{"id":"c019421b-4b84-43c5-859e-aaae62404e82","arxiv_id":"2412.00781","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Energy-minimizing maps into trees have, near every frequency-3/2 point, free interfaces composed of three C^{1,α} surfaces meeting at 120 degrees along a C^{1,α} (d-2)-dimensional boundary.","lead":"This paper proves that near a triple junction singularity, the free interface of energy-minimizing harmonic maps into trees is made of three smooth surfaces meeting at 120 degrees. It is the first such structure theorem for these free boundaries in dimensions higher than two, and it also settles a spherical optimal partition conjecture.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 6.3 applies the epiperimetric inequality (Theorem 1.2) directly to non-homogeneous rescalings u_{x0,r}; the required homogeneous-extension step and verification of its hypotheses are missing.","rationale":"The reader's weakest_assumption identifies Theorem 1.2; I agree that the result collapses without it. However, after reviewing the contradiction proof of Theorem 1.2, I found no clear error: the equality ∑||w_i||²_H^1 = ||ξ||²_H^1 is valid since |∇|A||=|∇A| almost everywhere; Step 3's use of a closest-point projection is justifiable by rotation invariance; and Lemma 5.3's sign is correct. The more pressing concrete gap is the unqualified application of Theorem 1.2 to non-homogeneous rescalings in Proposition 6.3, whose proof is only a reference to [OV24, Prop. 8.1] and does not carry out the homogeneous-extension reduction. This is fixable, so the verdict should remain CONDITIONAL. Separately, the classification Proposition 3.5 has the disconnected-phase gap noted by the reader: the sentence 'By Lemma 3.4, there is a point x0 in (F(u)\\F1(u))∩∂B1' is a non-sequitur unless the disconnected-phase alternative is ruled out; this is also likely fixable by invoking [ST15, Lemma 4.2], but as written it is a genuine logical gap.","tokens_in":27317,"tokens_out":30574,"duration_ms":256139,"concrete_test":"Write out the omitted Weiss reduction: for each x0∈K and r2, and it introduces an epiperimetric-inequality method in this setting. The paper is carefully structured, with explicit statements of the main inequality, a detailed contradiction proof in the spirit of Weiss, a linearized problem, and a no-holes lemma. The claimed applications to optimal partitions and to the min-max spherical partition problem are natural and would be important corollaries. The result is plausibly true, and the proof strategy is a serious contribution; however, as detailed below, several load-bearing steps in the transition from the epiperimetric inequality to the regularity theorem are not sufficiently justified in the current manuscript.","major_comments":[{"comment":"Theorem 1.2 applies only to 3/2-homogeneous functions c, but Proposition 6.3 states that “we can apply the epiperimetric inequality to all the rescaling of the form u_{x0,r}(x)=r^{-3/2}u(x0+rx).” The functions u_{x0,r} are not homogeneous. The standard Weiss argument would require: (i) for each x0,r, form the 3/2-homogeneous extension c_{x0,r} of the trace of u_{x0,r} on ∂B1; (ii) verify uniformly, in x0 and r, the Hausdorff and H^1 closeness hypotheses of Theorem 1.2 for c_{x0,r}; (iii) apply the inequality to c_{x0,r}; and (iv) use minimality to compare W_{3/2}(u_{x0,r}) with W_{3/2}(c_{x0,r}). None of these steps appears in the manuscript. This is not a cosmetic omission: Proposition 6.3 is the bridge from the epiperimetric inequality to the uniqueness of the blow-up, the rate r^α, the oscillation estimate, the no-holes lemma, and the final C^{1,α} regularity. Without a complete proof of this step, the main theorem is not established.","section":"§6.2, Lemma 6.5"},{"comment":"The proof of the frequency gap from above applies Theorem 1.2 to a sequence u_n ∈ M_{γ_n}(R^d;N) of γ_n-homogeneous minimizers with γ_n → 3/2. Theorem 1.2 requires the input c to be 3/2-homogeneous; the functions u_n have degree γ_n > 3/2, so they do not satisfy the hypothesis. One cannot