Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Mechanisms and Stability of Li Dynamics in Amorphous Li-Ti-P-S-Based Mixed Ionic-Electronic Conductors: A Machine Learning Molecular Dynamics Study

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Machine-learned molecular dynamics shows that lithium transport in amorphous Ti-doped Li-P-S proceeds by free-volume hopping, with 10–20% Ti doping giving the most stable transport channels.

desk verdict Useful MLFF-MD transport data for Ti-doped LPS, but the entropy-based stability argument is not supported by the calculations. read the letter →

arxiv 2506.11199 v2 pith:AZR3L4QA submitted 2025-06-12 cond-mat.mtrl-sci physics.comp-ph

classification cond-mat.mtrl-sciphysics.comp-ph
keywords machinelearningforcefieldmoleculardynamicslithiumsolidelectrolytemixedionic-electronicconductorLi-Ti-P-Sionicconductivityconfigurationalentropyfree-volumediffusion
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that in an amorphous lithium titanium phosphorus sulfide mixed ionic-electronic conductor, lithium ions move by a free-volume diffusion mechanism: they hop through voids in the disordered structure rather than along fixed lattice pathways. Using a machine-learning force field trained on ab initio molecular dynamics, the authors simulate 12,000-atom cells at three titanium doping levels and six temperatures, and their computed ionic conductivities and activation energies match recent experiments. They further claim that 10% and 20% Ti doping produce the most stable transport channels, because those compositions show higher configurational entropy of disordered Li-S coordination polyhedra, which they equate with lower Gibbs free energy. If correct, the work identifies an optimal doping window for the material and demonstrates that large-scale machine-learning molecular dynamics can explain transport mechanisms in amorphous solid electrolytes.

What carries the argument

The load-bearing machinery is a deep learning molecular dynamics workflow: a neural-network potential trained on ab initio MD data, then used in 3 ns NVT simulations of 12,000-atom amorphous cells at three Ti concentrations and six temperatures. Transport is quantified from mean-square displacement via the Nernst-Einstein relation, and channel stability is quantified from the Li-S coordination-number distribution $P(n_c)$ computed by k-nearest-neighbor counting. The identity that carries the stability argument is Eq. (9), $S_{\mathrm{config}} = -k_B \sum_{n_c} P(n_c)\ln P(n_c)$, supplemented by a vibrational-entropy term; the paper argues that higher $S_{\mathrm{config}}$ lowers the configurational Gibbs free energy and hence makes the 10% and 20% Ti channels more favorable.

What would settle it

A concrete check would be to compute, with the same force field or with DFT, the enthalpy and total entropy of amorphous Li-Ti-P-S at 0%, 10%, 20%, and 30% Ti; if 10–20% Ti does not give the lowest Gibbs free energy, or if the coordination-number entropy does not track the free-energy ordering, the channel-stability conclusion fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that amorphous lithium titanium phosphorus sulfide conducts lithium not through fixed crystalline pathways but through free-volume diffusion: individual Li ions hop among voids in the disordered structure, and the hopping medium is a set of disordered Li-S_n polyhedra with n ranging from 1 to 6. The authors report that a machine-learning force field trained on ab initio molecular dynamics reproduces experimentally measured ionic conductivities and activation energies across six temperatures, and that the lowest activation energy occurs at 10% Ti (0.3 eV), followed closely by 20% Ti (0.32 eV), while 0% and 30% Ti give higher barriers. They interpret the stability of the transport channels through coordination statistics: at 10% and 20% Ti, more than half of the Li atoms sit in four-coordinated S environments, whereas at 0% and 30% Ti the majority are three-coordinated. The accompanying configurational entropy of the coordination-number distribution is highest at 10% and next highest at 20% Ti, which the authors read as a decrease in Gibbs free energy and therefore a thermodynamically stabilized channel.

Load-bearing premise

The stability claim rests on treating the Shannon entropy of the Li-S coordination-number distribution as the dominant term in the Gibbs free energy, with enthalpy and other entropy contributions neglected.

