{"id":"5730b093-c59b-40bf-b444-2bf13ed53388","arxiv_id":"2507.20635","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Level-set grain-growth simulations of diffusion-welded interfaces introduce two normalized crossing criteria and quantify how grain size and second-phase particle density, shape, and evolution control interface crossing.","lead":"This paper simulates, in 2D, how grain boundaries migrate across a diffusion-welded seam, and proposes two image-based measures of how much 'crossing' has occurred. The simulations show that smaller starting grains cross sooner and more completely, while dense or elongated particles at the seam pin boundaries and can even cause re-pinning.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"C1(ε) saturates and is ε-sensitive; the grain-size-independent crossing measure underlying the rankings is not established.","rationale":"The reader's weakest-assumption analysis focuses on the curvature-flow kinetic law and the undisclosed dissolution model. Those are real external-correctness risks. The concern raised here is more load-bearing because it targets the paper's central contribution—the C1 criterion itself—using evidence internal to the manuscript: the authors acknowledge C1 is highly ε-sensitive, fix ε without justification, and report a Case I in which C1 reaches 0.91 while their own C2 and pixel-profile analyses show the interface is still detectable. If C1 saturates or is grain-size-dependent through the ε/grain-size ratio, then the simulated rankings of Cases I-VII are not a reliable measure of interface crossing, and the central claim fails regardless of whether the kinetic law is exactly right. The concern does not require rejecting the simulations or impugning the authors; it requires a robustness demonstration and, ideally, experimental validation. Since the reader already issued CONDITIONAL, this concern reinforces that verdict rather than moving it: the paper should be accepted only if the ε-robustness of the rankings and the criterion's sensitivity to grain size are demonstrated, or if the claims are softened accordingly.","tokens_in":15503,"tokens_out":7353,"duration_ms":87890,"concrete_test":"Re-run the post-processing for all seven cases with ε = 1, 2, 3, 5, 10 pixels, and with ε scaled to keep ε/dASTM constant (about 13 px for Case I if Case II uses 3 px); report pixel calibration for each RVE. If the relative order of final C1 values changes, or if Case II no longer exceeds Case I at any ε, the headline ranking is ε-dependent. Also superimpose C1(t), C2(t), and the line-pixel profile for Case I: if C1 exceeds 0.9 while C2 remains above 1.3 or a localized interface peak persists, C1 does not measure crossing completion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The rankings depend on C1(ε), yet the paper shows C1 is highly sensitive to ε (Fig. 6) and sets ε = 3 pixels post hoc without a physical rule. In Case I, C1(ε=3) reaches 0.91 while C2 and the line-pixel profile still show a residual interface (§3.3, Figs. 8-9), so the criterion saturates before crossing is complete. Grain-size independence is not established either: ε is fixed in pixels, and if pixel size is not calibrated identically across RVEs, the physical strip width relative to initial grain size differs between Case I (105 µm) and Case II (24 µm); normalizing by Cref does not cancel this dependence because migration distances and boundary curvature scale with grain size. The final values (Case II 1.00 vs Case I 0.91; Cases III-VII 0.48-0.95) may therefore change with ε, and the headline ranking—fine grains and no obstacles are most favorable, elongated obstacles are hardest to bypass—could be an artifact of the chosen threshold rather than a robust material ranking.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies grain-boundary crossing of the initial bonding interface during diffusion welding using two-dimensional level-set grain-growth simulations. It introduces two normalized criteria, C1(ε) (interface-boundary length in a strip normalized by a reference line) and C2 (grain-count ratio), and applies them to seven simulated cases varying initial grain size, second-phase particle density, shape, and dissolution behavior. The main conclusions are that fine initial microstructures without obstacles cross most favorably, circular low-density particles are least disruptive, elongated interface-aligned particles are harder to bypass, and high obstacle density can cause re-pinning of grain boundaries.","tokens_in":15694,"tokens_out":5262,"duration_ms":60648,"significance":"If the central claims hold, the paper provides a useful computational framework for ranking diffusion-welding conditions with respect to interface crossing, complementing existing void-closure models. The use of a full-field level-set framework with a reduced mobility calibrated against external 316L heat-treatment data gives the kinetics some experimental anchoring, and the parametric campaign systematically varies grain size, obstacle density, shape, and evolution. The main caveat is that the crossing metric itself is not robustly established and no experimental crossing measurement is used as ground truth, so the quantitative rankings remain simulation-level predictions rather than validated material rankings.","major_comments":[{"comment":"The central metric C1(ε) is not robust: Fig. 6 shows strong sensitivity to ε with no clear plateau except at ε=1, and §3.3 reports that for Case I the ε=3 curve \"hardly reaches 1\" (final value 0.91 in Table 1) while C2 still