Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics

Fariborz Kargar; Konstantin Klyukin; Lucas Lindsay; Tom Berlijn; Tribhuwan Pandey; Valentina Nesterova

REVIEW 3 major objections 5 minor 55 references

Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A universal machine-learning interatomic potential, benchmarked only against density-functional displacement energies, reproduces the full heating–cooling sequence of charge-density-wave phases in monolayer 1T-TaS2 in 2028-atom molecular dy

desk verdict First zero-shot universal MLIP to drive a finite-T CDW transition in 1T-TaS2; the physics story is coherent, but the surrogate's error budget leaves the domain/hysteresis claims on a validation gap. read the letter →

arxiv 2607.22316 v1 pith:TMBNNAIZ submitted 2026-07-24 cond-mat.mtrl-sci

Valentina Nesterova , Tribhuwan Pandey , Tom Berlijn , Fariborz Kargar , Lucas Lindsay , Konstantin Klyukin This is my paper · ORCID

classification cond-mat.mtrl-sci

keywords chargedensitywavesmonolayer1T-TaS2universalmachinelearninginteratomicpotentialmoleculardynamicsStar-of-Daviddistortionthermalhysteresistemperature-dependenteffectiveCDWchiralitydomains

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

The paper claims that a universal machine-learning interatomic potential, benchmarked only against DFT displacement energies, is accurate enough to drive 2028-atom molecular dynamics that reproduce the full heating sequence of charge-density-wave phases in monolayer 1T-TaS2—from the low-temperature Star-of-David commensurate phase through the nearly commensurate and incommensurate states to the high-temperature undistorted hexagonal metal. It further claims that cooling from high temperature yields a thermal hysteresis of roughly 30 K and freezes the system into a multi-domain state in which the two mirror-related CDW chiralities nucleate independently and persist to low temperature. Temperature-dependent effective potential phonons extracted from the trajectories show the CDW soft modes becoming dynamically stable as temperature rises, offering a vibrational signature of the transitions. A sympathetic reader would care because these transitions control transport switching in 1T-TaS2 devices, and the result suggests that zero-shot universal potentials can access finite-temperature phenomena previously out of reach for first-principles MD.

What carries the argument

The central object is the Star-of-David (SoD) structural motif—13 Ta atoms contracting around a central Ta—used both as the physical fingerprint of CDW order and as a quantitative order parameter. The argument is carried by a geometric SoD-counting metric applied to time-averaged MD trajectories: a Ta atom is a SoD center if its six nearest neighbors all lie within 3.23 Å (temperature-adjusted), its second shell within 5.8 Å contains at least three Ta atoms, and at least three of the first-shell neighbours show the characteristic short-long bond pattern. This metric converts a 2028-atom trajectory into a number that tracks the fraction of Ta atoms in CDW clusters through the phase sequence.

What would settle it

Compute UMA-s-1p1 energies for a set of domain-wall- and NCCDW-containing configurations drawn from the MD trajectories and compare against DFT energies for the same configurations; if the mean error in these unbenchmarked environments approaches or exceeds the 0.027 eV/f.u. NM-CCDW enthalpy difference, the simulated transition sequence, hysteresis width, and multi-domain freezing cannot be trusted as intrinsic lattice physics.

Watch

Extended reading notes

Core claim

On the paper's own terms: a pretrained universal MLIP (UMA-s-1p1), selected by benchmarking against DFT displacement energies for the normal-metallic and commensurate CDW phases, is used in molecular dynamics on 2028-atom supercells of monolayer 1T-TaS2. Counting Ta atoms that belong to Star-of-David clusters yields a quantitative order parameter that drops sharply near 290 K on heating, passes through a nearly commensurate state of coexisting ordered and disordered regions, then an incommensurate-like state of transient clusters, and reaches zero in the hexagonal phase by about 500 K. Cooling reverses the transition about 30 K lower, and the low-temperature state is a multi-domain texture o

Load-bearing premise

The load-bearing premise is that UMA-s-1p1's energy surface is accurate in exactly the configurations that decide the results—intermediate NCCDW-like states, domain boundaries, and the nucleation events during heating and cooling—even though the published benchmark samples only small random displacements from the two end phases.

Editorial extensions

If this is right

A single universal potential can resolve the experimentally known CDW phase sequence in 1T-TaS2 at 2028-atom scale without material-specific training.
The computed ~290 K heating transition and ~30 K hysteresis match phenomenological observations, suggesting such simulations can locate approximate transition temperatures in CDW monolayers.
The persistence of α/β chiral domains on cooling provides a structural explanation for the suppressed long-range CCDW order reported in monolayer crystals.
TDEP phonons show that the high-temperature hexagonal phase is stabilized by anharmonic fluctuations, not by static harmonic stability.
The geometric SoD-counting metric can serve as a general order parameter for mixed CDW states in other TMDC monolayers.

