{"id":"682aeee0-1525-4bfa-aa70-fc74827395fd","arxiv_id":"2607.21031","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"BV functions and finite-perimeter sets on arbitrary Riemannian manifolds admit the full Euclidean structure theory (differentiation, De Giorgi, Federer, traces, Gauss–Green, strict interior approximation) without completeness or global curvature bounds.","lead":"The paper extends the classical BV/geometric-measure toolbox — perimeters, reduced boundaries, structure theorems of De Giorgi and Federer — from Euclidean space to arbitrary Riemannian manifolds, dropping completeness and global curvature assumptions. This enables boundary-value, capillarity, and shape-optimization problems to be treated intrinsically on curved spaces, and it is applied here to a mixed-boundary capillarity Γ-convergence problem.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Vitali-relation premise in Remark 4.1(2) is the load-bearing foundation; it is asserted with citations rather than proved, and if closed balls are not a Vitali relation for every Radon measure on non-complete or curvature-unbounded manifolds, Theorems 4.2–4.3 and all of Sections 5–6 collapse.","rationale":"The reader’s weakest-assumption analysis already identifies the Vitali-relation premise as load-bearing, and I agree that this is the point where the paper is thinnest. However, I do not see an internal contradiction or an obvious counterexample in the manuscript itself: the local directional-limitedness argument is a standard route, and the subsequent chart-localization steps are largely consistent. The paper’s own statement that the Besicovitch lemma is generally false makes the Vitali claim nontrivial, and the noncompact-ball/infinite-measure subtlety in non-complete manifolds is not addressed. Thus the concern is real but not demonstrated fatal: it points to a gap in justification rather than a proven false theorem. The intersection-perimeter transfer at (6.9)/(6.13) is a second asserted step, but it is less foundational because it would follow from the structure theory if the Vitali premise is secured. I therefore keep the reader’s CONDITIONAL verdict unchanged: the central claims are likely correct, but the load-bearing Vitali assumption should be fully pinned down in a revision or referee round.","tokens_in":42488,"tokens_out":45782,"duration_ms":515447,"concrete_test":"Verify Federer’s Theorem 2.8.18 against the following concrete non-complete Riemannian manifold: M = (0,1) with metric g = (1-x)^(-3/2) dx^2, smooth on M but incomplete at x=1. For the Radon measure µ = vol_g, a closed ball centered near x=0 can have infinite µ-measure once it reaches 1, so closed balls are not finite-on-bounded sets. Check directly whether the family V = {(x, B̄(x,r)) : x∈M, r>0} satisfies the paper’s Definition 4 Vitali property for µ, and whether the conclusion of Theorem 4.2(i) holds for a measure ν with density (1-x)^(-1/2) with respect to µ. If the cited Federer theorem requires finite-on-bounded measures or globally compact balls, the premise of Remark 4.1(2) — and hence the central claim — needs an additional localizing argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper’s entire structure theory rests on Theorem 4.2 (Lebesgue–Besicovitch–Federer differentiation), whose only non-Euclidean input is the claim in Remark 4.1(2) that closed Riemannian balls form a Vitali relation for every Radon measure on an arbitrary smooth Riemannian manifold. The paper itself notes that the Besicovitch covering lemma is generally false on arbitrary Riemannian manifolds, so this Vitali relation is doing all the work. The justification is a one-paragraph appeal to directional limitedness on relatively compact sets and to [22, 2.8.9, 2.8.18] / [11, Def. 2.5]. This is plausible, but it is not demonstrated: the hypotheses of Federer’s theorem are not stated, and the non-complete setting introduces a real subtlety — a closed Riemannian ball need not be compact, so a locally finite Radon measure can assign infinite mass to a ball (e.g. on (0,1) with a metric that blows up near 1). The standard Vitali/differentiation theorems are usually stated for measures finite on bounded sets or for compact balls; whether the cited Federer theorem covers this case is exactly the question. Since Definition 5.1 of the reduced boundary, Theorem 5.15, and Theorem 5.19 all depend on Theorem 4.3, this is the single most load-bearing point. A secondary load-bearing transfer is the unproved Riemannian version of the intersection-perimeter identity [43, Thm. 16.3] used in (6.9)/(6.13); it is probably routine once De Giorgi is available, but it compounds the sense that the most foundational analytic premise is asserted rather than proved.