{"id":"82fb7805-fab1-425c-9cd8-c24ff08cddbf","arxiv_id":"2506.22273","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A phase field model with a geodesic distance penalty between curves Gamma-converges to Plateau's problem for a single curve on a cylinder, and gives numerical approximations in wider cases.","lead":"The paper introduces a phase field energy that adds a curve-to-curve geodesic distance penalty to the usual diffuse interface energy, and proves a Gamma-convergence result toward Plateau's minimal surface problem in a cylindrical test case. The motivation is a numerically tractable way to compute soap-film-like minimal surfaces, including non-orientable ones.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The p<∞ liminf proof chooses αε=√cε, so the Chebyshev bound gives H²(Kε)≤C cε^{1−p/2}, which does not tend to zero when p≥2; since p=2 is exactly the case used in the numerics, the p<∞ branch of Theorem 1.1 is unproved as written.","rationale":"The reader's verdict is already CONDITIONAL, and my concern supports that verdict without moving it to a stronger or weaker position. I agree with the reader's identification of Proposition 2.1 as an external regularity input, but I regard the Chebyshev exponent error in §2.5.3 as the more directly load-bearing issue for the stated central claim. The theorem explicitly covers p ∈ [1,∞], and the p<∞ proof as written does not establish the vanishing of H²(Kε) for p ≥ 2; the numerical method uses p = 2, so this is not an irrelevant endpoint. The proposed replacement αε = cε^{1/(2p)} appears to repair the estimate, so the concern is a proof gap rather than a counterexample. The p = ∞ argument is substantially self-contained, and the limsup construction is standard modulo Proposition 2.1, so credit is due for the main structure. The numerical section's reliance on non-optimal geodesics is separately noted but does not affect the theoretical theorem; it only weakens the paper's numerical claims. My recommendation remains CONDITIONAL: the paper should be accepted only after the p<∞ Chebyshev step is corrected or explicitly re-derived with a suitable αε.","tokens_in":28570,"tokens_out":15279,"duration_ms":176763,"concrete_test":"Re-derive §2.5.3 Step 2 with αε := cε^{1/(2p)} instead of √cε. Verify that (a) H²(S_ℓε ∩ {uε ≥ αε}) ≤ C√cε → 0, (b) sε = αε − αε²/2 → 0, and (c) the perimeter comparison P(Aε ∪ Uε) ≥ P(Aε) − P(Uε) holds for BV sets, e.g. via P(Aε) ≤ P(Aε ∪ Uε) + P(Uε). If all three hold, the p<∞ liminf proof is repairable and the conditional verdict stands; if not, the central claim fails for p ≥ 2.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In §2.5.3, Step 2, the energy bound only gives ∫_{S_ℓε} |uε|^p dH² ≤ C cε. Chebyshev's inequality then yields H²(S_ℓε ∩ {uε ≥ α}) ≤ C cε / α^p. The text chooses αε := √cε, which gives H²(Kε) ≤ C cε^{1−p/2}. This tends to 0 only when p < 2; for p = 2 the bound is merely O(1) and for p > 2 it diverges. Consequently, the assertion in (28) that H²(Kε) → 0 is not justified for the parameter range p ∈ [2,∞), and the subsequent use of Lemma 2.2 to obtain P(Uε) → 0 also fails. Step 3 then relies on (31), the equality of the liminf of P({gε ≤ tε}) and P({gε ≤ tε} ∪ Uε); without P(Uε) → 0 this equality is unsupported. Since Section 3 explicitly minimizes the p = 2 functional, this is not a peripheral endpoint of the theorem. The p = ∞ branch is largely self-contained up to Proposition 2.1, and the limsup construction appears sound. The gap is likely repairable by taking αε = cε^{1/(2p)}, which yields C√cε → 0 and still gives sε → 0, but as written the p<∞ half of Theorem 2.2 and hence Theorem 1.1 is incomplete.