{"id":"0c244727-8b4c-4fc5-9606-4857e5ed4960","arxiv_id":"2608.05239","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a new frequency set larger than Brjuno-Rüssmann, renormalized Gevrey-γ perturbations of twist maps still admit real-analytic invariant graphs, with a λ-independent threshold in the dissipative case.","lead":"This paper studies when elliptic invariant circles of twist maps survive small perturbations, using a direct KAM method with tree expansions. It claims persistence for frequencies beyond the classical Brjuno condition, and a damping-rate independent smallness threshold in dissipative maps.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1's key resummation estimate (32) rests entirely on an unproved cancellation lemma (Lemma A.4) cited to [Gen15, Lemma 3.16]; if that cancellation fails for q_m-renormalized trees, the central claim has no support.","rationale":"The reader's weakest_assumption and my independent reading converge on the same point: Lemma A.4 is the pivotal unproved ingredient. The manuscript's Appendix A says it provides a self-contained treatment of resonance renormalization, but the actual proof of Lemma A.4 is a one-sentence citation to [Gen15, Lemma 3.16]. The paper itself lists three ways its setting differs from [Gen15] (q_m-dependent scales, threshold κ(n), mixed propagator), and none of these differences is shown to preserve the algebraic cancellation. Without Lemma A.4, the resummation step replacing P_n factors by N_n^R factors is unsupported, so the counting lemma cannot control the accumulation of small divisors and the lower bound on the radius of convergence in Section 3.2.5 collapses. I do not see an independent flaw in the arithmetic construction of ∆α in Proposition 3.1 or in the Gevrey-norm comparison in Sections 3.2.6–3.2.7; those parts read coherently. There are secondary gaps noted by the reader — the γ=0 case is handled heuristically and the uniqueness claim in Theorem 2 is not proved — but they are less central than Lemma A.4. A concrete enumeration test of Lemma A.4 on a minimal resonance would settle the matter. Until that check is done, a conditional verdict is appropriate; if the check fails, the verdict would need to move to REJECT.","tokens_in":28967,"tokens_out":11703,"duration_ms":116087,"concrete_test":"Check Lemma A.4 in the minimal nontrivial q_m-renormalized case: take a first-generation resonance V of order 2 with one entering and one exiting line, node labels multiples of q_m, scales n and n+1, and total momentum below κ(n). Enumerate the trees in F_V(ϑ) exactly as defined by the two line reattachments and the mode-label inversion, compute the localized factors L_V Val(ϑ') from equations (61)–(65), and verify that their sum is identically zero. If the sum is not zero, or if some configurations violate the condition M(T)<κ(n), Lemma A.4 fails and estimate (32) has no basis; if the sum is zero, repeat the check for a chain of two nested resonances to confirm the induction in Proposition A.6.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the transition from (31) to (32) in Section 3.2.2: the factor ∏(768 q_{m+n}+1)^{2 P_n(ϑ)} coming from lines exiting clusters is asserted to be replaceable by ∏(768 q_{m+n}+1)^{2 N_n^R(ϑ)} after resummation. This replacement removes the divergent P_n contribution and is exactly what makes the counting lemma (Lemma 2.14) applicable, yielding the radius bound in Section 3.2.5. Its only support is Appendix A, where the key cancellation is Lemma A.4, whose proof is a citation to [Gen15, Lemma 3.16] with the words 'identical telescoping argument.' The Appendix claims to be self-contained but does not verify the three modifications it itself lists: (i) the multi-scale decomposition is q_m-dependent; (ii) the resonance threshold is κ(n), not q_{m+n}; (iii) the propagator has both hyperbolic and elliptic components. In particular, for trees whose node labels are restricted to multiples of q_m, the resonance family F_V(ϑ) must be closed under the line reattachments and mode-label inversions and must preserve the resonance condition M(T)<κ(n); this is not demonstrated. If the localized factors do not cancel in this restricted class, estimate (32) is unjustified, the radius bound in Section 3.2.5 is not established, and Theorem 1 does not follow. Since Theorem 1 is the paper's central new claim — that analytic conjugacy can persist for a frequency set strictly larger than the Brjuno–Rüssmann set — this is the single load-bearing concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies persistence of invariant graphs for standard-like area-preserving and dissipative twist maps under Gevrey perturbations. The main result, Theorem 1, constructs, for each frequency in an