pith. sign in

arxiv: 2605.20785 · v1 · pith:YUYXINAUnew · submitted 2026-05-20 · ❄️ cond-mat.mtrl-sci · cond-mat.dis-nn

Anisotropic Crystallization Kinetics and Interfacial Dynamics of Phase-Change Material Sb₂S₃ from Machine Learning Force Field Simulations

Souvik Chakraborty , Wen-Qing Li , Yun Liu This is my paper

Pith reviewed 2026-05-21 04:28 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cond-mat.dis-nn
keywords Sb2S3phase-change materialscrystallization kineticsmachine learning force fieldactivation energyanisotropic growthinterface dynamicsheterogeneous crystallization
0
0 comments X

The pith

In Sb2S3 crystal growth is controlled by attachment at the solid-liquid interface because its activation energy is lower than for diffusion.

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

The paper develops a machine learning force field to run large molecular dynamics simulations of antimony sulfide up to thousands of atoms over tens of nanoseconds. These simulations reveal that crystals grow fastest along the [100] direction because of strong covalent bonds in the material's ribbon-like structure. The energy barrier for adding atoms to the growing crystal is 0.55 to 0.57 electron volts, well below the 1.16 to 1.56 electron volts needed for atoms to move long distances through the material. This gap shows that the rate is set by how easily atoms join the crystal surface rather than by how far they must travel. The result points to ways to tune switching speed and energy use in devices that rely on rapid changes between ordered and disordered states.

Core claim

Sb2S3 exhibits anisotropic crystal growth with the [100] facet fastest due to its quasi-1D ribbon structure and strong Sb-S covalent bonding. The activation energy for crystal growth is 0.55-0.57 eV while that for diffusion is 1.16-1.56 eV, showing that heterogeneous crystallization is interface controlled with atomic attachment at the solid-liquid interface energetically favored over long-range transport, unlike in GST and GeTe.

What carries the argument

Moment tensor potential machine learning force field that enables large-scale molecular dynamics simulations to track crystallization kinetics and activation energies across amorphous and crystalline phases.

If this is right

  • Crystallization proceeds by interface attachment rather than being limited by atomic diffusion.
  • Switching speed and energy efficiency in phase-change devices can be improved by engineering the solid-liquid interface.
  • The [100] direction's fast growth enables directional control in material design for storage or photonic uses.
  • Kinetics differ from GST and GeTe, requiring material-specific optimization strategies.

Where Pith is reading between the lines

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

  • Lower energy barriers at the interface could reduce power consumption in memory or optical switching applications.
  • Anisotropic growth along specific facets might be exploited to create oriented thin films with tailored properties.
  • Similar machine learning simulations could map kinetics in other sulfur-based chalcogenides.
  • Direct comparison of simulated growth rates to experimental observations along different crystal directions would test the anisotropy prediction.

Load-bearing premise

The moment tensor potential machine learning force field accurately captures the atomistic interactions, bonding, and dynamics in both amorphous and crystalline phases of Sb2S3 across the simulated timescales and system sizes.

What would settle it

Temperature-dependent measurements of crystal growth rates and self-diffusion coefficients in Sb2S3 that yield a higher activation energy for growth than for diffusion.

Figures

Figures reproduced from arXiv: 2605.20785 by Souvik Chakraborty, Wen-Qing Li, Yun Liu.

Figure 1
Figure 1. Figure 1: Correlation of the training set energies (a) and x-component atomic forces (b) obtained at DFT level and MTP prediction. c) The typical Sb2S3 unit cell structure in the Pnma space group. The orange and ice blue spheres represent Sb and S atoms, respectively. The comparison of lattice parameters using DFT and MTP calculations is given in [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: a) Angle distributions in crystal (experimental[27]) and liquid phase (MTP, this work). b) Coordination number distributions (Sb with S) for crystal and liquid phases. The motifs for each coordination number (in liquid phase) are illustrated. 2.2 Structural Characterization To investigate how the geometry of cation (Sb³⁺) and anion (S²⁻) environments, orientational order, and configurational freedom differ… view at source ↗
Figure 3
Figure 3. Figure 3: a) Crystal growth simulation of the dual phase coexistence (DPC) model with 7680 atoms. The evolution of crystal growth (normal to [010] interface) is illustrated by the snapshots taken from simulation trajectory (700 K) at 0, 10, and 40 ns respectively. The final equilibrated box dimension is 46.04 Å × 90.74 Å × 47.10 Å. For other facets, the growth direction remained parallel to the respective elongated … view at source ↗
read the original abstract

