{"id":"5835f72d-d2ec-4f46-9519-a6e8cfd91eaf","arxiv_id":"2509.01707","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Degenerate cylindrical singular sets of mean curvature flow are locally contained in C^{2,α} submanifolds, with curvature determined by the flow's asymptotic profile.","lead":"This paper shows that degenerate cylindrical singularities of mean curvature flow sit on nearly smooth k-dimensional slices of spacetime, and that the slice's curvature is read from the flow's asymptotic shape. A smart generalist should care because it is a step toward proving that singular sets of melting hypersurfaces are as regular as the surfaces themselves.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uniform Whitney data for S_k(M)^+ depend on unverified quantitative uniqueness [CM25]; if it fails, Theorem 1.1(iii) does not follow.","rationale":"The reader's weakest_assumption correctly identifies the quantitative uniqueness of cylindrical tangent flow [CM25] as the key external input. Examining the proof, Lemma 5.1 Step 1 uses [CM25] to assert that the spine of the tangent flow at any nearby p̄ is close to the spine at p, and Section 6.3 uses it to obtain a uniform δ-L2 closeness of the RMCF over [-1,+∞) for all p̄ in a neighborhood. These uniform estimates are exactly what makes the Whitney data consistent; without them, the error terms in the asymptotic expansions (5.6) cannot be controlled uniformly, and the proof of (5.13) and hence the location estimate (5.2) break down. The rest of the argument, including the asymptotic profile Theorem 4.1 and the Whitney extension Proposition 6.6, appears internally consistent. Since the paper relies on a preprint for the most delicate quantitative step, the correct verdict is conditional on verification of [CM25], matching the reader's assessment. No adjustment is needed.","tokens_in":62069,"tokens_out":19588,"duration_ms":205436,"concrete_test":"Check the precise statement in [CM25] (arXiv:2502.03634) that is invoked in Lemma 5.1 Step 1 and Section 6.3: it must provide, for every ε>0 a δ>0 such that for any unit-regular Brakke flow with entropy ≤ ε^{-1}, if the RMCF at p is δ-L2 close to Cn,k on [-1,+∞), then for every p̄ with parabolic distance d_P(p,p̄) ≤ δ, the spine of the tangent flow at p̄ is within Ψ(δ|n,ε) of the spine of Cn,k, and the RMCF based at p̄ is Ψ(δ|n,ε)-L2 close to Cn,k on [0,1]. If [CM25] only yields this for a single tangent flow or under stronger hypotheses (e.g., a uniform Gaussian density gap), then the estimate (5.3) and the uniform L2-closeness used to apply Theorem 4.1 to every p̄ are not justified, and the location estimate (5.2) is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim Theorem 1.1(iii) rests on the location estimate Lemma 5.1, whose proof uses Colding–Minicozzi [CM25] in two places: (a) Step 1 asserts that for any p̄ in S_k(M)^+ with parabolic distance r ≤ δ, the spine of the tangent flow at p̄ is within angle Ψ(δ|n,ε) of the spine at p, and (b) Section 6.3 asserts that after rescaling, the RMCF based at p̄ is uniformly δ5.1-L2 close to Cn,k on τ ∈ [-1,+∞) for all p̄ in a fixed parabolic neighborhood of p. These are quantitative, uniform statements about a family of singularities. [CM25] is a preprint (arXiv:2502.03634) and the paper does not state precisely which theorem of [CM25] yields these estimates, nor does it supply a proof. If [CM25] only gives closeness of the tangent flow at a single point when the Gaussian density is within a small tolerance, and not uniform L2-closeness over an entire time interval for all nearby singularities, then the asymptotic expansions (5.6) cannot be obtained with the uniform error term needed for (5.13); consequently the Whitney data (5.2) are not uniform and Proposition 6.6 cannot be applied. This is not an internal inconsistency but a dependency on an external, unverified result that is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the k-cylindrical singular set of a unit-regular Brakke mean curvature flow with finite entropy in R^{n+1}. It introduces a relative L2-nonconcentration estimate with respect to a family of 'low spherical flows' and uses it, together with the quantitative uniqueness of Colding–Minicozzi [CM25] and the Whitney extension theorem, to prove Theorem 1.1: for every 1≤k≤n−1, the nondegenerate part S_k(M)^0 has parabolic Hausdorff dimension at most k−1, the degenerate part S_k(M)^+ is relatively closed in S_k(M), and locally S_k(M)^+ is contained in a k-dimensional C^{2,α}-submanifold for α<min{1,2/(n−k)}. It also proves a quantitative asymptotic profile theorem (Theorem 1.4) and uses it to show that, when S_k(M)^+ is a submanifold, its second fundamental form is determined by the leading W_{1/2}-eigenfunction of the rescaled flow.","tokens_in":62437,"tokens_out":4994,"duration_ms":60107,"significance":"If the main theorem holds, it