{"id":"e3b25761-6b8f-4052-bae0-b91e1f42af76","arxiv_id":"2607.09356","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"A first-order FE² homogenization of phase-field theory, with micro–macro links for both the order parameter and its gradient, matches DNS evolution of averaged fields on notched and holed specimens.","lead":"The paper derives a consistent two-scale homogenization theory for phase-field models by applying a Hill–Mandel condition to Gurtin’s microforces, then implements it as FE². It lets engineers evolve averaged microstructure fields on component-scale meshes without resolving every interface.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified beyond the reader's already-flagged RVE-size caveat.","rationale":"The reader’s weakest-assumption diagnosis is precise and already sufficient for a CONDITIONAL verdict. The theory is a clean first-order extension of classical Hill–Mandel homogenization to Gurtin microforces; the numerical examples demonstrate that the scheme is implementable and qualitatively predictive. Because the paper itself flags the need for “broader and more systematic investigations” of RVE size (§4) and supplies no quantitative error norms, the accuracy claim cannot yet be treated as fully established. No additional internal inconsistency or stronger empirical failure mode appears on a second pass, so the verdict and confidence level remain unchanged.","tokens_in":20966,"tokens_out":549,"duration_ms":7597,"concrete_test":"Re-run the notched Allen–Cahn FE² case of Fig. 6 with three RVE sizes (ℓ_Γ/ℓ_RVE = 0.05, 0.1, 0.2) under identical macroscopic mesh and initialization; report L2(¯η_FE2 – ¯η_DNS) and interface-arrival time at the notch root. If the error norms stay within a factor of two across the three sizes, the pragmatic window is adequate; if they grow systematically, the accuracy claim needs a quantitative RVE-selection criterion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Hill–Mandel microforce condition yields a well-posed RVE BVP with rigorous micro–macro maps for η and ∇η, and FE² recovers macro-averaged evolution vs DNS with “reasonable accuracy”) is internally consistent. The variational construction in §2.3 (Lagrangians (12),(19),(25), Euler–Lagrange (14)–(17), power equivalence (9) and definitions (22)) is standard and free of hidden contradictions; the three classical BCs all satisfy the same macrohomogeneity relations. Empirical support is qualitative visual agreement on two examples (§3.1.4, §3.2.3) under deliberately matched initializations and moderate scale separation. The only soft spot is exactly the one the reader already isolates: the pragmatic choice ℓ_Γ/ℓ_RVE=0.1 (§3.1.3, Fig. 5) whose coarsening kinetics are known to be scale-dependent, so that “reasonable accuracy” remains unquantified. No deeper load-bearing flaw (e.g., inconsistency of the microforce residual, failure of the DirectFE² thickness scaling, or breakdown of the heat-transfer analogy) is visible in the text.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript derives a first-order homogenization framework for phase-field models by enforcing a Hill–Mandel-type micro-homogeneity condition written in Gurtin’s microforces. From the associated Lagrangian with kinematic constraints on the averaged order parameter and its gradient, it obtains a well-posed RVE problem (uniform microtraction, linear, and periodic BCs) together with consistent micro–macro maps for η̄, K̄, π̄ and ξ̄, fully coupled to mechanics. The theory is implemented via DirectFE² in Abaqus (heat-transfer analogy) and illustrated on a pure Allen–Cahn model and a stress-driven martensitic transformation model, with visual comparison to fully resolved DNS on notched and plate-with-hole geometries. The authors conclude that the scheme recovers the spatial and temporal evolution of the macroscopically averaged fields with reasonable accuracy.","tokens_in":21359,"tokens_out":1074,"duration_ms":24081,"significance":"A rigorous scale-bridging theory for non-conserved phase-field evolution that retains the macroscopic gradient of the order parameter has been missing; prior work either treated only mechanical coupling (Kochmann et al.) or used asymptotic/heuristic treatments limited to linear fracture PDEs. The variational construction in §2.3 is clean, recovers the expected residual microforce and the three classical BC families, and correctly distinguishes zeroth- from first-order phase-field homogenization. If the accuracy claim holds under broader conditions, the framework supplies a legitimate route to MOR/ML surrogates at the RVE level and thereby to engineering-scale phase-field simulations. The dual-example validation and the explicit acknowledgment of scale-separation limitations are strengths.","major_comments":[{"comment":"The central empirical claim (abstract; §3–4) that FE² “reliably predict[s] … with reasonable accuracy” rests solely on visual comparison of contour plots (Figs. 6, 8, 12). No quantitative error measures (e.g., L2 or H1 norms of η̄ versus the DNS average, interface-location error, or transformation-fraction histories) are reported. Without such metrics the accuracy statement cannot be assessed, especially given the known sensitivity of coarsening kinetics to the averaging window.","section":"§3.1.4, §3.2.3, Figs. 