{"id":"9ebba2b3-53f8-4fd6-a5a7-172201db8191","arxiv_id":"2507.16510","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For contraction ratios λ up to 2−√3 the circle meets Kλ×Kλ only in two corners; above it the intersection can be non-trivial or even a continuum, and Legendre symbols control intersections with 1/n².","lead":"The paper finds when the unit circle cuts through the product of two self-similar Cantor sets. A sharp threshold at 2−√3 is proved, answering a 2023 question by Yu, and a number-theoretic rule describes when 1/n² lies in such sets.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1(iii) is proved only for λ ≥ 0.407494, while the stated threshold is 0.407493; the determining inequality in (2.11) is tight on the gap and may fail there.","rationale":"The negative answer to Yu's Question 1.2 is Theorem 1.1(i) and is fully proved in Proposition 2.1, independent of computer checks; that core result is not threatened. The nontriviality and continuum parts are secondary but are stated in the abstract and constitute most of Theorem 1.1. The reader's CONDITIONAL verdict is appropriate. My stress-test sharpens the weakest point: the gap between the theorem's threshold 0.407493 and the proof's coverage 0.407494 is real, and the active inequality (first of (2.11)) is exactly what determines the threshold, so this is not just a cosmetic rounding issue. Numerical evaluation suggests P may be slightly negative at 0.407493, meaning the stated threshold could be false for the proposed construction. I also noticed a separate gap in Lemma 2.4: the claim that all subinterval pairs satisfy a' < b' is false (e.g., λ=0.4, a=0.5, b=0.51, n=4, k=5), but the lemma's conclusion still appears true and the proof can be repaired by handling a' > b' via symmetry; this is less central because it does not affect the numerical thresholds. Machine-checked proofs or formal verification are absent, but the parameter count is zero and the main small-λ proof is analytic. Overall, the paper's central contribution is likely correct; the conditional verdict stands pending the threshold patch.","tokens_in":16628,"tokens_out":32324,"duration_ms":275438,"concrete_test":"Compute the exact value of P(λ)=λ^5−λ^3−λ^2+3λ−1 on [0.407493, 0.407494] via Sturm's theorem or high-precision interval arithmetic (e.g., 50-digit precision), and similarly check the remaining three inequalities in (2.11) on the same interval. If min P < 0, find the root λ*; then restate Theorem 1.1(iii) with λ* (or the next safe decimal) instead of 0.407493, or supply a new interval choice for [0.407493, λ*). If min P ≥ 0 on the whole gap, replace 0.407494 by 0.407493 in Proposition 2.10(ii) and confirm the other three inequalities, which closes the gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The continuum claim for 0.407493 ≤ λ < 1/2 rests on Proposition 2.9 applied to the intervals in Proposition 2.10. Proposition 2.10(ii) verifies the required inequalities (2.11) only for 0.407494 ≤ λ < 0.415, and Proposition 2.10(i) starts at 0.415. No argument is given for 0.407493 ≤ λ < 0.407494. This is not merely a missing certificate: the first inequality in (2.11), after multiplying by λ>0, is equivalent to P(λ)=λ^5−λ^3−λ^2+3λ−1 ≥ 0. Evaluating P at 0.407493 gives a value very close to zero and plausibly negative, while P(0.407494) > 0, so this inequality is the active constraint and its root lies inside the stated interval. If P is negative on part of [0.407493, 0.407494), the specific construction in Proposition 2.10 does not satisfy the hypothesis of Proposition 2.9, and no alternative construction is supplied; the theorem's threshold is then too low. If P is nonnegative throughout, the proof can be patched by a one-line change, but as written Theorem 1.1(iii) is not established on that interval. The other numerical verifications (Propositions 2.5 and 2.10) are also stated without code or interval certificates, but the threshold mismatch is the concrete, localized gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the intersection of the unit