{"id":"9b19b24d-d9dc-4fde-8443-bf44cafb8640","arxiv_id":"2603.02111","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The refined-direction horizontal Kakeya operator on H₁(F_q) has exact ℓᵘ→ℓᵛ norm growth max{1/v, 1−1/u, 2/v−1/u, 1+2/v−3/u}, driven by a sharp, Fourier-proved ℓ²→ℓ² bound q^{1/2}.","lead":"This paper determines exact growth exponents for Kakeya-type maximal operators built from horizontal lines in finite Heisenberg groups, and introduces a refined 'direction plus slope' parameter that yields a sharp square-root ℓ² bound in rank one. It offers a purely Fourier-analytic alternative to the polynomial method in this finite-field model, with consequences for the size of Heisenberg Kakeya sets.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader identified the quadratic-fiber estimate (48) as the most load-bearing step in the proof of Theorem 1.6. I agree that this is the key step, but my stress-test confirms it is correct. The derivation of (48) is transparent: after Fourier transform in γ and orthogonality in m, the bound reduces to counting solutions of Q_ρ(x)=t; because Q_ρ is a nonzero quadratic polynomial when ξ≠0, each fiber has at most 2 elements, giving the factor 2. This does not rely on any hidden assumption beyond ξ≠0. I also verified that the zero-frequency component is properly controlled by the planar ℓ² estimate, that the interpolation in Theorem 1.7 covers all regions with the four extremal exponents, and that the lower-bound test functions are legitimate. The only issue found is a harmless constant typo in (50), where the correct coefficient is 3/q rather than 5/q; this does not change any exponent or the ACCEPT verdict. Thus no load-bearing concern lands.","tokens_in":31769,"tokens_out":33944,"duration_ms":260467,"concrete_test":"Run a brute-force verification for small odd q (e.g., q=5,7,11): generate a random complex-valued f on H_1(F_q), fix ξ∈F_q^*, compute U_ξ(m,γ) and \\hat f, and check that (48) holds with constant 2. Additionally, for q=5, compute the spectral norm of a randomly chosen line-family operator T (one line per refined direction) via SVD on the q^3×q^2 incidence matrix and confirm ∥T∥≤C q^{1/2} with C≈3-4.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the central argument, I find no load-bearing objection. Theorem 1.6 rests on the bound (48): for ξ≠0, Σ_{m,γ}|U_ξ(m,γ)|² ≤ 2q Σ_{x,y}|\\hat f(x,y;ξ)|². The proof is sound: Plancherel in γ, the change of variables to Q_ρ(x)=ξx²−ρx, orthogonality in m forces summation over fibers Q_ρ^{-1}(t), and the degree-2 bound |Q_ρ^{-1}(t)|≤2 (valid in any field when ξ≠0) converts this into the claimed 2q factor. I also checked the zero-frequency reduction (43), which correctly reduces to the planar bound of Lemma 2.5, and the interpolation in Theorem 1.7; the four test functions match the four extremal terms. The only flaw I found is cosmetic: in (50) the constant should be 3/q, not 5/q, from adding (48) and (49); this only shrinks the final constant and does not affect any exponent. Since the contested quadratic-fiber step is in fact correct, the ACCEPT verdict stands.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies two discrete Kakeya maximal operators on finite Heisenberg groups H_n(F_q): the direction-only operator M_Hn and the refined-direction operator M^rd_Hn whose parameter records both the projective horizontal direction and the central slope. It determines the exact mixed-norm growth exponent for M_Hn in every rank (Theorems 1.4 and 1.5) and, in rank one, proves the sharp ℓ^2→ℓ^2 estimate ||M^rd_H1 F||_{ℓ^2(D_1)} ≤ C q^{1/2} ||F||_{ℓ^2(H_1(F_q))} (Theorem 1.6), derives the complete exponent formula for M^rd_H1 (Theorem 1.7), and obtains lower bounds for Heisenberg Kakeya sets (Theorem 1.8) and moment bounds (Theorem 1.9). The proof is Fourier-analytic: the zero-frequency term reduces to a planar TT* estimate, while nonzero central frequencies are controlled by Plancherel, character orthogonality, and the fact that the quadratic Q_ρ(x)=ξx^2−ρx has fibers of size at most two when ξ≠0. Lower bounds are given by explicit test functions. Section 11 contains