{"id":"8c81d616-ced8-41e4-a46b-599ad15a421f","arxiv_id":"2506.17028","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For higher-order Sobolev inequalities with the sharp constant and only an L^2 remainder, validity is governed by the manifold's scalar curvature and dimension, with a complete answer for k=2.","lead":"The paper determines when the sharp higher-order Sobolev inequality on a closed curved space remains true when the correction term is the smallest possible L^2 norm. It finds that, unlike the first-order case, the answer depends on the manifold's geometry, with a complete answer in the fourth-order case.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The validity half of The dichotomy rests on the uniform fast decay of the Green's function in Theorem 8, and the proof of that estimate is compressed at precisely the Neumann-series recurrence.","rationale":"The reader's weakest assumption correctly identifies the fast-decay Green's function estimate as the most load-bearing technical input: it is not a restatement of the Sobolev inequality being classified, and the validity proofs in Section 5 pass through it at multiple points. My stress-test agrees with that identification and makes it more specific: the risk is not merely that [32] is unpublished, but that the alpha-power bookkeeping in the Neumann series of Section 7.4 is asserted rather than fully demonstrated. The construction of G_alpha in (74), the recurrence (75), and the cancellation yielding (77) are the exact places where a missing power of alpha would propagate into Theorem 8 and then into the pointwise control of Theorem 6 and the Pohozaev estimates (39)-(45). I found no internal contradiction in the paper and no reason to believe the estimate is false; the concern is about the completeness and verifiability of the proof at a step that the central claim genuinely depends on. Because the paper gives a plausible but compressed proof, the right verdict remains CONDITIONAL: the authors should either expand the Neumann-series argument or provide the public version of [32]. My review does not change the reader's verdict.","tokens_in":32784,"tokens_out":36006,"duration_ms":372101,"concrete_test":"In the Euclidean model Delta_xi^k + alpha^{2k} on R^n, take k=2, n=5 and k=2, n=6, and compute explicitly H^1=-f and H^2(x)=int H^1(x,z) f(z,0) dz with f as in (72). Verify that H^2 satisfies the claimed bound in (75) uniformly in alpha>=1, including the alpha-power for i=E(n/2)+1. Then repeat for the next iteration and check that the residual H^{N+1} in (74)-(76) is cancelled by phi_alpha with the stated alpha exponent in Lemma 7.2. If an extra positive power of alpha survives in H^2 or in H^{N+1}, estimate (69) fails at the stated rate, and the O(theta_alpha), O(theta'_alpha) bounds in Section 5 would need to be recomputed; that would directly test the validity dichotomy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The validity direction of Theorems 4 and 5 is proved by contradiction: starting from a minimizing sequence u_alpha below the Euclidean threshold, Theorem 6 supplies the pointwise control (12). That control is proved via Theorem 7 and Proposition 6.4, whose Green's representation step (59)-(60) uses the uniform Green's function bound (69). The same estimate is then used in Section 5 to bound the Pohozaev error terms III_alpha, II_alpha, IV_alpha by O(theta_alpha)+O(theta'_alpha) with theta_alpha, theta'_alpha as in (39), and the contradictions in Cases 1-3 require these errors to be of strictly lower order than the leading positive terms (alpha mu_alpha)^{n-2k} and mu_alpha^2 (or mu_alpha^2 ln(1/(alpha mu_alpha))). If (69) held only with a slower alpha-decay, the gain (alpha_i mu_i)^tau in (60) would be lost, the error terms in (41) and (45) could absorb the dominant right-hand side, and the sign contradiction that proves validity would disappear. The paper does provide a proof of (69) in Section 7, but the load-bearing step is the recursive Neumann-series bound (75): the alpha-powers for i>E(n/2) and the cancellation leading to (77) are asserted via 'Iterating Lemma 7.1' and 'standard cancellation', with a reference to the unpublished note [32]. Since