{"id":"8a9d7af6-2fd9-4589-8550-ee07d6a43c50","arxiv_id":"2608.08423","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"An independent proof establishes the sharp log-free L^2 to L^{10/3} endpoint estimate for the two-dimensional Hermite operator.","lead":"This paper proves a sharp bound, with no logarithmic loss, for Hermite eigenfunction projections in two dimensions. A generalist reader might care because it closes an open problem in harmonic analysis and offers a reusable technique for removing logarithmic factors in spectral projection estimates.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 2.1's full-annulus bounds may not follow from the ℓ∞-over-pieces Koch–Tataru input if an annulus contains many pieces.","rationale":"Read in good faith, the internal argument of the paper is otherwise coherent. The polar decomposition, Liouville–Green representation, weighted van der Corput/TT* estimate, forbidden-region decay, and final region decomposition all check out. The obvious textual issues—the line r(u)=√λ(1-u), which should be √λ√(1-u), and the sign statement in Lemma 4.2—are not load-bearing: all subsequent formulas are consistent with u=1-|x|²/λ, and Proposition 5.2 only needs constant sign, not a particular sign. The genuinely load-bearing step is the transfer from the ℓ∞-over-pieces Koch–Tataru estimate (2.2) to full-annulus bounds (2.3)–(2.5). If the pieces covering a full annulus number O(1), the proof works; if they number λ^α, the L^p summation over the circle adds a factor λ^{α/p}, breaking the sharp exponent. This sharpens the reader's flagged external assumption to the piece-counting point, so the reader's conditional verdict remains appropriate pending this check.","tokens_in":13954,"tokens_out":48380,"duration_ms":457477,"concrete_test":"Inspect [12, Theorem 3(a)] and the definition of its decomposition, then count the number of pieces intersecting (i) A_in_μ at μ=C*λ^{-2/3} and (ii) {|u|≤4C*λ^{-2/3}}, for λ=2N+2 as N→∞. If the count is O(1), the proof stands. If it grows as λ^α, replace the left side of (2.3)–(2.5) by the ℓ^p sum over the pieces and recompute the exponents; if the resulting L^2→L^{10/3} bound is λ^{-1/10+β} with β>0, Theorem 1.1 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2 converts the Koch–Tataru estimate (2.2), stated as an ℓ∞_ρ L^p norm over pieces, into the full-annulus bounds (2.3)–(2.5). The only justification is the sentence that each annulus meets only a bounded number of pieces in the Koch–Tataru turning-point decomposition. For the turning zone {|u|≤Au*} and for A_in_μ with μ≳u*, the set is a full circle of radius ρ=√λ. If the pieces have the natural λ^{-1/2} tangential scale, the number of pieces in such an annulus grows like λ^{1/2}(μ/u*) (and like λ^{1/2} for the turning zone), so an ℓ^p summation over pieces would multiply the bounds by a positive power of λ and destroy the λ^{-1/10} endpoint. The paper does not define the pieces or prove the bounded-overlap claim. Estimates (2.3)–(2.5) feed directly into Propositions 6.1 and 7.1 and into Section 8, so Theorem 1.1 depends on this unverified step. If the assertion is instead correct—for example, if the pieces in the Koch–Tataru theorem are the full annuli or are chosen with polar symmetry—the argument closes. The text alone does not settle which is the case.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves the sharp log-free endpoint estimate for two-dimensional Hermite spectral projections: for λ = 2N+2 and every f in L²(R²), the projection Π_λ satisfies ∥Π_λ f∥_{L^{10/3}} ≲ λ^{-1/10}∥f∥_{L²}. The proof uses a polar-coordinate spectral decomposition, the Koch–Tataru localized eigenfunction bounds, a uniform Liouville–Green representation for radial modes away from turning points, a curvature estimate for the resulting phases, a weighted van der Corput exponential-sum lemma, and a weighted TT* argument over radial scales. The remaining modes are controlled by the Koch–Tataru annulus bounds and by rapid decay in the forbidden region. The paper also shows sharpness via the