{"id":"e34b0014-2b2c-4e46-b002-0ac787c27630","arxiv_id":"2504.14880","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For suitable harmonic map flows, every time slice of the singular set is (n-2)-rectifiable with uniform Minkowski content estimates, and excluding harmonic and quasi-harmonic 2-spheres yields a sharp L^{3,∞} bound on the gradient.","lead":"Suitable solutions of the harmonic map heat flow are shown to have singular sets whose every time slice is rectifiable, with a uniform bound on how much space the slice can fill. A second result gives a sharp spatial bound on the gradient when the target manifold has no special harmonic spheres.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Quantitative stratification has an off-by-one error: Definition 3.2 uses (k,ε)-symmetry where Lemma 3.16 and identity (3.2) require (k+1,ε)-symmetry; under the literal definition the central identity is false.","rationale":"The reader identified the imported localized monotonicity formula, Proposition 2.4, as the weakest assumption. I agree that this lemma is imported and that the paper's estimates depend heavily on it. However, I found a more elementary internal inconsistency that directly threatens the link between the analytic strata and the quantitative strata used for all later estimates. Even if Proposition 2.4 is granted, the proof of identity (3.2) fails under the literal wording of Definition 3.2. The text of Lemma 3.16 and Lemma 5.7 makes the intended definition clear, so this is very likely a typo rather than a mathematical falsehood. Still, a conditional acceptance should require the correction of Definition 3.2 and a verification that (3.2) is proved for the corrected definition. This does not change the overall verdict: the paper remains a serious, technically substantial contribution whose central argument is plausible but not fully checked as written. I also note other textual issues, such as the reference to a nonexistent Lemma 3.20 in the proof of Theorem 1.14 and the occasional inconsistent notation, but the off-by-one in the quantitative stratification is more load-bearing because it affects the identification of the very sets whose rectifiability is claimed.","tokens_in":43707,"tokens_out":30094,"duration_ms":268289,"concrete_test":"Check the asserted identity (3.2) against the literal Definition 3.2 for a model suitable solution that is a static harmonic map independent of one spatial coordinate, e.g. u(x_1,...,x_n,t)=V(x_1,...,x_{n-1}) with V a nontrivial stationary harmonic map. At a point of sing(V)×R, the tangent measure is spatially (n−2)-symmetric but not (n−1)-symmetric, so the point belongs to the analytic stratum Σ^{n−2}(u,t). Under the literal definition, at sufficiently small scales u is approximately (n−2,ε)-symmetric, hence the point is not in Σ^{n−2}_{ε;0,R}(u,t) for any ε; this contradicts (3.2) and shows the proof of Lemma 3.16/Theorem 1.14 uses the wrong symmetry order unless Definition 3.2 is changed to 'not spatially (k+1,ε)-symmetric'.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definition 3.2 defines Σ^k_{ε;r,R}(u,t) as the set of points for which u is not spatially (k,ε)-symmetric at any scale s∈[r,R]. But the analytic stratum Σ^k(u,t) in Definition 1.9 is defined by the absence of spatial (k+1)-symmetric tangent measures, so the quantitative set must approximate the stronger condition 'no (k+1,ε)-symmetry'. The surrounding text confirms this: Lemma 3.16 infers from x∉Σ^{n-2}_{ε;0,ε} the existence of a scale with (n−1,ε)-symmetry, and Lemma 5.7 infers from x∈Σ^k_{ε;Rr} the failure of (k+1,ε)-symmetry. Both inferences are valid only if the definiens is 'not spatially (k+1,ε)-symmetric'. Under the literal Definition 3.2, the second half of the proof of (3.2) in Remark 3.4 produces only a k-symmetric tangent measure (because the approximating measures μ_i are k-symmetric), which is insufficient to conclude x∉Σ^k(u,t). A concrete failure mode is a solution whose only tangent symmetry is exactly k-dimensional: such a point lies in Σ^k(u,t), yet it is approximately k-symmetric at small scales and therefore lies outside every Σ^k_{ε;0,R}(u,t) as the definition is written. Thus (3.2) is false under the literal text, and the reduction of Theorem 1.14 to the quantitative estimates in Theorem 3.6 does not go through. The intended definition is visible from context, but the manuscript must be corrected; otherwise the central rectifiability claim is unproved.