Pith. sign in

REVIEW 4 major objections 5 minor 114 references

Diverse polymorphs and phase transitions in van der Waals In$_2$Se$_3$

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A machine-learning potential trained on 5,262 DFT structures maps In2Se3's polymorph landscape, resolving the β″ structure and the phase diagram.

desk verdict Solid MLP study of In2Se3 with genuinely new polymorph structures and a strain-driven transition, but the phase diagram and transition temperatures rest on unquantified MLP extrapolation and kinetic MD, so they should be presented as semi-quantitative predictions rather than first-principles results. read the letter →

arxiv 2506.21248 v1 pith:3LQSO5FZ submitted 2025-06-26 cond-mat.mtrl-sci

Mingfeng Liu , Jiantao Wang , Peitao Liu , Qiang Wang , Zhibo Liu , Yan Sun , Xing-Qiu Chen This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords In2Se3vanderWaalslayeredmaterialsmachine-learningpotentialpolymorphicphasetransitionsferroelectric–antiferroelectrictemperature–pressurediagrammoleculardynamicsMexican-hatenergysurface
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 machine-learning potential trained on 5,262 density-functional-theory structures can map the polymorphic energy landscape of van der Waals In2Se3 accurately enough to resolve an 11 meV per formula unit energy difference between competing β″ structures, uncover previously unknown β′ superstructure variants, and simulate phase transitions over nanosecond timescales. If true, this yields the first ab initio-quality temperature–pressure phase diagram for In2Se3 and a strain-based route to switch between antiferroelectric and nonpolar phases. The broader payoff would be atomic-scale understanding of how nucleation and domain-wall motion drive transitions among layered ferroic phases.

What carries the argument

The central object is the moment tensor machine-learning potential (MLP), a fitted interatomic potential that reproduces DFT energies, forces, and stresses for In2Se3 and runs molecular dynamics on supercells of tens of thousands of atoms. The mechanism it exposes is the Mexican-hat-like potential energy surface of the β phase: the central Se atoms sit on a circular energy ridge with many shallow minima, so freezing the flat imaginary phonon branches at different wavevectors yields the family of β′ nanostripe superstructures and the β″ phase. Thermodynamic integration and reversible scaling turn MLP-mapped free energies into the temperature–pressure phase diagram.

What would settle it

Directly recompute with DFT the total energies of the proposed Type-II and Type-I β″ structures and the reported β′ superstructure variants using the same exchange-correlation functional; if Type-II is not lower by roughly 11 meV per formula unit, or if the β′ variants are not dynamically stable in the MLP phonon spectra, the claim fails. Experimentally, measure the β′↔β″ transition under controlled uniaxial strain in few-layer In2Se3 and compare the critical strains and polarization changes to the MD predictions.

Watch

Extended reading notes

Core claim

The central discovery is that In2Se3's β polymorph is a high-symmetry parent phase whose central Se atoms sit on a Mexican-hat-like potential energy surface, so freezing its flat imaginary phonon modes at different wavevectors generates the family of β′ nanostripe superstructures and the β″ phase. The paper identifies a β″-1T structure (Type-II) that is 11 meV per formula unit lower in energy than the previously proposed Type-I structure, supports this with simulated STM images that better match one set of experiments, and reports several previously unreported β′ superstructure variants across the 1T, 2H, and 3R polytypes. Using MLP-driven molecular dynamics, it observes α→β→β′→β″ transformations, constructs a temperature–pressure phase diagram via thermodynamic integration, and finds reversible β′↔β″ transitions under uniaxial strain with accompanying polarization changes.

Load-bearing premise

The machine-learning potential remains accurate outside the configuration space it was trained on, down to energy differences of about 11 meV per formula unit and the free-energy differences that set transition temperatures.

Editorial extensions

If this is right

  • The MLP places the β″-2H, β″-3R, and β″-1T Type-II structures as dynamically stable phases, with Type-II 11 meV/f.u. more stable than the previously proposed model.
  • Several β′ nanostripe superstructures (2d-2d, 3d-3d, 4d-4d, 5d-5d, and mixed variants) are dynamically stable and nearly degenerate in energy, explaining observed coexistence of ferroelectric and antiferroelectric contrast in experiments.
  • The computed temperature–pressure phase diagram gives α stable at low temperature and pressure, β at high temperature and pressure, and β′ in between, with the α→β boundary at about 683 K at ambient pressure and about 0.9 GPa at 300 K, both consistent with experimental data.
  • Molecular dynamics shows that all these phase transitions proceed by in-plane nucleation followed by layer-by-layer growth along the c axis, including the formation and motion of 0°, 60°, and 120° domain walls.
  • Uniaxial strain reversibly converts β′ to β″ (a-axis strains of 3.81–4.02%) and back (b-axis strains near 0.59%), with ferroelectric polarization appearing and disappearing during the transitions.
  • The strain-induced β′↔β″ transition decouples phase switching from electric fields, pointing toward mechanically programmable ferroelectric devices.

