{"id":"92b0316f-1bc8-4b75-9430-ea4b8e8dd2fe","arxiv_id":"2506.18093","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Countable oscillator systems are non-wandering, continuous systems with absolutely continuous measures are wandering, and a Fourier condition extends wandering to a class of singular Bernoulli measures.","lead":"Countably many oscillators always return near their starting point, while a continuum of oscillators with a smooth spread of frequencies never does: every point wanders. This paper also gives a Fourier-transform condition that forces this wandering behavior for certain singular measures, including Bernoulli measures.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.1's dense-set conclusion is unsupported: condition (3.5) is checked only for the unweighted Bernoulli measure, never for mu_u = |u|^2 mu_eta.","rationale":"The reader's weakest assumption is exactly the one I find most load-bearing: the Fourier decay condition on weighted measures is the only input distinguishing the singular Bernoulli case from a generic measure, and Section 4 checks it only for the unweighted measure. I do not see a fatal flaw in Theorem 3.5 itself: the norm identity ||Phi_t v - v||^2 = 2||mu_u|| - \\hat mu_u(t) - \\hat mu_u(-t) would give the wandering conclusion directly once (3.5) holds, so the stated conditional theorem is sound and the proof can be repaired around the isometric-embedding issue. The countable non-wandering and absolutely-continuous wandering theorems are also independently plausible. The problem is confined to the dense-set Bernoulli application. Since the omitted argument is a lemma about restrictions of Bernoulli measures that may well be true by self-similarity, the correct disposition is to keep the reader's CONDITIONAL verdict rather than reject: the central mechanism is conditional on a condition that is verified in only the trivial case. My concrete test is designed to determine whether the missing lemma holds.","tokens_in":18532,"tokens_out":30206,"duration_ms":323901,"concrete_test":"Test the missing lemma for eta = 1/3: take u = 2 on the first-digit cylinder A and u = 1 on A^c, so mu_u = 4 mu|_A + mu|_{A^c}. Using the self-similarity of the Cantor measure, write \\widehat{mu|_A}(t) = (1/3) \\hat mu(t/3), compute \\widehat{mu_u}(t) in closed form, and determine whether limsup_{t->infty} |\\widehat{mu_u}(t)| < ||mu_u||. If equality is possible, Theorem 4.1 is in danger; if a strict gap appears, repeat the same cylinder estimate at all levels to obtain a uniform q < 1, which would prove (3.5) for all step functions and repair the dense-set conclusion.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central mechanism, Theorem 3.5, is a conditional statement: it gives wandering and non-transitivity only when the weighted measure mu_u = |u|^2 mu satisfies limsup |widehat(mu_u)(xi)| < ||mu_u||. Section 4 must therefore verify this inequality for a dense set of amplitude profiles u in L2(R, mu_eta). It does not. The product estimate (4.2) applies to mu_eta itself, i.e. to constant amplitude |u| = 1; it says nothing about |u|^2 mu_eta, whose Fourier transform is not a product and cannot be controlled by the k1 factor argument. For singular Bernoulli measures one cannot appeal to Riemann-Lebesgue: restrictions to cylinder sets can have positive limsup, so the step from mu_eta to mu_u is genuinely nontrivial. The text's 'Therefore, from Theorem 3.5' before Theorem 4.1 supplies no argument that (3.5) holds on a dense set; constant-modulus functions are not dense in L2, so the dense-set assertion is not a corollary of the unweighted check. This is the load-bearing gap in the Bernoulli application.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the linear flow Phi_t u(x) = e^{itx} u(x) on L^2(R, mu) arising from a system of harmonic oscillators, together with the invariant tori T_u = {v : |v| = |u| mu-a.e.}. The central abstract result is a Fourier criterion (Theorem 3.5): if the weighted measure mu_u = |u|^2 mu satisfies limsup |mu_u-hat(xi)| < ||mu_u||, then the restricted flow on T_u has no transitive trajectory and every point is wandering. The paper then applies this criterion to Bernoulli measures and claims in Theorem 4.1 that for every eta in (0,1) a dense set of profiles u in L^2(R, mu_eta) has this wandering, non-dense-trajectory behavior. It also proves that all points are non-wandering for countable (point-measure) oscillator systems (Theorem 7.2), that all points are wandering for absolutely continuous continual systems (Theorem 7.3), and derives non-periodicity and non-transitivity statements for absolutely