substantially sharpens the known Lipschitz/C^1 containment of the cylindrical singular set (Colding–Minicozzi) to C^{2,α}, and it gives a concrete geometric interpretation of the leading eigenfunction: the curvature of the singular set is read off from the asymptotic profile of the rescaled flow. The proof strategy is original: it corrects the L2-distance by low spherical flows and converts a location estimate at nearby singularities into uniform Whitney data. The paper is carefully organized and contains many quantitative statements, including graphical radius estimates and detailed asymptotic expansions. Its main risk is the heavy reliance on unpublished or companion preprints, especially [CM25], whose precise content is not stated and which is load-bearing for the central regularity theorem.","major_comments":[{"comment":"The proof of the location estimate (5.2) uses [CM25] in two places: Step 1 asserts that for every p̄ in a parabolic neighborhood of p the spine of the tangent flow at p̄ is within Ψ(δ|n,ε) of the spine at p, and Section 6.3 asserts that after rescaling the RMCF based at p̄ is uniformly δ_{5.1}-L2 close to C_{n,k} over τ∈[-1,+∞). The paper does not state which theorem of [CM25] gives these uniformity statements. If [CM25] only provides single-point tangent-flow uniqueness under Gaussian-density closeness, the uniform family version needed for the expansions (5.6) and for the uniform error in (5.13) is not justified. This uniformity is load-bearing for the Whitney data (5.2) and hence for Proposition 6.6 and Theorem 1.1(iii).","section":"Section 5, Lemma 5.1 and Section 6.3"},{"comment":"Lemma A.3 is an a priori estimate with two different exponential weights, stated without proof and justified only by saying the proof is essentially the same as in [Str20]. This lemma is used to prove Lemma A.4, which in turn constructs the low spherical flows used throughout Section 3 and Theorem 1.4. Because the norm ∥·∥_{ℓ,σ,η} in (A.3) differs from the norms in [Str20], the adaptation is not completely formal. The proof, or at least a precise statement of the corresponding result in [Str20], should be supplied.","section":"Appendix A, Lemma A.3"}],"minor_comments":[{"comment":"The title page contains a line break in 'CUR V A TURE'; should read 'CURVATURE'.","section":"Title/Abstract"},{"comment":"'Corollary (3.15)' should be 'Corollary 3.15'.","section":"Section 4.2"},{"comment":"Typo 'fi nises' should be 'finishes'.","section":"Section 6.1"},{"comment":"The preprint [CM25] is cited repeatedly without a theorem number in Section 5; please add the precise version and statement used, or note the relevant theorem number once it is available.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The central result appears plausible and the argument is well structured, but the main theorem depends on a precise uniform quantitative uniqueness statement from [CM25] that is not formulated in the paper. For a journal submission, I would want the authors to either include the exact statement and proof of the needed [CM25] corollary or cite a published version, and to fill the gap in Lemma A.3. This is a fixable issue rather than a reason to reject, but it blocks acceptance on the current version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The main result, Theorem 1.1(iii), genuinely improves the known Lipschitz/C1 containment of Colding–Minicozzi to C^{2,α} for the degenerate k-cylindrical stratum, and Remark 1.7's curvature formula is new. The relative L2-nonconcentration trick—modding out low spherical flows—is a real idea, and the Whitney-data approach is well executed. If the inputs hold, the theorem follows.\n\nBut the inputs are the issue. The location estimate Lemma 5.1 depends on [CM25] in two specific places: uniform closeness of the tangent flow's spine for all nearby singularities, and uniform L2-closeness over a time interval. The paper does not state which theorem of [CM25] gives this, and [CM25] is a preprint. If it only gives single-point closeness, the asymptotic expansions (5.6) lack the uniform error term needed for (5.13), and the Whitney data won't be uniform. This is load-bearing, not decoration. The stress-test note is on target.\n\nThere are also smaller gaps: Lemma A.3 is stated without proof, and several key ingredients come from the authors' own preprints [SWX25, SX25b]. This is self-citation, not circularity, but it means the paper cannot be fully evaluated as a standalone unit. Case (i) of Theorem 1.4 is explicitly credited to [SX25b], so the novelty is concentrated in the degenerate case—which is fine, but worth keeping in mind.