{"id":"ccabeaa6-12be-4848-809a-566a550db8f3","arxiv_id":"1908.05952","paper_version":4,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A convex body with k-th mean curvature at least the sharp volume-to-surface constant at almost all normal points must be a ball, and constant curvature outside a small singular set also forces a ball.","lead":"Here is a proof that a convex shape whose boundary is curved enough at almost every point must be a ball, with a sharp curvature threshold. The result extends classical rigidity theorems to highly non-smooth shapes and solves an open problem about convex hypersurfaces with small singular sets.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof's load-bearing step is property (11) in Theorem 4.1, an absolute continuity of the normal-bundle measure asserted from [San20a, 5.6]; this is false for arbitrary closed sets (e.g., a line segment) and the paper does not show condition (10) implies it.","rationale":"The reader correctly identified the reliance on the author's prior work [San20a] as the weakest assumption. I agree that the proof is internally consistent if those cited results hold. However, I can make the concern more concrete: property (11) is not true for arbitrary closed sets, as the line-segment example shows. The paper's citation of [San20a, 5.6] must therefore depend on condition (10) in a nontrivial way, and the paper does not spell out that dependency. This raises the verification burden but does not constitute a demonstrated mathematical error. The central theorem 5.4 and its proof through Theorems 5.3 and 4.1 are otherwise well-structured: the Newton-McLaurin step, the equality-case argument, and the application of Hartman's theorem and [HHL04] are sound. Since [San20a] is published peer-reviewed work, the dependence is acceptable; I do not see a reason to change the reader's ACCEPT verdict, but the correctness risk remains medium for the stated reason.","tokens_in":17496,"tokens_out":33876,"duration_ms":310657,"concrete_test":"Derive property (11) directly from the definitions in Section 2 and the mean-convexity condition (10), without invoking [San20a, 5.6]. If the derivation fails without additional assumptions, Theorem 4.1 has a gap. As a complementary check, compute H^n(N(C)|S) for C equal to the complement of a convex body with a flat face (e.g., a cube in R^3) and S the edge set; if this quantity is positive, the absolute continuity used in the proof is invalid.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4.1 depends crucially on equation (11): H^n(N(C)|S)=0 whenever H^n(S)=0. This is used to pass from null sets on the base ∂+C to null sets on the unit normal bundle N(C), both in Claim 3 and in the later coarea arguments. The paper derives (11) from [San20a, 5.6] after condition (10) has been used to rule out lower-dimensional strata. However, the property is not true for arbitrary closed sets. For example, if C is a line segment in R^2, the unit normal bundle over the endpoints has positive H^1 measure even though the endpoints have zero H^1 measure. The mean-convexity condition (10) excludes such low-dimensional strata, but the paper does not state the precise hypotheses of [San20a, 5.6] nor prove that condition (10) suffices. If [San20a, 5.6] requires more than condition (10), the proof of Theorem 4.1 has a hidden assumption. This is the single most load-bearing point because the entire volume estimate and equality-case rigidity rest on it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a sharp isoperimetric-type inequality for closed sets satisfying a generalized mean-convexity condition expressed in terms of the trace of the generalized second fundamental form on the unit normal bundle (Theorem 4.1, stated as Theorem 1.3 in the introduction). This inequality is then specialized to convex bodies: Theorem 5.3 gives an isoperimetric inequality involving the pointwise first mean curvature, and Theorem 5.4 characterizes the ball as the unique convex body whose pointwise k-th mean curvature is bounded below by a geometric constant depending only on volume and boundary area. The paper also answers an open question from FLW19 by proving that a convex body whose boundary is C^{1,1} outside a singular set of sufficiently small Hausdorff dimension and whose k-th mean curvature is constant almost everywhere must be a ball (Theorem 5.8). The proofs rely on a substantial apparatus of generalized principal curvatures, approximate mean curvature vectors, and area formulas on unit normal bundles, largely developed in the author's prior work.","tokens_in":17705,"tokens_out":19389,"duration_ms":185586,"significance":"If the results are correct, they constitute a significant contribution: Theorem 5.4 provides a nonsmooth, sharp generalization of the classical Liebmann–Süss and Alexandrov characterizations of the sphere, and the equality case in the general isoperimetric inequality yields a rigidity statement for generalized mean-convex sets. The paper explicitly checks the sharpness