{"id":"8856e9a9-e085-4178-87f2-c9a379c5c330","arxiv_id":"2607.13181","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Maximizers exist, and every maximizing sequence converges modulo symmetries, for L^p Fourier extension from the hyperbolic paraboloid at every exponent where the estimate is known.","lead":"This paper proves that for the sharp L^p Fourier extension inequality on the hyperbolic paraboloid — a saddle-shaped surface in three dimensions — optimal functions attaining the best constant always exist, whenever the inequality itself is known to hold. It is the last major negatively curved surface in the sharp-restriction program to receive this treatment, and the proof introduces an interpolation tool of independent interest.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 5.2 is the load-bearing unproved import; its property (ii) is malformed and the final Cauchy argument needs an exact L^p defect bound that the printed statement does not supply.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: Proposition 5.2 is imported without proof, and its stated property (ii) is not verifiable from the manuscript. My closer look at the proof of Theorem 1.1 confirms that this is not a cosmetic issue. The final argument hinges on the exact form of the L^p defect of compactness inequality: the chain around display (5.1) requires a bound on Σ∥ϕ_j∥_p^{\\tilde p} (or an exact statement that yields this after accounting for frequency localization), not the garbled Σ∥ϕ_j∥_{\\tilde p}^p that appears in the printed text. If the imported proposition has the weaker or different form, the main profile's extension norm would only be bounded below by a constant fraction of A_p that degrades as the localization radius tends to infinity, breaking the proof of (5.1) and the subsequent Cauchy argument. I also note the transfer from [33] (elliptic paraboloid) to the hyperbolic paraboloid is nontrivial; although the L^2 profile decomposition for the hyperbolic Schrödinger equation is available, the L^p upgrade may depend on the specific geometry, and the paper does not supply the details. No other internal inconsistency emerged: the range-localization and frequency-localization arguments (§3–4) are detailed and plausible, and the conditional framing of Theorem 1.1 is appropriate. The reader's CONDITIONAL verdict remains appropriate; the authors should fix the statement of Proposition 5.2, provide a proof or precise reference, and clarify the hyperbolic-paraboloid adaptation.","tokens_in":18288,"tokens_out":19640,"duration_ms":164825,"concrete_test":"Obtain the exact statement and proof of [33, Prop. 4.1]. (1) Verify that property (ii), as used at display (5.1), should be liminf_n(∥f_n∥_p^{\\tilde p} - Σ_j ∥ϕ_j∥_p^{\\tilde p}) ≥ 0, not the garbled printed form; if the true form is Σ∥ϕ_j∥_{\\tilde p}^p ≤1, trace the R-dependence and show (5.1) fails. (2) Check each step of the proof of [33, Prop. 4.1] for reliance on the elliptic parabolic phase e^{itΔ}; in particular, confirm that the only analytic input is Theorem 5.1 ([9, Thm 12]) and the refined Strichartz/bilinear estimates already proven in this paper (§3, [32]). If any step needs an L^{\\tilde p} extension bound with \\tilde p>p0, or a paraboloid-specific orthogonality, then the transfer is invalid and Theorem 1.1 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 5.2 is imported from [33] without proof, and the arXiv rendering of its property (ii) is malformed: it mixes p and \\tilde p inside/outside the sum. The proof of Theorem 1.1 uses this property at display (5.1) through the chain A_p^q - o_m(1) ≤ Σ_j ∥Eϕ^{m,j}∥_q^q ≤ ... ≤ A_p^q max_j ∥ϕ^{m,j}∥_p^{q−\\tilde p}. For that chain to close, property (ii) must provide a bound of the form liminf_n(∥f_n∥_p^{\\tilde p} - Σ_j ∥ϕ_j∥_p^{\\tilde p}) ≥ 0 (so Σ∥ϕ_j∥_p^{\\tilde p} ≤ 1), with \\tilde p = max(p,p') ≥ 2. If, as printed, the property instead bounds Σ∥ϕ_j∥_{\\tilde p}^p, then bounding ∥Eϕ_j∥_q ≤ A_p∥ϕ_j∥_p and using support/boundedness loses a factor C_R that grows with the localization radius R=m; the main profile's extension norm would only be ≥ c_m A_p with c_m→0, so (5.1) and the subsequent uniform-convexity/Cauchy argument collapse. Moreover, [33] is for the elliptic paraboloid; the transfer to the hyperbolic paraboloid is asserted but not shown, and it is not immediate that the L^2 profile decomposition [9, Thm 12] upgrades to the L^p frequency-localized decomposition without using elliptic-specific phase estimates. The paper explicitly says 'we will omit some details, such as the proof of Proposition 5.2' — so the central theorem rests on an unverified import.