{"id":"9700b482-78ea-46f0-864d-2ef5a430af79","arxiv_id":"1908.07276","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Heating BaTiO3 nanoparticles causes paraelectric cubic regions to nucleate at the crystal surface as ~1 nm clusters that then grow toward the bulk.","lead":"This paper uses atomic-resolution electron microscopy to watch BaTiO3 crystals lose their ferroelectric phase while being heated. It reports that the loss starts with tiny paraelectric regions that form at the crystal surface and then grow inward.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Projection averaging through the 50-nm crystal makes the claimed single-unit-cell cubic-nucleus classification unsupported; the paper supplies no simulation or detection-limit analysis showing a 0.8-nm surface nucleus is visible.","rationale":"The reader's weakest assumption identified the HRTEM projection problem, and I agree that it is the most load-bearing point. The paper's headline contribution is an atomic-scale observation of nucleation and growth: paraelectric cubic nuclei of one to two unit cells appearing near the surface and growing inward. If those unit-cell classifications are not valid, the entire narrative falls, regardless of any thermodynamic calculation. The text itself admits the fundamental difficulty: at 383 K, when roughly half the cells are cubic, the authors say clear distinction between phases is hindered because TEM averages atomic locations over the full crystal thickness. A 0.8-nm surface nucleus in a 50-nm projection is a far more extreme version of the same averaging problem, and no evidence is presented that its signal survives. A multislice simulation with realistic noise would settle this directly. In addition, the quantitative claims are internally inconsistent: the nucleation barrier computed from the stated inputs in Eq. (2) is 8.6 eV rather than 2.13 eV; the SI strain energy of 1.67 eV is not obtainable from S = ½VEε² with the given values and contradicts the 4.0 eV cited in the main text; and the depolarization term uses a spontaneous polarization value that is orders of magnitude too small. These are objective, checkable errors that independently justify rejection of the paper as written. I therefore recommend keeping the reader's REJECT verdict, with the projection-averaging concern identified as the single most load-bearing issue.","tokens_in":10236,"tokens_out":8745,"duration_ms":95140,"concrete_test":"Run multislice HRTEM image simulations of a 50-nm-thick BaTiO3 slab under the stated conditions (300 keV, double-corrected, relevant defocus and dose), comparing a fully tetragonal crystal with the same crystal containing a single 0.8-nm cubic nucleus at the top, bottom, and side surfaces. Apply the same Ti-displacement extraction and per-column classification threshold used in the paper, including realistic noise, and determine whether the nucleus is detectable at the 22±5 pm level. If the simulated projected displacement maps cannot distinguish the nucleus, the surface-nucleation observation is unsupported; if they can, the qualitative claim survives and the main remaining obstacle is the incorrect energy analysis.","verdict_should_be":"REJECT","load_bearing_attack":"The central narrative requires classifying individual unit cells as tetragonal or cubic from a 22±5 pm Ti off-centering measured in HRTEM projections of a 50-nm-thick BaTiO3 crystal. A cubic nucleus of 'a couple of unit cells' (0.8 nm) at the surface contributes, at best, about 1–2% of the projected thickness in the affected columns, so the projected Ti-displacement signal changes by only a fraction of a picometer if the projection is a thickness average—far below the 5 pm spread quoted for the tetragonal displacement and far below any plausible per-column classification threshold. The authors do not provide multislice simulations, an exit-wave reconstruction, or a detection-threshold calibration for an isolated 0.8-nm nucleus embedded in 50 nm of material. Their own Figure 3f caption concedes that at ~50% cubic content the two phases can no longer be clearly distinguished because the TEM 'averages out atomic location over the entire crystal thickness,' yet no argument is given for why this same averaging does not destroy the signal at the 2–14% nucleation stage. If this unit-cell classification is invalid, the surface-nucleation and sidewalk-growth mechanism, the reported critical nucleus size a* = 0.8 nm, and all derived energies lose their factual basis. Separately, the quantitative analysis is internally inconsistent: Eq. (2) with a* = 0.8 nm and γ = 1.07 eV per unit-cell area gives 8.6 eV, not 2.13 eV; the SI strain energy of 1.67 eV does not follow from S = ½VEε² with the stated lattice parameters and modulus, and it contradicts the main-text value of 4.0 eV; and the SI uses P_s = 1.67×10⁻⁷ C/m², orders of magnitude below the known spontaneous polarization of BaTiO3. These arithmetic issues reinforce the rejection, but the projection problem is the more load-bearing concern because it threatens the qualitative observation itself.