{"id":"52f5855a-aa49-4bf1-a996-4621594b6ef0","arxiv_id":"1908.07053","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every real analytic surface of revolution in R^3, an l4 decoupling inequality holds with a partition into essentially flat boxes of possibly different sizes.","lead":"This paper proves a new decoupling inequality for real analytic surfaces of revolution in three dimensions, including the torus and a class of perturbed cones. It shows that a broad conjecture in Fourier decoupling theory holds for an entire family of curved surfaces.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader identified the rescaling/curvature uniformity as the weakest assumption, and I agree that this is the most delicate point in the argument. However, after checking the geometry of the rescaled surfaces in both the quasi-torus and perturbed-cone cases, the required nondegeneracy holds uniformly, with constants that depend only on the fixed analytic function γ. The assertion 'S_ref has both principal curvatures ∼1' is imprecise for large n (one principal curvature can be exponentially small in n), but Theorem 1 only needs a positive lower bound on Gaussian curvature, which is present. The omitted proof of Lemma 6's derivative estimates is a real gap in exposition, but the estimates themselves are true and can be supplied by routine repeated differentiation; they do not change the validity of the central claim. No circularity, internal inconsistency, or counterexample was found. Therefore the reader's ACCEPT verdict stands without modification.","tokens_in":8728,"tokens_out":46798,"duration_ms":435952,"concrete_test":"Independently derive the leading-order expansion of ψ_k for a general n in the perturbed-cone case (Section 4) and compute det Hess(ψ_k) explicitly; verify it is bounded below by a constant c(n) > 0 uniformly for η1, η2 in the stated domains and for all k. Also differentiate the implicit equation defining φ in Lemma 6 up to order 3 to confirm the asserted bounds |D1^p D2^q φ| ≲ min{s_k^{n-p-q}, 1} for p+q ≤ 3; if these bounds fail, the C^3 uniformity needed for Theorem 1 would break.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is the l4 decoupling for all real analytic surfaces of revolution, and the proof's most delicate step is the application of Theorem 1 to the anisotropically rescaled caps θ_{k,new}; this requires the rescaled surface to have Gaussian curvature bounded away from zero uniformly in k and a C^3 norm bounded independently of k. I checked this condition in both new cases. For the quasi-torus (Section 3), the rescaled surface tends to η3 = c(η1^2+2η2)^n with η2 ∈ [1/2,1]; the Hessian determinant is positive and bounded below by a constant depending only on n, so the curvature is uniformly nonzero. For the perturbed cone (Section 4), the computation Hess(ψ_k) = s_k^{2-n} Hess(ψ) is correct, and the expansion of ψ shows Hess(ψ) ≍ s_k^{n-2}, so the curvature of θ_{k,new} stays comparable to 1 uniformly in k; the C^3 bounds in Lemma 6 follow from analyticity of γ and the smallness of s_k. The only genuine soft spot is that Lemma 6's derivative estimate is asserted as 'immediate' rather than proved; this is an omitted detail, not a demonstrated flaw. The summation over O(log 1/δ) dyadic pieces is handled by the standard δ^{-ε} loss. Thus I find no load-bearing concern that would undermine the theorem.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes Conjecture 4 for all real analytic surfaces of revolution in R^3: for every such surface Sγ and every δ>0, it constructs a partition of the δ-neighborhood into essentially flat boxes (of possibly different dimensions) such that the l^4 decoupling inequality holds with constant δ^{-ε}|Pδ(S)|^{1/4}. The proof splits the surface according to the zeros of γ'γ''. Away from these zeros, the classical nonzero-curvature decoupling (Theorem 1) applies; near a zero, a case analysis is made. If γ' vanishes to finite order, the surface is a 'quasi-torus' and the proof uses a cylinder decoupling followed by an anisotropic rescaling and another application of Theorem 1; if γ'' vanishes while γ' does not, the surface is a 'perturbed cone' and the proof uses a cone decoupling followed by a similar rescaling. The two new cases include the torus and the perturbed cone.","tokens_in":8874,"tokens_out":38155,"duration_ms":319676,"significance":"If the proof is correct, this is a substantial advance in decoupling theory for surfaces with vanishing curvature. It confirms the flexible l^4 decoupling conjecture for a natural class of surfaces and introduces a two-scale partition that may be instructive for the general case. The argument is honest: the main ingredients, Theorems 1 and 3, are prior published results used as black boxes, and no free parameters or normalizations are introduced to force the conclusion. The novel rescaling steps are the heart of the paper and are, on inspection, sound. The construction