Pith. sign in

REVIEW 4 major objections 5 minor 22 references

Suitability of ReaxFF potential for MD modelling of lithium across low and high temperatures

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read For MD models of lithium, ReaxFF is reliable below 800 K but loses about 10% of the density by 1000 K and overestimates diffusion, drifting into an artificially open liquid structure.

desk verdict Useful, controlled benchmark of Li potentials, but the headline ReaxFF claim is attached to an unspecified parameter file, so the paper needs a revision before the high-T failure can be taken as a property of ReaxFF rather than of one parameterization. read the letter →

arxiv 2509.00807 v1 pith:FVMZZQP2 submitted 2025-08-31 physics.chem-ph

Predrag S. Krstic , Swarit Dwivedi , Adri C.T. van Duin , Bruce E. Koel This is my paper · ORCID
classification physics.chem-ph PACS 61.25.Mv71.15.Pd
keywords ReaxFFlithiummoleculardynamicsinteratomicpotentialbenchmarkliquiddensityself-diffusionradialdistributionfunctionplasma-facingmaterials
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

Lithium matters for two applied problems — liquid plasma-facing walls in fusion reactors and battery electrodes — and molecular dynamics is how its atomic-scale behavior gets modeled, so the choice of interatomic potential is a practical question with real consequences. The paper asks whether ReaxFF, the reactive force field with bond-breaking and charge-transfer capability that makes it attractive for lithium chemistry, stays trustworthy across the 100–1000 K range those applications cover, and it benchmarks ReaxFF against experimental density, diffusivity, and structure data, against DFT molecular dynamics, and against two other potentials under controlled cooling and equilibration protocols. The answer is a temperature ceiling: ReaxFF captures glassy and liquid lithium well below roughly 800 K, but above that it underestimates density by about 10% at 1000 K and overestimates self-diffusion, with the second shell of the radial distribution function nearly vanishing — an open, low-density-liquid-like structure the authors treat as a likely artifact or exaggeration rather than established physics. The practical upshot is that ReaxFF remains a qualitative tool for disordered lithium at high temperature but needs experimental or ab initio anchoring for any quantitative claim there.

What carries the argument

The argument runs on four observables: density versus temperature; self-diffusion from the mean-squared displacement via the Einstein relation; radial distribution functions g(r) with 0.01–0.05 Å binning; and coordination numbers from integrating g(r) to its first minimum. The density curve does the heavy lifting: experiment and the 2NN-MEAM potential agree across the range, DFT overshoots by under 10%, and ReaxFF goes into a growing deficit above 800 K. The RDF's second-shell amplitude names the defect — at 1000 K the second peak nearly vanishes, the signature of the open low-density-liquid-like state the paper attributes to ReaxFF's softened high-temperature cohesion.

What would settle it

Equilibrate ReaxFF lithium at 900 and 1000 K starting from a melted crystal or a slowly annealed liquid seed — not a quenched glass — and measure the equilibrium density against Yakimovich's experimental curve. If the roughly 10% deficit persists across preparation routes, the divergence is intrinsic to ReaxFF; if it shrinks, the paper's 800 K boundary is partly an artifact of its quench protocol. In parallel, an ab initio MD run at 1000 K that reproduces the vanishing second RDF shell would confirm the low-density-liquid-like structure is physical rather than a ReaxFF artifact.

Watch

Extended reading notes

Core claim

ReaxFF, the reactive force field prized for lithium's bond-breaking chemistry, reproduces glassy and liquid lithium well from 100 K up to about 800 K — matching experimental RDFs, coordination numbers, density, and diffusivity, and never spuriously crystallizing. Above 800 K it under-binds the liquid: density falls to about 10% below experiment at 1000 K, self-diffusion runs too fast, coordination drops below 12, and the second RDF shell nearly vanishes. The paper reads this as ReaxFF's angular flexibility and weakened short-range cohesion letting the hot liquid relax into an artificially open, low-density-liquid-like structure, and it cautions that this LDL-like phase should not be treated

Load-bearing premise

The load-bearing assumption is that a sample rapidly quenched from 1100 K into a glass and relaxed for 100 ps stands in for lithium in equilibrium at the target temperature — the paper compares these glassy and supercooled structures against experimental data for stable liquid and crystalline lithium, so if the quench history is not fully erased, part of the reported divergence above 800 K is a preparation artifact rather than a property of ReaxFF.

