{"id":"38b118f2-4c7d-4b94-9abd-178a803ae2de","arxiv_id":"2507.16638","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"The ferromagnetic phase of FeRh nucleates at the surface as roughly 30 nm columnar domains during the laser-driven transition, according to the authors' ultrafast X-ray sonography analysis.","lead":"Using ultrafast X-ray diffraction and laser-generated sound pulses, the authors tracked where the ferromagnetic phase appears inside a 44 nm FeRh film after laser excitation. They conclude it nucleates at the surface in narrow columnar domains about 30 nm wide, a spatial detail hidden in earlier averaging measurements.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Scenario ranking and the 30 nm domain size are never tested in one self-consistent model: the global chi-square comparison omits the in-plane expansion broadening that later determines the domain diameter, so the central claim may rest on an incomplete forward model.","rationale":"The paper's advance is a new way to turn strain-pulse UXRD data into 3D phase-heterogeneity information. The load-bearing quantity is therefore not just the 30 nm number, but the combination 'near-surface + columnar + 30 nm'. The reader's verdict already flags the 1D elastic model and the 5-site domain-size model as separate assumptions. My stress-test identifies a stronger version of the same worry: the two halves of the claim are not just parameter-sensitive, they are computed with two mutually inconsistent forward models, and the scenario ranking never sees the broadening that later determines the diameter. This makes the central claim internally underdetermined until the two models are merged. Secondary concerns — the ad hoc 5.5 nm dead layer, the optimized optical penetration depth, the 1.3 factor for scenario IV, and the absence of error bars on the chi-square values — are real but would only shift parameters within the 1D framework; the omission of the lateral broadening is a structural gap in the argument. The proposed test is a single simulation change: add the already-constructed 5-site broadening to the already-used udkm1Dsim intensities and see whether scenario III still wins. This is inexpensive, uses only the paper's own models, and directly settles the internal consistency of the claim. I therefore do not change the reader's CONDITIONAL verdict, but the condition should explicitly include this self-consistency test.","tokens_in":17334,"tokens_out":11777,"duration_ms":133181,"concrete_test":"Re-run the global chi-square scenario comparison of Fig. 4d with the 5-site in-plane-expansion broadening (simultaneously fitting lFM) incorporated into the FM component of the simulated sonograms for scenarios I–IV. If scenario III remains the global minimum with a 30-nm-scale lFM, the central claim is self-consistently supported; if scenario I or IV becomes competitive, or lFM shifts beyond the stated 10 nm uncertainty, the conclusion is an artifact of using two incompatible forward models.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim — FM nucleates at the surface in narrow, vertically extended 30 nm columns — is assembled from two disjoint analyses. The scenario ranking (Fig. 4d) compares experiment with sonograms produced by udkm1Dsim, a 1D elastic model in which lateral heterogeneity enters only through coverage/weighting, not through the in-plane lattice expansion that lateral nucleation unlocks (Methods, 'Modelling of sonograms'). The 30 nm diameter is then estimated from the residual FM peak broadening of 0.06 Å^-1 (Extended Data E1c) using a separate 5-site stochastic model of in-plane expansion (Methods, 'Estimating the in-plane domain dimension'). Because the global chi-square used to identify scenario III is computed from model intensities that explicitly omit this broadening, the very signal that fixes the domain size contributes to the chi-square residual during scenario selection instead of being part of the model. The paper acknowledges this omission: the domain-dimension discussion states that the additional broadening is 'not captured by our one-dimensional model'. Consequently, there is no self-consistent check that the geometry of scenario III plus 30 nm domains reproduces the full measured I(t,qz), and the scenario ranking could be biased by a model that is known to be incomplete in exactly the channel later used to extract the headline number.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces ultrafast X-ray sonography, a pump-probe technique that combines time-resolved hard-X-ray diffraction with laser-launched strain pulses to image the spatial heterogeneity of a laser-induced phase transition. The method is demonstrated on the antiferromagnetic-to-ferromagnetic transition in a 44 nm FeRh film. The authors compare measured sonograms with one-dimensional elastic simulations for five nucleation scenarios and, using a global chi-square residual, select scenario III: the FM phase nucleates as columns in the near-surface region and coalesces into a continuous layer. They then estimate the in-plane domain diameter to be approximately 30 nm from an additional FM Bragg peak broadening that is not captured by their 1D model, using