{"id":"1b0eee20-4ace-449b-ab64-840a94dfe75e","arxiv_id":"1909.02118","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Commutators of Hilbert transforms along parabolic and monomial curves are bounded when the symbol lies in a curve-adapted BMO space, and a new testing BMO space gives a partial converse.","lead":"This paper proves that commutators of the parabolic Hilbert transform with a symbol in a parabolic BMO space are bounded on L^p, and introduces a new testing BMO condition for the converse. It also shows curvature matters: for Hilbert transforms along lines, the analogous BMO spaces only overlap but neither contains the other.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The lower-bound proof claims E_Q is a rectangle; the swept region is not, so the stated verification of |E_Q| ~ |Q| in (3.3) is invalid as written.","rationale":"The reader's conditional verdict is appropriate: the main theorems are plausible and the proof strategy is standard, but there are gaps. The most concrete, checkable gap I found is the false geometric assertion in the proof of Theorem 1.2: E_Q is not a rectangle, and the displayed dimensions are dimensionally inconsistent. This directly affects the claimed uniform bound in the testing-BMO lower bound. However, the underlying area estimate is true and provable by a short coarea computation, so the theorem survives and the gap is repairable. I also agree with the reader's stated concern about the parabolic transfer of the Bényi and Hytönen-Perez-Rela lemmas in Section 2; that is the main risk for the upper bound, but it is a standard extension and not contradicted by anything in the paper. I did not find a fatal flaw in the Cauchy integral trick itself, and the lower-bound strategy is sound once (3.3) is established correctly. Because all identified issues are addressable gaps rather than demonstrated false conclusions, the reader's CONDITIONAL verdict should stand unchanged.","tokens_in":12076,"tokens_out":34657,"duration_ms":358716,"concrete_test":"Take Q = [0,1] x [0,1] (so l=1). For each u in [8,10], write y1 = -u and compute the length L(u) of the set of t for which x1 = t-u lies in [0,1]; the vertical fiber is the union over t in that set of intervals [-t^2, 1-t^2]. Show the fiber length is (u+1)^2 - 80 for u in [8,9] and 101 - u^2 for u in [9,10]. Integrate to get |E_Q| = 21, and compare to the paper's rectangle area 40. This verifies that (3.3) holds, but only via a corrected coarea computation, not the stated rectangle argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 3, after defining E_Q = {x - gamma(t) : x in Q, 9l(Q) <= t <= 10l(Q)}, the proof says 'by considering the flows of the corners ... we can explicitly compute E_Q: it is a (non-parabolic) rectangle whose base length is ... and whose height is ...' and concludes (3.3), |E_Q| ~ |Q|. This assertion is false. For Q = [0,l] x [0,l^2], the set E_Q is not a rectangle: over the first-coordinate value -8l, the only admissible t is t = 9l, so the vertical fiber has length l^2, whereas over -9l the fiber has length about 20l^2. The true region is a swept parallelogram-like set with triangular end fibers. Moreover, the displayed height has the wrong units: it should be 20l^2, not 20l, so the rectangle-area computation does not even give the claimed comparison. Theorem 1.2 relies on (3.3) to make the constant in the testing-BMO lower bound independent of Q, so this is a genuine gap in the proof of the necessity half of the central chain. The estimate itself is still true: for l = 1 the exact area is 21, still comparable to |Q| = 1. Thus the theorem is likely repairable, but the written proof of (3.3) is not correct.