{"id":"e3f03086-4945-4563-a065-e6ca3e81c570","arxiv_id":"1908.09373","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Aggregation-diffusion energies on domains with boundaries have an existence threshold set by the effective volume dimension, and asymmetric unbounded domains admit no minimizer without external confinement.","lead":"Swarm equilibria in regions with boundaries are governed by the number of directions in which the region extends to infinity, not just its full dimension. The paper shows that in asymmetric unbounded regions a diffusing swarm has no stable resting state unless an external force confines it.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.1 is false for fD=0: finite-volume unbounded domains satisfy its hypothesis but admit minimizers, so the sharp-threshold claim needs an infinite-volume or fD>0 assumption.","rationale":"The reader's verdict was CONDITIONAL, with the weakest assumption identified as the trace-theorem gap in Theorem 5.3. That is a real proof gap: the divergence theorem is applied to possibly non-Lipschitz unbounded domains using only Assumption 2(ii), and ρ is not shown to lie in the Sobolev space needed for classical traces. However, the more decisive and easily checkable defect is Theorem 5.1 at fD=0. This is not merely a missing regularity hypothesis: the theorem as stated is false, with a concrete two-dimensional finite-volume domain and K=0,V=0 satisfying every stated assumption and contradicting the conclusion. This matters because the paper's central claim (i) is that diffusion-dominated spreading depends on fD; the fD=0 case shows a second condition, infinite volume (or sufficiently fast volume growth), is needed. The fix is mild -- add fD>0 or infinite volume to Theorem 5.1 -- and the product-domain sharp threshold in Theorem 6.1 (where fD=d-m≥1) is unaffected. For that reason the overall verdict need not change from CONDITIONAL, but the manuscript should be revised to restrict Theorem 5.1 and state the corrected dichotomy. Agreement is partial because the reader's weakest assumption concerned Theorem 5.3, while my load-bearing concern is the counterexample to Theorem 5.1; I still regard the trace-theorem point as a legitimate secondary issue that also warrants a repair.","tokens_in":29078,"tokens_out":19191,"duration_ms":204452,"concrete_test":"Verify the counterexample to Theorem 5.1: fix ν>0 and take D = {(x,y)∈R²: x≥0, 0≤y≤e^{-x}}, K=0, V=0. Confirm fD=0 by noting |D∩Br(x)|≤1 for all r,x and |D∩Br(0)|→1 as r→∞. Confirm condition (5.1) holds with any δ0∈(0,1), C0=0, R0=1. Then compute Eν[ρ]=ν∫_D ρ log ρ for ρ∈P(D) and apply Jensen's inequality to show inf Eν = 0, attained by ρ≡1 (since |D|=1), contradicting the theorem's conclusion. If the authors intend Theorem 5.1 only for domains with fD>0, re-run the proof with an explicit infinite-volume assumption and check whether any later results (Theorems 6.3, 6.4, or the 'sharp threshold' discussion) rely on the unrestricted fD=0 case.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 5.1 (Section 5.1) asserts that if fD=0 and K(x) is bounded above at infinity, then Eν is unbounded below and no global minimizer exists. The proof constructs Dn = D ∩ Bn(xn) with |Dn| ≥ C n^{fD} and derives Eν[µn] ≤ -δ0 fD ν log n + C. When fD=0 this lower bound on |Dn| gives only |Dn| ≥ C, so the logarithmic divergence disappears and the proof collapses. Moreover, the statement is false for finite-volume unbounded domains. Let D = {(x,y): x≥0, 0≤y≤e^{-x}} ⊂ R², so |D|=1, D is closed and connected, and VD(r) ≤ 1 with fD=0. Take K=0, V=0. Then condition (5.1) holds vacuously (0 ≤ 2(1-δ0)·0·ν log|x|), yet by Jensen's inequality Eν[ρ] = ν∫ρ log ρ ≥ -ν log|D| = 0, and the uniform density ρ≡1 on D is a global minimizer. Thus the claimed dependence of diffusion-dominated spreading on the effective volume dimension is incomplete: fD=0 conflates finite-volume domains (where spreading cannot drive Eν to -∞) with sub-polynomial-volume domains where it can. The central sharp-threshold statement in the abstract and in the discussion around Theorems 5.1 and 6.1 is therefore not valid as stated for the full class of domains allowed by Assumption 2.