Pith. sign in

REVIEW 4 major objections 5 minor 31 references

Real-time dynamics of the two-step charge-density-wave transition in bulk 1T-TaS$_2$

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read The 200 K step of the CDW transition in 1T-TaS2 shifts Star-of-David clusters while preserving their distortion; the 350 K step melts that distortion, and the paper proves this with machine-learning molecular dynamics.

desk verdict A serious computational kinetics study with a clean mechanistic story, but the RDF data as presented seem to undermine the central claim that the 200 K transition preserves SoD amplitude; needs direct order parameter and MLFF validation before I'd trust the distinction. read the letter →

arxiv 2608.02456 v1 pith:FT7VWCKV submitted 2026-08-03 cond-mat.mtrl-sci

Daeheon Kim , Jinseok Oh , Kahyeon Koh , Yejun Cho , Angel Rubio , Noejung Park , Dongbin Shin This is my paper · ORCID
classification cond-mat.mtrl-sci PACS 71.45.Lr71.30.+h63.22.-m
keywords chargedensitywave1T-TaS2Star-of-Daviddistortionmachinelearningforcefieldmoleculardynamicsstackingorderdomainwallsmetal-insulatortransition
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper uses machine-learning-force-field molecular dynamics to watch the 200 K and 350 K transitions in bulk 1T-TaS2 in real time. It claims the two steps are mechanistically distinct: above 200 K, Star-of-David clusters translate by transiently dissolving and re-forming around shifted centers, preserving their local amplitude while randomizing interlayer stacking and nucleating domain walls. Only near 350 K does the Star-of-David distortion itself collapse. This resolves a long-standing question about whether the transitions are driven by amplitude collapse or lattice rearrangement, and it explains why the 200 K transition disappears in thin films.

What carries the argument

The central mechanism is the deformation-formation process: a low-barrier path for SoD cluster translation (108 meV, versus 207 meV for layer sliding) where the cluster's internal distortion weakens just enough for the cluster to re-center, then re-forms. The authors access this dynamics with a NequIP machine-learning force field trained on PBE-based AIMD trajectories, enabling 5-ns simulations of 1404-atom supercells that capture collective coherent breathing motions and domain-wall nucleation.

What would settle it

Measure the intensity of the sqrt13 x sqrt13 superlattice reflection and the interlayer stacking correlation in the same time-resolved diffraction experiment while heating through 200 K: if the superlattice peak weakens or the peak width broadens at 200 K, the claim that the SoD amplitude is preserved would be falsified; only a loss of stacking correlation without superlattice attenuation would support the paper's mechanism.

Watch

Extended reading notes

Core claim

The paper establishes that the commensurate-to-near-commensurate transition (200 K) is a lattice-rearrangement event: Star-of-David clusters shift coherently through a deformation-formation process in which the local displacement pattern temporarily dissolves and re-forms around a new center, leaving the SoD amplitude intact. This shifting randomizes the interlayer stacking order and creates metastable in-plane domain walls, both of which contribute to metallization. The near-commensurate-to-incommensurate transition (350 K) is instead an amplitude-collapse event, where the SoD distortion fully melts into a symmetric triangular lattice. The two transitions therefore have different order para

Load-bearing premise

The entire dynamical picture rests on the machine-learning force field faithfully reproducing the density-functional-theory potential-energy surface, especially the relative heights of the 108 meV shifting barrier and the 0.35 eV/SoD collapse barrier; if strong electron correlations alter these barriers, the mechanistic ordering of the two transitions could change.

Editorial extensions

If this is right

  • If correct, the metallization at the 200 K transition is driven primarily by randomization of interlayer stacking, not by in-plane domain walls alone, explaining why thin films (which lack out-of-plane dispersion) do not show the transition.
  • The 350 K transition is a genuine melting of the SoD distortion, independent of stacking configuration, consistent with its appearance in all sample thicknesses.
  • The transition time for SoD shifting grows exponentially with lattice size because coherence of the breathing motion becomes harder to achieve, suggesting that experimentally observed domain-wall dynamics are nucleation-limited.
  • Light-induced hidden phases should retain the intralayer SoD amplitude while losing interlayer correlations, pointing to a coherent deformation-formation process as the underlying route.
  • Cavity-modified transition temperatures can be interpreted as renormalization of the 108 meV shifting barrier rather than a change in the electronic correlation strength.