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a phase-field approximation of a Reifenberg-type Plateau problem in R^3. The functional combines an Ambrosio-Tortorelli term with a p-geodesic distance penalty between a prescribed boundary curve and a point on it, generalizing earlier Steiner-problem approximations. The main theoretical result, Theorem 1.1, states that in a cylinder, for p in [1,∞], any quasi-minimizing sequence of the p-energy produces level sets that converge in L1 to a solution of the Plateau problem. The proof is organized as a Γ-limsup/liminf pair (Theorems 2.1 and 2.2), with the liminf proof using a separation argument based on the Borsuk theorem and an averaging argument over level sets. The numerical section develops a gradient-flow scheme using a Willmore-Cahn-Hilliard variant and fast-marching approximations of the geodesic term, with experiments for one, two, three, and six boundary curves, including the cube example.","tokens_in":28957,"tokens_out":10631,"duration_ms":104488,"significance":"If the main theorem is correct, this is a valuable contribution: it provides a new phase-field model for Plateau's problem in the Reifenberg sense, with a topological penalty that is not reduced to a current boundary constraint, and it gives the first Γ-convergence analysis in a nontrivial geometric setting. The proof is genuinely original in its use of separation and level-set averaging, and no fitted parameters or circular arguments are involved. The numerical experiments are visually convincing and suggest that the model is practically useful. However, the manuscript currently contains a load-bearing gap in the p<∞ part of the liminf proof, and it depends at a central point on an unpublished regularity result. These issues need to be resolved before the claims can be accepted as stated.","major_comments":[{"comment":"The Chebyshev bound displayed in Step 2 is not a consequence of the energy estimate. From ∫_{S_ℓε} |uε|^p dH² ≤ C cε one obtains H²(S_ℓε ∩ {uε ≥ α}) ≤ C cε / α^p, not (C cε/α)^p as written. With the chosen αε = √cε, the correct bound gives H²(Kε) ≤ C cε^{1−p/2}, which fails to tend to zero for p ≥ 2 and diverges for p > 2. Consequently assertion (28), the conclusion P(Uε) → 0, and equality (31) in Step 3 are not established for p ∈ [2,∞), which includes the value p = 2 used throughout the numerical section. The gap appears repairable, for example by taking αε = cε^{1/(2p)}, which gives H²(Kε) = O(√cε) and still sε → 0; but as written the p<∞ half of Theorem 2.2, and hence Theorem 1.1, is incomplete.","section":"§2.5.3, Step 2 (Eq. (28))"},{"comment":"The limsup inequality relies on the Ahlfors regularity of the essential boundary through (12)–(13), while the liminf proof uses the graph structure and the equality ∂∗Ω = ∂Ω stated in Proposition 2.1. The proof of this proposition is not contained in the manuscript; it is attributed to the unpublished manuscript [Mac25] ('in preparation'). Since these properties are load-bearing for both inequalities, the paper is not self-contained at a central point. The authors should provide a proof of Proposition 2.1, or at least a detailed self-contained argument in an appendix, or replace the reference by a publicly available verifiable source.","section":"Proposition 2.1, used in §2.4 (Eq. (13)) and §2.5"}],"minor_comments":[{"comment":"In the p<∞ part of the limsup proof, the integral of |uε|^p over K equals H²(K)(kε^p + δε), not H²(K)(kε + δε) as displayed; the conclusion (15) is unchanged because kε = cε² makes the extra factor vanish, but the displayed equality is inaccurate.","section":"§2.4, Eq. (15)"},{"comment":"The displayed identity '∂∗Ω1ε ∩ ∂∗Ω1ε ∩ C0 = ∅' should almost certainly read '∂∗Ω1ε ∩ ∂∗Ω2ε ∩ C0 = ∅'.","section":"§2.5.2, Step 4, Eq. (22)"},{"comment":"The statement asserts sε = O(cε), but in the p<∞ proof sε = αε − αε²/2 with αε = √cε (and after the proposed fix, αε = cε^{1/(2p)}); in neither case is sε = O(cε) for p ≥ 2. The theorem should state sε → 0, or give the exact order.","section":"Theorem 1.1"},{"comment":"The numerical section explicitly acknowledges (Remark 3.3 and §3.2.2) that the computed geodesics are not optimal and reports no quantitative error metrics; the cylinder-instead-of-catenoid example is admitted. The numerical claims are therefore qualitative, and the paper should add convergence or error data, or state more cautiously that these are heuristic illustrations.","section":"§3.2, numerical experiments"},{"comment":"The statement that replacing ATε by the Willmore-Cahn-Hilliard energy Pε preserves the Γ-convergence result is made without proof or reference; since the numerical minimization uses Pε, a precise statement or reference would be