{"id":"06892893-189e-4faf-b1e2-0c723dc5404d","arxiv_id":"2412.10581","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every ancient noncollapsed mean curvature flow in R^4 is one of the known shrinkers, bowls, ovals, or Hoffman-Ilmanen-Martin-White translators.","lead":"This paper completes the classification of all possible ancient noncollapsed mean curvature flows in four-dimensional space. It gives a definitive list of every singularity shape that can appear, closing a long-running research program in geometric analysis.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader identified Hamilton's Harnack inequality as the weakest assumption. I agree that this is the step with the least direct justification inside the paper, but the cited references appear to cover the noncollapsed ancient setting, and the uses in Sections 10 and 11 are standard. I did not find a circular step, a hidden parameter, or an omitted proof that would change the verdict. The proof is long and relies on several prior results, including the unpublished [CDD+], but that is a dependency rather than a detected flaw. The reader's ACCEPT verdict with moderate confidence remains appropriate.","tokens_in":50639,"tokens_out":19221,"duration_ms":169214,"concrete_test":"Check the hypotheses of the Harnack inequality in [DS23, BLL23] at each invocation in Sections 10-11, in particular that the ancient noncollapsed flows considered satisfy the conditions needed for (43) and for the rigidity case that yields H(P_h) = 1. If any invocation uses a noncompact or non-graphical solution outside the theorem's scope, that would mark a gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After a full read, I do not identify a load-bearing concern that would invalidate the central classification. The most sensitive input is Hamilton's Harnack inequality (B_tH >= 0), used after (43) and in Sections 10-11 to compare cap speeds and to force selfsimilarity via the rigidity case of Harnack; however, the authors cite [Ham95, DS23, BLL23] for exactly this setting, and the applications are consistent with the stated hypotheses. The proof is technically dense, but the main bootstraps (Merle-Zaag switch dynamics, anisotropic propagation of smallness, differential neck theorem, and the final area-width comparison) are internally coherent. The only caveat is reliance on the unpublished [CDD+] for the slow-convergence case (Theorem 1.6), which is a dependency but not an identified error.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper completes the classification programme for ancient noncollapsed mean curvature flows in R^4. After reducing to the case where the tangent flow at -infinity is a bubble-sheet with mixed convergence, the authors introduce a localized slope function and prove a differential neck theorem (Theorems 1.10 and 9.9) showing that the relevant slope is, up to a universal error, a e^{tau/2} with a nonzero constant a(M). They then use this theorem to rule out compact exotic ovals and to show that every noncompact strictly convex ancient noncollapsed flow is selfsimilarly translating. Combining this with the prior fast-convergence, slow-convergence, and translator classifications, they obtain the complete list in Theorem 1.2 and the canonical-neighborhood consequence in Corollary 1.3.","tokens_in":50717,"tokens_out":7764,"duration_ms":82557,"significance":"If correct, this is a decisive result: it settles the classification of ancient noncollapsed mean curvature flows in R^4 and hence gives a complete list of possible blowup limits of mean-convex flows in four dimensions, answering questions of White and Wang. The paper contains a substantial amount of novel hard analysis: Merle-Zaag switch dynamics for the slope function, anisotropic barriers and propagation of smallness estimates, and a quantitatively sharp differential neck theorem. The main theorem is not an input to the proof: the constant a(M) is derived and shown to be nonzero by contradiction, and the final rigidity step is obtained from the cited Harnack inequality rather than from assuming the classification. The proof is explicit about the exponents involved, and I found no internal contradiction or circular dependence. The principal caveat is the reliance on the companion paper [CDD+] for the slow-convergence case, which the manuscript states is to appear; I did not find any indication that this dependence is circular, but the journal should confirm that all cited 'to appear' results are in final form.","major_comments":[],"minor_comments":[{"comment":"The inequality H(p,t) <= <x,nu>/|t| is stated for a flow normalized so that 0 is in K_0; for a general