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a quantitative stratification theory for suitable solutions of the harmonic map flow, based on spatial symmetry of tangent measures rather than tangent flows. The main result, Theorem 1.14, asserts that for every time slice the singular set coincides with the top spatial stratum, that each stratum Σ^k(u,t) is k-rectifiable for 1≤k≤n−2, and that the (n−2)-dimensional Minkowski content of sing(u)∩{t} is uniformly bounded. Theorem 1.20 adds a sharp L^{3,∞} gradient bound under the assumption that the target admits neither harmonic nor quasi-harmonic 2-spheres. The proofs adapt the Naber–Valtorta quantitative stratification and Reifenberg-rectifiability machinery to the parabolic heat-flow setting, using localized backward-heat-kernel densities, scale-restricted quantitative strata, L^2-best estimates, and a tree/covering argument. The central new ingredient is the parabolic L^2-best estimate of Proposition 5.1, which links density differences to k-dimensional displacements.","tokens_in":44053,"tokens_out":8676,"duration_ms":72502,"significance":"If the stated results hold, this is a substantial advance: it upgrades earlier almost-everywhere-time-slice rectifiability statements (Lin–Wang, Chen–Struwe) to every time slice, with optimal rectifiability of each stratum and uniform Minkowski content estimates, and it yields a sharp L^{3,∞} regularity bound under target assumptions. The adaptation to the non-nested parabolic domains via localized monotone densities and the L^2-best estimate is a genuine technical innovation. The paper carefully identifies its dependence on external results (Naber–Valtorta, Azzam–Tolsa, Lin–Wang compactness) and states falsifiable quantitative claims. However, the manuscript currently contains an index error in the definition of quantitative stratifications and a missing referenced lemma; these are load-bearing and must be corrected before the claims can be certified.","major_comments":[{"comment":"The quantitative stratum Σ^k_{ε;r,R}(u,t) is defined by failure of spatial (k,ε)-symmetry, but the surrounding text requires failure of spatial (k+1,ε)-symmetry. Under the literal definition, the parenthetical 'in other words' after (3.1), the proof of identity (3.2), Lemma 3.16, and Lemma 5.7 are all false: the approximating measures μ_i in Remark 3.4 are only k-symmetric, so the limiting tangent measure is at best k-symmetric and gives no information about Σ^k(u,t), whose definition uses (k+1)-symmetry. A concrete failure mode is a point whose only tangent-measure symmetry is exactly k-dimensional: such a point belongs to Σ^k(u,t) but is approximately k-symmetric at all small scales and therefore lies outside every Σ^k_{ε;0,R}(u,t) as written. This off-by-one error is load-bearing because it connects the analytic strata of Definition 1.9 to the quantitative estimates of Theorem 3.6 and hence to Theorem 1.14. The intended definition is visible from the text, so the fix is local: amend Definition 3.2 to require failure of spatial (k+1,ε)-symmetry, and propagate the index shift through Remark 3.4, Lemma 3.16, Lemma 5.7, and all proofs that invoke them.","section":"Definition 3.2 and Remark 3.4"},{"comment":"The Minkowski content estimate (1.11) is asserted to be 'a direct consequence of Theorem 3.6 and Lemma 3.20', but no Lemma 3.20 exists anywhere in the manuscript: Section 3 ends at Lemma 3.18 and no later section introduces Lemma 3.20. Since (1.11) is one of the two central conclusions of Theorem 1.14, the proof is incomplete as written. The authors should supply the missing lemma with a full proof, or replace the reference by an existing statement and verify that the constant C(Λ,n,N) is independent of t.","section":"Section 7.2, proof of Theorem 1.14"},{"comment":"The inclusion {x∈B_1 : r_u((x,t))<εr} ⊂ Σ^{n−3}_{ε;0}(u,t) is attributed to Lemma 3.16, but Lemma 3.16 only yields the weaker inclusion into Σ^{n−2}_{ε;0,ε} (or its corrected analogue), because it assumes spatial (n−1,ε)-symmetry. The stronger inclusion into Σ^{n−3}_{ε;0} is exactly what Lemma 3.18 provides under the assumption that N admits no harmonic or quasi-harmonic 2-spheres. As it stands, the proof of Theorem 1.20 cites the wrong lemma for the decisive step that produces the r^3 volume bound; replace the reference and verify that the quantitative stratum index matches the (n−2,ε)-symmetry hypothesis of Lemma 3.18.","section":"Section 7.3, proof of Theorem 1.20"}],"minor_comments":[{"comment":"In the display defining the lower Minkowski content, 'Mink r(E)' appears; this should be 'Min^α_r(E)' for consistency with the preceding line.","section":"Definition 1.10"},{"comment":"The notation H^k_s is used for the Hausdorff content at scale s, but it is not defined; please define the s-scale Hausdorff content before using it in the proof of (3.7).","section":"Section 7.1"},{"comment":"The statement writes 'γ=γ∈(0,1)'; this should