{"id":"ce575605-1d89-4c5d-8d4b-e8cf332363db","arxiv_id":"2412.03942","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Radial Fourier restriction to the sphere satisfies a dimension-free endpoint Stein-Tomas bound, general functions lose only d^{1/2}, and best constants tend to zero or infinity according to the Riesz region.","lead":"This paper studies the best constant in the Fourier restriction inequality to the unit sphere as the dimension of the ambient space grows. It proves that for radial functions the endpoint Stein-Tomas constant is bounded between two universal numbers independent of dimension, while general functions pay at most a square-root factor in dimension.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 7's L² bound (3.6) is false: the delta identity omits the Jacobian 1/√(1−|x'|²), making the Fourier multiplier K unbounded, so U(τ) is not bounded on L² and the proof of (1.8) collapses.","rationale":"The reader's verdict accepts the paper on the strength of the Bessel-based radial arguments and their internally coherent presentation. The weakest_assumption flagged by the reader, uniformity of the Krasikov error bound in Proposition 3, is a real but secondary risk, and it only affects the radial statements. The more serious problem is in the proof of the general estimate (1.8). Lemma 7's L² bound is a central tool, and its proof contains an identifiable computational error: the delta integral over S^{d-1} is evaluated without the 1/√(1−|x'|²) Jacobian, so the multiplier is unbounded rather than bounded by 2. This makes U(τ) unbounded on L² for every τ, including τ=0, so the interpolation step that yields the O(d^{1/2}) general bound is invalid. The failure is independent of Bessel asymptotics and cannot be fixed by a constant-readjustment; either the operator U(τ) needs a different analysis or the claimed general bound lacks proof. Because this supports both Theorem 2(1.8) and the general-function part of Theorem 1, the current manuscript should not be accepted as proving those claims. I would move the verdict to REJECT, while noting that the radial results may survive after a separate verification of Proposition 3.","tokens_in":15989,"tokens_out":21017,"duration_ms":183267,"concrete_test":"Recompute K in Lemma 7 with the surface measure written as dσ(ω)=dω'/√(1−|ω'|²). For d=2 and τ=0, the multiplier is exactly 2/√(1−x²) on |x|<1. Let ĝ_N(x)=√N 1_{[1−2/N,1−1/N]}(x), so ‖ĝ_N‖₂=1. Then ‖U(0)g_N‖₂²=∫4|ĝ_N(x)|²/(1−x²)dx ≥4N ln 2, so the ratio ‖U(0)g_N‖₂/‖g_N‖₂ tends to infinity, directly contradicting (3.6). More generally, the unbounded multiplier shows U(0) is not bounded on L².","verdict_should_be":"REJECT","load_bearing_attack":"The load-bearing step is Lemma 7 in §3.1, which underpins Theorem 2(1.8). In the proof of (3.6), the Fourier multiplier of U(τ) is computed as K(x',τ)=∫_{S^{d-1}} δ(ω'+x') e^{-2πiτ ω_d} dσ(ω), and it is asserted to equal 2 cos(2πτ√(1−|x'|²)) on |x'|≤1. This equality omits the Jacobian of the parametrization of S^{d-1} by hemispheres. Writing dσ(ω)=dω'/√(1−|ω'|²) gives K(x',τ)=2 cos(2πτ√(1−|x'|²))/√(1−|x'|²), which is unbounded as |x'|→1. For τ=0 the multiplier is 2/√(1−|x'|²), which is not in L∞. By Plancherel, ‖U(0)g‖₂=‖(2/√(1−|·|²))ĝ‖₂, so taking ĝ supported near |x'|=1 gives an arbitrarily large ratio and contradicts (3.6). Thus the Riesz–Thorin interpolation (3.8) and the proof of (1.8) fail. Since Theorem 1(1.5) interpolates from (1.8), the general-function dimension-free claim is not established. This is independent of the Proposition 3 Bessel lower-bound concern.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the asymptotic behavior, as the dimension d tends to infinity, of the best constants in Fourier restriction inequalities to the unit sphere, both for general functions and for radial functions. The main results are: a uniform two-sided L^p estimate for Bessel functions (Proposition 3); a dimension-free endpoint Stein--Tomas bound for radial functions together with an O(d^{1/2}) bound for general functions (Theorem 2); and asymptotic results for the restriction constants in three Riesz-exponent regions (Theorem 1). The proofs use exact formulas for radial constants, fractional integration, interpolation, and a refined Bessel-function approximation of Krasikov.","tokens_in":16283,"tokens_out":21174,"duration_ms":210396,"significance":"The radial results are potentially significant: Theorem 2(1.9), the dimension-free boundedness of the radial