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports variable-temperature high-resolution transmission electron microscopy (HRTEM) observations of BaTiO3 nanocrystals (~50 nm) across the ferroelectric-to-paraelectric transition. The authors claim that the transition begins with the emergence of paraelectric cubic nuclei of roughly two unit cells (0.8 nm) near the crystal surface, followed by inward 'sidewalk' growth of the cubic phase. They further estimate a nucleation energy barrier of 2.13 eV and argue that mechanical strain dominates the barrier, comparing their findings to models of electric-field-driven domain nucleation. The manuscript presents atomic-scale maps of Ti off-centering displacements and proposes a nucleation-and-growth mechanism for the transition.","tokens_in":10560,"tokens_out":7226,"duration_ms":67744,"significance":"If substantiated, the observation of unit-cell-scale surface nucleation during a temperature-driven ferroic transition would be a significant advance, providing direct experimental evidence for the microscopic mechanism of a ferroelectric-to-paraelectric transition and connecting it to nucleation theory of domain switching. The qualitative images are suggestive and the experimental setup (in-situ heating of an intermediate-size crystal) is well chosen. However, the quantitative core of the paper is not reliable: the reported barrier and energy decompositions contain arithmetical errors and internal contradictions, and the claimed single-unit-cell classification is not supported by an analysis of projection averaging. The potential significance is high, but the manuscript in its present form does not support its central quantitative or qualitative claims.","major_comments":[{"comment":"The numerical evaluation of Eq. (2) is inconsistent with the stated parameters. With a critical nucleus size a* = 0.8 nm and an interfacial energy gamma = 1.07 eV per unit-cell area, the barrier is 2(a*)²gamma = 8.56 eV (equivalently, 8 unit-cell areas times 1.07 eV), not 2.13 eV as reported. The reported value corresponds to a* = 0.4 nm (one unit cell), which contradicts the observation of a nucleus 'of the size of a couple of unit cells'. This is a load-bearing quantitative error in the central claim of a 2.13 eV nucleation barrier.","section":"Discussions, Eq. (2); SI 'Nucleation barrier calculation'"},{"comment":"The reported strain energy is internally inconsistent. The main text states 4.0 eV, the SI states 1.67 eV, and direct evaluation of the SI formula S = (1/2)VEepsilon² with V = (0.8 nm)³, E = 40 GPa, and epsilon = (cT - aC)/cT ≈ 1.67e-3 gives S ≈ 1.8e-4 eV—three to four orders of magnitude smaller than either reported value. The SI value 1.67 eV would require epsilon ≈ 0.16, which is far outside the lattice mismatch. Consequently, the conclusion that mechanical strain dominates the nucleation barrier and is much larger than the electric contribution is not supported by the accompanying calculation.","section":"Discussions, 'mechanical and electric strain'; SI 'Nuclei strain energy calculation'"},{"comment":"The central experimental claim—classification of individual unit cells as tetragonal or cubic—rests on a 22±5 pm Ti off-centering displacement measured from HRTEM projections of a ~50-nm-thick crystal. A 0.8 nm cubic nucleus at the surface represents only a few percent of the projected column, so the projected displacement signal would shift by much less than 1 pm, far below the measurement spread. The authors themselves state in the Fig. 3f caption that at about 50% cubic content the two phases can no longer be clearly distinguished because of thickness averaging, yet they claim to detect cubic phase at 2–14% volume fraction. No multislice simulations, exit-wave reconstruction, or detection-limit analysis is provided to show that a 0.8 nm surface nucleus is visible within a 50-nm projection. Without such support, the unit-cell-level nucleation narrative, the critical nucleus size a*, and all derived energies lack a validated experimental basis.","section":"Results, Fig. 3 caption; Fig. 4; SI 'Ion-displacement mapping'"},{"comment":"The depolarization-energy calculation uses a spontaneous polarization Ps = 1.67e-7 C/m², which is about six orders of magnitude smaller than the known value for BaTiO3 (approximately 0.26 C/m²). With the given formula and parameters, the depolarization energy evaluates to about 1e-13 eV, not 0.06 eV as stated in the SI and main text. This invalidates the quantitative claim that the electric contribution is