{"id":"08ad25cc-205c-482e-bbde-8d37a5b09f6e","arxiv_id":"2506.20215","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Fractional partition perimeters Gamma-converge as s approaches 1/2 to the local partition perimeter with surface tensions replaced by their shortest-path relaxation.","lead":"This paper proves that certain nonlocal 'fractional' perimeter energies defined on partitions of space converge, in a variational sense, to a local interfacial energy. The limit uses relaxed surface tension coefficients, so when the original coefficients violate the triangle inequality the limiting energy can nucleate new phases.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central Gamma-convergence claim is well supported; the only load-bearing external input is the standard half-space minimality theorem.","rationale":"The reader's weakest assumption is precisely the reliance on the two-phase half-space minimality theorem from [3] and its extension to the multi-phase cell problem. I examined the logical chain that uses this assumption: the blow-up lower bound defines an abstract matrix Γ; Section 4 refines it to a cell formula Γ; Section 5 proves the cell formula is solved by the half-space via Lemma 5.3 (max-flow min-cut) and Theorem 5.2. The max-flow argument is valid: capacities are symmetric, the cut defines a two-phase competitor, and the triangle inequality for the relaxed matrix σbar is exactly what is needed to compare the cut value with the original energy. The path decomposition avoids opposite arcs, so the capacity bound is used correctly. The approximation lemmas in Section 4 are technical but internally consistent; terms involving (1-2s) times constants or L1-differences vanish in the limit, and the averaging arguments select good slices without hidden assumptions. The compactness and upper-bound arguments are standard and properly cited. Thus the central claim holds up under scrutiny, and I have no concrete concern that would change the ACCEPT verdict.","tokens_in":39215,"tokens_out":21877,"duration_ms":211105,"concrete_test":"Verify that [3, Proposition 17] (and its original source [11]) holds for the fractional 2s-perimeter with kernel |x-y|^{-(n+2s)} for every s in (0,1/2), for the cube Q, and for boundary datum H outside Q. If the theorem is stated only for a restricted range of s or for balls, the reduction in Lemma 5.3/Theorem 5.2 would require a supplementary proof; if it covers the stated setting, the lower bound is complete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. I reviewed the Gamma-liminf chain (Sections 3-5) and the max-flow/min-cut replacement lemma (Lemma 5.3). The proof that the multi-phase cell problem is solved by the half-space partition reduces correctly to the two-phase half-space minimality of [3, Prop. 17]; the path-decomposition step uses only the triangle inequality satisfied by the relaxed matrix σbar, and the capacity symmetry is consistent with the functional. The approximation lemmas 4.2-4.3 contain no apparent gap: all error terms either vanish after multiplication by (1-2s) or are controlled by the L1 convergence. The use of [3, Prop. 17] is the single most load-bearing external input, but it is a published theorem and the reduction to it is mathematically sound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that, as s→1/2^-, the family of fractional multi-phase perimeter energies (1-2s)P^σ_{2s}(·,Ω) Γ-converges, in the L^1(Ω)-sense for partitions, to ω_{n-1}P^{\\barσ}_1(·,Ω), where \\barσ is the shortest-path relaxation of the surface-tension matrix σ. This holds for an arbitrary symmetric matrix with positive off-diagonal coefficients, without any triangle inequality assumption. The authors also prove a compactness theorem and a convergence result for local minimizers, showing that limits of local minimizers are local minimizers of the relaxed local functional. The proof strategy combines a blow-up lower bound, a cell-formula reduction through gluing lemmas, a half-space minimality theorem obtained by a max-flow/min-cut replacement argument, and an upper bound via polyhedral approximation and relaxation. The main novelty is that the limit functional is the lower-semicontinuous envelope of the classical multi-phase perimeter, so coefficients