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an intrinsic theory of BV functions and finite-perimeter sets on arbitrary smooth Riemannian manifolds, without assuming completeness or global curvature bounds. The framework combines the covector-measure characterization of BV from [30] with Federer's metric-space differentiation theory and Euclidean geometric measure theory in charts. The main results are: a Lebesgue–Besicovitch–Federer differentiation theorem for Radon and covector measures (Theorems 4.2 and 4.3); a Riemannian reduced boundary defined by weak-* blow-up limits (Definition 5.1); De Giorgi's structure theorem giving rectifiability of the reduced boundary, the identity |D1_E| = H^{m-1} restricted to ∂*E, and half-space blow-ups (Theorem 5.15); Federer's characterization of the measure-theoretic boundary (Theorem 5.19); trace and Gauss–Green theorems on domains (Theorem 6.4, Corollary 6.6); a strict interior approximation result for finite-perimeter sets respecting a Dirichlet boundary portion (Theorem 6.12); and a Γ-convergence result for capillarity functionals with mixed boundary conditions (Theorem 6.14). The proof strategy is to reduce to Euclidean theorems chartwise and to use local doubling and Poincaré properties on relatively compact sets.","tokens_in":42732,"tokens_out":32947,"duration_ms":326935,"significance":"If the technical foundations hold, this paper makes a substantial portion of Euclidean BV theory available on arbitrary Riemannian manifolds, filling a gap between the Euclidean setting and the more restrictive metric-measure-space framework. The localization approach is natural, and the paper has several strengths: it explicitly preserves Riemannian objects such as polar vector fields, normal vectors, blow-ups via the exponential map, and Hausdorff measures; it gives intrinsic statements rather than relying on an ambient Euclidean embedding; and the main chain of deductions — from differentiation to De Giorgi/Federer structure theorems and then to boundary-value applications — is coherent and largely explicit. The paper also correctly identifies a genuinely Riemannian subtlety: pointwise densities are not invariant under general charts (Example 5.17), only under normal charts (Corollary 5.18). However, two load-bearing transfer points are not fully demonstrated in the manuscript: the Vitali-relation premise for closed Riemannian balls on arbitrary non-complete manifolds, and the verbatim Riemannian transfer of the intersection-perimeter identity from [43, Thm 16.3]. These need to","major_comments":[{"comment":"The entire differentiation theory, and hence Definition 5.1 and all of Sections 5–6, rests on the claim that the family of closed Riemannian balls V={(x,\\bar B(x,r))} is a Vitali relation for every Radon measure on an arbitrary smooth Riemannian manifold. The justification is a one-paragraph citation to [22, 2.8.9, 2.8.18] and [11, Def. 2.5], but the hypotheses of the cited theorem are not stated and the non-complete case is not addressed. In particular, if the manifold is not complete a closed ball may be non-compact, and a locally finite Radon measure can have infinite mass on it, so the quotients in Theorem 4.2 are not a priori defined for all r>0. The issue is likely repairable by restricting to radii below the local injectivity radius (where closed balls are compact) and proving the Vitali property for this subfamily, but this must be written out. This is the single most load-bearin","section":"Section 4, Remark 4.1(2)"},{"comment":"The proof of Theorem 6.12 transfers the Euclidean intersection-perimeter identity [43, Theorem 16.3] to Riemannian manifolds by saying 'the proof applies verbatim.' This identity involves density sets E^(1), normal-vector coincidence sets {n_E=n_F}, and H^{m-1}-null-set assertions, all of which need the fine structure theory from Section 5. Since the density of a set at a boundary point is not invariant under arbitrary charts (Example 5.17), the transfer is not completely automatic. The manuscript should provide a proof, or at least a precise Riemannian statement of the identity used in (6.9) and (6.13), with the necessary chart-invariance and null-set facts justified. This is load-bearing because Theorem 6.12 is the basis for the recovery sequence in the Γ-convergence result, Theorem 