{"id":"356fdeea-2505-489e-94dc-7bdda238192a","arxiv_id":"1908.07711","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For monic centered hyperbolic polynomials of any degree, the paper derives a second-order expansion of Lyapunov exponents under weighted Lyubich measures and states a sign-symmetry between real and imaginary coefficient derivatives.","lead":"This paper derives a coefficient expansion of Lyapunov exponents for hyperbolic polynomials on Julia sets, weighted by a family of measures. It claims a real-imaginary symmetry of the second derivatives, but the proof covers only a limiting case.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof computes derivatives of the p_j→1 limiting expansion (5.3) and then asserts the same derivatives for every fixed probability vector p; no justification for exchanging limit and derivative is given, and the authors' own equidistributed-case observation already contradicts the method.","rationale":"The reader's weakest assumption identifies precisely the load-bearing gap: the proof differentiates a second-order expansion obtained only in the limit p_j→1 and uses it to assert exact second derivatives for every strictly positive probability vector p. My independent reading of Sections 5 and 6 confirms this. The concern is not a stylistic objection; it is a missing exchange of limits and derivatives, with no uniformity estimate for the remainder in (5.1). Moreover, the p=(1/2,1/2) case shows the method cannot be salvaged by continuity alone, because at the equidistributed measure the paper's own statement that all integrals in (5.1) vanish gives different second coefficients from those in (5.3). Thus the central theorem, as stated and as proved, is not supported. The reader's REJECT verdict is therefore appropriate and I see no reason to change it.","tokens_in":9457,"tokens_out":14821,"duration_ms":196217,"concrete_test":"For d=2, take a fixed strictly positive p, e.g. p=(1/2,1/2), and compute the second-order coefficient of Λ_p(c) around c=0 directly from the expansion (5.1), using the explicit Fourier coefficients a_n = ∏_{ℓ=1}^∞ (p_1 + p_2 e^{2π i n/2^ℓ}) of the weighted Lyubich measure on S¹. Compare ∂²Λ_p/∂α² at c=0 with the value 3/2 obtained by differentiating (5.3). At p=(1/2,1/2) the paper's own vanishing-integral observation already yields 0 rather than 3/2; repeating the calculation at p=(0.9,0.1) determines whether the coefficients depend on p, which would settle the validity of the p-independence claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5 derives equations (5.2) and (5.3) only in the limit 'as p_j ↑ 1'. Section 6 then directly differentiates these limiting expressions and concludes Theorems 1.1 and 1.2 for every strictly positive probability vector p. This inference is invalid unless one proves that differentiation with respect to the coefficients commutes with the limit p_j→1 and that the O(Σξ_r≥3) remainder in (5.1) has controlled derivatives. Neither is established, and the measure μ_p itself depends on p, so fixed-p Taylor coefficients cannot be read off from the p_j→1 limit without uniformity in p and in the coefficients. The gap is not merely formal: for d=2 and p=(1/2,1/2), the weighted Lyubich measure is the equidistributed Lyubich measure, and Section 5 explicitly observes that every integral in (5.1) vanishes for that measure. Hence the paper's own expansion gives Λ = -log2 + O(|A|^3), so ∂²Λ/∂α² at A=0 is 0, whereas differentiation of (5.2) gives 3/2. This shows the claimed p-independence of the differentiated coefficients is unsupported and in fact contradicted by the paper's own statements when the limit is not taken.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies a monic, centred hyperbolic polynomial P(z)=z^d+Σ_{r=0}^{d-2}(α_r+iβ_r)z^r with |α_r+iβ_r|<1 and considers its Lyapunov exponent Λ_{μ_p}(P) = -∫_{J_P} log|P'| dμ_p, where μ_p is the weighted Lyubich measure associated with a strictly positive probability vector p=(p_1,...,p_d). Using the analytic conjugacy Φ between z^d and P, the paper derives an expansion of Λ_{μ_p} in the coefficients A_r and states two theorems: Theorem 1.1 claims that ∂²Λ/∂α_r∂α_s = -∂²Λ/∂β_r∂β_s for every r,s and every p, and Theorem 1.2 compares first and second derivatives for real versus complex coefficient polynomials. The proofs in Section 6 differentiate the limiting expansions (5.2) and (5.3), which are obtained only as p_j↑1, and apply the resulting identities to all p.","tokens_in":9791,"tokens_out":11938,"duration_ms":312587,"significance":"If correct, the p-independence of the second-order mixed derivatives in Theorem 