indicates 1.5 times more grains at the interface than at the reference line and the line-pixel profile still shows a residual interface feature (Figs. 8–9). The criterion therefore saturates before crossing is complete, and the choice ε=3 pixels is made post hoc (\"this parameter was fixed to ε = 3 pixels\") without a physical rule. Because every cross-case comparison in Section 4 and Table 1 is based on final C1(ε=3) values, the rankings (fine grains favorable; elongated particles harder to bypass) may be threshold artifacts rather than material rankings.","section":"§3.3, Eq. (10), Fig. 6"},{"comment":"The claimed grain-size independence of C1 is not established. ε is fixed in pixels, but the physical width of the analysis strip relative to the initial grain size differs between Case I (105 µm) and Case II (24 µm) unless the pixel calibration is identical, which is not stated; normalizing by Cref does not remove this dependence because both Cint and Cref depend on grain size and ε. The paper should either compute C1 with ε scaled by initial grain size or by a fixed physical length calibrated identically in all RVEs, or show convergence of the relative rankings with respect to ε before using C1 to rank cases with different grain sizes.","section":"§3.2, Eq. (10)"},{"comment":"The calibration equation is incomplete as written: (µγ)1 = texp/t0 has dimensions of time and cannot by itself yield a reduced mobility; the relation must involve the Burke–Turnbull constant α, the initial and final mean radii, and the reference simulation mobility (µγ)0. Since the reduced mobility sets the absolute time scale for every crossing simulation, the missing formula prevents reproduction of the kinetics; please give the complete least-squares calibration used for the 316L datasets.","section":"§3.1, Eq. (8)"},{"comment":"Case VII is not independently checkable: the pore-dissolution kinetics are described only as \"a realistic kinetics model—undisclosed here for confidentiality reasons.\" Since the conclusion that dynamic dissolution improves crossing but does not fully recover the obstacle-free case depends on this model, the governing equation and parameters must be provided or at least benchmarked against a public analytical or experimental case; otherwise this case should be removed from the central claims.","section":"§4.4"},{"comment":"The entire campaign uses the curvature-flow law v = −µγκn (Eq. 1) with homogeneous γ and a mobility calibrated against grain-growth heat-treatment data (Eq. 8). As the authors acknowledge, this law is a first-order approximation, and welding interfaces may involve anisotropy, torque, stored energy, or solute drag. This is not by itself an error, but it means the predicted crossing kinetics and rankings are conditional on the bulk grain-growth law; a direct experimental check of at least one simulated ranking (e.g., fine vs. coarse initial grain size) is needed before the results can be read as robust predictions of diffusion-welded material behavior.","section":"§2.1 and §3.1"}],"minor_comments":[{"comment":"In the sentence \"the same RVE and initial obstacle distribution as in Case VII were used,\" the case reference should almost certainly be Case VI, since Case VII is the dynamic-obstacle variant of Case VI.","section":"§4.4"},{"comment":"In Eq. (11), the symbol Nint is used both for the interface count and for the reference-line count; the reference-line count should be denoted Nref to match the surrounding text.","section":"§3.2, Eq. (11)"},{"comment":"Several typos should be corrected, including \"emphazised\" (§3.1), \"inital\" (§1), \"thez total number\" (§3.3), and \"partic les\" (Fig. 17 caption).","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits a computational materials science venue and the qualitative trends are plausible, but the central metric and calibration need substantive work. I would not reject on novelty grounds; the main concerns are the ε-sensitivity of C1, the incomplete calibration equation (Eq. 8), and the undisclosed kinetics in Case VII, all of which are addressable within revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Christophe—quick take on arXiv:2507.20635. The genuinely new thing here is the pair of measurement criteria: C1, a grain-size-normalized version of Bouquet's C_int, and C2, a grain-number ratio at the interface vs a reference line. The parametric level-set campaign (fine vs coarse grains; circular vs elongated particles; density series; dissolving particles) is a sensible first pass at ranking crossing scenarios. The qualitative trends—fine grains cross faster, dense particles pin, elongated particles aligned with the interface are harder to bypass—are plausible and consistent with Zener-pinning physics. Credit where due: the authors are upfront that the curvature-flow law (Eq. 1) is a first-order approximation, and they calibrate reduced mobility against an external 316L dataset rather than fitting the crossing outcomes.\n\nThat said, the central metric has a real problem. C1(ε) is highly sensitive to the strip thickness ε (their Fig. 6), and ε=3 pixels is chosen post hoc with no physical rule. In Case I, C1 saturates at 0.91 while C2 and the line-pixel profile still show a residual interface—so the criterion is declaring crossing before it is complete. The claim that C1 removes grain-size dependence is also shaky: ε is fixed in pixels, and the paper doesn't show that the physical strip width relative to grain size is the same across RVEs. Normalizing by Cref doesn't obviously cancel grain-size-dependent migration distances and curvature. So the final rankings (Case II 1.00 vs Case I 0.91; Cases III–VII 0.48–0.95) may be threshold artifacts rather than robust material rankings.