Reading between the lines

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

If the potential's accuracy generalizes to unbenchmarked domain-wall and mixed-state configurations, the same approach could be applied to other strongly coupled CDW materials (e.g., 1T-TaSe2, NbTe2) to map finite-temperature phase diagrams at near-first-principles cost.
The independence of α and β nucleation suggests a route to study chirality-controlled domains as bit-like objects; strain or defect pinning, which the paper lists as beyond scope, might be used to engineer domain sizes in simulations.
A direct test: run the same heating-cooling protocol with a second universal MLIP that passes the same flat-benchmark; if the hysteresis width or multi-domain persistence changes qualitatively, the current results may be potential-specific rather than material-intrinsic.
The SoD-counting metric could be applied to experimental diffuse scattering or STM topographies to infer domain fractions, connecting simulation order parameters to measurements.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Worth your time. This is the first paper in its cited set to take a universal, zero-shot MLIP (UMA-s-1p1, no fine-tuning) and run 2028-atom MD through the 1T-TaS2 CDW sequence: SoD order melts near 290 K on heating, an NCCDW-like coexistence appears, the hexagonal phase is restored by ~500 K, cooling shows ~30 K hysteresis, and the layer freezes into persistent α/β chiral domains. That is a genuinely new demonstration — earlier MLIP-CDW studies on NbSe2 used system-specific training or nonthermal tuning. The paper also does a lot right: it benchmarks two uMLIPs against PBE displacement energies, uses independent metrics (Ta–Ta g(r) splitting, SoD fraction, TDEP soft-mode stabilization) that tell a consistent story, deposits analysis scripts and trajectory videos on Figshare, and states its limitations about quantitative Tc, cutoff radii, and missing SOC/magnetism.

The central weakness is the error budget. The benchmark reports MAE 0.048 eV/f.u. on random displacements from the two endpoint structures, while the NM–CCDW enthalpy difference is 0.027 eV/f.u. So the surrogate's stated error exceeds the energy scale that encodes relative phase stability. The MD results that matter — the NCCDW-like boundaries, the ~30 K hysteresis, the α/β domain walls that freeze in — come from configurations that are never benchmarked: partially formed SoD clusters, interdomain compressed bonds, transient ICDW-like motifs. That is a validation gap, not an internal inconsistency. The authors reproduce ΔH for the relaxed endpoints, which is encouraging, but two endpoints do not certify the energy surface along the transition paths. Second, the hysteresis and domain-multiplication narrative rests on single heating and cooling trajectories with no error bars; the SoD order parameter is a hand-thresholded geometric count, not an electronic order parameter. Third, the abstract's phrase 'reproduce the experimentally observed phase transition sequence' is stronger than the authors' own limitation statement warrants, especially since monolayer-specific experimental Tc values are not pinned down.

None of this kills the paper. The internal logic is coherent, the authors are honest, and the qualitative physics is plausible. But it is a proof-of-concept, not a definitive quantitative study.

Who should read it: anyone working on MLIPs for correlated TMDCs, or on finite-T CDW simulations. It deserves serious peer review. The referee should push on the error-budget issue and ask for at least one repeated trajectory or an electronic-structure check of a few domain-wall snapshots. I'd send it to review with expectations of heavy revision.

Referee Report

3 major / 5 minor

Summary. The paper benchmarks two universal machine-learning interatomic potentials (MACE-MP-0 and UMA-s-1p1) against PBE and PBE+U displacement energies for monolayer 1T-TaS2, selects UMA-s-1p1, and then uses it for 2028-atom molecular dynamics simulations. On heating from the commensurate Star-of-David (SoD) phase, the authors observe a sharp loss of SoD order near 290 K, with an intermediate nearly commensurate CDW-like state, followed by an incommensurate-like regime of transient clusters and finally the undistorted hexagonal phase near 500 K. On cooling, the transition occurs near 260 K, giving a ~30 K hysteresis; the system freezes into a multi-domain state with α and β chiral domains. Temperature-dependent effective potential (TDEP) phonon calculations from the MD trajectories show progressive stabilization of soft modes with temperature. The central claim is that a zero-shot universal MLIP, benchmarked only against DFT displacement energies, reproduces the experimentally observed CDW phase transition sequence, including the intermediate coexistence, hysteresis, and multi-domain freezing.

