Reply to Smallenburg: Near-melting nucleation and the exponential growth of hard-sphere nucleation times

Roseanna N. Zia

REVIEW 4 major objections 5 minor 9 references

Reply to Smallenburg: Near-melting nucleation and the exponential growth of hard-sphere nucleation times

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Hard-sphere crystallization waits grow 6-fold per 0.1% density drop, making spontaneous coexistence reachable only near melting.

desk verdict A mostly correct reply that overreaches in its table: the qualitative exponential-growth claim stands, but the constant-slope extrapolation is not a valid predictive benchmark. read the letter →

arxiv 2608.07644 v1 pith:EYN3VLL3 submitted 2026-08-07 cond-mat.soft

Roseanna N. Zia This is my paper · ORCID

classification cond-mat.soft

keywords hard-spherecolloidsnucleationbarriervolumefractionentropy-exchangemechanismcrystalfluid-crystalcoexistencesupersaturationsimulationtimescales

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 reply to a Comment defends a sharp claim about pristine hard-sphere simulations: spontaneous fluid-crystal coexistence is dynamically accessible only in a narrow window very close to the melting volume fraction, around 53%. The author argues that the Comment's successful nucleation run at a single near-melting packing fraction is the easy limiting case, not evidence that coexistence is readily achievable across the coexistence region. The central evidence is an exponential growth law: the nucleation time increases by roughly a factor of six for every 0.1% decrease in volume fraction, rising from about 1.5 seconds at φ=0.535 to 16,000 years at φ=0.520 and effectively infinite by φ≈0.505. If this is right, the long-observed absence of spontaneous coexistence in simulations is the expected outcome of the entropy-exchange mechanism, not a numerical artifact.

What carries the argument

The load-bearing object is Frenkel's entropy-exchange mechanism, defined as the competition between configurational and vibrational entropy that sets the nucleation barrier and makes it rise sharply as supersaturation increases. The reply quantifies this with a mapping from the atomic Lennard-Jones system to colloidal hard spheres, treating the hexagonal-close-packed volume fraction φ_HCP=0.74 as the analog of absolute zero, with melting at φ_M=0.543 and freezing at φ_F=0.4918; this maps 20% supersaturation to φ≈0.541, or φ≈0.536 by an independent estimate. The exponential growth law itself comes from a linear fit to the reduced nucleation rates measured in the 2004 hard-sphere study, giving roughly a 6-fold increase in nucleation time per 0.1% decrease in volume fraction, which the reply extrapolates across the coexistence region.

What would settle it

A pristine, unbiased hard-sphere or nearly hard-sphere colloid simulation at φ=0.52 that produces a stable crystal nucleus within, say, a month of simulated time would contradict the predicted 16,000-year waiting time; conversely, rate measurements at φ=0.525, 0.52, and 0.515 that show the 6-fold-per-0.1% factor persisting would support the extrapolation.

Watch

Extended reading notes

Core claim

The paper's central claim is that the nucleation time of monodisperse, purely repulsive hard spheres grows exponentially as the system moves away from the melting point: a 0.1% reduction in volume fraction multiplies the waiting time by about six. Using the absolute crystallization rate measured near φ=0.534 as a baseline, the author converts that growth law into a table of laboratory waiting times for a colloidal-scale simulation, showing that nucleation takes about 1.5 seconds at φ=0.535, 2.9 hours at φ=0.53, 16,000 years at φ=0.52, and becomes practically unattainable at lower densities. The same entropy-exchange mechanism that produces this barrier also explains why the Comment's simulation at φ=0.5325 succeeded: that state point sits just inside the narrow near-melting window where the barrier is low enough for spontaneous nucleation on feasible timescales. The reply concludes that the Comment's broader takeaway, that equilibrium coexistence is 'readily achievable,' is an overgeneralization from a single favorable state point.

Load-bearing premise

The exponential waiting-time table assumes that the growth rate measured near φ=0.534 continues unchanged all the way down to φ=0.495, over an 11-order-of-magnitude range, with no error bars on the fit.

Editorial extensions

If this is right

Spontaneous, unbiased coexistence simulations of hard spheres are feasible only above roughly φ=0.53; below that, waiting times exceed what any current simulation can reach.
A single near-melting state point cannot be used to infer that equilibrium coexistence is generally easy to reach; the Comment's own data (8 of 50 slab runs and 11 of 50 cubic runs crystallized) support the narrow-window picture.
Colloidal-scale simulations nucleate far faster than atomic-scale ones near melting, about 1.5 seconds versus years, but retain the same exponential sensitivity to volume fraction.
The lever-rule prediction of about 80% final crystal fraction requires simulation durations much longer than used in the Comment, so scattered final fractions likely reflect incomplete convergence rather than the equilibrium value.

