{"id":"6cb1a533-2694-434c-9d4d-d9d278efbd23","arxiv_id":"2411.14186","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Singular S1-valued harmonic maps with prescribed codimension-2 singular set exist on closed manifolds, and three variational relaxations share the same renormalised interaction energy.","lead":"This paper proves which sets can be singularities of maps from a manifold to a circle that are harmonic everywhere else, and shows that three ways of relaxing the energy give the same renormalized interaction energy between parts of the singular set. The result extends a classical theory of point vortices to higher-dimensional singular sets.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The fractional lower bound (4.31) relies on the unproved claim that the Γ-average of a degree-one map still has degree +1; averaging over Γ can destroy the winding, so the H^s expansion (4.22) is not fully established as written.","rationale":"I read Theorem B as the central claim: all three relaxations of the Dirichlet energy produce the same renormalised energy W_M(Γ) after removing the divergent leading term. The δ-tube relaxation is treated in detail in §3.2 and Theorem 2, and the p-energy section reduces to comparison with u_δ and the fibre degree estimate. The fractional argument in §4.5 is the least secure. The step passing from degree-one u_s on B_δ × Γ to a degree-one averaged map \\bar u_s is not merely compressed; arithmetic averaging of S^1-valued maps over Γ can annihilate the winding. The explicit map e^{i(arg z + θ)} has degree +1 on every Γ-fibre while its Γ-average is identically zero, so the sentence 'By the degree condition on u_s, it follows that \\bar u_s has degree +1' is false as a general implication. Since (4.31) is the lower bound that produces the 1/(2-2s) singularity in the H^s expansion, the written proof of (4.22) has a genuine gap. I nevertheless do not think this warrants rejection: the gap appears repairable by integrating the fibrewise [CFP24] lower bound over Γ instead of first averaging, and no circularity or fitted-parameter issue is present. This is exactly the reader's identified weakest assumption, so the appropriate verdict remains CONDITIONAL; my read does not change the reader's verdict.","tokens_in":31423,"tokens_out":13160,"duration_ms":137152,"concrete_test":"Take M = B_1 × S^1, Γ = {0} × S^1, and the H^s-admissible competitor u(z,θ) = e^{i(arg z + θ)}. Compute its Γ-average: \\bar u(z) = 0 for every z, so the asserted implication 'degree +1 on Γ fibers implies degree +1 of the averaged map' fails for this explicit map. Then check the repair route: using the spectral identity (4.33), prove that [u_s]^2_{H^s(B_δ×Γ)} ≥ ∫_Γ [u_s(·,y)]^2_{H^s(B_δ)} dy and combine this with the fiberwise [CFP24] lower bound for each degree-one fiber. If this yields exactly (4.31), the theorem survives with a corrected proof; if not, the H^s expansion as stated lacks a valid lower bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §4.5, after deriving the inequality [u_s]^2_{H^s(B_δ×Γ)} ≥ H^{n-2}(Γ) [\\bar u_s]^2_{H^s(B_δ)}, the proof states: 'By the degree condition on u_s, it follows that \\bar u_s has degree +1' and then applies [CFP24, Lemma 3.6 and Proposition 4.1]. This implication is not a consequence of the degree condition. The averaged map \\bar u_s(z) = \\fint_Γ u_s(z,·) takes values in the closed unit disk, not necessarily in S^1, and its degree is not even defined unless it avoids 0. Averaging S^1-valued degree-one maps over the parameter space can cancel winding: on M = B_1 × S^1 with Γ = {0} × S^1, the admissible map u(z,θ) = e^{i(arg z + θ)} has degree +1 on every fiber {θ}, but its Γ-average vanishes identically. Hence the lower bound (4.31) does not follow from the displayed argument. This is load-bearing because (4.31) is the only step producing the leading constant 2π H^{n-2}(Γ)/(2-2s) in the fractional energy expansion; the upper-bound half of the proof does not by itself force that constant. The gap is likely repairable, for instance by applying the fiberwise [CFP24] bound to u_s(·,y) for each y ∈ Γ and then integrating over Γ using the spectral identity (4.33), but that replacement is not present in the manuscript. The related compressed step 'one sees that [ω_2] = [d^*ψ]' in Proposition 4.1 is also not fully written out, but it is less central than the fractional lower bound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies S1-valued harmonic maps with prescribed codimension-two singular set on a closed oriented Riemannian