circle S with the planar Cantor product K_λ × K_λ, where K_λ is the self-similar set generated by {λx, λx+1−λ}. Theorem 1.1 asserts: (i) for 0<λ≤2−√3 the intersection is exactly {(0,1),(1,0)}; (ii) for 0.330384≤λ<1/2 the intersection is non-trivial; (iii) for 0.407493≤λ<1/2 it has cardinality continuum. Remark 1.2(b) gives an explicit decreasing sequence λ_n↘2−√3 with non-trivial intersections, showing sharpness. The paper also contains a complete description of K_λ×K_λ intersected with monomial curves (Theorem 1.3) and, via quadratic reciprocity, a dichotomy for {1/n²}∩K_{m,D} (Theorem 1.5). The proofs are constructive and use interval-covering lemmas, a binary branching argument, and classical number theory.","tokens_in":16933,"tokens_out":27639,"duration_ms":249214,"significance":"If the results are correct, Theorem 1.1(i) with the explicit sharpness sequence resolves the negative direction of Yu's Question 1.2, and the interval-covering method leading to a continuum cardinality is a useful contribution. The proof of Theorem 1.1(i) is a clean induction with verifiable estimates, and the number-theoretic Theorem 1.5 is elegant and appears sound. However, the full strength of the paper depends on several polynomial inequalities that are only checked numerically without certificates, and one of those checks leaves a concrete gap in the proof of Theorem 1.1(iii) as stated. The central contribution is not compromised, but the manuscript needs revision before acceptance.","major_comments":[{"comment":"Proposition 2.10(ii) is proved only for 0.407494≤λ<0.415, while Theorem 1.1(iii) states the threshold 0.407493. The interval [0.407493,0.407494) is not covered by any argument. This is not merely a missing certificate: after multiplying by λ>0, the first inequality in (2.11) is equivalent to P(λ)=λ^5−λ^3−λ^2+3λ−1≥0. At the stated endpoint one has P(0.407493)<0 (approximately −2.6×10^{-7}), while P(0.407494)>0, so the root of P lies inside the claimed range. Consequently the chosen intervals do not satisfy the hypothesis of Proposition 2.9 on [0.407493,0.407494), and no alternative construction is supplied. The authors must either raise the threshold in Theorem 1.1(iii) to 0.407494 and adjust Proposition 2.10 accordingly, or provide a new argument covering the missing interval.","section":"Proposition 2.10, Theorem 1.1(iii), Eq. (2.11)"},{"comment":"The polynomial inequalities that determine the numerical thresholds are asserted to be 'easily checked' or 'checked with the assistance of computers' without any reproducible verification. In Proposition 2.5, the reduction of (2.3) to a single evaluation at λ=0.330384 relies on an unproved monotonicity assertion, and the companion polynomial inequalities are not examined. In Proposition 2.10, the inequalities (2.10) and (2.11) are non-trivial on their intervals and are load-bearing for the continuum claim. The authors should supply explicit Sturm sequences, factorizations, or interval-arithmetic certificates, or reproducible code, so that the claimed thresholds are verifiable.","section":"Propositions 2.5 and 2.10, Eqs. (2.3), (2.10), (2.11)"},{"comment":"The proof states that the sequence {β_k(a_k,b_k)} is increasing to (a+λ^n)^2+(b+λ^n)^2, but the displayed computation only proves β_k(a_k,b_k)>a_{k+1}^2+b_{k+1}^2, which by itself does not imply monotonicity in k. The monotonicity is true and can be shown directly from the formula, but it should be justified explicitly, or the argument should be rephrased so that only the overlap inequality is used.","section":"Lemma 2.8"}],"minor_comments":[{"comment":"The author name 'W ANG' in the running title should read 'WANG'.","section":"Title page / author line"},{"comment":"'Guass's law of reciprocity' is a typo for 'Gauss's law of reciprocity'.","section":"Section 4, after Theorem 4.2"},{"comment":"The conclusion that (α_n(a,b),β_n(a,b))⊂G(I,J) is justified mainly by reference to Figure 2; a short analytic explanation of why every point in the union of the two middle intervals lies in at least two of the four interval images