Pith. sign in

REVIEW 5 major objections 5 minor 19 references

Non uniform expansion and additive noise imply random horseshoe

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Additive noise plus positive Lyapunov exponents force random horseshoes in every open set.

desk verdict A plausible and important claim, but the horseshoe construction stops at a limit statement and never proves the defining conditions; deserves peer review, not acceptance as is. read the letter →

arxiv 2501.11656 v2 pith:PR67FMZS submitted 2025-01-20 math.DS math.PR

Jeroen S.W. Lamb , Giuseppe Tenaglia , Dmitry Turaev This is my paper · ORCID
classification math.DSmath.PR MSC 37H2037B1037D2560J05
keywords randomhorseshoepositiveLyapunovexponentssymbolicdynamicsdynamicalsystemsadditivenoisenon-uniformlyexpandingmapsYoungtimeslargedeviations
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves that a non-uniformly expanding map driven by additive noise must contain random horseshoes, and that these horseshoes are dense: every open set in the ergodic component contains two disjoint balls whose forward images repeatedly cover a common reference set with uniform expansion, yielding a set in one-to-one correspondence with binary sequences. The result matters because random systems typically have no periodic orbits, so the classical deterministic route to horseshoes via hyperbolic periodic points is unavailable; the authors replace it with a purely probabilistic construction. The proof combines ergodicity, large-deviation estimates for Lyapunov averages and critical-set recurrences, and modified Young times that mark when a small neighbourhood is expanded diffeomorphically onto a fixed reference ball $J$. In short, the theorem transfers the classical dense-horseshoe picture from deterministic diffeomorphisms to dissipative random systems under essentially no assumptions beyond diffusive additive noise and an ergodic stationary measure with $\lambda = \int \log \|df^{-1}\|\,d\mu < 0$, i.e. all Lyapunov exponents positive.

What carries the argument

The central objects are the random $\kappa$-horseshoe of Definition 2.1 and the modified Young times of Section 5. A Young time for $(\omega,x)$ is a hyperbolic time followed by an event in which some sub-ball of a large ball is mapped into a fixed reference set $J$ while staying away from the critical set; Proposition 5.4 shows that at such a time $f^n_\omega$ expands a small neighbourhood of $x$ diffeomorphically onto $J$ with uniform derivative and distortion control. The argument also rests on the annealed large-deviation estimates of Proposition 3.1, obtained by the spectral method for Markov chains applied to the family of transfer operators $P_\theta$, and on Proposition 4.1, which uses ergodicity to show that balls of size at least $\varepsilon$ have probability at least $\rho$ of eventually covering $J$. The horseshoe construction takes $M$ small balls of size about $|J|/M$, forms independent sequences of their Young return times, and uses an inclusion-exclusion bound to force two of those sequences to intersect infinitely often with positive lower density; Proposition 6.3 converts the intersection density into the required upper bound on $n_k(\omega)/k$.

What would settle it

Exhibit a map satisfying (H1)--(H2) whose ergodic stationary measure $\mu$ has an open hole $A$ inside its component $Y$, and check whether two balls in $A$ can still form a $\kappa$-horseshoe; if they cannot, the full-support assumption is necessary and the proof's unstated assumption is exposed.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2.2: under Hypotheses (H1)--(H2), there exists $\kappa > 0$ such that the set of $\kappa$-horseshoes is dense. In Definition 2.1, a pair of disjoint balls $(I_0,I_1)$ is a random $\kappa$-horseshoe when there exist random times $n_k(\omega)$ with $\limsup_{k\to\infty} n_k(\omega)/k = \mathbb{E}[n_0]$, the image of each ball under $f^{n_{k+1}-n_k}_{\theta^{n_k}\omega}$ contains $I_0 \cup I_1$, and sub-balls inside each $I_i$ are mapped diffeomorphically onto $I_j$ with inverse derivative norm below $\kappa^{-1}$. The set $P(\omega)$ of points whose orbit at every $n_k$ lies in $I_0 \cup I_1$ is then hyperbolic and in one-to-one correspondence with the full shift on two symbols. The proof establishes this via annealed large-deviation bounds for $\log \|df^{-1}\|$ and for close approaches to the critical set, an 'eventually onto' statement for large balls with uniform probability, and exponential tail control on the Young times that mark good expansion events. The underlying hypotheses are only that the critical set is a nice codimension-one manifold with power-law derivative bounds and that an ergodic stationary measure $\mu$ exists with $\lambda < 0$, meaning all Lyapunov exponents are positive.

