{"id":"69a8a41f-66e8-413e-b83b-b25a3da862da","arxiv_id":"2504.15238","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Swap Monte Carlo is extended to a size-polydisperse diatomic molecular model and equilibrates it down to near the experimental glass transition with an estimated 1000 to 1,000,000 times speedup.","lead":"A new simulation recipe lets supercooled diatomic molecules be cooled much faster in the computer than before. It gives glass researchers a molecular model with realistic rotations that can reach temperatures close to the glass transition.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline speedup range rests on unreported VFT/parabolic extrapolations of τMD from T≈0.35–0.37 to T=0.29; without fit parameters, fitting window, and uncertainties, the 10^3–10^6 claim is not independently verifiable.","rationale":"The reader's weakest assumption and my independent read converge: the speedup range is the least secure part of the central claim. I considered alternatives—actual wall-clock speedup including swap overhead, possible growth of size-resolved dynamic asymmetry below the lowest directly simulated temperatures, and whether the swap dynamics decorrelates configurations sufficiently—but those are either secondary or supported by the paper's checks. The absence of fit parameters and raw data is a concrete, fixable gap. Since the reader already conditions acceptance on providing these details, my recommendation is to keep the CONDITIONAL verdict. The concern does not force rejection: the model, swap implementation, and equilibrium validation are plausible and internally consistent.","tokens_in":10830,"tokens_out":6445,"duration_ms":60480,"concrete_test":"Recompute the extrapolation from the Fig. 2 data: fit τMD(T) for δ=5% and δ=10% with both the VFT and parabolic forms over the same temperature range used by the authors, and report parameters, residuals, and 95% prediction intervals at T=0.29. Then recompute τMD(0.29)/τswap(0.29) and the implied Tg values. If the two fitted forms disagree at T=0.29 by more than the stated decade-wide uncertainty, or if the prediction intervals span more than two decades, the abstract's speedup range should be revised or explicitly labeled as fit-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—an estimated speedup of 10^3–10^6 at T=0.29—is not measured but extrapolated. In Section II the authors state that τMD(T=0.29) exceeds the resolvable MD time scale and that they therefore apply both the Vogel-Fulcher-Tamann equation and the parabolic law, leading to a range of values. However, the paper does not report the fitted parameters, the fitting window, the number of points, residuals, or the uncertainty of the extrapolation. The extrapolation bridges from directly measured MD relaxation times near T=0.35–0.37 (Fig. 4) to roughly ten decades at the estimated Tg; two different three-parameter empirical forms can easily bracket a 10^3–10^6 range because neither is constrained at low T. The same extrapolation underlies the Tg estimates via the τMD(Tg)/τ0=10^12 criterion, so errors propagate into both headline numbers. The equilibrium checks in Fig. 3 and the direct observation that size-resolved relaxation-time ratios saturate below 1.2/1.5 are credible; the weak point is specifically the unquantified extrapolation. Requiring the fit details is a reproducibility issue, not a challenge to the swap algorithm itself.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a size-polydisperse asymmetric-dumbbell (ASD) molecular model, based on earlier toluene-model parameters, and shows that a Swap Monte Carlo scheme that alternates short MD trajectories with swaps of molecular size parameters can equilibrate this model at temperatures down to T=0.29 in reduced units, near the estimated experimental glass-transition temperature. The central claims are (i) an estimated speedup of 10^3–10^6 relative to standard MD at the lowest temperature, depending on polydispersity, and (ii) that size-resolved orientational relaxation times differ by only small factors (≤1.2 for 5% polydispersity and ≤1.5 for 10%), in contrast to strongly polydisperse point-particle swap models. The authors support the equilibrium nature of the Swap-generated configurations by showing absence of aging in P2(t) and continuity of the potential energy as a function of temperature. They also report consistency between rotational and translational relaxation and provide additional correlation functions and