{"id":"bfd8977e-f287-4021-a3cd-2db9120541e2","arxiv_id":"2608.05589","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"The paper introduces hS-IGA, a hybrid global B-spline / local Lagrange s-method that eliminates coupling-integrand discontinuities and reduces integration points by about 81% in 2D and 95.6% in 3D dynamic crack benchmarks.","lead":"A hybrid finite element strategy is proposed for dynamic crack propagation, combining smooth B-spline functions in the coarse global mesh with standard finite elements in the fine crack-tip mesh. It removes a known quadrature bottleneck in the s-method and cuts integration points by about 81% in 2D and 95.6% in 3D benchmarks while preserving accuracy of stress intensity factors and local stress.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fixed 3x3 Gauss for coupling integrands is only C0 at global element boundaries; accuracy is demonstrated only empirically, and no quadrature error estimate covers worst-case overlaps.","rationale":"The reader's weakest assumption identifies exactly the fixed-Gauss accuracy over local elements spanning multiple global IGA elements, and the present concern sharpens it: continuity alone does not imply that 3x3 Gauss quadrature is accurate, because the stiffness coupling integrand is only C0 (with a kink) at global knot boundaries. This is load-bearing because the method's entire efficiency gain is premised on removing recursive subdivision while retaining accuracy. The paper provides strong empirical evidence on the selected benchmarks, including verification against analytical and refined-FEA reference solutions and consistent large integration-point reductions in both 2D and 3D, so the concern does not justify rejection. It does, however, justify keeping the verdict CONDITIONAL: the central claim should be accompanied by a quadrature error estimate or a worst-case overlap test before the method is adopted for new problems. No ad hominem or outside-consensus issue is raised; the critique is internal to the correctness argument for fixed Gauss quadrature.","tokens_in":54965,"tokens_out":6716,"duration_ms":76085,"concrete_test":"Assemble the 2D coupling stiffness entries K_GL of Eq. (46) for a single local element that is bisected by a global knot line at parametric position x0, using the paper's fixed 3x3 Gauss rule and a reference high-order subdivided quadrature (e.g., 20-point Gauss on each sub-element). Sweep x0 across the local element, including x0 near a Gauss point and near the element edge, and record the maximum relative error in any K_GL entry. If the max relative error exceeds 1%, the fixed quadrature is not uniformly accurate for arbitrary overlaps, and the paper's central claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that standard 3x3 (2D) or 3x3x3 (3D) Gauss quadrature over each local element be accurate even when the local element crosses global IGA element boundaries. The paper's argument establishes continuity of the coupling integrand, but for the stiffness-type coupling matrices in Eqs. (46)-(47) the integrand contains first derivatives of the quadratic B-spline global basis. At a global knot with multiplicity 1, that derivative is continuous but not differentiable, so the integrand has a kink inside the local element. Fixed Gauss-Legendre quadrature is exact only for polynomials on the whole element; an internal kink introduces algebraic quadrature error that depends on the overlap position, and no a priori error estimate is provided. The verification in Sections 4 and 5 is empirical, and the recommended parameter conditions (r_GL=4, a_L=2.5h_G, l_L=1.2h_G or sqrt(2)h_G, H_L=1.8h_G) are tuned using the same benchmarks. If some local-element/global-knot configuration places the kink unfavorably relative to the Gauss points, the evaluated DSIF or local stress could degrade even though the integrand is continuous. This would invalidate the central efficiency claim that recursive subdivision can be omitted entirely. The paper does not test such worst-case alignments, and no code or data are provided to probe them independently.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a hybrid s-version isogeometric strategy (hS-IGA) for dynamic crack propagation analyses, in which the global mesh is discretized with quadratic B-spline basis functions while the local crack-domain mesh retains standard Lagrange basis functions. The motivation is that the global B-spline basis removes the derivative discontinuities that, in the conventional Lagrange-based s-method, make the global-local coupling integrands discontinuous inside local elements and force recursive subdivision for numerical integration. The formulation is presented in Sections 2 and 3, and verification is carried out on two-dimensional stationary and dynamic straight-crack benchmarks and three-dimensional stationary and dynamically propagating circular-crack benchmarks. The results are compared with the standard FEM, the conventional s-method, and available analytical or high-fidelity FEA