{"id":"2e00f23b-9a13-4a77-90b4-9724a35e42e8","arxiv_id":"2607.27493","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A dual-parameter persistence summary I, integrating Betti-1 counts over shape-distance and spatial-scale, rises monotonically with strain and jumps sharply between 8% and 12% in four EBSD ice microstructures.","lead":"A new analysis combines shape similarity and topological data analysis to quantify how grains are arranged in material microstructures. Applied to ice samples strained by different amounts, it reports a jump in a topological 'texture' score between 8% and 12% strain.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"I is not normalized for grain count; the 8–12% jump matches a >2× increase in centroid density, so the claimed topological transition may be a density artifact.","rationale":"Reading the paper in good faith, the methodological novelty—combining SST shape distances with a Rips bifiltration and a scalar integral summary—is reasonable and potentially useful for microstructure metrology. The empirical demonstration, however, is the load-bearing part of the paper: the claim that I reveals a sharp transition in topological texture coinciding with DRX onset. That claim rests on four strain levels, each from a single EBSD image, with no error bars and no control for the number of grains. The reader's weakest-assumption analysis correctly isolates the most serious confound: β1 in a Rips complex is not scale-invariant with respect to point count. For a planar point process, expected Betti numbers scale with intensity, so when Fan et al. report that grain number density more than doubles across the exact strain interval where I jumps, a density-driven explanation is the null hypothesis that must be excluded. The paper contains no such null model and no normalization. This is not an ad hominem or a complaint about disagreement with prior consensus; it is an internal-control gap in the experimental argument. The proposed concrete test—randomized centroids with identical counts and shape-distance distributions—would settle whether I carries topological information beyond density. Until that test is run, the central quantitative claim is unsupported, so the REJECT verdict stands. I also note in passing that the formal definition of the sublevel Rips bifiltration in Section 3.1 appears garbled in the rendered text, but that is a presentation issue rather than the primary scientific concern.","tokens_in":18965,"tokens_out":3875,"duration_ms":40798,"concrete_test":"Generate null ensembles for each of PIL254, PIL184, PIL185, and PIL255: place the same number of grain centroids uniformly at random inside the same scan mask (or use a Poisson-disk/hard-core process matching the observed minimum spacing), assign shape distances by random permutation of the empirical SST distances, recompute the Eq. (3) integral I for ≥1000 realizations, and form a null distribution for each strain level. If PIL185's observed I lies inside the null interval for its grain count, the 8–12% jump is explained by density. Complement this by recomputing Fig. 12 with I normalized by N (or by the null mean) and checking whether the sharp transition at 12% persists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Section 4.1, Fig. 12) is that I, the double integral of β1 in Eq. (3), reveals a sharp topological transition between 8% and 12% strain. However, I is computed from a Rips complex on grain centroids, and for any stationary point process the expected Betti-1 curve scales linearly with intensity: E[β1(ε)] ≈ λ·C(ε). Doubling the number density of grains therefore roughly doubles I even if the spatial pattern is unchanged. The authors themselves cite Fan et al. [42] reporting that between 8% and 12% strain the number density of grains more than doubles—exactly where I jumps. The paper provides no normalization by grain count, no comparison with randomized centroid configurations at the same N, and no uncertainty quantification (only one EBSD image per strain level). Without such controls, the observed rise in I is equally consistent with a pure density effect; the inference that it measures 'necklace topology' or DRX onset is unsupported. This is not a claim that the method is wrong in principle, only that the headline empirical result has not been separated from a known confound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a dual-parameter topological summary for microstructure images: a Rips bifiltration on grain centroids with a second axis given by sublevel sets of a Separable Shape Tensor (SST) shape distance from an intrinsic mean. It defines a scalar I as the double integral of the Betti-1 count over the two-parameter domain (Eq. 3). The method is demonstrated on four EBSD ice samples strained from 0.03 to 0.20. The central claim (Sections 1.2, 4.1, Fig. 12) is that I reveals a sharp transition in topological texture between 8% and 12% strain, coinciding with the onset of dynamic