Editorial extensions

If this is right

  • If the central claim is right, the optimal Ti doping for this amorphous MIEC lies near 10–20%, where activation energy is lowest and the Li-S channel network is most stable; 30% Ti overdoping and 0% Ti both degrade transport.
  • Free-volume diffusion implies that amorphous disorder is a design lever: increasing the diversity of Li-S coordination environments should raise configurational entropy and improve ionic transport.
  • The machine-learning force field approach becomes a validated tool for screening other amorphous sulfide electrolytes at 12,000-atom scale, where direct ab initio molecular dynamics would be prohibitively expensive.
  • Computed conductivities and activation energies matching experiment strengthen confidence that the machine-learned force field captures the relevant physics, not just the training set.

Reading between the lines

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

  • Editorial inference: the entropy-stability ranking could be checked by computing the enthalpy of each composition; if enthalpy differences are large enough to outweigh temperature times the configurational entropy, the claimed 10–20% window may not survive a full free-energy comparison.
  • Editorial inference: the coordination-entropy descriptor is cheap to compute from any trajectory and could serve as a screening metric for other amorphous solid electrolytes, not just Ti-doped LPS.
  • Editorial inference: because the paper does not compute electronic conductivity, the MIEC label rests on prior experiments; a simulation-based test of how Ti doping affects electron transport in the same 10–20% window would complete the picture.
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.

Referee Report

4 major / 5 minor

Summary. The manuscript develops a DeePMD-based machine-learned force field (MLFF) for amorphous Li-Ti-P-S mixed ionic-electronic conductors, trained on AIMD trajectories at 500-900 K, and uses it to run 12,000-atom DLMD simulations at 300-500 K for compositions with 0%, 10%, 20%, and 30% TiS2. From these simulations the authors extract Li MSDs, diffusion coefficients, Nernst-Einstein ionic conductivities, Arrhenius activation energies, Li-S coordination-number distributions, and a Shannon-entropy-based 'configurational entropy.' They report that the computed conductivities and activation energies agree with their own experimental values, that Li transport occurs by free-volume diffusion through disordered Li-S polyhedra, and that 10% and 20% Ti doping stabilize the transport channels by increasing configurational entropy and lowering the Gibbs free energy. The paper concludes that 10-20% Ti doping is optimal for this MIEC.

Significance. If the results are fully supported, the paper would be a useful demonstration that a DeePMD force field trained on high-temperature AIMD data can be applied to large-scale MD simulations of an amorphous sulfide MIEC, and it would identify a doping window (10-20% Ti) for optimizing Li transport. The transport calculations use a standard and appropriate Arrhenius/Nernst-Einstein framework, and the direct comparison with the authors' prior experimental work [13] is a strength. The main conceptual contribution, however, is the thermodynamic stability argument based on coordination-number entropy, and this is currently not established. The paper would also be strengthened by quantitative error reporting and by validation of the MLFF in the production temperature range. The computational pipeline itself is potentially valuable, but the central stability claim needs substantial revision before the manuscript can be accepted.

