{"id":"4636fdee-74c4-4c67-a13f-00becb3d6fd3","arxiv_id":"2509.06323","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In hydrogen-charged 1.5 GPa dual-phase steel, once a crack outruns hydrogen, sustained brittle {100} ferrite cleavage occurs, explaining why ductile ferrite fails to arrest embrittlement.","lead":"A hydrogen-charged 1.5 GPa ferrite-martensite steel fractures in the elastic range at 900 MPa; soft ferrite patches blunt early cracks but cannot stop fast brittle growth. The authors explain this as hydrogen-induced fast fracture, where a rapidly moving crack outruns the hydrogen and keeps propagating because it is too fast for ductile blunting.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The HIFF claim rests on an unmeasured local crack velocity crossing the 5.4 m/s threshold; the observed fracture morphology could also arise from crack-length-dependent stress state.","rationale":"I read the paper in good faith. The experimental characterization is substantial: site-specific FIB lift-out, TEM and TKD, EBSD, fractography, and comparison with uncharged and U-bend tests. The authors are also appropriately cautious in noting that hydride formation has never been observed and that the river-pattern mechanism needs further validation. However, the load-bearing step of the central mechanistic claim is the existence of a crack velocity above Vcrit during stage-B propagation. This velocity is never measured; it is inferred from fracture morphology. The reader's weakest_assumption identifies exactly this point, and I agree with it. The morphological sequence is plausible but not unique: the same center-to-surface brittle-to-ductile transition is the classic plane-strain to plane-stress transition, and the authors use that explanation themselves for the shear lip. Their K-based argument establishes only that velocity should increase with crack length, not that the threshold is crossed in this specimen at the observed transition. The exclusion of stress partitioning is based on the low nominal failure stress, but local crack-tip stresses are not quantified. Because this condition is central and testable, the appropriate verdict remains CONDITIONAL. I see no basis for rejection; the data and microscopy are valuable, and the proposed velocity measurement would either support or refute the HIFF interpretation.","tokens_in":23321,"tokens_out":5567,"duration_ms":57203,"concrete_test":"Measure crack velocity directly during hydrogen-precharged slow-strain-rate tests (1e-4/s) using a high-bandwidth method, e.g., direct-current potential drop with at least 1 MHz sampling or high-speed imaging at 100,000+ fps with digital image correlation, and record local velocity versus crack length from surface initiation to failure. Compare the velocity at the location where the fracture surface transitions from ferrite blunting/mixed-mode to {100} ferrite cleavage with the 5.4 m/s threshold. If the measured velocity is below 5.4 m/s at that transition, the HIFF stage-B explanation is not supported and the center-region cleavage requires another mechanism; if it exceeds 5.4 m/s, the central claim is strengthened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central conclusion (Section 5; Sections 4.3-4.4) requires that, after hydrogen supply to the crack tip is depleted, the crack continues at v > Vcrit = 5.4 m/s (the BCC-Fe threshold taken from Shishvan et al. [52]) so that blunting by dislocation emission is suppressed and {100} ferrite cleavage persists. No direct measurement of crack velocity is reported. Velocity is inferred from the spatial sequence of fracture morphologies: ferrite blunting near initiation and {100} ferrite cleavage after significant crack growth. That inference is underdetermined. The same center-versus-surface transition is expected from the plane-strain/plane-stress change, an explanation the authors themselves invoke in Section 4.4 for the shear-lip region, and it could also reflect hydrogen distribution or pre-existing martensite damage. The K ∝ (crack length)^0.5 argument shows only a monotonic trend, not that the 5.4 m/s threshold is actually crossed at the crack length where cleavage appears. The authors exclude the stress-partitioning mechanism used for the uncharged case because nominal failure stress is low (900 MPa), but they do not quantify local crack-tip stress intensity at the transition; a long crack can produce high crack-tip driving force even at low nominal stress. If the local velocity never exceeds Vcrit, stage-B of HIFF cannot explain the sustained {100} ferrite cleavage, and the central claim is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental study of hydrogen embrittlement in a 1.5 GPa ferrite-martensite dual-phase (DP) steel. Hydrogen pre-charging to 3.8 mass ppm diffusible hydrogen, followed by slow strain-rate tensile testing at 10^-4 s^-1, produces fracture within the elastic regime at 900 MPa. Fractography, site-specific TEM/TKD, and EBSD reveal: surface crack initiation at prior austenite grain boundaries with intergranular and quasi-cleavage features; ferrite