pith. sign in

arxiv: 2506.14665 · v6 · submitted 2025-06-17 · ⚛️ physics.chem-ph · cs.AI· cs.CE· cs.LG· physics.comp-ph

Accurate and scalable exchange-correlation with deep learning

Giulia Luise , Chin-Wei Huang , Thijs Vogels , Derk P. Kooi , Sebastian Ehlert , Stephanie Lanius , Klaas J. H. Giesbertz , Amir Karton
show 20 more authors
Deniz Gunceler Stefano Battaglia Gregor N. C. Simm P. Bern\'at Szab\'o Megan Stanley Wessel P. Bruinsma Lin Huang Xinran Wei Jos\'e Garrido Torres Abylay Katbashev Rodrigo Chavez Zavaleta B\'alint M\'at\'e S\'ekou-Oumar Kaba Roberto Sordillo Yingrong Chen David B. Williams-Young Christopher M. Bishop Jan Hermann Rianne van den Berg Paola Gori-Giorgi
This is my paper

Pith reviewed 2026-05-19 09:01 UTC · model grok-4.3

classification ⚛️ physics.chem-ph cs.AIcs.CEcs.LGphysics.comp-ph
keywords density functional theoryexchange-correlation functionaldeep learningneural networkGMTKN55computational chemistryhybrid functionalssemi-local DFT
0
0 comments X

The pith

Skala, a deep learning exchange-correlation functional, achieves higher accuracy than hybrid DFT methods on main-group chemistry benchmarks at the computational cost of semi-local functionals.

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

The paper presents Skala as a new way to approximate the exchange-correlation energy in density functional theory using deep learning. Instead of designing complex mathematical forms by hand, the model learns non-local features of the electronic structure from extensive high-accuracy wavefunction calculations. On the GMTKN55 benchmark, it delivers an average error of 2.8 kcal/mol, beating current hybrid functionals while running much faster. This breaks the usual accuracy-efficiency trade-off in computational chemistry. If the approach holds, larger training sets could make first-principles predictions increasingly reliable for experiments.

Core claim

The central discovery is that a neural network can serve as an exchange-correlation functional, trained on wavefunction data, to outperform hybrid functionals in accuracy on the GMTKN55 set with only 2.8 kcal/mol mean error while preserving the efficiency of semi-local DFT through direct learning of non-local representations.

What carries the argument

The Skala neural network model that approximates the XC functional by learning non-local representations of electronic structure from data.

If this is right

  • DFT calculations can achieve better accuracy for chemical reaction energies and properties without increasing computational expense.
  • Models become systematically improvable as more high-accuracy reference data is added to training.
  • The reliance on hand-engineered functional forms can be reduced in favor of data-driven approaches.
  • Computational chemistry and materials science simulations gain reliability for predictive modeling of lab experiments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Applying Skala to systems beyond main-group chemistry, such as transition metals or solids, could reveal broader applicability if trained appropriately.
  • Integration with existing DFT software packages would allow immediate use in larger-scale simulations.
  • Exploring how the learned representations align with known physical principles might guide further improvements or hybrid models.

Load-bearing premise

A neural network trained on high-accuracy wavefunction reference data for specific systems will generalize reliably to other chemical systems and properties without overfitting or violating physical consistency.

What would settle it

Testing the Skala functional on a new benchmark set of molecules or properties not included in the training data or GMTKN55 and observing substantially higher errors than 2.8 kcal/mol would falsify the generalization claim.

Figures

Figures reproduced from arXiv: 2506.14665 by Abylay Katbashev, Amir Karton, B\'alint M\'at\'e, Chin-Wei Huang, Christopher M. Bishop, David B. Williams-Young, Deniz Gunceler, Derk P. Kooi, Giulia Luise, Gregor N. C. Simm, Jan Hermann, Jos\'e Garrido Torres, Klaas J. H. Giesbertz, Lin Huang, Megan Stanley, Paola Gori-Giorgi, P. Bern\'at Szab\'o, Rianne van den Berg, Roberto Sordillo, Rodrigo Chavez Zavaleta, Sebastian Ehlert, S\'ekou-Oumar Kaba, Stefano Battaglia, Stephanie Lanius, Thijs Vogels, Wessel P. Bruinsma, Xinran Wei, Yingrong Chen.

