{"id":"50594749-13dc-4b1b-aac5-83d64c593e66","arxiv_id":"1908.08152","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Short-time tracking of a single colloidal tracer reveals a logarithmic mean squared displacement phase between two stable glass plateaus, reported as the first 3D marginal glass.","lead":"By tracking a single colloidal particle inside a slowly densifying glass, the authors observed its short-time motion pass through liquid, stable glass, a predicted 'marginal' glass phase, and back to stable glass. This is presented as the first direct 3D thermal evidence for a subtle phase transition predicted by replica mean-field theory.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The marginal-phase assignment rests on the log-shaped MSD of one tracer while the sample is still densifying; without fixed-density controls and replicate tracers, the logarithmic segment cannot be uniquely attributed to the Gardner phase.","rationale":"The reader's weakest assumption is exactly the load-bearing point: the logarithmic MSD of one tracer is treated as an unambiguous Gardner signature. I agree. The paper offers some independent support—the tracking and drift-subtraction methods were used previously, and the authors cite simulations in which logarithmic MSDs appear in the marginal phase—but that support does not close the specific gap. The experiment never stops densifying, never varies temperature, never measures local density, and never gives an uncertainty on the shape classification. The 'return to stable glass' is inferred from the same single tracer later showing a plateau; without a local density or replicate tracers, that could be a deeply aged or locally jammed cage rather than a thermodynamic reentrant phase transition. The data-available-on-request policy is a limitation but not grounds for rejection. Because the core physics claim is plausible and the flaw is evidentiary rather than mathematical, I recommend no change from the reader's CONDITIONAL verdict: the paper should be accepted only if the missing replicates, error analysis, and a fixed-density control are supplied. I found no internal contradiction sufficient to warrant rejection, and disagreement with the mean-field prediction itself is not the issue here.","tokens_in":7756,"tokens_out":8561,"duration_ms":86376,"concrete_test":"Perform a density-matched, fixed-volume-fraction control: density-match the solvent so sedimentation stops, prepare fresh samples at volume fractions corresponding to the putative plateau, logarithmic, and reentrant regimes (estimating the local volume fraction near each tracer by confocal or tomographic imaging), and track at least 20 tracer particles per state point over the same 1.72 s window. If the ensemble-averaged MSD at fixed density does not show a statistically preferred logarithmic branch between the two plateaus (e.g., BIC preferring Eq. (2) over plateau and power-law models), then the Fig. 4 sequence is an artifact of ongoing sedimentation or aging rather than evidence for a marginal glass phase.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The decisive step is classifying the blue MSD curves in Fig. 4 as logarithmic via Eq. (2) and mapping them to the mean-field phase diagram in Fig. 1. Every subsequent claim about a reentrant marginal glass depends on that classification. Two gaps make it insecure. First, the system is not at fixed state points: Fig. 2 states that the density is still monotonically increasing even at the longest times, so every 20-minute MSD is recorded while the glass is aging and compacting. Logarithmic growth of the MSD is a generic signature of slow caging, compaction, or aging in a dense suspension, and the paper does not compare Eq. (2) against aging or compaction models. Second, the sequence of curves comes from a single tracer particle, with no error bars, no ensemble average over tracers, and no local volume fraction measurement near the tracer. The text refers to 'across all samples' and to supplemental runs, but the criterion for a 'good logarithmic fit' is not quantified, and no alternative functional forms (plateau, power law, crossover) are statistically compared. The timing descriptions are also inconsistent: the Fig. 4 caption says logarithmic from 320 to 400 min, while the text says after 340 min and later says after about 380 min, which further weakens the post-hoc phase assignment. Thus the strongest claim—direct experimental evidence of a marginal glass in 3D—is underdetermined by the data as presented. This is not an internal mathematical contradiction; it is a mismatch between the strength of the phase claim and the evidence supplied.