Reading between the lines

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

The 6-fold-per-0.1% law is asserted from a linear fit over an 11-order-of-magnitude extrapolation; a direct measurement of nucleation rates at intermediate volume fractions, say φ=0.525 to 0.53, would test whether the growth rate itself changes with supersaturation.
The supersaturation mapping crucially depends on assigning HCP as the zero-motion reference for colloids; if that reference were revised, the nominal 20% supersaturation point would shift, moving the predicted fast-nucleation window without changing the exponential structure.
If the entropy-exchange argument is correct, reported spontaneous coexistence in other entropy-dominated systems should likewise cluster near their melting points, and any claim of rapid coexistence far from melting would warrant scrutiny.
The reply's table assumes a fixed system volume; because nucleation rates scale with volume, a larger simulation box could in principle bring lower-volume-fraction nucleation into reach, a testable prediction.

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.

Desk Editor's Note

read the letter

The reply makes a solid qualitative point: Smallenburg's single simulation at φ=0.5325, right next to melting, does not license the claim that coexistence is "readily achievable" throughout the coexistence region. The exponential growth of nucleation times as you move away from 53% is real—Auer and Frenkel measured it—and the paper is right to push back on that overgeneralization. It also credits its own earlier Perspective for being unclear on this point, which is honest.

What's useful here: the paper frames φ≈0.53 as the practical departure point for accessible spontaneous nucleation, translates Auer and Frenkel's reduced rates into dimensional lab times for a colloidal system, and works through what that means for simulation design. That interpretive work is a service, even if the underlying law is not new.

Now the soft spots, in proportion. Table 1 is the main problem. It takes a constant "6-fold per 0.1%" slope, fit near φ=0.534, and extends it linearly down to φ=0.495—eleven orders of magnitude. No fit details, no error bars, no residuals. Worse, the constant-slope assumption is inconsistent with classical nucleation theory, which gives log I* ∝ -A/(φ-φ_F)^2, so the logarithmic slope grows as you approach freezing. That means the true waiting times at φ=0.52 and below are even longer than the table says. So the qualitative conclusion—that spontaneous phase separation is dynamically inaccessible through most of the coexistence region—survives and is actually strengthened. But the specific numbers in Table 1 are not a reliable benchmark. They should be presented as illustrative, not predictive, or the fit should be replaced with a proper CNT-based extrapolation with uncertainties.

Two smaller issues. The supersaturation mapping "% supercooling = (φ-φ_F)/(φ_HCP-φ_F)" is asserted without derivation; it may be reasonable, but it needs a citation or a few lines of justification. And for a Reply, the paper never actually cites Smallenburg's Comment in the reference list. That's a noticeable omission, even if the journal version will include it.

Who is this for? People working on hard-sphere or colloidal phase behavior and anyone designing coexistence simulations. It's a short, argumentative reply, not a major new result. But the caution it delivers is important, and the qualitative argument is grounded in prior published rates.

Recommendation: it deserves peer review as a reply, because it corrects a potential misconception and the central claim holds up. A serious referee should, however, insist that Table 1 be reframed or recomputed before publication. The paper is fine to publish after that revision.

Referee Report

4 major / 5 minor

Summary. This manuscript is a Reply to a Comment by Smallenburg on the authors' recent Perspective. The Reply argues that Smallenburg's conclusion that equilibrium hard-sphere coexistence is 'readily achievable' in simulation is an overgeneralization from a single high-density state point (φ=0.5325). The authors reiterate their central claim that nucleation times in monodisperse, purely repulsive hard spheres grow exponentially as the volume fraction decreases from about φ=0.53, based on Auer and Frenkel's measured nucleation rates. They present Table 1, which lists waiting times from 1.5 seconds at φ=0.535 to 'never' at φ=0.495, obtained by extrapolating a 6-fold increase per 0.1% volume-fraction decrease. They also propose a mapping between atomic supersaturation and colloidal volume fraction, identifying 20% supersaturation with φ≈0.541 or 0.536, and they critique Smallenburg's simulation statistics as not establishing the theoretical 80% crystal fraction.

