{"id":"004b6537-5602-4bdc-b92c-b46d1da45db2","arxiv_id":"2505.07129","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"The paper gives explicit constructions of half-line and whole-line Schrödinger operators with prescribed spectral fractal dimensions, but Theorem 1.1's proof has a false step.","lead":"This math paper constructs Schrödinger operators with unusual spectral fractal dimensions, including a whole-line operator that is 'large' while every half-line cut is 'small'. A central proof in the first construction contains a false inequality, so the paper's full set of claims is not supported as written.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1's proof rests on ω(L)≥L^L; for the §3.1 potential this product is only O(L), so the packing-dimension-zero conclusion is unsupported.","rationale":"The reader's rejection is justified by the failure of the proof of Theorem 1.1. The specific false inequality (k+1)^{k+1}≥L^L is concrete, but the deeper problem is that the proof's central estimate (3.1.6) requires ω(L)≥L^L, while the construction yields ω(L)=O(L): the Gram determinant bound shows the max/min product is linear in L, independent of how large the sparse potential values are chosen. Thus the packing-dimension-zero claim is not just under-proved; the proposed mechanism cannot work for this operator. Theorems 1.3 and 1.4 rely on different constructions and arguments, so they may well be correct, but the paper as submitted contains a load-bearing error in one of its three abstract-level results. The rejection stands; a repair would need a substantially different potential or a different lower-bound mechanism for ω(L).","tokens_in":17592,"tokens_out":47140,"duration_ms":471784,"concrete_test":"Analytic check: take L=8, E=0, and the §3.1 potential with V(4)=A and V(n)=0 otherwise. Let v_k=(u^{(1)}(k),u^{(2)}(k)) for the two normalized solutions, form A_8=∑_{k=1}^8 v_k v_k^T, and compute ω(8)=√det A_8 via Cauchy–Binet. Because the free transfer is a rotation, every 2×2 minor W_{ij} is O(1) with the A-terms cancelling, so det A_8=O(64) independent of A. Thus ω(8)=O(8), while 8^8≈1.7×10^7; this directly falsifies the inequality ω(L)≥L^L used in (3.1.6). The same cancellation holds on every single-spike interval, showing no choice of V(n) in the current construction can make the product super-polynomial.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.1 depends on (3.1.6): it requires ω(L(ε))=1/ε ≥ L(ε)^{L(ε)} to convert Proposition 2.18 and (2.3.11) into Im mθ(E+iε)≥ε^{-t}. The paper tries to get this from Corollary 3.2, but that corollary only lower-bounds max_η∥u_η∥_L by (k+1)^{k+1} with k=⌊√L⌋; the final inequality (k+1)^{k+1}≥L^L is false (e.g. L=8,k=2: 27<8^8). More fundamentally, the proof needs the product ω(L)=max·min, and Corollary 3.2 gives no lower bound on the minimum. This is not a repairable numerical slip: for the constructed potential (zero background, free elliptic propagation between sparse spikes), the Gram matrix A_L=∑_{k=1}^L v_k v_k^T has determinant O(L²) independent of the spike heights, because the 2×2 minors W_{ij} in the Cauchy–Binet expansion are O(1). Hence ω(L)=√det A_L=O(L), not ≥L^L. Estimate (3.1.6) therefore fails for the operator actually constructed, and the packing-dimension-zero conclusion of Theorem 1.1 does not follow. Theorems 1.3–1.4 are supported by separate arguments, but the present version of Theorem 1.1 is unsound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs one-dimensional Schr\\\"odinger operators with unusual fractal continuity properties of their spectral measures. It claims three main results: (1) a half-line operator with essential spectrum [-2,2] whose spectral measure has packing dimension zero for every boundary condition (Theorem 1.1); (2) a whole-line operator whose spectral measure has Hausdorff dimension one while every half-line restriction has Hausdorff dimension zero for every boundary condition (Theorem 1.3); and (3) for the same operator, a Borel set carrying positive whole-line spectral measure but zero measure with respect to every positive half-line restriction (Theorem 1.4). The proofs use transfer matrices, subordinacy theory, Borel-transform estimates from [17], and a construction with sparse large potential spikes. The paper is written in a conventional spectral-theory style, but several load-bearing estimates in the proofs of the