{"id":"7033d85d-248c-4cc1-a14f-59dff3cda645","arxiv_id":"2507.17167","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For any growth function φ, the sets of reals whose product of ℓ consecutive prime partial quotients exceeds φ infinitely often have Hausdorff dimension equal to the unrestricted case, and zero-one Lebesgue measure dictated by a log log φ-weighted series.","lead":"This paper proves that for continued fractions, making infinitely many partial quotients both large and prime changes the Lebesgue measure zero-one threshold but leaves Hausdorff dimension unchanged. It also gives new estimates for the tail of the almost prime zeta function, a tool of independent interest.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: central dimension equality is supported by the proof once minor statement typos are corrected.","rationale":"The reader's weakest_assumption singles out Proposition 2.4 as load-bearing. My read finds that assumption true and well-supported by classical explicit prime-gap estimates, so it is not a source of real risk. The proof has several correctable typos and one short 'first case' that should be labeled 'B_phi < infinity' with C > B_phi, but no step that would invalidate Theorem 1.10 or Theorem 1.8. The main constructive lower bound is a faithful specialization of Huang-Wu-Xu with prime cylinders substituted, and the counting estimates are consistent once the obvious typo in Proposition 2.3 is reversed. I therefore do not see a load-bearing objection; the conditional verdict can remain because the manuscript would benefit from cleanup, but the mathematical claim appears sound.","tokens_in":25617,"tokens_out":41776,"duration_ms":411163,"concrete_test":"Verify Proposition 2.4 with Dusart's explicit prime-gap theorem: for x > e^20, log^2 x > 400 and x/(25 log^2 x) < x/1000, so applying the theorem to 0.999x yields a prime in [0.999x, x). Then re-run the Section 6.6 finite-B_phi argument with C > B_phi and N_C = {n: log phi(n)/n < log C}; if pressure-function continuity gives dim_H E'_ell(phi) >= t_B^(ell), the central equality is confirmed in the 1 < B_phi < infinity regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I checked the proof of Theorem 1.10, the central claim. The lower-bound construction in Sections 6.1-6.5 for exponential phi_B(n)=B^n uses Proposition 2.4 to obtain primes at both ends of each interval [alpha^n, 2alpha^n]; applying it to x/0.999 and 2x gives q2 - q1 of size x, so Lemma 6.2 and Proposition 6.4 are justified. Proposition 2.4 itself follows from standard explicit prime-gap bounds (e.g., Dusart), so the reader's flagged assumption is not fragile. The proof of Theorem 1.10 for general finite B_phi is indicated in Section 6.6 by 'letting C -> B_phi'; the line 'First assume B_phi = 1' should read 'B_phi < infinity', and one must choose C > B_phi instead of C > 1, a routine correction. The remaining blemishes are exposition typos: Proposition 2.3's fraction should be gamma^n/(n log gamma), Theorem 1.7's base case says s > 2 but the argument works for s > 1, and Omega(ell) is used as exactly ell primes despite the 'at most' wording. None of these break the central argument; the dimension equality and the zero-one law stand.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the metric theory of continued fractions whose partial quotients are restricted to primes and are required to be large. For functions φ: N → R_{>0}, it defines E'_ℓ(φ) as the set of x for which a product of ℓ consecutive prime partial quotients is at least φ(n) infinitely often, and E''_ℓ(φ) for the analogous condition holding for all n. The main results are a zero-one law for the Lebesgue measure of E'_ℓ(φ) (Theorem 1.8), an equality of Hausdorff dimensions dim_H E'_ℓ(φ) = dim_H E_ℓ(φ) with the unrestricted set (Theorem 1.10), a Łuczak-type dimension theorem for E''_ℓ and E'_ℓ in the double-exponential case (Theorem 1.11), and an asymptotic for the tail of the almost prime zeta function (Theorem 1.7). The proofs follow the Cantor-set and pressure-function framework of Huang-Wu-Xu, replacing counts over all integers by counts over primes via the Prime Number Theorem, Mertens' theorem, and an explicit prime-gap result.","tokens_in":25832,"tokens_out":20711,"duration_ms":197142,"significance":"If the results hold, Theorem 1.10 is a clean and natural statement: imposing primality on the large partial quotients does not change the Hausdorff dimension, and Theorem 1.8 gives the exact prime analogue of the Huang-Wu-Xu zero-one law. The paper is careful and constructive: the Cantor-set lower bound is built explicitly, the estimates are traced back to standard number-theoretic inputs, and there are no