Pith. sign in

REVIEW 1 major objections 5 minor 28 references

Materials Database from All-electron Hybrid Functional DFT Calculations

T0 review · 1 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper constructs a database of 7,024 inorganic materials with all-electron hybrid-functional (HSE06) total energies and uses it to assess thermodynamic and electrochemical stability and to train interpretable AI models for band gaps.

desk verdict A genuinely useful all-electron HSE06 database with stability metrics; the PBEsol-geometry single-point approximation is a real but acknowledged limitation that deserves scrutiny, not rejection. read the letter →

arxiv 2504.20812 v1 pith:AUPREF26 submitted 2025-04-29 cond-mat.mtrl-sci

Akhil S. Nair , Lucas Foppa , Matthias Scheffler This is my paper · ORCID
classification cond-mat.mtrl-sci PACS 71.15.Mb71.20.-b
keywords materialsdatabasehybridfunctionalHSE06all-electronDFTthermodynamicstabilityPourbaixdiagrambandgappredictionSISSO
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

Most materials databases rely on GGA functionals, which misjudge localized electronic states in transition-metal oxides and underestimate band gaps. This paper presents an open database of 7,024 inorganic materials computed with the all-electron, range-separated hybrid functional HSE06, alongside the PBEsol results used for geometries. For each material the database reports total energies, band gaps, formation energies, decomposition energies from convex hull phase diagrams, and Pourbaix decomposition energies at oxygen-evolution-reaction conditions (pH 0, U = 1.23 V). These data show systematic differences from GGA: HSE06 lowers formation energies, raises band gaps (experimental band-gap error drops from 1.35 eV to 0.62 eV), and flips some stability classifications, e.g. AgRhO2 is acid-stable at HSE06 but not with PBEsol. The authors also train a SISSO model on the database that predicts HSE06 band gaps, with 90% of errors within 7.5% of the dataset's band-gap spread.

What carries the argument

The central object is the database itself, produced by a two-stage workflow: geometry optimization with the PBEsol functional, chosen for accurate lattice constants, followed by single-point HSE06 energy and electronic-structure calculations, both performed all-electron with numerically atom-centered basis functions. HSE06 is a range-separated hybrid functional that mixes a fraction of exact exchange with semilocal exchange, correcting the band-gap underestimation of GGA functionals. Stability analysis is carried out by constructing convex hull phase diagrams and Pourbaix diagrams; the load-bearing quantities are the decomposition energies $\Delta H_d$ and $\Delta G^{\mathrm{OER}}_{\mathrm{pbx}}$, which turn total energies into statements about thermodynamic and acid electrochemical stability. The AI demonstration uses SISSO, a compressed-sensing symbolic-regression method, to find a compact descriptor for the HSE06 band gap.

What would settle it

Take a sample of roughly 50 oxides from the database, fully relax them with HSE06, and recompute their convex-hull and Pourbaix decomposition energies; if many materials change stable/unstable classification or decomposition energies shift by more than about 0.1 eV/atom, the single-point-on-PBEsol assumption is the weak link. An experimental check would compare the database's acid-stability predictions for the reported oxide set against measured dissolution or corrosion behavior at pH 0 and 1.23 V.

Watch

Extended reading notes

Core claim

The paper sets out to establish that an openly accessible, all-electron hybrid-functional database can serve as a reliable basis for stability analysis and for training AI models for materials properties. The central results are a collection of 7,024 materials with HSE06 single-point energies on PBEsol-relaxed structures; quantitative stability metrics, namely the convex-hull decomposition energy $\Delta H_d$ and the Pourbaix decomposition energy $\Delta G^{\mathrm{OER}}_{\mathrm{pbx}}$, computed at both functional levels; and an interpretable SISSO descriptor that predicts HSE06 band gaps from PBEsol-computed features. The reported comparisons show a mean absolute deviation of 0.15 eV/atom between PBEsol and HSE06 formation energies, 0.77 eV for band gaps, and 0.27 eV/atom for Pourbaix decomposition energies; under the chosen criterion, 255 materials are acid-stable with HSE06 versus 222 with PBEsol. The paper presents examples where the two functionals identify different critical decomposition reactions and even reverse stability, such as AgRhO2.

Load-bearing premise

The load-bearing premise is that PBEsol-relaxed geometries are close enough to HSE06 equilibrium geometries that single-point HSE06 energies give trustworthy relative stabilities; if the two functionals prefer different geometries for some oxide class, convex-hull and Pourbaix decomposition energies could shift and stability verdicts could flip.