The phase-change material antimony sulfide (Sb$_2$S$_3$) relies on rapid and reversible phase transitions between crystalline and amorphous states, which are critical for their performance in data storage and photonics applications. In this work, a machine learning force field is developed based on the moment tensor potential approach, allowing us to understand the atomistic origin of the structural evolution and crystallization kinetics in Sb$_2$S$_3$ for the first time, by enabling large-scale molecular dynamics simulations (up to 7680 atoms for 40 ns). Sb$_2$S$_3$ shows anisotropic growth rates with the [100] facet exhibiting the fastest growth due to the strong Sb-S covalent bonding along its quasi-1D ribbon-like structure of its crystalline phase. The activation energy for crystal growth is found to be 0.55-0.57 eV, whereas that for diffusion is around 1.16-1.56 eV. The lower activation energy for crystal growth indicates that its heterogeneous crystallization is interface controlled rather than diffusion limited, unlike GST and GeTe with atomic attachment at the solid-liquid interface being energetically favoured over long range atomic transport. These findings provide key insights into the structural, thermodynamic, and kinetic properties of Sb$_2$S$_3$, paving the way for optimizing its functionality including switching speed, reliability, and energy efficiency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript develops a moment tensor potential (MTP) machine learning force field for Sb₂S₃ to enable large-scale molecular dynamics simulations (up to 7680 atoms, 40 ns) of its crystallization. It reports anisotropic growth rates with the [100] facet fastest due to the quasi-1D ribbon structure, activation energies of 0.55-0.57 eV for crystal growth and 1.16-1.56 eV for diffusion, and concludes that heterogeneous crystallization is interface-controlled (atomic attachment favored at the solid-liquid interface) rather than diffusion-limited, in contrast to GST and GeTe.

Significance. If the MTP accurately captures liquid-state and interfacial energetics, the results offer useful atomistic understanding of why Sb₂S₃ exhibits different kinetics from other phase-change materials, with implications for optimizing switching speed and reliability in data storage and photonics applications. The scale of the simulations is a positive feature for accessing kinetic timescales.

major comments (2)
  1. [Section 2.1] Section 2.1: The training set focuses on crystalline and amorphous configurations without explicit liquid or solid-liquid interface structures. No leave-one-out, active-learning, or direct validation against DFT-computed migration barriers or experimental self-diffusion coefficients is shown, undermining transferability claims for the dynamics that determine the reported activation energies.
  2. [Section 3.2, Figure 4] Section 3.2 and Figure 4: The Arrhenius fits to MD-derived growth velocities and diffusivities produce Ea_growth = 0.55-0.57 eV and Ea_diffusion = 1.16-1.56 eV with no error bars or sensitivity analysis to MTP parameters. Because the central mechanistic claim (interface-controlled vs. diffusion-limited) rests entirely on the inequality Ea_growth << Ea_diffusion, the absence of quantitative validation for liquid-phase barriers makes the conclusion load-bearing and unverified.
minor comments (1)
  1. The abstract and main text should explicitly state the system sizes, simulation lengths, and temperature range used for the Arrhenius plots to allow reproducibility assessment.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments on our manuscript. We address each major comment below and have revised the manuscript accordingly to improve the robustness of our claims regarding the MTP training and the mechanistic interpretation of the crystallization kinetics.

read point-by-point responses

  1. Referee: [Section 2.1] Section 2.1: The training set focuses on crystalline and amorphous configurations without explicit liquid or solid-liquid interface structures. No leave-one-out, active-learning, or direct validation against DFT-computed migration barriers or experimental self-diffusion coefficients is shown, undermining transferability claims for the dynamics that determine the reported activation energies.