simply rescale them to be 3/2-homogeneous, and the proof gives no alternative construction of a 3/2-homogeneous function to which the inequality can be applied. Since Lemma 6.5 is used in the no-holes lemma and in the proof of Theorem 1.1, this gap also affects the main result. A careful justification is needed.","section":"§6.2, Lemma 6.5"},{"comment":"The proof begins with “By Lemma 3.4, there is a point x0 ∈ (F(u)\\F_1(u)) ∩ ∂B1,” but Lemma 3.4 is a conditional statement: it assumes that all free boundary points on ∂B1 are regular and that the sets Ω_i^u ∩ ∂B1 are connected, and then concludes that σ·u is harmonic and hence γ is an integer. The correct argument would be to apply Lemma 3.4 under the contrary assumption and derive a contradiction from γ=3/2, thereby obtaining a singular point. As written, the invocation of Lemma 3.4 is a non-sequitur. The intended reasoning is probably repairable, but it needs to be stated explicitly because Proposition 3.5 is the classification of all 3/2-homogeneous blow-ups.","section":"§3.2, Proposition 3.5"}],"minor_comments":[{"comment":"There are several typos and duplicated words, for example “is a a smooth” and “of of” in the introduction, and “Ord u = 1” should be consistently typeset as “Ord_u”. A careful proofreading pass is needed.","section":"§1"},{"comment":"The Weiss energy is defined as W_γ(u,x0,r) := H(u,x0,r)/r^{2γ}(N(u,x0,r)−γ), but the notation “N” in this formula overloads the number of components N. The authors use N for both the Almgren frequency and the number of phases; this should be clarified.","section":"§2.2"},{"comment":"The notation W_γ(d_{Σ_N}(u,v)) is not defined: W_γ was introduced for vector-valued maps in H^1(B1;Σ_N), while d_{Σ_N}(u,v) is a scalar function. The proof appears to interpret W_γ of the distance as a sum over the components of v, but this abuse of notation should be explained or replaced by a properly defined functional.","section":"§5, Lemma 5.3"},{"comment":"In the construction of the competitor, the interpolation function t=2(r−1/2) is used, and the notation Ω_i is redefined as Ω_i∩B1 without explicitly noting the change; this is a minor clarity issue but should be stated to avoid confusion.","section":"§5, Step 1 of Theorem 1.2"},{"comment":"The paper relies on the authors’ own preprint [OV24] for epsilon-regularity and for the standard rate-of-convergence argument; this is acceptable, but since [OV24] is unpublished, the dependence should be highlighted in the introduction.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses an important problem and introduces a promising new technique, and the gaps identified above appear to be repairable rather than fatal. In particular, Proposition 6.3 needs a full treatment of the homogeneous-extension step, and Lemma 6.5 needs a correct application of the epiperimetric inequality. The classification argument in Proposition 3.5 should also be rewritten. I would encourage the editors to seek a revised version with these points addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Three things you should know. First, this is a genuine step forward: a new epiperimetric inequality for the 3/2-Weiss energy, and a structure theorem for triple junctions in harmonic maps into trees and optimal partitions in any dimension d≥2. Second, it catches a real error in Caffarelli–Lin's uniqueness proof (missing term in [CL10, Lemma 3.2]) and supplies the corrected identity (Lemma 5.3). Third, the paper as written has two gaps that a referee should send back, but both look repairable.\n\nThe epiperimetric inequality (Theorem 1.2) is proved by contradiction, with the linearization, orthogonality conditions, and competitor construction well laid out. The classification of 3/2-homogeneous blow-ups (Proposition 3.5) is a corollary of [ST15] plus a topological lemma. The final theorem is the first of its kind in d>2, so the significance is high.