{"id":"38805771-cc25-47fa-b7e4-4b170328b04a","arxiv_id":"2411.17325","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For bounded C^{1,1} domains, every BV function u satisfies the trace inequality with constant 1 on the total variation term, and this fails for some C^{1,α} domains, 0<α<1.","lead":"This paper proves that the trace inequality for functions of bounded variation holds with the best possible constant 1 when the domain has a C^{1,1} boundary, and shows that mildly rougher C^{1,α} boundaries are not enough. The result clarifies when a widely used lower-semicontinuity property for 1-Laplacian boundary value problems is valid.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the proof's reliance on C^{1,1} positive reach and Jacobian control in Theorem 7 is standard and adequately justified.","rationale":"The reader's verdict of ACCEPT with high confidence is reasonable. The central claim is the optimal trace inequality with constant 1 on bounded C^{1,1} domains. I scrutinized the local chart argument in Theorem 7, the C^{1,1} adaptation, the strict-topology approximation, and the counterexamples in Section 4. The proof of the local estimate (3.13) is valid: for C^{1,1} boundaries, positive-reach geometry gives a Lipschitz normal field and bi-Lipschitz charts; the Jacobian positivity and bounded ratio follow from L∞ coefficients and a uniform lower bound on J0. The counterexamples are correctly computed and demonstrate the failure for C^{1,α}, α<1, while their borderline behavior at α=1 matches the theorem. There is no load-bearing gap. The only caveats are minor technical details (choosing ε small enough so that φ_i support is compactly inside V_i, and choosing charts with compact closure for the uniform J0 bound), which a specialist can fill without altering the result. Hence the verdict remains unchanged.","tokens_in":14388,"tokens_out":29975,"duration_ms":271372,"concrete_test":"Verify the borderline case α=1 in the construction of Example 9: for Ω with graph ψ(t)=t^2/2 (C^{1,1} but not C^2 at 0), compute the cap sequence u_n=χ_{E_n}; if the ratio (∫∂Ω|u_n| - ∫Ω|D u_n|)/∫Ω|u_n| stays bounded as r_n→0, the C^{1,1} threshold is consistent with Theorem 7. Additionally, for a non-smooth C^{1,1} example where the normal is only Lipschitz (e.g., a graph with f'(s)=1-|s|), compute the chart Jacobian J(s,t) and its t-derivative numerically to confirm J≥c>0 and |∂_tJ|/J≤C for small ε.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I read Theorem 7 as claiming that the trace inequality (3.11) holds with the optimal constant C1=1 on every bounded C^{1,1} domain. The only place where the argument could fail is the change-of-coordinates step (3.12)-(3.18), since it requires the chart maps h_i(s,t)=g_i(s)+t n(g_i(s)) to be bi-Lipschitz with positive Jacobian and bounded ratio |∂_tJ|/J uniformly for small t. For a C^{1,1} boundary this is supplied by positive reach (Theorem 2): n is Lipschitz, so Dhi exists a.e. in s, J(s,t) is a polynomial in t with coefficients in L∞, and J(s,0)=|g_{s1}∧...∧g_{sN-1}| is bounded below on compact chart domains; hence a uniform ε>0 exists with J≥c>0 and |∂_tJ|/J bounded. The rest of the proof (partition of unity, product rule, strict-topology approximation via BV-version of Meyer-Serrin) is standard. Example 9 correctly shows that for α<1 the excess ∫∂Ω|u_n| - ∫Ω|Du_n| has order r^{N-1+2α}, which dominates the L1 mass r^{N+α}, so C1=1 is impossible; for α=1 the two orders coincide, so no contradiction arises. I found no internal inconsistency or missing essential step that would change the verdict.