{"id":"e199cd12-9ff4-4d9e-8100-7ce726a99ee5","arxiv_id":"1908.02871","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Axially symmetric mean curvature flow with Neumann boundary develops only type I singularities at the first singular time.","lead":"This paper claims that every singularity that forms when an axially symmetric surface with right-angle boundary data shrinks under mean curvature flow is a type I singularity. If true, it would complete the classification started by Huisken, but the proof has serious gaps.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.1 invalid at the catenoid-limit contradiction: the convergence index I0 depends on the point, so fixing i and letting the catenoid coordinate tend to infinity is not justified.","rationale":"The paper's central claim, Theorem 1.1, rests on showing that all singularities must fall into the type I case. The proof splits singularity formation into three cases and uses Theorem 6.1 to rule out bounded mean curvature and Theorem 6.2 to rule out |A|^2/H^2→∞. Both theorems share the same catenoid-limit contradiction. The reader's weakest_assumption correctly identifies the flaw: the pointwise convergence estimate has an index I0(l) that depends on the point, and the proof illegitimately fixes i and then lets l run to infinity on the catenoid. This is not a minor gap; it is the exact step that produces the contradiction. Without an additional argument controlling the convergence rate uniformly on expanding domains, the contradiction does not follow. The paper also relies on an unproved non-cylindrical maximum principle in Proposition 7.3 and on unpublished preprints for key estimates, but the most load-bearing and clearly identifiable failure is the quantifier exchange in Theorem 6.1. The static corollary about H<0 may be true independently, but the proof given for the main theorem is incomplete. Hence the reader's REJECT verdict is justified, and no change to the verdict is needed.","tokens_in":8177,"tokens_out":12995,"duration_ms":140839,"concrete_test":"Formalize the step: for each R>0, compact convergence gives I0(R). The contradiction would require choosing R_i→∞ with i>I0(R_i) and α_i^{-1} cosh(2c^{-1}R_i)→∞. Test whether the stated convergence argument provides such a sequence. Concretely, take any sequence of rescaled flows converging smoothly to the catenoid on compact sets and compute or bound I0(R)=min{i: sup_{|\\x1|≤R}|\\tilde_v_i\\tilde_y_i - \\hat_v\\hat_y|<ε}. Show that for every fixed i the set of points with I0(l)≤i is bounded, which blocks the 'for fixed i' contradiction. The paper does not supply the required rate control; without it, Theorem 6.1 is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 6.1 (Section 6) rescales at a supposed bounded-H singularity, obtains a stationary limit, identifies it with the catenoid, and then tries to contradict Lemma 3.2. The critical step reads: 'For any ε>0 and for any l∈M^2 we can find I0∈N such that ... for all i>I0' and then 'For fixed i, the left-hand side can be made as large as we like.' This is a quantifier error. The threshold I0 depends on l (equivalently on the catenoid coordinate \\x1); convergence to the catenoid is only asserted on compact subsets (Section 5). Fixing i first and sending \\x1→∞ requires the inequality to hold for that single i at arbitrarily large \\x1, which is not established and is generally false for smooth convergence on compact sets. Also, after dividing by α_i, the inequality is consistent with the rescaled form of Lemma 3.2 (which gives \\tilde_v_i\\tilde_y_i ≤ α_i c), so for fixed \\x1 the left-hand side tends to 0 as i→∞; the claimed contradiction depends entirely on the invalid exchange of limits. Since Theorem 6.2 invokes the same argument, the exclusion of both the bounded-H case and the |A|^2/H^2→∞ case is not proved. Theorem 1.1 therefore lacks its main structural step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies smooth, compact, axially symmetric mean curvature flow in R^3 with Neumann boundary on two coaxial planes. The announced theorem (Theorem 1.1) asserts that at the first singular time all singularities are of type I, without any sign assumption on mean curvature. The proof strategy is to rule out (a) singularities with bounded mean curvature and (b) singularities where |A|^2/H^2 tends to infinity