of partitions with boxes of different scales is an additional conceptual contribution.","major_comments":[{"comment":"The assertion that S_ref has both principal curvatures ~1 is load-bearing, because Theorem 1 is applied to θ_{k,new} only after this condition is verified, but the computation is not shown. For the model η3=(2n)^{-1}(η1^2+2η2)^n, the Hessian determinant is positive and bounded below on the relevant domain η2∈[1/2,1], |η1|≲1, with a lower bound depending only on n; this is the key point. Please include this computation (or a precise statement and proof) so that the reader can verify the uniform nondegeneracy of the curvature of θ_{k,new}.","section":"Section 3, after Eq. (10) and the definition of S_ref"},{"comment":"The proof of Lemma 6 is only a sketch: the derivative bound on φ is said to be 'quite immediate' and the consequent bound on ψ_k 'immediate'. Since this lemma is the sole verification that the rescaled surfaces θ_{k,new} have C^3 norms bounded independently of k, a hypothesis of Theorem 1, the proof should be expanded. In particular, one should demonstrate how repeated differentiation of expressions of the form (sqrt(...)-1)^m yields terms of size at most (s_k)^{m-p-q} (or O(1) after division by s_k^n), taking into account the implicit dependence of φ on ψ through the equation ψ = ξ'_1^2/(4ξ'_2) + φ.","section":"Section 4, Lemma 6"}],"minor_comments":[{"comment":"The leading coefficient in the expansion of η3 is 1/2^n, not 1/(2n) as written; this typo does not affect the argument since any positive constant would yield the same curvature conclusion.","section":"Section 3, equation after (9)"},{"comment":"The phrase 'due to symmetry' is imprecise: the surface need not be symmetric under r → 2-r. What is meant is that the analysis for r<1 is identical with |r-1| in place of r-1, so it suffices to present the right half; please clarify.","section":"Section 3, 'Due to symmetry'"},{"comment":"The change of variables omits mention of the Jacobian determinant of L_k; the determinant cancels in the decoupling inequality, but this should be stated for completeness.","section":"Section 4, change of variables after (11)"},{"comment":"The abstract contains a typo: 'surf aces' should read 'surfaces'.","section":"Section 1, Abstract"},{"comment":"The final assembly of the inequalities for the individual S_k into the global inequality for Sγ is not written out; since the number of dyadic scales is O(log 1/δ), the resulting logarithmic loss is absorbed into δ^{-ε}. A sentence making this explicit would help the reader.","section":"Sections 3 and 4, final assembly"},{"comment":"The notation 'Hess(ψ_k) = (s_k)^{2-n} Hess(ψ)' is potentially confusing: Hess usually denotes the matrix, while the displayed formula concerns the determinant. Please write 'det Hess' or 'the Hessian determinant'.","section":"Section 4, Hessian computation"}],"recommendation":"minor_revision","confidential_remarks":"This is a natural and well-executed successor to the Bourgain–Demeter decoupling theorems. The main concerns are purely presentational: two key verifications (the curvature of S_ref and the derivative bounds of Lemma 6) are asserted rather than proved. Both are true and easily expanded, so I recommend minor revision rather than major revision. No concerns about novelty, attribution, or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe one thing to know: this paper proves Conjecture 4 for all real analytic surfaces of revolution, and the proof is essentially correct. The torus and perturbed cone are genuinely new examples, and the machinery is the Bourgain–Demeter decoupling toolbox with an important twist: the δ-neighborhood is partitioned into essentially flat boxes of different scales, not uniform ones. That twist is the real contribution, and it works.\n\nWhat it does well: the case analysis based on principal curvatures is clean. The two new cases—quasi-torus (angular curvature vanishes) and perturbed cone (radial curvature vanishes)—are handled by parabolic rescaling that sends each small curved cap to a surface of curvature ~1, where the classical nonzero-curvature decoupling applies. I checked the two delicate rescaling steps. For the quasi-torus, the rescaled surface tends to η3 = c(η1^2 + 2η2)^n, whose Hessian determinant is positive and bounded below uniformly in n. For the perturbed cone, the computation Hess(ψ_k) = s_k^{2−n} Hess(ψ) is correct, and the expansion of ψ shows the curvature of the rescaled cap stays comparable to 1. The derivative estimates in Lemma 6 are plausible: they follow from analyticity of γ and the scale conditions, though the proof says “immediate” where a few lines would have helped. No fitted parameters, no circularity: Theorems 1 and 3 are independent, published results used as black boxes, and the self-citation is appropriate.