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the trace inequality for BV functions on bounded domains, aiming at the best constant C1=1 in (1.1). The main result (Theorem 7) establishes that for every bounded C^{1,1} domain Ω there exists C(Ω) such that ∫∂Ω|u| ≤ ∫Ω|Du| + C∫Ω|u| for all u in BV(Ω). The proof uses a tubular neighborhood coordinate system based on the signed distance function, a partition of unity, and a first-order computation showing that the Jacobian J(s,t) satisfies the needed a priori estimate. The paper also proves that for C^1 domains the constant C1 can be taken arbitrarily close to 1 (Theorem 4), shows that C1 cannot be below 1 (Lemma 3), and constructs explicit examples showing that C^{1,α} regularity with 0<α<1 does not suffice (Example 9) and that the associated lower semicontinuity property fails in that setting (Example 10).","tokens_in":14623,"tokens_out":11236,"duration_ms":101788,"significance":"If accepted, the paper resolves a natural question on the exact constant in the BV trace inequality for smooth domains, showing C1=1 for C^{1,1} domains and providing elementary counterexamples for lower Hölder regularity. The paper is rigorous and self-contained: the proof of Theorem 7 is based on explicit local computations, the dependence of the constant on the charts and the mean curvature is made explicit, and the lower-bound examples are explicit and verifiable. The link between positivity of the reach and C^{1,1} regularity (Theorem 1) is valuable, and the paper clarifies the implicit use of C1=1 in Modica's semicontinuity theorem. The examples are carefully chosen to separate the C^{1,1} threshold from C^{1,α}.","major_comments":[],"minor_comments":[{"comment":"In the proof of Theorem 7 for the C^{1,1} case, the quantity C_i = sup_{U_i×(0,ε)} |∂tJ|/J is used; this is finite only if the chart domains are chosen to be bounded and the Jacobian J(s,0) is uniformly positive on the support of the partition of unity. Since ∂Ω is compact, such a finite sub-atlas exists, but this should be stated explicitly, otherwise the supremum may be infinite for a chart with unbounded parameter domain.","section":"Section 2.2 / Section 3"},{"comment":"In the definition of h_i(s,t) = g_i(s) + t n(g(si)), the argument of n should be g_i(s), not g(si); the current notation is a typo.","section":"Equation (2.4)"},{"comment":"After (3.18), the sentence 'the change of variables x = h_i(x,t) must be performed' should read x = h_i(s,t).","section":"Equation (3.18)"},{"comment":"The introduction's phrase 'optimum smoothness requirements' suggests a necessary-and-sufficient characterization, whereas the paper actually establishes sufficiency of C^{1,1} and non-sufficiency of the class C^{1,α} for every α<1; it does not rule out that some individual C^{1,α} domains satisfy the inequality. The wording should be softened to avoid overclaiming.","section":"Introduction / Section 4"},{"comment":"In the definition of E_n, the sequence r_n is introduced without an explicit monotonicity or convergence statement; for clarity, state that r_n → 0+ as n → ∞.","section":"Section 4, Example 9"},{"comment":"The sentence 'founded by MCIN/AEI' should read 'funded by MCIN/AEI'.","section":"Declarations"}],"recommendation":"minor_revision","confidential_remarks":"The paper is within the scope of the journal and makes a solid contribution. The only concerns are local presentation issues and an implicit compactness point in the atlas construction; none affects the central claim. I recommend minor revision rather than immediate acceptance in order to address these clarifications."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper closes a real gap. Modica assumed C1=1 for smooth domains; Giusti only suggested it; Motron claimed piecewise C^1 and is wrong. Sabina de Lis proves the constant-1 inequality for bounded C^{1,1} domains and gives explicit C^{1,α} domains, 0<α<1, where the inequality fails. The examples are simple and convincing; they also show lower semicontinuity of J(u)=∫Ω|Du| + ∫∂Ω F(x,u) fails for C^{1,α}, so the C^{1,1} assumption is not a technical artifact. This is a genuine within-field advance, not a big conceptual leap.