6–8, 12"},{"comment":"Section 3.1.3 and Fig. 5 demonstrate that the macroscopic coarsening time depends strongly on the ratio ℓ_Γ/ℓ_RVE; the authors themselves note that “broader and more systematic investigations are required” (§4). Yet the full-structure FE² examples adopt the single pragmatic value 0.1 without a sensitivity study of the predicted macro fields. Because the RVE size is an essential free parameter of the theory (it sets the macroscopic length that can be resolved), the load-bearing claim that the chosen window still represents the DNS evolution needs at least a limited parametric check on the notched or plate-with-hole problems.","section":"§3.1.3, Fig. 5; §4"}],"minor_comments":[{"comment":"The shared stochastic initialization between DNS and every RVE is methodologically fair for a controlled comparison, but should be stated more prominently as a deliberate validation choice rather than a general predictive setting.","section":"§3.1.1, §3.1.4"},{"comment":"Notation for the macroscopic gradient switches between K̄ and ∇_X η̄; a single consistent symbol would improve readability.","section":"§2.3, Eq. (8)"},{"comment":"The computational-cost comparison (DNS 20 600 s vs FE² 4100 s) is useful; reporting the corresponding wall-clock ratio for the pure Allen–Cahn notched example (where scale separation is lower) would complete the picture.","section":"§3.2.3"},{"comment":"A few typographical issues remain (e.g., “andvalidatedagainst”, missing spaces after commas in the abstract and keywords).","section":"Abstract, Keywords"},{"comment":"Figure 7 caption refers to “final state t/τ = 428” but does not state the macroscopic mesh sizes used; adding element counts or h/ℓ would make the convergence claim self-contained.","section":"Fig. 7"}],"recommendation":"major_revision","confidential_remarks":"The theoretical contribution is genuine and clearly distinguished from Kochmann et al. and from the fracture-homogenization literature. The two major comments are fixable within a revision (add error norms and a short RVE-size sensitivity on one of the existing geometries). I would not reject on novelty or correctness grounds; the manuscript is a solid candidate once the accuracy claim is quantified."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The real advance here is the first-order theory: they write a Hill–Mandel condition on Gurtin’s microforce power, pull out the macroscopic averages for both η and ∇η, and get a well-posed RVE problem with the three classical BCs (uniform microtraction, linear, periodic) all falling out of the same Lagrangian. Kochmann et al. is correctly called zeroth-order because it dropped the macro gradient; this paper keeps it and shows the residual microforce that appears in the RVE balance. That derivation in §2.3 is clean and standard variational work—no hidden contradictions.\n\nImplementation is DirectFE2 with the heat-transfer analogy, which is pragmatic and works. The two examples (pure Allen–Cahn and the Rajendran martensite model) show that the macro-averaged fields track the DNS patterns and rough timing under matched initializations. They are honest that coarsening speed depends on ℓ_Γ/ℓ_RVE (Fig. 5) and that their choice of 0.1 is pragmatic; the conclusions correctly flag the need for better RVE guidelines.\n\nSoft spots are exactly the ones the reader flagged and no worse: “reasonable accuracy” is visual only, no L2 norms or quantitative error tables, and under the moderate scale separation they used the wall-clock gain is modest (factor of five on the plate-with-hole). That does not break the central claim; it just means the paper is a methods foundation rather than a finished production tool. Citations engage the right literature and do not over-claim.\n\nThis is for people who already do FE2 or phase-field multiscale work and want a thermodynamically consistent way to carry the order-parameter gradient. The math is solid, the data are open-loop DNS comparisons with literature parameters, and the thinking is clear. I would send it to peer review; the theory section alone is worth having in the literature. Engage with it if you care about scale-bridging microstructure evolution.","headline":"Solid first-order phase-field homogenization via Gurtin microforces; theory is clean, DNS agreement is real but only qualitative, and RVE sizing remains the open practical issue.","tokens_in":21904,"tokens_out":509,"would_cite":true,"duration_ms":15828,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A Hill–Mandel condition written with Gurtin microforces yields a well-posed RVE problem and rigorous micro–macro links for both the order parameter and its gradient, so FE² can track macro phase evolution without resolving every interface.","keywords":["phase-field theory","microforce theory","computational homogenization","multi-scale simulation","FE2","Allen-Cahn","martensitic phase transformation"],"falsifier":"Run the same notched-structure or plate-with-hole problem with identical initial patterns but systematically varied RVE-to-interface-thickness ratios; if the FE² macroscopic order-parameter history then diverges from the fully resolved DNS outside the authors’ chosen ratio of 0.1, the claimed predictive accuracy fails.","tokens_in":21867,"feed_emoji":"🔬","tokens_out":1001,"duration_ms":19525,"temperature":0.7,"pith_summary":"Phase-field