Pith. sign in

REVIEW 4 major objections 5 minor 80 references

Fine-tuning a universal machine-learning interatomic potential on about 100 compound-specific DFT frames removes the systematic force softening and makes MD-derived EXAFS spectra of WS2 and MoS2 match experiment at 300 K.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Fine-tuning CHGNet with about 100 DFT frames for WS2 and MoS2 removes systematic softening and aligns MD-derived EXAFS spectra with experiment.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection A careful, useful benchmark showing ~100 fine-tuning frames fixes CHGNet's softening and buys DFT-level EXAFS agreement, with minor caveats about the PBE-D3 reference and missing error bars. the 4 major comments →

arxiv 2509.08498 v1 pith:WVJ6TFK5 submitted 2025-09-10 cond-mat.mtrl-sci

Benchmarking CHGNet Universal Machine Learning Interatomic Potential Against DFT and EXAFS: Case of Layered WS2 and MoS2

Pjotrs \v{Z}guns , Inga Pudza , Alexei Kuzmin This is my paper
classification cond-mat.mtrl-sci
keywords universal machine learning interatomic potentialsCHGNetfine-tuningEXAFSthermal disorderWS2MoS2van der Waals interactions
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Universal machine-learning interatomic potentials promise near-ab initio accuracy, but they are trained mainly on near-equilibrium structures and tend to underestimate forces—a 'softening' that shows up as exaggerated thermal vibration. This paper tests that effect for two isostructural layered materials, WS2 and MoS2, by comparing the CHGNet potential against DFT and against experimental EXAFS spectra, which are sensitive to the spread of bond lengths and angles at 300 K. It argues that fine-tuning CHGNet on compound-specific DFT relaxation trajectories corrects the softening, and that around one hundred DFT frames is enough to reach DFT-level force accuracy and reproduce the measured spectra. If true, this gives practitioners a cheap, practical recipe: take a general-purpose potential, refine it on about a hundred DFT calculations, and use it for reliable thermal-disorder simulations. The paper also finds that adding more DFT data does not automatically improve agreement with experiment, because the chosen exchange-correlation functional fixes the target structure.

Core claim

The central claim is that the vanilla CHGNet potential systematically underestimates interatomic forces in WS2 and MoS2—force parity slopes near 0.6 and a force MAE of 337 meV/Å for WS2—and that this softening produces overly damped theoretical EXAFS spectra at high wavenumber. Fine-tuning on relaxation trajectories computed with PBE-D3, even with a single displaced DFT structure, reduces the force error by roughly half and restores near-unity parity slopes; about 100 structures brings the force MAE to around 100 meV/Å, comparable to routine ab initio MD. Against both DFT references and experimental W L3-edge and Mo K-edge EXAFS spectra at 300 K, the fine-tuned potential reproduces the measu

What carries the argument

The load-bearing pieces are the universal potential itself and the MD-EXAFS validation pipeline. The potential is a graph neural network pre-trained on a large database of relaxation trajectories; it predicts energies, forces, stresses, and magnetic moments. The tested mechanism is fine-tuning: short relaxation trajectories are generated at seven isotropic strains with 0.2 Å random atomic displacements, and the model's weights are updated on energies, forces, and stresses with most layers frozen. The fine-tuned model then drives 300 K molecular dynamics; the MD trajectories yield radial distribution functions and mean-square relative displacements (bond-length variances), and configuration-a

Load-bearing premise

The central results assume that PBE-D3 DFT is an accurate reference for the thermal forces and equilibrium structure of WS2 and MoS2, and that roughly seven thousand short relaxation snapshots with 0.2 Å random displacements at seven isotropic strains cover the 300 K configurational space; if PBE-D3 misdescribes the interlayer van der Waals coupling that controls EXAFS damping, the '100 frames is enough' recommendation would be an artifact of that functional choice.

What would settle it

The general recipe can be tested directly by applying the same ~100-frame PBE-D3 fine-tuning protocol to a third layered dichalcogenide, say WSe2, and requiring both a held-out force MAE near 100 meV/Å and an MD-derived EXAFS envelope error no larger than the fine-tuned WS2 case against experimental 300 K spectra; failure on either target would show that the 'about 100 frames' recipe does not transfer.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A practical protocol emerges: for a new compound, fine-tune a universal potential on about 100 DFT relaxation frames before running production MD, rather than training a bespoke MLIP from scratch.
  • Thermal-disorder properties such as EXAFS spectra, mean-square relative displacements, and elastic constants become accessible with near-DFT accuracy at a fraction of the cost of ab initio MD.
  • MD-EXAFS becomes a standard experimental validation tool for machine-learning potentials, with EXAFS damping serving as a direct fingerprint of force softening.
  • Fine-tuning can teach a potential interactions—here van der Waals forces—that were absent from its pre-training data, extending universal-potential applicability to weakly bonded layered materials.
  • Choosing the right exchange-correlation functional matters more than simply adding DFT frames: refining to a functional that misdescribes the lattice will improve agreement with that functional while worsening agreement with experiment.