\n\nThe proof is honest local geometry plus partition of unity. The chart h_i(s,t)=g_i(s)+t n(s) with J(s,t)=det Dh_i is standard; for C^{1,1} boundary, n is Lipschitz, J(s,t) is polynomial in t with L∞ coefficients, and J(s,0) is bounded below on compact chart domains, so the required uniform estimates on J and |∂tJ|/J hold. The C^2 case is presented first and then extended by Rademacher; that is the right order. The correction of Motron's piecewise C^1 claim is backed by the cone example showing C1≥√(1+L^2)>1. The author also computes the normal curvature blow-up for the C^{1,α} example, which explains why the positive-reach machinery fails there.\n\nSoft spots are minor and mostly implicit. The proof would be easier to check if the uniform lower bound on J(s,t) and the finiteness of C_i=sup|∂tJ|/J were stated explicitly rather than left in the phrase \"J(s,t) can be assumed positive for ε small.\" Similarly, the change-of-variables formula for Lipschitz homeomorphisms is cited but the details that h_i is bi-Lipschitz with Lipschitz inverse are only in Theorem 2. A referee should ask for those two sentences. The paper is a bit dense in Section 3; the polynomial-in-t observation for J deserves a display. None of this threatens the argument.\n\nCitation pattern is fine: Modica, Giusti, and Motron are discussed accurately, and the self-citations are for computational details and do not carry the proof.\n\nBottom line: the paper deserves a serious referee. I would send it out. The results are right, the counterexamples are explicit, and the sharpness at C^{1,1} versus C^{1,α} is exactly the kind of precision people working on 1-Laplacian boundary value problems need. I would not insist on major changes.","headline":"Sharp BV trace inequality with constant 1 on C^{1,1} domains, with matching counterexamples; the main result is correct and the regularity threshold is essentially optimal.","tokens_in":15251,"tokens_out":4224,"would_cite":true,"duration_ms":37669,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46E35","26D10","35J92"],"pacs":[],"model":"deepseek-v4-flash","headline":"On bounded $C^{1,1}$ domains, the BV trace inequality holds with the sharp constant 1.","keywords":["trace inequality","functions of bounded variation","sets of finite perimeter","C^{1,1} domain","positive reach","lower semicontinuity","one-Laplacian","boundary trace"],"falsifier":"Take the $C^{1,\\alpha}$ domain of Example 9 and the sets $E_n=\\{x_N>|x'|^{1+\\alpha}/(1+\\alpha),\\,x_N<\\psi(r_n)\\}$ with $r_n\\to0$. If the claimed optimal inequality held there, then $\\int_{\\partial\\Omega}|u_n|-\\int_\\Omega|Du_n|$ would have to be bounded by $C\\int_\\Omega|u_n|$; the paper's asymptotic gives the left side of order $r_n^{N-1+2\\alpha}$ and the right side of order $r_n^{N+\\alpha}$, so the ratio tends to infinity because $\\alpha<1$. Recomputing that expansion is the direct check.","tokens_in":14122,"feed_emoji":"📐","tokens_out":10608,"duration_ms":83486,"temperature":0.7,"pith_summary":"This paper proves that the boundary trace inequality for functions of bounded variation, $\\int_{\\partial\\Omega}|u| \\le C_1\\int_\\Omega|Du| + C_2\\int_\\Omega|u|$, holds on bounded $C^{1,1}$ domains with the sharp coefficient $C_1 = 1$, and that this is the best possible value. The regularity assumption is essentially optimal: for every $0<\\alpha<1$ there are $C^{1,\\alpha}$ domains where the same inequality with $C_1=1$ fails for every finite $C_2$. The result matters because variational problems for the one-Laplacian and for phase transitions with boundary contact terms need exactly this constant-one inequality to preserve lower semicontinuity of the energy. The paper also identifies the precise regularity hidden behind a widely used 'smooth domain' assumption in an earlier semicontinuity theorem.","feed_headline":"Best trace constant is 1 on C^{1,1} domains","feed_subtitle":"The sharp regularity threshold is C^{1,1}; slightly weaker C^{1,α} boundaries defeat the inequality.","key_machinery":"The load-bearing object is the tubular coordinate system built from