{"id":"420fb8ba-d4e0-4dc5-ab36-0d0ebca4d392","arxiv_id":"2509.09945","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For subcritical almost Mathieu operators with Diophantine frequency, each exceptional lower-scaling-index level set has Hausdorff dimension 0 but log-Hausdorff dimension 1.","lead":"Rare exceptional energies in the absolutely continuous spectrum of the subcritical almost Mathieu operator are invisible to classical Hausdorff dimension yet full-size in a logarithmic gauge, and the paper computes both descriptions exactly. The proof reduces the spectral question to a Diophantine approximation problem for irrational rotations and solves it with a Cantor construction.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 2.1, the imported two-sided IDS estimate, is the sole bridge between spectral level sets and Diophantine sets; the lower bound c ε^(3/2) is not located or proved in the text, and if it is absent from Avila [1], Theorems 1.1–1.2 collapse.","rationale":"The reader's weakest_assumption is exactly Proposition 2.1, and my reading agrees: this two-sided estimate is the only bridge connecting the spectral level sets Σ(β) to the Diophantine sets D(δ). Without the lower bound c ε^{3/2}, the comparison in Proposition 1.1 lacks both directions, so the zero-infinity dichotomy of Theorem 1.2 and the consequent Theorem 1.1 would not be established. The other issues noted by the reader—the cover refinement in Proposition 1.1 and small constant slips in the mass-distribution construction—are real but appear patchable; they do not by themselves threaten the main conclusion as directly as the unverified import. The Cantor-set construction for D(δ) in Section 3 is detailed and largely self-contained, and I did not find a fatal internal inconsistency there. Therefore I do not recommend changing the reader's CONDITIONAL verdict, but the condition should explicitly require verification of the lower half of Proposition 2.1 from Avila [1] (or a proof supplied in the manuscript).","tokens_in":17460,"tokens_out":39335,"duration_ms":697776,"concrete_test":"Locate and transcribe the exact statement in Avila's paper arXiv:0810.2965 that is cited as Proposition 2.1. Determine whether it contains the lower bound c ε^{3/2} for all E in the spectrum with a uniform c, or whether it contains only the upper 1/2-Hölder bound. If the lower bound is not present, independently re-derive it from the reducibility and quantitative continuity estimates of [1]; a single Diophantine pair (λ,α) for which the bound fails would invalidate Proposition 1.1. A supporting numerical probe, though not conclusive, is to compute finite-volume IDS for a Diophantine α, e.g., α=(√5−1)/2 and λ=0.5, over a fine grid of energies E, and check whether the ratio [N(E+ε)−N(E−ε)]/ε^{3/2} stays bounded below uniformly in E and ε=10^{-k}; if it decays to zero along some energies, the claimed uniform lower bound is false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central reduction is Proposition 1.1, which asserts H^{ω_s}(F(δ)) and H^{ω_s}(D(δ)) are comparable under the gauge ω_s. Its only quantitative input is Proposition 2.1, attributed to Avila [1]: for α∈DC and 0<λ<1, c ε^{3/2} ≤ N(E+ε)-N(E-ε) ≤ c^{-1} ε^{1/2} for every E in the spectrum and 0<ε<1. The upper bound is a standard 1/2-Hölder estimate, but the lower bound c ε^{3/2} is the crucial non-standard ingredient: it is what lets the proof bound the preimage intervals I_i in Proposition 1.1 and obtain the two-sided comparison (1.3). If the lower bound fails, or is only an upper estimate, then the cover arguments in Section 2 do not go through, and neither Theorem 1.2 nor Theorem 1.1 follows. The paper gives no proof and no theorem number or specific location within [1], and the notation c is reused in Section 3 for a different annulus parameter, making the citation easy to misread. Since the construction of the Cantor set C⊂D(δ) in Section 3 is independent of the spectral input and appears internally sound, the imported IDS estimate is the single most load-bearing unverified premise.