\n\nOther soft spots are minor-to-moderate: no experimental validation of either criterion (the only external data calibrate the mobility), one run per configuration with no error bars, proprietary software with no shipped parameters, and an undisclosed 'realistic kinetics model' for pore dissolution in Case VII—that one is a genuine gap because the reader can't check whether the dissolution rate drives the result. These are fixable weaknesses, not fatal ones. The simulation framework itself is established and well-cited.\n\nBottom line: this is a serious paper for the diffusion-welding community, one that deserves referee time. It probably needs a major revision to either justify ε physically or report results across an ε range, and to add some experimental or literature-based crossing measurement for the criteria to land. But the qualitative conclusions are worth taking seriously even if the quantitative C1 values don't yet carry the weight the title implies. I'd send it out.","headline":"A useful simulation paper with two genuinely new crossing criteria, but the headline rankings rest on an epsilon-sensitive metric that is not convincingly grain-size independent.","tokens_in":16255,"tokens_out":1887,"would_cite":false,"duration_ms":19868,"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":"A normalized crossing criterion together with level-set grain-growth simulations ranks diffusion-welding configurations, with fine grains and obstacle-free interfaces crossing best.","keywords":["diffusion welding","interface crossing","grain growth","level-set method","second-phase particles","Zener pinning","pore dissolution","316L stainless steel"],"falsifier":"Take a diffusion-welded 316L or titanium sample with known initial grain size and known particle population at the bond plane, section it after the same thermal cycle, measure $C_1(\\varepsilon)$ on several fields of view, and compare with the simulated curves: if the coarse-grain case crosses as fast as the fine-grain case, or if elongated particles are crossed more often than circular ones at equal linear fraction, the central ranking is falsified.","tokens_in":15288,"feed_emoji":"🔬","tokens_out":4674,"duration_ms":54412,"temperature":0.7,"pith_summary":"This paper tries to establish that the disappearance of a diffusion-welded interface—grain boundaries crossing the original bond plane—can be simulated at the polycrystalline scale and measured by a grain-size-independent normalized criterion. The authors build two-dimensional full-field level-set simulations of grain growth on both sides of a welding interface, with and without second-phase particles, and rank the configurations by how completely and how fast the interface is crossed. They report that fine initial grains cross fastest, that circular obstacles at low density hinder crossing least, that elongated obstacles aligned with the interface are hardest to bypass, and that dense obstacle populations can re-pin boundaries after an initial improvement. If the approach is right, microstructural features that determine weld quality can be ranked by simulation before expensive welding trials.","feed_headline":"Fine grains and clean interfaces cross fastest in diffusion-weld model","feed_subtitle":"A normalized criterion lets level-set grain-growth simulations compare obstacle density and shape for bonding quality.","key_machinery":"The load-bearing object is the level-set description of the polycrystal: each grain is a signed distance function $\\phi_i(x,t)$, and grain-boundary motion obeys $\\vec{v} = -\\mu\\gamma\\kappa\\vec{n}$, so with a distance function the convection equation becomes a heat-type equation $\\partial_t\\phi_i - \\mu\\gamma\\Delta\\phi_i = 0$. A calibrated reduced mobility $\\mu\\gamma$ is fitted by matching simulated 316L grain growth to Burke-Turnbull data. Obstacles are inserted either as holes in the mesh (static particles) or as an additional level-set population with an evolution law (dissolving pores). The crossing state is read from thresholded images: $C_1(\\varepsilon) = L_{int}(\\varepsilon)/L_{ref}(\\varepsilon)$ counts remaining boundary length in an $\\varepsilon = 3$-pixel strip around the interface normalized by a reference line; $C_2$ counts grains crossing the interface relative to a reference line. The criterion and the mobility law together carry the ranking argument.","core_discovery":"The central claim is that interface crossing during diffusion welding can be captured by coupling the curvature-driven level-set formulation of grain growth with two quantitative criteria. Criterion $C_1(\\varepsilon)$ divides the remaining grain-boundary length in a thin strip around the original bonding interface by the same measurement on a reference line far from the interface, removing the dependence on grain size that plagued the earlier criterion $C_{int}$. Criterion $C_2$ compares grain counts at the interface and a reference line, but proved noise-sensitive and was not adopted. The simulation campaign then shows a clear ordering: a fine initial microstructure without obstacles reaches the highest crossing ($C_1 = 1.00$), coarse grains without obstacles lag ($0.91$), circular particles at $2.6\\%$, $6.5\\%$, and $12.5\\%$ linear fraction give $0.92$, $0.74$, and $0.48$, elongated particles at $6.9\\%$ give $0.87$, and the same elongated population that dissolves