Where Pith is reading between the lines

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

  • If the ~100-frame scaling extends to other universal potentials and compounds, this protocol offers a template for turning a generic foundation model into a compound-specific force field with only a few hours of DFT data.
  • The observation that fine-tuning on a single strain state matched experiment slightly better than seven strain states suggests a practical refinement: tune first at the experimental lattice parameters, then add multiple strains only if stress accuracy is needed.
  • The intralayer-interlayer MSRD gap of about 0.005 Ų, consistent with prior EXAFS work on layered dichalcogenides, hints that fine-tuned universal potentials could replace ab initio MD in quantitative studies of two-dimensional materials—an implication the paper leaves for future work.
  • Active learning or smarter snapshot selection might push the needed dataset below 100 frames; the paper notes this possibility but does not test it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper benchmarks the universal machine-learning interatomic potential CHGNet against DFT and experimental EXAFS for layered 2Hc-WS2 and 2Hc-MoS2. The authors find that vanilla CHGNet systematically underestimates forces (parity slopes ~0.6), overestimates the van der Waals gap and lattice constant c (17.54 Å vs. 12.32 Å), and predicts excessive EXAFS damping. Fine-tuning CHGNet on as few as ~100 compound-specific DFT relaxation frames reduces force MAEs to ~100 meV/Å, corrects the parity slopes, and brings MD-derived EXAFS spectra into substantially better agreement with experiment. The paper recommends ~100 frames as a practical fine-tuning dataset size, claiming this is more data-efficient than training MLIPs from scratch and sufficient to reach DFT-level accuracy for thermal disorder.

Significance. If the central claim holds, the paper provides a useful, practical recommendation for fine-tuning uMLIPs with minimal DFT data and demonstrates EXAFS as a sensitive external benchmark for thermal disorder. The study has notable strengths: it uses two isostructural compounds, a systematic progression of fine-tuning set sizes (1, 10, 50, 70, 100, 350, 2000, 7000 frames), separated test sets, parity-slope diagnostics, and public deposition of the benchmark data on Zenodo. The comparison against state-of-the-art classical potentials and other uMLIPs adds context. The main limitation is that the quantitative '100 frames' recommendation is tied to the assumption that PBE-D3 is the correct reference for EXAFS-relevant dynamics, and the EXAFS error analysis lacks uncertainty quantification.

major comments (4)
  1. [§3.2 / Figure 5] The EXAFS MSE bar chart has no error bars, and the central recommendation ('~100 frames suffice') is based on differences between single MSE values that appear small relative to the visible scatter. For example, the observation that fine-tuning on a single strain state is slightly better than seven strain states, and that errors 'stabilize' after 50 frames, is not statistically supported. Please provide uncertainties (e.g., bootstrap over training seeds, multiple fine-tuning runs, or over EXAFS k-ranges) and test whether the 50 vs 100 vs 350 frame differences are significant.
  2. [§3.2 / Figure 4] The claim that vanilla CHGNet overestimates EXAFS damping due to systematic softening is confounded by its severely overestimated c lattice (17.54 Å vs 12.32 Å experimental). The overestimated interlayer spacing removes or shifts interlayer coordination shells and directly changes the multiple-scattering contributions that dominate high-k damping. To attribute the EXAFS difference to force softening, the authors should either run vanilla CHGNet MD with the correct experimental lattice parameters or compare RDF widths/MSRDs at fixed geometry. This is load-bearing for the EXAFS-based diagnosis of softening, even though the force benchmarks independently show softening.
  3. [§3.2 / Figure 3 and §3.2 discussion] The ~100-frame recommendation rests on the uncontrolled trade-off between improving force accuracy and the slow convergence of z_vdW toward the PBE-D3 value. The paper itself states that increasing frames beyond ~50 leads to a slight but steady increase in EXAFS error, attributed to z_vdW drifting away from the experimental value. This means the optimal frame count is an artifact of the chosen functional's equilibrium structure, not of force accuracy alone. The authors should directly compare the fine-tuned model's MD-EXAFS to PBE-D3 MD-EXAFS (i.e., DFT-MD-EXAFS) and to MSRDs/RDF widths from PBE-D3 MD, so the reader can see whether the fine-tuned model reproduces the reference DFT dynamics. Without that, the '100 frames' guidance is not robust to the choice of exchange-correlation functional (e.g., r2SCAN or MBD), which is acknowledged but not tested.
  4. [§3.2 / Figure 5] The MSE metric is computed over a k-range of 3–16 Å^{-1}, but the spectra in Figure 4(a) show the largest discrepancies between vanilla and fine-tuned CHGNet only for k > 7 Å^{-1}. A single integrated MSE can obscure where the disagreement lies. Please report k-resolved error curves or at least a decomposition into low-k and high-k contributions, and justify the chosen k-range. This is relevant to the conclusion that '~100 frames suffice to reproduce EXAFS spectra'.
minor comments (5)
  1. [Abstract / §2.1] Typo: 'Perdew–Burke–Enrzerhoff' should be 'Ernzerhof'.
  2. [Figure 5] The inset (seven strain states) is difficult to read; please increase font size and add axis labels. Also, the x-axis label 'N_DFT frames = 0' is written as text; use a consistent numerical axis.
  3. [§3.1] The terms 'frames' and 'structures' are used interchangeably. Clarify that one relaxation trajectory frame corresponds to one structure, and specify whether the 7000 frames include the initial displaced configurations.
  4. [§3.2 / Figure 4(d)] The MSRD gap of 0.005 Å^2 between intralayer and interlayer pairs is reported without uncertainty. Since the RDF Gaussian fitting procedure (Figure S11) involves multiple overlapping peaks, please provide error estimates for these MSRD values.
  5. [References] Reference to the MACE paper (ref. 20) uses 'arXiv preprint' without a journal citation; if a peer-reviewed version exists, it should be cited. Also, the link to Zenodo (ref. in Supporting Information) should be checked for permanence.