ancient solution one should explicitly translate in space-time. The intended meaning is clear from the context, but a one-line clarification would prevent misreading.","section":"Section 2, after (43)"},{"comment":"The quantity omega(tau) is defined with sup over tau' <= min{tau, tau_sp(X)}, whereas the introductory version in (13) uses sup over tau' <= tau. Please align the two definitions or explain why the switch-time truncation is harmless there.","section":"Section 6.1, near (189)"},{"comment":"The text refers to 'Proposition 11.3 (eccentricity scale)', but the statement with that content is Corollary 11.3. The cross-reference should be corrected.","section":"Section 11.3, after (323)"},{"comment":"The effective range in which the strong error estimate is really useful is tau <= tau_* - (7/4) log max{1,Z}^2, as the authors explain in the paragraph after (7). It would help the reader if this range were displayed as part of the theorem statement itself.","section":"Theorem 1.10"},{"comment":"In the comparison argument involving the function phi_s, the notation t in (337) and (338) is used both for the time variable and for the fixed time T?delta; writing the fixed time as t_0 or s would remove the ambiguity.","section":"Section 11.3, proof of Claim 11.14"}],"recommendation":"accept","confidential_remarks":"My residual concern is external rather than internal: Theorem 1.2 depends on [CDD+] and on [DH], both currently listed as 'to appear'. I could not verify those inputs independently, but the manuscript is explicit about the dependence and I found no sign that the present paper assumes its own conclusion. If the editors are satisfied that the companion papers are in final form, acceptance is appropriate; otherwise the theorem should be marked conditional on those papers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: the theorem is the real thing. It closes the mixed-convergence case that the five predecessor papers explicitly left open, and the list in Theorem 1.2 is the expected one. The new machinery—switch dynamics, anisotropic barriers, propagation of smallness, and especially the differential neck theorem with the explicit 7/3 and 2/3 exponents—is substantial. These are not cosmetic refinements; they are what let the authors detect an exponentially small slope under a much larger quadratic term. The proof is organized well: reductions to prior classifications, then the mixed case, bounded bubble-sheet scale, then unbounded. I read the differential neck theorem and the final rigidity arguments closely and did not find a circular step or a fitted constant. The differential neck constant a(M) is derived and shown nonzero by contradiction, not assumed. The paper also honestly flags its own dependency on previous work, including the slow-convergence case from [CDD+], which is only listed as to appear.\n\nSoft spots, in proportion. The biggest is external dependency: the slow-convergence case relies on [CDD+], which is not yet publicly available in the same verified form. The authors state Theorem 1.6 as established; the referee will have to trust that chain or chase the preprint. Second, the applicability of Hamilton's Harnack inequality B_t H >= 0 in this noncompact ancient setting is cited to [Ham95] plus [DS23, BLL23]. That is standard in the program, and the way it is used in Sections 10–11—to pin H(P_h)=1 and invoke rigidity—is consistent with the stated hypotheses. Still, it is the load-bearing input most worth checking. Third, the paper is dense; independent verification of the bootstrap constants and maximum-principle estimates will take an expert weeks. That is a cost, not a defect. I did not see invented entities or free parameters. The citation pattern is heavy on the same group's earlier papers, but this is a sequential program and the dependencies are explicit rather than hidden.\n\nWho is this for? Geometric analysts working on MCF singularities, and anyone building surgery or canonical neighborhood theorems in R4. A serious referee should be assigned. The main theorem deserves a strong journal slot. If I were refereeing, I would ask for the [CDD+] dependency to be stated clearly in the introduction—it is, but only as 'to appear'—and for a short note on the Harnack hypotheses. No structural revision needed. I would bring it to reading group and cite it. Verdict: engage.","headline":"Completes the R4 noncollapsed singularity classification with a serious new analytic toolbox; technically heavy but no apparent load-bearing flaw.","tokens_in":51249,"tokens_out":2202,"would_cite":true,"duration_ms":22116,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C44","53E10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Any