every validity conclusion depends on this alpha-power bookkeeping, this is the most load-bearing unverified step in the argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the sharp higher-order Sobolev inequality H^2_k(M^n) -> L^{2n/(n-2k)}(M^n) on closed Riemannian manifolds, with the remainder taken in L^2 instead of H^{k-1}_2. The central results are Theorem 1 and the more precise Theorems 4 and 5: for k>1 the validity of the optimal inequality is no longer independent of geometry, as it is for k=1. In particular, the inequality always holds when n=2k+1; it holds for n=2k+2 and n=2k+3 exactly when the scalar curvature R_g is nonpositive everywhere; for n>=2k+2 it holds when R_g<0 everywhere and fails when R_g>0 somewhere; flat tori satisfy it; and for k>2, n>=2k+4, Ricci-flat non-flat manifolds fail. Theorem 2 gives the complete k=2 classification. The proofs combine test-function estimates for the non-validity part with a pointwise blow-up analysis (Theorems 3, 6, 7) and a Pohozaev-type identity; a central technical ingredient is a uniform Green's function estimate for Delta^k_g + alpha^{2k} (Theorem 8).","tokens_in":32985,"tokens_out":9583,"duration_ms":93859,"significance":"If the results are correct, they are a significant contribution to geometric analysis. They show that the validity of the optimal Sobolev inequality with an L^2-remainder is a genuinely geometric question for k>1, in sharp contrast with the k=1 and with the H^{k-1}_2-remainder case. The complete classification for k=2 (Theorem 2) and the sharpness of the conditions in low dimensions are strong and potentially influential. The paper also introduces a pointwise control theorem for polyharmonic equations that may be useful beyond this problem. However, the proof is heavily dependent on several technical inputs whose full proofs are deferred to other works, including an unpublished note, and one central estimate appears to contain a sign error as printed.","major_comments":[{"comment":"Propositions 6.2 and 6.3 are stated without proof, with the phrase \"Following the same arguments as in [4,30,31]\". These propositions provide the uniform invertibility of the linearized operator L_{\\alpha,\\nu} and the existence and uniqueness of the correction \\phi_{\\alpha,\\nu}. They are directly used in the proof of Theorem 7 and then in the identification of u_\\alpha = V_{\\alpha,\\nu} + \\phi_\\alpha in Theorem 6. Since the present setting includes the term \\alpha^{2k} and a different kernel than any of the cited papers, the deferred proof is load-bearing: the authors should either include a full proof or point to precise statements (with equation numbers) in [4,30,31] that cover exactly this case.","section":"§6, Propositions 6.2 and 6.3"},{"comment":"The uniform Green's function estimate (69) is a crucial technical ingredient: it enters the pointwise bound (60) in Proposition 6.4 and the estimates of III_\\alpha, II_\\alpha, IV_\\alpha in Section 5. The proof in Section 7, Step 4, asserts the bounds (75) for i > E(n/2) by \"Iterating Lemma 7.1\" and the identity (77) by \"standard cancellation\", referring to the unpublished note [32]. The \\alpha-power bookkeeping in (75) and the cancellation in (77) are exactly what produces the fast decay in (69). Without a complete proof of these steps, the validity half of Theorems 4 and 5 is not fully established within the manuscript.","section":"§7, Step 4, equations (75) and (77)"},{"comment":"As printed, the pointwise control (5) and its rescaled version (33) contain (1+\\alpha d)^p in the numerator. This grows with \\alpha and contradicts the announced fast decay at infinity; it is also incompatible with the subsequent use in the proof of Lemma 5.1 and with the estimate IV_\\alpha = O(\\mu^{n-2k}/\\alpha^{2q}) in §5.1. The exponent must be negative. This sign error affects the central estimate of the paper and should be corrected before the proof can be considered rigorous.","section":"§1, Theorem 3, Eq. (5); §5.1, Eq. (33)"}],"minor_comments":[{"comment":"In the definition of \\varphi_\\alpha, the notation \\exp^{-1}_{x_\\alpha} should specify the metric (g or g_{x_\\alpha}) to avoid ambiguity.","section":"§5.4, Lemma 