Significance. If correct, this is a valuable demonstration that universal MLIPs can capture finite-temperature lattice transformations in a correlated CDW material at a scale inaccessible to direct DFT-MD. The paper's strengths include explicit DFT benchmarking, a quantitative structural order parameter for SoD order, 2028-atom MD simulations, TDEP phonon analysis, and public availability of analysis scripts via Figshare. The zero-shot aspect is notable and would extend recent system-specific MLIP studies of NbSe2. However, the central claim depends on the surrogate's accuracy in configurations that are not benchmarked, and the reported energy error is comparable to the phase-stability energy scale. The qualitative domain-multiplication and hysteresis conclusions require additional validation or a more cautious framing.

major comments (3)

[Section III, Fig. 2; Section II.B] The benchmark that justifies using UMA-s-1p1 samples only random displacements of 0.01–0.15 Å from the relaxed NM and CCDW reference structures. The mean absolute error reported for UMA-s-1p1, 0.048 eV/f.u., is larger than the NM–CCDW enthalpy difference reported in the same section, ΔH = 0.0272 eV/f.u., which sets the phase-stability scale. The intermediate NCCDW-like states, α/β domain walls, interdomain regions with compressed ~3.2 Å bonds, and transient ICDW-like clusters shown in Fig. 4 and Figs. S7–S8 are never benchmarked against DFT. If the MLIP error in those unbenchmarked environments is comparable to the domain-wall or coexistence energies, the 290 K melting, ~30 K hysteresis, and persistent multi-domain state could be artifacts of the surrogate rather than physical. The paper's limitation statement about quantitative transition temperatures does not cover these qualitative co
[Section II.C] The SoD order parameter is purely geometric: it classifies Ta atoms based on Ta–Ta distances and coordination shells. The paper repeatedly refers to 'CDW' phases, but no electronic charge modulation is measured in the MD simulations. The geometric SoD pattern may not uniquely correspond to the electronic CDW order, especially at finite temperature and at domain boundaries where the classification uses an adaptive tolerance and accepts partially formed patterns. To support the claim that the simulations reproduce CDW phase transitions, the relation between the geometric metric and the actual charge-density wave should be established, e.g., by computing electronic structure or charge localization on representative snapshots, or by reframing the conclusions as referring to the lattice distortion only.
[Section II.C, Fig. 4] The hysteresis and multi-domain freezing are central conclusions, but the protocol uses 50 ps equilibration at each temperature and sequentially initializes each temperature from the previous final frame. No convergence tests with longer equilibration, independent initial conditions, or different supercell sizes are reported. The persistent multi-domain state at low temperature could result from slow kinetics in the 2028-atom cell rather than from a thermodynamically stable or long-lived physical state. The paper does acknowledge that the multi-domain structure persists 'at the cooling conditions considered here,' but this caveat does not fully address the concern because the claim is presented as reproducing experimental observations. Additional tests of equilibration time and initial-condition dependence would materially strengthen the hysteresis and domain-persistence claims.

minor comments (5)

[Global] The model name is spelled inconsistently ('UMA-s-1p1' vs. 'UMA s-1p1'). Please standardize.
[Abstract / Fig. 4] The abstract and Fig. 4 use 'pink line' and 'blue line' for heating and cooling; many readers may see grayscale printouts. Use distinct line styles or symbols as well.
[Section III] The phrase 'part of Ta atoms' should be 'fraction of Ta atoms' or similar.
[Section II.D] For TDEP phonons, the text states IFCs are extracted for the 'NM cell structure' from MD trajectories. At low temperatures the simulated system is in the CCDW state; the procedure for mapping those trajectories onto the NM cell should be described more explicitly to avoid ambiguity.
[References] Reference [32] is cited as arXiv:2506.23971v2; please update to the published or final version if available. Reference [20] appears to be a journal article; please provide complete details.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the transition sequence, hysteresis, and domain structures are emergent outputs of a pretrained universal MLIP benchmarked against DFT, not fitted inputs.

full rationale

The paper's central results (temperature-dependent SoD fraction, CCDW-to-NCCDW-to-ICDW-to-NM sequence, ~30 K hysteresis, persistent α/β multi-domains, and TDEP soft-mode stabilization) are generated by MD using UMA-s-1p1, a universal MLIP trained on OMat-24 by independent authors before this study. The only fitting-like step is benchmarking UMA-s-1p1 against DFT displacement energies for NM and CCDW structures, but that benchmark does not use the claimed output quantities as targets; the MD outcomes are emergent from dynamics. The SoD-counting metric is a structural diagnostic defined from the DFT-relaxed CCDW geometry, not a parameter fitted to reproduce the transition or hysteresis; it merely labels configurations that the MD trajectories produce. No load-bearing self-citation or uniqueness argument is invoked: chiral-domain references and TDEP citations are external, and the chosen potential is justified by direct DFT comparison. The main substantive concern is validation coverage — the benchmark does not sample NCCDW/domain-wall configurations that dominate the qualitative claims — but that is an accuracy/robustness risk, not a circularity. Therefore the score is 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on (i) the surrogate potential's transfer accuracy, which is benchmarked on a narrow set of displaced structures with an MAE exceeding the phase-stability enthalpy difference, and (ii) the geometric identification of CDW order from Ta positions alone. No new physical entities are introduced; the α/β chiral variants and Star-of-David clusters are known entities from prior literature. The free parameters are the hand-chosen thresholds of the SoD classifier plus the externally fitted weights of the universal potential.