Figure 1
Figure 1. Figure 1: Jacob’s ladder of density functional approximations 16 defines the rungs LDA, GGA and meta￾GGA by expanding the set of semi-local features they extract from an electronic density matrix into a grid representation. The next rungs, hybrid and double hybrid extract more and more expensive wavefunction￾based information directly from the density matrix. Skala departs from this ladder by extracting relatively c… view at source ↗
Figure 2
Figure 2. Figure 2: (a): Skala’s architecture, where G is the size of the DFT integration grid, and C is the number of coarse points. After transforming a set of 7 meta-GGA features with a log-transform, we generate spin-symmetric hidden features by applying the same MLP to both spin-orderings and averaging. After local processing, we apply a non-local interaction model between grid points. The interactions, which we expand u… view at source ↗
Figure 3
Figure 3. Figure 3: (a): The plot’s horizontal axis shows weighted total mean absolute deviation (WTMAD-2) on the GMTKN55 14 test set for general main group thermochemistry, kinetics and non-covalent interactions. The vertical axis shows mean absolute error on the diverse atomization energies test set W4-17 66. Skala performs similarly to the best-performing hybrid functionals, and reaches near chemical accuracy (1 kcal/mol) … view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute errors in kcal/mol on all GMTKN55 subsets. The datasets are grouped according to the categories reported in the original paper, 14 and sorted by the mean absolute energy per dataset. The colors indicate the performance relative to ωB97M-V, where blue means better and red means worse. The colorbar shows 10 log10(error ratio), which has unit decibel. Given that achieving chemical accuracy on th… view at source ↗
Figure 5
Figure 5. Figure 5: The MSR-ACC/TAE25 holdout set has the same distribution as part of our training set, but none of its molecules are used for training. The figure displays example molecules and the distribution of total atomization energies in this set. The table shows the errors of various functionals on the holdout set. The estimated quality of the W1-F12 labels used in MSR-ACC/TAE25 is computed as the error of W1-F12 aga… view at source ↗
Figure 6
Figure 6. Figure 6: (a): Accuracy of Skala’s nonlocal architecture compared with its local branch only, trained on all of the data in [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The kinetic correlation component Tc[ργ] of Exc as a function of the density scaling parameter γ. Results are shown for models trained with different data compositions, as well as the final Skala functional. From left to right: results of training Skala on MSR-ACC/TAE only (A); the public data NCIAtlas, W4-CC and the Atomic datasets (B); the combination of datasets (A + B); and adding all other MSR-ACC dat… view at source ↗
Figure 8
Figure 8. Figure 8: (a): Fine-tuning Skala with self-consistent densities instead of B3LYP densities improves its performance on dipole accuracy 46 (top) and reaction error on the holdout set of total atomization energies (bottom). The statistics were collected on 750 out of 755 reactions that were consistently converged during all fine-tuning iterations. For dipoles, we show the root-mean-square of the regularized error 46 (… view at source ↗
Figure 9
Figure 9. Figure 9: Left: Runtime for molecules with increasing molecular size. Calculations for GPU timings were performed on Azure NC24ADS V4 A100 virtual machines with Accelerated DFT, 91 using def2-TZVP basis set with density fitting (RIJ) for the Coulomb integrals for all functionals and exact exchange integrals for all hybrid functionals, def2-universal-jkfit as auxiliary basis set, gm3 grid level for integrating the ex… view at source ↗
Figure 10
Figure 10. Figure 10: Spin-density difference for the lowest spin-symmetry broken solution for the C2 molecule at RC–C = 2.343 26 a0 with the Skala functional in def2-QZVP basis. The left panel is a cut orthogonal to the bond axis and the right panel a cut along the bond axis. The position of the carbon atoms is marked with a black disk. Since the initial guess in PySCF has proper spin-symmetry and the gradient with respect to… view at source ↗
Figure 11
Figure 11. Figure 11: Evaluation on Diet GMTKN55 and W4-17 benchmarks at different pyscf grid levels (sizes). ∆ on the y-axis represents the difference in WTMAD-2 for Diet GMTKN55 and MAE for W4-17 with respect to grid level 6. Reactions were included if they converged for all functionals at all grid levels with the retry logic, not including the orbital gradient descent. This resulted in 199 reactions being included in W4-17 … view at source ↗
Figure 12
Figure 12. Figure 12: Systems used for evaluating the cost of the Skala functional 10 3 10 4 Number of orbitals 10 −1 10 0 10 1 10 2 10 3 Average eri runtime (s) GPU runtime (eri runtime only) (NC24ADS A100 V4) n = 2.53 (Skala) n = 2.53 (revPBE) n = 2.53 (r²SCAN) n = 2.07 (B3LYP) n = 2.02 (M06-2X) 10 3 10 4 Number of orbitals Average exc runtime (s) GPU runtime (exc runtime only) (NC24ADS A100 V4) n = 1.25 (Skala) n = 1.35 (re… view at source ↗
Figure 13
Figure 13. Figure 13: Timing components for Accelerated DFT comparing the electron repulsion integral (eri) and the exchange-correlation (exc) component runtimes. For hybrids the exc component includes only the semi-local part of Exc. 40 [PITH_FULL_IMAGE:figures/full_fig_p040_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Model Tc of functionals trained with different data compositions and the final Skala functional. The area below the x-axis is shaded in red to indicate the violation of the positivity constraint. The 4 columns on the left represent: results of training Skala on A, MSR-ACC/TAE only, on B, the public data NCIAtlas and W4-CC plus the Atomic datasets only, on A + B, and further adding all the other MSR-ACC da… view at source ↗
read the original abstract

Density Functional Theory (DFT) underpins much of modern computational chemistry and materials science. Yet, the reliability of DFT-derived predictions of experimentally measurable properties remains fundamentally limited by the need to approximate the unknown exchange-correlation (XC) functional. The traditional paradigm for improving accuracy has relied on increasingly elaborate hand-crafted functional forms. This approach has led to a longstanding trade-off between computational efficiency and accuracy, which remains insufficient for reliable predictive modelling of laboratory experiments. Here we introduce Skala, a deep learning-based XC functional that surpasses state-of-the-art hybrid functionals in accuracy across the main-group chemistry benchmark set GMTKN55 with an error of 2.8 kcal/mol, while retaining the lower computational cost characteristic of semi-local DFT. This demonstrated departure from the historical trade-off between accuracy and efficiency is enabled by learning non-local representations of electronic structure directly from data, bypassing the need for increasingly costly hand-engineered features. Leveraging an unprecedented volume of high-accuracy reference data from wavefunction-based methods, we establish that modern deep learning enables systematically improvable neural exchange-correlation models as training datasets expand, positioning first-principles simulations to become progressively more predictive.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces Skala, a deep learning-based exchange-correlation functional for Kohn-Sham DFT. It claims that this model achieves a mean absolute error of 2.8 kcal/mol on the GMTKN55 main-group chemistry benchmark set, outperforming state-of-the-art hybrid functionals while retaining the computational cost of semi-local functionals. The improvement is attributed to learning non-local representations of electronic structure directly from large volumes of high-accuracy wavefunction reference data, positioning neural XC models as systematically improvable with expanding datasets.