main theorems are not justified as written.","tokens_in":17938,"tokens_out":5461,"duration_ms":57407,"significance":"If Theorems 1.3 and 1.4 held, they would provide striking illustrations of the possible discontinuity between whole-line and half-line spectral fractal dimensions, and they would confirm a claim attributed to [16]. The constructions are not obtained by curve fitting or parameter search; they are explicit inductive potential constructions built on subordinacy theory, which is a genuine strength. However, the first headline result, Theorem 1.1, is central to the paper's advertised contribution, and its proof contains a plainly false inequality and an unjustified product estimate. The proof of Theorem 1.3 also appears to use Proposition 4.3 in a way that does not follow from the stated dichotomy. These are not presentation issues; they affect the validity of the paper's main claims.","major_comments":[{"comment":"The final step of Corollary 3.2 states that 'clearly (k+1)^{k+1} ≥ L^L', where k is chosen with k^2 ≤ L < (k+1)^2. This inequality is false for L ≥ 4; for instance L=8, k=2 gives 3^3=27 < 8^8. Therefore the corollary does not establish max_{u∈Sol(E)} ∥u∥_L ≥ L^L, and the subsequent use of this bound in the proof of Theorem 1.1 is invalid.","section":"§3.1, Corollary 3.2"},{"comment":"The proof needs ω(L(ε)) ≥ L(ε)^{L(ε)}, where ω(L) = max_η ∥u_η∥_L · min_η ∥u_η∥_L. Corollary 3.2, even if corrected, would only lower-bound the maximum factor; the proof provides no lower bound on the minimum factor. Without a bound on the product, the chain leading to Im m_θ(E+iε) ≥ ε^{-t} collapses. Moreover, for the potential actually constructed (zero background with sparse spikes), a Gram-matrix determinant estimate indicates that ω(L) grows only polynomially in L, so the required super-exponential bound cannot hold for this example. Thus Theorem 1.1 is not proved.","section":"§3.1, proof of Theorem 1.1, Eq. (3.1.6)–(3.1.7)"},{"comment":"The construction ensures that for every γ and every E, either the positive half-line m-functions are bounded by γ^{-(1-α)} or the negative half-line m-functions are bounded by γ^{-(1-α)}. Proposition 4.3, as stated, bounds the whole-line Borel transform M by the supremum of the positive half-line m-functions only. The proof does not explain how a bound on the negative half-line m-functions can be used to bound M. Without such an argument, the conclusion lim sup_{ε→0} ε^{1-α}|M(E+iε)| < ∞ does not follow from the stated dichotomy.","section":"§4.2, proof of Theorem 1.3"},{"comment":"The proof of Theorem 4.4 relies on the identity ψ(j) = lim_{ε→0} M_{j1}(E+iε)/M_1(E+iε) for the subordinate solution at energies in eS. This is stated as 'shown in the appendix of [23]', a separate preprint by the author, and no proof is included in the present paper. Since this step is essential for defining the measurable function θ(E) and the sets A_1,A_2, the argument is incomplete as written; please either include a proof or a precise, independently verified statement of the cited result.","section":"§4.3, Theorem 4.4 and proof of Theorem 1.4"}],"minor_comments":[{"comment":"There are several typographical issues, including inconsistent umlaut rendering in 'Schr\\\"odinger' and malformed displayed equations in Lemma 2.11 and in the proof of Theorem 1.3; these should be corrected in a revision.","section":"Throughout"},{"comment":"Proposition 2.19 gives ∥u_θ∥_L ≤ C(E)L^{1/2} ln L; the text writes L^{1+ε}, which is acceptable for small ε but should be justified explicitly.","section":"§3.1, proof of Theorem 1.1, Eq. (3.1.5)"},{"comment":"The statement of Proposition 4.3 is one-sided, bounding M by the positive half-line m-functions. A comment explaining whether a symmetric bound with negative half-line m-functions also holds would help the reader, and would be needed for the argument in Theorem 1.3.","section":"§4.1, Proposition 4.3"}],"recommendation":"reject","confidential_remarks":"The paper advertises three main results, but the proof of the first is invalid for a concrete reason (Corollary 3.2 contains a false inequality), and the proof of the third has a gap in its use of Proposition 4.3. The dependence on the author's own preprint [23] for a key step in Theorem 4.4 is also a concern. Given the number and centrality of the gaps, I cannot support acceptance in the present form; rejection seems appropriate, though a substantially revised manuscript that repairs these arguments could be reconsidered."