free parameters (5)

SoD 1NN short-bond threshold = 3.23 Å
Hand-chosen cutoff below which Ta–Ta bonds count as SoD short bonds; based on the DFT-relaxed CCDW reference geometry (Methods II.C).
SoD 2NN detection radius = 5.8 Å
Second coordination shell radius used to count neighbors in the SoD classifier; hand-chosen (Methods II.C).
SoD motif acceptance count = 3 of 6 atoms
Requiring at least three (not all six) atoms to show the short/long-bond motif lets the classifier admit partial clusters at domain boundaries; hand-chosen to match the authors' picture of grain boundaries (Methods II.C).
SoD tolerance rule (RMSD-adaptive) = mean Ta RMSD over trajectory window
Bond-distance tolerances adapt to thermal RMSD; the choice of averaging window (15 ps) and the adaptive rule affect how many SoDs are detected at each temperature.
UMA-s-1p1 model parameters = trained on OMat-24 (external, not fit here)
The surrogate's weights are fitted to a large DFT dataset. The central claim depends on this fitted model's transfer accuracy; the model was not fine-tuned for TaS2.

assumptions (5)

domain assumption PBE without Hubbard U and without spin polarization gives an adequate energy landscape for the lattice dynamics simulated here.
The MLIPs are trained on no-U, nonmagnetic DFT; the paper checks robustness with U = 2.5 eV and spin polarization for selected structures (Figs. S3, S5), but all production MD runs operate on the no-U, nonmagnetic surface (Section III).
domain assumption UMA-s-1p1 generalizes from OMat-24 training data to the 1T-TaS2 thermal ensemble, including unbenchmarked domain-wall and NCCDW configurations.
The benchmark dataset covers only ±0.01–0.15 Å displacements of the NM and CCDW reference cells (Section II.B); the intermediate mixed-phase and domain-wall configurations that drive the central results are not in the benchmark set.
domain assumption Classical Langevin MD with 1 fs timestep and 50–100 ps equilibration captures the relevant anharmonic lattice dynamics; nuclear quantum effects are negligible above 60 K.
Standard practice for this class of simulation but not justified in the text; relevant to the reported 290/260 K transition temperatures and to the TDEP mode-stabilization conclusion (Section II.C).
domain assumption The geometric Star-of-David pattern is an adequate order parameter for the electronic CDW phases (CCDW/NCCDW/ICDW/NM).
The MLIP carries no electronic degrees of freedom; phase identity is assigned from Ta–Ta distances only (Methods II.C), while the experimentally defined phases are electronic and lattice CDW states.
domain assumption The 2028-atom supercell is large enough that finite-size effects do not control the observed transition temperatures, hysteresis width, or domain sizes.
The cell was chosen to accommodate both α and β chiralities (Fig. S6); no systematic supercell-size convergence study for the MD results is reported.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics." pith.science (2026). https://pith.science/paper/TMBNNAIZ

@misc{pith2026260722316,
  author       = {Pith},
  title        = {Pith review of: Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TMBNNAIZ}},
  note         = {Machine review of arXiv:2607.22316}
}

read the original abstract

Charge-density-wave (CDW) phases in 1T transition-metal dichalcogenides arise from strong electron-phonon coupling and accompanying lattice instabilities. Capturing their temperature-dependent structural evolution using conventional first-principles molecular dynamics (MD) remains challenging because of the large supercells and extensive finite-temperature sampling required. Here, we combine density functional theory (DFT), universal machine-learning interatomic potentials (MLIPs), MD, and temperature-dependent effective potential phonon calculations to investigate the structural and vibrational signatures of CDW transitions in monolayer 1T-TaS2. Benchmarking against DFT displacement energies identifies UMA-s-1p1 universal machine learning potentials with sufficient accuracy for subsequent finite-temperature simulations. Our results show that large-scale MD simulations reproduce the experimentally observed phase transition sequence from the low-temperature Star-of-David (SoD) distorted structure to the high-temperature primitive hexagonal structure, as quantified by the number of Ta atoms attributed to SoDs. Heating-cooling cycles exhibit thermal hysteresis, and upon cooling, the system freezes into a multi-domain state in which {\alpha} and \b{eta} CDW chiralities nucleate independently and persist to the lowest temperatures. These findings demonstrate that carefully benchmarked universal MLIPs can provide a scalable framework for finite-temperature studies of CDW materials.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 1 canonical work pages