by rescaling and obtaining a catenoid limit; the remaining case |A|^2/H^2 bounded is then handled by Huisken-type estimates. The paper also presents a static corollary that no such axially symmetric surface with Neumann boundary can have H<0 everywhere.","tokens_in":8397,"tokens_out":7205,"duration_ms":80658,"significance":"If the classification were proved, it would strengthen Huisken's type-I result in an axially symmetric setting with free boundary and would be a useful contribution, with a clean corollary excluding H<0 everywhere. The paper is well organized and includes explicit proofs for several auxiliary estimates, such as the height bound and the curvature estimate in Proposition 3.4. However, the central proof currently relies on a limit-exchange step that is not justified in Theorem 6.1, and several foundational lemmas are cited to unpublished preprints. The main theorem is therefore not established in the submitted form.","major_comments":[{"comment":"The advertised contradiction from the catenoid comparison rests on an unjustified exchange of quantifiers. The paper establishes, for every ε>0 and every l in M^2, an index I0(l) such that for all i>I0(l) one has v_hat(l)y_hat(l)-ε ≤ v_tilde_i(l,τ0)y_tilde_i(l,τ0), and hence (v_hat(l)y_hat(l)-ε)/α_i ≤ c. The next sentence, however, fixes i and lets l run to infinity on the catenoid, claiming the left-hand side becomes arbitrarily large. This requires the inequality for one fixed i at arbitrarily large l, whereas the convergence statement in Section 5 is only on compact subsets and I0(l) may grow with l. For each fixed l the divided inequality is perfectly consistent with Lemma 3.2 because α_i tends to infinity, so the contradiction does not follow. Since Theorem 6.2 repeats the same argument, the exclusion of both the bounded-mean-curvature case and the case |A|^2/H^2→∞ is not proved; these exclusions are the main structural step in Theorem 1.1.","section":"Section 6, proof of Theorem 6.1"},{"comment":"Lemma 3.2, the gradient estimate yv≤c, and the rescaling procedure are quoted from the unpublished preprint [8], while Proposition 4.1 is quoted from [5]. These results are load-bearing: Lemma 3.2 underlies Proposition 3.4 and the compactness argument in Section 5, and Proposition 4.1 is essential to restrict singularities to Ω∖Ω_tilde^-. As submitted, a substantial part of the argument is inaccessible to the reader; the authors should include proofs of these results or replace the references with published versions.","section":"Sections 3 and 5"},{"comment":"The non-cylindrical maximum principle is stated without proof and without a precise theorem reference, although it is used in Proposition 7.1 to pass from the PDE for q/H to a boundary supremum over a moving space-time domain. Since the hypotheses involve a vector field a that is only assumed continuous near maxima and the boundary set δV is nonstandard, this is not a routine citation as written. The type I estimate for the remaining case is therefore incomplete.","section":"Appendix, Proposition 7.3"}],"minor_comments":[{"comment":"The bulleted case list is typeset incorrectly (showing '/Bullet') and the three cases are not labelled; this impedes reading.","section":"Section 5"},{"comment":"The notation τ0 ∈ (-α_{I0}^2 t_{I0},0) is unclear because I0 is chosen in the same sentence as τ0; the order of quantifiers should be stated more carefully.","section":"Section 6, proof of Theorem 6.1"},{"comment":"The comparison argument involving the enclosing cylinder should specify the comparison principle and the boundary conditions being used, since the cylinder does not have the same Neumann boundary data as the evolving surface.","section":"Remark 7.2"}],"recommendation":"reject","confidential_remarks":"Given the limit-exchange error in Theorem 6.1, the paper should not be accepted in its current form. The reliance on two unpublished preprints by the same authors is a further barrier to verification. If the authors can supply a corrected limit argument and complete proofs of the cited estimates, the result would merit reconsideration."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the target result is real and worth having: removing Huisken's H>0 assumption in the axially symmetric Neumann-boundary setting would complete the singularity-type classification for surfaces of revolution. Second, the paper as written does not prove it. Theorem 6.1 has a quantifier error at the catenoid-limit contradiction, and since Theorem 6.2 leans on the same step, the central type I classification is unsupported.