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, the headline splitting result (Theorem 1.3: a whole-line operator with Hausdorff dimension 1 while every half-line restriction has dimension 0) is new, plausibly correct, and confirms a claim that was announced but apparently never written out. Theorem 1.4 is a nice consequence and the half-line construction in Section 3.2 is clean. Second, Theorem 1.1, one of the three abstract promises, is not proved. The reader's rejection is justified.\n\nWhat is good: the line construction is transparent — alternating long free stretches with finite-rank perturbations, using Corollary 3.5 to control half-line m-functions on nested scales. The proof of Theorem 4.4 is inventive, and the use of the author's own preprint [23] for measurability of theta(E) is an independent general result, not a circular dependency. Theorems 1.3 and 1.4 appear to be supported by separate arguments that do not touch the broken part of Theorem 1.1. Credit where it is due: this is a real step forward on the Jitomirskaya–Last line/half-line question.\n\nWhere it falls apart: the proof of Theorem 1.1 needs Im m(E+i epsilon) >= epsilon^{-t} for arbitrarily small t > 0. Via Proposition 2.18 this requires omega(L(ε)) = (max_η ||u_η||_L)(min_η ||u_η||_L) to be at least L(ε)^{L(ε)}. Corollary 3.2 purports to give this, but its last inequality, (k+1)^{k+1} >= L^L, is false already at L=8. And even a correct lower bound on the maximum would not do: the proof needs the product, and Corollary 3.2 says nothing about the minimum. The stress-test note goes further and says that for the constructed potential the product is actually only O(L). That stronger claim checks out: after each large spike both solutions are amplified by the same factor and become nearly parallel, so the Gram determinant of the two solution vectors over [1,L] stays O(L^2); hence omega(L)=O(L), not super-polynomial. Estimate (3.1.6) therefore fails for the operator that was built. This is not a minor typo; it is a load-bearing gap in one of the three headline results.\n\nBottom line: the paper deserves a serious referee and a major revision, not a desk rejection. Theorem 1.3/1.4 are important enough that the correct part should not be lost. But Theorem 1.1 as written is unsound, and the current version should not be accepted.","headline":"The whole-line construction in Theorems 1.3 and 1.4 is genuinely worth attention, but Theorem 1.1 is not proved: Corollary 3.2 rests on a false inequality and, more fundamentally, the product omega(L) that the argument needs is O(L), not L^L, for the potential actually constructed.","tokens_in":18494,"tokens_out":13893,"would_cite":false,"duration_ms":149174,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47B36","34L40","81Q10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper constructs one-dimensional Schrödinger operators whose spectral measures are zero-dimensional on the half line for every boundary condition, while a whole-line variant has a one-dimensional spectral measure whose every half-line…","keywords":["Schrödinger operators","spectral measures","packing dimension","Hausdorff dimension","singular continuous spectrum","subordinacy theory","rank-one perturbations","sparse potentials"],"falsifier":"Take the constructed half-line potential at scale $L=4$ (so $k=2$): the claimed inequality reads $3^3=27\\ge4^4=256$, which is false. Computing the actual maximum and minimum of $\\|u_\\eta\\|_L$ over boundary conditions at such scales will show whether $\\omega(L)\\ge L^L$ holds; if the product is smaller, the estimate (3.1.6) in the proof of Theorem 1.1 does not follow.","tokens_in":17345,"feed_emoji":"⚛️","tokens_out":9745,"duration_ms":91587,"temperature":0.7,"pith_summary":"This paper constructs explicit one-dimensional discrete Schrödinger operators that show how the fractal dimension of spectral measures can change when an operator is restricted to a half line. The half-line construction uses potentials supported on very sparse square-indexed sites to force every boundary condition's spectral measure to have packing dimension zero on $[-2,2]$. The whole-line construction alternates sparse peaks on the positive and negative sides, producing a spectral measure of Hausdorff dimension one while every half-line