[1]

Star‐of‐David

*Corresponding authors: lindsaylr@ornl.gov and klyukin@auburn.edu Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics Valentina Nesterova,1 Tribhuwan Pandey2, Tom Berlijn3,4, Fariborz Kargar,1 Lucas Lindsay3*, Konstantin Klyukin1* 1 Department of Mechanical and Materials Engineering, Auburn Univers...

2020
[2]

Yoshida, R

M. Yoshida, R. Suzuki, Y. Zhang, M. Nakano, and Y. Iwasa, Memristive phase switching in two-dimensional 1T-TaS2 crystals, Science Advances 1, e1500606 (2015)

2015
[3]

Manzeli, D

S. Manzeli, D. Ovchinnikov, D. Pasquier, O. V. Yazyev, and A. Kis, 2D transition metal dichalcogenides, Nat Rev Mater 2, 17033 (2017)

2017
[4]

Nie et al., Tailoring exciton dynamics of monolayer transition metal dichalcogenides by interfacial electron-phonon coupling, Commun Phys 2, 103 (2019)

Z. Nie et al., Tailoring exciton dynamics of monolayer transition metal dichalcogenides by interfacial electron-phonon coupling, Commun Phys 2, 103 (2019)

2019
[5]

S. Yin, K. He, B. Barut, M. D. Randle, R. Dixit, J. Nathawat, D. Adinehloo, V. Perebeinos, J. E. Han, and J. P. Bird, Real-Time Observation of Slowed Charge Density Wave Dynamics in Thinned 1T-TaS2, Advanced Physics Research 3, 2400033 (2024)

2024
[6]

for the NM and CCDW phases (Figure 2). Atomic displacements ranging from 0.01 to 0.15 Å, that correspond to the temperature fluctuations near phase transition region (Figure S4), are used to sample the local energy landscape associated with thermally accessible vibrational distortions. MACE-MP (Figure 2a) systematically underestimates displacement energie...
[7]

G. Liu, B. Debnath, T. R. Pope, T. T. Salguero, R. K. Lake, and A. A. Balandin, A charge-density-wave oscillator based on an integrated tantalum disulfide-boron nitride-graphene device operating at room temperature, Nat Nanotechnol 11, 845 (2016)

2016
[8]

W. Liu, Z. Duan, C. Zhang, X. X. Hu, J. B. Cao, L. J. Liu, and L. Lin, Experimental observations and density functional simulations on the structural transition behavior of a two-dimensional transition-metal dichalcogenide, Sci Rep 10, 18255 (2020)

2020

Show all 55 references

[9]

Salgado, A

R. Salgado, A. Mohammadzadeh, F. Kargar, A. Geremew, C.-Y. Huang, M. A. Bloodgood, S. Rumyantsev, T. T. Salguero, and A. A. Balandin, Low-frequency noise spectroscopy of charge-density-wave phase transitions in vertical quasi-2D 1T-TaS2 devices, Appl. Phys. Express 12, 037001 (2019)

2019
[10]

J.-A. Yan, M. A. D. Cruz, B. Cook, and K. Varga, Structural, electronic and vibrational properties of few-layer 2H- and 1T-TaSe2, Sci Rep 5, 16646 (2015)

2015
[11]

Pasquier and O

D. Pasquier and O. V. Yazyev, Ab initio theory of magnetism in two-dimensional 1T-TaS2, Phys. Rev. B 105, L081106 (2022)

2022
[12]

Lazar, J

P. Lazar, J. Martincová, and M. Otyepka, Structure, dynamical stability, and electronic properties of phases in TaS2 from a high-level quantum mechanical calculation, Phys. Rev. B 92, 224104 (2015)

2015
[13]

Mohammadzadeh, A

A. Mohammadzadeh, A. Rehman, F. Kargar, S. Rumyantsev, J. M. Smulko, W. Knap, R. K. Lake, and A. A. Balandin, Room temperature depinning of the charge-density waves in quasi-two-dimensional 1T-TaS2 devices, Appl. Phys. Lett. 118, 223101 (2021)

2021
[14]

Khitun, G

A. Khitun, G. Liu, and A. A. Balandin, Two-Dimensional Oscillatory Neural Network Based on Room-Temperature Charge-Density-Wave Devices, IEEE Transactions on Nanotechnology 16, 860 (2017)