Load-bearing premise

The proof needs the ergodic stationary measure to visit every open subset of its ergodic component within a fixed time, so a reference ball can be placed inside any prescribed open set $A$; this full-support property is used in Lemma 4.2 but not stated in Hypothesis (H2).

Editorial extensions

If this is right

  • Every open set inside the ergodic component contains two disjoint balls forming a random $\kappa$-horseshoe, so the symbolic subsystem $P(\omega)$ is present at arbitrarily small scales.
  • The set $P(\omega)$ is hyperbolic and in one-to-one correspondence with binary sequences, giving a random analogue of the classical horseshoe that does not rely on periodic orbits.
  • The expected waiting time for a Young time of a ball of size $|I|$ is of order $\log(1/|I|)$; consequently smaller balls wait only logarithmically longer to be expanded onto the reference set.
  • The large-deviation estimates hold uniformly over the state space outside a Lebesgue-exceptional set, so the annealed control of Lyapunov averages and critical-set recurrences is exponentially strong.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the proof's reliance on the annealed transition kernel suggests that the additive form of the noise is not essential; any absolutely continuous perturbation with full support and controlled density should yield the same density of random horseshoes, with different large-deviation constants.
  • Beyond the paper, the full-support condition flagged in the proof may be automatic for diffusive additive noise: a noise density bounded away from zero should make the transition kernel irreducible on components, forcing $\operatorname{supp}\mu = Y$ and closing the gap in Hypothesis (H2).
  • Beyond the paper, the binary coding of $P(\omega)$ suggests that quenched positive entropy and a random symbolic dynamics description follow from the same construction; the expected return time $\mathbb{E}[n_0]$ in Definition 2.1 is the natural rate to extract, though the paper does not compute it.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes a notion of random horseshoe (Definition 2.1) and claims in Theorem 2.2 that, for non-uniformly expanding random dynamical systems with additive noise satisfying hypotheses (H1)-(H2), the set of κ-horseshoes is dense. The proof strategy is: (i) establish large deviation estimates for Birkhoff averages and recurrences to the critical set (Section 3); (ii) prove annealed properties of orbits of large balls (Section 4); (iii) introduce Young times and derive tail estimates for their occurrence (Section 5); and (iv) use a combinatorial argument to find two balls with infinitely many common Young times, which are then asserted to yield a random horseshoe (Section 6). The main theorem is presented as a generalization of Katok's theorem to random systems, avoiding the need for periodic orbits.

Significance. If the main theorem were established, it would be a substantial contribution to the random dynamical systems literature, providing a random analogue of Katok's dense horseshoes under only positivity of Lyapunov exponents and additive noise. The paper introduces a useful definition of random horseshoe and attempts to connect Young times to random symbolic dynamics. However, the central proof is incomplete: Section 6.2 stops without verifying the defining conditions of a random horseshoe, and the probabilistic argument for common Young times is flawed. Several auxiliary results are taken from self-authored preprints, making verification difficult. Thus, the claimed theorem is not established, though some of the ancillary estimates (e.g., the large deviation results) may be of independent interest.