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that, conditional on the boundedness of the Fourier extension operator E from L^{p_0}(R^2) to L^{2p_0'}(R^{1+2}) for some 1<p_0<3, all nonendpoint inequalities \\|Ef\\|_{2p'} ≤ A_p\\|f\\|_p with 1<p<p_0 admit extremizers, and every maximizing sequence is precompact modulo the dilation/translation/modulation symmetries of the hyperbolic paraboloid. The argument combines a new sharpened Hunt–Marcinkiewicz interpolation lemma (Lemma 3.4), a range-localization and tile-selection routine (Section 3), a frequency-localization result for near-maximizers (Proposition 4.1, with the bilinear Lemma 4.2), and an L^p-based profile decomposition imported from [33] (Proposition 5.2). The theorem is explicitly conditional on the known or conjectured range of boundedness, and the proof is largely self-contained except for that imported decomposition.","tokens_in":18618,"tokens_out":28298,"duration_ms":237769,"significance":"If the result holds, it provides the first extremizer existence and modular precompactness theorem for Fourier extension from a negatively curved surface (the hyperbolic paraboloid) beyond the Stein–Tomas range, up to the currently known threshold p<11/4 and conditionally beyond. The sharpened Hunt–Marcinkiewicz lemma (Lemma 3.4) and the ‘funky characteristic function’ range-localization argument are substantive technical novelties of independent interest. The paper is honest about its conditional nature and about the fact that Proposition 5.2 is taken from [33]. However, the central proof depends critically on that imported decomposition, and the printed version contains several malformed displays that must be corrected before the argument can be verified.","major_comments":[{"comment":"The L^p profile decomposition is the load-bearing import from [33], yet property (ii) is malformed: the displayed inequality \"liminf_n(\\|f_n\\|_p - (Σ_{j=1}^{J_0} \\|φ_j\\|_{tilde p}^p)^{1/tilde p} ≥ 0\" mixes L^p and L^{tilde p} norms and the parentheses are unbalanced. The proof of Theorem 1.1 then uses the decomposition to conclude Σ\\|φ_j\\|_{tilde p}^p ≤ 1 and bounds \\|Eφ_j\\|_q via A_{tilde p}; the final conclusion requires \\|φ^m\\|_p ≥ 1-o_m(1). Please state the correct defect relation and either prove its adaptation to the hyperbolic-paraboloid phase or give a precise reference that covers this case. As printed, the central argument is not verifiable.","section":"§5, Proposition 5.2 and its use in Theorem 1.1"},{"comment":"The chain A_p^q - o_m(1) ≤ Σ\\|Eφ^{m,j}\\|_q^q ≤ A_{tilde p}^p max_j \\|Eφ^{m,j}\\|_q^{q-tilde p} Σ\\|φ^{m,j}\\|_{tilde p}^p ≤ A_{tilde p}^p max_j \\|Eφ^{m,j}\\|_q^{q-tilde p} ≤ A_p^q max_j \\|φ^{m,j}\\|_p^{q-tilde p} is internally inconsistent: the middle terms use A_{tilde p} and L^{tilde p} norms while the final bound uses A_p and L^p norms. For the conclusion \\|φ^m\\|_p ≥ 1-o_m(1) to follow, one expects a bound of the form Σ\\|φ_j\\|_p^{tilde p} ≤ 1 (or with matching powers). Please rewrite this display and reconcile it with the corrected Proposition 5.2(ii).","section":"§5, proof of Theorem 1.1 (display before (5.1))"},{"comment":"The assertion that Q_i^n is nonempty for at most two values of i with i ≥ -l_n (and analogously τ_i^n with i ≥ k_n) appears false when the bad tile overlaps the unit square: the dyadic annuli {|ξ_2-ζ_{n,2}| ∼ 2^{l_n+i}} intersect [0,1] for O(-l_n) values of i, and similarly there are O(k_n) nonempty vertical strips. The subsequent summation over i,i' ≥ C depends on this counting statement. The final estimate may still hold because the bilinear exponent 2-4/q-2/r is negative for r sufficiently close to (q/2)', but the written proof needs a corrected accounting of the number of nonempty strips.","section":"§4, Lemma 4.2, Case 2"}],"minor_comments":[{"comment":"The right-hand side of the second estimate reads \"A_p - o_m(1) ≤ \\|Eφ^m\\|_p\"; the target norm should be \\|Eφ^m\\|_q. Also the earlier display starts with \"A_q^p\" where the context requires \"A_p^q\".","section":"§5, Eq. (5.1)"},{"comment":"Please fix the unbalanced parentheses in the statement of property (ii). The current rendering is not a well-formed inequality.","section":"§5, Proposition 5.2(ii)"},{"comment":"The phrase using \"X ∼= Y\" for X=CY is nonstandard; consider using \"X = C Y\" or define the notation more explicitly.","section":"§1.6, terminology"},{"comment":"Typo: \"Hunt–Marcinkiwicz\" should be \"Hunt–Marcinkiewicz\".","section":"Keywords"},{"comment":"[23] is cited with an access date in April 2026; if this is a preprint/lecture note, please provide a more stable reference or archive link.","section":"References"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The paper proves the first existence/precompactness result for maximizers of L^p Fourier extension from the hyperbolic paraboloid in R^3, conditionally on boundedness of the extension operator at some p0. That closes a genuine gap in the sharp restriction program, and the conditional framing is honest: if boundedness holds, you get extremizers and compactness modulo symmetries for every p<p0.