Editorial extensions

If this is right

  • AI models trained on this database inherit hybrid-functional accuracy for band gaps and stability, avoiding the systematic GGA errors that affect transition-metal oxides.
  • The database provides a concrete shortlist of acid-stable oxides for oxygen-evolution electrocatalysis: 255 materials satisfy the $\Delta G^{\mathrm{OER}}_{\mathrm{pbx}} \le 0.1$ eV/atom criterion at HSE06.
  • Functional choice matters for stability screening: PBEsol and HSE06 decomposition energies differ with a mean absolute deviation of 0.27 eV/atom for Pourbaix diagrams and can reverse individual stability classifications.
  • Because each material carries both PBEsol and HSE06 values, users can estimate functional-induced uncertainty in any property derived from the database.
  • The SISSO demonstration shows that HSE06 band gaps can be estimated from inexpensive PBEsol and composition features, with 90% of test errors within 7.5% of the band-gap spread in the dataset.

Reading between the lines

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

  • Extending the workflow to full HSE06 geometry relaxation for a subset would quantify how much of the reported stability data depends on the PBEsol-geometry assumption; defect and band-edge properties are the most likely to be affected.
  • The Pourbaix analysis covers only pH 0 and U = 1.23 V; computing the full pH-potential grid for the acid-stable subset could reveal passivation or corrosion regimes that the current OER-focused criterion misses.
  • The SISSO descriptor includes the PBEsol band gap as an input feature, so its applicability to unexplored chemistries without a cheap band-gap estimate is an open question; retraining without that feature would test how much predictive power remains.
  • The same pipeline could be run with other reference structure sets or exchange-correlation functionals to build an ensemble of databases, giving a quantitative handle on how stability predictions depend on the initial structure filter.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 5 minor

Summary. This paper presents a database of 7,024 inorganic materials with all-electron hybrid functional (HSE06) total energies, band gaps, formation energies, and stability metrics, computed on PBEsol-relaxed geometries using FHI-aims. The authors construct convex hull and Pourbaix diagrams for stability analysis, and demonstrate a SISSO-based machine-learning model for HSE06 band gaps. The database is made publicly available through NOMAD and figshare.

Significance. The database is a valuable open resource that extends hybrid-functional accuracy to a broad chemical space using an all-electron approach, with publicly accessible data repos and a transparent workflow. The band-gap benchmarking against experiment (MAE 0.62 eV for HSE06 vs 1.35 eV for PBEsol on 121 materials) provides concrete evidence of improved electronic property prediction. The interpretable SISSO model with nested cross-validation is a useful illustration of how the database can support AI models. If the geometry-sensitivity concern is addressed, the database and its stability metrics would be a strong community resource.

major comments (1)
  1. [Section IV, convex hull and Pourbaix analysis] The central stability claims rely on HSE06 single-point energies evaluated on PBEsol-relaxed geometries. The paper's own examples show that differences between PBEsol and HSE06 at fixed geometry can flip stability classifications (e.g., Li2Al is stable with PBEsol but unstable by 4 meV/atom with HSE06; Co(PtO3)2 is unstable by 11 meV/atom with PBEsol but stable with HSE06). However, these examples do not quantify the effect of geometry relaxation, and the argument based on lattice-constant accuracy does not directly address relative energy shifts. Since the acid-stability criterion is 0.1 eV/atom and the PBEsol–HSE06 MAD for Pourbaix decomposition energies is 0.27 eV/atom, geometry-induced shifts of tens of meV/atom could change stability assignments. The authors should perform HSE06 relaxation for a representative subset of near-hull and Pourbaix-relevant materials and report how many stability classifications change. Without this test, the stability labels are HSE06 energetics projected onto PBEsol geometries, and the magnitude of the projection error is unknown.
minor comments (5)
  1. [Section II] Typographical errors: "he data" should be "the data", and "Crystral" should be "Crystal" in "Inorganic Crystral Structure Database".
  2. [Fig. 7 caption] The word "Distriution" should be "Distribution".
  3. [Section IV] The statement "A MAD of 0.27 eV/atom is observed between the two methods" for Pourbaix decomposition energies would benefit from explicit clarification that this is the mean absolute deviation across all materials, as it can be misread as a per-reaction value.
  4. [Equation (1) and descriptor definitions] The equation uses the symbol \langle NVAC\rangle, but the text only defines \langle NVAL\rangle as the number of valence orbitals; please define NVAC and clarify the notation.
  5. [Section IV, SISSO model] The SISSO model uses the PBEsol band gap as a feature to predict the HSE06 band gap; stating explicitly that this is a delta-learning or correction approach would improve clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the HSE06 calculations, stability evaluations, and SISSO band-gap model are self-contained or externally benchmarked; self-citations are not load-bearing.