2017
[15]

W. Wang, D. Dietzel, and A. Schirmeisen, Lattice Discontinuities of 1T-TaS2 across First Order Charge Density Wave Phase Transitions, Sci Rep 9, 7066 (2019)

2019
[16]

Husain, R

S. Husain, R. Gupta, P. Kumar, N. Behera, R. Brucas, S. Chaudhary, A. Kumar, and P. Svedlindh, Probing Charge Density Wave Effects in 1T-TaS2 Monolayer/Ni81Fe19 Heterostructure: A Spin Dynamics Approach, ACS Appl. Electron. Mater. 3, 3321 (2021)

2021
[17]

Gan, L.-H

L.-Y. Gan, L.-H. Zhang, Q. Zhang, C.-S. Guo, U. Schwingenschlögl, and Y. Zhao, Strain tuning of the charge density wave in monolayer and bilayer 1T-TaS2, Physical Chemistry Chemical Physics 18, 3080 (2016)

2016
[18]

A. K. Geremew, S. Rumyantsev, F. Kargar, B. Debnath, A. Nosek, M. A. Bloodgood, M. Bockrath, T. T. Salguero, R. K. Lake, and A. A. Balandin, Bias-Voltage Driven Switching of the Charge-Density-Wave and Normal Metallic Phases in 1T-TaS2 Thin-Film Devices, ACS Nano 13, 7231 (2019)

2019
[19]

Barani, T

Z. Barani, T. Geremew, M. Stokey, N. Sesing, M. Taheri, M. J. Hilfiker, F. Kargar, M. Schubert, T. T. Salguero, and A. A. Balandin, Quantum Composites with Charge-Density-Wave Fillers, Advanced Materials 35, 2209708 (2023)

2023
[20]

Rivano et al., Exploring charge density waves in two-dimensional NbSe2 with machine learning, Npj Comput Mater (2026)

N. Rivano et al., Exploring charge density waves in two-dimensional NbSe2 with machine learning, Npj Comput Mater (2026)

2026
[21]

Benić, F

L. Benić, F. Grasselli, C. Ben Mahmoud, D. Novko, and I. Lončarić, Machine Learning Model for Efficient Nonthermal Tuning of the Charge Density Wave in Monolayer NbSe2, J. Chem. Theory Comput. 21, 8130 (2025)

2025
[22]

Y. D. Wang, W. L. Yao, Z. M. Xin, T. T. Han, Z. G. Wang, L. Chen, C. Cai, Y. Li, and Y. Zhang, Band insulator to Mott insulator transition in 1T-TaS2, Nat Commun 11, 4215 (2020)

2020
[23]

Kresse and J

G. Kresse and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996)

1996
[24]

Kresse and J

G. Kresse and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996)

1996
[25]

Nakata et al., Robust charge-density wave strengthened by electron correlations in monolayer 1T-TaSe2 and 1T-NbSe2, Nat Commun 12, 5873 (2021)

Y. Nakata et al., Robust charge-density wave strengthened by electron correlations in monolayer 1T-TaSe2 and 1T-NbSe2, Nat Commun 12, 5873 (2021)

2021
[26]

J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Phys. Rev. Lett. 77, 3865 (1996)

1996
[27]

S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P. Sutton, Electron-energy-loss spectra and the structural stability of nickel oxide: An LSDA+U study, Phys. Rev. B 57, 1505 (1998)

1998
[28]

Wang et al., Dualistic insulator states in 1T-TaS2 crystals, Nat Commun 15, 3425 (2024)

Y. Wang et al., Dualistic insulator states in 1T-TaS2 crystals, Nat Commun 15, 3425 (2024)

2024
[29]

Darancet, A

P. Darancet, A. J. Millis, and C. A. Marianetti, Three-dimensional metallic and two-dimensional insulating behavior in octahedral tantalum dichalcogenides, Phys. Rev. B 90, 045134 (2014)

2014
[30]

Hafner, Ab-initio simulations of materials using VASP: Density-functional theory and beyond, Journal of Computational Chemistry 29, 2044 (2008)

J. Hafner, Ab-initio simulations of materials using VASP: Density-functional theory and beyond, Journal of Computational Chemistry 29, 2044 (2008)
[31]

D. P. Kovács, I. Batatia, E. S. Arany, and G. Csányi, Evaluation of the MACE force field architecture: From medicinal chemistry to materials science, J. Chem. Phys. 159, 044118 (2023)

2023
[32]

B. M. Wood et al., UMA: A Family of Universal Models for Atoms, https://arxiv.org/abs/2506.23971v2
[33]

The UMA s-1p1 model was selected for all subsequent molecular dynamics simulations based on its overall accuracy

