{"id":"da321472-d9be-4bb7-ba50-b259ccb17b6f","arxiv_id":"2606.27034","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.5,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For the Cantor measure µ, the set of points with ||2^n x|| < n^{-τ} infinitely often has measure zero when τ > 2-γ and full measure when τ < (1-γ)/2.","lead":"The paper proves sharper zero-one laws for how well points on the middle-third Cantor set can be approximated by dyadic rationals. It narrows the gap toward a long-standing conjecture of Velani by improving both the convergence and divergence exponents via new averaged Fourier estimates.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The manuscript’s central claims rest on a single, cleanly proved averaged Fourier bound that exploits the exact multiplicative order of 2 modulo powers of 3. That bound is correctly transferred into first- and second-moment estimates and then into standard Borel–Cantelli arguments. The reader’s identification of Lemma 6 as the weakest (yet still secure) assumption is accurate; no deeper gap or unjustified step is present. The quantitative distance to Velani’s threshold τ=1 remains large, but that is a limitation of scope, not of correctness. Concurrent independent work is disclosed. Consequently the ACCEPT verdict with high confidence stands unchanged.","tokens_in":10898,"tokens_out":476,"duration_ms":4722,"concrete_test":"Re-derive the constant in the multiplicity bound of Lemma 6 Case 2: verify that #N(y) ≤ 1 + H/(2·3^{K-a-1}) ≪ 3^a uniformly for 3^K ≤ H < 3^{K+1}, then recompute the resulting sum 3^a · 2^{K-a} and confirm it is still ≪ H^γ 3^{(1-γ)a}. If the implicit constant remains absolute, the claim is unaffected.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly isolates Lemma 6 as the load-bearing input. Its proof is elementary and self-contained: the order of 2 mod 3^{K-a} is exactly 2·3^{K-a-1}, so residues of q0 2^n fill the units with multiplicity ≪ 3^a; the resulting sum of PK is then bounded by Lemma 4 (the inductive 3-adic averaging of cosines). Both the first-moment estimate (Lemma 10) and the bilinear second-moment estimate (Lemma 7) inherit the same absolute constant and the same exponent γ+(1-γ)min(ν3,K). The subsequent Borel–Cantelli arguments (coarse-to-fine transfer via Ahlfors regularity for convergence; Markov on the L2 discrepancy for divergence) introduce no further free parameters or hidden cancellations. No internal inconsistency or unjustified absolute-constant claim appears.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The paper studies dyadic Diophantine approximation on the middle-third Cantor set C with its natural measure μ (Hausdorff dimension γ = log 2 / log 3). For the sets of points x ∈ C with ||2^n x|| < n^{-τ} for infinitely many n, it proves a zero law when τ > 2 − γ ≈ 1.369 and a full law when τ < (1 − γ)/2 ≈ 0.185. These improve the previously known ranges (roughly τ ≳ 1.55 for zero and τ ≲ 0.01 for full) and give partial progress toward Velani’s conjectured zero–one law with threshold τ = 1. The proofs rest on new averaged and bilinear bounds for the Fourier transform of μ along dyadic orbits, obtained from the product formula for μ̂, the multiplicative order of 2 modulo powers of 3, and an inductive cosine-averaging lemma; these are combined with smooth majorants/minorants, coefficient sums weighted by 3-adic valuations, Ahlfors regularity (coarse-to-fine transfer), and Borel–Cantelli / L^2 Markov arguments.","tokens_in":11077,"tokens_out":834,"duration_ms":18689,"significance":"The work makes concrete, quantitative progress on both sides of a well-known conjecture in metric Diophantine approximation on fractals. The averaged Fourier estimates (Lemmas 6–7) are elementary, self-contained, and of independent interest; they exploit the special pair of bases (2, 3) more sharply than the classical Schmidt-type arguments they draw on. The concurrent independent preprint of Dai–Li–Wang–Wu is properly flagged. There are no free parameters, no numerical fitting, and the arguments reduce cleanly to absolute-constant bounds and standard measure-theoretic tools. The results are therefore a solid, citable advance even though the conjectural threshold τ = 1 remains open.","major_comments":[],"minor_comments":[{"comment":"The exponent β := 1 − γ is used in Lemma 6 (and thereafter) before it is defined. Introduce β = 1 − γ explicitly at the first appearance (or in the introduction) so that the statements of Lemmas 6–7 and the coefficient-sum lemmas are self-contained.","section":null},{"comment":"Several OCR/typesetting artefacts remain in the front matter and early sections (e.g., “A VERAGED”, “DY ADIC”, “for sq∈Z”, occasional subscript glitches such as a±_ℓ,R,y). A careful proofreading pass is needed.","section":null},{"comment":"In the statement of Lemma 10 the sum is bounded by ≪ N^γ, while the displayed calculation ends with “= 2 N^γ”; the absolute constant is harmless but the wording “= 2 N^γ” should be replaced by “≪ N^γ” for consistency with the rest of the paper.","section":null},{"comment":"The paper notes that the methods adapt to inhomogeneous and asymptotic-counting statements but does not pursue them. A brief remark on the precise range that would follow for the inhomogeneous problem (or a pointer to where the extra terms appear) would help the reader assess the scope of the technique.","section":null},{"comment":"References [BHZ26] and [DLWW26] are listed with 2026 dates; if these are still preprints, the arXiv identifiers (already given for some) should be made uniform for all unpublished items.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is technically sound and the concurrent independent work is disclosed appropriately. The contribution is incremental but cleanly executed; it is a good fit for a solid number-theory or fractal-geometry journal. I see no reason to delay acceptance beyond the minor presentation fixes listed above."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The paper improves the published ranges for Velani’s zero-one problem on dyadic approximation in the middle-third Cantor set: zero measure for τ > 2 − γ ≈ 1.37 (down from ≈ 1.55) and full measure for τ < (1 − γ)/2 ≈ 0.185 (up from Baker’s 0.01). Both statements are proved by elementary Fourier analysis of the Cantor measure.\n\nWhat is new is the pair of averaged estimates (Lemmas 6–7). They use the exact order of 2 modulo 3^r together with a short inductive cosine-sum bound (Lemma 4) to get a saving of 3^{(1−γ) min(ν3(q),K)} over the trivial length-H bound. Once those estimates are in hand, the rest is standard: smooth majorants/minorants of the indicator, coefficient sums controlled by the same 3-adic weights, Ahlfors regularity for the coarse-to-fine transfer, and first- and second-moment Borel–Cantelli. The proofs are self-contained and short; the concurrent Dai–Li–Wang–Wu preprint is flagged as independent.\n\nThe soft spots are real but limited. The gap to the conjectured threshold τ = 1 remains large, so the advance is quantitative rather than conceptual. The absolute constants in the Fourier averages are not tracked, and the method does not yet reach the critical exponent. Neither issue undermines the stated theorems.\n\nThis is for specialists working on metric Diophantine approximation on fractals or on Fourier decay of self-similar measures. The math is solid, the citations are appropriate, and there are no free parameters or circular steps. I would send it to referees without hesitation; a serious journal in the area should take it.","headline":"Clean elementary improvement of the known zero-one ranges for dyadic approximation on the middle-third Cantor set; the averaged Fourier bounds are the real content.","tokens_in":11702,"tokens_out":466,"would_cite":true,"duration_ms":3702,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11J83","28A80","11K60","37A45"],"pacs":[],"model":"grok-4.5","headline":"Averaged Fourier bounds on the Cantor measure push dyadic approximation thresholds closer to the conjectured zero-one law.","keywords":["middle-third Cantor set","dyadic approximation","Cantor measure","Fourier transform","metric Diophantine approximation","zero-one law","Velani conjecture"],"falsifier":"Compute or rigorously bound the sum of |µ̂(q 2^n)| over intervals of length H = 3^K for a sequence of q with fixed 3-adic valuation; if the sum is asymptotically larger than H^γ 3^{(1−γ) min(ν_3(q),K)}, the main theorems fail.","tokens_in":11773,"feed_emoji":"📐","tokens_out":969,"duration_ms":8724,"temperature":0.7,"pith_summary":"The paper studies how well points of the middle-third Cantor set can be approximated by dyadic rationals, measured with respect to the natural Cantor probability measure. Velani's conjecture predicts a sharp zero-one law: almost every Cantor point is approximable to order