{"id":"27e73c1e-1073-476a-9610-d4c8ebf640c1","arxiv_id":"2411.17540","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Annealing ultrastable vapor-deposited MMT glasses at 0.98Tg increases the storage dielectric susceptibility without changing the loss, interpreted as boson-peak recovery, while swap Monte Carlo simulations show no pre-transformation softening.","lead":"Annealing ultrastable vapor-deposited glass films just below their glass transition temperature for six hours causes the storage part of the dielectric response to rise without a matching rise in the loss part. The authors interpret this as an early recovery of the boson peak, a high-frequency vibration signature, before the glass fully transforms, and they note that supercomputer simulations of stable glasses do not show this effect.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The boson-peak recovery is not independently measured: the 2% χ′ rise is reconstructed using glycerol's spectral shape and indomethacin's suppression, with the 7% recovery amplitude chosen to match the data, so the experiment/simulation contrast rests on an unverified high-frequency assignment.","rationale":"Reader's weakest assumption matches my own: the central speculative identification of the boson peak is not independently measured and the quantitative reconstruction is parameterized to fit the data. The paper contains multiple explicit admission tokens: 'speculatively identified' (§IV.A), and the statement that any process >10^5 Hz would be a candidate. These are flagged in full text and weigh against accepting the strong conclusion. The simulation comparison is clean but cannot rescue the experimental interpretation; the MD MSD curves probe translational vibrational dynamics, while the experimental signal could in principle arise from intramolecular or other high-frequency modes. The absence of error bars and the lack of a live drift reference for the isothermal 2% signal makes direct confirmation necessary. However, the observation itself is interesting and the authors are transparent about the speculative nature; a conditional verdict is appropriate pending the broadband check. I therefore keep the reader's CONDITIONAL verdict and agree with the identified weakest assumption.","tokens_in":20989,"tokens_out":6054,"duration_ms":60037,"concrete_test":"Measure the complex dielectric susceptibility of a fresh 0.80Tg MMT film before and after the same 6 h, 0.98Tg anneal over a broadband window extending at least from 10^5 Hz into the GHz-THz range (e.g., by waveguide/THz-TDS or by repeating the anneal on a substrate compatible with such measurements). If the K-K-predicted high-frequency loss increase beneath the reconstructed boson peak is absent—or if its amplitude is inconsistent with the fitted 7% recovery—the boson-peak interpretation in Fig. 6 is falsified. In the same run, a covered-electrode reference should be recorded continuously during annealing to rule out a 2% substrate drift artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central conclusion—that stable MMT glasses soften (boson peak partially recovers) before the α relaxation—depends on the Kramers-Kronig inference in §IV.A that an increase in χ′ at low frequencies with no change in χ″ is produced by a process above the 10^5 Hz measurement window. That inference is secure only if the effect is real and if the high-frequency process is the boson peak. Neither is directly established. The boson peak is labeled 'speculatively identified' in §IV.A. The quantitative reconstruction imports the dielectric boson-peak line shape from glycerol and the 34% stable-glass suppression from indomethacin, then chooses a 7% recovery to reproduce the measured 2% χ′ change; this is a fit, not a prediction. Any other process above 10^5 Hz would produce the same K-K signature, and the paper explicitly concedes this. Additionally, the constant-background subtraction (§II.A) could in principle mistake a slow substrate/drift artifact for film response, since no live reference channel was used during the six-hour isothermal anneals. If either the drift-control or the direct high-frequency check fails, the observed phenomenon loses its link to boson-peak recovery and the claimed contrast with simulation becomes uninterpretable.