reference solutions. The headline quantitative claims are that the proposed method reduces the number of integration points required for coupling integration by approximately 81% in the two-dimensional dynamic benchmark and 95.6% in the three-dimensional dynamic benchmark relative to the conventional s-method, while preserving comparable accuracy in the dynamic stress intensity factor and local stress.","tokens_in":55281,"tokens_out":4407,"duration_ms":48995,"significance":"The proposed method is potentially valuable: it combines the modelling advantages of the s-method (independent global and local meshes, localized refinement, natural crack representation by Lagrange elements) with the smoother global approximation offered by IGA. If the central claim holds, the method removes a serious practical bottleneck of the conventional s-method without requiring a full IGA local discretization, which would be attractive for fracture-oriented global-local simulations. The verification is extensive and, on the reported benchmarks, convincing: convergence studies are shown for displacement error, the fine-mesh FEA reference histories are checked against Broberg's analytical DSIF and local-stress solutions with maximum errors below 0.9% and 2.6%, and comparisons against both the conventional s-method and standard FEM are systematic. The paper also makes several honest limitations explicit, including restriction to planar linear-elastic crack problems.","major_comments":[{"comment":"The continuity argument for the coupling integrands establishes only C0 continuity, not the polynomial smoothness that fixed Gauss quadrature would need for exactness. For quadratic B-splines with single-multiplicity interior knots, the first derivatives entering B_G are continuous but piecewise linear, so a stiffness coupling integrand containing B_G^T D B_L has a kink (a discontinuity in its derivative) when a local element crosses a global knot. Fixed 3x3 or 3x3x3 Gauss-Legendre quadrature over that local element has no a priori error estimate for such an integrand, and the error depends on where the kink falls relative to the Gauss points. The verification in Sections 4 and 5 is empirical and is performed on the same benchmark families used to select the recommended mesh conditions. I recommend adding a quadrature-convergence study for representative worst-case configurations (e.g., a local element whose global-knot kink is near a Gauss point, and a local element spanning a global knot on the crack path) or providing an explicit quadrature error estimate; without this, the strong claim that recursive subdivision can be omitted entirely is not fully supported for problems outside the calibrated benchmark set.","section":"§3.2.2, Eqs. (46)-(47)"},{"comment":"The recommended parameter conditions (r_GL=4, a_L=2.5h_G, l_L=1.2h_G in 2D and sqrt(2)h_G in 3D, H_L=1.8h_G) are obtained from systematic sweeps on the same benchmarks that are later used to demonstrate accuracy. Consequently, the quoted accuracy ranges (within 1.8% for DSIF and 4.0% for local stress in 2D; within 1.3% and 2.5% in 3D) are partly a measure of how well the method fits the calibration benchmarks rather than of predictive performance on new problems. I recommend adding at least one out-of-sample configuration (a different domain geometry, crack-front curvature, or loading history) or a sensitivity analysis around the recommended parameters, so that the claimed generality of the method is not conflated with benchmark-specific tuning.","section":"§4.2.2 and §5.2.2"},{"comment":"Computational efficiency is assessed exclusively by counting integration points, not by wall-clock time. The hS-IGA method replaces recursive subdivision with a per-point global parametric mapping (Newton-Raphson iteration) and a global-element identification step; these operations also have a cost. A reduction in integration-point count is therefore not automatically equivalent to a proportional reduction in assembly time. The wording in the abstract and conclusions ('computationally efficient', 'computational cost') should be qualified either by adding measured assembly or total solve times or by explicitly stating that integration-point count is used as a proxy for assembly cost.","section":"§4.2.2, §5.2.2, Figs. 33 and 49"}],"minor_comments":[{"comment":"The L-shaped convergence curves under fixed h_L are described but not explained; a sentence interpreting why the global mesh begins to limit the attainable accuracy would help the reader.","section":"§4.1.2, Fig. 18"},{"comment":"The relative L2 error norm is written as a ratio of integral expressions; please check that the notation matches the intended definition of a norm ratio and clarify the denominator in the text.","section":"Eq. (54)"},{"comment":"The integration-point reductions of 81% and 95.6% are quoted without stating the reference model (the conventional s-method with the same global-to-local element-size ratio). This context is given later in the text and should be stated in the abstract