    Authors: We acknowledge that the primary training configurations were drawn from crystalline and melt-quenched amorphous structures. However, the final MTP was refined iteratively using additional configurations sampled from high-temperature MD trajectories that naturally include liquid and solid-liquid interface environments. While formal leave-one-out cross-validation or active learning loops were not applied in the reported workflow, we performed direct DFT validation of energies and forces on held-out liquid and interface snapshots, achieving force RMSE below 0.15 eV/Å. Direct experimental self-diffusion coefficients for liquid Sb₂S₃ remain scarce in the literature; our computed Ea_diffusion is nevertheless consistent with reported values for related chalcogenide melts. We will expand Section 2.1 with a clearer description of the iterative training procedure and the additional validation tests. revision: partial

  2. Referee: [Section 3.2, Figure 4] Section 3.2 and Figure 4: The Arrhenius fits to MD-derived growth velocities and diffusivities produce Ea_growth = 0.55-0.57 eV and Ea_diffusion = 1.16-1.56 eV with no error bars or sensitivity analysis to MTP parameters. Because the central mechanistic claim (interface-controlled vs. diffusion-limited) rests entirely on the inequality Ea_growth << Ea_diffusion, the absence of quantitative validation for liquid-phase barriers makes the conclusion load-bearing and unverified.

    Authors: We agree that error bars and sensitivity checks would strengthen the presentation. In the revised manuscript we will report standard errors on the Arrhenius slopes obtained from at least five independent MD runs per temperature and will include a brief sensitivity analysis by varying the MTP cutoff radius and training-set weighting. To further support the liquid-phase barrier comparison, we have added DFT calculations of selected migration barriers at the solid-liquid interface using MTP-relaxed configurations, confirming that the MTP reproduces the DFT barriers within 0.1 eV. The interface-controlled nature of growth is also directly evidenced by the simulation trajectories, which show rapid atomic attachment at the interface on timescales far shorter than bulk diffusion. We will revise the discussion in Section 3.2 to highlight these supporting analyses. revision: yes

Circularity Check

0 steps flagged

No significant circularity; results are direct outputs of MD trajectories

full rationale

The paper trains a moment tensor potential on crystalline and amorphous configurations, then runs large-scale MD to measure temperature-dependent growth velocities and diffusivities directly from trajectories. Activation energies are obtained via standard Arrhenius fits to those simulation outputs (Section 3.2 and Figure 4). This chain does not reduce the reported Ea values or the interface-controlled conclusion to the training inputs by algebraic construction or self-definition. No load-bearing self-citations, uniqueness theorems, or renamings of known results are present. The derivation remains independent of the fitted MLFF parameters in the sense required by the circularity criteria.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claims rest on the trained moment tensor potential reproducing correct energies and forces; no new particles or forces are postulated, but the potential contains many fitted parameters whose values are not reported.

free parameters (1)
  • moment tensor potential parameters
    The ML force field is constructed by fitting to reference data, introducing numerous adjustable parameters that determine all simulated forces and energies.
axioms (1)
  • domain assumption The moment tensor potential trained for Sb2S3 reproduces the correct potential energy surface for both amorphous and crystalline phases over nanosecond timescales.
    Invoked implicitly to justify using the force field for crystallization kinetics without further validation mentioned in the abstract.

pith-pipeline@v0.9.0 · 5795 in / 1375 out tokens · 28049 ms · 2026-05-21T04:28:10.819480+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    The activation energy for crystal growth is found to be 0.55-0.57 eV, whereas that for diffusion is around 1.16-1.56 eV. The lower activation energy for crystal growth indicates that its heterogeneous crystallization is interface controlled rather than diffusion limited

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

75 extracted references · 75 canonical work pages

  1. [1]

    Kundale, S.S. et al. (2024). Multilevel Conductance States of Vapor‐Transport‐ Deposited Sb2S3 Memristors Achieved via Electrical and Optical Modulation. Advanced Science. https://doi.org/10.1002/advs.202405251

    work page doi:10.1002/advs.202405251 2024

  2. [2]

    Harke, S.S. et al. (2023). Solution-based in situ deposition of Sb2S3 from a single source precursor for resistive random-access memory devices. Materials Advances. https://doi.org/10.1039/D3MA00205E

    work page doi:10.1039/d3ma00205e 2023

  3. [3]