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript describes isothermal annealing experiments on vapor-deposited glasses of methyl-m-toluate (MMT) at 0.98Tg for up to six hours, monitored by dielectric spectroscopy. Films deposited at 0.95Tg fully rejuvenate, while 0.90Tg films partially rejuvenate, as judged by changes in the dielectric susceptibility and in the onset temperature of the alpha relaxation measured on subsequent heating. For the most stable films (deposited at 0.80Tg and 0.85Tg), the onset temperature is essentially unchanged, but the storage component χ' increases by roughly 2% at 20 Hz (about 20% of the way to the equilibrium value) with no measurable change in the loss component χ''. The authors interpret this as a partial recovery of the boson peak, conclude that the boson peak recovers faster than the stable-glass-to-liquid transformation, and compare with swap Monte Carlo simulations of ultrastable two-dimensional glasses, which show no evolution of the vibrational dynamics of the glass particles before transformation. The paper also reports KWW fits to the onset-temperature evolution and a Tool-Narayanaswamy-Moynihan analysis to estimate the extremely long relaxation times of the stable glasses at their deposition temperatures.","tokens_in":21320,"tokens_out":14804,"duration_ms":118138,"significance":"If the experimental observation is robust and the assignment to the boson peak is correct, the result is significant: it would demonstrate that the vibrational properties of an ultrastable glass evolve on timescales orders of magnitude shorter than the alpha-rejuvenation time, with direct implications for models of stable-glass softening and for the physical picture of rejuvenation. The simulation part is a strength: the MSD and van Hove analyses are clearly presented and constitute a clean, reproducible null result that is a valuable point of comparison. The authors are also commendably explicit about the speculative nature of the boson-peak identification. However, the significance of the central claim is currently limited because the key experimental effect is small and lacks reported uncertainty and drift controls, and because the boson-peak recovery is reconstructed by fitting rather than measured independently; as a result, the claimed contrast between experiment and simulation is not yet on the same footing as the simulation evidence.","major_comments":[{"comment":"The central interpretation rests on a speculative identification of the high-frequency process as the boson peak. The text states that the process is 'speculatively identified' and that the authors have an 'inability to directly observe the boson peak in this work.' The quantitative reconstruction in Figure 6 imports the boson-peak line shape from glycerol (Ref. 59) and the 34% suppression in stable PVD indomethacin (Ref. 49), and then chooses a 7% recovery of the boson-peak intensity so that the Kramers-Kronig calculation reproduces the measured 2% rise in χ'. This is a fit, not a prediction; the agreement of Figure 6 with the data is built in by construction. The paper itself concedes that 'any process (not necessarily the boson peak) occurring at a frequency greater than 10^5 Hz would be a possible candidate.' Consequently, the abstract and conclusion claim that 'the boson peak recovers more quickly than the transformation of stable glass into supercooled liquid' is not established by the reported measurements.","section":"IV.A and Figure 6"},{"comment":"The constant-background subtraction used in the isothermal annealing experiments assumes that the substrate contribution to the dielectric response is time-independent over the full six-hour anneal. The background is taken as the average of a ten-minute baseline measured at the annealing temperature before film deposition, and this constant value is subtracted from all data collected during the anneal. No live reference channel is used during isothermal measurements; the two-channel subtraction is applied only during temperature ramps. A slow drift of the substrate or instrument response over 10^4-10^5 s could produce a monotonic change in χ' of a few percent with negligible change in χ'', comparable to the reported effect. A control experiment without a film, or simultaneous monitoring of the covered electrode pair during the anneal, is needed to exclude this artifact.","section":"II.A and Section III.A"},{"comment":"The key experimental result—a roughly 2% increase in the normalized storage susceptibility at 20 Hz for the 0.80Tg and 0.85Tg glasses with no change in loss—is presented without error bars, replicate runs, or a statistical test. Figure 2 and Supplemental Figure 8 show single traces for each deposition temperature, and the text does not quantify the run-to-run scatter. Given the small magnitude of the effect, the reader cannot assess whether the reported 'significant' increases exceed experimental uncertainty. The same lack of uncertainty reporting applies to the null changes in χ'' and in the onset temperature, which are central to the interpretation.","section":"Section III.A, Figure 2"},{"comment":"The simulation null result is not demonstrated to be sensitive to the small boson-peak recovery inferred in the experiment. The authors state that 'any modification of the glass would affect the peak near t ~ 0.5 or the height of the plateau' in the mean-squared displacement, but the MSD is an integral over the vibrational density of states, and