\n\nOn balance, the paper is coherent on its own terms, and the central strategy is sound. The concern is an external dependency, not an internal contradiction. That's why my verdict is conditional rather than reject. I'd send it to a serious referee, but with instructions to closely check (a) the exact quantitative uniqueness statement in [CM25] and whether it actually provides the uniform family version, and (b) the absorption step from (5.13) to (5.17). Also ask the authors to prove or precisely cite the needed theorem from [CM25], and to provide the proof of Lemma A.3 or move it to an accessible appendix.\n\nWho should read it? People working on MCF singular sets, particularly those interested in fine strata. It is not a desk reject.","headline":"Strong step on degenerate cylindrical singular sets, but the uniform Whitney estimates ride on an unstated quantitative uniqueness result from a preprint; referee should verify that before accepting.","tokens_in":62928,"tokens_out":2803,"would_cite":true,"duration_ms":28606,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C44","35B65","35K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"Cylindrical singular sets of mean curvature flow are locally C^{2,α} after removing a lower-dimensional subset, and their curvature is determined by the leading eigenfunction of the linearized flow.","keywords":["mean curvature flow","cylindrical singular set","C^{2,α} regularity","tangent flow","rescaled mean curvature flow","Jacobi operator","L² non-concentration","Whitney extension theorem"],"falsifier":"Take a mean curvature flow with two degenerate k-cylindrical singularities approaching a known one, rescale so their parabolic distance is r, and measure the angle between their spines and the quadratic form q. If the quantities r⁻²|Δt|, r⁻¹|Δx − q(Δy)|, ‖ℓ − ∇q(Δy)‖, and r‖Q₂ − Q₁‖ exceed C r^{2γ⁺_{n,k}−10ε} for a sequence r → 0, the Whitney data are incompatible and the C^{2,α} conclusion fails.","tokens_in":61990,"feed_emoji":"🌊","tokens_out":8519,"duration_ms":92412,"temperature":0.7,"pith_summary":"This paper studies the set of points where a mean curvature flow collapses like a round cylinder R^k × S^{n−k}. It proves that after discarding a lower-dimensional subset S_k^0, the remaining degenerate k-cylindrical singular set S_k^+ is locally contained in a k-dimensional C^{2,α}-submanifold; when S_k^+ is itself a submanifold, its curvature is explicitly determined by the leading eigenfunction of the linearized flow. A key consequence is that singular sets, which satisfy no PDE by themselves, inherit a smooth structure from the local dynamics of the singularity. The proof yields a trichotomy for the asymptotic profile near each cylindrical singularity and a precise location estimate that forces nearby singularities to lie on a single surface.","feed_headline":"Remove a lower stratum and cylindrical singular loci are C2,alpha","feed_subtitle":"The curvature of each degenerate cylindrical singularity is encoded in the leading eigenfunction of the linearized flow.","key_machinery":"The engine is a relative L²-distance non-concentration estimate: the weighted Gaussian L² distance from the rescaled flow to a low spherical flow cannot concentrate near the cylinder, yielding a discrete monotonicity of the decay order N_u(τ,M) = log(d_u(τ,M(τ))/d_u(τ+1,M(τ+1))). This forces the decay order to converge to an eigenvalue of the Jacobi operator −L_{n,k} = −Δ_{C_{n,k}} + (1/2)⟨y,∇_y⟩ − 1, after quotienting out the slow spherical eigenmodes that do not affect the geometry of the singular set. The resulting asymptotic-profile trichotomy is converted, through translation, dilation, and rotation unwinding, into a location estimate comparing the spines and curvature forms at two near","core_discovery":"The central claim is Theorem 1.1: for each 1 ≤ k ≤ n−1, the k-cylindrical singular set S_k(M) splits as S_k^0 ∪ S_k^+, where S_k^0 has parabolic Hausdorff dimension at most k−1 (and is isolated when k=1), S_k^+ is relatively closed in S_k(M), and for every α < min{1, 2/(n−k)}, each point of S_k^+ has a neighborhood in which S_k^+ lies inside a k-dimensional C^{2,α}-submanifold. The theorem also bounds the time-image of S_k^+ and, when S_k^+ is a submanifold, identifies its second fundamental form: it is twice the quadratic coefficient appearing in the leading eigenfunction ψ ∈ W_{1/2} of the Jacobi operator, via ψ(θ,y) = ⟨q(y) − 2 tr q, θ/|θ|⟩ plus cubic terms. Thus the curvature of the sing","pith_inferences":["If the quantitative uniqueness comparison at the heart of the proof fails only mildly, the regularity exponent α may drop while containment in a Lipschitz or C^{1,1} surface might survive; a direct test is to push two degenerate singularities together in a rotationally symmetric flow and measure spine-angle versus distance.","If the paper's conjecture on super-exponential decay is true—namely that such flows coincide with a low spherical flow—then case (iii) carries no geometric data, and the curvature formula would extend to all degenerate singularities by continuity.","In semilinear heat