Non-existence of Lyapunov exponents in the Newhouse domain

Shin Kiriki, Xiaolong Li, Yushi Nakano, Teruhiko Soma · 2026

Read Non-existence of Lyapunov exponents in the Newhouse domain by Shin Kiriki, Xiaolong Li, Yushi Nakano, Teruhiko Soma (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-10913.txt.

Introduction

Kiriki and three colleagues show that Lyapunov exponents can fail to exist throughout a dense subset of the Newhouse domain of surface maps. For any map in that subset, exponents fail for all points in some open set and every nonzero tangent vector in some open cone. This makes non-existence a persistent phenomenon under robust homoclinic tangencies, not an isolated quirk. The proof builds maps with particular oscillatory return times near a homoclinic tangency, using Newhouse theory and recent work on Lyapunov irregularity.

Catalog record

CREATORShin Kiriki, Xiaolong Li, Yushi Nakano, Teruhiko Soma
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE74845
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.10913

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Non-existence of Lyapunov exponents in the Newhouse domain Shin Kiriki, Xiaolong Li, Yushi Nakano, Teruhiko Soma Abstract . We show that within the Newhouse domain of C r C^{r} surface diffeomorphisms ( r ∈ [ 2 , ∞ ) r\in[2,\infty) ), there exists a dense subset 𝒟 \mathcal{D} such that for any f ∈ 𝒟 f\in\mathcal{D} , Lyapunov exponents fail to exist for all points in some open set U U and all nonzero tangent vectors in some open cone V ⊂ ℝ 2 V\subset\mathbb{R}^{2} . This demonstrates that the non-existence of Lyapunov exponents is a persistent phenomenon in the setting of robust homoclinic tangencies. The proof relies on constructing diffeomorphisms exhibiting specific oscillatory return times near a homoclinic tangency, incorporating techniques from Newhouse theory and recent results on Lyapunov irregularity, alongside several refinements and new arguments. 2020 Mathematics Subject Classification 37C05, 37C40, 37C29 1. Introduction Dynamical systems with a robust homoclinic tangency – those in the so-called Newhouse domain – are a major source of extremely non-hyperbolic dynamics and have long been known to exhibit a rich variety of complex dynamical phenomena, such as infinitely many sinks [ 17 ] , infinitely many strange attractors [ 6 ] , super-exponential growth of the number of periodic points [ 7 ] , universal dynamics [ 23 ] , historic wandering domains [ 13 , 12 ] and pluripotency [ 9 ] . This paper focuses on statistical irregularities associated with one of the most fundamental quantities in dynamical systems: the Lyapunov exponent, which quantifies the sensitivity of an orbit to initial conditions and is widely used as an indicator of chaos in the natural sciences. We show that, within the Newhouse domain, Lyapunov exponents persistently fail to exist on open sets of points and tangent vectors: Theorem A . Given a closed surface M and r\in[2,\infty) , there is a dense subset \mathcal{D} of the Newhouse domain of \mathrm{Diff}^{r}(M) such that for any f\in\mathcal{D} , there exist nonempty open sets U\subset M and V\subset\mathbb{R}^{2} , under the identification of TU with U\times\mathbb{R}^{2} , such that the Lyapunov exponent \displaystyle\lim_{n\to+\infty}\frac{\log||Df^{n}(x)v||}{n} does not exist for all x\in U and nonzero v\in V . In light of the widely believed dichotomy suggesting that ‘generic’ surface diffeomorphisms are either hyperbolic or contain homoclinic tangencies, together with its higher-dimensional counterpart involving dynamics with heterodimensional cycles [ 21 , 4 , 20 ] , our result provides significant insight into the statistical properties of non-hyperbolic dynamics. It highlights that complex behavior regarding Lyapunov exponents, specifically their non-existence, widely occurs within these non-hyperbolic systems, potentially contributing to a more complete understanding of dynamics beyond uniform hyperbolicity. Despite the fundamental importance of Lyapunov exponents in characterizing dynamical behavior and their applications in various scientific fields (cf. [ 24 ] ), the phenomenon where these exponents fail to converge for observable points or vectors (i.e. those in a positive Lebesgue measure set) has remained relatively less explored. Early work by Ott and Yorke [ 18 ] presented the first example of such non-convergence in a surface flow with an attracting homoclinic loop. More recently, our previous work [ 10 ] aimed to uncover the underlying mechanisms driving the non-existence of Lyapunov exponents for certain surface diffeomorphisms. This mechanism was subsequently adapted to demonstrate the