continuous systems in Section 6.","tokens_in":18798,"tokens_out":27399,"duration_ms":284482,"significance":"The countable-versus-continual dichotomy is conceptually appealing, and the conditional Fourier criterion (Theorem 3.5) is a genuinely useful sufficient condition that unifies the absolutely continuous and singular cases. The non-wandering/wandering results in Section 7 are the main new contributions and appear correct in substance. However, the advertised application to a dense set of Bernoulli amplitude profiles is not established by the arguments in Section 4, and that dense-set claim is the principal new singular-measure result. If the gap is repaired or the statement is weakened to an existence result, the paper would make a solid contribution to the ergodic theory of infinite-dimensional linear flows.","major_comments":[{"comment":"The dense-set conclusion of Theorem 4.1 is not supported by the preceding argument. The paragraph before the theorem verifies, using (4.1) and (4.2), only the unweighted inequality limsup |mu_eta-hat(xi)| < 1, i.e. the case of constant amplitude u = 1. Theorem 3.5 requires, for each u in the claimed dense set, the weighted inequality limsup |(|u|^2 mu_eta)-hat(xi)| < || |u|^2 mu_eta ||. The Fourier transform of |u|^2 mu_eta is not a product of the factors in (4.1), and the k1-factor estimate does not control it. Since functions of constant modulus are not dense in L^2(R, mu_eta), the sentence 'Therefore, from the Theorem 3.5 we obtain the following statement' is a non sequitur. The authors must either prove condition (3.5) for a dense set of amplitude profiles, for example by controlling restrictions of mu_eta as Corollary 3.7 indicates, or they must replace Theorem 4.1 with an existence statement for at least one nonconstant torus.","section":"Section 4"},{"comment":"The transfer from Lemma 3.2 to the torus T_u is not fully written. One needs the isometric conjugacy phi in L^2(mu_u) to phi u in T_u, which intertwines the two flows; as written, the proof simply asserts that Lemma 3.2 applies to the restricted flow. This is a local but necessary step for the central criterion and should be made explicit.","section":"Theorem 3.5"}],"minor_comments":[{"comment":"Conditions (3.1), (3.5), (3.6), and Corollary 3.7 should be stated with limsup rather than lim, since the relevant limits need not exist.","section":"Section 3"},{"comment":"The proof writes v(x) = phi(x) |u(x)|, but Lemma 3.4 gives v(x) = phi(x) u(x); the phase of u should be absorbed into phi.","section":"Theorem 3.5"},{"comment":"The statement should exclude u = 0, because the zero vector is fixed and hence non-wandering; the assumption of a non-degenerate torus should be made explicit.","section":"Theorem 7.3"},{"comment":"The proof invokes Poincare recurrence for the finite-dimensional projection without noting that every point of a linear flow on a finite-dimensional torus is non-wandering; the notation F_m (tail subspace) and F_m^perp is confusing and should be clarified.","section":"Theorem 7.2"},{"comment":"The sentence that Bernoulli measures 'provide explicit examples of singular continuous measures' needs a qualification, since mu_{1/2} is absolutely continuous.","section":"Section 4"},{"comment":"Reference [31] is the same arXiv preprint as the present paper; this self-reference should be replaced by a proper announcement or removed.","section":"References"},{"comment":"There are typographical issues: 'This research was founded' should be 'funded', 'Results of of the present paper' has a duplicated word, and 'Theorem 1.2 is also known as the Weyl-Kronecker theorem' should refer to Theorem 1.5.","section":"General"},{"comment":"The displayed lower bound after Eq. (3.12) should be an inequality rather than an equality; the constant sigma/16 is a lower bound after absorbing the factor 2 in the preceding estimate.","section":"Theorem 3.6"},{"comment":"The proof that every point of T_u, and not only u itself, is wandering is omitted; the same norm computation with v in place of u supplies it and should be stated.","section":"Lemma 3.2"}],"recommendation":"major_revision","confidential_remarks":"The main obstacle is the unsupported dense-set claim in Theorem 4.1. If the authors replace it by an existence statement or supply the missing verification, the paper would be suitable for publication; the conditional theorems and the countable/ac dichotomy are otherwise sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper's main structural claims are right, and the countable-versus-continual wandering dichotomy is a genuinely useful addition to the infinite-dimensional torus literature. But the Bernoulli application is oversold. Theorem 4.1 claims a dense set of profiles for which all points wander, and the proof only checks the Fourier condition for the unweighted Bernoulli measure. That does not get you the dense set.