manifold. Theorem A/Theorem 1 classifies all such distributionally harmonic maps for any admissible integral current Γ=∂Σ, parametrising them by H1(M,2πZ) via the Hodge decomposition of the pulled-back form ju. For embedded C1,1 Γ, the paper defines a renormalised energy WM(Γ) from the regular part of the Green's potential of the Hodge Laplacian and proves in Proposition 3.1 the Dirichlet-energy expansion on Mδ. Theorem B/Theorem 3 claims matching asymptotic expansions for the tube-removed, p-energy (p↑2), and fractional Hs (s↑1) relaxations, with common leading term 2πH^{n-2}(Γ) times the relevant divergence and common renormalised energy WM(Γ)+e(uα). The proofs compare the relaxed minimisers to the fixed harmonic map uα; the tube-removed case uses a De Giorgi iteration, and the p/fractional cases compare with uδ.","tokens_in":31800,"tokens_out":10291,"duration_ms":98085,"significance":"If the central claims are made rigorous, this is an important contribution: it gives a Hodge-theoretic classification of singular harmonic maps, and it shows that three different relaxations share the same renormalised interaction energy, providing a higher-dimensional analogue of the Bethuel–Brezis–Hélein theory. The explicit expression in R3 (Eq. (1.2)) answers a question of Rivière and connects the renormalised energy to magnetic inductance; the fractional case also connects to a recently introduced nonlocal area in codimension two. The paper's structural clarity and its intrinsic Green's-function formulation are genuine strengths. However, the proof of the fractional expansion currently contains a load-bearing gap, so the significance is conditional on the repair described below.","major_comments":[{"comment":"The lower bound (4.31) is not established by the displayed argument. After deriving [us]^2_{Hs(Bδ×Γ)} ≥ H^{n-2}(Γ)[ū_s]^2_{Hs(Bδ)}, the proof states that the degree condition on us implies that the Γ-average ū_s has degree +1, and then applies [CFP24, Lemma 3.6 and Proposition 4.1]. This implication is false as stated: ū_s(z)=average over Γ of us(z,·) is merely a unit-disk-valued map, its degree is not even defined unless it avoids 0, and averaging S1-valued degree-one maps can destroy the winding. For instance, u(z,θ)=e^{i(arg z+θ)} on B1×S1 has degree +1 on every fiber but its Γ-average is identically zero. Since (4.31) is the only step in the proof of Theorem 3 that produces the leading constant 2πH^{n-2}(Γ)/(2-2s), expansion (4.22) is not proven as written. The gap is likely repairable, for example by applying the fiberwise [CFP24] lower bound to us(·,y) and integrating over Γ using the spectral identity (4.33), but that argument must be supplied.","section":"§4.5, Eq. (4.31)"},{"comment":"The identification [ω2]=[d*ψ] is asserted with the phrase 'one sees that' after invoking the degree condition and properties of the distributional degree. This step is load-bearing because Proposition 4.1 justifies the reduction of the tube-removed minimisation (4.5) to the sectors W^{1,2}_α, and the same sector decomposition is reused in §4.2–4.3 for the p- and fractional relaxations. Please provide the full cohomological argument, including the relation between deg(u,Γ), the class of ω2 in H1(M\\Γ), and the class of d*ψ.","section":"§4.1, Proposition 4.1"}],"minor_comments":[{"comment":"In the double minimisation display, 'min_{u∈Hs_α} Ep(u)' should read 'min_{u∈Hs_α} Es(u)', since the surrounding discussion concerns the fractional energy Es.","section":"§4.3"},{"comment":"Just before Eq. (4.31), the sentence 'By the degree condition on us, if follows' contains a typo; it should be 'it follows'.","section":"§4.5"},{"comment":"The existence of the model map u* with u*|Tδ = z/|z| is imported from [BD24, Appendix A] without stating the precise hypotheses; adding a short lemma stating the model-map result would improve self-containedness.","section":"Lemma 3.3(4)"},{"comment":"The cut-off ξ and the radius δ0 in the definition of S should be tied explicitly to the Fermi-coordinate tube; as written, the choice of cut-off is implicit and the invariance of WM(Γ) under this choice deserves at least a sentence.","section":"Lemma 3.4"}],"recommendation":"major_revision","confidential_remarks":"This is a strong manuscript and the main technical concern is local to §4.5, but it is genuine: the fractional lower bound (4.31) is not proven by the averaging argument displayed in the text. I do not see grounds for rejection, since the defect appears repairable and the surrounding framework is coherent. I would ask the author to supply the missing degree-preservation or fiberwise argument for the fractional case, and to expand the compressed cohomology step in Proposition 4.1."