{"id":"9b2e41a9-0c61-44c7-9ddc-2190cfc8a41b","arxiv_id":"2607.14824","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"For every d≥2, there exist Kakeya sets whose δ-neighborhood volume is O(|log δ|^{-(d-1)}), improving all prior constructions.","lead":"This paper builds needle-like sets that point in every direction but occupy vanishingly little space, in any dimension, improving the best known construction. The new sets shrink at the conjectured optimal logarithmic rate, and the construction sharpens necessary conditions for certain Fourier multipliers.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified — Theorem 1.3 proof is internally consistent; the flagged Proposition 5.4 concern rests on a misreading of δ_n.","rationale":"The reader's weakest assumption centers on Proposition 5.4 being overclaimed for arbitrary nondecreasing f, and on the geometric hypotheses. On the geometric side, the dyadic cube and CFK simplex examples are explicit and satisfy all required properties; the induction in Proposition 4.1 preserves them under affine identifications. On Proposition 5.4, the proof's algebra is internally coherent once δ_n is read as 2^{-2^n}; the suspicious inequalities become identities/dominations rather than failures, and the final patching step works for arbitrary nondecreasing f. The volume computation in §4.2 checks out, and the use of Lemma 2.5 in Lemma 5.3 is plausible and constants are absorbed. I therefore find no load-bearing concern that would invalidate Theorem 1.3, and no reason to move the verdict away from the conditional status already assigned.","tokens_in":23858,"tokens_out":41633,"duration_ms":302951,"concrete_test":"Recompute the key inequalities in Proposition 5.4 under the intended double-exponential definition δ_n=2^{-2^n}: verify δ_n ≤ ε_n for n ≥ 1 + log_2 log_2(1+√(d−1)), verify δ_n/ε_n ≤ 2√(d−1)δ_{n−1}, and verify √δ_n = δ_{n+1}^{1/4}. If these fail under the literal parse δ_n=2^{-2n}, the proof needs revision; they hold under the double-exponential parse.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I traced the central construction and patching argument. Proposition 4.1's induction and the §4.2 volume optimization are internally consistent; the dyadic cube satisfies condition (3) with c=1/2 and is affinely separated, and the CFK simplex does the same for q≥d. The reader's concern about Proposition 5.4 does not land: the proof is consistent if δ_n is read as the double exponential 2^{-2^n}, which the threshold condition 'n ≥ 1 + log_2 log_2(1+√(d−1))' confirms. Under that reading, ε_n=(δ_{n−1}−δ_n)/√(d−1) yields δ_n/ε_n ≈ √(d−1)·δ_n/δ_{n−1} ≤ 2√(d−1)δ_{n−1}, and √δ_n = δ_{n+1}^{1/4}, so the final monotonicity estimate works for arbitrary nondecreasing f. Lemma 2.5 is stated with only a proof sketch, but it is a standard geometric containment fact and its constant is absorbed; this is an exposition gap, not a correctness risk. I find no load-bearing flaw in the proof of |N_δ(E)| ≤ C |log δ|^{-(d−1)}.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs, for every d≥2, a Kakeya set E⊂R^d whose δ-neighbourhood has volume O(|log δ|^{-(d-1)}), improving on the product constructions that gave |log δ|^{-1} in odd dimensions and |log δ|^{-d/2} in even dimensions. The construction is a higher-dimensional analogue of the Perron tree: starting from a (d−1)-polytope Σ with a fair subdivision satisfying a homothety-containment condition (3) and an affine-separation property, the authors iteratively translate pyramids toward a chosen point o and obtain an n-level tree of volume O(n^{-(d-1)}). They verify the required geometric hypotheses for the dyadic cube (Example 3.2) and for the Coxeter–Freudenthal–Kuhn simplicial subdivision (Proposition B.1). By inserting δ-tubes into the cells (Lemma 2.4), they prove a version with pairwise disjoint α-translates (Theorem 1.4), and by a patching argument (Proposition 5.4) they convert the scale-wise Kakeya sets into a single Kakeya set with the stated bound (Theorem 1.3). Section 6 uses Theorem 1.4, through an external result of de la Salle, to derive Besov-regularity conclusions for radial Fourier multipliers (Corollary 1.5).","tokens_in":24127,"tokens_out":28483,"duration_ms":216436,"significance":"If correct, this is the first quantitative improvement over Cartesian-product Kakeya sets in arbitrary dimension, and the exponent (d−1) is the conjecturally optimal logarithmic decay. The proof is self-contained and elementary, and it introduces a clean general framework—fair subdivisions