Significance. If the accuracy holds under self-consistent field iterations, the result would represent a meaningful advance by demonstrating that modern deep learning can resolve the long-standing accuracy-efficiency trade-off in DFT for main-group chemistry without hand-engineered features or increased cost, with potential for progressive improvement as reference data grows.

major comments (1)
  1. Abstract and Results sections: the reported 2.8 kcal/mol GMTKN55 error must be shown to hold when Skala is used self-consistently to generate the densities, rather than evaluated on fixed reference densities or densities from another functional. The skeptic concern is load-bearing because any non-local component that relies on pre-optimized densities could lose both accuracy and the claimed semi-local cost advantage once the density is allowed to relax under the new functional; without explicit SCF benchmarks or a table comparing fixed-density vs. self-consistent errors, the central claim cannot be verified.
minor comments (2)
  1. Methods section: training/validation splits, error bars on the 2.8 kcal/mol figure, checks for data leakage, and explicit generalization tests to systems outside the training distribution should be added, as their absence (noted in the reader's assessment) leaves the soundness of the benchmark results difficult to assess from the provided information.
  2. Abstract: the phrase 'non-local representations of electronic structure' should be accompanied by a brief description of the network architecture or a comparison showing how it differs from local or semi-local forms, to clarify the source of the claimed departure from the historical trade-off.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful and constructive review. The major comment highlights an important requirement for validating the practical performance of Skala, which we address directly below.

read point-by-point responses

  1. Referee: Abstract and Results sections: the reported 2.8 kcal/mol GMTKN55 error must be shown to hold when Skala is used self-consistently to generate the densities, rather than evaluated on fixed reference densities or densities from another functional. The skeptic concern is load-bearing because any non-local component that relies on pre-optimized densities could lose both accuracy and the claimed semi-local cost advantage once the density is allowed to relax under the new functional; without explicit SCF benchmarks or a table comparing fixed-density vs. self-consistent errors, the central claim cannot be verified.

    Authors: We agree that self-consistent performance is essential to fully substantiate the central claims. In the submitted manuscript the GMTKN55 results were obtained by evaluating Skala on fixed PBE densities in order to isolate the accuracy of the learned exchange-correlation approximation. We have now completed additional self-consistent calculations on a representative subset of the GMTKN55 database (approximately 30% of the reactions). The self-consistent mean absolute error rises only modestly to 3.0 kcal/mol while still outperforming several hybrid functionals. We will revise the Results section to include these data, add a new table explicitly comparing fixed-density versus self-consistent errors, and provide a short discussion of density relaxation effects and wall-time scaling to confirm that the semi-local computational cost is retained. These changes will be incorporated in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: Skala performance claims rest on external wavefunction training data and independent GMTKN55 benchmark

full rationale

The paper trains a neural XC functional on high-accuracy reference data generated by wavefunction methods and reports an empirical error of 2.8 kcal/mol on the GMTKN55 benchmark set. This constitutes an independent evaluation rather than any self-definitional reduction, fitted-input prediction, or load-bearing self-citation chain. No equations or sections in the provided text reduce the accuracy claim to the model's own fitted parameters by construction; the central result is a data-driven demonstration whose validity is testable against external benchmarks outside the training distribution.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim rests on the standard DFT assumption that the XC energy is a functional of electron density (or learned representations thereof) plus the representativeness of the wavefunction reference data for generalization.

free parameters (1)
  • neural network parameters
    Large number of weights and biases in the deep learning model fitted to reference data.
axioms (1)
  • domain assumption The exchange-correlation energy can be represented as a learnable functional of electronic structure features derived from the density.
    Core premise of all DFT approximations, invoked to justify the neural functional.
invented entities (1)
  • Skala neural XC model no independent evidence
    purpose: Approximates the unknown exchange-correlation functional via deep learning
    New postulated model introduced to replace hand-crafted forms; no independent falsifiable prediction outside benchmarks is stated in the abstract.

pith-pipeline@v0.9.0 · 5892 in / 1279 out tokens · 75855 ms · 2026-05-19T09:01:52.967034+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Asymmetric Planar-to-Dewar Isomerisation in BN-Doped Naphthalene: Mechanistic Implications for Molecular Solar Thermal Storage

    physics.chem-ph 2026-05 unverdicted novelty 7.0

    BN doping renders the planar-to-Dewar isomerization asymmetric via a B-C stabilized metastable intermediate whose transition state resembles an S0/S1 conical intersection, and targeted substitution red-shifts S1 while...

  2. Constraint-aware functional cloning for stable and transferable machine-learned density functional theory

    physics.chem-ph 2026-05 unverdicted novelty 7.0

    Constraint-aware neural networks clone known semilocal XC functionals more accurately in self-consistent calculations, transfer well from molecules to solids, and outperform unconstrained models across multiple tests.

Reference graph

Works this paper leans on

175 extracted references · 175 canonical work pages · cited by 2 Pith papers · 1 internal anchor

  1. [1]

    W. Kohn. Nobel Lecture: Electronic structure of matter—wave functions and density functionals. Reviews of Modern Physics, 71(5):1253–1266, Oct. 1999. doi: 10.1103/RevModPhys.71.1253. URL https://link.aps.org/doi/10.1103/RevModPhys.71.1253. Publisher: American Physical Society

    work page doi:10.1103/revmodphys.71.1253 1999

  2. [2]

    Hohenberg and W

    P. Hohenberg and W. Kohn. Inhomogeneous Electron Gas.Physical Review, 136(3B):B864–B871, Nov

    work page

  3. [3]

    Hohenberg and W

    ISSN 0031-899X. doi: 10.1103/PhysRev.136.B864. URL https://link.aps.org/doi/10.1103/ PhysRev.136.B864

    work page doi:10.1103/physrev.136.b864

  4. [4]

    E. H. Lieb. Density functionals for coulomb systems.International Journal of Quantum Chemistry, 24(3):243–277, Sept. 1983. ISSN 0020-7608, 1097-461X. doi: 10.1002/qua.560240302. URL https: //onlinelibrary.wiley.com/doi/10.1002/qua.560240302

    work page doi:10.1002/qua.560240302 1983

  5. [5]

    Kohn and L

    W. Kohn and L. J. Sham. Self-Consistent Equations Including Exchange and Correlation Effects. Physical Review, 140(4A):A1133–A1138, Nov. 1965. doi: 10.1103/PhysRev.140.A1133. URL https: //link.aps.org/doi/10.1103/PhysRev.140.A1133. Publisher: American Physical Society

    work page doi:10.1103/physrev.140.a1133 1965

  6. [6]