Editorial extensions

If this is right

  • Below 800 K, ReaxFF can be used without external benchmarking for relative structure and dynamics of glassy and liquid lithium, including fusion-relevant low-temperature surface chemistry.
  • Above 800 K, ReaxFF-based predictions of density-linked properties — thermal expansion, pressure response, transport — carry an error of order 10% and must be checked against experiment or ab initio MD before being trusted.
  • The open, low-density-liquid-like structure seen at 1000 K should not be reported as a liquid-liquid transition without confirmation from neutron diffraction or DFT-MD; the paper itself flags the over-interpretation risk.
  • For crystalline and ordered regimes, 2NN-MEAM matches experimental density and diffusivity closely and is the safer default, while SNAP fails on density and crystallizes too readily to be useful here.
  • The quench-and-equilibrate protocol is part of the result: since every sample begins as a glass quenched from 1100 K, the preparation route must be reported alongside potential choice in any lithium MD study.

Reading between the lines

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

  • Because every ReaxFF sample starts as a glass quenched from 1100 K and never crystallizes, the 470 K and 725 K 'liquid' structures are metastable supercooled states being compared against equilibrium molten-lithium experiments; the paper does not separate how much of the high-temperature divergence is the potential versus the quench protocol.
  • A direct way to separate the two: initialize ReaxFF at 900–1000 K from a melted, slowly equilibrated crystal rather than a quenched glass; a shrinking density deficit would pin part of the 800 K boundary on sample preparation, while a persistent deficit would confirm the defect is intrinsic to the parameterization.
  • The results implicitly define a target for future ReaxFF lithium parameterization — add high-temperature cohesive strength or soften short-range repulsion until the 1000 K density and diffusivity match experiment, then verify the low-temperature glassy agreement survives.
  • Since the motivation is plasma-facing lithium in contact with hydrogen isotopes, the natural next benchmark is Li–H mixtures: if the same open-structure bias carries into lithium hydride systems, fusion-relevant plasma-material interaction predictions inherit the high-temperature error.
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 benchmarks three classical interatomic potentials (ReaxFF, Cui 2NN-MEAM, and Zuo SNAP) in molecular dynamics simulations of lithium from 100 to 1000 K. Using configurations rapidly quenched from 1100 K and NPT equilibration, it reports density, self-diffusivity, radial distribution functions, and coordination numbers. The central claim is that ReaxFF matches experimental behavior below roughly 800 K but underestimates density by about 10% at 1000 K and overestimates diffusivity, producing an overly disordered, low-density liquid-like state. The authors conclude that ReaxFF is a useful qualitative tool for disordered lithium but not quantitatively reliable at high temperatures, and they caution against overinterpreting possible liquid-liquid transition signatures without validation.

Significance. If the central claim holds, the paper provides practically useful guidance for choosing interatomic potentials for lithium simulations in fusion and battery contexts. The study uses independent experimental benchmarks (density, diffusivity, RDF) and includes an honest discussion of limitations. Strengths include the multi-potential comparison, the explicit connection to experimental data, and the clear statement that ReaxFF's high-temperature failure is parameterization-dependent. However, the central result is currently attached to an unidentified ReaxFF parameter file, and the statistical basis is a single quenched trajectory per temperature. Both points must be addressed before the claim can be considered reproducible and generalizable.