full rationale

The paper's derivation chain is self-contained. Database construction is a direct computational workflow: structures are taken from ICSD/MP, optimized with PBEsol, and then evaluated with HSE06 single-point calculations; the HSE06 energies are first-principles outputs, not re-expressions of any input assumption. The technical validation compares PBEsol and HSE06 formation energies and band gaps, and the HSE06 band gaps are benchmarked against experimental data for 121 binary materials (Borlido et al.), showing a reduction in MAE from 1.35 eV to 0.62 eV. The stability analyses use convex-hull phase diagrams and Pourbaix diagrams constructed from the computed energies; the representative oxides RuO2 and AgRhO2 were identified in a prior same-group study, but the diagrams and decomposition energies shown here are recomputed with both PBEsol and HSE06 and are not assumed from that citation. The SISSO model for HSE06 band gaps is a fitted regression model, explicitly including Egap,PBEsol as one feature among several composition-weighted elemental properties; it is evaluated by nested cross-validation, so its reported accuracy is a generalization estimate rather than a refit of the training set. No equation in the paper reduces to its own input, and no fitted parameter is renamed as a prediction. The PBEsol-geometry assumption for HSE06 single points is a methodological choice supported by prior lattice-constant studies; the paper explicitly discloses in its limitations that HSE06-relaxed geometries may be needed for defects and band-edge alignments, so the assumption is not hidden or circular. The self-citations (refs. 7, 15, 21) support efficiency claims, prior lattice-constant benchmarks, and the selection of example oxides, but they do not define the database's target quantities. Therefore no circular step is identified, and the paper's central claims stand on independent computation and external validation.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central results rest on standard DFT methodology, the domain assumption that PBEsol geometries suffice for HSE06 energetics, and external chemical potential data for Pourbaix diagrams. The only explicit free parameters are the SISSO model coefficients, which are fitted to the database. No new physical entities are introduced.

free parameters (1)
  • SISSO model coefficients = c0=2.45 eV, a0=-0.07 ÅeV, a1=-0.08 ÅeV^2, a2=-1.05
    These coefficients are fitted to the HSE06 band gaps of 3,091 materials in the database using nested cross-validation; they are the free parameters of the demonstration AI model.
assumptions (4)
  • domain assumption HSE06 single-point energies on PBEsol-optimized geometries accurately represent equilibrium energetics
    Section II states geometry optimizations use PBEsol and HSE06 is a single-point evaluation; the paper justifies this by citing previous studies showing small HSE06 lattice constant corrections.
  • domain assumption The chemical potentials of aqueous ions from the ASE database are valid for Pourbaix diagram construction
    Section IV figure caption states ion chemical potentials are retrieved from the ASE database; errors in these would affect electrochemical stability predictions.
  • domain assumption Materials Project GGA/GGA+U energies are suitable for selecting the lowest-energy polymorph from ICSD entries
    Section II uses MP-id matching with lowest energy/atom to filter ICSD duplicates; this assumes MP energies correctly rank polymorphs.
  • ad hoc to paper The 0.1 eV/atom threshold defines acid stability
    Section IV uses this criterion for counting acid-stable materials; it is a chosen threshold, not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Materials Database from All-electron Hybrid Functional DFT Calculations." pith.science (2026). https://pith.science/paper/AUPREF26

@misc{pith2026250420812,
  author       = {Pith},
  title        = {Pith review of: Materials Database from All-electron Hybrid Functional DFT Calculations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AUPREF26}},
  note         = {Machine review of arXiv:2504.20812}
}
read the original abstract

Materials databases built from calculations based on density functional approximations play an important role in the discovery of materials with improved properties. Most databases thus constructed rely on the generalized gradient approximation (GGA) for electron exchange and correlation. This limits the reliability of these databases, as well as the artificial intelligence (AI) models trained on them, for certain classes of materials and properties which are not well described by GGA. In this paper, we describe a database of 7,024 inorganic materials presenting diverse structures and compositions generated using hybrid functional calculations enabled by their efficient implementation in the all-electron code FHI-aims. The database is used to evaluate the thermodynamic and electrochemical stability of oxides relevant to catalysis and energy related applications. We illustrate how the database can be used to train AI models for material properties using the sure-independence screening and sparsifying operator (SISSO) approach.