Both uMLIPs were benchmarked here against DFT PBE with both U = 0 and U = 2.5 eV, and the relative energies were computed on a dataset of structures generated by applying random displacements in the range 0.01-0.15 Å to both the high temperature NM and the low temperature CCDW...
[34]

Cooling and heating simulations used to demonstrate hysteresis behavior were performed as a series of consecutive MD calculations for these 2028 atom cells

with a timestep of 1 fs. Cooling and heating simulations used to demonstrate hysteresis behavior were performed as a series of consecutive MD calculations for these 2028 atom cells. For cooling, the initial temperature was set to 500 K, for heating – 60 K. Each subsequent temp...
[35]

At each simulated temperature, the Ta atomic positions were averaged over the last 15 ps of simulation to obtain a representative averaged distribution of atoms

Star of David (SoD) patterns were identified during the analysis of MD trajectories of the large 2028 atom supercells, using Ta–Ta distances from the DFT-relaxed CCDW cell as the reference for the ideal SoD geometry. At each simulated temperature, the Ta atomic positions were ...
[36]

Temperature-Dependent Effective Potential (TDEP) for Phonons

D. Temperature-Dependent Effective Potential (TDEP) for Phonons. Temperature-dependent phonon dispersions were extracted from MD trajectories using the TDEP method. [37,38] The last 15 ps were used for the 1T-TaS2 system with 2028 atom supercells, every 10th snapshot was used....
[37]

Above the transition temperature, the dispersion resembles that of the conventional metallic phase, with no unstable modes remaining in the primitive Brillouin zone

At low temperatures (blue curves), several phonon branches exhibit pronounced softening and become imaginary (shown as negative values in Figure 5), indicating the dynamical instability of the NM phase toward CDW distortion. Above the transition temperature, the dispersion res...
[38]

P. E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50, 17953 (1994)

1994
[39]

Togo, First-principles Phonon Calculations with Phonopy and Phono3py, J

A. Togo, First-principles Phonon Calculations with Phonopy and Phono3py, J. Phys. Soc. Jpn. 92, 012001 (2023)

2023
[40]

Cavignac et al., AI-Driven expansion and application of the Alexandria database, J

T. Cavignac et al., AI-Driven expansion and application of the Alexandria database, J. Phys. Mater. 9, 025014 (2026)

2026
[41]

Barroso-Luque, M

L. Barroso-Luque, M. Shuaibi, X. Fu, B. M. Wood, M. Dzamba, M. Gao, A. Rizvi, C. L. Zitnick, and Z. W. Ulissi, Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models, arXiv:2410.12771

2024 arXiv
[42]

J. A. Izaguirre, D. P. Catarello, J. M. Wozniak, and R. D. Skeel, Langevin stabilization of molecular dynamics, J. Chem. Phys. 114, 2090 (2001)
[43]

Hjorth Larsen et al., The atomic simulation environment—a Python library for working with atoms, J

A. Hjorth Larsen et al., The atomic simulation environment—a Python library for working with atoms, J. Phys.: Condens. Matter 29, 273002 (2017)

2017
[44]

The Ta–Ta bond-length distributions provide additional insight into the local structure of the multidomain state

The interlayer coupling [44,45] and defect-mediated pinning [13,46] may substantially alter domain nucleation and phase stability in thin 1T-TaS2 films, but these effects are beyond the scope of the present study. The Ta–Ta bond-length distributions provide additional insight ...
[45]

Valentina Nesterova, Tribhuwan Pandey, Tom Berlijn, Fariborz Kargar, Lucas Lindsay, and Konstantin Klyukin, Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 - Supplemental Materials at Figshare, https://doi.org/10.6084/m9.figshare.33040211

doi
[46]

Knoop et al., TDEP:Temperature Dependent Effective Potentials, Journal of Open Source Software 9, (2024)

F. Knoop et al., TDEP:Temperature Dependent Effective Potentials, Journal of Open Source Software 9, (2024)

2024
[47]

Hellman, P

O. Hellman, P. Steneteg, I. A. Abrikosov, and S. I. Simak, Temperature dependent effective potential method for accurate free energy calculations of solids, Phys. Rev. B 87, 104111 (2013)

2013
[48]

Jain et al., Commentary: The Materials Project: A materials genome approach to accelerating materials innovation, APL Mater

A. Jain et al., Commentary: The Materials Project: A materials genome approach to accelerating materials innovation, APL Mater. 1, 011002 (2013)

2013
[49]

M. M. Ghahremanpour, P. J. van Maaren, and D. van der Spoel, The Alexandria library, a quantum-chemical database of molecular properties for force field development, Sci Data 5, 180062 (2018)