Circularity Check

0 steps flagged

No significant circularity: EXAFS is an external benchmark and the DFT fine-tuning target is independent of it.

full rationale

The paper's central claim is that fine-tuning CHGNet with compound-specific PBE-D3 DFT data reduces the systematic force softening of the vanilla uMLIP and improves agreement between MD-derived and experimental EXAFS spectra. The EXAFS data are measured independently (Section 2.5) and are never used in fine-tuning, which is performed only on DFT energies, forces, and stresses (Section 2.2). The comparison to EXAFS is therefore an external benchmark, not a fitted input. The DFT reference (PBE-D3) is chosen based on reproducing experimental lattice parameters (Table S1, cited in Section 3.2), not on EXAFS agreement, so the benchmark chain is not circular. The paper even explicitly acknowledges the relevant limitation: 'Expanding the DFT dataset improves uMLIP accuracy relative to the chosen exchange-correlation functional, but does not guarantee better agreement with experiments' (Section 3.2). This shows the authors distinguish DFT-level accuracy from experimental fidelity rather than conflating them. The self-citations (e.g., refs. 51, 52, 66, 67) are methodological tooling for the MD-EXAFS approach and previous layered-compound studies; they are not load-bearing uniqueness arguments or ansatz smuggling. All quantitative comparisons—force MAEs, parity slopes, EXAFS MSEs, RDF widths, and MSRDs—are computed from the paper's own simulations and the external EXAFS experiment. No derivation step reduces by construction to its own input, and no fitted parameter is renamed as a prediction. The paper is self-contained against an external experimental benchmark, so the circularity score is 0.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities or ad hoc forces. The central claim rests on the choice of DFT functional, the representativeness of the fine-tuning set, and the accuracy of the EXAFS forward model. Two analysis-level parameters (E0 and the energy shift) are standard but technically fitted to align spectra or energies.

free parameters (2)
  • EXAFS threshold energy E0 = chosen to align experimental and theoretical spectra
    Section 2.5; a standard but non-unique choice that affects the quantitative MSE comparison in Figure 5.
  • Constant energy shift for CHGNet energies = 86 meV/atom (vanilla), 1.5 meV/atom (fine-tuned)
    Figure 2 caption; applied to correct systematic offset before computing MAE, standard practice but technically a fitted alignment.
axioms (4)
  • domain assumption PBE-D3 DFT is an accurate reference for forces, stresses, and equilibrium structure of WS2/MoS2, including vdW interactions.
    The fine-tuning target and all benchmark comparisons use PBE-D3; the paper justifies the functional via lattice parameter agreement (Table S1), but if PBE-D3 misdescribes anharmonicity or interlayer coupling, the improved EXAFS agreement would be an artifact.
  • domain assumption The fine-tuning dataset (relaxation trajectories from 7 isotropic strains with 0.2 Å displacements, 10 steps) is representative of the 300 K MD configurational space.
    Section 2.2; the model is tested on some MD snapshots, but the training distribution may not cover long-wavelength anharmonic modes relevant to EXAFS damping.
  • domain assumption FEFF8.5L multiple-scattering calculations with muffin-tin potential and 8 Å path length accurately convert MD geometries to EXAFS spectra.
    Standard methodology (refs 27, 28, 51, 52), but any systematic error in the EXAFS forward model would affect the comparison.
  • domain assumption 50 ps production MD at 300 K with 1 fs timestep yields converged thermal averages for MSRDs and EXAFS.
    Section 2.3; no autocorrelation analysis or convergence test is shown.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Benchmarking CHGNet Universal Machine Learning Interatomic Potential Against DFT and EXAFS: Case of Layered WS2 and MoS2." pith.science (2026). https://pith.science/paper/WVJ6TFK5