helpful.","section":"§3.1.2"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a worthwhile problem and the p=∞ branch, modulo Proposition 2.1, is convincing. The p<∞ gap is significant but likely fixable with a different choice of αε, so I recommend major revision rather than rejection. The dependence on the unpublished manuscript [Mac25] should be resolved as well. The numerical experiments are attractive but would benefit from quantitative validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about arXiv:2506.22273. First, the core idea is genuinely useful: combining Ambrosio-Tortorelli with a geodesic-distance penalty between curves extends the Steiner pipeline to Plateau's problem in a natural way, and the paper works out a nontrivial Gamma-convergence result in the cylinder case. Second, there is a real gap in the p<∞ liminf proof, and it hits exactly the parameter used in the numerics.\n\nThe limsup construction is standard and clean, and the p=∞ liminf is a genuinely clever piece of work: the level-set selection, the separation argument via Borsuk's theorem, and the construction of two competitors from the sublevel sets are all new and convincing. Credit is due there.\n\nThe soft spot is §2.5.3. The text claims Tchebychev gives H²(S_ℓε ∩ {uε ≥ α}) ≤ (C cε/α)^p, which is not the correct form of the inequality. Using the correct bound, H²(...) ≤ C cε / α^p, the choice αε = √cε yields C cε^{1−p/2}, which does not go to zero for p ≥ 2. Since the numerical section minimizes the p=2 functional, this is not a peripheral endpoint. The step that needs H²(Kε)→0, the subsequent perimeter bound (29), and the equality (31) all collapse for p≥2. The authors can repair this by taking αε = cε^{1/(2p)}, which gives H²(Kε) ≤ C√cε → 0 while sε → 0. So the gap is real but not deep.\n\nTwo other concerns are worth noting. Proposition 2.1, the Ahlfors regularity and boundary-to-topological-boundary identification for the Plateau minimizer, is imported from the unpublished manuscript [Mac25]. That is a structural dependency, not a self-citation problem, but the authors should make its status clearer. The numerical section is honest about using non-optimal geodesics and provides no error metrics; that is acceptable for a proof of concept, but it means the numerics support the model's plausibility rather than the proof's specifics.\n\nWho gets value from this? Researchers in phase-field methods for geometric variational problems, and people interested in non-oriented minimal surfaces. It does not resolve a major open problem, but it provides a workable variational approximation and a credible partial justification. I would send it to a serious referee. With the Chebyshev fix and a little tightening around [Mac25], it should be publishable.","headline":"The model is a real step toward phase-field Plateau approximations, and the cylinder-case Gamma-convergence proof is mostly sound, but the p<∞ branch has a genuine Chebyshev gap that leaves Theorem 1.1 unproved for the numerically relevant p=2; the fix is easy, so it deserves peer review.","tokens_in":29470,"tokens_out":3895,"would_cite":true,"duration_ms":41543,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q05","49Q20","49J45"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a single Lipschitz curve on a cylinder's edge, quasi-minimizers of an Ambrosio-Tortorelli energy with a geodesic distance penalty converge to a Plateau minimal surface.","keywords":["phase field approximation","Plateau's problem","Gamma-convergence","geodesic distance penalty","minimal surfaces","Ambrosio-Tortorelli energy","soap films","topological constraint"],"falsifier":"Run the numerical scheme in the exact one-curve cylinder setting and measure the perimeter in $C_0$ of the $L^1$ limit of the selected level-set component; if for a quasi-minimizing sequence that perimeter is strictly larger than the perimeter of the graph minimizer, Theorem 1.1 fails. A more targeted check is whether the Minkowski content identity $\\lim_{r\\to0}\\mathcal{L}^3(K_r)/(2r)=H^2(K)$ holds for the essential boundary $K$ of the cylinder minimizer; a violation there breaks the limsup construction.","tokens_in":28395,"feed_emoji":"🫧","tokens_out":6778,"duration_ms":67947,"temperature":0.7,"pith_summary":"The