examples separating affine Kakeya from refined-direction horizontal Kakeya and an explicit outlook toward a new proof of the affine Kakeya theorem in F_q^3.","tokens_in":32020,"tokens_out":11065,"duration_ms":101302,"significance":"If correct, this is a substantial and clean result: it gives the first sharp ℓ^2 bound for the refined-direction horizontal Kakeya maximal operator in the finite Heisenberg group, by purely Fourier-analytic means, without the polynomial method. The zero-frequency reduction to the planar Kakeya estimate is elegant, and the nonzero-frequency argument uses just a degree-two fiber bound, making the mechanism transparent. The exact mixed-norm exponent formula is supported by four matching lower-bound examples. The paper is also careful to separate established theorems from an explicitly labeled outlook; the higher-rank refined-direction estimate and the affine Kakeya program are not claimed as proved. The self-contained TT* proof of the planar input is a welcome feature.","major_comments":[],"minor_comments":[{"comment":"The constant in the displayed bound for ||T_ξ f||^2 should be 3/q, not 5/q: adding (48) and (49) gives (2q+q)Σ|\\hat f|^2 / q^2 = 3/q · Σ|\\hat f|^2. This only affects the numerical constant and does not change any exponent.","section":"Section 8, Eq. (50)"},{"comment":"The first inequality in (30) is tautological as written: ||TG||_{ℓ^r} ≤ (q+1)^{1/r} ||TG||_{ℓ^r}. It should read ||TG||_{ℓ^r} ≤ |P^{2n-1}|^{1/r} ||TG||_{ℓ^∞} ≤ |P^{2n-1}|^{1/r} ||G||_{ℓ^1}. The same typo appears in the proof of Lemma 4.5.","section":"Section 7, Eq. (30) (and analogous display in Lemma 4.5)"},{"comment":"The text explicitly defers the sharp n≥2 refined-direction estimate to a 'forthcoming arXiv revision' and Section 11.2 is presented as an outlook, not a theorem. This is acceptable, but the introduction should state more prominently that the refined-direction results are rank-one only and that the affine Kakeya reduction is a research program, to avoid any impression that those statements are proved here.","section":"Section 1, after Theorem 1.7; Section 11.2"},{"comment":"The conclusion that a full-direction horizontal Heisenberg Kakeya set has size ≳q^3 uses m=q, since a full line has q points. The derivation from (12) with m=q is immediate, but writing it explicitly would help the reader; with m=1 the bound (12) would only give |E|≳q.","section":"Theorem 1.8"},{"comment":"In the last two displays the target space of T is P^1(F_q), not F_q^2. As printed, 'ℓ^2(F_q^2)→ℓ^2(F_q^2)' is a typo and should be 'ℓ^2(F_q^2)→ℓ^2(P^1(F_q))'.","section":"Lemma 2.5, final lines"}],"recommendation":"minor_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know about this one. The main result is real: the refined-direction horizontal Kakeya operator on H_1(F_q) has exact mixed-norm exponent A^rd_1(u,v) = max{1/v, 1−1/u, 2/v−1/u, 1+2/v−3/u}, and the sharp ℓ²→ℓ² bound is q^{1/2}. The refined-direction parameter, which records the central slope, is a natural way to capture the Heisenberg twist, and I don't think it appears in the prior literature. The proof of Theorem 1.6 is the core achievement: a clean zero/nonzero frequency decomposition, Plancherel in the central variable, and the at-most-two-solutions bound for the quadratic Q_ρ(x)=ξx²−ρx. The zero-frequency term reduces exactly to the planar TT* estimate, which the paper proves self-containedly. The lower bounds come from explicit test functions, one for each term in the max formula, so there is no fitted exponent business. The general-rank projective-direction result is a benchmark that follows honestly from Ellenberg–Oberlin–Tao.\n\nI checked the stated weak spot — the fiber bound (52) — and it is correct: a quadratic over F_q takes each value at most twice when ξ≠0. The proof of (48) is solid. The stress-test note flags one minor typo: in (50), the constant should be 3/q, not 5/q, since it comes from adding (48) and (49). That only improves the final constant and changes no exponent. So the paper's central argument holds up.