fitted parameters or circular reductions. The almost prime zeta tail estimate of Theorem 1.7 is also of independent interest. However, the written proof contains statement/proof mismatches, most notably a possible gap in the central finite-B_φ case of Theorem 1.10 and an incomplete reconciliation between the definition of Ω(ℓ) and the quantity actually estimated in Section 3.","major_comments":[{"comment":"The theorem is stated for P_ℓ(s;M) with Ω(ℓ) defined as the set of positive integers with 'at most ℓ' prime factors, but the proof establishes the estimate only for S(ℓ,M,1,s), the ordered sum over products of exactly ℓ primes. These two objects are not identical, and the passage between them is not written. Please either redefine Ω(ℓ) as the set of integers with exactly ℓ (not necessarily distinct) prime factors and add the comparison S(ℓ,M,1,s) ≍ ∑_{Ω(n)=ℓ, n≥M} n^{-s}, or, if 'at most ℓ' is intended, add the argument that the sums over products of fewer than ℓ primes are of strictly smaller order and do not affect the asymptotic. Since Theorem 1.8 invokes Theorem 1.7, this point must be made explicit.","section":"Section 3, Theorem 1.7"},{"comment":"The written proof of Theorem 1.10 only discusses the cases B_φ = 1 and B_φ = +∞. The case 1 < B_φ < ∞, which is precisely the case prepared by the construction in Sections 6.1-6.5, is missing: the paragraph beginning 'First, assume that B_φ = 1' runs only for C > 1, and the next paragraph jumps to B_φ = +∞. The natural repair is to run the same argument for any finite B_φ > 1, defining N_C = {n : log φ(n)/n < log C} for C > B_φ, using E'_ℓ(φ_C; N_C) ⊆ E'_ℓ(φ), the equality dim_H E'_ℓ(φ_C; N_C) = dim_H E'_ℓ(φ_C) from Lemma 6.6, and then letting C ↓ B_φ. This is routine, but as it stands the central theorem has a gap in its main case and the section needs to be rewritten.","section":"Section 6.6, proof of Theorem 1.10"},{"comment":"The inclusion E''_ℓ(φ_{e,b_φ+δ}) ⊆ E'_ℓ(φ) is asserted without proof. It is true, but the justification is not immediate from the displayed limit comparison: one must observe that liminf_n (log log φ(n))/n = log b_φ < log(b_φ + δ) implies that φ_{e,b_φ+δ}(n) ≤ φ(n) for infinitely many n, so any x whose all-block product exceeds the faster-growing function still satisfies the infinitely-often condition for φ. Please supply this argument. In addition, the case b_φ < 1 is not discussed; if it is excluded by the earlier reduction to Lebesgue-null sets, that reduction should be stated explicitly.","section":"Section 6.6, B_φ = +∞ subcase"}],"minor_comments":[{"comment":"The displayed formula appears to have the fraction inverted: it should read #(P ∩ [γ^n, 2γ^n]) = c_n(γ) γ^n / (n log γ), since the Prime Number Theorem gives a count asymptotic to γ^n / (n log γ). With the printed formula, c_n(γ) → 1 would force the count to tend to 0, contradicting the intended meaning and the later usage in Section 6.","section":"Proposition 2.3"},{"comment":"The base case is introduced with 'Consider an arbitrary s > 2', but the argument via Karamata's theorem works for every s > 1. The statement of Theorem 1.7 also says s > 1, so the proof should be adjusted to s > 1 or the statement should be restricted.","section":"Section 3, proof of Theorem 1.7"},{"comment":"Proposition 1.12 is stated as a result in the introduction, but no proof appears anywhere in the paper, and the surrounding text contains two different definitions labelled S'_ℓ(A_0, ..., A_{ℓ-1}) as well as an undefined S_ℓ. The authors should either prove Proposition 1.12, provide a reference for it, or reformulate it as a remark with a proof sketch; as written, it is an unsupported claim.","section":"Introduction, Proposition 1.12"},{"comment":"The notation for the Łuczak-type functions is inconsistent: the introduction uses φ_{b,c}, while Theorem 1.11 and Section 5 use φ_{c,b}; Section 6.6 also uses φ_{e,b_φ+δ} in a way that mixes the two conventions. Please unify the notation.","section":"Notation throughout"},{"comment":"The inequalities in condition (24) contain several parentheses and index expressions that are hard to parse, for example '2ℓN ℓ log α0⋯ log αi−2 (α0(α1⋯αi)2αi+1)sδN'. Please rewrite the condition cleanly and verify the indices, since the equality claimed immediately after it is used to justify the measure estimates.","section":"Section 6.1.1, condition (24)"},{"comment":"There is a spelling error: 'infinte' should read 'infinite'. Also, in the B_φ = 1 paragraph, the phrase 'letting C → B_φ' only makes sense after the correction described in the major comments; in the current text B_φ = 1 and C → 1 is what is needed.","section":"Section 6.6"}],"recommendation":"major_revision","confidential_remarks":"I believe the mathematical content is very likely correct and the paper will be acceptable after a careful revision. The main risk is that a reader of the current version cannot verify Theorem 1.10 for the central range 1 < B_φ < ∞ from the written proof, and Theorem 1.7's definition/proof mismatch affects the foundation of Theorem 1.8. I would not reject the paper; rather, I would ask the authors to rewrite Section 6.6, to reconcile the definition of Ω(ℓ) with the quantity proved in Section 3, and to clean up the smaller statement and notation issues listed above."