{"id":"b3f317cb-d76e-4821-ab13-f4ae2f39a2ef","arxiv_id":"2607.16005","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A gradient-porous plate is homogenized by layer-wise Hashin-Shtrikman estimates assembled through a variational sum, which the authors report reproduces laminate theory and matches FFT to within 10%.","lead":"The paper proposes a two-step scheme: homogenize each slice of a graded porous plate with Hashin-Shtrikman bounds, then rebuild the plate by a variational sum of slice energies. The scheme is tested against FFT full-field simulations for linear and quadratic porosity gradients, with claimed agreement below 10% and a large CPU saving.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed <10% error vs FFT is not quantitatively supported in the manuscript.","rationale":"The reader's weakest_assumption focuses on the ergodicity hypothesis in the microstructure generation. That is a legitimate concern about generalization, but the more immediate load-bearing weakness is that the central numerical claim (error <10%) is not even supported for the specific cases shown. The paper's own statement that the variational sum reproduces laminate theory suggests the formula (14) is plausible, and the FFT comparison is an external benchmark, but the reported comparison lacks any quantitative error measure. I agree with the reader that ergodicity is a real premise, and the paper does admit deviation without it, but the first thing a careful reader needs is the actual error numbers. Without them, the central claim is unverified. The reader's verdict of CONDITIONAL is appropriate: the missing quantitative data and the ergodicity caveat are addressable. Therefore, my read does not change the verdict; it strengthens the basis for CONDITIONAL. I mark agreement as partial because I identify a different primary concern than the reader, though both are valid.","tokens_in":9176,"tokens_out":4341,"duration_ms":49466,"concrete_test":"Request the authors to provide the numerical data behind Figures 4–6 (or a reproducibility package) and compute the relative error |ψ_model − ψ_FFT|/ψ_FFT for every porosity level and for both linear and quadratic cases. If any point exceeds 10%, the claimed accuracy is false; if all points are below 10% with a defined error metric and error bars, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim includes a specific quantitative threshold: the method reproduces FFT full-field results with 'error measurements ... still less than (<10%)'. However, the Results section provides no quantitative error analysis. There are no tables of relative errors, no error bars on the plotted data, and no definition of the error metric (e.g., L2 relative error on the energy, pointwise error, etc.). Figures 4–6 show two curves that appear close, but visual inspection is not a substitute for a reported error. The claim of '50x CPU-time reduction' is likewise stated without profiling methodology. Without these data, an independent reader cannot verify the paper's headline claim, and the method's practical value (fast but accurate) rests directly on that claim. The manuscript's own 'uncertain about including Fig.6' note further indicates that the numerical section is in an unfinished state. This is load-bearing because the central claim is primarily an accuracy claim, and the evidence for it is absent.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a two-step homogenization scheme for a heterogeneous stratified medium whose properties vary along one direction. In the first step, the 3D domain is sliced into thin layers and each layer is homogenized using Hashin–Shtrikman estimates; in the second step, a 'variational sum' reassembles the layer energies into a macroscopic plate energy, culminating in Eq. (14). The method is applied to a plate with a porosity gradient, for linear and quadratic porosity profiles, and the results are compared with FFT full-field simulations, with a claimed error below 10% and a 50-fold CPU-time reduction. The Γ-convergence basis of the variational sum is imported from the authors' earlier papers [10,11,13] rather than derived in this manuscript.","tokens_in":9353,"tokens_out":4122,"duration_ms":44773,"significance":"If Eq. (14) is valid and the numerical accuracy claim holds, the proposed scheme is a fast approximate homogenization tool for graded plates. The paper has real strengths: the two-step logic is clear, the FFT comparison is an appropriate external benchmark, no constants are fitted to the FFT data, and the philosophy of preserving the gradient in the macroscopic law is sensible. However, the central formula is not derived here, and the quantitative validation that would make the headline claim credible is not reported. Therefore the significance cannot be assessed from the manuscript as it stands; the paper is more a programmatic presentation than a verified method.","major_comments":[{"comment":"The central claim that the model error is 'still less than (<10%)' is not quantitatively supported. No error metric is defined (e.g., relative L2 energy error, pointwise error, or component-wise error), no numerical values or tables are given, and no error bars or statistical variability