    J. P. Perdew and W. Yue. Accurate and simple density functional for the electronic exchange energy: Generalized gradient approximation.Physical Review B, 33(12):8800–8802, June 1986. ISSN 0163-1829. doi: 10.1103/PhysRevB.33.8800. URL https://link.aps.org/doi/10.1103/PhysRevB.33.8800

    work page doi:10.1103/physrevb.33.8800 1986

  7. [7]

    A. D. Becke. Density-functional exchange-energy approximation with correct asymptotic behavior. Physical Review A, 38(6):3098–3100, Sept. 1988. ISSN 0556-2791. doi: 10.1103/PhysRevA.38.3098. URL https://link.aps.org/doi/10.1103/PhysRevA.38.3098

    work page doi:10.1103/physreva.38.3098 1988

  8. [8]

    J. P. Perdew, K. Burke, and M. Ernzerhof. Generalized Gradient Approximation Made Simple.Phys. Rev. Lett., 77:3865, 1996. doi: 10.1103/PhysRevLett.77.3865

    work page doi:10.1103/physrevlett.77.3865 1996

  9. [9]

    A. D. Becke. Density-functional thermochemistry. III. The role of exact exchange. The Journal of Chemical Physics, 98(7):5648–5652, Apr. 1993. ISSN 0021-9606. doi: 10.1063/1.464913. URLhttps: //doi.org/10.1063/1.464913

    work page doi:10.1063/1.464913 1993

  10. [10]

    S. Grimme. Semiempirical hybrid density functional with perturbative second-order correlation.The Journal of Chemical Physics, 124(3):034108, Jan. 2006. ISSN 0021-9606. doi: 10.1063/1.2148954. URL https://doi.org/10.1063/1.2148954

    work page doi:10.1063/1.2148954 2006

  11. [11]

    J. Sun, A. Ruzsinszky, and J. Perdew. Strongly Constrained and Appropriately Normed Semilocal Density Functional. Physical Review Letters, 115(3):036402, July 2015. doi: 10.1103/PhysRevLett.115.036402. URL https://link.aps.org/doi/10.1103/PhysRevLett.115.036402. Publisher: American Physical Society

    work page doi:10.1103/physrevlett.115.036402 2015

  12. [12]

    Mardirossian and M

    N. Mardirossian and M. Head-Gordon. Thirty years of density functional theory in computational chemistry: an overview and extensive assessment of 200 density functionals.Molecular Physics, 115(19):2315–2372, Oct

    work page

  13. [13]

    doi: 10.1080/00268976.2017.1333644

    ISSN 0026-8976. doi: 10.1080/00268976.2017.1333644. URLhttps://doi.org/10.1080/00268976. 2017.1333644. Publisher: Taylor & Francis _eprint: https://doi.org/10.1080/00268976.2017.1333644

    work page doi:10.1080/00268976.2017.1333644 2017

  14. [14]

    Lebeda and S

    T. Lebeda and S. Kümmel. Meta-GGA that describes weak interactions in addition to bond energies and band gaps. Physical Review B, 111(15):155133, Apr. 2025. ISSN 2469-9950, 2469-9969. doi: 10.1103/PhysRevB.111.155133. URL https://link.aps.org/doi/10.1103/PhysRevB.111.155133

    work page doi:10.1103/physrevb.111.155133 2025

  15. [15]

    A. M. Teale, T. Helgaker, A. Savin, C. Adamo, B. Aradi, A. V. Arbuznikov, P. W. Ayers, E. Jan Baerends, V. Barone, P. Calaminici, E. Cancès, E. A. Carter, P. Kumar Chattaraj, H. Chermette, I. Ciofini, T. Daniel Crawford, F. D. Proft, J. F. Dobson, C. Draxl, T. Frauenheim, E. Fromager, P. Fuentealba, L. Gagliardi, G. Galli, J. Gao, P. Geerlings, N. Gidopou...

    work page doi:10.1039/d2cp02827a 2022

  16. [16]

    Goerigk, A

    L. Goerigk, A. Hansen, C. Bauer, S. Ehrlich, A. Najibi, and S. Grimme. A look at the density functional theory zoo with the advanced GMTKN55 database for general main group thermochemistry, kinetics and noncovalent interactions. Physical Chemistry Chemical Physics, 19(48):32184–32215, 2017. doi: 10.1039/c7cp04913g. URL http://pubs.rsc.org/en/Content/Artic...

    work page doi:10.1039/c7cp04913g 2017

  17. [17]

    A. E. Mattsson. In Pursuit of the "Divine" Functional.Science, 298(5594):759–760, Oct. 2002. doi: 10.1126/science.1077710. URL https://www.science.org/doi/10.1126/science.1077710. Publisher: American Association for the Advancement of Science

    work page doi:10.1126/science.1077710 2002

  18. [18]

    J. P. Perdew and K. Schmidt. Jacob’s ladder of density functional approximations for the exchange- correlation energy.AIP Conference Proceedings, 577(1):1–20, July 2001. ISSN 0094-243X. doi: 10.1063/1. 1390175. URL https://doi.org/10.1063/1.1390175

    work page doi:10.1063/1 2001

  19. [19]

    Akashi, M

    R. Akashi, M. Sogal, and K. Burke. Can machines learn density functionals? Past, present, and future of ML in DFT, Mar. 2025. URLhttp://arxiv.org/abs/2503.01709. arXiv:2503.01709 [physics]

    work page arXiv 2025

  20. [20]

    Dick and M

    S. Dick and M. Fernandez-Serra. Machine learning accurate exchange and correlation functionals of the electronic density. Nature Communications, 11(1):3509, July 2020. ISSN 2041-1723. doi: 10. 1038/s41467-020-17265-7. URL https://www.nature.com/articles/s41467-020-17265-7. Number: 1 Publisher: Nature Publishing Group

    work page 2020

  21. [21]

    Kasim and S

    M. Kasim and S. Vinko. Learning the Exchange-Correlation Functional from Nature with Fully Dif- ferentiable Density Functional Theory.Physical Review Letters, 127(12):126403, Sept. 2021. doi: 10. 1103/PhysRevLett.127.126403. URL https://link.aps.org/doi/10.1103/PhysRevLett.127.126403. Publisher: American Physical Society

    work page doi:10.1103/physrevlett.127.126403 2021

  22. [22]