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a single-particle tracking experiment in a sedimenting colloidal glass. The authors measure the short-time mean squared displacement (MSD) of one tracer colloid every 20 minutes and classify the MSD shape into four regimes as the sample densifies: subdiffusive liquid, stable glass (plateau), marginal glass (logarithmic growth), and reentrant stable glass (plateau). They interpret this sequence as direct experimental evidence for the marginal/Gardner glass phase predicted by replica mean-field theory and compare it qualitatively with the reentrant phase diagram for soft spheres.","tokens_in":8057,"tokens_out":5351,"duration_ms":52816,"significance":"If substantiated, the observation of a logarithmic MSD segment between two plateau regimes in a three-dimensional thermal colloidal glass would be a notable experimental confirmation of replica-theory predictions and would validate the short-time MSD as a probe of deep-glass phases. The experimental design has clear strengths: high-speed imaging at 64,000 fps, sub-pixel localization with a stated 1.5 nm localization error, explicit drift subtraction, and a defined free-colloid comparison for identifying the cage-onset lag time. The claim is timely and the technique is potentially powerful. However, as presented the phase assignment is underdetermined by the data: it rests on one tracer, on an unquantified goodness-of-fit criterion, and on measurements taken while the sample is still densifying. The central conclusion therefore needs substantially stronger quantitative and statistical support.","major_comments":[{"comment":"The assignment of the blue MSD curves to the marginal phase rests entirely on the curves being \"well fit\" by the logarithmic form in Eq. (2), but the classification is not quantitative: no goodness-of-fit measure, no comparison against plateau, power-law, stretched-exponential, or aging-based functional forms, and no uncertainty in the fitted parameters a and x* is given. The problem is compounded by the fact that the sample is never at a fixed state point: Fig. 2 states that the density increases monotonically even at the longest times, so every 20-minute MSD is recorded while the glass is aging and compacting. Logarithmic MSD growth is a generic signature of slow caging, compaction, or aging in dense suspensions. To make the Gardner-phase claim load-bearing, the authors need either fixed-density control measurements or a quantitative model comparison that distinguishes hierarchical sub-basin exploration from ordinary aging.","section":"§2, Fig. 4, Eq. (2)"},{"comment":"The entire phase sequence is inferred from a single tracer particle. The paper reports no error bars on the MSD curves, no ensemble average over multiple tracer particles, no replicate MSDs at the same elapsed time, and no local volume-fraction measurement in the tracer's neighborhood. The text refers to behavior \"across all samples\" and to supplemental runs, but the supporting data are not shown or quantified in the main text. Because every subsequent claim about the reentrant marginal glass depends on the shape classification of these particular curves, the analysis needs to be repeated over several tracers and reported with statistical uncertainty and an explicit, quantified criterion for what constitutes a \"good logarithmic fit.\"","section":"§2, Figs. 4 and 5"},{"comment":"The time intervals defining the marginal phase are internally inconsistent: the Fig. 4 caption states that the MSD is logarithmic from 320 to 400 minutes, whereas the text says the plateau-to-logarithm transition occurs \"after about 340 minute\" and later \"after about 380 minutes\"; the reported duration of the marginal phase is \"about 80 minutes\" in one sentence and \"an average of 165 minutes\" in the next. Since the phase boundaries are defined by these intervals, a table listing elapsed time, fitted parameters, and fit-quality measures for each run is needed to remove the ambiguity.","section":"Fig. 4 caption and §2"},{"comment":"The comparison to the theoretical phase diagram is only qualitative and the density axis is not calibrated. Fig. 1 is a sketch adapted from Refs. [6,14], the experiment measures only a global change in density from PIV (Fig. 2), and the local packing fraction around the tracked tracer is not measured. The claim of agreement with the mean-field reentrant phase diagram would require converting the global densification curve into a local volume fraction at the tracer's position and locating the phase boundaries with uncertainties; as it stands, the comparison is an interpretation rather than a quantitative test.","section":"§2 and Fig. 1"}],"minor_comments":[{"comment":"Equation (1) has a typographical imbalance in parentheses: the expression contains an unmatched closing parenthesis after the first term. Please correct the typography.","section":"Eq. (1)"},{"comment":"The color coding of the four regimes is difficult to parse in printed or grayscale versions; please add distinct line styles or markers in addition to color.","section":"Fig. 4"},{"comment":"The statement that \"the transition from ballistic to logarithmic behavior in the MSD has never been observed\" should be qualified with respect to Ref. [28], which reports Gardner-phase evidence in a quasi-two-dimensional granular system; the distinction between that prior work and the present experiment should be made explicit.","section":"Introduction"},{"comment":"Because the entire inference rests on a single tracked trajectory, please consider depositing the raw tracked positions, the analysis scripts, and the fitted parameters in a permanent public repository rather than providing them only upon request.