endpoint Stein--Tomas constant, and the radial asymptotics in Theorem 1(1.6)--(1.7) are new and rest on explicit, checkable computations: the exact radial norm formula (3.15), the Stirling and Hardy--Littlewood--Sobolev constant computations (3.10)--(3.11), and the uniform two-sided Bessel estimates of Proposition 3. The general O(d^{1/2}) bound in (1.8) would also be a strong step toward the conjectured dimension-free general endpoint inequality, but as written its proof contains a serious error. The paper is self-contained except for standard external results, and the Bessel proposition is of independent interest.","major_comments":[{"comment":"The computation of the multiplier K(x',τ) in Lemma 7 is incorrect. Parametrizing the sphere by its two hemispheres with ω' = -x' gives dσ(ω) = dω'/√(1-|ω'|²), so the delta identity yields K(x',τ) = 2 cos(2πτ√(1-|x'|²)) / √(1-|x'|²) for |x'|<1 (and 0 for |x'|>1), not 2 cos(2πτ√(1-|x'|²)). Thus |K| is unbounded. For τ=0, K = 2/√(1-|x'|²) is not in L^∞(R^{d-1}), so by Plancherel the inequality ‖U(0)g‖_2 ≤ 2‖g‖_2 fails: taking ĝ supported in an annulus {1-ε<|x'|<1} with ‖ĝ‖_2=1 gives ‖K(·,0)ĝ‖_2 / ‖ĝ‖_2 ~ √log(1/ε) → ∞. Consequently (3.6) is false, and the Riesz–Thorin step (3.8) together with the fractional-integration estimate (3.9) is unsupported. The claimed bound (1.8) therefore lacks a valid proof as written.","section":"§3.1, Eq. (3.6)"},{"comment":"The proof of Theorem 1(1.5) interpolates from the general estimate (1.8) via (4.4). Since the proof of (1.8) depends on Lemma 7's false L^2 bound, the general-function result (1.8) and its consequences for region A are not established by the manuscript. The radial results (1.6), (1.7), and (1.9) are independent of Lemma 7 and appear sound. The authors should either supply a corrected proof of the O(d^{1/2}) bound for general functions or explicitly separate the unproved general statements from the proved radial ones.","section":"§4.1, Eq. (4.4)"}],"minor_comments":[{"comment":"In (3.14), the factor should be σ(S^{d-1})^{-1/p}, not σ(S^{d-1})^{1/p}; with the displayed factor, identity (3.15) does not follow. The subsequent identity (3.15) is correct, so this appears to be a typo, but it should be fixed.","section":"§3.2, Eq. (3.14)"},{"comment":"The proof of (1.9) is written for sufficiently large d, after Proposition 3 is invoked. The statement claims the bound for every d ≥ 2; a sentence should explain that the finitely many small dimensions are absorbed into the absolute constants a and A.","section":"§3.2, proof of (1.9)"},{"comment":"In the lower-bound proof, the reduction to (2.13) uses the estimate 16ν+4π ≤ 24ν for ν ≥ 2; this is correct but should be stated explicitly. Also, the notation '(30)^{αp}' in (2.11) is used before the constant 30 is introduced; a brief clarification would help.","section":"§2, proof of Proposition 3"}],"recommendation":"major_revision","confidential_remarks":"The radial half of the paper is solid and likely publishable on its own, but the general O(d^{1/2}) bound rests on a false L^2 estimate in Lemma 7. This is not a presentation issue: the operator U(τ) is genuinely unbounded on L^2 for the stated kernel. I would recommend inviting a revision that either proves (1.8) with a correct substitute for the L^2 bound, or removes the unproved general-function statements from the main results and revises the abstract and title accordingly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's what I make of this paper. The radial-function results are genuinely interesting: the exact formula (3.15) for the radial constant, the two-sided Bessel estimate Proposition 3, and the asymptotic classification in the two Riesz regions are solid work, assuming Proposition 3's uniform error bound from Krasikov holds up. The paper is honest about what is new and doesn't overclaim.\n\nBut there is a load-bearing flaw in the proof of the general endpoint bound (1.8). Lemma 7 claims that the Fourier multiplier of U(τ) is 2 cos(2πτ√(1−|x'|²)) on the unit ball, and hence bounded. That identity is wrong: the delta computation omits the Jacobian of the sphere parametrization. With dσ(ω) = dω'/√(1−|ω'|²), the multiplier is 2 cos(...)/√(1−|x'|²), which is unbounded at |x'|=1. So U(τ) is not bounded on L², and the Riesz–Thorin interpolation leading to (1.8) collapses. Since (1.5) interpolates from (1.8), that part of Theorem 1 is unproven too.\n\nI checked the computation carefully, so I think the stress-test note is right. This is not a subtle point; it's a textbook Jacobian.