the inward unit normal field $n$ on $\\partial\\Omega$. On a $C^{1,1}$ boundary the normal is Lipschitz, so the charts $h_i(s,t)=g_i(s)+t n(g_i(s))$ are bi-Lipschitz onto a uniform neighborhood, the Jacobian $J(s,t)=\\det Dh_i(s,t)$ stays positive, and the ratio $|\\partial_t J|/J$ is bounded by a constant related to the mean curvature of the boundary, which is exactly the positive-reach property. The trace inequality is obtained by writing the surface integral as $\\int |u_i(s,0)|J(s,0)\\,ds$, using the identity $\\partial_t(u_iJ)=\\partial_tu_i J+u_i\\partial_tJ$ to trade the boundary term for volume integrals, and absorbing $|\\partial_tJ|/J$ into the $\\int_\\Omega|u|$ term. The boundedness of this ratio is precisely where the argument uses positive reach; without a Lipschitz normal the local estimate breaks, and the $C^{1,\\alpha}$ counterexamples exhibit that breakdown.","core_discovery":"On the paper's own terms, the central claim is Theorem 7: if $\\Omega\\subset\\mathbb{R}^N$ is a bounded $C^{1,1}$ domain, then there exists a constant $C=C(\\Omega)$ such that $\\int_{\\partial\\Omega}|u| \\le \\int_\\Omega|Du| + C\\int_\\Omega|u|$ for every $u\\in BV(\\Omega)$. The coefficient of the total variation is exactly $1$, and Lemma 3 shows that $C_1\\ge 1$ for any competing inequality, so the constant is optimal. The proof decomposes $u$ with a partition of unity, uses the normal-coordinate charts $h_i(s,t)=g_i(s)+t n(g_i(s))$, and converts the boundary integral into interior integrals through $\\partial_t(u_iJ)=\\partial_tu_iJ+u_i\\partial_tJ$, with boundedness of $|\\partial_tJ|/J$ supplied by the Lipschitz normal field. Example 9 then shows optimality by constructing $C^{1,\\alpha}$ cusp domains, with $0<\\alpha<1$, where the characteristic function of a thin layer has $\\int_{\\partial\\Omega}|u_n|-\\int_\\Omega|Du_n|$ of order $r_n^{N-1+2\\alpha}$ while $\\int_\\Omega|u_n|$ is only of order $r_n^{N+\\alpha}$, so no finite $C$ can close the gap. Example 10 transfers this failure to the lower semicontinuity of $J_-(u)=\\int_\\Omega|Du|-\\int_{\\partial\\Omega}|u|$.","pith_inferences":["Beyond the paper's own statement, the proof's mechanism suggests that the same $C_1=1$ result should extend to any bounded domain whose boundary has positive reach, including some piecewise-smooth boundaries with no reentrant corners.","For the one-Laplacian and phase-transition literature that invokes lower semicontinuity under a vague 'smooth domain' hypothesis, the paper's sharpened threshold suggests those arguments should either assume $C^{1,1}$ or prove the inequality directly for their specific domains.","A quantitative version of Theorem 7 is likely available: the constant $C$ can be chosen in terms of the supremum of $|(N-1)H|$ and the chart data, so domains in a family with uniformly controlled mean curvature would satisfy the inequality with a uniform constant."],"forward_implications":["On every bounded $C^{1,1}$ domain, the trace inequality holds with $C_1=1$; since $C_1\\ge1$ is forced, this is the optimal constant for all such domains.","The lower semicontinuity of $J(u)=\\int_\\Omega|Du|+\\int_{\\partial\\Omega}F(x,u)$ for functions $F$ with Lipschitz constant at most $1$ in $u$ is valid on $C^{1,1}$ domains, and Example 10 shows that $C^{1,\\alpha}$ regularity does not suffice.","The passage from $W^{1,1}$ to $BV$ by strict approximation preserves the inequality with the same constants, so the result applies to characteristic functions of sets of finite perimeter.","On merely $C^1$ domains the inequality holds with $C_1$ arbitrarily close to $1$ but not necessarily equal to $1$, and on piecewise-$C^1$ domains $C_1$ can be forced to exceed $1$."],"supporting_citations":[{"why":"Supplies the BV trace operator and its continuity in the strict topology, used to pass from smooth functions to BV.","marker":"[2]"},{"why":"Provides the Meyer-Serrin approximation for BV functions and the Lipschitz