{"id":"6bb29a0e-117c-4fea-906f-9dc4be5d4eb1","arxiv_id":"2411.15983","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Shear coupling, not anisotropy or triple junction drag, explains why grain boundary velocities in polycrystals are virtually uncorrelated with boundary curvature.","lead":"Using large-scale phase field simulations, this paper shows that the long-held picture of grain growth as motion driven by boundary curvature fails when grain boundary migration is coupled to shear deformation, a coupling that generates internal stresses. The simulations reproduce the weak velocity-curvature correlation seen in recent nickel, iron, and aluminum experiments, and point to shear coupling as the main culprit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central quantitative claim rests on an uncalibrated scaling factor S=1/30; because the elastic driving force scales as S², whether shear coupling is the main cause of the experimental velocity–curvature decorrelation is not yet settled.","rationale":"The paper makes a strong qualitative case and has genuine independent support: the PF model with measured bicrystal shear-coupling factors reproduces both microstructure evolution and stress fields in the idealized MD benchmarks (Fig. 5) without tunable physical parameters, and the control simulations with anisotropic energy, anisotropic mobility, and TJ drag leave COD an order of magnitude too high. What is less secure is the quantitative bridge from this validated mechanism to the experimental scatter. That bridge is built with S = 1/30, a single parameter applied to every GB and described in the SI as arbitrary. Because the self-stress and the elastic migration force both depend on β, the decorrelation effect scales as S²; the computed COD values in Fig. 4g,h are therefore extremely sensitive to this choice. Without a direct measurement or at least a sensitivity analysis of S, the experiments could either be reproduced with a different S or fail to be reproduced with a smaller S. The same uncertainty is amplified by the acknowledged neglect of bulk plastic relaxation, since stresses near μ/10 would be plastically relieved in real metals. This is exactly the weakest assumption identified by the reader; my read does not move the verdict, but it sharpens the condition: the central attribution requires a quantitative check on S.","tokens_in":22564,"tokens_out":5554,"duration_ms":51272,"concrete_test":"Rerun the Fig. 4g,h polycrystal simulations with S = 1/100, 1/30, 1/10, and 1/3 in Eq. (S-12), holding all other parameters fixed, and recompute the coefficient of determination for v vs κ. If COD at S = 1/100 is much larger than the experimental range (0.036–0.334), or if a narrow S window is required to match experiments, the main-cause claim is not robust; a complementary run with a von Mises yield cap on the elastic stress would show whether plastic relaxation changes the conclusion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's attribution of the v–κ decorrelation to shear coupling is conditional on the magnitude of the elastic driving term. In the polycrystal simulations every GB is assigned β = S β_bicrystal with S = 1/30 (SI §S-II C, Eq. S-12), introduced as 'chosen arbitrarily.' The self-stress in Eq. (S-3) is proportional to β, and the elastic driving force in Eq. (S-9) is τ·β, so the whole effect scales as S². A factor-of-three reduction in S cuts the elastic driving force by an order of magnitude; the high-COD curvature-flow limit should reappear. Yet no direct polycrystalline measurement of β is provided to pin S; the cited qualitative evidence only says polycrystal coupling is smaller than in bicrystals. The MD benchmarks in Fig. 5 use measured bicrystal β and validate the mechanism, but they do not constrain the statistical value of S in a 1000-grain polycrystal. The acknowledged absence of bulk plastic relaxation points the same way: simulated stresses reach ~μ/10 near TJs, and plastic relief would act as an additional reduction of the effective elastic driving force. Thus the central quantitative claim—that shear coupling, not anisotropy or TJ drag, is the main cause—rests on an uncalibrated parameter and remains conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that grain growth in polycrystals is not mean curvature flow because shear coupling between GB migration and shear deformation generates internal stresses that add an elastic driving force (τ·β) to the GB equation of motion. Using a diffuse-interface (phase-field) implementation of a disconnection-mediated GB model, the authors simulate 2D polycrystalline grain growth with and without shear coupling, quantify the velocity–curvature correlation (COD), compare it with experimental and MD data for Ni, α-Fe, and Al, and test alternative explanations (anisotropic GB energy/mobility and triple-junction drag). They also validate the model against idealized MD microstructures from Thomas et al. and conclude that internal stresses from shear coupling are the main cause of the observed failure of curvature flow and of the von Neumann–Mullins relation.","tokens_in":22765,"tokens_out":3705,"duration_ms":38618,"significance":"If the central claim is established, this is a significant result: it would imply that classical and weighted mean-curvature descriptions of grain