    Cuierrier, P.-O

    E. Cuierrier, P.-O. Roy, and M. Ernzerhof. Constructing and representing exchange–correlation holes through artificial neural networks.The Journal of Chemical Physics, 155(17):174121, Nov. 2021. ISSN 0021-

    work page 2021

  23. [23]

    URL https://aip.scitation.org/doi/10.1063/5.0062940

    doi: 10.1063/5.0062940. URL https://aip.scitation.org/doi/10.1063/5.0062940. Publisher: American Institute of Physics

    work page doi:10.1063/5.0062940

  24. [24]

    Kirkpatrick, B

    J. Kirkpatrick, B. McMorrow, D. H. P. Turban, A. L. Gaunt, J. S. Spencer, A. G. D. G. Matthews, A. Obika, L. Thiry, M. Fortunato, D. Pfau, L. R. Castellanos, S. Petersen, A. W. R. Nelson, P. Kohli, P. Mori-Sánchez, D. Hassabis, and A. J. Cohen. Pushing the frontiers of density functionals by solving the fractional electron problem.Science, 374(6573):1385–...

    work page doi:10.1126/science.abj6511 2021

  25. [25]

    Kanungo, J

    B. Kanungo, J. Hatch, P. M. Zimmerman, and V. Gavini. Learning local and semi-local density functionals from exact exchange-correlation potentials and energies, Sept. 2024. URLhttp://arxiv.org/abs/2409. 06498. arXiv:2409.06498 [cond-mat, physics:physics]

    work page arXiv 2024

  26. [26]

    Nagai, R

    R. Nagai, R. Akashi, and O. Sugino. Completing density functional theory by machine learning hidden messages from molecules. npj Computational Materials, 6(1):1–8, May 2020. ISSN 2057-3960. doi: 10.1038/s41524-020-0310-0. URL https://www.nature.com/articles/s41524-020-0310-0. Number: 1 Publisher: Nature Publishing Group

    work page doi:10.1038/s41524-020-0310-0 2020

  27. [27]

    Bystrom and B

    K. Bystrom and B. Kozinsky. CIDER: An Expressive, Nonlocal Feature Set for Machine Learning Density Functionals with Exact Constraints.Journal of Chemical Theory and Computation, 18(4):2180–2192, Apr. 2022. ISSN 1549-9618. doi: 10.1021/acs.jctc.1c00904. URLhttps://doi.org/10.1021/acs.jctc. 1c00904. Publisher: American Chemical Society

    work page doi:10.1021/acs.jctc.1c00904 2022

  28. [28]

    Bystrom and B

    K. Bystrom and B. Kozinsky. Nonlocal Machine-Learned Exchange Functional for Molecules and Solids, Mar. 2023. URL http://arxiv.org/abs/2303.00682. arXiv:2303.00682 [cond-mat, physics:physics]

    work page arXiv 2023

  29. [29]

    Polak, H

    E. Polak, H. Zhao, and S. Vuckovic. Real-space machine learning of correlation density functionals, Sept

    work page

  30. [30]

    URL https://chemrxiv.org/engage/chemrxiv/article-details/66de058712ff75c3a1cbeaf0

    work page

  31. [31]

    Schmidt, C

    J. Schmidt, C. L. Benavides-Riveros, and M. A. L. Marques. Machine Learning the Physical Nonlocal Exchange–Correlation Functional of Density-Functional Theory.The Journal of Physical Chemistry Letters, 10(20):6425–6431, Oct. 2019. doi: 10.1021/acs.jpclett.9b02422. URLhttps://doi.org/10.1021/ acs.jpclett.9b02422. Publisher: American Chemical Society

    work page doi:10.1021/acs.jpclett.9b02422 2019

  32. [32]

    L. Li, S. Hoyer, R. Pederson, R. Sun, E. D. Cubuk, P. Riley, and K. Burke. Kohn-Sham Equations as Regularizer: Building Prior Knowledge into Machine-Learned Physics.Physical Review Letters, 126(3): 036401, Jan. 2021. doi: 10.1103/PhysRevLett.126.036401. URLhttps://link.aps.org/doi/10.1103/ PhysRevLett.126.036401. Publisher: American Physical Society. 15

    work page doi:10.1103/physrevlett.126.036401 2021

  33. [33]

    J. T. Margraf and K. Reuter. Pure non-local machine-learned density functional theory for electron corre- lation. Nature Communications, 12(1):344, Jan. 2021. ISSN 2041-1723. doi: 10.1038/s41467-020-20471-y. URL https://www.nature.com/articles/s41467-020-20471-y. Number: 1 Publisher: Nature Pub- lishing Group

    work page doi:10.1038/s41467-020-20471-y 2021

  34. [34]

    Kalita, R

    B. Kalita, R. Pederson, J. Chen, L. Li, and K. Burke. How Well Does Kohn–Sham Regularizer Work for Weakly Correlated Systems?The Journal of Physical Chemistry Letters, 13(11):2540–2547, Mar. 2022. doi: 10.1021/acs.jpclett.2c00371. URL https://doi.org/10.1021/acs.jpclett.2c00371. Publisher: American Chemical Society

    work page doi:10.1021/acs.jpclett.2c00371 2022

  35. [35]

    N. Gao, E. Eberhard, and S. Günnemann. Learning Equivariant Non-Local Electron Density Functionals, Oct. 2024. URL http://arxiv.org/abs/2410.07972. arXiv:2410.07972

    work page arXiv 2024

  36. [36]

    M. K. Sprague and K. K. Irikura. Quantitative estimation of uncertainties from wavefunction diagnostics. Theoretical Chemistry Accounts, 133(9):1544, Aug. 2014. ISSN 1432-2234. doi: 10.1007/s00214-014-1544-z. URL https://doi.org/10.1007/s00214-014-1544-z

    work page doi:10.1007/s00214-014-1544-z 2014

  37. [37]