@misc{pith2026250908498,
  author       = {Pith},
  title        = {Pith review of: Benchmarking CHGNet Universal Machine Learning Interatomic Potential Against DFT and EXAFS: Case of Layered WS2 and MoS2},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WVJ6TFK5}},
  note         = {Machine review of arXiv:2509.08498}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Universal machine learning interatomic potentials (uMLIPs) deliver near ab initio accuracy in energy and force calculations at low computational cost, making them invaluable for materials modeling. Although uMLIPs are pre-trained on vast ab initio datasets, rigorous validation remains essential for their ongoing adoption. In this study, we use the CHGNet uMLIP to model thermal disorder in isostructural layered 2Hc-WS2 and 2Hc-MoS2, benchmarking it against ab initio data and extended X-ray absorption fine structure (EXAFS) spectra, which capture thermal variations in bond lengths and angles. Fine-tuning CHGNet with compound-specific ab initio (DFT) data mitigates the systematic softening (i.e., force underestimation) typical of uMLIPs and simultaneously improves alignment between molecular dynamics-derived and experimental EXAFS spectra. While fine-tuning with a single DFT structure is viable, using ~100 structures is recommended to accurately reproduce EXAFS spectra and achieve DFT-level accuracy. Benchmarking the CHGNet uMLIP against both DFT and experimental EXAFS data reinforces confidence in its performance and provides guidance for determining optimal fine-tuning dataset sizes.

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

80 extracted references · 2 linked inside Pith

  1. [1]

    Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces

    Behler, J.; Parrinello, M. Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces. Phys. Rev. Lett. 2007, 98, 146401

    2007

  2. [2]

    P.; Payne, M

    Bart\'ok, A. P.; Payne, M. C.; Kondor, R.; Cs\'anyi, G. Gaussian Approximation Potentials: The Accuracy of Quantum Mechanics, without the Electrons. Phys. Rev. Lett. 2010, 104, 136403

    2010

  3. [3]

    P.; Swiler, L

    Thompson, A. P.; Swiler, L. P.; Trott, C. R.; Foiles, S. M.; Tucker, G. J. Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials. J. Comput. Phys. 2015, 285, 316--330

    2015

  4. [4]

    Shapeev, A. V. Moment tensor potentials: A class of systematically improvable interatomic potentials. Multiscale Model. Simul. 2016, 14, 1153--1173

    2016

  5. [5]

    S.; Isayev, O.; Roitberg, A

    Smith, J. S.; Isayev, O.; Roitberg, A. E. ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chem. Sci. 2017, 8, 3192--3203

    2017

  6. [6]

    DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics

    Wang, H.; Zhang, L.; Han, J.; Weinan, E. DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics. Comput. Phys. Commun. 2018, 228, 178--184

    2018

  7. [7]

    Atomic cluster expansion for accurate and transferable interatomic potentials

    Drautz, R. Atomic cluster expansion for accurate and transferable interatomic potentials. Phys. Rev. B 2019, 99, 014104

    2019

  8. [8]

    P.; Kermode, J.; Bernstein, N.; Cs\'anyi, G

    Bart\'ok, A. P.; Kermode, J.; Bernstein, N.; Cs\'anyi, G. Machine Learning a General-Purpose Interatomic Potential for Silicon. Phys. Rev. X 2018, 8, 041048

    2018

  9. [9]

    On-the-fly machine learning force field generation: Application to melting points

    Jinnouchi, R.; Karsai, F.; Kresse, G. On-the-fly machine learning force field generation: Application to melting points. Phys. Rev. B 2019, 100, 014105

    2019

  10. [10]

    Phase Transitions of Hybrid Perovskites Simulated by Machine-Learning Force Fields Trained on the Fly with Bayesian Inference

    Jinnouchi, R.; Lahnsteiner, J.; Karsai, F.; Kresse, G.; Bokdam, M. Phase Transitions of Hybrid Perovskites Simulated by Machine-Learning Force Fields Trained on the Fly with Bayesian Inference. Phys. Rev. Lett. 2019, 122, 225701

    2019

  11. [11]

    Phase transitions in inorganic halide perovskites from machine-learned potentials

    Fransson, E.; Wiktor, J.; Erhart, P. Phase transitions in inorganic halide perovskites from machine-learned potentials. J. Phys. Chem. C 2023, 127, 13773--13781

    2023

  12. [12]

    Accessing thermal conductivity of complex compounds by machine learning interatomic potentials

    Korotaev, P.; Novoselov, I.; Yanilkin, A.; Shapeev, A. Accessing thermal conductivity of complex compounds by machine learning interatomic potentials. Phys. Rev. B 2019, 100, 144308

    2019

  13. [13]

    Thermal transport and phase transitions of zirconia by on-the-fly machine-learned interatomic potentials

    Verdi, C.; Karsai, F.; Liu, P.; Jinnouchi, R.; Kresse, G. Thermal transport and phase transitions of zirconia by on-the-fly machine-learned interatomic potentials. npj Comput. Mater. 2021, 7, 156

    2021

  14. [14]

