pith. sign in

arxiv: 2604.12877 · v2 · submitted 2026-04-14 · ⚛️ physics.chem-ph

Fidelity of Machine Learned Potentials: Quantitative Assessment for Protonated Oxalate

Chen Qu , Paul L. Houston , Qi Yu , Apurba Nandi , Joel M. Bowman , Valerii Andreichev , Silvan K\"aser , Markus Meuwly This is my paper

Pith reviewed 2026-05-10 13:45 UTC · model grok-4.3

classification ⚛️ physics.chem-ph
keywords machine learned potentialspotential energy surfacesprotonated oxalatevibrational energiestunneling splittingsfidelity assessmentpermutationally invariant polynomialsneural network potentials
0
0 comments X

The pith

Two different machine-learned potentials for protonated oxalate agree closely on vibrational energies, infrared spectra, and hydrogen tunneling splittings.

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

The paper compares two machine-learned potential energy surfaces trained on the same electronic structure data for protonated oxalate, one using permutationally invariant polynomials and the other a message-passing neural network. It goes beyond ordinary fitting errors by running demanding stress tests that include variational vibrational calculations, infrared spectra, and three independent methods for computing tunneling splittings in hydrogen transfer. These tests involve roughly a billion energy evaluations spread across the full 15-dimensional configuration space. The two surfaces produce results in excellent agreement with each other on all quantities examined.

Core claim

Two widely applied machine-learned potential energy surfaces, one based on permutationally invariant polynomial regression and the other on the PhysNet neural network approach, are trained on the same dataset for protonated oxalate. When these surfaces are used in VSCF/VCI calculations of vibrational energies and wavefunctions, in direct comparison of infrared spectra, and in tunneling splitting calculations via ring polymer instanton theory, diffusion Monte Carlo, and the Q_im path method, the two surfaces yield results in excellent agreement with each other.

What carries the argument

Stress tests consisting of VSCF/VCI vibrational calculations, ring polymer instanton theory, diffusion Monte Carlo simulations, and the Q_im path method that evaluate the potentials on roughly one billion points across the 15-dimensional space.

If this is right

  • Either potential can be used interchangeably for computing spectroscopic properties and tunneling rates in this system.
  • Agreement on these quantities indicates that both approaches faithfully capture the potential beyond the training set points.
  • The stress-test protocol provides a practical way to check fidelity for other machine-learned surfaces in high-dimensional spaces.
  • The results support the use of either regression method for similar polyatomic molecules where extensive dynamical calculations are needed.

Where Pith is reading between the lines

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

  • If similar agreement appears for other molecules, the need to maintain multiple independent machine-learned surfaces for cross-validation could decrease.
  • The protocol could be extended to larger systems or to properties such as reaction rates that also sample distant regions of configuration space.
  • Disagreement on only one of the three tunneling methods would point to a specific limitation in how a given potential handles rare but important geometries.

Load-bearing premise

The selected stress tests are sufficient to reveal any meaningful differences in how the two potentials represent the underlying surface throughout the full 15-dimensional space.

What would settle it

A clear numerical discrepancy, larger than method-specific uncertainties, appearing in the tunneling splitting values or in the vibrational energy levels computed from the two potentials.

Figures

Figures reproduced from arXiv: 2604.12877 by Apurba Nandi, Chen Qu, Joel M. Bowman, Markus Meuwly, Paul L. Houston, Qi Yu, Silvan K\"aser, Valerii Andreichev.

Figure 1
Figure 1. Figure 1: FIG. 1. Structures of protonated oxalate anion, OxH, at the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Panel A: Comparison between PES2025 (based on PhysNet) and the PIP-based PES. Panel B: Comparison between [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. IR spectrum for OxH calculated at the respective [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Two dimensional potential energy surface (colored [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

There has been a veritable explosion of methods and software to perform machine-learned regression on datasets of electronic energies and forces to develop high-dimensional machine learned potential energy surfaces (ML-PESs). A major, but not deeply-studied aspect is how well different ML-PESs represent the same dataset on which they are trained, beyond the standard fitting precision metrics. Here, this is examined in detail using several ''stress tests'', for two widely applied machine-learned potential approaches. One is based on permutationally invariant polynomial (PIP) linear least square regression and the other is the message-passing neural network PhysNet approach. These potentials and dipole moment surfaces are used in VSCF/VCI calculations of vibrational energies and wavefunctions. The energies from the two PESs are directly compared as are the IR spectra. In addition, tunneling splittings for the hydrogen transfer between two equivalent structures are reported from using three methods: ring polymer instanton theory, diffusion Monte Carlo simulations, and the $Q_{im}$ path method. These calculations require the evaluation of on the order of one billion energies that are widely dispersed in the 15-dimensional configurational space. The two PESs yield results for these quantities in excellent agreement with each other.