2018
[50]

Y. A. Gerasimenko, P. Karpov, I. Vaskivskyi, S. Brazovskii, and D. Mihailovic, Intertwined chiral charge orders and topological stabilization of the light-induced state of a prototypical transition metal dichalcogenide, Npj Quantum Mater. 4, 32 (2019)

2019
[51]

Zhao et al., Spectroscopic visualization and phase manipulation of chiral charge density waves in 1T-TaS2, Nat Commun 14, 2223 (2023)

Y. Zhao et al., Spectroscopic visualization and phase manipulation of chiral charge density waves in 1T-TaS2, Nat Commun 14, 2223 (2023)

2023
[52]

Yoshida, Y

M. Yoshida, Y. Zhang, J. Ye, R. Suzuki, Y. Imai, S. Kimura, A. Fujiwara, and Y. Iwasa, Controlling charge-density-wave states in nano-thick crystals of 1T-TaS2, Sci Rep 4, 7302 (2014)

2014
[53]

Stahl, M

Q. Stahl, M. Kusch, F. Heinsch, G. Garbarino, N. Kretzschmar, K. Hanff, K. Rossnagel, J. Geck, and T. Ritschel, Collapse of layer dimerization in the photo-induced hidden state of 1T-TaS2, Nat Commun 11, 1247 (2020)

2020
[54]

J. O. Brown, M. Taheri, F. Kargar, R. Salgado, T. Geremew, S. Rumyantsev, R. K. Lake, and A. A. Balandin, Current fluctuations and domain depinning in quasi-two-dimensional charge-density-wave 1T-TaS2 thin films, Appl. Phys. Rev. 10, 041401 (2023)

2023
[55]

T. J. Boerner, S. Deems, T. R. Furlani, S. L. Knuth, and J. Towns, ACCESS: Advancing Innovation: NSF’s Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support, in Practice and Experience in Advanced Research Computing 2023: Computing for the Common Good (Associ...

2023

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Star‐of‐David

*Corresponding authors: lindsaylr@ornl.gov and klyukin@auburn.edu Charge-Density-Wave Phase Transitions in Monolayer 1T-TaS2 from Universal Machine Learning Molecular Dynamics Valentina Nesterova,1 Tribhuwan Pandey2, Tom Berlijn3,4, Fariborz Kargar,1 Lucas Lindsay3*, Konstantin Klyukin1* 1 Department of Mechanical and Materials Engineering, Auburn Univers...

2020

[2] [2]

Yoshida, R

M. Yoshida, R. Suzuki, Y. Zhang, M. Nakano, and Y. Iwasa, Memristive phase switching in two-dimensional 1T-TaS2 crystals, Science Advances 1, e1500606 (2015)

2015

[3] [3]

Manzeli, D

S. Manzeli, D. Ovchinnikov, D. Pasquier, O. V. Yazyev, and A. Kis, 2D transition metal dichalcogenides, Nat Rev Mater 2, 17033 (2017)

2017

[4] [4]

Nie et al., Tailoring exciton dynamics of monolayer transition metal dichalcogenides by interfacial electron-phonon coupling, Commun Phys 2, 103 (2019)

Z. Nie et al., Tailoring exciton dynamics of monolayer transition metal dichalcogenides by interfacial electron-phonon coupling, Commun Phys 2, 103 (2019)

2019

[5] [5]

S. Yin, K. He, B. Barut, M. D. Randle, R. Dixit, J. Nathawat, D. Adinehloo, V. Perebeinos, J. E. Han, and J. P. Bird, Real-Time Observation of Slowed Charge Density Wave Dynamics in Thinned 1T-TaS2, Advanced Physics Research 3, 2400033 (2024)

2024

[6] [6]

for the NM and CCDW phases (Figure 2). Atomic displacements ranging from 0.01 to 0.15 Å, that correspond to the temperature fluctuations near phase transition region (Figure S4), are used to sample the local energy landscape associated with thermally accessible vibrational distortions. MACE-MP (Figure 2a) systematically underestimates displacement energie...

[7] [7]

G. Liu, B. Debnath, T. R. Pope, T. T. Salguero, R. K. Lake, and A. A. Balandin, A charge-density-wave oscillator based on an integrated tantalum disulfide-boron nitride-graphene device operating at room temperature, Nat Nanotechnol 11, 845 (2016)

2016

[8] [8]

W. Liu, Z. Duan, C. Zhang, X. X. Hu, J. B. Cao, L. J. Liu, and L. Lin, Experimental observations and density functional simulations on the structural transition behavior of a two-dimensional transition-metal dichalcogenide, Sci Rep 10, 18255 (2020)

2020