major comments (4)
  1. [Section 3.5.2 (Eqs. 8-10)] The central stability conclusion is not supported by the entropy analysis as presented. Equation (9) defines S_config as the Shannon entropy of the single-particle Li-S coordination-number distribution P(n_c). For an amorphous network, the configurational entropy that enters the Gibbs free energy counts accessible microscopic configurations, including many-body correlations among coordination environments, and one cannot simply identify it with the spread of a local coordination histogram. Moreover, the text states 'When S_config increases, ΔG_config decreases' and concludes that 10% and 20% Ti doping 'reduce the Gibbs free energy,' but ΔG = ΔH - TΔS, and no enthalpy difference ΔH is computed anywhere in the paper. The assertions that translational and electronic entropy contributions are negligible are not computed; only Svib (Eq. 10) is estimated, and Svib is reported to be of order 1-3 m k_B, i.e., negligible compared with k_B, so it cannot compensate for the missing enthalpy term. Unless an enthalpy term is computed or the claim is reframed as a structural descriptor rather than a thermodynamic stability argument, the 10/20% stabilization conclusion in Sections 3.5.2 and 4 is unsupported.
  2. [Sections 2.2-2.3 and 3.1] The MLFF is trained exclusively on AIMD trajectories at 500, 600, 700, 800, and 900 K, but all production DLMD simulations are performed at 300-500 K. The force/energy MAEs in Table 1 measure accuracy on the training distribution and do not establish accuracy in the extrapolation regime. A direct check of the MLFF against AIMD at 300-400 K, for example by comparing forces and energies on short AIMD trajectories or by running short AIMD simulations at low temperature and comparing with DLMD predictions, is needed before the quantitative conductivity and activation-energy values can be taken as reliable.
  3. [Section 3.3, Fig. 4] The quantitative comparison with experiment is presented without any statistical uncertainty. There are no error bars on σ(T), Ea, σ0, or the Shannon entropy, and the text does not state how many independent MD runs or trajectory blocks were used to estimate D. The reported composition ranking, with Ea = 0.30 eV for 10% Ti and 0.32 eV for 20% Ti versus higher values for 0% and 30% Ti, may lie within the statistical noise of a single 3 ns trajectory per state point. Block averaging or multiple independent simulations are required to support the claim that 10% and 20% Ti are optimum.
  4. [Section 3.4, Fig. 5] The free-volume diffusion mechanism is inferred primarily from visual inspection of the trajectory of a single Li ion per composition, without a quantitative free-volume analysis. The manuscript does not compute void-size distributions, hopping distances, or the correlation between Li displacements and local free-volume regions, and the conclusion in Fig. 5 that 10% and 20% Ti 'cover more area in 2D or volume in 3D' is not backed by a statistical or ensemble-averaged metric. A quantitative analysis of free volume and hopping statistics is needed to support the mechanistic claim.
minor comments (5)
  1. [Section 3.3, Eq. (4)] The text states that σ0 is determined from the y-intercept of a ln(σT) versus 1/T plot, but Eq. (4) is written as σ = σ0 exp(-Ea/kBT); if the plot uses σT, the pre-factor should be defined consistently, and the Arrhenius expression should be given in the same form.
  2. [Fig. 6 caption] The caption uses the label '(a)' twice, once for the 2D ball-and-stick model and once for the LPS:Ti00% heatmap; the panel labels (b)-(e) also appear to be misaligned with the described composition order.
  3. [Fig. 5 caption] The word 'psudo-boundaries' should be 'pseudo-boundaries,' and the caption should clarify that the cubic boundaries are only guides to visualize a single Li-ion trajectory.
  4. [Section 3.5.2] The sentence stating that configurational entropy 'explicitly refers to the degree of disorder in the Li-S polyhedra' is a definition introduced only after Eq. (9); it should appear before the entropy is computed.
  5. [References] References [29] and [30] appear to be the same DeePMD-kit citation in two forms; one should be removed and the other consistently formatted.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: ionic conductivities and activation energies come from MLFF-MD trajectories and are not fitted to the experimental values used for comparison; the only self-citation (Ref. 13) is a minor validation benchmark, and the Sec. 3.5.2 entropy/free-energy argument is under-supported rather than circular.

full rationale

The derivation chain for the main quantitative results is: DFT-AIMD trajectories -> DeePMD-kit MLFF training -> LAMMPS NVT MD -> MSD (Eq. 1) -> D (Eq. 2) -> Nernst-Einstein sigma (Eq. 3) -> Arrhenius Ea (Eq. 4). No experimental conductivity or activation energy enters this chain; the 'consistent with our recent experimental values' statements in Sec. 3.3 and Fig. 4 are validation comparisons to the authors' own paper [13], not fitted inputs. Since the MLFF is trained on ab initio data and the test error is reported on separate samples (Table 1, Fig. 1f-j), the conductivity prediction is not a fit renamed as a prediction. The transport-mechanism conclusion (free-volume diffusion, disordered Li-S polyhedra, Sec. 3.4) follows from inspecting MD trajectories and is likewise independent of the comparison data. The self-citation to Ref. [13] is real evidence (experimental measurements) and does not make the central claim circular, though the lack of an independent experimental benchmark keeps the validation at the level of a minor self-citation. The stability argument in Sec. 3.5.2 is the weakest part, but it is not circular: Eq. 9 defines Sconfig as a Shannon entropy of P(nc), and the paper then asserts 'When Sconfig increases, Delta Gconfig decreases' without computing the enthalpy in Delta G = H - T Stotal. That is an unsupported thermodynamic inference (a correctness/evidence gap), not a derivation that reduces to its own inputs by construction. Overall, no load-bearing step in the paper is equivalent to its input by definition, so circularity is low.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard transport equations, a transferability assumption for the MLFF, and a nonstandard identification of Shannon entropy of coordination numbers with thermodynamic configurational entropy. No free parameters are fit to the experimental target values; the activation energy and pre-exponential are fitted to simulated conductivities, and the coordination cutoff is an analysis choice.