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is interesting and the experimental technique is well motivated, but the evidence presented is thinner than the abstract claims. I recommend that the editor require the quantitative and statistical support described in the major comments before considering publication; if replicate trajectories or fixed-density controls cannot be supplied, the conclusions should be softened from \"direct experimental evidence\" to \"observations compatible with.\" The reference list is appropriate and I see no citation-practice or scope concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here is the short version: this paper is worth reading and worth sending to referees, but the main claim is not yet supported by the evidence as presented. The genuinely new thing is the assay and the reported sequence: stable glass, marginal glass, reentrant stable glass in a sedimenting 3D thermal colloidal glass, identified via the short-time single-tracer MSD. Prior experimental work was a 2D granular system (Seguin-Dauchot) and indirect dielectric measurements (Geirhos et al.); nobody has claimed this reentrant sequence in a 3D thermal colloidal system. The idea of probing the ballistic-to-caged crossover is clever and avoids the long-time-equilibration problem.\n\nWhat the paper does well: it is honest about the contested status of the Gardner transition, citing the simulations that argue it may be avoided in finite dimensions (refs 12 and 27). It explicitly states that the system is still densifying throughout, and the density never stops changing. The references are appropriate and include competing work; there is no self-citation problem beyond citing their own earlier method paper, which is relevant. No circularity: the phase diagram is taken from theory and used as a benchmark; the log-fit parameters are descriptive.\n\nSoft spots, in proportion. The decisive classification is of the blue MSD curves as logarithmic via Eq. (2). The system is still compacting at all measured times; logarithmic growth is a generic symptom of slow aging or cage compaction in dense suspensions. The paper offers no comparison against aging or power-law models, no error bars, no replicate tracers, no local packing fraction near the tracer, and no quantitative criterion for what counts as a 'good logarithmic fit.' There are also internal timing inconsistencies: the Fig. 4 caption says logarithmic from 320 to 400 min, while the text says 'after about 340' and later 'after about 380.' These could be minor editing slips, but they matter because the phase is assigned post-hoc to a window of elapsed time. Also, 'these are the only curves to demonstrate a good logarithmic fit' is asserted rather than shown. The data availability is stated, but 'upon request' is weak; there is no repository.\n\nMy read: the central sequence might well be real, but the current evidence underdetermines it. This is a plausibility argument, not a demonstration. For peer review: send it out, but expect heavy revision. The paper deserves referee time because it is the first experimental report of this reentrant sequence and it could sharpen the Gardner debate. I would ask for quantitative fits with alternative models, an ensemble of tracers, and at least one fixed-density control or a local density measurement near the tracer.","headline":"A clever short-time MSD assay reports the first 3D thermal colloidal marginal-glass sequence, but the central phase claim rests on one tracer's log-shaped MSD taken while the sample is still densifying.","tokens_in":8601,"tokens_out":2206,"would_cite":false,"duration_ms":22784,"reading_group":"maybe","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 claims that a densifying colloidal glass passes through a predicted 'marginal' phase, visible as logarithmic growth in a single tracer particle's motion.","keywords":["marginal glass","Gardner phase","colloidal glass","mean squared displacement","replica theory of glasses","stable glass","reentrant glass transition","single-particle tracking"],"falsifier":"Track several tracer colloids in the same sedimenting sample and measure the local packing fraction around each one. If the marginal-glass claim is right, the logarithmic regime should begin for every tracer when its local density crosses the same value, and the logarithmic slope should be reproducible across tracers at matched density; if the logarithmic shape depends on the chosen tracer or on global sedimentation time rather than local density, the central claim fails. A second check is to compare a sedimenting sample with its drift subtracted using a fixed reference and with an intentionally