Pith.

Chemical Physics · arXiv:2611.16420 v1 · submitted 12 Nov 2026 · 119 references

Compensated Grid-Based Hybrid Exchange - A Simple, Accurate, Efficient, and Differentiable Recipe for Large-Scale Molecular Systems

Edward G. Hohenstein, Melisa Alkan, Xiaojie Wu, Robert M. Parrish

The paper says

“A compensated grid-based approach is developed for hybrid exchange in large-scale molecular systems. The method uses a symmetrically overlap-fitted grid-based exchange approach, with fitting performed in the orthonormalized atomic orbital basis.”

more of the abstract

This fitting is augmented with a superposition-of-atomic-densities difference-density (dSCF) partitioning of the exchange energy, yielding the finished compensated grid-based exchange approach. The complete method is found to have orders of magnitude reduction of absolute and relative error properties versus raw grid-based exchange, and remains analytically differentiable. The method is easy to implement and provides efficiency gains over GPU integral-direct techniques for high quality and/or augmented basis sets. Library functions to help users implement the method in their own quantum chemistry codes are provided in the recently-shipped cuEST v0.3 library.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up. Well-executed engineering: compensated Grid-K works as advertised for large systems in TZ/QZ basis sets, with honest limitations stated. This line is wrong

5 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 119 references say.

5 minor objections

The paper says, at Theoretical sections (Gradients)

Pith says: minor objection 1 of 5

The gradient derivation is thorough but dense; a brief table summarizing the four gradient contributions (weights, collocation, potentials, overlap response) and their pseudo-densities (Eqs. 34-36, 40-44, 59-61) would improve readability for implementers.

The paper says, at Figures 4, 5, 8

Pith says: minor objection 2 of 5

The axis labels and tick-mark font sizes in Figures 4, 5, and 8 appear small; increasing them would improve legibility in print.

show 3 more

The paper says, at Table I caption

Pith says: minor objection 3 of 5

The systematic shift of +0.04 kcal/mol in the DF-J interaction-error mean is attributed to density fitting of the Coulomb term; a sentence in the caption or main text clarifying that this shift is not a Grid-K error (and likely removable by using more complete auxiliary sets or integral-direct J) would prevent misinterpretation.

The paper says, at §Smaller grids – the SG-0(−Δ) series

Pith says: minor objection 4 of 5

The heuristic grid-trimming procedure (SG-0(-Δ)) is clever but ad hoc; the paper acknowledges the need for a thorough grid optimization study. Consider adding a one-sentence note that future work may replace this heuristic with an automatic quality metric (e.g., monitoring the smallest eigenvalue of Ω).

The paper says, at Numerous places

Pith says: minor objection 5 of 5

A few minor typos: in Eq. (8) the symbol '≈≡' appears; consider '≈' or '≡'. In the gradient section, 'advections' is used where 'derivatives' or 'responses' might be clearer. These do not affect comprehension.

What would settle it

Run the symmetric overlap fitting on a system with a very coarse grid (e.g., SG-1 or SG-0(-4)) for a molecule with diffuse functions (e.g., water dimer in aug-cc-pVTZ) and verify that the metric Ω eigen value check flags negative values as expected, confirming the safety mechanism works.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up confidence high
  2. Skeptic partly agrees 15 s · 40,626 tokens read, 1,731 written
  3. Editor wrote the verdict above 9 s · 41,329 tokens read, 941 written
  4. Referee accept: 5 minor 16 s · 40,895 tokens read, 1,807 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

This is a clean paper. It takes two existing ideas—dSCF difference-density partitioning and symmetric overlap fitting in an orthogonalized basis—and puts them together for the first time with a full analytic gradient derivation and a GPU implementation. The benchmarks are thorough: protein–ligand interaction energies, GMTKN55, geometry optimizations, finite-difference gradient checks. The accuracy numbers are genuine: sub-0.01 kcal/mol relative errors on SG-0, near-microhartree absolute errors, and 10–50× speedups over integral-direct on a single B200 GPU. The authors give clear credit to prior work (semiJK, aCOSx, dSCF), state their own limitations, and ship code in cuEST v0.3. That is a solid contribution.

The soft spots are real but not damaging. The method only shows its full value in polarized triple- and quadruple-zeta basis sets; in double-zeta sets the overlap fit adds little, and the authors say so. The gradient depends on solving a Sylvester equation for the metric response, which is elegant but adds implementation complexity. The grids are not yet optimized for augmented basis sets—the authors flag this and call for further study. The reference implementation is proprietary (NVIDIA cuEST), though the library is freely downloadable. All authors are NVIDIA employees, which is disclosed; the conflict does not affect the science but is worth noting for context.

Who should read this: anyone doing hybrid DFT on large molecules (500+ atoms) with medium-to-large basis sets, especially those using GPUs. It is not a paradigm shift—seminumerical exchange is decades old—but it is a concrete, well-tested improvement that makes routine high-accuracy hybrid DFT on 1000-atom systems practical today.