a separate five-site stochastic model. The fluence dependence of the scenario selection is also presented, showing scenario III for fluences above 5.2 mJ/cm2 and scenario IV for the lowest fluence.","tokens_in":17646,"tokens_out":5749,"duration_ms":60549,"significance":"If correct, this work introduces a broadly applicable and minimally invasive method for probing nanoscale phase heterogeneity in ultrafast phase transitions, and it offers a way to reconcile conflicting FeRh results in the literature. The strength of the paper is that the scenario comparison is a genuine model selection against measured intensity maps, not forced by construction, and the fluence-dependent consistency provides a meaningful internal check. The principal weakness is that the quantitative support for the headline 30 nm domain size is obtained from a different forward model than the one used to rank the scenarios, and the model used for ranking is explicitly incomplete in exactly the channel used for the size estimate. This is a load-bearing issue for the central claim, but it is testable and addressable within the scope of the manuscript.","major_comments":[{"comment":"The central claim — FM nucleates in near-surface columns of about 30 nm diameter — is assembled from two disjoint forward models. The sonogram simulations used for the global chi-square ranking (Fig. 4d) are produced by udkm1Dsim, a 1D elastic model in which lateral heterogeneity enters only through incoherent averaging over coverage, not through in-plane lattice expansion; the Methods explicitly state that the additional FM Bragg peak broadening 'is not captured by our modelling.' That same residual broadening is then used to infer the 30 nm diameter in a separate 5-site stochastic model. Consequently, the 30 nm parameter has no effect on the simulated I(t,qz) used to select scenario III, and the unmodeled broadening contributes to the chi-square residual as a missing signal rather than as a model prediction. There is no self-consistent check that the geometry of scenario III plus 30 nm domains reproduces the complete measured I(t,qz). I request that the authors either incorporate the in-plane broadening into the forward model and re-run the scenario comparison, or demonstrate explicitly that the scenario ranking is unchanged when the FM peak broadening is excluded from or added to the residual.","section":"Modelling of sonograms / Estimating the in-plane domain dimension"},{"comment":"The 'global residual χ2' is a sum of squared normalized-intensity differences without an explicit noise model. The measured I(t,qz) is presented as a single train-averaged curve, and no error bars or confidence bands are provided. The paper asserts that the global analysis 'identifies scenario III to optimally describe' the data, but without a statistical measure (e.g., a reduced chi-square with estimated uncertainties or a likelihood-ratio test between scenarios), the separation between scenarios I, III, and IV may not be significant. Given that the distinction hinges on subtle FM-peak position shifts at early delays, please provide an uncertainty estimate for the chi-square values or otherwise quantify whether the scenario ranking is robust to data noise.","section":"Ultrafast domain nucleation in FeRh / Fig. 4"},{"comment":"The scenario ranking relies on a calibrated 1D elastic model whose free choices are not fully sensitivity-tested. The Methods state 'we use essentially the already calibrated parameters [26,34]' and 'only the optical penetration depth is optimised,' while also introducing an ad hoc assumption that the bottom-most 5.5 nm of FeRh does not contribute to the Bragg peak. Other parameters, such as the 0.6% phase-expansion amplitude, the fixed nucleation time τ=8 ps, and the scenario-IV thickness coefficient, enter the simulation. I ask for a sensitivity analysis showing how the scenario ranking and the inferred V*FM values in Fig. 4d and Extended Data Fig. E3 respond to plausible variations of these parameters; without it, the quantitative support for scenario III as the unique optimum is not fully established.","section":"Modelling of sonograms"}],"minor_comments":[{"comment":"The sentence 'It is the very same same sample as in a previous publication [18]' contains a duplicated word ('same same') that should be corrected.","section":"Methods (Sample growth and characterisation)"},{"comment":"The axis label in Fig. 4a reads 'qz (Å)' but should be 'qz (Å⁻¹)' to be consistent with the rest of the text and figures.","section":"Fig. 4a"},{"comment":"The phrase 'even larger as scenario II' should read 'even larger than scenario II.'","section":"Extended Data Fig. E3 caption"},{"comment":"The estimate of the uncertainty (±10 nm) is based on two stated sources, but the text does not explain how the 0.06 Å⁻¹ broadening value itself is determined from the fitted width in Extended Data Fig. E1c; a brief description of that measurement would aid reproducibility.","section":"Estimating the in-plane domain dimension"}],"recommendation":"major_revision","confidential_remarks":"The paper is a good fit for the journal and the experimental effort is substantial. The main methodological innovation is compelling, and the scenario comparison is not circular. However, the disjoint forward models for scenario selection and domain sizing create a real gap in the evidence for the headline claim. This is addressable by additional modelling and analysis, so I recommend major revision rather than rejection; no new experiment is required to close the gap."