\n\nThe main soft spots are structural, not mathematical. The higher-rank refined-direction problem is left open, and the paper says so plainly; the affine-Kakeya reduction in Section 11 is a promising research program but not a theorem — the cross-slice incidence estimate is still missing. Those are honest limitations, not hidden flaws. The examples in Section 11 are useful and correctly separate affine Kakeya from refined-direction Heisenberg Kakeya. The citation pattern looks appropriate; the Ellenberg–Oberlin–Tao dependence in Section 7 is properly attributed and the rank-one part is self-contained.\n\nWho gets value: anyone working on finite-field Kakeya, maximal operators over nonabelian groups, or the Heisenberg analogue more broadly. It deserves a serious referee. I would send it to peer review rather than desk reject, and in my own work I would cite the refined-direction operator and the sharp ℓ² bound.","headline":"This is a genuine contribution: exact mixed-norm exponents for a new refined-direction Heisenberg Kakeya operator, with a sharp Fourier-analytic ℓ² bound that holds up under checking.","tokens_in":32521,"tokens_out":1297,"would_cite":true,"duration_ms":13867,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B25","11T71","42B25","43A80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes exact mixed-norm growth exponents for horizontal Kakeya maximal operators on finite Heisenberg groups, including a sharp q^{1/2} ℓ² bound for the refined-direction operator in rank one.","keywords":["finite Heisenberg group","Kakeya maximal operator","refined directions","horizontal lines","finite fields","sharp exponents","Fourier analysis"],"falsifier":"For q = 5, 7, or 9, compute the supremum over functions F of ||M^rd F||_{ℓ²(D₁)}/||F||_{ℓ²} by exhaustive search on small supports; Theorem 1.6 asserts this grows at most like C q^{1/2}, while the point mass forces at least (q+1)^{1/2}. Finding any F whose ratio grows like q^{1/2+ε} would refute the sharp bound.","tokens_in":31662,"feed_emoji":"📐","tokens_out":8452,"duration_ms":77939,"temperature":0.7,"pith_summary":"This paper proves exact growth exponents for two Kakeya maximal operators associated with horizontal lines in finite Heisenberg groups over odd prime fields. For the operator that records only the projective spatial direction, it determines the mixed-norm exponent in every rank as max{(2n−1)/v, 1−1/u, 1+(2n−1)/v−2n/u}. For the finer refined-direction operator in rank one, which also records the central slope of a horizontal line, it proves the sharp ℓ²→ℓ² bound with growth q^{1/2} and derives the full exponent A^rd₁(u,v) = max{1/v, 1−1/u, 2/v−1/u, 1+2/v−3/u}. A direct consequence is that every full-direction horizontal Heisenberg Kakeya set has size at least a constant times q³. The proof is Fourier-analytic, avoiding the polynomial method; a key step is a bounded-fiber property of an explicit quadratic map.","feed_headline":"Sharp q^{1/2} bound for refined Heisenberg Kakeya operator","feed_subtitle":"Exact mixed-norm exponents follow, and every full-direction Kakeya set must have size at least q³.","key_machinery":"The central object is the refined direction set D_n = P^{2n}(F_q) \\ {[0:⋯:0:1]}, which parametrizes horizontal lines by projective spatial direction plus central slope; in rank one it has q²+q elements. The proof linearizes M^rd by selecting one line per refined direction and applies a central Fourier transform. The zero-frequency piece maps to the planar Kakeya operator M₂, whose ℓ² norm is bounded by √(2q) via a TT* argument with two-point intersections of lines of distinct directions. The nonzero-frequency pieces are bounded using character orthogonality: for each fixed ξ ∈ F_q^*, the relevant character sum decouples into the quadratic polynomial Q_ρ(x) = ξx² − ρx, and the key counting bo","core_discovery":"The central discovery is the sharp refined-direction estimate in H₁(F_q): if M^rd is the operator that, for each refined direction ω = [a:b:c], takes the maximal sum of |F| over horizontal lines with that direction, then ||M^rd F||_{ℓ²(D₁)} ≤ C q^{1/2} ||F||_{ℓ²(H₁(F_q))}, and the exponent 1/2 cannot be reduced. Combining this with ℓ¹ and ℓ∞ endpoints and