plus affine separation—that may be of independent use. The explicit verification for the cube and the CFK simplex, together with the accompanying Jupyter notebook, are valuable. I also examined the concern raised in the review about Proposition 5.4: it does not appear to land. With δ_n=2^{-2^n}, one has δ_n/ε_n = √(d−1)δ_{n−1}/(1−δ_{n−1}) ≤ 2√(d−1)δ_{n−1}=O(√δ_n), and the final estimate uses only the monotonicity of f, not any slow-variation assumption. Thus the patching argument is consistent for arbitrary non-decreasing f. The remaining issues are expositional and local, not correctness risks.","major_comments":[{"comment":"The induction step applies the induction hypothesis to the (n−1)-iterated partition of a first-level cell Σ_{i_n}, but the affine-separation hypothesis is stated only for the original family (Σ_i,x_i). The proof does not justify that affine separation is inherited by the iterated partitions under the affine maps defining them. This is needed in (ii) for the case i_n≠j_n, and hence for the disjointness conclusion in Theorem 4.2. The fix is short—if φ separates Σ_p and Σ_q, then φ∘A_i^{-1} separates A_i(Σ_p) and A_i(Σ_q) for the affine map A_i defining Σ_{i_n}—but it should be stated explicitly. As written, the induction is incomplete at this point.","section":"§4.1, Proposition 4.1"},{"comment":"Lemma 2.5 is a key geometric input for Lemma 5.3, where it converts the Perron-tree volume bound into a Kakeya set containing unit segments in an open set of directions. The proof is dismissed with “proven in an analogous way” and no details. The uniformity of the constant C_{Σ,v} over all fair partitions is not entirely immediate and should be written out. This is a completeness gap, not an apparent correctness issue, but it needs to be filled for the proof of Theorem 1.3 to be self-contained.","section":"§2, Lemma 2.5"},{"comment":"The proof uses the asserted scaling property f_d^α(δ) ≤ (δ'/δ)^{d−1} f_d^α(δ') for 0<δ≤δ' with no proof or reference. This is used to pass from the dyadic sequence δ_n to arbitrary δ. The property is true and easy to justify by subdividing each δ'-tube into (δ'/δ)^{d−1} δ-tubes, whose α-translates lie inside the original α-translates. Please add the argument or a reference, since the assertion is otherwise unsupported.","section":"§5.1, proof of Theorem 1.4"}],"minor_comments":[{"comment":"The notation “4√δ” (e.g. \\(C_d 4\\sqrt{\\delta}\\)) is ambiguous. From the proof in §5.2 it should be the fourth root \\(\\sqrt[4]{\\delta}\\), not \\(4\\sqrt{\\delta}\\). Please use unambiguous notation throughout.","section":"Theorems 1.3 and 5.4"},{"comment":"The apex notation “v=(0,1)” should be \\(v=(0,\\dots,0,1)\\in\\mathbb{R}^d\\), and the homothety notation “\\(C\\operatorname{conv}(\\Sigma_i,v)\\)” should specify the center (presumably \\(v\\)) to avoid ambiguity.","section":"§5.2, Lemma 5.3"},{"comment":"The set \\(G_{2^{-n}}\\) constructed as a finite union of translates of open neighborhoods is open, while Kakeya sets are defined as compact. Taking closures preserves the volume bound and the property of containing the required segments; the text should say this explicitly.","section":"§5.2, after Lemma 5.3"},{"comment":"The threshold \\(n\\ge 1+\\log_2\\log_2(1+\\sqrt{d-1})\\) is stated without explanation of where it comes from and without tracking the dimension constants. The phrase “for an appropriate value of \\(C_d\\)” is vague; a brief sentence explaining how finitely many small \\(n\\) are absorbed would improve readability.","section":"§5.2, Proposition 5.4"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript appears to be a genuine advance and the central derivation is sound. The reported concern about Proposition 5.4 is, in my reading, not valid. The remaining issues are expositional: a missing proof for Lemma 2.5, an unstated invariance fact in Proposition 4.1, and a few notation ambiguities. These can be fixed without changing the argument. The paper is a good fit for the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this paper works. The main theorem, Kakeya sets in R^d with δ-neighborhood volume O(|logδ|^{-(d-1)}), is a genuine improvement over product constructions, and the proof is in good shape. The reader's worry about Proposition 5.4 doesn't survive a careful look: with δ_n = 2^{-2^n}, the patching argument goes through for arbitrary nondecreasing f, because δ_{n-1} = sqrt(δ_n) and the constants absorb the rest. That proposition is stated more generally than it needs to be, but it's correct as stated.