negligible and calls into question the reported decomposition of the nucleation barrier.","section":"SI, 'Nuclei depolarization energy calculation'"}],"minor_comments":[{"comment":"The text says 'a* equals the length of two-unit cells (80 pm)'. Since 0.8 nm = 800 pm, this appears to be a typographical error, but it should be corrected to avoid confusion about the reported critical nucleus size.","section":"Discussions, nucleus size statement"},{"comment":"Figure 4 reports the cubic-phase fraction as a function of temperature without any error bars or statement of the number of unit cells or images analyzed. Given the central role of these fractions in the nucleation-and-growth narrative, the authors should provide a statistical characterization.","section":"Results, Fig. 4"},{"comment":"The sidewalk growth is initially described as progressing along [011], but the authors then argue that [001] is more plausible. This ambiguity should be resolved rather than left open, especially as it relates to the strain-mismatch argument.","section":"Discussion, growth direction"},{"comment":"The sentence 'At the nucleation energy barrier, the change in free energy becomes nil' is imprecise; it is the derivative of the free energy with respect to nucleus size that vanishes at the critical nucleus, not the free-energy change itself. Please correct the wording.","section":"SI, 'Nucleation barrier calculation'"},{"comment":"The authors assume a linear temperature change from 373 K to 383 K over 50 s but provide no calibration or justification for this assumption. A brief note on the heating-cell calibration would strengthen the temporal assignment of the observed structural changes.","section":"Experimental, temperature ramp"}],"recommendation":"reject","confidential_remarks":"The manuscript presents an eye-catching observation, but the central experimental claim is currently unsupported by a quantitative detection-limit analysis, and the energy calculations contain arithmetical errors and internal contradictions that would change the main conclusions if corrected. In particular, the projection-averaging issue implies that additional simulations or independent experiments are needed to validate the unit-cell-scale classification, which goes beyond a routine revision. The paper is not suitable for publication in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version: this paper has a genuinely new observation—paraelectric nuclei appearing at the surface during the temperature-driven transition in BaTiO3—but the current manuscript doesn't support that observation at the level it claims. The unit-cell classification is vulnerable to projection averaging through 50 nm of crystal, and the energy analysis has arithmetic errors that break the quantitative conclusions.\n\nWhat's good: The idea of directly mapping dipole dynamics during a ferroic transition at atomic scale is a real step beyond static domain mapping or mesoscale PFM. The XRD confirms the expected tetragonal-to-cubic transition, and the large-scale images show a plausible sequence of surface nucleation and inward growth. The authors also explicitly note that at ~50% cubic content their technique can no longer distinguish phases, which is honest.\n\nThe soft spots are serious. First, the signal: a 0.8 nm nucleus near the surface contributes a few percent of the projected thickness in HRTEM. The Ti off-centering is 22±5 pm; the expected change from a small cubic region is a fraction of a picometer after thickness averaging. That's below their own noise. The authors provide no multislice simulations or detection-threshold calibration for this geometry. Their own admission that the method fails at 50% cubic because of thickness averaging raises the question why the same averaging doesn't wash out the signal at the 2–14% stage.\n\nSecond, the quantitative core is internally inconsistent. Using their stated a*=0.8 nm and γ=1.07 eV per unit-cell area, Eq. (2) gives 8.56 eV, not 2.13. The SI gives 1.67 eV for strain energy, the main text says 4.0 eV, and neither follows from S=½VEε² with their stated parameters. The depolarization term uses P_s = 1.67×10⁻⁷ C/m², orders of magnitude below known BaTiO3 values. These aren't typos; they're load-bearing for the claim that mechanical interactions dominate.