a 7% recovery of the boson-peak intensity (the value used in the experimental reconstruction) would change the MSD plateau by an amount that may be far smaller than the scatter among the curves in Figure 8. Without a quantitative sensitivity estimate (for example, the expected change in the MSD plateau or the Debye-Waller factor for the experimental 7% boson-peak recovery, compared with the numerical noise), the simulation's 'no softening' conclusion does not directly contradict the experimental observation, and the paper's central 'contrast between experiment and simulation' is not established on the simulation side either.","section":"IV.B and Figure 8"}],"minor_comments":[{"comment":"Equation numbering is duplicated: the TNM equation in Section IV.C is labeled Eq. (1), but Eq. (1) is already used for the mean-squared displacement in Section II.B. The later equation should be renumbered.","section":"Throughout"},{"comment":"The spectral range '105 Hz to 100 Hz' should be written as 10^5 Hz to 10^2 Hz (or, if intended literally, '100 Hz to 10^5 Hz' with superscript notation); the current notation is ambiguous.","section":"II.A"},{"comment":"The estimated transformation time of the 0.80Tg glass at 0.98Tg is quoted as roughly 10^6 s in the main text, while Supplemental Figure 9 states 'almost 250 ks'; these differ by a factor of four and should be reconciled.","section":"III.B and Supplemental Figure 9"},{"comment":"The supplemental-figure cross-references are inconsistent: Section IV.A cites Supplemental Figure 8 for the ~10^6 s rejuvenation time, but this estimate is presented with Supplemental Figure 9 in Section III.B; Supplemental Figure 8 contains the second-frequency annealing data.","section":"IV.A and III.B"},{"comment":"The experimental temperature upjump is described as 'roughly 10%', but the ratio 0.98Tg/0.80Tg corresponds to an increase of about 22%; the basis for the 10% figure should be stated.","section":"IV.B"},{"comment":"The caption for the supplemental van Hove function is labeled 'Figure 1' instead of 'Supplemental Figure 10'.","section":"SI, van Hove figure"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is from an experienced group and the simulation analysis is of high quality. My main reservation is that the central experimental claim is built on a speculative identification and a reconstruction that is a fit, and the small effect is presented without the error analysis and drift controls that the magnitude demands. The authors are candid about the speculation, but the abstract and conclusion state the boson-peak conclusion more strongly than the evidence supports. A revision that adds replicate statistics and a drift control, and that either directly tests the high-frequency process or clearly reframes the conclusion as an observation of a high-frequency process with the boson-peak identification as one hypothesis, would bring the paper's claims in line with its evidence. The journal's readership would find the experimental result interesting even without the strong boson-peak claim, and the simulation null result would remain a useful counterpoint. I have no concerns about citation practices or overlap."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The paper has one genuinely new observation: in the most stable vapor-deposited MMT glasses, isothermal annealing at 0.98Tg for six hours increases the storage component of the dielectric susceptibility by about 20% of the way to equilibrium, with no measurable change in the loss component and no change in the alpha-onset temperature. I think that observation is real. The simulation part is also clean: the MSD and van Hove analyses show that glass particles in a swap-MC ultrastable glass do not soften before transformation begins. That contrast is the paper's real contribution.\n\nThe soft spot is the interpretation. The Kramers-Kronig argument that a low-frequency rise in chi' with no change in chi'' implies a change in some process above the measurement window is fine, provided the effect is real. But the attribution to the boson peak is explicitly speculative, and the quantitative reconstruction is a fit, not a prediction: the line shape comes from glycerol, the suppression from indomethacin, and the 7% recovery is chosen to reproduce the measured 2%. Any high-frequency process would give the same signature, as the authors concede. So the central claim—boson peak recovers faster than the alpha relaxation—is not directly supported.\n\nTwo experimental gaps matter. The 2% signal in Figure 2 has no error bars, and the constant-background subtraction has no drift control over the six-hour anneals. The covered electrode pair used for temperature ramping was not used as a live reference during isothermal runs, so a slow substrate drift is not ruled out.