{"id":"a879024b-4eb1-4512-8188-f6831e52e22d","arxiv_id":"2501.11928","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On the Heisenberg group, the dyadic elliptic maximal operator is L^p bounded for all 2-by-2 matrices A except A = cI (one-parameter case) and A = diag(c, c*2^(2a)) (two-parameter case).","lead":"This paper proves exactly when averaging a function over ellipses at dyadic scales, with a Heisenberg-group twist, yields a bounded maximal operator on L^p spaces. It finds a complete classification: boundedness fails only for two precise families of 2-by-2 matrices.","discovery_kind":"new_application","skeptic_critique":null,"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies lacunary elliptic maximal operators E^1_A and E^2_A on the Heisenberg group H^1 identified with R^3, defined as suprema over dyadic scales of averages over ellipses determined by a matrix A. It claims: E^2_J is bounded on L^p(H^1) for 1<p≤∞ for the skew-symmetric matrix J (Theorem 1.2); E^1_A is bounded iff A is not a nonzero scalar multiple of the identity; and E^2_A is bounded iff A is not of the form diag(c, c·2^{2a}) for c≠0 and a∈Z (Theorems 1.1 and 1.3). The strategy is to decompose the elliptic averages by Littlewood-Paley projections, prove L^2 decay estimates for oscillatory integral operators via the group Fourier transform, and then bootstrap to all p>1 using a vector-valued maximal inequality (Lemma 5.1). The extension to arbitrary A is based on a twisting identity (Lemma 6.1) that reduces the analysis to the skew-symmetric Heisenberg case plus additional quadratic phase terms.","tokens_in":19153,"tokens_out":4849,"duration_ms":49930,"significance":"If correct, the result gives a complete and clean characterization of L^p boundedness for these one- and two-parameter dyadic elliptic maximal operators on H^1, going beyond previous work on lacunary spherical maximal functions and identifying exactly which linear distortions break boundedness. The approach is attractive and largely self-contained: it uses explicit computations with the group Fourier transform, Plancherel, van der Corput estimates, and Littlewood-Paley theory, with no fitted parameters. However, the technical core of the proof is not yet fully demonstrated; the major gaps listed below concern precisely the estimates that carry the main theorems, so the significance of the paper depends on closing those gaps.","major_comments":[{"comment":"The central L^2 decay estimate (4.3) is not proved in the full parameter range required. The proof of (4.9) is carried out only under the hypothesis 2^{ℓ1} ≥ 2^{10}(2^{ℓ2} + |λ2^{k1+k2}|). In the complementary regime, where 2^{ℓ1} is comparable to or smaller than 2^{ℓ2} + |λ2^{k1+k2}|, the text provides no estimate; the bound (4.8), used alone, gives only 2^{-s/4} min{2^{-ℓ2/4}, |λ2^{k1+k2}|^{-1/4}}, which does not imply the claimed uniform decay 2^{-(s+ℓ1+ℓ2)/4}. Since (4.3) underpins Proposition 3.2 and hence Theorems 1.2 and 1.3, this is a load-bearing gap and not a mere presentation issue.","section":"§4.2, Eq. (4.9) and Eq. (4.3)"},{"comment":"The bootstrap argument for the full range 1<p<∞ rests entirely on Lemma 5.1, which is introduced with the phrase 'In the spirit of Nagel, Stein and Wainger [18]' but is neither proved in the text nor derived from [18] with the required strength. The lemma asserts a quantitative implication from scalar L^p maximal boundedness to an L^q vector-valued inequality for all q with 1/q < (1/2)(1+1/p); this is a strong statement and it is not an immediate consequence of the cited paper as written. Without a proof of Lemma 5.1, the transition from p>4/3 to all p>1 in Section 5 is unsupported, and with it the stated range of Theorems 1.2 and 1.3.","section":"§5, Lemma 5.1"},{"comment":"The proof of the symmetric case of Theorem 1.3 is incomplete. The case analysis after (7.13), stated as 'When b/d /∈ 22Z and e ≠ 0' versus 'either ... not both', omits several possibilities (for instance b=0 or d=0), and the notation 22Z is ambiguous. In addition, the claimed bound (7.13) uses a minimum of two Fourier decay estimates, while the support condition on the projections yields control only through their maximum; the passage from (7.13) to the square-function estimate (7.14) is not justified. The invocations of the 'Lifting Lemma' from page 484 of [19] and of the shifted maximal operators from [20, 23] are also too terse to verify. Since this subsection is where the sufficiency of Theorem 1.3 is proved for symmetric A, the gap matters.","section":"§7.2, proof of (7.12)"}],"minor_comments":[{"comment":"The label 'Proof of Proposition 7.1' in Section 4.1 is a misprint: the statement being proved there is Proposition 4.1, while Proposition 7.1 appears later in Section 7; the cross-reference should be corrected.","section":"§4.1, opening line"},{"comment":"The definition L^{ν,loc}_k f = ∑∞_k L^ν_j * f has an ambiguous summation limit; it should read ∑_{j≥k} L^ν_j * f, and similarly the limits in the decomposition (3.2) should be stated explicitly.","section":"§2, definition of L^{ν,loc}_k"},{"comment":"The vector-valued norm is written as (∑_{k1,k2} |ζ_K * L^2_{k2-ℓ2} L^1_{k1-ℓ1} f|^2)^{1/2}, but the accompanying sentence says it suffices to estimate sup_K |...