Pith. sign in

REVIEW 2 major objections 4 minor 22 references

Physical exact conditions as regularizers for exchange-correlation in solids and surface chemistry

T0 review · 2 major / 4 minor · reviewed 2026-07-10 · grok-4.5

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

Pith's one-line read Physical exact conditions regularize empirical XC functionals so they predict surface reaction barriers well instead of overfitting binding energies.

desk verdict Solid short note: free-electron-gas constraints act as useful regularizers so empirical XC fits for surface binding do not wreck barriers; new MCML/VCML numbers on SBH17 make the point cleanly. read the letter →

arxiv 2607.06729 v1 pith:RBBCZ4Y2 submitted 2026-07-07 cond-mat.mtrl-sci physics.chem-ph

Johannes Voss This is my paper · ORCID
classification cond-mat.mtrl-sciphysics.chem-ph
keywords densityfunctionaltheoryexchange-correlationsurfacesciencecatalysismeta-GGAdissociativechemisorptionregularization
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

This paper shows that when exchange-correlation functionals used in density functional theory are fitted to surface binding energies without physics constraints, they can become worse at predicting surface reaction barriers than simpler general-purpose functionals. By contrast, functionals that are empirically trained on similar data but are also forced to obey exact free-electron-gas limits (the local-density approximation limit and the correct exchange gradient expansion) recover competitive barrier heights. The author demonstrates this on a standard set of 17 dissociative chemisorption barriers on transition-metal surfaces, where the constrained meta-GGAs match the best existing functionals while purely data-driven ones rank among the worst. The practical message is that those exact conditions act as regularizers that prevent overfitting to thermodynamic ground states and keep the functional useful for the kinetic quantities that matter in catalysis.

What carries the argument

Physical exact conditions (LDA limit and exchange gradient expansion) used as regularizers during empirical training of meta-GGA functionals such as MCML and VCML-rVV10.

What would settle it

Re-optimize every transition-state geometry with each functional and recompute the SBH17 barriers; if the ranking of constrained versus unconstrained functionals reverses or the MAE gap disappears, the central claim fails.

Watch

Extended reading notes

Core claim

Empirical exchange-correlation functionals that are optimized for surface binding energies yet still satisfy the local-density approximation limit and the second-order exchange gradient expansion of the free-electron gas yield DFT-optimal average errors for surface barrier heights; purely empirical functionals that omit these constraints overfit ground-state binding and produce significantly larger barrier errors.

Load-bearing premise

The transition-state geometries taken from an earlier protocol and never re-relaxed with the functional under test remain accurate enough that geometry errors do not dominate the reported energy differences.

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

2 major / 4 minor

Summary. The manuscript argues that empirical optimization of exchange-correlation (XC) functionals against surface binding energies can degrade accuracy for surface reaction barriers unless exact physical conditions (LDA limit and exchange gradient expansion of the homogeneous electron gas) are enforced as regularizers. Using the SBH17 set of 17 dissociative-chemisorption barriers, the constrained meta-GGAs MCML and VCML-rVV10 (previously trained on surface data by the same group) yield MAEs of 0.101 eV and 0.141 eV, ranking near PBE and better than unconstrained empirical functionals such as BEEF-vdW-DF2 and RPBE that share similar training data. The authors conclude that the free-electron-gas constraints prevent overfitting to thermodynamic ground states and thereby improve transferability to barriers.

Significance. If robust, the result supplies concrete design guidance for XC functionals used in heterogeneous catalysis: physical constraints improve transferability from adsorption energies to barriers. The work is transparent, re-uses the published SBH17 benchmark and literature MAEs, and supplies new VASP calculations for two constrained functionals under a uniform protocol. Explicit strengths are the side-by-side ranking (Fig. 2, Table 1) that isolates the effect of constraints and the clear demonstration that pure data-driven regularization is insufficient.