free parameters (3)
  • Li-S coordination cutoff radius rcut = 2.5 to 3.1 Å (scanned, no single value chosen)
    The coordination numbers P(nc), configurational entropy (Eq. 9), and the stability conclusion all depend on this hand-chosen cutoff range; the paper reports trends across rcut rather than a physically justified single cutoff.
  • Activation energy Ea (Arrhenius fit) = 0.30 eV (10%), 0.32 eV (20%), higher for 0% and 30% (exact values not stated)
    Ea is obtained by least-squares fitting of ln(sigma*T) versus 1/T via Eq. 4; the comparison of these fitted values to experiment is the central validation of the doping trend.
  • Pre-exponential factor sigma0 (Arrhenius fit) = not reported numerically
    sigma0 is a fitted intercept in the Arrhenius analysis and is not given; without it the fit cannot be fully reconstructed.
assumptions (5)
  • domain assumption Nernst-Einstein relation with Z=1 and no Haven ratio correction (Eq. 3)
    The conversion from diffusivity to ionic conductivity assumes all Li contribute to conduction as independent charge carriers; correlated motion or incomplete dissociation would shift sigma.
  • domain assumption Single Arrhenius process over the simulated temperature range (Eq. 4)
    Activation energy is extracted from a linear ln(sigma*T) versus 1/T fit, assuming one thermally activated mechanism from 300-500 K.
  • domain assumption MLFF trained on ~100-atom AIMD at 500-900 K transfers to 12,000-atom cells at 300-500 K
    No AIMD or experimental validation of the low-temperature diffusivity is shown; if the potential extrapolates poorly, all conductivity numbers are affected.
  • ad hoc to paper Shannon entropy of P(nc) is treated as configurational entropy controlling Gibbs free energy
    Eq. 9 is an information entropy of coordination number counts, not a thermodynamic configurational entropy; the paper further neglects enthalpy in the Delta G argument without justification.
  • domain assumption PBE-type DFT provides the ground truth labels for MLFF training
    All MLFF accuracy is relative to the chosen DFT functional; no convergence or functional-error analysis is presented.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mechanisms and Stability of Li Dynamics in Amorphous Li-Ti-P-S-Based Mixed Ionic-Electronic Conductors: A Machine Learning Molecular Dynamics Study." pith.science (2026). https://pith.science/paper/AZR3L4QA

@misc{pith2026250611199,
  author       = {Pith},
  title        = {Pith review of: Mechanisms and Stability of Li Dynamics in Amorphous Li-Ti-P-S-Based Mixed Ionic-Electronic Conductors: A Machine Learning Molecular Dynamics Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AZR3L4QA}},
  note         = {Machine review of arXiv:2506.11199}
}
abstract

Mixed ionic-electronic conductors (MIECs) exhibit both high ionic and electronic conductivity to improve the battery performance. In this work, we investigate the mechanism and stability of transport channels in our recently developed MIEC material, amorphous Ti-doped lithium phosphorus sulfide (LPS), using molecular dynamics (MD) simulations with a 99\% accurate machine-learning force field (MLFF) trained on \textit{ab-initio} MD data. The achieved MLFF helps efficient large-scale MD simulations on LPS with three Ti concentrations (10\%, 20\%, and 30\%) and six temperatures (25$^\mathrm{o}$C to 225$^\mathrm{o}$C) to calculate ionic conductivity, activation energy, Li-ion transport mechanism, and configurational entropy. Results show that ionic conductivities and activation energies are consistent with our recent experimental values. Moreover, Li-ion transport occurs via free-volume diffusion facilitated by the formation of disordered Li-S polyhedra. The enhanced stability of transport channels at 10\% and 20\% Ti doping, compared to 0\% and 30\%, is observed by analyzing the vibrational and configurational entropy of these disordered Li-S polyhedra. Overall, this study highlights the utility of MLFF-based large-scale MD simulations in explaining the transport mechanism and the stability of Li-ion in Ti-doped LPS electrolyte with significant computational efficiency.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 22 canonical work pages

  1. [13]

    D. Wang, B. Gwalani, D. Wierzbicki, V. Singh, L.J. Jhang, T. Rojas, R. Kou, M. Liao, L. Ye, H. Jiang, S. Shan, A. Silver, A.T. Ngo, Y. Du, X. Li, D. Wang, Overcoming the conversion reaction limitation at three-phase interfaces using mixed conductors towards energy-dense solid- state li–s batteries. Nature Materials (2025). https://doi.org/10.1038/s41563-0...