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the experiment is real, the sonography concept is clever, and the fluence series gives the scenario comparison real teeth. The paper claims FM nucleates near the surface as roughly 30 nm columnar domains that coalesce into a layer, and this is not just a rehash of prior volume-fraction measurements. The four-fluence consistency, the use of already calibrated parameters on the same sample, and the explicit listing of ad hoc choices (5.5 nm dead layer, optimized optical penetration, the 1.3 factor for scenario IV) all speak well of the authors.\n\nThe soft spot the stress-test note identifies is genuine and central. The global chi-square that selects scenario III is computed from a 1D model that, by the authors' own admission, does not capture the in-plane expansion broadening. That same broadening is then used, in a separate five-site stochastic model, to extract the 30 nm domain diameter. So there is no forward model that checks whether scenario III plus 30 nm domains reproduces the full measured I(t,qz). The scenario ranking could in principle be biased by this omission, and the headline number is an indirect inference rather than a self-consistent fit. The authors are honest that the broadening is \"not captured by our one-dimensional model,\" but the paper still presents the 30 nm number as a result without showing that the scenario choice survives when that channel is modeled.\n\nOther issues are secondary but real. The chi-square rankings have no uncertainty estimates. The dead-layer thickness and optimized penetration depth are not varied to show the scenario ranking is robust. The claimed 10 nm uncertainty on the domain size looks like a judgment call rather than a propagated error. No data or code are deposited, only \"available upon request.\"\n\nNone of this falsifies the central claim. The qualitative exclusions—pure in-plane heterogeneity and backside nucleation—look solid, and scenario III winning over I and IV is plausible. But the paper stops one step short of the self-consistency check that would make the 30 nm result convincing.\n\nThis paper deserves a serious referee. I would send it out and ask for a revision that either incorporates the in-plane domain broadening into the global chi-square fit or explicitly demonstrates that the scenario ranking is unchanged when that broadening is added, plus uncertainty estimates and deposited code/data. The idea is worth publishing; the central number needs to be earned in one model.","headline":"A genuinely new spatial claim about FeRh nucleation, but the 30 nm column diameter and the scenario ranking rest on two disjoint models that are never checked in one self-consistent forward calculation.","tokens_in":18209,"tokens_out":2220,"would_cite":true,"duration_ms":27299,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["61.05.cp","75.30.Kz"],"model":"deepseek-v4-flash","headline":"Strain pulses launched by the driving laser act as an internal sonar probe, showing that FeRh's ferromagnetic phase nucleates at the surface in columnar domains roughly 30 nm wide that coalesce into a layer.","keywords":["phase transitions","phase heterogeneity","ultrafast X-ray diffraction","X-ray sonography","FeRh","magneto-structural phase transition","strain pulses","domain nucleation"],"falsifier":"A real-space imaging measurement with nanometre resolution — for example single-shot coherent X-ray imaging or time-resolved X-ray nanodiffraction on the same FeRh film at the same $7.7\\,\\text{mJ cm}^{-2}$ fluence — that directly shows whether the ferromagnetic phase appears as roughly 30 nm wide columnar domains at the surface that coalesce into a layer, or instead as a uniform surface layer or as full-thickness columns. A second check would vary the film's mosaic grain size and test whether the inferred domain diameter follows the grain size, confirming that the 30 nm value is set by microstructure.","tokens_in":17136,"feed_emoji":"🔬","tokens_out":15366,"duration_ms":142783,"temperature":0.7,"pith_summary":"The paper introduces ultrafast X-ray sonography: a propagating strain pulse, launched by the same laser that drives a phase transition, is used as a non-invasive structural probe that locates where in a thin film the new phase appears and how it spreads. Applied to the antiferromagnetic-to-ferromagnetic transition in FeRh, the method shows the ferromagnetic phase nucleating at the surface in narrow, vertically extended columnar domains with a diameter of roughly 30 nm, which then coalesce into a continuous layer. A sympathetic reader should care because this is the kind of three-dimensional, growth-resolved detail that ordinary ultrafast diffraction averages away, and because the picture reconciles the fast structural rise with the slower magnetization rise seen in earlier FeRh