{"id":"7da99d1d-b077-49e0-ba98-02f98e591b67","arxiv_id":"2607.17588","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Sobolev trace deficit controls the max{2,p}-th power of the gradient distance to the trace-bubble manifold for all 1<p<n.","lead":"This paper proves that any function on the upper half-space that nearly attains the best possible boundary-versus-energy ratio must be close, in gradient norm, to one of a known family of ideal functions, with the gap controlled by the max{2,p}-th power of the distance. It is the sharp half-space analogue of the whole-space Sobolev stability theorem and covers the full range 1<p<n.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.19's no-zero-eigenvalue conclusion is not established: two eigenfunctions below Λ* fix δ2 as the second eigenvalue but allow a later δ3=0; Proposition 3.15's L^{-1} step therefore has a gap.","rationale":"Agree with the reader's weakest_assumption. The rest of the proof is detailed and follows [19] closely; the remainder estimates and compactness arguments are not the weak point. The only genuinely unsupported inference is the exclusion of 0 from the discrete spectrum of L in Lemma 3.19. As written, the two explicit eigenfunctions determine the first two eigenvalues below Λ* and exclude a second negative eigenvalue, but they say nothing about the possibility of a later zero eigenvalue with a two-nodal eigenfunction. Because Proposition 3.15 needs L_+^{-1} bounded, this gap is load-bearing for the equality classification and hence for the main theorem. No issue of circularity, fraud, or post-hoc fitting; it is an internal proof gap. The reader's conditional verdict is appropriate, so no adjustment.","tokens_in":51294,"tokens_out":25309,"duration_ms":209553,"concrete_test":"Run a high-accuracy spectral/shooting computation of the singular Sturm-Liouville problem (3.51) for representative parameters (e.g., n=3, p=3/2 and n=4, p=3), imposing the natural endpoint conditions of Claim 3.18, and list all eigenvalues below Λ* in (-∞, Λ*). If 0 is among them, Proposition 3.15's use of L^{-1} is invalid. If no zero is found in these cases, the remaining issue is the missing proof of absence, not a numerical counterexample; an analytic resolution would be to show the zero-energy solution of Lf=0 has too many zeros or is not in V.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Lemma 3.19, Step 4, the proof exhibits a positive eigenfunction a with δ1=-np/(p-1)<0 and a one-zero eigenfunction g with δ2>0, both below Λ*. By Lemma 3.17 these indeed force δ2 to be the second discrete eigenvalue below Λ*, and they rule out another negative eigenvalue. They do not, however, rule out a third discrete eigenvalue δ3=0: an eigenfunction for δ3 would have two interior zeros, which is fully consistent with g having one zero. The next sentence 'Consequently ... no zero eigenvalue' therefore does not follow. This is not a cosmetic point: Proposition 3.15 explicitly uses invertibility of L ('L^{-1}1=-b_0', bounded L_+^{-1}) to convert the Cauchy-Schwarz equality into the sharp one-dimensional inequality; if 0∈σ(L), L_+^{-1} is unbounded and the argument collapses. Since Proposition 3.15 feeds Theorem 3.10, the sectorial classification, Theorem 2.2, and hence the spectral gap in Theorem 1.1, this unsupported exclusion is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Theorem 1.1: for every n≥3 and 1<p<n, the Sobolev trace deficit δ_T(u) controls the max{2,p}-th power of the gradient distance d_T(u,M_T) to the trace-bubble manifold, with the exponent sharp. The proof combines a spectral nondegeneracy theorem for the linearized weighted Steklov problem at a trace bubble with a nonlinear Taylor-remainder analysis in the spirit of Figalli–Zhang. The spectral part is reduced by a conformal transformation to a weighted problem on a ball, decomposed into spherical-harmonic sectors; the axisymmetric sector is handled by a Pohozaev identity and a singular Sturm–Liouville analysis.","tokens_in":51560,"tokens_out":14999,"duration_ms":118149,"significance":"If the result holds, it closes the full-range sharp gradient stability question for the classical