would make the argument self-contained.","section":"Lemma 2.6 and Figure 2"},{"comment":"The displayed containment is stated as 'easy to check', but the verification is not given; since the containment involves the ternary expansions of 1/121 and its shifts, a one-sentence explanation would be helpful.","section":"Question 5.2"},{"comment":"The abstract says the methods 'extend beyond the unit circle and remain effective for many nonlinear curves'; Section 3 treats only monomial curves and the Pythagorean cone, so the wording is broader than the demonstrated scope.","section":"Abstract and Introduction"}],"recommendation":"major_revision","confidential_remarks":"The main gap is localized to the interval [0.407493,0.407494) in Theorem 1.1(iii) and to the lack of certificates for the computer-checked polynomial inequalities. If these are repaired, the paper is likely suitable for publication. I found no circularity, parameter fitting, or other integrity concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version. The paper has one solid, new result that will stick: for λ ≤ 2−√3, the circle meets Kλ×Kλ in exactly the two trivial points, and the threshold is sharp via an explicit sequence. That alone answers Yu's question for λ=1/5 and is a nice piece of work. The number-theory section (Theorem 1.5) is also rigorous and interesting, connecting Cantor set intersections with the Legendre symbol.\n\nNow the soft spot. The claimed continuum result for λ ≥ 0.407493 is not proved. The proof in Proposition 2.10 only covers λ ≥ 0.407494, and the missing interval [0.407493, 0.407494) is not a harmless epsilon: the first inequality in (2.11) reduces to P(λ) = λ^5−λ^3−λ^2+3λ−1 ≥ 0, and P(0.407493) is negative. So the construction in the proof fails at the stated endpoint, and no alternative is supplied. The theorem statement needs to be corrected (either raise the threshold to 0.407494 or find a different construction). This is a specific, localized gap, not a sign of deeper trouble.\n\nTwo smaller notes. The other computer-assisted checks (Propositions 2.5 and 2.10) are not backed by code or interval certificates. In this context, a rigorous proof requires exact polynomial certificates; the phrase \"easily checked\" is not enough. Also, Proposition 2.5's verification of (2.3) at a single λ is fine because the function is monotone, but the algebra is omitted; that's minor.\n\nBottom line: the paper is worth refereeing. The core triviality result is clean and correct, and the number theory is solid. The continuum claim needs a patch, and the numerical checks need to be made rigorous. I'd send it to a good referee with a note to look at the threshold mismatch. Once patched, it will be a useful paper.\n\nWho's it for: fractal geometers and people working on digit expansions and quadratic reciprocity. I'd probably bring it to a reading group to discuss the gap, but I wouldn't rely on the continuum theorem until it's fixed.","headline":"Clean triviality result with sharp threshold, but the continuum claim has a real gap; needs a patch before publication.","tokens_in":17479,"tokens_out":4887,"would_cite":true,"duration_ms":43646,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","28A78"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves a sharp dichotomy: for every $0<\\lambda\\le 2-\\sqrt{3}$, the unit circle meets $K_\\lambda\\times K_\\lambda$ only at $(0,1)$ and $(1,0)$, and the bound cannot be improved.","keywords":["Cantor set","self-similar set","unit circle intersection","iterated function system","cardinality continuum","Legendre symbol","quadratic reciprocity","missing digits"],"falsifier":"Evaluate the four inequalities in (2.11) at $\\lambda=0.407493$ with exact interval arithmetic: if any one fails, the theorem's threshold is not covered by the proof as written; if all four hold, the small gap closes and the stated constant is verified.","tokens_in":16438,"feed_emoji":"⭕","tokens_out":11655,"duration_ms":103826,"temperature":0.7,"pith_summary":"The paper proves a sharp dichotomy for the self-similar