{"id":"f36ed8f3-a321-4357-b774-3fe9c4c6cce4","arxiv_id":"2510.12593","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Tin growth on Sb-doped MnBi2Te4 shows a crossover from conserved KPZ to temporally-correlated-noise KPZ scaling at t≈5.4 min, attributed to adatom escape.","lead":"In tin films grown on a doped bismuth-telluride crystal, the surface-roughening behavior abruptly changes at a critical time, switching from one universal growth law to another. The paper argues this is the first experimental sign of a predicted effect where noise that is correlated in time, caused by atoms escaping the film, rewrites the growth statistics.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"MD crossover is imposed by a substrate switch rather than emerging from a single self-contained model; the escape-noise mechanism is therefore not independently established.","rationale":"After reading the manuscript in good faith, the scaling analysis and the two-regime fits are internally plausible. The experimental values (beta1=0.21, alpha_loc1=0.71, 1/z1=0.31; beta2=0.66, alpha_loc2=0.87, 1/z2=0.74) match the cited theoretical predictions within error, and the crossover at 5.4 min is consistently seen in roughness, HHCF, correlation length, structure factor, and AES onset. That consistency is real supporting evidence. However, the paper's distinctive claim is not just an exponent crossover; it is a microscopic mechanism: adatom escape generating temporally correlated noise. That mechanism is supported almost entirely by MD and AES. The MD's use of a substrate switch at the experimentally known transition time creates a circularity: the simulation is not allowed to fail after 7.8 min on the MBST analog because the experiment says the late-time regime is on stanene. Whether that switch is physically legitimate depends on the untested assumption that the full-coverage stanene layer, which the simulation cannot form, would stabilize the same growth. This is the load-bearing condition. The AES slope change is a second leg of evidence, but it is also indirect; morphological shadowing and attenuation can change Auger ratios without a change in sticking. Thus, if the two-substrate protocol is not self-contained, the microscopic-origin claim is unsupported, and the paper reduces to an empirical crossover consistent with TCN-KPZ but not uniquely diagnostic. I do not think this is fatal to the experimental data; it requires a testable revision. The reader already rated the paper CONDITIONAL; my read does not change that verdict.","tokens_in":19290,"tokens_out":5608,"duration_ms":53736,"concrete_test":"Run a single continuous MD simulation of Sn deposition on an H-passivated Bi2Te3/MBST slab (or an MBST slab that naturally develops a full stanene layer) from t=0 through the equivalent of >150 min without switching substrate, using the same LJ potential, flux, and substrate temperature. Compute the height-height correlation exponents and the escape-noise correlation function over the full timeline; if a cKPZ-to-TCN-KPZ crossover with theta~0.45 appears naturally at the experimental rescaled time, the mechanism is internal. If the crossover appears only when the substrate is manually changed to stanene, the MD protocol is imposing the result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that escape-induced temporally correlated noise drives the cKPZ-to-TCN-KPZ crossover rests on the MD simulation being a faithful, self-contained model. The manuscript states that 'reproducing the experimental crossover required two substrates' and that 'we switched the substrate to stanene' after growth on Bi2Te3 fails beyond t=7.8 min. A fully covered stanene layer is never achieved on Bi2Te3 in the simulation, even though experiment shows such a layer forms before the crossover. Thus the simulation's late-time TCN-KPZ behavior is not derived from the same initial condition as the experiment; the switch is motivated by the very experimental outcome the simulation is supposed to explain. The escape-noise correlations, fitted with theta_avg=0.48 (Fig. 4P), are measured in this post-switch stanene model, so they may reflect the model's Sn/stanene interaction rather than the actual MBST/Sn system. If the two-substrate protocol is an encoding of the experimental timeline, the 'microscopic origin' claim collapses even if the experimental exponents stand. The AES slope change is also indirect: the Sn/Te ratio decrease after t=5.4 could be a geometric attenuation effect of 3D islands/facets, not necessarily a change in sticking probability.