violating the triangle inequality are relaxed in the limit, which the authors interpret as phase nucleation.","tokens_in":39318,"tokens_out":27364,"duration_ms":265978,"significance":"If correct, this result settles the sharp-interface asymptotics for multi-phase fractional perimeter functionals with arbitrary surface tensions, a question that is central to the variational analysis of nonlocal minimal clusters and to threshold-dynamics models for grain growth. The identification of the relaxed matrix \\barσ as the shortest-path metric closure is clean and makes the relaxation mechanism explicit. The proof is unusually detailed: the key estimates in Lemmas 3.5, 4.2, 4.3 and 5.3 are written out, the gluing construction in Section 6 is carried through carefully, and the reliance on external results is limited to standard facts such as the two-phase half-space minimality theorem of [3] and the polyhedral approximation lemma of [5]. The max-flow/min-cut reduction of the multi-phase half-space problem to the two-phase case is an elegant and potentially reusable idea. The paper contains no circular reasoning: the relaxed coefficients are characterized independently in Lemma 3.4, and the Γ-limit is then derived from that characterization together with external minimality results.","major_comments":[],"minor_comments":[{"comment":"In the displayed chain after the four-phase decomposition, the coefficient of P_{2s}(E_1∪E_4,Q) is written as (α*−α̃_6); it should be (α*−α̃_7). As printed, the sum of the coefficients does not equal α̃_1+α̃_2−α̃_6−α̃_7, and the displayed lower bound would not yield P^σ_{2s}(H_{12},Q).","section":"§5.2.2, four-phase case"},{"comment":"The uniqueness claim for H_{ij} does not follow immediately from uniqueness of H in the two-phase problem: Lemma 5.3 gives a competitor F with P^σ(F)≤P^σ(E), and equality in that inequality only implies F=H_{ij}, not E=H_{ij}. The remark should either supply the additional argument or be weakened to an assertion of uniqueness up to the equality cases in the max-flow/min-cut reduction.","section":"§5.4, Remark 5.4"},{"comment":"The original matrix σ and its relaxation \\barσ are typographically very similar throughout Sections 3–5, and in several displayed formulas both appear in the same line (for example in Lemma 3.4, Proposition 4.1, and Lemma 5.3). Please use a clearly distinguishable notation, such as \\barσ or \\underlineσ, consistently in all statements and proofs.","section":"§3–§5, notation"},{"comment":"The displayed line starting with 'lim inf (1−2s_k)(1−2s_k)Eσ...' contains a duplicated factor (1−2s_k); it should read lim inf (1−2s_k)Eσ_{2s_k}(E_k,A_{1−δ,1}).","section":"Lemma 4.2, before (4.2)"},{"comment":"In the definition of the competitor F after choosing the minimal cut, the clause '∅ for k≥3' is ambiguous because k is also used as the sequence index; it should read 'F_l=∅ for l=3,...,m' or use a different symbol.","section":"Lemma 5.3, definition of F"},{"comment":"Lemma 3.7 is imported from [5] with only a sketch of proof. Since it is load-bearing for the upper bound, please mark it explicitly as a quoted result with the precise reference to [5, Lemma 3.1] and either omit the sketch or label it as a sketch of the adaptation.","section":"Lemma 3.7"}],"recommendation":"minor_revision","confidential_remarks":"I agree with the reader's positive assessment. The mathematical core is sound: the Γ-convergence chain is complete, the cell formula is computed correctly, and the max-flow/min-cut replacement lemma is a genuine contribution. The issues I found are local presentation problems and one unsupported uniqueness remark; none affects the main theorems. A minor revision addressing the notation and the typos should suffice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it proves the Gamma-convergence of (1-2s) times the fractional sigma-perimeter, for partitions, to the local perimeter with the relaxed coefficient matrix sigma-bar. That relaxed matrix is the shortest-path distance induced by sigma, and the main theorems (1.2 and 1.3) pin down both the Gamma-limit and the behavior of local minimizers. This is genuinely new: previous work covered the single-phase case or additive coefficients, and the relaxation effect for non-additive matrices was open.