growth are insufficient for real polycrystals and that mechanical driving forces must be included in microstructure evolution models. The paper has notable strengths: the idealized-microstructure MD benchmarks use measured bicrystal shear-coupling factors with no adjustable parameters; the comparison of velocity–curvature scatter across experiments, MD, and the present simulations is systematic; and the paper explicitly tests—and finds wanting—anisotropic energy, anisotropic mobility, and triple-junction drag as alternative explanations. These features make the paper valuable regardless of the quantitative calibration issue discussed below.","major_comments":[{"comment":"The polycrystalline simulations assign every GB a shear-coupling factor β = S·β_bicrystal with S = 1/30, introduced as 'chosen arbitrarily.' Because the elastic driving force in Eq. (2) is τ·β and the self-stress in Eq. (S-3) is proportional to β, the entire elastic driving contribution scales as S². A factor-of-three reduction in S therefore reduces the elastic driving force by roughly an order of magnitude, and the predicted velocity–curvature correlation should move substantially back toward the curvature-flow limit. The reported COD values in Figs. 4g,h are thus not a parameter-free prediction but a demonstration conditional on the order of magnitude of S. The manuscript needs either a direct quantitative constraint on polycrystalline shear-coupling factors, a sensitivity analysis in S, or a clearly stated bounds argument showing that the conclusions are robust over the plausible range of S.","section":"SI §S-II C, Eq. (S-12)"},{"comment":"The MD benchmarks in Fig. 5 use shear-coupling factors measured in bicrystal MD simulations and validate the mechanism in small, idealized microstructures with a small number of GB types. These benchmarks do not constrain the statistical value of S in a 1000-grain polycrystalline network, where constraints from neighboring grains and triple junctions activate additional disconnection modes and reduce the effective coupling. The cited qualitative evidence that polycrystalline coupling is smaller than bicrystal coupling supports the direction of the correction but not the specific factor 1/30. As written, the central quantitative claim rests on this uncalibrated factor.","section":"Main text, Discussion (Fig. 5) and SI §S-II D"},{"comment":"The authors acknowledge that bulk plastic deformation is not included in the simulations and that simulated stresses reach values as large as μ/10, comparable to the theoretical strength. Plastic relaxation in real polycrystals would reduce the internal stresses and hence reduce the elastic driving force τ·β in Eq. (2). This is not a minor detail: it is another mechanism—independent of S—that could lower the effective decorrelation strength. The paper should discuss quantitatively whether the conclusion that shear coupling is the 'main' cause survives when plastic relaxation is present, or whether the claimed dominance should be restricted to microstructures without significant intragranular plasticity.","section":"Main text, text following Fig. 2"},{"comment":"The comparison between the 2D shear-coupling simulations in Figs. 3b and the 3D SrTiO3 experimental data in Fig. 3d is qualitative rather than quantitative. In 3D the appropriate von Neumann–Mullins-type statement involves the mean width of the grain, not simply the number of neighbors, as the authors themselves note in the footnote. The large scatter in the 3D experiment is suggestive and consistent with the authors' mechanism, but it cannot by itself establish that shear coupling is the cause of the 3D von Neumann–Mullins failure. This comparison should be framed as a qualitative analog, not as a direct validation.","section":"Main text, Fig. 3 and accompanying footnote"}],"minor_comments":[{"comment":"The phrase 'quantitively answer' should be 'quantitatively answer.'","section":"Introduction, first paragraph"},{"comment":"M is described as an 'intrinsic disconnection mobility tensor,' but throughout the simulations it is used as a scalar mobility. Please clarify whether the tensor character is retained and what off-diagonal components would mean, or state that the isotropic/scalar limit is used.","section":"Main text, Eq. (2) and Methods"},{"comment":"Panels (a,b) plot δR2/δt while panels (c,d) plot δR3/δt, but the axes are labeled identically except for the variable name; the different physical dimensions and scales make direct visual comparison difficult. Please add normalizations or a clearer statement that the two rows are separate quantities.","section":"Fig. 3 caption and axes"},{"comment":"The sentence introducing S says it is 'chosen arbitrarily' but 'of the correct order of magnitude compared with experimental observations.' If no additional constraint can be provided, at least identify which experimental observations set this order of magnitude and how a factor-of-1/30 estimate was obtained.","section":"SI §S-II C"},{"comment":"In Figs. 2e and 2h the color or line assignment of distributions at different times is not explicitly explained in the caption; the statement 'the lightest and darkest lines are for the initial and final microstructures' should be complemented by a scale or legend for intermediate times.","section":"Main text, Fig. 2 panels"}],"recommendation":"major_revision","confidential_remarks":"The paper's core mechanism is credible and the MD validation is a real strength, but the main quantitative claim about the dominance of shear coupling depends on the arbitrarily chosen S=1/30 factor and on the omission of bulk plastic relaxation. I recommend major revision rather than rejection, because the issue is a calibration and robustness question that could be addressed with additional simulations and sensitivity analysis rather than a fundamental flaw in the model. I did not find evidence of any authorship or citation concerns, and the topic fits the journal's scope well."