as well.","section":"Abstract and §4.2.2, §5.2.2"},{"comment":"No data or code availability statement is included; providing the meshes, parameter sets, and reference FEA histories would materially improve reproducibility of the verification results.","section":"General"},{"comment":"Several convergence and parameter-sweep figures use similar marker styles and are difficult to read in grayscale; increasing marker size and using distinct line styles would improve clarity.","section":"Figs. 16, 26-29, 43-46"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope and the verification effort is substantial. The main technical risk is the quadrature claim: the paper shows that the coupling integrand is continuous, but fixed Gauss quadrature on a kinked, non-polynomial integrand needs either an a priori estimate or a dedicated worst-case convergence study. The benchmark-derived mesh conditions and the absence of wall-clock timing are secondary but should be addressed. I do not see grounds for rejection; the concerns are fixable within revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid computational mechanics paper with an honest incremental advance. The hybrid split—B-splines in the global mesh, Lagrange in the local crack mesh—is genuinely new relative to the full-IGA S-IGA and the higher-continuity s-version it cites. The motivation is sound: the coupling-integrand bottleneck in the conventional s-method comes from the global side, so fixing the global basis removes the need for recursive subdivision while keeping the local Lagrange basis for crack representation and post-processing.\n\nWhat it does well: verification is extensive and honestly benchmarked. Two-dimensional stationary convergence against FEM and the s-method, 2D dynamic against Broberg with reported errors below 0.9% for DSIF and 2.6% for local stress, 3D stationary against Sneddon, and 3D dynamic against an axisymmetric FEA reference. The parameter sweeps over crack velocity are thorough. The integration-point reductions of roughly 81% in 2D and 95.6% in 3D are consistent with the proposed mechanism, and the paper does not oversell them as wall-clock savings.\n\nThe soft spots are real but not fatal. Efficiency is counted in integration points, not actual time, so the headline claim is a proxy for cost. No code or data is provided, which makes independent probing harder. The recommended mesh conditions (r_GL=4, a_L=2.5h_G, l_L=1.2h_G or sqrt(2)h_G, H_L=1.8h_G) are tuned on the same benchmarks; that is mild fitting, not circularity, because the evaluated quantities are validated against independent analytical or reference solutions. The largest theoretical gap is exactly what the stress-test note flags: the coupling integrand is C0 but not C1 at global knots, so fixed 3x3 Gauss quadrature has an algebraic error that depends on where the kink falls inside the local element. The paper demonstrates accuracy only empirically. I do not think this breaks the central claim—the benchmarks cover many overlap configurations and the errors are small—but an a priori error estimate or a worst-case-alignment test would make the efficiency claim transferable to new problems.\n\nThis paper is for computational fracture mechanics researchers who use overlay methods and for IGA practitioners interested in coupling integration. It deserves a serious referee: the method is clearly formulated, the verification is broad, and the advance is real even if incremental. Send it to review, with the request that the authors add a discussion of quadrature error for non-smooth integrands and, if possible, wall-clock comparisons against the conventional s-method.","headline":"A solid, honest incremental advance: B-spline global basis removes the s-method's coupling-integration bottleneck, with strong benchmark verification; the main gap is that quadrature robustness is shown empirically, not theoretically.","tokens_in":55848,"tokens_out":1628,"would_cite":true,"duration_ms":17133,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74S05","65N30","74R10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Replacing only the global mesh with B-splines makes global–local coupling integrands continuous, so standard Gauss quadrature suffices and cuts dynamic crack-analysis integration points by roughly 81% in 2D and 95.6% in 3D benchmarks.","keywords":["hybrid s-version isogeometric analysis","dynamic crack propagation","global-local coupling","B-spline basis functions","Lagrange local mesh","dynamic stress intensity factor","local stress","Gauss quadrature"],"falsifier":"Compute the coupling stiffness and mass entries for a local element whose support intersects a global knot span and compare the fixed 3×3 or 3×3×3 Gauss result against a highly subdivided high-order reference quadrature; if the relative error exceeds the level needed to keep the dynamic stress intensity factor and local stress within the reported few percent for some valid overlap configuration, the efficiency claim would not transfer outside the tested