swap-acceptance rates in the appendices.","tokens_in":11150,"tokens_out":3081,"duration_ms":30878,"significance":"If the central claims hold, this is a useful extension of Swap Monte Carlo from point particles to molecular degrees of freedom, enabling future numerical studies of rotational dynamics in deeply supercooled molecular liquids. The paper has clear strengths: the equilibrium checks in Fig. 3 (absence of aging and no kink in the energy curve) are appropriate and directly support the validity of the Swap-equilibrated configurations; the size-resolved P2(t) data in Fig. 4 provide direct evidence that low polydispersity does not introduce large dynamic heterogeneity; and the appendices extend the main conclusions to translational and first-order orientational correlators. The main weakness is that the headline speedup and Tg estimates rely on an extrapolation of τMD(T) whose fitting details are not reported, making the quantitative claim not independently reproducible as written.","major_comments":[{"comment":"The central quantitative claims—the speedup of 10^3–10^6 at T=0.29 and the Tg estimates—depend on extrapolating τMD(T) from the lowest measured temperatures (near T=0.35–0.37 for the two polydispersities) down to T=0.29 using the VFT equation and the parabolic law. The manuscript does not report the fitted parameters, the fitting window, the number of data points, residuals, or any uncertainty estimate for either extrapolation. Since both functional forms are three-parameter empirical expressions and neither is constrained in the extrapolated regime, the stated ranges are not independently verifiable. Please add a table or paragraph with the full fit details, and also assess the sensitivity of the speedup and Tg estimates to the fitting window or to the inclusion/exclusion of individual data points.","section":"Section II, paragraph beginning \"To estimate the maximum speedup...\""},{"comment":"The Tg estimates, obtained via the criterion τMD(Tg)/τ0 = 10^12, inherit the same extrapolation uncertainty as the speedup estimates. The reported ranges Tg(δ=10%) ≈ 0.265–0.285 and Tg(δ=5%) ≈ 0.250–0.270 are presented without error bars or a discussion of how the spread between VFT and parabolic extrapolations translates into an uncertainty interval. This should be stated explicitly, and the same fit documentation requested above should be used to quantify the uncertainty in Tg.","section":"Section II, same paragraph"},{"comment":"The phrase \"maximum speedup achieved in this study\" in the abstract and Section II is stronger than what is actually measured: the speedup at T=0.29 is not directly measured but is obtained by combining measured τSwap(T=0.29) with extrapolated τMD(T=0.29). The wording \"estimated\" is used, which is appropriate, but the separation between directly measured speedups (at the lowest temperature where both τSwap and τMD are available) and extrapolated speedups should be made explicit in the text, ideally with the direct speedup values reported as well.","section":"Section II, Fig. 2 and its discussion"}],"minor_comments":[{"comment":"The definition of polydispersity δ in Eq. (2) is stated for the A-particle size distribution, but the caption of Fig. 1(a) merely says \"particle-size distribution\"; please clarify in the caption that the shown distribution is of the molecular (A-particle) sizes, not of all interaction sites.","section":"Section II, Eq. (2) and Fig. 1(a)"},{"comment":"The text states that \"The MD units used refer to the size and energy of the A particles of the monodisperse model.\" Please specify the complete set of MD units (length σ_A, energy ε_A, mass m_A, and the temperature unit ε_A/k_B) so that the reduced units are unambiguous.","section":"Section II, simulation details"},{"comment":"The sentence \"resulting in acceptance rates of 15-30% (SI)\" appears to conflict with Fig. 8, which shows acceptance rates decreasing with temperature and falling below 15% at the lowest temperatures. Please check the reported range or clarify whether the 15–30% range refers only to the higher-temperature portion of the data.","section":"Section II, simulation details"},{"comment":"The y-axis label in Fig. 8 reads \"acceptance rate swaps\"; for readability, write \"swap acceptance rate\" and add a unit or clarify it is dimensionless, as is conventional for Metropolis acceptance probabilities.","section":"Appendix B, Fig. 8"},{"comment":"Reference [7] is given as \"G. L. Hunter and E. R. Weeks\" but should include the co-authors of \"The physics of the colloidal glass transition\" (the article has an additional author, M. D. Ediger or similar, depending on the journal version). Please verify the complete author list and also check reference [58] for the same issue.