{"id":"b7512f88-4268-4984-a040-b6b738774846","arxiv_id":"1908.01448","paper_version":5,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The Hardy space H^1_{FIO}(R^n) for Fourier integral operators equals its Littlewood-Paley g-function, maximal-function, and G*_alpha versions for alpha > 2, with equivalent norms.","lead":"Mathematicians proved that a function space used to study wave equations can be measured in three equivalent ways, settling an open question from earlier work. The new descriptions make the space easier to use in harmonic analysis and in the study of wave propagation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The hard inclusion H^1_FIO,G ⊆ H^1_FIO turns on the off-singularity kernel bound of Corollary 2.3, which is imported without proof from [15, Prop 3.6] and [12, Thm 5.1]; this is the single load-bearing external input, and it is plausible but unverified within the paper.","rationale":"The paper gives several equivalent characterizations of H^1_FIO, answering [15, Remark 4.3]. I traced the main chain: Proposition 3.1 reduces H^1_FIO,G to the Hardy-space property of φ_ω(D)f; Proposition 3.2 supplies the Sobolev embedding into W^{-(n-1)/4,1}; Lemmas 3.4–3.5 and Proposition 3.6 convert the square function into a Peetre-type maximal estimate; Theorem 3.8 then proves the g-function characterization, with Theorem 4.1 and Theorem 5.2 following by classical H^1 characterizations and the change-of-aperture lemma. The internal inequalities check out: the annulus sums in (3.21) converge because αr>n, the vector-valued Hardy–Littlewood inequality applies with p=1/r>1, and the geometric series in Theorem 5.2 requires exactly α>2. The one place the manuscript leans on an unproved input is Corollary 2.3, the off-singularity kernel bound with fast dyadic decay; Proposition 3.6 would fail without it. This is flagged explicitly in the paper as a proof deferred to [15, Proposition 3.6] and [12, Theorem 5.1]. Since those bounds are published and the arguments are said to rely only on the wave-packet estimates established in Lemma 2.2, I treat this as a verification burden rather than a demonstrated flaw. My read therefore does not change the reader's ACCEPT verdict; it strengthens the recommendation to check the cited off-singularity bound independently.","tokens_in":29175,"tokens_out":14429,"duration_ms":132505,"concrete_test":"Independently re-derive the bound in Corollary 2.3 from Lemma 2.2, following the integration-by-parts argument in [15, Proposition 3.6] (or [12, Theorem 5.1]). Concretely, compute F^{-1}(θ_{ω,σ}χ_{ν,τ})(x−y) and verify that for every N there is C_N, uniform in ω,ν,σ,τ,x,y, with |K^{ω,ν}_{σ,τ}(x,y)| ≤ C_N min(σ/τ,τ/σ)^N ρ^{-n}(1+ρ^{-1}d(x,ω;y,ν)^2)^{-N}, where ρ=min(σ,τ). If the proof goes through, the lacunary estimate in Proposition 3.6 is sound; if any step fails, test the specific case σ=2^{-l}, τ=2^{-j} used there and check whether the j-summation in (3.21) still converges with the exponent available.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.4 is built on Theorem 3.8, and the hard inclusion H^1_FIO,G(R^n) ⊆ H^1_FIO(R^n) is proved through Proposition 3.6. Proposition 3.6 uses Corollary 2.3 to control the kernel K^{ω,ν}_{σ,τ} of θ_{ω,σ}(D)χ_{ν,τ}(D); the factor min(σ/τ,τ/σ)^N makes the lacunary series over j converge, and the off-diagonal factor (1+ρ^{-1}d^2)^{-N} feeds the Hardy–Littlewood maximal estimate (3.21). Corollary 2.3 is stated without proof here; the proof is said to repeat [15, Proposition 3.6] and [12, Theorem 5.1]. These are prior works by the same author group, not independently checkable inside this manuscript. If the decay in min(σ/τ,τ/σ) were weaker than N for every N, or if the bound had an additional σ,τ-dependent constant not absorbable into C_N, the summation in Proposition 3.6 would diverge and the embedding would fail. I found no internal error in the traced estimates of Sections 3–5; the concern is therefore about dependence on a published but not reproved bound, not about an internal inconsistency.