are reported. The plotted curves are visually close, but visual inspection is not a measurement. Furthermore, the exact loading / macroscopic strain used in the FFT calculation and the precise energy quantity plotted (the caption says normalized energy density, but the normalization denominator is not specified) are not stated. This is load-bearing because the practical value of the method rests on the accuracy claim. The concluding sentence 'uncertain about including Fig.6' reinforces that the numerical section is not in a finished state.","section":"§3.1.2, Figures 4–6"},{"comment":"Equation (14) is not derived in this manuscript; it is asserted via Γ-convergence from the authors' own prior work [10,11,13], and Eq. (11) is invoked rather than proved. The paper itself states in §3 'the variational sum, resulting in (14), allows us to find the same result as that obtained from the laminate theory.' This raises two issues: (i) the central derivation is not independently established here; (ii) if Eq. (14) is equivalent to classical laminate theory, the novelty of the variational-sum route needs to be articulated. The authors should either provide a self-contained proof or a clear derivation of (14) in the present setting, and state precisely what is added beyond standard lamination.","section":"§2 and §3, Eq. (14)"},{"comment":"The ergodicity hypothesis is load-bearing and is not verified. The manuscript states that without this assumption the macroscopic behavior will deviate from the actual behavior, but the microstructure generation is described only as a random n×n cell assembly with 'slight perturbation from the center of the inclusion' and no quantitative statistical characterization. No number of realizations, two-point correlation functions, or sensitivity analyses are reported. Since the FFT reference is generated from the same class of microstructures, the comparison may be self-consistent but does not establish that the slice-level Hashin–Shtrikman estimate is representative for steep gradients, clustered pores, or connected pores, all of which the paper mentions as future extensions.","section":"§3 and §3.1.1"},{"comment":"The 50-fold CPU-time reduction is presented without a profiling methodology. The paper states 'parallelized computation on 32 cores in 500 seconds CPU' versus 'our script completes the computation in less than 10 seconds CPU,' but it is not clear whether these are wall-clock times or CPU-seconds, what hardware/software versions were used, how many runs were averaged, or what operations are included in the 'script' time. Without this information, the speed-up claim cannot be reproduced or assessed. The same section also does not report the values of n, θ0, or the number of inclusions used in the comparison, which are needed to interpret the figures.","section":"§3.1.2 and Conclusions"}],"minor_comments":[{"comment":"The regularization parameter λ is introduced but never specified or discussed. If λ is a fixed constant, the term γn tends to zero in the limit, but the rate of convergence depends on λ; a sentence on its role and typical value would help.","section":"Eq. (2)"},{"comment":"Reference [3] is cited both for non-local approaches and for the variational sum of 2D energy, but the reference is Eringen's nonlocal polar continua; this appears to be a citation error. Reference [19] lists Mechanics of Materials but the DOI points to a Journal of Materials Processing Technology article, which is inconsistent with the bibliographic entry.","section":"References [3] and [19]"},{"comment":"Fig. 3 caption states '2.2% of porosity' while the text discusses varying porosity fractions; the relationship is unclear. Also, the solver name is spelled 'FoXtroT' in one place and 'FoxTrot' in another.","section":"Figures and captions"},{"comment":"The wording 'in the non-linear case' is misleading: the paper treats linear elasticity with nonlinear (quadratic) porosity variation, not a materially nonlinear problem. Replacing 'non-linear case' with 'quadratic porosity profile' would avoid confusion.","section":"Conclusion, §4"},{"comment":"The notation A_i and D_i is used for layer quantities, but in Eq. (13) the same symbols appear with a different meaning; the difference between 'homogenized layer tensors' and 'integrals over the thickness' should be made explicit.","section":"Eqs. (15)–(16)"}],"recommendation":"major_revision","confidential_remarks":"The central result of the paper is heavily dependent on self-citations [10,11,13] and the manuscript explicitly acknowledges that the variational sum reproduces laminate-theory results. Given that the quantitative validation for the headline <10% accuracy claim is absent, I would not recommend acceptance before the authors supply a self-contained derivation or a precise reference to a proof, and a full numerical validation with error metrics and profiling data. I also draw the editor's attention to the unfinished sentence in the conclusions ('uncertain about including Fig.6'), which suggests the manuscript was submitted before the numerical section was completed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper's actual novelty is thin. The energy in (14) is explicitly acknowledged to coincide with laminate theory, so the homogenization result itself is not new. What is new is the claim that this layered variational scheme gives a cheap (<10% error, 50x faster than FFT) way to estimate the effective energy of a porosity-gradient plate. That claim is exactly where the evidence falls short.