Reading between the lines

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

  • A direct experimental test could use time-resolved electron or X-ray diffraction just above 200 K: if the superlattice peak intensity stays constant while the interlayer stacking correlation decays, the deformation-formation picture is confirmed.
  • The reliance on a PBE-trained force field implies that strong-correlation renormalization of the 0.35 eV/SoD collapse barrier might shift the absolute temperature of the second transition, even though the qualitative two-step ordering is robust.
  • The exponential size dependence of the shifting time suggests that in real crystals, the CCDW-NCCDW transition might be governed by rare, locally coherent events rather than global cooperative motion, which could be probed by carefully measuring pre-transitional dynamics in nanoscale samples.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper reports AIMD and machine-learning force-field molecular dynamics simulations of bulk 1T-TaS2, using up to 1404-atom cells and 5 ns trajectories, to characterize the microscopic mechanism of the two-step CDW transition. The central claim is that the two transitions are mechanistically distinct: the CCDW-NCCDW transition at ~200 K proceeds by coherent SoD shifts via a deformation-formation process that preserves the local SoD amplitude while randomizing interlayer stacking and nucleating domain walls; the NCCDW-ICCDW transition near 350 K involves collapse of the SoD amplitude itself. Evidence includes AIMD at 100 and 175 K, NEB barriers (108 meV for deformation-formation vs 207 meV for layer sliding; 79 meV for domain wall), MLFF-MD at 200-300 K showing SoD shifts with size-dependent transition times, and RDF/DOS analyses.

Significance. If the conclusions hold, this would resolve a long-standing debate on the 1T-TaS2 transition sequence, provide a microscopic explanation of the thickness dependence of the CCDW-NCCDW transition, and give a concrete framework for light-induced hidden phases and cavity-modified transitions. The paper's use of a first-principles-trained MLFF to reach previously inaccessible time/length scales is appropriate, and the distinction between deformation-formation and sliding mechanisms via NEB is a strong point. However, the absence of MLFF validation on extended cells and the inconsistency between the claimed preserved SoD amplitude and Fig. 4e currently prevent full confidence; the central claim is plausible but not yet fully demonstrated.

major comments (4)
  1. [Thermally driven SoD deformation and the NCCDW-ICCDW transition, Fig. 4e] The central claim that the 200 K transition preserves the local SoD amplitude is difficult to reconcile with Fig. 4e, which shows the first Ta-Ta RDF peak shifting from 3.18 A at 100 K to 3.28 A at 250 K, the latter identified by the authors as the Ta-Ta distance of the undistorted triangular lattice. If the SoD short bonds were preserved, this shortest-peak position should remain near 3.18 A. The text attributes the shift to 'SoD shifts and thermal expansion,' but no direct SoD amplitude order parameter is reported to substantiate that the short-bond population survives. Please provide a quantitative mode-amplitude or bond-population analysis as a function of temperature, or reconcile the RDF shift with the proposed mechanism.
  2. [Methods: Machine-learning force field and large-scale molecular dynamics] The MLFF is the workhorse for the central dynamics claims (5 ns, 1404 atoms), yet the manuscript reports no validation of the NequIP model: no training/validation force- or energy-RMSE, no held-out test set, and no direct comparison of MLFF-MD with AIMD on the extended cells or even on the smaller 2 sqrt(13) x sqrt(13) x 2 cell. The transition times and the RDF evolution are therefore only as reliable as the surrogate. Please report held-out errors, a comparison of MLFF and AIMD observables (e.g., RDF, transition events) in the training cell, and if possible a short MLFF-vs-AIMD check on a subset of extended-cell configurations.
  3. [Results: Stacking-order randomization, Fig. 3d] The claim that the SoD transition time 'increases exponentially with lattice size and temperature' is based on what appear to be single trajectories for a few superlattice sizes and temperatures. No error bars, replicate runs, or first-passage-time statistics are provided. Given that the size dependence is used to argue about experimental observability, this quantitative claim needs at least 3-5 independent runs per condition or an explicit statement that these are single-event estimates.
  4. [Methods: DFT calculations / NEB] The MLFF and NEB barriers are obtained with a PBE-type functional without Hubbard corrections, while the electronic-structure discussion uses on-site Coulomb interactions (Ref. [18]). In a strongly correlated system like 1T-TaS2, semi-local PBE can misestimate relative barriers; the ordering of 108 meV vs 0.35 eV/SoD is load-bearing for the two-step mechanism. Please provide PBE+U (or hybrid-functional) NEB barriers for the deformation-formation and SoD-collapse paths, or otherwise demonstrate that the barrier ordering is robust to the treatment of correlations.
minor comments (5)
  1. [Abstract/Introduction] There is a typographical artifact 'T¿200 K' in the Introduction; it should be 'T > 200 K'. Please check for other non-ASCII artifacts.
  2. [Results: AIMD, Fig. 2b] The text describes single SoD-shift events at specific times (5-9 ps). Please clarify in the caption or text that these are representative single trajectories, not ensemble averages, and state how many AIMD runs were performed.
  3. [Results: Fig. 3d] The horizontal axis 'Transition time (ns)' is on a logarithmic scale; please make this explicit in the axis label or caption. Also specify whether 'exponential with lattice size' means with the number of SoDs or with the linear supercell dimension.
  4. [Methods: DFT calculations] Please state explicitly whether the VASP AIMD training data and the Quantum Espresso DFT/NEB calculations use the same functional, pseudopotential scheme, and dispersion corrections. If they differ, the MLFF training and the NEB barrier comparison may not be internally consistent.
  5. [Data availability] Provide a direct URL or accession link to the Figshare repository in addition to the DOI, for reader convenience.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the two-step dynamics emerge from MLFF simulations trained on first-principles AIMD, not from fitting the target mechanism.