change-of-variables formula used in the C^{1,1} case.","marker":"[8]"},{"why":"Introduces the reach of a boundary point and the tubular neighborhood structure underlying the coordinate system.","marker":"[9]"},{"why":"States the suggestion that bounded mean curvature yields C1=1, which the paper makes precise and proves.","marker":"[12]"},{"why":"Gives the regularity of the signed distance to a C^k hypersurface, used for the tubular coordinates.","marker":"[15]"},{"why":"Supplies the equivalence between positive reach and C^{1,1} boundary regularity.","marker":"[18]"},{"why":"Contains the lower semicontinuity theorem whose proof requires C1=1, motivating the whole paper.","marker":"[21]"},{"why":"Establishes that the infimum of C1 is 1 for C^1 domains via an epsilon argument, which Theorem 7 sharpens.","marker":"[22]"}],"fun_headline_variants":["Optimal trace constant 1 requires C^{1,1} regularity","Trace constant 1 proved sharp on C^{1,1} domains","C^{1,1} domains: best trace constant is exactly 1","Sharp trace bound: constant 1, but only on C^{1,1}","Trace inequality optimum: constant 1 on smooth C^{1,1}"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof collapses if the boundary's unit normal field is not Lipschitz, because then the normal-coordinate map has no uniform positive Jacobian with bounded logarithmic derivative; the $C^{1,\\alpha}$ examples show this is not a removable technicality.","fun_headline_variants_meta":{"raw":{"variants":["Optimal trace constant 1 requires C^{1,1} regularity","Trace constant 1 proved sharp on C^{1,1} domains","C^{1,1} domains: best trace constant is exactly 1","Sharp trace bound: constant 1, but only on C^{1,1}","Trace inequality optimum: constant 1 on smooth C^{1,1}"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001374,"raw_usage":{"total_tokens":5550,"prompt_tokens":909,"completion_tokens":4641,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":525,"completion_tokens_details":{"reasoning_tokens":4541}},"tokens_in":525,"tokens_out":4641,"duration_ms":29420,"temperature":1.0,"reasoning_tokens":4541,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:15:41.259633+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the $C^{1,\\alpha}$ domain of Example 9 and the sets $E_n=\\{x_N>|x'|^{1+\\alpha}/(1+\\alpha),\\,x_N<\\psi(r_n)\\}$ with $r_n\\to0$. If the claimed optimal inequality held there, then $\\int_{\\partial\\Omega}|u_n|-\\int_\\Omega|Du_n|$ would have to be bounded by $C\\int_\\Omega|u_n|$; the paper's asymptotic gives the left side of order $r_n^{N-1+2\\alpha}$ and the right side of order $r_n^{N+\\alpha}$, so the ratio tends to infinity because $\\alpha<1$. Recomputing that expansion is the direct check.","supporting_citations":[{"cited_title":"Ambrosio, N","cited_arxiv_id":null,"evidence_quote":"Supplies the BV trace operator and its continuity in the strict topology, used to pass from smooth functions to BV."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Meyer-Serrin approximation for BV functions and the Lipschitz change-of-variables formula used in the C^{1,1} case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the reach of a boundary point and the tubular neighborhood structure underlying the coordinate system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the suggestion that bounded mean curvature yields C1=1, which the paper makes precise and proves."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the regularity of the signed distance to a C^k hypersurface, used for the tubular coordinates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the equivalence between positive reach and C^{1,1} boundary regularity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contains the lower semicontinuity theorem whose proof requires C1=1, motivating the whole paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that the infimum of C1 is 1 for C^1 domains via an epsilon argument, which Theorem 7 sharpens."}],"review_version":1}