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the energy functional (1.1) for an aggregation model with linear diffusion in domains with boundaries, and asks when global minimizers exist. The main theoretical contributions are (i) a non-existence theorem, Theorem 5.1, relating diffusion-dominated spreading to an 'effective volume dimension' f_D of the domain; (ii) a new 'escaping mass' phenomenon illustrated by the half-line example in Theorem 5.2 and encoded in the boundary necessary condition (5.5) of Theorem 5.3; and (iii) existence results, Theorems 6.1, 6.3, and 6.4, under growth conditions on K and confining conditions on V. The paper also proposes a fixed-point iteration for computing critical points and reports several numerical experiments, including examples with multiple critical points. The authors claim that the domain geometry, through f_D, determines the sharp threshold for existence versus non-existence, and that external potentials are necessary to confine swarms in general domains.","tokens_in":29443,"tokens_out":9514,"duration_ms":105115,"significance":"If the main results were valid as stated, the paper would extend the free-space sharp conditions of Carrillo, Delgadino, and Patacchini [13] to bounded and unbounded domains with boundaries, and would identify a genuinely new domain-dependent mechanism for non-existence. The paper is careful in several places: it gives a self-contained proof that minimizers have full support, it provides explicit benchmark computations for the half-line, and the numerical method is described in reproducible detail. However, the central sharp-threshold claim is not valid as stated for f_D=0, and the boundary-trace argument in Theorem 5.3 requires stronger domain regularity than Assumption 2 provides. These are load-bearing issues, so the current version cannot be accepted without substantial revision.","major_comments":[{"comment":"The statement of Theorem 5.1 is false when f_D=0, because condition (5.1) then reduces to a mere upper bound on K and the proof's estimate |D_n| >= C n^{f_D} degenerates to a constant lower bound; the logarithmic divergence -delta0 f_D nu log(n) disappears. A concrete counterexample is D = {(x,y) : x>=0, 0<=y<=e^{-x}} subset R^2, with K=0 and V=0. This domain is closed, connected, has |D|=1, and satisfies Assumption 2; moreover f_D=0 since VD(r) <= 1 for all r. Condition (5.1) holds with C0=0 because 0 <= 0 for all |x|. Yet by Jensen's inequality, E_nu[rho] = nu integral rho log rho >= -nu log|D| = 0 for every probability density rho on D, and the uniform density rho=1_D attains equality, so a global minimizer exists. Thus the claimed sharp dependence on f_D is incomplete: f_D=0 does not distinguish finite-volume unbounded domains, where spreading cannot drive the energy to -infinity, from domains with slower-than-polynomial volume growth where it can. The theorem and the associated discussion in the abstract should either assume f_D>0 or replace f_D by a more refined volume-growth exponent.","section":"Section 5.1, Theorem 5.1"},{"comment":"The proof of Theorem 5.3 applies the divergence theorem and classical trace theorems [21, Ch. 5] to the possibly unbounded domain D under Assumption 2(ii), which only requires the outward normal to exist almost everywhere on the boundary. Classical trace theorems require Lipschitz or more regular boundaries, and for unbounded domains one also needs uniform control of the exhaustion sets A_n used in the proof. In addition, the validity of the step 'integral over D of grad rho = boundary integral' requires grad rho in L^1(D), which is not established from the Euler-Lagrange equation under the stated assumptions. Consequently the necessary condition (5.5), and with it the claim in Remark 5.2 that external forces are necessary to confine the swarm in general domains, is not proved for the full class of domains admitted by Assumption 2. The authors should either strengthen Assumption 2 (for instance, assume D is Lipschitz and impose the integrability needed for the trace theorem) or give a proof that the trace of rho exists and is integrable on the boundary under their weaker assumptions.","section":"Section 5.2, Theorem 5.3"},{"comment":"Corollary 3.1 states an 'if and only if' relationship between the Euler-Lagrange equation (3.3) and fixed points of the map T in (3.8) without proof and without specifying conditions under which Z(mu) is finite and the convolution K*mu is sufficiently regular to justify the exponential representation. This result is used in Theorem 5.2 and Theorem 6.2, so it is not purely cosmetic. A short argument or a precise citation should be supplied, including integrability assumptions on K*mu and V.","section":"Section 3, Corollary 3.1"}],"minor_comments":[{"comment":"The definition of f_D as a supremum does not automatically imply the stated consequence (2.13) at s=f_D; for example, a supremum of an open set of exponents need not be attained. Please either define f_D as the largest exponent for which (2.13) holds, or add an assumption that the supremum is attained, or alter (2.13) to hold for every s0 that might have empty interior, so either the domain should be assumed open with nonempty interior (or otherwise