highest-angular-momentum eigenfunction. The main theorem is the same result already announced in the contemporaneous preprint [10], but the proof strategy here is independent and structurally different.","tokens_in":14274,"tokens_out":35470,"duration_ms":379191,"significance":"If the proof is correct, the paper gives a clean resolution of the two-dimensional endpoint eigenfunction problem by a different mechanism from the asymmetric-localization approach of Jeong–Lee–Ryu, and the weighted TT* argument in Proposition 5.2 is potentially reusable. The proof is largely self-contained beyond the Koch–Tataru input: the Liouville–Green representation, the forbidden-region decay, the exponential-sum estimates, and the TT* coupling are all explicit, with universal constants and no fitted parameters. However, the derivation of the full-annulus bounds from the Koch–Tataru estimate (2.2) is not sufficiently justified as written, and this step is load-bearing for the rest of the paper. The novelty overlap with [10] should also be weighed, though the independent derivation is a point in the manuscript's favor.","major_comments":[{"comment":"Proposition 2.1 is derived from the Koch–Tataru estimate (2.2) by asserting that each annulus A_in_μ, A_out_μ, and the set {|u|≤Au*} meets only a bounded number of pieces of the Koch–Tataru turning-point decomposition, but the pieces are never defined. This is not a purely cosmetic issue: if the pieces are the standard ρ^{-1/3} Cartesian cubes (the literal reading of ℓ∞_ρ L^p), then an annulus A_in_μ with μ≫u* has y-width λ^{2/3}μ and contains roughly λ^{5/6}μ cubes across its radial thickness, so the displayed bounds (2.3)–(2.5) would gain a positive power of λ after summation over pieces, and the exterior summability in (2.9) would not close. If instead the pieces are dyadic annuli in the variable y, or are the full annuli themselves, the bounded-overlap claim is true. The manuscript must either quote from [12] the precise theorem that yields the full-annulus bounds directly, or define the pieces and prove the bounded-overlap assertion. Since estimates (2.3)–(2.8) are used in (2.9), in (7.4), and in the final covering argument in Section 8, this gap is load-bearing for Theorem 1.1.","section":"Section 2, Proposition 2.1 and (2.2)–(2.8)"}],"minor_comments":[{"comment":"The definition r(u)=√λ(1-u) conflicts with u=1-|x|²/λ and with the factorization (3.5). The correct relation is r(u)=√λ(1-u)^{1/2}; the later computation |rdr|=(λ/2)du in Section 6 is consistent with the square-root definition but not with r(u)=√λ(1-u). This appears to be a typo, but it should be corrected because the phase integrals and the change of variables depend on it.","section":"Section 3, after (3.5)"},{"comment":"The sign assertion in Lemma 4.2 is wrong: for t<s, the second derivative ∂²_ξ(S(u_t,ξ)-S(u_s,ξ)) is positive, not negative, because ∂²_ξ p_ξ<0 and r(u) is decreasing. The magnitude bound |∂²|≈|t-s| is correct, and the sign is immaterial for the subsequent van der Corput application since Lemma 5.1 only requires a constant sign. Still, the statement 'this derivative has the sign of t−s' and the last sentence of the proof should be corrected.","section":"Lemma 4.2"},{"comment":"The sentence 'Here the middle estimate follows because the indicated set meets only O(1) turning-point pieces' explicitly justifies bounded overlap only for {|u|≤Au*}. The first and third displayed estimates in the proof are obtained by the same type of restriction, so the manuscript should either give the analogous justification for A_in_μ and A_out_μ or state that these are the Koch–Tataru annulus estimates themselves.","section":"Section 2, proof of Proposition 2.1"},{"comment":"The notation ℓ∞_ρ L^p is introduced but the parameter ρ and the pieces of the decomposition are not precisely defined in the manuscript. A short summary of the relevant Koch–Tataru definitions, or at least a precise quotation of [12, Theorem 3(a)], would remove a significant obstacle for the reader.