recrystallization and necklacing as described in ASTM E1181.","tokens_in":19172,"tokens_out":6516,"duration_ms":54823,"significance":"If the empirical claim were established, the framework would be a meaningful contribution: it combines shape and spatial arrangement in one interpretable summary, with no labels or threshold choices. The construction of the two-parameter filtration from SST distances is original and clearly explained, and the paper is generally readable. However, the demonstration is too weak to support the headline result: only four strain levels with one image each are used, no uncertainty quantification is provided, and the key scalar is not separated from a known confound with grain centroid density. As it stands, I is not a validated measure of necklacing topology.","major_comments":[{"comment":"The central claim that I measures topological texture is confounded by grain count. I is a double integral of β1 over a Rips filtration on grain centroids. For any stationary point process, expected β1 is proportional to intensity; doubling the number of centroids roughly doubles I even if the spatial pattern is unchanged. The paper itself cites Fan et al. [42] reporting that grain number density more than doubles between 8% and 12% strain, exactly where I jumps. No normalization by grain count, no comparison with randomized configurations of the same N, and no spatial thinning are provided. The observed rise in I is therefore equally consistent with a density effect, and the claim of a topological transition is unsupported.","section":"Section 4.1, Fig. 12, Eq. (3)"},{"comment":"The statistical evidence is very limited. Only one EBSD scan is used per strain level, so Figure 12 has four points and no error bars. The 'sharp transition' rests on a comparison of one image at 8% and one at 12%. The 'plateau' between 12% and 20% also rests on two points. To support the monotonic trend and the location of a transition, the authors need replicated scans, bootstrap over subregions, or at least a statement of measurement uncertainty; otherwise the scalar I cannot be distinguished from sample-to-sample variation.","section":"Section 4.1, Fig. 12"},{"comment":"The identification of β1 cycles with necklacing is not validated. The idealized Figure 5 shows one mechanism by which rings of fine grains create persistent holes, but Rips complexes on random point clouds also produce β1, especially as density increases. The paper provides no synthetic experiments with known necklace versus random arrangements at matched density. Consequently, the mechanistic interpretation in Section 4.1 (core-and-mantle structure, saturation of necklace topology) and the stronger claim in Section 5 that I is a 'direct, quantitative proxy' for creep and ductility are not supported by the evidence presented.","section":"Section 2.1, Fig. 5; Section 4.1"}],"minor_comments":[{"comment":"Typos: 'difficult' (Section 1), 'recrystalization' and 'bewteen' (Section 4.1). Please proofread.","section":"General/typos"},{"comment":"The dimensions in the product manifold Gr(d,q) × S^d_++ and the role of the low-dimensional submanifold principal directions are not fully defined. Since a,b,d and the number of principal directions are user choices, a sensitivity study or explicit default values are needed for reproducibility.","section":"Eq. (1), Section 2.2"},{"comment":"The authors note that the construction 'may not necessarily reflect standard two-parameter TDA.' This caveat is important and should be moved earlier (e.g., Section 3.1) so that readers do not mistake the scalar I for a multiparameter persistence invariant such as a rank invariant.","section":"Section 5"},{"comment":"The histograms are not clearly captioned; it is difficult to read which distance is plotted in each panel. Please label directly.","section":"Figure 7"},{"comment":"Javaplex and public data are cited, but the grid resolution used for the bifiltration, the weights a,b in Eq. (1), and the SST rank d are not reported. For a methods paper, this information should be given so the scalar I can be reproduced.","section":"Reproducibility"}],"recommendation":"reject","confidential_remarks":"The methodological core (SST + two-parameter Rips) is promising and could be published after substantial additional work. My rejection is based on the unaddressed density confound and the lack of any uncertainty quantification; with only four images and no null models, the headline transition is not established. A revised submission that includes normalization/randomized controls and more extensive validation would be worth reconsidering."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper has a real kernel: combining sublevel-Rips bifiltrations with SST product-manifold shape distances to summarize necklace-type grain structures. The specific construction is not in the literature, and it is presented clearly, with the standard ingredients properly attributed. The visual bifiltration plots are informative, and the paper is honest that the two-parameter setup is a tailored variant of multiparameter persistence.