\n\nWhat is genuinely new is the frequency and range localization for negative curvature. The sharpened Hunt–Marcinkiewicz lemma (Lemma 3.4) is a nice piece of work, and Lemma 4.2 — the asymptotic orthogonality of good and bad pieces with tracked constants — is proved in detail. These are the hard parts, and they look solid to me. The paper also avoids circularity: the boundedness assumption is an input, not derived from the maximizer, and the heavy citations to Stovall's earlier papers are to independent established theorems.\n\nThe soft spot is the same one the stress-tester flags: Proposition 5.2, the L^p frequency-localized profile decomposition, is imported from [33] with 'we will omit some details,' and the printed statement of property (ii) is malformed — the norm/exponent inside the sum doesn't match the power outside, and the missing parenthesis makes the liminf condition unverifiable. Since the final Cauchy argument runs through that property, a referee cannot check the proof as it stands. I don't think this is fatal: the L^2 profile decomposition [9] is already for the hyperbolic Schrödinger equation, so the paraboloid-based upgrade in [33] should transfer without much trouble, but the authors need to state the property correctly and either prove it or provide a detailed adaptation. The typesetting in display (5.1) is also hard to parse, and the dependence on the 2026 preprint [19] for the current record range should be explicit.\n\nNet: this is a substantial, likely correct contribution. It deserves a serious referee, not a desk reject. My advice to the editor: send it out, but require the authors to fix the statement and proof of Prop 5.2 and clean up the typos before publication.","headline":"Solid and likely correct: first extremizer existence for the hyperbolic paraboloid, conditional on boundedness, with a real but fixable gap in the imported profile decomposition.","tokens_in":19210,"tokens_out":7301,"would_cite":true,"duration_ms":55023,"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":"For the hyperbolic paraboloid, every nonendpoint L^p extension inequality below the boundedness range is maximizable, with maximizing sequences precompact modulo symmetries.","keywords":["Fourier extension","hyperbolic paraboloid","maximizers","precompactness","sharp constants","Hunt–Marcinkiewicz interpolation","profile decomposition","restriction theory"],"falsifier":"Find an $L^p$-normalized maximizing sequence for some $p<p_0$ whose $L^p$-profile decomposition (Proposition 5.2) has two nonzero profiles with nonzero extension norms; the claimed convergence to a single profile would be contradicted, since the proof's uniform-convexity step forces all but one profile to vanish.","tokens_in":18087,"feed_emoji":"📐","tokens_out":9285,"duration_ms":89466,"temperature":0.7,"texified_at":"2026-08-05T21:20:08.730209+00:00","pith_summary":"The paper proves that the sharp $L^p$ Fourier extension inequality for the hyperbolic paraboloid in $\\mathbb{R}^3$ has actual maximizers whenever the exponent $p$ lies strictly below any exponent for which the extension operator is known to be bounded. It also proves that every norm-one maximizing sequence has a subsequence converging, after dilations, frequency translations, and modulations, to a maximizer. This addresses a class of negatively curved surfaces where earlier existence proofs, built for the paraboloid, cone, and sphere, did not transfer because bilinear-to-linear arguments lose too much. The proof rests on a sharpened Hunt–Marcinkiewicz interpolation lemma, a tile-based frequency localization showing that near-maximizers concentrate on a single tile, and an imported $L^p$-profile decomposition whose pieces have additive extension norms.","texify_model":"deepseek-v4-flash","texify_usage":{"total_tokens":5207,"prompt_tokens":827,"completion_tokens":4380,"prompt_tokens_details":{"cached_tokens":0},"prompt_cache_hit_tokens":0,"prompt_cache_miss_tokens":827,"completion_tokens_details":{"reasoning_tokens":3559}},"feed_headline":"Hyperbolic paraboloid gains maximizers at every nonendpoint exponent","feed_subtitle":"A sharpened interpolation lemma and tile localization prove maximizing sequences converge modulo symmetries.","key_machinery":"The central mechanism is a two-step localization. First, a sharpened Hunt–Marcinkiewicz interpolation lemma (Lemma 3.4) converts the operator's restricted weak-type bounds into an estimate controlling $\\|Ef\\|_q$ by the largest contribution of $f$'s dyadic-level sets on dyadic tiles (Proposition 3.1). A tile-selection argument (Proposition 3.5) then shows that near-maximizers must have a 'first tile' carrying a non-negligible share of the norm, and Proposition 4.1 upgrades this to: after composing with a symmetry, a near-maximizer is