example (the union of two antipodal spherical caps) by computation, and the equality-case analysis is detailed and uses a polynomial-volume argument to conclude that the set has positive reach and that its level sets are spheres. The reliance on prior work is heavy but not circular: the convex-body case includes a self-contained proof of the required absolute continuity of the normal-bundle measure (Lemma 5.1), and the general theorem is an independent contribution. The manuscript is careful in its statements and the main line of argument is credible.","major_comments":[{"comment":"In Theorem 5.8, when s=0 and k=n, the conclusion C_{n-k}(K,B)=\\lambda C_n(K,B) becomes C_0(K,B)=\\lambda H^n(B), and the proof invokes [Sch79, Satz 1.2]. However, the introduction explicitly states Schneider's theorem for indices k=1,\\ldots,n-1, and it is not immediate that the index k=0 is covered. The author should clarify the exact statement of the cited theorem or supply an alternative argument for the C_0 case; otherwise the claim for s=0, k=n is not justified as written.","section":"Section 5, proof of Theorem 5.8"}],"minor_comments":[{"comment":"There are several typographical and language issues (e.g., 'difference', 'a such characterization', 'Clos(B)'); these do not affect the mathematics but should be corrected in a revision.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript relies substantially on the author's own prior papers [San19, San20a, MS19], particularly for the generalized curvature framework and the area formula. This is not a circularity problem, but it means a referee cannot fully verify the main theorem without consulting those works; the editor may wish to ensure the cited results are in their published form. The main convex-body theorem appears sound, and the requested changes are local: stating the cited theorem behind (11) and clarifying the index range in Theorem 5.8."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper settles the natural singular analogue of the Liebmann-Suss/Aleksandrov problem for pointwise k-th mean curvature: a pointwise a.e. lower bound forces the body to be a ball. That is a genuinely new result, and it is surprising precisely because pointwise curvature is generically degenerate on typical convex bodies. The proof goes through a new isoperimetric inequality for arbitrary convex bodies (the smooth case was Ros/Montiel-Ros), and the paper also answers the open question from Fang-Lai-Wo about constant higher mean curvature on C^{1,1} convex bodies with small singular sets.\n\nThe structure is right. The general theorem for closed sets (Theorem 1.3) is the core; the convex-body results are corollaries. The equality case is worked out in detail, the two-caps example in Remark 5.5 is computed explicitly, and the stability statement Theorem 5.6 is a decent bonus. The paper is honest about what it assumes.\n\nThe soft spots are concentrated in the load-bearing citations. Theorem 4.1 depends on the author's own prior machinery: the generalized area formula, the trace identity (12), and the absolute-continuity property (11) - that H^n(N(C)|S) = 0 when H^n(S) = 0. The stress-test worry about a line segment does not land: under condition (10) the lower-dimensional strata are excluded, since the infinite principal curvature over them carries positive measure on the normal bundle and (10) forbids exactly that. So the counterexample fails the hypothesis. The broader point does stand: the step from (10) to (11) is a one-sentence citation to [San20a, 5.3, 5.6], and a referee will need to confirm that condition (10) really puts you inside those hypotheses. The cited papers are published in solid journals, so this is a verification task rather than a fatal flaw. I would call it medium risk.\n\nMinor note: the two-caps example does not satisfy the hypothesis of Theorem 5.4 - the flat caps have zero pointwise curvature. Its role is to show the constant is the natural one, and the computation is correct; the exposition could make that role clearer.\n\nWho gets value: convex geometers and anyone working on rigidity from curvature measures. If the referee checks the San20a citations, I expect the paper to hold up. It deserves a serious referee. I would send it out.","headline":"A genuinely new sharp rigidity theorem for pointwise k-th mean curvature of convex bodies, proved through a general isoperimetric inequality; the main risk is the heavy reliance on the author's own prior technical machinery, which a referee should verify.","tokens_in":18254,"tokens_out":11809,"would_cite":true,"duration_ms":109630,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52A20","52A40","53A07"],"pacs":[],"model":"deepseek-v4-flash","headline":"A sharp pointwise curvature condition forces every convex body to be a Euclidean ball.","keywords":["convex bodies","pointwise k-th mean curvature","isoperimetric inequality","characterization of the ball","generalized principal curvatures","singular convex hypersurfaces","curvature measures","positive