\n\nThe soft spot is exactly what the reader flagged: the scalar summary I is computed from Rips complexes on grain centroids, and Betti-1 counts in a Rips complex scale with point density. For any point process with intensity lambda, expected Betti-1 is roughly lambda times a scale-dependent term. The paper cites Fan et al. reporting that grain number density more than doubles between 8% and 12% strain, which is precisely where I jumps. Without normalizing by grain count, or comparing to randomized configurations with the same N, the sharp increase in I cannot be attributed to necklace topology. The authors even mention the density increase as supporting evidence, which suggests they are aware of it, but they do not treat it as a confound. That is a load-bearing flaw in the central quantitative claim.\n\nThat said, the method itself is not invalid. The bifiltration surfaces in Figure 10 show differences in structure, not merely in height, and a shape-conditioned sublevel Rips filtration may genuinely separate spatial arrangement from shape distribution. The missing controls are straightforward: intensity normalization, null models (e.g., Poisson or hard-core processes with matched centroids), and some uncertainty quantification. With those, the paper could make a compelling case.\n\nAlso note the data is publicly available, and the processing pipeline uses common tools, so the analysis is reproducible.\n\nMy take: the paper deserves a serious referee, but the current version should not be accepted as-is. The scarcity of data (four strain levels, one image each) is a minor concern for a methods demonstration, but it amplifies the need for statistical controls. If you see this in peer review, ask for the null-model comparison and a clear discussion of what I does and does not measure. It is a promising framework, currently over-claimed.","headline":"Promising TDA+SST framework, but the headline scalar summary is likely confounded by grain density; needs null models and normalization before the claims hold.","tokens_in":19717,"tokens_out":2112,"would_cite":false,"duration_ms":22590,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["55N31"],"pacs":[],"model":"deepseek-v4-flash","headline":"A scalar bipersistence summary I, integrating Betti-1 counts over a shape-distance and spatial-scale filtration, sharply distinguishes low- from high-strain ice microstructures and detects onset of dynamic recrystallization.","keywords":["topological data analysis","persistent homology","multiparameter persistence","shape analysis","separable shape tensors","electron backscatter diffraction","necklacing","dynamic recrystallization"],"falsifier":"Take each ice-sample scan, keep the same number of grain centroids and the same image domain, but place the centroids uniformly at random (or as a Poisson process) with the observed shape-distance values assigned to them; compute the same bifiltration and I. If the sharp jump in I between 8% and 12% strain largely persists under random placement, the measure is tracking density or domain effects rather than ring topology; if the jump disappears, the necklace interpretation is supported.","tokens_in":18783,"feed_emoji":"❄️","tokens_out":9184,"duration_ms":74354,"temperature":0.7,"pith_summary":"The paper sets out to quantify the spatial arrangement of grain shapes in polycrystalline materials, not just their average size or shape. It builds a two-parameter filtration on grain centroids: one axis is the usual spatial scale of a Rips complex, the other is how far each grain's shape sits from the mean shape in a manifold of Separable Shape Tensors. Counting the persistent one-dimensional loops (Betti-1) over this grid yields a surface, and integrating the surface gives a single scalar I. Applied to EBSD scans of ice deformed to increasing axial strain, I rises gently up to 8% strain and then jumps sharply by 12%, matching the strain range where dynamic recrystallization becomes the dominant grain-refinement mechanism. If this holds, the framework gives a quantitative, interpretable measure of 'necklacing' — fine grains arranged in rings around coarse grains — that existing standards describe only qualitatively.","feed_headline":"Topology score catches recrystallization onset in ice","feed_subtitle":"This integral turns ring-like fine-grain patterns into a number tracking strain history.","key_machinery":"The central object is the sublevel Rips bifiltration S↑_{ε,ℓ} = φ^{-1}((-∞,ℓ]) ∩ Rips(ε), built on grain centroids, where φ is the distance of each grain's separable shape tensor from the intrinsic mean on the product manifold Gr(d,q) × S^d_{++}. The first parameter ε is the Rips radius controlling spatial connectivity; the second ℓ is a shape-distance threshold selecting grains similar to the mean. Tracking Betti-1 counts β1(ε,ℓ) across the two-parameter grid, and