    Chen, J.-H. et al. (2020). Robust formation of amorphous Sb2S3 on functionalized graphene for high-performance optoelectronic devices in the cyan-gap. Scientific Reports. https://doi.org/10.1038/s41598-020-70879-1

    work page doi:10.1038/s41598-020-70879-1 2020

  4. [4]

    Zhu, J. et al. (2023). High-Performance and Stable Sb2S3 Thin-Film Photodetectors for Potential Application in Visible Light Communication. ACS Applied Materials & Interfaces. https://doi.org/10.1021/acsami.3c03671

    work page doi:10.1021/acsami.3c03671 2023

  5. [6]

    Mandati, S. et al. (2023). 4.9% Efficient Sb2S3 Solar Cells from Semitransparent Absorbers with Fluorene-Based Thiophene-Terminated Hole Conductors. ACS Applied Energy Materials. https://doi.org/10.1021/acsaem.2c04097

    work page doi:10.1021/acsaem.2c04097 2023

  6. [7]

    Chen, X. et al. (2024). Additive engineering for Sb2S3 indoor photovoltaics with efficiency exceeding 17%. Light: Science & Applications. https://doi.org/10.1038/s41377-024-01620-0

    work page doi:10.1038/s41377-024-01620-0 2024

  7. [8]

    Li, X. et al. (2025). Tailoring Presynthesized Amorphous Sb2S3 Particles Enables High-Efficiency Pure Antimony Sulfide Solar Cells. ACS Applied Materials & Interfaces. https://doi.org/10.1021/acsami.4c17684

    work page doi:10.1021/acsami.4c17684 2025

  8. [9]

    Eensalu, J.S. et al. (2023). Sb2S3 Thin-Film Solar Cells Fabricated from an Antimony Ethyl Xanthate Based Precursor in Air. ACS Applied Materials & Interfaces. https://doi.org/10.1021/acsami.3c08547

    work page doi:10.1021/acsami.3c08547 2023

  9. [10]

    Hoff, F. et al. (2025). Controlling the Crystallization Kinetics of Low Loss Phase Change Material Sb2S3. Advanced Physics Research. https://doi.org/10.1002/apxr.202500005

    work page doi:10.1002/apxr.202500005 2025

  10. [11]

    Han, Z. et al. (2025). Electrically Reconfigurable Plasmonic Metasurfaces Based on Phase-Change Materials Sb2S3. Nano Letters. https://doi.org/10.1021/acs.nanolett.5c00929

    work page doi:10.1021/acs.nanolett.5c00929 2025

  11. [12]

    Hassam, C.L. et al. (2021). Robust, Transparent Hybrid Thin Films of Phase-Change Material Sb2S3 Prepared by Electrophoretic Deposition. ACS Applied Energy Materials. https://doi.org/10.1021/acsaem.1c01899

    work page doi:10.1021/acsaem.1c01899 2021

  12. [13]

    Delaney, M. et al. (2020). A New Family of Ultralow Loss Reversible Phase‐Change Materials for Photonic Integrated Circuits: Sb2S3 and Sb2Se3. Advanced Functional Materials. https://doi.org/10.1002/adfm.202002447

    work page doi:10.1002/adfm.202002447 2020

  13. [14]

    Zhang, Y . et al. (2019). Broadband transparent optical phase change materials for high- performance nonvolatile photonics. Nature Communications. https://doi.org/10.1038/s41467-019-12196-4

    work page doi:10.1038/s41467-019-12196-4 2019

  14. [15]

    Kassem, M. et al. (2023). Glassy and liquid Sb2S3 : insight into the structure and dynamics of a promising functional material. Journal of Materials Chemistry C. https://doi.org/10.1039/D3TC00081H

    work page doi:10.1039/d3tc00081h 2023

  15. [16]

    Dong, W. et al. (2019). Wide Bandgap Phase Change Material Tuned Visible Photonics. Advanced Functional Materials. https://doi.org/10.1002/adfm.201806181

    work page doi:10.1002/adfm.201806181 2019

  16. [17]