\n\nSoft spots, in proportion: the paper is compressed. Lemma 6's derivative estimate is asserted rather than proved; a referee will want the routine but explicit computation. The smallness conditions on the intervals Δ_i are never made fully explicit—there is a hand-wave that “various restrictions will become apparent,” which is fine for a specialist but mildly annoying. The result is limited to surfaces of revolution, so the full Conjecture 4 remains open; the paper says this clearly. The final remarks about possible l2 upgrades and the open Conjecture 7 are honest and not a flaw.\n\nWho it is for: decoupling theorists and anyone working on restriction/decoupling for surfaces with degenerate curvature. A serious referee should engage with it, mainly to verify the two rescalings and the derivative bounds. I would accept it for peer review and expect acceptance after minor revision.\n\nBottom line: send it to the journal. The main theorem is a real step forward, and the argument holds up.","headline":"Credible proof of the l4 decoupling conjecture for all real analytic surfaces of revolution, with the variable-scale partition idea as the real new content; the few terse spots are minor, not fatal.","tokens_in":9487,"tokens_out":1974,"would_cite":true,"duration_ms":21029,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that every real analytic surface of revolution in $\\mathbb{R}^3$ satisfies the $\\ell^4$ decoupling conjecture: its $\\delta$-neighborhood admits an essentially flat partition, of possibly mixed box sizes, giving an $L^4$…","keywords":["decoupling inequalities","surfaces of revolution","real analytic surfaces","l4 decoupling","Fourier restriction","torus","perturbed cone","Gaussian curvature"],"falsifier":"For a profile such as $\\gamma(r)=r+(r-1)^3$, compute the determinant of the Hessian of the rescaled profile $\\psi_k$ on $|\\eta_1|,|\\eta_2|\\le 1$; the proof requires it to be comparable to $1$ uniformly in $k$. A real analytic $\\gamma$ with $\\gamma''(1)=0$, $\\gamma'''(1)\\ne 0$ for which this determinant tends to $0$, or an explicit function $f$ with Fourier support in $N_\\delta(S_\\gamma)$ violating the stated $\\ell^4$ inequality, would refute the main claim.","tokens_in":8433,"feed_emoji":"🍩","tokens_out":11034,"duration_ms":105529,"temperature":0.7,"pith_summary":"The paper establishes that Conjecture 4, the $\\ell^4$ decoupling conjecture for real analytic surfaces, holds for every surface of revolution in $\\mathbb{R}^3$. For each $\\delta$, the $\\delta$-neighborhood of such a surface can be partitioned into essentially flat boxes, possibly of different dimensions, so that the $L^4$ norm of any function whose Fourier transform is supported in that neighborhood is controlled by the $\\ell^4$ sum of the norms of the Fourier pieces, up to a factor $\\delta^{-\\epsilon}$. This covers new examples for which the classical nonzero-curvature decoupling theorem does not apply: the torus, whose angular curvature vanishes along a circle, and the perturbed cone, whose radial curvature vanishes. The proof splits the surface into a cone region, quasi-torus regions, and perturbed-cone regions, and treats each with a rescaling that reduces it to the known nonzero-curvature case.","feed_headline":"L4 decoupling proven for every real analytic surface of revolution","feed_subtitle":"Torus and perturbed cones included: boxes of different scales decouple with only an epsilon loss.","key_machinery":"The argument is carried by the Gaussian-curvature formula for surfaces of revolution, $K(r)=\\gamma'(r)\\gamma''(r)/(r(1+\\gamma'(r)^2)^2)$, together with a two-stage decouple-then-rescale mechanism. The formula locates the finitely many radii where classical nonzero-curvature decoupling fails. Around each such radius the paper decomposes the surface into dyadic annuli $U_k$, applies cone or cylindrical decoupling at a coarse scale, and then applies a linear rescaling $L_k$ that stretches each curved cap so that its equation approaches a reference surface with both principal curvatures comparable to $1$ and uniformly controlled third derivatives. The known nonzero-curvature decoupling is then applied on the rescaled surface, and the boxes are pulled back to the original coordinates. The quantitative checks that keep the rescaling uniform are Lemma 6, bounding the derivatives of the rescaled profile, and the Hessian identity $\\mathrm{Hess}(\\psi_k)=(s_k)^{2-n}\\,\\mathrm{Hess}(\\psi)$, which keeps the rescaled Gaussian curvature bounded away from zero.","core_discovery":"The central claim is Theorem 5: for every real analytic profile $\\gamma$ on $[1/2,2]$, the surface $S_\\gamma=\\{(\\xi_1,\\xi_2,\\gamma(\\sqrt{\\xi_1^2+\\xi_2^2}))\\}$ satisfies the $\\ell^4$ decoupling inequality with a partition into essentially flat boxes of possibly different sizes, with an $\\epsilon$ loss $\\delta^{-\\epsilon}$ and a factor $|P_\\delta(S)|^{1/4}$. Curvature degenerates exactly where $\\gamma'$ or $\\gamma''$ vanishes, and real analyticity ensures only finitely many such radii. Away