Reading between the lines

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

  • If the MLP extrapolates as claimed, other layered III2-VI3 compounds with flat imaginary phonon bands and Mexican-hat-like surfaces should show similarly large families of closely spaced metastable polar and antiferroelectric superstructures.
  • The strain-induced β′↔β″ transition suggests a mechanical-writing scheme for ferroelectric memories without applied electric fields; the paper leaves device implementation unexplored.
  • Because the validation set covers only a fraction of the phase space represented by the training dataset, an independent check of transition barriers and nucleation rates (for example, comparing MD-derived kinetics against experimental hysteresis) would be needed before relying on the predicted transition temperatures quantitatively.
  • The coexistence of Type-I and Type-II β″ structures in different STM experiments hints that both may appear depending on growth or quench history, a hypothesis testable by controlled cooling-rate experiments.
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 develops a moment tensor machine-learning potential (MTP-MLP) for bulk van der Waals In2Se3, trained by active learning on 5,262 DFT structures, and validates it against lattice parameters, energy differences, energy-volume curves, and phonon dispersions. Using this MLP, the authors identify a Type-II structure for the beta'' polymorph that is 11 meV/f.u. lower in energy than the literature Type-I model, report numerous new beta' superstructure variants with dynamic stability, and interpret them through flat imaginary phonon bands and a Mexican-hat-like potential energy surface. Large-scale MLP molecular dynamics is used to observe alpha-to-beta, beta-to-beta', and beta'-to-beta'' transitions, to construct a temperature-pressure phase diagram by thermodynamic integration and reversible scaling supplemented by direct MD, and to predict a reversible strain-induced beta'-to-beta'' transition with accompanying polarization modulation. The manuscript concludes with proposals for strain-programmable electronic devices based on these transitions.

Significance. If the MLP is accurate in the regions where it is applied, the paper is significant: it unifies many experimentally observed polymorphs through a common Mexican-hat PES, provides a candidate corrected structure for the beta'' phase, reports an ab initio temperature-pressure phase diagram, and predicts a strain-driven reversible transition that could be exploited in devices. The paper has real strengths: multi-pronged validation against DFT (lattice parameters, energies, phonons, STM simulations), reproduction of experimentally known transitions, inclusion of additional InSe structures to improve transferability, and use of thermodynamic integration rather than direct fitting of transition temperatures. However, the load-bearing premise is extrapolative accuracy of the MLP in configurations not well represented by the validation set. The reported validation metrics characterize interpolation on a subset of the training manifold, and energy differences among competing variants are only a few meV/f.u.; without DFT checks of the specific new structures, strained states, and transition trajectories, the new phases and phase boundaries could be MLP artifacts.

major comments (4)
  1. [Development of MLP; Supplementary Fig. 1a] The validation set is explicitly described as covering only a fraction of the phase space represented by the training dataset, and its RMSEs are lower than the training RMSEs. These metrics therefore demonstrate interpolation accuracy on the training manifold, not extrapolation accuracy to unsampled configurations. Since the central discoveries (the beta'' Type-II structure 11 meV/f.u. below Type-I, the new beta' variants with meV-scale energy differences, and the 3-4% strained states) live precisely in such unsampled regions, the paper should supply DFT calculations for these specific structures and for representative strained MD configurations, or an uncertainty metric such as the active-learning extrapolation grade along the MD trajectories. As written, the claim that these phases and transitions are not MLP artifacts is not fully supported.
  2. [Methods, 'Thermodynamic phase diagram calculations'] The beta-beta' boundary in Fig. 2c is not obtained by thermodynamic integration or reversible scaling; the Methods state that this transition is determined from five heating and five cooling MD runs in 20 K steps with 100 ps equilibration and 1 ns production per step. This is a finite-rate kinetic estimate, not an equilibrium free-energy boundary, and the near-overlap of heating and cooling curves in Fig. 4a does not by itself remove hysteresis. Please report the spread over independent seeds and, ideally, compute the free-energy difference with an order-parameter-based thermodynamic integration scheme that can handle the interchangeable nature of the two phases.
  3. [Thermodynamic phase diagram, Fig. 2c] The quantitative agreement with experiment is weaker than the text suggests: the ambient-pressure alpha-to-beta transition temperature is computed as 683 K versus the cited experimental 473 K, and the later beta'-to-beta'' estimate is 277 K versus the cited 180 K. The phrase 'showing good agreement with experiments' should be qualified, and convergence checks (system size, heating rate, finite-size corrections) should be reported. If the overestimates are attributed to the PBE-D3 functional or to kinetic effects, this should be stated explicitly in the text.
  4. [Dynamical polymorphic phase transitions, Fig. 6 and Supplementary Figs. 13-14] The strain-induced beta'-to-beta'' transition is simulated at a strain rate of 0.01 ns^-1 with critical strains of 3.81-4.02% in one direction and 0.59% in the reverse direction, and the associated polarization modulation is computed entirely from MLP trajectories. No DFT benchmark is reported for strained cells at these strains. Because the competing variants are separated by only a few meV/f.u., representative strained configurations at the endpoints and at the nucleation strain should be revalidated by DFT before the strain-driven transition and the polarization-switching claims are accepted as quantitative predictions.
minor comments (5)
  1. [Introduction] The phrase 'pays the way for the design' should read 'paves the way for the design'.
  2. [Dynamical polymorphic phase transitions, text accompanying Fig. 6] The sentence describing the reverse transition says 'the reverse beta'' to beta'' phase transition'; this should read 'beta'' to beta' phase transition'.
  3. [Affiliation and Code availability] There are small typographical errors in the text: 'T echnology' in the affiliation, 'MLlP' in the code availability section, and inconsistent spacing in 'V ASP'; these should be corrected.
  4. [Supplementary Fig. 12 caption] The heating rate is 0.039 K/ps and the cooling rate is 0.39 K/ps, differing by an order of magnitude; the main text should state both rates when quoting the averaged beta'-beta'' transition temperature of 277 K, since the asymmetry likely affects the averaged value.
  5. [Fig. 2 caption and axes] Fig. 2a uses absolute energies (approximately -19.3 eV/f.u.) while Fig. 2b uses enthalpies relative to alpha phase; for consistency, the captions should explicitly define the reference state for each panel.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the MLP is trained on DFT energies/forces, the phase diagram is computed by thermodynamic integration rather than fitted to transition temperatures, and newly identified polymorphs are revalidated with DFT and compared with external experiments.