  2. [1]

    S. Li, Z. Yang, S.B. Wang, M. Ye, H. He, X. Zhang, C.W. Nan, S. Wang, Sulfide-based composite solid electrolyte films for all-solid-state batteries. Communications Materials 5(1), 44 (2024). https://doi.org/10.1038/s43246-024-00482-8. URL https://doi.org/10.1038/s43246-024-00482-8

  3. [2]

    S. Wang, Y. Zhang, X. Zhang, T. Liu, Y.H. Lin, Y. Shen, L. Li, C.W. Nan, High-conductivity argyrodite li6ps5cl solid electrolytes prepared via optimized sintering processes for all-solid-state lithium–sulfur batteries. ACS applied materials & interfaces 10(49), 42279–42285 (2018)

  4. [3]

    Kamaya, K

    N. Kamaya, K. Homma, Y. Yamakawa, M. Hirayama, R. Kanno, M. Yonemura, T. Kamiyama, Y. Kato, S. Hama, K. Kawamoto, A. Mitsui, A lithium superionic conductor. Nature Materi- als 10(9), 682–686 (2011). https://doi.org/10.1038/nmat3066. URL https://doi.org/10.1038/ nmat3066

  5. [4]

    Adeli, J.D

    P. Adeli, J.D. Bazak, K.H. Park, I. Kochetkov, A. Huq, G.R. Goward, L.F. Nazar, Boosting solid-state diffusivity and conductivity in lithium superionic argyrodites by halide substitution. Angewandte Chemie International Edition 58(26), 8681–8686 (2019). https://doi.org/https: //doi.org/10.1002/anie.201814222. URL https://onlinelibrary.wiley.com/doi/abs/10...

  6. [5]

    Y. Kato, S. Hori, T. Saito, K. Suzuki, M. Hirayama, A. Mitsui, M. Yonemura, H. Iba, R. Kanno, High-power all-solid-state batteries using sulfide superionic conductors. Nature Energy 1(4), 16030 (2016). https://doi.org/10.1038/nenergy.2016.30. URL https://doi.org/10.1038/nenergy. 2016.30

  7. [6]

    Liang, Y

    Z. Liang, Y. Xiang, K. Wang, J. Zhu, Y. Jin, H. Wang, B. Zheng, Z. Chen, M. Tao, X. Liu, Y. Wu, R. Fu, C. Wang, M. Winter, Y. Yang, Understanding the failure process of sulfide- based all-solid-state lithium batteries via operando nuclear magnetic resonance spectroscopy. Nature Communications 14(1), 259 (2023). https://doi.org/10.1038/s41467-023-35920-7. ...

  8. [7]

    D. Cao, Y. Zhang, T. Ji, X. Zhao, E. Cakmak, S. Ozcan, M. Geiwitz, J. Bilheux, K. Xu, Y. Wang, K.S. Burch, Q.H. Tu, H. Zhu, Li dynamics in mixed ionic-electronic conducting inter- layer of all-solid-state li-metal batteries. Nano Letters 24(5), 1544–1552 (2024). https://doi. org/10.1021/acs.nanolett.3c04072. URL https://doi.org/10.1021/acs.nanolett.3c0407...

Show all 34 references
  1. [8]

    S.Y. Kim, J. Li, Porous mixed ionic electronic conductor interlayers for solid- state batteries. Energy Material Advances 2021 (2021). https://doi.org/10. 34133/2021/1519569. URL https://spj.science.org/doi/abs/10.34133/2021/1519569. https://spj.science.org/doi/pdf/10.34133/20...