immobilized tracer; if the logarithmic regime disappears under those controls, it was an artifact.","tokens_in":7540,"feed_emoji":"🔬","tokens_out":7142,"duration_ms":66617,"temperature":0.7,"pith_summary":"This paper claims that a single tracer colloid in a slowly densifying three-dimensional colloidal glass acts as a probe of the glass phase diagram, and that the probe's mean squared displacement reveals four successive regimes: a sub-diffusive liquid, a stable glass with a flat plateau, a marginal glass whose motion grows logarithmically in time, and a return to stable glass at the highest densities. The measurements are taken in the short-time ballistic-to-cage regime, which sidesteps the slow aging that makes long-time glass measurements difficult. The authors interpret the logarithmic regime as the signature of the marginal, or Gardner, glass phase predicted by replica mean-field theory, and they take the overall sequence as direct experimental evidence that this phase exists in three-dimensional thermal colloids. A sympathetic reader would care because the result would carry the Gardner transition out of infinite-dimensional theory and simulation and into laboratory matter.","feed_headline":"A tracer bead finds the predicted 'marginal glass' phase in 3D","feed_subtitle":"A slowly settling colloidal glass shows stable, marginal, then stable motion—evidence the Gardner transition is real in 3D.","key_machinery":"The central object is the drift-subtracted mean squared displacement of a single tracer particle, computed from 64,000-frames-per-second video and corrected for localization noise. Its shape is classified against a crossover time $\\tau^*$ where the tracer first feels its neighbors: beyond $\\tau^*$, a flat plateau means a smooth energy basin (stable glass), while a logarithmic law $\\mathrm{MSD}(\\tau) = a \\log(\\tau/\\tau^*) + x^*$ means a basin hierarchically split into sub-basins (marginal glass). Slow gravity-driven densification sweeps the packing fraction upward by about four percent over hundreds of minutes at fixed temperature, so one continuous trajectory of MSD curves maps out successive phases. The logarithmic fit parameter $a$ is the quantitative handle on how fast the system explores the sub-basin hierarchy.","core_discovery":"The paper's central claim is that the short-time mean squared displacement (MSD) of one tracer colloid in a sedimenting colloidal glass is a phase assay, and by that assay a fixed-temperature system traverses stable glass, marginal glass, and stable glass again as density rises. At the lowest densities the MSD is ballistic at short times and sub-diffusive at long times, the liquid signal. After roughly 180 minutes of sedimentation a flat plateau appears and shrinks with increasing density; this is the smooth-basin signature of a stable glass. The plateau does not shrink to zero. Instead, from roughly 340 to 420 minutes the MSD is well fit by a logarithm growing toward the earlier plateau value, which the paper identifies with exploration of a hierarchy of sub-basins within sub-basins, the marginal phase. At the highest densities the logarithmic regime vanishes and a plateau reappears at almost the same height, the re-entrant stable glass. The paper compares this sequence to the mean-field phase diagram for soft-sphere glasses and finds agreement.","pith_inferences":["One extension the authors leave implicit is that the same short-time MSD protocol could be applied to many tracers simultaneously; if the Gardner interpretation is right, each tracer should enter the logarithmic regime at the same local density, and the spread of those onsets would measure the sharpness of the transition.","A natural further test is to examine higher-order statistics, such as the non-Gaussian parameter or displacement kurtosis, over the same trajectories; a hierarchical sub-basin landscape would predict intermittent rearrangements whose statistics differ from a smooth cage.","The protocol could be transferred to other glass formers, including emulsions or granular systems with thermalized grains, to see whether the same logarithmic regime appears whenever a glass is densified at fixed temperature."],"forward_implications":["If the marginal phase is real in three-dimensional colloids, then the replica-theory phase diagram, including the re-entrant stable-glass branch, applies to laboratory materials rather than only to infinite-dimensional models.","The short-time MSD protocol gives a practical assay for locating the Gardner transition: repeating the sedimentation experiment at other temperatures would map the boundaries of the marginal phase.","The logarithmic growth observed here implies that the tracer's cage is not a smooth bowl but a nested hierarchy of barriers, so caged dynamics in dense colloids should retain long-time memory of the sub-basin structure.","Because the re-entrant plateau appears within the same sample without a change in temperature, the stable-to-marginal-to-stable sequence is a density-driven transition, and