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Hardy space H^1_FIO(R^n) introduced by Smith and by Hassell, Portal, and Rozendaal for Fourier integral operators. Its main result (Theorem 1.4) asserts that, for alpha > 2, H^1_FIO(R^n) coincides, with equivalent norms, with the Littlewood--Paley g-function space H^1_FIO,G(R^n), the maximal-function space H^1_FIO,max(R^n), and the Lusin-type square-function space H^1_FIO,G*_alpha(R^n). The proof of the hardest inclusion H^1_FIO,G subset H^1_FIO is carried out in Section 3 via a Peetre-type maximal function, a technical lemma of Rychkov type, off-singularity kernel bounds, and vector-valued Hardy--Littlewood maximal estimates. Section 4 gives the maximal-function characterization, Section 5 gives the G*_alpha characterization, and Section 6 applies these results to Fourier multiplier boundedness on H^p_FIO and to sharp norm comparisons for frequency-localized functions. The paper answers an open question posed in [15, Remark 4.3].","tokens_in":29288,"tokens_out":27616,"duration_ms":257441,"significance":"If correct, the result completes the p = 1 case of the characterization program started in [15] for 1 < p < infinity and provides practically useful descriptions of H^1_FIO in terms of more classical harmonic-analysis objects. The proof is nontrivial, and the main chain of estimates is internally consistent: I checked the support decomposition in Proposition 3.6, the application of Corollary 2.3, and the summation and maximal-function arguments in Theorem 3.8. The applications in Section 6 are reasonable first demonstrations of the utility of the new characterizations. The paper is also transparent about its dependence on the authors' earlier work for the off-singularity kernel bound; the reference is precise and the proof strategy is indicated, so I do not regard this as an internal gap.","major_comments":[],"minor_comments":[{"comment":"The off-singularity bound (2.12) is the single most load-bearing estimate in the paper, since it drives the lacunary summation in Proposition 3.6, but its proof here is only a reference to [15, Proposition 3.6] and [12, Theorem 5.1]. I suggest including the full proof or at least a precise statement of the cited result, so that the present paper is self-contained at this key point; the current sentence \"repeat the arguments\" is acceptable as a citation but would benefit from a short explanation of the mechanism.","section":"Section 2.3, Corollary 2.3"},{"comment":"In the display following the choice of r and N, the exponent on 2^{-|j-l|} is written as N, but Proposition 3.6's proof produces the exponent (N-alpha)r before the final relabeling of N. This is not an error, but the notation should be clarified by using distinct constants (for example N_0 and N) to avoid confusing the reader.","section":"Section 3.2, Theorem 3.8"},{"comment":"The adjoint computation for m(D) is difficult to read in the current typesetting (the display appears as \"m(D)^*g(x) = m(D)g(x) = m(D)\\tilde g(-x)\"). Please state explicitly the convention for the distributional duality pairing used in (1.2) and rewrite the identity in a way that is unambiguous about complex conjugation and the action on g.","section":"Section 6.1, Theorem 6.1"},{"comment":"The use of V(B_\\sqrt{\\sigma}(x,omega)) \\simeq \\sigma^n may surprise readers, since Lemma 2.1 states that balls of radius tau < 1 have volume of order tau^{2n}. A brief parenthetical noting that the radius here is \\sqrt{\\sigma} and hence the volume is (\\sqrt{\\sigma})^{2n} = \\sigma^n would remove any ambiguity.","section":"Section 5, Theorem 5.2"},{"comment":"The preprint typesetting contains several typographical artifacts, including \"m aximal\" in the abstract, \"\\greaterorsimilar\" and \"\\nelement\" in place of symbols, and assorted spacing issues. Please ensure the final published version is typeset cleanly.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"The paper is well within the scope of the journal and the central result is significant. The only issue I would flag for the editor is the paper's reliance on the authors' own prior work for the off-singularity bound in Corollary 2.3; if the editorial policy requires proofs to be self-contained at load-bearing points, the authors should be asked to add an appendix or a detailed proof sketch. I do not see grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good paper, worth refereeing. It answers the p=1 endpoint left open in Rozendaal's paper, and it does so with a method that is not just a re-run of the duality proof used for 1<p<∞. The direct argument through the Peetre-type maximal function is the right tool: at p=1, duality against H^∞_FIO is unavailable, and the paper avoids it. I traced the decomposition in Proposition 3.6, the use of Corollary 2.3, the annulus estimates behind (3.21), and the aperture change in Theorem 5.2. Everything I checked is internally consistent, and the α>2 condition comes honestly from a convergent geometric series.