  2. [9]

    Kızılaslan, R

    A. Kızılaslan, R. Kızılaslan, A. Miura, K. Tadanaga, Mixed ion-electron conductive materials: A path to higher energy density all-solid-state lithium-ion batteries. Nano Today60, 102556 (2025). https://doi.org/https://doi.org/10.1016/j.nantod.2024.102556. URL https://www.scien...

  3. [10]

    Y. Lin, S. Fang, D. Su, K.S. Brinkman, F. Chen, Enhancing grain boundary ionic conductivity in mixed ionic–electronic conductors. Nature Communications 6(1), 6824 (2015). https://doi. org/10.1038/ncomms7824. URL https://doi.org/10.1038/ncomms7824

  4. [11]

    C. Ma, W. Jiang, Q. Duan, D. Ning, M. Wang, J. Wang, B. Chen, H. Jiang, C. Yang, W. Wu, Superdense lithium deposition via mixed ionic/electronic conduc- tive interfaces implanted in vivo/vitro for stable lithium metal batteries. Advanced Energy Materials 14(25), 2400202 (2024)...

  5. [12]

    Conlin, H

    P. Conlin, H. Kim, Y. Hu, C. Liang, M. Cho, K. Cho, Mechanism of mixed conduc- tivity in crystalline and amorphous lithium lanthanum titanate. Solid State Ionics 386, 116029 (2022). https://doi.org/https://doi.org/10.1016/j.ssi.2022.116029. URL https://www. sciencedirect.com/s...

  6. [14]

    Selvaraj, V

    S.C. Selvaraj, V. Koverga, A.T. Ngo, Exploring li-ion transport properties of li3ticl6: A machine learning molecular dynamics study. Journal of The Electrochemical Society 171(5), 050544 (2024). https://doi.org/10.1149/1945-7111/ad4ac9. URL https://dx.doi.org/10.1149/ 1945-7111/ad4ac9

  7. [15]

    Chandrasekaran, P

    S.S. Chandrasekaran, P. Murugan, Structural and electronic properties of solid-state (limpo4 |γ- li3po4)[010] electrochemical interface (m=fe and co). Applied Surface Science 418, 17–21 (2017). https://doi.org/10.1016/j.apsusc.2017.02.176. URL https://www.sciencedirect.com/ sc...

  8. [16]

    Chandrasekaran, M.R

    S.S. Chandrasekaran, M.R. Ponnaiah, P. Murugan, P. Saravanan, Magnetic and electronic prop- erties of hard|soft magnetic interface in (yco5|co)[0001] and (yfe5|co)[0001] superlattices. Journal of Magnetism and Magnetic Materials 418, 92–98 (2016). https://doi.org/10.1016/j.jmm...

  9. [17]

    J. Peng, J. Du, L. Wang, The study of ionic conductivity of the li10gep2s12 type solid state electrolyte by an extrapolation method and a deep-learning method. Journal of Physics: Confer- ence Series 2557(1), 012034 (2023). https://doi.org/10.1088/1742-6596/2557/1/012034. URL ...

  10. [18]

    Zhang, L

    W. Zhang, L. Zhou, T. Yan, M. Chen, Speciation of la3+–cl– complexes in hydrothermal fluids from deep potential molecular dynamics. The Journal of Physical Chemistry B 127(41), 8926– 8937 (2023). https://doi.org/10.1021/acs.jpcb.3c05428. URL https://doi.org/10.1021/acs.jpcb. 3...

  11. [19]

    Hirel, Atomsk: A tool for manipulating and converting atomic data files

    P. Hirel, Atomsk: A tool for manipulating and converting atomic data files. Computer Physics Communications 197, 212–219 (2015). https://doi.org/https://doi.org/10.1016/j.cpc.2015.07

  12. [20]

    URL https://www.sciencedirect.com/science/article/pii/S0010465515002817

  13. [21]

    Kresse, J

    G. Kresse, J. Hafner, Ab initio molecular dynamics for liquid metals. Phys. Rev. B 47, 558– 561 (1993). https://doi.org/10.1103/PhysRevB.47.558. URL https://link.aps.org/doi/10.1103/ PhysRevB.47.558

  14. [22]

    Kresse, J

    G. Kresse, J. Hafner, Ab initio molecular-dynamics simulation of the liquid-metal–amorphous- semiconductor transition in germanium. Phys. Rev. B 49, 14251–14269 (1994). https://doi.org/ 10.1103/PhysRevB.49.14251. URL https://link.aps.org/doi/10.1103/PhysRevB.49.14251