Cantor set $K_\\lambda$: when $0<\\lambda\\le 2-\\sqrt{3}$, the only points of $K_\\lambda\\times K_\\lambda$ lying on the unit circle $x^2+y^2=1$ are $(0,1)$ and $(1,0)$, and this cutoff cannot be raised because a sequence of parameters approaching $2-\\sqrt{3}$ from above already forces extra intersections. For larger parameters the picture reverses: at $\\lambda\\ge 0.330384$ the intersection is non-trivial, and at $\\lambda\\ge 0.407493$ it has the cardinality of the continuum. In a separate arithmetic strand, the paper shows that for prime bases $m$ and digit sets satisfying a Legendre-symbol condition, the intersection of the Cantor set with $\\{1/n^2:n\\in\\mathbb{N}\\}$ is either empty or exactly $\\{1/m^{2\\ell}:\\ell\\in\\mathbb{N}\\}$. Together these results give a negative answer to a 2023 open question about missing-digit points near the circle, since $1/5$ lies inside the trivial range.","feed_headline":"One number decides when Cantor squares hit the unit circle","feed_subtitle":"For λ≤2−√3 only two points of Kλ×Kλ lie on x^2+y^2=1; just above, the set becomes uncountable.","key_machinery":"The workhorse is the interval-image calculus for $g(x,y)=x^2+y^2$ on the basic intervals of $K_\\lambda$. Lemma 2.4 shows that when two basic intervals satisfy the ratio condition $\\frac{1-2\\lambda}{\\lambda}(a+\\lambda^n)\\le b\\le \\frac{a}{1-2\\lambda}$, the image $g(I\\times J)$ is exactly the interval image of the union of the next-level subrectangles, so the circle condition can be checked by finite interval overlap. Lemmas 2.6–2.8 locate the set $G(I,J)$ of points covered twice by images of finer subintervals; Proposition 2.9 turns this double-covering set into a full binary tree, yielding continuum many solutions. For the sequence theorem, the load-bearing object is the period of the base-$m$ expansion of $1/n^2$: reducing the period equation modulo $m$ gives $(x_q n)^2\\equiv -x_q\\pmod m$, so a Legendre-symbol condition on the digit set forces the final period digit outside $D$ and hence forces $n$ to be a power of $m$.","core_discovery":"The central claim is a sharp phase transition for $\\mathbb{S}\\cap(K_\\lambda\\times K_\\lambda)$: trivial, with only $(0,1)$ and $(1,0)$, exactly up to $\\lambda=2-\\sqrt{3}$, and non-trivial for $\\lambda\\ge 0.330384$. The proof of triviality runs by induction on the cylinder level of the first coordinate, using the fact that $\\lambda^2-4\\lambda+1\\ge 0$ on this range to rule out every nonzero solution. The richness result is stronger: for $\\lambda\\ge 0.407493$ the intersection has cardinality continuum, and the proof constructs $2^{\\aleph_0}$ solutions by a binary branching argument on interval images. Sharpness is shown by the explicit roots of $\\lambda^{2k+2}+\\lambda^2-4\\lambda+1=0$, which decrease to $2-\\sqrt{3}$ and each give a non-trivial intersection. The arithmetic companion result is a dichotomy for $\\{1/n^2:n\\in\\mathbb{N}\\}\\cap K_{m,D}$ in terms of the Legendre symbol $(-a/m)_L$ over the nonzero digit set, yielding an empty intersection or exactly $\\{1/m^{2\\ell}\\}$.","pith_inferences":["The double-covering interval technique should transfer to any smooth curve whose interval images obey similar overlap inequalities; ellipses and hyperbolas are a direct next test.","The arithmetic dichotomy is probably one instance of a broader principle: for sequences such as $\\{1/n^k\\}$, the intersection with a self-similar set should be governed by a character condition on the digit set modulo the base, and the Legendre-symbol case is the quadratic instance.","The small unverified interval around $\\lambda=0.407493$ can be settled by a short exact interval-arithmetic computation; a reader with a computer algebra system can confirm or correct the stated threshold in minutes."],"forward_implications":["Since $1/5<2-\\sqrt{3}$, the unit circle meets $K_{1/5}\\times K_{1/5}$ only at $(0,1)$ and $(1,0)$, settling the 2023 missing-digits question in the negative for that parameter.","The cutoff $2-\\sqrt{3}$ is optimal: the explicit sequence