\n\nNow the gaps. Proposition 3.5 says 'By Lemma 3.4, there is a point x0 in (F(u)\\F1(u))∩∂B1.' Lemma 3.4 requires the free boundary on the sphere to be all regular and each Ω_i∩∂B1 connected. Those hypotheses are not verified; the disconnected-phase case is ignored. It can probably be fixed with a short argument (a disconnected cone phase would force a singular point on the sphere), but as written it is a non-sequitur.\n\nBigger issue: Proposition 6.3 claims the epiperimetric inequality applies directly to non-homogeneous rescalings u_{x0,r}. Theorem 1.2 only applies to 3/2-homogeneous functions. The standard Weiss argument requires passing to the homogeneous extension of the trace on ∂B1, then comparing Weiss energies via minimality. None of that appears here; the proof just cites [OV24, Prop 8.1]. The stress-tester is right that this bridge is missing. The final regularity theorem collapses without it. Again, likely fixable by importing the missing argument, but it is not in the manuscript.\n\nAlso note the paper leans on [OV24] for epsilon-regularity and the graph argument, so it is not self-contained; that is acceptable if the preprint is solid, but the referee should check the dependency.\n\nBottom line: this is a serious paper that deserves a serious referee. It is not ready as is; the two gaps above need to be addressed. If the authors fill them, this becomes a strong publication in a top analysis journal.","headline":"Real advance with a repairable gap: the new epiperimetric inequality is solid, but two steps in the regularity proof are missing as written.","tokens_in":28119,"tokens_out":4520,"would_cite":true,"duration_ms":39351,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R35","49Q10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the free interface of an energy-minimizing harmonic map around any point of frequency $3/2$ is exactly three $C^{1,\\alpha}$-smooth surfaces, sharing a common $C^{1,\\alpha}$ boundary and meeting at $120^\\circ$ angles.","keywords":["triple junction","epiperimetric inequality","harmonic maps into singular spaces","free interface regularity","optimal partition problems","Almgren frequency","Y-configuration","Weiss energy"],"falsifier":"Set $u$ equal to a small perturbation of the model $Y$ on $\\partial B_1$, for example by adding a tiny fourth component supported away from the junction, and compute $W_{3/2}$ of the best competitor built by the paper's interpolation rule; if for some sequence of perturbations approaching $Y$ the infimum of $W_{3/2}(u)-W_{3/2}(c)$ is not bounded below by $-\\varepsilon W_{3/2}(c)$ with $\\varepsilon>0$ fixed, the epiperimetric inequality, and with it Theorem 1.1, would be false.","tokens_in":27126,"feed_emoji":"📐","tokens_out":12334,"duration_ms":106775,"temperature":0.7,"pith_summary":"This paper proves a structural theorem for the singular set of energy-minimizing harmonic maps into singular spaces, the model target being a space $\\Sigma_N$ made of $N$ half-lines glued at a common origin; the same result applies to spectral optimal partition problems. The main claim (Theorem 1.1) is that around any point of the lowest singular frequency, $\\gamma(u,x)=3/2$, the zero set of the map is exactly three $C^{1,\\alpha}$-smooth hypersurfaces that share a common $(d-2)$-dimensional $C^{1,\\alpha}$ boundary and meet along it at $120^\\circ$ angles. The proof is carried by a new epiperimetric inequality (Theorem 1.2): near the model triple junction $Y$, any $3/2$-homogeneous map that is close to $Y$ in Hausdorff and $H^1$ senses can be replaced, with unchanged boundary values, by a competitor of strictly lower $3/2$-Weiss energy, with a fixed multiplicative gain. From this inequality the paper derives uniqueness of the $Y$-shaped blow-up with a quantitative rate, a gap separating the frequency $3/2$ from all higher frequencies, and the full $C^{1,\\alpha}$ regularity of the interface. A corollary settles the $Y$-configuration as the unique minimizer of the three-partition min-max problem on the sphere for $p=+\\infty$ in every dimension $d\\ge 3$.","feed_headline":"In harmonic maps, every triple junction is three surfaces at 120°","feed_subtitle":"A new energy-drop