\n\nThe proof is thorough. The compactness argument is clean, the blow-up lower bound is standard but executed carefully, and the upper bound via polyhedral partitions and the separate relaxation-by-hand is a nice choice. The most interesting piece is the max-flow min-cut replacement lemma (Lemma 5.3), which reduces the multi-phase half-space minimality problem to the two-phase result from Ambrosio–De Philippis–Martinazzi. I checked the path-decomposition step and the use of the triangle inequality for the relaxed matrix; it works. The gluing argument in Section 6 is also handled with the right amount of care, using the generalized co-area formula without pretending the partition constraints disappear.\n\nSoft spots: the proof leans heavily on [3, Proposition 17] (half-space minimality for the two-phase fractional perimeter). That is a published, standard theorem, and the reduction to it is sound, but it is load-bearing. If that result ever turned out to have a gap, parts of this paper would need revisiting. That is not a criticism of this paper, just a statement about what is imported. There are also a few typos and minor notational inconsistencies (e.g., some subscripts appear as '1,h' instead of '1,k'), but nothing that obscures the mathematics.\n\nOverall, the central claim is well supported. The proof is detailed enough that I found no hidden assumption or post-hoc fitting. The authors also give proper credit to prior work and do not overclaim; the introduction is honest about what is new and what is borrowed.\n\nThis paper deserves a serious referee. I would recommend accepting it after minor revisions, and I would bring it to our reading group. It is a solid contribution to the calculus of variations and will be useful for anyone working on nonlocal perimeter models or grain growth.","headline":"A careful, self-contained Gamma-convergence result for multi-phase fractional perimeters with arbitrary surface tensions; the relaxed limit is identified correctly and the proof holds up.","tokens_in":39862,"tokens_out":1058,"would_cite":true,"duration_ms":14371,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q20","49J45","35R11"],"pacs":[],"model":"deepseek-v4-flash","headline":"The fractional perimeter functionals for multiphase partitions converge to a local energy with relaxed surface-tension coefficients.","keywords":["Gamma-convergence","fractional perimeter","partitions","nonlocal perimeter","relaxation","surface tension","half-space minimality","max-flow min-cut"],"falsifier":"Compute, for a five-chamber matrix that violates the triangle inequality but is not $\\ell^1$-embeddable, the minimum of $(1-2s)P^\\sigma_{2s}(E,Q)$ among partitions of the unit cube that agree with the half-space partition outside $Q$, for a sequence $s\\to 1/2^-$; if the value does not approach $\\omega_{n-1}\\bar\\sigma_{ij}$, or if any competitor strictly beats the half-space partition for some $s$, the central claim collapses.","tokens_in":38976,"feed_emoji":"📐","tokens_out":7910,"duration_ms":72378,"temperature":0.7,"pith_summary":"This paper establishes a sharp-interface limit for a family of nonlocal energies defined on partitions of space into $m\\ge 3$ measurable chambers. For any matrix $\\sigma$ of positive surface-tension coefficients, the rescaled fractional $\\sigma$-perimeter $(1-2s)P^\\sigma_{2s}(\\cdot,\\Omega)$ is shown to $\\Gamma$-converge as $s\\to 1/2^-$ to $\\omega_{n-1}$ times the classical perimeter functional $P^{\\bar\\sigma}_1(\\cdot,\\Omega)$, where $\\bar\\sigma$ is the largest componentwise-lower matrix satisfying the triangle inequality. Because $P^{\\bar\\sigma}_1$ is the lower semicontinuous envelope of $P^\\sigma_1$, the limit is well posed even when $\\sigma$ itself violates the triangle inequality. The paper also proves that limits of local minimizers of the nonlocal energies are local minimizers of the relaxed local energy. A sympathetic reader should care because this pins down how interfacial energies with incompatible surface tensions relax through phase nucleation as the interaction range shrinks.","feed_headline":"Nonlocal partition energies sharpen into relaxed surface tensions","feed_subtitle":"The limit replaces broken triangle inequalities by shortest-path