n^{-τ} infinitely often precisely when τ ≤ 1. Previous work only controlled the extremes τ ≳ 1.55 (convergence) and τ ≲ 0.01 (divergence). By establishing new averaged estimates for the Fourier transform of the Cantor measure along the orbit q·2^n, the paper improves both sides: the measure of the limsup set is zero for every τ > 2 − γ ≈ 1.369 and is full for every τ < (1 − γ)/2 ≈ 0.185. The same estimates also yield the corresponding statements for the power functions ψ(2^n) = n^{-τ}. A sympathetic reader cares because the dyadic and triadic bases are multiplicatively independent, so the problem is genuinely arithmetic rather than geometric; each improvement of the exponents narrows the gap between what the Fourier method can prove and the conjectured threshold.","feed_headline":"Cantor dyadic approximation gaps shrink to 1.37 and 0.185","feed_subtitle":"New averaged Fourier bounds push both sides of Velani's zero-one law closer to the conjectured threshold of 1.","key_machinery":"Averaged Fourier estimates (Lemma 6 and the bilinear form Lemma 7): for H ≍ 3^K the sum of |µ̂(q 2^n)| over any interval of length H is ≪ H^γ 3^{(1−γ) min(ν_3(q),K)}. The bound is obtained by combining the exact multiplicative order of 2 modulo powers of 3 with a finite averaging argument over residue classes; both the first-moment convergence proof and the second-moment divergence proof reduce to these estimates.","core_discovery":"For the middle-third Cantor measure µ and γ = log 2 / log 3, the set of points x in C that satisfy ||2^n x|| < n^{-τ} for infinitely many n has µ-measure zero whenever τ > 2 − γ, and has full µ-measure whenever τ < (1 − γ)/2. These are the first quantitative improvements of the known ranges on both the convergence and divergence sides of Velani's zero-one conjecture for dyadic approximation on C.","pith_inferences":["The remaining gap between 0.185 and 1.369 still leaves room for a method that exploits more of the multiplicative independence of 2 and 3 than pure Fourier averaging.","The same order-of-2-modulo-3^r technique should transfer to other self-similar measures whose contraction ratios are powers of an odd integer.","If the bilinear estimate can be sharpened by a logarithmic factor, the divergence side would reach τ < 1 − γ, closing half the remaining gap."],"forward_implications":["The zero-measure threshold for power-law dyadic approximation on C drops from roughly 1.55 to 2 − γ ≈ 1.369.","The full-measure threshold rises from 0.01 to (1 − γ)/2 ≈ 0.185.","The same averaged estimates adapt immediately to inhomogeneous approximation and to asymptotic counting statements of Baker type.","Any further improvement of the averaged Fourier exponent would automatically tighten both sides of the zero-one law."],"fun_headline_variants":["Cantor µ vanishes for dyadic rates above 2-γ","Full Cantor measure for rates below (1-γ)/2","Averaged Fourier bounds tighten both Velani sides","Zero-one progress: µ=0 past 2-γ, µ=1 under (1-γ)/2","New dyadic thresholds on middle-third Cantor measure"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The absolute-constant averaged Fourier bound must hold uniformly for every nonzero integer q; if the power of 3 that multiplies the valuation of q cannot be controlled that sharply, both the zero and full-measure statements collapse.","fun_headline_variants_meta":{"raw":{"variants":["Cantor µ vanishes for dyadic rates above 2-γ","Full Cantor measure for rates below (1-γ)/2","Averaged Fourier bounds tighten both Velani sides","Zero-one progress: µ=0 past 2-γ, µ=1 under (1-γ)/2","New dyadic thresholds on middle-third Cantor measure"]},"model":"grok-4.5","effort":"low","cost_usd":0.007306,"raw_usage":{"total_tokens":1771,"prompt_tokens":734,"num_sources_used":0,"completion_tokens":79,"cost_in_usd_ticks":73060000,"prompt_tokens_details":{"text_tokens":734,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":958,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":734,"tokens_out":79,"duration_ms":7767,"temperature":1.0,"reasoning_tokens":958,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-12T11:50:58.571730+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Compute or rigorously bound the sum of |µ̂(q 2^n)| over intervals of length H = 3^K for a sequence of q with fixed 3-adic valuation; if the sum is asymptotically larger than H^γ 3^{(1−γ) min(ν_3(q),K)}, the main theorems fail.","supporting_citations":[],"review_version":2}