\n\nWhat the paper does well: it lays out a clear two-step scheme (slice-level Hashin-Shtrikman, then variational sum along the thickness), and it tests the scheme against an independent FFT solver rather than only against its own approximations. No constants are fitted to the FFT results. It also honestly notes that the variational sum reproduces classical laminate theory, which is a useful sanity check.\n\nWhere it is soft: the manuscript does not actually provide the promised error analysis. The 'less than 10%' claim is one sentence with no definition of the error metric, no per-case table, no error bars. The '50x CPU time' claim is likewise a pair of wall-clock numbers with no profiling methodology. The FFT comparison covers two porosity profiles (linear and quadratic), which is a thin basis for a general claim. The central limit formula (14) is imported from the authors' own prior Gamma-convergence papers and not re-derived or restated precisely. The regularization parameter lambda is never assigned, and Eq. (3) defines E_n in terms of itself. The conclusion even contains a note 'uncertain about including Fig.6', which makes the paper look unfinished. The ergodicity assumption is stated and flagged by the authors, so I won't call it hidden, but it is load-bearing: with steep gradients or clustering pores, the slice-level HS estimate and the FFT reference would not be representative.\n\nThese are all addressable. The idea is sensible and the external benchmark is a plus. The paper is the kind of engineering-oriented contribution that can be useful after substantial revision, but right now the accuracy and speedup claims are unsubstantiated.\n\nWho it's for: someone working on functionally graded materials who wants a quick approximate tool and is willing to take the accuracy on faith. I wouldn't cite it in my own work until the numbers appear. Send it to peer review if the venue tolerates major revision; otherwise a desk reject with an invitation to resubmit with real error data would also be defensible.","headline":"A plausible fast approximate homogenization scheme for gradient porous plates, but the accuracy and speedup claims are not backed by the data actually reported.","tokens_in":9877,"tokens_out":3004,"would_cite":false,"duration_ms":32380,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74Q05","74K20","49J45"],"pacs":[],"model":"deepseek-v4-flash","headline":"A two-step homogenization scheme claims to capture the elastic energy of porosity-gradient plates within 10% of full-field FFT simulations, at roughly 1/50th the compute time.","keywords":["homogenization","porosity gradient","Hashin-Shtrikman bounds","variational sum","thin plate theory","Gamma-convergence","FFT full-field simulation"],"falsifier":"Generate a graded-porosity plate where the porosity changes substantially within a single slice, or where pores cluster or percolate across slices, run the FFT full-field solver and equation 14 on the same microstructure, and check whether the normalized energy difference stays below the claimed 10%; finding the gap widen past 10% in those regimes would falsify the claim's scope. A simpler arithmetic check: set the porosity constant, θ(x3)=θ0; then Eq. 14 should reduce exactly to the classical plate stiffness from a single Hashin–Shtrikman homogenization.","tokens_in":9016,"feed_emoji":"⚙️","tokens_out":7773,"duration_ms":69547,"temperature":0.7,"pith_summary":"The paper tries to establish that a stratified material whose properties vary along one direction can be homogenized by slicing it into thin 2D plates, estimating each plate with Hashin–Shtrikman bounds, and reassembling the 3D energy through a variational sum. Applied to a thin plate with a porosity gradient, this yields a closed-form macroscopic plate energy (equation 14) that the authors compare with FFT full-field simulations for linear and quadratic porosity profiles. They report the model within 10% of the FFT energies and about 50 times faster (10 s vs 500 s on 32 cores for a 400×400×400 voxel case). The paper itself notes that the variational sum recovers classical lamination-theory results, so the claimed contribution is a fast, explicit, gradient-preserving reconstruction rather than a new bound. A self-acknowledged limit is the ergodicity assumption within each slice, and the conclusions contain an unresolved aside about whether the comparison figure against Hashin–Shtrikman and Voigt bounds should be kept; that figure supports the claim that classical schemes miss the gradient, so the aside is about presentation rather than the load-bearing argument.","feed_headline":"Fast scheme matches graded-porosity plate stiffness within 10%","feed_subtitle":"Slicing a porosity-gradient plate, homogenizing each slice, and summing variationally reproduces full-field energies at 1/50th cost.","key_machinery":"The variational sum