\n\nNone of this kills the paper. The observation is new, the simulations are well done, and the authors are honest about what is speculative. The experiment/simulation contrast is worth publishing even if the boson-peak label is provisional. But the paper needs revision: repeat measurements with error bars, a drift control, and a conclusion that names a generic high-frequency process rather than asserting boson-peak recovery.\n\nI would send it to peer review. A good referee can help tighten this. I'd probably bring it to reading group and would cite it with a caveat.","headline":"A real experimental observation with a provisional interpretation; the simulation contrast is clean but rests on an unverified boson-peak identification.","tokens_in":21838,"tokens_out":3761,"would_cite":true,"duration_ms":71285,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["64.70.P-","77.22.Gm"],"model":"deepseek-v4-flash","headline":"Annealing stable glasses raises their dielectric storage without touching loss, a sign the boson peak recovers before the glass transforms.","keywords":["vapor-deposited glasses","rejuvenation","boson peak","dielectric spectroscopy","swap Monte Carlo","kinetic stability","glass transition","methyl-m-toluate"],"falsifier":"Directly measure the boson peak, for example by inelastic neutron or Brillouin light scattering, on a 0.80Tg MMT film before and after six hours of annealing at 0.98Tg; if the boson-peak intensity does not increase by roughly 7% while chi' rises 2%, the proposed mechanism fails.","tokens_in":20827,"feed_emoji":"🔬","tokens_out":4445,"duration_ms":37983,"temperature":0.7,"pith_summary":"This paper asks what happens in the earliest stages of a stable glass turning back into a supercooled liquid. By annealing vapor-deposited glasses of methyl-m-toluate just below their glass transition for six hours and measuring the dielectric response, the authors find that the most stable glasses increase their storage susceptibility by about 2% with no change in the loss. They interpret the rise as a partial recovery of the boson peak, a high-frequency vibrational excess common to all glasses, while the alpha relaxation that signals full transformation remains untouched. In contrast, swap Monte Carlo simulations of ultrastable glasses show no change in the vibrational dynamics of the glass until the alpha relaxation begins. The paper thus claims that stable glass properties can evolve before the glass transforms, and that current simulations miss this early softening.","feed_headline":"Boson peak recovers before a stable glass transforms","feed_subtitle":"Annealing experiments find a 2% storage rise with no loss change, pointing to early softening that simulations do not capture.","key_machinery":"The central object is the complex dielectric susceptibility and the Kramers-Kronig relation that links a high-frequency absorption (the boson peak) to a flat contribution in the storage component chi' over a lower-frequency window, with negligible change in the loss component chi''. In the simulations, the mean-squared displacement of particles classified as glassy, decomposed from the liquid particles, serves as the probe of vibrational dynamics; the swap Monte Carlo algorithm generates the ultrastable configurations. The boson peak is defined as the excess in the vibrational density of states relative to the Debye crystal, appearing in dielectric spectra around 100 to 1000 GHz.","core_discovery":"For highly stable MMT glasses deposited at 0.85Tg and 0.80Tg, annealing at 0.98Tg for six hours (about 2% of the estimated transformation time) leaves the onset temperature of the alpha relaxation unchanged, but raises the storage component of the dielectric susceptibility by roughly 20% of the way to the equilibrium value, uniformly across the measured spectrum, with no corresponding rise in the loss component. The authors argue from Kramers-Kronig relations that this pattern implies a high-frequency process above the experimental window is recovering, and they speculatively identify it as the boson peak. Using the boson peak spectral shape from glycerol and a 34% suppression from indomethacin, they reconstruct the dielectric response and find that a 7% recovery of the boson peak intensity reproduces the measured 2% increase in chi' at 20 Hz. In swap Monte Carlo simulations of a two-dimensional polydisperse system, the mean-squared displacement of particles that remain in the glass state is identical for waiting times corresponding to liquid fractions up to 5%, indicating no softening of the glass prior to transformation. The paper concludes that the boson peak recovers more quickly than the transformation of stable glass into supercooled liquid, and that the current simulations do not capture this early evolution of stable glass properties.","pith_inferences":["If the boson peak interpretation holds, the early softening should be observable with direct high-frequency probes such as neutron