\n\nWho should read this? The ferroelectrics community, mostly as a cautionary example of what in-situ TEM can and cannot resolve. If the projection issue is resolved with simulations and the energy numbers are redone, the qualitative surface-nucleation picture could be an important result. As written, the paper deserves a serious referee but needs major revision before publication. My vote: send to review, but with a strong request for multislice modeling and corrected calculations.","headline":"Genuinely new observation, but the projection-averaging problem undermines the unit-cell classification and the energy numbers do not add up.","tokens_in":11163,"tokens_out":5088,"would_cite":false,"duration_ms":49873,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The ferroelectric-to-paraelectric transition in BaTiO3 starts with ~0.8-nm cubic nuclei at the crystal surface and then grows inward.","keywords":["ferroelectric","BaTiO3","paraelectric phase transition","nucleation and growth","high-resolution transmission electron microscopy","dipole-moment mapping","Curie temperature","surface nucleation"],"falsifier":"Cool the same 50-nm BaTiO3 particle back through the transition while imaging: a true thermodynamic surface-nucleation mechanism should produce cubic islands at the surface on heating and make them shrink from the bulk side on cooling at the same temperatures, whereas a beam-induced or projection artifact would not show symmetric reversibility and would depend on electron dose or sample tilt. Alternatively, compare the per-unit-cell HRTEM classification with 4D-STEM nanodiffraction maps on the same area to see whether the 'cubic' unit cells are genuinely cubic rather than a projection artifact.","tokens_in":10028,"feed_emoji":"🔬","tokens_out":8487,"duration_ms":79658,"temperature":0.7,"pith_summary":"This paper establishes that the ferroelectric-to-paraelectric transition in 50-nm BaTiO3 crystals is a surface-initiated nucleation-and-growth process. Variable-temperature high-resolution electron microscopy shows that the crystal is fully tetragonal below the transition, and the first paraelectric cubic unit cells appear near the surface as nuclei only about two unit cells (0.8 nm) across. With further heating these nuclei merge and the cubic phase advances into the bulk as a sidewalk. From the observed critical nucleus size, the authors extract a nucleation barrier of 2.13 eV and show that mechanical strain, not depolarization energy, dominates that barrier. The result matters because it puts the thermal phase transition in the same quantitative nucleation language used for electric-field-driven domain switching.","feed_headline":"BaTiO3's transition to paraelectric begins with 0.8 nm surface nuclei","feed_subtitle":"High-resolution electron microscopy shows cubic islands nucleating at the surface, then growing inward like a sidewalk.","key_machinery":"The carrying object is the critical cubic nucleus of paraelectric BaTiO3 embedded in the tetragonal ferroelectric host. The argument combines atomic-scale classification of unit cells from Ti off-centering displacements with the classical nucleation balance $\\Delta G = -a^3 \\Delta G_V + 6a^2\\gamma$, whose maximum yields $\\Delta G^* = 2(a^*)^2\\gamma$ at the critical size $a^*$. To assign the barrier's origin, the paper adds an elastic strain energy $S = \\tfrac{1}{2}VE\\epsilon^2$ for a 0.8-nm cubic box in the tetragonal matrix and a standard electrostatic depolarization estimate, showing that the mechanical term dominates. These pieces together translate the visual observation of surface nuclei into a quantitative barrier.","core_discovery":"The central claim is that the tetragonal-to-cubic transition in BaTiO3 proceeds by inhomogeneous surface nucleation of the paraelectric phase, followed by inward growth. The evidence is atomic-scale mapping of the Ti off-centering displacement (22 ± 5 pm in the tetragonal phase), which lets the authors label individual unit cells as tetragonal or cubic in HRTEM images taken while the temperature is slowly raised from 373 K to 383 K. First cubic nuclei of about 1 nm appear at 375 K near the surface; more appear by 377 K; by 379–383 K they merge and a cubic sidewalk grows inward, converting about 40% of the material by 383 K. Using classical nucleation theory with a critical nucleus size of two unit cells and a surface energy of 1.07 eV per unit-cell area, the paper obtains $\\Delta G^* = 2.13$ eV for the nucleation barrier, and an elastic strain energy estimate (about 1.7–4 eV) far exceeding the 0.06 eV depolarization term. The authors therefore conclude that mechanical interactions dominate the nucleation barrier.","pith_inferences":["If surface nucleation is generic, then ferroelectric films and nanoparticles should consistently lose polarization from surfaces and interfaces first; this could be tested by comparing faceted particles with different exposed facets in the same in-situ TEM experiment.","Reducing surface defect density or changing surface termination should suppress nucleation and raise the apparent Curie temperature; a controlled comparison of as-prepared versus surface-passivated BaTiO3 particles would test this directly.","The single-unit-cell classification from ~22 pm displacements could be validated against a thickness-independent