{"id":"32a116e5-e604-42da-882a-b383d0704423","arxiv_id":"2411.18439","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The exact simultaneous approximation set in R^2 has Hausdorff dimension 3/(λ+1); the paper supplies an elementary proof for a result already established for all n by Bandi and De Saxcé.","lead":"The paper gives a short proof that the set of real pairs with a prescribed exact rational approximation rate has Hausdorff dimension 3/(λ+1). It matters as a clean completion of a case left open by one method, though the formula itself was already known by a heavier method.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 2.3's divergence estimate contains a reversed inequality: (psi(q)/q)^s >= q^{-(1-epsilon)} does not follow from the definition of lambda and is generally false; the correct lower bound is q^{-(1-epsilon)-s epsilon}, which still diverges, so the proof is fixable.","rationale":"The reader's weakest_assumption correctly identifies the reversed inequality in Proposition 2.3. I have verified the algebra: the definition of lambda gives psi(q) >= q^{-(lambda+epsilon)}, not psi(q) >= q^{-lambda}, so (psi(q)/q)^s is bounded below by q^{-s(lambda+1+epsilon)} = q^{-(1-epsilon)-s epsilon}, which is smaller than the claimed q^{-(1-epsilon)}. This is a genuine gap in the written proof. However, the gap is easily fixed by using the slightly weaker lower bound; the divergence of the corrected series is immediate because the exponent epsilon(1-s) is positive and Q(x1) is infinite. I also considered the reader's other concern about the Mass Transference Principle applied to the sparse, non-monotone function Psi. The cited lemma from Beresnevich-Velani does not require monotonicity of the approximating function; it only requires the divergence of sum f(psi(n)/n) phi(n), so the MTP application is valid. Thus the central theorem is correct, and the paper's original contribution is the new proof. As the proof as printed contains a false inequality, the paper should be accepted only after the authors correct the estimate in Proposition 2.3.","tokens_in":3749,"tokens_out":18331,"duration_ms":153980,"concrete_test":"Independently re-derive the chain in Proposition 2.3 using the two-sided bound q^{-(lambda+epsilon)} <= psi(q) <= q^{-(lambda-epsilon)} implied by lambda = lim -log psi(q)/log q. With s = (1-epsilon)/(lambda+1), compute the minimum of (psi(q)/q)^s as q^{-s(lambda+1+epsilon)} = q^{-(1-epsilon)-s epsilon} and verify it is strictly smaller than q^{-(1-epsilon)}. Then check that the corrected series sum_{q in Q(x1)} q^{-(1-epsilon)-s epsilon} phi(q) diverges because the terms grow like q^{epsilon(1-s)}/log q, so the conclusion dim_H W(x1,1,psi) >= 1/(lambda+1) survives.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Proposition 2.3, the chain is: sum_{q in Q(x1)} (psi(q)/q)^s phi(q) >= sum_{q in Q(x1)} q^{-(1-epsilon)} phi(q) >> sum q^epsilon / log q = infinity, with s = (1-epsilon)/(lambda+1). The first inequality is claimed to hold by the definition of lambda. But lambda = lim -log psi(q)/log q gives both bounds q^{-(lambda+epsilon)} <= psi(q) <= q^{-(lambda-epsilon)} for sufficiently large q. To lower-bound (psi/q)^s one must use the smallest possible psi, q^{-(lambda+epsilon)}; then (psi/q)^s >= q^{-s(lambda+1+epsilon)} = q^{-(1-epsilon)-s epsilon}. This exponent is strictly less than -(1-epsilon), because s epsilon > 0, so q^{-(1-epsilon)-s epsilon} < q^{-(1-epsilon)} for q > 1. Hence the displayed inequality is reversed; the correct lower bound is much smaller than q^{-(1-epsilon)} and the estimate as written is false. The error is not fatal to the theorem: replacing q^{-(1-epsilon)} with q^{-(1-epsilon)-s epsilon} gives a series whose