Significance. If the exponential sensitivity claim is correct, it has practical consequences for the design of hard-sphere simulations: unbiased spontaneous nucleation would be feasible only in a narrow window very close to melting. The Reply credibly emphasizes that even a modest shift in volume fraction changes nucleation times by many orders of magnitude, a point grounded in established data (Auer and Frenkel, Ref. [6]). The manuscript also honestly acknowledges that the original Perspective was unclear about the range over which the long-time claim applies. However, the quantitative predictions in Table 1 and the supersaturation mapping are not adequately supported, and these are the load-bearing elements of the Reply's quantitative argument.

major comments (4)

[Table 1 and the paragraph beginning 'Now, Frenkel's study further shows...'] Table 1 is presented as the central quantitative evidence for 'astronomically long' nucleation times, but it is based on a linear fit to Fig. 11 of Ref. [6] with no documented fit range, no residuals, and no error bars. The table extrapolates a constant 6-fold-per-0.1% factor over 11 orders of magnitude, from φ=0.5342 to φ=0.495. Classical nucleation theory predicts that the logarithmic slope should increase as the freezing point is approached, so a constant-slope fit from the melting-side data is not a reliable predictor at lower volume fractions. The specific entries in Table 1 should therefore be labeled as an illustrative extrapolation, or the table should be restricted to the volume-fraction range over which the cited linear fit is actually valid.
[Paragraph beginning 'It is straightforward to translate...'] The mapping %supercooling = (φ−φ_F)/(φ_HCP−φ_F) is asserted without derivation or comparison to any direct measurement. The physical analogy is not self-evident: Brownian motion does not vanish at φ_HCP, and the choice of φ_HCP as the analogue of absolute zero is not justified. This mapping is used to identify 20% supersaturation with φ≈0.541 (or 0.536) and underlies the statement that spontaneous phase separation is dynamically accessible only down to about 53% volume fraction. In its current form the mapping is an unsupported assertion; the Reply should either provide a derivation and justification or rely on the direct Auer–Frenkel data instead.
[Paragraphs 2 and 4 (laboratory-time estimates)] The Reply gives two widely different laboratory-time estimates for the authors' simulation box: about 2 years using the atomic nucleation rate of 1 nucleus cm⁻³ s⁻¹ at 20% supersaturation, and 9.1 seconds using the Auer–Frenkel colloidal rate at φ=0.5342. The discrepancy spans several orders of magnitude and is not reconciled. The phrase 'faster for colloids' is insufficient to explain the gap, and it is not clear which estimate is appropriate for which volume fraction. The Reply should clarify the relationship between the atomic and colloidal rates and avoid juxtaposing the two estimates without explicit reconciliation.
[Penultimate paragraph (critique of Smallenburg's data)] The Reply dismisses the significance of Smallenburg's simulation results by noting that only 8 of 50 slab runs and 11 of 50 cubic runs produced crystallization, with final crystal fractions ranging from 56% to 100%. However, no statistical context is provided: there is no definition of what constitutes 'any crystallization,' no information on simulation duration, and no error bars on the fractions. Without such context, the statement that the results 'do not establish that the 80% theoretical value ... has been achieved consistently or spontaneously' is itself unsupported. This critique is a secondary point in the Reply but, as written, it does not meet the rigor expected for dismissing a commentator's simulation evidence.

minor comments (5)

[Throughout] The manuscript contains several typographical errors and formatting glitches, including 'reportsnear-melting pointsimulations' (missing spaces), '10 −16cm3' (missing superscripts), 'ϕ= 0.5342%of' (stray percent sign), and 'Table inFigure 1shows' (no Figure 1 is actually included; only Table 1 appears). These should be corrected before publication.
[Paragraph 2, reference [5]] The text says 'Independent calculations cited by ten Wolde et al.' but the reference given is to Kelton (Ref. [5]), not to ten Wolde et al. This should be clarified to indicate the original source of the φ=0.536 value.
[Table 1 caption] The caption states that the table is 'calculated based on a linear fit to the data from Figure 11 in Ref. [6]' but does not give the fit parameters, the range of the fit, or the goodness of fit. Even if the fit details remain in the text, the caption should at least state the confidence interval or the range of validity.
[Terminology] The Reply uses the term 'supercooling' for colloids, but the relevant variable is volume-fraction supersaturation. Consider using a consistent term such as 'supersaturation' throughout, and number the defining equation for clarity.
[Language] The phrase 'spotty results' in the penultimate paragraph is informal and should be replaced with a more neutral description in a formal journal reply.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central exponential-growth claim rests on external Auer-Frenkel data, with the authors' own results used only as corroboration.