    H.; Li, X.; Ong, S

    Qi, J.; Banerjee, S.; Zuo, Y.; Chen, C.; Zhu, Z.; Chandrappa, M. H.; Li, X.; Ong, S. P. Bridging the gap between simulated and experimental ionic conductivities in lithium superionic conductors. Mater. Today Phys. 2021, 21, 100463

    2021

  15. [15]

    S.; Xiong, G.; Li, J.; Haile, S

    Z guns, P.; Klyukin, K.; Wang, L. S.; Xiong, G.; Li, J.; Haile, S. M.; Yildiz, B. Uncovering fast solid-acid proton conductors based on dynamics of polyanion groups and proton bonding strength. Energy Environ. Sci. 2024, 17, 5730--5742

    2024

  16. [16]

    Atomistic line graph neural network for improved materials property predictions

    Choudhary, K.; DeCost, B. Atomistic line graph neural network for improved materials property predictions. npj Comput. Mater. 2021, 7, 185

    2021

  17. [17]

    Chen, C.; Ong, S. P. A universal graph deep learning interatomic potential for the periodic table. Nat. Comput. Sci. 2022, 2, 718--728

    2022

  18. [18]

    J.; Ceder, G

    Deng, B.; Zhong, P.; Jun, K.; Riebesell, J.; Han, K.; Bartel, C. J.; Ceder, G. CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling. Nat. Mach. Intell. 2023, 5, 1031--1041

    2023

  19. [19]

    S.; Aykol, M.; Cheon, G.; Cubuk, E

    Merchant, A.; Batzner, S.; Schoenholz, S. S.; Aykol, M.; Cheon, G.; Cubuk, E. D. Scaling deep learning for materials discovery. Nature 2023, 624, 80--85

    2023

  20. [20]

    M.; Kov \'a cs, D

    Batatia, I.; Benner, P.; Chiang, Y.; Elena, A. M.; Kov \'a cs, D. P.; Riebesell, J.; Advincula, X. R.; Asta, M.; Baldwin, W. J.; Bernstein, N.; others A foundation model for atomistic materials chemistry. arXiv preprint arXiv:2401.00096 2023,

    Pith/arXiv arXiv 2023

  21. [21]

    arXiv preprint arXiv:2405.04967 2024,

    Yang, H.; Hu, C.; Zhou, Y.; Liu, X.; Shi, Y.; Li, J.; Li, G.; Chen, Z.; Chen, S.; Zeni, C.; others Mattersim: A deep learning atomistic model across elements, temperatures and pressures. arXiv preprint arXiv:2405.04967 2024,

    Pith/arXiv arXiv 2024

  22. [22]

    A.; Ceder, G

    Deng, B.; Choi, Y.; Zhong, P.; Riebesell, J.; Anand, S.; Li, Z.; Jun, K.; Persson, K. A.; Ceder, G. Systematic softening in universal machine learning interatomic potentials. npj Computational Materials 2025, 11, 9

    2025

  23. [23]

    D.; Gardner, J

    Morrow, J. D.; Gardner, J. L. A.; Deringer, V. L. How to validate machine-learned interatomic potentials. J. Chem. Phys. 2023, 158, 121501

    2023

  24. [24]

    Systematic assessment of various universal machine-learning interatomic potentials

    Yu, H.; Giantomassi, M.; Materzanini, G.; Wang, J.; Rignanese, G.-M. Systematic assessment of various universal machine-learning interatomic potentials. MGE Advances 2024, 2, e58

    2024

  25. [25]

    Sensitivity of Extended X-Ray-Absorption Fine Structure to Thermal Expansion

    Dalba, G.; Fornasini, P.; Grisenti, R.; Purans, J. Sensitivity of Extended X-Ray-Absorption Fine Structure to Thermal Expansion. Phys. Rev. Lett. 1999, 82, 4240--4243

    1999

  26. [26]

    D.; Dalba, G.; Grisenti, R.; De Panfilis, S.; Kuzmin, A.; Ozhogin, V

    Purans, J.; Afify, N. D.; Dalba, G.; Grisenti, R.; De Panfilis, S.; Kuzmin, A.; Ozhogin, V. I.; Rocca, F.; Sanson, A.; Tiutiunnikov, S. I.; Fornasini, P. Isotopic effect in extended X-ray-absorption fine structure of germanium . Phys. Rev. Lett. 2008, 100, 055901

    2008

  27. [27]

    J.; Albers, R

    Rehr, J. J.; Albers, R. C. Theoretical approaches to X-ray absorption fine structure . Rev. Mod. Phys. 2000, 72, 621--654

    2000

  28. [28]

    L.; Ravel, B.; Rehr, J

    Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Real-space multiple-scattering calculation and interpretation of x-ray-absorption near-edge structure . Phys. Rev. B 1998, 58, 7565--7576

    1998

  29. [29]

    The use of X-ray absorption spectra for validation of classical force-field models