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies commutators [b,H_γ] of the parabolic Hilbert transform along γ(t)=(t,t^2), proving an upper bound when the symbol b lies in a parabolic BMO space (Theorem 1.1) and a lower bound in terms of a newly introduced 'testing' BMO space adapted to the curve (Theorem 1.2). The upper bound is obtained via the Coifman–Rochberg–Weiss Cauchy integral trick combined with the sparse domination theorem of Cladek and Ou and weighted estimates for H_γ. The lower bound is derived by testing the commutator on characteristic functions of parabolic cubes and their flow sets E_Q. The results are then extended to monomial curves in R^n (Theorems 4.1 and 4.2) and local torsion-free curves (Theorem 4.3), and the final section contrasts the parabolic setting with Hilbert transforms along straight lines.","tokens_in":12338,"tokens_out":9488,"duration_ms":94523,"significance":"If correct, the paper provides a natural BMO-type characterization for commutators of a genuinely non-Calderón–Zygmund singular integral, which is a valuable contribution to the harmonic analysis of Hilbert transforms along curves. The use of the Cauchy integral trick and Cladek–Ou sparse domination is methodologically sound, and the introduction of a testing BMO space is a useful new idea that gives a concrete necessary condition. The paper is clearly written and the main structural claims are plausible. However, two load-bearing points in the written proofs need repair: the geometric computation of the flow set E_Q in Section 3 is incorrect as stated, and the transfer of two classical weighted/BMO results to the parabolic-cube setting in Section 2 is asserted but not proved. These issues are likely fixable, but they currently prevent the central theorems from being fully established.","major_comments":[{"comment":"The assertion that E_Q is a rectangle is not correct, and the displayed area computation has wrong units. For Q=[0,ℓ]×[0,ℓ^2], the set E_Q={x−γ(t): x∈Q, 9ℓ≤t≤10ℓ} is a swept region whose vertical fiber over a fixed first coordinate u is an interval of length ℓ^2+(t_max^2−t_min^2), where t_min and t_max are the endpoints of the admissible t-range; the region has triangular end fibers and is not a rectangle. In particular, the claimed height 20ℓ should be of order 20ℓ^2. The comparison |E_Q|∼|Q| is nevertheless true — for ℓ=1, integrating the fiber lengths gives area 21 — so Theorem 1.2 is likely repairable, but the proof of (3.3) as written must be replaced by a correct computation.","section":"Section 3, Eq. (3.3)"},{"comment":"The proof of Theorem 1.1 depends on two unproved assertions: that the exponentiation lemma of Bényi, Martell, Moen, Stachura, and Torres [1, Lemma 3.5] holds for parabolic cubes with the same constant 4, and that the reverse Hölder theorem of Hytönen, Pérez, and Rela [6, Theorem 2.3] 'immediately translates' to the parabolic setting. These transfers are used to obtain the uniform bounds on [w]_{A_{2/r}} and [w]_{RH_{(s'/2)'}} that feed into the weighted estimate (2.3); without them, the upper bound in Theorem 1.1 is not established. Please provide either a proof of the parabolic versions or a precise reference where they appear.","section":"Section 2, Eqs. (2.4)–(2.5)"},{"comment":"In the passage leading to (2.7), the mixed characteristic is written as [w]_{A_{p/r}}[w]_{A_{(s'/p)'}}, but the surrounding text describes a reverse Hölder characteristic, and the L^2 estimate in (2.3) uses [w]_{A_{2/r}}[w]_{RH_{(s'/2)'}}. As written, the displayed formula is inconsistent with the argument, and the parameter selection cannot be checked. Please state the correct Cladek–Ou L^p estimate and make the notation align with the reverse Hölder characteristic used in the text.","section":"Section 2, L^p case, around Eq. (2.7)"}],"minor_comments":[{"comment":"The vertices of the acceptable parameter triangle are listed inconsistently: the L^2 case states (0,0), (1,0), (2/3,1/3), while the L^p case states (0,0), (1,1), (2/3,1/3). Please clarify which set of vertices is correct.","section":"Section 2, parameter triangle"},{"comment":"In the proof of Proposition 3.2, the assertion that one can find a small parabolic cube Q̃⊆Q with E_{Q̃}⊆R is stated without justification. Since this is the key geometric input of the proof, please provide a short argument or a reference.","section":"Section 3, Proposition 3.2"},{"comment":"The parameter ǫ is introduced first as the radius of the integration contour and later redefined as 2/r−1 or p/r−1. This reuse of notation is confusing; consider using different symbols for the contour radius and the exponent gap.","section":"Section 2, Cauchy integral trick"},{"comment":"Reference [1] is cited as an arXiv preprint; if a published version now exists, it would be helpful to update the citation.