\n\nWhat is new: Theorem 1.4 gives the g-function, maximal-function, and G*_alpha characterizations of H^1_FIO for α>2, with equivalence of norms. This is genuinely new at p=1, and the comparison with [15, Remark 4.3] is accurate. Section 6 adds a multiplier theorem and some explicit norm computations; these are useful but illustrative, not the main event. The citation pattern is healthy: the paper leans on [12] and [15] because it extends them, not for padding.\n\nSoft spots, in proportion: the one genuinely load-bearing input not proved here is Corollary 2.3, the off-singularity kernel bound, whose proof is delegated to [15, Prop 3.6] and [12, Thm 5.1]. That is prior work by the same group, currently in press or preprint, and the proof is said to follow from Lemma 2.2 by integration by parts. The paper is honest about this, and I do not see evidence of circularity or a missing hypothesis. Still, a careful referee should verify that the dyadic decay in min(σ/τ,τ/σ)^N is exactly what those references supply, since the lacunary summation in Proposition 3.6 collapses without it. This is a real dependency, not a flaw in the argument. The stress-test concern about Corollary 2.3 is fair to flag, but on reading the paper it does not land as a fatal objection. Minor: p=∞ is left open in Remark 4.2, and the authors say so explicitly; that is the natural boundary of the method, not a defect.\n\nWho is this for? Harmonic analysts working on Fourier integral operators and adapted Hardy spaces. It deserves a serious referee; I would send it to peer review rather than desk reject, and if the imported bound checks out, it should be published. I would cite it.","headline":"Solid endpoint result answering the p=1 Hardy-space question for FIOs, with a clean direct proof and only one load-bearing imported bound to watch.","tokens_in":30080,"tokens_out":2897,"would_cite":true,"duration_ms":30081,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B35","35S30","42B30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the Hardy space for Fourier integral operators on R^n is characterized by Littlewood–Paley g functions, maximal functions, and Lusin-type square functions, with equivalent norms.","keywords":["Fourier integral operators","Hardy spaces","Littlewood-Paley g function","maximal function characterization","wave packets","conical square function","parabolic frequency localization","Lusin area function"],"falsifier":"A concrete way to test the central claim is to exhibit $f\\in \\mathcal{S}'(\\mathbb{R}^n)$ with $q(D)f\\in L^1$ and $G(f)\\in L^1(S^*(\\mathbb{R}^n))$ but with the conical square function $S(f)$ infinite; Theorem 3.8 says no such $f$ exists, so one would refute the main theorem. Alternatively, compute the kernel bound (2.12) explicitly for $n=2$ with $\\sigma=2^{-k}$ and $\\tau=2^{-l}$, and look for a failure of the $\\min(\\sigma/\\tau,\\tau/\\sigma)^N$ decay, which would break the engine of the proof.","tokens_in":28775,"feed_emoji":"📐","tokens_out":8996,"duration_ms":83930,"temperature":0.7,"pith_summary":"The paper proves that $H^1_{FIO}(\\mathbb{R}^n)$, the Hardy space adapted to Fourier integral operators, can be described by tools from classical harmonic analysis. Specifically, an $f$ lies in $H^1_{FIO}(\\mathbb{R}^n)$ exactly when its low-frequency part $q(D)f$ is in $L^1$ and, for almost every direction $\\omega$, the parabolically localized piece $\\phi_\\omega(D)f$ lies in the classical Hardy space $H^1(\\mathbb{R}^n)$, with finite integral of those $H^1$ norms over the sphere. The same conclusion holds for the Littlewood–Paley $g$-function description, for a maximal-function description, and for the Lusin-type square function $G^*_\\alpha$ when $\\alpha>2$, all with equivalent norms. This answers an open question from earlier work on these spaces, and it matters because the new descriptions make $L^1$-based harmonic analysis tools available for studying Fourier integral operators and wave equations with rough coefficients.","feed_headline":"Littlewood–Paley tools now characterize the FIO Hardy space","feed_subtitle":"Three norm-equivalent descriptions—g function, maximal function, area function—open L^1 harmonic analysis for Fourier integral operators.","key_machinery":"The machinery consists of wave packet projections $\\theta_{\\omega,\\sigma}(D)$, $\\chi_{\\omega,\\sigma}(D)$, and $\\phi_\\omega(D)$, where $\\theta_{\\omega,\\sigma}$ localizes to frequencies of size $\\sigma^{-1}$ inside a parabolic cone around the direction $\\omega$. The hard inclusion is carried by a Peetre-type maximal function $M^*_\\alpha$ and a lacunary decomposition identity. The load-bearing estimate is Corollary 2.3: the kernel of $\\theta_{\\omega,\\sigma}(D)\\chi_{\\nu,\\tau}(D)$ decays like $\\min(\\sigma/\\tau,\\tau/\\sigma)^N$ for every $N$, which makes dyadic-lacunary sums converge. A geometric-convolution lemma (Lemma 3.4) converts power-type inequalities into $L^r$ estimates, Lemma 3.7 controls the resulting weighted sums, and the vector-valued Hardy–Littlewood maximal operator completes the argument.","core_discovery":"The central discovery is that, for $p=1$, the Hardy space for Fourier integral operators $H^1_{FIO}(\\mathbb{R}^n)$ coincides with $H^1_{FIO,G}(\\mathbb{R}^n)$, $H^1_{FIO,\\max}(\\mathbb{R}^n)$, and $H^1_{FIO,G^*_\\alpha}(\\mathbb{R}^n)$ for $\\alpha>2$, with equivalence of norms. Equivalently, an $f$ belongs to $H^1_{FIO}(\\mathbb{R}^n)$ if and only if $q(D)f$ is in $L^1(\\mathbb{R}^n)$, $\\phi_\\omega(D)f$ is in the classical Hardy space $H^1(\\mathbb{R}^n)$ for almost every $\\omega\\in S^{n-1}$, and the integral over $S^{n-1}$ of $\\|\\phi_\\omega(D)f\\|_{H^1}\\,d\\omega$ is finite. The difficult inclusion $H^1_{FIO,G}\\subseteq H^1_{FIO}$ is proved directly rather than by duality, using parabolic frequency localizations and a maximal function of Peetre type.","pith_inferences":["Because the characterization reduces $H^1_{FIO}$ to an average of classical $H^1$ norms of directionally localized pieces, many $L^1$-based harmonic analysis tools—maximal inequalities, Riesz transforms, interpolation—can be transplanted to Fourier integral operator problems without re-proving them for the conical square function.","A natural test is whether the endpoint $p=\\infty$ admits a BMO description of the same shape; the paper proves only the easy inclusion there, so the converse is a plausible but unproved extension.","The same off-singularity decay that powers the proof suggests the characterizations should survive for more general phases or rough-coefficient wave equations, provided the kernel bound remains valid.","The explicit norm computation for wave packets makes these spaces amenable to numerical checks: one can approximate $\\phi_\\omega(D)f$ and compare the directional $H^1$ average against the conical square function on test functions."],"forward_implications":["The conical square function in the definition of $H^1_{FIO}$ can be replaced by the simpler vertical Littlewood–Paley $g$ function, giving $H^1_{FIO}(\\mathbb{R}^n)=H^1_{FIO,G}(\\mathbb{R}^n)$ with comparable norms.","An $f$ belongs to $H^1_{FIO}(\\mathbb{R}^n)$ precisely when its parabolically localized pieces $\\phi_\\omega(D)f$ are in $H^1(\\mathbb{R}^n)$ for almost every direction and the average of their $H^1$ norms over the sphere is finite.","For $\\alpha>2$, the Lusin-type square function $G^*_\\alpha$ gives a norm-equivalent description, so changes of aperture are harmless in $H^1_{FIO}$.","Any bounded Fourier multiplier that maps the local Hardy space $H^1(\\mathbb{R}^n)$ to $L^1(\\mathbb{R}^n)$ is automatically bounded on $H^p_{FIO}(\\mathbb{R}^n)$ for every $1\\le p\\le \\infty$, extending earlier smooth-symbol results.","For functions with frequency support in a dyadic-parabolic region, the $H^p_{FIO}$ norm is equivalent to a shifted Sobolev norm, and the shift $s_p=\\frac{n-1}{2}|\\tfrac12-\\tfrac1p|$ is optimal."],"supporting_citations":[{"why":"Supplies