1.1 would be a clean and checkable rigidity statement for weighted equilibrium measures on polynomial Julia sets, and it would naturally extend the d=2,3 observations of the authors' earlier paper [12]. The computation is self-contained from the conjugacy equation (2.2) and contains no fitted parameters, so the claim is in principle falsifiable. The obstacle is that the proof as written derives exact coefficient identities from a one-sided limit p_j↑1 and never controls the remainder in the expansion; because μ_p itself depends on p, the passage from the limiting expansion to fixed-p second derivatives is the crux and is not justified.","major_comments":[{"comment":"Equations (5.2) and (5.3) are explicitly obtained only 'as p_j ↑ 1', yet the proof of Theorems 1.1 and 1.2 differentiates these expressions and concludes identities for every strictly positive probability vector. Since the measure μ_p depends on p, the second derivatives at a fixed p are not obtained by differentiating the p_j→1 limit unless one proves that the limit is uniform in the coefficients and commutes with differentiation in A_r. No such uniformity or interchange argument is given. This is the central gap of the paper.","section":"Section 5, Eqs. (5.2)-(5.3); Section 6"},{"comment":"The proof never controls the remainder O(Σξ_r≥3). Theorem 1.1 concerns exact second derivatives on the full domain |A_r|²<1, so the remainder must be shown to be twice differentiable with second derivatives converging to 0 as A→0. The functions φ_r, φ_{r²}, φ_{rs} in (4.3)-(4.5) are infinite series, and the paper does not justify their convergence or their termwise differentiability; therefore the 'direct differentiation' in Section 6 is not a valid derivation.","section":"Section 5, Eq. (5.1)"},{"comment":"The paper states that every integral in (5.1) vanishes when the measure is the equidistributed Lyubich measure. For d=2 and p=(1/2,1/2), the weighted Lyubich measure coincides with the equidistributed measure, so the expansion gives Λ_{μ_p}(P)=-log 2+O(|A_0|^3) at A_0=0. The second-order coefficient obtained by differentiating (5.2) is 3/2 in that case. This does not by itself disprove the identity in Theorem 1.1, but it shows that the p-independent coefficient values used in the proof are not available at p=(1/2,1/2), and it makes the p-dependence of the coefficients concrete rather than a mere formal obstruction.","section":"Section 5, paragraph after Eq. (5.1)"},{"comment":"The limiting values of the integrals are asserted without derivation, for both P_R and P_C. These limits are the only quantitative input that produces (5.2) and (5.3), and they involve the complicated infinite series φ_r, φ_{r²}, and φ_{rs}. No calculation, no interchange of sums and integrals, and no convergence argument is presented, so the central computation cannot be checked from the manuscript.","section":"Section 5, displayed limiting values"}],"minor_comments":[{"comment":"The sum over s≠r has coefficient (d−r−s)/(d−1)^2, but differentiating the cross term in (5.2) gives (d−r−s+1)/(d−1)^2; the displayed formula should be corrected.","section":"Section 6, displayed first derivative"},{"comment":"The summation notation '∑_{r=0}^{d−3}∑_{r<s=1}^{d−2}' is malformed; this should be written with an explicit sum over 0≤r<s≤d−2.","section":"Equation (5.1)"},{"comment":"The theorem uses the notation Λ_μ and Λ_{μ_p} inconsistently; the measure should be named once and the p-dependence made explicit throughout.","section":"Theorem 1.1"},{"comment":"The text contains numerous typographical errors (for example, 'ponit s', 'senstivity', 'exponen ts', 'co efficient'); the manuscript needs a careful proofreading pass.","section":"Abstract and Introduction"},{"comment":"The integrand uses Φ(z) although the conjugacy was denoted Φ_P; the notation should specify which conjugacy is integrated and with respect to which variable.","section":"Equation (3.4)"}],"recommendation":"reject","confidential_remarks":"To the editor: The paper is built on the authors' own prior work, so there is no novelty-disclosure concern. My main worry is that the gap in Sections 5–6 is not a presentation issue: the fixed-p statement does not follow from a p_j→1 expansion, and the paper's own equidistributed observation shows that the second-order coefficients are p-dependent. If the authors can prove a uniform-in-p expansion or restrict the theorem to the p_j→1 limit, a substantially revised manuscript might be worth reconsidering."