration of time-dependent gw cal- culations,” arXiv preprint arXiv:2502.05635 (2025)

    2025 arXiv

  18. [27]

    Attention is all you need,

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin, “Attention is all you need,” Advances in neural information processing systems 30 (2017). 19

    2017

  19. [28]

    Recent progress of the computational 2d materials database (c2db),

    Morten Niklas Gjerding, Alireza Taghizadeh, Asbjørn Rasmussen, Sajid Ali, Fabian Bertoldo, Thorsten Deilmann, Nikolaj Rørbæk Knøsgaard, Mads Kruse, Ask Hjorth Larsen, Simone Manti, et al., “Recent progress of the computational 2d materials database (c2db),” 2D Materials8, 0440...

    2021

  20. [29]

    The computational 2d materials database: high-throughput modeling and discovery of atomically thin crystals,

    Sten Haastrup, Mikkel Strange, Mohnish Pandey, Thorsten Deilmann, Per S Schmidt, Nicki F Hinsche, Morten N Gjerding, Daniele Torelli, Peter M Larsen, Anders C Riis-Jensen,et al., “The computational 2d materials database: high-throughput modeling and discovery of atomically thi...

    2018

  21. [30]

    Electron–electron and electron-hole interactions in small semiconductor crystallites: The size dependence of the lowest excited electronic state,

    Louis E Brus, “Electron–electron and electron-hole interactions in small semiconductor crystallites: The size dependence of the lowest excited electronic state,” The Journal of chemical physics80, 4403–4409 (1984)

    1984

  22. [31]

    Optical absorption of insulators and the electron-hole interaction: An ab initio calculation,

    Lorin X Benedict, Eric L Shirley, and Robert B Bohn, “Optical absorption of insulators and the electron-hole interaction: An ab initio calculation,” Physical review letters80, 4514 (1998)

    1998

  23. [32]

    Electron-hole excitations in semiconductors and insulators,

    Michael Rohlfing and Steven G. Louie, “Electron-hole excitations in semiconductors and insulators,” Phys. Rev. Lett.81, 2312–2315 (1998)

    1998

  24. [33]

    Hot charge-transfer exci- tons set the time limit for charge separation at donor/acceptor interfaces in organic photovoltaics,

    Askat E Jailaubekov, Adam P Willard, John R Tritsch, Wai-Lun Chan, Na Sai, Raluca Gearba, Loren G Kaake, Kenrick J Williams, Kevin Leung, Peter J Rossky,et al., “Hot charge-transfer exci- tons set the time limit for charge separation at donor/acceptor interfaces in organic pho...

    2013

  25. [34]

    Role of microstructure in the electron–hole interaction of hybrid lead halide perovskites,

    Giulia Grancini, Ajay Ram Srimath Kandada, Jarvist M Frost, Alex J Barker, Michele De Bastiani, Marina Gandini, Sergio Marras, Guglielmo Lanzani, Aron Walsh, and Annamaria Petrozza, “Role of microstructure in the electron–hole interaction of hybrid lead halide perovskites,” Na...

    2015

  26. [35]

    New method for calculating the one-particle green’s function with application to the electron-gas problem,

    Lars Hedin, “New method for calculating the one-particle green’s function with application to the electron-gas problem,” Physical Review139, A796 (1965)

    1965

  27. [36]

    Effects of electron-electron and electron-phonon interactions on the one-electron states of solids,

    Lars Hedin and Stig Lundqvist, “Effects of electron-electron and electron-phonon interactions on the one-electron states of solids,” inSolid state physics , Vol. 23 (Elsevier, 1970) pp. 1–181

    1970

  28. [37]

    Quantum theory of the dielectric constant in real solids,

    Stephen L Adler, “Quantum theory of the dielectric constant in real solids,” Physical Review126, 413 (1962)

    1962

  29. [38]

    Dielectric constant with local field effects included,

    Nathan Wiser, “Dielectric constant with local field effects included,” Physical Review129, 62 (1963). 20

    1963

  30. [39]

    General e (2)-equivariant steerable cnns,

    Maurice Weiler and Gabriele Cesa, “General e (2)-equivariant steerable cnns,” Advances in neural information processing systems32 (2019)

    2019

  31. [40]

    Unsupervised representation learning of kohn–sham states and consequences for downstream predictions of many-body effects,

    Bowen Hou, Jinyuan Wu, and Diana Y Qiu, “Unsupervised representation learning of kohn–sham states and consequences for downstream predictions of many-body effects,” Nature Communications 15, 9481 (2024)

    2024

  32. [41]

    General E(2)-Equivariant Steerable CNNs,

    Maurice Weiler and Gabriele Cesa, “General E(2)-Equivariant Steerable CNNs,” inConference on Neural Information Processing Systems (NeurIPS) (2019)

    2019

  33. [42]

    Atomic positional embedding-based transformer model for predicting the density of states of crys- talline materials,

    Yaning Cui, Kang Chen, Lingyao Zhang, Haotian Wang, Lei Bai, David Elliston, and Wei Ren, “Atomic positional embedding-based transformer model for predicting the density of states of crys- talline materials,” The Journal of Physical Chemistry Letters14, 7924–7930 (2023)

    2023

  34. [43]

    Optical spectrum of mos 2: many-body effects and diversity of exciton states,

    Diana Y Qiu, Felipe H Da Jornada, and Steven G Louie, “Optical spectrum of mos 2: many-body effects and diversity of exciton states,” Physical review letters111, 216805 (2013)

    2013

  35. [44]

    Optical spectrum ofmos2: Many-body effects and diversity of exciton states,

    Diana Y. Qiu, Felipe H. da Jornada, and Steven G. Louie, “Optical spectrum ofmos2: Many-body effects and diversity of exciton states,” Phys. Rev. Lett.111, 216805 (2013)

    2013

  36. [45]

    Berkeleygw: A massively parallel computer package for the calculation of the quasiparticle and optical properties of materials and nanostructures,

    JackDeslippe, GeorgySamsonidze, DavidA.Strubbe, ManishJain, MarvinL.Cohen, andStevenG. Louie, “Berkeleygw: A massively parallel computer package for the calculation of the quasiparticle and optical properties of materials and nanostructures,” Computer Physics Communications183...

    2012

  37. [46]

    Univer- sal ensemble-embedding graph neural network for direct prediction of optical spectra from crystal structures,

    Nguyen Tuan Hung, Ryotaro Okabe, Abhijatmedhi Chotrattanapituk, and Mingda Li, “Univer- sal ensemble-embedding graph neural network for direct prediction of optical spectra from crystal structures,” Advanced Materials36, 2409175 (2024)

    2024

  38. [47]

    Screening and many-body effects in two-dimensional crystals: Monolayer mos 2,

    Diana Y Qiu, Felipe H Da Jornada, and Steven G Louie, “Screening and many-body effects in two-dimensional crystals: Monolayer mos 2,” Physical Review B93, 235435 (2016). 21

    2016

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.