{"id":"48fdc0a2-49b3-4614-b5ec-0bf9aed1e998","arxiv_id":"2607.16167","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Reactive molecular dynamics identifies a fluorine-assisted, chromium-depletion-driven grain boundary migration mechanism in NiCr in molten FLiNaK, suppressed at the coherent Σ3 twin boundary.","lead":"Simulations of nickel-chromium surfaces in molten fluoride salt show that open grain boundaries concentrate fluorine, shed chromium fastest, and then creep into the intact alloy, while a tight twin boundary stays put. The result gives a concrete target for grain-boundary engineering of alloys that must survive molten-salt reactors.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Apparent GB migration may be an erosion/reclassification artifact: CNA-based boundary tracking cannot distinguish true migration from preferential corrosion at Σ5 boundaries.","rationale":"The reader's weakest assumption focuses on force-field transferability, with the CNA migration-vs-erosion issue listed as secondary. I agree the force-field concern is real, but I find the more immediately load-bearing issue to be the internal validity of the migration metric itself. If the GB 'shift' is an erosion/reclassification artifact, the central DIGM claim fails regardless of force-field accuracy. The paper provides no marker-based or lattice-registration validation that the two grains truly displace relative to each other, and the qualitative FCC wake argument is insufficient because preferential Cr removal from a stationary GB would also leave a Ni-enriched FCC-like wake. This is not a rejection—the simulations may well show genuine DIGM—but it requires a concrete discriminator before the mechanistic claim can be accepted. The reader's conditional verdict already captures this need; my concern sharpens it and suggests a specific test. Hence unchanged verdict.","tokens_in":12387,"tokens_out":3632,"duration_ms":37019,"concrete_test":"Perform lattice-registration analysis on the Σ5(012) and Σ5(013) trajectories: for each frame, fit the ideal FCC lattice orientation of each grain (e.g., using OVITO's lattice analysis or a local orientational order parameter) and compute the relative displacement of the two grains across the GB. Compare this registered relative grain displacement to the CNA-based GB position shift reported in Fig. 6a. If the CNA shift exceeds the registered shift by more than ~1 Å, the apparent migration is dominated by erosion/reclassification, not true boundary motion. Alternatively, embed a few weakly interacting marker atoms (e.g., W) in the GB core and track their positions; stationary markers with a shifting CNA boundary would confirm the artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central DIGM claim rests on reported GB shifts of ~3.56 Å (Σ5(012)) and ~3.89 Å (Σ5(013)) versus 0.45–1.60 Å in no-salt controls (Fig. 6a, Table S2). However, the boundary position is computed as the mean coordinate of atoms classified as non-FCC by common neighbor analysis (Sec. 2.2), with only the top/bottom 5% of the cell excluded. Under salt exposure, the same Σ5 boundaries exhibit strong GB-localized dissolution (Fig. 2c,d), F penetration (Fig. 5c,d), and pit-like recession (Fig. 3). The CNA metric will shift if boundary atoms are preferentially dissolved, if adjacent atoms become disordered by vacancy accumulation, or if the GB plane roughens—even if neither grain actually grows relative to the other. The paper's only discriminator is the qualitative claim that the wake remains FCC and Ni-enriched (Fig. 6b,c), but preferential Cr dissolution from a stationary disordered GB core would also produce a Cr-depleted, Ni-enriched, FCC wake. Without an orientation-independent marker of the two grain lattices, the measured 'migration' is confounded with localized etching. Additionally, Fig. 6a reports no error bars or statistical significance; with 8 seeds, the salt-vs-no-salt difference must be shown to exceed seed-to-seed variation. If the apparent shifts are largely erosion/reclassification artifacts, the proposed mechanism—fluorine localization driving Cr dissolution, vacancy supersaturation, and genuine boundary motion—is not uniquely established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses reactive molecular dynamics (ReaxFF) with the FLiNaK–NiCr force field to simulate corrosion of Ni75Cr25 bicrystals containing four grain boundaries (Σ3(111), Σ11(113), Σ5(012), Σ5(013)) and compares them to bulk (110) and (100) surfaces with and without molten FLiNaK. The central claims are: (i) corrosion is initially controlled by surface orientation, while GB character controls spatial localization at longer times; (ii) Σ5 boundaries localize F adsorption, Cr dissolution, and vacancy accumulation, leading to pit-like recession; and (iii) this localized dealloying drives corrosion-assisted GB migration, interpreted as dealloying-driven diffusion-induced grain boundary migration (DIGM), whereas the coherent Σ3 twin suppresses this coupling. The authors propose low-energy special boundaries as a GB-engineering strategy for corrosion