ancient noncollapsed mean curvature flow in $\\mathbb{R}^4$ belongs to one explicit list of standard shrinkers, bowls, ovals, and two one-parameter families.","keywords":["mean curvature flow","ancient noncollapsed solutions","singularity classification","R^4","bubble-sheet","differential neck theorem","selfsimilar translation","exotic ovals"],"falsifier":"Construct an ancient noncollapsed solution in $\\mathbb{R}^4$ whose tangent flow at $-\\infty$ is the round bubble-sheet and whose bubble-sheet function satisfies the mixed-convergence expansion (5), but whose level-set eccentricity grows faster than $\\log|t|$ or whose cap distance grows faster than $C|t|$; such a solution would violate the cap-distance estimate and the no-exotic-oval conclusion. A more direct check is numerical: evolve a strictly convex hypersurface in $\\mathbb{R}^4$ that is very long in the $x_1$ direction under mean curvature flow and look for a forward-time decrease of the mean-curvature speed at the caps, since any decrease of $H$ with time contradicts the Harnack input used in Theorem 11.12.","tokens_in":50411,"feed_emoji":"🫧","tokens_out":10191,"duration_ms":89464,"temperature":0.7,"pith_summary":"This paper completes the classification of noncollapsed singularities of mean curvature flow in four dimensions. It proves that every ancient noncollapsed solution, the possible blowup limits obtained by zooming into a singularity, is one of a short explicit list: standard shrinking spheres, cylinders and planes, bowls, ovals, or one of the known one-parameter families of translators and ovals. The hard case is mixed convergence, where the flow approaches a round bubble-sheet quickly in one direction and quadratically in another; the paper introduces a differential neck theorem that detects the tiny slope responsible for translation. If correct, this gives a complete catalog of singularities for mean-convex flows in $\\mathbb{R}^4$ and a canonical neighborhood theorem around high-curvature points.","feed_headline":"Every noncollapsed singularity in R4 is now on one list","feed_subtitle":"The complete catalog of ancient solutions settles the classification program for mean-convex flows in four dimensions.","key_machinery":"The load-bearing object is the differential neck theorem: for every space-time point $X$, the $x_1$-derivative $u_1$ of the bubble-sheet profile, localized near the graphical radius $\\rho(\\tau)$, equals $a e^{\\tau/2}$ plus a controlled error, with a nonzero constant $a = a(M)$ attached to the whole flow. The proof builds this theorem from a Merle-Zaag switch dynamics for the slope's spectral coefficients, elongated anisotropic barriers obtained by rotating 2d shrinkers around ellipses, and propagation-of-smallness estimates that push decay of $|\\nu_1|$ from a central strip to the whole level set; Hamilton's Harnack inequality then turns the slope control into selfsimilarity.","core_discovery":"Any ancient noncollapsed mean curvature flow in $\\mathbb{R}^4$ is, up to scaling and rigid motion, one of the following: the standard shrinkers $S^3$, $\\mathbb{R}\\times S^2$, $\\mathbb{R}^2\\times S^1$, or $\\mathbb{R}^3$; the 3d-bowl or $\\mathbb{R}\\times$ 2d-bowl; the $\\mathbb{Z}_2\\times O_3$-symmetric 3d-oval, the $O_2\\times O_2$-symmetric 3d-oval, or $\\mathbb{R}\\times$ 2d-oval; or one of the one-parameter families of translators and of 3d-ovals. The proof splits the bubble-sheet analysis into fast, slow, and mixed convergence; the mixed case is handled by a differential neck theorem showing that the slope of the bubble-sheet profile is a nonzero exponential, which forces every noncompact strictly convex solution to be selfsimilarly translating and rules out exotic ovals.","pith_inferences":[],"forward_implications":["Every blowup limit of mean-convex mean curvature flow in $\\mathbb{R}^4$, and in any 4-manifold, is on the classified list, giving a canonical neighborhood theorem: above a curvature threshold every point sees one of the listed models at scale $H^{-1}$.","Exotic ovals do not exist: there is no compact ancient noncollapsed solution with mixed bubble-sheet convergence, closing the last potential compact model.","Every noncompact strictly convex ancient noncollapsed flow in $\\mathbb{R}^4$ is selfsimilarly translating, so non-selfsimilar ancient solutions such as slowly accelerating bowls are ruled out.","The classification is sharp: all listed examples are known to exist, so the set of noncollapsed singularity models in $\\mathbb{R}^4$ is now exactly this finite plus one-parameter list.","The $\\mathbb{R}^4$ canonical neighborhood theorem