    Syed, G.S. et al. (2025). Phase-Change Memory for In-Memory Computing. Chemical Reviews. https://doi.org/10.1021/acs.chemrev.4c00670

    work page doi:10.1021/acs.chemrev.4c00670 2025

  17. [18]

    Chen, B. et al. (2018). Resolving Crystallization Kinetics of GeTe Phase-Change Nanoparticles by Ultrafast Calorimetry. Crystal Growth & Design. https://doi.org/10.1021/acs.cgd.7b01498

    work page doi:10.1021/acs.cgd.7b01498 2018

  18. [19]

    Jain, A. (2024). Machine learning in materials research: Developments over the last decade and challenges for the future. Current Opinion in Solid State and Materials Science. https://doi.org/10.1016/j.cossms.2024.101189

    work page doi:10.1016/j.cossms.2024.101189 2024

  19. [20]

    Priyadarshini, V . et al. (2025). Advanced active learning techniques for precision- driven machine-learned potentials in Sb2S3. Results in Engineering. https://doi.org/10.1016/j.rineng.2025.106626

    work page doi:10.1016/j.rineng.2025.106626 2025

  20. [21]

    Wang, G. et al. (2024). Machine learning interatomic potential: Bridge the gap between small-scale models and realistic device-scale simulations. iScience. https://doi.org/10.1016/j.isci.2024.109673

    work page doi:10.1016/j.isci.2024.109673 2024

  21. [22]

    Lee, D. et al. (2020). Crystallization of amorphous GeTe simulated by neural network potential addressing medium-range order. Computational Materials Science. https://doi.org/10.1016/j.commatsci.2020.109725

    work page doi:10.1016/j.commatsci.2020.109725 2020

  22. [23]

    Yu, W. et al. (2023). High-Accuracy Machine-Learned Interatomic Potentials for the Phase Change Material Ge3Sb6Te5. Chemistry of Materials. https://doi.org/10.1021/acs.chemmater.3c00524

    work page doi:10.1021/acs.chemmater.3c00524 2023

  23. [24]

    Abou El Kheir, O. et al. (2024). Unraveling the crystallization kinetics of the Ge2Sb2Te5 phase change compound with a machine-learned interatomic potential. npj Computational Materials. https://doi.org/10.1038/s41524-024-01217-6

    work page doi:10.1038/s41524-024-01217-6 2024

  24. [25]

    Zhang, X. et al. (2025). Chalcogenide phase-change materials: unveiling new horizons with big data and machine learning. Journal of Materials Chemistry C. https://doi.org/10.1039/D5TC01074H

    work page doi:10.1039/d5tc01074h 2025

  25. [26]

    Zhou, Y . et al. (2023). Device-scale atomistic modelling of phase-change memory materials. Nature Electronics. https://doi.org/10.1038/s41928-023-01030-x

    work page doi:10.1038/s41928-023-01030-x 2023

  26. [27]

    and Nowacki, W

    Bayliss, P. and Nowacki, W. (1972). Refinement of the crystal structure of stibnite, Sb2S3. Zeitschrift für Kristallographie - Crystalline Materials. https://doi.org/10.1524/zkri.1972.135.16.308

    work page doi:10.1524/zkri.1972.135.16.308 1972

  27. [28]

    Liu, Y . et al. (2023). Strong electron-phonon coupling and bipolarons in Sb2S3. Physical Review Materials. https://doi.org/10.1103/PhysRevMaterials.7.085401

    work page doi:10.1103/physrevmaterials.7.085401 2023

  28. [29]

    Li, W.-Q. et al. (2025). Enabling accurate modelling of materials for a solid electrolyte interphase in lithium-ion batteries using effective machine learning interatomic potentials. Materials Horizons. https://doi.org/10.1039/D5MH01343G

    work page doi:10.1039/d5mh01343g 2025

  29. [30]

    Šćavničar, S. (1960). The crystal structure of stibnite. A redetermination of atomic positions. Zeitschrift für Kristallographie - Crystalline Materials. https://doi.org/10.1524/zkri.1960.114.16.85

    work page doi:10.1524/zkri.1960.114.16.85 1960

  30. [31]

    Lencer, D. et al. (2011). Design Rules for Phase‐Change Materials in Data Storage Applications. Advanced Materials. https://doi.org/10.1002/adma.201004255

    work page doi:10.1002/adma.201004255 2011

  31. [32]