major comments (2)
  1. [Approach / Table 1] Approach section and Table 1: Barrier heights are obtained with fixed transition-state geometries taken from the ‘medium’ protocol of Ref. 7 (optimized once with a different functional and never re-relaxed). While the protocol is applied uniformly and is computationally efficient, functional-dependent geometry errors remain uncontrolled and could systematically shift the energy differences that underlie the MAE ranking in Fig. 2. A short sensitivity test (re-optimization of a representative subset of the 17 systems) or an explicit error estimate would be needed to confirm that the performance ordering is driven by the XC energy evaluation rather than geometry mismatch.
  2. [Results / Fig. 2] Results section and Fig. 2: The central claim that the LDA limit and exchange gradient expansion act as essential regularizers rests on the contrast between MCML/VCML-rVV10 (constrained, good barriers) and BEEF-vdW-DF2 (similar training data, unconstrained, poor barriers). Other differences in functional form (meta-GGA versus GGA+vdW, presence of non-local correlation kernels) are not isolated. A brief quantification or discussion of how much of the MAE gap is attributable specifically to the two free-electron-gas constraints versus these other design choices would make the regularization interpretation more definitive.
minor comments (4)
  1. [Approach] Approach: typographical error ‘withk-point spacings’ (missing space).
  2. [Table 1 / Results] Table 1 and Results: inconsistent spacing around minus signs (e.g., ‘0.006−0.002’) and occasional space in ‘N 2’.
  3. [Fig. 2] Figure 2 caption is dense; a short legend or color coding distinguishing constrained versus unconstrained functionals would improve readability.
  4. [References] Reference [11] contains a raw URL that should be cleaned or replaced by a DOI.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor self-citation of author-group constrained functionals (MCML/VCML-rVV10); barrier MAEs are independent new computations vs external SBH17, with no reduction by construction.

  1. self citation load bearing [Approach §2 and Results §3 (Table 1, Fig. 2, and surrounding text)]
    "As XC functionals, we employ the surface chemistry-optimized and constrained meta-GGAs MCML 10 and VCML-rVV10. 11 … The MCML and VCML-rVV10 functionals with similar training data sets as BEEF-vdW-DF2 do not lead to as low chemisorption energy errors for late transition-metals,11 but the incorporation of physics constraints appears here to avoid an overfit to surface binding energies and lead to functionals with competitive accuracy for surface barriers despite not being explicitly optimized for them."

    The two functionals whose good barrier performance is used to illustrate the regularizing power of exact conditions were constructed and fitted by the same author (and collaborators) in the cited prior papers. While the barrier numbers themselves are newly computed against an external benchmark, the paper’s contrast between “constrained empirical” and “purely empirical” therefore rests on self-developed tools rather than independently published constrained functionals.

full rationale

The paper's central claim—that LDA-limit and exchange-gradient-expansion constraints regularize empirical XC training and improve surface barriers relative to unconstrained fits—is supported by fresh VASP calculations of the 17 SBH17 barriers (Table 1) and the resulting MAE ranking (Fig. 2). SBH17 references and all other functional MAEs are taken from the external Ref. 7; the barriers themselves were never part of the training sets of MCML or VCML-rVV10. The only self-referential element is the choice of those two functionals (developed by the same author/group in Refs. 10–11) as the exemplars of “constrained empirical” XC. That citation is normal sequential work and does not force the reported MAEs or the ranking by construction. No equation equates a fitted parameter to a predicted barrier, no uniqueness theorem is imported, and no ansatz is smuggled. The fixed-geometry protocol of Ref. 7 is a methodological assumption, not a circularity. Hence only a low score for the non-load-bearing self-citation of the test functionals.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central claim rests on standard DFT machinery, the homogeneous-electron-gas exact conditions already codified in the literature, and the numerical protocol of Ref. 7. No new free parameters are fitted in this work; the only ad-hoc modeling choice is the reuse of fixed transition-state geometries.

assumptions (3)
  • domain assumption The local-density approximation and second-order exchange gradient expansion of the homogeneous electron gas are exact limits that any transferable XC functional should recover.
    Invoked throughout the Introduction and Summary as the physical regularizers whose presence or absence explains the barrier performance difference.
  • ad hoc to paper Barrier heights computed with fixed transition-state geometries taken from a single reference functional (the ‘medium’ protocol of Ref. 7) are sufficiently accurate for ranking XC functionals.
    Stated in the Approach section; geometries are never re-optimized with MCML or VCML-rVV10.
  • domain assumption Mean absolute error on the 17-system SBH17 set is a representative metric of surface-barrier accuracy for late-transition-metal catalysis.
    Used as the sole quantitative figure of merit in Results and Figure 2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physical exact conditions as regularizers for exchange-correlation in solids and surface chemistry." pith.science (2026). https://pith.science/paper/RBBCZ4Y2

@misc{pith2026260706729,
  author       = {Pith},
  title        = {Pith review of: Physical exact conditions as regularizers for exchange-correlation in solids and surface chemistry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RBBCZ4Y2}},
  note         = {Machine review of arXiv:2607.06729}
}
read the original abstract

Density functional theory (DFT) often is the method of choice for simulating the electronic properties of extended solids and surfaces from first principles due to a favorable compromise between accuracy and computational cost. In the field of heterogeneous catalysis, DFT is indispensable for deriving mechanistic insights and understanding trends in surface reactivity. The accuracy of DFT for surface reaction energetics depends strongly on the exchange-correlation (XC) approximation. We show here that optimization of such XC functionals for surface binding energies can lead to a worse description of surface reaction barriers, unless important physical exact conditions are fulfilled.