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a referee's time. The paper proves a zero-one law for products of ℓ consecutive prime partial quotients (Theorem 1.8) and, more substantially, that restricting infinitely many partial quotients to be prime does not change the Hausdorff dimension for arbitrary growth functions (Theorem 1.10). That is a genuine extension of Huang-Wu-Xu and Schindler-Zweimüller. The new analytic ingredient is the tail estimate for the almost prime zeta function (Theorem 1.7), and the lower-bound construction in Section 6 is a careful adaptation of the HWX Cantor set with prime-counting input. I checked the central dimension argument; it holds. The flagged dependence on Proposition 2.4 (explicit prime in [0.999x,x)) is not fragile—such interval primes follow from standard explicit PNT bounds like Dusart—so that worry is overblown.\n\nThe soft spots are real but minor. Theorem 1.7 is stated for the almost prime zeta function over Ω(ℓ) ('at most ℓ' prime factors), but the proof establishes the estimate for the ordered sum over exactly ℓ primes. The at-most version follows because lower ℓ terms are dominated, but the proof should say so. The base case of the induction says s>2; the argument works for s>1. Proposition 2.3 has the fraction inverted (should be γ^n/(n log γ)). And in the proof of Theorem 1.10, the line 'First assume B_φ = 1' should read 'B_φ < ∞', with C > B_φ rather than C > 1; without that correction the finite-growth case is skipped. These are routine fixes, not load-bearing flaws. There is no circularity and no fitting; the citation pattern is normal, including the recent preprint on Chen's theorem, which is only used as a convenient source for a standard explicit prime gap.\n\nWho is this for? Specialists in metric Diophantine approximation and continued fractions. The paper is long and technical, but proofs are detailed enough to verify, and the main results are new and correctly stated after small corrections. Send it to a serious referee; it will need minor revisions, not a rewrite.","headline":"Solid extension of Huang-Wu-Xu to prime partial quotients; main theorems hold, with fixable typos and one statement-proof mismatch in Theorem 1.7.","tokens_in":26420,"tokens_out":4808,"would_cite":true,"duration_ms":48314,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11A55","11K50","28A80","11N05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that requiring infinitely many partial quotients of a continued fraction to be prime leaves the Hausdorff dimension of the associated sets unchanged, and it determines their Lebesgue measure through a zero-one law driven…","keywords":["continued fractions","prime partial quotients","Hausdorff dimension","Lebesgue measure","zero-one law","almost prime zeta function","Borel–Bernstein theorem","Diophantine approximation"],"falsifier":"Exhibiting a single $x>e^{20}$ with no prime in $[0.999x,x)$ would falsify the explicit prime-gap input and invalidate the gap estimates behind the dimension lower bound. Short of that, a direct computation of $\\dim_H E'_2(\\varphi)$ for $\\varphi(n)=2^n$ that disagrees with the predicted pressure value $t_2^{(2)}$ would refute Theorem 1.10.","tokens_in":25387,"feed_emoji":"🔢","tokens_out":10957,"duration_ms":102650,"temperature":0.7,"pith_summary":"The paper studies sets of real numbers whose continued-fraction partial quotients are, infinitely often, both large and prime. Its main claim is that this primality restriction does not change the Hausdorff dimension: for every function $\\varphi$ and every $\\ell$, the dimension of $E'_\\ell(\\varphi)$ equals that of the unrestricted set $E_\\ell(\\varphi)$. It also proves a zero-one law for Lebesgue measure, with the threshold series $\\sum (\\log\\log\\varphi(n))^{\\ell-1}/(\\varphi(n)\\log\\varphi(n))$. The engine behind both results is a new asymptotic for the tail of the almost prime zeta function. If the paper is right, the dimension formulas for large partial quotients carry over verbatim to prime partial quotients, while the Lebesgue-measure threshold changes by replacing one