early gives $0.95$.","pith_inferences":["Editorial extension: an experimentalist could apply the $C_1$ protocol directly to micrographs of welded cross-sections, since it requires only two cut lines and pixel counting; this would test the ranking without needing the simulation.","Editorial extension: the paper's proposed future parameter—critical grain size as a function of precipitate spacing—could make the crossing results a design rule analogous to the Zener limit for grain growth.","Editorial extension: because the simulations are 2D and use isotropic curvature flow, the ranking of obstacle shapes may differ in 3D, where particle bypass involves different topology; this is an inference, not the paper's claim.","Editorial extension: the undisclosed evolution law for dissolving pores is a real reproducibility gap; a public version of that law would let the dynamic-obstacle result be independently checked."],"forward_implications":["The $C_1(\\varepsilon)$ criterion gives a grain-size-independent measure of interface crossing that can be applied to both simulated and experimental images.","Fine initial grain size accelerates crossing before the temperature plateau because smaller boundary curvature radii create higher capillary pressure.","Higher linear fractions of circular second-phase particles pin grain boundaries and can cause re-pinning, making crossing deteriorate after an initial rise.","Elongated particles aligned with the interface are harder to bypass, and the number of pinning points matters more than the occupied interface fraction.","If obstacles dissolve early in the thermal cycle, crossing improves but still does not reach the obstacle-free level."],"supporting_citations":[{"why":"Supplies the stage description of diffusion welding (contact, void closure, crossing) that motivates the study.","marker":"[7]"},{"why":"Introduces the earlier normed crossing criterion $C_{int}(\\varepsilon)$ that this paper modifies into $C_1$.","marker":"[19]"},{"why":"Provides the finite-element level-set framework for grain growth in the presence of second-phase particles.","marker":"[32]"},{"why":"Extends the level-set approach to an evolving population of second-phase particles, used for the dissolving-obstacle case.","marker":"[35]"},{"why":"Reviews the level-set numerical treatment and its limitations for solid-state microstructural evolution, supporting the chosen methods.","marker":"[37]"},{"why":"Supplies the Burke-Turnbull grain-growth law used to calibrate the reduced mobility against 316L heat-treatment data.","marker":"[55]"}],"fun_headline_variants":["Fine grains fastest to cross weld interface in model","Obstacles slow diffusion weld interface crossing","Level-set model: clean fine grains beat particles in weld bond","Weld bond quality: grain size and particles ranked by new criterion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole ranking depends on the assumption that grain boundaries in the welded metal move at a speed proportional to their curvature, with one average boundary energy and a fitted mobility constant—if that is not how these interfaces move, the predicted crossing order could change.","fun_headline_variants_meta":{"raw":{"variants":["Fine grains fastest to cross weld interface in model","Obstacles slow diffusion weld interface crossing","Level-set model: clean fine grains beat particles in weld bond","Weld bond quality: grain size and particles ranked by new criterion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000139,"raw_usage":{"total_tokens":1099,"prompt_tokens":829,"completion_tokens":270,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":445,"completion_tokens_details":{"reasoning_tokens":205}},"tokens_in":445,"tokens_out":270,"duration_ms":4894,"temperature":1.0,"reasoning_tokens":205,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T13:24:45.814844+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a diffusion-welded 316L or titanium sample with known initial grain size and known particle population at the bond plane, section it after the same thermal cycle, measure $C_1(\\varepsilon)$ on several fields of view, and compare with the simulated curves: if the coarse-grain case crosses as fast as the fine-grain case, or if elongated particles are crossed more often than circular ones at equal linear fraction, the central ranking is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the stage description of diffusion welding (contact, void closure, crossing) that motivates the study."},{"cited_title":"Bouquet, Etude de la formation des joints soudés par diffusion : application aux échangeurs de chaleur compacts, Ph.D","cited_arxiv_id":null,"evidence_quote":"Introduces the earlier normed crossing criterion $C_{int}(\\varepsilon)$ that this paper modifies into $C_1$."},{"cited_title":"Agnoli, M","cited_arxiv_id":null,"evidence_quote":"Provides the finite-element level-set framework for grain growth in the presence of second-phase particles."},{"cited_title":"Alvarado, S","cited_arxiv_id":null,"evidence_quote":"Extends the level-set approach to an evolving population of second-phase particles, used for the dissolving-obstacle case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reviews the level-set numerical treatment and its limitations for solid-state microstructural evolution, supporting the chosen methods."},{"cited_title":"Burke, D","cited_arxiv_id":null,"evidence_quote":"Supplies the Burke-Turnbull grain-growth law used to calibrate the reduced mobility against 316L heat-treatment data."}],"review_version":1}