My recommendation: accept for peer review. The paper deserves a serious referee. The derivations are correct, the benchmarks are reproducible (floats and structures in SI), and the claims are proportionate to the evidence.

The strongest objection

After careful review, the central claim that compensated Grid-K achieves orders-of-magnitude error reduction while remaining differentiable and efficient is well supported. The derivations are rigorous, the implementation is available, benchmarks cover multiple basis sets and system sizes, and limitations (grid quality for the symmetric overlap fit, reduced benefit in double-zeta basis sets, need for custom grids with augmented functions) are openly stated. The paper's weakest assumption—that the quadrature metric Ω remains positive definite—is explicitly monitored and triggers an error if violated, so it does not undermine the method's applicability within its stated scope. No internal inconsistency or unacknowledged risk was found.

The reader correctly identified positive definiteness of Ω as an assumption. However, the paper acknowledges this limitation (§Grid-Based Exchange, after Eq. 25) and bakes in a runtime check that errors out on negative eigenvalues. This makes the method self-correcting rather than fragile. My analysis found no other load-bearing gaps. The benchmarks are extensive, the gradient is verified to machine precision via finite differences, and the code is publicly available. Therefore the original ACCEPT verdict should stand.

The referee's report

If the reported accuracy and speedups hold, compensated Grid-K removes one of the last major computational bottlenecks in large-scale hybrid DFT. The combination of dSCF and symmetric overlap fitting is conceptually clean, yields a fully variational and differentiable energy, and the extensive benchmarks (multiple functionals, basis sets, grids, system sizes, gradient checks) leave little doubt about correctness. The method is particularly impactful for triple- and quadruple-zeta basis sets with high angular momentum, where integral-direct exchange is most expensive and where the overlap fit is most effective. The open-source library implementation and the detailed gradient derivation should enable rapid adoption.

Pith's reading of the paper (a machine's summary, not the authors')

The paper introduces Compensated Grid-K, a method that combines two independent corrections—symmetric overlap fitting in an orthogonalized basis and a superposition-of-atomic-densities difference-density (dSCF) energy partition—to the raw grid-based hybrid exchange scheme. These corrections reduce absolute and relative errors of grid-based exchange by orders of magnitude on standard coarse quadrature grids, achieving near-microhartree total energy accuracy and sub-0.01 kcal/mol relative energy errors for large molecular systems. The method remains analytically differentiable and is implemented on GPUs, providing speedups of 10–50× over integral-direct algorithms in high-quality basis sets. If correct, this enables hybrid density functional theory for systems previously limited by the cost of the exact exchange term.

What carries the argument. The key mechanism is the combination of two orthogonal corrections to the raw grid-based exchange (Grid-K) scheme. First, the collocation matrix is dressed by a global transfer matrix derived from the symmetric inverse square root of the quadrature metric in the canonically orthogonalized basis, ensuring that the grid integrates the overlap exactly. Second, the exchange energy is partitioned via the dSCF approach: the SAD part is evaluated analytically using standard integral-direct exchange, while the difference-density part is approximated on the grid, drastically reducing the magnitude of the quantity to be approximated. The method is differentiable, and the analytic gradient is derived, including the response of the transfer matrix via a single Sylvester equation.

Core claim. The central claim is that by simultaneously applying a symmetric overlap fit (performed in the canonically orthogonalized atomic-orbital basis to exactly reproduce the analytical overlap) and a dSCF partition (which evaluates the exchange energy of the superposition of atomic densities analytically and approximates only the small quadratic term in the difference density), standard Becke molecular quadrature grids of SG-0 quality can approximate hybrid exchange with drastically reduced error. For ωB97M-V/def2-TZVP on the SG-0 grid, the absolute energy error drops from 7.4 mEh to 16 μEh and the interaction energy error from 0.19 kcal/mol to 0.002 kcal/mol, while retaining differentiability. The method is most effective in polarized triple-zeta and quadruple-zeta basis sets and is demonstrated on systems of over 1000 atoms.

The premise it rests on. The method assumes that, for a given basis set and grid, the symmetric overlap fit in the orthogonalized basis produces a positive-definite metric matrix $\Omega$; negative eigenvalues would indicate an inadequate grid, and the authors acknowledge that standard grids like SG-0 and SG-1 generally satisfy this condition but that augmented basis sets may require customized grids.

If the paper is right (Pith's inferences, not the authors')

Hybrid DFT calculations in large basis sets (def2-TZVP, def2-QZVP) on systems with thousands of atoms become routinely tractable on a single GPU, with exchange errors below chemical accuracy.

The analytic differentiability of the method enables exact geometry optimizations and vibrational frequency calculations without the approximations needed for post-hoc corrections.

Density-fitted Coulomb builds combined with compensated Grid-K yield the highest overall speedups, with the exchange term becoming the dominant remaining cost.

The per-atom error of compensated Grid-K is approximately constant for a given basis set, allowing users to anticipate total energy errors simply from atom count.