logarithm by a double logarithm.","feed_headline":"Prime partial quotients don't shrink continued-fraction dimension","feed_subtitle":"A new zero-one law and dimension formula show primality preserves fractal dimension while shifting the measure threshold.","key_machinery":"The central analytic object is the almost prime zeta tail $P_\\ell(s;M)$, the sum of $k^{-s}$ over integers $k\\ge M$ having at most $\\ell$ prime factors; the asymptotic of Theorem 1.7 converts prime-counting information into interval-measure estimates. The dimension argument is carried by a Cantor-set construction on fundamental intervals whose distinguished blocks are filled by primes in dyadic ranges, with gaps controlled by the explicit short-interval prime bound of Proposition 2.4 and lengths controlled by the denominator recursion for continued fractions. The Mass Distribution Principle then turns H\\\"older estimates for the constructed measure into the lower bound $s(1-\\delta)-\\delta$, and letting $\\delta\\to 0$ and $s\\to t_B^{(\\ell)}$ yields the equality in Theorem 1.10.","core_discovery":"On the paper's own terms, the discovery is that the metric theory of continued fractions with large partial quotients is unchanged when the allowed partial quotients are restricted to primes. Theorem 1.10 states $\\dim_H E'_\\ell(\\varphi)=\\dim_H E_\\ell(\\varphi)$ for every $\\varphi:\\mathbb{N}\\to\\mathbb{R}_{>0}$ and $\\ell\\in\\mathbb{N}$, so the full dimension formula of the unrestricted case, expressed through the pressure function $t_B^{(\\ell)}$, transfers verbatim to the prime setting. Theorem 1.8 gives $m(E'_\\ell(\\varphi))=0$ or $1$ according as the series $\\sum (\\log\\log\\varphi(n))^{\\ell-1}/(\\varphi(n)\\log\\varphi(n))$ converges or diverges, the prime analogue of the Borel-Bernstein-type law in which the factor $\\log\\varphi(n)$ is replaced by $\\log\\log\\varphi(n)$ and an extra $\\log\\varphi(n)$ appears in the denominator. The supporting analytic input is Theorem 1.7: for fixed $\\ell$ and $s>1$, the tail $P_\\ell(s;M)=\\sum_{k\\in\\Omega(\\ell),\\,k\\ge M}k^{-s}$ is comparable to $(\\log\\log M)^{\\ell-1}/(M^{s-1}\\log M)$.","pith_inferences":["A natural extension, not pursued here, would replace the primes by any sparse set $A$ whose counting function grows regularly, such as squares or powers of $2$; the same mechanism would predict dimension equality whenever every interval $[x,(1+\\varepsilon)x]$ eventually contains an element of $A$.","The contrast between the measure thresholds in the unrestricted and prime cases suggests a heuristic reading: Hausdorff dimension is blind to a $1/\\log p$ density factor, whereas Lebesgue measure is not; this interpretation is an editorial gloss, not a claim of the paper.","A testable intermediate step is to prove the analogue of Proposition 1.12 when the positions of the large prime quotients are prescribed by a sparse infinite set $\\mathcal{N}$; the flexible-block lemma of Section 6.5 already contains most of the technology needed.","Because the proof leans only on a short-interval prime bound with gaps of size $O(x)$ and on the prime-counting rate from the Prime Number Theorem, the dimension equality would likely survive for any allowed-quotient set with the same two features, making primality one instance of a broader phenomenon."],"forward_implications":["The Lebesgue measure of $E'_\\ell(\\varphi)$ is now fully determined for $\\varphi:\\mathbb{N}\\to[3,\\infty)$: the set is null or full according to the convergence or divergence of $\\sum (\\log\\log\\varphi(n))^{\\ell-1}/(\\varphi(n)\\log\\varphi(n))$.","Every Hausdorff dimension formula proved for unrestricted large partial quotients, including the value $1/(b+1)$ for exponential thresholds, holds verbatim when the large partial quotients are required to be prime.","The zero-one law extends to $F_\\ell(\\varphi)$, the version where the threshold is evaluated at the denominator $q_n$ of the convergent rather than at $n$.","Sets with exponentially growing blocks of prime partial quotients have the same Hausdorff dimension as their unrestricted counterparts, as stated in Proposition 1.12.","The tail asymptotic for the almost prime zeta function at $s=2$ is the quantitative reason the measure law takes the form it does, and it is a self-contained number-theoretic byproduct."],"supporting_citations":[{"why":"Supplies the unrestricted Lebesgue-measure and Hausdorff-dimension theorems that this paper extends, together with the pressure-function and Cantor-set machinery used in the proof of Theorem 1.10.","marker":"[21]"},{"why":"Gives the $\\ell=1$ prime Borel-Bernstein law and the tail estimate for the prime zeta function at $s=2$ that