5.1"},{"comment":"The constants c_n^{(1)} and c_n^{(2)} are not explicitly defined; stating their positivity and independence of \\alpha would help the reader.","section":"§5, equations (42) and (45)"},{"comment":"The definition of \\theta_\\alpha for n<2k+4 is compact; a one-line derivation of the case n=2k+1, where \\theta_\\alpha = \\mu/\\alpha^3, would improve readability.","section":"§5.1, equation (39)"},{"comment":"Some \"straightforward computations\" (e.g., (46) and (47)) are used to justify the size of the error terms; providing a brief indication of the computation would increase verifiability.","section":"§6, equations (46)-(50)"}],"recommendation":"major_revision","confidential_remarks":"I recommend major revision rather than rejection. The main theorems appear plausible and the overall strategy is coherent, but the manuscript as it stands relies on several unproved or under-proved technical ingredients, including Propositions 6.2-6.3 and the Neumann-series construction for the Green's function in Section 7, the latter pointing to an unpublished note. The sign error in (5)/(33) is easily corrected but is nonetheless a load-bearing issue. The heavy use of the authors' own previous works is not circular, but the referee should insist that the final version be self-contained on these points."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper answers a genuine open question: for k≥2, the optimal Sobolev inequality with L2 remainder is not automatic, and validity depends on curvature and dimension. The main theorems are new — no prior work characterized (I_opt,k) for k>1 — and the dichotomy is sharp for k=2 and in small dimensions. The proof is a substantial analytic tour de force: test-function expansions for non-validity, a blow-up analysis for validity, and a pointwise control theorem for polyharmonic solutions with fast decay. The general framework looks coherent and the conclusions align with the known k=1 case. I found no fatal mathematical contradiction.\n\nThe main soft spot is exactly where the stress-test note points: Section 7.4. The uniform Green's function estimate (69) is load-bearing for the validity half, since the gain (α_i μ_i)^τ in the pointwise control depends on the α-decay. The proof of (69) relies on the Neumann-series recursion (75) and the cancellation (77), and the crucial α-power bookkeeping for i>E(n/2) is asserted via \"Iterating Lemma 7.1\" and \"standard cancellation,\" with a reference to the unpublished note [32]. The published literature cited for the same technique does not contain this exact construction. This is a gap in verifiability, not a demonstrated error, but it is a genuine gap: a referee cannot check the most load-bearing estimate without the unpublished note or an expanded proof.\n\nTwo smaller issues: Propositions 6.2 and 6.3 are deferred to earlier papers by the same circle. Those papers are published, so this is acceptable, but the statements should be precise and the appendix should say which arguments carry over verbatim. Also, the paper is long and dense, which will make reviewing labor-intensive.\n\nThe central argument holds up as far as I can trace it. The result is significant for geometric analysis and deserves a serious referee. My recommendation: send it to peer review with the requirement that the authors either make [32] public or give a self-contained proof of the Neumann-series estimates and the cancellation in Section 7.4. The \"conditional\" verdict from the reader is the right call.","headline":"A genuinely new dichotomy for high-order Sobolev inequalities with L2 remainder, with a real but localized soft spot in the Green's function proof.","tokens_in":33678,"tokens_out":4479,"would_cite":true,"duration_ms":43735,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35J35","35J60","35B44","35J08","58J05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Higher-order Sobolev inequalities with the optimal constant are governed by scalar curvature, except in a high-dimensional regime where scalar curvature no longer decides.","keywords":["optimal Sobolev inequality","higher-order Sobolev inequality","L2 remainder","scalar curvature","blow-up analysis","polyharmonic