    Kuzmin, A.; Anspoks, A.; Kalinko, A.; Timoshenko, J. The use of X-ray absorption spectra for validation of classical force-field models . Z. Phys. Chem. 2016, 230, 537--549

    2016

  30. [30]

    V.; Bocharov, D.; Kuzmin, A

    Shapeev, A. V.; Bocharov, D.; Kuzmin, A. Validation of moment tensor potentials for fcc and bcc metals using EXAFS spectra. Comput. Mater. Sci. 2022, 210, 111028

    2022

  31. [31]

    P.; Hautier, G.; Chen, W.; Richards, W

    Jain, A.; Ong, S. P.; Hautier, G.; Chen, W.; Richards, W. D.; Dacek, S.; Cholia, S.; Gunter, D.; Skinner, D.; Ceder, G.; Persson, K. A. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation . APL Mater. 2013, 1, 011002

    2013

  32. [32]

    https://matbench-discovery.materialsproject.org, Accessed: 2024-11-22

    Matbench Discovery. https://matbench-discovery.materialsproject.org, Accessed: 2024-11-22

    2024

  33. [33]

    Crystal structures of tungsten disulfide and diselenide

    Schutte, W.; De Boer , J.; Jellinek, F. Crystal structures of tungsten disulfide and diselenide . J. Solid State Chem. 1987, 70, 207--209

    1987

  34. [34]

    Phase transitions between polytypes and intralayer superstructures in transition metal dichalcogenides

    Katzke, H.; Tol\'edano, P.; Depmeier, W. Phase transitions between polytypes and intralayer superstructures in transition metal dichalcogenides. Phys. Rev. B 2004, 69, 134111

    2004

  35. [35]

    Anisotropic mean-square displacements (MSD) in single-crystals of 2H- and 3R-MoS _2

    Sch \"o nfeld, B.; Huang, J.; Moss, S. Anisotropic mean-square displacements (MSD) in single-crystals of 2H- and 3R-MoS _2 . Acta Cryst. B 1983, 39, 404--407

    1983

  36. [36]

    Bl\"ochl, P. E. Projector augmented-wave method. Phys. Rev. B 1994, 50, 17953--17979

    1994

  37. [37]

    Ab initio molecular dynamics for liquid metals

    Kresse, G.; Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B 1993, 47, 558--561

    1993

  38. [38]

    Ab initio molecular-dynamics simulation of the liquid-metal--amorphous-semiconductor transition in germanium

    Kresse, G.; Hafner, J. Ab initio molecular-dynamics simulation of the liquid-metal--amorphous-semiconductor transition in germanium. Phys. Rev. B 1994, 49, 14251--14269

    1994

  39. [39]

    Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set

    Kresse, G.; Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comp. Mater. Sci. 1996, 6, 15--50

    1996

  40. [40]

    Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set

    Kresse, G.; Furthm\"uller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 1996, 54, 11169--11186

    1996

  41. [41]

    From ultrasoft pseudopotentials to the projector augmented-wave method

    Kresse, G.; Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 1999, 59, 1758--1775

    1999

  42. [42]

    P.; Burke, K.; Ernzerhof, M

    Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865--3868

    1996

  43. [43]

    P.; Burke, K.; Ernzerhof, M

    Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple [Phys. Rev. Lett. 77, 3865 (1996)]. Phys. Rev. Lett. 1997, 78, 1396--1396

    1996

  44. [44]

    A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu

    Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu . J. Chem. Phys. 2010, 132, 154104

    2010

  45. [45]

    Abraham, F. F. Computational statistical mechanics methodology, applications and supercomputing . Adv. Phys. 1986, 35, 1--111

    1986

  46. [46]

    Hoover, W. G. Canonical dynamics: Equilibrium phase-space distributions . Phys. Rev. A 1985, 31, 1695--1697

    1985

  47. [47]

    H.; Mortensen, J

    Larsen, A. H.; Mortensen, J. J.; Blomqvist, J.; Castelli, I. E.; Christensen, R.; Du ak, M.; Friis, J.; Groves, M. N.; Hammer, B.; Hargus, C.; others The atomic simulation environment—a Python library for working with atoms. J. Phys.: Condens. Matter 2017, 29, 273002

    2017

  48. [48]

    V.; Lukyanov, S

    Bandura, A. V.; Lukyanov, S. I.; Domnin, A. V.; Kuruch, D. D.; Evarestov, R. A. Density functional and force field modeling of multi-walled WS _2 nanotubes . Comput. Theor. Chem. 2023, 1229, 114333

    2023

  49. [49]

    Gale, J. D. GULP: A computer program for the symmetry-adapted simulation of solids . J. Chem. Soc. Faraday Trans. 1997, 93, 629--637

    1997

  50. [50]

    D.; Rohl, A

    Gale, J. D.; Rohl, A. L. The General Utility Lattice Program (GULP) . Mol. Simul. 2003, 29, 291--341

    2003

  51. [51]