Figures

Figures reproduced from arXiv: 2504.20812 by the authors.

Figure 1
Figure 1. FIG. 1: Computational workflow employed for curating the database. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: (a) Periodic table style heatmap showing the frequency of elements constituting the materials present in the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3: Comparison of (a) formation energies and (b) band gaps between PBEsol and HSE06 for the materials in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: FIG. 4: Convex hull phase diagram of Li-Al (upper panel) and Co-Pt-O (lower panel) chemical systems calculated [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5: Comparison of PBEsol vs. HSE06 decomposition energies from convex hull phase diagrams (left) and [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6: Ru (upper panel) and Ag-Rh (lower panel) Pourbaix diagram calculated using PBEsol (left) and HSE06 [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7: Distriution of nested cross validation test errors [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 19 canonical work pages

  1. [1]

    A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, et al. , Commentary: The materials project: A materials genome approach to accelerating materials innovation, APL materials 1 (2013)

    work page 2013

  2. [2]

    Kirklin, J

    S. Kirklin, J. E. Saal, B. Meredig, A. Thompson, J. W. Doak, M. Aykol, S. R¨ uhl, and C. Wolverton, The open quantum materials database (oqmd): assessing the accu- racy of dft formation energies, npj Computational Mate- rials 1, 1 (2015)

    work page 2015

  3. [3]

    Curtarolo, W

    S. Curtarolo, W. Setyawan, G. L. Hart, M. Jahnatek, R. V. Chepulskii, R. H. Taylor, S. Wang, J. Xue, K. Yang, O. Levy, et al., Aflow: An automatic framework for high- throughput materials discovery, Computational Materi- als Science 58, 218 (2012)

    work page 2012

  4. [4]

    Batatia, P

    I. Batatia, P. Benner, Y. Chiang, A. M. Elena, D. P. Kov´ acs, J. Riebesell, X. R. Advincula, M. Asta, M. Avaylon, W. J. Baldwin, et al. , A foundation model for atomistic materials chemistry, arXiv preprint arXiv:2401.00096 (2023)

    arXiv 2023

  5. [5]

    Merchant, S

    A. Merchant, S. Batzner, S. S. Schoenholz, M. Aykol, G. Cheon, and E. D. Cubuk, Scaling deep learning for materials discovery, Nature 624, 80 (2023)

    2023

  6. [6]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient approximation made simple, Physical review let- ters 77, 3865 (1996)

    1996

  7. [7]

    Zhang, A

    G.-X. Zhang, A. M. Reilly, A. Tkatchenko, and M. Schef- fler, Performance of various density-functional approxi- mations for cohesive properties of 64 bulk solids, New Journal of Physics 20, 063020 (2018)

    work page 2018

  8. [8]

    L. Wang, T. Maxisch, and G. Ceder, Oxidation ener- gies of transition metal oxides within the gga+ u frame- work, Physical Review B—Condensed Matter and Mate- rials Physics 73, 195107 (2006)

    work page 2006

Show all 28 references
  1. [9]

    Schmidt, H.-C

    J. Schmidt, H.-C. Wang, T. F. Cerqueira, S. Botti, and M. A. Marques, A dataset of 175k stable and metastable materials calculated with the pbesol and scan functionals, Scientific Data 9, 64 (2022)

    2022

  2. [10]

    Kingsbury, A

    R. Kingsbury, A. S. Gupta, C. J. Bartel, J. M. Munro, S. Dwaraknath, M. Horton, and K. A. Persson, Perfor- mance comparison of r 2 scan and scan metagga den- sity functionals for solid materials via an automated, high-throughput computational workflow, Physical Re- view Materi...