The success of dSCF and symmetric overlap fitting in Grid-K suggests that similar compensation strategies could be applied to grid-based approximations in post-Hartree–Fock methods (e.g., MP2 or CCSD), where the magnitude of the approximated quantity can also be reduced via a reference partition.

For heavily augmented basis sets (e.g., aug-cc-pVQZ) the authors acknowledge that customized grids may be needed; a systematic grid optimization for such cases would be a natural extension, ideally yielding robust grids of SG-0(−2) size or smaller.

The global symmetric overlap fit requires a Sylvester solve that is O(N³) in the number of orthogonal orbitals; future work may investigate local or iterative alternatives to scale to systems beyond tens of thousands of basis functions.

The method still leaves the exchange term as roughly 60% of the total SCF time on the coarsest grids, indicating room for further algorithmic improvements in the grid-based step itself.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

A compensated grid-based approach is developed for hybrid exchange in large-scale molecular systems. The method uses a symmetrically overlap-fitted grid-based exchange approach, with fitting performed in the orthonormalized atomic orbital basis. This fitting is augmented with a superposition-of-atomic-densities difference-density (dSCF) partitioning of the exchange energy, yielding the finished compensated grid-based exchange approach. The complete method is found to have orders of magnitude reduction of absolute and relative error properties versus raw grid-based exchange, and remains analytically differentiable. The method is easy to implement and provides efficiency gains over GPU integral-direct techniques for high quality and/or augmented basis sets. Library functions to help users implement the method in their own quantum chemistry codes are provided in the recently-shipped cuEST v0.3 library.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.97

“The complete method is found to have orders of magnitude reduction of absolute and relative error properties versus raw grid-based exchange, and remains analytically differentiable.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.10 · human participants 0.01 · controlled comparison 0.66 · randomized assignment 0.01 · states quantitative result 0.16 · compares with prior results 0.91 · mentions code or data availability 0.90

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
A compensated grid-based approach is developed for hybrid exchange in large-scale molecular systems. contribution 0.99 0.92 0.18 asserted 1.00
The method uses a symmetrically overlap-fitted grid-based exchange approach, with fitting performed in the orthonormalized atomic orbital basis. method 0.99 0.71 0.05 asserted 1.00
This fitting is augmented with a superposition-of-atomic-densities difference-density (dSCF) partitioning of the exchange energy, yielding the finished compensated grid-based exchange approach. method 0.98 0.83 0.08 asserted 0.99
The complete method is found to have orders of magnitude reduction of absolute and relative error properties versus raw grid-based exchange, and remains analytically differentiable. result 1.00 0.93 0.53 asserted 1.00
The method is easy to implement and provides efficiency gains over GPU integral-direct techniques for high quality and/or augmented basis sets. result 0.85 0.79 0.15 asserted 1.00
Library functions to help users implement the method in their own quantum chemistry codes are provided in the recently-shipped cuEST v0.3 library. contribution 0.94 0.60 0.03 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from simulation or computation 0.97

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 4,916 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (11)
Figure 1 of the paper
Figure 1. FIG. 1: Gramicidin A dimer
Figure 2 of the paper
Figure 2. FIG. 2: Gramicidin A dimer
Figure 3 of the paper
Figure 3. FIG. 3: GNNQQNY steric zipper
Figure 4 of the paper
Figure 4. FIG. 4: Magnitude of the signed error in the
Figure 5 of the paper
Figure 5. FIG. 5: Magnitude of the signed error per atom in the
Figure 6 of the paper
Figure 6. FIG. 6: Signed errors in
Figure 7 of the paper
Figure 7. FIG. 7: Signed errors in B3LYP relative energies computed with Grid-K variants with respect to analytic two-electron integrals
Figure 8 of the paper
Figure 8. FIG. 8: RMSD between the interatomic distance matrices (over all unique atom pairs) of geometries optimized with Grid-K
Figure 9 of the paper
Figure 9. FIG. 9: Wall time for a B3LYP energy and analytic gradient on a single NVIDIA B200 GPU for clusters of 5 to 100 molecules
Figure 10 of the paper
Figure 10. FIG. 10: Speedup of the B3LYP energy and analytic gradient wall time over integral-direct J and K for the benzene crystal
Figure 11 of the paper
Figure 11. FIG. 11: Decomposition of the compensated Grid-K B3LYP energy and analytic gradient wall time on a single NVIDIA B200

Figures are the authors', reproduced from the paper.