Figures

Figures reproduced from arXiv: 2607.06729 by the authors.

Figure 1
Figure 1. Depiction of slab geometries used for surface barrier height com [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Different XC functionals with and without non-local correlation [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [1]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett.77, 3865 (1996)

    work page 1996

  2. [2]

    J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria, 6 L. A. Constantin, X. Zhou, and K. Burke, Phys. Rev. Lett.100, 136406 (2008)

    work page 2008

  3. [3]

    J. Tao, J. P. Perdew, V. N. Staroverov, and G. E. Scuseria, Phys. Rev. Lett.91, 146401 (2003)

    work page 2003

  4. [4]

    J. Sun, A. Ruzsinszky, and J. P. Perdew, Phys. Rev. Lett.115, 036402 (2015)

    work page 2015

  5. [5]

    Wellendorff, K

    J. Wellendorff, K. T. Lundgaard, A. Møgelhøj, V. Petzold, D. D. Landis, J. K. Nørskov, T. Bligaard, and K. W. Jacobsen, Phys. Rev. B85, 235149 (2012)

    work page 2012

  6. [6]

    Mallikarjun Sharada, R

    S. Mallikarjun Sharada, R. K. B. Karlsson, Y. Maimaiti, J. Voss, and T. Bligaard, Phys. Rev. B100, 035439 (2019)

    work page 2019

  7. [7]

    Tchakoua, N

    T. Tchakoua, N. Gerrits, E. W. F. Smeets, and G.-J. Kroes, Journal of Chemical Theory and Computation19, 245 (2023)

    work page 2023

  8. [8]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Phys. Rev. B54, 11169 (1996)

    work page 1996

Show all 22 references
  1. [9]

    Kresse and D

    G. Kresse and D. Joubert, Phys. Rev. B59, 1758 (1999)

    1999

  2. [10]

    Brown, Y

    K. Brown, Y. Maimaiti, K. Trepte, T. Bligaard, and J. Voss, J. Comput. Chem.42, 2004 (2021)

    2004

  3. [11]

    Trepte and J

    K. Trepte and J. Voss, J. Comput. Chem.43, 1104 (2022), https://onlinelibrary.wiley.com/doi/pdf/10.1002/jcc.26872

    2022 doi

  4. [12]

    M. Dion, H. Rydberg, E. Schr¨ oder, D. C. Langreth, and B. I. Lundqvist, Phys. Rev. Lett.92, 246401 (2004)

    2004

  5. [13]

    K. Lee, E. D. Murray, L. Kong, B. I. Lundqvist, and D. C. Langreth, Phys. Rev. B82, 081101(R) (2010)

    2010

  6. [14]

    Sabatini, T

    R. Sabatini, T. Gorni, and S. de Gironcoli, Phys. Rev. B87, 041108(R) (2013)

    2013

  7. [15]

    O. A. Vydrov and T. Van Voorhis, J. Chem. Phys.133, 244103 (2010)

    2010

  8. [16]

    J. Sun, R. Haunschild, B. Xiao, I. W. Bulik, G. E. Scuseria, and J. P. Perdew, J. Chem. Phys.138, 044113 (2013). 7

    2013

  9. [17]

    G. K. H. Madsen, Phys. Rev. B75, 195108 (2007)

    2007

  10. [18]

    D´ ıaz, E

    C. D´ ıaz, E. Pijper, R. A. Olsen, H. F. Busnengo, D. J. Auerbach, and G. J. Kroes, Science326, 832 (2009)

    2009

  11. [19]

    Klimeˇ s, D

    J. Klimeˇ s, D. R. Bowler, and A. Michaelides, J. Phys.: Condens. Matter 22, 022201 (2009)

    2009

  12. [20]

    J. P. Perdew, A. Ruzsinszky, G. I. Csonka, L. A. Constantin, and J. Sun, Phys. Rev. Lett.103, 026403 (2009)

    2009

  13. [21]

    E. W. F. Smeets, J. Voss, and G.-J. Kroes, J. Phys. Chem. A123, 5395 (2019)

    2019

  14. [22]

    Hammer, L

    B. Hammer, L. B. Hansen, and J. K. Nørskov, Phys. Rev. B59, 7413 (1999). 8

    1999

Pith tools

Reviewed July 10, 2026 · model on record in the stance chip above.