\n\nWhat the paper does well: it correctly identifies that singularities in regions of negative mean curvature cannot occur, and the a priori estimates in Sections 3–4 are mostly standard but competently assembled. The rescaling setup follows [8] and is fine. The final section 7 is a careful adaptation of Huisken's type I argument to a Neumann boundary, modulo the stated non-cylindrical maximum principle.\n\nThe gap: after rescaling to a stationary catenoid limit, the proof says that for any ε and any point l there is I0(l) such that for all i > I0(l) the lower bound holds. Then it says for fixed i the left-hand side can be made arbitrarily large. That only works if you can take l large while keeping i fixed, but the convergence to the catenoid is only on compact sets, and I0(l) can spoil exactly that. Fixing i first gives only a finite, l-dependent region. So the contradiction does not follow. The same flaw propagates into Theorem 6.2. This is not a cosmetic issue; it is the step that forces the bounded-H case to be type I.\n\nTwo structural weaknesses are also worth noting. Several key estimates—Lemma 3.2, the rescaling procedure, the maximum principle—are cited to unpublished preprints [5,8] by the same authors. That is not circular, but it makes verification dependent on material the reader cannot see. And the non-cylindrical maximum principle is stated in an appendix without proof; the authors point to [5] for details, which is the same inaccessible source.\n\nWho this is for: someone working on geometric flows, especially mean curvature flow of surfaces of revolution, might want to see the approach, but should not rely on the conclusion. A referee could reasonably send it back for major revision: repair Theorem 6.1, prove or properly cite the key lemmas, and make the maximum principle argument self-contained. The paper deserves a serious referee, but my own judgment is that it is not currently acceptable.\n\nRecommendation: send to peer review with a request for major revision. The core idea is worthwhile; the proof is not there yet.","headline":"The target result is genuine and attractive, but a quantifier error at the catenoid-limit contradiction in Theorem 6.1 breaks the main proof, so the paper is not acceptable as written.","tokens_in":8969,"tokens_out":3333,"would_cite":false,"duration_ms":35967,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C44","35K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that every singularity at the first singular time of a smooth axially symmetric mean curvature flow in R^3 with Neumann boundary is of type I: curvature blows up at most like C/(T-t), with no sign restriction on mean…","keywords":["mean curvature flow","axially symmetric surfaces","Neumann boundary","type I singularity","singularity classification","parabolic rescaling","catenoid","negative mean curvature"],"falsifier":"Compute, along the rescaled sequence $\\tilde M_{i,\\tau}$, the quantity $\\sup_l \\tilde v_i(l,0)\\tilde y_i(l,0)$ for a sequence of points moving out along the axis; if the supremum approaches the catenoid's unbounded value while the gradient bound $\\tilde v_i\\tilde y_i \\le C$ holds, the contradiction claimed in Theorem 6.1 fails. More directly, exhibiting one sequence of bounded-mean-curvature flows whose rescaled limit is a catenoid and for which the lower bound $\\hat v(l)\\hat y(l)-\\epsilon$ holds only for $i>I_0(l)$ with $I_0(l)$ unbounded in $l$ would settle the proof's gap.","tokens_in":7919,"feed_emoji":"🌀","tokens_out":13894,"duration_ms":131579,"temperature":0.7,"pith_summary":"This paper studies mean curvature flow of smooth, compact, axially symmetric surfaces in $\\mathbb{R}^3$ whose boundary meets two parallel planes orthogonally. It aims to prove that at the first singular time every singularity is of type I, meaning the maximum curvature blows up no faster than $C/(T-t)$. If true, this is a complete classification of singularity types for this symmetry class, and it removes the positive-mean-curvature assumption that previous work required. A direct corollary is that no axially symmetric surface with