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

0 major / 3 minor

Summary. The manuscript examines the fidelity of two machine-learned potential energy surfaces (ML-PESs) for protonated oxalate (a 15-dimensional system) trained on the same electronic structure dataset: a permutationally invariant polynomial (PIP) linear least-squares model and the PhysNet message-passing neural network. Beyond standard fitting metrics, the authors apply multiple stress tests—VSCF/VCI calculations of vibrational energies and IR spectra, plus tunneling splittings obtained independently via ring-polymer instanton theory, diffusion Monte Carlo, and the Q_im path method—requiring ~10^9 energy evaluations widely dispersed across configuration space. The central claim is that the two PESs (and associated dipole surfaces) produce results in excellent quantitative agreement with each other on all tested observables.

Significance. If the reported agreement holds, the work provides concrete evidence that distinct ML architectures can faithfully represent the same underlying PES for high-dimensional quantum dynamics properties, including tunneling, without being limited by fitting precision alone. The multi-method cross-validation (~10^9 points) and use of independent quantum methods constitute a genuine advance over typical ML-PES validation. This strengthens confidence in deploying ML-PESs for spectroscopy and reaction dynamics in complex molecular systems.

minor comments (3)
  1. [Abstract/Introduction] The abstract and introduction would benefit from a brief explicit statement of the training-set size, energy range, and force RMSE values for both models to allow immediate comparison with the stress-test results.
  2. [VSCF/VCI calculations] In the VSCF/VCI section, clarify whether the same vibrational basis and mode-coupling scheme were used for both PESs or whether any re-optimization was performed; this would strengthen the direct comparability of the reported energies and spectra.
  3. [Figures and Tables] Figure captions for the IR spectra and tunneling-splitting tables should include the precise number of points sampled in each method and the coordinate ranges covered, to substantiate the claim of wide dispersion in the 15D space.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript and for recommending acceptance. The referee's summary correctly captures the central claim and the extensive stress-testing performed across vibrational properties and tunneling splittings.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's central claim is that two independently trained ML-PESs (PIP linear least-squares and PhysNet neural network) produce excellent agreement on VSCF/VCI vibrational energies/IR spectra and on tunneling splittings computed via three external methods (ring-polymer instanton, DMC, Q_im path). These stress tests evaluate ~10^9 points dispersed across the 15D space and are not derived from the PES fitting procedure itself. No load-bearing step reduces the agreement result to a fitted parameter, self-definition, or self-citation chain; the multi-method cross-validation supplies independent evidence. Minor self-citations to prior PIP or PhysNet work exist but are not invoked to justify the fidelity conclusion.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The assessment rests on standard quantum chemistry methods for validation rather than new postulates; the ML models themselves contain many fitted parameters but these are not introduced by the paper.

axioms (2)
  • domain assumption VSCF/VCI calculations accurately reflect differences in PES quality for vibrational energies and wavefunctions
    Invoked to compare the two PESs via spectra
  • domain assumption Ring polymer instanton theory, diffusion Monte Carlo, and Q_im path method provide reliable tunneling splittings from a given PES
    Used as independent checks requiring ~1 billion energy evaluations

pith-pipeline@v0.9.0 · 5548 in / 1289 out tokens · 47807 ms · 2026-05-10T13:45:55.928823+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

73 extracted references · 73 canonical work pages · 1 internal anchor

  1. [1]