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper says shear coupling — not anisotropic energy, mobility, or triple-junction drag — is why grain boundaries in polycrystals don't follow curvature flow. The claim is bold and mostly convincing in mechanism, but the quantitative link to the experimental scatter of the velocity–curvature correlation depends on an uncalibrated scaling factor S=1/30.\n\nWhat's actually new: the authors run large-scale phase-field grain growth simulations with a disconnection-based equation of motion that includes the elastic driving force τ·β. With shear coupling, they reproduce the breakdown of the von Neumann–Mullins relation and the weak v–κ correlation seen in Ni, Fe, and MD Al, and they show that anisotropy and TJ drag alone leave COD around 0.86–0.99, far from the experimental <0.35. That ruling out is a useful result in itself.\n\nThe strongest evidence is the MD benchmark in Fig. 5. For two idealized microstructures, the PF model with measured bicrystal coupling factors reproduces the MD shrinkage behavior and stress fields with no adjustable physical parameters. That gives real confidence the model captures the mechanism.\n\nNow the soft spots. The polycrystal simulations assign every GB a coupling factor β = S·β_bicrystal with S=1/30, introduced in the SI as 'chosen arbitrarily.' Because the elastic driving force scales as τ·β and the self-stress scales as β, the whole decorrelation effect scales as S². A factor-of-three reduction in S cuts the effect by an order of magnitude and the high-COD curvature-flow limit should reappear. The paper cites qualitative evidence that polycrystal coupling is smaller than bicrystal, but there is no direct measurement constraining S. The shear-modulus variation (μ=25 vs 75 GPa) is a useful sensitivity test, but it varies the whole elastic response, not the statistical distribution of β. So the central attribution — that shear coupling is the main cause — is conditional on S being in the right ballpark. Also, the simulations are 2D while two of the key experimental comparisons (Ni, SrTiO3) are 3D; the 3D argument is made by analogy. The acknowledged neglect of bulk plastic relaxation matters in the same direction: simulated stresses reach ~μ/10 near triple junctions, and real metals would likely relax some of that, further reducing the effective driving force.\n\nNet: the paper is serious, honest about its limitations, and makes a strong qualitative case. The quantitative claim is not settled. I'd send it to a thoughtful referee, with instructions to focus on S and on whether the 'mainly originates' language can be supported. If the authors add a sensitivity analysis on S or directly measure polycrystalline β, the paper could become a landmark. As is, it's a valuable contribution that deserves review, not a desk reject.","headline":"A well-validated mechanism and a strong MD benchmark, but the central quantitative claim leans on an arbitrary scaling factor; still deserves serious review.","tokens_in":23332,"tokens_out":2876,"would_cite":true,"duration_ms":26532,"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":"The paper shows that grain growth departs from classical curvature flow because grain boundaries shear as they migrate, and the internal stresses this generates, not anisotropy or triple-junction drag, are the main source of the observed…","keywords":["grain growth","shear coupling","mean curvature flow","von Neumann-Mullins relation","phase-field simulation","disconnections","internal stress","grain boundary migration"],"falsifier":"Measure the shear-coupling factors of many individual grain boundaries within a growing polycrystal, for example by combining in-situ diffraction mapping of grain rotation and shear with boundary migration tracking during annealing, and use the measured distribution rather than $S=1/30$ in the same model. If the measured coupling is an order of magnitude smaller, the predicted velocity–curvature decorrelation should largely disappear; if it is close to the scaled values, the claim is supported.","tokens_in":22301,"feed_emoji":"⚙️","tokens_out":7844,"duration_ms":68571,"temperature":0.7,"pith_summary":"The paper argues that grain growth in polycrystals is not mean curvature flow, and that the failures of the classical description are caused mainly by the shear deformation that accompanies grain-boundary migration. When grain boundaries move, they also shear, and in a constraining polycrystalline network this generates internal stresses that add a mechanical driving force to the usual capillarity force. Using large-scale phase-field simulations built on a disconnection-based model