paper introduces a phase field model for Plateau's problem, the search for a least-area surface spanning prescribed boundary curves. The energy combines the standard Ambrosio-Tortorelli approximation of area with a geodesic distance penalty that connects boundary curves through homotopies, so the topological spanning condition is enforced only in the limit. The main theorem treats a single Lipschitz graph curve on the edge of a cylinder: any quasi-minimizing sequence produces level sets whose selected connected component converges in $L^1$ to a solution of Plateau's problem. The numerical section shows the same scheme approximating catenoids, two disks joined by a tube, the singular cube film, and non-orientable surfaces.","feed_headline":"Geodesic penalty lets phase fields find minimal spanning surfaces","feed_subtitle":"Quasi-minimizers of the new energy converge in L1 to least-area surfaces that span the boundary curve.","key_machinery":"The central object is the functional $F^p_\\varepsilon(u)=\\varepsilon\\int_C|\\nabla u|^2\\,dx+\\frac{1}{4\\varepsilon}\\int_C(1-u)^2\\,dx+\\frac{1}{c_\\varepsilon}d^p_u(\\gamma,\\gamma_0)$, where $d^p_u$ is the $p$-geodesic distance between curves, defined as the infimum of $\\int_{S_\\ell}(|u|^p+\\delta_\\varepsilon)\\,dH^2$ (or of the sup norm when $p=\\infty$) over Lipschitz homotopies $\\ell$ joining $\\gamma$ to $\\gamma_0$, with $S_\\ell=\\ell([0,1]\\times S^1)$. The geodesic term is the carrier of the topological constraint: when it is small, the surface $S_\\ell$ lies in a region where $u$ is small, and the separation property of such surfaces forces the selected level set to separate the cylinder. The proof then uses the co-area formula on $g_\\varepsilon=u_\\varepsilon-u_\\varepsilon^2/2$, an averaging lemma to pick a good level $t_\\varepsilon$, and the two competitors extracted from the separated components to obtain the liminf bound.","core_discovery":"The central claim is Theorem 1.1: for $p\\in[1,\\infty]$, if the prescribed curve $\\gamma$ is the graph of a Lipschitz function on the lateral boundary of a cylinder and $\\gamma_0$ is a constant curve inside it, then any quasi-minimizing sequence $u_\\varepsilon$ of $F^p_\\varepsilon$ yields a level set $\\{u_\\varepsilon-u_\\varepsilon^2/2>t_\\varepsilon\\}$ whose component containing the upper part of the cylinder converges in $L^1$, up to a subsequence, to a solution of Plateau's problem. The proof is a $\\Gamma$-convergence style argument: the limsup inequality builds a recovery sequence from the optimal Modica-Mortola profile around the minimizer's boundary, while the liminf inequality selects a suitable level set, proves it separates the cylinder, and reads off two competitors whose perimeters force the limit to be minimal. The paper therefore claims that topology can be prescribed by a penalty term at the limit, removing the need for an explicit current or divergence constraint.","pith_inferences":["The cylinder and single-curve assumptions look technical rather than essential; a natural conjecture is that the same level-set argument works on any domain where admissible homotopies separate the boundary and the limit minimizer is Ahlfors regular up to the boundary.","The numerical construction uses non-optimal geodesics and still converges to good films, which suggests the geodesic penalty is doing less topological work than the analysis requires; testing the exact energy with optimal geodesics in the cylinder would isolate how much slack the numerical relaxation introduces.","The choice of connection graph $I_\\gamma$ (which curve is paired with which point or curve) appears to select which Plateau solution is reached, so variants that learn $I_\\gamma$ during the flow could target or avoid singular solutions such as the tube connecting two disks.","Proving a $\\Gamma$-limit for the Willmore-Cahn-Hilliard variant and for multiple curves would close the gap between the analysis and the numerics; numerical benchmarks with measured perimeters could serve as the first evidence."],"forward_implications":["Quasi-minimizers of $F^p_\\varepsilon$ give a computable route to