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an STM, AES, and MD study of Sn thin-film growth on Sb-doped MnBi2Te4. It claims a sharp dynamical crossover at t=5.4 min from conserved KPZ (cKPZ) behavior, with exponents β1=0.21±0.03, α_loc1=0.71±0.03, 1/z1=0.31±0.07, to a temporally correlated noise KPZ (TCN-KPZ) regime, with β2=0.66±0.13, α_loc2=0.87±0.07, 1/z2=0.74±0.05, matching Ref. 25 with θ≈0.45. The authors further present MD simulations that, after switching from a Bi2Te3 to a stanene substrate, reproduce the two regimes and show escape-induced temporally correlated noise with θ_avg=0.48±0.01, which they identify as the microscopic origin of the crossover.","tokens_in":19722,"tokens_out":4823,"duration_ms":44296,"significance":"If the scaling analysis and microscopic attribution hold, this would be a notable first experimental realization of temporal noise correlations changing the universality class of a growing interface. The paper has clear strengths: atomically resolved STM over a wide time range, multiple independent observables (ω, α_loc, 1/z, α_s, AES) all changing near the same time, and a plausible MD mechanism. The experimental exponent extraction is, however, built on a single cumulative growth sequence with no replicate noise realizations, and the MD crossover is imposed by a two-substrate protocol rather than emerging from one self-contained model. The experimental scaling observations are significant, but the mechanistic conclusion as stated is stronger than the evidence supports.","major_comments":[{"comment":"The central mechanistic claim — that escape-induced temporally correlated noise drives the cKPZ-to-TCN-KPZ crossover — is not independently established. The text states that 'reproducing the experimental crossover required two substrates' and 'Guided by experimental observations, we switched the substrate to stanene.' Fig. S12 shows separate early-time Sn/Bi2Te3 and late-time Sn/stanene runs, and the paper admits that a fully covered stanene layer is never achieved on Bi2Te3. Thus the late-time exponents and the escape-noise correlations (Fig. 4O,P; θ_avg=0.48) are obtained from a model chosen to reproduce the experimental outcome, not from a simulation that spontaneously crosses over. This circularity affects the 'microscopic origin' claim. The authors should either provide a single-substrate simulation that produces the crossover without external switching, or explicitly reframe the MD","section":"MD simulations, Fig. 4, Fig. S12, Methods 'Simulation details'"},{"comment":"All experimental exponents are extracted from one cumulative growth sequence; each growth time is a single realization, and no uncertainties from independent runs or bootstrap resampling are provided. The crossover time t=5.4 min is selected from the same data and then used to group points into two regimes, so the reported fit uncertainties understate the uncertainty in the exponents and the sharpness of the crossover. Please add error estimates based on repeated measurements or bootstrapping, and test whether a two-power-law model with t_c as a free parameter is preferred over a single power law (e.g., via F-test, AIC, or chi-square comparison).","section":"Fig. 1P, Fig. 2E,G, Table 1"},{"comment":"The interpretation of the AES Sn/Te intensity-ratio slope change as a reduction in sticking probability is not unique. The same time range also shows a morphological transition from 2D islands to high-aspect-ratio mounds and faceted grains. A decreasing Sn/Te ratio can arise from geometric attenuation and shadowing of Auger electrons by 3D features, or from changes in escape depth, independent of any change in sticking or desorption. Without a quantitative transport model, angle-resolved AES, or another control separating coverage from morphology, the AES results do not provide independent support for the escape mechanism.","section":"AES data, Fig. 2J"},{"comment":"The structure-factor collapse is obtained with exponents that differ from those extracted by direct fits: the early regime uses 1/z1=0.36, α1=0.58 versus measured 1/z1=0.31±0.07, α1=0.68±0.18; the late regime uses 1/z2=0.75, α2=0.88 with an extra ad hoc effective time offset t′=1.2 min. This additional parameter and the discrepancy from the directly measured exponents weaken the claim of universality collapse. Please justify the offset and the exponent choices, or perform a data-driven collapse that does not introduce free parameters beyond the measured exponents.","section":"Fig. 3F,G, structure-factor collapse"}],"minor_comments":[{"comment":"Typos and language: 'surface to vaccum' in the roughness definition; 'satisfies the the condition' in the spectral scaling section; 'Normalished ACF' in Fig. S7 caption.","section":"General"},{"comment":"The TCN-KPZ model row is missing the α_loc value; including it would allow a direct comparison with the experimental and simulated α_loc2 values.","section":"Table 1"},{"comment":"The mapping of MD deposition flux ('one atom per ten MD steps') to the experimental 0.83 ML/min is not derived; please state the assumptions used in the timescale conversion.","section":"Methods (Simulation details)"},{"comment":"The claim λ1=0 is based on only a few early-time points and a fit with large scatter; the conclusion that the early regime is the Lai–Das Sarma model rather than the Das Sarma–Tamborenea model would benefit from a more detailed analysis of the local slope evolution.","section":"Fig. S8 and local slope exponent"}],"recommendation":"major_revision","confidential_remarks":"I would not accept the paper in its current form because the MD-supported microscopic mechanism is, by the authors' own admission, obtained via a substrate switch that is guided by the experimental crossover. The experimental scaling observations are potentially valuable, but the mechanistic narrative needs either a self-contained single-substrate simulation or a careful downgrade of the claim. The lack of error bars from replicate measurements is a further concern that should be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe experimental story is better than the simulation story, and you should read the paper for the former, not the latter. This is the first clean experimental dataset I've seen that shows a kinetic roughening crossover from conserved KPZ to TCN-KPZ. The authors extract the usual scaling exponents (beta, alpha_loc, 1/z, alpha_s) from STM and they all break at the same growth time, ~5.4 min. The early-time values match Lai–Das Sarma conserved KPZ; the late-time values land near the temporally-correlated-noise KPZ predictions of Song and Xia. That is a genuine potential milestone.