full rationale

The derivation chain is not circular. The MLP is fitted to DFT energies, forces, and stresses for 5,262 sampled structures, and the central predictions (polymorph energy ordering, phonon stability, phase boundaries, strain-induced transitions) are outputs of that model rather than quantities that were themselves used as fitting labels. The temperature-pressure phase diagram is obtained from thermodynamic integration and reversible scaling between a harmonic reference Hamiltonian and the MLP Hamiltonian (Methods, Eqs. 1-6), so the reported phase boundaries are computed free-energy differences, not re-fitted transition temperatures; the β-β′ boundary uses direct heating/cooling MD as a kinetic estimate, but this is not a fit to the experimental transition points. The newly proposed β″ Type-II structure and the β′ variants are re-evaluated with independent DFT total energies and phonon calculations (Supplementary Figs. 4, 6, 7), and the claimed 11 meV/f.u. energy lowering is a DFT result rather than an MLP-only claim. External benchmarks include experimental lattice parameters, transition temperatures, STM images, and domain-wall geometries. Self-citations to prior MLP methodology (Refs. 71-73, 80) and to the authors' previous HAADF-STEM stacking-fault observation (Ref. 85) provide methodological context or experimental comparison data and are not load-bearing circular arguments. The one substantive caveat—that the validation set covers only part of the training phase space and the MLP may extrapolate imperfectly to the strained configurations in Fig. 6—is an accuracy and extrapolation risk, not a circularity by construction, and therefore does not raise the circularity score under the specified criteria.

Assumptions & free parameters 1 free parameters · 4 assumptions · 3 invented entities

The central claims rest on a fitted MLP, not on a first-principles derivation. The MLP coefficients are the dominant free parameter; the DFT-PBE-D3 functional is the assumed ground truth. New metastable phases and the beta'' Type-II structure are introduced as candidate realizations of the polymorphic phase space, with partial external support from STM comparisons and phonon stability. The absence of released training data and potential weights makes it impossible to independently audit the fitted surface.

free parameters (1)
  • Moment tensor potential coefficients
    The MLP is fitted to 5,262 DFT training structures; all predicted energies, phase boundaries, and transition mechanisms depend on this fit. The coefficients are not provided in the preprint.
assumptions (4)
  • domain assumption DFT-PBE-D3 energies, forces, and stresses are the ground truth for the MLP and for all validation comparisons.
    The MLP is trained and tested against PBE-D3 data (Methods, First-principles calculations); any DFT functional error propagates into the predictions.
  • domain assumption The MLP accurately represents the potential energy surface in regions not directly sampled by the training set, including energy differences as small as 11 meV/f.u.
    The validity of the predicted beta'' Type-II structure and the phase boundaries relies on the MLP's extrapolative accuracy; the validation set covers only a fraction of the training phase space.
  • standard math The thermodynamic integration and reversible scaling methods yield converged free energies for the solid phases.
    These are established statistical mechanics methods (Refs. 83-84); the paper uses them with 10^6 integration steps and harmonic references (Methods).
  • domain assumption Harmonic phonon calculations at 0 K, including the flat imaginary modes of beta-2H, correctly indicate the existence of metastable polymorphs.
    Used to motivate and verify the stability of newly identified phases (Fig. 1b and Supplementary Figs. 4, 6).
invented entities (3)
  • beta''-2H Type-II structure (lower energy by 11 meV/f.u. than literature Type-I) independent evidence
    purpose: To resolve the atomic structure of the experimentally observed beta'' polymorph.
    Dynamic stability from phonons (Supplementary Fig. 7k-m), DFT energy comparison, and agreement with the STM image of Chen et al. (Supplementary Fig. 7j) provide evidence outside the training set.
  • Previously unreported beta' superstructure variants (e.g., 2d-3d/2d-3d, 3d-4d/3d-4d, 4d-5d/4d-5d, etc.) independent evidence
    purpose: To explain the variety of nanostripe orderings in beta' In2Se3.
    Predicted periodic lengths match experimental values for several variants (Supplementary Table 2), and the variants are dynamically stable (Supplementary Fig. 4).
  • Chirally-ordered metastable phase of central Se atoms
    purpose: To illustrate the diversity of metastable states on the Mexican-hat PES.
    No experimental observation or external falsifiable prediction is provided; it is a simulation prediction only.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diverse polymorphs and phase transitions in van der Waals In$_2$Se$_3$." pith.science (2026). https://pith.science/paper/3LQSO5FZ