    1250.13 1.70018 1241.07 27.9534 90.8098 3.45034 1213.17 29.7036 49.193 5.20051 1167.59 31.4538 20.038 6.9507 1106.19 33.2041 3.61036 8.70089 1031.33 34.2042 0. 10.4511 945.749 42.7052 135.185 12.2013 852.367 51.2561 541.933 13.9515 754.143 59.8069 1165.57 15.7017 653.949 68.3575 1958.58 17.452 554.487 76.9081 2889.29 19.2022 458.221 85.4587 3937.51 20.952...

    work page 2041

  2. [2]

    Meuwly, WIREs Computational Molecular Science9, e1386 (2019)

    M. Meuwly, WIREs Computational Molecular Science9, e1386 (2019)

    work page 2019

  3. [3]

    S.ManzhosandT.J.Carrington,Chem.Rev.121,10187 (2021)

    work page 2021

  4. [5]

    O. T. Unke, S. Chmiela, H. E. Sauceda, M. Gastegger, I. Poltavsky, K. T. Schütt, A. Tkatchenko, and K.-R. Müller, Chem. Rev.121, 10142 (2021)

    work page 2021

  5. [6]

    D. Lu, H. Wang, M. Chen, L. Lin, R. Car, E. Weinan, W. Jia, and L. Zhang, Comput. Phys. Commun.259, 107624 (2021)

    work page 2021

  6. [7]

    Zhang, J

    L. Zhang, J. Han, H. Wang, R. Car, and W. E, Phys. Rev. Lett.120, 143001 (2018)

    work page 2018

  7. [8]

    O. T. Unke, M. Stöhr, S. Ganscha, T. Unterthiner, H. Maennel, S. Kashubin, D. Ahlin, M. Gastegger, L. Medrano Sandonas, J. T. Berryman,et al., Sci. Adv. 10, eadn4397 (2024)

    work page 2024

  8. [9]

    Zaverkin, M

    V. Zaverkin, M. Ferraz, F. Alesiani, and M. Niepert, arXiv preprint arXiv:2508.10841 (2025)

    work page arXiv 2025

  9. [10]

    Musaelian, S

    A. Musaelian, S. Batzner, A. Johansson, L. Sun, C. J. Owen, M. Kornbluth, and B. Kozinsky, Nat. Comm.14, 579 (2023)

    work page 2023

  10. [11]

    Kozinsky, A

    B. Kozinsky, A. Musaelian, A. Johansson, and S. Batzner, inProceedings of the International Con- ference for High Performance Computing, Networking, Storage and Analysis, SC ’23 (Association for Comput- ing Machinery, New York, NY, USA, 2023)

    work page 2023

  11. [12]

    Kadau, T

    K. Kadau, T. C. Germann, and P. S. Lomdahl, Int. J. Mod. Phys. C17, 1755 (2006)

    work page 2006

  12. [13]

    T. C. Germann and K. Kadau, Int. J. Mod. Phys. C19, 1315 (2008)

    work page 2008

  13. [14]

    Shibuta, S

    Y. Shibuta, S. Sakane, E. Miyoshi, S. Okita, T. Takaki, and M. Ohno, Nat. Comm.8, 10 (2017)

    work page 2017

  14. [15]

    Casalino, C

    L. Casalino, C. Seitz, J. Lederhofer, Y. Tsybovsky, I. A. Wilson, M. Kanekiyo, and R. E. Amaro, ACS Cent. Sci. 8, 1646 (2022)

    work page 2022

  15. [16]

    Santos, S

    K. Santos, S. Moore, T. Oppelstrup, A. Sharifian, I. Sharapov, A. Thompson, D. Z. Kalchev, D. Perez, R. Schreiber, S. Pakin,et al., inSC24: International Conference for High Performance Computing, Network- ing, Storage and Analysis(IEEE, 2024) pp. 1–13

    work page 2024

  16. [17]

    Ugarte La Torre, J

    D. Ugarte La Torre, J. Jung, and Y. Sugita, inPro- ceedings of the International Conference on High Per- formance Computing in Asia-Pacific Region(2025) pp. 1–12

    work page 2025

  17. [18]

    Lek Chaton and M

    K. Lek Chaton and M. Meuwly, J. Comput. Chem.46, e70162 (2025)

    work page 2025

  18. [19]

    S.-L. J. Lahey and C. N. Rowley, Chem. Sci.11, 2362 (2020)

    work page 2020

  19. [20]