    Kuzmin, A.; Evarestov, R. A. Quantum mechanics-molecular dynamics approach to the interpretation of X-ray absorption spectra . J. Phys.: Condens. Matter 2009, 21, 055401

    2009

  52. [52]

    Treatment of disorder effects in X-ray absorption spectra beyond the conventional approach

    Kuzmin, A.; Timoshenko, J.; Kalinko, A.; Jonane, I.; Anspoks, A. Treatment of disorder effects in X-ray absorption spectra beyond the conventional approach . Rad. Phys. Chem. 2020, 175, 108112

    2020

  53. [53]

    I.; Rehr, J

    Zabinsky, S. I.; Rehr, J. J.; Ankudinov, A.; Albers, R. C.; Eller, M. J. Multiple-scattering calculations of X-ray-absorption spectra . Phys. Rev. B 1995, 52, 2995--3009

    1995

  54. [54]

    A practical introduction to multiple scattering theory

    Ravel, B. A practical introduction to multiple scattering theory . J. Alloys Compd. 2005, 401, 118--126

    2005

  55. [55]

    J.; Kas, J

    Rehr, J. J.; Kas, J. J.; Prange, M. P.; Sorini, A. P.; Takimoto, Y.; Vila, F. Ab initio theory and calculations of X-ray spectra . C. R. Phys. 2009, 10, 548--559

    2009

  56. [56]

    Hedin, L.; Lundqvist, B. I. Explicit local exchange-correlation potentials . J. Phys. C: Solid State Phys. 1971, 4, 2064

    1971

  57. [57]

    A beamline for bulk sample x-ray absorption spectroscopy at the high brilliance storage ring PETRA III

    Welter, E.; Chernikov, R.; Herrmann, M.; Nemausat, R. A beamline for bulk sample x-ray absorption spectroscopy at the high brilliance storage ring PETRA III . AIP Conf. Proc. 2019, 2054, 040002

    2019

  58. [58]

    XAESA v0.07

    Kalinko, A. XAESA v0.07 . 2023; https://gitlab.desy.de/aleksandr.kalinko/xaesa

    2023

  59. [59]

    EXAFS and XANES analysis of oxides at the nanoscale

    Kuzmin, A.; Chaboy, J. EXAFS and XANES analysis of oxides at the nanoscale . IUCrJ 2014, 1, 571--589

    2014

  60. [60]

    Effect of Cation Disorder on Lithium Transport in Halide Superionic Conductors

    Zhong, P.; Gupta, S.; Deng, B.; Jun, K.; Ceder, G. Effect of Cation Disorder on Lithium Transport in Halide Superionic Conductors. ACS Energy Lett. 2024, 9, 2775--2781

    2024

  61. [61]

    Beni, G.; Platzman, P. M. Temperature and polarization dependence of extended X-ray absorption fine-structure spectra . Phys. Rev. B 1976, 14, 1514--1518

    1976

  62. [62]

    Temperature dependence of the mean square relative displacements of nearest-neighbour atoms derived from EXAFS spectra

    Bohmer, W.; Rabe, P. Temperature dependence of the mean square relative displacements of nearest-neighbour atoms derived from EXAFS spectra . J. Phys. C: Solid State Physics 1979, 12, 2465

    1979

  63. [63]

    Van der Waals heterostructures and devices

    Liu, Y.; Weiss, N.; Duan, X.; Cheng, H.-C.; Huang, Y.; Duan, X. Van der Waals heterostructures and devices . Nat. Rev. Mater. 2016, 1, 16042

    2016

  64. [64]

    Interlayer coupling in two-dimensional semiconductor materials

    Shi, Z.; Wang, X.; Sun, Y.; Li, Y.; Zhang, L. Interlayer coupling in two-dimensional semiconductor materials . Semicond. Sci. Technol. 2018, 33, 093001

    2018

  65. [65]

    Recent progress in the synthesis of novel two-dimensional van der Waals materials

    Bian, R.; Li, C.; Liu, Q.; Cao, G.; Fu, Q.; Meng, P.; Zhou, J.; Liu, F.; Liu, Z. Recent progress in the synthesis of novel two-dimensional van der Waals materials . Natl. Sci. Rev. 2021, 9, nwab164

    2021

  66. [66]

    Unraveling the interlayer and intralayer coupling in two-dimensional layered MoS _2 by X-ray absorption spectroscopy and ab initio molecular dynamics simulations

    Pudza, I.; Bocharov, D.; Anspoks, A.; Krack, M.; Kalinko, A.; Welter, E.; Kuzmin, A. Unraveling the interlayer and intralayer coupling in two-dimensional layered MoS _2 by X-ray absorption spectroscopy and ab initio molecular dynamics simulations . Mater. Today Commun. 2023, 35, 106359

    2023

  67. [67]

    Temperature-dependent local structure and lattice dynamics of 1T-TiSe _2 and 1T-VSe _2 probed by X-ray absorption spectroscopy