restriction has Hausdorff dimension zero. The final theorem converts this dimensional gap into a measure-theoretic statement: a Borel set carries positive whole-line spectral measure yet is invisible to every positive half-line boundary-condition measure.","feed_headline":"Whole-line spectrum has dimension one; half-line cuts hit zero","feed_subtitle":"A single operator family shows why spectral fractal dimension is not inherited by half-line restrictions.","key_machinery":"The machinery is subordinacy theory for one-dimensional Schrödinger operators, mediated by the Borel transform $m_\\theta$ of the spectral measures. The half-line argument uses transfer matrices to make solution norms large at sparse scales, together with a recent proposition (Proposition 2.10) that converts a lower bound on $\\liminf_{\\varepsilon\\to0} \\varepsilon^{1-\\eta}\\operatorname{Im} m_\\theta(E+i\\varepsilon)$ into an upper bound on the upper local dimension of $\\mu_\\theta$; this is the mechanism that would force packing dimension zero. The whole-line construction relies on Proposition 4.3, which bounds the whole-line Borel transform by the supremum of the two half-line Borel transforms, so alternating sparse peaks on the two sides keep the whole-line $m$-function controlled at every scale while each half-line $m$-function is large only on its own sparse scales. The invisible-set theorem uses Kac's theorem identifying the Radon–Nikodym derivative of the positive half-line measure inside the sum of the two half-line measures, letting the paper separate the energies according to which side carries the spectral mass.","core_discovery":"The paper's central claim is a pair of constructions. On the half line, a potential that vanishes except at sites $k^2$ with sufficiently tall values is claimed to have essential spectrum $[-2,2]$, with the proof organized around showing that the Borel transforms of all rank-one perturbations grow faster than any power of $1/\\varepsilon$ on the essential spectrum; if that growth holds, every spectral measure has packing dimension zero. On the whole line, the construction alternates between adding a large peak on the right and on the left, with the sites chosen so that the half-line $m$-functions recover after each change. The resulting operator is claimed to have a spectral measure with Hausdorff dimension one, while each of its half-line restrictions has Hausdorff dimension zero for every boundary condition. The paper also proves a general structural consequence: any line operator with this dimensional gap admits a Borel set with positive whole-line spectral measure that is null for all positive half-line boundary conditions.","pith_inferences":["A natural testable extension is to run the same alternating sparse-peak construction for continuum Schrödinger operators, where the same subordinacy and Borel-transform tools are available; if the phenomenon persists, the dimension discontinuity under restriction is not an artefact of the discrete lattice.","The invisible-set theorem concerns positive half-line restrictions; the same partition argument should also produce a set invisible to negative half-line restrictions, and the paper leaves open whether a single set can be invisible to both sides simultaneously.","Varying the growth of the sparse peaks could presumably tune the spectral dimension of the whole-line measure to any value in $(0,1)$ while keeping half-line restrictions zero-dimensional, producing a full family of dimension-gap examples."],"forward_implications":["For the constructed half-line family, every rank-one perturbation has a spectral measure of packing dimension zero on $[-2,2]$; since packing dimension dominates Hausdorff dimension, these measures are also zero-dimensional there.","The whole-line example shows that a line operator can have a one-dimensional spectral measure while all half-line restrictions are zero-dimensional, so local fractal dimension is not inherited under restriction to half lines.","Theorem 1.4 yields a Borel set of positive whole-line spectral measure that is annihilated by every positive half-line boundary-condition spectral measure.","All constructions keep the essential spectrum equal to $[-2,2]$, so the phenomena occur on a full spectral interval rather than on a thin exceptional set."],"supporting_citations":[{"why":"supplies Proposition 2.10, the local-dimension bound