Estimation of topological entropy in random dynamical systems

https://doi.org/10.1016/j.jde.2019.12.030Get rights and content

Abstract

For Cr(r>1) random dynamical systems F on a compact smooth Riemannian manifold M, the fiber topological entropy is bounded above by an integral formula. Particularly, for the C random dynamical systems, the integral formula coincides with the fiber topological entropy.

Introduction

In 1965, Adler, Konheim and McAndrew [1] introduced a notion called topological entropy into the study of topological dynamical systems. Afterwards, it turned out to be a canonical conjugate invariant, which measures the exponential complexity of dynamical systems. As one of the most critical invariants to categorize different dynamical systems, calculating the topological entropy in different dynamical systems is of great interest to people. However, calculation through various definitions is difficult. In [18], [6], [17], Goodwyn, Dinaburg, and Goodman proved a fundamental relationship between topological entropy and measure-theoretic entropy introduced by Kolmogorov in [9], which is called variational principle. The variational principle gives an estimation of topological entropy by the optimal measure-theoretic entropy of invariant measures, but finding all invariant measures is hardly feasible. In this paper, we mainly concern the estimation of fiber topological entropy of differentiable random dynamical systems (RDS).

Let (Ω,F,P) be a complete probability space, and θ be an invertible P-invariant map on it, then (Ω,F,P,θ) is called a metric dynamical system. Let M be a compact smooth Riemannian manifold without boundary, Diffr(M) be the space of Cr(r>1) diffeomorphisms on M. Let F:ΩDiffr(M), ωFω be a Borel measurable map. We consider the RDS F generated by {Fω,ωΩ} over the base space (Ω,F,P,θ), and letΘ:Ω×MΩ×M,(ω,x)(θω,Fωx) be the corresponding skew product transformation.

For nN and a positive random variable ε=ε(ω) define a family of metrics dε,nω on Mω by the formuladε,nω=max0i<n{d(Fωix,Fωiy)ε(θiω)1,x,yMω}, where d(,) is the Riemannian metric on M. Denote by B(ω,x)n(ε) the closed ball in Mω centered at x of radius 1 under metric dε,nω. We say x,yMω are (ω,ε,n)-close if dε,nω(x,y)1. A set QMω is called (ω,ε,n)-separated set if for x,yQ, xy implies dε,nω(x,y)>1.

Due to compactness of Mω, there exists a smallest natural number s(ω,ε,n) such that Card(Q)s(ω,ε,n)< for every (ω,ε,n)-separated set Q. Furthermore, there exists a maximal (ω,ε,n)-separated set Q(ω,ε,n) such that for any xMωQ, Q{x} is not (ω,ε,n)-separated set anymore. Therefore, MωxQB(ω,x)n(ε).

The fiber topological entropy is defined ashtop(F)=htop(r)(Θ)=limε0limsupn1nlogs(ω,ϵ,n)dP.

In deterministic dynamical systems, the estimation of topological entropy has a long history. In [3], [19], [20], and [22], the topological entropy for tori is the sum of logarithm of eigenvalues with modulo greater than 1. Afterwards, Bowen gave the same formula for entropy of an endomorphism of Lie group in [5]. Sacksteder and Shub made an attempt in [21] to estimate the topological entropy for differentiable map on a compact manifold. Then Przytycki proved that the topological entropy of a differentiable map on a compact manifold is bounded above by the (exponential) growth rate of an integral formula in [23]. Newhouse extended the result to non-invertible maps in [16]. In [25], Yomdin proved an inequality of volume growth and entropy, also pointed out that the topological entropy is equal to the volume growth for C maps. In [10], Kozlovski combined the techniques from Przytycki, Newhouse, and Yomdin to estimate the topological entropy by the growth rate of the integral formula Przytycki used. In the case of RDSs, there are several proceeding works. In [7], Kifer and Yomdin extended the work of volume growth and entropy for a sequence of independent random smooth maps of a manifold. In [27], Zhu proved the growth in volume, fundamental groups, homological groups are bounded by topological entropy for smooth RDSs. In this paper, we have the following theorem, which estimates fiber topological entropy of RDSs by an integral formula.

Theorem 1

Let F be a Borel measurable mapping from Ω to Diff2(M) satisfies Ωlog+DFωC2dP<+, we have the following estimate:htop(F)limsupn1nlog((DxFωn)dx)dP(ω). Furthermore, if F:ΩDiff(M) satisfies Ωlog+DFωCdP<+, thenhtop(F)=limsupn1nlog((DxFωn)dx)dP.