extended set Δ^α and each Gevrey-γ perturbation with γ < ε/α, a sequence of area-preserving maps whose Gevrey-α distance to the integrable twist map tends to 0 while their Gevrey-(α−ε) distance tends to ∞, and each such map admits an invariant graph real-analytically conjugate to a rotation of frequency ω. The dissipative results, Theorem 2 and its corollaries, claim persistence with a uniform smallness threshold independent of λ, including uniqueness of the invariant graph. The proofs are based on a tree expansion with a multi-scale decomposition, a counting lemma for resonances, and a renormalization procedure in Appendix A.","tokens_in":29350,"tokens_out":20982,"duration_ms":220487,"significance":"If correct, Theorem 1 would be a genuinely interesting extension beyond the classical Brjuno–Rüssmann framework: for every α>1 and ω∈Δ^α∖BR_α, sufficiently regular Gevrey perturbations need not be small in the ambient analytic topology, because δ_m can be chosen so that the α-norm is small while the (α−ε)-norm blows up. The paper also contains useful concrete ingredients: Lemma 2.6 gives a uniform propagator lower bound for all λ∈(0,1], Lemma 2.14 is an explicit counting lemma, and Section 5 provides a constructive example verifying Remark 1.5. However, the principal estimate (32) rests on a resonance-cancellation lemma that is not proved in the paper, and the tree-summation lemma (Lemma 3.3) is only sketched. These gaps prevent the main theorems from being fully established as written.","major_comments":[{"comment":"The transition from (31) to (32) is the decisive step: the factor ∏(768 q_{m+n}+1)^{2P_n(ϑ)} is replaced by ∏(768 q_{m+n}+1)^{2N_n^R(ϑ)}, which makes the counting lemma applicable and yields the radius bound in Section 3.2.5. The only support cited for this replacement is Lemma A.4, whose proof is delegated to [Gen15, Lemma 3.16] with the phrase 'identical telescoping argument'. Appendix A itself lists three modifications of the [BG01]/[Gen15] setting—the m-dependent multi-scale decomposition, the threshold κ(n), and the mixed hyperbolic/elliptic propagator—but it does not prove that the resonance family F_V(ϑ) is closed under line reattachments and mode-label inversions for trees whose Fourier labels are restricted to multiples of q_m, nor that condition (5) of Definition 2.9 (M(T)<κ(n)) is preserved under these operations. Without this verification, estimate (32) is unsupported, and the radius bound in Section 3.2.5 does not follow. Since Theorem 1 is the central new claim, this gap must be repaired.","section":"Section 3.2.2, Eq. (32); Appendix A, Lemma A.4"},{"comment":"The summation over the enlarged tree set T^*_{ν,k} is the other load-bearing step: it converts the per-tree estimates (35)–(36) into the coefficient bound (37) from which the radius and δ_m in (38) are chosen. Lemma 3.3 is only a sketch. The stated constant C_5=4·17·C_3C_4C_6 refers to a C_6 that has not been defined at that point, and the proof does not explain how the resonance renormalization of Appendix A affects the count of renormalized trees; in particular, the simple factor 4^k·17^{#lines} does not account for the multiplicities and equivalence classes in Eq. (58) and the counting in Appendix A.8. A complete proof of Lemma 3.3, or a replacement that counts the renormalized trees rigorously, is needed before (37) can be used.","section":"Section 3.2.4, Lemma 3.3"},{"comment":"Theorem 2 asserts that the invariant graph with Property A_ω is unique, but the proof in Section 4.1 establishes only convergence of the Lindstedt series to one real-analytic solution of Eq. (11). No uniqueness argument is given, and uniqueness is not automatic for nonlinear equations of this type, especially because the graph is not claimed to be 1-normally hyperbolic uniformly in λ. The statement should be weakened to existence, or uniqueness should be proved.","section":"Theorem 2 / Section 4.1"},{"comment":"Both theorems are stated for arbitrary φ∈G^γ(T) (resp. ψ∈C^ω_s(T)), but the model (1) assumes ∫φ=0, and the recurrence (13) has γ(0)=0; a nonzero constant Fourier mode of φ cannot be absorbed by a mean-zero u, so the Lindstedt series does not solve Eq. (11). The same issue affects Theorem 2 and Corollary 4. The statements need either the mean-zero hypothesis or an argument that adjusts ω (as in Eq. (54)) before the tree expansion is applied.","section":"Theorem 1 and Theorem 2 statements"}],"minor_comments":[{"comment":"The assertion that both BR_α and Δ^α have Hausdorff dimension zero is incorrect: the Brjuno–Rüssmann set has full Lebesgue measure, hence Hausdorff dimension 1. This does not affect the proof of Theorem 1, but the statement should be corrected.","section":"Section 1.1.2"},{"comment":"The case γ=0 