integrating them to define I = ∫∫ β1 dε dℓ, converts the spatial pattern of similar-shaped grains into a single number. The key mechanism is that a ring of fine grains surrounding a coarse grain completes a 1-cycle at relatively s","core_discovery":"The central claim is that a bifiltration combining spatial Rips complexes with shape-distance sublevel sets — using the separable shape-tensor distance from the intrinsic mean as the second parameter — turns the heuristic notion of necklacing into a measurable topological signature: a plateau of elevated Betti-1 counts over a range of spatial scales, visible as an island in the (ε, ℓ) contour map. The scalar I = ∫∫ β1(ε,ℓ) dε dℓ accumulates this persistence and increases monotonically with strain, with a sharp transition between 8% and 12% strain that the authors identify with the onset of dynamic recrystallization as the dominant grain-refinement mechanism. The authors further argue that th","pith_inferences":["The paper does not normalize β1 or I by the number of grain centroids; since Rips Betti numbers grow with point count, a comparison against random point configurations matched for count and domain would clarify whether I measures ring topology or simply grain density. (Editorial extension.)","If the necklace plateau is the mechanistic signature, the same bifiltration could be used to detect percolation-like transitions in other spatial networks where a second attribute (shape, intensity, orientation) conditions connectivity.","The shape-distance axis could be replaced by other grain attributes — orientation, phase, or local misorientation — yielding bifiltrations tuned to different microstructural patterns, such as banding or clustering by crystallographic orientation.","A natural testable extension is to normalize I by the integral over a null ensemble of Poisson-distributed centroids; the strain transition's persistence under this normalization would separate topology from density effects."],"forward_implications":["Necklacing and other duplex grain arrangements can be assigned a quantitative scalar I, or region-specific integrals, providing a basis for ordering or clustering microstructure samples without manual grading.","Because the construction uses only grain centroids and shape distances, it can be applied to metals, ceramics, and composites imaged by EBSD or optical microscopy, where standard test methods currently rely on qualitative pattern descriptions.","The sharp rise in I between 8% and 12% strain localizes the strain range in which dynamic recrystallization becomes the dominant grain-refinement mechanism, offering a quantitative fingerprint of processing history.","Region-restricted integrals of β1 can be tuned to isolate specific topological patterns — necklacing, banding, duplex distributions — allowing standards to be extended with quantitative pattern-specific measures.","The scalar I and bifiltration surfaces can serve as reduced-order inputs or calibration targets for phase-field and cellular automata simulations of grain growth and recrystallization."],"fun_headline_variants":["Topological fingerprint tracks strain and recrystallization onset","Bifiltration Betti numbers signal recrystallization start","New shape-aware topology detects recrystallization threshold","Persistence integral marks strain-driven grain refinement"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The key load-bearing premise is that the scalar I captures necklace topology rather than merely reflecting the number of grain centroids: the grain count roughly doubles over the same 8%–12% strain range where I jumps, and the paper does not control for point density in its summary.","fun_headline_variants_meta":{"raw":{"variants":["Topological fingerprint tracks strain and recrystallization onset","Bifiltration Betti numbers signal recrystallization start","New shape-aware topology detects recrystallization threshold","Persistence integral marks strain-driven grain refinement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00053,"raw_usage":{"total_tokens":2379,"prompt_tokens":720,"completion_tokens":1659,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":464,"completion_tokens_details":{"reasoning_tokens":1598}},"tokens_in":464,"tokens_out":1659,"duration_ms":11984,"temperature":1.0,"reasoning_tokens":1598,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T06:50:10.177566+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take each ice-sample scan, keep the same number of grain centroids and the same image domain, but place the centroids uniformly at random (or as a Poisson process) with the observed shape-distance values assigned to them; compute the same bifiltration and I. If the sharp jump in I between 8% and 12% strain largely persists under random placement, the measure is tracking density or domain effects rather than ring topology; if the jump disappears, the necklace interpretation is supported.","supporting_citations":[],"review_version":1}