terms are q^{epsilon - s epsilon}/log q = q^{epsilon(1-s)}/log q, which diverges for any infinite Q(x1) because the exponent is positive. Thus the MTP conclusion still follows after a small epsilon loss, but the proof as printed is not sound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that for a function ψ satisfying that qψ(q) is non-increasing and tends to 0 at infinity, with λ = lim_{q→∞} −log ψ(q)/log q, the Hausdorff dimension of the set Exact(2,ψ) of vectors in R² having exact simultaneous approximation order ψ equals 3/(λ+1). The proof combines the one-dimensional exact-order result of Bugeaud and Bugeaud–Moreira with a fibre inclusion inspired by Fregoli, Marstrands' slicing lemma, and the mass transference principle for Duffin–Schaeffer sets. The upper bound is obtained from Rynne's formula for the simultaneous well-approximable set W(2,ψ).","tokens_in":4104,"tokens_out":15300,"duration_ms":125910,"significance":"If correct, the paper fills the n=2 gap in Fregoli's fibre-based method and provides an elementary proof of the exact-order dimension for the simultaneous case in R². The result itself is not entirely new—Bandi–De Saxcé have treated all n via parametric geometry—but the proof here is short and uses different, standard tools. The main ingredients are cited correctly and the overall strategy is natural. However, there is a local but load-bearing error in the divergence estimate in Proposition 2.3, which must be repaired before the proof is sound.","major_comments":[{"comment":"The displayed chain of inequalities in the proof of Proposition 2.3 contains a reversed inequality. Since λ = lim_{q→∞} (−log ψ(q))/log q, for every ε>0 and all sufficiently large q one has q^{−(λ+ε)} ≤ ψ(q) ≤ q^{−(λ−ε)}. Therefore (ψ(q)/q)^s ≥ q^{−s(λ+1+ε)} = q^{−(1−ε)−sε}, which is strictly smaller than the claimed bound q^{−(1−ε)} for q>1. Thus the inequality (ψ(q)/q)^s ≥ q^{−(1−ε)} is false. This is load-bearing because the divergence of ∑_{q∈Q(x1)} (ψ(q)/q)^s φ(q) is what triggers the mass transference principle. The error is local and repairable: replacing q^{−(1−ε)} with q^{−(1−ε)−sε} yields the lower bound ∑_{q∈Q(x1)} q^{ε(1−s)}/log q, which diverges because Q(x1) is infinite and ε(1−s)>0. The proposition's conclusion then follows, but the proof as printed is not sound.","section":"§2, Proposition 2.3"}],"minor_comments":[{"comment":"The definition of W*(1,ψ) contains a stray symbol γ: '||qx − γ||_*' should read '||qx||_*'.","section":"Introduction, definition of W*(1,ψ)"},{"comment":"In the statement of Lemma 2.2 the summation is written as ∑_{q∈N} f(ψ(n)/n) φ(n), mixing the variables q and n; it should be ∑_{q∈N} f(ψ(q)/q) φ(q). Also, since the lemma imposes monotonicity of x^{-1}f(x), the intended direction (non-increasing or non-decreasing) should be specified, as it is relevant to the application f(x)=x^s.","section":"Lemma 2.2"},{"comment":"The inclusion W*(1,Ψ) ⊂ W(x1,1,ψ) is only stated as 'readily checked'. A short verification would improve readability: for any c<1, the exactness of x1 implies that all but finitely many q satisfy ||qx1|| ≥ cψ(q), so points in W*(1,Ψ) cannot lie in W(2,cψ) and therefore lie in Exact(2,ψ).","section":"§2, Proposition 2.3"},{"comment":"The upper bound is derived from Rynne's formula (1.1). If Rynne's theorem is stated under a different monotonicity assumption (for instance, ψ decreasing rather than qψ(q) non-increasing), the exact hypothesis should be cited. Alternatively, the needed inequality dim_H Exact(2,ψ) ≤ 3/(λ+1) follows from the standard covering construction for W(2,ψ) and does not require monotonicity beyond the asymptotic of ψ and ψ(q) ≤ 1/(2q).","section":"§1, upper bound"}],"recommendation":"major_revision","confidential_remarks":"The reversed inequality in Proposition 2.3 is the only serious technical defect I found; it appears to be a slip rather than a fundamental flaw, and it is easily repaired. The result is already known from Bandi–De Saxcé's preprint, so the value of this paper lies in the elementary proof for n=2. I recommend major revision rather than rejection. The authors should correct the inequality and clarify the hypotheses of Lemma 2.2."