the vertical-versus-conical square function inequality that gives the easy inclusion $H^1_{FIO}\\subseteq H^1_{FIO,G}$.","marker":"[4]"},{"why":"Introduced the parabolic wave packets and the question for $p=1$ that this paper answers, and provides the off-singularity kernel bound quoted as Corollary 2.3.","marker":"[15]"},{"why":"Gives the original definition and theory of $H^p_{FIO}$, including Sobolev embeddings and boundedness results used throughout.","marker":"[12]"},{"why":"Provides the geometric-convolution lemma (Lemma 3.4) used to turn power-type maximal estimates into $L^r$ bounds.","marker":"[16]"},{"why":"Supplies classical Littlewood–Paley $g$-function and Riesz transform characterizations of $H^1$ used in Propositions 3.1 and Theorem 6.1.","marker":"[20]"},{"why":"Supplies the maximal function characterization of $H^1$ used to prove the maximal function description of $H^1_{FIO}$.","marker":"[11]"}],"fun_headline_variants":["FIO Hardy space equivalent to Littlewood-Paley versions","Littlewood-Paley g and maximal functions characterize FIO H^1","Three norm-equivalent definitions of FIO Hardy space","FIO H^1 pinned down by Littlewood-Paley tools","Answering Rozendaal: FIO Hardy space has multiple equivalent norms"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on an imported, unproved kernel estimate: when two wave-packet projections at frequency scales $\\sigma$ and $\\tau$ are composed, their kernel decays by a fixed power $\\min(\\sigma/\\tau,\\tau/\\sigma)^N$ for every $N$. If that decay were weaker, the lacunary summation driving the main $L^1$ estimate would not converge, and the inclusions would not follow by this route.","fun_headline_variants_meta":{"raw":{"variants":["FIO Hardy space equivalent to Littlewood-Paley versions","Littlewood-Paley g and maximal functions characterize FIO H^1","Three norm-equivalent definitions of FIO Hardy space","FIO H^1 pinned down by Littlewood-Paley tools","Answering Rozendaal: FIO Hardy space has multiple equivalent norms"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00123,"raw_usage":{"total_tokens":5033,"prompt_tokens":902,"completion_tokens":4131,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":4043}},"tokens_in":518,"tokens_out":4131,"duration_ms":29522,"temperature":1.0,"reasoning_tokens":4043,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:17:14.084351+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete way to test the central claim is to exhibit $f\\in \\mathcal{S}'(\\mathbb{R}^n)$ with $q(D)f\\in L^1$ and $G(f)\\in L^1(S^*(\\mathbb{R}^n))$ but with the conical square function $S(f)$ infinite; Theorem 3.8 says no such $f$ exists, so one would refute the main theorem. Alternatively, compute the kernel bound (2.12) explicitly for $n=2$ with $\\sigma=2^{-k}$ and $\\tau=2^{-l}$, and look for a failure of the $\\min(\\sigma/\\tau,\\tau/\\sigma)^N$ decay, which would break the engine of the proof.","supporting_citations":[{"cited_title":"Auscher, S","cited_arxiv_id":null,"evidence_quote":"Supplies the vertical-versus-conical square function inequality that gives the easy inclusion $H^1_{FIO}\\subseteq H^1_{FIO,G}$."},{"cited_title":"Characterizations of Hardy spaces for Fourier integral operators","cited_arxiv_id":"1907.02680","evidence_quote":"Introduced the parabolic wave packets and the question for $p=1$ that this paper answers, and provides the off-singularity kernel bound quoted as Corollary 2.3."},{"cited_title":"Off-singularity bounds and Hardy spaces for Fourier integral operators","cited_arxiv_id":"1811.11376","evidence_quote":"Gives the original definition and theory of $H^p_{FIO}$, including Sobolev embeddings and boundedness results used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the geometric-convolution lemma (Lemma 3.4) used to turn power-type maximal estimates into $L^r$ bounds."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies classical Littlewood–Paley $g$-function and Riesz transform characterizations of $H^1$ used in Propositions 3.1 and Theorem 6.1."},{"cited_title":"Grafakos","cited_arxiv_id":null,"evidence_quote":"Supplies the maximal function characterization of $H^1$ used to prove the maximal function description of $H^1_{FIO}$."}],"review_version":1}