blunting and crack arrest near the initiation region; mixed-mode fracture ahead of initiation; and predominantly brittle {100} ferrite cleavage after significant crack growth. The authors interpret these observations using the hydrogen-induced fast fracture (HIFF) model of Shishvan et al., replacing the hydride sub-mechanism with hydrogen-enhanced decohesion at PAGBs, and propose that once the crack velocity exceeds Vcrit = 5.4 m/s, cleavage in ferrite persists even after hydrogen depletion at the crack tip. They also propose a mechanism for river-pattern formation via crack deflection along {110} micro-deformation bands and {121} twins.","tokens_in":23551,"tokens_out":7247,"duration_ms":61804,"significance":"If the central claim is correct, the work identifies a qualitatively important regime for hydrogen embrittlement in DP steels: ferrite's crack-arresting ability is effective only for short cracks, and long-crack propagation is governed by fast fracture rather than by direct hydrogen-assisted crack-tip processes. The experimental dataset is rich and internally consistent; the HIFF interpretation is applied without fitting parameters, and the proposed river-pattern mechanism is falsifiable. However, the load-bearing assertion that the local crack velocity actually exceeds the literature value of Vcrit is not directly measured, so the significance depends on whether the authors can strengthen this inference with quantitative analysis or a direct velocity measurement.","major_comments":[{"comment":"The central stage-B claim is supported only by an unmeasured crack velocity. The authors infer v > Vcrit = 5.4 m/s from the monotonic relation K ∝ (crack length)^0.5, but this establishes only that K and hence a characteristic crack speed increase with crack length; it does not establish that the local crack speed exceeds the BCC-Fe threshold at the crack length where {100} ferrite cleavage first appears (Figs. 5 and 6). The observed transition from mixed-mode to predominantly brittle fracture is also consistent with an increasing crack-tip driving force and constraint with crack growth, which would occur with or without hydrogen assistance. To make stage-B load-bearing, the authors should either measure the crack speed (e.g., by potential drop, acoustic emission, or high-speed imaging) or provide a quantitative fracture-mechanics estimate of the local stress intensity at the transition crack length and show that it is sufficient to drive ferrite cleavage under the observed 900 MPa nominal stress. Without this, the sustained {100} cleavage after hydrogen depletion could be caused by the crack-length-dependent stress state rather than by v > Vcrit.","section":"Section 4.4"},{"comment":"The exclusion of the stress-partitioning mechanism is not quantitative. The authors argue that the high strain rate at the crack tip cannot be due to stress partitioning because the hydrogen-charged specimen fails at a nominal stress of 900 MPa, whereas the uncharged specimen's cleavage was attributed to partitioning with martensite stresses of ~1700 MPa. However, crack-tip fields scale with the stress intensity factor, not with the nominal stress alone; a long crack at 900 MPa can produce a larger K than a short crack at 1700 MPa, and the authors themselves use this K scaling for the velocity argument. The authors should estimate the local K at the observed mixed-to-cleavage transition and show that the resulting crack-tip strain rates are below those required for cleavage unless HIFF is invoked. In addition, the high dislocation density observed beneath the {100} cleavage (Section 3.4, Fig. 9a) indicates substantial plasticity, which the HIFF model must accommodate; the paper should specify how much dislocation emission is compatible with v > Vcrit in the Shishvan framework.","section":"Section 4.4"},{"comment":"The proposed stage-A hydrogen supply mechanism is in tension with the paper's own diffusion argument. Section 4.2 states that at brittle crack speeds (~1000 m/s) hydrogen cannot diffuse ahead of the crack tip over relevant distances, and Section 4.3 replaces Shishvan's hydride-assisted growth with hydrogen diffusion to the PAGB crack tip to sustain v > Vcrit. The authors should clarify whether the hydrogen that sustains stage A is pre-existing at the PAGB (accumulated during the loading phase) or must arrive by diffusion during crack growth. If it must arrive by diffusion, the paper should quantify the hydrogen-enriched zone size and show that it can sustain the intergranular crack path at high velocity; if it is pre-existing, the length of the PAGB segment over which stage A operates should be specified so that the transition to stage B can be identified. Currently this transition is not tied to any measured microstructural length.","section":"Section 4.3"}],"minor_comments":[{"comment":"The uncharged specimen is tested only at 10^-3 s^-1, so the charged 10^-4 s^-1 test is compared without a corresponding uncharged baseline at the same strain rate. Consider adding an uncharged 10^-4 s^-1 test or discussing this limitation