References (119)
  1. [1] A. D. Becke, J. Chem. Phys.98, 1372 (1993), URL https://doi.org/10.1063/1.464304 on Pithdoi:10.1063/1.464304
  2. [2] A. D. Becke, J. Chem. Phys.98, 5648 (1993), URL https://doi.org/10.1063/1.464913 on Pithdoi:10.1063/1.464913
  3. [3] P. J. Stephens, F. J. Devlin, C. F. Chabalowski, and M. J. Frisch, J. Phys. Chem.98, 11623 (1994), URL https://doi.org/10.1021/j100096a001 on Pithdoi:10.1021/j100096a001
  4. [4] A. Savin, inRecent Developments and Applications of Modern Density Functional Theory, edited by J. M. Seminario (Elsevier, Amsterdam, 1996), pp. 327– 357, URLhttps://doi.org/10.1016/S1380-7323(96) 80091-4 on Pithdoi:10.1016/s1380-7323(96
  5. [5] P. M. W. Gill, R. D. Adamson, and J. A. Pople, Mol. Phys.88, 1005 (1996), URLhttps://doi.org/ 10.1080/00268979609484488 on Pithdoi:10.1080/00268979609484488
  6. [6] H. Iikura, T. Tsuneda, T. Yanai, and K. Hirao, J. Chem. Phys.115, 3540 (2001), URLhttps://doi.org/10. 1063/1.1383587 on Pith
  7. [7] J.-D. Chai and M. Head-Gordon, J. Chem. Phys. 128, 084106 (2008), URLhttps://doi.org/10.1063/ 1.2834918 on Pith
  8. [8] O. A. Vydrov and G. E. Scuseria, J. Chem. Phys. 125, 234109 (2006), URLhttps://doi.org/10.1063/ 1.2409292 on Pith
  9. [9] N. Mardirossian and M. Head-Gordon, J. Chem. Phys. 144, 214110 (2016), URLhttps://doi.org/10.1063/ 1.4952647 on Pith
  10. [10] J. Liang and M. Head-Gordon, arXiv:2603.23466 (2026), URLhttps://arxiv.org/abs/2603.23466 on PithPith's reviewarXiv
  11. [11] L. Goerigk, A. Hansen, C. Bauer, S. Ehrlich, A. Najibi, and S. Grimme, Phys. Chem. Chem. Phys.19, 32184 (2017), URLhttps://doi.org/10.1039/C7CP04913G on Pithdoi:10.1039/c7cp04913g
  12. [12] A. Dreuw, J. L. Weisman, and M. Head-Gordon, J. Chem. Phys.119, 2943 (2003), URLhttps://doi.org/ 10.1063/1.1590951 on Pithdoi:10.1063/1.1590951
  13. [13] Y. Tawada, T. Tsuneda, S. Yanagisawa, T. Yanai, and K. Hirao, J. Chem. Phys.120, 8425 (2004), URLhttps: //doi.org/10.1063/1.1688752 on Pithdoi:10.1063/1.1688752
  14. [14] J. L. Whitten, J. Chem. Phys.58, 4496 (1973), URL https://doi.org/10.1063/1.1679012 on Pithdoi:10.1063/1.1679012
  15. [15] B. I. Dunlap, J. W. D. Connolly, and J. R. Sabin, J. Chem. Phys.71, 3396 (1979), URLhttps://doi.org/ 10.1063/1.438728 on Pithdoi:10.1063/1.438728
  16. [16] O. Vahtras, J. Alml¨ of, and M. W. Feyereisen, Chem. Phys. Lett.213, 514 (1993), URLhttps://doi.org/ 10.1016/0009-2614(93)89151-7 on Pithdoi:10.1016/0009-2614(93)89151-7
  17. [17] F. Weigend, Phys. Chem. Chem. Phys.4, 4285 (2002), URLhttps://doi.org/10.1039/b204199p on Pithdoi:10.1039/b204199p
  18. [18] R. Polly, H.-J. Werner, F. R. Manby, and P. J. Knowles, Mol. Phys.102, 2311 (2004), URLhttps://doi.org/ 10.1080/0026897042000274801 on Pithdoi:10.1080/0026897042000274801
  19. [19] S. G. Balasubramani, G. P. Chen, S. Coriani, M. Diedenhofen, M. S. Frank, Y. J. Franzke, F. Furche, R. Grotjahn, M. E. Harding, C. H¨ attig, et al., J. Chem. Phys.152, 184107 (2020), URLhttps://doi.org/10. 1063/5.0004635 on Pith
  20. [20] H.-J. Werner, P. J. Knowles, F. R. Manby, J. A. Black, K. Doll, A. Heßelmann, D. Kats, A. K¨ ohn, T. Ko- rona, D. A. Kreplin, et al., J. Chem. Phys.152, 144107 (2020), URLhttps://doi.org/10.1063/5.0005081 on Pithdoi:10.1063/5.0005081
show 80 more