resistance.","tokens_in":12595,"tokens_out":2649,"duration_ms":26167,"significance":"If the migration claim is correct, the paper provides a concrete atomistic mechanism connecting GB character to molten-salt corrosion and DIGM, with direct implications for GB engineering of Ni–Cr alloys. The simulation design is careful: it includes no-salt controls, 8 independent seeds per system, two surface orientations to deconfound surface and GB effects, and a 1 ns extension that supports the delayed GB-character crossover. The spatial correlation between F coverage, Cr dissolution, and GB-localized damage is a useful contribution regardless of the migration interpretation. However, the paper's central quantitative claim—enhanced GB migration—rests on a boundary-tracking metric that may conflate true migration with preferential corrosion/reclassification at the GB, and the force field itself is developed by the same group and encodes the chemistry that the paper then reports as discovered. Both issues need to be addressed before the mechanism can be considered established.","major_comments":[{"comment":"The GB position is defined as the mean coordinate of atoms identified as non-FCC (or HCP for Σ3) by CNA, with only the top/bottom 5% of the slab excluded. This metric cannot distinguish true GB migration from preferential dissolution and structural disordering at the boundary. The Σ5 systems show GB-centered dissolution (Fig. 2c,d), F penetration (Fig. 5c,d), vacancy accumulation, and pit-like recession (Fig. 3). If boundary atoms are preferentially removed or neighboring lattice atoms become disordered by vacancy accumulation, the CNA centroid will shift even if neither grain grows relative to the other. The authors argue that the wake remains FCC and Ni-enriched (Fig. 6b,c), but preferential Cr dissolution from a stationary disordered GB core would leave exactly the same signature. To support the DIGM claim, the authors must use an orientation-independent marker of the two grain lattic","section":"§2.2, Fig. 6"},{"comment":"Figure 6a reports mean GB displacements without error bars or statistical significance. With 8 seeds per system, the reported salt-induced shifts (~3.56 Å for Σ5(012), ~3.89 Å for Σ5(013)) must be shown to exceed seed-to-seed variation, and compared against the no-salt controls (0.45–1.60 Å) with a proper statistical test. Currently, the reader cannot assess whether the 'substantial enhancement' is robust or whether the apparent differences are within thermal/statistical noise. Please report per-seed data, confidence intervals, and an appropriate significance test for each GB and condition.","section":"Fig. 6a, Table S2"},{"comment":"The mechanism's pre-conditions—that F binds Cr preferentially, that Σ5 GB sites bind F more strongly, and that Cr dissolution from GB sites has a lower barrier—are largely encoded in the ReaxFF force field developed by members of this team (ref [10]) and in the same group's DFT study (ref [23]). The paper presents no benchmark against independent, non-training data for GB vacancy mobility, Cr–F binding at GB sites, or GB migration kinetics. This is a correctness-risk concern: if the potential mis-weights Cr–F vs Ni–F interactions near GB seams, the entire chain from F localization to vacancy supersaturation to migration could be a force-field artifact. I am not asking for a new force field, but the authors should provide explicit validation or sensitivity tests (e.g., against DFT for representative GB dissolution barriers, or comparisons with measured GB migration rates) and clearly stat","section":"§2.1/§3.3"}],"minor_comments":[{"comment":"The caption should state explicitly that no error bars are shown and refer to the per-seed data in Table S2.","section":"Fig. 6a caption"},{"comment":"The dissolution criterion (metal-neighbor count below three) and F-adsorption criterion (bonded to two or more non-dissolved metals) are presented as thresholds but their sensitivity is not discussed. Please add a brief sensitivity statement or cite a prior calibration.","section":"§2.2"},{"comment":"The phrase 'vacancy supersaturation' is used to explain the driving force, but no direct vacancy concentration profile is reported. A quantitative vacancy-density profile near the GB would strengthen the mechanistic claim.","section":"§3.3"},{"comment":"Several references are to the same author team's prior work (refs [9,10,23,53]) and are used to justify both the method and the interpretation. It would be helpful to explicitly distinguish validated, literature-supported facts from assumptions inherited from those studies.