reach"],"falsifier":"If the theorem is right, the inequality must fail on the flat equatorial region of any non-spherical convex body of revolution. For instance, take the ellipsoid with semi-axes $1,1,2$ in $\\mathbb R^3$ and compute the pointwise mean curvature $H_1$ at an equatorial point, comparing it with $2\\mathcal H^2(\\partial K)/(3\\mathcal L^3(K))$; the comparison must come out negative. Finding any non-ball body on which the pointwise inequality is nonnegative almost everywhere would refute the main theorem.","tokens_in":17265,"feed_emoji":"🔵","tokens_out":14233,"duration_ms":135514,"temperature":0.7,"pith_summary":"The paper proves a sharp rigidity theorem: if a convex body in $\\mathbb R^{n+1}$ has pointwise $k$-th mean curvature at least $\\bigl(\\mathcal H^n(\\partial K)/((n+1)\\mathcal L^{n+1}(K))\\bigr)^k\\binom{n}{k}$ at almost every boundary point, then the body must be a Euclidean ball. This matters because it extends classical smooth rigidity results to a pointwise notion of curvature that imposes no regularity on the boundary. The proof rests on a new isoperimetric inequality for arbitrary convex bodies, $\\mathcal L^{n+1}(K) \\le \\frac{n}{n+1}\\int_{\\partial K} \\frac{1}{H_1(K,x)}\\,d\\mathcal H^n x$, whose equality case forces the ball. The same inequality is proved for a broad class of closed sets with a generalized mean-convexity condition, and the paper also answers an open conjecture about convex bodies with small singular sets and constant higher-order mean curvature.","feed_headline":"A curvature bound forces convex bodies to be balls","feed_subtitle":"A sharp pointwise inequality on the k-th mean curvature singles out the ball, with no smoothness required.","key_machinery":"The engine of the proof is the generalized curvature calculus for arbitrary closed sets. For a closed set $C$ one forms its generalized unit normal bundle $\\mathcal N(C)$ and the generalized principal curvatures $\\kappa_{C,1},\\dots,\\kappa_{C,n}$; for a convex body these coincide at normal points with the eigenvalues of the second-order differential of the convex function representing the boundary. Setting $C=\\mathbb R^{n+1}\\setminus\\operatorname{int}(K)$, the mean-convexity condition $\\sum_i\\kappa_{C,i}\\le 0$ a.e. holds automatically, and the trace of the generalized second fundamental form is identified with $-h\\cdot\\eta$, where $h$ is the approximate mean curvature vector of the positive boundary $\\partial_+ C$. The volume estimate comes from applying the generalized area formula to the map $(z,\\eta,t)\\mapsto z+t\\eta$ over the normal bundle, using the arithmetic-geometric mean inequality to bound the Jacobian; the equality case then shows that the parallel hypersurfaces $S(C,r)$ are umbilical $C^{1,1}$ hypersurfaces, all principal curvatures equal almost everywhere, which by the paper's Theorem 3.1 forces them to be spheres, and polynomial volume growth forces positive reach, yielding the finite union of balls.","core_discovery":"Let $K$ be a convex body in $\\mathbb R^{n+1}$, and let $H_k(K,x)$ be the pointwise $k$-th mean curvature defined at normal boundary points. The central theorem states that if, for some $k=1,\\dots,n$, the inequality $H_k(K,x) \\ge \\bigl(\\mathcal H^n(\\partial K)/((n+1)\\mathcal L^{n+1}(K))\\bigr)^k\\binom{n}{k}$ holds for $\\mathcal H^n$-almost every $x\\in\\partial K$, then $K$ is a ball. The constant is sharp: the convex body formed by two proper antipodal spherical caps has $H_k$ equal to $\\binom{n}{k}$ on almost all of its boundary, but its volume-to-area ratio makes the right-hand side larger than $\\binom{n}{k}$, so the hypothesis barely fails. The theorem is deduced from an isoperimetric inequality for arbitrary convex bodies, $\\mathcal L^{n+1}(K) \\le \\frac{n}{n+1}\\int_{\\partial K} 1/H_1(K,x)\\,d\\mathcal H^n x$, whose equality case, under a boundedness assumption on $H_1$, forces $K$ to be a ball. This inequality is proved in a wider setting: for any closed set $C$ whose generalized principal curvatures satisfy $\\sum_i \\kappa_{C,i} \\le 0$ almost everywhere on the generalized unit normal bundle, the volume of $\\mathbb R^{n+1}\\setminus C$ is bounded by $\\frac{n}{n+1}\\int_{\\partial_+ C} 1/|h|\\,d\\mathcal H^n$, where $h$ is the approximate mean curvature vector of the positive boundary; equality plus boundedness of $|h|$ implies the complement is a finite union of disjoint open balls. A final result resolves the open conjecture from [FLW19]: a convex body of class $C^{1,1}$ outside a singular set of zero $\\mathcal H^s$ measure with constant $k$-th mean curvature on its regular part, $1\\le k\\le n-s$, must be a ball.","pith_inferences":["The same isoperimetric mechanism suggests a quantitative stability route: the gap $\\mathcal L^{n+1}(K)-\\frac{n}{n+1}\\int_{\\partial K}1/H_1\\,d\\mathcal H^n$ is a natural, scale-dependent defect that vanishes only for balls, so a quantitative lower bound in terms of a shape asymmetry is a plausible next step.","Because the proof