\n\nWhat's actually new: the higher-dimensional Perron tree built from fair subdivisions of the base polytope, and the volume bound with exponent d−1. The paper is explicit about debts: the cube-based version reduces to Keich's 2D formulas, and the authors note that versions were in the air among experts. That's honest. The real contribution is a clean, general framework (condition (3), affine separation, the induction in Proposition 4.1) that covers both the dyadic cube and the Coxeter-Freudenthal-Kuhn simplex subdivision, and the tube-disjointness version in Theorem 1.4, which drives the harmonic analysis application.\n\nI traced the volume computation and the cone-disjointness arguments; they check out. Lemma 2.4 is fine. Lemma 2.5 is given with a proof sketch, which is acceptable for a standard containment fact but terse; an editor might ask for two more lines. The CFK simplex section requires q≥d, which is fine here and they flag it. The Besov/multiplier corollaries come out of the cited [dlS26] black box, so their validity is contingent on that preprint, but that is not a defect in this paper.\n\nSoft spots, in proportion: the presentation occasionally under-specifies notation (the δ_n in Prop 5.4) and the novelty should be stated carefully—this is not the first construction in spirit, but the first with this exponent. Both are minor. I found no load-bearing gap.\n\nBottom line: worth a serious referee. I'd accept for review and expect it to be published after small expository fixes.","headline":"The higher-dimensional Perron tree construction is correct and gives the expected d−1 log-exponent; the reader's main worry about Proposition 5.4 does not land.","tokens_in":24660,"tokens_out":3202,"would_cite":true,"duration_ms":28078,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A78","42B15","42B25"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs Kakeya sets in R^d whose δ-neighbourhood volume is at most C |log δ|^{-(d-1)}, the conjectured optimal decay.","keywords":["Kakeya sets","Perron tree","fair subdivision","δ-tubes","Fourier multipliers","Besov spaces","Minkowski dimension","reverse Littlewood-Paley"],"falsifier":"Take a specific fair subdivision that does not satisfy the corner-containment condition (3), for instance a barycentric subdivision, which lacks bounded aspect ratios, and compute the volume of the union of the translated pyramids after n iterations; finding it grows faster than n^{-(d-1)} would show the hypothesis is essential. Alternatively, exhibit a fair subdivision whose reflected cones overlap after the prescribed translations, which would destroy the disjointness in Theorem 1.4.","tokens_in":23725,"feed_emoji":"📐","tokens_out":4876,"duration_ms":40428,"temperature":0.7,"pith_summary":"The paper generalizes the two-dimensional Perron tree construction to every dimension d≥2, producing Kakeya sets whose δ-neighbourhood volumes decay like |log δ|^{-(d-1)}. Previous product constructions only gave |log δ|^{-1} in odd dimensions and |log δ|^{-d/2} in even dimensions, so this is a genuine improvement and matches the best bound allowed by the reverse Littlewood–Paley conjecture. The engine is a hierarchical rearrangement: a fair subdivision of the base polytope into congruent pieces, with each iteration translating whole blocks toward one point so the union shrinks while the reflected opposite cones stay disjoint. The same construction yields families of δ-tubes with pairwise disjoint translates and small union, and consequently gives new necessary conditions for L^p boundedness of radial Fourier multipliers in terms of Besov spaces with logarithmic smoothness.","feed_headline":"Kakeya sets attain near-optimal thinness in every dimension","feed_subtitle":"A new construction generalizes the Perron tree, giving δ-neighborhood volume C|log δ|^{-(d-1)} — the conjectured best possible.","key_machinery":"The fair subdivision and the affine-separation condition. A fair partition of the base Σ into N congruent pieces (images of N^{-1/(d−1)}Σ under a compact group) keeps aspect ratios from degenerating, so each pyramid conv(Σ_i,v) contains a δ-tube of width comparable to N^{-1/(d−1)}. Condition (3) says each piece lies in a homothetic copy of Σ centered at a distinguished point x_i; translating the piece toward o then keeps it inside a shrinking copy of Σ. Affine separation of the pairs (Σ_i,x_i) ensures the reflected cones Cone^-(Σ_i,v) have disjoint interiors after translation, which is what makes the tubes' translates