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. Theorem A is clean and genuinely new: it classifies S1-valued harmonic maps with prescribed admissible singular set by H^1(M,2πZ), and the diagram chase linking integrality of the singular current to integrality of d*ψ is elegant. Theorem B unifies three relaxations—tube-removed Dirichlet energy, p-energy, and fractional H^s—around the same renormalised energy W_M(Γ), with an extra sector term e(u_α). That unification is the real contribution, and the filament formula (1.2) matches the known vortex-filament interaction and the electromagnetic inductance picture. The definition of W_M(Γ) via the Green's function of the Hodge Laplacian, with the log kernel subtracted in Fermi coordinates, is concrete and checkable.\n\nThe classification and the tube expansion are mostly solid. The p-harmonic expansion is also reasonable. But the fractional expansion has a load-bearing gap. In §4.5 the proof says 'by the degree condition on u_s, it follows that \\bar u_s has degree +1' and applies [CFP24]. That implication is false in general. The Γ-average of an S1-valued map takes values in the unit disk, and averaging over the parameter space can kill winding. On B_1×S^1, u(z,θ)=e^{i(arg z+θ)} has degree +1 on every fiber but its Γ-average is identically zero. So the lower bound (4.31) is not established as written. This matters because (4.31) is the only step forcing the leading constant 2πH^{n-2}(Γ)/(2-2s); the upper bound alone doesn't pin it down. The gap looks repairable—apply the fiberwise [CFP24] bound to u_s(·,y) for a.e. y and integrate over Γ using the spectral identity—but that argument is not in the manuscript.\n\nA smaller compressed step is the identification [ω_2]=[d*ψ] in Proposition 4.1, introduced with 'one sees that'. Plausible, but it deserves a few lines. Not a major issue.\n\nOverall the program is coherent, the setup is honest, and there is no circular reasoning. But the fractional theorem is not fully proved as written. This deserves a serious referee who asks for that lower bound to be fixed. If the repair lands, it is a substantial contribution.\n\nFor whom: geometric analysts working on singular harmonic maps, Ginzburg–Landau and vortex filaments, p-harmonic and fractional relaxations, and min-max for codimension-two minimal submanifolds. I'd bring it to reading group, and I'd cite the classification. Recommendation: send to peer review.","headline":"A genuinely new classification and a mostly convincing renormalised-energy unification, but the fractional expansion in Theorem B has a repairable gap in the lower bound.","tokens_in":32352,"tokens_out":4199,"would_cite":true,"duration_ms":35135,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["58E20","49Q20","49J45"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves a complete classification of circle-valued harmonic maps with prescribed codimension-two singular set and shows that three variational relaxations converge to a common renormalised interaction energy after subtracting a…","keywords":["singular harmonic maps","renormalised energy","distributional Jacobian","fractional Sobolev maps","p-harmonic maps","Hodge Laplacian","codimension-two singularities","vortex filaments"],"falsifier":"Take $M=S^1\\times B^2$ with $\\Gamma=S^1\\times\\{0\\}$ and consider the admissible map $u_s(y,z)=e^{iy}z/|z|$, which has degree $+1$ around $\\Gamma$; its average over $\\Gamma$ is $(1/2\\pi)\\int_0^{2\\pi}e^{iy}dy\\, z/|z|=0$, so the averaged map has degree $0$. If such a map can be realised as, or perturbed into, a competitor for the fractional problem with the same energy asymptotics, the asserted step behind the lower bound (4.31) fails; if not, the missing proof must use a structural property that rules out this averaging cancellation.","tokens_in":31145,"feed_emoji":"🌀","tokens_out":11646,"duration_ms":105075,"temperature":0.7,"pith_summary":"This paper proves that a codimension-two singular set for maps from a closed Riemannian manifold to the circle carries a finite, metric-dependent interaction energy that appears in all three natural ways of making the singular problem variational. The first theorem classifies the maps that are harmonic away from the prescribed singular set and distributionally harmonic across it: they form the set $\\mathcal{H}(M,\\llbracket\\Gamma\\rrbracket)$, parametrised by