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"Editor: the main theorem is already claimed in the contemporaneous preprint [10], so the novelty decision rests on whether the independent proof by a different method is of sufficient interest. The mathematical argument appears plausible and the internal estimates are mostly coherent, but I was unable to verify the conversion from the Koch–Tataru ℓ∞_ρ L^p estimate to the full-annulus bounds in Section 2 without knowing the exact definition of the 'pieces'. I would ask an expert in the Koch–Tataru paper to check this point, as it is the main gate for the rest of the proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main thing you should know: the theorem is not new. Jeong–Lee–Ryu already proved the same sharp endpoint bound in a concurrent preprint, and the authors acknowledge this up front. What this paper offers is an independent proof strategy — polar decomposition, a Liouville–Green representation, discrete van der Corput, and a weighted TT* argument that couples radial scales. That is a genuine contribution, even if the result itself is known.\n\nThe proof is careful and unusually complete. The Liouville–Green lemma is proved with explicit error bounds, the forbidden-region decay is worked out with the right exponents, and the weighted TT* proposition is stated cleanly and proven. The external Koch–Tataru input is explicitly recorded, and the authors do not hide their dependence on it. All of that is good practice.\n\nThe most serious question is in Section 2. Proposition 2.1 converts the Koch–Tataru ℓ∞-over-pieces bound (2.2) into the full-annulus bounds (2.3)–(2.5), justified only by a sentence that each annulus meets a bounded number of pieces. The text never defines the pieces or proves this claim. For the turning-point set {|u|≤Au*}, a natural piece decomposition would give many pieces along the circle, and then an ℓ^p summation would multiply the bound by a positive power of λ — potentially ruining the endpoint. I suspect the claim may be true if the pieces are chosen with polar symmetry or if the ℓ∞_ρ norm is taken over pieces that are the whole annuli, but the paper does not say so. This needs to be fixed before the proof can be accepted.\n\nThere is also a minor sign error in Lemma 4.2: the second derivative of the phase difference has the sign of s−t, not t−s. That is harmless because only constant sign is used, but it should be corrected. The definition of r(u) seems clear to me (it is given in Section 3), though repeating it near first use would help the reader.\n\nWho is this for? People working on eigenfunction estimates and harmonic analysis will find the technique valuable. It deserves a serious referee — not because the theorem is new, but because the independent proof is substantive and the Section 2 issue is exactly the kind of thing a knowledgeable referee should check. I would send it to a referee who knows the Koch–Tataru paper and the concurrent Jeong–Lee–Ryu work.","headline":"Independent proof of a theorem already proved elsewhere; the argument is mostly solid but has an unverified piece-counting claim in Section 2.","tokens_in":14764,"tokens_out":6720,"would_cite":true,"duration_ms":71600,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B99","42C10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For the two-dimensional Hermite operator, the spectral projection at eigenvalue λ=2N+2 maps L² to L^{10/3} with norm O(λ^{-1/10}) and no logarithmic factor.","keywords":["Hermite operator","spectral projection","endpoint eigenfunction estimates","critical exponent","Liouville–Green representation","van der Corput estimates","weighted TT* argument","L^{10/3} bounds"],"falsifier":"One concrete check: on a single turning-point annulus with μ between C_* $λ^{{-2/3}}$ and u₀, evaluate ||1_{A_in_μ} Π_λ f||_{$L^{{10/3}}$} for highly concentrated data f; if this ratio ever exceeds C $λ^{{-1/10}}$ ||f||₂ for arbitrarily large λ, the localized input fails and the proof cannot close. More directly, any sequence with ||Π_λ f||_{$L^{{10/3}}$} / ||f||₂ ≥ c $λ^{{-1/10+ε}}$ would disprove the theorem.","tokens_in":13776,"feed_emoji":"📐","tokens_out":8807,"duration_ms":88269,"temperature":0.7,"pith_summary":"The paper establishes the sharp endpoint eigenfunction estimate for the Hermite operator H=-Δ+|x|² on R²: for every eigenvalue λ=2N+2, the spectral projection Π_λ satisfies ||Π_λ f||_{$L^{{10/3}}$(R²)} ≤ C $λ^{{-1/10}}$ ||f||_{L²(R²)}. The exponent $λ^{{-1/10}}$ is optimal; the highest-angular-momentum eigenfunction shows that it cannot be improved. The earlier annulus-by-annulus approach gave only $λ^{{-1/10}}$(log λ)^{3/10} at the critical exponent, and the paper removes the logarithmic factor by coupling all radial scales rather than summing dyadic annuli independently. A weighted TT* argument, applied to the radial modes from the polar spectral decomposition, is the mechanism that performs the coupling.","feed_headline":"Hermite eigenfunction decay hits optimal power, no log loss","feed_subtitle":"For the Hermite operator in two dimensions, the L2-to-L10/3 projection norm is O(lambda^{-1/10}) at every eigenspace.","key_machinery":"The central mechanism is the polar spectral decomposition g(r,θ)=(2π)^{-1/2} ∑_{m∈M_N} c_m R_{N,m}(r)$e^{{imθ}}$, with the mode-dependent turning parameter ν_m defined by ν_m(1-ν_m)=(m²-1/4)/λ². For non-glancing modes, a uniform Liouville–Green representation writes each radial mode as an amplitude times $e^{{±i S(u,m)}}$, and the curvature lemma says the second derivative of the phase difference between two radial scales is comparable to |t-s|. Feeding this curvature into a weighted discrete van der Corput estimate and then into a weighted TT* proposition couples all radial scales with kernel |t-s|^{-1/5}, bypassing the logarithmic loss from dyadic summation.","core_discovery":"On its own terms, the paper proves Theorem 1.1: there is an absolute constant C such that for every λ in 2N+2 and every f in L²(R²), the projection onto the λ-eigenspace obeys ||Π_λ f||_{$L^{{10/3}}$(R²)} ≤ C $λ^{{-1/10}}$ ||f||_{L²(R²)}. The proof treats the interior collar C_* $λ^{{-2/3}}$ << u << 1, where u=1-|x|²/λ, by expanding g=Π_λ f in the polar Laguerre-Hermite basis. Each angular mode has a turning point at u=ν_m; modes with u >> ν_m are non-glancing and are represented by a Liouville–Green oscillatory phase, then controlled by a weighted exponential-sum estimate. Modes near their turning point or deep in the forbidden region are handled by localized spectral projection bounds and by rapid decay. The five resulting estimates cover R² and combine to give the theorem. Sharpness is shown by the eigenfunction G_N(x₁,x₂)=(x₁+i x₂)^N $e^{{-|x|²/2}}$/√(π N!), whose $L^{{10/3}}$ norm is approximately $N^{{-1/10}}$.","pith_inferences":["Inference: Because the TT* kernel estimate only uses curvature of the radial phase difference, the same argument could remove logarithmic losses at critical exponents for radial potentials with a single turning circle, such as |x|^{2k} generalizations of the Hermite operator.","Inference: The sharpness example concentrates on the turning circle, suggesting that improved local inequalities or restriction-type estimates for eigenfunctions may hold on that circle with the same λ^{-1/10} rate.","Inference: The boundedness of the angular Fourier multiplier via the M. Riesz theorem suggests the argument may extend to other orthonormal systems whose angular projections are uniformly bounded in L^p on the circle.","Inference: A testable extension is to perturb the Hermite operator by a bounded potential; if the localized annulus bounds remain true with the same powers, the same coupling would give the endpoint estimate for the perturbed eigenfunctions."],"forward_implications":["At the critical exponent p=10/3 in dimension two, the Hermite spectral projection has the sharp λ^{-1/10} decay with no logarithmic factor.","The highest-angular-momentum eigenfunction G_N is extremal up to constants, so the estimate cannot be improved.","The proof gives a scale-coupling route to the endpoint bound that does not require recursive space-time