regular) or the statement should be adjusted.","section":"Proof of Theorem 3.1"},{"comment":"The wedge and paraboloid examples are only sketched; the sign computations of the boundary integrals should be written out explicitly, since the claim that N_d 'cannot be zero' is central to the illustration of Remark 5.2.","section":"Section 5.3, Examples"},{"comment":"The L1-Lipschitz bound (7.5) is cited to the first author's master's thesis; if this bound is used to justify the choice tau_c=O(nu), it would help to state the exact hypotheses under which it holds or to label the choice as heuristic.","section":"Section 7, Eq. (7.5)"}],"recommendation":"major_revision","confidential_remarks":"The f_D=0 counterexample is a genuine counterexample to the stated Theorem 5.1 and directly affects the abstract's sharp-condition claim. The trace-theorem issue in Theorem 5.3 is also substantive but fixable. The paper's core idea---that domain geometry enters the spreading threshold through a volume-growth exponent and that boundary asymmetries create metastable translation---seems worth pursuing, and the explicit half-line computations are valuable. I recommend major revision rather than rejection because both problems can be addressed by adding hypotheses and adjusting the statements, without changing the overall framework."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one. The core idea is good: replace the free-space dimension d in the log-growth threshold with an effective volume dimension fD, and identify a second non-existence mechanism — metastable mass translation — that forces external confinement in asymmetric domains. The product-domain sharp threshold (Theorem 6.1) and the escaping-mass necessary condition (5.5) are real contributions, and the numerical examples support the intuition.\n\nBut there is a load-bearing flaw in Theorem 5.1. As stated, it says that if fD=0 and K is bounded above at infinity (so condition (5.1) holds vacuously), the energy is unbounded below. That's false. Take D = {(x,y): x≥0, 0≤y≤e^{-x}} ⊂ R², which is closed, connected, has |D|=1, fD=0, and satisfies Assumption 2. With K=0, V=0, condition (5.1) is satisfied (RHS is a constant), yet Eν[ρ] = ν∫ρ log ρ is bounded below by −ν log|D| = 0, and the uniform density is a global minimizer. The proof collapses because the lower bound |D∩Bn(xn)| ≥ C n^{fD} gives only a constant when fD=0, so the log divergence disappears. The sharp-threshold claim needs an additional assumption like fD>0 or infinite volume; otherwise it conflates finite-volume unbounded domains (where spreading cannot drive E to −∞) with sub-polynomial-volume domains where it can.\n\nThe secondary issue is Theorem 5.3. The proof invokes classical trace theorems to justify ∫_D ∇ρ = ∫_{∂D} n ρ, but Assumption 2 only gives an a.e. outward normal. Trace theorems typically require at least Lipschitz boundary. If D is not regular enough, the necessary condition (5.5) — and the claim that external forces are necessary — is not established for the full domain class. This is a correctable gap, but it should be fixed by either strengthening Assumption 2 or replacing the trace argument with a distributional one.\n\nMinor points: Corollary 3.1's 'if and only if' is asserted without proof (the 'only if' direction needs a uniqueness argument for the fixed point), and the numerics are described but no code is shipped, so reproducibility is limited.\n\nOverall: the paper deserves a serious referee. The main idea is original and likely right for the right domain class, but Theorem 5.1's overstatement is a genuine mathematical error that needs repair before the sharp-threshold claim can stand as stated.","headline":"The effective-volume-dimension idea is good and the escaping-mass phenomenon is real, but Theorem 5.1 as stated is false for fD=0, and the sharp-threshold claim needs a corrected assumption.","tokens_in":29951,"tokens_out":2882,"would_cite":true,"duration_ms":25530,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A15","35Q92","49Q22"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that in domains with boundaries, the existence of swarm equilibria is controlled by an effective volume dimension and by whether the domain is symmetric enough to prevent mass from escaping to infinity.","keywords":["aggregation-diffusion","swarm equilibria","domain boundaries","effective volume dimension","energy minimizers","metastability","Wasserstein gradient flow","nonlocal interaction"],"falsifier":"Take the half-line D=[0,∞), K(x)=$x^{2}$/2, V=0, and compute the energy E_ν on the family ρ_c of truncated Gaussians (as in the proof of Theorem 5.2). The proof claims c↦E_ν[ρ_c] is strictly decreasing with no critical point; numerically