  15. [23]

    Bl¨ ochl, Projector augmented-wave method

    P.E. Bl¨ ochl, Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994). https: //doi.org/10.1103/PhysRevB.50.17953. URL https://link.aps.org/doi/10.1103/PhysRevB.50. 17953

  16. [24]

    Perdew, K

    J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865–3868 (1996). https://doi.org/10.1103/PhysRevLett.77.3865. URL https: //link.aps.org/doi/10.1103/PhysRevLett.77.3865

  17. [25]

    Burke, J.P

    K. Burke, J.P. Perdew, Y. Wang, Derivation of a Generalized Gradient Approximation: The PW91 Density Functional (Springer US, Boston, MA, 1998), pp. 81–111. https://doi.org/10. 1007/978-1-4899-0316-7 7. URL https://doi.org/10.1007/978-1-4899-0316-7 7

  18. [26]

    J. Zeng, D. Zhang, D. Lu, P. Mo, Z. Li, Y. Chen, M. Rynik, L. Huang, Z. Li, S. Shi, et al., Deepmd-kit v2: A software package for deep potential models. arXiv preprint arXiv:2304.09409 (2023)

  19. [27]

    Abadi, A

    M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G.S. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Isard, Y. Jia, R. Joze- fowicz, L. Kaiser, M. Kudlur, J. Levenberg, D. Man´ e, R. Monga, S. Moore, D. Murray, C. Olah, M...

  20. [28]

    Zhang, J

    L. Zhang, J. Han, H. Wang, W.A. Saidi, R. Car, E. Weinan, End-to-End Symmetry Preserving Inter-Atomic Potential Energy Model for Finite and Extended Systems, in Proceedings of the 17 32nd International Conference on Neural Information Processing Systems (Curran Associates Inc....

  21. [29]

    Thompson, H.M

    A.P. Thompson, H.M. Aktulga, R. Berger, D.S. Bolintineanu, W.M. Brown, P.S. Crozier, P.J. in ’t Veld, A. Kohlmeyer, S.G. Moore, T.D. Nguyen, R. Shan, M.J. Stevens, J. Tranchida, C. Trott, S.J. Plimpton, Lammps - a flexible simulation tool for particle-based materials mod- elin...

  22. [31]

    J. Zeng, D. Zhang, D. Lu, P. Mo, Z. Li, Y. Chen, M. Rynik, L. Huang, Z. Li, S. Shi, Y. Wang, H. Ye, P. Tuo, J. Yang, Y. Ding, Y. Li, D. Tisi, Q. Zeng, H. Bao, Y. Xia, J. Huang, K. Muraoka, Y. Wang, J. Chang, F. Yuan, S.L. Bore, C. Cai, Y. Lin, B. Wang, J. Xu, J.X. Zhu, C. Luo,...

  23. [32]

    H. Wang, L. Zhang, J. Han, W. E, Deepmd-kit: A deep learning package for many-body potential energy representation and molecular dynamics. Computer Physics Communications 228, 178–184 (2018). https://doi.org/https://doi.org/10.1016/j.cpc.2018.03.016. URL https: //www.sciencedi...

  24. [33]

    Momma, F

    K. Momma, F. Izumi, Vesta: a three-dimensional visualization system for electronic and struc- tural analysis. Journal of Applied Crystallography 41(3), 653–658 (2008). https://doi.org/https: //doi.org/10.1107/S0021889808012016. URL https://onlinelibrary.wiley.com/doi/abs/10.11...

  25. [34]

    Stukowski, Visualization and analysis of atomistic simulation data with ovito–the open visu- alization tool

    A. Stukowski, Visualization and analysis of atomistic simulation data with ovito–the open visu- alization tool. Modelling and Simulation in Materials Science and Engineering 18(1), 015012 (2009). https://doi.org/10.1088/0965-0393/18/1/015012. URL https://dx.doi.org/10.1088/ 09...

  26. [35]

    EV ANS, G.P

    D.J. EV ANS, G.P. MORRISS, Statistical Mechanics of Nonequilibrium Liquids (ANU Press, 2007). URL http://www.jstor.org/stable/j.ctt24h99q 18

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.