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses a timely and interesting question, and the overall strategy is credible. The two main gaps — the geometric computation of E_Q and the unproved transfer of weighted/BMO results to parabolic cubes — are localized and seem repairable within the scope of the paper. If the authors can supply a correct proof of (3.3) and either prove or properly reference the parabolic versions of the exponentiation and reverse Hölder results, I would support publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nQuick take on Bongers–Guo–Li–Wick, arXiv:1909.02118. The paper does something genuinely new: it proves L^p commutator bounds for the parabolic Hilbert transform H_γ when the symbol lies in the parabolic BMO space BMO_γ, and introduces a testing-BMO condition that is necessary for L^2 boundedness. The upper bound uses the Cauchy integral trick on top of Cladek–Ou sparse domination and weighted estimates; that part is a clean, standard template, and it works. The testing-BMO lower bound is the more original piece, and the line-versus-curve comparison in Section 5 is a nice observation that curvature is essential. No circular reasoning anywhere.\n\nThe main soft spot is the proof of (3.3), the estimate |E_Q| ~ |Q|. The paper claims E_Q is a rectangle with base 2ℓ and height 20ℓ. That is not right: E_Q is a swept region, not a rectangle, and the height should be 20ℓ², not 20ℓ, so the displayed computation does not give the claimed comparison. The estimate itself is still true—for ℓ=1 the exact area is 21—so the theorem is likely repairable, but the written proof of that key step is incorrect. The referee should insist on a correct geometric verification.\n\nTwo smaller issues. The transfer of the BMO-to-A_p exponentiation lemma (Bényi et al.) and the reverse-Hölder theorem (Hytönen–Pérez–Rela) to parabolic cubes is asserted with “only a change of constants” but not proved. That is probably fine, but it should be written out or given a precise reference. Proposition 5.3, the existence of a planar BMO function with unbounded fiberwise BMO, is only sketched; the construction of ψ with slowly blowing-up averages is handwavy. Again, plausible, but needs a real proof.\n\nWho is this for? Harmonic analysts working on singular integrals along curves, commutators, and weighted theory. It deserves a serious referee: the main theorems are new, the approach is sound overall, and the gaps are addressable. I would send it to review, and I'd expect the authors to fix the geometric issue and fill in the parabolic-cube lemma transfers.\n\nRegards.","headline":"Genuinely new commutator bounds for parabolic Hilbert transforms; upper bound clean, lower bound has a repairable but real geometric gap.","tokens_in":12859,"tokens_out":3501,"would_cite":true,"duration_ms":31902,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B20","42B25","42B35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that commutators of Hilbert transforms along monomial curves are bounded on L^p whenever the symbol belongs to the corresponding non-isotropic BMO space, and gives a partial converse through a new testing BMO space.","keywords":["commutators","Hilbert transform along curves","parabolic BMO","testing BMO","sparse domination","Muckenhoupt weights","reverse Holder","monomial curves"],"falsifier":"Find a parabolic-cube weight w with [w]_{A_{2/r}} bounded by a fixed constant but [w]_{RH_{1+$\\sigma$}} arbitrarily large for a fixed small $\\sigma$; that would falsify the transferred reverse-Holder estimate (2.5) used in the proof. Alternatively, construct b in testing BMO for the parabola for which [b,H_gamma] fails to be bounded on $L^{2}$, which would show the lower-bound theorem cannot be reversed.","tokens_in":11877,"feed_emoji":"📐","tokens_out":4835,"duration_ms":44545,"temperature":0.7,"pith_summary":"The paper establishes that the commutator of a Hilbert transform taken along a curved monomial path, such as a parabola (t, $t^{2}$), is controlled by a natural parabolic version of BMO: if the symbol b has bounded mean oscillation over parabolic cubes, then [b,H_gamma] is bounded on L^p($R^{2}$) for every 1