6.14.","section":"Section 6.2, Theorem 6.12, especially (6.9) and (6.13)"},{"comment":"The chartwise compatibility of the reduced boundary is central for transferring Euclidean results, but the proof relies on the claim that the limit is independent of the choice of Vitali relation and that images of Euclidean balls under a chart form a Vitali relation. This is plausible for small r, where charts are bi-Lipschitz, but the local-to-global passage is not detailed. In particular, the proof should specify that the limits are taken over r small enough for the Euclidean ball to lie inside the chart domain and for the chart to be bi-Lipschitz there, and should justify that the two normalization sequences give the same limit. This is not a fatal flaw, but it needs to be made explicit because the rest of Section 5 hinges on this lemma.","section":"Section 5, Lemma 5.2"}],"minor_comments":[{"comment":"The identity 'D^h u[X] = sqrt(det A) D^g u[X]' is a shorthand for the functional identity involving multiplication of covector measures by the scalar function sqrt(det A); as written it may be misread as a pointwise scalar action on X. Since the paper defines multiplication fν by (fν)[X]=ν[fX], this is consistent, but a parenthetical clarification would help.","section":"Remark 3.8(2)"},{"comment":"The sentence 'The generalization ... presents itself with two major difficulties. Firstly ... Secondly ...' is clear, but Theorem 4.2 is stated as a 'summary' of [22, 2.9.5–2.9.10] without giving the statement of the Vitali-relation theorem used. Adding the explicit statement of the relevant Federer theorem would improve readability.","section":"Section 4, opening paragraph"},{"comment":"The constants C_v(r0) and the application of inequality (5.28) are somewhat terse. In particular, the expression C_v(r0)=(1/L(r0))^{2m} exp((m-1)K r0) is not clearly derived from (5.28). A short derivation would remove ambiguity.","section":"Corollary 5.18 proof"},{"comment":"The lower semicontinuity estimate uses a product formula for liminf of (1-β_n)P(E_n;Ω∪γ)+β_n P(E_n); this is correct but should be justified briefly, since liminf of a sum with converging coefficients is being split termwise.","section":"Theorem 6.14 proof"},{"comment":"There are a few typographical issues visible in the text: unmatched parentheses in the definition of the total variation after (3.1), 'Annals de la Faculté des sciences de Toulouse' in the references, and some equation numbers referenced only by 'above' or 'as before'. A careful proofreading pass is recommended.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The core concern is whether the Vitali-relation premise, Remark 4.1(2), is indeed a standard theorem applicable to every Radon measure on a non-complete Riemannian manifold. The authors should be asked to provide a self-contained proof or a very precise citation with all hypotheses checked. If that premise is valid, the paper is likely correct and would be a valuable contribution; if not, much of the structure theory collapses. The secondary issue is the unproved Riemannian transfer of the intersection-perimeter identity; this is more routine but should be fixed in revision. I would not reject on the current evidence, but the revision must close these gaps."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it claims: it extends De Giorgi, Federer, trace, Gauss–Green, and strict interior approximation to arbitrary Riemannian manifolds without completeness or global curvature bounds. The chart-localization machinery is genuinely useful, the metric-change computations (Lemma 3.7, (3.10), Corollary 5.10) are careful, and the applications to boundary value problems and capillarity are honest, with limitations flagged in Remark 6.5 and after Theorem 6.12. There is no fitted data and no circular reliance on the theorems being proved; the reductions to Euclidean results and to Federer's metric-space differentiation machine are legitimate.