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper proves Theorem 1.1, but the statement is already in Bandi–De Saxcé [1], which the authors cite. The novelty is the proof, not the result. The proof is short and genuinely simpler than the parametric-geometry route: fiber slicing plus the one-dimensional exact-order result plus MTP. For the n=2 case this is a clean, readable argument.\n\nThe proof as printed has a real slip in Proposition 2.3. The chain\n(ψ(q)/q)^s φ(q) ≥ q^{-(1−ε)} φ(q)\ndoes not follow from the definition of λ. Since λ = lim −log ψ/log q, for large q we have ψ(q) ≥ q^{-(λ+ε)} and ψ(q) ≤ q^{-(λ−ε)}. To lower-bound (ψ/q)^s you need the smallest ψ, which gives (ψ/q)^s ≥ q^{-s(λ+1+ε)} = q^{-(1−ε)−sε}. That exponent is less than −(1−ε), not greater, so the displayed inequality is reversed. The good news: the corrected bound still diverges. The series becomes ∑_{q∈Q(x1)} q^{-(1−ε)−sε} φ(q) ≫ ∑ q^{ε−sε}/log q = ∑ q^{ε(1−s)}/log q, which diverges for any infinite Q(x1) because s<1. So the MTP conclusion survives after replacing the wrong exponent. The error is fixable and the theorem stands, but the written argument is not sound as it stands.\n\nA smaller point: the MTP lemma is cited without checking that the sparse, non-monotone Ψ satisfies its hypotheses. In practice MTP does not require monotonicity of the approximating function, so this is more an exposition gap than a mathematical hole, but it should be addressed.\n\nThe paper is honest: it cites [1] and explicitly says the proof is different. The abstract's \"fill a gap\" refers to Fregoli's approach, which is fair. The citation pattern is clean, no self-citation issue.\n\nWho is this for? Anyone working on exact approximation or wanting an elementary proof of the n=2 dimension formula. The paper does not change the landscape, but it is a useful pedagogical and technical contribution. I would send it to a referee if the authors fix the inequality and add a remark that the result is already known; the referee should check the corrected exponent.\n\nRecommendation: engage with it after revision.","headline":"Elementary proof of a known dimension formula; the printed divergence estimate has a fixable inequality slip, but the argument's core is sound.","tokens_in":4598,"tokens_out":3440,"would_cite":false,"duration_ms":29406,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","11K55","11J83"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the set of pairs in the plane with exact simultaneous approximation order ψ has Hausdorff dimension 3/(λ+1), the same as the full well-approximable set.","keywords":["Hausdorff dimension","exact approximation order","simultaneous Diophantine approximation","Mass Transference Principle","Marstrand slicing lemma","fiber dimension","well-approximable set"],"falsifier":"Take $\\psi(q)=q^{-\\lambda}$ for some $\\lambda>1$ and an explicit $x_1\\in\\mathrm{Exact}(1,\\psi)$ whose approximating denominator set $Q(x_1)$ is known, then evaluate the series $\\sum_{q\\in Q(x_1)}(\\psi(q)/q)^s\\varphi(q)$ at $s=(1-\\varepsilon)/(\\lambda+1)$. A convergent value for any such series would break the proof's divergence claim; a direct computation of $\\dim_H \\mathrm{Exact}(2,\\psi)$ differing from $3/(\\lambda+1)$ for any $\\lambda$ would disprove the theorem.","tokens_in":3548,"feed_emoji":"📐","tokens_out":11734,"duration_ms":86373,"temperature":0.7,"pith_summary":"This paper proves that in the plane, the set of pairs whose simultaneous Diophantine approximation by rationals has exactly the prescribed error order $\\psi$ has Hausdorff dimension $3/(\\lambda+1)$, where $\\lambda$ is the growth exponent of $\\psi$. This is the same dimension as the well-approximable set $W(2,\\psi)$, so requiring the approximation to be exact rather than merely at least as good does not shrink the fractal size. The result fills the last unproved case $n=2$ in the family of dimension formulas for Exact($n,\\psi$), after the one-dimensional case and the cases $n\\ge 3$ were known. The proof splits into a lower bound using Marstrand's slicing lemma together with a mass transference argument, and an upper bound inherited from the dimension formula for $W(2,\\psi)$.","feed_headline":"Exact simultaneous approximants in R² get full dimension 3/(λ+1)","feed_subtitle":"This closes the n=2 case: exactness does not shrink the Hausdorff dimension of well-approximable pairs.","key_machinery":"The proof rests on two tools. Marstrand's Slicing Lemma says that if a set $E\\subset X\\times Y$ has sections of dimension at least $t$ over a base of dimension $s$, then $\\dim_H E\\ge s+t$. The Mass Transference Principle says that for a dimension function $f$ with $f(x)/x$ monotonic, if the series $\\sum_q f(\\psi(q)/q)\\varphi(q)$ diverges, then the Hausdorff $f$-measure of the set of numbers with a coprime approximation of quality $\\psi$ is full. The paper's twist is to feed the principle the auxiliary function $\\Psi(q)=\\psi(q)$ for $q\\in Q(x_1)$ and $0$ otherwise, where $Q(x_1)$ is the set of denominators for which $x_1$ is $\\psi$-approximable; this proves the fiber lower bound. The divergence of the relevant series then follows from the definition of $\\lambda$ and the standard estimate for products over primes.","core_discovery":"The central claim is Theorem 1.1: if $q\\psi(q)$ is non-increasing and tends to zero at infinity, then $\\dim_H \\mathrm{Exact}(2,\\psi)=3/(\\lambda+1)$, where $\\lambda := \\lim_{q\\to\\infty} -\\log \\psi(q)/\\log q$. The discovery is that exactness does not reduce the dimension in two dimensions: for every $x_1\\in \\mathrm{Exact}(1,\\psi)$, the fiber of $x_2$ for which $(x_1,x_2)$ has exact order $\\psi$ has dimension at least $1/(\\lambda+1)$, and Marstrand's slicing lemma converts this fiber lower bound into the full lower bound. The upper bound is immediate from the inclusion $\\mathrm{Exact}(2,\\psi)\\subset W(2,\\psi)$ and the known dimension of $W(2,\\psi)$. The argument works by thinning $\\psi$ to an auxiliary function $\\Psi$ supported only on the denominators $q$ for which $qx_1$ is already $\\psi$-close, so that the mass transference principle can be applied to a subset of the fiber.","pith_inferences":["If the sparse-support trick can be made fully rigorous, the same induction through fibers could establish the exact-dimension formula for all $n$ without relying on parametric geometry of numbers.","The equality of dimensions suggests a general principle: in simultaneous Diophantine approximation, exactness is dimension-neutral, deleting the faster-than-$\\psi$ approximation layers without changing the fractal exponent.","A testable extension is to allow coordinate-dependent error functions $\\psi_1,\\psi_2$; the natural conjecture is that $\\dim_H\\mathrm{Exact}(2,(\\psi_1,\\psi_2))$ is governed by the slower error function."],"forward_implications":["For $n=2$, the set of exactly approximable pairs has the same Hausdorff dimension as the full well-approximable