non-existence of Lyapunov exponents in a class of two-dimensional piecewise expanding maps [ 15 ] . This paper builds upon and integrates the techniques developed in the previous work [ 10 ] with the rich technical machinery accumulated since the era of Newhouse, Palis and Takens [ 19 ] for studying the Newhouse domain. In particular, it leverages the methods for constructing diffeomorphisms with prescribed return time properties of wandering domains developed by Kiriki and Soma [ 13 ] , which were themselves inspired by earlier work of Colli and Vargas [ 5 ] . Furthermore, we introduce several new arguments and refinements. Specifically, accurately estimating certain newly arising off-diagonal terms (cf. the (1,2)-entry of ( 4.20 )) in the derivative matrices presents significant challenges. This necessitated modifications to the arguments, the development of a new ‘Modified Rectangle Lemma’ (Section 3 ), an inductive analysis of the products of Jacobian matrices for the lower bounds needed to control the relevant coefficients despite the presence of newly arising off-diagonal terms (Section 5 ), and a fine-tuning of the arguments both in [ 13 ] and [ 10 ] to establish the required properties, including specific conditions on strong dissipativity, and exponential oscillations of return times (conditions detailed in Section 2 ). Another observation in our construction suggests that the mechanism leading to the non-existence of Lyapunov exponents identified here operates independently of Birkhoff irregularity (non-existence of time averages). Indeed, one can show the convergence of time averages for any continuous observable at every point within the open set U of Theorem A on which the Lyapunov exponent does not exist (see Remark 2.6 ), emphasizing that these two quantities are independent. We close this introduction by presenting an application of Theorem A to the Hénon family, i.e., diffeomorphisms on \mathbb{R}^{2} given by f_{a,b}(x,y)=(1-ax^{2}+by,x) with (a,b)\in\mathbb{R}^{2} , in sharp contrast to pioneering works [ 1 , 14 , 3 , 2 ] on the existence and positivity of Lyapunov exponents of the Hénon family. The following is an immediate consequence of Theorem A . Corollary 1.1 . There is an open set \mathcal{O} in \mathbb{R}^{2} with (2,0)\in\overline{\mathcal{O}} such that, for every (a,b)\in\mathcal{O} , the Hénon map f_{a,b} can be C^{r} -approximated by diffeomorphisms with a nonempty open set on which Lyapunov exponents do not exist. 2. Non-existence of Lyapunov exponents in abstract settings Throughout this paper, let r be a fixed real number with r\geq 2 , M a C^{r} closed surface and \mbox{Diff}^{r}(M) the space of all C^{r} diffeomorphisms of M endowed with the C^{r} topology. A guiding principle of this paper is that one important source of oscillation of Lyapunov exponents is the interaction between hyperbolic dynamics and tangential returns. The former produces strong anisotropic growth, while the latter reconfigures the expanding and contracting directions. The following conditions are designed to capture this hybrid mechanism in an explicit and quantitative form. Definition 2.1 . We say that f\in\mathrm{Diff}^{r}(M) satisfies DH , H , QT and OE if there are two sequences (n_{k}^{H})_{k\geq 1} , (n_{k}^{T})_{k\geq 1} of positive integers with \lim\limits_{k\to\infty}n_{k}^{H}=+\infty and a sequence of points (\zeta_{k})_{k\geq 1} of M such that the following conditions hold. DH (Dominance of hyperbolic return times) For each k\in\mathbb{N} , with the notation n_{k}=n_{k}^{H}+n_{k}^{T} , we have f^{n_{k}}(\zeta_{k})=\zeta_{k+1}\quad\text{and}\quad\lim_{k\to\infty}\frac{n_{k}^{H}}{n_{k}}=1. H (Strongly dissipative hyperbolic region) There exist an open set K\subset M and a C^{r} coordinate on K in which f(x,y)=(\lambda_{s}x,\,\lambda_{u}y) with constants 0<\lambda_{s}<1<\lambda_{u} satisfying \lambda_{s}\lambda_{u}^{3}<1. Furthermore, there exists a nonempty closed set K^{\prime}\subset K such that f^{j}(\zeta_{k})\in K^{\prime} for every k\geq 1 and 0\leq j\leq n_{k}^{H} . QT (Quadratic tangential return) For each k\geq 1 , writing \zeta_{k+1}^{\prime}=f^{n_{k}^{H}}(\zeta_{k}) , we have Df^{n_{k}^{T}}(\zeta_{k+1}^{\prime})v^{u}\in\mathbb{R}v^{s}. Moreover, there exists a constant C_{T}>0 , independent of k , such that for every line segment \gamma\subset\ell^{u}(\zeta_{k+1}^{\prime})\cap K containing \zeta_{k+1}^{\prime} , \bigl|\pi^{u}_{k+1}(f^{n_{k}^{T}}(\gamma))\bigr|\leq C_{T}\bigl|\pi^{s}_{k+1}(f^{n_{k}^{T}}(\gamma))\bigr|^{2}. Here v^{s}=(1,0) and v^{u}=(0,1) are the horizontal and vertical unit vectors, respectively, \ell^{\sigma}(z) denotes the lines in K through z parallel to v^{\sigma} (

Read the complete text