probe such as 4D-STEM nanodiffraction on the same particle; agreement would strengthen the method, while disagreement would point to projection artifacts.","The model implies a size limit: once the crystal is small enough that two unit cells are a significant fraction of the particle, the distinction between nucleation and growth should blur, connecting the observed mechanism to the known size dependence of the ferroelectric transition."],"forward_implications":["The tetragonal–cubic transition in intermediate-size BaTiO3 single crystals is a nucleation-and-growth process with a well-defined critical nucleus, so phase coexistence over a roughly 10 K window is an expected part of the transition rather than a sample artifact.","The critical nucleus is only about two unit cells, close to theoretical predictions for electric-field-driven domain-wall nucleation, suggesting that thermal and field-driven ferroelectric nucleation may share a common underlying mechanism.","Because nuclei appear at the surface, surface termination, roughness, and defects should control where and when the transition starts in nanoscale ferroelectrics.","Mechanical strain dominates the nucleation barrier (elastic terms of roughly 1.7–4 eV versus 0.06 eV depolarization), so strain engineering should be able to raise or lower the effective transition temperature by changing the cost of forming a cubic nucleus.","The cubic sidewalk grows preferentially along the direction that best matches the long tetragonal lattice parameter, giving the paraelectric growth a crystallographic direction that minimizes misfit energy."],"supporting_citations":[{"why":"Supplies the size-dependent phase diagram with a tetragonal–cubic coexistence regime that motivates the choice of 50-nm crystals.","marker":"[27]"},{"why":"Shows very fine BaTiO3 nanoparticles suppress the ferroic transition, establishing why an intermediate size is needed.","marker":"[28]"},{"why":"Provides the prior mesoscopic-scale observation of a BaTiO3 phase transition with slow temperature ramping, the approach this study extends to atomic scale.","marker":"[34]"},{"why":"Documents the high defect concentration and surface structure of BaTiO3 used to explain why nucleation occurs at the surface.","marker":"[35]"},{"why":"Provides the domain-nucleation model and predicted ~2 eV activation energy that the measured 2.13 eV barrier is compared with.","marker":"[40]"},{"why":"Supplies the electrostatic depolarization model used to estimate the small 0.06 eV electric contribution.","marker":"[41]"},{"why":"Supplies the BaTiO3 surface energy of 1.07 eV per unit cell used to compute the nucleation barrier.","marker":"[42]"},{"why":"Supplies the Young's modulus value of 40 GPa for 50-nm BaTiO3 used in the elastic strain energy estimate.","marker":"[43]"}],"fun_headline_variants":["Surface nuclei trigger BaTiO3's paraelectric transition","Atomic mapping: BaTiO3 cubic phase nucleates at surface","Mechanical strain sets BaTiO3 nucleation barrier","BaTiO3 transition starts with ~1 nm surface islands"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole narrative depends on the assumption that a ~22 pm Ti off-centering displacement extracted from TEM images averaged through a 50-nm-thick crystal can correctly classify each unit cell as tetragonal or cubic; if that classification fails, the observed surface nuclei and sidewalk are not actually distinct phase regions.","fun_headline_variants_meta":{"raw":{"variants":["Surface nuclei trigger BaTiO3's paraelectric transition","Atomic mapping: BaTiO3 cubic phase nucleates at surface","Mechanical strain sets BaTiO3 nucleation barrier","BaTiO3 transition starts with ~1 nm surface islands"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000502,"raw_usage":{"total_tokens":2454,"prompt_tokens":946,"completion_tokens":1508,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":562,"completion_tokens_details":{"reasoning_tokens":1443}},"tokens_in":562,"tokens_out":1508,"duration_ms":12500,"temperature":1.0,"reasoning_tokens":1443,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:21:41.407966+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Cool the same 50-nm BaTiO3 particle back through the transition while imaging: a true thermodynamic surface-nucleation mechanism should produce cubic islands at the surface on heating and make them shrink from the bulk side on cooling at the same temperatures, whereas a beam-induced or projection artifact would not show symmetric reversibility and would depend on electron dose or sample tilt. Alternatively, compare the per-unit-cell HRTEM classification with 4D-STEM nanodiffraction maps on the same area to see whether the 'cubic' unit cells are genuinely cubic rather than a projection artifact.","supporting_citations":[],"review_version":1}