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the level sets of the lower local dimension of the absolutely continuous spectral measure of the subcritical almost Mathieu operator with Diophantine frequency. It claims (Theorem 1.1) that the classical multifractal spectrum is f_mu(beta)=1 for beta=1 and f_mu(beta)=0 for beta in [1/2,1), and (Theorem 1.2) that in the logarithmic gauge one has H^{omega_s}(Sigma_{lambda,alpha}(beta))=0 for s>1 and =infinity for s<=1, so the logarithmic Hausdorff dimension of every exceptional level set is 1. The proof consists of two parts: a spectral reduction (Proposition 1.1) that compares the level sets of the IDS with the Diophantine sets D(delta) via a two-sided IDS regularity estimate, and a purely Diophantine construction (Theorem 1.3) that analyzes D(delta) using continued fractions, covering arguments, and a mass distribution on a nested Cantor set.","tokens_in":17800,"tokens_out":36497,"duration_ms":320776,"significance":"If the central estimates hold, the paper gives a complete answer to a refined version of the Tang-Kohmoto conjecture for the subcritical almost Mathieu operator: the absolutely continuous component is dominated by energies of scaling index 1, while the exceptional sets, invisible to classical Hausdorff dimension, are logarithmically as large as the spectrum. The Diophantine result Theorem 1.3 is of independent interest and is proved by an explicit, parameter-free construction with no fitting or normalization. The arithmetic core in Sections 3.1-3.6 is coherent, uses standard continued-fraction separation, and the mass-distribution argument is checkable with modest constant adjustments. The main risks are the imported IDS estimate and one unproved covering refinement in the spectral reduction.","major_comments":[{"comment":"The two-sided IDS estimate c epsilon^{3/2} <= N(E+epsilon)-N(E-epsilon) <= c^{-1} epsilon^{1/2} is the only quantitative bridge between the spectral level sets and the Diophantine sets D(delta), and the lower bound is used critically in the first inequality of (1.3) at the step omega_s(|N(I_i)|) <= omega_s(c^{-1}|I_i|^{1/2}). The paper cites [1] but gives no proof, no theorem number, and no page or section within [1]. If the lower bound is not actually established in [1] for every E in the spectrum and every 0<epsilon<1, then Proposition 1.1 and hence Theorems 1.1 and 1.2 do not follow. Please provide a self-contained proof of the lower bound or an exact pointer to the statement in [1], and state explicitly how the cited result implies the uniform version used here.","section":"Section 2, Proposition 2.1"},{"comment":"The proof fixes a countable cover (J_i) of D(delta) by intervals with |N^{-1}(J_i)|<1/2, but no argument is given that such a refinement exists. This condition is not a consequence of smallness of |J_i| when N has flat pieces: if J_i contains a gap label k alpha mod Z, then N^{-1}(J_i) contains the whole corresponding spectral gap, whose length is not controlled by |J_i|. Since later bounds such as |I_i| <= (c/6)^{2/3} depend on |N^{-1}(J_i)|<1/2, the estimate H^{omega_s}(F(delta)) <= 3^{s+1} H^{omega_s}(D(delta)) is not established as written. A lemma on refining covers of D(delta), or a separate treatment of delta=infinity, is needed.","section":"Section 2, proof of the second inequality of Proposition 1.1"}],"minor_comments":[{"comment":"The sentence 'Theorem 1.1 follows directly from Theorem 1.2' skips a comparison step: from H^{omega_s}(Sigma(beta))=0 for s>1 one obtains H^{r^t}(Sigma(beta))=0 for every t>0 because r^t <= C_s omega_s(r) for small r, and the s<=1 part alone does not imply zero classical Hausdorff dimension. Please spell out this argument.","section":"Section 3.8"},{"comment":"The constant c in Proposition 2.1 is reused as the inner-radius parameter of the annuli in Section 3.2; these are different constants with different roles. Using different letters would prevent confusion when checking the constants in the proof of Proposition 1.1.","section":"Sections 1.2 and 3.2"},{"comment":"The reduction to t_0 in [1/2,2/3] is not justified; when t_0 in [1/3,1/2), the argument should use the radius E_i^* - E_i^1 rather than E_i^2 - E_i^*. A short sentence explaining this symmetric case would make the proof complete.","section":"Section 2, Case 1 of the proof of Proposition 1.1"},{"comment":"The convention that delta(alpha,phi)=infinity if phi is congruent to k alpha mod Z should be reconciled with the limsup definition: for Diophantine alpha the orbit points themselves have limsup of (-log ||phi-k alpha||)/|k| equal to 0, not infinity. Since the identification F(infinity)=N^{-1}({k alpha}) is used through (1.2) for beta=1/2, the intended definition should be stated explicitly and consistently.","section":"Section 1, definition of D(delta)"},{"comment":"The