{"id":"60907239-ba88-402f-b392-af73969042ef","arxiv_id":"1908.09322","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"If a bounded extension operator from L^1_p(Ω) to L^1_q(R^n) exists with n < q ≤ p, then Ω must satisfy a generalized (p,q)-measure density inequality and a weak equivalence between its intrinsic and Euclidean metrics.","lead":"This paper derives necessary conditions for extending Sobolev functions from a domain to all of Euclidean space when the integrability exponent decreases. The conditions are formulated through a set function associated with the extension operator and yield lower bounds on the operator's norm.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Countable additivity of Phi is unproved: Theorem 2.1 relies on disjoint supports of grad(E(f_k)), which nonlocal extension operators need not have for A_k not contained in Omega; Theorems 2.5 and 2.8 depend on this.","rationale":"Reading the paper in good faith, the main density condition in Theorem 2.4 is plausible: it follows from inequality (2.1), Holder, and Sobolev embedding, and the Holder-cusp example supports sharpness. The reader's weakest-assumption analysis is correct, and I agree with it. The flaw is not stylistic: without a valid proof that Phi is countably additive, or at least superadditive on the boundary-crossing balls used in differentiating K(x), the derivation of Theorem 2.5 is missing its key analytic step. Theorem 2.8 is explicitly listed without proof, so it also cannot be accepted on the present evidence. I do not see a problem with Theorem 2.4 itself that would force rejection; the necessary condition may well be true after minor technical repairs. Hence the appropriate verdict remains conditional, i.e., no change from the reader. The issue is confined to the argument, not to the author's intent.","tokens_in":8345,"tokens_out":26479,"duration_ms":259008,"concrete_test":"Construct a finite-dimensional test of superadditivity. Let Omega be the unit ball in R^n, choose two disjoint open balls A_1, A_2 with A_i cap Omega nonempty and A_i \\ Omega nonempty. Let E_0 be the local extension (zero outside Omega). Since p > n, point evaluations are bounded on L1_p(Omega). Define T(f) = (ell_1(f) - ell_2(f)) chi_O, where O subset A_2 \\ Omega is a smooth bump away from Omega and ell_i are point evaluations at points in A_i cap Omega. Set E = E_0 + T; this is a bounded linear extension operator. Choose bump functions f_1, f_2 supported in A_1 cap Omega and A_2 cap Omega with ell_1(f_1) = ell_2(f_2) = a. For g = f_1 + f_2, grad(E(g)) cancels on O, while the denominator norm is unchanged. Compute Phi(A_1), Phi(A_2), and Phi(A_1 union A_2) for this operator, numerically or in a 2x2 matrix model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing gap is in the proof of Theorem 2.1. The sentence 'since the sets where grad(E(f_k)) do not vanish are disjoint' is the only step that converts linearity of E and disjointness of the supports of f_k in Omega into the lower bound norm(sum grad(E(f_k)))_Lq(A) >= (sum norm(grad(E(f_k)))_Lq(A_k)^q)^(1/q), which then yields sum Phi(A_k) <= Phi(union A_k). The supports of f_k are disjoint in Omega, so grad(E(f_k)) = 0 on A_j cap Omega for j != k; but the extension operator is not assumed to be local or support-preserving. For disjoint open sets A_k that cross the boundary of Omega, grad(E(f_k)) may be nonzero and may overlap with grad(E(f_j)) in A_j \\ Omega, and a bounded nonlocal perturbation of a local extension operator can make these gradients cancel. Hence countable, or even finite, superadditivity of Phi does not follow from the given argument. This is load-bearing because Theorem 2.5 invokes Phi' and integral_U Phi' = Phi(U) to pass from the density inequality to the L^alpha-norm estimate for K(x), and Theorem 2.8 is stated without proof but belongs to the same norm-estimate circle. The main density Theorem 2.4 uses only the direct inequality (2.1), so it may survive a repair; however, the paper's norm lower estimates are not justified as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies bounded linear extension operators E: L^1_p(Ω) → L^1_q(R^n) with q < p (decreasing integrability). It introduces a set function Φ(A), defined as the supremum of (‖Ef‖_{L^1_q(A)}/‖f‖_{L^1_p(A∩Ω)})^κ with 1/κ = 1/q − 1/p, and claims that Φ is a bounded, monotone, countably additive set function. On this basis it derives a capacitary inequality (Theorem 2.3), a generalized (p,q)-measure density condition (Theorem 2.4), an intrinsic-metric comparison (Theorem 2.7), and lower bounds on the operator norm in terms of the integral regularity function K and the metric distortion M (Theorems 2.5 and 2.8). The density condition is argued to be sharp on Hölder cusp domains.","tokens_in":8629,"tokens_out":10256,"duration_ms":96916,"significance":"If the main claims are correct, Theorem 2.4 would provide a sharp necessary condition for Sobolev extension with decreasing integrability, generalizing the classical measure density condition and matching known sufficient conditions for cusp domains. The set-function framework is natural, and the explicit constant dependence in the density inequality is a strength. However, the proof of countable additivity of Φ is currently incomplete, and the norm-estimate theorems 2.5 and 2.8 depend on that missing step; the sharpness discussion for the density condition is more robust because it relies only on the direct inequality (2.1).","major_comments":[{"comment":"In