is the engine: the domain is cut into n thin plates of thickness 1/n; each slice's 2D plate energy uses membrane and bending tensors obtained from Hashin–Shtrikman estimates on the local porosity; a regularization term γ_n penalizes energy jumps between neighboring slices and vanishes as n→∞, so the discrete sum Γ-converges to the integral energy ψ0 in equation 14. The passage from slice-wise estimates to a continuous thickness-dependent stiffness A(x3), D(x3) is what lets the model keep the gradient information that constant-porosity bounds discard.","core_discovery":"The central claim is that the macroscopic energy of a porosity-gradient plate is captured by ψ0 = (1/|O|) ∫_O [ (1/2) d A(x3) d + x3^2 (1/2) κ D(x3) κ ] dV, where A and D are membrane and bending stiffness tensors obtained by Hashin–Shtrikman homogenization of each slice and the integral over the volume reassembles the layers. The authors treat this as a two-step scale transition: first reduce each 3D sliced layer to a homogeneous 2D plate, then let the slice thickness go to zero so the discrete plate energies converge to a single variational energy. They test the formula against an FFT full-field solver on a 400×400×400 voxel porous plate with linearly and quadratically graded porosity and","pith_inferences":["If equation 14 holds, the same slice-then-sum structure should transfer to thermal, electrical, or diffusion transport through graded media, where the variational assumptions are even easier to satisfy than in elasticity.","The ergodicity boundary is testable: generate microstructures with clustered or connected pores at the same overall porosity and compare Eq. 14 to FFT; the reported <10% error should degrade as slice-level representativeness fails.","The closing suggestion to make θ a tensor points at a concrete extension: replace the scalar porosity profile by a direction-dependent or spatially varying one, which would widen the method toward additively manufactured parts with oriented pore textures.","Since the variational sum equals laminate theory, the practical value may lie in giving designers a closed-form energy they can differentiate for stiffness, which makes gradient-property optimization (e.g., tailoring porosity profiles) computationally cheap."],"forward_implications":["For unidirectional porosity gradients, the membrane and bending response of a thin plate can be evaluated from slice-wise Hashin–Shtrikman estimates without building a 3D mesh of the pores.","The same variational-sum pipeline should apply to other physical behaviors (the paper notes linear elasticity is one example), as long as the slice energy is convex and Lipschitz.","A 400×400×400 voxel comparison completes in about 10 s on one CPU versus 500 s on 32 cores for the FFT solver, opening the door to routine parametric studies of graded plates.","Because the formula reproduces lamination theory, it connects classical laminate calculations to microstructure-informed estimates in a single expression."],"fun_headline_variants":["Two-step homogenization captures graded porosity plate energy within 10%","Slice-wise homogenization matches gradient plate stiffness at 1/50th cost","Variational sum of slice models predicts porosity-gradient plate response","Gradient property homogenization: slice, bound, sum, match full-field","Hashin-Shtrikman slices plus variational sum for graded plates"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that, within each thin slice, the pore distribution is statistically homogeneous (ergodic) and the Hashin–Shtrikman slice estimate is representative; the paper states that if this fails the macroscopic behavior will deviate, and steep gradients, clustering, or connected pores would break it.","fun_headline_variants_meta":{"raw":{"variants":["Two-step homogenization captures graded porosity plate energy within 10%","Slice-wise homogenization matches gradient plate stiffness at 1/50th cost","Variational sum of slice models predicts porosity-gradient plate response","Gradient property homogenization: slice, bound, sum, match full-field","Hashin-Shtrikman slices plus variational sum for graded plates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001021,"raw_usage":{"total_tokens":4117,"prompt_tokens":687,"completion_tokens":3430,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":431,"completion_tokens_details":{"reasoning_tokens":3337}},"tokens_in":431,"tokens_out":3430,"duration_ms":24746,"temperature":1.0,"reasoning_tokens":3337,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T21:37:11.501416+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate a graded-porosity plate where the porosity changes substantially within a single slice, or where pores cluster or percolate across slices, run the FFT full-field solver and equation 14 on the same microstructure, and check whether the normalized energy difference stays below the claimed 10%; finding the gap widen past 10% in those regimes would falsify the claim's scope. A simpler arithmetic check: set the porosity constant, θ(x3)=θ0; then Eq. 14 should reduce exactly to the classical plate stiffness from a single Hashin–Shtrikman homogenization.","supporting_citations":[],"review_version":1}