coefficients, letting new phases nucleate.","key_machinery":"The load-bearing object is the asymptotic cell formula: for the unit cube $Q$ and upper half-space $H$, the coefficients $\\Gamma_{ij}$ are defined by the lowest possible rescaled energies of partitions that converge to the two-chamber configuration $(H,H^c)$ inside $Q$. The paper reworks this cell formula until it is solved by the half-space partition itself, using a replacement lemma: given any multi-phase competitor, the max-flow min-cut theorem on the complete directed graph whose vertices are the chambers and whose edge capacities are the pairwise interaction energies produces a two-phase competitor of no greater energy. This reduces the multi-phase cell problem to the known two-phase half-space minimality for the fractional perimeter. The upper bound is carried out by polyhedral partitions and a direct computation of the pointwise limit, together with a relaxation step that replaces $\\sigma$ by $\\bar\\sigma$; the convergence of minimizers uses a co-area gluing construction to compare competitors without breaking the partition constraint.","core_discovery":"The central discovery is that the relaxation already visible in local partition perimeters appears automatically in the sharp-interface limit of the nonlocal ones. The proof identifies the limiting matrix explicitly: $\\bar\\sigma_{ij}$ is the infimum over chains $i=i_0,i_1,\\dots,i_H=j$ of $\\sigma_{i_0i_1}+\\cdots+\\sigma_{i_{H-1}i_H}$, equivalently the metric closure of $\\sigma$. In the cell problem for a pair of chambers, the half-space partition is the unique minimizer, and the cell energy equals $\\omega_{n-1}\\bar\\sigma_{ij}$; the multi-phase case is reduced to this two-phase statement by a replacement argument based on the max-flow min-cut theorem. Consequently the $\\Gamma$-limit is $\\omega_{n-1}P^{\\bar\\sigma}_1$, not $\\omega_{n-1}P^\\sigma_1$, and any deficiency created by the failure of the triangle inequality is healed by nucleating intermediate phases in the limit.","pith_inferences":["Inference: By analogy with known threshold-dynamics schemes, the $\\Gamma$-convergence here suggests that curvature-driven network motions dissipating $P^\\sigma_{2s}$ should converge, in the vanishing-interaction limit, to the relaxation flow of $\\omega_{n-1}P^{\\bar\\sigma}_1$; the paper supplies the variational half of that statement but does not construct the flow itself.","Inference: The replacement lemma is likely robust: for any pairwise interaction kernel for which the two-phase half-space minimality holds, the same max-flow min-cut reduction should yield a relaxed $\\Gamma$-limit with the metric-closure coefficients, so the phenomenon is not specific to the fractional kernel.","Inference: A testable quantitative prediction is that for a fixed small $s$ and a matrix violating the triangle inequality, minimizers should form thin layers of intermediate phases near interfaces, with layer width tending to zero as $s\\to 1/2^-$ and the energy gap to $\\omega_{n-1}P^{\\bar\\sigma}_1$ of order $(1/2-s)$; direct numerical simulation could verify the scaling."],"forward_implications":["If the theorem is correct, the sharp-interface limit of fractional multiphase perimeters is a local functional even for coefficients that violate the triangle inequality, and the limiting coefficients are the shortest-path relaxed ones.","Any sequence of local minimizers of the nonlocal energies converges in $L^1_{\\mathrm{loc}}$ to a local minimizer of $\\omega_{n-1}P^{\\bar\\sigma}_1$, with convergence of the rescaled energies on domains whose boundary has zero relaxed perimeter.","The explicit shortest-path formula for $\\bar\\sigma$ makes the relaxation process computable: an interface between phases $i$ and $j$ may split into a chain of intermediate phases whose total surface tension is cheaper.","Compactness holds for equi-bounded sequences: limits of finite-energy partitions are Caccioppoli partitions, so existence and regularity theory for the relaxed problem become available.","For $m=3$ and $m=4$, the triangle inequality forces additive or nearly additive structure, so the half-space minimality is proved by explicit decompositions; for general $m$ it follows from the network-flow