full rationale

The derivation chain is self-contained: DFT (PBE+D3) is used to generate AIMD trajectories and NEB barriers; an MLFF is trained on those first-principles data; then large-scale MD is run with the MLFF and analyzed via time-dependent Ta-Ta distances, RDFs, and standard deviations. The central claim—that the CCDW–NCCDW transition is a stacking/domain-wall rearrangement preserving local SoD amplitude, while the NCCDW–ICCDW transition collapses the SoD distortion—is not used as a training label or fitting target. No experimental transition temperature is inserted as an input; 200 K and 350 K are external phase boundaries but are not parameters of the MLFF or of the mechanistic analysis. The AIMD training data do contain 175 K SoD-shifting events, so the MLFF's observation of the same process in larger cells is partly a consistency check; however, the mechanistic distinction is evaluated from RDF statistics and geometry in the 1404-atom MD, not from a fitted relation. The only self-citation, Ref. [18] (Shin et al.), is invoked for supplementary electronic-structure details with on-site Coulomb interactions; it is not load-bearing for the dynamical mechanism. Potential PBE accuracy limitations regarding correlated 1T-TaS2 are a correctness/transferability risk, not circularity. The RDF first-peak shift to 3.28 Å at 250 K noted by skeptics is an internal-consistency or interpretation issue, but it does not constitute a reduction of a prediction to an input.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical entities are proposed. The two quantitative inputs are a DFT-trained surrogate potential (a fit to first-principles data) and an empirical scaling relation inferred from four MD points. All other assumptions are standard DFT/MLFF domain assumptions.

free parameters (2)
  • MLFF/NequIP neural network weights
    Trained on >5000 AIMD (PBE) configurations at 100-500 K. The potential-energy surface of the simulated system is entirely determined by this fit; errors in it propagate to all barriers and dynamics.
  • SoD-shift time vs cell-size scaling exponent
    Exponential increase of transition time with superlattice size (Fig. 3d) is inferred from a few MD runs; used to extrapolate to experimental STM timescales, but no functional form or uncertainty is given.
assumptions (4)
  • domain assumption PBE + Grimme-D3 DFT is an adequate ground truth for the CDW potential-energy surface of 1T-TaS2 despite strong electronic correlations
    MLFF is trained exclusively on PBE AIMD; the central mechanistic distinction (amplitude-preserving shifts vs amplitude collapse) is obtained from this surface. See Methods 'DFT calculations'.
  • domain assumption Trained MLFF transfers to 1404-atom supercells and nanosecond trajectories
    No validation of MLFF against AIMD on the extended 9x2 cell is shown; transferability is assumed.
  • domain assumption AIMD training set at 100-500 K covers all relevant configurational phase space, including transition states
    The training data are short trajectories from 2-layer cells; if the transition-state region is undersampled, the MLFF barriers could be distorted.
  • domain assumption AL stacking is the low-temperature ground state
    Used as the MD initial condition; supported by prior DFT (Refs [12,13]) but not re-derived here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Real-time dynamics of the two-step charge-density-wave transition in bulk 1T-TaS$_2$." pith.science (2026). https://pith.science/paper/FT7VWCKV

@misc{pith2026260802456,
  author       = {Pith},
  title        = {Pith review of: Real-time dynamics of the two-step charge-density-wave transition in bulk 1T-TaS$_2$},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FT7VWCKV}},
  note         = {Machine review of arXiv:2608.02456}
}
read the original abstract