be 'γ=γ(ε,Λ,n,N)∈(0,1)' to make the dependence explicit.","section":"Lemma 3.12"},{"comment":"The sentence 'Furthermore, [40] and implies that' contains a dangling 'and implies' with no reference; either complete the sentence or insert the intended citation.","section":"Introduction, after (1.8)"}],"recommendation":"major_revision","confidential_remarks":"The paper is ambitious and likely correct in intention, but the quantitative-stratification index shift and the missing Lemma 3.20 are currently load-bearing gaps. I recommend asking the authors to fix the definition of Σ^k_{ε;r,R}, re-check every statement that uses 'spatially (k,ε)-symmetric' versus 'spatially (k+1,ε)-symmetric', and supply the missing lemma or replace the reference. After those corrections, the paper deserves a careful verification round."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read on the Fu-Wang-Wu-Zhang paper. It does deliver something new: for suitable harmonic map flows, every time slice of the singular set is rectifiable, not just almost every slice, and you get Minkowski content estimates on the top stratum. The adaptation of Naber-Valtorta to the parabolic setting is real work; the parabolic L2-best approximation (Prop 5.1) is the genuinely new ingredient, and the scale-restricted quantitative stratification is a sensible tool. I think the main architecture is sound.\n\nThat said, there is a definition-level problem that has to be fixed. Definition 3.2 defines the quantitative strata using 'not spatially (k,eps)-symmetric', but every application in the paper, and the 'in other words' line right there, needs 'not spatially (k+1,eps)-symmetric'. Under the literal definition, the compactness argument in Remark 3.4 only produces k-symmetric tangents, so identity (3.2) is false; the reduction of the analytic theorem to the quantitative estimates fails as written. The intended definition is unambiguous from context, so it's a typo, but it's a central typo and the manuscript must be corrected.\n\nOther issues: 'Lemma 3.20' is cited but doesn't exist; should be Lemma 3.16, which is the one giving the singular set as the top quantitative stratum. The n=2 case is not handled: the rectifiability of the 0-strata is outside Proposition 5.1's k>=1 range, so the claim for sing(u) cap {t} in 2D needs separate treatment. And several load-bearing quantitative lemmas are pinned to unpublished preprints ([59, Lemma 6.3], [28]); a referee will want those checked.\n\nNone of these are fatal in the sense of the main idea being wrong. The off-by-one is fixable, and the rest is verification and edge-case work. But the paper as it stands is not ready for acceptance without revision.\n\nWho should read it: specialists in geometric measure theory and parabolic PDE. It is a proper contribution to the singular-set literature.\n\nRecommendation: send it to referees, but flag the off-by-one and the n=2 gap. It deserves careful refereeing, not a desk reject.","headline":"Genuinely new per-time-slice rectifiability for harmonic map flows, but the manuscript has an off-by-one error in the key definition that must be fixed before the main result is valid as stated.","tokens_in":44610,"tokens_out":5342,"would_cite":true,"duration_ms":44375,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K55","58E20","35B65"],"pacs":[],"model":"deepseek-v4-flash","headline":"Suitable solutions of the harmonic map flow have time-sliced singular sets that are countably rectifiable, with optimal dimension bounds.","keywords":["harmonic map flow","suitable solutions","singular set","rectifiability","tangent measures","quantitative stratification","Minkowski content","Lorentz regularity"],"falsifier":"Compute, for an explicit backwardly self-similar quasi-harmonic sphere solution such as $u(x,t)=\\psi(x/\\sqrt{-t})$ with $\\psi$ a quasi-harmonic sphere, the localized difference $W(u,X,R,r)+C_1(R-r)$. If for some $R>r$ the expression is negative, Proposition 2.4 is false and the whole argument collapses. Alternatively, construct or numerically approximate a suitable solution whose singular set at one fixed time has box-counting dimension greater than $n-2$; that would directly refute the Minkowski-dimension conclusion of Theorem 1.14.","tokens_in":43489,"feed_emoji":"🌀","tokens_out":8161,"duration_ms":71112,"temperature":0.7,"pith_summary":"The paper proves a sharp geometric description of where singularities can form in suitable solutions of the harmonic map flow. It shows that at every fixed time, the singular set is countably $(n-2)$-rectifiable and has Minkowski dimension at most $n-2$, with a uniform constant depending only on the energy bound, the dimension, and the target manifold. This