@misc{pith2026250621248,
  author       = {Pith},
  title        = {Pith review of: Diverse polymorphs and phase transitions in van der Waals In$_2$Se$_3$},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3LQSO5FZ}},
  note         = {Machine review of arXiv:2506.21248}
}
abstract

Van der Waals In$_2$Se$_3$ has garnered significant attention due to its unique properties and wide applications associated with its rich polymorphs and polymorphic phase transitions. Despite extensive studies, the vast complex polymorphic phase space remains largely unexplored, and the underlying microscopic mechanism for their phase transformations remains elusive. Here, we develop a highly accurate, efficient, and reliable machine-learning potential (MLP), which not only facilitates accurate exploration of the intricate potential energy surface (PES), but also enables us to conduct large-scale molecular dynamics (MD) simulations with first-principles accuracy. We identify the accurate structure of the $\beta''$ polymorph and uncover several previously unreported $\beta'$ polymorph variants exhibiting dynamic stability and competing energies, which are elucidated by characteristic flat imaginary phonon bands and the distinctive Mexican-hat-like PES in the $\beta$ polymorph. Through the MLP-accelerated MD simulations, we directly observe the polymorphic phase transformations among the $\alpha$, $\beta$, $\beta'$, and $\beta''$ polymorphs under varying temperature and pressure conditions, and build for the first time an ab initio temperature-pressure phase diagram, showing good agreement with experiments. Furthermore, our MD simulations reveal a novel strain-induced reversible phase transition between the $\beta'$ and $\beta''$ polymorphs. This work not only unveils diverse polymorphs in van der Waals In$_2$Se$_3$, but also provides crucial atomic insights into their phase transitions, opening new avenues for the design of novel functional electronic devices.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

114 extracted references · 79 canonical work pages

  1. [1]

    Ding, W. et al. Prediction of intrinsic two-dimensional ferro- electrics in In 2Se3 and other III 2-VI3 van der Waals materials. Nature Communications 8, 14956 (2017)

    2017

  2. [2]

    Zhou, Y . et al. Out-of-plane piezoelectricity and ferroelectric- ity in layered α -In2Se3 nanoflakes. Nano Letters 17, 5508–5513 (2017)

    2017

  3. [3]

    Xiao, J. et al. Intrinsic two-dimensional ferroelectricity with dipole locking. Phys. Rev. Lett. 120, 227601 (2018)

    2018

  4. [4]

    Cui, C. et al. Intercorrelated in-plane and out-of-plane ferro- electricity in ultrathin two-dimensional layered semiconduc- tor In2Se3. Nano Letters 18, 1253–1258 (2018)

    2018

  5. [5]

    Poh, S. M. et al. Molecular-beam epitaxy of two-dimensional In2Se3 and its giant electroresistance switching in ferroresis- tive memory junction. Nano Letters 18, 6340–6346 (2018)

    2018

  6. [6]

    Xue, F. et al. Room-temperature ferroelectricity in hexagonally layered α -In2Se3 nanoflakes down to the monolayer limit. Ad- vanced Functional Materials 28, 1803738 (2018)

    2018

  7. [7]

    Io, W. F. et al. Temperature- and thickness-dependence of ro- bust out-of-plane ferroelectricity in CVD grown ultrathin van der Waals α -In2Se3 layers. Nano Research 13, 1897–1902 (2020)

    2020

  8. [8]

    Lv , B. et al. Layer-dependent ferroelectricity in 2H-stacked few-layer α -In2Se3. Mater. Horiz. 8, 1472–1480 (2021)

    2021

Show all 114 references
  1. [9]

    Bai, L. et al. Intrinsic ferroelectric switching in two- dimensional α -In2Se3. ACS Nano 18, 26103–26114 (2024)

    2024

  2. [10]

    Zheng, C. et al. Room temperature in-plane ferroelectricity in van der Waals In 2Se3. Science Advances 4, eaar7720 (2018)

    2018

  3. [12]

    Xu, C. et al. Two-dimensional ferroelasticity in van der Waals β ’-In2Se3. Nature Communications 12, 3665 (2021)

    2021

  4. [13]

    & Seidel, J

    Zhang, D., Schoenherr, P ., Sharma, P . & Seidel, J. Ferroelec- tric order in van der Waals layered materials. Nature Reviews Materials 8, 25–40 (2023)

    2023

  5. [14]

    & Zhou, J

    Wu, Z., Liu, K., Mu, X. & Zhou, J. Renormalizing antiferroelec- tric nanostripes in β -In2Se3 via optomechanics. The Journal of Physical Chemistry Letters 14, 677–684 (2023)

    2023

  6. [15]

    Spellberg, J. L. et al. Electronic structure orientation as a map of in-plane antiferroelectricity in β -In2Se3. Science Advances 10, eado2136 (2024)

    2024

  7. [16]

    Wang, L. et al. In-plane ferrielectric order in van der Waals β -In2Se3. ACS Nano 18, 809–818 (2024)

    2024

  8. [17]

    Li, S. et al. Van der Waals ferroelectrics: Theories, materi- als, and device applications. Advanced Materials 36, 2301472 (2024)