    Gastegger, K

    M. Gastegger, K. T. Schütt, and K.-R. Müller, Chem. Sci.12, 11473 (2021)

    work page 2021

  20. [21]

    T. J. Inizan, T. Plé, O. Adjoua, P. Ren, H. Gökcan, O. Isayev, L. Lagardère, and J.-P. Piquemal, Chem. Sci. 14, 5438 (2023)

    work page 2023

  21. [22]

    K. Song, S. Käser, K. Töpfer, L. I. Vazquez-Salazar, and M. Meuwly, J. Chem. Phys.159(2023)

    work page 2023

  22. [23]

    Zinovjev, J

    K. Zinovjev, J. Chem. Theory Comput.19, 1888 (2023)

    work page 2023

  23. [24]

    Zinovjev, L

    K. Zinovjev, L. Hedges, R. Montagud Andreu, C. Woods, I. Tuñón, and M. W. van der Kamp, J. Chem. Theory Comput.20, 4514 (2024)

    work page 2024

  24. [25]

    Kalayan, I

    J. Kalayan, I. Ramzan, C. D. Williams, R. A. Bryce, and N. A. Burton, J. Comput. Chem.45, 1143 (2024)

    work page 2024

  25. [26]

    Poltavsky, A

    I. Poltavsky, A. Charkin-Gorbulin, M. Puleva, G. Fon- seca, I. Batatia, N. J. Browning, S. Chmiela, M. Cui, J. T. Frank, S. Heinen, B. Huang, S. Käser, A. Kabylda, D. Khan, C. Müller, A. J. A. Price, K. Riedmiller, K. Töpfer, T. W. Ko, M. Meuwly, M. Rupp, G. Csányi, O. A. von Lilienfeld, J. T. Margraf, K.-R. Müller, and A. Tkatchenko, Chem. Sci.16, 3720 (2025)

    work page 2025

  26. [27]

    Poltavsky, M

    I. Poltavsky, M. Puleva, A. Charkin-Gorbulin, G. Fon- seca, I. Batatia, N. J. Browning, S. Chmiela, M. Cui, J. T. Frank, S. Heinen, B. Huang, S. Käser, A. Kabylda, D. Khan, C. Müller, A. J. A. Price, K. Riedmiller, K. Töpfer, T. W. Ko, M. Meuwly, M. Rupp, G. Csányi, O. A. von Lilienfeld, J. T. Margraf, K.-R. Müller, and A. Tkatchenko, Chem. Sci.16, 3738 (2025)

    work page 2025

  27. [28]

    J. Xia, Y. Zhang, and B. Jiang, Chem. Soc. Rev. 10.1039/D5CS00104H (2025), dOI: 10.1039/D5CS00104H

    work page doi:10.1039/d5cs00104h 2025

  28. [29]

    J. M. Bowman, C. Qu, R. Conte, A. Nandi, P. L. Hous- ton, and Q. Yu, J. Chem. Phys.162, 180901 (2025)

    work page 2025

  29. [30]

    Käser, L

    S. Käser, L. I. Vazquez-Salazar, M. Meuwly, and K. Töpfer, Digit. Discov.2, 28 (2023)

    work page 2023

  30. [31]

    Hwang, S

    W. Hwang, S. L. Austin, A. Blondel, E. D. Boittier, S. Boresch, M. Buck, J. Buckner, A. Caflisch, H.-T. Chang, X. Cheng, Y. K. Choi, J.-W. Chu, M. F. Crow- ley, Q. Cui, A. Damjanovic, Y. Deng, M. Devereux, X. Ding, M. F. Feig, J. Gao, D. R. Glowacki, J. E. I. Gonzales, M. B. Hamaneh, E. D. Harder, R. L. Hayes, J. Huang, Y. Huang, P. S. Hudson, W. Im, S. M...

    work page 2024

  31. [32]

    M. E. Taylor and P. Stone, J. Mach. Learn. Res.10, 1633 (2009)

    work page 2009

  32. [33]

    S. J. Pan and Q. Yang, IEEE Trans. Knowl. Data Eng. 22, 1345 (2009)

    work page 2009

  33. [34]