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is well written and the core algorithmic contribution is plausible and supported by appropriate equilibrium checks and size-resolved dynamic data. However, the headline numbers (speedup and Tg) are not reproducible without the details of the VFT and parabolic extrapolations. I would recommend that the editor require the authors to report fit parameters, fitting windows, and uncertainty estimates before acceptance. The paper is within the scope of the journal, and the requested additions are local rather than requiring new physics."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: this is a useful paper. It extends Swap MC from point particles to asymmetric dumbbells and shows the trick still works at 5–10% polydispersity, where size-resolved dynamics stay close (ratios ≤1.2/1.5). That last point is the real result; it answers the Pihlajamaa concern about 23% polydispersity systems and makes the model usable for molecular-glass questions. The equilibrium checks are the right ones: no aging in first/second half P2(t), continuous energy curve matching MD, and the model/code/temperature set are built on the earlier toluene-like ASD model. I believe the swap acceleration is genuine; Fig. 2 and the acceptance rates support it.\n\nThe soft spot is exactly where the stress-test note points. The maximum speedup 10^3–10^6 and the Tg estimates both rest on VFT and parabolic-law extrapolations of tau_MD from T ≈ 0.35–0.37 down to 0.29. The paper says it applied both forms and got a range, but does not give the fit parameters, fitting window, number of points, or residuals. Two three-parameter forms bracket 10^3–10^6 precisely because neither is constrained at low T. This is a reproducibility and presentation issue, not a hidden flaw in the algorithm. The fix is easy: report the fits, show the extrapolation curves, and label the speedup range as an estimate. Also, no data or scripts are deposited; given that the model is meant as a community tool, input files or a table of tau values would help. Both are minor in the sense that the core claim—Swap equilibrates a diatomic molecular model at low polydispersity—is directly demonstrated.\n\nWho it's for: computational glass physicists who want a molecular analog of the Ninarello–Berthier–Coslovich model, and experimental groups comparing dielectric or rotational data. It deserves a serious referee. I would want the fit details sorted out before publication, but that is revision, not rejection.","headline":"First credible Swap MC extension to non-spherical molecules, with a real but contained weakness: the headline speedup is extrapolated from fits the paper does not report.","tokens_in":11643,"tokens_out":1626,"would_cite":true,"duration_ms":15703,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper shows a simple polydisperse dumbbell liquid can be equilibrated with Swap Monte Carlo down to temperatures near the experimental glass transition, with estimated speedups of 10^3–10^6 over standard molecular dynamics.","keywords":["swap Monte Carlo","molecular glass former","asymmetric dumbbells","polydispersity","supercooled liquids","orientational relaxation","glass transition","molecular dynamics simulation"],"falsifier":"Run standard MD on the δ=10% system at T=0.31, measure the orientational relaxation time directly, and compare it with the VFT and parabolic extrapolations anchored at higher temperatures; a measured value well below the extrapolated band would invalidate the estimated $10^{4}$–$10^{6}$ speedup at T=0.29, while agreement would support it.","tokens_in":10634,"feed_emoji":"🧊","tokens_out":6467,"duration_ms":53315,"temperature":0.7,"pith_summary":"This paper introduces a minimal molecular model—size-polydisperse asymmetric dumbbells—that can be equilibrated with Swap Monte Carlo at temperatures low enough to approach the experimental glass transition, something previously possible only for spherical point particles. The authors report estimated speedups of $10^{3}$–$10^{6}$ over standard molecular dynamics at 5–10% polydispersity, and show that the orientational relaxation of the largest and smallest molecules differs by at most factors of 1.2 