\n\nWhat they do well: The experimental analysis is fairly thorough. They use HHCF, autocorrelation, structure factors, data collapse, and a local slope exponent that rules out the Das Sarma–Tamborenea class. The AES data showing a drop in Sn/Te ratio at the same time is a nice supporting piece. The crossover is observed in multiple independent observables, so the internal consistency is real.\n\nNow the soft spots. First, each exponent is extracted from a single growth sequence. No replicate depositions, no error bars on the crossover time itself. The uncertainties quoted are fitting errors, not run-to-run scatter. Since the crossover time is picked from the same data, there is some circularity; the multiple-observable agreement mitigates it, but does not eliminate it.\n\nSecond, the MD simulation is not a self-contained model of the experiment. The text admits they needed two substrates: Bi2Te3 for the early stage and stanene for the late stage, switching at the experimentally observed transition. The Bi2Te3 simulation never achieves a full stanene layer; they switch because that is what the experiment does. The late-time TCN-KPZ exponents and the escape-noise correlation theta=0.48 are then extracted from the post-switch stanene model. So the claim that escape-induced noise is the microscopic origin of the crossover is not derived from a single faithful simulation — it's engineered by construction. That is the load-bearing flaw of the paper, and it should be the main target of peer review.\n\nThird, the AES interpretation is plausible but not unique. The Sn/Te slope drop could simply reflect 3D island/facet formation shadowing the Te signal, not a change in sticking probability.\n\nAlso, code and data are not deposited; the MD parameters are described, but \"available on request\" is not enough.\n\nNet: The experimental crossover deserves a serious referee and, if it holds up, it's an important result. The microscopic mechanism is a hypothesis, not an established finding. My recommendation: send to peer review, but explicitly ask the reviewers to assess whether the two-substrate MD is a faithful realization or an encoding of the answer. The authors should be required to either run a single self-contained simulation or soften the mechanistic claim.","headline":"Experimental crossover from cKPZ to TCN-KPZ is plausible and worth refereeing; the MD 'microscopic origin' is imposed by a substrate switch and should not be accepted as is.","tokens_in":20211,"tokens_out":2592,"would_cite":true,"duration_ms":22787,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C24","82C31"],"pacs":["68.55.-a","81.15.Aa","05.40.-a"],"model":"deepseek-v4-flash","headline":"During tin film growth on a doped manganese bismuth telluride substrate, a sharp crossover at 5.4 minutes moves the growing interface from conserved KPZ scaling to a regime governed by temporally correlated noise, with adatom escape identif","keywords":["kinetic roughening","KPZ universality","temporally correlated noise","conserved KPZ","thin film growth","scanning tunneling microscopy","molecular dynamics","adatom escape"],"falsifier":"A direct measurement of the Sn sticking or desorption rate during growth that shows no decrease at t ≈ 5.4 min would falsify the escape-noise mechanism; alternatively, an MD simulation on a single substrate including hydrogen passivation that reproduces both scaling regimes without any particle escape would also falsify it.","tokens_in":19199,"feed_emoji":"⚛️","tokens_out":4664,"duration_ms":39565,"temperature":0.7,"pith_summary":"The paper reports the first experimental realization of a universality crossover during thin-film growth, from the conserved Kardar–Parisi–Zhang (cKPZ) class to the temporally correlated noise KPZ (TCN-KPZ) class. Scanning tunneling microscopy of tin deposited on antimony-doped MnBi2Te4 shows that early growth forms two-dimensional islands and stanene layers with exponents matching cKPZ, while after 5.4 minutes the surface develops clustered mounds and faceted grains with exponents matching TCN-KPZ for a noise correlation exponent near 0.45. The crossover coincides with a measured drop in tin sticking, and molecular dynamics simulations show that escaping adatoms generate temporally correlated noise. If correct, this establishes that temporal noise correlations can fundamentally alter the scaling class of a growing interface, linking atomistic desorption to emergent universal behavior.","feed_headline":"Escaping atoms flip