the proof of Theorem 2.1, the passage from the linearity of E and the disjointness of the supports of f_k in Ω to the lower bound for ‖E(g_N)‖_{L^q(∪A_k)} uses the sentence “since the sets where ∇E(f_k) do not vanish are disjoint.” For an arbitrary bounded linear extension operator E that is not assumed to be local or support-preserving, the gradients of E(f_k) need not have disjoint supports in R^n even when the supports of f_k are disjoint in Ω; the gradients may overlap outside Ω and can even cancel under nonlocal perturbations of a local extension operator. Therefore the displayed lower bound and the resulting superadditivity of Φ are not established. This gap is load-bearing because Theorem 2.5 and Theorem 2.8 rely on Φ being a countably additive measure.","section":"§2.1, Theorem 2.1"},{"comment":"The proof of Theorem 2.5 invokes a “Lebesgue type differentiability theorem” and the identity ∫_U Φ′(x) dx = Φ(U) for bounded open sets U. This step requires that the set function Φ, at least on the relevant balls, is absolutely continuous with respect to Lebesgue measure or is otherwise known to have a density. Countable additivity alone does not imply absolute continuity, and no argument is supplied to rule out a singular component of Φ. Since this differentiation step is essential for deriving the L^α estimate of K and the lower bound for ‖E‖ in Theorem 2.5, the proof is incomplete even if the countable additivity gap in Theorem 2.1 were repaired.","section":"§2.2, Theorem 2.5"},{"comment":"Theorem 2.8 is stated as one of the paper’s main results, giving the lower norm estimate ‖E‖ ≥ C_0 ‖M‖_{L^α(Ω)}^{1−n/q}, but the proof is omitted; the text merely says the inequality (2.4) “leads to” the estimate. Because this theorem belongs to the same circle of norm estimates whose other justifications are already in doubt, an explicit proof (or a clear statement that it is quoted from [23] rather than proved here) is required.","section":"§2.3, Theorem 2.8"},{"comment":"The operator norm in the introduction is defined for E : W^1_p(Ω) → W^1_q(R^n) using the full W^1 norms, while all subsequent theorems and the definition of Φ use homogeneous L^1_p and L^1_q seminorms. No argument is given that a bounded operator on W^1 spaces induces a bounded operator on the homogeneous spaces with comparable norm, or that the constants in Theorem 2.4 can be made independent of the diameter of the balls. The manuscript should state precisely which norm is meant in the hypotheses of Theorems 2.1–2.8 and how the homogeneous-space bound follows from the W^1 bound.","section":"§1 and §2"}],"minor_comments":[{"comment":"There are several typos, including “Let there exists a a continuous” in Theorems 2.1 and 2.3, and “the greatest lower bond” should be “greatest lower bound”.","section":"Throughout"},{"comment":"The construction of f_k assumes Φ(A_k) > 0; if some Φ(A_k) = 0, a separate limiting argument is needed. The normalization ‖f_k‖^p = Φ(A_k)(1−ε/2^k) cannot be performed when Φ(A_k) = 0.","section":"§2.1, Theorem 2.1"},{"comment":"The proof fixes x,y ∈ Ω and sets r = |x−y|, but the theorem states the inequality for all x ∈ Ω̄. A limiting argument handling boundary points is not supplied.","section":"§2.2, Theorem 2.4"},{"comment":"In the displayed inequality (2.5), the integral ∫_{B(x,R)} |∇f|^p dz should be over B(x,R) ∩ Ω, since f is a priori defined only on Ω. Also, Lemma 2.6 gives supp(f) ⊂ B(x,d_Ω(x,y)), and the distinction between the Euclidean ball and its intersection with Ω should be made explicit.","section":"§2.3, Theorem 2.7"},{"comment":"The sharpness discussion for the Hölder cusp derives the range 1 ≤ q < 2p/(α+1) without explicitly bounding Φ(B(0,r)); the argument would be clearer if it stated that Φ(B(0,r)) is bounded by a constant independent of r (for instance by ‖E‖^κ), which is what makes the exponent comparison valid.","section":"§2.2, sharpness example"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read of Ukhlov's paper. The headline result—Theorem 2.4, the generalized (p,q)-measure density condition—is a genuine extension of the p=q condition from [9] to qn. The sharpness check on the Hölder cusp matches known sufficient conditions, so the exponent in the density inequality looks right. If that were all the paper did, I'd be satisfied.\n\nThe trouble starts with Theorem 2.1, countable additivity of Φ. The proof says the sets where ∇E(f_k) don't vanish are disjoint because the f_k have disjoint supports in Ω. That's true for the f_k, but not for their extensions: E is not assumed local or support-preserving, and the gradients can overlap outside Ω. For open sets A_k that cross ∂Ω, nothing prevents ∇E(f_k) and ∇E(f_j) from overlapping, even canceling. So finite superadditivity doesn't follow. This is not a cosmetic gap: Theorems 2.5 and 2.8 use Φ's additivity to differentiate Φ and pass to L^α norm estimates for the density ratio K(x) and the metric distortion M(x). Theorem 2.8 is also stated without proof. The core density Theorem 2.4 only uses (2.1), so it may survive a repair, but the norm lower bounds are not justified as written.\n\nThere's also a more conceptual issue: Φ is built from the operator E itself, so the necessary condition in Theorem 2.4 cannot be checked from Ω alone. That's not fatal—it's still a constraint—but it weakens the geometric force of the result.\n\nI want to be fair: the paper is clear, the idea of using set functions tied to operator norms is a reasonable way to handle q