{"id":"2d9a7111-d5e7-4e3e-88b9-8c835aae158c","arxiv_id":"2607.23003","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A J-integral implementation in a three-dimensional quasicontinuum method is validated against linear-elastic theory and virtual crack extension, with a Monte Carlo sensitivity test of the integration domain.","lead":"This paper adds the J-integral, a standard measure of how strongly a crack is driven to grow, to the quasicontinuum method and validates it against two established numerical approaches. The main draw is a practical tool: large multiscale fracture simulations can now report a crack-driving force with a domain-quality check.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Inner q=1 contour placed at the A/C interface means J samples elements perturbed by ghost-force corrections; Regime-3 validation is only mutual J/VCE consistency, not an independent measure of G.","rationale":"The reader identified as weakest assumption the claim that a domain integral evaluated only in the continuum region, with q=1 on the atomistic/continuum interface, measures the energy release rate of the full coupled system. My stress-test converges on the same point, with a sharper mechanism: ghost-force corrections at the interface act as external dead loads, so the Eshelby tensor is not necessarily divergence-free in the elements adjacent to the inner contour, and the standard proof of Rice path-independence does not apply exactly. The manuscript's own honesty about Regime 2 being only a code-verification step and about the lack of a closed-form solution in Regime 3 supports the need for a direct test. Because the concern is precisely the one that motivated the CONDITIONAL verdict, and because the proposed checks are straightforward and well within the authors' computational reach, I recommend no change to the reader's verdict. The concern is not fatal: it is a missing validation, not a demonstrated contradiction, and the paper's numerical claims remain plausible. No ad hominem or manufactured issue is involved; the critique targets a specific technical assumption and a specific missing comparison.","tokens_in":21102,"tokens_out":6266,"duration_ms":76331,"concrete_test":"On the same relaxed configuration used in Fig. 10, recompute J with the inner q=1 contour moved outward by one, two, and five element layers from the atomistic/continuum interface, keeping the outer contour and mesh fixed. If J changes by more than the MCMC-reported standard deviation (0.0006 J/m^2 at K_I = 0.9 MPa m^{1/2}), the integration domain is contaminated by interface/ghost-force effects and the central claim is not supported. As a stronger cross-check, build a fully atomistic reference model of the same small geometry (same K-field boundary conditions, same Stillinger-Weber potential), relax it, and compute J via Hardy estimates or atomic summation; compare directly with the QC domain-integral J.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 sets the inner contour (q=1) of the J-integral domain exactly on the atomistic/continuum interface, so the gradient of q samples elements adjacent to that interface. In QC, the interface carries ghost-force corrections f_ghost (Eq. 10), which act as external dead loads. At a relaxed state these are balanced only at repatom nodes; they do not imply div P = 0 in the continuum sense, and therefore the Eshelby tensor is not guaranteed to be divergence-free in the very annulus that contributes to the domain integral. Rice path-independence, which is the premise for equality J = G of the full coupled system, can then fail. Section 4.5 does not close this gap: the VCE comparison shares the same q-field and the same QC energy infrastructure; agreement with LEFM in Regimes 1 and 2 involves neither the interface nor relaxation; and the three-point-bending match to 2γs (Table 1) is a single point computed with the same contested J. The central claim therefore rests on an unverified path-independence assumption across a region whose equilibrium is perturbed by ghost forces. The paper is honest about the absence of a closed-form solution in Regime 3, but does not supply the direct numerical check needed to establish the claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an implementation of the domain-form J-integral in the three-dimensional quasicontinuum (QC3D) method. The J-integral is evaluated from Cauchy–Born continuum fields (stress and strain energy density) in the local (continuum) region of a relaxed QC model, with the integration domain bounded internally by the atomistic/continuum interface. Validation is organized in three regimes: (1) linear elasticity, where the anisotropic LEFM K-field is prescribed and Hooke's law is used; (2) the same prescribed K-field evaluated through the nonlinear Cauchy–Born relation without relaxation; and (3) the same nonlinear relation with full atomic relaxation. In each regime the J-integral is compared with the virtual crack extension (VCE) method; agreement is exact in Regime 1 and within about 3% in Regime 3. The paper also introduces an MCMC sampling scheme over admissible q-fields to quantify sensitivity of the computed J to the choice of interpolation function, and demonstrates the method on a three-point bending test of silicon, reporting J_c = 0.1886 J/m^2 versus 2γ_s = 0.1814 J/m^2 (3.85% error).","tokens_in":21300,"tokens_out":4832,"duration_ms":55604,"significance":"If the central