major comments (5)
  1. [Section 6.2, equations (6.9)-(6.13)] The proof stops at the limit statement (6.13) and never verifies the conditions (2.8)-(2.11) of Definition 2.1. In particular, (2.9) requires that f^{n_{k+1}-n_k}_{θ^{n_k}(ω)}(I_i) ⊇ I_0 ∪ I_1 for both i, which is not shown: a common Young time only gives that at time n_k a sub-ball of I_i maps onto the reference set J; it gives no control on the image of the whole ball I_i under the subsequent segment of the composition. Similarly, no sub-balls J(k,ω)_{i,j} satisfying (2.10) and (2.11) are constructed. Therefore Theorem 2.2 is not proven.
  2. [Section 6.2, paragraph after (6.10)] The sequence n^{i,j}_k of common times of two renewal processes is not itself a renewal process with i.i.d. increments. Proposition 6.1, which is invoked to obtain the law of large numbers (6.13), applies only to sums of i.i.d. nonnegative random variables. The waiting times between successive common events of two renewal processes depend on the phases of both processes, and are not i.i.d. Hence the equality (6.11) and the limit (6.13) are unjustified.
  3. [Section 6.1, Proposition 6.1] The proof of Proposition 6.1 is flawed: it uses the Central Limit Theorem to conclude an almost-sure limit of the averages of truncated variables. The correct tool is the Strong Law of Large Numbers (or the monotone convergence theorem applied to the SLLN for truncated variables). While the statement of the proposition is true, the error matters because the proposition is the only justification for (6.13), and the sequence to which it is applied is not i.i.d. in the first place.
  4. [Section 4, Lemma 4.2] The proof assumes that the orbit of the chosen point (ω,x) under the ergodic measure μ visits every element of a δ-partition of Y and reaches a reference ball Δ_c within a fixed time N. This requires that μ have full support on Y (or at least that the support intersects every partition element). Hypothesis (H2) only postulates the existence of a unique ergodic measure on Y, not that its support is all of Y. Without full support, the Birkhoff ergodic theorem does not imply visits to every partition element, and the time N may be infinite on a set of positive measure. This is a load-bearing gap in Proposition 4.1.
  5. [Section 5, Propositions 5.4, 5.5, 5.7] Several key results are taken from self-authored preprints [3] and [12] without proof or even statement. Proposition 5.4 cites [3, Lemma 4.1]; Proposition 5.5 uses [3, Proposition 3.5] and [3, Theorem 3.7]; Proposition 5.7 relies on [12, Proposition 4.3]. Since these results are central to the Young-times estimates, the paper is not self-contained, and the referee cannot verify the main theorem without access to the full proofs of these preprints. At minimum, the relevant statements should be included as lemmas with proofs or clear references to published versions.
minor comments (5)
  1. [Section 4, definition of E_J(I,N,ι)] There is a typo: 'an ball' should be 'a ball'.
  2. [Section 4, proof of Proposition 4.1] The text says 'by Proposition 4.2' but should refer to Lemma 4.2.
  3. [Definition 2.1] The condition (2.8) involves E[n_0], but n_0 is not defined before taking the limit, and the meaning of its expectation is unclear. This should be clarified.
  4. [Section 6.2, equations (6.3)-(6.8)] The notation H^n_i is used before the set is defined, and the phrase 'there exists indices i(n,ω) and j(n,ω)' should be 'there exist indices'.
  5. [Throughout] The paper mixes notation S^1 and X (e.g., Definition 5.1 and Proposition 5.2 mention S^1 while the setup is X ⊂ R^n). This should be made consistent.

Circularity Check

3 steps flagged · score 5.0 of 10

Main theorem leans on same-author preprints for Young-time statistics; no construction-level circularity but load-bearing self-citations.

  1. self citation load bearing [Section 5, Proposition 5.4]
    "The importance of Young times is due to the fact that if n is a Young time, then f n ω uniformly expands a small neighbourhood of x right into J, keeping uniform distortion bounds, as stated in the following proposition (see [3, Lemma 4.1])."

    This proposition is the mechanism that turns a Young time into the uniform expansion onto the reference set J that later becomes the random horseshoe return. It is not proved here; it is imported from [3], an arXiv preprint coauthored by the present author Tenaglia, which was developed for predominantly expanding maps with the large-expanding-region and large-noise hypotheses (i)-(ii). The paper does not verify that [3, Lemma 4.1] holds under (H1)-(H2), nor is [3] machine-checked or otherwise independent. Thus the central expansion/distortion input of the horseshoe construction is a load-bearing self-citation.

  2. self citation load bearing [Section 5, Proposition 5.5]
    "By [3, Proposition 3.5] and Proposition 3.1, there exists u > 0 there exists a set En ⊂ Y with m(En) ≤ Ce−γn, such that if x /∈ En and m ≥ n P {|Sm(ω, x)| ≤ um} ≤ Ce−γm. ... The result now follows following the lines of [3, Theorem 3.7]."

    The density of N-sparse hyperbolic times (via |S_m|) and the resulting counting/tail estimate for Young times are taken from [3], a same-author preprint, rather than derived from (H1)-(H2) in this paper. Proposition 5.5 is the input to Proposition 5.7's stopping-time tail E[m] ≈ log(1/|I|), which Section 6.2 uses to choose M and force two balls to share infinitely many Young times. The step is load-bearing and rests on self-citation.

1 more flagged steps
  1. self citation load bearing [Section 5, Proposition 5.7]
    "Then, the first part of Proposition 5.7 follows from the lines of [12, Proposition 4.3]."