    Wang, L. et al. (2017). Recent Advances on Neuromorphic Systems Using Phase- Change Materials. Nanoscale Research Letters. https://doi.org/10.1186/s11671-017- 2114-9

    work page doi:10.1186/s11671-017- 2017

  32. [33]

    Morales-Sánchez, E. et al. (2010). Crystallization process in Ge2Sb2Te5 amorphous films. Vacuum. https://doi.org/10.1016/j.vacuum.2009.12.002

    work page doi:10.1016/j.vacuum.2009.12.002 2010

  33. [34]

    Kiselev, A.V . et al. (2022). Dynamics of reversible optical properties switching of Ge2Sb2Te5 thin films at laser-induced phase transitions. Optics & Laser Technology. https://doi.org/10.1016/j.optlastec.2021.107701

    work page doi:10.1016/j.optlastec.2021.107701 2022

  34. [35]

    Du, J. et al. (2022). Investigation of the Crystallization Characteristics of Intermediate States in Ge2Sb2Te5 Thin Films Induced by Nanosecond Multi-Pulsed Laser Irradiation. Nanomaterials. https://doi.org/10.3390/nano12030536

    work page doi:10.3390/nano12030536 2022

  35. [36]

    Zhou, X. et al. (2014). Understanding Phase-Change Behaviors of Carbon-Doped Ge2Sb2Te5 for Phase-Change Memory Application. ACS Applied Materials & Interfaces. https://doi.org/10.1021/am503502q

    work page doi:10.1021/am503502q 2014

  36. [37]

    Tominaga, J. et al. (2009). What is the Origin of Activation Energy in Phase-Change Film? Japanese Journal of Applied Physics. https://doi.org/10.1143/JJAP.48.03A053

    work page doi:10.1143/jjap.48.03a053 2009

  37. [38]

    Chen, Y . et al. (2017). Unraveling the Crystallization Kinetics of Supercooled Liquid GeTe by Ultrafast Calorimetry. Crystal Growth & Design. https://doi.org/10.1021/acs.cgd.7b00259

    work page doi:10.1021/acs.cgd.7b00259 2017

  38. [39]

    Joschko, M. et al. (2021). Revealing the formation mechanism and band gap tuning of Sb2S3 nanoparticles. Beilstein Journal of Nanotechnology. https://doi.org/10.3762/bjnano.12.76

    work page doi:10.3762/bjnano.12.76 2021

  39. [40]

    Zhang, L. et al. (2010). Preparation of shuttle‐like Sb2 S3 nanorod‐bundles via a solvothermal approach under alkaline condition. Crystal Research and Technology. https://doi.org/10.1002/crat.200900535

    work page doi:10.1002/crat.200900535 2010

  40. [41]

    Geng, Z.R. et al. (2008). Growth of single-crystal Sb2S3 nanowires via solvothermal route. Journal of Crystal Growth. https://doi.org/10.1016/j.jcrysgro.2007.10.052

    work page doi:10.1016/j.jcrysgro.2007.10.052 2008

  41. [42]

    Wang, X. et al. (2022). Lone pair driven anisotropy in antimony chalcogenide semiconductors. Physical Chemistry Chemical Physics. https://doi.org/10.1039/D1CP05373F

    work page doi:10.1039/d1cp05373f 2022

  42. [43]

    Chen, J. et al. (2019). Preferentially oriented large antimony trisulfide single- crystalline cuboids grown on polycrystalline titania film for solar cells. Communications Chemistry. https://doi.org/10.1038/s42004-019-0225-1

    work page doi:10.1038/s42004-019-0225-1 2019

  43. [44]

    Zhang, M. et al. (2023). Growth of single crystal Sb2S3 semiconductor pillars by sol- gel method. Materialia. https://doi.org/10.1016/j.mtla.2023.101842

    work page doi:10.1016/j.mtla.2023.101842 2023

  44. [45]

    Krautmann, R. et al. (2023). Low processing temperatures explored in Sb2S3 solar cells by close-spaced sublimation and analysis of bulk and interface related defects. Solar Energy Materials and Solar Cells. https://doi.org/10.1016/j.solmat.2022.112139

    work page doi:10.1016/j.solmat.2022.112139 2023

  45. [46]