Sobolev trace inequality, extending the whole-space theorem of Figalli–Zhang and the p=2 theorem of Ho. The paper is technically substantial: it proves a weighted compact trace embedding, identifies the first two eigenspaces of the linearized Steklov problem in a setting without full radial symmetry, and treats the ranges p<2 and p≥2 with exact remainder estimates. The sharpness examples are explicit. The potential gap in Lemma 3.19 flagged by the stress test does not land: because g has exactly one interior zero, Lemma 3.17 identifies it as the second eigenfunction below Λ*, so δ2>0 is the second eigenvalue; any later eigenvalue is larger than δ2, ruling out a zero eigenvalue. Thus the invertibility of L on the relevant subspace is justified.","major_comments":[{"comment":"The concern that a zero eigenvalue could occur as a later discrete eigenvalue below Λ* is unfounded. The function g is shown to have exactly one interior zero and δ2<Λ*. By Lemma 3.17(2), the i-th eigenfunction below Λ* has exactly i−1 interior zeros, so g must be the second eigenfunction. Therefore δ2 is the second eigenvalue, and any subsequent eigenvalue satisfies δ3≥δ2>0. A zero eigenvalue would have to lie between δ1 and δ2, which is impossible. Consequently the use of invertibility of L and boundedness of L_+^{-1} in Proposition 3.15 is justified.","section":"Lemma 3.19, Step 4"}],"minor_comments":[{"comment":"The notation for the two critical exponents is inconsistent: the text uses p∗ and p ∗ in a way that is visually confusable. Since the interior Sobolev exponent and the trace exponent play very different roles, please use unambiguous symbols such as p^* and p_* throughout.","section":"Section 2.1"},{"comment":"The symbol Ẋ^{1,p}(R^n_+) appears in the abstract and Theorem 1.1; this appears to be a typo for the homogeneous Sobolev space \\dot W^{1,p}(R^n_+). Please correct it.","section":"Abstract and Section 1"},{"comment":"The optimality example with the diagonal matrix A_i = diag(1,...,1,1+1/i) should specify more explicitly how A_i acts on the half-space variables, and the choice of the sequence ε_i with v(x_i)≪ε_i≪1 should be spelled out. The argument is clear, but a few more details would improve readability.","section":"Remark 1.2"},{"comment":"The symbol Q is used both for the quadratic form Q[f] and for the coefficient function Q(t)=-pN(t)ρ(t) in equations (3.47)–(3.51). This is confusing; please use a different letter, e.g. q(t), for the potential coefficient.","section":"Section 3.2.5"},{"comment":"The verification that the coefficients C_1 and C_2 vanish after substituting m=m_*, c=c_*, δ=δ_2 is asserted rather than displayed. A short factorization of C_1 and C_2 (or a remark that the computation is a direct symbolic simplification) would make the proof easier to check.","section":"Lemma 3.19, Step 2"},{"comment":"The assertion that choosing v∈M_T with ||∇v||_{L^p}≤2||∇u||_{L^p} gives a universal upper bound on d_T(u,M_T) is slightly misleading; the needed bound d_T≤C follows more directly from taking bubbles with arbitrarily small amplitude. Please adjust the wording.","section":"Section 5, Claim 5.4"}],"recommendation":"minor_revision","confidential_remarks":"The paper is technically dense but sound. The previously flagged concern about Lemma 3.19 does not survive scrutiny once the ordering of eigenvalues below Λ* is taken into account. I recommend publication after minor editorial revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves a genuinely new sharp gradient stability estimate for the Sobolev trace inequality for all 1<p<n, and the framework is a real adaptation of Figalli–Zhang to the trace setting. The sector decomposition and the singular Sturm–Liouville analysis are nontrivial, and the Pohozaev identity and explicit eigenfunctions are nice pieces of work. If the spectral gap is fixed, this is a significant contribution.