experiments. The approach is claimed to be sample-agnostic and applicable to any phase transition whose coexisting phases have distinguishable diffraction signatures.","feed_headline":"30 nm surface columns: where FeRh's new phase is born","feed_subtitle":"Laser-made strain pulses act as built-in sonar, revealing where a phase transition starts and how it grows.","key_machinery":"The central object is the phase-specific strain response induced by a propagating bipolar strain pulse. Because a layer's average strain is nonzero only while the compressive and expansive halves of the pulse are unbalanced inside it, the layer's Bragg peak shifts in a characteristic timing pattern, and that timing decodes the location of each phase: the phase thickness follows from $d_{\\mathrm{phase}} = \\Delta t_{\\mathrm{phase}} / v_s$, and the volume fraction splits into thickness and in-plane coverage through $V_{\\mathrm{phase}}(t) = D_{\\mathrm{phase}}(t)\\, A_{\\mathrm{phase}}(t)$. In the FeRh experiment the optical pump simultaneously drives the phase transition and launches the strain pulse, so no dedicated transducer is needed. The quantitative scenario selection solves the linear one-dimensional elastic wave equation with thermophysical parameters calibrated on earlier measurements of the same sample, feeds the resulting strain into a dynamical X-ray scattering calculation, and averages incoherently over stochastic nucleation delays.","core_discovery":"On the paper's own terms, the discovery is that the ferromagnetic phase of FeRh created by an intense femtosecond laser pulse does not appear as a uniform surface layer, nor as columns spanning the full film thickness, but as narrow columnar domains about 30 nm in diameter that nucleate in the near-surface region and later merge into a continuous layer. This identification comes from comparing the full experimental sonogram — diffracted X-ray intensity as a function of both delay time and out-of-plane reciprocal coordinate — with simulated sonograms for five distinct nucleation scenarios; a global $\\chi^2$ analysis selects the near-surface-column scenario (III) across all pump-probe delays, and the same scenario fits three of the four laser fluences studied, while the lowest fluence leaves the film laterally heterogeneous with partial in-plane coverage. The paper further finds that the depth of the ferromagnetic phase tracks the depth at which the optical excitation overcomes the equilibrium transition threshold, revealing the thermal character of the transition even on its non-equilibrium pathway.","pith_inferences":["Because the depth information is carried by the strain-pulse transit time, the technique will discriminate best in films whose thickness is comparable to or larger than the pulse's spatial extent; for a few-nanometre layer the sonogram would essentially collapse, and the surface-versus-bulk distinction that carries the FeRh conclusion would be lost.","The match between the 30 nm domain diameter and the 25 nm mosaic grain size suggests a testable prediction the paper does not make: engineering the grain size of FeRh films through substrate choice, annealing, or ion bombardment should move the inferred column diameter, which would show that the value is set by microstructure rather than by intrinsic physics.","The broken-in-plane-symmetry argument — heterogeneous nucleation unlocks an in-plane expansion that broadens the out-of-plane Bragg peak — could serve as a general, probe-agnostic indicator of lateral phase coexistence in other materials, even where the coexisting phases have nearly identical out-of-plane lattice constants.","If the FeRh picture is correct, reported 'switching speeds' of magnetostructural transitions can differ by an order of magnitude depending on whether the probe reads lattice volume or magnetization; comparing the two signals is itself a diagnostic of near-surface columnar nucleation in other magnetic materials."],"forward_implications":["The conflicting FeRh observations in the literature are reconciled: the structural Bragg signal rises on the 8 ps nucleation timescale because near-surface columnar formation changes lattice volume quickly, while the macroscopic magnetization rises more slowly because the freshly nucleated domains start with magnetization along different magnetic easy axes and only align later through domain-wall ","Because the pump that drives the transition also launches the probing strain pulse, the method needs no added transducer and almost no sample preparation, and it works for any material whose coexisting phases give distinguishable diffraction signatures.","The fluence series shows the non-equilibrium pathway still obeys a thermal logic: the near-surface extent of the ferromagnetic phase equals the depth at which the optical excitation crosses the equilibrium transition threshold, and only the lowest studied fluence leaves the film laterally heterogeneous with partial in-plane coverage.","The similar sizes of the nucleating ferromagnetic domains (about 30 nm) and the film's mosaic crystallites (about 25 nm) point to structural granularity as a key factor controlling where and how fast long-range ferromagnetic order