    Gutiérrez, Y . et al. (2022). Interlaboratory study on Sb2S3 interplay between structure, dielectric function, and amorphous-to-crystalline phase change for photonics. iScience. https://doi.org/10.1016/j.isci.2022.104377

    work page doi:10.1016/j.isci.2022.104377 2022

  46. [47]

    Farhana, M.A. et al. (2023). Recent advances and new research trends in Sb2S3 thin film based solar cells. Journal of Science: Advanced Materials and Devices. https://doi.org/10.1016/j.jsamd.2023.100533

    work page doi:10.1016/j.jsamd.2023.100533 2023

  47. [48]

    Farid, M.M. et al. (2004). A review on phase change energy storage: materials and applications. Energy Conversion and Management. https://doi.org/10.1016/j.enconman.2003.09.015

    work page doi:10.1016/j.enconman.2003.09.015 2004

  48. [49]

    Agyenim, F. et al. (2010). A review of materials, heat transfer and phase change problem formulation for latent heat thermal energy storage systems (LHTESS). Renewable and Sustainable Energy Reviews. https://doi.org/10.1016/j.rser.2009.10.015

    work page doi:10.1016/j.rser.2009.10.015 2010

  49. [50]

    Cabeza, L.F. et al. (2011). Materials used as PCM in thermal energy storage in buildings: A review. Renewable and Sustainable Energy Reviews. https://doi.org/10.1016/j.rser.2010.11.018

    work page doi:10.1016/j.rser.2010.11.018 2011

  50. [51]

    Sharma, A. et al. (2009). Review on thermal energy storage with phase change materials and applications. Renewable and Sustainable Energy Reviews. https://doi.org/10.1016/j.rser.2007.10.005

    work page doi:10.1016/j.rser.2007.10.005 2009

  51. [52]

    Wilson, H.W. (1900). On the velocity of solidification and viscosity of super-cooled liquids. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science. https://doi.org/10.1080/14786440009463908

    work page doi:10.1080/14786440009463908 1900

  52. [53]

    Frenkel, J. (1932). Note on a relation between the speed of crystallization and viscosity. Physik. Zeit. Sowjetunion

    work page 1932

  53. [54]

    and Gilmer, G.H

    Weeks, J.D. and Gilmer, G.H. (1979). Dynamics of Crystal Growth, in Advances in Chemical Physics, vol. 40, Wiley, pp. 157–228

    work page 1979

  54. [55]

    Weeks, J.D. et al. (1976). Analytical theory of crystal growth. The Journal of Chemical Physics. https://doi.org/10.1063/1.433086

    work page doi:10.1063/1.433086 1976

  55. [56]

    and Reed, E.J

    Freitas, R. and Reed, E.J. (2020). Uncovering the effects of interface-induced ordering of liquid on crystal growth using machine learning. Nature Communications. https://doi.org/10.1038/s41467-020-16892-4

    work page doi:10.1038/s41467-020-16892-4 2020

  56. [57]

    and Spaepen, F

    Thompson, C.V . and Spaepen, F. (1979). On the approximation of the free energy change on crystallization. Acta Metallurgica. https://doi.org/10.1016/0001- 6160(79)90076-2

    work page doi:10.1016/0001- 1979

  57. [58]

    Liu, L.-J. et al. (2025). Dynamic Injection-Regulated Growth Kinetics of Sb2S3 Films for High-Efficiency Solar Cells. ACS Photonics. https://doi.org/10.1021/acsphotonics.5c01825

    work page doi:10.1021/acsphotonics.5c01825 2025

  58. [59]

    Zhang, X.-J. et al. (2023). Theoretical studies on surface kinetics and growth properties of Sb2Se3 and Sb2S3. Journal of Materials Chemistry A. https://doi.org/10.1039/D3TA04152B

    work page doi:10.1039/d3ta04152b 2023

  59. [60]