almost entirely supported in a fixed ball and has bounded amplitude. With that frequency localization in hand, the proof imports an $L^p$-based profile decomposition $f$","core_discovery":"Let $\\Sigma = \\{(\\xi_1\\xi_2, \\xi_1, \\xi_2) : \\xi \\in \\mathbb{R}^2\\}$ be the hyperbolic paraboloid and $Ef(t,x) = \\int_{\\mathbb{R}^2} e^{i(t,x)\\cdot(\\xi_1\\xi_2,\\xi)} f(\\xi) \\, d\\xi$ its Fourier extension operator. The main theorem assumes $E$ is bounded from $L^{p_0}(\\mathbb{R}^2)$ into $L^{2p_0'}(\\mathbb{R}^{1+2})$ for some $1<p_0<3$. Then for every $1<p<p_0$ the sharp constant $A_p = \\sup_{\\|f\\|_p=1} \\|Ef\\|_{2p'}$ is attained: there is a nonzero $f \\in L^p$ with $\\|Ef\\|_{2p'} = A_p \\|f\\|_p$. Moreover, any $L^p$-normalized maximizing sequence has a subsequence that, after precomposing with dilations, frequency translations, and modulations, converges in $L^p$ to a maximizer. In the known boundedness range $p_0 < 11/4$ the theorem is unconditional; for $11/4 \\le p_0 < 3$ it is conditional on future improvement","pith_inferences":["The proof's reliance on an imported profile decomposition suggests a modular strategy: the same frequency-localization and interpolation steps could, in principle, be adapted to any hypersurface that admits a suitable L^p-profile decomposition and a refined Strichartz inequality.","Because maximizing sequences converge to a single profile, the sharp constant A_p could be approached numerically by solving a low-dimensional variational problem over the symmetry group of a known maximizer, rather than over all of L^p.","If the restriction conjecture for the hyperbolic paraboloid (boundedness for all p<3) is eventually proved, Theorem 1.1 would immediately yield maximizers and compactness for every p in (1,3), closing the gap at the endpoint behavior p→3.","The malformed condition (ii) in the printed profile decomposition is a testable detail: a reader can compare against the original statement and verify whether the exponent should be \\tilde p in the sum; if the original statement differs, the present proof may need a small adjustment in the Cauchy step."],"forward_implications":["Existence: for every p below the boundedness threshold, the sharp constant A_p is attained; no loss of compactness occurs at nonendpoint exponents.","Structure of near-extremizers: every maximizing sequence is asymptotically a single modulated, dilated, and translated copy of a fixed maximizer, giving a complete modulo-symmetry description.","Duality: the analogous precompactness and existence statement holds for the restriction operator mapping L^{2p'} to L^p (Remark 1.2).","Robustness: any future improvement of the boundedness range will automatically deliver maximizers and compactness for all newly covered exponents, since Theorem 1.1 is conditional only on boundedness.","New interpolation tool: the sharpened Hunt–Marcinkiewicz lemma provides a general way to pass from restricted weak-type bounds to tile-localized estimates, which may be useful for other Fourier extension problems."],"fun_headline_variants":["Maximizers exist for L^p extension from hyperbolic paraboloid","Hyperbolic paraboloid gains optimal functions in 3D","Extension maximizers proven for hyperbolic paraboloid","Sharp constants attained for hyperbolic paraboloid inequality","Modulo symmetries, maximizing sequences converge"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof relies on an imported $L^p$-profile decomposition for frequency-localized sequences that is stated without proof and whose printed norm identity appears malformed; if that decomposition fails in the symmetry setting of the hyperbolic paraboloid, the main theorem does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Maximizers exist for L^p extension from hyperbolic paraboloid","Hyperbolic paraboloid gains optimal functions in 3D","Extension maximizers proven for hyperbolic paraboloid","Sharp constants attained for hyperbolic paraboloid inequality","Modulo symmetries, maximizing sequences converge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000173,"raw_usage":{"total_tokens":1069,"prompt_tokens":651,"completion_tokens":418,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":395,"completion_tokens_details":{"reasoning_tokens":354}},"tokens_in":395,"tokens_out":418,"duration_ms":4130,"temperature":1.0,"reasoning_tokens":354,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T05:57:32.270479+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find an $L^p$-normalized maximizing sequence for some $p<p_0$ whose $L^p$-profile decomposition (Proposition 5.2) has two nonzero profiles with nonzero extension norms; the claimed convergence to a single profile would be contradicted, since the proof's uniform-convexity step forces all but one profile to vanish.","supporting_citations":[],"review_version":1}