or Brillouin light scattering on the same MMT films, before any shift in the onset temperature appears.","The discrepancy between experiment and simulation may stem from the much larger temperature up-jump used in simulations (about threefold versus ten percent); simulations with smaller up-jumps, if computationally feasible, might reveal a slow softening that currently falls outside the numerical time window.","The constant-background subtraction method assumes the substrate response does not drift over six hours; a control experiment with an uncoated electrode pair would confirm that the 2% rise is not an instrumental artifact."],"forward_implications":["The alpha-relaxation onset temperature is not a complete measure of a stable glass's state; other properties can evolve while the onset stays fixed.","The boson peak can serve as a sensitive early indicator of rejuvenation in vapor-deposited glasses, detectable through changes in dielectric storage.","Swap Monte Carlo simulations of polydisperse disk glasses do not reproduce this early softening, suggesting that current coarse-grained models miss physics relevant to molecular glasses.","The KWW and Tool-Narayanaswamy-Moynihan analysis provides a way to estimate extremely long structural relaxation times (up to 10^10 s) from short sub-Tg annealing experiments."],"supporting_citations":[{"why":"Supplies the glycerol dielectric spectrum used to model the boson peak contribution to the susceptibility.","marker":"[59]"},{"why":"Supplies the 34% boson-peak suppression value for stable vapor-deposited indomethacin glasses.","marker":"[49]"},{"why":"Provides the simulation method and results for two-step devitrification of ultrastable glasses that the present simulations extend.","marker":"[39]"},{"why":"Reports softening of capped stable glasses in nanocalorimetry, the qualitative precedent the paper distinguishes from.","marker":"[65]"},{"why":"Provides VFT parameters and the fictive-temperature partitioning x=0.32 used in the relaxation-time estimates.","marker":"[72]"},{"why":"Characterizes secondary relaxations in MMT stable glasses, establishing the baseline dielectric behavior of this material.","marker":"[34]"}],"fun_headline_variants":["Boson peak revives before stable glass transforms","Simulations miss early boson peak revival in stable glass","Dielectric rise signals early boson peak recovery","Stable glass softens before transformation, simulations don't show","Experiments show boson peak recovery precedes glass transformation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The observed rise in the storage component without a corresponding rise in loss is caused by a partial recovery of the boson peak, with the boson peak's intensity and spectral shape estimated from other glass-formers (indomethacin and glycerol) rather than measured in methyl-m-toluate.","fun_headline_variants_meta":{"raw":{"variants":["Boson peak revives before stable glass transforms","Simulations miss early boson peak revival in stable glass","Dielectric rise signals early boson peak recovery","Stable glass softens before transformation, simulations don't show","Experiments show boson peak recovery precedes glass transformation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000959,"raw_usage":{"total_tokens":4147,"prompt_tokens":1065,"completion_tokens":3082,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":681,"completion_tokens_details":{"reasoning_tokens":3005}},"tokens_in":681,"tokens_out":3082,"duration_ms":21845,"temperature":1.0,"reasoning_tokens":3005,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:59:37.738058+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly measure the boson peak, for example by inelastic neutron or Brillouin light scattering, on a 0.80Tg MMT film before and after six hours of annealing at 0.98Tg; if the boson-peak intensity does not increase by roughly 7% while chi' rises 2%, the proposed mechanism fails.","supporting_citations":[{"cited_title":"Perez-Castaneda, C","cited_arxiv_id":null,"evidence_quote":"Supplies the 34% boson-peak suppression value for stable vapor-deposited indomethacin glasses."},{"cited_title":"Herrero, C","cited_arxiv_id":null,"evidence_quote":"Provides the simulation method and results for two-step devitrification of ultrastable glasses that the present simulations extend."},{"cited_title":"Vila-Costa, J","cited_arxiv_id":null,"evidence_quote":"Reports softening of capped stable glasses in nanocalorimetry, the qualitative precedent the paper distinguishes from."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides VFT parameters and the fictive-temperature partitioning x=0.32 used in the relaxation-time estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Characterizes secondary relaxations in MMT stable glasses, establishing the baseline dielectric behavior of this material."}],"review_version":1}