models capture microstructure evolution well, yet resolving every diffuse interface on an engineering component is almost always too expensive. This paper supplies the missing consistent homogenization: by requiring power equivalence between scales in the language of Gurtin microforces, it derives a well-posed boundary-value problem on each representative volume element and the corresponding micro–macro relations for the order parameter and its gradient. The resulting first-order theory is implemented as an FE² scheme and checked against fully resolved simulations for a minimal Allen–Cahn model and for stress-driven martensitic transformation. In both cases the macroscopic averaged fields evolve with the correct spatial pattern and temporal pace. A reader who needs microstructure-informed component predictions therefore gains a thermodynamically grounded route that keeps the fine-scale physics while discarding the need for a global fine mesh.","feed_headline":"FE² tracks phase evolution without meshing every interface","feed_subtitle":"A microforce Hill–Mandel condition gives RVE problems that match DNS averages for Allen–Cahn and martensite.","key_machinery":"The Hill–Mandel-type micro-homogeneity condition written with Gurtin microforces (internal power equality that identifies the macroscopic microstress and microforce). It simultaneously supplies the missing energetic link and the residual microforce that appears in the microscopic balance, thereby closing a first-order phase-field homogenization theory.","core_discovery":"Enforcing a Hill–Mandel-type condition of micro-homogeneity formulated in terms of Gurtin’s microforces produces a well-posed RVE problem whose three classical boundary conditions (uniform microtraction, linear order-parameter, periodic) all satisfy energetic equivalence; the associated micro–macro maps transfer both the order parameter and its gradient, and the resulting FE² scheme recovers the spatial and temporal evolution of the macroscopically averaged fields against direct numerical simulation with reasonable accuracy.","pith_inferences":["Because coarsening time depends strongly on the RVE-to-interface ratio, practical use on polycrystals will require grain-size-based RVE selection rules before the method can be trusted for quantitative kinetics.","The identical microforce Hill–Mandel construction should transfer to conserved Cahn–Hilliard dynamics and to phase-field fracture, both of which still lack fully consistent gradient-scale bridges.","Direct embedding of RVEs already works in plane problems via out-of-plane thickness scaling; a genuine three-dimensional embedding would remove that restriction and broaden industrial applicability."],"forward_implications":["Engineering components can be simulated with phase-field physics without meshing every diffuse interface.","Existing mechanical FE² pipelines can be extended to order-parameter kinetics by adding the microforce balance and the three classical RVE boundary conditions.","Model-order reduction or machine-learning surrogates become applicable at the RVE level, opening routes to further multi-order-of-magnitude speed-ups.","The same first-order construction immediately suggests multi-order-parameter, higher-gradient or micromorphic enrichments once the scalar theory is in place."],"fun_headline_variants":["FE² homogenizes phase fields via microforce Hill-Mandel condition","Scale-bridging FE² tracks microstructure without resolving every interface","Microforce-based RVE problems enable accurate FE² phase evolution","FE² scheme recovers macro averages for Allen-Cahn and martensite phases","Consistent homogenization maps order parameter gradients in phase-field FE²"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The claim assumes that a pragmatically chosen RVE size still functions as a faithful averaging window whose internal coarsening kinetics correctly stand in for the macroscopic evolution.","fun_headline_variants_meta":{"raw":{"variants":["FE² homogenizes phase fields via microforce Hill-Mandel condition","Scale-bridging FE² tracks microstructure without resolving every interface","Microforce-based RVE problems enable accurate FE² phase evolution","FE² scheme recovers macro averages for Allen-Cahn and martensite phases","Consistent homogenization maps order parameter gradients in phase-field FE²"]},"model":"grok-4.5","effort":"low","cost_usd":0.009494,"raw_usage":{"total_tokens":2119,"prompt_tokens":757,"num_sources_used":0,"completion_tokens":80,"cost_in_usd_ticks":94940000,"prompt_tokens_details":{"text_tokens":757,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1282,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":757,"tokens_out":80,"duration_ms":11525,"temperature":1.0,"reasoning_tokens":1282,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T03:37:57.271465+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Run the same notched-structure or plate-with-hole problem with identical initial patterns but systematically varied RVE-to-interface-thickness ratios; if the FE² macroscopic order-parameter history then diverges from the fully resolved DNS outside the authors’ chosen ratio of 0.1, the claimed predictive accuracy fails.","supporting_citations":[],"review_version":1}