major comments (4)
  1. [§2, refs [5-7]] The central claim is that 'ReaxFF' underestimates Li density above 800 K, but ReaxFF is a framework, not a single potential. Ref [5] is a LiF parameterization, ref [6] a Li-battery ReaxFF, and ref [7] a lithiated-sulfur parameterization; the manuscript never identifies which LAMMPS ffield file, pair_coeff settings, charges, or training data were used. Figures 3-5 therefore report results for an unidentified potential. Section 4 itself attributes the high-temperature failure to 'the force field's parameterization,' making the parameter file a load-bearing variable. The exact parameterization and input scripts must be given; otherwise the result cannot be reproduced, tested, or generalized beyond one unspecified file.
  2. [§2, Figs. 3-4] Each temperature is represented by a single quenched trajectory, a single 100 ps NPT equilibration, and a single 100 ps production run. No error bars or repeat simulations are reported. Diffusivities extracted from MSD slopes over 100 ps carry statistical uncertainty, and the reported ~10% density discrepancy at 1000 K could plausibly lie within run-to-run scatter for a glassy sample. The authors should provide block-error estimates, multiple independent quenches, or a convergence check to demonstrate that the high-temperature divergence is not an artifact of one thermal history.
  3. [§3.2, Fig. 6] The ReaxFF samples are prepared by rapid quenching from 1100 K and are described as glassy/amorphous or supercooled, yet they are compared against equilibrium liquid experimental RDFs at 470 K and 725 K (Olbrich et al.) and against liquid DFT-MD results. A glassy or supercooled sample is not the same thermodynamic state as an equilibrium liquid at the same nominal temperature. This conflation is load-bearing for the conclusion that ReaxFF 'diverges' above 800 K: the observed discrepancies could partly stem from preparation state rather than potential error. The authors should either demonstrate equilibration to the equilibrium liquid (e.g., density plateau and RDF convergence over longer or repeated runs) or compare with deliberately glassy/amorphous benchmarks.
  4. [§3.3, Eq. (2), Fig. 9] Coordination numbers are obtained by integrating the RDF to the first minimum, but the authors note that at high temperatures the minimum is shallow and wide, and Eq. (2) depends sensitively on the chosen cutoff rc. The choice of rc is a free parameter and can change CN by several units when the first minimum is ambiguous. Before interpreting ReaxFF's 'drop below 12' at 1000 K as a signature of an LDL-like state, a sensitivity analysis of CN to rc is needed. In addition, the experimental CN benchmarks cited (Salmon, Waseda) are for equilibrium or supercooled liquid lithium, not necessarily the same glassy state produced by the rapid-quench protocol used here.
minor comments (5)
  1. [§3.3, Eq. (2)] Typo: 'rc is he cutoff' should be 'rc is the cutoff radius'.
  2. [§3.3, LDL paragraph] The text says 'confirmed at 1000 K (see Fig. R4)', but no Figure R4 exists; this should be Fig. 5.
  3. [Fig. 4 caption] The caption cites 'DFT-MD [6]', but the DFT-MD results discussed in the text are from Chen et al. [15]. Please correct the reference.
  4. [Abstract and §2] The abstract states that 'ensemble and cooling protocols' are 'carefully controlled,' but the paper uses one rapid-quench protocol per temperature. Please temper the wording to match the actual protocol.
  5. [Fig. 1 caption] Minor grammar: 'Thermodynamics features' should be 'Thermodynamic features'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; benchmark conclusions are anchored to independent experimental and DFT data.

full rationale

The paper makes no derivation claim: it is a benchmark study. Its headline result (ReaxFF underestimates liquid Li density by ~10% at 1000 K and overestimates diffusivity) is presented as a comparison against independent measurements. Section 3.1 states the densities are 'compared with experimental measurements from Yakimovich et al.' and the diffusivities are 'compared with experimental results of Yakimovich and Murday.' These external benchmarks are not constructed from the force fields nor from any self-citation. The RDF and coordination analyses are likewise compared with independent neutron/X-ray diffraction data (Olbrich et al., Salmon et al.) and with DFT-MD (Chen et al.). The only overlapping-author citation, ref. [11] (Krstic/Ostrowski/Dwivedi), is used only to justify the quench protocol ('This method ensured that initial conditions were consistently disordered and free from crystalline bias... [10,11]') and does not carry the density/diffusivity conclusions. The paper's Discussion even attributes the high-temperature divergence to 'the force field's parameterization' and explicitly warns that the LDL-like features may be artifacts, which is the opposite of circular reasoning. The skeptical issue that no specific ReaxFF parameter file or training set is identified, so the claim is attached to an unnamed ReaxFF instance, is a reproducibility/under-specification concern, not circularity: nothing in the paper defines the ReaxFF density in terms of the experimental density. No load-bearing reduction of the conclusion to its own inputs is exhibited, so no circular step rises above the reporting threshold.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