interpolation determines all mixed-norm exponents A^rd₁(u,v) exactly; the formula is the maximum of four terms, each forced by an explicit test function. For the coarser operator parameterized only by projective directions, the paper determines the exact exponent in every rank, A_n(u,v) = max{(2n−1)/v, 1−1/u, 1+(2n−1)/v−2n/u}; in rank one t","pith_inferences":["Because the nonzero-frequency bound relies only on the at-most-two fiber size of a quadratic map, the same refined-direction strategy may transfer to other two-step nilpotent groups or Heisenberg-type twists whose central twist is a polynomial of bounded degree; a higher-degree twist would break this particular mechanism.","The paper's proposed route to affine Kakeya in F_q³ suggests a sharper structural question: whether different µ-slices of non-horizontal Kakeya lines can be forced to overlap. A positive answer there would yield a Fourier-analytic proof of the q³ lower bound without polynomial vanishing.","The rank-one refined bound is sharp, but the analogous ℓ²→ℓ^{2n} estimate in H_n for n ≥ 2 remains open; the paper's straightforward adaptation only gives growth q^{n/2}, leaving a gap to the point-mass obstruction scale q^{(2n−1)/2n}. Testing intermediate n would indicate what new mechanism is needed.","The paper's examples show that being an affine Kakeya set in F_q³ and being a full refined-direction Heisenberg Kakeya set are incomparable; this reframes the affine Kakeya lower bound as a question about non-horizontal lines with constrained basepoints."],"forward_implications":["Full-direction horizontal Heisenberg Kakeya sets in H₁(F_q) have size at least a constant times q³, giving a sharp nonabelian analogue of the finite-field Kakeya lower bound.","If a set E meets, for every refined direction in a set Ω, a horizontal line in at least m points, then |E| ≳ m²|Ω|/q; for Ω = D₁ this is the q³ lower bound.","The refined line-intersection function satisfies the higher-moment bound Σ_{ω∈D₁} M_E(ω)^s ≲ q |E|^{s−1} for every 2 ≤ s < ∞.","The exact mixed-norm exponent for the projective-direction operator in every rank is max{(2n−1)/v, 1−1/u, 1+(2n−1)/v−2n/u}; for the refined operator in rank one it is max{1/v, 1−1/u, 2/v−1/u, 1+2/v−3/u}.","The sharp ℓ² exponent 1/2 and the sharp ℓ³→ℓ³ exponent 2/3 for the refined operator are both attained by explicit test functions, so no smaller growth constant is possible."],"fun_headline_variants":["Sharp q^{1/2} estimate for refined Heisenberg Kakeya","Exact mixed-norm exponents for Heisenberg Kakeya operators","Kakeya sets in H₁(F_q) have size at least q³, sharp bound","Fourier-only proof of sharp refined Heisenberg Kakeya estimate"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The estimate for nonzero central frequencies rests on the algebraic fact that the quadratic map Q_ρ(x) = ξx² − ρx attains each value in F_q at most twice; if that fiber-size bound failed—say under a higher-degree Heisenberg twist—the q^{1/2} ℓ² bound would no longer follow from this argument.","fun_headline_variants_meta":{"raw":{"variants":["Sharp q^{1/2} estimate for refined Heisenberg Kakeya","Exact mixed-norm exponents for Heisenberg Kakeya operators","Kakeya sets in H₁(F_q) have size at least q³, sharp bound","Fourier-only proof of sharp refined Heisenberg Kakeya estimate"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000847,"raw_usage":{"total_tokens":3640,"prompt_tokens":979,"completion_tokens":2661,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":723,"completion_tokens_details":{"reasoning_tokens":2577}},"tokens_in":723,"tokens_out":2661,"duration_ms":17979,"temperature":1.0,"reasoning_tokens":2577,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T19:27:52.690699+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For q = 5, 7, or 9, compute the supremum over functions F of ||M^rd F||_{ℓ²(D₁)}/||F||_{ℓ²} by exhaustive search on small supports; Theorem 1.6 asserts this grows at most like C q^{1/2}, while the point mass forces at least (q+1)^{1/2}. Finding any F whose ratio grows like q^{1/2+ε} would refute the sharp bound.","supporting_citations":[],"review_version":1}