    2022

  3. [11]

    S. Kim, M. Lee, C. Hong, Y. Yoon, H. An, D. Lee, W. Jeong, D. Yoo, Y. Kang, Y. Youn, et al. , A band- gap database for semiconducting inorganic materials cal- culated with hybrid functional, Scientific Data 7, 387 (2020)

    2020

  4. [12]

    M. Liu, A. Gopakumar, V. I. Hegde, J. He, and C. Wolverton, High-throughput hybrid-functional dft cal- culations of bandgaps and formation energies and multi- 8 fidelity learning with uncertainty quantification, Physical Review Materials 8, 043803 (2024)

    2024

  5. [13]

    Lejaeghere, G

    K. Lejaeghere, G. Bihlmayer, T. Bj¨ orkman, P. Blaha, S. Bl¨ ugel, V. Blum, D. Caliste, I. E. Castelli, S. J. Clark, A. Dal Corso, et al. , Reproducibility in density func- tional theory calculations of solids, Science 351, aad3000 (2016)

    2016

  6. [14]

    M. Mazdziarz, Uncertainty of dft calculated mechanical and structural properties of solids due to incompatibility of pseudopotentials and exchange–correlation function- als, Journal of Chemical Theory and Computation 20, 9734 (2024)

    2024

  7. [15]

    Kokott, F

    S. Kokott, F. Merz, Y. Yao, C. Carbogno, M. Rossi, V. Havu, M. Rampp, M. Scheffler, and V. Blum, Effi- cient all-electron hybrid density functionals for atomistic simulations beyond 10,000 atoms, The Journal of Chem- ical Physics 161, 024112 (2024)

    2024

  8. [16]

    Ouyang, S

    R. Ouyang, S. Curtarolo, E. Ahmetcik, M. Scheffler, and L. M. Ghiringhelli, Sisso: A compressed-sensing method for identifying the best low-dimensional descriptor in an immensity of offered candidates, Physical Review Mate- rials 2, 083802 (2018)

    2018

  9. [17]

    J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria, L. A. Constantin, X. Zhou, and K. Burke, Restoring the density-gradient expansion for exchange in solids and surfaces, Physical review letters 100, 136406 (2008)

    2008

  10. [18]

    G. I. Csonka, J. P. Perdew, A. Ruzsinszky, P. H. Philipsen, S. Leb` egue, J. Paier, O. A. Vydrov, and J. G. ´Angy´ an, Assessing the performance of recent density functionals for bulk solids, Physical Review B—Condensed Matter and Materials Physics 79, 155107 (2009)

    2009

  11. [19]

    V. L. Chevrier, S. P. Ong, R. Armiento, M. K. Chan, and G. Ceder, Hybrid density functional calculations of re- dox potentials and formation energies of transition metal compounds, Physical Review B 82, 075122 (2010)

    2010

  12. [20]

    A. J. Garza and G. E. Scuseria, Predicting band gaps with hybrid density functionals, The journal of physical chemistry letters 7, 4165 (2016)

    2016

  13. [21]

    A. S. Nair, L. Foppa, and M. Scheffler, Materials- discovery workflows guided by symbolic regression: Iden- tifying acid-stable oxides for electrocatalysis, arXiv preprint arXiv:2412.05947 (2024)

    2024 arXiv

  14. [22]

    https://gitlab.com/taskblaster/taskblaster

  15. [23]

    A. Togo, K. Shinohara, and I. Tanaka, Spglib: a software library for crystal symmetry search, Science and Technol- ogy of Advanced Materials: Methods 4, 2384822 (2024)

    2024

  16. [24]

    Borlido, T

    P. Borlido, T. Aull, A. W. Huran, F. Tran, M. A. Mar- ques, and S. Botti, Large-scale benchmark of exchange– correlation functionals for the determination of electronic band gaps of solids, Journal of chemical theory and com- putation 15, 5069 (2019)

    2019

  17. [25]

    M. K. Horton, J. H. Montoya, M. Liu, and K. A. Persson, High-throughput prediction of the ground-state collinear magnetic order of inorganic materials using density func- tional theory, npj Computational Materials 5, 64 (2019)

    2019

  18. [26]

    R. M. Nieminen, Issues in first-principles calculations for defects in semiconductors and oxides, Modelling and Sim- ulation in Materials Science and Engineering 17, 084001 (2009)

    2009

  19. [27]

    Pourbaix, Atlas of electrochemical equilibria in aque- ous solutions, NACE (1966)

    M. Pourbaix, Atlas of electrochemical equilibria in aque- ous solutions, NACE (1966)

    1966

  20. [28]

    K. A. Persson, B. Waldwick, P. Lazic, and G. Ceder, Pre- diction of solid-aqueous equilibria: Scheme to combine first-principles calculations of solids with experimental aqueous states, Physical Review B 85, 235438 (2012)

    2012

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.