    2024

  9. [18]

    Sreekumar, R. et al. Different phases of indium selenide pre- pared by annealing In/Se bilayer at various temperatures: Characterization studies. Solar Energy Materials and Solar Cells 90, 2908–2917 (2006)

    2006

  10. [19]

    Küpers, M. et al. Controlled crystal growth of indium selenide, In2Se3, and the crystal structures of α -In2Se3. Inorganic Chem- istry 57, 11775–11781 (2018)

    2018

  11. [20]

    Liu, L. et al. Atomically resolving polymorphs and crystal structures of In 2Se3. Chemistry of Materials 31, 10143–10149 (2019)

    2019

  12. [21]

    & Wang, Z

    Li, J., Li, H., Niu, X. & Wang, Z. Low-dimensional In 2Se3 com- pounds: From material preparations to device applications. ACS Nano 15, 18683–18707 (2021)

    2021

  13. [22]

    Huang, Y .-T. et al. Two-dimensional In 2Se3: A rising ad- vanced material for ferroelectric data storage. InfoMat 4, e12341 (2022)

    2022

  14. [23]

    Han, W. et al. Recent advances of phase transition and fer- roelectric device in two-dimensional In 2Se3. Applied Physics Reviews 11, 021314 (2024)

    2024

  15. [24]

    Tan, C. K. Y ., Fu, W. & Loh, K. P . Polymorphism and ferroelec- tricity in indium(III) selenide. Chemical Reviews 123, 8701–8717 (2023)

    2023

  16. [25]

    Ullah, K., Li, Q., Li, T. & Gu, T. Melting-free integrated pho- tonic memory with layered polymorphs. Nanophotonics 13, 2089–2099 (2024)

    2024

  17. [26]

    & Amelinckx, S

    van Landuyt, J., van Tendeloo, G. & Amelinckx, S. Phase tran- sitions in In 2Se3 as studied by electron microscopy and elec- tron diffraction. physica status solidi (a) 30, 299–314 (1975)

    1975

  18. [27]

    New results on the phase transformations of In2Se3

    Manolikas, C. New results on the phase transformations of In2Se3. Journal of Solid State Chemistry 74, 319–328 (1988)

    1988

  19. [28]

    Y ., Sigeo Soeda, S

    Jiping Ye, J. Y ., Sigeo Soeda, S. S., Yoshio Nakamura, Y . N. & Osamu Nittono, O. N. Crystal structures and phase transfor- mation in In2Se3 compound semiconductor. Japanese Journal of Applied Physics 37, 4264 (1998)

    1998

  20. [29]

    Tao, X. & Gu, Y . Crystalline–crystalline phase transformation in two-dimensional In 2Se3 thin layers. Nano Letters 13, 3501– 3505 (2013)

    2013

  21. [30]

    Zhang, F. et al. Atomic-scale observation of reversible ther- mally driven phase transformation in 2D In 2Se3. ACS Nano 13, 8004–8011 (2019). 10

    2019

  22. [31]

    Lyu, F. et al. Temperature-driven α -β phase transformation and enhanced electronic property of 2H α -In2Se3. ACS Applied Materials & Interfaces 14, 23637–23644 (2022)

    2022

  23. [32]

    Wu, J. et al. Reversible thermally driven phase change of lay- ered In 2Se3 for integrated photonics. Nano Letters 23, 6440– 6448 (2023)

    2023

  24. [33]

    Igo, J., Gabel, M., Yu, Z.-G., Yang, L. & Gu, Y . Photodefined in- plane heterostructures in two-dimensional In 2Se3 nanolayers for ultrathin photodiodes. ACS Applied Nano Materials 2, 6774– 6782 (2019)

    2019

  25. [34]

    Li, T. et al. Structural phase transitions between layered in- dium selenide for integrated photonic memory . Advanced Ma- terials 34, 2108261 (2022)

    2022

  26. [35]

    & Zhou, Y

    Wan, S., Peng, Q., Wu, Z. & Zhou, Y . Nonvolatile ferroelec- tric memory with lateral β /α /β In2Se3 heterojunctions. ACS Applied Materials & Interfaces 14, 25693–25700 (2022)

    2022

  27. [36]

    Guo, J. et al. Femtosecond laser manipulation of multistage phase switching in two-dimensional In 2Se3 visualized via an in situ transmission electron microscope. ACS Nano 19, 13264– 13272 (2025)

    2025

  28. [38]

    Zhang, Z. et al. Atomic visualization and switching of ferro- electric order in β -In2Se3 films at the single layer limit. Ad- vanced Materials 34, 2106951 (2022)

    2022

  29. [39]

    Zhang, F. et al. Atomic-scale manipulation of polar domain boundaries in monolayer ferroelectric In 2Se3. Nature Commu- nications 15, 718 (2024)

    2024

  30. [40]

    Wu, Y . et al. Stacking selected polarization switching and phase transition in vdw ferroelectric α -In2Se3 junction de- vices. Nature Communications 15, 10481 (2024)

    2024

  31. [41]

    Zhang, J. et al. Interlayer reconstruction phase transition in van der waals materials. Nature Materials 24, 369–376 (2025)

    2025

  32. [42]