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper has a solid computational design and the corrosion-localization results are likely robust. The main issue is the interpretation of the GB-displacement measurement: the CNA centroid is too easily confounded with preferential etching at the boundary. This can likely be fixed by re-analyzing the trajectories with lattice-orientation-based grain tracking and by providing proper statistics. The force-field provenance is a second, more fundamental concern, but it is a standard limitation of ReaxFF studies and can be addressed with targeted validation. I recommend major revision rather than rejection, because the central claim is defensible in principle but currently not uniquely established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the two-regime corrosion picture holds up: surface orientation drives early dissolution, and GB character controls where corrosion localizes later. The F-coverage and dissolution profiles for Σ5 versus Σ3 are clean, the controls are sensible, and the 1 ns extension showing delayed divergence from bulk is a nice touch. If you take only the localization story, this is a useful paper.\n\nBut the headline claim—corrosion-driven GB migration—is not yet established. The boundary position is defined as the mean coordinate of CNA non-FCC/HCP atoms. Under salt exposure, the same Σ5 boundaries are dissolving, roughening, and pitting. If GB-core atoms are preferentially removed, or if vacancy accumulation disorders neighboring lattice, the mean coordinate of the non-FCC set shifts even if neither grain actually consumes the other. The Ni-enriched, FCC wake is exactly what you'd expect from selective Cr etching of a stationary boundary, so it doesn't discriminate. There are no marker planes or lattice registration to show that one grain has grown at the expense of the other, and Fig 6a shows no error bars despite 8 seeds. As it stands, the migration numbers could be dissolution artifacts.\n\nThe mechanism itself—F localization, preferential Cr removal, vacancy supersaturation, boundary motion—is plausible and consistent with the experimental DIGM literature. But the evidence is correlational, and the interpretation leans on the same group's ReaxFF and DFT. The force field was fitted by these authors, and the supporting DFT premise comes from their own prior work. That's not disqualifying by itself, but there's no independent check of the potential in GB environments, and no perturbation runs to isolate the vacancy-supersaturation driving force.\n\nSo the central claim is under-supported, not disproven. I'd send it to peer review, but only with the expectation of major revision: the authors need a better boundary-tracking method, uncertainty quantification, and ideally an independent force-field benchmark. The localization results deserve publication; the migration claim needs more work. I wouldn't cite it for migration in its current form.","headline":"The localization story is solid, but the migration claim is confounded by the CNA tracking metric and needs much stronger evidence before it can be believed.","tokens_in":13232,"tokens_out":3347,"would_cite":false,"duration_ms":31303,"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":"Simulations show molten salt drives NiCr grain boundaries to migrate, while coherent twin boundaries stay put.","keywords":["molten salt corrosion","NiCr alloys","FLiNaK","grain boundary migration","diffusion-induced grain boundary migration","reactive molecular dynamics","dealloying","grain boundary engineering"],"falsifier":"An independent first-principles calculation of fluorine and chromium binding energies at the Σ5(012) and Σ5(013) boundary sites that disagrees with the force-field predictions, or a bicrystal experiment in which a high-angle boundary in Ni–20Cr immersed in molten FLiNaK at 800 °C does not migrate more than a twin boundary while chromium is depleted, would settle whether the claimed mechanism is real.","tokens_in":12112,"feed_emoji":"🧂","tokens_out":6286,"duration_ms":48974,"temperature":0.7,"pith_summary":"The paper sets out to show that grain boundaries in nickel–chromium alloys are not passive corrosion pathways in molten FLiNaK salt but dynamic interfaces that move as the salt attacks them. Using reactive molecular dynamics, it compares four types of boundaries and finds that open high-energy boundaries such as Σ5 concentrate fluorine adsorption and selectively dissolve chromium, leaving a vacancy-rich, Cr-depleted zone that pushes the boundary into the intact grain. With salt present, the Σ5 boundaries shift roughly 3.6–3.9 Å over 500 ps, versus 0.45–1.60 Å without salt. The coherent Σ3 twin, by contrast, stays nearly fixed. If correct, the result connects atomistic corrosion chemistry to the wavy, channel-like intergranular damage seen in experiments and points to twin-boundary engineering as a corrosion-resistance strategy.","feed_headline":"Molten salt drives NiCr grain boundaries to migrate; twins resist","feed_subtitle":"Fluorine strips chromium at open boundaries, driving them into the alloy while coherent twins stay put.","key_machinery":"The load-bearing machinery is a dealloying-driven form of diffusion-induced grain boundary migration (DIGM), a named mechanism in which a moving grain boundary leaves behind a compositionally altered wake. Here the salt continuously removes chromium rather than adding solute, so the boundary advances away from the Cr-depleted, vacancy-rich region. The simulations are carried out with a reactive force field for FLiNaK on Ni–Cr that allows metal–fluorine bonds to form and break, and boundary positions are tracked with common-neighbor