provides the natural input for mean-convex flow with surgery in $\\mathbb{R}^4$ and for the mean-convex neighborhood conjecture in four dimensions.","The nonzero constant $a(M)$ in the differential neck theorem gives a computable invariant of a singularity: its sign fixes the translation direction and its size the translation speed, which may be measurable from time-slice data.","The anisotropic propagation-of-smallness estimates look transferable to 4d Ricci flow $\\kappa$-solutions, where a similar mixed-mode problem appears in curvature profiles; an analogous differential neck theorem could organize the 4d Ricci singularity classification.","A concrete testable extension is to use the theorem's canonical neighborhood structure to construct surgeries in $\\mathbb{R}^4$ with uniform neck parameters, since the classified models specify exactly which necks and caps can appear."],"supporting_citations":[{"why":"Supplies the Harnack inequality controlling the mean-curvature speed at earlier times, which drives the final rigidity argument for selfsimilar translation.","marker":"[Ham95]"},{"why":"Supplies the normal form and Ornstein-Uhlenbeck evolution for the bubble-sheet function, the starting point for the mixed convergence analysis.","marker":"[DH24]"},{"why":"Provides the 2d oval asymptotics and the barrier construction used to build elongated cylindrical regions in the present proof.","marker":"[ADS19]"},{"why":"Classifies two-convex ancient solutions in $\\mathbb{R}^3$ and gives uniqueness of the 2d oval, used to identify rescaled limits such as $\\mathbb{R}\\times$ 2d-oval.","marker":"[ADS20]"},{"why":"Gives uniqueness of convex ancient solutions in $\\mathbb{R}^3$ and the neck-improvement idea underlying the bubble-sheet symmetry analysis.","marker":"[BC19]"},{"why":"Settles the fast-convergence case (no wings), leaving only the mixed-convergence case for the present paper.","marker":"[CHH24]"},{"why":"Settles the slow-convergence case of bubble-sheet ovals, so the classification reduces to the mixed case.","marker":"[CDD+]"},{"why":"Constructs the one-parameter family of translators that the classification must include and that the differential neck theorem controls.","marker":"[HIMW19]"},{"why":"Constructs the one-parameter family of 3d-ovals that the classification must include.","marker":"[DH]"},{"why":"Reduces membership in the translator family to proving that a solution is noncompact and selfsimilarly translating.","marker":"[CHH23]"}],"fun_headline_variants":["All ancient noncollapsed flows in R4 now classified","R4 mean curvature flow: full classification of ancient solutions","Ancient noncollapsed MCF singularities in R4 fully listed","Complete catalog of ancient noncollapsed flows in R4","MCF in R4: every ancient noncollapsed solution identified"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes Hamilton's Harnack monotonicity for the mean-curvature speed of ancient noncollapsed flows in $\\mathbb{R}^4$; if the speed could decrease going forward in time, the final argument that cap speeds are constant across levels and hence the flow translates self-similarly would break.","fun_headline_variants_meta":{"raw":{"variants":["All ancient noncollapsed flows in R4 now classified","R4 mean curvature flow: full classification of ancient solutions","Ancient noncollapsed MCF singularities in R4 fully listed","Complete catalog of ancient noncollapsed flows in R4","MCF in R4: every ancient noncollapsed solution identified"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00046,"raw_usage":{"total_tokens":2374,"prompt_tokens":1089,"completion_tokens":1285,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":1214}},"tokens_in":705,"tokens_out":1285,"duration_ms":9499,"temperature":1.0,"reasoning_tokens":1214,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T15:49:37.762155+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct an ancient noncollapsed solution in $\\mathbb{R}^4$ whose tangent flow at $-\\infty$ is the round bubble-sheet and whose bubble-sheet function satisfies the mixed-convergence expansion (5), but whose level-set eccentricity grows faster than $\\log|t|$ or whose cap distance grows faster than $C|t|$; such a solution would violate the cap-distance estimate and the no-exotic-oval conclusion. A more direct check is numerical: evolve a strictly convex hypersurface in $\\mathbb{R}^4$ that is very long in the $x_1$ direction under mean curvature flow and look for a forward-time decrease of the mean-curvature speed at the caps, since any decrease of $H$ with time contradicts the Harnack input used in Theorem 11.12.","supporting_citations":[],"review_version":1}