\n\nThe soft spots are real but mostly cosmetic. The main one is Remark 4.1(2): the claim that closed Riemannian balls form a Vitali relation for every Radon measure is the foundation for Theorems 4.2–4.3, and it is justified by citation rather than proof. The stress-test worry about infinite-mass balls on non-complete manifolds does not actually bite for the differentiation theorem, because local finiteness guarantees that every point has a neighborhood of finite measure, hence balls of sufficiently small radius have finite measure; the Vitali covering property can be restricted to such balls. But the cited hypotheses of Federer 2.8.18 should be stated, and the directionally-limited condition on arbitrary Riemannian manifolds deserves a short proof or a precise reference. A second gap is the verbatim transfer of the intersection-perimeter identity (Maggi 16.3) to Riemannian manifolds at (6.9)/(6.13); it is probably routine once De Giorgi is available, but it should be spelled out because Theorem 6.12 and the Gamma-convergence result rest on it. Minor: the closing arguments of Theorem 5.19 rely on 'same arguments as before' and a 'straightforward covering argument' — acceptable but slightly compressed.\n\nWho gets value from this: anyone working in BV theory on manifolds, geometric measure theory in non-compact settings, or variational problems with rough domains. The paper is dense and long, but the structure is clear. It deserves a serious referee, not a desk rejection. Recommendation: send it to review, and ask the referee to pin down the Vitali-relation justification and the intersection-perimeter transfer. Those are the only load-bearing points that could genuinely shift the verdict.","headline":"The paper plausibly delivers the full Euclidean BV/finite-perimeter toolbox on arbitrary Riemannian manifolds; the load-bearing Vitali-relation premise is asserted rather than proved, but the non-compact-ball worry is weaker than it looks.","tokens_in":665,"tokens_out":1422,"would_cite":true,"duration_ms":50831,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q15","26B30","28A75","58J32"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the Euclidean theory of bounded-variation functions and finite-perimeter sets—including De Giorgi's and Federer's structure theorems, Gauss–Green formulas, and strict interior approximation with Dirichlet boundary con","keywords":["BV functions","finite perimeter sets","Riemannian manifolds","De Giorgi structure theorem","Federer theorem","reduced boundary","covector measures","Gamma-convergence capillarity"],"falsifier":"Take a specific non-complete or curvature-unbounded Riemannian manifold and a Radon measure; compute the limsup and liminf of ball averages of a covector measure. If they differ on a set of positive measure, the Vitali relation fails and Theorems 4.2–4.3 collapse. Alternatively, construct two finite-perimeter sets on a manifold where P(E∪F)+P(E∩F) > P(E)+P(F), which would falsify the verbatim transfer of the Euclidean intersection-perimeter identity used in Theorem 6.12 and break the approximation theorem.","tokens_in":42173,"feed_emoji":"📐","tokens_out":4169,"duration_ms":38844,"temperature":0.7,"pith_summary":"The paper aims to transfer the full Euclidean BV and finite-perimeter structure theory to arbitrary smooth Riemannian manifolds, dropping the usual crutches of global curvature bounds or completeness. It does so by replacing the Besicovitch covering lemma (which fails on general manifolds) with a Vitali-relation argument for closed Riemannian balls, and by introducing covector measures as the intrinsic analog of vector-valued measures. The payoff is that De Giorgi's theorem (reduced boundaries are rectifiable, perimeter equals codimension-1 Hausdorff measure), Federer's theorem (measure-theoretic boundary points have density 1/2 almost everywhere), trace and Gauss–Green theorems, and a strict interior approximation result all hold in this generality. The authors demonstrate the power by proving Gamma-convergence of mixed-boundary capillarity functionals on rough domains in manifolds. A careful reader will see that the whole edifice rests on one key analytic premise: closed balls form a Vitali relation for every Radon measure.","feed_headline":"BV structure theory holds on any Riemannian manifold","feed_subtitle":"De Giorgi, Federer, Gauss–Green, and capillarity Gamma-convergence without global curvature or completeness assumptions.","key_machinery":"The load-bearing mechanism is the Vitali relation of closed Riemannian balls: following Federer's general differentiation theory, the paper shows that for every Radon measure on an arbitrary manifold, the family of closed balls (x, ¯B(x,r)) is a Vitali relation, because the Riemannian distance is directionally limited on relatively compact sets. This substitutes for the Besicovitch covering lemma, which is generally false on arbitrary Riemannian manifolds. Around this, the paper builds covector measures (pairs of a Radon measure and a Borel section of the cotangent bundle, with polar decomposition) as the Riemannian analog