gauge function omega_s(r)=(-log r)^{-s} is undefined at r=1; since Hausdorff covers use intervals of diameter at most epsilon<1, this is harmless, but it would be cleaner to state that omega_s is defined on [0,1) or to specify a value at 1.","section":"Section 3.6 and definition of omega_s"},{"comment":"In the definition delta_k = min{delta, log log k}, the case delta=infinity should be interpreted as min{infinity, log log k}=log log k; stating this explicitly would avoid ambiguity for readers of the construction.","section":"Section 3.4, proof of Lemma 3.4"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially a significant contribution, but the verdict hinges on the provenance of Proposition 2.1. If the lower bound c epsilon^{3/2} is not in Avila's paper [1], the spectral reduction collapses and the manuscript should not be published in its present form. If it is in [1], the missing pointer and the cover-refinement gap in Proposition 1.1 are fixable in revision. I was unable to verify the exact statement of [1] from the text alone."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"A couple things you should know before spending time on arXiv:2509.09945. The genuinely new result is Theorem 1.3, and it is the part worth refereeing. The zero-infinity dichotomy for the exact-approximation sets D(δ) under the logarithmic gauge, for every irrational rotation, is not a rerun of known shrinking-target results, and the Cantor-set construction in Section 3 is real work. If Theorem 1.3 is correct, then the transfer to the AMO level sets via Proposition 1.1 gives a complete multifractal description in both power and logarithmic gauges, which is what the Tang–Kohmoto heuristics called for. That transfer is plausible, and the paper deserves a careful referee.\n\nThe arithmetic core—separation lemma, covering counts, mass distribution—checks out in broad outline. I did not find a real gap there. The upper bound H^{ω_s}(D(δ))=0 for s>1 is a straightforward Borel–Cantelli argument; the lower bound is the subtle part, and the mass construction looks coherent. There are minor slips: the constant c in Section 3 has nothing to do with the c of Proposition 2.1; some indices in Proposition 3.1 look typo'd (q_i should be q_l); and the proof of Theorem 1.1 is compressed.\n\nThe real soft spot is Proposition 2.1. The paper needs the lower bound c ε^{3/2} ≤ N(E+ε)-N(E-ε), uniformly over the spectrum, and it cites [1] without a theorem number or a proof. The upper bound ε^{1/2} is standard, but the lower bound is doing all the work in Proposition 1.1. If the estimate is not actually in Avila's paper, the bridge between spectral level sets and the Diophantine sets D(δ) does not stand. This is not a manufactured concern; the stress-test note has the right target. My guess is the estimate is true and available from Avila's reducibility machinery, but the authors need to state it precisely and either prove it or point to the exact statement. I would not desk-reject over this, but I would make it a condition of acceptance.\n\nAlso, in Proposition 1.1, the restriction to covers with |N^{-1}(J_i)|<1/2 is asserted with no justification. It is probably fixable by splitting intervals, but as written it is a gap in the presentation. Same for restricting covers of F(δ) to intervals of diameter ≤ c^6; that one is standard, but the preimage restriction is the piece to watch.\n\nWho is this for? People working on spectral measures, multifractal analysis, or metric Diophantine approximation. It is a solid paper with one load-bearing citation to verify. Send it to a referee who can check Proposition 2.1; if that holds, Theorem 1.2 has a good chance of being right. I would accept it for peer review.","headline":"The Cantor-set construction behind the new Theorem 1.3 is solid and worth refereeing; before accepting, verify that Avila actually proves the c ε^{3/2} lower bound cited as Proposition 2.1.","tokens_in":18315,"tokens_out":10444,"would_cite":true,"duration_ms":95444,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A78","28A80","37C45","47B36","11K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the exceptional energy level sets of the absolutely continuous spectral measure of the subcritical almost Mathieu operator, though classically invisible, are logarithmically as large as the entire spectrum, with a…","keywords":["almost Mathieu operator","absolutely continuous spectral measure","multifractal