of boundary motion, the authors reproduce the weak velocity–curvature correlation and the breakdown of the von Neumann–Mullins relation seen in experiments on nickel and iron and in atomistic simulations of aluminum. If correct, the result says that predicting microstructure evolution requires coupling mechanics to capillarity, not just anisotropic curvature flow.","feed_headline":"Shear coupling explains why grain growth is not curvature flow","feed_subtitle":"With shear coupling included, simulations reproduce the velocity-curvature scatter seen in nickel, iron, and aluminum.","key_machinery":"The load-bearing object is the disconnection: a line defect on a grain boundary that carries both a step, so it moves the boundary, and a Burgers vector, so it shears the boundary. The ratio of shear rate to normal migration rate is the shear-coupling factor $\\beta = v_\\parallel/v_\\perp$, assigned by misorientation to each boundary. In a polycrystal, boundaries cannot shear freely, so disconnection flow builds an internal stress field; the machinery is the phase-field discretization of the equation of motion $v = M(\\Gamma\\kappa + \\boldsymbol{\\tau}\\cdot\\boldsymbol{\\beta} + \\psi)\\hat n$, where the elastic term $\\boldsymbol{\\tau}\\cdot\\boldsymbol{\\beta}$ enters through an added force density $f_{\\rm elastic}=\\boldsymbol{\\tau}\\cdot\\boldsymbol{\\beta}_{ij}|\\nabla\\eta_j|(\\hat n(\\eta_i)\\cdot\\hat n(\\eta_j))$. This couples each boundary's velocity to the resolved shear stress from all other boundaries' disconnections, which is what decorrelates velocity from curvature.","core_discovery":"The central claim is that the well-documented failure of mean curvature flow and the von Neumann–Mullins relation in real grain growth is not an anisotropy or triple-junction drag effect, but the direct consequence of shear-coupled grain-boundary migration. The paper proposes the equation of motion $v = M(\\Gamma\\kappa + \\boldsymbol{\\tau}\\cdot\\boldsymbol{\\beta} + \\psi)\\hat n$ in place of $v = M\\gamma\\kappa$: the new term $\\boldsymbol{\\tau}\\cdot\\boldsymbol{\\beta}$ is the elastic driving force exerted by internal shear stresses on the disconnections that mediate boundary motion. Because those stresses are generated by the boundaries' own motion and are spatially and temporally inhomogeneous, boundaries can migrate faster, slower, or even opposite to their curvature. The authors support this with two-dimensional phase-field simulations that, with shear coupling included, reproduce both the scatter in grain size change versus neighbor number and the weak velocity–curvature correlation of experiments and molecular dynamics, while simulations with anisotropic energy, anisotropic mobility, or triple-junction drag alone do not.","pith_inferences":["Beyond the paper, the same mechanism may unify previously separate observations—stress-driven grain growth, grain rotation, and grain-boundary sliding—as different manifestations of one disconnection-mediated, shear-coupled kinetics.","Beyond the paper, a direct quantitative test would be to measure the actual distribution of shear-coupling factors of individual boundaries inside a growing polycrystal, for example from in-situ lattice-rotation and strain maps, and feed that distribution into the same model without the arbitrary $S=1/30$ scaling.","Beyond the paper, the results imply that microstructure evolution codes for annealing and thermomechanical processing should solve for the internal stress field concurrently with boundary motion; residual stresses are not just a response to grain growth but also a driver of it.","Beyond the paper, the prediction could be checked by comparing grain growth in high-modulus and low-modulus materials at the same homologous temperature, since the model predicts weaker velocity–curvature correlation in higher-modulus materials."],"forward_implications":["The von Neumann–Mullins relation, and its anisotropic and three-dimensional extensions, is not a valid quantitative description of polycrystalline grain growth; grain size changes are not reliably predicted by neighbor count.","Grain boundaries should frequently migrate against their local curvature when internal stresses are strong enough, a signature that has been seen experimentally and appears in the simulations only when shear coupling is included.","Grain growth proceeds faster with shear coupling (roughly 50 percent faster in the two-dimensional simulations here) and dissipates elastic energy, so stored internal stress is part of the coarsening process itself.","Anisotropic grain-boundary energy, anisotropic mobility, and triple-junction drag are real effects but are too weak to explain the observed scatter; quantitative models of microstructure evolution should include mechanical coupling.","At sufficiently high temperatures, where multiple disconnection modes weaken net shear coupling, grain growth should approach curvature flow, but most experimentally relevant temperatures are far from that limit."],"supporting_citations":[{"why":"Supplies the experimental Ni polycrystal dataset showing grain-boundary velocity and curvature are nearly uncorrelated, plus the method used to extract velocities and curvatures.","marker":"[18]"},{"why":"Provides experimental α-Fe data showing weak velocity–curvature correlation