of vector-valued measures, and a chartwise localization lemma (Lemma","core_discovery":"The central claim is that the classical analytic machinery of BV functions transfers to arbitrary Riemannian manifolds. Concretely, Theorem 4.2 establishes Lebesgue–Besicovitch–Federer differentiation for Radon measures and Theorem 4.3 for covector measures, using only that closed metric balls form a Vitali relation. Theorem 5.15 (De Giorgi) states that the reduced boundary of a finite-perimeter set is H^{m-1}-rectifiable, the perimeter measure equals H^{m-1} restricted to the reduced boundary, and blow-ups converge to half-spaces orthogonal to the measure-theoretic inner normal. Theorem 5.19 (Federer) shows that H^{m-1}-almost every point of the measure-theoretic boundary has density 1/2 an","pith_inferences":["If the Vitali-relation argument is sound, the same structure theorems should extend to any metric space whose closed balls form a Vitali relation for every Radon measure, such as certain sub-Riemannian or Alexandrov spaces; the paper does not assert this, but the machinery appears transferable.","The strict interior approximation theorem may serve as a general template for constructing recovery sequences in Gamma-convergence problems beyond capillarity, including generalized Cheeger or free-boundary problems on manifolds.","The paper's metric-independence results (essential boundary is topological, while the density-1/2 set is Riemannian) suggest that any purely metric definition of reduced boundary must handle this dependence explicitly.","A direct testable extension: check whether the Euclidean intersection-perimeter identity P(E∪F)+P(E∩F)≤P(E)+P(F) used in Theorem 6.12 remains valid with the same constants on manifolds with non-negative Ricci curvature, where geodesic boundaries may deform intersections and unions."],"forward_implications":["De Giorgi's and Federer's structure theorems hold on any smooth Riemannian manifold, so perimeter measures are always H^{m-1} on the reduced boundary and the measure-theoretic boundary is rectifiable up to null sets.","Trace and Gauss–Green theorems hold for BV functions on finite-perimeter strong extension domains, providing the toolset for boundary value problems in manifolds.","Finite-perimeter sets in rough domains can be strictly approximated from the interior while keeping a prescribed Dirichlet boundary portion untouched, enabling relaxation and Gamma-convergence for mixed-boundary variational problems.","Capillarity functionals with mixed boundary conditions admit minimizers and Gamma-converge as adhesion parameters vary, on arbitrary Riemannian manifolds."],"fun_headline_variants":["BV theory on any Riemannian manifold, no completeness needed","BV structure theorems for every Riemannian manifold","De Giorgi-Federer works without curvature or completeness","No global bounds needed for BV theory on manifolds"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that closed balls form a Vitali relation for every Radon measure on an arbitrary Riemannian manifold, since the usual Besicovitch covering lemma is false there; the intersection-perimeter identity used in the approximation theorem is also assumed to transfer verbatim from Euclidean space.","fun_headline_variants_meta":{"raw":{"variants":["BV theory on any Riemannian manifold, no completeness needed","BV structure theorems for every Riemannian manifold","De Giorgi-Federer works without curvature or completeness","No global bounds needed for BV theory on manifolds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000634,"raw_usage":{"total_tokens":2771,"prompt_tokens":762,"completion_tokens":2009,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":506,"completion_tokens_details":{"reasoning_tokens":1946}},"tokens_in":506,"tokens_out":2009,"duration_ms":15205,"temperature":1.0,"reasoning_tokens":1946,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T08:41:55.247558+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a specific non-complete or curvature-unbounded Riemannian manifold and a Radon measure; compute the limsup and liminf of ball averages of a covector measure. If they differ on a set of positive measure, the Vitali relation fails and Theorems 4.2–4.3 collapse. Alternatively, construct two finite-perimeter sets on a manifold where P(E∪F)+P(E∩F) > P(E)+P(F), which would falsify the verbatim transfer of the Euclidean intersection-perimeter identity used in Theorem 6.12 and break the approximation theorem.","supporting_citations":[],"review_version":1}