spectrum","log-Hausdorff dimension","local dimension","Diophantine approximation","integrated density of states","continued fractions"],"falsifier":"Fix $\\alpha=(\\sqrt{5}-1)/2$ and $\\lambda=1/2$. Numerically approximate the $\\omega_1$-Hausdorff measure of the set of energies with lower local dimension $1/2$ by optimizing covers with interval lengths below $\\varepsilon$; the theorem predicts the $\\varepsilon\\to0$ sums diverge to $\\infty$. A finite upper bound for any such cover sequence at $\\beta=1/2$, or any observed local dimension outside $[1/2,1]$, would refute Theorem 1.2.","tokens_in":17288,"feed_emoji":"📏","tokens_out":9746,"duration_ms":81032,"temperature":0.7,"pith_summary":"The paper determines the full multifractal spectrum of the absolutely continuous spectral measure of the subcritical almost Mathieu operator at Diophantine frequencies. It proves that the local scaling dimension is 1 almost everywhere, while every exceptional level set $\\Sigma_{\\lambda,\\alpha}(\\beta)$ with $\\beta\\in[1/2,1)$ has classical Hausdorff dimension 0. In the finer logarithmic gauge, however, each such set has log-Hausdorff dimension 1, with $H^{\\omega_s}(\\Sigma_{\\lambda,\\alpha}(\\beta))=\\infty$ for $s\\le 1$ and $0$ for $s>1$. This sharpens the Tang–Kohmoto conjecture: the exceptional energies are negligible under power-law rulers yet as large as the spectrum under logarithmic rulers.","feed_headline":"Spectral outliers fill a full logarithmic dimension","feed_subtitle":"For subcritical almost Mathieu operators, exceptional energy sets vanish in power law but fill the full log dimension.","key_machinery":"The machinery has three layers. First, the resonance strength $\\delta(\\alpha,\\varphi)=\\limsup_{|k|\\to\\infty}-\\log\\|\\varphi-k\\alpha\\|_{\\mathbb{R}/\\mathbb{Z}}/|k|$ defines Diophantine sets $D(\\delta)=\\{x:\\delta(\\alpha,x)=\\delta\\}$, and the identity $\\Sigma_{\\lambda,\\alpha}(\\beta)=F\\bigl(\\frac{\\beta\\log\\lambda}{1-2\\beta}\\bigr)$ from [47] ties spectral level sets to these Diophantine sets through the integrated density of states. Second, Proposition 1.1 transfers $\\omega_s$-Hausdorff measures between $D(\\delta)$ and $F(\\delta)$ using the two-sided Hölder bounds $c\\varepsilon^{3/2}\\le N(E+\\varepsilon)-N(E-\\varepsilon)\\le c^{-1}\\varepsilon^{1/2}$ on the IDS. Third, for $D(\\delta)$ itself, a Cantor subset is built from annuli centered at $k\\alpha$ at continued-fraction scales, using separation and uniform distribution of the orbit, and a mass-distribution argument shows $H^{\\omega_1}(C)=\\infty$.","core_discovery":"The central discovery is a zero-infinity dichotomy for the exceptional level sets of the absolutely continuous spectral measure. For $0<\\lambda<1$ and $\\alpha\\in\\mathrm{DC}$, the level set $\\Sigma_{\\lambda,\\alpha}(\\beta)=\\{E\\in\\Sigma_{\\lambda,\\alpha}: \\underline{d}_{\\mu}(E)=\\beta\\}$ satisfies $H^{\\omega_s}(\\Sigma_{\\lambda,\\alpha}(\\beta))=0$ for $s>1$ and $H^{\\omega_s}(\\Sigma_{\\lambda,\\alpha}(\\beta))=\\infty$ for $s\\le 1$, for every $\\beta\\in[1/2,1)$, hence $\\dim_{H,\\log}\\Sigma_{\\lambda,\\alpha}(\\beta)=1$. In classical gauges, the same level sets have Hausdorff dimension 0 for $\\beta<1$ and dimension 1 only at $\\beta=1$. The proof identifies these spectral level sets with sets of points on the circle that are approximated by the rotation orbit $\\{k\\alpha\\}$ with prescribed exponential rate, then computes the logarithmic size of those Diophantine sets for every irrational frequency.","pith_inferences":["A natural extension is to replace the exponential rate by a general shrinking function $\\psi(k)$ and ask whether the corresponding exact-approximation set still has log-Hausdorff dimension 1; the annulus construction here suggests the answer may be governed solely by the logarithmic decay rate.","Because Theorem 1.3 holds for every irrational $\\alpha$ but Theorem 1.2 needs Diophantine $\\alpha$, the only obstacle to extending the spectral result is the IDS regularity in Proposition 2.1; improving that two-sided bound for Liouville or other frequencies would transfer the full dichotomy.","The zero-infinity threshold at $s=1$ suggests a further refinement by iterated-logarithmic gauges such as $(\\log\\log 1/r)^{-t}$, which may separate the $\\omega_1$-infinite level sets into a finer hierarchy.","If the measure-transfer argument of Proposition 1.1 is robust, the same multifractal picture may hold for any