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper extends the authors' earlier quadratic and cubic computation to arbitrary degree d. The explicit coefficient functions φr, φr2, φrs and the second-order expansion for the Lyapunov exponent are real computations, and the sign symmetry ∂²Λ/∂α_r∂α_s = −∂²Λ/∂β_r∂β_s is a clean observation. If the goal was just to document that computation, this is mostly fine.\n\nThe problem is the theorem as stated. Section 5 explicitly evaluates the integrals only \"when one of the p_j ↑ 1\" and gives formulas (5.2)–(5.3) with an arrow. Section 6 differentiates those limiting formulas and asserts the second derivatives for every strictly positive probability vector p. That inference needs uniformity in p and control of the O(≥3) remainder; neither is supplied. The weighted measure itself depends on p, so there is no reason the coefficient of |A|² in a p-dependent expansion can be read off from the p→degenerate limit.\n\nAnd it is not just a missing technical hypothesis—the paper's own statements show the claim is false as written. For d=2 and p=(1/2,1/2), the weighted Lyubich measure is the equidistributed Lyubich measure. Section 5 notes that every integral in (5.1) vanishes for that measure, so Λ = −log2 + O(|A|³) and ∂²Λ/∂α² at A=0 is 0. Differentiating (5.2) gives 3/2. The same issue appears for cubic.\n\nThe integral evaluations in Section 5 are also listed without derivation, and the O(≥3) terms in (5.1) are never controlled, so even in the limiting case the expansion is not rigorously established beyond formal power series. That said, the coefficient functions from the conjugacy equation are derived cleanly and appear correct as formal expansions.\n\nBottom line: as a research claim, Theorem 1.1 is not supported; the p-independence in particular is contradicted by the equidistributed case. There may be a salvageable restricted result—for example, a version holding only in the p_j→1 limit, or with a careful statement about which derivatives survive. The computation itself could be useful to someone who wants explicit expansions for Lyapunov exponents of polynomials.\n\nFor peer review: I would send it out—the computation is nontrivial and the flaw is the kind a referee could catch and fix. But I would not cite the theorem in its present form. My recommendation: engage with the computation, not the theorem; expect major revision.","headline":"The paper computes a degree-d expansion for Lyapunov exponents and proves a sign symmetry only by differentiating a p_j→1 limit, and that limit exchange is false—their own equidistributed case contradicts it.","tokens_in":10251,"tokens_out":2226,"would_cite":false,"duration_ms":491777,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37B25","37F15","37F10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For hyperbolic polynomials of any degree, the Lyapunov exponent's second derivatives in real and imaginary coefficient directions are opposite, under every weighted Lyubich measure.","keywords":["Lyapunov exponents","weighted Lyubich's measures","Julia sets","hyperbolic polynomials","pressure function","second derivatives","coefficient dependence","complex dynamics"],"falsifier":"Compute the Lyapunov exponent numerically for a concrete cubic, say $z^3+(\\alpha_1+i\\beta_1)z+(\\alpha_0+i\\beta_0)$ with small real coefficients, using the weighted Lyubich measure approximated by $\\sum_{\\eta}p_{\\eta_1}\\cdots p_{\\eta_n}\\delta_\\omega$ over the $d^n$ preimages of a generic point, with weights such as $(1/2,1/4,1/4)$ rather than any weight near 1. Estimate $\\partial^2\\Lambda/\\partial\\alpha_1^2$ and $\\partial^2\\Lambda/\\partial\\beta_1^2$ by finite differences and check whether they equal $+1/4$ and $-1/4$, with the sum exactly zero; if the sum is nonzero or depends on the chosen weights, the theorem's 'irrespective of the probability vector' clause fails.","tokens_in":9250,"feed_emoji":"🔄","tokens_out":12586,"duration_ms":110127,"temperature":0.7,"pith_summary":"The paper sets out to prove that the Lyapunov exponent of a monic, centred, hyperbolic polynomial—computed on its Julia set with respect to a weighted Lyubich's measure—has a rigid second-order symmetry: perturbing any coefficient's real part bends the exponent exactly as much as perturbing its imaginary part bends it in the opposite direction. The setting is $P(z)=z^d+\\sum_{r=0}^{d-2}(\\alpha_r+i\\beta_r)z^r$ with $\\alpha_r^2+\\beta_r^2<1$ and bounded critical orbit, and the weighted Lyubich measure assigns probabilities $p_1,\\dots,p_d$ to the $d$ preimage branches of a generic point. By expanding the topological conjugacy between $P$ and $z^d$ in powers of the coefficients and evaluating the resulting integrals in the limit $p_j\\uparrow 