is left as 'the obvious interpretation'; since 1/γ is undefined in (27) and the subsequent estimates use quantities such as (γ/s_1)^{γk}, the trigonometric-polynomial case should be treated separately or by a limiting argument.","section":"Section 3.2.1, Eq. (27)"},{"comment":"The displayed definition of a_j in the construction of the continued fraction coefficients is garbled; the floor and exponent notation is not typeset correctly, making the construction difficult to verify.","section":"Proposition 3.1"},{"comment":"The bound (77) is asserted to follow from [BG01, Lemma 5.1] after the statement e^{4k}; since the appendix is announced as self-contained, either the combinatorial argument should be given or the precise reference with a verification of the present setting should be supplied.","section":"Appendix A.8"}],"recommendation":"major_revision","confidential_remarks":"The reliance on [HLW26] and [SW26] for background and sharpness seems acceptable. The main concern is that the paper's central estimate (32) depends on Lemma A.4, which is not proved but only cited to [Gen15], despite the appendix claiming self-containedness. If the authors cannot supply the missing verification, the recommendation should be reconsidered. The other load-bearing issues are Lemma 3.3 and the unproved uniqueness in Theorem 2."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline is that this paper has two genuinely new results worth a careful read: the frequency set Δ^α strictly containing the Brjuno–Rüssmann set, with an explicit uncountable family in the complement, and the λ-independent smallness threshold for dissipative twist maps. The arithmetic construction in Prop 3.1 is explicit and looks right. The two-regime split in Section 3.2.3 (large vs small momentum) and the use of the counting lemma give a coherent strategy, and the dissipative propagator bound (Lemma 2.6) that combines the hyperbolic and elliptic components is a nice observation. Credit where due: the paper is not a repackaging of [Gen15]; the renormalized perturbation φ(q_m x) and the δ_m scaling are new, and the theorem statements are sharper than what the cited literature gives. The citation pattern looks honest; the two self-cited preprints are used for context, not to establish the main estimates.\n\nThe soft spot is real, and it is load-bearing. Estimate (32), the transition that removes the P_n factors after resummation, is the step that makes the counting lemma applicable. Its support is Appendix A, where Lemma A.4 — the cancellation of localized resonance factors — is not proved but cited to [Gen15, Lemma 3.16] with the phrase 'identical telescoping argument.' The appendix lists three differences from [Gen15]: the q_m-dependent multi-scale decomposition, the κ(n) threshold, and the two-component propagator. It does not verify that the cancellation survives the restriction of node labels to multiples of q_m, nor that the resonance family is closed under the line reattachments and mode-label inversions in that restricted class. The stress-test note is accurate: if that cancellation fails, estimate (32) collapses and Theorem 1 does not follow. This is not a minor typo; it is the central proof step.\n\nThe other gaps are smaller. Lemma 3.3 is a sketch, though the bound is plausible. Theorem 2 asserts uniqueness of the invariant graph, but the proof only shows existence of the convergent Lindstedt series; uniqueness is not demonstrated. The γ=0 case rests on an 'obvious interpretation' of (27), which should be stated properly.\n\nNet verdict: the paper deserves a serious referee. The ideas are serious, the statements are new, and the structure is coherent. But the authors should be asked either to prove Lemma A.4 in the q_m-renormalized setting or to state it as an explicit assumption with a complete reference, and to justify or drop the uniqueness claim. If the gap is filled, this is a solid contribution to KAM theory; if not, the main theorem is unsupported.\n\nRecommendation: send to peer review, with the appendix as the primary focus.","headline":"New results and a coherent framework, but the central resummation step rests on an unproved cancellation lemma; worthy of refereeing, not of acceptance as-is.","tokens_in":29910,"tokens_out":2958,"would_cite":true,"duration_ms":29425,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37J40","37E40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that invariant graphs with real-analytic conjugacy to a rigid rotation can persist for frequencies strictly outside the classical Brjuno–Rüssmann set, as long as the perturbation is sufficiently Gevrey-smooth and is…","keywords":["direct KAM method","invariant graphs","twist maps","Property A_ω","Brjuno–Rüssmann condition","Gevrey