\n\nBut the reader's and stress-test's complaint is correct, and it is the central soft spot. In Lemma 3.19, Step 4, the proof exhibits two eigenfunctions: a positive one with δ1<0 and g with one interior zero and δ2>0, both below Λ*. By Lemma 3.17 that indeed makes δ2 the second eigenvalue and rules out another negative eigenvalue below Λ*. However, Lemma 3.17 also says the third eigenvalue, if present, has an eigenfunction with two interior zeros. g having one zero does not contradict that. Therefore δ3=0 is not excluded, and the sentence 'Consequently ... no zero eigenvalue' does not follow.\n\nThis is not cosmetic. Proposition 3.15 explicitly uses L^{-1}, including boundedness of L_+^{-1}, to convert the Cauchy–Schwarz equality into the sharp one-dimensional inequality. If 0 is an eigenvalue, that inverse is unbounded and the argument collapses. Proposition 3.15 feeds Theorem 3.10, the sectorial classification, Theorem 2.2, and hence Corollary 2.6 and the final Theorem 1.1. So the gap is load-bearing.\n\nThe rest of the paper is careful and mostly convincing. The compactness lemmas, the nonlinear remainder estimates, and the modulation argument are handled in detail. The citation pattern looks honest and the novelty claim is credible. The concern is not with the overall strategy but with one unsupported inference in a key lemma.\n\nCould the gap be repaired? Possibly, with a direct nodal argument showing 0 is not an eigenvalue, or by proving the needed inequality without inverting L. But that repair is not in the manuscript.\n\nThis paper deserves a serious referee and a careful revision, not a desk rejection. I would send it to peer review, and I would tell the authors to fix Lemma 3.19 or replace the L^{-1} step before it is accepted.","headline":"Genuinely new sharp trace stability result, but the proof has a load-bearing gap: Lemma 3.19 does not rule out a zero eigenvalue, and Proposition 3.15 relies on invertibility.","tokens_in":52051,"tokens_out":2784,"would_cite":false,"duration_ms":26022,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A23","35J92","46E35","58J50"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that near-equality in the Sobolev trace inequality forces closeness to trace bubbles in gradient norm, with the optimal power max{2,p}.","keywords":["Sobolev trace inequality","quantitative stability","gradient distance","trace bubbles","spectral nondegeneracy","weighted boundary eigenvalue problem","upper half-space","concentration-compactness"],"falsifier":"For fixed n≥3 and p in (1,n), compute the discrete spectrum of the one-dimensional operator with the explicit weights appearing in the paper, below the threshold p + (n-1)^2/(4(p-1)). If any eigenvalue equals 0 besides the eigenfunctions already identified, or if the number of negative eigenvalues is not exactly one, then the classification of the second eigenspace, and hence Theorem 1.1, is not established.","tokens_in":51178,"feed_emoji":"📐","tokens_out":7148,"duration_ms":63929,"temperature":0.7,"pith_summary":"The paper proves a quantitative stability theorem for the sharp Sobolev trace inequality on the upper half-space: when the trace deficit is small, the function must be near the manifold of trace bubbles in gradient norm, and the deficit controls the gradient distance raised to the power max{2,p}. This settles sharp gradient stability for the trace inequality for every n≥3 and 1<p<n, matching the exponent already known for the whole-space Sobolev inequality. The exponent is optimal, as two explicit perturbation families show that no smaller power works. The load-bearing step is a spectral nondegeneracy theorem for the linearized weighted boundary eigenvalue problem around a trace bubble.","feed_headline":"Trace Sobolev stability gains sharp gradient power max(2,p)","feed_subtitle":"Near-equality forces closeness to trace bubbles at the optimal deficit power for all 1<p<n.","key_machinery":"The central object is the linearized weighted boundary eigenvalue problem ∫ A_v ∇φ·∇ψ dx = μ ∫ |v|^{p*-2} φψ dy, where A_v = |∇v|^{p-2}(I + (p-2) ∇v/|∇v| ⊗ ∇v/|∇v|) is the linearization matrix of the p-Laplacian. The spectral nondegeneracy assertion — the first two eigenspaces are exactly the tangent modes of the bubble manifold — supplies a positive spectral gap on the tangent-orthogonal complement. The proof transforms the half-space to a ball, decomposes the eigenfunction into angular