explicitly.","section":"Section 3.2 and Fig. 2"},{"comment":"There is a typo in the caption: 'Prescence' should be 'Presence'.","section":"Fig. S2 caption"},{"comment":"The prime notation for subfigures (e.g., Fig. 8f' and Fig. 9f') is nonstandard; the authors should define the prime notation in the captions or use separate labels to avoid ambiguity.","section":"Figures 8 and 9"},{"comment":"The table has empty cells for crack initiation in some columns; consider filling them with 'not observed' or a dash to avoid implying missing data.","section":"Table 2"},{"comment":"The paper does not report the hydrogen content after the tensile test; since slow strain-rate tests take time, some hydrogen may desorb during loading. State whether the 3.8 mass ppm is the pre-test content and discuss possible loss.","section":"Section 3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript builds on the authors' previous work [5,33], and the self-citations are appropriate given the continuity of the study. The main concern is the unmeasured crack velocity underpinning the central HIFF claim; a quantitative K estimate at the transition crack length or a direct velocity measurement would make the argument load-bearing rather than interpretive. If the authors can supply this, the paper would be a strong contribution to the hydrogen embrittlement literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a careful experimental paper with genuinely new observations. The authors show, with site-specific TEM and TKD, that ferrite arrests hydrogen-assisted cracks at initiation in a 1.5 GPa ferrite-martensite DP steel, but after significant crack growth the same ferrite undergoes {100} cleavage. That observation chain—fractography, EBSD, TKD, and TEM—is mutually consistent and well documented. It is the first such characterization for this steel under tensile loading, and it gives the community a clear microstructural sequence to explain.\n\nThe paper does a lot right. The comparison to their own prior work on the uncharged steel and on U-bend delayed fracture is useful and honest. They also flag exactly where the Shishvan model has a weak point (the unobserved hydride sub-mechanism) and propose a reasonable HEDE-based alternative for stage A. The river-pattern discussion is openly speculative and marked as needing MD validation. That is the right amount of caution.\n\nThe soft spot is the central load-bearing claim: the crack velocity exceeds the 5.4 m/s critical threshold from Shishvan et al. The velocity is never measured. The K ∝ (crack length)^0.5 argument only shows a monotonic increase, not that the threshold is crossed where cleavage appears. The stress-test note makes a fair point: the center-versus-surface transition from brittle to ductile could also be the plane-strain to plane-stress transition. The authors invoke that explanation for the shear-lipped region, but do not rule it out for the central cleavage. They exclude stress partitioning from the uncharged case because nominal failure stress is low (900 MPa), but a long crack can develop high local crack-tip driving force even at low nominal stress. If the local velocity never exceeds Vcrit, stage B of the HIFF model cannot explain the sustained ferrite cleavage, and the mechanism loses its footing.\n\nMinor points: the tensile data appear to come from single tests with no error bars, and the hydrogen content is reported from one TDS measurement. These are small issues, not central.\n\nWho this is for: people working on hydrogen embrittlement of advanced high-strength steels, especially those interested in ferrite's crack-arresting role and in fast-fracture mechanisms. The paper deserves a serious referee. I would send it to review, but request direct velocity evidence—acoustic emission, striation spacing, or at least a quantitative local-K estimate showing consistency with the 5.4 m/s threshold—and a direct discussion of the stress-state alternative.","headline":"First site-specific TEM/TKD evidence of ferrite blunting then {100} cleavage in a 1.5 GPa DP steel, wrapped in a plausible but unmeasured HIFF story.","tokens_in":24155,"tokens_out":1856,"would_cite":true,"duration_ms":19188,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Hydrogen embrittlement in a 1.5 GPa dual-phase steel is governed by a crack-velocity threshold, not by hydrogen acting at the crack tip.","keywords":["hydrogen embrittlement","dual-phase steel","hydrogen-induced fast fracture","ferrite cleavage","prior austenite grain boundary","crack blunting","river pattern","hydrogen-enhanced decohesion"],"falsifier":"Directly measure the running crack velocity during the stage where ferrite cleaves on {100}, for example by high-speed imaging, electrical potential drop, or acoustic emission, in a hydrogen-precharged tensile specimen. If the crack is moving slower than about 5.4 m/s while {100} cleavage continues after hydrogen depletion, the central claim fails; the same test should also confirm that hydrogen is actually depleted at the crack tip at that