emerges."],"supporting_citations":[{"why":"Earlier ultrafast X-ray diffraction on FeRh whose fast structural rise and slow magnetization behaviour this paper's columnar-nucleation picture reconciles.","marker":"[17]"},{"why":"Prior experiment on the very same FeRh sample; supplies the calibrated thermophysical parameters and the 8 ps nucleation timescale used in the model.","marker":"[18]"},{"why":"Measurement of the slower macroscopic magnetization rise in FeRh that the near-surface columnar domain picture is invoked to explain.","marker":"[25]"},{"why":"Earlier weak-excitation strain measurements on FeRh that calibrated the model's thermophysical parameters and pinned the 0.6% volume expansion of the FM phase.","marker":"[26]"},{"why":"The concepts and use cases of picosecond ultrasonics with X-rays on which the sonography method is built.","marker":"[27]"},{"why":"Static characterization of the FeRh film, including the roughly 25 nm mosaic crystallite size used to estimate the 30 nm domain diameter.","marker":"[32]"},{"why":"Establishes the non-thermal-electron pathway and the 8 ps nucleation timescale that enters Eq. (3) for the FM volume fraction.","marker":"[34]"},{"why":"The simulation toolbox used to solve the 1D elastic wave equation and compute the modelled sonograms.","marker":"[41]"}],"fun_headline_variants":["FeRh phase birth mapped: 30 nm surface columns","X-ray sonography reveals FeRh's 30 nm nucleation columns","Ultrafast sonogram: FeRh's new phase starts in 30 nm columns","30 nm columns: FeRh's laser-driven phase nucleates at surface","FeRh's magnetic phase emerges as narrow 30 nm surface columns"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The scenario ranking and the roughly 30 nm column diameter stand on a one-dimensional elastic model whose thermophysical parameters were calibrated in earlier experiments on the same sample — including the choice to ignore the bottommost 5.5 nm of the FeRh layer — and on the assumption that the extra broadening of the ferromagnetic Bragg peak comes entirely from in-plane expansion as captured by a simplified five-site stochastic model.","fun_headline_variants_meta":{"raw":{"variants":["FeRh phase birth mapped: 30 nm surface columns","X-ray sonography reveals FeRh's 30 nm nucleation columns","Ultrafast sonogram: FeRh's new phase starts in 30 nm columns","30 nm columns: FeRh's laser-driven phase nucleates at surface","FeRh's magnetic phase emerges as narrow 30 nm surface columns"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000194,"raw_usage":{"total_tokens":1342,"prompt_tokens":923,"completion_tokens":419,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":325}},"tokens_in":539,"tokens_out":419,"duration_ms":4713,"temperature":1.0,"reasoning_tokens":325,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:05:18.045366+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A real-space imaging measurement with nanometre resolution — for example single-shot coherent X-ray imaging or time-resolved X-ray nanodiffraction on the same FeRh film at the same $7.7\\,\\text{mJ cm}^{-2}$ fluence — that directly shows whether the ferromagnetic phase appears as roughly 30 nm wide columnar domains at the surface that coalesce into a layer, or instead as a uniform surface layer or as full-thickness columns. A second check would vary the film's mosaic grain size and test whether the inferred domain diameter follows the grain size, confirming that the 30 nm value is set by microstructure.","supporting_citations":[{"cited_title":"Physical Review Letters 108(8), 087201 (2012) https://doi.org/10","cited_arxiv_id":null,"evidence_quote":"Earlier ultrafast X-ray diffraction on FeRh whose fast structural rise and slow magnetization behaviour this paper's columnar-nucleation picture reconciles."},{"cited_title":"APL Materials 12(5) (2024) https://doi.org/ 10.1063/5.0206095","cited_arxiv_id":null,"evidence_quote":"Prior experiment on the very same FeRh sample; supplies the calibrated thermophysical parameters and the 8 ps nucleation timescale used in the model."},{"cited_title":"Nature Com- munications 13(1), 2998 (2022) https: //doi.org/10.1038/s41467-022-30591-2","cited_arxiv_id":null,"evidence_quote":"Measurement of the slower macroscopic magnetization rise in FeRh that the near-surface columnar domain picture is invoked to explain."},{"cited_title":"Advanced Functional Materials 34(32), 2313014 (2024) https://doi.org/10.1002/adfm.202313014","cited_arxiv_id":null,"evidence_quote":"Earlier weak-excitation strain measurements on FeRh that calibrated the model's thermophysical parameters and pinned the 0.6% volume expansion of the FM phase."},{"cited_title":"Physical Review B 101(17), 174413 (2020) https://doi.org/10","cited_arxiv_id":null,"evidence_quote":"Static characterization of the FeRh film, including the roughly 25 nm mosaic crystallite size used to estimate the 30 nm domain diameter."},{"cited_title":"Communica- tions Physics 8(1), 140 (2025) https: //doi.org/10.1038/s42005-025-02066-5","cited_arxiv_id":null,"evidence_quote":"Establishes the non-thermal-electron pathway and the 8 ps nucleation timescale that enters Eq. (3) for the FM volume fraction."}],"review_version":1}