    M., Teklemichael, S

    Rasmussen, A. M., Teklemichael, S. T., Mafi, E., Gu, Y . & Mc- Cluskey , M. D. Pressure-induced phase transformation of In2Se3. Applied Physics Letters 102, 062105 (2013)

    2013

  33. [43]

    Ke, F. et al. Interlayer-glide-driven isosymmetric phase transi- tion in compressed In 2Se3. Applied Physics Letters 104, 212102 (2014)

    2014

  34. [44]

    & Yang, L

    Zhao, J. & Yang, L. Structure evolutions and metallic tran- sitions in In 2Se3 under high pressure. The Journal of Physical Chemistry C 118, 5445–5452 (2014)

    2014

  35. [45]

    Vilaplana, R. et al. Experimental and theoretical studies on α -In2Se3 at high pressure. Inorganic Chemistry 57, 8241–8252 (2018)

    2018

  36. [46]

    Tang, L. et al. Giant piezoresistivity in a van der Waals mate- rial induced by intralayer atomic motions. Nature Communica- tions 14, 1519 (2023)

    2023

  37. [47]

    Dong, J. et al. Influence of van der Waals epitaxy on phase transformation behaviors in 2D heterostructure. Applied Physics Letters 116, 021602 (2020)

    2020

  38. [48]

    Zheng, X. et al. Phase and polarization modulation in two-dimensional In 2Se3 via in situ transmission electron mi- croscopy .Science Advances 8, eabo0773 (2022)

    2022

  39. [49]

    Han, W. et al. Phase-controllable large-area two-dimensional In2Se3 and ferroelectric heterophase junction. Nature Nan- otechnology 18, 55–63 (2023)

    2023

  40. [50]

    & Loh, K

    Chen, Z., Sun, M., Li, H., Huang, B. & Loh, K. P . Oscillatory order–disorder transition during layer-by-layer growth of in- dium selenide. Nano Letters 23, 1077–1084 (2023)

    2023

  41. [51]

    & Zou, J

    Han, G., Chen, Z.-G., Drennan, J. & Zou, J. Indium selenides: Structural characteristics, synthesis and their thermoelectric performances. Small 10, 2747–2765 (2014)

    2014

  42. [52]

    Wang, Q. et al. Phase-defined van der Waals schottky junc- tions with significantly enhanced thermoelectric properties. The Journal of Physical Chemistry Letters 8, 2887–2894 (2017)

    2017

  43. [53]

    Lin, M. et al. Controlled growth of atomically thin In 2Se3 flakes by van der waals epitaxy . Journal of the American Chemi- cal Society 135, 13274–13277 (2013)

    2013

  44. [54]

    Zhai, T. et al. Fabrication of high-quality In 2Se3 nanowire arrays toward high-performance visible-light photodetectors. ACS Nano 4, 1596–1602 (2010)

    2010

  45. [55]

    L., Li, Y ., Gao, J., Wang, S

    Li, Q. L., Li, Y ., Gao, J., Wang, S. D. & Sun, X. H. High perfor- mance single In 2Se3 nanowire photodetector. Applied Physics Letters 99, 243105 (2011)

    2011

  46. [56]

    Xue, F. et al. Optoelectronic ferroelectric domain-wall memo- ries made from a single van der Waals ferroelectric. Advanced Functional Materials 30, 2004206 (2020)

    2020

  47. [57]

    Yang, J. et al. Ultrasensitive ferroelectric semiconductor pho- totransistors for photon-level detection. Advanced Functional Materials 32, 2205468 (2022)

    2022

  48. [58]

    Wang, S. et al. Two-dimensional ferroelectric channel transis- tors integrating ultra-fast memory and neural computing. Na- ture Communications 12, 53 (2021)

    2021

  49. [59]

    Liu, K. et al. An optoelectronic synapse based on α -In2Se3 with controllable temporal dynamics for multimode and multiscale reservoir computing. Nature Electronics 5, 761–773 (2022)

    2022

  50. [60]

    T., Meister, S., Zhang, X

    Peng, H., Schoen, D. T., Meister, S., Zhang, X. F. & Cui, Y . Synthesis and phase transformation of In 2Se3 and CuInSe 2 nanowires. Journal of the American Chemical Society 129, 34–35 (2007)

    2007

  51. [61]

    Li, W. et al. In2Se3, In 2Te3, and In 2(Se,Te)3 alloys as photo- voltaic materials. The Journal of Physical Chemistry Letters 13, 12026–12031 (2022)

    2022

  52. [62]

    & Tran, L

    Lee, H., Kang, D.-H. & Tran, L. Indium selenide (In 2Se3) thin film for phase-change memory . Materials Science and Engineer- ing: B 119, 196–201 (2005)

    2005

  53. [63]

    Yu, B. et al. Indium selenide nanowire phase-change memory . Applied Physics Letters 91, 133119 (2007)

    2007

  54. [64]

    Huang, Y .-T. et al. Dynamic observation of phase transforma- tion behaviors in indium(III) selenide nanowire based phase change memory .ACS Nano 8, 9457–9462 (2014). 11

    2014

  55. [65]

    Choi, M. S. et al. Electrically driven reversible phase changes in layered In 2Se3 crystalline film. Advanced Materials 29, 1703568 (2017)