one-dimensional quasiperiodic operator whose absolutely continuous spectral measure satisfies identical two-sided IDS power bounds."],"forward_implications":["Every exceptional level set $\\Sigma_{\\lambda,\\alpha}(\\beta)$, $\\beta\\in[1/2,1)$, carries infinite $\\omega_1$-Hausdorff measure, so the log-Hausdorff multifractal spectrum is a flat value 1 across the whole range of attainable lower local dimensions.","The $\\beta=1$ level set has full Lebesgue measure because the spectral measure is absolutely continuous, confirming that the spectrum's absolutely continuous component is dominated by points with trivial scaling index.","The critical gauge exponent is $s=1$ uniformly in $\\beta$: changing the logarithmic power by any positive amount flips the Hausdorff measure between zero and infinity.","As a standalone arithmetic statement, for every irrational rotation the set of points approximated by the orbit with a prescribed exponential rate has log-Hausdorff dimension 1.","The classical multifractal spectrum is the two-point function $f_\\mu(1)=1$ and $f_\\mu(\\beta)=0$ for $\\beta\\in[1/2,1)$."],"supporting_citations":[{"why":"Supplies pure absolute continuity of the spectrum and Proposition 2.1, the two-sided Hölder bounds on the IDS that drive the measure transfer.","marker":"[1]"},{"why":"Supplies the identity $\\Sigma_{\\lambda,\\alpha}(\\beta)=F(\\beta\\log\\lambda/(1-2\\beta))$ and the bounds $\\underline{d}_{\\mu}(E)\\in[1/2,1]$, converting spectral level sets into Diophantine level sets.","marker":"[47]"},{"why":"Supplies the continued-fraction separation lemma for denominators $q_n$, used to keep resonant orbit points apart at the relevant scales.","marker":"[38]"},{"why":"Supplies the discrepancy estimate for irrational rotations used to count how many orbit points fall into each annulus.","marker":"[41]"},{"why":"Supplies the mass distribution principle used to prove $H^{\\omega_1}(C)=\\infty$ for the Cantor subset.","marker":"[24]"}],"fun_headline_variants":["Subcritical AMO: exceptional sets fill log dimension","Zero-infinity dichotomy in spectral log-dimension","Log-dimension 1 for exceptional energies of AMO","Spectral outliers: zero power, full logarithmic measure","AMO multifractality: log-Hausdorff completes the picture"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on a uniform two-sided power-law control of the integrated density of states at every spectrum energy; if some energy had intervals whose IDS increment fell outside the $c\\varepsilon^{3/2}$ to $c^{-1}\\varepsilon^{1/2}$ range at arbitrarily small scales, the bridge from spectral level sets to Diophantine sets would break.","fun_headline_variants_meta":{"raw":{"variants":["Subcritical AMO: exceptional sets fill log dimension","Zero-infinity dichotomy in spectral log-dimension","Log-dimension 1 for exceptional energies of AMO","Spectral outliers: zero power, full logarithmic measure","AMO multifractality: log-Hausdorff completes the picture"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000847,"raw_usage":{"total_tokens":3652,"prompt_tokens":880,"completion_tokens":2772,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":2692}},"tokens_in":496,"tokens_out":2772,"duration_ms":18191,"temperature":1.0,"reasoning_tokens":2692,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:00:48.179279+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix $\\alpha=(\\sqrt{5}-1)/2$ and $\\lambda=1/2$. Numerically approximate the $\\omega_1$-Hausdorff measure of the set of energies with lower local dimension $1/2$ by optimizing covers with interval lengths below $\\varepsilon$; the theorem predicts the $\\varepsilon\\to0$ sums diverge to $\\infty$. A finite upper bound for any such cover sequence at $\\beta=1/2$, or any observed local dimension outside $[1/2,1]$, would refute Theorem 1.2.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the continued-fraction separation lemma for denominators $q_n$, used to keep resonant orbit points apart at the relevant scales."},{"cited_title":"Kuipers and H","cited_arxiv_id":null,"evidence_quote":"Supplies the discrepancy estimate for irrational rotations used to count how many orbit points fall into each annulus."},{"cited_title":"Falconer , Fractal geometry , John Wiley & Sons, Ltd., Chichester, 1990","cited_arxiv_id":null,"evidence_quote":"Supplies the mass distribution principle used to prove $H^{\\omega_1}(C)=\\infty$ for the Cantor subset."}],"review_version":2}