of roots of $\\lambda^{2k+2}+\\lambda^2-4\\lambda+1=0$ decreases to it and each root produces a non-trivial circle intersection.","For every $\\lambda\\ge 0.407493$, the intersection $\\mathbb{S}\\cap(K_\\lambda\\times K_\\lambda)$ has the cardinality of the continuum; the binary-branching proof actually yields continuum many points on every circle $x^2+y^2=r$ in a suitable interval.","Under the Legendre-symbol condition, any element of $\\{1/n^2:n\\in\\mathbb{N}\\}$ that lies in $K_{m,D}$ must have $n$ a pure power of $m$, so the sequence intersection is either empty or $\\{1/m^{2\\ell}:\\ell\\in\\mathbb{N}\\}$.","For $0<\\lambda\\le 1/3$ and $2\\le q\\le 1/\\lambda-1$, the polynomial curve $y=x^q$ meets $K_\\lambda\\times K_\\lambda$ exactly at $(\\lambda^k,\\lambda^{qk})$ for $k\\in\\mathbb{N}$, together with $(0,0)$ and $(1,1)$."],"supporting_citations":[{"why":"Defines iterated function systems and self-similar sets, which is the construction that produces $K_\\lambda$.","marker":"[6]"},{"why":"Poses the missing-digits question whose $\\lambda=1/5$ case the paper answers in the negative.","marker":"[17]"},{"why":"Supplies the proposition on purely periodic base-$m$ expansions used in the proof of Theorem 1.5.","marker":"[2]"},{"why":"Supplies the Legendre-symbol properties and the quadratic reciprocity law used in Theorem 1.5 and Corollary 4.3.","marker":"[5]"}],"fun_headline_variants":["Cantor squares on unit circle: sharp cutoff at 2−√3","Only two intersections for λ≤2−√3, then uncountably many","Sharp phase transition: Cantor set meets circle at 2−√3","Unit circle hits Cantor product: trivial up to 2−√3","Sharp answer to Yu's problem: Cantor circle intersection"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the handful of polynomial inequalities in Propositions 2.5 and 2.10, which are asserted to be 'easily checked with the assistance of computers', really hold on the stated intervals; no code or interval certificates are supplied, and Proposition 2.10 only verifies its inequalities for $\\lambda\\ge 0.407494$ while the theorem claims the threshold $0.407493$, leaving the small interval $[0.407493,0.407494)$ without a stated verification.","fun_headline_variants_meta":{"raw":{"variants":["Cantor squares on unit circle: sharp cutoff at 2−√3","Only two intersections for λ≤2−√3, then uncountably many","Sharp phase transition: Cantor set meets circle at 2−√3","Unit circle hits Cantor product: trivial up to 2−√3","Sharp answer to Yu's problem: Cantor circle intersection"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000836,"raw_usage":{"total_tokens":3769,"prompt_tokens":1193,"completion_tokens":2576,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":809,"completion_tokens_details":{"reasoning_tokens":2478}},"tokens_in":809,"tokens_out":2576,"duration_ms":20605,"temperature":1.0,"reasoning_tokens":2478,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:07:48.254389+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the four inequalities in (2.11) at $\\lambda=0.407493$ with exact interval arithmetic: if any one fails, the theorem's threshold is not covered by the proof as written; if all four hold, the small gap closes and the stated constant is verified.","supporting_citations":[{"cited_title":"Fractals and self-similarity","cited_arxiv_id":null,"evidence_quote":"Defines iterated function systems and self-similar sets, which is the construction that produces $K_\\lambda$."},{"cited_title":"Missing digits points near manifolds","cited_arxiv_id":"2309.00130","evidence_quote":"Poses the missing-digits question whose $\\lambda=1/5$ case the paper answers in the negative."},{"cited_title":"Ergodic Theory of Numbers","cited_arxiv_id":null,"evidence_quote":"Supplies the proposition on purely periodic base-$m$ expansions used in the proof of Theorem 1.5."},{"cited_title":"An introduction to the theory of numbers","cited_arxiv_id":null,"evidence_quote":"Supplies the Legendre-symbol properties and the quadratic reciprocity law used in Theorem 1.5 and Corollary 4.3."}],"review_version":1}