No new model parameters are fitted in this paper; the central claim depends on the quality of the chosen interatomic potentials, the experimental benchmarks, and the equilibration protocol.

free parameters (1)
  • Coordination-number cutoff rc = First minimum of g(r), varies with temperature
    Chosen by hand for each temperature; the paper acknowledges CN sensitivity to rc in Sec. 3.3.
assumptions (4)
  • domain assumption The experimental density and diffusivity data (Yakimovich, Murday, Olbrich, Salmon) are accurate ground truth.
    Used as benchmarks in Figs. 3, 4, and 6.
  • domain assumption Quenching from 1100 K, followed by 100 ps NPT equilibration, produces a representative state for the target temperature.
    Protocol in Sec. 2; at 470-725 K the sample is supercooled liquid, not equilibrium liquid, yet compared to equilibrium experimental RDFs.
  • ad hoc to paper The ReaxFF potential used corresponds to the intended lithium parameterization.
    Section 2 cites ReaxFF for Li systems but does not specify the exact parameter file or version; results depend on it.
  • standard math The Einstein relation and RDF integration are valid for extracting diffusivity and coordination in these systems.
    Eqs. 1 and 2, Secs. 2 and 3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Suitability of ReaxFF potential for MD modelling of lithium across low and high temperatures." pith.science (2026). https://pith.science/paper/FVMZZQP2

@misc{pith2026250900807,
  author       = {Pith},
  title        = {Pith review of: Suitability of ReaxFF potential for MD modelling of lithium across low and high temperatures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FVMZZQP2}},
  note         = {Machine review of arXiv:2509.00807}
}
read the original abstract

Modeling lithium's atomic-scale behavior is critical for its roles in plasma-facing fusion components and lithium-ion batteries, yet it remains challenging across phase regimes. This study benchmarks ReaxFF, 2NN-MEAM, and SNAP potentials from 100 to 1000 K using molecular dynamics, with ensemble and cooling protocols carefully controlled. Compared to reliable experimental data, ReaxFF diverges above 800 K, underestimating density by ~10% at 1000 K and overestimating diffusivity, yielding anomalously low-density liquid behavior. Radial distribution functions and coordination profiles further reveal possible excessive disorder above 800 K. While ReaxFF offers qualitative insight into disordered lithium, its quantitative reliability diminishes at high temperatures, requiring validation against ab initio or experimental benchmarks. These findings inform potential selections for fusion- and battery-relevant simulations, underscoring the sensitivity of glassy phase modeling to potential choices and thermal histories.

Figures

Figures reproduced from arXiv: 2509.00807 by the authors.

Figure 2
Figure 2. a) Diffusivity is derived from MSD(t) slope during the stability of PEA and temperature; crystallization is seen at ≤200K (c and d), while amorphous features (e) predominate at higher temperatures. We do not pursue further analysis of the Zuo SNAP potential, since the resultant density as a function of temperature does not agree well with the experiment in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 8
Figure 8. Comparisons of RDFs calculated by ReaxFF with those from Cui’s 2NN [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [5]

    De Angelis M., et al., Development and validation of a ReaxFF potential for lithium fluoride systems, Sci. Rep. 13(2023) 50978

    work page 2023

  2. [6]

    Shi S., et al., Applications of ReaxFF to lithium battery systems, Front. Chem. 8 (2021) 634379

    work page 2021

  3. [7]

    Islam M.M., et al., ReaxFF molecular dynamics simulations of lithiated sulfur cathodes, Phys. Chem. Chem. Phys. 17(2015) 31217–31226

    work page 2015

  4. [1]

    basis for the ITER divertor, Nucl

    Pitts R.A., et al., Phys. basis for the ITER divertor, Nucl. Fusion 51 (2011) 073007

    work page 2011

  5. [2]