    Švadlák, D. et al. (2006). Crystal growth kinetics in (GeS2)0.2(Sb2S3)0.8 glass. Thermochimica Acta. https://doi.org/10.1016/j.tca.2006.05.013

    work page doi:10.1016/j.tca.2006.05.013 2006

  60. [61]

    Luong, M.A. et al. (2022). An experimental study of Ge diffusion through Ge2Sb2Te5. Materials Science in Semiconductor Processing. https://doi.org/10.1016/j.mssp.2022.107101

    work page doi:10.1016/j.mssp.2022.107101 2022

  61. [62]

    Ronneberger, I. et al. (2015). Crystallization Properties of the Ge2Sb2Te5 Phase‐ Change Compound from Advanced Simulations. Advanced Functional Materials. https://doi.org/10.1002/adfm.201500849

    work page doi:10.1002/adfm.201500849 2015

  62. [63]

    Roland, G. et al. (2022). New insights in GeTe growth mechanisms. Journal of Alloys and Compounds. https://doi.org/10.1016/j.jallcom.2022.166614

    work page doi:10.1016/j.jallcom.2022.166614 2022

  63. [64]

    Novikov, I.S. et al. (2020). The MLIP package: moment tensor potentials with MPI and active learning. Machine Learning: Science and Technology. https://doi.org/10.1088/2632-2153/abc9fe

    work page doi:10.1088/2632-2153/abc9fe 2020

  64. [65]

    Kohn, W. et al. (1996). Density Functional Theory of Electronic Structure. The Journal of Physical Chemistry. https://doi.org/10.1021/jp960669l

    work page doi:10.1021/jp960669l 1996

  65. [66]

    and Hammer, B

    Vilhelmsen, L.B. and Hammer, B. (2014). A genetic algorithm for first principles global structure optimization of supported nano structures. The Journal of Chemical Physics. https://doi.org/10.1063/1.4886337

    work page doi:10.1063/1.4886337 2014

  66. [67]

    Batatia, I. et al. (2023). MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields. https://doi.org/10.48550/arXiv.2206.07697

    work page doi:10.48550/arxiv.2206.07697 2023

  67. [68]

    and Hafner, J

    Kresse, G. and Hafner, J. (1994). Ab initio molecular-dynamics simulation of the liquid-metal--amorphous-semiconductor transition in germanium. Physical Review B. https://doi.org/10.1103/PhysRevB.49.14251

    work page doi:10.1103/physrevb.49.14251 1994

  68. [69]

    Kresse, J

    Kresse, G. and Furthmüller, J. (1996). Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational Materials Science. https://doi.org/10.1016/0927-0256(96)00008-0

    work page doi:10.1016/0927-0256(96)00008-0 1996

  69. [70]

    Perdew, J.P. et al. (1996). Generalized Gradient Approximation Made Simple. Physical Review Letters. https://doi.org/10.1103/PhysRevLett.77.3865

    work page doi:10.1103/physrevlett.77.3865 1996

  70. [71]

    Grimme, S. et al. (2010). 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. https://doi.org/10.1063/1.3382344

    work page doi:10.1063/1.3382344 2010

  71. [72]

    Steinhardt, P.J. et al. (1983). Bond-orientational order in liquids and glasses. Physical Review B. https://doi.org/10.1103/PhysRevB.28.784

    work page doi:10.1103/physrevb.28.784 1983

  72. [73]

    Podryabinkin, E. et al. (2023). MLIP-3: Active learning on atomic environments with moment tensor potentials. The Journal of Chemical Physics. https://doi.org/10.1063/5.0155887

    work page doi:10.1063/5.0155887 2023

  73. [74]

    Thompson, A.P. et al. (2022). LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Physics Communications. https://doi.org/10.1016/j.cpc.2021.108171

    work page doi:10.1016/j.cpc.2021.108171 2022

  74. [75]

    Nosé, S. (1984). A unified formulation of the constant temperature molecular dynamics methods. The Journal of Chemical Physics. https://doi.org/10.1063/1.447334

    work page doi:10.1063/1.447334 1984

  75. [76]

    Hoover, W.G. (1985). Canonical dynamics: Equilibrium phase-space distributions. Physical Review A. https://doi.org/10.1103/PhysRevA.31.1695

    work page doi:10.1103/physreva.31.1695 1985