upgrades earlier results that controlled the singular set only for almost every time slice. The proof works through the spatial symmetry of tangent measures rather than through uniqueness of tangent flows, and a second theorem gives a uniform $L^{3,\\infty}$ bound on $\\nabla u$ when the target admits neither harmonic nor quasi-harmonic $2$-spheres.","feed_headline":"Every time slice of the singular set is rectifiable","feed_subtitle":"For suitable solutions, every frozen time slice of the singular set is rectifiable, with sharp Minkowski-content bounds.","key_machinery":"The machine is the pair consisting of a localized density and a quantitative displacement. The localized density is $\\Phi(u,X_0,\\rho)=\\frac12\\int_{T_\\rho(X_0)}\\varphi_{x_0}^2|\\nabla u|^2 G_{X_0}\\,dxdt$, built from a cutoff $\\varphi$ and the backward heat kernel, whose monotonicity formula supplies the scale-comparison estimates. The new $L^2$-best estimate, Proposition 5.1, bounds the $k$-dimensional displacement $D^k_\\mu(x_0,r)$ of any finite measure $\\mu$ by the integral of $W(u,(y,t_0),2r,r/2)+r$ against $\\mu$, whenever the flow is spatially $(0,\\delta)$-symmetric at scale $2r$ but not spatially $(k+1,\\varepsilon)$-symmetric. This converts density drops across scales into geometric closeness to $k$-planes and feeds the Reifenberg-type covering arguments that yield $k$-rectifiability and Minkowski-content bounds.","core_discovery":"The central discovery is that the stratified singular set of a suitable solution is not only small but geometrically regular at each instant. Defining $\\Sigma_k(u,t)$ as the set of spatial points where no tangent measure is backwardly self-similar and invariant along a $(k+1)$-dimensional subspace, the paper proves $\\Sigma_{n-2}(u,t)\\times\\{t\\}=\\mathrm{sing}(u)\\cap\\{t\\}$, that each $\\Sigma_k(u,t)$ is $k$-rectifiable for $1\\le k\\le n-2$, and that $\\mathrm{Min}_r^{n-2}(\\mathrm{sing}(u)\\cap\\{t\\})$ is uniformly bounded. The key novelty is a parabolic $L^2$-best estimate linking differences of localized backward-heat-kernel energy densities to the $k$-dimensional displacement of a measure, which makes the Reifenberg machinery available without nesting properties of integral domains or unique continuation. Under the additional hypothesis that the target manifold has no harmonic or quasi-harmonic $2$-spheres, the same machinery yields a uniform $L^{3,\\infty}$ bound on $\\nabla u$ in $B_1$ over all times.","pith_inferences":["The paper does not claim it, but the same monotonicity-and-displacement mechanism should give every-time-slice rectifiability of the concentration set for other parabolic systems with localized backward-heat-kernel monotonicity, such as supercritical semilinear heat equations and mean-curvature flow, even without uniqueness of tangent flows.","One could test whether the $L^{3,\\infty}$ exponent in Theorem 1.20 is optimal: on targets containing quasi-harmonic $2$-spheres, explicit shrinking-soliton solutions should saturate the bound, while on sphere-free targets the argument suggests stronger pointwise decay near singularities.","A direct numerical or analytic check of Proposition 2.4 on one explicit shrinking bubble solution would validate the engine of the proof; the paper itself delegates that monotonicity formula to an existing lemma.","Because the stratification uses tangent measures instead of tangent flows, the results should also hold for weak solutions not known to be unique, provided the suitable-solution conditions are satisfied."],"forward_implications":["At any fixed time $t$, the singular set $\\mathrm{sing}(u)\\cap\\{t\\}$ is countably $(n-2)$-rectifiable, not merely for almost every $t$.","For each $k\\in\\{1,\\dots,n-2\\}$, the $k$-th spatial stratum $\\Sigma_k(u,t)$ is $k$-rectifiable, matching the known Hausdorff-dimension upper bound and making the stratification optimal.","The Minkowski dimension of every singular time slice is at most $n-2$, with the uniform $r$-content bound $\\mathrm{Min}_r^{n-2}(\\mathrm{sing}(u)\\cap\\{t\\})\\le C$ depending only on $\\Lambda$, $n$, and $N$.","When $N$ has no harmonic or quasi-harmonic $2$-spheres, $\\sup_t\\|\\nabla u(\\cdot,t)\\|_{L^{3,\\infty}(B_1)}\\le C$, and hence $\\nabla u(\\cdot,t)\\in L^p(B_1)$ uniformly for $2<p<3$.","The tangent-flow strata $S_k(u,t)$ are also $k$-rectifiable, and the same arguments extend to solutions obtained as limits of Ginzburg-Landau approximations."],"supporting_citations":[{"why":"Supplies the quantitative-stratification and Reifenberg-rectifiable framework that the paper adapts to the flow setting.","marker":"[44]"},{"why":"Contains the localized monotonicity formula used as Proposition 2.4, on which all scale-comparison and W-difference estimates rest.","marker":"[56]"},{"why":"Establishes the tangent-measure stratification for heat-flow and Ginzburg-Landau limits and the a.e.