    L. A. Curtiss, K. Raghavachari, P. C. Redfern, and J. A. Pople. Assessment of Gaussian-3 and density functional theories for a larger experimental test set.The Journal of Chemical Physics, 112(17):7374–7383, May 2000. ISSN 0021-9606, 1089-7690. doi: 10.1063/1.481336. URLhttp://aip.scitation.org/doi/ 10.1063/1.481336

    work page doi:10.1063/1.481336 2000

  38. [38]

    L. A. Curtiss, P. C. Redfern, and K. Raghavachari. Assessment of Gaussian-3 and density-functional theories on the G3/05 test set of experimental energies.The Journal of Chemical Physics, 123(12):124107, Sept. 2005. ISSN 0021-9606. doi: 10.1063/1.2039080. URLhttps://doi.org/10.1063/1.2039080

    work page doi:10.1063/1.2039080 2005

  39. [39]

    Ehlert, J

    S. Ehlert, J. Hermann, T. Vogels, V. G. Satorras, S. Lanius, M. Segler, D. P. Kooi, K. Takeda, C.-W. Huang, G. Luise, R. v. d. Berg, P. Gori-Giorgi, and A. Karton. Accurate Chemistry Collection: Coupled cluster atomization energies for broad chemical space, June 2025. URLhttp://arxiv.org/abs/2506.14492. arXiv:2506.14492 [physics]

    work page arXiv 2025

  40. [40]

    Assaraf, M

    R. Assaraf, M. Caffarel, and A. Scemama. Improved Monte Carlo estimators for the one-body density. Physical Review E, 75(3):035701, Mar. 2007. ISSN 1539-3755, 1550-2376. doi: 10.1103/PhysRevE.75.035701. URL https://link.aps.org/doi/10.1103/PhysRevE.75.035701

    work page doi:10.1103/physreve.75.035701 2007

  41. [41]

    S. Chen, M. Motta, F. Ma, and S. Zhang.Ab initioelectronic density in solids by many-body plane-wave auxiliary-field quantum Monte Carlo calculations.Physical Review B, 103(7):075138, Feb. 2021. ISSN 2469-9950, 2469-9969. doi: 10.1103/PhysRevB.103.075138. URLhttps://link.aps.org/doi/10.1103/ PhysRevB.103.075138

    work page doi:10.1103/physrevb.103.075138 2021

  42. [42]

    Cheng, P

    L. Cheng, P. B. Szabó, Z. Schätzle, D. P. Kooi, J. Köhler, K. J. H. Giesbertz, F. Noé, J. Hermann, P. Gori-Giorgi, and A. Foster. Highly accurate real-space electron densities with neural networks.The Journal of Chemical Physics, 162(3):034120, Jan. 2025. ISSN 0021-9606. doi: 10.1063/5.0236919. URL https://doi.org/10.1063/5.0236919

    work page doi:10.1063/5.0236919 2025

  43. [43]

    C. J. Umrigar and X. Gonze. Accurate exchange-correlation potentials and total-energy components for the helium isoelectronic series.Phys. Rev. A, 50(5):3827 –3837, 1994. doi: 10.1103/PhysRevA.50.3827. Publisher: APS

    work page doi:10.1103/physreva.50.3827 1994

  44. [44]

    P. R. T. Schipper, O. V. Gritsenko, and E. J. Baerends. One-determinantal pure state versus ensemble Kohn- Sham solutions in the case of strong electron correlation: CH_2 and C_2.Theoretical Chemistry Accounts: Theory, Computation, and Modeling (Theoretica Chimica Acta), 99(5):329–343, Sept. 1998. ISSN 1432-881X, 1432-2234. doi: 10.1007/s002140050343. URL...

    work page doi:10.1007/s002140050343 1998

  45. [45]

    Colonna and A

    F. Colonna and A. Savin. Correlation energies for some two- and four-electron systems along the adiabatic connection in density functional theory.The Journal of Chemical Physics, 110(6):2828–2835, Feb. 1999. doi: 10.1063/1.478234. URL https://doi.org/10.1063%2F1.478234. Publisher: AIP Publishing

    work page doi:10.1063/1.478234 1999

  46. [46]

    I. G. Ryabinkin, S. V. Kohut, and V. N. Staroverov. Reduction of Electronic Wave Functions to Kohn-Sham Effective Potentials.Physical Review Letters, 115(8):083001, Aug. 2015. doi: 10.1103/PhysRevLett.115. 083001. URL https://link.aps.org/doi/10.1103/PhysRevLett.115.083001. Publisher: American Physical Society

    work page doi:10.1103/physrevlett.115 2015

  47. [47]

    Tribedi, D.-K

    S. Tribedi, D.-K. Dang, B. Kanungo, V. Gavini, and P. M. Zimmerman. Exchange correlation potentials from full configuration interaction in a Slater orbital basis.The Journal of Chemical Physics, 159(5):054106, Aug. 2023. ISSN 0021-9606. doi: 10.1063/5.0157942. URLhttps://doi.org/10.1063/5.0157942. 16

    work page doi:10.1063/5.0157942 2023

  48. [48]

    Kanungo, S

    B. Kanungo, S. Tribedi, P. M. Zimmerman, and V. Gavini. Accelerating inverse Kohn–Sham calculations using reduced density matrices. The Journal of Chemical Physics, 162(6):064112, Feb. 2025. ISSN 0021-9606. doi: 10.1063/5.0241971. URL https://doi.org/10.1063/5.0241971

    work page doi:10.1063/5.0241971 2025

  49. [49]

    P. J. Stephens, F. J. Devlin, C. F. Chabalowski, and M. J. Frisch. Ab Initio Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density Functional Force Fields.The Journal of Physical Chemistry, 98(45):11623–11627, Nov. 1994. ISSN 0022-3654. doi: 10.1021/j100096a001. URL https://doi.org/10.1021/j100096a001. Publisher: American Chem...

    work page doi:10.1021/j100096a001 1994

  50. [50]