  1. [21] D. G. A. Smith, L. A. Burns, A. C. Simmonett, R. M. Parrish, M. C. Schieber, R. Galvelis, P. Kraus, H. Kruse, R. Di Remigio, A. Alenaizan, et al., J. Chem. Phys. 152, 184108 (2020), URLhttps://doi.org/10.1063/ 5.0006002 on Pith
  2. [22] F. Neese, WIREs Comput. Mol. Sci.12, e1606 (2022), URLhttps://doi.org/10.1002/wcms.1606 on Pithdoi:10.1002/wcms.1606
  3. [23] E. Epifanovsky, A. T. B. Gilbert, X. Feng, J. Lee, Y. Mao, N. Mardirossian, P. Pokhilko, A. F. White, M. P. Coons, A. L. Dempwolff, et al., J. Chem. Phys. 155, 084801 (2021), URLhttps://doi.org/10.1063/ 5.0055522 on Pith
  4. [24] Q. Sun, X. Zhang, S. Banerjee, P. Bao, M. Barbry, N. S. Blunt, N. A. Bogdanov, G. H. Booth, J. Chen, Z.-H. Cui, et al., J. Chem. Phys.153, 024109 (2020), URL https://doi.org/10.1063/5.0006074 on Pithdoi:10.1063/5.0006074
  5. [25] X. Wu, Q. Sun, Z. Pu, T. Zheng, W. Ma, W. Yan, Y. Xia, Z. Wu, M. Huo, X. Li, et al., WIREs Comput. Mol. Sci.15, e70008 (2025), URLhttps://doi.org/ 10.1002/wcms.70008 on Pithdoi:10.1002/wcms.70008
  6. [26] R. Li, Q. Sun, X. Zhang, and G. K.-L. Chan, J. Phys. Chem. A129, 1459 (2025), URLhttps://doi.org/10. 1021/acs.jpca.4c05876 on Pith
  7. [27] R. Wang, Y. Wang, L. Lu, D. Hait, and T. J. Mart ´ ınez, J. Chem. Theory Comput.22, 3398 (2026), URLhttps: //doi.org/10.1021/acs.jctc.5c02079 on Pithdoi:10.1021/acs.jctc.5c02079
  8. [28] NVIDIA Corporation,cuEST: Accelerating quantum chemistry on NVIDIA GPUs,https://developer. nvidia.com/cuest(2026) on Pith
  9. [29] K. Ozaki, T. Ogita, S. Oishi, and S. M. Rump, Numer. Algorithms59, 95 (2012), URLhttps://doi.org/10. 1007/s11075-011-9478-1 on Pith
  10. [30] K. Ozaki, Y. Uchino, and T. Imamura, Int. J. High Per- form. Comput. Appl. (2026), URLhttps://doi.org/ 10.1177/10943420261467787 on Pithdoi:10.1177/10943420261467787
  11. [31] A. Schwarz, A. Anders, C. Brower, H. Bayraktar, J. Gunnels, K. Clark, R. G. Xu, S. Rodriguez, S. Cay- rols, P. Tabaszewski, et al., inProceedings of the Su- 19 percomputing Asia and International Conference on High Performance Computing in Asia Pacific Region (SCA/HPCAsia 2026... on Pithdoi:10.1145/3773656.3773670
  12. [32] H. Huang, W. Shao, and J. Hammond, J. Phys. Chem. A130, 3483 (2026), URLhttps://doi.org/10.1021/ acs.jpca.6c00225 on Pith
  13. [33] J. Alml¨ of, K. Faegri, and K. Korsell, J. Comp. Chem.3, 385 (1982), ISSN 1096-987X, URLhttps://doi.org/ 10.1002/jcc.540030314 on Pithdoi:10.1002/jcc.540030314
  14. [35] L. E. McMurchie and E. R. Davidson, J. Comput. Phys.26, 218 (1978), URLhttps://doi.org/10.1016/ 0021-9991(78)90092-X on Pith
  15. [37] S. Obara and A. Saika, J. Chem. Phys.84, 3963 (1986), URLhttps://doi.org/10.1063/1.450106 on Pithdoi:10.1063/1.450106
  16. [39] P. M. W. Gill, M. Head-Gordon, and J. A. Pople, J. Phys. Chem.94, 5564 (1990), URLhttps://doi.org/ 10.1021/j100377a031 on Pithdoi:10.1021/j100377a031
  17. [40] P. M. W. Gill and J. A. Pople, Int. J. Quantum Chem. 40, 753 (1991), URLhttps://doi.org/10.1002/qua. 560400605 on Pithdoi:10.1002/qua
  18. [41] P. M. Gill, B. G. Johnson, and J. A. Pople, Chem. Phys. Lett.217, 65 (1994), URLhttps://doi.org/10.1016/ 0009-2614(93)E1340-M on Pith
  19. [43] C. Ochsenfeld, C. A. White, and M. Head-Gordon, J. Chem. Phys.109, 1663 (1998), URLhttps://doi.org/ 10.1063/1.476741 on Pithdoi:10.1063/1.476741
  20. [44] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, et al.,Gaus- sian 16, Revision C.01, Gaussian, Inc., Wallingford, CT (2016) on Pith