from lower Borel-transform estimates used in the half-line proof","marker":"[17]"},{"why":"supplies power-law subordinacy for half-line operators and the result that positive Lyapunov exponent gives zero-dimensional restrictions","marker":"[15]"},{"why":"supplies the subordinacy estimate relating $\\|u_\\theta\\|_L/\\|v_\\theta\\|_L$ to $|m_\\theta(E+i\\varepsilon)|$ and the lower bound on $\\operatorname{Im} m_\\theta$","marker":"[20]"},{"why":"proves Proposition 4.3, the bound on the whole-line Borel transform by the supremum of half-line Borel transforms","marker":"[8]"},{"why":"provides the line-operator subordinacy framework whose stated construction the whole-line example confirms","marker":"[16]"},{"why":"gives the almost-everywhere solution-norm bound used in the final estimate of Theorem 1.1","marker":"[24]"},{"why":"supplies the equivalence between local-density and Borel-transform criteria for continuum dimensions","marker":"[6]"},{"why":"supplies the decomposition of measures into Hausdorff and packing singular and continuous parts used in the final arguments","marker":"[2]"}],"fun_headline_variants":["Fractal dimension of spectrum drops to zero on half-line restrictions","Half-line cuts zero out spectral fractal dimension","Whole-line fractal dimension one, half-line zero","Borel set with whole-line mass is null for every half-line cut","Alternating potentials create spectral dimension gap"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The half-line packing-dimension-zero theorem rests on the assumption that the product of the largest and smallest normalized solution norms at scale $L$ is at least $L^L$, justified by the inequality $(k+1)^{k+1}\\ge L^L$, which fails for $L\\ge4$ and gives no lower bound on the smaller norm.","fun_headline_variants_meta":{"raw":{"variants":["Fractal dimension of spectrum drops to zero on half-line restrictions","Half-line cuts zero out spectral fractal dimension","Whole-line fractal dimension one, half-line zero","Borel set with whole-line mass is null for every half-line cut","Alternating potentials create spectral dimension gap"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000499,"raw_usage":{"total_tokens":2390,"prompt_tokens":841,"completion_tokens":1549,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":457,"completion_tokens_details":{"reasoning_tokens":1472}},"tokens_in":457,"tokens_out":1549,"duration_ms":11969,"temperature":1.0,"reasoning_tokens":1472,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T22:25:08.557361+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the constructed half-line potential at scale $L=4$ (so $k=2$): the claimed inequality reads $3^3=27\\ge4^4=256$, which is false. Computing the actual maximum and minimum of $\\|u_\\eta\\|_L$ over boundary conditions at such scales will show whether $\\omega(L)\\ge L^L$ holds; if the product is smaller, the estimate (3.1.6) in the proof of Theorem 1.1 does not follow.","supporting_citations":[{"cited_title":"Lower bounds on concentration through Borel transforms and quantitative singularity of spectral measures near the arithmetic transition","cited_arxiv_id":"2501.12153","evidence_quote":"supplies Proposition 2.10, the local-dimension bound from lower Borel-transform estimates used in the half-line proof"},{"cited_title":"Jitomirskaya and Y","cited_arxiv_id":null,"evidence_quote":"supplies power-law subordinacy for half-line operators and the result that positive Lyapunov exponent gives zero-dimensional restrictions"},{"cited_title":"Damanik, R","cited_arxiv_id":null,"evidence_quote":"proves Proposition 4.3, the bound on the whole-line Borel transform by the supremum of half-line Borel transforms"},{"cited_title":"Jitomirskaya and Y","cited_arxiv_id":null,"evidence_quote":"provides the line-operator subordinacy framework whose stated construction the whole-line example confirms"},{"cited_title":"Last and B","cited_arxiv_id":null,"evidence_quote":"gives the almost-everywhere solution-norm bound used in the final estimate of Theorem 1.1"},{"cited_title":"del Rio, S","cited_arxiv_id":null,"evidence_quote":"supplies the equivalence between local-density and Borel-transform criteria for continuum dimensions"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the decomposition of measures into Hausdorff and packing singular and continuous parts used in the final arguments"}],"review_version":1}