and 1.5, respectively. A sympathetic reader would care because this gives a route to simulating the rotational degrees of freedom of deeply supercooled molecular liquids directly, rather than inferring them from atomic models. The paper also confirms the swapped configurations are true equilibrium states by showing the absence of aging and continuity of potential energy with temperature.","feed_headline":"Swap Monte Carlo reaches glass-transition depths for diatomic liquids","feed_subtitle":"A 5–10% polydisperse dumbbell liquid relaxes 1,000 to 1,000,000 times faster with swaps.","key_machinery":"The load-bearing object is the size-polydisperse asymmetric dumbbell (ASD) model with molecule-level swaps: each molecule is a harmonic-bonded pair of a large A particle and a small B particle, and polydispersity enters by scaling both sizes of a molecule by the same factor drawn uniformly from [1-Δ,1+Δ] with 500 discrete types. The swap algorithm alternates short NVT molecular-dynamics runs (t=0.32) with 2N attempts to exchange the size parameters of randomly chosen molecule pairs, accepted by Metropolis rules. The argument is carried by comparing orientational relaxation times from P2(t) during swap and standard MD, and by extrapolating the MD relaxation times using the VFT and parabolic formulas to estimate the speedup at T=0.29 and the glass-transition temperature via the criterion τMD(Tg)/τ0 = $10^{12}$. The machinery also includes checks for equilibrium: absence of aging in P2(t) halves and continuity of potential energy across temperatures.","core_discovery":"The central claim is that Swap Monte Carlo works for a molecular liquid when the swapped property is the molecular size: each molecule is an asymmetric dumbbell of large and small particles (modeling toluene-like chemistry), and the two particle sizes of a given molecule are scaled together so that the whole molecule swaps size with another molecule. With only 5–10% size polydispersity, the algorithm accelerates both rotational and translational relaxation by the same large factor, reaching T=0.29 in reduced units, near the estimated Tg of 0.25–0.285. The size-resolved orientational autocorrelation functions P2(t) show that the smallest and largest molecules relax nearly alike—a ratio of relaxation times at most 1.2 for 5% and 1.5 for 10% polydispersity—in contrast to the factor-of-50 differences reported for highly polydisperse point-particle systems. The authors therefore propose the model as a minimal, realistic system for studying deeply supercooled molecular liquids in silico.","pith_inferences":["The reported speedup range depends on extrapolating τMD from higher temperatures with VFT and parabolic fits; a direct measurement of τMD at T≈0.31–0.35, even approximate, would significantly narrow the 10^3–10^6 uncertainty.","If the near-equality of size-resolved relaxation persists at the estimated Tg, the model becomes a test bed for whether rotational and translational decoupling observed experimentally in molecular liquids arises from intrinsic molecular asymmetry rather than from polydispersity.","The successful removal of polydispersity at a shifted temperature (as in Fig. 5) hints that one could equilibrate with swaps and then 'de-polydisperse' to study a monodisperse molecular liquid at a renormalized temperature, effectively using swap MC as a preparation step rather than as the production ensemble."],"forward_implications":["Equilibrium configurations of a molecular liquid can now be generated at temperatures where τMD/τ0 ≈ 10^6–10^12, enabling direct simulation of the ultraviscous regime with rotations.","The model can serve as a baseline for comparing simulated dielectric relaxation spectra with experiments on molecular glass formers, e.g., checking for the predicted 1/√ω high-frequency decay of the α process.","The small size-resolved dynamical differences indicate that polydispersity artifacts seen in point-particle swap models do not contaminate this molecular model at 5–10% polydispersity.","Because swaps work at only 5% polydispersity, the approach may be extended to more complex molecules such as trimers, although efficiency is expected to drop for larger molecules."],"supporting_citations":[{"why":"Introduced the continuous size-polydisperse swap model that this paper adapts to molecules, and the equilibration protocol's theoretical basis.","marker":"[19]"},{"why":"Established efficient swap algorithms for molecular-dynamics-based equilibration, including