benchmarks.","tokens_in":54760,"feed_emoji":"💥","tokens_out":6177,"duration_ms":56233,"temperature":0.7,"pith_summary":"The paper proposes a hybrid s-version isogeometric analysis (hS-IGA) for dynamic crack propagation: B-spline basis functions in the coarse global mesh, Lagrange elements in the local crack mesh. The B-spline global basis removes the derivative discontinuities that, in the conventional Lagrange-based s-method, force recursive subdivision of integration domains. As a result, the coupling terms can be evaluated with standard 3×3 or 3×3×3 Gauss quadrature, preserving the global–local modelling advantages while cutting integration points by approximately 81% in the two-dimensional dynamic benchmark and 95.6% in the three-dimensional dynamic benchmark. The paper verifies this on stationary and propagating straight cracks in 2D and circular cracks in 3D, showing accurate dynamic stress intensity factors and local stresses, comparable degrees-of-freedom efficiency to the s-method, and reduced sensitivity near the local-domain boundary.","feed_headline":"Hybrid IGA trims crack-propagation integration points by up to 95.6%","feed_subtitle":"Smoother global basis removes recursive subdivision in global–local coupling while keeping Lagrange local crack meshes.","key_machinery":"The load-bearing object is the hybrid basis assignment: quadratic B-spline basis functions in tensor-product form in the global IGA mesh, and Lagrange basis functions in the local mesh. The B-splines' higher inter-element continuity makes the first derivatives of the global basis, and hence the coupling integrands in the stiffness and mass coupling matrices $\\mathbf{K}_{GL}$, $\\mathbf{K}_{LG}$, $\\mathbf{M}_{GL}$, $\\mathbf{M}_{LG}$, continuous inside local elements. This makes fixed $3\\times 3$ or $3\\times 3\\times 3$ Gauss quadrature accurate without recursive subdivision, while the Lagrange local mesh keeps the crack as a $C^0$-but-not-$C^1$ line and permits nodal force release, moving local mesh updates, domain-integral $J$ evaluation, and nodal-force-based local stress post-processing.","core_discovery":"The central discovery is that the global approximation's inter-element continuity, not the local mesh, is the cause of the s-method's coupling-integration bottleneck. Replacing the global Lagrange basis with quadratic B-splines, which are $C^1$ continuous across knot spans, makes the coupling integrands continuous even when a local element spans several global elements. The local mesh remains Lagrange-based, which naturally represents the $C^0$-but-not-$C^1$ crack-tip displacement field and supports the nodal-force-release technique, local mesh update, and fracture post-processing. With this split, standard Gauss quadrature suffices without recursive subdivision; benchmark comparisons against Broberg's and Sneddon's solutions show accuracy comparable to conventional FEM and the s-method while integration points drop by roughly 81% in 2D dynamic propagation and 95.6% in 3D dynamic propagation.","pith_inferences":["If the continuity argument is general, the same hybrid basis assignment could reduce integration cost in any superposition or overlay method with non-matching meshes, not only fracture, by using spline bases on the coarser background mesh.","The empirical parameter sweeps suggest a quadrature-accuracy threshold; a systematic error estimate for Gauss quadrature of products of B-splines and Lagrange functions would show whether the savings transfer to curved crack fronts, elastoplasticity, or adaptive local meshes.","The paper restricts verification to linear-elastic planar cracks; the extension to 3D elastoplastic fracture that the conclusion mentions would test whether the continuity benefit survives material nonlinearity and more complex crack-front geometry."],"forward_implications":["If the central claim is correct, dynamic crack propagation analyses can retain the s-method's degrees-of-freedom advantage while eliminating the coupling-integration bottleneck, with roughly 81% fewer integration points in the 2D dynamic benchmark and 95.6% fewer in the 3D dynamic benchmark.","The near-crack fracture quantities, dynamic stress intensity factor and local stress, match analytical or reference solutions within a few percent in both 2D and 3D, including a wide crack-velocity range from 200 to 1500 m/s.","The accuracy of local stress becomes governed mainly by the local element size $h_L$ rather than the global element size $h_G$, removing the mesh-interface sensitivity near $\\Gamma_{GL}$ that appears in the conventional s-method.","The parameter conditions identified from the benchmark sweeps, such as $r_{GL} \\geq 4$, $a_L = 2.5 h_G$, $l_L = 1.2 h_G$ (or $\\sqrt{2} h_G$ in 3D), and $H_L = 1.8 h_G$, provide a starting recipe for configuring future hS-IGA crack analyses."],"supporting_citations":[{"why":"Supplies the two-dimensional s-method dynamic crack propagation baseline, benchmark setup, and previous parameter conditions against which hS-IGA is verified.","marker":"[12]"},{"why":"Provides