    J. S. Smith, B. T. Nebgen, R. Zubatyuk, N. Lubbers, C. Devereux, K. Barros, S. Tretiak, O. Isayev, and A. E. Roitberg, Nat. Commun.10, 1 (2019)

    work page 2019

  34. [35]

    Käser and M

    S. Käser and M. Meuwly, Phys. Chem. Chem. Phys.24, 5269 (2022)

    work page 2022

  35. [36]

    Käser, J

    S. Käser, J. O. Richardson, and M. Meuwly, J. Chem. Theory Comput.18, 6840 (2022)

    work page 2022

  36. [37]

    Werner, P

    H.-J. Werner, P. J. Knowles, G. Knizia, F. R. Manby, M. Schütz, P. Celani, W. Györffy, D. Kats, T. Korona, R. Lindh, A. Mitrushenkov, G. Rauhut, K. R. Shamasun- dar, T.B.Adler, R.D.Amos, S.J.Bennie, A.Bernhards- son, A. Berning, D. L. Cooper, M. J. O. Deegan, A. J. Dobbyn, F. Eckert, E. Goll, C. Hampel, A. Hesselmann, G. Hetzer, T. Hrenar, G. Jansen, C. K...

    work page 2019

  37. [38]

    Käser, D

    S. Käser, D. Koner, and M. Meuwly, arXiv e- prints , arXiv:2411.18121 (2024), arXiv:2411.18121 [physics.chem-ph]

    work page arXiv 2024

  38. [39]

    Andreichev, S

    V. Andreichev, S. Käser, E. L. Bocanegra, M. Salik, M. A. Johnson, and M. Meuwly, Phys. Chem. Chem. Phys. (2025)

    work page 2025

  39. [40]

    B. J. Braams and J. M. Bowman, Int. Rev. Phys. Chem. 28, 577 (2009)

    work page 2009

  40. [41]

    Xie and J

    Z. Xie and J. M. Bowman, J. Chem. Theory Comput.6, 26 (2010)

    work page 2010

  41. [42]

    Msa software with gradients,https://github.com/ szquchen/MSA-2.0(2019), accessed: 2019-01-20

    work page 2019

  42. [43]

    P. L. Houston, C. Qu, Q. Yu, R. Conte, A. Nandi, J. K. Li, and J. M. Bowman, J. Chem. Phys.158, 044109 (2023)

    work page 2023

  43. [44]

    Ramakrishnan, P

    R. Ramakrishnan, P. Dral, M. Rupp, and O. A. von Lilienfeld, J. Chem. Theory Comput.11, 2087 (2015)

    work page 2087

  44. [45]

    Nandi, C

    A. Nandi, C. Qu, P. L. Houston, R. Conte, and J. M. Bowman, J. Chem. Phys.154, 051102 (2021)

    work page 2021

  45. [46]

    Nandi, G

    A. Nandi, G. Laude, S. S. Khire, N. D. Gurav, C. Qu, R. Conte, Q. Yu, S. Li, P. L. Houston, S. R. Gadre, J. O. Richardson, F. A. Evangelista, and J. M. Bowman, J. Amer. Chem. Soc.145, 9655 (2023)

    work page 2023

  46. [47]

    J. M. Bowman, C. Qu, R. Conte, A. Nandi, P. L. Hous- ton, and Q. Yu, J. Chem. Theory and Comput.19, 1 (2023)

    work page 2023

  47. [48]

    Liu and J

    Y. Liu and J. Li, J. Phys. Chem. Letts.13, 4729 (2022)

    work page 2022

  48. [49]

    P. L. Houston, C. Qu, Q. Yu, P. Pandey, R. Conte, A. Nandi, J. M. Bowman, and S. G. Kukolich, J. Chem. Theory Comput.20, 1821 (2024)

    work page 2024

  49. [50]

    Huang, B

    X. Huang, B. J. Braams, and J. M. Bowman, J. Chem. Phys.122, 044308 (2005)

    work page 2005

  50. [52]

    Griewank and A

    A. Griewank and A. Walther,Evaluating Derivatives: Principles and Techniques of Algorithmic Differentia- tion, 2nd ed. (Society for Industrial and Applied Mathe- matics (SIAM), Philadelphia, PA, 2008)

    work page 2008

  51. [53]