argument."],"supporting_citations":[{"why":"Supplies the single-phase Gamma-convergence and the two-phase half-space minimality that the multi-phase argument imports.","marker":"[3]"},{"why":"Establishes the two-phase minimality of half-spaces for the fractional perimeter, which the replacement lemma extends to partitions.","marker":"[11]"},{"why":"Supplies the max-flow min-cut replacement construction for partitions that reduces competitors to two phases.","marker":"[24]"},{"why":"Characterizes lower semicontinuity and relaxation of the local partition functional, yielding the definition of the relaxed coefficient matrix.","marker":"[1]"},{"why":"Provides the polyhedral approximation density used in the upper-bound and relaxation steps.","marker":"[5]"},{"why":"Motivates the nonlocal relaxation via threshold dynamics and gives the additive and l1 representations used in special cases.","marker":"[21]"}],"fun_headline_variants":["Broken triangle inequalities heal in nonlocal partition limits","Nonlocal partition energies converge to local perimeter with relaxed costs","Missing triangle inequality? Nonlocal partition limit nucleates new phases","Relaxed surface tensions appear as Gamma-limit of nonlocal partitions","Metric closure repairs nonlocal partition energy convergence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof imports from the two-phase theory the fact that a half-space minimizes the fractional perimeter among sets agreeing with it outside a small cube; if that statement failed for some $s\\in(0,1/2)$, the cell constant would not equal $\\omega_{n-1}\\bar\\sigma_{ij}$ and the lower bound would not match the upper bound.","fun_headline_variants_meta":{"raw":{"variants":["Broken triangle inequalities heal in nonlocal partition limits","Nonlocal partition energies converge to local perimeter with relaxed costs","Missing triangle inequality? Nonlocal partition limit nucleates new phases","Relaxed surface tensions appear as Gamma-limit of nonlocal partitions","Metric closure repairs nonlocal partition energy convergence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000668,"raw_usage":{"total_tokens":2977,"prompt_tokens":806,"completion_tokens":2171,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":422,"completion_tokens_details":{"reasoning_tokens":2090}},"tokens_in":422,"tokens_out":2171,"duration_ms":15939,"temperature":1.0,"reasoning_tokens":2090,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:21:08.228679+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for a five-chamber matrix that violates the triangle inequality but is not $\\ell^1$-embeddable, the minimum of $(1-2s)P^\\sigma_{2s}(E,Q)$ among partitions of the unit cube that agree with the half-space partition outside $Q$, for a sequence $s\\to 1/2^-$; if the value does not approach $\\omega_{n-1}\\bar\\sigma_{ij}$, or if any competitor strictly beats the half-space partition for some $s$, the central claim collapses.","supporting_citations":[{"cited_title":"Ambrosio, G","cited_arxiv_id":null,"evidence_quote":"Supplies the single-phase Gamma-convergence and the two-phase half-space minimality that the multi-phase argument imports."},{"cited_title":"Caffarelli, J.-M","cited_arxiv_id":null,"evidence_quote":"Establishes the two-phase minimality of half-spaces for the fractional perimeter, which the replacement lemma extends to partitions."},{"cited_title":"Leonardi : Infiltrations in immiscible fluids systems, Proc","cited_arxiv_id":null,"evidence_quote":"Supplies the max-flow min-cut replacement construction for partitions that reduces competitors to two phases."},{"cited_title":"Ambrosio, A","cited_arxiv_id":null,"evidence_quote":"Characterizes lower semicontinuity and relaxation of the local partition functional, yielding the definition of the relaxed coefficient matrix."},{"cited_title":"Baldo : Minimal interface criterion for phase transitions in mixtures of Cahn-Hilliard fluids","cited_arxiv_id":null,"evidence_quote":"Provides the polyhedral approximation density used in the upper-bound and relaxation steps."},{"cited_title":"Esedoglu, F","cited_arxiv_id":null,"evidence_quote":"Motivates the nonlocal relaxation via threshold dynamics and gives the additive and l1 representations used in special cases."}],"review_version":2}