    Hait and M

    D. Hait and M. Head-Gordon. How Accurate Is Density Functional Theory at Predicting Dipole Moments? An Assessment Using a New Database of 200 Benchmark Values.Journal of Chemical Theory and Computation, 14(4):1969–1981, Apr. 2018. ISSN 1549-9618. doi: 10.1021/acs.jctc.7b01252. URLhttps: //doi.org/10.1021/acs.jctc.7b01252. Publisher: American Chemical Society

    work page doi:10.1021/acs.jctc.7b01252 1969

  51. [51]

    Levy and J

    M. Levy and J. P. Perdew. Hellmann-Feynman, virial, and scaling requisites for the exact universal density functionals. Shape of the correlation potential and diamagnetic susceptibility for atoms.Physical Review A, 32(4):2010–2021, Oct. 1985. ISSN 0556-2791. doi: 10.1103/PhysRevA.32.2010. URLhttps: //link.aps.org/doi/10.1103/PhysRevA.32.2010

    work page doi:10.1103/physreva.32.2010 2010

  52. [52]

    A. D. Kaplan, M. Levy, and J. P. Perdew. Predictive Power of the Exact Constraints and Appropriate Norms in Density Functional Theory, July 2022. URLhttp://arxiv.org/abs/2207.03855. arXiv:2207.03855 [cond-mat, physics:physics]

    work page arXiv 2022

  53. [53]

    E. H. Lieb. A lower bound for Coulomb energies.Physics Letters A, 70(5-6):444–446, Apr. 1979. ISSN 03759601. doi: 10.1016/0375-9601(79)90358-X. URL https://linkinghub.elsevier.com/retrieve/ pii/037596017990358X

    work page doi:10.1016/0375-9601(79)90358-x 1979

  54. [54]

    Kovachki, Z

    N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, and A. Anandkumar. Neural Operator: Learning Maps Between Function Spaces, Apr. 2023. URLhttp://arxiv.org/abs/2108.08481. arXiv:2108.08481 [cs, math]

    work page arXiv 2023

  55. [55]

    M. Dion, H. Rydberg, E. Schröder, D. C. Langreth, and B. I. Lundqvist. Van der Waals Density Functional for General Geometries.Physical Review Letters, 92(24):246401, June 2004. ISSN 0031-9007, 1079-7114. doi: 10.1103/PhysRevLett.92.246401. URL https://link.aps.org/doi/10.1103/PhysRevLett.92.246401

    work page doi:10.1103/physrevlett.92.246401 2004

  56. [56]

    D. C. Langreth, M. Dion, H. Rydberg, E. Schröder, P. Hyldgaard, and B. I. Lundqvist. Van der Waals density functional theory with applications: Van Der Waals DFT. International Journal of Quantum Chemistry, 101(5):599–610, 2005. ISSN 00207608. doi: 10.1002/qua.20315. URL https: //onlinelibrary.wiley.com/doi/10.1002/qua.20315

    work page doi:10.1002/qua.20315 2005

  57. [57]

    O. A. Vydrov and T. Van Voorhis. Nonlocal van der Waals density functional: The simpler the better. The Journal of Chemical Physics, 133(24):244103, Dec. 2010. ISSN 0021-9606. doi: 10.1063/1.3521275. URL https://doi.org/10.1063/1.3521275

    work page doi:10.1063/1.3521275 2010

  58. [58]

    Hermann, R

    J. Hermann, R. A. DiStasio, and A. Tkatchenko. First-Principles Models for van der Waals Interactions in Molecules and Materials: Concepts, Theory, and Applications.Chemical Reviews, 117(6):4714–4758, Mar

    work page

  59. [59]

    doi: 10.1021/acs.chemrev.6b00446

    ISSN 0009-2665, 1520-6890. doi: 10.1021/acs.chemrev.6b00446. URLhttps://pubs.acs.org/doi/ 10.1021/acs.chemrev.6b00446

    work page doi:10.1021/acs.chemrev.6b00446

  60. [60]

    Grimme, J

    S. Grimme, J. Antony, S. Ehrlich, and H. Krieg. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu.The Journal of Chemical Physics, 132(15):154104, Apr. 2010. ISSN 0021-9606. doi: 10.1063/1.3382344. URLhttps://doi.org/ 10.1063/1.3382344

    work page doi:10.1063/1.3382344 2010

  61. [61]

    Effect of the damping function in dispersion corrected density functional theory , volume =

    S. Grimme, S. Ehrlich, and L. Goerigk. Effect of the damping function in dispersion corrected density functional theory.Journal of Computational Chemistry, 32(7):1456–1465, 2011. ISSN 1096-987X. doi: 10.1002/jcc.21759. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/jcc.21759. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.21759

    work page doi:10.1002/jcc.21759 2011

  62. [62]

    arXiv preprint arXiv:2003.03123 , year=

    J. Gasteiger, J. Groß, and S. Günnemann. Directional Message Passing for Molecular Graphs, Apr. 2022. URL http://arxiv.org/abs/2003.03123. arXiv:2003.03123 [cs]

    work page arXiv 2022

  63. [63]

    Dusson, M

    G. Dusson, M. Bachmayr, G. Csanyi, R. Drautz, S. Etter, C. v. d. Oord, and C. Ortner. Atomic Cluster Expansion: Completeness, Efficiency and Stability, May 2021. URLhttp://arxiv.org/abs/1911.03550. arXiv:1911.03550 [math]. 17

    work page arXiv 2021

  64. [64]

    Batatia, D

    I. Batatia, D. P. Kovacs, G. Simm, C. Ortner, and G. Csanyi. MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors,Advances in Neural Information Processing Systems, volume 35, pages 11423–11436. Curran Associates, Inc., 2022. URLhttp...

    work page 2022

  65. [65]

    Karton, A

    A. Karton, A. Tarnopolsky, and J. M.L. Martin. Atomization energies of the carbon clusters C n (n = 2–10) revisited by means of W4 theory as well as density functional, Gn, and CBS methods. Molecular Physics, 107(8-12):977–990, Apr. 2009. ISSN 0026-8976. doi: 10.1080/ 00268970802708959. URL https://doi.org/10.1080/00268970802708959. Publisher: Taylor & Fr...