    Pudza, I.; Polyakov, B.; Pudzs, K.; Welter, E.; Kuzmin, A. Temperature-dependent local structure and lattice dynamics of 1T-TiSe _2 and 1T-VSe _2 probed by X-ray absorption spectroscopy . Physica B 2024, 685, 415995

    2024

  68. [68]

    H.; Graf, M

    Jeong, I.-K.; Heffner, R. H.; Graf, M. J.; Billinge, S. J. L. Lattice dynamics and correlated atomic motion from the atomic pair distribution function . Phys. Rev. B 2003, 67, 104301

    2003

  69. [69]

    Advanced approach to the local structure reconstruction and theory validation on the example of the W L _3 -edge extended X-ray absorption fine structure of tungsten

    Jonane, I.; Anspoks, A.; Kuzmin, A. Advanced approach to the local structure reconstruction and theory validation on the example of the W L _3 -edge extended X-ray absorption fine structure of tungsten . Model. Simul. Mater. Sci. Eng. 2018, 26, 025004

    2018

  70. [70]

    An evolutionary many-objective optimization algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints

    Deb, K.; Jain, H. An evolutionary many-objective optimization algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints. IEEE transactions on evolutionary computation 2013, 18, 577--601

    2013

  71. [71]

    A.; Wenny, M

    Goodwin, Z. A.; Wenny, M. B.; Yang, J. H.; Cepellotti, A.; Ding, J.; Bystrom, K.; Duschatko, B. R.; Johansson, A.; Sun, L.; Batzner, S.; others Transferability and accuracy of ionic liquid simulations with equivariant machine learning interatomic potentials. J. Phys. Chem. Lett. 2024, 15, 7539--7547

    2024

  72. [72]

    u ller, D.; Steinwart, I.; K\

    Zaverkin, V.; Holzm\" u ller, D.; Steinwart, I.; K\" a stner, J. Fast and sample-efficient interatomic neural network potentials for molecules and materials based on Gaussian moments. J. Chem. Theory Comput. 2021, 17, 6658--6670

    2021

  73. [73]

    MLIP-3: Active learning on atomic environments with moment tensor potentials

    Podryabinkin, E.; Garifullin, K.; Shapeev, A.; Novikov, I. MLIP-3: Active learning on atomic environments with moment tensor potentials. J. Chem. Phys. 2023, 159

    2023

  74. [74]

    W.; Wood, B

    Qi, J.; Ko, T. W.; Wood, B. C.; Pham, T. A.; Ong, S. P. Robust training of machine learning interatomic potentials with dimensionality reduction and stratified sampling. npj Comput. Mater. 2024, 10, 43

    2024

  75. [75]

    k-Means Clustering in Fingerprint-Based Configuration Selection for Fitting Interatomic Potentials

    Lebeda, M.; Drahokoupil, J.; L \"o bel, L.; Vlčák, P. k-Means Clustering in Fingerprint-Based Configuration Selection for Fitting Interatomic Potentials. J. Chem. Theory Comput. 2024, 20, 10676--10683

    2024

  76. [76]

    F.; Romero, A

    Schmidt, J.; Cerqueira, T. F.; Romero, A. H.; Loew, A.; J \"a ger, F.; Wang, H.-C.; Botti, S.; Marques, M. A. Improving machine-learning models in materials science through large datasets. Mater. Today Phys. 2024, 48, 101560

    2024

  77. [77]

    Effect of the damping function in dispersion corrected density functional theory

    Grimme, S.; Ehrlich, S.; Goerigk, L. Effect of the damping function in dispersion corrected density functional theory. J. Comput. Chem. 2011, 32, 1456--1465

    2011

  78. [78]

    L.; DeCost, B

    Joress, H.; Ravel, B.; Anber, E.; Hollenbach, J.; Sur, D.; Hattrick-Simpers, J.; Taheri, M. L.; DeCost, B. Why is EXAFS for complex concentrated alloys so hard? Challenges and opportunities for measuring ordering with X-ray absorption spectroscopy. Matter 2023,

    2023

  79. [79]

    L.; Anber, E.; Barnett, A.; Billinge, S.; Birbilis, N.; DeCost, B.; Foley, D

    Taheri, M. L.; Anber, E.; Barnett, A.; Billinge, S.; Birbilis, N.; DeCost, B.; Foley, D. L.; Holcombe, E.; Hollenbach, J.; Joress, H.; others Understanding and leveraging short-range order in compositionally complex alloys. MRS Bulletin 2023, 48, 1280--1291

    2023

  80. [80]

    yes", ctrl-chapter-title =

    Bocharov, D.; Pudza, I.; Klementiev, K.; Krack, M.; Kuzmin, A. Study of high-temperature behaviour of ZnO by ab initio molecular dynamics simulations and X-ray absorption spectroscopy. Materials 2021, 14, 5206 mcitethebibliography achemso-v20.tex0000664000000000000000000013054315044675172012345 0ustar rootroot [journal=jctcce,manuscript=article] achemso c...

    arXiv 2021

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.