    [12] is the authors' own prior arXiv preprint on predominantly expanding circle RDS; its Proposition 4.3 supplies the exponential tail for the first Young-time stopping time and hence (5.5) E[m] ≈ log(1/|I|). This estimate is what makes the Bonferroni argument in Section 6.2 work (M V_M > 1). Importing it from a non-independent, non-machine-checked self-citation means the horseshoe existence is not derived from Hypotheses (H1)-(H2) alone in the present text.

full rationale

The paper is not circular in the 'prediction equals fitted input' sense: Theorem 2.2 is a genuine transfer from a Lyapunov/ergodic hypothesis to a topological horseshoe conclusion, and Section 3's large-deviations argument is largely self-contained via standard Nagaev and spectral references. However, the Young-time machinery that carries the construction is repeatedly imported from two same-author preprints ([3], with coauthor Tenaglia; [12], all three current authors). Proposition 5.4, Proposition 5.5, and Proposition 5.7 each rest on unproved lemmas from [3] or [12]; these are not machine-checked or independently falsified, and the paper does not re-prove them under (H1)-(H2). Since (5.5) is the quantitative input for the Bonferroni/common-times step in Section 6.2, the central existence claim depends on this self-citation chain. Separately, the passage from common Young times to Definition 2.1's (2.9)-(2.11), and the SLLN for the common-time process in (6.13), are not justified in the text; those are correctness gaps rather than circularity and are not counted in the score. Overall, there is substantial independent content but with load-bearing unverified self-citations, so the circularity score is 5.

Assumptions & free parameters 0 free parameters · 5 assumptions · 1 invented entities

The central claim rests on the structural hypotheses (H1)-(H2), on an unstated full-support assumption, on the spectral method for transfer operators, and crucially on results imported from the authors' own preprints [3] and [12]. The ledger shows the proof is not self-contained.

assumptions (5)
  • domain assumption Hypothesis (H1): structural conditions on the critical set C and derivative bounds (2.3)-(2.5)
    The theorem only applies to maps satisfying these conditions; they are used throughout for large deviations and hyperbolic times.
  • domain assumption Hypothesis (H2): existence of an ergodic stationary component Y with unique ergodic measure μ and λ = ∫ log ||df^{-1}|| dμ < 0
    This is the main condition on the dynamics; the proof builds horseshoes inside Y.
  • domain assumption Full support of μ on Y (supp μ = Y)
    Not stated in (H1)-(H2) but required in Lemma 4.2 to hit every element of a partition in bounded time and to place the reference set J inside any open set A.
  • standard math Spectral gap for the annealed transfer operator Pθ and the Nagaev method
    Lemmas 3.3 and 3.4 invoke [4, Proposition 2.3] and [18, Lemma 1.3] without explicitly verifying their hypotheses; these yield the large deviation estimates on which the Young-time bounds rest.
  • domain assumption Results from self-authored preprints [3] and [12]: random Young tower construction, hyperbolic-time and Young-time estimates
    Proposition 5.4 cites [3, Lemma 4.1], Proposition 5.5 uses [3, Proposition 3.5] and [3, Theorem 3.7], and Proposition 5.7 invokes [12, Proposition 4.3]. These are not proved in the present text and are not independently verified.
invented entities (1)
  • Random horseshoe (Definition 2.1)
    purpose: To formalize a Smale horseshoe for random dynamical systems where periodic points generically do not exist, using a sequence of positive-measure return times.
    This is a new mathematical construct introduced as a definition; no experimental or independent physical evidence is applicable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Non uniform expansion and additive noise imply random horseshoe." pith.science (2026). https://pith.science/paper/PR67FMZS

@misc{pith2026250111656,
  author       = {Pith},
  title        = {Pith review of: Non uniform expansion and additive noise imply random horseshoe},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PR67FMZS}},
  note         = {Machine review of arXiv:2501.11656}
}
read the original abstract

We propose a notion of random horseshoe and prove density of random horseshoes for non uniformly expanding random dynamical systems with additive noise

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 18 canonical work pages

  1. [3]

    Random young tow ers and quenched decay of correlations for predominantly expanding multimodal circle maps

    Matheus M Castro and Giuseppe Tenaglia. Random young tow ers and quenched decay of correlations for predominantly expanding multimodal circle maps. arXiv preprint arXiv:2303.16345 , 2023

    arXiv 2023

  2. [12]

    Horseshoes for a class of nonuniformly expanding random dynamical systems on the circle

    Jeroen SW Lamb, Giuseppe Tenaglia, and Dmitry Turaev. H orsehoes for a class of nonuniformly expanding random dynamical systems on the circle. arXiv preprint arXiv:2304.03685 , 2023

    work page Pith review arXiv 2023

  3. [1]

    J. F. Alves. Nonuniformly Hyperbolic Attractors–Geometric and Probabilistic Aspects. Springer Monographs in Mathematics. Springer, Cham, [2020] ©2020

    work page 2020

  4. [2]