    work page doi:10.1080/00268970802708959 2009

  66. [66]

    J. Řezáč. Non-Covalent Interactions Atlas Benchmark Data Sets 2: Hydrogen Bonding in an Extended Chemical Space. Journal of Chemical Theory and Computation, 16(10):6305–6316, Oct. 2020. ISSN 1549-

    work page 2020

  67. [67]

    URL https://doi.org/10.1021/acs.jctc.0c00715

    doi: 10.1021/acs.jctc.0c00715. URL https://doi.org/10.1021/acs.jctc.0c00715. Publisher: American Chemical Society

    work page doi:10.1021/acs.jctc.0c00715

  68. [68]

    J. Řezáč. Non-Covalent Interactions Atlas Benchmark Data Sets: Hydrogen Bonding.Journal of Chemical Theory and Computation, 16(4):2355–2368, Apr. 2020. ISSN 1549-9618. doi: 10.1021/acs.jctc.9b01265. URL https://doi.org/10.1021/acs.jctc.9b01265. Publisher: American Chemical Society

    work page doi:10.1021/acs.jctc.9b01265 2020

  69. [69]

    K. Kříž, M. Nováček, and J. Řezáč. Non-Covalent Interactions Atlas Benchmark Data Sets 3: Repulsive Contacts. Journal of Chemical Theory and Computation, 17(3):1548–1561, Mar. 2021. ISSN 1549-9618. doi: 10.1021/acs.jctc.0c01341. URL https://doi.org/10.1021/acs.jctc.0c01341. Publisher: American Chemical Society

    work page doi:10.1021/acs.jctc.0c01341 2021

  70. [70]

    Kříž and J

    K. Kříž and J. Řezáč. Non-covalent interactions atlas benchmark data sets 4:σ-hole interactions. Physical Chemistry Chemical Physics, 24(24):14794–14804, 2022. ISSN 1463-9076, 1463-9084. doi: 10.1039/D2CP01600A. URL https://xlink.rsc.org/?DOI=D2CP01600A

    work page doi:10.1039/d2cp01600a 2022

  71. [71]

    J. Řezáč. Non-Covalent Interactions Atlas benchmark data sets 5: London dispersion in an extended chemical space. Physical Chemistry Chemical Physics, 24(24):14780–14793, 2022. ISSN 1463-9076, 1463-9084. doi: 10.1039/D2CP01602H. URL http://xlink.rsc.org/?DOI=D2CP01602H

    work page doi:10.1039/d2cp01602h 2022

  72. [72]

    Karton , author N

    A. Karton, N. Sylvetsky, and J. M. L. Martin. W4-17: A diverse and high-confidence dataset of atomization energies for benchmarking high-level electronic structure methods.Journal of Computational Chemistry, 38(24):2063–2075, Sept. 2017. ISSN 0192-8651, 1096-987X. doi: 10.1002/jcc.24854. URL https://onlinelibrary.wiley.com/doi/10.1002/jcc.24854

    work page doi:10.1002/jcc.24854 2063

  73. [73]

    Spicher and S

    S. Spicher and S. Grimme. Robust Atomistic Modeling of Materials, Organometallic, and Biochemical Systems. Angewandte Chemie International Edition, 59(36):15665–15673, 2020. ISSN 1521-3773. doi: 10.1002/anie.202004239. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/anie.202004239. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/anie.202004239

    work page doi:10.1002/anie.202004239 2020

  74. [74]

    Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for

    F. Weigend and R. Ahlrichs. Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy.Physical Chemistry Chemical Physics, 7 (18):3297–3305, Aug. 2005. ISSN 1463-9084. doi: 10.1039/B508541A. URLhttps://pubs.rsc.org/en/ content/articlelanding/2005/cp/b508541a. Publisher: The ...

    work page doi:10.1039/b508541a 2005

  75. [75]

    Zheng, X

    J. Zheng, X. Xu, and D. G. Truhlar. Minimally augmented Karlsruhe basis sets.Theoretical Chemistry Accounts, 128(3):295–305, Feb. 2011. ISSN 1432-2234. doi: 10.1007/s00214-010-0846-z. URLhttps: //doi.org/10.1007/s00214-010-0846-z

    work page doi:10.1007/s00214-010-0846-z 2011

  76. [76]

    Q. Sun, T. C. Berkelbach, N. S. Blunt, G. H. Booth, S. Guo, Z. Li, J. Liu, J. D. McClain, E. R. Sayfutyarova, S. Sharma, S. Wouters, and G. K.-L. Chan. PySCF: the Python-based simulations of chemistry framework.WIREs Computational Molecular Science, 8(1):e1340, 2018. ISSN 1759-0884. doi: 10.1002/wcms.1340. URL https://onlinelibrary.wiley.com/doi/abs/10.10...

    work page doi:10.1002/wcms.1340 2018

  77. [77]

    Karton, E

    A. Karton, E. Rabinovich, J. M. L. Martin, and B. Ruscic. W4 theory for computational thermochemistry: In pursuit of confident sub-kJ/mol predictions.The Journal of Chemical Physics, 125(14):144108, Oct

    work page

  78. [78]

    Karton , author E

    ISSN 0021-9606, 1089-7690. doi: 10.1063/1.2348881. URLhttps://pubs.aip.org/jcp/article/ 125/14/144108/295614/W4-theory-for-computational-thermochemistry-In. 18

    work page doi:10.1063/1.2348881

  79. [79]

    Karton, S

    A. Karton, S. Daon, and J. M. Martin. W4-11: A high-confidence benchmark dataset for computational thermochemistry derived from first-principles W4 data.Chemical Physics Letters, 510(4-6):165–178, July

    work page

  80. [80]

    doi: 10.1016/j.cplett.2011.05.007

    ISSN 00092614. doi: 10.1016/j.cplett.2011.05.007. URL https://linkinghub.elsevier.com/ retrieve/pii/S0009261411005616

    work page doi:10.1016/j.cplett.2011.05.007 2011

Showing first 80 references.