regularity","Lindstedt series","resonance renormalization"],"falsifier":"Take a single Fourier mode $\\varphi(x)=e^{2\\pi i q_m x}$ and a frequency $\\omega\\in\\Delta^\\alpha$, and compute the sum over the resonance family of the localized resonance factors in (61)–(64) for a resonance with two entering lines and total momentum zero, with $M(T)<\\kappa(n)$; if this sum is nonzero for any admissible tree, the cancellation lemma is false and the proof of Theorem 1 breaks down.","tokens_in":28732,"feed_emoji":"🌀","tokens_out":10545,"duration_ms":88507,"temperature":0.7,"pith_summary":"The paper asks when the strongest possible invariant object of a twist map—a graph on which the dynamics is real-analytically conjugate to a rigid rotation—survives perturbation. In the area-preserving case, it exhibits a frequency set $\\Delta^\\alpha$ strictly larger than the classical Brjuno–Rüssmann set and proves that for any Gevrey-$\\gamma$ perturbation with $\\gamma<\\varepsilon/\\alpha$, a carefully renormalized sequence of standard-like maps admits such an invariant graph, even though the map diverges in every Gevrey-$(\\alpha-\\varepsilon)$ norm. In the dissipative case, it proves that an analytic perturbation whose size is below a threshold independent of the dissipation parameter $\\lambda$ still leaves a unique invariant graph with the same strongest dynamics. The paper thus claims that enhancing the regularity of the perturbation enlarges the admissible frequency set and removes the normal-hyperbolicity dependence of the persistence threshold.","feed_headline":"Invariant graphs survive beyond the Brjuno–Rüssmann barrier","feed_subtitle":"Gevrey-smooth renormalized perturbations keep rotation-conjugate graphs intact, even when stronger norms diverge.","key_machinery":"The machinery is the parameterized direct KAM method built on Lindstedt series for the conjugacy equation $D^2_{\\lambda,\\omega}u=\\varepsilon\\phi(\\theta+u)$, organized as tree expansions whose lines carry Fourier momenta and propagators $1/\\gamma(\\nu)$ with $\\gamma(\\nu)=(1+\\lambda)(\\cos 2\\pi\\omega\\nu-1)+i(1-\\lambda)\\sin 2\\pi\\omega\\nu$. A multi-scale decomposition splits each propagator according to the size of $\\|\\omega\\nu\\|$, clusters and resonances are defined through the momentum threshold $\\kappa(n)=\\min\\{k:q_{mk}\\ge q_{m+n}\\}$, and a counting lemma (the Siegel–Brjuno estimate) bounds the number of lines on each scale. The decisive step is resonance renormalization: localized resonance factors cancel exactly, so the factors counting lines that exit clusters can be replaced by resonance-count factors, yielding the refined estimate (32). For the dissipative case, the two-component lower bound on $|\\gamma(\\nu)|$ gives a radius of convergence $\\rho(\\omega)\\ge\\max\\{\\Lambda(1-\\lambda),\\Omega(1+\\lambda)\\}$, which is what makes the persistence threshold independent of $\\lambda$.","core_discovery":"On the paper's own terms, the central discovery is that the direct KAM method, implemented through Lindstedt series, tree expansions, multi-scale decompositions, and renormalization of resonances, can prove persistence of invariant graphs with Property $A_\\omega$ under conditions that classical KAM and normally hyperbolic invariant manifold theory do not cover. Theorem 1 constructs, for each $\\omega\\in\\Delta^\\alpha$ and each Gevrey-$\\gamma$ function $\\varphi$ with $0\\le\\gamma<\\varepsilon/\\alpha$, a sequence of perturbation amplitudes $\\delta_m$ such that the map $f_m(x,y)=(x+\\omega+y+\\delta_m\\varphi(q_m x),y+\\delta_m\\varphi(q_m x))$ has an invariant graph that is real-analytically conjugate to rotation by $\\omega$, with $\\|f_m-T\\|_{\\alpha,L}\\to0$ and $\\|f_m-T\\|_{\\alpha-\\varepsilon,L}\\to\\infty$. Theorem 2 shows that for a Brjuno frequency and an analytic perturbation, the invariant graph persists with a perturbation size that is uniform in $\\lambda\\in(0,1]$, and Corollary 4 extends this to trigonometric polynomial perturbations controlled in $C^1$.","pith_inferences":["The renormalized-perturbation mechanism suggests a general principle: regularity of the perturbation can compensate for arithmetic difficulties of the frequency, and the same tree-resummation technology could yield analogous compensation for higher-dimensional tori or quasi-periodic forcing.","The divergence of $\\|f_m-T\\|_{\\alpha-\\varepsilon,L}$ despite persistence of the analytic invariant graph means the invariant object is robust while the map itself escapes every stronger Gevrey neighbourhood; this sharpens the usual KAM picture in which persistence requires closeness in the working topology.","A direct numerical evaluation of the cancellation in Lemma A.4 for small resonances with