analysis, which identifies disordered atoms at high-angle boundaries and HCP atoms at the coherent twin. The combination lets the authors separate surface-orienta","core_discovery":"The central discovery is a corrosion-driven grain boundary migration mechanism that the authors describe as a dealloying-driven form of diffusion-induced grain boundary migration (DIGM). In the simulations, fluorine from FLiNaK penetrates the open structure of Σ5(012) and Σ5(013) boundaries, forms Cr–F bonds preferentially at boundary-adjacent sites, and weakens Cr–Ni metallic coordination, promoting selective chromium removal. The resulting chromium depletion and vacancy supersaturation along the boundary drive the boundary to migrate away from the corroded zone into the undisturbed material, leaving behind a Ni-enriched, Cr-depleted wake that retains the FCC lattice. This migration is grai","pith_inferences":["A testable extension: if the mechanism is correct, bicrystal corrosion experiments should reveal a Cr-depleted, Ni-enriched band just ahead of migrating high-angle boundaries at early times, before wormhole channels fully develop.","The same coupled dealloying–DIGM picture may apply to other molten halide salts and to other alloys with a selectively dissolved reactive component, since the ingredients—salt-driven selective removal, vacancy accumulation, and boundary openness—are generic.","The paper's two conditions suggest a screening metric: grain-boundary excess volume combined with fluorine binding energy at boundary sites could rank boundary types for corrosion resistance without running full reactive simulations.","Because surface orientation controls the initial dissolution rate, twin-boundary engineering may be partially masked in components exposing reactive (110) textures; surface crystallography and boundary character likely need to be optimized together."],"forward_implications":["High-energy Σ5 boundaries act as localized corrosion channels: they concentrate fluorine adsorption, chromium dissolution, and vacancy accumulation at the boundary plane, so pit-like damage develops at the boundary rather than uniformly across the surface.","Corrosion-assisted boundary migration leaves a Ni-enriched, Cr-depleted wake with the same FCC lattice, altering near-boundary composition and making the swept region less susceptible to further dissolution.","The coherent Σ3(111) twin suppresses both fluorine localization and boundary migration even when its exposed (110)-type surface is intrinsically reactive, making increased twin fraction a candidate grain-boundary-engineering strategy for corrosion resistance.","The planar grain-boundary migration captured here is a plausible early-stage precursor to the one-dimensional wormhole channels seen experimentally, implying that intergranular attack may become self-sustaining as the boundary moves while the salt continues to supply fluorine and drain chromium.","Two simultaneous conditions appear necessary for this degradation path: sufficient fluorine activity in the salt and a structurally open boundary; raising the fraction of low-energy special boundaries should suppress corrosion-assisted migration."],"fun_headline_variants":["Corrosion makes NiCr grain boundaries migrate; twins stay put","Fluorine attack moves NiCr grain boundaries into the alloy","Cr depletion drives boundary migration in molten FLiNaK","Molten salt corrosion induces grain boundary migration in NiCr","NiCr boundaries wander under fluorine; coherent twins resist"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The mechanism rests on the reactive force field for FLiNaK on Ni–Cr being accurate specifically in grain-boundary environments: if it mis-weights how strongly fluorine binds chromium versus nickel at boundary sites, the entire localization→dissolution→migration chain could be an artifact of the potential.","fun_headline_variants_meta":{"raw":{"variants":["Corrosion makes NiCr grain boundaries migrate; twins stay put","Fluorine attack moves NiCr grain boundaries into the alloy","Cr depletion drives boundary migration in molten FLiNaK","Molten salt corrosion induces grain boundary migration in NiCr","NiCr boundaries wander under fluorine; coherent twins resist"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000625,"raw_usage":{"total_tokens":2700,"prompt_tokens":684,"completion_tokens":2016,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":428,"completion_tokens_details":{"reasoning_tokens":1948}},"tokens_in":428,"tokens_out":2016,"duration_ms":15115,"temperature":1.0,"reasoning_tokens":1948,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T21:09:12.028484+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An independent first-principles calculation of fluorine and chromium binding energies at the Σ5(012) and Σ5(013) boundary sites that disagrees with the force-field predictions, or a bicrystal experiment in which a high-angle boundary in Ni–20Cr immersed in molten FLiNaK at 800 °C does not migrate more than a twin boundary while chromium is depleted, would settle whether the claimed mechanism is real.","supporting_citations":[],"review_version":1}