a growing film between two universality classes","feed_subtitle":"Tin growth obeys conserved KPZ scaling until adatom escape triggers correlated noise and a switch to faceted morphology.","key_machinery":"The key object is the interface-height scaling analysis: the local roughness exponent α_loc, growth exponent β, and dynamic exponent z are extracted from STM topographies via height–height correlation functions and the lateral correlation length, while the spectral roughness exponent α_s comes from the structure factor. These are compared to the Lai–Das Sarma conserved-KPZ model and to a KPZ equation with temporally correlated noise of the form ⟨ηη⟩ = 2D δ_{r,r'} |t−t'|^{2θ−1} exp(−|t−t'|²/t₀²). The microscopic mechanism is adatom escape: MD particle tracking shows all deposited atoms stick before the crossover, while afterward a fraction escape, producing the correlated noise; AES confirms","core_discovery":"The central claim is that a growing Sn film on 30% Sb-doped MnBi2Te4 undergoes a sharp dynamical crossover at t = 5.4 min between two distinct universality classes. Early growth follows conserved KPZ scaling, with experimental exponents β1 = 0.21 ± 0.03, α_loc1 = 0.71 ± 0.03, and 1/z1 = 0.31 ± 0.07, consistent with the Lai–Das Sarma conserved KPZ model. Beyond the crossover, the exponents become β2 = 0.66 ± 0.13, α_loc2 = 0.87 ± 0.07, and 1/z2 = 0.74 ± 0.05, matching TCN-KPZ theory with θ ≈ 0.45. Auger electron spectroscopy shows a sharp reduction in Sn sticking probability at the same time, and molecular dynamics simulations with particle tracking reveal that adatom escape generates tempora","pith_inferences":["If escape-induced noise is generic, similar cKPZ-to-TCN-KPZ crossovers should appear in other vapor-deposited films where desorption increases with coverage or temperature; this could be tested by varying deposition rate and substrate temperature while tracking exponents.","The two-substrate MD protocol suggests a concrete test: a single-substrate simulation that includes residual hydrogen passivation should reproduce both regimes without a substrate switch; if it cannot, the microscopic mechanism is less secure.","The finite correlation time t₀ in the noise correlation implies that TCN-KPZ scaling may be universal only at intermediate timescales, with observable deviations at very long times depending on t₀.","The crossover coincides with full first-layer coverage, hinting that the onset of desorption is tied to a change in binding environment (from substrate-supported stanene to Sn-on-Sn), a hypothesis that could be probed by measuring desorption rates on films of different thickness."],"forward_implications":["Temporal noise correlations are experimentally shown to change the universality class of an interface, closing a long-standing gap between theory and experiment.","Controlling sticking or desorption during growth becomes a tunable handle for selecting between layered (cKPZ) and faceted (TCN-KPZ) morphologies.","The measured exponents provide a benchmark for TCN-KPZ theory in (2+1) dimensions at θ ≈ 0.45.","The hybrid noise correlation (power law at short lags, exponential at long lags) implies that real noise sources have a finite correlation time, refining predictions for experimental systems.","The combined method of STM scaling, AES, and MD particle tracking can be applied to other thin-film systems to identify analogous universality crossovers."],"fun_headline_variants":["Adatom escape triggers universality crossover in Sn film growth","Film growth switches universality class when atoms escape","Time-correlated noise from escaping atoms flips growth scaling","Escape-induced noise changes film's universality class mid-growth"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The molecular dynamics simulation reproduces the crossover only by switching substrates partway through (from Bi2Te3 to stanene) guided by experiment; if this two-substrate model is not a faithful representation of the real system, the claim that escape-induced noise is the microscopic origin is unsupported even if the measured exponents stand.","fun_headline_variants_meta":{"raw":{"variants":["Adatom escape triggers universality crossover in Sn film growth","Film growth switches universality class when atoms escape","Time-correlated noise from escaping atoms flips growth scaling","Escape-induced noise changes film's universality class mid-growth"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000678,"raw_usage":{"total_tokens":2939,"prompt_tokens":783,"completion_tokens":2156,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":527,"completion_tokens_details":{"reasoning_tokens":2099}},"tokens_in":527,"tokens_out":2156,"duration_ms":12365,"temperature":1.0,"reasoning_tokens":2099,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T09:54:25.471958+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct measurement of the Sn sticking or desorption rate during growth that shows no decrease at t ≈ 5.4 min would falsify the escape-noise mechanism; alternatively, an MD simulation on a single substrate including hydrogen passivation that reproduces both scaling regimes without any particle escape would also falsify it.","supporting_citations":[],"review_version":1}