operator","Green's function","closed Riemannian manifold"],"falsifier":"Take the flat torus $\\mathbb{T}^n$, $n>2k$, and compute the Green's function for $\\Delta^k+\\alpha^{2k}$ in Fourier series; if the bound $|G_\\alpha(x,y)|\\le C\\,d(x,y)^{2k-n}/(1+\\alpha^q d(x,y)^q)$ fails for some $q$, then Theorem 8 is false and the validity proofs built on it collapse, while verifying it for all $q$ would remove the main technical risk.","tokens_in":32459,"feed_emoji":"📐","tokens_out":7127,"duration_ms":68689,"temperature":0.7,"pith_summary":"For the classical Sobolev inequality on a closed manifold, the optimal Euclidean constant can be attained with an $L^2$ remainder term regardless of geometry. This paper claims that for higher-order versions ($k\\ge2$) the situation changes: the optimal inequality $(I_{\\mathrm{opt},k})$ holds or fails depending on the manifold's dimension and curvature. The sharpest result is for $k=2$: the inequality holds exactly when $n=5$ or the scalar curvature is non-positive everywhere. For general $k$, negative scalar curvature everywhere guarantees validity, positive scalar curvature somewhere forbids it in high enough dimensions, and in high dimensions scalar curvature alone no longer decides the question. A reader should care because this pinpoints precisely where geometry enters an apparently constant-only problem.","feed_headline":"For k=2, optimal Sobolev inequality holds iff n=5 or R≤0","feed_subtitle":"Higher-order versions depend on scalar curvature, unlike the classical case k=1.","key_machinery":"The argument is carried by a sharp pointwise description of blowing-up solutions. The paper considers sequences $u_\\alpha$ of solutions to $\\Delta_g^k u_\\alpha+\\alpha^{2k}u_\\alpha=|u_\\alpha|^{2^\\star-2}u_\\alpha$ with energy just below the Euclidean threshold, and proves (Theorem 6) that each such sequence is a Euclidean bubble $U$ plus a small remainder controlled by\n$$(\\mu_\\$\\alpha$+d(x_\\$\\alpha$,x))^l\\,|\\nabla^l(u_\\$\\alpha$-V_{\\$\\alpha$,\\nu_\\$\\alpha$})| \\le C_{p,\\tau}\\,\\frac{(\\$\\alpha$(\\mu_\\$\\alpha$+d(x_\\$\\alpha$,x)))^\\tau}{(1+\\$\\alpha$ d(x_\\$\\alpha$,x))^p}\\,B_{x_\\$\\alpha$,\\mu_\\$\\alpha$}(x).$$\nThe fast decay in $\\alpha d$ comes from a Green's-function estimate (Theorem 8) for the operator $\\Delta_g^k+\\alpha^{2k}$. Feeding this control into a Pohozaev identity produces terms whose relative sizes decide validity; separate test-function expansions produce the non-validity obstructions.","core_discovery":"The paper's central discovery is a dichotomy: the inequality\n$$\\left(\\int_M |u|^{2^\\star}\\,dv_g\\right)^{2/2^\\star} \\le K(n,k)\\int_M (\\$Delta_g^{{k/2}}$u)^2\\,dv_g + B\\int_M $u^{2}$\\,dv_g$$\nwith the Euclidean optimal constant $K(n,k)$ and an $L^2$ remainder holds on a closed $n$-manifold precisely under geometric conditions. For $k=2$ and $n\\ge5$, it holds if and only if $n=5$ or $R_g\\le0$ everywhere. More generally, for $n=2k+1$ it always holds; for $n=2k+2$ or $2k+3$, and for $k=2$ with $n\\ge6$, it holds exactly when $R_g\\le0$; for $n\\ge2k+2$ it holds when $R_g<0$ everywhere and fails when $R_g>0$ somewhere. The paper also shows that for $k\\ge3$ and $n\\ge2k+4$, a Ricci-flat non-flat manifold fails while a flat torus holds, so scalar curvature alone cannot decide validity in high dimensions.","pith_inferences":["If the pointwise-control machinery extends to $H^l$ remainders for $1\\le l\\le k-2$, the same dichotomy should appear with shifted critical dimensions; this is an extension the authors flag but leave unproved.","The Ricci-flat failure suggests that on scalar-flat manifolds with non-vanishing Weyl tensor, validity should depend on the vanishing order of the Weyl tensor at the worst point, mirroring known second-order results; this is not established here.","A numerical or asymptotic check of the Fourier Green's function on the flat torus could test the sharpness of the decay rate in Theorem 8 and thereby the robustness of the validity proof."],"forward_implications":["For second-order inequalities ($k=2$) the validity question is now completely settled in every dimension $n\\ge5$.","Every closed manifold of dimension $2k+1$ satisfies the optimal inequality regardless of curvature.","Positive scalar curvature at a single point blocks the inequality whenever $n\\ge2k+2$ and $k>1$.","Negative scalar curvature everywhere permits the inequality for all $k>1$ and $n\\ge2k+2$.","In high dimensions ($k\\ge3$, $n\\ge2k+4$) curvature data beyond scalar curvature, such as the Weyl tensor, controls validity."],"supporting_citations":[{"why":"Supplies the bubble-decomposition and pointwise-control technique for higher-order Sobolev inequalities that Theorem 6 refines.","marker":"[4]"},{"why":"Constructs Green's functions for polyharmonic operators with diverging coefficients, providing the basis for Theorem 8.","marker":"[5]"},{"why":"Proves the best constant in the Sobolev inequality can be approached, yielding the minimizing sequence used in the blow-up argument.","marker":"[25]"},{"why":"Provides the Pohozaev-Pucci-Serrin computations and conformal expansions used to estimate the blow-up terms.","marker":"[33]"},{"why":"Identifies the explicit extremal $U$ and the value of the Euclidean constant $K(n,k)$ that serve as the bubble profile.","marker":"[36]"},{"why":"Classifies positive solutions of the critical polyharmonic equation, ensuring the limit profile is the standard bubble.","marker":"[37]"},{"why":"Supplies the formula for the tensor $T_g$ appearing in the test-function expansions that produce the non-validity obstructions.","marker":"[26]"}],"fun_headline_variants":["k=2 Sobolev: only dimension and scalar curvature decide","k>1 Sobolev: geometry's call, not just dimension","For k=2 Sobolev, the gate is n=5 or R≤0"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the sharp pointwise estimate (12) for the difference between a solution and its approximating bubble, which relies on the fast decay (69) of the Green's function of $\\Delta_g^k+\\alpha^{2k}$; if that decay is slower than claimed, the contradiction arguments proving validity close too slowly or not at all.","fun_headline_variants_meta":{"raw":{"variants":["k=2 Sobolev: only dimension and scalar curvature decide","k>1 Sobolev: geometry's call, not just dimension","For k=2 Sobolev, the gate is n=5 or R≤0"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001998,"raw_usage":{"total_tokens":7775,"prompt_tokens":901,"completion_tokens":6874,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":6809}},"tokens_in":517,"tokens_out":6874,"duration_ms":52997,"temperature":1.0,"reasoning_tokens":6809,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:14:02.437267+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the flat torus $\\mathbb{T}^n$, $n>2k$, and compute the Green's function for $\\Delta^k+\\alpha^{2k}$ in Fourier series; if the bound $|G_\\alpha(x,y)|\\le C\\,d(x,y)^{2k-n}/(1+\\alpha^q d(x,y)^q)$ fails for some $q$, then Theorem 8 is false and the validity proofs built on it collapse, while verifying it for all $q$ would remove the main technical risk.","supporting_citations":[{"cited_title":"London Math","cited_arxiv_id":null,"evidence_quote":"Supplies the bubble-decomposition and pointwise-control technique for higher-order Sobolev inequalities that Theorem 6 refines."},{"cited_title":"Differential Equations 419 (2025), 370–417","cited_arxiv_id":null,"evidence_quote":"Constructs Green's functions for polyharmonic operators with diverging coefficients, providing the basis for Theorem 8."},{"cited_title":"Differential Equations 261 (2016), no","cited_arxiv_id":null,"evidence_quote":"Proves the best constant in the Sobolev inequality can be approached, yielding the minimizing sequence used in the blow-up argument."},{"cited_title":"Swanson, The best Sobolev constant , Appl","cited_arxiv_id":null,"evidence_quote":"Identifies the explicit extremal $U$ and the value of the Euclidean constant $K(n,k)$ that serve as the bubble profile."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Classifies positive solutions of the critical polyharmonic equation, ensuring the limit profile is the standard bubble."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the formula for the tensor $T_g$ appearing in the test-function expansions that produce the non-validity obstructions."}],"review_version":1}