The charge-density wave (CDW) of bulk 1T-TaS2 is built from Star-of-David (SoD) clusters tiling a sqrt{13} x sqrt{13} superlattice, and it melts through a two-step sequence accompanied by order-of-magnitude changes in resistivity. Whether these steps proceed by collapse of the SoD amplitude or by rearrangement of the SoD lattice has remained unresolved, because the relevant dynamics occur on length and time scales beyond the reach of ab initio molecular dynamics. Here we follow the CDW transitions in real time using a machine-learning force field trained on first-principles data, giving access to 1404-atom supercells over 5 ns. The two steps are mechanistically distinct. Above 200 K, SoD clusters translate coherently by transiently dissolving and re-forming about shifted centers, a deformation-formation process that preserves the local SoD amplitude while randomizing the interlayer stacking order and nucleating domain walls. Only near 350 K does the SoD distortion itself collapse. These results provide microscopic support for the recently proposed two-step model of the CDW transition and offer a framework for interpreting light-induced hidden phases and cavity-modified transition temperatures in 1T-TaS2.

Figures

Figures reproduced from arXiv: 2608.02456 by the authors.

Figure 1
Figure 1. Temperature-dependent electrical resistivity and band structure of 1T-TaS2 depending on the stacking con￾figurations. a Temperature-dependent electronic resistivity in thin film TaS2. The data are adapted from those published in Ref. [16]. b-d Band structure of 1T-TaS2 in (b) monolayer, (c) AL stacking, and (d) L stacking configurations. Insets show schematic images for (a) 1T-TaS2 in a primitive lattice without CDW… view at source ↗
Figure 3
Figure 3. Classical MD simulation with MLFF for the ther￾mal dynamics in the extended superlattices. a-b Schematic geometry of 9√ 13 × 2 √ 13 × 2 super lattice with AL stacking for 1T-TaS2 at (a) its ground geometry and (b) thermally excited ge￾ometry with domain wall formation at 0.2 ns with thermostat at 200 K. c Time profile of the averaged nearest Ta-Ta distance in the MD with a thermostat at 200 K. d Transition time for … view at source ↗
Figure 4
Figure 4. Deformation of SoD structure achieved by the higher thermal energy. a The atomic geometry of 1T-TaS2 with SoD formation at ground-state geometry. b Radial distribu￾tion function for Ta-Ta atomic positions at 100 K. c Triangular lattice of 1T-TaS2 prior to the SoD formation. d Radial distri￾bution function for Ta-Ta atomic positions at 400 K. e Tempera￾ture dependence of the peak position of the radial distribution f… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references

  1. [18]

    Shin, D.et al.Identification of the mott insulating charge density wave state in 1T-TaS 2.Phys. Rev. Lett.126, 196406 (2021)

    2021

  2. [1]

    Mater.7, 960–965 (2008)

    Sipos, B.et al.From mott state to superconductivity in 1T-TaS2.Nat. Mater.7, 960–965 (2008)

    2008

  3. [2]

    Appl.4, 7 (2020)

    Martino, E.et al.Preferential out-of-plane conduction and quasi-one-dimensional electronic states in layered 1T-TaS2.npj 2D Mater. Appl.4, 7 (2020)

    2020

  4. [3]

    Yu, Y.et al.Gate-tunable phase transitions in thin flakes of 1T-TaS2.Nature Nanotechnology10, 270–276 (2015)

    2015

  5. [4]

    Jarc, G.et al.Cavity-mediated thermal control of metal- to-insulator transition in 1T-TaS2.Nature622, 487–492 (2023)

    2023

  6. [5]

    Fujii, D.et al.Electronic states of domain structure in 1T-TaS 2−x Sex observed by STM/STS.Journal of Physics: Conference Series969, 012041 (2018)

    2018

  7. [6]

    W., Cho, G

    Park, J. W., Cho, G. Y., Lee, J. & Yeom, H. W. Emer- gent honeycomb network of topological excitations in correlated charge density wave.Nature Communications 10, 4038 (2019)

    2019

  8. [7]

    A., Karpov, P., Vaskivskyi, I., Bra- zovskii, S

    Gerasimenko, Y. A., Karpov, P., Vaskivskyi, I., Bra- zovskii, S. & Mihailovic, D. Intertwined chiral charge orders and topological stabilization of the light-induced state of a prototypical transition metal dichalcogenide. npj Quantum Materials4, 32 (2019)

    2019

Show all 31 references
  1. [8]

    Commun.7, 10956 (2016)

    Ma, L.et al.A metallic mosaic phase and the origin of mott-insulating state in 1T-TaS 2.Nat. Commun.7, 10956 (2016)