equations, curvature of the blow-up set originates from rotation of nondegenerate directions; here it comes from spherical eigenmodes. This suggests analogous C^{2,α} regularity for other geometric flows may come from tracing slow spherical modes rather than spine directions.","The dimensional bound dim_H t(S_k^+) ≤ k/(3+α) could be tested on explicit neckpinch families; a family saturating it would show the estimate is sharp and that the α-limit is natural."],"forward_implications":["The degenerate k-cylindrical singular set, despite being defined by a nonlocal collapse condition and satisfying no PDE, is locally trapped in a smooth surface of dimension k, so counting or integrating over it is justified.","When the singular set is already known to be a k-dimensional submanifold, its curvature is not free: it equals 2A_i determined by the ψ ∈ W_{1/2} eigenmode, and vanishing of those coefficients forces the second fundamental form to vanish.","In R³ for mean-convex or genus-zero starting surfaces, and in higher dimensions for 2-convex initial data, the only nondegenerate lower stratum is isolated, so the singular set near the top stratum is a C^{2,α} curve, ruling out any non-C^{2,α} curve as a singular set.","The parabolic Hausdorff dimension of S_k^0 is at most k−1, and the image of S_k^+ under the time function has Hausdorff dimension at most k/(3 + min{1, 2/(n−k)}).","The graphical radius of the rescaled flow grows at least exponentially in the degenerate case, with rate e^{(γ−ε)τ/[2(γ+1)]} set by the γ-eigenmode, matching the known neckpinch rate up to ε."],"supporting_citations":[{"why":"Supplies the original L²-distance non-concentration property and decay order, which this paper modifies into a relative version modulo low spherical modes.","marker":"[SWX25]"},{"why":"Provides the C¹ normal form theorem and the algebraic asymptotic profile used for the lower stratum S_k^0 and case (i) of the trichotomy.","marker":"[SX25b]"},{"why":"Provides the quantitative uniqueness of cylindrical tangent flow that controls spines and L²-closeness at nearby singularities; this is the load-bearing comparison for the location estimate.","marker":"[CM25]"},{"why":"Gives uniqueness of blowups and Lojasiewicz-type estimates ensuring the rescaled flow is graphical over the cylinder.","marker":"[CM15]"},{"why":"Supplies the Whitney-data strategy for extracting C^{ℓ,α} submanifolds from location estimates, adapted here to the parabolic setting.","marker":"[FS19]"},{"why":"Supplies Brakke–White ε-regularity, used to pass from L²-closeness to smooth graphical control.","marker":"[Whi05]"},{"why":"Introduces the rescaled mean curvature flow and monotonicity formula on which all the asymptotic analysis is built.","marker":"[Hui90]"},{"why":"Provides the nonlinear semigroup construction of flows with prescribed slow spherical asymptotics, adapted in Appendix A for low spherical flows.","marker":"[Str20]"}],"fun_headline_variants":["Cylindrical singular sets are C^{2,α} after removing a thin lower part","MCF: cylindrical singularities locally smooth up to a lower-dimensional set","Curvature of cylindrical singularities from leading eigenfunction","New L^2 non-concentration yields C^{2,α} for cylindrical singular sets","Cylindrical singular sets: C^{2,α} except on small parabolic pieces"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The proof assumes that two nearby degenerate cylindrical singularities have tangent-flow spines that stay uniformly close to each other over a time interval that does not shrink to zero; if that quantitative comparison fails, the location estimate and the C^{2,α} conclusion do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Cylindrical singular sets are C^{2,α} after removing a thin lower part","MCF: cylindrical singularities locally smooth up to a lower-dimensional set","Curvature of cylindrical singularities from leading eigenfunction","New L^2 non-concentration yields C^{2,α} for cylindrical singular sets","Cylindrical singular sets: C^{2,α} except on small parabolic pieces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1386,"prompt_tokens":763,"completion_tokens":623,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":517}},"tokens_in":507,"tokens_out":623,"duration_ms":6045,"temperature":1.0,"reasoning_tokens":517,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T12:16:33.338422+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a mean curvature flow with two degenerate k-cylindrical singularities approaching a known one, rescale so their parabolic distance is r, and measure the angle between their spines and the quadratic form q. If the quantities r⁻²|Δt|, r⁻¹|Δx − q(Δy)|, ‖ℓ − ∇q(Δy)‖, and r‖Q₂ − Q₁‖ exceed C r^{2γ⁺_{n,k}−10ε} for a sequence r → 0, the Whitney data are incompatible and the C^{2,α} conclusion fails.","supporting_citations":[],"review_version":1}