and boundaries migrating opposite to curvature, a key observational target.","marker":"[19]"},{"why":"Provides molecular dynamics results for nanocrystalline Al with weak velocity–curvature correlation, used as the atomistic comparison.","marker":"[20]"},{"why":"Provides MD simulations of idealized microstructures showing shear coupling and internal stresses control which grains shrink, used to benchmark the phase-field model.","marker":"[24]"},{"why":"Derives the grain-boundary equation of motion $v=M(\\Gamma\\kappa+\\tau\\cdot\\beta+\\psi)\\hat n$ that Eq. (2) of this paper adopts.","marker":"[25]"},{"why":"Sets out the continuum disconnection-mediated model of interface migration on which the simulations are based.","marker":"[26]"},{"why":"Provides the diffuse-interface phase-field formulation of that continuum model, which the present simulations build on.","marker":"[27]"},{"why":"Gives the unified disconnection/grain-boundary kinetics framework and the misorientation-dependent shear-coupling factor expression used to assign couplings.","marker":"[23]"},{"why":"Supplies the 3D curvature-flow simulation results and SrTiO3 experimental grain-growth data compared in the von Neumann–Mullins failure analysis.","marker":"[35]"},{"why":"Establishes the coupling of grain-boundary motion to shear deformation, the basic physical effect the explanation rests on.","marker":"[22]"}],"fun_headline_variants":["Shear coupling, not curvature, governs grain growth","Grain growth's curvature failure traced to shear coupling","Why grain growth defies curvature: shear coupling","Shear coupling rewrites grain growth equations","Grain growth's true driver: shear-coupled boundaries"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the shear-coupling factors of grain boundaries inside a polycrystal are comparable to the scaled bicrystal values used here, with the scaling factor $S=1/30$ chosen arbitrarily in the paper; if real polycrystalline shear coupling is much weaker, the simulated decorrelation between velocity and curvature would shrink.","fun_headline_variants_meta":{"raw":{"variants":["Shear coupling, not curvature, governs grain growth","Grain growth's curvature failure traced to shear coupling","Why grain growth defies curvature: shear coupling","Shear coupling rewrites grain growth equations","Grain growth's true driver: shear-coupled boundaries"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000732,"raw_usage":{"total_tokens":3254,"prompt_tokens":901,"completion_tokens":2353,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":2280}},"tokens_in":517,"tokens_out":2353,"duration_ms":15822,"temperature":1.0,"reasoning_tokens":2280,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:39:14.802679+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the shear-coupling factors of many individual grain boundaries within a growing polycrystal, for example by combining in-situ diffraction mapping of grain rotation and shear with boundary migration tracking during annealing, and use the measured distribution rather than $S=1/30$ in the same model. If the measured coupling is an order of magnitude smaller, the predicted velocity–curvature decorrelation should largely disappear; if it is close to the scaled values, the claim is supported.","supporting_citations":[{"cited_title":"Bhattacharya, Y.-F","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental Ni polycrystal dataset showing grain-boundary velocity and curvature are nearly uncorrelated, plus the method used to extract velocities and curvatures."},{"cited_title":"Xu, Y.-F","cited_arxiv_id":null,"evidence_quote":"Provides experimental α-Fe data showing weak velocity–curvature correlation and boundaries migrating opposite to curvature, a key observational target."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides molecular dynamics results for nanocrystalline Al with weak velocity–curvature correlation, used as the atomistic comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides MD simulations of idealized microstructures showing shear coupling and internal stresses control which grains shrink, used to benchmark the phase-field model."},{"cited_title":"Zhang, J","cited_arxiv_id":null,"evidence_quote":"Derives the grain-boundary equation of motion $v=M(\\Gamma\\kappa+\\tau\\cdot\\beta+\\psi)\\hat n$ that Eq. (2) of this paper adopts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Sets out the continuum disconnection-mediated model of interface migration on which the simulations are based."},{"cited_title":"Salvalaglio, D","cited_arxiv_id":null,"evidence_quote":"Provides the diffuse-interface phase-field formulation of that continuum model, which the present simulations build on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the unified disconnection/grain-boundary kinetics framework and the misorientation-dependent shear-coupling factor expression used to assign couplings."},{"cited_title":"Muralikrishnan, H","cited_arxiv_id":null,"evidence_quote":"Supplies the 3D curvature-flow simulation results and SrTiO3 experimental grain-growth data compared in the von Neumann–Mullins failure analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the coupling of grain-boundary motion to shear deformation, the basic physical effect the explanation rests on."}],"review_version":1}