    Adam: A Method for Stochastic Optimization

    D. Kingma and J. Ba, arXiv preprint arXiv:1412.6980 (2014)

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  52. [54]

    S. J. Reddi, S. Kale, and S. Kumar, On the convergence of adam and beyond (2019), arXiv:1904.09237 [cs.LG]

    work page Pith review arXiv 2019

  53. [55]

    O. T. Unke and M. Meuwly, J. Chem. Theory Comput. 15, 3678 (2019)

    work page 2019

  54. [56]

    Boittier, K

    E. Boittier, K. Töpfer, M. Devereux, and M. Meuwly, Journal of Chemical Theory and Computation20, 8088 (2024)

    work page 2024

  55. [57]

    E. D. Boittier and M. Meuwly, arXiv preprint arXiv:2602.07234 (2026)

    work page arXiv 2026

  56. [58]

    Käser, D

    S. Käser, D. Koner, A. S. Christensen, O. A. von Lilien- feld, and M. Meuwly, J. Phys. Chem. A124, 8853 (2020)

    work page 2020

  57. [59]

    164, 061103 (2026)

    C.Qu, P.L.Houston,andJ.M.Bowman,J.Chem.Phys. 164, 061103 (2026)

    work page 2026

  58. [60]

    Käser, J

    S. Käser, J. O. Richardson, and M. Meuwly, J. Chem. Theory Comput. , arXiv:2407.21366 (2025)

    work page arXiv 2025

  59. [61]

    Nandi, G

    A. Nandi, G. Laude, S. S. Khire, N. D. Gurav, C. Qu, R. Conte, Q. Yu, S. Li, P. L. Houston, S. R. Gadre,et al., J. Am. Chem. Soc.145, 9655 (2023)

    work page 2023

  60. [62]

    C. Qu, A. Nandi, P. L. Houston, and J. M. Bowman, J. Phys. Chem. Axx, xx (2026)

    work page 2026

  61. [63]

    J. M. Bowman, S. Carter, and X. Huang, Int. Rev. Phys. Chem.22, 533 (2003)

    work page 2003

  62. [64]

    Yu and et al.,MULTIMODE, the n-mode Represen- tation of the Potential and Illustrations to IR Spectra of Glycine and Two Protonated Water Clusters., edited by J

    Q. Yu and et al.,MULTIMODE, the n-mode Represen- tation of the Potential and Illustrations to IR Spectra of Glycine and Two Protonated Water Clusters., edited by J. M. Bowman (Vibrational Dynamics of Molecules, World Scientific, 2022)

    work page 2022

  63. [65]

    C. Qu, T. C. Allison, P. L. Houston, R. Conte, A. Nandi, and J. M. Bowman, J. Chem. Phys.xxx, xxxx (2026), in revision

    work page 2026

  64. [66]

    J. B. Anderson, J. Chem. Phys.63, 1499 (1975)

    work page 1975

  65. [67]

    Kosztin, B

    I. Kosztin, B. Faber, and K. Schulten, Am. J. Phys.64, 633 (1996)

    work page 1996

  66. [68]

    J. B. Anderson, J. Chem. Phys.65, 4121 (1976)

    work page 1976

  67. [69]

    Y. Wang, B. J. Braams, J. M. Bowman, S. Carter, and D. P. Tew, J. Chem. Phys.128, 224314 (2008)

    work page 2008

  68. [70]

    J. O. Richardson and S. C. Althorpe, J. Chem. Phys. 134, 054109 (2011)

    work page 2011

  69. [71]

    J. O. Richardson, Int. Rev. Phys. Chem.37, 171 (2018)

    work page 2018

  70. [72]

    J. O. Richardson, Intern. Rev. Phys. Chem.37, 171 (2018)

    work page 2018

  71. [73]

    J. E. Lawrence, J. Dušek, and J. O. Richardson, J. Chem. Phys.159, 014111 (2023)

    work page 2023

  72. [74]

    Kamarchik, Y

    E. Kamarchik, Y. Wang, and J. Bowman, J. Phys. Chem. A113, 7556 (2009)

    work page 2009

  73. [75]

    D. T. Colbert and W. H. Miller, J. Chem. Phys.96, 1982 (1992)

    work page 1982