this Neumann boundary condition can have $H<0$ everywhere.","feed_headline":"All singularities in axially symmetric mean curvature flow are type I","feed_subtitle":"A complete singularity classification for axially symmetric Neumann surfaces, with no sign restriction on curvature.","key_machinery":"Three estimates carry the argument: the height bound $y\\ge c$ in negative mean curvature regions, the gradient bound $yv \\le C$, and the curvature ratio bound $k/p \\le C$. On top of these, the paper uses the parabolic rescaling (5.2) centered at a maximum-curvature point, which preserves axial symmetry and yields uniform $C^\\infty$ convergence to a limit flow. For the non-type-I cases the limit is stationary and hence the catenoid, the minimal surface of revolution obtained by rotating $\\hat y = c\\cosh(\\hat x_1/c)$ about the axis; the product $\\hat v \\hat y$ is unbounded along the axis, while the gradient estimate makes the rescaled product bounded, producing the contradiction. For the type-I case the key object is the evolution equation for $q/H$, combined with the non-cylindrical maximum principle (Proposition 7.3), which turns the boundary and neighborhood information into the bound $|q| \\le C p$ and hence the type-I curvature bound.","core_discovery":"The paper's central claim is Theorem 1.1: in the axially symmetric Neumann-boundary setting, no condition on the sign of the mean curvature is needed, and all singularities at the first singular time are type I. The strategy is to show that singularities cannot occur in regions of negative mean curvature, where direct estimates keep the full curvature tensor bounded. If a singularity occurs while the mean curvature is bounded, or while $|A|^2/H^2\\to \\infty$, a parabolic rescaling around the maximum-curvature point produces a stationary limit, which must be the catenoid; the catenoid's height-gradient product $\\hat v \\hat y$ grows without bound along the axis, contradicting the gradient estimate $yv \\le C$. The only remaining case has $H\\to\\infty$ with $|A|^2/H^2$ bounded, and there an adaptation of the $q/H$ evolution argument from [10] yields $|A|^2 \\le C/(T-t)$.","pith_inferences":["If the quantifier gap in Theorem 6.1 is repaired by a uniform counterpart of the pointwise lower bound, the classification would be fully established; the natural route is to control $I_0(l)$ through an integral or maximum-principle estimate rather than pointwise convergence.","The same height-gradient-catenoid obstruction should rule out type II singularities in higher-dimensional axially symmetric flows with boundary, because the unbounded $\\hat v \\hat y$ product is a one-dimensional feature of the catenoid profile.","A numerical test could evolve many axially symmetric initial surfaces with systematically negative mean curvature under Neumann boundary conditions and check whether the flow develops a singularity; the paper predicts none until the global extinction time.","In the volume-preserving axially symmetric problem studied by the same authors, the same split into negative-curvature regions and a catenoid limit may give a singularity classification without assuming volume preservation."],"forward_implications":["At the first singular time the curvature satisfies $|A|^2 \\le C/(T-t)$, so no type II blow-up can occur in this symmetry class.","No singularity can occur while the mean curvature stays bounded, and in negative mean curvature regions the full curvature tensor is bounded.","Singularities are confined to regions where $H\\to\\infty$ and $|A|^2/H^2$ remains bounded.","No smooth axially symmetric surface with Neumann boundary in $\\mathbb{R}^3$ can have $H<0$ everywhere, a static statement independent of the flow.","The classification covers initial data whose mean curvature changes sign and stays negative up to the singular time, extending the earlier positive-mean-curvature result to that broader class."],"supporting_citations":[{"why":"Supplies the positive-mean-curvature type-I estimate, including the $q/H$ maximum principle and $k/p$ bound that section 7 adapts to Neumann boundary.","marker":"[10]"},{"why":"Provides the gradient estimate $yv\\le C$ and the maximum-curvature rescaling procedure used throughout the blow-up analysis.","marker":"[8]"},{"why":"Gives the negative-mean-curvature curvature estimates and boundary