moment.","tokens_in":23088,"feed_emoji":"⚙️","tokens_out":5829,"duration_ms":49111,"temperature":0.7,"pith_summary":"This paper tries to establish that, in a hydrogen-precharged 1.5 GPa ferrite-martensite dual-phase steel, hydrogen embrittlement is driven by a hydrogen-induced fast-fracture mechanism rather than by hydrogen directly weakening the advancing crack tip. Precharged specimens failed within the elastic regime at 900 MPa after slow-strain tensile testing, and site-specific electron microscopy showed that ferrite arrests short cracks but later fractures by {100} cleavage. The authors explain the sustained brittle propagation with a critical crack-velocity threshold: once the crack outruns about 5.4 metres per second for body-centred cubic iron, dislocation emission can no longer blunt the tip, so brittle cleavage continues even after local hydrogen is depleted. If true, this means ductile ferrite cannot reliably stop hydrogen embrittlement once fast fracture takes over, and fractographic estimates of brittle area fraction by themselves can mislead.","feed_headline":"Crack speed, not hydrogen, drives brittle failure in 1.5 GPa steel","feed_subtitle":"Hydrogen pre-charged steel fails in the elastic regime; ductile ferrite turns brittle once crack speed passes a threshold.","key_machinery":"The central mechanism is the hydrogen-induced fast-fracture (HIFF) model, applied in two stages. In stage A, hydrogen diffuses to prior austenite grain boundaries, weakens them by hydrogen-enhanced decohesion, and sustains fast brittle crack growth in martensite through continued hydrogen supply to the crack tip. In stage B, once the crack velocity exceeds the critical threshold Vcrit ≈ 5.4 m/s for BCC iron, the high crack-tip strain rate prevents effective blunting by dislocation emission, so {100} ferrite cleavage persists even after hydrogen at the tip is exhausted. The paper modifies the original HIFF picture by replacing hydride-assisted crack initiation with hydrogen-enhanced decohesion at grain boundaries, and it uses the velocity threshold to explain why ferrite arrests short cracks but cleaves after long crack growth.","core_discovery":"The paper reports that hydrogen pre-charging to 3.8 mass ppm diffusible hydrogen makes a 1.5 GPa dual-phase steel fracture at 900 MPa with only 0.57% elongation, well inside the elastic regime. The fracture path begins with intergranular cracking along prior austenite grain boundaries, attributed to hydrogen-enhanced decohesion assisted by hydrogen diffusion during loading. Ferrite blunts sub-surface cracks during initiation and early growth, but after significant crack propagation the same ferrite undergoes {100} cleavage, as identified by site-specific transmission electron microscopy and transmission Kikuchi diffraction. The authors propose that hydrogen-assisted crack growth in martensite accelerates the crack beyond a critical velocity, Vcrit ≈ 5.4 m/s for BCC iron; above this speed, crack-tip blunting by dislocation emission is suppressed, so brittle {100} ferrite cleavage can continue even after the hydrogen supply to the crack tip is depleted. The river-pattern serrations on the {100} surface are explained by crack deflection along {110} micro deformation bands and {121} twins, with the connecting surfaces being non-crystallographic coalescence features.","pith_inferences":["If the velocity threshold is the real controlling parameter, then the same ferrite grain can act as an arrestor early in fracture and as a cleavage path later; the crack length at which ferrite switches roles should be predictable from applied stress and the threshold, which could be mapped experimentally.","Direct crack-speed measurement in similar pre-charged high-strength steels would test whether the inferred threshold is real; observing {100} cleavage below roughly 5.4 m/s would force an alternative explanation.","The HIFF picture implies that trapping strategies that only remove diffusible hydrogen may not stop fast fracture once the crack is long; what matters is preventing the crack from ever reaching the critical velocity.","Because the model replaces hydride-assisted initiation with hydrogen-enhanced decohesion at prior austenite grain boundaries, steels with engineered grain-boundary segregation would be predicted to raise the stress needed to cross the threshold; this is testable."],"forward_implications":["A ductile ferrite phase does not guarantee hydrogen resistance: ferrite arrests cracks only while the crack is short and its velocity is below the critical threshold.","In a hydrogen-precharged specimen, brittle fracture can begin within the elastic regime at 900 MPa, so design margins for high-strength steels must account for fast-fracture initiation at prior austenite grain boundaries.","Strain-rate sensitivity of hydrogen embrittlement in this steel can be understood through hydrogen diffusion kinetics to grain boundaries rather than through hydride formation at cavities.","The {100} ferrite cleavage surface forms river-pattern serrations by crack deflection along {110} micro deformation bands and {121} twins; the connecting surfaces