\n\nWhere the paper is solid: Theorem 7.2 (countable systems: every point non-wandering) is correct, and the proof via finite-dimensional Poincare recurrence plus a tail estimate is clean. Theorem 7.3 (absolutely continuous measure: every point wandering) follows from Riemann-Lebesgue and is also correct. Theorem 3.5 is a correct conditional statement: if the Fourier transform of mu_u = |u|^2 mu has limsup strictly below total variation, then the invariant torus has neither transitivity nor non-wandering points. The product estimate for the Bernoulli measures themselves (inequality (4.2)) is a nice explicit argument.\n\nThe soft spot is the step to Theorem 4.1. Condition (3.5) needs to hold for each mu_u, and for u with nonconstant modulus the Fourier transform of |u|^2 mu_eta is not the same as the product in (4.1). Constant-modulus profiles are not dense in L^2, so checking the unweighted measure does not yield a dense set. The text simply says \"Therefore, from Theorem 3.5\" and moves on. This is the one load-bearing gap. To fix it, either prove (3.5) for a dense set of profiles (which is not obvious and may be false), or weaken the theorem to a statement about the specific torus of constant amplitude. There is also a minor issue: Theorem 3.6's condition (3.6) is strong and is not verified for the Bernoulli measures, but it is not used in the main application.\n\nThe paper inherits some results from the authors' earlier [30], and that is fine; the new dichotomy results are proved independently. The citation pattern is not a problem.\n\nVerdict: this deserves a serious referee. The dichotomy results and the Fourier criterion are worth publishing after revision. The Bernoulli application should be corrected or downgraded.","headline":"The countable-vs-continual wandering dichotomy is a real contribution, but the Bernoulli application in Theorem 4.1 is oversold: the Fourier condition is only checked for the unweighted measure, not for the weighted measures that the dense-set conclusion needs.","tokens_in":19295,"tokens_out":5354,"would_cite":true,"duration_ms":54002,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28D05","37A05","47A35","37K10","37N20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that for linear flows on infinite-dimensional tori, the type of the underlying measure decides recurrence: countable systems keep every point non-wandering, while absolutely continuous and singular Bernoulli systems can…","keywords":["infinite-dimensional tori","harmonic oscillator systems","wandering points","topological transitivity","singular Bernoulli measures","Fourier transform of a measure","linear flow","non-wandering dichotomy"],"falsifier":"For a Bernoulli measure $\\mu_\\eta$ and a nonconstant profile $u\\in L^2(\\mathbb{R},\\mu_\\eta)$ (for instance $u(x)=1+\\varepsilon x$ on a set of positive $\\mu_\\eta$-mass), compute $\\limsup_{\\xi\\to\\infty}|\\widehat{|u|^2\\mu_\\eta}(\\xi)|$. If for some such $u$ this limsup equals $\\|u\\|_{L^2(\\mu_\\eta)}^2$, then the sufficient condition in Theorem 3.5 fails for that torus, and the dense-set assertion in Theorem 4.1 would require an additional argument; if instead the limsup is strictly smaller, the paper's hypothesis extends to weighted profiles without further assumptions.","tokens_in":18359,"feed_emoji":"🌀","tokens_out":6714,"duration_ms":67398,"temperature":0.7,"pith_summary":"This paper asks how trajectories of a linear flow on an infinite-dimensional torus behave when the system is a continuum of independent harmonic oscillators rather than a finite or countable set. The authors claim a sharp dichotomy: for a countable oscillator system every point on a non-degenerate invariant torus is non-wandering, whereas for a continuous system whose parameter measure is absolutely continuous every such point is wandering. For singular continuous measures they identify a sufficient condition on the Fourier transform of the measure that also forces every point to be wandering and rules out transitive trajectories, and they exhibit Bernoulli measures satisfying it. The interest is that these behaviors are genuinely infinite-dimensional: finite-dimensional linear flows only see periodic or dense orbits, while here a third, non-recurrent type of trajectory becomes generic. If correct, the results give a spectral criterion for recurrence in systems that include every isolated quantum system via Koopman–von Neumann equivalence.","feed_headline":"Countable oscillator systems recur; continuous ones drift away","feed_subtitle":"Measure type decides the fate of orbits: countable systems return, while continuous and Bernoulli systems can have every point wandering.","key_machinery":"The central objects are the invariant torus $T_r=\\{u\\in L^2(\\mathbb{R},\\mu): |u(x)|=r(x)\\ \\mu$-a.e.