treatment that section 4 generalizes to this setting.","marker":"[5]"},{"why":"Establishes that axially symmetric singularities are finite, occur on the axis, and provides the space-time neighbourhood structure and height lower bound.","marker":"[1]"},{"why":"Provides the evolution equations and parabolic derivative estimates used to get uniform $C^\\infty$ convergence of the rescaled flows.","marker":"[9]"},{"why":"Underlies the non-cylindrical maximum principle used to control boundary terms in Proposition 7.1.","marker":"[7]"},{"why":"Extends the parabolic maximum principle to non-cylindrical domains, which the appendix adapts for the boundary terms.","marker":"[12]"}],"fun_headline_variants":["All singularities type I in axially symmetric Neumann flow","Axially symmetric Neumann MCF: singularities always type I","No sign condition needed for type I singularities in Neumann flow","Axially symmetric Neumann flow: every singularity is type I"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument in Theorem 6.1 requires a pointwise lower bound on the rescaled surfaces to remain valid as the spatial point is taken to infinity along the catenoid, even though convergence is only established on compact subsets.","fun_headline_variants_meta":{"raw":{"variants":["All singularities type I in axially symmetric Neumann flow","Axially symmetric Neumann MCF: singularities always type I","No sign condition needed for type I singularities in Neumann flow","Axially symmetric Neumann flow: every singularity is type I"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000584,"raw_usage":{"total_tokens":2654,"prompt_tokens":759,"completion_tokens":1895,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":375,"completion_tokens_details":{"reasoning_tokens":1826}},"tokens_in":375,"tokens_out":1895,"duration_ms":17271,"temperature":1.0,"reasoning_tokens":1826,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:32:42.785332+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, along the rescaled sequence $\\tilde M_{i,\\tau}$, the quantity $\\sup_l \\tilde v_i(l,0)\\tilde y_i(l,0)$ for a sequence of points moving out along the axis; if the supremum approaches the catenoid's unbounded value while the gradient bound $\\tilde v_i\\tilde y_i \\le C$ holds, the contradiction claimed in Theorem 6.1 fails. More directly, exhibiting one sequence of bounded-mean-curvature flows whose rescaled limit is a catenoid and for which the lower bound $\\hat v(l)\\hat y(l)-\\epsilon$ holds only for $i>I_0(l)$ with $I_0(l)$ unbounded in $l$ would settle the proof's gap.","supporting_citations":[{"cited_title":"Asymptotic behavior for singularities of the mean curva ture flow","cited_arxiv_id":null,"evidence_quote":"Supplies the positive-mean-curvature type-I estimate, including the $q/H$ maximum principle and $k/p$ bound that section 7 adapts to Neumann boundary."},{"cited_title":"On the extension of axially symmet- ric volume preserving mean curvature flow","cited_arxiv_id":null,"evidence_quote":"Provides the gradient estimate $yv\\le C$ and the maximum-curvature rescaling procedure used throughout the blow-up analysis."},{"cited_title":"Singularities of axially symmetric volume preserving mean curvature flow","cited_arxiv_id":null,"evidence_quote":"Gives the negative-mean-curvature curvature estimates and boundary treatment that section 4 generalizes to this setting."},{"cited_title":"B., AND GIGA , Y","cited_arxiv_id":null,"evidence_quote":"Establishes that axially symmetric singularities are finite, occur on the axis, and provides the space-time neighbourhood structure and height lower bound."},{"cited_title":"Flow by mean curvature of convex surfaces into spheres","cited_arxiv_id":null,"evidence_quote":"Provides the evolution equations and parabolic derivative estimates used to get uniform $C^\\infty$ convergence of the rescaled flows."},{"cited_title":"Regularity theory for mean curvature flow","cited_arxiv_id":null,"evidence_quote":"Underlies the non-cylindrical maximum principle used to control boundary terms in Proposition 7.1."},{"cited_title":"Principes du maximum paraboliques pour des domaines (x, t) non- cylindriques","cited_arxiv_id":null,"evidence_quote":"Extends the parabolic maximum principle to non-cylindrical domains, which the appendix adapts for the boundary terms."}],"review_version":1}