the integral cohomology $H^1(M,2\\pi\\mathbb{Z})$, so the topology of the ambient manifold controls the multiplicity. The second theorem gives asymptotic expansions for three minimisation problems --- removing a $\\delta$-tube around $\\Gamma$ and prescribing the degree, minimising the $p$-energy as $p\\to 2$, and minimising the $H^s$-energy as $s\\to 1$ --- whose common finite part is the same renormalised energy $W_M(\\Gamma)$, plus a sector-dependent desingularised energy. This matters because it converts a topologically forced singularity into a computable cost: a divergent area term of order $2\\pi\\mathcal{H}^{n-2}(\\Gamma)/(2-p)$ and then a finite interaction between different parts of the singular set, with an explicit integral formula in Euclidean space.","feed_headline":"Three energy relaxations share one renormalised energy","feed_subtitle":"Tube-removed, p-energy, and fractional minimisers all expose the same finite interaction after the area blow-up.","key_machinery":"The load-bearing identity is the Hodge decomposition of the real one-form $ju=u^*(d\\theta)=d\\phi+d^*\\psi+\\omega$: the $d^*\\psi$ part encodes the singular set through $dd^*\\psi=2\\pi\\star\\llbracket\\Gamma\\rrbracket$, the $d\\phi$ part records harmonicity (it must vanish modulo constants for globally distributionally harmonic maps), and the harmonic part $\\omega$ carries the topological sector. The renormalised energy is built from the regular part of the Green's potential: with $A=2\\pi\\int_\\Gamma G_M^{n-2}(\\cdot,y)\\vec\\Gamma(y)$ and $S=-\\xi(|z|)\\log|z|\\vec\\Gamma$ the subtracted logarithmic singularity in Fermi coordinates, one defines $W_M(\\Gamma)=-2\\pi\\int_\\Gamma(A+S)$. The proof machinery also includes a De Giorgi-type regularity estimate for the boundary-value problem solved by the phase in the tube-removed relaxation, which controls the difference between the true minimiser and the model harmonic map.","core_discovery":"On its own terms, the paper establishes two structural facts. First, whenever the singular set $\\Gamma$ is an admissible integral current that bounds an oriented hypersurface, the singular harmonic maps are exactly the maps whose angle form satisfies $ju_\\alpha=d^*\\psi+\\omega_\\alpha$, where $\\psi$ is the unique two-form with $dd^*\\psi=2\\pi\\star\\llbracket\\Gamma\\rrbracket$ and $\\omega_\\alpha$ ranges over the affine lattice $[d^*\\psi]+H^1(M,2\\pi\\mathbb{Z})$; this proves Theorem A and its integral-current version. Second, for embedded $C^{1,1}$ singular sets of codimension two, the three relaxed energies have the common expansion with divergent leading term $2\\pi\\mathcal{H}^{n-2}(\\Gamma)\\log(1/\\delta)$, $2\\pi\\mathcal{H}^{n-2}(\\Gamma)/(2-p)$, or $2\\pi\\mathcal{H}^{n-2}(\\Gamma)/(2-2s)$, followed by the same renormalised energy $W_M(\\Gamma)$ and the sector energy $e(u_\\alpha)$. The renormalised energy is defined intrinsically from the Green's function of the Hodge Laplacian, and in $\\mathbb{R}^3$ the interaction between two curves $\\gamma_1,\\gamma_2$ reduces to the double integral of $\\dot\\gamma_1(x)\\cdot\\dot\\gamma_2(y)|x-y|^{-1}$.","pith_inferences":["A direct extrapolation is that $W_M(\\Gamma)$ defines an intrinsic self-interaction for arbitrary codimension-two cycles; in flat space it should equal a desingularised Coulomb-type integral, and computing it numerically for knotted curves would test the diagonal subtraction procedure explicitly.","If the commonality of $W_M$ survives the conjectured mixed $W^{s,p}$ relaxations, then multiplying the fractional expansion by $(1-s)$ would write nonlocal codimension-two area as local area plus $W_M$, suggesting that the nonlocal-to-local limit inherits a universal finite interaction term.","A topological subtlety left implicit is that averaging the fractional minimiser over $\\Gamma$ could in principle reduce its degree; checking maps of the form $e^{i\\varphi(y)}z/|z|$ on a tube over a circle would clarify whether the lower bound needs an additional hypothesis or a stronger structural property of minimisers."],"forward_implications":["For two disjoint embedded $C^{1,1}$ curves $\\gamma_1,\\gamma_2\\subset\\mathbb{R}^3$, the interaction is $\\int_{\\gamma_1}\\int_{\\gamma_2}\\dot\\gamma_1(x)\\cdot\\dot\\gamma_2(y)\\,|x-y|^{-1}\\,d\\mathcal{H}^1(x)d\\mathcal{H}^1(y)$, so perpendicular filaments do not interact and parallel filaments interact with orientation-dependent sign.","The same finite $W_M(\\Gamma)$ controls the tube-removed, $p$-harmonic, and fractional minimisers, so the choice of