converts $H_k$ bounds into $H_1$ bounds through the Newton-McLaurin inequality, analogous characterizations may hold for other symmetric functions of the principal curvatures whenever the same inequality chain applies.","The sharp threshold could serve as a scale-invariant sphericity diagnostic in shape analysis: any convex body violating the pointwise inequality is certified non-spherical, and the theorem guarantees no non-ball can satisfy it.","Since the mean-convexity condition also covers complements of mean-convex level-set flows and sets of finite perimeter with bounded distributional mean curvature, the equality analysis may yield rigidity statements for those broader classes, which the paper notes but does not explore."],"forward_implications":["A purely pointwise curvature bound, carrying no implicit regularity, now suffices to force a convex body to be a ball; the two-cap example shows the threshold cannot be lowered.","Every convex body satisfies the new isoperimetric inequality $\\mathcal L^{n+1}(K) \\le \\frac{n}{n+1}\\int_{\\partial K} 1/H_1(K,x)\\,d\\mathcal H^n x$, with equality characterizing balls under the boundedness condition $H_1\\le q$.","For generalized mean-convex closed sets, equality in the sharp volume bound forces the complement to be a finite union of disjoint open balls of radius at least $n/q$.","The open problem from [FLW19] is settled: a convex body that is $C^{1,1}$ outside a singular set of zero $\\mathcal H^s$ measure and has constant $k$-th mean curvature on its regular part, with $1\\le k\\le n-s$, must be a ball.","The $L^1$ stability statement shows that a Hausdorff limit of convex bodies whose pointwise $H_k$ approaches the threshold in $L^1$ is necessarily a ball."],"supporting_citations":[{"why":"Builds the generalized unit normal bundle, generalized principal curvatures, and the generalized area formula that convert normal-bundle integrals into volume estimates.","marker":"[San20a]"},{"why":"Provides the stratum decomposition showing the positive boundary is countably $\\mathcal H^n$-rectifiable of class 2.","marker":"[MS19]"},{"why":"Supplies the second-order approximate differentiability used to define approximate second fundamental forms.","marker":"[San19]"},{"why":"Yields the notion of reach and the normal-bundle facts used in the equality case.","marker":"[Fed59]"},{"why":"Establishes that polynomial parallel volume forces positive reach, the step that turns the equality analysis into a finite union of balls.","marker":"[HHL04]"},{"why":"Gives the curvature-measure rigidity theorem invoked to conclude the ball in the constant-$H_k$ corollary.","marker":"[Sch79]"},{"why":"Proves a weaker version of the isoperimetric inequality for viscosity mean-convex sets, which the paper improves and extends.","marker":"[DM19]"},{"why":"States the open conjecture about singular convex hypersurfaces that the paper answers.","marker":"[FLW19]"}],"fun_headline_variants":["A curvature bound leaves only the ball as convex body","Sharp curvature inequality implies ball, no smoothness required","Ball is unique convex body with a k-mean curvature bound","Pointwise curvature bound uniquely identifies the ball","No smoothness needed: bound forces convex bodies to balls"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes the validity of the generalized curvature formalism for arbitrary closed sets, specifically the identity linking the trace of the generalized second fundamental form to the approximate mean curvature vector and the generalized area formula used to pass from normal-bundle integrals to volume; the paper cites these tools from earlier work rather than reproving them.","fun_headline_variants_meta":{"raw":{"variants":["A curvature bound leaves only the ball as convex body","Sharp curvature inequality implies ball, no smoothness required","Ball is unique convex body with a k-mean curvature bound","Pointwise curvature bound uniquely identifies the ball","No smoothness needed: bound forces convex bodies to balls"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000792,"raw_usage":{"total_tokens":3576,"prompt_tokens":1119,"completion_tokens":2457,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":735,"completion_tokens_details":{"reasoning_tokens":2380}},"tokens_in":735,"tokens_out":2457,"duration_ms":15884,"temperature":1.0,"reasoning_tokens":2380,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:00:31.689961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If the theorem is right, the inequality must fail on the flat equatorial region of any non-spherical convex body of revolution. For instance, take the ellipsoid with semi-axes $1,1,2$ in $\\mathbb R^3$ and compute the pointwise mean curvature $H_1$ at an equatorial point, comparing it with $2\\mathcal H^2(\\partial K)/(3\\mathcal L^3(K))$; the comparison must come out negative. Finding any non-ball body on which the pointwise inequality is nonnegative almost everywhere would refute the main theorem.","supporting_citations":[],"review_version":1}