localization or iterative almost-orthogonality.","The exterior annulus estimate remains summable, and the five-region decomposition shows the full eigenspace projection decays at the endpoint rate."],"supporting_citations":[{"why":"Supplies the localized spectral projection bounds on turning-point, interior, and exterior annuli that are the external input for the annulus estimates.","marker":"[12]"},{"why":"Provides the Laguerre normalization and Hecke–Bochner formula underlying the polar spectral basis.","marker":"[20]"},{"why":"Supplies the discrete van der Corput estimate used in the weighted exponential-sum lemma.","marker":"[8]"},{"why":"Supplies the M. Riesz theorem used to bound the angular Fourier projection multiplier.","marker":"[7]"},{"why":"Supplies the Liouville–Green (WKB) approximation construction used for the radial modes.","marker":"[16]"},{"why":"Provides error bounds for the Liouville–Green approximation used in the radial representation.","marker":"[17]"},{"why":"Supplies the Laguerre orthogonality relation used to normalize the radial modes.","marker":"[18]"}],"fun_headline_variants":["2D Hermite eigenfunctions hit optimal L^10/3 decay","No log loss in Hermite projection bound for 2D","Sharp Hermite estimate: L^2 to L^10/3 is lambda^-1/10","Optimal endpoint bound for 2D Hermite spectral projections"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the localized spectral projection bounds taken as input: on each turning-point annulus of width comparable to $λ^{{-2/3}}$, the L²-to-$L^{{10/3}}$ norm is $λ^{{-1/10}}$, and on interior annuli the L²-to-L² norm is $μ^{{1/4}}$; if those localized bounds were weaker, the dyadic summation and near-mode control would not close.","fun_headline_variants_meta":{"raw":{"variants":["2D Hermite eigenfunctions hit optimal L^10/3 decay","No log loss in Hermite projection bound for 2D","Sharp Hermite estimate: L^2 to L^10/3 is lambda^-1/10","Optimal endpoint bound for 2D Hermite spectral projections"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000735,"raw_usage":{"total_tokens":3278,"prompt_tokens":930,"completion_tokens":2348,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":546,"completion_tokens_details":{"reasoning_tokens":2267}},"tokens_in":546,"tokens_out":2348,"duration_ms":15570,"temperature":1.0,"reasoning_tokens":2267,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:37:16.009999+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"One concrete check: on a single turning-point annulus with μ between C_* $λ^{{-2/3}}$ and u₀, evaluate ||1_{A_in_μ} Π_λ f||_{$L^{{10/3}}$} for highly concentrated data f; if this ratio ever exceeds C $λ^{{-1/10}}$ ||f||₂ for arbitrarily large λ, the localized input fails and the proof cannot close. More directly, any sequence with ||Π_λ f||_{$L^{{10/3}}$} / ||f||₂ ≥ c $λ^{{-1/10+ε}}$ would disprove the theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the discrete van der Corput estimate used in the weighted exponential-sum lemma."},{"cited_title":"Grafakos,Classical Fourier Analysis, third edition, Graduate Texts in Mathematics, Vol","cited_arxiv_id":null,"evidence_quote":"Supplies the M. Riesz theorem used to bound the angular Fourier projection multiplier."},{"cited_title":"Koch and D","cited_arxiv_id":null,"evidence_quote":"Supplies the localized spectral projection bounds on turning-point, interior, and exterior annuli that are the external input for the annulus estimates."},{"cited_title":"Thangavelu,Lectures on Hermite and Laguerre Expansions, Mathematical Notes, Vol","cited_arxiv_id":null,"evidence_quote":"Provides the Laguerre normalization and Hecke–Bochner formula underlying the polar spectral basis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides error bounds for the Liouville–Green approximation used in the radial representation."},{"cited_title":"Szegő,Orthogonal Polynomials, fourth edition, AMS Colloquium Publications, Vol","cited_arxiv_id":null,"evidence_quote":"Supplies the Laguerre orthogonality relation used to normalize the radial modes."}],"review_version":1}