claim holds, this is a valuable contribution: it would allow users of multiscale codes such as QC3D to extract a standard fracture mechanics parameter (the energy release rate) directly from continuum fields in the local region, bridging atomistic crack-tip processes and continuum driving forces. The validation hierarchy is thoughtfully designed: Regime 1 is a genuine exact benchmark, Regime 2 is honestly labeled as code verification rather than physics validation, and the authors state plainly that no closed-form solution exists for Regime 3. The MCMC uncertainty quantification is novel and the detailed-balance proof in Appendix A is a useful addition. The paper is also transparent about the limitations of the interatomic potential and the quasistatic plane-strain scope. However, the load-bearing assertion—that the domain integral evaluated in the continuum region measures the energy release rate of the fully coupled atomistic/continuum system—is not independently established; the Regime 3 validation relies on consistency with VCE, which shares the same q-field and the same energy functional.","major_comments":[{"comment":"The inner contour (q=1) is placed exactly on the atomistic/continuum interface, so the integrand samples the annulus of elements adjacent to that interface. In QC, ghost forces f_ghost act as external dead loads at the interface. A relaxed QC state balances these forces only at repatom nodes; it does not imply div P = 0 in the continuum sense, and therefore the Eshelby tensor is not guaranteed divergence-free in the region that contributes most to the domain integral. Rice's path independence, which is the premise for J = G of the full coupled system, is thus an unverified assumption. The Regime 3 validation does not close this gap because VCE uses the same q-field and the same QC energy functional. A direct numerical check is needed: compute J for a family of inner contours moved progressively into the continuum (away from the interface) and show convergence; or compare with a global en","section":"Sec. 3.1, Eq. (14); Sec. 2.2, Eq. (10)"},{"comment":"The claim that VCE provides an 'independently constructed' measure of G is overstated. By the deLorenzi equivalence cited by the authors, VCE reduces to the domain J-integral for elastic equilibrium, and the implementation here uses the same q-field and the same energy functional as the J computation. The 3% agreement in Regime 3 is therefore primarily a consistency check between two implementations, not a validation against an external reference. The only external anchor in Regime 3 is the LEFM limit at small loads, which is insensitive to interface and ghost-force effects. The paper should either soften the wording or supply an independent benchmark, e.g., a fully atomistic simulation of a small model, or an energy-vs-crack-length curve from multiple relaxed simulations with actual crack advance.","section":"Sec. 4.2, Sec. 4.5, Fig. 10"},{"comment":"The three-point bending comparison to 2γ_s is a useful end-to-end sanity check, but it is a single data point and J_c is computed with the same contested domain integral. The surface energy γ_s is not defined or computed in the manuscript (cleavage plane, potential, methodology), so the reader cannot assess whether the 3.85% agreement is meaningful. This check does not address path independence across the atomistic/continuum interface. Please provide details of the γ_s calculation and, if possible, report the MCMC sensitivity for the bending configuration.","section":"Sec. 6, Table 1"}],"minor_comments":[{"comment":"The text says 'as in Region 2' where 'Regime 2' is meant. Please make terminology consistent.","section":"Sec. 4.5"},{"comment":"The display equations have unbalanced parentheses/brackets; please check the formatting, especially the summation terms.","section":"Eq. (18), Eq. (20)"},{"comment":"The sentence 'all three quantities drop suddenly together as the crack propagates' is ambiguous: J is normally defined for a stationary crack. Clarify whether J is reported for the last stable configuration before cleavage and what the post-drop values represent.","section":"Sec. 4.5, Fig. 10"},{"comment":"The MCMC standard deviation is described as providing 'rigorous error bars' in Sec. 5.2. Since the target distribution is uniform over q-fields and the chain is generated with all proposals accepted, the result is a sensitivity measure, not a rigorous statistical error bar on J. Please temper the wording.","section":"Sec. 5"},{"comment":"No data availability or code availability statement is included. Given the reproducibility emphasis of the field, please add one.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is honest and technically careful in many respects, and the central idea is sound. The main weakness is that the Regime 3 validation does not independently establish the equality J = G for the relaxed QC system, because VCE shares the same q-field and energy functional and the inner contour sits on the ghost-force-perturbed interface. The recommended additional tests—inner-contour shift and an independent energy-derivative calculation—are feasible within the scope of the manuscript and would substantially strengthen the claim. I would be willing to see a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first domain-integral J implementation inside QC3D, and it is a useful, honestly validated piece of engineering. It deserves a serious referee, but not unconditional acceptance as it stands.