from these, the known nonzero-curvature decoupling gives the needed inequality. Near each degenerate radius one of three models occurs: a genuine cone, a quasi-torus (angular curvature zero), or a perturbed cone (radial curvature zero). For the torus-like case the argument first decouples via the cylinder theorem on a small annulus and then rescales each cap by a linear map $L_k$ so that the cap becomes a reference surface with both principal curvatures comparable to $1$; for the cone-like case the same is achieved with the cone decoupling followed by a rotation and rescaling, with a Hessian computation showing the rescaled Gaussian curvature is comparable to $1$. The resulting partition mixes boxes of several scales, and the proof is uniform in the profile's derivatives.","pith_inferences":["Because real analyticity enters only to guarantee finitely many zeros of $\\gamma'\\gamma''$, a testable extension would replace analyticity by the requirement that the curvature-vanishing set is finite, with each degenerate model admitting a rescaling of the same type.","The two-stage decouple-then-rescale pattern suggests a general recipe for $\\ell^4$ decoupling on other surfaces: embed each scale into a known model (cone or cylinder), then normalize curvature by a linear map. Applying this to surfaces with degenerate curves rather than points would be a natural next step.","The uniform control after rescaling suggests that explicit constants depending only on the analytic profile and the number of degenerate radii could be extracted, although the paper does not optimize them."],"forward_implications":["The torus and every rotationally symmetric quasi-torus satisfy an $\\ell^4$ decoupling; on the positively curved exterior of the torus the argument upgrades to the stronger $\\ell^2$ decoupling.","The perturbed cone family, with profiles $\\gamma(r)=r+(r-1)^n+\\cdots$ for $n\\ge 3$, satisfies an $\\ell^4$ decoupling whose essentially flat boxes are smaller than the cone's usual length-one plates.","For every real analytic surface of revolution, a single essentially flat partition of the $\\delta$-neighborhood works for all functions at once, and the boxes generally have several different scales.","The $\\ell^4$ exponent is the natural endpoint for negatively curved pieces such as the inside of the torus; the proof does not claim $\\ell^2$ decoupling there."],"supporting_citations":[{"why":"Supplies the nonzero-Gaussian-curvature decoupling (Theorem 1) and the cone and cylinder decouplings (Theorem 3) that seed every nondegenerate scale and both rescaling arguments.","marker":"[1]"},{"why":"Supplies the companion nonzero-curvature decoupling theorem (Theorem 1) applied after each linear rescaling, so the final decoupling step depends on it.","marker":"[2]"}],"fun_headline_variants":["L4 decoupling for all real analytic surfaces of revolution","Torus and perturbed cone: L4 decoupling for surfaces of revolution","Every real analytic surface of revolution satisfies L4 decoupling","Real analytic surfaces of revolution: L4 decoupling for torus and cone"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole proof depends on the fact that every flattened cap can be stretched into a reference surface whose curvature is genuinely nonzero and whose shape is well controlled; if any rescaled cap stayed flat, the known decoupling theorem could not be applied and the argument would collapse.","fun_headline_variants_meta":{"raw":{"variants":["L4 decoupling for all real analytic surfaces of revolution","Torus and perturbed cone: L4 decoupling for surfaces of revolution","Every real analytic surface of revolution satisfies L4 decoupling","Real analytic surfaces of revolution: L4 decoupling for torus and cone"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001186,"raw_usage":{"total_tokens":4849,"prompt_tokens":850,"completion_tokens":3999,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":466,"completion_tokens_details":{"reasoning_tokens":3924}},"tokens_in":466,"tokens_out":3999,"duration_ms":29071,"temperature":1.0,"reasoning_tokens":3924,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:27:49.723190+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a profile such as $\\gamma(r)=r+(r-1)^3$, compute the determinant of the Hessian of the rescaled profile $\\psi_k$ on $|\\eta_1|,|\\eta_2|\\le 1$; the proof requires it to be comparable to $1$ uniformly in $k$. A real analytic $\\gamma$ with $\\gamma''(1)=0$, $\\gamma'''(1)\\ne 0$ for which this determinant tends to $0$, or an explicit function $f$ with Fourier support in $N_\\delta(S_\\gamma)$ violating the stated $\\ell^4$ inequality, would refute the main claim.","supporting_citations":[{"cited_title":"and Demeter, C","cited_arxiv_id":null,"evidence_quote":"Supplies the companion nonzero-curvature decoupling theorem (Theorem 1) applied after each linear rescaling, so the final decoupling step depends on it."}],"review_version":1}