\\}$ and the effective measure $\\mu_u=|u|^2\\mu$, whose Fourier transform $\\widehat{\\mu_u}(\\xi)=\\int |u(x)|^2 $e^{{-i\\xi x}}$\\,d\\mu(x)$ carries the argument. The lever is Theorem 3.5: if the Fourier transform's magnitude at infinity stays strictly below the total mass of $\\mu_u$, then no orbit can return to any fixed neighborhood of its starting point, so every point is wandering and the torus is not topologically transitive. The countable/continuous dichotomy is read off from whether this Fourier decay happens: it fails for point measures, holds for absolutely continuous measures by the Riemann–Lebesgue lemma, and holds for Bernoulli measures through their infinite cosine product.","core_discovery":"On the torus $T_u=\\{v\\in L^2(\\mathbb{R},\\mu): |v|=|u|\\ \\mu$-a.e.\\}$ with flow $\\Phi_t v(x)=$e^{{itx}}$v(x)$, define the effective measure $\\mu_u=|u|^2\\mu$. Theorem 3.5 proves that whenever $\\lim_{\\xi\\to\\infty}|\\widehat{\\mu_u}(\\xi)|<\\|\\mu_u\\|$, the restricted flow is not topologically transitive and every point of $T_u$ is wandering. The countable case escapes this because the Fourier transform of a point measure is an almost periodic function whose limsup can equal its total mass; Theorem 7.2 uses finite-dimensional Poincaré recurrence on high modes to show every point is non-wandering. The absolutely continuous case satisfies the condition by the Riemann–Lebesgue lemma, giving wandering for all points (Theorem 7.3). For Bernoulli measures $\\mu_\\eta$ with parameter $\\eta\\in(0,1)$, the product formula $\\widehat{\\mu_\\eta}(t)=\\prod_{k=1}^\\infty \\cos(2\\pi t\\eta^k)$ forces $\\lim_{t\\to\\infty}|\\widehat{\\mu_\\eta}(t)|<1$, and Theorem 4.1 asserts that a dense set of profiles $u$ have every point of $T_u$ wandering and no trajectory dense in $T_u$.","pith_inferences":["Editorial inference: the Fourier-decay condition is essentially a Rajchman-type spectral property for the weighted measure; if one could show $\\lim_{\\xi\\to\\infty}|\\widehat{|u|^2\\mu}(\\xi)|=0$ for every profile in a dense set, the Bernoulli result would follow cleanly, but the paper only proves the unweighted case, so the dense-set statement in Theorem 4.1 rests on an unverified extension.","Editorial inference: the same criterion suggests a measurable-rigidity hierarchy in which recurrence type is classified by the set of frequencies where the spectral measure has atoms or lacks Fourier decay; countable systems sit at the atomic end, absolutely continuous measures at the Rajchman end, and Bernoulli convolutions interpolate between them.","Editorial inference: a natural testable extension is to non-Bernoulli self-similar measures (for example Cantor measures with other digit sets) and to profiles $u$ with $|u|^2$ comparable to such measures; wherever the weighted Fourier limsup drops below the weighted mass, the same wandering conclusion should hold.","Editorial inference: because every isolated quantum system is unitarily equivalent to such an oscillator system, the dichotomy would say that continuous-spectrum quantum dynamics disperses (all states wandering) while discrete-spectrum dynamics is recurrent; checking this against explicit Schrödinger evolutions could connect the theorem to quantum recurrence statements."],"forward_implications":["For countable oscillator systems with rationally commensurable but unbounded reciprocal frequencies, trajectories can be neither periodic nor dense, yet every point is still non-wandering; recurrence and transitivity separate cleanly in infinite dimensions.","For continuous systems with locally integrable density, every point on a non-degenerate invariant torus is wandering, so no nontrivial orbit returns to its starting neighborhood; the generic trajectory is neither periodic nor transitive.","For Bernoulli measures $\\mu_\\eta$ with $\\eta\\in(0,1)$, the paper's criterion excludes dense trajectories and non-wandering points on every non-degenerate torus, including the singular continuous cases $\\eta\\neq 1/2$.","The Fourier-decay condition (3.5) gives a quantitative sufficient test that can be checked directly from the spectral measure of the oscillator system, without solving individual trajectories."],"supporting_citations":[{"why":"Supplies the countable-oscillator results on periodic, dense, and Type III trajectories that the present paper extends.","marker":"[30]"},{"why":"Establishes the Koopman–von Neumann style representation of any Hamiltonian system with invariant measure as a system of harmonic oscillators.","marker":"[37]"},{"why":"Provides Jacobi's theorem on periodicity and density of trajectories on finite-dimensional tori, the classical benchmark the infinite-dimensional results are measured against.","marker":"[1]"},{"why":"Defines resonance and strong rational commensurability used in the periodicity criteria.","marker":"[9]"},{"why":"Extends Weyl's ergodicity theorem to infinite-dimensional tori and provides context for the transitivity criteria.","marker":"[21]"},{"why":"Gives the classical treatment of Bernoulli measures whose infinite cosine product representation appears in equation (4.1).","marker":"[16]"},{"why":"Supplies the definition of non-wandering point used in Section 7.","marker":"[10]"},{"why":"Provides the spectral theorem and the point/absolutely-continuous/singular decomposition of measures underlying the canonical triplet representation.","marker":"[28]"}],"fun_headline_variants":["Measure type decides recurrence in oscillator flows","Fourier decay traps orbits: non-wandering vs wandering","Countable tori recur, continuous tori wander","Orbit recurrence flips with oscillator measure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the claim that for the relevant amplitude profiles the Fourier transform of the weighted measure $|u|^2\\mu$ stays, in magnitude, strictly below its total mass at infinity—and the paper demonstrates this only for unweighted Bernoulli measures, not for the general profiles its dense-set theorem asserts.","fun_headline_variants_meta":{"raw":{"variants":["Measure type decides recurrence in oscillator flows","Fourier decay traps orbits: non-wandering vs wandering","Countable tori recur, continuous tori wander","Orbit recurrence flips with oscillator measure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00019,"raw_usage":{"total_tokens":1371,"prompt_tokens":1005,"completion_tokens":366,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":621,"completion_tokens_details":{"reasoning_tokens":306}},"tokens_in":621,"tokens_out":366,"duration_ms":3943,"temperature":1.0,"reasoning_tokens":306,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:57:51.024989+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a Bernoulli measure $\\mu_\\eta$ and a nonconstant profile $u\\in L^2(\\mathbb{R},\\mu_\\eta)$ (for instance $u(x)=1+\\varepsilon x$ on a set of positive $\\mu_\\eta$-mass), compute $\\limsup_{\\xi\\to\\infty}|\\widehat{|u|^2\\mu_\\eta}(\\xi)|$. If for some such $u$ this limsup equals $\\|u\\|_{L^2(\\mu_\\eta)}^2$, then the sufficient condition in Theorem 3.5 fails for that torus, and the dense-set assertion in Theorem 4.1 would require an additional argument; if instead the limsup is strictly smaller, the paper's hypothesis extends to weighted profiles without further assumptions.","supporting_citations":[{"cited_title":"Zh., & Volovich, I","cited_arxiv_id":null,"evidence_quote":"Supplies the countable-oscillator results on periodic, dense, and Type III trajectories that the present paper extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the Koopman–von Neumann style representation of any Hamiltonian system with invariant measure as a system of harmonic oscillators."},{"cited_title":"I., & Avez, A","cited_arxiv_id":null,"evidence_quote":"Provides Jacobi's theorem on periodicity and density of trajectories on finite-dimensional tori, the classical benchmark the infinite-dimensional results are measured against."},{"cited_title":"S., & Fischler, S","cited_arxiv_id":null,"evidence_quote":"Defines resonance and strong rational commensurability used in the periodicity criteria."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extends Weyl's ergodicity theorem to infinite-dimensional tori and provides context for the transitivity criteria."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the classical treatment of Bernoulli measures whose infinite cosine product representation appears in equation (4.1)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the definition of non-wandering point used in Section 7."},{"cited_title":"(1977).Methods of modern mathematical physics, Vol","cited_arxiv_id":null,"evidence_quote":"Provides the spectral theorem and the point/absolutely-continuous/singular decomposition of measures underlying the canonical triplet representation."}],"review_version":2}