\n\nWhat's new: the J domain integral is assembled from Cauchy-Born continuum fields in the local region of a relaxed QC3D model, which nobody had done before; the VCE adaptation that runs two relaxed simulations is also new; and the MCMC sampling over q-fields gives users a practical diagnostic for integration-domain quality. The staging of the validation is the paper's real strength. Regime 1 is a true independent LEFM benchmark. Regime 2 is explicitly labeled a code-verification step, with the authors noting that J/VCE agreement there follows from calculus alone. Regime 3 is the intended operating point, and J and VCE agree within 3% up to cleavage. The three-point bending demonstration, matching Griffith's criterion within 4% while cutting the atom count by a factor of 34, makes the utility concrete.\n\nThe soft spots are real but addressable. The main one concerns the inner q=1 contour, placed exactly on the atomistic/continuum interface. Ghost forces are balanced only at repatom nodes, so the Eshelby tensor is not guaranteed divergence-free in the annulus that contributes most to the domain integral. Rice path independence is precisely what would justify J = G for the coupled system, and the paper never checks it directly. The VCE comparison does not fully close that gap: VCE uses the same q-field, the same QC energy, and the same interatomic potential, so it is mutual consistency rather than independent confirmation. A direct path-independence test, or an atomistic J computed from the same geometry, would settle it. These are addressable rather than fatal; I don't think the central result is wrong.\n\nSecond, the claimed \"uncertainty quantification\" is oversold. The MCMC procedure samples an arbitrary uniform distribution over q-fields; the resulting standard deviation is a sensitivity measure for the integration domain, not a statistical uncertainty on the fracture driving force. The detailed-balance argument is fine but trivial, since every proposed move is accepted under a uniform target. Third, no code or data are shipped, so the numbers cannot be independently reproduced from the preprint.\n\nWho this is for: QC users and people doing multiscale fracture who need a standard driving-force postprocessor and a mesh-quality diagnostic. I would send it to peer review and ask for the path-independence check, code/data release, and more careful UQ language. Reading group: maybe. I would cite it if I worked in this area.","headline":"Useful first QC3D J-integral implementation with honest staged validation, but the central path-independence claim needs a direct check and the UQ label oversells the MCMC sensitivity analysis.","tokens_in":21917,"tokens_out":3330,"would_cite":true,"duration_ms":34483,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74R10","74S05","65C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the J-integral evaluated only in the continuum region of a relaxed quasicontinuum model equals the coupled system's energy release rate, matching virtual crack extension within 3% and Griffith's criterion within 4%.","keywords":["J-integral","quasicontinuum method","energy release rate","Cauchy-Born rule","Griffith criterion","uncertainty quantification","Markov chain Monte Carlo","fracture mechanics"],"falsifier":"Run the same relaxed quasicontinuum fracture test with the inner integration contour moved into the atomistic region, or with the domain substantially enlarged or shrunk, and check whether J shifts by more than the MCMC-reported standard deviation; alternatively, compute an independently constructed atomistic J-integral on a model with a much larger atomistic region fully enclosing the crack tip and compare — if either test gives a discrepancy beyond the quoted ~3%, the claim that the continuum-only domain integral equals the true energy release rate fails.","tokens_in":20844,"feed_emoji":"💥","tokens_out":8446,"duration_ms":74961,"temperature":0.7,"pith_summary":"The paper sets out to prove that the J-integral — the standard fracture-mechanics measure of energy release rate — can be computed in a multiscale atomistic/continuum model using only the coarse-grained continuum region, and that the result equals the true energy release rate of the coupled system. If that holds, fracture toughness can be extracted from a single relaxed multiscale configuration, even though the atomistic crack-tip core is excluded from the integration domain. The authors validate this in three escalating regimes: small-strain linear elasticity against exact LEFM, nonlinear Cauchy-Born response against virtual crack extension, and fully relaxed nonlinear response where J and VCE agree to within about 3% at the point of fracture instability. A three-point bending test of silicon then shows the critical J reproducing the Griffith energy balance within 3.85%. They also introduce a Markov chain Monte Carlo sampler over the J-integral's q-field that gives statistical error bars on J and, as the paper argues, applies to conventional finite element fracture computations as well.","feed_headline":"Continuum-only J-integral predicts fracture energy within 4%","feed_subtitle":"A coarse-grained fracture integral captures atomistic crack behavior, with MCMC error bars for any mesh.","key_machinery":"The load-bearing machinery is the volume form of the J-integral, written in terms of the Eshelby stress