\n\nThe reader's worry about Proposition 3's lower bound is more of a potential fragility than an actual contradiction. I didn't find a concrete mistake there. The real problem is Lemma 7.\n\nWhat this means: the paper contains a correct and novel treatment of radial restriction constants, but the headline general-function dimension-free estimate is not established. As it stands, the paper should not be accepted. A referee could verify the radial part and the Bessel analysis, and the authors might repair the general argument (or weaken the claim). I'd send it out for review, but with a clear flag that the main theorem's proof fails.","headline":"Radial restriction results are solid, but the proof of the O(d^{1/2}) general bound fails at Lemma 7's missing Jacobian.","tokens_in":16882,"tokens_out":10350,"would_cite":false,"duration_ms":85997,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Radial endpoint Stein–Tomas constants are dimension-free on spheres; general ones grow at most like √d.","keywords":["Fourier restriction","dimension-free estimates","Stein–Tomas inequality","Bessel functions","radial functions","high-dimensional analysis","best constants"],"falsifier":"Numerically evaluate the integral in the lower bound (1.13) for large ν with α near −1/2 and p near 2; if the ratio to the claimed ν-power falls below the absolute constant, the bound fails. Alternatively, test the error estimate in Lemma 6 directly for large ν and r just above 2ν to see whether the ($r^{2}$−μ)^{−3/4} bound still holds uniformly.","tokens_in":15742,"feed_emoji":"📐","tokens_out":6883,"duration_ms":63638,"temperature":0.7,"pith_summary":"This paper addresses how the best constants in Fourier restriction to the sphere behave as the dimension d tends to infinity. At the endpoint Stein–Tomas exponent p_*=2(d+1)/(d+3), it proves that the restriction constant for radial functions is bounded above and below by absolute constants independent of d, while the constant for general functions is at most C $d^{{1/2}}$. In the subcritical region A, the best constants for general functions tend to zero as d→∞; for radial functions in region B the constants tend to zero at the line q=p' and diverge to infinity when q>p'. These results settle the dimension dependence of these operator norms and mark progress toward the conjecture that constant functions are global maximizers.","feed_headline":"Radial Stein–Tomas constant is dimension-free on spheres","feed_subtitle":"For Fourier restriction to the sphere, the best endpoint constant for radial functions stays between two absolute constants in every…","key_machinery":"The load-bearing object is a two-sided uniform L^p estimate for Bessel functions (Proposition 3) that tracks the dependence on the order ν=d/2−1. It controls the weighted integrals ∫|J_ν(r)r^α|^p dr from above and below with explicit ν-powers, refining a known asymptotic estimate so that the radial restriction constant can be computed up to absolute factors. For the general estimate, a fractional-integration argument in the direction transverse to the sphere reduces the problem to the one-dimensional sharp Hardy–Littlewood–Sobolev inequality, whose constant is tracked carefully in d.","core_discovery":"The paper's central discovery is that, at the endpoint Stein–Tomas exponent, the radial restriction operator is dimension-free: there are absolute constants 0<a<A<∞ with a ≤ R_{$S^{{d-1}}$}(p_*→2;rad) ≤ A for every d≥2. The corresponding constant for general functions satisfies R_{$S^{{d-1}}$}(p_*→2) ≤ C $d^{{1/2}}$ with C absolute, and the authors show this square-root loss is the only price of generality in the endpoint estimate. These bounds yield a phase portrait as d→∞: in region A the best constants decay to zero, in region B radial constants decay to zero at q=p' and blow up when q>p'. The proof is built from an exact integral expression for the radial constant in terms of a Bessel function and from a fractional-integration argument transverse to the sphere.","pith_inferences":["If the constant-function maximizer conjecture (1.10) holds, the general endpoint constant would coincide with the radial one and hence be dimension-free; the paper's bounds are consistent with that.","The same Bessel-based machinery could be applied to other rotation-invariant hypersurfaces, where similar dimension thresholds might be decided by uniform asymptotics of the associated special