are non-crystallographic coalescence features.","Microstructure design should aim to disrupt the continuous martensite network and trap diffusible hydrogen, for example through carbides, fine MnS inclusions, or Cu precipitates, so the crack never reaches the critical velocity."],"supporting_citations":[{"why":"Supplies the hydrogen-induced fast-fracture model, the two-stage crack growth picture, and the critical velocity threshold Vcrit ≈ 5.4 m/s for BCC iron.","marker":"[52]"},{"why":"Earlier study of the same steel without hydrogen showing {100} ferrite cleavage at 5% ductility; used as the baseline and to exclude stress partitioning as the cause of fast fracture here.","marker":"[33]"},{"why":"U-bend delayed-fracture study of the same steel where gradual hydrogen charging keeps crack velocity below the threshold and ferrite arrests cracks; provides the contrast case.","marker":"[5]"},{"why":"Extension of the fast-fracture model to strain-rate sensitivity; used to argue the observed strain-rate dependence reflects hydrogen diffusion kinetics.","marker":"[56]"},{"why":"Reports tangled versus linear dislocation morphologies in hydrogen-charged ferrite; used to rule out direct hydrogen-assisted cleavage and hydrogen-enhanced plasticity in ferrite.","marker":"[48]"},{"why":"Explains cleavage-front breakthrough at a grain boundary, used to account for diverging river patterns on the {100} ferrite surface.","marker":"[58]"},{"why":"Gives the ~1000 m/s brittle crack speed in steel, supporting the argument that hydrogen cannot diffuse ahead of a fast-running crack.","marker":"[49]"}],"fun_headline_variants":["Crack speed threshold, not hydrogen, triggers brittleness in DP steel","Hydrogen sets the stage, but crack speed drives fracture in DP steel","In 1.5 GPa DP steel, fast cracks turn ductile ferrite brittle","Brittle failure in DP steel: crack velocity, not hydrogen, is the switch","Hydrogen initiates, crack speed sustains brittle fracture in steel"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The crack actually runs faster than the critical velocity (roughly 5.4 m/s for body-centred cubic iron), a value inferred from fracture morphology and the model rather than measured; if the local crack speed stays below that threshold, the proposed explanation for sustained brittle cleavage after hydrogen depletion collapses.","fun_headline_variants_meta":{"raw":{"variants":["Crack speed threshold, not hydrogen, triggers brittleness in DP steel","Hydrogen sets the stage, but crack speed drives fracture in DP steel","In 1.5 GPa DP steel, fast cracks turn ductile ferrite brittle","Brittle failure in DP steel: crack velocity, not hydrogen, is the switch","Hydrogen initiates, crack speed sustains brittle fracture in steel"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000594,"raw_usage":{"total_tokens":2847,"prompt_tokens":1074,"completion_tokens":1773,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":690,"completion_tokens_details":{"reasoning_tokens":1686}},"tokens_in":690,"tokens_out":1773,"duration_ms":9385,"temperature":1.0,"reasoning_tokens":1686,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:17:24.971926+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly measure the running crack velocity during the stage where ferrite cleaves on {100}, for example by high-speed imaging, electrical potential drop, or acoustic emission, in a hydrogen-precharged tensile specimen. If the crack is moving slower than about 5.4 m/s while {100} cleavage continues after hydrogen depletion, the central claim fails; the same test should also confirm that hydrogen is actually depleted at the crack tip at that moment.","supporting_citations":[{"cited_title":"Asari, S","cited_arxiv_id":null,"evidence_quote":"Supplies the hydrogen-induced fast-fracture model, the two-stage crack growth picture, and the critical velocity threshold Vcrit ≈ 5.4 m/s for BCC iron."},{"cited_title":"Kwon, S.-P","cited_arxiv_id":null,"evidence_quote":"Earlier study of the same steel without hydrogen showing {100} ferrite cleavage at 5% ductility; used as the baseline and to exclude stress partitioning as the cause of fast fracture here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"U-bend delayed-fracture study of the same steel where gradual hydrogen charging keeps crack velocity below the threshold and ferrite arrests cracks; provides the contrast case."},{"cited_title":"Shishvan, G","cited_arxiv_id":null,"evidence_quote":"Extension of the fast-fracture model to strain-rate sensitivity; used to argue the observed strain-rate dependence reflects hydrogen diffusion kinetics."},{"cited_title":"Varanasi, M","cited_arxiv_id":null,"evidence_quote":"Reports tangled versus linear dislocation morphologies in hydrogen-charged ferrite; used to rule out direct hydrogen-assisted cleavage and hydrogen-enhanced plasticity in ferrite."},{"cited_title":"Morooka, O","cited_arxiv_id":null,"evidence_quote":"Explains cleavage-front breakthrough at a grain boundary, used to account for diverging river patterns on the {100} ferrite surface."}],"review_version":2}