    Alves and Helder Vilarinho

    Jose F. Alves and Helder Vilarinho. Strong stochastic st ability for non-uniformly expanding maps. Ergodic Theory and Dynamical Systems , 33(3):647–692, 2013

    work page 2013

  5. [4]

    Limit theorems in dynamical systems using the spectral method

    S´ ebastien Gou¨ ezel. Limit theorems in dynamical systems using the spectral method. Hyperbolic dynamics, fluctuations and large deviations , 89:161–193, 2015

    work page 2015

  6. [5]

    Convergence of markov processes

    Martin Hairer. Convergence of markov processes. Lecture notes, https: // www. hairer. org/ notes/ Convergence. pdf, 2010. EXISTENCE OF HORSESHOES 17

    work page 2010

  7. [6]

    Horseshoes for Anosov systems on fibers driven by an equicontinuous system

    W en Huang and Zeng Lian. Horseshoes for Anosov systems on fibers driven by an equicontinuous system. Acta Mathematica Sinica , 38(1):281–290, 2022

    work page 2022

  8. [7]

    Ergodic theory of Random Anosov systems mixing on fibers

    W en Huang, Zeng Lian, and Kening Lu. Ergodic theory of ran dom anosov systems mixing on fibers. arXiv preprint arXiv:1612.08394, 2016

    work page Pith review arXiv 2016

Show all 19 references
  1. [8]

    Entropy, chaos, and weak horsesh oe for infinite-dimensional random dynamical systems

    W en Huang and Kening Lu. Entropy, chaos, and weak horsesh oe for infinite-dimensional random dynamical systems. Communications on Pure and Applied Mathematics , 70(10):1987–2036, 2017

    1987

  2. [9]

    Full-horseshoes for the gal erkin truncations of 2d navier-stokes equation with degenerate stochastic forcing

    W en Huang and Jianhua Zhang. Full-horseshoes for the gal erkin truncations of 2d navier-stokes equation with degenerate stochastic forcing. arXiv preprint arXiv:2303.05027 , 2023

    2023 arXiv

  3. [10]

    Observable full-horsesho es for lagrangian flows advected by stochastic 2d navier-stokes equations

    W en Huang and Jianhua Zhang. Observable full-horsesho es for lagrangian flows advected by stochastic 2d navier-stokes equations. arXiv preprint arXiv:2311.05193 , 2023

    2023 arXiv

  4. [11]

    Lyapunov exponents, entropy and period ic orbits for diffeomorphisms

    Anatole Katok. Lyapunov exponents, entropy and period ic orbits for diffeomorphisms. Publications Math´ ematiques de l’IH´ES, 51:137–173, 1980

    1980

  5. [13]

    Existence of periodic orbits and h orseshoes for mappings in a separable banach space

    Zeng Lian and Xiao Ma. Existence of periodic orbits and h orseshoes for mappings in a separable banach space. Journal of Differential Equations , 269(12):11694–11738, 2020

    2020

  6. [14]

    Positive Lyapunov expone nt by a random perturbation

    Zeng Lian and Mikko Stenlund. Positive Lyapunov expone nt by a random perturbation. Dynamical Systems, 27:239 – 252, 2012

    2012

  7. [15]

    Lyapunov exponents, peri odic orbits and horseshoes for mappings of hilbert spaces

    Zeng Lian and Lai-Sang Young. Lyapunov exponents, peri odic orbits and horseshoes for mappings of hilbert spaces. Annales Henri Poincar´ e, 12(6):1081–1108, September 2011

    2011

  8. [16]

    Lyapunov exponents, peri odic orbits, and horseshoes for semiflows on hilbert spaces

    Zeng Lian and Lai-Sang Young. Lyapunov exponents, peri odic orbits, and horseshoes for semiflows on hilbert spaces. Journal of the American Mathematical Society , 25:637–665, 2012

    2012

  9. [17]

    Existence of periodic orbits and horseshoes fo r semiflows on a separable banach space

    Xiao Ma. Existence of periodic orbits and horseshoes fo r semiflows on a separable banach space. Calculus of Variations and Partial Differential Equations , 61(6):217, October 2022

    2022

  10. [18]

    S. V. Nagaev. Some limit theorems for stationary markov chains. Theory of Probability & Its Applications , 2(4):378–406, 1957

    1957

  11. [19]

    James R. Norris. Markov chains, volume 2 of Cambridge Series in Statistical and Probabilistic Mathema tics. Cambridge University Press, Cambridge, 1998. Reprint of 19 97 original. 2International Research Center for Neurointelligence, The Un iversity of Tokyo, Tokyo 113- 0033, ...

    1998

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.