    2017

  56. [66]

    Huang, Y .-T. et al. Complex charge density waves in simple electronic systems of two-dimensional III 2-VI3 materials. Na- ture Communications 15, 9983 (2024)

    2024

  57. [67]

    Huang, Y .-T. et al. Mexican-hat potential energy surface in two-dimensional III2-VI3 materials and the importance of en- tropy barrier in ultrafast reversible ferroelectric phase change. Applied Physics Reviews 8, 031413 (2021)

    2021

  58. [68]

    & Pantelides, S

    Liu, J. & Pantelides, S. T. Pyroelectric response and temperature-induced α -β phase transitions in α -In2Se3 and other α -III2VI3 (III=Al, Ga, In; VI=S, Se) monolayers. 2D Ma- terials 6, 025001 (2019)

    2019

  59. [69]

    & Pourfath, M

    Soleimani, M. & Pourfath, M. Ferroelectricity and phase tran- sitions in In 2Se3 van der Waals material. Nanoscale 12, 22688– 22697 (2020)

    2020

  60. [70]

    Wu, J. et al. Accurate force field of two-dimensional ferro- electrics from deep learning. Phys. Rev. B 104, 174107 (2021)

    2021

  61. [71]

    Liu, P . et al. Combining machine learning and many-body cal- culations: coverage-dependent adsorption of CO on Rh(111). Physical Review Letters 130, 078001 (2023)

    2023

  62. [72]

    Liu, M. et al. Layer-by-layer phase transformation in Ti3O5 re- vealed by machine-learning molecular dynamics simulations. Nature Communications 15, 3079 (2024)

    2024

  63. [73]

    Cao, Y . et al. Quantum delocalization enables water dissocia- tion on Ru(0001). arXiv.2412.00484 (2024)

    2024 arXiv

  64. [74]

    & Bok- dam, M

    Jinnouchi, R., Lahnsteiner, J., Karsai, F., Kresse, G. & Bok- dam, M. Phase transitions of hybrid perovskites simulated by machine-learning force fields trained on the fly with bayesian inference. Physical Review Letters 122, 225701 (2019)

    2019

  65. [75]

    & Kresse, G

    Jinnouchi, R., Karsai, F. & Kresse, G. On-the-fly machine learn- ing force field generation: application to melting points. Phys- ical Review B 100, 014105 (2019)

    2019

  66. [76]

    Shapeev , A. V . Moment tensor potentials: a class of systemat- ically improvable interatomic potentials. Multiscale Modeling & Simulation 14, 1153–1173 (2016)

    2016

  67. [77]

    Podryabinkin, E. V . & Shapeev , A. V . Active learning of lin- early parametrized interatomic potentials. Computational Ma- terials Science 140, 171–180 (2017)

    2017

  68. [78]

    S., Gubaev , K., Podryabinkin, E

    Novikov , I. S., Gubaev , K., Podryabinkin, E. V . & Shapeev , A. V . The MLIP package: moment tensor potentials with MPI and active learning. Machine Learning: Science and T echnology 2, 025002 (2021)

    2021

  69. [79]

    Cheng, B. et al. Mapping materials and molecules. Accounts of Chemical Research 53, 1981–1991 (2020)

    2020

  70. [80]

    Wang, J. et al. Efficient moment tensor machine-learning in- teratomic potential for accurate description of defects in ni-al alloys. 2411.01282 (2024)

    2024 arXiv

  71. [81]

    Soft modes and structural phase transi- tions

    V enkataraman, G. Soft modes and structural phase transi- tions. Bulletin of Materials Science 1, 129–170 (1979)

    1979

  72. [82]

    Pallikara, I., Kayastha, P ., Skelton, J. M. & Whalley , L. D. The physical significance of imaginary phonon modes in crystals. Electronic Structure 4, 033002 (2022)

    2022

  73. [83]

    Understanding Molecular Simulation (Elsevier, 2002)

    2002

  74. [84]

    & Yip, S

    de Koning, M., Antonelli, A. & Yip, S. Optimized Free- Energy Evaluation Using a Single Reversible-Scaling Simula- tion. Phys. Rev. Lett. 83, 3973–3977 (1999)

    1999

  75. [86]

    Felton, J. et al. Probing and manipulating the Mexican hat- shaped valence band of In 2Se3. Nature Communications 16, 922 (2025)

    2025

  76. [87]

    Zhou, S., Tao, X. & Gu, Y . Thickness-dependent thermal con- ductivity of suspended two-dimensional single-crystal In 2Se3 layers grown by chemical vapor deposition. The Journal of Physical Chemistry C 120, 4753–4758 (2016)

    2016

  77. [88]

    & Liu, S

    Zhou, W. & Liu, S. Two-dimensional ferroelectric crys- tal with temperature-invariant ultralow thermal conductivity . arXiv:2501.09990 (2025)

    2025 arXiv

  78. [89]

    Li, W. et al. Large disparity between optical and fundamen- tal band gaps in layered In 2Se3. Physical Review B 98, 165134 (2018)

    2018

  79. [90]

    & Furthmüller, J

    Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical Review B 54, 11169–11186 (1996)

    1996

  80. [91]

    P ., Burke, K

    Perdew, J. P ., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Physical Review Letters 77, 3865– 3868 (1996)