set $W(2,\\psi)$.","Combined with the known cases $n=1$ and $n\\ge 3$, the formula $\\dim_H\\mathrm{Exact}(n,\\psi)=(n+1)/(\\lambda+1)$ now holds for every dimension $n\\ge 1$ under the same monotonicity assumption.","For every $x_1\\in\\mathrm{Exact}(1,\\psi)$, the fiber of $x_2$ with $(x_1,x_2)\\in\\mathrm{Exact}(2,\\psi)$ has Hausdorff dimension at least $1/(\\lambda+1)$.","The upper bound requires no new work: it follows from the inclusion $\\mathrm{Exact}(2,\\psi)\\subset W(2,\\psi)$ and the known dimension of $W(2,\\psi)$."],"supporting_citations":[{"why":"Supplies the Mass Transference Principle used to derive the fiber lower bound from a divergent series.","marker":"[2]"},{"why":"Establishes the one-dimensional exact approximation order dimension used to identify the base set Exact(1,ψ).","marker":"[3, 4, 5]"},{"why":"Provides Marstrand's Slicing Lemma that combines the base dimension with the fiber dimension.","marker":"[8]"},{"why":"Proves the higher-dimensional case n≥3, leaving n=2 as the gap this paper fills.","marker":"[9]"},{"why":"Gives the dimension formula for the well-approximable set W(2,ψ) that supplies the upper bound.","marker":"[16]"}],"fun_headline_variants":["Exactness doesn't shrink dimension in R²","Exact R² pairs: full dimension 3/(λ+1) proven","Dimension of exact simultaneous approximants in R² settled","No dimension loss for exact approximation in R²"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The lower-bound argument applies the Mass Transference Principle to the auxiliary function $\\Psi$, which is zero except on the possibly very sparse set $Q(x_1)$ of denominators that approximate $x_1$. The paper does not spell out that this sparse, non-monotone function meets the principle's hypotheses, and the displayed divergence estimate in Proposition 2.3 contains a reversed inequality.","fun_headline_variants_meta":{"raw":{"variants":["Exactness doesn't shrink dimension in R²","Exact R² pairs: full dimension 3/(λ+1) proven","Dimension of exact simultaneous approximants in R² settled","No dimension loss for exact approximation in R²"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000886,"raw_usage":{"total_tokens":3765,"prompt_tokens":822,"completion_tokens":2943,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":438,"completion_tokens_details":{"reasoning_tokens":2875}},"tokens_in":438,"tokens_out":2943,"duration_ms":19322,"temperature":1.0,"reasoning_tokens":2875,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:13:08.463687+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $\\psi(q)=q^{-\\lambda}$ for some $\\lambda>1$ and an explicit $x_1\\in\\mathrm{Exact}(1,\\psi)$ whose approximating denominator set $Q(x_1)$ is known, then evaluate the series $\\sum_{q\\in Q(x_1)}(\\psi(q)/q)^s\\varphi(q)$ at $s=(1-\\varepsilon)/(\\lambda+1)$. A convergent value for any such series would break the proof's divergence claim; a direct computation of $\\dim_H \\mathrm{Exact}(2,\\psi)$ differing from $3/(\\lambda+1)$ for any $\\lambda$ would disprove the theorem.","supporting_citations":[{"cited_title":"Beresnevich and S","cited_arxiv_id":null,"evidence_quote":"Supplies the Mass Transference Principle used to derive the fiber lower bound from a divergent series."},{"cited_title":"Falconer, Fractal geometry","cited_arxiv_id":null,"evidence_quote":"Provides Marstrand's Slicing Lemma that combines the base dimension with the fiber dimension."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the higher-dimensional case n≥3, leaving n=2 as the gap this paper fills."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the dimension formula for the well-approximable set W(2,ψ) that supplies the upper bound."}],"review_version":1}