least-area surfaces: minimizers of the phase field energy approximate Plateau solutions without imposing a divergence or current constraint on the field.","In the cylinder setting, any algorithm that decreases the energy sufficiently fast eventually produces the minimal surface, because the energy gap to the optimum controls the $L^1$ distance of the selected level-set component.","The same geodesic-penalty design extends formally to several boundary curves and to non-oriented films, as the numerical experiments with catenoids, tubes, and the cube suggest.","The method is a higher-dimensional analogue of the Steiner phase field approximation: shortest connection between points is replaced by least-area homotopy between curves."],"supporting_citations":[{"why":"Introduces the geodesic distance penalty idea for Steiner's problem that this paper generalizes to Plateau's problem.","marker":"[LS14]"},{"why":"Gives the Steiner approximation and the liminf argument whose one-dimensional projection method cannot be reused in higher dimensions.","marker":"[BLS15]"},{"why":"Supplies the regularity of Plateau minimizers used in Proposition 2.1: open representative, essential equals topological boundary almost everywhere, and Ahlfors regularity up to the boundary.","marker":"[Mac25]"},{"why":"Provides the sets-of-finite-perimeter framework, the direct method, perimeter lower semicontinuity, and the isoperimetric inequality used in the proofs.","marker":"[Mag12]"},{"why":"Is the Reifenberg homological spanning formulation that motivates the paper's topological constraint.","marker":"[Rei60]"},{"why":"Supplies the numerical relaxation and fast-marching schemes for Steiner's problem that are adapted here for Plateau's problem.","marker":"[BBL20]"},{"why":"Is the alternative phase field approximation via divergence constraints that this paper contrasts with the geodesic penalty approach.","marker":"[CFM19a]"},{"why":"Provides the Willmore-Cahn-Hilliard energy used in the numerical model to improve regularity and accelerate optimization.","marker":"[BCM24]"}],"fun_headline_variants":["Geodesic penalty drives phase fields to minimal spanning surfaces","Phase field meets Plateau: geodesic penalty yields minimal surfaces","Single penalty term enforces topology in phase field minimal surfaces","Achieving Plateau's problem via a geodesic phase field energy","New model unifies phase field and geodesic distance for minimal surfaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the regularity of the ideal minimal surface: its boundary is a graph, its essential boundary coincides with its topological boundary almost everywhere, and its area in small balls is bounded above and below by a constant times the radius squared; if that regularity fails, the recovery sequence used in the limsup inequality may not have the claimed energy.","fun_headline_variants_meta":{"raw":{"variants":["Geodesic penalty drives phase fields to minimal spanning surfaces","Phase field meets Plateau: geodesic penalty yields minimal surfaces","Single penalty term enforces topology in phase field minimal surfaces","Achieving Plateau's problem via a geodesic phase field energy","New model unifies phase field and geodesic distance for minimal surfaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000244,"raw_usage":{"total_tokens":1497,"prompt_tokens":876,"completion_tokens":621,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":537}},"tokens_in":492,"tokens_out":621,"duration_ms":7110,"temperature":1.0,"reasoning_tokens":537,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:08:32.313818+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the numerical scheme in the exact one-curve cylinder setting and measure the perimeter in $C_0$ of the $L^1$ limit of the selected level-set component; if for a quasi-minimizing sequence that perimeter is strictly larger than the perimeter of the graph minimizer, Theorem 1.1 fails. A more targeted check is whether the Minkowski content identity $\\lim_{r\\to0}\\mathcal{L}^3(K_r)/(2r)=H^2(K)$ holds for the essential boundary $K$ of the cylinder minimizer; a violation there breaks the limsup construction.","supporting_citations":[],"review_version":1}