    1996

  81. [92]

    Blöchl, P . E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994)

    1994

  82. [93]

    & Joubert, D

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

    1999

  83. [94]

    & Krieg, H

    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. The Journal of Chemical Physics 132, 154104 (2010)

    2010

  84. [95]

    & Tanaka, I

    Togo, A. & Tanaka, I. First principles phonon calculations in materials science. Scripta Materialia 108, 1–5 (2015)

    2015

  85. [96]

    Sheppard, D., Xiao, P ., Chemelewski, W., Johnson, D. D. & Henkelman, G. A generalized solid-state nudged elastic band method. The Journal of Chemical Physics 136, 074103 (2012)

    2012

  86. [97]

    https://github.com/orest-d/p4vasp

  87. [98]

    & Kresse, G

    Jinnouchi, R., Karsai, F., V erdi, C., Asahi, R. & Kresse, G. Descriptors representing two- and three-body atomic distri- butions and their effects on the accuracy of machine-learned inter-atomic potentials. J. Chem. Phys. 152, 234102 (2020)

    2020

  88. [99]

    P ., Kondor, R

    Bartók, A. P ., Kondor, R. & Csányi, G. On representing chem- ical environments. Physical Review B 87, 184115 (2013). 12

    2013

  89. [100]

    Thompson, A. P . et al. Lammps-a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Physics Communications 271, 108171 (2022)

    2022

  90. [101]

    Hoover, W. G. Canonical dynamics: equilibrium phase-space distributions. Physical Review A 31, 1695–1697 (1985)

    1985

  91. [102]

    Constant temperature molecular dynamics meth- ods

    Shuichi, N. Constant temperature molecular dynamics meth- ods. Progress of Theoretical Physics Supplement 103, 1–46 (1991)

    1991

  92. [103]

    & Rahman, A

    Parrinello, M. & Rahman, A. Polymorphic transitions in single crystals: a new molecular dynamics method. Journal of Applied Physics 52, 7182–7190 (1981)

    1981

  93. [104]

    Visualization and analysis of atomistic simula- tion data with OVITO-the Open Visualization Tool

    Stukowski, A. Visualization and analysis of atomistic simula- tion data with OVITO-the Open Visualization Tool. Modelling Simul. Mater. Sci. Eng. 18, 015012 (2010)

    2010

  94. [105]

    Monacelli, L. et al. The stochastic self-consistent harmonic ap- proximation: Calculating vibrational properties of materials with full quantum and anharmonic effects. J. Phys.: Condens. Matter 33, 363001 (2021)

    2021

  95. [106]

    & Antonelli, A

    De Koning, M. & Antonelli, A. Einstein crystal as a reference system in free energy estimation using adiabatic switching. Phys. Rev. E 53, 465–474 (1996)

    1996

  96. [107]

    M., Trizac, E., Pronk, S

    Polson, J. M., Trizac, E., Pronk, S. & Frenkel, D. Finite-size corrections to the free energies of crystalline solids. The Journal of Chemical Physics 112, 5339–5342 (2000)

    2000

  97. [108]

    Diverse polymorphs and phase transitions in van der Waals In 2Se3

    Cheng, B. & Ceriotti, M. Computing the absolute Gibbs free energy in atomistic simulations: Applications to defects in solids. Phys. Rev. B 97, 054102 (2018). Acknowledgements This work is supported by the National Natural Science Foundation of China (Grants No. 52422112, No. ...

    2018

  98. [109]

    Zhang, F. et al. Atomic-scale observation of reversible thermally driven phase transformation in 2D In 2Se3. ACS Nano 13, 8004–8011 (2019)

    2019

  99. [110]

    Chen, Z. et al. Atomic imaging of electrically switchable striped domains in β -In2Se3. Advanced Science 8, 2100713 (2021)

    2021

  100. [111]

    Zhang, F. et al. Atomic-scale manipulation of polar domain boundaries in monolayer ferroelectric In 2Se3. Nature Communications 15, 718 (2024)

    2024

  101. [112]

    Wang, Q. et al. Electron microscopy study of stacking defects in β -In2Se3. Journal of Materials Research 38, 330–336 (2024)

    2024

  102. [113]

    Tan, C. K. Y ., Fu, W. & Loh, K. P . Polymorphism and ferroelectricity in indium(III) selenide. Chemical Reviews 123, 8701–8717 (2023)

    2023

  103. [114]

    & Amelinckx, S

    van Landuyt, J., van Tendeloo, G. & Amelinckx, S. Phase transitions in In 2Se3 as studied by electron microscopy and electron diffraction. physica status solidi (a) 30, 299–314 (1975)

    1975

  104. [115]

    Lin, M. et al. Controlled growth of atomically thin In 2Se3 flakes by van der waals epitaxy . Journal of the American Chemical Society 135, 13274–13277 (2013)

    2013

  105. [116]

    Xu, C. et al. Two-dimensional antiferroelectricity in nanostripe-ordered In2Se3. Phys. Rev. Lett. 125, 047601 (2020)

    2020

  106. [117]

    Xu, C. et al. Two-dimensional ferroelasticity in van der Waals β ’-In2Se3. Nature Communications 12, 3665 (2021). 19

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.