its sharpness can be tested by measuring the logarithmic slope as a function of elapsed time."],"supporting_citations":[{"why":"Supplies the mean-field prediction of a re-entrant stable glass and the phase diagram the experiment is compared against.","marker":"[6]"},{"why":"Gives the mean-field marginally stable phases and the re-entrant Gardner phase diagram used as the reference for the observed sequence.","marker":"[14]"},{"why":"Derives how a hierarchy of sub-basins merges into logarithmic growth of the mean squared displacement.","marker":"[7]"},{"why":"Reviews Gardner physics and the logarithmic signatures expected in amorphous solids.","marker":"[8]"},{"why":"Provides numerical MSDs with logarithmic-like aging in colloidal and granular glasses that the experiment mirrors.","marker":"[13]"},{"why":"Supplies the earlier experimental evidence of a Gardner phase in a two-dimensional granular glass that this experiment extends to three dimensions.","marker":"[28]"},{"why":"Gives previous experimental evidence compatible with a Gardner transition in molecular glasses, supporting the phase's physical reality.","marker":"[29]"},{"why":"Validates the short-time tracking and localization-error-subtraction method used to measure the MSD below the nominal noise floor.","marker":"[32]"}],"fun_headline_variants":["First 3D sighting of the elusive 'marginal glass' phase","Tracer colloid reveals predicted marginal glass phase","Experimental evidence of marginal glass in 3D colloids","Short-time dynamics expose the marginal glass phase","Colloidal glass shows predicted 'marginal' phase"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a logarithmically growing mean squared displacement in one tracer particle is an unambiguous fingerprint of the marginal glass phase, rather than a product of ordinary aging, polydispersity, residual drift, or the tracer's particular local cage.","fun_headline_variants_meta":{"raw":{"variants":["First 3D sighting of the elusive 'marginal glass' phase","Tracer colloid reveals predicted marginal glass phase","Experimental evidence of marginal glass in 3D colloids","Short-time dynamics expose the marginal glass phase","Colloidal glass shows predicted 'marginal' phase"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001002,"raw_usage":{"total_tokens":4250,"prompt_tokens":963,"completion_tokens":3287,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":3206}},"tokens_in":579,"tokens_out":3287,"duration_ms":21728,"temperature":1.0,"reasoning_tokens":3206,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:47:50.111620+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Track several tracer colloids in the same sedimenting sample and measure the local packing fraction around each one. If the marginal-glass claim is right, the logarithmic regime should begin for every tracer when its local density crosses the same value, and the logarithmic slope should be reproducible across tracers at matched density; if the logarithmic shape depends on the chosen tracer or on global sedimentation time rather than local density, the central claim fails. A second check is to compare a sedimenting sample with its drift subtracted using a fixed reference and with an intentionally immobilized tracer; if the logarithmic regime disappears under those controls, it was an artifact.","supporting_citations":[{"cited_title":"Nuclear Physics B 257:747–765","cited_arxiv_id":null,"evidence_quote":"Supplies the mean-field prediction of a re-entrant stable glass and the phase diagram the experiment is compared against."},{"cited_title":"(2016) Growing timescales and lengthscales characterizing vibrations of amorphous solids","cited_arxiv_id":null,"evidence_quote":"Gives the mean-field marginally stable phases and the re-entrant Gardner phase diagram used as the reference for the observed sequence."},{"cited_title":"The Journal of Chemical Physics 131(2):024504","cited_arxiv_id":null,"evidence_quote":"Derives how a hierarchy of sub-basins merges into logarithmic growth of the mean squared displacement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reviews Gardner physics and the logarithmic signatures expected in amorphous solids."},{"cited_title":"(2015) Numerical detection of the Gardner transition in a mean-field glass former","cited_arxiv_id":null,"evidence_quote":"Provides numerical MSDs with logarithmic-like aging in colloidal and granular glasses that the experiment mirrors."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the earlier experimental evidence of a Gardner phase in a two-dimensional granular glass that this experiment extends to three dimensions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives previous experimental evidence compatible with a Gardner transition in molecular glasses, supporting the phase's physical reality."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Validates the short-time tracking and localization-error-subtraction method used to measure the MSD below the nominal noise floor."}],"review_version":1}