  21. [45] G. M. J. Barca, C. Bertoni, L. Carrington, D. Datta, N. De Silva, J. E. Deustua, D. G. Fedorov, J. R. Gour, A. O. Gunina, E. Guidez, et al., J. Chem. Phys. 152, 154102 (2020), URLhttps://doi.org/10.1063/ 5.0005188 on Pith
  22. [46] I. S. Ufimtsev and T. J. Mart ´ ınez, J. Chem. Theory Comput.4, 222 (2008), URLhttps://doi.org/10. 1021/ct700268q on Pith
  23. [47] I. S. Ufimtsev and T. J. Mart ´ ınez, J. Chem. Theory Comput.5, 1004 (2009), URLhttps://doi.org/10. 1021/ct800526s on Pith
  24. [48] A. Asadchev, V. Allada, J. Felder, B. M. Bode, M. S. Gordon, and T. L. Windus, J. Chem. Theory Com- put.6, 696 (2010), URLhttps://doi.org/10.1021/ ct9005079 on Pith
  25. [49] A. Asadchev and M. S. Gordon, J. Chem. Theory Com- put.8, 4166 (2012), URLhttps://doi.org/10.1021/ ct300526w on Pith
  26. [50] Y. Miao and K. M. Merz, J. Chem. Theory Com- put.9, 965 (2013), URLhttps://doi.org/10.1021/ ct300754n on Pith
  27. [51] M. Manathunga, Y. Miao, D. Mu, A. W. G¨ otz, and K. M. Merz, J. Chem. Theory Comput.16, 4315 (2020), URLhttps://doi.org/10.1021/acs.jctc.0c00290 on Pithdoi:10.1021/acs.jctc.0c00290
  28. [52] G. M. J. Barca, D. L. Poole, J. L. G. Vallejo, M. Alkan, C. Bertoni, A. P. Rendell, and M. S. Gordon, inSC20: International Conference for High Performance Com- puting, Networking, Storage and Analysis(IEEE, 2020), pp. 1–14, URLhttps://doi.org/10.1109/SC41405. 2020.00085 on Pithdoi:10.1109/sc41405
  29. [53] G. M. J. Barca, M. Alkan, J. L. Galvez-Vallejo, D. L. Poole, A. P. Rendell, and M. S. Gordon, J. Chem. The- ory Comput.17, 7486 (2021), URLhttps://doi.org/ 10.1021/acs.jctc.1c00720 on Pithdoi:10.1021/acs.jctc.1c00720
  30. [54] F. Ju, X. Wei, L. Huang, A. J. Jenkins, L. Xia, J. Zhang, J. Zhu, H. Yang, B. Shao, P. Dai, et al., J. Chem. Theory Comput.20, 10838 (2024), URLhttps://doi.org/10. 1021/acs.jctc.4c00940 on Pith
  31. [55] X. Wu, Q. Sun, and Y. Wang, J. Chem. Phys. 164, 194102 (2026), URLhttps://doi.org/10.1063/ 5.0314659 on Pith
  32. [56] R. A. Friesner, Chem. Phys. Lett.116, 39 (1985), URL https://doi.org/10.1016/0009-2614(85)80121-4 on Pithdoi:10.1016/0009-2614(85)80121-4
  33. [57] R. A. Friesner, J. Chem. Phys.85, 1462 (1986), URL https://doi.org/10.1063/1.451237 on Pithdoi:10.1063/1.451237
  34. [58] R. A. Friesner, J. Chem. Phys.86, 3522 (1987), URL https://doi.org/10.1063/1.451955 on Pithdoi:10.1063/1.451955
  35. [59] A. D. Bochevarov, E. Harder, T. F. Hughes, J. R. Green- wood, D. A. Braden, D. M. Philipp, D. Rinaldo, M. D. Halls, J. Zhang, and R. A. Friesner, Int. J. Quantum Chem.113, 2110 (2013), URLhttps://doi.org/10. 1002/qua.24481 on Pith
  36. [60] M. N. Ringnalda, M. Belhadj, and R. A. Friesner, J. Chem. Phys.93, 3397 (1990), URLhttps://doi.org/ 10.1063/1.458819 on Pithdoi:10.1063/1.458819
  37. [61] B. H. Greeley, T. V. Russo, D. T. Mainz, R. A. Friesner, J.-M. Langlois, W. A. Goddard, R. E. Donnelly, and M. N. Ringnalda, J. Chem. Phys.101, 4028 (1994), URLhttps://doi.org/10.1063/1.467520 on Pithdoi:10.1063/1.467520
  38. [62] F. Neese, F. Wennmohs, A. Hansen, and U. Becker, Chem. Phys.356, 98 (2009), URLhttps://doi.org/ 10.1016/j.chemphys.2008.10.036 on Pithdoi:10.1016/j.chemphys.2008.10.036
  39. [63] B. Helmich-Paris, B. de Souza, F. Neese, and R. Izs´ ak, J. Chem. Phys.155, 104109 (2021), URLhttps://doi. org/10.1063/5.0058766 on Pithdoi:10.1063/5.0058766