|; the connection between the sup-norm and the square function should be made explicit.","section":"§3, display before (3.3)"},{"comment":"In the proof of Proposition 4.1, the sentence 'the second part follows from 2^m ≤ 2^{-s+10}' occurs in Case 2, where the hypothesis is 2^m ≥ 2^{-s+10}; the comparison of the two terms in the minimum of (4.4) should be written out carefully.","section":"§4.1, Case 2"},{"comment":"The notation 'b/d /∈ 22Z' is ambiguous; it presumably means b/d ∉ 2^{2Z}, but the present typesetting makes the condition hard to parse and should be clarified.","section":"§7.2, after (7.13)"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of the journal and the overall strategy is attractive, but the current version has a substantial gap in the key L^2 decay estimate (4.3) and relies on an unproved bootstrap lemma (Lemma 5.1). These issues are localized enough that a serious revision could fix them; I recommend major revision rather than rejection. I would also encourage the authors to provide a self-contained proof of Lemma 5.1 or a precise derivation from the cited literature, since that lemma is the load-bearing mechanism for the full p-range."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere is my read. The paper proves a genuinely new classification: for 2x2 matrices A, the one-parameter lacunary elliptic maximal operator E^1_A is bounded on L^p(H1) unless A is a nonzero scalar multiple of the identity, and the two-parameter E^2_A is bounded unless A is diag(c, c 2^{2a}). That clean statement is not in the earlier radial work, and it includes a surprising unbounded case. The overall architecture is sound: Fourier transform on the Heisenberg group, reduction to oscillatory integral estimates, then a bootstrap. The unboundedness examples are simple and convincing.\n\nThe soft spot the reader flagged is real. The key L2 estimate (4.3) is only proved under a frequency-dominance condition: (4.9) assumes 2^{ℓ1} ≥ 2^{10}(2^{ℓ2} + |λ 2^{k1+k2}|). When that inequality fails, the paper never says what happens. The step from (4.8) to the claimed ℓ1 decay is not justified in that regime. This hole sits under Proposition 3.2, so Theorem 1.2 and the general matrix result rest on it.\n\nSecond, Lemma 5.1, the vector-valued inequality used to push p>4/3 down to p>1, is stated with 'in the spirit of Nagel, Stein and Wainger' and no proof. This is a load-bearing lemma. If it is not true as stated, the whole 1<p<∞ range collapses. The authors may have a proof in mind, but it is not in the paper.\n\nMinor: the section labelled 'Proof of Proposition 7.1' in Section 4 comes before Proposition 7.1 is even stated; that needs renumbering. Also a few typos.\n\nThe core result looks plausible and the strategy is the right one. But as written, the proof of the main theorem is incomplete in two places. I would not cite it as a theorem until those are fixed. I would still send it to a serious referee, because the question is natural, the classification is new, and the gaps look fixable. If the referee can extract a proof of (4.9) in the missing case and a proof of Lemma 5.1, the paper would be a nice contribution.\n\nFor your reading group, I'd say maybe—the main result is worth knowing, but the incomplete proof makes it awkward for a seminar.\n\nRecommendation: engage with it, but require substantial revision before acceptance.","headline":"Genuinely new classification result, but the proof has a real gap in the key L2 estimate and an unproved bootstrap lemma; worth refereeing, not yet citable as a theorem.","tokens_in":19633,"tokens_out":3381,"would_cite":false,"duration_ms":33152,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B25","43A80","22E30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that lacunary elliptic maximal operators on the Heisenberg group are L^p-bounded for 1<p<∞ precisely when the defining matrix A avoids two explicit exceptional forms.","keywords":["lacunary maximal operator","elliptic maximal operator","Heisenberg group","group Fourier transform","Littlewood-Paley theory","oscillatory integral decay","L^p boundedness","bootstrap vector-valued inequality"],"falsifier":"Take the measure ζ in Lemma 5.1 to be the ellipse measure from (3.1) and choose f_K as the Littlewood–Paley pieces $L^{2}$_{k2−ℓ2}$L^{1}$_{k1−ℓ1}f for a sequence of disjoint bump functions; if the square-function estimate (5.5) fails for some q≤4/3, the bootstrap argument cannot reach p≤4/3 and the claimed full range is disproved.","tokens_in":18975,"feed_emoji":"📐","tokens_out":8320,"duration_ms":77733,"temperature":0.7,"pith_summary":"This paper establishes a complete characterization of when lacunary (dyadic) elliptic maximal operators on the three-dimensional Heisenberg group are bounded on L^p. For 1<p<∞, the one-parameter operator $E^{1}$_A, which takes a supremum over dyadic ellipses whose two axes scale together, is bounded if and only if the 2×2 matrix A is not a nonzero scalar multiple of the identity. The two-parameter operator $E^{2}$_A, whose axes scale independently, is bounded if and only if A is not of the diagonal form diag(c, c·$2^{{2a}}$) with c≠0 and a∈Z. This matters because such dyadic maximal operators are rarely unbounded, and here the curvature induced by A governs the dichotomy completely. The proof combines the group Fourier transform, Littlewood–Paley theory on the Heisenberg group, and decay estimates for oscillatory integral operators.","feed_headline":"Elliptic maximal operators bounded except for two exceptional matrices","feed_subtitle":"One-parameter averages fail exactly for scalar matrices; two-parameter fail for diag(c,c·2^{2a}).","key_machinery":"The load-bearing object is the measure ζ_K carried by a dyadic ellipse segment, defined by ⟨ζ_K,f⟩ = ∫_{π/4}^{3π/4} f($2^{{k1}}$ cosθ, $2^{{k2}}$ sinθ, 0)dθ, with K=(k1,k2) ∈ $Z^{2}$, whose convolution with f on the Heisenberg group gives the elliptic average. The proof splits this convolution into local and global Littlewood–Paley pieces using the Heisenberg Littlewood–Paley projections L^ν_j, then reduces the main term to estimating oscillatory integral operators with phase Φ_A(x,y) = (e(x−y)+(x+y))√(1−(x−y)^2) + b $2^{{2k1}}$(x−y)^2/$2^{{k1+k2}}$ + d $2^{{2k2}}$(1−(x−y)^2)/$2^{{k1+k2}}$ for A with entries (b,e; e,d) plus a skew part. Uniform decay estimates for these phases (via a van der Corput-type lemma) give the $L^{2}$ bounds, and a bootstrap vector-valued maximal inequality extends them to all p>1. In the symmetric case A_w=0, the same decay is obtained through the Euclidean Fourier transform of the ellipse measure and reduces to known multiparameter maximal estimates.","core_discovery":"The central discovery is that L^p boundedness of the lacunary elliptic maximal operators on $H^{1}$ is decided entirely by the shape of the matrix A, with exactly two exceptional families. Specifically, Theorem 1.3 states that for every 1<p<∞, $E^{1}$_A is bounded on L^p($R^{3}$) if and only if A is not cI for any nonzero real c, and $E^{2}$_A is bounded if and only if A is not diag(c, c·$2^{{2a}}$) for any c≠0 and integer a. The paper also proves the positive estimate for all skew-symmetric matrices J (Theorem 1.2) and the matching unboundedness results (Theorem 1.1) for the exceptional cases. The forward direction — boundedness — is obtained by writing the operator as a convolution with an ellipse-carried measure, applying the group Fourier transform, and proving uniform decay for the resulting oscillatory integral operators; the converse is shown by explicit test functions that produce unbounded norms.","pith_inferences":["If the bootstrap Lemma 5.1 is not valid as stated, the claimed range 1<p<∞ would shrink to p>4/3, since that is what the preceding estimates establish directly.","The same two exceptional families likely appear for the analogous maximal operators on higher-dimensional Heisenberg groups, with the scalar-identity condition replaced by appropriate block-scalar conditions; the single-variable proof here does not settle that case.","A concrete testable extension is to replace the dyadic scales 2^{k1},2^{k2} by a lacunary sequence with ratio r>1; the exceptional set of A may depend on r, and the boundedness threshold might shift.","The unboundedness for A=cI reflects the vanishing of curvature of the circular averages along the center of the Heisenberg group; this suggests a general principle that dyadic maximal operators on groups fail to be bounded exactly when the averaging family lies along a degenerate (flat) submanifold."],"forward_implications":["For any skew-symmetric J, the two-parameter lacunary elliptic maximal operator is bounded on L^p(H^1) for every 1<p≤∞.","For every A outside the two exceptional families and every 1<p<∞, both E^1_A and E^2_A are bounded on L^p(R^3).","The exceptional matrices produce genuine unboundedness on every L^p with 0<p<∞, so the characterization in Theorem 1.3 is sharp in both directions.","The decay