-time rectifiability that this paper upgrades to every time.","marker":"[40]"},{"why":"Provides the compactness-based approach and the L2-best-estimate template that let the authors avoid nesting and unique continuation.","marker":"[21]"},{"why":"Introduces quantitative stratification, the scale-uniform notion the paper refines with upper-scale restrictions.","marker":"[8]"},{"why":"Gives the partial-regularity criterion used to identify the regular set and to bound regularity scales.","marker":"[42]"},{"why":"Constructs suitable weak solutions via Ginzburg-Landau approximation and proves the concentration-set partial regularity underlying the heat-flow setting.","marker":"[12]"}],"fun_headline_variants":["Theorem: singular set slices are rectifiable for suitable solutions","Tangent measures yield rectifiable singular strata in harmonic flow","Stratification via tangent measures: each time slice is rectifiable","Harmonic map flows: singular set rectifiable on each time slice","Sharp regularity for suitable harmonic map flows via tangent measures"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the localized monotonicity formula for the cut-off backward-heat-kernel densities, Proposition 2.4, including its explicit $C_1(R-r)$ error term; if that inequality fails or the error term cannot be controlled for all suitable solutions, no scale comparison, $L^2$-best estimate, covering argument, or rectifiability conclusion survives.","fun_headline_variants_meta":{"raw":{"variants":["Theorem: singular set slices are rectifiable for suitable solutions","Tangent measures yield rectifiable singular strata in harmonic flow","Stratification via tangent measures: each time slice is rectifiable","Harmonic map flows: singular set rectifiable on each time slice","Sharp regularity for suitable harmonic map flows via tangent measures"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000565,"raw_usage":{"total_tokens":2648,"prompt_tokens":885,"completion_tokens":1763,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":1680}},"tokens_in":501,"tokens_out":1763,"duration_ms":10517,"temperature":1.0,"reasoning_tokens":1680,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:38:51.511648+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for an explicit backwardly self-similar quasi-harmonic sphere solution such as $u(x,t)=\\psi(x/\\sqrt{-t})$ with $\\psi$ a quasi-harmonic sphere, the localized difference $W(u,X,R,r)+C_1(R-r)$. If for some $R>r$ the expression is negative, Proposition 2.4 is false and the whole argument collapses. Alternatively, construct or numerically approximate a suitable solution whose singular set at one fixed time has box-counting dimension greater than $n-2$; that would directly refute the Minkowski-dimension conclusion of Theorem 1.14.","supporting_citations":[{"cited_title":"Naber and D","cited_arxiv_id":null,"evidence_quote":"Supplies the quantitative-stratification and Reifenberg-rectifiable framework that the paper adapts to the flow setting."},{"cited_title":"Wang, Heat flow of harmonic maps whose gradients belong toL n x L∞ t ,Archive for rational mechanics and analysis,188(2008), 351-369","cited_arxiv_id":null,"evidence_quote":"Contains the localized monotonicity formula used as Proposition 2.4, on which all scale-comparison and W-difference estimates rest."},{"cited_title":"Lin and C","cited_arxiv_id":null,"evidence_quote":"Establishes the tangent-measure stratification for heat-flow and Ginzburg-Landau limits and the a.e.-time rectifiability that this paper upgrades to every time."},{"cited_title":"Edelen and M","cited_arxiv_id":null,"evidence_quote":"Provides the compactness-based approach and the L2-best-estimate template that let the authors avoid nesting and unique continuation."},{"cited_title":"Cheeger and A","cited_arxiv_id":null,"evidence_quote":"Introduces quantitative stratification, the scale-uniform notion the paper refines with upper-scale restrictions."},{"cited_title":"Liu, Partial regularity for weak heat flows into a general compact Riemannian manifold,Archive for Rational Mechanics and Analysis,168(2003), 131-163","cited_arxiv_id":null,"evidence_quote":"Gives the partial-regularity criterion used to identify the regular set and to bound regularity scales."},{"cited_title":"Chen and M","cited_arxiv_id":null,"evidence_quote":"Constructs suitable weak solutions via Ginzburg-Landau approximation and proves the concentration-set partial regularity underlying the heat-flow setting."}],"review_version":1}