tensor — an energy-momentum quantity defined from the strain energy density and the stress times displacement gradient — and summed element by element over linear tetrahedra in the continuum region. The scalar weight function q of the classical domain integral is promoted to a vector field pointing along the crack direction, so the same q-values serve both the J computation and the nodal perturbation in the virtual crack extension validation. The Cauchy-Born rule makes the continuum stress and energy density inherit the interatomic potential, so the excluded atomistic core and the integrat","core_discovery":"The central claim is that a domain-integral form of the J-integral — evaluated entirely in the continuum region of a three-dimensional quasicontinuum model, using stress and strain energy density derived from the interatomic potential through the Cauchy-Born rule — correctly measures the energy release rate of the full atomistic/continuum system. The integration domain is bounded internally by the atomistic/continuum interface, so the atomistic crack tip is excluded, and the weighting function q is set to unity on that interface and zero on the outer boundary. In a fully relaxed, finite-strain silicon model under mode I loading, the continuum-domain J and an independently constructed virtual","pith_inferences":["The vector-q formulation for an arbitrary crack direction suggests a natural route to mixed-mode and curved crack fronts via piecewise integration along straight segments, an extension the paper leaves implicit.","The MCMC spread could be repurposed as an adaptive-mesh criterion: large local sensitivity of J to q would flag under-resolved regions and drive refinement or domain repositioning automatically.","Because the integration domain stops at the atomistic/continuum interface, residual ghost forces there could contribute a small systematic error to J; comparing against a ghost-force-free formulation or a much larger atomistic region would isolate that contribution, which the paper does not do.","The ~3% J–VCE gap at the final stable load step is attributed to VCE's forward-difference error, but testing a central-difference VCE across several crack-extension sizes would show whether the residual is purely numerical or reflects physical lattice trapping."],"forward_implications":["J can be obtained from a single equilibrated multiscale configuration, avoiding the two-simulation cost of virtual crack extension at every load step.","The critical J at cleavage directly yields the material's fracture toughness for the interatomic potential, enabling direct comparison with the Griffith criterion and potential-driven toughness predictions.","The MCMC standard deviation serves as a quantitative certificate of integration-domain quality, usable in ordinary finite element codes for mesh-convergence verification.","Agreement across linear elastic, nonlinear elastic, and fully relaxed regimes means the method spans the practical range from LEFM-dominated remote loading to atomistically relaxed crack tips.","The three-point bending demonstration shows structural-scale simulations are feasible: 358k repatoms reproduce a model that would require about 12 million atoms fully atomistically."],"fun_headline_variants":["J-integral in quasicontinuum: crack energy with MCMC errors","Continuum-region J integral for multiscale fracture, uncertainty-quantified","QC3D J-integral validated across linear and nonlinear regimes","Fracture energy from continuum fields with quantified uncertainty","Multiscale J-integral: atomistic accuracy without atomistic integration"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that a J-integral evaluated only in the continuum region — with the atomistic crack-tip core excluded and the interface subject to ghost-force corrections — still equals the energy release rate of the entire coupled atomistic/continuum system; the paper tests this only indirectly through agreement with a second numerical method.","fun_headline_variants_meta":{"raw":{"variants":["J-integral in quasicontinuum: crack energy with MCMC errors","Continuum-region J integral for multiscale fracture, uncertainty-quantified","QC3D J-integral validated across linear and nonlinear regimes","Fracture energy from continuum fields with quantified uncertainty","Multiscale J-integral: atomistic accuracy without atomistic integration"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00026,"raw_usage":{"total_tokens":1464,"prompt_tokens":823,"completion_tokens":641,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":567,"completion_tokens_details":{"reasoning_tokens":546}},"tokens_in":567,"tokens_out":641,"duration_ms":6345,"temperature":1.0,"reasoning_tokens":546,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T03:52:05.141277+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same relaxed quasicontinuum fracture test with the inner integration contour moved into the atomistic region, or with the domain substantially enlarged or shrunk, and check whether J shifts by more than the MCMC-reported standard deviation; alternatively, compute an independently constructed atomistic J-integral on a model with a much larger atomistic region fully enclosing the crack tip and compare — if either test gives a discrepancy beyond the quoted ~3%, the claim that the continuum-only domain integral equals the true energy release rate fails.","supporting_citations":[],"review_version":1}