momenta supported on multiples of $q_m$ would test the key step before a complete analytic proof is supplied.","The boundary case $\\gamma=\\varepsilon/\\alpha$ is left open; testing perturbations with Gevrey exponent exactly at the threshold could reveal whether the persistence phenomenon has a phase transition."],"forward_implications":["For frequencies in $\\Delta^\\alpha\\setminus BR^\\alpha$, an uncountable set, there exist Gevrey perturbations of class $\\gamma<\\varepsilon/\\alpha$ that keep the rotation-conjugate invariant graph while the same perturbation is invisible in every stronger Gevrey topology.","The explicit choice $\\delta_m=\\exp(-c_0q_m^{\\sigma})$ with $\\sigma=1/(\\alpha(1-\\gamma))$ ties the allowed perturbation size to the arithmetic of the frequency, and the condition $\\gamma<\\varepsilon/\\alpha$ is exactly what makes the $\\alpha$-norm converge and the $(\\alpha-\\varepsilon)$-norm diverge.","In the dissipative setting, the uniform radius $\\rho(\\omega)\\ge\\Omega(1+\\lambda)$ gives a persistence threshold for analytic perturbations that does not vanish as $\\lambda\\to1^-$, which normal hyperbolicity alone does not provide.","For trigonometric polynomial perturbations, a $C^1$ smallness condition with a threshold independent of $\\lambda$ suffices for the strongest dynamics to persist at Brjuno frequencies."],"supporting_citations":[{"why":"supplies the direct KAM tree-expansion framework for invariant curves of symplectic twist maps, including the cancellation lemma (Lemma 3.16) on which the resummation step depends.","marker":"[Gen15]"},{"why":"introduces the Bryuno-function tree expansion and the resonance machinery that the paper adapts to $q_m$-renormalized perturbations.","marker":"[BG01]"},{"why":"provides the multi-scale decomposition and the scale-constant choices used in the propagator bounds.","marker":"[BG02]"},{"why":"proves the Brjuno–Rüssmann persistence result for Gevrey perturbations that the paper's $\\Delta^\\alpha$ result extends.","marker":"[BF19]"},{"why":"shows analytic conjugacy can be destroyed for non-Brjuno frequencies, the baseline against which the new positive result is measured.","marker":"[Bou19]"},{"why":"constructs examples of analytic destruction of invariant circles that delimit the gap the renormalized construction fills.","marker":"[For94]"},{"why":"gives the normally hyperbolic invariant manifold theorem whose $\\lambda$-dependent smallness threshold Theorem 2 goes beyond.","marker":"[HPS77]"},{"why":"provides the sharp $C^1$-norm threshold $(1-\\sqrt\\lambda)^2$ for normally hyperbolic persistence and the counterexample used to verify Remark 1.5.","marker":"[HL W26]"},{"why":"supplies the Herman–Mather formula for conformally symplectic twist maps used to match the frequency in the dissipative construction.","marker":"[SW26]"},{"why":"gives the Hausdorff-dimension-zero result used to contrast the cardinality of $\\Delta^\\alpha\\setminus BR^\\alpha$ with its meagreness.","marker":"[Jar29]"}],"fun_headline_variants":["Invariant graphs persist even as norms diverge","Direct KAM overcomes limits of classical theory","Gevrey perturbations extend graph survival past KAM","Rotation-conjugate graphs beat Brjuno barrier"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole argument rests on a cancellation among renormalized trees that the paper borrows from an earlier work; if that cancellation fails for trees whose Fourier labels are restricted to multiples of $q_m$ with the threshold $\\kappa(n)$, the small-divisor estimates and the convergence radius collapse.","fun_headline_variants_meta":{"raw":{"variants":["Invariant graphs persist even as norms diverge","Direct KAM overcomes limits of classical theory","Gevrey perturbations extend graph survival past KAM","Rotation-conjugate graphs beat Brjuno barrier"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000377,"raw_usage":{"total_tokens":1972,"prompt_tokens":878,"completion_tokens":1094,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":494,"completion_tokens_details":{"reasoning_tokens":1034}},"tokens_in":494,"tokens_out":1094,"duration_ms":11233,"temperature":1.0,"reasoning_tokens":1034,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T17:29:06.351796+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a single Fourier mode $\\varphi(x)=e^{2\\pi i q_m x}$ and a frequency $\\omega\\in\\Delta^\\alpha$, and compute the sum over the resonance family of the localized resonance factors in (61)–(64) for a resonance with two entering lines and total momentum zero, with $M(T)<\\kappa(n)$; if this sum is nonzero for any admissible tree, the cancellation lemma is false and the proof of Theorem 1 breaks down.","supporting_citations":[],"review_version":1}