full rationale

The reply's central quantitative claim—that hard-sphere nucleation times grow by about 6-fold per 0.1% volume-fraction decrease below melting—is imported from Auer and Frenkel (Ref. [6]), not derived from the authors' own simulations or fitted to their own target data. Table 1 is explicitly described as 'calculated based on a linear fit to the data from Figure 11 in Ref. [6]', i.e., an external benchmark. The 20%-supersaturation mapping uses a stated formula with phi_F = 0.4918, phi_HCP = 0.74, and phi_M = 0.543, together with rates from ten Wolde et al. (Ref. [4]) and Kelton (Ref. [5]); these are independent sources rather than self-citations. The authors' prior work (Refs. [1] and [3]) is cited to show that spontaneous nucleation was observed near phi = 0.535 and that Smallenburg reproduced it, but those citations are corroborative: removing them would not alter the exponential-growth curve taken from Ref. [6]. No equation in the reply is defined in terms of the claimed result, and no fitted parameter is relabeled as a prediction. The legitimate weakness is extrapolation: a constant 6-fold/0.1% slope fitted near phi = 0.5342 is extended down to phi = 0.495, an 11-order-of-magnitude range, and the supersaturation formula is asserted rather than derived. These are accuracy and robustness concerns about a borrowed external trend, not circularity. Accordingly, the paper is self-contained against external benchmarks; the only minor self-citation appears as supporting context, not as the load-bearing derivation.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central quantitative table depends on two externally fitted numbers from Auer and Frenkel (the 6-fold growth factor and the baseline rate). Three domain assumptions are asserted: the atomic-to-colloidal supersaturation mapping, the validity of the linear extrapolation over a huge range, and the direct applicability of bulk rates to a finite simulation box.

free parameters (2)

Exponential growth factor per 0.1% volume fraction = 6-fold (dimensionless)
Derived from a linear fit to Figure 11 of Ref. [6] (Auer and Frenkel 2004); used as the basis for all entries in Table 1.
Baseline reduced nucleation rate at φ=0.5342 = I* = 10^-9 (dimensionless)
Reported in Ref. [6] and used to anchor the absolute nucleation times in Table 1 via I = I* σ^5 / D0.

assumptions (3)

domain assumption Percentage supersaturation for hard spheres is (φ - φ_F)/(φ_HCP - φ_F), with φ_F=0.4918, φ_HCP=0.74, φ_M=0.543.
Used to map atomic undercooling to colloidal volume fraction and to identify φ≈0.53 as the departure point; no derivation or validation is given.
domain assumption Auer and Frenkel's measured 6-fold per 0.1% growth rate remains valid over the full range φ=0.540 to 0.495.
Table 1 relies on this linear extrapolation over 11 orders of magnitude in waiting time.
domain assumption Laboratory waiting times computed from Auer and Frenkel's rates apply directly to a 2,048,000-particle colloidal simulation box of volume 1.6×10^-8 cm^3.
The conversion assumes the simulation box samples nuclei at the bulk rate and that Stokes-Einstein diffusion with D0 corresponds to the stated particle size and temperature.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Reply to Smallenburg: Near-melting nucleation and the exponential growth of hard-sphere nucleation times." pith.science (2026). https://pith.science/paper/EYN3VLL3

@misc{pith2026260807644,
  author       = {Pith},
  title        = {Pith review of: Reply to Smallenburg: Near-melting nucleation and the exponential growth of hard-sphere nucleation times},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EYN3VLL3}},
  note         = {Machine review of arXiv:2608.07644}
}

read the original abstract

Smallenburg reports near-melting point simulations and observes a well-predicted spontaneous nucleation with mixed, finite-time morphology, in a Comment on our recent Perspective. The Comment's incorrect broader takeaway --- that equilibrium coexistence is "readily achievable" --- rests on an untested generalization from a single state point near the phase envelope, and misses entirely the intriguing role played by Frenkel's underlying mechanism. We reiterate the salient point missed by the Comment: the nucleation time grows astronomically with just tenths of a percent of volume fraction away from 53%. This phenomenology emerges from the entropy exchange mechanism Frenkel described, which predicts that spontaneous phase separation is dynamically accessible only down to about 53% volume fraction from the melting point, and astronomically long waiting times through most of the remaining phase envelope. We provide here calculations to address the potential misconception created by the Comment.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages

[6]

Numerical prediction of absolute crystallization rates in hard-sphere colloids.The Journal of Chemical Physics, 120(6):3015–3029, 2004

Stefan Auer and Daan Frenkel. Numerical prediction of absolute crystallization rates in hard-sphere colloids.The Journal of Chemical Physics, 120(6):3015–3029, 2004

work page 2004
[1]

The elusive fluid-and-crystal coexis- tence state in simulations of monodisperse, hard-sphere colloids.AIChE Journal, 72(6):e70275, 2026