inequality proves the singular set near 3/2-frequency points is a smooth edge, with consequences for optimal partitions.","key_machinery":"The engine of the proof is a new epiperimetric inequality for the $3/2$-Weiss energy $W_{3/2}(u)=\\sum_i\\int_{B_1}|\\nabla u_i|^2\\,dx -\\frac{3}{2}\\sum_i\\int_{\\partial B_1}u_i^2\\,dS$ near the model triple junction $Y$, the $3/2$-homogeneous configuration whose three non-zero components are supported on three sectors of angle $2\\pi/3$ and equal $r^{3/2}|\\cos(3\\theta/2)|$. The inequality asserts that if a $3/2$-homogeneous map $c$ is sufficiently close to $Y$, both in Hausdorff distance of the supports and in $H^1$ distance measured with the metric $d_{\\Sigma_N}$, then there is a competitor $u$ with the same boundary values and $W_{3/2}(u)\\le (1-\\varepsilon)W_{3/2}(c)$, with $\\varepsilon>0$ depending only on the dimension. The proof is by contradiction: it rescales by the $H^1$ distance, extracts a limit of the linearized maps, uses a classification of $3/2$-homogeneous solutions of the linearized problem via a reduction to a thin-obstacle-type equation, and exploits a Weiss-energy linearization identity containing a positive interaction term $\\beta$ that forces the limit to vanish. Once the inequality is available, it yields a quantitative uniqueness of the $Y$-blow-up at every $3/2$-frequency point, a frequency gap from above, and a topological no-holes argument that forces the three interfaces to persist up to the singular boundary; the $C^{1,\\alpha}$ regularity then follows by iteration.","core_discovery":"The central discovery, stated as Theorem 1.1, is that the singular set of an energy-minimizing map $u:B_1\\to\\Sigma_N$ splits as $\\mathrm{Sing}(u)=F_{3/2}(u)\\cup\\{x\\in F(u):\\gamma(u,x)\\ge 3/2+\\delta_d\\}$, where $\\delta_d>0$ depends only on the dimension. The stratum $F_{3/2}(u)$, consisting of points of frequency $3/2$, is open inside the singular set and is locally a $(d-2)$-dimensional $C^{1,\\alpha}$-smooth manifold. Around every point $x\\in F_{3/2}(u)$, the whole free interface $F(u)$ is composed of three $(d-1)$-dimensional surfaces $\\Gamma_{12},\\Gamma_{23},\\Gamma_{31}$, each $C^{1,\\alpha}$-regular up to the common boundary $F_{3/2}(u)$, and the three surfaces meet at $120^\\circ$ angles. In the case of maps into a locally finite metric tree, the same conclusion holds for the fiber over any vertex, and in the optimal partition problem it holds for the free interface of the vector of first eigenfunctions.","pith_inferences":["Beyond the paper: the same epiperimetric strategy could be directed at the next frequency stratum, since Theorem 1.1 does not classify homogeneous blow-ups with frequencies between $3/2$ and $3/2+\\delta_d$; a natural test is whether iterating the linearization argument forces all such frequencies to be integer or half-integer.","Beyond the paper: for optimal partitions, the theorem suggests that numerical methods can represent the singular set near a triple junction by tracking a single smooth $(d-2)$-dimensional edge with three attached surfaces, rather than treating the interface as an arbitrary rectifiable set.","Beyond the paper: the uniqueness of the $Y$-configuration for the $p=+\\infty$ problem is consistent with, but does not prove, the full three-partition min-max conjecture for finite $p$; computing the second variation of the sum functional near $Y$ for $p<\\infty$ would be a concrete step toward that conjecture."],"forward_implications":["Around every point of frequency $3/2$, the free interface is locally a $Y$-shaped union of three $C^{1,\\alpha}$ surfaces with common $C^{1,\\alpha}$ boundary, meeting at $120^\\circ$; there are no other frequencies between $3/2$ and $3/2+\\delta_d$.","For energy-minimizing maps into any locally finite metric tree, the fiber over a vertex has the same structure, so the triple-junction description is not tied to the model target $\\Sigma_N$.","For minimizers of the spectral optimal partition problem, the singular set of the free interface is locally a smooth $(d-2)$-dimensional manifold, and the regular parts are three smooth $(d-1)$-dimensional surfaces meeting at $120^\\circ$.","The uniqueness of the blow-up at frequency $3/2$ is quantitative: rescalings converge to the $Y$-configuration with a power rate $r^\\alpha$."],"supporting_citations":[{"why":"Establishes the decomposition of the nodal set into regular and singular strata via the frequency function, the baseline that Theorem 1.1 refines.","marker":"[GS92]"},{"why":"Gives the rectifiability and dimension bounds for the singular set of harmonic maps to trees that the present theorem upgrades to full structural regularity.","marker":"[Sun03]"},{"why":"Provides the complete two-dimensional description of the nodal set around singular points, which the paper extends to all dimensions.","marker":"[CTV03]"},{"why":"Supplies the minimality of the frequency $3/2$ and the classification ingredients for $3/2$-homogeneous solutions used in Proposition 3.5.","marker":"[ST15]"},{"why":"Contributes the contradiction framework for proving epiperimetric inequalities from the obstacle problem that the paper adapts to the $3/2$-Weiss energy.","marker":"[Wei99]"},{"why":"Gives the classical triple-junction theorem for area-minimizing surfaces that the paper transfers to harmonic-map free interfaces.","marker":"[Tay76]"},{"why":"Provides the topological persistence argument (no-holes) that forces the three interfaces to survive up to the singular boundary.","marker":"[Sim93]"},{"why":"Supplies the epsilon-regularity lemma for frequency-1 free boundary points and the rate-of-convergence template used in the final regularity argument.","marker":"[OV24]"},{"why":"Provides the classification of $3/2$-homogeneous solutions of the thin-obstacle linearized problem used in Step 3 of the epiperimetric proof.","marker":"[FS16]"},{"why":"Gives the epiperimetric-inequality approach for the thin obstacle problem whose linearized structure the paper imports via Proposition 4.1.","marker":"[GPS16]"}],"fun_headline_variants":["Harmonic map triple junctions: three smooth surfaces at 120°","Every triple junction in harmonic maps is three 120° surfaces","New proof: singular set near frequency 3/2 is a smooth 120° Y","Triple points in harmonic maps are smooth, always 120° angles","Spectral partitions get smooth 120° interfaces from harmonic maps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the new epiperimetric inequality of Theorem 1.2: any $3/2$-homogeneous map sufficiently close to the model triple junction $Y$ in Hausdorff distance and in $H^1$ norm can be strictly improved by a fixed percentage in the $3/2$-Weiss energy among maps with the same boundary values, and every later conclusion in Theorem 1.1 would collapse without it.","fun_headline_variants_meta":{"raw":{"variants":["Harmonic map triple junctions: three smooth surfaces at 120°","Every triple junction in harmonic maps is three 120° surfaces","New proof: singular set near frequency 3/2 is a smooth 120° Y","Triple points in harmonic maps are smooth, always 120° angles","Spectral partitions get smooth 120° interfaces from harmonic maps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000191,"raw_usage":{"total_tokens":1332,"prompt_tokens":923,"completion_tokens":409,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":314}},"tokens_in":539,"tokens_out":409,"duration_ms":4434,"temperature":1.0,"reasoning_tokens":314,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:01:12.299148+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Set $u$ equal to a small perturbation of the model $Y$ on $\\partial B_1$, for example by adding a tiny fourth component supported away from the junction, and compute $W_{3/2}$ of the best competitor built by the paper's interpolation rule; if for some sequence of perturbations approaching $Y$ the infimum of $W_{3/2}(u)-W_{3/2}(c)$ is not bounded below by $-\\varepsilon W_{3/2}(c)$ with $\\varepsilon>0$ fixed, the epiperimetric inequality, and with it Theorem 1.1, would be false.","supporting_citations":[],"review_version":1}