sectors, reduces the axisymmetric sector, via an integration-by-parts identity with the dilation derivative, to a sharp one-dimensional inequality, and then analyzes the spectrum of that one-dimensional oper","core_discovery":"The central assertion is Theorem 1.1: for n≥3 and 1<p<n there exists c(n,p)>0 such that every u with nonzero trace satisfies δ_T(u) ≥ c d_T(u,M_T)^{max{2,p}}. Here δ_T is the deficit of the sharp trace inequality and d_T is the normalized gradient distance to the manifold of trace bubbles. The paper proves the exponent is optimal and identifies the mechanism: the second variation around every positive trace bubble is coercive on the tangent-orthogonal complement, because the first two eigenspaces of the linearized weighted boundary eigenvalue problem are exactly the amplitude, dilation, and tangential translation modes. With this spectral gap in hand, the proof uses exact remainder estimates","pith_inferences":["Editorial inference: the sector decomposition plus one-dimensional reduction may transfer to other non-radially symmetric weighted boundary eigenvalue problems where full separation of variables fails.","Editorial inference: a direct numerical spectrum computation of the one-dimensional operator would settle the residual zero-eigenvalue question; the weights are explicit, so this is feasible for fixed n and p.","Editorial inference: if the sharp gradient exponent transfers through the known reduction for fractional trace inequalities, the same max{2,p} power would be expected there; that is a testable extension, not a claim of this paper."],"forward_implications":["For every p in (1,n), near-equality sequences in the Sobolev trace inequality converge to the trace-bubble manifold in the W^{1,p} gradient norm.","The gradient distance is controlled with the sharp power: squared for p≤2 and p-th power for p≥2; no smaller exponent can work.","The spectral nondegeneracy theorem gives quantitative coercivity of the second variation around every positive trace bubble, a standard ingredient for stability of critical points and bubble decompositions.","The trace-setting exponent matches the whole-space gradient stability exponent, so the two sharp Sobolev stability results are consistent in form."],"fun_headline_variants":["Trace stability hits sharp power max(2,p)","Spectral gap yields optimal trace Sobolev stability","Trace bubbles dominate stability at sharp exponent","Optimal deficit power for Sobolev trace inequality","Stability exponent max(2,p) proven sharp for traces"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The stability estimate rests on the claim that a certain one-dimensional operator has exactly one negative eigenvalue and no zero eigenvalue; the paper exhibits two eigenfunctions with opposite-sign eigenvalues but does not explicitly rule out a zero eigenvalue appearing later among the discrete eigenvalues before the inverse of the operator is used.","fun_headline_variants_meta":{"raw":{"variants":["Trace stability hits sharp power max(2,p)","Spectral gap yields optimal trace Sobolev stability","Trace bubbles dominate stability at sharp exponent","Optimal deficit power for Sobolev trace inequality","Stability exponent max(2,p) proven sharp for traces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":9e-05,"raw_usage":{"total_tokens":753,"prompt_tokens":610,"completion_tokens":143,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":354,"completion_tokens_details":{"reasoning_tokens":82}},"tokens_in":354,"tokens_out":143,"duration_ms":2166,"temperature":1.0,"reasoning_tokens":82,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T17:33:51.672900+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For fixed n≥3 and p in (1,n), compute the discrete spectrum of the one-dimensional operator with the explicit weights appearing in the paper, below the threshold p + (n-1)^2/(4(p-1)). If any eigenvalue equals 0 besides the eigenfunctions already identified, or if the number of negative eigenvalues is not exactly one, then the classification of the second eigenspace, and hence Theorem 1.1, is not established.","supporting_citations":[],"review_version":1}