    2016

  2. [9]

    Stojchevska, L.et al.Ultrafast Switching to a Stable Hidden Quantum State in an Electronic Crystal.Science 344, 177–180 (2014)

    2014

  3. [10]

    de la Torre, A.et al.Dynamic phase transition in 1T-TaS2 via a thermal quench.Nature Physics21, 1267–1274 (2025)

    2025

  4. [11]

    & Iwasa, Y

    Yoshida, M., Gokuden, T., Suzuki, R., Nakano, M. & Iwasa, Y. Current switching of electronic structures in two-dimensional 1T-TaS 2 crystals.Phys. Rev. B95, 121405(R) (2017)

    2017

  5. [12]

    & Geck, J

    Ritschel, T., Berger, H. & Geck, J. Stacking-driven gap formation in layered 1T-TaS2.Phys. Rev. B98, 195134 (2018)

    2018

  6. [13]

    H., Goh, J

    Lee, S. H., Goh, J. S. & Cho, D. Origin of the insulating phase and first-order metal-insulator transition in 1T- TaS2.Phys. Rev. Lett.122, 106404 (2019)

    2019

  7. [14]

    Dong, J.et al.Electronic dispersion, correlations and stacking in the photoexcited state of 1T-TaS 2.2D Ma- terials10, 045001 (2023)

    2023

  8. [15]

    J., Yoshida, M., Hanaguri, T

    Butler, C. J., Yoshida, M., Hanaguri, T. & Iwasa, Y. Mottness versus unit-cell doubling as the driver of the insulating state in 1T-TaS 2.Nat. Commun.11, 2477 (2020)

    2020

  9. [16]

    Yoshida, M.et al.Controlling charge-density-wave states in nano-thick crystals of 1T-TaS 2.Scientific Reports4, 7302 (2014)

    2014

  10. [17]

    Ritschel, T.et al.Orbital textures and charge density waves in transition metal dichalcogenides.Nat. Phys. 11, 328–331 (2015)

    2015

  11. [19]

    Wang, Y.et al.Band insulator to mott insulator tran- sition in 1T-TaS 2.Nature communications11, 4215 (2020)

    2020

  12. [20]

    Wang, Y.et al.Dualistic insulator states in 1T-TaS 2 crystals.Nature Communications15, 3425 (2024)

    2024

  13. [21]

    npj Quantum Materials7, 8 (2022)

    Zhang, W.et al.Visualizing the evolution from Mott insulator to Anderson insulator in Ti-doped 1T-TaS 2. npj Quantum Materials7, 8 (2022)

    2022

  14. [22]

    Yang, W., Kim, D., Kim, H. K. & Kim, T.-H. Thick- ness dependent charge density wave networks on thin 1T-TaS2.Journal of the Korean Physical Society82, 293–297 (2023)

    2023

  15. [23]

    Zhang, W.et al.Reconciling the bulk metallic and sur- face insulating state in 1T-TaSe 2.Physical Review B 105, 035110 (2022)

    2022

  16. [24]

    Wu, Z.et al.Effect of stacking order on the electronic state of 1T-TaS2.Physical Review B105, 035109 (2022)

    2022

  17. [25]

    W., Lee, J

    Park, J. W., Lee, J. & Yeom, H. W. Stacking and spin order in a van der Waals Mott insulator 1T-TaS 2.Com- munications Materials4, 99 (2023)

    2023

  18. [26]

    Giannozzi, P.et al.Advanced capabilities for materi- als modeling with quantum espresso.J. Phys. Condens. Matter29(2017)

    2017

  19. [27]

    P., Burke, K

    Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple.Phys. Rev. Lett. 77, 3865–3868 (1996)

    1996

  20. [28]

    & Krieg, H

    Grimme, S., Antony, J., Ehrlich, S. & Krieg, H. A con- sistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 el- ements H-Pu.The Journal of Chemical Physics132, 154104 (2010)

    2010

  21. [29]

    P.et al.LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales.Comp

    Thompson, A. P.et al.LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales.Comp. Phys. Comm.271, 108171 (2022). 6

    2022

  22. [30]

    Batzner, S.et al.E(3)-equivariant graph neural net- works for data-efficient and accurate interatomic poten- tials.Nature Communications13, 2453 (2022)

    2022

  23. [31]

    CUI: Advanced Imaging of Matter

    Kresse, G. & Furthm¨ uller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set.Phys. Rev. B54, 11169–11186 (1996). Acknowledgments Funding We acknowledge support by the Max Planck Institute New York City Center for Non-Equilibri...

    1996

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.