estimates used in the proof give quantitative exponential control in the lacunary separation parameters ℓ1,ℓ2, which is what makes the summation over the Littlewood–Paley pieces converge."],"supporting_citations":[{"why":"Supplies the group-Fourier-transform strategy for lacunary spherical maximal functions on Heisenberg groups, which the present proof adapts.","marker":"[6]"},{"why":"Gives the lacunary spherical maximal function estimate on the Heisenberg group that the authors extend from spherical to elliptic averages.","marker":"[13]"},{"why":"Provides the Littlewood–Paley theorem on the Heisenberg group used to decompose the elliptic averages.","marker":"[22]"},{"why":"Provides the multiparameter maximal operator boundedness theorem used in the symmetric case and the Fourier-decay condition.","marker":"[14]"},{"why":"Invoked as the source of the vector-valued maximal inequality (Lemma 5.1) that bootstraps the L^p range to all p>1; the paper does not prove the lemma.","marker":"[18]"},{"why":"Supplies the oscillatory-integral tools (van der Corput-type lemma and the Lifting Lemma) used throughout the decay estimates.","marker":"[19]"},{"why":"Supplies the shifted maximal operator estimates used to handle the symmetric case and the summation over ℓ1,ℓ2.","marker":"[23]"}],"fun_headline_variants":["Two matrix families determine boundedness of elliptic maximal operators","Elliptic maximal operators on H^1: only two exceptional matrices","Two matrix shapes break L^p boundedness for lacunary elliptic maximal operators","Boundedness of elliptic maximal operators pinned by two matrix families","Lacunary elliptic maximal operators: boundedness iff not in two matrix families"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim that the L^p bound holds for all 1<p<∞ rests on an unproved vector-valued maximal inequality (Lemma 5.1) that the paper says is 'in the spirit of' a classical differentiation result but does not derive; if that lemma is false, the established range is only p>4/3.","fun_headline_variants_meta":{"raw":{"variants":["Two matrix families determine boundedness of elliptic maximal operators","Elliptic maximal operators on H^1: only two exceptional matrices","Two matrix shapes break L^p boundedness for lacunary elliptic maximal operators","Boundedness of elliptic maximal operators pinned by two matrix families","Lacunary elliptic maximal operators: boundedness iff not in two matrix families"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001551,"raw_usage":{"total_tokens":6150,"prompt_tokens":843,"completion_tokens":5307,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":459,"completion_tokens_details":{"reasoning_tokens":5224}},"tokens_in":459,"tokens_out":5307,"duration_ms":37159,"temperature":1.0,"reasoning_tokens":5224,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:43:17.791735+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the measure ζ in Lemma 5.1 to be the ellipse measure from (3.1) and choose f_K as the Littlewood–Paley pieces $L^{2}$_{k2−ℓ2}$L^{1}$_{k1−ℓ1}f for a sequence of disjoint bump functions; if the square-function estimate (5.5) fails for some q≤4/3, the bootstrap argument cannot reach p≤4/3 and the claimed full range is disproved.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the group-Fourier-transform strategy for lacunary spherical maximal functions on Heisenberg groups, which the present proof adapts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the lacunary spherical maximal function estimate on the Heisenberg group that the authors extend from spherical to elliptic averages."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Littlewood–Paley theorem on the Heisenberg group used to decompose the elliptic averages."},{"cited_title":", Multiparameter singular integrals and maximal functions , Ann","cited_arxiv_id":null,"evidence_quote":"Provides the multiparameter maximal operator boundedness theorem used in the symmetric case and the Fourier-decay condition."},{"cited_title":"and W ainger, Stephen , Differentiation in lacunary directions , Proc","cited_arxiv_id":null,"evidence_quote":"Invoked as the source of the vector-valued maximal inequality (Lemma 5.1) that bootstraps the L^p range to all p>1; the paper does not prove the lemma."},{"cited_title":", Harmonic analysis: real-variable methods, orthogonality , and oscillatory integrals , Princeton Mathematical Series, vol","cited_arxiv_id":null,"evidence_quote":"Supplies the oscillatory-integral tools (van der Corput-type lemma and the Lifting Lemma) used throughout the decay estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the shifted maximal operator estimates used to handle the symmetric case and the summation over ℓ1,ℓ2."}],"review_version":1}