1$, the paper obtains an explicit second-order expansion whose cross-derivatives satisfy $\\partial^2\\Lambda/\\partial\\alpha_r\\partial\\alpha_s=-\\partial^2\\Lambda/\\partial\\beta_r\\partial\\beta_s$. If correct, this quantifies a well-behaved pattern inside a quantity that usually measures instability, and it extends a phenomenon previously seen only for quadratic and cubic polynomials to all degrees.","feed_headline":"Lyapunov exponents obey a real–imaginary sign flip","feed_subtitle":"At second order, real and imaginary coefficient perturbations bend the Lyapunov exponent oppositely, for any degree.","key_machinery":"The load-bearing object is the topological conjugacy $\\Phi_P\\colon S^1\\to J_P$ between the map $Q(z)=z^d$ on the unit circle and the hyperbolic polynomial $P$, defined by $P\\circ\\Phi_P=\\Phi_P\\circ Q$. Because $P$ is hyperbolic and all its coefficients have modulus below 1, this conjugacy depends analytically on the coefficients, and the paper expands it as\n$$\\Phi_P(z)=z+\\sum_{\\xi_{d-2}+\\cdots+\\xi_0\\ge1}\\varphi_{(\\xi_{d-2},\\dots,\\xi_0)}(z)A_{d-2}^{\\xi_{d-2}}\\cdots $A_0^{{\\xi_0}}$,$$\nthen solves for the coefficient functions $\\varphi_r,\\varphi_{r^2},\\varphi_{rs}$ by matching powers in the defining equation. The Lyapunov exponent is re-expressed through the pressure identity as $-\\log d-\\int_{S^1}\\log|\\Phi_P(z)|\\,d\\nu$, so the whole computation reduces to integrals of these coefficient functions against the weighted Lyubich measure; those integrals are evaluated in the limit $p_j\\uparrow 1$, yielding the second-order expansion (5.3) that carries the theorem.","core_discovery":"The central claim is an identity of mixed partial derivatives. For the polynomial $P$ above and the weighted Lyubich measure $\\mu_{\\vec p}$ built from any strictly positive probability vector $\\vec p$, the Lyapunov exponent $\\Lambda_{\\mu_{\\vec p}}(P)=-\\int_{J_P}\\log|P'|\\,d\\mu_{\\vec p}$ satisfies\n$$\\frac{\\$partial^{2}$\\Lambda_{\\mu_{\\vec p}}}{\\partial\\alpha_r\\partial\\alpha_s}=-\\frac{\\$partial^{2}$\\Lambda_{\\mu_{\\vec p}}}{\\partial\\beta_r\\partial\\beta_s}$$\nfor all $0\\le r,s\\le d-2$. The proof computes, as one weight $p_j$ approaches 1, the expansion\n$$\\Lambda\\to -\\log d+\\sum_{r=0}^{d-2}\\frac{\\alpha_r}{d-1}+\\sum_{r=0}^{d-2}\\frac{d-2r+1}{2(d-1)^2}(\\$alpha_r^{2}$-\\$beta_r^{2}$)+\\sum_{0\\le r<s\\le d-2}\\frac{d-r-s+1}{(d-1)^2}(\\alpha_r\\alpha_s-\\beta_r\\beta_s),$$\nand the identity is read off term by term. The paper also shows that using only real coefficients leaves every first and second $\\alpha$-derivative unchanged, so complex coefficients enter the quadratic part solely through the $\\beta^2$ and $\\beta_r\\beta_s$ terms, with signs opposite to their $\\alpha$ counterparts.","pith_inferences":["If the second-order expansion is exact, the Hessian of $\\Lambda$ on the $2(d-1)$-dimensional real coefficient space has block form with real block $H$ and imaginary block $-H$, so the quadratic part is the real part of a holomorphic quadratic form $\\sum_{r\\le s}c_{rs}A_rA_s$; the paper does not state this geometric reading explicitly.","A numerical finite-difference check of the second derivatives for weights far from the vertex (for example uniform weights) would test whether the 'every strictly positive probability vector' clause holds beyond the $p_j\\uparrow 1$ limit in which the expansion is derived.","The same coefficient-expansion machinery could be applied to third derivatives; if the sign-flip pattern persists in the next order, the symmetry may reflect a general real–imaginary structure of the Lyapunov exponent as a function of holomorphic coefficients."],"forward_implications":["For a quadratic polynomial $z^2+\\alpha+i\\beta$, the expansion reduces to $-\\log 2+\\alpha+\\frac32(\\alpha^2-\\beta^2)$, so the second $\\alpha$-derivative is $3$ and the second $\\beta$-derivative is $-3$.","For a cubic polynomial $z^3+(\\alpha_1+i\\beta_1)z+(\\alpha_0+i\\beta_0)$, the identity holds simultaneously for both coefficients, with diagonal contributions $\\frac12(\\alpha_0^2-\\beta_0^2)$ and $\\frac14(\\alpha_1^2-\\beta_1^2)$ and a cross contribution $\\frac34(\\alpha_1\\alpha_0-\\beta_1\\beta_0)$.","The statement is made for every strictly positive probability vector, so the sign-flip identity is presented as a feature of the whole family of weighted Lyubich measures, not of a special choice of weights.","Real–complex agreement of first and second $\\alpha$-derivatives means that, up to second order, the imaginary part of each coefficient affects the Lyapunov exponent only through terms of the opposite