J Galen Wang, Umesh Dhumal, Monica EA Zakhari, and Roseanna N Zia. The elusive fluid-and-crystal coexis- tence state in simulations of monodisperse, hard-sphere colloids.AIChE Journal, 72(6):e70275, 2026

work page 2026
[2]

Order through disorder: entropy strikes back.Phys

Daan Frenkel. Order through disorder: entropy strikes back.Phys. World, 6(2):24–25, feb 1993

work page 1993
[3]

J Galen Wang, Umesh Dhumal, Monica EA Zakhari, and Roseanna N Zia. Frenkel’s entropy-exchange mechanism in monodisperse, nearly hard-sphere colloids: minimal perturbations to access fluid–crystal coexistence.Journal of Fluid Mechanics, 1031:A13, 2026

work page 2026
[4]

Numerical calculation of the rate of crystal nu- cleation in a lennard-jones system at moderate undercool- ing.The Journal of Chemical Physics, 104(24):9932– 9947, 1996

Pieter R ten Wolde, Maria J Ruiz-Montero, and Daan Frenkel. Numerical calculation of the rate of crystal nu- cleation in a lennard-jones system at moderate undercool- ing.The Journal of Chemical Physics, 104(24):9932– 9947, 1996

work page 1996
[5]

Crystal nucleation in liquids and glasses

Kenneth F Kelton. Crystal nucleation in liquids and glasses. InSolid State Physics, volume 45, pages 75–

work page
[7]

Phase transition for a hard sphere system.The Journal of Chemical Physics, 27(5):1208, 1957

Berni Julian Alder, Thomas Everett Wainwright, et al. Phase transition for a hard sphere system.The Journal of Chemical Physics, 27(5):1208, 1957

work page 1957
[8]

Studies in molecular dynamics

Berni J Alder and Thomas Everett Wainwright. Studies in molecular dynamics. i. general method.The Journal of Chemical Physics, 31(2):459–466, 1959

work page 1959

Show all 9 references

[9]

Studies in molecular dynamics

Berni Julian Alder and Thomas Everett Wainwright. Studies in molecular dynamics. ii. behavior of a small number of elastic spheres.The Journal of Chemical Physics, 33(5):1439–1451, 1960. 3

1960

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [6]

Numerical prediction of absolute crystallization rates in hard-sphere colloids.The Journal of Chemical Physics, 120(6):3015–3029, 2004

Stefan Auer and Daan Frenkel. Numerical prediction of absolute crystallization rates in hard-sphere colloids.The Journal of Chemical Physics, 120(6):3015–3029, 2004

work page 2004

[2] [1]

The elusive fluid-and-crystal coexis- tence state in simulations of monodisperse, hard-sphere colloids.AIChE Journal, 72(6):e70275, 2026

J Galen Wang, Umesh Dhumal, Monica EA Zakhari, and Roseanna N Zia. The elusive fluid-and-crystal coexis- tence state in simulations of monodisperse, hard-sphere colloids.AIChE Journal, 72(6):e70275, 2026

work page 2026

[3] [2]

Order through disorder: entropy strikes back.Phys

Daan Frenkel. Order through disorder: entropy strikes back.Phys. World, 6(2):24–25, feb 1993

work page 1993

[4] [3]

J Galen Wang, Umesh Dhumal, Monica EA Zakhari, and Roseanna N Zia. Frenkel’s entropy-exchange mechanism in monodisperse, nearly hard-sphere colloids: minimal perturbations to access fluid–crystal coexistence.Journal of Fluid Mechanics, 1031:A13, 2026

work page 2026

[5] [4]

Numerical calculation of the rate of crystal nu- cleation in a lennard-jones system at moderate undercool- ing.The Journal of Chemical Physics, 104(24):9932– 9947, 1996

Pieter R ten Wolde, Maria J Ruiz-Montero, and Daan Frenkel. Numerical calculation of the rate of crystal nu- cleation in a lennard-jones system at moderate undercool- ing.The Journal of Chemical Physics, 104(24):9932– 9947, 1996

work page 1996

[6] [5]

Crystal nucleation in liquids and glasses

Kenneth F Kelton. Crystal nucleation in liquids and glasses. InSolid State Physics, volume 45, pages 75–

work page

[7] [7]

Phase transition for a hard sphere system.The Journal of Chemical Physics, 27(5):1208, 1957

Berni Julian Alder, Thomas Everett Wainwright, et al. Phase transition for a hard sphere system.The Journal of Chemical Physics, 27(5):1208, 1957

work page 1957

[8] [8]

Studies in molecular dynamics

Berni J Alder and Thomas Everett Wainwright. Studies in molecular dynamics. i. general method.The Journal of Chemical Physics, 31(2):459–466, 1959

work page 1959