    Baldwin T., et al., Liquid lithium divertor development for NSTX, J. Nucl. Mater. 390–391 (2009) 886–889

    work page 2009

  6. [3]

    Shimada M., et al., Plasma–material interactions in fusion devices, J. Nucl. Mater. 438 (2013) S996–S1001

    work page 2013

  7. [4]

    Plasma Phys

    Schmidt K., von Toussaint U., Atomic-scale modeling of plasma–wall interactions, Contrib. Plasma Phys. 50 (2010) 700–705

    work page 2010

  8. [8]

    Kresse G., Hafner J., Ab initio molecular dynamics for liquid metals, Phys. Rev. B 47 (1993) 558–561

    work page 1993

Show all 22 references
  1. [10]

    38 (2000)

    Yakimovich K.A., Mozgovoi A.G., Experimental investigation of the density and surface tension of molten lithium at temperatures up to 1300 K, High Temp. 38 (2000)

    2000

  2. [11]

    Krstic P.S., Ostrowski E.T., Dwivedi S., Processes at lithium-hydride/deuteride surfaces upon low energy impact of H/D, Front. Phys. 11 (2023) 1105194

    2023

  3. [12]

    Zuo Y., Chen C., Li X., Deng Z., Chen Y., Behler J., Csányi G., Shapeev A.V., Thompson A.P., Wood M.A., Ong S.P., Performance and cost assessment of machine learning interatomic potentials, J. Phys. Chem. A 124 (2020) 731–745

    2020

  4. [13]

    Cui Z., Gao F., Cui Z., Qu J., Developing a second nearest neighbor modified embedded atom method interatomic potential for lithium, Model. Simul. Mater. Sci. Eng. 20 (2012) 015014. 15

    2012

  5. [14]

    Power Sources 207 (2012) 150–159

    Cui Z., Gao F., Cui Z., A second nearest-neighbor embedded atom method interatomic potential for lithium–silicon (Li–Si) alloys, J. Power Sources 207 (2012) 150–159

    2012

  6. [15]

    Fusion 56 (2016) 016020

    Chen M., Abrams T., Jaworski M., et al., Rock-salt structure lithium deuteride formation in liquid lithium with high concentrations of deuterium: a first-principles molecular dynamics study, Nucl. Fusion 56 (2016) 016020

    2016

  7. [16]

    Baskes M.I., Modified embedded-atom potentials for cubic materials and impurities, Phys. Rev. B 46 (1992) 2727–2742

    1992

  8. [17]

    Kim Y., Jung I., Lee B., Atomistic modeling of pure Li and Mg–Li system, Model. Simul. Mater. Sci. Eng. 20 (2012) 035005

    2012

  9. [18]

    51 (2013) 626–639

    Belashchenko D.K., Impact compression of alkali metal: computer aided simulation, High Temp. 51 (2013) 626–639

    2013

  10. [19]

    Murday J.S., Cotts R.M., Self-diffusion in liquid lithium, J. Chem. Phys. 48 (1968)

    1968

  11. [20]

    Naturforsch

    Olbrich H., Ruppersberg H., Steeb S., Structure of molten lithium, Z. Naturforsch. A 38 (1983) 1328–1336

    1983

  12. [21]

    Salmon P.S., Petri I., de Jong P.H.K., Verkerk P., Fischer H.E., Howells W.S., Structure of liquid lithium, J. Phys. Condens. Matter 16 (2004) 195–222

    2004

  13. [22]

    Mater.19 (2020) 1339–1345

    Wang X., Pawar G., Li Y., et al., Glassy Li metal anode for high-performance rechargeable Li batteries, Nat. Mater.19 (2020) 1339–1345

    2020

  14. [23]

    Blagoveshchenskii N.M., Noikov A., Savostin V., Self-diffusion in liquid lithium from coherent quasielastic neutron scattering, Physica B 407 (2012)

    2012

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.