Note that (DxFωn) is a mapping between exterior algebras of the tangent space TxMω and TFωnxMθnω induced by DxFωn. For kZ+ and Ck diffeomorphisms, DFωCk=sups=1,,kDsFω, and DFωC=supsZ+DsFω, where DsFω:=supxMDxsFω.

The structure of this paper is as follows. Firstly, several preliminaries of RDSs. In Section 3, we illuminate that the fiber topological entropy is bounded above by the growth rate of the integral formula, by using the Multiplicative Ergodic Theorem, Lyapunov charts and variational principle. In Section 4, firstly, for the Ck RDSs, the fiber topological entropy plus an k correction is bounded below by the growth rate of the integral formula. The main technique is a modification of Yomdin's work to the RDSs. Then the second part of Theorem 1 holds by letting k tend to infinity and combining the result in Section 3.

Section snippets

Preliminaries

In this section, we will demonstrate the Multiplicative Ergodic Theorem and Lyapunov charts for random dynamical systems for completeness of this paper.

Let F be a Cr RDS over (Ω,F,P,θ), μ be a Θ-invariant probability of F with marginal P on Ω, i.e. it can be disintegrated into dμ=dμωdP. The disintegration μω, for ωΩ is called the sample measures of μ and μω is invariant in the sense of Fωμω=μθω, which means for each Borel measurable subset AMθω, we have μω(Fω1A)=μθω(A). In the following, we

The upper bound of topological entropy of Ck RDS

In this section, we consider the first part of Theorem 1. We rephrase the first part of Theorem 1 in the following lemma for convenience.

Lemma 5

Let F be a Borel measurable mapping from Ω to Diff2(M) satisfies Ωlog+DFωC2dP<+, we have the following estimate:htop(F)limsupn1nlog((DxFωn)dx)dP(ω).

Proof of Lemma 5: For the fiber topological entropyhtop(F)=htop(r)(Θ)=limε0limsupn1nlogs(ω,ϵ,n)dP, where s(ω,ϵ,n) is the upper bound of the cardinality of (ω,ϵ,n)-separated set in each fiber Mω

The lower bound of topological entropy of C RDS

In this section, our goal is to estimate the upper bound of the integral formula for RDSs. There are two steps of the estimation, locally and globally. In other words, we estimate the integral formula on a Bowen ball, then on the whole space, which gives a lower bound of the fiber topological entropy for C RDSs.

Let M be a compact smooth manifold with Riemannian metric d. The random dynamical system F:ΩDiffk(M) be Borel measurable, where k1. And assume thatR˜(F):=Ωlog+DFωdP(ω)<+, where D

Acknowledgement

The author would like to thank Prof. Wen Huang, Prof. Zeng Lian, and Dr. Rui Gao for their insights and opinions. Many thanks to the referee for the careful reading of the manuscript and valuable comments and suggestions. The author is partially supported by NNSF of China (11971455) and The Fundamental Research Funds for the Central Universities (WK3470000014).

References (28)

  • Yu. Kifer et al.

    Random dynamics

  • R. Sacksteder et al.

    Entropy of a differentiable map

    Adv. Math.

    (1978)
  • R.L. Adler et al.

    Topological entropy

    Trans. Am. Math. Soc.

    (1965)
  • L. Arnold

    Random Dynamical Systems

    (1998)
  • D.Z. Arov

    Calculation of entropy for a class of group endomorphisms

    Zap. Meh.-Mat. Fak. Har'kov. Gos. Univ. i Har'kov. Mat. Obšč.

    (1964)
  • L. Barreira et al.

    Introduction to Smooth Ergodic Theory

    (2013)
  • R. Bowen

    Entropy for group automorphism and homogenous spaces

    Trans. Am. Math. Soc.

    (1971)
  • E.I. Dinaburg

    The relation between topological entropy and metric entropy

    Sov. Math.

    (1970)
  • Yu. Kifer et al.

    Volume Growth and Topological Entropy for Random Transformations

    (1988)
  • A.N. Kolmogorov

    A new metric invariant of transient dynamical systems and automorphisms of Lebesgue spaces

    Dokl. Akad. Sci. SSSR

    (1958)
  • O.S. Kozlovski

    An integral formula for topological entropy of C maps

    Ergod. Theory Dyn. Syst.

    (1998)
  • Z. Lian et al.

    Lyapunov exponents and invariant manifolds for random dynamical systems in a Banach space

    Mem. Am. Math. Soc.

    (2010)
  • Z. Lian, X. Ma, Existence of periodic orbits and horseshoes for mappings in a separable Banach space, preprint,...
  • Z. Lian et al.

    Lyapunov exponents, periodic orbits and horseshoes for mappings of Hilbert spaces

    Ann. Henri Poincaré

    (2011)
  • Cited by (1)

    View full text