disjoint in Theorem 1.4.","core_discovery":"For every d≥2 there exists a Kakeya set E⊂R^d and a constant C such that for every δ∈(0,1), |N_δ(E)| ≤ C |log δ|^{-(d-1)}. The construction is an iterated Perron tree: starting from a (d−1)-dimensional polytope with a fair subdivision satisfying a corner-containment condition and an affine-separation property, one translates the pyramids conv(Σ_i,v) toward a point o in the base so that after n iterations the union has volume O(n^{-(d-1)}). The bound is conjecturally optimal, as the reverse Littlewood–Paley conjecture would imply a matching lower bound. The paper also proves a tube-disjointness version (Theorem 1.4) and derives that radial Fourier multipliers that are L^p-bounded force the fu","pith_inferences":["Editorial inference: the construction should extend to any self-similar tiling whose cells are congruent to a scaled copy of the base and whose corner structure permits affine separation; testing other Coxeter-type tessellations could reveal which polytopes yield the same exponent.","Editorial inference: the disjointness of reflected cones is stronger than needed for the volume bound; a quantitative version measuring how often cones overlap might yield bounds for Kakeya maximal operators rather than just for a single set.","Editorial inference: if the conjectured lower bound holds, these Kakeya sets are exactly as thin as possible, so the obstruction to proving the Kakeya conjecture is not the existence of very thin sets but the L^p behavior of maximal operators; this reframes where the difficulty lies."],"forward_implications":["There exist Kakeya sets in R^d whose δ-neighbourhood volume is at most C|log δ|^{-(d-1)}, improving on all previously known explicit constructions for d≥3.","For every α>0, there are δ-tubes whose α-translates are pairwise disjoint yet whose union volume is at most C_{d,α}|log δ|^{-(d-1)} times the sum of the tube volumes.","Any radial Fourier multiplier bounded on L^p(R^d) must have symbol satisfying Besov regularity B^{0,(d-1)|1/p-1/2|}_{∞,∞} in logarithmic scale; in particular, logarithmic Bochner–Riesz multipliers with exponent below (d−1)|1/p−1/2| are unbounded.","Improving the exponent (d−1) in Theorem 1.4 would yield stronger necessary conditions for logarithmic Bochner–Riesz multipliers and would have further consequences for noncommutative L^p approximation properties.","The construction works for any fair polytopal partition satisfying the two geometric hypotheses; the dyadic cube and the Coxeter–Freudenthal–Kuhn simplex subdivision are concrete instances."],"fun_headline_variants":["Kakeya sets reach conjectured thinness in every dimension","Perron tree generalized: near-optimal Kakeya in any d","Thinnest Kakeya sets yet, now in arbitrary dimension","Tight Kakeya construction: volume drops like log^{-(d-1)}","Kakeya sets: near-optimal bound achieved for all d"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction depends on partitioning the base polytope into congruent pieces that can all be pushed toward one point while staying inside a shrinking copy of the whole, with reflected cones that never overlap; if such a partition is unavailable, the volume bound n^{-(d-1)} and the disjointness of tubes both fail.","fun_headline_variants_meta":{"raw":{"variants":["Kakeya sets reach conjectured thinness in every dimension","Perron tree generalized: near-optimal Kakeya in any d","Thinnest Kakeya sets yet, now in arbitrary dimension","Tight Kakeya construction: volume drops like log^{-(d-1)}","Kakeya sets: near-optimal bound achieved for all d"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000185,"raw_usage":{"total_tokens":1111,"prompt_tokens":652,"completion_tokens":459,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":364}},"tokens_in":396,"tokens_out":459,"duration_ms":4760,"temperature":1.0,"reasoning_tokens":364,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T00:58:39.993362+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a specific fair subdivision that does not satisfy the corner-containment condition (3), for instance a barycentric subdivision, which lacks bounded aspect ratios, and compute the volume of the union of the translated pyramids after n iterations; finding it grows faster than n^{-(d-1)} would show the hypothesis is essential. Alternatively, exhibit a fair subdivision whose reflected cones overlap after the prescribed translations, which would destroy the disjointness in Theorem 1.4.","supporting_citations":[],"review_version":1}