finding a stationary c would refute the non-existence claim in that example.","tokens_in":28851,"feed_emoji":"🐝","tokens_out":7474,"duration_ms":68801,"temperature":0.7,"pith_summary":"This paper studies equilibrium states of a continuous swarm that moves by nonlocal attraction or repulsion plus linear diffusion, confined to a spatial domain with a boundary. It argues that two features of the domain geometry control whether an equilibrium exists: the effective volume dimension f_D, which counts how many independent directions extend to infinity, replaces the ambient dimension in the sharp balance between diffusion and attraction; and asymmetry of the domain creates metastable mass translation, so that without an external potential the swarm's center of mass drifts to infinity and no global minimizer exists. In a class of domains of the form F×$R^{{d−m}}$, the paper proves a sharp threshold: attraction growing faster than 2 f_D ν log|x| gives existence, while slower growth makes the energy unbounded below. For general domains it derives a necessary boundary-force balance, and shows that a confining force restores existence. The upshot is that in unbounded domains with boundaries, diffusion plus nonlocal attraction alone is not enough to confine a swarm: the boundary itself pushes it away.","feed_headline":"Effective volume dimension decides if a swarm spreads to infinity","feed_subtitle":"In asymmetric domains, only an external force can stop the swarm from sliding off to infinity.","key_machinery":"The key object is the effective volume dimension f_D, defined through V_D(r)=sup_{x∈D}|D∩B_r(x)| and f_D=sup{s: V_D(r)≳r^s}; it measures the number of directions in which the domain extends to infinity and appears in place of d in the logarithmic threshold for K. The other load-bearing tool is the Euler-Lagrange equation K∗ρ+ν log ρ+V=λ together with the fixed-point map T(µ)=Z(µ)^{-1}exp(−(K∗µ+V)/ν), whose fixed points are exactly the critical points with full support. From the Euler-Lagrange equation, the paper derives the boundary condition (5.5) by taking gradients, using that ∇K's antisymmetry cancels the interaction term, and applying the divergence theorem; this condition is what exposes the escaping-mass phenomenon.","core_discovery":"The central claim is that in domains with boundaries, the condition for existence of global minimizers of the energy E_ν = 1/2∫∫K(x−y)dµdµ + ν∫ρ log ρ + ∫V dµ no longer reduces to boundedness from below, as it does in free space. Instead, two geometry-dependent obstructions appear. First, diffusion-dominated spreading is controlled by the effective volume dimension f_D: if K(x) grows no faster than 2(1−δ)f_Dν log|x| at infinity, the energy is unbounded below, while for domains D=F×$R^{{d−m}}$, growth above 2(1+δ)f_Dν log|x| forces existence. Second, asymmetric unbounded domains exhibit an escaping-mass phenomenon: the necessary condition ν∫_{∂D}nρ dS = −∫_D ρ∇V dx, obtained by differentiating the Euler-Lagrange equation and using antisymmetry of ∇K, cannot be satisfied for V=0 because the density is strictly positive on the boundary, so critical points—and hence minimizers—do not exist. Thus external forces are necessary to confine the swarm in general domains, in contrast to free space.","pith_inferences":["The effective volume dimension may be computable for more general domains than F×R^{d−m} (e.g., parabolic or spiral regions), giving a concrete way to predict when a swarm in an obstacle field must be confined externally; measuring V_D(r) numerically would provide a test.","The escaping-mass mechanism suggests a design principle for experimental or engineered swarm confinement: a domain whose boundary has zero average outward normal along the region where the swarm sits (e.g., a periodic channel) can equilibrate without external forcing, while any symmetry-breaking defect will pump the swarm toward the boundary.","One could test whether the necessary boundary condition (5.5) survives the addition of nonlinear diffusion; if it does, the conclusion that boundaries alone cannot confine a swarm may extend to degenerate-diffusion models, where compactly supported equilibria are otherwise expected.","The numerical continuation-from-large-ν heuristic, which found the lower-energy four-aggregate state, suggests a general annealing procedure for non-convex interaction energies on bounded domains; whether it always lands on a global minimizer is not proved here and is a natural next question."],"forward_implications":["If the sharp threshold holds, then in a cylinder-like domain D=F×R^{d−m}, attraction growing exactly as 2f_Dν log|x| is the borderline: slightly weaker attraction blows the energy to −∞, slightly