  40. [64] R. Izs´ ak and F. Neese, J. Chem. Phys.135, 144105 (2011), URLhttps://doi.org/10.1063/1.3646921 on Pithdoi:10.1063/1.3646921
  41. [65] C. Holzer, J. Chem. Phys.153, 184115 (2020), URL https://doi.org/10.1063/5.0022755 on Pithdoi:10.1063/5.0022755
  42. [66] R. Izs´ ak, F. Neese, and W. Klopper, J. Chem. Phys. 139, 094111 (2013), URLhttps://doi.org/10.1063/ 1.4819264 on Pith
  43. [67] P. Plessow and F. Weigend, J. Comput. Chem.33, 810 (2012), URLhttps://doi.org/10.1002/jcc.22901 on Pithdoi:10.1002/jcc.22901
  44. [68] H. Laqua, J. Kussmann, and C. Ochsenfeld, J. Chem. Theory Comput.14, 3451 (2018), URLhttps://doi. org/10.1021/acs.jctc.8b00062 on Pithdoi:10.1021/acs.jctc.8b00062
  45. [69] H. Bahmann and M. Kaupp, J. Chem. Theory Com- put.11, 1540 (2015), URLhttps://doi.org/10.1021/ ct501137x on Pith
  46. [71] H. Laqua, T. H. Thompson, J. Kussmann, and 20 C. Ochsenfeld, J. Chem. Theory Comput.16, 1456 (2020), URLhttps://doi.org/10.1021/acs.jctc. 9b00860 on Pithdoi:10.1021/acs.jctc
  47. [72] H. Laqua, J. Kussmann, and C. Ochsenfeld, J. Chem. Phys.154, 214116 (2021), URLhttps://doi.org/10. 1063/5.0045084 on Pith
  48. [73] J. DiCenso, H. Laqua, and M. Head-Gordon, arXiv:2610.21075 (2026), URLhttps://arxiv.org/ abs/2610.21075 on PithPith's reviewarXiv
  49. [74] D. B. Williams-Young, A. Asadchev, D. T. Popovici, D. Clark, J. Waldrop, T. L. Windus, E. F. Valeev, and W. A. de Jong, J. Chem. Phys.158, 234104 (2023), URLhttps://doi.org/10.1063/5.0151070 on Pithdoi:10.1063/5.0151070
  50. [75] H. Laqua, J. Kussmann, and C. Ochsenfeld, J. Chem. Phys.149, 204111 (2018), URLhttps://doi.org/10. 1063/1.5049435 on Pith
  51. [76] Q. Sun, M. R. Hermes, X. Wu, H. Zhai, X. Zhang, A. M. Ahmed, J. J. Aucar, O. J. Backhouse, S. Baner- jee, P. Bao, et al., arXiv:2603.14155 (2026), URLhttps: //arxiv.org/abs/2603.14155 on PithPith's reviewarXiv
  52. [77] Y. Zhang, R. Lei, B. Suo, and W. Liu, J. Phys. Chem. A129, 1492 (2025), URLhttps://doi.org/10.1021/ acs.jpca.4c07454 on Pith
  53. [78] Y. Zhang, B. Suo, Z. Wang, N. Zhang, Z. Li, Y. Lei, W. Zou, J. Gao, D. Peng, Z. Pu, et al., J. Chem. Phys. 152, 064113 (2020), URLhttps://doi.org/10.1063/ 1.5143173 on Pith
  54. [79] E. G. Hohenstein, R. M. Parrish, and T. J. Mart ´ ınez, J. Chem. Phys.137, 044103 (pages 10) (2012), URL https://doi.org/10.1063/1.4732310 on Pithdoi:10.1063/1.4732310
  55. [80] R. M. Parrish, E. G. Hohenstein, T. J. Mart ´ ınez, and C. D. Sherrill, J. Chem. Phys.137, 224106 (2012), URL https://doi.org/10.1063/1.4768233 on Pithdoi:10.1063/1.4768233
  56. [81] E. G. Hohenstein, R. M. Parrish, C. D. Sherrill, and T. J. Mart ´ ınez, J. Chem. Phys.137, 221101 (2012), URLhttps://doi.org/10.1063/1.4768241 on Pithdoi:10.1063/1.4768241
  57. [82] R. M. Parrish, E. G. Hohenstein, N. F. Schunck, C. D. Sherrill, and T. J. Mart ´ ınez, Phys. Rev. Lett. 111, 132505 (2013), URLhttps://doi.org/10.1103/ PhysRevLett.111.132505 on Pith
  58. [83] J. Lu and L. Ying, J. Comput. Phys.302, 329 (2015), URLhttps://doi.org/10.1016/j.jcp.2015.09.014 on Pithdoi:10.1016/j.jcp.2015.09.014
  59. [84] W. Hu, L. Lin, and C. Yang, J. Chem. Theory Comput. 13, 5420 (2017), URLhttps://doi.org/10.1021/acs. jctc.7b00807 on Pithdoi:10.1021/acs
  60. [85] J. Lee, L. Lin, and M. Head-Gordon, J. Chem. Theory Comput.16, 243 (2020), URLhttps://doi.org/10. 1021/acs.jctc.9b00820 on Pith