the three-dimensional s-method framework, the local-stress evaluation procedure, and the reference FEA used for dynamic boundary conditions.","marker":"[13]"},{"why":"Introduces isogeometric analysis and the higher-continuity spline basis that the proposed method exploits in the global mesh.","marker":"[20]"},{"why":"Describes the full IGA overlay formulation (S-IGA) that the paper contrasts with its hybrid choice of keeping a Lagrange local mesh.","marker":"[26]"},{"why":"Documents the recursive subdivision technique for coupling integration that the new method avoids, forming the cost baseline.","marker":"[19]"},{"why":"Defines the original s-version FEM concept of superimposed meshes on which the global–local superposition framework rests.","marker":"[28]"},{"why":"Provides Broberg's analytical solutions for the dynamic stress intensity factor and local stress used as reference truth in the dynamic benchmarks.","marker":"[48]"},{"why":"Gives Sneddon's exact solution for the stationary circular crack, used to set boundary conditions and verify the 3D stationary results.","marker":"[58]"},{"why":"Establishes the nodal force release technique and Rayleigh damping stabilization adopted for dynamic crack propagation in the local mesh.","marker":"[38]"},{"why":"Justifies the $p+1$ Gauss quadrature rule used for the quadratic B-spline global basis in both 2D and 3D.","marker":"[21]"}],"fun_headline_variants":["Hybrid IGA cuts crack integration points up to 95.6% in 3D","Smooth global basis removes recursive crack coupling in IGA","95.6% fewer integration points: hybrid IGA for dynamic cracks","B-splines globally, Lagrange locally: lean dynamic crack IGA","Hybrid s-IGA: B-spline global mesh breaks coupling bottleneck"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The assumption that fixed 3×3 or 3×3×3 Gauss quadrature over local elements that span several global IGA elements integrates the coupling integrands accurately enough; the paper demonstrates this on the chosen benchmarks but provides no quadrature-error bound, and the recommended mesh parameters are derived from the same benchmark sweeps.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid IGA cuts crack integration points up to 95.6% in 3D","Smooth global basis removes recursive crack coupling in IGA","95.6% fewer integration points: hybrid IGA for dynamic cracks","B-splines globally, Lagrange locally: lean dynamic crack IGA","Hybrid s-IGA: B-spline global mesh breaks coupling bottleneck"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000981,"raw_usage":{"total_tokens":4210,"prompt_tokens":1033,"completion_tokens":3177,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":649,"completion_tokens_details":{"reasoning_tokens":3078}},"tokens_in":649,"tokens_out":3177,"duration_ms":24191,"temperature":1.0,"reasoning_tokens":3078,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T05:57:31.568419+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the coupling stiffness and mass entries for a local element whose support intersects a global knot span and compare the fixed 3×3 or 3×3×3 Gauss result against a highly subdivided high-order reference quadrature; if the relative error exceeds the level needed to keep the dynamic stress intensity factor and local stress within the reported few percent for some valid overlap configuration, the efficiency claim would not transfer outside the tested benchmarks.","supporting_citations":[{"cited_title":"Kishi, Y","cited_arxiv_id":null,"evidence_quote":"Supplies the two-dimensional s-method dynamic crack propagation baseline, benchmark setup, and previous parameter conditions against which hS-IGA is verified."},{"cited_title":"Shibanuma, K","cited_arxiv_id":null,"evidence_quote":"Provides the three-dimensional s-method framework, the local-stress evaluation procedure, and the reference FEA used for dynamic boundary conditions."},{"cited_title":"Tsuchiyama, Y","cited_arxiv_id":null,"evidence_quote":"Describes the full IGA overlay formulation (S-IGA) that the paper contrasts with its hybrid choice of keeping a Lagrange local mesh."},{"cited_title":"Okada, C.T","cited_arxiv_id":null,"evidence_quote":"Documents the recursive subdivision technique for coupling integration that the new method avoids, forming the cost baseline."},{"cited_title":"Fish, The s-version of the finite element method, Comput","cited_arxiv_id":null,"evidence_quote":"Defines the original s-version FEM concept of superimposed meshes on which the global–local superposition framework rests."},{"cited_title":"Bertram Broberg, Cracks and Fracture, Elsevi er, 1999","cited_arxiv_id":null,"evidence_quote":"Provides Broberg's analytical solutions for the dynamic stress intensity factor and local stress used as reference truth in the dynamic benchmarks."},{"cited_title":"Yanagimoto, K","cited_arxiv_id":null,"evidence_quote":"Establishes the nodal force release technique and Rayleigh damping stabilization adopted for dynamic crack propagation in the local mesh."}],"review_version":1}