acceptance-rate behavior used here.","marker":"[20]"},{"why":"Provided the deterministic molecule-size sampling procedure that avoids finite-size effects in the polydisperse distribution.","marker":"[22]"},{"why":"Documented factor-~50 size-resolved relaxation differences in high-polydispersity point-particle systems, the benchmark the 1.2–1.5 ratios are contrasted against.","marker":"[50]"},{"why":"Supplies the monodisperse asymmetric-dumbbell interaction parameters and the toluene-like minimal-model rationale adopted for the A/B particles.","marker":"[53–55]"},{"why":"VFT equation used to extrapolate standard-MD relaxation times to T=0.29 for the speedup estimate.","marker":"[59]"},{"why":"Parabolic law used as the second extrapolation model, defining the lower/upper range of the speedup estimate.","marker":"[60]"},{"why":"The GPU-accelerated simulation package in which all MD and swap simulations were run.","marker":"[57]"}],"fun_headline_variants":["Swap Monte Carlo extends to diatomic molecules via size swaps","Dumbbell size swaps: 1,000–1,000,000× faster equilibration","Molecular Swap MC: supercooled diatomic liquids relax uniformly","Beyond point particles: Swap Monte Carlo for dumbbell liquids","Size-polydisperse dumbbells: Swap MC reaches glassy depths"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative speedup and glass-transition estimates rest on assuming that the standard formulas used to extrapolate the molecular-dynamics relaxation time from warmer temperatures down to T=0.29 remain accurate in that unmeasured range.","fun_headline_variants_meta":{"raw":{"variants":["Swap Monte Carlo extends to diatomic molecules via size swaps","Dumbbell size swaps: 1,000–1,000,000× faster equilibration","Molecular Swap MC: supercooled diatomic liquids relax uniformly","Beyond point particles: Swap Monte Carlo for dumbbell liquids","Size-polydisperse dumbbells: Swap MC reaches glassy depths"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000832,"raw_usage":{"total_tokens":3588,"prompt_tokens":854,"completion_tokens":2734,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":470,"completion_tokens_details":{"reasoning_tokens":2637}},"tokens_in":470,"tokens_out":2734,"duration_ms":17963,"temperature":1.0,"reasoning_tokens":2637,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:29:45.805909+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run standard MD on the δ=10% system at T=0.31, measure the orientational relaxation time directly, and compare it with the VFT and parabolic extrapolations anchored at higher temperatures; a measured value well below the extrapolated band would invalidate the estimated $10^{4}$–$10^{6}$ speedup at T=0.29, while agreement would support it.","supporting_citations":[{"cited_title":"Models and algorithms for the next generation of glass transition studies,","cited_arxiv_id":null,"evidence_quote":"Introduced the continuous size-polydisperse swap model that this paper adapts to molecules, and the equilibration protocol's theoretical basis."},{"cited_title":"Efficient swap algorithms for molecular dynamics simulations of equilibrium supercooled liquids,","cited_arxiv_id":null,"evidence_quote":"Established efficient swap algorithms for molecular-dynamics-based equilibration, including acceptance-rate behavior used here."},{"cited_title":"Understanding the swap Monte Carlo algorithm in a size-polydisperse model glass- former,","cited_arxiv_id":null,"evidence_quote":"Provided the deterministic molecule-size sampling procedure that avoids finite-size effects in the polydisperse distribution."},{"cited_title":"Influence of polydispersity on the relaxation mechanisms of glassy liquids,","cited_arxiv_id":null,"evidence_quote":"Documented factor-~50 size-resolved relaxation differences in high-polydispersity point-particle systems, the benchmark the 1.2–1.5 ratios are contrasted against."},{"cited_title":"Das temperaturabhangigkeitsgesetz der viskositat von flussigkeiten,","cited_arxiv_id":null,"evidence_quote":"VFT equation used to extrapolate standard-MD relaxation times to T=0.29 for the speedup estimate."},{"cited_title":"Corresponding states of structural glass formers,","cited_arxiv_id":null,"evidence_quote":"Parabolic law used as the second extrapolation model, defining the lower/upper range of the speedup estimate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The GPU-accelerated simulation package in which all MD and swap simulations were run."}],"review_version":1}