stronger yields a global minimizer.","For any unbounded domain without translation symmetry (half-space, wedge, paraboloid), zero external force means no global minimizer exists, no matter how strong the short-range attraction is—only an external potential can confine the swarm.","The boundary condition (5.5) gives a precise stationary-state test: an equilibrium requires the boundary flux ν∫∂D nρ dS to exactly cancel the force ∫_D ρ∇V dx; in simulations, this can be checked pointwise.","Since every minimizer must have full support D (Theorem 3.1), linear diffusion prevents compactly supported equilibria even when attraction is very strong; compact support emerges only in the p→∞ singular limit.","Adding a coercive confining potential V (growing to +∞ at infinity), or a periodic/translation-invariant domain structure, restores existence for general domains (Theorems 6.3 and 6.4)."],"supporting_citations":[{"why":"Supplies the free-space existence theory, the logarithmic Hardy-Littlewood-Sobolev inequality, and the tightness-up-to-translations lemma that the domain-geometry results extend and modify.","marker":"[13]"},{"why":"Provides the classical trace theorem used to turn the integrated gradient of the Euler-Lagrange equation into the boundary condition (5.5).","marker":"[21]"},{"why":"Establishes the Wasserstein gradient-flow formulation that identifies critical points of E_ν with steady states of the aggregation-diffusion PDE.","marker":"[1]"},{"why":"Provides the variational technique (perturbation inside the Wasserstein ball) used to derive the Euler-Lagrange equation.","marker":"[2]"},{"why":"Shows the fixed-point map (3.8) can be used to construct stationary states, motivating the numerical method and the fixed-point characterization in Corollary 3.1.","marker":"[5]"},{"why":"Supplies the L1-Lipschitz bound on the fixed-point map that guides the relaxation parameter τ_c in the numerical scheme.","marker":"[37]"}],"fun_headline_variants":["Boundary shape decides if a swarm spreads to infinity","Domain symmetry dictates swarm confinement needs","Effective dimension sets swarm spreading threshold","Asymmetric domains force external swarm confinement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the necessary boundary condition assumes the classical trace theorem applies to the possibly unbounded domain, so if the boundary is too irregular for that theorem, the conclusion that external forces are required to confine a swarm is not established.","fun_headline_variants_meta":{"raw":{"variants":["Boundary shape decides if a swarm spreads to infinity","Domain symmetry dictates swarm confinement needs","Effective dimension sets swarm spreading threshold","Asymmetric domains force external swarm confinement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000125,"raw_usage":{"total_tokens":1103,"prompt_tokens":936,"completion_tokens":167,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":115}},"tokens_in":552,"tokens_out":167,"duration_ms":2275,"temperature":1.0,"reasoning_tokens":115,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:15:46.947925+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the half-line D=[0,∞), K(x)=$x^{2}$/2, V=0, and compute the energy E_ν on the family ρ_c of truncated Gaussians (as in the proof of Theorem 5.2). The proof claims c↦E_ν[ρ_c] is strictly decreasing with no critical point; numerically finding a stationary c would refute the non-existence claim in that example.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the free-space existence theory, the logarithmic Hardy-Littlewood-Sobolev inequality, and the tightness-up-to-translations lemma that the domain-geometry results extend and modify."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the classical trace theorem used to turn the integrated gradient of the Euler-Lagrange equation into the boundary condition (5.5)."},{"cited_title":"Ambrosio, N","cited_arxiv_id":null,"evidence_quote":"Establishes the Wasserstein gradient-flow formulation that identifies critical points of E_ν with steady states of the aggregation-diffusion PDE."},{"cited_title":"Balagu´ e, J","cited_arxiv_id":null,"evidence_quote":"Provides the variational technique (perturbation inside the Wasserstein ball) used to derive the Euler-Lagrange equation."},{"cited_title":"Benachour, B","cited_arxiv_id":null,"evidence_quote":"Shows the fixed-point map (3.8) can be used to construct stationary states, motivating the numerical method and the fixed-point characterization in Corollary 3.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the L1-Lipschitz bound on the fixed-point map that guides the relaxation parameter τ_c in the numerical scheme."}],"review_version":1}