functions.","The O(d^{1/2}) growth for general functions suggests the loss in the transverse direction is the only obstruction to dimension-free restriction, since radial functions see no such loss."],"forward_implications":["The radial endpoint Stein–Tomas inequality is dimension-free: the same absolute constants work for every d≥2, so radial restriction estimates do not degrade in high dimensions.","The general endpoint constant has at most square-root growth in d; if the conjectured equality with constant functions holds, it would actually be dimension-free.","Best constants in region A tend to zero as d→∞, so the operator norms shrink rather than stabilize in high dimensions.","For radial functions the boundary q=p' in region B is sharp: constants tend to zero on the boundary and diverge above it."],"supporting_citations":[{"why":"Supplies the Bessel approximation with explicit error term used in the proof of the uniform lower bound.","marker":"[18]"},{"why":"Contains the weighted L^p Bessel estimate that Proposition 3 refines with uniform constants.","marker":"[29]"},{"why":"Gives the sharp one-dimensional Hardy–Littlewood–Sobolev inequality whose constant produces the d^{1/2} dependence in the general endpoint bound.","marker":"[19]"},{"why":"Provides the fractional integration method for the Stein–Tomas inequality that is adapted and tracked in dimension.","marker":"[21]"},{"why":"Provides the uniform pointwise Bessel estimates used for the upper bounds in Proposition 3.","marker":"[10]"},{"why":"Gives the pointwise bound (2.4) used to control the Fourier transform of surface measure.","marker":"[23]"}],"fun_headline_variants":["Dimension-free radial Stein–Tomas constant on spheres","Radial restriction constant stays bounded in all dimensions","Sphere Fourier restriction: radial bound is dimension-free","General sphere restriction: only sqrt(d) dependence"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The uniform lower bound in Proposition 3 requires that the error term in the Bessel approximation lemma, bounded by ($r^{2}$−μ)^{−3/4}, hold with an absolute constant independent of ν; any positive power of ν in that error would break the lower bounds and the divergence statements.","fun_headline_variants_meta":{"raw":{"variants":["Dimension-free radial Stein–Tomas constant on spheres","Radial restriction constant stays bounded in all dimensions","Sphere Fourier restriction: radial bound is dimension-free","General sphere restriction: only sqrt(d) dependence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001396,"raw_usage":{"total_tokens":5625,"prompt_tokens":904,"completion_tokens":4721,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":4661}},"tokens_in":520,"tokens_out":4721,"duration_ms":37728,"temperature":1.0,"reasoning_tokens":4661,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T21:56:54.430693+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically evaluate the integral in the lower bound (1.13) for large ν with α near −1/2 and p near 2; if the ratio to the claimed ν-power falls below the absolute constant, the bound fails. Alternatively, test the error estimate in Lemma 6 directly for large ν and r just above 2ν to see whether the ($r^{2}$−μ)^{−3/4} bound still holds uniformly.","supporting_citations":[{"cited_title":"Krasikov, Approximations for the Bessel and Airy functions with an exp licit error term","cited_arxiv_id":null,"evidence_quote":"Supplies the Bessel approximation with explicit error term used in the proof of the uniform lower bound."},{"cited_title":"Stempak, A weighted uniform Lp-estimate of Bessel functions: a note on a paper of Guo","cited_arxiv_id":null,"evidence_quote":"Contains the weighted L^p Bessel estimate that Proposition 3 refines with uniform constants."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the sharp one-dimensional Hardy–Littlewood–Sobolev inequality whose constant produces the d^{1/2} dependence in the general endpoint bound."},{"cited_title":"Muscalu, W","cited_arxiv_id":null,"evidence_quote":"Provides the fractional integration method for the Stein–Tomas inequality that is adapted and tracked in dimension."},{"cited_title":"Carneiro, D","cited_arxiv_id":null,"evidence_quote":"Provides the uniform pointwise Bessel estimates used for the upper bounds in Proposition 3."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the pointwise bound (2.4) used to control the Fourier transform of surface measure."}],"review_version":1}