Theorem 1.7 generalizes to almost primes.","marker":"[37]"},{"why":"Provides the explicit short-interval prime statement, a prime in $[0.999x,x)$ for $x>e^{20}$, which controls the gaps between fundamental intervals in the dimension construction.","marker":"[1]"},{"why":"Provides the quantitative Borel-Cantelli lemma for the Gauss map that yields the zero-one law for the Lebesgue measure in Theorem 1.8.","marker":"[27]"},{"why":"Supplies explicit upper and lower bounds for the prime-counting function, used to count primes in dyadic ranges in the prime analogue of the exponential-threshold lower bound.","marker":"[36]"},{"why":"Establishes the Hausdorff dimension of $E_1(\\varphi)$ for general $\\varphi$, the base case whose pressure formula is extended to products and to primes.","marker":"[41]"},{"why":"Supplies the regular-variation theorem used to turn prime-counting asymptotics into the tail estimate for the prime zeta function in Theorem 1.7.","marker":"[34]"},{"why":"Provides the Cantor-set lower-bound lemma used to obtain the dimension $1/(b+1)$ in the prime analogue of the classical exponential-threshold result.","marker":"[13]"}],"fun_headline_variants":["Primes preserve continued-fraction dimension","Prime-only quotients keep fractal dimension","Zero-one law for prime continued fractions","Prime partial quotients: dimension same, measure threshold shifts","Dimension formula extends to prime quotients"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that every interval $[0.999x,x)$ with $x>e^{20}$ contains at least one prime; if consecutive primes could be spaced more widely than one tenth of one percent of $x$, the Cantor blocks would have gaps too large for the dimension lower bound to survive.","fun_headline_variants_meta":{"raw":{"variants":["Primes preserve continued-fraction dimension","Prime-only quotients keep fractal dimension","Zero-one law for prime continued fractions","Prime partial quotients: dimension same, measure threshold shifts","Dimension formula extends to prime quotients"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000806,"raw_usage":{"total_tokens":3509,"prompt_tokens":885,"completion_tokens":2624,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":2559}},"tokens_in":501,"tokens_out":2624,"duration_ms":25654,"temperature":1.0,"reasoning_tokens":2559,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T14:55:49.249857+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibiting a single $x>e^{20}$ with no prime in $[0.999x,x)$ would falsify the explicit prime-gap input and invalidate the gap estimates behind the dimension lower bound. Short of that, a direct computation of $\\dim_H E'_2(\\varphi)$ for $\\varphi(n)=2^n$ that disagrees with the predicted pressure value $t_2^{(2)}$ would refute Theorem 1.10.","supporting_citations":[{"cited_title":"Metric properties of the product of consecutive partial quotients in continued fractions","cited_arxiv_id":null,"evidence_quote":"Supplies the unrestricted Lebesgue-measure and Hausdorff-dimension theorems that this paper extends, together with the pressure-function and Cantor-set machinery used in the proof of Theorem 1.10."},{"cited_title":"I., AND ZWEIM ¨ULLER , R","cited_arxiv_id":null,"evidence_quote":"Gives the $\\ell=1$ prime Borel-Bernstein law and the tail estimate for the prime zeta function at $s=2$ that Theorem 1.7 generalizes to almost primes."},{"cited_title":"A zero-one law for improvements to Dirichlet’s Theorem","cited_arxiv_id":null,"evidence_quote":"Provides the quantitative Borel-Cantelli lemma for the Gauss map that yields the zero-one law for the Lebesgue measure in Theorem 1.8."},{"cited_title":"Explicit bounds for some functions of prime numbers","cited_arxiv_id":null,"evidence_quote":"Supplies explicit upper and lower bounds for the prime-counting function, used to count primes in dyadic ranges in the prime analogue of the exponential-threshold lower bound."},{"cited_title":"Hausdorff dimension of certain sets arising in continued fraction expansions","cited_arxiv_id":null,"evidence_quote":"Establishes the Hausdorff dimension of $E_1(\\varphi)$ for general $\\varphi$, the base case whose pressure formula is extended to products and to primes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the regular-variation theorem used to turn prime-counting asymptotics into the tail estimate for the prime zeta function in Theorem 1.7."},{"cited_title":"On the fractional dimension of sets of continued fractions","cited_arxiv_id":null,"evidence_quote":"Provides the Cantor-set lower-bound lemma used to obtain the dimension $1/(b+1)$ in the prime analogue of the classical exponential-threshold result."}],"review_version":1}