sign."],"supporting_citations":[{"why":"the earlier quadratic and cubic computation whose observed sign flip this paper generalizes, and the source of the conjugacy-based method","marker":"[12]"},{"why":"structural stability of hyperbolic polynomial families, giving the topological conjugacy on which the whole expansion rests","marker":"[8]"},{"why":"convergence of preimage counting measures to the Lyubich measure, which underpins the definition of weighted Lyubich measures","marker":"[13]"},{"why":"thermodynamic formalism and the analytic dependence of the conjugacy on coefficients, used to justify the power-series expansion","marker":"[15]"},{"why":"complex dynamics background for the Julia set and for the analytic dependence of the conjugacy","marker":"[4]"},{"why":"the variational and pressure formalism used to rewrite the Lyapunov exponent as an integral involving the conjugacy","marker":"[14]"}],"fun_headline_variants":["Lyapunov mixed second derivatives: real and imaginary signs oppose","For any degree, real and imaginary parts of Lyapunov exponent oppose","Complex quadratic terms in Lyapunov exponent are sign-opposite to real","Second-order real and imaginary perturbations bend Lyapunov oppositely"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that the second-order expansion computed in the limit $p_j\\to1$ can be differentiated to give the exact second derivatives of the Lyapunov exponent for every strictly positive probability vector and for every coefficient in the domain $|A_r|^2<1$, meaning the discarded higher-order coefficient terms in (5.1) do not contribute to those derivatives.","fun_headline_variants_meta":{"raw":{"variants":["Lyapunov mixed second derivatives: real and imaginary signs oppose","For any degree, real and imaginary parts of Lyapunov exponent oppose","Complex quadratic terms in Lyapunov exponent are sign-opposite to real","Second-order real and imaginary perturbations bend Lyapunov oppositely"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001312,"raw_usage":{"total_tokens":5330,"prompt_tokens":914,"completion_tokens":4416,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":530,"completion_tokens_details":{"reasoning_tokens":4339}},"tokens_in":530,"tokens_out":4416,"duration_ms":32253,"temperature":1.0,"reasoning_tokens":4339,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:01:18.362655+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the Lyapunov exponent numerically for a concrete cubic, say $z^3+(\\alpha_1+i\\beta_1)z+(\\alpha_0+i\\beta_0)$ with small real coefficients, using the weighted Lyubich measure approximated by $\\sum_{\\eta}p_{\\eta_1}\\cdots p_{\\eta_n}\\delta_\\omega$ over the $d^n$ preimages of a generic point, with weights such as $(1/2,1/4,1/4)$ rather than any weight near 1. Estimate $\\partial^2\\Lambda/\\partial\\alpha_1^2$ and $\\partial^2\\Lambda/\\partial\\beta_1^2$ by finite differences and check whether they equal $+1/4$ and $-1/4$, with the sum exactly zero; if the sum is nonzero or depends on the chosen weights, the theorem's 'irrespective of the probability vector' clause fails.","supporting_citations":[{"cited_title":"The dependence of Lyapunov exponents of poly- nomials on their coefficients","cited_arxiv_id":null,"evidence_quote":"the earlier quadratic and cubic computation whose observed sign flip this paper generalizes, and the source of the conjugacy-based method"},{"cited_title":"The dynamics of rational transforms: the topological picture","cited_arxiv_id":null,"evidence_quote":"structural stability of hyperbolic polynomial families, giving the topological conjugacy on which the whole expansion rests"},{"cited_title":", Rational iteration: complex analytic dynamical systems , De Gruyter studies in Mathematics, 16, Walter de Gruyter and Co., Berlin, (1993)","cited_arxiv_id":null,"evidence_quote":"convergence of preimage counting measures to the Lyubich measure, which underpins the definition of weighted Lyubich measures"},{"cited_title":", Formalisme thermodynamique et syst´ emes dynamiques holom or- phes”, Panoramas et Synth´ eses,4, (1996)","cited_arxiv_id":null,"evidence_quote":"thermodynamic formalism and the analytic dependence of the conjugacy on coefficients, used to justify the power-series expansion"},{"cited_title":"and Gamelin, T.W","cited_arxiv_id":null,"evidence_quote":"complex dynamics background for the Julia set and for the analytic dependence of the conjugacy"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"the variational and pressure formalism used to rewrite the Lyapunov exponent as an integral involving the conjugacy"}],"review_version":1}