  61. [86] R. M. Parrish, C. D. Sherrill, E. G. Hohenstein, S. I. L. Kokkila, and T. J. Mart ´ ınez, J. Chem. Phys. 140, 181102 (2014), URLhttps://doi.org/10.1063/ 1.4876016 on Pith
  62. [87] E. G. Hohenstein, S. I. L. Kokkila, R. M. Parrish, and T. J. Mart ´ ınez, J. Chem. Phys.138, 124111 (2013), URLhttps://doi.org/10.1063/1.4795514 on Pithdoi:10.1063/1.4795514
  63. [88] E. G. Hohenstein, S. I. L. Kokkila, R. M. Parrish, and T. J. Mart ´ ınez, J. Phys. Chem. B117, 12972 (2013), URLhttps://doi.org/10.1021/jp4021905 on Pithdoi:10.1021/jp4021905
  64. [89] C. Song and T. J. Mart ´ ınez, J. Chem. Phys.144, 174111 (2016), URLhttps://doi.org/10.1063/1.4948438 on Pithdoi:10.1063/1.4948438
  65. [90] C. Song and T. J. Mart ´ ınez, J. Chem. Phys.146, 034104 (2017), URLhttps://doi.org/10.1063/1.4973840 on Pithdoi:10.1063/1.4973840
  66. [91] C. Song and T. J. Mart ´ ınez, J. Chem. Phys.149, 044108 (2018), URLhttps://doi.org/10.1063/1.5037283 on Pithdoi:10.1063/1.5037283
  67. [92] C. Song and T. J. Mart ´ ınez, J. Chem. Phys.152, 234113 (2020), URLhttps://doi.org/10.1063/5.0007417 on Pithdoi:10.1063/5.0007417
  68. [93] A. E. Hillers-Bendtsen and T. J. Mart ´ ınez, J. Phys. Chem. Lett.16, 4734 (2025), URLhttps://doi.org/ 10.1021/acs.jpclett.5c00405 on Pithdoi:10.1021/acs.jpclett.5c00405
  69. [94] A. E. Hillers-Bendtsen and T. J. Mart ´ ınez, J. Chem. Theory Comput.21, 11595 (2025), URLhttps://doi. org/10.1021/acs.jctc.5c01421 on Pithdoi:10.1021/acs.jctc.5c01421
  70. [95] R. M. Parrish, F. Liu, and T. J. Mart ´ ınez, J. Chem. Phys.144, 131101 (2016), URLhttps://doi.org/10. 1063/1.4945277 on Pith
  71. [96] A. D. Becke and R. M. Dickson, J. Chem. Phys.89, 2993 (1988), URLhttps://doi.org/10.1063/1.455005 on Pithdoi:10.1063/1.455005
  72. [97] G. Knizia, J. Chem. Theory Comput.9, 4834 (2013), URLhttps://doi.org/10.1021/ct400687b on Pithdoi:10.1021/ct400687b
  73. [98] T. H. Dunning, J. Chem. Phys.90, 1007 (1989), URL https://doi.org/10.1063/1.456153 on Pithdoi:10.1063/1.456153
  74. [99] T. Tsuchiya, M. Abe, T. Nakajima, and K. Hirao, J. Chem. Phys.115, 4463 (2001), URLhttps://doi.org/ 10.1063/1.1390515 on Pithdoi:10.1063/1.1390515
  75. [100] A. D. Becke, J. Chem. Phys.88, 2547 (1988), URL https://doi.org/10.1063/1.454033 on Pithdoi:10.1063/1.454033
  76. [101] V. Mironov, A. Paul, D. Morozov, R. Mal- laev, M. Perelshtein, and R. Ellerbrock, ChemRxiv:10.26434/chemrxiv-2025-tt68b (2026), URL https://doi.org/10.26434/chemrxiv-2025-tt68b on Pithdoi:10.26434/chemrxiv-2025-tt68b
  77. [102] Y. Yamaguchi, Y. Osamura, J. D. Goddard, and H. F. Schaefer, III,A New Dimension to Quantum Chem- istry: Analytic Derivative Methods in Ab Initio Molecu- lar Electronic Structure Theory, vol. 29 ofInternational Series of Monographs on Chemistry(Oxford University Press, New York, 1994) on Pith
  78. [103] F. Weigend, J. Comput. Chem.29, 167 (2008), URL https://doi.org/10.1002/jcc.20702 on Pithdoi:10.1002/jcc.20702
  79. [104] NVIDIA Corporation,cuBLAS library,https://docs. nvidia.com/cuda/cublas/(2026) on Pith
  80. [105] NVIDIA Corporation,cuSOL VER library,https:// docs.nvidia.com/cuda/cusolver/(2026) on Pith

The first 100 of 119 references are listed here. All of them are in this paper's citation record.

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment