{"id":"e8b4289c-8090-4903-9cf8-623ec6d86d63","arxiv_id":"2506.02855","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A global topological conjugacy is established between a nonlinearly perturbed nonautonomous system and its linear part, assuming a nonuniform μ-dichotomy and a small, time-decaying Lipschitz perturbation.","lead":"This paper proves a global linearization theorem for nonautonomous differential equations with unbounded Lipschitz perturbations, under a generalized hyperbolic splitting called nonuniform μ-dichotomy. If correct, it extends a classical result by Pugh, which required bounded perturbations, to systems where the perturbing force can grow with the state.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 26's continuity proof for F_s at x0=0 uses the unproved bound μ(ℓ_s(τ,x0))^{sign(ℓ_s)ε} ≤ B; as x0→0 the crossing time ℓ_s→−∞ and μ(ℓ_s)^{-ε}→∞, so the estimate (35) cannot hold and the homeomorphism is not established.","rationale":"The central claim of Theorem 13 is the global topological conjugacy, and the proof reduces it to the splitting lemma plus the two linearization lemmas, Lemma 26 and Lemma 27. Lemma 26 is the stable half, and its only argument for continuity of F_s at x0=0 is the estimate (35), which imports the asserted bound on μ(ℓ_s)^{sign(ℓ_s)ε}. That bound is false in the scalar linear example with f=0, which satisfies every stated hypothesis. This is an internal estimate failure, not merely a deviation from the existing literature. I independently verified the reader's weakest assumption and agree that it is the same load-bearing concern. Because the flaw is local and the overall strategy may be salvageable with a sharper variation-of-constants argument, the reader's CONDITIONAL verdict is appropriate and my read does not change it.","tokens_in":34311,"tokens_out":8909,"duration_ms":93078,"concrete_test":"Use the minimal admissible example satisfying the hypotheses of Lemma 26: n=1, μ(t)=e^t, A(t)=-1, f(t,x)=0, so D=1, λ_s=-1, ν=ω=θ=0, and the strict quadratic Lyapunov function from Proposition 19 is V(t,x)=-c|x| for some c>0. For x0>0 the stable solution is x_s(t)=e^{-(t-τ)}x0, and the crossing time defined by V(ℓ_s,x_s(ℓ_s))=-1 is ℓ_s(τ,x0)=τ+ln(c x0)→−∞ as x0→0. Hence μ(ℓ_s)^{sign(ℓ_s)ε}=e^{-εℓ_s}→∞ for ε>0, directly contradicting the bound used in (35). This settles that the proof's premise fails; a repaired argument must avoid bounding that factor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Lemma 26, after inequality (33) the text asserts 'μ(ℓ_s(τ,x0))^{sign(ℓ_s(τ,x0))ε} ≤ B for any given τ ∈ R and some constant B>0' and then uses this to conclude ∥F_s(τ,x0)∥→0 from (35). This assertion is not a consequence of the hypotheses and is false in natural admissible cases. For initial data with 0<∥x0∥<μ(τ)^{-sign(τ)ε}/C, the proof itself shows ℓ_s(τ,x0)<τ, and because V(τ,x0)→0 as x0→0, the crossing time must tend to −∞. Since μ(t)→0 as t→−∞ and sign(ℓ_s)<0 for ℓ_s<0, we have μ(ℓ_s)^{sign(ℓ_s)ε}=μ(ℓ_s)^{-ε}→∞ whenever ε>0. The strict quadratic Lyapunov function constructed in Proposition 19 has ε>0, so this factor cannot be bounded uniformly. Consequently inequality (35) does not prove continuity of F_s at 0, and without that continuity F_s is not proved to be a homeomorphism. Theorem 13 relies on F_s as one of the two linearizing factors, so the central conjugacy is not established by the written proof. The gap appears repairable, but it is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper aims to prove a global topological linearization theorem for nonautonomous ODEs x' = A(t)x + f(t,x) under the assumption that the linear part admits a nonuniform μ-dichotomy and that f is a Lipschitz perturbation with a time-decaying Lipschitz constant φ(t), so that f need not be globally bounded. The proof strategy is to construct global stable and unstable invariant manifolds and foliations (Lemmas 16-17), use them to split the system into contractive and expansive parts (Lemma 24), develop strict quadratic Lyapunov functions for nonuniform μ-dichotomies (Propositions 18-23), and then define linearizing homeomorphisms Fs and Fu via crossing times of the Lyapunov function level sets (Lemmas 26-27). The final conjugacy is obtained by composing these homeomorphisms with the splitting homeomorphism S. The main theorem is Theorem 13.","tokens_in":34620,"tokens_out":7041,"duration_ms":64483,"significance":"If Theorem 13 is correct, it is a meaningful advance: it removes the global boundedness assumption on the perturbation that has been standard since Pugh and Palmer, and it places the result in the framework of nonuniform μ-dichotomies, covering polynomial and other non-exponential growth rates. The paper provides substantial technical machinery, including explicit Lyapunov function constructions, invariant foliations, and a splitting lemma, and it makes the dependence on the dichotomy rates and the smallness of δf explicit. However, the proof as written contains a load-bearing gap in the continuity proof of the stable linearizing map at the origin, so the main theorem is not yet established.","major_comments":[{"comment":"The proof of continuity of F_s(τ,·) at x0=0 relies on the assertion 'μ(ℓ_s(τ,x0))^{sign(ℓ_s(τ,x0))ε} ≤ B' just before (35), but this bound is not a consequence of the hypotheses and is in fact false for the Lyapunov function used. For 0 < ∥x0∥ < μ(τ)^{-sign(τ)ε}/C, the proof itself shows ℓ_s(τ,x0) < τ; moreover, because V(τ,x0)→0 as x0→0 and V(t,x_s(t,τ,x0)) is strictly increasing in t with limit -∞ as t→−∞, the crossing time ℓ_s(τ,x0) tends to −∞ as x0→0. Since μ(t)→0 as t→−∞ and sign(ℓ_s)=-1 for ℓ_s<0, the factor equals μ(ℓ_s)^{-ε}, which is unbounded whenever ε>0; Proposition 19's strict quadratic Lyapunov function indeed has ε>0. Hence inequality (35) does not follow, and the claimed continuity of F_s at 0 is unproved. Because F_s is one of the two homeomorphisms used to build the conjugacy G = F∘S in Theorem 13, this is a load-bearing gap. The gap appears repairable, for example by a direct estimate of ∥Ψ_s(τ,ℓ_s) x_s(ℓ_s,τ,x0)∥ using the dichotomy rates and the lower bound from Lemma 12, but the present proof is incomplete.","section":"Section 5, proof of Lemma 26, between (33) and (35)"}],"minor_comments":[{"comment":"The exponents sign(t-s) in the statement of Lemma 12 and in its proof refer to an undefined variable s; they should be sign(t-τ). Also, the phrase 'regarding t < τ one can be obtained in a similar way' should read 'the case t < τ can be handled similarly'.","section":"Section 2, Lemma 12"},{"comment":"In the definition of L_s(τ,x0) and in equation (40), the argument of Ψ_s(κ_s, τ) is missing; it should be Ψ_s(κ_s(τ,x0), τ)x0. The same notational omission occurs in the displayed formula for L_s(τ,x0).","section":"Section 5, proof of Lemma 26"},{"comment":"The sentence 'x0 ∈ Rn, 0<∥x0∥<μ(τ)^{-sign(τ)ε}/C =⇒ ℓ_s(τ,x0)<τ' should be phrased as a quantifier: for every x0 ∈ Rn with 0<∥x0∥<μ(τ)^{-sign(τ)ε}/C, we have ℓ_s(τ,x0)<τ. As written, the comma after Rn makes the statement ambiguous.","section":"Section 5, proof of Lemma 26"},{"comment":"The statement of Theorem 13 does not explicitly require μ to be differentiable, although φ(t) in (10) and several computations in Section 5 use μ'(t). The authors should state that μ is a differentiable growth rate in the main theorem or define φ(t) without differentiability.","section":"Section 3.1, Theorem 13"}],"recommendation":"major_revision","confidential_remarks":"The gap in Lemma 26 is the key obstacle; the rest of the proof appears carefully developed and the claimed result is plausible. I recommend major revision and would be willing to review a corrected version. The authors should also double-check that the continuity argument for L_s at x0=0 does not suffer from the same unbounded-factor issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper tackles a real gap: Pugh–Palmer-type global linearization needed bounded perturbations, and the recent unbounded results covered only contractive or uniform hyperbolic systems. Extending this to nonuniform μ-dichotomy is a substantive step, and the Lyapunov-function machinery developed here (strict quadratic Lyapunov functions, the splitting lemma, the invariant foliations) is serious work. I believe the strategy is essentially right and the result is likely true with repairs.\n\nThe soft spot is exactly where the stress-test note points. In Lemma 26, the proof that F_s is continuous at x0=0 needs the bound μ(ℓ_s)^{sign(ℓ_s)ε} ≤ B. That assertion is dropped in without proof, and it is not a minor omission: for small x0 the crossing time ℓ_s(τ,x0) goes to −∞, μ(ℓ_s) goes to 0, and with sign(ℓ_s)<0 the factor μ(ℓ_s)^{-ε} blows up. So inequality (35) does not follow, and without continuity of F_s at 0 the homeomorphism is not established. Since both linearizing maps depend on this step, Theorem 13 is not proved as written. The gap looks repairable—one can imagine rescaling the crossing-time level or arguing differently—but it is load-bearing.\n\nTwo smaller points. The lower-bound estimate in Lemma 12 is stated for x(t,τ,·) but the proof only yields a lower bound for x(τ,t,·); the asymmetry is sloppy. That said, the proof of Lemma 26 uses only the upper bound, so this is not fatal. The reader's complaint about Proposition 19's choice of ϵ does not hold up: the identity μ(τ)^{-sign(τ)ϵ}=μ(|τ|)^{-ϵ} is correct for the growth-rate convention in the paper.\n\nVerdict: this deserves a serious referee. The contribution is significant and the framework is worth engaging with, but the proof needs a substantive fix in Lemma 26 before the theorem is accepted. For a specialist in nonautonomous linearization, this is worth careful reading; for a general audience, the gap matters.","headline":"Plausible, significant extension of global linearization to nonuniform μ-dichotomies, but the proof of the key continuity estimate in Lemma 26 has a load-bearing gap that makes the main theorem unproven as written.","tokens_in":35182,"tokens_out":4600,"would_cite":false,"duration_ms":42563,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37D25","37C86","37C60"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves a global linearization theorem for nonautonomous systems with unbounded perturbations, under a nonuniform μ-dichotomy, using strict quadratic Lyapunov functions and crossing-time maps.","keywords":["global linearization","nonuniform μ-dichotomy","Lyapunov functions","topological conjugacy","unbounded perturbation","nonautonomous systems","invariant foliations","crossing time"],"falsifier":"Inspect the scalar linear equation $x' = -x$ (a nonuniform $\\mu$-dichotomy with $\\mu(t)=e^t$ and $f=0$): as the initial value $x_0$ tends to 0, the crossing time $\\ell_s$ defined by $V(\\ell_s, x_s(\\ell_s)) = -1$ satisfies $e^{\\ell_s} \\asymp x_0$, so $\\ell_s \\to -\\infty$ and $\\mu(\\ell_s)^{\\mathrm{sign}(\\ell_s)\\epsilon} = e^{-\\epsilon\\ell_s}$ is unbounded for any $\\epsilon>0$, directly violating the bound used in inequality (35) of Lemma 26 and showing the continuity argument at $x_0=0$ is not justified as written.","tokens_in":34066,"feed_emoji":"🔄","tokens_out":11764,"duration_ms":102664,"temperature":0.7,"pith_summary":"This paper aims to extend Pugh's global linearization theorem from bounded to unbounded perturbations of linear nonautonomous systems. The authors try to prove that if the linear part admits a nonuniform μ-dichotomy, a hyperbolicity condition whose rates are set by a growth function μ(t), and if the nonlinearity is Lipschitz with a sufficiently small rate function, then the nonlinear flow is globally topologically conjugate to the linear flow. This would matter because all earlier global linearization theorems required the perturbation to be globally bounded, which is rarely true in applications. The result is designed to cover exponential, polynomial, and other non-exponential dichotomy rates, with perturbations that may grow linearly in the state but are controlled by a small Lipschitz rate.","feed_headline":"Unbounded perturbations still allow global linearization","feed_subtitle":"Conjugates nonlinear nonautonomous flows to their linear parts with no boundedness assumption.","key_machinery":"The paper's engine is the family of strict quadratic Lyapunov functions $V(t,x) = -\\mathrm{sign}\\,U(t,x)\\sqrt{|U(t,x)|}$ built from a time-dependent symmetric operator $S(t)$ that integrates the stable and unstable evolutions of the linear part. These functions characterize nonuniform $\\mu$-dichotomy: existence of such a Lyapunov function is proved from the dichotomy, and conversely the dichotomy follows from the Lyapunov function together with bounded-growth estimates. The linearization map is then defined through the crossing time $\\ell_s(\\tau,x_0)$ at which a trajectory of the contractive part reaches the level set $V=-1$, together with the analogous map for the expansive part; these crossing-time maps conjugate the nonlinear flow to the linear flow. A splitting lemma decouples the hyperbolic system into contractive and expansive subsystems before the crossing-time construction is applied.","core_discovery":"The paper's central claim is Theorem 13: for the nonautonomous nonlinear system $x' = A(t)x + f(t,x)$ with $f$ a $C^0$-Carathéodory function satisfying $f(t,0)=0$ and a Lipschitz estimate with rate $\\varphi(t) = \\delta_f \\mu(t)^{-1-\\mathrm{sign}(t)\\theta}\\mu'(t)$ for a small constant $\\delta_f$, if the linear part admits a nonuniform $\\mu$-dichotomy (Definition 8) and if $\\theta \\ge \\max\\{\\nu,\\omega\\}$, $\\lambda_s < \\nu-\\theta$, and $\\lambda_u > \\theta-\\omega$, then the nonlinear system is globally topologically conjugate to its linear part. This is the first global linearization result that allows the perturbation to be unbounded while the hyperbolicity is merely nonuniform, subsuming exponential dichotomies and polynomial-type dichotomies as special cases. The autonomous corollary recovers Pugh's theorem with the boundedness hypothesis on the perturbation removed.","pith_inferences":["It would be natural to try the same Lyapunov-plus-crossing-time construction in discrete time, extending global unbounded linearization to nonautonomous maps with nonuniform $\\mu$-dichotomies.","The explicit quadratic Lyapunov operators $S(t)$ could support Hölder or smooth linearization results under additional regularity assumptions on $A$ and $f$, a direction this paper does not explore.","The crossing-time formulation suggests a concrete numerical test: approximate the conjugacy by integrating trajectories to the Lyapunov unit sphere and recording the renormalized state; this would also check the continuity of the constructed maps at the origin."],"forward_implications":["Corollary 14 provides a global conjugacy for nonuniform exponential dichotomies with unbounded perturbations, a case not covered by earlier bounded-perturbation theorems.","Corollary 15 recovers Pugh's autonomous linearization theorem while dropping the boundedness hypothesis on the perturbation; only a small Lipschitz constant is needed.","Systems with nonuniform polynomial dichotomies, corresponding to growth functions like $\\mu(t)=t+1$ for $t\\ge 0$ and its reciprocal for $t\\le 0$, are included, so global topological linearization is now available for a wider class of rates.","The strict quadratic Lyapunov functions established in Propositions 19 and 23 provide explicit forward-invariant cone estimates that are themselves usable for stability and reducibility questions beyond linearization."],"supporting_citations":[{"why":"Supplies the classical global linearization theorem being generalized: bounded small-Lipschitz perturbations of hyperbolic linear isomorphisms are topologically conjugate to the linear part.","marker":"[41]"},{"why":"Extends Pugh's theorem to nonautonomous systems with exponential dichotomy and globally bounded perturbations; it is the baseline the new result removes the boundedness from.","marker":"[38]"},{"why":"Introduces the nonuniform $\\mu$-dichotomy and its spectrum theory, the hyperbolicity notion on which the whole paper rests.","marker":"[43]"},{"why":"Establishes the relation between quadratic Lyapunov functions and nonuniform exponential dichotomies, the model for the Lyapunov machinery developed here.","marker":"[7]"},{"why":"Proves a partially unbounded linearization result for uniform hyperbolic nonautonomous systems; the crossing-time construction is inspired by and extends this work.","marker":"[45]"},{"why":"Provides unbounded linearization for contractive systems via the crossing-time/sphere-intersection technique that the proof adapts to the contractive and expansive parts.","marker":"[14]"},{"why":"Supplies the Lyapunov-Perron invariant manifold method used in the splitting lemma to decouple the system into contractive and expansive components.","marker":"[46]"}],"fun_headline_variants":["Linearization survives unbounded perturbations","Global linearization for unbounded nonautonomous systems","Nonuniform μ-dichotomy unlocks unbounded linearization","Pugh's theorem generalized: no boundedness needed"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Lyapunov function evaluated at the trajectory's crossing time stays bounded as the initial data goes to zero, a bound the proof asserts without support even though the crossing time escapes to $-\\infty$ where $\\mu(t)$ collapses to zero.","fun_headline_variants_meta":{"raw":{"variants":["Linearization survives unbounded perturbations","Global linearization for unbounded nonautonomous systems","Nonuniform μ-dichotomy unlocks unbounded linearization","Pugh's theorem generalized: no boundedness needed"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000635,"raw_usage":{"total_tokens":2954,"prompt_tokens":993,"completion_tokens":1961,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":609,"completion_tokens_details":{"reasoning_tokens":1901}},"tokens_in":609,"tokens_out":1961,"duration_ms":14533,"temperature":1.0,"reasoning_tokens":1901,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:17:45.619531+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inspect the scalar linear equation $x' = -x$ (a nonuniform $\\mu$-dichotomy with $\\mu(t)=e^t$ and $f=0$): as the initial value $x_0$ tends to 0, the crossing time $\\ell_s$ defined by $V(\\ell_s, x_s(\\ell_s)) = -1$ satisfies $e^{\\ell_s} \\asymp x_0$, so $\\ell_s \\to -\\infty$ and $\\mu(\\ell_s)^{\\mathrm{sign}(\\ell_s)\\epsilon} = e^{-\\epsilon\\ell_s}$ is unbounded for any $\\epsilon>0$, directly violating the bound used in inequality (35) of Lemma 26 and showing the continuity argument at $x_0=0$ is not justified as written.","supporting_citations":[{"cited_title":"Pugh, On a theorem of P","cited_arxiv_id":null,"evidence_quote":"Supplies the classical global linearization theorem being generalized: bounded small-Lipschitz perturbations of hyperbolic linear isomorphisms are topologically conjugate to the linear part."},{"cited_title":"Palmer, A generalization of Hartman’s linearization theorem, J","cited_arxiv_id":null,"evidence_quote":"Extends Pugh's theorem to nonautonomous systems with exponential dichotomy and globally bounded perturbations; it is the baseline the new result removes the boundedness from."},{"cited_title":"Silva, Nonuniform µ-dichotomy spectrum and kinematic similarity, J","cited_arxiv_id":null,"evidence_quote":"Introduces the nonuniform $\\mu$-dichotomy and its spectrum theory, the hyperbolicity notion on which the whole paper rests."},{"cited_title":"Barreira, C","cited_arxiv_id":null,"evidence_quote":"Establishes the relation between quadratic Lyapunov functions and nonuniform exponential dichotomies, the model for the Lyapunov machinery developed here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves a partially unbounded linearization result for uniform hyperbolic nonautonomous systems; the crossing-time construction is inspired by and extends this work."},{"cited_title":"Casta˜ neda, G","cited_arxiv_id":null,"evidence_quote":"Provides unbounded linearization for contractive systems via the crossing-time/sphere-intersection technique that the proof adapts to the contractive and expansive parts."},{"cited_title":"Zhang, Generalized exponential dichotomies and invariant manifolds for differential equations, Adv","cited_arxiv_id":null,"evidence_quote":"Supplies the Lyapunov-Perron invariant manifold method used in the splitting lemma to decouple the system into contractive and expansive components."}],"review_version":1}