relaxation changes only the divergent prefactor and the sector term $e(u_\\alpha)$, not the geometric interaction.","If $\\mathcal{H}(M,\\llbracket\\Gamma\\rrbracket)$ contains a strict minimiser of the desingularised energy $e$, then for small $\\delta$ the tube-removed minimiser lies in that sector, so the asymptotic expansion selects the least-energetic harmonic representative.","The second and third expansions give a large-cost formula for prescribing a distributional Jacobian: the least $p$-energy and $H^s$-energy needed to force $Ju=Jv$ diverge like $2\\pi\\mathcal{H}^{n-2}(\\Gamma)/(2-p)$ and $2\\pi\\mathcal{H}^{n-2}(\\Gamma)/(2-2s)$.","The paper conjectures that mixed $W^{s,p}$ relaxations with $sp\\to 2$ obey the analogous expansion with $2-sp$ in the denominator, which would place the fractional and $p$-harmonic results in one family."],"supporting_citations":[{"why":"introduces canonical harmonic maps to the circle and the desingularised-energy limit $E_\\epsilon-2\\pi|\\deg|\\log\\epsilon|\\to W$ that the tube-removed relaxation extends.","marker":"[BBH94]"},{"why":"shows the same renormalised energy appears for $p$-harmonic maps with $p<2$ in planar domains, the model for the $p$-relaxation.","marker":"[HL95]"},{"why":"provides the renormalised energy for vortices on Riemannian surfaces and the flat-torus example showing the affine lattice $[d^*\\psi]+H^1(M,2\\pi\\mathbb{Z})$ can be non-integral.","marker":"[IJ21]"},{"why":"supplies the theory of the distributional Jacobian, the Factorisation theorem used to define singular sets for $H^s$ maps, and the least-energy-to-prescribe-Jacobian formulation.","marker":"[BM21]"},{"why":"gives the prescribed-singularities framework and the necessity of the boundary condition $\\Gamma=\\partial\\Sigma$, used for admissibility and the interpretation of $Ju$.","marker":"[ABO03]"},{"why":"proves $\\Gamma$-convergence of a fractional area in codimension two and supplies the $H^s$ estimates for degree-one maps on disks used in the fractional lower and upper bounds.","marker":"[CFP24]"},{"why":"provides Fermi-coordinate parametrisation of tubes and the expansion of the volume element used throughout the asymptotics.","marker":"[Gra04]"},{"why":"gives the fundamental solution and Green's operator for the Hodge Laplacian on forms, from which $A$ and $W_M(\\Gamma)$ are built.","marker":"[War83, Sco95]"},{"why":"raises the question on line-vortex interaction that formula (1.2) answers, situating the renormalised energy in the $\\mathrm{U}(1)$-Higgs context.","marker":"[Riv96]"}],"fun_headline_variants":["Three relaxations, one renormalised energy for singular harmonic maps","One renormalised energy governs three harmonic-map limits","Any oriented boundary is a harmonic map singular set","Tube, p-energy, fractional: all share the same renormalised energy","Renormalised energy emerges from three distinct cut-off schemes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The fractional lower bound assumes that averaging the minimiser over the singular set $\\Gamma$ preserves its degree-one winding, and this is asserted rather than proved.","fun_headline_variants_meta":{"raw":{"variants":["Three relaxations, one renormalised energy for singular harmonic maps","One renormalised energy governs three harmonic-map limits","Any oriented boundary is a harmonic map singular set","Tube, p-energy, fractional: all share the same renormalised energy","Renormalised energy emerges from three distinct cut-off schemes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000904,"raw_usage":{"total_tokens":3907,"prompt_tokens":982,"completion_tokens":2925,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":598,"completion_tokens_details":{"reasoning_tokens":2841}},"tokens_in":598,"tokens_out":2925,"duration_ms":21904,"temperature":1.0,"reasoning_tokens":2841,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:26:30.404095+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $M=S^1\\times B^2$ with $\\Gamma=S^1\\times\\{0\\}$ and consider the admissible map $u_s(y,z)=e^{iy}z/|z|$, which has degree $+1$ around $\\Gamma$; its average over $\\Gamma$ is $(1/2\\pi)\\int_0^{2\\pi}e^{iy}dy\\, z/|z|=0$, so the averaged map has degree $0$. If such a map can be realised as, or perturbed into, a competitor for the fractional problem with the same energy asymptotics, the asserted step behind the lower bound (4.31) fails; if not, the missing proof must use a structural property that rules out this averaging cancellation.","supporting_citations":[],"review_version":1}